Commit | Line | Data |
---|---|---|
e65f17fe DN |
1 | ;;; sregex.el --- symbolic regular expressions |
2 | ||
d59c3137 | 3 | ;; Copyright (C) 1997, 1998, 2000, 2001, 2002, 2003, 2004, |
f0fa15c5 | 4 | ;; 2005, 2006, 2007 Free Software Foundation, Inc. |
e65f17fe DN |
5 | |
6 | ;; Author: Bob Glickstein <bobg+sregex@zanshin.com> | |
7 | ;; Maintainer: Bob Glickstein <bobg+sregex@zanshin.com> | |
c2413c67 | 8 | ;; Keywords: extensions |
e65f17fe DN |
9 | |
10 | ;; This file is part of GNU Emacs. | |
11 | ||
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 2, or (at your option) | |
15 | ;; any later version. | |
16 | ||
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. | |
21 | ||
22 | ;; You should have received a copy of the GNU General Public License | |
23 | ;; along with GNU Emacs; see the file COPYING. If not, write to the | |
3a35cf56 LK |
24 | ;; Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, |
25 | ;; Boston, MA 02110-1301, USA. | |
e65f17fe DN |
26 | |
27 | ;;; Commentary: | |
28 | ||
29 | ;; This package allows you to write regular expressions using a | |
30 | ;; totally new, Lisp-like syntax. | |
31 | ||
32 | ;; A "symbolic regular expression" (sregex for short) is a Lisp form | |
33 | ;; that, when evaluated, produces the string form of the specified | |
34 | ;; regular expression. Here's a simple example: | |
35 | ||
36 | ;; (sregexq (or "Bob" "Robert")) => "Bob\\|Robert" | |
37 | ||
38 | ;; As you can see, an sregex is specified by placing one or more | |
39 | ;; special clauses in a call to `sregexq'. The clause in this case is | |
40 | ;; the `or' of two strings (not to be confused with the Lisp function | |
41 | ;; `or'). The list of allowable clauses appears below. | |
42 | ||
43 | ;; With sregex, it is never necessary to "escape" magic characters | |
44 | ;; that are meant to be taken literally; that happens automatically. | |
45 | ;; For example: | |
46 | ||
47 | ;; (sregexq "M*A*S*H") => "M\\*A\\*S\\*H" | |
48 | ||
49 | ;; It is also unnecessary to "group" parts of the expression together | |
50 | ;; to overcome operator precedence; that also happens automatically. | |
51 | ;; For example: | |
52 | ||
40aeecad | 53 | ;; (sregexq (opt (or "Bob" "Robert"))) => "\\(?:Bob\\|Robert\\)?" |
e65f17fe DN |
54 | |
55 | ;; It *is* possible to group parts of the expression in order to refer | |
56 | ;; to them with numbered backreferences: | |
57 | ||
58 | ;; (sregexq (group (or "Go" "Run")) | |
59 | ;; ", Spot, " | |
60 | ;; (backref 1)) => "\\(Go\\|Run\\), Spot, \\1" | |
61 | ||
e65f17fe DN |
62 | ;; `sregexq' is a macro. Each time it is used, it constructs a simple |
63 | ;; Lisp expression that then invokes a moderately complex engine to | |
64 | ;; interpret the sregex and render the string form. Because of this, | |
65 | ;; I don't recommend sprinkling calls to `sregexq' throughout your | |
66 | ;; code, the way one normally does with string regexes (which are | |
67 | ;; cheap to evaluate). Instead, it's wiser to precompute the regexes | |
68 | ;; you need wherever possible instead of repeatedly constructing the | |
69 | ;; same ones over and over. Example: | |
70 | ||
71 | ;; (let ((field-regex (sregexq (opt "resent-") | |
72 | ;; (or "to" "cc" "bcc")))) | |
73 | ;; ... | |
74 | ;; (while ... | |
75 | ;; ... | |
76 | ;; (re-search-forward field-regex ...) | |
77 | ;; ...)) | |
78 | ||
79 | ;; The arguments to `sregexq' are automatically quoted, but the | |
80 | ;; flipside of this is that it is not straightforward to include | |
81 | ;; computed (i.e., non-constant) values in `sregexq' expressions. So | |
82 | ;; `sregex' is a function that is like `sregexq' but which does not | |
83 | ;; automatically quote its values. Literal sregex clauses must be | |
84 | ;; explicitly quoted like so: | |
85 | ||
86 | ;; (sregex '(or "Bob" "Robert")) => "Bob\\|Robert" | |
87 | ||
88 | ;; but computed clauses can be included easily, allowing for the reuse | |
89 | ;; of common clauses: | |
90 | ||
91 | ;; (let ((dotstar '(0+ any)) | |
92 | ;; (whitespace '(1+ (syntax ?-))) | |
93 | ;; (digits '(1+ (char (?0 . ?9))))) | |
94 | ;; (sregex 'bol dotstar ":" whitespace digits)) => "^.*:\\s-+[0-9]+" | |
95 | ||
e65f17fe DN |
96 | ;; To use this package in a Lisp program, simply (require 'sregex). |
97 | ||
98 | ;; Here are the clauses allowed in an `sregex' or `sregexq' | |
99 | ;; expression: | |
100 | ||
101 | ;; - a string | |
102 | ;; This stands for the literal string. If it contains | |
103 | ;; metacharacters, they will be escaped in the resulting regex | |
104 | ;; (using `regexp-quote'). | |
105 | ||
106 | ;; - the symbol `any' | |
107 | ;; This stands for ".", a regex matching any character except | |
108 | ;; newline. | |
109 | ||
110 | ;; - the symbol `bol' | |
111 | ;; Stands for "^", matching the empty string at the beginning of a line | |
112 | ||
113 | ;; - the symbol `eol' | |
114 | ;; Stands for "$", matching the empty string at the end of a line | |
115 | ||
116 | ;; - (group CLAUSE ...) | |
117 | ;; Groups the given CLAUSEs using "\\(" and "\\)". | |
118 | ||
119 | ;; - (sequence CLAUSE ...) | |
120 | ||
40aeecad SM |
121 | ;; Groups the given CLAUSEs; may or may not use "\\(?:" and "\\)". |
122 | ;; Clauses grouped by `sequence' do not count for purposes of | |
e65f17fe DN |
123 | ;; numbering backreferences. Use `sequence' in situations like |
124 | ;; this: | |
125 | ||
126 | ;; (sregexq (or "dog" "cat" | |
127 | ;; (sequence (opt "sea ") "monkey"))) | |
40aeecad | 128 | ;; => "dog\\|cat\\|\\(?:sea \\)?monkey" |
e65f17fe DN |
129 | |
130 | ;; where a single `or' alternate needs to contain multiple | |
131 | ;; subclauses. | |
132 | ||
133 | ;; - (backref N) | |
134 | ;; Matches the same string previously matched by the Nth "group" in | |
40aeecad | 135 | ;; the same sregex. N is a positive integer. |
e65f17fe DN |
136 | |
137 | ;; - (or CLAUSE ...) | |
138 | ;; Matches any one of the CLAUSEs by separating them with "\\|". | |
139 | ||
140 | ;; - (0+ CLAUSE ...) | |
141 | ;; Concatenates the given CLAUSEs and matches zero or more | |
142 | ;; occurrences by appending "*". | |
143 | ||
144 | ;; - (1+ CLAUSE ...) | |
145 | ;; Concatenates the given CLAUSEs and matches one or more | |
146 | ;; occurrences by appending "+". | |
147 | ||
148 | ;; - (opt CLAUSE ...) | |
149 | ;; Concatenates the given CLAUSEs and matches zero or one occurrence | |
150 | ;; by appending "?". | |
151 | ||
152 | ;; - (repeat MIN MAX CLAUSE ...) | |
153 | ;; Concatenates the given CLAUSEs and constructs a regex matching at | |
154 | ;; least MIN occurrences and at most MAX occurrences. MIN must be a | |
155 | ;; non-negative integer. MAX must be a non-negative integer greater | |
156 | ;; than or equal to MIN; or MAX can be nil to mean "infinity." | |
157 | ||
158 | ;; - (char CHAR-CLAUSE ...) | |
159 | ;; Creates a "character class" matching one character from the given | |
160 | ;; set. See below for how to construct a CHAR-CLAUSE. | |
161 | ||
162 | ;; - (not-char CHAR-CLAUSE ...) | |
163 | ;; Creates a "character class" matching any one character not in the | |
164 | ;; given set. See below for how to construct a CHAR-CLAUSE. | |
165 | ||
166 | ;; - the symbol `bot' | |
167 | ;; Stands for "\\`", matching the empty string at the beginning of | |
168 | ;; text (beginning of a string or of a buffer). | |
169 | ||
170 | ;; - the symbol `eot' | |
171 | ;; Stands for "\\'", matching the empty string at the end of text. | |
172 | ||
173 | ;; - the symbol `point' | |
174 | ;; Stands for "\\=", matching the empty string at point. | |
175 | ||
176 | ;; - the symbol `word-boundary' | |
177 | ;; Stands for "\\b", matching the empty string at the beginning or | |
178 | ;; end of a word. | |
179 | ||
180 | ;; - the symbol `not-word-boundary' | |
181 | ;; Stands for "\\B", matching the empty string not at the beginning | |
182 | ;; or end of a word. | |
183 | ||
184 | ;; - the symbol `bow' | |
185 | ;; Stands for "\\<", matching the empty string at the beginning of a | |
186 | ;; word. | |
187 | ||
188 | ;; - the symbol `eow' | |
189 | ;; Stands for "\\>", matching the empty string at the end of a word. | |
190 | ||
191 | ;; - the symbol `wordchar' | |
192 | ;; Stands for the regex "\\w", matching a word-constituent character | |
193 | ;; (as determined by the current syntax table) | |
194 | ||
195 | ;; - the symbol `not-wordchar' | |
196 | ;; Stands for the regex "\\W", matching a non-word-constituent | |
197 | ;; character. | |
198 | ||
199 | ;; - (syntax CODE) | |
200 | ;; Stands for the regex "\\sCODE", where CODE is a syntax table code | |
201 | ;; (a single character). Matches any character with the requested | |
202 | ;; syntax. | |
203 | ||
204 | ;; - (not-syntax CODE) | |
205 | ;; Stands for the regex "\\SCODE", where CODE is a syntax table code | |
206 | ;; (a single character). Matches any character without the | |
207 | ;; requested syntax. | |
208 | ||
209 | ;; - (regex REGEX) | |
210 | ;; This is a "trapdoor" for including ordinary regular expression | |
211 | ;; strings in the result. Some regular expressions are clearer when | |
212 | ;; written the old way: "[a-z]" vs. (sregexq (char (?a . ?z))), for | |
213 | ;; instance. However, see the note under "Bugs," below. | |
214 | ||
215 | ;; Each CHAR-CLAUSE that is passed to (char ...) and (not-char ...) | |
216 | ;; has one of the following forms: | |
217 | ||
218 | ;; - a character | |
219 | ;; Adds that character to the set. | |
220 | ||
221 | ;; - a string | |
222 | ;; Adds all the characters in the string to the set. | |
223 | ||
224 | ;; - A pair (MIN . MAX) | |
225 | ;; Where MIN and MAX are characters, adds the range of characters | |
226 | ;; from MIN through MAX to the set. | |
227 | ||
228 | ;;; To do: | |
229 | ||
e65f17fe DN |
230 | ;; An earlier version of this package could optionally translate the |
231 | ;; symbolic regex into other languages' syntaxes, e.g. Perl. For | |
232 | ;; instance, with Perl syntax selected, (sregexq (or "ab" "cd")) would | |
233 | ;; yield "ab|cd" instead of "ab\\|cd". It might be useful to restore | |
234 | ;; such a facility. | |
235 | ||
40aeecad SM |
236 | ;; - handle multibyte chars in sregex--char-aux |
237 | ;; - add support for character classes ([:blank:], ...) | |
238 | ;; - add support for non-greedy operators *? and +? | |
239 | ;; - bug: (sregexq (opt (opt ?a))) returns "a??" which is a non-greedy "a?" | |
e65f17fe | 240 | |
40aeecad | 241 | ;;; Bugs: |
e65f17fe DN |
242 | |
243 | ;;; Code: | |
244 | ||
40aeecad | 245 | (eval-when-compile (require 'cl)) |
e65f17fe | 246 | |
40aeecad SM |
247 | ;; Compatibility code for when we didn't have shy-groups |
248 | (defvar sregex--current-sregex nil) | |
249 | (defun sregex-info () nil) | |
250 | (defmacro sregex-save-match-data (&rest forms) (cons 'save-match-data forms)) | |
251 | (defun sregex-replace-match (r &optional f l str subexp x) | |
252 | (replace-match r f l str subexp)) | |
253 | (defun sregex-match-string (c &optional i x) (match-string c i)) | |
e65f17fe | 254 | (defun sregex-match-string-no-properties (count &optional in-string sregex) |
40aeecad SM |
255 | (match-string-no-properties count in-string)) |
256 | (defun sregex-match-beginning (count &optional sregex) (match-beginning count)) | |
257 | (defun sregex-match-end (count &optional sregex) (match-end count)) | |
258 | (defun sregex-match-data (&optional sregex) (match-data)) | |
259 | (defun sregex-backref-num (n &optional sregex) n) | |
260 | ||
e65f17fe DN |
261 | |
262 | (defun sregex (&rest exps) | |
263 | "Symbolic regular expression interpreter. | |
264 | This is exactly like `sregexq' (q.v.) except that it evaluates all its | |
265 | arguments, so literal sregex clauses must be quoted. For example: | |
266 | ||
267 | (sregex '(or \"Bob\" \"Robert\")) => \"Bob\\\\|Robert\" | |
268 | ||
269 | An argument-evaluating sregex interpreter lets you reuse sregex | |
270 | subexpressions: | |
271 | ||
272 | (let ((dotstar '(0+ any)) | |
273 | (whitespace '(1+ (syntax ?-))) | |
274 | (digits '(1+ (char (?0 . ?9))))) | |
275 | (sregex 'bol dotstar \":\" whitespace digits)) => \"^.*:\\\\s-+[0-9]+\"" | |
40aeecad | 276 | (sregex--sequence exps nil)) |
e65f17fe DN |
277 | |
278 | (defmacro sregexq (&rest exps) | |
279 | "Symbolic regular expression interpreter. | |
280 | This macro allows you to specify a regular expression (regexp) in | |
281 | symbolic form, and converts it into the string form required by Emacs's | |
282 | regex functions such as `re-search-forward' and `looking-at'. Here is | |
283 | a simple example: | |
284 | ||
285 | (sregexq (or \"Bob\" \"Robert\")) => \"Bob\\\\|Robert\" | |
286 | ||
287 | As you can see, an sregex is specified by placing one or more special | |
288 | clauses in a call to `sregexq'. The clause in this case is the `or' | |
289 | of two strings (not to be confused with the Lisp function `or'). The | |
290 | list of allowable clauses appears below. | |
291 | ||
292 | With `sregex', it is never necessary to \"escape\" magic characters | |
293 | that are meant to be taken literally; that happens automatically. | |
294 | For example: | |
295 | ||
296 | (sregexq \"M*A*S*H\") => \"M\\\\*A\\\\*S\\\\*H\" | |
297 | ||
298 | It is also unnecessary to \"group\" parts of the expression together | |
299 | to overcome operator precedence; that also happens automatically. | |
300 | For example: | |
301 | ||
302 | (sregexq (opt (or \"Bob\" \"Robert\"))) => \"\\\\(Bob\\\\|Robert\\\\)?\" | |
303 | ||
304 | It *is* possible to group parts of the expression in order to refer | |
305 | to them with numbered backreferences: | |
306 | ||
307 | (sregexq (group (or \"Go\" \"Run\")) | |
308 | \", Spot, \" | |
309 | (backref 1)) => \"\\\\(Go\\\\|Run\\\\), Spot, \\\\1\" | |
310 | ||
311 | If `sregexq' needs to introduce its own grouping parentheses, it will | |
312 | automatically renumber your backreferences: | |
313 | ||
314 | (sregexq (opt \"resent-\") | |
315 | (group (or \"to\" \"cc\" \"bcc\")) | |
316 | \": \" | |
317 | (backref 1)) => \"\\\\(resent-\\\\)?\\\\(to\\\\|cc\\\\|bcc\\\\): \\\\2\" | |
318 | ||
319 | `sregexq' is a macro. Each time it is used, it constructs a simple | |
320 | Lisp expression that then invokes a moderately complex engine to | |
321 | interpret the sregex and render the string form. Because of this, I | |
322 | don't recommend sprinkling calls to `sregexq' throughout your code, | |
323 | the way one normally does with string regexes (which are cheap to | |
324 | evaluate). Instead, it's wiser to precompute the regexes you need | |
325 | wherever possible instead of repeatedly constructing the same ones | |
326 | over and over. Example: | |
327 | ||
328 | (let ((field-regex (sregexq (opt \"resent-\") | |
329 | (or \"to\" \"cc\" \"bcc\")))) | |
330 | ... | |
331 | (while ... | |
332 | ... | |
333 | (re-search-forward field-regex ...) | |
334 | ...)) | |
335 | ||
336 | The arguments to `sregexq' are automatically quoted, but the | |
337 | flipside of this is that it is not straightforward to include | |
338 | computed (i.e., non-constant) values in `sregexq' expressions. So | |
339 | `sregex' is a function that is like `sregexq' but which does not | |
340 | automatically quote its values. Literal sregex clauses must be | |
341 | explicitly quoted like so: | |
342 | ||
343 | (sregex '(or \"Bob\" \"Robert\")) => \"Bob\\\\|Robert\" | |
344 | ||
345 | but computed clauses can be included easily, allowing for the reuse | |
346 | of common clauses: | |
347 | ||
348 | (let ((dotstar '(0+ any)) | |
349 | (whitespace '(1+ (syntax ?-))) | |
350 | (digits '(1+ (char (?0 . ?9))))) | |
351 | (sregex 'bol dotstar \":\" whitespace digits)) => \"^.*:\\\\s-+[0-9]+\" | |
352 | ||
353 | Here are the clauses allowed in an `sregex' or `sregexq' expression: | |
354 | ||
355 | - a string | |
356 | This stands for the literal string. If it contains | |
357 | metacharacters, they will be escaped in the resulting regex | |
358 | (using `regexp-quote'). | |
359 | ||
360 | - the symbol `any' | |
361 | This stands for \".\", a regex matching any character except | |
362 | newline. | |
363 | ||
364 | - the symbol `bol' | |
365 | Stands for \"^\", matching the empty string at the beginning of a line | |
366 | ||
367 | - the symbol `eol' | |
368 | Stands for \"$\", matching the empty string at the end of a line | |
369 | ||
370 | - (group CLAUSE ...) | |
371 | Groups the given CLAUSEs using \"\\\\(\" and \"\\\\)\". | |
372 | ||
373 | - (sequence CLAUSE ...) | |
374 | ||
375 | Groups the given CLAUSEs; may or may not use \"\\\\(\" and \"\\\\)\". | |
40aeecad | 376 | Clauses grouped by `sequence' do not count for purposes of |
e65f17fe DN |
377 | numbering backreferences. Use `sequence' in situations like |
378 | this: | |
379 | ||
380 | (sregexq (or \"dog\" \"cat\" | |
381 | (sequence (opt \"sea \") \"monkey\"))) | |
40aeecad | 382 | => \"dog\\\\|cat\\\\|\\\\(?:sea \\\\)?monkey\" |
e65f17fe DN |
383 | |
384 | where a single `or' alternate needs to contain multiple | |
385 | subclauses. | |
386 | ||
387 | - (backref N) | |
388 | Matches the same string previously matched by the Nth \"group\" in | |
40aeecad | 389 | the same sregex. N is a positive integer. |
e65f17fe DN |
390 | |
391 | - (or CLAUSE ...) | |
392 | Matches any one of the CLAUSEs by separating them with \"\\\\|\". | |
393 | ||
394 | - (0+ CLAUSE ...) | |
395 | Concatenates the given CLAUSEs and matches zero or more | |
396 | occurrences by appending \"*\". | |
397 | ||
398 | - (1+ CLAUSE ...) | |
399 | Concatenates the given CLAUSEs and matches one or more | |
400 | occurrences by appending \"+\". | |
401 | ||
402 | - (opt CLAUSE ...) | |
403 | Concatenates the given CLAUSEs and matches zero or one occurrence | |
404 | by appending \"?\". | |
405 | ||
406 | - (repeat MIN MAX CLAUSE ...) | |
407 | Concatenates the given CLAUSEs and constructs a regex matching at | |
408 | least MIN occurrences and at most MAX occurrences. MIN must be a | |
409 | non-negative integer. MAX must be a non-negative integer greater | |
410 | than or equal to MIN; or MAX can be nil to mean \"infinity.\" | |
411 | ||
412 | - (char CHAR-CLAUSE ...) | |
413 | Creates a \"character class\" matching one character from the given | |
414 | set. See below for how to construct a CHAR-CLAUSE. | |
415 | ||
416 | - (not-char CHAR-CLAUSE ...) | |
417 | Creates a \"character class\" matching any one character not in the | |
418 | given set. See below for how to construct a CHAR-CLAUSE. | |
419 | ||
420 | - the symbol `bot' | |
421 | Stands for \"\\\\`\", matching the empty string at the beginning of | |
422 | text (beginning of a string or of a buffer). | |
423 | ||
424 | - the symbol `eot' | |
425 | Stands for \"\\\\'\", matching the empty string at the end of text. | |
426 | ||
427 | - the symbol `point' | |
faf9e188 | 428 | Stands for \"\\\\=\\=\", matching the empty string at point. |
e65f17fe DN |
429 | |
430 | - the symbol `word-boundary' | |
431 | Stands for \"\\\\b\", matching the empty string at the beginning or | |
432 | end of a word. | |
433 | ||
434 | - the symbol `not-word-boundary' | |
435 | Stands for \"\\\\B\", matching the empty string not at the beginning | |
436 | or end of a word. | |
437 | ||
438 | - the symbol `bow' | |
92fbf041 | 439 | Stands for \"\\\\=\\<\", matching the empty string at the beginning of a |
e65f17fe DN |
440 | word. |
441 | ||
442 | - the symbol `eow' | |
92fbf041 | 443 | Stands for \"\\\\=\\>\", matching the empty string at the end of a word. |
e65f17fe DN |
444 | |
445 | - the symbol `wordchar' | |
446 | Stands for the regex \"\\\\w\", matching a word-constituent character | |
447 | (as determined by the current syntax table) | |
448 | ||
449 | - the symbol `not-wordchar' | |
450 | Stands for the regex \"\\\\W\", matching a non-word-constituent | |
451 | character. | |
452 | ||
453 | - (syntax CODE) | |
454 | Stands for the regex \"\\\\sCODE\", where CODE is a syntax table code | |
455 | (a single character). Matches any character with the requested | |
456 | syntax. | |
457 | ||
458 | - (not-syntax CODE) | |
459 | Stands for the regex \"\\\\SCODE\", where CODE is a syntax table code | |
460 | (a single character). Matches any character without the | |
461 | requested syntax. | |
462 | ||
463 | - (regex REGEX) | |
464 | This is a \"trapdoor\" for including ordinary regular expression | |
465 | strings in the result. Some regular expressions are clearer when | |
466 | written the old way: \"[a-z]\" vs. (sregexq (char (?a . ?z))), for | |
40aeecad | 467 | instance. |
e65f17fe DN |
468 | |
469 | Each CHAR-CLAUSE that is passed to (char ...) and (not-char ...) | |
470 | has one of the following forms: | |
471 | ||
472 | - a character | |
473 | Adds that character to the set. | |
474 | ||
475 | - a string | |
476 | Adds all the characters in the string to the set. | |
477 | ||
478 | - A pair (MIN . MAX) | |
479 | Where MIN and MAX are characters, adds the range of characters | |
480 | from MIN through MAX to the set." | |
481 | `(apply 'sregex ',exps)) | |
482 | ||
483 | (defun sregex--engine (exp combine) | |
40aeecad SM |
484 | (cond |
485 | ((stringp exp) | |
486 | (if (and combine | |
487 | (eq combine 'suffix) | |
488 | (/= (length exp) 1)) | |
489 | (concat "\\(?:" (regexp-quote exp) "\\)") | |
490 | (regexp-quote exp))) | |
491 | ((symbolp exp) | |
492 | (ecase exp | |
493 | (any ".") | |
494 | (bol "^") | |
495 | (eol "$") | |
496 | (wordchar "\\w") | |
497 | (not-wordchar "\\W") | |
498 | (bot "\\`") | |
499 | (eot "\\'") | |
500 | (point "\\=") | |
501 | (word-boundary "\\b") | |
502 | (not-word-boundary "\\B") | |
503 | (bow "\\<") | |
504 | (eow "\\>"))) | |
505 | ((consp exp) | |
506 | (funcall (intern (concat "sregex--" | |
507 | (symbol-name (car exp)))) | |
508 | (cdr exp) | |
509 | combine)) | |
510 | (t (error "Invalid expression: %s" exp)))) | |
e65f17fe DN |
511 | |
512 | (defun sregex--sequence (exps combine) | |
40aeecad SM |
513 | (if (= (length exps) 1) (sregex--engine (car exps) combine) |
514 | (let ((re (mapconcat | |
515 | (lambda (e) (sregex--engine e 'concat)) | |
516 | exps ""))) | |
e65f17fe | 517 | (if (eq combine 'suffix) |
40aeecad SM |
518 | (concat "\\(?:" re "\\)") |
519 | re)))) | |
e65f17fe DN |
520 | |
521 | (defun sregex--or (exps combine) | |
40aeecad SM |
522 | (if (= (length exps) 1) (sregex--engine (car exps) combine) |
523 | (let ((re (mapconcat | |
524 | (lambda (e) (sregex--engine e 'or)) | |
525 | exps "\\|"))) | |
526 | (if (not (eq combine 'or)) | |
527 | (concat "\\(?:" re "\\)") | |
528 | re)))) | |
529 | ||
530 | (defun sregex--group (exps combine) (concat "\\(" (sregex--sequence exps nil) "\\)")) | |
531 | ||
532 | (defun sregex--backref (exps combine) (concat "\\" (int-to-string (car exps)))) | |
533 | (defun sregex--opt (exps combine) (concat (sregex--sequence exps 'suffix) "?")) | |
534 | (defun sregex--0+ (exps combine) (concat (sregex--sequence exps 'suffix) "*")) | |
535 | (defun sregex--1+ (exps combine) (concat (sregex--sequence exps 'suffix) "+")) | |
536 | ||
537 | (defun sregex--char (exps combine) (sregex--char-aux nil exps)) | |
538 | (defun sregex--not-char (exps combine) (sregex--char-aux t exps)) | |
539 | ||
540 | (defun sregex--syntax (exps combine) (format "\\s%c" (car exps))) | |
541 | (defun sregex--not-syntax (exps combine) (format "\\S%c" (car exps))) | |
542 | ||
543 | (defun sregex--regex (exps combine) | |
544 | (if combine (concat "\\(?:" (car exps) "\\)") (car exps))) | |
545 | ||
546 | (defun sregex--repeat (exps combine) | |
547 | (let* ((min (or (pop exps) 0)) | |
548 | (minstr (number-to-string min)) | |
549 | (max (pop exps))) | |
550 | (concat (sregex--sequence exps 'suffix) | |
551 | (concat "\\{" minstr "," | |
552 | (when max (number-to-string max)) "\\}")))) | |
553 | ||
554 | (defun sregex--char-range (start end) | |
555 | (let ((startc (char-to-string start)) | |
556 | (endc (char-to-string end))) | |
557 | (cond | |
558 | ((> end (+ start 2)) (concat startc "-" endc)) | |
559 | ((> end (+ start 1)) (concat startc (char-to-string (1+ start)) endc)) | |
560 | ((> end start) (concat startc endc)) | |
561 | (t startc)))) | |
e65f17fe DN |
562 | |
563 | (defun sregex--char-aux (complement args) | |
40aeecad SM |
564 | ;; regex-opt does the same, we should join effort. |
565 | (let ((chars (make-bool-vector 256 nil))) ; Yeah, right! | |
566 | (dolist (arg args) | |
567 | (cond ((integerp arg) (aset chars arg t)) | |
568 | ((stringp arg) (mapcar (lambda (c) (aset chars c t)) arg)) | |
569 | ((consp arg) | |
570 | (let ((start (car arg)) | |
571 | (end (cdr arg))) | |
572 | (when (> start end) | |
573 | (let ((tmp start)) (setq start end) (setq end tmp))) | |
574 | ;; now start <= end | |
575 | (let ((i start)) | |
576 | (while (<= i end) | |
577 | (aset chars i t) | |
578 | (setq i (1+ i)))))))) | |
e65f17fe | 579 | ;; now chars is a map of the characters in the class |
40aeecad SM |
580 | (let ((caret (aref chars ?^)) |
581 | (dash (aref chars ?-)) | |
582 | (class (if (aref chars ?\]) "]" ""))) | |
e65f17fe | 583 | (aset chars ?^ nil) |
40aeecad SM |
584 | (aset chars ?- nil) |
585 | (aset chars ?\] nil) | |
586 | ||
587 | (let (start end) | |
588 | (dotimes (i 256) | |
589 | (if (aref chars i) | |
590 | (progn | |
591 | (unless start (setq start i)) | |
592 | (setq end i) | |
593 | (aset chars i nil)) | |
594 | (when start | |
595 | (setq class (concat class (sregex--char-range start end))) | |
596 | (setq start nil)))) | |
597 | (if start | |
598 | (setq class (concat class (sregex--char-range start end))))) | |
599 | ||
600 | (if (> (length class) 0) | |
601 | (setq class (concat class (if caret "^") (if dash "-"))) | |
602 | (setq class (concat class (if dash "-") (if caret "^")))) | |
603 | (if (and (not complement) (= (length class) 1)) | |
604 | (regexp-quote class) | |
605 | (concat "[" (if complement "^") class "]"))))) | |
e65f17fe DN |
606 | |
607 | (provide 'sregex) | |
608 | ||
ab5796a9 | 609 | ;;; arch-tag: 460c1f5a-eb6e-42ec-a451-ffac78bdf492 |
e65f17fe | 610 | ;;; sregex.el ends here |