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
7 @documentencoding UTF-8
13 @include emacsver.texi
15 @c ================ How to Print a Book in Various Sizes ================
17 @c This book can be printed in any of three different sizes.
18 @c Set the following @-commands appropriately.
28 @c European A4 size paper:
33 @c (Note: if you edit the book so as to change the length of the
34 @c table of contents, you may have to change the value of `pageno' below.)
36 @c <<<< For hard copy printing, this file is now
37 @c set for smallbook, which works for all sizes
38 @c of paper, and with PostScript figures >>>>
46 @c ================ Included Figures ================
48 @c If you clear this, the figures will be printed as ASCII diagrams
49 @c rather than PostScript/PDF.
50 @c (This is not relevant to Info, since Info only handles ASCII.)
51 @set print-postscript-figures
52 @c clear print-postscript-figures
54 @comment %**end of header
56 @c per rms and peterb, use 10pt fonts for the main text, mostly to
57 @c save on paper cost.
58 @c Do this inside @tex for now, so current makeinfo does not complain.
64 \global\hbadness=6666 % don't worry about not-too-underfull boxes
67 @c These refer to the printed book sold by the FSF.
68 @set edition-number 3.10
69 @set update-date 28 October 2009
71 @c For next or subsequent edition:
72 @c create function using with-output-to-temp-buffer
73 @c create a major mode, with keymaps
74 @c run an asynchronous process, like grep or diff
76 @c For 8.5 by 11 inch format: do not use such a small amount of
77 @c whitespace between paragraphs as smallbook format
80 \global\parskip 6pt plus 1pt
84 @c For all sized formats: print within-book cross
85 @c reference with ``...'' rather than [...]
87 @c This works with the texinfo.tex file, version 2003-05-04.08,
88 @c in the Texinfo version 4.6 of the 2003 Jun 13 distribution.
91 \if \xrefprintnodename
92 \global\def\xrefprintnodename#1{\unskip, ``#1''}
94 \global\def\xrefprintnodename#1{ ``#1''}
96 % \global\def\xrefprintnodename#1{, ``#1''}
99 @c ----------------------------------------------------
101 @dircategory Emacs lisp
103 * Emacs Lisp Intro: (eintr). A simple introduction to Emacs Lisp programming.
107 This is an @cite{Introduction to Programming in Emacs Lisp}, for
108 people who are not programmers.
111 Edition @value{edition-number}, @value{update-date}
114 Distributed with Emacs version @value{EMACSVER}.
117 Copyright @copyright{} 1990--1995, 1997, 2001--2013 Free Software
124 GNU Press, @hfill @uref{http://www.fsf.org/licensing/gnu-press/}@*
125 a division of the @hfill email: @email{sales@@fsf.org}@*
126 Free Software Foundation, Inc. @hfill Tel: +1 (617) 542-5942@*
127 51 Franklin Street, Fifth Floor @hfill Fax: +1 (617) 542-2652@*
128 Boston, MA 02110-1301 USA
132 Printed copies available from @uref{http://shop.fsf.org/}. Published by:
135 GNU Press, http://www.fsf.org/licensing/gnu-press/
136 a division of the email: sales@@fsf.org
137 Free Software Foundation, Inc. Tel: +1 (617) 542-5942
138 51 Franklin Street, Fifth Floor Fax: +1 (617) 542-2652
139 Boston, MA 02110-1301 USA
147 Permission is granted to copy, distribute and/or modify this document
148 under the terms of the GNU Free Documentation License, Version 1.3 or
149 any later version published by the Free Software Foundation; there
150 being no Invariant Section, with the Front-Cover Texts being ``A GNU
151 Manual'', and with the Back-Cover Texts as in (a) below. A copy of
152 the license is included in the section entitled ``GNU Free
153 Documentation License''.
155 (a) The FSF's Back-Cover Text is: ``You have the freedom to
156 copy and modify this GNU manual. Buying copies from the FSF
157 supports it in developing GNU and promoting software freedom.''
161 @c half title; two lines here, so do not use `shorttitlepage'
164 \hbox{}\vskip 1.5in \chaprm \centerline{An Introduction to}%
166 {\begingroup\hbox{}\vskip 0.25in \chaprm%
167 \centerline{Programming in Emacs Lisp}%
168 \endgroup\page\hbox{}\page}
173 @center @titlefont{An Introduction to}
175 @center @titlefont{Programming in Emacs Lisp}
177 @center Revised Third Edition
179 @center by Robert J. Chassell
182 @vskip 0pt plus 1filll
188 @evenheading @thispage @| @| @thischapter
189 @oddheading @thissection @| @| @thispage
193 @c Keep T.O.C. short by tightening up for largebook
196 \global\parskip 2pt plus 1pt
197 \global\advance\baselineskip by -1pt
207 @top An Introduction to Programming in Emacs Lisp
211 <p>The homepage for GNU Emacs is at
212 <a href="/software/emacs/">http://www.gnu.org/software/emacs/</a>.<br>
213 To view this manual in other formats, click
214 <a href="/software/emacs/manual/eintr.html">here</a>.
220 This master menu first lists each chapter and index; then it lists
221 every node in every chapter.
224 @c >>>> Set pageno appropriately <<<<
226 @c The first page of the Preface is a roman numeral; it is the first
227 @c right handed page after the Table of Contents; hence the following
228 @c setting must be for an odd negative number.
231 @c global@pageno = -11
234 @set COUNT-WORDS count-words-example
235 @c Length of variable name chosen so that things still line up when expanded.
238 * Preface:: What to look for.
239 * List Processing:: What is Lisp?
240 * Practicing Evaluation:: Running several programs.
241 * Writing Defuns:: How to write function definitions.
242 * Buffer Walk Through:: Exploring a few buffer-related functions.
243 * More Complex:: A few, even more complex functions.
244 * Narrowing & Widening:: Restricting your and Emacs attention to
246 * car cdr & cons:: Fundamental functions in Lisp.
247 * Cutting & Storing Text:: Removing text and saving it.
248 * List Implementation:: How lists are implemented in the computer.
249 * Yanking:: Pasting stored text.
250 * Loops & Recursion:: How to repeat a process.
251 * Regexp Search:: Regular expression searches.
252 * Counting Words:: A review of repetition and regexps.
253 * Words in a defun:: Counting words in a @code{defun}.
254 * Readying a Graph:: A prototype graph printing function.
255 * Emacs Initialization:: How to write a @file{.emacs} file.
256 * Debugging:: How to run the Emacs Lisp debuggers.
257 * Conclusion:: Now you have the basics.
258 * the-the:: An appendix: how to find reduplicated words.
259 * Kill Ring:: An appendix: how the kill ring works.
260 * Full Graph:: How to create a graph with labeled axes.
261 * Free Software and Free Manuals::
262 * GNU Free Documentation License::
267 --- The Detailed Node Listing ---
271 * Why:: Why learn Emacs Lisp?
272 * On Reading this Text:: Read, gain familiarity, pick up habits....
273 * Who You Are:: For whom this is written.
275 * Note for Novices:: You can read this as a novice.
280 * Lisp Lists:: What are lists?
281 * Run a Program:: Any list in Lisp is a program ready to run.
282 * Making Errors:: Generating an error message.
283 * Names & Definitions:: Names of symbols and function definitions.
284 * Lisp Interpreter:: What the Lisp interpreter does.
285 * Evaluation:: Running a program.
286 * Variables:: Returning a value from a variable.
287 * Arguments:: Passing information to a function.
288 * set & setq:: Setting the value of a variable.
289 * Summary:: The major points.
290 * Error Message Exercises::
294 * Numbers Lists:: List have numbers, other lists, in them.
295 * Lisp Atoms:: Elemental entities.
296 * Whitespace in Lists:: Formatting lists to be readable.
297 * Typing Lists:: How GNU Emacs helps you type lists.
301 * Complications:: Variables, Special forms, Lists within.
302 * Byte Compiling:: Specially processing code for speed.
306 * How the Interpreter Acts:: Returns and Side Effects...
307 * Evaluating Inner Lists:: Lists within lists...
311 * fill-column Example::
312 * Void Function:: The error message for a symbol
314 * Void Variable:: The error message for a symbol without a value.
318 * Data types:: Types of data passed to a function.
319 * Args as Variable or List:: An argument can be the value
320 of a variable or list.
321 * Variable Number of Arguments:: Some functions may take a
322 variable number of arguments.
323 * Wrong Type of Argument:: Passing an argument of the wrong type
325 * message:: A useful function for sending messages.
327 Setting the Value of a Variable
329 * Using set:: Setting values.
330 * Using setq:: Setting a quoted value.
331 * Counting:: Using @code{setq} to count.
333 Practicing Evaluation
335 * How to Evaluate:: Typing editing commands or @kbd{C-x C-e}
337 * Buffer Names:: Buffers and files are different.
338 * Getting Buffers:: Getting a buffer itself, not merely its name.
339 * Switching Buffers:: How to change to another buffer.
340 * Buffer Size & Locations:: Where point is located and the size of
342 * Evaluation Exercise::
344 How To Write Function Definitions
346 * Primitive Functions::
347 * defun:: The @code{defun} macro.
348 * Install:: Install a function definition.
349 * Interactive:: Making a function interactive.
350 * Interactive Options:: Different options for @code{interactive}.
351 * Permanent Installation:: Installing code permanently.
352 * let:: Creating and initializing local variables.
354 * else:: If--then--else expressions.
355 * Truth & Falsehood:: What Lisp considers false and true.
356 * save-excursion:: Keeping track of point, mark, and buffer.
360 Install a Function Definition
362 * Effect of installation::
363 * Change a defun:: How to change a function definition.
365 Make a Function Interactive
367 * Interactive multiply-by-seven:: An overview.
368 * multiply-by-seven in detail:: The interactive version.
372 * Prevent confusion::
373 * Parts of let Expression::
374 * Sample let Expression::
375 * Uninitialized let Variables::
377 The @code{if} Special Form
379 * if in more detail::
380 * type-of-animal in detail:: An example of an @code{if} expression.
382 Truth and Falsehood in Emacs Lisp
384 * nil explained:: @code{nil} has two meanings.
386 @code{save-excursion}
388 * Point and mark:: A review of various locations.
389 * Template for save-excursion::
391 A Few Buffer--Related Functions
393 * Finding More:: How to find more information.
394 * simplified-beginning-of-buffer:: Shows @code{goto-char},
395 @code{point-min}, and @code{push-mark}.
396 * mark-whole-buffer:: Almost the same as @code{beginning-of-buffer}.
397 * append-to-buffer:: Uses @code{save-excursion} and
398 @code{insert-buffer-substring}.
399 * Buffer Related Review:: Review.
402 The Definition of @code{mark-whole-buffer}
404 * mark-whole-buffer overview::
405 * Body of mark-whole-buffer:: Only three lines of code.
407 The Definition of @code{append-to-buffer}
409 * append-to-buffer overview::
410 * append interactive:: A two part interactive expression.
411 * append-to-buffer body:: Incorporates a @code{let} expression.
412 * append save-excursion:: How the @code{save-excursion} works.
414 A Few More Complex Functions
416 * copy-to-buffer:: With @code{set-buffer}, @code{get-buffer-create}.
417 * insert-buffer:: Read-only, and with @code{or}.
418 * beginning-of-buffer:: Shows @code{goto-char},
419 @code{point-min}, and @code{push-mark}.
420 * Second Buffer Related Review::
421 * optional Exercise::
423 The Definition of @code{insert-buffer}
425 * insert-buffer code::
426 * insert-buffer interactive:: When you can read, but not write.
427 * insert-buffer body:: The body has an @code{or} and a @code{let}.
428 * if & or:: Using an @code{if} instead of an @code{or}.
429 * Insert or:: How the @code{or} expression works.
430 * Insert let:: Two @code{save-excursion} expressions.
431 * New insert-buffer::
433 The Interactive Expression in @code{insert-buffer}
435 * Read-only buffer:: When a buffer cannot be modified.
436 * b for interactive:: An existing buffer or else its name.
438 Complete Definition of @code{beginning-of-buffer}
440 * Optional Arguments::
441 * beginning-of-buffer opt arg:: Example with optional argument.
442 * beginning-of-buffer complete::
444 @code{beginning-of-buffer} with an Argument
446 * Disentangle beginning-of-buffer::
447 * Large buffer case::
448 * Small buffer case::
450 Narrowing and Widening
452 * Narrowing advantages:: The advantages of narrowing
453 * save-restriction:: The @code{save-restriction} special form.
454 * what-line:: The number of the line that point is on.
457 @code{car}, @code{cdr}, @code{cons}: Fundamental Functions
459 * Strange Names:: An historical aside: why the strange names?
460 * car & cdr:: Functions for extracting part of a list.
461 * cons:: Constructing a list.
462 * nthcdr:: Calling @code{cdr} repeatedly.
464 * setcar:: Changing the first element of a list.
465 * setcdr:: Changing the rest of a list.
471 * length:: How to find the length of a list.
473 Cutting and Storing Text
475 * Storing Text:: Text is stored in a list.
476 * zap-to-char:: Cutting out text up to a character.
477 * kill-region:: Cutting text out of a region.
478 * copy-region-as-kill:: A definition for copying text.
479 * Digression into C:: Minor note on C programming language macros.
480 * defvar:: How to give a variable an initial value.
481 * cons & search-fwd Review::
486 * Complete zap-to-char:: The complete implementation.
487 * zap-to-char interactive:: A three part interactive expression.
488 * zap-to-char body:: A short overview.
489 * search-forward:: How to search for a string.
490 * progn:: The @code{progn} special form.
491 * Summing up zap-to-char:: Using @code{point} and @code{search-forward}.
495 * Complete kill-region:: The function definition.
496 * condition-case:: Dealing with a problem.
499 @code{copy-region-as-kill}
501 * Complete copy-region-as-kill:: The complete function definition.
502 * copy-region-as-kill body:: The body of @code{copy-region-as-kill}.
504 The Body of @code{copy-region-as-kill}
506 * last-command & this-command::
507 * kill-append function::
508 * kill-new function::
510 Initializing a Variable with @code{defvar}
512 * See variable current value::
513 * defvar and asterisk::
515 How Lists are Implemented
518 * Symbols as Chest:: Exploring a powerful metaphor.
523 * Kill Ring Overview::
524 * kill-ring-yank-pointer:: The kill ring is a list.
525 * yank nthcdr Exercises:: The @code{kill-ring-yank-pointer} variable.
529 * while:: Causing a stretch of code to repeat.
531 * Recursion:: Causing a function to call itself.
536 * Looping with while:: Repeat so long as test returns true.
537 * Loop Example:: A @code{while} loop that uses a list.
538 * print-elements-of-list:: Uses @code{while}, @code{car}, @code{cdr}.
539 * Incrementing Loop:: A loop with an incrementing counter.
540 * Incrementing Loop Details::
541 * Decrementing Loop:: A loop with a decrementing counter.
543 Details of an Incrementing Loop
545 * Incrementing Example:: Counting pebbles in a triangle.
546 * Inc Example parts:: The parts of the function definition.
547 * Inc Example altogether:: Putting the function definition together.
549 Loop with a Decrementing Counter
551 * Decrementing Example:: More pebbles on the beach.
552 * Dec Example parts:: The parts of the function definition.
553 * Dec Example altogether:: Putting the function definition together.
555 Save your time: @code{dolist} and @code{dotimes}
562 * Building Robots:: Same model, different serial number ...
563 * Recursive Definition Parts:: Walk until you stop ...
564 * Recursion with list:: Using a list as the test whether to recurse.
565 * Recursive triangle function::
566 * Recursion with cond::
567 * Recursive Patterns:: Often used templates.
568 * No Deferment:: Don't store up work ...
569 * No deferment solution::
571 Recursion in Place of a Counter
573 * Recursive Example arg of 1 or 2::
574 * Recursive Example arg of 3 or 4::
582 Regular Expression Searches
584 * sentence-end:: The regular expression for @code{sentence-end}.
585 * re-search-forward:: Very similar to @code{search-forward}.
586 * forward-sentence:: A straightforward example of regexp search.
587 * forward-paragraph:: A somewhat complex example.
588 * etags:: How to create your own @file{TAGS} table.
590 * re-search Exercises::
592 @code{forward-sentence}
594 * Complete forward-sentence::
595 * fwd-sentence while loops:: Two @code{while} loops.
596 * fwd-sentence re-search:: A regular expression search.
598 @code{forward-paragraph}: a Goldmine of Functions
600 * forward-paragraph in brief:: Key parts of the function definition.
601 * fwd-para let:: The @code{let*} expression.
602 * fwd-para while:: The forward motion @code{while} loop.
604 Counting: Repetition and Regexps
607 * @value{COUNT-WORDS}:: Use a regexp, but find a problem.
608 * recursive-count-words:: Start with case of no words in region.
609 * Counting Exercise::
611 The @code{@value{COUNT-WORDS}} Function
613 * Design @value{COUNT-WORDS}:: The definition using a @code{while} loop.
614 * Whitespace Bug:: The Whitespace Bug in @code{@value{COUNT-WORDS}}.
616 Counting Words in a @code{defun}
618 * Divide and Conquer::
619 * Words and Symbols:: What to count?
620 * Syntax:: What constitutes a word or symbol?
621 * count-words-in-defun:: Very like @code{@value{COUNT-WORDS}}.
622 * Several defuns:: Counting several defuns in a file.
623 * Find a File:: Do you want to look at a file?
624 * lengths-list-file:: A list of the lengths of many definitions.
625 * Several files:: Counting in definitions in different files.
626 * Several files recursively:: Recursively counting in different files.
627 * Prepare the data:: Prepare the data for display in a graph.
629 Count Words in @code{defuns} in Different Files
631 * lengths-list-many-files:: Return a list of the lengths of defuns.
632 * append:: Attach one list to another.
634 Prepare the Data for Display in a Graph
636 * Data for Display in Detail::
637 * Sorting:: Sorting lists.
638 * Files List:: Making a list of files.
639 * Counting function definitions::
643 * Columns of a graph::
644 * graph-body-print:: How to print the body of a graph.
645 * recursive-graph-body-print::
647 * Line Graph Exercise::
649 Your @file{.emacs} File
651 * Default Configuration::
652 * Site-wide Init:: You can write site-wide init files.
653 * defcustom:: Emacs will write code for you.
654 * Beginning init File:: How to write a @file{.emacs} init file.
655 * Text and Auto-fill:: Automatically wrap lines.
656 * Mail Aliases:: Use abbreviations for email addresses.
657 * Indent Tabs Mode:: Don't use tabs with @TeX{}
658 * Keybindings:: Create some personal keybindings.
659 * Keymaps:: More about key binding.
660 * Loading Files:: Load (i.e., evaluate) files automatically.
661 * Autoload:: Make functions available.
662 * Simple Extension:: Define a function; bind it to a key.
663 * X11 Colors:: Colors in X.
665 * Mode Line:: How to customize your mode line.
669 * debug:: How to use the built-in debugger.
670 * debug-on-entry:: Start debugging when you call a function.
671 * debug-on-quit:: Start debugging when you quit with @kbd{C-g}.
672 * edebug:: How to use Edebug, a source level debugger.
673 * Debugging Exercises::
675 Handling the Kill Ring
677 * What the Kill Ring Does::
679 * yank:: Paste a copy of a clipped element.
680 * yank-pop:: Insert element pointed to.
683 The @code{current-kill} Function
685 * Code for current-kill::
686 * Understanding current-kill::
688 @code{current-kill} in Outline
690 * Body of current-kill::
691 * Digression concerning error:: How to mislead humans, but not computers.
692 * Determining the Element::
694 A Graph with Labeled Axes
697 * print-graph Varlist:: @code{let} expression in @code{print-graph}.
698 * print-Y-axis:: Print a label for the vertical axis.
699 * print-X-axis:: Print a horizontal label.
700 * Print Whole Graph:: The function to print a complete graph.
702 The @code{print-Y-axis} Function
704 * print-Y-axis in Detail::
705 * Height of label:: What height for the Y axis?
706 * Compute a Remainder:: How to compute the remainder of a division.
707 * Y Axis Element:: Construct a line for the Y axis.
708 * Y-axis-column:: Generate a list of Y axis labels.
709 * print-Y-axis Penultimate:: A not quite final version.
711 The @code{print-X-axis} Function
713 * Similarities differences:: Much like @code{print-Y-axis}, but not exactly.
714 * X Axis Tic Marks:: Create tic marks for the horizontal axis.
716 Printing the Whole Graph
718 * The final version:: A few changes.
719 * Test print-graph:: Run a short test.
720 * Graphing words in defuns:: Executing the final code.
721 * lambda:: How to write an anonymous function.
722 * mapcar:: Apply a function to elements of a list.
723 * Another Bug:: Yet another bug @dots{} most insidious.
724 * Final printed graph:: The graph itself!
732 Most of the GNU Emacs integrated environment is written in the programming
733 language called Emacs Lisp. The code written in this programming
734 language is the software---the sets of instructions---that tell the
735 computer what to do when you give it commands. Emacs is designed so
736 that you can write new code in Emacs Lisp and easily install it as an
737 extension to the editor.
739 (GNU Emacs is sometimes called an ``extensible editor'', but it does
740 much more than provide editing capabilities. It is better to refer to
741 Emacs as an ``extensible computing environment''. However, that
742 phrase is quite a mouthful. It is easier to refer to Emacs simply as
743 an editor. Moreover, everything you do in Emacs---find the Mayan date
744 and phases of the moon, simplify polynomials, debug code, manage
745 files, read letters, write books---all these activities are kinds of
746 editing in the most general sense of the word.)
749 * Why:: Why learn Emacs Lisp?
750 * On Reading this Text:: Read, gain familiarity, pick up habits....
751 * Who You Are:: For whom this is written.
753 * Note for Novices:: You can read this as a novice.
759 @unnumberedsec Why Study Emacs Lisp?
762 Although Emacs Lisp is usually thought of in association only with Emacs,
763 it is a full computer programming language. You can use Emacs Lisp as
764 you would any other programming language.
766 Perhaps you want to understand programming; perhaps you want to extend
767 Emacs; or perhaps you want to become a programmer. This introduction to
768 Emacs Lisp is designed to get you started: to guide you in learning the
769 fundamentals of programming, and more importantly, to show you how you
770 can teach yourself to go further.
772 @node On Reading this Text
773 @unnumberedsec On Reading this Text
775 All through this document, you will see little sample programs you can
776 run inside of Emacs. If you read this document in Info inside of GNU
777 Emacs, you can run the programs as they appear. (This is easy to do and
778 is explained when the examples are presented.) Alternatively, you can
779 read this introduction as a printed book while sitting beside a computer
780 running Emacs. (This is what I like to do; I like printed books.) If
781 you don't have a running Emacs beside you, you can still read this book,
782 but in this case, it is best to treat it as a novel or as a travel guide
783 to a country not yet visited: interesting, but not the same as being
786 Much of this introduction is dedicated to walkthroughs or guided tours
787 of code used in GNU Emacs. These tours are designed for two purposes:
788 first, to give you familiarity with real, working code (code you use
789 every day); and, second, to give you familiarity with the way Emacs
790 works. It is interesting to see how a working environment is
793 hope that you will pick up the habit of browsing through source code.
794 You can learn from it and mine it for ideas. Having GNU Emacs is like
795 having a dragon's cave of treasures.
797 In addition to learning about Emacs as an editor and Emacs Lisp as a
798 programming language, the examples and guided tours will give you an
799 opportunity to get acquainted with Emacs as a Lisp programming
800 environment. GNU Emacs supports programming and provides tools that
801 you will want to become comfortable using, such as @kbd{M-.} (the key
802 which invokes the @code{find-tag} command). You will also learn about
803 buffers and other objects that are part of the environment.
804 Learning about these features of Emacs is like learning new routes
805 around your home town.
808 In addition, I have written several programs as extended examples.
809 Although these are examples, the programs are real. I use them.
810 Other people use them. You may use them. Beyond the fragments of
811 programs used for illustrations, there is very little in here that is
812 `just for teaching purposes'; what you see is used. This is a great
813 advantage of Emacs Lisp: it is easy to learn to use it for work.
816 Finally, I hope to convey some of the skills for using Emacs to
817 learn aspects of programming that you don't know. You can often use
818 Emacs to help you understand what puzzles you or to find out how to do
819 something new. This self-reliance is not only a pleasure, but an
823 @unnumberedsec For Whom This is Written
825 This text is written as an elementary introduction for people who are
826 not programmers. If you are a programmer, you may not be satisfied with
827 this primer. The reason is that you may have become expert at reading
828 reference manuals and be put off by the way this text is organized.
830 An expert programmer who reviewed this text said to me:
833 @i{I prefer to learn from reference manuals. I ``dive into'' each
834 paragraph, and ``come up for air'' between paragraphs.}
836 @i{When I get to the end of a paragraph, I assume that that subject is
837 done, finished, that I know everything I need (with the
838 possible exception of the case when the next paragraph starts talking
839 about it in more detail). I expect that a well written reference manual
840 will not have a lot of redundancy, and that it will have excellent
841 pointers to the (one) place where the information I want is.}
844 This introduction is not written for this person!
846 Firstly, I try to say everything at least three times: first, to
847 introduce it; second, to show it in context; and third, to show it in a
848 different context, or to review it.
850 Secondly, I hardly ever put all the information about a subject in one
851 place, much less in one paragraph. To my way of thinking, that imposes
852 too heavy a burden on the reader. Instead I try to explain only what
853 you need to know at the time. (Sometimes I include a little extra
854 information so you won't be surprised later when the additional
855 information is formally introduced.)
857 When you read this text, you are not expected to learn everything the
858 first time. Frequently, you need only make, as it were, a `nodding
859 acquaintance' with some of the items mentioned. My hope is that I have
860 structured the text and given you enough hints that you will be alert to
861 what is important, and concentrate on it.
863 You will need to ``dive into'' some paragraphs; there is no other way
864 to read them. But I have tried to keep down the number of such
865 paragraphs. This book is intended as an approachable hill, rather than
866 as a daunting mountain.
868 This introduction to @cite{Programming in Emacs Lisp} has a companion
871 @cite{The GNU Emacs Lisp Reference Manual}.
874 @ref{Top, , The GNU Emacs Lisp Reference Manual, elisp, The GNU
875 Emacs Lisp Reference Manual}.
877 The reference manual has more detail than this introduction. In the
878 reference manual, all the information about one topic is concentrated
879 in one place. You should turn to it if you are like the programmer
880 quoted above. And, of course, after you have read this
881 @cite{Introduction}, you will find the @cite{Reference Manual} useful
882 when you are writing your own programs.
885 @unnumberedsec Lisp History
888 Lisp was first developed in the late 1950s at the Massachusetts
889 Institute of Technology for research in artificial intelligence. The
890 great power of the Lisp language makes it superior for other purposes as
891 well, such as writing editor commands and integrated environments.
895 GNU Emacs Lisp is largely inspired by Maclisp, which was written at MIT
896 in the 1960s. It is somewhat inspired by Common Lisp, which became a
897 standard in the 1980s. However, Emacs Lisp is much simpler than Common
898 Lisp. (The standard Emacs distribution contains an optional extensions
899 file, @file{cl.el}, that adds many Common Lisp features to Emacs Lisp.)
901 @node Note for Novices
902 @unnumberedsec A Note for Novices
904 If you don't know GNU Emacs, you can still read this document
905 profitably. However, I recommend you learn Emacs, if only to learn to
906 move around your computer screen. You can teach yourself how to use
907 Emacs with the on-line tutorial. To use it, type @kbd{C-h t}. (This
908 means you press and release the @key{CTRL} key and the @kbd{h} at the
909 same time, and then press and release @kbd{t}.)
911 Also, I often refer to one of Emacs's standard commands by listing the
912 keys which you press to invoke the command and then giving the name of
913 the command in parentheses, like this: @kbd{M-C-\}
914 (@code{indent-region}). What this means is that the
915 @code{indent-region} command is customarily invoked by typing
916 @kbd{M-C-\}. (You can, if you wish, change the keys that are typed to
917 invoke the command; this is called @dfn{rebinding}. @xref{Keymaps, ,
918 Keymaps}.) The abbreviation @kbd{M-C-\} means that you type your
919 @key{META} key, @key{CTRL} key and @key{\} key all at the same time.
920 (On many modern keyboards the @key{META} key is labeled
922 Sometimes a combination like this is called a keychord, since it is
923 similar to the way you play a chord on a piano. If your keyboard does
924 not have a @key{META} key, the @key{ESC} key prefix is used in place
925 of it. In this case, @kbd{M-C-\} means that you press and release your
926 @key{ESC} key and then type the @key{CTRL} key and the @key{\} key at
927 the same time. But usually @kbd{M-C-\} means press the @key{CTRL} key
928 along with the key that is labeled @key{ALT} and, at the same time,
929 press the @key{\} key.
931 In addition to typing a lone keychord, you can prefix what you type
932 with @kbd{C-u}, which is called the `universal argument'. The
933 @kbd{C-u} keychord passes an argument to the subsequent command.
934 Thus, to indent a region of plain text by 6 spaces, mark the region,
935 and then type @w{@kbd{C-u 6 M-C-\}}. (If you do not specify a number,
936 Emacs either passes the number 4 to the command or otherwise runs the
937 command differently than it would otherwise.) @xref{Arguments, ,
938 Numeric Arguments, emacs, The GNU Emacs Manual}.
940 If you are reading this in Info using GNU Emacs, you can read through
941 this whole document just by pressing the space bar, @key{SPC}.
942 (To learn about Info, type @kbd{C-h i} and then select Info.)
944 A note on terminology: when I use the word Lisp alone, I often am
945 referring to the various dialects of Lisp in general, but when I speak
946 of Emacs Lisp, I am referring to GNU Emacs Lisp in particular.
949 @unnumberedsec Thank You
951 My thanks to all who helped me with this book. My especial thanks to
952 @r{Jim Blandy}, @r{Noah Friedman}, @w{Jim Kingdon}, @r{Roland
953 McGrath}, @w{Frank Ritter}, @w{Randy Smith}, @w{Richard M.
954 Stallman}, and @w{Melissa Weisshaus}. My thanks also go to both
955 @w{Philip Johnson} and @w{David Stampe} for their patient
956 encouragement. My mistakes are my own.
968 @c ================ Beginning of main text ================
970 @c Start main text on right-hand (verso) page
973 \par\vfill\supereject
976 \par\vfill\supereject
978 \par\vfill\supereject
980 \par\vfill\supereject
984 @c Note: this resetting of the page number back to 1 causes TeX to gripe
985 @c about already having seen page numbers 1-4 before (in the preface):
986 @c pdfTeX warning (ext4): destination with the same identifier (name{1})
987 @c has been already used, duplicate ignored
988 @c I guess that is harmless (what happens if a later part of the text
989 @c makes a link to something in the first 4 pages though?).
990 @c E.g., note that the Emacs manual has a preface, but does not bother
991 @c resetting the page numbers back to 1 after that.
994 @evenheading @thispage @| @| @thischapter
995 @oddheading @thissection @| @| @thispage
999 @node List Processing
1000 @chapter List Processing
1002 To the untutored eye, Lisp is a strange programming language. In Lisp
1003 code there are parentheses everywhere. Some people even claim that
1004 the name stands for `Lots of Isolated Silly Parentheses'. But the
1005 claim is unwarranted. Lisp stands for LISt Processing, and the
1006 programming language handles @emph{lists} (and lists of lists) by
1007 putting them between parentheses. The parentheses mark the boundaries
1008 of the list. Sometimes a list is preceded by a single apostrophe or
1009 quotation mark, @samp{'}@footnote{The single apostrophe or quotation
1010 mark is an abbreviation for the function @code{quote}; you need not
1011 think about functions now; functions are defined in @ref{Making
1012 Errors, , Generate an Error Message}.} Lists are the basis of Lisp.
1015 * Lisp Lists:: What are lists?
1016 * Run a Program:: Any list in Lisp is a program ready to run.
1017 * Making Errors:: Generating an error message.
1018 * Names & Definitions:: Names of symbols and function definitions.
1019 * Lisp Interpreter:: What the Lisp interpreter does.
1020 * Evaluation:: Running a program.
1021 * Variables:: Returning a value from a variable.
1022 * Arguments:: Passing information to a function.
1023 * set & setq:: Setting the value of a variable.
1024 * Summary:: The major points.
1025 * Error Message Exercises::
1032 In Lisp, a list looks like this: @code{'(rose violet daisy buttercup)}.
1033 This list is preceded by a single apostrophe. It could just as well be
1034 written as follows, which looks more like the kind of list you are likely
1035 to be familiar with:
1047 The elements of this list are the names of the four different flowers,
1048 separated from each other by whitespace and surrounded by parentheses,
1049 like flowers in a field with a stone wall around them.
1050 @cindex Flowers in a field
1053 * Numbers Lists:: List have numbers, other lists, in them.
1054 * Lisp Atoms:: Elemental entities.
1055 * Whitespace in Lists:: Formatting lists to be readable.
1056 * Typing Lists:: How GNU Emacs helps you type lists.
1061 @unnumberedsubsec Numbers, Lists inside of Lists
1064 Lists can also have numbers in them, as in this list: @code{(+ 2 2)}.
1065 This list has a plus-sign, @samp{+}, followed by two @samp{2}s, each
1066 separated by whitespace.
1068 In Lisp, both data and programs are represented the same way; that is,
1069 they are both lists of words, numbers, or other lists, separated by
1070 whitespace and surrounded by parentheses. (Since a program looks like
1071 data, one program may easily serve as data for another; this is a very
1072 powerful feature of Lisp.) (Incidentally, these two parenthetical
1073 remarks are @emph{not} Lisp lists, because they contain @samp{;} and
1074 @samp{.} as punctuation marks.)
1077 Here is another list, this time with a list inside of it:
1080 '(this list has (a list inside of it))
1083 The components of this list are the words @samp{this}, @samp{list},
1084 @samp{has}, and the list @samp{(a list inside of it)}. The interior
1085 list is made up of the words @samp{a}, @samp{list}, @samp{inside},
1086 @samp{of}, @samp{it}.
1089 @subsection Lisp Atoms
1092 In Lisp, what we have been calling words are called @dfn{atoms}. This
1093 term comes from the historical meaning of the word atom, which means
1094 `indivisible'. As far as Lisp is concerned, the words we have been
1095 using in the lists cannot be divided into any smaller parts and still
1096 mean the same thing as part of a program; likewise with numbers and
1097 single character symbols like @samp{+}. On the other hand, unlike an
1098 ancient atom, a list can be split into parts. (@xref{car cdr & cons,
1099 , @code{car} @code{cdr} & @code{cons} Fundamental Functions}.)
1101 In a list, atoms are separated from each other by whitespace. They can be
1102 right next to a parenthesis.
1104 @cindex @samp{empty list} defined
1105 Technically speaking, a list in Lisp consists of parentheses surrounding
1106 atoms separated by whitespace or surrounding other lists or surrounding
1107 both atoms and other lists. A list can have just one atom in it or
1108 have nothing in it at all. A list with nothing in it looks like this:
1109 @code{()}, and is called the @dfn{empty list}. Unlike anything else, an
1110 empty list is considered both an atom and a list at the same time.
1112 @cindex Symbolic expressions, introduced
1113 @cindex @samp{expression} defined
1114 @cindex @samp{form} defined
1115 The printed representation of both atoms and lists are called
1116 @dfn{symbolic expressions} or, more concisely, @dfn{s-expressions}.
1117 The word @dfn{expression} by itself can refer to either the printed
1118 representation, or to the atom or list as it is held internally in the
1119 computer. Often, people use the term @dfn{expression}
1120 indiscriminately. (Also, in many texts, the word @dfn{form} is used
1121 as a synonym for expression.)
1123 Incidentally, the atoms that make up our universe were named such when
1124 they were thought to be indivisible; but it has been found that physical
1125 atoms are not indivisible. Parts can split off an atom or it can
1126 fission into two parts of roughly equal size. Physical atoms were named
1127 prematurely, before their truer nature was found. In Lisp, certain
1128 kinds of atom, such as an array, can be separated into parts; but the
1129 mechanism for doing this is different from the mechanism for splitting a
1130 list. As far as list operations are concerned, the atoms of a list are
1133 As in English, the meanings of the component letters of a Lisp atom
1134 are different from the meaning the letters make as a word. For
1135 example, the word for the South American sloth, the @samp{ai}, is
1136 completely different from the two words, @samp{a}, and @samp{i}.
1138 There are many kinds of atom in nature but only a few in Lisp: for
1139 example, @dfn{numbers}, such as 37, 511, or 1729, and @dfn{symbols}, such
1140 as @samp{+}, @samp{foo}, or @samp{forward-line}. The words we have
1141 listed in the examples above are all symbols. In everyday Lisp
1142 conversation, the word ``atom'' is not often used, because programmers
1143 usually try to be more specific about what kind of atom they are dealing
1144 with. Lisp programming is mostly about symbols (and sometimes numbers)
1145 within lists. (Incidentally, the preceding three word parenthetical
1146 remark is a proper list in Lisp, since it consists of atoms, which in
1147 this case are symbols, separated by whitespace and enclosed by
1148 parentheses, without any non-Lisp punctuation.)
1151 Text between double quotation marks---even sentences or
1152 paragraphs---is also an atom. Here is an example:
1153 @cindex Text between double quotation marks
1156 '(this list includes "text between quotation marks.")
1159 @cindex @samp{string} defined
1161 In Lisp, all of the quoted text including the punctuation mark and the
1162 blank spaces is a single atom. This kind of atom is called a
1163 @dfn{string} (for `string of characters') and is the sort of thing that
1164 is used for messages that a computer can print for a human to read.
1165 Strings are a different kind of atom than numbers or symbols and are
1168 @node Whitespace in Lists
1169 @subsection Whitespace in Lists
1170 @cindex Whitespace in lists
1173 The amount of whitespace in a list does not matter. From the point of view
1174 of the Lisp language,
1185 is exactly the same as this:
1188 '(this list looks like this)
1191 Both examples show what to Lisp is the same list, the list made up of
1192 the symbols @samp{this}, @samp{list}, @samp{looks}, @samp{like}, and
1193 @samp{this} in that order.
1195 Extra whitespace and newlines are designed to make a list more readable
1196 by humans. When Lisp reads the expression, it gets rid of all the extra
1197 whitespace (but it needs to have at least one space between atoms in
1198 order to tell them apart.)
1200 Odd as it seems, the examples we have seen cover almost all of what Lisp
1201 lists look like! Every other list in Lisp looks more or less like one
1202 of these examples, except that the list may be longer and more complex.
1203 In brief, a list is between parentheses, a string is between quotation
1204 marks, a symbol looks like a word, and a number looks like a number.
1205 (For certain situations, square brackets, dots and a few other special
1206 characters may be used; however, we will go quite far without them.)
1209 @subsection GNU Emacs Helps You Type Lists
1210 @cindex Help typing lists
1211 @cindex Formatting help
1213 When you type a Lisp expression in GNU Emacs using either Lisp
1214 Interaction mode or Emacs Lisp mode, you have available to you several
1215 commands to format the Lisp expression so it is easy to read. For
1216 example, pressing the @key{TAB} key automatically indents the line the
1217 cursor is on by the right amount. A command to properly indent the
1218 code in a region is customarily bound to @kbd{M-C-\}. Indentation is
1219 designed so that you can see which elements of a list belong to which
1220 list---elements of a sub-list are indented more than the elements of
1223 In addition, when you type a closing parenthesis, Emacs momentarily
1224 jumps the cursor back to the matching opening parenthesis, so you can
1225 see which one it is. This is very useful, since every list you type
1226 in Lisp must have its closing parenthesis match its opening
1227 parenthesis. (@xref{Major Modes, , Major Modes, emacs, The GNU Emacs
1228 Manual}, for more information about Emacs's modes.)
1231 @section Run a Program
1232 @cindex Run a program
1233 @cindex Program, running one
1235 @cindex @samp{evaluate} defined
1236 A list in Lisp---any list---is a program ready to run. If you run it
1237 (for which the Lisp jargon is @dfn{evaluate}), the computer will do one
1238 of three things: do nothing except return to you the list itself; send
1239 you an error message; or, treat the first symbol in the list as a
1240 command to do something. (Usually, of course, it is the last of these
1241 three things that you really want!)
1243 @c use code for the single apostrophe, not samp.
1244 The single apostrophe, @code{'}, that I put in front of some of the
1245 example lists in preceding sections is called a @dfn{quote}; when it
1246 precedes a list, it tells Lisp to do nothing with the list, other than
1247 take it as it is written. But if there is no quote preceding a list,
1248 the first item of the list is special: it is a command for the computer
1249 to obey. (In Lisp, these commands are called @emph{functions}.) The list
1250 @code{(+ 2 2)} shown above did not have a quote in front of it, so Lisp
1251 understands that the @code{+} is an instruction to do something with the
1252 rest of the list: add the numbers that follow.
1255 If you are reading this inside of GNU Emacs in Info, here is how you can
1256 evaluate such a list: place your cursor immediately after the right
1257 hand parenthesis of the following list and then type @kbd{C-x C-e}:
1263 @c use code for the number four, not samp.
1265 You will see the number @code{4} appear in the echo area. (In the
1266 jargon, what you have just done is ``evaluate the list.'' The echo area
1267 is the line at the bottom of the screen that displays or ``echoes''
1268 text.) Now try the same thing with a quoted list: place the cursor
1269 right after the following list and type @kbd{C-x C-e}:
1272 '(this is a quoted list)
1276 You will see @code{(this is a quoted list)} appear in the echo area.
1278 @cindex Lisp interpreter, explained
1279 @cindex Interpreter, Lisp, explained
1280 In both cases, what you are doing is giving a command to the program
1281 inside of GNU Emacs called the @dfn{Lisp interpreter}---giving the
1282 interpreter a command to evaluate the expression. The name of the Lisp
1283 interpreter comes from the word for the task done by a human who comes
1284 up with the meaning of an expression---who ``interprets'' it.
1286 You can also evaluate an atom that is not part of a list---one that is
1287 not surrounded by parentheses; again, the Lisp interpreter translates
1288 from the humanly readable expression to the language of the computer.
1289 But before discussing this (@pxref{Variables}), we will discuss what the
1290 Lisp interpreter does when you make an error.
1293 @section Generate an Error Message
1294 @cindex Generate an error message
1295 @cindex Error message generation
1297 Partly so you won't worry if you do it accidentally, we will now give
1298 a command to the Lisp interpreter that generates an error message.
1299 This is a harmless activity; and indeed, we will often try to generate
1300 error messages intentionally. Once you understand the jargon, error
1301 messages can be informative. Instead of being called ``error''
1302 messages, they should be called ``help'' messages. They are like
1303 signposts to a traveler in a strange country; deciphering them can be
1304 hard, but once understood, they can point the way.
1306 The error message is generated by a built-in GNU Emacs debugger. We
1307 will `enter the debugger'. You get out of the debugger by typing @code{q}.
1309 What we will do is evaluate a list that is not quoted and does not
1310 have a meaningful command as its first element. Here is a list almost
1311 exactly the same as the one we just used, but without the single-quote
1312 in front of it. Position the cursor right after it and type @kbd{C-x
1316 (this is an unquoted list)
1321 What you see depends on which version of Emacs you are running. GNU
1322 Emacs version 22 provides more information than version 20 and before.
1323 First, the more recent result of generating an error; then the
1324 earlier, version 20 result.
1328 In GNU Emacs version 22, a @file{*Backtrace*} window will open up and
1329 you will see the following in it:
1332 A @file{*Backtrace*} window will open up and you should see the
1337 ---------- Buffer: *Backtrace* ----------
1338 Debugger entered--Lisp error: (void-function this)
1339 (this is an unquoted list)
1340 eval((this is an unquoted list))
1341 eval-last-sexp-1(nil)
1343 call-interactively(eval-last-sexp)
1344 ---------- Buffer: *Backtrace* ----------
1350 Your cursor will be in this window (you may have to wait a few seconds
1351 before it becomes visible). To quit the debugger and make the
1352 debugger window go away, type:
1359 Please type @kbd{q} right now, so you become confident that you can
1360 get out of the debugger. Then, type @kbd{C-x C-e} again to re-enter
1363 @cindex @samp{function} defined
1364 Based on what we already know, we can almost read this error message.
1366 You read the @file{*Backtrace*} buffer from the bottom up; it tells
1367 you what Emacs did. When you typed @kbd{C-x C-e}, you made an
1368 interactive call to the command @code{eval-last-sexp}. @code{eval} is
1369 an abbreviation for `evaluate' and @code{sexp} is an abbreviation for
1370 `symbolic expression'. The command means `evaluate last symbolic
1371 expression', which is the expression just before your cursor.
1373 Each line above tells you what the Lisp interpreter evaluated next.
1374 The most recent action is at the top. The buffer is called the
1375 @file{*Backtrace*} buffer because it enables you to track Emacs
1379 At the top of the @file{*Backtrace*} buffer, you see the line:
1382 Debugger entered--Lisp error: (void-function this)
1386 The Lisp interpreter tried to evaluate the first atom of the list, the
1387 word @samp{this}. It is this action that generated the error message
1388 @samp{void-function this}.
1390 The message contains the words @samp{void-function} and @samp{this}.
1392 @cindex @samp{function} defined
1393 The word @samp{function} was mentioned once before. It is a very
1394 important word. For our purposes, we can define it by saying that a
1395 @dfn{function} is a set of instructions to the computer that tell the
1396 computer to do something.
1398 Now we can begin to understand the error message: @samp{void-function
1399 this}. The function (that is, the word @samp{this}) does not have a
1400 definition of any set of instructions for the computer to carry out.
1402 The slightly odd word, @samp{void-function}, is designed to cover the
1403 way Emacs Lisp is implemented, which is that when a symbol does not
1404 have a function definition attached to it, the place that should
1405 contain the instructions is `void'.
1407 On the other hand, since we were able to add 2 plus 2 successfully, by
1408 evaluating @code{(+ 2 2)}, we can infer that the symbol @code{+} must
1409 have a set of instructions for the computer to obey and those
1410 instructions must be to add the numbers that follow the @code{+}.
1412 It is possible to prevent Emacs entering the debugger in cases like
1413 this. We do not explain how to do that here, but we will mention what
1414 the result looks like, because you may encounter a similar situation
1415 if there is a bug in some Emacs code that you are using. In such
1416 cases, you will see only one line of error message; it will appear in
1417 the echo area and look like this:
1420 Symbol's function definition is void:@: this
1425 (Also, your terminal may beep at you---some do, some don't; and others
1426 blink. This is just a device to get your attention.)
1428 The message goes away as soon as you type a key, even just to
1431 We know the meaning of the word @samp{Symbol}. It refers to the first
1432 atom of the list, the word @samp{this}. The word @samp{function}
1433 refers to the instructions that tell the computer what to do.
1434 (Technically, the symbol tells the computer where to find the
1435 instructions, but this is a complication we can ignore for the
1438 The error message can be understood: @samp{Symbol's function
1439 definition is void:@: this}. The symbol (that is, the word
1440 @samp{this}) lacks instructions for the computer to carry out.
1442 @node Names & Definitions
1443 @section Symbol Names and Function Definitions
1444 @cindex Symbol names
1446 We can articulate another characteristic of Lisp based on what we have
1447 discussed so far---an important characteristic: a symbol, like
1448 @code{+}, is not itself the set of instructions for the computer to
1449 carry out. Instead, the symbol is used, perhaps temporarily, as a way
1450 of locating the definition or set of instructions. What we see is the
1451 name through which the instructions can be found. Names of people
1452 work the same way. I can be referred to as @samp{Bob}; however, I am
1453 not the letters @samp{B}, @samp{o}, @samp{b} but am, or was, the
1454 consciousness consistently associated with a particular life-form.
1455 The name is not me, but it can be used to refer to me.
1457 In Lisp, one set of instructions can be attached to several names.
1458 For example, the computer instructions for adding numbers can be
1459 linked to the symbol @code{plus} as well as to the symbol @code{+}
1460 (and are in some dialects of Lisp). Among humans, I can be referred
1461 to as @samp{Robert} as well as @samp{Bob} and by other words as well.
1463 On the other hand, a symbol can have only one function definition
1464 attached to it at a time. Otherwise, the computer would be confused as
1465 to which definition to use. If this were the case among people, only
1466 one person in the world could be named @samp{Bob}. However, the function
1467 definition to which the name refers can be changed readily.
1468 (@xref{Install, , Install a Function Definition}.)
1470 Since Emacs Lisp is large, it is customary to name symbols in a way
1471 that identifies the part of Emacs to which the function belongs.
1472 Thus, all the names for functions that deal with Texinfo start with
1473 @samp{texinfo-} and those for functions that deal with reading mail
1474 start with @samp{rmail-}.
1476 @node Lisp Interpreter
1477 @section The Lisp Interpreter
1478 @cindex Lisp interpreter, what it does
1479 @cindex Interpreter, what it does
1481 Based on what we have seen, we can now start to figure out what the
1482 Lisp interpreter does when we command it to evaluate a list.
1483 First, it looks to see whether there is a quote before the list; if
1484 there is, the interpreter just gives us the list. On the other
1485 hand, if there is no quote, the interpreter looks at the first element
1486 in the list and sees whether it has a function definition. If it does,
1487 the interpreter carries out the instructions in the function definition.
1488 Otherwise, the interpreter prints an error message.
1490 This is how Lisp works. Simple. There are added complications which we
1491 will get to in a minute, but these are the fundamentals. Of course, to
1492 write Lisp programs, you need to know how to write function definitions
1493 and attach them to names, and how to do this without confusing either
1494 yourself or the computer.
1497 * Complications:: Variables, Special forms, Lists within.
1498 * Byte Compiling:: Specially processing code for speed.
1503 @unnumberedsubsec Complications
1506 Now, for the first complication. In addition to lists, the Lisp
1507 interpreter can evaluate a symbol that is not quoted and does not have
1508 parentheses around it. The Lisp interpreter will attempt to determine
1509 the symbol's value as a @dfn{variable}. This situation is described
1510 in the section on variables. (@xref{Variables}.)
1512 @cindex Special form
1513 The second complication occurs because some functions are unusual and
1514 do not work in the usual manner. Those that don't are called
1515 @dfn{special forms}. They are used for special jobs, like defining a
1516 function, and there are not many of them. In the next few chapters,
1517 you will be introduced to several of the more important special forms.
1519 As well as special forms, there are also @dfn{macros}. A macro
1520 is a construct defined in Lisp, which differs from a function in that it
1521 translates a Lisp expression into another expression that is to be
1522 evaluated in place of the original expression. (@xref{Lisp macro}.)
1524 For the purposes of this introduction, you do not need to worry too much
1525 about whether something is a special form, macro, or ordinary function.
1526 For example, @code{if} is a special form (@pxref{if}), but @code{when}
1527 is a macro (@pxref{Lisp macro}). In earlier versions of Emacs,
1528 @code{defun} was a special form, but now it is a macro (@pxref{defun}).
1529 It still behaves in the same way.
1531 The final complication is this: if the function that the
1532 Lisp interpreter is looking at is not a special form, and if it is part
1533 of a list, the Lisp interpreter looks to see whether the list has a list
1534 inside of it. If there is an inner list, the Lisp interpreter first
1535 figures out what it should do with the inside list, and then it works on
1536 the outside list. If there is yet another list embedded inside the
1537 inner list, it works on that one first, and so on. It always works on
1538 the innermost list first. The interpreter works on the innermost list
1539 first, to evaluate the result of that list. The result may be
1540 used by the enclosing expression.
1542 Otherwise, the interpreter works left to right, from one expression to
1545 @node Byte Compiling
1546 @subsection Byte Compiling
1547 @cindex Byte compiling
1549 One other aspect of interpreting: the Lisp interpreter is able to
1550 interpret two kinds of entity: humanly readable code, on which we will
1551 focus exclusively, and specially processed code, called @dfn{byte
1552 compiled} code, which is not humanly readable. Byte compiled code
1553 runs faster than humanly readable code.
1555 You can transform humanly readable code into byte compiled code by
1556 running one of the compile commands such as @code{byte-compile-file}.
1557 Byte compiled code is usually stored in a file that ends with a
1558 @file{.elc} extension rather than a @file{.el} extension. You will
1559 see both kinds of file in the @file{emacs/lisp} directory; the files
1560 to read are those with @file{.el} extensions.
1562 As a practical matter, for most things you might do to customize or
1563 extend Emacs, you do not need to byte compile; and I will not discuss
1564 the topic here. @xref{Byte Compilation, , Byte Compilation, elisp,
1565 The GNU Emacs Lisp Reference Manual}, for a full description of byte
1572 When the Lisp interpreter works on an expression, the term for the
1573 activity is called @dfn{evaluation}. We say that the interpreter
1574 `evaluates the expression'. I've used this term several times before.
1575 The word comes from its use in everyday language, `to ascertain the
1576 value or amount of; to appraise', according to @cite{Webster's New
1577 Collegiate Dictionary}.
1580 * How the Interpreter Acts:: Returns and Side Effects...
1581 * Evaluating Inner Lists:: Lists within lists...
1585 @node How the Interpreter Acts
1586 @unnumberedsubsec How the Lisp Interpreter Acts
1589 @cindex @samp{returned value} explained
1590 After evaluating an expression, the Lisp interpreter will most likely
1591 @dfn{return} the value that the computer produces by carrying out the
1592 instructions it found in the function definition, or perhaps it will
1593 give up on that function and produce an error message. (The interpreter
1594 may also find itself tossed, so to speak, to a different function or it
1595 may attempt to repeat continually what it is doing for ever and ever in
1596 what is called an `infinite loop'. These actions are less common; and
1597 we can ignore them.) Most frequently, the interpreter returns a value.
1599 @cindex @samp{side effect} defined
1600 At the same time the interpreter returns a value, it may do something
1601 else as well, such as move a cursor or copy a file; this other kind of
1602 action is called a @dfn{side effect}. Actions that we humans think are
1603 important, such as printing results, are often ``side effects'' to the
1604 Lisp interpreter. The jargon can sound peculiar, but it turns out that
1605 it is fairly easy to learn to use side effects.
1607 In summary, evaluating a symbolic expression most commonly causes the
1608 Lisp interpreter to return a value and perhaps carry out a side effect;
1609 or else produce an error.
1611 @node Evaluating Inner Lists
1612 @subsection Evaluating Inner Lists
1613 @cindex Inner list evaluation
1614 @cindex Evaluating inner lists
1616 If evaluation applies to a list that is inside another list, the outer
1617 list may use the value returned by the first evaluation as information
1618 when the outer list is evaluated. This explains why inner expressions
1619 are evaluated first: the values they return are used by the outer
1623 We can investigate this process by evaluating another addition example.
1624 Place your cursor after the following expression and type @kbd{C-x C-e}:
1631 The number 8 will appear in the echo area.
1633 What happens is that the Lisp interpreter first evaluates the inner
1634 expression, @code{(+ 3 3)}, for which the value 6 is returned; then it
1635 evaluates the outer expression as if it were written @code{(+ 2 6)}, which
1636 returns the value 8. Since there are no more enclosing expressions to
1637 evaluate, the interpreter prints that value in the echo area.
1639 Now it is easy to understand the name of the command invoked by the
1640 keystrokes @kbd{C-x C-e}: the name is @code{eval-last-sexp}. The
1641 letters @code{sexp} are an abbreviation for `symbolic expression', and
1642 @code{eval} is an abbreviation for `evaluate'. The command means
1643 `evaluate last symbolic expression'.
1645 As an experiment, you can try evaluating the expression by putting the
1646 cursor at the beginning of the next line immediately following the
1647 expression, or inside the expression.
1650 Here is another copy of the expression:
1657 If you place the cursor at the beginning of the blank line that
1658 immediately follows the expression and type @kbd{C-x C-e}, you will
1659 still get the value 8 printed in the echo area. Now try putting the
1660 cursor inside the expression. If you put it right after the next to
1661 last parenthesis (so it appears to sit on top of the last parenthesis),
1662 you will get a 6 printed in the echo area! This is because the command
1663 evaluates the expression @code{(+ 3 3)}.
1665 Now put the cursor immediately after a number. Type @kbd{C-x C-e} and
1666 you will get the number itself. In Lisp, if you evaluate a number, you
1667 get the number itself---this is how numbers differ from symbols. If you
1668 evaluate a list starting with a symbol like @code{+}, you will get a
1669 value returned that is the result of the computer carrying out the
1670 instructions in the function definition attached to that name. If a
1671 symbol by itself is evaluated, something different happens, as we will
1672 see in the next section.
1678 In Emacs Lisp, a symbol can have a value attached to it just as it can
1679 have a function definition attached to it. The two are different.
1680 The function definition is a set of instructions that a computer will
1681 obey. A value, on the other hand, is something, such as number or a
1682 name, that can vary (which is why such a symbol is called a variable).
1683 The value of a symbol can be any expression in Lisp, such as a symbol,
1684 number, list, or string. A symbol that has a value is often called a
1687 A symbol can have both a function definition and a value attached to
1688 it at the same time. Or it can have just one or the other.
1689 The two are separate. This is somewhat similar
1690 to the way the name Cambridge can refer to the city in Massachusetts
1691 and have some information attached to the name as well, such as
1692 ``great programming center''.
1695 (Incidentally, in Emacs Lisp, a symbol can have two
1696 other things attached to it, too: a property list and a documentation
1697 string; these are discussed later.)
1700 Another way to think about this is to imagine a symbol as being a chest
1701 of drawers. The function definition is put in one drawer, the value in
1702 another, and so on. What is put in the drawer holding the value can be
1703 changed without affecting the contents of the drawer holding the
1704 function definition, and vice-verse.
1707 * fill-column Example::
1708 * Void Function:: The error message for a symbol
1710 * Void Variable:: The error message for a symbol without a value.
1714 @node fill-column Example
1715 @unnumberedsubsec @code{fill-column}, an Example Variable
1718 @findex fill-column, @r{an example variable}
1719 @cindex Example variable, @code{fill-column}
1720 @cindex Variable, example of, @code{fill-column}
1721 The variable @code{fill-column} illustrates a symbol with a value
1722 attached to it: in every GNU Emacs buffer, this symbol is set to some
1723 value, usually 72 or 70, but sometimes to some other value. To find the
1724 value of this symbol, evaluate it by itself. If you are reading this in
1725 Info inside of GNU Emacs, you can do this by putting the cursor after
1726 the symbol and typing @kbd{C-x C-e}:
1733 After I typed @kbd{C-x C-e}, Emacs printed the number 72 in my echo
1734 area. This is the value for which @code{fill-column} is set for me as I
1735 write this. It may be different for you in your Info buffer. Notice
1736 that the value returned as a variable is printed in exactly the same way
1737 as the value returned by a function carrying out its instructions. From
1738 the point of view of the Lisp interpreter, a value returned is a value
1739 returned. What kind of expression it came from ceases to matter once
1742 A symbol can have any value attached to it or, to use the jargon, we can
1743 @dfn{bind} the variable to a value: to a number, such as 72; to a
1744 string, @code{"such as this"}; to a list, such as @code{(spruce pine
1745 oak)}; we can even bind a variable to a function definition.
1747 A symbol can be bound to a value in several ways. @xref{set & setq, ,
1748 Setting the Value of a Variable}, for information about one way to do
1752 @subsection Error Message for a Symbol Without a Function
1753 @cindex Symbol without function error
1754 @cindex Error for symbol without function
1756 When we evaluated @code{fill-column} to find its value as a variable,
1757 we did not place parentheses around the word. This is because we did
1758 not intend to use it as a function name.
1760 If @code{fill-column} were the first or only element of a list, the
1761 Lisp interpreter would attempt to find the function definition
1762 attached to it. But @code{fill-column} has no function definition.
1763 Try evaluating this:
1771 You will create a @file{*Backtrace*} buffer that says:
1775 ---------- Buffer: *Backtrace* ----------
1776 Debugger entered--Lisp error: (void-function fill-column)
1779 eval-last-sexp-1(nil)
1781 call-interactively(eval-last-sexp)
1782 ---------- Buffer: *Backtrace* ----------
1787 (Remember, to quit the debugger and make the debugger window go away,
1788 type @kbd{q} in the @file{*Backtrace*} buffer.)
1792 In GNU Emacs 20 and before, you will produce an error message that says:
1795 Symbol's function definition is void:@: fill-column
1799 (The message will go away as soon as you move the cursor or type
1804 @subsection Error Message for a Symbol Without a Value
1805 @cindex Symbol without value error
1806 @cindex Error for symbol without value
1808 If you attempt to evaluate a symbol that does not have a value bound to
1809 it, you will receive an error message. You can see this by
1810 experimenting with our 2 plus 2 addition. In the following expression,
1811 put your cursor right after the @code{+}, before the first number 2,
1820 In GNU Emacs 22, you will create a @file{*Backtrace*} buffer that
1825 ---------- Buffer: *Backtrace* ----------
1826 Debugger entered--Lisp error: (void-variable +)
1828 eval-last-sexp-1(nil)
1830 call-interactively(eval-last-sexp)
1831 ---------- Buffer: *Backtrace* ----------
1836 (Again, you can quit the debugger by
1837 typing @kbd{q} in the @file{*Backtrace*} buffer.)
1839 This backtrace is different from the very first error message we saw,
1840 which said, @samp{Debugger entered--Lisp error: (void-function this)}.
1841 In this case, the function does not have a value as a variable; while
1842 in the other error message, the function (the word `this') did not
1845 In this experiment with the @code{+}, what we did was cause the Lisp
1846 interpreter to evaluate the @code{+} and look for the value of the
1847 variable instead of the function definition. We did this by placing the
1848 cursor right after the symbol rather than after the parenthesis of the
1849 enclosing list as we did before. As a consequence, the Lisp interpreter
1850 evaluated the preceding s-expression, which in this case was
1853 Since @code{+} does not have a value bound to it, just the function
1854 definition, the error message reported that the symbol's value as a
1859 In GNU Emacs version 20 and before, your error message will say:
1862 Symbol's value as variable is void:@: +
1866 The meaning is the same as in GNU Emacs 22.
1872 @cindex Passing information to functions
1874 To see how information is passed to functions, let's look again at
1875 our old standby, the addition of two plus two. In Lisp, this is written
1882 If you evaluate this expression, the number 4 will appear in your echo
1883 area. What the Lisp interpreter does is add the numbers that follow
1886 @cindex @samp{argument} defined
1887 The numbers added by @code{+} are called the @dfn{arguments} of the
1888 function @code{+}. These numbers are the information that is given to
1889 or @dfn{passed} to the function.
1891 The word `argument' comes from the way it is used in mathematics and
1892 does not refer to a disputation between two people; instead it refers to
1893 the information presented to the function, in this case, to the
1894 @code{+}. In Lisp, the arguments to a function are the atoms or lists
1895 that follow the function. The values returned by the evaluation of
1896 these atoms or lists are passed to the function. Different functions
1897 require different numbers of arguments; some functions require none at
1898 all.@footnote{It is curious to track the path by which the word `argument'
1899 came to have two different meanings, one in mathematics and the other in
1900 everyday English. According to the @cite{Oxford English Dictionary},
1901 the word derives from the Latin for @samp{to make clear, prove}; thus it
1902 came to mean, by one thread of derivation, `the evidence offered as
1903 proof', which is to say, `the information offered', which led to its
1904 meaning in Lisp. But in the other thread of derivation, it came to mean
1905 `to assert in a manner against which others may make counter
1906 assertions', which led to the meaning of the word as a disputation.
1907 (Note here that the English word has two different definitions attached
1908 to it at the same time. By contrast, in Emacs Lisp, a symbol cannot
1909 have two different function definitions at the same time.)}
1912 * Data types:: Types of data passed to a function.
1913 * Args as Variable or List:: An argument can be the value
1914 of a variable or list.
1915 * Variable Number of Arguments:: Some functions may take a
1916 variable number of arguments.
1917 * Wrong Type of Argument:: Passing an argument of the wrong type
1919 * message:: A useful function for sending messages.
1923 @subsection Arguments' Data Types
1925 @cindex Types of data
1926 @cindex Arguments' data types
1928 The type of data that should be passed to a function depends on what
1929 kind of information it uses. The arguments to a function such as
1930 @code{+} must have values that are numbers, since @code{+} adds numbers.
1931 Other functions use different kinds of data for their arguments.
1935 For example, the @code{concat} function links together or unites two or
1936 more strings of text to produce a string. The arguments are strings.
1937 Concatenating the two character strings @code{abc}, @code{def} produces
1938 the single string @code{abcdef}. This can be seen by evaluating the
1942 (concat "abc" "def")
1946 The value produced by evaluating this expression is @code{"abcdef"}.
1948 A function such as @code{substring} uses both a string and numbers as
1949 arguments. The function returns a part of the string, a substring of
1950 the first argument. This function takes three arguments. Its first
1951 argument is the string of characters, the second and third arguments are
1952 numbers that indicate the beginning and end of the substring. The
1953 numbers are a count of the number of characters (including spaces and
1954 punctuation) from the beginning of the string.
1957 For example, if you evaluate the following:
1960 (substring "The quick brown fox jumped." 16 19)
1964 you will see @code{"fox"} appear in the echo area. The arguments are the
1965 string and the two numbers.
1967 Note that the string passed to @code{substring} is a single atom even
1968 though it is made up of several words separated by spaces. Lisp counts
1969 everything between the two quotation marks as part of the string,
1970 including the spaces. You can think of the @code{substring} function as
1971 a kind of `atom smasher' since it takes an otherwise indivisible atom
1972 and extracts a part. However, @code{substring} is only able to extract
1973 a substring from an argument that is a string, not from another type of
1974 atom such as a number or symbol.
1976 @node Args as Variable or List
1977 @subsection An Argument as the Value of a Variable or List
1979 An argument can be a symbol that returns a value when it is evaluated.
1980 For example, when the symbol @code{fill-column} by itself is evaluated,
1981 it returns a number. This number can be used in an addition.
1984 Position the cursor after the following expression and type @kbd{C-x
1992 The value will be a number two more than what you get by evaluating
1993 @code{fill-column} alone. For me, this is 74, because my value of
1994 @code{fill-column} is 72.
1996 As we have just seen, an argument can be a symbol that returns a value
1997 when evaluated. In addition, an argument can be a list that returns a
1998 value when it is evaluated. For example, in the following expression,
1999 the arguments to the function @code{concat} are the strings
2000 @w{@code{"The "}} and @w{@code{" red foxes."}} and the list
2001 @code{(number-to-string (+ 2 fill-column))}.
2003 @c For GNU Emacs 22, need number-to-string
2005 (concat "The " (number-to-string (+ 2 fill-column)) " red foxes.")
2009 If you evaluate this expression---and if, as with my Emacs,
2010 @code{fill-column} evaluates to 72---@code{"The 74 red foxes."} will
2011 appear in the echo area. (Note that you must put spaces after the
2012 word @samp{The} and before the word @samp{red} so they will appear in
2013 the final string. The function @code{number-to-string} converts the
2014 integer that the addition function returns to a string.
2015 @code{number-to-string} is also known as @code{int-to-string}.)
2017 @node Variable Number of Arguments
2018 @subsection Variable Number of Arguments
2019 @cindex Variable number of arguments
2020 @cindex Arguments, variable number of
2022 Some functions, such as @code{concat}, @code{+} or @code{*}, take any
2023 number of arguments. (The @code{*} is the symbol for multiplication.)
2024 This can be seen by evaluating each of the following expressions in
2025 the usual way. What you will see in the echo area is printed in this
2026 text after @samp{@result{}}, which you may read as `evaluates to'.
2029 In the first set, the functions have no arguments:
2040 In this set, the functions have one argument each:
2051 In this set, the functions have three arguments each:
2055 (+ 3 4 5) @result{} 12
2057 (* 3 4 5) @result{} 60
2061 @node Wrong Type of Argument
2062 @subsection Using the Wrong Type Object as an Argument
2063 @cindex Wrong type of argument
2064 @cindex Argument, wrong type of
2066 When a function is passed an argument of the wrong type, the Lisp
2067 interpreter produces an error message. For example, the @code{+}
2068 function expects the values of its arguments to be numbers. As an
2069 experiment we can pass it the quoted symbol @code{hello} instead of a
2070 number. Position the cursor after the following expression and type
2078 When you do this you will generate an error message. What has happened
2079 is that @code{+} has tried to add the 2 to the value returned by
2080 @code{'hello}, but the value returned by @code{'hello} is the symbol
2081 @code{hello}, not a number. Only numbers can be added. So @code{+}
2082 could not carry out its addition.
2085 You will create and enter a @file{*Backtrace*} buffer that says:
2090 ---------- Buffer: *Backtrace* ----------
2091 Debugger entered--Lisp error:
2092 (wrong-type-argument number-or-marker-p hello)
2094 eval((+ 2 (quote hello)))
2095 eval-last-sexp-1(nil)
2097 call-interactively(eval-last-sexp)
2098 ---------- Buffer: *Backtrace* ----------
2103 As usual, the error message tries to be helpful and makes sense after you
2104 learn how to read it.@footnote{@code{(quote hello)} is an expansion of
2105 the abbreviation @code{'hello}.}
2107 The first part of the error message is straightforward; it says
2108 @samp{wrong type argument}. Next comes the mysterious jargon word
2109 @w{@samp{number-or-marker-p}}. This word is trying to tell you what
2110 kind of argument the @code{+} expected.
2112 The symbol @code{number-or-marker-p} says that the Lisp interpreter is
2113 trying to determine whether the information presented it (the value of
2114 the argument) is a number or a marker (a special object representing a
2115 buffer position). What it does is test to see whether the @code{+} is
2116 being given numbers to add. It also tests to see whether the
2117 argument is something called a marker, which is a specific feature of
2118 Emacs Lisp. (In Emacs, locations in a buffer are recorded as markers.
2119 When the mark is set with the @kbd{C-@@} or @kbd{C-@key{SPC}} command,
2120 its position is kept as a marker. The mark can be considered a
2121 number---the number of characters the location is from the beginning
2122 of the buffer.) In Emacs Lisp, @code{+} can be used to add the
2123 numeric value of marker positions as numbers.
2125 The @samp{p} of @code{number-or-marker-p} is the embodiment of a
2126 practice started in the early days of Lisp programming. The @samp{p}
2127 stands for `predicate'. In the jargon used by the early Lisp
2128 researchers, a predicate refers to a function to determine whether some
2129 property is true or false. So the @samp{p} tells us that
2130 @code{number-or-marker-p} is the name of a function that determines
2131 whether it is true or false that the argument supplied is a number or
2132 a marker. Other Lisp symbols that end in @samp{p} include @code{zerop},
2133 a function that tests whether its argument has the value of zero, and
2134 @code{listp}, a function that tests whether its argument is a list.
2136 Finally, the last part of the error message is the symbol @code{hello}.
2137 This is the value of the argument that was passed to @code{+}. If the
2138 addition had been passed the correct type of object, the value passed
2139 would have been a number, such as 37, rather than a symbol like
2140 @code{hello}. But then you would not have got the error message.
2144 In GNU Emacs version 20 and before, the echo area displays an error
2148 Wrong type argument:@: number-or-marker-p, hello
2151 This says, in different words, the same as the top line of the
2152 @file{*Backtrace*} buffer.
2156 @subsection The @code{message} Function
2159 Like @code{+}, the @code{message} function takes a variable number of
2160 arguments. It is used to send messages to the user and is so useful
2161 that we will describe it here.
2164 A message is printed in the echo area. For example, you can print a
2165 message in your echo area by evaluating the following list:
2168 (message "This message appears in the echo area!")
2171 The whole string between double quotation marks is a single argument
2172 and is printed @i{in toto}. (Note that in this example, the message
2173 itself will appear in the echo area within double quotes; that is
2174 because you see the value returned by the @code{message} function. In
2175 most uses of @code{message} in programs that you write, the text will
2176 be printed in the echo area as a side-effect, without the quotes.
2177 @xref{multiply-by-seven in detail, , @code{multiply-by-seven} in
2178 detail}, for an example of this.)
2180 However, if there is a @samp{%s} in the quoted string of characters, the
2181 @code{message} function does not print the @samp{%s} as such, but looks
2182 to the argument that follows the string. It evaluates the second
2183 argument and prints the value at the location in the string where the
2187 You can see this by positioning the cursor after the following
2188 expression and typing @kbd{C-x C-e}:
2191 (message "The name of this buffer is: %s." (buffer-name))
2195 In Info, @code{"The name of this buffer is: *info*."} will appear in the
2196 echo area. The function @code{buffer-name} returns the name of the
2197 buffer as a string, which the @code{message} function inserts in place
2200 To print a value as an integer, use @samp{%d} in the same way as
2201 @samp{%s}. For example, to print a message in the echo area that
2202 states the value of the @code{fill-column}, evaluate the following:
2205 (message "The value of fill-column is %d." fill-column)
2209 On my system, when I evaluate this list, @code{"The value of
2210 fill-column is 72."} appears in my echo area@footnote{Actually, you
2211 can use @code{%s} to print a number. It is non-specific. @code{%d}
2212 prints only the part of a number left of a decimal point, and not
2213 anything that is not a number.}.
2215 If there is more than one @samp{%s} in the quoted string, the value of
2216 the first argument following the quoted string is printed at the
2217 location of the first @samp{%s} and the value of the second argument is
2218 printed at the location of the second @samp{%s}, and so on.
2221 For example, if you evaluate the following,
2225 (message "There are %d %s in the office!"
2226 (- fill-column 14) "pink elephants")
2231 a rather whimsical message will appear in your echo area. On my system
2232 it says, @code{"There are 58 pink elephants in the office!"}.
2234 The expression @code{(- fill-column 14)} is evaluated and the resulting
2235 number is inserted in place of the @samp{%d}; and the string in double
2236 quotes, @code{"pink elephants"}, is treated as a single argument and
2237 inserted in place of the @samp{%s}. (That is to say, a string between
2238 double quotes evaluates to itself, like a number.)
2240 Finally, here is a somewhat complex example that not only illustrates
2241 the computation of a number, but also shows how you can use an
2242 expression within an expression to generate the text that is substituted
2247 (message "He saw %d %s"
2251 "The quick brown foxes jumped." 16 21)
2256 In this example, @code{message} has three arguments: the string,
2257 @code{"He saw %d %s"}, the expression, @code{(- fill-column 32)}, and
2258 the expression beginning with the function @code{concat}. The value
2259 resulting from the evaluation of @code{(- fill-column 32)} is inserted
2260 in place of the @samp{%d}; and the value returned by the expression
2261 beginning with @code{concat} is inserted in place of the @samp{%s}.
2263 When your fill column is 70 and you evaluate the expression, the
2264 message @code{"He saw 38 red foxes leaping."} appears in your echo
2268 @section Setting the Value of a Variable
2269 @cindex Variable, setting value
2270 @cindex Setting value of variable
2272 @cindex @samp{bind} defined
2273 There are several ways by which a variable can be given a value. One of
2274 the ways is to use either the function @code{set} or the function
2275 @code{setq}. Another way is to use @code{let} (@pxref{let}). (The
2276 jargon for this process is to @dfn{bind} a variable to a value.)
2278 The following sections not only describe how @code{set} and @code{setq}
2279 work but also illustrate how arguments are passed.
2282 * Using set:: Setting values.
2283 * Using setq:: Setting a quoted value.
2284 * Counting:: Using @code{setq} to count.
2288 @subsection Using @code{set}
2291 To set the value of the symbol @code{flowers} to the list @code{'(rose
2292 violet daisy buttercup)}, evaluate the following expression by
2293 positioning the cursor after the expression and typing @kbd{C-x C-e}.
2296 (set 'flowers '(rose violet daisy buttercup))
2300 The list @code{(rose violet daisy buttercup)} will appear in the echo
2301 area. This is what is @emph{returned} by the @code{set} function. As a
2302 side effect, the symbol @code{flowers} is bound to the list; that is,
2303 the symbol @code{flowers}, which can be viewed as a variable, is given
2304 the list as its value. (This process, by the way, illustrates how a
2305 side effect to the Lisp interpreter, setting the value, can be the
2306 primary effect that we humans are interested in. This is because every
2307 Lisp function must return a value if it does not get an error, but it
2308 will only have a side effect if it is designed to have one.)
2310 After evaluating the @code{set} expression, you can evaluate the symbol
2311 @code{flowers} and it will return the value you just set. Here is the
2312 symbol. Place your cursor after it and type @kbd{C-x C-e}.
2319 When you evaluate @code{flowers}, the list
2320 @code{(rose violet daisy buttercup)} appears in the echo area.
2322 Incidentally, if you evaluate @code{'flowers}, the variable with a quote
2323 in front of it, what you will see in the echo area is the symbol itself,
2324 @code{flowers}. Here is the quoted symbol, so you can try this:
2330 Note also, that when you use @code{set}, you need to quote both
2331 arguments to @code{set}, unless you want them evaluated. Since we do
2332 not want either argument evaluated, neither the variable
2333 @code{flowers} nor the list @code{(rose violet daisy buttercup)}, both
2334 are quoted. (When you use @code{set} without quoting its first
2335 argument, the first argument is evaluated before anything else is
2336 done. If you did this and @code{flowers} did not have a value
2337 already, you would get an error message that the @samp{Symbol's value
2338 as variable is void}; on the other hand, if @code{flowers} did return
2339 a value after it was evaluated, the @code{set} would attempt to set
2340 the value that was returned. There are situations where this is the
2341 right thing for the function to do; but such situations are rare.)
2344 @subsection Using @code{setq}
2347 As a practical matter, you almost always quote the first argument to
2348 @code{set}. The combination of @code{set} and a quoted first argument
2349 is so common that it has its own name: the special form @code{setq}.
2350 This special form is just like @code{set} except that the first argument
2351 is quoted automatically, so you don't need to type the quote mark
2352 yourself. Also, as an added convenience, @code{setq} permits you to set
2353 several different variables to different values, all in one expression.
2355 To set the value of the variable @code{carnivores} to the list
2356 @code{'(lion tiger leopard)} using @code{setq}, the following expression
2360 (setq carnivores '(lion tiger leopard))
2364 This is exactly the same as using @code{set} except the first argument
2365 is automatically quoted by @code{setq}. (The @samp{q} in @code{setq}
2366 means @code{quote}.)
2369 With @code{set}, the expression would look like this:
2372 (set 'carnivores '(lion tiger leopard))
2375 Also, @code{setq} can be used to assign different values to
2376 different variables. The first argument is bound to the value
2377 of the second argument, the third argument is bound to the value of the
2378 fourth argument, and so on. For example, you could use the following to
2379 assign a list of trees to the symbol @code{trees} and a list of herbivores
2380 to the symbol @code{herbivores}:
2384 (setq trees '(pine fir oak maple)
2385 herbivores '(gazelle antelope zebra))
2390 (The expression could just as well have been on one line, but it might
2391 not have fit on a page; and humans find it easier to read nicely
2394 Although I have been using the term `assign', there is another way of
2395 thinking about the workings of @code{set} and @code{setq}; and that is to
2396 say that @code{set} and @code{setq} make the symbol @emph{point} to the
2397 list. This latter way of thinking is very common and in forthcoming
2398 chapters we shall come upon at least one symbol that has `pointer' as
2399 part of its name. The name is chosen because the symbol has a value,
2400 specifically a list, attached to it; or, expressed another way,
2401 the symbol is set to ``point'' to the list.
2404 @subsection Counting
2407 Here is an example that shows how to use @code{setq} in a counter. You
2408 might use this to count how many times a part of your program repeats
2409 itself. First set a variable to zero; then add one to the number each
2410 time the program repeats itself. To do this, you need a variable that
2411 serves as a counter, and two expressions: an initial @code{setq}
2412 expression that sets the counter variable to zero; and a second
2413 @code{setq} expression that increments the counter each time it is
2418 (setq counter 0) ; @r{Let's call this the initializer.}
2420 (setq counter (+ counter 1)) ; @r{This is the incrementer.}
2422 counter ; @r{This is the counter.}
2427 (The text following the @samp{;} are comments. @xref{Change a
2428 defun, , Change a Function Definition}.)
2430 If you evaluate the first of these expressions, the initializer,
2431 @code{(setq counter 0)}, and then evaluate the third expression,
2432 @code{counter}, the number @code{0} will appear in the echo area. If
2433 you then evaluate the second expression, the incrementer, @code{(setq
2434 counter (+ counter 1))}, the counter will get the value 1. So if you
2435 again evaluate @code{counter}, the number @code{1} will appear in the
2436 echo area. Each time you evaluate the second expression, the value of
2437 the counter will be incremented.
2439 When you evaluate the incrementer, @code{(setq counter (+ counter 1))},
2440 the Lisp interpreter first evaluates the innermost list; this is the
2441 addition. In order to evaluate this list, it must evaluate the variable
2442 @code{counter} and the number @code{1}. When it evaluates the variable
2443 @code{counter}, it receives its current value. It passes this value and
2444 the number @code{1} to the @code{+} which adds them together. The sum
2445 is then returned as the value of the inner list and passed to the
2446 @code{setq} which sets the variable @code{counter} to this new value.
2447 Thus, the value of the variable, @code{counter}, is changed.
2452 Learning Lisp is like climbing a hill in which the first part is the
2453 steepest. You have now climbed the most difficult part; what remains
2454 becomes easier as you progress onwards.
2462 Lisp programs are made up of expressions, which are lists or single atoms.
2465 Lists are made up of zero or more atoms or inner lists, separated by whitespace and
2466 surrounded by parentheses. A list can be empty.
2469 Atoms are multi-character symbols, like @code{forward-paragraph}, single
2470 character symbols like @code{+}, strings of characters between double
2471 quotation marks, or numbers.
2474 A number evaluates to itself.
2477 A string between double quotes also evaluates to itself.
2480 When you evaluate a symbol by itself, its value is returned.
2483 When you evaluate a list, the Lisp interpreter looks at the first symbol
2484 in the list and then at the function definition bound to that symbol.
2485 Then the instructions in the function definition are carried out.
2488 A single quotation mark,
2495 , tells the Lisp interpreter that it should
2496 return the following expression as written, and not evaluate it as it
2497 would if the quote were not there.
2500 Arguments are the information passed to a function. The arguments to a
2501 function are computed by evaluating the rest of the elements of the list
2502 of which the function is the first element.
2505 A function always returns a value when it is evaluated (unless it gets
2506 an error); in addition, it may also carry out some action called a
2507 ``side effect''. In many cases, a function's primary purpose is to
2508 create a side effect.
2511 @node Error Message Exercises
2514 A few simple exercises:
2518 Generate an error message by evaluating an appropriate symbol that is
2519 not within parentheses.
2522 Generate an error message by evaluating an appropriate symbol that is
2523 between parentheses.
2526 Create a counter that increments by two rather than one.
2529 Write an expression that prints a message in the echo area when
2533 @node Practicing Evaluation
2534 @chapter Practicing Evaluation
2535 @cindex Practicing evaluation
2536 @cindex Evaluation practice
2538 Before learning how to write a function definition in Emacs Lisp, it is
2539 useful to spend a little time evaluating various expressions that have
2540 already been written. These expressions will be lists with the
2541 functions as their first (and often only) element. Since some of the
2542 functions associated with buffers are both simple and interesting, we
2543 will start with those. In this section, we will evaluate a few of
2544 these. In another section, we will study the code of several other
2545 buffer-related functions, to see how they were written.
2548 * How to Evaluate:: Typing editing commands or @kbd{C-x C-e}
2550 * Buffer Names:: Buffers and files are different.
2551 * Getting Buffers:: Getting a buffer itself, not merely its name.
2552 * Switching Buffers:: How to change to another buffer.
2553 * Buffer Size & Locations:: Where point is located and the size of
2555 * Evaluation Exercise::
2559 @node How to Evaluate
2560 @unnumberedsec How to Evaluate
2563 @i{Whenever you give an editing command} to Emacs Lisp, such as the
2564 command to move the cursor or to scroll the screen, @i{you are evaluating
2565 an expression,} the first element of which is a function. @i{This is
2568 @cindex @samp{interactive function} defined
2569 @cindex @samp{command} defined
2570 When you type keys, you cause the Lisp interpreter to evaluate an
2571 expression and that is how you get your results. Even typing plain text
2572 involves evaluating an Emacs Lisp function, in this case, one that uses
2573 @code{self-insert-command}, which simply inserts the character you
2574 typed. The functions you evaluate by typing keystrokes are called
2575 @dfn{interactive} functions, or @dfn{commands}; how you make a function
2576 interactive will be illustrated in the chapter on how to write function
2577 definitions. @xref{Interactive, , Making a Function Interactive}.
2579 In addition to typing keyboard commands, we have seen a second way to
2580 evaluate an expression: by positioning the cursor after a list and
2581 typing @kbd{C-x C-e}. This is what we will do in the rest of this
2582 section. There are other ways to evaluate an expression as well; these
2583 will be described as we come to them.
2585 Besides being used for practicing evaluation, the functions shown in the
2586 next few sections are important in their own right. A study of these
2587 functions makes clear the distinction between buffers and files, how to
2588 switch to a buffer, and how to determine a location within it.
2591 @section Buffer Names
2593 @findex buffer-file-name
2595 The two functions, @code{buffer-name} and @code{buffer-file-name}, show
2596 the difference between a file and a buffer. When you evaluate the
2597 following expression, @code{(buffer-name)}, the name of the buffer
2598 appears in the echo area. When you evaluate @code{(buffer-file-name)},
2599 the name of the file to which the buffer refers appears in the echo
2600 area. Usually, the name returned by @code{(buffer-name)} is the same as
2601 the name of the file to which it refers, and the name returned by
2602 @code{(buffer-file-name)} is the full path-name of the file.
2604 A file and a buffer are two different entities. A file is information
2605 recorded permanently in the computer (unless you delete it). A buffer,
2606 on the other hand, is information inside of Emacs that will vanish at
2607 the end of the editing session (or when you kill the buffer). Usually,
2608 a buffer contains information that you have copied from a file; we say
2609 the buffer is @dfn{visiting} that file. This copy is what you work on
2610 and modify. Changes to the buffer do not change the file, until you
2611 save the buffer. When you save the buffer, the buffer is copied to the file
2612 and is thus saved permanently.
2615 If you are reading this in Info inside of GNU Emacs, you can evaluate
2616 each of the following expressions by positioning the cursor after it and
2617 typing @kbd{C-x C-e}.
2628 When I do this in Info, the value returned by evaluating
2629 @code{(buffer-name)} is @file{"*info*"}, and the value returned by
2630 evaluating @code{(buffer-file-name)} is @file{nil}.
2632 On the other hand, while I am writing this document, the value
2633 returned by evaluating @code{(buffer-name)} is
2634 @file{"introduction.texinfo"}, and the value returned by evaluating
2635 @code{(buffer-file-name)} is
2636 @file{"/gnu/work/intro/introduction.texinfo"}.
2638 @cindex @code{nil}, history of word
2639 The former is the name of the buffer and the latter is the name of the
2640 file. In Info, the buffer name is @file{"*info*"}. Info does not
2641 point to any file, so the result of evaluating
2642 @code{(buffer-file-name)} is @file{nil}. The symbol @code{nil} is
2643 from the Latin word for `nothing'; in this case, it means that the
2644 buffer is not associated with any file. (In Lisp, @code{nil} is also
2645 used to mean `false' and is a synonym for the empty list, @code{()}.)
2647 When I am writing, the name of my buffer is
2648 @file{"introduction.texinfo"}. The name of the file to which it
2649 points is @file{"/gnu/work/intro/introduction.texinfo"}.
2651 (In the expressions, the parentheses tell the Lisp interpreter to
2652 treat @w{@code{buffer-name}} and @w{@code{buffer-file-name}} as
2653 functions; without the parentheses, the interpreter would attempt to
2654 evaluate the symbols as variables. @xref{Variables}.)
2656 In spite of the distinction between files and buffers, you will often
2657 find that people refer to a file when they mean a buffer and vice-verse.
2658 Indeed, most people say, ``I am editing a file,'' rather than saying,
2659 ``I am editing a buffer which I will soon save to a file.'' It is
2660 almost always clear from context what people mean. When dealing with
2661 computer programs, however, it is important to keep the distinction in mind,
2662 since the computer is not as smart as a person.
2664 @cindex Buffer, history of word
2665 The word `buffer', by the way, comes from the meaning of the word as a
2666 cushion that deadens the force of a collision. In early computers, a
2667 buffer cushioned the interaction between files and the computer's
2668 central processing unit. The drums or tapes that held a file and the
2669 central processing unit were pieces of equipment that were very
2670 different from each other, working at their own speeds, in spurts. The
2671 buffer made it possible for them to work together effectively.
2672 Eventually, the buffer grew from being an intermediary, a temporary
2673 holding place, to being the place where work is done. This
2674 transformation is rather like that of a small seaport that grew into a
2675 great city: once it was merely the place where cargo was warehoused
2676 temporarily before being loaded onto ships; then it became a business
2677 and cultural center in its own right.
2679 Not all buffers are associated with files. For example, a
2680 @file{*scratch*} buffer does not visit any file. Similarly, a
2681 @file{*Help*} buffer is not associated with any file.
2683 In the old days, when you lacked a @file{~/.emacs} file and started an
2684 Emacs session by typing the command @code{emacs} alone, without naming
2685 any files, Emacs started with the @file{*scratch*} buffer visible.
2686 Nowadays, you will see a splash screen. You can follow one of the
2687 commands suggested on the splash screen, visit a file, or press the
2688 spacebar to reach the @file{*scratch*} buffer.
2690 If you switch to the @file{*scratch*} buffer, type
2691 @code{(buffer-name)}, position the cursor after it, and then type
2692 @kbd{C-x C-e} to evaluate the expression. The name @code{"*scratch*"}
2693 will be returned and will appear in the echo area. @code{"*scratch*"}
2694 is the name of the buffer. When you type @code{(buffer-file-name)} in
2695 the @file{*scratch*} buffer and evaluate that, @code{nil} will appear
2696 in the echo area, just as it does when you evaluate
2697 @code{(buffer-file-name)} in Info.
2699 Incidentally, if you are in the @file{*scratch*} buffer and want the
2700 value returned by an expression to appear in the @file{*scratch*}
2701 buffer itself rather than in the echo area, type @kbd{C-u C-x C-e}
2702 instead of @kbd{C-x C-e}. This causes the value returned to appear
2703 after the expression. The buffer will look like this:
2706 (buffer-name)"*scratch*"
2710 You cannot do this in Info since Info is read-only and it will not allow
2711 you to change the contents of the buffer. But you can do this in any
2712 buffer you can edit; and when you write code or documentation (such as
2713 this book), this feature is very useful.
2715 @node Getting Buffers
2716 @section Getting Buffers
2717 @findex current-buffer
2718 @findex other-buffer
2719 @cindex Getting a buffer
2721 The @code{buffer-name} function returns the @emph{name} of the buffer;
2722 to get the buffer @emph{itself}, a different function is needed: the
2723 @code{current-buffer} function. If you use this function in code, what
2724 you get is the buffer itself.
2726 A name and the object or entity to which the name refers are different
2727 from each other. You are not your name. You are a person to whom
2728 others refer by name. If you ask to speak to George and someone hands you
2729 a card with the letters @samp{G}, @samp{e}, @samp{o}, @samp{r},
2730 @samp{g}, and @samp{e} written on it, you might be amused, but you would
2731 not be satisfied. You do not want to speak to the name, but to the
2732 person to whom the name refers. A buffer is similar: the name of the
2733 scratch buffer is @file{*scratch*}, but the name is not the buffer. To
2734 get a buffer itself, you need to use a function such as
2735 @code{current-buffer}.
2737 However, there is a slight complication: if you evaluate
2738 @code{current-buffer} in an expression on its own, as we will do here,
2739 what you see is a printed representation of the name of the buffer
2740 without the contents of the buffer. Emacs works this way for two
2741 reasons: the buffer may be thousands of lines long---too long to be
2742 conveniently displayed; and, another buffer may have the same contents
2743 but a different name, and it is important to distinguish between them.
2746 Here is an expression containing the function:
2753 If you evaluate this expression in Info in Emacs in the usual way,
2754 @file{#<buffer *info*>} will appear in the echo area. The special
2755 format indicates that the buffer itself is being returned, rather than
2758 Incidentally, while you can type a number or symbol into a program, you
2759 cannot do that with the printed representation of a buffer: the only way
2760 to get a buffer itself is with a function such as @code{current-buffer}.
2762 A related function is @code{other-buffer}. This returns the most
2763 recently selected buffer other than the one you are in currently, not
2764 a printed representation of its name. If you have recently switched
2765 back and forth from the @file{*scratch*} buffer, @code{other-buffer}
2766 will return that buffer.
2769 You can see this by evaluating the expression:
2776 You should see @file{#<buffer *scratch*>} appear in the echo area, or
2777 the name of whatever other buffer you switched back from most
2778 recently@footnote{Actually, by default, if the buffer from which you
2779 just switched is visible to you in another window, @code{other-buffer}
2780 will choose the most recent buffer that you cannot see; this is a
2781 subtlety that I often forget.}.
2783 @node Switching Buffers
2784 @section Switching Buffers
2785 @findex switch-to-buffer
2787 @cindex Switching to a buffer
2789 The @code{other-buffer} function actually provides a buffer when it is
2790 used as an argument to a function that requires one. We can see this
2791 by using @code{other-buffer} and @code{switch-to-buffer} to switch to a
2794 But first, a brief introduction to the @code{switch-to-buffer}
2795 function. When you switched back and forth from Info to the
2796 @file{*scratch*} buffer to evaluate @code{(buffer-name)}, you most
2797 likely typed @kbd{C-x b} and then typed @file{*scratch*}@footnote{Or
2798 rather, to save typing, you probably only typed @kbd{RET} if the
2799 default buffer was @file{*scratch*}, or if it was different, then you
2800 typed just part of the name, such as @code{*sc}, pressed your
2801 @kbd{TAB} key to cause it to expand to the full name, and then typed
2802 @kbd{RET}.} when prompted in the minibuffer for the name of
2803 the buffer to which you wanted to switch. The keystrokes, @kbd{C-x
2804 b}, cause the Lisp interpreter to evaluate the interactive function
2805 @code{switch-to-buffer}. As we said before, this is how Emacs works:
2806 different keystrokes call or run different functions. For example,
2807 @kbd{C-f} calls @code{forward-char}, @kbd{M-e} calls
2808 @code{forward-sentence}, and so on.
2810 By writing @code{switch-to-buffer} in an expression, and giving it a
2811 buffer to switch to, we can switch buffers just the way @kbd{C-x b}
2815 (switch-to-buffer (other-buffer))
2819 The symbol @code{switch-to-buffer} is the first element of the list,
2820 so the Lisp interpreter will treat it as a function and carry out the
2821 instructions that are attached to it. But before doing that, the
2822 interpreter will note that @code{other-buffer} is inside parentheses
2823 and work on that symbol first. @code{other-buffer} is the first (and
2824 in this case, the only) element of this list, so the Lisp interpreter
2825 calls or runs the function. It returns another buffer. Next, the
2826 interpreter runs @code{switch-to-buffer}, passing to it, as an
2827 argument, the other buffer, which is what Emacs will switch to. If
2828 you are reading this in Info, try this now. Evaluate the expression.
2829 (To get back, type @kbd{C-x b @key{RET}}.)@footnote{Remember, this
2830 expression will move you to your most recent other buffer that you
2831 cannot see. If you really want to go to your most recently selected
2832 buffer, even if you can still see it, you need to evaluate the
2833 following more complex expression:
2836 (switch-to-buffer (other-buffer (current-buffer) t))
2840 In this case, the first argument to @code{other-buffer} tells it which
2841 buffer to skip---the current one---and the second argument tells
2842 @code{other-buffer} it is OK to switch to a visible buffer.
2843 In regular use, @code{switch-to-buffer} takes you to an invisible
2844 window since you would most likely use @kbd{C-x o} (@code{other-window})
2845 to go to another visible buffer.}
2847 In the programming examples in later sections of this document, you will
2848 see the function @code{set-buffer} more often than
2849 @code{switch-to-buffer}. This is because of a difference between
2850 computer programs and humans: humans have eyes and expect to see the
2851 buffer on which they are working on their computer terminals. This is
2852 so obvious, it almost goes without saying. However, programs do not
2853 have eyes. When a computer program works on a buffer, that buffer does
2854 not need to be visible on the screen.
2856 @code{switch-to-buffer} is designed for humans and does two different
2857 things: it switches the buffer to which Emacs's attention is directed; and
2858 it switches the buffer displayed in the window to the new buffer.
2859 @code{set-buffer}, on the other hand, does only one thing: it switches
2860 the attention of the computer program to a different buffer. The buffer
2861 on the screen remains unchanged (of course, normally nothing happens
2862 there until the command finishes running).
2864 @cindex @samp{call} defined
2865 Also, we have just introduced another jargon term, the word @dfn{call}.
2866 When you evaluate a list in which the first symbol is a function, you
2867 are calling that function. The use of the term comes from the notion of
2868 the function as an entity that can do something for you if you `call'
2869 it---just as a plumber is an entity who can fix a leak if you call him
2872 @node Buffer Size & Locations
2873 @section Buffer Size and the Location of Point
2874 @cindex Size of buffer
2876 @cindex Point location
2877 @cindex Location of point
2879 Finally, let's look at several rather simple functions,
2880 @code{buffer-size}, @code{point}, @code{point-min}, and
2881 @code{point-max}. These give information about the size of a buffer and
2882 the location of point within it.
2884 The function @code{buffer-size} tells you the size of the current
2885 buffer; that is, the function returns a count of the number of
2886 characters in the buffer.
2893 You can evaluate this in the usual way, by positioning the
2894 cursor after the expression and typing @kbd{C-x C-e}.
2896 @cindex @samp{point} defined
2897 In Emacs, the current position of the cursor is called @dfn{point}.
2898 The expression @code{(point)} returns a number that tells you where the
2899 cursor is located as a count of the number of characters from the
2900 beginning of the buffer up to point.
2903 You can see the character count for point in this buffer by evaluating
2904 the following expression in the usual way:
2911 As I write this, the value of @code{point} is 65724. The @code{point}
2912 function is frequently used in some of the examples later in this
2916 The value of point depends, of course, on its location within the
2917 buffer. If you evaluate point in this spot, the number will be larger:
2924 For me, the value of point in this location is 66043, which means that
2925 there are 319 characters (including spaces) between the two
2926 expressions. (Doubtless, you will see different numbers, since I will
2927 have edited this since I first evaluated point.)
2929 @cindex @samp{narrowing} defined
2930 The function @code{point-min} is somewhat similar to @code{point}, but
2931 it returns the value of the minimum permissible value of point in the
2932 current buffer. This is the number 1 unless @dfn{narrowing} is in
2933 effect. (Narrowing is a mechanism whereby you can restrict yourself,
2934 or a program, to operations on just a part of a buffer.
2935 @xref{Narrowing & Widening, , Narrowing and Widening}.) Likewise, the
2936 function @code{point-max} returns the value of the maximum permissible
2937 value of point in the current buffer.
2939 @node Evaluation Exercise
2942 Find a file with which you are working and move towards its middle.
2943 Find its buffer name, file name, length, and your position in the file.
2945 @node Writing Defuns
2946 @chapter How To Write Function Definitions
2947 @cindex Definition writing
2948 @cindex Function definition writing
2949 @cindex Writing a function definition
2951 When the Lisp interpreter evaluates a list, it looks to see whether the
2952 first symbol on the list has a function definition attached to it; or,
2953 put another way, whether the symbol points to a function definition. If
2954 it does, the computer carries out the instructions in the definition. A
2955 symbol that has a function definition is called, simply, a function
2956 (although, properly speaking, the definition is the function and the
2957 symbol refers to it.)
2960 * Primitive Functions::
2961 * defun:: The @code{defun} macro.
2962 * Install:: Install a function definition.
2963 * Interactive:: Making a function interactive.
2964 * Interactive Options:: Different options for @code{interactive}.
2965 * Permanent Installation:: Installing code permanently.
2966 * let:: Creating and initializing local variables.
2968 * else:: If--then--else expressions.
2969 * Truth & Falsehood:: What Lisp considers false and true.
2970 * save-excursion:: Keeping track of point, mark, and buffer.
2976 @node Primitive Functions
2977 @unnumberedsec An Aside about Primitive Functions
2979 @cindex Primitive functions
2980 @cindex Functions, primitive
2982 @cindex C language primitives
2983 @cindex Primitives written in C
2984 All functions are defined in terms of other functions, except for a few
2985 @dfn{primitive} functions that are written in the C programming
2986 language. When you write functions' definitions, you will write them in
2987 Emacs Lisp and use other functions as your building blocks. Some of the
2988 functions you will use will themselves be written in Emacs Lisp (perhaps
2989 by you) and some will be primitives written in C@. The primitive
2990 functions are used exactly like those written in Emacs Lisp and behave
2991 like them. They are written in C so we can easily run GNU Emacs on any
2992 computer that has sufficient power and can run C.
2994 Let me re-emphasize this: when you write code in Emacs Lisp, you do not
2995 distinguish between the use of functions written in C and the use of
2996 functions written in Emacs Lisp. The difference is irrelevant. I
2997 mention the distinction only because it is interesting to know. Indeed,
2998 unless you investigate, you won't know whether an already-written
2999 function is written in Emacs Lisp or C.
3002 @section The @code{defun} Macro
3005 @cindex @samp{function definition} defined
3006 In Lisp, a symbol such as @code{mark-whole-buffer} has code attached to
3007 it that tells the computer what to do when the function is called.
3008 This code is called the @dfn{function definition} and is created by
3009 evaluating a Lisp expression that starts with the symbol @code{defun}
3010 (which is an abbreviation for @emph{define function}).
3012 In subsequent sections, we will look at function definitions from the
3013 Emacs source code, such as @code{mark-whole-buffer}. In this section,
3014 we will describe a simple function definition so you can see how it
3015 looks. This function definition uses arithmetic because it makes for a
3016 simple example. Some people dislike examples using arithmetic; however,
3017 if you are such a person, do not despair. Hardly any of the code we
3018 will study in the remainder of this introduction involves arithmetic or
3019 mathematics. The examples mostly involve text in one way or another.
3021 A function definition has up to five parts following the word
3026 The name of the symbol to which the function definition should be
3030 A list of the arguments that will be passed to the function. If no
3031 arguments will be passed to the function, this is an empty list,
3035 Documentation describing the function. (Technically optional, but
3036 strongly recommended.)
3039 Optionally, an expression to make the function interactive so you can
3040 use it by typing @kbd{M-x} and then the name of the function; or by
3041 typing an appropriate key or keychord.
3043 @cindex @samp{body} defined
3045 The code that instructs the computer what to do: the @dfn{body} of the
3046 function definition.
3049 It is helpful to think of the five parts of a function definition as
3050 being organized in a template, with slots for each part:
3054 (defun @var{function-name} (@var{arguments}@dots{})
3055 "@var{optional-documentation}@dots{}"
3056 (interactive @var{argument-passing-info}) ; @r{optional}
3061 As an example, here is the code for a function that multiplies its
3062 argument by 7. (This example is not interactive. @xref{Interactive,
3063 , Making a Function Interactive}, for that information.)
3067 (defun multiply-by-seven (number)
3068 "Multiply NUMBER by seven."
3073 This definition begins with a parenthesis and the symbol @code{defun},
3074 followed by the name of the function.
3076 @cindex @samp{argument list} defined
3077 The name of the function is followed by a list that contains the
3078 arguments that will be passed to the function. This list is called
3079 the @dfn{argument list}. In this example, the list has only one
3080 element, the symbol, @code{number}. When the function is used, the
3081 symbol will be bound to the value that is used as the argument to the
3084 Instead of choosing the word @code{number} for the name of the argument,
3085 I could have picked any other name. For example, I could have chosen
3086 the word @code{multiplicand}. I picked the word `number' because it
3087 tells what kind of value is intended for this slot; but I could just as
3088 well have chosen the word `multiplicand' to indicate the role that the
3089 value placed in this slot will play in the workings of the function. I
3090 could have called it @code{foogle}, but that would have been a bad
3091 choice because it would not tell humans what it means. The choice of
3092 name is up to the programmer and should be chosen to make the meaning of
3095 Indeed, you can choose any name you wish for a symbol in an argument
3096 list, even the name of a symbol used in some other function: the name
3097 you use in an argument list is private to that particular definition.
3098 In that definition, the name refers to a different entity than any use
3099 of the same name outside the function definition. Suppose you have a
3100 nick-name `Shorty' in your family; when your family members refer to
3101 `Shorty', they mean you. But outside your family, in a movie, for
3102 example, the name `Shorty' refers to someone else. Because a name in an
3103 argument list is private to the function definition, you can change the
3104 value of such a symbol inside the body of a function without changing
3105 its value outside the function. The effect is similar to that produced
3106 by a @code{let} expression. (@xref{let, , @code{let}}.)
3109 Note also that we discuss the word `number' in two different ways: as a
3110 symbol that appears in the code, and as the name of something that will
3111 be replaced by a something else during the evaluation of the function.
3112 In the first case, @code{number} is a symbol, not a number; it happens
3113 that within the function, it is a variable who value is the number in
3114 question, but our primary interest in it is as a symbol. On the other
3115 hand, when we are talking about the function, our interest is that we
3116 will substitute a number for the word @var{number}. To keep this
3117 distinction clear, we use different typography for the two
3118 circumstances. When we talk about this function, or about how it works,
3119 we refer to this number by writing @var{number}. In the function
3120 itself, we refer to it by writing @code{number}.
3123 The argument list is followed by the documentation string that
3124 describes the function. This is what you see when you type
3125 @w{@kbd{C-h f}} and the name of a function. Incidentally, when you
3126 write a documentation string like this, you should make the first line
3127 a complete sentence since some commands, such as @code{apropos}, print
3128 only the first line of a multi-line documentation string. Also, you
3129 should not indent the second line of a documentation string, if you
3130 have one, because that looks odd when you use @kbd{C-h f}
3131 (@code{describe-function}). The documentation string is optional, but
3132 it is so useful, it should be included in almost every function you
3135 @findex * @r{(multiplication)}
3136 The third line of the example consists of the body of the function
3137 definition. (Most functions' definitions, of course, are longer than
3138 this.) In this function, the body is the list, @code{(* 7 number)}, which
3139 says to multiply the value of @var{number} by 7. (In Emacs Lisp,
3140 @code{*} is the function for multiplication, just as @code{+} is the
3141 function for addition.)
3143 When you use the @code{multiply-by-seven} function, the argument
3144 @code{number} evaluates to the actual number you want used. Here is an
3145 example that shows how @code{multiply-by-seven} is used; but don't try
3146 to evaluate this yet!
3149 (multiply-by-seven 3)
3153 The symbol @code{number}, specified in the function definition in the
3154 next section, is given or ``bound to'' the value 3 in the actual use of
3155 the function. Note that although @code{number} was inside parentheses
3156 in the function definition, the argument passed to the
3157 @code{multiply-by-seven} function is not in parentheses. The
3158 parentheses are written in the function definition so the computer can
3159 figure out where the argument list ends and the rest of the function
3162 If you evaluate this example, you are likely to get an error message.
3163 (Go ahead, try it!) This is because we have written the function
3164 definition, but not yet told the computer about the definition---we have
3165 not yet installed (or `loaded') the function definition in Emacs.
3166 Installing a function is the process that tells the Lisp interpreter the
3167 definition of the function. Installation is described in the next
3171 @section Install a Function Definition
3172 @cindex Install a Function Definition
3173 @cindex Definition installation
3174 @cindex Function definition installation
3176 If you are reading this inside of Info in Emacs, you can try out the
3177 @code{multiply-by-seven} function by first evaluating the function
3178 definition and then evaluating @code{(multiply-by-seven 3)}. A copy of
3179 the function definition follows. Place the cursor after the last
3180 parenthesis of the function definition and type @kbd{C-x C-e}. When you
3181 do this, @code{multiply-by-seven} will appear in the echo area. (What
3182 this means is that when a function definition is evaluated, the value it
3183 returns is the name of the defined function.) At the same time, this
3184 action installs the function definition.
3188 (defun multiply-by-seven (number)
3189 "Multiply NUMBER by seven."
3195 By evaluating this @code{defun}, you have just installed
3196 @code{multiply-by-seven} in Emacs. The function is now just as much a
3197 part of Emacs as @code{forward-word} or any other editing function you
3198 use. (@code{multiply-by-seven} will stay installed until you quit
3199 Emacs. To reload code automatically whenever you start Emacs, see
3200 @ref{Permanent Installation, , Installing Code Permanently}.)
3203 * Effect of installation::
3204 * Change a defun:: How to change a function definition.
3208 @node Effect of installation
3209 @unnumberedsubsec The effect of installation
3212 You can see the effect of installing @code{multiply-by-seven} by
3213 evaluating the following sample. Place the cursor after the following
3214 expression and type @kbd{C-x C-e}. The number 21 will appear in the
3218 (multiply-by-seven 3)
3221 If you wish, you can read the documentation for the function by typing
3222 @kbd{C-h f} (@code{describe-function}) and then the name of the
3223 function, @code{multiply-by-seven}. When you do this, a
3224 @file{*Help*} window will appear on your screen that says:
3228 multiply-by-seven is a Lisp function.
3229 (multiply-by-seven NUMBER)
3231 Multiply NUMBER by seven.
3236 (To return to a single window on your screen, type @kbd{C-x 1}.)
3238 @node Change a defun
3239 @subsection Change a Function Definition
3240 @cindex Changing a function definition
3241 @cindex Function definition, how to change
3242 @cindex Definition, how to change
3244 If you want to change the code in @code{multiply-by-seven}, just rewrite
3245 it. To install the new version in place of the old one, evaluate the
3246 function definition again. This is how you modify code in Emacs. It is
3249 As an example, you can change the @code{multiply-by-seven} function to
3250 add the number to itself seven times instead of multiplying the number
3251 by seven. It produces the same answer, but by a different path. At
3252 the same time, we will add a comment to the code; a comment is text
3253 that the Lisp interpreter ignores, but that a human reader may find
3254 useful or enlightening. The comment is that this is the ``second
3259 (defun multiply-by-seven (number) ; @r{Second version.}
3260 "Multiply NUMBER by seven."
3261 (+ number number number number number number number))
3265 @cindex Comments in Lisp code
3266 The comment follows a semicolon, @samp{;}. In Lisp, everything on a
3267 line that follows a semicolon is a comment. The end of the line is the
3268 end of the comment. To stretch a comment over two or more lines, begin
3269 each line with a semicolon.
3271 @xref{Beginning init File, , Beginning a @file{.emacs}
3272 File}, and @ref{Comments, , Comments, elisp, The GNU Emacs Lisp
3273 Reference Manual}, for more about comments.
3275 You can install this version of the @code{multiply-by-seven} function by
3276 evaluating it in the same way you evaluated the first function: place
3277 the cursor after the last parenthesis and type @kbd{C-x C-e}.
3279 In summary, this is how you write code in Emacs Lisp: you write a
3280 function; install it; test it; and then make fixes or enhancements and
3284 @section Make a Function Interactive
3285 @cindex Interactive functions
3288 You make a function interactive by placing a list that begins with
3289 the special form @code{interactive} immediately after the
3290 documentation. A user can invoke an interactive function by typing
3291 @kbd{M-x} and then the name of the function; or by typing the keys to
3292 which it is bound, for example, by typing @kbd{C-n} for
3293 @code{next-line} or @kbd{C-x h} for @code{mark-whole-buffer}.
3295 Interestingly, when you call an interactive function interactively,
3296 the value returned is not automatically displayed in the echo area.
3297 This is because you often call an interactive function for its side
3298 effects, such as moving forward by a word or line, and not for the
3299 value returned. If the returned value were displayed in the echo area
3300 each time you typed a key, it would be very distracting.
3303 * Interactive multiply-by-seven:: An overview.
3304 * multiply-by-seven in detail:: The interactive version.
3308 @node Interactive multiply-by-seven
3309 @unnumberedsubsec An Interactive @code{multiply-by-seven}, An Overview
3312 Both the use of the special form @code{interactive} and one way to
3313 display a value in the echo area can be illustrated by creating an
3314 interactive version of @code{multiply-by-seven}.
3321 (defun multiply-by-seven (number) ; @r{Interactive version.}
3322 "Multiply NUMBER by seven."
3324 (message "The result is %d" (* 7 number)))
3329 You can install this code by placing your cursor after it and typing
3330 @kbd{C-x C-e}. The name of the function will appear in your echo area.
3331 Then, you can use this code by typing @kbd{C-u} and a number and then
3332 typing @kbd{M-x multiply-by-seven} and pressing @key{RET}. The phrase
3333 @samp{The result is @dots{}} followed by the product will appear in the
3336 Speaking more generally, you invoke a function like this in either of two
3341 By typing a prefix argument that contains the number to be passed, and
3342 then typing @kbd{M-x} and the name of the function, as with
3343 @kbd{C-u 3 M-x forward-sentence}; or,
3346 By typing whatever key or keychord the function is bound to, as with
3351 Both the examples just mentioned work identically to move point forward
3352 three sentences. (Since @code{multiply-by-seven} is not bound to a key,
3353 it could not be used as an example of key binding.)
3355 (@xref{Keybindings, , Some Keybindings}, to learn how to bind a command
3358 A prefix argument is passed to an interactive function by typing the
3359 @key{META} key followed by a number, for example, @kbd{M-3 M-e}, or by
3360 typing @kbd{C-u} and then a number, for example, @kbd{C-u 3 M-e} (if you
3361 type @kbd{C-u} without a number, it defaults to 4).
3363 @node multiply-by-seven in detail
3364 @subsection An Interactive @code{multiply-by-seven}
3366 Let's look at the use of the special form @code{interactive} and then at
3367 the function @code{message} in the interactive version of
3368 @code{multiply-by-seven}. You will recall that the function definition
3373 (defun multiply-by-seven (number) ; @r{Interactive version.}
3374 "Multiply NUMBER by seven."
3376 (message "The result is %d" (* 7 number)))
3380 In this function, the expression, @code{(interactive "p")}, is a list of
3381 two elements. The @code{"p"} tells Emacs to pass the prefix argument to
3382 the function and use its value for the argument of the function.
3385 The argument will be a number. This means that the symbol
3386 @code{number} will be bound to a number in the line:
3389 (message "The result is %d" (* 7 number))
3394 For example, if your prefix argument is 5, the Lisp interpreter will
3395 evaluate the line as if it were:
3398 (message "The result is %d" (* 7 5))
3402 (If you are reading this in GNU Emacs, you can evaluate this expression
3403 yourself.) First, the interpreter will evaluate the inner list, which
3404 is @code{(* 7 5)}. This returns a value of 35. Next, it
3405 will evaluate the outer list, passing the values of the second and
3406 subsequent elements of the list to the function @code{message}.
3408 As we have seen, @code{message} is an Emacs Lisp function especially
3409 designed for sending a one line message to a user. (@xref{message, ,
3410 The @code{message} function}.) In summary, the @code{message}
3411 function prints its first argument in the echo area as is, except for
3412 occurrences of @samp{%d} or @samp{%s} (and various other %-sequences
3413 which we have not mentioned). When it sees a control sequence, the
3414 function looks to the second or subsequent arguments and prints the
3415 value of the argument in the location in the string where the control
3416 sequence is located.
3418 In the interactive @code{multiply-by-seven} function, the control string
3419 is @samp{%d}, which requires a number, and the value returned by
3420 evaluating @code{(* 7 5)} is the number 35. Consequently, the number 35
3421 is printed in place of the @samp{%d} and the message is @samp{The result
3424 (Note that when you call the function @code{multiply-by-seven}, the
3425 message is printed without quotes, but when you call @code{message}, the
3426 text is printed in double quotes. This is because the value returned by
3427 @code{message} is what appears in the echo area when you evaluate an
3428 expression whose first element is @code{message}; but when embedded in a
3429 function, @code{message} prints the text as a side effect without
3432 @node Interactive Options
3433 @section Different Options for @code{interactive}
3434 @cindex Options for @code{interactive}
3435 @cindex Interactive options
3437 In the example, @code{multiply-by-seven} used @code{"p"} as the
3438 argument to @code{interactive}. This argument told Emacs to interpret
3439 your typing either @kbd{C-u} followed by a number or @key{META}
3440 followed by a number as a command to pass that number to the function
3441 as its argument. Emacs has more than twenty characters predefined for
3442 use with @code{interactive}. In almost every case, one of these
3443 options will enable you to pass the right information interactively to
3444 a function. (@xref{Interactive Codes, , Code Characters for
3445 @code{interactive}, elisp, The GNU Emacs Lisp Reference Manual}.)
3448 Consider the function @code{zap-to-char}. Its interactive expression
3452 (interactive "p\ncZap to char: ")
3455 The first part of the argument to @code{interactive} is @samp{p}, with
3456 which you are already familiar. This argument tells Emacs to
3457 interpret a `prefix', as a number to be passed to the function. You
3458 can specify a prefix either by typing @kbd{C-u} followed by a number
3459 or by typing @key{META} followed by a number. The prefix is the
3460 number of specified characters. Thus, if your prefix is three and the
3461 specified character is @samp{x}, then you will delete all the text up
3462 to and including the third next @samp{x}. If you do not set a prefix,
3463 then you delete all the text up to and including the specified
3464 character, but no more.
3466 The @samp{c} tells the function the name of the character to which to delete.
3468 More formally, a function with two or more arguments can have
3469 information passed to each argument by adding parts to the string that
3470 follows @code{interactive}. When you do this, the information is
3471 passed to each argument in the same order it is specified in the
3472 @code{interactive} list. In the string, each part is separated from
3473 the next part by a @samp{\n}, which is a newline. For example, you
3474 can follow @samp{p} with a @samp{\n} and an @samp{cZap to char:@: }.
3475 This causes Emacs to pass the value of the prefix argument (if there
3476 is one) and the character.
3478 In this case, the function definition looks like the following, where
3479 @code{arg} and @code{char} are the symbols to which @code{interactive}
3480 binds the prefix argument and the specified character:
3484 (defun @var{name-of-function} (arg char)
3485 "@var{documentation}@dots{}"
3486 (interactive "p\ncZap to char: ")
3487 @var{body-of-function}@dots{})
3492 (The space after the colon in the prompt makes it look better when you
3493 are prompted. @xref{copy-to-buffer, , The Definition of
3494 @code{copy-to-buffer}}, for an example.)
3496 When a function does not take arguments, @code{interactive} does not
3497 require any. Such a function contains the simple expression
3498 @code{(interactive)}. The @code{mark-whole-buffer} function is like
3501 Alternatively, if the special letter-codes are not right for your
3502 application, you can pass your own arguments to @code{interactive} as
3505 @xref{append-to-buffer, , The Definition of @code{append-to-buffer}},
3506 for an example. @xref{Using Interactive, , Using @code{Interactive},
3507 elisp, The GNU Emacs Lisp Reference Manual}, for a more complete
3508 explanation about this technique.
3510 @node Permanent Installation
3511 @section Install Code Permanently
3512 @cindex Install code permanently
3513 @cindex Permanent code installation
3514 @cindex Code installation
3516 When you install a function definition by evaluating it, it will stay
3517 installed until you quit Emacs. The next time you start a new session
3518 of Emacs, the function will not be installed unless you evaluate the
3519 function definition again.
3521 At some point, you may want to have code installed automatically
3522 whenever you start a new session of Emacs. There are several ways of
3527 If you have code that is just for yourself, you can put the code for the
3528 function definition in your @file{.emacs} initialization file. When you
3529 start Emacs, your @file{.emacs} file is automatically evaluated and all
3530 the function definitions within it are installed.
3531 @xref{Emacs Initialization, , Your @file{.emacs} File}.
3534 Alternatively, you can put the function definitions that you want
3535 installed in one or more files of their own and use the @code{load}
3536 function to cause Emacs to evaluate and thereby install each of the
3537 functions in the files.
3538 @xref{Loading Files, , Loading Files}.
3541 Thirdly, if you have code that your whole site will use, it is usual
3542 to put it in a file called @file{site-init.el} that is loaded when
3543 Emacs is built. This makes the code available to everyone who uses
3544 your machine. (See the @file{INSTALL} file that is part of the Emacs
3548 Finally, if you have code that everyone who uses Emacs may want, you
3549 can post it on a computer network or send a copy to the Free Software
3550 Foundation. (When you do this, please license the code and its
3551 documentation under a license that permits other people to run, copy,
3552 study, modify, and redistribute the code and which protects you from
3553 having your work taken from you.) If you send a copy of your code to
3554 the Free Software Foundation, and properly protect yourself and
3555 others, it may be included in the next release of Emacs. In large
3556 part, this is how Emacs has grown over the past years, by donations.
3562 The @code{let} expression is a special form in Lisp that you will need
3563 to use in most function definitions.
3565 @code{let} is used to attach or bind a symbol to a value in such a way
3566 that the Lisp interpreter will not confuse the variable with a
3567 variable of the same name that is not part of the function.
3569 To understand why the @code{let} special form is necessary, consider
3570 the situation in which you own a home that you generally refer to as
3571 `the house', as in the sentence, ``The house needs painting.'' If you
3572 are visiting a friend and your host refers to `the house', he is
3573 likely to be referring to @emph{his} house, not yours, that is, to a
3576 If your friend is referring to his house and you think he is referring
3577 to your house, you may be in for some confusion. The same thing could
3578 happen in Lisp if a variable that is used inside of one function has
3579 the same name as a variable that is used inside of another function,
3580 and the two are not intended to refer to the same value. The
3581 @code{let} special form prevents this kind of confusion.
3584 * Prevent confusion::
3585 * Parts of let Expression::
3586 * Sample let Expression::
3587 * Uninitialized let Variables::
3591 @node Prevent confusion
3592 @unnumberedsubsec @code{let} Prevents Confusion
3595 @cindex @samp{local variable} defined
3596 @cindex @samp{variable, local}, defined
3597 The @code{let} special form prevents confusion. @code{let} creates a
3598 name for a @dfn{local variable} that overshadows any use of the same
3599 name outside the @code{let} expression. This is like understanding
3600 that whenever your host refers to `the house', he means his house, not
3601 yours. (Symbols used in argument lists work the same way.
3602 @xref{defun, , The @code{defun} Macro}.)
3604 Local variables created by a @code{let} expression retain their value
3605 @emph{only} within the @code{let} expression itself (and within
3606 expressions called within the @code{let} expression); the local
3607 variables have no effect outside the @code{let} expression.
3609 Another way to think about @code{let} is that it is like a @code{setq}
3610 that is temporary and local. The values set by @code{let} are
3611 automatically undone when the @code{let} is finished. The setting
3612 only affects expressions that are inside the bounds of the @code{let}
3613 expression. In computer science jargon, we would say ``the binding of
3614 a symbol is visible only in functions called in the @code{let} form;
3615 in Emacs Lisp, scoping is dynamic, not lexical.''
3617 @code{let} can create more than one variable at once. Also,
3618 @code{let} gives each variable it creates an initial value, either a
3619 value specified by you, or @code{nil}. (In the jargon, this is called
3620 `binding the variable to the value'.) After @code{let} has created
3621 and bound the variables, it executes the code in the body of the
3622 @code{let}, and returns the value of the last expression in the body,
3623 as the value of the whole @code{let} expression. (`Execute' is a jargon
3624 term that means to evaluate a list; it comes from the use of the word
3625 meaning `to give practical effect to' (@cite{Oxford English
3626 Dictionary}). Since you evaluate an expression to perform an action,
3627 `execute' has evolved as a synonym to `evaluate'.)
3629 @node Parts of let Expression
3630 @subsection The Parts of a @code{let} Expression
3631 @cindex @code{let} expression, parts of
3632 @cindex Parts of @code{let} expression
3634 @cindex @samp{varlist} defined
3635 A @code{let} expression is a list of three parts. The first part is
3636 the symbol @code{let}. The second part is a list, called a
3637 @dfn{varlist}, each element of which is either a symbol by itself or a
3638 two-element list, the first element of which is a symbol. The third
3639 part of the @code{let} expression is the body of the @code{let}. The
3640 body usually consists of one or more lists.
3643 A template for a @code{let} expression looks like this:
3646 (let @var{varlist} @var{body}@dots{})
3650 The symbols in the varlist are the variables that are given initial
3651 values by the @code{let} special form. Symbols by themselves are given
3652 the initial value of @code{nil}; and each symbol that is the first
3653 element of a two-element list is bound to the value that is returned
3654 when the Lisp interpreter evaluates the second element.
3656 Thus, a varlist might look like this: @code{(thread (needles 3))}. In
3657 this case, in a @code{let} expression, Emacs binds the symbol
3658 @code{thread} to an initial value of @code{nil}, and binds the symbol
3659 @code{needles} to an initial value of 3.
3661 When you write a @code{let} expression, what you do is put the
3662 appropriate expressions in the slots of the @code{let} expression
3665 If the varlist is composed of two-element lists, as is often the case,
3666 the template for the @code{let} expression looks like this:
3670 (let ((@var{variable} @var{value})
3671 (@var{variable} @var{value})
3677 @node Sample let Expression
3678 @subsection Sample @code{let} Expression
3679 @cindex Sample @code{let} expression
3680 @cindex @code{let} expression sample
3682 The following expression creates and gives initial values
3683 to the two variables @code{zebra} and @code{tiger}. The body of the
3684 @code{let} expression is a list which calls the @code{message} function.
3688 (let ((zebra 'stripes)
3690 (message "One kind of animal has %s and another is %s."
3695 Here, the varlist is @code{((zebra 'stripes) (tiger 'fierce))}.
3697 The two variables are @code{zebra} and @code{tiger}. Each variable is
3698 the first element of a two-element list and each value is the second
3699 element of its two-element list. In the varlist, Emacs binds the
3700 variable @code{zebra} to the value @code{stripes}@footnote{According
3701 to Jared Diamond in @cite{Guns, Germs, and Steel}, ``@dots{} zebras
3702 become impossibly dangerous as they grow older'' but the claim here is
3703 that they do not become fierce like a tiger. (1997, W. W. Norton and
3704 Co., ISBN 0-393-03894-2, page 171)}, and binds the
3705 variable @code{tiger} to the value @code{fierce}. In this example,
3706 both values are symbols preceded by a quote. The values could just as
3707 well have been another list or a string. The body of the @code{let}
3708 follows after the list holding the variables. In this example, the
3709 body is a list that uses the @code{message} function to print a string
3713 You may evaluate the example in the usual fashion, by placing the
3714 cursor after the last parenthesis and typing @kbd{C-x C-e}. When you do
3715 this, the following will appear in the echo area:
3718 "One kind of animal has stripes and another is fierce."
3721 As we have seen before, the @code{message} function prints its first
3722 argument, except for @samp{%s}. In this example, the value of the variable
3723 @code{zebra} is printed at the location of the first @samp{%s} and the
3724 value of the variable @code{tiger} is printed at the location of the
3727 @node Uninitialized let Variables
3728 @subsection Uninitialized Variables in a @code{let} Statement
3729 @cindex Uninitialized @code{let} variables
3730 @cindex @code{let} variables uninitialized
3732 If you do not bind the variables in a @code{let} statement to specific
3733 initial values, they will automatically be bound to an initial value of
3734 @code{nil}, as in the following expression:
3743 "Here are %d variables with %s, %s, and %s value."
3744 birch pine fir oak))
3749 Here, the varlist is @code{((birch 3) pine fir (oak 'some))}.
3752 If you evaluate this expression in the usual way, the following will
3753 appear in your echo area:
3756 "Here are 3 variables with nil, nil, and some value."
3760 In this example, Emacs binds the symbol @code{birch} to the number 3,
3761 binds the symbols @code{pine} and @code{fir} to @code{nil}, and binds
3762 the symbol @code{oak} to the value @code{some}.
3764 Note that in the first part of the @code{let}, the variables @code{pine}
3765 and @code{fir} stand alone as atoms that are not surrounded by
3766 parentheses; this is because they are being bound to @code{nil}, the
3767 empty list. But @code{oak} is bound to @code{some} and so is a part of
3768 the list @code{(oak 'some)}. Similarly, @code{birch} is bound to the
3769 number 3 and so is in a list with that number. (Since a number
3770 evaluates to itself, the number does not need to be quoted. Also, the
3771 number is printed in the message using a @samp{%d} rather than a
3772 @samp{%s}.) The four variables as a group are put into a list to
3773 delimit them from the body of the @code{let}.
3776 @section The @code{if} Special Form
3778 @cindex Conditional with @code{if}
3780 A third special form, in addition to @code{defun} and @code{let}, is the
3781 conditional @code{if}. This form is used to instruct the computer to
3782 make decisions. You can write function definitions without using
3783 @code{if}, but it is used often enough, and is important enough, to be
3784 included here. It is used, for example, in the code for the
3785 function @code{beginning-of-buffer}.
3787 The basic idea behind an @code{if}, is that ``@emph{if} a test is true,
3788 @emph{then} an expression is evaluated.'' If the test is not true, the
3789 expression is not evaluated. For example, you might make a decision
3790 such as, ``if it is warm and sunny, then go to the beach!''
3793 * if in more detail::
3794 * type-of-animal in detail:: An example of an @code{if} expression.
3798 @node if in more detail
3799 @unnumberedsubsec @code{if} in more detail
3802 @cindex @samp{if-part} defined
3803 @cindex @samp{then-part} defined
3804 An @code{if} expression written in Lisp does not use the word `then';
3805 the test and the action are the second and third elements of the list
3806 whose first element is @code{if}. Nonetheless, the test part of an
3807 @code{if} expression is often called the @dfn{if-part} and the second
3808 argument is often called the @dfn{then-part}.
3810 Also, when an @code{if} expression is written, the true-or-false-test
3811 is usually written on the same line as the symbol @code{if}, but the
3812 action to carry out if the test is true, the ``then-part'', is written
3813 on the second and subsequent lines. This makes the @code{if}
3814 expression easier to read.
3818 (if @var{true-or-false-test}
3819 @var{action-to-carry-out-if-test-is-true})
3824 The true-or-false-test will be an expression that
3825 is evaluated by the Lisp interpreter.
3827 Here is an example that you can evaluate in the usual manner. The test
3828 is whether the number 5 is greater than the number 4. Since it is, the
3829 message @samp{5 is greater than 4!} will be printed.
3833 (if (> 5 4) ; @r{if-part}
3834 (message "5 is greater than 4!")) ; @r{then-part}
3839 (The function @code{>} tests whether its first argument is greater than
3840 its second argument and returns true if it is.)
3841 @findex > (greater than)
3843 Of course, in actual use, the test in an @code{if} expression will not
3844 be fixed for all time as it is by the expression @code{(> 5 4)}.
3845 Instead, at least one of the variables used in the test will be bound to
3846 a value that is not known ahead of time. (If the value were known ahead
3847 of time, we would not need to run the test!)
3849 For example, the value may be bound to an argument of a function
3850 definition. In the following function definition, the character of the
3851 animal is a value that is passed to the function. If the value bound to
3852 @code{characteristic} is @code{fierce}, then the message, @samp{It's a
3853 tiger!} will be printed; otherwise, @code{nil} will be returned.
3857 (defun type-of-animal (characteristic)
3858 "Print message in echo area depending on CHARACTERISTIC.
3859 If the CHARACTERISTIC is the symbol `fierce',
3860 then warn of a tiger."
3861 (if (equal characteristic 'fierce)
3862 (message "It's a tiger!")))
3868 If you are reading this inside of GNU Emacs, you can evaluate the
3869 function definition in the usual way to install it in Emacs, and then you
3870 can evaluate the following two expressions to see the results:
3874 (type-of-animal 'fierce)
3876 (type-of-animal 'zebra)
3881 @c Following sentences rewritten to prevent overfull hbox.
3883 When you evaluate @code{(type-of-animal 'fierce)}, you will see the
3884 following message printed in the echo area: @code{"It's a tiger!"}; and
3885 when you evaluate @code{(type-of-animal 'zebra)} you will see @code{nil}
3886 printed in the echo area.
3888 @node type-of-animal in detail
3889 @subsection The @code{type-of-animal} Function in Detail
3891 Let's look at the @code{type-of-animal} function in detail.
3893 The function definition for @code{type-of-animal} was written by filling
3894 the slots of two templates, one for a function definition as a whole, and
3895 a second for an @code{if} expression.
3898 The template for every function that is not interactive is:
3902 (defun @var{name-of-function} (@var{argument-list})
3903 "@var{documentation}@dots{}"
3909 The parts of the function that match this template look like this:
3913 (defun type-of-animal (characteristic)
3914 "Print message in echo area depending on CHARACTERISTIC.
3915 If the CHARACTERISTIC is the symbol `fierce',
3916 then warn of a tiger."
3917 @var{body: the} @code{if} @var{expression})
3921 The name of function is @code{type-of-animal}; it is passed the value
3922 of one argument. The argument list is followed by a multi-line
3923 documentation string. The documentation string is included in the
3924 example because it is a good habit to write documentation string for
3925 every function definition. The body of the function definition
3926 consists of the @code{if} expression.
3929 The template for an @code{if} expression looks like this:
3933 (if @var{true-or-false-test}
3934 @var{action-to-carry-out-if-the-test-returns-true})
3939 In the @code{type-of-animal} function, the code for the @code{if}
3944 (if (equal characteristic 'fierce)
3945 (message "It's a tiger!")))
3950 Here, the true-or-false-test is the expression:
3953 (equal characteristic 'fierce)
3957 In Lisp, @code{equal} is a function that determines whether its first
3958 argument is equal to its second argument. The second argument is the
3959 quoted symbol @code{'fierce} and the first argument is the value of the
3960 symbol @code{characteristic}---in other words, the argument passed to
3963 In the first exercise of @code{type-of-animal}, the argument
3964 @code{fierce} is passed to @code{type-of-animal}. Since @code{fierce}
3965 is equal to @code{fierce}, the expression, @code{(equal characteristic
3966 'fierce)}, returns a value of true. When this happens, the @code{if}
3967 evaluates the second argument or then-part of the @code{if}:
3968 @code{(message "It's tiger!")}.
3970 On the other hand, in the second exercise of @code{type-of-animal}, the
3971 argument @code{zebra} is passed to @code{type-of-animal}. @code{zebra}
3972 is not equal to @code{fierce}, so the then-part is not evaluated and
3973 @code{nil} is returned by the @code{if} expression.
3976 @section If--then--else Expressions
3979 An @code{if} expression may have an optional third argument, called
3980 the @dfn{else-part}, for the case when the true-or-false-test returns
3981 false. When this happens, the second argument or then-part of the
3982 overall @code{if} expression is @emph{not} evaluated, but the third or
3983 else-part @emph{is} evaluated. You might think of this as the cloudy
3984 day alternative for the decision ``if it is warm and sunny, then go to
3985 the beach, else read a book!''.
3987 The word ``else'' is not written in the Lisp code; the else-part of an
3988 @code{if} expression comes after the then-part. In the written Lisp, the
3989 else-part is usually written to start on a line of its own and is
3990 indented less than the then-part:
3994 (if @var{true-or-false-test}
3995 @var{action-to-carry-out-if-the-test-returns-true}
3996 @var{action-to-carry-out-if-the-test-returns-false})
4000 For example, the following @code{if} expression prints the message @samp{4
4001 is not greater than 5!} when you evaluate it in the usual way:
4005 (if (> 4 5) ; @r{if-part}
4006 (message "4 falsely greater than 5!") ; @r{then-part}
4007 (message "4 is not greater than 5!")) ; @r{else-part}
4012 Note that the different levels of indentation make it easy to
4013 distinguish the then-part from the else-part. (GNU Emacs has several
4014 commands that automatically indent @code{if} expressions correctly.
4015 @xref{Typing Lists, , GNU Emacs Helps You Type Lists}.)
4017 We can extend the @code{type-of-animal} function to include an
4018 else-part by simply incorporating an additional part to the @code{if}
4022 You can see the consequences of doing this if you evaluate the following
4023 version of the @code{type-of-animal} function definition to install it
4024 and then evaluate the two subsequent expressions to pass different
4025 arguments to the function.
4029 (defun type-of-animal (characteristic) ; @r{Second version.}
4030 "Print message in echo area depending on CHARACTERISTIC.
4031 If the CHARACTERISTIC is the symbol `fierce',
4032 then warn of a tiger;
4033 else say it's not fierce."
4034 (if (equal characteristic 'fierce)
4035 (message "It's a tiger!")
4036 (message "It's not fierce!")))
4043 (type-of-animal 'fierce)
4045 (type-of-animal 'zebra)
4050 @c Following sentence rewritten to prevent overfull hbox.
4052 When you evaluate @code{(type-of-animal 'fierce)}, you will see the
4053 following message printed in the echo area: @code{"It's a tiger!"}; but
4054 when you evaluate @code{(type-of-animal 'zebra)}, you will see
4055 @code{"It's not fierce!"}.
4057 (Of course, if the @var{characteristic} were @code{ferocious}, the
4058 message @code{"It's not fierce!"} would be printed; and it would be
4059 misleading! When you write code, you need to take into account the
4060 possibility that some such argument will be tested by the @code{if}
4061 and write your program accordingly.)
4063 @node Truth & Falsehood
4064 @section Truth and Falsehood in Emacs Lisp
4065 @cindex Truth and falsehood in Emacs Lisp
4066 @cindex Falsehood and truth in Emacs Lisp
4069 There is an important aspect to the truth test in an @code{if}
4070 expression. So far, we have spoken of `true' and `false' as values of
4071 predicates as if they were new kinds of Emacs Lisp objects. In fact,
4072 `false' is just our old friend @code{nil}. Anything else---anything
4075 The expression that tests for truth is interpreted as @dfn{true}
4076 if the result of evaluating it is a value that is not @code{nil}. In
4077 other words, the result of the test is considered true if the value
4078 returned is a number such as 47, a string such as @code{"hello"}, or a
4079 symbol (other than @code{nil}) such as @code{flowers}, or a list (so
4080 long as it is not empty), or even a buffer!
4083 * nil explained:: @code{nil} has two meanings.
4088 @unnumberedsubsec An explanation of @code{nil}
4091 Before illustrating a test for truth, we need an explanation of @code{nil}.
4093 In Emacs Lisp, the symbol @code{nil} has two meanings. First, it means the
4094 empty list. Second, it means false and is the value returned when a
4095 true-or-false-test tests false. @code{nil} can be written as an empty
4096 list, @code{()}, or as @code{nil}. As far as the Lisp interpreter is
4097 concerned, @code{()} and @code{nil} are the same. Humans, however, tend
4098 to use @code{nil} for false and @code{()} for the empty list.
4100 In Emacs Lisp, any value that is not @code{nil}---is not the empty
4101 list---is considered true. This means that if an evaluation returns
4102 something that is not an empty list, an @code{if} expression will test
4103 true. For example, if a number is put in the slot for the test, it
4104 will be evaluated and will return itself, since that is what numbers
4105 do when evaluated. In this conditional, the @code{if} expression will
4106 test true. The expression tests false only when @code{nil}, an empty
4107 list, is returned by evaluating the expression.
4109 You can see this by evaluating the two expressions in the following examples.
4111 In the first example, the number 4 is evaluated as the test in the
4112 @code{if} expression and returns itself; consequently, the then-part
4113 of the expression is evaluated and returned: @samp{true} appears in
4114 the echo area. In the second example, the @code{nil} indicates false;
4115 consequently, the else-part of the expression is evaluated and
4116 returned: @samp{false} appears in the echo area.
4133 Incidentally, if some other useful value is not available for a test that
4134 returns true, then the Lisp interpreter will return the symbol @code{t}
4135 for true. For example, the expression @code{(> 5 4)} returns @code{t}
4136 when evaluated, as you can see by evaluating it in the usual way:
4144 On the other hand, this function returns @code{nil} if the test is false.
4150 @node save-excursion
4151 @section @code{save-excursion}
4152 @findex save-excursion
4153 @cindex Region, what it is
4154 @cindex Preserving point, mark, and buffer
4155 @cindex Point, mark, buffer preservation
4159 The @code{save-excursion} function is the third and final special form
4160 that we will discuss in this chapter.
4162 In Emacs Lisp programs used for editing, the @code{save-excursion}
4163 function is very common. It saves the location of point and mark,
4164 executes the body of the function, and then restores point and mark to
4165 their previous positions if their locations were changed. Its primary
4166 purpose is to keep the user from being surprised and disturbed by
4167 unexpected movement of point or mark.
4170 * Point and mark:: A review of various locations.
4171 * Template for save-excursion::
4175 @node Point and mark
4176 @unnumberedsubsec Point and Mark
4179 Before discussing @code{save-excursion}, however, it may be useful
4180 first to review what point and mark are in GNU Emacs. @dfn{Point} is
4181 the current location of the cursor. Wherever the cursor
4182 is, that is point. More precisely, on terminals where the cursor
4183 appears to be on top of a character, point is immediately before the
4184 character. In Emacs Lisp, point is an integer. The first character in
4185 a buffer is number one, the second is number two, and so on. The
4186 function @code{point} returns the current position of the cursor as a
4187 number. Each buffer has its own value for point.
4189 The @dfn{mark} is another position in the buffer; its value can be set
4190 with a command such as @kbd{C-@key{SPC}} (@code{set-mark-command}). If
4191 a mark has been set, you can use the command @kbd{C-x C-x}
4192 (@code{exchange-point-and-mark}) to cause the cursor to jump to the mark
4193 and set the mark to be the previous position of point. In addition, if
4194 you set another mark, the position of the previous mark is saved in the
4195 mark ring. Many mark positions can be saved this way. You can jump the
4196 cursor to a saved mark by typing @kbd{C-u C-@key{SPC}} one or more
4199 The part of the buffer between point and mark is called @dfn{the
4200 region}. Numerous commands work on the region, including
4201 @code{center-region}, @code{count-lines-region}, @code{kill-region}, and
4202 @code{print-region}.
4204 The @code{save-excursion} special form saves the locations of point and
4205 mark and restores those positions after the code within the body of the
4206 special form is evaluated by the Lisp interpreter. Thus, if point were
4207 in the beginning of a piece of text and some code moved point to the end
4208 of the buffer, the @code{save-excursion} would put point back to where
4209 it was before, after the expressions in the body of the function were
4212 In Emacs, a function frequently moves point as part of its internal
4213 workings even though a user would not expect this. For example,
4214 @code{count-lines-region} moves point. To prevent the user from being
4215 bothered by jumps that are both unexpected and (from the user's point of
4216 view) unnecessary, @code{save-excursion} is often used to keep point and
4217 mark in the location expected by the user. The use of
4218 @code{save-excursion} is good housekeeping.
4220 To make sure the house stays clean, @code{save-excursion} restores the
4221 values of point and mark even if something goes wrong in the code inside
4222 of it (or, to be more precise and to use the proper jargon, ``in case of
4223 abnormal exit''). This feature is very helpful.
4225 In addition to recording the values of point and mark,
4226 @code{save-excursion} keeps track of the current buffer, and restores
4227 it, too. This means you can write code that will change the buffer and
4228 have @code{save-excursion} switch you back to the original buffer.
4229 This is how @code{save-excursion} is used in @code{append-to-buffer}.
4230 (@xref{append-to-buffer, , The Definition of @code{append-to-buffer}}.)
4232 @node Template for save-excursion
4233 @subsection Template for a @code{save-excursion} Expression
4236 The template for code using @code{save-excursion} is simple:
4246 The body of the function is one or more expressions that will be
4247 evaluated in sequence by the Lisp interpreter. If there is more than
4248 one expression in the body, the value of the last one will be returned
4249 as the value of the @code{save-excursion} function. The other
4250 expressions in the body are evaluated only for their side effects; and
4251 @code{save-excursion} itself is used only for its side effect (which
4252 is restoring the positions of point and mark).
4255 In more detail, the template for a @code{save-excursion} expression
4261 @var{first-expression-in-body}
4262 @var{second-expression-in-body}
4263 @var{third-expression-in-body}
4265 @var{last-expression-in-body})
4270 An expression, of course, may be a symbol on its own or a list.
4272 In Emacs Lisp code, a @code{save-excursion} expression often occurs
4273 within the body of a @code{let} expression. It looks like this:
4286 In the last few chapters we have introduced a macro and a fair number
4287 of functions and special forms. Here they are described in brief,
4288 along with a few similar functions that have not been mentioned yet.
4291 @item eval-last-sexp
4292 Evaluate the last symbolic expression before the current location of
4293 point. The value is printed in the echo area unless the function is
4294 invoked with an argument; in that case, the output is printed in the
4295 current buffer. This command is normally bound to @kbd{C-x C-e}.
4298 Define function. This macro has up to five parts: the name, a
4299 template for the arguments that will be passed to the function,
4300 documentation, an optional interactive declaration, and the body of
4304 For example, in an early version of Emacs, the function definition was
4305 as follows. (It is slightly more complex now that it seeks the first
4306 non-whitespace character rather than the first visible character.)
4310 (defun back-to-indentation ()
4311 "Move point to first visible character on line."
4313 (beginning-of-line 1)
4314 (skip-chars-forward " \t"))
4321 (defun backward-to-indentation (&optional arg)
4322 "Move backward ARG lines and position at first nonblank character."
4324 (forward-line (- (or arg 1)))
4325 (skip-chars-forward " \t"))
4327 (defun back-to-indentation ()
4328 "Move point to the first non-whitespace character on this line."
4330 (beginning-of-line 1)
4331 (skip-syntax-forward " " (line-end-position))
4332 ;; Move back over chars that have whitespace syntax but have the p flag.
4333 (backward-prefix-chars))
4337 Declare to the interpreter that the function can be used
4338 interactively. This special form may be followed by a string with one
4339 or more parts that pass the information to the arguments of the
4340 function, in sequence. These parts may also tell the interpreter to
4341 prompt for information. Parts of the string are separated by
4342 newlines, @samp{\n}.
4345 Common code characters are:
4349 The name of an existing buffer.
4352 The name of an existing file.
4355 The numeric prefix argument. (Note that this `p' is lower case.)
4358 Point and the mark, as two numeric arguments, smallest first. This
4359 is the only code letter that specifies two successive arguments
4363 @xref{Interactive Codes, , Code Characters for @samp{interactive},
4364 elisp, The GNU Emacs Lisp Reference Manual}, for a complete list of
4368 Declare that a list of variables is for use within the body of the
4369 @code{let} and give them an initial value, either @code{nil} or a
4370 specified value; then evaluate the rest of the expressions in the body
4371 of the @code{let} and return the value of the last one. Inside the
4372 body of the @code{let}, the Lisp interpreter does not see the values of
4373 the variables of the same names that are bound outside of the
4381 (let ((foo (buffer-name))
4382 (bar (buffer-size)))
4384 "This buffer is %s and has %d characters."
4389 @item save-excursion
4390 Record the values of point and mark and the current buffer before
4391 evaluating the body of this special form. Restore the values of point
4392 and mark and buffer afterward.
4399 (message "We are %d characters into this buffer."
4402 (goto-char (point-min)) (point))))
4407 Evaluate the first argument to the function; if it is true, evaluate
4408 the second argument; else evaluate the third argument, if there is one.
4410 The @code{if} special form is called a @dfn{conditional}. There are
4411 other conditionals in Emacs Lisp, but @code{if} is perhaps the most
4419 (if (= 22 emacs-major-version)
4420 (message "This is version 22 Emacs")
4421 (message "This is not version 22 Emacs"))
4430 The @code{<} function tests whether its first argument is smaller than
4431 its second argument. A corresponding function, @code{>}, tests whether
4432 the first argument is greater than the second. Likewise, @code{<=}
4433 tests whether the first argument is less than or equal to the second and
4434 @code{>=} tests whether the first argument is greater than or equal to
4435 the second. In all cases, both arguments must be numbers or markers
4436 (markers indicate positions in buffers).
4440 The @code{=} function tests whether two arguments, both numbers or
4446 Test whether two objects are the same. @code{equal} uses one meaning
4447 of the word `same' and @code{eq} uses another: @code{equal} returns
4448 true if the two objects have a similar structure and contents, such as
4449 two copies of the same book. On the other hand, @code{eq}, returns
4450 true if both arguments are actually the same object.
4459 The @code{string-lessp} function tests whether its first argument is
4460 smaller than the second argument. A shorter, alternative name for the
4461 same function (a @code{defalias}) is @code{string<}.
4463 The arguments to @code{string-lessp} must be strings or symbols; the
4464 ordering is lexicographic, so case is significant. The print names of
4465 symbols are used instead of the symbols themselves.
4467 @cindex @samp{empty string} defined
4468 An empty string, @samp{""}, a string with no characters in it, is
4469 smaller than any string of characters.
4471 @code{string-equal} provides the corresponding test for equality. Its
4472 shorter, alternative name is @code{string=}. There are no string test
4473 functions that correspond to @var{>}, @code{>=}, or @code{<=}.
4476 Print a message in the echo area. The first argument is a string that
4477 can contain @samp{%s}, @samp{%d}, or @samp{%c} to print the value of
4478 arguments that follow the string. The argument used by @samp{%s} must
4479 be a string or a symbol; the argument used by @samp{%d} must be a
4480 number. The argument used by @samp{%c} must be an @sc{ascii} code
4481 number; it will be printed as the character with that @sc{ascii} code.
4482 (Various other %-sequences have not been mentioned.)
4486 The @code{setq} function sets the value of its first argument to the
4487 value of the second argument. The first argument is automatically
4488 quoted by @code{setq}. It does the same for succeeding pairs of
4489 arguments. Another function, @code{set}, takes only two arguments and
4490 evaluates both of them before setting the value returned by its first
4491 argument to the value returned by its second argument.
4494 Without an argument, return the name of the buffer, as a string.
4496 @item buffer-file-name
4497 Without an argument, return the name of the file the buffer is
4500 @item current-buffer
4501 Return the buffer in which Emacs is active; it may not be
4502 the buffer that is visible on the screen.
4505 Return the most recently selected buffer (other than the buffer passed
4506 to @code{other-buffer} as an argument and other than the current
4509 @item switch-to-buffer
4510 Select a buffer for Emacs to be active in and display it in the current
4511 window so users can look at it. Usually bound to @kbd{C-x b}.
4514 Switch Emacs's attention to a buffer on which programs will run. Don't
4515 alter what the window is showing.
4518 Return the number of characters in the current buffer.
4521 Return the value of the current position of the cursor, as an
4522 integer counting the number of characters from the beginning of the
4526 Return the minimum permissible value of point in
4527 the current buffer. This is 1, unless narrowing is in effect.
4530 Return the value of the maximum permissible value of point in the
4531 current buffer. This is the end of the buffer, unless narrowing is in
4536 @node defun Exercises
4541 Write a non-interactive function that doubles the value of its
4542 argument, a number. Make that function interactive.
4545 Write a function that tests whether the current value of
4546 @code{fill-column} is greater than the argument passed to the function,
4547 and if so, prints an appropriate message.
4550 @node Buffer Walk Through
4551 @chapter A Few Buffer--Related Functions
4553 In this chapter we study in detail several of the functions used in GNU
4554 Emacs. This is called a ``walk-through''. These functions are used as
4555 examples of Lisp code, but are not imaginary examples; with the
4556 exception of the first, simplified function definition, these functions
4557 show the actual code used in GNU Emacs. You can learn a great deal from
4558 these definitions. The functions described here are all related to
4559 buffers. Later, we will study other functions.
4562 * Finding More:: How to find more information.
4563 * simplified-beginning-of-buffer:: Shows @code{goto-char},
4564 @code{point-min}, and @code{push-mark}.
4565 * mark-whole-buffer:: Almost the same as @code{beginning-of-buffer}.
4566 * append-to-buffer:: Uses @code{save-excursion} and
4567 @code{insert-buffer-substring}.
4568 * Buffer Related Review:: Review.
4569 * Buffer Exercises::
4573 @section Finding More Information
4575 @findex describe-function, @r{introduced}
4576 @cindex Find function documentation
4577 In this walk-through, I will describe each new function as we come to
4578 it, sometimes in detail and sometimes briefly. If you are interested,
4579 you can get the full documentation of any Emacs Lisp function at any
4580 time by typing @kbd{C-h f} and then the name of the function (and then
4581 @key{RET}). Similarly, you can get the full documentation for a
4582 variable by typing @kbd{C-h v} and then the name of the variable (and
4585 @cindex Find source of function
4586 @c In version 22, tells location both of C and of Emacs Lisp
4587 Also, @code{describe-function} will tell you the location of the
4588 function definition.
4590 Put point into the name of the file that contains the function and
4591 press the @key{RET} key. In this case, @key{RET} means
4592 @code{push-button} rather than `return' or `enter'. Emacs will take
4593 you directly to the function definition.
4598 If you move point over the file name and press
4599 the @key{RET} key, which in this case means @code{help-follow} rather
4600 than `return' or `enter', Emacs will take you directly to the function
4604 More generally, if you want to see a function in its original source
4605 file, you can use the @code{find-tag} function to jump to it.
4606 @code{find-tag} works with a wide variety of languages, not just
4607 Lisp, and C, and it works with non-programming text as well. For
4608 example, @code{find-tag} will jump to the various nodes in the
4609 Texinfo source file of this document.
4610 The @code{find-tag} function depends on `tags tables' that record
4611 the locations of the functions, variables, and other items to which
4612 @code{find-tag} jumps.
4614 To use the @code{find-tag} command, type @kbd{M-.} (i.e., press the
4615 period key while holding down the @key{META} key, or else type the
4616 @key{ESC} key and then type the period key), and then, at the prompt,
4617 type in the name of the function whose source code you want to see,
4618 such as @code{mark-whole-buffer}, and then type @key{RET}. Emacs will
4619 switch buffers and display the source code for the function on your
4620 screen. To switch back to your current buffer, type @kbd{C-x b
4621 @key{RET}}. (On some keyboards, the @key{META} key is labeled
4624 @c !!! 22.1.1 tags table location in this paragraph
4625 @cindex TAGS table, specifying
4627 Depending on how the initial default values of your copy of Emacs are
4628 set, you may also need to specify the location of your `tags table',
4629 which is a file called @file{TAGS}. For example, if you are
4630 interested in Emacs sources, the tags table you will most likely want,
4631 if it has already been created for you, will be in a subdirectory of
4632 the @file{/usr/local/share/emacs/} directory; thus you would use the
4633 @code{M-x visit-tags-table} command and specify a pathname such as
4634 @file{/usr/local/share/emacs/22.1.1/lisp/TAGS}. If the tags table
4635 has not already been created, you will have to create it yourself. It
4636 will be in a file such as @file{/usr/local/src/emacs/src/TAGS}.
4639 To create a @file{TAGS} file in a specific directory, switch to that
4640 directory in Emacs using @kbd{M-x cd} command, or list the directory
4641 with @kbd{C-x d} (@code{dired}). Then run the compile command, with
4642 @w{@code{etags *.el}} as the command to execute:
4645 M-x compile RET etags *.el RET
4648 For more information, see @ref{etags, , Create Your Own @file{TAGS} File}.
4650 After you become more familiar with Emacs Lisp, you will find that you will
4651 frequently use @code{find-tag} to navigate your way around source code;
4652 and you will create your own @file{TAGS} tables.
4654 @cindex Library, as term for `file'
4655 Incidentally, the files that contain Lisp code are conventionally
4656 called @dfn{libraries}. The metaphor is derived from that of a
4657 specialized library, such as a law library or an engineering library,
4658 rather than a general library. Each library, or file, contains
4659 functions that relate to a particular topic or activity, such as
4660 @file{abbrev.el} for handling abbreviations and other typing
4661 shortcuts, and @file{help.el} for on-line help. (Sometimes several
4662 libraries provide code for a single activity, as the various
4663 @file{rmail@dots{}} files provide code for reading electronic mail.)
4664 In @cite{The GNU Emacs Manual}, you will see sentences such as ``The
4665 @kbd{C-h p} command lets you search the standard Emacs Lisp libraries
4666 by topic keywords.''
4668 @node simplified-beginning-of-buffer
4669 @section A Simplified @code{beginning-of-buffer} Definition
4670 @findex simplified-beginning-of-buffer
4672 The @code{beginning-of-buffer} command is a good function to start with
4673 since you are likely to be familiar with it and it is easy to
4674 understand. Used as an interactive command, @code{beginning-of-buffer}
4675 moves the cursor to the beginning of the buffer, leaving the mark at the
4676 previous position. It is generally bound to @kbd{M-<}.
4678 In this section, we will discuss a shortened version of the function
4679 that shows how it is most frequently used. This shortened function
4680 works as written, but it does not contain the code for a complex option.
4681 In another section, we will describe the entire function.
4682 (@xref{beginning-of-buffer, , Complete Definition of
4683 @code{beginning-of-buffer}}.)
4685 Before looking at the code, let's consider what the function
4686 definition has to contain: it must include an expression that makes
4687 the function interactive so it can be called by typing @kbd{M-x
4688 beginning-of-buffer} or by typing a keychord such as @kbd{M-<}; it
4689 must include code to leave a mark at the original position in the
4690 buffer; and it must include code to move the cursor to the beginning
4694 Here is the complete text of the shortened version of the function:
4698 (defun simplified-beginning-of-buffer ()
4699 "Move point to the beginning of the buffer;
4700 leave mark at previous position."
4703 (goto-char (point-min)))
4707 Like all function definitions, this definition has five parts following
4708 the macro @code{defun}:
4712 The name: in this example, @code{simplified-beginning-of-buffer}.
4715 A list of the arguments: in this example, an empty list, @code{()},
4718 The documentation string.
4721 The interactive expression.
4728 In this function definition, the argument list is empty; this means that
4729 this function does not require any arguments. (When we look at the
4730 definition for the complete function, we will see that it may be passed
4731 an optional argument.)
4733 The interactive expression tells Emacs that the function is intended to
4734 be used interactively. In this example, @code{interactive} does not have
4735 an argument because @code{simplified-beginning-of-buffer} does not
4739 The body of the function consists of the two lines:
4744 (goto-char (point-min))
4748 The first of these lines is the expression, @code{(push-mark)}. When
4749 this expression is evaluated by the Lisp interpreter, it sets a mark at
4750 the current position of the cursor, wherever that may be. The position
4751 of this mark is saved in the mark ring.
4753 The next line is @code{(goto-char (point-min))}. This expression
4754 jumps the cursor to the minimum point in the buffer, that is, to the
4755 beginning of the buffer (or to the beginning of the accessible portion
4756 of the buffer if it is narrowed. @xref{Narrowing & Widening, ,
4757 Narrowing and Widening}.)
4759 The @code{push-mark} command sets a mark at the place where the cursor
4760 was located before it was moved to the beginning of the buffer by the
4761 @code{(goto-char (point-min))} expression. Consequently, you can, if
4762 you wish, go back to where you were originally by typing @kbd{C-x C-x}.
4764 That is all there is to the function definition!
4766 @findex describe-function
4767 When you are reading code such as this and come upon an unfamiliar
4768 function, such as @code{goto-char}, you can find out what it does by
4769 using the @code{describe-function} command. To use this command, type
4770 @kbd{C-h f} and then type in the name of the function and press
4771 @key{RET}. The @code{describe-function} command will print the
4772 function's documentation string in a @file{*Help*} window. For
4773 example, the documentation for @code{goto-char} is:
4777 Set point to POSITION, a number or marker.
4778 Beginning of buffer is position (point-min), end is (point-max).
4783 The function's one argument is the desired position.
4786 (The prompt for @code{describe-function} will offer you the symbol
4787 under or preceding the cursor, so you can save typing by positioning
4788 the cursor right over or after the function and then typing @kbd{C-h f
4791 The @code{end-of-buffer} function definition is written in the same way as
4792 the @code{beginning-of-buffer} definition except that the body of the
4793 function contains the expression @code{(goto-char (point-max))} in place
4794 of @code{(goto-char (point-min))}.
4796 @node mark-whole-buffer
4797 @section The Definition of @code{mark-whole-buffer}
4798 @findex mark-whole-buffer
4800 The @code{mark-whole-buffer} function is no harder to understand than the
4801 @code{simplified-beginning-of-buffer} function. In this case, however,
4802 we will look at the complete function, not a shortened version.
4804 The @code{mark-whole-buffer} function is not as commonly used as the
4805 @code{beginning-of-buffer} function, but is useful nonetheless: it
4806 marks a whole buffer as a region by putting point at the beginning and
4807 a mark at the end of the buffer. It is generally bound to @kbd{C-x
4811 * mark-whole-buffer overview::
4812 * Body of mark-whole-buffer:: Only three lines of code.
4816 @node mark-whole-buffer overview
4817 @unnumberedsubsec An overview of @code{mark-whole-buffer}
4821 In GNU Emacs 22, the code for the complete function looks like this:
4825 (defun mark-whole-buffer ()
4826 "Put point at beginning and mark at end of buffer.
4827 You probably should not use this function in Lisp programs;
4828 it is usually a mistake for a Lisp function to use any subroutine
4829 that uses or sets the mark."
4832 (push-mark (point-max) nil t)
4833 (goto-char (point-min)))
4838 Like all other functions, the @code{mark-whole-buffer} function fits
4839 into the template for a function definition. The template looks like
4844 (defun @var{name-of-function} (@var{argument-list})
4845 "@var{documentation}@dots{}"
4846 (@var{interactive-expression}@dots{})
4851 Here is how the function works: the name of the function is
4852 @code{mark-whole-buffer}; it is followed by an empty argument list,
4853 @samp{()}, which means that the function does not require arguments.
4854 The documentation comes next.
4856 The next line is an @code{(interactive)} expression that tells Emacs
4857 that the function will be used interactively. These details are similar
4858 to the @code{simplified-beginning-of-buffer} function described in the
4862 @node Body of mark-whole-buffer
4863 @subsection Body of @code{mark-whole-buffer}
4865 The body of the @code{mark-whole-buffer} function consists of three
4872 (push-mark (point-max) nil t)
4873 (goto-char (point-min))
4877 The first of these lines is the expression, @code{(push-mark (point))}.
4879 This line does exactly the same job as the first line of the body of
4880 the @code{simplified-beginning-of-buffer} function, which is written
4881 @code{(push-mark)}. In both cases, the Lisp interpreter sets a mark
4882 at the current position of the cursor.
4884 I don't know why the expression in @code{mark-whole-buffer} is written
4885 @code{(push-mark (point))} and the expression in
4886 @code{beginning-of-buffer} is written @code{(push-mark)}. Perhaps
4887 whoever wrote the code did not know that the arguments for
4888 @code{push-mark} are optional and that if @code{push-mark} is not
4889 passed an argument, the function automatically sets mark at the
4890 location of point by default. Or perhaps the expression was written
4891 so as to parallel the structure of the next line. In any case, the
4892 line causes Emacs to determine the position of point and set a mark
4895 In earlier versions of GNU Emacs, the next line of
4896 @code{mark-whole-buffer} was @code{(push-mark (point-max))}. This
4897 expression sets a mark at the point in the buffer that has the highest
4898 number. This will be the end of the buffer (or, if the buffer is
4899 narrowed, the end of the accessible portion of the buffer.
4900 @xref{Narrowing & Widening, , Narrowing and Widening}, for more about
4901 narrowing.) After this mark has been set, the previous mark, the one
4902 set at point, is no longer set, but Emacs remembers its position, just
4903 as all other recent marks are always remembered. This means that you
4904 can, if you wish, go back to that position by typing @kbd{C-u
4908 In GNU Emacs 22, the @code{(point-max)} is slightly more complicated.
4912 (push-mark (point-max) nil t)
4916 The expression works nearly the same as before. It sets a mark at the
4917 highest numbered place in the buffer that it can. However, in this
4918 version, @code{push-mark} has two additional arguments. The second
4919 argument to @code{push-mark} is @code{nil}. This tells the function
4920 it @emph{should} display a message that says `Mark set' when it pushes
4921 the mark. The third argument is @code{t}. This tells
4922 @code{push-mark} to activate the mark when Transient Mark mode is
4923 turned on. Transient Mark mode highlights the currently active
4924 region. It is often turned off.
4926 Finally, the last line of the function is @code{(goto-char
4927 (point-min)))}. This is written exactly the same way as it is written
4928 in @code{beginning-of-buffer}. The expression moves the cursor to
4929 the minimum point in the buffer, that is, to the beginning of the buffer
4930 (or to the beginning of the accessible portion of the buffer). As a
4931 result of this, point is placed at the beginning of the buffer and mark
4932 is set at the end of the buffer. The whole buffer is, therefore, the
4935 @node append-to-buffer
4936 @section The Definition of @code{append-to-buffer}
4937 @findex append-to-buffer
4939 The @code{append-to-buffer} command is more complex than the
4940 @code{mark-whole-buffer} command. What it does is copy the region
4941 (that is, the part of the buffer between point and mark) from the
4942 current buffer to a specified buffer.
4945 * append-to-buffer overview::
4946 * append interactive:: A two part interactive expression.
4947 * append-to-buffer body:: Incorporates a @code{let} expression.
4948 * append save-excursion:: How the @code{save-excursion} works.
4952 @node append-to-buffer overview
4953 @unnumberedsubsec An Overview of @code{append-to-buffer}
4956 @findex insert-buffer-substring
4957 The @code{append-to-buffer} command uses the
4958 @code{insert-buffer-substring} function to copy the region.
4959 @code{insert-buffer-substring} is described by its name: it takes a
4960 string of characters from part of a buffer, a ``substring'', and
4961 inserts them into another buffer.
4963 Most of @code{append-to-buffer} is
4964 concerned with setting up the conditions for
4965 @code{insert-buffer-substring} to work: the code must specify both the
4966 buffer to which the text will go, the window it comes from and goes
4967 to, and the region that will be copied.
4970 Here is the complete text of the function:
4974 (defun append-to-buffer (buffer start end)
4975 "Append to specified buffer the text of the region.
4976 It is inserted into that buffer before its point.
4980 When calling from a program, give three arguments:
4981 BUFFER (or buffer name), START and END.
4982 START and END specify the portion of the current buffer to be copied."
4984 (list (read-buffer "Append to buffer: " (other-buffer
4985 (current-buffer) t))
4986 (region-beginning) (region-end)))
4989 (let ((oldbuf (current-buffer)))
4991 (let* ((append-to (get-buffer-create buffer))
4992 (windows (get-buffer-window-list append-to t t))
4994 (set-buffer append-to)
4995 (setq point (point))
4996 (barf-if-buffer-read-only)
4997 (insert-buffer-substring oldbuf start end)
4998 (dolist (window windows)
4999 (when (= (window-point window) point)
5000 (set-window-point window (point))))))))
5004 The function can be understood by looking at it as a series of
5005 filled-in templates.
5007 The outermost template is for the function definition. In this
5008 function, it looks like this (with several slots filled in):
5012 (defun append-to-buffer (buffer start end)
5013 "@var{documentation}@dots{}"
5014 (interactive @dots{})
5019 The first line of the function includes its name and three arguments.
5020 The arguments are the @code{buffer} to which the text will be copied, and
5021 the @code{start} and @code{end} of the region in the current buffer that
5024 The next part of the function is the documentation, which is clear and
5025 complete. As is conventional, the three arguments are written in
5026 upper case so you will notice them easily. Even better, they are
5027 described in the same order as in the argument list.
5029 Note that the documentation distinguishes between a buffer and its
5030 name. (The function can handle either.)
5032 @node append interactive
5033 @subsection The @code{append-to-buffer} Interactive Expression
5035 Since the @code{append-to-buffer} function will be used interactively,
5036 the function must have an @code{interactive} expression. (For a
5037 review of @code{interactive}, see @ref{Interactive, , Making a
5038 Function Interactive}.) The expression reads as follows:
5044 "Append to buffer: "
5045 (other-buffer (current-buffer) t))
5052 This expression is not one with letters standing for parts, as
5053 described earlier. Instead, it starts a list with these parts:
5055 The first part of the list is an expression to read the name of a
5056 buffer and return it as a string. That is @code{read-buffer}. The
5057 function requires a prompt as its first argument, @samp{"Append to
5058 buffer: "}. Its second argument tells the command what value to
5059 provide if you don't specify anything.
5061 In this case that second argument is an expression containing the
5062 function @code{other-buffer}, an exception, and a @samp{t}, standing
5065 The first argument to @code{other-buffer}, the exception, is yet
5066 another function, @code{current-buffer}. That is not going to be
5067 returned. The second argument is the symbol for true, @code{t}. that
5068 tells @code{other-buffer} that it may show visible buffers (except in
5069 this case, it will not show the current buffer, which makes sense).
5072 The expression looks like this:
5075 (other-buffer (current-buffer) t)
5078 The second and third arguments to the @code{list} expression are
5079 @code{(region-beginning)} and @code{(region-end)}. These two
5080 functions specify the beginning and end of the text to be appended.
5083 Originally, the command used the letters @samp{B} and @samp{r}.
5084 The whole @code{interactive} expression looked like this:
5087 (interactive "BAppend to buffer:@: \nr")
5091 But when that was done, the default value of the buffer switched to
5092 was invisible. That was not wanted.
5094 (The prompt was separated from the second argument with a newline,
5095 @samp{\n}. It was followed by an @samp{r} that told Emacs to bind the
5096 two arguments that follow the symbol @code{buffer} in the function's
5097 argument list (that is, @code{start} and @code{end}) to the values of
5098 point and mark. That argument worked fine.)
5100 @node append-to-buffer body
5101 @subsection The Body of @code{append-to-buffer}
5104 in GNU Emacs 22 in /usr/local/src/emacs/lisp/simple.el
5106 (defun append-to-buffer (buffer start end)
5107 "Append to specified buffer the text of the region.
5108 It is inserted into that buffer before its point.
5110 When calling from a program, give three arguments:
5111 BUFFER (or buffer name), START and END.
5112 START and END specify the portion of the current buffer to be copied."
5114 (list (read-buffer "Append to buffer: " (other-buffer (current-buffer) t))
5115 (region-beginning) (region-end)))
5116 (let ((oldbuf (current-buffer)))
5118 (let* ((append-to (get-buffer-create buffer))
5119 (windows (get-buffer-window-list append-to t t))
5121 (set-buffer append-to)
5122 (setq point (point))
5123 (barf-if-buffer-read-only)
5124 (insert-buffer-substring oldbuf start end)
5125 (dolist (window windows)
5126 (when (= (window-point window) point)
5127 (set-window-point window (point))))))))
5130 The body of the @code{append-to-buffer} function begins with @code{let}.
5132 As we have seen before (@pxref{let, , @code{let}}), the purpose of a
5133 @code{let} expression is to create and give initial values to one or
5134 more variables that will only be used within the body of the
5135 @code{let}. This means that such a variable will not be confused with
5136 any variable of the same name outside the @code{let} expression.
5138 We can see how the @code{let} expression fits into the function as a
5139 whole by showing a template for @code{append-to-buffer} with the
5140 @code{let} expression in outline:
5144 (defun append-to-buffer (buffer start end)
5145 "@var{documentation}@dots{}"
5146 (interactive @dots{})
5147 (let ((@var{variable} @var{value}))
5152 The @code{let} expression has three elements:
5156 The symbol @code{let};
5159 A varlist containing, in this case, a single two-element list,
5160 @code{(@var{variable} @var{value})};
5163 The body of the @code{let} expression.
5167 In the @code{append-to-buffer} function, the varlist looks like this:
5170 (oldbuf (current-buffer))
5174 In this part of the @code{let} expression, the one variable,
5175 @code{oldbuf}, is bound to the value returned by the
5176 @code{(current-buffer)} expression. The variable, @code{oldbuf}, is
5177 used to keep track of the buffer in which you are working and from
5178 which you will copy.
5180 The element or elements of a varlist are surrounded by a set of
5181 parentheses so the Lisp interpreter can distinguish the varlist from
5182 the body of the @code{let}. As a consequence, the two-element list
5183 within the varlist is surrounded by a circumscribing set of parentheses.
5184 The line looks like this:
5188 (let ((oldbuf (current-buffer)))
5194 The two parentheses before @code{oldbuf} might surprise you if you did
5195 not realize that the first parenthesis before @code{oldbuf} marks the
5196 boundary of the varlist and the second parenthesis marks the beginning
5197 of the two-element list, @code{(oldbuf (current-buffer))}.
5199 @node append save-excursion
5200 @subsection @code{save-excursion} in @code{append-to-buffer}
5202 The body of the @code{let} expression in @code{append-to-buffer}
5203 consists of a @code{save-excursion} expression.
5205 The @code{save-excursion} function saves the locations of point and
5206 mark, and restores them to those positions after the expressions in the
5207 body of the @code{save-excursion} complete execution. In addition,
5208 @code{save-excursion} keeps track of the original buffer, and
5209 restores it. This is how @code{save-excursion} is used in
5210 @code{append-to-buffer}.
5213 @cindex Indentation for formatting
5214 @cindex Formatting convention
5215 Incidentally, it is worth noting here that a Lisp function is normally
5216 formatted so that everything that is enclosed in a multi-line spread is
5217 indented more to the right than the first symbol. In this function
5218 definition, the @code{let} is indented more than the @code{defun}, and
5219 the @code{save-excursion} is indented more than the @code{let}, like
5235 This formatting convention makes it easy to see that the lines in
5236 the body of the @code{save-excursion} are enclosed by the parentheses
5237 associated with @code{save-excursion}, just as the
5238 @code{save-excursion} itself is enclosed by the parentheses associated
5239 with the @code{let}:
5243 (let ((oldbuf (current-buffer)))
5246 (set-buffer @dots{})
5247 (insert-buffer-substring oldbuf start end)
5253 The use of the @code{save-excursion} function can be viewed as a process
5254 of filling in the slots of a template:
5259 @var{first-expression-in-body}
5260 @var{second-expression-in-body}
5262 @var{last-expression-in-body})
5268 In this function, the body of the @code{save-excursion} contains only
5269 one expression, the @code{let*} expression. You know about a
5270 @code{let} function. The @code{let*} function is different. It has a
5271 @samp{*} in its name. It enables Emacs to set each variable in its
5272 varlist in sequence, one after another.
5274 Its critical feature is that variables later in the varlist can make
5275 use of the values to which Emacs set variables earlier in the varlist.
5276 @xref{fwd-para let, , The @code{let*} expression}.
5278 We will skip functions like @code{let*} and focus on two: the
5279 @code{set-buffer} function and the @code{insert-buffer-substring}
5283 In the old days, the @code{set-buffer} expression was simply
5286 (set-buffer (get-buffer-create buffer))
5294 (set-buffer append-to)
5298 @code{append-to} is bound to @code{(get-buffer-create buffer)} earlier
5299 on in the @code{let*} expression. That extra binding would not be
5300 necessary except for that @code{append-to} is used later in the
5301 varlist as an argument to @code{get-buffer-window-list}.
5306 (let ((oldbuf (current-buffer)))
5308 (let* ((append-to (get-buffer-create buffer))
5309 (windows (get-buffer-window-list append-to t t))
5311 (set-buffer append-to)
5312 (setq point (point))
5313 (barf-if-buffer-read-only)
5314 (insert-buffer-substring oldbuf start end)
5315 (dolist (window windows)
5316 (when (= (window-point window) point)
5317 (set-window-point window (point))))))))
5320 The @code{append-to-buffer} function definition inserts text from the
5321 buffer in which you are currently to a named buffer. It happens that
5322 @code{insert-buffer-substring} copies text from another buffer to the
5323 current buffer, just the reverse---that is why the
5324 @code{append-to-buffer} definition starts out with a @code{let} that
5325 binds the local symbol @code{oldbuf} to the value returned by
5326 @code{current-buffer}.
5329 The @code{insert-buffer-substring} expression looks like this:
5332 (insert-buffer-substring oldbuf start end)
5336 The @code{insert-buffer-substring} function copies a string
5337 @emph{from} the buffer specified as its first argument and inserts the
5338 string into the present buffer. In this case, the argument to
5339 @code{insert-buffer-substring} is the value of the variable created
5340 and bound by the @code{let}, namely the value of @code{oldbuf}, which
5341 was the current buffer when you gave the @code{append-to-buffer}
5344 After @code{insert-buffer-substring} has done its work,
5345 @code{save-excursion} will restore the action to the original buffer
5346 and @code{append-to-buffer} will have done its job.
5349 Written in skeletal form, the workings of the body look like this:
5353 (let (@var{bind-}@code{oldbuf}@var{-to-value-of-}@code{current-buffer})
5354 (save-excursion ; @r{Keep track of buffer.}
5356 @var{insert-substring-from-}@code{oldbuf}@var{-into-buffer})
5358 @var{change-back-to-original-buffer-when-finished}
5359 @var{let-the-local-meaning-of-}@code{oldbuf}@var{-disappear-when-finished}
5363 In summary, @code{append-to-buffer} works as follows: it saves the
5364 value of the current buffer in the variable called @code{oldbuf}. It
5365 gets the new buffer (creating one if need be) and switches Emacs's
5366 attention to it. Using the value of @code{oldbuf}, it inserts the
5367 region of text from the old buffer into the new buffer; and then using
5368 @code{save-excursion}, it brings you back to your original buffer.
5370 In looking at @code{append-to-buffer}, you have explored a fairly
5371 complex function. It shows how to use @code{let} and
5372 @code{save-excursion}, and how to change to and come back from another
5373 buffer. Many function definitions use @code{let},
5374 @code{save-excursion}, and @code{set-buffer} this way.
5376 @node Buffer Related Review
5379 Here is a brief summary of the various functions discussed in this chapter.
5382 @item describe-function
5383 @itemx describe-variable
5384 Print the documentation for a function or variable.
5385 Conventionally bound to @kbd{C-h f} and @kbd{C-h v}.
5388 Find the file containing the source for a function or variable and
5389 switch buffers to it, positioning point at the beginning of the item.
5390 Conventionally bound to @kbd{M-.} (that's a period following the
5393 @item save-excursion
5394 Save the location of point and mark and restore their values after the
5395 arguments to @code{save-excursion} have been evaluated. Also, remember
5396 the current buffer and return to it.
5399 Set mark at a location and record the value of the previous mark on the
5400 mark ring. The mark is a location in the buffer that will keep its
5401 relative position even if text is added to or removed from the buffer.
5404 Set point to the location specified by the value of the argument, which
5405 can be a number, a marker, or an expression that returns the number of
5406 a position, such as @code{(point-min)}.
5408 @item insert-buffer-substring
5409 Copy a region of text from a buffer that is passed to the function as
5410 an argument and insert the region into the current buffer.
5412 @item mark-whole-buffer
5413 Mark the whole buffer as a region. Normally bound to @kbd{C-x h}.
5416 Switch the attention of Emacs to another buffer, but do not change the
5417 window being displayed. Used when the program rather than a human is
5418 to work on a different buffer.
5420 @item get-buffer-create
5422 Find a named buffer or create one if a buffer of that name does not
5423 exist. The @code{get-buffer} function returns @code{nil} if the named
5424 buffer does not exist.
5428 @node Buffer Exercises
5433 Write your own @code{simplified-end-of-buffer} function definition;
5434 then test it to see whether it works.
5437 Use @code{if} and @code{get-buffer} to write a function that prints a
5438 message telling you whether a buffer exists.
5441 Using @code{find-tag}, find the source for the @code{copy-to-buffer}
5446 @chapter A Few More Complex Functions
5448 In this chapter, we build on what we have learned in previous chapters
5449 by looking at more complex functions. The @code{copy-to-buffer}
5450 function illustrates use of two @code{save-excursion} expressions in
5451 one definition, while the @code{insert-buffer} function illustrates
5452 use of an asterisk in an @code{interactive} expression, use of
5453 @code{or}, and the important distinction between a name and the object
5454 to which the name refers.
5457 * copy-to-buffer:: With @code{set-buffer}, @code{get-buffer-create}.
5458 * insert-buffer:: Read-only, and with @code{or}.
5459 * beginning-of-buffer:: Shows @code{goto-char},
5460 @code{point-min}, and @code{push-mark}.
5461 * Second Buffer Related Review::
5462 * optional Exercise::
5465 @node copy-to-buffer
5466 @section The Definition of @code{copy-to-buffer}
5467 @findex copy-to-buffer
5469 After understanding how @code{append-to-buffer} works, it is easy to
5470 understand @code{copy-to-buffer}. This function copies text into a
5471 buffer, but instead of adding to the second buffer, it replaces all the
5472 previous text in the second buffer.
5475 The body of @code{copy-to-buffer} looks like this,
5480 (interactive "BCopy to buffer: \nr")
5481 (let ((oldbuf (current-buffer)))
5482 (with-current-buffer (get-buffer-create buffer)
5483 (barf-if-buffer-read-only)
5486 (insert-buffer-substring oldbuf start end)))))
5490 The @code{copy-to-buffer} function has a simpler @code{interactive}
5491 expression than @code{append-to-buffer}.
5494 The definition then says
5497 (with-current-buffer (get-buffer-create buffer) @dots{}
5500 First, look at the earliest inner expression; that is evaluated first.
5501 That expression starts with @code{get-buffer-create buffer}. The
5502 function tells the computer to use the buffer with the name specified
5503 as the one to which you are copying, or if such a buffer does not
5504 exist, to create it. Then, the @code{with-current-buffer} function
5505 evaluates its body with that buffer temporarily current.
5507 (This demonstrates another way to shift the computer's attention but
5508 not the user's. The @code{append-to-buffer} function showed how to do
5509 the same with @code{save-excursion} and @code{set-buffer}.
5510 @code{with-current-buffer} is a newer, and arguably easier,
5513 The @code{barf-if-buffer-read-only} function sends you an error
5514 message saying the buffer is read-only if you cannot modify it.
5516 The next line has the @code{erase-buffer} function as its sole
5517 contents. That function erases the buffer.
5519 Finally, the last two lines contain the @code{save-excursion}
5520 expression with @code{insert-buffer-substring} as its body.
5521 The @code{insert-buffer-substring} expression copies the text from
5522 the buffer you are in (and you have not seen the computer shift its
5523 attention, so you don't know that that buffer is now called
5526 Incidentally, this is what is meant by `replacement'. To replace text,
5527 Emacs erases the previous text and then inserts new text.
5530 In outline, the body of @code{copy-to-buffer} looks like this:
5534 (let (@var{bind-}@code{oldbuf}@var{-to-value-of-}@code{current-buffer})
5535 (@var{with-the-buffer-you-are-copying-to}
5536 (@var{but-do-not-erase-or-copy-to-a-read-only-buffer})
5539 @var{insert-substring-from-}@code{oldbuf}@var{-into-buffer})))
5544 @section The Definition of @code{insert-buffer}
5545 @findex insert-buffer
5547 @code{insert-buffer} is yet another buffer-related function. This
5548 command copies another buffer @emph{into} the current buffer. It is the
5549 reverse of @code{append-to-buffer} or @code{copy-to-buffer}, since they
5550 copy a region of text @emph{from} the current buffer to another buffer.
5552 Here is a discussion based on the original code. The code was
5553 simplified in 2003 and is harder to understand.
5555 (@xref{New insert-buffer, , New Body for @code{insert-buffer}}, to see
5556 a discussion of the new body.)
5558 In addition, this code illustrates the use of @code{interactive} with a
5559 buffer that might be @dfn{read-only} and the important distinction
5560 between the name of an object and the object actually referred to.
5563 * insert-buffer code::
5564 * insert-buffer interactive:: When you can read, but not write.
5565 * insert-buffer body:: The body has an @code{or} and a @code{let}.
5566 * if & or:: Using an @code{if} instead of an @code{or}.
5567 * Insert or:: How the @code{or} expression works.
5568 * Insert let:: Two @code{save-excursion} expressions.
5569 * New insert-buffer::
5573 @node insert-buffer code
5574 @unnumberedsubsec The Code for @code{insert-buffer}
5578 Here is the earlier code:
5582 (defun insert-buffer (buffer)
5583 "Insert after point the contents of BUFFER.
5584 Puts mark after the inserted text.
5585 BUFFER may be a buffer or a buffer name."
5586 (interactive "*bInsert buffer:@: ")
5589 (or (bufferp buffer)
5590 (setq buffer (get-buffer buffer)))
5591 (let (start end newmark)
5595 (setq start (point-min) end (point-max)))
5598 (insert-buffer-substring buffer start end)
5599 (setq newmark (point)))
5600 (push-mark newmark)))
5605 As with other function definitions, you can use a template to see an
5606 outline of the function:
5610 (defun insert-buffer (buffer)
5611 "@var{documentation}@dots{}"
5612 (interactive "*bInsert buffer:@: ")
5617 @node insert-buffer interactive
5618 @subsection The Interactive Expression in @code{insert-buffer}
5619 @findex interactive, @r{example use of}
5621 In @code{insert-buffer}, the argument to the @code{interactive}
5622 declaration has two parts, an asterisk, @samp{*}, and @samp{bInsert
5626 * Read-only buffer:: When a buffer cannot be modified.
5627 * b for interactive:: An existing buffer or else its name.
5630 @node Read-only buffer
5631 @unnumberedsubsubsec A Read-only Buffer
5632 @cindex Read-only buffer
5633 @cindex Asterisk for read-only buffer
5634 @findex * @r{for read-only buffer}
5636 The asterisk is for the situation when the current buffer is a
5637 read-only buffer---a buffer that cannot be modified. If
5638 @code{insert-buffer} is called when the current buffer is read-only, a
5639 message to this effect is printed in the echo area and the terminal
5640 may beep or blink at you; you will not be permitted to insert anything
5641 into current buffer. The asterisk does not need to be followed by a
5642 newline to separate it from the next argument.
5644 @node b for interactive
5645 @unnumberedsubsubsec @samp{b} in an Interactive Expression
5647 The next argument in the interactive expression starts with a lower
5648 case @samp{b}. (This is different from the code for
5649 @code{append-to-buffer}, which uses an upper-case @samp{B}.
5650 @xref{append-to-buffer, , The Definition of @code{append-to-buffer}}.)
5651 The lower-case @samp{b} tells the Lisp interpreter that the argument
5652 for @code{insert-buffer} should be an existing buffer or else its
5653 name. (The upper-case @samp{B} option provides for the possibility
5654 that the buffer does not exist.) Emacs will prompt you for the name
5655 of the buffer, offering you a default buffer, with name completion
5656 enabled. If the buffer does not exist, you receive a message that
5657 says ``No match''; your terminal may beep at you as well.
5659 The new and simplified code generates a list for @code{interactive}.
5660 It uses the @code{barf-if-buffer-read-only} and @code{read-buffer}
5661 functions with which we are already familiar and the @code{progn}
5662 special form with which we are not. (It will be described later.)
5664 @node insert-buffer body
5665 @subsection The Body of the @code{insert-buffer} Function
5667 The body of the @code{insert-buffer} function has two major parts: an
5668 @code{or} expression and a @code{let} expression. The purpose of the
5669 @code{or} expression is to ensure that the argument @code{buffer} is
5670 bound to a buffer and not just the name of a buffer. The body of the
5671 @code{let} expression contains the code which copies the other buffer
5672 into the current buffer.
5675 In outline, the two expressions fit into the @code{insert-buffer}
5680 (defun insert-buffer (buffer)
5681 "@var{documentation}@dots{}"
5682 (interactive "*bInsert buffer:@: ")
5687 (let (@var{varlist})
5688 @var{body-of-}@code{let}@dots{} )
5692 To understand how the @code{or} expression ensures that the argument
5693 @code{buffer} is bound to a buffer and not to the name of a buffer, it
5694 is first necessary to understand the @code{or} function.
5696 Before doing this, let me rewrite this part of the function using
5697 @code{if} so that you can see what is done in a manner that will be familiar.
5700 @subsection @code{insert-buffer} With an @code{if} Instead of an @code{or}
5702 The job to be done is to make sure the value of @code{buffer} is a
5703 buffer itself and not the name of a buffer. If the value is the name,
5704 then the buffer itself must be got.
5706 You can imagine yourself at a conference where an usher is wandering
5707 around holding a list with your name on it and looking for you: the
5708 usher is ``bound'' to your name, not to you; but when the usher finds
5709 you and takes your arm, the usher becomes ``bound'' to you.
5712 In Lisp, you might describe this situation like this:
5716 (if (not (holding-on-to-guest))
5717 (find-and-take-arm-of-guest))
5721 We want to do the same thing with a buffer---if we do not have the
5722 buffer itself, we want to get it.
5725 Using a predicate called @code{bufferp} that tells us whether we have a
5726 buffer (rather than its name), we can write the code like this:
5730 (if (not (bufferp buffer)) ; @r{if-part}
5731 (setq buffer (get-buffer buffer))) ; @r{then-part}
5736 Here, the true-or-false-test of the @code{if} expression is
5737 @w{@code{(not (bufferp buffer))}}; and the then-part is the expression
5738 @w{@code{(setq buffer (get-buffer buffer))}}.
5740 In the test, the function @code{bufferp} returns true if its argument is
5741 a buffer---but false if its argument is the name of the buffer. (The
5742 last character of the function name @code{bufferp} is the character
5743 @samp{p}; as we saw earlier, such use of @samp{p} is a convention that
5744 indicates that the function is a predicate, which is a term that means
5745 that the function will determine whether some property is true or false.
5746 @xref{Wrong Type of Argument, , Using the Wrong Type Object as an
5750 The function @code{not} precedes the expression @code{(bufferp buffer)},
5751 so the true-or-false-test looks like this:
5754 (not (bufferp buffer))
5758 @code{not} is a function that returns true if its argument is false
5759 and false if its argument is true. So if @code{(bufferp buffer)}
5760 returns true, the @code{not} expression returns false and vice-verse:
5761 what is ``not true'' is false and what is ``not false'' is true.
5763 Using this test, the @code{if} expression works as follows: when the
5764 value of the variable @code{buffer} is actually a buffer rather than
5765 its name, the true-or-false-test returns false and the @code{if}
5766 expression does not evaluate the then-part. This is fine, since we do
5767 not need to do anything to the variable @code{buffer} if it really is
5770 On the other hand, when the value of @code{buffer} is not a buffer
5771 itself, but the name of a buffer, the true-or-false-test returns true
5772 and the then-part of the expression is evaluated. In this case, the
5773 then-part is @code{(setq buffer (get-buffer buffer))}. This
5774 expression uses the @code{get-buffer} function to return an actual
5775 buffer itself, given its name. The @code{setq} then sets the variable
5776 @code{buffer} to the value of the buffer itself, replacing its previous
5777 value (which was the name of the buffer).
5780 @subsection The @code{or} in the Body
5782 The purpose of the @code{or} expression in the @code{insert-buffer}
5783 function is to ensure that the argument @code{buffer} is bound to a
5784 buffer and not just to the name of a buffer. The previous section shows
5785 how the job could have been done using an @code{if} expression.
5786 However, the @code{insert-buffer} function actually uses @code{or}.
5787 To understand this, it is necessary to understand how @code{or} works.
5790 An @code{or} function can have any number of arguments. It evaluates
5791 each argument in turn and returns the value of the first of its
5792 arguments that is not @code{nil}. Also, and this is a crucial feature
5793 of @code{or}, it does not evaluate any subsequent arguments after
5794 returning the first non-@code{nil} value.
5797 The @code{or} expression looks like this:
5801 (or (bufferp buffer)
5802 (setq buffer (get-buffer buffer)))
5807 The first argument to @code{or} is the expression @code{(bufferp buffer)}.
5808 This expression returns true (a non-@code{nil} value) if the buffer is
5809 actually a buffer, and not just the name of a buffer. In the @code{or}
5810 expression, if this is the case, the @code{or} expression returns this
5811 true value and does not evaluate the next expression---and this is fine
5812 with us, since we do not want to do anything to the value of
5813 @code{buffer} if it really is a buffer.
5815 On the other hand, if the value of @code{(bufferp buffer)} is @code{nil},
5816 which it will be if the value of @code{buffer} is the name of a buffer,
5817 the Lisp interpreter evaluates the next element of the @code{or}
5818 expression. This is the expression @code{(setq buffer (get-buffer
5819 buffer))}. This expression returns a non-@code{nil} value, which
5820 is the value to which it sets the variable @code{buffer}---and this
5821 value is a buffer itself, not the name of a buffer.
5823 The result of all this is that the symbol @code{buffer} is always
5824 bound to a buffer itself rather than to the name of a buffer. All
5825 this is necessary because the @code{set-buffer} function in a
5826 following line only works with a buffer itself, not with the name to a
5830 Incidentally, using @code{or}, the situation with the usher would be
5834 (or (holding-on-to-guest) (find-and-take-arm-of-guest))
5838 @subsection The @code{let} Expression in @code{insert-buffer}
5840 After ensuring that the variable @code{buffer} refers to a buffer itself
5841 and not just to the name of a buffer, the @code{insert-buffer function}
5842 continues with a @code{let} expression. This specifies three local
5843 variables, @code{start}, @code{end}, and @code{newmark} and binds them
5844 to the initial value @code{nil}. These variables are used inside the
5845 remainder of the @code{let} and temporarily hide any other occurrence of
5846 variables of the same name in Emacs until the end of the @code{let}.
5849 The body of the @code{let} contains two @code{save-excursion}
5850 expressions. First, we will look at the inner @code{save-excursion}
5851 expression in detail. The expression looks like this:
5857 (setq start (point-min) end (point-max)))
5862 The expression @code{(set-buffer buffer)} changes Emacs's attention
5863 from the current buffer to the one from which the text will copied.
5864 In that buffer, the variables @code{start} and @code{end} are set to
5865 the beginning and end of the buffer, using the commands
5866 @code{point-min} and @code{point-max}. Note that we have here an
5867 illustration of how @code{setq} is able to set two variables in the
5868 same expression. The first argument of @code{setq} is set to the
5869 value of its second, and its third argument is set to the value of its
5872 After the body of the inner @code{save-excursion} is evaluated, the
5873 @code{save-excursion} restores the original buffer, but @code{start} and
5874 @code{end} remain set to the values of the beginning and end of the
5875 buffer from which the text will be copied.
5878 The outer @code{save-excursion} expression looks like this:
5883 (@var{inner-}@code{save-excursion}@var{-expression}
5884 (@var{go-to-new-buffer-and-set-}@code{start}@var{-and-}@code{end})
5885 (insert-buffer-substring buffer start end)
5886 (setq newmark (point)))
5891 The @code{insert-buffer-substring} function copies the text
5892 @emph{into} the current buffer @emph{from} the region indicated by
5893 @code{start} and @code{end} in @code{buffer}. Since the whole of the
5894 second buffer lies between @code{start} and @code{end}, the whole of
5895 the second buffer is copied into the buffer you are editing. Next,
5896 the value of point, which will be at the end of the inserted text, is
5897 recorded in the variable @code{newmark}.
5899 After the body of the outer @code{save-excursion} is evaluated, point
5900 and mark are relocated to their original places.
5902 However, it is convenient to locate a mark at the end of the newly
5903 inserted text and locate point at its beginning. The @code{newmark}
5904 variable records the end of the inserted text. In the last line of
5905 the @code{let} expression, the @code{(push-mark newmark)} expression
5906 function sets a mark to this location. (The previous location of the
5907 mark is still accessible; it is recorded on the mark ring and you can
5908 go back to it with @kbd{C-u C-@key{SPC}}.) Meanwhile, point is
5909 located at the beginning of the inserted text, which is where it was
5910 before you called the insert function, the position of which was saved
5911 by the first @code{save-excursion}.
5914 The whole @code{let} expression looks like this:
5918 (let (start end newmark)
5922 (setq start (point-min) end (point-max)))
5923 (insert-buffer-substring buffer start end)
5924 (setq newmark (point)))
5925 (push-mark newmark))
5929 Like the @code{append-to-buffer} function, the @code{insert-buffer}
5930 function uses @code{let}, @code{save-excursion}, and
5931 @code{set-buffer}. In addition, the function illustrates one way to
5932 use @code{or}. All these functions are building blocks that we will
5933 find and use again and again.
5935 @node New insert-buffer
5936 @subsection New Body for @code{insert-buffer}
5937 @findex insert-buffer, new version body
5938 @findex new version body for insert-buffer
5940 The body in the GNU Emacs 22 version is more confusing than the original.
5943 It consists of two expressions,
5949 (insert-buffer-substring (get-buffer buffer))
5957 except, and this is what confuses novices, very important work is done
5958 inside the @code{push-mark} expression.
5960 The @code{get-buffer} function returns a buffer with the name
5961 provided. You will note that the function is @emph{not} called
5962 @code{get-buffer-create}; it does not create a buffer if one does not
5963 already exist. The buffer returned by @code{get-buffer}, an existing
5964 buffer, is passed to @code{insert-buffer-substring}, which inserts the
5965 whole of the buffer (since you did not specify anything else).
5967 The location into which the buffer is inserted is recorded by
5968 @code{push-mark}. Then the function returns @code{nil}, the value of
5969 its last command. Put another way, the @code{insert-buffer} function
5970 exists only to produce a side effect, inserting another buffer, not to
5973 @node beginning-of-buffer
5974 @section Complete Definition of @code{beginning-of-buffer}
5975 @findex beginning-of-buffer
5977 The basic structure of the @code{beginning-of-buffer} function has
5978 already been discussed. (@xref{simplified-beginning-of-buffer, , A
5979 Simplified @code{beginning-of-buffer} Definition}.)
5980 This section describes the complex part of the definition.
5982 As previously described, when invoked without an argument,
5983 @code{beginning-of-buffer} moves the cursor to the beginning of the
5984 buffer (in truth, the beginning of the accessible portion of the
5985 buffer), leaving the mark at the previous position. However, when the
5986 command is invoked with a number between one and ten, the function
5987 considers that number to be a fraction of the length of the buffer,
5988 measured in tenths, and Emacs moves the cursor that fraction of the
5989 way from the beginning of the buffer. Thus, you can either call this
5990 function with the key command @kbd{M-<}, which will move the cursor to
5991 the beginning of the buffer, or with a key command such as @kbd{C-u 7
5992 M-<} which will move the cursor to a point 70% of the way through the
5993 buffer. If a number bigger than ten is used for the argument, it
5994 moves to the end of the buffer.
5996 The @code{beginning-of-buffer} function can be called with or without an
5997 argument. The use of the argument is optional.
6000 * Optional Arguments::
6001 * beginning-of-buffer opt arg:: Example with optional argument.
6002 * beginning-of-buffer complete::
6005 @node Optional Arguments
6006 @subsection Optional Arguments
6008 Unless told otherwise, Lisp expects that a function with an argument in
6009 its function definition will be called with a value for that argument.
6010 If that does not happen, you get an error and a message that says
6011 @samp{Wrong number of arguments}.
6013 @cindex Optional arguments
6016 However, optional arguments are a feature of Lisp: a particular
6017 @dfn{keyword} is used to tell the Lisp interpreter that an argument is
6018 optional. The keyword is @code{&optional}. (The @samp{&} in front of
6019 @samp{optional} is part of the keyword.) In a function definition, if
6020 an argument follows the keyword @code{&optional}, no value need be
6021 passed to that argument when the function is called.
6024 The first line of the function definition of @code{beginning-of-buffer}
6025 therefore looks like this:
6028 (defun beginning-of-buffer (&optional arg)
6032 In outline, the whole function looks like this:
6036 (defun beginning-of-buffer (&optional arg)
6037 "@var{documentation}@dots{}"
6039 (or (@var{is-the-argument-a-cons-cell} arg)
6040 (and @var{are-both-transient-mark-mode-and-mark-active-true})
6042 (let (@var{determine-size-and-set-it})
6044 (@var{if-there-is-an-argument}
6045 @var{figure-out-where-to-go}
6052 The function is similar to the @code{simplified-beginning-of-buffer}
6053 function except that the @code{interactive} expression has @code{"P"}
6054 as an argument and the @code{goto-char} function is followed by an
6055 if-then-else expression that figures out where to put the cursor if
6056 there is an argument that is not a cons cell.
6058 (Since I do not explain a cons cell for many more chapters, please
6059 consider ignoring the function @code{consp}. @xref{List
6060 Implementation, , How Lists are Implemented}, and @ref{Cons Cell Type,
6061 , Cons Cell and List Types, elisp, The GNU Emacs Lisp Reference
6064 The @code{"P"} in the @code{interactive} expression tells Emacs to
6065 pass a prefix argument, if there is one, to the function in raw form.
6066 A prefix argument is made by typing the @key{META} key followed by a
6067 number, or by typing @kbd{C-u} and then a number. (If you don't type
6068 a number, @kbd{C-u} defaults to a cons cell with a 4. A lowercase
6069 @code{"p"} in the @code{interactive} expression causes the function to
6070 convert a prefix arg to a number.)
6072 The true-or-false-test of the @code{if} expression looks complex, but
6073 it is not: it checks whether @code{arg} has a value that is not
6074 @code{nil} and whether it is a cons cell. (That is what @code{consp}
6075 does; it checks whether its argument is a cons cell.) If @code{arg}
6076 has a value that is not @code{nil} (and is not a cons cell), which
6077 will be the case if @code{beginning-of-buffer} is called with a
6078 numeric argument, then this true-or-false-test will return true and
6079 the then-part of the @code{if} expression will be evaluated. On the
6080 other hand, if @code{beginning-of-buffer} is not called with an
6081 argument, the value of @code{arg} will be @code{nil} and the else-part
6082 of the @code{if} expression will be evaluated. The else-part is
6083 simply @code{point-min}, and when this is the outcome, the whole
6084 @code{goto-char} expression is @code{(goto-char (point-min))}, which
6085 is how we saw the @code{beginning-of-buffer} function in its
6088 @node beginning-of-buffer opt arg
6089 @subsection @code{beginning-of-buffer} with an Argument
6091 When @code{beginning-of-buffer} is called with an argument, an
6092 expression is evaluated which calculates what value to pass to
6093 @code{goto-char}. This expression is rather complicated at first sight.
6094 It includes an inner @code{if} expression and much arithmetic. It looks
6099 (if (> (buffer-size) 10000)
6100 ;; @r{Avoid overflow for large buffer sizes!}
6101 (* (prefix-numeric-value arg)
6106 size (prefix-numeric-value arg))) 10)))
6111 * Disentangle beginning-of-buffer::
6112 * Large buffer case::
6113 * Small buffer case::
6117 @node Disentangle beginning-of-buffer
6118 @unnumberedsubsubsec Disentangle @code{beginning-of-buffer}
6121 Like other complex-looking expressions, the conditional expression
6122 within @code{beginning-of-buffer} can be disentangled by looking at it
6123 as parts of a template, in this case, the template for an if-then-else
6124 expression. In skeletal form, the expression looks like this:
6128 (if (@var{buffer-is-large}
6129 @var{divide-buffer-size-by-10-and-multiply-by-arg}
6130 @var{else-use-alternate-calculation}
6134 The true-or-false-test of this inner @code{if} expression checks the
6135 size of the buffer. The reason for this is that the old version 18
6136 Emacs used numbers that are no bigger than eight million or so and in
6137 the computation that followed, the programmer feared that Emacs might
6138 try to use over-large numbers if the buffer were large. The term
6139 `overflow', mentioned in the comment, means numbers that are over
6140 large. More recent versions of Emacs use larger numbers, but this
6141 code has not been touched, if only because people now look at buffers
6142 that are far, far larger than ever before.
6144 There are two cases: if the buffer is large and if it is not.
6146 @node Large buffer case
6147 @unnumberedsubsubsec What happens in a large buffer
6149 In @code{beginning-of-buffer}, the inner @code{if} expression tests
6150 whether the size of the buffer is greater than 10,000 characters. To do
6151 this, it uses the @code{>} function and the computation of @code{size}
6152 that comes from the let expression.
6154 In the old days, the function @code{buffer-size} was used. Not only
6155 was that function called several times, it gave the size of the whole
6156 buffer, not the accessible part. The computation makes much more
6157 sense when it handles just the accessible part. (@xref{Narrowing &
6158 Widening, , Narrowing and Widening}, for more information on focusing
6159 attention to an `accessible' part.)
6162 The line looks like this:
6170 When the buffer is large, the then-part of the @code{if} expression is
6171 evaluated. It reads like this (after formatting for easy reading):
6176 (prefix-numeric-value arg)
6182 This expression is a multiplication, with two arguments to the function
6185 The first argument is @code{(prefix-numeric-value arg)}. When
6186 @code{"P"} is used as the argument for @code{interactive}, the value
6187 passed to the function as its argument is passed a ``raw prefix
6188 argument'', and not a number. (It is a number in a list.) To perform
6189 the arithmetic, a conversion is necessary, and
6190 @code{prefix-numeric-value} does the job.
6192 @findex / @r{(division)}
6194 The second argument is @code{(/ size 10)}. This expression divides
6195 the numeric value by ten---the numeric value of the size of the
6196 accessible portion of the buffer. This produces a number that tells
6197 how many characters make up one tenth of the buffer size. (In Lisp,
6198 @code{/} is used for division, just as @code{*} is used for
6202 In the multiplication expression as a whole, this amount is multiplied
6203 by the value of the prefix argument---the multiplication looks like this:
6207 (* @var{numeric-value-of-prefix-arg}
6208 @var{number-of-characters-in-one-tenth-of-the-accessible-buffer})
6213 If, for example, the prefix argument is @samp{7}, the one-tenth value
6214 will be multiplied by 7 to give a position 70% of the way through.
6217 The result of all this is that if the accessible portion of the buffer
6218 is large, the @code{goto-char} expression reads like this:
6222 (goto-char (* (prefix-numeric-value arg)
6227 This puts the cursor where we want it.
6229 @node Small buffer case
6230 @unnumberedsubsubsec What happens in a small buffer
6232 If the buffer contains fewer than 10,000 characters, a slightly
6233 different computation is performed. You might think this is not
6234 necessary, since the first computation could do the job. However, in
6235 a small buffer, the first method may not put the cursor on exactly the
6236 desired line; the second method does a better job.
6239 The code looks like this:
6241 @c Keep this on one line.
6243 (/ (+ 10 (* size (prefix-numeric-value arg))) 10))
6248 This is code in which you figure out what happens by discovering how the
6249 functions are embedded in parentheses. It is easier to read if you
6250 reformat it with each expression indented more deeply than its
6251 enclosing expression:
6259 (prefix-numeric-value arg)))
6266 Looking at parentheses, we see that the innermost operation is
6267 @code{(prefix-numeric-value arg)}, which converts the raw argument to
6268 a number. In the following expression, this number is multiplied by
6269 the size of the accessible portion of the buffer:
6272 (* size (prefix-numeric-value arg))
6276 This multiplication creates a number that may be larger than the size of
6277 the buffer---seven times larger if the argument is 7, for example. Ten
6278 is then added to this number and finally the large number is divided by
6279 ten to provide a value that is one character larger than the percentage
6280 position in the buffer.
6282 The number that results from all this is passed to @code{goto-char} and
6283 the cursor is moved to that point.
6286 @node beginning-of-buffer complete
6287 @subsection The Complete @code{beginning-of-buffer}
6290 Here is the complete text of the @code{beginning-of-buffer} function:
6296 (defun beginning-of-buffer (&optional arg)
6297 "Move point to the beginning of the buffer;
6298 leave mark at previous position.
6299 With \\[universal-argument] prefix,
6300 do not set mark at previous position.
6302 put point N/10 of the way from the beginning.
6304 If the buffer is narrowed,
6305 this command uses the beginning and size
6306 of the accessible part of the buffer.
6310 Don't use this command in Lisp programs!
6311 \(goto-char (point-min)) is faster
6312 and avoids clobbering the mark."
6315 (and transient-mark-mode mark-active)
6319 (let ((size (- (point-max) (point-min))))
6320 (goto-char (if (and arg (not (consp arg)))
6323 ;; Avoid overflow for large buffer sizes!
6324 (* (prefix-numeric-value arg)
6326 (/ (+ 10 (* size (prefix-numeric-value arg)))
6329 (if (and arg (not (consp arg))) (forward-line 1)))
6334 From before GNU Emacs 22
6337 (defun beginning-of-buffer (&optional arg)
6338 "Move point to the beginning of the buffer;
6339 leave mark at previous position.
6340 With arg N, put point N/10 of the way
6341 from the true beginning.
6344 Don't use this in Lisp programs!
6345 \(goto-char (point-min)) is faster
6346 and does not set the mark."
6353 (if (> (buffer-size) 10000)
6354 ;; @r{Avoid overflow for large buffer sizes!}
6355 (* (prefix-numeric-value arg)
6356 (/ (buffer-size) 10))
6359 (/ (+ 10 (* (buffer-size)
6360 (prefix-numeric-value arg)))
6363 (if arg (forward-line 1)))
6369 Except for two small points, the previous discussion shows how this
6370 function works. The first point deals with a detail in the
6371 documentation string, and the second point concerns the last line of
6375 In the documentation string, there is reference to an expression:
6378 \\[universal-argument]
6382 A @samp{\\} is used before the first square bracket of this
6383 expression. This @samp{\\} tells the Lisp interpreter to substitute
6384 whatever key is currently bound to the @samp{[@dots{}]}. In the case
6385 of @code{universal-argument}, that is usually @kbd{C-u}, but it might
6386 be different. (@xref{Documentation Tips, , Tips for Documentation
6387 Strings, elisp, The GNU Emacs Lisp Reference Manual}, for more
6391 Finally, the last line of the @code{beginning-of-buffer} command says
6392 to move point to the beginning of the next line if the command is
6393 invoked with an argument:
6396 (if (and arg (not (consp arg))) (forward-line 1))
6400 This puts the cursor at the beginning of the first line after the
6401 appropriate tenths position in the buffer. This is a flourish that
6402 means that the cursor is always located @emph{at least} the requested
6403 tenths of the way through the buffer, which is a nicety that is,
6404 perhaps, not necessary, but which, if it did not occur, would be sure
6405 to draw complaints. (The @code{(not (consp arg))} portion is so that
6406 if you specify the command with a @kbd{C-u}, but without a number,
6407 that is to say, if the `raw prefix argument' is simply a cons cell,
6408 the command does not put you at the beginning of the second line.)
6410 @node Second Buffer Related Review
6413 Here is a brief summary of some of the topics covered in this chapter.
6417 Evaluate each argument in sequence, and return the value of the first
6418 argument that is not @code{nil}; if none return a value that is not
6419 @code{nil}, return @code{nil}. In brief, return the first true value
6420 of the arguments; return a true value if one @emph{or} any of the
6424 Evaluate each argument in sequence, and if any are @code{nil}, return
6425 @code{nil}; if none are @code{nil}, return the value of the last
6426 argument. In brief, return a true value only if all the arguments are
6427 true; return a true value if one @emph{and} each of the others is
6431 A keyword used to indicate that an argument to a function definition
6432 is optional; this means that the function can be evaluated without the
6433 argument, if desired.
6435 @item prefix-numeric-value
6436 Convert the `raw prefix argument' produced by @code{(interactive
6437 "P")} to a numeric value.
6440 Move point forward to the beginning of the next line, or if the argument
6441 is greater than one, forward that many lines. If it can't move as far
6442 forward as it is supposed to, @code{forward-line} goes forward as far as
6443 it can and then returns a count of the number of additional lines it was
6444 supposed to move but couldn't.
6447 Delete the entire contents of the current buffer.
6450 Return @code{t} if its argument is a buffer; otherwise return @code{nil}.
6453 @node optional Exercise
6454 @section @code{optional} Argument Exercise
6456 Write an interactive function with an optional argument that tests
6457 whether its argument, a number, is greater than or equal to, or else,
6458 less than the value of @code{fill-column}, and tells you which, in a
6459 message. However, if you do not pass an argument to the function, use
6460 56 as a default value.
6462 @node Narrowing & Widening
6463 @chapter Narrowing and Widening
6464 @cindex Focusing attention (narrowing)
6468 Narrowing is a feature of Emacs that makes it possible for you to focus
6469 on a specific part of a buffer, and work without accidentally changing
6470 other parts. Narrowing is normally disabled since it can confuse
6474 * Narrowing advantages:: The advantages of narrowing
6475 * save-restriction:: The @code{save-restriction} special form.
6476 * what-line:: The number of the line that point is on.
6481 @node Narrowing advantages
6482 @unnumberedsec The Advantages of Narrowing
6485 With narrowing, the rest of a buffer is made invisible, as if it weren't
6486 there. This is an advantage if, for example, you want to replace a word
6487 in one part of a buffer but not in another: you narrow to the part you want
6488 and the replacement is carried out only in that section, not in the rest
6489 of the buffer. Searches will only work within a narrowed region, not
6490 outside of one, so if you are fixing a part of a document, you can keep
6491 yourself from accidentally finding parts you do not need to fix by
6492 narrowing just to the region you want.
6493 (The key binding for @code{narrow-to-region} is @kbd{C-x n n}.)
6495 However, narrowing does make the rest of the buffer invisible, which
6496 can scare people who inadvertently invoke narrowing and think they
6497 have deleted a part of their file. Moreover, the @code{undo} command
6498 (which is usually bound to @kbd{C-x u}) does not turn off narrowing
6499 (nor should it), so people can become quite desperate if they do not
6500 know that they can return the rest of a buffer to visibility with the
6501 @code{widen} command.
6502 (The key binding for @code{widen} is @kbd{C-x n w}.)
6504 Narrowing is just as useful to the Lisp interpreter as to a human.
6505 Often, an Emacs Lisp function is designed to work on just part of a
6506 buffer; or conversely, an Emacs Lisp function needs to work on all of a
6507 buffer that has been narrowed. The @code{what-line} function, for
6508 example, removes the narrowing from a buffer, if it has any narrowing
6509 and when it has finished its job, restores the narrowing to what it was.
6510 On the other hand, the @code{count-lines} function
6511 uses narrowing to restrict itself to just that portion
6512 of the buffer in which it is interested and then restores the previous
6515 @node save-restriction
6516 @section The @code{save-restriction} Special Form
6517 @findex save-restriction
6519 In Emacs Lisp, you can use the @code{save-restriction} special form to
6520 keep track of whatever narrowing is in effect, if any. When the Lisp
6521 interpreter meets with @code{save-restriction}, it executes the code
6522 in the body of the @code{save-restriction} expression, and then undoes
6523 any changes to narrowing that the code caused. If, for example, the
6524 buffer is narrowed and the code that follows @code{save-restriction}
6525 gets rid of the narrowing, @code{save-restriction} returns the buffer
6526 to its narrowed region afterwards. In the @code{what-line} command,
6527 any narrowing the buffer may have is undone by the @code{widen}
6528 command that immediately follows the @code{save-restriction} command.
6529 Any original narrowing is restored just before the completion of the
6533 The template for a @code{save-restriction} expression is simple:
6543 The body of the @code{save-restriction} is one or more expressions that
6544 will be evaluated in sequence by the Lisp interpreter.
6546 Finally, a point to note: when you use both @code{save-excursion} and
6547 @code{save-restriction}, one right after the other, you should use
6548 @code{save-excursion} outermost. If you write them in reverse order,
6549 you may fail to record narrowing in the buffer to which Emacs switches
6550 after calling @code{save-excursion}. Thus, when written together,
6551 @code{save-excursion} and @code{save-restriction} should be written
6562 In other circumstances, when not written together, the
6563 @code{save-excursion} and @code{save-restriction} special forms must
6564 be written in the order appropriate to the function.
6580 /usr/local/src/emacs/lisp/simple.el
6583 "Print the current buffer line number and narrowed line number of point."
6585 (let ((start (point-min))
6586 (n (line-number-at-pos)))
6588 (message "Line %d" n)
6592 (message "line %d (narrowed line %d)"
6593 (+ n (line-number-at-pos start) -1) n))))))
6595 (defun line-number-at-pos (&optional pos)
6596 "Return (narrowed) buffer line number at position POS.
6597 If POS is nil, use current buffer location.
6598 Counting starts at (point-min), so the value refers
6599 to the contents of the accessible portion of the buffer."
6600 (let ((opoint (or pos (point))) start)
6602 (goto-char (point-min))
6603 (setq start (point))
6606 (1+ (count-lines start (point))))))
6608 (defun count-lines (start end)
6609 "Return number of lines between START and END.
6610 This is usually the number of newlines between them,
6611 but can be one more if START is not equal to END
6612 and the greater of them is not at the start of a line."
6615 (narrow-to-region start end)
6616 (goto-char (point-min))
6617 (if (eq selective-display t)
6620 (while (re-search-forward "[\n\C-m]" nil t 40)
6621 (setq done (+ 40 done)))
6622 (while (re-search-forward "[\n\C-m]" nil t 1)
6623 (setq done (+ 1 done)))
6624 (goto-char (point-max))
6625 (if (and (/= start end)
6629 (- (buffer-size) (forward-line (buffer-size)))))))
6633 @section @code{what-line}
6635 @cindex Widening, example of
6637 The @code{what-line} command tells you the number of the line in which
6638 the cursor is located. The function illustrates the use of the
6639 @code{save-restriction} and @code{save-excursion} commands. Here is the
6640 original text of the function:
6645 "Print the current line number (in the buffer) of point."
6652 (1+ (count-lines 1 (point)))))))
6656 (In recent versions of GNU Emacs, the @code{what-line} function has
6657 been expanded to tell you your line number in a narrowed buffer as
6658 well as your line number in a widened buffer. The recent version is
6659 more complex than the version shown here. If you feel adventurous,
6660 you might want to look at it after figuring out how this version
6661 works. You will probably need to use @kbd{C-h f}
6662 (@code{describe-function}). The newer version uses a conditional to
6663 determine whether the buffer has been narrowed.
6665 (Also, it uses @code{line-number-at-pos}, which among other simple
6666 expressions, such as @code{(goto-char (point-min))}, moves point to
6667 the beginning of the current line with @code{(forward-line 0)} rather
6668 than @code{beginning-of-line}.)
6670 The @code{what-line} function as shown here has a documentation line
6671 and is interactive, as you would expect. The next two lines use the
6672 functions @code{save-restriction} and @code{widen}.
6674 The @code{save-restriction} special form notes whatever narrowing is in
6675 effect, if any, in the current buffer and restores that narrowing after
6676 the code in the body of the @code{save-restriction} has been evaluated.
6678 The @code{save-restriction} special form is followed by @code{widen}.
6679 This function undoes any narrowing the current buffer may have had
6680 when @code{what-line} was called. (The narrowing that was there is
6681 the narrowing that @code{save-restriction} remembers.) This widening
6682 makes it possible for the line counting commands to count from the
6683 beginning of the buffer. Otherwise, they would have been limited to
6684 counting within the accessible region. Any original narrowing is
6685 restored just before the completion of the function by the
6686 @code{save-restriction} special form.
6688 The call to @code{widen} is followed by @code{save-excursion}, which
6689 saves the location of the cursor (i.e., of point) and of the mark, and
6690 restores them after the code in the body of the @code{save-excursion}
6691 uses the @code{beginning-of-line} function to move point.
6693 (Note that the @code{(widen)} expression comes between the
6694 @code{save-restriction} and @code{save-excursion} special forms. When
6695 you write the two @code{save- @dots{}} expressions in sequence, write
6696 @code{save-excursion} outermost.)
6699 The last two lines of the @code{what-line} function are functions to
6700 count the number of lines in the buffer and then print the number in the
6706 (1+ (count-lines 1 (point)))))))
6710 The @code{message} function prints a one-line message at the bottom of
6711 the Emacs screen. The first argument is inside of quotation marks and
6712 is printed as a string of characters. However, it may contain a
6713 @samp{%d} expression to print a following argument. @samp{%d} prints
6714 the argument as a decimal, so the message will say something such as
6718 The number that is printed in place of the @samp{%d} is computed by the
6719 last line of the function:
6722 (1+ (count-lines 1 (point)))
6728 (defun count-lines (start end)
6729 "Return number of lines between START and END.
6730 This is usually the number of newlines between them,
6731 but can be one more if START is not equal to END
6732 and the greater of them is not at the start of a line."
6735 (narrow-to-region start end)
6736 (goto-char (point-min))
6737 (if (eq selective-display t)
6740 (while (re-search-forward "[\n\C-m]" nil t 40)
6741 (setq done (+ 40 done)))
6742 (while (re-search-forward "[\n\C-m]" nil t 1)
6743 (setq done (+ 1 done)))
6744 (goto-char (point-max))
6745 (if (and (/= start end)
6749 (- (buffer-size) (forward-line (buffer-size)))))))
6753 What this does is count the lines from the first position of the
6754 buffer, indicated by the @code{1}, up to @code{(point)}, and then add
6755 one to that number. (The @code{1+} function adds one to its
6756 argument.) We add one to it because line 2 has only one line before
6757 it, and @code{count-lines} counts only the lines @emph{before} the
6760 After @code{count-lines} has done its job, and the message has been
6761 printed in the echo area, the @code{save-excursion} restores point and
6762 mark to their original positions; and @code{save-restriction} restores
6763 the original narrowing, if any.
6765 @node narrow Exercise
6766 @section Exercise with Narrowing
6768 Write a function that will display the first 60 characters of the
6769 current buffer, even if you have narrowed the buffer to its latter
6770 half so that the first line is inaccessible. Restore point, mark, and
6771 narrowing. For this exercise, you need to use a whole potpourri of
6772 functions, including @code{save-restriction}, @code{widen},
6773 @code{goto-char}, @code{point-min}, @code{message}, and
6774 @code{buffer-substring}.
6776 @cindex Properties, mention of @code{buffer-substring-no-properties}
6777 (@code{buffer-substring} is a previously unmentioned function you will
6778 have to investigate yourself; or perhaps you will have to use
6779 @code{buffer-substring-no-properties} or
6780 @code{filter-buffer-substring} @dots{}, yet other functions. Text
6781 properties are a feature otherwise not discussed here. @xref{Text
6782 Properties, , Text Properties, elisp, The GNU Emacs Lisp Reference
6785 Additionally, do you really need @code{goto-char} or @code{point-min}?
6786 Or can you write the function without them?
6788 @node car cdr & cons
6789 @chapter @code{car}, @code{cdr}, @code{cons}: Fundamental Functions
6790 @findex car, @r{introduced}
6791 @findex cdr, @r{introduced}
6793 In Lisp, @code{car}, @code{cdr}, and @code{cons} are fundamental
6794 functions. The @code{cons} function is used to construct lists, and
6795 the @code{car} and @code{cdr} functions are used to take them apart.
6797 In the walk through of the @code{copy-region-as-kill} function, we
6798 will see @code{cons} as well as two variants on @code{cdr},
6799 namely, @code{setcdr} and @code{nthcdr}. (@xref{copy-region-as-kill}.)
6802 * Strange Names:: An historical aside: why the strange names?
6803 * car & cdr:: Functions for extracting part of a list.
6804 * cons:: Constructing a list.
6805 * nthcdr:: Calling @code{cdr} repeatedly.
6807 * setcar:: Changing the first element of a list.
6808 * setcdr:: Changing the rest of a list.
6814 @unnumberedsec Strange Names
6817 The name of the @code{cons} function is not unreasonable: it is an
6818 abbreviation of the word `construct'. The origins of the names for
6819 @code{car} and @code{cdr}, on the other hand, are esoteric: @code{car}
6820 is an acronym from the phrase `Contents of the Address part of the
6821 Register'; and @code{cdr} (pronounced `could-er') is an acronym from
6822 the phrase `Contents of the Decrement part of the Register'. These
6823 phrases refer to specific pieces of hardware on the very early
6824 computer on which the original Lisp was developed. Besides being
6825 obsolete, the phrases have been completely irrelevant for more than 25
6826 years to anyone thinking about Lisp. Nonetheless, although a few
6827 brave scholars have begun to use more reasonable names for these
6828 functions, the old terms are still in use. In particular, since the
6829 terms are used in the Emacs Lisp source code, we will use them in this
6833 @section @code{car} and @code{cdr}
6835 The @sc{car} of a list is, quite simply, the first item in the list.
6836 Thus the @sc{car} of the list @code{(rose violet daisy buttercup)} is
6840 If you are reading this in Info in GNU Emacs, you can see this by
6841 evaluating the following:
6844 (car '(rose violet daisy buttercup))
6848 After evaluating the expression, @code{rose} will appear in the echo
6851 Clearly, a more reasonable name for the @code{car} function would be
6852 @code{first} and this is often suggested.
6854 @code{car} does not remove the first item from the list; it only reports
6855 what it is. After @code{car} has been applied to a list, the list is
6856 still the same as it was. In the jargon, @code{car} is
6857 `non-destructive'. This feature turns out to be important.
6859 The @sc{cdr} of a list is the rest of the list, that is, the
6860 @code{cdr} function returns the part of the list that follows the
6861 first item. Thus, while the @sc{car} of the list @code{'(rose violet
6862 daisy buttercup)} is @code{rose}, the rest of the list, the value
6863 returned by the @code{cdr} function, is @code{(violet daisy
6867 You can see this by evaluating the following in the usual way:
6870 (cdr '(rose violet daisy buttercup))
6874 When you evaluate this, @code{(violet daisy buttercup)} will appear in
6877 Like @code{car}, @code{cdr} does not remove any elements from the
6878 list---it just returns a report of what the second and subsequent
6881 Incidentally, in the example, the list of flowers is quoted. If it were
6882 not, the Lisp interpreter would try to evaluate the list by calling
6883 @code{rose} as a function. In this example, we do not want to do that.
6885 Clearly, a more reasonable name for @code{cdr} would be @code{rest}.
6887 (There is a lesson here: when you name new functions, consider very
6888 carefully what you are doing, since you may be stuck with the names
6889 for far longer than you expect. The reason this document perpetuates
6890 these names is that the Emacs Lisp source code uses them, and if I did
6891 not use them, you would have a hard time reading the code; but do,
6892 please, try to avoid using these terms yourself. The people who come
6893 after you will be grateful to you.)
6895 When @code{car} and @code{cdr} are applied to a list made up of symbols,
6896 such as the list @code{(pine fir oak maple)}, the element of the list
6897 returned by the function @code{car} is the symbol @code{pine} without
6898 any parentheses around it. @code{pine} is the first element in the
6899 list. However, the @sc{cdr} of the list is a list itself, @code{(fir
6900 oak maple)}, as you can see by evaluating the following expressions in
6905 (car '(pine fir oak maple))
6907 (cdr '(pine fir oak maple))
6911 On the other hand, in a list of lists, the first element is itself a
6912 list. @code{car} returns this first element as a list. For example,
6913 the following list contains three sub-lists, a list of carnivores, a
6914 list of herbivores and a list of sea mammals:
6918 (car '((lion tiger cheetah)
6919 (gazelle antelope zebra)
6920 (whale dolphin seal)))
6925 In this example, the first element or @sc{car} of the list is the list of
6926 carnivores, @code{(lion tiger cheetah)}, and the rest of the list is
6927 @code{((gazelle antelope zebra) (whale dolphin seal))}.
6931 (cdr '((lion tiger cheetah)
6932 (gazelle antelope zebra)
6933 (whale dolphin seal)))
6937 It is worth saying again that @code{car} and @code{cdr} are
6938 non-destructive---that is, they do not modify or change lists to which
6939 they are applied. This is very important for how they are used.
6941 Also, in the first chapter, in the discussion about atoms, I said that
6942 in Lisp, ``certain kinds of atom, such as an array, can be separated
6943 into parts; but the mechanism for doing this is different from the
6944 mechanism for splitting a list. As far as Lisp is concerned, the
6945 atoms of a list are unsplittable.'' (@xref{Lisp Atoms}.) The
6946 @code{car} and @code{cdr} functions are used for splitting lists and
6947 are considered fundamental to Lisp. Since they cannot split or gain
6948 access to the parts of an array, an array is considered an atom.
6949 Conversely, the other fundamental function, @code{cons}, can put
6950 together or construct a list, but not an array. (Arrays are handled
6951 by array-specific functions. @xref{Arrays, , Arrays, elisp, The GNU
6952 Emacs Lisp Reference Manual}.)
6955 @section @code{cons}
6956 @findex cons, @r{introduced}
6958 The @code{cons} function constructs lists; it is the inverse of
6959 @code{car} and @code{cdr}. For example, @code{cons} can be used to make
6960 a four element list from the three element list, @code{(fir oak maple)}:
6963 (cons 'pine '(fir oak maple))
6968 After evaluating this list, you will see
6971 (pine fir oak maple)
6975 appear in the echo area. @code{cons} causes the creation of a new
6976 list in which the element is followed by the elements of the original
6979 We often say that `@code{cons} puts a new element at the beginning of
6980 a list; it attaches or pushes elements onto the list', but this
6981 phrasing can be misleading, since @code{cons} does not change an
6982 existing list, but creates a new one.
6984 Like @code{car} and @code{cdr}, @code{cons} is non-destructive.
6988 * length:: How to find the length of a list.
6993 @unnumberedsubsec Build a list
6996 @code{cons} must have a list to attach to.@footnote{Actually, you can
6997 @code{cons} an element to an atom to produce a dotted pair. Dotted
6998 pairs are not discussed here; see @ref{Dotted Pair Notation, , Dotted
6999 Pair Notation, elisp, The GNU Emacs Lisp Reference Manual}.} You
7000 cannot start from absolutely nothing. If you are building a list, you
7001 need to provide at least an empty list at the beginning. Here is a
7002 series of @code{cons} expressions that build up a list of flowers. If
7003 you are reading this in Info in GNU Emacs, you can evaluate each of
7004 the expressions in the usual way; the value is printed in this text
7005 after @samp{@result{}}, which you may read as `evaluates to'.
7009 (cons 'buttercup ())
7010 @result{} (buttercup)
7014 (cons 'daisy '(buttercup))
7015 @result{} (daisy buttercup)
7019 (cons 'violet '(daisy buttercup))
7020 @result{} (violet daisy buttercup)
7024 (cons 'rose '(violet daisy buttercup))
7025 @result{} (rose violet daisy buttercup)
7030 In the first example, the empty list is shown as @code{()} and a list
7031 made up of @code{buttercup} followed by the empty list is constructed.
7032 As you can see, the empty list is not shown in the list that was
7033 constructed. All that you see is @code{(buttercup)}. The empty list is
7034 not counted as an element of a list because there is nothing in an empty
7035 list. Generally speaking, an empty list is invisible.
7037 The second example, @code{(cons 'daisy '(buttercup))} constructs a new,
7038 two element list by putting @code{daisy} in front of @code{buttercup};
7039 and the third example constructs a three element list by putting
7040 @code{violet} in front of @code{daisy} and @code{buttercup}.
7043 @subsection Find the Length of a List: @code{length}
7046 You can find out how many elements there are in a list by using the Lisp
7047 function @code{length}, as in the following examples:
7051 (length '(buttercup))
7056 (length '(daisy buttercup))
7061 (length (cons 'violet '(daisy buttercup)))
7067 In the third example, the @code{cons} function is used to construct a
7068 three element list which is then passed to the @code{length} function as
7072 We can also use @code{length} to count the number of elements in an
7083 As you would expect, the number of elements in an empty list is zero.
7085 An interesting experiment is to find out what happens if you try to find
7086 the length of no list at all; that is, if you try to call @code{length}
7087 without giving it an argument, not even an empty list:
7095 What you see, if you evaluate this, is the error message
7098 Lisp error: (wrong-number-of-arguments length 0)
7102 This means that the function receives the wrong number of
7103 arguments, zero, when it expects some other number of arguments. In
7104 this case, one argument is expected, the argument being a list whose
7105 length the function is measuring. (Note that @emph{one} list is
7106 @emph{one} argument, even if the list has many elements inside it.)
7108 The part of the error message that says @samp{length} is the name of
7112 @code{length} is still a subroutine, but you need C-h f to discover that.
7114 In an earlier version:
7115 This is written with a special notation, @samp{#<subr},
7116 that indicates that the function @code{length} is one of the primitive
7117 functions written in C rather than in Emacs Lisp. (@samp{subr} is an
7118 abbreviation for `subroutine'.) @xref{What Is a Function, , What Is a
7119 Function?, elisp , The GNU Emacs Lisp Reference Manual}, for more
7124 @section @code{nthcdr}
7127 The @code{nthcdr} function is associated with the @code{cdr} function.
7128 What it does is take the @sc{cdr} of a list repeatedly.
7130 If you take the @sc{cdr} of the list @code{(pine fir
7131 oak maple)}, you will be returned the list @code{(fir oak maple)}. If you
7132 repeat this on what was returned, you will be returned the list
7133 @code{(oak maple)}. (Of course, repeated @sc{cdr}ing on the original
7134 list will just give you the original @sc{cdr} since the function does
7135 not change the list. You need to evaluate the @sc{cdr} of the
7136 @sc{cdr} and so on.) If you continue this, eventually you will be
7137 returned an empty list, which in this case, instead of being shown as
7138 @code{()} is shown as @code{nil}.
7141 For review, here is a series of repeated @sc{cdr}s, the text following
7142 the @samp{@result{}} shows what is returned.
7146 (cdr '(pine fir oak maple))
7147 @result{}(fir oak maple)
7151 (cdr '(fir oak maple))
7152 @result{} (oak maple)
7177 You can also do several @sc{cdr}s without printing the values in
7182 (cdr (cdr '(pine fir oak maple)))
7183 @result{} (oak maple)
7188 In this example, the Lisp interpreter evaluates the innermost list first.
7189 The innermost list is quoted, so it just passes the list as it is to the
7190 innermost @code{cdr}. This @code{cdr} passes a list made up of the
7191 second and subsequent elements of the list to the outermost @code{cdr},
7192 which produces a list composed of the third and subsequent elements of
7193 the original list. In this example, the @code{cdr} function is repeated
7194 and returns a list that consists of the original list without its
7197 The @code{nthcdr} function does the same as repeating the call to
7198 @code{cdr}. In the following example, the argument 2 is passed to the
7199 function @code{nthcdr}, along with the list, and the value returned is
7200 the list without its first two items, which is exactly the same
7201 as repeating @code{cdr} twice on the list:
7205 (nthcdr 2 '(pine fir oak maple))
7206 @result{} (oak maple)
7211 Using the original four element list, we can see what happens when
7212 various numeric arguments are passed to @code{nthcdr}, including 0, 1,
7217 ;; @r{Leave the list as it was.}
7218 (nthcdr 0 '(pine fir oak maple))
7219 @result{} (pine fir oak maple)
7223 ;; @r{Return a copy without the first element.}
7224 (nthcdr 1 '(pine fir oak maple))
7225 @result{} (fir oak maple)
7229 ;; @r{Return a copy of the list without three elements.}
7230 (nthcdr 3 '(pine fir oak maple))
7235 ;; @r{Return a copy lacking all four elements.}
7236 (nthcdr 4 '(pine fir oak maple))
7241 ;; @r{Return a copy lacking all elements.}
7242 (nthcdr 5 '(pine fir oak maple))
7251 The @code{nthcdr} function takes the @sc{cdr} of a list repeatedly.
7252 The @code{nth} function takes the @sc{car} of the result returned by
7253 @code{nthcdr}. It returns the Nth element of the list.
7256 Thus, if it were not defined in C for speed, the definition of
7257 @code{nth} would be:
7262 "Returns the Nth element of LIST.
7263 N counts from zero. If LIST is not that long, nil is returned."
7264 (car (nthcdr n list)))
7269 (Originally, @code{nth} was defined in Emacs Lisp in @file{subr.el},
7270 but its definition was redone in C in the 1980s.)
7272 The @code{nth} function returns a single element of a list.
7273 This can be very convenient.
7275 Note that the elements are numbered from zero, not one. That is to
7276 say, the first element of a list, its @sc{car} is the zeroth element.
7277 This is called `zero-based' counting and often bothers people who
7278 are accustomed to the first element in a list being number one, which
7286 (nth 0 '("one" "two" "three"))
7289 (nth 1 '("one" "two" "three"))
7294 It is worth mentioning that @code{nth}, like @code{nthcdr} and
7295 @code{cdr}, does not change the original list---the function is
7296 non-destructive. This is in sharp contrast to the @code{setcar} and
7297 @code{setcdr} functions.
7300 @section @code{setcar}
7303 As you might guess from their names, the @code{setcar} and @code{setcdr}
7304 functions set the @sc{car} or the @sc{cdr} of a list to a new value.
7305 They actually change the original list, unlike @code{car} and @code{cdr}
7306 which leave the original list as it was. One way to find out how this
7307 works is to experiment. We will start with the @code{setcar} function.
7310 First, we can make a list and then set the value of a variable to the
7311 list, using the @code{setq} function. Here is a list of animals:
7314 (setq animals '(antelope giraffe lion tiger))
7318 If you are reading this in Info inside of GNU Emacs, you can evaluate
7319 this expression in the usual fashion, by positioning the cursor after
7320 the expression and typing @kbd{C-x C-e}. (I'm doing this right here
7321 as I write this. This is one of the advantages of having the
7322 interpreter built into the computing environment. Incidentally, when
7323 there is nothing on the line after the final parentheses, such as a
7324 comment, point can be on the next line. Thus, if your cursor is in
7325 the first column of the next line, you do not need to move it.
7326 Indeed, Emacs permits any amount of white space after the final
7330 When we evaluate the variable @code{animals}, we see that it is bound to
7331 the list @code{(antelope giraffe lion tiger)}:
7336 @result{} (antelope giraffe lion tiger)
7341 Put another way, the variable @code{animals} points to the list
7342 @code{(antelope giraffe lion tiger)}.
7344 Next, evaluate the function @code{setcar} while passing it two
7345 arguments, the variable @code{animals} and the quoted symbol
7346 @code{hippopotamus}; this is done by writing the three element list
7347 @code{(setcar animals 'hippopotamus)} and then evaluating it in the
7351 (setcar animals 'hippopotamus)
7356 After evaluating this expression, evaluate the variable @code{animals}
7357 again. You will see that the list of animals has changed:
7362 @result{} (hippopotamus giraffe lion tiger)
7367 The first element on the list, @code{antelope} is replaced by
7368 @code{hippopotamus}.
7370 So we can see that @code{setcar} did not add a new element to the list
7371 as @code{cons} would have; it replaced @code{antelope} with
7372 @code{hippopotamus}; it @emph{changed} the list.
7375 @section @code{setcdr}
7378 The @code{setcdr} function is similar to the @code{setcar} function,
7379 except that the function replaces the second and subsequent elements of
7380 a list rather than the first element.
7382 (To see how to change the last element of a list, look ahead to
7383 @ref{kill-new function, , The @code{kill-new} function}, which uses
7384 the @code{nthcdr} and @code{setcdr} functions.)
7387 To see how this works, set the value of the variable to a list of
7388 domesticated animals by evaluating the following expression:
7391 (setq domesticated-animals '(horse cow sheep goat))
7396 If you now evaluate the list, you will be returned the list
7397 @code{(horse cow sheep goat)}:
7401 domesticated-animals
7402 @result{} (horse cow sheep goat)
7407 Next, evaluate @code{setcdr} with two arguments, the name of the
7408 variable which has a list as its value, and the list to which the
7409 @sc{cdr} of the first list will be set;
7412 (setcdr domesticated-animals '(cat dog))
7416 If you evaluate this expression, the list @code{(cat dog)} will appear
7417 in the echo area. This is the value returned by the function. The
7418 result we are interested in is the ``side effect'', which we can see by
7419 evaluating the variable @code{domesticated-animals}:
7423 domesticated-animals
7424 @result{} (horse cat dog)
7429 Indeed, the list is changed from @code{(horse cow sheep goat)} to
7430 @code{(horse cat dog)}. The @sc{cdr} of the list is changed from
7431 @code{(cow sheep goat)} to @code{(cat dog)}.
7436 Construct a list of four birds by evaluating several expressions with
7437 @code{cons}. Find out what happens when you @code{cons} a list onto
7438 itself. Replace the first element of the list of four birds with a
7439 fish. Replace the rest of that list with a list of other fish.
7441 @node Cutting & Storing Text
7442 @chapter Cutting and Storing Text
7443 @cindex Cutting and storing text
7444 @cindex Storing and cutting text
7445 @cindex Killing text
7446 @cindex Clipping text
7447 @cindex Erasing text
7448 @cindex Deleting text
7450 Whenever you cut or clip text out of a buffer with a `kill' command in
7451 GNU Emacs, it is stored in a list and you can bring it back with a
7454 (The use of the word `kill' in Emacs for processes which specifically
7455 @emph{do not} destroy the values of the entities is an unfortunate
7456 historical accident. A much more appropriate word would be `clip' since
7457 that is what the kill commands do; they clip text out of a buffer and
7458 put it into storage from which it can be brought back. I have often
7459 been tempted to replace globally all occurrences of `kill' in the Emacs
7460 sources with `clip' and all occurrences of `killed' with `clipped'.)
7463 * Storing Text:: Text is stored in a list.
7464 * zap-to-char:: Cutting out text up to a character.
7465 * kill-region:: Cutting text out of a region.
7466 * copy-region-as-kill:: A definition for copying text.
7467 * Digression into C:: Minor note on C programming language macros.
7468 * defvar:: How to give a variable an initial value.
7469 * cons & search-fwd Review::
7470 * search Exercises::
7475 @unnumberedsec Storing Text in a List
7478 When text is cut out of a buffer, it is stored on a list. Successive
7479 pieces of text are stored on the list successively, so the list might
7483 ("a piece of text" "previous piece")
7488 The function @code{cons} can be used to create a new list from a piece
7489 of text (an `atom', to use the jargon) and an existing list, like
7494 (cons "another piece"
7495 '("a piece of text" "previous piece"))
7501 If you evaluate this expression, a list of three elements will appear in
7505 ("another piece" "a piece of text" "previous piece")
7508 With the @code{car} and @code{nthcdr} functions, you can retrieve
7509 whichever piece of text you want. For example, in the following code,
7510 @code{nthcdr 1 @dots{}} returns the list with the first item removed;
7511 and the @code{car} returns the first element of that remainder---the
7512 second element of the original list:
7516 (car (nthcdr 1 '("another piece"
7519 @result{} "a piece of text"
7523 The actual functions in Emacs are more complex than this, of course.
7524 The code for cutting and retrieving text has to be written so that
7525 Emacs can figure out which element in the list you want---the first,
7526 second, third, or whatever. In addition, when you get to the end of
7527 the list, Emacs should give you the first element of the list, rather
7528 than nothing at all.
7530 The list that holds the pieces of text is called the @dfn{kill ring}.
7531 This chapter leads up to a description of the kill ring and how it is
7532 used by first tracing how the @code{zap-to-char} function works. This
7533 function uses (or `calls') a function that invokes a function that
7534 manipulates the kill ring. Thus, before reaching the mountains, we
7535 climb the foothills.
7537 A subsequent chapter describes how text that is cut from the buffer is
7538 retrieved. @xref{Yanking, , Yanking Text Back}.
7541 @section @code{zap-to-char}
7544 Let us look at the interactive @code{zap-to-char} function.
7547 * Complete zap-to-char:: The complete implementation.
7548 * zap-to-char interactive:: A three part interactive expression.
7549 * zap-to-char body:: A short overview.
7550 * search-forward:: How to search for a string.
7551 * progn:: The @code{progn} special form.
7552 * Summing up zap-to-char:: Using @code{point} and @code{search-forward}.
7556 @node Complete zap-to-char
7557 @unnumberedsubsec The Complete @code{zap-to-char} Implementation
7560 The @code{zap-to-char} function removes the text in the region between
7561 the location of the cursor (i.e., of point) up to and including the
7562 next occurrence of a specified character. The text that
7563 @code{zap-to-char} removes is put in the kill ring; and it can be
7564 retrieved from the kill ring by typing @kbd{C-y} (@code{yank}). If
7565 the command is given an argument, it removes text through that number
7566 of occurrences. Thus, if the cursor were at the beginning of this
7567 sentence and the character were @samp{s}, @samp{Thus} would be
7568 removed. If the argument were two, @samp{Thus, if the curs} would be
7569 removed, up to and including the @samp{s} in @samp{cursor}.
7571 If the specified character is not found, @code{zap-to-char} will say
7572 ``Search failed'', tell you the character you typed, and not remove
7575 In order to determine how much text to remove, @code{zap-to-char} uses
7576 a search function. Searches are used extensively in code that
7577 manipulates text, and we will focus attention on them as well as on the
7581 @c GNU Emacs version 19
7582 (defun zap-to-char (arg char) ; version 19 implementation
7583 "Kill up to and including ARG'th occurrence of CHAR.
7584 Goes backward if ARG is negative; error if CHAR not found."
7585 (interactive "*p\ncZap to char: ")
7586 (kill-region (point)
7589 (char-to-string char) nil nil arg)
7594 Here is the complete text of the version 22 implementation of the function:
7599 (defun zap-to-char (arg char)
7600 "Kill up to and including ARG'th occurrence of CHAR.
7601 Case is ignored if `case-fold-search' is non-nil in the current buffer.
7602 Goes backward if ARG is negative; error if CHAR not found."
7603 (interactive "p\ncZap to char: ")
7604 (if (char-table-p translation-table-for-input)
7605 (setq char (or (aref translation-table-for-input char) char)))
7606 (kill-region (point) (progn
7607 (search-forward (char-to-string char)
7613 The documentation is thorough. You do need to know the jargon meaning
7616 @node zap-to-char interactive
7617 @subsection The @code{interactive} Expression
7620 The interactive expression in the @code{zap-to-char} command looks like
7624 (interactive "p\ncZap to char: ")
7627 The part within quotation marks, @code{"p\ncZap to char:@: "}, specifies
7628 two different things. First, and most simply, is the @samp{p}.
7629 This part is separated from the next part by a newline, @samp{\n}.
7630 The @samp{p} means that the first argument to the function will be
7631 passed the value of a `processed prefix'. The prefix argument is
7632 passed by typing @kbd{C-u} and a number, or @kbd{M-} and a number. If
7633 the function is called interactively without a prefix, 1 is passed to
7636 The second part of @code{"p\ncZap to char:@: "} is
7637 @samp{cZap to char:@: }. In this part, the lower case @samp{c}
7638 indicates that @code{interactive} expects a prompt and that the
7639 argument will be a character. The prompt follows the @samp{c} and is
7640 the string @samp{Zap to char:@: } (with a space after the colon to
7643 What all this does is prepare the arguments to @code{zap-to-char} so they
7644 are of the right type, and give the user a prompt.
7646 In a read-only buffer, the @code{zap-to-char} function copies the text
7647 to the kill ring, but does not remove it. The echo area displays a
7648 message saying that the buffer is read-only. Also, the terminal may
7649 beep or blink at you.
7651 @node zap-to-char body
7652 @subsection The Body of @code{zap-to-char}
7654 The body of the @code{zap-to-char} function contains the code that
7655 kills (that is, removes) the text in the region from the current
7656 position of the cursor up to and including the specified character.
7658 The first part of the code looks like this:
7661 (if (char-table-p translation-table-for-input)
7662 (setq char (or (aref translation-table-for-input char) char)))
7663 (kill-region (point) (progn
7664 (search-forward (char-to-string char) nil nil arg)
7669 @code{char-table-p} is an hitherto unseen function. It determines
7670 whether its argument is a character table. When it is, it sets the
7671 character passed to @code{zap-to-char} to one of them, if that
7672 character exists, or to the character itself. (This becomes important
7673 for certain characters in non-European languages. The @code{aref}
7674 function extracts an element from an array. It is an array-specific
7675 function that is not described in this document. @xref{Arrays, ,
7676 Arrays, elisp, The GNU Emacs Lisp Reference Manual}.)
7679 @code{(point)} is the current position of the cursor.
7681 The next part of the code is an expression using @code{progn}. The body
7682 of the @code{progn} consists of calls to @code{search-forward} and
7685 It is easier to understand how @code{progn} works after learning about
7686 @code{search-forward}, so we will look at @code{search-forward} and
7687 then at @code{progn}.
7689 @node search-forward
7690 @subsection The @code{search-forward} Function
7691 @findex search-forward
7693 The @code{search-forward} function is used to locate the
7694 zapped-for-character in @code{zap-to-char}. If the search is
7695 successful, @code{search-forward} leaves point immediately after the
7696 last character in the target string. (In @code{zap-to-char}, the
7697 target string is just one character long. @code{zap-to-char} uses the
7698 function @code{char-to-string} to ensure that the computer treats that
7699 character as a string.) If the search is backwards,
7700 @code{search-forward} leaves point just before the first character in
7701 the target. Also, @code{search-forward} returns @code{t} for true.
7702 (Moving point is therefore a `side effect'.)
7705 In @code{zap-to-char}, the @code{search-forward} function looks like this:
7708 (search-forward (char-to-string char) nil nil arg)
7711 The @code{search-forward} function takes four arguments:
7715 The first argument is the target, what is searched for. This must be a
7716 string, such as @samp{"z"}.
7718 As it happens, the argument passed to @code{zap-to-char} is a single
7719 character. Because of the way computers are built, the Lisp
7720 interpreter may treat a single character as being different from a
7721 string of characters. Inside the computer, a single character has a
7722 different electronic format than a string of one character. (A single
7723 character can often be recorded in the computer using exactly one
7724 byte; but a string may be longer, and the computer needs to be ready
7725 for this.) Since the @code{search-forward} function searches for a
7726 string, the character that the @code{zap-to-char} function receives as
7727 its argument must be converted inside the computer from one format to
7728 the other; otherwise the @code{search-forward} function will fail.
7729 The @code{char-to-string} function is used to make this conversion.
7732 The second argument bounds the search; it is specified as a position in
7733 the buffer. In this case, the search can go to the end of the buffer,
7734 so no bound is set and the second argument is @code{nil}.
7737 The third argument tells the function what it should do if the search
7738 fails---it can signal an error (and print a message) or it can return
7739 @code{nil}. A @code{nil} as the third argument causes the function to
7740 signal an error when the search fails.
7743 The fourth argument to @code{search-forward} is the repeat count---how
7744 many occurrences of the string to look for. This argument is optional
7745 and if the function is called without a repeat count, this argument is
7746 passed the value 1. If this argument is negative, the search goes
7751 In template form, a @code{search-forward} expression looks like this:
7755 (search-forward "@var{target-string}"
7756 @var{limit-of-search}
7757 @var{what-to-do-if-search-fails}
7762 We will look at @code{progn} next.
7765 @subsection The @code{progn} Special Form
7768 @code{progn} is a special form that causes each of its arguments to be
7769 evaluated in sequence and then returns the value of the last one. The
7770 preceding expressions are evaluated only for the side effects they
7771 perform. The values produced by them are discarded.
7774 The template for a @code{progn} expression is very simple:
7783 In @code{zap-to-char}, the @code{progn} expression has to do two things:
7784 put point in exactly the right position; and return the location of
7785 point so that @code{kill-region} will know how far to kill to.
7787 The first argument to the @code{progn} is @code{search-forward}. When
7788 @code{search-forward} finds the string, the function leaves point
7789 immediately after the last character in the target string. (In this
7790 case the target string is just one character long.) If the search is
7791 backwards, @code{search-forward} leaves point just before the first
7792 character in the target. The movement of point is a side effect.
7794 The second and last argument to @code{progn} is the expression
7795 @code{(point)}. This expression returns the value of point, which in
7796 this case will be the location to which it has been moved by
7797 @code{search-forward}. (In the source, a line that tells the function
7798 to go to the previous character, if it is going forward, was commented
7799 out in 1999; I don't remember whether that feature or mis-feature was
7800 ever a part of the distributed source.) The value of @code{point} is
7801 returned by the @code{progn} expression and is passed to
7802 @code{kill-region} as @code{kill-region}'s second argument.
7804 @node Summing up zap-to-char
7805 @subsection Summing up @code{zap-to-char}
7807 Now that we have seen how @code{search-forward} and @code{progn} work,
7808 we can see how the @code{zap-to-char} function works as a whole.
7810 The first argument to @code{kill-region} is the position of the cursor
7811 when the @code{zap-to-char} command is given---the value of point at
7812 that time. Within the @code{progn}, the search function then moves
7813 point to just after the zapped-to-character and @code{point} returns the
7814 value of this location. The @code{kill-region} function puts together
7815 these two values of point, the first one as the beginning of the region
7816 and the second one as the end of the region, and removes the region.
7818 The @code{progn} special form is necessary because the
7819 @code{kill-region} command takes two arguments; and it would fail if
7820 @code{search-forward} and @code{point} expressions were written in
7821 sequence as two additional arguments. The @code{progn} expression is
7822 a single argument to @code{kill-region} and returns the one value that
7823 @code{kill-region} needs for its second argument.
7826 @section @code{kill-region}
7829 The @code{zap-to-char} function uses the @code{kill-region} function.
7830 This function clips text from a region and copies that text to
7831 the kill ring, from which it may be retrieved.
7836 (defun kill-region (beg end &optional yank-handler)
7837 "Kill (\"cut\") text between point and mark.
7838 This deletes the text from the buffer and saves it in the kill ring.
7839 The command \\[yank] can retrieve it from there.
7840 \(If you want to kill and then yank immediately, use \\[kill-ring-save].)
7842 If you want to append the killed region to the last killed text,
7843 use \\[append-next-kill] before \\[kill-region].
7845 If the buffer is read-only, Emacs will beep and refrain from deleting
7846 the text, but put the text in the kill ring anyway. This means that
7847 you can use the killing commands to copy text from a read-only buffer.
7849 This is the primitive for programs to kill text (as opposed to deleting it).
7850 Supply two arguments, character positions indicating the stretch of text
7852 Any command that calls this function is a \"kill command\".
7853 If the previous command was also a kill command,
7854 the text killed this time appends to the text killed last time
7855 to make one entry in the kill ring.
7857 In Lisp code, optional third arg YANK-HANDLER, if non-nil,
7858 specifies the yank-handler text property to be set on the killed
7859 text. See `insert-for-yank'."
7860 ;; Pass point first, then mark, because the order matters
7861 ;; when calling kill-append.
7862 (interactive (list (point) (mark)))
7863 (unless (and beg end)
7864 (error "The mark is not set now, so there is no region"))
7866 (let ((string (filter-buffer-substring beg end t)))
7867 (when string ;STRING is nil if BEG = END
7868 ;; Add that string to the kill ring, one way or another.
7869 (if (eq last-command 'kill-region)
7870 (kill-append string (< end beg) yank-handler)
7871 (kill-new string nil yank-handler)))
7872 (when (or string (eq last-command 'kill-region))
7873 (setq this-command 'kill-region))
7875 ((buffer-read-only text-read-only)
7876 ;; The code above failed because the buffer, or some of the characters
7877 ;; in the region, are read-only.
7878 ;; We should beep, in case the user just isn't aware of this.
7879 ;; However, there's no harm in putting
7880 ;; the region's text in the kill ring, anyway.
7881 (copy-region-as-kill beg end)
7882 ;; Set this-command now, so it will be set even if we get an error.
7883 (setq this-command 'kill-region)
7884 ;; This should barf, if appropriate, and give us the correct error.
7885 (if kill-read-only-ok
7886 (progn (message "Read only text copied to kill ring") nil)
7887 ;; Signal an error if the buffer is read-only.
7888 (barf-if-buffer-read-only)
7889 ;; If the buffer isn't read-only, the text is.
7890 (signal 'text-read-only (list (current-buffer)))))))
7893 The Emacs 22 version of that function uses @code{condition-case} and
7894 @code{copy-region-as-kill}, both of which we will explain.
7895 @code{condition-case} is an important special form.
7897 In essence, the @code{kill-region} function calls
7898 @code{condition-case}, which takes three arguments. In this function,
7899 the first argument does nothing. The second argument contains the
7900 code that does the work when all goes well. The third argument
7901 contains the code that is called in the event of an error.
7904 * Complete kill-region:: The function definition.
7905 * condition-case:: Dealing with a problem.
7910 @node Complete kill-region
7911 @unnumberedsubsec The Complete @code{kill-region} Definition
7915 We will go through the @code{condition-case} code in a moment. First,
7916 let us look at the definition of @code{kill-region}, with comments
7922 (defun kill-region (beg end)
7923 "Kill (\"cut\") text between point and mark.
7924 This deletes the text from the buffer and saves it in the kill ring.
7925 The command \\[yank] can retrieve it from there. @dots{} "
7929 ;; @bullet{} Since order matters, pass point first.
7930 (interactive (list (point) (mark)))
7931 ;; @bullet{} And tell us if we cannot cut the text.
7932 ;; `unless' is an `if' without a then-part.
7933 (unless (and beg end)
7934 (error "The mark is not set now, so there is no region"))
7938 ;; @bullet{} `condition-case' takes three arguments.
7939 ;; If the first argument is nil, as it is here,
7940 ;; information about the error signal is not
7941 ;; stored for use by another function.
7946 ;; @bullet{} The second argument to `condition-case' tells the
7947 ;; Lisp interpreter what to do when all goes well.
7951 ;; It starts with a `let' function that extracts the string
7952 ;; and tests whether it exists. If so (that is what the
7953 ;; `when' checks), it calls an `if' function that determines
7954 ;; whether the previous command was another call to
7955 ;; `kill-region'; if it was, then the new text is appended to
7956 ;; the previous text; if not, then a different function,
7957 ;; `kill-new', is called.
7961 ;; The `kill-append' function concatenates the new string and
7962 ;; the old. The `kill-new' function inserts text into a new
7963 ;; item in the kill ring.
7967 ;; `when' is an `if' without an else-part. The second `when'
7968 ;; again checks whether the current string exists; in
7969 ;; addition, it checks whether the previous command was
7970 ;; another call to `kill-region'. If one or the other
7971 ;; condition is true, then it sets the current command to
7972 ;; be `kill-region'.
7975 (let ((string (filter-buffer-substring beg end t)))
7976 (when string ;STRING is nil if BEG = END
7977 ;; Add that string to the kill ring, one way or another.
7978 (if (eq last-command 'kill-region)
7981 ;; @minus{} `yank-handler' is an optional argument to
7982 ;; `kill-region' that tells the `kill-append' and
7983 ;; `kill-new' functions how deal with properties
7984 ;; added to the text, such as `bold' or `italics'.
7985 (kill-append string (< end beg) yank-handler)
7986 (kill-new string nil yank-handler)))
7987 (when (or string (eq last-command 'kill-region))
7988 (setq this-command 'kill-region))
7993 ;; @bullet{} The third argument to `condition-case' tells the interpreter
7994 ;; what to do with an error.
7997 ;; The third argument has a conditions part and a body part.
7998 ;; If the conditions are met (in this case,
7999 ;; if text or buffer are read-only)
8000 ;; then the body is executed.
8003 ;; The first part of the third argument is the following:
8004 ((buffer-read-only text-read-only) ;; the if-part
8005 ;; @dots{} the then-part
8006 (copy-region-as-kill beg end)
8009 ;; Next, also as part of the then-part, set this-command, so
8010 ;; it will be set in an error
8011 (setq this-command 'kill-region)
8012 ;; Finally, in the then-part, send a message if you may copy
8013 ;; the text to the kill ring without signaling an error, but
8014 ;; don't if you may not.
8017 (if kill-read-only-ok
8018 (progn (message "Read only text copied to kill ring") nil)
8019 (barf-if-buffer-read-only)
8020 ;; If the buffer isn't read-only, the text is.
8021 (signal 'text-read-only (list (current-buffer)))))
8029 (defun kill-region (beg end)
8030 "Kill between point and mark.
8031 The text is deleted but saved in the kill ring."
8036 ;; 1. `condition-case' takes three arguments.
8037 ;; If the first argument is nil, as it is here,
8038 ;; information about the error signal is not
8039 ;; stored for use by another function.
8044 ;; 2. The second argument to `condition-case'
8045 ;; tells the Lisp interpreter what to do when all goes well.
8049 ;; The `delete-and-extract-region' function usually does the
8050 ;; work. If the beginning and ending of the region are both
8051 ;; the same, then the variable `string' will be empty, or nil
8052 (let ((string (delete-and-extract-region beg end)))
8056 ;; `when' is an `if' clause that cannot take an `else-part'.
8057 ;; Emacs normally sets the value of `last-command' to the
8058 ;; previous command.
8061 ;; `kill-append' concatenates the new string and the old.
8062 ;; `kill-new' inserts text into a new item in the kill ring.
8064 (if (eq last-command 'kill-region)
8065 ;; if true, prepend string
8066 (kill-append string (< end beg))
8068 (setq this-command 'kill-region))
8072 ;; 3. The third argument to `condition-case' tells the interpreter
8073 ;; what to do with an error.
8076 ;; The third argument has a conditions part and a body part.
8077 ;; If the conditions are met (in this case,
8078 ;; if text or buffer are read-only)
8079 ;; then the body is executed.
8082 ((buffer-read-only text-read-only) ;; this is the if-part
8084 (copy-region-as-kill beg end)
8087 (if kill-read-only-ok ;; usually this variable is nil
8088 (message "Read only text copied to kill ring")
8089 ;; or else, signal an error if the buffer is read-only;
8090 (barf-if-buffer-read-only)
8091 ;; and, in any case, signal that the text is read-only.
8092 (signal 'text-read-only (list (current-buffer)))))))
8097 @node condition-case
8098 @subsection @code{condition-case}
8099 @findex condition-case
8101 As we have seen earlier (@pxref{Making Errors, , Generate an Error
8102 Message}), when the Emacs Lisp interpreter has trouble evaluating an
8103 expression, it provides you with help; in the jargon, this is called
8104 ``signaling an error''. Usually, the computer stops the program and
8105 shows you a message.
8107 However, some programs undertake complicated actions. They should not
8108 simply stop on an error. In the @code{kill-region} function, the most
8109 likely error is that you will try to kill text that is read-only and
8110 cannot be removed. So the @code{kill-region} function contains code
8111 to handle this circumstance. This code, which makes up the body of
8112 the @code{kill-region} function, is inside of a @code{condition-case}
8116 The template for @code{condition-case} looks like this:
8123 @var{error-handler}@dots{})
8127 The second argument, @var{bodyform}, is straightforward. The
8128 @code{condition-case} special form causes the Lisp interpreter to
8129 evaluate the code in @var{bodyform}. If no error occurs, the special
8130 form returns the code's value and produces the side-effects, if any.
8132 In short, the @var{bodyform} part of a @code{condition-case}
8133 expression determines what should happen when everything works
8136 However, if an error occurs, among its other actions, the function
8137 generating the error signal will define one or more error condition
8140 An error handler is the third argument to @code{condition case}.
8141 An error handler has two parts, a @var{condition-name} and a
8142 @var{body}. If the @var{condition-name} part of an error handler
8143 matches a condition name generated by an error, then the @var{body}
8144 part of the error handler is run.
8146 As you will expect, the @var{condition-name} part of an error handler
8147 may be either a single condition name or a list of condition names.
8149 Also, a complete @code{condition-case} expression may contain more
8150 than one error handler. When an error occurs, the first applicable
8153 Lastly, the first argument to the @code{condition-case} expression,
8154 the @var{var} argument, is sometimes bound to a variable that
8155 contains information about the error. However, if that argument is
8156 nil, as is the case in @code{kill-region}, that information is
8160 In brief, in the @code{kill-region} function, the code
8161 @code{condition-case} works like this:
8165 @var{If no errors}, @var{run only this code}
8166 @var{but}, @var{if errors}, @var{run this other code}.
8173 copy-region-as-kill is short, 12 lines, and uses
8174 filter-buffer-substring, which is longer, 39 lines
8175 and has delete-and-extract-region in it.
8176 delete-and-extract-region is written in C.
8178 see Initializing a Variable with @code{defvar}
8180 Initializing a Variable with @code{defvar} includes line 8350
8184 @subsection Lisp macro
8188 The part of the @code{condition-case} expression that is evaluated in
8189 the expectation that all goes well has a @code{when}. The code uses
8190 @code{when} to determine whether the @code{string} variable points to
8193 A @code{when} expression is simply a programmers' convenience. It is
8194 an @code{if} without the possibility of an else clause. In your mind,
8195 you can replace @code{when} with @code{if} and understand what goes
8196 on. That is what the Lisp interpreter does.
8198 Technically speaking, @code{when} is a Lisp macro. A Lisp macro
8199 enables you to define new control constructs and other language
8200 features. It tells the interpreter how to compute another Lisp
8201 expression which will in turn compute the value. In this case, the
8202 `other expression' is an @code{if} expression.
8204 The @code{kill-region} function definition also has an @code{unless}
8205 macro; it is the converse of @code{when}. The @code{unless} macro is
8206 an @code{if} without a then clause
8208 For more about Lisp macros, see @ref{Macros, , Macros, elisp, The GNU
8209 Emacs Lisp Reference Manual}. The C programming language also
8210 provides macros. These are different, but also useful.
8213 We will briefly look at C macros in
8214 @ref{Digression into C}.
8218 Regarding the @code{when} macro, in the @code{condition-case}
8219 expression, when the string has content, then another conditional
8220 expression is executed. This is an @code{if} with both a then-part
8225 (if (eq last-command 'kill-region)
8226 (kill-append string (< end beg) yank-handler)
8227 (kill-new string nil yank-handler))
8231 The then-part is evaluated if the previous command was another call to
8232 @code{kill-region}; if not, the else-part is evaluated.
8234 @code{yank-handler} is an optional argument to @code{kill-region} that
8235 tells the @code{kill-append} and @code{kill-new} functions how deal
8236 with properties added to the text, such as `bold' or `italics'.
8238 @code{last-command} is a variable that comes with Emacs that we have
8239 not seen before. Normally, whenever a function is executed, Emacs
8240 sets the value of @code{last-command} to the previous command.
8243 In this segment of the definition, the @code{if} expression checks
8244 whether the previous command was @code{kill-region}. If it was,
8247 (kill-append string (< end beg) yank-handler)
8251 concatenates a copy of the newly clipped text to the just previously
8252 clipped text in the kill ring.
8254 @node copy-region-as-kill
8255 @section @code{copy-region-as-kill}
8256 @findex copy-region-as-kill
8259 The @code{copy-region-as-kill} function copies a region of text from a
8260 buffer and (via either @code{kill-append} or @code{kill-new}) saves it
8261 in the @code{kill-ring}.
8263 If you call @code{copy-region-as-kill} immediately after a
8264 @code{kill-region} command, Emacs appends the newly copied text to the
8265 previously copied text. This means that if you yank back the text, you
8266 get it all, from both this and the previous operation. On the other
8267 hand, if some other command precedes the @code{copy-region-as-kill},
8268 the function copies the text into a separate entry in the kill ring.
8271 * Complete copy-region-as-kill:: The complete function definition.
8272 * copy-region-as-kill body:: The body of @code{copy-region-as-kill}.
8276 @node Complete copy-region-as-kill
8277 @unnumberedsubsec The complete @code{copy-region-as-kill} function definition
8281 Here is the complete text of the version 22 @code{copy-region-as-kill}
8286 (defun copy-region-as-kill (beg end)
8287 "Save the region as if killed, but don't kill it.
8288 In Transient Mark mode, deactivate the mark.
8289 If `interprogram-cut-function' is non-nil, also save the text for a window
8290 system cut and paste."
8294 (if (eq last-command 'kill-region)
8295 (kill-append (filter-buffer-substring beg end) (< end beg))
8296 (kill-new (filter-buffer-substring beg end)))
8299 (if transient-mark-mode
8300 (setq deactivate-mark t))
8306 As usual, this function can be divided into its component parts:
8310 (defun copy-region-as-kill (@var{argument-list})
8311 "@var{documentation}@dots{}"
8317 The arguments are @code{beg} and @code{end} and the function is
8318 interactive with @code{"r"}, so the two arguments must refer to the
8319 beginning and end of the region. If you have been reading through this
8320 document from the beginning, understanding these parts of a function is
8321 almost becoming routine.
8323 The documentation is somewhat confusing unless you remember that the
8324 word `kill' has a meaning different from usual. The `Transient Mark'
8325 and @code{interprogram-cut-function} comments explain certain
8328 After you once set a mark, a buffer always contains a region. If you
8329 wish, you can use Transient Mark mode to highlight the region
8330 temporarily. (No one wants to highlight the region all the time, so
8331 Transient Mark mode highlights it only at appropriate times. Many
8332 people turn off Transient Mark mode, so the region is never
8335 Also, a windowing system allows you to copy, cut, and paste among
8336 different programs. In the X windowing system, for example, the
8337 @code{interprogram-cut-function} function is @code{x-select-text},
8338 which works with the windowing system's equivalent of the Emacs kill
8341 The body of the @code{copy-region-as-kill} function starts with an
8342 @code{if} clause. What this clause does is distinguish between two
8343 different situations: whether or not this command is executed
8344 immediately after a previous @code{kill-region} command. In the first
8345 case, the new region is appended to the previously copied text.
8346 Otherwise, it is inserted into the beginning of the kill ring as a
8347 separate piece of text from the previous piece.
8349 The last two lines of the function prevent the region from lighting up
8350 if Transient Mark mode is turned on.
8352 The body of @code{copy-region-as-kill} merits discussion in detail.
8354 @node copy-region-as-kill body
8355 @subsection The Body of @code{copy-region-as-kill}
8357 The @code{copy-region-as-kill} function works in much the same way as
8358 the @code{kill-region} function. Both are written so that two or more
8359 kills in a row combine their text into a single entry. If you yank
8360 back the text from the kill ring, you get it all in one piece.
8361 Moreover, kills that kill forward from the current position of the
8362 cursor are added to the end of the previously copied text and commands
8363 that copy text backwards add it to the beginning of the previously
8364 copied text. This way, the words in the text stay in the proper
8367 Like @code{kill-region}, the @code{copy-region-as-kill} function makes
8368 use of the @code{last-command} variable that keeps track of the
8369 previous Emacs command.
8372 * last-command & this-command::
8373 * kill-append function::
8374 * kill-new function::
8378 @node last-command & this-command
8379 @unnumberedsubsubsec @code{last-command} and @code{this-command}
8382 Normally, whenever a function is executed, Emacs sets the value of
8383 @code{this-command} to the function being executed (which in this case
8384 would be @code{copy-region-as-kill}). At the same time, Emacs sets
8385 the value of @code{last-command} to the previous value of
8386 @code{this-command}.
8388 In the first part of the body of the @code{copy-region-as-kill}
8389 function, an @code{if} expression determines whether the value of
8390 @code{last-command} is @code{kill-region}. If so, the then-part of
8391 the @code{if} expression is evaluated; it uses the @code{kill-append}
8392 function to concatenate the text copied at this call to the function
8393 with the text already in the first element (the @sc{car}) of the kill
8394 ring. On the other hand, if the value of @code{last-command} is not
8395 @code{kill-region}, then the @code{copy-region-as-kill} function
8396 attaches a new element to the kill ring using the @code{kill-new}
8400 The @code{if} expression reads as follows; it uses @code{eq}:
8404 (if (eq last-command 'kill-region)
8406 (kill-append (filter-buffer-substring beg end) (< end beg))
8408 (kill-new (filter-buffer-substring beg end)))
8412 @findex filter-buffer-substring
8413 (The @code{filter-buffer-substring} function returns a filtered
8414 substring of the buffer, if any. Optionally---the arguments are not
8415 here, so neither is done---the function may delete the initial text or
8416 return the text without its properties; this function is a replacement
8417 for the older @code{buffer-substring} function, which came before text
8418 properties were implemented.)
8420 @findex eq @r{(example of use)}
8422 The @code{eq} function tests whether its first argument is the same Lisp
8423 object as its second argument. The @code{eq} function is similar to the
8424 @code{equal} function in that it is used to test for equality, but
8425 differs in that it determines whether two representations are actually
8426 the same object inside the computer, but with different names.
8427 @code{equal} determines whether the structure and contents of two
8428 expressions are the same.
8430 If the previous command was @code{kill-region}, then the Emacs Lisp
8431 interpreter calls the @code{kill-append} function
8433 @node kill-append function
8434 @unnumberedsubsubsec The @code{kill-append} function
8438 The @code{kill-append} function looks like this:
8443 (defun kill-append (string before-p &optional yank-handler)
8444 "Append STRING to the end of the latest kill in the kill ring.
8445 If BEFORE-P is non-nil, prepend STRING to the kill.
8447 (let* ((cur (car kill-ring)))
8448 (kill-new (if before-p (concat string cur) (concat cur string))
8449 (or (= (length cur) 0)
8451 (get-text-property 0 'yank-handler cur)))
8458 (defun kill-append (string before-p)
8459 "Append STRING to the end of the latest kill in the kill ring.
8460 If BEFORE-P is non-nil, prepend STRING to the kill.
8461 If `interprogram-cut-function' is set, pass the resulting kill to
8463 (kill-new (if before-p
8464 (concat string (car kill-ring))
8465 (concat (car kill-ring) string))
8470 The @code{kill-append} function is fairly straightforward. It uses
8471 the @code{kill-new} function, which we will discuss in more detail in
8474 (Also, the function provides an optional argument called
8475 @code{yank-handler}; when invoked, this argument tells the function
8476 how to deal with properties added to the text, such as `bold' or
8479 @c !!! bug in GNU Emacs 22 version of kill-append ?
8480 It has a @code{let*} function to set the value of the first element of
8481 the kill ring to @code{cur}. (I do not know why the function does not
8482 use @code{let} instead; only one value is set in the expression.
8483 Perhaps this is a bug that produces no problems?)
8485 Consider the conditional that is one of the two arguments to
8486 @code{kill-new}. It uses @code{concat} to concatenate the new text to
8487 the @sc{car} of the kill ring. Whether it prepends or appends the
8488 text depends on the results of an @code{if} expression:
8492 (if before-p ; @r{if-part}
8493 (concat string cur) ; @r{then-part}
8494 (concat cur string)) ; @r{else-part}
8499 If the region being killed is before the region that was killed in the
8500 last command, then it should be prepended before the material that was
8501 saved in the previous kill; and conversely, if the killed text follows
8502 what was just killed, it should be appended after the previous text.
8503 The @code{if} expression depends on the predicate @code{before-p} to
8504 decide whether the newly saved text should be put before or after the
8505 previously saved text.
8507 The symbol @code{before-p} is the name of one of the arguments to
8508 @code{kill-append}. When the @code{kill-append} function is
8509 evaluated, it is bound to the value returned by evaluating the actual
8510 argument. In this case, this is the expression @code{(< end beg)}.
8511 This expression does not directly determine whether the killed text in
8512 this command is located before or after the kill text of the last
8513 command; what it does is determine whether the value of the variable
8514 @code{end} is less than the value of the variable @code{beg}. If it
8515 is, it means that the user is most likely heading towards the
8516 beginning of the buffer. Also, the result of evaluating the predicate
8517 expression, @code{(< end beg)}, will be true and the text will be
8518 prepended before the previous text. On the other hand, if the value of
8519 the variable @code{end} is greater than the value of the variable
8520 @code{beg}, the text will be appended after the previous text.
8523 When the newly saved text will be prepended, then the string with the new
8524 text will be concatenated before the old text:
8532 But if the text will be appended, it will be concatenated
8536 (concat cur string))
8539 To understand how this works, we first need to review the
8540 @code{concat} function. The @code{concat} function links together or
8541 unites two strings of text. The result is a string. For example:
8545 (concat "abc" "def")
8551 (car '("first element" "second element")))
8552 @result{} "new first element"
8555 '("first element" "second element")) " modified")
8556 @result{} "first element modified"
8560 We can now make sense of @code{kill-append}: it modifies the contents
8561 of the kill ring. The kill ring is a list, each element of which is
8562 saved text. The @code{kill-append} function uses the @code{kill-new}
8563 function which in turn uses the @code{setcar} function.
8565 @node kill-new function
8566 @unnumberedsubsubsec The @code{kill-new} function
8569 @c in GNU Emacs 22, additional documentation to kill-new:
8571 Optional third arguments YANK-HANDLER controls how the STRING is later
8572 inserted into a buffer; see `insert-for-yank' for details.
8573 When a yank handler is specified, STRING must be non-empty (the yank
8574 handler, if non-nil, is stored as a `yank-handler' text property on STRING).
8576 When the yank handler has a non-nil PARAM element, the original STRING
8577 argument is not used by `insert-for-yank'. However, since Lisp code
8578 may access and use elements from the kill ring directly, the STRING
8579 argument should still be a \"useful\" string for such uses."
8582 The @code{kill-new} function looks like this:
8586 (defun kill-new (string &optional replace yank-handler)
8587 "Make STRING the latest kill in the kill ring.
8588 Set `kill-ring-yank-pointer' to point to it.
8590 If `interprogram-cut-function' is non-nil, apply it to STRING.
8591 Optional second argument REPLACE non-nil means that STRING will replace
8592 the front of the kill ring, rather than being added to the list.
8596 (if (> (length string) 0)
8598 (put-text-property 0 (length string)
8599 'yank-handler yank-handler string))
8601 (signal 'args-out-of-range
8602 (list string "yank-handler specified for empty string"))))
8605 (if (fboundp 'menu-bar-update-yank-menu)
8606 (menu-bar-update-yank-menu string (and replace (car kill-ring))))
8609 (if (and replace kill-ring)
8610 (setcar kill-ring string)
8611 (push string kill-ring)
8612 (if (> (length kill-ring) kill-ring-max)
8613 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil)))
8616 (setq kill-ring-yank-pointer kill-ring)
8617 (if interprogram-cut-function
8618 (funcall interprogram-cut-function string (not replace))))
8623 (defun kill-new (string &optional replace)
8624 "Make STRING the latest kill in the kill ring.
8625 Set the kill-ring-yank pointer to point to it.
8626 If `interprogram-cut-function' is non-nil, apply it to STRING.
8627 Optional second argument REPLACE non-nil means that STRING will replace
8628 the front of the kill ring, rather than being added to the list."
8629 (and (fboundp 'menu-bar-update-yank-menu)
8630 (menu-bar-update-yank-menu string (and replace (car kill-ring))))
8631 (if (and replace kill-ring)
8632 (setcar kill-ring string)
8633 (setq kill-ring (cons string kill-ring))
8634 (if (> (length kill-ring) kill-ring-max)
8635 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil)))
8636 (setq kill-ring-yank-pointer kill-ring)
8637 (if interprogram-cut-function
8638 (funcall interprogram-cut-function string (not replace))))
8641 (Notice that the function is not interactive.)
8643 As usual, we can look at this function in parts.
8645 The function definition has an optional @code{yank-handler} argument,
8646 which when invoked tells the function how to deal with properties
8647 added to the text, such as `bold' or `italics'. We will skip that.
8650 The first line of the documentation makes sense:
8653 Make STRING the latest kill in the kill ring.
8657 Let's skip over the rest of the documentation for the moment.
8660 Also, let's skip over the initial @code{if} expression and those lines
8661 of code involving @code{menu-bar-update-yank-menu}. We will explain
8665 The critical lines are these:
8669 (if (and replace kill-ring)
8671 (setcar kill-ring string)
8675 (push string kill-ring)
8678 (setq kill-ring (cons string kill-ring))
8679 (if (> (length kill-ring) kill-ring-max)
8680 ;; @r{avoid overly long kill ring}
8681 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil)))
8684 (setq kill-ring-yank-pointer kill-ring)
8685 (if interprogram-cut-function
8686 (funcall interprogram-cut-function string (not replace))))
8690 The conditional test is @w{@code{(and replace kill-ring)}}.
8691 This will be true when two conditions are met: the kill ring has
8692 something in it, and the @code{replace} variable is true.
8695 When the @code{kill-append} function sets @code{replace} to be true
8696 and when the kill ring has at least one item in it, the @code{setcar}
8697 expression is executed:
8700 (setcar kill-ring string)
8703 The @code{setcar} function actually changes the first element of the
8704 @code{kill-ring} list to the value of @code{string}. It replaces the
8708 On the other hand, if the kill ring is empty, or replace is false, the
8709 else-part of the condition is executed:
8712 (push string kill-ring)
8717 @code{push} puts its first argument onto the second. It is similar to
8721 (setq kill-ring (cons string kill-ring))
8729 (add-to-list kill-ring string)
8733 When it is false, the expression first constructs a new version of the
8734 kill ring by prepending @code{string} to the existing kill ring as a
8735 new element (that is what the @code{push} does). Then it executes a
8736 second @code{if} clause. This second @code{if} clause keeps the kill
8737 ring from growing too long.
8739 Let's look at these two expressions in order.
8741 The @code{push} line of the else-part sets the new value of the kill
8742 ring to what results from adding the string being killed to the old
8745 We can see how this works with an example.
8751 (setq example-list '("here is a clause" "another clause"))
8756 After evaluating this expression with @kbd{C-x C-e}, you can evaluate
8757 @code{example-list} and see what it returns:
8762 @result{} ("here is a clause" "another clause")
8768 Now, we can add a new element on to this list by evaluating the
8769 following expression:
8770 @findex push, @r{example}
8773 (push "a third clause" example-list)
8778 When we evaluate @code{example-list}, we find its value is:
8783 @result{} ("a third clause" "here is a clause" "another clause")
8788 Thus, the third clause is added to the list by @code{push}.
8791 Now for the second part of the @code{if} clause. This expression
8792 keeps the kill ring from growing too long. It looks like this:
8796 (if (> (length kill-ring) kill-ring-max)
8797 (setcdr (nthcdr (1- kill-ring-max) kill-ring) nil))
8801 The code checks whether the length of the kill ring is greater than
8802 the maximum permitted length. This is the value of
8803 @code{kill-ring-max} (which is 60, by default). If the length of the
8804 kill ring is too long, then this code sets the last element of the
8805 kill ring to @code{nil}. It does this by using two functions,
8806 @code{nthcdr} and @code{setcdr}.
8808 We looked at @code{setcdr} earlier (@pxref{setcdr, , @code{setcdr}}).
8809 It sets the @sc{cdr} of a list, just as @code{setcar} sets the
8810 @sc{car} of a list. In this case, however, @code{setcdr} will not be
8811 setting the @sc{cdr} of the whole kill ring; the @code{nthcdr}
8812 function is used to cause it to set the @sc{cdr} of the next to last
8813 element of the kill ring---this means that since the @sc{cdr} of the
8814 next to last element is the last element of the kill ring, it will set
8815 the last element of the kill ring.
8817 @findex nthcdr, @r{example}
8818 The @code{nthcdr} function works by repeatedly taking the @sc{cdr} of a
8819 list---it takes the @sc{cdr} of the @sc{cdr} of the @sc{cdr}
8820 @dots{} It does this @var{N} times and returns the results.
8821 (@xref{nthcdr, , @code{nthcdr}}.)
8823 @findex setcdr, @r{example}
8824 Thus, if we had a four element list that was supposed to be three
8825 elements long, we could set the @sc{cdr} of the next to last element
8826 to @code{nil}, and thereby shorten the list. (If you set the last
8827 element to some other value than @code{nil}, which you could do, then
8828 you would not have shortened the list. @xref{setcdr, ,
8831 You can see shortening by evaluating the following three expressions
8832 in turn. First set the value of @code{trees} to @code{(maple oak pine
8833 birch)}, then set the @sc{cdr} of its second @sc{cdr} to @code{nil}
8834 and then find the value of @code{trees}:
8838 (setq trees '(maple oak pine birch))
8839 @result{} (maple oak pine birch)
8843 (setcdr (nthcdr 2 trees) nil)
8847 @result{} (maple oak pine)
8852 (The value returned by the @code{setcdr} expression is @code{nil} since
8853 that is what the @sc{cdr} is set to.)
8855 To repeat, in @code{kill-new}, the @code{nthcdr} function takes the
8856 @sc{cdr} a number of times that is one less than the maximum permitted
8857 size of the kill ring and @code{setcdr} sets the @sc{cdr} of that
8858 element (which will be the rest of the elements in the kill ring) to
8859 @code{nil}. This prevents the kill ring from growing too long.
8862 The next to last expression in the @code{kill-new} function is
8865 (setq kill-ring-yank-pointer kill-ring)
8868 The @code{kill-ring-yank-pointer} is a global variable that is set to be
8869 the @code{kill-ring}.
8871 Even though the @code{kill-ring-yank-pointer} is called a
8872 @samp{pointer}, it is a variable just like the kill ring. However, the
8873 name has been chosen to help humans understand how the variable is used.
8876 Now, to return to an early expression in the body of the function:
8880 (if (fboundp 'menu-bar-update-yank-menu)
8881 (menu-bar-update-yank-menu string (and replace (car kill-ring))))
8886 It starts with an @code{if} expression
8888 In this case, the expression tests first to see whether
8889 @code{menu-bar-update-yank-menu} exists as a function, and if so,
8890 calls it. The @code{fboundp} function returns true if the symbol it
8891 is testing has a function definition that `is not void'. If the
8892 symbol's function definition were void, we would receive an error
8893 message, as we did when we created errors intentionally (@pxref{Making
8894 Errors, , Generate an Error Message}).
8897 The then-part contains an expression whose first element is the
8898 function @code{and}.
8901 The @code{and} special form evaluates each of its arguments until one
8902 of the arguments returns a value of @code{nil}, in which case the
8903 @code{and} expression returns @code{nil}; however, if none of the
8904 arguments returns a value of @code{nil}, the value resulting from
8905 evaluating the last argument is returned. (Since such a value is not
8906 @code{nil}, it is considered true in Emacs Lisp.) In other words, an
8907 @code{and} expression returns a true value only if all its arguments
8908 are true. (@xref{Second Buffer Related Review}.)
8910 The expression determines whether the second argument to
8911 @code{menu-bar-update-yank-menu} is true or not.
8913 ;; If we're supposed to be extending an existing string, and that
8914 ;; string really is at the front of the menu, then update it in place.
8917 @code{menu-bar-update-yank-menu} is one of the functions that make it
8918 possible to use the `Select and Paste' menu in the Edit item of a menu
8919 bar; using a mouse, you can look at the various pieces of text you
8920 have saved and select one piece to paste.
8922 The last expression in the @code{kill-new} function adds the newly
8923 copied string to whatever facility exists for copying and pasting
8924 among different programs running in a windowing system. In the X
8925 Windowing system, for example, the @code{x-select-text} function takes
8926 the string and stores it in memory operated by X@. You can paste the
8927 string in another program, such as an Xterm.
8930 The expression looks like this:
8934 (if interprogram-cut-function
8935 (funcall interprogram-cut-function string (not replace))))
8939 If an @code{interprogram-cut-function} exists, then Emacs executes
8940 @code{funcall}, which in turn calls its first argument as a function
8941 and passes the remaining arguments to it. (Incidentally, as far as I
8942 can see, this @code{if} expression could be replaced by an @code{and}
8943 expression similar to the one in the first part of the function.)
8945 We are not going to discuss windowing systems and other programs
8946 further, but merely note that this is a mechanism that enables GNU
8947 Emacs to work easily and well with other programs.
8949 This code for placing text in the kill ring, either concatenated with
8950 an existing element or as a new element, leads us to the code for
8951 bringing back text that has been cut out of the buffer---the yank
8952 commands. However, before discussing the yank commands, it is better
8953 to learn how lists are implemented in a computer. This will make
8954 clear such mysteries as the use of the term `pointer'. But before
8955 that, we will digress into C.
8958 @c is this true in Emacs 22? Does not seems to be
8960 (If the @w{@code{(< end beg))}}
8961 expression is true, @code{kill-append} prepends the string to the just
8962 previously clipped text. For a detailed discussion, see
8963 @ref{kill-append function, , The @code{kill-append} function}.)
8965 If you then yank back the text, i.e., `paste' it, you get both
8966 pieces of text at once. That way, if you delete two words in a row,
8967 and then yank them back, you get both words, in their proper order,
8968 with one yank. (The @w{@code{(< end beg))}} expression makes sure the
8971 On the other hand, if the previous command is not @code{kill-region},
8972 then the @code{kill-new} function is called, which adds the text to
8973 the kill ring as the latest item, and sets the
8974 @code{kill-ring-yank-pointer} variable to point to it.
8978 @c Evidently, changed for Emacs 22. The zap-to-char command does not
8979 @c use the delete-and-extract-region function
8981 2006 Oct 26, the Digression into C is now OK but should come after
8982 copy-region-as-kill and filter-buffer-substring
8986 copy-region-as-kill is short, 12 lines, and uses
8987 filter-buffer-substring, which is longer, 39 lines
8988 and has delete-and-extract-region in it.
8989 delete-and-extract-region is written in C.
8991 see Initializing a Variable with @code{defvar}
8994 @node Digression into C
8995 @section Digression into C
8996 @findex delete-and-extract-region
8997 @cindex C, a digression into
8998 @cindex Digression into C
9000 The @code{copy-region-as-kill} function (@pxref{copy-region-as-kill, ,
9001 @code{copy-region-as-kill}}) uses the @code{filter-buffer-substring}
9002 function, which in turn uses the @code{delete-and-extract-region}
9003 function. It removes the contents of a region and you cannot get them
9006 Unlike the other code discussed here, the
9007 @code{delete-and-extract-region} function is not written in Emacs
9008 Lisp; it is written in C and is one of the primitives of the GNU Emacs
9009 system. Since it is very simple, I will digress briefly from Lisp and
9012 @c GNU Emacs 24 in src/editfns.c
9013 @c the DEFUN for delete-and-extract-region
9016 Like many of the other Emacs primitives,
9017 @code{delete-and-extract-region} is written as an instance of a C
9018 macro, a macro being a template for code. The complete macro looks
9023 DEFUN ("delete-and-extract-region", Fdelete_and_extract_region,
9024 Sdelete_and_extract_region, 2, 2, 0,
9025 doc: /* Delete the text between START and END and return it. */)
9026 (Lisp_Object start, Lisp_Object end)
9028 validate_region (&start, &end);
9029 if (XINT (start) == XINT (end))
9030 return empty_unibyte_string;
9031 return del_range_1 (XINT (start), XINT (end), 1, 1);
9036 Without going into the details of the macro writing process, let me
9037 point out that this macro starts with the word @code{DEFUN}. The word
9038 @code{DEFUN} was chosen since the code serves the same purpose as
9039 @code{defun} does in Lisp. (The @code{DEFUN} C macro is defined in
9040 @file{emacs/src/lisp.h}.)
9042 The word @code{DEFUN} is followed by seven parts inside of
9047 The first part is the name given to the function in Lisp,
9048 @code{delete-and-extract-region}.
9051 The second part is the name of the function in C,
9052 @code{Fdelete_and_extract_region}. By convention, it starts with
9053 @samp{F}. Since C does not use hyphens in names, underscores are used
9057 The third part is the name for the C constant structure that records
9058 information on this function for internal use. It is the name of the
9059 function in C but begins with an @samp{S} instead of an @samp{F}.
9062 The fourth and fifth parts specify the minimum and maximum number of
9063 arguments the function can have. This function demands exactly 2
9067 The sixth part is nearly like the argument that follows the
9068 @code{interactive} declaration in a function written in Lisp: a letter
9069 followed, perhaps, by a prompt. The only difference from the Lisp is
9070 when the macro is called with no arguments. Then you write a @code{0}
9071 (which is a `null string'), as in this macro.
9073 If you were to specify arguments, you would place them between
9074 quotation marks. The C macro for @code{goto-char} includes
9075 @code{"NGoto char: "} in this position to indicate that the function
9076 expects a raw prefix, in this case, a numerical location in a buffer,
9077 and provides a prompt.
9080 The seventh part is a documentation string, just like the one for a
9081 function written in Emacs Lisp. This is written as a C comment. (When
9082 you build Emacs, the program @command{lib-src/make-docfile} extracts
9083 these comments and uses them to make the ``real'' documentation.)
9087 In a C macro, the formal parameters come next, with a statement of
9088 what kind of object they are, followed by what might be called the `body'
9089 of the macro. For @code{delete-and-extract-region} the `body'
9090 consists of the following four lines:
9094 validate_region (&start, &end);
9095 if (XINT (start) == XINT (end))
9096 return empty_unibyte_string;
9097 return del_range_1 (XINT (start), XINT (end), 1, 1);
9101 The @code{validate_region} function checks whether the values
9102 passed as the beginning and end of the region are the proper type and
9103 are within range. If the beginning and end positions are the same,
9104 then return an empty string.
9106 The @code{del_range_1} function actually deletes the text. It is a
9107 complex function we will not look into. It updates the buffer and
9108 does other things. However, it is worth looking at the two arguments
9109 passed to @code{del_range}. These are @w{@code{XINT (start)}} and
9110 @w{@code{XINT (end)}}.
9112 As far as the C language is concerned, @code{start} and @code{end} are
9113 two integers that mark the beginning and end of the region to be
9114 deleted@footnote{More precisely, and requiring more expert knowledge
9115 to understand, the two integers are of type `Lisp_Object', which can
9116 also be a C union instead of an integer type.}.
9118 In early versions of Emacs, these two numbers were thirty-two bits
9119 long, but the code is slowly being generalized to handle other
9120 lengths. Three of the available bits are used to specify the type of
9121 information; the remaining bits are used as `content'.
9123 @samp{XINT} is a C macro that extracts the relevant number from the
9124 longer collection of bits; the three other bits are discarded.
9127 The command in @code{delete-and-extract-region} looks like this:
9130 del_range_1 (XINT (start), XINT (end), 1, 1);
9134 It deletes the region between the beginning position, @code{start},
9135 and the ending position, @code{end}.
9137 From the point of view of the person writing Lisp, Emacs is all very
9138 simple; but hidden underneath is a great deal of complexity to make it
9142 @section Initializing a Variable with @code{defvar}
9144 @cindex Initializing a variable
9145 @cindex Variable initialization
9150 copy-region-as-kill is short, 12 lines, and uses
9151 filter-buffer-substring, which is longer, 39 lines
9152 and has delete-and-extract-region in it.
9153 delete-and-extract-region is written in C.
9155 see Initializing a Variable with @code{defvar}
9159 The @code{copy-region-as-kill} function is written in Emacs Lisp. Two
9160 functions within it, @code{kill-append} and @code{kill-new}, copy a
9161 region in a buffer and save it in a variable called the
9162 @code{kill-ring}. This section describes how the @code{kill-ring}
9163 variable is created and initialized using the @code{defvar} special
9166 (Again we note that the term @code{kill-ring} is a misnomer. The text
9167 that is clipped out of the buffer can be brought back; it is not a ring
9168 of corpses, but a ring of resurrectable text.)
9170 In Emacs Lisp, a variable such as the @code{kill-ring} is created and
9171 given an initial value by using the @code{defvar} special form. The
9172 name comes from ``define variable''.
9174 The @code{defvar} special form is similar to @code{setq} in that it sets
9175 the value of a variable. It is unlike @code{setq} in two ways: first,
9176 it only sets the value of the variable if the variable does not already
9177 have a value. If the variable already has a value, @code{defvar} does
9178 not override the existing value. Second, @code{defvar} has a
9179 documentation string.
9181 (There is a related macro, @code{defcustom}, designed for variables
9182 that people customize. It has more features than @code{defvar}.
9183 (@xref{defcustom, , Setting Variables with @code{defcustom}}.)
9186 * See variable current value::
9187 * defvar and asterisk::
9191 @node See variable current value
9192 @unnumberedsubsec Seeing the Current Value of a Variable
9195 You can see the current value of a variable, any variable, by using
9196 the @code{describe-variable} function, which is usually invoked by
9197 typing @kbd{C-h v}. If you type @kbd{C-h v} and then @code{kill-ring}
9198 (followed by @key{RET}) when prompted, you will see what is in your
9199 current kill ring---this may be quite a lot! Conversely, if you have
9200 been doing nothing this Emacs session except read this document, you
9201 may have nothing in it. Also, you will see the documentation for
9207 List of killed text sequences.
9208 Since the kill ring is supposed to interact nicely with cut-and-paste
9209 facilities offered by window systems, use of this variable should
9212 interact nicely with `interprogram-cut-function' and
9213 `interprogram-paste-function'. The functions `kill-new',
9214 `kill-append', and `current-kill' are supposed to implement this
9215 interaction; you may want to use them instead of manipulating the kill
9221 The kill ring is defined by a @code{defvar} in the following way:
9225 (defvar kill-ring nil
9226 "List of killed text sequences.
9232 In this variable definition, the variable is given an initial value of
9233 @code{nil}, which makes sense, since if you have saved nothing, you want
9234 nothing back if you give a @code{yank} command. The documentation
9235 string is written just like the documentation string of a @code{defun}.
9236 As with the documentation string of the @code{defun}, the first line of
9237 the documentation should be a complete sentence, since some commands,
9238 like @code{apropos}, print only the first line of documentation.
9239 Succeeding lines should not be indented; otherwise they look odd when
9240 you use @kbd{C-h v} (@code{describe-variable}).
9242 @node defvar and asterisk
9243 @subsection @code{defvar} and an asterisk
9244 @findex defvar @r{for a user customizable variable}
9245 @findex defvar @r{with an asterisk}
9247 In the past, Emacs used the @code{defvar} special form both for
9248 internal variables that you would not expect a user to change and for
9249 variables that you do expect a user to change. Although you can still
9250 use @code{defvar} for user customizable variables, please use
9251 @code{defcustom} instead, since it provides a path into
9252 the Customization commands. (@xref{defcustom, , Specifying Variables
9253 using @code{defcustom}}.)
9255 When you specified a variable using the @code{defvar} special form,
9256 you could distinguish a variable that a user might want to change from
9257 others by typing an asterisk, @samp{*}, in the first column of its
9258 documentation string. For example:
9262 (defvar shell-command-default-error-buffer nil
9263 "*Buffer name for `shell-command' @dots{} error output.
9268 @findex set-variable
9270 You could (and still can) use the @code{set-variable} command to
9271 change the value of @code{shell-command-default-error-buffer}
9272 temporarily. However, options set using @code{set-variable} are set
9273 only for the duration of your editing session. The new values are not
9274 saved between sessions. Each time Emacs starts, it reads the original
9275 value, unless you change the value within your @file{.emacs} file,
9276 either by setting it manually or by using @code{customize}.
9277 @xref{Emacs Initialization, , Your @file{.emacs} File}.
9279 For me, the major use of the @code{set-variable} command is to suggest
9280 variables that I might want to set in my @file{.emacs} file. There
9281 are now more than 700 such variables, far too many to remember
9282 readily. Fortunately, you can press @key{TAB} after calling the
9283 @code{M-x set-variable} command to see the list of variables.
9284 (@xref{Examining, , Examining and Setting Variables, emacs,
9285 The GNU Emacs Manual}.)
9288 @node cons & search-fwd Review
9291 Here is a brief summary of some recently introduced functions.
9296 @code{car} returns the first element of a list; @code{cdr} returns the
9297 second and subsequent elements of a list.
9304 (car '(1 2 3 4 5 6 7))
9306 (cdr '(1 2 3 4 5 6 7))
9307 @result{} (2 3 4 5 6 7)
9312 @code{cons} constructs a list by prepending its first argument to its
9326 @code{funcall} evaluates its first argument as a function. It passes
9327 its remaining arguments to its first argument.
9330 Return the result of taking @sc{cdr} `n' times on a list.
9338 The `rest of the rest', as it were.
9345 (nthcdr 3 '(1 2 3 4 5 6 7))
9352 @code{setcar} changes the first element of a list; @code{setcdr}
9353 changes the second and subsequent elements of a list.
9360 (setq triple '(1 2 3))
9367 (setcdr triple '("foo" "bar"))
9370 @result{} (37 "foo" "bar")
9375 Evaluate each argument in sequence and then return the value of the
9388 @item save-restriction
9389 Record whatever narrowing is in effect in the current buffer, if any,
9390 and restore that narrowing after evaluating the arguments.
9392 @item search-forward
9393 Search for a string, and if the string is found, move point. With a
9394 regular expression, use the similar @code{re-search-forward}.
9395 (@xref{Regexp Search, , Regular Expression Searches}, for an
9396 explanation of regular expression patterns and searches.)
9400 @code{search-forward} and @code{re-search-forward} take four
9405 The string or regular expression to search for.
9408 Optionally, the limit of the search.
9411 Optionally, what to do if the search fails, return @code{nil} or an
9415 Optionally, how many times to repeat the search; if negative, the
9416 search goes backwards.
9420 @itemx delete-and-extract-region
9421 @itemx copy-region-as-kill
9423 @code{kill-region} cuts the text between point and mark from the
9424 buffer and stores that text in the kill ring, so you can get it back
9427 @code{copy-region-as-kill} copies the text between point and mark into
9428 the kill ring, from which you can get it by yanking. The function
9429 does not cut or remove the text from the buffer.
9432 @code{delete-and-extract-region} removes the text between point and
9433 mark from the buffer and throws it away. You cannot get it back.
9434 (This is not an interactive command.)
9437 @node search Exercises
9438 @section Searching Exercises
9442 Write an interactive function that searches for a string. If the
9443 search finds the string, leave point after it and display a message
9444 that says ``Found!''. (Do not use @code{search-forward} for the name
9445 of this function; if you do, you will overwrite the existing version of
9446 @code{search-forward} that comes with Emacs. Use a name such as
9447 @code{test-search} instead.)
9450 Write a function that prints the third element of the kill ring in the
9451 echo area, if any; if the kill ring does not contain a third element,
9452 print an appropriate message.
9455 @node List Implementation
9456 @chapter How Lists are Implemented
9457 @cindex Lists in a computer
9459 In Lisp, atoms are recorded in a straightforward fashion; if the
9460 implementation is not straightforward in practice, it is, nonetheless,
9461 straightforward in theory. The atom @samp{rose}, for example, is
9462 recorded as the four contiguous letters @samp{r}, @samp{o}, @samp{s},
9463 @samp{e}. A list, on the other hand, is kept differently. The mechanism
9464 is equally simple, but it takes a moment to get used to the idea. A
9465 list is kept using a series of pairs of pointers. In the series, the
9466 first pointer in each pair points to an atom or to another list, and the
9467 second pointer in each pair points to the next pair, or to the symbol
9468 @code{nil}, which marks the end of the list.
9470 A pointer itself is quite simply the electronic address of what is
9471 pointed to. Hence, a list is kept as a series of electronic addresses.
9474 * Lists diagrammed::
9475 * Symbols as Chest:: Exploring a powerful metaphor.
9480 @node Lists diagrammed
9481 @unnumberedsec Lists diagrammed
9484 For example, the list @code{(rose violet buttercup)} has three elements,
9485 @samp{rose}, @samp{violet}, and @samp{buttercup}. In the computer, the
9486 electronic address of @samp{rose} is recorded in a segment of computer
9487 memory along with the address that gives the electronic address of where
9488 the atom @samp{violet} is located; and that address (the one that tells
9489 where @samp{violet} is located) is kept along with an address that tells
9490 where the address for the atom @samp{buttercup} is located.
9493 This sounds more complicated than it is and is easier seen in a diagram:
9495 @c clear print-postscript-figures
9496 @c !!! cons-cell-diagram #1
9500 ___ ___ ___ ___ ___ ___
9501 |___|___|--> |___|___|--> |___|___|--> nil
9504 --> rose --> violet --> buttercup
9508 @ifset print-postscript-figures
9511 @center @image{cons-1}
9515 @ifclear print-postscript-figures
9519 ___ ___ ___ ___ ___ ___
9520 |___|___|--> |___|___|--> |___|___|--> nil
9523 --> rose --> violet --> buttercup
9530 In the diagram, each box represents a word of computer memory that
9531 holds a Lisp object, usually in the form of a memory address. The boxes,
9532 i.e., the addresses, are in pairs. Each arrow points to what the address
9533 is the address of, either an atom or another pair of addresses. The
9534 first box is the electronic address of @samp{rose} and the arrow points
9535 to @samp{rose}; the second box is the address of the next pair of boxes,
9536 the first part of which is the address of @samp{violet} and the second
9537 part of which is the address of the next pair. The very last box
9538 points to the symbol @code{nil}, which marks the end of the list.
9541 When a variable is set to a list with a function such as @code{setq},
9542 it stores the address of the first box in the variable. Thus,
9543 evaluation of the expression
9546 (setq bouquet '(rose violet buttercup))
9551 creates a situation like this:
9553 @c cons-cell-diagram #2
9559 | ___ ___ ___ ___ ___ ___
9560 --> |___|___|--> |___|___|--> |___|___|--> nil
9563 --> rose --> violet --> buttercup
9567 @ifset print-postscript-figures
9570 @center @image{cons-2}
9574 @ifclear print-postscript-figures
9580 | ___ ___ ___ ___ ___ ___
9581 --> |___|___|--> |___|___|--> |___|___|--> nil
9584 --> rose --> violet --> buttercup
9591 In this example, the symbol @code{bouquet} holds the address of the first
9595 This same list can be illustrated in a different sort of box notation
9598 @c cons-cell-diagram #2a
9604 | -------------- --------------- ----------------
9605 | | car | cdr | | car | cdr | | car | cdr |
9606 -->| rose | o------->| violet | o------->| butter- | nil |
9607 | | | | | | | cup | |
9608 -------------- --------------- ----------------
9612 @ifset print-postscript-figures
9615 @center @image{cons-2a}
9619 @ifclear print-postscript-figures
9625 | -------------- --------------- ----------------
9626 | | car | cdr | | car | cdr | | car | cdr |
9627 -->| rose | o------->| violet | o------->| butter- | nil |
9628 | | | | | | | cup | |
9629 -------------- --------------- ----------------
9635 (Symbols consist of more than pairs of addresses, but the structure of
9636 a symbol is made up of addresses. Indeed, the symbol @code{bouquet}
9637 consists of a group of address-boxes, one of which is the address of
9638 the printed word @samp{bouquet}, a second of which is the address of a
9639 function definition attached to the symbol, if any, a third of which
9640 is the address of the first pair of address-boxes for the list
9641 @code{(rose violet buttercup)}, and so on. Here we are showing that
9642 the symbol's third address-box points to the first pair of
9643 address-boxes for the list.)
9645 If a symbol is set to the @sc{cdr} of a list, the list itself is not
9646 changed; the symbol simply has an address further down the list. (In
9647 the jargon, @sc{car} and @sc{cdr} are `non-destructive'.) Thus,
9648 evaluation of the following expression
9651 (setq flowers (cdr bouquet))
9658 @c cons-cell-diagram #3
9665 | ___ ___ | ___ ___ ___ ___
9666 --> | | | --> | | | | | |
9667 |___|___|----> |___|___|--> |___|___|--> nil
9670 --> rose --> violet --> buttercup
9675 @ifset print-postscript-figures
9678 @center @image{cons-3}
9682 @ifclear print-postscript-figures
9689 | ___ ___ | ___ ___ ___ ___
9690 --> | | | --> | | | | | |
9691 |___|___|----> |___|___|--> |___|___|--> nil
9694 --> rose --> violet --> buttercup
9702 The value of @code{flowers} is @code{(violet buttercup)}, which is
9703 to say, the symbol @code{flowers} holds the address of the pair of
9704 address-boxes, the first of which holds the address of @code{violet},
9705 and the second of which holds the address of @code{buttercup}.
9707 A pair of address-boxes is called a @dfn{cons cell} or @dfn{dotted
9708 pair}. @xref{Cons Cell Type, , Cons Cell and List Types, elisp, The GNU Emacs Lisp
9709 Reference Manual}, and @ref{Dotted Pair Notation, , Dotted Pair
9710 Notation, elisp, The GNU Emacs Lisp Reference Manual}, for more
9711 information about cons cells and dotted pairs.
9714 The function @code{cons} adds a new pair of addresses to the front of
9715 a series of addresses like that shown above. For example, evaluating
9719 (setq bouquet (cons 'lily bouquet))
9726 @c cons-cell-diagram #4
9733 | ___ ___ ___ ___ | ___ ___ ___ ___
9734 --> | | | | | | --> | | | | | |
9735 |___|___|----> |___|___|----> |___|___|---->|___|___|--> nil
9738 --> lily --> rose --> violet --> buttercup
9743 @ifset print-postscript-figures
9746 @center @image{cons-4}
9750 @ifclear print-postscript-figures
9757 | ___ ___ ___ ___ | ___ ___ ___ ___
9758 --> | | | | | | --> | | | | | |
9759 |___|___|----> |___|___|----> |___|___|---->|___|___|--> nil
9762 --> lily --> rose --> violet --> buttercup
9771 However, this does not change the value of the symbol
9772 @code{flowers}, as you can see by evaluating the following,
9775 (eq (cdr (cdr bouquet)) flowers)
9779 which returns @code{t} for true.
9781 Until it is reset, @code{flowers} still has the value
9782 @code{(violet buttercup)}; that is, it has the address of the cons
9783 cell whose first address is of @code{violet}. Also, this does not
9784 alter any of the pre-existing cons cells; they are all still there.
9786 Thus, in Lisp, to get the @sc{cdr} of a list, you just get the address
9787 of the next cons cell in the series; to get the @sc{car} of a list,
9788 you get the address of the first element of the list; to @code{cons} a
9789 new element on a list, you add a new cons cell to the front of the list.
9790 That is all there is to it! The underlying structure of Lisp is
9793 And what does the last address in a series of cons cells refer to? It
9794 is the address of the empty list, of @code{nil}.
9796 In summary, when a Lisp variable is set to a value, it is provided with
9797 the address of the list to which the variable refers.
9799 @node Symbols as Chest
9800 @section Symbols as a Chest of Drawers
9801 @cindex Symbols as a Chest of Drawers
9802 @cindex Chest of Drawers, metaphor for a symbol
9803 @cindex Drawers, Chest of, metaphor for a symbol
9805 In an earlier section, I suggested that you might imagine a symbol as
9806 being a chest of drawers. The function definition is put in one
9807 drawer, the value in another, and so on. What is put in the drawer
9808 holding the value can be changed without affecting the contents of the
9809 drawer holding the function definition, and vice-verse.
9811 Actually, what is put in each drawer is the address of the value or
9812 function definition. It is as if you found an old chest in the attic,
9813 and in one of its drawers you found a map giving you directions to
9814 where the buried treasure lies.
9816 (In addition to its name, symbol definition, and variable value, a
9817 symbol has a `drawer' for a @dfn{property list} which can be used to
9818 record other information. Property lists are not discussed here; see
9819 @ref{Property Lists, , Property Lists, elisp, The GNU Emacs Lisp
9823 Here is a fanciful representation:
9825 @c chest-of-drawers diagram
9830 Chest of Drawers Contents of Drawers
9834 ---------------------
9835 | directions to | [map to]
9836 | symbol name | bouquet
9838 +---------------------+
9840 | symbol definition | [none]
9842 +---------------------+
9843 | directions to | [map to]
9844 | variable value | (rose violet buttercup)
9846 +---------------------+
9848 | property list | [not described here]
9850 +---------------------+
9856 @ifset print-postscript-figures
9859 @center @image{drawers}
9863 @ifclear print-postscript-figures
9868 Chest of Drawers Contents of Drawers
9872 ---------------------
9873 | directions to | [map to]
9874 | symbol name | bouquet
9876 +---------------------+
9878 | symbol definition | [none]
9880 +---------------------+
9881 | directions to | [map to]
9882 | variable value | (rose violet buttercup)
9884 +---------------------+
9886 | property list | [not described here]
9888 +---------------------+
9899 Set @code{flowers} to @code{violet} and @code{buttercup}. Cons two
9900 more flowers on to this list and set this new list to
9901 @code{more-flowers}. Set the @sc{car} of @code{flowers} to a fish.
9902 What does the @code{more-flowers} list now contain?
9905 @chapter Yanking Text Back
9907 @cindex Text retrieval
9908 @cindex Retrieving text
9909 @cindex Pasting text
9911 Whenever you cut text out of a buffer with a `kill' command in GNU Emacs,
9912 you can bring it back with a `yank' command. The text that is cut out of
9913 the buffer is put in the kill ring and the yank commands insert the
9914 appropriate contents of the kill ring back into a buffer (not necessarily
9915 the original buffer).
9917 A simple @kbd{C-y} (@code{yank}) command inserts the first item from
9918 the kill ring into the current buffer. If the @kbd{C-y} command is
9919 followed immediately by @kbd{M-y}, the first element is replaced by
9920 the second element. Successive @kbd{M-y} commands replace the second
9921 element with the third, fourth, or fifth element, and so on. When the
9922 last element in the kill ring is reached, it is replaced by the first
9923 element and the cycle is repeated. (Thus the kill ring is called a
9924 `ring' rather than just a `list'. However, the actual data structure
9925 that holds the text is a list.
9926 @xref{Kill Ring, , Handling the Kill Ring}, for the details of how the
9927 list is handled as a ring.)
9930 * Kill Ring Overview::
9931 * kill-ring-yank-pointer:: The kill ring is a list.
9932 * yank nthcdr Exercises:: The @code{kill-ring-yank-pointer} variable.
9935 @node Kill Ring Overview
9936 @section Kill Ring Overview
9937 @cindex Kill ring overview
9939 The kill ring is a list of textual strings. This is what it looks like:
9942 ("some text" "a different piece of text" "yet more text")
9945 If this were the contents of my kill ring and I pressed @kbd{C-y}, the
9946 string of characters saying @samp{some text} would be inserted in this
9947 buffer where my cursor is located.
9949 The @code{yank} command is also used for duplicating text by copying it.
9950 The copied text is not cut from the buffer, but a copy of it is put on the
9951 kill ring and is inserted by yanking it back.
9953 Three functions are used for bringing text back from the kill ring:
9954 @code{yank}, which is usually bound to @kbd{C-y}; @code{yank-pop},
9955 which is usually bound to @kbd{M-y}; and @code{rotate-yank-pointer},
9956 which is used by the two other functions.
9958 These functions refer to the kill ring through a variable called the
9959 @code{kill-ring-yank-pointer}. Indeed, the insertion code for both the
9960 @code{yank} and @code{yank-pop} functions is:
9963 (insert (car kill-ring-yank-pointer))
9967 (Well, no more. In GNU Emacs 22, the function has been replaced by
9968 @code{insert-for-yank} which calls @code{insert-for-yank-1}
9969 repetitively for each @code{yank-handler} segment. In turn,
9970 @code{insert-for-yank-1} strips text properties from the inserted text
9971 according to @code{yank-excluded-properties}. Otherwise, it is just
9972 like @code{insert}. We will stick with plain @code{insert} since it
9973 is easier to understand.)
9975 To begin to understand how @code{yank} and @code{yank-pop} work, it is
9976 first necessary to look at the @code{kill-ring-yank-pointer} variable.
9978 @node kill-ring-yank-pointer
9979 @section The @code{kill-ring-yank-pointer} Variable
9981 @code{kill-ring-yank-pointer} is a variable, just as @code{kill-ring} is
9982 a variable. It points to something by being bound to the value of what
9983 it points to, like any other Lisp variable.
9986 Thus, if the value of the kill ring is:
9989 ("some text" "a different piece of text" "yet more text")
9994 and the @code{kill-ring-yank-pointer} points to the second clause, the
9995 value of @code{kill-ring-yank-pointer} is:
9998 ("a different piece of text" "yet more text")
10001 As explained in the previous chapter (@pxref{List Implementation}), the
10002 computer does not keep two different copies of the text being pointed to
10003 by both the @code{kill-ring} and the @code{kill-ring-yank-pointer}. The
10004 words ``a different piece of text'' and ``yet more text'' are not
10005 duplicated. Instead, the two Lisp variables point to the same pieces of
10006 text. Here is a diagram:
10008 @c cons-cell-diagram #5
10012 kill-ring kill-ring-yank-pointer
10014 | ___ ___ | ___ ___ ___ ___
10015 ---> | | | --> | | | | | |
10016 |___|___|----> |___|___|--> |___|___|--> nil
10019 | | --> "yet more text"
10021 | --> "a different piece of text"
10028 @ifset print-postscript-figures
10031 @center @image{cons-5}
10035 @ifclear print-postscript-figures
10039 kill-ring kill-ring-yank-pointer
10041 | ___ ___ | ___ ___ ___ ___
10042 ---> | | | --> | | | | | |
10043 |___|___|----> |___|___|--> |___|___|--> nil
10046 | | --> "yet more text"
10048 | --> "a different piece of text
10057 Both the variable @code{kill-ring} and the variable
10058 @code{kill-ring-yank-pointer} are pointers. But the kill ring itself is
10059 usually described as if it were actually what it is composed of. The
10060 @code{kill-ring} is spoken of as if it were the list rather than that it
10061 points to the list. Conversely, the @code{kill-ring-yank-pointer} is
10062 spoken of as pointing to a list.
10064 These two ways of talking about the same thing sound confusing at first but
10065 make sense on reflection. The kill ring is generally thought of as the
10066 complete structure of data that holds the information of what has recently
10067 been cut out of the Emacs buffers. The @code{kill-ring-yank-pointer}
10068 on the other hand, serves to indicate---that is, to `point to'---that part
10069 of the kill ring of which the first element (the @sc{car}) will be
10073 In GNU Emacs 22, the @code{kill-new} function calls
10075 @code{(setq kill-ring-yank-pointer kill-ring)}
10077 (defun rotate-yank-pointer (arg)
10078 "Rotate the yanking point in the kill ring.
10079 With argument, rotate that many kills forward (or backward, if negative)."
10081 (current-kill arg))
10083 (defun current-kill (n &optional do-not-move)
10084 "Rotate the yanking point by N places, and then return that kill.
10085 If N is zero, `interprogram-paste-function' is set, and calling it
10086 returns a string, then that string is added to the front of the
10087 kill ring and returned as the latest kill.
10088 If optional arg DO-NOT-MOVE is non-nil, then don't actually move the
10089 yanking point; just return the Nth kill forward."
10090 (let ((interprogram-paste (and (= n 0)
10091 interprogram-paste-function
10092 (funcall interprogram-paste-function))))
10093 (if interprogram-paste
10095 ;; Disable the interprogram cut function when we add the new
10096 ;; text to the kill ring, so Emacs doesn't try to own the
10097 ;; selection, with identical text.
10098 (let ((interprogram-cut-function nil))
10099 (kill-new interprogram-paste))
10100 interprogram-paste)
10101 (or kill-ring (error "Kill ring is empty"))
10102 (let ((ARGth-kill-element
10103 (nthcdr (mod (- n (length kill-ring-yank-pointer))
10104 (length kill-ring))
10107 (setq kill-ring-yank-pointer ARGth-kill-element))
10108 (car ARGth-kill-element)))))
10113 @node yank nthcdr Exercises
10114 @section Exercises with @code{yank} and @code{nthcdr}
10118 Using @kbd{C-h v} (@code{describe-variable}), look at the value of
10119 your kill ring. Add several items to your kill ring; look at its
10120 value again. Using @kbd{M-y} (@code{yank-pop)}, move all the way
10121 around the kill ring. How many items were in your kill ring? Find
10122 the value of @code{kill-ring-max}. Was your kill ring full, or could
10123 you have kept more blocks of text within it?
10126 Using @code{nthcdr} and @code{car}, construct a series of expressions
10127 to return the first, second, third, and fourth elements of a list.
10130 @node Loops & Recursion
10131 @chapter Loops and Recursion
10132 @cindex Loops and recursion
10133 @cindex Recursion and loops
10134 @cindex Repetition (loops)
10136 Emacs Lisp has two primary ways to cause an expression, or a series of
10137 expressions, to be evaluated repeatedly: one uses a @code{while}
10138 loop, and the other uses @dfn{recursion}.
10140 Repetition can be very valuable. For example, to move forward four
10141 sentences, you need only write a program that will move forward one
10142 sentence and then repeat the process four times. Since a computer does
10143 not get bored or tired, such repetitive action does not have the
10144 deleterious effects that excessive or the wrong kinds of repetition can
10147 People mostly write Emacs Lisp functions using @code{while} loops and
10148 their kin; but you can use recursion, which provides a very powerful
10149 way to think about and then to solve problems@footnote{You can write
10150 recursive functions to be frugal or wasteful of mental or computer
10151 resources; as it happens, methods that people find easy---that are
10152 frugal of `mental resources'---sometimes use considerable computer
10153 resources. Emacs was designed to run on machines that we now consider
10154 limited and its default settings are conservative. You may want to
10155 increase the values of @code{max-specpdl-size} and
10156 @code{max-lisp-eval-depth}. In my @file{.emacs} file, I set them to
10157 15 and 30 times their default value.}.
10160 * while:: Causing a stretch of code to repeat.
10162 * Recursion:: Causing a function to call itself.
10163 * Looping exercise::
10167 @section @code{while}
10171 The @code{while} special form tests whether the value returned by
10172 evaluating its first argument is true or false. This is similar to what
10173 the Lisp interpreter does with an @code{if}; what the interpreter does
10174 next, however, is different.
10176 In a @code{while} expression, if the value returned by evaluating the
10177 first argument is false, the Lisp interpreter skips the rest of the
10178 expression (the @dfn{body} of the expression) and does not evaluate it.
10179 However, if the value is true, the Lisp interpreter evaluates the body
10180 of the expression and then again tests whether the first argument to
10181 @code{while} is true or false. If the value returned by evaluating the
10182 first argument is again true, the Lisp interpreter again evaluates the
10183 body of the expression.
10186 The template for a @code{while} expression looks like this:
10190 (while @var{true-or-false-test}
10196 * Looping with while:: Repeat so long as test returns true.
10197 * Loop Example:: A @code{while} loop that uses a list.
10198 * print-elements-of-list:: Uses @code{while}, @code{car}, @code{cdr}.
10199 * Incrementing Loop:: A loop with an incrementing counter.
10200 * Incrementing Loop Details::
10201 * Decrementing Loop:: A loop with a decrementing counter.
10205 @node Looping with while
10206 @unnumberedsubsec Looping with @code{while}
10209 So long as the true-or-false-test of the @code{while} expression
10210 returns a true value when it is evaluated, the body is repeatedly
10211 evaluated. This process is called a loop since the Lisp interpreter
10212 repeats the same thing again and again, like an airplane doing a loop.
10213 When the result of evaluating the true-or-false-test is false, the
10214 Lisp interpreter does not evaluate the rest of the @code{while}
10215 expression and `exits the loop'.
10217 Clearly, if the value returned by evaluating the first argument to
10218 @code{while} is always true, the body following will be evaluated
10219 again and again @dots{} and again @dots{} forever. Conversely, if the
10220 value returned is never true, the expressions in the body will never
10221 be evaluated. The craft of writing a @code{while} loop consists of
10222 choosing a mechanism such that the true-or-false-test returns true
10223 just the number of times that you want the subsequent expressions to
10224 be evaluated, and then have the test return false.
10226 The value returned by evaluating a @code{while} is the value of the
10227 true-or-false-test. An interesting consequence of this is that a
10228 @code{while} loop that evaluates without error will return @code{nil}
10229 or false regardless of whether it has looped 1 or 100 times or none at
10230 all. A @code{while} expression that evaluates successfully never
10231 returns a true value! What this means is that @code{while} is always
10232 evaluated for its side effects, which is to say, the consequences of
10233 evaluating the expressions within the body of the @code{while} loop.
10234 This makes sense. It is not the mere act of looping that is desired,
10235 but the consequences of what happens when the expressions in the loop
10236 are repeatedly evaluated.
10239 @subsection A @code{while} Loop and a List
10241 A common way to control a @code{while} loop is to test whether a list
10242 has any elements. If it does, the loop is repeated; but if it does not,
10243 the repetition is ended. Since this is an important technique, we will
10244 create a short example to illustrate it.
10246 A simple way to test whether a list has elements is to evaluate the
10247 list: if it has no elements, it is an empty list and will return the
10248 empty list, @code{()}, which is a synonym for @code{nil} or false. On
10249 the other hand, a list with elements will return those elements when it
10250 is evaluated. Since Emacs Lisp considers as true any value that is not
10251 @code{nil}, a list that returns elements will test true in a
10255 For example, you can set the variable @code{empty-list} to @code{nil} by
10256 evaluating the following @code{setq} expression:
10259 (setq empty-list ())
10263 After evaluating the @code{setq} expression, you can evaluate the
10264 variable @code{empty-list} in the usual way, by placing the cursor after
10265 the symbol and typing @kbd{C-x C-e}; @code{nil} will appear in your
10272 On the other hand, if you set a variable to be a list with elements, the
10273 list will appear when you evaluate the variable, as you can see by
10274 evaluating the following two expressions:
10278 (setq animals '(gazelle giraffe lion tiger))
10284 Thus, to create a @code{while} loop that tests whether there are any
10285 items in the list @code{animals}, the first part of the loop will be
10296 When the @code{while} tests its first argument, the variable
10297 @code{animals} is evaluated. It returns a list. So long as the list
10298 has elements, the @code{while} considers the results of the test to be
10299 true; but when the list is empty, it considers the results of the test
10302 To prevent the @code{while} loop from running forever, some mechanism
10303 needs to be provided to empty the list eventually. An oft-used
10304 technique is to have one of the subsequent forms in the @code{while}
10305 expression set the value of the list to be the @sc{cdr} of the list.
10306 Each time the @code{cdr} function is evaluated, the list will be made
10307 shorter, until eventually only the empty list will be left. At this
10308 point, the test of the @code{while} loop will return false, and the
10309 arguments to the @code{while} will no longer be evaluated.
10311 For example, the list of animals bound to the variable @code{animals}
10312 can be set to be the @sc{cdr} of the original list with the
10313 following expression:
10316 (setq animals (cdr animals))
10320 If you have evaluated the previous expressions and then evaluate this
10321 expression, you will see @code{(giraffe lion tiger)} appear in the echo
10322 area. If you evaluate the expression again, @code{(lion tiger)} will
10323 appear in the echo area. If you evaluate it again and yet again,
10324 @code{(tiger)} appears and then the empty list, shown by @code{nil}.
10326 A template for a @code{while} loop that uses the @code{cdr} function
10327 repeatedly to cause the true-or-false-test eventually to test false
10332 (while @var{test-whether-list-is-empty}
10334 @var{set-list-to-cdr-of-list})
10338 This test and use of @code{cdr} can be put together in a function that
10339 goes through a list and prints each element of the list on a line of its
10342 @node print-elements-of-list
10343 @subsection An Example: @code{print-elements-of-list}
10344 @findex print-elements-of-list
10346 The @code{print-elements-of-list} function illustrates a @code{while}
10349 @cindex @file{*scratch*} buffer
10350 The function requires several lines for its output. If you are
10351 reading this in a recent instance of GNU Emacs,
10352 @c GNU Emacs 21, GNU Emacs 22, or a later version,
10353 you can evaluate the following expression inside of Info, as usual.
10355 If you are using an earlier version of Emacs, you need to copy the
10356 necessary expressions to your @file{*scratch*} buffer and evaluate
10357 them there. This is because the echo area had only one line in the
10360 You can copy the expressions by marking the beginning of the region
10361 with @kbd{C-@key{SPC}} (@code{set-mark-command}), moving the cursor to
10362 the end of the region and then copying the region using @kbd{M-w}
10363 (@code{kill-ring-save}, which calls @code{copy-region-as-kill} and
10364 then provides visual feedback). In the @file{*scratch*}
10365 buffer, you can yank the expressions back by typing @kbd{C-y}
10368 After you have copied the expressions to the @file{*scratch*} buffer,
10369 evaluate each expression in turn. Be sure to evaluate the last
10370 expression, @code{(print-elements-of-list animals)}, by typing
10371 @kbd{C-u C-x C-e}, that is, by giving an argument to
10372 @code{eval-last-sexp}. This will cause the result of the evaluation
10373 to be printed in the @file{*scratch*} buffer instead of being printed
10374 in the echo area. (Otherwise you will see something like this in your
10375 echo area: @code{^Jgazelle^J^Jgiraffe^J^Jlion^J^Jtiger^Jnil}, in which
10376 each @samp{^J} stands for a `newline'.)
10379 In a recent instance of GNU Emacs, you can evaluate these expressions
10380 directly in the Info buffer, and the echo area will grow to show the
10385 (setq animals '(gazelle giraffe lion tiger))
10387 (defun print-elements-of-list (list)
10388 "Print each element of LIST on a line of its own."
10391 (setq list (cdr list))))
10393 (print-elements-of-list animals)
10399 When you evaluate the three expressions in sequence, you will see
10415 Each element of the list is printed on a line of its own (that is what
10416 the function @code{print} does) and then the value returned by the
10417 function is printed. Since the last expression in the function is the
10418 @code{while} loop, and since @code{while} loops always return
10419 @code{nil}, a @code{nil} is printed after the last element of the list.
10421 @node Incrementing Loop
10422 @subsection A Loop with an Incrementing Counter
10424 A loop is not useful unless it stops when it ought. Besides
10425 controlling a loop with a list, a common way of stopping a loop is to
10426 write the first argument as a test that returns false when the correct
10427 number of repetitions are complete. This means that the loop must
10428 have a counter---an expression that counts how many times the loop
10432 @node Incrementing Loop Details
10433 @unnumberedsubsec Details of an Incrementing Loop
10436 The test for a loop with an incrementing counter can be an expression
10437 such as @code{(< count desired-number)} which returns @code{t} for
10438 true if the value of @code{count} is less than the
10439 @code{desired-number} of repetitions and @code{nil} for false if the
10440 value of @code{count} is equal to or is greater than the
10441 @code{desired-number}. The expression that increments the count can
10442 be a simple @code{setq} such as @code{(setq count (1+ count))}, where
10443 @code{1+} is a built-in function in Emacs Lisp that adds 1 to its
10444 argument. (The expression @w{@code{(1+ count)}} has the same result
10445 as @w{@code{(+ count 1)}}, but is easier for a human to read.)
10448 The template for a @code{while} loop controlled by an incrementing
10449 counter looks like this:
10453 @var{set-count-to-initial-value}
10454 (while (< count desired-number) ; @r{true-or-false-test}
10456 (setq count (1+ count))) ; @r{incrementer}
10461 Note that you need to set the initial value of @code{count}; usually it
10465 * Incrementing Example:: Counting pebbles in a triangle.
10466 * Inc Example parts:: The parts of the function definition.
10467 * Inc Example altogether:: Putting the function definition together.
10470 @node Incrementing Example
10471 @unnumberedsubsubsec Example with incrementing counter
10473 Suppose you are playing on the beach and decide to make a triangle of
10474 pebbles, putting one pebble in the first row, two in the second row,
10475 three in the third row and so on, like this:
10493 @bullet{} @bullet{}
10494 @bullet{} @bullet{} @bullet{}
10495 @bullet{} @bullet{} @bullet{} @bullet{}
10502 (About 2500 years ago, Pythagoras and others developed the beginnings of
10503 number theory by considering questions such as this.)
10505 Suppose you want to know how many pebbles you will need to make a
10506 triangle with 7 rows?
10508 Clearly, what you need to do is add up the numbers from 1 to 7. There
10509 are two ways to do this; start with the smallest number, one, and add up
10510 the list in sequence, 1, 2, 3, 4 and so on; or start with the largest
10511 number and add the list going down: 7, 6, 5, 4 and so on. Because both
10512 mechanisms illustrate common ways of writing @code{while} loops, we will
10513 create two examples, one counting up and the other counting down. In
10514 this first example, we will start with 1 and add 2, 3, 4 and so on.
10516 If you are just adding up a short list of numbers, the easiest way to do
10517 it is to add up all the numbers at once. However, if you do not know
10518 ahead of time how many numbers your list will have, or if you want to be
10519 prepared for a very long list, then you need to design your addition so
10520 that what you do is repeat a simple process many times instead of doing
10521 a more complex process once.
10523 For example, instead of adding up all the pebbles all at once, what you
10524 can do is add the number of pebbles in the first row, 1, to the number
10525 in the second row, 2, and then add the total of those two rows to the
10526 third row, 3. Then you can add the number in the fourth row, 4, to the
10527 total of the first three rows; and so on.
10529 The critical characteristic of the process is that each repetitive
10530 action is simple. In this case, at each step we add only two numbers,
10531 the number of pebbles in the row and the total already found. This
10532 process of adding two numbers is repeated again and again until the last
10533 row has been added to the total of all the preceding rows. In a more
10534 complex loop the repetitive action might not be so simple, but it will
10535 be simpler than doing everything all at once.
10537 @node Inc Example parts
10538 @unnumberedsubsubsec The parts of the function definition
10540 The preceding analysis gives us the bones of our function definition:
10541 first, we will need a variable that we can call @code{total} that will
10542 be the total number of pebbles. This will be the value returned by
10545 Second, we know that the function will require an argument: this
10546 argument will be the total number of rows in the triangle. It can be
10547 called @code{number-of-rows}.
10549 Finally, we need a variable to use as a counter. We could call this
10550 variable @code{counter}, but a better name is @code{row-number}. That
10551 is because what the counter does in this function is count rows, and a
10552 program should be written to be as understandable as possible.
10554 When the Lisp interpreter first starts evaluating the expressions in the
10555 function, the value of @code{total} should be set to zero, since we have
10556 not added anything to it. Then the function should add the number of
10557 pebbles in the first row to the total, and then add the number of
10558 pebbles in the second to the total, and then add the number of
10559 pebbles in the third row to the total, and so on, until there are no
10560 more rows left to add.
10562 Both @code{total} and @code{row-number} are used only inside the
10563 function, so they can be declared as local variables with @code{let}
10564 and given initial values. Clearly, the initial value for @code{total}
10565 should be 0. The initial value of @code{row-number} should be 1,
10566 since we start with the first row. This means that the @code{let}
10567 statement will look like this:
10577 After the internal variables are declared and bound to their initial
10578 values, we can begin the @code{while} loop. The expression that serves
10579 as the test should return a value of @code{t} for true so long as the
10580 @code{row-number} is less than or equal to the @code{number-of-rows}.
10581 (If the expression tests true only so long as the row number is less
10582 than the number of rows in the triangle, the last row will never be
10583 added to the total; hence the row number has to be either less than or
10584 equal to the number of rows.)
10587 @findex <= @r{(less than or equal)}
10588 Lisp provides the @code{<=} function that returns true if the value of
10589 its first argument is less than or equal to the value of its second
10590 argument and false otherwise. So the expression that the @code{while}
10591 will evaluate as its test should look like this:
10594 (<= row-number number-of-rows)
10597 The total number of pebbles can be found by repeatedly adding the number
10598 of pebbles in a row to the total already found. Since the number of
10599 pebbles in the row is equal to the row number, the total can be found by
10600 adding the row number to the total. (Clearly, in a more complex
10601 situation, the number of pebbles in the row might be related to the row
10602 number in a more complicated way; if this were the case, the row number
10603 would be replaced by the appropriate expression.)
10606 (setq total (+ total row-number))
10610 What this does is set the new value of @code{total} to be equal to the
10611 sum of adding the number of pebbles in the row to the previous total.
10613 After setting the value of @code{total}, the conditions need to be
10614 established for the next repetition of the loop, if there is one. This
10615 is done by incrementing the value of the @code{row-number} variable,
10616 which serves as a counter. After the @code{row-number} variable has
10617 been incremented, the true-or-false-test at the beginning of the
10618 @code{while} loop tests whether its value is still less than or equal to
10619 the value of the @code{number-of-rows} and if it is, adds the new value
10620 of the @code{row-number} variable to the @code{total} of the previous
10621 repetition of the loop.
10624 The built-in Emacs Lisp function @code{1+} adds 1 to a number, so the
10625 @code{row-number} variable can be incremented with this expression:
10628 (setq row-number (1+ row-number))
10631 @node Inc Example altogether
10632 @unnumberedsubsubsec Putting the function definition together
10634 We have created the parts for the function definition; now we need to
10638 First, the contents of the @code{while} expression:
10642 (while (<= row-number number-of-rows) ; @r{true-or-false-test}
10643 (setq total (+ total row-number))
10644 (setq row-number (1+ row-number))) ; @r{incrementer}
10648 Along with the @code{let} expression varlist, this very nearly
10649 completes the body of the function definition. However, it requires
10650 one final element, the need for which is somewhat subtle.
10652 The final touch is to place the variable @code{total} on a line by
10653 itself after the @code{while} expression. Otherwise, the value returned
10654 by the whole function is the value of the last expression that is
10655 evaluated in the body of the @code{let}, and this is the value
10656 returned by the @code{while}, which is always @code{nil}.
10658 This may not be evident at first sight. It almost looks as if the
10659 incrementing expression is the last expression of the whole function.
10660 But that expression is part of the body of the @code{while}; it is the
10661 last element of the list that starts with the symbol @code{while}.
10662 Moreover, the whole of the @code{while} loop is a list within the body
10666 In outline, the function will look like this:
10670 (defun @var{name-of-function} (@var{argument-list})
10671 "@var{documentation}@dots{}"
10672 (let (@var{varlist})
10673 (while (@var{true-or-false-test})
10674 @var{body-of-while}@dots{} )
10675 @dots{} )) ; @r{Need final expression here.}
10679 The result of evaluating the @code{let} is what is going to be returned
10680 by the @code{defun} since the @code{let} is not embedded within any
10681 containing list, except for the @code{defun} as a whole. However, if
10682 the @code{while} is the last element of the @code{let} expression, the
10683 function will always return @code{nil}. This is not what we want!
10684 Instead, what we want is the value of the variable @code{total}. This
10685 is returned by simply placing the symbol as the last element of the list
10686 starting with @code{let}. It gets evaluated after the preceding
10687 elements of the list are evaluated, which means it gets evaluated after
10688 it has been assigned the correct value for the total.
10690 It may be easier to see this by printing the list starting with
10691 @code{let} all on one line. This format makes it evident that the
10692 @var{varlist} and @code{while} expressions are the second and third
10693 elements of the list starting with @code{let}, and the @code{total} is
10698 (let (@var{varlist}) (while (@var{true-or-false-test}) @var{body-of-while}@dots{} ) total)
10703 Putting everything together, the @code{triangle} function definition
10708 (defun triangle (number-of-rows) ; @r{Version with}
10709 ; @r{ incrementing counter.}
10710 "Add up the number of pebbles in a triangle.
10711 The first row has one pebble, the second row two pebbles,
10712 the third row three pebbles, and so on.
10713 The argument is NUMBER-OF-ROWS."
10718 (while (<= row-number number-of-rows)
10719 (setq total (+ total row-number))
10720 (setq row-number (1+ row-number)))
10726 After you have installed @code{triangle} by evaluating the function, you
10727 can try it out. Here are two examples:
10738 The sum of the first four numbers is 10 and the sum of the first seven
10741 @node Decrementing Loop
10742 @subsection Loop with a Decrementing Counter
10744 Another common way to write a @code{while} loop is to write the test
10745 so that it determines whether a counter is greater than zero. So long
10746 as the counter is greater than zero, the loop is repeated. But when
10747 the counter is equal to or less than zero, the loop is stopped. For
10748 this to work, the counter has to start out greater than zero and then
10749 be made smaller and smaller by a form that is evaluated
10752 The test will be an expression such as @code{(> counter 0)} which
10753 returns @code{t} for true if the value of @code{counter} is greater
10754 than zero, and @code{nil} for false if the value of @code{counter} is
10755 equal to or less than zero. The expression that makes the number
10756 smaller and smaller can be a simple @code{setq} such as @code{(setq
10757 counter (1- counter))}, where @code{1-} is a built-in function in
10758 Emacs Lisp that subtracts 1 from its argument.
10761 The template for a decrementing @code{while} loop looks like this:
10765 (while (> counter 0) ; @r{true-or-false-test}
10767 (setq counter (1- counter))) ; @r{decrementer}
10772 * Decrementing Example:: More pebbles on the beach.
10773 * Dec Example parts:: The parts of the function definition.
10774 * Dec Example altogether:: Putting the function definition together.
10777 @node Decrementing Example
10778 @unnumberedsubsubsec Example with decrementing counter
10780 To illustrate a loop with a decrementing counter, we will rewrite the
10781 @code{triangle} function so the counter decreases to zero.
10783 This is the reverse of the earlier version of the function. In this
10784 case, to find out how many pebbles are needed to make a triangle with
10785 3 rows, add the number of pebbles in the third row, 3, to the number
10786 in the preceding row, 2, and then add the total of those two rows to
10787 the row that precedes them, which is 1.
10789 Likewise, to find the number of pebbles in a triangle with 7 rows, add
10790 the number of pebbles in the seventh row, 7, to the number in the
10791 preceding row, which is 6, and then add the total of those two rows to
10792 the row that precedes them, which is 5, and so on. As in the previous
10793 example, each addition only involves adding two numbers, the total of
10794 the rows already added up and the number of pebbles in the row that is
10795 being added to the total. This process of adding two numbers is
10796 repeated again and again until there are no more pebbles to add.
10798 We know how many pebbles to start with: the number of pebbles in the
10799 last row is equal to the number of rows. If the triangle has seven
10800 rows, the number of pebbles in the last row is 7. Likewise, we know how
10801 many pebbles are in the preceding row: it is one less than the number in
10804 @node Dec Example parts
10805 @unnumberedsubsubsec The parts of the function definition
10807 We start with three variables: the total number of rows in the
10808 triangle; the number of pebbles in a row; and the total number of
10809 pebbles, which is what we want to calculate. These variables can be
10810 named @code{number-of-rows}, @code{number-of-pebbles-in-row}, and
10811 @code{total}, respectively.
10813 Both @code{total} and @code{number-of-pebbles-in-row} are used only
10814 inside the function and are declared with @code{let}. The initial
10815 value of @code{total} should, of course, be zero. However, the
10816 initial value of @code{number-of-pebbles-in-row} should be equal to
10817 the number of rows in the triangle, since the addition will start with
10821 This means that the beginning of the @code{let} expression will look
10827 (number-of-pebbles-in-row number-of-rows))
10832 The total number of pebbles can be found by repeatedly adding the number
10833 of pebbles in a row to the total already found, that is, by repeatedly
10834 evaluating the following expression:
10837 (setq total (+ total number-of-pebbles-in-row))
10841 After the @code{number-of-pebbles-in-row} is added to the @code{total},
10842 the @code{number-of-pebbles-in-row} should be decremented by one, since
10843 the next time the loop repeats, the preceding row will be
10844 added to the total.
10846 The number of pebbles in a preceding row is one less than the number of
10847 pebbles in a row, so the built-in Emacs Lisp function @code{1-} can be
10848 used to compute the number of pebbles in the preceding row. This can be
10849 done with the following expression:
10853 (setq number-of-pebbles-in-row
10854 (1- number-of-pebbles-in-row))
10858 Finally, we know that the @code{while} loop should stop making repeated
10859 additions when there are no pebbles in a row. So the test for
10860 the @code{while} loop is simply:
10863 (while (> number-of-pebbles-in-row 0)
10866 @node Dec Example altogether
10867 @unnumberedsubsubsec Putting the function definition together
10869 We can put these expressions together to create a function definition
10870 that works. However, on examination, we find that one of the local
10871 variables is unneeded!
10874 The function definition looks like this:
10878 ;;; @r{First subtractive version.}
10879 (defun triangle (number-of-rows)
10880 "Add up the number of pebbles in a triangle."
10882 (number-of-pebbles-in-row number-of-rows))
10883 (while (> number-of-pebbles-in-row 0)
10884 (setq total (+ total number-of-pebbles-in-row))
10885 (setq number-of-pebbles-in-row
10886 (1- number-of-pebbles-in-row)))
10891 As written, this function works.
10893 However, we do not need @code{number-of-pebbles-in-row}.
10895 @cindex Argument as local variable
10896 When the @code{triangle} function is evaluated, the symbol
10897 @code{number-of-rows} will be bound to a number, giving it an initial
10898 value. That number can be changed in the body of the function as if
10899 it were a local variable, without any fear that such a change will
10900 effect the value of the variable outside of the function. This is a
10901 very useful characteristic of Lisp; it means that the variable
10902 @code{number-of-rows} can be used anywhere in the function where
10903 @code{number-of-pebbles-in-row} is used.
10906 Here is a second version of the function written a bit more cleanly:
10910 (defun triangle (number) ; @r{Second version.}
10911 "Return sum of numbers 1 through NUMBER inclusive."
10913 (while (> number 0)
10914 (setq total (+ total number))
10915 (setq number (1- number)))
10920 In brief, a properly written @code{while} loop will consist of three parts:
10924 A test that will return false after the loop has repeated itself the
10925 correct number of times.
10928 An expression the evaluation of which will return the value desired
10929 after being repeatedly evaluated.
10932 An expression to change the value passed to the true-or-false-test so
10933 that the test returns false after the loop has repeated itself the right
10937 @node dolist dotimes
10938 @section Save your time: @code{dolist} and @code{dotimes}
10940 In addition to @code{while}, both @code{dolist} and @code{dotimes}
10941 provide for looping. Sometimes these are quicker to write than the
10942 equivalent @code{while} loop. Both are Lisp macros. (@xref{Macros, ,
10943 Macros, elisp, The GNU Emacs Lisp Reference Manual}. )
10945 @code{dolist} works like a @code{while} loop that `@sc{cdr}s down a
10946 list': @code{dolist} automatically shortens the list each time it
10947 loops---takes the @sc{cdr} of the list---and binds the @sc{car} of
10948 each shorter version of the list to the first of its arguments.
10950 @code{dotimes} loops a specific number of times: you specify the number.
10958 @unnumberedsubsec The @code{dolist} Macro
10961 Suppose, for example, you want to reverse a list, so that
10962 ``first'' ``second'' ``third'' becomes ``third'' ``second'' ``first''.
10965 In practice, you would use the @code{reverse} function, like this:
10969 (setq animals '(gazelle giraffe lion tiger))
10977 Here is how you could reverse the list using a @code{while} loop:
10981 (setq animals '(gazelle giraffe lion tiger))
10983 (defun reverse-list-with-while (list)
10984 "Using while, reverse the order of LIST."
10985 (let (value) ; make sure list starts empty
10987 (setq value (cons (car list) value))
10988 (setq list (cdr list)))
10991 (reverse-list-with-while animals)
10997 And here is how you could use the @code{dolist} macro:
11001 (setq animals '(gazelle giraffe lion tiger))
11003 (defun reverse-list-with-dolist (list)
11004 "Using dolist, reverse the order of LIST."
11005 (let (value) ; make sure list starts empty
11006 (dolist (element list value)
11007 (setq value (cons element value)))))
11009 (reverse-list-with-dolist animals)
11015 In Info, you can place your cursor after the closing parenthesis of
11016 each expression and type @kbd{C-x C-e}; in each case, you should see
11019 (tiger lion giraffe gazelle)
11025 For this example, the existing @code{reverse} function is obviously best.
11026 The @code{while} loop is just like our first example (@pxref{Loop
11027 Example, , A @code{while} Loop and a List}). The @code{while} first
11028 checks whether the list has elements; if so, it constructs a new list
11029 by adding the first element of the list to the existing list (which in
11030 the first iteration of the loop is @code{nil}). Since the second
11031 element is prepended in front of the first element, and the third
11032 element is prepended in front of the second element, the list is reversed.
11034 In the expression using a @code{while} loop,
11035 the @w{@code{(setq list (cdr list))}}
11036 expression shortens the list, so the @code{while} loop eventually
11037 stops. In addition, it provides the @code{cons} expression with a new
11038 first element by creating a new and shorter list at each repetition of
11041 The @code{dolist} expression does very much the same as the
11042 @code{while} expression, except that the @code{dolist} macro does some
11043 of the work you have to do when writing a @code{while} expression.
11045 Like a @code{while} loop, a @code{dolist} loops. What is different is
11046 that it automatically shortens the list each time it loops---it
11047 `@sc{cdr}s down the list' on its own---and it automatically binds
11048 the @sc{car} of each shorter version of the list to the first of its
11051 In the example, the @sc{car} of each shorter version of the list is
11052 referred to using the symbol @samp{element}, the list itself is called
11053 @samp{list}, and the value returned is called @samp{value}. The
11054 remainder of the @code{dolist} expression is the body.
11056 The @code{dolist} expression binds the @sc{car} of each shorter
11057 version of the list to @code{element} and then evaluates the body of
11058 the expression; and repeats the loop. The result is returned in
11062 @unnumberedsubsec The @code{dotimes} Macro
11065 The @code{dotimes} macro is similar to @code{dolist}, except that it
11066 loops a specific number of times.
11068 The first argument to @code{dotimes} is assigned the numbers 0, 1, 2
11069 and so forth each time around the loop, and the value of the third
11070 argument is returned. You need to provide the value of the second
11071 argument, which is how many times the macro loops.
11074 For example, the following binds the numbers from 0 up to, but not
11075 including, the number 3 to the first argument, @var{number}, and then
11076 constructs a list of the three numbers. (The first number is 0, the
11077 second number is 1, and the third number is 2; this makes a total of
11078 three numbers in all, starting with zero as the first number.)
11082 (let (value) ; otherwise a value is a void variable
11083 (dotimes (number 3 value)
11084 (setq value (cons number value))))
11091 @code{dotimes} returns @code{value}, so the way to use
11092 @code{dotimes} is to operate on some expression @var{number} number of
11093 times and then return the result, either as a list or an atom.
11096 Here is an example of a @code{defun} that uses @code{dotimes} to add
11097 up the number of pebbles in a triangle.
11101 (defun triangle-using-dotimes (number-of-rows)
11102 "Using dotimes, add up the number of pebbles in a triangle."
11103 (let ((total 0)) ; otherwise a total is a void variable
11104 (dotimes (number number-of-rows total)
11105 (setq total (+ total (1+ number))))))
11107 (triangle-using-dotimes 4)
11115 A recursive function contains code that tells the Lisp interpreter to
11116 call a program that runs exactly like itself, but with slightly
11117 different arguments. The code runs exactly the same because it has
11118 the same name. However, even though the program has the same name, it
11119 is not the same entity. It is different. In the jargon, it is a
11120 different `instance'.
11122 Eventually, if the program is written correctly, the `slightly
11123 different arguments' will become sufficiently different from the first
11124 arguments that the final instance will stop.
11127 * Building Robots:: Same model, different serial number ...
11128 * Recursive Definition Parts:: Walk until you stop ...
11129 * Recursion with list:: Using a list as the test whether to recurse.
11130 * Recursive triangle function::
11131 * Recursion with cond::
11132 * Recursive Patterns:: Often used templates.
11133 * No Deferment:: Don't store up work ...
11134 * No deferment solution::
11137 @node Building Robots
11138 @subsection Building Robots: Extending the Metaphor
11139 @cindex Building robots
11140 @cindex Robots, building
11142 It is sometimes helpful to think of a running program as a robot that
11143 does a job. In doing its job, a recursive function calls on a second
11144 robot to help it. The second robot is identical to the first in every
11145 way, except that the second robot helps the first and has been
11146 passed different arguments than the first.
11148 In a recursive function, the second robot may call a third; and the
11149 third may call a fourth, and so on. Each of these is a different
11150 entity; but all are clones.
11152 Since each robot has slightly different instructions---the arguments
11153 will differ from one robot to the next---the last robot should know
11156 Let's expand on the metaphor in which a computer program is a robot.
11158 A function definition provides the blueprints for a robot. When you
11159 install a function definition, that is, when you evaluate a
11160 @code{defun} macro, you install the necessary equipment to build
11161 robots. It is as if you were in a factory, setting up an assembly
11162 line. Robots with the same name are built according to the same
11163 blueprints. So they have, as it were, the same `model number', but a
11164 different `serial number'.
11166 We often say that a recursive function `calls itself'. What we mean
11167 is that the instructions in a recursive function cause the Lisp
11168 interpreter to run a different function that has the same name and
11169 does the same job as the first, but with different arguments.
11171 It is important that the arguments differ from one instance to the
11172 next; otherwise, the process will never stop.
11174 @node Recursive Definition Parts
11175 @subsection The Parts of a Recursive Definition
11176 @cindex Parts of a Recursive Definition
11177 @cindex Recursive Definition Parts
11179 A recursive function typically contains a conditional expression which
11184 A true-or-false-test that determines whether the function is called
11185 again, here called the @dfn{do-again-test}.
11188 The name of the function. When this name is called, a new instance of
11189 the function---a new robot, as it were---is created and told what to do.
11192 An expression that returns a different value each time the function is
11193 called, here called the @dfn{next-step-expression}. Consequently, the
11194 argument (or arguments) passed to the new instance of the function
11195 will be different from that passed to the previous instance. This
11196 causes the conditional expression, the @dfn{do-again-test}, to test
11197 false after the correct number of repetitions.
11200 Recursive functions can be much simpler than any other kind of
11201 function. Indeed, when people first start to use them, they often look
11202 so mysteriously simple as to be incomprehensible. Like riding a
11203 bicycle, reading a recursive function definition takes a certain knack
11204 which is hard at first but then seems simple.
11207 There are several different common recursive patterns. A very simple
11208 pattern looks like this:
11212 (defun @var{name-of-recursive-function} (@var{argument-list})
11213 "@var{documentation}@dots{}"
11214 (if @var{do-again-test}
11216 (@var{name-of-recursive-function}
11217 @var{next-step-expression})))
11221 Each time a recursive function is evaluated, a new instance of it is
11222 created and told what to do. The arguments tell the instance what to do.
11224 An argument is bound to the value of the next-step-expression. Each
11225 instance runs with a different value of the next-step-expression.
11227 The value in the next-step-expression is used in the do-again-test.
11229 The value returned by the next-step-expression is passed to the new
11230 instance of the function, which evaluates it (or some
11231 transmogrification of it) to determine whether to continue or stop.
11232 The next-step-expression is designed so that the do-again-test returns
11233 false when the function should no longer be repeated.
11235 The do-again-test is sometimes called the @dfn{stop condition},
11236 since it stops the repetitions when it tests false.
11238 @node Recursion with list
11239 @subsection Recursion with a List
11241 The example of a @code{while} loop that printed the elements of a list
11242 of numbers can be written recursively. Here is the code, including
11243 an expression to set the value of the variable @code{animals} to a list.
11245 If you are reading this in Info in Emacs, you can evaluate this
11246 expression directly in Info. Otherwise, you must copy the example
11247 to the @file{*scratch*} buffer and evaluate each expression there.
11248 Use @kbd{C-u C-x C-e} to evaluate the
11249 @code{(print-elements-recursively animals)} expression so that the
11250 results are printed in the buffer; otherwise the Lisp interpreter will
11251 try to squeeze the results into the one line of the echo area.
11253 Also, place your cursor immediately after the last closing parenthesis
11254 of the @code{print-elements-recursively} function, before the comment.
11255 Otherwise, the Lisp interpreter will try to evaluate the comment.
11257 @findex print-elements-recursively
11260 (setq animals '(gazelle giraffe lion tiger))
11262 (defun print-elements-recursively (list)
11263 "Print each element of LIST on a line of its own.
11265 (when list ; @r{do-again-test}
11266 (print (car list)) ; @r{body}
11267 (print-elements-recursively ; @r{recursive call}
11268 (cdr list)))) ; @r{next-step-expression}
11270 (print-elements-recursively animals)
11274 The @code{print-elements-recursively} function first tests whether
11275 there is any content in the list; if there is, the function prints the
11276 first element of the list, the @sc{car} of the list. Then the
11277 function `invokes itself', but gives itself as its argument, not the
11278 whole list, but the second and subsequent elements of the list, the
11279 @sc{cdr} of the list.
11281 Put another way, if the list is not empty, the function invokes
11282 another instance of code that is similar to the initial code, but is a
11283 different thread of execution, with different arguments than the first
11286 Put in yet another way, if the list is not empty, the first robot
11287 assembles a second robot and tells it what to do; the second robot is
11288 a different individual from the first, but is the same model.
11290 When the second evaluation occurs, the @code{when} expression is
11291 evaluated and if true, prints the first element of the list it
11292 receives as its argument (which is the second element of the original
11293 list). Then the function `calls itself' with the @sc{cdr} of the list
11294 it is invoked with, which (the second time around) is the @sc{cdr} of
11295 the @sc{cdr} of the original list.
11297 Note that although we say that the function `calls itself', what we
11298 mean is that the Lisp interpreter assembles and instructs a new
11299 instance of the program. The new instance is a clone of the first,
11300 but is a separate individual.
11302 Each time the function `invokes itself', it invokes itself on a
11303 shorter version of the original list. It creates a new instance that
11304 works on a shorter list.
11306 Eventually, the function invokes itself on an empty list. It creates
11307 a new instance whose argument is @code{nil}. The conditional expression
11308 tests the value of @code{list}. Since the value of @code{list} is
11309 @code{nil}, the @code{when} expression tests false so the then-part is
11310 not evaluated. The function as a whole then returns @code{nil}.
11313 When you evaluate the expression @code{(print-elements-recursively
11314 animals)} in the @file{*scratch*} buffer, you see this result:
11330 @node Recursive triangle function
11331 @subsection Recursion in Place of a Counter
11332 @findex triangle-recursively
11335 The @code{triangle} function described in a previous section can also
11336 be written recursively. It looks like this:
11340 (defun triangle-recursively (number)
11341 "Return the sum of the numbers 1 through NUMBER inclusive.
11343 (if (= number 1) ; @r{do-again-test}
11345 (+ number ; @r{else-part}
11346 (triangle-recursively ; @r{recursive call}
11347 (1- number))))) ; @r{next-step-expression}
11349 (triangle-recursively 7)
11354 You can install this function by evaluating it and then try it by
11355 evaluating @code{(triangle-recursively 7)}. (Remember to put your
11356 cursor immediately after the last parenthesis of the function
11357 definition, before the comment.) The function evaluates to 28.
11359 To understand how this function works, let's consider what happens in the
11360 various cases when the function is passed 1, 2, 3, or 4 as the value of
11364 * Recursive Example arg of 1 or 2::
11365 * Recursive Example arg of 3 or 4::
11369 @node Recursive Example arg of 1 or 2
11370 @unnumberedsubsubsec An argument of 1 or 2
11373 First, what happens if the value of the argument is 1?
11375 The function has an @code{if} expression after the documentation
11376 string. It tests whether the value of @code{number} is equal to 1; if
11377 so, Emacs evaluates the then-part of the @code{if} expression, which
11378 returns the number 1 as the value of the function. (A triangle with
11379 one row has one pebble in it.)
11381 Suppose, however, that the value of the argument is 2. In this case,
11382 Emacs evaluates the else-part of the @code{if} expression.
11385 The else-part consists of an addition, the recursive call to
11386 @code{triangle-recursively} and a decrementing action; and it looks like
11390 (+ number (triangle-recursively (1- number)))
11393 When Emacs evaluates this expression, the innermost expression is
11394 evaluated first; then the other parts in sequence. Here are the steps
11398 @item Step 1 @w{ } Evaluate the innermost expression.
11400 The innermost expression is @code{(1- number)} so Emacs decrements the
11401 value of @code{number} from 2 to 1.
11403 @item Step 2 @w{ } Evaluate the @code{triangle-recursively} function.
11405 The Lisp interpreter creates an individual instance of
11406 @code{triangle-recursively}. It does not matter that this function is
11407 contained within itself. Emacs passes the result Step 1 as the
11408 argument used by this instance of the @code{triangle-recursively}
11411 In this case, Emacs evaluates @code{triangle-recursively} with an
11412 argument of 1. This means that this evaluation of
11413 @code{triangle-recursively} returns 1.
11415 @item Step 3 @w{ } Evaluate the value of @code{number}.
11417 The variable @code{number} is the second element of the list that
11418 starts with @code{+}; its value is 2.
11420 @item Step 4 @w{ } Evaluate the @code{+} expression.
11422 The @code{+} expression receives two arguments, the first
11423 from the evaluation of @code{number} (Step 3) and the second from the
11424 evaluation of @code{triangle-recursively} (Step 2).
11426 The result of the addition is the sum of 2 plus 1, and the number 3 is
11427 returned, which is correct. A triangle with two rows has three
11431 @node Recursive Example arg of 3 or 4
11432 @unnumberedsubsubsec An argument of 3 or 4
11434 Suppose that @code{triangle-recursively} is called with an argument of
11438 @item Step 1 @w{ } Evaluate the do-again-test.
11440 The @code{if} expression is evaluated first. This is the do-again
11441 test and returns false, so the else-part of the @code{if} expression
11442 is evaluated. (Note that in this example, the do-again-test causes
11443 the function to call itself when it tests false, not when it tests
11446 @item Step 2 @w{ } Evaluate the innermost expression of the else-part.
11448 The innermost expression of the else-part is evaluated, which decrements
11449 3 to 2. This is the next-step-expression.
11451 @item Step 3 @w{ } Evaluate the @code{triangle-recursively} function.
11453 The number 2 is passed to the @code{triangle-recursively} function.
11455 We already know what happens when Emacs evaluates @code{triangle-recursively} with
11456 an argument of 2. After going through the sequence of actions described
11457 earlier, it returns a value of 3. So that is what will happen here.
11459 @item Step 4 @w{ } Evaluate the addition.
11461 3 will be passed as an argument to the addition and will be added to the
11462 number with which the function was called, which is 3.
11466 The value returned by the function as a whole will be 6.
11468 Now that we know what will happen when @code{triangle-recursively} is
11469 called with an argument of 3, it is evident what will happen if it is
11470 called with an argument of 4:
11474 In the recursive call, the evaluation of
11477 (triangle-recursively (1- 4))
11482 will return the value of evaluating
11485 (triangle-recursively 3)
11489 which is 6 and this value will be added to 4 by the addition in the
11494 The value returned by the function as a whole will be 10.
11496 Each time @code{triangle-recursively} is evaluated, it evaluates a
11497 version of itself---a different instance of itself---with a smaller
11498 argument, until the argument is small enough so that it does not
11501 Note that this particular design for a recursive function
11502 requires that operations be deferred.
11504 Before @code{(triangle-recursively 7)} can calculate its answer, it
11505 must call @code{(triangle-recursively 6)}; and before
11506 @code{(triangle-recursively 6)} can calculate its answer, it must call
11507 @code{(triangle-recursively 5)}; and so on. That is to say, the
11508 calculation that @code{(triangle-recursively 7)} makes must be
11509 deferred until @code{(triangle-recursively 6)} makes its calculation;
11510 and @code{(triangle-recursively 6)} must defer until
11511 @code{(triangle-recursively 5)} completes; and so on.
11513 If each of these instances of @code{triangle-recursively} are thought
11514 of as different robots, the first robot must wait for the second to
11515 complete its job, which must wait until the third completes, and so
11518 There is a way around this kind of waiting, which we will discuss in
11519 @ref{No Deferment, , Recursion without Deferments}.
11521 @node Recursion with cond
11522 @subsection Recursion Example Using @code{cond}
11525 The version of @code{triangle-recursively} described earlier is written
11526 with the @code{if} special form. It can also be written using another
11527 special form called @code{cond}. The name of the special form
11528 @code{cond} is an abbreviation of the word @samp{conditional}.
11530 Although the @code{cond} special form is not used as often in the
11531 Emacs Lisp sources as @code{if}, it is used often enough to justify
11535 The template for a @code{cond} expression looks like this:
11545 where the @var{body} is a series of lists.
11548 Written out more fully, the template looks like this:
11553 (@var{first-true-or-false-test} @var{first-consequent})
11554 (@var{second-true-or-false-test} @var{second-consequent})
11555 (@var{third-true-or-false-test} @var{third-consequent})
11560 When the Lisp interpreter evaluates the @code{cond} expression, it
11561 evaluates the first element (the @sc{car} or true-or-false-test) of
11562 the first expression in a series of expressions within the body of the
11565 If the true-or-false-test returns @code{nil} the rest of that
11566 expression, the consequent, is skipped and the true-or-false-test of the
11567 next expression is evaluated. When an expression is found whose
11568 true-or-false-test returns a value that is not @code{nil}, the
11569 consequent of that expression is evaluated. The consequent can be one
11570 or more expressions. If the consequent consists of more than one
11571 expression, the expressions are evaluated in sequence and the value of
11572 the last one is returned. If the expression does not have a consequent,
11573 the value of the true-or-false-test is returned.
11575 If none of the true-or-false-tests test true, the @code{cond} expression
11576 returns @code{nil}.
11579 Written using @code{cond}, the @code{triangle} function looks like this:
11583 (defun triangle-using-cond (number)
11584 (cond ((<= number 0) 0)
11587 (+ number (triangle-using-cond (1- number))))))
11592 In this example, the @code{cond} returns 0 if the number is less than or
11593 equal to 0, it returns 1 if the number is 1 and it evaluates @code{(+
11594 number (triangle-using-cond (1- number)))} if the number is greater than
11597 @node Recursive Patterns
11598 @subsection Recursive Patterns
11599 @cindex Recursive Patterns
11601 Here are three common recursive patterns. Each involves a list.
11602 Recursion does not need to involve lists, but Lisp is designed for lists
11603 and this provides a sense of its primal capabilities.
11612 @unnumberedsubsubsec Recursive Pattern: @emph{every}
11613 @cindex Every, type of recursive pattern
11614 @cindex Recursive pattern: every
11616 In the @code{every} recursive pattern, an action is performed on every
11620 The basic pattern is:
11624 If a list be empty, return @code{nil}.
11626 Else, act on the beginning of the list (the @sc{car} of the list)
11629 through a recursive call by the function on the rest (the
11630 @sc{cdr}) of the list,
11632 and, optionally, combine the acted-on element, using @code{cons},
11633 with the results of acting on the rest.
11642 (defun square-each (numbers-list)
11643 "Square each of a NUMBERS LIST, recursively."
11644 (if (not numbers-list) ; do-again-test
11647 (* (car numbers-list) (car numbers-list))
11648 (square-each (cdr numbers-list))))) ; next-step-expression
11652 (square-each '(1 2 3))
11659 If @code{numbers-list} is empty, do nothing. But if it has content,
11660 construct a list combining the square of the first number in the list
11661 with the result of the recursive call.
11663 (The example follows the pattern exactly: @code{nil} is returned if
11664 the numbers' list is empty. In practice, you would write the
11665 conditional so it carries out the action when the numbers' list is not
11668 The @code{print-elements-recursively} function (@pxref{Recursion with
11669 list, , Recursion with a List}) is another example of an @code{every}
11670 pattern, except in this case, rather than bring the results together
11671 using @code{cons}, we print each element of output.
11674 The @code{print-elements-recursively} function looks like this:
11678 (setq animals '(gazelle giraffe lion tiger))
11682 (defun print-elements-recursively (list)
11683 "Print each element of LIST on a line of its own.
11685 (when list ; @r{do-again-test}
11686 (print (car list)) ; @r{body}
11687 (print-elements-recursively ; @r{recursive call}
11688 (cdr list)))) ; @r{next-step-expression}
11690 (print-elements-recursively animals)
11695 The pattern for @code{print-elements-recursively} is:
11699 When the list is empty, do nothing.
11701 But when the list has at least one element,
11704 act on the beginning of the list (the @sc{car} of the list),
11706 and make a recursive call on the rest (the @sc{cdr}) of the list.
11711 @unnumberedsubsubsec Recursive Pattern: @emph{accumulate}
11712 @cindex Accumulate, type of recursive pattern
11713 @cindex Recursive pattern: accumulate
11715 Another recursive pattern is called the @code{accumulate} pattern. In
11716 the @code{accumulate} recursive pattern, an action is performed on
11717 every element of a list and the result of that action is accumulated
11718 with the results of performing the action on the other elements.
11720 This is very like the `every' pattern using @code{cons}, except that
11721 @code{cons} is not used, but some other combiner.
11728 If a list be empty, return zero or some other constant.
11730 Else, act on the beginning of the list (the @sc{car} of the list),
11733 and combine that acted-on element, using @code{+} or
11734 some other combining function, with
11736 a recursive call by the function on the rest (the @sc{cdr}) of the list.
11741 Here is an example:
11745 (defun add-elements (numbers-list)
11746 "Add the elements of NUMBERS-LIST together."
11747 (if (not numbers-list)
11749 (+ (car numbers-list) (add-elements (cdr numbers-list)))))
11753 (add-elements '(1 2 3 4))
11758 @xref{Files List, , Making a List of Files}, for an example of the
11759 accumulate pattern.
11762 @unnumberedsubsubsec Recursive Pattern: @emph{keep}
11763 @cindex Keep, type of recursive pattern
11764 @cindex Recursive pattern: keep
11766 A third recursive pattern is called the @code{keep} pattern.
11767 In the @code{keep} recursive pattern, each element of a list is tested;
11768 the element is acted on and the results are kept only if the element
11771 Again, this is very like the `every' pattern, except the element is
11772 skipped unless it meets a criterion.
11775 The pattern has three parts:
11779 If a list be empty, return @code{nil}.
11781 Else, if the beginning of the list (the @sc{car} of the list) passes
11785 act on that element and combine it, using @code{cons} with
11787 a recursive call by the function on the rest (the @sc{cdr}) of the list.
11790 Otherwise, if the beginning of the list (the @sc{car} of the list) fails
11794 skip on that element,
11796 and, recursively call the function on the rest (the @sc{cdr}) of the list.
11801 Here is an example that uses @code{cond}:
11805 (defun keep-three-letter-words (word-list)
11806 "Keep three letter words in WORD-LIST."
11808 ;; First do-again-test: stop-condition
11809 ((not word-list) nil)
11811 ;; Second do-again-test: when to act
11812 ((eq 3 (length (symbol-name (car word-list))))
11813 ;; combine acted-on element with recursive call on shorter list
11814 (cons (car word-list) (keep-three-letter-words (cdr word-list))))
11816 ;; Third do-again-test: when to skip element;
11817 ;; recursively call shorter list with next-step expression
11818 (t (keep-three-letter-words (cdr word-list)))))
11822 (keep-three-letter-words '(one two three four five six))
11823 @result{} (one two six)
11827 It goes without saying that you need not use @code{nil} as the test for
11828 when to stop; and you can, of course, combine these patterns.
11831 @subsection Recursion without Deferments
11832 @cindex Deferment in recursion
11833 @cindex Recursion without Deferments
11835 Let's consider again what happens with the @code{triangle-recursively}
11836 function. We will find that the intermediate calculations are
11837 deferred until all can be done.
11840 Here is the function definition:
11844 (defun triangle-recursively (number)
11845 "Return the sum of the numbers 1 through NUMBER inclusive.
11847 (if (= number 1) ; @r{do-again-test}
11849 (+ number ; @r{else-part}
11850 (triangle-recursively ; @r{recursive call}
11851 (1- number))))) ; @r{next-step-expression}
11855 What happens when we call this function with a argument of 7?
11857 The first instance of the @code{triangle-recursively} function adds
11858 the number 7 to the value returned by a second instance of
11859 @code{triangle-recursively}, an instance that has been passed an
11860 argument of 6. That is to say, the first calculation is:
11863 (+ 7 (triangle-recursively 6))
11867 The first instance of @code{triangle-recursively}---you may want to
11868 think of it as a little robot---cannot complete its job. It must hand
11869 off the calculation for @code{(triangle-recursively 6)} to a second
11870 instance of the program, to a second robot. This second individual is
11871 completely different from the first one; it is, in the jargon, a
11872 `different instantiation'. Or, put another way, it is a different
11873 robot. It is the same model as the first; it calculates triangle
11874 numbers recursively; but it has a different serial number.
11876 And what does @code{(triangle-recursively 6)} return? It returns the
11877 number 6 added to the value returned by evaluating
11878 @code{triangle-recursively} with an argument of 5. Using the robot
11879 metaphor, it asks yet another robot to help it.
11885 (+ 7 6 (triangle-recursively 5))
11889 And what happens next?
11892 (+ 7 6 5 (triangle-recursively 4))
11895 Each time @code{triangle-recursively} is called, except for the last
11896 time, it creates another instance of the program---another robot---and
11897 asks it to make a calculation.
11900 Eventually, the full addition is set up and performed:
11906 This design for the function defers the calculation of the first step
11907 until the second can be done, and defers that until the third can be
11908 done, and so on. Each deferment means the computer must remember what
11909 is being waited on. This is not a problem when there are only a few
11910 steps, as in this example. But it can be a problem when there are
11913 @node No deferment solution
11914 @subsection No Deferment Solution
11915 @cindex No deferment solution
11916 @cindex Defermentless solution
11917 @cindex Solution without deferment
11919 The solution to the problem of deferred operations is to write in a
11920 manner that does not defer operations@footnote{The phrase @dfn{tail
11921 recursive} is used to describe such a process, one that uses
11922 `constant space'.}. This requires
11923 writing to a different pattern, often one that involves writing two
11924 function definitions, an `initialization' function and a `helper'
11927 The `initialization' function sets up the job; the `helper' function
11931 Here are the two function definitions for adding up numbers. They are
11932 so simple, I find them hard to understand.
11936 (defun triangle-initialization (number)
11937 "Return the sum of the numbers 1 through NUMBER inclusive.
11938 This is the `initialization' component of a two function
11939 duo that uses recursion."
11940 (triangle-recursive-helper 0 0 number))
11946 (defun triangle-recursive-helper (sum counter number)
11947 "Return SUM, using COUNTER, through NUMBER inclusive.
11948 This is the `helper' component of a two function duo
11949 that uses recursion."
11950 (if (> counter number)
11952 (triangle-recursive-helper (+ sum counter) ; @r{sum}
11953 (1+ counter) ; @r{counter}
11954 number))) ; @r{number}
11959 Install both function definitions by evaluating them, then call
11960 @code{triangle-initialization} with 2 rows:
11964 (triangle-initialization 2)
11969 The `initialization' function calls the first instance of the `helper'
11970 function with three arguments: zero, zero, and a number which is the
11971 number of rows in the triangle.
11973 The first two arguments passed to the `helper' function are
11974 initialization values. These values are changed when
11975 @code{triangle-recursive-helper} invokes new instances.@footnote{The
11976 jargon is mildly confusing: @code{triangle-recursive-helper} uses a
11977 process that is iterative in a procedure that is recursive. The
11978 process is called iterative because the computer need only record the
11979 three values, @code{sum}, @code{counter}, and @code{number}; the
11980 procedure is recursive because the function `calls itself'. On the
11981 other hand, both the process and the procedure used by
11982 @code{triangle-recursively} are called recursive. The word
11983 `recursive' has different meanings in the two contexts.}
11985 Let's see what happens when we have a triangle that has one row. (This
11986 triangle will have one pebble in it!)
11989 @code{triangle-initialization} will call its helper with
11990 the arguments @w{@code{0 0 1}}. That function will run the conditional
11991 test whether @code{(> counter number)}:
11999 and find that the result is false, so it will invoke
12000 the else-part of the @code{if} clause:
12004 (triangle-recursive-helper
12005 (+ sum counter) ; @r{sum plus counter} @result{} @r{sum}
12006 (1+ counter) ; @r{increment counter} @result{} @r{counter}
12007 number) ; @r{number stays the same}
12013 which will first compute:
12017 (triangle-recursive-helper (+ 0 0) ; @r{sum}
12018 (1+ 0) ; @r{counter}
12022 (triangle-recursive-helper 0 1 1)
12026 Again, @code{(> counter number)} will be false, so again, the Lisp
12027 interpreter will evaluate @code{triangle-recursive-helper}, creating a
12028 new instance with new arguments.
12031 This new instance will be;
12035 (triangle-recursive-helper
12036 (+ sum counter) ; @r{sum plus counter} @result{} @r{sum}
12037 (1+ counter) ; @r{increment counter} @result{} @r{counter}
12038 number) ; @r{number stays the same}
12042 (triangle-recursive-helper 1 2 1)
12046 In this case, the @code{(> counter number)} test will be true! So the
12047 instance will return the value of the sum, which will be 1, as
12050 Now, let's pass @code{triangle-initialization} an argument
12051 of 2, to find out how many pebbles there are in a triangle with two rows.
12053 That function calls @code{(triangle-recursive-helper 0 0 2)}.
12056 In stages, the instances called will be:
12060 @r{sum counter number}
12061 (triangle-recursive-helper 0 1 2)
12063 (triangle-recursive-helper 1 2 2)
12065 (triangle-recursive-helper 3 3 2)
12069 When the last instance is called, the @code{(> counter number)} test
12070 will be true, so the instance will return the value of @code{sum},
12073 This kind of pattern helps when you are writing functions that can use
12074 many resources in a computer.
12077 @node Looping exercise
12078 @section Looping Exercise
12082 Write a function similar to @code{triangle} in which each row has a
12083 value which is the square of the row number. Use a @code{while} loop.
12086 Write a function similar to @code{triangle} that multiplies instead of
12090 Rewrite these two functions recursively. Rewrite these functions
12093 @c comma in printed title causes problem in Info cross reference
12095 Write a function for Texinfo mode that creates an index entry at the
12096 beginning of a paragraph for every @samp{@@dfn} within the paragraph.
12097 (In a Texinfo file, @samp{@@dfn} marks a definition. This book is
12098 written in Texinfo.)
12100 Many of the functions you will need are described in two of the
12101 previous chapters, @ref{Cutting & Storing Text, , Cutting and Storing
12102 Text}, and @ref{Yanking, , Yanking Text Back}. If you use
12103 @code{forward-paragraph} to put the index entry at the beginning of
12104 the paragraph, you will have to use @w{@kbd{C-h f}}
12105 (@code{describe-function}) to find out how to make the command go
12108 For more information, see
12110 @ref{Indicating, , Indicating Definitions, texinfo}.
12113 @ref{Indicating, , Indicating, texinfo, Texinfo Manual}, which goes to
12114 a Texinfo manual in the current directory. Or, if you are on the
12116 @uref{http://www.gnu.org/software/texinfo/manual/texinfo/}
12119 ``Indicating Definitions, Commands, etc.'' in @cite{Texinfo, The GNU
12120 Documentation Format}.
12124 @node Regexp Search
12125 @chapter Regular Expression Searches
12126 @cindex Searches, illustrating
12127 @cindex Regular expression searches
12128 @cindex Patterns, searching for
12129 @cindex Motion by sentence and paragraph
12130 @cindex Sentences, movement by
12131 @cindex Paragraphs, movement by
12133 Regular expression searches are used extensively in GNU Emacs. The
12134 two functions, @code{forward-sentence} and @code{forward-paragraph},
12135 illustrate these searches well. They use regular expressions to find
12136 where to move point. The phrase `regular expression' is often written
12139 Regular expression searches are described in @ref{Regexp Search, ,
12140 Regular Expression Search, emacs, The GNU Emacs Manual}, as well as in
12141 @ref{Regular Expressions, , , elisp, The GNU Emacs Lisp Reference
12142 Manual}. In writing this chapter, I am presuming that you have at
12143 least a mild acquaintance with them. The major point to remember is
12144 that regular expressions permit you to search for patterns as well as
12145 for literal strings of characters. For example, the code in
12146 @code{forward-sentence} searches for the pattern of possible
12147 characters that could mark the end of a sentence, and moves point to
12150 Before looking at the code for the @code{forward-sentence} function, it
12151 is worth considering what the pattern that marks the end of a sentence
12152 must be. The pattern is discussed in the next section; following that
12153 is a description of the regular expression search function,
12154 @code{re-search-forward}. The @code{forward-sentence} function
12155 is described in the section following. Finally, the
12156 @code{forward-paragraph} function is described in the last section of
12157 this chapter. @code{forward-paragraph} is a complex function that
12158 introduces several new features.
12161 * sentence-end:: The regular expression for @code{sentence-end}.
12162 * re-search-forward:: Very similar to @code{search-forward}.
12163 * forward-sentence:: A straightforward example of regexp search.
12164 * forward-paragraph:: A somewhat complex example.
12165 * etags:: How to create your own @file{TAGS} table.
12167 * re-search Exercises::
12171 @section The Regular Expression for @code{sentence-end}
12172 @findex sentence-end
12174 The symbol @code{sentence-end} is bound to the pattern that marks the
12175 end of a sentence. What should this regular expression be?
12177 Clearly, a sentence may be ended by a period, a question mark, or an
12178 exclamation mark. Indeed, in English, only clauses that end with one
12179 of those three characters should be considered the end of a sentence.
12180 This means that the pattern should include the character set:
12186 However, we do not want @code{forward-sentence} merely to jump to a
12187 period, a question mark, or an exclamation mark, because such a character
12188 might be used in the middle of a sentence. A period, for example, is
12189 used after abbreviations. So other information is needed.
12191 According to convention, you type two spaces after every sentence, but
12192 only one space after a period, a question mark, or an exclamation mark in
12193 the body of a sentence. So a period, a question mark, or an exclamation
12194 mark followed by two spaces is a good indicator of an end of sentence.
12195 However, in a file, the two spaces may instead be a tab or the end of a
12196 line. This means that the regular expression should include these three
12197 items as alternatives.
12200 This group of alternatives will look like this:
12211 Here, @samp{$} indicates the end of the line, and I have pointed out
12212 where the tab and two spaces are inserted in the expression. Both are
12213 inserted by putting the actual characters into the expression.
12215 Two backslashes, @samp{\\}, are required before the parentheses and
12216 vertical bars: the first backslash quotes the following backslash in
12217 Emacs; and the second indicates that the following character, the
12218 parenthesis or the vertical bar, is special.
12221 Also, a sentence may be followed by one or more carriage returns, like
12232 Like tabs and spaces, a carriage return is inserted into a regular
12233 expression by inserting it literally. The asterisk indicates that the
12234 @key{RET} is repeated zero or more times.
12236 But a sentence end does not consist only of a period, a question mark or
12237 an exclamation mark followed by appropriate space: a closing quotation
12238 mark or a closing brace of some kind may precede the space. Indeed more
12239 than one such mark or brace may precede the space. These require a
12240 expression that looks like this:
12246 In this expression, the first @samp{]} is the first character in the
12247 expression; the second character is @samp{"}, which is preceded by a
12248 @samp{\} to tell Emacs the @samp{"} is @emph{not} special. The last
12249 three characters are @samp{'}, @samp{)}, and @samp{@}}.
12251 All this suggests what the regular expression pattern for matching the
12252 end of a sentence should be; and, indeed, if we evaluate
12253 @code{sentence-end} we find that it returns the following value:
12258 @result{} "[.?!][]\"')@}]*\\($\\| \\| \\)[
12264 (Well, not in GNU Emacs 22; that is because of an effort to make the
12265 process simpler and to handle more glyphs and languages. When the
12266 value of @code{sentence-end} is @code{nil}, then use the value defined
12267 by the function @code{sentence-end}. (Here is a use of the difference
12268 between a value and a function in Emacs Lisp.) The function returns a
12269 value constructed from the variables @code{sentence-end-base},
12270 @code{sentence-end-double-space}, @code{sentence-end-without-period},
12271 and @code{sentence-end-without-space}. The critical variable is
12272 @code{sentence-end-base}; its global value is similar to the one
12273 described above but it also contains two additional quotation marks.
12274 These have differing degrees of curliness. The
12275 @code{sentence-end-without-period} variable, when true, tells Emacs
12276 that a sentence may end without a period, such as text in Thai.)
12280 (Note that here the @key{TAB}, two spaces, and @key{RET} are shown
12281 literally in the pattern.)
12283 This regular expression can be deciphered as follows:
12287 The first part of the pattern is the three characters, a period, a question
12288 mark and an exclamation mark, within square brackets. The pattern must
12289 begin with one or other of these characters.
12292 The second part of the pattern is the group of closing braces and
12293 quotation marks, which can appear zero or more times. These may follow
12294 the period, question mark or exclamation mark. In a regular expression,
12295 the backslash, @samp{\}, followed by the double quotation mark,
12296 @samp{"}, indicates the class of string-quote characters. Usually, the
12297 double quotation mark is the only character in this class. The
12298 asterisk, @samp{*}, indicates that the items in the previous group (the
12299 group surrounded by square brackets, @samp{[]}) may be repeated zero or
12302 @item \\($\\| \\| \\)
12303 The third part of the pattern is one or other of: either the end of a
12304 line, or two blank spaces, or a tab. The double back-slashes are used
12305 to prevent Emacs from reading the parentheses and vertical bars as part
12306 of the search pattern; the parentheses are used to mark the group and
12307 the vertical bars are used to indicated that the patterns to either side
12308 of them are alternatives. The dollar sign is used to indicate the end
12309 of a line and both the two spaces and the tab are each inserted as is to
12310 indicate what they are.
12313 Finally, the last part of the pattern indicates that the end of the line
12314 or the whitespace following the period, question mark or exclamation
12315 mark may, but need not, be followed by one or more carriage returns. In
12316 the pattern, the carriage return is inserted as an actual carriage
12317 return between square brackets but here it is shown as @key{RET}.
12321 @node re-search-forward
12322 @section The @code{re-search-forward} Function
12323 @findex re-search-forward
12325 The @code{re-search-forward} function is very like the
12326 @code{search-forward} function. (@xref{search-forward, , The
12327 @code{search-forward} Function}.)
12329 @code{re-search-forward} searches for a regular expression. If the
12330 search is successful, it leaves point immediately after the last
12331 character in the target. If the search is backwards, it leaves point
12332 just before the first character in the target. You may tell
12333 @code{re-search-forward} to return @code{t} for true. (Moving point
12334 is therefore a `side effect'.)
12336 Like @code{search-forward}, the @code{re-search-forward} function takes
12341 The first argument is the regular expression that the function searches
12342 for. The regular expression will be a string between quotation marks.
12345 The optional second argument limits how far the function will search; it is a
12346 bound, which is specified as a position in the buffer.
12349 The optional third argument specifies how the function responds to
12350 failure: @code{nil} as the third argument causes the function to
12351 signal an error (and print a message) when the search fails; any other
12352 value causes it to return @code{nil} if the search fails and @code{t}
12353 if the search succeeds.
12356 The optional fourth argument is the repeat count. A negative repeat
12357 count causes @code{re-search-forward} to search backwards.
12361 The template for @code{re-search-forward} looks like this:
12365 (re-search-forward "@var{regular-expression}"
12366 @var{limit-of-search}
12367 @var{what-to-do-if-search-fails}
12368 @var{repeat-count})
12372 The second, third, and fourth arguments are optional. However, if you
12373 want to pass a value to either or both of the last two arguments, you
12374 must also pass a value to all the preceding arguments. Otherwise, the
12375 Lisp interpreter will mistake which argument you are passing the value
12379 In the @code{forward-sentence} function, the regular expression will be
12380 the value of the variable @code{sentence-end}. In simple form, that is:
12384 "[.?!][]\"')@}]*\\($\\| \\| \\)[
12390 The limit of the search will be the end of the paragraph (since a
12391 sentence cannot go beyond a paragraph). If the search fails, the
12392 function will return @code{nil}; and the repeat count will be provided
12393 by the argument to the @code{forward-sentence} function.
12395 @node forward-sentence
12396 @section @code{forward-sentence}
12397 @findex forward-sentence
12399 The command to move the cursor forward a sentence is a straightforward
12400 illustration of how to use regular expression searches in Emacs Lisp.
12401 Indeed, the function looks longer and more complicated than it is; this
12402 is because the function is designed to go backwards as well as forwards;
12403 and, optionally, over more than one sentence. The function is usually
12404 bound to the key command @kbd{M-e}.
12407 * Complete forward-sentence::
12408 * fwd-sentence while loops:: Two @code{while} loops.
12409 * fwd-sentence re-search:: A regular expression search.
12413 @node Complete forward-sentence
12414 @unnumberedsubsec Complete @code{forward-sentence} function definition
12418 Here is the code for @code{forward-sentence}:
12423 (defun forward-sentence (&optional arg)
12424 "Move forward to next `sentence-end'. With argument, repeat.
12425 With negative argument, move backward repeatedly to `sentence-beginning'.
12427 The variable `sentence-end' is a regular expression that matches ends of
12428 sentences. Also, every paragraph boundary terminates sentences as well."
12432 (or arg (setq arg 1))
12433 (let ((opoint (point))
12434 (sentence-end (sentence-end)))
12436 (let ((pos (point))
12437 (par-beg (save-excursion (start-of-paragraph-text) (point))))
12438 (if (and (re-search-backward sentence-end par-beg t)
12439 (or (< (match-end 0) pos)
12440 (re-search-backward sentence-end par-beg t)))
12441 (goto-char (match-end 0))
12442 (goto-char par-beg)))
12443 (setq arg (1+ arg)))
12447 (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
12448 (if (re-search-forward sentence-end par-end t)
12449 (skip-chars-backward " \t\n")
12450 (goto-char par-end)))
12451 (setq arg (1- arg)))
12452 (constrain-to-field nil opoint t)))
12460 (defun forward-sentence (&optional arg)
12461 "Move forward to next sentence-end. With argument, repeat.
12462 With negative argument, move backward repeatedly to sentence-beginning.
12463 Sentence ends are identified by the value of sentence-end
12464 treated as a regular expression. Also, every paragraph boundary
12465 terminates sentences as well."
12469 (or arg (setq arg 1))
12472 (save-excursion (start-of-paragraph-text) (point))))
12473 (if (re-search-backward
12474 (concat sentence-end "[^ \t\n]") par-beg t)
12475 (goto-char (1- (match-end 0)))
12476 (goto-char par-beg)))
12477 (setq arg (1+ arg)))
12480 (save-excursion (end-of-paragraph-text) (point))))
12481 (if (re-search-forward sentence-end par-end t)
12482 (skip-chars-backward " \t\n")
12483 (goto-char par-end)))
12484 (setq arg (1- arg))))
12489 The function looks long at first sight and it is best to look at its
12490 skeleton first, and then its muscle. The way to see the skeleton is to
12491 look at the expressions that start in the left-most columns:
12495 (defun forward-sentence (&optional arg)
12496 "@var{documentation}@dots{}"
12498 (or arg (setq arg 1))
12499 (let ((opoint (point)) (sentence-end (sentence-end)))
12501 (let ((pos (point))
12502 (par-beg (save-excursion (start-of-paragraph-text) (point))))
12503 @var{rest-of-body-of-while-loop-when-going-backwards}
12505 (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
12506 @var{rest-of-body-of-while-loop-when-going-forwards}
12507 @var{handle-forms-and-equivalent}
12511 This looks much simpler! The function definition consists of
12512 documentation, an @code{interactive} expression, an @code{or}
12513 expression, a @code{let} expression, and @code{while} loops.
12515 Let's look at each of these parts in turn.
12517 We note that the documentation is thorough and understandable.
12519 The function has an @code{interactive "p"} declaration. This means
12520 that the processed prefix argument, if any, is passed to the
12521 function as its argument. (This will be a number.) If the function
12522 is not passed an argument (it is optional) then the argument
12523 @code{arg} will be bound to 1.
12525 When @code{forward-sentence} is called non-interactively without an
12526 argument, @code{arg} is bound to @code{nil}. The @code{or} expression
12527 handles this. What it does is either leave the value of @code{arg} as
12528 it is, but only if @code{arg} is bound to a value; or it sets the
12529 value of @code{arg} to 1, in the case when @code{arg} is bound to
12532 Next is a @code{let}. That specifies the values of two local
12533 variables, @code{point} and @code{sentence-end}. The local value of
12534 point, from before the search, is used in the
12535 @code{constrain-to-field} function which handles forms and
12536 equivalents. The @code{sentence-end} variable is set by the
12537 @code{sentence-end} function.
12539 @node fwd-sentence while loops
12540 @unnumberedsubsec The @code{while} loops
12542 Two @code{while} loops follow. The first @code{while} has a
12543 true-or-false-test that tests true if the prefix argument for
12544 @code{forward-sentence} is a negative number. This is for going
12545 backwards. The body of this loop is similar to the body of the second
12546 @code{while} clause, but it is not exactly the same. We will skip
12547 this @code{while} loop and concentrate on the second @code{while}
12551 The second @code{while} loop is for moving point forward. Its skeleton
12556 (while (> arg 0) ; @r{true-or-false-test}
12558 (if (@var{true-or-false-test})
12561 (setq arg (1- arg)))) ; @code{while} @r{loop decrementer}
12565 The @code{while} loop is of the decrementing kind.
12566 (@xref{Decrementing Loop, , A Loop with a Decrementing Counter}.) It
12567 has a true-or-false-test that tests true so long as the counter (in
12568 this case, the variable @code{arg}) is greater than zero; and it has a
12569 decrementer that subtracts 1 from the value of the counter every time
12572 If no prefix argument is given to @code{forward-sentence}, which is
12573 the most common way the command is used, this @code{while} loop will
12574 run once, since the value of @code{arg} will be 1.
12576 The body of the @code{while} loop consists of a @code{let} expression,
12577 which creates and binds a local variable, and has, as its body, an
12578 @code{if} expression.
12581 The body of the @code{while} loop looks like this:
12586 (save-excursion (end-of-paragraph-text) (point))))
12587 (if (re-search-forward sentence-end par-end t)
12588 (skip-chars-backward " \t\n")
12589 (goto-char par-end)))
12593 The @code{let} expression creates and binds the local variable
12594 @code{par-end}. As we shall see, this local variable is designed to
12595 provide a bound or limit to the regular expression search. If the
12596 search fails to find a proper sentence ending in the paragraph, it will
12597 stop on reaching the end of the paragraph.
12599 But first, let us examine how @code{par-end} is bound to the value of
12600 the end of the paragraph. What happens is that the @code{let} sets the
12601 value of @code{par-end} to the value returned when the Lisp interpreter
12602 evaluates the expression
12606 (save-excursion (end-of-paragraph-text) (point))
12611 In this expression, @code{(end-of-paragraph-text)} moves point to the
12612 end of the paragraph, @code{(point)} returns the value of point, and then
12613 @code{save-excursion} restores point to its original position. Thus,
12614 the @code{let} binds @code{par-end} to the value returned by the
12615 @code{save-excursion} expression, which is the position of the end of
12616 the paragraph. (The @code{end-of-paragraph-text} function uses
12617 @code{forward-paragraph}, which we will discuss shortly.)
12620 Emacs next evaluates the body of the @code{let}, which is an @code{if}
12621 expression that looks like this:
12625 (if (re-search-forward sentence-end par-end t) ; @r{if-part}
12626 (skip-chars-backward " \t\n") ; @r{then-part}
12627 (goto-char par-end))) ; @r{else-part}
12631 The @code{if} tests whether its first argument is true and if so,
12632 evaluates its then-part; otherwise, the Emacs Lisp interpreter
12633 evaluates the else-part. The true-or-false-test of the @code{if}
12634 expression is the regular expression search.
12636 It may seem odd to have what looks like the `real work' of
12637 the @code{forward-sentence} function buried here, but this is a common
12638 way this kind of operation is carried out in Lisp.
12640 @node fwd-sentence re-search
12641 @unnumberedsubsec The regular expression search
12643 The @code{re-search-forward} function searches for the end of the
12644 sentence, that is, for the pattern defined by the @code{sentence-end}
12645 regular expression. If the pattern is found---if the end of the sentence is
12646 found---then the @code{re-search-forward} function does two things:
12650 The @code{re-search-forward} function carries out a side effect, which
12651 is to move point to the end of the occurrence found.
12654 The @code{re-search-forward} function returns a value of true. This is
12655 the value received by the @code{if}, and means that the search was
12660 The side effect, the movement of point, is completed before the
12661 @code{if} function is handed the value returned by the successful
12662 conclusion of the search.
12664 When the @code{if} function receives the value of true from a successful
12665 call to @code{re-search-forward}, the @code{if} evaluates the then-part,
12666 which is the expression @code{(skip-chars-backward " \t\n")}. This
12667 expression moves backwards over any blank spaces, tabs or carriage
12668 returns until a printed character is found and then leaves point after
12669 the character. Since point has already been moved to the end of the
12670 pattern that marks the end of the sentence, this action leaves point
12671 right after the closing printed character of the sentence, which is
12674 On the other hand, if the @code{re-search-forward} function fails to
12675 find a pattern marking the end of the sentence, the function returns
12676 false. The false then causes the @code{if} to evaluate its third
12677 argument, which is @code{(goto-char par-end)}: it moves point to the
12678 end of the paragraph.
12680 (And if the text is in a form or equivalent, and point may not move
12681 fully, then the @code{constrain-to-field} function comes into play.)
12683 Regular expression searches are exceptionally useful and the pattern
12684 illustrated by @code{re-search-forward}, in which the search is the
12685 test of an @code{if} expression, is handy. You will see or write code
12686 incorporating this pattern often.
12688 @node forward-paragraph
12689 @section @code{forward-paragraph}: a Goldmine of Functions
12690 @findex forward-paragraph
12694 (defun forward-paragraph (&optional arg)
12695 "Move forward to end of paragraph.
12696 With argument ARG, do it ARG times;
12697 a negative argument ARG = -N means move backward N paragraphs.
12699 A line which `paragraph-start' matches either separates paragraphs
12700 \(if `paragraph-separate' matches it also) or is the first line of a paragraph.
12701 A paragraph end is the beginning of a line which is not part of the paragraph
12702 to which the end of the previous line belongs, or the end of the buffer.
12703 Returns the count of paragraphs left to move."
12705 (or arg (setq arg 1))
12706 (let* ((opoint (point))
12707 (fill-prefix-regexp
12708 (and fill-prefix (not (equal fill-prefix ""))
12709 (not paragraph-ignore-fill-prefix)
12710 (regexp-quote fill-prefix)))
12711 ;; Remove ^ from paragraph-start and paragraph-sep if they are there.
12712 ;; These regexps shouldn't be anchored, because we look for them
12713 ;; starting at the left-margin. This allows paragraph commands to
12714 ;; work normally with indented text.
12715 ;; This hack will not find problem cases like "whatever\\|^something".
12716 (parstart (if (and (not (equal "" paragraph-start))
12717 (equal ?^ (aref paragraph-start 0)))
12718 (substring paragraph-start 1)
12720 (parsep (if (and (not (equal "" paragraph-separate))
12721 (equal ?^ (aref paragraph-separate 0)))
12722 (substring paragraph-separate 1)
12723 paragraph-separate))
12725 (if fill-prefix-regexp
12726 (concat parsep "\\|"
12727 fill-prefix-regexp "[ \t]*$")
12729 ;; This is used for searching.
12730 (sp-parstart (concat "^[ \t]*\\(?:" parstart "\\|" parsep "\\)"))
12732 (while (and (< arg 0) (not (bobp)))
12733 (if (and (not (looking-at parsep))
12734 (re-search-backward "^\n" (max (1- (point)) (point-min)) t)
12735 (looking-at parsep))
12736 (setq arg (1+ arg))
12737 (setq start (point))
12738 ;; Move back over paragraph-separating lines.
12739 (forward-char -1) (beginning-of-line)
12740 (while (and (not (bobp))
12741 (progn (move-to-left-margin)
12742 (looking-at parsep)))
12746 (setq arg (1+ arg))
12747 ;; Go to end of the previous (non-separating) line.
12749 ;; Search back for line that starts or separates paragraphs.
12750 (if (if fill-prefix-regexp
12751 ;; There is a fill prefix; it overrides parstart.
12752 (let (multiple-lines)
12753 (while (and (progn (beginning-of-line) (not (bobp)))
12754 (progn (move-to-left-margin)
12755 (not (looking-at parsep)))
12756 (looking-at fill-prefix-regexp))
12757 (unless (= (point) start)
12758 (setq multiple-lines t))
12760 (move-to-left-margin)
12761 ;; This deleted code caused a long hanging-indent line
12762 ;; not to be filled together with the following lines.
12763 ;; ;; Don't move back over a line before the paragraph
12764 ;; ;; which doesn't start with fill-prefix
12765 ;; ;; unless that is the only line we've moved over.
12766 ;; (and (not (looking-at fill-prefix-regexp))
12768 ;; (forward-line 1))
12770 (while (and (re-search-backward sp-parstart nil 1)
12771 (setq found-start t)
12772 ;; Found a candidate, but need to check if it is a
12774 (progn (setq start (point))
12775 (move-to-left-margin)
12776 (not (looking-at parsep)))
12777 (not (and (looking-at parstart)
12778 (or (not use-hard-newlines)
12781 (1- start) 'hard)))))
12782 (setq found-start nil)
12787 ;; Move forward over paragraph separators.
12788 ;; We know this cannot reach the place we started
12789 ;; because we know we moved back over a non-separator.
12790 (while (and (not (eobp))
12791 (progn (move-to-left-margin)
12792 (looking-at parsep)))
12794 ;; If line before paragraph is just margin, back up to there.
12796 (if (> (current-column) (current-left-margin))
12798 (skip-chars-backward " \t")
12800 (forward-line 1))))
12801 ;; No starter or separator line => use buffer beg.
12802 (goto-char (point-min))))))
12804 (while (and (> arg 0) (not (eobp)))
12805 ;; Move forward over separator lines...
12806 (while (and (not (eobp))
12807 (progn (move-to-left-margin) (not (eobp)))
12808 (looking-at parsep))
12810 (unless (eobp) (setq arg (1- arg)))
12811 ;; ... and one more line.
12813 (if fill-prefix-regexp
12814 ;; There is a fill prefix; it overrides parstart.
12815 (while (and (not (eobp))
12816 (progn (move-to-left-margin) (not (eobp)))
12817 (not (looking-at parsep))
12818 (looking-at fill-prefix-regexp))
12820 (while (and (re-search-forward sp-parstart nil 1)
12821 (progn (setq start (match-beginning 0))
12824 (progn (move-to-left-margin)
12825 (not (looking-at parsep)))
12826 (or (not (looking-at parstart))
12827 (and use-hard-newlines
12828 (not (get-text-property (1- start) 'hard)))))
12830 (if (< (point) (point-max))
12831 (goto-char start))))
12832 (constrain-to-field nil opoint t)
12833 ;; Return the number of steps that could not be done.
12837 The @code{forward-paragraph} function moves point forward to the end
12838 of the paragraph. It is usually bound to @kbd{M-@}} and makes use of a
12839 number of functions that are important in themselves, including
12840 @code{let*}, @code{match-beginning}, and @code{looking-at}.
12842 The function definition for @code{forward-paragraph} is considerably
12843 longer than the function definition for @code{forward-sentence}
12844 because it works with a paragraph, each line of which may begin with a
12847 A fill prefix consists of a string of characters that are repeated at
12848 the beginning of each line. For example, in Lisp code, it is a
12849 convention to start each line of a paragraph-long comment with
12850 @samp{;;; }. In Text mode, four blank spaces make up another common
12851 fill prefix, creating an indented paragraph. (@xref{Fill Prefix, , ,
12852 emacs, The GNU Emacs Manual}, for more information about fill
12855 The existence of a fill prefix means that in addition to being able to
12856 find the end of a paragraph whose lines begin on the left-most
12857 column, the @code{forward-paragraph} function must be able to find the
12858 end of a paragraph when all or many of the lines in the buffer begin
12859 with the fill prefix.
12861 Moreover, it is sometimes practical to ignore a fill prefix that
12862 exists, especially when blank lines separate paragraphs.
12863 This is an added complication.
12866 * forward-paragraph in brief:: Key parts of the function definition.
12867 * fwd-para let:: The @code{let*} expression.
12868 * fwd-para while:: The forward motion @code{while} loop.
12872 @node forward-paragraph in brief
12873 @unnumberedsubsec Shortened @code{forward-paragraph} function definition
12876 Rather than print all of the @code{forward-paragraph} function, we
12877 will only print parts of it. Read without preparation, the function
12881 In outline, the function looks like this:
12885 (defun forward-paragraph (&optional arg)
12886 "@var{documentation}@dots{}"
12888 (or arg (setq arg 1))
12891 (while (and (< arg 0) (not (bobp))) ; @r{backward-moving-code}
12893 (while (and (> arg 0) (not (eobp))) ; @r{forward-moving-code}
12898 The first parts of the function are routine: the function's argument
12899 list consists of one optional argument. Documentation follows.
12901 The lower case @samp{p} in the @code{interactive} declaration means
12902 that the processed prefix argument, if any, is passed to the function.
12903 This will be a number, and is the repeat count of how many paragraphs
12904 point will move. The @code{or} expression in the next line handles
12905 the common case when no argument is passed to the function, which occurs
12906 if the function is called from other code rather than interactively.
12907 This case was described earlier. (@xref{forward-sentence, The
12908 @code{forward-sentence} function}.) Now we reach the end of the
12909 familiar part of this function.
12912 @unnumberedsubsec The @code{let*} expression
12914 The next line of the @code{forward-paragraph} function begins a
12915 @code{let*} expression. This is a different than @code{let}. The
12916 symbol is @code{let*} not @code{let}.
12918 The @code{let*} special form is like @code{let} except that Emacs sets
12919 each variable in sequence, one after another, and variables in the
12920 latter part of the varlist can make use of the values to which Emacs
12921 set variables in the earlier part of the varlist.
12924 ( refappend save-excursion, , code save-excursion in code append-to-buffer .)
12927 (@ref{append save-excursion, , @code{save-excursion} in @code{append-to-buffer}}.)
12929 In the @code{let*} expression in this function, Emacs binds a total of
12930 seven variables: @code{opoint}, @code{fill-prefix-regexp},
12931 @code{parstart}, @code{parsep}, @code{sp-parstart}, @code{start}, and
12932 @code{found-start}.
12934 The variable @code{parsep} appears twice, first, to remove instances
12935 of @samp{^}, and second, to handle fill prefixes.
12937 The variable @code{opoint} is just the value of @code{point}. As you
12938 can guess, it is used in a @code{constrain-to-field} expression, just
12939 as in @code{forward-sentence}.
12941 The variable @code{fill-prefix-regexp} is set to the value returned by
12942 evaluating the following list:
12947 (not (equal fill-prefix ""))
12948 (not paragraph-ignore-fill-prefix)
12949 (regexp-quote fill-prefix))
12954 This is an expression whose first element is the @code{and} special form.
12956 As we learned earlier (@pxref{kill-new function, , The @code{kill-new}
12957 function}), the @code{and} special form evaluates each of its
12958 arguments until one of the arguments returns a value of @code{nil}, in
12959 which case the @code{and} expression returns @code{nil}; however, if
12960 none of the arguments returns a value of @code{nil}, the value
12961 resulting from evaluating the last argument is returned. (Since such
12962 a value is not @code{nil}, it is considered true in Lisp.) In other
12963 words, an @code{and} expression returns a true value only if all its
12964 arguments are true.
12967 In this case, the variable @code{fill-prefix-regexp} is bound to a
12968 non-@code{nil} value only if the following four expressions produce a
12969 true (i.e., a non-@code{nil}) value when they are evaluated; otherwise,
12970 @code{fill-prefix-regexp} is bound to @code{nil}.
12974 When this variable is evaluated, the value of the fill prefix, if any,
12975 is returned. If there is no fill prefix, this variable returns
12978 @item (not (equal fill-prefix "")
12979 This expression checks whether an existing fill prefix is an empty
12980 string, that is, a string with no characters in it. An empty string is
12981 not a useful fill prefix.
12983 @item (not paragraph-ignore-fill-prefix)
12984 This expression returns @code{nil} if the variable
12985 @code{paragraph-ignore-fill-prefix} has been turned on by being set to a
12986 true value such as @code{t}.
12988 @item (regexp-quote fill-prefix)
12989 This is the last argument to the @code{and} special form. If all the
12990 arguments to the @code{and} are true, the value resulting from
12991 evaluating this expression will be returned by the @code{and} expression
12992 and bound to the variable @code{fill-prefix-regexp},
12995 @findex regexp-quote
12997 The result of evaluating this @code{and} expression successfully is that
12998 @code{fill-prefix-regexp} will be bound to the value of
12999 @code{fill-prefix} as modified by the @code{regexp-quote} function.
13000 What @code{regexp-quote} does is read a string and return a regular
13001 expression that will exactly match the string and match nothing else.
13002 This means that @code{fill-prefix-regexp} will be set to a value that
13003 will exactly match the fill prefix if the fill prefix exists.
13004 Otherwise, the variable will be set to @code{nil}.
13006 The next two local variables in the @code{let*} expression are
13007 designed to remove instances of @samp{^} from @code{parstart} and
13008 @code{parsep}, the local variables which indicate the paragraph start
13009 and the paragraph separator. The next expression sets @code{parsep}
13010 again. That is to handle fill prefixes.
13012 This is the setting that requires the definition call @code{let*}
13013 rather than @code{let}. The true-or-false-test for the @code{if}
13014 depends on whether the variable @code{fill-prefix-regexp} evaluates to
13015 @code{nil} or some other value.
13017 If @code{fill-prefix-regexp} does not have a value, Emacs evaluates
13018 the else-part of the @code{if} expression and binds @code{parsep} to
13019 its local value. (@code{parsep} is a regular expression that matches
13020 what separates paragraphs.)
13022 But if @code{fill-prefix-regexp} does have a value, Emacs evaluates
13023 the then-part of the @code{if} expression and binds @code{parsep} to a
13024 regular expression that includes the @code{fill-prefix-regexp} as part
13027 Specifically, @code{parsep} is set to the original value of the
13028 paragraph separate regular expression concatenated with an alternative
13029 expression that consists of the @code{fill-prefix-regexp} followed by
13030 optional whitespace to the end of the line. The whitespace is defined
13031 by @w{@code{"[ \t]*$"}}.) The @samp{\\|} defines this portion of the
13032 regexp as an alternative to @code{parsep}.
13034 According to a comment in the code, the next local variable,
13035 @code{sp-parstart}, is used for searching, and then the final two,
13036 @code{start} and @code{found-start}, are set to @code{nil}.
13038 Now we get into the body of the @code{let*}. The first part of the body
13039 of the @code{let*} deals with the case when the function is given a
13040 negative argument and is therefore moving backwards. We will skip this
13043 @node fwd-para while
13044 @unnumberedsubsec The forward motion @code{while} loop
13046 The second part of the body of the @code{let*} deals with forward
13047 motion. It is a @code{while} loop that repeats itself so long as the
13048 value of @code{arg} is greater than zero. In the most common use of
13049 the function, the value of the argument is 1, so the body of the
13050 @code{while} loop is evaluated exactly once, and the cursor moves
13051 forward one paragraph.
13054 (while (and (> arg 0) (not (eobp)))
13056 ;; Move forward over separator lines...
13057 (while (and (not (eobp))
13058 (progn (move-to-left-margin) (not (eobp)))
13059 (looking-at parsep))
13061 (unless (eobp) (setq arg (1- arg)))
13062 ;; ... and one more line.
13065 (if fill-prefix-regexp
13066 ;; There is a fill prefix; it overrides parstart.
13067 (while (and (not (eobp))
13068 (progn (move-to-left-margin) (not (eobp)))
13069 (not (looking-at parsep))
13070 (looking-at fill-prefix-regexp))
13073 (while (and (re-search-forward sp-parstart nil 1)
13074 (progn (setq start (match-beginning 0))
13077 (progn (move-to-left-margin)
13078 (not (looking-at parsep)))
13079 (or (not (looking-at parstart))
13080 (and use-hard-newlines
13081 (not (get-text-property (1- start) 'hard)))))
13084 (if (< (point) (point-max))
13085 (goto-char start))))
13088 This part handles three situations: when point is between paragraphs,
13089 when there is a fill prefix and when there is no fill prefix.
13092 The @code{while} loop looks like this:
13096 ;; @r{going forwards and not at the end of the buffer}
13097 (while (and (> arg 0) (not (eobp)))
13099 ;; @r{between paragraphs}
13100 ;; Move forward over separator lines...
13101 (while (and (not (eobp))
13102 (progn (move-to-left-margin) (not (eobp)))
13103 (looking-at parsep))
13105 ;; @r{This decrements the loop}
13106 (unless (eobp) (setq arg (1- arg)))
13107 ;; ... and one more line.
13112 (if fill-prefix-regexp
13113 ;; There is a fill prefix; it overrides parstart;
13114 ;; we go forward line by line
13115 (while (and (not (eobp))
13116 (progn (move-to-left-margin) (not (eobp)))
13117 (not (looking-at parsep))
13118 (looking-at fill-prefix-regexp))
13123 ;; There is no fill prefix;
13124 ;; we go forward character by character
13125 (while (and (re-search-forward sp-parstart nil 1)
13126 (progn (setq start (match-beginning 0))
13129 (progn (move-to-left-margin)
13130 (not (looking-at parsep)))
13131 (or (not (looking-at parstart))
13132 (and use-hard-newlines
13133 (not (get-text-property (1- start) 'hard)))))
13138 ;; and if there is no fill prefix and if we are not at the end,
13139 ;; go to whatever was found in the regular expression search
13141 (if (< (point) (point-max))
13142 (goto-char start))))
13147 We can see that this is a decrementing counter @code{while} loop,
13148 using the expression @code{(setq arg (1- arg))} as the decrementer.
13149 That expression is not far from the @code{while}, but is hidden in
13150 another Lisp macro, an @code{unless} macro. Unless we are at the end
13151 of the buffer---that is what the @code{eobp} function determines; it
13152 is an abbreviation of @samp{End Of Buffer P}---we decrease the value
13153 of @code{arg} by one.
13155 (If we are at the end of the buffer, we cannot go forward any more and
13156 the next loop of the @code{while} expression will test false since the
13157 test is an @code{and} with @code{(not (eobp))}. The @code{not}
13158 function means exactly as you expect; it is another name for
13159 @code{null}, a function that returns true when its argument is false.)
13161 Interestingly, the loop count is not decremented until we leave the
13162 space between paragraphs, unless we come to the end of buffer or stop
13163 seeing the local value of the paragraph separator.
13165 That second @code{while} also has a @code{(move-to-left-margin)}
13166 expression. The function is self-explanatory. It is inside a
13167 @code{progn} expression and not the last element of its body, so it is
13168 only invoked for its side effect, which is to move point to the left
13169 margin of the current line.
13172 The @code{looking-at} function is also self-explanatory; it returns
13173 true if the text after point matches the regular expression given as
13176 The rest of the body of the loop looks difficult at first, but makes
13177 sense as you come to understand it.
13180 First consider what happens if there is a fill prefix:
13184 (if fill-prefix-regexp
13185 ;; There is a fill prefix; it overrides parstart;
13186 ;; we go forward line by line
13187 (while (and (not (eobp))
13188 (progn (move-to-left-margin) (not (eobp)))
13189 (not (looking-at parsep))
13190 (looking-at fill-prefix-regexp))
13196 This expression moves point forward line by line so long
13197 as four conditions are true:
13201 Point is not at the end of the buffer.
13204 We can move to the left margin of the text and are
13205 not at the end of the buffer.
13208 The text following point does not separate paragraphs.
13211 The pattern following point is the fill prefix regular expression.
13214 The last condition may be puzzling, until you remember that point was
13215 moved to the beginning of the line early in the @code{forward-paragraph}
13216 function. This means that if the text has a fill prefix, the
13217 @code{looking-at} function will see it.
13220 Consider what happens when there is no fill prefix.
13224 (while (and (re-search-forward sp-parstart nil 1)
13225 (progn (setq start (match-beginning 0))
13228 (progn (move-to-left-margin)
13229 (not (looking-at parsep)))
13230 (or (not (looking-at parstart))
13231 (and use-hard-newlines
13232 (not (get-text-property (1- start) 'hard)))))
13238 This @code{while} loop has us searching forward for
13239 @code{sp-parstart}, which is the combination of possible whitespace
13240 with a the local value of the start of a paragraph or of a paragraph
13241 separator. (The latter two are within an expression starting
13242 @code{\(?:} so that they are not referenced by the
13243 @code{match-beginning} function.)
13246 The two expressions,
13250 (setq start (match-beginning 0))
13256 mean go to the start of the text matched by the regular expression
13259 The @code{(match-beginning 0)} expression is new. It returns a number
13260 specifying the location of the start of the text that was matched by
13263 The @code{match-beginning} function is used here because of a
13264 characteristic of a forward search: a successful forward search,
13265 regardless of whether it is a plain search or a regular expression
13266 search, moves point to the end of the text that is found. In this
13267 case, a successful search moves point to the end of the pattern for
13268 @code{sp-parstart}.
13270 However, we want to put point at the end of the current paragraph, not
13271 somewhere else. Indeed, since the search possibly includes the
13272 paragraph separator, point may end up at the beginning of the next one
13273 unless we use an expression that includes @code{match-beginning}.
13275 @findex match-beginning
13276 When given an argument of 0, @code{match-beginning} returns the
13277 position that is the start of the text matched by the most recent
13278 search. In this case, the most recent search looks for
13279 @code{sp-parstart}. The @code{(match-beginning 0)} expression returns
13280 the beginning position of that pattern, rather than the end position
13283 (Incidentally, when passed a positive number as an argument, the
13284 @code{match-beginning} function returns the location of point at that
13285 parenthesized expression in the last search unless that parenthesized
13286 expression begins with @code{\(?:}. I don't know why @code{\(?:}
13287 appears here since the argument is 0.)
13290 The last expression when there is no fill prefix is
13294 (if (< (point) (point-max))
13295 (goto-char start))))
13300 This says that if there is no fill prefix and if we are not at the
13301 end, point should move to the beginning of whatever was found by the
13302 regular expression search for @code{sp-parstart}.
13304 The full definition for the @code{forward-paragraph} function not only
13305 includes code for going forwards, but also code for going backwards.
13307 If you are reading this inside of GNU Emacs and you want to see the
13308 whole function, you can type @kbd{C-h f} (@code{describe-function})
13309 and the name of the function. This gives you the function
13310 documentation and the name of the library containing the function's
13311 source. Place point over the name of the library and press the RET
13312 key; you will be taken directly to the source. (Be sure to install
13313 your sources! Without them, you are like a person who tries to drive
13314 a car with his eyes shut!)
13317 @section Create Your Own @file{TAGS} File
13319 @cindex @file{TAGS} file, create own
13321 Besides @kbd{C-h f} (@code{describe-function}), another way to see the
13322 source of a function is to type @kbd{M-.} (@code{find-tag}) and the
13323 name of the function when prompted for it. This is a good habit to
13324 get into. The @kbd{M-.} (@code{find-tag}) command takes you directly
13325 to the source for a function, variable, or node. The function depends
13326 on tags tables to tell it where to go.
13328 If the @code{find-tag} function first asks you for the name of a
13329 @file{TAGS} table, give it the name of a @file{TAGS} file such as
13330 @file{/usr/local/src/emacs/src/TAGS}. (The exact path to your
13331 @file{TAGS} file depends on how your copy of Emacs was installed. I
13332 just told you the location that provides both my C and my Emacs Lisp
13335 You can also create your own @file{TAGS} file for directories that
13338 You often need to build and install tags tables yourself. They are
13339 not built automatically. A tags table is called a @file{TAGS} file;
13340 the name is in upper case letters.
13342 You can create a @file{TAGS} file by calling the @code{etags} program
13343 that comes as a part of the Emacs distribution. Usually, @code{etags}
13344 is compiled and installed when Emacs is built. (@code{etags} is not
13345 an Emacs Lisp function or a part of Emacs; it is a C program.)
13348 To create a @file{TAGS} file, first switch to the directory in which
13349 you want to create the file. In Emacs you can do this with the
13350 @kbd{M-x cd} command, or by visiting a file in the directory, or by
13351 listing the directory with @kbd{C-x d} (@code{dired}). Then run the
13352 compile command, with @w{@code{etags *.el}} as the command to execute
13355 M-x compile RET etags *.el RET
13359 to create a @file{TAGS} file for Emacs Lisp.
13361 For example, if you have a large number of files in your
13362 @file{~/emacs} directory, as I do---I have 137 @file{.el} files in it,
13363 of which I load 12---you can create a @file{TAGS} file for the Emacs
13364 Lisp files in that directory.
13367 The @code{etags} program takes all the usual shell `wildcards'. For
13368 example, if you have two directories for which you want a single
13369 @file{TAGS} file, type @w{@code{etags *.el ../elisp/*.el}}, where
13370 @file{../elisp/} is the second directory:
13373 M-x compile RET etags *.el ../elisp/*.el RET
13380 M-x compile RET etags --help RET
13384 to see a list of the options accepted by @code{etags} as well as a
13385 list of supported languages.
13387 The @code{etags} program handles more than 20 languages, including
13388 Emacs Lisp, Common Lisp, Scheme, C, C++, Ada, Fortran, HTML, Java,
13389 LaTeX, Pascal, Perl, PostScript, Python, TeX, Texinfo, makefiles, and
13390 most assemblers. The program has no switches for specifying the
13391 language; it recognizes the language in an input file according to its
13392 file name and contents.
13394 @file{etags} is very helpful when you are writing code yourself and
13395 want to refer back to functions you have already written. Just run
13396 @code{etags} again at intervals as you write new functions, so they
13397 become part of the @file{TAGS} file.
13399 If you think an appropriate @file{TAGS} file already exists for what
13400 you want, but do not know where it is, you can use the @code{locate}
13401 program to attempt to find it.
13403 Type @w{@kbd{M-x locate @key{RET} TAGS @key{RET}}} and Emacs will list
13404 for you the full path names of all your @file{TAGS} files. On my
13405 system, this command lists 34 @file{TAGS} files. On the other hand, a
13406 `plain vanilla' system I recently installed did not contain any
13409 If the tags table you want has been created, you can use the @code{M-x
13410 visit-tags-table} command to specify it. Otherwise, you will need to
13411 create the tag table yourself and then use @code{M-x
13414 @subsubheading Building Tags in the Emacs sources
13415 @cindex Building Tags in the Emacs sources
13416 @cindex Tags in the Emacs sources
13419 The GNU Emacs sources come with a @file{Makefile} that contains a
13420 sophisticated @code{etags} command that creates, collects, and merges
13421 tags tables from all over the Emacs sources and puts the information
13422 into one @file{TAGS} file in the @file{src/} directory. (The
13423 @file{src/} directory is below the top level of your Emacs directory.)
13426 To build this @file{TAGS} file, go to the top level of your Emacs
13427 source directory and run the compile command @code{make tags}:
13430 M-x compile RET make tags RET
13434 (The @code{make tags} command works well with the GNU Emacs sources,
13435 as well as with some other source packages.)
13437 For more information, see @ref{Tags, , Tag Tables, emacs, The GNU Emacs
13440 @node Regexp Review
13443 Here is a brief summary of some recently introduced functions.
13447 Repeatedly evaluate the body of the expression so long as the first
13448 element of the body tests true. Then return @code{nil}. (The
13449 expression is evaluated only for its side effects.)
13458 (insert (format "foo is %d.\n" foo))
13459 (setq foo (1- foo))))
13461 @result{} foo is 2.
13468 (The @code{insert} function inserts its arguments at point; the
13469 @code{format} function returns a string formatted from its arguments
13470 the way @code{message} formats its arguments; @code{\n} produces a new
13473 @item re-search-forward
13474 Search for a pattern, and if the pattern is found, move point to rest
13478 Takes four arguments, like @code{search-forward}:
13482 A regular expression that specifies the pattern to search for.
13483 (Remember to put quotation marks around this argument!)
13486 Optionally, the limit of the search.
13489 Optionally, what to do if the search fails, return @code{nil} or an
13493 Optionally, how many times to repeat the search; if negative, the
13494 search goes backwards.
13498 Bind some variables locally to particular values,
13499 and then evaluate the remaining arguments, returning the value of the
13500 last one. While binding the local variables, use the local values of
13501 variables bound earlier, if any.
13510 (message "`bar' is %d." bar))
13511 @result{} `bar' is 21.
13515 @item match-beginning
13516 Return the position of the start of the text found by the last regular
13520 Return @code{t} for true if the text after point matches the argument,
13521 which should be a regular expression.
13524 Return @code{t} for true if point is at the end of the accessible part
13525 of a buffer. The end of the accessible part is the end of the buffer
13526 if the buffer is not narrowed; it is the end of the narrowed part if
13527 the buffer is narrowed.
13531 @node re-search Exercises
13532 @section Exercises with @code{re-search-forward}
13536 Write a function to search for a regular expression that matches two
13537 or more blank lines in sequence.
13540 Write a function to search for duplicated words, such as `the the'.
13541 @xref{Regexps, , Syntax of Regular Expressions, emacs, The GNU Emacs
13542 Manual}, for information on how to write a regexp (a regular
13543 expression) to match a string that is composed of two identical
13544 halves. You can devise several regexps; some are better than others.
13545 The function I use is described in an appendix, along with several
13546 regexps. @xref{the-the, , @code{the-the} Duplicated Words Function}.
13549 @node Counting Words
13550 @chapter Counting via Repetition and Regexps
13551 @cindex Repetition for word counting
13552 @cindex Regular expressions for word counting
13554 Repetition and regular expression searches are powerful tools that you
13555 often use when you write code in Emacs Lisp. This chapter illustrates
13556 the use of regular expression searches through the construction of
13557 word count commands using @code{while} loops and recursion.
13560 * Why Count Words::
13561 * @value{COUNT-WORDS}:: Use a regexp, but find a problem.
13562 * recursive-count-words:: Start with case of no words in region.
13563 * Counting Exercise::
13567 @node Why Count Words
13568 @unnumberedsec Counting words
13571 The standard Emacs distribution contains functions for counting the
13572 number of lines and words within a region.
13574 Certain types of writing ask you to count words. Thus, if you write
13575 an essay, you may be limited to 800 words; if you write a novel, you
13576 may discipline yourself to write 1000 words a day. It seems odd, but
13577 for a long time, Emacs lacked a word count command. Perhaps people used
13578 Emacs mostly for code or types of documentation that did not require
13579 word counts; or perhaps they restricted themselves to the operating
13580 system word count command, @code{wc}. Alternatively, people may have
13581 followed the publishers' convention and computed a word count by
13582 dividing the number of characters in a document by five.
13584 There are many ways to implement a command to count words. Here are
13585 some examples, which you may wish to compare with the standard Emacs
13586 command, @code{count-words-region}.
13588 @node @value{COUNT-WORDS}
13589 @section The @code{@value{COUNT-WORDS}} Function
13590 @findex @value{COUNT-WORDS}
13592 A word count command could count words in a line, paragraph, region,
13593 or buffer. What should the command cover? You could design the
13594 command to count the number of words in a complete buffer. However,
13595 the Emacs tradition encourages flexibility---you may want to count
13596 words in just a section, rather than all of a buffer. So it makes
13597 more sense to design the command to count the number of words in a
13598 region. Once you have a command to count words in a region, you can,
13599 if you wish, count words in a whole buffer by marking it with
13600 @w{@kbd{C-x h}} (@code{mark-whole-buffer}).
13602 Clearly, counting words is a repetitive act: starting from the
13603 beginning of the region, you count the first word, then the second
13604 word, then the third word, and so on, until you reach the end of the
13605 region. This means that word counting is ideally suited to recursion
13606 or to a @code{while} loop.
13609 * Design @value{COUNT-WORDS}:: The definition using a @code{while} loop.
13610 * Whitespace Bug:: The Whitespace Bug in @code{@value{COUNT-WORDS}}.
13614 @node Design @value{COUNT-WORDS}
13615 @unnumberedsubsec Designing @code{@value{COUNT-WORDS}}
13618 First, we will implement the word count command with a @code{while}
13619 loop, then with recursion. The command will, of course, be
13623 The template for an interactive function definition is, as always:
13627 (defun @var{name-of-function} (@var{argument-list})
13628 "@var{documentation}@dots{}"
13629 (@var{interactive-expression}@dots{})
13634 What we need to do is fill in the slots.
13636 The name of the function should be self-explanatory and similar to the
13637 existing @code{count-lines-region} name. This makes the name easier
13638 to remember. @code{count-words-region} is the obvious choice. Since
13639 that name is now used for the standard Emacs command to count words, we
13640 will name our implementation @code{@value{COUNT-WORDS}}.
13642 The function counts words within a region. This means that the
13643 argument list must contain symbols that are bound to the two
13644 positions, the beginning and end of the region. These two positions
13645 can be called @samp{beginning} and @samp{end} respectively. The first
13646 line of the documentation should be a single sentence, since that is
13647 all that is printed as documentation by a command such as
13648 @code{apropos}. The interactive expression will be of the form
13649 @samp{(interactive "r")}, since that will cause Emacs to pass the
13650 beginning and end of the region to the function's argument list. All
13653 The body of the function needs to be written to do three tasks:
13654 first, to set up conditions under which the @code{while} loop can
13655 count words, second, to run the @code{while} loop, and third, to send
13656 a message to the user.
13658 When a user calls @code{@value{COUNT-WORDS}}, point may be at the
13659 beginning or the end of the region. However, the counting process
13660 must start at the beginning of the region. This means we will want
13661 to put point there if it is not already there. Executing
13662 @code{(goto-char beginning)} ensures this. Of course, we will want to
13663 return point to its expected position when the function finishes its
13664 work. For this reason, the body must be enclosed in a
13665 @code{save-excursion} expression.
13667 The central part of the body of the function consists of a
13668 @code{while} loop in which one expression jumps point forward word by
13669 word, and another expression counts those jumps. The true-or-false-test
13670 of the @code{while} loop should test true so long as point should jump
13671 forward, and false when point is at the end of the region.
13673 We could use @code{(forward-word 1)} as the expression for moving point
13674 forward word by word, but it is easier to see what Emacs identifies as a
13675 `word' if we use a regular expression search.
13677 A regular expression search that finds the pattern for which it is
13678 searching leaves point after the last character matched. This means
13679 that a succession of successful word searches will move point forward
13682 As a practical matter, we want the regular expression search to jump
13683 over whitespace and punctuation between words as well as over the
13684 words themselves. A regexp that refuses to jump over interword
13685 whitespace would never jump more than one word! This means that
13686 the regexp should include the whitespace and punctuation that follows
13687 a word, if any, as well as the word itself. (A word may end a buffer
13688 and not have any following whitespace or punctuation, so that part of
13689 the regexp must be optional.)
13691 Thus, what we want for the regexp is a pattern defining one or more
13692 word constituent characters followed, optionally, by one or more
13693 characters that are not word constituents. The regular expression for
13701 The buffer's syntax table determines which characters are and are not
13702 word constituents. For more information about syntax,
13703 @pxref{Syntax Tables, , Syntax Tables, elisp, The GNU Emacs Lisp
13707 The search expression looks like this:
13710 (re-search-forward "\\w+\\W*")
13714 (Note that paired backslashes precede the @samp{w} and @samp{W}. A
13715 single backslash has special meaning to the Emacs Lisp interpreter.
13716 It indicates that the following character is interpreted differently
13717 than usual. For example, the two characters, @samp{\n}, stand for
13718 @samp{newline}, rather than for a backslash followed by @samp{n}. Two
13719 backslashes in a row stand for an ordinary, `unspecial' backslash, so
13720 Emacs Lisp interpreter ends of seeing a single backslash followed by a
13721 letter. So it discovers the letter is special.)
13723 We need a counter to count how many words there are; this variable
13724 must first be set to 0 and then incremented each time Emacs goes
13725 around the @code{while} loop. The incrementing expression is simply:
13728 (setq count (1+ count))
13731 Finally, we want to tell the user how many words there are in the
13732 region. The @code{message} function is intended for presenting this
13733 kind of information to the user. The message has to be phrased so
13734 that it reads properly regardless of how many words there are in the
13735 region: we don't want to say that ``there are 1 words in the region''.
13736 The conflict between singular and plural is ungrammatical. We can
13737 solve this problem by using a conditional expression that evaluates
13738 different messages depending on the number of words in the region.
13739 There are three possibilities: no words in the region, one word in the
13740 region, and more than one word. This means that the @code{cond}
13741 special form is appropriate.
13744 All this leads to the following function definition:
13748 ;;; @r{First version; has bugs!}
13749 (defun @value{COUNT-WORDS} (beginning end)
13750 "Print number of words in the region.
13751 Words are defined as at least one word-constituent
13752 character followed by at least one character that
13753 is not a word-constituent. The buffer's syntax
13754 table determines which characters these are."
13756 (message "Counting words in region ... ")
13760 ;;; @r{1. Set up appropriate conditions.}
13762 (goto-char beginning)
13767 ;;; @r{2. Run the} while @r{loop.}
13768 (while (< (point) end)
13769 (re-search-forward "\\w+\\W*")
13770 (setq count (1+ count)))
13774 ;;; @r{3. Send a message to the user.}
13775 (cond ((zerop count)
13777 "The region does NOT have any words."))
13780 "The region has 1 word."))
13783 "The region has %d words." count))))))
13788 As written, the function works, but not in all circumstances.
13790 @node Whitespace Bug
13791 @subsection The Whitespace Bug in @code{@value{COUNT-WORDS}}
13793 The @code{@value{COUNT-WORDS}} command described in the preceding
13794 section has two bugs, or rather, one bug with two manifestations.
13795 First, if you mark a region containing only whitespace in the middle
13796 of some text, the @code{@value{COUNT-WORDS}} command tells you that the
13797 region contains one word! Second, if you mark a region containing
13798 only whitespace at the end of the buffer or the accessible portion of
13799 a narrowed buffer, the command displays an error message that looks
13803 Search failed: "\\w+\\W*"
13806 If you are reading this in Info in GNU Emacs, you can test for these
13809 First, evaluate the function in the usual manner to install it.
13811 Here is a copy of the definition. Place your cursor after the closing
13812 parenthesis and type @kbd{C-x C-e} to install it.
13816 ;; @r{First version; has bugs!}
13817 (defun @value{COUNT-WORDS} (beginning end)
13818 "Print number of words in the region.
13819 Words are defined as at least one word-constituent character followed
13820 by at least one character that is not a word-constituent. The buffer's
13821 syntax table determines which characters these are."
13825 (message "Counting words in region ... ")
13829 ;;; @r{1. Set up appropriate conditions.}
13831 (goto-char beginning)
13836 ;;; @r{2. Run the} while @r{loop.}
13837 (while (< (point) end)
13838 (re-search-forward "\\w+\\W*")
13839 (setq count (1+ count)))
13843 ;;; @r{3. Send a message to the user.}
13844 (cond ((zerop count)
13845 (message "The region does NOT have any words."))
13846 ((= 1 count) (message "The region has 1 word."))
13847 (t (message "The region has %d words." count))))))
13853 If you wish, you can also install this keybinding by evaluating it:
13856 (global-set-key "\C-c=" '@value{COUNT-WORDS})
13859 To conduct the first test, set mark and point to the beginning and end
13860 of the following line and then type @kbd{C-c =} (or @kbd{M-x
13861 @value{COUNT-WORDS}} if you have not bound @kbd{C-c =}):
13868 Emacs will tell you, correctly, that the region has three words.
13870 Repeat the test, but place mark at the beginning of the line and place
13871 point just @emph{before} the word @samp{one}. Again type the command
13872 @kbd{C-c =} (or @kbd{M-x @value{COUNT-WORDS}}). Emacs should tell you
13873 that the region has no words, since it is composed only of the
13874 whitespace at the beginning of the line. But instead Emacs tells you
13875 that the region has one word!
13877 For the third test, copy the sample line to the end of the
13878 @file{*scratch*} buffer and then type several spaces at the end of the
13879 line. Place mark right after the word @samp{three} and point at the
13880 end of line. (The end of the line will be the end of the buffer.)
13881 Type @kbd{C-c =} (or @kbd{M-x @value{COUNT-WORDS}}) as you did before.
13882 Again, Emacs should tell you that the region has no words, since it is
13883 composed only of the whitespace at the end of the line. Instead,
13884 Emacs displays an error message saying @samp{Search failed}.
13886 The two bugs stem from the same problem.
13888 Consider the first manifestation of the bug, in which the command
13889 tells you that the whitespace at the beginning of the line contains
13890 one word. What happens is this: The @code{M-x @value{COUNT-WORDS}}
13891 command moves point to the beginning of the region. The @code{while}
13892 tests whether the value of point is smaller than the value of
13893 @code{end}, which it is. Consequently, the regular expression search
13894 looks for and finds the first word. It leaves point after the word.
13895 @code{count} is set to one. The @code{while} loop repeats; but this
13896 time the value of point is larger than the value of @code{end}, the
13897 loop is exited; and the function displays a message saying the number
13898 of words in the region is one. In brief, the regular expression
13899 search looks for and finds the word even though it is outside
13902 In the second manifestation of the bug, the region is whitespace at
13903 the end of the buffer. Emacs says @samp{Search failed}. What happens
13904 is that the true-or-false-test in the @code{while} loop tests true, so
13905 the search expression is executed. But since there are no more words
13906 in the buffer, the search fails.
13908 In both manifestations of the bug, the search extends or attempts to
13909 extend outside of the region.
13911 The solution is to limit the search to the region---this is a fairly
13912 simple action, but as you may have come to expect, it is not quite as
13913 simple as you might think.
13915 As we have seen, the @code{re-search-forward} function takes a search
13916 pattern as its first argument. But in addition to this first,
13917 mandatory argument, it accepts three optional arguments. The optional
13918 second argument bounds the search. The optional third argument, if
13919 @code{t}, causes the function to return @code{nil} rather than signal
13920 an error if the search fails. The optional fourth argument is a
13921 repeat count. (In Emacs, you can see a function's documentation by
13922 typing @kbd{C-h f}, the name of the function, and then @key{RET}.)
13924 In the @code{@value{COUNT-WORDS}} definition, the value of the end of
13925 the region is held by the variable @code{end} which is passed as an
13926 argument to the function. Thus, we can add @code{end} as an argument
13927 to the regular expression search expression:
13930 (re-search-forward "\\w+\\W*" end)
13933 However, if you make only this change to the @code{@value{COUNT-WORDS}}
13934 definition and then test the new version of the definition on a
13935 stretch of whitespace, you will receive an error message saying
13936 @samp{Search failed}.
13938 What happens is this: the search is limited to the region, and fails
13939 as you expect because there are no word-constituent characters in the
13940 region. Since it fails, we receive an error message. But we do not
13941 want to receive an error message in this case; we want to receive the
13942 message that "The region does NOT have any words."
13944 The solution to this problem is to provide @code{re-search-forward}
13945 with a third argument of @code{t}, which causes the function to return
13946 @code{nil} rather than signal an error if the search fails.
13948 However, if you make this change and try it, you will see the message
13949 ``Counting words in region ... '' and @dots{} you will keep on seeing
13950 that message @dots{}, until you type @kbd{C-g} (@code{keyboard-quit}).
13952 Here is what happens: the search is limited to the region, as before,
13953 and it fails because there are no word-constituent characters in the
13954 region, as expected. Consequently, the @code{re-search-forward}
13955 expression returns @code{nil}. It does nothing else. In particular,
13956 it does not move point, which it does as a side effect if it finds the
13957 search target. After the @code{re-search-forward} expression returns
13958 @code{nil}, the next expression in the @code{while} loop is evaluated.
13959 This expression increments the count. Then the loop repeats. The
13960 true-or-false-test tests true because the value of point is still less
13961 than the value of end, since the @code{re-search-forward} expression
13962 did not move point. @dots{} and the cycle repeats @dots{}
13964 The @code{@value{COUNT-WORDS}} definition requires yet another
13965 modification, to cause the true-or-false-test of the @code{while} loop
13966 to test false if the search fails. Put another way, there are two
13967 conditions that must be satisfied in the true-or-false-test before the
13968 word count variable is incremented: point must still be within the
13969 region and the search expression must have found a word to count.
13971 Since both the first condition and the second condition must be true
13972 together, the two expressions, the region test and the search
13973 expression, can be joined with an @code{and} special form and embedded in
13974 the @code{while} loop as the true-or-false-test, like this:
13977 (and (< (point) end) (re-search-forward "\\w+\\W*" end t))
13980 @c colon in printed section title causes problem in Info cross reference
13981 @c also trouble with an overfull hbox
13984 (For information about @code{and}, see
13985 @ref{kill-new function, , The @code{kill-new} function}.)
13989 (@xref{kill-new function, , The @code{kill-new} function}, for
13990 information about @code{and}.)
13993 The @code{re-search-forward} expression returns @code{t} if the search
13994 succeeds and as a side effect moves point. Consequently, as words are
13995 found, point is moved through the region. When the search expression
13996 fails to find another word, or when point reaches the end of the
13997 region, the true-or-false-test tests false, the @code{while} loop
13998 exits, and the @code{@value{COUNT-WORDS}} function displays one or
13999 other of its messages.
14001 After incorporating these final changes, the @code{@value{COUNT-WORDS}}
14002 works without bugs (or at least, without bugs that I have found!).
14003 Here is what it looks like:
14007 ;;; @r{Final version:} @code{while}
14008 (defun @value{COUNT-WORDS} (beginning end)
14009 "Print number of words in the region."
14011 (message "Counting words in region ... ")
14015 ;;; @r{1. Set up appropriate conditions.}
14018 (goto-char beginning)
14022 ;;; @r{2. Run the} while @r{loop.}
14023 (while (and (< (point) end)
14024 (re-search-forward "\\w+\\W*" end t))
14025 (setq count (1+ count)))
14029 ;;; @r{3. Send a message to the user.}
14030 (cond ((zerop count)
14032 "The region does NOT have any words."))
14035 "The region has 1 word."))
14038 "The region has %d words." count))))))
14042 @node recursive-count-words
14043 @section Count Words Recursively
14044 @cindex Count words recursively
14045 @cindex Recursively counting words
14046 @cindex Words, counted recursively
14048 You can write the function for counting words recursively as well as
14049 with a @code{while} loop. Let's see how this is done.
14051 First, we need to recognize that the @code{@value{COUNT-WORDS}}
14052 function has three jobs: it sets up the appropriate conditions for
14053 counting to occur; it counts the words in the region; and it sends a
14054 message to the user telling how many words there are.
14056 If we write a single recursive function to do everything, we will
14057 receive a message for every recursive call. If the region contains 13
14058 words, we will receive thirteen messages, one right after the other.
14059 We don't want this! Instead, we must write two functions to do the
14060 job, one of which (the recursive function) will be used inside of the
14061 other. One function will set up the conditions and display the
14062 message; the other will return the word count.
14064 Let us start with the function that causes the message to be displayed.
14065 We can continue to call this @code{@value{COUNT-WORDS}}.
14067 This is the function that the user will call. It will be interactive.
14068 Indeed, it will be similar to our previous versions of this
14069 function, except that it will call @code{recursive-count-words} to
14070 determine how many words are in the region.
14073 We can readily construct a template for this function, based on our
14078 ;; @r{Recursive version; uses regular expression search}
14079 (defun @value{COUNT-WORDS} (beginning end)
14080 "@var{documentation}@dots{}"
14081 (@var{interactive-expression}@dots{})
14085 ;;; @r{1. Set up appropriate conditions.}
14086 (@var{explanatory message})
14087 (@var{set-up functions}@dots{}
14091 ;;; @r{2. Count the words.}
14092 @var{recursive call}
14096 ;;; @r{3. Send a message to the user.}
14097 @var{message providing word count}))
14101 The definition looks straightforward, except that somehow the count
14102 returned by the recursive call must be passed to the message
14103 displaying the word count. A little thought suggests that this can be
14104 done by making use of a @code{let} expression: we can bind a variable
14105 in the varlist of a @code{let} expression to the number of words in
14106 the region, as returned by the recursive call; and then the
14107 @code{cond} expression, using binding, can display the value to the
14110 Often, one thinks of the binding within a @code{let} expression as
14111 somehow secondary to the `primary' work of a function. But in this
14112 case, what you might consider the `primary' job of the function,
14113 counting words, is done within the @code{let} expression.
14116 Using @code{let}, the function definition looks like this:
14120 (defun @value{COUNT-WORDS} (beginning end)
14121 "Print number of words in the region."
14126 ;;; @r{1. Set up appropriate conditions.}
14127 (message "Counting words in region ... ")
14129 (goto-char beginning)
14133 ;;; @r{2. Count the words.}
14134 (let ((count (recursive-count-words end)))
14138 ;;; @r{3. Send a message to the user.}
14139 (cond ((zerop count)
14141 "The region does NOT have any words."))
14144 "The region has 1 word."))
14147 "The region has %d words." count))))))
14151 Next, we need to write the recursive counting function.
14153 A recursive function has at least three parts: the `do-again-test', the
14154 `next-step-expression', and the recursive call.
14156 The do-again-test determines whether the function will or will not be
14157 called again. Since we are counting words in a region and can use a
14158 function that moves point forward for every word, the do-again-test
14159 can check whether point is still within the region. The do-again-test
14160 should find the value of point and determine whether point is before,
14161 at, or after the value of the end of the region. We can use the
14162 @code{point} function to locate point. Clearly, we must pass the
14163 value of the end of the region to the recursive counting function as an
14166 In addition, the do-again-test should also test whether the search finds a
14167 word. If it does not, the function should not call itself again.
14169 The next-step-expression changes a value so that when the recursive
14170 function is supposed to stop calling itself, it stops. More
14171 precisely, the next-step-expression changes a value so that at the
14172 right time, the do-again-test stops the recursive function from
14173 calling itself again. In this case, the next-step-expression can be
14174 the expression that moves point forward, word by word.
14176 The third part of a recursive function is the recursive call.
14178 Somewhere, also, we also need a part that does the `work' of the
14179 function, a part that does the counting. A vital part!
14182 But already, we have an outline of the recursive counting function:
14186 (defun recursive-count-words (region-end)
14187 "@var{documentation}@dots{}"
14188 @var{do-again-test}
14189 @var{next-step-expression}
14190 @var{recursive call})
14194 Now we need to fill in the slots. Let's start with the simplest cases
14195 first: if point is at or beyond the end of the region, there cannot
14196 be any words in the region, so the function should return zero.
14197 Likewise, if the search fails, there are no words to count, so the
14198 function should return zero.
14200 On the other hand, if point is within the region and the search
14201 succeeds, the function should call itself again.
14204 Thus, the do-again-test should look like this:
14208 (and (< (point) region-end)
14209 (re-search-forward "\\w+\\W*" region-end t))
14213 Note that the search expression is part of the do-again-test---the
14214 function returns @code{t} if its search succeeds and @code{nil} if it
14215 fails. (@xref{Whitespace Bug, , The Whitespace Bug in
14216 @code{@value{COUNT-WORDS}}}, for an explanation of how
14217 @code{re-search-forward} works.)
14219 The do-again-test is the true-or-false test of an @code{if} clause.
14220 Clearly, if the do-again-test succeeds, the then-part of the @code{if}
14221 clause should call the function again; but if it fails, the else-part
14222 should return zero since either point is outside the region or the
14223 search failed because there were no words to find.
14225 But before considering the recursive call, we need to consider the
14226 next-step-expression. What is it? Interestingly, it is the search
14227 part of the do-again-test.
14229 In addition to returning @code{t} or @code{nil} for the
14230 do-again-test, @code{re-search-forward} moves point forward as a side
14231 effect of a successful search. This is the action that changes the
14232 value of point so that the recursive function stops calling itself
14233 when point completes its movement through the region. Consequently,
14234 the @code{re-search-forward} expression is the next-step-expression.
14237 In outline, then, the body of the @code{recursive-count-words}
14238 function looks like this:
14242 (if @var{do-again-test-and-next-step-combined}
14244 @var{recursive-call-returning-count}
14250 How to incorporate the mechanism that counts?
14252 If you are not used to writing recursive functions, a question like
14253 this can be troublesome. But it can and should be approached
14256 We know that the counting mechanism should be associated in some way
14257 with the recursive call. Indeed, since the next-step-expression moves
14258 point forward by one word, and since a recursive call is made for
14259 each word, the counting mechanism must be an expression that adds one
14260 to the value returned by a call to @code{recursive-count-words}.
14263 Consider several cases:
14267 If there are two words in the region, the function should return
14268 a value resulting from adding one to the value returned when it counts
14269 the first word, plus the number returned when it counts the remaining
14270 words in the region, which in this case is one.
14273 If there is one word in the region, the function should return
14274 a value resulting from adding one to the value returned when it counts
14275 that word, plus the number returned when it counts the remaining
14276 words in the region, which in this case is zero.
14279 If there are no words in the region, the function should return zero.
14282 From the sketch we can see that the else-part of the @code{if} returns
14283 zero for the case of no words. This means that the then-part of the
14284 @code{if} must return a value resulting from adding one to the value
14285 returned from a count of the remaining words.
14288 The expression will look like this, where @code{1+} is a function that
14289 adds one to its argument.
14292 (1+ (recursive-count-words region-end))
14296 The whole @code{recursive-count-words} function will then look like
14301 (defun recursive-count-words (region-end)
14302 "@var{documentation}@dots{}"
14304 ;;; @r{1. do-again-test}
14305 (if (and (< (point) region-end)
14306 (re-search-forward "\\w+\\W*" region-end t))
14310 ;;; @r{2. then-part: the recursive call}
14311 (1+ (recursive-count-words region-end))
14313 ;;; @r{3. else-part}
14319 Let's examine how this works:
14321 If there are no words in the region, the else part of the @code{if}
14322 expression is evaluated and consequently the function returns zero.
14324 If there is one word in the region, the value of point is less than
14325 the value of @code{region-end} and the search succeeds. In this case,
14326 the true-or-false-test of the @code{if} expression tests true, and the
14327 then-part of the @code{if} expression is evaluated. The counting
14328 expression is evaluated. This expression returns a value (which will
14329 be the value returned by the whole function) that is the sum of one
14330 added to the value returned by a recursive call.
14332 Meanwhile, the next-step-expression has caused point to jump over the
14333 first (and in this case only) word in the region. This means that
14334 when @code{(recursive-count-words region-end)} is evaluated a second
14335 time, as a result of the recursive call, the value of point will be
14336 equal to or greater than the value of region end. So this time,
14337 @code{recursive-count-words} will return zero. The zero will be added
14338 to one, and the original evaluation of @code{recursive-count-words}
14339 will return one plus zero, which is one, which is the correct amount.
14341 Clearly, if there are two words in the region, the first call to
14342 @code{recursive-count-words} returns one added to the value returned
14343 by calling @code{recursive-count-words} on a region containing the
14344 remaining word---that is, it adds one to one, producing two, which is
14345 the correct amount.
14347 Similarly, if there are three words in the region, the first call to
14348 @code{recursive-count-words} returns one added to the value returned
14349 by calling @code{recursive-count-words} on a region containing the
14350 remaining two words---and so on and so on.
14354 With full documentation the two functions look like this:
14358 The recursive function:
14360 @findex recursive-count-words
14363 (defun recursive-count-words (region-end)
14364 "Number of words between point and REGION-END."
14368 ;;; @r{1. do-again-test}
14369 (if (and (< (point) region-end)
14370 (re-search-forward "\\w+\\W*" region-end t))
14374 ;;; @r{2. then-part: the recursive call}
14375 (1+ (recursive-count-words region-end))
14377 ;;; @r{3. else-part}
14388 ;;; @r{Recursive version}
14389 (defun @value{COUNT-WORDS} (beginning end)
14390 "Print number of words in the region.
14394 Words are defined as at least one word-constituent
14395 character followed by at least one character that is
14396 not a word-constituent. The buffer's syntax table
14397 determines which characters these are."
14401 (message "Counting words in region ... ")
14403 (goto-char beginning)
14404 (let ((count (recursive-count-words end)))
14407 (cond ((zerop count)
14409 "The region does NOT have any words."))
14413 (message "The region has 1 word."))
14416 "The region has %d words." count))))))
14420 @node Counting Exercise
14421 @section Exercise: Counting Punctuation
14423 Using a @code{while} loop, write a function to count the number of
14424 punctuation marks in a region---period, comma, semicolon, colon,
14425 exclamation mark, and question mark. Do the same using recursion.
14427 @node Words in a defun
14428 @chapter Counting Words in a @code{defun}
14429 @cindex Counting words in a @code{defun}
14430 @cindex Word counting in a @code{defun}
14432 Our next project is to count the number of words in a function
14433 definition. Clearly, this can be done using some variant of
14434 @code{@value{COUNT-WORDS}}. @xref{Counting Words, , Counting via
14435 Repetition and Regexps}. If we are just going to count the words in
14436 one definition, it is easy enough to mark the definition with the
14437 @kbd{C-M-h} (@code{mark-defun}) command, and then call
14438 @code{@value{COUNT-WORDS}}.
14440 However, I am more ambitious: I want to count the words and symbols in
14441 every definition in the Emacs sources and then print a graph that
14442 shows how many functions there are of each length: how many contain 40
14443 to 49 words or symbols, how many contain 50 to 59 words or symbols,
14444 and so on. I have often been curious how long a typical function is,
14445 and this will tell.
14448 * Divide and Conquer::
14449 * Words and Symbols:: What to count?
14450 * Syntax:: What constitutes a word or symbol?
14451 * count-words-in-defun:: Very like @code{@value{COUNT-WORDS}}.
14452 * Several defuns:: Counting several defuns in a file.
14453 * Find a File:: Do you want to look at a file?
14454 * lengths-list-file:: A list of the lengths of many definitions.
14455 * Several files:: Counting in definitions in different files.
14456 * Several files recursively:: Recursively counting in different files.
14457 * Prepare the data:: Prepare the data for display in a graph.
14461 @node Divide and Conquer
14462 @unnumberedsec Divide and Conquer
14465 Described in one phrase, the histogram project is daunting; but
14466 divided into numerous small steps, each of which we can take one at a
14467 time, the project becomes less fearsome. Let us consider what the
14472 First, write a function to count the words in one definition. This
14473 includes the problem of handling symbols as well as words.
14476 Second, write a function to list the numbers of words in each function
14477 in a file. This function can use the @code{count-words-in-defun}
14481 Third, write a function to list the numbers of words in each function
14482 in each of several files. This entails automatically finding the
14483 various files, switching to them, and counting the words in the
14484 definitions within them.
14487 Fourth, write a function to convert the list of numbers that we
14488 created in step three to a form that will be suitable for printing as
14492 Fifth, write a function to print the results as a graph.
14495 This is quite a project! But if we take each step slowly, it will not
14498 @node Words and Symbols
14499 @section What to Count?
14500 @cindex Words and symbols in defun
14502 When we first start thinking about how to count the words in a
14503 function definition, the first question is (or ought to be) what are
14504 we going to count? When we speak of `words' with respect to a Lisp
14505 function definition, we are actually speaking, in large part, of
14506 `symbols'. For example, the following @code{multiply-by-seven}
14507 function contains the five symbols @code{defun},
14508 @code{multiply-by-seven}, @code{number}, @code{*}, and @code{7}. In
14509 addition, in the documentation string, it contains the four words
14510 @samp{Multiply}, @samp{NUMBER}, @samp{by}, and @samp{seven}. The
14511 symbol @samp{number} is repeated, so the definition contains a total
14512 of ten words and symbols.
14516 (defun multiply-by-seven (number)
14517 "Multiply NUMBER by seven."
14523 However, if we mark the @code{multiply-by-seven} definition with
14524 @kbd{C-M-h} (@code{mark-defun}), and then call
14525 @code{@value{COUNT-WORDS}} on it, we will find that
14526 @code{@value{COUNT-WORDS}} claims the definition has eleven words, not
14527 ten! Something is wrong!
14529 The problem is twofold: @code{@value{COUNT-WORDS}} does not count the
14530 @samp{*} as a word, and it counts the single symbol,
14531 @code{multiply-by-seven}, as containing three words. The hyphens are
14532 treated as if they were interword spaces rather than intraword
14533 connectors: @samp{multiply-by-seven} is counted as if it were written
14534 @samp{multiply by seven}.
14536 The cause of this confusion is the regular expression search within
14537 the @code{@value{COUNT-WORDS}} definition that moves point forward word
14538 by word. In the canonical version of @code{@value{COUNT-WORDS}}, the
14546 This regular expression is a pattern defining one or more word
14547 constituent characters possibly followed by one or more characters
14548 that are not word constituents. What is meant by `word constituent
14549 characters' brings us to the issue of syntax, which is worth a section
14553 @section What Constitutes a Word or Symbol?
14554 @cindex Syntax categories and tables
14556 Emacs treats different characters as belonging to different
14557 @dfn{syntax categories}. For example, the regular expression,
14558 @samp{\\w+}, is a pattern specifying one or more @emph{word
14559 constituent} characters. Word constituent characters are members of
14560 one syntax category. Other syntax categories include the class of
14561 punctuation characters, such as the period and the comma, and the
14562 class of whitespace characters, such as the blank space and the tab
14563 character. (For more information, @pxref{Syntax Tables, , Syntax
14564 Tables, elisp, The GNU Emacs Lisp Reference Manual}.)
14566 Syntax tables specify which characters belong to which categories.
14567 Usually, a hyphen is not specified as a `word constituent character'.
14568 Instead, it is specified as being in the `class of characters that are
14569 part of symbol names but not words.' This means that the
14570 @code{@value{COUNT-WORDS}} function treats it in the same way it treats
14571 an interword white space, which is why @code{@value{COUNT-WORDS}}
14572 counts @samp{multiply-by-seven} as three words.
14574 There are two ways to cause Emacs to count @samp{multiply-by-seven} as
14575 one symbol: modify the syntax table or modify the regular expression.
14577 We could redefine a hyphen as a word constituent character by
14578 modifying the syntax table that Emacs keeps for each mode. This
14579 action would serve our purpose, except that a hyphen is merely the
14580 most common character within symbols that is not typically a word
14581 constituent character; there are others, too.
14583 Alternatively, we can redefine the regexp used in the
14584 @code{@value{COUNT-WORDS}} definition so as to include symbols. This
14585 procedure has the merit of clarity, but the task is a little tricky.
14588 The first part is simple enough: the pattern must match ``at least one
14589 character that is a word or symbol constituent''. Thus:
14592 "\\(\\w\\|\\s_\\)+"
14596 The @samp{\\(} is the first part of the grouping construct that
14597 includes the @samp{\\w} and the @samp{\\s_} as alternatives, separated
14598 by the @samp{\\|}. The @samp{\\w} matches any word-constituent
14599 character and the @samp{\\s_} matches any character that is part of a
14600 symbol name but not a word-constituent character. The @samp{+}
14601 following the group indicates that the word or symbol constituent
14602 characters must be matched at least once.
14604 However, the second part of the regexp is more difficult to design.
14605 What we want is to follow the first part with ``optionally one or more
14606 characters that are not constituents of a word or symbol''. At first,
14607 I thought I could define this with the following:
14610 "\\(\\W\\|\\S_\\)*"
14614 The upper case @samp{W} and @samp{S} match characters that are
14615 @emph{not} word or symbol constituents. Unfortunately, this
14616 expression matches any character that is either not a word constituent
14617 or not a symbol constituent. This matches any character!
14619 I then noticed that every word or symbol in my test region was
14620 followed by white space (blank space, tab, or newline). So I tried
14621 placing a pattern to match one or more blank spaces after the pattern
14622 for one or more word or symbol constituents. This failed, too. Words
14623 and symbols are often separated by whitespace, but in actual code
14624 parentheses may follow symbols and punctuation may follow words. So
14625 finally, I designed a pattern in which the word or symbol constituents
14626 are followed optionally by characters that are not white space and
14627 then followed optionally by white space.
14630 Here is the full regular expression:
14633 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*"
14636 @node count-words-in-defun
14637 @section The @code{count-words-in-defun} Function
14638 @cindex Counting words in a @code{defun}
14640 We have seen that there are several ways to write a
14641 @code{count-words-region} function. To write a
14642 @code{count-words-in-defun}, we need merely adapt one of these
14645 The version that uses a @code{while} loop is easy to understand, so I
14646 am going to adapt that. Because @code{count-words-in-defun} will be
14647 part of a more complex program, it need not be interactive and it need
14648 not display a message but just return the count. These considerations
14649 simplify the definition a little.
14651 On the other hand, @code{count-words-in-defun} will be used within a
14652 buffer that contains function definitions. Consequently, it is
14653 reasonable to ask that the function determine whether it is called
14654 when point is within a function definition, and if it is, to return
14655 the count for that definition. This adds complexity to the
14656 definition, but saves us from needing to pass arguments to the
14660 These considerations lead us to prepare the following template:
14664 (defun count-words-in-defun ()
14665 "@var{documentation}@dots{}"
14666 (@var{set up}@dots{}
14667 (@var{while loop}@dots{})
14668 @var{return count})
14673 As usual, our job is to fill in the slots.
14677 We are presuming that this function will be called within a buffer
14678 containing function definitions. Point will either be within a
14679 function definition or not. For @code{count-words-in-defun} to work,
14680 point must move to the beginning of the definition, a counter must
14681 start at zero, and the counting loop must stop when point reaches the
14682 end of the definition.
14684 The @code{beginning-of-defun} function searches backwards for an
14685 opening delimiter such as a @samp{(} at the beginning of a line, and
14686 moves point to that position, or else to the limit of the search. In
14687 practice, this means that @code{beginning-of-defun} moves point to the
14688 beginning of an enclosing or preceding function definition, or else to
14689 the beginning of the buffer. We can use @code{beginning-of-defun} to
14690 place point where we wish to start.
14692 The @code{while} loop requires a counter to keep track of the words or
14693 symbols being counted. A @code{let} expression can be used to create
14694 a local variable for this purpose, and bind it to an initial value of zero.
14696 The @code{end-of-defun} function works like @code{beginning-of-defun}
14697 except that it moves point to the end of the definition.
14698 @code{end-of-defun} can be used as part of an expression that
14699 determines the position of the end of the definition.
14701 The set up for @code{count-words-in-defun} takes shape rapidly: first
14702 we move point to the beginning of the definition, then we create a
14703 local variable to hold the count, and finally, we record the position
14704 of the end of the definition so the @code{while} loop will know when to stop
14708 The code looks like this:
14712 (beginning-of-defun)
14714 (end (save-excursion (end-of-defun) (point))))
14719 The code is simple. The only slight complication is likely to concern
14720 @code{end}: it is bound to the position of the end of the definition
14721 by a @code{save-excursion} expression that returns the value of point
14722 after @code{end-of-defun} temporarily moves it to the end of the
14725 The second part of the @code{count-words-in-defun}, after the set up,
14726 is the @code{while} loop.
14728 The loop must contain an expression that jumps point forward word by
14729 word and symbol by symbol, and another expression that counts the
14730 jumps. The true-or-false-test for the @code{while} loop should test
14731 true so long as point should jump forward, and false when point is at
14732 the end of the definition. We have already redefined the regular
14733 expression for this, so the loop is straightforward:
14737 (while (and (< (point) end)
14739 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*" end t))
14740 (setq count (1+ count)))
14744 The third part of the function definition returns the count of words
14745 and symbols. This part is the last expression within the body of the
14746 @code{let} expression, and can be, very simply, the local variable
14747 @code{count}, which when evaluated returns the count.
14750 Put together, the @code{count-words-in-defun} definition looks like this:
14752 @findex count-words-in-defun
14755 (defun count-words-in-defun ()
14756 "Return the number of words and symbols in a defun."
14757 (beginning-of-defun)
14759 (end (save-excursion (end-of-defun) (point))))
14763 (and (< (point) end)
14765 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*"
14767 (setq count (1+ count)))
14772 How to test this? The function is not interactive, but it is easy to
14773 put a wrapper around the function to make it interactive; we can use
14774 almost the same code as for the recursive version of
14775 @code{@value{COUNT-WORDS}}:
14779 ;;; @r{Interactive version.}
14780 (defun count-words-defun ()
14781 "Number of words and symbols in a function definition."
14784 "Counting words and symbols in function definition ... ")
14787 (let ((count (count-words-in-defun)))
14791 "The definition does NOT have any words or symbols."))
14796 "The definition has 1 word or symbol."))
14799 "The definition has %d words or symbols." count)))))
14805 Let's re-use @kbd{C-c =} as a convenient keybinding:
14808 (global-set-key "\C-c=" 'count-words-defun)
14811 Now we can try out @code{count-words-defun}: install both
14812 @code{count-words-in-defun} and @code{count-words-defun}, and set the
14813 keybinding, and then place the cursor within the following definition:
14817 (defun multiply-by-seven (number)
14818 "Multiply NUMBER by seven."
14825 Success! The definition has 10 words and symbols.
14827 The next problem is to count the numbers of words and symbols in
14828 several definitions within a single file.
14830 @node Several defuns
14831 @section Count Several @code{defuns} Within a File
14833 A file such as @file{simple.el} may have a hundred or more function
14834 definitions within it. Our long term goal is to collect statistics on
14835 many files, but as a first step, our immediate goal is to collect
14836 statistics on one file.
14838 The information will be a series of numbers, each number being the
14839 length of a function definition. We can store the numbers in a list.
14841 We know that we will want to incorporate the information regarding one
14842 file with information about many other files; this means that the
14843 function for counting definition lengths within one file need only
14844 return the list of lengths. It need not and should not display any
14847 The word count commands contain one expression to jump point forward
14848 word by word and another expression to count the jumps. The function
14849 to return the lengths of definitions can be designed to work the same
14850 way, with one expression to jump point forward definition by
14851 definition and another expression to construct the lengths' list.
14853 This statement of the problem makes it elementary to write the
14854 function definition. Clearly, we will start the count at the
14855 beginning of the file, so the first command will be @code{(goto-char
14856 (point-min))}. Next, we start the @code{while} loop; and the
14857 true-or-false test of the loop can be a regular expression search for
14858 the next function definition---so long as the search succeeds, point
14859 is moved forward and then the body of the loop is evaluated. The body
14860 needs an expression that constructs the lengths' list. @code{cons},
14861 the list construction command, can be used to create the list. That
14862 is almost all there is to it.
14865 Here is what this fragment of code looks like:
14869 (goto-char (point-min))
14870 (while (re-search-forward "^(defun" nil t)
14872 (cons (count-words-in-defun) lengths-list)))
14876 What we have left out is the mechanism for finding the file that
14877 contains the function definitions.
14879 In previous examples, we either used this, the Info file, or we
14880 switched back and forth to some other buffer, such as the
14881 @file{*scratch*} buffer.
14883 Finding a file is a new process that we have not yet discussed.
14886 @section Find a File
14887 @cindex Find a File
14889 To find a file in Emacs, you use the @kbd{C-x C-f} (@code{find-file})
14890 command. This command is almost, but not quite right for the lengths
14894 Let's look at the source for @code{find-file}:
14898 (defun find-file (filename)
14899 "Edit file FILENAME.
14900 Switch to a buffer visiting file FILENAME,
14901 creating one if none already exists."
14902 (interactive "FFind file: ")
14903 (switch-to-buffer (find-file-noselect filename)))
14908 (The most recent version of the @code{find-file} function definition
14909 permits you to specify optional wildcards to visit multiple files; that
14910 makes the definition more complex and we will not discuss it here,
14911 since it is not relevant. You can see its source using either
14912 @kbd{M-.} (@code{find-tag}) or @kbd{C-h f} (@code{describe-function}).)
14916 (defun find-file (filename &optional wildcards)
14917 "Edit file FILENAME.
14918 Switch to a buffer visiting file FILENAME,
14919 creating one if none already exists.
14920 Interactively, the default if you just type RET is the current directory,
14921 but the visited file name is available through the minibuffer history:
14922 type M-n to pull it into the minibuffer.
14924 Interactively, or if WILDCARDS is non-nil in a call from Lisp,
14925 expand wildcards (if any) and visit multiple files. You can
14926 suppress wildcard expansion by setting `find-file-wildcards' to nil.
14928 To visit a file without any kind of conversion and without
14929 automatically choosing a major mode, use \\[find-file-literally]."
14930 (interactive (find-file-read-args "Find file: " nil))
14931 (let ((value (find-file-noselect filename nil nil wildcards)))
14933 (mapcar 'switch-to-buffer (nreverse value))
14934 (switch-to-buffer value))))
14937 The definition I am showing possesses short but complete documentation
14938 and an interactive specification that prompts you for a file name when
14939 you use the command interactively. The body of the definition
14940 contains two functions, @code{find-file-noselect} and
14941 @code{switch-to-buffer}.
14943 According to its documentation as shown by @kbd{C-h f} (the
14944 @code{describe-function} command), the @code{find-file-noselect}
14945 function reads the named file into a buffer and returns the buffer.
14946 (Its most recent version includes an optional wildcards argument,
14947 too, as well as another to read a file literally and an other you
14948 suppress warning messages. These optional arguments are irrelevant.)
14950 However, the @code{find-file-noselect} function does not select the
14951 buffer in which it puts the file. Emacs does not switch its attention
14952 (or yours if you are using @code{find-file-noselect}) to the selected
14953 buffer. That is what @code{switch-to-buffer} does: it switches the
14954 buffer to which Emacs attention is directed; and it switches the
14955 buffer displayed in the window to the new buffer. We have discussed
14956 buffer switching elsewhere. (@xref{Switching Buffers}.)
14958 In this histogram project, we do not need to display each file on the
14959 screen as the program determines the length of each definition within
14960 it. Instead of employing @code{switch-to-buffer}, we can work with
14961 @code{set-buffer}, which redirects the attention of the computer
14962 program to a different buffer but does not redisplay it on the screen.
14963 So instead of calling on @code{find-file} to do the job, we must write
14964 our own expression.
14966 The task is easy: use @code{find-file-noselect} and @code{set-buffer}.
14968 @node lengths-list-file
14969 @section @code{lengths-list-file} in Detail
14971 The core of the @code{lengths-list-file} function is a @code{while}
14972 loop containing a function to move point forward `defun by defun' and
14973 a function to count the number of words and symbols in each defun.
14974 This core must be surrounded by functions that do various other tasks,
14975 including finding the file, and ensuring that point starts out at the
14976 beginning of the file. The function definition looks like this:
14977 @findex lengths-list-file
14981 (defun lengths-list-file (filename)
14982 "Return list of definitions' lengths within FILE.
14983 The returned list is a list of numbers.
14984 Each number is the number of words or
14985 symbols in one function definition."
14988 (message "Working on `%s' ... " filename)
14990 (let ((buffer (find-file-noselect filename))
14992 (set-buffer buffer)
14993 (setq buffer-read-only t)
14995 (goto-char (point-min))
14996 (while (re-search-forward "^(defun" nil t)
14998 (cons (count-words-in-defun) lengths-list)))
14999 (kill-buffer buffer)
15005 The function is passed one argument, the name of the file on which it
15006 will work. It has four lines of documentation, but no interactive
15007 specification. Since people worry that a computer is broken if they
15008 don't see anything going on, the first line of the body is a
15011 The next line contains a @code{save-excursion} that returns Emacs's
15012 attention to the current buffer when the function completes. This is
15013 useful in case you embed this function in another function that
15014 presumes point is restored to the original buffer.
15016 In the varlist of the @code{let} expression, Emacs finds the file and
15017 binds the local variable @code{buffer} to the buffer containing the
15018 file. At the same time, Emacs creates @code{lengths-list} as a local
15021 Next, Emacs switches its attention to the buffer.
15023 In the following line, Emacs makes the buffer read-only. Ideally,
15024 this line is not necessary. None of the functions for counting words
15025 and symbols in a function definition should change the buffer.
15026 Besides, the buffer is not going to be saved, even if it were changed.
15027 This line is entirely the consequence of great, perhaps excessive,
15028 caution. The reason for the caution is that this function and those
15029 it calls work on the sources for Emacs and it is inconvenient if they
15030 are inadvertently modified. It goes without saying that I did not
15031 realize a need for this line until an experiment went awry and started
15032 to modify my Emacs source files @dots{}
15034 Next comes a call to widen the buffer if it is narrowed. This
15035 function is usually not needed---Emacs creates a fresh buffer if none
15036 already exists; but if a buffer visiting the file already exists Emacs
15037 returns that one. In this case, the buffer may be narrowed and must
15038 be widened. If we wanted to be fully `user-friendly', we would
15039 arrange to save the restriction and the location of point, but we
15042 The @code{(goto-char (point-min))} expression moves point to the
15043 beginning of the buffer.
15045 Then comes a @code{while} loop in which the `work' of the function is
15046 carried out. In the loop, Emacs determines the length of each
15047 definition and constructs a lengths' list containing the information.
15049 Emacs kills the buffer after working through it. This is to save
15050 space inside of Emacs. My version of GNU Emacs 19 contained over 300
15051 source files of interest; GNU Emacs 22 contains over a thousand source
15052 files. Another function will apply @code{lengths-list-file} to each
15055 Finally, the last expression within the @code{let} expression is the
15056 @code{lengths-list} variable; its value is returned as the value of
15057 the whole function.
15059 You can try this function by installing it in the usual fashion. Then
15060 place your cursor after the following expression and type @kbd{C-x
15061 C-e} (@code{eval-last-sexp}).
15063 @c !!! 22.1.1 lisp sources location here
15066 "/usr/local/share/emacs/22.1.1/lisp/emacs-lisp/debug.el")
15070 (You may need to change the pathname of the file; the one here is for
15071 GNU Emacs version 22.1.1. To change the expression, copy it to
15072 the @file{*scratch*} buffer and edit it.
15076 (Also, to see the full length of the list, rather than a truncated
15077 version, you may have to evaluate the following:
15080 (custom-set-variables '(eval-expression-print-length nil))
15084 (@xref{defcustom, , Specifying Variables using @code{defcustom}}.
15085 Then evaluate the @code{lengths-list-file} expression.)
15088 The lengths' list for @file{debug.el} takes less than a second to
15089 produce and looks like this in GNU Emacs 22:
15092 (83 113 105 144 289 22 30 97 48 89 25 52 52 88 28 29 77 49 43 290 232 587)
15096 (Using my old machine, the version 19 lengths' list for @file{debug.el}
15097 took seven seconds to produce and looked like this:
15100 (75 41 80 62 20 45 44 68 45 12 34 235)
15103 (The newer version of @file{debug.el} contains more defuns than the
15104 earlier one; and my new machine is much faster than the old one.)
15106 Note that the length of the last definition in the file is first in
15109 @node Several files
15110 @section Count Words in @code{defuns} in Different Files
15112 In the previous section, we created a function that returns a list of
15113 the lengths of each definition in a file. Now, we want to define a
15114 function to return a master list of the lengths of the definitions in
15117 Working on each of a list of files is a repetitious act, so we can use
15118 either a @code{while} loop or recursion.
15121 * lengths-list-many-files:: Return a list of the lengths of defuns.
15122 * append:: Attach one list to another.
15126 @node lengths-list-many-files
15127 @unnumberedsubsec Determine the lengths of @code{defuns}
15130 The design using a @code{while} loop is routine. The argument passed
15131 the function is a list of files. As we saw earlier (@pxref{Loop
15132 Example}), you can write a @code{while} loop so that the body of the
15133 loop is evaluated if such a list contains elements, but to exit the
15134 loop if the list is empty. For this design to work, the body of the
15135 loop must contain an expression that shortens the list each time the
15136 body is evaluated, so that eventually the list is empty. The usual
15137 technique is to set the value of the list to the value of the @sc{cdr}
15138 of the list each time the body is evaluated.
15141 The template looks like this:
15145 (while @var{test-whether-list-is-empty}
15147 @var{set-list-to-cdr-of-list})
15151 Also, we remember that a @code{while} loop returns @code{nil} (the
15152 result of evaluating the true-or-false-test), not the result of any
15153 evaluation within its body. (The evaluations within the body of the
15154 loop are done for their side effects.) However, the expression that
15155 sets the lengths' list is part of the body---and that is the value
15156 that we want returned by the function as a whole. To do this, we
15157 enclose the @code{while} loop within a @code{let} expression, and
15158 arrange that the last element of the @code{let} expression contains
15159 the value of the lengths' list. (@xref{Incrementing Example, , Loop
15160 Example with an Incrementing Counter}.)
15162 @findex lengths-list-many-files
15164 These considerations lead us directly to the function itself:
15168 ;;; @r{Use @code{while} loop.}
15169 (defun lengths-list-many-files (list-of-files)
15170 "Return list of lengths of defuns in LIST-OF-FILES."
15173 (let (lengths-list)
15175 ;;; @r{true-or-false-test}
15176 (while list-of-files
15181 ;;; @r{Generate a lengths' list.}
15183 (expand-file-name (car list-of-files)))))
15187 ;;; @r{Make files' list shorter.}
15188 (setq list-of-files (cdr list-of-files)))
15190 ;;; @r{Return final value of lengths' list.}
15195 @code{expand-file-name} is a built-in function that converts a file
15196 name to the absolute, long, path name form. The function employs the
15197 name of the directory in which the function is called.
15199 @c !!! 22.1.1 lisp sources location here
15201 Thus, if @code{expand-file-name} is called on @code{debug.el} when
15202 Emacs is visiting the
15203 @file{/usr/local/share/emacs/22.1.1/lisp/emacs-lisp/} directory,
15213 @c !!! 22.1.1 lisp sources location here
15215 /usr/local/share/emacs/22.1.1/lisp/emacs-lisp/debug.el
15218 The only other new element of this function definition is the as yet
15219 unstudied function @code{append}, which merits a short section for
15223 @subsection The @code{append} Function
15226 The @code{append} function attaches one list to another. Thus,
15229 (append '(1 2 3 4) '(5 6 7 8))
15240 This is exactly how we want to attach two lengths' lists produced by
15241 @code{lengths-list-file} to each other. The results contrast with
15245 (cons '(1 2 3 4) '(5 6 7 8))
15250 which constructs a new list in which the first argument to @code{cons}
15251 becomes the first element of the new list:
15254 ((1 2 3 4) 5 6 7 8)
15257 @node Several files recursively
15258 @section Recursively Count Words in Different Files
15260 Besides a @code{while} loop, you can work on each of a list of files
15261 with recursion. A recursive version of @code{lengths-list-many-files}
15262 is short and simple.
15264 The recursive function has the usual parts: the `do-again-test', the
15265 `next-step-expression', and the recursive call. The `do-again-test'
15266 determines whether the function should call itself again, which it
15267 will do if the @code{list-of-files} contains any remaining elements;
15268 the `next-step-expression' resets the @code{list-of-files} to the
15269 @sc{cdr} of itself, so eventually the list will be empty; and the
15270 recursive call calls itself on the shorter list. The complete
15271 function is shorter than this description!
15272 @findex recursive-lengths-list-many-files
15276 (defun recursive-lengths-list-many-files (list-of-files)
15277 "Return list of lengths of each defun in LIST-OF-FILES."
15278 (if list-of-files ; @r{do-again-test}
15281 (expand-file-name (car list-of-files)))
15282 (recursive-lengths-list-many-files
15283 (cdr list-of-files)))))
15288 In a sentence, the function returns the lengths' list for the first of
15289 the @code{list-of-files} appended to the result of calling itself on
15290 the rest of the @code{list-of-files}.
15292 Here is a test of @code{recursive-lengths-list-many-files}, along with
15293 the results of running @code{lengths-list-file} on each of the files
15296 Install @code{recursive-lengths-list-many-files} and
15297 @code{lengths-list-file}, if necessary, and then evaluate the
15298 following expressions. You may need to change the files' pathnames;
15299 those here work when this Info file and the Emacs sources are located
15300 in their customary places. To change the expressions, copy them to
15301 the @file{*scratch*} buffer, edit them, and then evaluate them.
15303 The results are shown after the @samp{@result{}}. (These results are
15304 for files from Emacs version 22.1.1; files from other versions of
15305 Emacs may produce different results.)
15307 @c !!! 22.1.1 lisp sources location here
15310 (cd "/usr/local/share/emacs/22.1.1/")
15312 (lengths-list-file "./lisp/macros.el")
15313 @result{} (283 263 480 90)
15317 (lengths-list-file "./lisp/mail/mailalias.el")
15318 @result{} (38 32 29 95 178 180 321 218 324)
15322 (lengths-list-file "./lisp/makesum.el")
15327 (recursive-lengths-list-many-files
15328 '("./lisp/macros.el"
15329 "./lisp/mail/mailalias.el"
15330 "./lisp/makesum.el"))
15331 @result{} (283 263 480 90 38 32 29 95 178 180 321 218 324 85 181)
15335 The @code{recursive-lengths-list-many-files} function produces the
15338 The next step is to prepare the data in the list for display in a graph.
15340 @node Prepare the data
15341 @section Prepare the Data for Display in a Graph
15343 The @code{recursive-lengths-list-many-files} function returns a list
15344 of numbers. Each number records the length of a function definition.
15345 What we need to do now is transform this data into a list of numbers
15346 suitable for generating a graph. The new list will tell how many
15347 functions definitions contain less than 10 words and
15348 symbols, how many contain between 10 and 19 words and symbols, how
15349 many contain between 20 and 29 words and symbols, and so on.
15351 In brief, we need to go through the lengths' list produced by the
15352 @code{recursive-lengths-list-many-files} function and count the number
15353 of defuns within each range of lengths, and produce a list of those
15357 * Data for Display in Detail::
15358 * Sorting:: Sorting lists.
15359 * Files List:: Making a list of files.
15360 * Counting function definitions::
15364 @node Data for Display in Detail
15365 @unnumberedsubsec The Data for Display in Detail
15368 Based on what we have done before, we can readily foresee that it
15369 should not be too hard to write a function that `@sc{cdr}s' down the
15370 lengths' list, looks at each element, determines which length range it
15371 is in, and increments a counter for that range.
15373 However, before beginning to write such a function, we should consider
15374 the advantages of sorting the lengths' list first, so the numbers are
15375 ordered from smallest to largest. First, sorting will make it easier
15376 to count the numbers in each range, since two adjacent numbers will
15377 either be in the same length range or in adjacent ranges. Second, by
15378 inspecting a sorted list, we can discover the highest and lowest
15379 number, and thereby determine the largest and smallest length range
15383 @subsection Sorting Lists
15386 Emacs contains a function to sort lists, called (as you might guess)
15387 @code{sort}. The @code{sort} function takes two arguments, the list
15388 to be sorted, and a predicate that determines whether the first of
15389 two list elements is ``less'' than the second.
15391 As we saw earlier (@pxref{Wrong Type of Argument, , Using the Wrong
15392 Type Object as an Argument}), a predicate is a function that
15393 determines whether some property is true or false. The @code{sort}
15394 function will reorder a list according to whatever property the
15395 predicate uses; this means that @code{sort} can be used to sort
15396 non-numeric lists by non-numeric criteria---it can, for example,
15397 alphabetize a list.
15400 The @code{<} function is used when sorting a numeric list. For example,
15403 (sort '(4 8 21 17 33 7 21 7) '<)
15411 (4 7 7 8 17 21 21 33)
15415 (Note that in this example, both the arguments are quoted so that the
15416 symbols are not evaluated before being passed to @code{sort} as
15419 Sorting the list returned by the
15420 @code{recursive-lengths-list-many-files} function is straightforward;
15421 it uses the @code{<} function:
15425 In GNU Emacs 22, eval
15427 (cd "/usr/local/share/emacs/22.0.50/")
15429 (recursive-lengths-list-many-files
15430 '("./lisp/macros.el"
15431 "./lisp/mail/mailalias.el"
15432 "./lisp/makesum.el"))
15440 (recursive-lengths-list-many-files
15441 '("./lisp/macros.el"
15442 "./lisp/mailalias.el"
15443 "./lisp/makesum.el"))
15453 (29 32 38 85 90 95 178 180 181 218 263 283 321 324 480)
15457 (Note that in this example, the first argument to @code{sort} is not
15458 quoted, since the expression must be evaluated so as to produce the
15459 list that is passed to @code{sort}.)
15462 @subsection Making a List of Files
15464 The @code{recursive-lengths-list-many-files} function requires a list
15465 of files as its argument. For our test examples, we constructed such
15466 a list by hand; but the Emacs Lisp source directory is too large for
15467 us to do for that. Instead, we will write a function to do the job
15468 for us. In this function, we will use both a @code{while} loop and a
15471 @findex directory-files
15472 We did not have to write a function like this for older versions of
15473 GNU Emacs, since they placed all the @samp{.el} files in one
15474 directory. Instead, we were able to use the @code{directory-files}
15475 function, which lists the names of files that match a specified
15476 pattern within a single directory.
15478 However, recent versions of Emacs place Emacs Lisp files in
15479 sub-directories of the top level @file{lisp} directory. This
15480 re-arrangement eases navigation. For example, all the mail related
15481 files are in a @file{lisp} sub-directory called @file{mail}. But at
15482 the same time, this arrangement forces us to create a file listing
15483 function that descends into the sub-directories.
15485 @findex files-in-below-directory
15486 We can create this function, called @code{files-in-below-directory},
15487 using familiar functions such as @code{car}, @code{nthcdr}, and
15488 @code{substring} in conjunction with an existing function called
15489 @code{directory-files-and-attributes}. This latter function not only
15490 lists all the filenames in a directory, including the names
15491 of sub-directories, but also their attributes.
15493 To restate our goal: to create a function that will enable us
15494 to feed filenames to @code{recursive-lengths-list-many-files}
15495 as a list that looks like this (but with more elements):
15499 ("./lisp/macros.el"
15500 "./lisp/mail/rmail.el"
15501 "./lisp/makesum.el")
15505 The @code{directory-files-and-attributes} function returns a list of
15506 lists. Each of the lists within the main list consists of 13
15507 elements. The first element is a string that contains the name of the
15508 file---which, in GNU/Linux, may be a `directory file', that is to
15509 say, a file with the special attributes of a directory. The second
15510 element of the list is @code{t} for a directory, a string
15511 for symbolic link (the string is the name linked to), or @code{nil}.
15513 For example, the first @samp{.el} file in the @file{lisp/} directory
15514 is @file{abbrev.el}. Its name is
15515 @file{/usr/local/share/emacs/22.1.1/lisp/abbrev.el} and it is not a
15516 directory or a symbolic link.
15519 This is how @code{directory-files-and-attributes} lists that file and
15531 (20615 27034 579989 697000)
15533 (20615 26327 734791 805000)
15545 On the other hand, @file{mail/} is a directory within the @file{lisp/}
15546 directory. The beginning of its listing looks like this:
15557 (To learn about the different attributes, look at the documentation of
15558 @code{file-attributes}. Bear in mind that the @code{file-attributes}
15559 function does not list the filename, so its first element is
15560 @code{directory-files-and-attributes}'s second element.)
15562 We will want our new function, @code{files-in-below-directory}, to
15563 list the @samp{.el} files in the directory it is told to check, and in
15564 any directories below that directory.
15566 This gives us a hint on how to construct
15567 @code{files-in-below-directory}: within a directory, the function
15568 should add @samp{.el} filenames to a list; and if, within a directory,
15569 the function comes upon a sub-directory, it should go into that
15570 sub-directory and repeat its actions.
15572 However, we should note that every directory contains a name that
15573 refers to itself, called @file{.}, (``dot'') and a name that refers to
15574 its parent directory, called @file{..} (``double dot''). (In
15575 @file{/}, the root directory, @file{..} refers to itself, since
15576 @file{/} has no parent.) Clearly, we do not want our
15577 @code{files-in-below-directory} function to enter those directories,
15578 since they always lead us, directly or indirectly, to the current
15581 Consequently, our @code{files-in-below-directory} function must do
15586 Check to see whether it is looking at a filename that ends in
15587 @samp{.el}; and if so, add its name to a list.
15590 Check to see whether it is looking at a filename that is the name of a
15591 directory; and if so,
15595 Check to see whether it is looking at @file{.} or @file{..}; and if
15599 Or else, go into that directory and repeat the process.
15603 Let's write a function definition to do these tasks. We will use a
15604 @code{while} loop to move from one filename to another within a
15605 directory, checking what needs to be done; and we will use a recursive
15606 call to repeat the actions on each sub-directory. The recursive
15607 pattern is `accumulate'
15608 (@pxref{Accumulate}),
15609 using @code{append} as the combiner.
15612 (directory-files "/usr/local/src/emacs/lisp/" t "\\.el$")
15613 (shell-command "find /usr/local/src/emacs/lisp/ -name '*.el'")
15615 (directory-files "/usr/local/share/emacs/22.1.1/lisp/" t "\\.el$")
15616 (shell-command "find /usr/local/share/emacs/22.1.1/lisp/ -name '*.el'")
15619 @c /usr/local/share/emacs/22.1.1/lisp/
15622 Here is the function:
15626 (defun files-in-below-directory (directory)
15627 "List the .el files in DIRECTORY and in its sub-directories."
15628 ;; Although the function will be used non-interactively,
15629 ;; it will be easier to test if we make it interactive.
15630 ;; The directory will have a name such as
15631 ;; "/usr/local/share/emacs/22.1.1/lisp/"
15632 (interactive "DDirectory name: ")
15635 (let (el-files-list
15636 (current-directory-list
15637 (directory-files-and-attributes directory t)))
15638 ;; while we are in the current directory
15639 (while current-directory-list
15643 ;; check to see whether filename ends in `.el'
15644 ;; and if so, append its name to a list.
15645 ((equal ".el" (substring (car (car current-directory-list)) -3))
15646 (setq el-files-list
15647 (cons (car (car current-directory-list)) el-files-list)))
15650 ;; check whether filename is that of a directory
15651 ((eq t (car (cdr (car current-directory-list))))
15652 ;; decide whether to skip or recurse
15655 (substring (car (car current-directory-list)) -1))
15656 ;; then do nothing since filename is that of
15657 ;; current directory or parent, "." or ".."
15661 ;; else descend into the directory and repeat the process
15662 (setq el-files-list
15664 (files-in-below-directory
15665 (car (car current-directory-list)))
15667 ;; move to the next filename in the list; this also
15668 ;; shortens the list so the while loop eventually comes to an end
15669 (setq current-directory-list (cdr current-directory-list)))
15670 ;; return the filenames
15675 @c (files-in-below-directory "/usr/local/src/emacs/lisp/")
15676 @c (files-in-below-directory "/usr/local/share/emacs/22.1.1/lisp/")
15678 The @code{files-in-below-directory} @code{directory-files} function
15679 takes one argument, the name of a directory.
15682 Thus, on my system,
15684 @c (length (files-in-below-directory "/usr/local/src/emacs/lisp/"))
15686 @c !!! 22.1.1 lisp sources location here
15690 (files-in-below-directory "/usr/local/share/emacs/22.1.1/lisp/"))
15695 tells me that in and below my Lisp sources directory are 1031
15698 @code{files-in-below-directory} returns a list in reverse alphabetical
15699 order. An expression to sort the list in alphabetical order looks
15705 (files-in-below-directory "/usr/local/share/emacs/22.1.1/lisp/")
15712 "Test how long it takes to find lengths of all sorted elisp defuns."
15713 (insert "\n" (current-time-string) "\n")
15716 (recursive-lengths-list-many-files
15717 (files-in-below-directory "/usr/local/src/emacs/lisp/"))
15719 (insert (format "%s" (current-time-string))))
15722 @node Counting function definitions
15723 @subsection Counting function definitions
15725 Our immediate goal is to generate a list that tells us how many
15726 function definitions contain fewer than 10 words and symbols, how many
15727 contain between 10 and 19 words and symbols, how many contain between
15728 20 and 29 words and symbols, and so on.
15730 With a sorted list of numbers, this is easy: count how many elements
15731 of the list are smaller than 10, then, after moving past the numbers
15732 just counted, count how many are smaller than 20, then, after moving
15733 past the numbers just counted, count how many are smaller than 30, and
15734 so on. Each of the numbers, 10, 20, 30, 40, and the like, is one
15735 larger than the top of that range. We can call the list of such
15736 numbers the @code{top-of-ranges} list.
15739 If we wished, we could generate this list automatically, but it is
15740 simpler to write a list manually. Here it is:
15741 @vindex top-of-ranges
15745 (defvar top-of-ranges
15748 110 120 130 140 150
15749 160 170 180 190 200
15750 210 220 230 240 250
15751 260 270 280 290 300)
15752 "List specifying ranges for `defuns-per-range'.")
15756 To change the ranges, we edit this list.
15758 Next, we need to write the function that creates the list of the
15759 number of definitions within each range. Clearly, this function must
15760 take the @code{sorted-lengths} and the @code{top-of-ranges} lists
15763 The @code{defuns-per-range} function must do two things again and
15764 again: it must count the number of definitions within a range
15765 specified by the current top-of-range value; and it must shift to the
15766 next higher value in the @code{top-of-ranges} list after counting the
15767 number of definitions in the current range. Since each of these
15768 actions is repetitive, we can use @code{while} loops for the job.
15769 One loop counts the number of definitions in the range defined by the
15770 current top-of-range value, and the other loop selects each of the
15771 top-of-range values in turn.
15773 Several entries of the @code{sorted-lengths} list are counted for each
15774 range; this means that the loop for the @code{sorted-lengths} list
15775 will be inside the loop for the @code{top-of-ranges} list, like a
15776 small gear inside a big gear.
15778 The inner loop counts the number of definitions within the range. It
15779 is a simple counting loop of the type we have seen before.
15780 (@xref{Incrementing Loop, , A loop with an incrementing counter}.)
15781 The true-or-false test of the loop tests whether the value from the
15782 @code{sorted-lengths} list is smaller than the current value of the
15783 top of the range. If it is, the function increments the counter and
15784 tests the next value from the @code{sorted-lengths} list.
15787 The inner loop looks like this:
15791 (while @var{length-element-smaller-than-top-of-range}
15792 (setq number-within-range (1+ number-within-range))
15793 (setq sorted-lengths (cdr sorted-lengths)))
15797 The outer loop must start with the lowest value of the
15798 @code{top-of-ranges} list, and then be set to each of the succeeding
15799 higher values in turn. This can be done with a loop like this:
15803 (while top-of-ranges
15804 @var{body-of-loop}@dots{}
15805 (setq top-of-ranges (cdr top-of-ranges)))
15810 Put together, the two loops look like this:
15814 (while top-of-ranges
15816 ;; @r{Count the number of elements within the current range.}
15817 (while @var{length-element-smaller-than-top-of-range}
15818 (setq number-within-range (1+ number-within-range))
15819 (setq sorted-lengths (cdr sorted-lengths)))
15821 ;; @r{Move to next range.}
15822 (setq top-of-ranges (cdr top-of-ranges)))
15826 In addition, in each circuit of the outer loop, Emacs should record
15827 the number of definitions within that range (the value of
15828 @code{number-within-range}) in a list. We can use @code{cons} for
15829 this purpose. (@xref{cons, , @code{cons}}.)
15831 The @code{cons} function works fine, except that the list it
15832 constructs will contain the number of definitions for the highest
15833 range at its beginning and the number of definitions for the lowest
15834 range at its end. This is because @code{cons} attaches new elements
15835 of the list to the beginning of the list, and since the two loops are
15836 working their way through the lengths' list from the lower end first,
15837 the @code{defuns-per-range-list} will end up largest number first.
15838 But we will want to print our graph with smallest values first and the
15839 larger later. The solution is to reverse the order of the
15840 @code{defuns-per-range-list}. We can do this using the
15841 @code{nreverse} function, which reverses the order of a list.
15848 (nreverse '(1 2 3 4))
15859 Note that the @code{nreverse} function is ``destructive''---that is,
15860 it changes the list to which it is applied; this contrasts with the
15861 @code{car} and @code{cdr} functions, which are non-destructive. In
15862 this case, we do not want the original @code{defuns-per-range-list},
15863 so it does not matter that it is destroyed. (The @code{reverse}
15864 function provides a reversed copy of a list, leaving the original list
15869 Put all together, the @code{defuns-per-range} looks like this:
15873 (defun defuns-per-range (sorted-lengths top-of-ranges)
15874 "SORTED-LENGTHS defuns in each TOP-OF-RANGES range."
15875 (let ((top-of-range (car top-of-ranges))
15876 (number-within-range 0)
15877 defuns-per-range-list)
15882 (while top-of-ranges
15888 ;; @r{Need number for numeric test.}
15889 (car sorted-lengths)
15890 (< (car sorted-lengths) top-of-range))
15894 ;; @r{Count number of definitions within current range.}
15895 (setq number-within-range (1+ number-within-range))
15896 (setq sorted-lengths (cdr sorted-lengths)))
15898 ;; @r{Exit inner loop but remain within outer loop.}
15902 (setq defuns-per-range-list
15903 (cons number-within-range defuns-per-range-list))
15904 (setq number-within-range 0) ; @r{Reset count to zero.}
15908 ;; @r{Move to next range.}
15909 (setq top-of-ranges (cdr top-of-ranges))
15910 ;; @r{Specify next top of range value.}
15911 (setq top-of-range (car top-of-ranges)))
15915 ;; @r{Exit outer loop and count the number of defuns larger than}
15916 ;; @r{ the largest top-of-range value.}
15917 (setq defuns-per-range-list
15919 (length sorted-lengths)
15920 defuns-per-range-list))
15924 ;; @r{Return a list of the number of definitions within each range,}
15925 ;; @r{ smallest to largest.}
15926 (nreverse defuns-per-range-list)))
15932 The function is straightforward except for one subtle feature. The
15933 true-or-false test of the inner loop looks like this:
15937 (and (car sorted-lengths)
15938 (< (car sorted-lengths) top-of-range))
15944 instead of like this:
15947 (< (car sorted-lengths) top-of-range)
15950 The purpose of the test is to determine whether the first item in the
15951 @code{sorted-lengths} list is less than the value of the top of the
15954 The simple version of the test works fine unless the
15955 @code{sorted-lengths} list has a @code{nil} value. In that case, the
15956 @code{(car sorted-lengths)} expression function returns
15957 @code{nil}. The @code{<} function cannot compare a number to
15958 @code{nil}, which is an empty list, so Emacs signals an error and
15959 stops the function from attempting to continue to execute.
15961 The @code{sorted-lengths} list always becomes @code{nil} when the
15962 counter reaches the end of the list. This means that any attempt to
15963 use the @code{defuns-per-range} function with the simple version of
15964 the test will fail.
15966 We solve the problem by using the @code{(car sorted-lengths)}
15967 expression in conjunction with the @code{and} expression. The
15968 @code{(car sorted-lengths)} expression returns a non-@code{nil}
15969 value so long as the list has at least one number within it, but
15970 returns @code{nil} if the list is empty. The @code{and} expression
15971 first evaluates the @code{(car sorted-lengths)} expression, and
15972 if it is @code{nil}, returns false @emph{without} evaluating the
15973 @code{<} expression. But if the @code{(car sorted-lengths)}
15974 expression returns a non-@code{nil} value, the @code{and} expression
15975 evaluates the @code{<} expression, and returns that value as the value
15976 of the @code{and} expression.
15978 @c colon in printed section title causes problem in Info cross reference
15979 This way, we avoid an error.
15982 (For information about @code{and}, see
15983 @ref{kill-new function, , The @code{kill-new} function}.)
15987 (@xref{kill-new function, , The @code{kill-new} function}, for
15988 information about @code{and}.)
15991 Here is a short test of the @code{defuns-per-range} function. First,
15992 evaluate the expression that binds (a shortened)
15993 @code{top-of-ranges} list to the list of values, then evaluate the
15994 expression for binding the @code{sorted-lengths} list, and then
15995 evaluate the @code{defuns-per-range} function.
15999 ;; @r{(Shorter list than we will use later.)}
16000 (setq top-of-ranges
16001 '(110 120 130 140 150
16002 160 170 180 190 200))
16004 (setq sorted-lengths
16005 '(85 86 110 116 122 129 154 176 179 200 265 300 300))
16007 (defuns-per-range sorted-lengths top-of-ranges)
16013 The list returned looks like this:
16016 (2 2 2 0 0 1 0 2 0 0 4)
16020 Indeed, there are two elements of the @code{sorted-lengths} list
16021 smaller than 110, two elements between 110 and 119, two elements
16022 between 120 and 129, and so on. There are four elements with a value
16025 @c The next step is to turn this numbers' list into a graph.
16026 @node Readying a Graph
16027 @chapter Readying a Graph
16028 @cindex Readying a graph
16029 @cindex Graph prototype
16030 @cindex Prototype graph
16031 @cindex Body of graph
16033 Our goal is to construct a graph showing the numbers of function
16034 definitions of various lengths in the Emacs lisp sources.
16036 As a practical matter, if you were creating a graph, you would
16037 probably use a program such as @code{gnuplot} to do the job.
16038 (@code{gnuplot} is nicely integrated into GNU Emacs.) In this case,
16039 however, we create one from scratch, and in the process we will
16040 re-acquaint ourselves with some of what we learned before and learn
16043 In this chapter, we will first write a simple graph printing function.
16044 This first definition will be a @dfn{prototype}, a rapidly written
16045 function that enables us to reconnoiter this unknown graph-making
16046 territory. We will discover dragons, or find that they are myth.
16047 After scouting the terrain, we will feel more confident and enhance
16048 the function to label the axes automatically.
16051 * Columns of a graph::
16052 * graph-body-print:: How to print the body of a graph.
16053 * recursive-graph-body-print::
16055 * Line Graph Exercise::
16059 @node Columns of a graph
16060 @unnumberedsec Printing the Columns of a Graph
16063 Since Emacs is designed to be flexible and work with all kinds of
16064 terminals, including character-only terminals, the graph will need to
16065 be made from one of the `typewriter' symbols. An asterisk will do; as
16066 we enhance the graph-printing function, we can make the choice of
16067 symbol a user option.
16069 We can call this function @code{graph-body-print}; it will take a
16070 @code{numbers-list} as its only argument. At this stage, we will not
16071 label the graph, but only print its body.
16073 The @code{graph-body-print} function inserts a vertical column of
16074 asterisks for each element in the @code{numbers-list}. The height of
16075 each line is determined by the value of that element of the
16076 @code{numbers-list}.
16078 Inserting columns is a repetitive act; that means that this function can
16079 be written either with a @code{while} loop or recursively.
16081 Our first challenge is to discover how to print a column of asterisks.
16082 Usually, in Emacs, we print characters onto a screen horizontally,
16083 line by line, by typing. We have two routes we can follow: write our
16084 own column-insertion function or discover whether one exists in Emacs.
16086 To see whether there is one in Emacs, we can use the @kbd{M-x apropos}
16087 command. This command is like the @kbd{C-h a} (@code{command-apropos})
16088 command, except that the latter finds only those functions that are
16089 commands. The @kbd{M-x apropos} command lists all symbols that match
16090 a regular expression, including functions that are not interactive.
16093 What we want to look for is some command that prints or inserts
16094 columns. Very likely, the name of the function will contain either
16095 the word `print' or the word `insert' or the word `column'.
16096 Therefore, we can simply type @kbd{M-x apropos RET
16097 print\|insert\|column RET} and look at the result. On my system, this
16098 command once too takes quite some time, and then produced a list of 79
16099 functions and variables. Now it does not take much time at all and
16100 produces a list of 211 functions and variables. Scanning down the
16101 list, the only function that looks as if it might do the job is
16102 @code{insert-rectangle}.
16105 Indeed, this is the function we want; its documentation says:
16110 Insert text of RECTANGLE with upper left corner at point.
16111 RECTANGLE's first line is inserted at point,
16112 its second line is inserted at a point vertically under point, etc.
16113 RECTANGLE should be a list of strings.
16114 After this command, the mark is at the upper left corner
16115 and point is at the lower right corner.
16119 We can run a quick test, to make sure it does what we expect of it.
16121 Here is the result of placing the cursor after the
16122 @code{insert-rectangle} expression and typing @kbd{C-u C-x C-e}
16123 (@code{eval-last-sexp}). The function inserts the strings
16124 @samp{"first"}, @samp{"second"}, and @samp{"third"} at and below
16125 point. Also the function returns @code{nil}.
16129 (insert-rectangle '("first" "second" "third"))first
16136 Of course, we won't be inserting the text of the
16137 @code{insert-rectangle} expression itself into the buffer in which we
16138 are making the graph, but will call the function from our program. We
16139 shall, however, have to make sure that point is in the buffer at the
16140 place where the @code{insert-rectangle} function will insert its
16143 If you are reading this in Info, you can see how this works by
16144 switching to another buffer, such as the @file{*scratch*} buffer,
16145 placing point somewhere in the buffer, typing @kbd{M-:}, typing the
16146 @code{insert-rectangle} expression into the minibuffer at the prompt,
16147 and then typing @key{RET}. This causes Emacs to evaluate the
16148 expression in the minibuffer, but to use as the value of point the
16149 position of point in the @file{*scratch*} buffer. (@kbd{M-:} is the
16150 keybinding for @code{eval-expression}. Also, @code{nil} does not
16151 appear in the @file{*scratch*} buffer since the expression is
16152 evaluated in the minibuffer.)
16154 We find when we do this that point ends up at the end of the last
16155 inserted line---that is to say, this function moves point as a
16156 side-effect. If we were to repeat the command, with point at this
16157 position, the next insertion would be below and to the right of the
16158 previous insertion. We don't want this! If we are going to make a
16159 bar graph, the columns need to be beside each other.
16161 So we discover that each cycle of the column-inserting @code{while}
16162 loop must reposition point to the place we want it, and that place
16163 will be at the top, not the bottom, of the column. Moreover, we
16164 remember that when we print a graph, we do not expect all the columns
16165 to be the same height. This means that the top of each column may be
16166 at a different height from the previous one. We cannot simply
16167 reposition point to the same line each time, but moved over to the
16168 right---or perhaps we can@dots{}
16170 We are planning to make the columns of the bar graph out of asterisks.
16171 The number of asterisks in the column is the number specified by the
16172 current element of the @code{numbers-list}. We need to construct a
16173 list of asterisks of the right length for each call to
16174 @code{insert-rectangle}. If this list consists solely of the requisite
16175 number of asterisks, then we will have position point the right number
16176 of lines above the base for the graph to print correctly. This could
16179 Alternatively, if we can figure out some way to pass
16180 @code{insert-rectangle} a list of the same length each time, then we
16181 can place point on the same line each time, but move it over one
16182 column to the right for each new column. If we do this, however, some
16183 of the entries in the list passed to @code{insert-rectangle} must be
16184 blanks rather than asterisks. For example, if the maximum height of
16185 the graph is 5, but the height of the column is 3, then
16186 @code{insert-rectangle} requires an argument that looks like this:
16189 (" " " " "*" "*" "*")
16192 This last proposal is not so difficult, so long as we can determine
16193 the column height. There are two ways for us to specify the column
16194 height: we can arbitrarily state what it will be, which would work
16195 fine for graphs of that height; or we can search through the list of
16196 numbers and use the maximum height of the list as the maximum height
16197 of the graph. If the latter operation were difficult, then the former
16198 procedure would be easiest, but there is a function built into Emacs
16199 that determines the maximum of its arguments. We can use that
16200 function. The function is called @code{max} and it returns the
16201 largest of all its arguments, which must be numbers. Thus, for
16209 returns 7. (A corresponding function called @code{min} returns the
16210 smallest of all its arguments.)
16214 However, we cannot simply call @code{max} on the @code{numbers-list};
16215 the @code{max} function expects numbers as its argument, not a list of
16216 numbers. Thus, the following expression,
16219 (max '(3 4 6 5 7 3))
16224 produces the following error message;
16227 Wrong type of argument: number-or-marker-p, (3 4 6 5 7 3)
16231 We need a function that passes a list of arguments to a function.
16232 This function is @code{apply}. This function `applies' its first
16233 argument (a function) to its remaining arguments, the last of which
16240 (apply 'max 3 4 7 3 '(4 8 5))
16246 (Incidentally, I don't know how you would learn of this function
16247 without a book such as this. It is possible to discover other
16248 functions, like @code{search-forward} or @code{insert-rectangle}, by
16249 guessing at a part of their names and then using @code{apropos}. Even
16250 though its base in metaphor is clear---`apply' its first argument to
16251 the rest---I doubt a novice would come up with that particular word
16252 when using @code{apropos} or other aid. Of course, I could be wrong;
16253 after all, the function was first named by someone who had to invent
16256 The second and subsequent arguments to @code{apply} are optional, so
16257 we can use @code{apply} to call a function and pass the elements of a
16258 list to it, like this, which also returns 8:
16261 (apply 'max '(4 8 5))
16264 This latter way is how we will use @code{apply}. The
16265 @code{recursive-lengths-list-many-files} function returns a numbers'
16266 list to which we can apply @code{max} (we could also apply @code{max} to
16267 the sorted numbers' list; it does not matter whether the list is
16271 Hence, the operation for finding the maximum height of the graph is this:
16274 (setq max-graph-height (apply 'max numbers-list))
16277 Now we can return to the question of how to create a list of strings
16278 for a column of the graph. Told the maximum height of the graph
16279 and the number of asterisks that should appear in the column, the
16280 function should return a list of strings for the
16281 @code{insert-rectangle} command to insert.
16283 Each column is made up of asterisks or blanks. Since the function is
16284 passed the value of the height of the column and the number of
16285 asterisks in the column, the number of blanks can be found by
16286 subtracting the number of asterisks from the height of the column.
16287 Given the number of blanks and the number of asterisks, two
16288 @code{while} loops can be used to construct the list:
16292 ;;; @r{First version.}
16293 (defun column-of-graph (max-graph-height actual-height)
16294 "Return list of strings that is one column of a graph."
16295 (let ((insert-list nil)
16296 (number-of-top-blanks
16297 (- max-graph-height actual-height)))
16301 ;; @r{Fill in asterisks.}
16302 (while (> actual-height 0)
16303 (setq insert-list (cons "*" insert-list))
16304 (setq actual-height (1- actual-height)))
16308 ;; @r{Fill in blanks.}
16309 (while (> number-of-top-blanks 0)
16310 (setq insert-list (cons " " insert-list))
16311 (setq number-of-top-blanks
16312 (1- number-of-top-blanks)))
16316 ;; @r{Return whole list.}
16321 If you install this function and then evaluate the following
16322 expression you will see that it returns the list as desired:
16325 (column-of-graph 5 3)
16333 (" " " " "*" "*" "*")
16336 As written, @code{column-of-graph} contains a major flaw: the symbols
16337 used for the blank and for the marked entries in the column are
16338 `hard-coded' as a space and asterisk. This is fine for a prototype,
16339 but you, or another user, may wish to use other symbols. For example,
16340 in testing the graph function, you many want to use a period in place
16341 of the space, to make sure the point is being repositioned properly
16342 each time the @code{insert-rectangle} function is called; or you might
16343 want to substitute a @samp{+} sign or other symbol for the asterisk.
16344 You might even want to make a graph-column that is more than one
16345 display column wide. The program should be more flexible. The way to
16346 do that is to replace the blank and the asterisk with two variables
16347 that we can call @code{graph-blank} and @code{graph-symbol} and define
16348 those variables separately.
16350 Also, the documentation is not well written. These considerations
16351 lead us to the second version of the function:
16355 (defvar graph-symbol "*"
16356 "String used as symbol in graph, usually an asterisk.")
16360 (defvar graph-blank " "
16361 "String used as blank in graph, usually a blank space.
16362 graph-blank must be the same number of columns wide
16368 (For an explanation of @code{defvar}, see
16369 @ref{defvar, , Initializing a Variable with @code{defvar}}.)
16373 ;;; @r{Second version.}
16374 (defun column-of-graph (max-graph-height actual-height)
16375 "Return MAX-GRAPH-HEIGHT strings; ACTUAL-HEIGHT are graph-symbols.
16379 The graph-symbols are contiguous entries at the end
16381 The list will be inserted as one column of a graph.
16382 The strings are either graph-blank or graph-symbol."
16386 (let ((insert-list nil)
16387 (number-of-top-blanks
16388 (- max-graph-height actual-height)))
16392 ;; @r{Fill in @code{graph-symbols}.}
16393 (while (> actual-height 0)
16394 (setq insert-list (cons graph-symbol insert-list))
16395 (setq actual-height (1- actual-height)))
16399 ;; @r{Fill in @code{graph-blanks}.}
16400 (while (> number-of-top-blanks 0)
16401 (setq insert-list (cons graph-blank insert-list))
16402 (setq number-of-top-blanks
16403 (1- number-of-top-blanks)))
16405 ;; @r{Return whole list.}
16410 If we wished, we could rewrite @code{column-of-graph} a third time to
16411 provide optionally for a line graph as well as for a bar graph. This
16412 would not be hard to do. One way to think of a line graph is that it
16413 is no more than a bar graph in which the part of each bar that is
16414 below the top is blank. To construct a column for a line graph, the
16415 function first constructs a list of blanks that is one shorter than
16416 the value, then it uses @code{cons} to attach a graph symbol to the
16417 list; then it uses @code{cons} again to attach the `top blanks' to
16420 It is easy to see how to write such a function, but since we don't
16421 need it, we will not do it. But the job could be done, and if it were
16422 done, it would be done with @code{column-of-graph}. Even more
16423 important, it is worth noting that few changes would have to be made
16424 anywhere else. The enhancement, if we ever wish to make it, is
16427 Now, finally, we come to our first actual graph printing function.
16428 This prints the body of a graph, not the labels for the vertical and
16429 horizontal axes, so we can call this @code{graph-body-print}.
16431 @node graph-body-print
16432 @section The @code{graph-body-print} Function
16433 @findex graph-body-print
16435 After our preparation in the preceding section, the
16436 @code{graph-body-print} function is straightforward. The function
16437 will print column after column of asterisks and blanks, using the
16438 elements of a numbers' list to specify the number of asterisks in each
16439 column. This is a repetitive act, which means we can use a
16440 decrementing @code{while} loop or recursive function for the job. In
16441 this section, we will write the definition using a @code{while} loop.
16443 The @code{column-of-graph} function requires the height of the graph
16444 as an argument, so we should determine and record that as a local variable.
16446 This leads us to the following template for the @code{while} loop
16447 version of this function:
16451 (defun graph-body-print (numbers-list)
16452 "@var{documentation}@dots{}"
16453 (let ((height @dots{}
16458 (while numbers-list
16459 @var{insert-columns-and-reposition-point}
16460 (setq numbers-list (cdr numbers-list)))))
16465 We need to fill in the slots of the template.
16467 Clearly, we can use the @code{(apply 'max numbers-list)} expression to
16468 determine the height of the graph.
16470 The @code{while} loop will cycle through the @code{numbers-list} one
16471 element at a time. As it is shortened by the @code{(setq numbers-list
16472 (cdr numbers-list))} expression, the @sc{car} of each instance of the
16473 list is the value of the argument for @code{column-of-graph}.
16475 At each cycle of the @code{while} loop, the @code{insert-rectangle}
16476 function inserts the list returned by @code{column-of-graph}. Since
16477 the @code{insert-rectangle} function moves point to the lower right of
16478 the inserted rectangle, we need to save the location of point at the
16479 time the rectangle is inserted, move back to that position after the
16480 rectangle is inserted, and then move horizontally to the next place
16481 from which @code{insert-rectangle} is called.
16483 If the inserted columns are one character wide, as they will be if
16484 single blanks and asterisks are used, the repositioning command is
16485 simply @code{(forward-char 1)}; however, the width of a column may be
16486 greater than one. This means that the repositioning command should be
16487 written @code{(forward-char symbol-width)}. The @code{symbol-width}
16488 itself is the length of a @code{graph-blank} and can be found using
16489 the expression @code{(length graph-blank)}. The best place to bind
16490 the @code{symbol-width} variable to the value of the width of graph
16491 column is in the varlist of the @code{let} expression.
16494 These considerations lead to the following function definition:
16498 (defun graph-body-print (numbers-list)
16499 "Print a bar graph of the NUMBERS-LIST.
16500 The numbers-list consists of the Y-axis values."
16502 (let ((height (apply 'max numbers-list))
16503 (symbol-width (length graph-blank))
16508 (while numbers-list
16509 (setq from-position (point))
16511 (column-of-graph height (car numbers-list)))
16512 (goto-char from-position)
16513 (forward-char symbol-width)
16516 ;; @r{Draw graph column by column.}
16518 (setq numbers-list (cdr numbers-list)))
16521 ;; @r{Place point for X axis labels.}
16522 (forward-line height)
16529 The one unexpected expression in this function is the
16530 @w{@code{(sit-for 0)}} expression in the @code{while} loop. This
16531 expression makes the graph printing operation more interesting to
16532 watch than it would be otherwise. The expression causes Emacs to
16533 `sit' or do nothing for a zero length of time and then redraw the
16534 screen. Placed here, it causes Emacs to redraw the screen column by
16535 column. Without it, Emacs would not redraw the screen until the
16538 We can test @code{graph-body-print} with a short list of numbers.
16542 Install @code{graph-symbol}, @code{graph-blank},
16543 @code{column-of-graph}, which are in
16545 @ref{Readying a Graph, , Readying a Graph},
16548 @ref{Columns of a graph},
16550 and @code{graph-body-print}.
16554 Copy the following expression:
16557 (graph-body-print '(1 2 3 4 6 4 3 5 7 6 5 2 3))
16561 Switch to the @file{*scratch*} buffer and place the cursor where you
16562 want the graph to start.
16565 Type @kbd{M-:} (@code{eval-expression}).
16568 Yank the @code{graph-body-print} expression into the minibuffer
16569 with @kbd{C-y} (@code{yank)}.
16572 Press @key{RET} to evaluate the @code{graph-body-print} expression.
16576 Emacs will print a graph like this:
16590 @node recursive-graph-body-print
16591 @section The @code{recursive-graph-body-print} Function
16592 @findex recursive-graph-body-print
16594 The @code{graph-body-print} function may also be written recursively.
16595 The recursive solution is divided into two parts: an outside `wrapper'
16596 that uses a @code{let} expression to determine the values of several
16597 variables that need only be found once, such as the maximum height of
16598 the graph, and an inside function that is called recursively to print
16602 The `wrapper' is uncomplicated:
16606 (defun recursive-graph-body-print (numbers-list)
16607 "Print a bar graph of the NUMBERS-LIST.
16608 The numbers-list consists of the Y-axis values."
16609 (let ((height (apply 'max numbers-list))
16610 (symbol-width (length graph-blank))
16612 (recursive-graph-body-print-internal
16619 The recursive function is a little more difficult. It has four parts:
16620 the `do-again-test', the printing code, the recursive call, and the
16621 `next-step-expression'. The `do-again-test' is a @code{when}
16622 expression that determines whether the @code{numbers-list} contains
16623 any remaining elements; if it does, the function prints one column of
16624 the graph using the printing code and calls itself again. The
16625 function calls itself again according to the value produced by the
16626 `next-step-expression' which causes the call to act on a shorter
16627 version of the @code{numbers-list}.
16631 (defun recursive-graph-body-print-internal
16632 (numbers-list height symbol-width)
16633 "Print a bar graph.
16634 Used within recursive-graph-body-print function."
16639 (setq from-position (point))
16641 (column-of-graph height (car numbers-list)))
16644 (goto-char from-position)
16645 (forward-char symbol-width)
16646 (sit-for 0) ; @r{Draw graph column by column.}
16647 (recursive-graph-body-print-internal
16648 (cdr numbers-list) height symbol-width)))
16653 After installation, this expression can be tested; here is a sample:
16656 (recursive-graph-body-print '(3 2 5 6 7 5 3 4 6 4 3 2 1))
16660 Here is what @code{recursive-graph-body-print} produces:
16674 Either of these two functions, @code{graph-body-print} or
16675 @code{recursive-graph-body-print}, create the body of a graph.
16678 @section Need for Printed Axes
16680 A graph needs printed axes, so you can orient yourself. For a do-once
16681 project, it may be reasonable to draw the axes by hand using Emacs's
16682 Picture mode; but a graph drawing function may be used more than once.
16684 For this reason, I have written enhancements to the basic
16685 @code{print-graph-body} function that automatically print labels for
16686 the horizontal and vertical axes. Since the label printing functions
16687 do not contain much new material, I have placed their description in
16688 an appendix. @xref{Full Graph, , A Graph with Labeled Axes}.
16690 @node Line Graph Exercise
16693 Write a line graph version of the graph printing functions.
16695 @node Emacs Initialization
16696 @chapter Your @file{.emacs} File
16697 @cindex @file{.emacs} file
16698 @cindex Customizing your @file{.emacs} file
16699 @cindex Initialization file
16701 ``You don't have to like Emacs to like it''---this seemingly
16702 paradoxical statement is the secret of GNU Emacs. The plain, `out of
16703 the box' Emacs is a generic tool. Most people who use it, customize
16704 it to suit themselves.
16706 GNU Emacs is mostly written in Emacs Lisp; this means that by writing
16707 expressions in Emacs Lisp you can change or extend Emacs.
16710 * Default Configuration::
16711 * Site-wide Init:: You can write site-wide init files.
16712 * defcustom:: Emacs will write code for you.
16713 * Beginning init File:: How to write a @file{.emacs} init file.
16714 * Text and Auto-fill:: Automatically wrap lines.
16715 * Mail Aliases:: Use abbreviations for email addresses.
16716 * Indent Tabs Mode:: Don't use tabs with @TeX{}
16717 * Keybindings:: Create some personal keybindings.
16718 * Keymaps:: More about key binding.
16719 * Loading Files:: Load (i.e., evaluate) files automatically.
16720 * Autoload:: Make functions available.
16721 * Simple Extension:: Define a function; bind it to a key.
16722 * X11 Colors:: Colors in X.
16724 * Mode Line:: How to customize your mode line.
16728 @node Default Configuration
16729 @unnumberedsec Emacs's Default Configuration
16732 There are those who appreciate Emacs's default configuration. After
16733 all, Emacs starts you in C mode when you edit a C file, starts you in
16734 Fortran mode when you edit a Fortran file, and starts you in
16735 Fundamental mode when you edit an unadorned file. This all makes
16736 sense, if you do not know who is going to use Emacs. Who knows what a
16737 person hopes to do with an unadorned file? Fundamental mode is the
16738 right default for such a file, just as C mode is the right default for
16739 editing C code. (Enough programming languages have syntaxes
16740 that enable them to share or nearly share features, so C mode is
16741 now provided by CC mode, the `C Collection'.)
16743 But when you do know who is going to use Emacs---you,
16744 yourself---then it makes sense to customize Emacs.
16746 For example, I seldom want Fundamental mode when I edit an
16747 otherwise undistinguished file; I want Text mode. This is why I
16748 customize Emacs: so it suits me.
16750 You can customize and extend Emacs by writing or adapting a
16751 @file{~/.emacs} file. This is your personal initialization file; its
16752 contents, written in Emacs Lisp, tell Emacs what to do.@footnote{You
16753 may also add @file{.el} to @file{~/.emacs} and call it a
16754 @file{~/.emacs.el} file. In the past, you were forbidden to type the
16755 extra keystrokes that the name @file{~/.emacs.el} requires, but now
16756 you may. The new format is consistent with the Emacs Lisp file
16757 naming conventions; the old format saves typing.}
16759 A @file{~/.emacs} file contains Emacs Lisp code. You can write this
16760 code yourself; or you can use Emacs's @code{customize} feature to write
16761 the code for you. You can combine your own expressions and
16762 auto-written Customize expressions in your @file{.emacs} file.
16764 (I myself prefer to write my own expressions, except for those,
16765 particularly fonts, that I find easier to manipulate using the
16766 @code{customize} command. I combine the two methods.)
16768 Most of this chapter is about writing expressions yourself. It
16769 describes a simple @file{.emacs} file; for more information, see
16770 @ref{Init File, , The Init File, emacs, The GNU Emacs Manual}, and
16771 @ref{Init File, , The Init File, elisp, The GNU Emacs Lisp Reference
16774 @node Site-wide Init
16775 @section Site-wide Initialization Files
16777 @cindex @file{default.el} init file
16778 @cindex @file{site-init.el} init file
16779 @cindex @file{site-load.el} init file
16780 In addition to your personal initialization file, Emacs automatically
16781 loads various site-wide initialization files, if they exist. These
16782 have the same form as your @file{.emacs} file, but are loaded by
16785 Two site-wide initialization files, @file{site-load.el} and
16786 @file{site-init.el}, are loaded into Emacs and then `dumped' if a
16787 `dumped' version of Emacs is created, as is most common. (Dumped
16788 copies of Emacs load more quickly. However, once a file is loaded and
16789 dumped, a change to it does not lead to a change in Emacs unless you
16790 load it yourself or re-dump Emacs. @xref{Building Emacs, , Building
16791 Emacs, elisp, The GNU Emacs Lisp Reference Manual}, and the
16792 @file{INSTALL} file.)
16794 Three other site-wide initialization files are loaded automatically
16795 each time you start Emacs, if they exist. These are
16796 @file{site-start.el}, which is loaded @emph{before} your @file{.emacs}
16797 file, and @file{default.el}, and the terminal type file, which are both
16798 loaded @emph{after} your @file{.emacs} file.
16800 Settings and definitions in your @file{.emacs} file will overwrite
16801 conflicting settings and definitions in a @file{site-start.el} file,
16802 if it exists; but the settings and definitions in a @file{default.el}
16803 or terminal type file will overwrite those in your @file{.emacs} file.
16804 (You can prevent interference from a terminal type file by setting
16805 @code{term-file-prefix} to @code{nil}. @xref{Simple Extension, , A
16806 Simple Extension}.)
16808 @c Rewritten to avoid overfull hbox.
16809 The @file{INSTALL} file that comes in the distribution contains
16810 descriptions of the @file{site-init.el} and @file{site-load.el} files.
16812 The @file{loadup.el}, @file{startup.el}, and @file{loaddefs.el} files
16813 control loading. These files are in the @file{lisp} directory of the
16814 Emacs distribution and are worth perusing.
16816 The @file{loaddefs.el} file contains a good many suggestions as to
16817 what to put into your own @file{.emacs} file, or into a site-wide
16818 initialization file.
16821 @section Specifying Variables using @code{defcustom}
16824 You can specify variables using @code{defcustom} so that you and
16825 others can then use Emacs's @code{customize} feature to set their
16826 values. (You cannot use @code{customize} to write function
16827 definitions; but you can write @code{defuns} in your @file{.emacs}
16828 file. Indeed, you can write any Lisp expression in your @file{.emacs}
16831 The @code{customize} feature depends on the @code{defcustom} macro.
16832 Although you can use @code{defvar} or @code{setq} for variables that
16833 users set, the @code{defcustom} macro is designed for the job.
16835 You can use your knowledge of @code{defvar} for writing the
16836 first three arguments for @code{defcustom}. The first argument to
16837 @code{defcustom} is the name of the variable. The second argument is
16838 the variable's initial value, if any; and this value is set only if
16839 the value has not already been set. The third argument is the
16842 The fourth and subsequent arguments to @code{defcustom} specify types
16843 and options; these are not featured in @code{defvar}. (These
16844 arguments are optional.)
16846 Each of these arguments consists of a keyword followed by a value.
16847 Each keyword starts with the colon character @samp{:}.
16850 For example, the customizable user option variable
16851 @code{text-mode-hook} looks like this:
16855 (defcustom text-mode-hook nil
16856 "Normal hook run when entering Text mode and many related modes."
16858 :options '(turn-on-auto-fill flyspell-mode)
16864 The name of the variable is @code{text-mode-hook}; it has no default
16865 value; and its documentation string tells you what it does.
16867 The @code{:type} keyword tells Emacs the kind of data to which
16868 @code{text-mode-hook} should be set and how to display the value in a
16869 Customization buffer.
16871 The @code{:options} keyword specifies a suggested list of values for
16872 the variable. Usually, @code{:options} applies to a hook.
16873 The list is only a suggestion; it is not exclusive; a person who sets
16874 the variable may set it to other values; the list shown following the
16875 @code{:options} keyword is intended to offer convenient choices to a
16878 Finally, the @code{:group} keyword tells the Emacs Customization
16879 command in which group the variable is located. This tells where to
16882 The @code{defcustom} macro recognizes more than a dozen keywords.
16883 For more information, see @ref{Customization, , Writing Customization
16884 Definitions, elisp, The GNU Emacs Lisp Reference Manual}.
16886 Consider @code{text-mode-hook} as an example.
16888 There are two ways to customize this variable. You can use the
16889 customization command or write the appropriate expressions yourself.
16892 Using the customization command, you can type:
16899 and find that the group for editing files of data is called `data'.
16900 Enter that group. Text Mode Hook is the first member. You can click
16901 on its various options, such as @code{turn-on-auto-fill}, to set the
16902 values. After you click on the button to
16905 Save for Future Sessions
16909 Emacs will write an expression into your @file{.emacs} file.
16910 It will look like this:
16914 (custom-set-variables
16915 ;; custom-set-variables was added by Custom.
16916 ;; If you edit it by hand, you could mess it up, so be careful.
16917 ;; Your init file should contain only one such instance.
16918 ;; If there is more than one, they won't work right.
16919 '(text-mode-hook (quote (turn-on-auto-fill text-mode-hook-identify))))
16924 (The @code{text-mode-hook-identify} function tells
16925 @code{toggle-text-mode-auto-fill} which buffers are in Text mode.
16926 It comes on automatically.)
16928 The @code{custom-set-variables} function works somewhat differently
16929 than a @code{setq}. While I have never learned the differences, I
16930 modify the @code{custom-set-variables} expressions in my @file{.emacs}
16931 file by hand: I make the changes in what appears to me to be a
16932 reasonable manner and have not had any problems. Others prefer to use
16933 the Customization command and let Emacs do the work for them.
16935 Another @code{custom-set-@dots{}} function is @code{custom-set-faces}.
16936 This function sets the various font faces. Over time, I have set a
16937 considerable number of faces. Some of the time, I re-set them using
16938 @code{customize}; other times, I simply edit the
16939 @code{custom-set-faces} expression in my @file{.emacs} file itself.
16941 The second way to customize your @code{text-mode-hook} is to set it
16942 yourself in your @file{.emacs} file using code that has nothing to do
16943 with the @code{custom-set-@dots{}} functions.
16946 When you do this, and later use @code{customize}, you will see a
16950 CHANGED outside Customize; operating on it here may be unreliable.
16954 This message is only a warning. If you click on the button to
16957 Save for Future Sessions
16961 Emacs will write a @code{custom-set-@dots{}} expression near the end
16962 of your @file{.emacs} file that will be evaluated after your
16963 hand-written expression. It will, therefore, overrule your
16964 hand-written expression. No harm will be done. When you do this,
16965 however, be careful to remember which expression is active; if you
16966 forget, you may confuse yourself.
16968 So long as you remember where the values are set, you will have no
16969 trouble. In any event, the values are always set in your
16970 initialization file, which is usually called @file{.emacs}.
16972 I myself use @code{customize} for hardly anything. Mostly, I write
16973 expressions myself.
16977 Incidentally, to be more complete concerning defines: @code{defsubst}
16978 defines an inline function. The syntax is just like that of
16979 @code{defun}. @code{defconst} defines a symbol as a constant. The
16980 intent is that neither programs nor users should ever change a value
16981 set by @code{defconst}. (You can change it; the value set is a
16982 variable; but please do not.)
16984 @node Beginning init File
16985 @section Beginning a @file{.emacs} File
16986 @cindex @file{.emacs} file, beginning of
16988 When you start Emacs, it loads your @file{.emacs} file unless you tell
16989 it not to by specifying @samp{-q} on the command line. (The
16990 @code{emacs -q} command gives you a plain, out-of-the-box Emacs.)
16992 A @file{.emacs} file contains Lisp expressions. Often, these are no
16993 more than expressions to set values; sometimes they are function
16996 @xref{Init File, , The Init File @file{~/.emacs}, emacs, The GNU Emacs
16997 Manual}, for a short description of initialization files.
16999 This chapter goes over some of the same ground, but is a walk among
17000 extracts from a complete, long-used @file{.emacs} file---my own.
17002 The first part of the file consists of comments: reminders to myself.
17003 By now, of course, I remember these things, but when I started, I did
17009 ;;;; Bob's .emacs file
17010 ; Robert J. Chassell
17011 ; 26 September 1985
17016 Look at that date! I started this file a long time ago. I have been
17017 adding to it ever since.
17021 ; Each section in this file is introduced by a
17022 ; line beginning with four semicolons; and each
17023 ; entry is introduced by a line beginning with
17024 ; three semicolons.
17029 This describes the usual conventions for comments in Emacs Lisp.
17030 Everything on a line that follows a semicolon is a comment. Two,
17031 three, and four semicolons are used as subsection and section markers.
17032 (@xref{Comments, ,, elisp, The GNU Emacs Lisp Reference Manual}, for
17033 more about comments.)
17038 ; Control-h is the help key;
17039 ; after typing control-h, type a letter to
17040 ; indicate the subject about which you want help.
17041 ; For an explanation of the help facility,
17042 ; type control-h two times in a row.
17047 Just remember: type @kbd{C-h} two times for help.
17051 ; To find out about any mode, type control-h m
17052 ; while in that mode. For example, to find out
17053 ; about mail mode, enter mail mode and then type
17059 `Mode help', as I call this, is very helpful. Usually, it tells you
17060 all you need to know.
17062 Of course, you don't need to include comments like these in your
17063 @file{.emacs} file. I included them in mine because I kept forgetting
17064 about Mode help or the conventions for comments---but I was able to
17065 remember to look here to remind myself.
17067 @node Text and Auto-fill
17068 @section Text and Auto Fill Mode
17070 Now we come to the part that `turns on' Text mode and
17075 ;;; Text mode and Auto Fill mode
17076 ;; The next two lines put Emacs into Text mode
17077 ;; and Auto Fill mode, and are for writers who
17078 ;; want to start writing prose rather than code.
17079 (setq-default major-mode 'text-mode)
17080 (add-hook 'text-mode-hook 'turn-on-auto-fill)
17084 Here is the first part of this @file{.emacs} file that does something
17085 besides remind a forgetful human!
17087 The first of the two lines in parentheses tells Emacs to turn on Text
17088 mode when you find a file, @emph{unless} that file should go into some
17089 other mode, such as C mode.
17091 @cindex Per-buffer, local variables list
17092 @cindex Local variables list, per-buffer,
17093 @cindex Automatic mode selection
17094 @cindex Mode selection, automatic
17095 When Emacs reads a file, it looks at the extension to the file name,
17096 if any. (The extension is the part that comes after a @samp{.}.) If
17097 the file ends with a @samp{.c} or @samp{.h} extension then Emacs turns
17098 on C mode. Also, Emacs looks at first nonblank line of the file; if
17099 the line says @w{@samp{-*- C -*-}}, Emacs turns on C mode. Emacs
17100 possesses a list of extensions and specifications that it uses
17101 automatically. In addition, Emacs looks near the last page for a
17102 per-buffer, ``local variables list'', if any.
17105 @xref{Choosing Modes, , How Major Modes are Chosen, emacs, The GNU
17108 @xref{File Variables, , Local Variables in Files, emacs, The GNU Emacs
17112 See sections ``How Major Modes are Chosen'' and ``Local Variables in
17113 Files'' in @cite{The GNU Emacs Manual}.
17116 Now, back to the @file{.emacs} file.
17119 Here is the line again; how does it work?
17121 @cindex Text Mode turned on
17123 (setq major-mode 'text-mode)
17127 This line is a short, but complete Emacs Lisp expression.
17129 We are already familiar with @code{setq}. It sets the following variable,
17130 @code{major-mode}, to the subsequent value, which is @code{text-mode}.
17131 The single quote mark before @code{text-mode} tells Emacs to deal directly
17132 with the @code{text-mode} symbol, not with whatever it might stand for.
17133 @xref{set & setq, , Setting the Value of a Variable},
17134 for a reminder of how @code{setq} works.
17135 The main point is that there is no difference between the procedure you
17136 use to set a value in your @file{.emacs} file and the procedure you use
17137 anywhere else in Emacs.
17140 Here is the next line:
17142 @cindex Auto Fill mode turned on
17145 (add-hook 'text-mode-hook 'turn-on-auto-fill)
17149 In this line, the @code{add-hook} command adds
17150 @code{turn-on-auto-fill} to the variable.
17152 @code{turn-on-auto-fill} is the name of a program, that, you guessed
17153 it!, turns on Auto Fill mode.
17155 Every time Emacs turns on Text mode, Emacs runs the commands `hooked'
17156 onto Text mode. So every time Emacs turns on Text mode, Emacs also
17157 turns on Auto Fill mode.
17159 In brief, the first line causes Emacs to enter Text mode when you edit a
17160 file, unless the file name extension, a first non-blank line, or local
17161 variables to tell Emacs otherwise.
17163 Text mode among other actions, sets the syntax table to work
17164 conveniently for writers. In Text mode, Emacs considers an apostrophe
17165 as part of a word like a letter; but Emacs does not consider a period
17166 or a space as part of a word. Thus, @kbd{M-f} moves you over
17167 @samp{it's}. On the other hand, in C mode, @kbd{M-f} stops just after
17168 the @samp{t} of @samp{it's}.
17170 The second line causes Emacs to turn on Auto Fill mode when it turns
17171 on Text mode. In Auto Fill mode, Emacs automatically breaks a line
17172 that is too wide and brings the excessively wide part of the line down
17173 to the next line. Emacs breaks lines between words, not within them.
17175 When Auto Fill mode is turned off, lines continue to the right as you
17176 type them. Depending on how you set the value of
17177 @code{truncate-lines}, the words you type either disappear off the
17178 right side of the screen, or else are shown, in a rather ugly and
17179 unreadable manner, as a continuation line on the screen.
17182 In addition, in this part of my @file{.emacs} file, I tell the Emacs
17183 fill commands to insert two spaces after a colon:
17186 (setq colon-double-space t)
17190 @section Mail Aliases
17192 Here is a @code{setq} that `turns on' mail aliases, along with more
17198 ; To enter mail mode, type `C-x m'
17199 ; To enter RMAIL (for reading mail),
17201 (setq mail-aliases t)
17205 @cindex Mail aliases
17207 This @code{setq} command sets the value of the variable
17208 @code{mail-aliases} to @code{t}. Since @code{t} means true, the line
17209 says, in effect, ``Yes, use mail aliases.''
17211 Mail aliases are convenient short names for long email addresses or
17212 for lists of email addresses. The file where you keep your `aliases'
17213 is @file{~/.mailrc}. You write an alias like this:
17216 alias geo george@@foobar.wiz.edu
17220 When you write a message to George, address it to @samp{geo}; the
17221 mailer will automatically expand @samp{geo} to the full address.
17223 @node Indent Tabs Mode
17224 @section Indent Tabs Mode
17225 @cindex Tabs, preventing
17226 @findex indent-tabs-mode
17228 By default, Emacs inserts tabs in place of multiple spaces when it
17229 formats a region. (For example, you might indent many lines of text
17230 all at once with the @code{indent-region} command.) Tabs look fine on
17231 a terminal or with ordinary printing, but they produce badly indented
17232 output when you use @TeX{} or Texinfo since @TeX{} ignores tabs.
17235 The following turns off Indent Tabs mode:
17239 ;;; Prevent Extraneous Tabs
17240 (setq-default indent-tabs-mode nil)
17244 Note that this line uses @code{setq-default} rather than the
17245 @code{setq} command that we have seen before. The @code{setq-default}
17246 command sets values only in buffers that do not have their own local
17247 values for the variable.
17250 @xref{Just Spaces, , Tabs vs. Spaces, emacs, The GNU Emacs Manual}.
17252 @xref{File Variables, , Local Variables in Files, emacs, The GNU Emacs
17256 See sections ``Tabs vs.@: Spaces'' and ``Local Variables in
17257 Files'' in @cite{The GNU Emacs Manual}.
17262 @section Some Keybindings
17264 Now for some personal keybindings:
17268 ;;; Compare windows
17269 (global-set-key "\C-cw" 'compare-windows)
17273 @findex compare-windows
17274 @code{compare-windows} is a nifty command that compares the text in
17275 your current window with text in the next window. It makes the
17276 comparison by starting at point in each window, moving over text in
17277 each window as far as they match. I use this command all the time.
17279 This also shows how to set a key globally, for all modes.
17281 @cindex Setting a key globally
17282 @cindex Global set key
17283 @cindex Key setting globally
17284 @findex global-set-key
17285 The command is @code{global-set-key}. It is followed by the
17286 keybinding. In a @file{.emacs} file, the keybinding is written as
17287 shown: @code{\C-c} stands for `control-c', which means `press the
17288 control key and the @key{c} key at the same time'. The @code{w} means
17289 `press the @key{w} key'. The keybinding is surrounded by double
17290 quotation marks. In documentation, you would write this as
17291 @w{@kbd{C-c w}}. (If you were binding a @key{META} key, such as
17292 @kbd{M-c}, rather than a @key{CTRL} key, you would write
17293 @w{@code{\M-c}} in your @file{.emacs} file. @xref{Init Rebinding, ,
17294 Rebinding Keys in Your Init File, emacs, The GNU Emacs Manual}, for
17297 The command invoked by the keys is @code{compare-windows}. Note that
17298 @code{compare-windows} is preceded by a single quote; otherwise, Emacs
17299 would first try to evaluate the symbol to determine its value.
17301 These three things, the double quotation marks, the backslash before
17302 the @samp{C}, and the single quote mark are necessary parts of
17303 keybinding that I tend to forget. Fortunately, I have come to
17304 remember that I should look at my existing @file{.emacs} file, and
17305 adapt what is there.
17307 As for the keybinding itself: @kbd{C-c w}. This combines the prefix
17308 key, @kbd{C-c}, with a single character, in this case, @kbd{w}. This
17309 set of keys, @kbd{C-c} followed by a single character, is strictly
17310 reserved for individuals' own use. (I call these `own' keys, since
17311 these are for my own use.) You should always be able to create such a
17312 keybinding for your own use without stomping on someone else's
17313 keybinding. If you ever write an extension to Emacs, please avoid
17314 taking any of these keys for public use. Create a key like @kbd{C-c
17315 C-w} instead. Otherwise, we will run out of `own' keys.
17318 Here is another keybinding, with a comment:
17322 ;;; Keybinding for `occur'
17323 ; I use occur a lot, so let's bind it to a key:
17324 (global-set-key "\C-co" 'occur)
17329 The @code{occur} command shows all the lines in the current buffer
17330 that contain a match for a regular expression. Matching lines are
17331 shown in a buffer called @file{*Occur*}. That buffer serves as a menu
17332 to jump to occurrences.
17334 @findex global-unset-key
17335 @cindex Unbinding key
17336 @cindex Key unbinding
17338 Here is how to unbind a key, so it does not
17344 (global-unset-key "\C-xf")
17348 There is a reason for this unbinding: I found I inadvertently typed
17349 @w{@kbd{C-x f}} when I meant to type @kbd{C-x C-f}. Rather than find a
17350 file, as I intended, I accidentally set the width for filled text,
17351 almost always to a width I did not want. Since I hardly ever reset my
17352 default width, I simply unbound the key.
17354 @findex list-buffers, @r{rebound}
17355 @findex buffer-menu, @r{bound to key}
17357 The following rebinds an existing key:
17361 ;;; Rebind `C-x C-b' for `buffer-menu'
17362 (global-set-key "\C-x\C-b" 'buffer-menu)
17366 By default, @kbd{C-x C-b} runs the
17367 @code{list-buffers} command. This command lists
17368 your buffers in @emph{another} window. Since I
17369 almost always want to do something in that
17370 window, I prefer the @code{buffer-menu}
17371 command, which not only lists the buffers,
17372 but moves point into that window.
17377 @cindex Rebinding keys
17379 Emacs uses @dfn{keymaps} to record which keys call which commands.
17380 When you use @code{global-set-key} to set the keybinding for a single
17381 command in all parts of Emacs, you are specifying the keybinding in
17382 @code{current-global-map}.
17384 Specific modes, such as C mode or Text mode, have their own keymaps;
17385 the mode-specific keymaps override the global map that is shared by
17388 The @code{global-set-key} function binds, or rebinds, the global
17389 keymap. For example, the following binds the key @kbd{C-x C-b} to the
17390 function @code{buffer-menu}:
17393 (global-set-key "\C-x\C-b" 'buffer-menu)
17396 Mode-specific keymaps are bound using the @code{define-key} function,
17397 which takes a specific keymap as an argument, as well as the key and
17398 the command. For example, my @file{.emacs} file contains the
17399 following expression to bind the @code{texinfo-insert-@@group} command
17400 to @kbd{C-c C-c g}:
17404 (define-key texinfo-mode-map "\C-c\C-cg" 'texinfo-insert-@@group)
17409 The @code{texinfo-insert-@@group} function itself is a little extension
17410 to Texinfo mode that inserts @samp{@@group} into a Texinfo file. I
17411 use this command all the time and prefer to type the three strokes
17412 @kbd{C-c C-c g} rather than the six strokes @kbd{@@ g r o u p}.
17413 (@samp{@@group} and its matching @samp{@@end group} are commands that
17414 keep all enclosed text together on one page; many multi-line examples
17415 in this book are surrounded by @samp{@@group @dots{} @@end group}.)
17418 Here is the @code{texinfo-insert-@@group} function definition:
17422 (defun texinfo-insert-@@group ()
17423 "Insert the string @@group in a Texinfo buffer."
17425 (beginning-of-line)
17426 (insert "@@group\n"))
17430 (Of course, I could have used Abbrev mode to save typing, rather than
17431 write a function to insert a word; but I prefer key strokes consistent
17432 with other Texinfo mode key bindings.)
17434 You will see numerous @code{define-key} expressions in
17435 @file{loaddefs.el} as well as in the various mode libraries, such as
17436 @file{cc-mode.el} and @file{lisp-mode.el}.
17438 @xref{Key Bindings, , Customizing Key Bindings, emacs, The GNU Emacs
17439 Manual}, and @ref{Keymaps, , Keymaps, elisp, The GNU Emacs Lisp
17440 Reference Manual}, for more information about keymaps.
17442 @node Loading Files
17443 @section Loading Files
17444 @cindex Loading files
17447 Many people in the GNU Emacs community have written extensions to
17448 Emacs. As time goes by, these extensions are often included in new
17449 releases. For example, the Calendar and Diary packages are now part
17450 of the standard GNU Emacs, as is Calc.
17452 You can use a @code{load} command to evaluate a complete file and
17453 thereby install all the functions and variables in the file into Emacs.
17456 @c (auto-compression-mode t)
17459 (load "~/emacs/slowsplit")
17462 This evaluates, i.e., loads, the @file{slowsplit.el} file or if it
17463 exists, the faster, byte compiled @file{slowsplit.elc} file from the
17464 @file{emacs} sub-directory of your home directory. The file contains
17465 the function @code{split-window-quietly}, which John Robinson wrote in
17468 The @code{split-window-quietly} function splits a window with the
17469 minimum of redisplay. I installed it in 1989 because it worked well
17470 with the slow 1200 baud terminals I was then using. Nowadays, I only
17471 occasionally come across such a slow connection, but I continue to use
17472 the function because I like the way it leaves the bottom half of a
17473 buffer in the lower of the new windows and the top half in the upper
17477 To replace the key binding for the default
17478 @code{split-window-vertically}, you must also unset that key and bind
17479 the keys to @code{split-window-quietly}, like this:
17483 (global-unset-key "\C-x2")
17484 (global-set-key "\C-x2" 'split-window-quietly)
17489 If you load many extensions, as I do, then instead of specifying the
17490 exact location of the extension file, as shown above, you can specify
17491 that directory as part of Emacs's @code{load-path}. Then, when Emacs
17492 loads a file, it will search that directory as well as its default
17493 list of directories. (The default list is specified in @file{paths.h}
17494 when Emacs is built.)
17497 The following command adds your @file{~/emacs} directory to the
17498 existing load path:
17502 ;;; Emacs Load Path
17503 (setq load-path (cons "~/emacs" load-path))
17507 Incidentally, @code{load-library} is an interactive interface to the
17508 @code{load} function. The complete function looks like this:
17510 @findex load-library
17513 (defun load-library (library)
17514 "Load the library named LIBRARY.
17515 This is an interface to the function `load'."
17517 (list (completing-read "Load library: "
17518 (apply-partially 'locate-file-completion-table
17520 (get-load-suffixes)))))
17525 The name of the function, @code{load-library}, comes from the use of
17526 `library' as a conventional synonym for `file'. The source for the
17527 @code{load-library} command is in the @file{files.el} library.
17529 Another interactive command that does a slightly different job is
17530 @code{load-file}. @xref{Lisp Libraries, , Libraries of Lisp Code for
17531 Emacs, emacs, The GNU Emacs Manual}, for information on the
17532 distinction between @code{load-library} and this command.
17535 @section Autoloading
17538 Instead of installing a function by loading the file that contains it,
17539 or by evaluating the function definition, you can make the function
17540 available but not actually install it until it is first called. This
17541 is called @dfn{autoloading}.
17543 When you execute an autoloaded function, Emacs automatically evaluates
17544 the file that contains the definition, and then calls the function.
17546 Emacs starts quicker with autoloaded functions, since their libraries
17547 are not loaded right away; but you need to wait a moment when you
17548 first use such a function, while its containing file is evaluated.
17550 Rarely used functions are frequently autoloaded. The
17551 @file{loaddefs.el} library contains hundreds of autoloaded functions,
17552 from @code{bookmark-set} to @code{wordstar-mode}. Of course, you may
17553 come to use a `rare' function frequently. When you do, you should
17554 load that function's file with a @code{load} expression in your
17555 @file{.emacs} file.
17557 In my @file{.emacs} file, I load 14 libraries that contain functions
17558 that would otherwise be autoloaded. (Actually, it would have been
17559 better to include these files in my `dumped' Emacs, but I forgot.
17560 @xref{Building Emacs, , Building Emacs, elisp, The GNU Emacs Lisp
17561 Reference Manual}, and the @file{INSTALL} file for more about
17564 You may also want to include autoloaded expressions in your @file{.emacs}
17565 file. @code{autoload} is a built-in function that takes up to five
17566 arguments, the final three of which are optional. The first argument
17567 is the name of the function to be autoloaded; the second is the name
17568 of the file to be loaded. The third argument is documentation for the
17569 function, and the fourth tells whether the function can be called
17570 interactively. The fifth argument tells what type of
17571 object---@code{autoload} can handle a keymap or macro as well as a
17572 function (the default is a function).
17575 Here is a typical example:
17579 (autoload 'html-helper-mode
17580 "html-helper-mode" "Edit HTML documents" t)
17585 (@code{html-helper-mode} is an older alternative to @code{html-mode},
17586 which is a standard part of the distribution.)
17589 This expression autoloads the @code{html-helper-mode} function. It
17590 takes it from the @file{html-helper-mode.el} file (or from the byte
17591 compiled version @file{html-helper-mode.elc}, if that exists.) The
17592 file must be located in a directory specified by @code{load-path}.
17593 The documentation says that this is a mode to help you edit documents
17594 written in the HyperText Markup Language. You can call this mode
17595 interactively by typing @kbd{M-x html-helper-mode}. (You need to
17596 duplicate the function's regular documentation in the autoload
17597 expression because the regular function is not yet loaded, so its
17598 documentation is not available.)
17600 @xref{Autoload, , Autoload, elisp, The GNU Emacs Lisp Reference
17601 Manual}, for more information.
17603 @node Simple Extension
17604 @section A Simple Extension: @code{line-to-top-of-window}
17605 @findex line-to-top-of-window
17606 @cindex Simple extension in @file{.emacs} file
17608 Here is a simple extension to Emacs that moves the line point is on to
17609 the top of the window. I use this all the time, to make text easier
17612 You can put the following code into a separate file and then load it
17613 from your @file{.emacs} file, or you can include it within your
17614 @file{.emacs} file.
17617 Here is the definition:
17621 ;;; Line to top of window;
17622 ;;; replace three keystroke sequence C-u 0 C-l
17623 (defun line-to-top-of-window ()
17624 "Move the line point is on to top of window."
17631 Now for the keybinding.
17633 Nowadays, function keys as well as mouse button events and
17634 non-@sc{ascii} characters are written within square brackets, without
17635 quotation marks. (In Emacs version 18 and before, you had to write
17636 different function key bindings for each different make of terminal.)
17638 I bind @code{line-to-top-of-window} to my @key{F6} function key like
17642 (global-set-key [f6] 'line-to-top-of-window)
17645 For more information, see @ref{Init Rebinding, , Rebinding Keys in
17646 Your Init File, emacs, The GNU Emacs Manual}.
17648 @cindex Conditional 'twixt two versions of Emacs
17649 @cindex Version of Emacs, choosing
17650 @cindex Emacs version, choosing
17651 If you run two versions of GNU Emacs, such as versions 22 and 23, and
17652 use one @file{.emacs} file, you can select which code to evaluate with
17653 the following conditional:
17658 ((= 22 emacs-major-version)
17659 ;; evaluate version 22 code
17661 ((= 23 emacs-major-version)
17662 ;; evaluate version 23 code
17667 For example, recent versions blink
17668 their cursors by default. I hate such blinking, as well as other
17669 features, so I placed the following in my @file{.emacs}
17670 file@footnote{When I start instances of Emacs that do not load my
17671 @file{.emacs} file or any site file, I also turn off blinking:
17674 emacs -q --no-site-file -eval '(blink-cursor-mode nil)'
17676 @exdent Or nowadays, using an even more sophisticated set of options,
17684 (when (>= emacs-major-version 21)
17685 (blink-cursor-mode 0)
17686 ;; Insert newline when you press `C-n' (next-line)
17687 ;; at the end of the buffer
17688 (setq next-line-add-newlines t)
17691 ;; Turn on image viewing
17692 (auto-image-file-mode t)
17695 ;; Turn on menu bar (this bar has text)
17696 ;; (Use numeric argument to turn on)
17700 ;; Turn off tool bar (this bar has icons)
17701 ;; (Use numeric argument to turn on)
17702 (tool-bar-mode nil)
17705 ;; Turn off tooltip mode for tool bar
17706 ;; (This mode causes icon explanations to pop up)
17707 ;; (Use numeric argument to turn on)
17709 ;; If tooltips turned on, make tips appear promptly
17710 (setq tooltip-delay 0.1) ; default is 0.7 second
17716 @section X11 Colors
17718 You can specify colors when you use Emacs with the MIT X Windowing
17721 I dislike the default colors and specify my own.
17724 Here are the expressions in my @file{.emacs}
17725 file that set values:
17729 ;; Set cursor color
17730 (set-cursor-color "white")
17733 (set-mouse-color "white")
17735 ;; Set foreground and background
17736 (set-foreground-color "white")
17737 (set-background-color "darkblue")
17741 ;;; Set highlighting colors for isearch and drag
17742 (set-face-foreground 'highlight "white")
17743 (set-face-background 'highlight "blue")
17747 (set-face-foreground 'region "cyan")
17748 (set-face-background 'region "blue")
17752 (set-face-foreground 'secondary-selection "skyblue")
17753 (set-face-background 'secondary-selection "darkblue")
17757 ;; Set calendar highlighting colors
17758 (setq calendar-load-hook
17760 (set-face-foreground 'diary-face "skyblue")
17761 (set-face-background 'holiday-face "slate blue")
17762 (set-face-foreground 'holiday-face "white")))
17766 The various shades of blue soothe my eye and prevent me from seeing
17767 the screen flicker.
17769 Alternatively, I could have set my specifications in various X
17770 initialization files. For example, I could set the foreground,
17771 background, cursor, and pointer (i.e., mouse) colors in my
17772 @file{~/.Xresources} file like this:
17776 Emacs*foreground: white
17777 Emacs*background: darkblue
17778 Emacs*cursorColor: white
17779 Emacs*pointerColor: white
17783 In any event, since it is not part of Emacs, I set the root color of
17784 my X window in my @file{~/.xinitrc} file, like this@footnote{I also
17785 run more modern window managers, such as Enlightenment, Gnome, or KDE;
17786 in those cases, I often specify an image rather than a plain color.}:
17789 xsetroot -solid Navy -fg white &
17793 @node Miscellaneous
17794 @section Miscellaneous Settings for a @file{.emacs} File
17797 Here are a few miscellaneous settings:
17802 Set the shape and color of the mouse cursor:
17806 ; Cursor shapes are defined in
17807 ; `/usr/include/X11/cursorfont.h';
17808 ; for example, the `target' cursor is number 128;
17809 ; the `top_left_arrow' cursor is number 132.
17813 (let ((mpointer (x-get-resource "*mpointer"
17814 "*emacs*mpointer")))
17815 ;; If you have not set your mouse pointer
17816 ;; then set it, otherwise leave as is:
17817 (if (eq mpointer nil)
17818 (setq mpointer "132")) ; top_left_arrow
17821 (setq x-pointer-shape (string-to-int mpointer))
17822 (set-mouse-color "white"))
17827 Or you can set the values of a variety of features in an alist, like
17833 default-frame-alist
17834 '((cursor-color . "white")
17835 (mouse-color . "white")
17836 (foreground-color . "white")
17837 (background-color . "DodgerBlue4")
17838 ;; (cursor-type . bar)
17839 (cursor-type . box)
17842 (tool-bar-lines . 0)
17843 (menu-bar-lines . 1)
17847 "-Misc-Fixed-Medium-R-Normal--20-200-75-75-C-100-ISO8859-1")
17853 Convert @kbd{@key{CTRL}-h} into @key{DEL} and @key{DEL}
17854 into @kbd{@key{CTRL}-h}.@*
17855 (Some older keyboards needed this, although I have not seen the
17860 ;; Translate `C-h' to <DEL>.
17861 ; (keyboard-translate ?\C-h ?\C-?)
17863 ;; Translate <DEL> to `C-h'.
17864 (keyboard-translate ?\C-? ?\C-h)
17868 @item Turn off a blinking cursor!
17872 (if (fboundp 'blink-cursor-mode)
17873 (blink-cursor-mode -1))
17878 or start GNU Emacs with the command @code{emacs -nbc}.
17881 @item When using `grep'@*
17882 @samp{-i}@w{ } Ignore case distinctions@*
17883 @samp{-n}@w{ } Prefix each line of output with line number@*
17884 @samp{-H}@w{ } Print the filename for each match.@*
17885 @samp{-e}@w{ } Protect patterns beginning with a hyphen character, @samp{-}
17888 (setq grep-command "grep -i -nH -e ")
17892 @c Evidently, no longer needed in GNU Emacs 22
17894 item Automatically uncompress compressed files when visiting them
17897 (load "uncompress")
17902 @item Find an existing buffer, even if it has a different name@*
17903 This avoids problems with symbolic links.
17906 (setq find-file-existing-other-name t)
17909 @item Set your language environment and default input method
17913 (set-language-environment "latin-1")
17914 ;; Remember you can enable or disable multilingual text input
17915 ;; with the @code{toggle-input-method'} (@kbd{C-\}) command
17916 (setq default-input-method "latin-1-prefix")
17920 If you want to write with Chinese `GB' characters, set this instead:
17924 (set-language-environment "Chinese-GB")
17925 (setq default-input-method "chinese-tonepy")
17930 @subsubheading Fixing Unpleasant Key Bindings
17931 @cindex Key bindings, fixing
17932 @cindex Bindings, key, fixing unpleasant
17934 Some systems bind keys unpleasantly. Sometimes, for example, the
17935 @key{CTRL} key appears in an awkward spot rather than at the far left
17938 Usually, when people fix these sorts of keybindings, they do not
17939 change their @file{~/.emacs} file. Instead, they bind the proper keys
17940 on their consoles with the @code{loadkeys} or @code{install-keymap}
17941 commands in their boot script and then include @code{xmodmap} commands
17942 in their @file{.xinitrc} or @file{.Xsession} file for X Windows.
17950 loadkeys /usr/share/keymaps/i386/qwerty/emacs2.kmap.gz
17952 install-keymap emacs2
17958 For a @file{.xinitrc} or @file{.Xsession} file when the @key{Caps
17959 Lock} key is at the far left of the home row:
17963 # Bind the key labeled `Caps Lock' to `Control'
17964 # (Such a broken user interface suggests that keyboard manufacturers
17965 # think that computers are typewriters from 1885.)
17967 xmodmap -e "clear Lock"
17968 xmodmap -e "add Control = Caps_Lock"
17974 In a @file{.xinitrc} or @file{.Xsession} file, to convert an @key{ALT}
17975 key to a @key{META} key:
17979 # Some ill designed keyboards have a key labeled ALT and no Meta
17980 xmodmap -e "keysym Alt_L = Meta_L Alt_L"
17986 @section A Modified Mode Line
17987 @vindex mode-line-format
17988 @cindex Mode line format
17990 Finally, a feature I really like: a modified mode line.
17992 When I work over a network, I forget which machine I am using. Also,
17993 I tend to I lose track of where I am, and which line point is on.
17995 So I reset my mode line to look like this:
17998 -:-- foo.texi rattlesnake:/home/bob/ Line 1 (Texinfo Fill) Top
18001 I am visiting a file called @file{foo.texi}, on my machine
18002 @file{rattlesnake} in my @file{/home/bob} buffer. I am on line 1, in
18003 Texinfo mode, and am at the top of the buffer.
18006 My @file{.emacs} file has a section that looks like this:
18010 ;; Set a Mode Line that tells me which machine, which directory,
18011 ;; and which line I am on, plus the other customary information.
18012 (setq-default mode-line-format
18016 "mouse-1: select window, mouse-2: delete others ..."))
18017 mode-line-mule-info
18019 mode-line-frame-identification
18023 mode-line-buffer-identification
18026 (system-name) 0 (string-match "\\..+" (system-name))))
18031 "mouse-1: select window, mouse-2: delete others ..."))
18032 (line-number-mode " Line %l ")
18038 "mouse-1: select window, mouse-2: delete others ..."))
18039 (:eval (mode-line-mode-name))
18042 #("%n" 0 2 (help-echo "mouse-2: widen" local-map (keymap ...)))
18051 Here, I redefine the default mode line. Most of the parts are from
18052 the original; but I make a few changes. I set the @emph{default} mode
18053 line format so as to permit various modes, such as Info, to override
18056 Many elements in the list are self-explanatory:
18057 @code{mode-line-modified} is a variable that tells whether the buffer
18058 has been modified, @code{mode-name} tells the name of the mode, and so
18059 on. However, the format looks complicated because of two features we
18060 have not discussed.
18062 @cindex Properties, in mode line example
18063 The first string in the mode line is a dash or hyphen, @samp{-}. In
18064 the old days, it would have been specified simply as @code{"-"}. But
18065 nowadays, Emacs can add properties to a string, such as highlighting
18066 or, as in this case, a help feature. If you place your mouse cursor
18067 over the hyphen, some help information appears (By default, you must
18068 wait seven-tenths of a second before the information appears. You can
18069 change that timing by changing the value of @code{tooltip-delay}.)
18072 The new string format has a special syntax:
18075 #("-" 0 1 (help-echo "mouse-1: select window, ..."))
18079 The @code{#(} begins a list. The first element of the list is the
18080 string itself, just one @samp{-}. The second and third
18081 elements specify the range over which the fourth element applies. A
18082 range starts @emph{after} a character, so a zero means the range
18083 starts just before the first character; a 1 means that the range ends
18084 just after the first character. The third element is the property for
18085 the range. It consists of a property list, a
18086 property name, in this case, @samp{help-echo}, followed by a value, in this
18087 case, a string. The second, third, and fourth elements of this new
18088 string format can be repeated.
18090 @xref{Text Properties, , Text Properties, elisp, The GNU Emacs Lisp
18091 Reference Manual}, and see @ref{Mode Line Format, , Mode Line Format,
18092 elisp, The GNU Emacs Lisp Reference Manual}, for more information.
18094 @code{mode-line-buffer-identification}
18095 displays the current buffer name. It is a list
18096 beginning @code{(#("%12b" 0 4 @dots{}}.
18097 The @code{#(} begins the list.
18099 The @samp{"%12b"} displays the current buffer name, using the
18100 @code{buffer-name} function with which we are familiar; the `12'
18101 specifies the maximum number of characters that will be displayed.
18102 When a name has fewer characters, whitespace is added to fill out to
18103 this number. (Buffer names can and often should be longer than 12
18104 characters; this length works well in a typical 80 column wide
18107 @code{:eval} says to evaluate the following form and use the result as
18108 a string to display. In this case, the expression displays the first
18109 component of the full system name. The end of the first component is
18110 a @samp{.} (`period'), so I use the @code{string-match} function to
18111 tell me the length of the first component. The substring from the
18112 zeroth character to that length is the name of the machine.
18115 This is the expression:
18120 (system-name) 0 (string-match "\\..+" (system-name))))
18124 @samp{%[} and @samp{%]} cause a pair of square brackets
18125 to appear for each recursive editing level. @samp{%n} says `Narrow'
18126 when narrowing is in effect. @samp{%P} tells you the percentage of
18127 the buffer that is above the bottom of the window, or `Top', `Bottom',
18128 or `All'. (A lower case @samp{p} tell you the percentage above the
18129 @emph{top} of the window.) @samp{%-} inserts enough dashes to fill
18132 Remember, ``You don't have to like Emacs to like it''---your own
18133 Emacs can have different colors, different commands, and different
18134 keys than a default Emacs.
18136 On the other hand, if you want to bring up a plain `out of the box'
18137 Emacs, with no customization, type:
18144 This will start an Emacs that does @emph{not} load your
18145 @file{~/.emacs} initialization file. A plain, default Emacs. Nothing
18152 GNU Emacs has two debuggers, @code{debug} and @code{edebug}. The
18153 first is built into the internals of Emacs and is always with you;
18154 the second requires that you instrument a function before you can use it.
18156 Both debuggers are described extensively in @ref{Debugging, ,
18157 Debugging Lisp Programs, elisp, The GNU Emacs Lisp Reference Manual}.
18158 In this chapter, I will walk through a short example of each.
18161 * debug:: How to use the built-in debugger.
18162 * debug-on-entry:: Start debugging when you call a function.
18163 * debug-on-quit:: Start debugging when you quit with @kbd{C-g}.
18164 * edebug:: How to use Edebug, a source level debugger.
18165 * Debugging Exercises::
18169 @section @code{debug}
18172 Suppose you have written a function definition that is intended to
18173 return the sum of the numbers 1 through a given number. (This is the
18174 @code{triangle} function discussed earlier. @xref{Decrementing
18175 Example, , Example with Decrementing Counter}, for a discussion.)
18176 @c xref{Decrementing Loop,, Loop with a Decrementing Counter}, for a discussion.)
18178 However, your function definition has a bug. You have mistyped
18179 @samp{1=} for @samp{1-}. Here is the broken definition:
18181 @findex triangle-bugged
18184 (defun triangle-bugged (number)
18185 "Return sum of numbers 1 through NUMBER inclusive."
18187 (while (> number 0)
18188 (setq total (+ total number))
18189 (setq number (1= number))) ; @r{Error here.}
18194 If you are reading this in Info, you can evaluate this definition in
18195 the normal fashion. You will see @code{triangle-bugged} appear in the
18199 Now evaluate the @code{triangle-bugged} function with an
18203 (triangle-bugged 4)
18207 In a recent GNU Emacs, you will create and enter a @file{*Backtrace*}
18213 ---------- Buffer: *Backtrace* ----------
18214 Debugger entered--Lisp error: (void-function 1=)
18216 (setq number (1= number))
18217 (while (> number 0) (setq total (+ total number))
18218 (setq number (1= number)))
18219 (let ((total 0)) (while (> number 0) (setq total ...)
18220 (setq number ...)) total)
18224 eval((triangle-bugged 4))
18225 eval-last-sexp-1(nil)
18226 eval-last-sexp(nil)
18227 call-interactively(eval-last-sexp)
18228 ---------- Buffer: *Backtrace* ----------
18233 (I have reformatted this example slightly; the debugger does not fold
18234 long lines. As usual, you can quit the debugger by typing @kbd{q} in
18235 the @file{*Backtrace*} buffer.)
18237 In practice, for a bug as simple as this, the `Lisp error' line will
18238 tell you what you need to know to correct the definition. The
18239 function @code{1=} is `void'.
18243 In GNU Emacs 20 and before, you will see:
18246 Symbol's function definition is void:@: 1=
18250 which has the same meaning as the @file{*Backtrace*} buffer line in
18254 However, suppose you are not quite certain what is going on?
18255 You can read the complete backtrace.
18257 In this case, you need to run a recent GNU Emacs, which automatically
18258 starts the debugger that puts you in the @file{*Backtrace*} buffer; or
18259 else, you need to start the debugger manually as described below.
18261 Read the @file{*Backtrace*} buffer from the bottom up; it tells you
18262 what Emacs did that led to the error. Emacs made an interactive call
18263 to @kbd{C-x C-e} (@code{eval-last-sexp}), which led to the evaluation
18264 of the @code{triangle-bugged} expression. Each line above tells you
18265 what the Lisp interpreter evaluated next.
18268 The third line from the top of the buffer is
18271 (setq number (1= number))
18275 Emacs tried to evaluate this expression; in order to do so, it tried
18276 to evaluate the inner expression shown on the second line from the
18285 This is where the error occurred; as the top line says:
18288 Debugger entered--Lisp error: (void-function 1=)
18292 You can correct the mistake, re-evaluate the function definition, and
18293 then run your test again.
18295 @node debug-on-entry
18296 @section @code{debug-on-entry}
18297 @findex debug-on-entry
18299 A recent GNU Emacs starts the debugger automatically when your
18300 function has an error.
18303 GNU Emacs version 20 and before did not; it simply
18304 presented you with an error message. You had to start the debugger
18308 Incidentally, you can start the debugger manually for all versions of
18309 Emacs; the advantage is that the debugger runs even if you do not have
18310 a bug in your code. Sometimes your code will be free of bugs!
18312 You can enter the debugger when you call the function by calling
18313 @code{debug-on-entry}.
18320 M-x debug-on-entry RET triangle-bugged RET
18325 Now, evaluate the following:
18328 (triangle-bugged 5)
18332 All versions of Emacs will create a @file{*Backtrace*} buffer and tell
18333 you that it is beginning to evaluate the @code{triangle-bugged}
18338 ---------- Buffer: *Backtrace* ----------
18339 Debugger entered--entering a function:
18340 * triangle-bugged(5)
18341 eval((triangle-bugged 5))
18344 eval-last-sexp-1(nil)
18345 eval-last-sexp(nil)
18346 call-interactively(eval-last-sexp)
18347 ---------- Buffer: *Backtrace* ----------
18351 In the @file{*Backtrace*} buffer, type @kbd{d}. Emacs will evaluate
18352 the first expression in @code{triangle-bugged}; the buffer will look
18357 ---------- Buffer: *Backtrace* ----------
18358 Debugger entered--beginning evaluation of function call form:
18359 * (let ((total 0)) (while (> number 0) (setq total ...)
18360 (setq number ...)) total)
18361 * triangle-bugged(5)
18362 eval((triangle-bugged 5))
18365 eval-last-sexp-1(nil)
18366 eval-last-sexp(nil)
18367 call-interactively(eval-last-sexp)
18368 ---------- Buffer: *Backtrace* ----------
18373 Now, type @kbd{d} again, eight times, slowly. Each time you type
18374 @kbd{d}, Emacs will evaluate another expression in the function
18378 Eventually, the buffer will look like this:
18382 ---------- Buffer: *Backtrace* ----------
18383 Debugger entered--beginning evaluation of function call form:
18384 * (setq number (1= number))
18385 * (while (> number 0) (setq total (+ total number))
18386 (setq number (1= number)))
18389 * (let ((total 0)) (while (> number 0) (setq total ...)
18390 (setq number ...)) total)
18391 * triangle-bugged(5)
18392 eval((triangle-bugged 5))
18395 eval-last-sexp-1(nil)
18396 eval-last-sexp(nil)
18397 call-interactively(eval-last-sexp)
18398 ---------- Buffer: *Backtrace* ----------
18404 Finally, after you type @kbd{d} two more times, Emacs will reach the
18405 error, and the top two lines of the @file{*Backtrace*} buffer will look
18410 ---------- Buffer: *Backtrace* ----------
18411 Debugger entered--Lisp error: (void-function 1=)
18414 ---------- Buffer: *Backtrace* ----------
18418 By typing @kbd{d}, you were able to step through the function.
18420 You can quit a @file{*Backtrace*} buffer by typing @kbd{q} in it; this
18421 quits the trace, but does not cancel @code{debug-on-entry}.
18423 @findex cancel-debug-on-entry
18424 To cancel the effect of @code{debug-on-entry}, call
18425 @code{cancel-debug-on-entry} and the name of the function, like this:
18428 M-x cancel-debug-on-entry RET triangle-bugged RET
18432 (If you are reading this in Info, cancel @code{debug-on-entry} now.)
18434 @node debug-on-quit
18435 @section @code{debug-on-quit} and @code{(debug)}
18437 In addition to setting @code{debug-on-error} or calling @code{debug-on-entry},
18438 there are two other ways to start @code{debug}.
18440 @findex debug-on-quit
18441 You can start @code{debug} whenever you type @kbd{C-g}
18442 (@code{keyboard-quit}) by setting the variable @code{debug-on-quit} to
18443 @code{t}. This is useful for debugging infinite loops.
18446 @cindex @code{(debug)} in code
18447 Or, you can insert a line that says @code{(debug)} into your code
18448 where you want the debugger to start, like this:
18452 (defun triangle-bugged (number)
18453 "Return sum of numbers 1 through NUMBER inclusive."
18455 (while (> number 0)
18456 (setq total (+ total number))
18457 (debug) ; @r{Start debugger.}
18458 (setq number (1= number))) ; @r{Error here.}
18463 The @code{debug} function is described in detail in @ref{Debugger, ,
18464 The Lisp Debugger, elisp, The GNU Emacs Lisp Reference Manual}.
18467 @section The @code{edebug} Source Level Debugger
18468 @cindex Source level debugger
18471 Edebug is a source level debugger. Edebug normally displays the
18472 source of the code you are debugging, with an arrow at the left that
18473 shows which line you are currently executing.
18475 You can walk through the execution of a function, line by line, or run
18476 quickly until reaching a @dfn{breakpoint} where execution stops.
18478 Edebug is described in @ref{Edebug, , , elisp, The GNU Emacs
18479 Lisp Reference Manual}.
18482 Here is a bugged function definition for @code{triangle-recursively}.
18483 @xref{Recursive triangle function, , Recursion in place of a counter},
18484 for a review of it.
18488 (defun triangle-recursively-bugged (number)
18489 "Return sum of numbers 1 through NUMBER inclusive.
18494 (triangle-recursively-bugged
18495 (1= number))))) ; @r{Error here.}
18500 Normally, you would install this definition by positioning your cursor
18501 after the function's closing parenthesis and typing @kbd{C-x C-e}
18502 (@code{eval-last-sexp}) or else by positioning your cursor within the
18503 definition and typing @kbd{C-M-x} (@code{eval-defun}). (By default,
18504 the @code{eval-defun} command works only in Emacs Lisp mode or in Lisp
18508 However, to prepare this function definition for Edebug, you must
18509 first @dfn{instrument} the code using a different command. You can do
18510 this by positioning your cursor within or just after the definition
18514 M-x edebug-defun RET
18518 This will cause Emacs to load Edebug automatically if it is not
18519 already loaded, and properly instrument the function.
18521 After instrumenting the function, place your cursor after the
18522 following expression and type @kbd{C-x C-e} (@code{eval-last-sexp}):
18525 (triangle-recursively-bugged 3)
18529 You will be jumped back to the source for
18530 @code{triangle-recursively-bugged} and the cursor positioned at the
18531 beginning of the @code{if} line of the function. Also, you will see
18532 an arrowhead at the left hand side of that line. The arrowhead marks
18533 the line where the function is executing. (In the following examples,
18534 we show the arrowhead with @samp{=>}; in a windowing system, you may
18535 see the arrowhead as a solid triangle in the window `fringe'.)
18538 =>@point{}(if (= number 1)
18543 In the example, the location of point is displayed with a star,
18544 @samp{@point{}} (in Info, it is displayed as @samp{-!-}).
18547 In the example, the location of point is displayed as @samp{@point{}}
18548 (in a printed book, it is displayed with a five pointed star).
18551 If you now press @key{SPC}, point will move to the next expression to
18552 be executed; the line will look like this:
18555 =>(if @point{}(= number 1)
18559 As you continue to press @key{SPC}, point will move from expression to
18560 expression. At the same time, whenever an expression returns a value,
18561 that value will be displayed in the echo area. For example, after you
18562 move point past @code{number}, you will see the following:
18565 Result: 3 (#o3, #x3, ?\C-c)
18569 This means the value of @code{number} is 3, which is octal three,
18570 hexadecimal three, and @sc{ascii} `control-c' (the third letter of the
18571 alphabet, in case you need to know this information).
18573 You can continue moving through the code until you reach the line with
18574 the error. Before evaluation, that line looks like this:
18577 => @point{}(1= number))))) ; @r{Error here.}
18582 When you press @key{SPC} once again, you will produce an error message
18586 Symbol's function definition is void:@: 1=
18592 Press @kbd{q} to quit Edebug.
18594 To remove instrumentation from a function definition, simply
18595 re-evaluate it with a command that does not instrument it.
18596 For example, you could place your cursor after the definition's
18597 closing parenthesis and type @kbd{C-x C-e}.
18599 Edebug does a great deal more than walk with you through a function.
18600 You can set it so it races through on its own, stopping only at an
18601 error or at specified stopping points; you can cause it to display the
18602 changing values of various expressions; you can find out how many
18603 times a function is called, and more.
18605 Edebug is described in @ref{Edebug, , , elisp, The GNU Emacs
18606 Lisp Reference Manual}.
18609 @node Debugging Exercises
18610 @section Debugging Exercises
18614 Install the @code{@value{COUNT-WORDS}} function and then cause it to
18615 enter the built-in debugger when you call it. Run the command on a
18616 region containing two words. You will need to press @kbd{d} a
18617 remarkable number of times. On your system, is a `hook' called after
18618 the command finishes? (For information on hooks, see @ref{Command
18619 Overview, , Command Loop Overview, elisp, The GNU Emacs Lisp Reference
18623 Copy @code{@value{COUNT-WORDS}} into the @file{*scratch*} buffer,
18624 instrument the function for Edebug, and walk through its execution.
18625 The function does not need to have a bug, although you can introduce
18626 one if you wish. If the function lacks a bug, the walk-through
18627 completes without problems.
18630 While running Edebug, type @kbd{?} to see a list of all the Edebug commands.
18631 (The @code{global-edebug-prefix} is usually @kbd{C-x X}, i.e.,
18632 @kbd{@key{CTRL}-x} followed by an upper case @kbd{X}; use this prefix
18633 for commands made outside of the Edebug debugging buffer.)
18636 In the Edebug debugging buffer, use the @kbd{p}
18637 (@code{edebug-bounce-point}) command to see where in the region the
18638 @code{@value{COUNT-WORDS}} is working.
18641 Move point to some spot further down the function and then type the
18642 @kbd{h} (@code{edebug-goto-here}) command to jump to that location.
18645 Use the @kbd{t} (@code{edebug-trace-mode}) command to cause Edebug to
18646 walk through the function on its own; use an upper case @kbd{T} for
18647 @code{edebug-Trace-fast-mode}.
18650 Set a breakpoint, then run Edebug in Trace mode until it reaches the
18655 @chapter Conclusion
18657 We have now reached the end of this Introduction. You have now
18658 learned enough about programming in Emacs Lisp to set values, to write
18659 simple @file{.emacs} files for yourself and your friends, and write
18660 simple customizations and extensions to Emacs.
18662 This is a place to stop. Or, if you wish, you can now go onward, and
18665 You have learned some of the basic nuts and bolts of programming. But
18666 only some. There are a great many more brackets and hinges that are
18667 easy to use that we have not touched.
18669 A path you can follow right now lies among the sources to GNU Emacs
18672 @cite{The GNU Emacs Lisp Reference Manual}.
18675 @ref{Top, , The GNU Emacs Lisp Reference Manual, elisp, The GNU
18676 Emacs Lisp Reference Manual}.
18679 The Emacs Lisp sources are an adventure. When you read the sources and
18680 come across a function or expression that is unfamiliar, you need to
18681 figure out or find out what it does.
18683 Go to the Reference Manual. It is a thorough, complete, and fairly
18684 easy-to-read description of Emacs Lisp. It is written not only for
18685 experts, but for people who know what you know. (The @cite{Reference
18686 Manual} comes with the standard GNU Emacs distribution. Like this
18687 introduction, it comes as a Texinfo source file, so you can read it
18688 on-line and as a typeset, printed book.)
18690 Go to the other on-line help that is part of GNU Emacs: the on-line
18691 documentation for all functions and variables, and @code{find-tag},
18692 the program that takes you to sources.
18694 Here is an example of how I explore the sources. Because of its name,
18695 @file{simple.el} is the file I looked at first, a long time ago. As
18696 it happens some of the functions in @file{simple.el} are complicated,
18697 or at least look complicated at first sight. The @code{open-line}
18698 function, for example, looks complicated.
18700 You may want to walk through this function slowly, as we did with the
18701 @code{forward-sentence} function. (@xref{forward-sentence, The
18702 @code{forward-sentence} function}.) Or you may want to skip that
18703 function and look at another, such as @code{split-line}. You don't
18704 need to read all the functions. According to
18705 @code{count-words-in-defun}, the @code{split-line} function contains
18706 102 words and symbols.
18708 Even though it is short, @code{split-line} contains expressions
18709 we have not studied: @code{skip-chars-forward}, @code{indent-to},
18710 @code{current-column} and @code{insert-and-inherit}.
18712 Consider the @code{skip-chars-forward} function. (It is part of the
18713 function definition for @code{back-to-indentation}, which is shown in
18714 @ref{Review, , Review}.)
18716 In GNU Emacs, you can find out more about @code{skip-chars-forward} by
18717 typing @kbd{C-h f} (@code{describe-function}) and the name of the
18718 function. This gives you the function documentation.
18720 You may be able to guess what is done by a well named function such as
18721 @code{indent-to}; or you can look it up, too. Incidentally, the
18722 @code{describe-function} function itself is in @file{help.el}; it is
18723 one of those long, but decipherable functions. You can look up
18724 @code{describe-function} using the @kbd{C-h f} command!
18726 In this instance, since the code is Lisp, the @file{*Help*} buffer
18727 contains the name of the library containing the function's source.
18728 You can put point over the name of the library and press the RET key,
18729 which in this situation is bound to @code{help-follow}, and be taken
18730 directly to the source, in the same way as @kbd{M-.}
18733 The definition for @code{describe-function} illustrates how to
18734 customize the @code{interactive} expression without using the standard
18735 character codes; and it shows how to create a temporary buffer.
18737 (The @code{indent-to} function is written in C rather than Emacs Lisp;
18738 it is a `built-in' function. @code{help-follow} takes you to its
18739 source as does @code{find-tag}, when properly set up.)
18741 You can look at a function's source using @code{find-tag}, which is
18742 bound to @kbd{M-.} Finally, you can find out what the Reference
18743 Manual has to say by visiting the manual in Info, and typing @kbd{i}
18744 (@code{Info-index}) and the name of the function, or by looking up the
18745 function in the index to a printed copy of the manual.
18747 Similarly, you can find out what is meant by
18748 @code{insert-and-inherit}.
18750 Other interesting source files include @file{paragraphs.el},
18751 @file{loaddefs.el}, and @file{loadup.el}. The @file{paragraphs.el}
18752 file includes short, easily understood functions as well as longer
18753 ones. The @file{loaddefs.el} file contains the many standard
18754 autoloads and many keymaps. I have never looked at it all; only at
18755 parts. @file{loadup.el} is the file that loads the standard parts of
18756 Emacs; it tells you a great deal about how Emacs is built.
18757 (@xref{Building Emacs, , Building Emacs, elisp, The GNU Emacs Lisp
18758 Reference Manual}, for more about building.)
18760 As I said, you have learned some nuts and bolts; however, and very
18761 importantly, we have hardly touched major aspects of programming; I
18762 have said nothing about how to sort information, except to use the
18763 predefined @code{sort} function; I have said nothing about how to store
18764 information, except to use variables and lists; I have said nothing
18765 about how to write programs that write programs. These are topics for
18766 another, and different kind of book, a different kind of learning.
18768 What you have done is learn enough for much practical work with GNU
18769 Emacs. What you have done is get started. This is the end of a
18772 @c ================ Appendix ================
18775 @appendix The @code{the-the} Function
18777 @cindex Duplicated words function
18778 @cindex Words, duplicated
18780 Sometimes when you you write text, you duplicate words---as with ``you
18781 you'' near the beginning of this sentence. I find that most
18782 frequently, I duplicate ``the''; hence, I call the function for
18783 detecting duplicated words, @code{the-the}.
18786 As a first step, you could use the following regular expression to
18787 search for duplicates:
18790 \\(\\w+[ \t\n]+\\)\\1
18794 This regexp matches one or more word-constituent characters followed
18795 by one or more spaces, tabs, or newlines. However, it does not detect
18796 duplicated words on different lines, since the ending of the first
18797 word, the end of the line, is different from the ending of the second
18798 word, a space. (For more information about regular expressions, see
18799 @ref{Regexp Search, , Regular Expression Searches}, as well as
18800 @ref{Regexps, , Syntax of Regular Expressions, emacs, The GNU Emacs
18801 Manual}, and @ref{Regular Expressions, , Regular Expressions, elisp,
18802 The GNU Emacs Lisp Reference Manual}.)
18804 You might try searching just for duplicated word-constituent
18805 characters but that does not work since the pattern detects doubles
18806 such as the two occurrences of `th' in `with the'.
18808 Another possible regexp searches for word-constituent characters
18809 followed by non-word-constituent characters, reduplicated. Here,
18810 @w{@samp{\\w+}} matches one or more word-constituent characters and
18811 @w{@samp{\\W*}} matches zero or more non-word-constituent characters.
18814 \\(\\(\\w+\\)\\W*\\)\\1
18820 Here is the pattern that I use. It is not perfect, but good enough.
18821 @w{@samp{\\b}} matches the empty string, provided it is at the beginning
18822 or end of a word; @w{@samp{[^@@ \n\t]+}} matches one or more occurrences of
18823 any characters that are @emph{not} an @@-sign, space, newline, or tab.
18826 \\b\\([^@@ \n\t]+\\)[ \n\t]+\\1\\b
18829 One can write more complicated expressions, but I found that this
18830 expression is good enough, so I use it.
18832 Here is the @code{the-the} function, as I include it in my
18833 @file{.emacs} file, along with a handy global key binding:
18838 "Search forward for for a duplicated word."
18840 (message "Searching for for duplicated words ...")
18844 ;; This regexp is not perfect
18845 ;; but is fairly good over all:
18846 (if (re-search-forward
18847 "\\b\\([^@@ \n\t]+\\)[ \n\t]+\\1\\b" nil 'move)
18848 (message "Found duplicated word.")
18849 (message "End of buffer")))
18853 ;; Bind `the-the' to C-c \
18854 (global-set-key "\C-c\\" 'the-the)
18863 one two two three four five
18868 You can substitute the other regular expressions shown above in the
18869 function definition and try each of them on this list.
18872 @appendix Handling the Kill Ring
18873 @cindex Kill ring handling
18874 @cindex Handling the kill ring
18875 @cindex Ring, making a list like a
18877 The kill ring is a list that is transformed into a ring by the
18878 workings of the @code{current-kill} function. The @code{yank} and
18879 @code{yank-pop} commands use the @code{current-kill} function.
18881 This appendix describes the @code{current-kill} function as well as
18882 both the @code{yank} and the @code{yank-pop} commands, but first,
18883 consider the workings of the kill ring.
18886 * What the Kill Ring Does::
18888 * yank:: Paste a copy of a clipped element.
18889 * yank-pop:: Insert element pointed to.
18894 @node What the Kill Ring Does
18895 @unnumberedsec What the Kill Ring Does
18899 The kill ring has a default maximum length of sixty items; this number
18900 is too large for an explanation. Instead, set it to four. Please
18901 evaluate the following:
18905 (setq old-kill-ring-max kill-ring-max)
18906 (setq kill-ring-max 4)
18911 Then, please copy each line of the following indented example into the
18912 kill ring. You may kill each line with @kbd{C-k} or mark it and copy
18916 (In a read-only buffer, such as the @file{*info*} buffer, the kill
18917 command, @kbd{C-k} (@code{kill-line}), will not remove the text,
18918 merely copy it to the kill ring. However, your machine may beep at
18919 you. Alternatively, for silence, you may copy the region of each line
18920 with the @kbd{M-w} (@code{kill-ring-save}) command. You must mark
18921 each line for this command to succeed, but it does not matter at which
18922 end you put point or mark.)
18926 Please invoke the calls in order, so that five elements attempt to
18927 fill the kill ring:
18932 second piece of text
18934 fourth line of text
18941 Then find the value of @code{kill-ring} by evaluating
18953 ("fifth bit of text" "fourth line of text"
18954 "third line" "second piece of text")
18959 The first element, @samp{first some text}, was dropped.
18962 To return to the old value for the length of the kill ring, evaluate:
18965 (setq kill-ring-max old-kill-ring-max)
18969 @appendixsec The @code{current-kill} Function
18970 @findex current-kill
18972 The @code{current-kill} function changes the element in the kill ring
18973 to which @code{kill-ring-yank-pointer} points. (Also, the
18974 @code{kill-new} function sets @code{kill-ring-yank-pointer} to point
18975 to the latest element of the kill ring. The @code{kill-new}
18976 function is used directly or indirectly by @code{kill-append},
18977 @code{copy-region-as-kill}, @code{kill-ring-save}, @code{kill-line},
18978 and @code{kill-region}.)
18981 * Code for current-kill::
18982 * Understanding current-kill::
18986 @node Code for current-kill
18987 @unnumberedsubsec The code for @code{current-kill}
18992 The @code{current-kill} function is used by @code{yank} and by
18993 @code{yank-pop}. Here is the code for @code{current-kill}:
18997 (defun current-kill (n &optional do-not-move)
18998 "Rotate the yanking point by N places, and then return that kill.
18999 If N is zero, `interprogram-paste-function' is set, and calling it
19000 returns a string, then that string is added to the front of the
19001 kill ring and returned as the latest kill.
19004 If optional arg DO-NOT-MOVE is non-nil, then don't actually move the
19005 yanking point; just return the Nth kill forward."
19006 (let ((interprogram-paste (and (= n 0)
19007 interprogram-paste-function
19008 (funcall interprogram-paste-function))))
19011 (if interprogram-paste
19013 ;; Disable the interprogram cut function when we add the new
19014 ;; text to the kill ring, so Emacs doesn't try to own the
19015 ;; selection, with identical text.
19016 (let ((interprogram-cut-function nil))
19017 (kill-new interprogram-paste))
19018 interprogram-paste)
19021 (or kill-ring (error "Kill ring is empty"))
19022 (let ((ARGth-kill-element
19023 (nthcdr (mod (- n (length kill-ring-yank-pointer))
19024 (length kill-ring))
19027 (setq kill-ring-yank-pointer ARGth-kill-element))
19028 (car ARGth-kill-element)))))
19032 Remember also that the @code{kill-new} function sets
19033 @code{kill-ring-yank-pointer} to the latest element of the kill
19034 ring, which means that all the functions that call it set the value
19035 indirectly: @code{kill-append}, @code{copy-region-as-kill},
19036 @code{kill-ring-save}, @code{kill-line}, and @code{kill-region}.
19039 Here is the line in @code{kill-new}, which is explained in
19040 @ref{kill-new function, , The @code{kill-new} function}.
19043 (setq kill-ring-yank-pointer kill-ring)
19047 @node Understanding current-kill
19048 @unnumberedsubsec @code{current-kill} in Outline
19051 The @code{current-kill} function looks complex, but as usual, it can
19052 be understood by taking it apart piece by piece. First look at it in
19057 (defun current-kill (n &optional do-not-move)
19058 "Rotate the yanking point by N places, and then return that kill."
19064 This function takes two arguments, one of which is optional. It has a
19065 documentation string. It is @emph{not} interactive.
19068 * Body of current-kill::
19069 * Digression concerning error:: How to mislead humans, but not computers.
19070 * Determining the Element::
19074 @node Body of current-kill
19075 @unnumberedsubsubsec The Body of @code{current-kill}
19078 The body of the function definition is a @code{let} expression, which
19079 itself has a body as well as a @var{varlist}.
19081 The @code{let} expression declares a variable that will be only usable
19082 within the bounds of this function. This variable is called
19083 @code{interprogram-paste} and is for copying to another program. It
19084 is not for copying within this instance of GNU Emacs. Most window
19085 systems provide a facility for interprogram pasting. Sadly, that
19086 facility usually provides only for the last element. Most windowing
19087 systems have not adopted a ring of many possibilities, even though
19088 Emacs has provided it for decades.
19090 The @code{if} expression has two parts, one if there exists
19091 @code{interprogram-paste} and one if not.
19094 Let us consider the `if not' or else-part of the @code{current-kill}
19095 function. (The then-part uses the @code{kill-new} function, which
19096 we have already described. @xref{kill-new function, , The
19097 @code{kill-new} function}.)
19101 (or kill-ring (error "Kill ring is empty"))
19102 (let ((ARGth-kill-element
19103 (nthcdr (mod (- n (length kill-ring-yank-pointer))
19104 (length kill-ring))
19107 (setq kill-ring-yank-pointer ARGth-kill-element))
19108 (car ARGth-kill-element))
19113 The code first checks whether the kill ring has content; otherwise it
19117 Note that the @code{or} expression is very similar to testing length
19124 (if (zerop (length kill-ring)) ; @r{if-part}
19125 (error "Kill ring is empty")) ; @r{then-part}
19131 If there is not anything in the kill ring, its length must be zero and
19132 an error message sent to the user: @samp{Kill ring is empty}. The
19133 @code{current-kill} function uses an @code{or} expression which is
19134 simpler. But an @code{if} expression reminds us what goes on.
19136 This @code{if} expression uses the function @code{zerop} which returns
19137 true if the value it is testing is zero. When @code{zerop} tests
19138 true, the then-part of the @code{if} is evaluated. The then-part is a
19139 list starting with the function @code{error}, which is a function that
19140 is similar to the @code{message} function
19141 (@pxref{message, , The @code{message} Function}) in that
19142 it prints a one-line message in the echo area. However, in addition
19143 to printing a message, @code{error} also stops evaluation of the
19144 function within which it is embedded. This means that the rest of the
19145 function will not be evaluated if the length of the kill ring is zero.
19147 Then the @code{current-kill} function selects the element to return.
19148 The selection depends on the number of places that @code{current-kill}
19149 rotates and on where @code{kill-ring-yank-pointer} points.
19151 Next, either the optional @code{do-not-move} argument is true or the
19152 current value of @code{kill-ring-yank-pointer} is set to point to the
19153 list. Finally, another expression returns the first element of the
19154 list even if the @code{do-not-move} argument is true.
19157 @node Digression concerning error
19158 @unnumberedsubsubsec Digression about the word `error'
19161 In my opinion, it is slightly misleading, at least to humans, to use
19162 the term `error' as the name of the @code{error} function. A better
19163 term would be `cancel'. Strictly speaking, of course, you cannot
19164 point to, much less rotate a pointer to a list that has no length, so
19165 from the point of view of the computer, the word `error' is correct.
19166 But a human expects to attempt this sort of thing, if only to find out
19167 whether the kill ring is full or empty. This is an act of
19170 From the human point of view, the act of exploration and discovery is
19171 not necessarily an error, and therefore should not be labeled as one,
19172 even in the bowels of a computer. As it is, the code in Emacs implies
19173 that a human who is acting virtuously, by exploring his or her
19174 environment, is making an error. This is bad. Even though the computer
19175 takes the same steps as it does when there is an `error', a term such as
19176 `cancel' would have a clearer connotation.
19179 @node Determining the Element
19180 @unnumberedsubsubsec Determining the Element
19183 Among other actions, the else-part of the @code{if} expression sets
19184 the value of @code{kill-ring-yank-pointer} to
19185 @code{ARGth-kill-element} when the kill ring has something in it and
19186 the value of @code{do-not-move} is @code{nil}.
19189 The code looks like this:
19193 (nthcdr (mod (- n (length kill-ring-yank-pointer))
19194 (length kill-ring))
19199 This needs some examination. Unless it is not supposed to move the
19200 pointer, the @code{current-kill} function changes where
19201 @code{kill-ring-yank-pointer} points.
19203 @w{@code{(setq kill-ring-yank-pointer ARGth-kill-element))}}
19204 expression does. Also, clearly, @code{ARGth-kill-element} is being
19205 set to be equal to some @sc{cdr} of the kill ring, using the
19206 @code{nthcdr} function that is described in an earlier section.
19207 (@xref{copy-region-as-kill}.) How does it do this?
19209 As we have seen before (@pxref{nthcdr}), the @code{nthcdr} function
19210 works by repeatedly taking the @sc{cdr} of a list---it takes the
19211 @sc{cdr} of the @sc{cdr} of the @sc{cdr} @dots{}
19214 The two following expressions produce the same result:
19218 (setq kill-ring-yank-pointer (cdr kill-ring))
19220 (setq kill-ring-yank-pointer (nthcdr 1 kill-ring))
19224 However, the @code{nthcdr} expression is more complicated. It uses
19225 the @code{mod} function to determine which @sc{cdr} to select.
19227 (You will remember to look at inner functions first; indeed, we will
19228 have to go inside the @code{mod}.)
19230 The @code{mod} function returns the value of its first argument modulo
19231 the second; that is to say, it returns the remainder after dividing
19232 the first argument by the second. The value returned has the same
19233 sign as the second argument.
19241 @result{} 0 ;; @r{because there is no remainder}
19248 In this case, the first argument is often smaller than the second.
19260 We can guess what the @code{-} function does. It is like @code{+} but
19261 subtracts instead of adds; the @code{-} function subtracts its second
19262 argument from its first. Also, we already know what the @code{length}
19263 function does (@pxref{length}). It returns the length of a list.
19265 And @code{n} is the name of the required argument to the
19266 @code{current-kill} function.
19269 So when the first argument to @code{nthcdr} is zero, the @code{nthcdr}
19270 expression returns the whole list, as you can see by evaluating the
19275 ;; kill-ring-yank-pointer @r{and} kill-ring @r{have a length of four}
19276 ;; @r{and} (mod (- 0 4) 4) @result{} 0
19277 (nthcdr (mod (- 0 4) 4)
19278 '("fourth line of text"
19280 "second piece of text"
19281 "first some text"))
19286 When the first argument to the @code{current-kill} function is one,
19287 the @code{nthcdr} expression returns the list without its first
19292 (nthcdr (mod (- 1 4) 4)
19293 '("fourth line of text"
19295 "second piece of text"
19296 "first some text"))
19300 @cindex @samp{global variable} defined
19301 @cindex @samp{variable, global}, defined
19302 Incidentally, both @code{kill-ring} and @code{kill-ring-yank-pointer}
19303 are @dfn{global variables}. That means that any expression in Emacs
19304 Lisp can access them. They are not like the local variables set by
19305 @code{let} or like the symbols in an argument list.
19306 Local variables can only be accessed
19307 within the @code{let} that defines them or the function that specifies
19308 them in an argument list (and within expressions called by them).
19311 @c texi2dvi fails when the name of the section is within ifnottex ...
19312 (@xref{Prevent confusion, , @code{let} Prevents Confusion}, and
19313 @ref{defun, , The @code{defun} Macro}.)
19317 @appendixsec @code{yank}
19320 After learning about @code{current-kill}, the code for the
19321 @code{yank} function is almost easy.
19323 The @code{yank} function does not use the
19324 @code{kill-ring-yank-pointer} variable directly. It calls
19325 @code{insert-for-yank} which calls @code{current-kill} which sets the
19326 @code{kill-ring-yank-pointer} variable.
19329 The code looks like this:
19334 (defun yank (&optional arg)
19335 "Reinsert (\"paste\") the last stretch of killed text.
19336 More precisely, reinsert the stretch of killed text most recently
19337 killed OR yanked. Put point at end, and set mark at beginning.
19338 With just \\[universal-argument] as argument, same but put point at
19339 beginning (and mark at end). With argument N, reinsert the Nth most
19340 recently killed stretch of killed text.
19342 When this command inserts killed text into the buffer, it honors
19343 `yank-excluded-properties' and `yank-handler' as described in the
19344 doc string for `insert-for-yank-1', which see.
19346 See also the command \\[yank-pop]."
19350 (setq yank-window-start (window-start))
19351 ;; If we don't get all the way thru, make last-command indicate that
19352 ;; for the following command.
19353 (setq this-command t)
19354 (push-mark (point))
19357 (insert-for-yank (current-kill (cond
19362 ;; This is like exchange-point-and-mark,
19363 ;; but doesn't activate the mark.
19364 ;; It is cleaner to avoid activation, even though the command
19365 ;; loop would deactivate the mark because we inserted text.
19366 (goto-char (prog1 (mark t)
19367 (set-marker (mark-marker) (point) (current-buffer)))))
19370 ;; If we do get all the way thru, make this-command indicate that.
19371 (if (eq this-command t)
19372 (setq this-command 'yank))
19377 The key expression is @code{insert-for-yank}, which inserts the string
19378 returned by @code{current-kill}, but removes some text properties from
19381 However, before getting to that expression, the function sets the value
19382 of @code{yank-window-start} to the position returned by the
19383 @code{(window-start)} expression, the position at which the display
19384 currently starts. The @code{yank} function also sets
19385 @code{this-command} and pushes the mark.
19387 After it yanks the appropriate element, if the optional argument is a
19388 @sc{cons} rather than a number or nothing, it puts point at beginning
19389 of the yanked text and mark at its end.
19391 (The @code{prog1} function is like @code{progn} but returns the value
19392 of its first argument rather than the value of its last argument. Its
19393 first argument is forced to return the buffer's mark as an integer.
19394 You can see the documentation for these functions by placing point
19395 over them in this buffer and then typing @kbd{C-h f}
19396 (@code{describe-function}) followed by a @kbd{RET}; the default is the
19399 The last part of the function tells what to do when it succeeds.
19402 @appendixsec @code{yank-pop}
19405 After understanding @code{yank} and @code{current-kill}, you know how
19406 to approach the @code{yank-pop} function. Leaving out the
19407 documentation to save space, it looks like this:
19412 (defun yank-pop (&optional arg)
19415 (if (not (eq last-command 'yank))
19416 (error "Previous command was not a yank"))
19419 (setq this-command 'yank)
19420 (unless arg (setq arg 1))
19421 (let ((inhibit-read-only t)
19422 (before (< (point) (mark t))))
19426 (funcall (or yank-undo-function 'delete-region) (point) (mark t))
19427 (funcall (or yank-undo-function 'delete-region) (mark t) (point)))
19428 (setq yank-undo-function nil)
19431 (set-marker (mark-marker) (point) (current-buffer))
19432 (insert-for-yank (current-kill arg))
19433 ;; Set the window start back where it was in the yank command,
19435 (set-window-start (selected-window) yank-window-start t)
19439 ;; This is like exchange-point-and-mark,
19440 ;; but doesn't activate the mark.
19441 ;; It is cleaner to avoid activation, even though the command
19442 ;; loop would deactivate the mark because we inserted text.
19443 (goto-char (prog1 (mark t)
19444 (set-marker (mark-marker)
19446 (current-buffer))))))
19451 The function is interactive with a small @samp{p} so the prefix
19452 argument is processed and passed to the function. The command can
19453 only be used after a previous yank; otherwise an error message is
19454 sent. This check uses the variable @code{last-command} which is set
19455 by @code{yank} and is discussed elsewhere.
19456 (@xref{copy-region-as-kill}.)
19458 The @code{let} clause sets the variable @code{before} to true or false
19459 depending whether point is before or after mark and then the region
19460 between point and mark is deleted. This is the region that was just
19461 inserted by the previous yank and it is this text that will be
19464 @code{funcall} calls its first argument as a function, passing
19465 remaining arguments to it. The first argument is whatever the
19466 @code{or} expression returns. The two remaining arguments are the
19467 positions of point and mark set by the preceding @code{yank} command.
19469 There is more, but that is the hardest part.
19472 @appendixsec The @file{ring.el} File
19473 @cindex @file{ring.el} file
19475 Interestingly, GNU Emacs posses a file called @file{ring.el} that
19476 provides many of the features we just discussed. But functions such
19477 as @code{kill-ring-yank-pointer} do not use this library, possibly
19478 because they were written earlier.
19481 @appendix A Graph with Labeled Axes
19483 Printed axes help you understand a graph. They convey scale. In an
19484 earlier chapter (@pxref{Readying a Graph, , Readying a Graph}), we
19485 wrote the code to print the body of a graph. Here we write the code
19486 for printing and labeling vertical and horizontal axes, along with the
19490 * Labeled Example::
19491 * print-graph Varlist:: @code{let} expression in @code{print-graph}.
19492 * print-Y-axis:: Print a label for the vertical axis.
19493 * print-X-axis:: Print a horizontal label.
19494 * Print Whole Graph:: The function to print a complete graph.
19498 @node Labeled Example
19499 @unnumberedsec Labeled Example Graph
19502 Since insertions fill a buffer to the right and below point, the new
19503 graph printing function should first print the Y or vertical axis,
19504 then the body of the graph, and finally the X or horizontal axis.
19505 This sequence lays out for us the contents of the function:
19515 Print body of graph.
19522 Here is an example of how a finished graph should look:
19535 1 - ****************
19542 In this graph, both the vertical and the horizontal axes are labeled
19543 with numbers. However, in some graphs, the horizontal axis is time
19544 and would be better labeled with months, like this:
19558 Indeed, with a little thought, we can easily come up with a variety of
19559 vertical and horizontal labeling schemes. Our task could become
19560 complicated. But complications breed confusion. Rather than permit
19561 this, it is better choose a simple labeling scheme for our first
19562 effort, and to modify or replace it later.
19565 These considerations suggest the following outline for the
19566 @code{print-graph} function:
19570 (defun print-graph (numbers-list)
19571 "@var{documentation}@dots{}"
19572 (let ((height @dots{}
19576 (print-Y-axis height @dots{} )
19577 (graph-body-print numbers-list)
19578 (print-X-axis @dots{} )))
19582 We can work on each part of the @code{print-graph} function definition
19585 @node print-graph Varlist
19586 @appendixsec The @code{print-graph} Varlist
19587 @cindex @code{print-graph} varlist
19589 In writing the @code{print-graph} function, the first task is to write
19590 the varlist in the @code{let} expression. (We will leave aside for the
19591 moment any thoughts about making the function interactive or about the
19592 contents of its documentation string.)
19594 The varlist should set several values. Clearly, the top of the label
19595 for the vertical axis must be at least the height of the graph, which
19596 means that we must obtain this information here. Note that the
19597 @code{print-graph-body} function also requires this information. There
19598 is no reason to calculate the height of the graph in two different
19599 places, so we should change @code{print-graph-body} from the way we
19600 defined it earlier to take advantage of the calculation.
19602 Similarly, both the function for printing the X axis labels and the
19603 @code{print-graph-body} function need to learn the value of the width of
19604 each symbol. We can perform the calculation here and change the
19605 definition for @code{print-graph-body} from the way we defined it in the
19608 The length of the label for the horizontal axis must be at least as long
19609 as the graph. However, this information is used only in the function
19610 that prints the horizontal axis, so it does not need to be calculated here.
19612 These thoughts lead us directly to the following form for the varlist
19613 in the @code{let} for @code{print-graph}:
19617 (let ((height (apply 'max numbers-list)) ; @r{First version.}
19618 (symbol-width (length graph-blank)))
19623 As we shall see, this expression is not quite right.
19627 @appendixsec The @code{print-Y-axis} Function
19628 @cindex Axis, print vertical
19629 @cindex Y axis printing
19630 @cindex Vertical axis printing
19631 @cindex Print vertical axis
19633 The job of the @code{print-Y-axis} function is to print a label for
19634 the vertical axis that looks like this:
19652 The function should be passed the height of the graph, and then should
19653 construct and insert the appropriate numbers and marks.
19656 * print-Y-axis in Detail::
19657 * Height of label:: What height for the Y axis?
19658 * Compute a Remainder:: How to compute the remainder of a division.
19659 * Y Axis Element:: Construct a line for the Y axis.
19660 * Y-axis-column:: Generate a list of Y axis labels.
19661 * print-Y-axis Penultimate:: A not quite final version.
19665 @node print-Y-axis in Detail
19666 @unnumberedsubsec The @code{print-Y-axis} Function in Detail
19669 It is easy enough to see in the figure what the Y axis label should
19670 look like; but to say in words, and then to write a function
19671 definition to do the job is another matter. It is not quite true to
19672 say that we want a number and a tic every five lines: there are only
19673 three lines between the @samp{1} and the @samp{5} (lines 2, 3, and 4),
19674 but four lines between the @samp{5} and the @samp{10} (lines 6, 7, 8,
19675 and 9). It is better to say that we want a number and a tic mark on
19676 the base line (number 1) and then that we want a number and a tic on
19677 the fifth line from the bottom and on every line that is a multiple of
19681 @node Height of label
19682 @unnumberedsubsec What height should the label be?
19685 The next issue is what height the label should be? Suppose the maximum
19686 height of tallest column of the graph is seven. Should the highest
19687 label on the Y axis be @samp{5 -}, and should the graph stick up above
19688 the label? Or should the highest label be @samp{7 -}, and mark the peak
19689 of the graph? Or should the highest label be @code{10 -}, which is a
19690 multiple of five, and be higher than the topmost value of the graph?
19692 The latter form is preferred. Most graphs are drawn within rectangles
19693 whose sides are an integral number of steps long---5, 10, 15, and so
19694 on for a step distance of five. But as soon as we decide to use a
19695 step height for the vertical axis, we discover that the simple
19696 expression in the varlist for computing the height is wrong. The
19697 expression is @code{(apply 'max numbers-list)}. This returns the
19698 precise height, not the maximum height plus whatever is necessary to
19699 round up to the nearest multiple of five. A more complex expression
19702 As usual in cases like this, a complex problem becomes simpler if it is
19703 divided into several smaller problems.
19705 First, consider the case when the highest value of the graph is an
19706 integral multiple of five---when it is 5, 10, 15, or some higher
19707 multiple of five. We can use this value as the Y axis height.
19709 A fairly simply way to determine whether a number is a multiple of
19710 five is to divide it by five and see if the division results in a
19711 remainder. If there is no remainder, the number is a multiple of
19712 five. Thus, seven divided by five has a remainder of two, and seven
19713 is not an integral multiple of five. Put in slightly different
19714 language, more reminiscent of the classroom, five goes into seven
19715 once, with a remainder of two. However, five goes into ten twice,
19716 with no remainder: ten is an integral multiple of five.
19718 @node Compute a Remainder
19719 @appendixsubsec Side Trip: Compute a Remainder
19721 @findex % @r{(remainder function)}
19722 @cindex Remainder function, @code{%}
19723 In Lisp, the function for computing a remainder is @code{%}. The
19724 function returns the remainder of its first argument divided by its
19725 second argument. As it happens, @code{%} is a function in Emacs Lisp
19726 that you cannot discover using @code{apropos}: you find nothing if you
19727 type @kbd{M-x apropos @key{RET} remainder @key{RET}}. The only way to
19728 learn of the existence of @code{%} is to read about it in a book such
19729 as this or in the Emacs Lisp sources.
19731 You can try the @code{%} function by evaluating the following two
19743 The first expression returns 2 and the second expression returns 0.
19745 To test whether the returned value is zero or some other number, we
19746 can use the @code{zerop} function. This function returns @code{t} if
19747 its argument, which must be a number, is zero.
19759 Thus, the following expression will return @code{t} if the height
19760 of the graph is evenly divisible by five:
19763 (zerop (% height 5))
19767 (The value of @code{height}, of course, can be found from @code{(apply
19768 'max numbers-list)}.)
19770 On the other hand, if the value of @code{height} is not a multiple of
19771 five, we want to reset the value to the next higher multiple of five.
19772 This is straightforward arithmetic using functions with which we are
19773 already familiar. First, we divide the value of @code{height} by five
19774 to determine how many times five goes into the number. Thus, five
19775 goes into twelve twice. If we add one to this quotient and multiply by
19776 five, we will obtain the value of the next multiple of five that is
19777 larger than the height. Five goes into twelve twice. Add one to two,
19778 and multiply by five; the result is fifteen, which is the next multiple
19779 of five that is higher than twelve. The Lisp expression for this is:
19782 (* (1+ (/ height 5)) 5)
19786 For example, if you evaluate the following, the result is 15:
19789 (* (1+ (/ 12 5)) 5)
19792 All through this discussion, we have been using `five' as the value
19793 for spacing labels on the Y axis; but we may want to use some other
19794 value. For generality, we should replace `five' with a variable to
19795 which we can assign a value. The best name I can think of for this
19796 variable is @code{Y-axis-label-spacing}.
19799 Using this term, and an @code{if} expression, we produce the
19804 (if (zerop (% height Y-axis-label-spacing))
19807 (* (1+ (/ height Y-axis-label-spacing))
19808 Y-axis-label-spacing))
19813 This expression returns the value of @code{height} itself if the height
19814 is an even multiple of the value of the @code{Y-axis-label-spacing} or
19815 else it computes and returns a value of @code{height} that is equal to
19816 the next higher multiple of the value of the @code{Y-axis-label-spacing}.
19818 We can now include this expression in the @code{let} expression of the
19819 @code{print-graph} function (after first setting the value of
19820 @code{Y-axis-label-spacing}):
19821 @vindex Y-axis-label-spacing
19825 (defvar Y-axis-label-spacing 5
19826 "Number of lines from one Y axis label to next.")
19831 (let* ((height (apply 'max numbers-list))
19832 (height-of-top-line
19833 (if (zerop (% height Y-axis-label-spacing))
19838 (* (1+ (/ height Y-axis-label-spacing))
19839 Y-axis-label-spacing)))
19840 (symbol-width (length graph-blank))))
19846 (Note use of the @code{let*} function: the initial value of height is
19847 computed once by the @code{(apply 'max numbers-list)} expression and
19848 then the resulting value of @code{height} is used to compute its
19849 final value. @xref{fwd-para let, , The @code{let*} expression}, for
19850 more about @code{let*}.)
19852 @node Y Axis Element
19853 @appendixsubsec Construct a Y Axis Element
19855 When we print the vertical axis, we want to insert strings such as
19856 @w{@samp{5 -}} and @w{@samp{10 - }} every five lines.
19857 Moreover, we want the numbers and dashes to line up, so shorter
19858 numbers must be padded with leading spaces. If some of the strings
19859 use two digit numbers, the strings with single digit numbers must
19860 include a leading blank space before the number.
19862 @findex number-to-string
19863 To figure out the length of the number, the @code{length} function is
19864 used. But the @code{length} function works only with a string, not with
19865 a number. So the number has to be converted from being a number to
19866 being a string. This is done with the @code{number-to-string} function.
19871 (length (number-to-string 35))
19874 (length (number-to-string 100))
19880 (@code{number-to-string} is also called @code{int-to-string}; you will
19881 see this alternative name in various sources.)
19883 In addition, in each label, each number is followed by a string such
19884 as @w{@samp{ - }}, which we will call the @code{Y-axis-tic} marker.
19885 This variable is defined with @code{defvar}:
19890 (defvar Y-axis-tic " - "
19891 "String that follows number in a Y axis label.")
19895 The length of the Y label is the sum of the length of the Y axis tic
19896 mark and the length of the number of the top of the graph.
19899 (length (concat (number-to-string height) Y-axis-tic)))
19902 This value will be calculated by the @code{print-graph} function in
19903 its varlist as @code{full-Y-label-width} and passed on. (Note that we
19904 did not think to include this in the varlist when we first proposed it.)
19906 To make a complete vertical axis label, a tic mark is concatenated
19907 with a number; and the two together may be preceded by one or more
19908 spaces depending on how long the number is. The label consists of
19909 three parts: the (optional) leading spaces, the number, and the tic
19910 mark. The function is passed the value of the number for the specific
19911 row, and the value of the width of the top line, which is calculated
19912 (just once) by @code{print-graph}.
19916 (defun Y-axis-element (number full-Y-label-width)
19917 "Construct a NUMBERed label element.
19918 A numbered element looks like this ` 5 - ',
19919 and is padded as needed so all line up with
19920 the element for the largest number."
19923 (let* ((leading-spaces
19924 (- full-Y-label-width
19926 (concat (number-to-string number)
19931 (make-string leading-spaces ? )
19932 (number-to-string number)
19937 The @code{Y-axis-element} function concatenates together the leading
19938 spaces, if any; the number, as a string; and the tic mark.
19940 To figure out how many leading spaces the label will need, the
19941 function subtracts the actual length of the label---the length of the
19942 number plus the length of the tic mark---from the desired label width.
19944 @findex make-string
19945 Blank spaces are inserted using the @code{make-string} function. This
19946 function takes two arguments: the first tells it how long the string
19947 will be and the second is a symbol for the character to insert, in a
19948 special format. The format is a question mark followed by a blank
19949 space, like this, @samp{? }. @xref{Character Type, , Character Type,
19950 elisp, The GNU Emacs Lisp Reference Manual}, for a description of the
19951 syntax for characters. (Of course, you might want to replace the
19952 blank space by some other character @dots{} You know what to do.)
19954 The @code{number-to-string} function is used in the concatenation
19955 expression, to convert the number to a string that is concatenated
19956 with the leading spaces and the tic mark.
19958 @node Y-axis-column
19959 @appendixsubsec Create a Y Axis Column
19961 The preceding functions provide all the tools needed to construct a
19962 function that generates a list of numbered and blank strings to insert
19963 as the label for the vertical axis:
19965 @findex Y-axis-column
19968 (defun Y-axis-column (height width-of-label)
19969 "Construct list of Y axis labels and blank strings.
19970 For HEIGHT of line above base and WIDTH-OF-LABEL."
19974 (while (> height 1)
19975 (if (zerop (% height Y-axis-label-spacing))
19976 ;; @r{Insert label.}
19979 (Y-axis-element height width-of-label)
19983 ;; @r{Else, insert blanks.}
19986 (make-string width-of-label ? )
19988 (setq height (1- height)))
19989 ;; @r{Insert base line.}
19991 (cons (Y-axis-element 1 width-of-label) Y-axis))
19992 (nreverse Y-axis)))
19996 In this function, we start with the value of @code{height} and
19997 repetitively subtract one from its value. After each subtraction, we
19998 test to see whether the value is an integral multiple of the
19999 @code{Y-axis-label-spacing}. If it is, we construct a numbered label
20000 using the @code{Y-axis-element} function; if not, we construct a
20001 blank label using the @code{make-string} function. The base line
20002 consists of the number one followed by a tic mark.
20005 @node print-Y-axis Penultimate
20006 @appendixsubsec The Not Quite Final Version of @code{print-Y-axis}
20008 The list constructed by the @code{Y-axis-column} function is passed to
20009 the @code{print-Y-axis} function, which inserts the list as a column.
20011 @findex print-Y-axis
20014 (defun print-Y-axis (height full-Y-label-width)
20015 "Insert Y axis using HEIGHT and FULL-Y-LABEL-WIDTH.
20016 Height must be the maximum height of the graph.
20017 Full width is the width of the highest label element."
20018 ;; Value of height and full-Y-label-width
20019 ;; are passed by `print-graph'.
20022 (let ((start (point)))
20024 (Y-axis-column height full-Y-label-width))
20025 ;; @r{Place point ready for inserting graph.}
20027 ;; @r{Move point forward by value of} full-Y-label-width
20028 (forward-char full-Y-label-width)))
20032 The @code{print-Y-axis} uses the @code{insert-rectangle} function to
20033 insert the Y axis labels created by the @code{Y-axis-column} function.
20034 In addition, it places point at the correct position for printing the body of
20037 You can test @code{print-Y-axis}:
20045 Y-axis-label-spacing
20054 Copy the following expression:
20057 (print-Y-axis 12 5)
20061 Switch to the @file{*scratch*} buffer and place the cursor where you
20062 want the axis labels to start.
20065 Type @kbd{M-:} (@code{eval-expression}).
20068 Yank the @code{graph-body-print} expression into the minibuffer
20069 with @kbd{C-y} (@code{yank)}.
20072 Press @key{RET} to evaluate the expression.
20075 Emacs will print labels vertically, the top one being @w{@samp{10 -@w{
20076 }}}. (The @code{print-graph} function will pass the value of
20077 @code{height-of-top-line}, which in this case will end up as 15,
20078 thereby getting rid of what might appear as a bug.)
20082 @appendixsec The @code{print-X-axis} Function
20083 @cindex Axis, print horizontal
20084 @cindex X axis printing
20085 @cindex Print horizontal axis
20086 @cindex Horizontal axis printing
20088 X axis labels are much like Y axis labels, except that the ticks are on a
20089 line above the numbers. Labels should look like this:
20098 The first tic is under the first column of the graph and is preceded by
20099 several blank spaces. These spaces provide room in rows above for the Y
20100 axis labels. The second, third, fourth, and subsequent ticks are all
20101 spaced equally, according to the value of @code{X-axis-label-spacing}.
20103 The second row of the X axis consists of numbers, preceded by several
20104 blank spaces and also separated according to the value of the variable
20105 @code{X-axis-label-spacing}.
20107 The value of the variable @code{X-axis-label-spacing} should itself be
20108 measured in units of @code{symbol-width}, since you may want to change
20109 the width of the symbols that you are using to print the body of the
20110 graph without changing the ways the graph is labeled.
20113 * Similarities differences:: Much like @code{print-Y-axis}, but not exactly.
20114 * X Axis Tic Marks:: Create tic marks for the horizontal axis.
20118 @node Similarities differences
20119 @unnumberedsubsec Similarities and differences
20122 The @code{print-X-axis} function is constructed in more or less the
20123 same fashion as the @code{print-Y-axis} function except that it has
20124 two lines: the line of tic marks and the numbers. We will write a
20125 separate function to print each line and then combine them within the
20126 @code{print-X-axis} function.
20128 This is a three step process:
20132 Write a function to print the X axis tic marks, @code{print-X-axis-tic-line}.
20135 Write a function to print the X numbers, @code{print-X-axis-numbered-line}.
20138 Write a function to print both lines, the @code{print-X-axis} function,
20139 using @code{print-X-axis-tic-line} and
20140 @code{print-X-axis-numbered-line}.
20143 @node X Axis Tic Marks
20144 @appendixsubsec X Axis Tic Marks
20146 The first function should print the X axis tic marks. We must specify
20147 the tic marks themselves and their spacing:
20151 (defvar X-axis-label-spacing
20152 (if (boundp 'graph-blank)
20153 (* 5 (length graph-blank)) 5)
20154 "Number of units from one X axis label to next.")
20159 (Note that the value of @code{graph-blank} is set by another
20160 @code{defvar}. The @code{boundp} predicate checks whether it has
20161 already been set; @code{boundp} returns @code{nil} if it has not. If
20162 @code{graph-blank} were unbound and we did not use this conditional
20163 construction, in a recent GNU Emacs, we would enter the debugger and
20164 see an error message saying @samp{@w{Debugger entered--Lisp error:}
20165 @w{(void-variable graph-blank)}}.)
20168 Here is the @code{defvar} for @code{X-axis-tic-symbol}:
20172 (defvar X-axis-tic-symbol "|"
20173 "String to insert to point to a column in X axis.")
20178 The goal is to make a line that looks like this:
20184 The first tic is indented so that it is under the first column, which is
20185 indented to provide space for the Y axis labels.
20187 A tic element consists of the blank spaces that stretch from one tic to
20188 the next plus a tic symbol. The number of blanks is determined by the
20189 width of the tic symbol and the @code{X-axis-label-spacing}.
20192 The code looks like this:
20196 ;;; X-axis-tic-element
20200 ;; @r{Make a string of blanks.}
20201 (- (* symbol-width X-axis-label-spacing)
20202 (length X-axis-tic-symbol))
20204 ;; @r{Concatenate blanks with tic symbol.}
20210 Next, we determine how many blanks are needed to indent the first tic
20211 mark to the first column of the graph. This uses the value of
20212 @code{full-Y-label-width} passed it by the @code{print-graph} function.
20215 The code to make @code{X-axis-leading-spaces}
20220 ;; X-axis-leading-spaces
20222 (make-string full-Y-label-width ? )
20227 We also need to determine the length of the horizontal axis, which is
20228 the length of the numbers list, and the number of ticks in the horizontal
20235 (length numbers-list)
20241 (* symbol-width X-axis-label-spacing)
20245 ;; number-of-X-ticks
20246 (if (zerop (% (X-length tic-width)))
20247 (/ (X-length tic-width))
20248 (1+ (/ (X-length tic-width))))
20253 All this leads us directly to the function for printing the X axis tic line:
20255 @findex print-X-axis-tic-line
20258 (defun print-X-axis-tic-line
20259 (number-of-X-tics X-axis-leading-spaces X-axis-tic-element)
20260 "Print ticks for X axis."
20261 (insert X-axis-leading-spaces)
20262 (insert X-axis-tic-symbol) ; @r{Under first column.}
20265 ;; @r{Insert second tic in the right spot.}
20268 (- (* symbol-width X-axis-label-spacing)
20269 ;; @r{Insert white space up to second tic symbol.}
20270 (* 2 (length X-axis-tic-symbol)))
20272 X-axis-tic-symbol))
20275 ;; @r{Insert remaining ticks.}
20276 (while (> number-of-X-tics 1)
20277 (insert X-axis-tic-element)
20278 (setq number-of-X-tics (1- number-of-X-tics))))
20282 The line of numbers is equally straightforward:
20285 First, we create a numbered element with blank spaces before each number:
20287 @findex X-axis-element
20290 (defun X-axis-element (number)
20291 "Construct a numbered X axis element."
20292 (let ((leading-spaces
20293 (- (* symbol-width X-axis-label-spacing)
20294 (length (number-to-string number)))))
20295 (concat (make-string leading-spaces ? )
20296 (number-to-string number))))
20300 Next, we create the function to print the numbered line, starting with
20301 the number ``1'' under the first column:
20303 @findex print-X-axis-numbered-line
20306 (defun print-X-axis-numbered-line
20307 (number-of-X-tics X-axis-leading-spaces)
20308 "Print line of X-axis numbers"
20309 (let ((number X-axis-label-spacing))
20310 (insert X-axis-leading-spaces)
20316 ;; @r{Insert white space up to next number.}
20317 (- (* symbol-width X-axis-label-spacing) 2)
20319 (number-to-string number)))
20322 ;; @r{Insert remaining numbers.}
20323 (setq number (+ number X-axis-label-spacing))
20324 (while (> number-of-X-tics 1)
20325 (insert (X-axis-element number))
20326 (setq number (+ number X-axis-label-spacing))
20327 (setq number-of-X-tics (1- number-of-X-tics)))))
20331 Finally, we need to write the @code{print-X-axis} that uses
20332 @code{print-X-axis-tic-line} and
20333 @code{print-X-axis-numbered-line}.
20335 The function must determine the local values of the variables used by both
20336 @code{print-X-axis-tic-line} and @code{print-X-axis-numbered-line}, and
20337 then it must call them. Also, it must print the carriage return that
20338 separates the two lines.
20340 The function consists of a varlist that specifies five local variables,
20341 and calls to each of the two line printing functions:
20343 @findex print-X-axis
20346 (defun print-X-axis (numbers-list)
20347 "Print X axis labels to length of NUMBERS-LIST."
20348 (let* ((leading-spaces
20349 (make-string full-Y-label-width ? ))
20352 ;; symbol-width @r{is provided by} graph-body-print
20353 (tic-width (* symbol-width X-axis-label-spacing))
20354 (X-length (length numbers-list))
20362 ;; @r{Make a string of blanks.}
20363 (- (* symbol-width X-axis-label-spacing)
20364 (length X-axis-tic-symbol))
20368 ;; @r{Concatenate blanks with tic symbol.}
20369 X-axis-tic-symbol))
20373 (if (zerop (% X-length tic-width))
20374 (/ X-length tic-width)
20375 (1+ (/ X-length tic-width)))))
20378 (print-X-axis-tic-line tic-number leading-spaces X-tic)
20380 (print-X-axis-numbered-line tic-number leading-spaces)))
20385 You can test @code{print-X-axis}:
20389 Install @code{X-axis-tic-symbol}, @code{X-axis-label-spacing},
20390 @code{print-X-axis-tic-line}, as well as @code{X-axis-element},
20391 @code{print-X-axis-numbered-line}, and @code{print-X-axis}.
20394 Copy the following expression:
20399 (let ((full-Y-label-width 5)
20402 '(1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16))))
20407 Switch to the @file{*scratch*} buffer and place the cursor where you
20408 want the axis labels to start.
20411 Type @kbd{M-:} (@code{eval-expression}).
20414 Yank the test expression into the minibuffer
20415 with @kbd{C-y} (@code{yank)}.
20418 Press @key{RET} to evaluate the expression.
20422 Emacs will print the horizontal axis like this:
20432 @node Print Whole Graph
20433 @appendixsec Printing the Whole Graph
20434 @cindex Printing the whole graph
20435 @cindex Whole graph printing
20436 @cindex Graph, printing all
20438 Now we are nearly ready to print the whole graph.
20440 The function to print the graph with the proper labels follows the
20441 outline we created earlier (@pxref{Full Graph, , A Graph with Labeled
20442 Axes}), but with additions.
20445 Here is the outline:
20449 (defun print-graph (numbers-list)
20450 "@var{documentation}@dots{}"
20451 (let ((height @dots{}
20455 (print-Y-axis height @dots{} )
20456 (graph-body-print numbers-list)
20457 (print-X-axis @dots{} )))
20462 * The final version:: A few changes.
20463 * Test print-graph:: Run a short test.
20464 * Graphing words in defuns:: Executing the final code.
20465 * lambda:: How to write an anonymous function.
20466 * mapcar:: Apply a function to elements of a list.
20467 * Another Bug:: Yet another bug @dots{} most insidious.
20468 * Final printed graph:: The graph itself!
20472 @node The final version
20473 @unnumberedsubsec Changes for the Final Version
20476 The final version is different from what we planned in two ways:
20477 first, it contains additional values calculated once in the varlist;
20478 second, it carries an option to specify the labels' increment per row.
20479 This latter feature turns out to be essential; otherwise, a graph may
20480 have more rows than fit on a display or on a sheet of paper.
20483 This new feature requires a change to the @code{Y-axis-column}
20484 function, to add @code{vertical-step} to it. The function looks like
20487 @findex Y-axis-column @r{Final version.}
20490 ;;; @r{Final version.}
20491 (defun Y-axis-column
20492 (height width-of-label &optional vertical-step)
20493 "Construct list of labels for Y axis.
20494 HEIGHT is maximum height of graph.
20495 WIDTH-OF-LABEL is maximum width of label.
20496 VERTICAL-STEP, an option, is a positive integer
20497 that specifies how much a Y axis label increments
20498 for each line. For example, a step of 5 means
20499 that each line is five units of the graph."
20503 (number-per-line (or vertical-step 1)))
20504 (while (> height 1)
20505 (if (zerop (% height Y-axis-label-spacing))
20508 ;; @r{Insert label.}
20512 (* height number-per-line)
20517 ;; @r{Else, insert blanks.}
20520 (make-string width-of-label ? )
20522 (setq height (1- height)))
20525 ;; @r{Insert base line.}
20526 (setq Y-axis (cons (Y-axis-element
20527 (or vertical-step 1)
20530 (nreverse Y-axis)))
20534 The values for the maximum height of graph and the width of a symbol
20535 are computed by @code{print-graph} in its @code{let} expression; so
20536 @code{graph-body-print} must be changed to accept them.
20538 @findex graph-body-print @r{Final version.}
20541 ;;; @r{Final version.}
20542 (defun graph-body-print (numbers-list height symbol-width)
20543 "Print a bar graph of the NUMBERS-LIST.
20544 The numbers-list consists of the Y-axis values.
20545 HEIGHT is maximum height of graph.
20546 SYMBOL-WIDTH is number of each column."
20549 (let (from-position)
20550 (while numbers-list
20551 (setq from-position (point))
20553 (column-of-graph height (car numbers-list)))
20554 (goto-char from-position)
20555 (forward-char symbol-width)
20558 ;; @r{Draw graph column by column.}
20560 (setq numbers-list (cdr numbers-list)))
20561 ;; @r{Place point for X axis labels.}
20562 (forward-line height)
20568 Finally, the code for the @code{print-graph} function:
20570 @findex print-graph @r{Final version.}
20573 ;;; @r{Final version.}
20575 (numbers-list &optional vertical-step)
20576 "Print labeled bar graph of the NUMBERS-LIST.
20577 The numbers-list consists of the Y-axis values.
20581 Optionally, VERTICAL-STEP, a positive integer,
20582 specifies how much a Y axis label increments for
20583 each line. For example, a step of 5 means that
20584 each row is five units."
20587 (let* ((symbol-width (length graph-blank))
20588 ;; @code{height} @r{is both the largest number}
20589 ;; @r{and the number with the most digits.}
20590 (height (apply 'max numbers-list))
20593 (height-of-top-line
20594 (if (zerop (% height Y-axis-label-spacing))
20597 (* (1+ (/ height Y-axis-label-spacing))
20598 Y-axis-label-spacing)))
20601 (vertical-step (or vertical-step 1))
20602 (full-Y-label-width
20608 (* height-of-top-line vertical-step))
20614 height-of-top-line full-Y-label-width vertical-step)
20618 numbers-list height-of-top-line symbol-width)
20619 (print-X-axis numbers-list)))
20623 @node Test print-graph
20624 @appendixsubsec Testing @code{print-graph}
20627 We can test the @code{print-graph} function with a short list of numbers:
20631 Install the final versions of @code{Y-axis-column},
20632 @code{graph-body-print}, and @code{print-graph} (in addition to the
20636 Copy the following expression:
20639 (print-graph '(3 2 5 6 7 5 3 4 6 4 3 2 1))
20643 Switch to the @file{*scratch*} buffer and place the cursor where you
20644 want the axis labels to start.
20647 Type @kbd{M-:} (@code{eval-expression}).
20650 Yank the test expression into the minibuffer
20651 with @kbd{C-y} (@code{yank)}.
20654 Press @key{RET} to evaluate the expression.
20658 Emacs will print a graph that looks like this:
20679 On the other hand, if you pass @code{print-graph} a
20680 @code{vertical-step} value of 2, by evaluating this expression:
20683 (print-graph '(3 2 5 6 7 5 3 4 6 4 3 2 1) 2)
20688 The graph looks like this:
20709 (A question: is the `2' on the bottom of the vertical axis a bug or a
20710 feature? If you think it is a bug, and should be a `1' instead, (or
20711 even a `0'), you can modify the sources.)
20713 @node Graphing words in defuns
20714 @appendixsubsec Graphing Numbers of Words and Symbols
20716 Now for the graph for which all this code was written: a graph that
20717 shows how many function definitions contain fewer than 10 words and
20718 symbols, how many contain between 10 and 19 words and symbols, how
20719 many contain between 20 and 29 words and symbols, and so on.
20721 This is a multi-step process. First make sure you have loaded all the
20725 It is a good idea to reset the value of @code{top-of-ranges} in case
20726 you have set it to some different value. You can evaluate the
20731 (setq top-of-ranges
20734 110 120 130 140 150
20735 160 170 180 190 200
20736 210 220 230 240 250
20737 260 270 280 290 300)
20742 Next create a list of the number of words and symbols in each range.
20746 Evaluate the following:
20750 (setq list-for-graph
20753 (recursive-lengths-list-many-files
20754 (directory-files "/usr/local/emacs/lisp"
20762 On my old machine, this took about an hour. It looked though 303 Lisp
20763 files in my copy of Emacs version 19.23. After all that computing,
20764 the @code{list-for-graph} had this value:
20768 (537 1027 955 785 594 483 349 292 224 199 166 120 116 99
20769 90 80 67 48 52 45 41 33 28 26 25 20 12 28 11 13 220)
20774 This means that my copy of Emacs had 537 function definitions with
20775 fewer than 10 words or symbols in them, 1,027 function definitions
20776 with 10 to 19 words or symbols in them, 955 function definitions with
20777 20 to 29 words or symbols in them, and so on.
20779 Clearly, just by looking at this list we can see that most function
20780 definitions contain ten to thirty words and symbols.
20782 Now for printing. We do @emph{not} want to print a graph that is
20783 1,030 lines high @dots{} Instead, we should print a graph that is
20784 fewer than twenty-five lines high. A graph that height can be
20785 displayed on almost any monitor, and easily printed on a sheet of paper.
20787 This means that each value in @code{list-for-graph} must be reduced to
20788 one-fiftieth its present value.
20790 Here is a short function to do just that, using two functions we have
20791 not yet seen, @code{mapcar} and @code{lambda}.
20795 (defun one-fiftieth (full-range)
20796 "Return list, each number one-fiftieth of previous."
20797 (mapcar (lambda (arg) (/ arg 50)) full-range))
20802 @appendixsubsec A @code{lambda} Expression: Useful Anonymity
20803 @cindex Anonymous function
20806 @code{lambda} is the symbol for an anonymous function, a function
20807 without a name. Every time you use an anonymous function, you need to
20808 include its whole body.
20815 (lambda (arg) (/ arg 50))
20819 is a function definition that says `return the value resulting from
20820 dividing whatever is passed to me as @code{arg} by 50'.
20823 Earlier, for example, we had a function @code{multiply-by-seven}; it
20824 multiplied its argument by 7. This function is similar, except it
20825 divides its argument by 50; and, it has no name. The anonymous
20826 equivalent of @code{multiply-by-seven} is:
20829 (lambda (number) (* 7 number))
20833 (@xref{defun, , The @code{defun} Macro}.)
20837 If we want to multiply 3 by 7, we can write:
20839 @c clear print-postscript-figures
20840 @c lambda example diagram #1
20844 (multiply-by-seven 3)
20845 \_______________/ ^
20851 @ifset print-postscript-figures
20854 @center @image{lambda-1}
20858 @ifclear print-postscript-figures
20862 (multiply-by-seven 3)
20863 \_______________/ ^
20872 This expression returns 21.
20876 Similarly, we can write:
20878 @c lambda example diagram #2
20882 ((lambda (number) (* 7 number)) 3)
20883 \____________________________/ ^
20885 anonymous function argument
20889 @ifset print-postscript-figures
20892 @center @image{lambda-2}
20896 @ifclear print-postscript-figures
20900 ((lambda (number) (* 7 number)) 3)
20901 \____________________________/ ^
20903 anonymous function argument
20911 If we want to divide 100 by 50, we can write:
20913 @c lambda example diagram #3
20917 ((lambda (arg) (/ arg 50)) 100)
20918 \______________________/ \_/
20920 anonymous function argument
20924 @ifset print-postscript-figures
20927 @center @image{lambda-3}
20931 @ifclear print-postscript-figures
20935 ((lambda (arg) (/ arg 50)) 100)
20936 \______________________/ \_/
20938 anonymous function argument
20945 This expression returns 2. The 100 is passed to the function, which
20946 divides that number by 50.
20948 @xref{Lambda Expressions, , Lambda Expressions, elisp, The GNU Emacs
20949 Lisp Reference Manual}, for more about @code{lambda}. Lisp and lambda
20950 expressions derive from the Lambda Calculus.
20953 @appendixsubsec The @code{mapcar} Function
20956 @code{mapcar} is a function that calls its first argument with each
20957 element of its second argument, in turn. The second argument must be
20960 The @samp{map} part of the name comes from the mathematical phrase,
20961 `mapping over a domain', meaning to apply a function to each of the
20962 elements in a domain. The mathematical phrase is based on the
20963 metaphor of a surveyor walking, one step at a time, over an area he is
20964 mapping. And @samp{car}, of course, comes from the Lisp notion of the
20973 (mapcar '1+ '(2 4 6))
20979 The function @code{1+} which adds one to its argument, is executed on
20980 @emph{each} element of the list, and a new list is returned.
20982 Contrast this with @code{apply}, which applies its first argument to
20984 (@xref{Readying a Graph, , Readying a Graph}, for a explanation of
20988 In the definition of @code{one-fiftieth}, the first argument is the
20989 anonymous function:
20992 (lambda (arg) (/ arg 50))
20996 and the second argument is @code{full-range}, which will be bound to
20997 @code{list-for-graph}.
21000 The whole expression looks like this:
21003 (mapcar (lambda (arg) (/ arg 50)) full-range))
21006 @xref{Mapping Functions, , Mapping Functions, elisp, The GNU Emacs
21007 Lisp Reference Manual}, for more about @code{mapcar}.
21009 Using the @code{one-fiftieth} function, we can generate a list in
21010 which each element is one-fiftieth the size of the corresponding
21011 element in @code{list-for-graph}.
21015 (setq fiftieth-list-for-graph
21016 (one-fiftieth list-for-graph))
21021 The resulting list looks like this:
21025 (10 20 19 15 11 9 6 5 4 3 3 2 2
21026 1 1 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 4)
21031 This, we are almost ready to print! (We also notice the loss of
21032 information: many of the higher ranges are 0, meaning that fewer than
21033 50 defuns had that many words or symbols---but not necessarily meaning
21034 that none had that many words or symbols.)
21037 @appendixsubsec Another Bug @dots{} Most Insidious
21038 @cindex Bug, most insidious type
21039 @cindex Insidious type of bug
21041 I said `almost ready to print'! Of course, there is a bug in the
21042 @code{print-graph} function @dots{} It has a @code{vertical-step}
21043 option, but not a @code{horizontal-step} option. The
21044 @code{top-of-range} scale goes from 10 to 300 by tens. But the
21045 @code{print-graph} function will print only by ones.
21047 This is a classic example of what some consider the most insidious
21048 type of bug, the bug of omission. This is not the kind of bug you can
21049 find by studying the code, for it is not in the code; it is an omitted
21050 feature. Your best actions are to try your program early and often;
21051 and try to arrange, as much as you can, to write code that is easy to
21052 understand and easy to change. Try to be aware, whenever you can,
21053 that whatever you have written, @emph{will} be rewritten, if not soon,
21054 eventually. A hard maxim to follow.
21056 It is the @code{print-X-axis-numbered-line} function that needs the
21057 work; and then the @code{print-X-axis} and the @code{print-graph}
21058 functions need to be adapted. Not much needs to be done; there is one
21059 nicety: the numbers ought to line up under the tic marks. This takes
21063 Here is the corrected @code{print-X-axis-numbered-line}:
21067 (defun print-X-axis-numbered-line
21068 (number-of-X-tics X-axis-leading-spaces
21069 &optional horizontal-step)
21070 "Print line of X-axis numbers"
21071 (let ((number X-axis-label-spacing)
21072 (horizontal-step (or horizontal-step 1)))
21075 (insert X-axis-leading-spaces)
21076 ;; @r{Delete extra leading spaces.}
21079 (length (number-to-string horizontal-step)))))
21084 ;; @r{Insert white space.}
21086 X-axis-label-spacing)
21089 (number-to-string horizontal-step)))
21093 (* number horizontal-step))))
21096 ;; @r{Insert remaining numbers.}
21097 (setq number (+ number X-axis-label-spacing))
21098 (while (> number-of-X-tics 1)
21099 (insert (X-axis-element
21100 (* number horizontal-step)))
21101 (setq number (+ number X-axis-label-spacing))
21102 (setq number-of-X-tics (1- number-of-X-tics)))))
21107 If you are reading this in Info, you can see the new versions of
21108 @code{print-X-axis} @code{print-graph} and evaluate them. If you are
21109 reading this in a printed book, you can see the changed lines here
21110 (the full text is too much to print).
21115 (defun print-X-axis (numbers-list horizontal-step)
21117 (print-X-axis-numbered-line
21118 tic-number leading-spaces horizontal-step))
21126 &optional vertical-step horizontal-step)
21128 (print-X-axis numbers-list horizontal-step))
21136 (defun print-X-axis (numbers-list horizontal-step)
21137 "Print X axis labels to length of NUMBERS-LIST.
21138 Optionally, HORIZONTAL-STEP, a positive integer,
21139 specifies how much an X axis label increments for
21143 ;; Value of symbol-width and full-Y-label-width
21144 ;; are passed by `print-graph'.
21145 (let* ((leading-spaces
21146 (make-string full-Y-label-width ? ))
21147 ;; symbol-width @r{is provided by} graph-body-print
21148 (tic-width (* symbol-width X-axis-label-spacing))
21149 (X-length (length numbers-list))
21155 ;; @r{Make a string of blanks.}
21156 (- (* symbol-width X-axis-label-spacing)
21157 (length X-axis-tic-symbol))
21161 ;; @r{Concatenate blanks with tic symbol.}
21162 X-axis-tic-symbol))
21164 (if (zerop (% X-length tic-width))
21165 (/ X-length tic-width)
21166 (1+ (/ X-length tic-width)))))
21170 (print-X-axis-tic-line
21171 tic-number leading-spaces X-tic)
21173 (print-X-axis-numbered-line
21174 tic-number leading-spaces horizontal-step)))
21181 (numbers-list &optional vertical-step horizontal-step)
21182 "Print labeled bar graph of the NUMBERS-LIST.
21183 The numbers-list consists of the Y-axis values.
21187 Optionally, VERTICAL-STEP, a positive integer,
21188 specifies how much a Y axis label increments for
21189 each line. For example, a step of 5 means that
21190 each row is five units.
21194 Optionally, HORIZONTAL-STEP, a positive integer,
21195 specifies how much an X axis label increments for
21197 (let* ((symbol-width (length graph-blank))
21198 ;; @code{height} @r{is both the largest number}
21199 ;; @r{and the number with the most digits.}
21200 (height (apply 'max numbers-list))
21203 (height-of-top-line
21204 (if (zerop (% height Y-axis-label-spacing))
21207 (* (1+ (/ height Y-axis-label-spacing))
21208 Y-axis-label-spacing)))
21211 (vertical-step (or vertical-step 1))
21212 (full-Y-label-width
21216 (* height-of-top-line vertical-step))
21221 height-of-top-line full-Y-label-width vertical-step)
21223 numbers-list height-of-top-line symbol-width)
21224 (print-X-axis numbers-list horizontal-step)))
21231 Graphing Definitions Re-listed
21234 Here are all the graphing definitions in their final form:
21238 (defvar top-of-ranges
21241 110 120 130 140 150
21242 160 170 180 190 200
21243 210 220 230 240 250)
21244 "List specifying ranges for `defuns-per-range'.")
21248 (defvar graph-symbol "*"
21249 "String used as symbol in graph, usually an asterisk.")
21253 (defvar graph-blank " "
21254 "String used as blank in graph, usually a blank space.
21255 graph-blank must be the same number of columns wide
21260 (defvar Y-axis-tic " - "
21261 "String that follows number in a Y axis label.")
21265 (defvar Y-axis-label-spacing 5
21266 "Number of lines from one Y axis label to next.")
21270 (defvar X-axis-tic-symbol "|"
21271 "String to insert to point to a column in X axis.")
21275 (defvar X-axis-label-spacing
21276 (if (boundp 'graph-blank)
21277 (* 5 (length graph-blank)) 5)
21278 "Number of units from one X axis label to next.")
21284 (defun count-words-in-defun ()
21285 "Return the number of words and symbols in a defun."
21286 (beginning-of-defun)
21288 (end (save-excursion (end-of-defun) (point))))
21293 (and (< (point) end)
21295 "\\(\\w\\|\\s_\\)+[^ \t\n]*[ \t\n]*"
21297 (setq count (1+ count)))
21304 (defun lengths-list-file (filename)
21305 "Return list of definitions' lengths within FILE.
21306 The returned list is a list of numbers.
21307 Each number is the number of words or
21308 symbols in one function definition."
21312 (message "Working on `%s' ... " filename)
21314 (let ((buffer (find-file-noselect filename))
21316 (set-buffer buffer)
21317 (setq buffer-read-only t)
21319 (goto-char (point-min))
21323 (while (re-search-forward "^(defun" nil t)
21325 (cons (count-words-in-defun) lengths-list)))
21326 (kill-buffer buffer)
21333 (defun lengths-list-many-files (list-of-files)
21334 "Return list of lengths of defuns in LIST-OF-FILES."
21335 (let (lengths-list)
21336 ;;; @r{true-or-false-test}
21337 (while list-of-files
21343 ;;; @r{Generate a lengths' list.}
21345 (expand-file-name (car list-of-files)))))
21346 ;;; @r{Make files' list shorter.}
21347 (setq list-of-files (cdr list-of-files)))
21348 ;;; @r{Return final value of lengths' list.}
21355 (defun defuns-per-range (sorted-lengths top-of-ranges)
21356 "SORTED-LENGTHS defuns in each TOP-OF-RANGES range."
21357 (let ((top-of-range (car top-of-ranges))
21358 (number-within-range 0)
21359 defuns-per-range-list)
21364 (while top-of-ranges
21368 ;; @r{Need number for numeric test.}
21369 (car sorted-lengths)
21370 (< (car sorted-lengths) top-of-range))
21372 ;; @r{Count number of definitions within current range.}
21373 (setq number-within-range (1+ number-within-range))
21374 (setq sorted-lengths (cdr sorted-lengths)))
21378 ;; @r{Exit inner loop but remain within outer loop.}
21380 (setq defuns-per-range-list
21381 (cons number-within-range defuns-per-range-list))
21382 (setq number-within-range 0) ; @r{Reset count to zero.}
21384 ;; @r{Move to next range.}
21385 (setq top-of-ranges (cdr top-of-ranges))
21386 ;; @r{Specify next top of range value.}
21387 (setq top-of-range (car top-of-ranges)))
21391 ;; @r{Exit outer loop and count the number of defuns larger than}
21392 ;; @r{ the largest top-of-range value.}
21393 (setq defuns-per-range-list
21395 (length sorted-lengths)
21396 defuns-per-range-list))
21398 ;; @r{Return a list of the number of definitions within each range,}
21399 ;; @r{ smallest to largest.}
21400 (nreverse defuns-per-range-list)))
21406 (defun column-of-graph (max-graph-height actual-height)
21407 "Return list of MAX-GRAPH-HEIGHT strings;
21408 ACTUAL-HEIGHT are graph-symbols.
21409 The graph-symbols are contiguous entries at the end
21411 The list will be inserted as one column of a graph.
21412 The strings are either graph-blank or graph-symbol."
21416 (let ((insert-list nil)
21417 (number-of-top-blanks
21418 (- max-graph-height actual-height)))
21420 ;; @r{Fill in @code{graph-symbols}.}
21421 (while (> actual-height 0)
21422 (setq insert-list (cons graph-symbol insert-list))
21423 (setq actual-height (1- actual-height)))
21427 ;; @r{Fill in @code{graph-blanks}.}
21428 (while (> number-of-top-blanks 0)
21429 (setq insert-list (cons graph-blank insert-list))
21430 (setq number-of-top-blanks
21431 (1- number-of-top-blanks)))
21433 ;; @r{Return whole list.}
21440 (defun Y-axis-element (number full-Y-label-width)
21441 "Construct a NUMBERed label element.
21442 A numbered element looks like this ` 5 - ',
21443 and is padded as needed so all line up with
21444 the element for the largest number."
21447 (let* ((leading-spaces
21448 (- full-Y-label-width
21450 (concat (number-to-string number)
21455 (make-string leading-spaces ? )
21456 (number-to-string number)
21463 (defun print-Y-axis
21464 (height full-Y-label-width &optional vertical-step)
21465 "Insert Y axis by HEIGHT and FULL-Y-LABEL-WIDTH.
21466 Height must be the maximum height of the graph.
21467 Full width is the width of the highest label element.
21468 Optionally, print according to VERTICAL-STEP."
21471 ;; Value of height and full-Y-label-width
21472 ;; are passed by `print-graph'.
21473 (let ((start (point)))
21475 (Y-axis-column height full-Y-label-width vertical-step))
21478 ;; @r{Place point ready for inserting graph.}
21480 ;; @r{Move point forward by value of} full-Y-label-width
21481 (forward-char full-Y-label-width)))
21487 (defun print-X-axis-tic-line
21488 (number-of-X-tics X-axis-leading-spaces X-axis-tic-element)
21489 "Print ticks for X axis."
21490 (insert X-axis-leading-spaces)
21491 (insert X-axis-tic-symbol) ; @r{Under first column.}
21494 ;; @r{Insert second tic in the right spot.}
21497 (- (* symbol-width X-axis-label-spacing)
21498 ;; @r{Insert white space up to second tic symbol.}
21499 (* 2 (length X-axis-tic-symbol)))
21501 X-axis-tic-symbol))
21504 ;; @r{Insert remaining ticks.}
21505 (while (> number-of-X-tics 1)
21506 (insert X-axis-tic-element)
21507 (setq number-of-X-tics (1- number-of-X-tics))))
21513 (defun X-axis-element (number)
21514 "Construct a numbered X axis element."
21515 (let ((leading-spaces
21516 (- (* symbol-width X-axis-label-spacing)
21517 (length (number-to-string number)))))
21518 (concat (make-string leading-spaces ? )
21519 (number-to-string number))))
21525 (defun graph-body-print (numbers-list height symbol-width)
21526 "Print a bar graph of the NUMBERS-LIST.
21527 The numbers-list consists of the Y-axis values.
21528 HEIGHT is maximum height of graph.
21529 SYMBOL-WIDTH is number of each column."
21532 (let (from-position)
21533 (while numbers-list
21534 (setq from-position (point))
21536 (column-of-graph height (car numbers-list)))
21537 (goto-char from-position)
21538 (forward-char symbol-width)
21541 ;; @r{Draw graph column by column.}
21543 (setq numbers-list (cdr numbers-list)))
21544 ;; @r{Place point for X axis labels.}
21545 (forward-line height)
21552 (defun Y-axis-column
21553 (height width-of-label &optional vertical-step)
21554 "Construct list of labels for Y axis.
21555 HEIGHT is maximum height of graph.
21556 WIDTH-OF-LABEL is maximum width of label.
21559 VERTICAL-STEP, an option, is a positive integer
21560 that specifies how much a Y axis label increments
21561 for each line. For example, a step of 5 means
21562 that each line is five units of the graph."
21564 (number-per-line (or vertical-step 1)))
21567 (while (> height 1)
21568 (if (zerop (% height Y-axis-label-spacing))
21569 ;; @r{Insert label.}
21573 (* height number-per-line)
21578 ;; @r{Else, insert blanks.}
21581 (make-string width-of-label ? )
21583 (setq height (1- height)))
21586 ;; @r{Insert base line.}
21587 (setq Y-axis (cons (Y-axis-element
21588 (or vertical-step 1)
21591 (nreverse Y-axis)))
21597 (defun print-X-axis-numbered-line
21598 (number-of-X-tics X-axis-leading-spaces
21599 &optional horizontal-step)
21600 "Print line of X-axis numbers"
21601 (let ((number X-axis-label-spacing)
21602 (horizontal-step (or horizontal-step 1)))
21605 (insert X-axis-leading-spaces)
21607 (delete-char (- (1- (length (number-to-string horizontal-step)))))
21610 ;; @r{Insert white space up to next number.}
21611 (- (* symbol-width X-axis-label-spacing)
21612 (1- (length (number-to-string horizontal-step)))
21615 (number-to-string (* number horizontal-step))))
21618 ;; @r{Insert remaining numbers.}
21619 (setq number (+ number X-axis-label-spacing))
21620 (while (> number-of-X-tics 1)
21621 (insert (X-axis-element (* number horizontal-step)))
21622 (setq number (+ number X-axis-label-spacing))
21623 (setq number-of-X-tics (1- number-of-X-tics)))))
21629 (defun print-X-axis (numbers-list horizontal-step)
21630 "Print X axis labels to length of NUMBERS-LIST.
21631 Optionally, HORIZONTAL-STEP, a positive integer,
21632 specifies how much an X axis label increments for
21636 ;; Value of symbol-width and full-Y-label-width
21637 ;; are passed by `print-graph'.
21638 (let* ((leading-spaces
21639 (make-string full-Y-label-width ? ))
21640 ;; symbol-width @r{is provided by} graph-body-print
21641 (tic-width (* symbol-width X-axis-label-spacing))
21642 (X-length (length numbers-list))
21648 ;; @r{Make a string of blanks.}
21649 (- (* symbol-width X-axis-label-spacing)
21650 (length X-axis-tic-symbol))
21654 ;; @r{Concatenate blanks with tic symbol.}
21655 X-axis-tic-symbol))
21657 (if (zerop (% X-length tic-width))
21658 (/ X-length tic-width)
21659 (1+ (/ X-length tic-width)))))
21663 (print-X-axis-tic-line
21664 tic-number leading-spaces X-tic)
21666 (print-X-axis-numbered-line
21667 tic-number leading-spaces horizontal-step)))
21673 (defun one-fiftieth (full-range)
21674 "Return list, each number of which is 1/50th previous."
21675 (mapcar (lambda (arg) (/ arg 50)) full-range))
21682 (numbers-list &optional vertical-step horizontal-step)
21683 "Print labeled bar graph of the NUMBERS-LIST.
21684 The numbers-list consists of the Y-axis values.
21688 Optionally, VERTICAL-STEP, a positive integer,
21689 specifies how much a Y axis label increments for
21690 each line. For example, a step of 5 means that
21691 each row is five units.
21695 Optionally, HORIZONTAL-STEP, a positive integer,
21696 specifies how much an X axis label increments for
21698 (let* ((symbol-width (length graph-blank))
21699 ;; @code{height} @r{is both the largest number}
21700 ;; @r{and the number with the most digits.}
21701 (height (apply 'max numbers-list))
21704 (height-of-top-line
21705 (if (zerop (% height Y-axis-label-spacing))
21708 (* (1+ (/ height Y-axis-label-spacing))
21709 Y-axis-label-spacing)))
21712 (vertical-step (or vertical-step 1))
21713 (full-Y-label-width
21717 (* height-of-top-line vertical-step))
21723 height-of-top-line full-Y-label-width vertical-step)
21725 numbers-list height-of-top-line symbol-width)
21726 (print-X-axis numbers-list horizontal-step)))
21733 @node Final printed graph
21734 @appendixsubsec The Printed Graph
21736 When made and installed, you can call the @code{print-graph} command
21742 (print-graph fiftieth-list-for-graph 50 10)
21772 50 - ***************** * *
21774 10 50 100 150 200 250 300 350
21781 The largest group of functions contain 10--19 words and symbols each.
21783 @node Free Software and Free Manuals
21784 @appendix Free Software and Free Manuals
21786 @strong{by Richard M. Stallman}
21789 The biggest deficiency in free operating systems is not in the
21790 software---it is the lack of good free manuals that we can include in
21791 these systems. Many of our most important programs do not come with
21792 full manuals. Documentation is an essential part of any software
21793 package; when an important free software package does not come with a
21794 free manual, that is a major gap. We have many such gaps today.
21796 Once upon a time, many years ago, I thought I would learn Perl. I got
21797 a copy of a free manual, but I found it hard to read. When I asked
21798 Perl users about alternatives, they told me that there were better
21799 introductory manuals---but those were not free.
21801 Why was this? The authors of the good manuals had written them for
21802 O'Reilly Associates, which published them with restrictive terms---no
21803 copying, no modification, source files not available---which exclude
21804 them from the free software community.
21806 That wasn't the first time this sort of thing has happened, and (to
21807 our community's great loss) it was far from the last. Proprietary
21808 manual publishers have enticed a great many authors to restrict their
21809 manuals since then. Many times I have heard a GNU user eagerly tell me
21810 about a manual that he is writing, with which he expects to help the
21811 GNU project---and then had my hopes dashed, as he proceeded to explain
21812 that he had signed a contract with a publisher that would restrict it
21813 so that we cannot use it.
21815 Given that writing good English is a rare skill among programmers, we
21816 can ill afford to lose manuals this way.
21818 Free documentation, like free software, is a matter of freedom, not
21819 price. The problem with these manuals was not that O'Reilly Associates
21820 charged a price for printed copies---that in itself is fine. The Free
21821 Software Foundation @uref{http://shop.fsf.org, sells printed copies} of
21822 free @uref{http://www.gnu.org/doc/doc.html, GNU manuals}, too.
21823 But GNU manuals are available in source code form, while these manuals
21824 are available only on paper. GNU manuals come with permission to copy
21825 and modify; the Perl manuals do not. These restrictions are the
21828 The criterion for a free manual is pretty much the same as for free
21829 software: it is a matter of giving all users certain
21830 freedoms. Redistribution (including commercial redistribution) must be
21831 permitted, so that the manual can accompany every copy of the program,
21832 on-line or on paper. Permission for modification is crucial too.
21834 As a general rule, I don't believe that it is essential for people to
21835 have permission to modify all sorts of articles and books. The issues
21836 for writings are not necessarily the same as those for software. For
21837 example, I don't think you or I are obliged to give permission to
21838 modify articles like this one, which describe our actions and our
21841 But there is a particular reason why the freedom to modify is crucial
21842 for documentation for free software. When people exercise their right
21843 to modify the software, and add or change its features, if they are
21844 conscientious they will change the manual too---so they can provide
21845 accurate and usable documentation with the modified program. A manual
21846 which forbids programmers to be conscientious and finish the job, or
21847 more precisely requires them to write a new manual from scratch if
21848 they change the program, does not fill our community's needs.
21850 While a blanket prohibition on modification is unacceptable, some
21851 kinds of limits on the method of modification pose no problem. For
21852 example, requirements to preserve the original author's copyright
21853 notice, the distribution terms, or the list of authors, are ok. It is
21854 also no problem to require modified versions to include notice that
21855 they were modified, even to have entire sections that may not be
21856 deleted or changed, as long as these sections deal with nontechnical
21857 topics. (Some GNU manuals have them.)
21859 These kinds of restrictions are not a problem because, as a practical
21860 matter, they don't stop the conscientious programmer from adapting the
21861 manual to fit the modified program. In other words, they don't block
21862 the free software community from making full use of the manual.
21864 However, it must be possible to modify all the technical content of
21865 the manual, and then distribute the result in all the usual media,
21866 through all the usual channels; otherwise, the restrictions do block
21867 the community, the manual is not free, and so we need another manual.
21869 Unfortunately, it is often hard to find someone to write another
21870 manual when a proprietary manual exists. The obstacle is that many
21871 users think that a proprietary manual is good enough---so they don't
21872 see the need to write a free manual. They do not see that the free
21873 operating system has a gap that needs filling.
21875 Why do users think that proprietary manuals are good enough? Some have
21876 not considered the issue. I hope this article will do something to
21879 Other users consider proprietary manuals acceptable for the same
21880 reason so many people consider proprietary software acceptable: they
21881 judge in purely practical terms, not using freedom as a
21882 criterion. These people are entitled to their opinions, but since
21883 those opinions spring from values which do not include freedom, they
21884 are no guide for those of us who do value freedom.
21886 Please spread the word about this issue. We continue to lose manuals
21887 to proprietary publishing. If we spread the word that proprietary
21888 manuals are not sufficient, perhaps the next person who wants to help
21889 GNU by writing documentation will realize, before it is too late, that
21890 he must above all make it free.
21892 We can also encourage commercial publishers to sell free, copylefted
21893 manuals instead of proprietary ones. One way you can help this is to
21894 check the distribution terms of a manual before you buy it, and prefer
21895 copylefted manuals to non-copylefted ones.
21899 Note: The Free Software Foundation maintains a page on its Web site
21900 that lists free books available from other publishers:@*
21901 @uref{http://www.gnu.org/doc/other-free-books.html}
21903 @node GNU Free Documentation License
21904 @appendix GNU Free Documentation License
21906 @cindex FDL, GNU Free Documentation License
21907 @include doclicense.texi
21913 MENU ENTRY: NODE NAME.
21919 @c Place biographical information on right-hand (verso) page
21922 \par\vfill\supereject
21924 \global\evenheadline={\hfil} \global\evenfootline={\hfil}
21925 \global\oddheadline={\hfil} \global\oddfootline={\hfil}
21928 % \par\vfill\supereject
21929 \global\evenheadline={\hfil} \global\evenfootline={\hfil}
21930 \global\oddheadline={\hfil} \global\oddfootline={\hfil}
21931 %\page\hbox{}%\page
21932 %\page\hbox{}%\page
21939 @c ================ Biographical information ================
21943 @center About the Author
21948 @node About the Author
21949 @unnumbered About the Author
21953 Robert J. Chassell has worked with GNU Emacs since 1985. He writes
21954 and edits, teaches Emacs and Emacs Lisp, and speaks throughout the
21955 world on software freedom. Chassell was a founding Director and
21956 Treasurer of the Free Software Foundation, Inc. He is co-author of
21957 the @cite{Texinfo} manual, and has edited more than a dozen other
21958 books. He graduated from Cambridge University, in England. He has an
21959 abiding interest in social and economic history and flies his own
21966 @c @c Prevent page number on blank verso, so eject it first.
21968 @c \par\vfill\supereject
21973 @c @evenheading @thispage @| @| @thistitle
21974 @c @oddheading @| @| @thispage