Try and fix the more obvious sources of bug#30635
[emacs.git] / lisp / obsolete / sregex.el
blob6ff5133ca027c7cdd6d61855baf95bbec4758269
1 ;;; sregex.el --- symbolic regular expressions
3 ;; Copyright (C) 1997-1998, 2000-2018 Free Software Foundation, Inc.
5 ;; Author: Bob Glickstein <bobg+sregex@zanshin.com>
6 ;; Maintainer: emacs-devel@gnu.org
7 ;; Keywords: extensions
8 ;; Obsolete-since: 24.1
10 ;; This file is part of GNU Emacs.
12 ;; GNU Emacs is free software: you can redistribute it and/or modify
13 ;; it under the terms of the GNU General Public License as published by
14 ;; the Free Software Foundation, either version 3 of the License, or
15 ;; (at your option) any later version.
17 ;; GNU Emacs is distributed in the hope that it will be useful,
18 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
19 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 ;; GNU General Public License for more details.
22 ;; You should have received a copy of the GNU General Public License
23 ;; along with GNU Emacs. If not, see <https://www.gnu.org/licenses/>.
25 ;;; Commentary:
27 ;; This package allows you to write regular expressions using a
28 ;; totally new, Lisp-like syntax.
30 ;; A "symbolic regular expression" (sregex for short) is a Lisp form
31 ;; that, when evaluated, produces the string form of the specified
32 ;; regular expression. Here's a simple example:
34 ;; (sregexq (or "Bob" "Robert")) => "Bob\\|Robert"
36 ;; As you can see, an sregex is specified by placing one or more
37 ;; special clauses in a call to `sregexq'. The clause in this case is
38 ;; the `or' of two strings (not to be confused with the Lisp function
39 ;; `or'). The list of allowable clauses appears below.
41 ;; With sregex, it is never necessary to "escape" magic characters
42 ;; that are meant to be taken literally; that happens automatically.
43 ;; For example:
45 ;; (sregexq "M*A*S*H") => "M\\*A\\*S\\*H"
47 ;; It is also unnecessary to "group" parts of the expression together
48 ;; to overcome operator precedence; that also happens automatically.
49 ;; For example:
51 ;; (sregexq (opt (or "Bob" "Robert"))) => "\\(?:Bob\\|Robert\\)?"
53 ;; It *is* possible to group parts of the expression in order to refer
54 ;; to them with numbered backreferences:
56 ;; (sregexq (group (or "Go" "Run"))
57 ;; ", Spot, "
58 ;; (backref 1)) => "\\(Go\\|Run\\), Spot, \\1"
60 ;; `sregexq' is a macro. Each time it is used, it constructs a simple
61 ;; Lisp expression that then invokes a moderately complex engine to
62 ;; interpret the sregex and render the string form. Because of this,
63 ;; I don't recommend sprinkling calls to `sregexq' throughout your
64 ;; code, the way one normally does with string regexes (which are
65 ;; cheap to evaluate). Instead, it's wiser to precompute the regexes
66 ;; you need wherever possible instead of repeatedly constructing the
67 ;; same ones over and over. Example:
69 ;; (let ((field-regex (sregexq (opt "resent-")
70 ;; (or "to" "cc" "bcc"))))
71 ;; ...
72 ;; (while ...
73 ;; ...
74 ;; (re-search-forward field-regex ...)
75 ;; ...))
77 ;; The arguments to `sregexq' are automatically quoted, but the
78 ;; flipside of this is that it is not straightforward to include
79 ;; computed (i.e., non-constant) values in `sregexq' expressions. So
80 ;; `sregex' is a function that is like `sregexq' but which does not
81 ;; automatically quote its values. Literal sregex clauses must be
82 ;; explicitly quoted like so:
84 ;; (sregex '(or "Bob" "Robert")) => "Bob\\|Robert"
86 ;; but computed clauses can be included easily, allowing for the reuse
87 ;; of common clauses:
89 ;; (let ((dotstar '(0+ any))
90 ;; (whitespace '(1+ (syntax ?-)))
91 ;; (digits '(1+ (char (?0 . ?9)))))
92 ;; (sregex 'bol dotstar ":" whitespace digits)) => "^.*:\\s-+[0-9]+"
94 ;; To use this package in a Lisp program, simply (require 'sregex).
96 ;; Here are the clauses allowed in an `sregex' or `sregexq'
97 ;; expression:
99 ;; - a string
100 ;; This stands for the literal string. If it contains
101 ;; metacharacters, they will be escaped in the resulting regex
102 ;; (using `regexp-quote').
104 ;; - the symbol `any'
105 ;; This stands for ".", a regex matching any character except
106 ;; newline.
108 ;; - the symbol `bol'
109 ;; Stands for "^", matching the empty string at the beginning of a line
111 ;; - the symbol `eol'
112 ;; Stands for "$", matching the empty string at the end of a line
114 ;; - (group CLAUSE ...)
115 ;; Groups the given CLAUSEs using "\\(" and "\\)".
117 ;; - (sequence CLAUSE ...)
119 ;; Groups the given CLAUSEs; may or may not use "\\(?:" and "\\)".
120 ;; Clauses grouped by `sequence' do not count for purposes of
121 ;; numbering backreferences. Use `sequence' in situations like
122 ;; this:
124 ;; (sregexq (or "dog" "cat"
125 ;; (sequence (opt "sea ") "monkey")))
126 ;; => "dog\\|cat\\|\\(?:sea \\)?monkey"
128 ;; where a single `or' alternate needs to contain multiple
129 ;; subclauses.
131 ;; - (backref N)
132 ;; Matches the same string previously matched by the Nth "group" in
133 ;; the same sregex. N is a positive integer.
135 ;; - (or CLAUSE ...)
136 ;; Matches any one of the CLAUSEs by separating them with "\\|".
138 ;; - (0+ CLAUSE ...)
139 ;; Concatenates the given CLAUSEs and matches zero or more
140 ;; occurrences by appending "*".
142 ;; - (1+ CLAUSE ...)
143 ;; Concatenates the given CLAUSEs and matches one or more
144 ;; occurrences by appending "+".
146 ;; - (opt CLAUSE ...)
147 ;; Concatenates the given CLAUSEs and matches zero or one occurrence
148 ;; by appending "?".
150 ;; - (repeat MIN MAX CLAUSE ...)
151 ;; Concatenates the given CLAUSEs and constructs a regex matching at
152 ;; least MIN occurrences and at most MAX occurrences. MIN must be a
153 ;; non-negative integer. MAX must be a non-negative integer greater
154 ;; than or equal to MIN; or MAX can be nil to mean "infinity."
156 ;; - (char CHAR-CLAUSE ...)
157 ;; Creates a "character class" matching one character from the given
158 ;; set. See below for how to construct a CHAR-CLAUSE.
160 ;; - (not-char CHAR-CLAUSE ...)
161 ;; Creates a "character class" matching any one character not in the
162 ;; given set. See below for how to construct a CHAR-CLAUSE.
164 ;; - the symbol `bot'
165 ;; Stands for "\\`", matching the empty string at the beginning of
166 ;; text (beginning of a string or of a buffer).
168 ;; - the symbol `eot'
169 ;; Stands for "\\'", matching the empty string at the end of text.
171 ;; - the symbol `point'
172 ;; Stands for "\\=", matching the empty string at point.
174 ;; - the symbol `word-boundary'
175 ;; Stands for "\\b", matching the empty string at the beginning or
176 ;; end of a word.
178 ;; - the symbol `not-word-boundary'
179 ;; Stands for "\\B", matching the empty string not at the beginning
180 ;; or end of a word.
182 ;; - the symbol `bow'
183 ;; Stands for "\\<", matching the empty string at the beginning of a
184 ;; word.
186 ;; - the symbol `eow'
187 ;; Stands for "\\>", matching the empty string at the end of a word.
189 ;; - the symbol `wordchar'
190 ;; Stands for the regex "\\w", matching a word-constituent character
191 ;; (as determined by the current syntax table)
193 ;; - the symbol `not-wordchar'
194 ;; Stands for the regex "\\W", matching a non-word-constituent
195 ;; character.
197 ;; - (syntax CODE)
198 ;; Stands for the regex "\\sCODE", where CODE is a syntax table code
199 ;; (a single character). Matches any character with the requested
200 ;; syntax.
202 ;; - (not-syntax CODE)
203 ;; Stands for the regex "\\SCODE", where CODE is a syntax table code
204 ;; (a single character). Matches any character without the
205 ;; requested syntax.
207 ;; - (regex REGEX)
208 ;; This is a "trapdoor" for including ordinary regular expression
209 ;; strings in the result. Some regular expressions are clearer when
210 ;; written the old way: "[a-z]" vs. (sregexq (char (?a . ?z))), for
211 ;; instance. However, see the note under "Bugs," below.
213 ;; Each CHAR-CLAUSE that is passed to (char ...) and (not-char ...)
214 ;; has one of the following forms:
216 ;; - a character
217 ;; Adds that character to the set.
219 ;; - a string
220 ;; Adds all the characters in the string to the set.
222 ;; - A pair (MIN . MAX)
223 ;; Where MIN and MAX are characters, adds the range of characters
224 ;; from MIN through MAX to the set.
226 ;;; To do:
228 ;; An earlier version of this package could optionally translate the
229 ;; symbolic regex into other languages' syntaxes, e.g. Perl. For
230 ;; instance, with Perl syntax selected, (sregexq (or "ab" "cd")) would
231 ;; yield "ab|cd" instead of "ab\\|cd". It might be useful to restore
232 ;; such a facility.
234 ;; - handle multibyte chars in sregex--char-aux
235 ;; - add support for character classes ([:blank:], ...)
236 ;; - add support for non-greedy operators *? and +?
237 ;; - bug: (sregexq (opt (opt ?a))) returns "a??" which is a non-greedy "a?"
239 ;;; Bugs:
241 ;;; Code:
243 (eval-when-compile (require 'cl))
245 ;; Compatibility code for when we didn't have shy-groups
246 (defvar sregex--current-sregex nil)
247 (defun sregex-info () nil)
248 (defmacro sregex-save-match-data (&rest forms) (cons 'save-match-data forms))
249 (defun sregex-replace-match (r &optional f l str subexp x)
250 (replace-match r f l str subexp))
251 (defun sregex-match-string (c &optional i x) (match-string c i))
252 (defun sregex-match-string-no-properties (count &optional in-string sregex)
253 (match-string-no-properties count in-string))
254 (defun sregex-match-beginning (count &optional sregex) (match-beginning count))
255 (defun sregex-match-end (count &optional sregex) (match-end count))
256 (defun sregex-match-data (&optional sregex) (match-data))
257 (defun sregex-backref-num (n &optional sregex) n)
260 (defun sregex (&rest exps)
261 "Symbolic regular expression interpreter.
262 This is exactly like `sregexq' (q.v.) except that it evaluates all its
263 arguments, so literal sregex clauses must be quoted. For example:
265 (sregex \\='(or \"Bob\" \"Robert\")) => \"Bob\\\\|Robert\"
267 An argument-evaluating sregex interpreter lets you reuse sregex
268 subexpressions:
270 (let ((dotstar \\='(0+ any))
271 (whitespace \\='(1+ (syntax ?-)))
272 (digits \\='(1+ (char (?0 . ?9)))))
273 (sregex \\='bol dotstar \":\" whitespace digits)) => \"^.*:\\\\s-+[0-9]+\""
274 (sregex--sequence exps nil))
276 (defmacro sregexq (&rest exps)
277 "Symbolic regular expression interpreter.
278 This macro allows you to specify a regular expression (regexp) in
279 symbolic form, and converts it into the string form required by Emacs's
280 regex functions such as `re-search-forward' and `looking-at'. Here is
281 a simple example:
283 (sregexq (or \"Bob\" \"Robert\")) => \"Bob\\\\|Robert\"
285 As you can see, an sregex is specified by placing one or more special
286 clauses in a call to `sregexq'. The clause in this case is the `or'
287 of two strings (not to be confused with the Lisp function `or'). The
288 list of allowable clauses appears below.
290 With `sregex', it is never necessary to \"escape\" magic characters
291 that are meant to be taken literally; that happens automatically.
292 For example:
294 (sregexq \"M*A*S*H\") => \"M\\\\*A\\\\*S\\\\*H\"
296 It is also unnecessary to \"group\" parts of the expression together
297 to overcome operator precedence; that also happens automatically.
298 For example:
300 (sregexq (opt (or \"Bob\" \"Robert\"))) => \"\\\\(Bob\\\\|Robert\\\\)?\"
302 It *is* possible to group parts of the expression in order to refer
303 to them with numbered backreferences:
305 (sregexq (group (or \"Go\" \"Run\"))
306 \", Spot, \"
307 (backref 1)) => \"\\\\(Go\\\\|Run\\\\), Spot, \\\\1\"
309 If `sregexq' needs to introduce its own grouping parentheses, it will
310 automatically renumber your backreferences:
312 (sregexq (opt \"resent-\")
313 (group (or \"to\" \"cc\" \"bcc\"))
314 \": \"
315 (backref 1)) => \"\\\\(resent-\\\\)?\\\\(to\\\\|cc\\\\|bcc\\\\): \\\\2\"
317 `sregexq' is a macro. Each time it is used, it constructs a simple
318 Lisp expression that then invokes a moderately complex engine to
319 interpret the sregex and render the string form. Because of this, I
320 don't recommend sprinkling calls to `sregexq' throughout your code,
321 the way one normally does with string regexes (which are cheap to
322 evaluate). Instead, it's wiser to precompute the regexes you need
323 wherever possible instead of repeatedly constructing the same ones
324 over and over. Example:
326 (let ((field-regex (sregexq (opt \"resent-\")
327 (or \"to\" \"cc\" \"bcc\"))))
329 (while ...
331 (re-search-forward field-regex ...)
332 ...))
334 The arguments to `sregexq' are automatically quoted, but the
335 flipside of this is that it is not straightforward to include
336 computed (i.e., non-constant) values in `sregexq' expressions. So
337 `sregex' is a function that is like `sregexq' but which does not
338 automatically quote its values. Literal sregex clauses must be
339 explicitly quoted like so:
341 (sregex \\='(or \"Bob\" \"Robert\")) => \"Bob\\\\|Robert\"
343 but computed clauses can be included easily, allowing for the reuse
344 of common clauses:
346 (let ((dotstar \\='(0+ any))
347 (whitespace \\='(1+ (syntax ?-)))
348 (digits \\='(1+ (char (?0 . ?9)))))
349 (sregex \\='bol dotstar \":\" whitespace digits)) => \"^.*:\\\\s-+[0-9]+\"
351 Here are the clauses allowed in an `sregex' or `sregexq' expression:
353 - a string
354 This stands for the literal string. If it contains
355 metacharacters, they will be escaped in the resulting regex
356 (using `regexp-quote').
358 - the symbol `any'
359 This stands for \".\", a regex matching any character except
360 newline.
362 - the symbol `bol'
363 Stands for \"^\", matching the empty string at the beginning of a line
365 - the symbol `eol'
366 Stands for \"$\", matching the empty string at the end of a line
368 - (group CLAUSE ...)
369 Groups the given CLAUSEs using \"\\\\(\" and \"\\\\)\".
371 - (sequence CLAUSE ...)
373 Groups the given CLAUSEs; may or may not use \"\\\\(\" and \"\\\\)\".
374 Clauses grouped by `sequence' do not count for purposes of
375 numbering backreferences. Use `sequence' in situations like
376 this:
378 (sregexq (or \"dog\" \"cat\"
379 (sequence (opt \"sea \") \"monkey\")))
380 => \"dog\\\\|cat\\\\|\\\\(?:sea \\\\)?monkey\"
382 where a single `or' alternate needs to contain multiple
383 subclauses.
385 - (backref N)
386 Matches the same string previously matched by the Nth \"group\" in
387 the same sregex. N is a positive integer.
389 - (or CLAUSE ...)
390 Matches any one of the CLAUSEs by separating them with \"\\\\|\".
392 - (0+ CLAUSE ...)
393 Concatenates the given CLAUSEs and matches zero or more
394 occurrences by appending \"*\".
396 - (1+ CLAUSE ...)
397 Concatenates the given CLAUSEs and matches one or more
398 occurrences by appending \"+\".
400 - (opt CLAUSE ...)
401 Concatenates the given CLAUSEs and matches zero or one occurrence
402 by appending \"?\".
404 - (repeat MIN MAX CLAUSE ...)
405 Concatenates the given CLAUSEs and constructs a regex matching at
406 least MIN occurrences and at most MAX occurrences. MIN must be a
407 non-negative integer. MAX must be a non-negative integer greater
408 than or equal to MIN; or MAX can be nil to mean \"infinity.\"
410 - (char CHAR-CLAUSE ...)
411 Creates a \"character class\" matching one character from the given
412 set. See below for how to construct a CHAR-CLAUSE.
414 - (not-char CHAR-CLAUSE ...)
415 Creates a \"character class\" matching any one character not in the
416 given set. See below for how to construct a CHAR-CLAUSE.
418 - the symbol `bot'
419 Stands for \"\\\\\\=`\", matching the empty string at the beginning of
420 text (beginning of a string or of a buffer).
422 - the symbol `eot'
423 Stands for \"\\\\'\", matching the empty string at the end of text.
425 - the symbol `point'
426 Stands for \"\\\\=\\=\", matching the empty string at point.
428 - the symbol `word-boundary'
429 Stands for \"\\\\b\", matching the empty string at the beginning or
430 end of a word.
432 - the symbol `not-word-boundary'
433 Stands for \"\\\\B\", matching the empty string not at the beginning
434 or end of a word.
436 - the symbol `bow'
437 Stands for \"\\\\=\\<\", matching the empty string at the beginning of a
438 word.
440 - the symbol `eow'
441 Stands for \"\\\\=\\>\", matching the empty string at the end of a word.
443 - the symbol `wordchar'
444 Stands for the regex \"\\\\w\", matching a word-constituent character
445 (as determined by the current syntax table)
447 - the symbol `not-wordchar'
448 Stands for the regex \"\\\\W\", matching a non-word-constituent
449 character.
451 - (syntax CODE)
452 Stands for the regex \"\\\\sCODE\", where CODE is a syntax table code
453 (a single character). Matches any character with the requested
454 syntax.
456 - (not-syntax CODE)
457 Stands for the regex \"\\\\SCODE\", where CODE is a syntax table code
458 (a single character). Matches any character without the
459 requested syntax.
461 - (regex REGEX)
462 This is a \"trapdoor\" for including ordinary regular expression
463 strings in the result. Some regular expressions are clearer when
464 written the old way: \"[a-z]\" vs. (sregexq (char (?a . ?z))), for
465 instance.
467 Each CHAR-CLAUSE that is passed to (char ...) and (not-char ...)
468 has one of the following forms:
470 - a character
471 Adds that character to the set.
473 - a string
474 Adds all the characters in the string to the set.
476 - A pair (MIN . MAX)
477 Where MIN and MAX are characters, adds the range of characters
478 from MIN through MAX to the set."
479 `(apply 'sregex ',exps))
481 (defun sregex--engine (exp combine)
482 (cond
483 ((stringp exp)
484 (if (and combine
485 (eq combine 'suffix)
486 (/= (length exp) 1))
487 (concat "\\(?:" (regexp-quote exp) "\\)")
488 (regexp-quote exp)))
489 ((symbolp exp)
490 (ecase exp
491 (any ".")
492 (bol "^")
493 (eol "$")
494 (wordchar "\\w")
495 (not-wordchar "\\W")
496 (bot "\\`")
497 (eot "\\'")
498 (point "\\=")
499 (word-boundary "\\b")
500 (not-word-boundary "\\B")
501 (bow "\\<")
502 (eow "\\>")))
503 ((consp exp)
504 (funcall (intern (concat "sregex--"
505 (symbol-name (car exp))))
506 (cdr exp)
507 combine))
508 (t (error "Invalid expression: %s" exp))))
510 (defun sregex--sequence (exps combine)
511 (if (= (length exps) 1) (sregex--engine (car exps) combine)
512 (let ((re (mapconcat
513 (lambda (e) (sregex--engine e 'concat))
514 exps "")))
515 (if (eq combine 'suffix)
516 (concat "\\(?:" re "\\)")
517 re))))
519 (defun sregex--or (exps combine)
520 (if (= (length exps) 1) (sregex--engine (car exps) combine)
521 (let ((re (mapconcat
522 (lambda (e) (sregex--engine e 'or))
523 exps "\\|")))
524 (if (not (eq combine 'or))
525 (concat "\\(?:" re "\\)")
526 re))))
528 (defun sregex--group (exps combine) (concat "\\(" (sregex--sequence exps nil) "\\)"))
530 (defun sregex--backref (exps combine) (concat "\\" (int-to-string (car exps))))
531 (defun sregex--opt (exps combine) (concat (sregex--sequence exps 'suffix) "?"))
532 (defun sregex--0+ (exps combine) (concat (sregex--sequence exps 'suffix) "*"))
533 (defun sregex--1+ (exps combine) (concat (sregex--sequence exps 'suffix) "+"))
535 (defun sregex--char (exps combine) (sregex--char-aux nil exps))
536 (defun sregex--not-char (exps combine) (sregex--char-aux t exps))
538 (defun sregex--syntax (exps combine) (format "\\s%c" (car exps)))
539 (defun sregex--not-syntax (exps combine) (format "\\S%c" (car exps)))
541 (defun sregex--regex (exps combine)
542 (if combine (concat "\\(?:" (car exps) "\\)") (car exps)))
544 (defun sregex--repeat (exps combine)
545 (let* ((min (or (pop exps) 0))
546 (minstr (number-to-string min))
547 (max (pop exps)))
548 (concat (sregex--sequence exps 'suffix)
549 (concat "\\{" minstr ","
550 (when max (number-to-string max)) "\\}"))))
552 (defun sregex--char-range (start end)
553 (let ((startc (char-to-string start))
554 (endc (char-to-string end)))
555 (cond
556 ((> end (+ start 2)) (concat startc "-" endc))
557 ((> end (+ start 1)) (concat startc (char-to-string (1+ start)) endc))
558 ((> end start) (concat startc endc))
559 (t startc))))
561 (defun sregex--char-aux (complement args)
562 ;; regex-opt does the same, we should join effort.
563 (let ((chars (make-bool-vector 256 nil))) ; Yeah, right!
564 (dolist (arg args)
565 (cond ((integerp arg) (aset chars arg t))
566 ((stringp arg) (mapc (lambda (c) (aset chars c t)) arg))
567 ((consp arg)
568 (let ((start (car arg))
569 (end (cdr arg)))
570 (when (> start end)
571 (let ((tmp start)) (setq start end) (setq end tmp)))
572 ;; now start <= end
573 (let ((i start))
574 (while (<= i end)
575 (aset chars i t)
576 (setq i (1+ i))))))))
577 ;; now chars is a map of the characters in the class
578 (let ((caret (aref chars ?^))
579 (dash (aref chars ?-))
580 (class (if (aref chars ?\]) "]" "")))
581 (aset chars ?^ nil)
582 (aset chars ?- nil)
583 (aset chars ?\] nil)
585 (let (start end)
586 (dotimes (i 256)
587 (if (aref chars i)
588 (progn
589 (unless start (setq start i))
590 (setq end i)
591 (aset chars i nil))
592 (when start
593 (setq class (concat class (sregex--char-range start end)))
594 (setq start nil))))
595 (if start
596 (setq class (concat class (sregex--char-range start end)))))
598 (if (> (length class) 0)
599 (setq class (concat class (if caret "^") (if dash "-")))
600 (setq class (concat class (if dash "-") (if caret "^"))))
601 (if (and (not complement) (= (length class) 1))
602 (regexp-quote class)
603 (concat "[" (if complement "^") class "]")))))
605 (provide 'sregex)
607 ;;; sregex.el ends here