Mercurial > vim
annotate src/regexp.c @ 4674:7099f98528b4
Added tag v7-3-1084 for changeset 05d57d7c2d55
author | Bram Moolenaar <bram@vim.org> |
---|---|
date | Fri, 31 May 2013 23:18:00 +0200 |
parents | 7a2be4a39423 |
children | 4d92b873acef |
rev | line source |
---|---|
7 | 1 /* vi:set ts=8 sts=4 sw=4: |
2 * | |
3 * Handling of regular expressions: vim_regcomp(), vim_regexec(), vim_regsub() | |
4 * | |
5 * NOTICE: | |
6 * | |
7 * This is NOT the original regular expression code as written by Henry | |
8 * Spencer. This code has been modified specifically for use with the VIM | |
9 * editor, and should not be used separately from Vim. If you want a good | |
10 * regular expression library, get the original code. The copyright notice | |
11 * that follows is from the original. | |
12 * | |
13 * END NOTICE | |
14 * | |
15 * Copyright (c) 1986 by University of Toronto. | |
16 * Written by Henry Spencer. Not derived from licensed software. | |
17 * | |
18 * Permission is granted to anyone to use this software for any | |
19 * purpose on any computer system, and to redistribute it freely, | |
20 * subject to the following restrictions: | |
21 * | |
22 * 1. The author is not responsible for the consequences of use of | |
23 * this software, no matter how awful, even if they arise | |
24 * from defects in it. | |
25 * | |
26 * 2. The origin of this software must not be misrepresented, either | |
27 * by explicit claim or by omission. | |
28 * | |
29 * 3. Altered versions must be plainly marked as such, and must not | |
30 * be misrepresented as being the original software. | |
31 * | |
32 * Beware that some of this code is subtly aware of the way operator | |
33 * precedence is structured in regular expressions. Serious changes in | |
34 * regular-expression syntax might require a total rethink. | |
35 * | |
24 | 36 * Changes have been made by Tony Andrews, Olaf 'Rhialto' Seibert, Robert |
37 * Webb, Ciaran McCreesh and Bram Moolenaar. | |
7 | 38 * Named character class support added by Walter Briscoe (1998 Jul 01) |
39 */ | |
40 | |
4444 | 41 /* Uncomment the first if you do not want to see debugging logs or files |
42 * related to regular expressions, even when compiling with -DDEBUG. | |
43 * Uncomment the second to get the regexp debugging. */ | |
44 /* #undef DEBUG */ | |
45 /* #define DEBUG */ | |
46 | |
7 | 47 #include "vim.h" |
48 | |
4444 | 49 #ifdef DEBUG |
50 /* show/save debugging data when BT engine is used */ | |
51 # define BT_REGEXP_DUMP | |
52 /* save the debugging data to a file instead of displaying it */ | |
53 # define BT_REGEXP_LOG | |
4460 | 54 # define BT_REGEXP_DEBUG_LOG |
55 # define BT_REGEXP_DEBUG_LOG_NAME "bt_regexp_debug.log" | |
4444 | 56 #endif |
7 | 57 |
58 /* | |
59 * The "internal use only" fields in regexp.h are present to pass info from | |
60 * compile to execute that permits the execute phase to run lots faster on | |
61 * simple cases. They are: | |
62 * | |
63 * regstart char that must begin a match; NUL if none obvious; Can be a | |
64 * multi-byte character. | |
65 * reganch is the match anchored (at beginning-of-line only)? | |
66 * regmust string (pointer into program) that match must include, or NULL | |
67 * regmlen length of regmust string | |
68 * regflags RF_ values or'ed together | |
69 * | |
70 * Regstart and reganch permit very fast decisions on suitable starting points | |
71 * for a match, cutting down the work a lot. Regmust permits fast rejection | |
72 * of lines that cannot possibly match. The regmust tests are costly enough | |
73 * that vim_regcomp() supplies a regmust only if the r.e. contains something | |
74 * potentially expensive (at present, the only such thing detected is * or + | |
75 * at the start of the r.e., which can involve a lot of backup). Regmlen is | |
76 * supplied because the test in vim_regexec() needs it and vim_regcomp() is | |
77 * computing it anyway. | |
78 */ | |
79 | |
80 /* | |
81 * Structure for regexp "program". This is essentially a linear encoding | |
82 * of a nondeterministic finite-state machine (aka syntax charts or | |
83 * "railroad normal form" in parsing technology). Each node is an opcode | |
84 * plus a "next" pointer, possibly plus an operand. "Next" pointers of | |
85 * all nodes except BRANCH and BRACES_COMPLEX implement concatenation; a "next" | |
86 * pointer with a BRANCH on both ends of it is connecting two alternatives. | |
87 * (Here we have one of the subtle syntax dependencies: an individual BRANCH | |
88 * (as opposed to a collection of them) is never concatenated with anything | |
89 * because of operator precedence). The "next" pointer of a BRACES_COMPLEX | |
167 | 90 * node points to the node after the stuff to be repeated. |
91 * The operand of some types of node is a literal string; for others, it is a | |
92 * node leading into a sub-FSM. In particular, the operand of a BRANCH node | |
93 * is the first node of the branch. | |
94 * (NB this is *not* a tree structure: the tail of the branch connects to the | |
95 * thing following the set of BRANCHes.) | |
7 | 96 * |
97 * pattern is coded like: | |
98 * | |
99 * +-----------------+ | |
100 * | V | |
101 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END | |
102 * | ^ | ^ | |
103 * +------+ +----------+ | |
104 * | |
105 * | |
106 * +------------------+ | |
107 * V | | |
108 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END | |
109 * | | ^ ^ | |
110 * | +---------------+ | | |
111 * +---------------------------------------------+ | |
112 * | |
113 * | |
167 | 114 * +----------------------+ |
115 * V | | |
233 | 116 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END |
856 | 117 * | | ^ ^ |
118 * | +-----------+ | | |
179 | 119 * +--------------------------------------------------+ |
167 | 120 * |
121 * | |
7 | 122 * +-------------------------+ |
123 * V | | |
124 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END | |
125 * | | ^ | |
126 * | +----------------+ | |
127 * +-----------------------------------------------+ | |
128 * | |
129 * | |
130 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END | |
131 * | | ^ ^ | |
132 * | +----------------+ | | |
133 * +--------------------------------+ | |
134 * | |
135 * +---------+ | |
136 * | V | |
137 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END | |
138 * | | | | ^ ^ | |
139 * | | | +-----+ | | |
140 * | | +----------------+ | | |
141 * | +---------------------------+ | | |
142 * +------------------------------------------------------+ | |
143 * | |
1209 | 144 * They all start with a BRANCH for "\|" alternatives, even when there is only |
7 | 145 * one alternative. |
146 */ | |
147 | |
148 /* | |
149 * The opcodes are: | |
150 */ | |
151 | |
152 /* definition number opnd? meaning */ | |
153 #define END 0 /* End of program or NOMATCH operand. */ | |
154 #define BOL 1 /* Match "" at beginning of line. */ | |
155 #define EOL 2 /* Match "" at end of line. */ | |
156 #define BRANCH 3 /* node Match this alternative, or the | |
157 * next... */ | |
158 #define BACK 4 /* Match "", "next" ptr points backward. */ | |
159 #define EXACTLY 5 /* str Match this string. */ | |
160 #define NOTHING 6 /* Match empty string. */ | |
161 #define STAR 7 /* node Match this (simple) thing 0 or more | |
162 * times. */ | |
163 #define PLUS 8 /* node Match this (simple) thing 1 or more | |
164 * times. */ | |
165 #define MATCH 9 /* node match the operand zero-width */ | |
166 #define NOMATCH 10 /* node check for no match with operand */ | |
167 #define BEHIND 11 /* node look behind for a match with operand */ | |
168 #define NOBEHIND 12 /* node look behind for no match with operand */ | |
169 #define SUBPAT 13 /* node match the operand here */ | |
170 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and | |
171 * n times (\{m,n\}). */ | |
172 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */ | |
173 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */ | |
174 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE | |
175 * and BRACE_COMPLEX. */ | |
176 #define NEWL 18 /* Match line-break */ | |
177 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */ | |
178 | |
179 | |
180 /* character classes: 20-48 normal, 50-78 include a line-break */ | |
181 #define ADD_NL 30 | |
182 #define FIRST_NL ANY + ADD_NL | |
183 #define ANY 20 /* Match any one character. */ | |
184 #define ANYOF 21 /* str Match any character in this string. */ | |
185 #define ANYBUT 22 /* str Match any character not in this | |
186 * string. */ | |
187 #define IDENT 23 /* Match identifier char */ | |
188 #define SIDENT 24 /* Match identifier char but no digit */ | |
189 #define KWORD 25 /* Match keyword char */ | |
190 #define SKWORD 26 /* Match word char but no digit */ | |
191 #define FNAME 27 /* Match file name char */ | |
192 #define SFNAME 28 /* Match file name char but no digit */ | |
193 #define PRINT 29 /* Match printable char */ | |
194 #define SPRINT 30 /* Match printable char but no digit */ | |
195 #define WHITE 31 /* Match whitespace char */ | |
196 #define NWHITE 32 /* Match non-whitespace char */ | |
197 #define DIGIT 33 /* Match digit char */ | |
198 #define NDIGIT 34 /* Match non-digit char */ | |
199 #define HEX 35 /* Match hex char */ | |
200 #define NHEX 36 /* Match non-hex char */ | |
201 #define OCTAL 37 /* Match octal char */ | |
202 #define NOCTAL 38 /* Match non-octal char */ | |
203 #define WORD 39 /* Match word char */ | |
204 #define NWORD 40 /* Match non-word char */ | |
205 #define HEAD 41 /* Match head char */ | |
206 #define NHEAD 42 /* Match non-head char */ | |
207 #define ALPHA 43 /* Match alpha char */ | |
208 #define NALPHA 44 /* Match non-alpha char */ | |
209 #define LOWER 45 /* Match lowercase char */ | |
210 #define NLOWER 46 /* Match non-lowercase char */ | |
211 #define UPPER 47 /* Match uppercase char */ | |
212 #define NUPPER 48 /* Match non-uppercase char */ | |
213 #define LAST_NL NUPPER + ADD_NL | |
214 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL) | |
215 | |
216 #define MOPEN 80 /* -89 Mark this point in input as start of | |
217 * \( subexpr. MOPEN + 0 marks start of | |
218 * match. */ | |
219 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks | |
220 * end of match. */ | |
221 #define BACKREF 100 /* -109 node Match same string again \1-\9 */ | |
222 | |
223 #ifdef FEAT_SYN_HL | |
224 # define ZOPEN 110 /* -119 Mark this point in input as start of | |
225 * \z( subexpr. */ | |
226 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */ | |
227 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */ | |
228 #endif | |
229 | |
230 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */ | |
231 | |
232 #define NOPEN 150 /* Mark this point in input as start of | |
233 \%( subexpr. */ | |
234 #define NCLOSE 151 /* Analogous to NOPEN. */ | |
235 | |
236 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */ | |
237 #define RE_BOF 201 /* Match "" at beginning of file. */ | |
238 #define RE_EOF 202 /* Match "" at end of file. */ | |
239 #define CURSOR 203 /* Match location of cursor. */ | |
240 | |
241 #define RE_LNUM 204 /* nr cmp Match line number */ | |
242 #define RE_COL 205 /* nr cmp Match column number */ | |
243 #define RE_VCOL 206 /* nr cmp Match virtual column number */ | |
244 | |
639 | 245 #define RE_MARK 207 /* mark cmp Match mark position */ |
246 #define RE_VISUAL 208 /* Match Visual area */ | |
247 | |
7 | 248 /* |
249 * Magic characters have a special meaning, they don't match literally. | |
250 * Magic characters are negative. This separates them from literal characters | |
251 * (possibly multi-byte). Only ASCII characters can be Magic. | |
252 */ | |
253 #define Magic(x) ((int)(x) - 256) | |
254 #define un_Magic(x) ((x) + 256) | |
255 #define is_Magic(x) ((x) < 0) | |
256 | |
257 static int no_Magic __ARGS((int x)); | |
258 static int toggle_Magic __ARGS((int x)); | |
259 | |
260 static int | |
261 no_Magic(x) | |
262 int x; | |
263 { | |
264 if (is_Magic(x)) | |
265 return un_Magic(x); | |
266 return x; | |
267 } | |
268 | |
269 static int | |
270 toggle_Magic(x) | |
271 int x; | |
272 { | |
273 if (is_Magic(x)) | |
274 return un_Magic(x); | |
275 return Magic(x); | |
276 } | |
277 | |
278 /* | |
279 * The first byte of the regexp internal "program" is actually this magic | |
280 * number; the start node begins in the second byte. It's used to catch the | |
281 * most severe mutilation of the program by the caller. | |
282 */ | |
283 | |
284 #define REGMAGIC 0234 | |
285 | |
286 /* | |
287 * Opcode notes: | |
288 * | |
289 * BRANCH The set of branches constituting a single choice are hooked | |
290 * together with their "next" pointers, since precedence prevents | |
291 * anything being concatenated to any individual branch. The | |
292 * "next" pointer of the last BRANCH in a choice points to the | |
293 * thing following the whole choice. This is also where the | |
294 * final "next" pointer of each individual branch points; each | |
295 * branch starts with the operand node of a BRANCH node. | |
296 * | |
297 * BACK Normal "next" pointers all implicitly point forward; BACK | |
298 * exists to make loop structures possible. | |
299 * | |
300 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular | |
301 * BRANCH structures using BACK. Simple cases (one character | |
302 * per match) are implemented with STAR and PLUS for speed | |
303 * and to minimize recursive plunges. | |
304 * | |
305 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX | |
306 * node, and defines the min and max limits to be used for that | |
307 * node. | |
308 * | |
309 * MOPEN,MCLOSE ...are numbered at compile time. | |
310 * ZOPEN,ZCLOSE ...ditto | |
311 */ | |
312 | |
313 /* | |
314 * A node is one char of opcode followed by two chars of "next" pointer. | |
315 * "Next" pointers are stored as two 8-bit bytes, high order first. The | |
316 * value is a positive offset from the opcode of the node containing it. | |
317 * An operand, if any, simply follows the node. (Note that much of the | |
318 * code generation knows about this implicit relationship.) | |
319 * | |
320 * Using two bytes for the "next" pointer is vast overkill for most things, | |
321 * but allows patterns to get big without disasters. | |
322 */ | |
323 #define OP(p) ((int)*(p)) | |
324 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377)) | |
325 #define OPERAND(p) ((p) + 3) | |
326 /* Obtain an operand that was stored as four bytes, MSB first. */ | |
327 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \ | |
328 + ((long)(p)[5] << 8) + (long)(p)[6]) | |
329 /* Obtain a second operand stored as four bytes. */ | |
330 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4) | |
331 /* Obtain a second single-byte operand stored after a four bytes operand. */ | |
332 #define OPERAND_CMP(p) (p)[7] | |
333 | |
334 /* | |
335 * Utility definitions. | |
336 */ | |
337 #define UCHARAT(p) ((int)*(char_u *)(p)) | |
338 | |
339 /* Used for an error (down from) vim_regcomp(): give the error message, set | |
340 * rc_did_emsg and return NULL */ | |
653 | 341 #define EMSG_RET_NULL(m) return (EMSG(m), rc_did_emsg = TRUE, (void *)NULL) |
308 | 342 #define EMSG_RET_FAIL(m) return (EMSG(m), rc_did_emsg = TRUE, FAIL) |
4444 | 343 #define EMSG2_RET_NULL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, (void *)NULL) |
344 #define EMSG2_RET_FAIL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, FAIL) | |
345 #define EMSG_ONE_RET_NULL EMSG2_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL) | |
7 | 346 |
347 #define MAX_LIMIT (32767L << 16L) | |
348 | |
349 static int re_multi_type __ARGS((int)); | |
350 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n)); | |
351 static char_u *cstrchr __ARGS((char_u *, int)); | |
352 | |
4444 | 353 #ifdef BT_REGEXP_DUMP |
354 static void regdump __ARGS((char_u *, bt_regprog_T *)); | |
355 #endif | |
7 | 356 #ifdef DEBUG |
357 static char_u *regprop __ARGS((char_u *)); | |
358 #endif | |
359 | |
4444 | 360 static char_u e_missingbracket[] = N_("E769: Missing ] after %s["); |
361 static char_u e_unmatchedpp[] = N_("E53: Unmatched %s%%("); | |
362 static char_u e_unmatchedp[] = N_("E54: Unmatched %s("); | |
363 static char_u e_unmatchedpar[] = N_("E55: Unmatched %s)"); | |
364 | |
7 | 365 #define NOT_MULTI 0 |
366 #define MULTI_ONE 1 | |
367 #define MULTI_MULT 2 | |
368 /* | |
369 * Return NOT_MULTI if c is not a "multi" operator. | |
370 * Return MULTI_ONE if c is a single "multi" operator. | |
371 * Return MULTI_MULT if c is a multi "multi" operator. | |
372 */ | |
373 static int | |
374 re_multi_type(c) | |
375 int c; | |
376 { | |
377 if (c == Magic('@') || c == Magic('=') || c == Magic('?')) | |
378 return MULTI_ONE; | |
379 if (c == Magic('*') || c == Magic('+') || c == Magic('{')) | |
380 return MULTI_MULT; | |
381 return NOT_MULTI; | |
382 } | |
383 | |
384 /* | |
385 * Flags to be passed up and down. | |
386 */ | |
387 #define HASWIDTH 0x1 /* Known never to match null string. */ | |
388 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */ | |
389 #define SPSTART 0x4 /* Starts with * or +. */ | |
390 #define HASNL 0x8 /* Contains some \n. */ | |
391 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */ | |
392 #define WORST 0 /* Worst case. */ | |
393 | |
394 /* | |
395 * When regcode is set to this value, code is not emitted and size is computed | |
396 * instead. | |
397 */ | |
398 #define JUST_CALC_SIZE ((char_u *) -1) | |
399 | |
359 | 400 static char_u *reg_prev_sub = NULL; |
401 | |
7 | 402 /* |
403 * REGEXP_INRANGE contains all characters which are always special in a [] | |
404 * range after '\'. | |
405 * REGEXP_ABBR contains all characters which act as abbreviations after '\'. | |
406 * These are: | |
407 * \n - New line (NL). | |
408 * \r - Carriage Return (CR). | |
409 * \t - Tab (TAB). | |
410 * \e - Escape (ESC). | |
411 * \b - Backspace (Ctrl_H). | |
24 | 412 * \d - Character code in decimal, eg \d123 |
413 * \o - Character code in octal, eg \o80 | |
414 * \x - Character code in hex, eg \x4a | |
415 * \u - Multibyte character code, eg \u20ac | |
416 * \U - Long multibyte character code, eg \U12345678 | |
7 | 417 */ |
418 static char_u REGEXP_INRANGE[] = "]^-n\\"; | |
24 | 419 static char_u REGEXP_ABBR[] = "nrtebdoxuU"; |
7 | 420 |
421 static int backslash_trans __ARGS((int c)); | |
167 | 422 static int get_char_class __ARGS((char_u **pp)); |
423 static int get_equi_class __ARGS((char_u **pp)); | |
424 static void reg_equi_class __ARGS((int c)); | |
425 static int get_coll_element __ARGS((char_u **pp)); | |
7 | 426 static char_u *skip_anyof __ARGS((char_u *p)); |
427 static void init_class_tab __ARGS((void)); | |
428 | |
429 /* | |
430 * Translate '\x' to its control character, except "\n", which is Magic. | |
431 */ | |
432 static int | |
433 backslash_trans(c) | |
434 int c; | |
435 { | |
436 switch (c) | |
437 { | |
438 case 'r': return CAR; | |
439 case 't': return TAB; | |
440 case 'e': return ESC; | |
441 case 'b': return BS; | |
442 } | |
443 return c; | |
444 } | |
445 | |
446 /* | |
167 | 447 * Check for a character class name "[:name:]". "pp" points to the '['. |
7 | 448 * Returns one of the CLASS_ items. CLASS_NONE means that no item was |
449 * recognized. Otherwise "pp" is advanced to after the item. | |
450 */ | |
451 static int | |
167 | 452 get_char_class(pp) |
7 | 453 char_u **pp; |
454 { | |
455 static const char *(class_names[]) = | |
456 { | |
457 "alnum:]", | |
458 #define CLASS_ALNUM 0 | |
459 "alpha:]", | |
460 #define CLASS_ALPHA 1 | |
461 "blank:]", | |
462 #define CLASS_BLANK 2 | |
463 "cntrl:]", | |
464 #define CLASS_CNTRL 3 | |
465 "digit:]", | |
466 #define CLASS_DIGIT 4 | |
467 "graph:]", | |
468 #define CLASS_GRAPH 5 | |
469 "lower:]", | |
470 #define CLASS_LOWER 6 | |
471 "print:]", | |
472 #define CLASS_PRINT 7 | |
473 "punct:]", | |
474 #define CLASS_PUNCT 8 | |
475 "space:]", | |
476 #define CLASS_SPACE 9 | |
477 "upper:]", | |
478 #define CLASS_UPPER 10 | |
479 "xdigit:]", | |
480 #define CLASS_XDIGIT 11 | |
481 "tab:]", | |
482 #define CLASS_TAB 12 | |
483 "return:]", | |
484 #define CLASS_RETURN 13 | |
485 "backspace:]", | |
486 #define CLASS_BACKSPACE 14 | |
487 "escape:]", | |
488 #define CLASS_ESCAPE 15 | |
489 }; | |
490 #define CLASS_NONE 99 | |
491 int i; | |
492 | |
493 if ((*pp)[1] == ':') | |
494 { | |
1877 | 495 for (i = 0; i < (int)(sizeof(class_names) / sizeof(*class_names)); ++i) |
7 | 496 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0) |
497 { | |
498 *pp += STRLEN(class_names[i]) + 2; | |
499 return i; | |
500 } | |
501 } | |
502 return CLASS_NONE; | |
503 } | |
504 | |
505 /* | |
506 * Specific version of character class functions. | |
507 * Using a table to keep this fast. | |
508 */ | |
509 static short class_tab[256]; | |
510 | |
511 #define RI_DIGIT 0x01 | |
512 #define RI_HEX 0x02 | |
513 #define RI_OCTAL 0x04 | |
514 #define RI_WORD 0x08 | |
515 #define RI_HEAD 0x10 | |
516 #define RI_ALPHA 0x20 | |
517 #define RI_LOWER 0x40 | |
518 #define RI_UPPER 0x80 | |
519 #define RI_WHITE 0x100 | |
520 | |
521 static void | |
522 init_class_tab() | |
523 { | |
524 int i; | |
525 static int done = FALSE; | |
526 | |
527 if (done) | |
528 return; | |
529 | |
530 for (i = 0; i < 256; ++i) | |
531 { | |
532 if (i >= '0' && i <= '7') | |
533 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD; | |
534 else if (i >= '8' && i <= '9') | |
535 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD; | |
536 else if (i >= 'a' && i <= 'f') | |
537 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER; | |
538 #ifdef EBCDIC | |
539 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r') | |
540 || (i >= 's' && i <= 'z')) | |
541 #else | |
542 else if (i >= 'g' && i <= 'z') | |
543 #endif | |
544 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER; | |
545 else if (i >= 'A' && i <= 'F') | |
546 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER; | |
547 #ifdef EBCDIC | |
548 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R') | |
549 || (i >= 'S' && i <= 'Z')) | |
550 #else | |
551 else if (i >= 'G' && i <= 'Z') | |
552 #endif | |
553 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER; | |
554 else if (i == '_') | |
555 class_tab[i] = RI_WORD + RI_HEAD; | |
556 else | |
557 class_tab[i] = 0; | |
558 } | |
559 class_tab[' '] |= RI_WHITE; | |
560 class_tab['\t'] |= RI_WHITE; | |
561 done = TRUE; | |
562 } | |
563 | |
564 #ifdef FEAT_MBYTE | |
565 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT)) | |
566 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX)) | |
567 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL)) | |
568 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD)) | |
569 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD)) | |
570 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA)) | |
571 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER)) | |
572 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER)) | |
573 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE)) | |
574 #else | |
575 # define ri_digit(c) (class_tab[c] & RI_DIGIT) | |
576 # define ri_hex(c) (class_tab[c] & RI_HEX) | |
577 # define ri_octal(c) (class_tab[c] & RI_OCTAL) | |
578 # define ri_word(c) (class_tab[c] & RI_WORD) | |
579 # define ri_head(c) (class_tab[c] & RI_HEAD) | |
580 # define ri_alpha(c) (class_tab[c] & RI_ALPHA) | |
581 # define ri_lower(c) (class_tab[c] & RI_LOWER) | |
582 # define ri_upper(c) (class_tab[c] & RI_UPPER) | |
583 # define ri_white(c) (class_tab[c] & RI_WHITE) | |
584 #endif | |
585 | |
586 /* flags for regflags */ | |
587 #define RF_ICASE 1 /* ignore case */ | |
588 #define RF_NOICASE 2 /* don't ignore case */ | |
589 #define RF_HASNL 4 /* can match a NL */ | |
590 #define RF_ICOMBINE 8 /* ignore combining characters */ | |
591 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */ | |
592 | |
593 /* | |
594 * Global work variables for vim_regcomp(). | |
595 */ | |
596 | |
597 static char_u *regparse; /* Input-scan pointer. */ | |
598 static int prevchr_len; /* byte length of previous char */ | |
599 static int num_complex_braces; /* Complex \{...} count */ | |
600 static int regnpar; /* () count. */ | |
601 #ifdef FEAT_SYN_HL | |
602 static int regnzpar; /* \z() count. */ | |
603 static int re_has_z; /* \z item detected */ | |
604 #endif | |
605 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */ | |
606 static long regsize; /* Code size. */ | |
2010 | 607 static int reg_toolong; /* TRUE when offset out of range */ |
7 | 608 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */ |
609 static unsigned regflags; /* RF_ flags for prog */ | |
610 static long brace_min[10]; /* Minimums for complex brace repeats */ | |
611 static long brace_max[10]; /* Maximums for complex brace repeats */ | |
612 static int brace_count[10]; /* Current counts for complex brace repeats */ | |
613 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
614 static int had_eol; /* TRUE when EOL found by vim_regcomp() */ | |
615 #endif | |
616 static int one_exactly = FALSE; /* only do one char for EXACTLY */ | |
617 | |
618 static int reg_magic; /* magicness of the pattern: */ | |
619 #define MAGIC_NONE 1 /* "\V" very unmagic */ | |
620 #define MAGIC_OFF 2 /* "\M" or 'magic' off */ | |
621 #define MAGIC_ON 3 /* "\m" or 'magic' */ | |
622 #define MAGIC_ALL 4 /* "\v" very magic */ | |
623 | |
624 static int reg_string; /* matching with a string instead of a buffer | |
625 line */ | |
481 | 626 static int reg_strict; /* "[abc" is illegal */ |
7 | 627 |
628 /* | |
629 * META contains all characters that may be magic, except '^' and '$'. | |
630 */ | |
631 | |
632 #ifdef EBCDIC | |
633 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~"; | |
634 #else | |
635 /* META[] is used often enough to justify turning it into a table. */ | |
636 static char_u META_flags[] = { | |
637 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, | |
638 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, | |
639 /* % & ( ) * + . */ | |
640 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, | |
641 /* 1 2 3 4 5 6 7 8 9 < = > ? */ | |
642 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, | |
643 /* @ A C D F H I K L M O */ | |
644 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, | |
645 /* P S U V W X Z [ _ */ | |
646 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1, | |
647 /* a c d f h i k l m n o */ | |
648 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, | |
649 /* p s u v w x z { | ~ */ | |
650 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1 | |
651 }; | |
652 #endif | |
653 | |
4444 | 654 static int curchr; /* currently parsed character */ |
655 /* Previous character. Note: prevchr is sometimes -1 when we are not at the | |
656 * start, eg in /[ ^I]^ the pattern was never found even if it existed, | |
657 * because ^ was taken to be magic -- webb */ | |
658 static int prevchr; | |
659 static int prevprevchr; /* previous-previous character */ | |
660 static int nextchr; /* used for ungetchr() */ | |
7 | 661 |
662 /* arguments for reg() */ | |
663 #define REG_NOPAREN 0 /* toplevel reg() */ | |
664 #define REG_PAREN 1 /* \(\) */ | |
665 #define REG_ZPAREN 2 /* \z(\) */ | |
666 #define REG_NPAREN 3 /* \%(\) */ | |
667 | |
668 /* | |
669 * Forward declarations for vim_regcomp()'s friends. | |
670 */ | |
671 static void initchr __ARGS((char_u *)); | |
672 static int getchr __ARGS((void)); | |
673 static void skipchr_keepstart __ARGS((void)); | |
674 static int peekchr __ARGS((void)); | |
675 static void skipchr __ARGS((void)); | |
676 static void ungetchr __ARGS((void)); | |
24 | 677 static int gethexchrs __ARGS((int maxinputlen)); |
678 static int getoctchrs __ARGS((void)); | |
679 static int getdecchrs __ARGS((void)); | |
680 static int coll_get_char __ARGS((void)); | |
7 | 681 static void regcomp_start __ARGS((char_u *expr, int flags)); |
682 static char_u *reg __ARGS((int, int *)); | |
683 static char_u *regbranch __ARGS((int *flagp)); | |
684 static char_u *regconcat __ARGS((int *flagp)); | |
685 static char_u *regpiece __ARGS((int *)); | |
686 static char_u *regatom __ARGS((int *)); | |
687 static char_u *regnode __ARGS((int)); | |
714 | 688 #ifdef FEAT_MBYTE |
689 static int use_multibytecode __ARGS((int c)); | |
690 #endif | |
7 | 691 static int prog_magic_wrong __ARGS((void)); |
692 static char_u *regnext __ARGS((char_u *)); | |
693 static void regc __ARGS((int b)); | |
694 #ifdef FEAT_MBYTE | |
695 static void regmbc __ARGS((int c)); | |
2974 | 696 # define REGMBC(x) regmbc(x); |
697 # define CASEMBC(x) case x: | |
167 | 698 #else |
699 # define regmbc(c) regc(c) | |
2974 | 700 # define REGMBC(x) |
701 # define CASEMBC(x) | |
7 | 702 #endif |
703 static void reginsert __ARGS((int, char_u *)); | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
704 static void reginsert_nr __ARGS((int op, long val, char_u *opnd)); |
7 | 705 static void reginsert_limits __ARGS((int, long, long, char_u *)); |
706 static char_u *re_put_long __ARGS((char_u *pr, long_u val)); | |
707 static int read_limits __ARGS((long *, long *)); | |
708 static void regtail __ARGS((char_u *, char_u *)); | |
709 static void regoptail __ARGS((char_u *, char_u *)); | |
710 | |
4444 | 711 static regengine_T bt_regengine; |
712 static regengine_T nfa_regengine; | |
713 | |
7 | 714 /* |
715 * Return TRUE if compiled regular expression "prog" can match a line break. | |
716 */ | |
717 int | |
718 re_multiline(prog) | |
719 regprog_T *prog; | |
720 { | |
721 return (prog->regflags & RF_HASNL); | |
722 } | |
723 | |
724 /* | |
725 * Return TRUE if compiled regular expression "prog" looks before the start | |
726 * position (pattern contains "\@<=" or "\@<!"). | |
727 */ | |
728 int | |
729 re_lookbehind(prog) | |
730 regprog_T *prog; | |
731 { | |
732 return (prog->regflags & RF_LOOKBH); | |
733 } | |
734 | |
735 /* | |
167 | 736 * Check for an equivalence class name "[=a=]". "pp" points to the '['. |
737 * Returns a character representing the class. Zero means that no item was | |
738 * recognized. Otherwise "pp" is advanced to after the item. | |
739 */ | |
740 static int | |
741 get_equi_class(pp) | |
742 char_u **pp; | |
743 { | |
744 int c; | |
745 int l = 1; | |
746 char_u *p = *pp; | |
747 | |
748 if (p[1] == '=') | |
749 { | |
750 #ifdef FEAT_MBYTE | |
751 if (has_mbyte) | |
474 | 752 l = (*mb_ptr2len)(p + 2); |
167 | 753 #endif |
754 if (p[l + 2] == '=' && p[l + 3] == ']') | |
755 { | |
756 #ifdef FEAT_MBYTE | |
757 if (has_mbyte) | |
758 c = mb_ptr2char(p + 2); | |
759 else | |
760 #endif | |
761 c = p[2]; | |
762 *pp += l + 4; | |
763 return c; | |
764 } | |
765 } | |
766 return 0; | |
767 } | |
768 | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
769 #ifdef EBCDIC |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
770 /* |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
771 * Table for equivalence class "c". (IBM-1047) |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
772 */ |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
773 char *EQUIVAL_CLASS_C[16] = { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
774 "A\x62\x63\x64\x65\x66\x67", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
775 "C\x68", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
776 "E\x71\x72\x73\x74", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
777 "I\x75\x76\x77\x78", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
778 "N\x69", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
779 "O\xEB\xEC\xED\xEE\xEF", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
780 "U\xFB\xFC\xFD\xFE", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
781 "Y\xBA", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
782 "a\x42\x43\x44\x45\x46\x47", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
783 "c\x48", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
784 "e\x51\x52\x53\x54", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
785 "i\x55\x56\x57\x58", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
786 "n\x49", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
787 "o\xCB\xCC\xCD\xCE\xCF", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
788 "u\xDB\xDC\xDD\xDE", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
789 "y\x8D\xDF", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
790 }; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
791 #endif |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
792 |
167 | 793 /* |
794 * Produce the bytes for equivalence class "c". | |
795 * Currently only handles latin1, latin9 and utf-8. | |
4444 | 796 * NOTE: When changing this function, also change nfa_emit_equi_class() |
167 | 797 */ |
798 static void | |
799 reg_equi_class(c) | |
800 int c; | |
801 { | |
802 #ifdef FEAT_MBYTE | |
803 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0 | |
492 | 804 || STRCMP(p_enc, "iso-8859-15") == 0) |
167 | 805 #endif |
806 { | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
807 #ifdef EBCDIC |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
808 int i; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
809 |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
810 /* This might be slower than switch/case below. */ |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
811 for (i = 0; i < 16; i++) |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
812 { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
813 if (vim_strchr(EQUIVAL_CLASS_C[i], c) != NULL) |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
814 { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
815 char *p = EQUIVAL_CLASS_C[i]; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
816 |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
817 while (*p != 0) |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
818 regmbc(*p++); |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
819 return; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
820 } |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
821 } |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
822 #else |
167 | 823 switch (c) |
824 { | |
236 | 825 case 'A': case '\300': case '\301': case '\302': |
2974 | 826 CASEMBC(0x100) CASEMBC(0x102) CASEMBC(0x104) CASEMBC(0x1cd) |
827 CASEMBC(0x1de) CASEMBC(0x1e0) CASEMBC(0x1ea2) | |
236 | 828 case '\303': case '\304': case '\305': |
829 regmbc('A'); regmbc('\300'); regmbc('\301'); | |
830 regmbc('\302'); regmbc('\303'); regmbc('\304'); | |
831 regmbc('\305'); | |
2974 | 832 REGMBC(0x100) REGMBC(0x102) REGMBC(0x104) |
833 REGMBC(0x1cd) REGMBC(0x1de) REGMBC(0x1e0) | |
834 REGMBC(0x1ea2) | |
835 return; | |
836 case 'B': CASEMBC(0x1e02) CASEMBC(0x1e06) | |
837 regmbc('B'); REGMBC(0x1e02) REGMBC(0x1e06) | |
167 | 838 return; |
236 | 839 case 'C': case '\307': |
2974 | 840 CASEMBC(0x106) CASEMBC(0x108) CASEMBC(0x10a) CASEMBC(0x10c) |
236 | 841 regmbc('C'); regmbc('\307'); |
2974 | 842 REGMBC(0x106) REGMBC(0x108) REGMBC(0x10a) |
843 REGMBC(0x10c) | |
844 return; | |
845 case 'D': CASEMBC(0x10e) CASEMBC(0x110) CASEMBC(0x1e0a) | |
846 CASEMBC(0x1e0e) CASEMBC(0x1e10) | |
847 regmbc('D'); REGMBC(0x10e) REGMBC(0x110) | |
848 REGMBC(0x1e0a) REGMBC(0x1e0e) REGMBC(0x1e10) | |
167 | 849 return; |
236 | 850 case 'E': case '\310': case '\311': case '\312': case '\313': |
2974 | 851 CASEMBC(0x112) CASEMBC(0x114) CASEMBC(0x116) CASEMBC(0x118) |
852 CASEMBC(0x11a) CASEMBC(0x1eba) CASEMBC(0x1ebc) | |
236 | 853 regmbc('E'); regmbc('\310'); regmbc('\311'); |
854 regmbc('\312'); regmbc('\313'); | |
2974 | 855 REGMBC(0x112) REGMBC(0x114) REGMBC(0x116) |
856 REGMBC(0x118) REGMBC(0x11a) REGMBC(0x1eba) | |
857 REGMBC(0x1ebc) | |
858 return; | |
859 case 'F': CASEMBC(0x1e1e) | |
860 regmbc('F'); REGMBC(0x1e1e) | |
861 return; | |
862 case 'G': CASEMBC(0x11c) CASEMBC(0x11e) CASEMBC(0x120) | |
863 CASEMBC(0x122) CASEMBC(0x1e4) CASEMBC(0x1e6) CASEMBC(0x1f4) | |
864 CASEMBC(0x1e20) | |
865 regmbc('G'); REGMBC(0x11c) REGMBC(0x11e) | |
866 REGMBC(0x120) REGMBC(0x122) REGMBC(0x1e4) | |
867 REGMBC(0x1e6) REGMBC(0x1f4) REGMBC(0x1e20) | |
868 return; | |
869 case 'H': CASEMBC(0x124) CASEMBC(0x126) CASEMBC(0x1e22) | |
870 CASEMBC(0x1e26) CASEMBC(0x1e28) | |
871 regmbc('H'); REGMBC(0x124) REGMBC(0x126) | |
872 REGMBC(0x1e22) REGMBC(0x1e26) REGMBC(0x1e28) | |
167 | 873 return; |
236 | 874 case 'I': case '\314': case '\315': case '\316': case '\317': |
2974 | 875 CASEMBC(0x128) CASEMBC(0x12a) CASEMBC(0x12c) CASEMBC(0x12e) |
876 CASEMBC(0x130) CASEMBC(0x1cf) CASEMBC(0x1ec8) | |
236 | 877 regmbc('I'); regmbc('\314'); regmbc('\315'); |
878 regmbc('\316'); regmbc('\317'); | |
2974 | 879 REGMBC(0x128) REGMBC(0x12a) REGMBC(0x12c) |
880 REGMBC(0x12e) REGMBC(0x130) REGMBC(0x1cf) | |
881 REGMBC(0x1ec8) | |
882 return; | |
883 case 'J': CASEMBC(0x134) | |
884 regmbc('J'); REGMBC(0x134) | |
885 return; | |
886 case 'K': CASEMBC(0x136) CASEMBC(0x1e8) CASEMBC(0x1e30) | |
887 CASEMBC(0x1e34) | |
888 regmbc('K'); REGMBC(0x136) REGMBC(0x1e8) | |
889 REGMBC(0x1e30) REGMBC(0x1e34) | |
890 return; | |
891 case 'L': CASEMBC(0x139) CASEMBC(0x13b) CASEMBC(0x13d) | |
892 CASEMBC(0x13f) CASEMBC(0x141) CASEMBC(0x1e3a) | |
893 regmbc('L'); REGMBC(0x139) REGMBC(0x13b) | |
894 REGMBC(0x13d) REGMBC(0x13f) REGMBC(0x141) | |
895 REGMBC(0x1e3a) | |
896 return; | |
897 case 'M': CASEMBC(0x1e3e) CASEMBC(0x1e40) | |
898 regmbc('M'); REGMBC(0x1e3e) REGMBC(0x1e40) | |
167 | 899 return; |
236 | 900 case 'N': case '\321': |
2974 | 901 CASEMBC(0x143) CASEMBC(0x145) CASEMBC(0x147) CASEMBC(0x1e44) |
902 CASEMBC(0x1e48) | |
236 | 903 regmbc('N'); regmbc('\321'); |
2974 | 904 REGMBC(0x143) REGMBC(0x145) REGMBC(0x147) |
905 REGMBC(0x1e44) REGMBC(0x1e48) | |
167 | 906 return; |
236 | 907 case 'O': case '\322': case '\323': case '\324': case '\325': |
2974 | 908 case '\326': case '\330': |
909 CASEMBC(0x14c) CASEMBC(0x14e) CASEMBC(0x150) CASEMBC(0x1a0) | |
910 CASEMBC(0x1d1) CASEMBC(0x1ea) CASEMBC(0x1ec) CASEMBC(0x1ece) | |
236 | 911 regmbc('O'); regmbc('\322'); regmbc('\323'); |
912 regmbc('\324'); regmbc('\325'); regmbc('\326'); | |
2974 | 913 regmbc('\330'); |
914 REGMBC(0x14c) REGMBC(0x14e) REGMBC(0x150) | |
915 REGMBC(0x1a0) REGMBC(0x1d1) REGMBC(0x1ea) | |
916 REGMBC(0x1ec) REGMBC(0x1ece) | |
917 return; | |
918 case 'P': case 0x1e54: case 0x1e56: | |
919 regmbc('P'); REGMBC(0x1e54) REGMBC(0x1e56) | |
920 return; | |
921 case 'R': CASEMBC(0x154) CASEMBC(0x156) CASEMBC(0x158) | |
922 CASEMBC(0x1e58) CASEMBC(0x1e5e) | |
923 regmbc('R'); REGMBC(0x154) REGMBC(0x156) REGMBC(0x158) | |
924 REGMBC(0x1e58) REGMBC(0x1e5e) | |
925 return; | |
926 case 'S': CASEMBC(0x15a) CASEMBC(0x15c) CASEMBC(0x15e) | |
927 CASEMBC(0x160) CASEMBC(0x1e60) | |
928 regmbc('S'); REGMBC(0x15a) REGMBC(0x15c) | |
929 REGMBC(0x15e) REGMBC(0x160) REGMBC(0x1e60) | |
930 return; | |
931 case 'T': CASEMBC(0x162) CASEMBC(0x164) CASEMBC(0x166) | |
932 CASEMBC(0x1e6a) CASEMBC(0x1e6e) | |
933 regmbc('T'); REGMBC(0x162) REGMBC(0x164) | |
934 REGMBC(0x166) REGMBC(0x1e6a) REGMBC(0x1e6e) | |
167 | 935 return; |
236 | 936 case 'U': case '\331': case '\332': case '\333': case '\334': |
2974 | 937 CASEMBC(0x168) CASEMBC(0x16a) CASEMBC(0x16c) CASEMBC(0x16e) |
938 CASEMBC(0x170) CASEMBC(0x172) CASEMBC(0x1af) CASEMBC(0x1d3) | |
939 CASEMBC(0x1ee6) | |
236 | 940 regmbc('U'); regmbc('\331'); regmbc('\332'); |
941 regmbc('\333'); regmbc('\334'); | |
2974 | 942 REGMBC(0x168) REGMBC(0x16a) REGMBC(0x16c) |
943 REGMBC(0x16e) REGMBC(0x170) REGMBC(0x172) | |
944 REGMBC(0x1af) REGMBC(0x1d3) REGMBC(0x1ee6) | |
945 return; | |
946 case 'V': CASEMBC(0x1e7c) | |
947 regmbc('V'); REGMBC(0x1e7c) | |
948 return; | |
949 case 'W': CASEMBC(0x174) CASEMBC(0x1e80) CASEMBC(0x1e82) | |
950 CASEMBC(0x1e84) CASEMBC(0x1e86) | |
951 regmbc('W'); REGMBC(0x174) REGMBC(0x1e80) | |
952 REGMBC(0x1e82) REGMBC(0x1e84) REGMBC(0x1e86) | |
953 return; | |
954 case 'X': CASEMBC(0x1e8a) CASEMBC(0x1e8c) | |
955 regmbc('X'); REGMBC(0x1e8a) REGMBC(0x1e8c) | |
167 | 956 return; |
236 | 957 case 'Y': case '\335': |
2974 | 958 CASEMBC(0x176) CASEMBC(0x178) CASEMBC(0x1e8e) CASEMBC(0x1ef2) |
959 CASEMBC(0x1ef6) CASEMBC(0x1ef8) | |
236 | 960 regmbc('Y'); regmbc('\335'); |
2974 | 961 REGMBC(0x176) REGMBC(0x178) REGMBC(0x1e8e) |
962 REGMBC(0x1ef2) REGMBC(0x1ef6) REGMBC(0x1ef8) | |
963 return; | |
964 case 'Z': CASEMBC(0x179) CASEMBC(0x17b) CASEMBC(0x17d) | |
965 CASEMBC(0x1b5) CASEMBC(0x1e90) CASEMBC(0x1e94) | |
966 regmbc('Z'); REGMBC(0x179) REGMBC(0x17b) | |
967 REGMBC(0x17d) REGMBC(0x1b5) REGMBC(0x1e90) | |
968 REGMBC(0x1e94) | |
167 | 969 return; |
236 | 970 case 'a': case '\340': case '\341': case '\342': |
971 case '\343': case '\344': case '\345': | |
2974 | 972 CASEMBC(0x101) CASEMBC(0x103) CASEMBC(0x105) CASEMBC(0x1ce) |
973 CASEMBC(0x1df) CASEMBC(0x1e1) CASEMBC(0x1ea3) | |
236 | 974 regmbc('a'); regmbc('\340'); regmbc('\341'); |
975 regmbc('\342'); regmbc('\343'); regmbc('\344'); | |
976 regmbc('\345'); | |
2974 | 977 REGMBC(0x101) REGMBC(0x103) REGMBC(0x105) |
978 REGMBC(0x1ce) REGMBC(0x1df) REGMBC(0x1e1) | |
979 REGMBC(0x1ea3) | |
980 return; | |
981 case 'b': CASEMBC(0x1e03) CASEMBC(0x1e07) | |
982 regmbc('b'); REGMBC(0x1e03) REGMBC(0x1e07) | |
167 | 983 return; |
236 | 984 case 'c': case '\347': |
2974 | 985 CASEMBC(0x107) CASEMBC(0x109) CASEMBC(0x10b) CASEMBC(0x10d) |
236 | 986 regmbc('c'); regmbc('\347'); |
2974 | 987 REGMBC(0x107) REGMBC(0x109) REGMBC(0x10b) |
988 REGMBC(0x10d) | |
989 return; | |
990 case 'd': CASEMBC(0x10f) CASEMBC(0x111) CASEMBC(0x1d0b) | |
991 CASEMBC(0x1e11) | |
992 regmbc('d'); REGMBC(0x10f) REGMBC(0x111) | |
993 REGMBC(0x1e0b) REGMBC(0x01e0f) REGMBC(0x1e11) | |
167 | 994 return; |
236 | 995 case 'e': case '\350': case '\351': case '\352': case '\353': |
2974 | 996 CASEMBC(0x113) CASEMBC(0x115) CASEMBC(0x117) CASEMBC(0x119) |
997 CASEMBC(0x11b) CASEMBC(0x1ebb) CASEMBC(0x1ebd) | |
236 | 998 regmbc('e'); regmbc('\350'); regmbc('\351'); |
999 regmbc('\352'); regmbc('\353'); | |
2974 | 1000 REGMBC(0x113) REGMBC(0x115) REGMBC(0x117) |
1001 REGMBC(0x119) REGMBC(0x11b) REGMBC(0x1ebb) | |
1002 REGMBC(0x1ebd) | |
1003 return; | |
1004 case 'f': CASEMBC(0x1e1f) | |
1005 regmbc('f'); REGMBC(0x1e1f) | |
1006 return; | |
1007 case 'g': CASEMBC(0x11d) CASEMBC(0x11f) CASEMBC(0x121) | |
1008 CASEMBC(0x123) CASEMBC(0x1e5) CASEMBC(0x1e7) CASEMBC(0x1f5) | |
1009 CASEMBC(0x1e21) | |
1010 regmbc('g'); REGMBC(0x11d) REGMBC(0x11f) | |
1011 REGMBC(0x121) REGMBC(0x123) REGMBC(0x1e5) | |
1012 REGMBC(0x1e7) REGMBC(0x1f5) REGMBC(0x1e21) | |
1013 return; | |
1014 case 'h': CASEMBC(0x125) CASEMBC(0x127) CASEMBC(0x1e23) | |
1015 CASEMBC(0x1e27) CASEMBC(0x1e29) CASEMBC(0x1e96) | |
1016 regmbc('h'); REGMBC(0x125) REGMBC(0x127) | |
1017 REGMBC(0x1e23) REGMBC(0x1e27) REGMBC(0x1e29) | |
1018 REGMBC(0x1e96) | |
167 | 1019 return; |
236 | 1020 case 'i': case '\354': case '\355': case '\356': case '\357': |
2974 | 1021 CASEMBC(0x129) CASEMBC(0x12b) CASEMBC(0x12d) CASEMBC(0x12f) |
1022 CASEMBC(0x1d0) CASEMBC(0x1ec9) | |
236 | 1023 regmbc('i'); regmbc('\354'); regmbc('\355'); |
1024 regmbc('\356'); regmbc('\357'); | |
2974 | 1025 REGMBC(0x129) REGMBC(0x12b) REGMBC(0x12d) |
1026 REGMBC(0x12f) REGMBC(0x1d0) REGMBC(0x1ec9) | |
1027 return; | |
1028 case 'j': CASEMBC(0x135) CASEMBC(0x1f0) | |
1029 regmbc('j'); REGMBC(0x135) REGMBC(0x1f0) | |
1030 return; | |
1031 case 'k': CASEMBC(0x137) CASEMBC(0x1e9) CASEMBC(0x1e31) | |
1032 CASEMBC(0x1e35) | |
1033 regmbc('k'); REGMBC(0x137) REGMBC(0x1e9) | |
1034 REGMBC(0x1e31) REGMBC(0x1e35) | |
1035 return; | |
1036 case 'l': CASEMBC(0x13a) CASEMBC(0x13c) CASEMBC(0x13e) | |
1037 CASEMBC(0x140) CASEMBC(0x142) CASEMBC(0x1e3b) | |
1038 regmbc('l'); REGMBC(0x13a) REGMBC(0x13c) | |
1039 REGMBC(0x13e) REGMBC(0x140) REGMBC(0x142) | |
1040 REGMBC(0x1e3b) | |
1041 return; | |
1042 case 'm': CASEMBC(0x1e3f) CASEMBC(0x1e41) | |
1043 regmbc('m'); REGMBC(0x1e3f) REGMBC(0x1e41) | |
167 | 1044 return; |
236 | 1045 case 'n': case '\361': |
2974 | 1046 CASEMBC(0x144) CASEMBC(0x146) CASEMBC(0x148) CASEMBC(0x149) |
1047 CASEMBC(0x1e45) CASEMBC(0x1e49) | |
236 | 1048 regmbc('n'); regmbc('\361'); |
2974 | 1049 REGMBC(0x144) REGMBC(0x146) REGMBC(0x148) |
1050 REGMBC(0x149) REGMBC(0x1e45) REGMBC(0x1e49) | |
167 | 1051 return; |
236 | 1052 case 'o': case '\362': case '\363': case '\364': case '\365': |
2974 | 1053 case '\366': case '\370': |
1054 CASEMBC(0x14d) CASEMBC(0x14f) CASEMBC(0x151) CASEMBC(0x1a1) | |
1055 CASEMBC(0x1d2) CASEMBC(0x1eb) CASEMBC(0x1ed) CASEMBC(0x1ecf) | |
236 | 1056 regmbc('o'); regmbc('\362'); regmbc('\363'); |
1057 regmbc('\364'); regmbc('\365'); regmbc('\366'); | |
2974 | 1058 regmbc('\370'); |
1059 REGMBC(0x14d) REGMBC(0x14f) REGMBC(0x151) | |
1060 REGMBC(0x1a1) REGMBC(0x1d2) REGMBC(0x1eb) | |
1061 REGMBC(0x1ed) REGMBC(0x1ecf) | |
1062 return; | |
1063 case 'p': CASEMBC(0x1e55) CASEMBC(0x1e57) | |
1064 regmbc('p'); REGMBC(0x1e55) REGMBC(0x1e57) | |
1065 return; | |
1066 case 'r': CASEMBC(0x155) CASEMBC(0x157) CASEMBC(0x159) | |
1067 CASEMBC(0x1e59) CASEMBC(0x1e5f) | |
1068 regmbc('r'); REGMBC(0x155) REGMBC(0x157) REGMBC(0x159) | |
1069 REGMBC(0x1e59) REGMBC(0x1e5f) | |
1070 return; | |
1071 case 's': CASEMBC(0x15b) CASEMBC(0x15d) CASEMBC(0x15f) | |
1072 CASEMBC(0x161) CASEMBC(0x1e61) | |
1073 regmbc('s'); REGMBC(0x15b) REGMBC(0x15d) | |
1074 REGMBC(0x15f) REGMBC(0x161) REGMBC(0x1e61) | |
1075 return; | |
1076 case 't': CASEMBC(0x163) CASEMBC(0x165) CASEMBC(0x167) | |
1077 CASEMBC(0x1e6b) CASEMBC(0x1e6f) CASEMBC(0x1e97) | |
1078 regmbc('t'); REGMBC(0x163) REGMBC(0x165) REGMBC(0x167) | |
1079 REGMBC(0x1e6b) REGMBC(0x1e6f) REGMBC(0x1e97) | |
167 | 1080 return; |
236 | 1081 case 'u': case '\371': case '\372': case '\373': case '\374': |
2974 | 1082 CASEMBC(0x169) CASEMBC(0x16b) CASEMBC(0x16d) CASEMBC(0x16f) |
1083 CASEMBC(0x171) CASEMBC(0x173) CASEMBC(0x1b0) CASEMBC(0x1d4) | |
1084 CASEMBC(0x1ee7) | |
236 | 1085 regmbc('u'); regmbc('\371'); regmbc('\372'); |
1086 regmbc('\373'); regmbc('\374'); | |
2974 | 1087 REGMBC(0x169) REGMBC(0x16b) REGMBC(0x16d) |
1088 REGMBC(0x16f) REGMBC(0x171) REGMBC(0x173) | |
1089 REGMBC(0x1b0) REGMBC(0x1d4) REGMBC(0x1ee7) | |
1090 return; | |
1091 case 'v': CASEMBC(0x1e7d) | |
1092 regmbc('v'); REGMBC(0x1e7d) | |
1093 return; | |
1094 case 'w': CASEMBC(0x175) CASEMBC(0x1e81) CASEMBC(0x1e83) | |
1095 CASEMBC(0x1e85) CASEMBC(0x1e87) CASEMBC(0x1e98) | |
1096 regmbc('w'); REGMBC(0x175) REGMBC(0x1e81) | |
1097 REGMBC(0x1e83) REGMBC(0x1e85) REGMBC(0x1e87) | |
1098 REGMBC(0x1e98) | |
1099 return; | |
1100 case 'x': CASEMBC(0x1e8b) CASEMBC(0x1e8d) | |
1101 regmbc('x'); REGMBC(0x1e8b) REGMBC(0x1e8d) | |
167 | 1102 return; |
236 | 1103 case 'y': case '\375': case '\377': |
2974 | 1104 CASEMBC(0x177) CASEMBC(0x1e8f) CASEMBC(0x1e99) |
1105 CASEMBC(0x1ef3) CASEMBC(0x1ef7) CASEMBC(0x1ef9) | |
236 | 1106 regmbc('y'); regmbc('\375'); regmbc('\377'); |
2974 | 1107 REGMBC(0x177) REGMBC(0x1e8f) REGMBC(0x1e99) |
1108 REGMBC(0x1ef3) REGMBC(0x1ef7) REGMBC(0x1ef9) | |
1109 return; | |
1110 case 'z': CASEMBC(0x17a) CASEMBC(0x17c) CASEMBC(0x17e) | |
1111 CASEMBC(0x1b6) CASEMBC(0x1e91) CASEMBC(0x1e95) | |
1112 regmbc('z'); REGMBC(0x17a) REGMBC(0x17c) | |
1113 REGMBC(0x17e) REGMBC(0x1b6) REGMBC(0x1e91) | |
1114 REGMBC(0x1e95) | |
167 | 1115 return; |
1116 } | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
1117 #endif |
167 | 1118 } |
1119 regmbc(c); | |
1120 } | |
1121 | |
1122 /* | |
1123 * Check for a collating element "[.a.]". "pp" points to the '['. | |
1124 * Returns a character. Zero means that no item was recognized. Otherwise | |
1125 * "pp" is advanced to after the item. | |
1126 * Currently only single characters are recognized! | |
1127 */ | |
1128 static int | |
1129 get_coll_element(pp) | |
1130 char_u **pp; | |
1131 { | |
1132 int c; | |
1133 int l = 1; | |
1134 char_u *p = *pp; | |
1135 | |
1136 if (p[1] == '.') | |
1137 { | |
1138 #ifdef FEAT_MBYTE | |
1139 if (has_mbyte) | |
474 | 1140 l = (*mb_ptr2len)(p + 2); |
167 | 1141 #endif |
1142 if (p[l + 2] == '.' && p[l + 3] == ']') | |
1143 { | |
1144 #ifdef FEAT_MBYTE | |
1145 if (has_mbyte) | |
1146 c = mb_ptr2char(p + 2); | |
1147 else | |
1148 #endif | |
1149 c = p[2]; | |
1150 *pp += l + 4; | |
1151 return c; | |
1152 } | |
1153 } | |
1154 return 0; | |
1155 } | |
1156 | |
1157 | |
1158 /* | |
1159 * Skip over a "[]" range. | |
1160 * "p" must point to the character after the '['. | |
1161 * The returned pointer is on the matching ']', or the terminating NUL. | |
1162 */ | |
1163 static char_u * | |
1164 skip_anyof(p) | |
1165 char_u *p; | |
1166 { | |
1167 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
1168 int cpo_bsl; /* 'cpoptions' contains '\' flag */ | |
1169 #ifdef FEAT_MBYTE | |
1170 int l; | |
1171 #endif | |
1172 | |
410 | 1173 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; |
1174 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
167 | 1175 |
1176 if (*p == '^') /* Complement of range. */ | |
1177 ++p; | |
1178 if (*p == ']' || *p == '-') | |
1179 ++p; | |
1180 while (*p != NUL && *p != ']') | |
1181 { | |
1182 #ifdef FEAT_MBYTE | |
474 | 1183 if (has_mbyte && (l = (*mb_ptr2len)(p)) > 1) |
167 | 1184 p += l; |
1185 else | |
1186 #endif | |
1187 if (*p == '-') | |
1188 { | |
1189 ++p; | |
1190 if (*p != ']' && *p != NUL) | |
1191 mb_ptr_adv(p); | |
1192 } | |
1193 else if (*p == '\\' | |
1194 && !cpo_bsl | |
1195 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL | |
1196 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL))) | |
1197 p += 2; | |
1198 else if (*p == '[') | |
1199 { | |
1200 if (get_char_class(&p) == CLASS_NONE | |
1201 && get_equi_class(&p) == 0 | |
1202 && get_coll_element(&p) == 0) | |
1203 ++p; /* It was not a class name */ | |
1204 } | |
1205 else | |
1206 ++p; | |
1207 } | |
1208 | |
1209 return p; | |
1210 } | |
1211 | |
1212 /* | |
7 | 1213 * Skip past regular expression. |
153 | 1214 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc). |
7 | 1215 * Take care of characters with a backslash in front of it. |
1216 * Skip strings inside [ and ]. | |
1217 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the | |
1218 * expression and change "\?" to "?". If "*newp" is not NULL the expression | |
1219 * is changed in-place. | |
1220 */ | |
1221 char_u * | |
1222 skip_regexp(startp, dirc, magic, newp) | |
1223 char_u *startp; | |
1224 int dirc; | |
1225 int magic; | |
1226 char_u **newp; | |
1227 { | |
1228 int mymagic; | |
1229 char_u *p = startp; | |
1230 | |
1231 if (magic) | |
1232 mymagic = MAGIC_ON; | |
1233 else | |
1234 mymagic = MAGIC_OFF; | |
1235 | |
39 | 1236 for (; p[0] != NUL; mb_ptr_adv(p)) |
7 | 1237 { |
1238 if (p[0] == dirc) /* found end of regexp */ | |
1239 break; | |
1240 if ((p[0] == '[' && mymagic >= MAGIC_ON) | |
1241 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF)) | |
1242 { | |
1243 p = skip_anyof(p + 1); | |
1244 if (p[0] == NUL) | |
1245 break; | |
1246 } | |
1247 else if (p[0] == '\\' && p[1] != NUL) | |
1248 { | |
1249 if (dirc == '?' && newp != NULL && p[1] == '?') | |
1250 { | |
1251 /* change "\?" to "?", make a copy first. */ | |
1252 if (*newp == NULL) | |
1253 { | |
1254 *newp = vim_strsave(startp); | |
1255 if (*newp != NULL) | |
1256 p = *newp + (p - startp); | |
1257 } | |
1258 if (*newp != NULL) | |
1621 | 1259 STRMOVE(p, p + 1); |
7 | 1260 else |
1261 ++p; | |
1262 } | |
1263 else | |
1264 ++p; /* skip next character */ | |
1265 if (*p == 'v') | |
1266 mymagic = MAGIC_ALL; | |
1267 else if (*p == 'V') | |
1268 mymagic = MAGIC_NONE; | |
1269 } | |
1270 } | |
1271 return p; | |
1272 } | |
1273 | |
4444 | 1274 static regprog_T *bt_regcomp __ARGS((char_u *expr, int re_flags)); |
1275 | |
7 | 1276 /* |
4444 | 1277 * bt_regcomp() - compile a regular expression into internal code for the |
1278 * traditional back track matcher. | |
41 | 1279 * Returns the program in allocated space. Returns NULL for an error. |
7 | 1280 * |
1281 * We can't allocate space until we know how big the compiled form will be, | |
1282 * but we can't compile it (and thus know how big it is) until we've got a | |
1283 * place to put the code. So we cheat: we compile it twice, once with code | |
1284 * generation turned off and size counting turned on, and once "for real". | |
1285 * This also means that we don't allocate space until we are sure that the | |
1286 * thing really will compile successfully, and we never have to move the | |
1287 * code and thus invalidate pointers into it. (Note that it has to be in | |
1288 * one piece because vim_free() must be able to free it all.) | |
1289 * | |
1290 * Whether upper/lower case is to be ignored is decided when executing the | |
1291 * program, it does not matter here. | |
1292 * | |
1293 * Beware that the optimization-preparation code in here knows about some | |
1294 * of the structure of the compiled regexp. | |
1295 * "re_flags": RE_MAGIC and/or RE_STRING. | |
1296 */ | |
4444 | 1297 static regprog_T * |
1298 bt_regcomp(expr, re_flags) | |
7 | 1299 char_u *expr; |
1300 int re_flags; | |
1301 { | |
4444 | 1302 bt_regprog_T *r; |
7 | 1303 char_u *scan; |
1304 char_u *longest; | |
1305 int len; | |
1306 int flags; | |
1307 | |
1308 if (expr == NULL) | |
1309 EMSG_RET_NULL(_(e_null)); | |
1310 | |
1311 init_class_tab(); | |
1312 | |
1313 /* | |
1314 * First pass: determine size, legality. | |
1315 */ | |
1316 regcomp_start(expr, re_flags); | |
1317 regcode = JUST_CALC_SIZE; | |
1318 regc(REGMAGIC); | |
1319 if (reg(REG_NOPAREN, &flags) == NULL) | |
1320 return NULL; | |
1321 | |
1322 /* Small enough for pointer-storage convention? */ | |
1323 #ifdef SMALL_MALLOC /* 16 bit storage allocation */ | |
1324 if (regsize >= 65536L - 256L) | |
1325 EMSG_RET_NULL(_("E339: Pattern too long")); | |
1326 #endif | |
1327 | |
1328 /* Allocate space. */ | |
4444 | 1329 r = (bt_regprog_T *)lalloc(sizeof(bt_regprog_T) + regsize, TRUE); |
7 | 1330 if (r == NULL) |
1331 return NULL; | |
1332 | |
1333 /* | |
1334 * Second pass: emit code. | |
1335 */ | |
1336 regcomp_start(expr, re_flags); | |
1337 regcode = r->program; | |
1338 regc(REGMAGIC); | |
2010 | 1339 if (reg(REG_NOPAREN, &flags) == NULL || reg_toolong) |
7 | 1340 { |
1341 vim_free(r); | |
2010 | 1342 if (reg_toolong) |
1343 EMSG_RET_NULL(_("E339: Pattern too long")); | |
7 | 1344 return NULL; |
1345 } | |
1346 | |
1347 /* Dig out information for optimizations. */ | |
1348 r->regstart = NUL; /* Worst-case defaults. */ | |
1349 r->reganch = 0; | |
1350 r->regmust = NULL; | |
1351 r->regmlen = 0; | |
1352 r->regflags = regflags; | |
1353 if (flags & HASNL) | |
1354 r->regflags |= RF_HASNL; | |
1355 if (flags & HASLOOKBH) | |
1356 r->regflags |= RF_LOOKBH; | |
1357 #ifdef FEAT_SYN_HL | |
1358 /* Remember whether this pattern has any \z specials in it. */ | |
1359 r->reghasz = re_has_z; | |
1360 #endif | |
1361 scan = r->program + 1; /* First BRANCH. */ | |
1362 if (OP(regnext(scan)) == END) /* Only one top-level choice. */ | |
1363 { | |
1364 scan = OPERAND(scan); | |
1365 | |
1366 /* Starting-point info. */ | |
1367 if (OP(scan) == BOL || OP(scan) == RE_BOF) | |
1368 { | |
1369 r->reganch++; | |
1370 scan = regnext(scan); | |
1371 } | |
1372 | |
1373 if (OP(scan) == EXACTLY) | |
1374 { | |
1375 #ifdef FEAT_MBYTE | |
1376 if (has_mbyte) | |
1377 r->regstart = (*mb_ptr2char)(OPERAND(scan)); | |
1378 else | |
1379 #endif | |
1380 r->regstart = *OPERAND(scan); | |
1381 } | |
1382 else if ((OP(scan) == BOW | |
1383 || OP(scan) == EOW | |
1384 || OP(scan) == NOTHING | |
1385 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN | |
1386 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE) | |
1387 && OP(regnext(scan)) == EXACTLY) | |
1388 { | |
1389 #ifdef FEAT_MBYTE | |
1390 if (has_mbyte) | |
1391 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan))); | |
1392 else | |
1393 #endif | |
1394 r->regstart = *OPERAND(regnext(scan)); | |
1395 } | |
1396 | |
1397 /* | |
1398 * If there's something expensive in the r.e., find the longest | |
1399 * literal string that must appear and make it the regmust. Resolve | |
1400 * ties in favor of later strings, since the regstart check works | |
1401 * with the beginning of the r.e. and avoiding duplication | |
1402 * strengthens checking. Not a strong reason, but sufficient in the | |
1403 * absence of others. | |
1404 */ | |
1405 /* | |
1406 * When the r.e. starts with BOW, it is faster to look for a regmust | |
1407 * first. Used a lot for "#" and "*" commands. (Added by mool). | |
1408 */ | |
1409 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW) | |
1410 && !(flags & HASNL)) | |
1411 { | |
1412 longest = NULL; | |
1413 len = 0; | |
1414 for (; scan != NULL; scan = regnext(scan)) | |
1415 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len) | |
1416 { | |
1417 longest = OPERAND(scan); | |
1418 len = (int)STRLEN(OPERAND(scan)); | |
1419 } | |
1420 r->regmust = longest; | |
1421 r->regmlen = len; | |
1422 } | |
1423 } | |
4444 | 1424 #ifdef BT_REGEXP_DUMP |
7 | 1425 regdump(expr, r); |
1426 #endif | |
4444 | 1427 r->engine = &bt_regengine; |
1428 return (regprog_T *)r; | |
7 | 1429 } |
1430 | |
1431 /* | |
1432 * Setup to parse the regexp. Used once to get the length and once to do it. | |
1433 */ | |
1434 static void | |
1435 regcomp_start(expr, re_flags) | |
1436 char_u *expr; | |
1437 int re_flags; /* see vim_regcomp() */ | |
1438 { | |
1439 initchr(expr); | |
1440 if (re_flags & RE_MAGIC) | |
1441 reg_magic = MAGIC_ON; | |
1442 else | |
1443 reg_magic = MAGIC_OFF; | |
1444 reg_string = (re_flags & RE_STRING); | |
481 | 1445 reg_strict = (re_flags & RE_STRICT); |
7 | 1446 |
1447 num_complex_braces = 0; | |
1448 regnpar = 1; | |
1449 vim_memset(had_endbrace, 0, sizeof(had_endbrace)); | |
1450 #ifdef FEAT_SYN_HL | |
1451 regnzpar = 1; | |
1452 re_has_z = 0; | |
1453 #endif | |
1454 regsize = 0L; | |
2010 | 1455 reg_toolong = FALSE; |
7 | 1456 regflags = 0; |
1457 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1458 had_eol = FALSE; | |
1459 #endif | |
1460 } | |
1461 | |
1462 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1463 /* | |
1464 * Check if during the previous call to vim_regcomp the EOL item "$" has been | |
1465 * found. This is messy, but it works fine. | |
1466 */ | |
1467 int | |
1468 vim_regcomp_had_eol() | |
1469 { | |
1470 return had_eol; | |
1471 } | |
1472 #endif | |
1473 | |
1474 /* | |
4444 | 1475 * Parse regular expression, i.e. main body or parenthesized thing. |
7 | 1476 * |
1477 * Caller must absorb opening parenthesis. | |
1478 * | |
1479 * Combining parenthesis handling with the base level of regular expression | |
1480 * is a trifle forced, but the need to tie the tails of the branches to what | |
1481 * follows makes it hard to avoid. | |
1482 */ | |
1483 static char_u * | |
1484 reg(paren, flagp) | |
1485 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */ | |
1486 int *flagp; | |
1487 { | |
1488 char_u *ret; | |
1489 char_u *br; | |
1490 char_u *ender; | |
1491 int parno = 0; | |
1492 int flags; | |
1493 | |
1494 *flagp = HASWIDTH; /* Tentatively. */ | |
1495 | |
1496 #ifdef FEAT_SYN_HL | |
1497 if (paren == REG_ZPAREN) | |
1498 { | |
1499 /* Make a ZOPEN node. */ | |
1500 if (regnzpar >= NSUBEXP) | |
1501 EMSG_RET_NULL(_("E50: Too many \\z(")); | |
1502 parno = regnzpar; | |
1503 regnzpar++; | |
1504 ret = regnode(ZOPEN + parno); | |
1505 } | |
1506 else | |
1507 #endif | |
1508 if (paren == REG_PAREN) | |
1509 { | |
1510 /* Make a MOPEN node. */ | |
1511 if (regnpar >= NSUBEXP) | |
4444 | 1512 EMSG2_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL); |
7 | 1513 parno = regnpar; |
1514 ++regnpar; | |
1515 ret = regnode(MOPEN + parno); | |
1516 } | |
1517 else if (paren == REG_NPAREN) | |
1518 { | |
1519 /* Make a NOPEN node. */ | |
1520 ret = regnode(NOPEN); | |
1521 } | |
1522 else | |
1523 ret = NULL; | |
1524 | |
1525 /* Pick up the branches, linking them together. */ | |
1526 br = regbranch(&flags); | |
1527 if (br == NULL) | |
1528 return NULL; | |
1529 if (ret != NULL) | |
1530 regtail(ret, br); /* [MZ]OPEN -> first. */ | |
1531 else | |
1532 ret = br; | |
1533 /* If one of the branches can be zero-width, the whole thing can. | |
1534 * If one of the branches has * at start or matches a line-break, the | |
1535 * whole thing can. */ | |
1536 if (!(flags & HASWIDTH)) | |
1537 *flagp &= ~HASWIDTH; | |
1538 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH); | |
1539 while (peekchr() == Magic('|')) | |
1540 { | |
1541 skipchr(); | |
1542 br = regbranch(&flags); | |
2010 | 1543 if (br == NULL || reg_toolong) |
7 | 1544 return NULL; |
1545 regtail(ret, br); /* BRANCH -> BRANCH. */ | |
1546 if (!(flags & HASWIDTH)) | |
1547 *flagp &= ~HASWIDTH; | |
1548 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH); | |
1549 } | |
1550 | |
1551 /* Make a closing node, and hook it on the end. */ | |
1552 ender = regnode( | |
1553 #ifdef FEAT_SYN_HL | |
1554 paren == REG_ZPAREN ? ZCLOSE + parno : | |
1555 #endif | |
1556 paren == REG_PAREN ? MCLOSE + parno : | |
1557 paren == REG_NPAREN ? NCLOSE : END); | |
1558 regtail(ret, ender); | |
1559 | |
1560 /* Hook the tails of the branches to the closing node. */ | |
1561 for (br = ret; br != NULL; br = regnext(br)) | |
1562 regoptail(br, ender); | |
1563 | |
1564 /* Check for proper termination. */ | |
1565 if (paren != REG_NOPAREN && getchr() != Magic(')')) | |
1566 { | |
1567 #ifdef FEAT_SYN_HL | |
1568 if (paren == REG_ZPAREN) | |
308 | 1569 EMSG_RET_NULL(_("E52: Unmatched \\z(")); |
7 | 1570 else |
1571 #endif | |
1572 if (paren == REG_NPAREN) | |
4444 | 1573 EMSG2_RET_NULL(_(e_unmatchedpp), reg_magic == MAGIC_ALL); |
7 | 1574 else |
4444 | 1575 EMSG2_RET_NULL(_(e_unmatchedp), reg_magic == MAGIC_ALL); |
7 | 1576 } |
1577 else if (paren == REG_NOPAREN && peekchr() != NUL) | |
1578 { | |
1579 if (curchr == Magic(')')) | |
4444 | 1580 EMSG2_RET_NULL(_(e_unmatchedpar), reg_magic == MAGIC_ALL); |
7 | 1581 else |
308 | 1582 EMSG_RET_NULL(_(e_trailing)); /* "Can't happen". */ |
7 | 1583 /* NOTREACHED */ |
1584 } | |
1585 /* | |
1586 * Here we set the flag allowing back references to this set of | |
1587 * parentheses. | |
1588 */ | |
1589 if (paren == REG_PAREN) | |
1590 had_endbrace[parno] = TRUE; /* have seen the close paren */ | |
1591 return ret; | |
1592 } | |
1593 | |
1594 /* | |
4444 | 1595 * Parse one alternative of an | operator. |
7 | 1596 * Implements the & operator. |
1597 */ | |
1598 static char_u * | |
1599 regbranch(flagp) | |
1600 int *flagp; | |
1601 { | |
1602 char_u *ret; | |
1603 char_u *chain = NULL; | |
1604 char_u *latest; | |
1605 int flags; | |
1606 | |
1607 *flagp = WORST | HASNL; /* Tentatively. */ | |
1608 | |
1609 ret = regnode(BRANCH); | |
1610 for (;;) | |
1611 { | |
1612 latest = regconcat(&flags); | |
1613 if (latest == NULL) | |
1614 return NULL; | |
1615 /* If one of the branches has width, the whole thing has. If one of | |
1616 * the branches anchors at start-of-line, the whole thing does. | |
1617 * If one of the branches uses look-behind, the whole thing does. */ | |
1618 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH); | |
1619 /* If one of the branches doesn't match a line-break, the whole thing | |
1620 * doesn't. */ | |
1621 *flagp &= ~HASNL | (flags & HASNL); | |
1622 if (chain != NULL) | |
1623 regtail(chain, latest); | |
1624 if (peekchr() != Magic('&')) | |
1625 break; | |
1626 skipchr(); | |
1627 regtail(latest, regnode(END)); /* operand ends */ | |
2010 | 1628 if (reg_toolong) |
1629 break; | |
7 | 1630 reginsert(MATCH, latest); |
1631 chain = latest; | |
1632 } | |
1633 | |
1634 return ret; | |
1635 } | |
1636 | |
1637 /* | |
4444 | 1638 * Parse one alternative of an | or & operator. |
7 | 1639 * Implements the concatenation operator. |
1640 */ | |
1641 static char_u * | |
1642 regconcat(flagp) | |
1643 int *flagp; | |
1644 { | |
1645 char_u *first = NULL; | |
1646 char_u *chain = NULL; | |
1647 char_u *latest; | |
1648 int flags; | |
1649 int cont = TRUE; | |
1650 | |
1651 *flagp = WORST; /* Tentatively. */ | |
1652 | |
1653 while (cont) | |
1654 { | |
1655 switch (peekchr()) | |
1656 { | |
1657 case NUL: | |
1658 case Magic('|'): | |
1659 case Magic('&'): | |
1660 case Magic(')'): | |
1661 cont = FALSE; | |
1662 break; | |
1663 case Magic('Z'): | |
1664 #ifdef FEAT_MBYTE | |
1665 regflags |= RF_ICOMBINE; | |
1666 #endif | |
1667 skipchr_keepstart(); | |
1668 break; | |
1669 case Magic('c'): | |
1670 regflags |= RF_ICASE; | |
1671 skipchr_keepstart(); | |
1672 break; | |
1673 case Magic('C'): | |
1674 regflags |= RF_NOICASE; | |
1675 skipchr_keepstart(); | |
1676 break; | |
1677 case Magic('v'): | |
1678 reg_magic = MAGIC_ALL; | |
1679 skipchr_keepstart(); | |
1680 curchr = -1; | |
1681 break; | |
1682 case Magic('m'): | |
1683 reg_magic = MAGIC_ON; | |
1684 skipchr_keepstart(); | |
1685 curchr = -1; | |
1686 break; | |
1687 case Magic('M'): | |
1688 reg_magic = MAGIC_OFF; | |
1689 skipchr_keepstart(); | |
1690 curchr = -1; | |
1691 break; | |
1692 case Magic('V'): | |
1693 reg_magic = MAGIC_NONE; | |
1694 skipchr_keepstart(); | |
1695 curchr = -1; | |
1696 break; | |
1697 default: | |
1698 latest = regpiece(&flags); | |
2010 | 1699 if (latest == NULL || reg_toolong) |
7 | 1700 return NULL; |
1701 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH); | |
1702 if (chain == NULL) /* First piece. */ | |
1703 *flagp |= flags & SPSTART; | |
1704 else | |
1705 regtail(chain, latest); | |
1706 chain = latest; | |
1707 if (first == NULL) | |
1708 first = latest; | |
1709 break; | |
1710 } | |
1711 } | |
1712 if (first == NULL) /* Loop ran zero times. */ | |
1713 first = regnode(NOTHING); | |
1714 return first; | |
1715 } | |
1716 | |
1717 /* | |
4444 | 1718 * Parse something followed by possible [*+=]. |
7 | 1719 * |
1720 * Note that the branching code sequences used for = and the general cases | |
1721 * of * and + are somewhat optimized: they use the same NOTHING node as | |
1722 * both the endmarker for their branch list and the body of the last branch. | |
1723 * It might seem that this node could be dispensed with entirely, but the | |
1724 * endmarker role is not redundant. | |
1725 */ | |
1726 static char_u * | |
1727 regpiece(flagp) | |
1728 int *flagp; | |
1729 { | |
1730 char_u *ret; | |
1731 int op; | |
1732 char_u *next; | |
1733 int flags; | |
1734 long minval; | |
1735 long maxval; | |
1736 | |
1737 ret = regatom(&flags); | |
1738 if (ret == NULL) | |
1739 return NULL; | |
1740 | |
1741 op = peekchr(); | |
1742 if (re_multi_type(op) == NOT_MULTI) | |
1743 { | |
1744 *flagp = flags; | |
1745 return ret; | |
1746 } | |
1747 /* default flags */ | |
1748 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH))); | |
1749 | |
1750 skipchr(); | |
1751 switch (op) | |
1752 { | |
1753 case Magic('*'): | |
1754 if (flags & SIMPLE) | |
1755 reginsert(STAR, ret); | |
1756 else | |
1757 { | |
1758 /* Emit x* as (x&|), where & means "self". */ | |
1759 reginsert(BRANCH, ret); /* Either x */ | |
1760 regoptail(ret, regnode(BACK)); /* and loop */ | |
1761 regoptail(ret, ret); /* back */ | |
1762 regtail(ret, regnode(BRANCH)); /* or */ | |
1763 regtail(ret, regnode(NOTHING)); /* null. */ | |
1764 } | |
1765 break; | |
1766 | |
1767 case Magic('+'): | |
1768 if (flags & SIMPLE) | |
1769 reginsert(PLUS, ret); | |
1770 else | |
1771 { | |
1772 /* Emit x+ as x(&|), where & means "self". */ | |
1773 next = regnode(BRANCH); /* Either */ | |
1774 regtail(ret, next); | |
233 | 1775 regtail(regnode(BACK), ret); /* loop back */ |
7 | 1776 regtail(next, regnode(BRANCH)); /* or */ |
1777 regtail(ret, regnode(NOTHING)); /* null. */ | |
1778 } | |
1779 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH))); | |
1780 break; | |
1781 | |
1782 case Magic('@'): | |
1783 { | |
1784 int lop = END; | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1785 int nr; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1786 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1787 nr = getdecchrs(); |
7 | 1788 switch (no_Magic(getchr())) |
1789 { | |
1790 case '=': lop = MATCH; break; /* \@= */ | |
1791 case '!': lop = NOMATCH; break; /* \@! */ | |
1792 case '>': lop = SUBPAT; break; /* \@> */ | |
1793 case '<': switch (no_Magic(getchr())) | |
1794 { | |
1795 case '=': lop = BEHIND; break; /* \@<= */ | |
1796 case '!': lop = NOBEHIND; break; /* \@<! */ | |
1797 } | |
1798 } | |
1799 if (lop == END) | |
4444 | 1800 EMSG2_RET_NULL(_("E59: invalid character after %s@"), |
7 | 1801 reg_magic == MAGIC_ALL); |
1802 /* Look behind must match with behind_pos. */ | |
1803 if (lop == BEHIND || lop == NOBEHIND) | |
1804 { | |
1805 regtail(ret, regnode(BHPOS)); | |
1806 *flagp |= HASLOOKBH; | |
1807 } | |
1808 regtail(ret, regnode(END)); /* operand ends */ | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1809 if (lop == BEHIND || lop == NOBEHIND) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1810 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1811 if (nr < 0) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1812 nr = 0; /* no limit is same as zero limit */ |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1813 reginsert_nr(lop, nr, ret); |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1814 } |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1815 else |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1816 reginsert(lop, ret); |
7 | 1817 break; |
1818 } | |
1819 | |
1820 case Magic('?'): | |
1821 case Magic('='): | |
1822 /* Emit x= as (x|) */ | |
1823 reginsert(BRANCH, ret); /* Either x */ | |
1824 regtail(ret, regnode(BRANCH)); /* or */ | |
1825 next = regnode(NOTHING); /* null. */ | |
1826 regtail(ret, next); | |
1827 regoptail(ret, next); | |
1828 break; | |
1829 | |
1830 case Magic('{'): | |
1831 if (!read_limits(&minval, &maxval)) | |
1832 return NULL; | |
1833 if (flags & SIMPLE) | |
1834 { | |
1835 reginsert(BRACE_SIMPLE, ret); | |
1836 reginsert_limits(BRACE_LIMITS, minval, maxval, ret); | |
1837 } | |
1838 else | |
1839 { | |
1840 if (num_complex_braces >= 10) | |
4444 | 1841 EMSG2_RET_NULL(_("E60: Too many complex %s{...}s"), |
7 | 1842 reg_magic == MAGIC_ALL); |
1843 reginsert(BRACE_COMPLEX + num_complex_braces, ret); | |
1844 regoptail(ret, regnode(BACK)); | |
1845 regoptail(ret, ret); | |
1846 reginsert_limits(BRACE_LIMITS, minval, maxval, ret); | |
1847 ++num_complex_braces; | |
1848 } | |
1849 if (minval > 0 && maxval > 0) | |
1850 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH))); | |
1851 break; | |
1852 } | |
1853 if (re_multi_type(peekchr()) != NOT_MULTI) | |
1854 { | |
1855 /* Can't have a multi follow a multi. */ | |
1856 if (peekchr() == Magic('*')) | |
1857 sprintf((char *)IObuff, _("E61: Nested %s*"), | |
1858 reg_magic >= MAGIC_ON ? "" : "\\"); | |
1859 else | |
1860 sprintf((char *)IObuff, _("E62: Nested %s%c"), | |
1861 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr())); | |
1862 EMSG_RET_NULL(IObuff); | |
1863 } | |
1864 | |
1865 return ret; | |
1866 } | |
1867 | |
4444 | 1868 /* When making changes to classchars also change nfa_classcodes. */ |
1869 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU"; | |
1870 static int classcodes[] = { | |
1871 ANY, IDENT, SIDENT, KWORD, SKWORD, | |
1872 FNAME, SFNAME, PRINT, SPRINT, | |
1873 WHITE, NWHITE, DIGIT, NDIGIT, | |
1874 HEX, NHEX, OCTAL, NOCTAL, | |
1875 WORD, NWORD, HEAD, NHEAD, | |
1876 ALPHA, NALPHA, LOWER, NLOWER, | |
1877 UPPER, NUPPER | |
1878 }; | |
1879 | |
7 | 1880 /* |
4444 | 1881 * Parse the lowest level. |
7 | 1882 * |
1883 * Optimization: gobbles an entire sequence of ordinary characters so that | |
1884 * it can turn them into a single node, which is smaller to store and | |
1885 * faster to run. Don't do this when one_exactly is set. | |
1886 */ | |
1887 static char_u * | |
1888 regatom(flagp) | |
1889 int *flagp; | |
1890 { | |
1891 char_u *ret; | |
1892 int flags; | |
1893 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
167 | 1894 int cpo_bsl; /* 'cpoptions' contains '\' flag */ |
7 | 1895 int c; |
1896 char_u *p; | |
1897 int extra = 0; | |
1898 | |
1899 *flagp = WORST; /* Tentatively. */ | |
410 | 1900 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; |
1901 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
7 | 1902 |
1903 c = getchr(); | |
1904 switch (c) | |
1905 { | |
1906 case Magic('^'): | |
1907 ret = regnode(BOL); | |
1908 break; | |
1909 | |
1910 case Magic('$'): | |
1911 ret = regnode(EOL); | |
1912 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1913 had_eol = TRUE; | |
1914 #endif | |
1915 break; | |
1916 | |
1917 case Magic('<'): | |
1918 ret = regnode(BOW); | |
1919 break; | |
1920 | |
1921 case Magic('>'): | |
1922 ret = regnode(EOW); | |
1923 break; | |
1924 | |
1925 case Magic('_'): | |
1926 c = no_Magic(getchr()); | |
1927 if (c == '^') /* "\_^" is start-of-line */ | |
1928 { | |
1929 ret = regnode(BOL); | |
1930 break; | |
1931 } | |
1932 if (c == '$') /* "\_$" is end-of-line */ | |
1933 { | |
1934 ret = regnode(EOL); | |
1935 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1936 had_eol = TRUE; | |
1937 #endif | |
1938 break; | |
1939 } | |
1940 | |
1941 extra = ADD_NL; | |
1942 *flagp |= HASNL; | |
1943 | |
1944 /* "\_[" is character range plus newline */ | |
1945 if (c == '[') | |
1946 goto collection; | |
1947 | |
1948 /* "\_x" is character class plus newline */ | |
1949 /*FALLTHROUGH*/ | |
1950 | |
1951 /* | |
1952 * Character classes. | |
1953 */ | |
1954 case Magic('.'): | |
1955 case Magic('i'): | |
1956 case Magic('I'): | |
1957 case Magic('k'): | |
1958 case Magic('K'): | |
1959 case Magic('f'): | |
1960 case Magic('F'): | |
1961 case Magic('p'): | |
1962 case Magic('P'): | |
1963 case Magic('s'): | |
1964 case Magic('S'): | |
1965 case Magic('d'): | |
1966 case Magic('D'): | |
1967 case Magic('x'): | |
1968 case Magic('X'): | |
1969 case Magic('o'): | |
1970 case Magic('O'): | |
1971 case Magic('w'): | |
1972 case Magic('W'): | |
1973 case Magic('h'): | |
1974 case Magic('H'): | |
1975 case Magic('a'): | |
1976 case Magic('A'): | |
1977 case Magic('l'): | |
1978 case Magic('L'): | |
1979 case Magic('u'): | |
1980 case Magic('U'): | |
1981 p = vim_strchr(classchars, no_Magic(c)); | |
1982 if (p == NULL) | |
1983 EMSG_RET_NULL(_("E63: invalid use of \\_")); | |
714 | 1984 #ifdef FEAT_MBYTE |
1985 /* When '.' is followed by a composing char ignore the dot, so that | |
1986 * the composing char is matched here. */ | |
1987 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr())) | |
1988 { | |
1989 c = getchr(); | |
1990 goto do_multibyte; | |
1991 } | |
1992 #endif | |
7 | 1993 ret = regnode(classcodes[p - classchars] + extra); |
1994 *flagp |= HASWIDTH | SIMPLE; | |
1995 break; | |
1996 | |
1997 case Magic('n'): | |
1998 if (reg_string) | |
1999 { | |
2000 /* In a string "\n" matches a newline character. */ | |
2001 ret = regnode(EXACTLY); | |
2002 regc(NL); | |
2003 regc(NUL); | |
2004 *flagp |= HASWIDTH | SIMPLE; | |
2005 } | |
2006 else | |
2007 { | |
2008 /* In buffer text "\n" matches the end of a line. */ | |
2009 ret = regnode(NEWL); | |
2010 *flagp |= HASWIDTH | HASNL; | |
2011 } | |
2012 break; | |
2013 | |
2014 case Magic('('): | |
2015 if (one_exactly) | |
2016 EMSG_ONE_RET_NULL; | |
2017 ret = reg(REG_PAREN, &flags); | |
2018 if (ret == NULL) | |
2019 return NULL; | |
2020 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH); | |
2021 break; | |
2022 | |
2023 case NUL: | |
2024 case Magic('|'): | |
2025 case Magic('&'): | |
2026 case Magic(')'): | |
1468 | 2027 if (one_exactly) |
2028 EMSG_ONE_RET_NULL; | |
7 | 2029 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */ |
2030 /* NOTREACHED */ | |
2031 | |
2032 case Magic('='): | |
2033 case Magic('?'): | |
2034 case Magic('+'): | |
2035 case Magic('@'): | |
2036 case Magic('{'): | |
2037 case Magic('*'): | |
2038 c = no_Magic(c); | |
2039 sprintf((char *)IObuff, _("E64: %s%c follows nothing"), | |
2040 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL) | |
2041 ? "" : "\\", c); | |
2042 EMSG_RET_NULL(IObuff); | |
2043 /* NOTREACHED */ | |
2044 | |
2045 case Magic('~'): /* previous substitute pattern */ | |
359 | 2046 if (reg_prev_sub != NULL) |
7 | 2047 { |
2048 char_u *lp; | |
2049 | |
2050 ret = regnode(EXACTLY); | |
2051 lp = reg_prev_sub; | |
2052 while (*lp != NUL) | |
2053 regc(*lp++); | |
2054 regc(NUL); | |
2055 if (*reg_prev_sub != NUL) | |
2056 { | |
2057 *flagp |= HASWIDTH; | |
2058 if ((lp - reg_prev_sub) == 1) | |
2059 *flagp |= SIMPLE; | |
2060 } | |
2061 } | |
2062 else | |
2063 EMSG_RET_NULL(_(e_nopresub)); | |
2064 break; | |
2065 | |
2066 case Magic('1'): | |
2067 case Magic('2'): | |
2068 case Magic('3'): | |
2069 case Magic('4'): | |
2070 case Magic('5'): | |
2071 case Magic('6'): | |
2072 case Magic('7'): | |
2073 case Magic('8'): | |
2074 case Magic('9'): | |
2075 { | |
2076 int refnum; | |
2077 | |
2078 refnum = c - Magic('0'); | |
2079 /* | |
2080 * Check if the back reference is legal. We must have seen the | |
2081 * close brace. | |
2082 * TODO: Should also check that we don't refer to something | |
2083 * that is repeated (+*=): what instance of the repetition | |
2084 * should we match? | |
2085 */ | |
2086 if (!had_endbrace[refnum]) | |
2087 { | |
2088 /* Trick: check if "@<=" or "@<!" follows, in which case | |
2089 * the \1 can appear before the referenced match. */ | |
2090 for (p = regparse; *p != NUL; ++p) | |
2091 if (p[0] == '@' && p[1] == '<' | |
2092 && (p[2] == '!' || p[2] == '=')) | |
2093 break; | |
2094 if (*p == NUL) | |
2095 EMSG_RET_NULL(_("E65: Illegal back reference")); | |
2096 } | |
2097 ret = regnode(BACKREF + refnum); | |
2098 } | |
2099 break; | |
2100 | |
2101 case Magic('z'): | |
2102 { | |
2103 c = no_Magic(getchr()); | |
2104 switch (c) | |
2105 { | |
741 | 2106 #ifdef FEAT_SYN_HL |
7 | 2107 case '(': if (reg_do_extmatch != REX_SET) |
2108 EMSG_RET_NULL(_("E66: \\z( not allowed here")); | |
2109 if (one_exactly) | |
2110 EMSG_ONE_RET_NULL; | |
2111 ret = reg(REG_ZPAREN, &flags); | |
2112 if (ret == NULL) | |
2113 return NULL; | |
2114 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH); | |
2115 re_has_z = REX_SET; | |
2116 break; | |
2117 | |
2118 case '1': | |
2119 case '2': | |
2120 case '3': | |
2121 case '4': | |
2122 case '5': | |
2123 case '6': | |
2124 case '7': | |
2125 case '8': | |
2126 case '9': if (reg_do_extmatch != REX_USE) | |
2127 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here")); | |
2128 ret = regnode(ZREF + c - '0'); | |
2129 re_has_z = REX_USE; | |
2130 break; | |
741 | 2131 #endif |
7 | 2132 |
2133 case 's': ret = regnode(MOPEN + 0); | |
2134 break; | |
2135 | |
2136 case 'e': ret = regnode(MCLOSE + 0); | |
2137 break; | |
2138 | |
2139 default: EMSG_RET_NULL(_("E68: Invalid character after \\z")); | |
2140 } | |
2141 } | |
2142 break; | |
2143 | |
2144 case Magic('%'): | |
2145 { | |
2146 c = no_Magic(getchr()); | |
2147 switch (c) | |
2148 { | |
2149 /* () without a back reference */ | |
2150 case '(': | |
2151 if (one_exactly) | |
2152 EMSG_ONE_RET_NULL; | |
2153 ret = reg(REG_NPAREN, &flags); | |
2154 if (ret == NULL) | |
2155 return NULL; | |
2156 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH); | |
2157 break; | |
2158 | |
2159 /* Catch \%^ and \%$ regardless of where they appear in the | |
2160 * pattern -- regardless of whether or not it makes sense. */ | |
2161 case '^': | |
2162 ret = regnode(RE_BOF); | |
2163 break; | |
2164 | |
2165 case '$': | |
2166 ret = regnode(RE_EOF); | |
2167 break; | |
2168 | |
2169 case '#': | |
2170 ret = regnode(CURSOR); | |
2171 break; | |
2172 | |
639 | 2173 case 'V': |
2174 ret = regnode(RE_VISUAL); | |
2175 break; | |
2176 | |
7 | 2177 /* \%[abc]: Emit as a list of branches, all ending at the last |
2178 * branch which matches nothing. */ | |
2179 case '[': | |
2180 if (one_exactly) /* doesn't nest */ | |
2181 EMSG_ONE_RET_NULL; | |
2182 { | |
2183 char_u *lastbranch; | |
2184 char_u *lastnode = NULL; | |
2185 char_u *br; | |
2186 | |
2187 ret = NULL; | |
2188 while ((c = getchr()) != ']') | |
2189 { | |
2190 if (c == NUL) | |
4444 | 2191 EMSG2_RET_NULL(_("E69: Missing ] after %s%%["), |
7 | 2192 reg_magic == MAGIC_ALL); |
2193 br = regnode(BRANCH); | |
2194 if (ret == NULL) | |
2195 ret = br; | |
2196 else | |
2197 regtail(lastnode, br); | |
2198 | |
2199 ungetchr(); | |
2200 one_exactly = TRUE; | |
2201 lastnode = regatom(flagp); | |
2202 one_exactly = FALSE; | |
2203 if (lastnode == NULL) | |
2204 return NULL; | |
2205 } | |
2206 if (ret == NULL) | |
4444 | 2207 EMSG2_RET_NULL(_("E70: Empty %s%%[]"), |
7 | 2208 reg_magic == MAGIC_ALL); |
2209 lastbranch = regnode(BRANCH); | |
2210 br = regnode(NOTHING); | |
2211 if (ret != JUST_CALC_SIZE) | |
2212 { | |
2213 regtail(lastnode, br); | |
2214 regtail(lastbranch, br); | |
2215 /* connect all branches to the NOTHING | |
2216 * branch at the end */ | |
2217 for (br = ret; br != lastnode; ) | |
2218 { | |
2219 if (OP(br) == BRANCH) | |
2220 { | |
2221 regtail(br, lastbranch); | |
2222 br = OPERAND(br); | |
2223 } | |
2224 else | |
2225 br = regnext(br); | |
2226 } | |
2227 } | |
1701 | 2228 *flagp &= ~(HASWIDTH | SIMPLE); |
7 | 2229 break; |
2230 } | |
2231 | |
24 | 2232 case 'd': /* %d123 decimal */ |
2233 case 'o': /* %o123 octal */ | |
2234 case 'x': /* %xab hex 2 */ | |
2235 case 'u': /* %uabcd hex 4 */ | |
2236 case 'U': /* %U1234abcd hex 8 */ | |
2237 { | |
2238 int i; | |
2239 | |
2240 switch (c) | |
2241 { | |
2242 case 'd': i = getdecchrs(); break; | |
2243 case 'o': i = getoctchrs(); break; | |
2244 case 'x': i = gethexchrs(2); break; | |
2245 case 'u': i = gethexchrs(4); break; | |
2246 case 'U': i = gethexchrs(8); break; | |
2247 default: i = -1; break; | |
2248 } | |
2249 | |
2250 if (i < 0) | |
4444 | 2251 EMSG2_RET_NULL( |
24 | 2252 _("E678: Invalid character after %s%%[dxouU]"), |
2253 reg_magic == MAGIC_ALL); | |
714 | 2254 #ifdef FEAT_MBYTE |
2255 if (use_multibytecode(i)) | |
2256 ret = regnode(MULTIBYTECODE); | |
2257 else | |
2258 #endif | |
2259 ret = regnode(EXACTLY); | |
24 | 2260 if (i == 0) |
2261 regc(0x0a); | |
2262 else | |
2263 #ifdef FEAT_MBYTE | |
2264 regmbc(i); | |
2265 #else | |
2266 regc(i); | |
2267 #endif | |
2268 regc(NUL); | |
2269 *flagp |= HASWIDTH; | |
2270 break; | |
2271 } | |
2272 | |
7 | 2273 default: |
639 | 2274 if (VIM_ISDIGIT(c) || c == '<' || c == '>' |
2275 || c == '\'') | |
7 | 2276 { |
2277 long_u n = 0; | |
2278 int cmp; | |
2279 | |
2280 cmp = c; | |
2281 if (cmp == '<' || cmp == '>') | |
2282 c = getchr(); | |
2283 while (VIM_ISDIGIT(c)) | |
2284 { | |
2285 n = n * 10 + (c - '0'); | |
2286 c = getchr(); | |
2287 } | |
639 | 2288 if (c == '\'' && n == 0) |
2289 { | |
2290 /* "\%'m", "\%<'m" and "\%>'m": Mark */ | |
2291 c = getchr(); | |
2292 ret = regnode(RE_MARK); | |
2293 if (ret == JUST_CALC_SIZE) | |
2294 regsize += 2; | |
2295 else | |
2296 { | |
2297 *regcode++ = c; | |
2298 *regcode++ = cmp; | |
2299 } | |
2300 break; | |
2301 } | |
2302 else if (c == 'l' || c == 'c' || c == 'v') | |
7 | 2303 { |
2304 if (c == 'l') | |
2305 ret = regnode(RE_LNUM); | |
2306 else if (c == 'c') | |
2307 ret = regnode(RE_COL); | |
2308 else | |
2309 ret = regnode(RE_VCOL); | |
2310 if (ret == JUST_CALC_SIZE) | |
2311 regsize += 5; | |
2312 else | |
2313 { | |
2314 /* put the number and the optional | |
2315 * comparator after the opcode */ | |
2316 regcode = re_put_long(regcode, n); | |
2317 *regcode++ = cmp; | |
2318 } | |
2319 break; | |
2320 } | |
2321 } | |
2322 | |
4444 | 2323 EMSG2_RET_NULL(_("E71: Invalid character after %s%%"), |
7 | 2324 reg_magic == MAGIC_ALL); |
2325 } | |
2326 } | |
2327 break; | |
2328 | |
2329 case Magic('['): | |
2330 collection: | |
2331 { | |
2332 char_u *lp; | |
2333 | |
2334 /* | |
2335 * If there is no matching ']', we assume the '[' is a normal | |
2336 * character. This makes 'incsearch' and ":help [" work. | |
2337 */ | |
2338 lp = skip_anyof(regparse); | |
2339 if (*lp == ']') /* there is a matching ']' */ | |
2340 { | |
2341 int startc = -1; /* > 0 when next '-' is a range */ | |
2342 int endc; | |
2343 | |
2344 /* | |
2345 * In a character class, different parsing rules apply. | |
2346 * Not even \ is special anymore, nothing is. | |
2347 */ | |
2348 if (*regparse == '^') /* Complement of range. */ | |
2349 { | |
2350 ret = regnode(ANYBUT + extra); | |
2351 regparse++; | |
2352 } | |
2353 else | |
2354 ret = regnode(ANYOF + extra); | |
2355 | |
2356 /* At the start ']' and '-' mean the literal character. */ | |
2357 if (*regparse == ']' || *regparse == '-') | |
167 | 2358 { |
2359 startc = *regparse; | |
7 | 2360 regc(*regparse++); |
167 | 2361 } |
7 | 2362 |
2363 while (*regparse != NUL && *regparse != ']') | |
2364 { | |
2365 if (*regparse == '-') | |
2366 { | |
2367 ++regparse; | |
2368 /* The '-' is not used for a range at the end and | |
2369 * after or before a '\n'. */ | |
2370 if (*regparse == ']' || *regparse == NUL | |
2371 || startc == -1 | |
2372 || (regparse[0] == '\\' && regparse[1] == 'n')) | |
2373 { | |
2374 regc('-'); | |
2375 startc = '-'; /* [--x] is a range */ | |
2376 } | |
2377 else | |
2378 { | |
167 | 2379 /* Also accept "a-[.z.]" */ |
2380 endc = 0; | |
2381 if (*regparse == '[') | |
2382 endc = get_coll_element(®parse); | |
2383 if (endc == 0) | |
2384 { | |
7 | 2385 #ifdef FEAT_MBYTE |
167 | 2386 if (has_mbyte) |
2387 endc = mb_ptr2char_adv(®parse); | |
2388 else | |
7 | 2389 #endif |
167 | 2390 endc = *regparse++; |
2391 } | |
24 | 2392 |
2393 /* Handle \o40, \x20 and \u20AC style sequences */ | |
167 | 2394 if (endc == '\\' && !cpo_lit && !cpo_bsl) |
24 | 2395 endc = coll_get_char(); |
2396 | |
7 | 2397 if (startc > endc) |
2398 EMSG_RET_NULL(_(e_invrange)); | |
2399 #ifdef FEAT_MBYTE | |
2400 if (has_mbyte && ((*mb_char2len)(startc) > 1 | |
2401 || (*mb_char2len)(endc) > 1)) | |
2402 { | |
2403 /* Limit to a range of 256 chars */ | |
2404 if (endc > startc + 256) | |
2405 EMSG_RET_NULL(_(e_invrange)); | |
2406 while (++startc <= endc) | |
2407 regmbc(startc); | |
2408 } | |
2409 else | |
2410 #endif | |
2411 { | |
2412 #ifdef EBCDIC | |
2413 int alpha_only = FALSE; | |
2414 | |
2415 /* for alphabetical range skip the gaps | |
2416 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */ | |
2417 if (isalpha(startc) && isalpha(endc)) | |
2418 alpha_only = TRUE; | |
2419 #endif | |
2420 while (++startc <= endc) | |
2421 #ifdef EBCDIC | |
2422 if (!alpha_only || isalpha(startc)) | |
2423 #endif | |
2424 regc(startc); | |
2425 } | |
2426 startc = -1; | |
2427 } | |
2428 } | |
2429 /* | |
2430 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim | |
2431 * accepts "\t", "\e", etc., but only when the 'l' flag in | |
2432 * 'cpoptions' is not included. | |
167 | 2433 * Posix doesn't recognize backslash at all. |
7 | 2434 */ |
2435 else if (*regparse == '\\' | |
167 | 2436 && !cpo_bsl |
7 | 2437 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL |
2438 || (!cpo_lit | |
2439 && vim_strchr(REGEXP_ABBR, | |
2440 regparse[1]) != NULL))) | |
2441 { | |
2442 regparse++; | |
2443 if (*regparse == 'n') | |
2444 { | |
2445 /* '\n' in range: also match NL */ | |
2446 if (ret != JUST_CALC_SIZE) | |
2447 { | |
4084 | 2448 /* Using \n inside [^] does not change what |
2449 * matches. "[^\n]" is the same as ".". */ | |
2450 if (*ret == ANYOF) | |
2451 { | |
7 | 2452 *ret = ANYOF + ADD_NL; |
4084 | 2453 *flagp |= HASNL; |
2454 } | |
7 | 2455 /* else: must have had a \n already */ |
2456 } | |
2457 regparse++; | |
2458 startc = -1; | |
2459 } | |
24 | 2460 else if (*regparse == 'd' |
2461 || *regparse == 'o' | |
2462 || *regparse == 'x' | |
2463 || *regparse == 'u' | |
2464 || *regparse == 'U') | |
2465 { | |
2466 startc = coll_get_char(); | |
2467 if (startc == 0) | |
2468 regc(0x0a); | |
2469 else | |
2470 #ifdef FEAT_MBYTE | |
2471 regmbc(startc); | |
2472 #else | |
2473 regc(startc); | |
2474 #endif | |
2475 } | |
7 | 2476 else |
2477 { | |
2478 startc = backslash_trans(*regparse++); | |
2479 regc(startc); | |
2480 } | |
2481 } | |
2482 else if (*regparse == '[') | |
2483 { | |
2484 int c_class; | |
2485 int cu; | |
2486 | |
167 | 2487 c_class = get_char_class(®parse); |
7 | 2488 startc = -1; |
2489 /* Characters assumed to be 8 bits! */ | |
2490 switch (c_class) | |
2491 { | |
2492 case CLASS_NONE: | |
167 | 2493 c_class = get_equi_class(®parse); |
2494 if (c_class != 0) | |
2495 { | |
2496 /* produce equivalence class */ | |
2497 reg_equi_class(c_class); | |
2498 } | |
2499 else if ((c_class = | |
2500 get_coll_element(®parse)) != 0) | |
2501 { | |
2502 /* produce a collating element */ | |
2503 regmbc(c_class); | |
2504 } | |
2505 else | |
2506 { | |
2507 /* literal '[', allow [[-x] as a range */ | |
2508 startc = *regparse++; | |
2509 regc(startc); | |
2510 } | |
7 | 2511 break; |
2512 case CLASS_ALNUM: | |
2513 for (cu = 1; cu <= 255; cu++) | |
2514 if (isalnum(cu)) | |
2515 regc(cu); | |
2516 break; | |
2517 case CLASS_ALPHA: | |
2518 for (cu = 1; cu <= 255; cu++) | |
2519 if (isalpha(cu)) | |
2520 regc(cu); | |
2521 break; | |
2522 case CLASS_BLANK: | |
2523 regc(' '); | |
2524 regc('\t'); | |
2525 break; | |
2526 case CLASS_CNTRL: | |
2527 for (cu = 1; cu <= 255; cu++) | |
2528 if (iscntrl(cu)) | |
2529 regc(cu); | |
2530 break; | |
2531 case CLASS_DIGIT: | |
2532 for (cu = 1; cu <= 255; cu++) | |
2533 if (VIM_ISDIGIT(cu)) | |
2534 regc(cu); | |
2535 break; | |
2536 case CLASS_GRAPH: | |
2537 for (cu = 1; cu <= 255; cu++) | |
2538 if (isgraph(cu)) | |
2539 regc(cu); | |
2540 break; | |
2541 case CLASS_LOWER: | |
2542 for (cu = 1; cu <= 255; cu++) | |
1347 | 2543 if (MB_ISLOWER(cu)) |
7 | 2544 regc(cu); |
2545 break; | |
2546 case CLASS_PRINT: | |
2547 for (cu = 1; cu <= 255; cu++) | |
2548 if (vim_isprintc(cu)) | |
2549 regc(cu); | |
2550 break; | |
2551 case CLASS_PUNCT: | |
2552 for (cu = 1; cu <= 255; cu++) | |
2553 if (ispunct(cu)) | |
2554 regc(cu); | |
2555 break; | |
2556 case CLASS_SPACE: | |
2557 for (cu = 9; cu <= 13; cu++) | |
2558 regc(cu); | |
2559 regc(' '); | |
2560 break; | |
2561 case CLASS_UPPER: | |
2562 for (cu = 1; cu <= 255; cu++) | |
1347 | 2563 if (MB_ISUPPER(cu)) |
7 | 2564 regc(cu); |
2565 break; | |
2566 case CLASS_XDIGIT: | |
2567 for (cu = 1; cu <= 255; cu++) | |
2568 if (vim_isxdigit(cu)) | |
2569 regc(cu); | |
2570 break; | |
2571 case CLASS_TAB: | |
2572 regc('\t'); | |
2573 break; | |
2574 case CLASS_RETURN: | |
2575 regc('\r'); | |
2576 break; | |
2577 case CLASS_BACKSPACE: | |
2578 regc('\b'); | |
2579 break; | |
2580 case CLASS_ESCAPE: | |
2581 regc('\033'); | |
2582 break; | |
2583 } | |
2584 } | |
2585 else | |
2586 { | |
2587 #ifdef FEAT_MBYTE | |
2588 if (has_mbyte) | |
2589 { | |
2590 int len; | |
2591 | |
2592 /* produce a multibyte character, including any | |
2593 * following composing characters */ | |
2594 startc = mb_ptr2char(regparse); | |
474 | 2595 len = (*mb_ptr2len)(regparse); |
7 | 2596 if (enc_utf8 && utf_char2len(startc) != len) |
2597 startc = -1; /* composing chars */ | |
2598 while (--len >= 0) | |
2599 regc(*regparse++); | |
2600 } | |
2601 else | |
2602 #endif | |
2603 { | |
2604 startc = *regparse++; | |
2605 regc(startc); | |
2606 } | |
2607 } | |
2608 } | |
2609 regc(NUL); | |
2610 prevchr_len = 1; /* last char was the ']' */ | |
2611 if (*regparse != ']') | |
2612 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */ | |
2613 skipchr(); /* let's be friends with the lexer again */ | |
2614 *flagp |= HASWIDTH | SIMPLE; | |
2615 break; | |
2616 } | |
481 | 2617 else if (reg_strict) |
4444 | 2618 EMSG2_RET_NULL(_(e_missingbracket), reg_magic > MAGIC_OFF); |
7 | 2619 } |
2620 /* FALLTHROUGH */ | |
2621 | |
2622 default: | |
2623 { | |
2624 int len; | |
2625 | |
2626 #ifdef FEAT_MBYTE | |
2627 /* A multi-byte character is handled as a separate atom if it's | |
714 | 2628 * before a multi and when it's a composing char. */ |
2629 if (use_multibytecode(c)) | |
7 | 2630 { |
714 | 2631 do_multibyte: |
7 | 2632 ret = regnode(MULTIBYTECODE); |
2633 regmbc(c); | |
2634 *flagp |= HASWIDTH | SIMPLE; | |
2635 break; | |
2636 } | |
2637 #endif | |
2638 | |
2639 ret = regnode(EXACTLY); | |
2640 | |
2641 /* | |
2642 * Append characters as long as: | |
2643 * - there is no following multi, we then need the character in | |
2644 * front of it as a single character operand | |
2645 * - not running into a Magic character | |
2646 * - "one_exactly" is not set | |
2647 * But always emit at least one character. Might be a Multi, | |
2648 * e.g., a "[" without matching "]". | |
2649 */ | |
2650 for (len = 0; c != NUL && (len == 0 | |
2651 || (re_multi_type(peekchr()) == NOT_MULTI | |
2652 && !one_exactly | |
2653 && !is_Magic(c))); ++len) | |
2654 { | |
2655 c = no_Magic(c); | |
2656 #ifdef FEAT_MBYTE | |
2657 if (has_mbyte) | |
2658 { | |
2659 regmbc(c); | |
2660 if (enc_utf8) | |
2661 { | |
2662 int l; | |
2663 | |
714 | 2664 /* Need to get composing character too. */ |
7 | 2665 for (;;) |
2666 { | |
714 | 2667 l = utf_ptr2len(regparse); |
2668 if (!UTF_COMPOSINGLIKE(regparse, regparse + l)) | |
7 | 2669 break; |
714 | 2670 regmbc(utf_ptr2char(regparse)); |
2671 skipchr(); | |
7 | 2672 } |
2673 } | |
2674 } | |
2675 else | |
2676 #endif | |
2677 regc(c); | |
2678 c = getchr(); | |
2679 } | |
2680 ungetchr(); | |
2681 | |
2682 regc(NUL); | |
2683 *flagp |= HASWIDTH; | |
2684 if (len == 1) | |
2685 *flagp |= SIMPLE; | |
2686 } | |
2687 break; | |
2688 } | |
2689 | |
2690 return ret; | |
2691 } | |
2692 | |
714 | 2693 #ifdef FEAT_MBYTE |
2694 /* | |
2695 * Return TRUE if MULTIBYTECODE should be used instead of EXACTLY for | |
2696 * character "c". | |
2697 */ | |
2698 static int | |
2699 use_multibytecode(c) | |
2700 int c; | |
2701 { | |
2702 return has_mbyte && (*mb_char2len)(c) > 1 | |
2703 && (re_multi_type(peekchr()) != NOT_MULTI | |
2704 || (enc_utf8 && utf_iscomposing(c))); | |
2705 } | |
2706 #endif | |
2707 | |
7 | 2708 /* |
4444 | 2709 * Emit a node. |
7 | 2710 * Return pointer to generated code. |
2711 */ | |
2712 static char_u * | |
2713 regnode(op) | |
2714 int op; | |
2715 { | |
2716 char_u *ret; | |
2717 | |
2718 ret = regcode; | |
2719 if (ret == JUST_CALC_SIZE) | |
2720 regsize += 3; | |
2721 else | |
2722 { | |
2723 *regcode++ = op; | |
2724 *regcode++ = NUL; /* Null "next" pointer. */ | |
2725 *regcode++ = NUL; | |
2726 } | |
2727 return ret; | |
2728 } | |
2729 | |
2730 /* | |
2731 * Emit (if appropriate) a byte of code | |
2732 */ | |
2733 static void | |
2734 regc(b) | |
2735 int b; | |
2736 { | |
2737 if (regcode == JUST_CALC_SIZE) | |
2738 regsize++; | |
2739 else | |
2740 *regcode++ = b; | |
2741 } | |
2742 | |
2743 #ifdef FEAT_MBYTE | |
2744 /* | |
2745 * Emit (if appropriate) a multi-byte character of code | |
2746 */ | |
2747 static void | |
2748 regmbc(c) | |
2749 int c; | |
2750 { | |
2974 | 2751 if (!has_mbyte && c > 0xff) |
2752 return; | |
7 | 2753 if (regcode == JUST_CALC_SIZE) |
2754 regsize += (*mb_char2len)(c); | |
2755 else | |
2756 regcode += (*mb_char2bytes)(c, regcode); | |
2757 } | |
2758 #endif | |
2759 | |
2760 /* | |
4444 | 2761 * Insert an operator in front of already-emitted operand |
7 | 2762 * |
2763 * Means relocating the operand. | |
2764 */ | |
2765 static void | |
2766 reginsert(op, opnd) | |
2767 int op; | |
2768 char_u *opnd; | |
2769 { | |
2770 char_u *src; | |
2771 char_u *dst; | |
2772 char_u *place; | |
2773 | |
2774 if (regcode == JUST_CALC_SIZE) | |
2775 { | |
2776 regsize += 3; | |
2777 return; | |
2778 } | |
2779 src = regcode; | |
2780 regcode += 3; | |
2781 dst = regcode; | |
2782 while (src > opnd) | |
2783 *--dst = *--src; | |
2784 | |
2785 place = opnd; /* Op node, where operand used to be. */ | |
2786 *place++ = op; | |
2787 *place++ = NUL; | |
2788 *place = NUL; | |
2789 } | |
2790 | |
2791 /* | |
4444 | 2792 * Insert an operator in front of already-emitted operand. |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2793 * Add a number to the operator. |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2794 */ |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2795 static void |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2796 reginsert_nr(op, val, opnd) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2797 int op; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2798 long val; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2799 char_u *opnd; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2800 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2801 char_u *src; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2802 char_u *dst; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2803 char_u *place; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2804 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2805 if (regcode == JUST_CALC_SIZE) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2806 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2807 regsize += 7; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2808 return; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2809 } |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2810 src = regcode; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2811 regcode += 7; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2812 dst = regcode; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2813 while (src > opnd) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2814 *--dst = *--src; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2815 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2816 place = opnd; /* Op node, where operand used to be. */ |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2817 *place++ = op; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2818 *place++ = NUL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2819 *place++ = NUL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2820 place = re_put_long(place, (long_u)val); |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2821 } |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2822 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2823 /* |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2824 * Insert an operator in front of already-emitted operand. |
7 | 2825 * The operator has the given limit values as operands. Also set next pointer. |
2826 * | |
2827 * Means relocating the operand. | |
2828 */ | |
2829 static void | |
2830 reginsert_limits(op, minval, maxval, opnd) | |
2831 int op; | |
2832 long minval; | |
2833 long maxval; | |
2834 char_u *opnd; | |
2835 { | |
2836 char_u *src; | |
2837 char_u *dst; | |
2838 char_u *place; | |
2839 | |
2840 if (regcode == JUST_CALC_SIZE) | |
2841 { | |
2842 regsize += 11; | |
2843 return; | |
2844 } | |
2845 src = regcode; | |
2846 regcode += 11; | |
2847 dst = regcode; | |
2848 while (src > opnd) | |
2849 *--dst = *--src; | |
2850 | |
2851 place = opnd; /* Op node, where operand used to be. */ | |
2852 *place++ = op; | |
2853 *place++ = NUL; | |
2854 *place++ = NUL; | |
2855 place = re_put_long(place, (long_u)minval); | |
2856 place = re_put_long(place, (long_u)maxval); | |
2857 regtail(opnd, place); | |
2858 } | |
2859 | |
2860 /* | |
2861 * Write a long as four bytes at "p" and return pointer to the next char. | |
2862 */ | |
2863 static char_u * | |
2864 re_put_long(p, val) | |
2865 char_u *p; | |
2866 long_u val; | |
2867 { | |
2868 *p++ = (char_u) ((val >> 24) & 0377); | |
2869 *p++ = (char_u) ((val >> 16) & 0377); | |
2870 *p++ = (char_u) ((val >> 8) & 0377); | |
2871 *p++ = (char_u) (val & 0377); | |
2872 return p; | |
2873 } | |
2874 | |
2875 /* | |
4444 | 2876 * Set the next-pointer at the end of a node chain. |
7 | 2877 */ |
2878 static void | |
2879 regtail(p, val) | |
2880 char_u *p; | |
2881 char_u *val; | |
2882 { | |
2883 char_u *scan; | |
2884 char_u *temp; | |
2885 int offset; | |
2886 | |
2887 if (p == JUST_CALC_SIZE) | |
2888 return; | |
2889 | |
2890 /* Find last node. */ | |
2891 scan = p; | |
2892 for (;;) | |
2893 { | |
2894 temp = regnext(scan); | |
2895 if (temp == NULL) | |
2896 break; | |
2897 scan = temp; | |
2898 } | |
2899 | |
233 | 2900 if (OP(scan) == BACK) |
7 | 2901 offset = (int)(scan - val); |
2902 else | |
2903 offset = (int)(val - scan); | |
2010 | 2904 /* When the offset uses more than 16 bits it can no longer fit in the two |
2974 | 2905 * bytes available. Use a global flag to avoid having to check return |
2010 | 2906 * values in too many places. */ |
2907 if (offset > 0xffff) | |
2908 reg_toolong = TRUE; | |
2909 else | |
2910 { | |
2911 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377); | |
2912 *(scan + 2) = (char_u) (offset & 0377); | |
2913 } | |
7 | 2914 } |
2915 | |
2916 /* | |
4444 | 2917 * Like regtail, on item after a BRANCH; nop if none. |
7 | 2918 */ |
2919 static void | |
2920 regoptail(p, val) | |
2921 char_u *p; | |
2922 char_u *val; | |
2923 { | |
2924 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */ | |
2925 if (p == NULL || p == JUST_CALC_SIZE | |
2926 || (OP(p) != BRANCH | |
2927 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9))) | |
2928 return; | |
2929 regtail(OPERAND(p), val); | |
2930 } | |
2931 | |
2932 /* | |
4444 | 2933 * Functions for getting characters from the regexp input. |
7 | 2934 */ |
2935 | |
2936 static int at_start; /* True when on the first character */ | |
2937 static int prev_at_start; /* True when on the second character */ | |
2938 | |
4444 | 2939 /* |
2940 * Start parsing at "str". | |
2941 */ | |
7 | 2942 static void |
2943 initchr(str) | |
2944 char_u *str; | |
2945 { | |
2946 regparse = str; | |
2947 prevchr_len = 0; | |
2948 curchr = prevprevchr = prevchr = nextchr = -1; | |
2949 at_start = TRUE; | |
2950 prev_at_start = FALSE; | |
2951 } | |
2952 | |
4444 | 2953 /* |
2954 * Get the next character without advancing. | |
2955 */ | |
7 | 2956 static int |
2957 peekchr() | |
2958 { | |
167 | 2959 static int after_slash = FALSE; |
2960 | |
7 | 2961 if (curchr == -1) |
2962 { | |
2963 switch (curchr = regparse[0]) | |
2964 { | |
2965 case '.': | |
2966 case '[': | |
2967 case '~': | |
2968 /* magic when 'magic' is on */ | |
2969 if (reg_magic >= MAGIC_ON) | |
2970 curchr = Magic(curchr); | |
2971 break; | |
2972 case '(': | |
2973 case ')': | |
2974 case '{': | |
2975 case '%': | |
2976 case '+': | |
2977 case '=': | |
2978 case '?': | |
2979 case '@': | |
2980 case '!': | |
2981 case '&': | |
2982 case '|': | |
2983 case '<': | |
2984 case '>': | |
2985 case '#': /* future ext. */ | |
2986 case '"': /* future ext. */ | |
2987 case '\'': /* future ext. */ | |
2988 case ',': /* future ext. */ | |
2989 case '-': /* future ext. */ | |
2990 case ':': /* future ext. */ | |
2991 case ';': /* future ext. */ | |
2992 case '`': /* future ext. */ | |
2993 case '/': /* Can't be used in / command */ | |
2994 /* magic only after "\v" */ | |
2995 if (reg_magic == MAGIC_ALL) | |
2996 curchr = Magic(curchr); | |
2997 break; | |
2998 case '*': | |
167 | 2999 /* * is not magic as the very first character, eg "?*ptr", when |
3000 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But | |
3001 * "\(\*" is not magic, thus must be magic if "after_slash" */ | |
3002 if (reg_magic >= MAGIC_ON | |
3003 && !at_start | |
3004 && !(prev_at_start && prevchr == Magic('^')) | |
3005 && (after_slash | |
3006 || (prevchr != Magic('(') | |
3007 && prevchr != Magic('&') | |
3008 && prevchr != Magic('|')))) | |
7 | 3009 curchr = Magic('*'); |
3010 break; | |
3011 case '^': | |
3012 /* '^' is only magic as the very first character and if it's after | |
3013 * "\(", "\|", "\&' or "\n" */ | |
3014 if (reg_magic >= MAGIC_OFF | |
3015 && (at_start | |
3016 || reg_magic == MAGIC_ALL | |
3017 || prevchr == Magic('(') | |
3018 || prevchr == Magic('|') | |
3019 || prevchr == Magic('&') | |
3020 || prevchr == Magic('n') | |
3021 || (no_Magic(prevchr) == '(' | |
3022 && prevprevchr == Magic('%')))) | |
3023 { | |
3024 curchr = Magic('^'); | |
3025 at_start = TRUE; | |
3026 prev_at_start = FALSE; | |
3027 } | |
3028 break; | |
3029 case '$': | |
3030 /* '$' is only magic as the very last char and if it's in front of | |
3031 * either "\|", "\)", "\&", or "\n" */ | |
3032 if (reg_magic >= MAGIC_OFF) | |
3033 { | |
3034 char_u *p = regparse + 1; | |
3035 | |
3036 /* ignore \c \C \m and \M after '$' */ | |
3037 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C' | |
3038 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z')) | |
3039 p += 2; | |
3040 if (p[0] == NUL | |
3041 || (p[0] == '\\' | |
3042 && (p[1] == '|' || p[1] == '&' || p[1] == ')' | |
3043 || p[1] == 'n')) | |
3044 || reg_magic == MAGIC_ALL) | |
3045 curchr = Magic('$'); | |
3046 } | |
3047 break; | |
3048 case '\\': | |
3049 { | |
3050 int c = regparse[1]; | |
3051 | |
3052 if (c == NUL) | |
3053 curchr = '\\'; /* trailing '\' */ | |
3054 else if ( | |
3055 #ifdef EBCDIC | |
3056 vim_strchr(META, c) | |
3057 #else | |
3058 c <= '~' && META_flags[c] | |
3059 #endif | |
3060 ) | |
3061 { | |
3062 /* | |
3063 * META contains everything that may be magic sometimes, | |
3064 * except ^ and $ ("\^" and "\$" are only magic after | |
3065 * "\v"). We now fetch the next character and toggle its | |
3066 * magicness. Therefore, \ is so meta-magic that it is | |
3067 * not in META. | |
3068 */ | |
3069 curchr = -1; | |
3070 prev_at_start = at_start; | |
3071 at_start = FALSE; /* be able to say "/\*ptr" */ | |
3072 ++regparse; | |
167 | 3073 ++after_slash; |
7 | 3074 peekchr(); |
3075 --regparse; | |
167 | 3076 --after_slash; |
7 | 3077 curchr = toggle_Magic(curchr); |
3078 } | |
3079 else if (vim_strchr(REGEXP_ABBR, c)) | |
3080 { | |
3081 /* | |
3082 * Handle abbreviations, like "\t" for TAB -- webb | |
3083 */ | |
3084 curchr = backslash_trans(c); | |
3085 } | |
3086 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^')) | |
3087 curchr = toggle_Magic(c); | |
3088 else | |
3089 { | |
3090 /* | |
3091 * Next character can never be (made) magic? | |
3092 * Then backslashing it won't do anything. | |
3093 */ | |
3094 #ifdef FEAT_MBYTE | |
3095 if (has_mbyte) | |
3096 curchr = (*mb_ptr2char)(regparse + 1); | |
3097 else | |
3098 #endif | |
3099 curchr = c; | |
3100 } | |
3101 break; | |
3102 } | |
3103 | |
3104 #ifdef FEAT_MBYTE | |
3105 default: | |
3106 if (has_mbyte) | |
3107 curchr = (*mb_ptr2char)(regparse); | |
3108 #endif | |
3109 } | |
3110 } | |
3111 | |
3112 return curchr; | |
3113 } | |
3114 | |
3115 /* | |
3116 * Eat one lexed character. Do this in a way that we can undo it. | |
3117 */ | |
3118 static void | |
3119 skipchr() | |
3120 { | |
3121 /* peekchr() eats a backslash, do the same here */ | |
3122 if (*regparse == '\\') | |
3123 prevchr_len = 1; | |
3124 else | |
3125 prevchr_len = 0; | |
3126 if (regparse[prevchr_len] != NUL) | |
3127 { | |
3128 #ifdef FEAT_MBYTE | |
714 | 3129 if (enc_utf8) |
1449 | 3130 /* exclude composing chars that mb_ptr2len does include */ |
3131 prevchr_len += utf_ptr2len(regparse + prevchr_len); | |
714 | 3132 else if (has_mbyte) |
474 | 3133 prevchr_len += (*mb_ptr2len)(regparse + prevchr_len); |
7 | 3134 else |
3135 #endif | |
3136 ++prevchr_len; | |
3137 } | |
3138 regparse += prevchr_len; | |
3139 prev_at_start = at_start; | |
3140 at_start = FALSE; | |
3141 prevprevchr = prevchr; | |
3142 prevchr = curchr; | |
3143 curchr = nextchr; /* use previously unget char, or -1 */ | |
3144 nextchr = -1; | |
3145 } | |
3146 | |
3147 /* | |
3148 * Skip a character while keeping the value of prev_at_start for at_start. | |
3149 * prevchr and prevprevchr are also kept. | |
3150 */ | |
3151 static void | |
3152 skipchr_keepstart() | |
3153 { | |
3154 int as = prev_at_start; | |
3155 int pr = prevchr; | |
3156 int prpr = prevprevchr; | |
3157 | |
3158 skipchr(); | |
3159 at_start = as; | |
3160 prevchr = pr; | |
3161 prevprevchr = prpr; | |
3162 } | |
3163 | |
4444 | 3164 /* |
3165 * Get the next character from the pattern. We know about magic and such, so | |
3166 * therefore we need a lexical analyzer. | |
3167 */ | |
7 | 3168 static int |
3169 getchr() | |
3170 { | |
3171 int chr = peekchr(); | |
3172 | |
3173 skipchr(); | |
3174 return chr; | |
3175 } | |
3176 | |
3177 /* | |
3178 * put character back. Works only once! | |
3179 */ | |
3180 static void | |
3181 ungetchr() | |
3182 { | |
3183 nextchr = curchr; | |
3184 curchr = prevchr; | |
3185 prevchr = prevprevchr; | |
3186 at_start = prev_at_start; | |
3187 prev_at_start = FALSE; | |
3188 | |
3189 /* Backup regparse, so that it's at the same position as before the | |
3190 * getchr(). */ | |
3191 regparse -= prevchr_len; | |
3192 } | |
3193 | |
3194 /* | |
29 | 3195 * Get and return the value of the hex string at the current position. |
3196 * Return -1 if there is no valid hex number. | |
3197 * The position is updated: | |
24 | 3198 * blahblah\%x20asdf |
856 | 3199 * before-^ ^-after |
24 | 3200 * The parameter controls the maximum number of input characters. This will be |
3201 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence. | |
3202 */ | |
3203 static int | |
3204 gethexchrs(maxinputlen) | |
3205 int maxinputlen; | |
3206 { | |
3207 int nr = 0; | |
3208 int c; | |
3209 int i; | |
3210 | |
3211 for (i = 0; i < maxinputlen; ++i) | |
3212 { | |
3213 c = regparse[0]; | |
3214 if (!vim_isxdigit(c)) | |
3215 break; | |
3216 nr <<= 4; | |
3217 nr |= hex2nr(c); | |
3218 ++regparse; | |
3219 } | |
3220 | |
3221 if (i == 0) | |
3222 return -1; | |
3223 return nr; | |
3224 } | |
3225 | |
3226 /* | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
3227 * Get and return the value of the decimal string immediately after the |
24 | 3228 * current position. Return -1 for invalid. Consumes all digits. |
3229 */ | |
3230 static int | |
3231 getdecchrs() | |
3232 { | |
3233 int nr = 0; | |
3234 int c; | |
3235 int i; | |
3236 | |
3237 for (i = 0; ; ++i) | |
3238 { | |
3239 c = regparse[0]; | |
3240 if (c < '0' || c > '9') | |
3241 break; | |
3242 nr *= 10; | |
3243 nr += c - '0'; | |
3244 ++regparse; | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
3245 curchr = -1; /* no longer valid */ |
24 | 3246 } |
3247 | |
3248 if (i == 0) | |
3249 return -1; | |
3250 return nr; | |
3251 } | |
3252 | |
3253 /* | |
3254 * get and return the value of the octal string immediately after the current | |
3255 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle | |
3256 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't | |
3257 * treat 8 or 9 as recognised characters. Position is updated: | |
3258 * blahblah\%o210asdf | |
856 | 3259 * before-^ ^-after |
24 | 3260 */ |
3261 static int | |
3262 getoctchrs() | |
3263 { | |
3264 int nr = 0; | |
3265 int c; | |
3266 int i; | |
3267 | |
3268 for (i = 0; i < 3 && nr < 040; ++i) | |
3269 { | |
3270 c = regparse[0]; | |
3271 if (c < '0' || c > '7') | |
3272 break; | |
3273 nr <<= 3; | |
3274 nr |= hex2nr(c); | |
3275 ++regparse; | |
3276 } | |
3277 | |
3278 if (i == 0) | |
3279 return -1; | |
3280 return nr; | |
3281 } | |
3282 | |
3283 /* | |
3284 * Get a number after a backslash that is inside []. | |
3285 * When nothing is recognized return a backslash. | |
3286 */ | |
3287 static int | |
3288 coll_get_char() | |
3289 { | |
3290 int nr = -1; | |
3291 | |
3292 switch (*regparse++) | |
3293 { | |
3294 case 'd': nr = getdecchrs(); break; | |
3295 case 'o': nr = getoctchrs(); break; | |
3296 case 'x': nr = gethexchrs(2); break; | |
3297 case 'u': nr = gethexchrs(4); break; | |
3298 case 'U': nr = gethexchrs(8); break; | |
3299 } | |
3300 if (nr < 0) | |
3301 { | |
3302 /* If getting the number fails be backwards compatible: the character | |
3303 * is a backslash. */ | |
3304 --regparse; | |
3305 nr = '\\'; | |
3306 } | |
3307 return nr; | |
3308 } | |
3309 | |
3310 /* | |
7 | 3311 * read_limits - Read two integers to be taken as a minimum and maximum. |
3312 * If the first character is '-', then the range is reversed. | |
3313 * Should end with 'end'. If minval is missing, zero is default, if maxval is | |
3314 * missing, a very big number is the default. | |
3315 */ | |
3316 static int | |
3317 read_limits(minval, maxval) | |
3318 long *minval; | |
3319 long *maxval; | |
3320 { | |
3321 int reverse = FALSE; | |
3322 char_u *first_char; | |
3323 long tmp; | |
3324 | |
3325 if (*regparse == '-') | |
3326 { | |
3327 /* Starts with '-', so reverse the range later */ | |
3328 regparse++; | |
3329 reverse = TRUE; | |
3330 } | |
3331 first_char = regparse; | |
3332 *minval = getdigits(®parse); | |
3333 if (*regparse == ',') /* There is a comma */ | |
3334 { | |
3335 if (vim_isdigit(*++regparse)) | |
3336 *maxval = getdigits(®parse); | |
3337 else | |
3338 *maxval = MAX_LIMIT; | |
3339 } | |
3340 else if (VIM_ISDIGIT(*first_char)) | |
3341 *maxval = *minval; /* It was \{n} or \{-n} */ | |
3342 else | |
3343 *maxval = MAX_LIMIT; /* It was \{} or \{-} */ | |
3344 if (*regparse == '\\') | |
3345 regparse++; /* Allow either \{...} or \{...\} */ | |
167 | 3346 if (*regparse != '}') |
7 | 3347 { |
3348 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"), | |
3349 reg_magic == MAGIC_ALL ? "" : "\\"); | |
3350 EMSG_RET_FAIL(IObuff); | |
3351 } | |
3352 | |
3353 /* | |
3354 * Reverse the range if there was a '-', or make sure it is in the right | |
3355 * order otherwise. | |
3356 */ | |
3357 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval)) | |
3358 { | |
3359 tmp = *minval; | |
3360 *minval = *maxval; | |
3361 *maxval = tmp; | |
3362 } | |
3363 skipchr(); /* let's be friends with the lexer again */ | |
3364 return OK; | |
3365 } | |
3366 | |
3367 /* | |
3368 * vim_regexec and friends | |
3369 */ | |
3370 | |
3371 /* | |
3372 * Global work variables for vim_regexec(). | |
3373 */ | |
3374 | |
3375 /* The current match-position is remembered with these variables: */ | |
3376 static linenr_T reglnum; /* line number, relative to first line */ | |
3377 static char_u *regline; /* start of current line */ | |
3378 static char_u *reginput; /* current input, points into "regline" */ | |
3379 | |
3380 static int need_clear_subexpr; /* subexpressions still need to be | |
3381 * cleared */ | |
3382 #ifdef FEAT_SYN_HL | |
3383 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions | |
3384 * still need to be cleared */ | |
3385 #endif | |
3386 | |
3387 /* | |
3388 * Structure used to save the current input state, when it needs to be | |
3389 * restored after trying a match. Used by reg_save() and reg_restore(). | |
233 | 3390 * Also stores the length of "backpos". |
7 | 3391 */ |
3392 typedef struct | |
3393 { | |
3394 union | |
3395 { | |
3396 char_u *ptr; /* reginput pointer, for single-line regexp */ | |
3397 lpos_T pos; /* reginput pos, for multi-line regexp */ | |
3398 } rs_u; | |
233 | 3399 int rs_len; |
7 | 3400 } regsave_T; |
3401 | |
3402 /* struct to save start/end pointer/position in for \(\) */ | |
3403 typedef struct | |
3404 { | |
3405 union | |
3406 { | |
3407 char_u *ptr; | |
3408 lpos_T pos; | |
3409 } se_u; | |
3410 } save_se_T; | |
3411 | |
1579 | 3412 /* used for BEHIND and NOBEHIND matching */ |
3413 typedef struct regbehind_S | |
3414 { | |
3415 regsave_T save_after; | |
3416 regsave_T save_behind; | |
1602 | 3417 int save_need_clear_subexpr; |
1579 | 3418 save_se_T save_start[NSUBEXP]; |
3419 save_se_T save_end[NSUBEXP]; | |
3420 } regbehind_T; | |
3421 | |
7 | 3422 static char_u *reg_getline __ARGS((linenr_T lnum)); |
4444 | 3423 static long bt_regexec_both __ARGS((char_u *line, colnr_T col, proftime_T *tm)); |
3424 static long regtry __ARGS((bt_regprog_T *prog, colnr_T col)); | |
7 | 3425 static void cleanup_subexpr __ARGS((void)); |
3426 #ifdef FEAT_SYN_HL | |
3427 static void cleanup_zsubexpr __ARGS((void)); | |
3428 #endif | |
1579 | 3429 static void save_subexpr __ARGS((regbehind_T *bp)); |
3430 static void restore_subexpr __ARGS((regbehind_T *bp)); | |
7 | 3431 static void reg_nextline __ARGS((void)); |
233 | 3432 static void reg_save __ARGS((regsave_T *save, garray_T *gap)); |
3433 static void reg_restore __ARGS((regsave_T *save, garray_T *gap)); | |
7 | 3434 static int reg_save_equal __ARGS((regsave_T *save)); |
3435 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp)); | |
3436 static void save_se_one __ARGS((save_se_T *savep, char_u **pp)); | |
3437 | |
3438 /* Save the sub-expressions before attempting a match. */ | |
3439 #define save_se(savep, posp, pp) \ | |
3440 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp)) | |
3441 | |
3442 /* After a failed match restore the sub-expressions. */ | |
3443 #define restore_se(savep, posp, pp) { \ | |
3444 if (REG_MULTI) \ | |
3445 *(posp) = (savep)->se_u.pos; \ | |
3446 else \ | |
3447 *(pp) = (savep)->se_u.ptr; } | |
3448 | |
3449 static int re_num_cmp __ARGS((long_u val, char_u *scan)); | |
180 | 3450 static int regmatch __ARGS((char_u *prog)); |
7 | 3451 static int regrepeat __ARGS((char_u *p, long maxcount)); |
3452 | |
3453 #ifdef DEBUG | |
3454 int regnarrate = 0; | |
3455 #endif | |
3456 | |
3457 /* | |
3458 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec(). | |
3459 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern | |
3460 * contains '\c' or '\C' the value is overruled. | |
3461 */ | |
3462 static int ireg_ic; | |
3463 | |
3464 #ifdef FEAT_MBYTE | |
3465 /* | |
3466 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag | |
3467 * in the regexp. Defaults to false, always. | |
3468 */ | |
3469 static int ireg_icombine; | |
3470 #endif | |
3471 | |
3472 /* | |
410 | 3473 * Copy of "rmm_maxcol": maximum column to search for a match. Zero when |
3474 * there is no maximum. | |
3475 */ | |
418 | 3476 static colnr_T ireg_maxcol; |
410 | 3477 |
3478 /* | |
7 | 3479 * Sometimes need to save a copy of a line. Since alloc()/free() is very |
3480 * slow, we keep one allocated piece of memory and only re-allocate it when | |
4444 | 3481 * it's too small. It's freed in bt_regexec_both() when finished. |
7 | 3482 */ |
1468 | 3483 static char_u *reg_tofree = NULL; |
7 | 3484 static unsigned reg_tofreelen; |
3485 | |
3486 /* | |
3487 * These variables are set when executing a regexp to speed up the execution. | |
1209 | 3488 * Which ones are set depends on whether a single-line or multi-line match is |
7 | 3489 * done: |
3490 * single-line multi-line | |
3491 * reg_match ®match_T NULL | |
3492 * reg_mmatch NULL ®mmatch_T | |
3493 * reg_startp reg_match->startp <invalid> | |
3494 * reg_endp reg_match->endp <invalid> | |
3495 * reg_startpos <invalid> reg_mmatch->startpos | |
3496 * reg_endpos <invalid> reg_mmatch->endpos | |
3497 * reg_win NULL window in which to search | |
4061 | 3498 * reg_buf curbuf buffer in which to search |
7 | 3499 * reg_firstlnum <invalid> first line in which to search |
3500 * reg_maxline 0 last line nr | |
3501 * reg_line_lbr FALSE or TRUE FALSE | |
3502 */ | |
3503 static regmatch_T *reg_match; | |
3504 static regmmatch_T *reg_mmatch; | |
3505 static char_u **reg_startp = NULL; | |
3506 static char_u **reg_endp = NULL; | |
3507 static lpos_T *reg_startpos = NULL; | |
3508 static lpos_T *reg_endpos = NULL; | |
3509 static win_T *reg_win; | |
3510 static buf_T *reg_buf; | |
3511 static linenr_T reg_firstlnum; | |
3512 static linenr_T reg_maxline; | |
3513 static int reg_line_lbr; /* "\n" in string is line break */ | |
3514 | |
270 | 3515 /* Values for rs_state in regitem_T. */ |
3516 typedef enum regstate_E | |
3517 { | |
3518 RS_NOPEN = 0 /* NOPEN and NCLOSE */ | |
3519 , RS_MOPEN /* MOPEN + [0-9] */ | |
3520 , RS_MCLOSE /* MCLOSE + [0-9] */ | |
3521 #ifdef FEAT_SYN_HL | |
3522 , RS_ZOPEN /* ZOPEN + [0-9] */ | |
3523 , RS_ZCLOSE /* ZCLOSE + [0-9] */ | |
3524 #endif | |
3525 , RS_BRANCH /* BRANCH */ | |
3526 , RS_BRCPLX_MORE /* BRACE_COMPLEX and trying one more match */ | |
3527 , RS_BRCPLX_LONG /* BRACE_COMPLEX and trying longest match */ | |
3528 , RS_BRCPLX_SHORT /* BRACE_COMPLEX and trying shortest match */ | |
3529 , RS_NOMATCH /* NOMATCH */ | |
3530 , RS_BEHIND1 /* BEHIND / NOBEHIND matching rest */ | |
3531 , RS_BEHIND2 /* BEHIND / NOBEHIND matching behind part */ | |
3532 , RS_STAR_LONG /* STAR/PLUS/BRACE_SIMPLE longest match */ | |
3533 , RS_STAR_SHORT /* STAR/PLUS/BRACE_SIMPLE shortest match */ | |
3534 } regstate_T; | |
3535 | |
3536 /* | |
3537 * When there are alternatives a regstate_T is put on the regstack to remember | |
3538 * what we are doing. | |
3539 * Before it may be another type of item, depending on rs_state, to remember | |
3540 * more things. | |
3541 */ | |
3542 typedef struct regitem_S | |
3543 { | |
3544 regstate_T rs_state; /* what we are doing, one of RS_ above */ | |
3545 char_u *rs_scan; /* current node in program */ | |
3546 union | |
3547 { | |
3548 save_se_T sesave; | |
3549 regsave_T regsave; | |
3550 } rs_un; /* room for saving reginput */ | |
1579 | 3551 short rs_no; /* submatch nr or BEHIND/NOBEHIND */ |
270 | 3552 } regitem_T; |
3553 | |
3554 static regitem_T *regstack_push __ARGS((regstate_T state, char_u *scan)); | |
3555 static void regstack_pop __ARGS((char_u **scan)); | |
3556 | |
3557 /* used for STAR, PLUS and BRACE_SIMPLE matching */ | |
3558 typedef struct regstar_S | |
3559 { | |
3560 int nextb; /* next byte */ | |
3561 int nextb_ic; /* next byte reverse case */ | |
3562 long count; | |
3563 long minval; | |
3564 long maxval; | |
3565 } regstar_T; | |
3566 | |
3567 /* used to store input position when a BACK was encountered, so that we now if | |
3568 * we made any progress since the last time. */ | |
3569 typedef struct backpos_S | |
3570 { | |
3571 char_u *bp_scan; /* "scan" where BACK was encountered */ | |
3572 regsave_T bp_pos; /* last input position */ | |
3573 } backpos_T; | |
3574 | |
3575 /* | |
1520 | 3576 * "regstack" and "backpos" are used by regmatch(). They are kept over calls |
3577 * to avoid invoking malloc() and free() often. | |
3578 * "regstack" is a stack with regitem_T items, sometimes preceded by regstar_T | |
3579 * or regbehind_T. | |
3580 * "backpos_T" is a table with backpos_T for BACK | |
270 | 3581 */ |
1520 | 3582 static garray_T regstack = {0, 0, 0, 0, NULL}; |
3583 static garray_T backpos = {0, 0, 0, 0, NULL}; | |
3584 | |
3585 /* | |
3586 * Both for regstack and backpos tables we use the following strategy of | |
3587 * allocation (to reduce malloc/free calls): | |
3588 * - Initial size is fairly small. | |
3589 * - When needed, the tables are grown bigger (8 times at first, double after | |
3590 * that). | |
3591 * - After executing the match we free the memory only if the array has grown. | |
3592 * Thus the memory is kept allocated when it's at the initial size. | |
3593 * This makes it fast while not keeping a lot of memory allocated. | |
3594 * A three times speed increase was observed when using many simple patterns. | |
3595 */ | |
3596 #define REGSTACK_INITIAL 2048 | |
3597 #define BACKPOS_INITIAL 64 | |
3598 | |
3599 #if defined(EXITFREE) || defined(PROTO) | |
3600 void | |
3601 free_regexp_stuff() | |
3602 { | |
3603 ga_clear(®stack); | |
3604 ga_clear(&backpos); | |
3605 vim_free(reg_tofree); | |
3606 vim_free(reg_prev_sub); | |
3607 } | |
3608 #endif | |
270 | 3609 |
7 | 3610 /* |
3611 * Get pointer to the line "lnum", which is relative to "reg_firstlnum". | |
3612 */ | |
3613 static char_u * | |
3614 reg_getline(lnum) | |
3615 linenr_T lnum; | |
3616 { | |
3617 /* when looking behind for a match/no-match lnum is negative. But we | |
3618 * can't go before line 1 */ | |
3619 if (reg_firstlnum + lnum < 1) | |
3620 return NULL; | |
482 | 3621 if (lnum > reg_maxline) |
481 | 3622 /* Must have matched the "\n" in the last line. */ |
3623 return (char_u *)""; | |
7 | 3624 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE); |
3625 } | |
3626 | |
3627 static regsave_T behind_pos; | |
3628 | |
3629 #ifdef FEAT_SYN_HL | |
3630 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */ | |
3631 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */ | |
3632 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */ | |
3633 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */ | |
3634 #endif | |
3635 | |
3636 /* TRUE if using multi-line regexp. */ | |
3637 #define REG_MULTI (reg_match == NULL) | |
3638 | |
4444 | 3639 static int bt_regexec __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); |
3640 | |
7 | 3641 /* |
3642 * Match a regexp against a string. | |
3643 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3644 * Uses curbuf for line count and 'iskeyword'. | |
3645 * | |
3646 * Return TRUE if there is a match, FALSE if not. | |
3647 */ | |
4444 | 3648 static int |
3649 bt_regexec(rmp, line, col) | |
7 | 3650 regmatch_T *rmp; |
3651 char_u *line; /* string to match against */ | |
3652 colnr_T col; /* column to start looking for match */ | |
3653 { | |
3654 reg_match = rmp; | |
3655 reg_mmatch = NULL; | |
3656 reg_maxline = 0; | |
3657 reg_line_lbr = FALSE; | |
4061 | 3658 reg_buf = curbuf; |
7 | 3659 reg_win = NULL; |
3660 ireg_ic = rmp->rm_ic; | |
3661 #ifdef FEAT_MBYTE | |
3662 ireg_icombine = FALSE; | |
3663 #endif | |
410 | 3664 ireg_maxcol = 0; |
4444 | 3665 return (bt_regexec_both(line, col, NULL) != 0); |
7 | 3666 } |
3667 | |
39 | 3668 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ |
3669 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
4444 | 3670 |
3671 static int bt_regexec_nl __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
3672 | |
7 | 3673 /* |
3674 * Like vim_regexec(), but consider a "\n" in "line" to be a line break. | |
3675 */ | |
4444 | 3676 static int |
3677 bt_regexec_nl(rmp, line, col) | |
7 | 3678 regmatch_T *rmp; |
3679 char_u *line; /* string to match against */ | |
3680 colnr_T col; /* column to start looking for match */ | |
3681 { | |
3682 reg_match = rmp; | |
3683 reg_mmatch = NULL; | |
3684 reg_maxline = 0; | |
3685 reg_line_lbr = TRUE; | |
4061 | 3686 reg_buf = curbuf; |
7 | 3687 reg_win = NULL; |
3688 ireg_ic = rmp->rm_ic; | |
3689 #ifdef FEAT_MBYTE | |
3690 ireg_icombine = FALSE; | |
3691 #endif | |
410 | 3692 ireg_maxcol = 0; |
4444 | 3693 return (bt_regexec_both(line, col, NULL) != 0); |
7 | 3694 } |
3695 #endif | |
3696 | |
4444 | 3697 static long bt_regexec_multi __ARGS((regmmatch_T *rmp, win_T *win, buf_T *buf, linenr_T lnum, colnr_T col, proftime_T *tm)); |
3698 | |
7 | 3699 /* |
3700 * Match a regexp against multiple lines. | |
3701 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3702 * Uses curbuf for line count and 'iskeyword'. | |
3703 * | |
3704 * Return zero if there is no match. Return number of lines contained in the | |
3705 * match otherwise. | |
3706 */ | |
4444 | 3707 static long |
3708 bt_regexec_multi(rmp, win, buf, lnum, col, tm) | |
7 | 3709 regmmatch_T *rmp; |
3710 win_T *win; /* window in which to search or NULL */ | |
3711 buf_T *buf; /* buffer in which to search */ | |
3712 linenr_T lnum; /* nr of line to start looking for match */ | |
3713 colnr_T col; /* column to start looking for match */ | |
1521 | 3714 proftime_T *tm; /* timeout limit or NULL */ |
7 | 3715 { |
3716 long r; | |
3717 | |
3718 reg_match = NULL; | |
3719 reg_mmatch = rmp; | |
3720 reg_buf = buf; | |
3721 reg_win = win; | |
3722 reg_firstlnum = lnum; | |
3723 reg_maxline = reg_buf->b_ml.ml_line_count - lnum; | |
3724 reg_line_lbr = FALSE; | |
3725 ireg_ic = rmp->rmm_ic; | |
3726 #ifdef FEAT_MBYTE | |
3727 ireg_icombine = FALSE; | |
3728 #endif | |
410 | 3729 ireg_maxcol = rmp->rmm_maxcol; |
7 | 3730 |
4444 | 3731 r = bt_regexec_both(NULL, col, tm); |
7 | 3732 |
3733 return r; | |
3734 } | |
3735 | |
3736 /* | |
3737 * Match a regexp against a string ("line" points to the string) or multiple | |
3738 * lines ("line" is NULL, use reg_getline()). | |
3739 */ | |
3740 static long | |
4444 | 3741 bt_regexec_both(line, col, tm) |
7 | 3742 char_u *line; |
3743 colnr_T col; /* column to start looking for match */ | |
1877 | 3744 proftime_T *tm UNUSED; /* timeout limit or NULL */ |
7 | 3745 { |
4444 | 3746 bt_regprog_T *prog; |
7 | 3747 char_u *s; |
189 | 3748 long retval = 0L; |
7 | 3749 |
1520 | 3750 /* Create "regstack" and "backpos" if they are not allocated yet. |
3751 * We allocate *_INITIAL amount of bytes first and then set the grow size | |
3752 * to much bigger value to avoid many malloc calls in case of deep regular | |
3753 * expressions. */ | |
3754 if (regstack.ga_data == NULL) | |
3755 { | |
3756 /* Use an item size of 1 byte, since we push different things | |
3757 * onto the regstack. */ | |
3758 ga_init2(®stack, 1, REGSTACK_INITIAL); | |
3759 ga_grow(®stack, REGSTACK_INITIAL); | |
3760 regstack.ga_growsize = REGSTACK_INITIAL * 8; | |
3761 } | |
3762 | |
3763 if (backpos.ga_data == NULL) | |
3764 { | |
3765 ga_init2(&backpos, sizeof(backpos_T), BACKPOS_INITIAL); | |
3766 ga_grow(&backpos, BACKPOS_INITIAL); | |
3767 backpos.ga_growsize = BACKPOS_INITIAL * 8; | |
3768 } | |
270 | 3769 |
7 | 3770 if (REG_MULTI) |
3771 { | |
4444 | 3772 prog = (bt_regprog_T *)reg_mmatch->regprog; |
7 | 3773 line = reg_getline((linenr_T)0); |
3774 reg_startpos = reg_mmatch->startpos; | |
3775 reg_endpos = reg_mmatch->endpos; | |
3776 } | |
3777 else | |
3778 { | |
4444 | 3779 prog = (bt_regprog_T *)reg_match->regprog; |
7 | 3780 reg_startp = reg_match->startp; |
3781 reg_endp = reg_match->endp; | |
3782 } | |
3783 | |
3784 /* Be paranoid... */ | |
3785 if (prog == NULL || line == NULL) | |
3786 { | |
3787 EMSG(_(e_null)); | |
3788 goto theend; | |
3789 } | |
3790 | |
3791 /* Check validity of program. */ | |
3792 if (prog_magic_wrong()) | |
3793 goto theend; | |
3794 | |
410 | 3795 /* If the start column is past the maximum column: no need to try. */ |
3796 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3797 goto theend; | |
3798 | |
7 | 3799 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */ |
3800 if (prog->regflags & RF_ICASE) | |
3801 ireg_ic = TRUE; | |
3802 else if (prog->regflags & RF_NOICASE) | |
3803 ireg_ic = FALSE; | |
3804 | |
3805 #ifdef FEAT_MBYTE | |
3806 /* If pattern contains "\Z" overrule value of ireg_icombine */ | |
3807 if (prog->regflags & RF_ICOMBINE) | |
3808 ireg_icombine = TRUE; | |
3809 #endif | |
3810 | |
3811 /* If there is a "must appear" string, look for it. */ | |
3812 if (prog->regmust != NULL) | |
3813 { | |
3814 int c; | |
3815 | |
3816 #ifdef FEAT_MBYTE | |
3817 if (has_mbyte) | |
3818 c = (*mb_ptr2char)(prog->regmust); | |
3819 else | |
3820 #endif | |
3821 c = *prog->regmust; | |
3822 s = line + col; | |
170 | 3823 |
3824 /* | |
3825 * This is used very often, esp. for ":global". Use three versions of | |
3826 * the loop to avoid overhead of conditions. | |
3827 */ | |
3828 if (!ireg_ic | |
3829 #ifdef FEAT_MBYTE | |
3830 && !has_mbyte | |
3831 #endif | |
3832 ) | |
3833 while ((s = vim_strbyte(s, c)) != NULL) | |
3834 { | |
3835 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3836 break; /* Found it. */ | |
3837 ++s; | |
3838 } | |
3839 #ifdef FEAT_MBYTE | |
3840 else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1)) | |
3841 while ((s = vim_strchr(s, c)) != NULL) | |
3842 { | |
3843 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3844 break; /* Found it. */ | |
3845 mb_ptr_adv(s); | |
3846 } | |
3847 #endif | |
3848 else | |
3849 while ((s = cstrchr(s, c)) != NULL) | |
3850 { | |
3851 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3852 break; /* Found it. */ | |
3853 mb_ptr_adv(s); | |
3854 } | |
7 | 3855 if (s == NULL) /* Not present. */ |
3856 goto theend; | |
3857 } | |
3858 | |
3859 regline = line; | |
3860 reglnum = 0; | |
2578 | 3861 reg_toolong = FALSE; |
7 | 3862 |
3863 /* Simplest case: Anchored match need be tried only once. */ | |
3864 if (prog->reganch) | |
3865 { | |
3866 int c; | |
3867 | |
3868 #ifdef FEAT_MBYTE | |
3869 if (has_mbyte) | |
3870 c = (*mb_ptr2char)(regline + col); | |
3871 else | |
3872 #endif | |
3873 c = regline[col]; | |
3874 if (prog->regstart == NUL | |
3875 || prog->regstart == c | |
3876 || (ireg_ic && (( | |
3877 #ifdef FEAT_MBYTE | |
3878 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c))) | |
3879 || (c < 255 && prog->regstart < 255 && | |
3880 #endif | |
1347 | 3881 MB_TOLOWER(prog->regstart) == MB_TOLOWER(c))))) |
7 | 3882 retval = regtry(prog, col); |
3883 else | |
3884 retval = 0; | |
3885 } | |
3886 else | |
3887 { | |
1521 | 3888 #ifdef FEAT_RELTIME |
3889 int tm_count = 0; | |
3890 #endif | |
7 | 3891 /* Messy cases: unanchored match. */ |
180 | 3892 while (!got_int) |
7 | 3893 { |
3894 if (prog->regstart != NUL) | |
3895 { | |
170 | 3896 /* Skip until the char we know it must start with. |
3897 * Used often, do some work to avoid call overhead. */ | |
3898 if (!ireg_ic | |
3899 #ifdef FEAT_MBYTE | |
3900 && !has_mbyte | |
3901 #endif | |
3902 ) | |
3903 s = vim_strbyte(regline + col, prog->regstart); | |
3904 else | |
3905 s = cstrchr(regline + col, prog->regstart); | |
7 | 3906 if (s == NULL) |
3907 { | |
3908 retval = 0; | |
3909 break; | |
3910 } | |
3911 col = (int)(s - regline); | |
3912 } | |
3913 | |
410 | 3914 /* Check for maximum column to try. */ |
3915 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3916 { | |
3917 retval = 0; | |
3918 break; | |
3919 } | |
3920 | |
7 | 3921 retval = regtry(prog, col); |
3922 if (retval > 0) | |
3923 break; | |
3924 | |
3925 /* if not currently on the first line, get it again */ | |
3926 if (reglnum != 0) | |
3927 { | |
481 | 3928 reglnum = 0; |
7 | 3929 regline = reg_getline((linenr_T)0); |
3930 } | |
3931 if (regline[col] == NUL) | |
3932 break; | |
3933 #ifdef FEAT_MBYTE | |
3934 if (has_mbyte) | |
474 | 3935 col += (*mb_ptr2len)(regline + col); |
7 | 3936 else |
3937 #endif | |
3938 ++col; | |
1521 | 3939 #ifdef FEAT_RELTIME |
3940 /* Check for timeout once in a twenty times to avoid overhead. */ | |
3941 if (tm != NULL && ++tm_count == 20) | |
3942 { | |
3943 tm_count = 0; | |
3944 if (profile_passed_limit(tm)) | |
3945 break; | |
3946 } | |
3947 #endif | |
7 | 3948 } |
3949 } | |
3950 | |
3951 theend: | |
1520 | 3952 /* Free "reg_tofree" when it's a bit big. |
3953 * Free regstack and backpos if they are bigger than their initial size. */ | |
3954 if (reg_tofreelen > 400) | |
3955 { | |
3956 vim_free(reg_tofree); | |
3957 reg_tofree = NULL; | |
3958 } | |
3959 if (regstack.ga_maxlen > REGSTACK_INITIAL) | |
3960 ga_clear(®stack); | |
3961 if (backpos.ga_maxlen > BACKPOS_INITIAL) | |
3962 ga_clear(&backpos); | |
270 | 3963 |
7 | 3964 return retval; |
3965 } | |
3966 | |
3967 #ifdef FEAT_SYN_HL | |
3968 static reg_extmatch_T *make_extmatch __ARGS((void)); | |
3969 | |
3970 /* | |
3971 * Create a new extmatch and mark it as referenced once. | |
3972 */ | |
3973 static reg_extmatch_T * | |
3974 make_extmatch() | |
3975 { | |
3976 reg_extmatch_T *em; | |
3977 | |
3978 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T)); | |
3979 if (em != NULL) | |
3980 em->refcnt = 1; | |
3981 return em; | |
3982 } | |
3983 | |
3984 /* | |
3985 * Add a reference to an extmatch. | |
3986 */ | |
3987 reg_extmatch_T * | |
3988 ref_extmatch(em) | |
3989 reg_extmatch_T *em; | |
3990 { | |
3991 if (em != NULL) | |
3992 em->refcnt++; | |
3993 return em; | |
3994 } | |
3995 | |
3996 /* | |
3997 * Remove a reference to an extmatch. If there are no references left, free | |
3998 * the info. | |
3999 */ | |
4000 void | |
4001 unref_extmatch(em) | |
4002 reg_extmatch_T *em; | |
4003 { | |
4004 int i; | |
4005 | |
4006 if (em != NULL && --em->refcnt <= 0) | |
4007 { | |
4008 for (i = 0; i < NSUBEXP; ++i) | |
4009 vim_free(em->matches[i]); | |
4010 vim_free(em); | |
4011 } | |
4012 } | |
4013 #endif | |
4014 | |
4015 /* | |
4016 * regtry - try match of "prog" with at regline["col"]. | |
4017 * Returns 0 for failure, number of lines contained in the match otherwise. | |
4018 */ | |
4019 static long | |
4020 regtry(prog, col) | |
4444 | 4021 bt_regprog_T *prog; |
7 | 4022 colnr_T col; |
4023 { | |
4024 reginput = regline + col; | |
4025 need_clear_subexpr = TRUE; | |
4026 #ifdef FEAT_SYN_HL | |
4027 /* Clear the external match subpointers if necessary. */ | |
4028 if (prog->reghasz == REX_SET) | |
4029 need_clear_zsubexpr = TRUE; | |
4030 #endif | |
4031 | |
189 | 4032 if (regmatch(prog->program + 1) == 0) |
4033 return 0; | |
4034 | |
4035 cleanup_subexpr(); | |
4036 if (REG_MULTI) | |
7 | 4037 { |
189 | 4038 if (reg_startpos[0].lnum < 0) |
7 | 4039 { |
189 | 4040 reg_startpos[0].lnum = 0; |
4041 reg_startpos[0].col = col; | |
4042 } | |
4043 if (reg_endpos[0].lnum < 0) | |
4044 { | |
4045 reg_endpos[0].lnum = reglnum; | |
4046 reg_endpos[0].col = (int)(reginput - regline); | |
7 | 4047 } |
4048 else | |
189 | 4049 /* Use line number of "\ze". */ |
4050 reglnum = reg_endpos[0].lnum; | |
4051 } | |
4052 else | |
4053 { | |
4054 if (reg_startp[0] == NULL) | |
4055 reg_startp[0] = regline + col; | |
4056 if (reg_endp[0] == NULL) | |
4057 reg_endp[0] = reginput; | |
4058 } | |
7 | 4059 #ifdef FEAT_SYN_HL |
189 | 4060 /* Package any found \z(...\) matches for export. Default is none. */ |
4061 unref_extmatch(re_extmatch_out); | |
4062 re_extmatch_out = NULL; | |
4063 | |
4064 if (prog->reghasz == REX_SET) | |
4065 { | |
4066 int i; | |
4067 | |
4068 cleanup_zsubexpr(); | |
4069 re_extmatch_out = make_extmatch(); | |
4070 for (i = 0; i < NSUBEXP; i++) | |
7 | 4071 { |
189 | 4072 if (REG_MULTI) |
7 | 4073 { |
189 | 4074 /* Only accept single line matches. */ |
4075 if (reg_startzpos[i].lnum >= 0 | |
4076 && reg_endzpos[i].lnum == reg_startzpos[i].lnum) | |
4077 re_extmatch_out->matches[i] = | |
4078 vim_strnsave(reg_getline(reg_startzpos[i].lnum) | |
7 | 4079 + reg_startzpos[i].col, |
189 | 4080 reg_endzpos[i].col - reg_startzpos[i].col); |
4081 } | |
4082 else | |
4083 { | |
4084 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL) | |
4085 re_extmatch_out->matches[i] = | |
7 | 4086 vim_strnsave(reg_startzp[i], |
189 | 4087 (int)(reg_endzp[i] - reg_startzp[i])); |
7 | 4088 } |
4089 } | |
189 | 4090 } |
7 | 4091 #endif |
189 | 4092 return 1 + reglnum; |
7 | 4093 } |
4094 | |
4095 #ifdef FEAT_MBYTE | |
4096 static int reg_prev_class __ARGS((void)); | |
4097 | |
4098 /* | |
4099 * Get class of previous character. | |
4100 */ | |
4101 static int | |
4102 reg_prev_class() | |
4103 { | |
4104 if (reginput > regline) | |
4069 | 4105 return mb_get_class_buf(reginput - 1 |
4106 - (*mb_head_off)(regline, reginput - 1), reg_buf); | |
7 | 4107 return -1; |
4108 } | |
4109 | |
4110 #endif | |
39 | 4111 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput) |
7 | 4112 |
4113 /* | |
4114 * The arguments from BRACE_LIMITS are stored here. They are actually local | |
4115 * to regmatch(), but they are here to reduce the amount of stack space used | |
4116 * (it can be called recursively many times). | |
4117 */ | |
4118 static long bl_minval; | |
4119 static long bl_maxval; | |
4120 | |
4121 /* | |
4122 * regmatch - main matching routine | |
4123 * | |
180 | 4124 * Conceptually the strategy is simple: Check to see whether the current node |
4125 * matches, push an item onto the regstack and loop to see whether the rest | |
4126 * matches, and then act accordingly. In practice we make some effort to | |
4127 * avoid using the regstack, in particular by going through "ordinary" nodes | |
4128 * (that don't need to know whether the rest of the match failed) by a nested | |
4129 * loop. | |
7 | 4130 * |
4131 * Returns TRUE when there is a match. Leaves reginput and reglnum just after | |
4132 * the last matched character. | |
4133 * Returns FALSE when there is no match. Leaves reginput and reglnum in an | |
4134 * undefined state! | |
4135 */ | |
4136 static int | |
180 | 4137 regmatch(scan) |
7 | 4138 char_u *scan; /* Current node. */ |
4139 { | |
180 | 4140 char_u *next; /* Next node. */ |
4141 int op; | |
4142 int c; | |
4143 regitem_T *rp; | |
4144 int no; | |
4145 int status; /* one of the RA_ values: */ | |
4146 #define RA_FAIL 1 /* something failed, abort */ | |
4147 #define RA_CONT 2 /* continue in inner loop */ | |
4148 #define RA_BREAK 3 /* break inner loop */ | |
4149 #define RA_MATCH 4 /* successful match */ | |
4150 #define RA_NOMATCH 5 /* didn't match */ | |
270 | 4151 |
1520 | 4152 /* Make "regstack" and "backpos" empty. They are allocated and freed in |
4444 | 4153 * bt_regexec_both() to reduce malloc()/free() calls. */ |
270 | 4154 regstack.ga_len = 0; |
4155 backpos.ga_len = 0; | |
233 | 4156 |
180 | 4157 /* |
233 | 4158 * Repeat until "regstack" is empty. |
180 | 4159 */ |
4160 for (;;) | |
4161 { | |
4444 | 4162 /* Some patterns may cause a long time to match, even though they are not |
7 | 4163 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */ |
4164 fast_breakcheck(); | |
4165 | |
4166 #ifdef DEBUG | |
4167 if (scan != NULL && regnarrate) | |
4168 { | |
4444 | 4169 mch_errmsg((char *)regprop(scan)); |
7 | 4170 mch_errmsg("(\n"); |
4171 } | |
4172 #endif | |
180 | 4173 |
4174 /* | |
233 | 4175 * Repeat for items that can be matched sequentially, without using the |
180 | 4176 * regstack. |
4177 */ | |
4178 for (;;) | |
7 | 4179 { |
180 | 4180 if (got_int || scan == NULL) |
4181 { | |
4182 status = RA_FAIL; | |
4183 break; | |
4184 } | |
4185 status = RA_CONT; | |
4186 | |
7 | 4187 #ifdef DEBUG |
4188 if (regnarrate) | |
4189 { | |
4444 | 4190 mch_errmsg((char *)regprop(scan)); |
7 | 4191 mch_errmsg("...\n"); |
4192 # ifdef FEAT_SYN_HL | |
4193 if (re_extmatch_in != NULL) | |
4194 { | |
4195 int i; | |
4196 | |
4197 mch_errmsg(_("External submatches:\n")); | |
4198 for (i = 0; i < NSUBEXP; i++) | |
4199 { | |
4200 mch_errmsg(" \""); | |
4201 if (re_extmatch_in->matches[i] != NULL) | |
4444 | 4202 mch_errmsg((char *)re_extmatch_in->matches[i]); |
7 | 4203 mch_errmsg("\"\n"); |
4204 } | |
4205 } | |
4206 # endif | |
4207 } | |
4208 #endif | |
4209 next = regnext(scan); | |
4210 | |
4211 op = OP(scan); | |
4212 /* Check for character class with NL added. */ | |
1018 | 4213 if (!reg_line_lbr && WITH_NL(op) && REG_MULTI |
4214 && *reginput == NUL && reglnum <= reg_maxline) | |
7 | 4215 { |
4216 reg_nextline(); | |
4217 } | |
4218 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n') | |
4219 { | |
4220 ADVANCE_REGINPUT(); | |
4221 } | |
4222 else | |
4223 { | |
4224 if (WITH_NL(op)) | |
180 | 4225 op -= ADD_NL; |
7 | 4226 #ifdef FEAT_MBYTE |
4227 if (has_mbyte) | |
4228 c = (*mb_ptr2char)(reginput); | |
4229 else | |
4230 #endif | |
4231 c = *reginput; | |
4232 switch (op) | |
4233 { | |
4234 case BOL: | |
4235 if (reginput != regline) | |
180 | 4236 status = RA_NOMATCH; |
7 | 4237 break; |
4238 | |
4239 case EOL: | |
4240 if (c != NUL) | |
180 | 4241 status = RA_NOMATCH; |
7 | 4242 break; |
4243 | |
4244 case RE_BOF: | |
1458 | 4245 /* We're not at the beginning of the file when below the first |
4246 * line where we started, not at the start of the line or we | |
4247 * didn't start at the first line of the buffer. */ | |
7 | 4248 if (reglnum != 0 || reginput != regline |
1458 | 4249 || (REG_MULTI && reg_firstlnum > 1)) |
180 | 4250 status = RA_NOMATCH; |
7 | 4251 break; |
4252 | |
4253 case RE_EOF: | |
4254 if (reglnum != reg_maxline || c != NUL) | |
180 | 4255 status = RA_NOMATCH; |
7 | 4256 break; |
4257 | |
4258 case CURSOR: | |
4259 /* Check if the buffer is in a window and compare the | |
4260 * reg_win->w_cursor position to the match position. */ | |
4261 if (reg_win == NULL | |
4262 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum) | |
4263 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col)) | |
180 | 4264 status = RA_NOMATCH; |
7 | 4265 break; |
4266 | |
639 | 4267 case RE_MARK: |
4268 /* Compare the mark position to the match position. NOTE: Always | |
4269 * uses the current buffer. */ | |
4270 { | |
4271 int mark = OPERAND(scan)[0]; | |
4272 int cmp = OPERAND(scan)[1]; | |
4273 pos_T *pos; | |
4274 | |
4043 | 4275 pos = getmark_buf(reg_buf, mark, FALSE); |
1148 | 4276 if (pos == NULL /* mark doesn't exist */ |
639 | 4277 || pos->lnum <= 0 /* mark isn't set (in curbuf) */ |
4278 || (pos->lnum == reglnum + reg_firstlnum | |
4279 ? (pos->col == (colnr_T)(reginput - regline) | |
4280 ? (cmp == '<' || cmp == '>') | |
4281 : (pos->col < (colnr_T)(reginput - regline) | |
4282 ? cmp != '>' | |
4283 : cmp != '<')) | |
4284 : (pos->lnum < reglnum + reg_firstlnum | |
4285 ? cmp != '>' | |
4286 : cmp != '<'))) | |
4287 status = RA_NOMATCH; | |
4288 } | |
4289 break; | |
4290 | |
4291 case RE_VISUAL: | |
4292 #ifdef FEAT_VISUAL | |
4293 /* Check if the buffer is the current buffer. and whether the | |
4294 * position is inside the Visual area. */ | |
4295 if (reg_buf != curbuf || VIsual.lnum == 0) | |
4296 status = RA_NOMATCH; | |
4297 else | |
4298 { | |
4299 pos_T top, bot; | |
4300 linenr_T lnum; | |
4301 colnr_T col; | |
4302 win_T *wp = reg_win == NULL ? curwin : reg_win; | |
4303 int mode; | |
4304 | |
4305 if (VIsual_active) | |
4306 { | |
4307 if (lt(VIsual, wp->w_cursor)) | |
4308 { | |
4309 top = VIsual; | |
4310 bot = wp->w_cursor; | |
4311 } | |
4312 else | |
4313 { | |
4314 top = wp->w_cursor; | |
4315 bot = VIsual; | |
4316 } | |
4317 mode = VIsual_mode; | |
4318 } | |
4319 else | |
4320 { | |
692 | 4321 if (lt(curbuf->b_visual.vi_start, curbuf->b_visual.vi_end)) |
685 | 4322 { |
692 | 4323 top = curbuf->b_visual.vi_start; |
4324 bot = curbuf->b_visual.vi_end; | |
685 | 4325 } |
4326 else | |
4327 { | |
692 | 4328 top = curbuf->b_visual.vi_end; |
4329 bot = curbuf->b_visual.vi_start; | |
685 | 4330 } |
692 | 4331 mode = curbuf->b_visual.vi_mode; |
639 | 4332 } |
4333 lnum = reglnum + reg_firstlnum; | |
4334 col = (colnr_T)(reginput - regline); | |
4335 if (lnum < top.lnum || lnum > bot.lnum) | |
4336 status = RA_NOMATCH; | |
4337 else if (mode == 'v') | |
4338 { | |
4339 if ((lnum == top.lnum && col < top.col) | |
4340 || (lnum == bot.lnum | |
4341 && col >= bot.col + (*p_sel != 'e'))) | |
4342 status = RA_NOMATCH; | |
4343 } | |
4344 else if (mode == Ctrl_V) | |
4345 { | |
4346 colnr_T start, end; | |
4347 colnr_T start2, end2; | |
944 | 4348 colnr_T cols; |
639 | 4349 |
4350 getvvcol(wp, &top, &start, NULL, &end); | |
4351 getvvcol(wp, &bot, &start2, NULL, &end2); | |
4352 if (start2 < start) | |
4353 start = start2; | |
4354 if (end2 > end) | |
4355 end = end2; | |
4356 if (top.col == MAXCOL || bot.col == MAXCOL) | |
4357 end = MAXCOL; | |
944 | 4358 cols = win_linetabsize(wp, |
639 | 4359 regline, (colnr_T)(reginput - regline)); |
944 | 4360 if (cols < start || cols > end - (*p_sel == 'e')) |
639 | 4361 status = RA_NOMATCH; |
4362 } | |
4363 } | |
4364 #else | |
4365 status = RA_NOMATCH; | |
4366 #endif | |
4367 break; | |
4368 | |
7 | 4369 case RE_LNUM: |
4370 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum), | |
4371 scan)) | |
180 | 4372 status = RA_NOMATCH; |
7 | 4373 break; |
4374 | |
4375 case RE_COL: | |
4376 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan)) | |
180 | 4377 status = RA_NOMATCH; |
7 | 4378 break; |
4379 | |
4380 case RE_VCOL: | |
4381 if (!re_num_cmp((long_u)win_linetabsize( | |
4382 reg_win == NULL ? curwin : reg_win, | |
4383 regline, (colnr_T)(reginput - regline)) + 1, scan)) | |
180 | 4384 status = RA_NOMATCH; |
7 | 4385 break; |
4386 | |
4387 case BOW: /* \<word; reginput points to w */ | |
4388 if (c == NUL) /* Can't match at end of line */ | |
180 | 4389 status = RA_NOMATCH; |
7 | 4390 #ifdef FEAT_MBYTE |
180 | 4391 else if (has_mbyte) |
7 | 4392 { |
4393 int this_class; | |
4394 | |
4395 /* Get class of current and previous char (if it exists). */ | |
4069 | 4396 this_class = mb_get_class_buf(reginput, reg_buf); |
7 | 4397 if (this_class <= 1) |
180 | 4398 status = RA_NOMATCH; /* not on a word at all */ |
4399 else if (reg_prev_class() == this_class) | |
4400 status = RA_NOMATCH; /* previous char is in same word */ | |
7 | 4401 } |
4402 #endif | |
4403 else | |
4404 { | |
4061 | 4405 if (!vim_iswordc_buf(c, reg_buf) || (reginput > regline |
4406 && vim_iswordc_buf(reginput[-1], reg_buf))) | |
180 | 4407 status = RA_NOMATCH; |
7 | 4408 } |
4409 break; | |
4410 | |
4411 case EOW: /* word\>; reginput points after d */ | |
4412 if (reginput == regline) /* Can't match at start of line */ | |
180 | 4413 status = RA_NOMATCH; |
7 | 4414 #ifdef FEAT_MBYTE |
180 | 4415 else if (has_mbyte) |
7 | 4416 { |
4417 int this_class, prev_class; | |
4418 | |
4419 /* Get class of current and previous char (if it exists). */ | |
4069 | 4420 this_class = mb_get_class_buf(reginput, reg_buf); |
7 | 4421 prev_class = reg_prev_class(); |
180 | 4422 if (this_class == prev_class |
4423 || prev_class == 0 || prev_class == 1) | |
4424 status = RA_NOMATCH; | |
7 | 4425 } |
180 | 4426 #endif |
7 | 4427 else |
4428 { | |
4043 | 4429 if (!vim_iswordc_buf(reginput[-1], reg_buf) |
4430 || (reginput[0] != NUL && vim_iswordc_buf(c, reg_buf))) | |
180 | 4431 status = RA_NOMATCH; |
7 | 4432 } |
4433 break; /* Matched with EOW */ | |
4434 | |
4435 case ANY: | |
4084 | 4436 /* ANY does not match new lines. */ |
7 | 4437 if (c == NUL) |
180 | 4438 status = RA_NOMATCH; |
4439 else | |
4440 ADVANCE_REGINPUT(); | |
7 | 4441 break; |
4442 | |
4443 case IDENT: | |
4444 if (!vim_isIDc(c)) | |
180 | 4445 status = RA_NOMATCH; |
4446 else | |
4447 ADVANCE_REGINPUT(); | |
7 | 4448 break; |
4449 | |
4450 case SIDENT: | |
4451 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c)) | |
180 | 4452 status = RA_NOMATCH; |
4453 else | |
4454 ADVANCE_REGINPUT(); | |
7 | 4455 break; |
4456 | |
4457 case KWORD: | |
4069 | 4458 if (!vim_iswordp_buf(reginput, reg_buf)) |
180 | 4459 status = RA_NOMATCH; |
4460 else | |
4461 ADVANCE_REGINPUT(); | |
7 | 4462 break; |
4463 | |
4464 case SKWORD: | |
4069 | 4465 if (VIM_ISDIGIT(*reginput) || !vim_iswordp_buf(reginput, reg_buf)) |
180 | 4466 status = RA_NOMATCH; |
4467 else | |
4468 ADVANCE_REGINPUT(); | |
7 | 4469 break; |
4470 | |
4471 case FNAME: | |
4472 if (!vim_isfilec(c)) | |
180 | 4473 status = RA_NOMATCH; |
4474 else | |
4475 ADVANCE_REGINPUT(); | |
7 | 4476 break; |
4477 | |
4478 case SFNAME: | |
4479 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c)) | |
180 | 4480 status = RA_NOMATCH; |
4481 else | |
4482 ADVANCE_REGINPUT(); | |
7 | 4483 break; |
4484 | |
4485 case PRINT: | |
4486 if (ptr2cells(reginput) != 1) | |
180 | 4487 status = RA_NOMATCH; |
4488 else | |
4489 ADVANCE_REGINPUT(); | |
7 | 4490 break; |
4491 | |
4492 case SPRINT: | |
4493 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1) | |
180 | 4494 status = RA_NOMATCH; |
4495 else | |
4496 ADVANCE_REGINPUT(); | |
7 | 4497 break; |
4498 | |
4499 case WHITE: | |
4500 if (!vim_iswhite(c)) | |
180 | 4501 status = RA_NOMATCH; |
4502 else | |
4503 ADVANCE_REGINPUT(); | |
7 | 4504 break; |
4505 | |
4506 case NWHITE: | |
4507 if (c == NUL || vim_iswhite(c)) | |
180 | 4508 status = RA_NOMATCH; |
4509 else | |
4510 ADVANCE_REGINPUT(); | |
7 | 4511 break; |
4512 | |
4513 case DIGIT: | |
4514 if (!ri_digit(c)) | |
180 | 4515 status = RA_NOMATCH; |
4516 else | |
4517 ADVANCE_REGINPUT(); | |
7 | 4518 break; |
4519 | |
4520 case NDIGIT: | |
4521 if (c == NUL || ri_digit(c)) | |
180 | 4522 status = RA_NOMATCH; |
4523 else | |
4524 ADVANCE_REGINPUT(); | |
7 | 4525 break; |
4526 | |
4527 case HEX: | |
4528 if (!ri_hex(c)) | |
180 | 4529 status = RA_NOMATCH; |
4530 else | |
4531 ADVANCE_REGINPUT(); | |
7 | 4532 break; |
4533 | |
4534 case NHEX: | |
4535 if (c == NUL || ri_hex(c)) | |
180 | 4536 status = RA_NOMATCH; |
4537 else | |
4538 ADVANCE_REGINPUT(); | |
7 | 4539 break; |
4540 | |
4541 case OCTAL: | |
4542 if (!ri_octal(c)) | |
180 | 4543 status = RA_NOMATCH; |
4544 else | |
4545 ADVANCE_REGINPUT(); | |
7 | 4546 break; |
4547 | |
4548 case NOCTAL: | |
4549 if (c == NUL || ri_octal(c)) | |
180 | 4550 status = RA_NOMATCH; |
4551 else | |
4552 ADVANCE_REGINPUT(); | |
7 | 4553 break; |
4554 | |
4555 case WORD: | |
4556 if (!ri_word(c)) | |
180 | 4557 status = RA_NOMATCH; |
4558 else | |
4559 ADVANCE_REGINPUT(); | |
7 | 4560 break; |
4561 | |
4562 case NWORD: | |
4563 if (c == NUL || ri_word(c)) | |
180 | 4564 status = RA_NOMATCH; |
4565 else | |
4566 ADVANCE_REGINPUT(); | |
7 | 4567 break; |
4568 | |
4569 case HEAD: | |
4570 if (!ri_head(c)) | |
180 | 4571 status = RA_NOMATCH; |
4572 else | |
4573 ADVANCE_REGINPUT(); | |
7 | 4574 break; |
4575 | |
4576 case NHEAD: | |
4577 if (c == NUL || ri_head(c)) | |
180 | 4578 status = RA_NOMATCH; |
4579 else | |
4580 ADVANCE_REGINPUT(); | |
7 | 4581 break; |
4582 | |
4583 case ALPHA: | |
4584 if (!ri_alpha(c)) | |
180 | 4585 status = RA_NOMATCH; |
4586 else | |
4587 ADVANCE_REGINPUT(); | |
7 | 4588 break; |
4589 | |
4590 case NALPHA: | |
4591 if (c == NUL || ri_alpha(c)) | |
180 | 4592 status = RA_NOMATCH; |
4593 else | |
4594 ADVANCE_REGINPUT(); | |
7 | 4595 break; |
4596 | |
4597 case LOWER: | |
4598 if (!ri_lower(c)) | |
180 | 4599 status = RA_NOMATCH; |
4600 else | |
4601 ADVANCE_REGINPUT(); | |
7 | 4602 break; |
4603 | |
4604 case NLOWER: | |
4605 if (c == NUL || ri_lower(c)) | |
180 | 4606 status = RA_NOMATCH; |
4607 else | |
4608 ADVANCE_REGINPUT(); | |
7 | 4609 break; |
4610 | |
4611 case UPPER: | |
4612 if (!ri_upper(c)) | |
180 | 4613 status = RA_NOMATCH; |
4614 else | |
4615 ADVANCE_REGINPUT(); | |
7 | 4616 break; |
4617 | |
4618 case NUPPER: | |
4619 if (c == NUL || ri_upper(c)) | |
180 | 4620 status = RA_NOMATCH; |
4621 else | |
4622 ADVANCE_REGINPUT(); | |
7 | 4623 break; |
4624 | |
4625 case EXACTLY: | |
4626 { | |
4627 int len; | |
4628 char_u *opnd; | |
4629 | |
4630 opnd = OPERAND(scan); | |
4631 /* Inline the first byte, for speed. */ | |
4632 if (*opnd != *reginput | |
4633 && (!ireg_ic || ( | |
4634 #ifdef FEAT_MBYTE | |
4635 !enc_utf8 && | |
4636 #endif | |
1347 | 4637 MB_TOLOWER(*opnd) != MB_TOLOWER(*reginput)))) |
180 | 4638 status = RA_NOMATCH; |
4639 else if (*opnd == NUL) | |
7 | 4640 { |
4641 /* match empty string always works; happens when "~" is | |
4642 * empty. */ | |
4643 } | |
4644 else if (opnd[1] == NUL | |
4645 #ifdef FEAT_MBYTE | |
4646 && !(enc_utf8 && ireg_ic) | |
4647 #endif | |
4648 ) | |
4649 ++reginput; /* matched a single char */ | |
4650 else | |
4651 { | |
4652 len = (int)STRLEN(opnd); | |
4653 /* Need to match first byte again for multi-byte. */ | |
4654 if (cstrncmp(opnd, reginput, &len) != 0) | |
180 | 4655 status = RA_NOMATCH; |
7 | 4656 #ifdef FEAT_MBYTE |
4657 /* Check for following composing character. */ | |
180 | 4658 else if (enc_utf8 |
4659 && UTF_COMPOSINGLIKE(reginput, reginput + len)) | |
7 | 4660 { |
4661 /* raaron: This code makes a composing character get | |
4662 * ignored, which is the correct behavior (sometimes) | |
4663 * for voweled Hebrew texts. */ | |
4664 if (!ireg_icombine) | |
180 | 4665 status = RA_NOMATCH; |
7 | 4666 } |
180 | 4667 #endif |
7 | 4668 else |
180 | 4669 reginput += len; |
7 | 4670 } |
4671 } | |
4672 break; | |
4673 | |
4674 case ANYOF: | |
4675 case ANYBUT: | |
4676 if (c == NUL) | |
180 | 4677 status = RA_NOMATCH; |
4678 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF)) | |
4679 status = RA_NOMATCH; | |
4680 else | |
4681 ADVANCE_REGINPUT(); | |
7 | 4682 break; |
4683 | |
4684 #ifdef FEAT_MBYTE | |
4685 case MULTIBYTECODE: | |
4686 if (has_mbyte) | |
4687 { | |
4688 int i, len; | |
4689 char_u *opnd; | |
944 | 4690 int opndc = 0, inpc; |
7 | 4691 |
4692 opnd = OPERAND(scan); | |
4693 /* Safety check (just in case 'encoding' was changed since | |
4694 * compiling the program). */ | |
474 | 4695 if ((len = (*mb_ptr2len)(opnd)) < 2) |
180 | 4696 { |
4697 status = RA_NOMATCH; | |
4698 break; | |
4699 } | |
714 | 4700 if (enc_utf8) |
4701 opndc = mb_ptr2char(opnd); | |
4702 if (enc_utf8 && utf_iscomposing(opndc)) | |
4703 { | |
4704 /* When only a composing char is given match at any | |
4705 * position where that composing char appears. */ | |
4706 status = RA_NOMATCH; | |
4707 for (i = 0; reginput[i] != NUL; i += utf_char2len(inpc)) | |
180 | 4708 { |
714 | 4709 inpc = mb_ptr2char(reginput + i); |
4710 if (!utf_iscomposing(inpc)) | |
4711 { | |
4712 if (i > 0) | |
4713 break; | |
4714 } | |
4715 else if (opndc == inpc) | |
4716 { | |
4717 /* Include all following composing chars. */ | |
4718 len = i + mb_ptr2len(reginput + i); | |
4719 status = RA_MATCH; | |
4720 break; | |
4721 } | |
180 | 4722 } |
714 | 4723 } |
4724 else | |
4725 for (i = 0; i < len; ++i) | |
4726 if (opnd[i] != reginput[i]) | |
4727 { | |
4728 status = RA_NOMATCH; | |
4729 break; | |
4730 } | |
7 | 4731 reginput += len; |
4732 } | |
4733 else | |
180 | 4734 status = RA_NOMATCH; |
7 | 4735 break; |
4736 #endif | |
4737 | |
4738 case NOTHING: | |
4739 break; | |
4740 | |
4741 case BACK: | |
233 | 4742 { |
4743 int i; | |
4744 backpos_T *bp; | |
4745 | |
4746 /* | |
4747 * When we run into BACK we need to check if we don't keep | |
4748 * looping without matching any input. The second and later | |
4749 * times a BACK is encountered it fails if the input is still | |
4750 * at the same position as the previous time. | |
4751 * The positions are stored in "backpos" and found by the | |
4752 * current value of "scan", the position in the RE program. | |
4753 */ | |
4754 bp = (backpos_T *)backpos.ga_data; | |
4755 for (i = 0; i < backpos.ga_len; ++i) | |
4756 if (bp[i].bp_scan == scan) | |
4757 break; | |
4758 if (i == backpos.ga_len) | |
4759 { | |
4760 /* First time at this BACK, make room to store the pos. */ | |
4761 if (ga_grow(&backpos, 1) == FAIL) | |
4762 status = RA_FAIL; | |
4763 else | |
4764 { | |
4765 /* get "ga_data" again, it may have changed */ | |
4766 bp = (backpos_T *)backpos.ga_data; | |
4767 bp[i].bp_scan = scan; | |
4768 ++backpos.ga_len; | |
4769 } | |
4770 } | |
4771 else if (reg_save_equal(&bp[i].bp_pos)) | |
4772 /* Still at same position as last time, fail. */ | |
4773 status = RA_NOMATCH; | |
4774 | |
4775 if (status != RA_FAIL && status != RA_NOMATCH) | |
4776 reg_save(&bp[i].bp_pos, &backpos); | |
4777 } | |
179 | 4778 break; |
4779 | |
7 | 4780 case MOPEN + 0: /* Match start: \zs */ |
4781 case MOPEN + 1: /* \( */ | |
4782 case MOPEN + 2: | |
4783 case MOPEN + 3: | |
4784 case MOPEN + 4: | |
4785 case MOPEN + 5: | |
4786 case MOPEN + 6: | |
4787 case MOPEN + 7: | |
4788 case MOPEN + 8: | |
4789 case MOPEN + 9: | |
4790 { | |
4791 no = op - MOPEN; | |
4792 cleanup_subexpr(); | |
270 | 4793 rp = regstack_push(RS_MOPEN, scan); |
180 | 4794 if (rp == NULL) |
4795 status = RA_FAIL; | |
4796 else | |
4797 { | |
4798 rp->rs_no = no; | |
4799 save_se(&rp->rs_un.sesave, ®_startpos[no], | |
4800 ®_startp[no]); | |
4801 /* We simply continue and handle the result when done. */ | |
4802 } | |
7 | 4803 } |
180 | 4804 break; |
7 | 4805 |
4806 case NOPEN: /* \%( */ | |
4807 case NCLOSE: /* \) after \%( */ | |
270 | 4808 if (regstack_push(RS_NOPEN, scan) == NULL) |
180 | 4809 status = RA_FAIL; |
4810 /* We simply continue and handle the result when done. */ | |
4811 break; | |
7 | 4812 |
4813 #ifdef FEAT_SYN_HL | |
4814 case ZOPEN + 1: | |
4815 case ZOPEN + 2: | |
4816 case ZOPEN + 3: | |
4817 case ZOPEN + 4: | |
4818 case ZOPEN + 5: | |
4819 case ZOPEN + 6: | |
4820 case ZOPEN + 7: | |
4821 case ZOPEN + 8: | |
4822 case ZOPEN + 9: | |
4823 { | |
4824 no = op - ZOPEN; | |
4825 cleanup_zsubexpr(); | |
270 | 4826 rp = regstack_push(RS_ZOPEN, scan); |
180 | 4827 if (rp == NULL) |
4828 status = RA_FAIL; | |
4829 else | |
4830 { | |
4831 rp->rs_no = no; | |
4832 save_se(&rp->rs_un.sesave, ®_startzpos[no], | |
4833 ®_startzp[no]); | |
4834 /* We simply continue and handle the result when done. */ | |
4835 } | |
7 | 4836 } |
180 | 4837 break; |
7 | 4838 #endif |
4839 | |
4840 case MCLOSE + 0: /* Match end: \ze */ | |
4841 case MCLOSE + 1: /* \) */ | |
4842 case MCLOSE + 2: | |
4843 case MCLOSE + 3: | |
4844 case MCLOSE + 4: | |
4845 case MCLOSE + 5: | |
4846 case MCLOSE + 6: | |
4847 case MCLOSE + 7: | |
4848 case MCLOSE + 8: | |
4849 case MCLOSE + 9: | |
4850 { | |
4851 no = op - MCLOSE; | |
4852 cleanup_subexpr(); | |
270 | 4853 rp = regstack_push(RS_MCLOSE, scan); |
180 | 4854 if (rp == NULL) |
4855 status = RA_FAIL; | |
4856 else | |
4857 { | |
4858 rp->rs_no = no; | |
4859 save_se(&rp->rs_un.sesave, ®_endpos[no], ®_endp[no]); | |
4860 /* We simply continue and handle the result when done. */ | |
4861 } | |
7 | 4862 } |
180 | 4863 break; |
7 | 4864 |
4865 #ifdef FEAT_SYN_HL | |
4866 case ZCLOSE + 1: /* \) after \z( */ | |
4867 case ZCLOSE + 2: | |
4868 case ZCLOSE + 3: | |
4869 case ZCLOSE + 4: | |
4870 case ZCLOSE + 5: | |
4871 case ZCLOSE + 6: | |
4872 case ZCLOSE + 7: | |
4873 case ZCLOSE + 8: | |
4874 case ZCLOSE + 9: | |
4875 { | |
4876 no = op - ZCLOSE; | |
4877 cleanup_zsubexpr(); | |
270 | 4878 rp = regstack_push(RS_ZCLOSE, scan); |
180 | 4879 if (rp == NULL) |
4880 status = RA_FAIL; | |
4881 else | |
4882 { | |
4883 rp->rs_no = no; | |
4884 save_se(&rp->rs_un.sesave, ®_endzpos[no], | |
4885 ®_endzp[no]); | |
4886 /* We simply continue and handle the result when done. */ | |
4887 } | |
7 | 4888 } |
180 | 4889 break; |
7 | 4890 #endif |
4891 | |
4892 case BACKREF + 1: | |
4893 case BACKREF + 2: | |
4894 case BACKREF + 3: | |
4895 case BACKREF + 4: | |
4896 case BACKREF + 5: | |
4897 case BACKREF + 6: | |
4898 case BACKREF + 7: | |
4899 case BACKREF + 8: | |
4900 case BACKREF + 9: | |
4901 { | |
4902 int len; | |
4903 linenr_T clnum; | |
4904 colnr_T ccol; | |
4905 char_u *p; | |
4906 | |
4907 no = op - BACKREF; | |
4908 cleanup_subexpr(); | |
4909 if (!REG_MULTI) /* Single-line regexp */ | |
4910 { | |
1815 | 4911 if (reg_startp[no] == NULL || reg_endp[no] == NULL) |
7 | 4912 { |
4913 /* Backref was not set: Match an empty string. */ | |
4914 len = 0; | |
4915 } | |
4916 else | |
4917 { | |
4918 /* Compare current input with back-ref in the same | |
4919 * line. */ | |
4920 len = (int)(reg_endp[no] - reg_startp[no]); | |
4921 if (cstrncmp(reg_startp[no], reginput, &len) != 0) | |
180 | 4922 status = RA_NOMATCH; |
7 | 4923 } |
4924 } | |
4925 else /* Multi-line regexp */ | |
4926 { | |
1815 | 4927 if (reg_startpos[no].lnum < 0 || reg_endpos[no].lnum < 0) |
7 | 4928 { |
4929 /* Backref was not set: Match an empty string. */ | |
4930 len = 0; | |
4931 } | |
4932 else | |
4933 { | |
4934 if (reg_startpos[no].lnum == reglnum | |
4935 && reg_endpos[no].lnum == reglnum) | |
4936 { | |
4937 /* Compare back-ref within the current line. */ | |
4938 len = reg_endpos[no].col - reg_startpos[no].col; | |
4939 if (cstrncmp(regline + reg_startpos[no].col, | |
4940 reginput, &len) != 0) | |
180 | 4941 status = RA_NOMATCH; |
7 | 4942 } |
4943 else | |
4944 { | |
4945 /* Messy situation: Need to compare between two | |
4946 * lines. */ | |
4947 ccol = reg_startpos[no].col; | |
4948 clnum = reg_startpos[no].lnum; | |
4949 for (;;) | |
4950 { | |
4951 /* Since getting one line may invalidate | |
4952 * the other, need to make copy. Slow! */ | |
4953 if (regline != reg_tofree) | |
4954 { | |
4955 len = (int)STRLEN(regline); | |
4956 if (reg_tofree == NULL | |
4957 || len >= (int)reg_tofreelen) | |
4958 { | |
4959 len += 50; /* get some extra */ | |
4960 vim_free(reg_tofree); | |
4961 reg_tofree = alloc(len); | |
4962 if (reg_tofree == NULL) | |
180 | 4963 { |
4964 status = RA_FAIL; /* outof memory!*/ | |
4965 break; | |
4966 } | |
7 | 4967 reg_tofreelen = len; |
4968 } | |
4969 STRCPY(reg_tofree, regline); | |
4970 reginput = reg_tofree | |
4971 + (reginput - regline); | |
4972 regline = reg_tofree; | |
4973 } | |
4974 | |
4975 /* Get the line to compare with. */ | |
4976 p = reg_getline(clnum); | |
4977 if (clnum == reg_endpos[no].lnum) | |
4978 len = reg_endpos[no].col - ccol; | |
4979 else | |
4980 len = (int)STRLEN(p + ccol); | |
4981 | |
4982 if (cstrncmp(p + ccol, reginput, &len) != 0) | |
180 | 4983 { |
4984 status = RA_NOMATCH; /* doesn't match */ | |
4985 break; | |
4986 } | |
7 | 4987 if (clnum == reg_endpos[no].lnum) |
4988 break; /* match and at end! */ | |
481 | 4989 if (reglnum >= reg_maxline) |
180 | 4990 { |
4991 status = RA_NOMATCH; /* text too short */ | |
4992 break; | |
4993 } | |
7 | 4994 |
4995 /* Advance to next line. */ | |
4996 reg_nextline(); | |
4997 ++clnum; | |
4998 ccol = 0; | |
180 | 4999 if (got_int) |
5000 { | |
5001 status = RA_FAIL; | |
5002 break; | |
5003 } | |
7 | 5004 } |
5005 | |
5006 /* found a match! Note that regline may now point | |
5007 * to a copy of the line, that should not matter. */ | |
5008 } | |
5009 } | |
5010 } | |
5011 | |
5012 /* Matched the backref, skip over it. */ | |
5013 reginput += len; | |
5014 } | |
5015 break; | |
5016 | |
5017 #ifdef FEAT_SYN_HL | |
5018 case ZREF + 1: | |
5019 case ZREF + 2: | |
5020 case ZREF + 3: | |
5021 case ZREF + 4: | |
5022 case ZREF + 5: | |
5023 case ZREF + 6: | |
5024 case ZREF + 7: | |
5025 case ZREF + 8: | |
5026 case ZREF + 9: | |
5027 { | |
5028 int len; | |
5029 | |
5030 cleanup_zsubexpr(); | |
5031 no = op - ZREF; | |
5032 if (re_extmatch_in != NULL | |
5033 && re_extmatch_in->matches[no] != NULL) | |
5034 { | |
5035 len = (int)STRLEN(re_extmatch_in->matches[no]); | |
5036 if (cstrncmp(re_extmatch_in->matches[no], | |
5037 reginput, &len) != 0) | |
180 | 5038 status = RA_NOMATCH; |
5039 else | |
5040 reginput += len; | |
7 | 5041 } |
5042 else | |
5043 { | |
5044 /* Backref was not set: Match an empty string. */ | |
5045 } | |
5046 } | |
5047 break; | |
5048 #endif | |
5049 | |
5050 case BRANCH: | |
5051 { | |
5052 if (OP(next) != BRANCH) /* No choice. */ | |
5053 next = OPERAND(scan); /* Avoid recursion. */ | |
5054 else | |
5055 { | |
270 | 5056 rp = regstack_push(RS_BRANCH, scan); |
180 | 5057 if (rp == NULL) |
5058 status = RA_FAIL; | |
5059 else | |
5060 status = RA_BREAK; /* rest is below */ | |
7 | 5061 } |
5062 } | |
5063 break; | |
5064 | |
5065 case BRACE_LIMITS: | |
5066 { | |
5067 if (OP(next) == BRACE_SIMPLE) | |
5068 { | |
5069 bl_minval = OPERAND_MIN(scan); | |
5070 bl_maxval = OPERAND_MAX(scan); | |
5071 } | |
5072 else if (OP(next) >= BRACE_COMPLEX | |
5073 && OP(next) < BRACE_COMPLEX + 10) | |
5074 { | |
5075 no = OP(next) - BRACE_COMPLEX; | |
5076 brace_min[no] = OPERAND_MIN(scan); | |
5077 brace_max[no] = OPERAND_MAX(scan); | |
5078 brace_count[no] = 0; | |
5079 } | |
5080 else | |
5081 { | |
5082 EMSG(_(e_internal)); /* Shouldn't happen */ | |
180 | 5083 status = RA_FAIL; |
7 | 5084 } |
5085 } | |
5086 break; | |
5087 | |
5088 case BRACE_COMPLEX + 0: | |
5089 case BRACE_COMPLEX + 1: | |
5090 case BRACE_COMPLEX + 2: | |
5091 case BRACE_COMPLEX + 3: | |
5092 case BRACE_COMPLEX + 4: | |
5093 case BRACE_COMPLEX + 5: | |
5094 case BRACE_COMPLEX + 6: | |
5095 case BRACE_COMPLEX + 7: | |
5096 case BRACE_COMPLEX + 8: | |
5097 case BRACE_COMPLEX + 9: | |
5098 { | |
5099 no = op - BRACE_COMPLEX; | |
5100 ++brace_count[no]; | |
5101 | |
5102 /* If not matched enough times yet, try one more */ | |
5103 if (brace_count[no] <= (brace_min[no] <= brace_max[no] | |
180 | 5104 ? brace_min[no] : brace_max[no])) |
7 | 5105 { |
270 | 5106 rp = regstack_push(RS_BRCPLX_MORE, scan); |
180 | 5107 if (rp == NULL) |
5108 status = RA_FAIL; | |
5109 else | |
5110 { | |
5111 rp->rs_no = no; | |
233 | 5112 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5113 next = OPERAND(scan); |
5114 /* We continue and handle the result when done. */ | |
5115 } | |
5116 break; | |
7 | 5117 } |
5118 | |
5119 /* If matched enough times, may try matching some more */ | |
5120 if (brace_min[no] <= brace_max[no]) | |
5121 { | |
5122 /* Range is the normal way around, use longest match */ | |
5123 if (brace_count[no] <= brace_max[no]) | |
5124 { | |
270 | 5125 rp = regstack_push(RS_BRCPLX_LONG, scan); |
180 | 5126 if (rp == NULL) |
5127 status = RA_FAIL; | |
5128 else | |
5129 { | |
5130 rp->rs_no = no; | |
233 | 5131 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5132 next = OPERAND(scan); |
5133 /* We continue and handle the result when done. */ | |
5134 } | |
7 | 5135 } |
5136 } | |
5137 else | |
5138 { | |
5139 /* Range is backwards, use shortest match first */ | |
5140 if (brace_count[no] <= brace_min[no]) | |
5141 { | |
270 | 5142 rp = regstack_push(RS_BRCPLX_SHORT, scan); |
180 | 5143 if (rp == NULL) |
5144 status = RA_FAIL; | |
5145 else | |
5146 { | |
233 | 5147 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5148 /* We continue and handle the result when done. */ |
5149 } | |
7 | 5150 } |
5151 } | |
5152 } | |
5153 break; | |
5154 | |
5155 case BRACE_SIMPLE: | |
5156 case STAR: | |
5157 case PLUS: | |
5158 { | |
180 | 5159 regstar_T rst; |
7 | 5160 |
5161 /* | |
5162 * Lookahead to avoid useless match attempts when we know | |
5163 * what character comes next. | |
5164 */ | |
5165 if (OP(next) == EXACTLY) | |
5166 { | |
180 | 5167 rst.nextb = *OPERAND(next); |
7 | 5168 if (ireg_ic) |
5169 { | |
1347 | 5170 if (MB_ISUPPER(rst.nextb)) |
5171 rst.nextb_ic = MB_TOLOWER(rst.nextb); | |
7 | 5172 else |
1347 | 5173 rst.nextb_ic = MB_TOUPPER(rst.nextb); |
7 | 5174 } |
5175 else | |
180 | 5176 rst.nextb_ic = rst.nextb; |
7 | 5177 } |
5178 else | |
5179 { | |
180 | 5180 rst.nextb = NUL; |
5181 rst.nextb_ic = NUL; | |
7 | 5182 } |
5183 if (op != BRACE_SIMPLE) | |
5184 { | |
180 | 5185 rst.minval = (op == STAR) ? 0 : 1; |
5186 rst.maxval = MAX_LIMIT; | |
7 | 5187 } |
5188 else | |
5189 { | |
180 | 5190 rst.minval = bl_minval; |
5191 rst.maxval = bl_maxval; | |
7 | 5192 } |
5193 | |
5194 /* | |
5195 * When maxval > minval, try matching as much as possible, up | |
5196 * to maxval. When maxval < minval, try matching at least the | |
5197 * minimal number (since the range is backwards, that's also | |
5198 * maxval!). | |
5199 */ | |
180 | 5200 rst.count = regrepeat(OPERAND(scan), rst.maxval); |
7 | 5201 if (got_int) |
180 | 5202 { |
5203 status = RA_FAIL; | |
5204 break; | |
5205 } | |
5206 if (rst.minval <= rst.maxval | |
5207 ? rst.count >= rst.minval : rst.count >= rst.maxval) | |
7 | 5208 { |
180 | 5209 /* It could match. Prepare for trying to match what |
5210 * follows. The code is below. Parameters are stored in | |
5211 * a regstar_T on the regstack. */ | |
212 | 5212 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5213 { |
5214 EMSG(_(e_maxmempat)); | |
5215 status = RA_FAIL; | |
5216 } | |
5217 else if (ga_grow(®stack, sizeof(regstar_T)) == FAIL) | |
180 | 5218 status = RA_FAIL; |
5219 else | |
7 | 5220 { |
180 | 5221 regstack.ga_len += sizeof(regstar_T); |
270 | 5222 rp = regstack_push(rst.minval <= rst.maxval |
233 | 5223 ? RS_STAR_LONG : RS_STAR_SHORT, scan); |
180 | 5224 if (rp == NULL) |
5225 status = RA_FAIL; | |
5226 else | |
7 | 5227 { |
180 | 5228 *(((regstar_T *)rp) - 1) = rst; |
5229 status = RA_BREAK; /* skip the restore bits */ | |
7 | 5230 } |
5231 } | |
5232 } | |
5233 else | |
180 | 5234 status = RA_NOMATCH; |
5235 | |
7 | 5236 } |
180 | 5237 break; |
7 | 5238 |
5239 case NOMATCH: | |
5240 case MATCH: | |
5241 case SUBPAT: | |
270 | 5242 rp = regstack_push(RS_NOMATCH, scan); |
180 | 5243 if (rp == NULL) |
5244 status = RA_FAIL; | |
5245 else | |
7 | 5246 { |
180 | 5247 rp->rs_no = op; |
233 | 5248 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5249 next = OPERAND(scan); |
5250 /* We continue and handle the result when done. */ | |
7 | 5251 } |
5252 break; | |
5253 | |
5254 case BEHIND: | |
5255 case NOBEHIND: | |
180 | 5256 /* Need a bit of room to store extra positions. */ |
212 | 5257 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5258 { |
5259 EMSG(_(e_maxmempat)); | |
5260 status = RA_FAIL; | |
5261 } | |
5262 else if (ga_grow(®stack, sizeof(regbehind_T)) == FAIL) | |
180 | 5263 status = RA_FAIL; |
5264 else | |
7 | 5265 { |
180 | 5266 regstack.ga_len += sizeof(regbehind_T); |
270 | 5267 rp = regstack_push(RS_BEHIND1, scan); |
180 | 5268 if (rp == NULL) |
5269 status = RA_FAIL; | |
5270 else | |
7 | 5271 { |
1579 | 5272 /* Need to save the subexpr to be able to restore them |
5273 * when there is a match but we don't use it. */ | |
5274 save_subexpr(((regbehind_T *)rp) - 1); | |
5275 | |
180 | 5276 rp->rs_no = op; |
233 | 5277 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5278 /* First try if what follows matches. If it does then we |
5279 * check the behind match by looping. */ | |
7 | 5280 } |
5281 } | |
180 | 5282 break; |
7 | 5283 |
5284 case BHPOS: | |
5285 if (REG_MULTI) | |
5286 { | |
5287 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline) | |
5288 || behind_pos.rs_u.pos.lnum != reglnum) | |
180 | 5289 status = RA_NOMATCH; |
7 | 5290 } |
5291 else if (behind_pos.rs_u.ptr != reginput) | |
180 | 5292 status = RA_NOMATCH; |
7 | 5293 break; |
5294 | |
5295 case NEWL: | |
1018 | 5296 if ((c != NUL || !REG_MULTI || reglnum > reg_maxline |
5297 || reg_line_lbr) && (c != '\n' || !reg_line_lbr)) | |
180 | 5298 status = RA_NOMATCH; |
5299 else if (reg_line_lbr) | |
7 | 5300 ADVANCE_REGINPUT(); |
5301 else | |
5302 reg_nextline(); | |
5303 break; | |
5304 | |
5305 case END: | |
180 | 5306 status = RA_MATCH; /* Success! */ |
5307 break; | |
7 | 5308 |
5309 default: | |
5310 EMSG(_(e_re_corr)); | |
5311 #ifdef DEBUG | |
5312 printf("Illegal op code %d\n", op); | |
5313 #endif | |
180 | 5314 status = RA_FAIL; |
5315 break; | |
7 | 5316 } |
5317 } | |
5318 | |
180 | 5319 /* If we can't continue sequentially, break the inner loop. */ |
5320 if (status != RA_CONT) | |
5321 break; | |
5322 | |
5323 /* Continue in inner loop, advance to next item. */ | |
7 | 5324 scan = next; |
180 | 5325 |
5326 } /* end of inner loop */ | |
7 | 5327 |
5328 /* | |
180 | 5329 * If there is something on the regstack execute the code for the state. |
233 | 5330 * If the state is popped then loop and use the older state. |
7 | 5331 */ |
180 | 5332 while (regstack.ga_len > 0 && status != RA_FAIL) |
5333 { | |
5334 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1; | |
5335 switch (rp->rs_state) | |
5336 { | |
5337 case RS_NOPEN: | |
5338 /* Result is passed on as-is, simply pop the state. */ | |
270 | 5339 regstack_pop(&scan); |
180 | 5340 break; |
5341 | |
5342 case RS_MOPEN: | |
5343 /* Pop the state. Restore pointers when there is no match. */ | |
5344 if (status == RA_NOMATCH) | |
5345 restore_se(&rp->rs_un.sesave, ®_startpos[rp->rs_no], | |
5346 ®_startp[rp->rs_no]); | |
270 | 5347 regstack_pop(&scan); |
180 | 5348 break; |
5349 | |
5350 #ifdef FEAT_SYN_HL | |
5351 case RS_ZOPEN: | |
5352 /* Pop the state. Restore pointers when there is no match. */ | |
5353 if (status == RA_NOMATCH) | |
5354 restore_se(&rp->rs_un.sesave, ®_startzpos[rp->rs_no], | |
5355 ®_startzp[rp->rs_no]); | |
270 | 5356 regstack_pop(&scan); |
180 | 5357 break; |
5358 #endif | |
5359 | |
5360 case RS_MCLOSE: | |
5361 /* Pop the state. Restore pointers when there is no match. */ | |
5362 if (status == RA_NOMATCH) | |
5363 restore_se(&rp->rs_un.sesave, ®_endpos[rp->rs_no], | |
5364 ®_endp[rp->rs_no]); | |
270 | 5365 regstack_pop(&scan); |
180 | 5366 break; |
5367 | |
5368 #ifdef FEAT_SYN_HL | |
5369 case RS_ZCLOSE: | |
5370 /* Pop the state. Restore pointers when there is no match. */ | |
5371 if (status == RA_NOMATCH) | |
5372 restore_se(&rp->rs_un.sesave, ®_endzpos[rp->rs_no], | |
5373 ®_endzp[rp->rs_no]); | |
270 | 5374 regstack_pop(&scan); |
180 | 5375 break; |
5376 #endif | |
5377 | |
5378 case RS_BRANCH: | |
5379 if (status == RA_MATCH) | |
5380 /* this branch matched, use it */ | |
270 | 5381 regstack_pop(&scan); |
180 | 5382 else |
5383 { | |
5384 if (status != RA_BREAK) | |
5385 { | |
5386 /* After a non-matching branch: try next one. */ | |
233 | 5387 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5388 scan = rp->rs_scan; |
5389 } | |
5390 if (scan == NULL || OP(scan) != BRANCH) | |
5391 { | |
5392 /* no more branches, didn't find a match */ | |
5393 status = RA_NOMATCH; | |
270 | 5394 regstack_pop(&scan); |
180 | 5395 } |
5396 else | |
5397 { | |
5398 /* Prepare to try a branch. */ | |
5399 rp->rs_scan = regnext(scan); | |
233 | 5400 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5401 scan = OPERAND(scan); |
5402 } | |
5403 } | |
5404 break; | |
5405 | |
5406 case RS_BRCPLX_MORE: | |
5407 /* Pop the state. Restore pointers when there is no match. */ | |
5408 if (status == RA_NOMATCH) | |
5409 { | |
233 | 5410 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5411 --brace_count[rp->rs_no]; /* decrement match count */ |
5412 } | |
270 | 5413 regstack_pop(&scan); |
180 | 5414 break; |
5415 | |
5416 case RS_BRCPLX_LONG: | |
5417 /* Pop the state. Restore pointers when there is no match. */ | |
5418 if (status == RA_NOMATCH) | |
5419 { | |
5420 /* There was no match, but we did find enough matches. */ | |
233 | 5421 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5422 --brace_count[rp->rs_no]; |
5423 /* continue with the items after "\{}" */ | |
5424 status = RA_CONT; | |
5425 } | |
270 | 5426 regstack_pop(&scan); |
180 | 5427 if (status == RA_CONT) |
5428 scan = regnext(scan); | |
5429 break; | |
5430 | |
5431 case RS_BRCPLX_SHORT: | |
5432 /* Pop the state. Restore pointers when there is no match. */ | |
5433 if (status == RA_NOMATCH) | |
5434 /* There was no match, try to match one more item. */ | |
233 | 5435 reg_restore(&rp->rs_un.regsave, &backpos); |
270 | 5436 regstack_pop(&scan); |
180 | 5437 if (status == RA_NOMATCH) |
5438 { | |
5439 scan = OPERAND(scan); | |
5440 status = RA_CONT; | |
5441 } | |
5442 break; | |
5443 | |
5444 case RS_NOMATCH: | |
5445 /* Pop the state. If the operand matches for NOMATCH or | |
5446 * doesn't match for MATCH/SUBPAT, we fail. Otherwise backup, | |
5447 * except for SUBPAT, and continue with the next item. */ | |
5448 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH)) | |
5449 status = RA_NOMATCH; | |
5450 else | |
5451 { | |
5452 status = RA_CONT; | |
233 | 5453 if (rp->rs_no != SUBPAT) /* zero-width */ |
5454 reg_restore(&rp->rs_un.regsave, &backpos); | |
180 | 5455 } |
270 | 5456 regstack_pop(&scan); |
180 | 5457 if (status == RA_CONT) |
5458 scan = regnext(scan); | |
5459 break; | |
5460 | |
5461 case RS_BEHIND1: | |
5462 if (status == RA_NOMATCH) | |
5463 { | |
270 | 5464 regstack_pop(&scan); |
180 | 5465 regstack.ga_len -= sizeof(regbehind_T); |
5466 } | |
5467 else | |
5468 { | |
5469 /* The stuff after BEHIND/NOBEHIND matches. Now try if | |
5470 * the behind part does (not) match before the current | |
5471 * position in the input. This must be done at every | |
5472 * position in the input and checking if the match ends at | |
5473 * the current position. */ | |
5474 | |
5475 /* save the position after the found match for next */ | |
233 | 5476 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos); |
180 | 5477 |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5478 /* Start looking for a match with operand at the current |
1209 | 5479 * position. Go back one character until we find the |
180 | 5480 * result, hitting the start of the line or the previous |
5481 * line (for multi-line matching). | |
5482 * Set behind_pos to where the match should end, BHPOS | |
5483 * will match it. Save the current value. */ | |
5484 (((regbehind_T *)rp) - 1)->save_behind = behind_pos; | |
5485 behind_pos = rp->rs_un.regsave; | |
5486 | |
5487 rp->rs_state = RS_BEHIND2; | |
5488 | |
233 | 5489 reg_restore(&rp->rs_un.regsave, &backpos); |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5490 scan = OPERAND(rp->rs_scan) + 4; |
180 | 5491 } |
5492 break; | |
5493 | |
5494 case RS_BEHIND2: | |
5495 /* | |
5496 * Looping for BEHIND / NOBEHIND match. | |
5497 */ | |
5498 if (status == RA_MATCH && reg_save_equal(&behind_pos)) | |
5499 { | |
5500 /* found a match that ends where "next" started */ | |
5501 behind_pos = (((regbehind_T *)rp) - 1)->save_behind; | |
5502 if (rp->rs_no == BEHIND) | |
233 | 5503 reg_restore(&(((regbehind_T *)rp) - 1)->save_after, |
5504 &backpos); | |
180 | 5505 else |
1579 | 5506 { |
5507 /* But we didn't want a match. Need to restore the | |
5508 * subexpr, because what follows matched, so they have | |
5509 * been set. */ | |
180 | 5510 status = RA_NOMATCH; |
1579 | 5511 restore_subexpr(((regbehind_T *)rp) - 1); |
5512 } | |
270 | 5513 regstack_pop(&scan); |
180 | 5514 regstack.ga_len -= sizeof(regbehind_T); |
5515 } | |
5516 else | |
5517 { | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5518 long limit; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5519 |
1579 | 5520 /* No match or a match that doesn't end where we want it: Go |
5521 * back one character. May go to previous line once. */ | |
180 | 5522 no = OK; |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5523 limit = OPERAND_MIN(rp->rs_scan); |
180 | 5524 if (REG_MULTI) |
5525 { | |
5526 if (rp->rs_un.regsave.rs_u.pos.col == 0) | |
5527 { | |
5528 if (rp->rs_un.regsave.rs_u.pos.lnum | |
5529 < behind_pos.rs_u.pos.lnum | |
5530 || reg_getline( | |
5531 --rp->rs_un.regsave.rs_u.pos.lnum) | |
5532 == NULL) | |
5533 no = FAIL; | |
5534 else | |
5535 { | |
233 | 5536 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5537 rp->rs_un.regsave.rs_u.pos.col = |
5538 (colnr_T)STRLEN(regline); | |
5539 } | |
5540 } | |
5541 else | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5542 { |
4176 | 5543 #ifdef FEAT_MBYTE |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5544 if (has_mbyte) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5545 rp->rs_un.regsave.rs_u.pos.col -= |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5546 (*mb_head_off)(regline, regline |
4176 | 5547 + rp->rs_un.regsave.rs_u.pos.col - 1) + 1; |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5548 else |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5549 #endif |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5550 --rp->rs_un.regsave.rs_u.pos.col; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5551 if (limit > 0 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5552 && ((rp->rs_un.regsave.rs_u.pos.lnum |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5553 < behind_pos.rs_u.pos.lnum |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5554 ? (colnr_T)STRLEN(regline) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5555 : behind_pos.rs_u.pos.col) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5556 - rp->rs_un.regsave.rs_u.pos.col > limit)) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5557 no = FAIL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5558 } |
180 | 5559 } |
5560 else | |
5561 { | |
5562 if (rp->rs_un.regsave.rs_u.ptr == regline) | |
5563 no = FAIL; | |
5564 else | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5565 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5566 mb_ptr_back(regline, rp->rs_un.regsave.rs_u.ptr); |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5567 if (limit > 0 && (long)(behind_pos.rs_u.ptr |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5568 - rp->rs_un.regsave.rs_u.ptr) > limit) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5569 no = FAIL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5570 } |
180 | 5571 } |
5572 if (no == OK) | |
5573 { | |
5574 /* Advanced, prepare for finding match again. */ | |
233 | 5575 reg_restore(&rp->rs_un.regsave, &backpos); |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5576 scan = OPERAND(rp->rs_scan) + 4; |
1579 | 5577 if (status == RA_MATCH) |
5578 { | |
5579 /* We did match, so subexpr may have been changed, | |
5580 * need to restore them for the next try. */ | |
5581 status = RA_NOMATCH; | |
5582 restore_subexpr(((regbehind_T *)rp) - 1); | |
5583 } | |
180 | 5584 } |
5585 else | |
5586 { | |
5587 /* Can't advance. For NOBEHIND that's a match. */ | |
5588 behind_pos = (((regbehind_T *)rp) - 1)->save_behind; | |
5589 if (rp->rs_no == NOBEHIND) | |
5590 { | |
233 | 5591 reg_restore(&(((regbehind_T *)rp) - 1)->save_after, |
5592 &backpos); | |
180 | 5593 status = RA_MATCH; |
5594 } | |
5595 else | |
1579 | 5596 { |
5597 /* We do want a proper match. Need to restore the | |
5598 * subexpr if we had a match, because they may have | |
5599 * been set. */ | |
5600 if (status == RA_MATCH) | |
5601 { | |
5602 status = RA_NOMATCH; | |
5603 restore_subexpr(((regbehind_T *)rp) - 1); | |
5604 } | |
5605 } | |
270 | 5606 regstack_pop(&scan); |
180 | 5607 regstack.ga_len -= sizeof(regbehind_T); |
5608 } | |
5609 } | |
5610 break; | |
5611 | |
5612 case RS_STAR_LONG: | |
5613 case RS_STAR_SHORT: | |
5614 { | |
5615 regstar_T *rst = ((regstar_T *)rp) - 1; | |
5616 | |
5617 if (status == RA_MATCH) | |
5618 { | |
270 | 5619 regstack_pop(&scan); |
180 | 5620 regstack.ga_len -= sizeof(regstar_T); |
5621 break; | |
5622 } | |
5623 | |
5624 /* Tried once already, restore input pointers. */ | |
5625 if (status != RA_BREAK) | |
233 | 5626 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5627 |
5628 /* Repeat until we found a position where it could match. */ | |
5629 for (;;) | |
5630 { | |
5631 if (status != RA_BREAK) | |
5632 { | |
5633 /* Tried first position already, advance. */ | |
5634 if (rp->rs_state == RS_STAR_LONG) | |
5635 { | |
685 | 5636 /* Trying for longest match, but couldn't or |
5637 * didn't match -- back up one char. */ | |
180 | 5638 if (--rst->count < rst->minval) |
5639 break; | |
5640 if (reginput == regline) | |
5641 { | |
5642 /* backup to last char of previous line */ | |
5643 --reglnum; | |
5644 regline = reg_getline(reglnum); | |
5645 /* Just in case regrepeat() didn't count | |
5646 * right. */ | |
5647 if (regline == NULL) | |
5648 break; | |
5649 reginput = regline + STRLEN(regline); | |
5650 fast_breakcheck(); | |
5651 } | |
5652 else | |
5653 mb_ptr_back(regline, reginput); | |
5654 } | |
5655 else | |
5656 { | |
5657 /* Range is backwards, use shortest match first. | |
5658 * Careful: maxval and minval are exchanged! | |
5659 * Couldn't or didn't match: try advancing one | |
5660 * char. */ | |
5661 if (rst->count == rst->minval | |
5662 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0) | |
5663 break; | |
5664 ++rst->count; | |
5665 } | |
5666 if (got_int) | |
5667 break; | |
5668 } | |
5669 else | |
5670 status = RA_NOMATCH; | |
5671 | |
5672 /* If it could match, try it. */ | |
5673 if (rst->nextb == NUL || *reginput == rst->nextb | |
5674 || *reginput == rst->nextb_ic) | |
5675 { | |
233 | 5676 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5677 scan = regnext(rp->rs_scan); |
5678 status = RA_CONT; | |
5679 break; | |
5680 } | |
5681 } | |
5682 if (status != RA_CONT) | |
5683 { | |
5684 /* Failed. */ | |
270 | 5685 regstack_pop(&scan); |
180 | 5686 regstack.ga_len -= sizeof(regstar_T); |
5687 status = RA_NOMATCH; | |
5688 } | |
5689 } | |
5690 break; | |
5691 } | |
5692 | |
685 | 5693 /* If we want to continue the inner loop or didn't pop a state |
5694 * continue matching loop */ | |
180 | 5695 if (status == RA_CONT || rp == (regitem_T *) |
5696 ((char *)regstack.ga_data + regstack.ga_len) - 1) | |
5697 break; | |
5698 } | |
5699 | |
189 | 5700 /* May need to continue with the inner loop, starting at "scan". */ |
180 | 5701 if (status == RA_CONT) |
5702 continue; | |
5703 | |
5704 /* | |
5705 * If the regstack is empty or something failed we are done. | |
5706 */ | |
5707 if (regstack.ga_len == 0 || status == RA_FAIL) | |
5708 { | |
5709 if (scan == NULL) | |
5710 { | |
5711 /* | |
5712 * We get here only if there's trouble -- normally "case END" is | |
5713 * the terminating point. | |
5714 */ | |
5715 EMSG(_(e_re_corr)); | |
7 | 5716 #ifdef DEBUG |
180 | 5717 printf("Premature EOL\n"); |
7 | 5718 #endif |
180 | 5719 } |
189 | 5720 if (status == RA_FAIL) |
5721 got_int = TRUE; | |
180 | 5722 return (status == RA_MATCH); |
5723 } | |
5724 | |
5725 } /* End of loop until the regstack is empty. */ | |
5726 | |
5727 /* NOTREACHED */ | |
5728 } | |
5729 | |
5730 /* | |
5731 * Push an item onto the regstack. | |
5732 * Returns pointer to new item. Returns NULL when out of memory. | |
5733 */ | |
5734 static regitem_T * | |
270 | 5735 regstack_push(state, scan) |
180 | 5736 regstate_T state; |
5737 char_u *scan; | |
5738 { | |
5739 regitem_T *rp; | |
5740 | |
270 | 5741 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5742 { |
5743 EMSG(_(e_maxmempat)); | |
5744 return NULL; | |
5745 } | |
270 | 5746 if (ga_grow(®stack, sizeof(regitem_T)) == FAIL) |
180 | 5747 return NULL; |
5748 | |
270 | 5749 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len); |
180 | 5750 rp->rs_state = state; |
5751 rp->rs_scan = scan; | |
5752 | |
270 | 5753 regstack.ga_len += sizeof(regitem_T); |
180 | 5754 return rp; |
5755 } | |
5756 | |
5757 /* | |
5758 * Pop an item from the regstack. | |
5759 */ | |
5760 static void | |
270 | 5761 regstack_pop(scan) |
180 | 5762 char_u **scan; |
5763 { | |
5764 regitem_T *rp; | |
5765 | |
270 | 5766 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1; |
180 | 5767 *scan = rp->rs_scan; |
5768 | |
270 | 5769 regstack.ga_len -= sizeof(regitem_T); |
7 | 5770 } |
5771 | |
5772 /* | |
5773 * regrepeat - repeatedly match something simple, return how many. | |
5774 * Advances reginput (and reglnum) to just after the matched chars. | |
5775 */ | |
5776 static int | |
5777 regrepeat(p, maxcount) | |
5778 char_u *p; | |
5779 long maxcount; /* maximum number of matches allowed */ | |
5780 { | |
5781 long count = 0; | |
5782 char_u *scan; | |
5783 char_u *opnd; | |
5784 int mask; | |
5785 int testval = 0; | |
5786 | |
5787 scan = reginput; /* Make local copy of reginput for speed. */ | |
5788 opnd = OPERAND(p); | |
5789 switch (OP(p)) | |
5790 { | |
5791 case ANY: | |
5792 case ANY + ADD_NL: | |
5793 while (count < maxcount) | |
5794 { | |
5795 /* Matching anything means we continue until end-of-line (or | |
5796 * end-of-file for ANY + ADD_NL), only limited by maxcount. */ | |
5797 while (*scan != NUL && count < maxcount) | |
5798 { | |
5799 ++count; | |
39 | 5800 mb_ptr_adv(scan); |
7 | 5801 } |
1018 | 5802 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5803 || reg_line_lbr || count == maxcount) | |
7 | 5804 break; |
5805 ++count; /* count the line-break */ | |
5806 reg_nextline(); | |
5807 scan = reginput; | |
5808 if (got_int) | |
5809 break; | |
5810 } | |
5811 break; | |
5812 | |
5813 case IDENT: | |
5814 case IDENT + ADD_NL: | |
5815 testval = TRUE; | |
5816 /*FALLTHROUGH*/ | |
5817 case SIDENT: | |
5818 case SIDENT + ADD_NL: | |
5819 while (count < maxcount) | |
5820 { | |
4466 | 5821 if (vim_isIDc(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan))) |
7 | 5822 { |
39 | 5823 mb_ptr_adv(scan); |
7 | 5824 } |
5825 else if (*scan == NUL) | |
5826 { | |
1018 | 5827 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5828 || reg_line_lbr) | |
7 | 5829 break; |
5830 reg_nextline(); | |
5831 scan = reginput; | |
5832 if (got_int) | |
5833 break; | |
5834 } | |
5835 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5836 ++scan; | |
5837 else | |
5838 break; | |
5839 ++count; | |
5840 } | |
5841 break; | |
5842 | |
5843 case KWORD: | |
5844 case KWORD + ADD_NL: | |
5845 testval = TRUE; | |
5846 /*FALLTHROUGH*/ | |
5847 case SKWORD: | |
5848 case SKWORD + ADD_NL: | |
5849 while (count < maxcount) | |
5850 { | |
4069 | 5851 if (vim_iswordp_buf(scan, reg_buf) |
5852 && (testval || !VIM_ISDIGIT(*scan))) | |
7 | 5853 { |
39 | 5854 mb_ptr_adv(scan); |
7 | 5855 } |
5856 else if (*scan == NUL) | |
5857 { | |
1018 | 5858 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5859 || reg_line_lbr) | |
7 | 5860 break; |
5861 reg_nextline(); | |
5862 scan = reginput; | |
5863 if (got_int) | |
5864 break; | |
5865 } | |
5866 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5867 ++scan; | |
5868 else | |
5869 break; | |
5870 ++count; | |
5871 } | |
5872 break; | |
5873 | |
5874 case FNAME: | |
5875 case FNAME + ADD_NL: | |
5876 testval = TRUE; | |
5877 /*FALLTHROUGH*/ | |
5878 case SFNAME: | |
5879 case SFNAME + ADD_NL: | |
5880 while (count < maxcount) | |
5881 { | |
4466 | 5882 if (vim_isfilec(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan))) |
7 | 5883 { |
39 | 5884 mb_ptr_adv(scan); |
7 | 5885 } |
5886 else if (*scan == NUL) | |
5887 { | |
1018 | 5888 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5889 || reg_line_lbr) | |
7 | 5890 break; |
5891 reg_nextline(); | |
5892 scan = reginput; | |
5893 if (got_int) | |
5894 break; | |
5895 } | |
5896 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5897 ++scan; | |
5898 else | |
5899 break; | |
5900 ++count; | |
5901 } | |
5902 break; | |
5903 | |
5904 case PRINT: | |
5905 case PRINT + ADD_NL: | |
5906 testval = TRUE; | |
5907 /*FALLTHROUGH*/ | |
5908 case SPRINT: | |
5909 case SPRINT + ADD_NL: | |
5910 while (count < maxcount) | |
5911 { | |
5912 if (*scan == NUL) | |
5913 { | |
1018 | 5914 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5915 || reg_line_lbr) | |
7 | 5916 break; |
5917 reg_nextline(); | |
5918 scan = reginput; | |
5919 if (got_int) | |
5920 break; | |
5921 } | |
5922 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan))) | |
5923 { | |
39 | 5924 mb_ptr_adv(scan); |
7 | 5925 } |
5926 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5927 ++scan; | |
5928 else | |
5929 break; | |
5930 ++count; | |
5931 } | |
5932 break; | |
5933 | |
5934 case WHITE: | |
5935 case WHITE + ADD_NL: | |
5936 testval = mask = RI_WHITE; | |
5937 do_class: | |
5938 while (count < maxcount) | |
5939 { | |
5940 #ifdef FEAT_MBYTE | |
5941 int l; | |
5942 #endif | |
5943 if (*scan == NUL) | |
5944 { | |
1018 | 5945 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5946 || reg_line_lbr) | |
7 | 5947 break; |
5948 reg_nextline(); | |
5949 scan = reginput; | |
5950 if (got_int) | |
5951 break; | |
5952 } | |
5953 #ifdef FEAT_MBYTE | |
474 | 5954 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1) |
7 | 5955 { |
5956 if (testval != 0) | |
5957 break; | |
5958 scan += l; | |
5959 } | |
5960 #endif | |
5961 else if ((class_tab[*scan] & mask) == testval) | |
5962 ++scan; | |
5963 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5964 ++scan; | |
5965 else | |
5966 break; | |
5967 ++count; | |
5968 } | |
5969 break; | |
5970 | |
5971 case NWHITE: | |
5972 case NWHITE + ADD_NL: | |
5973 mask = RI_WHITE; | |
5974 goto do_class; | |
5975 case DIGIT: | |
5976 case DIGIT + ADD_NL: | |
5977 testval = mask = RI_DIGIT; | |
5978 goto do_class; | |
5979 case NDIGIT: | |
5980 case NDIGIT + ADD_NL: | |
5981 mask = RI_DIGIT; | |
5982 goto do_class; | |
5983 case HEX: | |
5984 case HEX + ADD_NL: | |
5985 testval = mask = RI_HEX; | |
5986 goto do_class; | |
5987 case NHEX: | |
5988 case NHEX + ADD_NL: | |
5989 mask = RI_HEX; | |
5990 goto do_class; | |
5991 case OCTAL: | |
5992 case OCTAL + ADD_NL: | |
5993 testval = mask = RI_OCTAL; | |
5994 goto do_class; | |
5995 case NOCTAL: | |
5996 case NOCTAL + ADD_NL: | |
5997 mask = RI_OCTAL; | |
5998 goto do_class; | |
5999 case WORD: | |
6000 case WORD + ADD_NL: | |
6001 testval = mask = RI_WORD; | |
6002 goto do_class; | |
6003 case NWORD: | |
6004 case NWORD + ADD_NL: | |
6005 mask = RI_WORD; | |
6006 goto do_class; | |
6007 case HEAD: | |
6008 case HEAD + ADD_NL: | |
6009 testval = mask = RI_HEAD; | |
6010 goto do_class; | |
6011 case NHEAD: | |
6012 case NHEAD + ADD_NL: | |
6013 mask = RI_HEAD; | |
6014 goto do_class; | |
6015 case ALPHA: | |
6016 case ALPHA + ADD_NL: | |
6017 testval = mask = RI_ALPHA; | |
6018 goto do_class; | |
6019 case NALPHA: | |
6020 case NALPHA + ADD_NL: | |
6021 mask = RI_ALPHA; | |
6022 goto do_class; | |
6023 case LOWER: | |
6024 case LOWER + ADD_NL: | |
6025 testval = mask = RI_LOWER; | |
6026 goto do_class; | |
6027 case NLOWER: | |
6028 case NLOWER + ADD_NL: | |
6029 mask = RI_LOWER; | |
6030 goto do_class; | |
6031 case UPPER: | |
6032 case UPPER + ADD_NL: | |
6033 testval = mask = RI_UPPER; | |
6034 goto do_class; | |
6035 case NUPPER: | |
6036 case NUPPER + ADD_NL: | |
6037 mask = RI_UPPER; | |
6038 goto do_class; | |
6039 | |
6040 case EXACTLY: | |
6041 { | |
6042 int cu, cl; | |
6043 | |
6044 /* This doesn't do a multi-byte character, because a MULTIBYTECODE | |
1347 | 6045 * would have been used for it. It does handle single-byte |
6046 * characters, such as latin1. */ | |
7 | 6047 if (ireg_ic) |
6048 { | |
1347 | 6049 cu = MB_TOUPPER(*opnd); |
6050 cl = MB_TOLOWER(*opnd); | |
7 | 6051 while (count < maxcount && (*scan == cu || *scan == cl)) |
6052 { | |
6053 count++; | |
6054 scan++; | |
6055 } | |
6056 } | |
6057 else | |
6058 { | |
6059 cu = *opnd; | |
6060 while (count < maxcount && *scan == cu) | |
6061 { | |
6062 count++; | |
6063 scan++; | |
6064 } | |
6065 } | |
6066 break; | |
6067 } | |
6068 | |
6069 #ifdef FEAT_MBYTE | |
6070 case MULTIBYTECODE: | |
6071 { | |
6072 int i, len, cf = 0; | |
6073 | |
6074 /* Safety check (just in case 'encoding' was changed since | |
6075 * compiling the program). */ | |
474 | 6076 if ((len = (*mb_ptr2len)(opnd)) > 1) |
7 | 6077 { |
6078 if (ireg_ic && enc_utf8) | |
6079 cf = utf_fold(utf_ptr2char(opnd)); | |
6080 while (count < maxcount) | |
6081 { | |
6082 for (i = 0; i < len; ++i) | |
6083 if (opnd[i] != scan[i]) | |
6084 break; | |
6085 if (i < len && (!ireg_ic || !enc_utf8 | |
6086 || utf_fold(utf_ptr2char(scan)) != cf)) | |
6087 break; | |
6088 scan += len; | |
6089 ++count; | |
6090 } | |
6091 } | |
6092 } | |
6093 break; | |
6094 #endif | |
6095 | |
6096 case ANYOF: | |
6097 case ANYOF + ADD_NL: | |
6098 testval = TRUE; | |
6099 /*FALLTHROUGH*/ | |
6100 | |
6101 case ANYBUT: | |
6102 case ANYBUT + ADD_NL: | |
6103 while (count < maxcount) | |
6104 { | |
6105 #ifdef FEAT_MBYTE | |
6106 int len; | |
6107 #endif | |
6108 if (*scan == NUL) | |
6109 { | |
1018 | 6110 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
6111 || reg_line_lbr) | |
7 | 6112 break; |
6113 reg_nextline(); | |
6114 scan = reginput; | |
6115 if (got_int) | |
6116 break; | |
6117 } | |
6118 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
6119 ++scan; | |
6120 #ifdef FEAT_MBYTE | |
474 | 6121 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1) |
7 | 6122 { |
6123 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval) | |
6124 break; | |
6125 scan += len; | |
6126 } | |
6127 #endif | |
6128 else | |
6129 { | |
6130 if ((cstrchr(opnd, *scan) == NULL) == testval) | |
6131 break; | |
6132 ++scan; | |
6133 } | |
6134 ++count; | |
6135 } | |
6136 break; | |
6137 | |
6138 case NEWL: | |
6139 while (count < maxcount | |
1018 | 6140 && ((*scan == NUL && reglnum <= reg_maxline && !reg_line_lbr |
6141 && REG_MULTI) || (*scan == '\n' && reg_line_lbr))) | |
7 | 6142 { |
6143 count++; | |
6144 if (reg_line_lbr) | |
6145 ADVANCE_REGINPUT(); | |
6146 else | |
6147 reg_nextline(); | |
6148 scan = reginput; | |
6149 if (got_int) | |
6150 break; | |
6151 } | |
6152 break; | |
6153 | |
6154 default: /* Oh dear. Called inappropriately. */ | |
6155 EMSG(_(e_re_corr)); | |
6156 #ifdef DEBUG | |
6157 printf("Called regrepeat with op code %d\n", OP(p)); | |
6158 #endif | |
6159 break; | |
6160 } | |
6161 | |
6162 reginput = scan; | |
6163 | |
6164 return (int)count; | |
6165 } | |
6166 | |
6167 /* | |
6168 * regnext - dig the "next" pointer out of a node | |
2010 | 6169 * Returns NULL when calculating size, when there is no next item and when |
6170 * there is an error. | |
7 | 6171 */ |
6172 static char_u * | |
6173 regnext(p) | |
6174 char_u *p; | |
6175 { | |
6176 int offset; | |
6177 | |
2010 | 6178 if (p == JUST_CALC_SIZE || reg_toolong) |
7 | 6179 return NULL; |
6180 | |
6181 offset = NEXT(p); | |
6182 if (offset == 0) | |
6183 return NULL; | |
6184 | |
233 | 6185 if (OP(p) == BACK) |
7 | 6186 return p - offset; |
6187 else | |
6188 return p + offset; | |
6189 } | |
6190 | |
6191 /* | |
6192 * Check the regexp program for its magic number. | |
6193 * Return TRUE if it's wrong. | |
6194 */ | |
6195 static int | |
6196 prog_magic_wrong() | |
6197 { | |
4444 | 6198 regprog_T *prog; |
6199 | |
6200 prog = REG_MULTI ? reg_mmatch->regprog : reg_match->regprog; | |
6201 if (prog->engine == &nfa_regengine) | |
6202 /* For NFA matcher we don't check the magic */ | |
6203 return FALSE; | |
6204 | |
6205 if (UCHARAT(((bt_regprog_T *)prog)->program) != REGMAGIC) | |
7 | 6206 { |
6207 EMSG(_(e_re_corr)); | |
6208 return TRUE; | |
6209 } | |
6210 return FALSE; | |
6211 } | |
6212 | |
6213 /* | |
6214 * Cleanup the subexpressions, if this wasn't done yet. | |
6215 * This construction is used to clear the subexpressions only when they are | |
6216 * used (to increase speed). | |
6217 */ | |
6218 static void | |
6219 cleanup_subexpr() | |
6220 { | |
6221 if (need_clear_subexpr) | |
6222 { | |
6223 if (REG_MULTI) | |
6224 { | |
6225 /* Use 0xff to set lnum to -1 */ | |
6226 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6227 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6228 } | |
6229 else | |
6230 { | |
6231 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP); | |
6232 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP); | |
6233 } | |
6234 need_clear_subexpr = FALSE; | |
6235 } | |
6236 } | |
6237 | |
6238 #ifdef FEAT_SYN_HL | |
6239 static void | |
6240 cleanup_zsubexpr() | |
6241 { | |
6242 if (need_clear_zsubexpr) | |
6243 { | |
6244 if (REG_MULTI) | |
6245 { | |
6246 /* Use 0xff to set lnum to -1 */ | |
6247 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6248 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6249 } | |
6250 else | |
6251 { | |
6252 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP); | |
6253 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP); | |
6254 } | |
6255 need_clear_zsubexpr = FALSE; | |
6256 } | |
6257 } | |
6258 #endif | |
6259 | |
6260 /* | |
1579 | 6261 * Save the current subexpr to "bp", so that they can be restored |
6262 * later by restore_subexpr(). | |
6263 */ | |
6264 static void | |
6265 save_subexpr(bp) | |
6266 regbehind_T *bp; | |
6267 { | |
6268 int i; | |
6269 | |
1602 | 6270 /* When "need_clear_subexpr" is set we don't need to save the values, only |
6271 * remember that this flag needs to be set again when restoring. */ | |
6272 bp->save_need_clear_subexpr = need_clear_subexpr; | |
6273 if (!need_clear_subexpr) | |
1579 | 6274 { |
1602 | 6275 for (i = 0; i < NSUBEXP; ++i) |
1579 | 6276 { |
1602 | 6277 if (REG_MULTI) |
6278 { | |
6279 bp->save_start[i].se_u.pos = reg_startpos[i]; | |
6280 bp->save_end[i].se_u.pos = reg_endpos[i]; | |
6281 } | |
6282 else | |
6283 { | |
6284 bp->save_start[i].se_u.ptr = reg_startp[i]; | |
6285 bp->save_end[i].se_u.ptr = reg_endp[i]; | |
6286 } | |
1579 | 6287 } |
6288 } | |
6289 } | |
6290 | |
6291 /* | |
6292 * Restore the subexpr from "bp". | |
6293 */ | |
6294 static void | |
6295 restore_subexpr(bp) | |
6296 regbehind_T *bp; | |
6297 { | |
6298 int i; | |
6299 | |
1602 | 6300 /* Only need to restore saved values when they are not to be cleared. */ |
6301 need_clear_subexpr = bp->save_need_clear_subexpr; | |
6302 if (!need_clear_subexpr) | |
1579 | 6303 { |
1602 | 6304 for (i = 0; i < NSUBEXP; ++i) |
1579 | 6305 { |
1602 | 6306 if (REG_MULTI) |
6307 { | |
6308 reg_startpos[i] = bp->save_start[i].se_u.pos; | |
6309 reg_endpos[i] = bp->save_end[i].se_u.pos; | |
6310 } | |
6311 else | |
6312 { | |
6313 reg_startp[i] = bp->save_start[i].se_u.ptr; | |
6314 reg_endp[i] = bp->save_end[i].se_u.ptr; | |
6315 } | |
1579 | 6316 } |
6317 } | |
6318 } | |
6319 | |
6320 /* | |
7 | 6321 * Advance reglnum, regline and reginput to the next line. |
6322 */ | |
6323 static void | |
6324 reg_nextline() | |
6325 { | |
6326 regline = reg_getline(++reglnum); | |
6327 reginput = regline; | |
6328 fast_breakcheck(); | |
6329 } | |
6330 | |
6331 /* | |
6332 * Save the input line and position in a regsave_T. | |
6333 */ | |
6334 static void | |
233 | 6335 reg_save(save, gap) |
7 | 6336 regsave_T *save; |
233 | 6337 garray_T *gap; |
7 | 6338 { |
6339 if (REG_MULTI) | |
6340 { | |
6341 save->rs_u.pos.col = (colnr_T)(reginput - regline); | |
6342 save->rs_u.pos.lnum = reglnum; | |
6343 } | |
6344 else | |
6345 save->rs_u.ptr = reginput; | |
233 | 6346 save->rs_len = gap->ga_len; |
7 | 6347 } |
6348 | |
6349 /* | |
6350 * Restore the input line and position from a regsave_T. | |
6351 */ | |
6352 static void | |
233 | 6353 reg_restore(save, gap) |
7 | 6354 regsave_T *save; |
233 | 6355 garray_T *gap; |
7 | 6356 { |
6357 if (REG_MULTI) | |
6358 { | |
6359 if (reglnum != save->rs_u.pos.lnum) | |
6360 { | |
6361 /* only call reg_getline() when the line number changed to save | |
6362 * a bit of time */ | |
6363 reglnum = save->rs_u.pos.lnum; | |
6364 regline = reg_getline(reglnum); | |
6365 } | |
6366 reginput = regline + save->rs_u.pos.col; | |
6367 } | |
6368 else | |
6369 reginput = save->rs_u.ptr; | |
233 | 6370 gap->ga_len = save->rs_len; |
7 | 6371 } |
6372 | |
6373 /* | |
6374 * Return TRUE if current position is equal to saved position. | |
6375 */ | |
6376 static int | |
6377 reg_save_equal(save) | |
6378 regsave_T *save; | |
6379 { | |
6380 if (REG_MULTI) | |
6381 return reglnum == save->rs_u.pos.lnum | |
6382 && reginput == regline + save->rs_u.pos.col; | |
6383 return reginput == save->rs_u.ptr; | |
6384 } | |
6385 | |
6386 /* | |
6387 * Tentatively set the sub-expression start to the current position (after | |
6388 * calling regmatch() they will have changed). Need to save the existing | |
6389 * values for when there is no match. | |
6390 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()), | |
6391 * depending on REG_MULTI. | |
6392 */ | |
6393 static void | |
6394 save_se_multi(savep, posp) | |
6395 save_se_T *savep; | |
6396 lpos_T *posp; | |
6397 { | |
6398 savep->se_u.pos = *posp; | |
6399 posp->lnum = reglnum; | |
6400 posp->col = (colnr_T)(reginput - regline); | |
6401 } | |
6402 | |
6403 static void | |
6404 save_se_one(savep, pp) | |
6405 save_se_T *savep; | |
6406 char_u **pp; | |
6407 { | |
6408 savep->se_u.ptr = *pp; | |
6409 *pp = reginput; | |
6410 } | |
6411 | |
6412 /* | |
6413 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL. | |
6414 */ | |
6415 static int | |
6416 re_num_cmp(val, scan) | |
6417 long_u val; | |
6418 char_u *scan; | |
6419 { | |
6420 long_u n = OPERAND_MIN(scan); | |
6421 | |
6422 if (OPERAND_CMP(scan) == '>') | |
6423 return val > n; | |
6424 if (OPERAND_CMP(scan) == '<') | |
6425 return val < n; | |
6426 return val == n; | |
6427 } | |
6428 | |
6429 | |
4444 | 6430 #ifdef BT_REGEXP_DUMP |
7 | 6431 |
6432 /* | |
6433 * regdump - dump a regexp onto stdout in vaguely comprehensible form | |
6434 */ | |
6435 static void | |
6436 regdump(pattern, r) | |
6437 char_u *pattern; | |
4444 | 6438 bt_regprog_T *r; |
7 | 6439 { |
6440 char_u *s; | |
6441 int op = EXACTLY; /* Arbitrary non-END op. */ | |
6442 char_u *next; | |
6443 char_u *end = NULL; | |
4444 | 6444 FILE *f; |
6445 | |
6446 #ifdef BT_REGEXP_LOG | |
6447 f = fopen("bt_regexp_log.log", "a"); | |
6448 #else | |
6449 f = stdout; | |
6450 #endif | |
6451 if (f == NULL) | |
6452 return; | |
6453 fprintf(f, "-------------------------------------\n\r\nregcomp(%s):\r\n", pattern); | |
7 | 6454 |
6455 s = r->program + 1; | |
6456 /* | |
6457 * Loop until we find the END that isn't before a referred next (an END | |
6458 * can also appear in a NOMATCH operand). | |
6459 */ | |
6460 while (op != END || s <= end) | |
6461 { | |
6462 op = OP(s); | |
4444 | 6463 fprintf(f, "%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */ |
7 | 6464 next = regnext(s); |
6465 if (next == NULL) /* Next ptr. */ | |
4444 | 6466 fprintf(f, "(0)"); |
7 | 6467 else |
4444 | 6468 fprintf(f, "(%d)", (int)((s - r->program) + (next - s))); |
7 | 6469 if (end < next) |
6470 end = next; | |
6471 if (op == BRACE_LIMITS) | |
6472 { | |
6473 /* Two short ints */ | |
4444 | 6474 fprintf(f, " minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s)); |
7 | 6475 s += 8; |
6476 } | |
6477 s += 3; | |
6478 if (op == ANYOF || op == ANYOF + ADD_NL | |
6479 || op == ANYBUT || op == ANYBUT + ADD_NL | |
6480 || op == EXACTLY) | |
6481 { | |
6482 /* Literal string, where present. */ | |
4444 | 6483 fprintf(f, "\nxxxxxxxxx\n"); |
7 | 6484 while (*s != NUL) |
4444 | 6485 fprintf(f, "%c", *s++); |
6486 fprintf(f, "\nxxxxxxxxx\n"); | |
7 | 6487 s++; |
6488 } | |
4444 | 6489 fprintf(f, "\r\n"); |
7 | 6490 } |
6491 | |
6492 /* Header fields of interest. */ | |
6493 if (r->regstart != NUL) | |
4444 | 6494 fprintf(f, "start `%s' 0x%x; ", r->regstart < 256 |
7 | 6495 ? (char *)transchar(r->regstart) |
6496 : "multibyte", r->regstart); | |
6497 if (r->reganch) | |
4444 | 6498 fprintf(f, "anchored; "); |
7 | 6499 if (r->regmust != NULL) |
4444 | 6500 fprintf(f, "must have \"%s\"", r->regmust); |
6501 fprintf(f, "\r\n"); | |
6502 | |
6503 #ifdef BT_REGEXP_LOG | |
6504 fclose(f); | |
6505 #endif | |
7 | 6506 } |
4444 | 6507 #endif /* BT_REGEXP_DUMP */ |
6508 | |
6509 #ifdef DEBUG | |
7 | 6510 /* |
6511 * regprop - printable representation of opcode | |
6512 */ | |
6513 static char_u * | |
6514 regprop(op) | |
6515 char_u *op; | |
6516 { | |
4444 | 6517 char *p; |
6518 static char buf[50]; | |
6519 | |
6520 STRCPY(buf, ":"); | |
6521 | |
6522 switch ((int) OP(op)) | |
7 | 6523 { |
6524 case BOL: | |
6525 p = "BOL"; | |
6526 break; | |
6527 case EOL: | |
6528 p = "EOL"; | |
6529 break; | |
6530 case RE_BOF: | |
6531 p = "BOF"; | |
6532 break; | |
6533 case RE_EOF: | |
6534 p = "EOF"; | |
6535 break; | |
6536 case CURSOR: | |
6537 p = "CURSOR"; | |
6538 break; | |
639 | 6539 case RE_VISUAL: |
6540 p = "RE_VISUAL"; | |
6541 break; | |
7 | 6542 case RE_LNUM: |
6543 p = "RE_LNUM"; | |
6544 break; | |
639 | 6545 case RE_MARK: |
6546 p = "RE_MARK"; | |
6547 break; | |
7 | 6548 case RE_COL: |
6549 p = "RE_COL"; | |
6550 break; | |
6551 case RE_VCOL: | |
6552 p = "RE_VCOL"; | |
6553 break; | |
6554 case BOW: | |
6555 p = "BOW"; | |
6556 break; | |
6557 case EOW: | |
6558 p = "EOW"; | |
6559 break; | |
6560 case ANY: | |
6561 p = "ANY"; | |
6562 break; | |
6563 case ANY + ADD_NL: | |
6564 p = "ANY+NL"; | |
6565 break; | |
6566 case ANYOF: | |
6567 p = "ANYOF"; | |
6568 break; | |
6569 case ANYOF + ADD_NL: | |
6570 p = "ANYOF+NL"; | |
6571 break; | |
6572 case ANYBUT: | |
6573 p = "ANYBUT"; | |
6574 break; | |
6575 case ANYBUT + ADD_NL: | |
6576 p = "ANYBUT+NL"; | |
6577 break; | |
6578 case IDENT: | |
6579 p = "IDENT"; | |
6580 break; | |
6581 case IDENT + ADD_NL: | |
6582 p = "IDENT+NL"; | |
6583 break; | |
6584 case SIDENT: | |
6585 p = "SIDENT"; | |
6586 break; | |
6587 case SIDENT + ADD_NL: | |
6588 p = "SIDENT+NL"; | |
6589 break; | |
6590 case KWORD: | |
6591 p = "KWORD"; | |
6592 break; | |
6593 case KWORD + ADD_NL: | |
6594 p = "KWORD+NL"; | |
6595 break; | |
6596 case SKWORD: | |
6597 p = "SKWORD"; | |
6598 break; | |
6599 case SKWORD + ADD_NL: | |
6600 p = "SKWORD+NL"; | |
6601 break; | |
6602 case FNAME: | |
6603 p = "FNAME"; | |
6604 break; | |
6605 case FNAME + ADD_NL: | |
6606 p = "FNAME+NL"; | |
6607 break; | |
6608 case SFNAME: | |
6609 p = "SFNAME"; | |
6610 break; | |
6611 case SFNAME + ADD_NL: | |
6612 p = "SFNAME+NL"; | |
6613 break; | |
6614 case PRINT: | |
6615 p = "PRINT"; | |
6616 break; | |
6617 case PRINT + ADD_NL: | |
6618 p = "PRINT+NL"; | |
6619 break; | |
6620 case SPRINT: | |
6621 p = "SPRINT"; | |
6622 break; | |
6623 case SPRINT + ADD_NL: | |
6624 p = "SPRINT+NL"; | |
6625 break; | |
6626 case WHITE: | |
6627 p = "WHITE"; | |
6628 break; | |
6629 case WHITE + ADD_NL: | |
6630 p = "WHITE+NL"; | |
6631 break; | |
6632 case NWHITE: | |
6633 p = "NWHITE"; | |
6634 break; | |
6635 case NWHITE + ADD_NL: | |
6636 p = "NWHITE+NL"; | |
6637 break; | |
6638 case DIGIT: | |
6639 p = "DIGIT"; | |
6640 break; | |
6641 case DIGIT + ADD_NL: | |
6642 p = "DIGIT+NL"; | |
6643 break; | |
6644 case NDIGIT: | |
6645 p = "NDIGIT"; | |
6646 break; | |
6647 case NDIGIT + ADD_NL: | |
6648 p = "NDIGIT+NL"; | |
6649 break; | |
6650 case HEX: | |
6651 p = "HEX"; | |
6652 break; | |
6653 case HEX + ADD_NL: | |
6654 p = "HEX+NL"; | |
6655 break; | |
6656 case NHEX: | |
6657 p = "NHEX"; | |
6658 break; | |
6659 case NHEX + ADD_NL: | |
6660 p = "NHEX+NL"; | |
6661 break; | |
6662 case OCTAL: | |
6663 p = "OCTAL"; | |
6664 break; | |
6665 case OCTAL + ADD_NL: | |
6666 p = "OCTAL+NL"; | |
6667 break; | |
6668 case NOCTAL: | |
6669 p = "NOCTAL"; | |
6670 break; | |
6671 case NOCTAL + ADD_NL: | |
6672 p = "NOCTAL+NL"; | |
6673 break; | |
6674 case WORD: | |
6675 p = "WORD"; | |
6676 break; | |
6677 case WORD + ADD_NL: | |
6678 p = "WORD+NL"; | |
6679 break; | |
6680 case NWORD: | |
6681 p = "NWORD"; | |
6682 break; | |
6683 case NWORD + ADD_NL: | |
6684 p = "NWORD+NL"; | |
6685 break; | |
6686 case HEAD: | |
6687 p = "HEAD"; | |
6688 break; | |
6689 case HEAD + ADD_NL: | |
6690 p = "HEAD+NL"; | |
6691 break; | |
6692 case NHEAD: | |
6693 p = "NHEAD"; | |
6694 break; | |
6695 case NHEAD + ADD_NL: | |
6696 p = "NHEAD+NL"; | |
6697 break; | |
6698 case ALPHA: | |
6699 p = "ALPHA"; | |
6700 break; | |
6701 case ALPHA + ADD_NL: | |
6702 p = "ALPHA+NL"; | |
6703 break; | |
6704 case NALPHA: | |
6705 p = "NALPHA"; | |
6706 break; | |
6707 case NALPHA + ADD_NL: | |
6708 p = "NALPHA+NL"; | |
6709 break; | |
6710 case LOWER: | |
6711 p = "LOWER"; | |
6712 break; | |
6713 case LOWER + ADD_NL: | |
6714 p = "LOWER+NL"; | |
6715 break; | |
6716 case NLOWER: | |
6717 p = "NLOWER"; | |
6718 break; | |
6719 case NLOWER + ADD_NL: | |
6720 p = "NLOWER+NL"; | |
6721 break; | |
6722 case UPPER: | |
6723 p = "UPPER"; | |
6724 break; | |
6725 case UPPER + ADD_NL: | |
6726 p = "UPPER+NL"; | |
6727 break; | |
6728 case NUPPER: | |
6729 p = "NUPPER"; | |
6730 break; | |
6731 case NUPPER + ADD_NL: | |
6732 p = "NUPPER+NL"; | |
6733 break; | |
6734 case BRANCH: | |
6735 p = "BRANCH"; | |
6736 break; | |
6737 case EXACTLY: | |
6738 p = "EXACTLY"; | |
6739 break; | |
6740 case NOTHING: | |
6741 p = "NOTHING"; | |
6742 break; | |
6743 case BACK: | |
6744 p = "BACK"; | |
6745 break; | |
6746 case END: | |
6747 p = "END"; | |
6748 break; | |
6749 case MOPEN + 0: | |
6750 p = "MATCH START"; | |
6751 break; | |
6752 case MOPEN + 1: | |
6753 case MOPEN + 2: | |
6754 case MOPEN + 3: | |
6755 case MOPEN + 4: | |
6756 case MOPEN + 5: | |
6757 case MOPEN + 6: | |
6758 case MOPEN + 7: | |
6759 case MOPEN + 8: | |
6760 case MOPEN + 9: | |
6761 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN); | |
6762 p = NULL; | |
6763 break; | |
6764 case MCLOSE + 0: | |
6765 p = "MATCH END"; | |
6766 break; | |
6767 case MCLOSE + 1: | |
6768 case MCLOSE + 2: | |
6769 case MCLOSE + 3: | |
6770 case MCLOSE + 4: | |
6771 case MCLOSE + 5: | |
6772 case MCLOSE + 6: | |
6773 case MCLOSE + 7: | |
6774 case MCLOSE + 8: | |
6775 case MCLOSE + 9: | |
6776 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE); | |
6777 p = NULL; | |
6778 break; | |
6779 case BACKREF + 1: | |
6780 case BACKREF + 2: | |
6781 case BACKREF + 3: | |
6782 case BACKREF + 4: | |
6783 case BACKREF + 5: | |
6784 case BACKREF + 6: | |
6785 case BACKREF + 7: | |
6786 case BACKREF + 8: | |
6787 case BACKREF + 9: | |
6788 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF); | |
6789 p = NULL; | |
6790 break; | |
6791 case NOPEN: | |
6792 p = "NOPEN"; | |
6793 break; | |
6794 case NCLOSE: | |
6795 p = "NCLOSE"; | |
6796 break; | |
6797 #ifdef FEAT_SYN_HL | |
6798 case ZOPEN + 1: | |
6799 case ZOPEN + 2: | |
6800 case ZOPEN + 3: | |
6801 case ZOPEN + 4: | |
6802 case ZOPEN + 5: | |
6803 case ZOPEN + 6: | |
6804 case ZOPEN + 7: | |
6805 case ZOPEN + 8: | |
6806 case ZOPEN + 9: | |
6807 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN); | |
6808 p = NULL; | |
6809 break; | |
6810 case ZCLOSE + 1: | |
6811 case ZCLOSE + 2: | |
6812 case ZCLOSE + 3: | |
6813 case ZCLOSE + 4: | |
6814 case ZCLOSE + 5: | |
6815 case ZCLOSE + 6: | |
6816 case ZCLOSE + 7: | |
6817 case ZCLOSE + 8: | |
6818 case ZCLOSE + 9: | |
6819 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE); | |
6820 p = NULL; | |
6821 break; | |
6822 case ZREF + 1: | |
6823 case ZREF + 2: | |
6824 case ZREF + 3: | |
6825 case ZREF + 4: | |
6826 case ZREF + 5: | |
6827 case ZREF + 6: | |
6828 case ZREF + 7: | |
6829 case ZREF + 8: | |
6830 case ZREF + 9: | |
6831 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF); | |
6832 p = NULL; | |
6833 break; | |
6834 #endif | |
6835 case STAR: | |
6836 p = "STAR"; | |
6837 break; | |
6838 case PLUS: | |
6839 p = "PLUS"; | |
6840 break; | |
6841 case NOMATCH: | |
6842 p = "NOMATCH"; | |
6843 break; | |
6844 case MATCH: | |
6845 p = "MATCH"; | |
6846 break; | |
6847 case BEHIND: | |
6848 p = "BEHIND"; | |
6849 break; | |
6850 case NOBEHIND: | |
6851 p = "NOBEHIND"; | |
6852 break; | |
6853 case SUBPAT: | |
6854 p = "SUBPAT"; | |
6855 break; | |
6856 case BRACE_LIMITS: | |
6857 p = "BRACE_LIMITS"; | |
6858 break; | |
6859 case BRACE_SIMPLE: | |
6860 p = "BRACE_SIMPLE"; | |
6861 break; | |
6862 case BRACE_COMPLEX + 0: | |
6863 case BRACE_COMPLEX + 1: | |
6864 case BRACE_COMPLEX + 2: | |
6865 case BRACE_COMPLEX + 3: | |
6866 case BRACE_COMPLEX + 4: | |
6867 case BRACE_COMPLEX + 5: | |
6868 case BRACE_COMPLEX + 6: | |
6869 case BRACE_COMPLEX + 7: | |
6870 case BRACE_COMPLEX + 8: | |
6871 case BRACE_COMPLEX + 9: | |
6872 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX); | |
6873 p = NULL; | |
6874 break; | |
6875 #ifdef FEAT_MBYTE | |
6876 case MULTIBYTECODE: | |
6877 p = "MULTIBYTECODE"; | |
6878 break; | |
6879 #endif | |
6880 case NEWL: | |
6881 p = "NEWL"; | |
6882 break; | |
6883 default: | |
6884 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op)); | |
6885 p = NULL; | |
6886 break; | |
6887 } | |
6888 if (p != NULL) | |
4444 | 6889 STRCAT(buf, p); |
6890 return (char_u *)buf; | |
7 | 6891 } |
4444 | 6892 #endif /* DEBUG */ |
7 | 6893 |
6894 #ifdef FEAT_MBYTE | |
6895 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3)); | |
6896 | |
6897 typedef struct | |
6898 { | |
6899 int a, b, c; | |
6900 } decomp_T; | |
6901 | |
6902 | |
6903 /* 0xfb20 - 0xfb4f */ | |
297 | 6904 static decomp_T decomp_table[0xfb4f-0xfb20+1] = |
7 | 6905 { |
6906 {0x5e2,0,0}, /* 0xfb20 alt ayin */ | |
6907 {0x5d0,0,0}, /* 0xfb21 alt alef */ | |
6908 {0x5d3,0,0}, /* 0xfb22 alt dalet */ | |
6909 {0x5d4,0,0}, /* 0xfb23 alt he */ | |
6910 {0x5db,0,0}, /* 0xfb24 alt kaf */ | |
6911 {0x5dc,0,0}, /* 0xfb25 alt lamed */ | |
6912 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */ | |
6913 {0x5e8,0,0}, /* 0xfb27 alt resh */ | |
6914 {0x5ea,0,0}, /* 0xfb28 alt tav */ | |
6915 {'+', 0, 0}, /* 0xfb29 alt plus */ | |
6916 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */ | |
6917 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */ | |
6918 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */ | |
6919 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */ | |
6920 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */ | |
6921 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */ | |
6922 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */ | |
6923 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */ | |
6924 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */ | |
6925 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */ | |
6926 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */ | |
6927 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */ | |
6928 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */ | |
6929 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */ | |
6930 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */ | |
6931 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */ | |
6932 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */ | |
6933 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */ | |
6934 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */ | |
6935 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */ | |
6936 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */ | |
6937 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */ | |
6938 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */ | |
6939 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */ | |
6940 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */ | |
6941 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */ | |
6942 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */ | |
6943 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */ | |
6944 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */ | |
6945 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */ | |
6946 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */ | |
6947 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */ | |
6948 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */ | |
6949 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */ | |
6950 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */ | |
6951 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */ | |
6952 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */ | |
6953 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */ | |
6954 }; | |
6955 | |
6956 static void | |
6957 mb_decompose(c, c1, c2, c3) | |
6958 int c, *c1, *c2, *c3; | |
6959 { | |
6960 decomp_T d; | |
6961 | |
4505
d037b9cbdaaa
updated for version 7.3.1000
Bram Moolenaar <bram@vim.org>
parents:
4466
diff
changeset
|
6962 if (c >= 0xfb20 && c <= 0xfb4f) |
7 | 6963 { |
6964 d = decomp_table[c - 0xfb20]; | |
6965 *c1 = d.a; | |
6966 *c2 = d.b; | |
6967 *c3 = d.c; | |
6968 } | |
6969 else | |
6970 { | |
6971 *c1 = c; | |
6972 *c2 = *c3 = 0; | |
6973 } | |
6974 } | |
6975 #endif | |
6976 | |
6977 /* | |
6978 * Compare two strings, ignore case if ireg_ic set. | |
6979 * Return 0 if strings match, non-zero otherwise. | |
6980 * Correct the length "*n" when composing characters are ignored. | |
6981 */ | |
6982 static int | |
6983 cstrncmp(s1, s2, n) | |
6984 char_u *s1, *s2; | |
6985 int *n; | |
6986 { | |
6987 int result; | |
6988 | |
6989 if (!ireg_ic) | |
6990 result = STRNCMP(s1, s2, *n); | |
6991 else | |
6992 result = MB_STRNICMP(s1, s2, *n); | |
6993 | |
6994 #ifdef FEAT_MBYTE | |
6995 /* if it failed and it's utf8 and we want to combineignore: */ | |
6996 if (result != 0 && enc_utf8 && ireg_icombine) | |
6997 { | |
6998 char_u *str1, *str2; | |
6999 int c1, c2, c11, c12; | |
7000 int junk; | |
7001 | |
7002 /* we have to handle the strcmp ourselves, since it is necessary to | |
7003 * deal with the composing characters by ignoring them: */ | |
7004 str1 = s1; | |
7005 str2 = s2; | |
7006 c1 = c2 = 0; | |
507 | 7007 while ((int)(str1 - s1) < *n) |
7 | 7008 { |
7009 c1 = mb_ptr2char_adv(&str1); | |
7010 c2 = mb_ptr2char_adv(&str2); | |
7011 | |
7012 /* decompose the character if necessary, into 'base' characters | |
7013 * because I don't care about Arabic, I will hard-code the Hebrew | |
7014 * which I *do* care about! So sue me... */ | |
7015 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2))) | |
7016 { | |
7017 /* decomposition necessary? */ | |
7018 mb_decompose(c1, &c11, &junk, &junk); | |
7019 mb_decompose(c2, &c12, &junk, &junk); | |
7020 c1 = c11; | |
7021 c2 = c12; | |
7022 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12))) | |
7023 break; | |
7024 } | |
7025 } | |
7026 result = c2 - c1; | |
7027 if (result == 0) | |
7028 *n = (int)(str2 - s2); | |
7029 } | |
7030 #endif | |
7031 | |
7032 return result; | |
7033 } | |
7034 | |
7035 /* | |
7036 * cstrchr: This function is used a lot for simple searches, keep it fast! | |
7037 */ | |
7038 static char_u * | |
7039 cstrchr(s, c) | |
7040 char_u *s; | |
7041 int c; | |
7042 { | |
7043 char_u *p; | |
7044 int cc; | |
7045 | |
7046 if (!ireg_ic | |
7047 #ifdef FEAT_MBYTE | |
7048 || (!enc_utf8 && mb_char2len(c) > 1) | |
7049 #endif | |
7050 ) | |
7051 return vim_strchr(s, c); | |
7052 | |
7053 /* tolower() and toupper() can be slow, comparing twice should be a lot | |
7054 * faster (esp. when using MS Visual C++!). | |
7055 * For UTF-8 need to use folded case. */ | |
7056 #ifdef FEAT_MBYTE | |
7057 if (enc_utf8 && c > 0x80) | |
7058 cc = utf_fold(c); | |
7059 else | |
7060 #endif | |
1347 | 7061 if (MB_ISUPPER(c)) |
7062 cc = MB_TOLOWER(c); | |
7063 else if (MB_ISLOWER(c)) | |
7064 cc = MB_TOUPPER(c); | |
7 | 7065 else |
7066 return vim_strchr(s, c); | |
7067 | |
7068 #ifdef FEAT_MBYTE | |
7069 if (has_mbyte) | |
7070 { | |
474 | 7071 for (p = s; *p != NUL; p += (*mb_ptr2len)(p)) |
7 | 7072 { |
7073 if (enc_utf8 && c > 0x80) | |
7074 { | |
7075 if (utf_fold(utf_ptr2char(p)) == cc) | |
7076 return p; | |
7077 } | |
7078 else if (*p == c || *p == cc) | |
7079 return p; | |
7080 } | |
7081 } | |
7082 else | |
7083 #endif | |
7084 /* Faster version for when there are no multi-byte characters. */ | |
7085 for (p = s; *p != NUL; ++p) | |
7086 if (*p == c || *p == cc) | |
7087 return p; | |
7088 | |
7089 return NULL; | |
7090 } | |
7091 | |
7092 /*************************************************************** | |
7093 * regsub stuff * | |
7094 ***************************************************************/ | |
7095 | |
7096 /* This stuff below really confuses cc on an SGI -- webb */ | |
7097 #ifdef __sgi | |
7098 # undef __ARGS | |
7099 # define __ARGS(x) () | |
7100 #endif | |
7101 | |
7102 /* | |
7103 * We should define ftpr as a pointer to a function returning a pointer to | |
7104 * a function returning a pointer to a function ... | |
7105 * This is impossible, so we declare a pointer to a function returning a | |
7106 * pointer to a function returning void. This should work for all compilers. | |
7107 */ | |
772 | 7108 typedef void (*(*fptr_T) __ARGS((int *, int)))(); |
7109 | |
7110 static fptr_T do_upper __ARGS((int *, int)); | |
7111 static fptr_T do_Upper __ARGS((int *, int)); | |
7112 static fptr_T do_lower __ARGS((int *, int)); | |
7113 static fptr_T do_Lower __ARGS((int *, int)); | |
7 | 7114 |
7115 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash)); | |
7116 | |
772 | 7117 static fptr_T |
7 | 7118 do_upper(d, c) |
772 | 7119 int *d; |
7120 int c; | |
7 | 7121 { |
772 | 7122 *d = MB_TOUPPER(c); |
7123 | |
7124 return (fptr_T)NULL; | |
7 | 7125 } |
7126 | |
772 | 7127 static fptr_T |
7128 do_Upper(d, c) | |
7129 int *d; | |
7 | 7130 int c; |
7131 { | |
772 | 7132 *d = MB_TOUPPER(c); |
7133 | |
7134 return (fptr_T)do_Upper; | |
7135 } | |
7136 | |
7137 static fptr_T | |
7138 do_lower(d, c) | |
7139 int *d; | |
7140 int c; | |
7141 { | |
7142 *d = MB_TOLOWER(c); | |
7143 | |
7144 return (fptr_T)NULL; | |
7145 } | |
7146 | |
7147 static fptr_T | |
7148 do_Lower(d, c) | |
7149 int *d; | |
7150 int c; | |
7151 { | |
7152 *d = MB_TOLOWER(c); | |
7153 | |
7154 return (fptr_T)do_Lower; | |
7 | 7155 } |
7156 | |
7157 /* | |
7158 * regtilde(): Replace tildes in the pattern by the old pattern. | |
7159 * | |
7160 * Short explanation of the tilde: It stands for the previous replacement | |
7161 * pattern. If that previous pattern also contains a ~ we should go back a | |
7162 * step further... But we insert the previous pattern into the current one | |
7163 * and remember that. | |
772 | 7164 * This still does not handle the case where "magic" changes. So require the |
7165 * user to keep his hands off of "magic". | |
7 | 7166 * |
7167 * The tildes are parsed once before the first call to vim_regsub(). | |
7168 */ | |
7169 char_u * | |
7170 regtilde(source, magic) | |
7171 char_u *source; | |
7172 int magic; | |
7173 { | |
7174 char_u *newsub = source; | |
7175 char_u *tmpsub; | |
7176 char_u *p; | |
7177 int len; | |
7178 int prevlen; | |
7179 | |
7180 for (p = newsub; *p; ++p) | |
7181 { | |
7182 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic)) | |
7183 { | |
7184 if (reg_prev_sub != NULL) | |
7185 { | |
7186 /* length = len(newsub) - 1 + len(prev_sub) + 1 */ | |
7187 prevlen = (int)STRLEN(reg_prev_sub); | |
7188 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen)); | |
7189 if (tmpsub != NULL) | |
7190 { | |
7191 /* copy prefix */ | |
7192 len = (int)(p - newsub); /* not including ~ */ | |
7193 mch_memmove(tmpsub, newsub, (size_t)len); | |
1209 | 7194 /* interpret tilde */ |
7 | 7195 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen); |
7196 /* copy postfix */ | |
7197 if (!magic) | |
7198 ++p; /* back off \ */ | |
7199 STRCPY(tmpsub + len + prevlen, p + 1); | |
7200 | |
7201 if (newsub != source) /* already allocated newsub */ | |
7202 vim_free(newsub); | |
7203 newsub = tmpsub; | |
7204 p = newsub + len + prevlen; | |
7205 } | |
7206 } | |
7207 else if (magic) | |
1621 | 7208 STRMOVE(p, p + 1); /* remove '~' */ |
7 | 7209 else |
1621 | 7210 STRMOVE(p, p + 2); /* remove '\~' */ |
7 | 7211 --p; |
7212 } | |
7213 else | |
7214 { | |
7215 if (*p == '\\' && p[1]) /* skip escaped characters */ | |
7216 ++p; | |
7217 #ifdef FEAT_MBYTE | |
7218 if (has_mbyte) | |
474 | 7219 p += (*mb_ptr2len)(p) - 1; |
7 | 7220 #endif |
7221 } | |
7222 } | |
7223 | |
7224 vim_free(reg_prev_sub); | |
7225 if (newsub != source) /* newsub was allocated, just keep it */ | |
7226 reg_prev_sub = newsub; | |
7227 else /* no ~ found, need to save newsub */ | |
7228 reg_prev_sub = vim_strsave(newsub); | |
7229 return newsub; | |
7230 } | |
7231 | |
7232 #ifdef FEAT_EVAL | |
7233 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */ | |
7234 | |
7235 /* These pointers are used instead of reg_match and reg_mmatch for | |
7236 * reg_submatch(). Needed for when the substitution string is an expression | |
7237 * that contains a call to substitute() and submatch(). */ | |
7238 static regmatch_T *submatch_match; | |
7239 static regmmatch_T *submatch_mmatch; | |
2011 | 7240 static linenr_T submatch_firstlnum; |
7241 static linenr_T submatch_maxline; | |
2904 | 7242 static int submatch_line_lbr; |
7 | 7243 #endif |
7244 | |
7245 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO) | |
7246 /* | |
7247 * vim_regsub() - perform substitutions after a vim_regexec() or | |
7248 * vim_regexec_multi() match. | |
7249 * | |
7250 * If "copy" is TRUE really copy into "dest". | |
7251 * If "copy" is FALSE nothing is copied, this is just to find out the length | |
7252 * of the result. | |
7253 * | |
7254 * If "backslash" is TRUE, a backslash will be removed later, need to double | |
7255 * them to keep them, and insert a backslash before a CR to avoid it being | |
7256 * replaced with a line break later. | |
7257 * | |
7258 * Note: The matched text must not change between the call of | |
7259 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back | |
7260 * references invalid! | |
7261 * | |
7262 * Returns the size of the replacement, including terminating NUL. | |
7263 */ | |
7264 int | |
7265 vim_regsub(rmp, source, dest, copy, magic, backslash) | |
7266 regmatch_T *rmp; | |
7267 char_u *source; | |
7268 char_u *dest; | |
7269 int copy; | |
7270 int magic; | |
7271 int backslash; | |
7272 { | |
7273 reg_match = rmp; | |
7274 reg_mmatch = NULL; | |
7275 reg_maxline = 0; | |
4061 | 7276 reg_buf = curbuf; |
7 | 7277 return vim_regsub_both(source, dest, copy, magic, backslash); |
7278 } | |
7279 #endif | |
7280 | |
7281 int | |
7282 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash) | |
7283 regmmatch_T *rmp; | |
7284 linenr_T lnum; | |
7285 char_u *source; | |
7286 char_u *dest; | |
7287 int copy; | |
7288 int magic; | |
7289 int backslash; | |
7290 { | |
7291 reg_match = NULL; | |
7292 reg_mmatch = rmp; | |
7293 reg_buf = curbuf; /* always works on the current buffer! */ | |
7294 reg_firstlnum = lnum; | |
7295 reg_maxline = curbuf->b_ml.ml_line_count - lnum; | |
7296 return vim_regsub_both(source, dest, copy, magic, backslash); | |
7297 } | |
7298 | |
7299 static int | |
7300 vim_regsub_both(source, dest, copy, magic, backslash) | |
7301 char_u *source; | |
7302 char_u *dest; | |
7303 int copy; | |
7304 int magic; | |
7305 int backslash; | |
7306 { | |
7307 char_u *src; | |
7308 char_u *dst; | |
7309 char_u *s; | |
7310 int c; | |
772 | 7311 int cc; |
7 | 7312 int no = -1; |
4244 | 7313 fptr_T func_all = (fptr_T)NULL; |
7314 fptr_T func_one = (fptr_T)NULL; | |
7 | 7315 linenr_T clnum = 0; /* init for GCC */ |
7316 int len = 0; /* init for GCC */ | |
7317 #ifdef FEAT_EVAL | |
7318 static char_u *eval_result = NULL; | |
7319 #endif | |
7320 | |
7321 /* Be paranoid... */ | |
7322 if (source == NULL || dest == NULL) | |
7323 { | |
7324 EMSG(_(e_null)); | |
7325 return 0; | |
7326 } | |
7327 if (prog_magic_wrong()) | |
7328 return 0; | |
7329 src = source; | |
7330 dst = dest; | |
7331 | |
7332 /* | |
7333 * When the substitute part starts with "\=" evaluate it as an expression. | |
7334 */ | |
7335 if (source[0] == '\\' && source[1] == '=' | |
7336 #ifdef FEAT_EVAL | |
7337 && !can_f_submatch /* can't do this recursively */ | |
7338 #endif | |
7339 ) | |
7340 { | |
7341 #ifdef FEAT_EVAL | |
7342 /* To make sure that the length doesn't change between checking the | |
7343 * length and copying the string, and to speed up things, the | |
7344 * resulting string is saved from the call with "copy" == FALSE to the | |
7345 * call with "copy" == TRUE. */ | |
7346 if (copy) | |
7347 { | |
7348 if (eval_result != NULL) | |
7349 { | |
7350 STRCPY(dest, eval_result); | |
7351 dst += STRLEN(eval_result); | |
7352 vim_free(eval_result); | |
7353 eval_result = NULL; | |
7354 } | |
7355 } | |
7356 else | |
7357 { | |
7358 win_T *save_reg_win; | |
7359 int save_ireg_ic; | |
7360 | |
7361 vim_free(eval_result); | |
7362 | |
7363 /* The expression may contain substitute(), which calls us | |
7364 * recursively. Make sure submatch() gets the text from the first | |
7365 * level. Don't need to save "reg_buf", because | |
7366 * vim_regexec_multi() can't be called recursively. */ | |
7367 submatch_match = reg_match; | |
7368 submatch_mmatch = reg_mmatch; | |
2011 | 7369 submatch_firstlnum = reg_firstlnum; |
7370 submatch_maxline = reg_maxline; | |
2904 | 7371 submatch_line_lbr = reg_line_lbr; |
7 | 7372 save_reg_win = reg_win; |
7373 save_ireg_ic = ireg_ic; | |
7374 can_f_submatch = TRUE; | |
7375 | |
714 | 7376 eval_result = eval_to_string(source + 2, NULL, TRUE); |
7 | 7377 if (eval_result != NULL) |
7378 { | |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7379 int had_backslash = FALSE; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7380 |
39 | 7381 for (s = eval_result; *s != NUL; mb_ptr_adv(s)) |
7 | 7382 { |
2904 | 7383 /* Change NL to CR, so that it becomes a line break, |
7384 * unless called from vim_regexec_nl(). | |
7 | 7385 * Skip over a backslashed character. */ |
2904 | 7386 if (*s == NL && !submatch_line_lbr) |
7 | 7387 *s = CAR; |
7388 else if (*s == '\\' && s[1] != NUL) | |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7389 { |
7 | 7390 ++s; |
2173 | 7391 /* Change NL to CR here too, so that this works: |
7392 * :s/abc\\\ndef/\="aaa\\\nbbb"/ on text: | |
7393 * abc\ | |
7394 * def | |
2904 | 7395 * Not when called from vim_regexec_nl(). |
2173 | 7396 */ |
2904 | 7397 if (*s == NL && !submatch_line_lbr) |
2173 | 7398 *s = CAR; |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7399 had_backslash = TRUE; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7400 } |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7401 } |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7402 if (had_backslash && backslash) |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7403 { |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7404 /* Backslashes will be consumed, need to double them. */ |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7405 s = vim_strsave_escaped(eval_result, (char_u *)"\\"); |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7406 if (s != NULL) |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7407 { |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7408 vim_free(eval_result); |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7409 eval_result = s; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7410 } |
7 | 7411 } |
7412 | |
7413 dst += STRLEN(eval_result); | |
7414 } | |
7415 | |
7416 reg_match = submatch_match; | |
7417 reg_mmatch = submatch_mmatch; | |
2011 | 7418 reg_firstlnum = submatch_firstlnum; |
7419 reg_maxline = submatch_maxline; | |
2904 | 7420 reg_line_lbr = submatch_line_lbr; |
7 | 7421 reg_win = save_reg_win; |
7422 ireg_ic = save_ireg_ic; | |
7423 can_f_submatch = FALSE; | |
7424 } | |
7425 #endif | |
7426 } | |
7427 else | |
7428 while ((c = *src++) != NUL) | |
7429 { | |
7430 if (c == '&' && magic) | |
7431 no = 0; | |
7432 else if (c == '\\' && *src != NUL) | |
7433 { | |
7434 if (*src == '&' && !magic) | |
7435 { | |
7436 ++src; | |
7437 no = 0; | |
7438 } | |
7439 else if ('0' <= *src && *src <= '9') | |
7440 { | |
7441 no = *src++ - '0'; | |
7442 } | |
7443 else if (vim_strchr((char_u *)"uUlLeE", *src)) | |
7444 { | |
7445 switch (*src++) | |
7446 { | |
4244 | 7447 case 'u': func_one = (fptr_T)do_upper; |
7 | 7448 continue; |
4244 | 7449 case 'U': func_all = (fptr_T)do_Upper; |
7 | 7450 continue; |
4244 | 7451 case 'l': func_one = (fptr_T)do_lower; |
7 | 7452 continue; |
4244 | 7453 case 'L': func_all = (fptr_T)do_Lower; |
7 | 7454 continue; |
7455 case 'e': | |
4244 | 7456 case 'E': func_one = func_all = (fptr_T)NULL; |
7 | 7457 continue; |
7458 } | |
7459 } | |
7460 } | |
7461 if (no < 0) /* Ordinary character. */ | |
7462 { | |
798 | 7463 if (c == K_SPECIAL && src[0] != NUL && src[1] != NUL) |
7464 { | |
1209 | 7465 /* Copy a special key as-is. */ |
798 | 7466 if (copy) |
7467 { | |
7468 *dst++ = c; | |
7469 *dst++ = *src++; | |
7470 *dst++ = *src++; | |
7471 } | |
7472 else | |
7473 { | |
7474 dst += 3; | |
7475 src += 2; | |
7476 } | |
7477 continue; | |
7478 } | |
7479 | |
7 | 7480 if (c == '\\' && *src != NUL) |
7481 { | |
7482 /* Check for abbreviations -- webb */ | |
7483 switch (*src) | |
7484 { | |
7485 case 'r': c = CAR; ++src; break; | |
7486 case 'n': c = NL; ++src; break; | |
7487 case 't': c = TAB; ++src; break; | |
7488 /* Oh no! \e already has meaning in subst pat :-( */ | |
7489 /* case 'e': c = ESC; ++src; break; */ | |
7490 case 'b': c = Ctrl_H; ++src; break; | |
7491 | |
7492 /* If "backslash" is TRUE the backslash will be removed | |
7493 * later. Used to insert a literal CR. */ | |
7494 default: if (backslash) | |
7495 { | |
7496 if (copy) | |
7497 *dst = '\\'; | |
7498 ++dst; | |
7499 } | |
7500 c = *src++; | |
7501 } | |
7502 } | |
798 | 7503 #ifdef FEAT_MBYTE |
7504 else if (has_mbyte) | |
7505 c = mb_ptr2char(src - 1); | |
7506 #endif | |
7 | 7507 |
7508 /* Write to buffer, if copy is set. */ | |
4244 | 7509 if (func_one != (fptr_T)NULL) |
7510 /* Turbo C complains without the typecast */ | |
7511 func_one = (fptr_T)(func_one(&cc, c)); | |
7512 else if (func_all != (fptr_T)NULL) | |
7513 /* Turbo C complains without the typecast */ | |
7514 func_all = (fptr_T)(func_all(&cc, c)); | |
7515 else /* just copy */ | |
772 | 7516 cc = c; |
7517 | |
7518 #ifdef FEAT_MBYTE | |
7519 if (has_mbyte) | |
7 | 7520 { |
2307
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7521 int totlen = mb_ptr2len(src - 1); |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7522 |
7 | 7523 if (copy) |
772 | 7524 mb_char2bytes(cc, dst); |
7525 dst += mb_char2len(cc) - 1; | |
2307
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7526 if (enc_utf8) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7527 { |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7528 int clen = utf_ptr2len(src - 1); |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7529 |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7530 /* If the character length is shorter than "totlen", there |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7531 * are composing characters; copy them as-is. */ |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7532 if (clen < totlen) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7533 { |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7534 if (copy) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7535 mch_memmove(dst + 1, src - 1 + clen, |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7536 (size_t)(totlen - clen)); |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7537 dst += totlen - clen; |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7538 } |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7539 } |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7540 src += totlen - 1; |
7 | 7541 } |
7542 else | |
7543 #endif | |
7544 if (copy) | |
772 | 7545 *dst = cc; |
7 | 7546 dst++; |
7547 } | |
7548 else | |
7549 { | |
7550 if (REG_MULTI) | |
7551 { | |
7552 clnum = reg_mmatch->startpos[no].lnum; | |
7553 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0) | |
7554 s = NULL; | |
7555 else | |
7556 { | |
7557 s = reg_getline(clnum) + reg_mmatch->startpos[no].col; | |
7558 if (reg_mmatch->endpos[no].lnum == clnum) | |
7559 len = reg_mmatch->endpos[no].col | |
7560 - reg_mmatch->startpos[no].col; | |
7561 else | |
7562 len = (int)STRLEN(s); | |
7563 } | |
7564 } | |
7565 else | |
7566 { | |
7567 s = reg_match->startp[no]; | |
7568 if (reg_match->endp[no] == NULL) | |
7569 s = NULL; | |
7570 else | |
7571 len = (int)(reg_match->endp[no] - s); | |
7572 } | |
7573 if (s != NULL) | |
7574 { | |
7575 for (;;) | |
7576 { | |
7577 if (len == 0) | |
7578 { | |
7579 if (REG_MULTI) | |
7580 { | |
7581 if (reg_mmatch->endpos[no].lnum == clnum) | |
7582 break; | |
7583 if (copy) | |
7584 *dst = CAR; | |
7585 ++dst; | |
7586 s = reg_getline(++clnum); | |
7587 if (reg_mmatch->endpos[no].lnum == clnum) | |
7588 len = reg_mmatch->endpos[no].col; | |
7589 else | |
7590 len = (int)STRLEN(s); | |
7591 } | |
7592 else | |
7593 break; | |
7594 } | |
7595 else if (*s == NUL) /* we hit NUL. */ | |
7596 { | |
7597 if (copy) | |
7598 EMSG(_(e_re_damg)); | |
7599 goto exit; | |
7600 } | |
7601 else | |
7602 { | |
7603 if (backslash && (*s == CAR || *s == '\\')) | |
7604 { | |
7605 /* | |
7606 * Insert a backslash in front of a CR, otherwise | |
7607 * it will be replaced by a line break. | |
7608 * Number of backslashes will be halved later, | |
7609 * double them here. | |
7610 */ | |
7611 if (copy) | |
7612 { | |
7613 dst[0] = '\\'; | |
7614 dst[1] = *s; | |
7615 } | |
7616 dst += 2; | |
7617 } | |
7618 else | |
7619 { | |
772 | 7620 #ifdef FEAT_MBYTE |
7621 if (has_mbyte) | |
7622 c = mb_ptr2char(s); | |
7623 else | |
7624 #endif | |
7625 c = *s; | |
7626 | |
4244 | 7627 if (func_one != (fptr_T)NULL) |
7628 /* Turbo C complains without the typecast */ | |
7629 func_one = (fptr_T)(func_one(&cc, c)); | |
7630 else if (func_all != (fptr_T)NULL) | |
7631 /* Turbo C complains without the typecast */ | |
7632 func_all = (fptr_T)(func_all(&cc, c)); | |
7633 else /* just copy */ | |
772 | 7634 cc = c; |
7635 | |
7636 #ifdef FEAT_MBYTE | |
7637 if (has_mbyte) | |
7 | 7638 { |
1332 | 7639 int l; |
7640 | |
7641 /* Copy composing characters separately, one | |
7642 * at a time. */ | |
7643 if (enc_utf8) | |
7644 l = utf_ptr2len(s) - 1; | |
7645 else | |
7646 l = mb_ptr2len(s) - 1; | |
772 | 7647 |
7648 s += l; | |
7649 len -= l; | |
7650 if (copy) | |
7651 mb_char2bytes(cc, dst); | |
7652 dst += mb_char2len(cc) - 1; | |
7 | 7653 } |
772 | 7654 else |
7655 #endif | |
7656 if (copy) | |
7657 *dst = cc; | |
7658 dst++; | |
7 | 7659 } |
772 | 7660 |
7 | 7661 ++s; |
7662 --len; | |
7663 } | |
7664 } | |
7665 } | |
7666 no = -1; | |
7667 } | |
7668 } | |
7669 if (copy) | |
7670 *dst = NUL; | |
7671 | |
7672 exit: | |
7673 return (int)((dst - dest) + 1); | |
7674 } | |
7675 | |
7676 #ifdef FEAT_EVAL | |
2012 | 7677 static char_u *reg_getline_submatch __ARGS((linenr_T lnum)); |
7678 | |
7 | 7679 /* |
2011 | 7680 * Call reg_getline() with the line numbers from the submatch. If a |
7681 * substitute() was used the reg_maxline and other values have been | |
7682 * overwritten. | |
7683 */ | |
7684 static char_u * | |
7685 reg_getline_submatch(lnum) | |
7686 linenr_T lnum; | |
7687 { | |
7688 char_u *s; | |
7689 linenr_T save_first = reg_firstlnum; | |
7690 linenr_T save_max = reg_maxline; | |
7691 | |
7692 reg_firstlnum = submatch_firstlnum; | |
7693 reg_maxline = submatch_maxline; | |
7694 | |
7695 s = reg_getline(lnum); | |
7696 | |
7697 reg_firstlnum = save_first; | |
7698 reg_maxline = save_max; | |
7699 return s; | |
7700 } | |
7701 | |
7702 /* | |
1209 | 7703 * Used for the submatch() function: get the string from the n'th submatch in |
7 | 7704 * allocated memory. |
7705 * Returns NULL when not in a ":s" command and for a non-existing submatch. | |
7706 */ | |
7707 char_u * | |
7708 reg_submatch(no) | |
7709 int no; | |
7710 { | |
7711 char_u *retval = NULL; | |
7712 char_u *s; | |
7713 int len; | |
7714 int round; | |
7715 linenr_T lnum; | |
7716 | |
840 | 7717 if (!can_f_submatch || no < 0) |
7 | 7718 return NULL; |
7719 | |
7720 if (submatch_match == NULL) | |
7721 { | |
7722 /* | |
7723 * First round: compute the length and allocate memory. | |
7724 * Second round: copy the text. | |
7725 */ | |
7726 for (round = 1; round <= 2; ++round) | |
7727 { | |
7728 lnum = submatch_mmatch->startpos[no].lnum; | |
7729 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0) | |
7730 return NULL; | |
7731 | |
2011 | 7732 s = reg_getline_submatch(lnum) + submatch_mmatch->startpos[no].col; |
7 | 7733 if (s == NULL) /* anti-crash check, cannot happen? */ |
7734 break; | |
7735 if (submatch_mmatch->endpos[no].lnum == lnum) | |
7736 { | |
7737 /* Within one line: take form start to end col. */ | |
7738 len = submatch_mmatch->endpos[no].col | |
7739 - submatch_mmatch->startpos[no].col; | |
7740 if (round == 2) | |
418 | 7741 vim_strncpy(retval, s, len); |
7 | 7742 ++len; |
7743 } | |
7744 else | |
7745 { | |
7746 /* Multiple lines: take start line from start col, middle | |
7747 * lines completely and end line up to end col. */ | |
7748 len = (int)STRLEN(s); | |
7749 if (round == 2) | |
7750 { | |
7751 STRCPY(retval, s); | |
7752 retval[len] = '\n'; | |
7753 } | |
7754 ++len; | |
7755 ++lnum; | |
7756 while (lnum < submatch_mmatch->endpos[no].lnum) | |
7757 { | |
2011 | 7758 s = reg_getline_submatch(lnum++); |
7 | 7759 if (round == 2) |
7760 STRCPY(retval + len, s); | |
7761 len += (int)STRLEN(s); | |
7762 if (round == 2) | |
7763 retval[len] = '\n'; | |
7764 ++len; | |
7765 } | |
7766 if (round == 2) | |
2011 | 7767 STRNCPY(retval + len, reg_getline_submatch(lnum), |
7 | 7768 submatch_mmatch->endpos[no].col); |
7769 len += submatch_mmatch->endpos[no].col; | |
7770 if (round == 2) | |
7771 retval[len] = NUL; | |
7772 ++len; | |
7773 } | |
7774 | |
840 | 7775 if (retval == NULL) |
7 | 7776 { |
7777 retval = lalloc((long_u)len, TRUE); | |
840 | 7778 if (retval == NULL) |
7 | 7779 return NULL; |
7780 } | |
7781 } | |
7782 } | |
7783 else | |
7784 { | |
1815 | 7785 s = submatch_match->startp[no]; |
7786 if (s == NULL || submatch_match->endp[no] == NULL) | |
7 | 7787 retval = NULL; |
7788 else | |
7789 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s)); | |
7790 } | |
7791 | |
7792 return retval; | |
7793 } | |
7794 #endif | |
4444 | 7795 |
7796 static regengine_T bt_regengine = | |
7797 { | |
7798 bt_regcomp, | |
7799 bt_regexec, | |
7800 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7801 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7802 bt_regexec_nl, | |
7803 #endif | |
7804 bt_regexec_multi | |
7805 #ifdef DEBUG | |
7806 ,(char_u *)"" | |
7807 #endif | |
7808 }; | |
7809 | |
7810 | |
7811 #include "regexp_nfa.c" | |
7812 | |
7813 static regengine_T nfa_regengine = | |
7814 { | |
7815 nfa_regcomp, | |
7816 nfa_regexec, | |
7817 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7818 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7819 nfa_regexec_nl, | |
7820 #endif | |
7821 nfa_regexec_multi | |
7822 #ifdef DEBUG | |
7823 ,(char_u *)"" | |
7824 #endif | |
7825 }; | |
7826 | |
7827 /* Which regexp engine to use? Needed for vim_regcomp(). | |
7828 * Must match with 'regexpengine'. */ | |
7829 static int regexp_engine = 0; | |
7830 #define AUTOMATIC_ENGINE 0 | |
7831 #define BACKTRACKING_ENGINE 1 | |
7832 #define NFA_ENGINE 2 | |
7833 #ifdef DEBUG | |
7834 static char_u regname[][30] = { | |
7835 "AUTOMATIC Regexp Engine", | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
7836 "BACKTRACKING Regexp Engine", |
4444 | 7837 "NFA Regexp Engine" |
7838 }; | |
7839 #endif | |
7840 | |
7841 /* | |
7842 * Compile a regular expression into internal code. | |
7843 * Returns the program in allocated memory. Returns NULL for an error. | |
7844 */ | |
7845 regprog_T * | |
7846 vim_regcomp(expr_arg, re_flags) | |
7847 char_u *expr_arg; | |
7848 int re_flags; | |
7849 { | |
7850 regprog_T *prog = NULL; | |
7851 char_u *expr = expr_arg; | |
7852 | |
7853 syntax_error = FALSE; | |
7854 regexp_engine = p_re; | |
7855 | |
7856 /* Check for prefix "\%#=", that sets the regexp engine */ | |
7857 if (STRNCMP(expr, "\\%#=", 4) == 0) | |
7858 { | |
7859 int newengine = expr[4] - '0'; | |
7860 | |
7861 if (newengine == AUTOMATIC_ENGINE | |
7862 || newengine == BACKTRACKING_ENGINE | |
7863 || newengine == NFA_ENGINE) | |
7864 { | |
7865 regexp_engine = expr[4] - '0'; | |
7866 expr += 5; | |
7867 #ifdef DEBUG | |
7868 EMSG3("New regexp mode selected (%d): %s", regexp_engine, | |
7869 regname[newengine]); | |
7870 #endif | |
7871 } | |
7872 else | |
7873 { | |
7874 EMSG(_("E864: \\%#= can only be followed by 0, 1, or 2. The automatic engine will be used ")); | |
7875 regexp_engine = AUTOMATIC_ENGINE; | |
7876 } | |
7877 } | |
7878 #ifdef DEBUG | |
7879 bt_regengine.expr = expr; | |
7880 nfa_regengine.expr = expr; | |
7881 #endif | |
7882 | |
7883 /* | |
7884 * First try the NFA engine, unless backtracking was requested. | |
7885 */ | |
7886 if (regexp_engine != BACKTRACKING_ENGINE) | |
7887 prog = nfa_regengine.regcomp(expr, re_flags); | |
7888 else | |
7889 prog = bt_regengine.regcomp(expr, re_flags); | |
7890 | |
7891 if (prog == NULL) /* error compiling regexp with initial engine */ | |
7892 { | |
4460 | 7893 #ifdef BT_REGEXP_DEBUG_LOG |
4444 | 7894 if (regexp_engine != BACKTRACKING_ENGINE) /* debugging log for NFA */ |
7895 { | |
7896 FILE *f; | |
4460 | 7897 f = fopen(BT_REGEXP_DEBUG_LOG_NAME, "a"); |
4444 | 7898 if (f) |
7899 { | |
7900 if (!syntax_error) | |
7901 fprintf(f, "NFA engine could not handle \"%s\"\n", expr); | |
7902 else | |
7903 fprintf(f, "Syntax error in \"%s\"\n", expr); | |
7904 fclose(f); | |
7905 } | |
7906 else | |
4460 | 7907 EMSG2("(NFA) Could not open \"%s\" to write !!!", |
7908 BT_REGEXP_DEBUG_LOG_NAME); | |
4444 | 7909 /* |
7910 if (syntax_error) | |
7911 EMSG("NFA Regexp: Syntax Error !"); | |
7912 */ | |
7913 } | |
7914 #endif | |
7915 /* | |
7916 * If NFA engine failed, then revert to the backtracking engine. | |
7917 * Except when there was a syntax error, which was properly handled by | |
7918 * NFA engine. | |
7919 */ | |
7920 if (regexp_engine == AUTOMATIC_ENGINE) | |
7921 if (!syntax_error) | |
7922 prog = bt_regengine.regcomp(expr, re_flags); | |
7923 | |
7924 } /* endif prog==NULL */ | |
7925 | |
7926 | |
7927 return prog; | |
7928 } | |
7929 | |
7930 /* | |
7931 * Match a regexp against a string. | |
7932 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
7933 * Uses curbuf for line count and 'iskeyword'. | |
7934 * | |
7935 * Return TRUE if there is a match, FALSE if not. | |
7936 */ | |
7937 int | |
7938 vim_regexec(rmp, line, col) | |
7939 regmatch_T *rmp; | |
7940 char_u *line; /* string to match against */ | |
7941 colnr_T col; /* column to start looking for match */ | |
7942 { | |
7943 return rmp->regprog->engine->regexec(rmp, line, col); | |
7944 } | |
7945 | |
7946 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7947 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7948 /* | |
7949 * Like vim_regexec(), but consider a "\n" in "line" to be a line break. | |
7950 */ | |
7951 int | |
7952 vim_regexec_nl(rmp, line, col) | |
7953 regmatch_T *rmp; | |
7954 char_u *line; | |
7955 colnr_T col; | |
7956 { | |
7957 return rmp->regprog->engine->regexec_nl(rmp, line, col); | |
7958 } | |
7959 #endif | |
7960 | |
7961 /* | |
7962 * Match a regexp against multiple lines. | |
7963 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
7964 * Uses curbuf for line count and 'iskeyword'. | |
7965 * | |
7966 * Return zero if there is no match. Return number of lines contained in the | |
7967 * match otherwise. | |
7968 */ | |
7969 long | |
7970 vim_regexec_multi(rmp, win, buf, lnum, col, tm) | |
7971 regmmatch_T *rmp; | |
7972 win_T *win; /* window in which to search or NULL */ | |
7973 buf_T *buf; /* buffer in which to search */ | |
7974 linenr_T lnum; /* nr of line to start looking for match */ | |
7975 colnr_T col; /* column to start looking for match */ | |
7976 proftime_T *tm; /* timeout limit or NULL */ | |
7977 { | |
7978 return rmp->regprog->engine->regexec_multi(rmp, win, buf, lnum, col, tm); | |
7979 } |