1 \input texinfo @c -*-texinfo-*-
2 @comment %**start of header
3 @setfilename ../../info/eintr
4 @c setfilename emacs-lisp-intro.info
5 @c sethtmlfilename emacs-lisp-intro.html
6 @settitle Programming in Emacs Lisp
11 @include emacsver.texi
13 @c ================ How to Print a Book in Various Sizes ================
15 @c This book can be printed in any of three different sizes.
16 @c Set the following @-commands appropriately.
26 @c European A4 size paper:
31 @c (Note: if you edit the book so as to change the length of the
32 @c table of contents, you may have to change the value of `pageno' below.)
34 @c <<<< For hard copy printing, this file is now
35 @c set for smallbook, which works for all sizes
36 @c of paper, and with PostScript figures >>>>
44 @c ================ Included Figures ================
46 @c If you clear this, the figures will be printed as ASCII diagrams
47 @c rather than PostScript/PDF.
48 @c (This is not relevant to Info, since Info only handles ASCII.)
49 @set print-postscript-figures
50 @c clear print-postscript-figures
52 @comment %**end of header
54 @c per rms and peterb, use 10pt fonts for the main text, mostly to
55 @c save on paper cost.
56 @c Do this inside @tex for now, so current makeinfo does not complain.
62 \global\hbadness=6666 % don't worry about not-too-underfull boxes
65 @c These refer to the printed book sold by the FSF.
66 @set edition-number 3.10
67 @set update-date 28 October 2009
69 @c For next or subsequent edition:
70 @c create function using with-output-to-temp-buffer
71 @c create a major mode, with keymaps
72 @c run an asynchronous process, like grep or diff
74 @c For 8.5 by 11 inch format: do not use such a small amount of
75 @c whitespace between paragraphs as smallbook format
78 \global\parskip 6pt plus 1pt
82 @c For all sized formats: print within-book cross
83 @c reference with ``...'' rather than [...]
85 @c This works with the texinfo.tex file, version 2003-05-04.08,
86 @c in the Texinfo version 4.6 of the 2003 Jun 13 distribution.
89 \if \xrefprintnodename
90 \global\def\xrefprintnodename#1{\unskip, ``#1''}
92 \global\def\xrefprintnodename#1{ ``#1''}
94 % \global\def\xrefprintnodename#1{, ``#1''}
97 @c ----------------------------------------------------
99 @dircategory GNU Emacs Lisp
101 * Emacs Lisp Intro: (eintr).
102 A simple introduction to Emacs Lisp programming.
106 This is an @cite{Introduction to Programming in Emacs Lisp}, for
107 people who are not programmers.
110 Edition @value{edition-number}, @value{update-date}
113 Distributed with Emacs version @value{EMACSVER}.
116 Copyright @copyright{} 1990--1995, 1997, 2001--2013 Free Software
123 GNU Press, @hfill @uref{http://www.fsf.org/licensing/gnu-press/}@*
124 a division of the @hfill email: @email{sales@@fsf.org}@*
125 Free Software Foundation, Inc. @hfill Tel: +1 (617) 542-5942@*
126 51 Franklin Street, Fifth Floor @hfill Fax: +1 (617) 542-2652@*
127 Boston, MA 02110-1301 USA
131 Printed copies available from @uref{http://shop.fsf.org/}. Published by:
134 GNU Press, http://www.fsf.org/licensing/gnu-press/
135 a division of the email: sales@@fsf.org
136 Free Software Foundation, Inc. Tel: +1 (617) 542-5942
137 51 Franklin Street, Fifth Floor Fax: +1 (617) 542-2652
138 Boston, MA 02110-1301 USA
145 Permission is granted to copy, distribute and/or modify this document
146 under the terms of the GNU Free Documentation License, Version 1.3 or
147 any later version published by the Free Software Foundation; there
148 being no Invariant Section, with the Front-Cover Texts being ``A GNU
149 Manual'', and with the Back-Cover Texts as in (a) below. A copy of
150 the license is included in the section entitled ``GNU Free
151 Documentation License''.
153 (a) The FSF's Back-Cover Text is: ``You have the freedom to
154 copy and modify this GNU manual. Buying copies from the FSF
155 supports it in developing GNU and promoting software freedom.''
158 @c half title; two lines here, so do not use `shorttitlepage'
161 \hbox{}\vskip 1.5in \chaprm \centerline{An Introduction to}%
163 {\begingroup\hbox{}\vskip 0.25in \chaprm%
164 \centerline{Programming in Emacs Lisp}%
165 \endgroup\page\hbox{}\page}
170 @center @titlefont{An Introduction to}
172 @center @titlefont{Programming in Emacs Lisp}
174 @center Revised Third Edition
176 @center by Robert J. Chassell
179 @vskip 0pt plus 1filll
185 @evenheading @thispage @| @| @thischapter
186 @oddheading @thissection @| @| @thispage
190 @c Keep T.O.C. short by tightening up for largebook
193 \global\parskip 2pt plus 1pt
194 \global\advance\baselineskip by -1pt
204 @top An Introduction to Programming in Emacs Lisp
208 <p>The homepage for GNU Emacs is at
209 <a href="/software/emacs/">http://www.gnu.org/software/emacs/</a>.<br>
210 To view this manual in other formats, click
211 <a href="/software/emacs/manual/eintr.html">here</a>.
217 This master menu first lists each chapter and index; then it lists
218 every node in every chapter.
221 @c >>>> Set pageno appropriately <<<<
223 @c The first page of the Preface is a roman numeral; it is the first
224 @c right handed page after the Table of Contents; hence the following
225 @c setting must be for an odd negative number.
228 @c global@pageno = -11
231 @set COUNT-WORDS count-words-example
232 @c Length of variable name chosen so that things still line up when expanded.
235 * Preface:: What to look for.
236 * List Processing:: What is Lisp?
237 * Practicing Evaluation:: Running several programs.
238 * Writing Defuns:: How to write function definitions.
239 * Buffer Walk Through:: Exploring a few buffer-related functions.
240 * More Complex:: A few, even more complex functions.
241 * Narrowing & Widening:: Restricting your and Emacs attention to
243 * car cdr & cons:: Fundamental functions in Lisp.
244 * Cutting & Storing Text:: Removing text and saving it.
245 * List Implementation:: How lists are implemented in the computer.
246 * Yanking:: Pasting stored text.
247 * Loops & Recursion:: How to repeat a process.
248 * Regexp Search:: Regular expression searches.
249 * Counting Words:: A review of repetition and regexps.
250 * Words in a defun:: Counting words in a @code{defun}.
251 * Readying a Graph:: A prototype graph printing function.
252 * Emacs Initialization:: How to write a @file{.emacs} file.
253 * Debugging:: How to run the Emacs Lisp debuggers.
254 * Conclusion:: Now you have the basics.
255 * the-the:: An appendix: how to find reduplicated words.
256 * Kill Ring:: An appendix: how the kill ring works.
257 * Full Graph:: How to create a graph with labeled axes.
258 * Free Software and Free Manuals::
259 * GNU Free Documentation License::
264 --- The Detailed Node Listing ---
268 * Why:: Why learn Emacs Lisp?
269 * On Reading this Text:: Read, gain familiarity, pick up habits....
270 * Who You Are:: For whom this is written.
272 * Note for Novices:: You can read this as a novice.
277 * Lisp Lists:: What are lists?
278 * Run a Program:: Any list in Lisp is a program ready to run.
279 * Making Errors:: Generating an error message.
280 * Names & Definitions:: Names of symbols and function definitions.
281 * Lisp Interpreter:: What the Lisp interpreter does.
282 * Evaluation:: Running a program.
283 * Variables:: Returning a value from a variable.
284 * Arguments:: Passing information to a function.
285 * set & setq:: Setting the value of a variable.
286 * Summary:: The major points.
287 * Error Message Exercises::
291 * Numbers Lists:: List have numbers, other lists, in them.
292 * Lisp Atoms:: Elemental entities.
293 * Whitespace in Lists:: Formatting lists to be readable.
294 * Typing Lists:: How GNU Emacs helps you type lists.
298 * Complications:: Variables, Special forms, Lists within.
299 * Byte Compiling:: Specially processing code for speed.
303 * How the Interpreter Acts:: Returns and Side Effects...
304 * Evaluating Inner Lists:: Lists within lists...
308 * fill-column Example::
309 * Void Function:: The error message for a symbol
311 * Void Variable:: The error message for a symbol without a value.
315 * Data types:: Types of data passed to a function.
316 * Args as Variable or List:: An argument can be the value
317 of a variable or list.
318 * Variable Number of Arguments:: Some functions may take a
319 variable number of arguments.
320 * Wrong Type of Argument:: Passing an argument of the wrong type
322 * message:: A useful function for sending messages.
324 Setting the Value of a Variable
326 * Using set:: Setting values.
327 * Using setq:: Setting a quoted value.
328 * Counting:: Using @code{setq} to count.
330 Practicing Evaluation
332 * How to Evaluate:: Typing editing commands or @kbd{C-x C-e}
334 * Buffer Names:: Buffers and files are different.
335 * Getting Buffers:: Getting a buffer itself, not merely its name.
336 * Switching Buffers:: How to change to another buffer.
337 * Buffer Size & Locations:: Where point is located and the size of
339 * Evaluation Exercise::
341 How To Write Function Definitions
343 * Primitive Functions::
344 * defun:: The @code{defun} macro.
345 * Install:: Install a function definition.
346 * Interactive:: Making a function interactive.
347 * Interactive Options:: Different options for @code{interactive}.
348 * Permanent Installation:: Installing code permanently.
349 * let:: Creating and initializing local variables.
351 * else:: If--then--else expressions.
352 * Truth & Falsehood:: What Lisp considers false and true.
353 * save-excursion:: Keeping track of point, mark, and buffer.
357 Install a Function Definition
359 * Effect of installation::
360 * Change a defun:: How to change a function definition.
362 Make a Function Interactive
364 * Interactive multiply-by-seven:: An overview.
365 * multiply-by-seven in detail:: The interactive version.
369 * Prevent confusion::
370 * Parts of let Expression::
371 * Sample let Expression::
372 * Uninitialized let Variables::
374 The @code{if} Special Form
376 * if in more detail::
377 * type-of-animal in detail:: An example of an @code{if} expression.
379 Truth and Falsehood in Emacs Lisp
381 * nil explained:: @code{nil} has two meanings.
383 @code{save-excursion}
385 * Point and mark:: A review of various locations.
386 * Template for save-excursion::
388 A Few Buffer--Related Functions
390 * Finding More:: How to find more information.
391 * simplified-beginning-of-buffer:: Shows @code{goto-char},
392 @code{point-min}, and @code{push-mark}.
393 * mark-whole-buffer:: Almost the same as @code{beginning-of-buffer}.
394 * append-to-buffer:: Uses @code{save-excursion} and
395 @code{insert-buffer-substring}.
396 * Buffer Related Review:: Review.
399 The Definition of @code{mark-whole-buffer}
401 * mark-whole-buffer overview::
402 * Body of mark-whole-buffer:: Only three lines of code.
404 The Definition of @code{append-to-buffer}
406 * append-to-buffer overview::
407 * append interactive:: A two part interactive expression.
408 * append-to-buffer body:: Incorporates a @code{let} expression.
409 * append save-excursion:: How the @code{save-excursion} works.
411 A Few More Complex Functions
413 * copy-to-buffer:: With @code{set-buffer}, @code{get-buffer-create}.
414 * insert-buffer:: Read-only, and with @code{or}.
415 * beginning-of-buffer:: Shows @code{goto-char},
416 @code{point-min}, and @code{push-mark}.
417 * Second Buffer Related Review::
418 * optional Exercise::
420 The Definition of @code{insert-buffer}
422 * insert-buffer code::
423 * insert-buffer interactive:: When you can read, but not write.
424 * insert-buffer body:: The body has an @code{or} and a @code{let}.
425 * if & or:: Using an @code{if} instead of an @code{or}.
426 * Insert or:: How the @code{or} expression works.
427 * Insert let:: Two @code{save-excursion} expressions.
428 * New insert-buffer::
430 The Interactive Expression in @code{insert-buffer}
432 * Read-only buffer:: When a buffer cannot be modified.
433 * b for interactive:: An existing buffer or else its name.
435 Complete Definition of @code{beginning-of-buffer}
437 * Optional Arguments::
438 * beginning-of-buffer opt arg:: Example with optional argument.
439 * beginning-of-buffer complete::
441 @code{beginning-of-buffer} with an Argument
443 * Disentangle beginning-of-buffer::
444 * Large buffer case::
445 * Small buffer case::
447 Narrowing and Widening
449 * Narrowing advantages:: The advantages of narrowing
450 * save-restriction:: The @code{save-restriction} special form.
451 * what-line:: The number of the line that point is on.
454 @code{car}, @code{cdr}, @code{cons}: Fundamental Functions
456 * Strange Names:: An historical aside: why the strange names?
457 * car & cdr:: Functions for extracting part of a list.
458 * cons:: Constructing a list.
459 * nthcdr:: Calling @code{cdr} repeatedly.
461 * setcar:: Changing the first element of a list.
462 * setcdr:: Changing the rest of a list.
468 * length:: How to find the length of a list.
470 Cutting and Storing Text
472 * Storing Text:: Text is stored in a list.
473 * zap-to-char:: Cutting out text up to a character.
474 * kill-region:: Cutting text out of a region.
475 * copy-region-as-kill:: A definition for copying text.
476 * Digression into C:: Minor note on C programming language macros.
477 * defvar:: How to give a variable an initial value.
478 * cons & search-fwd Review::
483 * Complete zap-to-char:: The complete implementation.
484 * zap-to-char interactive:: A three part interactive expression.
485 * zap-to-char body:: A short overview.
486 * search-forward:: How to search for a string.
487 * progn:: The @code{progn} special form.
488 * Summing up zap-to-char:: Using @code{point} and @code{search-forward}.
492 * Complete kill-region:: The function definition.
493 * condition-case:: Dealing with a problem.
496 @code{copy-region-as-kill}
498 * Complete copy-region-as-kill:: The complete function definition.
499 * copy-region-as-kill body:: The body of @code{copy-region-as-kill}.
501 The Body of @code{copy-region-as-kill}
503 * last-command & this-command::
504 * kill-append function::
505 * kill-new function::
507 Initializing a Variable with @code{defvar}
509 * See variable current value::
510 * defvar and asterisk::
512 How Lists are Implemented
515 * Symbols as Chest:: Exploring a powerful metaphor.
520 * Kill Ring Overview::
521 * kill-ring-yank-pointer:: The kill ring is a list.
522 * yank nthcdr Exercises:: The @code{kill-ring-yank-pointer} variable.
526 * while:: Causing a stretch of code to repeat.
528 * Recursion:: Causing a function to call itself.
533 * Looping with while:: Repeat so long as test returns true.
534 * Loop Example:: A @code{while} loop that uses a list.
535 * print-elements-of-list:: Uses @code{while}, @code{car}, @code{cdr}.
536 * Incrementing Loop:: A loop with an incrementing counter.
537 * Incrementing Loop Details::
538 * Decrementing Loop:: A loop with a decrementing counter.
540 Details of an Incrementing Loop
542 * Incrementing Example:: Counting pebbles in a triangle.
543 * Inc Example parts:: The parts of the function definition.
544 * Inc Example altogether:: Putting the function definition together.
546 Loop with a Decrementing Counter
548 * Decrementing Example:: More pebbles on the beach.
549 * Dec Example parts:: The parts of the function definition.
550 * Dec Example altogether:: Putting the function definition together.
552 Save your time: @code{dolist} and @code{dotimes}
559 * Building Robots:: Same model, different serial number ...
560 * Recursive Definition Parts:: Walk until you stop ...
561 * Recursion with list:: Using a list as the test whether to recurse.
562 * Recursive triangle function::
563 * Recursion with cond::
564 * Recursive Patterns:: Often used templates.
565 * No Deferment:: Don't store up work ...
566 * No deferment solution::
568 Recursion in Place of a Counter
570 * Recursive Example arg of 1 or 2::
571 * Recursive Example arg of 3 or 4::
579 Regular Expression Searches
581 * sentence-end:: The regular expression for @code{sentence-end}.
582 * re-search-forward:: Very similar to @code{search-forward}.
583 * forward-sentence:: A straightforward example of regexp search.
584 * forward-paragraph:: A somewhat complex example.
585 * etags:: How to create your own @file{TAGS} table.
587 * re-search Exercises::
589 @code{forward-sentence}
591 * Complete forward-sentence::
592 * fwd-sentence while loops:: Two @code{while} loops.
593 * fwd-sentence re-search:: A regular expression search.
595 @code{forward-paragraph}: a Goldmine of Functions
597 * forward-paragraph in brief:: Key parts of the function definition.
598 * fwd-para let:: The @code{let*} expression.
599 * fwd-para while:: The forward motion @code{while} loop.
601 Counting: Repetition and Regexps
604 * @value{COUNT-WORDS}:: Use a regexp, but find a problem.
605 * recursive-count-words:: Start with case of no words in region.
606 * Counting Exercise::
608 The @code{@value{COUNT-WORDS}} Function
610 * Design @value{COUNT-WORDS}:: The definition using a @code{while} loop.
611 * Whitespace Bug:: The Whitespace Bug in @code{@value{COUNT-WORDS}}.
613 Counting Words in a @code{defun}
615 * Divide and Conquer::
616 * Words and Symbols:: What to count?
617 * Syntax:: What constitutes a word or symbol?
618 * count-words-in-defun:: Very like @code{@value{COUNT-WORDS}}.
619 * Several defuns:: Counting several defuns in a file.
620 * Find a File:: Do you want to look at a file?
621 * lengths-list-file:: A list of the lengths of many definitions.
622 * Several files:: Counting in definitions in different files.
623 * Several files recursively:: Recursively counting in different files.
624 * Prepare the data:: Prepare the data for display in a graph.
626 Count Words in @code{defuns} in Different Files
628 * lengths-list-many-files:: Return a list of the lengths of defuns.
629 * append:: Attach one list to another.
631 Prepare the Data for Display in a Graph
633 * Data for Display in Detail::
634 * Sorting:: Sorting lists.
635 * Files List:: Making a list of files.
636 * Counting function definitions::
640 * Columns of a graph::
641 * graph-body-print:: How to print the body of a graph.
642 * recursive-graph-body-print::
644 * Line Graph Exercise::
646 Your @file{.emacs} File
648 * Default Configuration::
649 * Site-wide Init:: You can write site-wide init files.
650 * defcustom:: Emacs will write code for you.
651 * Beginning init File:: How to write a @file{.emacs} init file.
652 * Text and Auto-fill:: Automatically wrap lines.
653 * Mail Aliases:: Use abbreviations for email addresses.
654 * Indent Tabs Mode:: Don't use tabs with @TeX{}
655 * Keybindings:: Create some personal keybindings.
656 * Keymaps:: More about key binding.
657 * Loading Files:: Load (i.e., evaluate) files automatically.
658 * Autoload:: Make functions available.
659 * Simple Extension:: Define a function; bind it to a key.
660 * X11 Colors:: Colors in X.
662 * Mode Line:: How to customize your mode line.
666 * debug:: How to use the built-in debugger.
667 * debug-on-entry:: Start debugging when you call a function.
668 * debug-on-quit:: Start debugging when you quit with @kbd{C-g}.
669 * edebug:: How to use Edebug, a source level debugger.
670 * Debugging Exercises::
672 Handling the Kill Ring
674 * What the Kill Ring Does::
676 * yank:: Paste a copy of a clipped element.
677 * yank-pop:: Insert element pointed to.
680 The @code{current-kill} Function
682 * Code for current-kill::
683 * Understanding current-kill::
685 @code{current-kill} in Outline
687 * Body of current-kill::
688 * Digression concerning error:: How to mislead humans, but not computers.
689 * Determining the Element::
691 A Graph with Labeled Axes
694 * print-graph Varlist:: @code{let} expression in @code{print-graph}.
695 * print-Y-axis:: Print a label for the vertical axis.
696 * print-X-axis:: Print a horizontal label.
697 * Print Whole Graph:: The function to print a complete graph.
699 The @code{print-Y-axis} Function
701 * print-Y-axis in Detail::
702 * Height of label:: What height for the Y axis?
703 * Compute a Remainder:: How to compute the remainder of a division.
704 * Y Axis Element:: Construct a line for the Y axis.
705 * Y-axis-column:: Generate a list of Y axis labels.
706 * print-Y-axis Penultimate:: A not quite final version.
708 The @code{print-X-axis} Function
710 * Similarities differences:: Much like @code{print-Y-axis}, but not exactly.
711 * X Axis Tic Marks:: Create tic marks for the horizontal axis.
713 Printing the Whole Graph
715 * The final version:: A few changes.
716 * Test print-graph:: Run a short test.
717 * Graphing words in defuns:: Executing the final code.
718 * lambda:: How to write an anonymous function.
719 * mapcar:: Apply a function to elements of a list.
720 * Another Bug:: Yet another bug @dots{} most insidious.
721 * Final printed graph:: The graph itself!
729 Most of the GNU Emacs integrated environment is written in the programming
730 language called Emacs Lisp. The code written in this programming
731 language is the software---the sets of instructions---that tell the
732 computer what to do when you give it commands. Emacs is designed so
733 that you can write new code in Emacs Lisp and easily install it as an
734 extension to the editor.
736 (GNU Emacs is sometimes called an ``extensible editor'', but it does
737 much more than provide editing capabilities. It is better to refer to
738 Emacs as an ``extensible computing environment''. However, that
739 phrase is quite a mouthful. It is easier to refer to Emacs simply as
740 an editor. Moreover, everything you do in Emacs---find the Mayan date
741 and phases of the moon, simplify polynomials, debug code, manage
742 files, read letters, write books---all these activities are kinds of
743 editing in the most general sense of the word.)
746 * Why:: Why learn Emacs Lisp?
747 * On Reading this Text:: Read, gain familiarity, pick up habits....
748 * Who You Are:: For whom this is written.
750 * Note for Novices:: You can read this as a novice.
756 @unnumberedsec Why Study Emacs Lisp?
759 Although Emacs Lisp is usually thought of in association only with Emacs,
760 it is a full computer programming language. You can use Emacs Lisp as
761 you would any other programming language.
763 Perhaps you want to understand programming; perhaps you want to extend
764 Emacs; or perhaps you want to become a programmer. This introduction to
765 Emacs Lisp is designed to get you started: to guide you in learning the
766 fundamentals of programming, and more importantly, to show you how you
767 can teach yourself to go further.
769 @node On Reading this Text
770 @unnumberedsec On Reading this Text
772 All through this document, you will see little sample programs you can
773 run inside of Emacs. If you read this document in Info inside of GNU
774 Emacs, you can run the programs as they appear. (This is easy to do and
775 is explained when the examples are presented.) Alternatively, you can
776 read this introduction as a printed book while sitting beside a computer
777 running Emacs. (This is what I like to do; I like printed books.) If
778 you don't have a running Emacs beside you, you can still read this book,
779 but in this case, it is best to treat it as a novel or as a travel guide
780 to a country not yet visited: interesting, but not the same as being
783 Much of this introduction is dedicated to walkthroughs or guided tours
784 of code used in GNU Emacs. These tours are designed for two purposes:
785 first, to give you familiarity with real, working code (code you use
786 every day); and, second, to give you familiarity with the way Emacs
787 works. It is interesting to see how a working environment is
790 hope that you will pick up the habit of browsing through source code.
791 You can learn from it and mine it for ideas. Having GNU Emacs is like
792 having a dragon's cave of treasures.
794 In addition to learning about Emacs as an editor and Emacs Lisp as a
795 programming language, the examples and guided tours will give you an
796 opportunity to get acquainted with Emacs as a Lisp programming
797 environment. GNU Emacs supports programming and provides tools that
798 you will want to become comfortable using, such as @kbd{M-.} (the key
799 which invokes the @code{find-tag} command). You will also learn about
800 buffers and other objects that are part of the environment.
801 Learning about these features of Emacs is like learning new routes
802 around your home town.
805 In addition, I have written several programs as extended examples.
806 Although these are examples, the programs are real. I use them.
807 Other people use them. You may use them. Beyond the fragments of
808 programs used for illustrations, there is very little in here that is
809 `just for teaching purposes'; what you see is used. This is a great
810 advantage of Emacs Lisp: it is easy to learn to use it for work.
813 Finally, I hope to convey some of the skills for using Emacs to
814 learn aspects of programming that you don't know. You can often use
815 Emacs to help you understand what puzzles you or to find out how to do
816 something new. This self-reliance is not only a pleasure, but an
820 @unnumberedsec For Whom This is Written
822 This text is written as an elementary introduction for people who are
823 not programmers. If you are a programmer, you may not be satisfied with
824 this primer. The reason is that you may have become expert at reading
825 reference manuals and be put off by the way this text is organized.
827 An expert programmer who reviewed this text said to me:
830 @i{I prefer to learn from reference manuals. I ``dive into'' each
831 paragraph, and ``come up for air'' between paragraphs.}
833 @i{When I get to the end of a paragraph, I assume that that subject is
834 done, finished, that I know everything I need (with the
835 possible exception of the case when the next paragraph starts talking
836 about it in more detail). I expect that a well written reference manual
837 will not have a lot of redundancy, and that it will have excellent
838 pointers to the (one) place where the information I want is.}
841 This introduction is not written for this person!
843 Firstly, I try to say everything at least three times: first, to
844 introduce it; second, to show it in context; and third, to show it in a
845 different context, or to review it.
847 Secondly, I hardly ever put all the information about a subject in one
848 place, much less in one paragraph. To my way of thinking, that imposes
849 too heavy a burden on the reader. Instead I try to explain only what
850 you need to know at the time. (Sometimes I include a little extra
851 information so you won't be surprised later when the additional
852 information is formally introduced.)
854 When you read this text, you are not expected to learn everything the
855 first time. Frequently, you need only make, as it were, a `nodding
856 acquaintance' with some of the items mentioned. My hope is that I have
857 structured the text and given you enough hints that you will be alert to
858 what is important, and concentrate on it.
860 You will need to ``dive into'' some paragraphs; there is no other way
861 to read them. But I have tried to keep down the number of such
862 paragraphs. This book is intended as an approachable hill, rather than
863 as a daunting mountain.
865 This introduction to @cite{Programming in Emacs Lisp} has a companion
868 @cite{The GNU Emacs Lisp Reference Manual}.
871 @ref{Top, , The GNU Emacs Lisp Reference Manual, elisp, The GNU
872 Emacs Lisp Reference Manual}.
874 The reference manual has more detail than this introduction. In the
875 reference manual, all the information about one topic is concentrated
876 in one place. You should turn to it if you are like the programmer
877 quoted above. And, of course, after you have read this
878 @cite{Introduction}, you will find the @cite{Reference Manual} useful
879 when you are writing your own programs.
882 @unnumberedsec Lisp History
885 Lisp was first developed in the late 1950s at the Massachusetts
886 Institute of Technology for research in artificial intelligence. The
887 great power of the Lisp language makes it superior for other purposes as
888 well, such as writing editor commands and integrated environments.
892 GNU Emacs Lisp is largely inspired by Maclisp, which was written at MIT
893 in the 1960s. It is somewhat inspired by Common Lisp, which became a
894 standard in the 1980s. However, Emacs Lisp is much simpler than Common
895 Lisp. (The standard Emacs distribution contains an optional extensions
896 file, @file{cl.el}, that adds many Common Lisp features to Emacs Lisp.)
898 @node Note for Novices
899 @unnumberedsec A Note for Novices
901 If you don't know GNU Emacs, you can still read this document
902 profitably. However, I recommend you learn Emacs, if only to learn to
903 move around your computer screen. You can teach yourself how to use
904 Emacs with the on-line tutorial. To use it, type @kbd{C-h t}. (This
905 means you press and release the @key{CTRL} key and the @kbd{h} at the
906 same time, and then press and release @kbd{t}.)
908 Also, I often refer to one of Emacs's standard commands by listing the
909 keys which you press to invoke the command and then giving the name of
910 the command in parentheses, like this: @kbd{M-C-\}
911 (@code{indent-region}). What this means is that the
912 @code{indent-region} command is customarily invoked by typing
913 @kbd{M-C-\}. (You can, if you wish, change the keys that are typed to
914 invoke the command; this is called @dfn{rebinding}. @xref{Keymaps, ,
915 Keymaps}.) The abbreviation @kbd{M-C-\} means that you type your
916 @key{META} key, @key{CTRL} key and @key{\} key all at the same time.
917 (On many modern keyboards the @key{META} key is labeled
919 Sometimes a combination like this is called a keychord, since it is
920 similar to the way you play a chord on a piano. If your keyboard does
921 not have a @key{META} key, the @key{ESC} key prefix is used in place
922 of it. In this case, @kbd{M-C-\} means that you press and release your
923 @key{ESC} key and then type the @key{CTRL} key and the @key{\} key at
924 the same time. But usually @kbd{M-C-\} means press the @key{CTRL} key
925 along with the key that is labeled @key{ALT} and, at the same time,
926 press the @key{\} key.
928 In addition to typing a lone keychord, you can prefix what you type
929 with @kbd{C-u}, which is called the `universal argument'. The
930 @kbd{C-u} keychord passes an argument to the subsequent command.
931 Thus, to indent a region of plain text by 6 spaces, mark the region,
932 and then type @w{@kbd{C-u 6 M-C-\}}. (If you do not specify a number,
933 Emacs either passes the number 4 to the command or otherwise runs the
934 command differently than it would otherwise.) @xref{Arguments, ,
935 Numeric Arguments, emacs, The GNU Emacs Manual}.
937 If you are reading this in Info using GNU Emacs, you can read through
938 this whole document just by pressing the space bar, @key{SPC}.
939 (To learn about Info, type @kbd{C-h i} and then select Info.)
941 A note on terminology: when I use the word Lisp alone, I often am
942 referring to the various dialects of Lisp in general, but when I speak
943 of Emacs Lisp, I am referring to GNU Emacs Lisp in particular.
946 @unnumberedsec Thank You
948 My thanks to all who helped me with this book. My especial thanks to
949 @r{Jim Blandy}, @r{Noah Friedman}, @w{Jim Kingdon}, @r{Roland
950 McGrath}, @w{Frank Ritter}, @w{Randy Smith}, @w{Richard M.
951 Stallman}, and @w{Melissa Weisshaus}. My thanks also go to both
952 @w{Philip Johnson} and @w{David Stampe} for their patient
953 encouragement. My mistakes are my own.
965 @c ================ Beginning of main text ================
967 @c Start main text on right-hand (verso) page
970 \par\vfill\supereject
973 \par\vfill\supereject
975 \par\vfill\supereject
977 \par\vfill\supereject
981 @c Note: this resetting of the page number back to 1 causes TeX to gripe
982 @c about already having seen page numbers 1-4 before (in the preface):
983 @c pdfTeX warning (ext4): destination with the same identifier (name{1})
984 @c has been already used, duplicate ignored
985 @c I guess that is harmless (what happens if a later part of the text
986 @c makes a link to something in the first 4 pages though?).
987 @c E.g., note that the Emacs manual has a preface, but does not bother
988 @c resetting the page numbers back to 1 after that.
991 @evenheading @thispage @| @| @thischapter
992 @oddheading @thissection @| @| @thispage
996 @node List Processing
997 @chapter List Processing
999 To the untutored eye, Lisp is a strange programming language. In Lisp
1000 code there are parentheses everywhere. Some people even claim that
1001 the name stands for `Lots of Isolated Silly Parentheses'. But the
1002 claim is unwarranted. Lisp stands for LISt Processing, and the
1003 programming language handles @emph{lists} (and lists of lists) by
1004 putting them between parentheses. The parentheses mark the boundaries
1005 of the list. Sometimes a list is preceded by a single apostrophe or
1006 quotation mark, @samp{'}@footnote{The single apostrophe or quotation
1007 mark is an abbreviation for the function @code{quote}; you need not
1008 think about functions now; functions are defined in @ref{Making
1009 Errors, , Generate an Error Message}.} Lists are the basis of Lisp.
1012 * Lisp Lists:: What are lists?
1013 * Run a Program:: Any list in Lisp is a program ready to run.
1014 * Making Errors:: Generating an error message.
1015 * Names & Definitions:: Names of symbols and function definitions.
1016 * Lisp Interpreter:: What the Lisp interpreter does.
1017 * Evaluation:: Running a program.
1018 * Variables:: Returning a value from a variable.
1019 * Arguments:: Passing information to a function.
1020 * set & setq:: Setting the value of a variable.
1021 * Summary:: The major points.
1022 * Error Message Exercises::
1029 In Lisp, a list looks like this: @code{'(rose violet daisy buttercup)}.
1030 This list is preceded by a single apostrophe. It could just as well be
1031 written as follows, which looks more like the kind of list you are likely
1032 to be familiar with:
1044 The elements of this list are the names of the four different flowers,
1045 separated from each other by whitespace and surrounded by parentheses,
1046 like flowers in a field with a stone wall around them.
1047 @cindex Flowers in a field
1050 * Numbers Lists:: List have numbers, other lists, in them.
1051 * Lisp Atoms:: Elemental entities.
1052 * Whitespace in Lists:: Formatting lists to be readable.
1053 * Typing Lists:: How GNU Emacs helps you type lists.
1058 @unnumberedsubsec Numbers, Lists inside of Lists
1061 Lists can also have numbers in them, as in this list: @code{(+ 2 2)}.
1062 This list has a plus-sign, @samp{+}, followed by two @samp{2}s, each
1063 separated by whitespace.
1065 In Lisp, both data and programs are represented the same way; that is,
1066 they are both lists of words, numbers, or other lists, separated by
1067 whitespace and surrounded by parentheses. (Since a program looks like
1068 data, one program may easily serve as data for another; this is a very
1069 powerful feature of Lisp.) (Incidentally, these two parenthetical
1070 remarks are @emph{not} Lisp lists, because they contain @samp{;} and
1071 @samp{.} as punctuation marks.)
1074 Here is another list, this time with a list inside of it:
1077 '(this list has (a list inside of it))
1080 The components of this list are the words @samp{this}, @samp{list},
1081 @samp{has}, and the list @samp{(a list inside of it)}. The interior
1082 list is made up of the words @samp{a}, @samp{list}, @samp{inside},
1083 @samp{of}, @samp{it}.
1086 @subsection Lisp Atoms
1089 In Lisp, what we have been calling words are called @dfn{atoms}. This
1090 term comes from the historical meaning of the word atom, which means
1091 `indivisible'. As far as Lisp is concerned, the words we have been
1092 using in the lists cannot be divided into any smaller parts and still
1093 mean the same thing as part of a program; likewise with numbers and
1094 single character symbols like @samp{+}. On the other hand, unlike an
1095 ancient atom, a list can be split into parts. (@xref{car cdr & cons,
1096 , @code{car} @code{cdr} & @code{cons} Fundamental Functions}.)
1098 In a list, atoms are separated from each other by whitespace. They can be
1099 right next to a parenthesis.
1101 @cindex @samp{empty list} defined
1102 Technically speaking, a list in Lisp consists of parentheses surrounding
1103 atoms separated by whitespace or surrounding other lists or surrounding
1104 both atoms and other lists. A list can have just one atom in it or
1105 have nothing in it at all. A list with nothing in it looks like this:
1106 @code{()}, and is called the @dfn{empty list}. Unlike anything else, an
1107 empty list is considered both an atom and a list at the same time.
1109 @cindex Symbolic expressions, introduced
1110 @cindex @samp{expression} defined
1111 @cindex @samp{form} defined
1112 The printed representation of both atoms and lists are called
1113 @dfn{symbolic expressions} or, more concisely, @dfn{s-expressions}.
1114 The word @dfn{expression} by itself can refer to either the printed
1115 representation, or to the atom or list as it is held internally in the
1116 computer. Often, people use the term @dfn{expression}
1117 indiscriminately. (Also, in many texts, the word @dfn{form} is used
1118 as a synonym for expression.)
1120 Incidentally, the atoms that make up our universe were named such when
1121 they were thought to be indivisible; but it has been found that physical
1122 atoms are not indivisible. Parts can split off an atom or it can
1123 fission into two parts of roughly equal size. Physical atoms were named
1124 prematurely, before their truer nature was found. In Lisp, certain
1125 kinds of atom, such as an array, can be separated into parts; but the
1126 mechanism for doing this is different from the mechanism for splitting a
1127 list. As far as list operations are concerned, the atoms of a list are
1130 As in English, the meanings of the component letters of a Lisp atom
1131 are different from the meaning the letters make as a word. For
1132 example, the word for the South American sloth, the @samp{ai}, is
1133 completely different from the two words, @samp{a}, and @samp{i}.
1135 There are many kinds of atom in nature but only a few in Lisp: for
1136 example, @dfn{numbers}, such as 37, 511, or 1729, and @dfn{symbols}, such
1137 as @samp{+}, @samp{foo}, or @samp{forward-line}. The words we have
1138 listed in the examples above are all symbols. In everyday Lisp
1139 conversation, the word ``atom'' is not often used, because programmers
1140 usually try to be more specific about what kind of atom they are dealing
1141 with. Lisp programming is mostly about symbols (and sometimes numbers)
1142 within lists. (Incidentally, the preceding three word parenthetical
1143 remark is a proper list in Lisp, since it consists of atoms, which in
1144 this case are symbols, separated by whitespace and enclosed by
1145 parentheses, without any non-Lisp punctuation.)
1148 Text between double quotation marks---even sentences or
1149 paragraphs---is also an atom. Here is an example:
1150 @cindex Text between double quotation marks
1153 '(this list includes "text between quotation marks.")
1156 @cindex @samp{string} defined
1158 In Lisp, all of the quoted text including the punctuation mark and the
1159 blank spaces is a single atom. This kind of atom is called a
1160 @dfn{string} (for `string of characters') and is the sort of thing that
1161 is used for messages that a computer can print for a human to read.
1162 Strings are a different kind of atom than numbers or symbols and are
1165 @node Whitespace in Lists
1166 @subsection Whitespace in Lists
1167 @cindex Whitespace in lists
1170 The amount of whitespace in a list does not matter. From the point of view
1171 of the Lisp language,
1182 is exactly the same as this:
1185 '(this list looks like this)
1188 Both examples show what to Lisp is the same list, the list made up of
1189 the symbols @samp{this}, @samp{list}, @samp{looks}, @samp{like}, and
1190 @samp{this} in that order.
1192 Extra whitespace and newlines are designed to make a list more readable
1193 by humans. When Lisp reads the expression, it gets rid of all the extra
1194 whitespace (but it needs to have at least one space between atoms in
1195 order to tell them apart.)
1197 Odd as it seems, the examples we have seen cover almost all of what Lisp
1198 lists look like! Every other list in Lisp looks more or less like one
1199 of these examples, except that the list may be longer and more complex.
1200 In brief, a list is between parentheses, a string is between quotation
1201 marks, a symbol looks like a word, and a number looks like a number.
1202 (For certain situations, square brackets, dots and a few other special
1203 characters may be used; however, we will go quite far without them.)
1206 @subsection GNU Emacs Helps You Type Lists
1207 @cindex Help typing lists
1208 @cindex Formatting help
1210 When you type a Lisp expression in GNU Emacs using either Lisp
1211 Interaction mode or Emacs Lisp mode, you have available to you several
1212 commands to format the Lisp expression so it is easy to read. For
1213 example, pressing the @key{TAB} key automatically indents the line the
1214 cursor is on by the right amount. A command to properly indent the
1215 code in a region is customarily bound to @kbd{M-C-\}. Indentation is
1216 designed so that you can see which elements of a list belong to which
1217 list---elements of a sub-list are indented more than the elements of
1220 In addition, when you type a closing parenthesis, Emacs momentarily
1221 jumps the cursor back to the matching opening parenthesis, so you can
1222 see which one it is. This is very useful, since every list you type
1223 in Lisp must have its closing parenthesis match its opening
1224 parenthesis. (@xref{Major Modes, , Major Modes, emacs, The GNU Emacs
1225 Manual}, for more information about Emacs's modes.)
1228 @section Run a Program
1229 @cindex Run a program
1230 @cindex Program, running one
1232 @cindex @samp{evaluate} defined
1233 A list in Lisp---any list---is a program ready to run. If you run it
1234 (for which the Lisp jargon is @dfn{evaluate}), the computer will do one
1235 of three things: do nothing except return to you the list itself; send
1236 you an error message; or, treat the first symbol in the list as a
1237 command to do something. (Usually, of course, it is the last of these
1238 three things that you really want!)
1240 @c use code for the single apostrophe, not samp.
1241 The single apostrophe, @code{'}, that I put in front of some of the
1242 example lists in preceding sections is called a @dfn{quote}; when it
1243 precedes a list, it tells Lisp to do nothing with the list, other than
1244 take it as it is written. But if there is no quote preceding a list,
1245 the first item of the list is special: it is a command for the computer
1246 to obey. (In Lisp, these commands are called @emph{functions}.) The list
1247 @code{(+ 2 2)} shown above did not have a quote in front of it, so Lisp
1248 understands that the @code{+} is an instruction to do something with the
1249 rest of the list: add the numbers that follow.
1252 If you are reading this inside of GNU Emacs in Info, here is how you can
1253 evaluate such a list: place your cursor immediately after the right
1254 hand parenthesis of the following list and then type @kbd{C-x C-e}:
1260 @c use code for the number four, not samp.
1262 You will see the number @code{4} appear in the echo area. (In the
1263 jargon, what you have just done is ``evaluate the list.'' The echo area
1264 is the line at the bottom of the screen that displays or ``echoes''
1265 text.) Now try the same thing with a quoted list: place the cursor
1266 right after the following list and type @kbd{C-x C-e}:
1269 '(this is a quoted list)
1273 You will see @code{(this is a quoted list)} appear in the echo area.
1275 @cindex Lisp interpreter, explained
1276 @cindex Interpreter, Lisp, explained
1277 In both cases, what you are doing is giving a command to the program
1278 inside of GNU Emacs called the @dfn{Lisp interpreter}---giving the
1279 interpreter a command to evaluate the expression. The name of the Lisp
1280 interpreter comes from the word for the task done by a human who comes
1281 up with the meaning of an expression---who ``interprets'' it.
1283 You can also evaluate an atom that is not part of a list---one that is
1284 not surrounded by parentheses; again, the Lisp interpreter translates
1285 from the humanly readable expression to the language of the computer.
1286 But before discussing this (@pxref{Variables}), we will discuss what the
1287 Lisp interpreter does when you make an error.
1290 @section Generate an Error Message
1291 @cindex Generate an error message
1292 @cindex Error message generation
1294 Partly so you won't worry if you do it accidentally, we will now give
1295 a command to the Lisp interpreter that generates an error message.
1296 This is a harmless activity; and indeed, we will often try to generate
1297 error messages intentionally. Once you understand the jargon, error
1298 messages can be informative. Instead of being called ``error''
1299 messages, they should be called ``help'' messages. They are like
1300 signposts to a traveler in a strange country; deciphering them can be
1301 hard, but once understood, they can point the way.
1303 The error message is generated by a built-in GNU Emacs debugger. We
1304 will `enter the debugger'. You get out of the debugger by typing @code{q}.
1306 What we will do is evaluate a list that is not quoted and does not
1307 have a meaningful command as its first element. Here is a list almost
1308 exactly the same as the one we just used, but without the single-quote
1309 in front of it. Position the cursor right after it and type @kbd{C-x
1313 (this is an unquoted list)
1318 What you see depends on which version of Emacs you are running. GNU
1319 Emacs version 22 provides more information than version 20 and before.
1320 First, the more recent result of generating an error; then the
1321 earlier, version 20 result.
1325 In GNU Emacs version 22, a @file{*Backtrace*} window will open up and
1326 you will see the following in it:
1329 A @file{*Backtrace*} window will open up and you should see the
1334 ---------- Buffer: *Backtrace* ----------
1335 Debugger entered--Lisp error: (void-function this)
1336 (this is an unquoted list)
1337 eval((this is an unquoted list))
1338 eval-last-sexp-1(nil)
1340 call-interactively(eval-last-sexp)
1341 ---------- Buffer: *Backtrace* ----------
1347 Your cursor will be in this window (you may have to wait a few seconds
1348 before it becomes visible). To quit the debugger and make the
1349 debugger window go away, type:
1356 Please type @kbd{q} right now, so you become confident that you can
1357 get out of the debugger. Then, type @kbd{C-x C-e} again to re-enter
1360 @cindex @samp{function} defined
1361 Based on what we already know, we can almost read this error message.
1363 You read the @file{*Backtrace*} buffer from the bottom up; it tells
1364 you what Emacs did. When you typed @kbd{C-x C-e}, you made an
1365 interactive call to the command @code{eval-last-sexp}. @code{eval} is
1366 an abbreviation for `evaluate' and @code{sexp} is an abbreviation for
1367 `symbolic expression'. The command means `evaluate last symbolic
1368 expression', which is the expression just before your cursor.
1370 Each line above tells you what the Lisp interpreter evaluated next.
1371 The most recent action is at the top. The buffer is called the
1372 @file{*Backtrace*} buffer because it enables you to track Emacs
1376 At the top of the @file{*Backtrace*} buffer, you see the line:
1379 Debugger entered--Lisp error: (void-function this)
1383 The Lisp interpreter tried to evaluate the first atom of the list, the
1384 word @samp{this}. It is this action that generated the error message
1385 @samp{void-function this}.
1387 The message contains the words @samp{void-function} and @samp{this}.
1389 @cindex @samp{function} defined
1390 The word @samp{function} was mentioned once before. It is a very
1391 important word. For our purposes, we can define it by saying that a
1392 @dfn{function} is a set of instructions to the computer that tell the
1393 computer to do something.
1395 Now we can begin to understand the error message: @samp{void-function
1396 this}. The function (that is, the word @samp{this}) does not have a
1397 definition of any set of instructions for the computer to carry out.
1399 The slightly odd word, @samp{void-function}, is designed to cover the
1400 way Emacs Lisp is implemented, which is that when a symbol does not
1401 have a function definition attached to it, the place that should
1402 contain the instructions is `void'.
1404 On the other hand, since we were able to add 2 plus 2 successfully, by
1405 evaluating @code{(+ 2 2)}, we can infer that the symbol @code{+} must
1406 have a set of instructions for the computer to obey and those
1407 instructions must be to add the numbers that follow the @code{+}.
1409 It is possible to prevent Emacs entering the debugger in cases like
1410 this. We do not explain how to do that here, but we will mention what
1411 the result looks like, because you may encounter a similar situation
1412 if there is a bug in some Emacs code that you are using. In such
1413 cases, you will see only one line of error message; it will appear in
1414 the echo area and look like this:
1417 Symbol's function definition is void:@: this
1422 (Also, your terminal may beep at you---some do, some don't; and others
1423 blink. This is just a device to get your attention.)
1425 The message goes away as soon as you type a key, even just to
1428 We know the meaning of the word @samp{Symbol}. It refers to the first
1429 atom of the list, the word @samp{this}. The word @samp{function}
1430 refers to the instructions that tell the computer what to do.
1431 (Technically, the symbol tells the computer where to find the
1432 instructions, but this is a complication we can ignore for the
1435 The error message can be understood: @samp{Symbol's function
1436 definition is void:@: this}. The symbol (that is, the word
1437 @samp{this}) lacks instructions for the computer to carry out.
1439 @node Names & Definitions
1440 @section Symbol Names and Function Definitions
1441 @cindex Symbol names
1443 We can articulate another characteristic of Lisp based on what we have
1444 discussed so far---an important characteristic: a symbol, like
1445 @code{+}, is not itself the set of instructions for the computer to
1446 carry out. Instead, the symbol is used, perhaps temporarily, as a way
1447 of locating the definition or set of instructions. What we see is the
1448 name through which the instructions can be found. Names of people
1449 work the same way. I can be referred to as @samp{Bob}; however, I am
1450 not the letters @samp{B}, @samp{o}, @samp{b} but am, or was, the
1451 consciousness consistently associated with a particular life-form.
1452 The name is not me, but it can be used to refer to me.
1454 In Lisp, one set of instructions can be attached to several names.
1455 For example, the computer instructions for adding numbers can be
1456 linked to the symbol @code{plus} as well as to the symbol @code{+}
1457 (and are in some dialects of Lisp). Among humans, I can be referred
1458 to as @samp{Robert} as well as @samp{Bob} and by other words as well.
1460 On the other hand, a symbol can have only one function definition
1461 attached to it at a time. Otherwise, the computer would be confused as
1462 to which definition to use. If this were the case among people, only
1463 one person in the world could be named @samp{Bob}. However, the function
1464 definition to which the name refers can be changed readily.
1465 (@xref{Install, , Install a Function Definition}.)
1467 Since Emacs Lisp is large, it is customary to name symbols in a way
1468 that identifies the part of Emacs to which the function belongs.
1469 Thus, all the names for functions that deal with Texinfo start with
1470 @samp{texinfo-} and those for functions that deal with reading mail
1471 start with @samp{rmail-}.
1473 @node Lisp Interpreter
1474 @section The Lisp Interpreter
1475 @cindex Lisp interpreter, what it does
1476 @cindex Interpreter, what it does
1478 Based on what we have seen, we can now start to figure out what the
1479 Lisp interpreter does when we command it to evaluate a list.
1480 First, it looks to see whether there is a quote before the list; if
1481 there is, the interpreter just gives us the list. On the other
1482 hand, if there is no quote, the interpreter looks at the first element
1483 in the list and sees whether it has a function definition. If it does,
1484 the interpreter carries out the instructions in the function definition.
1485 Otherwise, the interpreter prints an error message.
1487 This is how Lisp works. Simple. There are added complications which we
1488 will get to in a minute, but these are the fundamentals. Of course, to
1489 write Lisp programs, you need to know how to write function definitions
1490 and attach them to names, and how to do this without confusing either
1491 yourself or the computer.
1494 * Complications:: Variables, Special forms, Lists within.
1495 * Byte Compiling:: Specially processing code for speed.
1500 @unnumberedsubsec Complications
1503 Now, for the first complication. In addition to lists, the Lisp
1504 interpreter can evaluate a symbol that is not quoted and does not have
1505 parentheses around it. The Lisp interpreter will attempt to determine
1506 the symbol's value as a @dfn{variable}. This situation is described
1507 in the section on variables. (@xref{Variables}.)
1509 @cindex Special form
1510 The second complication occurs because some functions are unusual and
1511 do not work in the usual manner. Those that don't are called
1512 @dfn{special forms}. They are used for special jobs, like defining a
1513 function, and there are not many of them. In the next few chapters,
1514 you will be introduced to several of the more important special forms.
1516 As well as special forms, there are also @dfn{macros}. A macro
1517 is a construct defined in Lisp, which differs from a function in that it
1518 translates a Lisp expression into another expression that is to be
1519 evaluated in place of the original expression. (@xref{Lisp macro}.)
1521 For the purposes of this introduction, you do not need to worry too much
1522 about whether something is a special form, macro, or ordinary function.
1523 For example, @code{if} is a special form (@pxref{if}), but @code{when}
1524 is a macro (@pxref{Lisp macro}). In earlier versions of Emacs,
1525 @code{defun} was a special form, but now it is a macro (@pxref{defun}).
1526 It still behaves in the same way.
1528 The final complication is this: if the function that the
1529 Lisp interpreter is looking at is not a special form, and if it is part
1530 of a list, the Lisp interpreter looks to see whether the list has a list
1531 inside of it. If there is an inner list, the Lisp interpreter first
1532 figures out what it should do with the inside list, and then it works on
1533 the outside list. If there is yet another list embedded inside the
1534 inner list, it works on that one first, and so on. It always works on
1535 the innermost list first. The interpreter works on the innermost list
1536 first, to evaluate the result of that list. The result may be
1537 used by the enclosing expression.
1539 Otherwise, the interpreter works left to right, from one expression to
1542 @node Byte Compiling
1543 @subsection Byte Compiling
1544 @cindex Byte compiling
1546 One other aspect of interpreting: the Lisp interpreter is able to
1547 interpret two kinds of entity: humanly readable code, on which we will
1548 focus exclusively, and specially processed code, called @dfn{byte
1549 compiled} code, which is not humanly readable. Byte compiled code
1550 runs faster than humanly readable code.
1552 You can transform humanly readable code into byte compiled code by
1553 running one of the compile commands such as @code{byte-compile-file}.
1554 Byte compiled code is usually stored in a file that ends with a
1555 @file{.elc} extension rather than a @file{.el} extension. You will
1556 see both kinds of file in the @file{emacs/lisp} directory; the files
1557 to read are those with @file{.el} extensions.
1559 As a practical matter, for most things you might do to customize or
1560 extend Emacs, you do not need to byte compile; and I will not discuss
1561 the topic here. @xref{Byte Compilation, , Byte Compilation, elisp,
1562 The GNU Emacs Lisp Reference Manual}, for a full description of byte
1569 When the Lisp interpreter works on an expression, the term for the
1570 activity is called @dfn{evaluation}. We say that the interpreter
1571 `evaluates the expression'. I've used this term several times before.
1572 The word comes from its use in everyday language, `to ascertain the
1573 value or amount of; to appraise', according to @cite{Webster's New
1574 Collegiate Dictionary}.
1577 * How the Interpreter Acts:: Returns and Side Effects...
1578 * Evaluating Inner Lists:: Lists within lists...
1582 @node How the Interpreter Acts
1583 @unnumberedsubsec How the Lisp Interpreter Acts
1586 @cindex @samp{returned value} explained
1587 After evaluating an expression, the Lisp interpreter will most likely
1588 @dfn{return} the value that the computer produces by carrying out the
1589 instructions it found in the function definition, or perhaps it will
1590 give up on that function and produce an error message. (The interpreter
1591 may also find itself tossed, so to speak, to a different function or it
1592 may attempt to repeat continually what it is doing for ever and ever in
1593 what is called an `infinite loop'. These actions are less common; and
1594 we can ignore them.) Most frequently, the interpreter returns a value.
1596 @cindex @samp{side effect} defined
1597 At the same time the interpreter returns a value, it may do something
1598 else as well, such as move a cursor or copy a file; this other kind of
1599 action is called a @dfn{side effect}. Actions that we humans think are
1600 important, such as printing results, are often ``side effects'' to the
1601 Lisp interpreter. The jargon can sound peculiar, but it turns out that
1602 it is fairly easy to learn to use side effects.
1604 In summary, evaluating a symbolic expression most commonly causes the
1605 Lisp interpreter to return a value and perhaps carry out a side effect;
1606 or else produce an error.
1608 @node Evaluating Inner Lists
1609 @subsection Evaluating Inner Lists
1610 @cindex Inner list evaluation
1611 @cindex Evaluating inner lists
1613 If evaluation applies to a list that is inside another list, the outer
1614 list may use the value returned by the first evaluation as information
1615 when the outer list is evaluated. This explains why inner expressions
1616 are evaluated first: the values they return are used by the outer
1620 We can investigate this process by evaluating another addition example.
1621 Place your cursor after the following expression and type @kbd{C-x C-e}:
1628 The number 8 will appear in the echo area.
1630 What happens is that the Lisp interpreter first evaluates the inner
1631 expression, @code{(+ 3 3)}, for which the value 6 is returned; then it
1632 evaluates the outer expression as if it were written @code{(+ 2 6)}, which
1633 returns the value 8. Since there are no more enclosing expressions to
1634 evaluate, the interpreter prints that value in the echo area.
1636 Now it is easy to understand the name of the command invoked by the
1637 keystrokes @kbd{C-x C-e}: the name is @code{eval-last-sexp}. The
1638 letters @code{sexp} are an abbreviation for `symbolic expression', and
1639 @code{eval} is an abbreviation for `evaluate'. The command means
1640 `evaluate last symbolic expression'.
1642 As an experiment, you can try evaluating the expression by putting the
1643 cursor at the beginning of the next line immediately following the
1644 expression, or inside the expression.
1647 Here is another copy of the expression:
1654 If you place the cursor at the beginning of the blank line that
1655 immediately follows the expression and type @kbd{C-x C-e}, you will
1656 still get the value 8 printed in the echo area. Now try putting the
1657 cursor inside the expression. If you put it right after the next to
1658 last parenthesis (so it appears to sit on top of the last parenthesis),
1659 you will get a 6 printed in the echo area! This is because the command
1660 evaluates the expression @code{(+ 3 3)}.
1662 Now put the cursor immediately after a number. Type @kbd{C-x C-e} and
1663 you will get the number itself. In Lisp, if you evaluate a number, you
1664 get the number itself---this is how numbers differ from symbols. If you
1665 evaluate a list starting with a symbol like @code{+}, you will get a
1666 value returned that is the result of the computer carrying out the
1667 instructions in the function definition attached to that name. If a
1668 symbol by itself is evaluated, something different happens, as we will
1669 see in the next section.
1675 In Emacs Lisp, a symbol can have a value attached to it just as it can
1676 have a function definition attached to it. The two are different.
1677 The function definition is a set of instructions that a computer will
1678 obey. A value, on the other hand, is something, such as number or a
1679 name, that can vary (which is why such a symbol is called a variable).
1680 The value of a symbol can be any expression in Lisp, such as a symbol,
1681 number, list, or string. A symbol that has a value is often called a
1684 A symbol can have both a function definition and a value attached to
1685 it at the same time. Or it can have just one or the other.
1686 The two are separate. This is somewhat similar
1687 to the way the name Cambridge can refer to the city in Massachusetts
1688 and have some information attached to the name as well, such as
1689 ``great programming center''.
1692 (Incidentally, in Emacs Lisp, a symbol can have two
1693 other things attached to it, too: a property list and a documentation
1694 string; these are discussed later.)
1697 Another way to think about this is to imagine a symbol as being a chest
1698 of drawers. The function definition is put in one drawer, the value in
1699 another, and so on. What is put in the drawer holding the value can be
1700 changed without affecting the contents of the drawer holding the
1701 function definition, and vice-verse.
1704 * fill-column Example::
1705 * Void Function:: The error message for a symbol
1707 * Void Variable:: The error message for a symbol without a value.
1711 @node fill-column Example
1712 @unnumberedsubsec @code{fill-column}, an Example Variable
1715 @findex fill-column, @r{an example variable}
1716 @cindex Example variable, @code{fill-column}
1717 @cindex Variable, example of, @code{fill-column}
1718 The variable @code{fill-column} illustrates a symbol with a value
1719 attached to it: in every GNU Emacs buffer, this symbol is set to some
1720 value, usually 72 or 70, but sometimes to some other value. To find the
1721 value of this symbol, evaluate it by itself. If you are reading this in
1722 Info inside of GNU Emacs, you can do this by putting the cursor after
1723 the symbol and typing @kbd{C-x C-e}:
1730 After I typed @kbd{C-x C-e}, Emacs printed the number 72 in my echo
1731 area. This is the value for which @code{fill-column} is set for me as I
1732 write this. It may be different for you in your Info buffer. Notice
1733 that the value returned as a variable is printed in exactly the same way
1734 as the value returned by a function carrying out its instructions. From
1735 the point of view of the Lisp interpreter, a value returned is a value
1736 returned. What kind of expression it came from ceases to matter once
1739 A symbol can have any value attached to it or, to use the jargon, we can
1740 @dfn{bind} the variable to a value: to a number, such as 72; to a
1741 string, @code{"such as this"}; to a list, such as @code{(spruce pine
1742 oak)}; we can even bind a variable to a function definition.
1744 A symbol can be bound to a value in several ways. @xref{set & setq, ,
1745 Setting the Value of a Variable}, for information about one way to do
1749 @subsection Error Message for a Symbol Without a Function
1750 @cindex Symbol without function error
1751 @cindex Error for symbol without function
1753 When we evaluated @code{fill-column} to find its value as a variable,
1754 we did not place parentheses around the word. This is because we did
1755 not intend to use it as a function name.
1757 If @code{fill-column} were the first or only element of a list, the
1758 Lisp interpreter would attempt to find the function definition
1759 attached to it. But @code{fill-column} has no function definition.
1760 Try evaluating this:
1768 You will create a @file{*Backtrace*} buffer that says:
1772 ---------- Buffer: *Backtrace* ----------
1773 Debugger entered--Lisp error: (void-function fill-column)
1776 eval-last-sexp-1(nil)
1778 call-interactively(eval-last-sexp)
1779 ---------- Buffer: *Backtrace* ----------
1784 (Remember, to quit the debugger and make the debugger window go away,
1785 type @kbd{q} in the @file{*Backtrace*} buffer.)
1789 In GNU Emacs 20 and before, you will produce an error message that says:
1792 Symbol's function definition is void:@: fill-column
1796 (The message will go away as soon as you move the cursor or type
1801 @subsection Error Message for a Symbol Without a Value
1802 @cindex Symbol without value error
1803 @cindex Error for symbol without value
1805 If you attempt to evaluate a symbol that does not have a value bound to
1806 it, you will receive an error message. You can see this by
1807 experimenting with our 2 plus 2 addition. In the following expression,
1808 put your cursor right after the @code{+}, before the first number 2,
1817 In GNU Emacs 22, you will create a @file{*Backtrace*} buffer that
1822 ---------- Buffer: *Backtrace* ----------
1823 Debugger entered--Lisp error: (void-variable +)
1825 eval-last-sexp-1(nil)
1827 call-interactively(eval-last-sexp)
1828 ---------- Buffer: *Backtrace* ----------
1833 (Again, you can quit the debugger by
1834 typing @kbd{q} in the @file{*Backtrace*} buffer.)
1836 This backtrace is different from the very first error message we saw,
1837 which said, @samp{Debugger entered--Lisp error: (void-function this)}.
1838 In this case, the function does not have a value as a variable; while
1839 in the other error message, the function (the word `this') did not
1842 In this experiment with the @code{+}, what we did was cause the Lisp
1843 interpreter to evaluate the @code{+} and look for the value of the
1844 variable instead of the function definition. We did this by placing the
1845 cursor right after the symbol rather than after the parenthesis of the
1846 enclosing list as we did before. As a consequence, the Lisp interpreter
1847 evaluated the preceding s-expression, which in this case was
1850 Since @code{+} does not have a value bound to it, just the function
1851 definition, the error message reported that the symbol's value as a
1856 In GNU Emacs version 20 and before, your error message will say:
1859 Symbol's value as variable is void:@: +
1863 The meaning is the same as in GNU Emacs 22.
1869 @cindex Passing information to functions
1871 To see how information is passed to functions, let's look again at
1872 our old standby, the addition of two plus two. In Lisp, this is written
1879 If you evaluate this expression, the number 4 will appear in your echo
1880 area. What the Lisp interpreter does is add the numbers that follow
1883 @cindex @samp{argument} defined
1884 The numbers added by @code{+} are called the @dfn{arguments} of the
1885 function @code{+}. These numbers are the information that is given to
1886 or @dfn{passed} to the function.
1888 The word `argument' comes from the way it is used in mathematics and
1889 does not refer to a disputation between two people; instead it refers to
1890 the information presented to the function, in this case, to the
1891 @code{+}. In Lisp, the arguments to a function are the atoms or lists
1892 that follow the function. The values returned by the evaluation of
1893 these atoms or lists are passed to the function. Different functions
1894 require different numbers of arguments; some functions require none at
1895 all.@footnote{It is curious to track the path by which the word `argument'
1896 came to have two different meanings, one in mathematics and the other in
1897 everyday English. According to the @cite{Oxford English Dictionary},
1898 the word derives from the Latin for @samp{to make clear, prove}; thus it
1899 came to mean, by one thread of derivation, `the evidence offered as
1900 proof', which is to say, `the information offered', which led to its
1901 meaning in Lisp. But in the other thread of derivation, it came to mean
1902 `to assert in a manner against which others may make counter
1903 assertions', which led to the meaning of the word as a disputation.
1904 (Note here that the English word has two different definitions attached
1905 to it at the same time. By contrast, in Emacs Lisp, a symbol cannot
1906 have two different function definitions at the same time.)}
1909 * Data types:: Types of data passed to a function.
1910 * Args as Variable or List:: An argument can be the value
1911 of a variable or list.
1912 * Variable Number of Arguments:: Some functions may take a
1913 variable number of arguments.
1914 * Wrong Type of Argument:: Passing an argument of the wrong type
1916 * message:: A useful function for sending messages.
1920 @subsection Arguments' Data Types
1922 @cindex Types of data
1923 @cindex Arguments' data types
1925 The type of data that should be passed to a function depends on what
1926 kind of information it uses. The arguments to a function such as
1927 @code{+} must have values that are numbers, since @code{+} adds numbers.
1928 Other functions use different kinds of data for their arguments.
1932 For example, the @code{concat} function links together or unites two or
1933 more strings of text to produce a string. The arguments are strings.
1934 Concatenating the two character strings @code{abc}, @code{def} produces
1935 the single string @code{abcdef}. This can be seen by evaluating the
1939 (concat "abc" "def")
1943 The value produced by evaluating this expression is @code{"abcdef"}.
1945 A function such as @code{substring} uses both a string and numbers as
1946 arguments. The function returns a part of the string, a substring of
1947 the first argument. This function takes three arguments. Its first
1948 argument is the string of characters, the second and third arguments are
1949 numbers that indicate the beginning and end of the substring. The
1950 numbers are a count of the number of characters (including spaces and
1951 punctuation) from the beginning of the string.
1954 For example, if you evaluate the following:
1957 (substring "The quick brown fox jumped." 16 19)
1961 you will see @code{"fox"} appear in the echo area. The arguments are the
1962 string and the two numbers.
1964 Note that the string passed to @code{substring} is a single atom even
1965 though it is made up of several words separated by spaces. Lisp counts
1966 everything between the two quotation marks as part of the string,
1967 including the spaces. You can think of the @code{substring} function as
1968 a kind of `atom smasher' since it takes an otherwise indivisible atom
1969 and extracts a part. However, @code{substring} is only able to extract
1970 a substring from an argument that is a string, not from another type of
1971 atom such as a number or symbol.
1973 @node Args as Variable or List
1974 @subsection An Argument as the Value of a Variable or List
1976 An argument can be a symbol that returns a value when it is evaluated.
1977 For example, when the symbol @code{fill-column} by itself is evaluated,
1978 it returns a number. This number can be used in an addition.
1981 Position the cursor after the following expression and type @kbd{C-x
1989 The value will be a number two more than what you get by evaluating
1990 @code{fill-column} alone. For me, this is 74, because my value of
1991 @code{fill-column} is 72.
1993 As we have just seen, an argument can be a symbol that returns a value
1994 when evaluated. In addition, an argument can be a list that returns a
1995 value when it is evaluated. For example, in the following expression,
1996 the arguments to the function @code{concat} are the strings
1997 @w{@code{"The "}} and @w{@code{" red foxes."}} and the list
1998 @code{(number-to-string (+ 2 fill-column))}.
2000 @c For GNU Emacs 22, need number-to-string
2002 (concat "The " (number-to-string (+ 2 fill-column)) " red foxes.")
2006 If you evaluate this expression---and if, as with my Emacs,
2007 @code{fill-column} evaluates to 72---@code{"The 74 red foxes."} will
2008 appear in the echo area. (Note that you must put spaces after the
2009 word @samp{The} and before the word @samp{red} so they will appear in
2010 the final string. The function @code{number-to-string} converts the
2011 integer that the addition function returns to a string.
2012 @code{number-to-string} is also known as @code{int-to-string}.)
2014 @node Variable Number of Arguments
2015 @subsection Variable Number of Arguments
2016 @cindex Variable number of arguments
2017 @cindex Arguments, variable number of
2019 Some functions, such as @code{concat}, @code{+} or @code{*}, take any
2020 number of arguments. (The @code{*} is the symbol for multiplication.)
2021 This can be seen by evaluating each of the following expressions in
2022 the usual way. What you will see in the echo area is printed in this
2023 text after @samp{@result{}}, which you may read as `evaluates to'.
2026 In the first set, the functions have no arguments:
2037 In this set, the functions have one argument each:
2048 In this set, the functions have three arguments each:
2052 (+ 3 4 5) @result{} 12
2054 (* 3 4 5) @result{} 60
2058 @node Wrong Type of Argument
2059 @subsection Using the Wrong Type Object as an Argument
2060 @cindex Wrong type of argument
2061 @cindex Argument, wrong type of
2063 When a function is passed an argument of the wrong type, the Lisp
2064 interpreter produces an error message. For example, the @code{+}
2065 function expects the values of its arguments to be numbers. As an
2066 experiment we can pass it the quoted symbol @code{hello} instead of a
2067 number. Position the cursor after the following expression and type
2075 When you do this you will generate an error message. What has happened
2076 is that @code{+} has tried to add the 2 to the value returned by
2077 @code{'hello}, but the value returned by @code{'hello} is the symbol
2078 @code{hello}, not a number. Only numbers can be added. So @code{+}
2079 could not carry out its addition.
2082 You will create and enter a @file{*Backtrace*} buffer that says:
2087 ---------- Buffer: *Backtrace* ----------
2088 Debugger entered--Lisp error:
2089 (wrong-type-argument number-or-marker-p hello)
2091 eval((+ 2 (quote hello)))
2092 eval-last-sexp-1(nil)
2094 call-interactively(eval-last-sexp)
2095 ---------- Buffer: *Backtrace* ----------
2100 As usual, the error message tries to be helpful and makes sense after you
2101 learn how to read it.@footnote{@code{(quote hello)} is an expansion of
2102 the abbreviation @code{'hello}.}
2104 The first part of the error message is straightforward; it says
2105 @samp{wrong type argument}. Next comes the mysterious jargon word
2106 @w{@samp{number-or-marker-p}}. This word is trying to tell you what
2107 kind of argument the @code{+} expected.
2109 The symbol @code{number-or-marker-p} says that the Lisp interpreter is
2110 trying to determine whether the information presented it (the value of
2111 the argument) is a number or a marker (a special object representing a
2112 buffer position). What it does is test to see whether the @code{+} is
2113 being given numbers to add. It also tests to see whether the
2114 argument is something called a marker, which is a specific feature of
2115 Emacs Lisp. (In Emacs, locations in a buffer are recorded as markers.
2116 When the mark is set with the @kbd{C-@@} or @kbd{C-@key{SPC}} command,
2117 its position is kept as a marker. The mark can be considered a
2118 number---the number of characters the location is from the beginning
2119 of the buffer.) In Emacs Lisp, @code{+} can be used to add the
2120 numeric value of marker positions as numbers.
2122 The @samp{p} of @code{number-or-marker-p} is the embodiment of a
2123 practice started in the early days of Lisp programming. The @samp{p}
2124 stands for `predicate'. In the jargon used by the early Lisp
2125 researchers, a predicate refers to a function to determine whether some
2126 property is true or false. So the @samp{p} tells us that
2127 @code{number-or-marker-p} is the name of a function that determines
2128 whether it is true or false that the argument supplied is a number or
2129 a marker. Other Lisp symbols that end in @samp{p} include @code{zerop},
2130 a function that tests whether its argument has the value of zero, and
2131 @code{listp}, a function that tests whether its argument is a list.
2133 Finally, the last part of the error message is the symbol @code{hello}.
2134 This is the value of the argument that was passed to @code{+}. If the
2135 addition had been passed the correct type of object, the value passed
2136 would have been a number, such as 37, rather than a symbol like
2137 @code{hello}. But then you would not have got the error message.
2141 In GNU Emacs version 20 and before, the echo area displays an error
2145 Wrong type argument:@: number-or-marker-p, hello
2148 This says, in different words, the same as the top line of the
2149 @file{*Backtrace*} buffer.
2153 @subsection The @code{message} Function
2156 Like @code{+}, the @code{message} function takes a variable number of
2157 arguments. It is used to send messages to the user and is so useful
2158 that we will describe it here.
2161 A message is printed in the echo area. For example, you can print a
2162 message in your echo area by evaluating the following list:
2165 (message "This message appears in the echo area!")
2168 The whole string between double quotation marks is a single argument
2169 and is printed @i{in toto}. (Note that in this example, the message
2170 itself will appear in the echo area within double quotes; that is
2171 because you see the value returned by the @code{message} function. In
2172 most uses of @code{message} in programs that you write, the text will
2173 be printed in the echo area as a side-effect, without the quotes.
2174 @xref{multiply-by-seven in detail, , @code{multiply-by-seven} in
2175 detail}, for an example of this.)
2177 However, if there is a @samp{%s} in the quoted string of characters, the
2178 @code{message} function does not print the @samp{%s} as such, but looks
2179 to the argument that follows the string. It evaluates the second
2180 argument and prints the value at the location in the string where the
2184 You can see this by positioning the cursor after the following
2185 expression and typing @kbd{C-x C-e}:
2188 (message "The name of this buffer is: %s." (buffer-name))
2192 In Info, @code{"The name of this buffer is: *info*."} will appear in the
2193 echo area. The function @code{buffer-name} returns the name of the
2194 buffer as a string, which the @code{message} function inserts in place
2197 To print a value as an integer, use @samp{%d} in the same way as
2198 @samp{%s}. For example, to print a message in the echo area that
2199 states the value of the @code{fill-column}, evaluate the following:
2202 (message "The value of fill-column is %d." fill-column)
2206 On my system, when I evaluate this list, @code{"The value of
2207 fill-column is 72."} appears in my echo area@footnote{Actually, you
2208 can use @code{%s} to print a number. It is non-specific. @code{%d}
2209 prints only the part of a number left of a decimal point, and not
2210 anything that is not a number.}.
2212 If there is more than one @samp{%s} in the quoted string, the value of
2213 the first argument following the quoted string is printed at the
2214 location of the first @samp{%s} and the value of the second argument is
2215 printed at the location of the second @samp{%s}, and so on.
2218 For example, if you evaluate the following,
2222 (message "There are %d %s in the office!"
2223 (- fill-column 14) "pink elephants")
2228 a rather whimsical message will appear in your echo area. On my system
2229 it says, @code{"There are 58 pink elephants in the office!"}.
2231 The expression @code{(- fill-column 14)} is evaluated and the resulting
2232 number is inserted in place of the @samp{%d}; and the string in double
2233 quotes, @code{"pink elephants"}, is treated as a single argument and
2234 inserted in place of the @samp{%s}. (That is to say, a string between
2235 double quotes evaluates to itself, like a number.)
2237 Finally, here is a somewhat complex example that not only illustrates
2238 the computation of a number, but also shows how you can use an
2239 expression within an expression to generate the text that is substituted
2244 (message "He saw %d %s"
2248 "The quick brown foxes jumped." 16 21)
2253 In this example, @code{message} has three arguments: the string,
2254 @code{"He saw %d %s"}, the expression, @code{(- fill-column 32)}, and
2255 the expression beginning with the function @code{concat}. The value
2256 resulting from the evaluation of @code{(- fill-column 32)} is inserted
2257 in place of the @samp{%d}; and the value returned by the expression
2258 beginning with @code{concat} is inserted in place of the @samp{%s}.
2260 When your fill column is 70 and you evaluate the expression, the
2261 message @code{"He saw 38 red foxes leaping."} appears in your echo
2265 @section Setting the Value of a Variable
2266 @cindex Variable, setting value
2267 @cindex Setting value of variable
2269 @cindex @samp{bind} defined
2270 There are several ways by which a variable can be given a value. One of
2271 the ways is to use either the function @code{set} or the function
2272 @code{setq}. Another way is to use @code{let} (@pxref{let}). (The
2273 jargon for this process is to @dfn{bind} a variable to a value.)
2275 The following sections not only describe how @code{set} and @code{setq}
2276 work but also illustrate how arguments are passed.
2279 * Using set:: Setting values.
2280 * Using setq:: Setting a quoted value.
2281 * Counting:: Using @code{setq} to count.
2285 @subsection Using @code{set}
2288 To set the value of the symbol @code{flowers} to the list @code{'(rose
2289 violet daisy buttercup)}, evaluate the following expression by
2290 positioning the cursor after the expression and typing @kbd{C-x C-e}.
2293 (set 'flowers '(rose violet daisy buttercup))
2297 The list @code{(rose violet daisy buttercup)} will appear in the echo
2298 area. This is what is @emph{returned} by the @code{set} function. As a
2299 side effect, the symbol @code{flowers} is bound to the list; that is,
2300 the symbol @code{flowers}, which can be viewed as a variable, is given
2301 the list as its value. (This process, by the way, illustrates how a
2302 side effect to the Lisp interpreter, setting the value, can be the
2303 primary effect that we humans are interested in. This is because every
2304 Lisp function must return a value if it does not get an error, but it
2305 will only have a side effect if it is designed to have one.)
2307 After evaluating the @code{set} expression, you can evaluate the symbol
2308 @code{flowers} and it will return the value you just set. Here is the
2309 symbol. Place your cursor after it and type @kbd{C-x C-e}.
2316 When you evaluate @code{flowers}, the list
2317 @code{(rose violet daisy buttercup)} appears in the echo area.
2319 Incidentally, if you evaluate @code{'flowers}, the variable with a quote
2320 in front of it, what you will see in the echo area is the symbol itself,
2321 @code{flowers}. Here is the quoted symbol, so you can try this:
2327 Note also, that when you use @code{set}, you need to quote both
2328 arguments to @code{set}, unless you want them evaluated. Since we do
2329 not want either argument evaluated, neither the variable
2330 @code{flowers} nor the list @code{(rose violet daisy buttercup)}, both
2331 are quoted. (When you use @code{set} without quoting its first
2332 argument, the first argument is evaluated before anything else is
2333 done. If you did this and @code{flowers} did not have a value
2334 already, you would get an error message that the @samp{Symbol's value
2335 as variable is void}; on the other hand, if @code{flowers} did return
2336 a value after it was evaluated, the @code{set} would attempt to set
2337 the value that was returned. There are situations where this is the
2338 right thing for the function to do; but such situations are rare.)
2341 @subsection Using @code{setq}
2344 As a practical matter, you almost always quote the first argument to
2345 @code{set}. The combination of @code{set} and a quoted first argument
2346 is so common that it has its own name: the special form @code{setq}.
2347 This special form is just like @code{set} except that the first argument
2348 is quoted automatically, so you don't need to type the quote mark
2349 yourself. Also, as an added convenience, @code{setq} permits you to set
2350 several different variables to different values, all in one expression.
2352 To set the value of the variable @code{carnivores} to the list
2353 @code{'(lion tiger leopard)} using @code{setq}, the following expression
2357 (setq carnivores '(lion tiger leopard))
2361 This is exactly the same as using @code{set} except the first argument
2362 is automatically quoted by @code{setq}. (The @samp{q} in @code{setq}
2363 means @code{quote}.)
2366 With @code{set}, the expression would look like this:
2369 (set 'carnivores '(lion tiger leopard))
2372 Also, @code{setq} can be used to assign different values to
2373 different variables. The first argument is bound to the value
2374 of the second argument, the third argument is bound to the value of the
2375 fourth argument, and so on. For example, you could use the following to
2376 assign a list of trees to the symbol @code{trees} and a list of herbivores
2377 to the symbol @code{herbivores}:
2381 (setq trees '(pine fir oak maple)
2382 herbivores '(gazelle antelope zebra))
2387 (The expression could just as well have been on one line, but it might
2388 not have fit on a page; and humans find it easier to read nicely
2391 Although I have been using the term `assign', there is another way of
2392 thinking about the workings of @code{set} and @code{setq}; and that is to
2393 say that @code{set} and @code{setq} make the symbol @emph{point} to the
2394 list. This latter way of thinking is very common and in forthcoming
2395 chapters we shall come upon at least one symbol that has `pointer' as
2396 part of its name. The name is chosen because the symbol has a value,
2397 specifically a list, attached to it; or, expressed another way,
2398 the symbol is set to ``point'' to the list.
2401 @subsection Counting
2404 Here is an example that shows how to use @code{setq} in a counter. You
2405 might use this to count how many times a part of your program repeats
2406 itself. First set a variable to zero; then add one to the number each
2407 time the program repeats itself. To do this, you need a variable that
2408 serves as a counter, and two expressions: an initial @code{setq}
2409 expression that sets the counter variable to zero; and a second
2410 @code{setq} expression that increments the counter each time it is
2415 (setq counter 0) ; @r{Let's call this the initializer.}
2417 (setq counter (+ counter 1)) ; @r{This is the incrementer.}
2419 counter ; @r{This is the counter.}
2424 (The text following the @samp{;} are comments. @xref{Change a
2425 defun, , Change a Function Definition}.)
2427 If you evaluate the first of these expressions, the initializer,
2428 @code{(setq counter 0)}, and then evaluate the third expression,
2429 @code{counter}, the number @code{0} will appear in the echo area. If
2430 you then evaluate the second expression, the incrementer, @code{(setq
2431 counter (+ counter 1))}, the counter will get the value 1. So if you
2432 again evaluate @code{counter}, the number @code{1} will appear in the
2433 echo area. Each time you evaluate the second expression, the value of
2434 the counter will be incremented.
2436 When you evaluate the incrementer, @code{(setq counter (+ counter 1))},
2437 the Lisp interpreter first evaluates the innermost list; this is the
2438 addition. In order to evaluate this list, it must evaluate the variable
2439 @code{counter} and the number @code{1}. When it evaluates the variable
2440 @code{counter}, it receives its current value. It passes this value and
2441 the number @code{1} to the @code{+} which adds them together. The sum
2442 is then returned as the value of the inner list and passed to the
2443 @code{setq} which sets the variable @code{counter} to this new value.
2444 Thus, the value of the variable, @code{counter}, is changed.
2449 Learning Lisp is like climbing a hill in which the first part is the
2450 steepest. You have now climbed the most difficult part; what remains
2451 becomes easier as you progress onwards.
2459 Lisp programs are made up of expressions, which are lists or single atoms.
2462 Lists are made up of zero or more atoms or inner lists, separated by whitespace and
2463 surrounded by parentheses. A list can be empty.
2466 Atoms are multi-character symbols, like @code{forward-paragraph}, single
2467 character symbols like @code{+}, strings of characters between double
2468 quotation marks, or numbers.
2471 A number evaluates to itself.
2474 A string between double quotes also evaluates to itself.
2477 When you evaluate a symbol by itself, its value is returned.
2480 When you evaluate a list, the Lisp interpreter looks at the first symbol
2481 in the list and then at the function definition bound to that symbol.
2482 Then the instructions in the function definition are carried out.
2485 A single quotation mark,
2492 , tells the Lisp interpreter that it should
2493 return the following expression as written, and not evaluate it as it
2494 would if the quote were not there.
2497 Arguments are the information passed to a function. The arguments to a
2498 function are computed by evaluating the rest of the elements of the list
2499 of which the function is the first element.
2502 A function always returns a value when it is evaluated (unless it gets
2503 an error); in addition, it may also carry out some action called a
2504 ``side effect''. In many cases, a function's primary purpose is to
2505 create a side effect.
2508 @node Error Message Exercises
2511 A few simple exercises:
2515 Generate an error message by evaluating an appropriate symbol that is
2516 not within parentheses.
2519 Generate an error message by evaluating an appropriate symbol that is
2520 between parentheses.
2523 Create a counter that increments by two rather than one.
2526 Write an expression that prints a message in the echo area when
2530 @node Practicing Evaluation
2531 @chapter Practicing Evaluation
2532 @cindex Practicing evaluation
2533 @cindex Evaluation practice
2535 Before learning how to write a function definition in Emacs Lisp, it is
2536 useful to spend a little time evaluating various expressions that have
2537 already been written. These expressions will be lists with the
2538 functions as their first (and often only) element. Since some of the
2539 functions associated with buffers are both simple and interesting, we
2540 will start with those. In this section, we will evaluate a few of
2541 these. In another section, we will study the code of several other
2542 buffer-related functions, to see how they were written.
2545 * How to Evaluate:: Typing editing commands or @kbd{C-x C-e}
2547 * Buffer Names:: Buffers and files are different.
2548 * Getting Buffers:: Getting a buffer itself, not merely its name.
2549 * Switching Buffers:: How to change to another buffer.
2550 * Buffer Size & Locations:: Where point is located and the size of
2552 * Evaluation Exercise::
2556 @node How to Evaluate
2557 @unnumberedsec How to Evaluate
2560 @i{Whenever you give an editing command} to Emacs Lisp, such as the
2561 command to move the cursor or to scroll the screen, @i{you are evaluating
2562 an expression,} the first element of which is a function. @i{This is
2565 @cindex @samp{interactive function} defined
2566 @cindex @samp{command} defined
2567 When you type keys, you cause the Lisp interpreter to evaluate an
2568 expression and that is how you get your results. Even typing plain text
2569 involves evaluating an Emacs Lisp function, in this case, one that uses
2570 @code{self-insert-command}, which simply inserts the character you
2571 typed. The functions you evaluate by typing keystrokes are called
2572 @dfn{interactive} functions, or @dfn{commands}; how you make a function
2573 interactive will be illustrated in the chapter on how to write function
2574 definitions. @xref{Interactive, , Making a Function Interactive}.
2576 In addition to typing keyboard commands, we have seen a second way to
2577 evaluate an expression: by positioning the cursor after a list and
2578 typing @kbd{C-x C-e}. This is what we will do in the rest of this
2579 section. There are other ways to evaluate an expression as well; these
2580 will be described as we come to them.
2582 Besides being used for practicing evaluation, the functions shown in the
2583 next few sections are important in their own right. A study of these
2584 functions makes clear the distinction between buffers and files, how to
2585 switch to a buffer, and how to determine a location within it.
2588 @section Buffer Names
2590 @findex buffer-file-name
2592 The two functions, @code{buffer-name} and @code{buffer-file-name}, show
2593 the difference between a file and a buffer. When you evaluate the
2594 following expression, @code{(buffer-name)}, the name of the buffer
2595 appears in the echo area. When you evaluate @code{(buffer-file-name)},
2596 the name of the file to which the buffer refers appears in the echo
2597 area. Usually, the name returned by @code{(buffer-name)} is the same as
2598 the name of the file to which it refers, and the name returned by
2599 @code{(buffer-file-name)} is the full path-name of the file.
2601 A file and a buffer are two different entities. A file is information
2602 recorded permanently in the computer (unless you delete it). A buffer,
2603 on the other hand, is information inside of Emacs that will vanish at
2604 the end of the editing session (or when you kill the buffer). Usually,
2605 a buffer contains information that you have copied from a file; we say
2606 the buffer is @dfn{visiting} that file. This copy is what you work on
2607 and modify. Changes to the buffer do not change the file, until you
2608 save the buffer. When you save the buffer, the buffer is copied to the file
2609 and is thus saved permanently.
2612 If you are reading this in Info inside of GNU Emacs, you can evaluate
2613 each of the following expressions by positioning the cursor after it and
2614 typing @kbd{C-x C-e}.
2625 When I do this in Info, the value returned by evaluating
2626 @code{(buffer-name)} is @file{"*info*"}, and the value returned by
2627 evaluating @code{(buffer-file-name)} is @file{nil}.
2629 On the other hand, while I am writing this document, the value
2630 returned by evaluating @code{(buffer-name)} is
2631 @file{"introduction.texinfo"}, and the value returned by evaluating
2632 @code{(buffer-file-name)} is
2633 @file{"/gnu/work/intro/introduction.texinfo"}.
2635 @cindex @code{nil}, history of word
2636 The former is the name of the buffer and the latter is the name of the
2637 file. In Info, the buffer name is @file{"*info*"}. Info does not
2638 point to any file, so the result of evaluating
2639 @code{(buffer-file-name)} is @file{nil}. The symbol @code{nil} is
2640 from the Latin word for `nothing'; in this case, it means that the
2641 buffer is not associated with any file. (In Lisp, @code{nil} is also
2642 used to mean `false' and is a synonym for the empty list, @code{()}.)
2644 When I am writing, the name of my buffer is
2645 @file{"introduction.texinfo"}. The name of the file to which it
2646 points is @file{"/gnu/work/intro/introduction.texinfo"}.
2648 (In the expressions, the parentheses tell the Lisp interpreter to
2649 treat @w{@code{buffer-name}} and @w{@code{buffer-file-name}} as
2650 functions; without the parentheses, the interpreter would attempt to
2651 evaluate the symbols as variables. @xref{Variables}.)
2653 In spite of the distinction between files and buffers, you will often
2654 find that people refer to a file when they mean a buffer and vice-verse.
2655 Indeed, most people say, ``I am editing a file,'' rather than saying,
2656 ``I am editing a buffer which I will soon save to a file.'' It is
2657 almost always clear from context what people mean. When dealing with
2658 computer programs, however, it is important to keep the distinction in mind,
2659 since the computer is not as smart as a person.
2661 @cindex Buffer, history of word
2662 The word `buffer', by the way, comes from the meaning of the word as a
2663 cushion that deadens the force of a collision. In early computers, a
2664 buffer cushioned the interaction between files and the computer's
2665 central processing unit. The drums or tapes that held a file and the
2666 central processing unit were pieces of equipment that were very
2667 different from each other, working at their own speeds, in spurts. The
2668 buffer made it possible for them to work together effectively.
2669 Eventually, the buffer grew from being an intermediary, a temporary
2670 holding place, to being the place where work is done. This
2671 transformation is rather like that of a small seaport that grew into a
2672 great city: once it was merely the place where cargo was warehoused
2673 temporarily before being loaded onto ships; then it became a business
2674 and cultural center in its own right.
2676 Not all buffers are associated with files. For example, a
2677 @file{*scratch*} buffer does not visit any file. Similarly, a
2678 @file{*Help*} buffer is not associated with any file.
2680 In the old days, when you lacked a @file{~/.emacs} file and started an
2681 Emacs session by typing the command @code{emacs} alone, without naming
2682 any files, Emacs started with the @file{*scratch*} buffer visible.
2683 Nowadays, you will see a splash screen. You can follow one of the
2684 commands suggested on the splash screen, visit a file, or press the
2685 spacebar to reach the @file{*scratch*} buffer.
2687 If you switch to the @file{*scratch*} buffer, type
2688 @code{(buffer-name)}, position the cursor after it, and then type
2689 @kbd{C-x C-e} to evaluate the expression. The name @code{"*scratch*"}
2690 will be returned and will appear in the echo area. @code{"*scratch*"}
2691 is the name of the buffer. When you type @code{(buffer-file-name)} in
2692 the @file{*scratch*} buffer and evaluate that, @code{nil} will appear
2693 in the echo area, just as it does when you evaluate
2694 @code{(buffer-file-name)} in Info.
2696 Incidentally, if you are in the @file{*scratch*} buffer and want the
2697 value returned by an expression to appear in the @file{*scratch*}
2698 buffer itself rather than in the echo area, type @kbd{C-u C-x C-e}
2699 instead of @kbd{C-x C-e}. This causes the value returned to appear
2700 after the expression. The buffer will look like this:
2703 (buffer-name)"*scratch*"
2707 You cannot do this in Info since Info is read-only and it will not allow
2708 you to change the contents of the buffer. But you can do this in any
2709 buffer you can edit; and when you write code or documentation (such as
2710 this book), this feature is very useful.
2712 @node Getting Buffers
2713 @section Getting Buffers
2714 @findex current-buffer
2715 @findex other-buffer
2716 @cindex Getting a buffer
2718 The @code{buffer-name} function returns the @emph{name} of the buffer;
2719 to get the buffer @emph{itself}, a different function is needed: the
2720 @code{current-buffer} function. If you use this function in code, what
2721 you get is the buffer itself.
2723 A name and the object or entity to which the name refers are different
2724 from each other. You are not your name. You are a person to whom
2725 others refer by name. If you ask to speak to George and someone hands you
2726 a card with the letters @samp{G}, @samp{e}, @samp{o}, @samp{r},
2727 @samp{g}, and @samp{e} written on it, you might be amused, but you would
2728 not be satisfied. You do not want to speak to the name, but to the
2729 person to whom the name refers. A buffer is similar: the name of the
2730 scratch buffer is @file{*scratch*}, but the name is not the buffer. To
2731 get a buffer itself, you need to use a function such as
2732 @code{current-buffer}.
2734 However, there is a slight complication: if you evaluate
2735 @code{current-buffer} in an expression on its own, as we will do here,
2736 what you see is a printed representation of the name of the buffer
2737 without the contents of the buffer. Emacs works this way for two
2738 reasons: the buffer may be thousands of lines long---too long to be
2739 conveniently displayed; and, another buffer may have the same contents
2740 but a different name, and it is important to distinguish between them.
2743 Here is an expression containing the function:
2750 If you evaluate this expression in Info in Emacs in the usual way,
2751 @file{#<buffer *info*>} will appear in the echo area. The special
2752 format indicates that the buffer itself is being returned, rather than
2755 Incidentally, while you can type a number or symbol into a program, you
2756 cannot do that with the printed representation of a buffer: the only way
2757 to get a buffer itself is with a function such as @code{current-buffer}.
2759 A related function is @code{other-buffer}. This returns the most
2760 recently selected buffer other than the one you are in currently, not
2761 a printed representation of its name. If you have recently switched
2762 back and forth from the @file{*scratch*} buffer, @code{other-buffer}
2763 will return that buffer.
2766 You can see this by evaluating the expression:
2773 You should see @file{#<buffer *scratch*>} appear in the echo area, or
2774 the name of whatever other buffer you switched back from most
2775 recently@footnote{Actually, by default, if the buffer from which you
2776 just switched is visible to you in another window, @code{other-buffer}
2777 will choose the most recent buffer that you cannot see; this is a
2778 subtlety that I often forget.}.
2780 @node Switching Buffers
2781 @section Switching Buffers
2782 @findex switch-to-buffer
2784 @cindex Switching to a buffer
2786 The @code{other-buffer} function actually provides a buffer when it is
2787 used as an argument to a function that requires one. We can see this
2788 by using @code{other-buffer} and @code{switch-to-buffer} to switch to a
2791 But first, a brief introduction to the @code{switch-to-buffer}
2792 function. When you switched back and forth from Info to the
2793 @file{*scratch*} buffer to evaluate @code{(buffer-name)}, you most
2794 likely typed @kbd{C-x b} and then typed @file{*scratch*}@footnote{Or
2795 rather, to save typing, you probably only typed @kbd{RET} if the
2796 default buffer was @file{*scratch*}, or if it was different, then you
2797 typed just part of the name, such as @code{*sc}, pressed your
2798 @kbd{TAB} key to cause it to expand to the full name, and then typed
2799 @kbd{RET}.} when prompted in the minibuffer for the name of
2800 the buffer to which you wanted to switch. The keystrokes, @kbd{C-x
2801 b}, cause the Lisp interpreter to evaluate the interactive function
2802 @code{switch-to-buffer}. As we said before, this is how Emacs works:
2803 different keystrokes call or run different functions. For example,
2804 @kbd{C-f} calls @code{forward-char}, @kbd{M-e} calls
2805 @code{forward-sentence}, and so on.
2807 By writing @code{switch-to-buffer} in an expression, and giving it a
2808 buffer to switch to, we can switch buffers just the way @kbd{C-x b}
2812 (switch-to-buffer (other-buffer))
2816 The symbol @code{switch-to-buffer} is the first element of the list,
2817 so the Lisp interpreter will treat it as a function and carry out the
2818 instructions that are attached to it. But before doing that, the
2819 interpreter will note that @code{other-buffer} is inside parentheses
2820 and work on that symbol first. @code{other-buffer} is the first (and
2821 in this case, the only) element of this list, so the Lisp interpreter
2822 calls or runs the function. It returns another buffer. Next, the
2823 interpreter runs @code{switch-to-buffer}, passing to it, as an
2824 argument, the other buffer, which is what Emacs will switch to. If
2825 you are reading this in Info, try this now. Evaluate the expression.
2826 (To get back, type @kbd{C-x b @key{RET}}.)@footnote{Remember, this
2827 expression will move you to your most recent other buffer that you
2828 cannot see. If you really want to go to your most recently selected
2829 buffer, even if you can still see it, you need to evaluate the
2830 following more complex expression:
2833 (switch-to-buffer (other-buffer (current-buffer) t))
2837 In this case, the first argument to @code{other-buffer} tells it which
2838 buffer to skip---the current one---and the second argument tells
2839 @code{other-buffer} it is OK to switch to a visible buffer.
2840 In regular use, @code{switch-to-buffer} takes you to an invisible
2841 window since you would most likely use @kbd{C-x o} (@code{other-window})
2842 to go to another visible buffer.}
2844 In the programming examples in later sections of this document, you will
2845 see the function @code{set-buffer} more often than
2846 @code{switch-to-buffer}. This is because of a difference between
2847 computer programs and humans: humans have eyes and expect to see the
2848 buffer on which they are working on their computer terminals. This is
2849 so obvious, it almost goes without saying. However, programs do not
2850 have eyes. When a computer program works on a buffer, that buffer does
2851 not need to be visible on the screen.
2853 @code{switch-to-buffer} is designed for humans and does two different
2854 things: it switches the buffer to which Emacs's attention is directed; and
2855 it switches the buffer displayed in the window to the new buffer.
2856 @code{set-buffer}, on the other hand, does only one thing: it switches
2857 the attention of the computer program to a different buffer. The buffer
2858 on the screen remains unchanged (of course, normally nothing happens
2859 there until the command finishes running).
2861 @cindex @samp{call} defined
2862 Also, we have just introduced another jargon term, the word @dfn{call}.
2863 When you evaluate a list in which the first symbol is a function, you
2864 are calling that function. The use of the term comes from the notion of
2865 the function as an entity that can do something for you if you `call'
2866 it---just as a plumber is an entity who can fix a leak if you call him
2869 @node Buffer Size & Locations
2870 @section Buffer Size and the Location of Point
2871 @cindex Size of buffer
2873 @cindex Point location
2874 @cindex Location of point
2876 Finally, let's look at several rather simple functions,
2877 @code{buffer-size}, @code{point}, @code{point-min}, and
2878 @code{point-max}. These give information about the size of a buffer and
2879 the location of point within it.
2881 The function @code{buffer-size} tells you the size of the current
2882 buffer; that is, the function returns a count of the number of
2883 characters in the buffer.
2890 You can evaluate this in the usual way, by positioning the
2891 cursor after the expression and typing @kbd{C-x C-e}.
2893 @cindex @samp{point} defined
2894 In Emacs, the current position of the cursor is called @dfn{point}.
2895 The expression @code{(point)} returns a number that tells you where the
2896 cursor is located as a count of the number of characters from the
2897 beginning of the buffer up to point.
2900 You can see the character count for point in this buffer by evaluating
2901 the following expression in the usual way:
2908 As I write this, the value of @code{point} is 65724. The @code{point}
2909 function is frequently used in some of the examples later in this
2913 The value of point depends, of course, on its location within the
2914 buffer. If you evaluate point in this spot, the number will be larger:
2921 For me, the value of point in this location is 66043, which means that
2922 there are 319 characters (including spaces) between the two
2923 expressions. (Doubtless, you will see different numbers, since I will
2924 have edited this since I first evaluated point.)
2926 @cindex @samp{narrowing} defined
2927 The function @code{point-min} is somewhat similar to @code{point}, but
2928 it returns the value of the minimum permissible value of point in the
2929 current buffer. This is the number 1 unless @dfn{narrowing} is in
2930 effect. (Narrowing is a mechanism whereby you can restrict yourself,
2931 or a program, to operations on just a part of a buffer.
2932 @xref{Narrowing & Widening, , Narrowing and Widening}.) Likewise, the
2933 function @code{point-max} returns the value of the maximum permissible
2934 value of point in the current buffer.
2936 @node Evaluation Exercise
2939 Find a file with which you are working and move towards its middle.
2940 Find its buffer name, file name, length, and your position in the file.
2942 @node Writing Defuns
2943 @chapter How To Write Function Definitions
2944 @cindex Definition writing
2945 @cindex Function definition writing
2946 @cindex Writing a function definition
2948 When the Lisp interpreter evaluates a list, it looks to see whether the
2949 first symbol on the list has a function definition attached to it; or,
2950 put another way, whether the symbol points to a function definition. If
2951 it does, the computer carries out the instructions in the definition. A
2952 symbol that has a function definition is called, simply, a function
2953 (although, properly speaking, the definition is the function and the
2954 symbol refers to it.)
2957 * Primitive Functions::
2958 * defun:: The @code{defun} macro.
2959 * Install:: Install a function definition.
2960 * Interactive:: Making a function interactive.
2961 * Interactive Options:: Different options for @code{interactive}.
2962 * Permanent Installation:: Installing code permanently.
2963 * let:: Creating and initializing local variables.
2965 * else:: If--then--else expressions.
2966 * Truth & Falsehood:: What Lisp considers false and true.
2967 * save-excursion:: Keeping track of point, mark, and buffer.
2973 @node Primitive Functions
2974 @unnumberedsec An Aside about Primitive Functions
2976 @cindex Primitive functions
2977 @cindex Functions, primitive
2979 @cindex C language primitives
2980 @cindex Primitives written in C
2981 All functions are defined in terms of other functions, except for a few
2982 @dfn{primitive} functions that are written in the C programming
2983 language. When you write functions' definitions, you will write them in
2984 Emacs Lisp and use other functions as your building blocks. Some of the
2985 functions you will use will themselves be written in Emacs Lisp (perhaps
2986 by you) and some will be primitives written in C@. The primitive
2987 functions are used exactly like those written in Emacs Lisp and behave
2988 like them. They are written in C so we can easily run GNU Emacs on any
2989 computer that has sufficient power and can run C.
2991 Let me re-emphasize this: when you write code in Emacs Lisp, you do not
2992 distinguish between the use of functions written in C and the use of
2993 functions written in Emacs Lisp. The difference is irrelevant. I
2994 mention the distinction only because it is interesting to know. Indeed,
2995 unless you investigate, you won't know whether an already-written
2996 function is written in Emacs Lisp or C.
2999 @section The @code{defun} Macro
3002 @cindex @samp{function definition} defined
3003 In Lisp, a symbol such as @code{mark-whole-buffer} has code attached to
3004 it that tells the computer what to do when the function is called.
3005 This code is called the @dfn{function definition} and is created by
3006 evaluating a Lisp expression that starts with the symbol @code{defun}
3007 (which is an abbreviation for @emph{define function}).
3009 In subsequent sections, we will look at function definitions from the
3010 Emacs source code, such as @code{mark-whole-buffer}. In this section,
3011 we will describe a simple function definition so you can see how it
3012 looks. This function definition uses arithmetic because it makes for a
3013 simple example. Some people dislike examples using arithmetic; however,
3014 if you are such a person, do not despair. Hardly any of the code we
3015 will study in the remainder of this introduction involves arithmetic or
3016 mathematics. The examples mostly involve text in one way or another.
3018 A function definition has up to five parts following the word
3023 The name of the symbol to which the function definition should be
3027 A list of the arguments that will be passed to the function. If no
3028 arguments will be passed to the function, this is an empty list,
3032 Documentation describing the function. (Technically optional, but
3033 strongly recommended.)
3036 Optionally, an expression to make the function interactive so you can
3037 use it by typing @kbd{M-x} and then the name of the function; or by
3038 typing an appropriate key or keychord.
3040 @cindex @samp{body} defined
3042 The code that instructs the computer what to do: the @dfn{body} of the
3043 function definition.
3046 It is helpful to think of the five parts of a function definition as
3047 being organized in a template, with slots for each part:
3051 (defun @var{function-name} (@var{arguments}@dots{})
3052 "@var{optional-documentation}@dots{}"
3053 (interactive @var{argument-passing-info}) ; @r{optional}
3058 As an example, here is the code for a function that multiplies its
3059 argument by 7. (This example is not interactive. @xref{Interactive,
3060 , Making a Function Interactive}, for that information.)
3064 (defun multiply-by-seven (number)
3065 "Multiply NUMBER by seven."
3070 This definition begins with a parenthesis and the symbol @code{defun},
3071 followed by the name of the function.
3073 @cindex @samp{argument list} defined
3074 The name of the function is followed by a list that contains the
3075 arguments that will be passed to the function. This list is called
3076 the @dfn{argument list}. In this example, the list has only one
3077 element, the symbol, @code{number}. When the function is used, the
3078 symbol will be bound to the value that is used as the argument to the
3081 Instead of choosing the word @code{number} for the name of the argument,
3082 I could have picked any other name. For example, I could have chosen
3083 the word @code{multiplicand}. I picked the word `number' because it
3084 tells what kind of value is intended for this slot; but I could just as
3085 well have chosen the word `multiplicand' to indicate the role that the
3086 value placed in this slot will play in the workings of the function. I
3087 could have called it @code{foogle}, but that would have been a bad
3088 choice because it would not tell humans what it means. The choice of
3089 name is up to the programmer and should be chosen to make the meaning of
3092 Indeed, you can choose any name you wish for a symbol in an argument
3093 list, even the name of a symbol used in some other function: the name
3094 you use in an argument list is private to that particular definition.
3095 In that definition, the name refers to a different entity than any use
3096 of the same name outside the function definition. Suppose you have a
3097 nick-name `Shorty' in your family; when your family members refer to
3098 `Shorty', they mean you. But outside your family, in a movie, for
3099 example, the name `Shorty' refers to someone else. Because a name in an
3100 argument list is private to the function definition, you can change the
3101 value of such a symbol inside the body of a function without changing
3102 its value outside the function. The effect is similar to that produced
3103 by a @code{let} expression. (@xref{let, , @code{let}}.)
3106 Note also that we discuss the word `number' in two different ways: as a
3107 symbol that appears in the code, and as the name of something that will
3108 be replaced by a something else during the evaluation of the function.
3109 In the first case, @code{number} is a symbol, not a number; it happens
3110 that within the function, it is a variable who value is the number in
3111 question, but our primary interest in it is as a symbol. On the other
3112 hand, when we are talking about the function, our interest is that we
3113 will substitute a number for the word @var{number}. To keep this
3114 distinction clear, we use different typography for the two
3115 circumstances. When we talk about this function, or about how it works,
3116 we refer to this number by writing @var{number}. In the function
3117 itself, we refer to it by writing @code{number}.
3120 The argument list is followed by the documentation string that
3121 describes the function. This is what you see when you type
3122 @w{@kbd{C-h f}} and the name of a function. Incidentally, when you
3123 write a documentation string like this, you should make the first line
3124 a complete sentence since some commands, such as @code{apropos}, print
3125 only the first line of a multi-line documentation string. Also, you
3126 should not indent the second line of a documentation string, if you
3127 have one, because that looks odd when you use @kbd{C-h f}
3128 (@code{describe-function}). The documentation string is optional, but
3129 it is so useful, it should be included in almost every function you
3132 @findex * @r{(multiplication)}
3133 The third line of the example consists of the body of the function
3134 definition. (Most functions' definitions, of course, are longer than
3135 this.) In this function, the body is the list, @code{(* 7 number)}, which
3136 says to multiply the value of @var{number} by 7. (In Emacs Lisp,
3137 @code{*} is the function for multiplication, just as @code{+} is the
3138 function for addition.)
3140 When you use the @code{multiply-by-seven} function, the argument
3141 @code{number} evaluates to the actual number you want used. Here is an
3142 example that shows how @code{multiply-by-seven} is used; but don't try
3143 to evaluate this yet!
3146 (multiply-by-seven 3)
3150 The symbol @code{number}, specified in the function definition in the
3151 next section, is given or ``bound to'' the value 3 in the actual use of
3152 the function. Note that although @code{number} was inside parentheses
3153 in the function definition, the argument passed to the
3154 @code{multiply-by-seven} function is not in parentheses. The
3155 parentheses are written in the function definition so the computer can
3156 figure out where the argument list ends and the rest of the function
3159 If you evaluate this example, you are likely to get an error message.
3160 (Go ahead, try it!) This is because we have written the function
3161 definition, but not yet told the computer about the definition---we have
3162 not yet installed (or `loaded') the function definition in Emacs.
3163 Installing a function is the process that tells the Lisp interpreter the
3164 definition of the function. Installation is described in the next
3168 @section Install a Function Definition
3169 @cindex Install a Function Definition
3170 @cindex Definition installation
3171 @cindex Function definition installation
3173 If you are reading this inside of Info in Emacs, you can try out the
3174 @code{multiply-by-seven} function by first evaluating the function
3175 definition and then evaluating @code{(multiply-by-seven 3)}. A copy of
3176 the function definition follows. Place the cursor after the last
3177 parenthesis of the function definition and type @kbd{C-x C-e}. When you
3178 do this, @code{multiply-by-seven} will appear in the echo area. (What
3179 this means is that when a function definition is evaluated, the value it
3180 returns is the name of the defined function.) At the same time, this
3181 action installs the function definition.
3185 (defun multiply-by-seven (number)
3186 "Multiply NUMBER by seven."
3192 By evaluating this @code{defun}, you have just installed
3193 @code{multiply-by-seven} in Emacs. The function is now just as much a
3194 part of Emacs as @code{forward-word} or any other editing function you
3195 use. (@code{multiply-by-seven} will stay installed until you quit
3196 Emacs. To reload code automatically whenever you start Emacs, see
3197 @ref{Permanent Installation, , Installing Code Permanently}.)
3200 * Effect of installation::
3201 * Change a defun:: How to change a function definition.
3205 @node Effect of installation
3206 @unnumberedsubsec The effect of installation
3209 You can see the effect of installing @code{multiply-by-seven} by
3210 evaluating the following sample. Place the cursor after the following
3211 expression and type @kbd{C-x C-e}. The number 21 will appear in the
3215 (multiply-by-seven 3)
3218 If you wish, you can read the documentation for the function by typing
3219 @kbd{C-h f} (@code{describe-function}) and then the name of the
3220 function, @code{multiply-by-seven}. When you do this, a
3221 @file{*Help*} window will appear on your screen that says:
3225 multiply-by-seven is a Lisp function.
3226 (multiply-by-seven NUMBER)
3228 Multiply NUMBER by seven.
3233 (To return to a single window on your screen, type @kbd{C-x 1}.)
3235 @node Change a defun
3236 @subsection Change a Function Definition
3237 @cindex Changing a function definition
3238 @cindex Function definition, how to change
3239 @cindex Definition, how to change
3241 If you want to change the code in @code{multiply-by-seven}, just rewrite
3242 it. To install the new version in place of the old one, evaluate the
3243 function definition again. This is how you modify code in Emacs. It is
3246 As an example, you can change the @code{multiply-by-seven} function to
3247 add the number to itself seven times instead of multiplying the number
3248 by seven. It produces the same answer, but by a different path. At
3249 the same time, we will add a comment to the code; a comment is text
3250 that the Lisp interpreter ignores, but that a human reader may find
3251 useful or enlightening. The comment is that this is the ``second
3256 (defun multiply-by-seven (number) ; @r{Second version.}
3257 "Multiply NUMBER by seven."
3258 (+ number number number number number number number))
3262 @cindex Comments in Lisp code
3263 The comment follows a semicolon, @samp{;}. In Lisp, everything on a
3264 line that follows a semicolon is a comment. The end of the line is the
3265 end of the comment. To stretch a comment over two or more lines, begin
3266 each line with a semicolon.
3268 @xref{Beginning init File, , Beginning a @file{.emacs}
3269 File}, and @ref{Comments, , Comments, elisp, The GNU Emacs Lisp
3270 Reference Manual}, for more about comments.
3272 You can install this version of the @code{multiply-by-seven} function by
3273 evaluating it in the same way you evaluated the first function: place
3274 the cursor after the last parenthesis and type @kbd{C-x C-e}.
3276 In summary, this is how you write code in Emacs Lisp: you write a
3277 function; install it; test it; and then make fixes or enhancements and
3281 @section Make a Function Interactive
3282 @cindex Interactive functions
3285 You make a function interactive by placing a list that begins with
3286 the special form @code{interactive} immediately after the
3287 documentation. A user can invoke an interactive function by typing
3288 @kbd{M-x} and then the name of the function; or by typing the keys to
3289 which it is bound, for example, by typing @kbd{C-n} for
3290 @code{next-line} or @kbd{C-x h} for @code{mark-whole-buffer}.
3292 Interestingly, when you call an interactive function interactively,
3293 the value returned is not automatically displayed in the echo area.
3294 This is because you often call an interactive function for its side
3295 effects, such as moving forward by a word or line, and not for the
3296 value returned. If the returned value were displayed in the echo area
3297 each time you typed a key, it would be very distracting.
3300 * Interactive multiply-by-seven:: An overview.
3301 * multiply-by-seven in detail:: The interactive version.
3305 @node Interactive multiply-by-seven
3306 @unnumberedsubsec An Interactive @code{multiply-by-seven}, An Overview
3309 Both the use of the special form @code{interactive} and one way to
3310 display a value in the echo area can be illustrated by creating an
3311 interactive version of @code{multiply-by-seven}.
3318 (defun multiply-by-seven (number) ; @r{Interactive version.}
3319 "Multiply NUMBER by seven."
3321 (message "The result is %d" (* 7 number)))
3326 You can install this code by placing your cursor after it and typing
3327 @kbd{C-x C-e}. The name of the function will appear in your echo area.
3328 Then, you can use this code by typing @kbd{C-u} and a number and then
3329 typing @kbd{M-x multiply-by-seven} and pressing @key{RET}. The phrase
3330 @samp{The result is @dots{}} followed by the product will appear in the
3333 Speaking more generally, you invoke a function like this in either of two
3338 By typing a prefix argument that contains the number to be passed, and
3339 then typing @kbd{M-x} and the name of the function, as with
3340 @kbd{C-u 3 M-x forward-sentence}; or,
3343 By typing whatever key or keychord the function is bound to, as with
3348 Both the examples just mentioned work identically to move point forward
3349 three sentences. (Since @code{multiply-by-seven} is not bound to a key,
3350 it could not be used as an example of key binding.)
3352 (@xref{Keybindings, , Some Keybindings}, to learn how to bind a command
3355 A prefix argument is passed to an interactive function by typing the
3356 @key{META} key followed by a number, for example, @kbd{M-3 M-e}, or by
3357 typing @kbd{C-u} and then a number, for example, @kbd{C-u 3 M-e} (if you
3358 type @kbd{C-u} without a number, it defaults to 4).
3360 @node multiply-by-seven in detail
3361 @subsection An Interactive @code{multiply-by-seven}
3363 Let's look at the use of the special form @code{interactive} and then at
3364 the function @code{message} in the interactive version of
3365 @code{multiply-by-seven}. You will recall that the function definition
3370 (defun multiply-by-seven (number) ; @r{Interactive version.}
3371 "Multiply NUMBER by seven."
3373 (message "The result is %d" (* 7 number)))
3377 In this function, the expression, @code{(interactive "p")}, is a list of
3378 two elements. The @code{"p"} tells Emacs to pass the prefix argument to
3379 the function and use its value for the argument of the function.
3382 The argument will be a number. This means that the symbol
3383 @code{number} will be bound to a number in the line:
3386 (message "The result is %d" (* 7 number))
3391 For example, if your prefix argument is 5, the Lisp interpreter will
3392 evaluate the line as if it were:
3395 (message "The result is %d" (* 7 5))
3399 (If you are reading this in GNU Emacs, you can evaluate this expression
3400 yourself.) First, the interpreter will evaluate the inner list, which
3401 is @code{(* 7 5)}. This returns a value of 35. Next, it
3402 will evaluate the outer list, passing the values of the second and
3403 subsequent elements of the list to the function @code{message}.
3405 As we have seen, @code{message} is an Emacs Lisp function especially
3406 designed for sending a one line message to a user. (@xref{message, ,
3407 The @code{message} function}.) In summary, the @code{message}
3408 function prints its first argument in the echo area as is, except for
3409 occurrences of @samp{%d} or @samp{%s} (and various other %-sequences
3410 which we have not mentioned). When it sees a control sequence, the
3411 function looks to the second or subsequent arguments and prints the
3412 value of the argument in the location in the string where the control
3413 sequence is located.
3415 In the interactive @code{multiply-by-seven} function, the control string
3416 is @samp{%d}, which requires a number, and the value returned by
3417 evaluating @code{(* 7 5)} is the number 35. Consequently, the number 35
3418 is printed in place of the @samp{%d} and the message is @samp{The result
3421 (Note that when you call the function @code{multiply-by-seven}, the
3422 message is printed without quotes, but when you call @code{message}, the
3423 text is printed in double quotes. This is because the value returned by
3424 @code{message} is what appears in the echo area when you evaluate an
3425 expression whose first element is @code{message}; but when embedded in a
3426 function, @code{message} prints the text as a side effect without
3429 @node Interactive Options
3430 @section Different Options for @code{interactive}
3431 @cindex Options for @code{interactive}
3432 @cindex Interactive options
3434 In the example, @code{multiply-by-seven} used @code{"p"} as the
3435 argument to @code{interactive}. This argument told Emacs to interpret
3436 your typing either @kbd{C-u} followed by a number or @key{META}
3437 followed by a number as a command to pass that number to the function
3438 as its argument. Emacs has more than twenty characters predefined for
3439 use with @code{interactive}. In almost every case, one of these
3440 options will enable you to pass the right information interactively to
3441 a function. (@xref{Interactive Codes, , Code Characters for
3442 @code{interactive}, elisp, The GNU Emacs Lisp Reference Manual}.)
3445 Consider the function @code{zap-to-char}. Its interactive expression
3449 (interactive "p\ncZap to char: ")
3452 The first part of the argument to @code{interactive} is @samp{p}, with
3453 which you are already familiar. This argument tells Emacs to
3454 interpret a `prefix', as a number to be passed to the function. You
3455 can specify a prefix either by typing @kbd{C-u} followed by a number
3456 or by typing @key{META} followed by a number. The prefix is the
3457 number of specified characters. Thus, if your prefix is three and the
3458 specified character is @samp{x}, then you will delete all the text up
3459 to and including the third next @samp{x}. If you do not set a prefix,
3460 then you delete all the text up to and including the specified
3461 character, but no more.
3463 The @samp{c} tells the function the name of the character to which to delete.
3465 More formally, a function with two or more arguments can have
3466 information passed to each argument by adding parts to the string that
3467 follows @code{interactive}. When you do this, the information is
3468 passed to each argument in the same order it is specified in the
3469 @code{interactive} list. In the string, each part is separated from
3470 the next part by a @samp{\n}, which is a newline. For example, you
3471 can follow @samp{p} with a @samp{\n} and an @samp{cZap to char:@: }.
3472 This causes Emacs to pass the value of the prefix argument (if there
3473 is one) and the character.
3475 In this case, the function definition looks like the following, where
3476 @code{arg} and @code{char} are the symbols to which @code{interactive}
3477 binds the prefix argument and the specified character:
3481 (defun @var{name-of-function} (arg char)
3482 "@var{documentation}@dots{}"
3483 (interactive "p\ncZap to char: ")
3484 @var{body-of-function}@dots{})
3489 (The space after the colon in the prompt makes it look better when you
3490 are prompted. @xref{copy-to-buffer, , The Definition of
3491 @code{copy-to-buffer}}, for an example.)
3493 When a function does not take arguments, @code{interactive} does not
3494 require any. Such a function contains the simple expression
3495 @code{(interactive)}. The @code{mark-whole-buffer} function is like
3498 Alternatively, if the special letter-codes are not right for your
3499 application, you can pass your own arguments to @code{interactive} as
3502 @xref{append-to-buffer, , The Definition of @code{append-to-buffer}},
3503 for an example. @xref{Using Interactive, , Using @code{Interactive},
3504 elisp, The GNU Emacs Lisp Reference Manual}, for a more complete
3505 explanation about this technique.
3507 @node Permanent Installation
3508 @section Install Code Permanently
3509 @cindex Install code permanently
3510 @cindex Permanent code installation
3511 @cindex Code installation
3513 When you install a function definition by evaluating it, it will stay
3514 installed until you quit Emacs. The next time you start a new session
3515 of Emacs, the function will not be installed unless you evaluate the
3516 function definition again.
3518 At some point, you may want to have code installed automatically
3519 whenever you start a new session of Emacs. There are several ways of
3524 If you have code that is just for yourself, you can put the code for the
3525 function definition in your @file{.emacs} initialization file. When you
3526 start Emacs, your @file{.emacs} file is automatically evaluated and all
3527 the function definitions within it are installed.
3528 @xref{Emacs Initialization, , Your @file{.emacs} File}.
3531 Alternatively, you can put the function definitions that you want
3532 installed in one or more files of their own and use the @code{load}
3533 function to cause Emacs to evaluate and thereby install each of the
3534 functions in the files.
3535 @xref{Loading Files, , Loading Files}.
3538 Thirdly, if you have code that your whole site will use, it is usual
3539 to put it in a file called @file{site-init.el} that is loaded when
3540 Emacs is built. This makes the code available to everyone who uses
3541 your machine. (See the @file{INSTALL} file that is part of the Emacs
3545 Finally, if you have code that everyone who uses Emacs may want, you
3546 can post it on a computer network or send a copy to the Free Software
3547 Foundation. (When you do this, please license the code and its
3548 documentation under a license that permits other people to run, copy,
3549 study, modify, and redistribute the code and which protects you from
3550 having your work taken from you.) If you send a copy of your code to
3551 the Free Software Foundation, and properly protect yourself and
3552 others, it may be included in the next release of Emacs. In large
3553 part, this is how Emacs has grown over the past years, by donations.
3559 The @code{let} expression is a special form in Lisp that you will need
3560 to use in most function definitions.
3562 @code{let} is used to attach or bind a symbol to a value in such a way
3563 that the Lisp interpreter will not confuse the variable with a
3564 variable of the same name that is not part of the function.
3566 To understand why the @code{let} special form is necessary, consider
3567 the situation in which you own a home that you generally refer to as
3568 `the house', as in the sentence, ``The house needs painting.'' If you
3569 are visiting a friend and your host refers to `the house', he is
3570 likely to be referring to @emph{his} house, not yours, that is, to a
3573 If your friend is referring to his house and you think he is referring
3574 to your house, you may be in for some confusion. The same thing could
3575 happen in Lisp if a variable that is used inside of one function has
3576 the same name as a variable that is used inside of another function,
3577 and the two are not intended to refer to the same value. The
3578 @code{let} special form prevents this kind of confusion.
3581 * Prevent confusion::
3582 * Parts of let Expression::
3583 * Sample let Expression::
3584 * Uninitialized let Variables::
3588 @node Prevent confusion
3589 @unnumberedsubsec @code{let} Prevents Confusion
3592 @cindex @samp{local variable} defined
3593 @cindex @samp{variable, local}, defined
3594 The @code{let} special form prevents confusion. @code{let} creates a
3595 name for a @dfn{local variable} that overshadows any use of the same
3596 name outside the @code{let} expression. This is like understanding
3597 that whenever your host refers to `the house', he means his house, not
3598 yours. (Symbols used in argument lists work the same way.
3599 @xref{defun, , The @code{defun} Macro}.)
3601 Local variables created by a @code{let} expression retain their value
3602 @emph{only} within the @code{let} expression itself (and within
3603 expressions called within the @code{let} expression); the local
3604 variables have no effect outside the @code{let} expression.
3606 Another way to think about @code{let} is that it is like a @code{setq}
3607 that is temporary and local. The values set by @code{let} are
3608 automatically undone when the @code{let} is finished. The setting
3609 only affects expressions that are inside the bounds of the @code{let}
3610 expression. In computer science jargon, we would say ``the binding of
3611 a symbol is visible only in functions called in the @code{let} form;
3612 in Emacs Lisp, scoping is dynamic, not lexical.''
3614 @code{let} can create more than one variable at once. Also,
3615 @code{let} gives each variable it creates an initial value, either a
3616 value specified by you, or @code{nil}. (In the jargon, this is called
3617 `binding the variable to the value'.) After @code{let} has created
3618 and bound the variables, it executes the code in the body of the
3619 @code{let}, and returns the value of the last expression in the body,
3620 as the value of the whole @code{let} expression. (`Execute' is a jargon
3621 term that means to evaluate a list; it comes from the use of the word
3622 meaning `to give practical effect to' (@cite{Oxford English
3623 Dictionary}). Since you evaluate an expression to perform an action,
3624 `execute' has evolved as a synonym to `evaluate'.)
3626 @node Parts of let Expression
3627 @subsection The Parts of a @code{let} Expression
3628 @cindex @code{let} expression, parts of
3629 @cindex Parts of @code{let} expression
3631 @cindex @samp{varlist} defined
3632 A @code{let} expression is a list of three parts. The first part is
3633 the symbol @code{let}. The second part is a list, called a
3634 @dfn{varlist}, each element of which is either a symbol by itself or a
3635 two-element list, the first element of which is a symbol. The third
3636 part of the @code{let} expression is the body of the @code{let}. The
3637 body usually consists of one or more lists.
3640 A template for a @code{let} expression looks like this:
3643 (let @var{varlist} @var{body}@dots{})
3647 The symbols in the varlist are the variables that are given initial
3648 values by the @code{let} special form. Symbols by themselves are given
3649 the initial value of @code{nil}; and each symbol that is the first
3650 element of a two-element list is bound to the value that is returned
3651 when the Lisp interpreter evaluates the second element.
3653 Thus, a varlist might look like this: @code{(thread (needles 3))}. In
3654 this case, in a @code{let} expression, Emacs binds the symbol
3655 @code{thread} to an initial value of @code{nil}, and binds the symbol
3656 @code{needles} to an initial value of 3.
3658 When you write a @code{let} expression, what you do is put the
3659 appropriate expressions in the slots of the @code{let} expression
3662 If the varlist is composed of two-element lists, as is often the case,
3663 the template for the @code{let} expression looks like this:
3667 (let ((@var{variable} @var{value})
3668 (@var{variable} @var{value})
3674 @node Sample let Expression
3675 @subsection Sample @code{let} Expression
3676 @cindex Sample @code{let} expression
3677 @cindex @code{let} expression sample
3679 The following expression creates and gives initial values
3680 to the two variables @code{zebra} and @code{tiger}. The body of the
3681 @code{let} expression is a list which calls the @code{message} function.
3685 (let ((zebra 'stripes)
3687 (message "One kind of animal has %s and another is %s."
3692 Here, the varlist is @code{((zebra 'stripes) (tiger 'fierce))}.
3694 The two variables are @code{zebra} and @code{tiger}. Each variable is
3695 the first element of a two-element list and each value is the second
3696 element of its two-element list. In the varlist, Emacs binds the
3697 variable @code{zebra} to the value @code{stripes}@footnote{According
3698 to Jared Diamond in @cite{Guns, Germs, and Steel}, ``@dots{} zebras
3699 become impossibly dangerous as they grow older'' but the claim here is
3700 that they do not become fierce like a tiger. (1997, W. W. Norton and
3701 Co., ISBN 0-393-03894-2, page 171)}, and binds the
3702 variable @code{tiger} to the value @code{fierce}. In this example,
3703 both values are symbols preceded by a quote. The values could just as
3704 well have been another list or a string. The body of the @code{let}
3705 follows after the list holding the variables. In this example, the
3706 body is a list that uses the @code{message} function to print a string
3710 You may evaluate the example in the usual fashion, by placing the
3711 cursor after the last parenthesis and typing @kbd{C-x C-e}. When you do
3712 this, the following will appear in the echo area:
3715 "One kind of animal has stripes and another is fierce."
3718 As we have seen before, the @code{message} function prints its first
3719 argument, except for @samp{%s}. In this example, the value of the variable
3720 @code{zebra} is printed at the location of the first @samp{%s} and the
3721 value of the variable @code{tiger} is printed at the location of the
3724 @node Uninitialized let Variables
3725 @subsection Uninitialized Variables in a @code{let} Statement
3726 @cindex Uninitialized @code{let} variables
3727 @cindex @code{let} variables uninitialized
3729 If you do not bind the variables in a @code{let} statement to specific
3730 initial values, they will automatically be bound to an initial value of
3731 @code{nil}, as in the following expression:
3740 "Here are %d variables with %s, %s, and %s value."
3741 birch pine fir oak))
3746 Here, the varlist is @code{((birch 3) pine fir (oak 'some))}.
3749 If you evaluate this expression in the usual way, the following will
3750 appear in your echo area:
3753 "Here are 3 variables with nil, nil, and some value."
3757 In this example, Emacs binds the symbol @code{birch} to the number 3,
3758 binds the symbols @code{pine} and @code{fir} to @code{nil}, and binds
3759 the symbol @code{oak} to the value @code{some}.
3761 Note that in the first part of the @code{let}, the variables @code{pine}
3762 and @code{fir} stand alone as atoms that are not surrounded by
3763 parentheses; this is because they are being bound to @code{nil}, the
3764 empty list. But @code{oak} is bound to @code{some} and so is a part of
3765 the list @code{(oak 'some)}. Similarly, @code{birch} is bound to the
3766 number 3 and so is in a list with that number. (Since a number
3767 evaluates to itself, the number does not need to be quoted. Also, the
3768 number is printed in the message using a @samp{%d} rather than a
3769 @samp{%s}.) The four variables as a group are put into a list to
3770 delimit them from the body of the @code{let}.
3773 @section The @code{if} Special Form
3775 @cindex Conditional with @code{if}
3777 A third special form, in addition to @code{defun} and @code{let}, is the
3778 conditional @code{if}. This form is used to instruct the computer to
3779 make decisions. You can write function definitions without using
3780 @code{if}, but it is used often enough, and is important enough, to be
3781 included here. It is used, for example, in the code for the
3782 function @code{beginning-of-buffer}.
3784 The basic idea behind an @code{if}, is that ``@emph{if} a test is true,
3785 @emph{then} an expression is evaluated.'' If the test is not true, the
3786 expression is not evaluated. For example, you might make a decision
3787 such as, ``if it is warm and sunny, then go to the beach!''
3790 * if in more detail::
3791 * type-of-animal in detail:: An example of an @code{if} expression.
3795 @node if in more detail
3796 @unnumberedsubsec @code{if} in more detail
3799 @cindex @samp{if-part} defined
3800 @cindex @samp{then-part} defined
3801 An @code{if} expression written in Lisp does not use the word `then';
3802 the test and the action are the second and third elements of the list
3803 whose first element is @code{if}. Nonetheless, the test part of an
3804 @code{if} expression is often called the @dfn{if-part} and the second
3805 argument is often called the @dfn{then-part}.
3807 Also, when an @code{if} expression is written, the true-or-false-test
3808 is usually written on the same line as the symbol @code{if}, but the
3809 action to carry out if the test is true, the ``then-part'', is written
3810 on the second and subsequent lines. This makes the @code{if}
3811 expression easier to read.
3815 (if @var{true-or-false-test}
3816 @var{action-to-carry-out-if-test-is-true})
3821 The true-or-false-test will be an expression that
3822 is evaluated by the Lisp interpreter.
3824 Here is an example that you can evaluate in the usual manner. The test
3825 is whether the number 5 is greater than the number 4. Since it is, the
3826 message @samp{5 is greater than 4!} will be printed.
3830 (if (> 5 4) ; @r{if-part}
3831 (message "5 is greater than 4!")) ; @r{then-part}
3836 (The function @code{>} tests whether its first argument is greater than
3837 its second argument and returns true if it is.)
3838 @findex > (greater than)
3840 Of course, in actual use, the test in an @code{if} expression will not
3841 be fixed for all time as it is by the expression @code{(> 5 4)}.
3842 Instead, at least one of the variables used in the test will be bound to
3843 a value that is not known ahead of time. (If the value were known ahead
3844 of time, we would not need to run the test!)
3846 For example, the value may be bound to an argument of a function
3847 definition. In the following function definition, the character of the
3848 animal is a value that is passed to the function. If the value bound to
3849 @code{characteristic} is @code{fierce}, then the message, @samp{It's a
3850 tiger!} will be printed; otherwise, @code{nil} will be returned.
3854 (defun type-of-animal (characteristic)
3855 "Print message in echo area depending on CHARACTERISTIC.
3856 If the CHARACTERISTIC is the symbol `fierce',
3857 then warn of a tiger."
3858 (if (equal characteristic 'fierce)
3859 (message "It's a tiger!")))
3865 If you are reading this inside of GNU Emacs, you can evaluate the
3866 function definition in the usual way to install it in Emacs, and then you
3867 can evaluate the following two expressions to see the results:
3871 (type-of-animal 'fierce)
3873 (type-of-animal 'zebra)
3878 @c Following sentences rewritten to prevent overfull hbox.
3880 When you evaluate @code{(type-of-animal 'fierce)}, you will see the
3881 following message printed in the echo area: @code{"It's a tiger!"}; and
3882 when you evaluate @code{(type-of-animal 'zebra)} you will see @code{nil}
3883 printed in the echo area.
3885 @node type-of-animal in detail
3886 @subsection The @code{type-of-animal} Function in Detail
3888 Let's look at the @code{type-of-animal} function in detail.
3890 The function definition for @code{type-of-animal} was written by filling
3891 the slots of two templates, one for a function definition as a whole, and
3892 a second for an @code{if} expression.
3895 The template for every function that is not interactive is:
3899 (defun @var{name-of-function} (@var{argument-list})
3900 "@var{documentation}@dots{}"
3906 The parts of the function that match this template look like this:
3910 (defun type-of-animal (characteristic)
3911 "Print message in echo area depending on CHARACTERISTIC.
3912 If the CHARACTERISTIC is the symbol `fierce',
3913 then warn of a tiger."
3914 @var{body: the} @code{if} @var{expression})
3918 The name of function is @code{type-of-animal}; it is passed the value
3919 of one argument. The argument list is followed by a multi-line
3920 documentation string. The documentation string is included in the
3921 example because it is a good habit to write documentation string for
3922 every function definition. The body of the function definition
3923 consists of the @code{if} expression.
3926 The template for an @code{if} expression looks like this:
3930 (if @var{true-or-false-test}
3931 @var{action-to-carry-out-if-the-test-returns-true})
3936 In the @code{type-of-animal} function, the code for the @code{if}
3941 (if (equal characteristic 'fierce)
3942 (message "It's a tiger!")))
3947 Here, the true-or-false-test is the expression:
3950 (equal characteristic 'fierce)
3954 In Lisp, @code{equal} is a function that determines whether its first
3955 argument is equal to its second argument. The second argument is the
3956 quoted symbol @code{'fierce} and the first argument is the value of the
3957 symbol @code{characteristic}---in other words, the argument passed to
3960 In the first exercise of @code{type-of-animal}, the argument
3961 @code{fierce} is passed to @code{type-of-animal}. Since @code{fierce}
3962 is equal to @code{fierce}, the expression, @code{(equal characteristic
3963 'fierce)}, returns a value of true. When this happens, the @code{if}
3964 evaluates the second argument or then-part of the @code{if}:
3965 @code{(message "It's tiger!")}.
3967 On the other hand, in the second exercise of @code{type-of-animal}, the
3968 argument @code{zebra} is passed to @code{type-of-animal}. @code{zebra}
3969 is not equal to @code{fierce}, so the then-part is not evaluated and
3970 @code{nil} is returned by the @code{if} expression.
3973 @section If--then--else Expressions
3976 An @code{if} expression may have an optional third argument, called
3977 the @dfn{else-part}, for the case when the true-or-false-test returns
3978 false. When this happens, the second argument or then-part of the
3979 overall @code{if} expression is @emph{not} evaluated, but the third or
3980 else-part @emph{is} evaluated. You might think of this as the cloudy
3981 day alternative for the decision ``if it is warm and sunny, then go to
3982 the beach, else read a book!''.
3984 The word ``else'' is not written in the Lisp code; the else-part of an
3985 @code{if} expression comes after the then-part. In the written Lisp, the
3986 else-part is usually written to start on a line of its own and is
3987 indented less than the then-part:
3991 (if @var{true-or-false-test}
3992 @var{action-to-carry-out-if-the-test-returns-true}
3993 @var{action-to-carry-out-if-the-test-returns-false})
3997 For example, the following @code{if} expression prints the message @samp{4
3998 is not greater than 5!} when you evaluate it in the usual way:
4002 (if (> 4 5) ; @r{if-part}
4003 (message "4 falsely greater than 5!") ; @r{then-part}
4004 (message "4 is not greater than 5!")) ; @r{else-part}
4009 Note that the different levels of indentation make it easy to
4010 distinguish the then-part from the else-part. (GNU Emacs has several
4011 commands that automatically indent @code{if} expressions correctly.
4012 @xref{Typing Lists, , GNU Emacs Helps You Type Lists}.)
4014 We can extend the @code{type-of-animal} function to include an
4015 else-part by simply incorporating an additional part to the @code{if}
4019 You can see the consequences of doing this if you evaluate the following
4020 version of the @code{type-of-animal} function definition to install it
4021 and then evaluate the two subsequent expressions to pass different
4022 arguments to the function.
4026 (defun type-of-animal (characteristic) ; @r{Second version.}
4027 "Print message in echo area depending on CHARACTERISTIC.
4028 If the CHARACTERISTIC is the symbol `fierce',
4029 then warn of a tiger;
4030 else say it's not fierce."
4031 (if (equal characteristic 'fierce)
4032 (message "It's a tiger!")
4033 (message "It's not fierce!")))
4040 (type-of-animal 'fierce)
4042 (type-of-animal 'zebra)
4047 @c Following sentence rewritten to prevent overfull hbox.
4049 When you evaluate @code{(type-of-animal 'fierce)}, you will see the
4050 following message printed in the echo area: @code{"It's a tiger!"}; but
4051 when you evaluate @code{(type-of-animal 'zebra)}, you will see
4052 @code{"It's not fierce!"}.
4054 (Of course, if the @var{characteristic} were @code{ferocious}, the
4055 message @code{"It's not fierce!"} would be printed; and it would be
4056 misleading! When you write code, you need to take into account the
4057 possibility that some such argument will be tested by the @code{if}
4058 and write your program accordingly.)
4060 @node Truth & Falsehood
4061 @section Truth and Falsehood in Emacs Lisp
4062 @cindex Truth and falsehood in Emacs Lisp
4063 @cindex Falsehood and truth in Emacs Lisp
4066 There is an important aspect to the truth test in an @code{if}
4067 expression. So far, we have spoken of `true' and `false' as values of
4068 predicates as if they were new kinds of Emacs Lisp objects. In fact,
4069 `false' is just our old friend @code{nil}. Anything else---anything
4072 The expression that tests for truth is interpreted as @dfn{true}
4073 if the result of evaluating it is a value that is not @code{nil}. In
4074 other words, the result of the test is considered true if the value
4075 returned is a number such as 47, a string such as @code{"hello"}, or a
4076 symbol (other than @code{nil}) such as @code{flowers}, or a list (so
4077 long as it is not empty), or even a buffer!
4080 * nil explained:: @code{nil} has two meanings.
4085 @unnumberedsubsec An explanation of @code{nil}
4088 Before illustrating a test for truth, we need an explanation of @code{nil}.
4090 In Emacs Lisp, the symbol @code{nil} has two meanings. First, it means the
4091 empty list. Second, it means false and is the value returned when a
4092 true-or-false-test tests false. @code{nil} can be written as an empty
4093 list, @code{()}, or as @code{nil}. As far as the Lisp interpreter is
4094 concerned, @code{()} and @code{nil} are the same. Humans, however, tend
4095 to use @code{nil} for false and @code{()} for the empty list.
4097 In Emacs Lisp, any value that is not @code{nil}---is not the empty
4098 list---is considered true. This means that if an evaluation returns
4099 something that is not an empty list, an @code{if} expression will test
4100 true. For example, if a number is put in the slot for the test, it
4101 will be evaluated and will return itself, since that is what numbers
4102 do when evaluated. In this conditional, the @code{if} expression will
4103 test true. The expression tests false only when @code{nil}, an empty
4104 list, is returned by evaluating the expression.
4106 You can see this by evaluating the two expressions in the following examples.
4108 In the first example, the number 4 is evaluated as the test in the
4109 @code{if} expression and returns itself; consequently, the then-part
4110 of the expression is evaluated and returned: @samp{true} appears in
4111 the echo area. In the second example, the @code{nil} indicates false;
4112 consequently, the else-part of the expression is evaluated and
4113 returned: @samp{false} appears in the echo area.
4130 Incidentally, if some other useful value is not available for a test that
4131 returns true, then the Lisp interpreter will return the symbol @code{t}
4132 for true. For example, the expression @code{(> 5 4)} returns @code{t}
4133 when evaluated, as you can see by evaluating it in the usual way:
4141 On the other hand, this function returns @code{nil} if the test is false.
4147 @node save-excursion
4148 @section @code{save-excursion}
4149 @findex save-excursion
4150 @cindex Region, what it is
4151 @cindex Preserving point, mark, and buffer
4152 @cindex Point, mark, buffer preservation
4156 The @code{save-excursion} function is the third and final special form
4157 that we will discuss in this chapter.
4159 In Emacs Lisp programs used for editing, the @code{save-excursion}
4160 function is very common. It saves the location of point and mark,
4161 executes the body of the function, and then restores point and mark to
4162 their previous positions if their locations were changed. Its primary
4163 purpose is to keep the user from being surprised and disturbed by
4164 unexpected movement of point or mark.
4167 * Point and mark:: A review of various locations.
4168 * Template for save-excursion::
4172 @node Point and mark
4173 @unnumberedsubsec Point and Mark
4176 Before discussing @code{save-excursion}, however, it may be useful
4177 first to review what point and mark are in GNU Emacs. @dfn{Point} is
4178 the current location of the cursor. Wherever the cursor
4179 is, that is point. More precisely, on terminals where the cursor
4180 appears to be on top of a character, point is immediately before the
4181 character. In Emacs Lisp, point is an integer. The first character in
4182 a buffer is number one, the second is number two, and so on. The
4183 function @code{point} returns the current position of the cursor as a
4184 number. Each buffer has its own value for point.
4186 The @dfn{mark} is another position in the buffer; its value can be set
4187 with a command such as @kbd{C-@key{SPC}} (@code{set-mark-command}). If
4188 a mark has been set, you can use the command @kbd{C-x C-x}
4189 (@code{exchange-point-and-mark}) to cause the cursor to jump to the mark
4190 and set the mark to be the previous position of point. In addition, if
4191 you set another mark, the position of the previous mark is saved in the
4192 mark ring. Many mark positions can be saved this way. You can jump the
4193 cursor to a saved mark by typing @kbd{C-u C-@key{SPC}} one or more
4196 The part of the buffer between point and mark is called @dfn{the
4197 region}. Numerous commands work on the region, including
4198 @code{center-region}, @code{count-lines-region}, @code{kill-region}, and
4199 @code{print-region}.
4201 The @code{save-excursion} special form saves the locations of point and
4202 mark and restores those positions after the code within the body of the
4203 special form is evaluated by the Lisp interpreter. Thus, if point were
4204 in the beginning of a piece of text and some code moved point to the end
4205 of the buffer, the @code{save-excursion} would put point back to where
4206 it was before, after the expressions in the body of the function were
4209 In Emacs, a function frequently moves point as part of its internal
4210 workings even though a user would not expect this. For example,
4211 @code{count-lines-region} moves point. To prevent the user from being
4212 bothered by jumps that are both unexpected and (from the user's point of
4213 view) unnecessary, @code{save-excursion} is often used to keep point and
4214 mark in the location expected by the user. The use of
4215 @code{save-excursion} is good housekeeping.
4217 To make sure the house stays clean, @code{save-excursion} restores the
4218 values of point and mark even if something goes wrong in the code inside
4219 of it (or, to be more precise and to use the proper jargon, ``in case of
4220 abnormal exit''). This feature is very helpful.
4222 In addition to recording the values of point and mark,
4223 @code{save-excursion} keeps track of the current buffer, and restores
4224 it, too. This means you can write code that will change the buffer and
4225 have @code{save-excursion} switch you back to the original buffer.
4226 This is how @code{save-excursion} is used in @code{append-to-buffer}.
4227 (@xref{append-to-buffer, , The Definition of @code{append-to-buffer}}.)
4229 @node Template for save-excursion
4230 @subsection Template for a @code{save-excursion} Expression
4233 The template for code using @code{save-excursion} is simple:
4243 The body of the function is one or more expressions that will be
4244 evaluated in sequence by the Lisp interpreter. If there is more than
4245 one expression in the body, the value of the last one will be returned
4246 as the value of the @code{save-excursion} function. The other
4247 expressions in the body are evaluated only for their side effects; and
4248 @code{save-excursion} itself is used only for its side effect (which
4249 is restoring the positions of point and mark).
4252 In more detail, the template for a @code{save-excursion} expression
4258 @var{first-expression-in-body}
4259 @var{second-expression-in-body}
4260 @var{third-expression-in-body}
4262 @var{last-expression-in-body})
4267 An expression, of course, may be a symbol on its own or a list.
4269 In Emacs Lisp code, a @code{save-excursion} expression often occurs
4270 within the body of a @code{let} expression. It looks like this:
4283 In the last few chapters we have introduced a macro and a fair number
4284 of functions and special forms. Here they are described in brief,
4285 along with a few similar functions that have not been mentioned yet.
4288 @item eval-last-sexp
4289 Evaluate the last symbolic expression before the current location of
4290 point. The value is printed in the echo area unless the function is
4291 invoked with an argument; in that case, the output is printed in the
4292 current buffer. This command is normally bound to @kbd{C-x C-e}.
4295 Define function. This macro has up to five parts: the name, a
4296 template for the arguments that will be passed to the function,
4297 documentation, an optional interactive declaration, and the body of
4301 For example, in an early version of Emacs, the function definition was
4302 as follows. (It is slightly more complex now that it seeks the first
4303 non-whitespace character rather than the first visible character.)
4307 (defun back-to-indentation ()
4308 "Move point to first visible character on line."
4310 (beginning-of-line 1)
4311 (skip-chars-forward " \t"))
4318 (defun backward-to-indentation (&optional arg)
4319 "Move backward ARG lines and position at first nonblank character."
4321 (forward-line (- (or arg 1)))
4322 (skip-chars-forward " \t"))
4324 (defun back-to-indentation ()
4325 "Move point to the first non-whitespace character on this line."
4327 (beginning-of-line 1)
4328 (skip-syntax-forward " " (line-end-position))
4329 ;; Move back over chars that have whitespace syntax but have the p flag.
4330 (backward-prefix-chars))
4334 Declare to the interpreter that the function can be used
4335 interactively. This special form may be followed by a string with one
4336 or more parts that pass the information to the arguments of the
4337 function, in sequence. These parts may also tell the interpreter to
4338 prompt for information. Parts of the string are separated by
4339 newlines, @samp{\n}.
4342 Common code characters are:
4346 The name of an existing buffer.
4349 The name of an existing file.
4352 The numeric prefix argument. (Note that this `p' is lower case.)
4355 Point and the mark, as two numeric arguments, smallest first. This
4356 is the only code letter that specifies two successive arguments
4360 @xref{Interactive Codes, , Code Characters for @samp{interactive},
4361 elisp, The GNU Emacs Lisp Reference Manual}, for a complete list of
4365 Declare that a list of variables is for use within the body of the
4366 @code{let} and give them an initial value, either @code{nil} or a
4367 specified value; then evaluate the rest of the expressions in the body
4368 of the @code{let} and return the value of the last one. Inside the
4369 body of the @code{let}, the Lisp interpreter does not see the values of
4370 the variables of the same names that are bound outside of the
4378 (let ((foo (buffer-name))
4379 (bar (buffer-size)))
4381 "This buffer is %s and has %d characters."
4386 @item save-excursion
4387 Record the values of point and mark and the current buffer before
4388 evaluating the body of this special form. Restore the values of point
4389 and mark and buffer afterward.
4396 (message "We are %d characters into this buffer."
4399 (goto-char (point-min)) (point))))
4404 Evaluate the first argument to the function; if it is true, evaluate
4405 the second argument; else evaluate the third argument, if there is one.
4407 The @code{if} special form is called a @dfn{conditional}. There are
4408 other conditionals in Emacs Lisp, but @code{if} is perhaps the most
4416 (if (= 22 emacs-major-version)
4417 (message "This is version 22 Emacs")
4418 (message "This is not version 22 Emacs"))
4427 The @code{<} function tests whether its first argument is smaller than
4428 its second argument. A corresponding function, @code{>}, tests whether
4429 the first argument is greater than the second. Likewise, @code{<=}
4430 tests whether the first argument is less than or equal to the second and
4431 @code{>=} tests whether the first argument is greater than or equal to
4432 the second. In all cases, both arguments must be numbers or markers
4433 (markers indicate positions in buffers).
4437 The @code{=} function tests whether two arguments, both numbers or
4443 Test whether two objects are the same. @code{equal} uses one meaning
4444 of the word `same' and @code{eq} uses another: @code{equal} returns
4445 true if the two objects have a similar structure and contents, such as
4446 two copies of the same book. On the other hand, @code{eq}, returns
4447 true if both arguments are actually the same object.
4456 The @code{string-lessp} function tests whether its first argument is
4457 smaller than the second argument. A shorter, alternative name for the
4458 same function (a @code{defalias}) is @code{string<}.
4460 The arguments to @code{string-lessp} must be strings or symbols; the
4461 ordering is lexicographic, so case is significant. The print names of
4462 symbols are used instead of the symbols themselves.
4464 @cindex @samp{empty string} defined
4465 An empty string, @samp{""}, a string with no characters in it, is
4466 smaller than any string of characters.
4468 @code{string-equal} provides the corresponding test for equality. Its
4469 shorter, alternative name is @code{string=}. There are no string test
4470 functions that correspond to @var{>}, @code{>=}, or @code{<=}.
4473 Print a message in the echo area. The first argument is a string that
4474 can contain @samp{%s}, @samp{%d}, or @samp{%c} to print the value of
4475 arguments that follow the string. The argument used by @samp{%s} must
4476 be a string or a symbol; the argument used by @samp{%d} must be a
4477 number. The argument used by @samp{%c} must be an @sc{ascii} code
4478 number; it will be printed as the character with that @sc{ascii} code.
4479 (Various other %-sequences have not been mentioned.)
4483 The @code{setq} function sets the value of its first argument to the
4484 value of the second argument. The first argument is automatically
4485 quoted by @code{setq}. It does the same for succeeding pairs of
4486 arguments. Another function, @code{set}, takes only two arguments and
4487 evaluates both of them before setting the value returned by its first
4488 argument to the value returned by its second argument.
4491 Without an argument, return the name of the buffer, as a string.
4493 @item buffer-file-name
4494 Without an argument, return the name of the file the buffer is
4497 @item current-buffer
4498 Return the buffer in which Emacs is active; it may not be
4499 the buffer that is visible on the screen.
4502 Return the most recently selected buffer (other than the buffer passed
4503 to @code{other-buffer} as an argument and other than the current
4506 @item switch-to-buffer
4507 Select a buffer for Emacs to be active in and display it in the current
4508 window so users can look at it. Usually bound to @kbd{C-x b}.
4511 Switch Emacs's attention to a buffer on which programs will run. Don't
4512 alter what the window is showing.
4515 Return the number of characters in the current buffer.
4518 Return the value of the current position of the cursor, as an
4519 integer counting the number of characters from the beginning of the
4523 Return the minimum permissible value of point in
4524 the current buffer. This is 1, unless narrowing is in effect.
4527 Return the value of the maximum permissible value of point in the
4528 current buffer. This is the end of the buffer, unless narrowing is in
4533 @node defun Exercises
4538 Write a non-interactive function that doubles the value of its
4539 argument, a number. Make that function interactive.
4542 Write a function that tests whether the current value of
4543 @code{fill-column} is greater than the argument passed to the function,
4544 and if so, prints an appropriate message.
4547 @node Buffer Walk Through
4548 @chapter A Few Buffer--Related Functions
4550 In this chapter we study in detail several of the functions used in GNU
4551 Emacs. This is called a ``walk-through''. These functions are used as
4552 examples of Lisp code, but are not imaginary examples; with the
4553 exception of the first, simplified function definition, these functions
4554 show the actual code used in GNU Emacs. You can learn a great deal from
4555 these definitions. The functions described here are all related to
4556 buffers. Later, we will study other functions.
4559 * Finding More:: How to find more information.
4560 * simplified-beginning-of-buffer:: Shows @code{goto-char},
4561 @code{point-min}, and @code{push-mark}.
4562 * mark-whole-buffer:: Almost the same as @code{beginning-of-buffer}.
4563 * append-to-buffer:: Uses @code{save-excursion} and
4564 @code{insert-buffer-substring}.
4565 * Buffer Related Review:: Review.
4566 * Buffer Exercises::
4570 @section Finding More Information
4572 @findex describe-function, @r{introduced}
4573 @cindex Find function documentation
4574 In this walk-through, I will describe each new function as we come to
4575 it, sometimes in detail and sometimes briefly. If you are interested,
4576 you can get the full documentation of any Emacs Lisp function at any
4577 time by typing @kbd{C-h f} and then the name of the function (and then
4578 @key{RET}). Similarly, you can get the full documentation for a
4579 variable by typing @kbd{C-h v} and then the name of the variable (and
4582 @cindex Find source of function
4583 @c In version 22, tells location both of C and of Emacs Lisp
4584 Also, @code{describe-function} will tell you the location of the
4585 function definition.
4587 Put point into the name of the file that contains the function and
4588 press the @key{RET} key. In this case, @key{RET} means
4589 @code{push-button} rather than `return' or `enter'. Emacs will take
4590 you directly to the function definition.
4595 If you move point over the file name and press
4596 the @key{RET} key, which in this case means @code{help-follow} rather
4597 than `return' or `enter', Emacs will take you directly to the function
4601 More generally, if you want to see a function in its original source
4602 file, you can use the @code{find-tag} function to jump to it.
4603 @code{find-tag} works with a wide variety of languages, not just
4604 Lisp, and C, and it works with non-programming text as well. For
4605 example, @code{find-tag} will jump to the various nodes in the
4606 Texinfo source file of this document.
4607 The @code{find-tag} function depends on `tags tables' that record
4608 the locations of the functions, variables, and other items to which
4609 @code{find-tag} jumps.
4611 To use the @code{find-tag} command, type @kbd{M-.} (i.e., press the
4612 period key while holding down the @key{META} key, or else type the
4613 @key{ESC} key and then type the period key), and then, at the prompt,
4614 type in the name of the function whose source code you want to see,
4615 such as @code{mark-whole-buffer}, and then type @key{RET}. Emacs will
4616 switch buffers and display the source code for the function on your
4617 screen. To switch back to your current buffer, type @kbd{C-x b
4618 @key{RET}}. (On some keyboards, the @key{META} key is labeled
4621 @c !!! 22.1.1 tags table location in this paragraph
4622 @cindex TAGS table, specifying
4624 Depending on how the initial default values of your copy of Emacs are
4625 set, you may also need to specify the location of your `tags table',
4626 which is a file called @file{TAGS}. For example, if you are
4627 interested in Emacs sources, the tags table you will most likely want,
4628 if it has already been created for you, will be in a subdirectory of
4629 the @file{/usr/local/share/emacs/} directory; thus you would use the
4630 @code{M-x visit-tags-table} command and specify a pathname such as
4631 @file{/usr/local/share/emacs/22.1.1/lisp/TAGS}. If the tags table
4632 has not already been created, you will have to create it yourself. It
4633 will be in a file such as @file{/usr/local/src/emacs/src/TAGS}.
4636 To create a @file{TAGS} file in a specific directory, switch to that
4637 directory in Emacs using @kbd{M-x cd} command, or list the directory
4638 with @kbd{C-x d} (@code{dired}). Then run the compile command, with
4639 @w{@code{etags *.el}} as the command to execute:
4642 M-x compile RET etags *.el RET
4645 For more information, see @ref{etags, , Create Your Own @file{TAGS} File}.
4647 After you become more familiar with Emacs Lisp, you will find that you will
4648 frequently use @code{find-tag} to navigate your way around source code;
4649 and you will create your own @file{TAGS} tables.
4651 @cindex Library, as term for `file'
4652 Incidentally, the files that contain Lisp code are conventionally
4653 called @dfn{libraries}. The metaphor is derived from that of a
4654 specialized library, such as a law library or an engineering library,
4655 rather than a general library. Each library, or file, contains
4656 functions that relate to a particular topic or activity, such as
4657 @file{abbrev.el} for handling abbreviations and other typing
4658 shortcuts, and @file{help.el} for on-line help. (Sometimes several
4659 libraries provide code for a single activity, as the various
4660 @file{rmail@dots{}} files provide code for reading electronic mail.)
4661 In @cite{The GNU Emacs Manual}, you will see sentences such as ``The
4662 @kbd{C-h p} command lets you search the standard Emacs Lisp libraries
4663 by topic keywords.''
4665 @node simplified-beginning-of-buffer
4666 @section A Simplified @code{beginning-of-buffer} Definition
4667 @findex simplified-beginning-of-buffer
4669 The @code{beginning-of-buffer} command is a good function to start with
4670 since you are likely to be familiar with it and it is easy to
4671 understand. Used as an interactive command, @code{beginning-of-buffer}
4672 moves the cursor to the beginning of the buffer, leaving the mark at the
4673 previous position. It is generally bound to @kbd{M-<}.
4675 In this section, we will discuss a shortened version of the function
4676 that shows how it is most frequently used. This shortened function
4677 works as written, but it does not contain the code for a complex option.
4678 In another section, we will describe the entire function.
4679 (@xref{beginning-of-buffer, , Complete Definition of
4680 @code{beginning-of-buffer}}.)
4682 Before looking at the code, let's consider what the function
4683 definition has to contain: it must include an expression that makes
4684 the function interactive so it can be called by typing @kbd{M-x
4685 beginning-of-buffer} or by typing a keychord such as @kbd{M-<}; it
4686 must include code to leave a mark at the original position in the
4687 buffer; and it must include code to move the cursor to the beginning
4691 Here is the complete text of the shortened version of the function:
4695 (defun simplified-beginning-of-buffer ()
4696 "Move point to the beginning of the buffer;
4697 leave mark at previous position."
4700 (goto-char (point-min)))
4704 Like all function definitions, this definition has five parts following
4705 the macro @code{defun}:
4709 The name: in this example, @code{simplified-beginning-of-buffer}.
4712 A list of the arguments: in this example, an empty list, @code{()},
4715 The documentation string.
4718 The interactive expression.
4725 In this function definition, the argument list is empty; this means that
4726 this function does not require any arguments. (When we look at the
4727 definition for the complete function, we will see that it may be passed
4728 an optional argument.)
4730 The interactive expression tells Emacs that the function is intended to
4731 be used interactively. In this example, @code{interactive} does not have
4732 an argument because @code{simplified-beginning-of-buffer} does not
4736 The body of the function consists of the two lines:
4741 (goto-char (point-min))
4745 The first of these lines is the expression, @code{(push-mark)}. When
4746 this expression is evaluated by the Lisp interpreter, it sets a mark at
4747 the current position of the cursor, wherever that may be. The position
4748 of this mark is saved in the mark ring.
4750 The next line is @code{(goto-char (point-min))}. This expression
4751 jumps the cursor to the minimum point in the buffer, that is, to the
4752 beginning of the buffer (or to the beginning of the accessible portion
4753 of the buffer if it is narrowed. @xref{Narrowing & Widening, ,
4754 Narrowing and Widening}.)
4756 The @code{push-mark} command sets a mark at the place where the cursor
4757 was located before it was moved to the beginning of the buffer by the
4758 @code{(goto-char (point-min))} expression. Consequently, you can, if
4759 you wish, go back to where you were originally by typing @kbd{C-x C-x}.
4761 That is all there is to the function definition!
4763 @findex describe-function
4764 When you are reading code such as this and come upon an unfamiliar
4765 function, such as @code{goto-char}, you can find out what it does by
4766 using the @code{describe-function} command. To use this command, type
4767 @kbd{C-h f} and then type in the name of the function and press
4768 @key{RET}. The @code{describe-function} command will print the
4769 function's documentation string in a @file{*Help*} window. For
4770 example, the documentation for @code{goto-char} is:
4774 Set point to POSITION, a number or marker.
4775 Beginning of buffer is position (point-min), end is (point-max).
4780 The function's one argument is the desired position.
4783 (The prompt for @code{describe-function} will offer you the symbol
4784 under or preceding the cursor, so you can save typing by positioning
4785 the cursor right over or after the function and then typing @kbd{C-h f
4788 The @code{end-of-buffer} function definition is written in the same way as
4789 the @code{beginning-of-buffer} definition except that the body of the
4790 function contains the expression @code{(goto-char (point-max))} in place
4791 of @code{(goto-char (point-min))}.
4793 @node mark-whole-buffer
4794 @section The Definition of @code{mark-whole-buffer}
4795 @findex mark-whole-buffer
4797 The @code{mark-whole-buffer} function is no harder to understand than the
4798 @code{simplified-beginning-of-buffer} function. In this case, however,
4799 we will look at the complete function, not a shortened version.
4801 The @code{mark-whole-buffer} function is not as commonly used as the
4802 @code{beginning-of-buffer} function, but is useful nonetheless: it
4803 marks a whole buffer as a region by putting point at the beginning and
4804 a mark at the end of the buffer. It is generally bound to @kbd{C-x
4808 * mark-whole-buffer overview::
4809 * Body of mark-whole-buffer:: Only three lines of code.
4813 @node mark-whole-buffer overview
4814 @unnumberedsubsec An overview of @code{mark-whole-buffer}
4818 In GNU Emacs 22, the code for the complete function looks like this:
4822 (defun mark-whole-buffer ()
4823 "Put point at beginning and mark at end of buffer.
4824 You probably should not use this function in Lisp programs;
4825 it is usually a mistake for a Lisp function to use any subroutine
4826 that uses or sets the mark."
4829 (push-mark (point-max) nil t)
4830 (goto-char (point-min)))
4835 Like all other functions, the @code{mark-whole-buffer} function fits
4836 into the template for a function definition. The template looks like
4841 (defun @var{name-of-function} (@var{argument-list})
4842 "@var{documentation}@dots{}"
4843 (@var{interactive-expression}@dots{})
4848 Here is how the function works: the name of the function is
4849 @code{mark-whole-buffer}; it is followed by an empty argument list,
4850 @samp{()}, which means that the function does not require arguments.
4851 The documentation comes next.
4853 The next line is an @code{(interactive)} expression that tells Emacs
4854 that the function will be used interactively. These details are similar
4855 to the @code{simplified-beginning-of-buffer} function described in the
4859 @node Body of mark-whole-buffer
4860 @subsection Body of @code{mark-whole-buffer}
4862 The body of the @code{mark-whole-buffer} function consists of three
4869 (push-mark (point-max) nil t)
4870 (goto-char (point-min))
4874 The first of these lines is the expression, @code{(push-mark (point))}.
4876 This line does exactly the same job as the first line of the body of
4877 the @code{simplified-beginning-of-buffer} function, which is written
4878 @code{(push-mark)}. In both cases, the Lisp interpreter sets a mark
4879 at the current position of the cursor.
4881 I don't know why the expression in @code{mark-whole-buffer} is written
4882 @code{(push-mark (point))} and the expression in
4883 @code{beginning-of-buffer} is written @code{(push-mark)}. Perhaps
4884 whoever wrote the code did not know that the arguments for
4885 @code{push-mark} are optional and that if @code{push-mark} is not
4886 passed an argument, the function automatically sets mark at the
4887 location of point by default. Or perhaps the expression was written
4888 so as to parallel the structure of the next line. In any case, the
4889 line causes Emacs to determine the position of point and set a mark
4892 In earlier versions of GNU Emacs, the next line of
4893 @code{mark-whole-buffer} was @code{(push-mark (point-max))}. This
4894 expression sets a mark at the point in the buffer that has the highest
4895 number. This will be the end of the buffer (or, if the buffer is
4896 narrowed, the end of the accessible portion of the buffer.
4897 @xref{Narrowing & Widening, , Narrowing and Widening}, for more about
4898 narrowing.) After this mark has been set, the previous mark, the one
4899 set at point, is no longer set, but Emacs remembers its position, just
4900 as all other recent marks are always remembered. This means that you
4901 can, if you wish, go back to that position by typing @kbd{C-u
4905 In GNU Emacs 22, the @code{(point-max)} is slightly more complicated.
4909 (push-mark (point-max) nil t)
4913 The expression works nearly the same as before. It sets a mark at the
4914 highest numbered place in the buffer that it can. However, in this
4915 version, @code{push-mark} has two additional arguments. The second
4916 argument to @code{push-mark} is @code{nil}. This tells the function
4917 it @emph{should} display a message that says `Mark set' when it pushes
4918 the mark. The third argument is @code{t}. This tells
4919 @code{push-mark} to activate the mark when Transient Mark mode is
4920 turned on. Transient Mark mode highlights the currently active
4921 region. It is often turned off.
4923 Finally, the last line of the function is @code{(goto-char
4924 (point-min)))}. This is written exactly the same way as it is written
4925 in @code{beginning-of-buffer}. The expression moves the cursor to
4926 the minimum point in the buffer, that is, to the beginning of the buffer
4927 (or to the beginning of the accessible portion of the buffer). As a
4928 result of this, point is placed at the beginning of the buffer and mark
4929 is set at the end of the buffer. The whole buffer is, therefore, the
4932 @node append-to-buffer
4933 @section The Definition of @code{append-to-buffer}
4934 @findex append-to-buffer
4936 The @code{append-to-buffer} command is more complex than the
4937 @code{mark-whole-buffer} command. What it does is copy the region
4938 (that is, the part of the buffer between point and mark) from the
4939 current buffer to a specified buffer.
4942 * append-to-buffer overview::
4943 * append interactive:: A two part interactive expression.
4944 * append-to-buffer body:: Incorporates a @code{let} expression.
4945 * append save-excursion:: How the @code{save-excursion} works.
4949 @node append-to-buffer overview
4950 @unnumberedsubsec An Overview of @code{append-to-buffer}
4953 @findex insert-buffer-substring
4954 The @code{append-to-buffer} command uses the
4955 @code{insert-buffer-substring} function to copy the region.
4956 @code{insert-buffer-substring} is described by its name: it takes a
4957 string of characters from part of a buffer, a ``substring'', and
4958 inserts them into another buffer.
4960 Most of @code{append-to-buffer} is
4961 concerned with setting up the conditions for
4962 @code{insert-buffer-substring} to work: the code must specify both the
4963 buffer to which the text will go, the window it comes from and goes
4964 to, and the region that will be copied.
4967 Here is the complete text of the function:
4971 (defun append-to-buffer (buffer start end)
4972 "Append to specified buffer the text of the region.
4973 It is inserted into that buffer before its point.
4977 When calling from a program, give three arguments:
4978 BUFFER (or buffer name), START and END.
4979 START and END specify the portion of the current buffer to be copied."
4981 (list (read-buffer "Append to buffer: " (other-buffer
4982 (current-buffer) t))
4983 (region-beginning) (region-end)))
4986 (let ((oldbuf (current-buffer)))
4988 (let* ((append-to (get-buffer-create buffer))
4989 (windows (get-buffer-window-list append-to t t))
4991 (set-buffer append-to)
4992 (setq point (point))
4993 (barf-if-buffer-read-only)
4994 (insert-buffer-substring oldbuf start end)
4995 (dolist (window windows)
4996 (when (= (window-point window) point)
4997 (set-window-point window (point))))))))
5001 The function can be understood by looking at it as a series of
5002 filled-in templates.
5004 The outermost template is for the function definition. In this
5005 function, it looks like this (with several slots filled in):
5009 (defun append-to-buffer (buffer start end)
5010 "@var{documentation}@dots{}"
5011 (interactive @dots{})
5016 The first line of the function includes its name and three arguments.
5017 The arguments are the @code{buffer} to which the text will be copied, and
5018 the @code{start} and @code{end} of the region in the current buffer that
5021 The next part of the function is the documentation, which is clear and
5022 complete. As is conventional, the three arguments are written in
5023 upper case so you will notice them easily. Even better, they are
5024 described in the same order as in the argument list.
5026 Note that the documentation distinguishes between a buffer and its
5027 name. (The function can handle either.)
5029 @node append interactive
5030 @subsection The @code{append-to-buffer} Interactive Expression
5032 Since the @code{append-to-buffer} function will be used interactively,
5033 the function must have an @code{interactive} expression. (For a
5034 review of @code{interactive}, see @ref{Interactive, , Making a
5035 Function Interactive}.) The expression reads as follows:
5041 "Append to buffer: "
5042 (other-buffer (current-buffer) t))
5049 This expression is not one with letters standing for parts, as
5050 described earlier. Instead, it starts a list with these parts:
5052 The first part of the list is an expression to read the name of a
5053 buffer and return it as a string. That is @code{read-buffer}. The
5054 function requires a prompt as its first argument, @samp{"Append to
5055 buffer: "}. Its second argument tells the command what value to
5056 provide if you don't specify anything.
5058 In this case that second argument is an expression containing the
5059 function @code{other-buffer}, an exception, and a @samp{t}, standing
5062 The first argument to @code{other-buffer}, the exception, is yet
5063 another function, @code{current-buffer}. That is not going to be
5064 returned. The second argument is the symbol for true, @code{t}. that
5065 tells @code{other-buffer} that it may show visible buffers (except in
5066 this case, it will not show the current buffer, which makes sense).
5069 The expression looks like this:
5072 (other-buffer (current-buffer) t)
5075 The second and third arguments to the @code{list} expression are
5076 @code{(region-beginning)} and @code{(region-end)}. These two
5077 functions specify the beginning and end of the text to be appended.
5080 Originally, the command used the letters @samp{B} and @samp{r}.
5081 The whole @code{interactive} expression looked like this:
5084 (interactive "BAppend to buffer:@: \nr")
5088 But when that was done, the default value of the buffer switched to
5089 was invisible. That was not wanted.
5091 (The prompt was separated from the second argument with a newline,
5092 @samp{\n}. It was followed by an @samp{r} that told Emacs to bind the
5093 two arguments that follow the symbol @code{buffer} in the function's
5094 argument list (that is, @code{start} and @code{end}) to the values of
5095 point and mark. That argument worked fine.)
5097 @node append-to-buffer body
5098 @subsection The Body of @code{append-to-buffer}
5101 in GNU Emacs 22 in /usr/local/src/emacs/lisp/simple.el
5103 (defun append-to-buffer (buffer start end)
5104 "Append to specified buffer the text of the region.
5105 It is inserted into that buffer before its point.
5107 When calling from a program, give three arguments:
5108 BUFFER (or buffer name), START and END.
5109 START and END specify the portion of the current buffer to be copied."
5111 (list (read-buffer "Append to buffer: " (other-buffer (current-buffer) t))
5112 (region-beginning) (region-end)))
5113 (let ((oldbuf (current-buffer)))
5115 (let* ((append-to (get-buffer-create buffer))
5116 (windows (get-buffer-window-list append-to t t))
5118 (set-buffer append-to)
5119 (setq point (point))
5120 (barf-if-buffer-read-only)
5121 (insert-buffer-substring oldbuf start end)
5122 (dolist (window windows)
5123 (when (= (window-point window) point)
5124 (set-window-point window (point))))))))
5127 The body of the @code{append-to-buffer} function begins with @code{let}.
5129 As we have seen before (@pxref{let, , @code{let}}), the purpose of a
5130 @code{let} expression is to create and give initial values to one or
5131 more variables that will only be used within the body of the
5132 @code{let}. This means that such a variable will not be confused with
5133 any variable of the same name outside the @code{let} expression.
5135 We can see how the @code{let} expression fits into the function as a
5136 whole by showing a template for @code{append-to-buffer} with the
5137 @code{let} expression in outline:
5141 (defun append-to-buffer (buffer start end)
5142 "@var{documentation}@dots{}"
5143 (interactive @dots{})
5144 (let ((@var{variable} @var{value}))
5149 The @code{let} expression has three elements:
5153 The symbol @code{let};
5156 A varlist containing, in this case, a single two-element list,
5157 @code{(@var{variable} @var{value})};
5160 The body of the @code{let} expression.
5164 In the @code{append-to-buffer} function, the varlist looks like this:
5167 (oldbuf (current-buffer))
5171 In this part of the @code{let} expression, the one variable,
5172 @code{oldbuf}, is bound to the value returned by the
5173 @code{(current-buffer)} expression. The variable, @code{oldbuf}, is
5174 used to keep track of the buffer in which you are working and from
5175 which you will copy.
5177 The element or elements of a varlist are surrounded by a set of
5178 parentheses so the Lisp interpreter can distinguish the varlist from
5179 the body of the @code{let}. As a consequence, the two-element list
5180 within the varlist is surrounded by a circumscribing set of parentheses.
5181 The line looks like this:
5185 (let ((oldbuf (current-buffer)))
5191 The two parentheses before @code{oldbuf} might surprise you if you did
5192 not realize that the first parenthesis before @code{oldbuf} marks the
5193 boundary of the varlist and the second parenthesis marks the beginning
5194 of the two-element list, @code{(oldbuf (current-buffer))}.
5196 @node append save-excursion
5197 @subsection @code{save-excursion} in @code{append-to-buffer}
5199 The body of the @code{let} expression in @code{append-to-buffer}
5200 consists of a @code{save-excursion} expression.
5202 The @code{save-excursion} function saves the locations of point and
5203 mark, and restores them to those positions after the expressions in the
5204 body of the @code{save-excursion} complete execution. In addition,
5205 @code{save-excursion} keeps track of the original buffer, and
5206 restores it. This is how @code{save-excursion} is used in
5207 @code{append-to-buffer}.
5210 @cindex Indentation for formatting
5211 @cindex Formatting convention
5212 Incidentally, it is worth noting here that a Lisp function is normally
5213 formatted so that everything that is enclosed in a multi-line spread is
5214 indented more to the right than the first symbol. In this function
5215 definition, the @code{let} is indented more than the @code{defun}, and
5216 the @code{save-excursion} is indented more than the @code{let}, like
5232 This formatting convention makes it easy to see that the lines in
5233 the body of the @code{save-excursion} are enclosed by the parentheses
5234 associated with @code{save-excursion}, just as the
5235 @code{save-excursion} itself is enclosed by the parentheses associated
5236 with the @code{let}:
5240 (let ((oldbuf (current-buffer)))
5243 (set-buffer @dots{})
5244 (insert-buffer-substring oldbuf start end)
5250 The use of the @code{save-excursion} function can be viewed as a process
5251 of filling in the slots of a template:
5256 @var{first-expression-in-body}
5257 @var{second-expression-in-body}
5259 @var{last-expression-in-body})
5265 In this function, the body of the @code{save-excursion} contains only
5266 one expression, the @code{let*} expression. You know about a
5267 @code{let} function. The @code{let*} function is different. It has a
5268 @samp{*} in its name. It enables Emacs to set each variable in its
5269 varlist in sequence, one after another.
5271 Its critical feature is that variables later in the varlist can make
5272 use of the values to which Emacs set variables earlier in the varlist.
5273 @xref{fwd-para let, , The @code{let*} expression}.
5275 We will skip functions like @code{let*} and focus on two: the
5276 @code{set-buffer} function and the @code{insert-buffer-substring}
5280 In the old days, the @code{set-buffer} expression was simply
5283 (set-buffer (get-buffer-create buffer))
5291 (set-buffer append-to)
5295 @code{append-to} is bound to @code{(get-buffer-create buffer)} earlier
5296 on in the @code{let*} expression. That extra binding would not be
5297 necessary except for that @code{append-to} is used later in the
5298 varlist as an argument to @code{get-buffer-window-list}.
5303 (let ((oldbuf (current-buffer)))
5305 (let* ((append-to (get-buffer-create buffer))
5306 (windows (get-buffer-window-list append-to t t))
5308 (set-buffer append-to)
5309 (setq point (point))
5310 (barf-if-buffer-read-only)
5311 (insert-buffer-substring oldbuf start end)
5312 (dolist (window windows)
5313 (when (= (window-point window) point)
5314 (set-window-point window (point))))))))
5317 The @code{append-to-buffer} function definition inserts text from the
5318 buffer in which you are currently to a named buffer. It happens that
5319 @code{insert-buffer-substring} copies text from another buffer to the
5320 current buffer, just the reverse---that is why the
5321 @code{append-to-buffer} definition starts out with a @code{let} that
5322 binds the local symbol @code{oldbuf} to the value returned by
5323 @code{current-buffer}.
5326 The @code{insert-buffer-substring} expression looks like this:
5329 (insert-buffer-substring oldbuf start end)
5333 The @code{insert-buffer-substring} function copies a string
5334 @emph{from} the buffer specified as its first argument and inserts the
5335 string into the present buffer. In this case, the argument to
5336 @code{insert-buffer-substring} is the value of the variable created
5337 and bound by the @code{let}, namely the value of @code{oldbuf}, which
5338 was the current buffer when you gave the @code{append-to-buffer}
5341 After @code{insert-buffer-substring} has done its work,
5342 @code{save-excursion} will restore the action to the original buffer
5343 and @code{append-to-buffer} will have done its job.
5346 Written in skeletal form, the workings of the body look like this:
5350 (let (@var{bind-}@code{oldbuf}@var{-to-value-of-}@code{current-buffer})
5351 (save-excursion ; @r{Keep track of buffer.}
5353 @var{insert-substring-from-}@code{oldbuf}@var{-into-buffer})
5355 @var{change-back-to-original-buffer-when-finished}
5356 @var{let-the-local-meaning-of-}@code{oldbuf}@var{-disappear-when-finished}
5360 In summary, @code{append-to-buffer} works as follows: it saves the
5361 value of the current buffer in the variable called @code{oldbuf}. It
5362 gets the new buffer (creating one if need be) and switches Emacs's
5363 attention to it. Using the value of @code{oldbuf}, it inserts the
5364 region of text from the old buffer into the new buffer; and then using
5365 @code{save-excursion}, it brings you back to your original buffer.
5367 In looking at @code{append-to-buffer}, you have explored a fairly
5368 complex function. It shows how to use @code{let} and
5369 @code{save-excursion}, and how to change to and come back from another
5370 buffer. Many function definitions use @code{let},
5371 @code{save-excursion}, and @code{set-buffer} this way.
5373 @node Buffer Related Review
5376 Here is a brief summary of the various functions discussed in this chapter.
5379 @item describe-function
5380 @itemx describe-variable
5381 Print the documentation for a function or variable.
5382 Conventionally bound to @kbd{C-h f} and @kbd{C-h v}.
5385 Find the file containing the source for a function or variable and
5386 switch buffers to it, positioning point at the beginning of the item.
5387 Conventionally bound to @kbd{M-.} (that's a period following the
5390 @item save-excursion
5391 Save the location of point and mark and restore their values after the
5392 arguments to @code{save-excursion} have been evaluated. Also, remember
5393 the current buffer and return to it.
5396 Set mark at a location and record the value of the previous mark on the
5397 mark ring. The mark is a location in the buffer that will keep its
5398 relative position even if text is added to or removed from the buffer.
5401 Set point to the location specified by the value of the argument, which
5402 can be a number, a marker, or an expression that returns the number of
5403 a position, such as @code{(point-min)}.
5405 @item insert-buffer-substring
5406 Copy a region of text from a buffer that is passed to the function as
5407 an argument and insert the region into the current buffer.
5409 @item mark-whole-buffer
5410 Mark the whole buffer as a region. Normally bound to @kbd{C-x h}.
5413 Switch the attention of Emacs to another buffer, but do not change the
5414 window being displayed. Used when the program rather than a human is
5415 to work on a different buffer.
5417 @item get-buffer-create
5419 Find a named buffer or create one if a buffer of that name does not
5420 exist. The @code{get-buffer} function returns @code{nil} if the named
5421 buffer does not exist.
5425 @node Buffer Exercises
5430 Write your own @code{simplified-end-of-buffer} function definition;
5431 then test it to see whether it works.
5434 Use @code{if} and @code{get-buffer} to write a function that prints a
5435 message telling you whether a buffer exists.
5438 Using @code{find-tag}, find the source for the @code{copy-to-buffer}
5443 @chapter A Few More Complex Functions
5445 In this chapter, we build on what we have learned in previous chapters
5446 by looking at more complex functions. The @code{copy-to-buffer}
5447 function illustrates use of two @code{save-excursion} expressions in
5448 one definition, while the @code{insert-buffer} function illustrates
5449 use of an asterisk in an @code{interactive} expression, use of
5450 @code{or}, and the important distinction between a name and the object
5451 to which the name refers.
5454 * copy-to-buffer:: With @code{set-buffer}, @code{get-buffer-create}.
5455 * insert-buffer:: Read-only, and with @code{or}.
5456 * beginning-of-buffer:: Shows @code{goto-char},
5457 @code{point-min}, and @code{push-mark}.
5458 * Second Buffer Related Review::
5459 * optional Exercise::
5462 @node copy-to-buffer
5463 @section The Definition of @code{copy-to-buffer}
5464 @findex copy-to-buffer
5466 After understanding how @code{append-to-buffer} works, it is easy to
5467 understand @code{copy-to-buffer}. This function copies text into a
5468 buffer, but instead of adding to the second buffer, it replaces all the
5469 previous text in the second buffer.
5472 The body of @code{copy-to-buffer} looks like this,
5477 (interactive "BCopy to buffer: \nr")
5478 (let ((oldbuf (current-buffer)))
5479 (with-current-buffer (get-buffer-create buffer)
5480 (barf-if-buffer-read-only)
5483 (insert-buffer-substring oldbuf start end)))))
5487 The @code{copy-to-buffer} function has a simpler @code{interactive}
5488 expression than @code{append-to-buffer}.
5491 The definition then says
5494 (with-current-buffer (get-buffer-create buffer) @dots{}
5497 First, look at the earliest inner expression; that is evaluated first.
5498 That expression starts with @code{get-buffer-create buffer}. The
5499 function tells the computer to use the buffer with the name specified
5500 as the one to which you are copying, or if such a buffer does not
5501 exist, to create it. Then, the @code{with-current-buffer} function
5502 evaluates its body with that buffer temporarily current.
5504 (This demonstrates another way to shift the computer's attention but
5505 not the user's. The @code{append-to-buffer} function showed how to do
5506 the same with @code{save-excursion} and @code{set-buffer}.
5507 @code{with-current-buffer} is a newer, and arguably easier,
5510 The @code{barf-if-buffer-read-only} function sends you an error
5511 message saying the buffer is read-only if you cannot modify it.
5513 The next line has the @code{erase-buffer} function as its sole
5514 contents. That function erases the buffer.
5516 Finally, the last two lines contain the @code{save-excursion}
5517 expression with @code{insert-buffer-substring} as its body.
5518 The @code{insert-buffer-substring} expression copies the text from
5519 the buffer you are in (and you have not seen the computer shift its
5520 attention, so you don't know that that buffer is now called
5523 Incidentally, this is what is meant by `replacement'. To replace text,
5524 Emacs erases the previous text and then inserts new text.
5527 In outline, the body of @code{copy-to-buffer} looks like this:
5531 (let (@var{bind-}@code{oldbuf}@var{-to-value-of-}@code{current-buffer})
5532 (@var{with-the-buffer-you-are-copying-to}
5533 (@var{but-do-not-erase-or-copy-to-a-read-only-buffer})
5536 @var{insert-substring-from-}@code{oldbuf}@var{-into-buffer})))
5541 @section The Definition of @code{insert-buffer}
5542 @findex insert-buffer
5544 @code{insert-buffer} is yet another buffer-related function. This
5545 command copies another buffer @emph{into} the current buffer. It is the
5546 reverse of @code{append-to-buffer} or @code{copy-to-buffer}, since they
5547 copy a region of text @emph{from} the current buffer to another buffer.
5549 Here is a discussion based on the original code. The code was
5550 simplified in 2003 and is harder to understand.
5552 (@xref{New insert-buffer, , New Body for @code{insert-buffer}}, to see
5553 a discussion of the new body.)
5555 In addition, this code illustrates the use of @code{interactive} with a
5556 buffer that might be @dfn{read-only} and the important distinction
5557 between the name of an object and the object actually referred to.
5560 * insert-buffer code::
5561 * insert-buffer interactive:: When you can read, but not write.
5562 * insert-buffer body:: The body has an @code{or} and a @code{let}.
5563 * if & or:: Using an @code{if} instead of an @code{or}.
5564 * Insert or:: How the @code{or} expression works.
5565 * Insert let:: Two @code{save-excursion} expressions.
5566 * New insert-buffer::
5570 @node insert-buffer code
5571 @unnumberedsubsec The Code for @code{insert-buffer}
5575 Here is the earlier code:
5579 (defun insert-buffer (buffer)
5580 "Insert after point the contents of BUFFER.
5581 Puts mark after the inserted text.
5582 BUFFER may be a buffer or a buffer name."
5583 (interactive "*bInsert buffer:@: ")
5586 (or (bufferp buffer)
5587 (setq buffer (get-buffer buffer)))
5588 (let (start end newmark)
5592 (setq start (point-min) end (point-max)))
5595 (insert-buffer-substring buffer start end)
5596 (setq newmark (point)))
5597 (push-mark newmark)))
5602 As with other function definitions, you can use a template to see an
5603 outline of the function:
5607 (defun insert-buffer (buffer)
5608 "@var{documentation}@dots{}"
5609 (interactive "*bInsert buffer:@: ")
5614 @node insert-buffer interactive
5615 @subsection The Interactive Expression in @code{insert-buffer}
5616 @findex interactive, @r{example use of}
5618 In @code{insert-buffer}, the argument to the @code{interactive}
5619 declaration has two parts, an asterisk, @samp{*}, and @samp{bInsert
5623 * Read-only buffer:: When a buffer cannot be modified.
5624 * b for interactive:: An existing buffer or else its name.
5627 @node Read-only buffer
5628 @unnumberedsubsubsec A Read-only Buffer
5629 @cindex Read-only buffer
5630 @cindex Asterisk for read-only buffer
5631 @findex * @r{for read-only buffer}
5633 The asterisk is for the situation when the current buffer is a
5634 read-only buffer---a buffer that cannot be modified. If
5635 @code{insert-buffer} is called when the current buffer is read-only, a
5636 message to this effect is printed in the echo area and the terminal
5637 may beep or blink at you; you will not be permitted to insert anything
5638 into current buffer. The asterisk does not need to be followed by a
5639 newline to separate it from the next argument.
5641 @node b for interactive
5642 @unnumberedsubsubsec @samp{b} in an Interactive Expression
5644 The next argument in the interactive expression starts with a lower
5645 case @samp{b}. (This is different from the code for
5646 @code{append-to-buffer}, which uses an upper-case @samp{B}.
5647 @xref{append-to-buffer, , The Definition of @code{append-to-buffer}}.)
5648 The lower-case @samp{b} tells the Lisp interpreter that the argument
5649 for @code{insert-buffer} should be an existing buffer or else its
5650 name. (The upper-case @samp{B} option provides for the possibility
5651 that the buffer does not exist.) Emacs will prompt you for the name
5652 of the buffer, offering you a default buffer, with name completion
5653 enabled. If the buffer does not exist, you receive a message that
5654 says ``No match''; your terminal may beep at you as well.
5656 The new and simplified code generates a list for @code{interactive}.
5657 It uses the @code{barf-if-buffer-read-only} and @code{read-buffer}
5658 functions with which we are already familiar and the @code{progn}
5659 special form with which we are not. (It will be described later.)
5661 @node insert-buffer body
5662 @subsection The Body of the @code{insert-buffer} Function
5664 The body of the @code{insert-buffer} function has two major parts: an
5665 @code{or} expression and a @code{let} expression. The purpose of the
5666 @code{or} expression is to ensure that the argument @code{buffer} is
5667 bound to a buffer and not just the name of a buffer. The body of the
5668 @code{let} expression contains the code which copies the other buffer
5669 into the current buffer.
5672 In outline, the two expressions fit into the @code{insert-buffer}
5677 (defun insert-buffer (buffer)
5678 "@var{documentation}@dots{}"
5679 (interactive "*bInsert buffer:@: ")
5684 (let (@var{varlist})
5685 @var{body-of-}@code{let}@dots{} )
5689 To understand how the @code{or} expression ensures that the argument
5690 @code{buffer} is bound to a buffer and not to the name of a buffer, it
5691 is first necessary to understand the @code{or} function.
5693 Before doing this, let me rewrite this part of the function using
5694 @code{if} so that you can see what is done in a manner that will be familiar.
5697 @subsection @code{insert-buffer} With an @code{if} Instead of an @code{or}
5699 The job to be done is to make sure the value of @code{buffer} is a
5700 buffer itself and not the name of a buffer. If the value is the name,
5701 then the buffer itself must be got.
5703 You can imagine yourself at a conference where an usher is wandering
5704 around holding a list with your name on it and looking for you: the
5705 usher is ``bound'' to your name, not to you; but when the usher finds
5706 you and takes your arm, the usher becomes ``bound'' to you.
5709 In Lisp, you might describe this situation like this:
5713 (if (not (holding-on-to-guest))
5714 (find-and-take-arm-of-guest))
5718 We want to do the same thing with a buffer---if we do not have the
5719 buffer itself, we want to get it.
5722 Using a predicate called @code{bufferp} that tells us whether we have a
5723 buffer (rather than its name), we can write the code like this:
5727 (if (not (bufferp buffer)) ; @r{if-part}
5728 (setq buffer (get-buffer buffer))) ; @r{then-part}
5733 Here, the true-or-false-test of the @code{if} expression is
5734 @w{@code{(not (bufferp buffer))}}; and the then-part is the expression
5735 @w{@code{(setq buffer (get-buffer buffer))}}.
5737 In the test, the function @code{bufferp} returns true if its argument is
5738 a buffer---but false if its argument is the name of the buffer. (The
5739 last character of the function name @code{bufferp} is the character
5740 @samp{p}; as we saw earlier, such use of @samp{p} is a convention that
5741 indicates that the function is a predicate, which is a term that means
5742 that the function will determine whether some property is true or false.
5743 @xref{Wrong Type of Argument, , Using the Wrong Type Object as an
5747 The function @code{not} precedes the expression @code{(bufferp buffer)},
5748 so the true-or-false-test looks like this:
5751 (not (bufferp buffer))
5755 @code{not} is a function that returns true if its argument is false
5756 and false if its argument is true. So if @code{(bufferp buffer)}
5757 returns true, the @code{not} expression returns false and vice-verse:
5758 what is ``not true'' is false and what is ``not false'' is true.
5760 Using this test, the @code{if} expression works as follows: when the
5761 value of the variable @code{buffer} is actually a buffer rather than
5762 its name, the true-or-false-test returns false and the @code{if}
5763 expression does not evaluate the then-part. This is fine, since we do
5764 not need to do anything to the variable @code{buffer} if it really is
5767 On the other hand, when the value of @code{buffer} is not a buffer
5768 itself, but the name of a buffer, the true-or-false-test returns true
5769 and the then-part of the expression is evaluated. In this case, the
5770 then-part is @code{(setq buffer (get-buffer buffer))}. This
5771 expression uses the @code{get-buffer} function to return an actual
5772 buffer itself, given its name. The @code{setq} then sets the variable
5773 @code{buffer} to the value of the buffer itself, replacing its previous
5774 value (which was the name of the buffer).
5777 @subsection The @code{or} in the Body
5779 The purpose of the @code{or} expression in the @code{insert-buffer}
5780 function is to ensure that the argument @code{buffer} is bound to a
5781 buffer and not just to the name of a buffer. The previous section shows
5782 how the job could have been done using an @code{if} expression.
5783 However, the @code{insert-buffer} function actually uses @code{or}.
5784 To understand this, it is necessary to understand how @code{or} works.
5787 An @code{or} function can have any number of arguments. It evaluates
5788 each argument in turn and returns the value of the first of its
5789 arguments that is not @code{nil}. Also, and this is a crucial feature
5790 of @code{or}, it does not evaluate any subsequent arguments after
5791 returning the first non-@code{nil} value.
5794 The @code{or} expression looks like this:
5798 (or (bufferp buffer)
5799 (setq buffer (get-buffer buffer)))
5804 The first argument to @code{or} is the expression @code{(bufferp buffer)}.
5805 This expression returns true (a non-@code{nil} value) if the buffer is
5806 actually a buffer, and not just the name of a buffer. In the @code{or}
5807 expression, if this is the case, the @code{or} expression returns this
5808 true value and does not evaluate the next expression---and this is fine
5809 with us, since we do not want to do anything to the value of
5810 @code{buffer} if it really is a buffer.
5812 On the other hand, if the value of @code{(bufferp buffer)} is @code{nil},
5813 which it will be if the value of @code{buffer} is the name of a buffer,
5814 the Lisp interpreter evaluates the next element of the @code{or}
5815 expression. This is the expression @code{(setq buffer (get-buffer
5816 buffer))}. This expression returns a non-@code{nil} value, which
5817 is the value to which it sets the variable @code{buffer}---and this
5818 value is a buffer itself, not the name of a buffer.
5820 The result of all this is that the symbol @code{buffer} is always
5821 bound to a buffer itself rather than to the name of a buffer. All
5822 this is necessary because the @code{set-buffer} function in a
5823 following line only works with a buffer itself, not with the name to a
5827 Incidentally, using @code{or}, the situation with the usher would be
5831 (or (holding-on-to-guest) (find-and-take-arm-of-guest))
5835 @subsection The @code{let} Expression in @code{insert-buffer}
5837 After ensuring that the variable @code{buffer} refers to a buffer itself
5838 and not just to the name of a buffer, the @code{insert-buffer function}
5839 continues with a @code{let} expression. This specifies three local
5840 variables, @code{start}, @code{end}, and @code{newmark} and binds them
5841 to the initial value @code{nil}. These variables are used inside the
5842 remainder of the @code{let} and temporarily hide any other occurrence of
5843 variables of the same name in Emacs until the end of the @code{let}.
5846 The body of the @code{let} contains two @code{save-excursion}
5847 expressions. First, we will look at the inner @code{save-excursion}
5848 expression in detail. The expression looks like this:
5854 (setq start (point-min) end (point-max)))
5859 The expression @code{(set-buffer buffer)} changes Emacs's attention
5860 from the current buffer to the one from which the text will copied.
5861 In that buffer, the variables @code{start} and @code{end} are set to
5862 the beginning and end of the buffer, using the commands
5863 @code{point-min} and @code{point-max}. Note that we have here an
5864 illustration of how @code{setq} is able to set two variables in the
5865 same expression. The first argument of @code{setq} is set to the
5866 value of its second, and its third argument is set to the value of its
5869 After the body of the inner @code{save-excursion} is evaluated, the
5870 @code{save-excursion} restores the original buffer, but @code{start} and
5871 @code{end} remain set to the values of the beginning and end of the
5872 buffer from which the text will be copied.
5875 The outer @code{save-excursion} expression looks like this:
5880 (@var{inner-}@code{save-excursion}@var{-expression}
5881 (@var{go-to-new-buffer-and-set-}@code{start}@var{-and-}@code{end})
5882 (insert-buffer-substring buffer start end)
5883 (setq newmark (point)))
5888 The @code{insert-buffer-substring} function copies the text
5889 @emph{into} the current buffer @emph{from} the region indicated by
5890 @code{start} and @code{end} in @code{buffer}. Since the whole of the
5891 second buffer lies between @code{start} and @code{end}, the whole of
5892 the second buffer is copied into the buffer you are editing. Next,
5893 the value of point, which will be at the end of the inserted text, is
5894 recorded in the variable @code{newmark}.
5896 After the body of the outer @code{save-excursion} is evaluated, point
5897 and mark are relocated to their original places.
5899 However, it is convenient to locate a mark at the end of the newly
5900 inserted text and locate point at its beginning. The @code{newmark}
5901 variable records the end of the inserted text. In the last line of
5902 the @code{let} expression, the @code{(push-mark newmark)} expression
5903 function sets a mark to this location. (The previous location of the
5904 mark is still accessible; it is recorded on the mark ring and you can
5905 go back to it with @kbd{C-u C-@key{SPC}}.) Meanwhile, point is
5906 located at the beginning of the inserted text, which is where it was
5907 before you called the insert function, the position of which was saved
5908 by the first @code{save-excursion}.
5911 The whole @code{let} expression looks like this:
5915 (let (start end newmark)
5919 (setq start (point-min) end (point-max)))
5920 (insert-buffer-substring buffer start end)
5921 (setq newmark (point)))
5922 (push-mark newmark))
5926 Like the @code{append-to-buffer} function, the @code{insert-buffer}
5927 function uses @code{let}, @code{save-excursion}, and
5928 @code{set-buffer}. In addition, the function illustrates one way to
5929 use @code{or}. All these functions are building blocks that we will
5930 find and use again and again.
5932 @node New insert-buffer
5933 @subsection New Body for @code{insert-buffer}
5934 @findex insert-buffer, new version body
5935 @findex new version body for insert-buffer
5937 The body in the GNU Emacs 22 version is more confusing than the original.
5940 It consists of two expressions,
5946 (insert-buffer-substring (get-buffer buffer))
5954 except, and this is what confuses novices, very important work is done
5955 inside the @code{push-mark} expression.
5957 The @code{get-buffer} function returns a buffer with the name
5958 provided. You will note that the function is @emph{not} called
5959 @code{get-buffer-create}; it does not create a buffer if one does not
5960 already exist. The buffer returned by @code{get-buffer}, an existing
5961 buffer, is passed to @code{insert-buffer-substring}, which inserts the
5962 whole of the buffer (since you did not specify anything else).
5964 The location into which the buffer is inserted is recorded by
5965 @code{push-mark}. Then the function returns @code{nil}, the value of
5966 its last command. Put another way, the @code{insert-buffer} function
5967 exists only to produce a side effect, inserting another buffer, not to
5970 @node beginning-of-buffer
5971 @section Complete Definition of @code{beginning-of-buffer}
5972 @findex beginning-of-buffer
5974 The basic structure of the @code{beginning-of-buffer} function has
5975 already been discussed. (@xref{simplified-beginning-of-buffer, , A
5976 Simplified @code{beginning-of-buffer} Definition}.)
5977 This section describes the complex part of the definition.
5979 As previously described, when invoked without an argument,
5980 @code{beginning-of-buffer} moves the cursor to the beginning of the
5981 buffer (in truth, the beginning of the accessible portion of the
5982 buffer), leaving the mark at the previous position. However, when the
5983 command is invoked with a number between one and ten, the function
5984 considers that number to be a fraction of the length of the buffer,
5985 measured in tenths, and Emacs moves the cursor that fraction of the
5986 way from the beginning of the buffer. Thus, you can either call this
5987 function with the key command @kbd{M-<}, which will move the cursor to
5988 the beginning of the buffer, or with a key command such as @kbd{C-u 7
5989 M-<} which will move the cursor to a point 70% of the way through the
5990 buffer. If a number bigger than ten is used for the argument, it
5991 moves to the end of the buffer.
5993 The @code{beginning-of-buffer} function can be called with or without an
5994 argument. The use of the argument is optional.
5997 * Optional Arguments::
5998 * beginning-of-buffer opt arg:: Example with optional argument.
5999 * beginning-of-buffer complete::
6002 @node Optional Arguments
6003 @subsection Optional Arguments
6005 Unless told otherwise, Lisp expects that a function with an argument in
6006 its function definition will be called with a value for that argument.
6007 If that does not happen, you get an error and a message that says
6008 @samp{Wrong number of arguments}.
6010 @cindex Optional arguments
6013 However, optional arguments are a feature of Lisp: a particular
6014 @dfn{keyword} is used to tell the Lisp interpreter that an argument is
6015 optional. The keyword is @code{&optional}. (The @samp{&} in front of
6016 @samp{optional} is part of the keyword.) In a function definition, if
6017 an argument follows the keyword @code{&optional}, no value need be
6018 passed to that argument when the function is called.
6021 The first line of the function definition of @code{beginning-of-buffer}
6022 therefore looks like this:
6025 (defun beginning-of-buffer (&optional arg)
6029 In outline, the whole function looks like this:
6033 (defun beginning-of-buffer (&optional arg)
6034 "@var{documentation}@dots{}"
6036 (or (@var{is-the-argument-a-cons-cell} arg)
6037 (and @var{are-both-transient-mark-mode-and-mark-active-true})
6039 (let (@var{determine-size-and-set-it})
6041 (@var{if-there-is-an-argument}
6042 @var{figure-out-where-to-go}
6049 The function is similar to the @code{simplified-beginning-of-buffer}
6050 function except that the @code{interactive} expression has @code{"P"}
6051 as an argument and the @code{goto-char} function is followed by an
6052 if-then-else expression that figures out where to put the cursor if
6053 there is an argument that is not a cons cell.
6055 (Since I do not explain a cons cell for many more chapters, please
6056 consider ignoring the function @code{consp}. @xref{List
6057 Implementation, , How Lists are Implemented}, and @ref{Cons Cell Type,
6058 , Cons Cell and List Types, elisp, The GNU Emacs Lisp Reference
6061 The @code{"P"} in the @code{interactive} expression tells Emacs to
6062 pass a prefix argument, if there is one, to the function in raw form.
6063 A prefix argument is made by typing the @key{META} key followed by a
6064 number, or by typing @kbd{C-u} and then a number. (If you don't type
6065 a number, @kbd{C-u} defaults to a cons cell with a 4. A lowercase
6066 @code{"p"} in the @code{interactive} expression causes the function to
6067 convert a prefix arg to a number.)
6069 The true-or-false-test of the @code{if} expression looks complex, but
6070 it is not: it checks whether @code{arg} has a value that is not
6071 @code{nil} and whether it is a cons cell. (That is what @code{consp}
6072 does; it checks whether its argument is a cons cell.) If @code{arg}
6073 has a value that is not @code{nil} (and is not a cons cell), which
6074 will be the case if @code{beginning-of-buffer} is called with a
6075 numeric argument, then this true-or-false-test will return true and
6076 the then-part of the @code{if} expression will be evaluated. On the
6077 other hand, if @code{beginning-of-buffer} is not called with an
6078 argument, the value of @code{arg} will be @code{nil} and the else-part
6079 of the @code{if} expression will be evaluated. The else-part is
6080 simply @code{point-min}, and when this is the outcome, the whole
6081 @code{goto-char} expression is @code{(goto-char (point-min))}, which
6082 is how we saw the @code{beginning-of-buffer} function in its
6085 @node beginning-of-buffer opt arg
6086 @subsection @code{beginning-of-buffer} with an Argument
6088 When @code{beginning-of-buffer} is called with an argument, an
6089 expression is evaluated which calculates what value to pass to
6090 @code{goto-char}. This expression is rather complicated at first sight.
6091 It includes an inner @code{if} expression and much arithmetic. It looks
6096 (if (> (buffer-size) 10000)
6097 ;; @r{Avoid overflow for large buffer sizes!}
6098 (* (prefix-numeric-value arg)
6103 size (prefix-numeric-value arg))) 10)))
6108 * Disentangle beginning-of-buffer::
6109 * Large buffer case::
6110 * Small buffer case::
6114 @node Disentangle beginning-of-buffer
6115 @unnumberedsubsubsec Disentangle @code{beginning-of-buffer}
6118 Like other complex-looking expressions, the conditional expression
6119 within @code{beginning-of-buffer} can be disentangled by looking at it
6120 as parts of a template, in this case, the template for an if-then-else
6121 expression. In skeletal form, the expression looks like this:
6125 (if (@var{buffer-is-large}
6126 @var{divide-buffer-size-by-10-and-multiply-by-arg}
6127 @var{else-use-alternate-calculation}
6131 The true-or-false-test of this inner @code{if} expression checks the
6132 size of the buffer. The reason for this is that the old version 18
6133 Emacs used numbers that are no bigger than eight million or so and in
6134 the computation that followed, the programmer feared that Emacs might
6135 try to use over-large numbers if the buffer were large. The term
6136 `overflow', mentioned in the comment, means numbers that are over
6137 large. More recent versions of Emacs use larger numbers, but this
6138 code has not been touched, if only because people now look at buffers
6139 that are far, far larger than ever before.
6141 There are two cases: if the buffer is large and if it is not.
6143 @node Large buffer case
6144 @unnumberedsubsubsec What happens in a large buffer
6146 In @code{beginning-of-buffer}, the inner @code{if} expression tests
6147 whether the size of the buffer is greater than 10,000 characters. To do
6148 this, it uses the @code{>} function and the computation of @code{size}
6149 that comes from the let expression.
6151 In the old days, the function @code{buffer-size} was used. Not only
6152 was that function called several times, it gave the size of the whole
6153 buffer, not the accessible part. The computation makes much more
6154 sense when it handles just the accessible part. (@xref{Narrowing &
6155 Widening, , Narrowing and Widening}, for more information on focusing
6156 attention to an `accessible' part.)
6159 The line looks like this:
6167 When the buffer is large, the then-part of the @code{if} expression is
6168 evaluated. It reads like this (after formatting for easy reading):
6173 (prefix-numeric-value arg)
6179 This expression is a multiplication, with two arguments to the function
6182 The first argument is @code{(prefix-numeric-value arg)}. When
6183 @code{"P"} is used as the argument for @code{interactive}, the value
6184 passed to the function as its argument is passed a ``raw prefix
6185 argument'', and not a number. (It is a number in a list.) To perform
6186 the arithmetic, a conversion is necessary, and
6187 @code{prefix-numeric-value} does the job.
6189 @findex / @r{(division)}
6191 The second argument is @code{(/ size 10)}. This expression divides
6192 the numeric value by ten---the numeric value of the size of the
6193 accessible portion of the buffer. This produces a number that tells
6194 how many characters make up one tenth of the buffer size. (In Lisp,
6195 @code{/} is used for division, just as @code{*} is used for
6199 In the multiplication expression as a whole, this amount is multiplied
6200 by the value of the prefix argument---the multiplication looks like this:
6204 (* @var{numeric-value-of-prefix-arg}
6205 @var{number-of-characters-in-one-tenth-of-the-accessible-buffer})
6210 If, for example, the prefix argument is @samp{7}, the one-tenth value
6211 will be multiplied by 7 to give a position 70% of the way through.
6214 The result of all this is that if the accessible portion of the buffer
6215 is large, the @code{goto-char} expression reads like this:
6219 (goto-char (* (prefix-numeric-value arg)
6224 This puts the cursor where we want it.
6226 @node Small buffer case
6227 @unnumberedsubsubsec What happens in a small buffer
6229 If the buffer contains fewer than 10,000 characters, a slightly
6230 different computation is performed. You might think this is not
6231 necessary, since the first computation could do the job. However, in
6232 a small buffer, the first method may not put the cursor on exactly the
6233 desired line; the second method does a better job.
6236 The code looks like this:
6238 @c Keep this on one line.
6240 (/ (+ 10 (* size (prefix-numeric-value arg))) 10))
6245 This is code in which you figure out what happens by discovering how the
6246 functions are embedded in parentheses. It is easier to read if you
6247 reformat it with each expression indented more deeply than its
6248 enclosing expression:
6256 (prefix-numeric-value arg)))
6263 Looking at parentheses, we see that the innermost operation is
6264 @code{(prefix-numeric-value arg)}, which converts the raw argument to
6265 a number. In the following expression, this number is multiplied by
6266 the size of the accessible portion of the buffer:
6269 (* size (prefix-numeric-value arg))
6273 This multiplication creates a number that may be larger than the size of
6274 the buffer---seven times larger if the argument is 7, for example. Ten
6275 is then added to this number and finally the large number is divided by
6276 ten to provide a value that is one character larger than the percentage
6277 position in the buffer.
6279 The number that results from all this is passed to @code{goto-char} and
6280 the cursor is moved to that point.
6283 @node beginning-of-buffer complete
6284 @subsection The Complete @code{beginning-of-buffer}
6287 Here is the complete text of the @code{beginning-of-buffer} function:
6293 (defun beginning-of-buffer (&optional arg)
6294 "Move point to the beginning of the buffer;
6295 leave mark at previous position.
6296 With \\[universal-argument] prefix,
6297 do not set mark at previous position.
6299 put point N/10 of the way from the beginning.
6301 If the buffer is narrowed,
6302 this command uses the beginning and size
6303 of the accessible part of the buffer.
6307 Don't use this command in Lisp programs!
6308 \(goto-char (point-min)) is faster
6309 and avoids clobbering the mark."
6312 (and transient-mark-mode mark-active)
6316 (let ((size (- (point-max) (point-min))))
6317 (goto-char (if (and arg (not (consp arg)))
6320 ;; Avoid overflow for large buffer sizes!
6321 (* (prefix-numeric-value arg)
6323 (/ (+ 10 (* size (prefix-numeric-value arg)))
6326 (if arg (forward-line 1)))
6331 From before GNU Emacs 22
6334 (defun beginning-of-buffer (&optional arg)
6335 "Move point to the beginning of the buffer;
6336 leave mark at previous position.
6337 With arg N, put point N/10 of the way
6338 from the true beginning.
6341 Don't use this in Lisp programs!
6342 \(goto-char (point-min)) is faster
6343 and does not set the mark."
6350 (if (> (buffer-size) 10000)
6351 ;; @r{Avoid overflow for large buffer sizes!}
6352 (* (prefix-numeric-value arg)
6353 (/ (buffer-size) 10))
6356 (/ (+ 10 (* (buffer-size)
6357 (prefix-numeric-value arg)))
6360 (if arg (forward-line 1)))
6366 Except for two small points, the previous discussion shows how this
6367 function works. The first point deals with a detail in the
6368 documentation string, and the second point concerns the last line of
6372 In the documentation string, there is reference to an expression:
6375 \\[universal-argument]
6379 A @samp{\\} is used before the first square bracket of this
6380 expression. This @samp{\\} tells the Lisp interpreter to substitute
6381 whatever key is currently bound to the @samp{[@dots{}]}. In the case
6382 of @code{universal-argument}, that is usually @kbd{C-u}, but it might
6383 be different. (@xref{Documentation Tips, , Tips for Documentation
6384 Strings, elisp, The GNU Emacs Lisp Reference Manual}, for more
6388 Finally, the last line of the @code{beginning-of-buffer} command says
6389 to move point to the beginning of the next line if the command is
6390 invoked with an argument:
6393 (if arg (forward-line 1)))
6397 This puts the cursor at the beginning of the first line after the
6398 appropriate tenths position in the buffer. This is a flourish that
6399 means that the cursor is always located @emph{at least} the requested
6400 tenths of the way through the buffer, which is a nicety that is,
6401 perhaps, not necessary, but which, if it did not occur, would be sure
6404 On the other hand, it also means that if you specify the command with
6405 a @kbd{C-u}, but without a number, that is to say, if the `raw prefix
6406 argument' is simply a cons cell, then the command puts you at the
6407 beginning of the second line @dots{} I don't know whether this is
6408 intended or whether no one has dealt with the code to avoid this
6411 @node Second Buffer Related Review
6414 Here is a brief summary of some of the topics covered in this chapter.
6418 Evaluate each argument in sequence, and return the value of the first
6419 argument that is not @code{nil}; if none return a value that is not
6420 @code{nil}, return @code{nil}. In brief, return the first true value
6421 of the arguments; return a true value if one @emph{or} any of the
6425 Evaluate each argument in sequence, and if any are @code{nil}, return
6426 @code{nil}; if none are @code{nil}, return the value of the last
6427 argument. In brief, return a true value only if all the arguments are
6428 true; return a true value if one @emph{and} each of the others is
6432 A keyword used to indicate that an argument to a function definition
6433 is optional; this means that the function can be evaluated without the
6434 argument, if desired.
6436 @item prefix-numeric-value
6437 Convert the `raw prefix argument' produced by @code{(interactive
6438 "P")} to a numeric value.
6441 Move point forward to the beginning of the next line, or if the argument
6442 is greater than one, forward that many lines. If it can't move as far
6443 forward as it is supposed to, @code{forward-line} goes forward as far as
6444 it can and then returns a count of the number of additional lines it was
6445 supposed to move but couldn't.
6448 Delete the entire contents of the current buffer.
6451 Return @code{t} if its argument is a buffer; otherwise return @code{nil}.
6454 @node optional Exercise
6455 @section @code{optional} Argument Exercise
6457 Write an interactive function with an optional argument that tests
6458 whether its argument, a number, is greater than or equal to, or else,
6459 less than the value of @code{fill-column}, and tells you which, in a
6460 message. However, if you do not pass an argument to the function, use
6461 56 as a default value.
6463 @node Narrowing & Widening
6464 @chapter Narrowing and Widening
6465 @cindex Focusing attention (narrowing)
6469 Narrowing is a feature of Emacs that makes it possible for you to focus
6470 on a specific part of a buffer, and work without accidentally changing
6471 other parts. Narrowing is normally disabled since it can confuse
6475 * Narrowing advantages:: The advantages of narrowing
6476 * save-restriction:: The @code{save-restriction} special form.
6477 * what-line:: The number of the line that point is on.
6482 @node Narrowing advantages
6483 @unnumberedsec The Advantages of Narrowing
6486 With narrowing, the rest of a buffer is made invisible, as if it weren't
6487 there. This is an advantage if, for example, you want to replace a word
6488 in one part of a buffer but not in another: you narrow to the part you want
6489 and the replacement is carried out only in that section, not in the rest
6490 of the buffer. Searches will only work within a narrowed region, not
6491 outside of one, so if you are fixing a part of a document, you can keep
6492 yourself from accidentally finding parts you do not need to fix by
6493 narrowing just to the region you want.
6494 (The key binding for @code{narrow-to-region} is @kbd{C-x n n}.)
6496 However, narrowing does make the rest of the buffer invisible, which
6497 can scare people who inadvertently invoke narrowing and think they
6498 have deleted a part of their file. Moreover, the @code{undo} command
6499 (which is usually bound to @kbd{C-x u}) does not turn off narrowing
6500 (nor should it), so people can become quite desperate if they do not
6501 know that they can return the rest of a buffer to visibility with the
6502 @code{widen} command.
6503 (The key binding for @code{widen} is @kbd{C-x n w}.)
6505 Narrowing is just as useful to the Lisp interpreter as to a human.
6506 Often, an Emacs Lisp function is designed to work on just part of a
6507 buffer; or conversely, an Emacs Lisp function needs to work on all of a
6508 buffer that has been narrowed. The @code{what-line} function, for
6509 example, removes the narrowing from a buffer, if it has any narrowing
6510 and when it has finished its job, restores the narrowing to what it was.
6511 On the other hand, the @code{count-lines} function
6512 uses narrowing to restrict itself to just that portion
6513 of the buffer in which it is interested and then restores the previous
6516 @node save-restriction
6517 @section The @code{save-restriction} Special Form
6518 @findex save-restriction
6520 In Emacs Lisp, you can use the @code{save-restriction} special form to
6521 keep track of whatever narrowing is in effect, if any. When the Lisp
6522 interpreter meets with @code{save-restriction}, it executes the code
6523 in the body of the @code{save-restriction} expression, and then undoes
6524 any changes to narrowing that the code caused. If, for example, the
6525 buffer is narrowed and the code that follows @code{save-restriction}
6526 gets rid of the narrowing, @code{save-restriction} returns the buffer
6527 to its narrowed region afterwards. In the @code{what-line} command,
6528 any narrowing the buffer may have is undone by the @code{widen}
6529 command that immediately follows the @code{save-restriction} command.
6530 Any original narrowing is restored just before the completion of the
6534 The template for a @code{save-restriction} expression is simple:
6544 The body of the @code{save-restriction} is one or more expressions that
6545 will be evaluated in sequence by the Lisp interpreter.
6547 Finally, a point to note: when you use both @code{save-excursion} and
6548 @code{save-restriction}, one right after the other, you should use
6549 @code{save-excursion} outermost. If you write them in reverse order,
6550 you may fail to record narrowing in the buffer to which Emacs switches
6551 after calling @code{save-excursion}. Thus, when written together,
6552 @code{save-excursion} and @code{save-restriction} should be written
6563 In other circumstances, when not written together, the
6564 @code{save-excursion} and @code{save-restriction} special forms must
6565 be written in the order appropriate to the function.
6581 /usr/local/src/emacs/lisp/simple.el
6584 "Print the current buffer line number and narrowed line number of point."
6586 (let ((start (point-min))
6587 (n (line-number-at-pos)))
6589 (message "Line %d" n)
6593 (message "line %d (narrowed line %d)"
6594 (+ n (line-number-at-pos start) -1) n))))))
6596 (defun line-number-at-pos (&optional pos)
6597 "Return (narrowed) buffer line number at position POS.
6598 If POS is nil, use current buffer location.
6599 Counting starts at (point-min), so the value refers
6600 to the contents of the accessible portion of the buffer."
6601 (let ((opoint (or pos (point))) start)
6603 (goto-char (point-min))
6604 (setq start (point))
6607 (1+ (count-lines start (point))))))
6609 (defun count-lines (start end)
6610 "Return number of lines between START and END.
6611 This is usually the number of newlines between them,
6612 but can be one more if START is not equal to END
6613 and the greater of them is not at the start of a line."
6616 (narrow-to-region start end)
6617 (goto-char (point-min))
6618 (if (eq selective-display t)
6621 (while (re-search-forward "[\n\C-m]" nil t 40)
6622 (setq done (+ 40 done)))
6623 (while (re-search-forward "[\n\C-m]" nil t 1)
6624 (setq done (+ 1 done)))
6625 (goto-char (point-max))
6626 (if (and (/= start end)
6630 (- (buffer-size) (forward-line (buffer-size)))))))
6634 @section @code{what-line}
6636 @cindex Widening, example of
6638 The @code{what-line} command tells you the number of the line in which
6639 the cursor is located. The function illustrates the use of the
6640 @code{save-restriction} and @code{save-excursion} commands. Here is the
6641 original text of the function:
6646 "Print the current line number (in the buffer) of point."
6653 (1+ (count-lines 1 (point)))))))
6657 (In recent versions of GNU Emacs, the @code{what-line} function has
6658 been expanded to tell you your line number in a narrowed buffer as
6659 well as your line number in a widened buffer. The recent version is
6660 more complex than the version shown here. If you feel adventurous,
6661 you might want to look at it after figuring out how this version
6662 works. You will probably need to use @kbd{C-h f}
6663 (@code{describe-function}). The newer version uses a conditional to
6664 determine whether the buffer has been narrowed.
6666 (Also, it uses @code{line-number-at-pos}, which among other simple
6667 expressions, such as @code{(goto-char (point-min))}, moves point to
6668 the beginning of the current line with @code{(forward-line 0)} rather
6669 than @code{beginning-of-line}.)
6671 The @code{what-line} function as shown here has a documentation line
6672 and is interactive, as you would expect. The next two lines use the
6673 functions @code{save-restriction} and @code{widen}.
6675 The @code{save-restriction} special form notes whatever narrowing is in
6676 effect, if any, in the current buffer and restores that narrowing after
6677 the code in the body of the @code{save-restriction} has been evaluated.
6679 The @code{save-restriction} special form is followed by @code{widen}.
6680 This function undoes any narrowing the current buffer may have had
6681 when @code{what-line} was called. (The narrowing that was there is
6682 the narrowing that @code{save-restriction} remembers.) This widening
6683 makes it possible for the line counting commands to count from the
6684 beginning of the buffer. Otherwise, they would have been limited to
6685 counting within the accessible region. Any original narrowing is
6686 restored just before the completion of the function by the
6687 @code{save-restriction} special form.
6689 The call to @code{widen} is followed by @code{save-excursion}, which
6690 saves the location of the cursor (i.e., of point) and of the mark, and
6691 restores them after the code in the body of the @code{save-excursion}
6692 uses the @code{beginning-of-line} function to move point.
6694 (Note that the @code{(widen)} expression comes between the
6695 @code{save-restriction} and @code{save-excursion} special forms. When
6696 you write the two @code{save- @dots{}} expressions in sequence, write
6697 @code{save-excursion} outermost.)
6700 The last two lines of the @code{what-line} function are functions to
6701 count the number of lines in the buffer and then print the number in the
6707 (1+ (count-lines 1 (point)))))))
6711 The @code{message} function prints a one-line message at the bottom of
6712 the Emacs screen. The first argument is inside of quotation marks and
6713 is printed as a string of characters. However, it may contain a
6714 @samp{%d} expression to print a following argument. @samp{%d} prints
6715 the argument as a decimal, so the message will say something such as
6719 The number that is printed in place of the @samp{%d} is computed by the
6720 last line of the function:
6723 (1+ (count-lines 1 (point)))
6729 (defun count-lines (start end)
6730 "Return number of lines between START and END.
6731 This is usually the number of newlines between them,
6732 but can be one more if START is not equal to END
6733 and the greater of them is not at the start of a line."
6736 (narrow-to-region start end)
6737 (goto-char (point-min))
6738 (if (eq selective-display t)
6741 (while (re-search-forward "[\n\C-m]" nil t 40)
6742 (setq done (+ 40 done)))
6743 (while (re-search-forward "[\n\C-m]" nil t 1)
6744 (setq done (+ 1 done)))
6745 (goto-char (point-max))
6746 (if (and (/= start end)
6750 (- (buffer-size) (forward-line (buffer-size)))))))
6754 What this does is count the lines from the first position of the
6755 buffer, indicated by the @code{1}, up to @code{(point)}, and then add
6756 one to that number. (The @code{1+} function adds one to its
6757 argument.) We add one to it because line 2 has only one line before
6758 it, and @code{count-lines} counts only the lines @emph{before} the
6761 After @code{count-lines} has done its job, and the message has been
6762 printed in the echo area, the @code{save-excursion} restores point and
6763 mark to their original positions; and @code{save-restriction} restores
6764 the original narrowing, if any.
6766 @node narrow Exercise
6767 @section Exercise with Narrowing
6769 Write a function that will display the first 60 characters of the
6770 current buffer, even if you have narrowed the buffer to its latter
6771 half so that the first line is inaccessible. Restore point, mark, and
6772 narrowing. For this exercise, you need to use a whole potpourri of
6773 functions, including @code{save-restriction}, @code{widen},
6774 @code{goto-char}, @code{point-min}, @code{message}, and
6775 @code{buffer-substring}.
6777 @cindex Properties, mention of @code{buffer-substring-no-properties}
6778 (@code{buffer-substring} is a previously unmentioned function you will
6779 have to investigate yourself; or perhaps you will have to use
6780 @code{buffer-substring-no-properties} or
6781 @code{filter-buffer-substring} @dots{}, yet other functions. Text
6782 properties are a feature otherwise not discussed here. @xref{Text
6783 Properties, , Text Properties, elisp, The GNU Emacs Lisp Reference
6786 Additionally, do you really need @code{goto-char} or @code{point-min}?
6787 Or can you write the function without them?
6789 @node car cdr & cons
6790 @chapter @code{car}, @code{cdr}, @code{cons}: Fundamental Functions
6791 @findex car, @r{introduced}
6792 @findex cdr, @r{introduced}
6794 In Lisp, @code{car}, @code{cdr}, and @code{cons} are fundamental
6795 functions. The @code{cons} function is used to construct lists, and
6796 the @code{car} and @code{cdr} functions are used to take them apart.
6798 In the walk through of the @code{copy-region-as-kill} function, we
6799 will see @code{cons} as well as two variants on @code{cdr},
6800 namely, @code{setcdr} and @code{nthcdr}. (@xref{copy-region-as-kill}.)
6803 * Strange Names:: An historical aside: why the strange names?
6804 * car & cdr:: Functions for extracting part of a list.
6805 * cons:: Constructing a list.
6806 * nthcdr:: Calling @code{cdr} repeatedly.
6808 * setcar:: Changing the first element of a list.
6809 * setcdr:: Changing the rest of a list.
6815 @unnumberedsec Strange Names
6818 The name of the @code{cons} function is not unreasonable: it is an
6819 abbreviation of the word `construct'. The origins of the names for
6820 @code{car} and @code{cdr}, on the other hand, are esoteric: @code{car}
6821 is an acronym from the phrase `Contents of the Address part of the
6822 Register'; and @code{cdr} (pronounced `could-er') is an acronym from
6823 the phrase `Contents of the Decrement part of the Register'. These
6824 phrases refer to specific pieces of hardware on the very early
6825 computer on which the original Lisp was developed. Besides being
6826 obsolete, the phrases have been completely irrelevant for more than 25
6827 years to anyone thinking about Lisp. Nonetheless, although a few
6828 brave scholars have begun to use more reasonable names for these
6829 functions, the old terms are still in use. In particular, since the
6830 terms are used in the Emacs Lisp source code, we will use them in this
6834 @section @code{car} and @code{cdr}
6836 The @sc{car} of a list is, quite simply, the first item in the list.
6837 Thus the @sc{car} of the list @code{(rose violet daisy buttercup)} is
6841 If you are reading this in Info in GNU Emacs, you can see this by
6842 evaluating the following:
6845 (car '(rose violet daisy buttercup))
6849 After evaluating the expression, @code{rose} will appear in the echo
6852 Clearly, a more reasonable name for the @code{car} function would be
6853 @code{first} and this is often suggested.
6855 @code{car} does not remove the first item from the list; it only reports
6856 what it is. After @code{car} has been applied to a list, the list is
6857 still the same as it was. In the jargon, @code{car} is
6858 `non-destructive'. This feature turns out to be important.
6860 The @sc{cdr} of a list is the rest of the list, that is, the
6861 @code{cdr} function returns the part of the list that follows the
6862 first item. Thus, while the @sc{car} of the list @code{'(rose violet
6863 daisy buttercup)} is @code{rose}, the rest of the list, the value
6864 returned by the @code{cdr} function, is @code{(violet daisy
6868 You can see this by evaluating the following in the usual way:
6871 (cdr '(rose violet daisy buttercup))
6875 When you evaluate this, @code{(violet daisy buttercup)} will appear in
6878 Like @code{car}, @code{cdr} does not remove any elements from the
6879 list---it just returns a report of what the second and subsequent
6882 Incidentally, in the example, the list of flowers is quoted. If it were
6883 not, the Lisp interpreter would try to evaluate the list by calling
6884 @code{rose} as a function. In this example, we do not want to do that.
6886 Clearly, a more reasonable name for @code{cdr} would be @code{rest}.
6888 (There is a lesson here: when you name new functions, consider very
6889 carefully what you are doing, since you may be stuck with the names
6890 for far longer than you expect. The reason this document perpetuates
6891 these names is that the Emacs Lisp source code uses them, and if I did
6892 not use them, you would have a hard time reading the code; but do,
6893 please, try to avoid using these terms yourself. The people who come
6894 after you will be grateful to you.)
6896 When @code{car} and @code{cdr} are applied to a list made up of symbols,
6897 such as the list @code{(pine fir oak maple)}, the element of the list
6898 returned by the function @code{car} is the symbol @code{pine} without
6899 any parentheses around it. @code{pine} is the first element in the
6900 list. However, the @sc{cdr} of the list is a list itself, @code{(fir
6901 oak maple)}, as you can see by evaluating the following expressions in
6906 (car '(pine fir oak maple))
6908 (cdr '(pine fir oak maple))
6912 On the other hand, in a list of lists, the first element is itself a
6913 list. @code{car} returns this first element as a list. For example,
6914 the following list contains three sub-lists, a list of carnivores, a
6915 list of herbivores and a list of sea mammals:
6919 (car '((lion tiger cheetah)
6920 (gazelle antelope zebra)
6921 (whale dolphin seal)))
6926 In this example, the first element or @sc{car} of the list is the list of
6927 carnivores, @code{(lion tiger cheetah)}, and the rest of the list is
6928 @code{((gazelle antelope zebra) (whale dolphin seal))}.
6932 (cdr '((lion tiger cheetah)
6933 (gazelle antelope zebra)
6934 (whale dolphin seal)))
6938 It is worth saying again that @code{car} and @code{cdr} are
6939 non-destructive---that is, they do not modify or change lists to which
6940 they are applied. This is very important for how they are used.
6942 Also, in the first chapter, in the discussion about atoms, I said that
6943 in Lisp, ``certain kinds of atom, such as an array, can be separated
6944 into parts; but the mechanism for doing this is different from the
6945 mechanism for splitting a list. As far as Lisp is concerned, the
6946 atoms of a list are unsplittable.'' (@xref{Lisp Atoms}.) The
6947 @code{car} and @code{cdr} functions are used for splitting lists and
6948 are considered fundamental to Lisp. Since they cannot split or gain
6949 access to the parts of an array, an array is considered an atom.
6950 Conversely, the other fundamental function, @code{cons}, can put
6951 together or construct a list, but not an array. (Arrays are handled
6952 by array-specific functions. @xref{Arrays, , Arrays, elisp, The GNU
6953 Emacs Lisp Reference Manual}.)
6956 @section @code{cons}
6957 @findex cons, @r{introduced}
6959 The @code{cons} function constructs lists; it is the inverse of
6960 @code{car} and @code{cdr}. For example, @code{cons} can be used to make
6961 a four element list from the three element list, @code{(fir oak maple)}:
6964 (cons 'pine '(fir oak maple))
6969 After evaluating this list, you will see
6972 (pine fir oak maple)
6976 appear in the echo area. @code{cons} causes the creation of a new
6977 list in which the element is followed by the elements of the original
6980 We often say that `@code{cons} puts a new element at the beginning of
6981 a list; it attaches or pushes elements onto the list', but this
6982 phrasing can be misleading, since @code{cons} does not change an
6983 existing list, but creates a new one.
6985 Like @code{car} and @code{cdr}, @code{cons} is non-destructive.
6989 * length:: How to find the length of a list.
6994 @unnumberedsubsec Build a list
6997 @code{cons} must have a list to attach to.@footnote{Actually, you can
6998 @code{cons} an element to an atom to produce a dotted pair. Dotted
6999 pairs are not discussed here; see @ref{Dotted Pair Notation, , Dotted
7000 Pair Notation, elisp, The GNU Emacs Lisp Reference Manual}.} You
7001 cannot start from absolutely nothing. If you are building a list, you
7002 need to provide at least an empty list at the beginning. Here is a
7003 series of @code{cons} expressions that build up a list of flowers. If
7004 you are reading this in Info in GNU Emacs, you can evaluate each of
7005 the expressions in the usual way; the value is printed in this text
7006 after @samp{@result{}}, which you may read as `evaluates to'.
7010 (cons 'buttercup ())
7011 @result{} (buttercup)
7015 (cons 'daisy '(buttercup))
7016 @result{} (daisy buttercup)
7020 (cons 'violet '(daisy buttercup))
7021 @result{} (violet daisy buttercup)
7025 (cons 'rose '(violet daisy buttercup))
7026 @result{} (rose violet daisy buttercup)
7031 In the first example, the empty list is shown as @code{()} and a list
7032 made up of @code{buttercup} followed by the empty list is constructed.
7033 As you can see, the empty list is not shown in the list that was
7034 constructed. All that you see is @code{(buttercup)}. The empty list is
7035 not counted as an element of a list because there is nothing in an empty
7036 list. Generally speaking, an empty list is invisible.
7038 The second example, @code{(cons 'daisy '(buttercup))} constructs a new,
7039 two element list by putting @code{daisy} in front of @code{buttercup};
7040 and the third example constructs a three element list by putting
7041 @code{violet} in front of @code{daisy} and @code{buttercup}.
7044 @subsection Find the Length of a List: @code{length}
7047 You can find out how many elements there are in a list by using the Lisp
7048 function @code{length}, as in the following examples:
7052 (length '(buttercup))
7057 (length '(daisy buttercup))
7062 (length (cons 'violet '(daisy buttercup)))
7068 In the third example, the @code{cons} function is used to construct a
7069 three element list which is then passed to the @code{length} function as
7073 We can also use @code{length} to count the number of elements in an
7084 As you would expect, the number of elements in an empty list is zero.
7086 An interesting experiment is to find out what happens if you try to find
7087 the length of no list at all; that is, if you try to call @code{length}
7088 without giving it an argument, not even an empty list:
7096 What you see, if you evaluate this, is the error message
7099 Lisp error: (wrong-number-of-arguments length 0)
7103 This means that the function receives the wrong number of
7104 arguments, zero, when it expects some other number of arguments. In
7105 this case, one argument is expected, the argument being a list whose
7106 length the function is measuring. (Note that @emph{one} list is
7107 @emph{one} argument, even if the list has many elements inside it.)
7109 The part of the error message that says @samp{length} is the name of
7113 @code{length} is still a subroutine, but you need C-h f to discover that.
7115 In an earlier version:
7116 This is written with a special notation, @samp{#<subr},
7117 that indicates that the function @code{length} is one of the primitive
7118 functions written in C rather than in Emacs Lisp. (@samp{subr} is an
7119 abbreviation for `subroutine'.) @xref{What Is a Function, , What Is a
7120 Function?, elisp , The GNU Emacs Lisp Reference Manual}, for more
7125 @section @code{nthcdr}
7128 The @code{nthcdr} function is associated with the @code{cdr} function.
7129 What it does is take the @sc{cdr} of a list repeatedly.
7131 If you take the @sc{cdr} of the list @code{(pine fir
7132 oak maple)}, you will be returned the list @code{(fir oak maple)}. If you
7133 repeat this on what was returned, you will be returned the list
7134 @code{(oak maple)}. (Of course, repeated @sc{cdr}ing on the original
7135 list will just give you the original @sc{cdr} since the function does
7136 not change the list. You need to evaluate the @sc{cdr} of the
7137 @sc{cdr} and so on.) If you continue this, eventually you will be
7138 returned an empty list, which in this case, instead of being shown as
7139 @code{()} is shown as @code{nil}.
7142 For review, here is a series of repeated @sc{cdr}s, the text following
7143 the @samp{@result{}} shows what is returned.
7147 (cdr '(pine fir oak maple))
7148 @result{}(fir oak maple)
7152 (cdr '(fir oak maple))
7153 @result{} (oak maple)
7178 You can also do several @sc{cdr}s without printing the values in
7183 (cdr (cdr '(pine fir oak maple)))
7184 @result{} (oak maple)
7189 In this example, the Lisp interpreter evaluates the innermost list first.
7190 The innermost list is quoted, so it just passes the list as it is to the
7191 innermost @code{cdr}. This @code{cdr} passes a list made up of the
7192 second and subsequent elements of the list to the outermost @code{cdr},
7193 which produces a list composed of the third and subsequent elements of
7194 the original list. In this example, the @code{cdr} function is repeated
7195 and returns a list that consists of the original list without its
7198 The @code{nthcdr} function does the same as repeating the call to
7199 @code{cdr}. In the following example, the argument 2 is passed to the
7200 function @code{nthcdr}, along with the list, and the value returned is
7201 the list without its first two items, which is exactly the same
7202 as repeating @code{cdr} twice on the list:
7206 (nthcdr 2 '(pine fir oak maple))
7207 @result{} (oak maple)
7212 Using the original four element list, we can see what happens when
7213 various numeric arguments are passed to @code{nthcdr}, including 0, 1,
7218 ;; @r{Leave the list as it was.}
7219 (nthcdr 0 '(pine fir oak maple))
7220 @result{} (pine fir oak maple)
7224 ;; @r{Return a copy without the first element.}
7225 (nthcdr 1 '(pine fir oak maple))
7226 @result{} (fir oak maple)
7230 ;; @r{Return a copy of the list without three elements.}
7231 (nthcdr 3 '(pine fir oak maple))
7236 ;; @r{Return a copy lacking all four elements.}
7237 (nthcdr 4 '(pine fir oak maple))
7242 ;; @r{Return a copy lacking all elements.}
7243 (nthcdr 5 '(pine fir oak maple))
7252 The @code{nthcdr} function takes the @sc{cdr} of a list repeatedly.
7253 The @code{nth} function takes the @sc{car} of the result returned by
7254 @code{nthcdr}. It returns the Nth element of the list.
7257 Thus, if it were not defined in C for speed, the definition of
7258 @code{nth} would be:
7263 "Returns the Nth element of LIST.
7264 N counts from zero. If LIST is not that long, nil is returned."
7265 (car (nthcdr n list)))
7270 (Originally, @code{nth} was defined in Emacs Lisp in @file{subr.el},
7271 but its definition was redone in C in the 1980s.)
7273 The @code{nth} function returns a single element of a list.
7274 This can be very convenient.
7276 Note that the elements are numbered from zero, not one. That is to
7277 say, the first element of a list, its @sc{car} is the zeroth element.
7278 This is called `zero-based' counting and often bothers people who
7279 are accustomed to the first element in a list being number one, which
7287 (nth 0 '("one" "two" "three"))
7290 (nth 1 '("one" "two" "three"))
7295 It is worth mentioning that @code{nth}, like @code{nthcdr} and
7296 @code{cdr}, does not change the original list---the function is
7297 non-destructive. This is in sharp contrast to the @code{setcar} and
7298 @code{setcdr} functions.
7301 @section @code{setcar}
7304 As you might guess from their names, the @code{setcar} and @code{setcdr}
7305 functions set the @sc{car} or the @sc{cdr} of a list to a new value.
7306 They actually change the original list, unlike @code{car} and @code{cdr}
7307 which leave the original list as it was. One way to find out how this
7308 works is to experiment. We will start with the @code{setcar} function.
7311 First, we can make a list and then set the value of a variable to the
7312 list, using the @code{setq} function. Here is a list of animals:
7315 (setq animals '(antelope giraffe lion tiger))
7319 If you are reading this in Info inside of GNU Emacs, you can evaluate
7320 this expression in the usual fashion, by positioning the cursor after
7321 the expression and typing @kbd{C-x C-e}. (I'm doing this right here
7322 as I write this. This is one of the advantages of having the
7323 interpreter built into the computing environment. Incidentally, when
7324 there is nothing on the line after the final parentheses, such as a
7325 comment, point can be on the next line. Thus, if your cursor is in
7326 the first column of the next line, you do not need to move it.
7327 Indeed, Emacs permits any amount of white space after the final
7331 When we evaluate the variable @code{animals}, we see that it is bound to
7332 the list @code{(antelope giraffe lion tiger)}:
7337 @result{} (antelope giraffe lion tiger)
7342 Put another way, the variable @code{animals} points to the list
7343 @code{(antelope giraffe lion tiger)}.
7345 Next, evaluate the function @code{setcar} while passing it two
7346 arguments, the variable @code{animals} and the quoted symbol
7347 @code{hippopotamus}; this is done by writing the three element list
7348 @code{(setcar animals 'hippopotamus)} and then evaluating it in the
7352 (setcar animals 'hippopotamus)
7357 After evaluating this expression, evaluate the variable @code{animals}
7358 again. You will see that the list of animals has changed:
7363 @result{} (hippopotamus giraffe lion tiger)
7368 The first element on the list, @code{antelope} is replaced by
7369 @code{hippopotamus}.
7371 So we can see that @code{setcar} did not add a new element to the list
7372 as @code{cons} would have; it replaced @code{antelope} with
7373 @code{hippopotamus}; it @emph{changed} the list.
7376 @section @code{setcdr}
7379 The @code{setcdr} function is similar to the @code{setcar} function,
7380 except that the function replaces the second and subsequent elements of
7381 a list rather than the first element.
7383 (To see how to change the last element of a list, look ahead to
7384 @ref{kill-new function, , The @code{kill-new} function}, which uses
7385 the @code{nthcdr} and @code{setcdr} functions.)
7388 To see how this works, set the value of the variable to a list of
7389 domesticated animals by evaluating the following expression:
7392 (setq domesticated-animals '(horse cow sheep goat))
7397 If you now evaluate the list, you will be returned the list
7398 @code{(horse cow sheep goat)}:
7402 domesticated-animals
7403 @result{} (horse cow sheep goat)
7408 Next, evaluate @code{setcdr} with two arguments, the name of the
7409 variable which has a list as its value, and the list to which the
7410 @sc{cdr} of the first list will be set;
7413 (setcdr domesticated-animals '(cat dog))
7417 If you evaluate this expression, the list @code{(cat dog)} will appear
7418 in the echo area. This is the value returned by the function. The
7419 result we are interested in is the ``side effect'', which we can see by
7420 evaluating the variable @code{domesticated-animals}:
7424 domesticated-animals
7425 @result{} (horse cat dog)
7430 Indeed, the list is changed from @code{(horse cow sheep goat)} to
7431 @code{(horse cat dog)}. The @sc{cdr} of the list is changed from
7432 @code{(cow sheep goat)} to @code{(cat dog)}.
7437 Construct a list of four birds by evaluating several expressions with
7438 @code{cons}. Find out what happens when you @code{cons} a list onto
7439 itself. Replace the first element of the list of four birds with a
7440 fish. Replace the rest of that list with a list of other fish.
7442 @node Cutting & Storing Text
7443 @chapter Cutting and Storing Text
7444 @cindex Cutting and storing text
7445 @cindex Storing and cutting text
7446 @cindex Killing text
7447 @cindex Clipping text
7448 @cindex Erasing text
7449 @cindex Deleting text
7451 Whenever you cut or clip text out of a buffer with a `kill' command in
7452 GNU Emacs, it is stored in a list and you can bring it back with a
7455 (The use of the word `kill' in Emacs for processes which specifically
7456 @emph{do not} destroy the values of the entities is an unfortunate
7457 historical accident. A much more appropriate word would be `clip' since
7458 that is what the kill commands do; they clip text out of a buffer and
7459 put it into storage from which it can be brought back. I have often
7460 been tempted to replace globally all occurrences of `kill' in the Emacs
7461 sources with `clip' and all occurrences of `killed' with `clipped'.)
7464 * Storing Text:: Text is stored in a list.
7465 * zap-to-char:: Cutting out text up to a character.
7466 * kill-region:: Cutting text out of a region.
7467 * copy-region-as-kill:: A definition for copying text.
7468 * Digression into C:: Minor note on C programming language macros.
7469 * defvar:: How to give a variable an initial value.
7470 * cons & search-fwd Review::
7471 * search Exercises::
7476 @unnumberedsec Storing Text in a List
7479 When text is cut out of a buffer, it is stored on a list. Successive
7480 pieces of text are stored on the list successively, so the list might
7484 ("a piece of text" "previous piece")
7489 The function @code{cons} can be used to create a new list from a piece
7490 of text (an `atom', to use the jargon) and an existing list, like
7495 (cons "another piece"
7496 '("a piece of text" "previous piece"))
7502 If you evaluate this expression, a list of three elements will appear in
7506 ("another piece" "a piece of text" "previous piece")
7509 With the @code{car} and @code{nthcdr} functions, you can retrieve
7510 whichever piece of text you want. For example, in the following code,
7511 @code{nthcdr 1 @dots{}} returns the list with the first item removed;
7512 and the @code{car} returns the first element of that remainder---the
7513 second element of the original list:
7517 (car (nthcdr 1 '("another piece"
7520 @result{} "a piece of text"
7524 The actual functions in Emacs are more complex than this, of course.
7525 The code for cutting and retrieving text has to be written so that
7526 Emacs can figure out which element in the list you want---the first,
7527 second, third, or whatever. In addition, when you get to the end of
7528 the list, Emacs should give you the first element of the list, rather
7529 than nothing at all.
7531 The list that holds the pieces of text is called the @dfn{kill ring}.
7532 This chapter leads up to a description of the kill ring and how it is
7533 used by first tracing how the @code{zap-to-char} function works. This
7534 function uses (or `calls') a function that invokes a function that
7535 manipulates the kill ring. Thus, before reaching the mountains, we
7536 climb the foothills.
7538 A subsequent chapter describes how text that is cut from the buffer is
7539 retrieved. @xref{Yanking, , Yanking Text Back}.
7542 @section @code{zap-to-char}
7545 Let us look at the interactive @code{zap-to-char} function.
7548 * Complete zap-to-char:: The complete implementation.
7549 * zap-to-char interactive:: A three part interactive expression.
7550 * zap-to-char body:: A short overview.
7551 * search-forward:: How to search for a string.
7552 * progn:: The @code{progn} special form.
7553 * Summing up zap-to-char:: Using @code{point} and @code{search-forward}.
7557 @node Complete zap-to-char
7558 @unnumberedsubsec The Complete @code{zap-to-char} Implementation
7561 The @code{zap-to-char} function removes the text in the region between
7562 the location of the cursor (i.e., of point) up to and including the
7563 next occurrence of a specified character. The text that
7564 @code{zap-to-char} removes is put in the kill ring; and it can be
7565 retrieved from the kill ring by typing @kbd{C-y} (@code{yank}). If
7566 the command is given an argument, it removes text through that number
7567 of occurrences. Thus, if the cursor were at the beginning of this
7568 sentence and the character were @samp{s}, @samp{Thus} would be
7569 removed. If the argument were two, @samp{Thus, if the curs} would be
7570 removed, up to and including the @samp{s} in @samp{cursor}.
7572 If the specified character is not found, @code{zap-to-char} will say
7573 ``Search failed'', tell you the character you typed, and not remove
7576 In order to determine how much text to remove, @code{zap-to-char} uses
7577 a search function. Searches are used extensively in code that
7578 manipulates text, and we will focus attention on them as well as on the
7582 @c GNU Emacs version 19
7583 (defun zap-to-char (arg char) ; version 19 implementation
7584 "Kill up to and including ARG'th occurrence of CHAR.
7585 Goes backward if ARG is negative; error if CHAR not found."
7586 (interactive "*p\ncZap to char: ")
7587 (kill-region (point)
7590 (char-to-string char) nil nil arg)
7595 Here is the complete text of the version 22 implementation of the function:
7600 (defun zap-to-char (arg char)
7601 "Kill up to and including ARG'th occurrence of CHAR.
7602 Case is ignored if `case-fold-search' is non-nil in the current buffer.
7603 Goes backward if ARG is negative; error if CHAR not found."
7604 (interactive "p\ncZap to char: ")
7605 (if (char-table-p translation-table-for-input)
7606 (setq char (or (aref translation-table-for-input char) char)))
7607 (kill-region (point) (progn
7608 (search-forward (char-to-string char)
7614 The documentation is thorough. You do need to know the jargon meaning
7617 @node zap-to-char interactive
7618 @subsection The @code{interactive} Expression
7621 The interactive expression in the @code{zap-to-char} command looks like
7625 (interactive "p\ncZap to char: ")
7628 The part within quotation marks, @code{"p\ncZap to char:@: "}, specifies
7629 two different things. First, and most simply, is the @samp{p}.
7630 This part is separated from the next part by a newline, @samp{\n}.
7631 The @samp{p} means that the first argument to the function will be
7632 passed the value of a `processed prefix'. The prefix argument is
7633 passed by typing @kbd{C-u} and a number, or @kbd{M-} and a number. If
7634 the function is called interactively without a prefix, 1 is passed to
7637 The second part of @code{"p\ncZap to char:@: "} is
7638 @samp{cZap to char:@: }. In this part, the lower case @samp{c}
7639 indicates that @code{interactive} expects a prompt and that the
7640 argument will be a character. The prompt follows the @samp{c} and is
7641 the string @samp{Zap to char:@: } (with a space after the colon to
7644 What all this does is prepare the arguments to @code{zap-to-char} so they
7645 are of the right type, and give the user a prompt.
7647 In a read-only buffer, the @code{zap-to-char} function copies the text
7648 to the kill ring, but does not remove it. The echo area displays a
7649 message saying that the buffer is read-only. Also, the terminal may
7650 beep or blink at you.
7652 @node zap-to-char body
7653 @subsection The Body of @code{zap-to-char}
7655 The body of the @code{zap-to-char} function contains the code that
7656 kills (that is, removes) the text in the region from the current
7657 position of the cursor up to and including the specified character.
7659 The first part of the code looks like this:
7662 (if (char-table-p translation-table-for-input)
7663 (setq char (or (aref translation-table-for-input char) char)))
7664 (kill-region (point) (progn
7665 (search-forward (char-to-string char) nil nil arg)
7670 @code{char-table-p} is an hitherto unseen function. It determines
7671 whether its argument is a character table. When it is, it sets the
7672 character passed to @code{zap-to-char} to one of them, if that
7673 character exists, or to the character itself. (This becomes important
7674 for certain characters in non-European languages. The @code{aref}
7675 function extracts an element from an array. It is an array-specific
7676 function that is not described in this document. @xref{Arrays, ,
7677 Arrays, elisp, The GNU Emacs Lisp Reference Manual}.)
7680 @code{(point)} is the current position of the cursor.
7682 The next part of the code is an expression using @code{progn}. The body
7683 of the @code{progn} consists of calls to @code{search-forward} and
7686 It is easier to understand how @code{progn} works after learning about
7687 @code{search-forward}, so we will look at @code{search-forward} and
7688 then at @code{progn}.
7690 @node search-forward
7691 @subsection The @code{search-forward} Function
7692 @findex search-forward
7694 The @code{search-forward} function is used to locate the
7695 zapped-for-character in @code{zap-to-char}. If the search is
7696 successful, @code{search-forward} leaves point immediately after the
7697 last character in the target string. (In @code{zap-to-char}, the
7698 target string is just one character long. @code{zap-to-char} uses the
7699 function @code{char-to-string} to ensure that the computer treats that
7700 character as a string.) If the search is backwards,
7701 @code{search-forward} leaves point just before the first character in
7702 the target. Also, @code{search-forward} returns @code{t} for true.
7703 (Moving point is therefore a `side effect'.)
7706 In @code{zap-to-char}, the @code{search-forward} function looks like this:
7709 (search-forward (char-to-string char) nil nil arg)
7712 The @code{search-forward} function takes four arguments:
7716 The first argument is the target, what is searched for. This must be a
7717 string, such as @samp{"z"}.
7719 As it happens, the argument passed to @code{zap-to-char} is a single
7720 character. Because of the way computers are built, the Lisp
7721 interpreter may treat a single character as being different from a
7722 string of characters. Inside the computer, a single character has a
7723 different electronic format than a string of one character. (A single
7724 character can often be recorded in the computer using exactly one
7725 byte; but a string may be longer, and the computer needs to be ready
7726 for this.) Since the @code{search-forward} function searches for a
7727 string, the character that the @code{zap-to-char} function receives as
7728 its argument must be converted inside the computer from one format to
7729 the other; otherwise the @code{search-forward} function will fail.
7730 The @code{char-to-string} function is used to make this conversion.
7733 The second argument bounds the search; it is specified as a position in
7734 the buffer. In this case, the search can go to the end of the buffer,
7735 so no bound is set and the second argument is @code{nil}.
7738 The third argument tells the function what it should do if the search
7739 fails---it can signal an error (and print a message) or it can return
7740 @code{nil}. A @code{nil} as the third argument causes the function to
7741 signal an error when the search fails.
7744 The fourth argument to @code{search-forward} is the repeat count---how
7745 many occurrences of the string to look for. This argument is optional
7746 and if the function is called without a repeat count, this argument is
7747 passed the value 1. If this argument is negative, the search goes
7752 In template form, a @code{search-forward} expression looks like this:
7756 (search-forward "@var{target-string}"
7757 @var{limit-of-search}
7758 @var{what-to-do-if-search-fails}
7763 We will look at @code{progn} next.
7766 @subsection The @code{progn} Special Form
7769 @code{progn} is a special form that causes each of its arguments to be
7770 evaluated in sequence and then returns the value of the last one. The
7771 preceding expressions are evaluated only for the side effects they
7772 perform. The values produced by them are discarded.
7775 The template for a @code{progn} expression is very simple:
7784 In @code{zap-to-char}, the @code{progn} expression has to do two things:
7785 put point in exactly the right position; and return the location of
7786 point so that @code{kill-region} will know how far to kill to.
7788 The first argument to the @code{progn} is @code{search-forward}. When
7789 @code{search-forward} finds the string, the function leaves point
7790 immediately after the last character in the target string. (In this
7791 case the target string is just one character long.) If the search is
7792 backwards, @code{search-forward} leaves point just before the first
7793 character in the target. The movement of point is a side effect.
7795 The second and last argument to @code{progn} is the expression
7796 @code{(point)}. This expression returns the value of point, which in
7797 this case will be the location to which it has been moved by
7798 @code{search-forward}. (In the source, a line that tells the function
7799 to go to the previous character, if it is going forward, was commented
7800 out in 1999; I don't remember whether that feature or mis-feature was
7801 ever a part of the distributed source.) The value of @code{point} is
7802 returned by the @code{progn} expression and is passed to
7803 @code{kill-region} as @code{kill-region}'s second argument.
7805 @node Summing up zap-to-char
7806 @subsection Summing up @code{zap-to-char}
7808 Now that we have seen how @code{search-forward} and @code{progn} work,
7809 we can see how the @code{zap-to-char} function works as a whole.
7811 The first argument to @code{kill-region} is the position of the cursor
7812 when the @code{zap-to-char} command is given---the value of point at
7813 that time. Within the @code{progn}, the search function then moves
7814 point to just after the zapped-to-character and @code{point} returns the
7815 value of this location. The @code{kill-region} function puts together
7816 these two values of point, the first one as the beginning of the region
7817 and the second one as the end of the region, and removes the region.
7819 The @code{progn} special form is necessary because the
7820 @code{kill-region} command takes two arguments; and it would fail if
7821 @code{search-forward} and @code{point} expressions were written in
7822 sequence as two additional arguments. The @code{progn} expression is
7823 a single argument to @code{kill-region} and returns the one value that
7824 @code{kill-region} needs for its second argument.
7827 @section @code{kill-region}
7830 The @code{zap-to-char} function uses the @code{kill-region} function.
7831 This function clips text from a region and copies that text to
7832 the kill ring, from which it may be retrieved.
7837 (defun kill-region (beg end &optional yank-handler)
7838 "Kill (\"cut\") text between point and mark.
7839 This deletes the text from the buffer and saves it in the kill ring.
7840 The command \\[yank] can retrieve it from there.
7841 \(If you want to kill and then yank immediately, use \\[kill-ring-save].)
7843 If you want to append the killed region to the last killed text,
7844 use \\[append-next-kill] before \\[kill-region].
7846 If the buffer is read-only, Emacs will beep and refrain from deleting
7847 the text, but put the text in the kill ring anyway. This means that
7848 you can use the killing commands to copy text from a read-only buffer.
7850 This is the primitive for programs to kill text (as opposed to deleting it).
7851 Supply two arguments, character positions indicating the stretch of text
7853 Any command that calls this function is a \"kill command\".
7854 If the previous command was also a kill command,
7855 the text killed this time appends to the text killed last time
7856 to make one entry in the kill ring.
7858 In Lisp code, optional third arg YANK-HANDLER, if non-nil,
7859 specifies the yank-handler text property to be set on the killed
7860 text. See `insert-for-yank'."
7861 ;; Pass point first, then mark, because the order matters
7862 ;; when calling kill-append.
7863 (interactive (list (point) (mark)))
7864 (unless (and beg end)
7865 (error "The mark is not set now, so there is no region"))
7867 (let ((string (filter-buffer-substring beg end t)))
7868 (when string ;STRING is nil if BEG = END
7869 ;; Add that string to the kill ring, one way or another.
7870 (if (eq last-command 'kill-region)
7871 (kill-append string (< end beg) yank-handler)
7872 (kill-new string nil yank-handler)))
7873 (when (or string (eq last-command 'kill-region))
7874 (setq this-command 'kill-region))
7876 ((buffer-read-only text-read-only)
7877 ;; The code above failed because the buffer, or some of the characters
7878 ;; in the region, are read-only.
7879 ;; We should beep, in case the user just isn't aware of this.
7880 ;; However, there's no harm in putting
7881 ;; the region's text in the kill ring, anyway.
7882 (copy-region-as-kill beg end)
7883 ;; Set this-command now, so it will be set even if we get an error.
7884 (setq this-command 'kill-region)
7885 ;; This should barf, if appropriate, and give us the correct error.
7886 (if kill-read-only-ok
7887 (progn (message "Read only text copied to kill ring") nil)
7888 ;; Signal an error if the buffer is read-only.
7889 (barf-if-buffer-read-only)
7890 ;; If the buffer isn't read-only, the text is.
7891 (signal 'text-read-only (list (current-buffer)))))))
7894 The Emacs 22 version of that function uses @code{condition-case} and
7895 @code{copy-region-as-kill}, both of which we will explain.
7896 @code{condition-case} is an important special form.
7898 In essence, the @code{kill-region} function calls
7899 @code{condition-case}, which takes three arguments. In this function,
7900 the first argument does nothing. The second argument contains the
7901 code that does the work when all goes well. The third argument
7902 contains the code that is called in the event of an error.
7905 * Complete kill-region:: The function definition.
7906 * condition-case:: Dealing with a problem.
7911 @node Complete kill-region
7912 @unnumberedsubsec The Complete @code{kill-region} Definition
7916 We will go through the @code{condition-case} code in a moment. First,
7917 let us look at the definition of @code{kill-region}, with comments
7923 (defun kill-region (beg end)
7924 "Kill (\"cut\") text between point and mark.
7925 This deletes the text from the buffer and saves it in the kill ring.
7926 The command \\[yank] can retrieve it from there. @dots{} "
7930 ;; @bullet{} Since order matters, pass point first.
7931 (interactive (list (point) (mark)))
7932 ;; @bullet{} And tell us if we cannot cut the text.
7933 ;; `unless' is an `if' without a then-part.
7934 (unless (and beg end)
7935 (error "The mark is not set now, so there is no region"))
7939 ;; @bullet{} `condition-case' takes three arguments.
7940 ;; If the first argument is nil, as it is here,
7941 ;; information about the error signal is not
7942 ;; stored for use by another function.
7947 ;; @bullet{} The second argument to `condition-case' tells the
7948 ;; Lisp interpreter what to do when all goes well.
7952 ;; It starts with a `let' function that extracts the string
7953 ;; and tests whether it exists. If so (that is what the
7954 ;; `when' checks), it calls an `if' function that determines
7955 ;; whether the previous command was another call to
7956 ;; `kill-region'; if it was, then the new text is appended to
7957 ;; the previous text; if not, then a different function,
7958 ;; `kill-new', is called.
7962 ;; The `kill-append' function concatenates the new string and
7963 ;; the old. The `kill-new' function inserts text into a new
7964 ;; item in the kill ring.
7968 ;; `when' is an `if' without an else-part. The second `when'
7969 ;; again checks whether the current string exists; in
7970 ;; addition, it checks whether the previous command was
7971 ;; another call to `kill-region'. If one or the other
7972 ;; condition is true, then it sets the current command to
7973 ;; be `kill-region'.
7976 (let ((string (filter-buffer-substring beg end t)))
7977 (when string ;STRING is nil if BEG = END
7978 ;; Add that string to the kill ring, one way or another.
7979 (if (eq last-command 'kill-region)
7982 ;; @minus{} `yank-handler' is an optional argument to
7983 ;; `kill-region' that tells the `kill-append' and
7984 ;; `kill-new' functions how deal with properties
7985 ;; added to the text, such as `bold' or `italics'.
7986 (kill-append string (< end beg) yank-handler)
7987 (kill-new string nil yank-handler)))
7988 (when (or string (eq last-command 'kill-region))
7989 (setq this-command 'kill-region))
7994 ;; @bullet{} The third argument to `condition-case' tells the interpreter
7995 ;; what to do with an error.
7998 ;; The third argument has a conditions part and a body part.
7999 ;; If the conditions are met (in this case,
8000 ;; if text or buffer are read-only)
8001 ;; then the body is executed.
8004 ;; The first part of the third argument is the following:
8005 ((buffer-read-only text-read-only) ;; the if-part
8006 ;; @dots{} the then-part
8007 (copy-region-as-kill beg end)
8010 ;; Next, also as part of the then-part, set this-command, so
8011 ;; it will be set in an error
8012 (setq this-command 'kill-region)
8013 ;; Finally, in the then-part, send a message if you may copy
8014 ;; the text to the kill ring without signaling an error, but
8015 ;; don't if you may not.
8018 (if kill-read-only-ok
8019 (progn (message "Read only text copied to kill ring") nil)
8020 (barf-if-buffer-read-only)
8021 ;; If the buffer isn't read-only, the text is.
8022 (signal 'text-read-only (list (current-buffer)))))
8030 (defun kill-region (beg end)
8031 "Kill between point and mark.
8032 The text is deleted but saved in the kill ring."
8037 ;; 1. `condition-case' takes three arguments.
8038 ;; If the first argument is nil, as it is here,
8039 ;; information about the error signal is not
8040 ;; stored for use by another function.
8045 ;; 2. The second argument to `condition-case'
8046 ;; tells the Lisp interpreter what to do when all goes well.
8050 ;; The `delete-and-extract-region' function usually does the
8051 ;; work. If the beginning and ending of the region are both
8052 ;; the same, then the variable `string' will be empty, or nil
8053 (let ((string (delete-and-extract-region beg end)))
8057 ;; `when' is an `if' clause that cannot take an `else-part'.
8058 ;; Emacs normally sets the value of `last-command' to the
8059 ;; previous command.
8062 ;; `kill-append' concatenates the new string and the old.
8063 ;; `kill-new' inserts text into a new item in the kill ring.
8065 (if (eq last-command 'kill-region)
8066 ;; if true, prepend string
8067 (kill-append string (< end beg))
8069 (setq this-command 'kill-region))
8073 ;; 3. The third argument to `condition-case' tells the interpreter
8074 ;; what to do with an error.
8077 ;; The third argument has a conditions part and a body part.
8078 ;; If the conditions are met (in this case,
8079 ;; if text or buffer are read-only)
8080 ;; then the body is executed.
8083 ((buffer-read-only text-read-only) ;; this is the if-part
8085 (copy-region-as-kill beg end)
8088 (if kill-read-only-ok ;; usually this variable is nil
8089 (message "Read only text copied to kill ring")
8090 ;; or else, signal an error if the buffer is read-only;
8091 (barf-if-buffer-read-only)
8092 ;; and, in any case, signal that the text is read-only.
8093 (signal 'text-read-only (list (current-buffer)))))))
8098 @node condition-case
8099 @subsection @code{condition-case}
8100 @findex condition-case
8102 As we have seen earlier (@pxref{Making Errors, , Generate an Error
8103 Message}), when the Emacs Lisp interpreter has trouble evaluating an
8104 expression, it provides you with help; in the jargon, this is called
8105 ``signaling an error''. Usually, the computer stops the program and
8106 shows you a message.
8108 However, some programs undertake complicated actions. They should not
8109 simply stop on an error. In the @code{kill-region} function, the most
8110 likely error is that you will try to kill text that is read-only and
8111 cannot be removed. So the @code{kill-region} function contains code
8112 to handle this circumstance. This code, which makes up the body of
8113 the @code{kill-region} function, is inside of a @code{condition-case}
8117 The template for @code{condition-case} looks like this:
8124 @var{error-handler}@dots{})
8128 The second argument, @var{bodyform}, is straightforward. The
8129 @code{condition-case} special form causes the Lisp interpreter to
8130 evaluate the code in @var{bodyform}. If no error occurs, the special
8131 form returns the code's value and produces the side-effects, if any.
8133 In short, the @var{bodyform} part of a @code{condition-case}
8134 expression determines what should happen when everything works
8137 However, if an error occurs, among its other actions, the function
8138 generating the error signal will define one or more error condition
8141 An error handler is the third argument to @code{condition case}.
8142 An error handler has two parts, a @var{condition-name} and a
8143 @var{body}. If the @var{condition-name} part of an error handler
8144 matches a condition name generated by an error, then the @var{body}
8145 part of the error handler is run.
8147 As you will expect, the @var{condition-name} part of an error handler
8148 may be either a single condition name or a list of condition names.
8150 Also, a complete @code{condition-case} expression may contain more
8151 than one error handler. When an error occurs, the first applicable
8154 Lastly, the first argument to the @code{condition-case} expression,
8155 the @var{var} argument, is sometimes bound to a variable that
8156 contains information about the error. However, if that argument is
8157 nil, as is the case in @code{kill-region}, that information is
8161 In brief, in the @code{kill-region} function, the code
8162 @code{condition-case} works like this:
8166 @var{If no errors}, @var{run only this code}
8167 @var{but}, @var{if errors}, @var{run this other code}.
8174 copy-region-as-kill is short, 12 lines, and uses
8175 filter-buffer-substring, which is longer, 39 lines
8176 and has delete-and-extract-region in it.
8177 delete-and-extract-region is written in C.
8179 see Initializing a Variable with @code{defvar}
8181 Initializing a Variable with @code{defvar} includes line 8350
8185 @subsection Lisp macro
8189 The part of the @code{condition-case} expression that is evaluated in
8190 the expectation that all goes well has a @code{when}. The code uses
8191 @code{when} to determine whether the @code{string} variable points to
8194 A @code{when} expression is simply a programmers' convenience. It is
8195 an @code{if} without the possibility of an else clause. In your mind,
8196 you can replace @code{when} with @code{if} and understand what goes
8197 on. That is what the Lisp interpreter does.
8199 Technically speaking, @code{when} is a Lisp macro. A Lisp macro
8200 enables you to define new control constructs and other language
8201 features. It tells the interpreter how to compute another Lisp
8202 expression which will in turn compute the value. In this case, the
8203 `other expression' is an @code{if} expression.
8205 The @code{kill-region} function definition also has an @code{unless}
8206 macro; it is the converse of @code{when}. The @code{unless} macro is
8207 an @code{if} without a then clause
8209 For more about Lisp macros, see @ref{Macros, , Macros, elisp, The GNU
8210 Emacs Lisp Reference Manual}. The C programming language also
8211 provides macros. These are different, but also useful.
8214 We will briefly look at C macros in
8215 @ref{Digression into C}.
8219 Regarding the @code{when} macro, in the @code{condition-case}
8220 expression, when the string has content, then another conditional
8221 expression is executed. This is an @code{if} with both a then-part
8226 (if (eq last-command 'kill-region)
8227 (kill-append string (< end beg) yank-handler)
8228 (kill-new string nil yank-handler))
8232 The then-part is evaluated if the previous command was another call to
8233 @code{kill-region}; if not, the else-part is evaluated.
8235 @code{yank-handler} is an optional argument to @code{kill-region} that
8236 tells the @code{kill-append} and @code{kill-new} functions how deal
8237 with properties added to the text, such as `bold' or `italics'.
8239 @code{last-command} is a variable that comes with Emacs that we have
8240 not seen before. Normally, whenever a function is executed, Emacs
8241 sets the value of @code{last-command} to the previous command.
8244 In this segment of the definition, the @code{if} expression checks
8245 whether the previous command was @code{kill-region}. If it was,
8248 (kill-append string (< end beg) yank-handler)
8252 concatenates a copy of the newly clipped text to the just previously
8253 clipped text in the kill ring.
8255 @node copy-region-as-kill
8256 @section @code{copy-region-as-kill}
8257 @findex copy-region-as-kill
8260 The @code{copy-region-as-kill} function copies a region of text from a
8261 buffer and (via either @code{kill-append} or @code{kill-new}) saves it
8262 in the @code{kill-ring}.
8264 If you call @code{copy-region-as-kill} immediately after a
8265 @code{kill-region} command, Emacs appends the newly copied text to the
8266 previously copied text. This means that if you yank back the text, you
8267 get it all, from both this and the previous operation. On the other
8268 hand, if some other command precedes the @code{copy-region-as-kill},
8269 the function copies the text into a separate entry in the kill ring.
8272 * Complete copy-region-as-kill:: The complete function definition.
8273 * copy-region-as-kill body:: The body of @code{copy-region-as-kill}.
8277 @node Complete copy-region-as-kill
8278 @unnumberedsubsec The complete @code{copy-region-as-kill} function definition
8282 Here is the complete text of the version 22 @code{copy-region-as-kill}
8287 (defun copy-region-as-kill (beg end)
8288 "Save the region as if killed, but don't kill it.
8289 In Transient Mark mode, deactivate the mark.
8290 If `interprogram-cut-function' is non-nil, also save the text for a window
8291 system cut and paste."
8295 (if (eq last-command 'kill-region)
8296 (kill-append (filter-buffer-substring beg end) (< end beg))
8297 (kill-new (filter-buffer-substring beg end)))
8300 (if transient-mark-mode
8301 (setq deactivate-mark t))
8307 As usual, this function can be divided into its component parts:
8311 (defun copy-region-as-kill (@var{argument-list})
8312 "@var{documentation}@dots{}"
8318 The arguments are @code{beg} and @code{end} and the function is
8319 interactive with @code{"r"}, so the two arguments must refer to the
8320 beginning and end of the region. If you have been reading through this
8321 document from the beginning, understanding these parts of a function is
8322 almost becoming routine.
8324 The documentation is somewhat confusing unless you remember that the
8325 word `kill' has a meaning different from usual. The `Transient Mark'
8326 and @code{interprogram-cut-function} comments explain certain
8329 After you once set a mark, a buffer always contains a region. If you
8330 wish, you can use Transient Mark mode to highlight the region
8331 temporarily. (No one wants to highlight the region all the time, so
8332 Transient Mark mode highlights it only at appropriate times. Many
8333 people turn off Transient Mark mode, so the region is never
8336 Also, a windowing system allows you to copy, cut, and paste among
8337 different programs. In the X windowing system, for example, the
8338 @code{interprogram-cut-function} function is @code{x-select-text},
8339 which works with the windowing system's equivalent of the Emacs kill
8342 The body of the @code{copy-region-as-kill} function starts with an
8343 @code{if} clause. What this clause does is distinguish between two
8344 different situations: whether or not this command is executed
8345 immediately after a previous @code{kill-region} command. In the first
8346 case, the new region is appended to the previously copied text.
8347 Otherwise, it is inserted into the beginning of the kill ring as a
8348 separate piece of text from the previous piece.
8350 The last two lines of the function prevent the region from lighting up
8351 if Transient Mark mode is turned on.
8353 The body of @code{copy-region-as-kill} merits discussion in detail.
8355 @node copy-region-as-kill body
8356 @subsection The Body of @code{copy-region-as-kill}
8358 The @code{copy-region-as-kill} function works in much the same way as
8359 the @code{kill-region} function. Both are written so that two or more
8360 kills in a row combine their text into a single entry. If you yank
8361 back the text from the kill ring, you get it all in one piece.
8362 Moreover, kills that kill forward from the current position of the
8363 cursor are added to the end of the previously copied text and commands
8364 that copy text backwards add it to the beginning of the previously
8365 copied text. This way, the words in the text stay in the proper
8368 Like @code{kill-region}, the @code{copy-region-as-kill} function makes
8369 use of the @code{last-command} variable that keeps track of the
8370 previous Emacs command.
8373 * last-command & this-command::
8374 * kill-append function::
8375 * kill-new function::
8379 @node last-command & this-command
8380 @unnumberedsubsubsec @code{last-command} and @code{this-command}
8383 Normally, whenever a function is executed, Emacs sets the value of
8384 @code{this-command} to the function being executed (which in this case
8385 would be @code{copy-region-as-kill}). At the same time, Emacs sets
8386 the value of @code{last-command} to the previous value of
8387 @code{this-command}.
8389 In the first part of the body of the @code{copy-region-as-kill}
8390 function, an @code{if} expression determines whether the value of
8391 @code{last-command} is @code{kill-region}. If so, the then-part of
8392 the @code{if} expression is evaluated; it uses the @code{kill-append}
8393 function to concatenate the text copied at this call to the function
8394 with the text already in the first element (the @sc{car}) of the kill
8395 ring. On the other hand, if the value of @code{last-command} is not
8396 @code{kill-region}, then the @code{copy-region-as-kill} function
8397 attaches a new element to the kill ring using the @code{kill-new}
8401 The @code{if} expression reads as follows; it uses @code{eq}:
8405 (if (eq last-command 'kill-region)
8407 (kill-append (filter-buffer-substring beg end) (< end beg))
8409 (kill-new (filter-buffer-substring beg end)))
8413 @findex filter-buffer-substring
8414 (The @code{filter-buffer-substring} function returns a filtered
8415 substring of the buffer, if any. Optionally---the arguments are not
8416 here, so neither is done---the function may delete the initial text or
8417 return the text without its properties; this function is a replacement
8418 for the older @code{buffer-substring} function, which came before text
8419 properties were implemented.)
8421 @findex eq @r{(example of use)}
8423 The @code{eq} function tests whether its first argument is the same Lisp
8424 object as its second argument. The @code{eq} function is similar to the
8425 @code{equal} function in that it is used to test for equality, but
8426 differs in that it determines whether two representations are actually
8427 the same object inside the computer, but with different names.
8428 @code{equal} determines whether the structure and contents of two
8429 expressions are the same.
8431 If the previous command was @code{kill-region}, then the Emacs Lisp
8432 interpreter calls the @code{kill-append} function
8434 @node kill-append function
8435 @unnumberedsubsubsec The @code{kill-append} function
8439 The @code{kill-append} function looks like this:
8444 (defun kill-append (string before-p &optional yank-handler)
8445 "Append STRING to the end of the latest kill in the kill ring.
8446 If BEFORE-P is non-nil, prepend STRING to the kill.
8448 (let* ((cur (car kill-ring)))
8449 (kill-new (if before-p (concat string cur) (concat cur string))
8450 (or (= (length cur) 0)
8452 (get-text-property 0 'yank-handler cur)))
8459 (defun kill-append (string before-p)
8460 "Append STRING to the end of the latest kill in the kill ring.
8461 If BEFORE-P is non-nil, prepend STRING to the kill.
8462 If `interprogram-cut-function' is set, pass the resulting kill to
8464 (kill-new (if before-p
8465 (concat string (car kill-ring))
8466 (concat (car kill-ring) string))
8471 The @code{kill-append} function is fairly straightforward. It uses
8472 the @code{kill-new} function, which we will discuss in more detail in
8475 (Also, the function provides an optional argument called
8476 @code{yank-handler}; when invoked, this argument tells the function
8477 how to deal with properties added to the text, such as `bold' or
8480 @c !!! bug in GNU Emacs 22 version of kill-append ?
8481 It has a @code{let*} function to set the value of the first element of
8482 the kill ring to @code{cur}. (I do not know why the function does not
8483 use @code{let} instead; only one value is set in the expression.
8484 Perhaps this is a bug that produces no problems?)
8486 Consider the conditional that is one of the two arguments to
8487 @code{kill-new}. It uses @code{concat} to concatenate the new text to
8488 the @sc{car} of the kill ring. Whether it prepends or appends the
8489 text depends on the results of an @code{if} expression:
8493 (if before-p ; @r{if-part}
8494 (concat string cur) ; @r{then-part}
8495 (concat cur string)) ; @r{else-part}
8500 If the region being killed is before the region that was killed in the
8501 last command, then it should be prepended before the material that was
8502 saved in the previous kill; and conversely, if the killed text follows
8503 what was just killed, it should be appended after the previous text.
8504 The @code{if} expression depends on the predicate @code{before-p} to
8505 decide whether the newly saved text should be put before or after the
8506 previously saved text.
8508 The symbol @code{before-p} is the name of one of the arguments to
8509 @code{kill-append}. When the @code{kill-append} function is
8510 evaluated, it is bound to the value returned by evaluating the actual
8511 argument. In this case, this is the expression @code{(< end beg)}.
8512 This expression does not directly determine whether the killed text in
8513 this command is located before or after the kill text of the last
8514 command; what it does is determine whether the value of the variable
8515 @code{end} is less than the value of the variable @code{beg}. If it
8516 is, it means that the user is most likely heading towards the
8517 beginning of the buffer. Also, the result of evaluating the predicate
8518 expression, @code{(< end beg)}, will be true and the text will be
8519 prepended before the previous text. On the other hand, if the value of
8520 the variable @code{end} is greater than the value of the variable
8521 @code{beg}, the text will be appended after the previous text.
8524 When the newly saved text will be prepended, then the string with the new
8525 text will be concatenated before the old text:
8533 But if the text will be appended, it will be concatenated
8537 (concat cur string))
8540 To understand how this works, we first need to review the
8541 @code{concat} function. The @code{concat} function links together or
8542 unites two strings of text. The result is a string. For example:
8546 (concat "abc" "def")
8552 (car '("first element" "second element")))
8553 @result{} "new first element"
8556 '("first element" "second element")) " modified")
8557 @result{} "first element modified"
8561 We can now make sense of @code{kill-append}: it modifies the contents
8562 of the kill ring. The kill ring is a list, each element of which is
8563 saved text. The @code{kill-append} function uses the @code{kill-new}
8564 function which in turn uses the @code{setcar} function.
8566 @node kill-new function
8567 @unnumberedsubsubsec The @code{kill-new} function
8570 @c in GNU Emacs 22, additional documentation to kill-new:
8572 Optional third arguments YANK-HANDLER controls how the STRING is later
8573 inserted into a buffer; see `insert-for-yank' for details.
8574 When a yank handler is specified, STRING must be non-empty (the yank
8575 handler, if non-nil, is stored as a `yank-handler' text property on STRING).
8577 When the yank handler has a non-nil PARAM element, the original STRING
8578 argument is not used by `insert-for-yank'. However, since Lisp code
8579 may access and use elements from the kill ring directly, the STRING
8580 argument should still be a \"useful\" string for such uses."
8583 The @code{kill-new} function looks like this:
8587 (defun kill-new (string &optional replace yank-handler)
8588 "Make STRING the latest kill in the kill ring.
8589 Set `kill-ring-yank-pointer' to point to it.
8591 If `interprogram-cut-function' is non-nil, apply it to STRING.
8592 Optional second argument REPLACE non-nil means that STRING will replace
8593 the front of the kill ring, rather than being added to the list.
8597 (if (> (length string) 0)
8599 (put-text-property 0 (length string)
8600 'yank-handler yank-handler string))
8602 (signal 'args-out-of-range
8603 (list string "yank-handler specified for empty string"))))
8606 (if (fboundp 'menu-bar-update-yank-menu)
8607 (menu-bar-update-yank-menu string (and replace (car kill-ring))))
8610 (if (and replace kill-ring)
8611 (setcar kill-ring string)
8612 (push string kill-ring)
8613 (if (> (length kill-ring) kill-ring-max)
8614 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil)))
8617 (setq kill-ring-yank-pointer kill-ring)
8618 (if interprogram-cut-function
8619 (funcall interprogram-cut-function string (not replace))))
8624 (defun kill-new (string &optional replace)
8625 "Make STRING the latest kill in the kill ring.
8626 Set the kill-ring-yank pointer to point to it.
8627 If `interprogram-cut-function' is non-nil, apply it to STRING.
8628 Optional second argument REPLACE non-nil means that STRING will replace
8629 the front of the kill ring, rather than being added to the list."
8630 (and (fboundp 'menu-bar-update-yank-menu)
8631 (menu-bar-update-yank-menu string (and replace (car kill-ring))))
8632 (if (and replace kill-ring)
8633 (setcar kill-ring string)
8634 (setq kill-ring (cons string kill-ring))
8635 (if (> (length kill-ring) kill-ring-max)
8636 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil)))
8637 (setq kill-ring-yank-pointer kill-ring)
8638 (if interprogram-cut-function
8639 (funcall interprogram-cut-function string (not replace))))
8642 (Notice that the function is not interactive.)
8644 As usual, we can look at this function in parts.
8646 The function definition has an optional @code{yank-handler} argument,
8647 which when invoked tells the function how to deal with properties
8648 added to the text, such as `bold' or `italics'. We will skip that.
8651 The first line of the documentation makes sense:
8654 Make STRING the latest kill in the kill ring.
8658 Let's skip over the rest of the documentation for the moment.
8661 Also, let's skip over the initial @code{if} expression and those lines
8662 of code involving @code{menu-bar-update-yank-menu}. We will explain
8666 The critical lines are these:
8670 (if (and replace kill-ring)
8672 (setcar kill-ring string)
8676 (push string kill-ring)
8679 (setq kill-ring (cons string kill-ring))
8680 (if (> (length kill-ring) kill-ring-max)
8681 ;; @r{avoid overly long kill ring}
8682 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil)))
8685 (setq kill-ring-yank-pointer kill-ring)
8686 (if interprogram-cut-function
8687 (funcall interprogram-cut-function string (not replace))))
8691 The conditional test is @w{@code{(and replace kill-ring)}}.
8692 This will be true when two conditions are met: the kill ring has
8693 something in it, and the @code{replace} variable is true.
8696 When the @code{kill-append} function sets @code{replace} to be true
8697 and when the kill ring has at least one item in it, the @code{setcar}
8698 expression is executed:
8701 (setcar kill-ring string)
8704 The @code{setcar} function actually changes the first element of the
8705 @code{kill-ring} list to the value of @code{string}. It replaces the
8709 On the other hand, if the kill ring is empty, or replace is false, the
8710 else-part of the condition is executed:
8713 (push string kill-ring)
8718 @code{push} puts its first argument onto the second. It is similar to
8722 (setq kill-ring (cons string kill-ring))
8730 (add-to-list kill-ring string)
8734 When it is false, the expression first constructs a new version of the
8735 kill ring by prepending @code{string} to the existing kill ring as a
8736 new element (that is what the @code{push} does). Then it executes a
8737 second @code{if} clause. This second @code{if} clause keeps the kill
8738 ring from growing too long.
8740 Let's look at these two expressions in order.
8742 The @code{push} line of the else-part sets the new value of the kill
8743 ring to what results from adding the string being killed to the old
8746 We can see how this works with an example.
8752 (setq example-list '("here is a clause" "another clause"))
8757 After evaluating this expression with @kbd{C-x C-e}, you can evaluate
8758 @code{example-list} and see what it returns:
8763 @result{} ("here is a clause" "another clause")
8769 Now, we can add a new element on to this list by evaluating the
8770 following expression:
8771 @findex push, @r{example}
8774 (push "a third clause" example-list)
8779 When we evaluate @code{example-list}, we find its value is:
8784 @result{} ("a third clause" "here is a clause" "another clause")
8789 Thus, the third clause is added to the list by @code{push}.
8792 Now for the second part of the @code{if} clause. This expression
8793 keeps the kill ring from growing too long. It looks like this:
8797 (if (> (length kill-ring) kill-ring-max)
8798 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil))
8802 The code checks whether the length of the kill ring is greater than
8803 the maximum permitted length. This is the value of
8804 @code{kill-ring-max} (which is 60, by default). If the length of the
8805 kill ring is too long, then this code sets the last element of the
8806 kill ring to @code{nil}. It does this by using two functions,
8807 @code{nthcdr} and @code{setcdr}.
8809 We looked at @code{setcdr} earlier (@pxref{setcdr, , @code{setcdr}}).
8810 It sets the @sc{cdr} of a list, just as @code{setcar} sets the
8811 @sc{car} of a list. In this case, however, @code{setcdr} will not be
8812 setting the @sc{cdr} of the whole kill ring; the @code{nthcdr}
8813 function is used to cause it to set the @sc{cdr} of the next to last
8814 element of the kill ring---this means that since the @sc{cdr} of the
8815 next to last element is the last element of the kill ring, it will set
8816 the last element of the kill ring.
8818 @findex nthcdr, @r{example}
8819 The @code{nthcdr} function works by repeatedly taking the @sc{cdr} of a
8820 list---it takes the @sc{cdr} of the @sc{cdr} of the @sc{cdr}
8821 @dots{} It does this @var{N} times and returns the results.
8822 (@xref{nthcdr, , @code{nthcdr}}.)
8824 @findex setcdr, @r{example}
8825 Thus, if we had a four element list that was supposed to be three
8826 elements long, we could set the @sc{cdr} of the next to last element
8827 to @code{nil}, and thereby shorten the list. (If you set the last
8828 element to some other value than @code{nil}, which you could do, then
8829 you would not have shortened the list. @xref{setcdr, ,
8832 You can see shortening by evaluating the following three expressions
8833 in turn. First set the value of @code{trees} to @code{(maple oak pine
8834 birch)}, then set the @sc{cdr} of its second @sc{cdr} to @code{nil}
8835 and then find the value of @code{trees}:
8839 (setq trees '(maple oak pine birch))
8840 @result{} (maple oak pine birch)
8844 (setcdr (nthcdr 2 trees) nil)
8848 @result{} (maple oak pine)
8853 (The value returned by the @code{setcdr} expression is @code{nil} since
8854 that is what the @sc{cdr} is set to.)
8856 To repeat, in @code{kill-new}, the @code{nthcdr} function takes the
8857 @sc{cdr} a number of times that is one less than the maximum permitted
8858 size of the kill ring and @code{setcdr} sets the @sc{cdr} of that
8859 element (which will be the rest of the elements in the kill ring) to
8860 @code{nil}. This prevents the kill ring from growing too long.
8863 The next to last expression in the @code{kill-new} function is
8866 (setq kill-ring-yank-pointer kill-ring)
8869 The @code{kill-ring-yank-pointer} is a global variable that is set to be
8870 the @code{kill-ring}.
8872 Even though the @code{kill-ring-yank-pointer} is called a
8873 @samp{pointer}, it is a variable just like the kill ring. However, the
8874 name has been chosen to help humans understand how the variable is used.
8877 Now, to return to an early expression in the body of the function:
8881 (if (fboundp 'menu-bar-update-yank-menu)
8882 (menu-bar-update-yank-menu string (and replace (car kill-ring))))
8887 It starts with an @code{if} expression
8889 In this case, the expression tests first to see whether
8890 @code{menu-bar-update-yank-menu} exists as a function, and if so,
8891 calls it. The @code{fboundp} function returns true if the symbol it
8892 is testing has a function definition that `is not void'. If the
8893 symbol's function definition were void, we would receive an error
8894 message, as we did when we created errors intentionally (@pxref{Making
8895 Errors, , Generate an Error Message}).
8898 The then-part contains an expression whose first element is the
8899 function @code{and}.
8902 The @code{and} special form evaluates each of its arguments until one
8903 of the arguments returns a value of @code{nil}, in which case the
8904 @code{and} expression returns @code{nil}; however, if none of the
8905 arguments returns a value of @code{nil}, the value resulting from
8906 evaluating the last argument is returned. (Since such a value is not
8907 @code{nil}, it is considered true in Emacs Lisp.) In other words, an
8908 @code{and} expression returns a true value only if all its arguments
8909 are true. (@xref{Second Buffer Related Review}.)
8911 The expression determines whether the second argument to
8912 @code{menu-bar-update-yank-menu} is true or not.
8914 ;; If we're supposed to be extending an existing string, and that
8915 ;; string really is at the front of the menu, then update it in place.
8918 @code{menu-bar-update-yank-menu} is one of the functions that make it
8919 possible to use the `Select and Paste' menu in the Edit item of a menu
8920 bar; using a mouse, you can look at the various pieces of text you
8921 have saved and select one piece to paste.
8923 The last expression in the @code{kill-new} function adds the newly
8924 copied string to whatever facility exists for copying and pasting
8925 among different programs running in a windowing system. In the X
8926 Windowing system, for example, the @code{x-select-text} function takes
8927 the string and stores it in memory operated by X@. You can paste the
8928 string in another program, such as an Xterm.
8931 The expression looks like this:
8935 (if interprogram-cut-function
8936 (funcall interprogram-cut-function string (not replace))))
8940 If an @code{interprogram-cut-function} exists, then Emacs executes
8941 @code{funcall}, which in turn calls its first argument as a function
8942 and passes the remaining arguments to it. (Incidentally, as far as I
8943 can see, this @code{if} expression could be replaced by an @code{and}
8944 expression similar to the one in the first part of the function.)
8946 We are not going to discuss windowing systems and other programs
8947 further, but merely note that this is a mechanism that enables GNU
8948 Emacs to work easily and well with other programs.
8950 This code for placing text in the kill ring, either concatenated with
8951 an existing element or as a new element, leads us to the code for
8952 bringing back text that has been cut out of the buffer---the yank
8953 commands. However, before discussing the yank commands, it is better
8954 to learn how lists are implemented in a computer. This will make
8955 clear such mysteries as the use of the term `pointer'. But before
8956 that, we will digress into C.
8959 @c is this true in Emacs 22? Does not seems to be
8961 (If the @w{@code{(< end beg))}}
8962 expression is true, @code{kill-append} prepends the string to the just
8963 previously clipped text. For a detailed discussion, see
8964 @ref{kill-append function, , The @code{kill-append} function}.)
8966 If you then yank back the text, i.e., `paste' it, you get both
8967 pieces of text at once. That way, if you delete two words in a row,
8968 and then yank them back, you get both words, in their proper order,
8969 with one yank. (The @w{@code{(< end beg))}} expression makes sure the
8972 On the other hand, if the previous command is not @code{kill-region},
8973 then the @code{kill-new} function is called, which adds the text to
8974 the kill ring as the latest item, and sets the
8975 @code{kill-ring-yank-pointer} variable to point to it.
8979 @c Evidently, changed for Emacs 22. The zap-to-char command does not
8980 @c use the delete-and-extract-region function
8982 2006 Oct 26, the Digression into C is now OK but should come after
8983 copy-region-as-kill and filter-buffer-substring
8987 copy-region-as-kill is short, 12 lines, and uses
8988 filter-buffer-substring, which is longer, 39 lines
8989 and has delete-and-extract-region in it.
8990 delete-and-extract-region is written in C.
8992 see Initializing a Variable with @code{defvar}
8995 @node Digression into C
8996 @section Digression into C
8997 @findex delete-and-extract-region
8998 @cindex C, a digression into
8999 @cindex Digression into C
9001 The @code{copy-region-as-kill} function (@pxref{copy-region-as-kill, ,
9002 @code{copy-region-as-kill}}) uses the @code{filter-buffer-substring}
9003 function, which in turn uses the @code{delete-and-extract-region}
9004 function. It removes the contents of a region and you cannot get them
9007 Unlike the other code discussed here, the
9008 @code{delete-and-extract-region} function is not written in Emacs
9009 Lisp; it is written in C and is one of the primitives of the GNU Emacs
9010 system. Since it is very simple, I will digress briefly from Lisp and
9013 @c GNU Emacs 24 in src/editfns.c
9014 @c the DEFUN for delete-and-extract-region
9017 Like many of the other Emacs primitives,
9018 @code{delete-and-extract-region} is written as an instance of a C
9019 macro, a macro being a template for code. The complete macro looks
9024 DEFUN ("delete-and-extract-region", Fdelete_and_extract_region,
9025 Sdelete_and_extract_region, 2, 2, 0,
9026 doc: /* Delete the text between START and END and return it. */)
9027 (Lisp_Object start, Lisp_Object end)
9029 validate_region (&start, &end);
9030 if (XINT (start) == XINT (end))
9031 return empty_unibyte_string;
9032 return del_range_1 (XINT (start), XINT (end), 1, 1);
9037 Without going into the details of the macro writing process, let me
9038 point out that this macro starts with the word @code{DEFUN}. The word
9039 @code{DEFUN} was chosen since the code serves the same purpose as
9040 @code{defun} does in Lisp. (The @code{DEFUN} C macro is defined in
9041 @file{emacs/src/lisp.h}.)
9043 The word @code{DEFUN} is followed by seven parts inside of
9048 The first part is the name given to the function in Lisp,
9049 @code{delete-and-extract-region}.
9052 The second part is the name of the function in C,
9053 @code{Fdelete_and_extract_region}. By convention, it starts with
9054 @samp{F}. Since C does not use hyphens in names, underscores are used
9058 The third part is the name for the C constant structure that records
9059 information on this function for internal use. It is the name of the
9060 function in C but begins with an @samp{S} instead of an @samp{F}.
9063 The fourth and fifth parts specify the minimum and maximum number of
9064 arguments the function can have. This function demands exactly 2
9068 The sixth part is nearly like the argument that follows the
9069 @code{interactive} declaration in a function written in Lisp: a letter
9070 followed, perhaps, by a prompt. The only difference from the Lisp is
9071 when the macro is called with no arguments. Then you write a @code{0}
9072 (which is a `null string'), as in this macro.
9074 If you were to specify arguments, you would place them between
9075 quotation marks. The C macro for @code{goto-char} includes
9076 @code{"NGoto char: "} in this position to indicate that the function
9077 expects a raw prefix, in this case, a numerical location in a buffer,
9078 and provides a prompt.
9081 The seventh part is a documentation string, just like the one for a
9082 function written in Emacs Lisp. This is written as a C comment. (When
9083 you build Emacs, the program @command{lib-src/make-docfile} extracts
9084 these comments and uses them to make the ``real'' documentation.)
9088 In a C macro, the formal parameters come next, with a statement of
9089 what kind of object they are, followed by what might be called the `body'
9090 of the macro. For @code{delete-and-extract-region} the `body'
9091 consists of the following four lines:
9095 validate_region (&start, &end);
9096 if (XINT (start) == XINT (end))
9097 return empty_unibyte_string;
9098 return del_range_1 (XINT (start), XINT (end), 1, 1);
9102 The @code{validate_region} function checks whether the values
9103 passed as the beginning and end of the region are the proper type and
9104 are within range. If the beginning and end positions are the same,
9105 then return an empty string.
9107 The @code{del_range_1} function actually deletes the text. It is a
9108 complex function we will not look into. It updates the buffer and
9109 does other things. However, it is worth looking at the two arguments
9110 passed to @code{del_range}. These are @w{@code{XINT (start)}} and
9111 @w{@code{XINT (end)}}.
9113 As far as the C language is concerned, @code{start} and @code{end} are
9114 two integers that mark the beginning and end of the region to be
9115 deleted@footnote{More precisely, and requiring more expert knowledge
9116 to understand, the two integers are of type `Lisp_Object', which can
9117 also be a C union instead of an integer type.}.
9119 In early versions of Emacs, these two numbers were thirty-two bits
9120 long, but the code is slowly being generalized to handle other
9121 lengths. Three of the available bits are used to specify the type of
9122 information; the remaining bits are used as `content'.
9124 @samp{XINT} is a C macro that extracts the relevant number from the
9125 longer collection of bits; the three other bits are discarded.
9128 The command in @code{delete-and-extract-region} looks like this:
9131 del_range_1 (XINT (start), XINT (end), 1, 1);
9135 It deletes the region between the beginning position, @code{start},
9136 and the ending position, @code{end}.
9138 From the point of view of the person writing Lisp, Emacs is all very
9139 simple; but hidden underneath is a great deal of complexity to make it
9143 @section Initializing a Variable with @code{defvar}
9145 @cindex Initializing a variable
9146 @cindex Variable initialization
9151 copy-region-as-kill is short, 12 lines, and uses
9152 filter-buffer-substring, which is longer, 39 lines
9153 and has delete-and-extract-region in it.
9154 delete-and-extract-region is written in C.
9156 see Initializing a Variable with @code{defvar}
9160 The @code{copy-region-as-kill} function is written in Emacs Lisp. Two
9161 functions within it, @code{kill-append} and @code{kill-new}, copy a
9162 region in a buffer and save it in a variable called the
9163 @code{kill-ring}. This section describes how the @code{kill-ring}
9164 variable is created and initialized using the @code{defvar} special
9167 (Again we note that the term @code{kill-ring} is a misnomer. The text
9168 that is clipped out of the buffer can be brought back; it is not a ring
9169 of corpses, but a ring of resurrectable text.)
9171 In Emacs Lisp, a variable such as the @code{kill-ring} is created and
9172 given an initial value by using the @code{defvar} special form. The
9173 name comes from ``define variable''.
9175 The @code{defvar} special form is similar to @code{setq} in that it sets
9176 the value of a variable. It is unlike @code{setq} in two ways: first,
9177 it only sets the value of the variable if the variable does not already
9178 have a value. If the variable already has a value, @code{defvar} does
9179 not override the existing value. Second, @code{defvar} has a
9180 documentation string.
9182 (There is a related macro, @code{defcustom}, designed for variables
9183 that people customize. It has more features than @code{defvar}.
9184 (@xref{defcustom, , Setting Variables with @code{defcustom}}.)
9187 * See variable current value::
9188 * defvar and asterisk::
9192 @node See variable current value
9193 @unnumberedsubsec Seeing the Current Value of a Variable
9196 You can see the current value of a variable, any variable, by using
9197 the @code{describe-variable} function, which is usually invoked by
9198 typing @kbd{C-h v}. If you type @kbd{C-h v} and then @code{kill-ring}
9199 (followed by @key{RET}) when prompted, you will see what is in your
9200 current kill ring---this may be quite a lot! Conversely, if you have
9201 been doing nothing this Emacs session except read this document, you
9202 may have nothing in it. Also, you will see the documentation for
9208 List of killed text sequences.
9209 Since the kill ring is supposed to interact nicely with cut-and-paste
9210 facilities offered by window systems, use of this variable should
9213 interact nicely with `interprogram-cut-function' and
9214 `interprogram-paste-function'. The functions `kill-new',
9215 `kill-append', and `current-kill' are supposed to implement this
9216 interaction; you may want to use them instead of manipulating the kill
9222 The kill ring is defined by a @code{defvar} in the following way:
9226 (defvar kill-ring nil
9227 "List of killed text sequences.
9233 In this variable definition, the variable is given an initial value of
9234 @code{nil}, which makes sense, since if you have saved nothing, you want
9235 nothing back if you give a @code{yank} command. The documentation
9236 string is written just like the documentation string of a @code{defun}.
9237 As with the documentation string of the @code{defun}, the first line of
9238 the documentation should be a complete sentence, since some commands,
9239 like @code{apropos}, print only the first line of documentation.
9240 Succeeding lines should not be indented; otherwise they look odd when
9241 you use @kbd{C-h v} (@code{describe-variable}).
9243 @node defvar and asterisk
9244 @subsection @code{defvar} and an asterisk
9245 @findex defvar @r{for a user customizable variable}
9246 @findex defvar @r{with an asterisk}
9248 In the past, Emacs used the @code{defvar} special form both for
9249 internal variables that you would not expect a user to change and for
9250 variables that you do expect a user to change. Although you can still
9251 use @code{defvar} for user customizable variables, please use
9252 @code{defcustom} instead, since it provides a path into
9253 the Customization commands. (@xref{defcustom, , Specifying Variables
9254 using @code{defcustom}}.)
9256 When you specified a variable using the @code{defvar} special form,
9257 you could distinguish a variable that a user might want to change from
9258 others by typing an asterisk, @samp{*}, in the first column of its
9259 documentation string. For example:
9263 (defvar shell-command-default-error-buffer nil
9264 "*Buffer name for `shell-command' @dots{} error output.
9269 @findex set-variable
9271 You could (and still can) use the @code{set-variable} command to
9272 change the value of @code{shell-command-default-error-buffer}
9273 temporarily. However, options set using @code{set-variable} are set
9274 only for the duration of your editing session. The new values are not
9275 saved between sessions. Each time Emacs starts, it reads the original
9276 value, unless you change the value within your @file{.emacs} file,
9277 either by setting it manually or by using @code{customize}.
9278 @xref{Emacs Initialization, , Your @file{.emacs} File}.
9280 For me, the major use of the @code{set-variable} command is to suggest
9281 variables that I might want to set in my @file{.emacs} file. There
9282 are now more than 700 such variables, far too many to remember
9283 readily. Fortunately, you can press @key{TAB} after calling the
9284 @code{M-x set-variable} command to see the list of variables.
9285 (@xref{Examining, , Examining and Setting Variables, emacs,
9286 The GNU Emacs Manual}.)
9289 @node cons & search-fwd Review
9292 Here is a brief summary of some recently introduced functions.
9297 @code{car} returns the first element of a list; @code{cdr} returns the
9298 second and subsequent elements of a list.
9305 (car '(1 2 3 4 5 6 7))
9307 (cdr '(1 2 3 4 5 6 7))
9308 @result{} (2 3 4 5 6 7)
9313 @code{cons} constructs a list by prepending its first argument to its
9327 @code{funcall} evaluates its first argument as a function. It passes
9328 its remaining arguments to its first argument.
9331 Return the result of taking @sc{cdr} `n' times on a list.
9339 The `rest of the rest', as it were.
9346 (nthcdr 3 '(1 2 3 4 5 6 7))
9353 @code{setcar} changes the first element of a list; @code{setcdr}
9354 changes the second and subsequent elements of a list.
9361 (setq triple '(1 2 3))
9368 (setcdr triple '("foo" "bar"))
9371 @result{} (37 "foo" "bar")
9376 Evaluate each argument in sequence and then return the value of the
9389 @item save-restriction
9390 Record whatever narrowing is in effect in the current buffer, if any,
9391 and restore that narrowing after evaluating the arguments.
9393 @item search-forward
9394 Search for a string, and if the string is found, move point. With a
9395 regular expression, use the similar @code{re-search-forward}.
9396 (@xref{Regexp Search, , Regular Expression Searches}, for an
9397 explanation of regular expression patterns and searches.)
9401 @code{search-forward} and @code{re-search-forward} take four
9406 The string or regular expression to search for.
9409 Optionally, the limit of the search.
9412 Optionally, what to do if the search fails, return @code{nil} or an
9416 Optionally, how many times to repeat the search; if negative, the
9417 search goes backwards.
9421 @itemx delete-and-extract-region
9422 @itemx copy-region-as-kill
9424 @code{kill-region} cuts the text between point and mark from the
9425 buffer and stores that text in the kill ring, so you can get it back
9428 @code{copy-region-as-kill} copies the text between point and mark into
9429 the kill ring, from which you can get it by yanking. The function
9430 does not cut or remove the text from the buffer.
9433 @code{delete-and-extract-region} removes the text between point and
9434 mark from the buffer and throws it away. You cannot get it back.
9435 (This is not an interactive command.)
9438 @node search Exercises
9439 @section Searching Exercises
9443 Write an interactive function that searches for a string. If the
9444 search finds the string, leave point after it and display a message
9445 that says ``Found!''. (Do not use @code{search-forward} for the name
9446 of this function; if you do, you will overwrite the existing version of
9447 @code{search-forward} that comes with Emacs. Use a name such as
9448 @code{test-search} instead.)
9451 Write a function that prints the third element of the kill ring in the
9452 echo area, if any; if the kill ring does not contain a third element,
9453 print an appropriate message.
9456 @node List Implementation
9457 @chapter How Lists are Implemented
9458 @cindex Lists in a computer
9460 In Lisp, atoms are recorded in a straightforward fashion; if the
9461 implementation is not straightforward in practice, it is, nonetheless,
9462 straightforward in theory. The atom @samp{rose}, for example, is
9463 recorded as the four contiguous letters @samp{r}, @samp{o}, @samp{s},
9464 @samp{e}. A list, on the other hand, is kept differently. The mechanism
9465 is equally simple, but it takes a moment to get used to the idea. A
9466 list is kept using a series of pairs of pointers. In the series, the
9467 first pointer in each pair points to an atom or to another list, and the
9468 second pointer in each pair points to the next pair, or to the symbol
9469 @code{nil}, which marks the end of the list.
9471 A pointer itself is quite simply the electronic address of what is
9472 pointed to. Hence, a list is kept as a series of electronic addresses.
9475 * Lists diagrammed::
9476 * Symbols as Chest:: Exploring a powerful metaphor.
9481 @node Lists diagrammed
9482 @unnumberedsec Lists diagrammed
9485 For example, the list @code{(rose violet buttercup)} has three elements,
9486 @samp{rose}, @samp{violet}, and @samp{buttercup}. In the computer, the
9487 electronic address of @samp{rose} is recorded in a segment of computer
9488 memory along with the address that gives the electronic address of where
9489 the atom @samp{violet} is located; and that address (the one that tells
9490 where @samp{violet} is located) is kept along with an address that tells
9491 where the address for the atom @samp{buttercup} is located.
9494 This sounds more complicated than it is and is easier seen in a diagram:
9496 @c clear print-postscript-figures
9497 @c !!! cons-cell-diagram #1
9501 ___ ___ ___ ___ ___ ___
9502 |___|___|--> |___|___|--> |___|___|--> nil
9505 --> rose --> violet --> buttercup
9509 @ifset print-postscript-figures
9512 @center @image{cons-1}
9516 @ifclear print-postscript-figures
9520 ___ ___ ___ ___ ___ ___
9521 |___|___|--> |___|___|--> |___|___|--> nil
9524 --> rose --> violet --> buttercup
9531 In the diagram, each box represents a word of computer memory that
9532 holds a Lisp object, usually in the form of a memory address. The boxes,
9533 i.e., the addresses, are in pairs. Each arrow points to what the address
9534 is the address of, either an atom or another pair of addresses. The
9535 first box is the electronic address of @samp{rose} and the arrow points
9536 to @samp{rose}; the second box is the address of the next pair of boxes,
9537 the first part of which is the address of @samp{violet} and the second
9538 part of which is the address of the next pair. The very last box
9539 points to the symbol @code{nil}, which marks the end of the list.
9542 When a variable is set to a list with a function such as @code{setq},
9543 it stores the address of the first box in the variable. Thus,
9544 evaluation of the expression
9547 (setq bouquet '(rose violet buttercup))
9552 creates a situation like this:
9554 @c cons-cell-diagram #2
9560 | ___ ___ ___ ___ ___ ___
9561 --> |___|___|--> |___|___|--> |___|___|--> nil
9564 --> rose --> violet --> buttercup
9568 @ifset print-postscript-figures
9571 @center @image{cons-2}
9575 @ifclear print-postscript-figures
9581 | ___ ___ ___ ___ ___ ___
9582 --> |___|___|--> |___|___|--> |___|___|--> nil
9585 --> rose --> violet --> buttercup
9592 In this example, the symbol @code{bouquet} holds the address of the first
9596 This same list can be illustrated in a different sort of box notation
9599 @c cons-cell-diagram #2a
9605 | -------------- --------------- ----------------
9606 | | car | cdr | | car | cdr | | car | cdr |
9607 -->| rose | o------->| violet | o------->| butter- | nil |
9608 | | | | | | | cup | |
9609 -------------- --------------- ----------------
9613 @ifset print-postscript-figures
9616 @center @image{cons-2a}
9620 @ifclear print-postscript-figures
9626 | -------------- --------------- ----------------
9627 | | car | cdr | | car | cdr | | car | cdr |
9628 -->| rose | o------->| violet | o------->| butter- | nil |
9629 | | | | | | | cup | |
9630 -------------- --------------- ----------------
9636 (Symbols consist of more than pairs of addresses, but the structure of
9637 a symbol is made up of addresses. Indeed, the symbol @code{bouquet}
9638 consists of a group of address-boxes, one of which is the address of
9639 the printed word @samp{bouquet}, a second of which is the address of a
9640 function definition attached to the symbol, if any, a third of which
9641 is the address of the first pair of address-boxes for the list
9642 @code{(rose violet buttercup)}, and so on. Here we are showing that
9643 the symbol's third address-box points to the first pair of
9644 address-boxes for the list.)
9646 If a symbol is set to the @sc{cdr} of a list, the list itself is not
9647 changed; the symbol simply has an address further down the list. (In
9648 the jargon, @sc{car} and @sc{cdr} are `non-destructive'.) Thus,
9649 evaluation of the following expression
9652 (setq flowers (cdr bouquet))
9659 @c cons-cell-diagram #3
9666 | ___ ___ | ___ ___ ___ ___
9667 --> | | | --> | | | | | |
9668 |___|___|----> |___|___|--> |___|___|--> nil
9671 --> rose --> violet --> buttercup
9676 @ifset print-postscript-figures
9679 @center @image{cons-3}
9683 @ifclear print-postscript-figures
9690 | ___ ___ | ___ ___ ___ ___
9691 --> | | | --> | | | | | |
9692 |___|___|----> |___|___|--> |___|___|--> nil
9695 --> rose --> violet --> buttercup
9703 The value of @code{flowers} is @code{(violet buttercup)}, which is
9704 to say, the symbol @code{flowers} holds the address of the pair of
9705 address-boxes, the first of which holds the address of @code{violet},
9706 and the second of which holds the address of @code{buttercup}.
9708 A pair of address-boxes is called a @dfn{cons cell} or @dfn{dotted
9709 pair}. @xref{Cons Cell Type, , Cons Cell and List Types, elisp, The GNU Emacs Lisp
9710 Reference Manual}, and @ref{Dotted Pair Notation, , Dotted Pair
9711 Notation, elisp, The GNU Emacs Lisp Reference Manual}, for more
9712 information about cons cells and dotted pairs.
9715 The function @code{cons} adds a new pair of addresses to the front of
9716 a series of addresses like that shown above. For example, evaluating
9720 (setq bouquet (cons 'lily bouquet))
9727 @c cons-cell-diagram #4
9734 | ___ ___ ___ ___ | ___ ___ ___ ___
9735 --> | | | | | | --> | | | | | |
9736 |___|___|----> |___|___|----> |___|___|---->|___|___|--> nil
9739 --> lily --> rose --> violet --> buttercup
9744 @ifset print-postscript-figures
9747 @center @image{cons-4}
9751 @ifclear print-postscript-figures
9758 | ___ ___ ___ ___ | ___ ___ ___ ___
9759 --> | | | | | | --> | | | | | |
9760 |___|___|----> |___|___|----> |___|___|---->|___|___|--> nil
9763 --> lily --> rose --> violet --> buttercup
9772 However, this does not change the value of the symbol
9773 @code{flowers}, as you can see by evaluating the following,
9776 (eq (cdr (cdr bouquet)) flowers)
9780 which returns @code{t} for true.
9782 Until it is reset, @code{flowers} still has the value
9783 @code{(violet buttercup)}; that is, it has the address of the cons
9784 cell whose first address is of @code{violet}. Also, this does not
9785 alter any of the pre-existing cons cells; they are all still there.
9787 Thus, in Lisp, to get the @sc{cdr} of a list, you just get the address
9788 of the next cons cell in the series; to get the @sc{car} of a list,
9789 you get the address of the first element of the list; to @code{cons} a
9790 new element on a list, you add a new cons cell to the front of the list.
9791 That is all there is to it! The underlying structure of Lisp is
9794 And what does the last address in a series of cons cells refer to? It
9795 is the address of the empty list, of @code{nil}.
9797 In summary, when a Lisp variable is set to a value, it is provided with
9798 the address of the list to which the variable refers.
9800 @node Symbols as Chest
9801 @section Symbols as a Chest of Drawers
9802 @cindex Symbols as a Chest of Drawers
9803 @cindex Chest of Drawers, metaphor for a symbol
9804 @cindex Drawers, Chest of, metaphor for a symbol
9806 In an earlier section, I suggested that you might imagine a symbol as
9807 being a chest of drawers. The function definition is put in one
9808 drawer, the value in another, and so on. What is put in the drawer
9809 holding the value can be changed without affecting the contents of the
9810 drawer holding the function definition, and vice-verse.
9812 Actually, what is put in each drawer is the address of the value or
9813 function definition. It is as if you found an old chest in the attic,
9814 and in one of its drawers you found a map giving you directions to
9815 where the buried treasure lies.
9817 (In addition to its name, symbol definition, and variable value, a
9818 symbol has a `drawer' for a @dfn{property list} which can be used to
9819 record other information. Property lists are not discussed here; see
9820 @ref{Property Lists, , Property Lists, elisp, The GNU Emacs Lisp
9824 Here is a fanciful representation:
9826 @c chest-of-drawers diagram
9831 Chest of Drawers Contents of Drawers
9835 ---------------------
9836 | directions to | [map to]
9837 | symbol name | bouquet
9839 +---------------------+
9841 | symbol definition | [none]
9843 +---------------------+
9844 | directions to | [map to]
9845 | variable value | (rose violet buttercup)
9847 +---------------------+
9849 | property list | [not described here]
9851 +---------------------+
9857 @ifset print-postscript-figures
9860 @center @image{drawers}
9864 @ifclear print-postscript-figures
9869 Chest of Drawers Contents of Drawers
9873 ---------------------
9874 | directions to | [map to]
9875 | symbol name | bouquet
9877 +---------------------+
9879 | symbol definition | [none]
9881 +---------------------+
9882 | directions to | [map to]
9883 | variable value | (rose violet buttercup)
9885 +---------------------+
9887 | property list | [not described here]
9889 +---------------------+
9900 Set @code{flowers} to @code{violet} and @code{buttercup}. Cons two
9901 more flowers on to this list and set this new list to
9902 @code{more-flowers}. Set the @sc{car} of @code{flowers} to a fish.
9903 What does the @code{more-flowers} list now contain?
9906 @chapter Yanking Text Back
9908 @cindex Text retrieval
9909 @cindex Retrieving text
9910 @cindex Pasting text
9912 Whenever you cut text out of a buffer with a `kill' command in GNU Emacs,
9913 you can bring it back with a `yank' command. The text that is cut out of
9914 the buffer is put in the kill ring and the yank commands insert the
9915 appropriate contents of the kill ring back into a buffer (not necessarily
9916 the original buffer).
9918 A simple @kbd{C-y} (@code{yank}) command inserts the first item from
9919 the kill ring into the current buffer. If the @kbd{C-y} command is
9920 followed immediately by @kbd{M-y}, the first element is replaced by
9921 the second element. Successive @kbd{M-y} commands replace the second
9922 element with the third, fourth, or fifth element, and so on. When the
9923 last element in the kill ring is reached, it is replaced by the first
9924 element and the cycle is repeated. (Thus the kill ring is called a
9925 `ring' rather than just a `list'. However, the actual data structure
9926 that holds the text is a list.
9927 @xref{Kill Ring, , Handling the Kill Ring}, for the details of how the
9928 list is handled as a ring.)
9931 * Kill Ring Overview::
9932 * kill-ring-yank-pointer:: The kill ring is a list.
9933 * yank nthcdr Exercises:: The @code{kill-ring-yank-pointer} variable.
9936 @node Kill Ring Overview
9937 @section Kill Ring Overview
9938 @cindex Kill ring overview
9940 The kill ring is a list of textual strings. This is what it looks like:
9943 ("some text" "a different piece of text" "yet more text")
9946 If this were the contents of my kill ring and I pressed @kbd{C-y}, the
9947 string of characters saying @samp{some text} would be inserted in this
9948 buffer where my cursor is located.
9950 The @code{yank} command is also used for duplicating text by copying it.
9951 The copied text is not cut from the buffer, but a copy of it is put on the
9952 kill ring and is inserted by yanking it back.
9954 Three functions are used for bringing text back from the kill ring:
9955 @code{yank}, which is usually bound to @kbd{C-y}; @code{yank-pop},
9956 which is usually bound to @kbd{M-y}; and @code{rotate-yank-pointer},
9957 which is used by the two other functions.
9959 These functions refer to the kill ring through a variable called the
9960 @code{kill-ring-yank-pointer}. Indeed, the insertion code for both the
9961 @code{yank} and @code{yank-pop} functions is:
9964 (insert (car kill-ring-yank-pointer))
9968 (Well, no more. In GNU Emacs 22, the function has been replaced by
9969 @code{insert-for-yank} which calls @code{insert-for-yank-1}
9970 repetitively for each @code{yank-handler} segment. In turn,
9971 @code{insert-for-yank-1} strips text properties from the inserted text
9972 according to @code{yank-excluded-properties}. Otherwise, it is just
9973 like @code{insert}. We will stick with plain @code{insert} since it
9974 is easier to understand.)
9976 To begin to understand how @code{yank} and @code{yank-pop} work, it is
9977 first necessary to look at the @code{kill-ring-yank-pointer} variable.
9979 @node kill-ring-yank-pointer
9980 @section The @code{kill-ring-yank-pointer} Variable
9982 @code{kill-ring-yank-pointer} is a variable, just as @code{kill-ring} is
9983 a variable. It points to something by being bound to the value of what
9984 it points to, like any other Lisp variable.
9987 Thus, if the value of the kill ring is:
9990 ("some text" "a different piece of text" "yet more text")
9995 and the @code{kill-ring-yank-pointer} points to the second clause, the
9996 value of @code{kill-ring-yank-pointer} is:
9999 ("a different piece of text" "yet more text")
10002 As explained in the previous chapter (@pxref{List Implementation}), the
10003 computer does not keep two different copies of the text being pointed to
10004 by both the @code{kill-ring} and the @code{kill-ring-yank-pointer}. The
10005 words ``a different piece of text'' and ``yet more text'' are not
10006 duplicated. Instead, the two Lisp variables point to the same pieces of
10007 text. Here is a diagram:
10009 @c cons-cell-diagram #5
10013 kill-ring kill-ring-yank-pointer
10015 | ___ ___ | ___ ___ ___ ___
10016 ---> | | | --> | | | | | |
10017 |___|___|----> |___|___|--> |___|___|--> nil
10020 | | --> "yet more text"
10022 | --> "a different piece of text"
10029 @ifset print-postscript-figures
10032 @center @image{cons-5}
10036 @ifclear print-postscript-figures
10040 kill-ring kill-ring-yank-pointer
10042 | ___ ___ | ___ ___ ___ ___
10043 ---> | | | --> | | | | | |
10044 |___|___|----> |___|___|--> |___|___|--> nil
10047 | | --> "yet more text"
10049 | --> "a different piece of text
10058 Both the variable @code{kill-ring} and the variable
10059 @code{kill-ring-yank-pointer} are pointers. But the kill ring itself is
10060 usually described as if it were actually what it is composed of. The
10061 @code{kill-ring} is spoken of as if it were the list rather than that it
10062 points to the list. Conversely, the @code{kill-ring-yank-pointer} is
10063 spoken of as pointing to a list.
10065 These two ways of talking about the same thing sound confusing at first but
10066 make sense on reflection. The kill ring is generally thought of as the
10067 complete structure of data that holds the information of what has recently
10068 been cut out of the Emacs buffers. The @code{kill-ring-yank-pointer}
10069 on the other hand, serves to indicate---that is, to `point to'---that part
10070 of the kill ring of which the first element (the @sc{car}) will be
10074 In GNU Emacs 22, the @code{kill-new} function calls
10076 @code{(setq kill-ring-yank-pointer kill-ring)}
10078 (defun rotate-yank-pointer (arg)
10079 "Rotate the yanking point in the kill ring.
10080 With argument, rotate that many kills forward (or backward, if negative)."
10082 (current-kill arg))
10084 (defun current-kill (n &optional do-not-move)
10085 "Rotate the yanking point by N places, and then return that kill.
10086 If N is zero, `interprogram-paste-function' is set, and calling it
10087 returns a string, then that string is added to the front of the
10088 kill ring and returned as the latest kill.
10089 If optional arg DO-NOT-MOVE is non-nil, then don't actually move the
10090 yanking point; just return the Nth kill forward."
10091 (let ((interprogram-paste (and (= n 0)
10092 interprogram-paste-function
10093 (funcall interprogram-paste-function))))
10094 (if interprogram-paste
10096 ;; Disable the interprogram cut function when we add the new
10097 ;; text to the kill ring, so Emacs doesn't try to own the
10098 ;; selection, with identical text.
10099 (let ((interprogram-cut-function nil))
10100 (kill-new interprogram-paste))
10101 interprogram-paste)
10102 (or kill-ring (error "Kill ring is empty"))
10103 (let ((ARGth-kill-element
10104 (nthcdr (mod (- n (length kill-ring-yank-pointer))
10105 (length kill-ring))
10108 (setq kill-ring-yank-pointer ARGth-kill-element))
10109 (car ARGth-kill-element)))))
10114 @node yank nthcdr Exercises
10115 @section Exercises with @code{yank} and @code{nthcdr}
10119 Using @kbd{C-h v} (@code{describe-variable}), look at the value of
10120 your kill ring. Add several items to your kill ring; look at its
10121 value again. Using @kbd{M-y} (@code{yank-pop)}, move all the way
10122 around the kill ring. How many items were in your kill ring? Find
10123 the value of @code{kill-ring-max}. Was your kill ring full, or could
10124 you have kept more blocks of text within it?
10127 Using @code{nthcdr} and @code{car}, construct a series of expressions
10128 to return the first, second, third, and fourth elements of a list.
10131 @node Loops & Recursion
10132 @chapter Loops and Recursion
10133 @cindex Loops and recursion
10134 @cindex Recursion and loops
10135 @cindex Repetition (loops)
10137 Emacs Lisp has two primary ways to cause an expression, or a series of
10138 expressions, to be evaluated repeatedly: one uses a @code{while}
10139 loop, and the other uses @dfn{recursion}.
10141 Repetition can be very valuable. For example, to move forward four
10142 sentences, you need only write a program that will move forward one
10143 sentence and then repeat the process four times. Since a computer does
10144 not get bored or tired, such repetitive action does not have the
10145 deleterious effects that excessive or the wrong kinds of repetition can
10148 People mostly write Emacs Lisp functions using @code{while} loops and
10149 their kin; but you can use recursion, which provides a very powerful
10150 way to think about and then to solve problems@footnote{You can write
10151 recursive functions to be frugal or wasteful of mental or computer
10152 resources; as it happens, methods that people find easy---that are
10153 frugal of `mental resources'---sometimes use considerable computer
10154 resources. Emacs was designed to run on machines that we now consider
10155 limited and its default settings are conservative. You may want to
10156 increase the values of @code{max-specpdl-size} and
10157 @code{max-lisp-eval-depth}. In my @file{.emacs} file, I set them to
10158 15 and 30 times their default value.}.
10161 * while:: Causing a stretch of code to repeat.
10163 * Recursion:: Causing a function to call itself.
10164 * Looping exercise::
10168 @section @code{while}
10172 The @code{while} special form tests whether the value returned by
10173 evaluating its first argument is true or false. This is similar to what
10174 the Lisp interpreter does with an @code{if}; what the interpreter does
10175 next, however, is different.
10177 In a @code{while} expression, if the value returned by evaluating the
10178 first argument is false, the Lisp interpreter skips the rest of the
10179 expression (the @dfn{body} of the expression) and does not evaluate it.
10180 However, if the value is true, the Lisp interpreter evaluates the body
10181 of the expression and then again tests whether the first argument to
10182 @code{while} is true or false. If the value returned by evaluating the
10183 first argument is again true, the Lisp interpreter again evaluates the
10184 body of the expression.
10187 The template for a @code{while} expression looks like this:
10191 (while @var{true-or-false-test}
10197 * Looping with while:: Repeat so long as test returns true.
10198 * Loop Example:: A @code{while} loop that uses a list.
10199 * print-elements-of-list:: Uses @code{while}, @code{car}, @code{cdr}.
10200 * Incrementing Loop:: A loop with an incrementing counter.
10201 * Incrementing Loop Details::
10202 * Decrementing Loop:: A loop with a decrementing counter.
10206 @node Looping with while
10207 @unnumberedsubsec Looping with @code{while}
10210 So long as the true-or-false-test of the @code{while} expression
10211 returns a true value when it is evaluated, the body is repeatedly
10212 evaluated. This process is called a loop since the Lisp interpreter
10213 repeats the same thing again and again, like an airplane doing a loop.
10214 When the result of evaluating the true-or-false-test is false, the
10215 Lisp interpreter does not evaluate the rest of the @code{while}
10216 expression and `exits the loop'.
10218 Clearly, if the value returned by evaluating the first argument to
10219 @code{while} is always true, the body following will be evaluated
10220 again and again @dots{} and again @dots{} forever. Conversely, if the
10221 value returned is never true, the expressions in the body will never
10222 be evaluated. The craft of writing a @code{while} loop consists of
10223 choosing a mechanism such that the true-or-false-test returns true
10224 just the number of times that you want the subsequent expressions to
10225 be evaluated, and then have the test return false.
10227 The value returned by evaluating a @code{while} is the value of the
10228 true-or-false-test. An interesting consequence of this is that a
10229 @code{while} loop that evaluates without error will return @code{nil}
10230 or false regardless of whether it has looped 1 or 100 times or none at
10231 all. A @code{while} expression that evaluates successfully never
10232 returns a true value! What this means is that @code{while} is always
10233 evaluated for its side effects, which is to say, the consequences of
10234 evaluating the expressions within the body of the @code{while} loop.
10235 This makes sense. It is not the mere act of looping that is desired,
10236 but the consequences of what happens when the expressions in the loop
10237 are repeatedly evaluated.
10240 @subsection A @code{while} Loop and a List
10242 A common way to control a @code{while} loop is to test whether a list
10243 has any elements. If it does, the loop is repeated; but if it does not,
10244 the repetition is ended. Since this is an important technique, we will
10245 create a short example to illustrate it.
10247 A simple way to test whether a list has elements is to evaluate the
10248 list: if it has no elements, it is an empty list and will return the
10249 empty list, @code{()}, which is a synonym for @code{nil} or false. On
10250 the other hand, a list with elements will return those elements when it
10251 is evaluated. Since Emacs Lisp considers as true any value that is not
10252 @code{nil}, a list that returns elements will test true in a
10256 For example, you can set the variable @code{empty-list} to @code{nil} by
10257 evaluating the following @code{setq} expression:
10260 (setq empty-list ())
10264 After evaluating the @code{setq} expression, you can evaluate the
10265 variable @code{empty-list} in the usual way, by placing the cursor after
10266 the symbol and typing @kbd{C-x C-e}; @code{nil} will appear in your
10273 On the other hand, if you set a variable to be a list with elements, the
10274 list will appear when you evaluate the variable, as you can see by
10275 evaluating the following two expressions:
10279 (setq animals '(gazelle giraffe lion tiger))
10285 Thus, to create a @code{while} loop that tests whether there are any
10286 items in the list @code{animals}, the first part of the loop will be
10297 When the @code{while} tests its first argument, the variable
10298 @code{animals} is evaluated. It returns a list. So long as the list
10299 has elements, the @code{while} considers the results of the test to be
10300 true; but when the list is empty, it considers the results of the test
10303 To prevent the @code{while} loop from running forever, some mechanism
10304 needs to be provided to empty the list eventually. An oft-used
10305 technique is to have one of the subsequent forms in the @code{while}
10306 expression set the value of the list to be the @sc{cdr} of the list.
10307 Each time the @code{cdr} function is evaluated, the list will be made
10308 shorter, until eventually only the empty list will be left. At this
10309 point, the test of the @code{while} loop will return false, and the
10310 arguments to the @code{while} will no longer be evaluated.
10312 For example, the list of animals bound to the variable @code{animals}
10313 can be set to be the @sc{cdr} of the original list with the
10314 following expression:
10317 (setq animals (cdr animals))
10321 If you have evaluated the previous expressions and then evaluate this
10322 expression, you will see @code{(giraffe lion tiger)} appear in the echo
10323 area. If you evaluate the expression again, @code{(lion tiger)} will
10324 appear in the echo area. If you evaluate it again and yet again,
10325 @code{(tiger)} appears and then the empty list, shown by @code{nil}.
10327 A template for a @code{while} loop that uses the @code{cdr} function
10328 repeatedly to cause the true-or-false-test eventually to test false
10333 (while @var{test-whether-list-is-empty}
10335 @var{set-list-to-cdr-of-list})
10339 This test and use of @code{cdr} can be put together in a function that
10340 goes through a list and prints each element of the list on a line of its
10343 @node print-elements-of-list
10344 @subsection An Example: @code{print-elements-of-list}
10345 @findex print-elements-of-list
10347 The @code{print-elements-of-list} function illustrates a @code{while}
10350 @cindex @file{*scratch*} buffer
10351 The function requires several lines for its output. If you are
10352 reading this in a recent instance of GNU Emacs,
10353 @c GNU Emacs 21, GNU Emacs 22, or a later version,
10354 you can evaluate the following expression inside of Info, as usual.
10356 If you are using an earlier version of Emacs, you need to copy the
10357 necessary expressions to your @file{*scratch*} buffer and evaluate
10358 them there. This is because the echo area had only one line in the
10361 You can copy the expressions by marking the beginning of the region
10362 with @kbd{C-@key{SPC}} (@code{set-mark-command}), moving the cursor to
10363 the end of the region and then copying the region using @kbd{M-w}
10364 (@code{kill-ring-save}, which calls @code{copy-region-as-kill} and
10365 then provides visual feedback). In the @file{*scratch*}
10366 buffer, you can yank the expressions back by typing @kbd{C-y}
10369 After you have copied the expressions to the @file{*scratch*} buffer,
10370 evaluate each expression in turn. Be sure to evaluate the last
10371 expression, @code{(print-elements-of-list animals)}, by typing
10372 @kbd{C-u C-x C-e}, that is, by giving an argument to
10373 @code{eval-last-sexp}. This will cause the result of the evaluation
10374 to be printed in the @file{*scratch*} buffer instead of being printed
10375 in the echo area. (Otherwise you will see something like this in your
10376 echo area: @code{^Jgazelle^J^Jgiraffe^J^Jlion^J^Jtiger^Jnil}, in which
10377 each @samp{^J} stands for a `newline'.)
10380 In a recent instance of GNU Emacs, you can evaluate these expressions
10381 directly in the Info buffer, and the echo area will grow to show the
10386 (setq animals '(gazelle giraffe lion tiger))
10388 (defun print-elements-of-list (list)
10389 "Print each element of LIST on a line of its own."
10392 (setq list (cdr list))))
10394 (print-elements-of-list animals)
10400 When you evaluate the three expressions in sequence, you will see
10416 Each element of the list is printed on a line of its own (that is what
10417 the function @code{print} does) and then the value returned by the
10418 function is printed. Since the last expression in the function is the
10419 @code{while} loop, and since @code{while} loops always return
10420 @code{nil}, a @code{nil} is printed after the last element of the list.
10422 @node Incrementing Loop
10423 @subsection A Loop with an Incrementing Counter
10425 A loop is not useful unless it stops when it ought. Besides
10426 controlling a loop with a list, a common way of stopping a loop is to
10427 write the first argument as a test that returns false when the correct
10428 number of repetitions are complete. This means that the loop must
10429 have a counter---an expression that counts how many times the loop
10433 @node Incrementing Loop Details
10434 @unnumberedsubsec Details of an Incrementing Loop
10437 The test for a loop with an incrementing counter can be an expression
10438 such as @code{(< count desired-number)} which returns @code{t} for
10439 true if the value of @code{count} is less than the
10440 @code{desired-number} of repetitions and @code{nil} for false if the
10441 value of @code{count} is equal to or is greater than the
10442 @code{desired-number}. The expression that increments the count can
10443 be a simple @code{setq} such as @code{(setq count (1+ count))}, where
10444 @code{1+} is a built-in function in Emacs Lisp that adds 1 to its
10445 argument. (The expression @w{@code{(1+ count)}} has the same result
10446 as @w{@code{(+ count 1)}}, but is easier for a human to read.)
10449 The template for a @code{while} loop controlled by an incrementing
10450 counter looks like this:
10454 @var{set-count-to-initial-value}
10455 (while (< count desired-number) ; @r{true-or-false-test}
10457 (setq count (1+ count))) ; @r{incrementer}
10462 Note that you need to set the initial value of @code{count}; usually it
10466 * Incrementing Example:: Counting pebbles in a triangle.
10467 * Inc Example parts:: The parts of the function definition.
10468 * Inc Example altogether:: Putting the function definition together.
10471 @node Incrementing Example
10472 @unnumberedsubsubsec Example with incrementing counter
10474 Suppose you are playing on the beach and decide to make a triangle of
10475 pebbles, putting one pebble in the first row, two in the second row,
10476 three in the third row and so on, like this:
10494 @bullet{} @bullet{}
10495 @bullet{} @bullet{} @bullet{}
10496 @bullet{} @bullet{} @bullet{} @bullet{}
10503 (About 2500 years ago, Pythagoras and others developed the beginnings of
10504 number theory by considering questions such as this.)
10506 Suppose you want to know how many pebbles you will need to make a
10507 triangle with 7 rows?
10509 Clearly, what you need to do is add up the numbers from 1 to 7. There
10510 are two ways to do this; start with the smallest number, one, and add up
10511 the list in sequence, 1, 2, 3, 4 and so on; or start with the largest
10512 number and add the list going down: 7, 6, 5, 4 and so on. Because both
10513 mechanisms illustrate common ways of writing @code{while} loops, we will
10514 create two examples, one counting up and the other counting down. In
10515 this first example, we will start with 1 and add 2, 3, 4 and so on.
10517 If you are just adding up a short list of numbers, the easiest way to do
10518 it is to add up all the numbers at once. However, if you do not know
10519 ahead of time how many numbers your list will have, or if you want to be
10520 prepared for a very long list, then you need to design your addition so
10521 that what you do is repeat a simple process many times instead of doing
10522 a more complex process once.
10524 For example, instead of adding up all the pebbles all at once, what you
10525 can do is add the number of pebbles in the first row, 1, to the number
10526 in the second row, 2, and then add the total of those two rows to the
10527 third row, 3. Then you can add the number in the fourth row, 4, to the
10528 total of the first three rows; and so on.
10530 The critical characteristic of the process is that each repetitive
10531 action is simple. In this case, at each step we add only two numbers,
10532 the number of pebbles in the row and the total already found. This
10533 process of adding two numbers is repeated again and again until the last
10534 row has been added to the total of all the preceding rows. In a more
10535 complex loop the repetitive action might not be so simple, but it will
10536 be simpler than doing everything all at once.
10538 @node Inc Example parts
10539 @unnumberedsubsubsec The parts of the function definition
10541 The preceding analysis gives us the bones of our function definition:
10542 first, we will need a variable that we can call @code{total} that will
10543 be the total number of pebbles. This will be the value returned by
10546 Second, we know that the function will require an argument: this
10547 argument will be the total number of rows in the triangle. It can be
10548 called @code{number-of-rows}.
10550 Finally, we need a variable to use as a counter. We could call this
10551 variable @code{counter}, but a better name is @code{row-number}. That
10552 is because what the counter does in this function is count rows, and a
10553 program should be written to be as understandable as possible.
10555 When the Lisp interpreter first starts evaluating the expressions in the
10556 function, the value of @code{total} should be set to zero, since we have
10557 not added anything to it. Then the function should add the number of
10558 pebbles in the first row to the total, and then add the number of
10559 pebbles in the second to the total, and then add the number of
10560 pebbles in the third row to the total, and so on, until there are no
10561 more rows left to add.
10563 Both @code{total} and @code{row-number} are used only inside the
10564 function, so they can be declared as local variables with @code{let}
10565 and given initial values. Clearly, the initial value for @code{total}
10566 should be 0. The initial value of @code{row-number} should be 1,
10567 since we start with the first row. This means that the @code{let}
10568 statement will look like this:
10578 After the internal variables are declared and bound to their initial
10579 values, we can begin the @code{while} loop. The expression that serves
10580 as the test should return a value of @code{t} for true so long as the
10581 @code{row-number} is less than or equal to the @code{number-of-rows}.
10582 (If the expression tests true only so long as the row number is less
10583 than the number of rows in the triangle, the last row will never be
10584 added to the total; hence the row number has to be either less than or
10585 equal to the number of rows.)
10588 @findex <= @r{(less than or equal)}
10589 Lisp provides the @code{<=} function that returns true if the value of
10590 its first argument is less than or equal to the value of its second
10591 argument and false otherwise. So the expression that the @code{while}
10592 will evaluate as its test should look like this:
10595 (<= row-number number-of-rows)
10598 The total number of pebbles can be found by repeatedly adding the number
10599 of pebbles in a row to the total already found. Since the number of
10600 pebbles in the row is equal to the row number, the total can be found by
10601 adding the row number to the total. (Clearly, in a more complex
10602 situation, the number of pebbles in the row might be related to the row
10603 number in a more complicated way; if this were the case, the row number
10604 would be replaced by the appropriate expression.)
10607 (setq total (+ total row-number))
10611 What this does is set the new value of @code{total} to be equal to the
10612 sum of adding the number of pebbles in the row to the previous total.
10614 After setting the value of @code{total}, the conditions need to be
10615 established for the next repetition of the loop, if there is one. This
10616 is done by incrementing the value of the @code{row-number} variable,
10617 which serves as a counter. After the @code{row-number} variable has
10618 been incremented, the true-or-false-test at the beginning of the
10619 @code{while} loop tests whether its value is still less than or equal to
10620 the value of the @code{number-of-rows} and if it is, adds the new value
10621 of the @code{row-number} variable to the @code{total} of the previous
10622 repetition of the loop.
10625 The built-in Emacs Lisp function @code{1+} adds 1 to a number, so the
10626 @code{row-number} variable can be incremented with this expression:
10629 (setq row-number (1+ row-number))
10632 @node Inc Example altogether
10633 @unnumberedsubsubsec Putting the function definition together
10635 We have created the parts for the function definition; now we need to
10639 First, the contents of the @code{while} expression:
10643 (while (<= row-number number-of-rows) ; @r{true-or-false-test}
10644 (setq total (+ total row-number))
10645 (setq row-number (1+ row-number))) ; @r{incrementer}
10649 Along with the @code{let} expression varlist, this very nearly
10650 completes the body of the function definition. However, it requires
10651 one final element, the need for which is somewhat subtle.
10653 The final touch is to place the variable @code{total} on a line by
10654 itself after the @code{while} expression. Otherwise, the value returned
10655 by the whole function is the value of the last expression that is
10656 evaluated in the body of the @code{let}, and this is the value
10657 returned by the @code{while}, which is always @code{nil}.
10659 This may not be evident at first sight. It almost looks as if the
10660 incrementing expression is the last expression of the whole function.
10661 But that expression is part of the body of the @code{while}; it is the
10662 last element of the list that starts with the symbol @code{while}.
10663 Moreover, the whole of the @code{while} loop is a list within the body
10667 In outline, the function will look like this:
10671 (defun @var{name-of-function} (@var{argument-list})
10672 "@var{documentation}@dots{}"
10673 (let (@var{varlist})
10674 (while (@var{true-or-false-test})
10675 @var{body-of-while}@dots{} )
10676 @dots{} )) ; @r{Need final expression here.}
10680 The result of evaluating the @code{let} is what is going to be returned
10681 by the @code{defun} since the @code{let} is not embedded within any
10682 containing list, except for the @code{defun} as a whole. However, if
10683 the @code{while} is the last element of the @code{let} expression, the
10684 function will always return @code{nil}. This is not what we want!
10685 Instead, what we want is the value of the variable @code{total}. This
10686 is returned by simply placing the symbol as the last element of the list
10687 starting with @code{let}. It gets evaluated after the preceding
10688 elements of the list are evaluated, which means it gets evaluated after
10689 it has been assigned the correct value for the total.
10691 It may be easier to see this by printing the list starting with
10692 @code{let} all on one line. This format makes it evident that the
10693 @var{varlist} and @code{while} expressions are the second and third
10694 elements of the list starting with @code{let}, and the @code{total} is
10699 (let (@var{varlist}) (while (@var{true-or-false-test}) @var{body-of-while}@dots{} ) total)
10704 Putting everything together, the @code{triangle} function definition
10709 (defun triangle (number-of-rows) ; @r{Version with}
10710 ; @r{ incrementing counter.}
10711 "Add up the number of pebbles in a triangle.
10712 The first row has one pebble, the second row two pebbles,
10713 the third row three pebbles, and so on.
10714 The argument is NUMBER-OF-ROWS."
10719 (while (<= row-number number-of-rows)
10720 (setq total (+ total row-number))
10721 (setq row-number (1+ row-number)))
10727 After you have installed @code{triangle} by evaluating the function, you
10728 can try it out. Here are two examples:
10739 The sum of the first four numbers is 10 and the sum of the first seven
10742 @node Decrementing Loop
10743 @subsection Loop with a Decrementing Counter
10745 Another common way to write a @code{while} loop is to write the test
10746 so that it determines whether a counter is greater than zero. So long
10747 as the counter is greater than zero, the loop is repeated. But when
10748 the counter is equal to or less than zero, the loop is stopped. For
10749 this to work, the counter has to start out greater than zero and then
10750 be made smaller and smaller by a form that is evaluated
10753 The test will be an expression such as @code{(> counter 0)} which
10754 returns @code{t} for true if the value of @code{counter} is greater
10755 than zero, and @code{nil} for false if the value of @code{counter} is
10756 equal to or less than zero. The expression that makes the number
10757 smaller and smaller can be a simple @code{setq} such as @code{(setq
10758 counter (1- counter))}, where @code{1-} is a built-in function in
10759 Emacs Lisp that subtracts 1 from its argument.
10762 The template for a decrementing @code{while} loop looks like this:
10766 (while (> counter 0) ; @r{true-or-false-test}
10768 (setq counter (1- counter))) ; @r{decrementer}
10773 * Decrementing Example:: More pebbles on the beach.
10774 * Dec Example parts:: The parts of the function definition.
10775 * Dec Example altogether:: Putting the function definition together.
10778 @node Decrementing Example
10779 @unnumberedsubsubsec Example with decrementing counter
10781 To illustrate a loop with a decrementing counter, we will rewrite the
10782 @code{triangle} function so the counter decreases to zero.
10784 This is the reverse of the earlier version of the function. In this
10785 case, to find out how many pebbles are needed to make a triangle with
10786 3 rows, add the number of pebbles in the third row, 3, to the number
10787 in the preceding row, 2, and then add the total of those two rows to
10788 the row that precedes them, which is 1.
10790 Likewise, to find the number of pebbles in a triangle with 7 rows, add
10791 the number of pebbles in the seventh row, 7, to the number in the
10792 preceding row, which is 6, and then add the total of those two rows to
10793 the row that precedes them, which is 5, and so on. As in the previous
10794 example, each addition only involves adding two numbers, the total of
10795 the rows already added up and the number of pebbles in the row that is
10796 being added to the total. This process of adding two numbers is
10797 repeated again and again until there are no more pebbles to add.
10799 We know how many pebbles to start with: the number of pebbles in the
10800 last row is equal to the number of rows. If the triangle has seven
10801 rows, the number of pebbles in the last row is 7. Likewise, we know how
10802 many pebbles are in the preceding row: it is one less than the number in
10805 @node Dec Example parts
10806 @unnumberedsubsubsec The parts of the function definition
10808 We start with three variables: the total number of rows in the
10809 triangle; the number of pebbles in a row; and the total number of
10810 pebbles, which is what we want to calculate. These variables can be
10811 named @code{number-of-rows}, @code{number-of-pebbles-in-row}, and
10812 @code{total}, respectively.
10814 Both @code{total} and @code{number-of-pebbles-in-row} are used only
10815 inside the function and are declared with @code{let}. The initial
10816 value of @code{total} should, of course, be zero. However, the
10817 initial value of @code{number-of-pebbles-in-row} should be equal to
10818 the number of rows in the triangle, since the addition will start with
10822 This means that the beginning of the @code{let} expression will look
10828 (number-of-pebbles-in-row number-of-rows))
10833 The total number of pebbles can be found by repeatedly adding the number
10834 of pebbles in a row to the total already found, that is, by repeatedly
10835 evaluating the following expression:
10838 (setq total (+ total number-of-pebbles-in-row))
10842 After the @code{number-of-pebbles-in-row} is added to the @code{total},
10843 the @code{number-of-pebbles-in-row} should be decremented by one, since
10844 the next time the loop repeats, the preceding row will be
10845 added to the total.
10847 The number of pebbles in a preceding row is one less than the number of
10848 pebbles in a row, so the built-in Emacs Lisp function @code{1-} can be
10849 used to compute the number of pebbles in the preceding row. This can be
10850 done with the following expression:
10854 (setq number-of-pebbles-in-row
10855 (1- number-of-pebbles-in-row))
10859 Finally, we know that the @code{while} loop should stop making repeated
10860 additions when there are no pebbles in a row. So the test for
10861 the @code{while} loop is simply:
10864 (while (> number-of-pebbles-in-row 0)
10867 @node Dec Example altogether
10868 @unnumberedsubsubsec Putting the function definition together
10870 We can put these expressions together to create a function definition
10871 that works. However, on examination, we find that one of the local
10872 variables is unneeded!
10875 The function definition looks like this:
10879 ;;; @r{First subtractive version.}
10880 (defun triangle (number-of-rows)
10881 "Add up the number of pebbles in a triangle."
10883 (number-of-pebbles-in-row number-of-rows))
10884 (while (> number-of-pebbles-in-row 0)
10885 (setq total (+ total number-of-pebbles-in-row))
10886 (setq number-of-pebbles-in-row
10887 (1- number-of-pebbles-in-row)))
10892 As written, this function works.
10894 However, we do not need @code{number-of-pebbles-in-row}.
10896 @cindex Argument as local variable
10897 When the @code{triangle} function is evaluated, the symbol
10898 @code{number-of-rows} will be bound to a number, giving it an initial
10899 value. That number can be changed in the body of the function as if
10900 it were a local variable, without any fear that such a change will
10901 effect the value of the variable outside of the function. This is a
10902 very useful characteristic of Lisp; it means that the variable
10903 @code{number-of-rows} can be used anywhere in the function where
10904 @code{number-of-pebbles-in-row} is used.
10907 Here is a second version of the function written a bit more cleanly:
10911 (defun triangle (number) ; @r{Second version.}
10912 "Return sum of numbers 1 through NUMBER inclusive."
10914 (while (> number 0)
10915 (setq total (+ total number))
10916 (setq number (1- number)))
10921 In brief, a properly written @code{while} loop will consist of three parts:
10925 A test that will return false after the loop has repeated itself the
10926 correct number of times.
10929 An expression the evaluation of which will return the value desired
10930 after being repeatedly evaluated.
10933 An expression to change the value passed to the true-or-false-test so
10934 that the test returns false after the loop has repeated itself the right
10938 @node dolist dotimes
10939 @section Save your time: @code{dolist} and @code{dotimes}
10941 In addition to @code{while}, both @code{dolist} and @code{dotimes}
10942 provide for looping. Sometimes these are quicker to write than the
10943 equivalent @code{while} loop. Both are Lisp macros. (@xref{Macros, ,
10944 Macros, elisp, The GNU Emacs Lisp Reference Manual}. )
10946 @code{dolist} works like a @code{while} loop that `@sc{cdr}s down a
10947 list': @code{dolist} automatically shortens the list each time it
10948 loops---takes the @sc{cdr} of the list---and binds the @sc{car} of
10949 each shorter version of the list to the first of its arguments.
10951 @code{dotimes} loops a specific number of times: you specify the number.
10959 @unnumberedsubsec The @code{dolist} Macro
10962 Suppose, for example, you want to reverse a list, so that
10963 ``first'' ``second'' ``third'' becomes ``third'' ``second'' ``first''.
10966 In practice, you would use the @code{reverse} function, like this:
10970 (setq animals '(gazelle giraffe lion tiger))
10978 Here is how you could reverse the list using a @code{while} loop:
10982 (setq animals '(gazelle giraffe lion tiger))
10984 (defun reverse-list-with-while (list)
10985 "Using while, reverse the order of LIST."
10986 (let (value) ; make sure list starts empty
10988 (setq value (cons (car list) value))
10989 (setq list (cdr list)))
10992 (reverse-list-with-while animals)
10998 And here is how you could use the @code{dolist} macro:
11002 (setq animals '(gazelle giraffe lion tiger))
11004 (defun reverse-list-with-dolist (list)
11005 "Using dolist, reverse the order of LIST."
11006 (let (value) ; make sure list starts empty
11007 (dolist (element list value)
11008 (setq value (cons element value)))))
11010 (reverse-list-with-dolist animals)
11016 In Info, you can place your cursor after the closing parenthesis of
11017 each expression and type @kbd{C-x C-e}; in each case, you should see
11020 (tiger lion giraffe gazelle)
11026 For this example, the existing @code{reverse} function is obviously best.
11027 The @code{while} loop is just like our first example (@pxref{Loop
11028 Example, , A @code{while} Loop and a List}). The @code{while} first
11029 checks whether the list has elements; if so, it constructs a new list
11030 by adding the first element of the list to the existing list (which in
11031 the first iteration of the loop is @code{nil}). Since the second
11032 element is prepended in front of the first element, and the third
11033 element is prepended in front of the second element, the list is reversed.
11035 In the expression using a @code{while} loop,
11036 the @w{@code{(setq list (cdr list))}}
11037 expression shortens the list, so the @code{while} loop eventually
11038 stops. In addition, it provides the @code{cons} expression with a new
11039 first element by creating a new and shorter list at each repetition of
11042 The @code{dolist} expression does very much the same as the
11043 @code{while} expression, except that the @code{dolist} macro does some
11044 of the work you have to do when writing a @code{while} expression.
11046 Like a @code{while} loop, a @code{dolist} loops. What is different is
11047 that it automatically shortens the list each time it loops---it
11048 `@sc{cdr}s down the list' on its own---and it automatically binds
11049 the @sc{car} of each shorter version of the list to the first of its
11052 In the example, the @sc{car} of each shorter version of the list is
11053 referred to using the symbol @samp{element}, the list itself is called
11054 @samp{list}, and the value returned is called @samp{value}. The
11055 remainder of the @code{dolist} expression is the body.
11057 The @code{dolist} expression binds the @sc{car} of each shorter
11058 version of the list to @code{element} and then evaluates the body of
11059 the expression; and repeats the loop. The result is returned in
11063 @unnumberedsubsec The @code{dotimes} Macro
11066 The @code{dotimes} macro is similar to @code{dolist}, except that it
11067 loops a specific number of times.
11069 The first argument to @code{dotimes} is assigned the numbers 0, 1, 2
11070 and so forth each time around the loop, and the value of the third
11071 argument is returned. You need to provide the value of the second
11072 argument, which is how many times the macro loops.
11075 For example, the following binds the numbers from 0 up to, but not
11076 including, the number 3 to the first argument, @var{number}, and then
11077 constructs a list of the three numbers. (The first number is 0, the
11078 second number is 1, and the third number is 2; this makes a total of
11079 three numbers in all, starting with zero as the first number.)
11083 (let (value) ; otherwise a value is a void variable
11084 (dotimes (number 3 value)
11085 (setq value (cons number value))))
11092 @code{dotimes} returns @code{value}, so the way to use
11093 @code{dotimes} is to operate on some expression @var{number} number of
11094 times and then return the result, either as a list or an atom.
11097 Here is an example of a @code{defun} that uses @code{dotimes} to add
11098 up the number of pebbles in a triangle.
11102 (defun triangle-using-dotimes (number-of-rows)
11103 "Using dotimes, add up the number of pebbles in a triangle."
11104 (let ((total 0)) ; otherwise a total is a void variable
11105 (dotimes (number number-of-rows total)
11106 (setq total (+ total (1+ number))))))
11108 (triangle-using-dotimes 4)
11116 A recursive function contains code that tells the Lisp interpreter to
11117 call a program that runs exactly like itself, but with slightly
11118 different arguments. The code runs exactly the same because it has
11119 the same name. However, even though the program has the same name, it
11120 is not the same entity. It is different. In the jargon, it is a
11121 different `instance'.
11123 Eventually, if the program is written correctly, the `slightly
11124 different arguments' will become sufficiently different from the first
11125 arguments that the final instance will stop.
11128 * Building Robots:: Same model, different serial number ...
11129 * Recursive Definition Parts:: Walk until you stop ...
11130 * Recursion with list:: Using a list as the test whether to recurse.
11131 * Recursive triangle function::
11132 * Recursion with cond::
11133 * Recursive Patterns:: Often used templates.
11134 * No Deferment:: Don't store up work ...
11135 * No deferment solution::
11138 @node Building Robots
11139 @subsection Building Robots: Extending the Metaphor
11140 @cindex Building robots
11141 @cindex Robots, building
11143 It is sometimes helpful to think of a running program as a robot that
11144 does a job. In doing its job, a recursive function calls on a second
11145 robot to help it. The second robot is identical to the first in every
11146 way, except that the second robot helps the first and has been
11147 passed different arguments than the first.
11149 In a recursive function, the second robot may call a third; and the
11150 third may call a fourth, and so on. Each of these is a different
11151 entity; but all are clones.
11153 Since each robot has slightly different instructions---the arguments
11154 will differ from one robot to the next---the last robot should know
11157 Let's expand on the metaphor in which a computer program is a robot.
11159 A function definition provides the blueprints for a robot. When you
11160 install a function definition, that is, when you evaluate a
11161 @code{defun} macro, you install the necessary equipment to build
11162 robots. It is as if you were in a factory, setting up an assembly
11163 line. Robots with the same name are built according to the same
11164 blueprints. So they have, as it were, the same `model number', but a
11165 different `serial number'.
11167 We often say that a recursive function `calls itself'. What we mean
11168 is that the instructions in a recursive function cause the Lisp
11169 interpreter to run a different function that has the same name and
11170 does the same job as the first, but with different arguments.
11172 It is important that the arguments differ from one instance to the
11173 next; otherwise, the process will never stop.
11175 @node Recursive Definition Parts
11176 @subsection The Parts of a Recursive Definition
11177 @cindex Parts of a Recursive Definition
11178 @cindex Recursive Definition Parts
11180 A recursive function typically contains a conditional expression which
11185 A true-or-false-test that determines whether the function is called
11186 again, here called the @dfn{do-again-test}.
11189 The name of the function. When this name is called, a new instance of
11190 the function---a new robot, as it were---is created and told what to do.
11193 An expression that returns a different value each time the function is
11194 called, here called the @dfn{next-step-expression}. Consequently, the
11195 argument (or arguments) passed to the new instance of the function
11196 will be different from that passed to the previous instance. This
11197 causes the conditional expression, the @dfn{do-again-test}, to test
11198 false after the correct number of repetitions.
11201 Recursive functions can be much simpler than any other kind of
11202 function. Indeed, when people first start to use them, they often look
11203 so mysteriously simple as to be incomprehensible. Like riding a
11204 bicycle, reading a recursive function definition takes a certain knack
11205 which is hard at first but then seems simple.
11208 There are several different common recursive patterns. A very simple
11209 pattern looks like this:
11213 (defun @var{name-of-recursive-function} (@var{argument-list})
11214 "@var{documentation}@dots{}"
11215 (if @var{do-again-test}
11217 (@var{name-of-recursive-function}
11218 @var{next-step-expression})))
11222 Each time a recursive function is evaluated, a new instance of it is
11223 created and told what to do. The arguments tell the instance what to do.
11225 An argument is bound to the value of the next-step-expression. Each
11226 instance runs with a different value of the next-step-expression.
11228 The value in the next-step-expression is used in the do-again-test.
11230 The value returned by the next-step-expression is passed to the new
11231 instance of the function, which evaluates it (or some
11232 transmogrification of it) to determine whether to continue or stop.
11233 The next-step-expression is designed so that the do-again-test returns
11234 false when the function should no longer be repeated.
11236 The do-again-test is sometimes called the @dfn{stop condition},
11237 since it stops the repetitions when it tests false.
11239 @node Recursion with list
11240 @subsection Recursion with a List
11242 The example of a @code{while} loop that printed the elements of a list
11243 of numbers can be written recursively. Here is the code, including
11244 an expression to set the value of the variable @code{animals} to a list.
11246 If you are reading this in Info in Emacs, you can evaluate this
11247 expression directly in Info. Otherwise, you must copy the example
11248 to the @file{*scratch*} buffer and evaluate each expression there.
11249 Use @kbd{C-u C-x C-e} to evaluate the
11250 @code{(print-elements-recursively animals)} expression so that the
11251 results are printed in the buffer; otherwise the Lisp interpreter will
11252 try to squeeze the results into the one line of the echo area.
11254 Also, place your cursor immediately after the last closing parenthesis
11255 of the @code{print-elements-recursively} function, before the comment.
11256 Otherwise, the Lisp interpreter will try to evaluate the comment.
11258 @findex print-elements-recursively
11261 (setq animals '(gazelle giraffe lion tiger))
11263 (defun print-elements-recursively (list)
11264 "Print each element of LIST on a line of its own.
11266 (when list ; @r{do-again-test}
11267 (print (car list)) ; @r{body}
11268 (print-elements-recursively ; @r{recursive call}
11269 (cdr list)))) ; @r{next-step-expression}
11271 (print-elements-recursively animals)
11275 The @code{print-elements-recursively} function first tests whether
11276 there is any content in the list; if there is, the function prints the
11277 first element of the list, the @sc{car} of the list. Then the
11278 function `invokes itself', but gives itself as its argument, not the
11279 whole list, but the second and subsequent elements of the list, the
11280 @sc{cdr} of the list.
11282 Put another way, if the list is not empty, the function invokes
11283 another instance of code that is similar to the initial code, but is a
11284 different thread of execution, with different arguments than the first
11287 Put in yet another way, if the list is not empty, the first robot
11288 assembles a second robot and tells it what to do; the second robot is
11289 a different individual from the first, but is the same model.
11291 When the second evaluation occurs, the @code{when} expression is
11292 evaluated and if true, prints the first element of the list it
11293 receives as its argument (which is the second element of the original
11294 list). Then the function `calls itself' with the @sc{cdr} of the list
11295 it is invoked with, which (the second time around) is the @sc{cdr} of
11296 the @sc{cdr} of the original list.
11298 Note that although we say that the function `calls itself', what we
11299 mean is that the Lisp interpreter assembles and instructs a new
11300 instance of the program. The new instance is a clone of the first,
11301 but is a separate individual.
11303 Each time the function `invokes itself', it invokes itself on a
11304 shorter version of the original list. It creates a new instance that
11305 works on a shorter list.
11307 Eventually, the function invokes itself on an empty list. It creates
11308 a new instance whose argument is @code{nil}. The conditional expression
11309 tests the value of @code{list}. Since the value of @code{list} is
11310 @code{nil}, the @code{when} expression tests false so the then-part is
11311 not evaluated. The function as a whole then returns @code{nil}.
11314 When you evaluate the expression @code{(print-elements-recursively
11315 animals)} in the @file{*scratch*} buffer, you see this result:
11331 @node Recursive triangle function
11332 @subsection Recursion in Place of a Counter
11333 @findex triangle-recursively
11336 The @code{triangle} function described in a previous section can also
11337 be written recursively. It looks like this:
11341 (defun triangle-recursively (number)
11342 "Return the sum of the numbers 1 through NUMBER inclusive.
11344 (if (= number 1) ; @r{do-again-test}
11346 (+ number ; @r{else-part}
11347 (triangle-recursively ; @r{recursive call}
11348 (1- number))))) ; @r{next-step-expression}
11350 (triangle-recursively 7)
11355 You can install this function by evaluating it and then try it by
11356 evaluating @code{(triangle-recursively 7)}. (Remember to put your
11357 cursor immediately after the last parenthesis of the function
11358 definition, before the comment.) The function evaluates to 28.
11360 To understand how this function works, let's consider what happens in the
11361 various cases when the function is passed 1, 2, 3, or 4 as the value of
11365 * Recursive Example arg of 1 or 2::
11366 * Recursive Example arg of 3 or 4::
11370 @node Recursive Example arg of 1 or 2
11371 @unnumberedsubsubsec An argument of 1 or 2
11374 First, what happens if the value of the argument is 1?
11376 The function has an @code{if} expression after the documentation
11377 string. It tests whether the value of @code{number} is equal to 1; if
11378 so, Emacs evaluates the then-part of the @code{if} expression, which
11379 returns the number 1 as the value of the function. (A triangle with
11380 one row has one pebble in it.)
11382 Suppose, however, that the value of the argument is 2. In this case,
11383 Emacs evaluates the else-part of the @code{if} expression.
11386 The else-part consists of an addition, the recursive call to
11387 @code{triangle-recursively} and a decrementing action; and it looks like
11391 (+ number (triangle-recursively (1- number)))
11394 When Emacs evaluates this expression, the innermost expression is
11395 evaluated first; then the other parts in sequence. Here are the steps
11399 @item Step 1 @w{ } Evaluate the innermost expression.
11401 The innermost expression is @code{(1- number)} so Emacs decrements the
11402 value of @code{number} from 2 to 1.
11404 @item Step 2 @w{ } Evaluate the @code{triangle-recursively} function.
11406 The Lisp interpreter creates an individual instance of
11407 @code{triangle-recursively}. It does not matter that this function is
11408 contained within itself. Emacs passes the result Step 1 as the
11409 argument used by this instance of the @code{triangle-recursively}
11412 In this case, Emacs evaluates @code{triangle-recursively} with an
11413 argument of 1. This means that this evaluation of
11414 @code{triangle-recursively} returns 1.
11416 @item Step 3 @w{ } Evaluate the value of @code{number}.
11418 The variable @code{number} is the second element of the list that
11419 starts with @code{+}; its value is 2.
11421 @item Step 4 @w{ } Evaluate the @code{+} expression.
11423 The @code{+} expression receives two arguments, the first
11424 from the evaluation of @code{number} (Step 3) and the second from the
11425 evaluation of @code{triangle-recursively} (Step 2).
11427 The result of the addition is the sum of 2 plus 1, and the number 3 is
11428 returned, which is correct. A triangle with two rows has three
11432 @node Recursive Example arg of 3 or 4
11433 @unnumberedsubsubsec An argument of 3 or 4
11435 Suppose that @code{triangle-recursively} is called with an argument of
11439 @item Step 1 @w{ } Evaluate the do-again-test.
11441 The @code{if} expression is evaluated first. This is the do-again
11442 test and returns false, so the else-part of the @code{if} expression
11443 is evaluated. (Note that in this example, the do-again-test causes
11444 the function to call itself when it tests false, not when it tests
11447 @item Step 2 @w{ } Evaluate the innermost expression of the else-part.
11449 The innermost expression of the else-part is evaluated, which decrements
11450 3 to 2. This is the next-step-expression.
11452 @item Step 3 @w{ } Evaluate the @code{triangle-recursively} function.
11454 The number 2 is passed to the @code{triangle-recursively} function.
11456 We already know what happens when Emacs evaluates @code{triangle-recursively} with
11457 an argument of 2. After going through the sequence of actions described
11458 earlier, it returns a value of 3. So that is what will happen here.
11460 @item Step 4 @w{ } Evaluate the addition.
11462 3 will be passed as an argument to the addition and will be added to the
11463 number with which the function was called, which is 3.
11467 The value returned by the function as a whole will be 6.
11469 Now that we know what will happen when @code{triangle-recursively} is
11470 called with an argument of 3, it is evident what will happen if it is
11471 called with an argument of 4:
11475 In the recursive call, the evaluation of
11478 (triangle-recursively (1- 4))
11483 will return the value of evaluating
11486 (triangle-recursively 3)
11490 which is 6 and this value will be added to 4 by the addition in the
11495 The value returned by the function as a whole will be 10.
11497 Each time @code{triangle-recursively} is evaluated, it evaluates a
11498 version of itself---a different instance of itself---with a smaller
11499 argument, until the argument is small enough so that it does not
11502 Note that this particular design for a recursive function
11503 requires that operations be deferred.
11505 Before @code{(triangle-recursively 7)} can calculate its answer, it
11506 must call @code{(triangle-recursively 6)}; and before
11507 @code{(triangle-recursively 6)} can calculate its answer, it must call
11508 @code{(triangle-recursively 5)}; and so on. That is to say, the
11509 calculation that @code{(triangle-recursively 7)} makes must be
11510 deferred until @code{(triangle-recursively 6)} makes its calculation;
11511 and @code{(triangle-recursively 6)} must defer until
11512 @code{(triangle-recursively 5)} completes; and so on.
11514 If each of these instances of @code{triangle-recursively} are thought
11515 of as different robots, the first robot must wait for the second to
11516 complete its job, which must wait until the third completes, and so
11519 There is a way around this kind of waiting, which we will discuss in
11520 @ref{No Deferment, , Recursion without Deferments}.
11522 @node Recursion with cond
11523 @subsection Recursion Example Using @code{cond}
11526 The version of @code{triangle-recursively} described earlier is written
11527 with the @code{if} special form. It can also be written using another
11528 special form called @code{cond}. The name of the special form
11529 @code{cond} is an abbreviation of the word @samp{conditional}.
11531 Although the @code{cond} special form is not used as often in the
11532 Emacs Lisp sources as @code{if}, it is used often enough to justify
11536 The template for a @code{cond} expression looks like this:
11546 where the @var{body} is a series of lists.
11549 Written out more fully, the template looks like this:
11554 (@var{first-true-or-false-test} @var{first-consequent})
11555 (@var{second-true-or-false-test} @var{second-consequent})
11556 (@var{third-true-or-false-test} @var{third-consequent})
11561 When the Lisp interpreter evaluates the @code{cond} expression, it
11562 evaluates the first element (the @sc{car} or true-or-false-test) of
11563 the first expression in a series of expressions within the body of the
11566 If the true-or-false-test returns @code{nil} the rest of that
11567 expression, the consequent, is skipped and the true-or-false-test of the
11568 next expression is evaluated. When an expression is found whose
11569 true-or-false-test returns a value that is not @code{nil}, the
11570 consequent of that expression is evaluated. The consequent can be one
11571 or more expressions. If the consequent consists of more than one
11572 expression, the expressions are evaluated in sequence and the value of
11573 the last one is returned. If the expression does not have a consequent,
11574 the value of the true-or-false-test is returned.
11576 If none of the true-or-false-tests test true, the @code{cond} expression
11577 returns @code{nil}.
11580 Written using @code{cond}, the @code{triangle} function looks like this:
11584 (defun triangle-using-cond (number)
11585 (cond ((<= number 0) 0)
11588 (+ number (triangle-using-cond (1- number))))))
11593 In this example, the @code{cond} returns 0 if the number is less than or
11594 equal to 0, it returns 1 if the number is 1 and it evaluates @code{(+
11595 number (triangle-using-cond (1- number)))} if the number is greater than
11598 @node Recursive Patterns
11599 @subsection Recursive Patterns
11600 @cindex Recursive Patterns
11602 Here are three common recursive patterns. Each involves a list.
11603 Recursion does not need to involve lists, but Lisp is designed for lists
11604 and this provides a sense of its primal capabilities.
11613 @unnumberedsubsubsec Recursive Pattern: @emph{every}
11614 @cindex Every, type of recursive pattern
11615 @cindex Recursive pattern: every
11617 In the @code{every} recursive pattern, an action is performed on every
11621 The basic pattern is:
11625 If a list be empty, return @code{nil}.
11627 Else, act on the beginning of the list (the @sc{car} of the list)
11630 through a recursive call by the function on the rest (the
11631 @sc{cdr}) of the list,
11633 and, optionally, combine the acted-on element, using @code{cons},
11634 with the results of acting on the rest.
11643 (defun square-each (numbers-list)
11644 "Square each of a NUMBERS LIST, recursively."
11645 (if (not numbers-list) ; do-again-test
11648 (* (car numbers-list) (car numbers-list))
11649 (square-each (cdr numbers-list))))) ; next-step-expression
11653 (square-each '(1 2 3))
11660 If @code{numbers-list} is empty, do nothing. But if it has content,
11661 construct a list combining the square of the first number in the list
11662 with the result of the recursive call.
11664 (The example follows the pattern exactly: @code{nil} is returned if
11665 the numbers' list is empty. In practice, you would write the
11666 conditional so it carries out the action when the numbers' list is not
11669 The @code{print-elements-recursively} function (@pxref{Recursion with
11670 list, , Recursion with a List}) is another example of an @code{every}
11671 pattern, except in this case, rather than bring the results together
11672 using @code{cons}, we print each element of output.
11675 The @code{print-elements-recursively} function looks like this:
11679 (setq animals '(gazelle giraffe lion tiger))
11683 (defun print-elements-recursively (list)
11684 "Print each element of LIST on a line of its own.
11686 (when list ; @r{do-again-test}
11687 (print (car list)) ; @r{body}
11688 (print-elements-recursively ; @r{recursive call}
11689 (cdr list)))) ; @r{next-step-expression}
11691 (print-elements-recursively animals)
11696 The pattern for @code{print-elements-recursively} is:
11700 When the list is empty, do nothing.
11702 But when the list has at least one element,
11705 act on the beginning of the list (the @sc{car} of the list),
11707 and make a recursive call on the rest (the @sc{cdr}) of the list.
11712 @unnumberedsubsubsec Recursive Pattern: @emph{accumulate}
11713 @cindex Accumulate, type of recursive pattern
11714 @cindex Recursive pattern: accumulate
11716 Another recursive pattern is called the @code{accumulate} pattern. In
11717 the @code{accumulate} recursive pattern, an action is performed on
11718 every element of a list and the result of that action is accumulated
11719 with the results of performing the action on the other elements.
11721 This is very like the `every' pattern using @code{cons}, except that
11722 @code{cons} is not used, but some other combiner.
11729 If a list be empty, return zero or some other constant.
11731 Else, act on the beginning of the list (the @sc{car} of the list),
11734 and combine that acted-on element, using @code{+} or
11735 some other combining function, with
11737 a recursive call by the function on the rest (the @sc{cdr}) of the list.
11742 Here is an example:
11746 (defun add-elements (numbers-list)
11747 "Add the elements of NUMBERS-LIST together."
11748 (if (not numbers-list)
11750 (+ (car numbers-list) (add-elements (cdr numbers-list)))))
11754 (add-elements '(1 2 3 4))
11759 @xref{Files List, , Making a List of Files}, for an example of the
11760 accumulate pattern.
11763 @unnumberedsubsubsec Recursive Pattern: @emph{keep}
11764 @cindex Keep, type of recursive pattern
11765 @cindex Recursive pattern: keep
11767 A third recursive pattern is called the @code{keep} pattern.
11768 In the @code{keep} recursive pattern, each element of a list is tested;
11769 the element is acted on and the results are kept only if the element
11772 Again, this is very like the `every' pattern, except the element is
11773 skipped unless it meets a criterion.
11776 The pattern has three parts:
11780 If a list be empty, return @code{nil}.
11782 Else, if the beginning of the list (the @sc{car} of the list) passes
11786 act on that element and combine it, using @code{cons} with
11788 a recursive call by the function on the rest (the @sc{cdr}) of the list.
11791 Otherwise, if the beginning of the list (the @sc{car} of the list) fails
11795 skip on that element,
11797 and, recursively call the function on the rest (the @sc{cdr}) of the list.
11802 Here is an example that uses @code{cond}:
11806 (defun keep-three-letter-words (word-list)
11807 "Keep three letter words in WORD-LIST."
11809 ;; First do-again-test: stop-condition
11810 ((not word-list) nil)
11812 ;; Second do-again-test: when to act
11813 ((eq 3 (length (symbol-name (car word-list))))
11814 ;; combine acted-on element with recursive call on shorter list
11815 (cons (car word-list) (keep-three-letter-words (cdr word-list))))
11817 ;; Third do-again-test: when to skip element;
11818 ;; recursively call shorter list with next-step expression
11819 (t (keep-three-letter-words (cdr word-list)))))
11823 (keep-three-letter-words '(one two three four five six))
11824 @result{} (one two six)
11828 It goes without saying that you need not use @code{nil} as the test for
11829 when to stop; and you can, of course, combine these patterns.
11832 @subsection Recursion without Deferments
11833 @cindex Deferment in recursion
11834 @cindex Recursion without Deferments
11836 Let's consider again what happens with the @code{triangle-recursively}
11837 function. We will find that the intermediate calculations are
11838 deferred until all can be done.
11841 Here is the function definition:
11845 (defun triangle-recursively (number)
11846 "Return the sum of the numbers 1 through NUMBER inclusive.
11848 (if (= number 1) ; @r{do-again-test}
11850 (+ number ; @r{else-part}
11851 (triangle-recursively ; @r{recursive call}
11852 (1- number))))) ; @r{next-step-expression}
11856 What happens when we call this function with a argument of 7?
11858 The first instance of the @code{triangle-recursively} function adds
11859 the number 7 to the value returned by a second instance of
11860 @code{triangle-recursively}, an instance that has been passed an
11861 argument of 6. That is to say, the first calculation is:
11864 (+ 7 (triangle-recursively 6))
11868 The first instance of @code{triangle-recursively}---you may want to
11869 think of it as a little robot---cannot complete its job. It must hand
11870 off the calculation for @code{(triangle-recursively 6)} to a second
11871 instance of the program, to a second robot. This second individual is
11872 completely different from the first one; it is, in the jargon, a
11873 `different instantiation'. Or, put another way, it is a different
11874 robot. It is the same model as the first; it calculates triangle
11875 numbers recursively; but it has a different serial number.
11877 And what does @code{(triangle-recursively 6)} return? It returns the
11878 number 6 added to the value returned by evaluating
11879 @code{triangle-recursively} with an argument of 5. Using the robot
11880 metaphor, it asks yet another robot to help it.
11886 (+ 7 6 (triangle-recursively 5))
11890 And what happens next?
11893 (+ 7 6 5 (triangle-recursively 4))
11896 Each time @code{triangle-recursively} is called, except for the last
11897 time, it creates another instance of the program---another robot---and
11898 asks it to make a calculation.
11901 Eventually, the full addition is set up and performed:
11907 This design for the function defers the calculation of the first step
11908 until the second can be done, and defers that until the third can be
11909 done, and so on. Each deferment means the computer must remember what
11910 is being waited on. This is not a problem when there are only a few
11911 steps, as in this example. But it can be a problem when there are
11914 @node No deferment solution
11915 @subsection No Deferment Solution
11916 @cindex No deferment solution
11917 @cindex Defermentless solution
11918 @cindex Solution without deferment
11920 The solution to the problem of deferred operations is to write in a
11921 manner that does not defer operations@footnote{The phrase @dfn{tail
11922 recursive} is used to describe such a process, one that uses
11923 `constant space'.}. This requires
11924 writing to a different pattern, often one that involves writing two
11925 function definitions, an `initialization' function and a `helper'
11928 The `initialization' function sets up the job; the `helper' function
11932 Here are the two function definitions for adding up numbers. They are
11933 so simple, I find them hard to understand.
11937 (defun triangle-initialization (number)
11938 "Return the sum of the numbers 1 through NUMBER inclusive.
11939 This is the `initialization' component of a two function
11940 duo that uses recursion."
11941 (triangle-recursive-helper 0 0 number))
11947 (defun triangle-recursive-helper (sum counter number)
11948 "Return SUM, using COUNTER, through NUMBER inclusive.
11949 This is the `helper' component of a two function duo
11950 that uses recursion."
11951 (if (> counter number)
11953 (triangle-recursive-helper (+ sum counter) ; @r{sum}
11954 (1+ counter) ; @r{counter}
11955 number))) ; @r{number}
11960 Install both function definitions by evaluating them, then call
11961 @code{triangle-initialization} with 2 rows:
11965 (triangle-initialization 2)
11970 The `initialization' function calls the first instance of the `helper'
11971 function with three arguments: zero, zero, and a number which is the
11972 number of rows in the triangle.
11974 The first two arguments passed to the `helper' function are
11975 initialization values. These values are changed when
11976 @code{triangle-recursive-helper} invokes new instances.@footnote{The
11977 jargon is mildly confusing: @code{triangle-recursive-helper} uses a
11978 process that is iterative in a procedure that is recursive. The
11979 process is called iterative because the computer need only record the
11980 three values, @code{sum}, @code{counter}, and @code{number}; the
11981 procedure is recursive because the function `calls itself'. On the
11982 other hand, both the process and the procedure used by
11983 @code{triangle-recursively} are called recursive. The word
11984 `recursive' has different meanings in the two contexts.}
11986 Let's see what happens when we have a triangle that has one row. (This
11987 triangle will have one pebble in it!)
11990 @code{triangle-initialization} will call its helper with
11991 the arguments @w{@code{0 0 1}}. That function will run the conditional
11992 test whether @code{(> counter number)}:
12000 and find that the result is false, so it will invoke
12001 the else-part of the @code{if} clause:
12005 (triangle-recursive-helper
12006 (+ sum counter) ; @r{sum plus counter} @result{} @r{sum}
12007 (1+ counter) ; @r{increment counter} @result{} @r{counter}
12008 number) ; @r{number stays the same}
12014 which will first compute:
12018 (triangle-recursive-helper (+ 0 0) ; @r{sum}
12019 (1+ 0) ; @r{counter}
12023 (triangle-recursive-helper 0 1 1)
12027 Again, @code{(> counter number)} will be false, so again, the Lisp
12028 interpreter will evaluate @code{triangle-recursive-helper}, creating a
12029 new instance with new arguments.
12032 This new instance will be;
12036 (triangle-recursive-helper
12037 (+ sum counter) ; @r{sum plus counter} @result{} @r{sum}
12038 (1+ counter) ; @r{increment counter} @result{} @r{counter}
12039 number) ; @r{number stays the same}
12043 (triangle-recursive-helper 1 2 1)
12047 In this case, the @code{(> counter number)} test will be true! So the
12048 instance will return the value of the sum, which will be 1, as
12051 Now, let's pass @code{triangle-initialization} an argument
12052 of 2, to find out how many pebbles there are in a triangle with two rows.
12054 That function calls @code{(triangle-recursive-helper 0 0 2)}.
12057 In stages, the instances called will be:
12061 @r{sum counter number}
12062 (triangle-recursive-helper 0 1 2)
12064 (triangle-recursive-helper 1 2 2)
12066 (triangle-recursive-helper 3 3 2)
12070 When the last instance is called, the @code{(> counter number)} test
12071 will be true, so the instance will return the value of @code{sum},
12074 This kind of pattern helps when you are writing functions that can use
12075 many resources in a computer.
12078 @node Looping exercise
12079 @section Looping Exercise
12083 Write a function similar to @code{triangle} in which each row has a
12084 value which is the square of the row number. Use a @code{while} loop.
12087 Write a function similar to @code{triangle} that multiplies instead of
12091 Rewrite these two functions recursively. Rewrite these functions
12094 @c comma in printed title causes problem in Info cross reference
12096 Write a function for Texinfo mode that creates an index entry at the
12097 beginning of a paragraph for every @samp{@@dfn} within the paragraph.
12098 (In a Texinfo file, @samp{@@dfn} marks a definition. This book is
12099 written in Texinfo.)
12101 Many of the functions you will need are described in two of the
12102 previous chapters, @ref{Cutting & Storing Text, , Cutting and Storing
12103 Text}, and @ref{Yanking, , Yanking Text Back}. If you use
12104 @code{forward-paragraph} to put the index entry at the beginning of
12105 the paragraph, you will have to use @w{@kbd{C-h f}}
12106 (@code{describe-function}) to find out how to make the command go
12109 For more information, see
12111 @ref{Indicating, , Indicating Definitions, texinfo}.
12114 @ref{Indicating, , Indicating, texinfo, Texinfo Manual}, which goes to
12115 a Texinfo manual in the current directory. Or, if you are on the
12117 @uref{http://www.gnu.org/software/texinfo/manual/texinfo/}
12120 ``Indicating Definitions, Commands, etc.'' in @cite{Texinfo, The GNU
12121 Documentation Format}.
12125 @node Regexp Search
12126 @chapter Regular Expression Searches
12127 @cindex Searches, illustrating
12128 @cindex Regular expression searches
12129 @cindex Patterns, searching for
12130 @cindex Motion by sentence and paragraph
12131 @cindex Sentences, movement by
12132 @cindex Paragraphs, movement by
12134 Regular expression searches are used extensively in GNU Emacs. The
12135 two functions, @code{forward-sentence} and @code{forward-paragraph},
12136 illustrate these searches well. They use regular expressions to find
12137 where to move point. The phrase `regular expression' is often written
12140 Regular expression searches are described in @ref{Regexp Search, ,
12141 Regular Expression Search, emacs, The GNU Emacs Manual}, as well as in
12142 @ref{Regular Expressions, , , elisp, The GNU Emacs Lisp Reference
12143 Manual}. In writing this chapter, I am presuming that you have at
12144 least a mild acquaintance with them. The major point to remember is
12145 that regular expressions permit you to search for patterns as well as
12146 for literal strings of characters. For example, the code in
12147 @code{forward-sentence} searches for the pattern of possible
12148 characters that could mark the end of a sentence, and moves point to
12151 Before looking at the code for the @code{forward-sentence} function, it
12152 is worth considering what the pattern that marks the end of a sentence
12153 must be. The pattern is discussed in the next section; following that
12154 is a description of the regular expression search function,
12155 @code{re-search-forward}. The @code{forward-sentence} function
12156 is described in the section following. Finally, the
12157 @code{forward-paragraph} function is described in the last section of
12158 this chapter. @code{forward-paragraph} is a complex function that
12159 introduces several new features.
12162 * sentence-end:: The regular expression for @code{sentence-end}.
12163 * re-search-forward:: Very similar to @code{search-forward}.
12164 * forward-sentence:: A straightforward example of regexp search.
12165 * forward-paragraph:: A somewhat complex example.
12166 * etags:: How to create your own @file{TAGS} table.
12168 * re-search Exercises::
12172 @section The Regular Expression for @code{sentence-end}
12173 @findex sentence-end
12175 The symbol @code{sentence-end} is bound to the pattern that marks the
12176 end of a sentence. What should this regular expression be?
12178 Clearly, a sentence may be ended by a period, a question mark, or an
12179 exclamation mark. Indeed, in English, only clauses that end with one
12180 of those three characters should be considered the end of a sentence.
12181 This means that the pattern should include the character set:
12187 However, we do not want @code{forward-sentence} merely to jump to a
12188 period, a question mark, or an exclamation mark, because such a character
12189 might be used in the middle of a sentence. A period, for example, is
12190 used after abbreviations. So other information is needed.
12192 According to convention, you type two spaces after every sentence, but
12193 only one space after a period, a question mark, or an exclamation mark in
12194 the body of a sentence. So a period, a question mark, or an exclamation
12195 mark followed by two spaces is a good indicator of an end of sentence.
12196 However, in a file, the two spaces may instead be a tab or the end of a
12197 line. This means that the regular expression should include these three
12198 items as alternatives.
12201 This group of alternatives will look like this:
12212 Here, @samp{$} indicates the end of the line, and I have pointed out
12213 where the tab and two spaces are inserted in the expression. Both are
12214 inserted by putting the actual characters into the expression.
12216 Two backslashes, @samp{\\}, are required before the parentheses and
12217 vertical bars: the first backslash quotes the following backslash in
12218 Emacs; and the second indicates that the following character, the
12219 parenthesis or the vertical bar, is special.
12222 Also, a sentence may be followed by one or more carriage returns, like
12233 Like tabs and spaces, a carriage return is inserted into a regular
12234 expression by inserting it literally. The asterisk indicates that the
12235 @key{RET} is repeated zero or more times.
12237 But a sentence end does not consist only of a period, a question mark or
12238 an exclamation mark followed by appropriate space: a closing quotation
12239 mark or a closing brace of some kind may precede the space. Indeed more
12240 than one such mark or brace may precede the space. These require a
12241 expression that looks like this:
12247 In this expression, the first @samp{]} is the first character in the
12248 expression; the second character is @samp{"}, which is preceded by a
12249 @samp{\} to tell Emacs the @samp{"} is @emph{not} special. The last
12250 three characters are @samp{'}, @samp{)}, and @samp{@}}.
12252 All this suggests what the regular expression pattern for matching the
12253 end of a sentence should be; and, indeed, if we evaluate
12254 @code{sentence-end} we find that it returns the following value:
12259 @result{} "[.?!][]\"')@}]*\\($\\| \\| \\)[
12265 (Well, not in GNU Emacs 22; that is because of an effort to make the
12266 process simpler and to handle more glyphs and languages. When the
12267 value of @code{sentence-end} is @code{nil}, then use the value defined
12268 by the function @code{sentence-end}. (Here is a use of the difference
12269 between a value and a function in Emacs Lisp.) The function returns a
12270 value constructed from the variables @code{sentence-end-base},
12271 @code{sentence-end-double-space}, @code{sentence-end-without-period},
12272 and @code{sentence-end-without-space}. The critical variable is
12273 @code{sentence-end-base}; its global value is similar to the one
12274 described above but it also contains two additional quotation marks.
12275 These have differing degrees of curliness. The
12276 @code{sentence-end-without-period} variable, when true, tells Emacs
12277 that a sentence may end without a period, such as text in Thai.)
12281 (Note that here the @key{TAB}, two spaces, and @key{RET} are shown
12282 literally in the pattern.)
12284 This regular expression can be deciphered as follows:
12288 The first part of the pattern is the three characters, a period, a question
12289 mark and an exclamation mark, within square brackets. The pattern must
12290 begin with one or other of these characters.
12293 The second part of the pattern is the group of closing braces and
12294 quotation marks, which can appear zero or more times. These may follow
12295 the period, question mark or exclamation mark. In a regular expression,
12296 the backslash, @samp{\}, followed by the double quotation mark,
12297 @samp{"}, indicates the class of string-quote characters. Usually, the
12298 double quotation mark is the only character in this class. The
12299 asterisk, @samp{*}, indicates that the items in the previous group (the
12300 group surrounded by square brackets, @samp{[]}) may be repeated zero or
12303 @item \\($\\| \\| \\)
12304 The third part of the pattern is one or other of: either the end of a
12305 line, or two blank spaces, or a tab. The double back-slashes are used
12306 to prevent Emacs from reading the parentheses and vertical bars as part
12307 of the search pattern; the parentheses are used to mark the group and
12308 the vertical bars are used to indicated that the patterns to either side
12309 of them are alternatives. The dollar sign is used to indicate the end
12310 of a line and both the two spaces and the tab are each inserted as is to
12311 indicate what they are.
12314 Finally, the last part of the pattern indicates that the end of the line
12315 or the whitespace following the period, question mark or exclamation
12316 mark may, but need not, be followed by one or more carriage returns. In
12317 the pattern, the carriage return is inserted as an actual carriage
12318 return between square brackets but here it is shown as @key{RET}.
12322 @node re-search-forward
12323 @section The @code{re-search-forward} Function
12324 @findex re-search-forward
12326 The @code{re-search-forward} function is very like the
12327 @code{search-forward} function. (@xref{search-forward, , The
12328 @code{search-forward} Function}.)
12330 @code{re-search-forward} searches for a regular expression. If the
12331 search is successful, it leaves point immediately after the last
12332 character in the target. If the search is backwards, it leaves point
12333 just before the first character in the target. You may tell
12334 @code{re-search-forward} to return @code{t} for true. (Moving point
12335 is therefore a `side effect'.)
12337 Like @code{search-forward}, the @code{re-search-forward} function takes
12342 The first argument is the regular expression that the function searches
12343 for. The regular expression will be a string between quotation marks.
12346 The optional second argument limits how far the function will search; it is a
12347 bound, which is specified as a position in the buffer.
12350 The optional third argument specifies how the function responds to
12351 failure: @code{nil} as the third argument causes the function to
12352 signal an error (and print a message) when the search fails; any other
12353 value causes it to return @code{nil} if the search fails and @code{t}
12354 if the search succeeds.
12357 The optional fourth argument is the repeat count. A negative repeat
12358 count causes @code{re-search-forward} to search backwards.
12362 The template for @code{re-search-forward} looks like this:
12366 (re-search-forward "@var{regular-expression}"
12367 @var{limit-of-search}
12368 @var{what-to-do-if-search-fails}
12369 @var{repeat-count})
12373 The second, third, and fourth arguments are optional. However, if you
12374 want to pass a value to either or both of the last two arguments, you
12375 must also pass a value to all the preceding arguments. Otherwise, the
12376 Lisp interpreter will mistake which argument you are passing the value
12380 In the @code{forward-sentence} function, the regular expression will be
12381 the value of the variable @code{sentence-end}. In simple form, that is:
12385 "[.?!][]\"')@}]*\\($\\| \\| \\)[
12391 The limit of the search will be the end of the paragraph (since a
12392 sentence cannot go beyond a paragraph). If the search fails, the
12393 function will return @code{nil}; and the repeat count will be provided
12394 by the argument to the @code{forward-sentence} function.
12396 @node forward-sentence
12397 @section @code{forward-sentence}
12398 @findex forward-sentence
12400 The command to move the cursor forward a sentence is a straightforward
12401 illustration of how to use regular expression searches in Emacs Lisp.
12402 Indeed, the function looks longer and more complicated than it is; this
12403 is because the function is designed to go backwards as well as forwards;
12404 and, optionally, over more than one sentence. The function is usually
12405 bound to the key command @kbd{M-e}.
12408 * Complete forward-sentence::
12409 * fwd-sentence while loops:: Two @code{while} loops.
12410 * fwd-sentence re-search:: A regular expression search.
12414 @node Complete forward-sentence
12415 @unnumberedsubsec Complete @code{forward-sentence} function definition
12419 Here is the code for @code{forward-sentence}:
12424 (defun forward-sentence (&optional arg)
12425 "Move forward to next `sentence-end'. With argument, repeat.
12426 With negative argument, move backward repeatedly to `sentence-beginning'.
12428 The variable `sentence-end' is a regular expression that matches ends of
12429 sentences. Also, every paragraph boundary terminates sentences as well."
12433 (or arg (setq arg 1))
12434 (let ((opoint (point))
12435 (sentence-end (sentence-end)))
12437 (let ((pos (point))
12438 (par-beg (save-excursion (start-of-paragraph-text) (point))))
12439 (if (and (re-search-backward sentence-end par-beg t)
12440 (or (< (match-end 0) pos)
12441 (re-search-backward sentence-end par-beg t)))
12442 (goto-char (match-end 0))
12443 (goto-char par-beg)))
12444 (setq arg (1+ arg)))
12448 (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
12449 (if (re-search-forward sentence-end par-end t)
12450 (skip-chars-backward " \t\n")
12451 (goto-char par-end)))
12452 (setq arg (1- arg)))
12453 (constrain-to-field nil opoint t)))
12461 (defun forward-sentence (&optional arg)
12462 "Move forward to next sentence-end. With argument, repeat.
12463 With negative argument, move backward repeatedly to sentence-beginning.
12464 Sentence ends are identified by the value of sentence-end
12465 treated as a regular expression. Also, every paragraph boundary
12466 terminates sentences as well."
12470 (or arg (setq arg 1))
12473 (save-excursion (start-of-paragraph-text) (point))))
12474 (if (re-search-backward
12475 (concat sentence-end "[^ \t\n]") par-beg t)
12476 (goto-char (1- (match-end 0)))
12477 (goto-char par-beg)))
12478 (setq arg (1+ arg)))
12481 (save-excursion (end-of-paragraph-text) (point))))
12482 (if (re-search-forward sentence-end par-end t)
12483 (skip-chars-backward " \t\n")
12484 (goto-char par-end)))
12485 (setq arg (1- arg))))
12490 The function looks long at first sight and it is best to look at its
12491 skeleton first, and then its muscle. The way to see the skeleton is to
12492 look at the expressions that start in the left-most columns:
12496 (defun forward-sentence (&optional arg)
12497 "@var{documentation}@dots{}"
12499 (or arg (setq arg 1))
12500 (let ((opoint (point)) (sentence-end (sentence-end)))
12502 (let ((pos (point))
12503 (par-beg (save-excursion (start-of-paragraph-text) (point))))
12504 @var{rest-of-body-of-while-loop-when-going-backwards}
12506 (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
12507 @var{rest-of-body-of-while-loop-when-going-forwards}
12508 @var{handle-forms-and-equivalent}
12512 This looks much simpler! The function definition consists of
12513 documentation, an @code{interactive} expression, an @code{or}
12514 expression, a @code{let} expression, and @code{while} loops.
12516 Let's look at each of these parts in turn.
12518 We note that the documentation is thorough and understandable.
12520 The function has an @code{interactive "p"} declaration. This means
12521 that the processed prefix argument, if any, is passed to the
12522 function as its argument. (This will be a number.) If the function
12523 is not passed an argument (it is optional) then the argument
12524 @code{arg} will be bound to 1.
12526 When @code{forward-sentence} is called non-interactively without an
12527 argument, @code{arg} is bound to @code{nil}. The @code{or} expression
12528 handles this. What it does is either leave the value of @code{arg} as
12529 it is, but only if @code{arg} is bound to a value; or it sets the
12530 value of @code{arg} to 1, in the case when @code{arg} is bound to
12533 Next is a @code{let}. That specifies the values of two local
12534 variables, @code{point} and @code{sentence-end}. The local value of
12535 point, from before the search, is used in the
12536 @code{constrain-to-field} function which handles forms and
12537 equivalents. The @code{sentence-end} variable is set by the
12538 @code{sentence-end} function.
12540 @node fwd-sentence while loops
12541 @unnumberedsubsec The @code{while} loops
12543 Two @code{while} loops follow. The first @code{while} has a
12544 true-or-false-test that tests true if the prefix argument for
12545 @code{forward-sentence} is a negative number. This is for going
12546 backwards. The body of this loop is similar to the body of the second
12547 @code{while} clause, but it is not exactly the same. We will skip
12548 this @code{while} loop and concentrate on the second @code{while}
12552 The second @code{while} loop is for moving point forward. Its skeleton
12557 (while (> arg 0) ; @r{true-or-false-test}
12559 (if (@var{true-or-false-test})
12562 (setq arg (1- arg)))) ; @code{while} @r{loop decrementer}
12566 The @code{while} loop is of the decrementing kind.
12567 (@xref{Decrementing Loop, , A Loop with a Decrementing Counter}.) It
12568 has a true-or-false-test that tests true so long as the counter (in
12569 this case, the variable @code{arg}) is greater than zero; and it has a
12570 decrementer that subtracts 1 from the value of the counter every time
12573 If no prefix argument is given to @code{forward-sentence}, which is
12574 the most common way the command is used, this @code{while} loop will
12575 run once, since the value of @code{arg} will be 1.
12577 The body of the @code{while} loop consists of a @code{let} expression,
12578 which creates and binds a local variable, and has, as its body, an
12579 @code{if} expression.
12582 The body of the @code{while} loop looks like this:
12587 (save-excursion (end-of-paragraph-text) (point))))
12588 (if (re-search-forward sentence-end par-end t)
12589 (skip-chars-backward " \t\n")
12590 (goto-char par-end)))
12594 The @code{let} expression creates and binds the local variable
12595 @code{par-end}. As we shall see, this local variable is designed to
12596 provide a bound or limit to the regular expression search. If the
12597 search fails to find a proper sentence ending in the paragraph, it will
12598 stop on reaching the end of the paragraph.
12600 But first, let us examine how @code{par-end} is bound to the value of
12601 the end of the paragraph. What happens is that the @code{let} sets the
12602 value of @code{par-end} to the value returned when the Lisp interpreter
12603 evaluates the expression
12607 (save-excursion (end-of-paragraph-text) (point))
12612 In this expression, @code{(end-of-paragraph-text)} moves point to the
12613 end of the paragraph, @code{(point)} returns the value of point, and then
12614 @code{save-excursion} restores point to its original position. Thus,
12615 the @code{let} binds @code{par-end} to the value returned by the
12616 @code{save-excursion} expression, which is the position of the end of
12617 the paragraph. (The @code{end-of-paragraph-text} function uses
12618 @code{forward-paragraph}, which we will discuss shortly.)
12621 Emacs next evaluates the body of the @code{let}, which is an @code{if}
12622 expression that looks like this:
12626 (if (re-search-forward sentence-end par-end t) ; @r{if-part}
12627 (skip-chars-backward " \t\n") ; @r{then-part}
12628 (goto-char par-end))) ; @r{else-part}
12632 The @code{if} tests whether its first argument is true and if so,
12633 evaluates its then-part; otherwise, the Emacs Lisp interpreter
12634 evaluates the else-part. The true-or-false-test of the @code{if}
12635 expression is the regular expression search.
12637 It may seem odd to have what looks like the `real work' of
12638 the @code{forward-sentence} function buried here, but this is a common
12639 way this kind of operation is carried out in Lisp.
12641 @node fwd-sentence re-search
12642 @unnumberedsubsec The regular expression search
12644 The @code{re-search-forward} function searches for the end of the
12645 sentence, that is, for the pattern defined by the @code{sentence-end}
12646 regular expression. If the pattern is found---if the end of the sentence is
12647 found---then the @code{re-search-forward} function does two things:
12651 The @code{re-search-forward} function carries out a side effect, which
12652 is to move point to the end of the occurrence found.
12655 The @code{re-search-forward} function returns a value of true. This is
12656 the value received by the @code{if}, and means that the search was
12661 The side effect, the movement of point, is completed before the
12662 @code{if} function is handed the value returned by the successful
12663 conclusion of the search.
12665 When the @code{if} function receives the value of true from a successful
12666 call to @code{re-search-forward}, the @code{if} evaluates the then-part,
12667 which is the expression @code{(skip-chars-backward " \t\n")}. This
12668 expression moves backwards over any blank spaces, tabs or carriage
12669 returns until a printed character is found and then leaves point after
12670 the character. Since point has already been moved to the end of the
12671 pattern that marks the end of the sentence, this action leaves point
12672 right after the closing printed character of the sentence, which is
12675 On the other hand, if the @code{re-search-forward} function fails to
12676 find a pattern marking the end of the sentence, the function returns
12677 false. The false then causes the @code{if} to evaluate its third
12678 argument, which is @code{(goto-char par-end)}: it moves point to the
12679 end of the paragraph.
12681 (And if the text is in a form or equivalent, and point may not move
12682 fully, then the @code{constrain-to-field} function comes into play.)
12684 Regular expression searches are exceptionally useful and the pattern
12685 illustrated by @code{re-search-forward}, in which the search is the
12686 test of an @code{if} expression, is handy. You will see or write code
12687 incorporating this pattern often.
12689 @node forward-paragraph
12690 @section @code{forward-paragraph}: a Goldmine of Functions
12691 @findex forward-paragraph
12695 (defun forward-paragraph (&optional arg)
12696 "Move forward to end of paragraph.
12697 With argument ARG, do it ARG times;
12698 a negative argument ARG = -N means move backward N paragraphs.
12700 A line which `paragraph-start' matches either separates paragraphs
12701 \(if `paragraph-separate' matches it also) or is the first line of a paragraph.
12702 A paragraph end is the beginning of a line which is not part of the paragraph
12703 to which the end of the previous line belongs, or the end of the buffer.
12704 Returns the count of paragraphs left to move."
12706 (or arg (setq arg 1))
12707 (let* ((opoint (point))
12708 (fill-prefix-regexp
12709 (and fill-prefix (not (equal fill-prefix ""))
12710 (not paragraph-ignore-fill-prefix)
12711 (regexp-quote fill-prefix)))
12712 ;; Remove ^ from paragraph-start and paragraph-sep if they are there.
12713 ;; These regexps shouldn't be anchored, because we look for them
12714 ;; starting at the left-margin. This allows paragraph commands to
12715 ;; work normally with indented text.
12716 ;; This hack will not find problem cases like "whatever\\|^something".
12717 (parstart (if (and (not (equal "" paragraph-start))
12718 (equal ?^ (aref paragraph-start 0)))
12719 (substring paragraph-start 1)
12721 (parsep (if (and (not (equal "" paragraph-separate))
12722 (equal ?^ (aref paragraph-separate 0)))
12723 (substring paragraph-separate 1)
12724 paragraph-separate))
12726 (if fill-prefix-regexp
12727 (concat parsep "\\|"
12728 fill-prefix-regexp "[ \t]*$")
12730 ;; This is used for searching.
12731 (sp-parstart (concat "^[ \t]*\\(?:" parstart "\\|" parsep "\\)"))
12733 (while (and (< arg 0) (not (bobp)))
12734 (if (and (not (looking-at parsep))
12735 (re-search-backward "^\n" (max (1- (point)) (point-min)) t)
12736 (looking-at parsep))
12737 (setq arg (1+ arg))
12738 (setq start (point))
12739 ;; Move back over paragraph-separating lines.
12740 (forward-char -1) (beginning-of-line)
12741 (while (and (not (bobp))
12742 (progn (move-to-left-margin)
12743 (looking-at parsep)))
12747 (setq arg (1+ arg))
12748 ;; Go to end of the previous (non-separating) line.
12750 ;; Search back for line that starts or separates paragraphs.
12751 (if (if fill-prefix-regexp
12752 ;; There is a fill prefix; it overrides parstart.
12753 (let (multiple-lines)
12754 (while (and (progn (beginning-of-line) (not (bobp)))
12755 (progn (move-to-left-margin)
12756 (not (looking-at parsep)))
12757 (looking-at fill-prefix-regexp))
12758 (unless (= (point) start)
12759 (setq multiple-lines t))
12761 (move-to-left-margin)
12762 ;; This deleted code caused a long hanging-indent line
12763 ;; not to be filled together with the following lines.
12764 ;; ;; Don't move back over a line before the paragraph
12765 ;; ;; which doesn't start with fill-prefix
12766 ;; ;; unless that is the only line we've moved over.
12767 ;; (and (not (looking-at fill-prefix-regexp))
12769 ;; (forward-line 1))
12771 (while (and (re-search-backward sp-parstart nil 1)
12772 (setq found-start t)
12773 ;; Found a candidate, but need to check if it is a
12775 (progn (setq start (point))
12776 (move-to-left-margin)
12777 (not (looking-at parsep)))
12778 (not (and (looking-at parstart)
12779 (or (not use-hard-newlines)
12782 (1- start) 'hard)))))
12783 (setq found-start nil)
12788 ;; Move forward over paragraph separators.
12789 ;; We know this cannot reach the place we started
12790 ;; because we know we moved back over a non-separator.
12791 (while (and (not (eobp))
12792 (progn (move-to-left-margin)
12793 (looking-at parsep)))
12795 ;; If line before paragraph is just margin, back up to there.
12797 (if (> (current-column) (current-left-margin))
12799 (skip-chars-backward " \t")
12801 (forward-line 1))))
12802 ;; No starter or separator line => use buffer beg.
12803 (goto-char (point-min))))))
12805 (while (and (> arg 0) (not (eobp)))
12806 ;; Move forward over separator lines...
12807 (while (and (not (eobp))
12808 (progn (move-to-left-margin) (not (eobp)))
12809 (looking-at parsep))
12811 (unless (eobp) (setq arg (1- arg)))
12812 ;; ... and one more line.
12814 (if fill-prefix-regexp
12815 ;; There is a fill prefix; it overrides parstart.
12816 (while (and (not (eobp))
12817 (progn (move-to-left-margin) (not (eobp)))
12818 (not (looking-at parsep))
12819 (looking-at fill-prefix-regexp))
12821 (while (and (re-search-forward sp-parstart nil 1)
12822 (progn (setq start (match-beginning 0))
12825 (progn (move-to-left-margin)
12826 (not (looking-at parsep)))
12827 (or (not (looking-at parstart))
12828 (and use-hard-newlines
12829 (not (get-text-property (1- start) 'hard)))))
12831 (if (< (point) (point-max))
12832 (goto-char start))))
12833 (constrain-to-field nil opoint t)
12834 ;; Return the number of steps that could not be done.
12838 The @code{forward-paragraph} function moves point forward to the end
12839 of the paragraph. It is usually bound to @kbd{M-@}} and makes use of a
12840 number of functions that are important in themselves, including
12841 @code{let*}, @code{match-beginning}, and @code{looking-at}.
12843 The function definition for @code{forward-paragraph} is considerably
12844 longer than the function definition for @code{forward-sentence}
12845 because it works with a paragraph, each line of which may begin with a
12848 A fill prefix consists of a string of characters that are repeated at
12849 the beginning of each line. For example, in Lisp code, it is a
12850 convention to start each line of a paragraph-long comment with
12851 @samp{;;; }. In Text mode, four blank spaces make up another common
12852 fill prefix, creating an indented paragraph. (@xref{Fill Prefix, , ,
12853 emacs, The GNU Emacs Manual}, for more information about fill
12856 The existence of a fill prefix means that in addition to being able to
12857 find the end of a paragraph whose lines begin on the left-most
12858 column, the @code{forward-paragraph} function must be able to find the
12859 end of a paragraph when all or many of the lines in the buffer begin
12860 with the fill prefix.
12862 Moreover, it is sometimes practical to ignore a fill prefix that
12863 exists, especially when blank lines separate paragraphs.
12864 This is an added complication.
12867 * forward-paragraph in brief:: Key parts of the function definition.
12868 * fwd-para let:: The @code{let*} expression.
12869 * fwd-para while:: The forward motion @code{while} loop.
12873 @node forward-paragraph in brief
12874 @unnumberedsubsec Shortened @code{forward-paragraph} function definition
12877 Rather than print all of the @code{forward-paragraph} function, we
12878 will only print parts of it. Read without preparation, the function
12882 In outline, the function looks like this:
12886 (defun forward-paragraph (&optional arg)
12887 "@var{documentation}@dots{}"
12889 (or arg (setq arg 1))
12892 (while (and (< arg 0) (not (bobp))) ; @r{backward-moving-code}
12894 (while (and (> arg 0) (not (eobp))) ; @r{forward-moving-code}
12899 The first parts of the function are routine: the function's argument
12900 list consists of one optional argument. Documentation follows.
12902 The lower case @samp{p} in the @code{interactive} declaration means
12903 that the processed prefix argument, if any, is passed to the function.
12904 This will be a number, and is the repeat count of how many paragraphs
12905 point will move. The @code{or} expression in the next line handles
12906 the common case when no argument is passed to the function, which occurs
12907 if the function is called from other code rather than interactively.
12908 This case was described earlier. (@xref{forward-sentence, The
12909 @code{forward-sentence} function}.) Now we reach the end of the
12910 familiar part of this function.
12913 @unnumberedsubsec The @code{let*} expression
12915 The next line of the @code{forward-paragraph} function begins a
12916 @code{let*} expression. This is a different than @code{let}. The
12917 symbol is @code{let*} not @code{let}.
12919 The @code{let*} special form is like @code{let} except that Emacs sets
12920 each variable in sequence, one after another, and variables in the
12921 latter part of the varlist can make use of the values to which Emacs
12922 set variables in the earlier part of the varlist.
12925 ( refappend save-excursion, , code save-excursion in code append-to-buffer .)
12928 (@ref{append save-excursion, , @code{save-excursion} in @code{append-to-buffer}}.)
12930 In the @code{let*} expression in this function, Emacs binds a total of
12931 seven variables: @code{opoint}, @code{fill-prefix-regexp},
12932 @code{parstart}, @code{parsep}, @code{sp-parstart}, @code{start}, and
12933 @code{found-start}.
12935 The variable @code{parsep} appears twice, first, to remove instances
12936 of @samp{^}, and second, to handle fill prefixes.
12938 The variable @code{opoint} is just the value of @code{point}. As you
12939 can guess, it is used in a @code{constrain-to-field} expression, just
12940 as in @code{forward-sentence}.
12942 The variable @code{fill-prefix-regexp} is set to the value returned by
12943 evaluating the following list:
12948 (not (equal fill-prefix ""))
12949 (not paragraph-ignore-fill-prefix)
12950 (regexp-quote fill-prefix))
12955 This is an expression whose first element is the @code{and} special form.
12957 As we learned earlier (@pxref{kill-new function, , The @code{kill-new}
12958 function}), the @code{and} special form evaluates each of its
12959 arguments until one of the arguments returns a value of @code{nil}, in
12960 which case the @code{and} expression returns @code{nil}; however, if
12961 none of the arguments returns a value of @code{nil}, the value
12962 resulting from evaluating the last argument is returned. (Since such
12963 a value is not @code{nil}, it is considered true in Lisp.) In other
12964 words, an @code{and} expression returns a true value only if all its
12965 arguments are true.
12968 In this case, the variable @code{fill-prefix-regexp} is bound to a
12969 non-@code{nil} value only if the following four expressions produce a
12970 true (i.e., a non-@code{nil}) value when they are evaluated; otherwise,
12971 @code{fill-prefix-regexp} is bound to @code{nil}.
12975 When this variable is evaluated, the value of the fill prefix, if any,
12976 is returned. If there is no fill prefix, this variable returns
12979 @item (not (equal fill-prefix "")
12980 This expression checks whether an existing fill prefix is an empty
12981 string, that is, a string with no characters in it. An empty string is
12982 not a useful fill prefix.
12984 @item (not paragraph-ignore-fill-prefix)
12985 This expression returns @code{nil} if the variable
12986 @code{paragraph-ignore-fill-prefix} has been turned on by being set to a
12987 true value such as @code{t}.
12989 @item (regexp-quote fill-prefix)
12990 This is the last argument to the @code{and} special form. If all the
12991 arguments to the @code{and} are true, the value resulting from
12992 evaluating this expression will be returned by the @code{and} expression
12993 and bound to the variable @code{fill-prefix-regexp},
12996 @findex regexp-quote
12998 The result of evaluating this @code{and} expression successfully is that
12999 @code{fill-prefix-regexp} will be bound to the value of
13000 @code{fill-prefix} as modified by the @code{regexp-quote} function.
13001 What @code{regexp-quote} does is read a string and return a regular
13002 expression that will exactly match the string and match nothing else.
13003 This means that @code{fill-prefix-regexp} will be set to a value that
13004 will exactly match the fill prefix if the fill prefix exists.
13005 Otherwise, the variable will be set to @code{nil}.
13007 The next two local variables in the @code{let*} expression are
13008 designed to remove instances of @samp{^} from @code{parstart} and
13009 @code{parsep}, the local variables which indicate the paragraph start
13010 and the paragraph separator. The next expression sets @code{parsep}
13011 again. That is to handle fill prefixes.
13013 This is the setting that requires the definition call @code{let*}
13014 rather than @code{let}. The true-or-false-test for the @code{if}
13015 depends on whether the variable @code{fill-prefix-regexp} evaluates to
13016 @code{nil} or some other value.
13018 If @code{fill-prefix-regexp} does not have a value, Emacs evaluates
13019 the else-part of the @code{if} expression and binds @code{parsep} to
13020 its local value. (@code{parsep} is a regular expression that matches
13021 what separates paragraphs.)
13023 But if @code{fill-prefix-regexp} does have a value, Emacs evaluates
13024 the then-part of the @code{if} expression and binds @code{parsep} to a
13025 regular expression that includes the @code{fill-prefix-regexp} as part
13028 Specifically, @code{parsep} is set to the original value of the
13029 paragraph separate regular expression concatenated with an alternative
13030 expression that consists of the @code{fill-prefix-regexp} followed by
13031 optional whitespace to the end of the line. The whitespace is defined
13032 by @w{@code{"[ \t]*$"}}.) The @samp{\\|} defines this portion of the
13033 regexp as an alternative to @code{parsep}.
13035 According to a comment in the code, the next local variable,
13036 @code{sp-parstart}, is used for searching, and then the final two,
13037 @code{start} and @code{found-start}, are set to @code{nil}.
13039 Now we get into the body of the @code{let*}. The first part of the body
13040 of the @code{let*} deals with the case when the function is given a
13041 negative argument and is therefore moving backwards. We will skip this
13044 @node fwd-para while
13045 @unnumberedsubsec The forward motion @code{while} loop
13047 The second part of the body of the @code{let*} deals with forward
13048 motion. It is a @code{while} loop that repeats itself so long as the
13049 value of @code{arg} is greater than zero. In the most common use of
13050 the function, the value of the argument is 1, so the body of the
13051 @code{while} loop is evaluated exactly once, and the cursor moves
13052 forward one paragraph.
13055 (while (and (> arg 0) (not (eobp)))
13057 ;; Move forward over separator lines...
13058 (while (and (not (eobp))
13059 (progn (move-to-left-margin) (not (eobp)))
13060 (looking-at parsep))
13062 (unless (eobp) (setq arg (1- arg)))
13063 ;; ... and one more line.
13066 (if fill-prefix-regexp
13067 ;; There is a fill prefix; it overrides parstart.
13068 (while (and (not (eobp))
13069 (progn (move-to-left-margin) (not (eobp)))
13070 (not (looking-at parsep))
13071 (looking-at fill-prefix-regexp))
13074 (while (and (re-search-forward sp-parstart nil 1)
13075 (progn (setq start (match-beginning 0))
13078 (progn (move-to-left-margin)
13079 (not (looking-at parsep)))
13080 (or (not (looking-at parstart))
13081 (and use-hard-newlines
13082 (not (get-text-property (1- start) 'hard)))))
13085 (if (< (point) (point-max))
13086 (goto-char start))))
13089 This part handles three situations: when point is between paragraphs,
13090 when there is a fill prefix and when there is no fill prefix.
13093 The @code{while} loop looks like this:
13097 ;; @r{going forwards and not at the end of the buffer}
13098 (while (and (> arg 0) (not (eobp)))
13100 ;; @r{between paragraphs}
13101 ;; Move forward over separator lines...
13102 (while (and (not (eobp))
13103 (progn (move-to-left-margin) (not (eobp)))
13104 (looking-at parsep))
13106 ;; @r{This decrements the loop}
13107 (unless (eobp) (setq arg (1- arg)))
13108 ;; ... and one more line.
13113 (if fill-prefix-regexp
13114 ;; There is a fill prefix; it overrides parstart;
13115 ;; we go forward line by line
13116 (while (and (not (eobp))
13117 (progn (move-to-left-margin) (not (eobp)))
13118 (not (looking-at parsep))
13119 (looking-at fill-prefix-regexp))
13124 ;; There is no fill prefix;
13125 ;; we go forward character by character
13126 (while (and (re-search-forward sp-parstart nil 1)
13127 (progn (setq start (match-beginning 0))
13130 (progn (move-to-left-margin)
13131 (not (looking-at parsep)))
13132 (or (not (looking-at parstart))
13133 (and use-hard-newlines
13134 (not (get-text-property (1- start) 'hard)))))
13139 ;; and if there is no fill prefix and if we are not at the end,
13140 ;; go to whatever was found in the regular expression search
13142 (if (< (point) (point-max))
13143 (goto-char start))))
13148 We can see that this is a decrementing counter @code{while} loop,
13149 using the expression @code{(setq arg (1- arg))} as the decrementer.
13150 That expression is not far from the @code{while}, but is hidden in
13151 another Lisp macro, an @code{unless} macro. Unless we are at the end
13152 of the buffer---that is what the @code{eobp} function determines; it
13153 is an abbreviation of @samp{End Of Buffer P}---we decrease the value
13154 of @code{arg} by one.
13156 (If we are at the end of the buffer, we cannot go forward any more and
13157 the next loop of the @code{while} expression will test false since the
13158 test is an @code{and} with @code{(not (eobp))}. The @code{not}
13159 function means exactly as you expect; it is another name for
13160 @code{null}, a function that returns true when its argument is false.)
13162 Interestingly, the loop count is not decremented until we leave the
13163 space between paragraphs, unless we come to the end of buffer or stop
13164 seeing the local value of the paragraph separator.
13166 That second @code{while} also has a @code{(move-to-left-margin)}
13167 expression. The function is self-explanatory. It is inside a
13168 @code{progn} expression and not the last element of its body, so it is
13169 only invoked for its side effect, which is to move point to the left
13170 margin of the current line.
13173 The @code{looking-at} function is also self-explanatory; it returns
13174 true if the text after point matches the regular expression given as
13177 The rest of the body of the loop looks difficult at first, but makes
13178 sense as you come to understand it.
13181 First consider what happens if there is a fill prefix:
13185 (if fill-prefix-regexp
13186 ;; There is a fill prefix; it overrides parstart;
13187 ;; we go forward line by line
13188 (while (and (not (eobp))
13189 (progn (move-to-left-margin) (not (eobp)))
13190 (not (looking-at parsep))
13191 (looking-at fill-prefix-regexp))
13197 This expression moves point forward line by line so long
13198 as four conditions are true:
13202 Point is not at the end of the buffer.
13205 We can move to the left margin of the text and are
13206 not at the end of the buffer.
13209 The text following point does not separate paragraphs.
13212 The pattern following point is the fill prefix regular expression.
13215 The last condition may be puzzling, until you remember that point was
13216 moved to the beginning of the line early in the @code{forward-paragraph}
13217 function. This means that if the text has a fill prefix, the
13218 @code{looking-at} function will see it.
13221 Consider what happens when there is no fill prefix.
13225 (while (and (re-search-forward sp-parstart nil 1)
13226 (progn (setq start (match-beginning 0))
13229 (progn (move-to-left-margin)
13230 (not (looking-at parsep)))
13231 (or (not (looking-at parstart))
13232 (and use-hard-newlines
13233 (not (get-text-property (1- start) 'hard)))))
13239 This @code{while} loop has us searching forward for
13240 @code{sp-parstart}, which is the combination of possible whitespace
13241 with a the local value of the start of a paragraph or of a paragraph
13242 separator. (The latter two are within an expression starting
13243 @code{\(?:} so that they are not referenced by the
13244 @code{match-beginning} function.)
13247 The two expressions,
13251 (setq start (match-beginning 0))
13257 mean go to the start of the text matched by the regular expression
13260 The @code{(match-beginning 0)} expression is new. It returns a number
13261 specifying the location of the start of the text that was matched by
13264 The @code{match-beginning} function is used here because of a
13265 characteristic of a forward search: a successful forward search,
13266 regardless of whether it is a plain search or a regular expression
13267 search, moves point to the end of the text that is found. In this
13268 case, a successful search moves point to the end of the pattern for
13269 @code{sp-parstart}.
13271 However, we want to put point at the end of the current paragraph, not
13272 somewhere else. Indeed, since the search possibly includes the
13273 paragraph separator, point may end up at the beginning of the next one
13274 unless we use an expression that includes @code{match-beginning}.
13276 @findex match-beginning
13277 When given an argument of 0, @code{match-beginning} returns the
13278 position that is the start of the text matched by the most recent
13279 search. In this case, the most recent search looks for
13280 @code{sp-parstart}. The @code{(match-beginning 0)} expression returns
13281 the beginning position of that pattern, rather than the end position
13284 (Incidentally, when passed a positive number as an argument, the
13285 @code{match-beginning} function returns the location of point at that
13286 parenthesized expression in the last search unless that parenthesized
13287 expression begins with @code{\(?:}. I don't know why @code{\(?:}
13288 appears here since the argument is 0.)
13291 The last expression when there is no fill prefix is
13295 (if (< (point) (point-max))
13296 (goto-char start))))
13301 This says that if there is no fill prefix and if we are not at the
13302 end, point should move to the beginning of whatever was found by the
13303 regular expression search for @code{sp-parstart}.
13305 The full definition for the @code{forward-paragraph} function not only
13306 includes code for going forwards, but also code for going backwards.
13308 If you are reading this inside of GNU Emacs and you want to see the
13309 whole function, you can type @kbd{C-h f} (@code{describe-function})
13310 and the name of the function. This gives you the function
13311 documentation and the name of the library containing the function's
13312 source. Place point over the name of the library and press the RET
13313 key; you will be taken directly to the source. (Be sure to install
13314 your sources! Without them, you are like a person who tries to drive
13315 a car with his eyes shut!)
13318 @section Create Your Own @file{TAGS} File
13320 @cindex @file{TAGS} file, create own
13322 Besides @kbd{C-h f} (@code{describe-function}), another way to see the
13323 source of a function is to type @kbd{M-.} (@code{find-tag}) and the
13324 name of the function when prompted for it. This is a good habit to
13325 get into. The @kbd{M-.} (@code{find-tag}) command takes you directly
13326 to the source for a function, variable, or node. The function depends
13327 on tags tables to tell it where to go.
13329 If the @code{find-tag} function first asks you for the name of a
13330 @file{TAGS} table, give it the name of a @file{TAGS} file such as
13331 @file{/usr/local/src/emacs/src/TAGS}. (The exact path to your
13332 @file{TAGS} file depends on how your copy of Emacs was installed. I
13333 just told you the location that provides both my C and my Emacs Lisp
13336 You can also create your own @file{TAGS} file for directories that
13339 You often need to build and install tags tables yourself. They are
13340 not built automatically. A tags table is called a @file{TAGS} file;
13341 the name is in upper case letters.
13343 You can create a @file{TAGS} file by calling the @code{etags} program
13344 that comes as a part of the Emacs distribution. Usually, @code{etags}
13345 is compiled and installed when Emacs is built. (@code{etags} is not
13346 an Emacs Lisp function or a part of Emacs; it is a C program.)
13349 To create a @file{TAGS} file, first switch to the directory in which
13350 you want to create the file. In Emacs you can do this with the
13351 @kbd{M-x cd} command, or by visiting a file in the directory, or by
13352 listing the directory with @kbd{C-x d} (@code{dired}). Then run the
13353 compile command, with @w{@code{etags *.el}} as the command to execute
13356 M-x compile RET etags *.el RET
13360 to create a @file{TAGS} file for Emacs Lisp.
13362 For example, if you have a large number of files in your
13363 @file{~/emacs} directory, as I do---I have 137 @file{.el} files in it,
13364 of which I load 12---you can create a @file{TAGS} file for the Emacs
13365 Lisp files in that directory.
13368 The @code{etags} program takes all the usual shell `wildcards'. For
13369 example, if you have two directories for which you want a single
13370 @file{TAGS} file, type @w{@code{etags *.el ../elisp/*.el}}, where
13371 @file{../elisp/} is the second directory:
13374 M-x compile RET etags *.el ../elisp/*.el RET
13381 M-x compile RET etags --help RET
13385 to see a list of the options accepted by @code{etags} as well as a
13386 list of supported languages.
13388 The @code{etags} program handles more than 20 languages, including
13389 Emacs Lisp, Common Lisp, Scheme, C, C++, Ada, Fortran, HTML, Java,
13390 LaTeX, Pascal, Perl, PostScript, Python, TeX, Texinfo, makefiles, and
13391 most assemblers. The program has no switches for specifying the
13392 language; it recognizes the language in an input file according to its
13393 file name and contents.
13395 @file{etags} is very helpful when you are writing code yourself and
13396 want to refer back to functions you have already written. Just run
13397 @code{etags} again at intervals as you write new functions, so they
13398 become part of the @file{TAGS} file.
13400 If you think an appropriate @file{TAGS} file already exists for what
13401 you want, but do not know where it is, you can use the @code{locate}
13402 program to attempt to find it.
13404 Type @w{@kbd{M-x locate @key{RET} TAGS @key{RET}}} and Emacs will list
13405 for you the full path names of all your @file{TAGS} files. On my
13406 system, this command lists 34 @file{TAGS} files. On the other hand, a
13407 `plain vanilla' system I recently installed did not contain any
13410 If the tags table you want has been created, you can use the @code{M-x
13411 visit-tags-table} command to specify it. Otherwise, you will need to
13412 create the tag table yourself and then use @code{M-x
13415 @subsubheading Building Tags in the Emacs sources
13416 @cindex Building Tags in the Emacs sources
13417 @cindex Tags in the Emacs sources
13420 The GNU Emacs sources come with a @file{Makefile} that contains a
13421 sophisticated @code{etags} command that creates, collects, and merges
13422 tags tables from all over the Emacs sources and puts the information
13423 into one @file{TAGS} file in the @file{src/} directory. (The
13424 @file{src/} directory is below the top level of your Emacs directory.)
13427 To build this @file{TAGS} file, go to the top level of your Emacs
13428 source directory and run the compile command @code{make tags}:
13431 M-x compile RET make tags RET
13435 (The @code{make tags} command works well with the GNU Emacs sources,
13436 as well as with some other source packages.)
13438 For more information, see @ref{Tags, , Tag Tables, emacs, The GNU Emacs
13441 @node Regexp Review
13444 Here is a brief summary of some recently introduced functions.
13448 Repeatedly evaluate the body of the expression so long as the first
13449 element of the body tests true. Then return @code{nil}. (The
13450 expression is evaluated only for its side effects.)
13459 (insert (format "foo is %d.\n" foo))
13460 (setq foo (1- foo))))
13462 @result{} foo is 2.
13469 (The @code{insert} function inserts its arguments at point; the
13470 @code{format} function returns a string formatted from its arguments
13471 the way @code{message} formats its arguments; @code{\n} produces a new
13474 @item re-search-forward
13475 Search for a pattern, and if the pattern is found, move point to rest
13479 Takes four arguments, like @code{search-forward}:
13483 A regular expression that specifies the pattern to search for.
13484 (Remember to put quotation marks around this argument!)
13487 Optionally, the limit of the search.
13490 Optionally, what to do if the search fails, return @code{nil} or an
13494 Optionally, how many times to repeat the search; if negative, the
13495 search goes backwards.
13499 Bind some variables locally to particular values,
13500 and then evaluate the remaining arguments, returning the value of the
13501 last one. While binding the local variables, use the local values of
13502 variables bound earlier, if any.
13511 (message "`bar' is %d." bar))
13512 @result{} `bar' is 21.
13516 @item match-beginning
13517 Return the position of the start of the text found by the last regular
13521 Return @code{t} for true if the text after point matches the argument,
13522 which should be a regular expression.
13525 Return @code{t} for true if point is at the end of the accessible part
13526 of a buffer. The end of the accessible part is the end of the buffer
13527 if the buffer is not narrowed; it is the end of the narrowed part if
13528 the buffer is narrowed.
13532 @node re-search Exercises
13533 @section Exercises with @code{re-search-forward}
13537 Write a function to search for a regular expression that matches two
13538 or more blank lines in sequence.
13541 Write a function to search for duplicated words, such as `the the'.
13542 @xref{Regexps, , Syntax of Regular Expressions, emacs, The GNU Emacs
13543 Manual}, for information on how to write a regexp (a regular
13544 expression) to match a string that is composed of two identical
13545 halves. You can devise several regexps; some are better than others.
13546 The function I use is described in an appendix, along with several
13547 regexps. @xref{the-the, , @code{the-the} Duplicated Words Function}.
13550 @node Counting Words
13551 @chapter Counting: Repetition and Regexps
13552 @cindex Repetition for word counting
13553 @cindex Regular expressions for word counting
13555 Repetition and regular expression searches are powerful tools that you
13556 often use when you write code in Emacs Lisp. This chapter illustrates
13557 the use of regular expression searches through the construction of
13558 word count commands using @code{while} loops and recursion.
13561 * Why Count Words::
13562 * @value{COUNT-WORDS}:: Use a regexp, but find a problem.
13563 * recursive-count-words:: Start with case of no words in region.
13564 * Counting Exercise::
13568 @node Why Count Words
13569 @unnumberedsec Counting words
13572 The standard Emacs distribution contains functions for counting the
13573 number of lines and words within a region.
13575 Certain types of writing ask you to count words. Thus, if you write
13576 an essay, you may be limited to 800 words; if you write a novel, you
13577 may discipline yourself to write 1000 words a day. It seems odd, but
13578 for a long time, Emacs lacked a word count command. Perhaps people used
13579 Emacs mostly for code or types of documentation that did not require
13580 word counts; or perhaps they restricted themselves to the operating
13581 system word count command, @code{wc}. Alternatively, people may have
13582 followed the publishers' convention and computed a word count by
13583 dividing the number of characters in a document by five.
13585 There are many ways to implement a command to count words. Here are
13586 some examples, which you may wish to compare with the standard Emacs
13587 command, @code{count-words-region}.
13589 @node @value{COUNT-WORDS}
13590 @section The @code{@value{COUNT-WORDS}} Function
13591 @findex @value{COUNT-WORDS}
13593 A word count command could count words in a line, paragraph, region,
13594 or buffer. What should the command cover? You could design the
13595 command to count the number of words in a complete buffer. However,
13596 the Emacs tradition encourages flexibility---you may want to count
13597 words in just a section, rather than all of a buffer. So it makes
13598 more sense to design the command to count the number of words in a
13599 region. Once you have a command to count words in a region, you can,
13600 if you wish, count words in a whole buffer by marking it with
13601 @w{@kbd{C-x h}} (@code{mark-whole-buffer}).
13603 Clearly, counting words is a repetitive act: starting from the
13604 beginning of the region, you count the first word, then the second
13605 word, then the third word, and so on, until you reach the end of the
13606 region. This means that word counting is ideally suited to recursion
13607 or to a @code{while} loop.
13610 * Design @value{COUNT-WORDS}:: The definition using a @code{while} loop.
13611 * Whitespace Bug:: The Whitespace Bug in @code{@value{COUNT-WORDS}}.
13615 @node Design @value{COUNT-WORDS}
13616 @unnumberedsubsec Designing @code{@value{COUNT-WORDS}}
13619 First, we will implement the word count command with a @code{while}
13620 loop, then with recursion. The command will, of course, be
13624 The template for an interactive function definition is, as always:
13628 (defun @var{name-of-function} (@var{argument-list})
13629 "@var{documentation}@dots{}"
13630 (@var{interactive-expression}@dots{})
13635 What we need to do is fill in the slots.
13637 The name of the function should be self-explanatory and similar to the
13638 existing @code{count-lines-region} name. This makes the name easier
13639 to remember. @code{count-words-region} is the obvious choice. Since
13640 that name is now used for the standard Emacs command to count words, we
13641 will name our implementation @code{@value{COUNT-WORDS}}.
13643 The function counts words within a region. This means that the
13644 argument list must contain symbols that are bound to the two
13645 positions, the beginning and end of the region. These two positions
13646 can be called @samp{beginning} and @samp{end} respectively. The first
13647 line of the documentation should be a single sentence, since that is
13648 all that is printed as documentation by a command such as
13649 @code{apropos}. The interactive expression will be of the form
13650 @samp{(interactive "r")}, since that will cause Emacs to pass the
13651 beginning and end of the region to the function's argument list. All
13654 The body of the function needs to be written to do three tasks:
13655 first, to set up conditions under which the @code{while} loop can
13656 count words, second, to run the @code{while} loop, and third, to send
13657 a message to the user.
13659 When a user calls @code{@value{COUNT-WORDS}}, point may be at the
13660 beginning or the end of the region. However, the counting process
13661 must start at the beginning of the region. This means we will want
13662 to put point there if it is not already there. Executing
13663 @code{(goto-char beginning)} ensures this. Of course, we will want to
13664 return point to its expected position when the function finishes its
13665 work. For this reason, the body must be enclosed in a
13666 @code{save-excursion} expression.
13668 The central part of the body of the function consists of a
13669 @code{while} loop in which one expression jumps point forward word by
13670 word, and another expression counts those jumps. The true-or-false-test
13671 of the @code{while} loop should test true so long as point should jump
13672 forward, and false when point is at the end of the region.
13674 We could use @code{(forward-word 1)} as the expression for moving point
13675 forward word by word, but it is easier to see what Emacs identifies as a
13676 `word' if we use a regular expression search.
13678 A regular expression search that finds the pattern for which it is
13679 searching leaves point after the last character matched. This means
13680 that a succession of successful word searches will move point forward
13683 As a practical matter, we want the regular expression search to jump
13684 over whitespace and punctuation between words as well as over the
13685 words themselves. A regexp that refuses to jump over interword
13686 whitespace would never jump more than one word! This means that
13687 the regexp should include the whitespace and punctuation that follows
13688 a word, if any, as well as the word itself. (A word may end a buffer
13689 and not have any following whitespace or punctuation, so that part of
13690 the regexp must be optional.)
13692 Thus, what we want for the regexp is a pattern defining one or more
13693 word constituent characters followed, optionally, by one or more
13694 characters that are not word constituents. The regular expression for
13702 The buffer's syntax table determines which characters are and are not
13703 word constituents. For more information about syntax,
13704 @pxref{Syntax Tables, , Syntax Tables, elisp, The GNU Emacs Lisp
13708 The search expression looks like this:
13711 (re-search-forward "\\w+\\W*")
13715 (Note that paired backslashes precede the @samp{w} and @samp{W}. A
13716 single backslash has special meaning to the Emacs Lisp interpreter.
13717 It indicates that the following character is interpreted differently
13718 than usual. For example, the two characters, @samp{\n}, stand for
13719 @samp{newline}, rather than for a backslash followed by @samp{n}. Two
13720 backslashes in a row stand for an ordinary, `unspecial' backslash, so
13721 Emacs Lisp interpreter ends of seeing a single backslash followed by a
13722 letter. So it discovers the letter is special.)
13724 We need a counter to count how many words there are; this variable
13725 must first be set to 0 and then incremented each time Emacs goes
13726 around the @code{while} loop. The incrementing expression is simply:
13729 (setq count (1+ count))
13732 Finally, we want to tell the user how many words there are in the
13733 region. The @code{message} function is intended for presenting this
13734 kind of information to the user. The message has to be phrased so
13735 that it reads properly regardless of how many words there are in the
13736 region: we don't want to say that ``there are 1 words in the region''.
13737 The conflict between singular and plural is ungrammatical. We can
13738 solve this problem by using a conditional expression that evaluates
13739 different messages depending on the number of words in the region.
13740 There are three possibilities: no words in the region, one word in the
13741 region, and more than one word. This means that the @code{cond}
13742 special form is appropriate.
13745 All this leads to the following function definition:
13749 ;;; @r{First version; has bugs!}
13750 (defun @value{COUNT-WORDS} (beginning end)
13751 "Print number of words in the region.
13752 Words are defined as at least one word-constituent
13753 character followed by at least one character that
13754 is not a word-constituent. The buffer's syntax
13755 table determines which characters these are."
13757 (message "Counting words in region ... ")
13761 ;;; @r{1. Set up appropriate conditions.}
13763 (goto-char beginning)
13768 ;;; @r{2. Run the} while @r{loop.}
13769 (while (< (point) end)
13770 (re-search-forward "\\w+\\W*")
13771 (setq count (1+ count)))
13775 ;;; @r{3. Send a message to the user.}
13776 (cond ((zerop count)
13778 "The region does NOT have any words."))
13781 "The region has 1 word."))
13784 "The region has %d words." count))))))
13789 As written, the function works, but not in all circumstances.
13791 @node Whitespace Bug
13792 @subsection The Whitespace Bug in @code{@value{COUNT-WORDS}}
13794 The @code{@value{COUNT-WORDS}} command described in the preceding
13795 section has two bugs, or rather, one bug with two manifestations.
13796 First, if you mark a region containing only whitespace in the middle
13797 of some text, the @code{@value{COUNT-WORDS}} command tells you that the
13798 region contains one word! Second, if you mark a region containing
13799 only whitespace at the end of the buffer or the accessible portion of
13800 a narrowed buffer, the command displays an error message that looks
13804 Search failed: "\\w+\\W*"
13807 If you are reading this in Info in GNU Emacs, you can test for these
13810 First, evaluate the function in the usual manner to install it.
13812 Here is a copy of the definition. Place your cursor after the closing
13813 parenthesis and type @kbd{C-x C-e} to install it.
13817 ;; @r{First version; has bugs!}
13818 (defun @value{COUNT-WORDS} (beginning end)
13819 "Print number of words in the region.
13820 Words are defined as at least one word-constituent character followed
13821 by at least one character that is not a word-constituent. The buffer's
13822 syntax table determines which characters these are."
13826 (message "Counting words in region ... ")
13830 ;;; @r{1. Set up appropriate conditions.}
13832 (goto-char beginning)
13837 ;;; @r{2. Run the} while @r{loop.}
13838 (while (< (point) end)
13839 (re-search-forward "\\w+\\W*")
13840 (setq count (1+ count)))
13844 ;;; @r{3. Send a message to the user.}
13845 (cond ((zerop count)
13846 (message "The region does NOT have any words."))
13847 ((= 1 count) (message "The region has 1 word."))
13848 (t (message "The region has %d words." count))))))
13854 If you wish, you can also install this keybinding by evaluating it:
13857 (global-set-key "\C-c=" '@value{COUNT-WORDS})
13860 To conduct the first test, set mark and point to the beginning and end
13861 of the following line and then type @kbd{C-c =} (or @kbd{M-x
13862 @value{COUNT-WORDS}} if you have not bound @kbd{C-c =}):
13869 Emacs will tell you, correctly, that the region has three words.
13871 Repeat the test, but place mark at the beginning of the line and place
13872 point just @emph{before} the word @samp{one}. Again type the command
13873 @kbd{C-c =} (or @kbd{M-x @value{COUNT-WORDS}}). Emacs should tell you
13874 that the region has no words, since it is composed only of the
13875 whitespace at the beginning of the line. But instead Emacs tells you
13876 that the region has one word!
13878 For the third test, copy the sample line to the end of the
13879 @file{*scratch*} buffer and then type several spaces at the end of the
13880 line. Place mark right after the word @samp{three} and point at the
13881 end of line. (The end of the line will be the end of the buffer.)
13882 Type @kbd{C-c =} (or @kbd{M-x @value{COUNT-WORDS}}) as you did before.
13883 Again, Emacs should tell you that the region has no words, since it is
13884 composed only of the whitespace at the end of the line. Instead,
13885 Emacs displays an error message saying @samp{Search failed}.
13887 The two bugs stem from the same problem.
13889 Consider the first manifestation of the bug, in which the command
13890 tells you that the whitespace at the beginning of the line contains
13891 one word. What happens is this: The @code{M-x @value{COUNT-WORDS}}
13892 command moves point to the beginning of the region. The @code{while}
13893 tests whether the value of point is smaller than the value of
13894 @code{end}, which it is. Consequently, the regular expression search
13895 looks for and finds the first word. It leaves point after the word.
13896 @code{count} is set to one. The @code{while} loop repeats; but this
13897 time the value of point is larger than the value of @code{end}, the
13898 loop is exited; and the function displays a message saying the number
13899 of words in the region is one. In brief, the regular expression
13900 search looks for and finds the word even though it is outside
13903 In the second manifestation of the bug, the region is whitespace at
13904 the end of the buffer. Emacs says @samp{Search failed}. What happens
13905 is that the true-or-false-test in the @code{while} loop tests true, so
13906 the search expression is executed. But since there are no more words
13907 in the buffer, the search fails.
13909 In both manifestations of the bug, the search extends or attempts to
13910 extend outside of the region.
13912 The solution is to limit the search to the region---this is a fairly
13913 simple action, but as you may have come to expect, it is not quite as
13914 simple as you might think.
13916 As we have seen, the @code{re-search-forward} function takes a search
13917 pattern as its first argument. But in addition to this first,
13918 mandatory argument, it accepts three optional arguments. The optional
13919 second argument bounds the search. The optional third argument, if
13920 @code{t}, causes the function to return @code{nil} rather than signal
13921 an error if the search fails. The optional fourth argument is a
13922 repeat count. (In Emacs, you can see a function's documentation by
13923 typing @kbd{C-h f}, the name of the function, and then @key{RET}.)
13925 In the @code{@value{COUNT-WORDS}} definition, the value of the end of
13926 the region is held by the variable @code{end} which is passed as an
13927 argument to the function. Thus, we can add @code{end} as an argument
13928 to the regular expression search expression:
13931 (re-search-forward "\\w+\\W*" end)
13934 However, if you make only this change to the @code{@value{COUNT-WORDS}}
13935 definition and then test the new version of the definition on a
13936 stretch of whitespace, you will receive an error message saying
13937 @samp{Search failed}.
13939 What happens is this: the search is limited to the region, and fails
13940 as you expect because there are no word-constituent characters in the
13941 region. Since it fails, we receive an error message. But we do not
13942 want to receive an error message in this case; we want to receive the
13943 message that "The region does NOT have any words."
13945 The solution to this problem is to provide @code{re-search-forward}
13946 with a third argument of @code{t}, which causes the function to return
13947 @code{nil} rather than signal an error if the search fails.
13949 However, if you make this change and try it, you will see the message
13950 ``Counting words in region ... '' and @dots{} you will keep on seeing
13951 that message @dots{}, until you type @kbd{C-g} (@code{keyboard-quit}).
13953 Here is what happens: the search is limited to the region, as before,
13954 and it fails because there are no word-constituent characters in the
13955 region, as expected. Consequently, the @code{re-search-forward}
13956 expression returns @code{nil}. It does nothing else. In particular,
13957 it does not move point, which it does as a side effect if it finds the
13958 search target. After the @code{re-search-forward} expression returns
13959 @code{nil}, the next expression in the @code{while} loop is evaluated.
13960 This expression increments the count. Then the loop repeats. The
13961 true-or-false-test tests true because the value of point is still less
13962 than the value of end, since the @code{re-search-forward} expression
13963 did not move point. @dots{} and the cycle repeats @dots{}
13965 The @code{@value{COUNT-WORDS}} definition requires yet another
13966 modification, to cause the true-or-false-test of the @code{while} loop
13967 to test false if the search fails. Put another way, there are two
13968 conditions that must be satisfied in the true-or-false-test before the
13969 word count variable is incremented: point must still be within the
13970 region and the search expression must have found a word to count.
13972 Since both the first condition and the second condition must be true
13973 together, the two expressions, the region test and the search
13974 expression, can be joined with an @code{and} special form and embedded in
13975 the @code{while} loop as the true-or-false-test, like this:
13978 (and (< (point) end) (re-search-forward "\\w+\\W*" end t))
13981 @c colon in printed section title causes problem in Info cross reference
13982 @c also trouble with an overfull hbox
13985 (For information about @code{and}, see
13986 @ref{kill-new function, , The @code{kill-new} function}.)
13990 (@xref{kill-new function, , The @code{kill-new} function}, for
13991 information about @code{and}.)
13994 The @code{re-search-forward} expression returns @code{t} if the search
13995 succeeds and as a side effect moves point. Consequently, as words are
13996 found, point is moved through the region. When the search expression
13997 fails to find another word, or when point reaches the end of the
13998 region, the true-or-false-test tests false, the @code{while} loop
13999 exits, and the @code{@value{COUNT-WORDS}} function displays one or
14000 other of its messages.
14002 After incorporating these final changes, the @code{@value{COUNT-WORDS}}
14003 works without bugs (or at least, without bugs that I have found!).
14004 Here is what it looks like:
14008 ;;; @r{Final version:} @code{while}
14009 (defun @value{COUNT-WORDS} (beginning end)
14010 "Print number of words in the region."
14012 (message "Counting words in region ... ")
14016 ;;; @r{1. Set up appropriate conditions.}
14019 (goto-char beginning)
14023 ;;; @r{2. Run the} while @r{loop.}
14024 (while (and (< (point) end)
14025 (re-search-forward "\\w+\\W*" end t))
14026 (setq count (1+ count)))
14030 ;;; @r{3. Send a message to the user.}
14031 (cond ((zerop count)
14033 "The region does NOT have any words."))
14036 "The region has 1 word."))
14039 "The region has %d words." count))))))
14043 @node recursive-count-words
14044 @section Count Words Recursively
14045 @cindex Count words recursively
14046 @cindex Recursively counting words
14047 @cindex Words, counted recursively
14049 You can write the function for counting words recursively as well as
14050 with a @code{while} loop. Let's see how this is done.
14052 First, we need to recognize that the @code{@value{COUNT-WORDS}}
14053 function has three jobs: it sets up the appropriate conditions for
14054 counting to occur; it counts the words in the region; and it sends a
14055 message to the user telling how many words there are.
14057 If we write a single recursive function to do everything, we will
14058 receive a message for every recursive call. If the region contains 13
14059 words, we will receive thirteen messages, one right after the other.
14060 We don't want this! Instead, we must write two functions to do the
14061 job, one of which (the recursive function) will be used inside of the
14062 other. One function will set up the conditions and display the
14063 message; the other will return the word count.
14065 Let us start with the function that causes the message to be displayed.
14066 We can continue to call this @code{@value{COUNT-WORDS}}.
14068 This is the function that the user will call. It will be interactive.
14069 Indeed, it will be similar to our previous versions of this
14070 function, except that it will call @code{recursive-count-words} to
14071 determine how many words are in the region.
14074 We can readily construct a template for this function, based on our
14079 ;; @r{Recursive version; uses regular expression search}
14080 (defun @value{COUNT-WORDS} (beginning end)
14081 "@var{documentation}@dots{}"
14082 (@var{interactive-expression}@dots{})
14086 ;;; @r{1. Set up appropriate conditions.}
14087 (@var{explanatory message})
14088 (@var{set-up functions}@dots{}
14092 ;;; @r{2. Count the words.}
14093 @var{recursive call}
14097 ;;; @r{3. Send a message to the user.}
14098 @var{message providing word count}))
14102 The definition looks straightforward, except that somehow the count
14103 returned by the recursive call must be passed to the message
14104 displaying the word count. A little thought suggests that this can be
14105 done by making use of a @code{let} expression: we can bind a variable
14106 in the varlist of a @code{let} expression to the number of words in
14107 the region, as returned by the recursive call; and then the
14108 @code{cond} expression, using binding, can display the value to the
14111 Often, one thinks of the binding within a @code{let} expression as
14112 somehow secondary to the `primary' work of a function. But in this
14113 case, what you might consider the `primary' job of the function,
14114 counting words, is done within the @code{let} expression.
14117 Using @code{let}, the function definition looks like this:
14121 (defun @value{COUNT-WORDS} (beginning end)
14122 "Print number of words in the region."
14127 ;;; @r{1. Set up appropriate conditions.}
14128 (message "Counting words in region ... ")
14130 (goto-char beginning)
14134 ;;; @r{2. Count the words.}
14135 (let ((count (recursive-count-words end)))
14139 ;;; @r{3. Send a message to the user.}
14140 (cond ((zerop count)
14142 "The region does NOT have any words."))
14145 "The region has 1 word."))
14148 "The region has %d words." count))))))
14152 Next, we need to write the recursive counting function.
14154 A recursive function has at least three parts: the `do-again-test', the
14155 `next-step-expression', and the recursive call.
14157 The do-again-test determines whether the function will or will not be
14158 called again. Since we are counting words in a region and can use a
14159 function that moves point forward for every word, the do-again-test
14160 can check whether point is still within the region. The do-again-test
14161 should find the value of point and determine whether point is before,
14162 at, or after the value of the end of the region. We can use the
14163 @code{point} function to locate point. Clearly, we must pass the
14164 value of the end of the region to the recursive counting function as an
14167 In addition, the do-again-test should also test whether the search finds a
14168 word. If it does not, the function should not call itself again.
14170 The next-step-expression changes a value so that when the recursive
14171 function is supposed to stop calling itself, it stops. More
14172 precisely, the next-step-expression changes a value so that at the
14173 right time, the do-again-test stops the recursive function from
14174 calling itself again. In this case, the next-step-expression can be
14175 the expression that moves point forward, word by word.
14177 The third part of a recursive function is the recursive call.
14179 Somewhere, also, we also need a part that does the `work' of the
14180 function, a part that does the counting. A vital part!
14183 But already, we have an outline of the recursive counting function:
14187 (defun recursive-count-words (region-end)
14188 "@var{documentation}@dots{}"
14189 @var{do-again-test}
14190 @var{next-step-expression}
14191 @var{recursive call})
14195 Now we need to fill in the slots. Let's start with the simplest cases
14196 first: if point is at or beyond the end of the region, there cannot
14197 be any words in the region, so the function should return zero.
14198 Likewise, if the search fails, there are no words to count, so the
14199 function should return zero.
14201 On the other hand, if point is within the region and the search
14202 succeeds, the function should call itself again.
14205 Thus, the do-again-test should look like this:
14209 (and (< (point) region-end)
14210 (re-search-forward "\\w+\\W*" region-end t))
14214 Note that the search expression is part of the do-again-test---the
14215 function returns @code{t} if its search succeeds and @code{nil} if it
14216 fails. (@xref{Whitespace Bug, , The Whitespace Bug in
14217 @code{@value{COUNT-WORDS}}}, for an explanation of how
14218 @code{re-search-forward} works.)
14220 The do-again-test is the true-or-false test of an @code{if} clause.
14221 Clearly, if the do-again-test succeeds, the then-part of the @code{if}
14222 clause should call the function again; but if it fails, the else-part
14223 should return zero since either point is outside the region or the
14224 search failed because there were no words to find.
14226 But before considering the recursive call, we need to consider the
14227 next-step-expression. What is it? Interestingly, it is the search
14228 part of the do-again-test.
14230 In addition to returning @code{t} or @code{nil} for the
14231 do-again-test, @code{re-search-forward} moves point forward as a side
14232 effect of a successful search. This is the action that changes the
14233 value of point so that the recursive function stops calling itself
14234 when point completes its movement through the region. Consequently,
14235 the @code{re-search-forward} expression is the next-step-expression.
14238 In outline, then, the body of the @code{recursive-count-words}
14239 function looks like this:
14243 (if @var{do-again-test-and-next-step-combined}
14245 @var{recursive-call-returning-count}
14251 How to incorporate the mechanism that counts?
14253 If you are not used to writing recursive functions, a question like
14254 this can be troublesome. But it can and should be approached
14257 We know that the counting mechanism should be associated in some way
14258 with the recursive call. Indeed, since the next-step-expression moves
14259 point forward by one word, and since a recursive call is made for
14260 each word, the counting mechanism must be an expression that adds one
14261 to the value returned by a call to @code{recursive-count-words}.
14264 Consider several cases:
14268 If there are two words in the region, the function should return
14269 a value resulting from adding one to the value returned when it counts
14270 the first word, plus the number returned when it counts the remaining
14271 words in the region, which in this case is one.
14274 If there is one word in the region, the function should return
14275 a value resulting from adding one to the value returned when it counts
14276 that word, plus the number returned when it counts the remaining
14277 words in the region, which in this case is zero.
14280 If there are no words in the region, the function should return zero.
14283 From the sketch we can see that the else-part of the @code{if} returns
14284 zero for the case of no words. This means that the then-part of the
14285 @code{if} must return a value resulting from adding one to the value
14286 returned from a count of the remaining words.
14289 The expression will look like this, where @code{1+} is a function that
14290 adds one to its argument.
14293 (1+ (recursive-count-words region-end))
14297 The whole @code{recursive-count-words} function will then look like
14302 (defun recursive-count-words (region-end)
14303 "@var{documentation}@dots{}"
14305 ;;; @r{1. do-again-test}
14306 (if (and (< (point) region-end)
14307 (re-search-forward "\\w+\\W*" region-end t))
14311 ;;; @r{2. then-part: the recursive call}
14312 (1+ (recursive-count-words region-end))
14314 ;;; @r{3. else-part}
14320 Let's examine how this works:
14322 If there are no words in the region, the else part of the @code{if}
14323 expression is evaluated and consequently the function returns zero.
14325 If there is one word in the region, the value of point is less than
14326 the value of @code{region-end} and the search succeeds. In this case,
14327 the true-or-false-test of the @code{if} expression tests true, and the
14328 then-part of the @code{if} expression is evaluated. The counting
14329 expression is evaluated. This expression returns a value (which will
14330 be the value returned by the whole function) that is the sum of one
14331 added to the value returned by a recursive call.
14333 Meanwhile, the next-step-expression has caused point to jump over the
14334 first (and in this case only) word in the region. This means that
14335 when @code{(recursive-count-words region-end)} is evaluated a second
14336 time, as a result of the recursive call, the value of point will be
14337 equal to or greater than the value of region end. So this time,
14338 @code{recursive-count-words} will return zero. The zero will be added
14339 to one, and the original evaluation of @code{recursive-count-words}
14340 will return one plus zero, which is one, which is the correct amount.
14342 Clearly, if there are two words in the region, the first call to
14343 @code{recursive-count-words} returns one added to the value returned
14344 by calling @code{recursive-count-words} on a region containing the
14345 remaining word---that is, it adds one to one, producing two, which is
14346 the correct amount.
14348 Similarly, if there are three words in the region, the first call to
14349 @code{recursive-count-words} returns one added to the value returned
14350 by calling @code{recursive-count-words} on a region containing the
14351 remaining two words---and so on and so on.
14355 With full documentation the two functions look like this:
14359 The recursive function:
14361 @findex recursive-count-words
14364 (defun recursive-count-words (region-end)
14365 "Number of words between point and REGION-END."
14369 ;;; @r{1. do-again-test}
14370 (if (and (< (point) region-end)
14371 (re-search-forward "\\w+\\W*" region-end t))
14375 ;;; @r{2. then-part: the recursive call}
14376 (1+ (recursive-count-words region-end))
14378 ;;; @r{3. else-part}
14389 ;;; @r{Recursive version}
14390 (defun @value{COUNT-WORDS} (beginning end)
14391 "Print number of words in the region.
14395 Words are defined as at least one word-constituent
14396 character followed by at least one character that is
14397 not a word-constituent. The buffer's syntax table
14398 determines which characters these are."
14402 (message "Counting words in region ... ")
14404 (goto-char beginning)
14405 (let ((count (recursive-count-words end)))
14408 (cond ((zerop count)
14410 "The region does NOT have any words."))
14414 (message "The region has 1 word."))
14417 "The region has %d words." count))))))
14421 @node Counting Exercise
14422 @section Exercise: Counting Punctuation
14424 Using a @code{while} loop, write a function to count the number of
14425 punctuation marks in a region---period, comma, semicolon, colon,
14426 exclamation mark, and question mark. Do the same using recursion.
14428 @node Words in a defun
14429 @chapter Counting Words in a @code{defun}
14430 @cindex Counting words in a @code{defun}
14431 @cindex Word counting in a @code{defun}
14433 Our next project is to count the number of words in a function
14434 definition. Clearly, this can be done using some variant of
14435 @code{@value{COUNT-WORDS}}. @xref{Counting Words, , Counting Words:
14436 Repetition and Regexps}. If we are just going to count the words in
14437 one definition, it is easy enough to mark the definition with the
14438 @kbd{C-M-h} (@code{mark-defun}) command, and then call
14439 @code{@value{COUNT-WORDS}}.
14441 However, I am more ambitious: I want to count the words and symbols in
14442 every definition in the Emacs sources and then print a graph that
14443 shows how many functions there are of each length: how many contain 40
14444 to 49 words or symbols, how many contain 50 to 59 words or symbols,
14445 and so on. I have often been curious how long a typical function is,
14446 and this will tell.
14449 * Divide and Conquer::
14450 * Words and Symbols:: What to count?
14451 * Syntax:: What constitutes a word or symbol?
14452 * count-words-in-defun:: Very like @code{@value{COUNT-WORDS}}.
14453 * Several defuns:: Counting several defuns in a file.
14454 * Find a File:: Do you want to look at a file?
14455 * lengths-list-file:: A list of the lengths of many definitions.
14456 * Several files:: Counting in definitions in different files.
14457 * Several files recursively:: Recursively counting in different files.
14458 * Prepare the data:: Prepare the data for display in a graph.
14462 @node Divide and Conquer
14463 @unnumberedsec Divide and Conquer
14466 Described in one phrase, the histogram project is daunting; but
14467 divided into numerous small steps, each of which we can take one at a
14468 time, the project becomes less fearsome. Let us consider what the
14473 First, write a function to count the words in one definition. This
14474 includes the problem of handling symbols as well as words.
14477 Second, write a function to list the numbers of words in each function
14478 in a file. This function can use the @code{count-words-in-defun}
14482 Third, write a function to list the numbers of words in each function
14483 in each of several files. This entails automatically finding the
14484 various files, switching to them, and counting the words in the
14485 definitions within them.
14488 Fourth, write a function to convert the list of numbers that we
14489 created in step three to a form that will be suitable for printing as
14493 Fifth, write a function to print the results as a graph.
14496 This is quite a project! But if we take each step slowly, it will not
14499 @node Words and Symbols
14500 @section What to Count?
14501 @cindex Words and symbols in defun
14503 When we first start thinking about how to count the words in a
14504 function definition, the first question is (or ought to be) what are
14505 we going to count? When we speak of `words' with respect to a Lisp
14506 function definition, we are actually speaking, in large part, of
14507 `symbols'. For example, the following @code{multiply-by-seven}
14508 function contains the five symbols @code{defun},
14509 @code{multiply-by-seven}, @code{number}, @code{*}, and @code{7}. In
14510 addition, in the documentation string, it contains the four words
14511 @samp{Multiply}, @samp{NUMBER}, @samp{by}, and @samp{seven}. The
14512 symbol @samp{number} is repeated, so the definition contains a total
14513 of ten words and symbols.
14517 (defun multiply-by-seven (number)
14518 "Multiply NUMBER by seven."
14524 However, if we mark the @code{multiply-by-seven} definition with
14525 @kbd{C-M-h} (@code{mark-defun}), and then call
14526 @code{@value{COUNT-WORDS}} on it, we will find that
14527 @code{@value{COUNT-WORDS}} claims the definition has eleven words, not
14528 ten! Something is wrong!
14530 The problem is twofold: @code{@value{COUNT-WORDS}} does not count the
14531 @samp{*} as a word, and it counts the single symbol,
14532 @code{multiply-by-seven}, as containing three words. The hyphens are
14533 treated as if they were interword spaces rather than intraword
14534 connectors: @samp{multiply-by-seven} is counted as if it were written
14535 @samp{multiply by seven}.
14537 The cause of this confusion is the regular expression search within
14538 the @code{@value{COUNT-WORDS}} definition that moves point forward word
14539 by word. In the canonical version of @code{@value{COUNT-WORDS}}, the
14547 This regular expression is a pattern defining one or more word
14548 constituent characters possibly followed by one or more characters
14549 that are not word constituents. What is meant by `word constituent
14550 characters' brings us to the issue of syntax, which is worth a section
14554 @section What Constitutes a Word or Symbol?
14555 @cindex Syntax categories and tables
14557 Emacs treats different characters as belonging to different
14558 @dfn{syntax categories}. For example, the regular expression,
14559 @samp{\\w+}, is a pattern specifying one or more @emph{word
14560 constituent} characters. Word constituent characters are members of
14561 one syntax category. Other syntax categories include the class of
14562 punctuation characters, such as the period and the comma, and the
14563 class of whitespace characters, such as the blank space and the tab
14564 character. (For more information, @pxref{Syntax Tables, , Syntax
14565 Tables, elisp, The GNU Emacs Lisp Reference Manual}.)
14567 Syntax tables specify which characters belong to which categories.
14568 Usually, a hyphen is not specified as a `word constituent character'.
14569 Instead, it is specified as being in the `class of characters that are
14570 part of symbol names but not words.' This means that the
14571 @code{@value{COUNT-WORDS}} function treats it in the same way it treats
14572 an interword white space, which is why @code{@value{COUNT-WORDS}}
14573 counts @samp{multiply-by-seven} as three words.
14575 There are two ways to cause Emacs to count @samp{multiply-by-seven} as
14576 one symbol: modify the syntax table or modify the regular expression.
14578 We could redefine a hyphen as a word constituent character by
14579 modifying the syntax table that Emacs keeps for each mode. This
14580 action would serve our purpose, except that a hyphen is merely the
14581 most common character within symbols that is not typically a word
14582 constituent character; there are others, too.
14584 Alternatively, we can redefine the regexp used in the
14585 @code{@value{COUNT-WORDS}} definition so as to include symbols. This
14586 procedure has the merit of clarity, but the task is a little tricky.
14589 The first part is simple enough: the pattern must match ``at least one
14590 character that is a word or symbol constituent''. Thus:
14593 "\\(\\w\\|\\s_\\)+"
14597 The @samp{\\(} is the first part of the grouping construct that
14598 includes the @samp{\\w} and the @samp{\\s_} as alternatives, separated
14599 by the @samp{\\|}. The @samp{\\w} matches any word-constituent
14600 character and the @samp{\\s_} matches any character that is part of a
14601 symbol name but not a word-constituent character. The @samp{+}
14602 following the group indicates that the word or symbol constituent
14603 characters must be matched at least once.
14605 However, the second part of the regexp is more difficult to design.
14606 What we want is to follow the first part with ``optionally one or more
14607 characters that are not constituents of a word or symbol''. At first,
14608 I thought I could define this with the following:
14611 "\\(\\W\\|\\S_\\)*"
14615 The upper case @samp{W} and @samp{S} match characters that are
14616 @emph{not} word or symbol constituents. Unfortunately, this
14617 expression matches any character that is either not a word constituent
14618 or not a symbol constituent. This matches any character!
14620 I then noticed that every word or symbol in my test region was
14621 followed by white space (blank space, tab, or newline). So I tried
14622 placing a pattern to match one or more blank spaces after the pattern
14623 for one or more word or symbol constituents. This failed, too. Words
14624 and symbols are often separated by whitespace, but in actual code
14625 parentheses may follow symbols and punctuation may follow words. So
14626 finally, I designed a pattern in which the word or symbol constituents
14627 are followed optionally by characters that are not white space and
14628 then followed optionally by white space.
14631 Here is the full regular expression:
14634 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*"
14637 @node count-words-in-defun
14638 @section The @code{count-words-in-defun} Function
14639 @cindex Counting words in a @code{defun}
14641 We have seen that there are several ways to write a
14642 @code{count-words-region} function. To write a
14643 @code{count-words-in-defun}, we need merely adapt one of these
14646 The version that uses a @code{while} loop is easy to understand, so I
14647 am going to adapt that. Because @code{count-words-in-defun} will be
14648 part of a more complex program, it need not be interactive and it need
14649 not display a message but just return the count. These considerations
14650 simplify the definition a little.
14652 On the other hand, @code{count-words-in-defun} will be used within a
14653 buffer that contains function definitions. Consequently, it is
14654 reasonable to ask that the function determine whether it is called
14655 when point is within a function definition, and if it is, to return
14656 the count for that definition. This adds complexity to the
14657 definition, but saves us from needing to pass arguments to the
14661 These considerations lead us to prepare the following template:
14665 (defun count-words-in-defun ()
14666 "@var{documentation}@dots{}"
14667 (@var{set up}@dots{}
14668 (@var{while loop}@dots{})
14669 @var{return count})
14674 As usual, our job is to fill in the slots.
14678 We are presuming that this function will be called within a buffer
14679 containing function definitions. Point will either be within a
14680 function definition or not. For @code{count-words-in-defun} to work,
14681 point must move to the beginning of the definition, a counter must
14682 start at zero, and the counting loop must stop when point reaches the
14683 end of the definition.
14685 The @code{beginning-of-defun} function searches backwards for an
14686 opening delimiter such as a @samp{(} at the beginning of a line, and
14687 moves point to that position, or else to the limit of the search. In
14688 practice, this means that @code{beginning-of-defun} moves point to the
14689 beginning of an enclosing or preceding function definition, or else to
14690 the beginning of the buffer. We can use @code{beginning-of-defun} to
14691 place point where we wish to start.
14693 The @code{while} loop requires a counter to keep track of the words or
14694 symbols being counted. A @code{let} expression can be used to create
14695 a local variable for this purpose, and bind it to an initial value of zero.
14697 The @code{end-of-defun} function works like @code{beginning-of-defun}
14698 except that it moves point to the end of the definition.
14699 @code{end-of-defun} can be used as part of an expression that
14700 determines the position of the end of the definition.
14702 The set up for @code{count-words-in-defun} takes shape rapidly: first
14703 we move point to the beginning of the definition, then we create a
14704 local variable to hold the count, and finally, we record the position
14705 of the end of the definition so the @code{while} loop will know when to stop
14709 The code looks like this:
14713 (beginning-of-defun)
14715 (end (save-excursion (end-of-defun) (point))))
14720 The code is simple. The only slight complication is likely to concern
14721 @code{end}: it is bound to the position of the end of the definition
14722 by a @code{save-excursion} expression that returns the value of point
14723 after @code{end-of-defun} temporarily moves it to the end of the
14726 The second part of the @code{count-words-in-defun}, after the set up,
14727 is the @code{while} loop.
14729 The loop must contain an expression that jumps point forward word by
14730 word and symbol by symbol, and another expression that counts the
14731 jumps. The true-or-false-test for the @code{while} loop should test
14732 true so long as point should jump forward, and false when point is at
14733 the end of the definition. We have already redefined the regular
14734 expression for this, so the loop is straightforward:
14738 (while (and (< (point) end)
14740 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*" end t))
14741 (setq count (1+ count)))
14745 The third part of the function definition returns the count of words
14746 and symbols. This part is the last expression within the body of the
14747 @code{let} expression, and can be, very simply, the local variable
14748 @code{count}, which when evaluated returns the count.
14751 Put together, the @code{count-words-in-defun} definition looks like this:
14753 @findex count-words-in-defun
14756 (defun count-words-in-defun ()
14757 "Return the number of words and symbols in a defun."
14758 (beginning-of-defun)
14760 (end (save-excursion (end-of-defun) (point))))
14764 (and (< (point) end)
14766 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*"
14768 (setq count (1+ count)))
14773 How to test this? The function is not interactive, but it is easy to
14774 put a wrapper around the function to make it interactive; we can use
14775 almost the same code as for the recursive version of
14776 @code{@value{COUNT-WORDS}}:
14780 ;;; @r{Interactive version.}
14781 (defun count-words-defun ()
14782 "Number of words and symbols in a function definition."
14785 "Counting words and symbols in function definition ... ")
14788 (let ((count (count-words-in-defun)))
14792 "The definition does NOT have any words or symbols."))
14797 "The definition has 1 word or symbol."))
14800 "The definition has %d words or symbols." count)))))
14806 Let's re-use @kbd{C-c =} as a convenient keybinding:
14809 (global-set-key "\C-c=" 'count-words-defun)
14812 Now we can try out @code{count-words-defun}: install both
14813 @code{count-words-in-defun} and @code{count-words-defun}, and set the
14814 keybinding, and then place the cursor within the following definition:
14818 (defun multiply-by-seven (number)
14819 "Multiply NUMBER by seven."
14826 Success! The definition has 10 words and symbols.
14828 The next problem is to count the numbers of words and symbols in
14829 several definitions within a single file.
14831 @node Several defuns
14832 @section Count Several @code{defuns} Within a File
14834 A file such as @file{simple.el} may have a hundred or more function
14835 definitions within it. Our long term goal is to collect statistics on
14836 many files, but as a first step, our immediate goal is to collect
14837 statistics on one file.
14839 The information will be a series of numbers, each number being the
14840 length of a function definition. We can store the numbers in a list.
14842 We know that we will want to incorporate the information regarding one
14843 file with information about many other files; this means that the
14844 function for counting definition lengths within one file need only
14845 return the list of lengths. It need not and should not display any
14848 The word count commands contain one expression to jump point forward
14849 word by word and another expression to count the jumps. The function
14850 to return the lengths of definitions can be designed to work the same
14851 way, with one expression to jump point forward definition by
14852 definition and another expression to construct the lengths' list.
14854 This statement of the problem makes it elementary to write the
14855 function definition. Clearly, we will start the count at the
14856 beginning of the file, so the first command will be @code{(goto-char
14857 (point-min))}. Next, we start the @code{while} loop; and the
14858 true-or-false test of the loop can be a regular expression search for
14859 the next function definition---so long as the search succeeds, point
14860 is moved forward and then the body of the loop is evaluated. The body
14861 needs an expression that constructs the lengths' list. @code{cons},
14862 the list construction command, can be used to create the list. That
14863 is almost all there is to it.
14866 Here is what this fragment of code looks like:
14870 (goto-char (point-min))
14871 (while (re-search-forward "^(defun" nil t)
14873 (cons (count-words-in-defun) lengths-list)))
14877 What we have left out is the mechanism for finding the file that
14878 contains the function definitions.
14880 In previous examples, we either used this, the Info file, or we
14881 switched back and forth to some other buffer, such as the
14882 @file{*scratch*} buffer.
14884 Finding a file is a new process that we have not yet discussed.
14887 @section Find a File
14888 @cindex Find a File
14890 To find a file in Emacs, you use the @kbd{C-x C-f} (@code{find-file})
14891 command. This command is almost, but not quite right for the lengths
14895 Let's look at the source for @code{find-file}:
14899 (defun find-file (filename)
14900 "Edit file FILENAME.
14901 Switch to a buffer visiting file FILENAME,
14902 creating one if none already exists."
14903 (interactive "FFind file: ")
14904 (switch-to-buffer (find-file-noselect filename)))
14909 (The most recent version of the @code{find-file} function definition
14910 permits you to specify optional wildcards to visit multiple files; that
14911 makes the definition more complex and we will not discuss it here,
14912 since it is not relevant. You can see its source using either
14913 @kbd{M-.} (@code{find-tag}) or @kbd{C-h f} (@code{describe-function}).)
14917 (defun find-file (filename &optional wildcards)
14918 "Edit file FILENAME.
14919 Switch to a buffer visiting file FILENAME,
14920 creating one if none already exists.
14921 Interactively, the default if you just type RET is the current directory,
14922 but the visited file name is available through the minibuffer history:
14923 type M-n to pull it into the minibuffer.
14925 Interactively, or if WILDCARDS is non-nil in a call from Lisp,
14926 expand wildcards (if any) and visit multiple files. You can
14927 suppress wildcard expansion by setting `find-file-wildcards' to nil.
14929 To visit a file without any kind of conversion and without
14930 automatically choosing a major mode, use \\[find-file-literally]."
14931 (interactive (find-file-read-args "Find file: " nil))
14932 (let ((value (find-file-noselect filename nil nil wildcards)))
14934 (mapcar 'switch-to-buffer (nreverse value))
14935 (switch-to-buffer value))))
14938 The definition I am showing possesses short but complete documentation
14939 and an interactive specification that prompts you for a file name when
14940 you use the command interactively. The body of the definition
14941 contains two functions, @code{find-file-noselect} and
14942 @code{switch-to-buffer}.
14944 According to its documentation as shown by @kbd{C-h f} (the
14945 @code{describe-function} command), the @code{find-file-noselect}
14946 function reads the named file into a buffer and returns the buffer.
14947 (Its most recent version includes an optional wildcards argument,
14948 too, as well as another to read a file literally and an other you
14949 suppress warning messages. These optional arguments are irrelevant.)
14951 However, the @code{find-file-noselect} function does not select the
14952 buffer in which it puts the file. Emacs does not switch its attention
14953 (or yours if you are using @code{find-file-noselect}) to the selected
14954 buffer. That is what @code{switch-to-buffer} does: it switches the
14955 buffer to which Emacs attention is directed; and it switches the
14956 buffer displayed in the window to the new buffer. We have discussed
14957 buffer switching elsewhere. (@xref{Switching Buffers}.)
14959 In this histogram project, we do not need to display each file on the
14960 screen as the program determines the length of each definition within
14961 it. Instead of employing @code{switch-to-buffer}, we can work with
14962 @code{set-buffer}, which redirects the attention of the computer
14963 program to a different buffer but does not redisplay it on the screen.
14964 So instead of calling on @code{find-file} to do the job, we must write
14965 our own expression.
14967 The task is easy: use @code{find-file-noselect} and @code{set-buffer}.
14969 @node lengths-list-file
14970 @section @code{lengths-list-file} in Detail
14972 The core of the @code{lengths-list-file} function is a @code{while}
14973 loop containing a function to move point forward `defun by defun' and
14974 a function to count the number of words and symbols in each defun.
14975 This core must be surrounded by functions that do various other tasks,
14976 including finding the file, and ensuring that point starts out at the
14977 beginning of the file. The function definition looks like this:
14978 @findex lengths-list-file
14982 (defun lengths-list-file (filename)
14983 "Return list of definitions' lengths within FILE.
14984 The returned list is a list of numbers.
14985 Each number is the number of words or
14986 symbols in one function definition."
14989 (message "Working on `%s' ... " filename)
14991 (let ((buffer (find-file-noselect filename))
14993 (set-buffer buffer)
14994 (setq buffer-read-only t)
14996 (goto-char (point-min))
14997 (while (re-search-forward "^(defun" nil t)
14999 (cons (count-words-in-defun) lengths-list)))
15000 (kill-buffer buffer)
15006 The function is passed one argument, the name of the file on which it
15007 will work. It has four lines of documentation, but no interactive
15008 specification. Since people worry that a computer is broken if they
15009 don't see anything going on, the first line of the body is a
15012 The next line contains a @code{save-excursion} that returns Emacs's
15013 attention to the current buffer when the function completes. This is
15014 useful in case you embed this function in another function that
15015 presumes point is restored to the original buffer.
15017 In the varlist of the @code{let} expression, Emacs finds the file and
15018 binds the local variable @code{buffer} to the buffer containing the
15019 file. At the same time, Emacs creates @code{lengths-list} as a local
15022 Next, Emacs switches its attention to the buffer.
15024 In the following line, Emacs makes the buffer read-only. Ideally,
15025 this line is not necessary. None of the functions for counting words
15026 and symbols in a function definition should change the buffer.
15027 Besides, the buffer is not going to be saved, even if it were changed.
15028 This line is entirely the consequence of great, perhaps excessive,
15029 caution. The reason for the caution is that this function and those
15030 it calls work on the sources for Emacs and it is inconvenient if they
15031 are inadvertently modified. It goes without saying that I did not
15032 realize a need for this line until an experiment went awry and started
15033 to modify my Emacs source files @dots{}
15035 Next comes a call to widen the buffer if it is narrowed. This
15036 function is usually not needed---Emacs creates a fresh buffer if none
15037 already exists; but if a buffer visiting the file already exists Emacs
15038 returns that one. In this case, the buffer may be narrowed and must
15039 be widened. If we wanted to be fully `user-friendly', we would
15040 arrange to save the restriction and the location of point, but we
15043 The @code{(goto-char (point-min))} expression moves point to the
15044 beginning of the buffer.
15046 Then comes a @code{while} loop in which the `work' of the function is
15047 carried out. In the loop, Emacs determines the length of each
15048 definition and constructs a lengths' list containing the information.
15050 Emacs kills the buffer after working through it. This is to save
15051 space inside of Emacs. My version of GNU Emacs 19 contained over 300
15052 source files of interest; GNU Emacs 22 contains over a thousand source
15053 files. Another function will apply @code{lengths-list-file} to each
15056 Finally, the last expression within the @code{let} expression is the
15057 @code{lengths-list} variable; its value is returned as the value of
15058 the whole function.
15060 You can try this function by installing it in the usual fashion. Then
15061 place your cursor after the following expression and type @kbd{C-x
15062 C-e} (@code{eval-last-sexp}).
15064 @c !!! 22.1.1 lisp sources location here
15067 "/usr/local/share/emacs/22.1.1/lisp/emacs-lisp/debug.el")
15071 (You may need to change the pathname of the file; the one here is for
15072 GNU Emacs version 22.1.1. To change the expression, copy it to
15073 the @file{*scratch*} buffer and edit it.
15077 (Also, to see the full length of the list, rather than a truncated
15078 version, you may have to evaluate the following:
15081 (custom-set-variables '(eval-expression-print-length nil))
15085 (@xref{defcustom, , Specifying Variables using @code{defcustom}}.
15086 Then evaluate the @code{lengths-list-file} expression.)
15089 The lengths' list for @file{debug.el} takes less than a second to
15090 produce and looks like this in GNU Emacs 22:
15093 (83 113 105 144 289 22 30 97 48 89 25 52 52 88 28 29 77 49 43 290 232 587)
15097 (Using my old machine, the version 19 lengths' list for @file{debug.el}
15098 took seven seconds to produce and looked like this:
15101 (75 41 80 62 20 45 44 68 45 12 34 235)
15104 (The newer version of @file{debug.el} contains more defuns than the
15105 earlier one; and my new machine is much faster than the old one.)
15107 Note that the length of the last definition in the file is first in
15110 @node Several files
15111 @section Count Words in @code{defuns} in Different Files
15113 In the previous section, we created a function that returns a list of
15114 the lengths of each definition in a file. Now, we want to define a
15115 function to return a master list of the lengths of the definitions in
15118 Working on each of a list of files is a repetitious act, so we can use
15119 either a @code{while} loop or recursion.
15122 * lengths-list-many-files:: Return a list of the lengths of defuns.
15123 * append:: Attach one list to another.
15127 @node lengths-list-many-files
15128 @unnumberedsubsec Determine the lengths of @code{defuns}
15131 The design using a @code{while} loop is routine. The argument passed
15132 the function is a list of files. As we saw earlier (@pxref{Loop
15133 Example}), you can write a @code{while} loop so that the body of the
15134 loop is evaluated if such a list contains elements, but to exit the
15135 loop if the list is empty. For this design to work, the body of the
15136 loop must contain an expression that shortens the list each time the
15137 body is evaluated, so that eventually the list is empty. The usual
15138 technique is to set the value of the list to the value of the @sc{cdr}
15139 of the list each time the body is evaluated.
15142 The template looks like this:
15146 (while @var{test-whether-list-is-empty}
15148 @var{set-list-to-cdr-of-list})
15152 Also, we remember that a @code{while} loop returns @code{nil} (the
15153 result of evaluating the true-or-false-test), not the result of any
15154 evaluation within its body. (The evaluations within the body of the
15155 loop are done for their side effects.) However, the expression that
15156 sets the lengths' list is part of the body---and that is the value
15157 that we want returned by the function as a whole. To do this, we
15158 enclose the @code{while} loop within a @code{let} expression, and
15159 arrange that the last element of the @code{let} expression contains
15160 the value of the lengths' list. (@xref{Incrementing Example, , Loop
15161 Example with an Incrementing Counter}.)
15163 @findex lengths-list-many-files
15165 These considerations lead us directly to the function itself:
15169 ;;; @r{Use @code{while} loop.}
15170 (defun lengths-list-many-files (list-of-files)
15171 "Return list of lengths of defuns in LIST-OF-FILES."
15174 (let (lengths-list)
15176 ;;; @r{true-or-false-test}
15177 (while list-of-files
15182 ;;; @r{Generate a lengths' list.}
15184 (expand-file-name (car list-of-files)))))
15188 ;;; @r{Make files' list shorter.}
15189 (setq list-of-files (cdr list-of-files)))
15191 ;;; @r{Return final value of lengths' list.}
15196 @code{expand-file-name} is a built-in function that converts a file
15197 name to the absolute, long, path name form. The function employs the
15198 name of the directory in which the function is called.
15200 @c !!! 22.1.1 lisp sources location here
15202 Thus, if @code{expand-file-name} is called on @code{debug.el} when
15203 Emacs is visiting the
15204 @file{/usr/local/share/emacs/22.1.1/lisp/emacs-lisp/} directory,
15214 @c !!! 22.1.1 lisp sources location here
15216 /usr/local/share/emacs/22.1.1/lisp/emacs-lisp/debug.el
15219 The only other new element of this function definition is the as yet
15220 unstudied function @code{append}, which merits a short section for
15224 @subsection The @code{append} Function
15227 The @code{append} function attaches one list to another. Thus,
15230 (append '(1 2 3 4) '(5 6 7 8))
15241 This is exactly how we want to attach two lengths' lists produced by
15242 @code{lengths-list-file} to each other. The results contrast with
15246 (cons '(1 2 3 4) '(5 6 7 8))
15251 which constructs a new list in which the first argument to @code{cons}
15252 becomes the first element of the new list:
15255 ((1 2 3 4) 5 6 7 8)
15258 @node Several files recursively
15259 @section Recursively Count Words in Different Files
15261 Besides a @code{while} loop, you can work on each of a list of files
15262 with recursion. A recursive version of @code{lengths-list-many-files}
15263 is short and simple.
15265 The recursive function has the usual parts: the `do-again-test', the
15266 `next-step-expression', and the recursive call. The `do-again-test'
15267 determines whether the function should call itself again, which it
15268 will do if the @code{list-of-files} contains any remaining elements;
15269 the `next-step-expression' resets the @code{list-of-files} to the
15270 @sc{cdr} of itself, so eventually the list will be empty; and the
15271 recursive call calls itself on the shorter list. The complete
15272 function is shorter than this description!
15273 @findex recursive-lengths-list-many-files
15277 (defun recursive-lengths-list-many-files (list-of-files)
15278 "Return list of lengths of each defun in LIST-OF-FILES."
15279 (if list-of-files ; @r{do-again-test}
15282 (expand-file-name (car list-of-files)))
15283 (recursive-lengths-list-many-files
15284 (cdr list-of-files)))))
15289 In a sentence, the function returns the lengths' list for the first of
15290 the @code{list-of-files} appended to the result of calling itself on
15291 the rest of the @code{list-of-files}.
15293 Here is a test of @code{recursive-lengths-list-many-files}, along with
15294 the results of running @code{lengths-list-file} on each of the files
15297 Install @code{recursive-lengths-list-many-files} and
15298 @code{lengths-list-file}, if necessary, and then evaluate the
15299 following expressions. You may need to change the files' pathnames;
15300 those here work when this Info file and the Emacs sources are located
15301 in their customary places. To change the expressions, copy them to
15302 the @file{*scratch*} buffer, edit them, and then evaluate them.
15304 The results are shown after the @samp{@result{}}. (These results are
15305 for files from Emacs version 22.1.1; files from other versions of
15306 Emacs may produce different results.)
15308 @c !!! 22.1.1 lisp sources location here
15311 (cd "/usr/local/share/emacs/22.1.1/")
15313 (lengths-list-file "./lisp/macros.el")
15314 @result{} (283 263 480 90)
15318 (lengths-list-file "./lisp/mail/mailalias.el")
15319 @result{} (38 32 29 95 178 180 321 218 324)
15323 (lengths-list-file "./lisp/makesum.el")
15328 (recursive-lengths-list-many-files
15329 '("./lisp/macros.el"
15330 "./lisp/mail/mailalias.el"
15331 "./lisp/makesum.el"))
15332 @result{} (283 263 480 90 38 32 29 95 178 180 321 218 324 85 181)
15336 The @code{recursive-lengths-list-many-files} function produces the
15339 The next step is to prepare the data in the list for display in a graph.
15341 @node Prepare the data
15342 @section Prepare the Data for Display in a Graph
15344 The @code{recursive-lengths-list-many-files} function returns a list
15345 of numbers. Each number records the length of a function definition.
15346 What we need to do now is transform this data into a list of numbers
15347 suitable for generating a graph. The new list will tell how many
15348 functions definitions contain less than 10 words and
15349 symbols, how many contain between 10 and 19 words and symbols, how
15350 many contain between 20 and 29 words and symbols, and so on.
15352 In brief, we need to go through the lengths' list produced by the
15353 @code{recursive-lengths-list-many-files} function and count the number
15354 of defuns within each range of lengths, and produce a list of those
15358 * Data for Display in Detail::
15359 * Sorting:: Sorting lists.
15360 * Files List:: Making a list of files.
15361 * Counting function definitions::
15365 @node Data for Display in Detail
15366 @unnumberedsubsec The Data for Display in Detail
15369 Based on what we have done before, we can readily foresee that it
15370 should not be too hard to write a function that `@sc{cdr}s' down the
15371 lengths' list, looks at each element, determines which length range it
15372 is in, and increments a counter for that range.
15374 However, before beginning to write such a function, we should consider
15375 the advantages of sorting the lengths' list first, so the numbers are
15376 ordered from smallest to largest. First, sorting will make it easier
15377 to count the numbers in each range, since two adjacent numbers will
15378 either be in the same length range or in adjacent ranges. Second, by
15379 inspecting a sorted list, we can discover the highest and lowest
15380 number, and thereby determine the largest and smallest length range
15384 @subsection Sorting Lists
15387 Emacs contains a function to sort lists, called (as you might guess)
15388 @code{sort}. The @code{sort} function takes two arguments, the list
15389 to be sorted, and a predicate that determines whether the first of
15390 two list elements is ``less'' than the second.
15392 As we saw earlier (@pxref{Wrong Type of Argument, , Using the Wrong
15393 Type Object as an Argument}), a predicate is a function that
15394 determines whether some property is true or false. The @code{sort}
15395 function will reorder a list according to whatever property the
15396 predicate uses; this means that @code{sort} can be used to sort
15397 non-numeric lists by non-numeric criteria---it can, for example,
15398 alphabetize a list.
15401 The @code{<} function is used when sorting a numeric list. For example,
15404 (sort '(4 8 21 17 33 7 21 7) '<)
15412 (4 7 7 8 17 21 21 33)
15416 (Note that in this example, both the arguments are quoted so that the
15417 symbols are not evaluated before being passed to @code{sort} as
15420 Sorting the list returned by the
15421 @code{recursive-lengths-list-many-files} function is straightforward;
15422 it uses the @code{<} function:
15426 In GNU Emacs 22, eval
15428 (cd "/usr/local/share/emacs/22.0.50/")
15430 (recursive-lengths-list-many-files
15431 '("./lisp/macros.el"
15432 "./lisp/mail/mailalias.el"
15433 "./lisp/makesum.el"))
15441 (recursive-lengths-list-many-files
15442 '("./lisp/macros.el"
15443 "./lisp/mailalias.el"
15444 "./lisp/makesum.el"))
15454 (29 32 38 85 90 95 178 180 181 218 263 283 321 324 480)
15458 (Note that in this example, the first argument to @code{sort} is not
15459 quoted, since the expression must be evaluated so as to produce the
15460 list that is passed to @code{sort}.)
15463 @subsection Making a List of Files
15465 The @code{recursive-lengths-list-many-files} function requires a list
15466 of files as its argument. For our test examples, we constructed such
15467 a list by hand; but the Emacs Lisp source directory is too large for
15468 us to do for that. Instead, we will write a function to do the job
15469 for us. In this function, we will use both a @code{while} loop and a
15472 @findex directory-files
15473 We did not have to write a function like this for older versions of
15474 GNU Emacs, since they placed all the @samp{.el} files in one
15475 directory. Instead, we were able to use the @code{directory-files}
15476 function, which lists the names of files that match a specified
15477 pattern within a single directory.
15479 However, recent versions of Emacs place Emacs Lisp files in
15480 sub-directories of the top level @file{lisp} directory. This
15481 re-arrangement eases navigation. For example, all the mail related
15482 files are in a @file{lisp} sub-directory called @file{mail}. But at
15483 the same time, this arrangement forces us to create a file listing
15484 function that descends into the sub-directories.
15486 @findex files-in-below-directory
15487 We can create this function, called @code{files-in-below-directory},
15488 using familiar functions such as @code{car}, @code{nthcdr}, and
15489 @code{substring} in conjunction with an existing function called
15490 @code{directory-files-and-attributes}. This latter function not only
15491 lists all the filenames in a directory, including the names
15492 of sub-directories, but also their attributes.
15494 To restate our goal: to create a function that will enable us
15495 to feed filenames to @code{recursive-lengths-list-many-files}
15496 as a list that looks like this (but with more elements):
15500 ("./lisp/macros.el"
15501 "./lisp/mail/rmail.el"
15502 "./lisp/makesum.el")
15506 The @code{directory-files-and-attributes} function returns a list of
15507 lists. Each of the lists within the main list consists of 13
15508 elements. The first element is a string that contains the name of the
15509 file---which, in GNU/Linux, may be a `directory file', that is to
15510 say, a file with the special attributes of a directory. The second
15511 element of the list is @code{t} for a directory, a string
15512 for symbolic link (the string is the name linked to), or @code{nil}.
15514 For example, the first @samp{.el} file in the @file{lisp/} directory
15515 is @file{abbrev.el}. Its name is
15516 @file{/usr/local/share/emacs/22.1.1/lisp/abbrev.el} and it is not a
15517 directory or a symbolic link.
15520 This is how @code{directory-files-and-attributes} lists that file and
15532 (20615 27034 579989 697000)
15534 (20615 26327 734791 805000)
15546 On the other hand, @file{mail/} is a directory within the @file{lisp/}
15547 directory. The beginning of its listing looks like this:
15558 (To learn about the different attributes, look at the documentation of
15559 @code{file-attributes}. Bear in mind that the @code{file-attributes}
15560 function does not list the filename, so its first element is
15561 @code{directory-files-and-attributes}'s second element.)
15563 We will want our new function, @code{files-in-below-directory}, to
15564 list the @samp{.el} files in the directory it is told to check, and in
15565 any directories below that directory.
15567 This gives us a hint on how to construct
15568 @code{files-in-below-directory}: within a directory, the function
15569 should add @samp{.el} filenames to a list; and if, within a directory,
15570 the function comes upon a sub-directory, it should go into that
15571 sub-directory and repeat its actions.
15573 However, we should note that every directory contains a name that
15574 refers to itself, called @file{.}, (``dot'') and a name that refers to
15575 its parent directory, called @file{..} (``double dot''). (In
15576 @file{/}, the root directory, @file{..} refers to itself, since
15577 @file{/} has no parent.) Clearly, we do not want our
15578 @code{files-in-below-directory} function to enter those directories,
15579 since they always lead us, directly or indirectly, to the current
15582 Consequently, our @code{files-in-below-directory} function must do
15587 Check to see whether it is looking at a filename that ends in
15588 @samp{.el}; and if so, add its name to a list.
15591 Check to see whether it is looking at a filename that is the name of a
15592 directory; and if so,
15596 Check to see whether it is looking at @file{.} or @file{..}; and if
15600 Or else, go into that directory and repeat the process.
15604 Let's write a function definition to do these tasks. We will use a
15605 @code{while} loop to move from one filename to another within a
15606 directory, checking what needs to be done; and we will use a recursive
15607 call to repeat the actions on each sub-directory. The recursive
15608 pattern is `accumulate'
15609 (@pxref{Accumulate, , Recursive Pattern: @emph{accumulate}}),
15610 using @code{append} as the combiner.
15613 (directory-files "/usr/local/src/emacs/lisp/" t "\\.el$")
15614 (shell-command "find /usr/local/src/emacs/lisp/ -name '*.el'")
15616 (directory-files "/usr/local/share/emacs/22.1.1/lisp/" t "\\.el$")
15617 (shell-command "find /usr/local/share/emacs/22.1.1/lisp/ -name '*.el'")
15620 @c /usr/local/share/emacs/22.1.1/lisp/
15623 Here is the function:
15627 (defun files-in-below-directory (directory)
15628 "List the .el files in DIRECTORY and in its sub-directories."
15629 ;; Although the function will be used non-interactively,
15630 ;; it will be easier to test if we make it interactive.
15631 ;; The directory will have a name such as
15632 ;; "/usr/local/share/emacs/22.1.1/lisp/"
15633 (interactive "DDirectory name: ")
15636 (let (el-files-list
15637 (current-directory-list
15638 (directory-files-and-attributes directory t)))
15639 ;; while we are in the current directory
15640 (while current-directory-list
15644 ;; check to see whether filename ends in `.el'
15645 ;; and if so, append its name to a list.
15646 ((equal ".el" (substring (car (car current-directory-list)) -3))
15647 (setq el-files-list
15648 (cons (car (car current-directory-list)) el-files-list)))
15651 ;; check whether filename is that of a directory
15652 ((eq t (car (cdr (car current-directory-list))))
15653 ;; decide whether to skip or recurse
15656 (substring (car (car current-directory-list)) -1))
15657 ;; then do nothing since filename is that of
15658 ;; current directory or parent, "." or ".."
15662 ;; else descend into the directory and repeat the process
15663 (setq el-files-list
15665 (files-in-below-directory
15666 (car (car current-directory-list)))
15668 ;; move to the next filename in the list; this also
15669 ;; shortens the list so the while loop eventually comes to an end
15670 (setq current-directory-list (cdr current-directory-list)))
15671 ;; return the filenames
15676 @c (files-in-below-directory "/usr/local/src/emacs/lisp/")
15677 @c (files-in-below-directory "/usr/local/share/emacs/22.1.1/lisp/")
15679 The @code{files-in-below-directory} @code{directory-files} function
15680 takes one argument, the name of a directory.
15683 Thus, on my system,
15685 @c (length (files-in-below-directory "/usr/local/src/emacs/lisp/"))
15687 @c !!! 22.1.1 lisp sources location here
15691 (files-in-below-directory "/usr/local/share/emacs/22.1.1/lisp/"))
15696 tells me that in and below my Lisp sources directory are 1031
15699 @code{files-in-below-directory} returns a list in reverse alphabetical
15700 order. An expression to sort the list in alphabetical order looks
15706 (files-in-below-directory "/usr/local/share/emacs/22.1.1/lisp/")
15713 "Test how long it takes to find lengths of all sorted elisp defuns."
15714 (insert "\n" (current-time-string) "\n")
15717 (recursive-lengths-list-many-files
15718 (files-in-below-directory "/usr/local/src/emacs/lisp/"))
15720 (insert (format "%s" (current-time-string))))
15723 @node Counting function definitions
15724 @subsection Counting function definitions
15726 Our immediate goal is to generate a list that tells us how many
15727 function definitions contain fewer than 10 words and symbols, how many
15728 contain between 10 and 19 words and symbols, how many contain between
15729 20 and 29 words and symbols, and so on.
15731 With a sorted list of numbers, this is easy: count how many elements
15732 of the list are smaller than 10, then, after moving past the numbers
15733 just counted, count how many are smaller than 20, then, after moving
15734 past the numbers just counted, count how many are smaller than 30, and
15735 so on. Each of the numbers, 10, 20, 30, 40, and the like, is one
15736 larger than the top of that range. We can call the list of such
15737 numbers the @code{top-of-ranges} list.
15740 If we wished, we could generate this list automatically, but it is
15741 simpler to write a list manually. Here it is:
15742 @vindex top-of-ranges
15746 (defvar top-of-ranges
15749 110 120 130 140 150
15750 160 170 180 190 200
15751 210 220 230 240 250
15752 260 270 280 290 300)
15753 "List specifying ranges for `defuns-per-range'.")
15757 To change the ranges, we edit this list.
15759 Next, we need to write the function that creates the list of the
15760 number of definitions within each range. Clearly, this function must
15761 take the @code{sorted-lengths} and the @code{top-of-ranges} lists
15764 The @code{defuns-per-range} function must do two things again and
15765 again: it must count the number of definitions within a range
15766 specified by the current top-of-range value; and it must shift to the
15767 next higher value in the @code{top-of-ranges} list after counting the
15768 number of definitions in the current range. Since each of these
15769 actions is repetitive, we can use @code{while} loops for the job.
15770 One loop counts the number of definitions in the range defined by the
15771 current top-of-range value, and the other loop selects each of the
15772 top-of-range values in turn.
15774 Several entries of the @code{sorted-lengths} list are counted for each
15775 range; this means that the loop for the @code{sorted-lengths} list
15776 will be inside the loop for the @code{top-of-ranges} list, like a
15777 small gear inside a big gear.
15779 The inner loop counts the number of definitions within the range. It
15780 is a simple counting loop of the type we have seen before.
15781 (@xref{Incrementing Loop, , A loop with an incrementing counter}.)
15782 The true-or-false test of the loop tests whether the value from the
15783 @code{sorted-lengths} list is smaller than the current value of the
15784 top of the range. If it is, the function increments the counter and
15785 tests the next value from the @code{sorted-lengths} list.
15788 The inner loop looks like this:
15792 (while @var{length-element-smaller-than-top-of-range}
15793 (setq number-within-range (1+ number-within-range))
15794 (setq sorted-lengths (cdr sorted-lengths)))
15798 The outer loop must start with the lowest value of the
15799 @code{top-of-ranges} list, and then be set to each of the succeeding
15800 higher values in turn. This can be done with a loop like this:
15804 (while top-of-ranges
15805 @var{body-of-loop}@dots{}
15806 (setq top-of-ranges (cdr top-of-ranges)))
15811 Put together, the two loops look like this:
15815 (while top-of-ranges
15817 ;; @r{Count the number of elements within the current range.}
15818 (while @var{length-element-smaller-than-top-of-range}
15819 (setq number-within-range (1+ number-within-range))
15820 (setq sorted-lengths (cdr sorted-lengths)))
15822 ;; @r{Move to next range.}
15823 (setq top-of-ranges (cdr top-of-ranges)))
15827 In addition, in each circuit of the outer loop, Emacs should record
15828 the number of definitions within that range (the value of
15829 @code{number-within-range}) in a list. We can use @code{cons} for
15830 this purpose. (@xref{cons, , @code{cons}}.)
15832 The @code{cons} function works fine, except that the list it
15833 constructs will contain the number of definitions for the highest
15834 range at its beginning and the number of definitions for the lowest
15835 range at its end. This is because @code{cons} attaches new elements
15836 of the list to the beginning of the list, and since the two loops are
15837 working their way through the lengths' list from the lower end first,
15838 the @code{defuns-per-range-list} will end up largest number first.
15839 But we will want to print our graph with smallest values first and the
15840 larger later. The solution is to reverse the order of the
15841 @code{defuns-per-range-list}. We can do this using the
15842 @code{nreverse} function, which reverses the order of a list.
15849 (nreverse '(1 2 3 4))
15860 Note that the @code{nreverse} function is ``destructive''---that is,
15861 it changes the list to which it is applied; this contrasts with the
15862 @code{car} and @code{cdr} functions, which are non-destructive. In
15863 this case, we do not want the original @code{defuns-per-range-list},
15864 so it does not matter that it is destroyed. (The @code{reverse}
15865 function provides a reversed copy of a list, leaving the original list
15870 Put all together, the @code{defuns-per-range} looks like this:
15874 (defun defuns-per-range (sorted-lengths top-of-ranges)
15875 "SORTED-LENGTHS defuns in each TOP-OF-RANGES range."
15876 (let ((top-of-range (car top-of-ranges))
15877 (number-within-range 0)
15878 defuns-per-range-list)
15883 (while top-of-ranges
15889 ;; @r{Need number for numeric test.}
15890 (car sorted-lengths)
15891 (< (car sorted-lengths) top-of-range))
15895 ;; @r{Count number of definitions within current range.}
15896 (setq number-within-range (1+ number-within-range))
15897 (setq sorted-lengths (cdr sorted-lengths)))
15899 ;; @r{Exit inner loop but remain within outer loop.}
15903 (setq defuns-per-range-list
15904 (cons number-within-range defuns-per-range-list))
15905 (setq number-within-range 0) ; @r{Reset count to zero.}
15909 ;; @r{Move to next range.}
15910 (setq top-of-ranges (cdr top-of-ranges))
15911 ;; @r{Specify next top of range value.}
15912 (setq top-of-range (car top-of-ranges)))
15916 ;; @r{Exit outer loop and count the number of defuns larger than}
15917 ;; @r{ the largest top-of-range value.}
15918 (setq defuns-per-range-list
15920 (length sorted-lengths)
15921 defuns-per-range-list))
15925 ;; @r{Return a list of the number of definitions within each range,}
15926 ;; @r{ smallest to largest.}
15927 (nreverse defuns-per-range-list)))
15933 The function is straightforward except for one subtle feature. The
15934 true-or-false test of the inner loop looks like this:
15938 (and (car sorted-lengths)
15939 (< (car sorted-lengths) top-of-range))
15945 instead of like this:
15948 (< (car sorted-lengths) top-of-range)
15951 The purpose of the test is to determine whether the first item in the
15952 @code{sorted-lengths} list is less than the value of the top of the
15955 The simple version of the test works fine unless the
15956 @code{sorted-lengths} list has a @code{nil} value. In that case, the
15957 @code{(car sorted-lengths)} expression function returns
15958 @code{nil}. The @code{<} function cannot compare a number to
15959 @code{nil}, which is an empty list, so Emacs signals an error and
15960 stops the function from attempting to continue to execute.
15962 The @code{sorted-lengths} list always becomes @code{nil} when the
15963 counter reaches the end of the list. This means that any attempt to
15964 use the @code{defuns-per-range} function with the simple version of
15965 the test will fail.
15967 We solve the problem by using the @code{(car sorted-lengths)}
15968 expression in conjunction with the @code{and} expression. The
15969 @code{(car sorted-lengths)} expression returns a non-@code{nil}
15970 value so long as the list has at least one number within it, but
15971 returns @code{nil} if the list is empty. The @code{and} expression
15972 first evaluates the @code{(car sorted-lengths)} expression, and
15973 if it is @code{nil}, returns false @emph{without} evaluating the
15974 @code{<} expression. But if the @code{(car sorted-lengths)}
15975 expression returns a non-@code{nil} value, the @code{and} expression
15976 evaluates the @code{<} expression, and returns that value as the value
15977 of the @code{and} expression.
15979 @c colon in printed section title causes problem in Info cross reference
15980 This way, we avoid an error.
15983 (For information about @code{and}, see
15984 @ref{kill-new function, , The @code{kill-new} function}.)
15988 (@xref{kill-new function, , The @code{kill-new} function}, for
15989 information about @code{and}.)
15992 Here is a short test of the @code{defuns-per-range} function. First,
15993 evaluate the expression that binds (a shortened)
15994 @code{top-of-ranges} list to the list of values, then evaluate the
15995 expression for binding the @code{sorted-lengths} list, and then
15996 evaluate the @code{defuns-per-range} function.
16000 ;; @r{(Shorter list than we will use later.)}
16001 (setq top-of-ranges
16002 '(110 120 130 140 150
16003 160 170 180 190 200))
16005 (setq sorted-lengths
16006 '(85 86 110 116 122 129 154 176 179 200 265 300 300))
16008 (defuns-per-range sorted-lengths top-of-ranges)
16014 The list returned looks like this:
16017 (2 2 2 0 0 1 0 2 0 0 4)
16021 Indeed, there are two elements of the @code{sorted-lengths} list
16022 smaller than 110, two elements between 110 and 119, two elements
16023 between 120 and 129, and so on. There are four elements with a value
16026 @c The next step is to turn this numbers' list into a graph.
16027 @node Readying a Graph
16028 @chapter Readying a Graph
16029 @cindex Readying a graph
16030 @cindex Graph prototype
16031 @cindex Prototype graph
16032 @cindex Body of graph
16034 Our goal is to construct a graph showing the numbers of function
16035 definitions of various lengths in the Emacs lisp sources.
16037 As a practical matter, if you were creating a graph, you would
16038 probably use a program such as @code{gnuplot} to do the job.
16039 (@code{gnuplot} is nicely integrated into GNU Emacs.) In this case,
16040 however, we create one from scratch, and in the process we will
16041 re-acquaint ourselves with some of what we learned before and learn
16044 In this chapter, we will first write a simple graph printing function.
16045 This first definition will be a @dfn{prototype}, a rapidly written
16046 function that enables us to reconnoiter this unknown graph-making
16047 territory. We will discover dragons, or find that they are myth.
16048 After scouting the terrain, we will feel more confident and enhance
16049 the function to label the axes automatically.
16052 * Columns of a graph::
16053 * graph-body-print:: How to print the body of a graph.
16054 * recursive-graph-body-print::
16056 * Line Graph Exercise::
16060 @node Columns of a graph
16061 @unnumberedsec Printing the Columns of a Graph
16064 Since Emacs is designed to be flexible and work with all kinds of
16065 terminals, including character-only terminals, the graph will need to
16066 be made from one of the `typewriter' symbols. An asterisk will do; as
16067 we enhance the graph-printing function, we can make the choice of
16068 symbol a user option.
16070 We can call this function @code{graph-body-print}; it will take a
16071 @code{numbers-list} as its only argument. At this stage, we will not
16072 label the graph, but only print its body.
16074 The @code{graph-body-print} function inserts a vertical column of
16075 asterisks for each element in the @code{numbers-list}. The height of
16076 each line is determined by the value of that element of the
16077 @code{numbers-list}.
16079 Inserting columns is a repetitive act; that means that this function can
16080 be written either with a @code{while} loop or recursively.
16082 Our first challenge is to discover how to print a column of asterisks.
16083 Usually, in Emacs, we print characters onto a screen horizontally,
16084 line by line, by typing. We have two routes we can follow: write our
16085 own column-insertion function or discover whether one exists in Emacs.
16087 To see whether there is one in Emacs, we can use the @kbd{M-x apropos}
16088 command. This command is like the @kbd{C-h a} (@code{command-apropos})
16089 command, except that the latter finds only those functions that are
16090 commands. The @kbd{M-x apropos} command lists all symbols that match
16091 a regular expression, including functions that are not interactive.
16094 What we want to look for is some command that prints or inserts
16095 columns. Very likely, the name of the function will contain either
16096 the word `print' or the word `insert' or the word `column'.
16097 Therefore, we can simply type @kbd{M-x apropos RET
16098 print\|insert\|column RET} and look at the result. On my system, this
16099 command once too takes quite some time, and then produced a list of 79
16100 functions and variables. Now it does not take much time at all and
16101 produces a list of 211 functions and variables. Scanning down the
16102 list, the only function that looks as if it might do the job is
16103 @code{insert-rectangle}.
16106 Indeed, this is the function we want; its documentation says:
16111 Insert text of RECTANGLE with upper left corner at point.
16112 RECTANGLE's first line is inserted at point,
16113 its second line is inserted at a point vertically under point, etc.
16114 RECTANGLE should be a list of strings.
16115 After this command, the mark is at the upper left corner
16116 and point is at the lower right corner.
16120 We can run a quick test, to make sure it does what we expect of it.
16122 Here is the result of placing the cursor after the
16123 @code{insert-rectangle} expression and typing @kbd{C-u C-x C-e}
16124 (@code{eval-last-sexp}). The function inserts the strings
16125 @samp{"first"}, @samp{"second"}, and @samp{"third"} at and below
16126 point. Also the function returns @code{nil}.
16130 (insert-rectangle '("first" "second" "third"))first
16137 Of course, we won't be inserting the text of the
16138 @code{insert-rectangle} expression itself into the buffer in which we
16139 are making the graph, but will call the function from our program. We
16140 shall, however, have to make sure that point is in the buffer at the
16141 place where the @code{insert-rectangle} function will insert its
16144 If you are reading this in Info, you can see how this works by
16145 switching to another buffer, such as the @file{*scratch*} buffer,
16146 placing point somewhere in the buffer, typing @kbd{M-:}, typing the
16147 @code{insert-rectangle} expression into the minibuffer at the prompt,
16148 and then typing @key{RET}. This causes Emacs to evaluate the
16149 expression in the minibuffer, but to use as the value of point the
16150 position of point in the @file{*scratch*} buffer. (@kbd{M-:} is the
16151 keybinding for @code{eval-expression}. Also, @code{nil} does not
16152 appear in the @file{*scratch*} buffer since the expression is
16153 evaluated in the minibuffer.)
16155 We find when we do this that point ends up at the end of the last
16156 inserted line---that is to say, this function moves point as a
16157 side-effect. If we were to repeat the command, with point at this
16158 position, the next insertion would be below and to the right of the
16159 previous insertion. We don't want this! If we are going to make a
16160 bar graph, the columns need to be beside each other.
16162 So we discover that each cycle of the column-inserting @code{while}
16163 loop must reposition point to the place we want it, and that place
16164 will be at the top, not the bottom, of the column. Moreover, we
16165 remember that when we print a graph, we do not expect all the columns
16166 to be the same height. This means that the top of each column may be
16167 at a different height from the previous one. We cannot simply
16168 reposition point to the same line each time, but moved over to the
16169 right---or perhaps we can@dots{}
16171 We are planning to make the columns of the bar graph out of asterisks.
16172 The number of asterisks in the column is the number specified by the
16173 current element of the @code{numbers-list}. We need to construct a
16174 list of asterisks of the right length for each call to
16175 @code{insert-rectangle}. If this list consists solely of the requisite
16176 number of asterisks, then we will have position point the right number
16177 of lines above the base for the graph to print correctly. This could
16180 Alternatively, if we can figure out some way to pass
16181 @code{insert-rectangle} a list of the same length each time, then we
16182 can place point on the same line each time, but move it over one
16183 column to the right for each new column. If we do this, however, some
16184 of the entries in the list passed to @code{insert-rectangle} must be
16185 blanks rather than asterisks. For example, if the maximum height of
16186 the graph is 5, but the height of the column is 3, then
16187 @code{insert-rectangle} requires an argument that looks like this:
16190 (" " " " "*" "*" "*")
16193 This last proposal is not so difficult, so long as we can determine
16194 the column height. There are two ways for us to specify the column
16195 height: we can arbitrarily state what it will be, which would work
16196 fine for graphs of that height; or we can search through the list of
16197 numbers and use the maximum height of the list as the maximum height
16198 of the graph. If the latter operation were difficult, then the former
16199 procedure would be easiest, but there is a function built into Emacs
16200 that determines the maximum of its arguments. We can use that
16201 function. The function is called @code{max} and it returns the
16202 largest of all its arguments, which must be numbers. Thus, for
16210 returns 7. (A corresponding function called @code{min} returns the
16211 smallest of all its arguments.)
16215 However, we cannot simply call @code{max} on the @code{numbers-list};
16216 the @code{max} function expects numbers as its argument, not a list of
16217 numbers. Thus, the following expression,
16220 (max '(3 4 6 5 7 3))
16225 produces the following error message;
16228 Wrong type of argument: number-or-marker-p, (3 4 6 5 7 3)
16232 We need a function that passes a list of arguments to a function.
16233 This function is @code{apply}. This function `applies' its first
16234 argument (a function) to its remaining arguments, the last of which
16241 (apply 'max 3 4 7 3 '(4 8 5))
16247 (Incidentally, I don't know how you would learn of this function
16248 without a book such as this. It is possible to discover other
16249 functions, like @code{search-forward} or @code{insert-rectangle}, by
16250 guessing at a part of their names and then using @code{apropos}. Even
16251 though its base in metaphor is clear---`apply' its first argument to
16252 the rest---I doubt a novice would come up with that particular word
16253 when using @code{apropos} or other aid. Of course, I could be wrong;
16254 after all, the function was first named by someone who had to invent
16257 The second and subsequent arguments to @code{apply} are optional, so
16258 we can use @code{apply} to call a function and pass the elements of a
16259 list to it, like this, which also returns 8:
16262 (apply 'max '(4 8 5))
16265 This latter way is how we will use @code{apply}. The
16266 @code{recursive-lengths-list-many-files} function returns a numbers'
16267 list to which we can apply @code{max} (we could also apply @code{max} to
16268 the sorted numbers' list; it does not matter whether the list is
16272 Hence, the operation for finding the maximum height of the graph is this:
16275 (setq max-graph-height (apply 'max numbers-list))
16278 Now we can return to the question of how to create a list of strings
16279 for a column of the graph. Told the maximum height of the graph
16280 and the number of asterisks that should appear in the column, the
16281 function should return a list of strings for the
16282 @code{insert-rectangle} command to insert.
16284 Each column is made up of asterisks or blanks. Since the function is
16285 passed the value of the height of the column and the number of
16286 asterisks in the column, the number of blanks can be found by
16287 subtracting the number of asterisks from the height of the column.
16288 Given the number of blanks and the number of asterisks, two
16289 @code{while} loops can be used to construct the list:
16293 ;;; @r{First version.}
16294 (defun column-of-graph (max-graph-height actual-height)
16295 "Return list of strings that is one column of a graph."
16296 (let ((insert-list nil)
16297 (number-of-top-blanks
16298 (- max-graph-height actual-height)))
16302 ;; @r{Fill in asterisks.}
16303 (while (> actual-height 0)
16304 (setq insert-list (cons "*" insert-list))
16305 (setq actual-height (1- actual-height)))
16309 ;; @r{Fill in blanks.}
16310 (while (> number-of-top-blanks 0)
16311 (setq insert-list (cons " " insert-list))
16312 (setq number-of-top-blanks
16313 (1- number-of-top-blanks)))
16317 ;; @r{Return whole list.}
16322 If you install this function and then evaluate the following
16323 expression you will see that it returns the list as desired:
16326 (column-of-graph 5 3)
16334 (" " " " "*" "*" "*")
16337 As written, @code{column-of-graph} contains a major flaw: the symbols
16338 used for the blank and for the marked entries in the column are
16339 `hard-coded' as a space and asterisk. This is fine for a prototype,
16340 but you, or another user, may wish to use other symbols. For example,
16341 in testing the graph function, you many want to use a period in place
16342 of the space, to make sure the point is being repositioned properly
16343 each time the @code{insert-rectangle} function is called; or you might
16344 want to substitute a @samp{+} sign or other symbol for the asterisk.
16345 You might even want to make a graph-column that is more than one
16346 display column wide. The program should be more flexible. The way to
16347 do that is to replace the blank and the asterisk with two variables
16348 that we can call @code{graph-blank} and @code{graph-symbol} and define
16349 those variables separately.
16351 Also, the documentation is not well written. These considerations
16352 lead us to the second version of the function:
16356 (defvar graph-symbol "*"
16357 "String used as symbol in graph, usually an asterisk.")
16361 (defvar graph-blank " "
16362 "String used as blank in graph, usually a blank space.
16363 graph-blank must be the same number of columns wide
16369 (For an explanation of @code{defvar}, see
16370 @ref{defvar, , Initializing a Variable with @code{defvar}}.)
16374 ;;; @r{Second version.}
16375 (defun column-of-graph (max-graph-height actual-height)
16376 "Return MAX-GRAPH-HEIGHT strings; ACTUAL-HEIGHT are graph-symbols.
16380 The graph-symbols are contiguous entries at the end
16382 The list will be inserted as one column of a graph.
16383 The strings are either graph-blank or graph-symbol."
16387 (let ((insert-list nil)
16388 (number-of-top-blanks
16389 (- max-graph-height actual-height)))
16393 ;; @r{Fill in @code{graph-symbols}.}
16394 (while (> actual-height 0)
16395 (setq insert-list (cons graph-symbol insert-list))
16396 (setq actual-height (1- actual-height)))
16400 ;; @r{Fill in @code{graph-blanks}.}
16401 (while (> number-of-top-blanks 0)
16402 (setq insert-list (cons graph-blank insert-list))
16403 (setq number-of-top-blanks
16404 (1- number-of-top-blanks)))
16406 ;; @r{Return whole list.}
16411 If we wished, we could rewrite @code{column-of-graph} a third time to
16412 provide optionally for a line graph as well as for a bar graph. This
16413 would not be hard to do. One way to think of a line graph is that it
16414 is no more than a bar graph in which the part of each bar that is
16415 below the top is blank. To construct a column for a line graph, the
16416 function first constructs a list of blanks that is one shorter than
16417 the value, then it uses @code{cons} to attach a graph symbol to the
16418 list; then it uses @code{cons} again to attach the `top blanks' to
16421 It is easy to see how to write such a function, but since we don't
16422 need it, we will not do it. But the job could be done, and if it were
16423 done, it would be done with @code{column-of-graph}. Even more
16424 important, it is worth noting that few changes would have to be made
16425 anywhere else. The enhancement, if we ever wish to make it, is
16428 Now, finally, we come to our first actual graph printing function.
16429 This prints the body of a graph, not the labels for the vertical and
16430 horizontal axes, so we can call this @code{graph-body-print}.
16432 @node graph-body-print
16433 @section The @code{graph-body-print} Function
16434 @findex graph-body-print
16436 After our preparation in the preceding section, the
16437 @code{graph-body-print} function is straightforward. The function
16438 will print column after column of asterisks and blanks, using the
16439 elements of a numbers' list to specify the number of asterisks in each
16440 column. This is a repetitive act, which means we can use a
16441 decrementing @code{while} loop or recursive function for the job. In
16442 this section, we will write the definition using a @code{while} loop.
16444 The @code{column-of-graph} function requires the height of the graph
16445 as an argument, so we should determine and record that as a local variable.
16447 This leads us to the following template for the @code{while} loop
16448 version of this function:
16452 (defun graph-body-print (numbers-list)
16453 "@var{documentation}@dots{}"
16454 (let ((height @dots{}
16459 (while numbers-list
16460 @var{insert-columns-and-reposition-point}
16461 (setq numbers-list (cdr numbers-list)))))
16466 We need to fill in the slots of the template.
16468 Clearly, we can use the @code{(apply 'max numbers-list)} expression to
16469 determine the height of the graph.
16471 The @code{while} loop will cycle through the @code{numbers-list} one
16472 element at a time. As it is shortened by the @code{(setq numbers-list
16473 (cdr numbers-list))} expression, the @sc{car} of each instance of the
16474 list is the value of the argument for @code{column-of-graph}.
16476 At each cycle of the @code{while} loop, the @code{insert-rectangle}
16477 function inserts the list returned by @code{column-of-graph}. Since
16478 the @code{insert-rectangle} function moves point to the lower right of
16479 the inserted rectangle, we need to save the location of point at the
16480 time the rectangle is inserted, move back to that position after the
16481 rectangle is inserted, and then move horizontally to the next place
16482 from which @code{insert-rectangle} is called.
16484 If the inserted columns are one character wide, as they will be if
16485 single blanks and asterisks are used, the repositioning command is
16486 simply @code{(forward-char 1)}; however, the width of a column may be
16487 greater than one. This means that the repositioning command should be
16488 written @code{(forward-char symbol-width)}. The @code{symbol-width}
16489 itself is the length of a @code{graph-blank} and can be found using
16490 the expression @code{(length graph-blank)}. The best place to bind
16491 the @code{symbol-width} variable to the value of the width of graph
16492 column is in the varlist of the @code{let} expression.
16495 These considerations lead to the following function definition:
16499 (defun graph-body-print (numbers-list)
16500 "Print a bar graph of the NUMBERS-LIST.
16501 The numbers-list consists of the Y-axis values."
16503 (let ((height (apply 'max numbers-list))
16504 (symbol-width (length graph-blank))
16509 (while numbers-list
16510 (setq from-position (point))
16512 (column-of-graph height (car numbers-list)))
16513 (goto-char from-position)
16514 (forward-char symbol-width)
16517 ;; @r{Draw graph column by column.}
16519 (setq numbers-list (cdr numbers-list)))
16522 ;; @r{Place point for X axis labels.}
16523 (forward-line height)
16530 The one unexpected expression in this function is the
16531 @w{@code{(sit-for 0)}} expression in the @code{while} loop. This
16532 expression makes the graph printing operation more interesting to
16533 watch than it would be otherwise. The expression causes Emacs to
16534 `sit' or do nothing for a zero length of time and then redraw the
16535 screen. Placed here, it causes Emacs to redraw the screen column by
16536 column. Without it, Emacs would not redraw the screen until the
16539 We can test @code{graph-body-print} with a short list of numbers.
16543 Install @code{graph-symbol}, @code{graph-blank},
16544 @code{column-of-graph}, which are in
16546 @ref{Readying a Graph, , Readying a Graph},
16549 @ref{Columns of a graph},
16551 and @code{graph-body-print}.
16555 Copy the following expression:
16558 (graph-body-print '(1 2 3 4 6 4 3 5 7 6 5 2 3))
16562 Switch to the @file{*scratch*} buffer and place the cursor where you
16563 want the graph to start.
16566 Type @kbd{M-:} (@code{eval-expression}).
16569 Yank the @code{graph-body-print} expression into the minibuffer
16570 with @kbd{C-y} (@code{yank)}.
16573 Press @key{RET} to evaluate the @code{graph-body-print} expression.
16577 Emacs will print a graph like this:
16591 @node recursive-graph-body-print
16592 @section The @code{recursive-graph-body-print} Function
16593 @findex recursive-graph-body-print
16595 The @code{graph-body-print} function may also be written recursively.
16596 The recursive solution is divided into two parts: an outside `wrapper'
16597 that uses a @code{let} expression to determine the values of several
16598 variables that need only be found once, such as the maximum height of
16599 the graph, and an inside function that is called recursively to print
16603 The `wrapper' is uncomplicated:
16607 (defun recursive-graph-body-print (numbers-list)
16608 "Print a bar graph of the NUMBERS-LIST.
16609 The numbers-list consists of the Y-axis values."
16610 (let ((height (apply 'max numbers-list))
16611 (symbol-width (length graph-blank))
16613 (recursive-graph-body-print-internal
16620 The recursive function is a little more difficult. It has four parts:
16621 the `do-again-test', the printing code, the recursive call, and the
16622 `next-step-expression'. The `do-again-test' is a @code{when}
16623 expression that determines whether the @code{numbers-list} contains
16624 any remaining elements; if it does, the function prints one column of
16625 the graph using the printing code and calls itself again. The
16626 function calls itself again according to the value produced by the
16627 `next-step-expression' which causes the call to act on a shorter
16628 version of the @code{numbers-list}.
16632 (defun recursive-graph-body-print-internal
16633 (numbers-list height symbol-width)
16634 "Print a bar graph.
16635 Used within recursive-graph-body-print function."
16640 (setq from-position (point))
16642 (column-of-graph height (car numbers-list)))
16645 (goto-char from-position)
16646 (forward-char symbol-width)
16647 (sit-for 0) ; @r{Draw graph column by column.}
16648 (recursive-graph-body-print-internal
16649 (cdr numbers-list) height symbol-width)))
16654 After installation, this expression can be tested; here is a sample:
16657 (recursive-graph-body-print '(3 2 5 6 7 5 3 4 6 4 3 2 1))
16661 Here is what @code{recursive-graph-body-print} produces:
16675 Either of these two functions, @code{graph-body-print} or
16676 @code{recursive-graph-body-print}, create the body of a graph.
16679 @section Need for Printed Axes
16681 A graph needs printed axes, so you can orient yourself. For a do-once
16682 project, it may be reasonable to draw the axes by hand using Emacs's
16683 Picture mode; but a graph drawing function may be used more than once.
16685 For this reason, I have written enhancements to the basic
16686 @code{print-graph-body} function that automatically print labels for
16687 the horizontal and vertical axes. Since the label printing functions
16688 do not contain much new material, I have placed their description in
16689 an appendix. @xref{Full Graph, , A Graph with Labeled Axes}.
16691 @node Line Graph Exercise
16694 Write a line graph version of the graph printing functions.
16696 @node Emacs Initialization
16697 @chapter Your @file{.emacs} File
16698 @cindex @file{.emacs} file
16699 @cindex Customizing your @file{.emacs} file
16700 @cindex Initialization file
16702 ``You don't have to like Emacs to like it''---this seemingly
16703 paradoxical statement is the secret of GNU Emacs. The plain, `out of
16704 the box' Emacs is a generic tool. Most people who use it, customize
16705 it to suit themselves.
16707 GNU Emacs is mostly written in Emacs Lisp; this means that by writing
16708 expressions in Emacs Lisp you can change or extend Emacs.
16711 * Default Configuration::
16712 * Site-wide Init:: You can write site-wide init files.
16713 * defcustom:: Emacs will write code for you.
16714 * Beginning init File:: How to write a @file{.emacs} init file.
16715 * Text and Auto-fill:: Automatically wrap lines.
16716 * Mail Aliases:: Use abbreviations for email addresses.
16717 * Indent Tabs Mode:: Don't use tabs with @TeX{}
16718 * Keybindings:: Create some personal keybindings.
16719 * Keymaps:: More about key binding.
16720 * Loading Files:: Load (i.e., evaluate) files automatically.
16721 * Autoload:: Make functions available.
16722 * Simple Extension:: Define a function; bind it to a key.
16723 * X11 Colors:: Colors in X.
16725 * Mode Line:: How to customize your mode line.
16729 @node Default Configuration
16730 @unnumberedsec Emacs's Default Configuration
16733 There are those who appreciate Emacs's default configuration. After
16734 all, Emacs starts you in C mode when you edit a C file, starts you in
16735 Fortran mode when you edit a Fortran file, and starts you in
16736 Fundamental mode when you edit an unadorned file. This all makes
16737 sense, if you do not know who is going to use Emacs. Who knows what a
16738 person hopes to do with an unadorned file? Fundamental mode is the
16739 right default for such a file, just as C mode is the right default for
16740 editing C code. (Enough programming languages have syntaxes
16741 that enable them to share or nearly share features, so C mode is
16742 now provided by CC mode, the `C Collection'.)
16744 But when you do know who is going to use Emacs---you,
16745 yourself---then it makes sense to customize Emacs.
16747 For example, I seldom want Fundamental mode when I edit an
16748 otherwise undistinguished file; I want Text mode. This is why I
16749 customize Emacs: so it suits me.
16751 You can customize and extend Emacs by writing or adapting a
16752 @file{~/.emacs} file. This is your personal initialization file; its
16753 contents, written in Emacs Lisp, tell Emacs what to do.@footnote{You
16754 may also add @file{.el} to @file{~/.emacs} and call it a
16755 @file{~/.emacs.el} file. In the past, you were forbidden to type the
16756 extra keystrokes that the name @file{~/.emacs.el} requires, but now
16757 you may. The new format is consistent with the Emacs Lisp file
16758 naming conventions; the old format saves typing.}
16760 A @file{~/.emacs} file contains Emacs Lisp code. You can write this
16761 code yourself; or you can use Emacs's @code{customize} feature to write
16762 the code for you. You can combine your own expressions and
16763 auto-written Customize expressions in your @file{.emacs} file.
16765 (I myself prefer to write my own expressions, except for those,
16766 particularly fonts, that I find easier to manipulate using the
16767 @code{customize} command. I combine the two methods.)
16769 Most of this chapter is about writing expressions yourself. It
16770 describes a simple @file{.emacs} file; for more information, see
16771 @ref{Init File, , The Init File, emacs, The GNU Emacs Manual}, and
16772 @ref{Init File, , The Init File, elisp, The GNU Emacs Lisp Reference
16775 @node Site-wide Init
16776 @section Site-wide Initialization Files
16778 @cindex @file{default.el} init file
16779 @cindex @file{site-init.el} init file
16780 @cindex @file{site-load.el} init file
16781 In addition to your personal initialization file, Emacs automatically
16782 loads various site-wide initialization files, if they exist. These
16783 have the same form as your @file{.emacs} file, but are loaded by
16786 Two site-wide initialization files, @file{site-load.el} and
16787 @file{site-init.el}, are loaded into Emacs and then `dumped' if a
16788 `dumped' version of Emacs is created, as is most common. (Dumped
16789 copies of Emacs load more quickly. However, once a file is loaded and
16790 dumped, a change to it does not lead to a change in Emacs unless you
16791 load it yourself or re-dump Emacs. @xref{Building Emacs, , Building
16792 Emacs, elisp, The GNU Emacs Lisp Reference Manual}, and the
16793 @file{INSTALL} file.)
16795 Three other site-wide initialization files are loaded automatically
16796 each time you start Emacs, if they exist. These are
16797 @file{site-start.el}, which is loaded @emph{before} your @file{.emacs}
16798 file, and @file{default.el}, and the terminal type file, which are both
16799 loaded @emph{after} your @file{.emacs} file.
16801 Settings and definitions in your @file{.emacs} file will overwrite
16802 conflicting settings and definitions in a @file{site-start.el} file,
16803 if it exists; but the settings and definitions in a @file{default.el}
16804 or terminal type file will overwrite those in your @file{.emacs} file.
16805 (You can prevent interference from a terminal type file by setting
16806 @code{term-file-prefix} to @code{nil}. @xref{Simple Extension, , A
16807 Simple Extension}.)
16809 @c Rewritten to avoid overfull hbox.
16810 The @file{INSTALL} file that comes in the distribution contains
16811 descriptions of the @file{site-init.el} and @file{site-load.el} files.
16813 The @file{loadup.el}, @file{startup.el}, and @file{loaddefs.el} files
16814 control loading. These files are in the @file{lisp} directory of the
16815 Emacs distribution and are worth perusing.
16817 The @file{loaddefs.el} file contains a good many suggestions as to
16818 what to put into your own @file{.emacs} file, or into a site-wide
16819 initialization file.
16822 @section Specifying Variables using @code{defcustom}
16825 You can specify variables using @code{defcustom} so that you and
16826 others can then use Emacs's @code{customize} feature to set their
16827 values. (You cannot use @code{customize} to write function
16828 definitions; but you can write @code{defuns} in your @file{.emacs}
16829 file. Indeed, you can write any Lisp expression in your @file{.emacs}
16832 The @code{customize} feature depends on the @code{defcustom} macro.
16833 Although you can use @code{defvar} or @code{setq} for variables that
16834 users set, the @code{defcustom} macro is designed for the job.
16836 You can use your knowledge of @code{defvar} for writing the
16837 first three arguments for @code{defcustom}. The first argument to
16838 @code{defcustom} is the name of the variable. The second argument is
16839 the variable's initial value, if any; and this value is set only if
16840 the value has not already been set. The third argument is the
16843 The fourth and subsequent arguments to @code{defcustom} specify types
16844 and options; these are not featured in @code{defvar}. (These
16845 arguments are optional.)
16847 Each of these arguments consists of a keyword followed by a value.
16848 Each keyword starts with the colon character @samp{:}.
16851 For example, the customizable user option variable
16852 @code{text-mode-hook} looks like this:
16856 (defcustom text-mode-hook nil
16857 "Normal hook run when entering Text mode and many related modes."
16859 :options '(turn-on-auto-fill flyspell-mode)
16865 The name of the variable is @code{text-mode-hook}; it has no default
16866 value; and its documentation string tells you what it does.
16868 The @code{:type} keyword tells Emacs the kind of data to which
16869 @code{text-mode-hook} should be set and how to display the value in a
16870 Customization buffer.
16872 The @code{:options} keyword specifies a suggested list of values for
16873 the variable. Usually, @code{:options} applies to a hook.
16874 The list is only a suggestion; it is not exclusive; a person who sets
16875 the variable may set it to other values; the list shown following the
16876 @code{:options} keyword is intended to offer convenient choices to a
16879 Finally, the @code{:group} keyword tells the Emacs Customization
16880 command in which group the variable is located. This tells where to
16883 The @code{defcustom} macro recognizes more than a dozen keywords.
16884 For more information, see @ref{Customization, , Writing Customization
16885 Definitions, elisp, The GNU Emacs Lisp Reference Manual}.
16887 Consider @code{text-mode-hook} as an example.
16889 There are two ways to customize this variable. You can use the
16890 customization command or write the appropriate expressions yourself.
16893 Using the customization command, you can type:
16900 and find that the group for editing files of data is called `data'.
16901 Enter that group. Text Mode Hook is the first member. You can click
16902 on its various options, such as @code{turn-on-auto-fill}, to set the
16903 values. After you click on the button to
16906 Save for Future Sessions
16910 Emacs will write an expression into your @file{.emacs} file.
16911 It will look like this:
16915 (custom-set-variables
16916 ;; custom-set-variables was added by Custom.
16917 ;; If you edit it by hand, you could mess it up, so be careful.
16918 ;; Your init file should contain only one such instance.
16919 ;; If there is more than one, they won't work right.
16920 '(text-mode-hook (quote (turn-on-auto-fill text-mode-hook-identify))))
16925 (The @code{text-mode-hook-identify} function tells
16926 @code{toggle-text-mode-auto-fill} which buffers are in Text mode.
16927 It comes on automatically.)
16929 The @code{custom-set-variables} function works somewhat differently
16930 than a @code{setq}. While I have never learned the differences, I
16931 modify the @code{custom-set-variables} expressions in my @file{.emacs}
16932 file by hand: I make the changes in what appears to me to be a
16933 reasonable manner and have not had any problems. Others prefer to use
16934 the Customization command and let Emacs do the work for them.
16936 Another @code{custom-set-@dots{}} function is @code{custom-set-faces}.
16937 This function sets the various font faces. Over time, I have set a
16938 considerable number of faces. Some of the time, I re-set them using
16939 @code{customize}; other times, I simply edit the
16940 @code{custom-set-faces} expression in my @file{.emacs} file itself.
16942 The second way to customize your @code{text-mode-hook} is to set it
16943 yourself in your @file{.emacs} file using code that has nothing to do
16944 with the @code{custom-set-@dots{}} functions.
16947 When you do this, and later use @code{customize}, you will see a
16951 CHANGED outside Customize; operating on it here may be unreliable.
16955 This message is only a warning. If you click on the button to
16958 Save for Future Sessions
16962 Emacs will write a @code{custom-set-@dots{}} expression near the end
16963 of your @file{.emacs} file that will be evaluated after your
16964 hand-written expression. It will, therefore, overrule your
16965 hand-written expression. No harm will be done. When you do this,
16966 however, be careful to remember which expression is active; if you
16967 forget, you may confuse yourself.
16969 So long as you remember where the values are set, you will have no
16970 trouble. In any event, the values are always set in your
16971 initialization file, which is usually called @file{.emacs}.
16973 I myself use @code{customize} for hardly anything. Mostly, I write
16974 expressions myself.
16978 Incidentally, to be more complete concerning defines: @code{defsubst}
16979 defines an inline function. The syntax is just like that of
16980 @code{defun}. @code{defconst} defines a symbol as a constant. The
16981 intent is that neither programs nor users should ever change a value
16982 set by @code{defconst}. (You can change it; the value set is a
16983 variable; but please do not.)
16985 @node Beginning init File
16986 @section Beginning a @file{.emacs} File
16987 @cindex @file{.emacs} file, beginning of
16989 When you start Emacs, it loads your @file{.emacs} file unless you tell
16990 it not to by specifying @samp{-q} on the command line. (The
16991 @code{emacs -q} command gives you a plain, out-of-the-box Emacs.)
16993 A @file{.emacs} file contains Lisp expressions. Often, these are no
16994 more than expressions to set values; sometimes they are function
16997 @xref{Init File, , The Init File @file{~/.emacs}, emacs, The GNU Emacs
16998 Manual}, for a short description of initialization files.
17000 This chapter goes over some of the same ground, but is a walk among
17001 extracts from a complete, long-used @file{.emacs} file---my own.
17003 The first part of the file consists of comments: reminders to myself.
17004 By now, of course, I remember these things, but when I started, I did
17010 ;;;; Bob's .emacs file
17011 ; Robert J. Chassell
17012 ; 26 September 1985
17017 Look at that date! I started this file a long time ago. I have been
17018 adding to it ever since.
17022 ; Each section in this file is introduced by a
17023 ; line beginning with four semicolons; and each
17024 ; entry is introduced by a line beginning with
17025 ; three semicolons.
17030 This describes the usual conventions for comments in Emacs Lisp.
17031 Everything on a line that follows a semicolon is a comment. Two,
17032 three, and four semicolons are used as subsection and section markers.
17033 (@xref{Comments, ,, elisp, The GNU Emacs Lisp Reference Manual}, for
17034 more about comments.)
17039 ; Control-h is the help key;
17040 ; after typing control-h, type a letter to
17041 ; indicate the subject about which you want help.
17042 ; For an explanation of the help facility,
17043 ; type control-h two times in a row.
17048 Just remember: type @kbd{C-h} two times for help.
17052 ; To find out about any mode, type control-h m
17053 ; while in that mode. For example, to find out
17054 ; about mail mode, enter mail mode and then type
17060 `Mode help', as I call this, is very helpful. Usually, it tells you
17061 all you need to know.
17063 Of course, you don't need to include comments like these in your
17064 @file{.emacs} file. I included them in mine because I kept forgetting
17065 about Mode help or the conventions for comments---but I was able to
17066 remember to look here to remind myself.
17068 @node Text and Auto-fill
17069 @section Text and Auto Fill Mode
17071 Now we come to the part that `turns on' Text mode and
17076 ;;; Text mode and Auto Fill mode
17077 ;; The next two lines put Emacs into Text mode
17078 ;; and Auto Fill mode, and are for writers who
17079 ;; want to start writing prose rather than code.
17080 (setq-default major-mode 'text-mode)
17081 (add-hook 'text-mode-hook 'turn-on-auto-fill)
17085 Here is the first part of this @file{.emacs} file that does something
17086 besides remind a forgetful human!
17088 The first of the two lines in parentheses tells Emacs to turn on Text
17089 mode when you find a file, @emph{unless} that file should go into some
17090 other mode, such as C mode.
17092 @cindex Per-buffer, local variables list
17093 @cindex Local variables list, per-buffer,
17094 @cindex Automatic mode selection
17095 @cindex Mode selection, automatic
17096 When Emacs reads a file, it looks at the extension to the file name,
17097 if any. (The extension is the part that comes after a @samp{.}.) If
17098 the file ends with a @samp{.c} or @samp{.h} extension then Emacs turns
17099 on C mode. Also, Emacs looks at first nonblank line of the file; if
17100 the line says @w{@samp{-*- C -*-}}, Emacs turns on C mode. Emacs
17101 possesses a list of extensions and specifications that it uses
17102 automatically. In addition, Emacs looks near the last page for a
17103 per-buffer, ``local variables list'', if any.
17106 @xref{Choosing Modes, , How Major Modes are Chosen, emacs, The GNU
17109 @xref{File Variables, , Local Variables in Files, emacs, The GNU Emacs
17113 See sections ``How Major Modes are Chosen'' and ``Local Variables in
17114 Files'' in @cite{The GNU Emacs Manual}.
17117 Now, back to the @file{.emacs} file.
17120 Here is the line again; how does it work?
17122 @cindex Text Mode turned on
17124 (setq major-mode 'text-mode)
17128 This line is a short, but complete Emacs Lisp expression.
17130 We are already familiar with @code{setq}. It sets the following variable,
17131 @code{major-mode}, to the subsequent value, which is @code{text-mode}.
17132 The single quote mark before @code{text-mode} tells Emacs to deal directly
17133 with the @code{text-mode} symbol, not with whatever it might stand for.
17134 @xref{set & setq, , Setting the Value of a Variable},
17135 for a reminder of how @code{setq} works.
17136 The main point is that there is no difference between the procedure you
17137 use to set a value in your @file{.emacs} file and the procedure you use
17138 anywhere else in Emacs.
17141 Here is the next line:
17143 @cindex Auto Fill mode turned on
17146 (add-hook 'text-mode-hook 'turn-on-auto-fill)
17150 In this line, the @code{add-hook} command adds
17151 @code{turn-on-auto-fill} to the variable.
17153 @code{turn-on-auto-fill} is the name of a program, that, you guessed
17154 it!, turns on Auto Fill mode.
17156 Every time Emacs turns on Text mode, Emacs runs the commands `hooked'
17157 onto Text mode. So every time Emacs turns on Text mode, Emacs also
17158 turns on Auto Fill mode.
17160 In brief, the first line causes Emacs to enter Text mode when you edit a
17161 file, unless the file name extension, a first non-blank line, or local
17162 variables to tell Emacs otherwise.
17164 Text mode among other actions, sets the syntax table to work
17165 conveniently for writers. In Text mode, Emacs considers an apostrophe
17166 as part of a word like a letter; but Emacs does not consider a period
17167 or a space as part of a word. Thus, @kbd{M-f} moves you over
17168 @samp{it's}. On the other hand, in C mode, @kbd{M-f} stops just after
17169 the @samp{t} of @samp{it's}.
17171 The second line causes Emacs to turn on Auto Fill mode when it turns
17172 on Text mode. In Auto Fill mode, Emacs automatically breaks a line
17173 that is too wide and brings the excessively wide part of the line down
17174 to the next line. Emacs breaks lines between words, not within them.
17176 When Auto Fill mode is turned off, lines continue to the right as you
17177 type them. Depending on how you set the value of
17178 @code{truncate-lines}, the words you type either disappear off the
17179 right side of the screen, or else are shown, in a rather ugly and
17180 unreadable manner, as a continuation line on the screen.
17183 In addition, in this part of my @file{.emacs} file, I tell the Emacs
17184 fill commands to insert two spaces after a colon:
17187 (setq colon-double-space t)
17191 @section Mail Aliases
17193 Here is a @code{setq} that `turns on' mail aliases, along with more
17199 ; To enter mail mode, type `C-x m'
17200 ; To enter RMAIL (for reading mail),
17202 (setq mail-aliases t)
17206 @cindex Mail aliases
17208 This @code{setq} command sets the value of the variable
17209 @code{mail-aliases} to @code{t}. Since @code{t} means true, the line
17210 says, in effect, ``Yes, use mail aliases.''
17212 Mail aliases are convenient short names for long email addresses or
17213 for lists of email addresses. The file where you keep your `aliases'
17214 is @file{~/.mailrc}. You write an alias like this:
17217 alias geo george@@foobar.wiz.edu
17221 When you write a message to George, address it to @samp{geo}; the
17222 mailer will automatically expand @samp{geo} to the full address.
17224 @node Indent Tabs Mode
17225 @section Indent Tabs Mode
17226 @cindex Tabs, preventing
17227 @findex indent-tabs-mode
17229 By default, Emacs inserts tabs in place of multiple spaces when it
17230 formats a region. (For example, you might indent many lines of text
17231 all at once with the @code{indent-region} command.) Tabs look fine on
17232 a terminal or with ordinary printing, but they produce badly indented
17233 output when you use @TeX{} or Texinfo since @TeX{} ignores tabs.
17236 The following turns off Indent Tabs mode:
17240 ;;; Prevent Extraneous Tabs
17241 (setq-default indent-tabs-mode nil)
17245 Note that this line uses @code{setq-default} rather than the
17246 @code{setq} command that we have seen before. The @code{setq-default}
17247 command sets values only in buffers that do not have their own local
17248 values for the variable.
17251 @xref{Just Spaces, , Tabs vs. Spaces, emacs, The GNU Emacs Manual}.
17253 @xref{File Variables, , Local Variables in Files, emacs, The GNU Emacs
17257 See sections ``Tabs vs.@: Spaces'' and ``Local Variables in
17258 Files'' in @cite{The GNU Emacs Manual}.
17263 @section Some Keybindings
17265 Now for some personal keybindings:
17269 ;;; Compare windows
17270 (global-set-key "\C-cw" 'compare-windows)
17274 @findex compare-windows
17275 @code{compare-windows} is a nifty command that compares the text in
17276 your current window with text in the next window. It makes the
17277 comparison by starting at point in each window, moving over text in
17278 each window as far as they match. I use this command all the time.
17280 This also shows how to set a key globally, for all modes.
17282 @cindex Setting a key globally
17283 @cindex Global set key
17284 @cindex Key setting globally
17285 @findex global-set-key
17286 The command is @code{global-set-key}. It is followed by the
17287 keybinding. In a @file{.emacs} file, the keybinding is written as
17288 shown: @code{\C-c} stands for `control-c', which means `press the
17289 control key and the @key{c} key at the same time'. The @code{w} means
17290 `press the @key{w} key'. The keybinding is surrounded by double
17291 quotation marks. In documentation, you would write this as
17292 @w{@kbd{C-c w}}. (If you were binding a @key{META} key, such as
17293 @kbd{M-c}, rather than a @key{CTRL} key, you would write
17294 @w{@code{\M-c}} in your @file{.emacs} file. @xref{Init Rebinding, ,
17295 Rebinding Keys in Your Init File, emacs, The GNU Emacs Manual}, for
17298 The command invoked by the keys is @code{compare-windows}. Note that
17299 @code{compare-windows} is preceded by a single quote; otherwise, Emacs
17300 would first try to evaluate the symbol to determine its value.
17302 These three things, the double quotation marks, the backslash before
17303 the @samp{C}, and the single quote mark are necessary parts of
17304 keybinding that I tend to forget. Fortunately, I have come to
17305 remember that I should look at my existing @file{.emacs} file, and
17306 adapt what is there.
17308 As for the keybinding itself: @kbd{C-c w}. This combines the prefix
17309 key, @kbd{C-c}, with a single character, in this case, @kbd{w}. This
17310 set of keys, @kbd{C-c} followed by a single character, is strictly
17311 reserved for individuals' own use. (I call these `own' keys, since
17312 these are for my own use.) You should always be able to create such a
17313 keybinding for your own use without stomping on someone else's
17314 keybinding. If you ever write an extension to Emacs, please avoid
17315 taking any of these keys for public use. Create a key like @kbd{C-c
17316 C-w} instead. Otherwise, we will run out of `own' keys.
17319 Here is another keybinding, with a comment:
17323 ;;; Keybinding for `occur'
17324 ; I use occur a lot, so let's bind it to a key:
17325 (global-set-key "\C-co" 'occur)
17330 The @code{occur} command shows all the lines in the current buffer
17331 that contain a match for a regular expression. Matching lines are
17332 shown in a buffer called @file{*Occur*}. That buffer serves as a menu
17333 to jump to occurrences.
17335 @findex global-unset-key
17336 @cindex Unbinding key
17337 @cindex Key unbinding
17339 Here is how to unbind a key, so it does not
17345 (global-unset-key "\C-xf")
17349 There is a reason for this unbinding: I found I inadvertently typed
17350 @w{@kbd{C-x f}} when I meant to type @kbd{C-x C-f}. Rather than find a
17351 file, as I intended, I accidentally set the width for filled text,
17352 almost always to a width I did not want. Since I hardly ever reset my
17353 default width, I simply unbound the key.
17355 @findex list-buffers, @r{rebound}
17356 @findex buffer-menu, @r{bound to key}
17358 The following rebinds an existing key:
17362 ;;; Rebind `C-x C-b' for `buffer-menu'
17363 (global-set-key "\C-x\C-b" 'buffer-menu)
17367 By default, @kbd{C-x C-b} runs the
17368 @code{list-buffers} command. This command lists
17369 your buffers in @emph{another} window. Since I
17370 almost always want to do something in that
17371 window, I prefer the @code{buffer-menu}
17372 command, which not only lists the buffers,
17373 but moves point into that window.
17378 @cindex Rebinding keys
17380 Emacs uses @dfn{keymaps} to record which keys call which commands.
17381 When you use @code{global-set-key} to set the keybinding for a single
17382 command in all parts of Emacs, you are specifying the keybinding in
17383 @code{current-global-map}.
17385 Specific modes, such as C mode or Text mode, have their own keymaps;
17386 the mode-specific keymaps override the global map that is shared by
17389 The @code{global-set-key} function binds, or rebinds, the global
17390 keymap. For example, the following binds the key @kbd{C-x C-b} to the
17391 function @code{buffer-menu}:
17394 (global-set-key "\C-x\C-b" 'buffer-menu)
17397 Mode-specific keymaps are bound using the @code{define-key} function,
17398 which takes a specific keymap as an argument, as well as the key and
17399 the command. For example, my @file{.emacs} file contains the
17400 following expression to bind the @code{texinfo-insert-@@group} command
17401 to @kbd{C-c C-c g}:
17405 (define-key texinfo-mode-map "\C-c\C-cg" 'texinfo-insert-@@group)
17410 The @code{texinfo-insert-@@group} function itself is a little extension
17411 to Texinfo mode that inserts @samp{@@group} into a Texinfo file. I
17412 use this command all the time and prefer to type the three strokes
17413 @kbd{C-c C-c g} rather than the six strokes @kbd{@@ g r o u p}.
17414 (@samp{@@group} and its matching @samp{@@end group} are commands that
17415 keep all enclosed text together on one page; many multi-line examples
17416 in this book are surrounded by @samp{@@group @dots{} @@end group}.)
17419 Here is the @code{texinfo-insert-@@group} function definition:
17423 (defun texinfo-insert-@@group ()
17424 "Insert the string @@group in a Texinfo buffer."
17426 (beginning-of-line)
17427 (insert "@@group\n"))
17431 (Of course, I could have used Abbrev mode to save typing, rather than
17432 write a function to insert a word; but I prefer key strokes consistent
17433 with other Texinfo mode key bindings.)
17435 You will see numerous @code{define-key} expressions in
17436 @file{loaddefs.el} as well as in the various mode libraries, such as
17437 @file{cc-mode.el} and @file{lisp-mode.el}.
17439 @xref{Key Bindings, , Customizing Key Bindings, emacs, The GNU Emacs
17440 Manual}, and @ref{Keymaps, , Keymaps, elisp, The GNU Emacs Lisp
17441 Reference Manual}, for more information about keymaps.
17443 @node Loading Files
17444 @section Loading Files
17445 @cindex Loading files
17448 Many people in the GNU Emacs community have written extensions to
17449 Emacs. As time goes by, these extensions are often included in new
17450 releases. For example, the Calendar and Diary packages are now part
17451 of the standard GNU Emacs, as is Calc.
17453 You can use a @code{load} command to evaluate a complete file and
17454 thereby install all the functions and variables in the file into Emacs.
17457 @c (auto-compression-mode t)
17460 (load "~/emacs/slowsplit")
17463 This evaluates, i.e., loads, the @file{slowsplit.el} file or if it
17464 exists, the faster, byte compiled @file{slowsplit.elc} file from the
17465 @file{emacs} sub-directory of your home directory. The file contains
17466 the function @code{split-window-quietly}, which John Robinson wrote in
17469 The @code{split-window-quietly} function splits a window with the
17470 minimum of redisplay. I installed it in 1989 because it worked well
17471 with the slow 1200 baud terminals I was then using. Nowadays, I only
17472 occasionally come across such a slow connection, but I continue to use
17473 the function because I like the way it leaves the bottom half of a
17474 buffer in the lower of the new windows and the top half in the upper
17478 To replace the key binding for the default
17479 @code{split-window-vertically}, you must also unset that key and bind
17480 the keys to @code{split-window-quietly}, like this:
17484 (global-unset-key "\C-x2")
17485 (global-set-key "\C-x2" 'split-window-quietly)
17490 If you load many extensions, as I do, then instead of specifying the
17491 exact location of the extension file, as shown above, you can specify
17492 that directory as part of Emacs's @code{load-path}. Then, when Emacs
17493 loads a file, it will search that directory as well as its default
17494 list of directories. (The default list is specified in @file{paths.h}
17495 when Emacs is built.)
17498 The following command adds your @file{~/emacs} directory to the
17499 existing load path:
17503 ;;; Emacs Load Path
17504 (setq load-path (cons "~/emacs" load-path))
17508 Incidentally, @code{load-library} is an interactive interface to the
17509 @code{load} function. The complete function looks like this:
17511 @findex load-library
17514 (defun load-library (library)
17515 "Load the library named LIBRARY.
17516 This is an interface to the function `load'."
17518 (list (completing-read "Load library: "
17519 (apply-partially 'locate-file-completion-table
17521 (get-load-suffixes)))))
17526 The name of the function, @code{load-library}, comes from the use of
17527 `library' as a conventional synonym for `file'. The source for the
17528 @code{load-library} command is in the @file{files.el} library.
17530 Another interactive command that does a slightly different job is
17531 @code{load-file}. @xref{Lisp Libraries, , Libraries of Lisp Code for
17532 Emacs, emacs, The GNU Emacs Manual}, for information on the
17533 distinction between @code{load-library} and this command.
17536 @section Autoloading
17539 Instead of installing a function by loading the file that contains it,
17540 or by evaluating the function definition, you can make the function
17541 available but not actually install it until it is first called. This
17542 is called @dfn{autoloading}.
17544 When you execute an autoloaded function, Emacs automatically evaluates
17545 the file that contains the definition, and then calls the function.
17547 Emacs starts quicker with autoloaded functions, since their libraries
17548 are not loaded right away; but you need to wait a moment when you
17549 first use such a function, while its containing file is evaluated.
17551 Rarely used functions are frequently autoloaded. The
17552 @file{loaddefs.el} library contains hundreds of autoloaded functions,
17553 from @code{bookmark-set} to @code{wordstar-mode}. Of course, you may
17554 come to use a `rare' function frequently. When you do, you should
17555 load that function's file with a @code{load} expression in your
17556 @file{.emacs} file.
17558 In my @file{.emacs} file, I load 14 libraries that contain functions
17559 that would otherwise be autoloaded. (Actually, it would have been
17560 better to include these files in my `dumped' Emacs, but I forgot.
17561 @xref{Building Emacs, , Building Emacs, elisp, The GNU Emacs Lisp
17562 Reference Manual}, and the @file{INSTALL} file for more about
17565 You may also want to include autoloaded expressions in your @file{.emacs}
17566 file. @code{autoload} is a built-in function that takes up to five
17567 arguments, the final three of which are optional. The first argument
17568 is the name of the function to be autoloaded; the second is the name
17569 of the file to be loaded. The third argument is documentation for the
17570 function, and the fourth tells whether the function can be called
17571 interactively. The fifth argument tells what type of
17572 object---@code{autoload} can handle a keymap or macro as well as a
17573 function (the default is a function).
17576 Here is a typical example:
17580 (autoload 'html-helper-mode
17581 "html-helper-mode" "Edit HTML documents" t)
17586 (@code{html-helper-mode} is an older alternative to @code{html-mode},
17587 which is a standard part of the distribution.)
17590 This expression autoloads the @code{html-helper-mode} function. It
17591 takes it from the @file{html-helper-mode.el} file (or from the byte
17592 compiled version @file{html-helper-mode.elc}, if that exists.) The
17593 file must be located in a directory specified by @code{load-path}.
17594 The documentation says that this is a mode to help you edit documents
17595 written in the HyperText Markup Language. You can call this mode
17596 interactively by typing @kbd{M-x html-helper-mode}. (You need to
17597 duplicate the function's regular documentation in the autoload
17598 expression because the regular function is not yet loaded, so its
17599 documentation is not available.)
17601 @xref{Autoload, , Autoload, elisp, The GNU Emacs Lisp Reference
17602 Manual}, for more information.
17604 @node Simple Extension
17605 @section A Simple Extension: @code{line-to-top-of-window}
17606 @findex line-to-top-of-window
17607 @cindex Simple extension in @file{.emacs} file
17609 Here is a simple extension to Emacs that moves the line point is on to
17610 the top of the window. I use this all the time, to make text easier
17613 You can put the following code into a separate file and then load it
17614 from your @file{.emacs} file, or you can include it within your
17615 @file{.emacs} file.
17618 Here is the definition:
17622 ;;; Line to top of window;
17623 ;;; replace three keystroke sequence C-u 0 C-l
17624 (defun line-to-top-of-window ()
17625 "Move the line point is on to top of window."
17632 Now for the keybinding.
17634 Nowadays, function keys as well as mouse button events and
17635 non-@sc{ascii} characters are written within square brackets, without
17636 quotation marks. (In Emacs version 18 and before, you had to write
17637 different function key bindings for each different make of terminal.)
17639 I bind @code{line-to-top-of-window} to my @key{F6} function key like
17643 (global-set-key [f6] 'line-to-top-of-window)
17646 For more information, see @ref{Init Rebinding, , Rebinding Keys in
17647 Your Init File, emacs, The GNU Emacs Manual}.
17649 @cindex Conditional 'twixt two versions of Emacs
17650 @cindex Version of Emacs, choosing
17651 @cindex Emacs version, choosing
17652 If you run two versions of GNU Emacs, such as versions 22 and 23, and
17653 use one @file{.emacs} file, you can select which code to evaluate with
17654 the following conditional:
17659 ((= 22 emacs-major-version)
17660 ;; evaluate version 22 code
17662 ((= 23 emacs-major-version)
17663 ;; evaluate version 23 code
17668 For example, recent versions blink
17669 their cursors by default. I hate such blinking, as well as other
17670 features, so I placed the following in my @file{.emacs}
17671 file@footnote{When I start instances of Emacs that do not load my
17672 @file{.emacs} file or any site file, I also turn off blinking:
17675 emacs -q --no-site-file -eval '(blink-cursor-mode nil)'
17677 @exdent Or nowadays, using an even more sophisticated set of options,
17685 (when (>= emacs-major-version 21)
17686 (blink-cursor-mode 0)
17687 ;; Insert newline when you press `C-n' (next-line)
17688 ;; at the end of the buffer
17689 (setq next-line-add-newlines t)
17692 ;; Turn on image viewing
17693 (auto-image-file-mode t)
17696 ;; Turn on menu bar (this bar has text)
17697 ;; (Use numeric argument to turn on)
17701 ;; Turn off tool bar (this bar has icons)
17702 ;; (Use numeric argument to turn on)
17703 (tool-bar-mode nil)
17706 ;; Turn off tooltip mode for tool bar
17707 ;; (This mode causes icon explanations to pop up)
17708 ;; (Use numeric argument to turn on)
17710 ;; If tooltips turned on, make tips appear promptly
17711 (setq tooltip-delay 0.1) ; default is 0.7 second
17717 @section X11 Colors
17719 You can specify colors when you use Emacs with the MIT X Windowing
17722 I dislike the default colors and specify my own.
17725 Here are the expressions in my @file{.emacs}
17726 file that set values:
17730 ;; Set cursor color
17731 (set-cursor-color "white")
17734 (set-mouse-color "white")
17736 ;; Set foreground and background
17737 (set-foreground-color "white")
17738 (set-background-color "darkblue")
17742 ;;; Set highlighting colors for isearch and drag
17743 (set-face-foreground 'highlight "white")
17744 (set-face-background 'highlight "blue")
17748 (set-face-foreground 'region "cyan")
17749 (set-face-background 'region "blue")
17753 (set-face-foreground 'secondary-selection "skyblue")
17754 (set-face-background 'secondary-selection "darkblue")
17758 ;; Set calendar highlighting colors
17759 (setq calendar-load-hook
17761 (set-face-foreground 'diary-face "skyblue")
17762 (set-face-background 'holiday-face "slate blue")
17763 (set-face-foreground 'holiday-face "white")))
17767 The various shades of blue soothe my eye and prevent me from seeing
17768 the screen flicker.
17770 Alternatively, I could have set my specifications in various X
17771 initialization files. For example, I could set the foreground,
17772 background, cursor, and pointer (i.e., mouse) colors in my
17773 @file{~/.Xresources} file like this:
17777 Emacs*foreground: white
17778 Emacs*background: darkblue
17779 Emacs*cursorColor: white
17780 Emacs*pointerColor: white
17784 In any event, since it is not part of Emacs, I set the root color of
17785 my X window in my @file{~/.xinitrc} file, like this@footnote{I also
17786 run more modern window managers, such as Enlightenment, Gnome, or KDE;
17787 in those cases, I often specify an image rather than a plain color.}:
17790 xsetroot -solid Navy -fg white &
17794 @node Miscellaneous
17795 @section Miscellaneous Settings for a @file{.emacs} File
17798 Here are a few miscellaneous settings:
17803 Set the shape and color of the mouse cursor:
17807 ; Cursor shapes are defined in
17808 ; `/usr/include/X11/cursorfont.h';
17809 ; for example, the `target' cursor is number 128;
17810 ; the `top_left_arrow' cursor is number 132.
17814 (let ((mpointer (x-get-resource "*mpointer"
17815 "*emacs*mpointer")))
17816 ;; If you have not set your mouse pointer
17817 ;; then set it, otherwise leave as is:
17818 (if (eq mpointer nil)
17819 (setq mpointer "132")) ; top_left_arrow
17822 (setq x-pointer-shape (string-to-int mpointer))
17823 (set-mouse-color "white"))
17828 Or you can set the values of a variety of features in an alist, like
17834 default-frame-alist
17835 '((cursor-color . "white")
17836 (mouse-color . "white")
17837 (foreground-color . "white")
17838 (background-color . "DodgerBlue4")
17839 ;; (cursor-type . bar)
17840 (cursor-type . box)
17843 (tool-bar-lines . 0)
17844 (menu-bar-lines . 1)
17848 "-Misc-Fixed-Medium-R-Normal--20-200-75-75-C-100-ISO8859-1")
17854 Convert @kbd{@key{CTRL}-h} into @key{DEL} and @key{DEL}
17855 into @kbd{@key{CTRL}-h}.@*
17856 (Some older keyboards needed this, although I have not seen the
17861 ;; Translate `C-h' to <DEL>.
17862 ; (keyboard-translate ?\C-h ?\C-?)
17864 ;; Translate <DEL> to `C-h'.
17865 (keyboard-translate ?\C-? ?\C-h)
17869 @item Turn off a blinking cursor!
17873 (if (fboundp 'blink-cursor-mode)
17874 (blink-cursor-mode -1))
17879 or start GNU Emacs with the command @code{emacs -nbc}.
17882 @item When using `grep'@*
17883 @samp{-i}@w{ } Ignore case distinctions@*
17884 @samp{-n}@w{ } Prefix each line of output with line number@*
17885 @samp{-H}@w{ } Print the filename for each match.@*
17886 @samp{-e}@w{ } Protect patterns beginning with a hyphen character, @samp{-}
17889 (setq grep-command "grep -i -nH -e ")
17893 @c Evidently, no longer needed in GNU Emacs 22
17895 item Automatically uncompress compressed files when visiting them
17898 (load "uncompress")
17903 @item Find an existing buffer, even if it has a different name@*
17904 This avoids problems with symbolic links.
17907 (setq find-file-existing-other-name t)
17910 @item Set your language environment and default input method
17914 (set-language-environment "latin-1")
17915 ;; Remember you can enable or disable multilingual text input
17916 ;; with the @code{toggle-input-method'} (@kbd{C-\}) command
17917 (setq default-input-method "latin-1-prefix")
17921 If you want to write with Chinese `GB' characters, set this instead:
17925 (set-language-environment "Chinese-GB")
17926 (setq default-input-method "chinese-tonepy")
17931 @subsubheading Fixing Unpleasant Key Bindings
17932 @cindex Key bindings, fixing
17933 @cindex Bindings, key, fixing unpleasant
17935 Some systems bind keys unpleasantly. Sometimes, for example, the
17936 @key{CTRL} key appears in an awkward spot rather than at the far left
17939 Usually, when people fix these sorts of keybindings, they do not
17940 change their @file{~/.emacs} file. Instead, they bind the proper keys
17941 on their consoles with the @code{loadkeys} or @code{install-keymap}
17942 commands in their boot script and then include @code{xmodmap} commands
17943 in their @file{.xinitrc} or @file{.Xsession} file for X Windows.
17951 loadkeys /usr/share/keymaps/i386/qwerty/emacs2.kmap.gz
17953 install-keymap emacs2
17959 For a @file{.xinitrc} or @file{.Xsession} file when the @key{Caps
17960 Lock} key is at the far left of the home row:
17964 # Bind the key labeled `Caps Lock' to `Control'
17965 # (Such a broken user interface suggests that keyboard manufacturers
17966 # think that computers are typewriters from 1885.)
17968 xmodmap -e "clear Lock"
17969 xmodmap -e "add Control = Caps_Lock"
17975 In a @file{.xinitrc} or @file{.Xsession} file, to convert an @key{ALT}
17976 key to a @key{META} key:
17980 # Some ill designed keyboards have a key labeled ALT and no Meta
17981 xmodmap -e "keysym Alt_L = Meta_L Alt_L"
17987 @section A Modified Mode Line
17988 @vindex mode-line-format
17989 @cindex Mode line format
17991 Finally, a feature I really like: a modified mode line.
17993 When I work over a network, I forget which machine I am using. Also,
17994 I tend to I lose track of where I am, and which line point is on.
17996 So I reset my mode line to look like this:
17999 -:-- foo.texi rattlesnake:/home/bob/ Line 1 (Texinfo Fill) Top
18002 I am visiting a file called @file{foo.texi}, on my machine
18003 @file{rattlesnake} in my @file{/home/bob} buffer. I am on line 1, in
18004 Texinfo mode, and am at the top of the buffer.
18007 My @file{.emacs} file has a section that looks like this:
18011 ;; Set a Mode Line that tells me which machine, which directory,
18012 ;; and which line I am on, plus the other customary information.
18013 (setq-default mode-line-format
18017 "mouse-1: select window, mouse-2: delete others ..."))
18018 mode-line-mule-info
18020 mode-line-frame-identification
18024 mode-line-buffer-identification
18027 (system-name) 0 (string-match "\\..+" (system-name))))
18032 "mouse-1: select window, mouse-2: delete others ..."))
18033 (line-number-mode " Line %l ")
18039 "mouse-1: select window, mouse-2: delete others ..."))
18040 (:eval (mode-line-mode-name))
18043 #("%n" 0 2 (help-echo "mouse-2: widen" local-map (keymap ...)))
18052 Here, I redefine the default mode line. Most of the parts are from
18053 the original; but I make a few changes. I set the @emph{default} mode
18054 line format so as to permit various modes, such as Info, to override
18057 Many elements in the list are self-explanatory:
18058 @code{mode-line-modified} is a variable that tells whether the buffer
18059 has been modified, @code{mode-name} tells the name of the mode, and so
18060 on. However, the format looks complicated because of two features we
18061 have not discussed.
18063 @cindex Properties, in mode line example
18064 The first string in the mode line is a dash or hyphen, @samp{-}. In
18065 the old days, it would have been specified simply as @code{"-"}. But
18066 nowadays, Emacs can add properties to a string, such as highlighting
18067 or, as in this case, a help feature. If you place your mouse cursor
18068 over the hyphen, some help information appears (By default, you must
18069 wait seven-tenths of a second before the information appears. You can
18070 change that timing by changing the value of @code{tooltip-delay}.)
18073 The new string format has a special syntax:
18076 #("-" 0 1 (help-echo "mouse-1: select window, ..."))
18080 The @code{#(} begins a list. The first element of the list is the
18081 string itself, just one @samp{-}. The second and third
18082 elements specify the range over which the fourth element applies. A
18083 range starts @emph{after} a character, so a zero means the range
18084 starts just before the first character; a 1 means that the range ends
18085 just after the first character. The third element is the property for
18086 the range. It consists of a property list, a
18087 property name, in this case, @samp{help-echo}, followed by a value, in this
18088 case, a string. The second, third, and fourth elements of this new
18089 string format can be repeated.
18091 @xref{Text Properties, , Text Properties, elisp, The GNU Emacs Lisp
18092 Reference Manual}, and see @ref{Mode Line Format, , Mode Line Format,
18093 elisp, The GNU Emacs Lisp Reference Manual}, for more information.
18095 @code{mode-line-buffer-identification}
18096 displays the current buffer name. It is a list
18097 beginning @code{(#("%12b" 0 4 @dots{}}.
18098 The @code{#(} begins the list.
18100 The @samp{"%12b"} displays the current buffer name, using the
18101 @code{buffer-name} function with which we are familiar; the `12'
18102 specifies the maximum number of characters that will be displayed.
18103 When a name has fewer characters, whitespace is added to fill out to
18104 this number. (Buffer names can and often should be longer than 12
18105 characters; this length works well in a typical 80 column wide
18108 @code{:eval} says to evaluate the following form and use the result as
18109 a string to display. In this case, the expression displays the first
18110 component of the full system name. The end of the first component is
18111 a @samp{.} (`period'), so I use the @code{string-match} function to
18112 tell me the length of the first component. The substring from the
18113 zeroth character to that length is the name of the machine.
18116 This is the expression:
18121 (system-name) 0 (string-match "\\..+" (system-name))))
18125 @samp{%[} and @samp{%]} cause a pair of square brackets
18126 to appear for each recursive editing level. @samp{%n} says `Narrow'
18127 when narrowing is in effect. @samp{%P} tells you the percentage of
18128 the buffer that is above the bottom of the window, or `Top', `Bottom',
18129 or `All'. (A lower case @samp{p} tell you the percentage above the
18130 @emph{top} of the window.) @samp{%-} inserts enough dashes to fill
18133 Remember, ``You don't have to like Emacs to like it''---your own
18134 Emacs can have different colors, different commands, and different
18135 keys than a default Emacs.
18137 On the other hand, if you want to bring up a plain `out of the box'
18138 Emacs, with no customization, type:
18145 This will start an Emacs that does @emph{not} load your
18146 @file{~/.emacs} initialization file. A plain, default Emacs. Nothing
18153 GNU Emacs has two debuggers, @code{debug} and @code{edebug}. The
18154 first is built into the internals of Emacs and is always with you;
18155 the second requires that you instrument a function before you can use it.
18157 Both debuggers are described extensively in @ref{Debugging, ,
18158 Debugging Lisp Programs, elisp, The GNU Emacs Lisp Reference Manual}.
18159 In this chapter, I will walk through a short example of each.
18162 * debug:: How to use the built-in debugger.
18163 * debug-on-entry:: Start debugging when you call a function.
18164 * debug-on-quit:: Start debugging when you quit with @kbd{C-g}.
18165 * edebug:: How to use Edebug, a source level debugger.
18166 * Debugging Exercises::
18170 @section @code{debug}
18173 Suppose you have written a function definition that is intended to
18174 return the sum of the numbers 1 through a given number. (This is the
18175 @code{triangle} function discussed earlier. @xref{Decrementing
18176 Example, , Example with Decrementing Counter}, for a discussion.)
18177 @c xref{Decrementing Loop,, Loop with a Decrementing Counter}, for a discussion.)
18179 However, your function definition has a bug. You have mistyped
18180 @samp{1=} for @samp{1-}. Here is the broken definition:
18182 @findex triangle-bugged
18185 (defun triangle-bugged (number)
18186 "Return sum of numbers 1 through NUMBER inclusive."
18188 (while (> number 0)
18189 (setq total (+ total number))
18190 (setq number (1= number))) ; @r{Error here.}
18195 If you are reading this in Info, you can evaluate this definition in
18196 the normal fashion. You will see @code{triangle-bugged} appear in the
18200 Now evaluate the @code{triangle-bugged} function with an
18204 (triangle-bugged 4)
18208 In a recent GNU Emacs, you will create and enter a @file{*Backtrace*}
18214 ---------- Buffer: *Backtrace* ----------
18215 Debugger entered--Lisp error: (void-function 1=)
18217 (setq number (1= number))
18218 (while (> number 0) (setq total (+ total number))
18219 (setq number (1= number)))
18220 (let ((total 0)) (while (> number 0) (setq total ...)
18221 (setq number ...)) total)
18225 eval((triangle-bugged 4))
18226 eval-last-sexp-1(nil)
18227 eval-last-sexp(nil)
18228 call-interactively(eval-last-sexp)
18229 ---------- Buffer: *Backtrace* ----------
18234 (I have reformatted this example slightly; the debugger does not fold
18235 long lines. As usual, you can quit the debugger by typing @kbd{q} in
18236 the @file{*Backtrace*} buffer.)
18238 In practice, for a bug as simple as this, the `Lisp error' line will
18239 tell you what you need to know to correct the definition. The
18240 function @code{1=} is `void'.
18244 In GNU Emacs 20 and before, you will see:
18247 Symbol's function definition is void:@: 1=
18251 which has the same meaning as the @file{*Backtrace*} buffer line in
18255 However, suppose you are not quite certain what is going on?
18256 You can read the complete backtrace.
18258 In this case, you need to run a recent GNU Emacs, which automatically
18259 starts the debugger that puts you in the @file{*Backtrace*} buffer; or
18260 else, you need to start the debugger manually as described below.
18262 Read the @file{*Backtrace*} buffer from the bottom up; it tells you
18263 what Emacs did that led to the error. Emacs made an interactive call
18264 to @kbd{C-x C-e} (@code{eval-last-sexp}), which led to the evaluation
18265 of the @code{triangle-bugged} expression. Each line above tells you
18266 what the Lisp interpreter evaluated next.
18269 The third line from the top of the buffer is
18272 (setq number (1= number))
18276 Emacs tried to evaluate this expression; in order to do so, it tried
18277 to evaluate the inner expression shown on the second line from the
18286 This is where the error occurred; as the top line says:
18289 Debugger entered--Lisp error: (void-function 1=)
18293 You can correct the mistake, re-evaluate the function definition, and
18294 then run your test again.
18296 @node debug-on-entry
18297 @section @code{debug-on-entry}
18298 @findex debug-on-entry
18300 A recent GNU Emacs starts the debugger automatically when your
18301 function has an error.
18304 GNU Emacs version 20 and before did not; it simply
18305 presented you with an error message. You had to start the debugger
18309 Incidentally, you can start the debugger manually for all versions of
18310 Emacs; the advantage is that the debugger runs even if you do not have
18311 a bug in your code. Sometimes your code will be free of bugs!
18313 You can enter the debugger when you call the function by calling
18314 @code{debug-on-entry}.
18321 M-x debug-on-entry RET triangle-bugged RET
18326 Now, evaluate the following:
18329 (triangle-bugged 5)
18333 All versions of Emacs will create a @file{*Backtrace*} buffer and tell
18334 you that it is beginning to evaluate the @code{triangle-bugged}
18339 ---------- Buffer: *Backtrace* ----------
18340 Debugger entered--entering a function:
18341 * triangle-bugged(5)
18342 eval((triangle-bugged 5))
18345 eval-last-sexp-1(nil)
18346 eval-last-sexp(nil)
18347 call-interactively(eval-last-sexp)
18348 ---------- Buffer: *Backtrace* ----------
18352 In the @file{*Backtrace*} buffer, type @kbd{d}. Emacs will evaluate
18353 the first expression in @code{triangle-bugged}; the buffer will look
18358 ---------- Buffer: *Backtrace* ----------
18359 Debugger entered--beginning evaluation of function call form:
18360 * (let ((total 0)) (while (> number 0) (setq total ...)
18361 (setq number ...)) total)
18362 * triangle-bugged(5)
18363 eval((triangle-bugged 5))
18366 eval-last-sexp-1(nil)
18367 eval-last-sexp(nil)
18368 call-interactively(eval-last-sexp)
18369 ---------- Buffer: *Backtrace* ----------
18374 Now, type @kbd{d} again, eight times, slowly. Each time you type
18375 @kbd{d}, Emacs will evaluate another expression in the function
18379 Eventually, the buffer will look like this:
18383 ---------- Buffer: *Backtrace* ----------
18384 Debugger entered--beginning evaluation of function call form:
18385 * (setq number (1= number))
18386 * (while (> number 0) (setq total (+ total number))
18387 (setq number (1= number)))
18390 * (let ((total 0)) (while (> number 0) (setq total ...)
18391 (setq number ...)) total)
18392 * triangle-bugged(5)
18393 eval((triangle-bugged 5))
18396 eval-last-sexp-1(nil)
18397 eval-last-sexp(nil)
18398 call-interactively(eval-last-sexp)
18399 ---------- Buffer: *Backtrace* ----------
18405 Finally, after you type @kbd{d} two more times, Emacs will reach the
18406 error, and the top two lines of the @file{*Backtrace*} buffer will look
18411 ---------- Buffer: *Backtrace* ----------
18412 Debugger entered--Lisp error: (void-function 1=)
18415 ---------- Buffer: *Backtrace* ----------
18419 By typing @kbd{d}, you were able to step through the function.
18421 You can quit a @file{*Backtrace*} buffer by typing @kbd{q} in it; this
18422 quits the trace, but does not cancel @code{debug-on-entry}.
18424 @findex cancel-debug-on-entry
18425 To cancel the effect of @code{debug-on-entry}, call
18426 @code{cancel-debug-on-entry} and the name of the function, like this:
18429 M-x cancel-debug-on-entry RET triangle-bugged RET
18433 (If you are reading this in Info, cancel @code{debug-on-entry} now.)
18435 @node debug-on-quit
18436 @section @code{debug-on-quit} and @code{(debug)}
18438 In addition to setting @code{debug-on-error} or calling @code{debug-on-entry},
18439 there are two other ways to start @code{debug}.
18441 @findex debug-on-quit
18442 You can start @code{debug} whenever you type @kbd{C-g}
18443 (@code{keyboard-quit}) by setting the variable @code{debug-on-quit} to
18444 @code{t}. This is useful for debugging infinite loops.
18447 @cindex @code{(debug)} in code
18448 Or, you can insert a line that says @code{(debug)} into your code
18449 where you want the debugger to start, like this:
18453 (defun triangle-bugged (number)
18454 "Return sum of numbers 1 through NUMBER inclusive."
18456 (while (> number 0)
18457 (setq total (+ total number))
18458 (debug) ; @r{Start debugger.}
18459 (setq number (1= number))) ; @r{Error here.}
18464 The @code{debug} function is described in detail in @ref{Debugger, ,
18465 The Lisp Debugger, elisp, The GNU Emacs Lisp Reference Manual}.
18468 @section The @code{edebug} Source Level Debugger
18469 @cindex Source level debugger
18472 Edebug is a source level debugger. Edebug normally displays the
18473 source of the code you are debugging, with an arrow at the left that
18474 shows which line you are currently executing.
18476 You can walk through the execution of a function, line by line, or run
18477 quickly until reaching a @dfn{breakpoint} where execution stops.
18479 Edebug is described in @ref{Edebug, , , elisp, The GNU Emacs
18480 Lisp Reference Manual}.
18483 Here is a bugged function definition for @code{triangle-recursively}.
18484 @xref{Recursive triangle function, , Recursion in place of a counter},
18485 for a review of it.
18489 (defun triangle-recursively-bugged (number)
18490 "Return sum of numbers 1 through NUMBER inclusive.
18495 (triangle-recursively-bugged
18496 (1= number))))) ; @r{Error here.}
18501 Normally, you would install this definition by positioning your cursor
18502 after the function's closing parenthesis and typing @kbd{C-x C-e}
18503 (@code{eval-last-sexp}) or else by positioning your cursor within the
18504 definition and typing @kbd{C-M-x} (@code{eval-defun}). (By default,
18505 the @code{eval-defun} command works only in Emacs Lisp mode or in Lisp
18509 However, to prepare this function definition for Edebug, you must
18510 first @dfn{instrument} the code using a different command. You can do
18511 this by positioning your cursor within or just after the definition
18515 M-x edebug-defun RET
18519 This will cause Emacs to load Edebug automatically if it is not
18520 already loaded, and properly instrument the function.
18522 After instrumenting the function, place your cursor after the
18523 following expression and type @kbd{C-x C-e} (@code{eval-last-sexp}):
18526 (triangle-recursively-bugged 3)
18530 You will be jumped back to the source for
18531 @code{triangle-recursively-bugged} and the cursor positioned at the
18532 beginning of the @code{if} line of the function. Also, you will see
18533 an arrowhead at the left hand side of that line. The arrowhead marks
18534 the line where the function is executing. (In the following examples,
18535 we show the arrowhead with @samp{=>}; in a windowing system, you may
18536 see the arrowhead as a solid triangle in the window `fringe'.)
18539 =>@point{}(if (= number 1)
18544 In the example, the location of point is displayed with a star,
18545 @samp{@point{}} (in Info, it is displayed as @samp{-!-}).
18548 In the example, the location of point is displayed as @samp{@point{}}
18549 (in a printed book, it is displayed with a five pointed star).
18552 If you now press @key{SPC}, point will move to the next expression to
18553 be executed; the line will look like this:
18556 =>(if @point{}(= number 1)
18560 As you continue to press @key{SPC}, point will move from expression to
18561 expression. At the same time, whenever an expression returns a value,
18562 that value will be displayed in the echo area. For example, after you
18563 move point past @code{number}, you will see the following:
18566 Result: 3 (#o3, #x3, ?\C-c)
18570 This means the value of @code{number} is 3, which is octal three,
18571 hexadecimal three, and @sc{ascii} `control-c' (the third letter of the
18572 alphabet, in case you need to know this information).
18574 You can continue moving through the code until you reach the line with
18575 the error. Before evaluation, that line looks like this:
18578 => @point{}(1= number))))) ; @r{Error here.}
18583 When you press @key{SPC} once again, you will produce an error message
18587 Symbol's function definition is void:@: 1=
18593 Press @kbd{q} to quit Edebug.
18595 To remove instrumentation from a function definition, simply
18596 re-evaluate it with a command that does not instrument it.
18597 For example, you could place your cursor after the definition's
18598 closing parenthesis and type @kbd{C-x C-e}.
18600 Edebug does a great deal more than walk with you through a function.
18601 You can set it so it races through on its own, stopping only at an
18602 error or at specified stopping points; you can cause it to display the
18603 changing values of various expressions; you can find out how many
18604 times a function is called, and more.
18606 Edebug is described in @ref{Edebug, , , elisp, The GNU Emacs
18607 Lisp Reference Manual}.
18610 @node Debugging Exercises
18611 @section Debugging Exercises
18615 Install the @code{@value{COUNT-WORDS}} function and then cause it to
18616 enter the built-in debugger when you call it. Run the command on a
18617 region containing two words. You will need to press @kbd{d} a
18618 remarkable number of times. On your system, is a `hook' called after
18619 the command finishes? (For information on hooks, see @ref{Command
18620 Overview, , Command Loop Overview, elisp, The GNU Emacs Lisp Reference
18624 Copy @code{@value{COUNT-WORDS}} into the @file{*scratch*} buffer,
18625 instrument the function for Edebug, and walk through its execution.
18626 The function does not need to have a bug, although you can introduce
18627 one if you wish. If the function lacks a bug, the walk-through
18628 completes without problems.
18631 While running Edebug, type @kbd{?} to see a list of all the Edebug commands.
18632 (The @code{global-edebug-prefix} is usually @kbd{C-x X}, i.e.,
18633 @kbd{@key{CTRL}-x} followed by an upper case @kbd{X}; use this prefix
18634 for commands made outside of the Edebug debugging buffer.)
18637 In the Edebug debugging buffer, use the @kbd{p}
18638 (@code{edebug-bounce-point}) command to see where in the region the
18639 @code{@value{COUNT-WORDS}} is working.
18642 Move point to some spot further down the function and then type the
18643 @kbd{h} (@code{edebug-goto-here}) command to jump to that location.
18646 Use the @kbd{t} (@code{edebug-trace-mode}) command to cause Edebug to
18647 walk through the function on its own; use an upper case @kbd{T} for
18648 @code{edebug-Trace-fast-mode}.
18651 Set a breakpoint, then run Edebug in Trace mode until it reaches the
18656 @chapter Conclusion
18658 We have now reached the end of this Introduction. You have now
18659 learned enough about programming in Emacs Lisp to set values, to write
18660 simple @file{.emacs} files for yourself and your friends, and write
18661 simple customizations and extensions to Emacs.
18663 This is a place to stop. Or, if you wish, you can now go onward, and
18666 You have learned some of the basic nuts and bolts of programming. But
18667 only some. There are a great many more brackets and hinges that are
18668 easy to use that we have not touched.
18670 A path you can follow right now lies among the sources to GNU Emacs
18673 @cite{The GNU Emacs Lisp Reference Manual}.
18676 @ref{Top, , The GNU Emacs Lisp Reference Manual, elisp, The GNU
18677 Emacs Lisp Reference Manual}.
18680 The Emacs Lisp sources are an adventure. When you read the sources and
18681 come across a function or expression that is unfamiliar, you need to
18682 figure out or find out what it does.
18684 Go to the Reference Manual. It is a thorough, complete, and fairly
18685 easy-to-read description of Emacs Lisp. It is written not only for
18686 experts, but for people who know what you know. (The @cite{Reference
18687 Manual} comes with the standard GNU Emacs distribution. Like this
18688 introduction, it comes as a Texinfo source file, so you can read it
18689 on-line and as a typeset, printed book.)
18691 Go to the other on-line help that is part of GNU Emacs: the on-line
18692 documentation for all functions and variables, and @code{find-tag},
18693 the program that takes you to sources.
18695 Here is an example of how I explore the sources. Because of its name,
18696 @file{simple.el} is the file I looked at first, a long time ago. As
18697 it happens some of the functions in @file{simple.el} are complicated,
18698 or at least look complicated at first sight. The @code{open-line}
18699 function, for example, looks complicated.
18701 You may want to walk through this function slowly, as we did with the
18702 @code{forward-sentence} function. (@xref{forward-sentence, The
18703 @code{forward-sentence} function}.) Or you may want to skip that
18704 function and look at another, such as @code{split-line}. You don't
18705 need to read all the functions. According to
18706 @code{count-words-in-defun}, the @code{split-line} function contains
18707 102 words and symbols.
18709 Even though it is short, @code{split-line} contains expressions
18710 we have not studied: @code{skip-chars-forward}, @code{indent-to},
18711 @code{current-column} and @code{insert-and-inherit}.
18713 Consider the @code{skip-chars-forward} function. (It is part of the
18714 function definition for @code{back-to-indentation}, which is shown in
18715 @ref{Review, , Review}.)
18717 In GNU Emacs, you can find out more about @code{skip-chars-forward} by
18718 typing @kbd{C-h f} (@code{describe-function}) and the name of the
18719 function. This gives you the function documentation.
18721 You may be able to guess what is done by a well named function such as
18722 @code{indent-to}; or you can look it up, too. Incidentally, the
18723 @code{describe-function} function itself is in @file{help.el}; it is
18724 one of those long, but decipherable functions. You can look up
18725 @code{describe-function} using the @kbd{C-h f} command!
18727 In this instance, since the code is Lisp, the @file{*Help*} buffer
18728 contains the name of the library containing the function's source.
18729 You can put point over the name of the library and press the RET key,
18730 which in this situation is bound to @code{help-follow}, and be taken
18731 directly to the source, in the same way as @kbd{M-.}
18734 The definition for @code{describe-function} illustrates how to
18735 customize the @code{interactive} expression without using the standard
18736 character codes; and it shows how to create a temporary buffer.
18738 (The @code{indent-to} function is written in C rather than Emacs Lisp;
18739 it is a `built-in' function. @code{help-follow} takes you to its
18740 source as does @code{find-tag}, when properly set up.)
18742 You can look at a function's source using @code{find-tag}, which is
18743 bound to @kbd{M-.} Finally, you can find out what the Reference
18744 Manual has to say by visiting the manual in Info, and typing @kbd{i}
18745 (@code{Info-index}) and the name of the function, or by looking up the
18746 function in the index to a printed copy of the manual.
18748 Similarly, you can find out what is meant by
18749 @code{insert-and-inherit}.
18751 Other interesting source files include @file{paragraphs.el},
18752 @file{loaddefs.el}, and @file{loadup.el}. The @file{paragraphs.el}
18753 file includes short, easily understood functions as well as longer
18754 ones. The @file{loaddefs.el} file contains the many standard
18755 autoloads and many keymaps. I have never looked at it all; only at
18756 parts. @file{loadup.el} is the file that loads the standard parts of
18757 Emacs; it tells you a great deal about how Emacs is built.
18758 (@xref{Building Emacs, , Building Emacs, elisp, The GNU Emacs Lisp
18759 Reference Manual}, for more about building.)
18761 As I said, you have learned some nuts and bolts; however, and very
18762 importantly, we have hardly touched major aspects of programming; I
18763 have said nothing about how to sort information, except to use the
18764 predefined @code{sort} function; I have said nothing about how to store
18765 information, except to use variables and lists; I have said nothing
18766 about how to write programs that write programs. These are topics for
18767 another, and different kind of book, a different kind of learning.
18769 What you have done is learn enough for much practical work with GNU
18770 Emacs. What you have done is get started. This is the end of a
18773 @c ================ Appendix ================
18776 @appendix The @code{the-the} Function
18778 @cindex Duplicated words function
18779 @cindex Words, duplicated
18781 Sometimes when you you write text, you duplicate words---as with ``you
18782 you'' near the beginning of this sentence. I find that most
18783 frequently, I duplicate ``the''; hence, I call the function for
18784 detecting duplicated words, @code{the-the}.
18787 As a first step, you could use the following regular expression to
18788 search for duplicates:
18791 \\(\\w+[ \t\n]+\\)\\1
18795 This regexp matches one or more word-constituent characters followed
18796 by one or more spaces, tabs, or newlines. However, it does not detect
18797 duplicated words on different lines, since the ending of the first
18798 word, the end of the line, is different from the ending of the second
18799 word, a space. (For more information about regular expressions, see
18800 @ref{Regexp Search, , Regular Expression Searches}, as well as
18801 @ref{Regexps, , Syntax of Regular Expressions, emacs, The GNU Emacs
18802 Manual}, and @ref{Regular Expressions, , Regular Expressions, elisp,
18803 The GNU Emacs Lisp Reference Manual}.)
18805 You might try searching just for duplicated word-constituent
18806 characters but that does not work since the pattern detects doubles
18807 such as the two occurrences of `th' in `with the'.
18809 Another possible regexp searches for word-constituent characters
18810 followed by non-word-constituent characters, reduplicated. Here,
18811 @w{@samp{\\w+}} matches one or more word-constituent characters and
18812 @w{@samp{\\W*}} matches zero or more non-word-constituent characters.
18815 \\(\\(\\w+\\)\\W*\\)\\1
18821 Here is the pattern that I use. It is not perfect, but good enough.
18822 @w{@samp{\\b}} matches the empty string, provided it is at the beginning
18823 or end of a word; @w{@samp{[^@@ \n\t]+}} matches one or more occurrences of
18824 any characters that are @emph{not} an @@-sign, space, newline, or tab.
18827 \\b\\([^@@ \n\t]+\\)[ \n\t]+\\1\\b
18830 One can write more complicated expressions, but I found that this
18831 expression is good enough, so I use it.
18833 Here is the @code{the-the} function, as I include it in my
18834 @file{.emacs} file, along with a handy global key binding:
18839 "Search forward for for a duplicated word."
18841 (message "Searching for for duplicated words ...")
18845 ;; This regexp is not perfect
18846 ;; but is fairly good over all:
18847 (if (re-search-forward
18848 "\\b\\([^@@ \n\t]+\\)[ \n\t]+\\1\\b" nil 'move)
18849 (message "Found duplicated word.")
18850 (message "End of buffer")))
18854 ;; Bind `the-the' to C-c \
18855 (global-set-key "\C-c\\" 'the-the)
18864 one two two three four five
18869 You can substitute the other regular expressions shown above in the
18870 function definition and try each of them on this list.
18873 @appendix Handling the Kill Ring
18874 @cindex Kill ring handling
18875 @cindex Handling the kill ring
18876 @cindex Ring, making a list like a
18878 The kill ring is a list that is transformed into a ring by the
18879 workings of the @code{current-kill} function. The @code{yank} and
18880 @code{yank-pop} commands use the @code{current-kill} function.
18882 This appendix describes the @code{current-kill} function as well as
18883 both the @code{yank} and the @code{yank-pop} commands, but first,
18884 consider the workings of the kill ring.
18887 * What the Kill Ring Does::
18889 * yank:: Paste a copy of a clipped element.
18890 * yank-pop:: Insert element pointed to.
18895 @node What the Kill Ring Does
18896 @unnumberedsec What the Kill Ring Does
18900 The kill ring has a default maximum length of sixty items; this number
18901 is too large for an explanation. Instead, set it to four. Please
18902 evaluate the following:
18906 (setq old-kill-ring-max kill-ring-max)
18907 (setq kill-ring-max 4)
18912 Then, please copy each line of the following indented example into the
18913 kill ring. You may kill each line with @kbd{C-k} or mark it and copy
18917 (In a read-only buffer, such as the @file{*info*} buffer, the kill
18918 command, @kbd{C-k} (@code{kill-line}), will not remove the text,
18919 merely copy it to the kill ring. However, your machine may beep at
18920 you. Alternatively, for silence, you may copy the region of each line
18921 with the @kbd{M-w} (@code{kill-ring-save}) command. You must mark
18922 each line for this command to succeed, but it does not matter at which
18923 end you put point or mark.)
18927 Please invoke the calls in order, so that five elements attempt to
18928 fill the kill ring:
18933 second piece of text
18935 fourth line of text
18942 Then find the value of @code{kill-ring} by evaluating
18954 ("fifth bit of text" "fourth line of text"
18955 "third line" "second piece of text")
18960 The first element, @samp{first some text}, was dropped.
18963 To return to the old value for the length of the kill ring, evaluate:
18966 (setq kill-ring-max old-kill-ring-max)
18970 @appendixsec The @code{current-kill} Function
18971 @findex current-kill
18973 The @code{current-kill} function changes the element in the kill ring
18974 to which @code{kill-ring-yank-pointer} points. (Also, the
18975 @code{kill-new} function sets @code{kill-ring-yank-pointer} to point
18976 to the latest element of the kill ring. The @code{kill-new}
18977 function is used directly or indirectly by @code{kill-append},
18978 @code{copy-region-as-kill}, @code{kill-ring-save}, @code{kill-line},
18979 and @code{kill-region}.)
18982 * Code for current-kill::
18983 * Understanding current-kill::
18987 @node Code for current-kill
18988 @unnumberedsubsec The code for @code{current-kill}
18993 The @code{current-kill} function is used by @code{yank} and by
18994 @code{yank-pop}. Here is the code for @code{current-kill}:
18998 (defun current-kill (n &optional do-not-move)
18999 "Rotate the yanking point by N places, and then return that kill.
19000 If N is zero, `interprogram-paste-function' is set, and calling it
19001 returns a string, then that string is added to the front of the
19002 kill ring and returned as the latest kill.
19005 If optional arg DO-NOT-MOVE is non-nil, then don't actually move the
19006 yanking point; just return the Nth kill forward."
19007 (let ((interprogram-paste (and (= n 0)
19008 interprogram-paste-function
19009 (funcall interprogram-paste-function))))
19012 (if interprogram-paste
19014 ;; Disable the interprogram cut function when we add the new
19015 ;; text to the kill ring, so Emacs doesn't try to own the
19016 ;; selection, with identical text.
19017 (let ((interprogram-cut-function nil))
19018 (kill-new interprogram-paste))
19019 interprogram-paste)
19022 (or kill-ring (error "Kill ring is empty"))
19023 (let ((ARGth-kill-element
19024 (nthcdr (mod (- n (length kill-ring-yank-pointer))
19025 (length kill-ring))
19028 (setq kill-ring-yank-pointer ARGth-kill-element))
19029 (car ARGth-kill-element)))))
19033 Remember also that the @code{kill-new} function sets
19034 @code{kill-ring-yank-pointer} to the latest element of the kill
19035 ring, which means that all the functions that call it set the value
19036 indirectly: @code{kill-append}, @code{copy-region-as-kill},
19037 @code{kill-ring-save}, @code{kill-line}, and @code{kill-region}.
19040 Here is the line in @code{kill-new}, which is explained in
19041 @ref{kill-new function, , The @code{kill-new} function}.
19044 (setq kill-ring-yank-pointer kill-ring)
19048 @node Understanding current-kill
19049 @unnumberedsubsec @code{current-kill} in Outline
19052 The @code{current-kill} function looks complex, but as usual, it can
19053 be understood by taking it apart piece by piece. First look at it in
19058 (defun current-kill (n &optional do-not-move)
19059 "Rotate the yanking point by N places, and then return that kill."
19065 This function takes two arguments, one of which is optional. It has a
19066 documentation string. It is @emph{not} interactive.
19069 * Body of current-kill::
19070 * Digression concerning error:: How to mislead humans, but not computers.
19071 * Determining the Element::
19075 @node Body of current-kill
19076 @unnumberedsubsubsec The Body of @code{current-kill}
19079 The body of the function definition is a @code{let} expression, which
19080 itself has a body as well as a @var{varlist}.
19082 The @code{let} expression declares a variable that will be only usable
19083 within the bounds of this function. This variable is called
19084 @code{interprogram-paste} and is for copying to another program. It
19085 is not for copying within this instance of GNU Emacs. Most window
19086 systems provide a facility for interprogram pasting. Sadly, that
19087 facility usually provides only for the last element. Most windowing
19088 systems have not adopted a ring of many possibilities, even though
19089 Emacs has provided it for decades.
19091 The @code{if} expression has two parts, one if there exists
19092 @code{interprogram-paste} and one if not.
19095 Let us consider the `if not' or else-part of the @code{current-kill}
19096 function. (The then-part uses the @code{kill-new} function, which
19097 we have already described. @xref{kill-new function, , The
19098 @code{kill-new} function}.)
19102 (or kill-ring (error "Kill ring is empty"))
19103 (let ((ARGth-kill-element
19104 (nthcdr (mod (- n (length kill-ring-yank-pointer))
19105 (length kill-ring))
19108 (setq kill-ring-yank-pointer ARGth-kill-element))
19109 (car ARGth-kill-element))
19114 The code first checks whether the kill ring has content; otherwise it
19118 Note that the @code{or} expression is very similar to testing length
19125 (if (zerop (length kill-ring)) ; @r{if-part}
19126 (error "Kill ring is empty")) ; @r{then-part}
19132 If there is not anything in the kill ring, its length must be zero and
19133 an error message sent to the user: @samp{Kill ring is empty}. The
19134 @code{current-kill} function uses an @code{or} expression which is
19135 simpler. But an @code{if} expression reminds us what goes on.
19137 This @code{if} expression uses the function @code{zerop} which returns
19138 true if the value it is testing is zero. When @code{zerop} tests
19139 true, the then-part of the @code{if} is evaluated. The then-part is a
19140 list starting with the function @code{error}, which is a function that
19141 is similar to the @code{message} function
19142 (@pxref{message, , The @code{message} Function}) in that
19143 it prints a one-line message in the echo area. However, in addition
19144 to printing a message, @code{error} also stops evaluation of the
19145 function within which it is embedded. This means that the rest of the
19146 function will not be evaluated if the length of the kill ring is zero.
19148 Then the @code{current-kill} function selects the element to return.
19149 The selection depends on the number of places that @code{current-kill}
19150 rotates and on where @code{kill-ring-yank-pointer} points.
19152 Next, either the optional @code{do-not-move} argument is true or the
19153 current value of @code{kill-ring-yank-pointer} is set to point to the
19154 list. Finally, another expression returns the first element of the
19155 list even if the @code{do-not-move} argument is true.
19158 @node Digression concerning error
19159 @unnumberedsubsubsec Digression about the word `error'
19162 In my opinion, it is slightly misleading, at least to humans, to use
19163 the term `error' as the name of the @code{error} function. A better
19164 term would be `cancel'. Strictly speaking, of course, you cannot
19165 point to, much less rotate a pointer to a list that has no length, so
19166 from the point of view of the computer, the word `error' is correct.
19167 But a human expects to attempt this sort of thing, if only to find out
19168 whether the kill ring is full or empty. This is an act of
19171 From the human point of view, the act of exploration and discovery is
19172 not necessarily an error, and therefore should not be labeled as one,
19173 even in the bowels of a computer. As it is, the code in Emacs implies
19174 that a human who is acting virtuously, by exploring his or her
19175 environment, is making an error. This is bad. Even though the computer
19176 takes the same steps as it does when there is an `error', a term such as
19177 `cancel' would have a clearer connotation.
19180 @node Determining the Element
19181 @unnumberedsubsubsec Determining the Element
19184 Among other actions, the else-part of the @code{if} expression sets
19185 the value of @code{kill-ring-yank-pointer} to
19186 @code{ARGth-kill-element} when the kill ring has something in it and
19187 the value of @code{do-not-move} is @code{nil}.
19190 The code looks like this:
19194 (nthcdr (mod (- n (length kill-ring-yank-pointer))
19195 (length kill-ring))
19200 This needs some examination. Unless it is not supposed to move the
19201 pointer, the @code{current-kill} function changes where
19202 @code{kill-ring-yank-pointer} points.
19204 @w{@code{(setq kill-ring-yank-pointer ARGth-kill-element))}}
19205 expression does. Also, clearly, @code{ARGth-kill-element} is being
19206 set to be equal to some @sc{cdr} of the kill ring, using the
19207 @code{nthcdr} function that is described in an earlier section.
19208 (@xref{copy-region-as-kill}.) How does it do this?
19210 As we have seen before (@pxref{nthcdr}), the @code{nthcdr} function
19211 works by repeatedly taking the @sc{cdr} of a list---it takes the
19212 @sc{cdr} of the @sc{cdr} of the @sc{cdr} @dots{}
19215 The two following expressions produce the same result:
19219 (setq kill-ring-yank-pointer (cdr kill-ring))
19221 (setq kill-ring-yank-pointer (nthcdr 1 kill-ring))
19225 However, the @code{nthcdr} expression is more complicated. It uses
19226 the @code{mod} function to determine which @sc{cdr} to select.
19228 (You will remember to look at inner functions first; indeed, we will
19229 have to go inside the @code{mod}.)
19231 The @code{mod} function returns the value of its first argument modulo
19232 the second; that is to say, it returns the remainder after dividing
19233 the first argument by the second. The value returned has the same
19234 sign as the second argument.
19242 @result{} 0 ;; @r{because there is no remainder}
19249 In this case, the first argument is often smaller than the second.
19261 We can guess what the @code{-} function does. It is like @code{+} but
19262 subtracts instead of adds; the @code{-} function subtracts its second
19263 argument from its first. Also, we already know what the @code{length}
19264 function does (@pxref{length}). It returns the length of a list.
19266 And @code{n} is the name of the required argument to the
19267 @code{current-kill} function.
19270 So when the first argument to @code{nthcdr} is zero, the @code{nthcdr}
19271 expression returns the whole list, as you can see by evaluating the
19276 ;; kill-ring-yank-pointer @r{and} kill-ring @r{have a length of four}
19277 ;; @r{and} (mod (- 0 4) 4) @result{} 0
19278 (nthcdr (mod (- 0 4) 4)
19279 '("fourth line of text"
19281 "second piece of text"
19282 "first some text"))
19287 When the first argument to the @code{current-kill} function is one,
19288 the @code{nthcdr} expression returns the list without its first
19293 (nthcdr (mod (- 1 4) 4)
19294 '("fourth line of text"
19296 "second piece of text"
19297 "first some text"))
19301 @cindex @samp{global variable} defined
19302 @cindex @samp{variable, global}, defined
19303 Incidentally, both @code{kill-ring} and @code{kill-ring-yank-pointer}
19304 are @dfn{global variables}. That means that any expression in Emacs
19305 Lisp can access them. They are not like the local variables set by
19306 @code{let} or like the symbols in an argument list.
19307 Local variables can only be accessed
19308 within the @code{let} that defines them or the function that specifies
19309 them in an argument list (and within expressions called by them).
19312 @c texi2dvi fails when the name of the section is within ifnottex ...
19313 (@xref{Prevent confusion, , @code{let} Prevents Confusion}, and
19314 @ref{defun, , The @code{defun} Macro}.)
19318 @appendixsec @code{yank}
19321 After learning about @code{current-kill}, the code for the
19322 @code{yank} function is almost easy.
19324 The @code{yank} function does not use the
19325 @code{kill-ring-yank-pointer} variable directly. It calls
19326 @code{insert-for-yank} which calls @code{current-kill} which sets the
19327 @code{kill-ring-yank-pointer} variable.
19330 The code looks like this:
19335 (defun yank (&optional arg)
19336 "Reinsert (\"paste\") the last stretch of killed text.
19337 More precisely, reinsert the stretch of killed text most recently
19338 killed OR yanked. Put point at end, and set mark at beginning.
19339 With just \\[universal-argument] as argument, same but put point at
19340 beginning (and mark at end). With argument N, reinsert the Nth most
19341 recently killed stretch of killed text.
19343 When this command inserts killed text into the buffer, it honors
19344 `yank-excluded-properties' and `yank-handler' as described in the
19345 doc string for `insert-for-yank-1', which see.
19347 See also the command \\[yank-pop]."
19351 (setq yank-window-start (window-start))
19352 ;; If we don't get all the way thru, make last-command indicate that
19353 ;; for the following command.
19354 (setq this-command t)
19355 (push-mark (point))
19358 (insert-for-yank (current-kill (cond
19363 ;; This is like exchange-point-and-mark,
19364 ;; but doesn't activate the mark.
19365 ;; It is cleaner to avoid activation, even though the command
19366 ;; loop would deactivate the mark because we inserted text.
19367 (goto-char (prog1 (mark t)
19368 (set-marker (mark-marker) (point) (current-buffer)))))
19371 ;; If we do get all the way thru, make this-command indicate that.
19372 (if (eq this-command t)
19373 (setq this-command 'yank))
19378 The key expression is @code{insert-for-yank}, which inserts the string
19379 returned by @code{current-kill}, but removes some text properties from
19382 However, before getting to that expression, the function sets the value
19383 of @code{yank-window-start} to the position returned by the
19384 @code{(window-start)} expression, the position at which the display
19385 currently starts. The @code{yank} function also sets
19386 @code{this-command} and pushes the mark.
19388 After it yanks the appropriate element, if the optional argument is a
19389 @sc{cons} rather than a number or nothing, it puts point at beginning
19390 of the yanked text and mark at its end.
19392 (The @code{prog1} function is like @code{progn} but returns the value
19393 of its first argument rather than the value of its last argument. Its
19394 first argument is forced to return the buffer's mark as an integer.
19395 You can see the documentation for these functions by placing point
19396 over them in this buffer and then typing @kbd{C-h f}
19397 (@code{describe-function}) followed by a @kbd{RET}; the default is the
19400 The last part of the function tells what to do when it succeeds.
19403 @appendixsec @code{yank-pop}
19406 After understanding @code{yank} and @code{current-kill}, you know how
19407 to approach the @code{yank-pop} function. Leaving out the
19408 documentation to save space, it looks like this:
19413 (defun yank-pop (&optional arg)
19416 (if (not (eq last-command 'yank))
19417 (error "Previous command was not a yank"))
19420 (setq this-command 'yank)
19421 (unless arg (setq arg 1))
19422 (let ((inhibit-read-only t)
19423 (before (< (point) (mark t))))
19427 (funcall (or yank-undo-function 'delete-region) (point) (mark t))
19428 (funcall (or yank-undo-function 'delete-region) (mark t) (point)))
19429 (setq yank-undo-function nil)
19432 (set-marker (mark-marker) (point) (current-buffer))
19433 (insert-for-yank (current-kill arg))
19434 ;; Set the window start back where it was in the yank command,
19436 (set-window-start (selected-window) yank-window-start t)
19440 ;; This is like exchange-point-and-mark,
19441 ;; but doesn't activate the mark.
19442 ;; It is cleaner to avoid activation, even though the command
19443 ;; loop would deactivate the mark because we inserted text.
19444 (goto-char (prog1 (mark t)
19445 (set-marker (mark-marker)
19447 (current-buffer))))))
19452 The function is interactive with a small @samp{p} so the prefix
19453 argument is processed and passed to the function. The command can
19454 only be used after a previous yank; otherwise an error message is
19455 sent. This check uses the variable @code{last-command} which is set
19456 by @code{yank} and is discussed elsewhere.
19457 (@xref{copy-region-as-kill}.)
19459 The @code{let} clause sets the variable @code{before} to true or false
19460 depending whether point is before or after mark and then the region
19461 between point and mark is deleted. This is the region that was just
19462 inserted by the previous yank and it is this text that will be
19465 @code{funcall} calls its first argument as a function, passing
19466 remaining arguments to it. The first argument is whatever the
19467 @code{or} expression returns. The two remaining arguments are the
19468 positions of point and mark set by the preceding @code{yank} command.
19470 There is more, but that is the hardest part.
19473 @appendixsec The @file{ring.el} File
19474 @cindex @file{ring.el} file
19476 Interestingly, GNU Emacs posses a file called @file{ring.el} that
19477 provides many of the features we just discussed. But functions such
19478 as @code{kill-ring-yank-pointer} do not use this library, possibly
19479 because they were written earlier.
19482 @appendix A Graph with Labeled Axes
19484 Printed axes help you understand a graph. They convey scale. In an
19485 earlier chapter (@pxref{Readying a Graph, , Readying a Graph}), we
19486 wrote the code to print the body of a graph. Here we write the code
19487 for printing and labeling vertical and horizontal axes, along with the
19491 * Labeled Example::
19492 * print-graph Varlist:: @code{let} expression in @code{print-graph}.
19493 * print-Y-axis:: Print a label for the vertical axis.
19494 * print-X-axis:: Print a horizontal label.
19495 * Print Whole Graph:: The function to print a complete graph.
19499 @node Labeled Example
19500 @unnumberedsec Labeled Example Graph
19503 Since insertions fill a buffer to the right and below point, the new
19504 graph printing function should first print the Y or vertical axis,
19505 then the body of the graph, and finally the X or horizontal axis.
19506 This sequence lays out for us the contents of the function:
19516 Print body of graph.
19523 Here is an example of how a finished graph should look:
19536 1 - ****************
19543 In this graph, both the vertical and the horizontal axes are labeled
19544 with numbers. However, in some graphs, the horizontal axis is time
19545 and would be better labeled with months, like this:
19559 Indeed, with a little thought, we can easily come up with a variety of
19560 vertical and horizontal labeling schemes. Our task could become
19561 complicated. But complications breed confusion. Rather than permit
19562 this, it is better choose a simple labeling scheme for our first
19563 effort, and to modify or replace it later.
19566 These considerations suggest the following outline for the
19567 @code{print-graph} function:
19571 (defun print-graph (numbers-list)
19572 "@var{documentation}@dots{}"
19573 (let ((height @dots{}
19577 (print-Y-axis height @dots{} )
19578 (graph-body-print numbers-list)
19579 (print-X-axis @dots{} )))
19583 We can work on each part of the @code{print-graph} function definition
19586 @node print-graph Varlist
19587 @appendixsec The @code{print-graph} Varlist
19588 @cindex @code{print-graph} varlist
19590 In writing the @code{print-graph} function, the first task is to write
19591 the varlist in the @code{let} expression. (We will leave aside for the
19592 moment any thoughts about making the function interactive or about the
19593 contents of its documentation string.)
19595 The varlist should set several values. Clearly, the top of the label
19596 for the vertical axis must be at least the height of the graph, which
19597 means that we must obtain this information here. Note that the
19598 @code{print-graph-body} function also requires this information. There
19599 is no reason to calculate the height of the graph in two different
19600 places, so we should change @code{print-graph-body} from the way we
19601 defined it earlier to take advantage of the calculation.
19603 Similarly, both the function for printing the X axis labels and the
19604 @code{print-graph-body} function need to learn the value of the width of
19605 each symbol. We can perform the calculation here and change the
19606 definition for @code{print-graph-body} from the way we defined it in the
19609 The length of the label for the horizontal axis must be at least as long
19610 as the graph. However, this information is used only in the function
19611 that prints the horizontal axis, so it does not need to be calculated here.
19613 These thoughts lead us directly to the following form for the varlist
19614 in the @code{let} for @code{print-graph}:
19618 (let ((height (apply 'max numbers-list)) ; @r{First version.}
19619 (symbol-width (length graph-blank)))
19624 As we shall see, this expression is not quite right.
19628 @appendixsec The @code{print-Y-axis} Function
19629 @cindex Axis, print vertical
19630 @cindex Y axis printing
19631 @cindex Vertical axis printing
19632 @cindex Print vertical axis
19634 The job of the @code{print-Y-axis} function is to print a label for
19635 the vertical axis that looks like this:
19653 The function should be passed the height of the graph, and then should
19654 construct and insert the appropriate numbers and marks.
19657 * print-Y-axis in Detail::
19658 * Height of label:: What height for the Y axis?
19659 * Compute a Remainder:: How to compute the remainder of a division.
19660 * Y Axis Element:: Construct a line for the Y axis.
19661 * Y-axis-column:: Generate a list of Y axis labels.
19662 * print-Y-axis Penultimate:: A not quite final version.
19666 @node print-Y-axis in Detail
19667 @unnumberedsubsec The @code{print-Y-axis} Function in Detail
19670 It is easy enough to see in the figure what the Y axis label should
19671 look like; but to say in words, and then to write a function
19672 definition to do the job is another matter. It is not quite true to
19673 say that we want a number and a tic every five lines: there are only
19674 three lines between the @samp{1} and the @samp{5} (lines 2, 3, and 4),
19675 but four lines between the @samp{5} and the @samp{10} (lines 6, 7, 8,
19676 and 9). It is better to say that we want a number and a tic mark on
19677 the base line (number 1) and then that we want a number and a tic on
19678 the fifth line from the bottom and on every line that is a multiple of
19682 @node Height of label
19683 @unnumberedsubsec What height should the label be?
19686 The next issue is what height the label should be? Suppose the maximum
19687 height of tallest column of the graph is seven. Should the highest
19688 label on the Y axis be @samp{5 -}, and should the graph stick up above
19689 the label? Or should the highest label be @samp{7 -}, and mark the peak
19690 of the graph? Or should the highest label be @code{10 -}, which is a
19691 multiple of five, and be higher than the topmost value of the graph?
19693 The latter form is preferred. Most graphs are drawn within rectangles
19694 whose sides are an integral number of steps long---5, 10, 15, and so
19695 on for a step distance of five. But as soon as we decide to use a
19696 step height for the vertical axis, we discover that the simple
19697 expression in the varlist for computing the height is wrong. The
19698 expression is @code{(apply 'max numbers-list)}. This returns the
19699 precise height, not the maximum height plus whatever is necessary to
19700 round up to the nearest multiple of five. A more complex expression
19703 As usual in cases like this, a complex problem becomes simpler if it is
19704 divided into several smaller problems.
19706 First, consider the case when the highest value of the graph is an
19707 integral multiple of five---when it is 5, 10, 15, or some higher
19708 multiple of five. We can use this value as the Y axis height.
19710 A fairly simply way to determine whether a number is a multiple of
19711 five is to divide it by five and see if the division results in a
19712 remainder. If there is no remainder, the number is a multiple of
19713 five. Thus, seven divided by five has a remainder of two, and seven
19714 is not an integral multiple of five. Put in slightly different
19715 language, more reminiscent of the classroom, five goes into seven
19716 once, with a remainder of two. However, five goes into ten twice,
19717 with no remainder: ten is an integral multiple of five.
19719 @node Compute a Remainder
19720 @appendixsubsec Side Trip: Compute a Remainder
19722 @findex % @r{(remainder function)}
19723 @cindex Remainder function, @code{%}
19724 In Lisp, the function for computing a remainder is @code{%}. The
19725 function returns the remainder of its first argument divided by its
19726 second argument. As it happens, @code{%} is a function in Emacs Lisp
19727 that you cannot discover using @code{apropos}: you find nothing if you
19728 type @kbd{M-x apropos @key{RET} remainder @key{RET}}. The only way to
19729 learn of the existence of @code{%} is to read about it in a book such
19730 as this or in the Emacs Lisp sources.
19732 You can try the @code{%} function by evaluating the following two
19744 The first expression returns 2 and the second expression returns 0.
19746 To test whether the returned value is zero or some other number, we
19747 can use the @code{zerop} function. This function returns @code{t} if
19748 its argument, which must be a number, is zero.
19760 Thus, the following expression will return @code{t} if the height
19761 of the graph is evenly divisible by five:
19764 (zerop (% height 5))
19768 (The value of @code{height}, of course, can be found from @code{(apply
19769 'max numbers-list)}.)
19771 On the other hand, if the value of @code{height} is not a multiple of
19772 five, we want to reset the value to the next higher multiple of five.
19773 This is straightforward arithmetic using functions with which we are
19774 already familiar. First, we divide the value of @code{height} by five
19775 to determine how many times five goes into the number. Thus, five
19776 goes into twelve twice. If we add one to this quotient and multiply by
19777 five, we will obtain the value of the next multiple of five that is
19778 larger than the height. Five goes into twelve twice. Add one to two,
19779 and multiply by five; the result is fifteen, which is the next multiple
19780 of five that is higher than twelve. The Lisp expression for this is:
19783 (* (1+ (/ height 5)) 5)
19787 For example, if you evaluate the following, the result is 15:
19790 (* (1+ (/ 12 5)) 5)
19793 All through this discussion, we have been using `five' as the value
19794 for spacing labels on the Y axis; but we may want to use some other
19795 value. For generality, we should replace `five' with a variable to
19796 which we can assign a value. The best name I can think of for this
19797 variable is @code{Y-axis-label-spacing}.
19800 Using this term, and an @code{if} expression, we produce the
19805 (if (zerop (% height Y-axis-label-spacing))
19808 (* (1+ (/ height Y-axis-label-spacing))
19809 Y-axis-label-spacing))
19814 This expression returns the value of @code{height} itself if the height
19815 is an even multiple of the value of the @code{Y-axis-label-spacing} or
19816 else it computes and returns a value of @code{height} that is equal to
19817 the next higher multiple of the value of the @code{Y-axis-label-spacing}.
19819 We can now include this expression in the @code{let} expression of the
19820 @code{print-graph} function (after first setting the value of
19821 @code{Y-axis-label-spacing}):
19822 @vindex Y-axis-label-spacing
19826 (defvar Y-axis-label-spacing 5
19827 "Number of lines from one Y axis label to next.")
19832 (let* ((height (apply 'max numbers-list))
19833 (height-of-top-line
19834 (if (zerop (% height Y-axis-label-spacing))
19839 (* (1+ (/ height Y-axis-label-spacing))
19840 Y-axis-label-spacing)))
19841 (symbol-width (length graph-blank))))
19847 (Note use of the @code{let*} function: the initial value of height is
19848 computed once by the @code{(apply 'max numbers-list)} expression and
19849 then the resulting value of @code{height} is used to compute its
19850 final value. @xref{fwd-para let, , The @code{let*} expression}, for
19851 more about @code{let*}.)
19853 @node Y Axis Element
19854 @appendixsubsec Construct a Y Axis Element
19856 When we print the vertical axis, we want to insert strings such as
19857 @w{@samp{5 -}} and @w{@samp{10 - }} every five lines.
19858 Moreover, we want the numbers and dashes to line up, so shorter
19859 numbers must be padded with leading spaces. If some of the strings
19860 use two digit numbers, the strings with single digit numbers must
19861 include a leading blank space before the number.
19863 @findex number-to-string
19864 To figure out the length of the number, the @code{length} function is
19865 used. But the @code{length} function works only with a string, not with
19866 a number. So the number has to be converted from being a number to
19867 being a string. This is done with the @code{number-to-string} function.
19872 (length (number-to-string 35))
19875 (length (number-to-string 100))
19881 (@code{number-to-string} is also called @code{int-to-string}; you will
19882 see this alternative name in various sources.)
19884 In addition, in each label, each number is followed by a string such
19885 as @w{@samp{ - }}, which we will call the @code{Y-axis-tic} marker.
19886 This variable is defined with @code{defvar}:
19891 (defvar Y-axis-tic " - "
19892 "String that follows number in a Y axis label.")
19896 The length of the Y label is the sum of the length of the Y axis tic
19897 mark and the length of the number of the top of the graph.
19900 (length (concat (number-to-string height) Y-axis-tic)))
19903 This value will be calculated by the @code{print-graph} function in
19904 its varlist as @code{full-Y-label-width} and passed on. (Note that we
19905 did not think to include this in the varlist when we first proposed it.)
19907 To make a complete vertical axis label, a tic mark is concatenated
19908 with a number; and the two together may be preceded by one or more
19909 spaces depending on how long the number is. The label consists of
19910 three parts: the (optional) leading spaces, the number, and the tic
19911 mark. The function is passed the value of the number for the specific
19912 row, and the value of the width of the top line, which is calculated
19913 (just once) by @code{print-graph}.
19917 (defun Y-axis-element (number full-Y-label-width)
19918 "Construct a NUMBERed label element.
19919 A numbered element looks like this ` 5 - ',
19920 and is padded as needed so all line up with
19921 the element for the largest number."
19924 (let* ((leading-spaces
19925 (- full-Y-label-width
19927 (concat (number-to-string number)
19932 (make-string leading-spaces ? )
19933 (number-to-string number)
19938 The @code{Y-axis-element} function concatenates together the leading
19939 spaces, if any; the number, as a string; and the tic mark.
19941 To figure out how many leading spaces the label will need, the
19942 function subtracts the actual length of the label---the length of the
19943 number plus the length of the tic mark---from the desired label width.
19945 @findex make-string
19946 Blank spaces are inserted using the @code{make-string} function. This
19947 function takes two arguments: the first tells it how long the string
19948 will be and the second is a symbol for the character to insert, in a
19949 special format. The format is a question mark followed by a blank
19950 space, like this, @samp{? }. @xref{Character Type, , Character Type,
19951 elisp, The GNU Emacs Lisp Reference Manual}, for a description of the
19952 syntax for characters. (Of course, you might want to replace the
19953 blank space by some other character @dots{} You know what to do.)
19955 The @code{number-to-string} function is used in the concatenation
19956 expression, to convert the number to a string that is concatenated
19957 with the leading spaces and the tic mark.
19959 @node Y-axis-column
19960 @appendixsubsec Create a Y Axis Column
19962 The preceding functions provide all the tools needed to construct a
19963 function that generates a list of numbered and blank strings to insert
19964 as the label for the vertical axis:
19966 @findex Y-axis-column
19969 (defun Y-axis-column (height width-of-label)
19970 "Construct list of Y axis labels and blank strings.
19971 For HEIGHT of line above base and WIDTH-OF-LABEL."
19975 (while (> height 1)
19976 (if (zerop (% height Y-axis-label-spacing))
19977 ;; @r{Insert label.}
19980 (Y-axis-element height width-of-label)
19984 ;; @r{Else, insert blanks.}
19987 (make-string width-of-label ? )
19989 (setq height (1- height)))
19990 ;; @r{Insert base line.}
19992 (cons (Y-axis-element 1 width-of-label) Y-axis))
19993 (nreverse Y-axis)))
19997 In this function, we start with the value of @code{height} and
19998 repetitively subtract one from its value. After each subtraction, we
19999 test to see whether the value is an integral multiple of the
20000 @code{Y-axis-label-spacing}. If it is, we construct a numbered label
20001 using the @code{Y-axis-element} function; if not, we construct a
20002 blank label using the @code{make-string} function. The base line
20003 consists of the number one followed by a tic mark.
20006 @node print-Y-axis Penultimate
20007 @appendixsubsec The Not Quite Final Version of @code{print-Y-axis}
20009 The list constructed by the @code{Y-axis-column} function is passed to
20010 the @code{print-Y-axis} function, which inserts the list as a column.
20012 @findex print-Y-axis
20015 (defun print-Y-axis (height full-Y-label-width)
20016 "Insert Y axis using HEIGHT and FULL-Y-LABEL-WIDTH.
20017 Height must be the maximum height of the graph.
20018 Full width is the width of the highest label element."
20019 ;; Value of height and full-Y-label-width
20020 ;; are passed by `print-graph'.
20023 (let ((start (point)))
20025 (Y-axis-column height full-Y-label-width))
20026 ;; @r{Place point ready for inserting graph.}
20028 ;; @r{Move point forward by value of} full-Y-label-width
20029 (forward-char full-Y-label-width)))
20033 The @code{print-Y-axis} uses the @code{insert-rectangle} function to
20034 insert the Y axis labels created by the @code{Y-axis-column} function.
20035 In addition, it places point at the correct position for printing the body of
20038 You can test @code{print-Y-axis}:
20046 Y-axis-label-spacing
20055 Copy the following expression:
20058 (print-Y-axis 12 5)
20062 Switch to the @file{*scratch*} buffer and place the cursor where you
20063 want the axis labels to start.
20066 Type @kbd{M-:} (@code{eval-expression}).
20069 Yank the @code{graph-body-print} expression into the minibuffer
20070 with @kbd{C-y} (@code{yank)}.
20073 Press @key{RET} to evaluate the expression.
20076 Emacs will print labels vertically, the top one being @w{@samp{10 -@w{
20077 }}}. (The @code{print-graph} function will pass the value of
20078 @code{height-of-top-line}, which in this case will end up as 15,
20079 thereby getting rid of what might appear as a bug.)
20083 @appendixsec The @code{print-X-axis} Function
20084 @cindex Axis, print horizontal
20085 @cindex X axis printing
20086 @cindex Print horizontal axis
20087 @cindex Horizontal axis printing
20089 X axis labels are much like Y axis labels, except that the ticks are on a
20090 line above the numbers. Labels should look like this:
20099 The first tic is under the first column of the graph and is preceded by
20100 several blank spaces. These spaces provide room in rows above for the Y
20101 axis labels. The second, third, fourth, and subsequent ticks are all
20102 spaced equally, according to the value of @code{X-axis-label-spacing}.
20104 The second row of the X axis consists of numbers, preceded by several
20105 blank spaces and also separated according to the value of the variable
20106 @code{X-axis-label-spacing}.
20108 The value of the variable @code{X-axis-label-spacing} should itself be
20109 measured in units of @code{symbol-width}, since you may want to change
20110 the width of the symbols that you are using to print the body of the
20111 graph without changing the ways the graph is labeled.
20114 * Similarities differences:: Much like @code{print-Y-axis}, but not exactly.
20115 * X Axis Tic Marks:: Create tic marks for the horizontal axis.
20119 @node Similarities differences
20120 @unnumberedsubsec Similarities and differences
20123 The @code{print-X-axis} function is constructed in more or less the
20124 same fashion as the @code{print-Y-axis} function except that it has
20125 two lines: the line of tic marks and the numbers. We will write a
20126 separate function to print each line and then combine them within the
20127 @code{print-X-axis} function.
20129 This is a three step process:
20133 Write a function to print the X axis tic marks, @code{print-X-axis-tic-line}.
20136 Write a function to print the X numbers, @code{print-X-axis-numbered-line}.
20139 Write a function to print both lines, the @code{print-X-axis} function,
20140 using @code{print-X-axis-tic-line} and
20141 @code{print-X-axis-numbered-line}.
20144 @node X Axis Tic Marks
20145 @appendixsubsec X Axis Tic Marks
20147 The first function should print the X axis tic marks. We must specify
20148 the tic marks themselves and their spacing:
20152 (defvar X-axis-label-spacing
20153 (if (boundp 'graph-blank)
20154 (* 5 (length graph-blank)) 5)
20155 "Number of units from one X axis label to next.")
20160 (Note that the value of @code{graph-blank} is set by another
20161 @code{defvar}. The @code{boundp} predicate checks whether it has
20162 already been set; @code{boundp} returns @code{nil} if it has not. If
20163 @code{graph-blank} were unbound and we did not use this conditional
20164 construction, in a recent GNU Emacs, we would enter the debugger and
20165 see an error message saying @samp{@w{Debugger entered--Lisp error:}
20166 @w{(void-variable graph-blank)}}.)
20169 Here is the @code{defvar} for @code{X-axis-tic-symbol}:
20173 (defvar X-axis-tic-symbol "|"
20174 "String to insert to point to a column in X axis.")
20179 The goal is to make a line that looks like this:
20185 The first tic is indented so that it is under the first column, which is
20186 indented to provide space for the Y axis labels.
20188 A tic element consists of the blank spaces that stretch from one tic to
20189 the next plus a tic symbol. The number of blanks is determined by the
20190 width of the tic symbol and the @code{X-axis-label-spacing}.
20193 The code looks like this:
20197 ;;; X-axis-tic-element
20201 ;; @r{Make a string of blanks.}
20202 (- (* symbol-width X-axis-label-spacing)
20203 (length X-axis-tic-symbol))
20205 ;; @r{Concatenate blanks with tic symbol.}
20211 Next, we determine how many blanks are needed to indent the first tic
20212 mark to the first column of the graph. This uses the value of
20213 @code{full-Y-label-width} passed it by the @code{print-graph} function.
20216 The code to make @code{X-axis-leading-spaces}
20221 ;; X-axis-leading-spaces
20223 (make-string full-Y-label-width ? )
20228 We also need to determine the length of the horizontal axis, which is
20229 the length of the numbers list, and the number of ticks in the horizontal
20236 (length numbers-list)
20242 (* symbol-width X-axis-label-spacing)
20246 ;; number-of-X-ticks
20247 (if (zerop (% (X-length tic-width)))
20248 (/ (X-length tic-width))
20249 (1+ (/ (X-length tic-width))))
20254 All this leads us directly to the function for printing the X axis tic line:
20256 @findex print-X-axis-tic-line
20259 (defun print-X-axis-tic-line
20260 (number-of-X-tics X-axis-leading-spaces X-axis-tic-element)
20261 "Print ticks for X axis."
20262 (insert X-axis-leading-spaces)
20263 (insert X-axis-tic-symbol) ; @r{Under first column.}
20266 ;; @r{Insert second tic in the right spot.}
20269 (- (* symbol-width X-axis-label-spacing)
20270 ;; @r{Insert white space up to second tic symbol.}
20271 (* 2 (length X-axis-tic-symbol)))
20273 X-axis-tic-symbol))
20276 ;; @r{Insert remaining ticks.}
20277 (while (> number-of-X-tics 1)
20278 (insert X-axis-tic-element)
20279 (setq number-of-X-tics (1- number-of-X-tics))))
20283 The line of numbers is equally straightforward:
20286 First, we create a numbered element with blank spaces before each number:
20288 @findex X-axis-element
20291 (defun X-axis-element (number)
20292 "Construct a numbered X axis element."
20293 (let ((leading-spaces
20294 (- (* symbol-width X-axis-label-spacing)
20295 (length (number-to-string number)))))
20296 (concat (make-string leading-spaces ? )
20297 (number-to-string number))))
20301 Next, we create the function to print the numbered line, starting with
20302 the number ``1'' under the first column:
20304 @findex print-X-axis-numbered-line
20307 (defun print-X-axis-numbered-line
20308 (number-of-X-tics X-axis-leading-spaces)
20309 "Print line of X-axis numbers"
20310 (let ((number X-axis-label-spacing))
20311 (insert X-axis-leading-spaces)
20317 ;; @r{Insert white space up to next number.}
20318 (- (* symbol-width X-axis-label-spacing) 2)
20320 (number-to-string number)))
20323 ;; @r{Insert remaining numbers.}
20324 (setq number (+ number X-axis-label-spacing))
20325 (while (> number-of-X-tics 1)
20326 (insert (X-axis-element number))
20327 (setq number (+ number X-axis-label-spacing))
20328 (setq number-of-X-tics (1- number-of-X-tics)))))
20332 Finally, we need to write the @code{print-X-axis} that uses
20333 @code{print-X-axis-tic-line} and
20334 @code{print-X-axis-numbered-line}.
20336 The function must determine the local values of the variables used by both
20337 @code{print-X-axis-tic-line} and @code{print-X-axis-numbered-line}, and
20338 then it must call them. Also, it must print the carriage return that
20339 separates the two lines.
20341 The function consists of a varlist that specifies five local variables,
20342 and calls to each of the two line printing functions:
20344 @findex print-X-axis
20347 (defun print-X-axis (numbers-list)
20348 "Print X axis labels to length of NUMBERS-LIST."
20349 (let* ((leading-spaces
20350 (make-string full-Y-label-width ? ))
20353 ;; symbol-width @r{is provided by} graph-body-print
20354 (tic-width (* symbol-width X-axis-label-spacing))
20355 (X-length (length numbers-list))
20363 ;; @r{Make a string of blanks.}
20364 (- (* symbol-width X-axis-label-spacing)
20365 (length X-axis-tic-symbol))
20369 ;; @r{Concatenate blanks with tic symbol.}
20370 X-axis-tic-symbol))
20374 (if (zerop (% X-length tic-width))
20375 (/ X-length tic-width)
20376 (1+ (/ X-length tic-width)))))
20379 (print-X-axis-tic-line tic-number leading-spaces X-tic)
20381 (print-X-axis-numbered-line tic-number leading-spaces)))
20386 You can test @code{print-X-axis}:
20390 Install @code{X-axis-tic-symbol}, @code{X-axis-label-spacing},
20391 @code{print-X-axis-tic-line}, as well as @code{X-axis-element},
20392 @code{print-X-axis-numbered-line}, and @code{print-X-axis}.
20395 Copy the following expression:
20400 (let ((full-Y-label-width 5)
20403 '(1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16))))
20408 Switch to the @file{*scratch*} buffer and place the cursor where you
20409 want the axis labels to start.
20412 Type @kbd{M-:} (@code{eval-expression}).
20415 Yank the test expression into the minibuffer
20416 with @kbd{C-y} (@code{yank)}.
20419 Press @key{RET} to evaluate the expression.
20423 Emacs will print the horizontal axis like this:
20433 @node Print Whole Graph
20434 @appendixsec Printing the Whole Graph
20435 @cindex Printing the whole graph
20436 @cindex Whole graph printing
20437 @cindex Graph, printing all
20439 Now we are nearly ready to print the whole graph.
20441 The function to print the graph with the proper labels follows the
20442 outline we created earlier (@pxref{Full Graph, , A Graph with Labeled
20443 Axes}), but with additions.
20446 Here is the outline:
20450 (defun print-graph (numbers-list)
20451 "@var{documentation}@dots{}"
20452 (let ((height @dots{}
20456 (print-Y-axis height @dots{} )
20457 (graph-body-print numbers-list)
20458 (print-X-axis @dots{} )))
20463 * The final version:: A few changes.
20464 * Test print-graph:: Run a short test.
20465 * Graphing words in defuns:: Executing the final code.
20466 * lambda:: How to write an anonymous function.
20467 * mapcar:: Apply a function to elements of a list.
20468 * Another Bug:: Yet another bug @dots{} most insidious.
20469 * Final printed graph:: The graph itself!
20473 @node The final version
20474 @unnumberedsubsec Changes for the Final Version
20477 The final version is different from what we planned in two ways:
20478 first, it contains additional values calculated once in the varlist;
20479 second, it carries an option to specify the labels' increment per row.
20480 This latter feature turns out to be essential; otherwise, a graph may
20481 have more rows than fit on a display or on a sheet of paper.
20484 This new feature requires a change to the @code{Y-axis-column}
20485 function, to add @code{vertical-step} to it. The function looks like
20488 @findex Y-axis-column @r{Final version.}
20491 ;;; @r{Final version.}
20492 (defun Y-axis-column
20493 (height width-of-label &optional vertical-step)
20494 "Construct list of labels for Y axis.
20495 HEIGHT is maximum height of graph.
20496 WIDTH-OF-LABEL is maximum width of label.
20497 VERTICAL-STEP, an option, is a positive integer
20498 that specifies how much a Y axis label increments
20499 for each line. For example, a step of 5 means
20500 that each line is five units of the graph."
20504 (number-per-line (or vertical-step 1)))
20505 (while (> height 1)
20506 (if (zerop (% height Y-axis-label-spacing))
20509 ;; @r{Insert label.}
20513 (* height number-per-line)
20518 ;; @r{Else, insert blanks.}
20521 (make-string width-of-label ? )
20523 (setq height (1- height)))
20526 ;; @r{Insert base line.}
20527 (setq Y-axis (cons (Y-axis-element
20528 (or vertical-step 1)
20531 (nreverse Y-axis)))
20535 The values for the maximum height of graph and the width of a symbol
20536 are computed by @code{print-graph} in its @code{let} expression; so
20537 @code{graph-body-print} must be changed to accept them.
20539 @findex graph-body-print @r{Final version.}
20542 ;;; @r{Final version.}
20543 (defun graph-body-print (numbers-list height symbol-width)
20544 "Print a bar graph of the NUMBERS-LIST.
20545 The numbers-list consists of the Y-axis values.
20546 HEIGHT is maximum height of graph.
20547 SYMBOL-WIDTH is number of each column."
20550 (let (from-position)
20551 (while numbers-list
20552 (setq from-position (point))
20554 (column-of-graph height (car numbers-list)))
20555 (goto-char from-position)
20556 (forward-char symbol-width)
20559 ;; @r{Draw graph column by column.}
20561 (setq numbers-list (cdr numbers-list)))
20562 ;; @r{Place point for X axis labels.}
20563 (forward-line height)
20569 Finally, the code for the @code{print-graph} function:
20571 @findex print-graph @r{Final version.}
20574 ;;; @r{Final version.}
20576 (numbers-list &optional vertical-step)
20577 "Print labeled bar graph of the NUMBERS-LIST.
20578 The numbers-list consists of the Y-axis values.
20582 Optionally, VERTICAL-STEP, a positive integer,
20583 specifies how much a Y axis label increments for
20584 each line. For example, a step of 5 means that
20585 each row is five units."
20588 (let* ((symbol-width (length graph-blank))
20589 ;; @code{height} @r{is both the largest number}
20590 ;; @r{and the number with the most digits.}
20591 (height (apply 'max numbers-list))
20594 (height-of-top-line
20595 (if (zerop (% height Y-axis-label-spacing))
20598 (* (1+ (/ height Y-axis-label-spacing))
20599 Y-axis-label-spacing)))
20602 (vertical-step (or vertical-step 1))
20603 (full-Y-label-width
20609 (* height-of-top-line vertical-step))
20615 height-of-top-line full-Y-label-width vertical-step)
20619 numbers-list height-of-top-line symbol-width)
20620 (print-X-axis numbers-list)))
20624 @node Test print-graph
20625 @appendixsubsec Testing @code{print-graph}
20628 We can test the @code{print-graph} function with a short list of numbers:
20632 Install the final versions of @code{Y-axis-column},
20633 @code{graph-body-print}, and @code{print-graph} (in addition to the
20637 Copy the following expression:
20640 (print-graph '(3 2 5 6 7 5 3 4 6 4 3 2 1))
20644 Switch to the @file{*scratch*} buffer and place the cursor where you
20645 want the axis labels to start.
20648 Type @kbd{M-:} (@code{eval-expression}).
20651 Yank the test expression into the minibuffer
20652 with @kbd{C-y} (@code{yank)}.
20655 Press @key{RET} to evaluate the expression.
20659 Emacs will print a graph that looks like this:
20680 On the other hand, if you pass @code{print-graph} a
20681 @code{vertical-step} value of 2, by evaluating this expression:
20684 (print-graph '(3 2 5 6 7 5 3 4 6 4 3 2 1) 2)
20689 The graph looks like this:
20710 (A question: is the `2' on the bottom of the vertical axis a bug or a
20711 feature? If you think it is a bug, and should be a `1' instead, (or
20712 even a `0'), you can modify the sources.)
20714 @node Graphing words in defuns
20715 @appendixsubsec Graphing Numbers of Words and Symbols
20717 Now for the graph for which all this code was written: a graph that
20718 shows how many function definitions contain fewer than 10 words and
20719 symbols, how many contain between 10 and 19 words and symbols, how
20720 many contain between 20 and 29 words and symbols, and so on.
20722 This is a multi-step process. First make sure you have loaded all the
20726 It is a good idea to reset the value of @code{top-of-ranges} in case
20727 you have set it to some different value. You can evaluate the
20732 (setq top-of-ranges
20735 110 120 130 140 150
20736 160 170 180 190 200
20737 210 220 230 240 250
20738 260 270 280 290 300)
20743 Next create a list of the number of words and symbols in each range.
20747 Evaluate the following:
20751 (setq list-for-graph
20754 (recursive-lengths-list-many-files
20755 (directory-files "/usr/local/emacs/lisp"
20763 On my old machine, this took about an hour. It looked though 303 Lisp
20764 files in my copy of Emacs version 19.23. After all that computing,
20765 the @code{list-for-graph} had this value:
20769 (537 1027 955 785 594 483 349 292 224 199 166 120 116 99
20770 90 80 67 48 52 45 41 33 28 26 25 20 12 28 11 13 220)
20775 This means that my copy of Emacs had 537 function definitions with
20776 fewer than 10 words or symbols in them, 1,027 function definitions
20777 with 10 to 19 words or symbols in them, 955 function definitions with
20778 20 to 29 words or symbols in them, and so on.
20780 Clearly, just by looking at this list we can see that most function
20781 definitions contain ten to thirty words and symbols.
20783 Now for printing. We do @emph{not} want to print a graph that is
20784 1,030 lines high @dots{} Instead, we should print a graph that is
20785 fewer than twenty-five lines high. A graph that height can be
20786 displayed on almost any monitor, and easily printed on a sheet of paper.
20788 This means that each value in @code{list-for-graph} must be reduced to
20789 one-fiftieth its present value.
20791 Here is a short function to do just that, using two functions we have
20792 not yet seen, @code{mapcar} and @code{lambda}.
20796 (defun one-fiftieth (full-range)
20797 "Return list, each number one-fiftieth of previous."
20798 (mapcar (lambda (arg) (/ arg 50)) full-range))
20803 @appendixsubsec A @code{lambda} Expression: Useful Anonymity
20804 @cindex Anonymous function
20807 @code{lambda} is the symbol for an anonymous function, a function
20808 without a name. Every time you use an anonymous function, you need to
20809 include its whole body.
20816 (lambda (arg) (/ arg 50))
20820 is a function definition that says `return the value resulting from
20821 dividing whatever is passed to me as @code{arg} by 50'.
20824 Earlier, for example, we had a function @code{multiply-by-seven}; it
20825 multiplied its argument by 7. This function is similar, except it
20826 divides its argument by 50; and, it has no name. The anonymous
20827 equivalent of @code{multiply-by-seven} is:
20830 (lambda (number) (* 7 number))
20834 (@xref{defun, , The @code{defun} Macro}.)
20838 If we want to multiply 3 by 7, we can write:
20840 @c clear print-postscript-figures
20841 @c lambda example diagram #1
20845 (multiply-by-seven 3)
20846 \_______________/ ^
20852 @ifset print-postscript-figures
20855 @center @image{lambda-1}
20859 @ifclear print-postscript-figures
20863 (multiply-by-seven 3)
20864 \_______________/ ^
20873 This expression returns 21.
20877 Similarly, we can write:
20879 @c lambda example diagram #2
20883 ((lambda (number) (* 7 number)) 3)
20884 \____________________________/ ^
20886 anonymous function argument
20890 @ifset print-postscript-figures
20893 @center @image{lambda-2}
20897 @ifclear print-postscript-figures
20901 ((lambda (number) (* 7 number)) 3)
20902 \____________________________/ ^
20904 anonymous function argument
20912 If we want to divide 100 by 50, we can write:
20914 @c lambda example diagram #3
20918 ((lambda (arg) (/ arg 50)) 100)
20919 \______________________/ \_/
20921 anonymous function argument
20925 @ifset print-postscript-figures
20928 @center @image{lambda-3}
20932 @ifclear print-postscript-figures
20936 ((lambda (arg) (/ arg 50)) 100)
20937 \______________________/ \_/
20939 anonymous function argument
20946 This expression returns 2. The 100 is passed to the function, which
20947 divides that number by 50.
20949 @xref{Lambda Expressions, , Lambda Expressions, elisp, The GNU Emacs
20950 Lisp Reference Manual}, for more about @code{lambda}. Lisp and lambda
20951 expressions derive from the Lambda Calculus.
20954 @appendixsubsec The @code{mapcar} Function
20957 @code{mapcar} is a function that calls its first argument with each
20958 element of its second argument, in turn. The second argument must be
20961 The @samp{map} part of the name comes from the mathematical phrase,
20962 `mapping over a domain', meaning to apply a function to each of the
20963 elements in a domain. The mathematical phrase is based on the
20964 metaphor of a surveyor walking, one step at a time, over an area he is
20965 mapping. And @samp{car}, of course, comes from the Lisp notion of the
20974 (mapcar '1+ '(2 4 6))
20980 The function @code{1+} which adds one to its argument, is executed on
20981 @emph{each} element of the list, and a new list is returned.
20983 Contrast this with @code{apply}, which applies its first argument to
20985 (@xref{Readying a Graph, , Readying a Graph}, for a explanation of
20989 In the definition of @code{one-fiftieth}, the first argument is the
20990 anonymous function:
20993 (lambda (arg) (/ arg 50))
20997 and the second argument is @code{full-range}, which will be bound to
20998 @code{list-for-graph}.
21001 The whole expression looks like this:
21004 (mapcar (lambda (arg) (/ arg 50)) full-range))
21007 @xref{Mapping Functions, , Mapping Functions, elisp, The GNU Emacs
21008 Lisp Reference Manual}, for more about @code{mapcar}.
21010 Using the @code{one-fiftieth} function, we can generate a list in
21011 which each element is one-fiftieth the size of the corresponding
21012 element in @code{list-for-graph}.
21016 (setq fiftieth-list-for-graph
21017 (one-fiftieth list-for-graph))
21022 The resulting list looks like this:
21026 (10 20 19 15 11 9 6 5 4 3 3 2 2
21027 1 1 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 4)
21032 This, we are almost ready to print! (We also notice the loss of
21033 information: many of the higher ranges are 0, meaning that fewer than
21034 50 defuns had that many words or symbols---but not necessarily meaning
21035 that none had that many words or symbols.)
21038 @appendixsubsec Another Bug @dots{} Most Insidious
21039 @cindex Bug, most insidious type
21040 @cindex Insidious type of bug
21042 I said `almost ready to print'! Of course, there is a bug in the
21043 @code{print-graph} function @dots{} It has a @code{vertical-step}
21044 option, but not a @code{horizontal-step} option. The
21045 @code{top-of-range} scale goes from 10 to 300 by tens. But the
21046 @code{print-graph} function will print only by ones.
21048 This is a classic example of what some consider the most insidious
21049 type of bug, the bug of omission. This is not the kind of bug you can
21050 find by studying the code, for it is not in the code; it is an omitted
21051 feature. Your best actions are to try your program early and often;
21052 and try to arrange, as much as you can, to write code that is easy to
21053 understand and easy to change. Try to be aware, whenever you can,
21054 that whatever you have written, @emph{will} be rewritten, if not soon,
21055 eventually. A hard maxim to follow.
21057 It is the @code{print-X-axis-numbered-line} function that needs the
21058 work; and then the @code{print-X-axis} and the @code{print-graph}
21059 functions need to be adapted. Not much needs to be done; there is one
21060 nicety: the numbers ought to line up under the tic marks. This takes
21064 Here is the corrected @code{print-X-axis-numbered-line}:
21068 (defun print-X-axis-numbered-line
21069 (number-of-X-tics X-axis-leading-spaces
21070 &optional horizontal-step)
21071 "Print line of X-axis numbers"
21072 (let ((number X-axis-label-spacing)
21073 (horizontal-step (or horizontal-step 1)))
21076 (insert X-axis-leading-spaces)
21077 ;; @r{Delete extra leading spaces.}
21080 (length (number-to-string horizontal-step)))))
21085 ;; @r{Insert white space.}
21087 X-axis-label-spacing)
21090 (number-to-string horizontal-step)))
21094 (* number horizontal-step))))
21097 ;; @r{Insert remaining numbers.}
21098 (setq number (+ number X-axis-label-spacing))
21099 (while (> number-of-X-tics 1)
21100 (insert (X-axis-element
21101 (* number horizontal-step)))
21102 (setq number (+ number X-axis-label-spacing))
21103 (setq number-of-X-tics (1- number-of-X-tics)))))
21108 If you are reading this in Info, you can see the new versions of
21109 @code{print-X-axis} @code{print-graph} and evaluate them. If you are
21110 reading this in a printed book, you can see the changed lines here
21111 (the full text is too much to print).
21116 (defun print-X-axis (numbers-list horizontal-step)
21118 (print-X-axis-numbered-line
21119 tic-number leading-spaces horizontal-step))
21127 &optional vertical-step horizontal-step)
21129 (print-X-axis numbers-list horizontal-step))
21137 (defun print-X-axis (numbers-list horizontal-step)
21138 "Print X axis labels to length of NUMBERS-LIST.
21139 Optionally, HORIZONTAL-STEP, a positive integer,
21140 specifies how much an X axis label increments for
21144 ;; Value of symbol-width and full-Y-label-width
21145 ;; are passed by `print-graph'.
21146 (let* ((leading-spaces
21147 (make-string full-Y-label-width ? ))
21148 ;; symbol-width @r{is provided by} graph-body-print
21149 (tic-width (* symbol-width X-axis-label-spacing))
21150 (X-length (length numbers-list))
21156 ;; @r{Make a string of blanks.}
21157 (- (* symbol-width X-axis-label-spacing)
21158 (length X-axis-tic-symbol))
21162 ;; @r{Concatenate blanks with tic symbol.}
21163 X-axis-tic-symbol))
21165 (if (zerop (% X-length tic-width))
21166 (/ X-length tic-width)
21167 (1+ (/ X-length tic-width)))))
21171 (print-X-axis-tic-line
21172 tic-number leading-spaces X-tic)
21174 (print-X-axis-numbered-line
21175 tic-number leading-spaces horizontal-step)))
21182 (numbers-list &optional vertical-step horizontal-step)
21183 "Print labeled bar graph of the NUMBERS-LIST.
21184 The numbers-list consists of the Y-axis values.
21188 Optionally, VERTICAL-STEP, a positive integer,
21189 specifies how much a Y axis label increments for
21190 each line. For example, a step of 5 means that
21191 each row is five units.
21195 Optionally, HORIZONTAL-STEP, a positive integer,
21196 specifies how much an X axis label increments for
21198 (let* ((symbol-width (length graph-blank))
21199 ;; @code{height} @r{is both the largest number}
21200 ;; @r{and the number with the most digits.}
21201 (height (apply 'max numbers-list))
21204 (height-of-top-line
21205 (if (zerop (% height Y-axis-label-spacing))
21208 (* (1+ (/ height Y-axis-label-spacing))
21209 Y-axis-label-spacing)))
21212 (vertical-step (or vertical-step 1))
21213 (full-Y-label-width
21217 (* height-of-top-line vertical-step))
21222 height-of-top-line full-Y-label-width vertical-step)
21224 numbers-list height-of-top-line symbol-width)
21225 (print-X-axis numbers-list horizontal-step)))
21232 Graphing Definitions Re-listed
21235 Here are all the graphing definitions in their final form:
21239 (defvar top-of-ranges
21242 110 120 130 140 150
21243 160 170 180 190 200
21244 210 220 230 240 250)
21245 "List specifying ranges for `defuns-per-range'.")
21249 (defvar graph-symbol "*"
21250 "String used as symbol in graph, usually an asterisk.")
21254 (defvar graph-blank " "
21255 "String used as blank in graph, usually a blank space.
21256 graph-blank must be the same number of columns wide
21261 (defvar Y-axis-tic " - "
21262 "String that follows number in a Y axis label.")
21266 (defvar Y-axis-label-spacing 5
21267 "Number of lines from one Y axis label to next.")
21271 (defvar X-axis-tic-symbol "|"
21272 "String to insert to point to a column in X axis.")
21276 (defvar X-axis-label-spacing
21277 (if (boundp 'graph-blank)
21278 (* 5 (length graph-blank)) 5)
21279 "Number of units from one X axis label to next.")
21285 (defun count-words-in-defun ()
21286 "Return the number of words and symbols in a defun."
21287 (beginning-of-defun)
21289 (end (save-excursion (end-of-defun) (point))))
21294 (and (< (point) end)
21296 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*"
21298 (setq count (1+ count)))
21305 (defun lengths-list-file (filename)
21306 "Return list of definitions' lengths within FILE.
21307 The returned list is a list of numbers.
21308 Each number is the number of words or
21309 symbols in one function definition."
21313 (message "Working on `%s' ... " filename)
21315 (let ((buffer (find-file-noselect filename))
21317 (set-buffer buffer)
21318 (setq buffer-read-only t)
21320 (goto-char (point-min))
21324 (while (re-search-forward "^(defun" nil t)
21326 (cons (count-words-in-defun) lengths-list)))
21327 (kill-buffer buffer)
21334 (defun lengths-list-many-files (list-of-files)
21335 "Return list of lengths of defuns in LIST-OF-FILES."
21336 (let (lengths-list)
21337 ;;; @r{true-or-false-test}
21338 (while list-of-files
21344 ;;; @r{Generate a lengths' list.}
21346 (expand-file-name (car list-of-files)))))
21347 ;;; @r{Make files' list shorter.}
21348 (setq list-of-files (cdr list-of-files)))
21349 ;;; @r{Return final value of lengths' list.}
21356 (defun defuns-per-range (sorted-lengths top-of-ranges)
21357 "SORTED-LENGTHS defuns in each TOP-OF-RANGES range."
21358 (let ((top-of-range (car top-of-ranges))
21359 (number-within-range 0)
21360 defuns-per-range-list)
21365 (while top-of-ranges
21369 ;; @r{Need number for numeric test.}
21370 (car sorted-lengths)
21371 (< (car sorted-lengths) top-of-range))
21373 ;; @r{Count number of definitions within current range.}
21374 (setq number-within-range (1+ number-within-range))
21375 (setq sorted-lengths (cdr sorted-lengths)))
21379 ;; @r{Exit inner loop but remain within outer loop.}
21381 (setq defuns-per-range-list
21382 (cons number-within-range defuns-per-range-list))
21383 (setq number-within-range 0) ; @r{Reset count to zero.}
21385 ;; @r{Move to next range.}
21386 (setq top-of-ranges (cdr top-of-ranges))
21387 ;; @r{Specify next top of range value.}
21388 (setq top-of-range (car top-of-ranges)))
21392 ;; @r{Exit outer loop and count the number of defuns larger than}
21393 ;; @r{ the largest top-of-range value.}
21394 (setq defuns-per-range-list
21396 (length sorted-lengths)
21397 defuns-per-range-list))
21399 ;; @r{Return a list of the number of definitions within each range,}
21400 ;; @r{ smallest to largest.}
21401 (nreverse defuns-per-range-list)))
21407 (defun column-of-graph (max-graph-height actual-height)
21408 "Return list of MAX-GRAPH-HEIGHT strings;
21409 ACTUAL-HEIGHT are graph-symbols.
21410 The graph-symbols are contiguous entries at the end
21412 The list will be inserted as one column of a graph.
21413 The strings are either graph-blank or graph-symbol."
21417 (let ((insert-list nil)
21418 (number-of-top-blanks
21419 (- max-graph-height actual-height)))
21421 ;; @r{Fill in @code{graph-symbols}.}
21422 (while (> actual-height 0)
21423 (setq insert-list (cons graph-symbol insert-list))
21424 (setq actual-height (1- actual-height)))
21428 ;; @r{Fill in @code{graph-blanks}.}
21429 (while (> number-of-top-blanks 0)
21430 (setq insert-list (cons graph-blank insert-list))
21431 (setq number-of-top-blanks
21432 (1- number-of-top-blanks)))
21434 ;; @r{Return whole list.}
21441 (defun Y-axis-element (number full-Y-label-width)
21442 "Construct a NUMBERed label element.
21443 A numbered element looks like this ` 5 - ',
21444 and is padded as needed so all line up with
21445 the element for the largest number."
21448 (let* ((leading-spaces
21449 (- full-Y-label-width
21451 (concat (number-to-string number)
21456 (make-string leading-spaces ? )
21457 (number-to-string number)
21464 (defun print-Y-axis
21465 (height full-Y-label-width &optional vertical-step)
21466 "Insert Y axis by HEIGHT and FULL-Y-LABEL-WIDTH.
21467 Height must be the maximum height of the graph.
21468 Full width is the width of the highest label element.
21469 Optionally, print according to VERTICAL-STEP."
21472 ;; Value of height and full-Y-label-width
21473 ;; are passed by `print-graph'.
21474 (let ((start (point)))
21476 (Y-axis-column height full-Y-label-width vertical-step))
21479 ;; @r{Place point ready for inserting graph.}
21481 ;; @r{Move point forward by value of} full-Y-label-width
21482 (forward-char full-Y-label-width)))
21488 (defun print-X-axis-tic-line
21489 (number-of-X-tics X-axis-leading-spaces X-axis-tic-element)
21490 "Print ticks for X axis."
21491 (insert X-axis-leading-spaces)
21492 (insert X-axis-tic-symbol) ; @r{Under first column.}
21495 ;; @r{Insert second tic in the right spot.}
21498 (- (* symbol-width X-axis-label-spacing)
21499 ;; @r{Insert white space up to second tic symbol.}
21500 (* 2 (length X-axis-tic-symbol)))
21502 X-axis-tic-symbol))
21505 ;; @r{Insert remaining ticks.}
21506 (while (> number-of-X-tics 1)
21507 (insert X-axis-tic-element)
21508 (setq number-of-X-tics (1- number-of-X-tics))))
21514 (defun X-axis-element (number)
21515 "Construct a numbered X axis element."
21516 (let ((leading-spaces
21517 (- (* symbol-width X-axis-label-spacing)
21518 (length (number-to-string number)))))
21519 (concat (make-string leading-spaces ? )
21520 (number-to-string number))))
21526 (defun graph-body-print (numbers-list height symbol-width)
21527 "Print a bar graph of the NUMBERS-LIST.
21528 The numbers-list consists of the Y-axis values.
21529 HEIGHT is maximum height of graph.
21530 SYMBOL-WIDTH is number of each column."
21533 (let (from-position)
21534 (while numbers-list
21535 (setq from-position (point))
21537 (column-of-graph height (car numbers-list)))
21538 (goto-char from-position)
21539 (forward-char symbol-width)
21542 ;; @r{Draw graph column by column.}
21544 (setq numbers-list (cdr numbers-list)))
21545 ;; @r{Place point for X axis labels.}
21546 (forward-line height)
21553 (defun Y-axis-column
21554 (height width-of-label &optional vertical-step)
21555 "Construct list of labels for Y axis.
21556 HEIGHT is maximum height of graph.
21557 WIDTH-OF-LABEL is maximum width of label.
21560 VERTICAL-STEP, an option, is a positive integer
21561 that specifies how much a Y axis label increments
21562 for each line. For example, a step of 5 means
21563 that each line is five units of the graph."
21565 (number-per-line (or vertical-step 1)))
21568 (while (> height 1)
21569 (if (zerop (% height Y-axis-label-spacing))
21570 ;; @r{Insert label.}
21574 (* height number-per-line)
21579 ;; @r{Else, insert blanks.}
21582 (make-string width-of-label ? )
21584 (setq height (1- height)))
21587 ;; @r{Insert base line.}
21588 (setq Y-axis (cons (Y-axis-element
21589 (or vertical-step 1)
21592 (nreverse Y-axis)))
21598 (defun print-X-axis-numbered-line
21599 (number-of-X-tics X-axis-leading-spaces
21600 &optional horizontal-step)
21601 "Print line of X-axis numbers"
21602 (let ((number X-axis-label-spacing)
21603 (horizontal-step (or horizontal-step 1)))
21606 (insert X-axis-leading-spaces)
21608 (delete-char (- (1- (length (number-to-string horizontal-step)))))
21611 ;; @r{Insert white space up to next number.}
21612 (- (* symbol-width X-axis-label-spacing)
21613 (1- (length (number-to-string horizontal-step)))
21616 (number-to-string (* number horizontal-step))))
21619 ;; @r{Insert remaining numbers.}
21620 (setq number (+ number X-axis-label-spacing))
21621 (while (> number-of-X-tics 1)
21622 (insert (X-axis-element (* number horizontal-step)))
21623 (setq number (+ number X-axis-label-spacing))
21624 (setq number-of-X-tics (1- number-of-X-tics)))))
21630 (defun print-X-axis (numbers-list horizontal-step)
21631 "Print X axis labels to length of NUMBERS-LIST.
21632 Optionally, HORIZONTAL-STEP, a positive integer,
21633 specifies how much an X axis label increments for
21637 ;; Value of symbol-width and full-Y-label-width
21638 ;; are passed by `print-graph'.
21639 (let* ((leading-spaces
21640 (make-string full-Y-label-width ? ))
21641 ;; symbol-width @r{is provided by} graph-body-print
21642 (tic-width (* symbol-width X-axis-label-spacing))
21643 (X-length (length numbers-list))
21649 ;; @r{Make a string of blanks.}
21650 (- (* symbol-width X-axis-label-spacing)
21651 (length X-axis-tic-symbol))
21655 ;; @r{Concatenate blanks with tic symbol.}
21656 X-axis-tic-symbol))
21658 (if (zerop (% X-length tic-width))
21659 (/ X-length tic-width)
21660 (1+ (/ X-length tic-width)))))
21664 (print-X-axis-tic-line
21665 tic-number leading-spaces X-tic)
21667 (print-X-axis-numbered-line
21668 tic-number leading-spaces horizontal-step)))
21674 (defun one-fiftieth (full-range)
21675 "Return list, each number of which is 1/50th previous."
21676 (mapcar (lambda (arg) (/ arg 50)) full-range))
21683 (numbers-list &optional vertical-step horizontal-step)
21684 "Print labeled bar graph of the NUMBERS-LIST.
21685 The numbers-list consists of the Y-axis values.
21689 Optionally, VERTICAL-STEP, a positive integer,
21690 specifies how much a Y axis label increments for
21691 each line. For example, a step of 5 means that
21692 each row is five units.
21696 Optionally, HORIZONTAL-STEP, a positive integer,
21697 specifies how much an X axis label increments for
21699 (let* ((symbol-width (length graph-blank))
21700 ;; @code{height} @r{is both the largest number}
21701 ;; @r{and the number with the most digits.}
21702 (height (apply 'max numbers-list))
21705 (height-of-top-line
21706 (if (zerop (% height Y-axis-label-spacing))
21709 (* (1+ (/ height Y-axis-label-spacing))
21710 Y-axis-label-spacing)))
21713 (vertical-step (or vertical-step 1))
21714 (full-Y-label-width
21718 (* height-of-top-line vertical-step))
21724 height-of-top-line full-Y-label-width vertical-step)
21726 numbers-list height-of-top-line symbol-width)
21727 (print-X-axis numbers-list horizontal-step)))
21734 @node Final printed graph
21735 @appendixsubsec The Printed Graph
21737 When made and installed, you can call the @code{print-graph} command
21743 (print-graph fiftieth-list-for-graph 50 10)
21773 50 - ***************** * *
21775 10 50 100 150 200 250 300 350
21782 The largest group of functions contain 10--19 words and symbols each.
21784 @node Free Software and Free Manuals
21785 @appendix Free Software and Free Manuals
21787 @strong{by Richard M. Stallman}
21790 The biggest deficiency in free operating systems is not in the
21791 software---it is the lack of good free manuals that we can include in
21792 these systems. Many of our most important programs do not come with
21793 full manuals. Documentation is an essential part of any software
21794 package; when an important free software package does not come with a
21795 free manual, that is a major gap. We have many such gaps today.
21797 Once upon a time, many years ago, I thought I would learn Perl. I got
21798 a copy of a free manual, but I found it hard to read. When I asked
21799 Perl users about alternatives, they told me that there were better
21800 introductory manuals---but those were not free.
21802 Why was this? The authors of the good manuals had written them for
21803 O'Reilly Associates, which published them with restrictive terms---no
21804 copying, no modification, source files not available---which exclude
21805 them from the free software community.
21807 That wasn't the first time this sort of thing has happened, and (to
21808 our community's great loss) it was far from the last. Proprietary
21809 manual publishers have enticed a great many authors to restrict their
21810 manuals since then. Many times I have heard a GNU user eagerly tell me
21811 about a manual that he is writing, with which he expects to help the
21812 GNU project---and then had my hopes dashed, as he proceeded to explain
21813 that he had signed a contract with a publisher that would restrict it
21814 so that we cannot use it.
21816 Given that writing good English is a rare skill among programmers, we
21817 can ill afford to lose manuals this way.
21819 Free documentation, like free software, is a matter of freedom, not
21820 price. The problem with these manuals was not that O'Reilly Associates
21821 charged a price for printed copies---that in itself is fine. The Free
21822 Software Foundation @uref{http://shop.fsf.org, sells printed copies} of
21823 free @uref{http://www.gnu.org/doc/doc.html, GNU manuals}, too.
21824 But GNU manuals are available in source code form, while these manuals
21825 are available only on paper. GNU manuals come with permission to copy
21826 and modify; the Perl manuals do not. These restrictions are the
21829 The criterion for a free manual is pretty much the same as for free
21830 software: it is a matter of giving all users certain
21831 freedoms. Redistribution (including commercial redistribution) must be
21832 permitted, so that the manual can accompany every copy of the program,
21833 on-line or on paper. Permission for modification is crucial too.
21835 As a general rule, I don't believe that it is essential for people to
21836 have permission to modify all sorts of articles and books. The issues
21837 for writings are not necessarily the same as those for software. For
21838 example, I don't think you or I are obliged to give permission to
21839 modify articles like this one, which describe our actions and our
21842 But there is a particular reason why the freedom to modify is crucial
21843 for documentation for free software. When people exercise their right
21844 to modify the software, and add or change its features, if they are
21845 conscientious they will change the manual too---so they can provide
21846 accurate and usable documentation with the modified program. A manual
21847 which forbids programmers to be conscientious and finish the job, or
21848 more precisely requires them to write a new manual from scratch if
21849 they change the program, does not fill our community's needs.
21851 While a blanket prohibition on modification is unacceptable, some
21852 kinds of limits on the method of modification pose no problem. For
21853 example, requirements to preserve the original author's copyright
21854 notice, the distribution terms, or the list of authors, are ok. It is
21855 also no problem to require modified versions to include notice that
21856 they were modified, even to have entire sections that may not be
21857 deleted or changed, as long as these sections deal with nontechnical
21858 topics. (Some GNU manuals have them.)
21860 These kinds of restrictions are not a problem because, as a practical
21861 matter, they don't stop the conscientious programmer from adapting the
21862 manual to fit the modified program. In other words, they don't block
21863 the free software community from making full use of the manual.
21865 However, it must be possible to modify all the technical content of
21866 the manual, and then distribute the result in all the usual media,
21867 through all the usual channels; otherwise, the restrictions do block
21868 the community, the manual is not free, and so we need another manual.
21870 Unfortunately, it is often hard to find someone to write another
21871 manual when a proprietary manual exists. The obstacle is that many
21872 users think that a proprietary manual is good enough---so they don't
21873 see the need to write a free manual. They do not see that the free
21874 operating system has a gap that needs filling.
21876 Why do users think that proprietary manuals are good enough? Some have
21877 not considered the issue. I hope this article will do something to
21880 Other users consider proprietary manuals acceptable for the same
21881 reason so many people consider proprietary software acceptable: they
21882 judge in purely practical terms, not using freedom as a
21883 criterion. These people are entitled to their opinions, but since
21884 those opinions spring from values which do not include freedom, they
21885 are no guide for those of us who do value freedom.
21887 Please spread the word about this issue. We continue to lose manuals
21888 to proprietary publishing. If we spread the word that proprietary
21889 manuals are not sufficient, perhaps the next person who wants to help
21890 GNU by writing documentation will realize, before it is too late, that
21891 he must above all make it free.
21893 We can also encourage commercial publishers to sell free, copylefted
21894 manuals instead of proprietary ones. One way you can help this is to
21895 check the distribution terms of a manual before you buy it, and prefer
21896 copylefted manuals to non-copylefted ones.
21900 Note: The Free Software Foundation maintains a page on its Web site
21901 that lists free books available from other publishers:@*
21902 @uref{http://www.gnu.org/doc/other-free-books.html}
21904 @node GNU Free Documentation License
21905 @appendix GNU Free Documentation License
21907 @cindex FDL, GNU Free Documentation License
21908 @include doclicense.texi
21914 MENU ENTRY: NODE NAME.
21920 @c Place biographical information on right-hand (verso) page
21923 \par\vfill\supereject
21925 \global\evenheadline={\hfil} \global\evenfootline={\hfil}
21926 \global\oddheadline={\hfil} \global\oddfootline={\hfil}
21929 % \par\vfill\supereject
21930 \global\evenheadline={\hfil} \global\evenfootline={\hfil}
21931 \global\oddheadline={\hfil} \global\oddfootline={\hfil}
21932 %\page\hbox{}%\page
21933 %\page\hbox{}%\page
21940 @c ================ Biographical information ================
21944 @center About the Author
21949 @node About the Author
21950 @unnumbered About the Author
21954 Robert J. Chassell has worked with GNU Emacs since 1985. He writes
21955 and edits, teaches Emacs and Emacs Lisp, and speaks throughout the
21956 world on software freedom. Chassell was a founding Director and
21957 Treasurer of the Free Software Foundation, Inc. He is co-author of
21958 the @cite{Texinfo} manual, and has edited more than a dozen other
21959 books. He graduated from Cambridge University, in England. He has an
21960 abiding interest in social and economic history and flies his own
21967 @c @c Prevent page number on blank verso, so eject it first.
21969 @c \par\vfill\supereject
21974 @c @evenheading @thispage @| @| @thistitle
21975 @c @oddheading @| @| @thispage