-;;; regexp-opt.el --- generate efficient regexps to match strings.
+;;; regexp-opt.el --- generate efficient regexps to match strings
-;; Copyright (C) 1994, 95, 96, 97, 98, 1999 Free Software Foundation, Inc.
+;; Copyright (C) 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2002,
+;; 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
;; Author: Simon Marshall <simon@gnu.org>
;; Maintainer: FSF
-;; Keywords: strings, regexps
-;; Version: 1.07
+;; Keywords: strings, regexps, extensions
;; This file is part of GNU Emacs.
;; You should have received a copy of the GNU General Public License
;; along with GNU Emacs; see the file COPYING. If not, write to the
-;; Free Software Foundation, Inc., 59 Temple Place - Suite 330,
-;; Boston, MA 02111-1307, USA.
+;; Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
+;; Boston, MA 02110-1301, USA.
;;; Commentary:
;;
;; Searching using the above example `regexp-opt' regexp takes approximately
;; two-thirds of the time taken using the equivalent `mapconcat' regexp.
-;;
-;; Note that this package will also find common suffix strings if this does not
-;; increase the number of grouping constructs. For example:
-;;
-;; (regexp-opt '("these" "those"))
-;; => "th[eo]se"
-;;
-;; but:
-;;
-;; (regexp-opt '("barfly" "housefly"))
-;; => "barfly\\|housefly" rather than "\\(bar\\|house\\)fly"
;; Since this package was written to produce efficient regexps, not regexps
;; efficiently, it is probably not a good idea to in-line too many calls in
;; Stefan Monnier.
;; No doubt `regexp-opt' doesn't always produce optimal regexps, so code, ideas
;; or any other information to improve things are welcome.
+;;
+;; One possible improvement would be to compile '("aa" "ab" "ba" "bb")
+;; into "[ab][ab]" rather than "a[ab]\\|b[ab]". I'm not sure it's worth
+;; it but if someone knows how to do it without going through too many
+;; contortions, I'm all ears.
\f
-;;; Code.
+;;; Code:
;;;###autoload
(defun regexp-opt (strings &optional paren)
is enclosed by at least one regexp grouping construct.
The returned regexp is typically more efficient than the equivalent regexp:
- (let ((open-paren (if PAREN \"\\\\(\" \"\")) (close-paren (if PAREN \"\\\\)\" \"\")))
- (concat open-paren (mapconcat 'regexp-quote STRINGS \"\\\\|\") close-paren))
+ (let ((open (if PAREN \"\\\\(\" \"\")) (close (if PAREN \"\\\\)\" \"\")))
+ (concat open (mapconcat 'regexp-quote STRINGS \"\\\\|\") close))
-but typically contains more regexp grouping constructs.
-Use `regexp-opt-depth' to count them."
+If PAREN is `words', then the resulting regexp is additionally surrounded
+by \\=\\< and \\>."
(save-match-data
;; Recurse on the sorted list.
- (let ((max-lisp-eval-depth (* 1024 1024))
- (completion-ignore-case nil))
- (regexp-opt-group (sort (copy-sequence strings) 'string-lessp) paren))))
+ (let* ((max-lisp-eval-depth (* 1024 1024))
+ (max-specpdl-size (* 1024 1024))
+ (completion-ignore-case nil)
+ (completion-regexp-list nil)
+ (words (eq paren 'words))
+ (open (cond ((stringp paren) paren) (paren "\\(")))
+ (sorted-strings (delete-dups
+ (sort (copy-sequence strings) 'string-lessp)))
+ (re (regexp-opt-group sorted-strings open)))
+ (if words (concat "\\<" re "\\>") re))))
;;;###autoload
(defun regexp-opt-depth (regexp)
"Return the depth of REGEXP.
-This means the number of regexp grouping constructs (parenthesised expressions)
-in REGEXP."
+This means the number of non-shy regexp grouping constructs
+\(parenthesized expressions) in REGEXP."
(save-match-data
;; Hack to signal an error if REGEXP does not have balanced parentheses.
(string-match regexp "")
;; Count the number of open parentheses in REGEXP.
- (let ((count 0) start)
- (while (string-match "\\\\(" regexp start)
- (setq count (1+ count) start (match-end 0)))
+ (let ((count 0) start last)
+ (while (string-match "\\\\(\\(\\?:\\)?" regexp start)
+ (setq start (match-end 0)) ; Start of next search.
+ (when (and (not (match-beginning 1))
+ (subregexp-context-p regexp (match-beginning 0) last))
+ ;; It's not a shy group and it's not inside brackets or after
+ ;; a backslash: it's really a group-open marker.
+ (setq last start) ; Speed up next regexp-opt-re-context-p.
+ (setq count (1+ count))))
count)))
\f
;;; Workhorse functions.
(eval-when-compile
(require 'cl))
-(unless (fboundp 'make-bool-vector)
- (defalias 'make-bool-vector 'make-vector))
-
(defun regexp-opt-group (strings &optional paren lax)
- ;;
- ;; Return a regexp to match a string in STRINGS.
+ ;; Return a regexp to match a string in the sorted list STRINGS.
;; If PAREN non-nil, output regexp parentheses around returned regexp.
;; If LAX non-nil, don't output parentheses if it doesn't require them.
;; Merges keywords to avoid backtracking in Emacs' regexp matcher.
- ;;
+
;; The basic idea is to find the shortest common prefix or suffix, remove it
;; and recurse. If there is no prefix, we divide the list into two so that
- ;; (at least) one half will have at least a one-character common prefix.
- ;;
+ ;; \(at least) one half will have at least a one-character common prefix.
+
;; Also we delay the addition of grouping parenthesis as long as possible
;; until we're sure we need them, and try to remove one-character sequences
;; so we can use character sets rather than grouping parenthesis.
- ;;
- (let* ((open-group (if paren "\\(" ""))
+ (let* ((open-group (cond ((stringp paren) paren) (paren "\\(?:") (t "")))
(close-group (if paren "\\)" ""))
(open-charset (if lax "" open-group))
- (close-charset (if lax "" close-group))
- (open-presuf open-charset)
- (close-presuf close-charset))
+ (close-charset (if lax "" close-group)))
(cond
;;
;; If there are no strings, just return the empty string.
(regexp-opt-group (cdr strings) t t) "?"
close-charset))
;;
- ;; If all are one-character strings, just return a character set.
- ((= (apply 'max (mapcar 'length strings)) 1)
- (concat open-charset
- (regexp-opt-charset strings)
- close-charset))
+ ;; If there are several one-char strings, use charsets
+ ((and (= (length (car strings)) 1)
+ (let ((strs (cdr strings)))
+ (while (and strs (/= (length (car strs)) 1))
+ (pop strs))
+ strs))
+ (let (letters rest)
+ ;; Collect one-char strings
+ (dolist (s strings)
+ (if (= (length s) 1) (push (string-to-char s) letters) (push s rest)))
+
+ (if rest
+ ;; several one-char strings: take them and recurse
+ ;; on the rest (first so as to match the longest).
+ (concat open-group
+ (regexp-opt-group (nreverse rest))
+ "\\|" (regexp-opt-charset letters)
+ close-group)
+ ;; all are one-char strings: just return a character set.
+ (concat open-charset
+ (regexp-opt-charset letters)
+ close-charset))))
;;
;; We have a list of different length strings.
(t
- (let ((prefix (try-completion "" (mapcar 'list strings)))
- (suffix (regexp-opt-try-suffix strings))
- (letters (let ((completion-regexp-list '("^.$")))
- (all-completions "" (mapcar 'list strings)))))
- (cond
- ;;
- ;; If there is a common prefix, remove it and recurse on the suffixes.
- ((> (length prefix) 0)
- (let* ((end (length prefix))
- (suffixes (mapcar (lambda (s) (substring s end)) strings)))
- (concat open-presuf
- (regexp-quote prefix) (regexp-opt-group suffixes t t)
- close-presuf)))
- ;;
- ;; If there is a common suffix, remove it and recurse on the prefixes.
- ((> (length suffix) (if lax
- 0
- (- (apply 'max (mapcar 'length strings)) 2)))
- (let* ((end (- (length suffix)))
- (prefixes (sort (mapcar (lambda (s) (substring s 0 end))
- strings)
- 'string-lessp)))
- (concat open-presuf
- (regexp-opt-group prefixes t t) (regexp-quote suffix)
- close-presuf)))
- ;;
- ;; If there are several one-character strings, remove them and recurse
- ;; on the rest (first so the final regexp finds the longest match).
- ((> (length letters) 1)
- (let ((rest (let ((completion-regexp-list '("^..+$")))
- (all-completions "" (mapcar 'list strings)))))
- (concat open-group
- (regexp-opt-group rest) "\\|" (regexp-opt-charset letters)
- close-group)))
- ;;
- ;; Otherwise, divide the list into those that start with a particular
- ;; letter and those that do not, and recurse on them.
- (t
- (let* ((char (substring (car strings) 0 1))
- (half1 (all-completions char (mapcar 'list strings)))
- (half2 (nthcdr (length half1) strings)))
- (concat open-group
- (regexp-opt-group half1) "\\|" (regexp-opt-group half2)
- close-group)))))))))
+ (let ((prefix (try-completion "" strings)))
+ (if (> (length prefix) 0)
+ ;; common prefix: take it and recurse on the suffixes.
+ (let* ((n (length prefix))
+ (suffixes (mapcar (lambda (s) (substring s n)) strings)))
+ (concat open-group
+ (regexp-quote prefix)
+ (regexp-opt-group suffixes t t)
+ close-group))
+
+ (let* ((sgnirts (mapcar (lambda (s)
+ (concat (nreverse (string-to-list s))))
+ strings))
+ (xiffus (try-completion "" sgnirts)))
+ (if (> (length xiffus) 0)
+ ;; common suffix: take it and recurse on the prefixes.
+ (let* ((n (- (length xiffus)))
+ (prefixes
+ ;; Sorting is necessary in cases such as ("ad" "d").
+ (sort (mapcar (lambda (s) (substring s 0 n)) strings)
+ 'string-lessp)))
+ (concat open-group
+ (regexp-opt-group prefixes t t)
+ (regexp-quote
+ (concat (nreverse (string-to-list xiffus))))
+ close-group))
+
+ ;; Otherwise, divide the list into those that start with a
+ ;; particular letter and those that do not, and recurse on them.
+ (let* ((char (char-to-string (string-to-char (car strings))))
+ (half1 (all-completions char strings))
+ (half2 (nthcdr (length half1) strings)))
+ (concat open-group
+ (regexp-opt-group half1)
+ "\\|" (regexp-opt-group half2)
+ close-group))))))))))
+
(defun regexp-opt-charset (chars)
;;
;; The basic idea is to find character ranges. Also we take care in the
;; position of character set meta characters in the character set regexp.
;;
- (let* ((charwidth 256) ; Yeah, right.
- (charmap (make-bool-vector charwidth nil))
+ (let* ((charmap (make-char-table 'case-table))
+ (start -1) (end -2)
(charset "")
(bracket "") (dash "") (caret ""))
;;
;; Make a character map but extract character set meta characters.
- (dolist (char (mapcar 'string-to-char chars))
+ (dolist (char chars)
(case char
(?\]
(setq bracket "]"))
(aset charmap char t))))
;;
;; Make a character set from the map using ranges where applicable.
- (dotimes (char charwidth)
- (let ((start char))
- (while (and (< char charwidth) (aref charmap char))
- (incf char))
- (cond ((> char (+ start 3))
- (setq charset (format "%s%c-%c" charset start (1- char))))
- ((> char start)
- (setq charset (format "%s%c" charset (setq char start)))))))
+ (map-char-table
+ (lambda (c v)
+ (when v
+ (if (= (1- c) end) (setq end c)
+ (if (> end (+ start 2))
+ (setq charset (format "%s%c-%c" charset start end))
+ (while (>= end start)
+ (setq charset (format "%s%c" charset start))
+ (incf start)))
+ (setq start c end c))))
+ charmap)
+ (when (>= end start)
+ (if (> end (+ start 2))
+ (setq charset (format "%s%c-%c" charset start end))
+ (while (>= end start)
+ (setq charset (format "%s%c" charset start))
+ (incf start))))
;;
;; Make sure a caret is not first and a dash is first or last.
(if (and (string-equal charset "") (string-equal bracket ""))
(concat "[" dash caret "]")
(concat "[" bracket charset caret dash "]"))))
-(defun regexp-opt-try-suffix (strings)
- ;;
- ;; Return common suffix of each string in STRINGS. See `try-completion'.
- ;;
- (let* ((chars (mapcar (lambda (s) (mapcar 'identity s)) strings))
- (srahc (mapcar 'reverse chars))
- (sgnirts (mapcar (lambda (c) (mapconcat 'char-to-string c "")) srahc))
- (xiffus (try-completion "" (mapcar 'list sgnirts))))
- (mapconcat 'char-to-string (reverse (mapcar 'identity xiffus)) "")))
-
(provide 'regexp-opt)
+;; arch-tag: 6c5a66f4-29af-4fd6-8c3b-4b554d5b4370
;;; regexp-opt.el ends here