Mercurial > vim
annotate src/regexp.c @ 4444:ccecb03e5e8b v7.3.970
updated for version 7.3.970
Problem: Syntax highlighting can be slow.
Solution: Include the NFA regexp engine. Add the 'regexpengine' option to
select which one is used. (various authors, including Ken Takata,
Andrei Aiordachioaie, Russ Cox, Xiaozhou Liua, Ian Young)
author | Bram Moolenaar <bram@vim.org> |
---|---|
date | Sun, 19 May 2013 19:40:29 +0200 |
parents | 7faeece39228 |
children | fe8a0a6a1c2a |
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 | |
54 #endif | |
7 | 55 |
56 /* | |
57 * The "internal use only" fields in regexp.h are present to pass info from | |
58 * compile to execute that permits the execute phase to run lots faster on | |
59 * simple cases. They are: | |
60 * | |
61 * regstart char that must begin a match; NUL if none obvious; Can be a | |
62 * multi-byte character. | |
63 * reganch is the match anchored (at beginning-of-line only)? | |
64 * regmust string (pointer into program) that match must include, or NULL | |
65 * regmlen length of regmust string | |
66 * regflags RF_ values or'ed together | |
67 * | |
68 * Regstart and reganch permit very fast decisions on suitable starting points | |
69 * for a match, cutting down the work a lot. Regmust permits fast rejection | |
70 * of lines that cannot possibly match. The regmust tests are costly enough | |
71 * that vim_regcomp() supplies a regmust only if the r.e. contains something | |
72 * potentially expensive (at present, the only such thing detected is * or + | |
73 * at the start of the r.e., which can involve a lot of backup). Regmlen is | |
74 * supplied because the test in vim_regexec() needs it and vim_regcomp() is | |
75 * computing it anyway. | |
76 */ | |
77 | |
78 /* | |
79 * Structure for regexp "program". This is essentially a linear encoding | |
80 * of a nondeterministic finite-state machine (aka syntax charts or | |
81 * "railroad normal form" in parsing technology). Each node is an opcode | |
82 * plus a "next" pointer, possibly plus an operand. "Next" pointers of | |
83 * all nodes except BRANCH and BRACES_COMPLEX implement concatenation; a "next" | |
84 * pointer with a BRANCH on both ends of it is connecting two alternatives. | |
85 * (Here we have one of the subtle syntax dependencies: an individual BRANCH | |
86 * (as opposed to a collection of them) is never concatenated with anything | |
87 * because of operator precedence). The "next" pointer of a BRACES_COMPLEX | |
167 | 88 * node points to the node after the stuff to be repeated. |
89 * The operand of some types of node is a literal string; for others, it is a | |
90 * node leading into a sub-FSM. In particular, the operand of a BRANCH node | |
91 * is the first node of the branch. | |
92 * (NB this is *not* a tree structure: the tail of the branch connects to the | |
93 * thing following the set of BRANCHes.) | |
7 | 94 * |
95 * pattern is coded like: | |
96 * | |
97 * +-----------------+ | |
98 * | V | |
99 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END | |
100 * | ^ | ^ | |
101 * +------+ +----------+ | |
102 * | |
103 * | |
104 * +------------------+ | |
105 * V | | |
106 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END | |
107 * | | ^ ^ | |
108 * | +---------------+ | | |
109 * +---------------------------------------------+ | |
110 * | |
111 * | |
167 | 112 * +----------------------+ |
113 * V | | |
233 | 114 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END |
856 | 115 * | | ^ ^ |
116 * | +-----------+ | | |
179 | 117 * +--------------------------------------------------+ |
167 | 118 * |
119 * | |
7 | 120 * +-------------------------+ |
121 * V | | |
122 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END | |
123 * | | ^ | |
124 * | +----------------+ | |
125 * +-----------------------------------------------+ | |
126 * | |
127 * | |
128 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END | |
129 * | | ^ ^ | |
130 * | +----------------+ | | |
131 * +--------------------------------+ | |
132 * | |
133 * +---------+ | |
134 * | V | |
135 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END | |
136 * | | | | ^ ^ | |
137 * | | | +-----+ | | |
138 * | | +----------------+ | | |
139 * | +---------------------------+ | | |
140 * +------------------------------------------------------+ | |
141 * | |
1209 | 142 * They all start with a BRANCH for "\|" alternatives, even when there is only |
7 | 143 * one alternative. |
144 */ | |
145 | |
146 /* | |
147 * The opcodes are: | |
148 */ | |
149 | |
150 /* definition number opnd? meaning */ | |
151 #define END 0 /* End of program or NOMATCH operand. */ | |
152 #define BOL 1 /* Match "" at beginning of line. */ | |
153 #define EOL 2 /* Match "" at end of line. */ | |
154 #define BRANCH 3 /* node Match this alternative, or the | |
155 * next... */ | |
156 #define BACK 4 /* Match "", "next" ptr points backward. */ | |
157 #define EXACTLY 5 /* str Match this string. */ | |
158 #define NOTHING 6 /* Match empty string. */ | |
159 #define STAR 7 /* node Match this (simple) thing 0 or more | |
160 * times. */ | |
161 #define PLUS 8 /* node Match this (simple) thing 1 or more | |
162 * times. */ | |
163 #define MATCH 9 /* node match the operand zero-width */ | |
164 #define NOMATCH 10 /* node check for no match with operand */ | |
165 #define BEHIND 11 /* node look behind for a match with operand */ | |
166 #define NOBEHIND 12 /* node look behind for no match with operand */ | |
167 #define SUBPAT 13 /* node match the operand here */ | |
168 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and | |
169 * n times (\{m,n\}). */ | |
170 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */ | |
171 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */ | |
172 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE | |
173 * and BRACE_COMPLEX. */ | |
174 #define NEWL 18 /* Match line-break */ | |
175 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */ | |
176 | |
177 | |
178 /* character classes: 20-48 normal, 50-78 include a line-break */ | |
179 #define ADD_NL 30 | |
180 #define FIRST_NL ANY + ADD_NL | |
181 #define ANY 20 /* Match any one character. */ | |
182 #define ANYOF 21 /* str Match any character in this string. */ | |
183 #define ANYBUT 22 /* str Match any character not in this | |
184 * string. */ | |
185 #define IDENT 23 /* Match identifier char */ | |
186 #define SIDENT 24 /* Match identifier char but no digit */ | |
187 #define KWORD 25 /* Match keyword char */ | |
188 #define SKWORD 26 /* Match word char but no digit */ | |
189 #define FNAME 27 /* Match file name char */ | |
190 #define SFNAME 28 /* Match file name char but no digit */ | |
191 #define PRINT 29 /* Match printable char */ | |
192 #define SPRINT 30 /* Match printable char but no digit */ | |
193 #define WHITE 31 /* Match whitespace char */ | |
194 #define NWHITE 32 /* Match non-whitespace char */ | |
195 #define DIGIT 33 /* Match digit char */ | |
196 #define NDIGIT 34 /* Match non-digit char */ | |
197 #define HEX 35 /* Match hex char */ | |
198 #define NHEX 36 /* Match non-hex char */ | |
199 #define OCTAL 37 /* Match octal char */ | |
200 #define NOCTAL 38 /* Match non-octal char */ | |
201 #define WORD 39 /* Match word char */ | |
202 #define NWORD 40 /* Match non-word char */ | |
203 #define HEAD 41 /* Match head char */ | |
204 #define NHEAD 42 /* Match non-head char */ | |
205 #define ALPHA 43 /* Match alpha char */ | |
206 #define NALPHA 44 /* Match non-alpha char */ | |
207 #define LOWER 45 /* Match lowercase char */ | |
208 #define NLOWER 46 /* Match non-lowercase char */ | |
209 #define UPPER 47 /* Match uppercase char */ | |
210 #define NUPPER 48 /* Match non-uppercase char */ | |
211 #define LAST_NL NUPPER + ADD_NL | |
212 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL) | |
213 | |
214 #define MOPEN 80 /* -89 Mark this point in input as start of | |
215 * \( subexpr. MOPEN + 0 marks start of | |
216 * match. */ | |
217 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks | |
218 * end of match. */ | |
219 #define BACKREF 100 /* -109 node Match same string again \1-\9 */ | |
220 | |
221 #ifdef FEAT_SYN_HL | |
222 # define ZOPEN 110 /* -119 Mark this point in input as start of | |
223 * \z( subexpr. */ | |
224 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */ | |
225 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */ | |
226 #endif | |
227 | |
228 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */ | |
229 | |
230 #define NOPEN 150 /* Mark this point in input as start of | |
231 \%( subexpr. */ | |
232 #define NCLOSE 151 /* Analogous to NOPEN. */ | |
233 | |
234 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */ | |
235 #define RE_BOF 201 /* Match "" at beginning of file. */ | |
236 #define RE_EOF 202 /* Match "" at end of file. */ | |
237 #define CURSOR 203 /* Match location of cursor. */ | |
238 | |
239 #define RE_LNUM 204 /* nr cmp Match line number */ | |
240 #define RE_COL 205 /* nr cmp Match column number */ | |
241 #define RE_VCOL 206 /* nr cmp Match virtual column number */ | |
242 | |
639 | 243 #define RE_MARK 207 /* mark cmp Match mark position */ |
244 #define RE_VISUAL 208 /* Match Visual area */ | |
245 | |
7 | 246 /* |
247 * Magic characters have a special meaning, they don't match literally. | |
248 * Magic characters are negative. This separates them from literal characters | |
249 * (possibly multi-byte). Only ASCII characters can be Magic. | |
250 */ | |
251 #define Magic(x) ((int)(x) - 256) | |
252 #define un_Magic(x) ((x) + 256) | |
253 #define is_Magic(x) ((x) < 0) | |
254 | |
255 static int no_Magic __ARGS((int x)); | |
256 static int toggle_Magic __ARGS((int x)); | |
257 | |
258 static int | |
259 no_Magic(x) | |
260 int x; | |
261 { | |
262 if (is_Magic(x)) | |
263 return un_Magic(x); | |
264 return x; | |
265 } | |
266 | |
267 static int | |
268 toggle_Magic(x) | |
269 int x; | |
270 { | |
271 if (is_Magic(x)) | |
272 return un_Magic(x); | |
273 return Magic(x); | |
274 } | |
275 | |
276 /* | |
277 * The first byte of the regexp internal "program" is actually this magic | |
278 * number; the start node begins in the second byte. It's used to catch the | |
279 * most severe mutilation of the program by the caller. | |
280 */ | |
281 | |
282 #define REGMAGIC 0234 | |
283 | |
284 /* | |
285 * Opcode notes: | |
286 * | |
287 * BRANCH The set of branches constituting a single choice are hooked | |
288 * together with their "next" pointers, since precedence prevents | |
289 * anything being concatenated to any individual branch. The | |
290 * "next" pointer of the last BRANCH in a choice points to the | |
291 * thing following the whole choice. This is also where the | |
292 * final "next" pointer of each individual branch points; each | |
293 * branch starts with the operand node of a BRANCH node. | |
294 * | |
295 * BACK Normal "next" pointers all implicitly point forward; BACK | |
296 * exists to make loop structures possible. | |
297 * | |
298 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular | |
299 * BRANCH structures using BACK. Simple cases (one character | |
300 * per match) are implemented with STAR and PLUS for speed | |
301 * and to minimize recursive plunges. | |
302 * | |
303 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX | |
304 * node, and defines the min and max limits to be used for that | |
305 * node. | |
306 * | |
307 * MOPEN,MCLOSE ...are numbered at compile time. | |
308 * ZOPEN,ZCLOSE ...ditto | |
309 */ | |
310 | |
311 /* | |
312 * A node is one char of opcode followed by two chars of "next" pointer. | |
313 * "Next" pointers are stored as two 8-bit bytes, high order first. The | |
314 * value is a positive offset from the opcode of the node containing it. | |
315 * An operand, if any, simply follows the node. (Note that much of the | |
316 * code generation knows about this implicit relationship.) | |
317 * | |
318 * Using two bytes for the "next" pointer is vast overkill for most things, | |
319 * but allows patterns to get big without disasters. | |
320 */ | |
321 #define OP(p) ((int)*(p)) | |
322 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377)) | |
323 #define OPERAND(p) ((p) + 3) | |
324 /* Obtain an operand that was stored as four bytes, MSB first. */ | |
325 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \ | |
326 + ((long)(p)[5] << 8) + (long)(p)[6]) | |
327 /* Obtain a second operand stored as four bytes. */ | |
328 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4) | |
329 /* Obtain a second single-byte operand stored after a four bytes operand. */ | |
330 #define OPERAND_CMP(p) (p)[7] | |
331 | |
332 /* | |
333 * Utility definitions. | |
334 */ | |
335 #define UCHARAT(p) ((int)*(char_u *)(p)) | |
336 | |
337 /* Used for an error (down from) vim_regcomp(): give the error message, set | |
338 * rc_did_emsg and return NULL */ | |
653 | 339 #define EMSG_RET_NULL(m) return (EMSG(m), rc_did_emsg = TRUE, (void *)NULL) |
308 | 340 #define EMSG_RET_FAIL(m) return (EMSG(m), rc_did_emsg = TRUE, FAIL) |
4444 | 341 #define EMSG2_RET_NULL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, (void *)NULL) |
342 #define EMSG2_RET_FAIL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, FAIL) | |
343 #define EMSG_ONE_RET_NULL EMSG2_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL) | |
7 | 344 |
345 #define MAX_LIMIT (32767L << 16L) | |
346 | |
347 static int re_multi_type __ARGS((int)); | |
348 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n)); | |
349 static char_u *cstrchr __ARGS((char_u *, int)); | |
350 | |
4444 | 351 #ifdef BT_REGEXP_DUMP |
352 static void regdump __ARGS((char_u *, bt_regprog_T *)); | |
353 #endif | |
7 | 354 #ifdef DEBUG |
355 static char_u *regprop __ARGS((char_u *)); | |
356 #endif | |
357 | |
4444 | 358 static char_u e_missingbracket[] = N_("E769: Missing ] after %s["); |
359 static char_u e_unmatchedpp[] = N_("E53: Unmatched %s%%("); | |
360 static char_u e_unmatchedp[] = N_("E54: Unmatched %s("); | |
361 static char_u e_unmatchedpar[] = N_("E55: Unmatched %s)"); | |
362 | |
7 | 363 #define NOT_MULTI 0 |
364 #define MULTI_ONE 1 | |
365 #define MULTI_MULT 2 | |
366 /* | |
367 * Return NOT_MULTI if c is not a "multi" operator. | |
368 * Return MULTI_ONE if c is a single "multi" operator. | |
369 * Return MULTI_MULT if c is a multi "multi" operator. | |
370 */ | |
371 static int | |
372 re_multi_type(c) | |
373 int c; | |
374 { | |
375 if (c == Magic('@') || c == Magic('=') || c == Magic('?')) | |
376 return MULTI_ONE; | |
377 if (c == Magic('*') || c == Magic('+') || c == Magic('{')) | |
378 return MULTI_MULT; | |
379 return NOT_MULTI; | |
380 } | |
381 | |
382 /* | |
383 * Flags to be passed up and down. | |
384 */ | |
385 #define HASWIDTH 0x1 /* Known never to match null string. */ | |
386 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */ | |
387 #define SPSTART 0x4 /* Starts with * or +. */ | |
388 #define HASNL 0x8 /* Contains some \n. */ | |
389 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */ | |
390 #define WORST 0 /* Worst case. */ | |
391 | |
392 /* | |
393 * When regcode is set to this value, code is not emitted and size is computed | |
394 * instead. | |
395 */ | |
396 #define JUST_CALC_SIZE ((char_u *) -1) | |
397 | |
359 | 398 static char_u *reg_prev_sub = NULL; |
399 | |
7 | 400 /* |
401 * REGEXP_INRANGE contains all characters which are always special in a [] | |
402 * range after '\'. | |
403 * REGEXP_ABBR contains all characters which act as abbreviations after '\'. | |
404 * These are: | |
405 * \n - New line (NL). | |
406 * \r - Carriage Return (CR). | |
407 * \t - Tab (TAB). | |
408 * \e - Escape (ESC). | |
409 * \b - Backspace (Ctrl_H). | |
24 | 410 * \d - Character code in decimal, eg \d123 |
411 * \o - Character code in octal, eg \o80 | |
412 * \x - Character code in hex, eg \x4a | |
413 * \u - Multibyte character code, eg \u20ac | |
414 * \U - Long multibyte character code, eg \U12345678 | |
7 | 415 */ |
416 static char_u REGEXP_INRANGE[] = "]^-n\\"; | |
24 | 417 static char_u REGEXP_ABBR[] = "nrtebdoxuU"; |
7 | 418 |
419 static int backslash_trans __ARGS((int c)); | |
167 | 420 static int get_char_class __ARGS((char_u **pp)); |
421 static int get_equi_class __ARGS((char_u **pp)); | |
422 static void reg_equi_class __ARGS((int c)); | |
423 static int get_coll_element __ARGS((char_u **pp)); | |
7 | 424 static char_u *skip_anyof __ARGS((char_u *p)); |
425 static void init_class_tab __ARGS((void)); | |
426 | |
427 /* | |
428 * Translate '\x' to its control character, except "\n", which is Magic. | |
429 */ | |
430 static int | |
431 backslash_trans(c) | |
432 int c; | |
433 { | |
434 switch (c) | |
435 { | |
436 case 'r': return CAR; | |
437 case 't': return TAB; | |
438 case 'e': return ESC; | |
439 case 'b': return BS; | |
440 } | |
441 return c; | |
442 } | |
443 | |
444 /* | |
167 | 445 * Check for a character class name "[:name:]". "pp" points to the '['. |
7 | 446 * Returns one of the CLASS_ items. CLASS_NONE means that no item was |
447 * recognized. Otherwise "pp" is advanced to after the item. | |
448 */ | |
449 static int | |
167 | 450 get_char_class(pp) |
7 | 451 char_u **pp; |
452 { | |
453 static const char *(class_names[]) = | |
454 { | |
455 "alnum:]", | |
456 #define CLASS_ALNUM 0 | |
457 "alpha:]", | |
458 #define CLASS_ALPHA 1 | |
459 "blank:]", | |
460 #define CLASS_BLANK 2 | |
461 "cntrl:]", | |
462 #define CLASS_CNTRL 3 | |
463 "digit:]", | |
464 #define CLASS_DIGIT 4 | |
465 "graph:]", | |
466 #define CLASS_GRAPH 5 | |
467 "lower:]", | |
468 #define CLASS_LOWER 6 | |
469 "print:]", | |
470 #define CLASS_PRINT 7 | |
471 "punct:]", | |
472 #define CLASS_PUNCT 8 | |
473 "space:]", | |
474 #define CLASS_SPACE 9 | |
475 "upper:]", | |
476 #define CLASS_UPPER 10 | |
477 "xdigit:]", | |
478 #define CLASS_XDIGIT 11 | |
479 "tab:]", | |
480 #define CLASS_TAB 12 | |
481 "return:]", | |
482 #define CLASS_RETURN 13 | |
483 "backspace:]", | |
484 #define CLASS_BACKSPACE 14 | |
485 "escape:]", | |
486 #define CLASS_ESCAPE 15 | |
487 }; | |
488 #define CLASS_NONE 99 | |
489 int i; | |
490 | |
491 if ((*pp)[1] == ':') | |
492 { | |
1877 | 493 for (i = 0; i < (int)(sizeof(class_names) / sizeof(*class_names)); ++i) |
7 | 494 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0) |
495 { | |
496 *pp += STRLEN(class_names[i]) + 2; | |
497 return i; | |
498 } | |
499 } | |
500 return CLASS_NONE; | |
501 } | |
502 | |
503 /* | |
504 * Specific version of character class functions. | |
505 * Using a table to keep this fast. | |
506 */ | |
507 static short class_tab[256]; | |
508 | |
509 #define RI_DIGIT 0x01 | |
510 #define RI_HEX 0x02 | |
511 #define RI_OCTAL 0x04 | |
512 #define RI_WORD 0x08 | |
513 #define RI_HEAD 0x10 | |
514 #define RI_ALPHA 0x20 | |
515 #define RI_LOWER 0x40 | |
516 #define RI_UPPER 0x80 | |
517 #define RI_WHITE 0x100 | |
518 | |
519 static void | |
520 init_class_tab() | |
521 { | |
522 int i; | |
523 static int done = FALSE; | |
524 | |
525 if (done) | |
526 return; | |
527 | |
528 for (i = 0; i < 256; ++i) | |
529 { | |
530 if (i >= '0' && i <= '7') | |
531 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD; | |
532 else if (i >= '8' && i <= '9') | |
533 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD; | |
534 else if (i >= 'a' && i <= 'f') | |
535 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER; | |
536 #ifdef EBCDIC | |
537 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r') | |
538 || (i >= 's' && i <= 'z')) | |
539 #else | |
540 else if (i >= 'g' && i <= 'z') | |
541 #endif | |
542 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER; | |
543 else if (i >= 'A' && i <= 'F') | |
544 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER; | |
545 #ifdef EBCDIC | |
546 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R') | |
547 || (i >= 'S' && i <= 'Z')) | |
548 #else | |
549 else if (i >= 'G' && i <= 'Z') | |
550 #endif | |
551 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER; | |
552 else if (i == '_') | |
553 class_tab[i] = RI_WORD + RI_HEAD; | |
554 else | |
555 class_tab[i] = 0; | |
556 } | |
557 class_tab[' '] |= RI_WHITE; | |
558 class_tab['\t'] |= RI_WHITE; | |
559 done = TRUE; | |
560 } | |
561 | |
562 #ifdef FEAT_MBYTE | |
563 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT)) | |
564 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX)) | |
565 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL)) | |
566 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD)) | |
567 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD)) | |
568 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA)) | |
569 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER)) | |
570 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER)) | |
571 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE)) | |
572 #else | |
573 # define ri_digit(c) (class_tab[c] & RI_DIGIT) | |
574 # define ri_hex(c) (class_tab[c] & RI_HEX) | |
575 # define ri_octal(c) (class_tab[c] & RI_OCTAL) | |
576 # define ri_word(c) (class_tab[c] & RI_WORD) | |
577 # define ri_head(c) (class_tab[c] & RI_HEAD) | |
578 # define ri_alpha(c) (class_tab[c] & RI_ALPHA) | |
579 # define ri_lower(c) (class_tab[c] & RI_LOWER) | |
580 # define ri_upper(c) (class_tab[c] & RI_UPPER) | |
581 # define ri_white(c) (class_tab[c] & RI_WHITE) | |
582 #endif | |
583 | |
584 /* flags for regflags */ | |
585 #define RF_ICASE 1 /* ignore case */ | |
586 #define RF_NOICASE 2 /* don't ignore case */ | |
587 #define RF_HASNL 4 /* can match a NL */ | |
588 #define RF_ICOMBINE 8 /* ignore combining characters */ | |
589 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */ | |
590 | |
591 /* | |
592 * Global work variables for vim_regcomp(). | |
593 */ | |
594 | |
595 static char_u *regparse; /* Input-scan pointer. */ | |
596 static int prevchr_len; /* byte length of previous char */ | |
597 static int num_complex_braces; /* Complex \{...} count */ | |
598 static int regnpar; /* () count. */ | |
599 #ifdef FEAT_SYN_HL | |
600 static int regnzpar; /* \z() count. */ | |
601 static int re_has_z; /* \z item detected */ | |
602 #endif | |
603 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */ | |
604 static long regsize; /* Code size. */ | |
2010 | 605 static int reg_toolong; /* TRUE when offset out of range */ |
7 | 606 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */ |
607 static unsigned regflags; /* RF_ flags for prog */ | |
608 static long brace_min[10]; /* Minimums for complex brace repeats */ | |
609 static long brace_max[10]; /* Maximums for complex brace repeats */ | |
610 static int brace_count[10]; /* Current counts for complex brace repeats */ | |
611 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
612 static int had_eol; /* TRUE when EOL found by vim_regcomp() */ | |
613 #endif | |
614 static int one_exactly = FALSE; /* only do one char for EXACTLY */ | |
615 | |
616 static int reg_magic; /* magicness of the pattern: */ | |
617 #define MAGIC_NONE 1 /* "\V" very unmagic */ | |
618 #define MAGIC_OFF 2 /* "\M" or 'magic' off */ | |
619 #define MAGIC_ON 3 /* "\m" or 'magic' */ | |
620 #define MAGIC_ALL 4 /* "\v" very magic */ | |
621 | |
622 static int reg_string; /* matching with a string instead of a buffer | |
623 line */ | |
481 | 624 static int reg_strict; /* "[abc" is illegal */ |
7 | 625 |
626 /* | |
627 * META contains all characters that may be magic, except '^' and '$'. | |
628 */ | |
629 | |
630 #ifdef EBCDIC | |
631 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~"; | |
632 #else | |
633 /* META[] is used often enough to justify turning it into a table. */ | |
634 static char_u META_flags[] = { | |
635 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, | |
636 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, | |
637 /* % & ( ) * + . */ | |
638 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, | |
639 /* 1 2 3 4 5 6 7 8 9 < = > ? */ | |
640 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, | |
641 /* @ A C D F H I K L M O */ | |
642 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, | |
643 /* P S U V W X Z [ _ */ | |
644 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1, | |
645 /* a c d f h i k l m n o */ | |
646 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, | |
647 /* p s u v w x z { | ~ */ | |
648 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1 | |
649 }; | |
650 #endif | |
651 | |
4444 | 652 static int curchr; /* currently parsed character */ |
653 /* Previous character. Note: prevchr is sometimes -1 when we are not at the | |
654 * start, eg in /[ ^I]^ the pattern was never found even if it existed, | |
655 * because ^ was taken to be magic -- webb */ | |
656 static int prevchr; | |
657 static int prevprevchr; /* previous-previous character */ | |
658 static int nextchr; /* used for ungetchr() */ | |
7 | 659 |
660 /* arguments for reg() */ | |
661 #define REG_NOPAREN 0 /* toplevel reg() */ | |
662 #define REG_PAREN 1 /* \(\) */ | |
663 #define REG_ZPAREN 2 /* \z(\) */ | |
664 #define REG_NPAREN 3 /* \%(\) */ | |
665 | |
666 /* | |
667 * Forward declarations for vim_regcomp()'s friends. | |
668 */ | |
669 static void initchr __ARGS((char_u *)); | |
670 static int getchr __ARGS((void)); | |
671 static void skipchr_keepstart __ARGS((void)); | |
672 static int peekchr __ARGS((void)); | |
673 static void skipchr __ARGS((void)); | |
674 static void ungetchr __ARGS((void)); | |
24 | 675 static int gethexchrs __ARGS((int maxinputlen)); |
676 static int getoctchrs __ARGS((void)); | |
677 static int getdecchrs __ARGS((void)); | |
678 static int coll_get_char __ARGS((void)); | |
7 | 679 static void regcomp_start __ARGS((char_u *expr, int flags)); |
680 static char_u *reg __ARGS((int, int *)); | |
681 static char_u *regbranch __ARGS((int *flagp)); | |
682 static char_u *regconcat __ARGS((int *flagp)); | |
683 static char_u *regpiece __ARGS((int *)); | |
684 static char_u *regatom __ARGS((int *)); | |
685 static char_u *regnode __ARGS((int)); | |
714 | 686 #ifdef FEAT_MBYTE |
687 static int use_multibytecode __ARGS((int c)); | |
688 #endif | |
7 | 689 static int prog_magic_wrong __ARGS((void)); |
690 static char_u *regnext __ARGS((char_u *)); | |
691 static void regc __ARGS((int b)); | |
692 #ifdef FEAT_MBYTE | |
693 static void regmbc __ARGS((int c)); | |
2974 | 694 # define REGMBC(x) regmbc(x); |
695 # define CASEMBC(x) case x: | |
167 | 696 #else |
697 # define regmbc(c) regc(c) | |
2974 | 698 # define REGMBC(x) |
699 # define CASEMBC(x) | |
7 | 700 #endif |
701 static void reginsert __ARGS((int, char_u *)); | |
702 static void reginsert_limits __ARGS((int, long, long, char_u *)); | |
703 static char_u *re_put_long __ARGS((char_u *pr, long_u val)); | |
704 static int read_limits __ARGS((long *, long *)); | |
705 static void regtail __ARGS((char_u *, char_u *)); | |
706 static void regoptail __ARGS((char_u *, char_u *)); | |
707 | |
4444 | 708 static regengine_T bt_regengine; |
709 static regengine_T nfa_regengine; | |
710 | |
7 | 711 /* |
712 * Return TRUE if compiled regular expression "prog" can match a line break. | |
713 */ | |
714 int | |
715 re_multiline(prog) | |
716 regprog_T *prog; | |
717 { | |
718 return (prog->regflags & RF_HASNL); | |
719 } | |
720 | |
721 /* | |
722 * Return TRUE if compiled regular expression "prog" looks before the start | |
723 * position (pattern contains "\@<=" or "\@<!"). | |
724 */ | |
725 int | |
726 re_lookbehind(prog) | |
727 regprog_T *prog; | |
728 { | |
729 return (prog->regflags & RF_LOOKBH); | |
730 } | |
731 | |
732 /* | |
167 | 733 * Check for an equivalence class name "[=a=]". "pp" points to the '['. |
734 * Returns a character representing the class. Zero means that no item was | |
735 * recognized. Otherwise "pp" is advanced to after the item. | |
736 */ | |
737 static int | |
738 get_equi_class(pp) | |
739 char_u **pp; | |
740 { | |
741 int c; | |
742 int l = 1; | |
743 char_u *p = *pp; | |
744 | |
745 if (p[1] == '=') | |
746 { | |
747 #ifdef FEAT_MBYTE | |
748 if (has_mbyte) | |
474 | 749 l = (*mb_ptr2len)(p + 2); |
167 | 750 #endif |
751 if (p[l + 2] == '=' && p[l + 3] == ']') | |
752 { | |
753 #ifdef FEAT_MBYTE | |
754 if (has_mbyte) | |
755 c = mb_ptr2char(p + 2); | |
756 else | |
757 #endif | |
758 c = p[2]; | |
759 *pp += l + 4; | |
760 return c; | |
761 } | |
762 } | |
763 return 0; | |
764 } | |
765 | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
766 #ifdef EBCDIC |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
767 /* |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
768 * 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
|
769 */ |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
770 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
|
771 "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
|
772 "C\x68", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
773 "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
|
774 "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
|
775 "N\x69", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
776 "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
|
777 "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
|
778 "Y\xBA", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
779 "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
|
780 "c\x48", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
781 "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
|
782 "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
|
783 "n\x49", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
784 "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
|
785 "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
|
786 "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
|
787 }; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
788 #endif |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
789 |
167 | 790 /* |
791 * Produce the bytes for equivalence class "c". | |
792 * Currently only handles latin1, latin9 and utf-8. | |
4444 | 793 * NOTE: When changing this function, also change nfa_emit_equi_class() |
167 | 794 */ |
795 static void | |
796 reg_equi_class(c) | |
797 int c; | |
798 { | |
799 #ifdef FEAT_MBYTE | |
800 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0 | |
492 | 801 || STRCMP(p_enc, "iso-8859-15") == 0) |
167 | 802 #endif |
803 { | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
804 #ifdef EBCDIC |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
805 int i; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
806 |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
807 /* 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
|
808 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
|
809 { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
810 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
|
811 { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
812 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
|
813 |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
814 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
|
815 regmbc(*p++); |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
816 return; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
817 } |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
818 } |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
819 #else |
167 | 820 switch (c) |
821 { | |
236 | 822 case 'A': case '\300': case '\301': case '\302': |
2974 | 823 CASEMBC(0x100) CASEMBC(0x102) CASEMBC(0x104) CASEMBC(0x1cd) |
824 CASEMBC(0x1de) CASEMBC(0x1e0) CASEMBC(0x1ea2) | |
236 | 825 case '\303': case '\304': case '\305': |
826 regmbc('A'); regmbc('\300'); regmbc('\301'); | |
827 regmbc('\302'); regmbc('\303'); regmbc('\304'); | |
828 regmbc('\305'); | |
2974 | 829 REGMBC(0x100) REGMBC(0x102) REGMBC(0x104) |
830 REGMBC(0x1cd) REGMBC(0x1de) REGMBC(0x1e0) | |
831 REGMBC(0x1ea2) | |
832 return; | |
833 case 'B': CASEMBC(0x1e02) CASEMBC(0x1e06) | |
834 regmbc('B'); REGMBC(0x1e02) REGMBC(0x1e06) | |
167 | 835 return; |
236 | 836 case 'C': case '\307': |
2974 | 837 CASEMBC(0x106) CASEMBC(0x108) CASEMBC(0x10a) CASEMBC(0x10c) |
236 | 838 regmbc('C'); regmbc('\307'); |
2974 | 839 REGMBC(0x106) REGMBC(0x108) REGMBC(0x10a) |
840 REGMBC(0x10c) | |
841 return; | |
842 case 'D': CASEMBC(0x10e) CASEMBC(0x110) CASEMBC(0x1e0a) | |
843 CASEMBC(0x1e0e) CASEMBC(0x1e10) | |
844 regmbc('D'); REGMBC(0x10e) REGMBC(0x110) | |
845 REGMBC(0x1e0a) REGMBC(0x1e0e) REGMBC(0x1e10) | |
167 | 846 return; |
236 | 847 case 'E': case '\310': case '\311': case '\312': case '\313': |
2974 | 848 CASEMBC(0x112) CASEMBC(0x114) CASEMBC(0x116) CASEMBC(0x118) |
849 CASEMBC(0x11a) CASEMBC(0x1eba) CASEMBC(0x1ebc) | |
236 | 850 regmbc('E'); regmbc('\310'); regmbc('\311'); |
851 regmbc('\312'); regmbc('\313'); | |
2974 | 852 REGMBC(0x112) REGMBC(0x114) REGMBC(0x116) |
853 REGMBC(0x118) REGMBC(0x11a) REGMBC(0x1eba) | |
854 REGMBC(0x1ebc) | |
855 return; | |
856 case 'F': CASEMBC(0x1e1e) | |
857 regmbc('F'); REGMBC(0x1e1e) | |
858 return; | |
859 case 'G': CASEMBC(0x11c) CASEMBC(0x11e) CASEMBC(0x120) | |
860 CASEMBC(0x122) CASEMBC(0x1e4) CASEMBC(0x1e6) CASEMBC(0x1f4) | |
861 CASEMBC(0x1e20) | |
862 regmbc('G'); REGMBC(0x11c) REGMBC(0x11e) | |
863 REGMBC(0x120) REGMBC(0x122) REGMBC(0x1e4) | |
864 REGMBC(0x1e6) REGMBC(0x1f4) REGMBC(0x1e20) | |
865 return; | |
866 case 'H': CASEMBC(0x124) CASEMBC(0x126) CASEMBC(0x1e22) | |
867 CASEMBC(0x1e26) CASEMBC(0x1e28) | |
868 regmbc('H'); REGMBC(0x124) REGMBC(0x126) | |
869 REGMBC(0x1e22) REGMBC(0x1e26) REGMBC(0x1e28) | |
167 | 870 return; |
236 | 871 case 'I': case '\314': case '\315': case '\316': case '\317': |
2974 | 872 CASEMBC(0x128) CASEMBC(0x12a) CASEMBC(0x12c) CASEMBC(0x12e) |
873 CASEMBC(0x130) CASEMBC(0x1cf) CASEMBC(0x1ec8) | |
236 | 874 regmbc('I'); regmbc('\314'); regmbc('\315'); |
875 regmbc('\316'); regmbc('\317'); | |
2974 | 876 REGMBC(0x128) REGMBC(0x12a) REGMBC(0x12c) |
877 REGMBC(0x12e) REGMBC(0x130) REGMBC(0x1cf) | |
878 REGMBC(0x1ec8) | |
879 return; | |
880 case 'J': CASEMBC(0x134) | |
881 regmbc('J'); REGMBC(0x134) | |
882 return; | |
883 case 'K': CASEMBC(0x136) CASEMBC(0x1e8) CASEMBC(0x1e30) | |
884 CASEMBC(0x1e34) | |
885 regmbc('K'); REGMBC(0x136) REGMBC(0x1e8) | |
886 REGMBC(0x1e30) REGMBC(0x1e34) | |
887 return; | |
888 case 'L': CASEMBC(0x139) CASEMBC(0x13b) CASEMBC(0x13d) | |
889 CASEMBC(0x13f) CASEMBC(0x141) CASEMBC(0x1e3a) | |
890 regmbc('L'); REGMBC(0x139) REGMBC(0x13b) | |
891 REGMBC(0x13d) REGMBC(0x13f) REGMBC(0x141) | |
892 REGMBC(0x1e3a) | |
893 return; | |
894 case 'M': CASEMBC(0x1e3e) CASEMBC(0x1e40) | |
895 regmbc('M'); REGMBC(0x1e3e) REGMBC(0x1e40) | |
167 | 896 return; |
236 | 897 case 'N': case '\321': |
2974 | 898 CASEMBC(0x143) CASEMBC(0x145) CASEMBC(0x147) CASEMBC(0x1e44) |
899 CASEMBC(0x1e48) | |
236 | 900 regmbc('N'); regmbc('\321'); |
2974 | 901 REGMBC(0x143) REGMBC(0x145) REGMBC(0x147) |
902 REGMBC(0x1e44) REGMBC(0x1e48) | |
167 | 903 return; |
236 | 904 case 'O': case '\322': case '\323': case '\324': case '\325': |
2974 | 905 case '\326': case '\330': |
906 CASEMBC(0x14c) CASEMBC(0x14e) CASEMBC(0x150) CASEMBC(0x1a0) | |
907 CASEMBC(0x1d1) CASEMBC(0x1ea) CASEMBC(0x1ec) CASEMBC(0x1ece) | |
236 | 908 regmbc('O'); regmbc('\322'); regmbc('\323'); |
909 regmbc('\324'); regmbc('\325'); regmbc('\326'); | |
2974 | 910 regmbc('\330'); |
911 REGMBC(0x14c) REGMBC(0x14e) REGMBC(0x150) | |
912 REGMBC(0x1a0) REGMBC(0x1d1) REGMBC(0x1ea) | |
913 REGMBC(0x1ec) REGMBC(0x1ece) | |
914 return; | |
915 case 'P': case 0x1e54: case 0x1e56: | |
916 regmbc('P'); REGMBC(0x1e54) REGMBC(0x1e56) | |
917 return; | |
918 case 'R': CASEMBC(0x154) CASEMBC(0x156) CASEMBC(0x158) | |
919 CASEMBC(0x1e58) CASEMBC(0x1e5e) | |
920 regmbc('R'); REGMBC(0x154) REGMBC(0x156) REGMBC(0x158) | |
921 REGMBC(0x1e58) REGMBC(0x1e5e) | |
922 return; | |
923 case 'S': CASEMBC(0x15a) CASEMBC(0x15c) CASEMBC(0x15e) | |
924 CASEMBC(0x160) CASEMBC(0x1e60) | |
925 regmbc('S'); REGMBC(0x15a) REGMBC(0x15c) | |
926 REGMBC(0x15e) REGMBC(0x160) REGMBC(0x1e60) | |
927 return; | |
928 case 'T': CASEMBC(0x162) CASEMBC(0x164) CASEMBC(0x166) | |
929 CASEMBC(0x1e6a) CASEMBC(0x1e6e) | |
930 regmbc('T'); REGMBC(0x162) REGMBC(0x164) | |
931 REGMBC(0x166) REGMBC(0x1e6a) REGMBC(0x1e6e) | |
167 | 932 return; |
236 | 933 case 'U': case '\331': case '\332': case '\333': case '\334': |
2974 | 934 CASEMBC(0x168) CASEMBC(0x16a) CASEMBC(0x16c) CASEMBC(0x16e) |
935 CASEMBC(0x170) CASEMBC(0x172) CASEMBC(0x1af) CASEMBC(0x1d3) | |
936 CASEMBC(0x1ee6) | |
236 | 937 regmbc('U'); regmbc('\331'); regmbc('\332'); |
938 regmbc('\333'); regmbc('\334'); | |
2974 | 939 REGMBC(0x168) REGMBC(0x16a) REGMBC(0x16c) |
940 REGMBC(0x16e) REGMBC(0x170) REGMBC(0x172) | |
941 REGMBC(0x1af) REGMBC(0x1d3) REGMBC(0x1ee6) | |
942 return; | |
943 case 'V': CASEMBC(0x1e7c) | |
944 regmbc('V'); REGMBC(0x1e7c) | |
945 return; | |
946 case 'W': CASEMBC(0x174) CASEMBC(0x1e80) CASEMBC(0x1e82) | |
947 CASEMBC(0x1e84) CASEMBC(0x1e86) | |
948 regmbc('W'); REGMBC(0x174) REGMBC(0x1e80) | |
949 REGMBC(0x1e82) REGMBC(0x1e84) REGMBC(0x1e86) | |
950 return; | |
951 case 'X': CASEMBC(0x1e8a) CASEMBC(0x1e8c) | |
952 regmbc('X'); REGMBC(0x1e8a) REGMBC(0x1e8c) | |
167 | 953 return; |
236 | 954 case 'Y': case '\335': |
2974 | 955 CASEMBC(0x176) CASEMBC(0x178) CASEMBC(0x1e8e) CASEMBC(0x1ef2) |
956 CASEMBC(0x1ef6) CASEMBC(0x1ef8) | |
236 | 957 regmbc('Y'); regmbc('\335'); |
2974 | 958 REGMBC(0x176) REGMBC(0x178) REGMBC(0x1e8e) |
959 REGMBC(0x1ef2) REGMBC(0x1ef6) REGMBC(0x1ef8) | |
960 return; | |
961 case 'Z': CASEMBC(0x179) CASEMBC(0x17b) CASEMBC(0x17d) | |
962 CASEMBC(0x1b5) CASEMBC(0x1e90) CASEMBC(0x1e94) | |
963 regmbc('Z'); REGMBC(0x179) REGMBC(0x17b) | |
964 REGMBC(0x17d) REGMBC(0x1b5) REGMBC(0x1e90) | |
965 REGMBC(0x1e94) | |
167 | 966 return; |
236 | 967 case 'a': case '\340': case '\341': case '\342': |
968 case '\343': case '\344': case '\345': | |
2974 | 969 CASEMBC(0x101) CASEMBC(0x103) CASEMBC(0x105) CASEMBC(0x1ce) |
970 CASEMBC(0x1df) CASEMBC(0x1e1) CASEMBC(0x1ea3) | |
236 | 971 regmbc('a'); regmbc('\340'); regmbc('\341'); |
972 regmbc('\342'); regmbc('\343'); regmbc('\344'); | |
973 regmbc('\345'); | |
2974 | 974 REGMBC(0x101) REGMBC(0x103) REGMBC(0x105) |
975 REGMBC(0x1ce) REGMBC(0x1df) REGMBC(0x1e1) | |
976 REGMBC(0x1ea3) | |
977 return; | |
978 case 'b': CASEMBC(0x1e03) CASEMBC(0x1e07) | |
979 regmbc('b'); REGMBC(0x1e03) REGMBC(0x1e07) | |
167 | 980 return; |
236 | 981 case 'c': case '\347': |
2974 | 982 CASEMBC(0x107) CASEMBC(0x109) CASEMBC(0x10b) CASEMBC(0x10d) |
236 | 983 regmbc('c'); regmbc('\347'); |
2974 | 984 REGMBC(0x107) REGMBC(0x109) REGMBC(0x10b) |
985 REGMBC(0x10d) | |
986 return; | |
987 case 'd': CASEMBC(0x10f) CASEMBC(0x111) CASEMBC(0x1d0b) | |
988 CASEMBC(0x1e11) | |
989 regmbc('d'); REGMBC(0x10f) REGMBC(0x111) | |
990 REGMBC(0x1e0b) REGMBC(0x01e0f) REGMBC(0x1e11) | |
167 | 991 return; |
236 | 992 case 'e': case '\350': case '\351': case '\352': case '\353': |
2974 | 993 CASEMBC(0x113) CASEMBC(0x115) CASEMBC(0x117) CASEMBC(0x119) |
994 CASEMBC(0x11b) CASEMBC(0x1ebb) CASEMBC(0x1ebd) | |
236 | 995 regmbc('e'); regmbc('\350'); regmbc('\351'); |
996 regmbc('\352'); regmbc('\353'); | |
2974 | 997 REGMBC(0x113) REGMBC(0x115) REGMBC(0x117) |
998 REGMBC(0x119) REGMBC(0x11b) REGMBC(0x1ebb) | |
999 REGMBC(0x1ebd) | |
1000 return; | |
1001 case 'f': CASEMBC(0x1e1f) | |
1002 regmbc('f'); REGMBC(0x1e1f) | |
1003 return; | |
1004 case 'g': CASEMBC(0x11d) CASEMBC(0x11f) CASEMBC(0x121) | |
1005 CASEMBC(0x123) CASEMBC(0x1e5) CASEMBC(0x1e7) CASEMBC(0x1f5) | |
1006 CASEMBC(0x1e21) | |
1007 regmbc('g'); REGMBC(0x11d) REGMBC(0x11f) | |
1008 REGMBC(0x121) REGMBC(0x123) REGMBC(0x1e5) | |
1009 REGMBC(0x1e7) REGMBC(0x1f5) REGMBC(0x1e21) | |
1010 return; | |
1011 case 'h': CASEMBC(0x125) CASEMBC(0x127) CASEMBC(0x1e23) | |
1012 CASEMBC(0x1e27) CASEMBC(0x1e29) CASEMBC(0x1e96) | |
1013 regmbc('h'); REGMBC(0x125) REGMBC(0x127) | |
1014 REGMBC(0x1e23) REGMBC(0x1e27) REGMBC(0x1e29) | |
1015 REGMBC(0x1e96) | |
167 | 1016 return; |
236 | 1017 case 'i': case '\354': case '\355': case '\356': case '\357': |
2974 | 1018 CASEMBC(0x129) CASEMBC(0x12b) CASEMBC(0x12d) CASEMBC(0x12f) |
1019 CASEMBC(0x1d0) CASEMBC(0x1ec9) | |
236 | 1020 regmbc('i'); regmbc('\354'); regmbc('\355'); |
1021 regmbc('\356'); regmbc('\357'); | |
2974 | 1022 REGMBC(0x129) REGMBC(0x12b) REGMBC(0x12d) |
1023 REGMBC(0x12f) REGMBC(0x1d0) REGMBC(0x1ec9) | |
1024 return; | |
1025 case 'j': CASEMBC(0x135) CASEMBC(0x1f0) | |
1026 regmbc('j'); REGMBC(0x135) REGMBC(0x1f0) | |
1027 return; | |
1028 case 'k': CASEMBC(0x137) CASEMBC(0x1e9) CASEMBC(0x1e31) | |
1029 CASEMBC(0x1e35) | |
1030 regmbc('k'); REGMBC(0x137) REGMBC(0x1e9) | |
1031 REGMBC(0x1e31) REGMBC(0x1e35) | |
1032 return; | |
1033 case 'l': CASEMBC(0x13a) CASEMBC(0x13c) CASEMBC(0x13e) | |
1034 CASEMBC(0x140) CASEMBC(0x142) CASEMBC(0x1e3b) | |
1035 regmbc('l'); REGMBC(0x13a) REGMBC(0x13c) | |
1036 REGMBC(0x13e) REGMBC(0x140) REGMBC(0x142) | |
1037 REGMBC(0x1e3b) | |
1038 return; | |
1039 case 'm': CASEMBC(0x1e3f) CASEMBC(0x1e41) | |
1040 regmbc('m'); REGMBC(0x1e3f) REGMBC(0x1e41) | |
167 | 1041 return; |
236 | 1042 case 'n': case '\361': |
2974 | 1043 CASEMBC(0x144) CASEMBC(0x146) CASEMBC(0x148) CASEMBC(0x149) |
1044 CASEMBC(0x1e45) CASEMBC(0x1e49) | |
236 | 1045 regmbc('n'); regmbc('\361'); |
2974 | 1046 REGMBC(0x144) REGMBC(0x146) REGMBC(0x148) |
1047 REGMBC(0x149) REGMBC(0x1e45) REGMBC(0x1e49) | |
167 | 1048 return; |
236 | 1049 case 'o': case '\362': case '\363': case '\364': case '\365': |
2974 | 1050 case '\366': case '\370': |
1051 CASEMBC(0x14d) CASEMBC(0x14f) CASEMBC(0x151) CASEMBC(0x1a1) | |
1052 CASEMBC(0x1d2) CASEMBC(0x1eb) CASEMBC(0x1ed) CASEMBC(0x1ecf) | |
236 | 1053 regmbc('o'); regmbc('\362'); regmbc('\363'); |
1054 regmbc('\364'); regmbc('\365'); regmbc('\366'); | |
2974 | 1055 regmbc('\370'); |
1056 REGMBC(0x14d) REGMBC(0x14f) REGMBC(0x151) | |
1057 REGMBC(0x1a1) REGMBC(0x1d2) REGMBC(0x1eb) | |
1058 REGMBC(0x1ed) REGMBC(0x1ecf) | |
1059 return; | |
1060 case 'p': CASEMBC(0x1e55) CASEMBC(0x1e57) | |
1061 regmbc('p'); REGMBC(0x1e55) REGMBC(0x1e57) | |
1062 return; | |
1063 case 'r': CASEMBC(0x155) CASEMBC(0x157) CASEMBC(0x159) | |
1064 CASEMBC(0x1e59) CASEMBC(0x1e5f) | |
1065 regmbc('r'); REGMBC(0x155) REGMBC(0x157) REGMBC(0x159) | |
1066 REGMBC(0x1e59) REGMBC(0x1e5f) | |
1067 return; | |
1068 case 's': CASEMBC(0x15b) CASEMBC(0x15d) CASEMBC(0x15f) | |
1069 CASEMBC(0x161) CASEMBC(0x1e61) | |
1070 regmbc('s'); REGMBC(0x15b) REGMBC(0x15d) | |
1071 REGMBC(0x15f) REGMBC(0x161) REGMBC(0x1e61) | |
1072 return; | |
1073 case 't': CASEMBC(0x163) CASEMBC(0x165) CASEMBC(0x167) | |
1074 CASEMBC(0x1e6b) CASEMBC(0x1e6f) CASEMBC(0x1e97) | |
1075 regmbc('t'); REGMBC(0x163) REGMBC(0x165) REGMBC(0x167) | |
1076 REGMBC(0x1e6b) REGMBC(0x1e6f) REGMBC(0x1e97) | |
167 | 1077 return; |
236 | 1078 case 'u': case '\371': case '\372': case '\373': case '\374': |
2974 | 1079 CASEMBC(0x169) CASEMBC(0x16b) CASEMBC(0x16d) CASEMBC(0x16f) |
1080 CASEMBC(0x171) CASEMBC(0x173) CASEMBC(0x1b0) CASEMBC(0x1d4) | |
1081 CASEMBC(0x1ee7) | |
236 | 1082 regmbc('u'); regmbc('\371'); regmbc('\372'); |
1083 regmbc('\373'); regmbc('\374'); | |
2974 | 1084 REGMBC(0x169) REGMBC(0x16b) REGMBC(0x16d) |
1085 REGMBC(0x16f) REGMBC(0x171) REGMBC(0x173) | |
1086 REGMBC(0x1b0) REGMBC(0x1d4) REGMBC(0x1ee7) | |
1087 return; | |
1088 case 'v': CASEMBC(0x1e7d) | |
1089 regmbc('v'); REGMBC(0x1e7d) | |
1090 return; | |
1091 case 'w': CASEMBC(0x175) CASEMBC(0x1e81) CASEMBC(0x1e83) | |
1092 CASEMBC(0x1e85) CASEMBC(0x1e87) CASEMBC(0x1e98) | |
1093 regmbc('w'); REGMBC(0x175) REGMBC(0x1e81) | |
1094 REGMBC(0x1e83) REGMBC(0x1e85) REGMBC(0x1e87) | |
1095 REGMBC(0x1e98) | |
1096 return; | |
1097 case 'x': CASEMBC(0x1e8b) CASEMBC(0x1e8d) | |
1098 regmbc('x'); REGMBC(0x1e8b) REGMBC(0x1e8d) | |
167 | 1099 return; |
236 | 1100 case 'y': case '\375': case '\377': |
2974 | 1101 CASEMBC(0x177) CASEMBC(0x1e8f) CASEMBC(0x1e99) |
1102 CASEMBC(0x1ef3) CASEMBC(0x1ef7) CASEMBC(0x1ef9) | |
236 | 1103 regmbc('y'); regmbc('\375'); regmbc('\377'); |
2974 | 1104 REGMBC(0x177) REGMBC(0x1e8f) REGMBC(0x1e99) |
1105 REGMBC(0x1ef3) REGMBC(0x1ef7) REGMBC(0x1ef9) | |
1106 return; | |
1107 case 'z': CASEMBC(0x17a) CASEMBC(0x17c) CASEMBC(0x17e) | |
1108 CASEMBC(0x1b6) CASEMBC(0x1e91) CASEMBC(0x1e95) | |
1109 regmbc('z'); REGMBC(0x17a) REGMBC(0x17c) | |
1110 REGMBC(0x17e) REGMBC(0x1b6) REGMBC(0x1e91) | |
1111 REGMBC(0x1e95) | |
167 | 1112 return; |
1113 } | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
1114 #endif |
167 | 1115 } |
1116 regmbc(c); | |
1117 } | |
1118 | |
1119 /* | |
1120 * Check for a collating element "[.a.]". "pp" points to the '['. | |
1121 * Returns a character. Zero means that no item was recognized. Otherwise | |
1122 * "pp" is advanced to after the item. | |
1123 * Currently only single characters are recognized! | |
1124 */ | |
1125 static int | |
1126 get_coll_element(pp) | |
1127 char_u **pp; | |
1128 { | |
1129 int c; | |
1130 int l = 1; | |
1131 char_u *p = *pp; | |
1132 | |
1133 if (p[1] == '.') | |
1134 { | |
1135 #ifdef FEAT_MBYTE | |
1136 if (has_mbyte) | |
474 | 1137 l = (*mb_ptr2len)(p + 2); |
167 | 1138 #endif |
1139 if (p[l + 2] == '.' && p[l + 3] == ']') | |
1140 { | |
1141 #ifdef FEAT_MBYTE | |
1142 if (has_mbyte) | |
1143 c = mb_ptr2char(p + 2); | |
1144 else | |
1145 #endif | |
1146 c = p[2]; | |
1147 *pp += l + 4; | |
1148 return c; | |
1149 } | |
1150 } | |
1151 return 0; | |
1152 } | |
1153 | |
1154 | |
1155 /* | |
1156 * Skip over a "[]" range. | |
1157 * "p" must point to the character after the '['. | |
1158 * The returned pointer is on the matching ']', or the terminating NUL. | |
1159 */ | |
1160 static char_u * | |
1161 skip_anyof(p) | |
1162 char_u *p; | |
1163 { | |
1164 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
1165 int cpo_bsl; /* 'cpoptions' contains '\' flag */ | |
1166 #ifdef FEAT_MBYTE | |
1167 int l; | |
1168 #endif | |
1169 | |
410 | 1170 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; |
1171 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
167 | 1172 |
1173 if (*p == '^') /* Complement of range. */ | |
1174 ++p; | |
1175 if (*p == ']' || *p == '-') | |
1176 ++p; | |
1177 while (*p != NUL && *p != ']') | |
1178 { | |
1179 #ifdef FEAT_MBYTE | |
474 | 1180 if (has_mbyte && (l = (*mb_ptr2len)(p)) > 1) |
167 | 1181 p += l; |
1182 else | |
1183 #endif | |
1184 if (*p == '-') | |
1185 { | |
1186 ++p; | |
1187 if (*p != ']' && *p != NUL) | |
1188 mb_ptr_adv(p); | |
1189 } | |
1190 else if (*p == '\\' | |
1191 && !cpo_bsl | |
1192 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL | |
1193 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL))) | |
1194 p += 2; | |
1195 else if (*p == '[') | |
1196 { | |
1197 if (get_char_class(&p) == CLASS_NONE | |
1198 && get_equi_class(&p) == 0 | |
1199 && get_coll_element(&p) == 0) | |
1200 ++p; /* It was not a class name */ | |
1201 } | |
1202 else | |
1203 ++p; | |
1204 } | |
1205 | |
1206 return p; | |
1207 } | |
1208 | |
1209 /* | |
7 | 1210 * Skip past regular expression. |
153 | 1211 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc). |
7 | 1212 * Take care of characters with a backslash in front of it. |
1213 * Skip strings inside [ and ]. | |
1214 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the | |
1215 * expression and change "\?" to "?". If "*newp" is not NULL the expression | |
1216 * is changed in-place. | |
1217 */ | |
1218 char_u * | |
1219 skip_regexp(startp, dirc, magic, newp) | |
1220 char_u *startp; | |
1221 int dirc; | |
1222 int magic; | |
1223 char_u **newp; | |
1224 { | |
1225 int mymagic; | |
1226 char_u *p = startp; | |
1227 | |
1228 if (magic) | |
1229 mymagic = MAGIC_ON; | |
1230 else | |
1231 mymagic = MAGIC_OFF; | |
1232 | |
39 | 1233 for (; p[0] != NUL; mb_ptr_adv(p)) |
7 | 1234 { |
1235 if (p[0] == dirc) /* found end of regexp */ | |
1236 break; | |
1237 if ((p[0] == '[' && mymagic >= MAGIC_ON) | |
1238 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF)) | |
1239 { | |
1240 p = skip_anyof(p + 1); | |
1241 if (p[0] == NUL) | |
1242 break; | |
1243 } | |
1244 else if (p[0] == '\\' && p[1] != NUL) | |
1245 { | |
1246 if (dirc == '?' && newp != NULL && p[1] == '?') | |
1247 { | |
1248 /* change "\?" to "?", make a copy first. */ | |
1249 if (*newp == NULL) | |
1250 { | |
1251 *newp = vim_strsave(startp); | |
1252 if (*newp != NULL) | |
1253 p = *newp + (p - startp); | |
1254 } | |
1255 if (*newp != NULL) | |
1621 | 1256 STRMOVE(p, p + 1); |
7 | 1257 else |
1258 ++p; | |
1259 } | |
1260 else | |
1261 ++p; /* skip next character */ | |
1262 if (*p == 'v') | |
1263 mymagic = MAGIC_ALL; | |
1264 else if (*p == 'V') | |
1265 mymagic = MAGIC_NONE; | |
1266 } | |
1267 } | |
1268 return p; | |
1269 } | |
1270 | |
4444 | 1271 static regprog_T *bt_regcomp __ARGS((char_u *expr, int re_flags)); |
1272 | |
7 | 1273 /* |
4444 | 1274 * bt_regcomp() - compile a regular expression into internal code for the |
1275 * traditional back track matcher. | |
41 | 1276 * Returns the program in allocated space. Returns NULL for an error. |
7 | 1277 * |
1278 * We can't allocate space until we know how big the compiled form will be, | |
1279 * but we can't compile it (and thus know how big it is) until we've got a | |
1280 * place to put the code. So we cheat: we compile it twice, once with code | |
1281 * generation turned off and size counting turned on, and once "for real". | |
1282 * This also means that we don't allocate space until we are sure that the | |
1283 * thing really will compile successfully, and we never have to move the | |
1284 * code and thus invalidate pointers into it. (Note that it has to be in | |
1285 * one piece because vim_free() must be able to free it all.) | |
1286 * | |
1287 * Whether upper/lower case is to be ignored is decided when executing the | |
1288 * program, it does not matter here. | |
1289 * | |
1290 * Beware that the optimization-preparation code in here knows about some | |
1291 * of the structure of the compiled regexp. | |
1292 * "re_flags": RE_MAGIC and/or RE_STRING. | |
1293 */ | |
4444 | 1294 static regprog_T * |
1295 bt_regcomp(expr, re_flags) | |
7 | 1296 char_u *expr; |
1297 int re_flags; | |
1298 { | |
4444 | 1299 bt_regprog_T *r; |
7 | 1300 char_u *scan; |
1301 char_u *longest; | |
1302 int len; | |
1303 int flags; | |
1304 | |
1305 if (expr == NULL) | |
1306 EMSG_RET_NULL(_(e_null)); | |
1307 | |
1308 init_class_tab(); | |
1309 | |
1310 /* | |
1311 * First pass: determine size, legality. | |
1312 */ | |
1313 regcomp_start(expr, re_flags); | |
1314 regcode = JUST_CALC_SIZE; | |
1315 regc(REGMAGIC); | |
1316 if (reg(REG_NOPAREN, &flags) == NULL) | |
1317 return NULL; | |
1318 | |
1319 /* Small enough for pointer-storage convention? */ | |
1320 #ifdef SMALL_MALLOC /* 16 bit storage allocation */ | |
1321 if (regsize >= 65536L - 256L) | |
1322 EMSG_RET_NULL(_("E339: Pattern too long")); | |
1323 #endif | |
1324 | |
1325 /* Allocate space. */ | |
4444 | 1326 r = (bt_regprog_T *)lalloc(sizeof(bt_regprog_T) + regsize, TRUE); |
7 | 1327 if (r == NULL) |
1328 return NULL; | |
1329 | |
1330 /* | |
1331 * Second pass: emit code. | |
1332 */ | |
1333 regcomp_start(expr, re_flags); | |
1334 regcode = r->program; | |
1335 regc(REGMAGIC); | |
2010 | 1336 if (reg(REG_NOPAREN, &flags) == NULL || reg_toolong) |
7 | 1337 { |
1338 vim_free(r); | |
2010 | 1339 if (reg_toolong) |
1340 EMSG_RET_NULL(_("E339: Pattern too long")); | |
7 | 1341 return NULL; |
1342 } | |
1343 | |
1344 /* Dig out information for optimizations. */ | |
1345 r->regstart = NUL; /* Worst-case defaults. */ | |
1346 r->reganch = 0; | |
1347 r->regmust = NULL; | |
1348 r->regmlen = 0; | |
1349 r->regflags = regflags; | |
1350 if (flags & HASNL) | |
1351 r->regflags |= RF_HASNL; | |
1352 if (flags & HASLOOKBH) | |
1353 r->regflags |= RF_LOOKBH; | |
1354 #ifdef FEAT_SYN_HL | |
1355 /* Remember whether this pattern has any \z specials in it. */ | |
1356 r->reghasz = re_has_z; | |
1357 #endif | |
1358 scan = r->program + 1; /* First BRANCH. */ | |
1359 if (OP(regnext(scan)) == END) /* Only one top-level choice. */ | |
1360 { | |
1361 scan = OPERAND(scan); | |
1362 | |
1363 /* Starting-point info. */ | |
1364 if (OP(scan) == BOL || OP(scan) == RE_BOF) | |
1365 { | |
1366 r->reganch++; | |
1367 scan = regnext(scan); | |
1368 } | |
1369 | |
1370 if (OP(scan) == EXACTLY) | |
1371 { | |
1372 #ifdef FEAT_MBYTE | |
1373 if (has_mbyte) | |
1374 r->regstart = (*mb_ptr2char)(OPERAND(scan)); | |
1375 else | |
1376 #endif | |
1377 r->regstart = *OPERAND(scan); | |
1378 } | |
1379 else if ((OP(scan) == BOW | |
1380 || OP(scan) == EOW | |
1381 || OP(scan) == NOTHING | |
1382 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN | |
1383 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE) | |
1384 && OP(regnext(scan)) == EXACTLY) | |
1385 { | |
1386 #ifdef FEAT_MBYTE | |
1387 if (has_mbyte) | |
1388 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan))); | |
1389 else | |
1390 #endif | |
1391 r->regstart = *OPERAND(regnext(scan)); | |
1392 } | |
1393 | |
1394 /* | |
1395 * If there's something expensive in the r.e., find the longest | |
1396 * literal string that must appear and make it the regmust. Resolve | |
1397 * ties in favor of later strings, since the regstart check works | |
1398 * with the beginning of the r.e. and avoiding duplication | |
1399 * strengthens checking. Not a strong reason, but sufficient in the | |
1400 * absence of others. | |
1401 */ | |
1402 /* | |
1403 * When the r.e. starts with BOW, it is faster to look for a regmust | |
1404 * first. Used a lot for "#" and "*" commands. (Added by mool). | |
1405 */ | |
1406 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW) | |
1407 && !(flags & HASNL)) | |
1408 { | |
1409 longest = NULL; | |
1410 len = 0; | |
1411 for (; scan != NULL; scan = regnext(scan)) | |
1412 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len) | |
1413 { | |
1414 longest = OPERAND(scan); | |
1415 len = (int)STRLEN(OPERAND(scan)); | |
1416 } | |
1417 r->regmust = longest; | |
1418 r->regmlen = len; | |
1419 } | |
1420 } | |
4444 | 1421 #ifdef BT_REGEXP_DUMP |
7 | 1422 regdump(expr, r); |
1423 #endif | |
4444 | 1424 r->engine = &bt_regengine; |
1425 return (regprog_T *)r; | |
7 | 1426 } |
1427 | |
1428 /* | |
1429 * Setup to parse the regexp. Used once to get the length and once to do it. | |
1430 */ | |
1431 static void | |
1432 regcomp_start(expr, re_flags) | |
1433 char_u *expr; | |
1434 int re_flags; /* see vim_regcomp() */ | |
1435 { | |
1436 initchr(expr); | |
1437 if (re_flags & RE_MAGIC) | |
1438 reg_magic = MAGIC_ON; | |
1439 else | |
1440 reg_magic = MAGIC_OFF; | |
1441 reg_string = (re_flags & RE_STRING); | |
481 | 1442 reg_strict = (re_flags & RE_STRICT); |
7 | 1443 |
1444 num_complex_braces = 0; | |
1445 regnpar = 1; | |
1446 vim_memset(had_endbrace, 0, sizeof(had_endbrace)); | |
1447 #ifdef FEAT_SYN_HL | |
1448 regnzpar = 1; | |
1449 re_has_z = 0; | |
1450 #endif | |
1451 regsize = 0L; | |
2010 | 1452 reg_toolong = FALSE; |
7 | 1453 regflags = 0; |
1454 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1455 had_eol = FALSE; | |
1456 #endif | |
1457 } | |
1458 | |
1459 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1460 /* | |
1461 * Check if during the previous call to vim_regcomp the EOL item "$" has been | |
1462 * found. This is messy, but it works fine. | |
1463 */ | |
1464 int | |
1465 vim_regcomp_had_eol() | |
1466 { | |
1467 return had_eol; | |
1468 } | |
1469 #endif | |
1470 | |
1471 /* | |
4444 | 1472 * Parse regular expression, i.e. main body or parenthesized thing. |
7 | 1473 * |
1474 * Caller must absorb opening parenthesis. | |
1475 * | |
1476 * Combining parenthesis handling with the base level of regular expression | |
1477 * is a trifle forced, but the need to tie the tails of the branches to what | |
1478 * follows makes it hard to avoid. | |
1479 */ | |
1480 static char_u * | |
1481 reg(paren, flagp) | |
1482 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */ | |
1483 int *flagp; | |
1484 { | |
1485 char_u *ret; | |
1486 char_u *br; | |
1487 char_u *ender; | |
1488 int parno = 0; | |
1489 int flags; | |
1490 | |
1491 *flagp = HASWIDTH; /* Tentatively. */ | |
1492 | |
1493 #ifdef FEAT_SYN_HL | |
1494 if (paren == REG_ZPAREN) | |
1495 { | |
1496 /* Make a ZOPEN node. */ | |
1497 if (regnzpar >= NSUBEXP) | |
1498 EMSG_RET_NULL(_("E50: Too many \\z(")); | |
1499 parno = regnzpar; | |
1500 regnzpar++; | |
1501 ret = regnode(ZOPEN + parno); | |
1502 } | |
1503 else | |
1504 #endif | |
1505 if (paren == REG_PAREN) | |
1506 { | |
1507 /* Make a MOPEN node. */ | |
1508 if (regnpar >= NSUBEXP) | |
4444 | 1509 EMSG2_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL); |
7 | 1510 parno = regnpar; |
1511 ++regnpar; | |
1512 ret = regnode(MOPEN + parno); | |
1513 } | |
1514 else if (paren == REG_NPAREN) | |
1515 { | |
1516 /* Make a NOPEN node. */ | |
1517 ret = regnode(NOPEN); | |
1518 } | |
1519 else | |
1520 ret = NULL; | |
1521 | |
1522 /* Pick up the branches, linking them together. */ | |
1523 br = regbranch(&flags); | |
1524 if (br == NULL) | |
1525 return NULL; | |
1526 if (ret != NULL) | |
1527 regtail(ret, br); /* [MZ]OPEN -> first. */ | |
1528 else | |
1529 ret = br; | |
1530 /* If one of the branches can be zero-width, the whole thing can. | |
1531 * If one of the branches has * at start or matches a line-break, the | |
1532 * whole thing can. */ | |
1533 if (!(flags & HASWIDTH)) | |
1534 *flagp &= ~HASWIDTH; | |
1535 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH); | |
1536 while (peekchr() == Magic('|')) | |
1537 { | |
1538 skipchr(); | |
1539 br = regbranch(&flags); | |
2010 | 1540 if (br == NULL || reg_toolong) |
7 | 1541 return NULL; |
1542 regtail(ret, br); /* BRANCH -> BRANCH. */ | |
1543 if (!(flags & HASWIDTH)) | |
1544 *flagp &= ~HASWIDTH; | |
1545 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH); | |
1546 } | |
1547 | |
1548 /* Make a closing node, and hook it on the end. */ | |
1549 ender = regnode( | |
1550 #ifdef FEAT_SYN_HL | |
1551 paren == REG_ZPAREN ? ZCLOSE + parno : | |
1552 #endif | |
1553 paren == REG_PAREN ? MCLOSE + parno : | |
1554 paren == REG_NPAREN ? NCLOSE : END); | |
1555 regtail(ret, ender); | |
1556 | |
1557 /* Hook the tails of the branches to the closing node. */ | |
1558 for (br = ret; br != NULL; br = regnext(br)) | |
1559 regoptail(br, ender); | |
1560 | |
1561 /* Check for proper termination. */ | |
1562 if (paren != REG_NOPAREN && getchr() != Magic(')')) | |
1563 { | |
1564 #ifdef FEAT_SYN_HL | |
1565 if (paren == REG_ZPAREN) | |
308 | 1566 EMSG_RET_NULL(_("E52: Unmatched \\z(")); |
7 | 1567 else |
1568 #endif | |
1569 if (paren == REG_NPAREN) | |
4444 | 1570 EMSG2_RET_NULL(_(e_unmatchedpp), reg_magic == MAGIC_ALL); |
7 | 1571 else |
4444 | 1572 EMSG2_RET_NULL(_(e_unmatchedp), reg_magic == MAGIC_ALL); |
7 | 1573 } |
1574 else if (paren == REG_NOPAREN && peekchr() != NUL) | |
1575 { | |
1576 if (curchr == Magic(')')) | |
4444 | 1577 EMSG2_RET_NULL(_(e_unmatchedpar), reg_magic == MAGIC_ALL); |
7 | 1578 else |
308 | 1579 EMSG_RET_NULL(_(e_trailing)); /* "Can't happen". */ |
7 | 1580 /* NOTREACHED */ |
1581 } | |
1582 /* | |
1583 * Here we set the flag allowing back references to this set of | |
1584 * parentheses. | |
1585 */ | |
1586 if (paren == REG_PAREN) | |
1587 had_endbrace[parno] = TRUE; /* have seen the close paren */ | |
1588 return ret; | |
1589 } | |
1590 | |
1591 /* | |
4444 | 1592 * Parse one alternative of an | operator. |
7 | 1593 * Implements the & operator. |
1594 */ | |
1595 static char_u * | |
1596 regbranch(flagp) | |
1597 int *flagp; | |
1598 { | |
1599 char_u *ret; | |
1600 char_u *chain = NULL; | |
1601 char_u *latest; | |
1602 int flags; | |
1603 | |
1604 *flagp = WORST | HASNL; /* Tentatively. */ | |
1605 | |
1606 ret = regnode(BRANCH); | |
1607 for (;;) | |
1608 { | |
1609 latest = regconcat(&flags); | |
1610 if (latest == NULL) | |
1611 return NULL; | |
1612 /* If one of the branches has width, the whole thing has. If one of | |
1613 * the branches anchors at start-of-line, the whole thing does. | |
1614 * If one of the branches uses look-behind, the whole thing does. */ | |
1615 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH); | |
1616 /* If one of the branches doesn't match a line-break, the whole thing | |
1617 * doesn't. */ | |
1618 *flagp &= ~HASNL | (flags & HASNL); | |
1619 if (chain != NULL) | |
1620 regtail(chain, latest); | |
1621 if (peekchr() != Magic('&')) | |
1622 break; | |
1623 skipchr(); | |
1624 regtail(latest, regnode(END)); /* operand ends */ | |
2010 | 1625 if (reg_toolong) |
1626 break; | |
7 | 1627 reginsert(MATCH, latest); |
1628 chain = latest; | |
1629 } | |
1630 | |
1631 return ret; | |
1632 } | |
1633 | |
1634 /* | |
4444 | 1635 * Parse one alternative of an | or & operator. |
7 | 1636 * Implements the concatenation operator. |
1637 */ | |
1638 static char_u * | |
1639 regconcat(flagp) | |
1640 int *flagp; | |
1641 { | |
1642 char_u *first = NULL; | |
1643 char_u *chain = NULL; | |
1644 char_u *latest; | |
1645 int flags; | |
1646 int cont = TRUE; | |
1647 | |
1648 *flagp = WORST; /* Tentatively. */ | |
1649 | |
1650 while (cont) | |
1651 { | |
1652 switch (peekchr()) | |
1653 { | |
1654 case NUL: | |
1655 case Magic('|'): | |
1656 case Magic('&'): | |
1657 case Magic(')'): | |
1658 cont = FALSE; | |
1659 break; | |
1660 case Magic('Z'): | |
1661 #ifdef FEAT_MBYTE | |
1662 regflags |= RF_ICOMBINE; | |
1663 #endif | |
1664 skipchr_keepstart(); | |
1665 break; | |
1666 case Magic('c'): | |
1667 regflags |= RF_ICASE; | |
1668 skipchr_keepstart(); | |
1669 break; | |
1670 case Magic('C'): | |
1671 regflags |= RF_NOICASE; | |
1672 skipchr_keepstart(); | |
1673 break; | |
1674 case Magic('v'): | |
1675 reg_magic = MAGIC_ALL; | |
1676 skipchr_keepstart(); | |
1677 curchr = -1; | |
1678 break; | |
1679 case Magic('m'): | |
1680 reg_magic = MAGIC_ON; | |
1681 skipchr_keepstart(); | |
1682 curchr = -1; | |
1683 break; | |
1684 case Magic('M'): | |
1685 reg_magic = MAGIC_OFF; | |
1686 skipchr_keepstart(); | |
1687 curchr = -1; | |
1688 break; | |
1689 case Magic('V'): | |
1690 reg_magic = MAGIC_NONE; | |
1691 skipchr_keepstart(); | |
1692 curchr = -1; | |
1693 break; | |
1694 default: | |
1695 latest = regpiece(&flags); | |
2010 | 1696 if (latest == NULL || reg_toolong) |
7 | 1697 return NULL; |
1698 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH); | |
1699 if (chain == NULL) /* First piece. */ | |
1700 *flagp |= flags & SPSTART; | |
1701 else | |
1702 regtail(chain, latest); | |
1703 chain = latest; | |
1704 if (first == NULL) | |
1705 first = latest; | |
1706 break; | |
1707 } | |
1708 } | |
1709 if (first == NULL) /* Loop ran zero times. */ | |
1710 first = regnode(NOTHING); | |
1711 return first; | |
1712 } | |
1713 | |
1714 /* | |
4444 | 1715 * Parse something followed by possible [*+=]. |
7 | 1716 * |
1717 * Note that the branching code sequences used for = and the general cases | |
1718 * of * and + are somewhat optimized: they use the same NOTHING node as | |
1719 * both the endmarker for their branch list and the body of the last branch. | |
1720 * It might seem that this node could be dispensed with entirely, but the | |
1721 * endmarker role is not redundant. | |
1722 */ | |
1723 static char_u * | |
1724 regpiece(flagp) | |
1725 int *flagp; | |
1726 { | |
1727 char_u *ret; | |
1728 int op; | |
1729 char_u *next; | |
1730 int flags; | |
1731 long minval; | |
1732 long maxval; | |
1733 | |
1734 ret = regatom(&flags); | |
1735 if (ret == NULL) | |
1736 return NULL; | |
1737 | |
1738 op = peekchr(); | |
1739 if (re_multi_type(op) == NOT_MULTI) | |
1740 { | |
1741 *flagp = flags; | |
1742 return ret; | |
1743 } | |
1744 /* default flags */ | |
1745 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH))); | |
1746 | |
1747 skipchr(); | |
1748 switch (op) | |
1749 { | |
1750 case Magic('*'): | |
1751 if (flags & SIMPLE) | |
1752 reginsert(STAR, ret); | |
1753 else | |
1754 { | |
1755 /* Emit x* as (x&|), where & means "self". */ | |
1756 reginsert(BRANCH, ret); /* Either x */ | |
1757 regoptail(ret, regnode(BACK)); /* and loop */ | |
1758 regoptail(ret, ret); /* back */ | |
1759 regtail(ret, regnode(BRANCH)); /* or */ | |
1760 regtail(ret, regnode(NOTHING)); /* null. */ | |
1761 } | |
1762 break; | |
1763 | |
1764 case Magic('+'): | |
1765 if (flags & SIMPLE) | |
1766 reginsert(PLUS, ret); | |
1767 else | |
1768 { | |
1769 /* Emit x+ as x(&|), where & means "self". */ | |
1770 next = regnode(BRANCH); /* Either */ | |
1771 regtail(ret, next); | |
233 | 1772 regtail(regnode(BACK), ret); /* loop back */ |
7 | 1773 regtail(next, regnode(BRANCH)); /* or */ |
1774 regtail(ret, regnode(NOTHING)); /* null. */ | |
1775 } | |
1776 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH))); | |
1777 break; | |
1778 | |
1779 case Magic('@'): | |
1780 { | |
1781 int lop = END; | |
1782 | |
1783 switch (no_Magic(getchr())) | |
1784 { | |
1785 case '=': lop = MATCH; break; /* \@= */ | |
1786 case '!': lop = NOMATCH; break; /* \@! */ | |
1787 case '>': lop = SUBPAT; break; /* \@> */ | |
1788 case '<': switch (no_Magic(getchr())) | |
1789 { | |
1790 case '=': lop = BEHIND; break; /* \@<= */ | |
1791 case '!': lop = NOBEHIND; break; /* \@<! */ | |
1792 } | |
1793 } | |
1794 if (lop == END) | |
4444 | 1795 EMSG2_RET_NULL(_("E59: invalid character after %s@"), |
7 | 1796 reg_magic == MAGIC_ALL); |
1797 /* Look behind must match with behind_pos. */ | |
1798 if (lop == BEHIND || lop == NOBEHIND) | |
1799 { | |
1800 regtail(ret, regnode(BHPOS)); | |
1801 *flagp |= HASLOOKBH; | |
1802 } | |
1803 regtail(ret, regnode(END)); /* operand ends */ | |
1804 reginsert(lop, ret); | |
1805 break; | |
1806 } | |
1807 | |
1808 case Magic('?'): | |
1809 case Magic('='): | |
1810 /* Emit x= as (x|) */ | |
1811 reginsert(BRANCH, ret); /* Either x */ | |
1812 regtail(ret, regnode(BRANCH)); /* or */ | |
1813 next = regnode(NOTHING); /* null. */ | |
1814 regtail(ret, next); | |
1815 regoptail(ret, next); | |
1816 break; | |
1817 | |
1818 case Magic('{'): | |
1819 if (!read_limits(&minval, &maxval)) | |
1820 return NULL; | |
1821 if (flags & SIMPLE) | |
1822 { | |
1823 reginsert(BRACE_SIMPLE, ret); | |
1824 reginsert_limits(BRACE_LIMITS, minval, maxval, ret); | |
1825 } | |
1826 else | |
1827 { | |
1828 if (num_complex_braces >= 10) | |
4444 | 1829 EMSG2_RET_NULL(_("E60: Too many complex %s{...}s"), |
7 | 1830 reg_magic == MAGIC_ALL); |
1831 reginsert(BRACE_COMPLEX + num_complex_braces, ret); | |
1832 regoptail(ret, regnode(BACK)); | |
1833 regoptail(ret, ret); | |
1834 reginsert_limits(BRACE_LIMITS, minval, maxval, ret); | |
1835 ++num_complex_braces; | |
1836 } | |
1837 if (minval > 0 && maxval > 0) | |
1838 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH))); | |
1839 break; | |
1840 } | |
1841 if (re_multi_type(peekchr()) != NOT_MULTI) | |
1842 { | |
1843 /* Can't have a multi follow a multi. */ | |
1844 if (peekchr() == Magic('*')) | |
1845 sprintf((char *)IObuff, _("E61: Nested %s*"), | |
1846 reg_magic >= MAGIC_ON ? "" : "\\"); | |
1847 else | |
1848 sprintf((char *)IObuff, _("E62: Nested %s%c"), | |
1849 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr())); | |
1850 EMSG_RET_NULL(IObuff); | |
1851 } | |
1852 | |
1853 return ret; | |
1854 } | |
1855 | |
4444 | 1856 /* When making changes to classchars also change nfa_classcodes. */ |
1857 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU"; | |
1858 static int classcodes[] = { | |
1859 ANY, IDENT, SIDENT, KWORD, SKWORD, | |
1860 FNAME, SFNAME, PRINT, SPRINT, | |
1861 WHITE, NWHITE, DIGIT, NDIGIT, | |
1862 HEX, NHEX, OCTAL, NOCTAL, | |
1863 WORD, NWORD, HEAD, NHEAD, | |
1864 ALPHA, NALPHA, LOWER, NLOWER, | |
1865 UPPER, NUPPER | |
1866 }; | |
1867 | |
7 | 1868 /* |
4444 | 1869 * Parse the lowest level. |
7 | 1870 * |
1871 * Optimization: gobbles an entire sequence of ordinary characters so that | |
1872 * it can turn them into a single node, which is smaller to store and | |
1873 * faster to run. Don't do this when one_exactly is set. | |
1874 */ | |
1875 static char_u * | |
1876 regatom(flagp) | |
1877 int *flagp; | |
1878 { | |
1879 char_u *ret; | |
1880 int flags; | |
1881 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
167 | 1882 int cpo_bsl; /* 'cpoptions' contains '\' flag */ |
7 | 1883 int c; |
1884 char_u *p; | |
1885 int extra = 0; | |
1886 | |
1887 *flagp = WORST; /* Tentatively. */ | |
410 | 1888 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; |
1889 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
7 | 1890 |
1891 c = getchr(); | |
1892 switch (c) | |
1893 { | |
1894 case Magic('^'): | |
1895 ret = regnode(BOL); | |
1896 break; | |
1897 | |
1898 case Magic('$'): | |
1899 ret = regnode(EOL); | |
1900 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1901 had_eol = TRUE; | |
1902 #endif | |
1903 break; | |
1904 | |
1905 case Magic('<'): | |
1906 ret = regnode(BOW); | |
1907 break; | |
1908 | |
1909 case Magic('>'): | |
1910 ret = regnode(EOW); | |
1911 break; | |
1912 | |
1913 case Magic('_'): | |
1914 c = no_Magic(getchr()); | |
1915 if (c == '^') /* "\_^" is start-of-line */ | |
1916 { | |
1917 ret = regnode(BOL); | |
1918 break; | |
1919 } | |
1920 if (c == '$') /* "\_$" is end-of-line */ | |
1921 { | |
1922 ret = regnode(EOL); | |
1923 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1924 had_eol = TRUE; | |
1925 #endif | |
1926 break; | |
1927 } | |
1928 | |
1929 extra = ADD_NL; | |
1930 *flagp |= HASNL; | |
1931 | |
1932 /* "\_[" is character range plus newline */ | |
1933 if (c == '[') | |
1934 goto collection; | |
1935 | |
1936 /* "\_x" is character class plus newline */ | |
1937 /*FALLTHROUGH*/ | |
1938 | |
1939 /* | |
1940 * Character classes. | |
1941 */ | |
1942 case Magic('.'): | |
1943 case Magic('i'): | |
1944 case Magic('I'): | |
1945 case Magic('k'): | |
1946 case Magic('K'): | |
1947 case Magic('f'): | |
1948 case Magic('F'): | |
1949 case Magic('p'): | |
1950 case Magic('P'): | |
1951 case Magic('s'): | |
1952 case Magic('S'): | |
1953 case Magic('d'): | |
1954 case Magic('D'): | |
1955 case Magic('x'): | |
1956 case Magic('X'): | |
1957 case Magic('o'): | |
1958 case Magic('O'): | |
1959 case Magic('w'): | |
1960 case Magic('W'): | |
1961 case Magic('h'): | |
1962 case Magic('H'): | |
1963 case Magic('a'): | |
1964 case Magic('A'): | |
1965 case Magic('l'): | |
1966 case Magic('L'): | |
1967 case Magic('u'): | |
1968 case Magic('U'): | |
1969 p = vim_strchr(classchars, no_Magic(c)); | |
1970 if (p == NULL) | |
1971 EMSG_RET_NULL(_("E63: invalid use of \\_")); | |
714 | 1972 #ifdef FEAT_MBYTE |
1973 /* When '.' is followed by a composing char ignore the dot, so that | |
1974 * the composing char is matched here. */ | |
1975 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr())) | |
1976 { | |
1977 c = getchr(); | |
1978 goto do_multibyte; | |
1979 } | |
1980 #endif | |
7 | 1981 ret = regnode(classcodes[p - classchars] + extra); |
1982 *flagp |= HASWIDTH | SIMPLE; | |
1983 break; | |
1984 | |
1985 case Magic('n'): | |
1986 if (reg_string) | |
1987 { | |
1988 /* In a string "\n" matches a newline character. */ | |
1989 ret = regnode(EXACTLY); | |
1990 regc(NL); | |
1991 regc(NUL); | |
1992 *flagp |= HASWIDTH | SIMPLE; | |
1993 } | |
1994 else | |
1995 { | |
1996 /* In buffer text "\n" matches the end of a line. */ | |
1997 ret = regnode(NEWL); | |
1998 *flagp |= HASWIDTH | HASNL; | |
1999 } | |
2000 break; | |
2001 | |
2002 case Magic('('): | |
2003 if (one_exactly) | |
2004 EMSG_ONE_RET_NULL; | |
2005 ret = reg(REG_PAREN, &flags); | |
2006 if (ret == NULL) | |
2007 return NULL; | |
2008 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH); | |
2009 break; | |
2010 | |
2011 case NUL: | |
2012 case Magic('|'): | |
2013 case Magic('&'): | |
2014 case Magic(')'): | |
1468 | 2015 if (one_exactly) |
2016 EMSG_ONE_RET_NULL; | |
7 | 2017 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */ |
2018 /* NOTREACHED */ | |
2019 | |
2020 case Magic('='): | |
2021 case Magic('?'): | |
2022 case Magic('+'): | |
2023 case Magic('@'): | |
2024 case Magic('{'): | |
2025 case Magic('*'): | |
2026 c = no_Magic(c); | |
2027 sprintf((char *)IObuff, _("E64: %s%c follows nothing"), | |
2028 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL) | |
2029 ? "" : "\\", c); | |
2030 EMSG_RET_NULL(IObuff); | |
2031 /* NOTREACHED */ | |
2032 | |
2033 case Magic('~'): /* previous substitute pattern */ | |
359 | 2034 if (reg_prev_sub != NULL) |
7 | 2035 { |
2036 char_u *lp; | |
2037 | |
2038 ret = regnode(EXACTLY); | |
2039 lp = reg_prev_sub; | |
2040 while (*lp != NUL) | |
2041 regc(*lp++); | |
2042 regc(NUL); | |
2043 if (*reg_prev_sub != NUL) | |
2044 { | |
2045 *flagp |= HASWIDTH; | |
2046 if ((lp - reg_prev_sub) == 1) | |
2047 *flagp |= SIMPLE; | |
2048 } | |
2049 } | |
2050 else | |
2051 EMSG_RET_NULL(_(e_nopresub)); | |
2052 break; | |
2053 | |
2054 case Magic('1'): | |
2055 case Magic('2'): | |
2056 case Magic('3'): | |
2057 case Magic('4'): | |
2058 case Magic('5'): | |
2059 case Magic('6'): | |
2060 case Magic('7'): | |
2061 case Magic('8'): | |
2062 case Magic('9'): | |
2063 { | |
2064 int refnum; | |
2065 | |
2066 refnum = c - Magic('0'); | |
2067 /* | |
2068 * Check if the back reference is legal. We must have seen the | |
2069 * close brace. | |
2070 * TODO: Should also check that we don't refer to something | |
2071 * that is repeated (+*=): what instance of the repetition | |
2072 * should we match? | |
2073 */ | |
2074 if (!had_endbrace[refnum]) | |
2075 { | |
2076 /* Trick: check if "@<=" or "@<!" follows, in which case | |
2077 * the \1 can appear before the referenced match. */ | |
2078 for (p = regparse; *p != NUL; ++p) | |
2079 if (p[0] == '@' && p[1] == '<' | |
2080 && (p[2] == '!' || p[2] == '=')) | |
2081 break; | |
2082 if (*p == NUL) | |
2083 EMSG_RET_NULL(_("E65: Illegal back reference")); | |
2084 } | |
2085 ret = regnode(BACKREF + refnum); | |
2086 } | |
2087 break; | |
2088 | |
2089 case Magic('z'): | |
2090 { | |
2091 c = no_Magic(getchr()); | |
2092 switch (c) | |
2093 { | |
741 | 2094 #ifdef FEAT_SYN_HL |
7 | 2095 case '(': if (reg_do_extmatch != REX_SET) |
2096 EMSG_RET_NULL(_("E66: \\z( not allowed here")); | |
2097 if (one_exactly) | |
2098 EMSG_ONE_RET_NULL; | |
2099 ret = reg(REG_ZPAREN, &flags); | |
2100 if (ret == NULL) | |
2101 return NULL; | |
2102 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH); | |
2103 re_has_z = REX_SET; | |
2104 break; | |
2105 | |
2106 case '1': | |
2107 case '2': | |
2108 case '3': | |
2109 case '4': | |
2110 case '5': | |
2111 case '6': | |
2112 case '7': | |
2113 case '8': | |
2114 case '9': if (reg_do_extmatch != REX_USE) | |
2115 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here")); | |
2116 ret = regnode(ZREF + c - '0'); | |
2117 re_has_z = REX_USE; | |
2118 break; | |
741 | 2119 #endif |
7 | 2120 |
2121 case 's': ret = regnode(MOPEN + 0); | |
2122 break; | |
2123 | |
2124 case 'e': ret = regnode(MCLOSE + 0); | |
2125 break; | |
2126 | |
2127 default: EMSG_RET_NULL(_("E68: Invalid character after \\z")); | |
2128 } | |
2129 } | |
2130 break; | |
2131 | |
2132 case Magic('%'): | |
2133 { | |
2134 c = no_Magic(getchr()); | |
2135 switch (c) | |
2136 { | |
2137 /* () without a back reference */ | |
2138 case '(': | |
2139 if (one_exactly) | |
2140 EMSG_ONE_RET_NULL; | |
2141 ret = reg(REG_NPAREN, &flags); | |
2142 if (ret == NULL) | |
2143 return NULL; | |
2144 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH); | |
2145 break; | |
2146 | |
2147 /* Catch \%^ and \%$ regardless of where they appear in the | |
2148 * pattern -- regardless of whether or not it makes sense. */ | |
2149 case '^': | |
2150 ret = regnode(RE_BOF); | |
2151 break; | |
2152 | |
2153 case '$': | |
2154 ret = regnode(RE_EOF); | |
2155 break; | |
2156 | |
2157 case '#': | |
2158 ret = regnode(CURSOR); | |
2159 break; | |
2160 | |
639 | 2161 case 'V': |
2162 ret = regnode(RE_VISUAL); | |
2163 break; | |
2164 | |
7 | 2165 /* \%[abc]: Emit as a list of branches, all ending at the last |
2166 * branch which matches nothing. */ | |
2167 case '[': | |
2168 if (one_exactly) /* doesn't nest */ | |
2169 EMSG_ONE_RET_NULL; | |
2170 { | |
2171 char_u *lastbranch; | |
2172 char_u *lastnode = NULL; | |
2173 char_u *br; | |
2174 | |
2175 ret = NULL; | |
2176 while ((c = getchr()) != ']') | |
2177 { | |
2178 if (c == NUL) | |
4444 | 2179 EMSG2_RET_NULL(_("E69: Missing ] after %s%%["), |
7 | 2180 reg_magic == MAGIC_ALL); |
2181 br = regnode(BRANCH); | |
2182 if (ret == NULL) | |
2183 ret = br; | |
2184 else | |
2185 regtail(lastnode, br); | |
2186 | |
2187 ungetchr(); | |
2188 one_exactly = TRUE; | |
2189 lastnode = regatom(flagp); | |
2190 one_exactly = FALSE; | |
2191 if (lastnode == NULL) | |
2192 return NULL; | |
2193 } | |
2194 if (ret == NULL) | |
4444 | 2195 EMSG2_RET_NULL(_("E70: Empty %s%%[]"), |
7 | 2196 reg_magic == MAGIC_ALL); |
2197 lastbranch = regnode(BRANCH); | |
2198 br = regnode(NOTHING); | |
2199 if (ret != JUST_CALC_SIZE) | |
2200 { | |
2201 regtail(lastnode, br); | |
2202 regtail(lastbranch, br); | |
2203 /* connect all branches to the NOTHING | |
2204 * branch at the end */ | |
2205 for (br = ret; br != lastnode; ) | |
2206 { | |
2207 if (OP(br) == BRANCH) | |
2208 { | |
2209 regtail(br, lastbranch); | |
2210 br = OPERAND(br); | |
2211 } | |
2212 else | |
2213 br = regnext(br); | |
2214 } | |
2215 } | |
1701 | 2216 *flagp &= ~(HASWIDTH | SIMPLE); |
7 | 2217 break; |
2218 } | |
2219 | |
24 | 2220 case 'd': /* %d123 decimal */ |
2221 case 'o': /* %o123 octal */ | |
2222 case 'x': /* %xab hex 2 */ | |
2223 case 'u': /* %uabcd hex 4 */ | |
2224 case 'U': /* %U1234abcd hex 8 */ | |
2225 { | |
2226 int i; | |
2227 | |
2228 switch (c) | |
2229 { | |
2230 case 'd': i = getdecchrs(); break; | |
2231 case 'o': i = getoctchrs(); break; | |
2232 case 'x': i = gethexchrs(2); break; | |
2233 case 'u': i = gethexchrs(4); break; | |
2234 case 'U': i = gethexchrs(8); break; | |
2235 default: i = -1; break; | |
2236 } | |
2237 | |
2238 if (i < 0) | |
4444 | 2239 EMSG2_RET_NULL( |
24 | 2240 _("E678: Invalid character after %s%%[dxouU]"), |
2241 reg_magic == MAGIC_ALL); | |
714 | 2242 #ifdef FEAT_MBYTE |
2243 if (use_multibytecode(i)) | |
2244 ret = regnode(MULTIBYTECODE); | |
2245 else | |
2246 #endif | |
2247 ret = regnode(EXACTLY); | |
24 | 2248 if (i == 0) |
2249 regc(0x0a); | |
2250 else | |
2251 #ifdef FEAT_MBYTE | |
2252 regmbc(i); | |
2253 #else | |
2254 regc(i); | |
2255 #endif | |
2256 regc(NUL); | |
2257 *flagp |= HASWIDTH; | |
2258 break; | |
2259 } | |
2260 | |
7 | 2261 default: |
639 | 2262 if (VIM_ISDIGIT(c) || c == '<' || c == '>' |
2263 || c == '\'') | |
7 | 2264 { |
2265 long_u n = 0; | |
2266 int cmp; | |
2267 | |
2268 cmp = c; | |
2269 if (cmp == '<' || cmp == '>') | |
2270 c = getchr(); | |
2271 while (VIM_ISDIGIT(c)) | |
2272 { | |
2273 n = n * 10 + (c - '0'); | |
2274 c = getchr(); | |
2275 } | |
639 | 2276 if (c == '\'' && n == 0) |
2277 { | |
2278 /* "\%'m", "\%<'m" and "\%>'m": Mark */ | |
2279 c = getchr(); | |
2280 ret = regnode(RE_MARK); | |
2281 if (ret == JUST_CALC_SIZE) | |
2282 regsize += 2; | |
2283 else | |
2284 { | |
2285 *regcode++ = c; | |
2286 *regcode++ = cmp; | |
2287 } | |
2288 break; | |
2289 } | |
2290 else if (c == 'l' || c == 'c' || c == 'v') | |
7 | 2291 { |
2292 if (c == 'l') | |
2293 ret = regnode(RE_LNUM); | |
2294 else if (c == 'c') | |
2295 ret = regnode(RE_COL); | |
2296 else | |
2297 ret = regnode(RE_VCOL); | |
2298 if (ret == JUST_CALC_SIZE) | |
2299 regsize += 5; | |
2300 else | |
2301 { | |
2302 /* put the number and the optional | |
2303 * comparator after the opcode */ | |
2304 regcode = re_put_long(regcode, n); | |
2305 *regcode++ = cmp; | |
2306 } | |
2307 break; | |
2308 } | |
2309 } | |
2310 | |
4444 | 2311 EMSG2_RET_NULL(_("E71: Invalid character after %s%%"), |
7 | 2312 reg_magic == MAGIC_ALL); |
2313 } | |
2314 } | |
2315 break; | |
2316 | |
2317 case Magic('['): | |
2318 collection: | |
2319 { | |
2320 char_u *lp; | |
2321 | |
2322 /* | |
2323 * If there is no matching ']', we assume the '[' is a normal | |
2324 * character. This makes 'incsearch' and ":help [" work. | |
2325 */ | |
2326 lp = skip_anyof(regparse); | |
2327 if (*lp == ']') /* there is a matching ']' */ | |
2328 { | |
2329 int startc = -1; /* > 0 when next '-' is a range */ | |
2330 int endc; | |
2331 | |
2332 /* | |
2333 * In a character class, different parsing rules apply. | |
2334 * Not even \ is special anymore, nothing is. | |
2335 */ | |
2336 if (*regparse == '^') /* Complement of range. */ | |
2337 { | |
2338 ret = regnode(ANYBUT + extra); | |
2339 regparse++; | |
2340 } | |
2341 else | |
2342 ret = regnode(ANYOF + extra); | |
2343 | |
2344 /* At the start ']' and '-' mean the literal character. */ | |
2345 if (*regparse == ']' || *regparse == '-') | |
167 | 2346 { |
2347 startc = *regparse; | |
7 | 2348 regc(*regparse++); |
167 | 2349 } |
7 | 2350 |
2351 while (*regparse != NUL && *regparse != ']') | |
2352 { | |
2353 if (*regparse == '-') | |
2354 { | |
2355 ++regparse; | |
2356 /* The '-' is not used for a range at the end and | |
2357 * after or before a '\n'. */ | |
2358 if (*regparse == ']' || *regparse == NUL | |
2359 || startc == -1 | |
2360 || (regparse[0] == '\\' && regparse[1] == 'n')) | |
2361 { | |
2362 regc('-'); | |
2363 startc = '-'; /* [--x] is a range */ | |
2364 } | |
2365 else | |
2366 { | |
167 | 2367 /* Also accept "a-[.z.]" */ |
2368 endc = 0; | |
2369 if (*regparse == '[') | |
2370 endc = get_coll_element(®parse); | |
2371 if (endc == 0) | |
2372 { | |
7 | 2373 #ifdef FEAT_MBYTE |
167 | 2374 if (has_mbyte) |
2375 endc = mb_ptr2char_adv(®parse); | |
2376 else | |
7 | 2377 #endif |
167 | 2378 endc = *regparse++; |
2379 } | |
24 | 2380 |
2381 /* Handle \o40, \x20 and \u20AC style sequences */ | |
167 | 2382 if (endc == '\\' && !cpo_lit && !cpo_bsl) |
24 | 2383 endc = coll_get_char(); |
2384 | |
7 | 2385 if (startc > endc) |
2386 EMSG_RET_NULL(_(e_invrange)); | |
2387 #ifdef FEAT_MBYTE | |
2388 if (has_mbyte && ((*mb_char2len)(startc) > 1 | |
2389 || (*mb_char2len)(endc) > 1)) | |
2390 { | |
2391 /* Limit to a range of 256 chars */ | |
2392 if (endc > startc + 256) | |
2393 EMSG_RET_NULL(_(e_invrange)); | |
2394 while (++startc <= endc) | |
2395 regmbc(startc); | |
2396 } | |
2397 else | |
2398 #endif | |
2399 { | |
2400 #ifdef EBCDIC | |
2401 int alpha_only = FALSE; | |
2402 | |
2403 /* for alphabetical range skip the gaps | |
2404 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */ | |
2405 if (isalpha(startc) && isalpha(endc)) | |
2406 alpha_only = TRUE; | |
2407 #endif | |
2408 while (++startc <= endc) | |
2409 #ifdef EBCDIC | |
2410 if (!alpha_only || isalpha(startc)) | |
2411 #endif | |
2412 regc(startc); | |
2413 } | |
2414 startc = -1; | |
2415 } | |
2416 } | |
2417 /* | |
2418 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim | |
2419 * accepts "\t", "\e", etc., but only when the 'l' flag in | |
2420 * 'cpoptions' is not included. | |
167 | 2421 * Posix doesn't recognize backslash at all. |
7 | 2422 */ |
2423 else if (*regparse == '\\' | |
167 | 2424 && !cpo_bsl |
7 | 2425 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL |
2426 || (!cpo_lit | |
2427 && vim_strchr(REGEXP_ABBR, | |
2428 regparse[1]) != NULL))) | |
2429 { | |
2430 regparse++; | |
2431 if (*regparse == 'n') | |
2432 { | |
2433 /* '\n' in range: also match NL */ | |
2434 if (ret != JUST_CALC_SIZE) | |
2435 { | |
4084 | 2436 /* Using \n inside [^] does not change what |
2437 * matches. "[^\n]" is the same as ".". */ | |
2438 if (*ret == ANYOF) | |
2439 { | |
7 | 2440 *ret = ANYOF + ADD_NL; |
4084 | 2441 *flagp |= HASNL; |
2442 } | |
7 | 2443 /* else: must have had a \n already */ |
2444 } | |
2445 regparse++; | |
2446 startc = -1; | |
2447 } | |
24 | 2448 else if (*regparse == 'd' |
2449 || *regparse == 'o' | |
2450 || *regparse == 'x' | |
2451 || *regparse == 'u' | |
2452 || *regparse == 'U') | |
2453 { | |
2454 startc = coll_get_char(); | |
2455 if (startc == 0) | |
2456 regc(0x0a); | |
2457 else | |
2458 #ifdef FEAT_MBYTE | |
2459 regmbc(startc); | |
2460 #else | |
2461 regc(startc); | |
2462 #endif | |
2463 } | |
7 | 2464 else |
2465 { | |
2466 startc = backslash_trans(*regparse++); | |
2467 regc(startc); | |
2468 } | |
2469 } | |
2470 else if (*regparse == '[') | |
2471 { | |
2472 int c_class; | |
2473 int cu; | |
2474 | |
167 | 2475 c_class = get_char_class(®parse); |
7 | 2476 startc = -1; |
2477 /* Characters assumed to be 8 bits! */ | |
2478 switch (c_class) | |
2479 { | |
2480 case CLASS_NONE: | |
167 | 2481 c_class = get_equi_class(®parse); |
2482 if (c_class != 0) | |
2483 { | |
2484 /* produce equivalence class */ | |
2485 reg_equi_class(c_class); | |
2486 } | |
2487 else if ((c_class = | |
2488 get_coll_element(®parse)) != 0) | |
2489 { | |
2490 /* produce a collating element */ | |
2491 regmbc(c_class); | |
2492 } | |
2493 else | |
2494 { | |
2495 /* literal '[', allow [[-x] as a range */ | |
2496 startc = *regparse++; | |
2497 regc(startc); | |
2498 } | |
7 | 2499 break; |
2500 case CLASS_ALNUM: | |
2501 for (cu = 1; cu <= 255; cu++) | |
2502 if (isalnum(cu)) | |
2503 regc(cu); | |
2504 break; | |
2505 case CLASS_ALPHA: | |
2506 for (cu = 1; cu <= 255; cu++) | |
2507 if (isalpha(cu)) | |
2508 regc(cu); | |
2509 break; | |
2510 case CLASS_BLANK: | |
2511 regc(' '); | |
2512 regc('\t'); | |
2513 break; | |
2514 case CLASS_CNTRL: | |
2515 for (cu = 1; cu <= 255; cu++) | |
2516 if (iscntrl(cu)) | |
2517 regc(cu); | |
2518 break; | |
2519 case CLASS_DIGIT: | |
2520 for (cu = 1; cu <= 255; cu++) | |
2521 if (VIM_ISDIGIT(cu)) | |
2522 regc(cu); | |
2523 break; | |
2524 case CLASS_GRAPH: | |
2525 for (cu = 1; cu <= 255; cu++) | |
2526 if (isgraph(cu)) | |
2527 regc(cu); | |
2528 break; | |
2529 case CLASS_LOWER: | |
2530 for (cu = 1; cu <= 255; cu++) | |
1347 | 2531 if (MB_ISLOWER(cu)) |
7 | 2532 regc(cu); |
2533 break; | |
2534 case CLASS_PRINT: | |
2535 for (cu = 1; cu <= 255; cu++) | |
2536 if (vim_isprintc(cu)) | |
2537 regc(cu); | |
2538 break; | |
2539 case CLASS_PUNCT: | |
2540 for (cu = 1; cu <= 255; cu++) | |
2541 if (ispunct(cu)) | |
2542 regc(cu); | |
2543 break; | |
2544 case CLASS_SPACE: | |
2545 for (cu = 9; cu <= 13; cu++) | |
2546 regc(cu); | |
2547 regc(' '); | |
2548 break; | |
2549 case CLASS_UPPER: | |
2550 for (cu = 1; cu <= 255; cu++) | |
1347 | 2551 if (MB_ISUPPER(cu)) |
7 | 2552 regc(cu); |
2553 break; | |
2554 case CLASS_XDIGIT: | |
2555 for (cu = 1; cu <= 255; cu++) | |
2556 if (vim_isxdigit(cu)) | |
2557 regc(cu); | |
2558 break; | |
2559 case CLASS_TAB: | |
2560 regc('\t'); | |
2561 break; | |
2562 case CLASS_RETURN: | |
2563 regc('\r'); | |
2564 break; | |
2565 case CLASS_BACKSPACE: | |
2566 regc('\b'); | |
2567 break; | |
2568 case CLASS_ESCAPE: | |
2569 regc('\033'); | |
2570 break; | |
2571 } | |
2572 } | |
2573 else | |
2574 { | |
2575 #ifdef FEAT_MBYTE | |
2576 if (has_mbyte) | |
2577 { | |
2578 int len; | |
2579 | |
2580 /* produce a multibyte character, including any | |
2581 * following composing characters */ | |
2582 startc = mb_ptr2char(regparse); | |
474 | 2583 len = (*mb_ptr2len)(regparse); |
7 | 2584 if (enc_utf8 && utf_char2len(startc) != len) |
2585 startc = -1; /* composing chars */ | |
2586 while (--len >= 0) | |
2587 regc(*regparse++); | |
2588 } | |
2589 else | |
2590 #endif | |
2591 { | |
2592 startc = *regparse++; | |
2593 regc(startc); | |
2594 } | |
2595 } | |
2596 } | |
2597 regc(NUL); | |
2598 prevchr_len = 1; /* last char was the ']' */ | |
2599 if (*regparse != ']') | |
2600 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */ | |
2601 skipchr(); /* let's be friends with the lexer again */ | |
2602 *flagp |= HASWIDTH | SIMPLE; | |
2603 break; | |
2604 } | |
481 | 2605 else if (reg_strict) |
4444 | 2606 EMSG2_RET_NULL(_(e_missingbracket), reg_magic > MAGIC_OFF); |
7 | 2607 } |
2608 /* FALLTHROUGH */ | |
2609 | |
2610 default: | |
2611 { | |
2612 int len; | |
2613 | |
2614 #ifdef FEAT_MBYTE | |
2615 /* A multi-byte character is handled as a separate atom if it's | |
714 | 2616 * before a multi and when it's a composing char. */ |
2617 if (use_multibytecode(c)) | |
7 | 2618 { |
714 | 2619 do_multibyte: |
7 | 2620 ret = regnode(MULTIBYTECODE); |
2621 regmbc(c); | |
2622 *flagp |= HASWIDTH | SIMPLE; | |
2623 break; | |
2624 } | |
2625 #endif | |
2626 | |
2627 ret = regnode(EXACTLY); | |
2628 | |
2629 /* | |
2630 * Append characters as long as: | |
2631 * - there is no following multi, we then need the character in | |
2632 * front of it as a single character operand | |
2633 * - not running into a Magic character | |
2634 * - "one_exactly" is not set | |
2635 * But always emit at least one character. Might be a Multi, | |
2636 * e.g., a "[" without matching "]". | |
2637 */ | |
2638 for (len = 0; c != NUL && (len == 0 | |
2639 || (re_multi_type(peekchr()) == NOT_MULTI | |
2640 && !one_exactly | |
2641 && !is_Magic(c))); ++len) | |
2642 { | |
2643 c = no_Magic(c); | |
2644 #ifdef FEAT_MBYTE | |
2645 if (has_mbyte) | |
2646 { | |
2647 regmbc(c); | |
2648 if (enc_utf8) | |
2649 { | |
2650 int l; | |
2651 | |
714 | 2652 /* Need to get composing character too. */ |
7 | 2653 for (;;) |
2654 { | |
714 | 2655 l = utf_ptr2len(regparse); |
2656 if (!UTF_COMPOSINGLIKE(regparse, regparse + l)) | |
7 | 2657 break; |
714 | 2658 regmbc(utf_ptr2char(regparse)); |
2659 skipchr(); | |
7 | 2660 } |
2661 } | |
2662 } | |
2663 else | |
2664 #endif | |
2665 regc(c); | |
2666 c = getchr(); | |
2667 } | |
2668 ungetchr(); | |
2669 | |
2670 regc(NUL); | |
2671 *flagp |= HASWIDTH; | |
2672 if (len == 1) | |
2673 *flagp |= SIMPLE; | |
2674 } | |
2675 break; | |
2676 } | |
2677 | |
2678 return ret; | |
2679 } | |
2680 | |
714 | 2681 #ifdef FEAT_MBYTE |
2682 /* | |
2683 * Return TRUE if MULTIBYTECODE should be used instead of EXACTLY for | |
2684 * character "c". | |
2685 */ | |
2686 static int | |
2687 use_multibytecode(c) | |
2688 int c; | |
2689 { | |
2690 return has_mbyte && (*mb_char2len)(c) > 1 | |
2691 && (re_multi_type(peekchr()) != NOT_MULTI | |
2692 || (enc_utf8 && utf_iscomposing(c))); | |
2693 } | |
2694 #endif | |
2695 | |
7 | 2696 /* |
4444 | 2697 * Emit a node. |
7 | 2698 * Return pointer to generated code. |
2699 */ | |
2700 static char_u * | |
2701 regnode(op) | |
2702 int op; | |
2703 { | |
2704 char_u *ret; | |
2705 | |
2706 ret = regcode; | |
2707 if (ret == JUST_CALC_SIZE) | |
2708 regsize += 3; | |
2709 else | |
2710 { | |
2711 *regcode++ = op; | |
2712 *regcode++ = NUL; /* Null "next" pointer. */ | |
2713 *regcode++ = NUL; | |
2714 } | |
2715 return ret; | |
2716 } | |
2717 | |
2718 /* | |
2719 * Emit (if appropriate) a byte of code | |
2720 */ | |
2721 static void | |
2722 regc(b) | |
2723 int b; | |
2724 { | |
2725 if (regcode == JUST_CALC_SIZE) | |
2726 regsize++; | |
2727 else | |
2728 *regcode++ = b; | |
2729 } | |
2730 | |
2731 #ifdef FEAT_MBYTE | |
2732 /* | |
2733 * Emit (if appropriate) a multi-byte character of code | |
2734 */ | |
2735 static void | |
2736 regmbc(c) | |
2737 int c; | |
2738 { | |
2974 | 2739 if (!has_mbyte && c > 0xff) |
2740 return; | |
7 | 2741 if (regcode == JUST_CALC_SIZE) |
2742 regsize += (*mb_char2len)(c); | |
2743 else | |
2744 regcode += (*mb_char2bytes)(c, regcode); | |
2745 } | |
2746 #endif | |
2747 | |
2748 /* | |
4444 | 2749 * Insert an operator in front of already-emitted operand |
7 | 2750 * |
2751 * Means relocating the operand. | |
2752 */ | |
2753 static void | |
2754 reginsert(op, opnd) | |
2755 int op; | |
2756 char_u *opnd; | |
2757 { | |
2758 char_u *src; | |
2759 char_u *dst; | |
2760 char_u *place; | |
2761 | |
2762 if (regcode == JUST_CALC_SIZE) | |
2763 { | |
2764 regsize += 3; | |
2765 return; | |
2766 } | |
2767 src = regcode; | |
2768 regcode += 3; | |
2769 dst = regcode; | |
2770 while (src > opnd) | |
2771 *--dst = *--src; | |
2772 | |
2773 place = opnd; /* Op node, where operand used to be. */ | |
2774 *place++ = op; | |
2775 *place++ = NUL; | |
2776 *place = NUL; | |
2777 } | |
2778 | |
2779 /* | |
4444 | 2780 * Insert an operator in front of already-emitted operand. |
7 | 2781 * The operator has the given limit values as operands. Also set next pointer. |
2782 * | |
2783 * Means relocating the operand. | |
2784 */ | |
2785 static void | |
2786 reginsert_limits(op, minval, maxval, opnd) | |
2787 int op; | |
2788 long minval; | |
2789 long maxval; | |
2790 char_u *opnd; | |
2791 { | |
2792 char_u *src; | |
2793 char_u *dst; | |
2794 char_u *place; | |
2795 | |
2796 if (regcode == JUST_CALC_SIZE) | |
2797 { | |
2798 regsize += 11; | |
2799 return; | |
2800 } | |
2801 src = regcode; | |
2802 regcode += 11; | |
2803 dst = regcode; | |
2804 while (src > opnd) | |
2805 *--dst = *--src; | |
2806 | |
2807 place = opnd; /* Op node, where operand used to be. */ | |
2808 *place++ = op; | |
2809 *place++ = NUL; | |
2810 *place++ = NUL; | |
2811 place = re_put_long(place, (long_u)minval); | |
2812 place = re_put_long(place, (long_u)maxval); | |
2813 regtail(opnd, place); | |
2814 } | |
2815 | |
2816 /* | |
2817 * Write a long as four bytes at "p" and return pointer to the next char. | |
2818 */ | |
2819 static char_u * | |
2820 re_put_long(p, val) | |
2821 char_u *p; | |
2822 long_u val; | |
2823 { | |
2824 *p++ = (char_u) ((val >> 24) & 0377); | |
2825 *p++ = (char_u) ((val >> 16) & 0377); | |
2826 *p++ = (char_u) ((val >> 8) & 0377); | |
2827 *p++ = (char_u) (val & 0377); | |
2828 return p; | |
2829 } | |
2830 | |
2831 /* | |
4444 | 2832 * Set the next-pointer at the end of a node chain. |
7 | 2833 */ |
2834 static void | |
2835 regtail(p, val) | |
2836 char_u *p; | |
2837 char_u *val; | |
2838 { | |
2839 char_u *scan; | |
2840 char_u *temp; | |
2841 int offset; | |
2842 | |
2843 if (p == JUST_CALC_SIZE) | |
2844 return; | |
2845 | |
2846 /* Find last node. */ | |
2847 scan = p; | |
2848 for (;;) | |
2849 { | |
2850 temp = regnext(scan); | |
2851 if (temp == NULL) | |
2852 break; | |
2853 scan = temp; | |
2854 } | |
2855 | |
233 | 2856 if (OP(scan) == BACK) |
7 | 2857 offset = (int)(scan - val); |
2858 else | |
2859 offset = (int)(val - scan); | |
2010 | 2860 /* When the offset uses more than 16 bits it can no longer fit in the two |
2974 | 2861 * bytes available. Use a global flag to avoid having to check return |
2010 | 2862 * values in too many places. */ |
2863 if (offset > 0xffff) | |
2864 reg_toolong = TRUE; | |
2865 else | |
2866 { | |
2867 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377); | |
2868 *(scan + 2) = (char_u) (offset & 0377); | |
2869 } | |
7 | 2870 } |
2871 | |
2872 /* | |
4444 | 2873 * Like regtail, on item after a BRANCH; nop if none. |
7 | 2874 */ |
2875 static void | |
2876 regoptail(p, val) | |
2877 char_u *p; | |
2878 char_u *val; | |
2879 { | |
2880 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */ | |
2881 if (p == NULL || p == JUST_CALC_SIZE | |
2882 || (OP(p) != BRANCH | |
2883 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9))) | |
2884 return; | |
2885 regtail(OPERAND(p), val); | |
2886 } | |
2887 | |
2888 /* | |
4444 | 2889 * Functions for getting characters from the regexp input. |
7 | 2890 */ |
2891 | |
2892 static int at_start; /* True when on the first character */ | |
2893 static int prev_at_start; /* True when on the second character */ | |
2894 | |
4444 | 2895 /* |
2896 * Start parsing at "str". | |
2897 */ | |
7 | 2898 static void |
2899 initchr(str) | |
2900 char_u *str; | |
2901 { | |
2902 regparse = str; | |
2903 prevchr_len = 0; | |
2904 curchr = prevprevchr = prevchr = nextchr = -1; | |
2905 at_start = TRUE; | |
2906 prev_at_start = FALSE; | |
2907 } | |
2908 | |
4444 | 2909 /* |
2910 * Get the next character without advancing. | |
2911 */ | |
7 | 2912 static int |
2913 peekchr() | |
2914 { | |
167 | 2915 static int after_slash = FALSE; |
2916 | |
7 | 2917 if (curchr == -1) |
2918 { | |
2919 switch (curchr = regparse[0]) | |
2920 { | |
2921 case '.': | |
2922 case '[': | |
2923 case '~': | |
2924 /* magic when 'magic' is on */ | |
2925 if (reg_magic >= MAGIC_ON) | |
2926 curchr = Magic(curchr); | |
2927 break; | |
2928 case '(': | |
2929 case ')': | |
2930 case '{': | |
2931 case '%': | |
2932 case '+': | |
2933 case '=': | |
2934 case '?': | |
2935 case '@': | |
2936 case '!': | |
2937 case '&': | |
2938 case '|': | |
2939 case '<': | |
2940 case '>': | |
2941 case '#': /* future ext. */ | |
2942 case '"': /* future ext. */ | |
2943 case '\'': /* future ext. */ | |
2944 case ',': /* future ext. */ | |
2945 case '-': /* future ext. */ | |
2946 case ':': /* future ext. */ | |
2947 case ';': /* future ext. */ | |
2948 case '`': /* future ext. */ | |
2949 case '/': /* Can't be used in / command */ | |
2950 /* magic only after "\v" */ | |
2951 if (reg_magic == MAGIC_ALL) | |
2952 curchr = Magic(curchr); | |
2953 break; | |
2954 case '*': | |
167 | 2955 /* * is not magic as the very first character, eg "?*ptr", when |
2956 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But | |
2957 * "\(\*" is not magic, thus must be magic if "after_slash" */ | |
2958 if (reg_magic >= MAGIC_ON | |
2959 && !at_start | |
2960 && !(prev_at_start && prevchr == Magic('^')) | |
2961 && (after_slash | |
2962 || (prevchr != Magic('(') | |
2963 && prevchr != Magic('&') | |
2964 && prevchr != Magic('|')))) | |
7 | 2965 curchr = Magic('*'); |
2966 break; | |
2967 case '^': | |
2968 /* '^' is only magic as the very first character and if it's after | |
2969 * "\(", "\|", "\&' or "\n" */ | |
2970 if (reg_magic >= MAGIC_OFF | |
2971 && (at_start | |
2972 || reg_magic == MAGIC_ALL | |
2973 || prevchr == Magic('(') | |
2974 || prevchr == Magic('|') | |
2975 || prevchr == Magic('&') | |
2976 || prevchr == Magic('n') | |
2977 || (no_Magic(prevchr) == '(' | |
2978 && prevprevchr == Magic('%')))) | |
2979 { | |
2980 curchr = Magic('^'); | |
2981 at_start = TRUE; | |
2982 prev_at_start = FALSE; | |
2983 } | |
2984 break; | |
2985 case '$': | |
2986 /* '$' is only magic as the very last char and if it's in front of | |
2987 * either "\|", "\)", "\&", or "\n" */ | |
2988 if (reg_magic >= MAGIC_OFF) | |
2989 { | |
2990 char_u *p = regparse + 1; | |
2991 | |
2992 /* ignore \c \C \m and \M after '$' */ | |
2993 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C' | |
2994 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z')) | |
2995 p += 2; | |
2996 if (p[0] == NUL | |
2997 || (p[0] == '\\' | |
2998 && (p[1] == '|' || p[1] == '&' || p[1] == ')' | |
2999 || p[1] == 'n')) | |
3000 || reg_magic == MAGIC_ALL) | |
3001 curchr = Magic('$'); | |
3002 } | |
3003 break; | |
3004 case '\\': | |
3005 { | |
3006 int c = regparse[1]; | |
3007 | |
3008 if (c == NUL) | |
3009 curchr = '\\'; /* trailing '\' */ | |
3010 else if ( | |
3011 #ifdef EBCDIC | |
3012 vim_strchr(META, c) | |
3013 #else | |
3014 c <= '~' && META_flags[c] | |
3015 #endif | |
3016 ) | |
3017 { | |
3018 /* | |
3019 * META contains everything that may be magic sometimes, | |
3020 * except ^ and $ ("\^" and "\$" are only magic after | |
3021 * "\v"). We now fetch the next character and toggle its | |
3022 * magicness. Therefore, \ is so meta-magic that it is | |
3023 * not in META. | |
3024 */ | |
3025 curchr = -1; | |
3026 prev_at_start = at_start; | |
3027 at_start = FALSE; /* be able to say "/\*ptr" */ | |
3028 ++regparse; | |
167 | 3029 ++after_slash; |
7 | 3030 peekchr(); |
3031 --regparse; | |
167 | 3032 --after_slash; |
7 | 3033 curchr = toggle_Magic(curchr); |
3034 } | |
3035 else if (vim_strchr(REGEXP_ABBR, c)) | |
3036 { | |
3037 /* | |
3038 * Handle abbreviations, like "\t" for TAB -- webb | |
3039 */ | |
3040 curchr = backslash_trans(c); | |
3041 } | |
3042 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^')) | |
3043 curchr = toggle_Magic(c); | |
3044 else | |
3045 { | |
3046 /* | |
3047 * Next character can never be (made) magic? | |
3048 * Then backslashing it won't do anything. | |
3049 */ | |
3050 #ifdef FEAT_MBYTE | |
3051 if (has_mbyte) | |
3052 curchr = (*mb_ptr2char)(regparse + 1); | |
3053 else | |
3054 #endif | |
3055 curchr = c; | |
3056 } | |
3057 break; | |
3058 } | |
3059 | |
3060 #ifdef FEAT_MBYTE | |
3061 default: | |
3062 if (has_mbyte) | |
3063 curchr = (*mb_ptr2char)(regparse); | |
3064 #endif | |
3065 } | |
3066 } | |
3067 | |
3068 return curchr; | |
3069 } | |
3070 | |
3071 /* | |
3072 * Eat one lexed character. Do this in a way that we can undo it. | |
3073 */ | |
3074 static void | |
3075 skipchr() | |
3076 { | |
3077 /* peekchr() eats a backslash, do the same here */ | |
3078 if (*regparse == '\\') | |
3079 prevchr_len = 1; | |
3080 else | |
3081 prevchr_len = 0; | |
3082 if (regparse[prevchr_len] != NUL) | |
3083 { | |
3084 #ifdef FEAT_MBYTE | |
714 | 3085 if (enc_utf8) |
1449 | 3086 /* exclude composing chars that mb_ptr2len does include */ |
3087 prevchr_len += utf_ptr2len(regparse + prevchr_len); | |
714 | 3088 else if (has_mbyte) |
474 | 3089 prevchr_len += (*mb_ptr2len)(regparse + prevchr_len); |
7 | 3090 else |
3091 #endif | |
3092 ++prevchr_len; | |
3093 } | |
3094 regparse += prevchr_len; | |
3095 prev_at_start = at_start; | |
3096 at_start = FALSE; | |
3097 prevprevchr = prevchr; | |
3098 prevchr = curchr; | |
3099 curchr = nextchr; /* use previously unget char, or -1 */ | |
3100 nextchr = -1; | |
3101 } | |
3102 | |
3103 /* | |
3104 * Skip a character while keeping the value of prev_at_start for at_start. | |
3105 * prevchr and prevprevchr are also kept. | |
3106 */ | |
3107 static void | |
3108 skipchr_keepstart() | |
3109 { | |
3110 int as = prev_at_start; | |
3111 int pr = prevchr; | |
3112 int prpr = prevprevchr; | |
3113 | |
3114 skipchr(); | |
3115 at_start = as; | |
3116 prevchr = pr; | |
3117 prevprevchr = prpr; | |
3118 } | |
3119 | |
4444 | 3120 /* |
3121 * Get the next character from the pattern. We know about magic and such, so | |
3122 * therefore we need a lexical analyzer. | |
3123 */ | |
7 | 3124 static int |
3125 getchr() | |
3126 { | |
3127 int chr = peekchr(); | |
3128 | |
3129 skipchr(); | |
3130 return chr; | |
3131 } | |
3132 | |
3133 /* | |
3134 * put character back. Works only once! | |
3135 */ | |
3136 static void | |
3137 ungetchr() | |
3138 { | |
3139 nextchr = curchr; | |
3140 curchr = prevchr; | |
3141 prevchr = prevprevchr; | |
3142 at_start = prev_at_start; | |
3143 prev_at_start = FALSE; | |
3144 | |
3145 /* Backup regparse, so that it's at the same position as before the | |
3146 * getchr(). */ | |
3147 regparse -= prevchr_len; | |
3148 } | |
3149 | |
3150 /* | |
29 | 3151 * Get and return the value of the hex string at the current position. |
3152 * Return -1 if there is no valid hex number. | |
3153 * The position is updated: | |
24 | 3154 * blahblah\%x20asdf |
856 | 3155 * before-^ ^-after |
24 | 3156 * The parameter controls the maximum number of input characters. This will be |
3157 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence. | |
3158 */ | |
3159 static int | |
3160 gethexchrs(maxinputlen) | |
3161 int maxinputlen; | |
3162 { | |
3163 int nr = 0; | |
3164 int c; | |
3165 int i; | |
3166 | |
3167 for (i = 0; i < maxinputlen; ++i) | |
3168 { | |
3169 c = regparse[0]; | |
3170 if (!vim_isxdigit(c)) | |
3171 break; | |
3172 nr <<= 4; | |
3173 nr |= hex2nr(c); | |
3174 ++regparse; | |
3175 } | |
3176 | |
3177 if (i == 0) | |
3178 return -1; | |
3179 return nr; | |
3180 } | |
3181 | |
3182 /* | |
3183 * get and return the value of the decimal string immediately after the | |
3184 * current position. Return -1 for invalid. Consumes all digits. | |
3185 */ | |
3186 static int | |
3187 getdecchrs() | |
3188 { | |
3189 int nr = 0; | |
3190 int c; | |
3191 int i; | |
3192 | |
3193 for (i = 0; ; ++i) | |
3194 { | |
3195 c = regparse[0]; | |
3196 if (c < '0' || c > '9') | |
3197 break; | |
3198 nr *= 10; | |
3199 nr += c - '0'; | |
3200 ++regparse; | |
3201 } | |
3202 | |
3203 if (i == 0) | |
3204 return -1; | |
3205 return nr; | |
3206 } | |
3207 | |
3208 /* | |
3209 * get and return the value of the octal string immediately after the current | |
3210 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle | |
3211 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't | |
3212 * treat 8 or 9 as recognised characters. Position is updated: | |
3213 * blahblah\%o210asdf | |
856 | 3214 * before-^ ^-after |
24 | 3215 */ |
3216 static int | |
3217 getoctchrs() | |
3218 { | |
3219 int nr = 0; | |
3220 int c; | |
3221 int i; | |
3222 | |
3223 for (i = 0; i < 3 && nr < 040; ++i) | |
3224 { | |
3225 c = regparse[0]; | |
3226 if (c < '0' || c > '7') | |
3227 break; | |
3228 nr <<= 3; | |
3229 nr |= hex2nr(c); | |
3230 ++regparse; | |
3231 } | |
3232 | |
3233 if (i == 0) | |
3234 return -1; | |
3235 return nr; | |
3236 } | |
3237 | |
3238 /* | |
3239 * Get a number after a backslash that is inside []. | |
3240 * When nothing is recognized return a backslash. | |
3241 */ | |
3242 static int | |
3243 coll_get_char() | |
3244 { | |
3245 int nr = -1; | |
3246 | |
3247 switch (*regparse++) | |
3248 { | |
3249 case 'd': nr = getdecchrs(); break; | |
3250 case 'o': nr = getoctchrs(); break; | |
3251 case 'x': nr = gethexchrs(2); break; | |
3252 case 'u': nr = gethexchrs(4); break; | |
3253 case 'U': nr = gethexchrs(8); break; | |
3254 } | |
3255 if (nr < 0) | |
3256 { | |
3257 /* If getting the number fails be backwards compatible: the character | |
3258 * is a backslash. */ | |
3259 --regparse; | |
3260 nr = '\\'; | |
3261 } | |
3262 return nr; | |
3263 } | |
3264 | |
3265 /* | |
7 | 3266 * read_limits - Read two integers to be taken as a minimum and maximum. |
3267 * If the first character is '-', then the range is reversed. | |
3268 * Should end with 'end'. If minval is missing, zero is default, if maxval is | |
3269 * missing, a very big number is the default. | |
3270 */ | |
3271 static int | |
3272 read_limits(minval, maxval) | |
3273 long *minval; | |
3274 long *maxval; | |
3275 { | |
3276 int reverse = FALSE; | |
3277 char_u *first_char; | |
3278 long tmp; | |
3279 | |
3280 if (*regparse == '-') | |
3281 { | |
3282 /* Starts with '-', so reverse the range later */ | |
3283 regparse++; | |
3284 reverse = TRUE; | |
3285 } | |
3286 first_char = regparse; | |
3287 *minval = getdigits(®parse); | |
3288 if (*regparse == ',') /* There is a comma */ | |
3289 { | |
3290 if (vim_isdigit(*++regparse)) | |
3291 *maxval = getdigits(®parse); | |
3292 else | |
3293 *maxval = MAX_LIMIT; | |
3294 } | |
3295 else if (VIM_ISDIGIT(*first_char)) | |
3296 *maxval = *minval; /* It was \{n} or \{-n} */ | |
3297 else | |
3298 *maxval = MAX_LIMIT; /* It was \{} or \{-} */ | |
3299 if (*regparse == '\\') | |
3300 regparse++; /* Allow either \{...} or \{...\} */ | |
167 | 3301 if (*regparse != '}') |
7 | 3302 { |
3303 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"), | |
3304 reg_magic == MAGIC_ALL ? "" : "\\"); | |
3305 EMSG_RET_FAIL(IObuff); | |
3306 } | |
3307 | |
3308 /* | |
3309 * Reverse the range if there was a '-', or make sure it is in the right | |
3310 * order otherwise. | |
3311 */ | |
3312 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval)) | |
3313 { | |
3314 tmp = *minval; | |
3315 *minval = *maxval; | |
3316 *maxval = tmp; | |
3317 } | |
3318 skipchr(); /* let's be friends with the lexer again */ | |
3319 return OK; | |
3320 } | |
3321 | |
3322 /* | |
3323 * vim_regexec and friends | |
3324 */ | |
3325 | |
3326 /* | |
3327 * Global work variables for vim_regexec(). | |
3328 */ | |
3329 | |
3330 /* The current match-position is remembered with these variables: */ | |
3331 static linenr_T reglnum; /* line number, relative to first line */ | |
3332 static char_u *regline; /* start of current line */ | |
3333 static char_u *reginput; /* current input, points into "regline" */ | |
3334 | |
3335 static int need_clear_subexpr; /* subexpressions still need to be | |
3336 * cleared */ | |
3337 #ifdef FEAT_SYN_HL | |
3338 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions | |
3339 * still need to be cleared */ | |
3340 #endif | |
3341 | |
3342 /* | |
3343 * Structure used to save the current input state, when it needs to be | |
3344 * restored after trying a match. Used by reg_save() and reg_restore(). | |
233 | 3345 * Also stores the length of "backpos". |
7 | 3346 */ |
3347 typedef struct | |
3348 { | |
3349 union | |
3350 { | |
3351 char_u *ptr; /* reginput pointer, for single-line regexp */ | |
3352 lpos_T pos; /* reginput pos, for multi-line regexp */ | |
3353 } rs_u; | |
233 | 3354 int rs_len; |
7 | 3355 } regsave_T; |
3356 | |
3357 /* struct to save start/end pointer/position in for \(\) */ | |
3358 typedef struct | |
3359 { | |
3360 union | |
3361 { | |
3362 char_u *ptr; | |
3363 lpos_T pos; | |
3364 } se_u; | |
3365 } save_se_T; | |
3366 | |
1579 | 3367 /* used for BEHIND and NOBEHIND matching */ |
3368 typedef struct regbehind_S | |
3369 { | |
3370 regsave_T save_after; | |
3371 regsave_T save_behind; | |
1602 | 3372 int save_need_clear_subexpr; |
1579 | 3373 save_se_T save_start[NSUBEXP]; |
3374 save_se_T save_end[NSUBEXP]; | |
3375 } regbehind_T; | |
3376 | |
7 | 3377 static char_u *reg_getline __ARGS((linenr_T lnum)); |
4444 | 3378 static long bt_regexec_both __ARGS((char_u *line, colnr_T col, proftime_T *tm)); |
3379 static long regtry __ARGS((bt_regprog_T *prog, colnr_T col)); | |
7 | 3380 static void cleanup_subexpr __ARGS((void)); |
3381 #ifdef FEAT_SYN_HL | |
3382 static void cleanup_zsubexpr __ARGS((void)); | |
3383 #endif | |
1579 | 3384 static void save_subexpr __ARGS((regbehind_T *bp)); |
3385 static void restore_subexpr __ARGS((regbehind_T *bp)); | |
7 | 3386 static void reg_nextline __ARGS((void)); |
233 | 3387 static void reg_save __ARGS((regsave_T *save, garray_T *gap)); |
3388 static void reg_restore __ARGS((regsave_T *save, garray_T *gap)); | |
7 | 3389 static int reg_save_equal __ARGS((regsave_T *save)); |
3390 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp)); | |
3391 static void save_se_one __ARGS((save_se_T *savep, char_u **pp)); | |
3392 | |
3393 /* Save the sub-expressions before attempting a match. */ | |
3394 #define save_se(savep, posp, pp) \ | |
3395 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp)) | |
3396 | |
3397 /* After a failed match restore the sub-expressions. */ | |
3398 #define restore_se(savep, posp, pp) { \ | |
3399 if (REG_MULTI) \ | |
3400 *(posp) = (savep)->se_u.pos; \ | |
3401 else \ | |
3402 *(pp) = (savep)->se_u.ptr; } | |
3403 | |
3404 static int re_num_cmp __ARGS((long_u val, char_u *scan)); | |
180 | 3405 static int regmatch __ARGS((char_u *prog)); |
7 | 3406 static int regrepeat __ARGS((char_u *p, long maxcount)); |
3407 | |
3408 #ifdef DEBUG | |
3409 int regnarrate = 0; | |
3410 #endif | |
3411 | |
3412 /* | |
3413 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec(). | |
3414 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern | |
3415 * contains '\c' or '\C' the value is overruled. | |
3416 */ | |
3417 static int ireg_ic; | |
3418 | |
3419 #ifdef FEAT_MBYTE | |
3420 /* | |
3421 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag | |
3422 * in the regexp. Defaults to false, always. | |
3423 */ | |
3424 static int ireg_icombine; | |
3425 #endif | |
3426 | |
3427 /* | |
410 | 3428 * Copy of "rmm_maxcol": maximum column to search for a match. Zero when |
3429 * there is no maximum. | |
3430 */ | |
418 | 3431 static colnr_T ireg_maxcol; |
410 | 3432 |
3433 /* | |
7 | 3434 * Sometimes need to save a copy of a line. Since alloc()/free() is very |
3435 * slow, we keep one allocated piece of memory and only re-allocate it when | |
4444 | 3436 * it's too small. It's freed in bt_regexec_both() when finished. |
7 | 3437 */ |
1468 | 3438 static char_u *reg_tofree = NULL; |
7 | 3439 static unsigned reg_tofreelen; |
3440 | |
3441 /* | |
3442 * These variables are set when executing a regexp to speed up the execution. | |
1209 | 3443 * Which ones are set depends on whether a single-line or multi-line match is |
7 | 3444 * done: |
3445 * single-line multi-line | |
3446 * reg_match ®match_T NULL | |
3447 * reg_mmatch NULL ®mmatch_T | |
3448 * reg_startp reg_match->startp <invalid> | |
3449 * reg_endp reg_match->endp <invalid> | |
3450 * reg_startpos <invalid> reg_mmatch->startpos | |
3451 * reg_endpos <invalid> reg_mmatch->endpos | |
3452 * reg_win NULL window in which to search | |
4061 | 3453 * reg_buf curbuf buffer in which to search |
7 | 3454 * reg_firstlnum <invalid> first line in which to search |
3455 * reg_maxline 0 last line nr | |
3456 * reg_line_lbr FALSE or TRUE FALSE | |
3457 */ | |
3458 static regmatch_T *reg_match; | |
3459 static regmmatch_T *reg_mmatch; | |
3460 static char_u **reg_startp = NULL; | |
3461 static char_u **reg_endp = NULL; | |
3462 static lpos_T *reg_startpos = NULL; | |
3463 static lpos_T *reg_endpos = NULL; | |
3464 static win_T *reg_win; | |
3465 static buf_T *reg_buf; | |
3466 static linenr_T reg_firstlnum; | |
3467 static linenr_T reg_maxline; | |
3468 static int reg_line_lbr; /* "\n" in string is line break */ | |
3469 | |
270 | 3470 /* Values for rs_state in regitem_T. */ |
3471 typedef enum regstate_E | |
3472 { | |
3473 RS_NOPEN = 0 /* NOPEN and NCLOSE */ | |
3474 , RS_MOPEN /* MOPEN + [0-9] */ | |
3475 , RS_MCLOSE /* MCLOSE + [0-9] */ | |
3476 #ifdef FEAT_SYN_HL | |
3477 , RS_ZOPEN /* ZOPEN + [0-9] */ | |
3478 , RS_ZCLOSE /* ZCLOSE + [0-9] */ | |
3479 #endif | |
3480 , RS_BRANCH /* BRANCH */ | |
3481 , RS_BRCPLX_MORE /* BRACE_COMPLEX and trying one more match */ | |
3482 , RS_BRCPLX_LONG /* BRACE_COMPLEX and trying longest match */ | |
3483 , RS_BRCPLX_SHORT /* BRACE_COMPLEX and trying shortest match */ | |
3484 , RS_NOMATCH /* NOMATCH */ | |
3485 , RS_BEHIND1 /* BEHIND / NOBEHIND matching rest */ | |
3486 , RS_BEHIND2 /* BEHIND / NOBEHIND matching behind part */ | |
3487 , RS_STAR_LONG /* STAR/PLUS/BRACE_SIMPLE longest match */ | |
3488 , RS_STAR_SHORT /* STAR/PLUS/BRACE_SIMPLE shortest match */ | |
3489 } regstate_T; | |
3490 | |
3491 /* | |
3492 * When there are alternatives a regstate_T is put on the regstack to remember | |
3493 * what we are doing. | |
3494 * Before it may be another type of item, depending on rs_state, to remember | |
3495 * more things. | |
3496 */ | |
3497 typedef struct regitem_S | |
3498 { | |
3499 regstate_T rs_state; /* what we are doing, one of RS_ above */ | |
3500 char_u *rs_scan; /* current node in program */ | |
3501 union | |
3502 { | |
3503 save_se_T sesave; | |
3504 regsave_T regsave; | |
3505 } rs_un; /* room for saving reginput */ | |
1579 | 3506 short rs_no; /* submatch nr or BEHIND/NOBEHIND */ |
270 | 3507 } regitem_T; |
3508 | |
3509 static regitem_T *regstack_push __ARGS((regstate_T state, char_u *scan)); | |
3510 static void regstack_pop __ARGS((char_u **scan)); | |
3511 | |
3512 /* used for STAR, PLUS and BRACE_SIMPLE matching */ | |
3513 typedef struct regstar_S | |
3514 { | |
3515 int nextb; /* next byte */ | |
3516 int nextb_ic; /* next byte reverse case */ | |
3517 long count; | |
3518 long minval; | |
3519 long maxval; | |
3520 } regstar_T; | |
3521 | |
3522 /* used to store input position when a BACK was encountered, so that we now if | |
3523 * we made any progress since the last time. */ | |
3524 typedef struct backpos_S | |
3525 { | |
3526 char_u *bp_scan; /* "scan" where BACK was encountered */ | |
3527 regsave_T bp_pos; /* last input position */ | |
3528 } backpos_T; | |
3529 | |
3530 /* | |
1520 | 3531 * "regstack" and "backpos" are used by regmatch(). They are kept over calls |
3532 * to avoid invoking malloc() and free() often. | |
3533 * "regstack" is a stack with regitem_T items, sometimes preceded by regstar_T | |
3534 * or regbehind_T. | |
3535 * "backpos_T" is a table with backpos_T for BACK | |
270 | 3536 */ |
1520 | 3537 static garray_T regstack = {0, 0, 0, 0, NULL}; |
3538 static garray_T backpos = {0, 0, 0, 0, NULL}; | |
3539 | |
3540 /* | |
3541 * Both for regstack and backpos tables we use the following strategy of | |
3542 * allocation (to reduce malloc/free calls): | |
3543 * - Initial size is fairly small. | |
3544 * - When needed, the tables are grown bigger (8 times at first, double after | |
3545 * that). | |
3546 * - After executing the match we free the memory only if the array has grown. | |
3547 * Thus the memory is kept allocated when it's at the initial size. | |
3548 * This makes it fast while not keeping a lot of memory allocated. | |
3549 * A three times speed increase was observed when using many simple patterns. | |
3550 */ | |
3551 #define REGSTACK_INITIAL 2048 | |
3552 #define BACKPOS_INITIAL 64 | |
3553 | |
3554 #if defined(EXITFREE) || defined(PROTO) | |
3555 void | |
3556 free_regexp_stuff() | |
3557 { | |
3558 ga_clear(®stack); | |
3559 ga_clear(&backpos); | |
3560 vim_free(reg_tofree); | |
3561 vim_free(reg_prev_sub); | |
3562 } | |
3563 #endif | |
270 | 3564 |
7 | 3565 /* |
3566 * Get pointer to the line "lnum", which is relative to "reg_firstlnum". | |
3567 */ | |
3568 static char_u * | |
3569 reg_getline(lnum) | |
3570 linenr_T lnum; | |
3571 { | |
3572 /* when looking behind for a match/no-match lnum is negative. But we | |
3573 * can't go before line 1 */ | |
3574 if (reg_firstlnum + lnum < 1) | |
3575 return NULL; | |
482 | 3576 if (lnum > reg_maxline) |
481 | 3577 /* Must have matched the "\n" in the last line. */ |
3578 return (char_u *)""; | |
7 | 3579 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE); |
3580 } | |
3581 | |
3582 static regsave_T behind_pos; | |
3583 | |
3584 #ifdef FEAT_SYN_HL | |
3585 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */ | |
3586 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */ | |
3587 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */ | |
3588 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */ | |
3589 #endif | |
3590 | |
3591 /* TRUE if using multi-line regexp. */ | |
3592 #define REG_MULTI (reg_match == NULL) | |
3593 | |
4444 | 3594 static int bt_regexec __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); |
3595 | |
7 | 3596 /* |
3597 * Match a regexp against a string. | |
3598 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3599 * Uses curbuf for line count and 'iskeyword'. | |
3600 * | |
3601 * Return TRUE if there is a match, FALSE if not. | |
3602 */ | |
4444 | 3603 static int |
3604 bt_regexec(rmp, line, col) | |
7 | 3605 regmatch_T *rmp; |
3606 char_u *line; /* string to match against */ | |
3607 colnr_T col; /* column to start looking for match */ | |
3608 { | |
3609 reg_match = rmp; | |
3610 reg_mmatch = NULL; | |
3611 reg_maxline = 0; | |
3612 reg_line_lbr = FALSE; | |
4061 | 3613 reg_buf = curbuf; |
7 | 3614 reg_win = NULL; |
3615 ireg_ic = rmp->rm_ic; | |
3616 #ifdef FEAT_MBYTE | |
3617 ireg_icombine = FALSE; | |
3618 #endif | |
410 | 3619 ireg_maxcol = 0; |
4444 | 3620 return (bt_regexec_both(line, col, NULL) != 0); |
7 | 3621 } |
3622 | |
39 | 3623 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ |
3624 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
4444 | 3625 |
3626 static int bt_regexec_nl __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
3627 | |
7 | 3628 /* |
3629 * Like vim_regexec(), but consider a "\n" in "line" to be a line break. | |
3630 */ | |
4444 | 3631 static int |
3632 bt_regexec_nl(rmp, line, col) | |
7 | 3633 regmatch_T *rmp; |
3634 char_u *line; /* string to match against */ | |
3635 colnr_T col; /* column to start looking for match */ | |
3636 { | |
3637 reg_match = rmp; | |
3638 reg_mmatch = NULL; | |
3639 reg_maxline = 0; | |
3640 reg_line_lbr = TRUE; | |
4061 | 3641 reg_buf = curbuf; |
7 | 3642 reg_win = NULL; |
3643 ireg_ic = rmp->rm_ic; | |
3644 #ifdef FEAT_MBYTE | |
3645 ireg_icombine = FALSE; | |
3646 #endif | |
410 | 3647 ireg_maxcol = 0; |
4444 | 3648 return (bt_regexec_both(line, col, NULL) != 0); |
7 | 3649 } |
3650 #endif | |
3651 | |
4444 | 3652 static long bt_regexec_multi __ARGS((regmmatch_T *rmp, win_T *win, buf_T *buf, linenr_T lnum, colnr_T col, proftime_T *tm)); |
3653 | |
7 | 3654 /* |
3655 * Match a regexp against multiple lines. | |
3656 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3657 * Uses curbuf for line count and 'iskeyword'. | |
3658 * | |
3659 * Return zero if there is no match. Return number of lines contained in the | |
3660 * match otherwise. | |
3661 */ | |
4444 | 3662 static long |
3663 bt_regexec_multi(rmp, win, buf, lnum, col, tm) | |
7 | 3664 regmmatch_T *rmp; |
3665 win_T *win; /* window in which to search or NULL */ | |
3666 buf_T *buf; /* buffer in which to search */ | |
3667 linenr_T lnum; /* nr of line to start looking for match */ | |
3668 colnr_T col; /* column to start looking for match */ | |
1521 | 3669 proftime_T *tm; /* timeout limit or NULL */ |
7 | 3670 { |
3671 long r; | |
3672 | |
3673 reg_match = NULL; | |
3674 reg_mmatch = rmp; | |
3675 reg_buf = buf; | |
3676 reg_win = win; | |
3677 reg_firstlnum = lnum; | |
3678 reg_maxline = reg_buf->b_ml.ml_line_count - lnum; | |
3679 reg_line_lbr = FALSE; | |
3680 ireg_ic = rmp->rmm_ic; | |
3681 #ifdef FEAT_MBYTE | |
3682 ireg_icombine = FALSE; | |
3683 #endif | |
410 | 3684 ireg_maxcol = rmp->rmm_maxcol; |
7 | 3685 |
4444 | 3686 r = bt_regexec_both(NULL, col, tm); |
7 | 3687 |
3688 return r; | |
3689 } | |
3690 | |
3691 /* | |
3692 * Match a regexp against a string ("line" points to the string) or multiple | |
3693 * lines ("line" is NULL, use reg_getline()). | |
3694 */ | |
3695 static long | |
4444 | 3696 bt_regexec_both(line, col, tm) |
7 | 3697 char_u *line; |
3698 colnr_T col; /* column to start looking for match */ | |
1877 | 3699 proftime_T *tm UNUSED; /* timeout limit or NULL */ |
7 | 3700 { |
4444 | 3701 bt_regprog_T *prog; |
7 | 3702 char_u *s; |
189 | 3703 long retval = 0L; |
7 | 3704 |
1520 | 3705 /* Create "regstack" and "backpos" if they are not allocated yet. |
3706 * We allocate *_INITIAL amount of bytes first and then set the grow size | |
3707 * to much bigger value to avoid many malloc calls in case of deep regular | |
3708 * expressions. */ | |
3709 if (regstack.ga_data == NULL) | |
3710 { | |
3711 /* Use an item size of 1 byte, since we push different things | |
3712 * onto the regstack. */ | |
3713 ga_init2(®stack, 1, REGSTACK_INITIAL); | |
3714 ga_grow(®stack, REGSTACK_INITIAL); | |
3715 regstack.ga_growsize = REGSTACK_INITIAL * 8; | |
3716 } | |
3717 | |
3718 if (backpos.ga_data == NULL) | |
3719 { | |
3720 ga_init2(&backpos, sizeof(backpos_T), BACKPOS_INITIAL); | |
3721 ga_grow(&backpos, BACKPOS_INITIAL); | |
3722 backpos.ga_growsize = BACKPOS_INITIAL * 8; | |
3723 } | |
270 | 3724 |
7 | 3725 if (REG_MULTI) |
3726 { | |
4444 | 3727 prog = (bt_regprog_T *)reg_mmatch->regprog; |
7 | 3728 line = reg_getline((linenr_T)0); |
3729 reg_startpos = reg_mmatch->startpos; | |
3730 reg_endpos = reg_mmatch->endpos; | |
3731 } | |
3732 else | |
3733 { | |
4444 | 3734 prog = (bt_regprog_T *)reg_match->regprog; |
7 | 3735 reg_startp = reg_match->startp; |
3736 reg_endp = reg_match->endp; | |
3737 } | |
3738 | |
3739 /* Be paranoid... */ | |
3740 if (prog == NULL || line == NULL) | |
3741 { | |
3742 EMSG(_(e_null)); | |
3743 goto theend; | |
3744 } | |
3745 | |
3746 /* Check validity of program. */ | |
3747 if (prog_magic_wrong()) | |
3748 goto theend; | |
3749 | |
410 | 3750 /* If the start column is past the maximum column: no need to try. */ |
3751 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3752 goto theend; | |
3753 | |
7 | 3754 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */ |
3755 if (prog->regflags & RF_ICASE) | |
3756 ireg_ic = TRUE; | |
3757 else if (prog->regflags & RF_NOICASE) | |
3758 ireg_ic = FALSE; | |
3759 | |
3760 #ifdef FEAT_MBYTE | |
3761 /* If pattern contains "\Z" overrule value of ireg_icombine */ | |
3762 if (prog->regflags & RF_ICOMBINE) | |
3763 ireg_icombine = TRUE; | |
3764 #endif | |
3765 | |
3766 /* If there is a "must appear" string, look for it. */ | |
3767 if (prog->regmust != NULL) | |
3768 { | |
3769 int c; | |
3770 | |
3771 #ifdef FEAT_MBYTE | |
3772 if (has_mbyte) | |
3773 c = (*mb_ptr2char)(prog->regmust); | |
3774 else | |
3775 #endif | |
3776 c = *prog->regmust; | |
3777 s = line + col; | |
170 | 3778 |
3779 /* | |
3780 * This is used very often, esp. for ":global". Use three versions of | |
3781 * the loop to avoid overhead of conditions. | |
3782 */ | |
3783 if (!ireg_ic | |
3784 #ifdef FEAT_MBYTE | |
3785 && !has_mbyte | |
3786 #endif | |
3787 ) | |
3788 while ((s = vim_strbyte(s, c)) != NULL) | |
3789 { | |
3790 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3791 break; /* Found it. */ | |
3792 ++s; | |
3793 } | |
3794 #ifdef FEAT_MBYTE | |
3795 else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1)) | |
3796 while ((s = vim_strchr(s, c)) != NULL) | |
3797 { | |
3798 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3799 break; /* Found it. */ | |
3800 mb_ptr_adv(s); | |
3801 } | |
3802 #endif | |
3803 else | |
3804 while ((s = cstrchr(s, c)) != NULL) | |
3805 { | |
3806 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3807 break; /* Found it. */ | |
3808 mb_ptr_adv(s); | |
3809 } | |
7 | 3810 if (s == NULL) /* Not present. */ |
3811 goto theend; | |
3812 } | |
3813 | |
3814 regline = line; | |
3815 reglnum = 0; | |
2578 | 3816 reg_toolong = FALSE; |
7 | 3817 |
3818 /* Simplest case: Anchored match need be tried only once. */ | |
3819 if (prog->reganch) | |
3820 { | |
3821 int c; | |
3822 | |
3823 #ifdef FEAT_MBYTE | |
3824 if (has_mbyte) | |
3825 c = (*mb_ptr2char)(regline + col); | |
3826 else | |
3827 #endif | |
3828 c = regline[col]; | |
3829 if (prog->regstart == NUL | |
3830 || prog->regstart == c | |
3831 || (ireg_ic && (( | |
3832 #ifdef FEAT_MBYTE | |
3833 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c))) | |
3834 || (c < 255 && prog->regstart < 255 && | |
3835 #endif | |
1347 | 3836 MB_TOLOWER(prog->regstart) == MB_TOLOWER(c))))) |
7 | 3837 retval = regtry(prog, col); |
3838 else | |
3839 retval = 0; | |
3840 } | |
3841 else | |
3842 { | |
1521 | 3843 #ifdef FEAT_RELTIME |
3844 int tm_count = 0; | |
3845 #endif | |
7 | 3846 /* Messy cases: unanchored match. */ |
180 | 3847 while (!got_int) |
7 | 3848 { |
3849 if (prog->regstart != NUL) | |
3850 { | |
170 | 3851 /* Skip until the char we know it must start with. |
3852 * Used often, do some work to avoid call overhead. */ | |
3853 if (!ireg_ic | |
3854 #ifdef FEAT_MBYTE | |
3855 && !has_mbyte | |
3856 #endif | |
3857 ) | |
3858 s = vim_strbyte(regline + col, prog->regstart); | |
3859 else | |
3860 s = cstrchr(regline + col, prog->regstart); | |
7 | 3861 if (s == NULL) |
3862 { | |
3863 retval = 0; | |
3864 break; | |
3865 } | |
3866 col = (int)(s - regline); | |
3867 } | |
3868 | |
410 | 3869 /* Check for maximum column to try. */ |
3870 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3871 { | |
3872 retval = 0; | |
3873 break; | |
3874 } | |
3875 | |
7 | 3876 retval = regtry(prog, col); |
3877 if (retval > 0) | |
3878 break; | |
3879 | |
3880 /* if not currently on the first line, get it again */ | |
3881 if (reglnum != 0) | |
3882 { | |
481 | 3883 reglnum = 0; |
7 | 3884 regline = reg_getline((linenr_T)0); |
3885 } | |
3886 if (regline[col] == NUL) | |
3887 break; | |
3888 #ifdef FEAT_MBYTE | |
3889 if (has_mbyte) | |
474 | 3890 col += (*mb_ptr2len)(regline + col); |
7 | 3891 else |
3892 #endif | |
3893 ++col; | |
1521 | 3894 #ifdef FEAT_RELTIME |
3895 /* Check for timeout once in a twenty times to avoid overhead. */ | |
3896 if (tm != NULL && ++tm_count == 20) | |
3897 { | |
3898 tm_count = 0; | |
3899 if (profile_passed_limit(tm)) | |
3900 break; | |
3901 } | |
3902 #endif | |
7 | 3903 } |
3904 } | |
3905 | |
3906 theend: | |
1520 | 3907 /* Free "reg_tofree" when it's a bit big. |
3908 * Free regstack and backpos if they are bigger than their initial size. */ | |
3909 if (reg_tofreelen > 400) | |
3910 { | |
3911 vim_free(reg_tofree); | |
3912 reg_tofree = NULL; | |
3913 } | |
3914 if (regstack.ga_maxlen > REGSTACK_INITIAL) | |
3915 ga_clear(®stack); | |
3916 if (backpos.ga_maxlen > BACKPOS_INITIAL) | |
3917 ga_clear(&backpos); | |
270 | 3918 |
7 | 3919 return retval; |
3920 } | |
3921 | |
3922 #ifdef FEAT_SYN_HL | |
3923 static reg_extmatch_T *make_extmatch __ARGS((void)); | |
3924 | |
3925 /* | |
3926 * Create a new extmatch and mark it as referenced once. | |
3927 */ | |
3928 static reg_extmatch_T * | |
3929 make_extmatch() | |
3930 { | |
3931 reg_extmatch_T *em; | |
3932 | |
3933 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T)); | |
3934 if (em != NULL) | |
3935 em->refcnt = 1; | |
3936 return em; | |
3937 } | |
3938 | |
3939 /* | |
3940 * Add a reference to an extmatch. | |
3941 */ | |
3942 reg_extmatch_T * | |
3943 ref_extmatch(em) | |
3944 reg_extmatch_T *em; | |
3945 { | |
3946 if (em != NULL) | |
3947 em->refcnt++; | |
3948 return em; | |
3949 } | |
3950 | |
3951 /* | |
3952 * Remove a reference to an extmatch. If there are no references left, free | |
3953 * the info. | |
3954 */ | |
3955 void | |
3956 unref_extmatch(em) | |
3957 reg_extmatch_T *em; | |
3958 { | |
3959 int i; | |
3960 | |
3961 if (em != NULL && --em->refcnt <= 0) | |
3962 { | |
3963 for (i = 0; i < NSUBEXP; ++i) | |
3964 vim_free(em->matches[i]); | |
3965 vim_free(em); | |
3966 } | |
3967 } | |
3968 #endif | |
3969 | |
3970 /* | |
3971 * regtry - try match of "prog" with at regline["col"]. | |
3972 * Returns 0 for failure, number of lines contained in the match otherwise. | |
3973 */ | |
3974 static long | |
3975 regtry(prog, col) | |
4444 | 3976 bt_regprog_T *prog; |
7 | 3977 colnr_T col; |
3978 { | |
3979 reginput = regline + col; | |
3980 need_clear_subexpr = TRUE; | |
3981 #ifdef FEAT_SYN_HL | |
3982 /* Clear the external match subpointers if necessary. */ | |
3983 if (prog->reghasz == REX_SET) | |
3984 need_clear_zsubexpr = TRUE; | |
3985 #endif | |
3986 | |
189 | 3987 if (regmatch(prog->program + 1) == 0) |
3988 return 0; | |
3989 | |
3990 cleanup_subexpr(); | |
3991 if (REG_MULTI) | |
7 | 3992 { |
189 | 3993 if (reg_startpos[0].lnum < 0) |
7 | 3994 { |
189 | 3995 reg_startpos[0].lnum = 0; |
3996 reg_startpos[0].col = col; | |
3997 } | |
3998 if (reg_endpos[0].lnum < 0) | |
3999 { | |
4000 reg_endpos[0].lnum = reglnum; | |
4001 reg_endpos[0].col = (int)(reginput - regline); | |
7 | 4002 } |
4003 else | |
189 | 4004 /* Use line number of "\ze". */ |
4005 reglnum = reg_endpos[0].lnum; | |
4006 } | |
4007 else | |
4008 { | |
4009 if (reg_startp[0] == NULL) | |
4010 reg_startp[0] = regline + col; | |
4011 if (reg_endp[0] == NULL) | |
4012 reg_endp[0] = reginput; | |
4013 } | |
7 | 4014 #ifdef FEAT_SYN_HL |
189 | 4015 /* Package any found \z(...\) matches for export. Default is none. */ |
4016 unref_extmatch(re_extmatch_out); | |
4017 re_extmatch_out = NULL; | |
4018 | |
4019 if (prog->reghasz == REX_SET) | |
4020 { | |
4021 int i; | |
4022 | |
4023 cleanup_zsubexpr(); | |
4024 re_extmatch_out = make_extmatch(); | |
4025 for (i = 0; i < NSUBEXP; i++) | |
7 | 4026 { |
189 | 4027 if (REG_MULTI) |
7 | 4028 { |
189 | 4029 /* Only accept single line matches. */ |
4030 if (reg_startzpos[i].lnum >= 0 | |
4031 && reg_endzpos[i].lnum == reg_startzpos[i].lnum) | |
4032 re_extmatch_out->matches[i] = | |
4033 vim_strnsave(reg_getline(reg_startzpos[i].lnum) | |
7 | 4034 + reg_startzpos[i].col, |
189 | 4035 reg_endzpos[i].col - reg_startzpos[i].col); |
4036 } | |
4037 else | |
4038 { | |
4039 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL) | |
4040 re_extmatch_out->matches[i] = | |
7 | 4041 vim_strnsave(reg_startzp[i], |
189 | 4042 (int)(reg_endzp[i] - reg_startzp[i])); |
7 | 4043 } |
4044 } | |
189 | 4045 } |
7 | 4046 #endif |
189 | 4047 return 1 + reglnum; |
7 | 4048 } |
4049 | |
4050 #ifdef FEAT_MBYTE | |
4051 static int reg_prev_class __ARGS((void)); | |
4052 | |
4053 /* | |
4054 * Get class of previous character. | |
4055 */ | |
4056 static int | |
4057 reg_prev_class() | |
4058 { | |
4059 if (reginput > regline) | |
4069 | 4060 return mb_get_class_buf(reginput - 1 |
4061 - (*mb_head_off)(regline, reginput - 1), reg_buf); | |
7 | 4062 return -1; |
4063 } | |
4064 | |
4065 #endif | |
39 | 4066 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput) |
7 | 4067 |
4068 /* | |
4069 * The arguments from BRACE_LIMITS are stored here. They are actually local | |
4070 * to regmatch(), but they are here to reduce the amount of stack space used | |
4071 * (it can be called recursively many times). | |
4072 */ | |
4073 static long bl_minval; | |
4074 static long bl_maxval; | |
4075 | |
4076 /* | |
4077 * regmatch - main matching routine | |
4078 * | |
180 | 4079 * Conceptually the strategy is simple: Check to see whether the current node |
4080 * matches, push an item onto the regstack and loop to see whether the rest | |
4081 * matches, and then act accordingly. In practice we make some effort to | |
4082 * avoid using the regstack, in particular by going through "ordinary" nodes | |
4083 * (that don't need to know whether the rest of the match failed) by a nested | |
4084 * loop. | |
7 | 4085 * |
4086 * Returns TRUE when there is a match. Leaves reginput and reglnum just after | |
4087 * the last matched character. | |
4088 * Returns FALSE when there is no match. Leaves reginput and reglnum in an | |
4089 * undefined state! | |
4090 */ | |
4091 static int | |
180 | 4092 regmatch(scan) |
7 | 4093 char_u *scan; /* Current node. */ |
4094 { | |
180 | 4095 char_u *next; /* Next node. */ |
4096 int op; | |
4097 int c; | |
4098 regitem_T *rp; | |
4099 int no; | |
4100 int status; /* one of the RA_ values: */ | |
4101 #define RA_FAIL 1 /* something failed, abort */ | |
4102 #define RA_CONT 2 /* continue in inner loop */ | |
4103 #define RA_BREAK 3 /* break inner loop */ | |
4104 #define RA_MATCH 4 /* successful match */ | |
4105 #define RA_NOMATCH 5 /* didn't match */ | |
270 | 4106 |
1520 | 4107 /* Make "regstack" and "backpos" empty. They are allocated and freed in |
4444 | 4108 * bt_regexec_both() to reduce malloc()/free() calls. */ |
270 | 4109 regstack.ga_len = 0; |
4110 backpos.ga_len = 0; | |
233 | 4111 |
180 | 4112 /* |
233 | 4113 * Repeat until "regstack" is empty. |
180 | 4114 */ |
4115 for (;;) | |
4116 { | |
4444 | 4117 /* Some patterns may cause a long time to match, even though they are not |
7 | 4118 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */ |
4119 fast_breakcheck(); | |
4120 | |
4121 #ifdef DEBUG | |
4122 if (scan != NULL && regnarrate) | |
4123 { | |
4444 | 4124 mch_errmsg((char *)regprop(scan)); |
7 | 4125 mch_errmsg("(\n"); |
4126 } | |
4127 #endif | |
180 | 4128 |
4129 /* | |
233 | 4130 * Repeat for items that can be matched sequentially, without using the |
180 | 4131 * regstack. |
4132 */ | |
4133 for (;;) | |
7 | 4134 { |
180 | 4135 if (got_int || scan == NULL) |
4136 { | |
4137 status = RA_FAIL; | |
4138 break; | |
4139 } | |
4140 status = RA_CONT; | |
4141 | |
7 | 4142 #ifdef DEBUG |
4143 if (regnarrate) | |
4144 { | |
4444 | 4145 mch_errmsg((char *)regprop(scan)); |
7 | 4146 mch_errmsg("...\n"); |
4147 # ifdef FEAT_SYN_HL | |
4148 if (re_extmatch_in != NULL) | |
4149 { | |
4150 int i; | |
4151 | |
4152 mch_errmsg(_("External submatches:\n")); | |
4153 for (i = 0; i < NSUBEXP; i++) | |
4154 { | |
4155 mch_errmsg(" \""); | |
4156 if (re_extmatch_in->matches[i] != NULL) | |
4444 | 4157 mch_errmsg((char *)re_extmatch_in->matches[i]); |
7 | 4158 mch_errmsg("\"\n"); |
4159 } | |
4160 } | |
4161 # endif | |
4162 } | |
4163 #endif | |
4164 next = regnext(scan); | |
4165 | |
4166 op = OP(scan); | |
4167 /* Check for character class with NL added. */ | |
1018 | 4168 if (!reg_line_lbr && WITH_NL(op) && REG_MULTI |
4169 && *reginput == NUL && reglnum <= reg_maxline) | |
7 | 4170 { |
4171 reg_nextline(); | |
4172 } | |
4173 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n') | |
4174 { | |
4175 ADVANCE_REGINPUT(); | |
4176 } | |
4177 else | |
4178 { | |
4179 if (WITH_NL(op)) | |
180 | 4180 op -= ADD_NL; |
7 | 4181 #ifdef FEAT_MBYTE |
4182 if (has_mbyte) | |
4183 c = (*mb_ptr2char)(reginput); | |
4184 else | |
4185 #endif | |
4186 c = *reginput; | |
4187 switch (op) | |
4188 { | |
4189 case BOL: | |
4190 if (reginput != regline) | |
180 | 4191 status = RA_NOMATCH; |
7 | 4192 break; |
4193 | |
4194 case EOL: | |
4195 if (c != NUL) | |
180 | 4196 status = RA_NOMATCH; |
7 | 4197 break; |
4198 | |
4199 case RE_BOF: | |
1458 | 4200 /* We're not at the beginning of the file when below the first |
4201 * line where we started, not at the start of the line or we | |
4202 * didn't start at the first line of the buffer. */ | |
7 | 4203 if (reglnum != 0 || reginput != regline |
1458 | 4204 || (REG_MULTI && reg_firstlnum > 1)) |
180 | 4205 status = RA_NOMATCH; |
7 | 4206 break; |
4207 | |
4208 case RE_EOF: | |
4209 if (reglnum != reg_maxline || c != NUL) | |
180 | 4210 status = RA_NOMATCH; |
7 | 4211 break; |
4212 | |
4213 case CURSOR: | |
4214 /* Check if the buffer is in a window and compare the | |
4215 * reg_win->w_cursor position to the match position. */ | |
4216 if (reg_win == NULL | |
4217 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum) | |
4218 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col)) | |
180 | 4219 status = RA_NOMATCH; |
7 | 4220 break; |
4221 | |
639 | 4222 case RE_MARK: |
4223 /* Compare the mark position to the match position. NOTE: Always | |
4224 * uses the current buffer. */ | |
4225 { | |
4226 int mark = OPERAND(scan)[0]; | |
4227 int cmp = OPERAND(scan)[1]; | |
4228 pos_T *pos; | |
4229 | |
4043 | 4230 pos = getmark_buf(reg_buf, mark, FALSE); |
1148 | 4231 if (pos == NULL /* mark doesn't exist */ |
639 | 4232 || pos->lnum <= 0 /* mark isn't set (in curbuf) */ |
4233 || (pos->lnum == reglnum + reg_firstlnum | |
4234 ? (pos->col == (colnr_T)(reginput - regline) | |
4235 ? (cmp == '<' || cmp == '>') | |
4236 : (pos->col < (colnr_T)(reginput - regline) | |
4237 ? cmp != '>' | |
4238 : cmp != '<')) | |
4239 : (pos->lnum < reglnum + reg_firstlnum | |
4240 ? cmp != '>' | |
4241 : cmp != '<'))) | |
4242 status = RA_NOMATCH; | |
4243 } | |
4244 break; | |
4245 | |
4246 case RE_VISUAL: | |
4247 #ifdef FEAT_VISUAL | |
4248 /* Check if the buffer is the current buffer. and whether the | |
4249 * position is inside the Visual area. */ | |
4250 if (reg_buf != curbuf || VIsual.lnum == 0) | |
4251 status = RA_NOMATCH; | |
4252 else | |
4253 { | |
4254 pos_T top, bot; | |
4255 linenr_T lnum; | |
4256 colnr_T col; | |
4257 win_T *wp = reg_win == NULL ? curwin : reg_win; | |
4258 int mode; | |
4259 | |
4260 if (VIsual_active) | |
4261 { | |
4262 if (lt(VIsual, wp->w_cursor)) | |
4263 { | |
4264 top = VIsual; | |
4265 bot = wp->w_cursor; | |
4266 } | |
4267 else | |
4268 { | |
4269 top = wp->w_cursor; | |
4270 bot = VIsual; | |
4271 } | |
4272 mode = VIsual_mode; | |
4273 } | |
4274 else | |
4275 { | |
692 | 4276 if (lt(curbuf->b_visual.vi_start, curbuf->b_visual.vi_end)) |
685 | 4277 { |
692 | 4278 top = curbuf->b_visual.vi_start; |
4279 bot = curbuf->b_visual.vi_end; | |
685 | 4280 } |
4281 else | |
4282 { | |
692 | 4283 top = curbuf->b_visual.vi_end; |
4284 bot = curbuf->b_visual.vi_start; | |
685 | 4285 } |
692 | 4286 mode = curbuf->b_visual.vi_mode; |
639 | 4287 } |
4288 lnum = reglnum + reg_firstlnum; | |
4289 col = (colnr_T)(reginput - regline); | |
4290 if (lnum < top.lnum || lnum > bot.lnum) | |
4291 status = RA_NOMATCH; | |
4292 else if (mode == 'v') | |
4293 { | |
4294 if ((lnum == top.lnum && col < top.col) | |
4295 || (lnum == bot.lnum | |
4296 && col >= bot.col + (*p_sel != 'e'))) | |
4297 status = RA_NOMATCH; | |
4298 } | |
4299 else if (mode == Ctrl_V) | |
4300 { | |
4301 colnr_T start, end; | |
4302 colnr_T start2, end2; | |
944 | 4303 colnr_T cols; |
639 | 4304 |
4305 getvvcol(wp, &top, &start, NULL, &end); | |
4306 getvvcol(wp, &bot, &start2, NULL, &end2); | |
4307 if (start2 < start) | |
4308 start = start2; | |
4309 if (end2 > end) | |
4310 end = end2; | |
4311 if (top.col == MAXCOL || bot.col == MAXCOL) | |
4312 end = MAXCOL; | |
944 | 4313 cols = win_linetabsize(wp, |
639 | 4314 regline, (colnr_T)(reginput - regline)); |
944 | 4315 if (cols < start || cols > end - (*p_sel == 'e')) |
639 | 4316 status = RA_NOMATCH; |
4317 } | |
4318 } | |
4319 #else | |
4320 status = RA_NOMATCH; | |
4321 #endif | |
4322 break; | |
4323 | |
7 | 4324 case RE_LNUM: |
4325 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum), | |
4326 scan)) | |
180 | 4327 status = RA_NOMATCH; |
7 | 4328 break; |
4329 | |
4330 case RE_COL: | |
4331 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan)) | |
180 | 4332 status = RA_NOMATCH; |
7 | 4333 break; |
4334 | |
4335 case RE_VCOL: | |
4336 if (!re_num_cmp((long_u)win_linetabsize( | |
4337 reg_win == NULL ? curwin : reg_win, | |
4338 regline, (colnr_T)(reginput - regline)) + 1, scan)) | |
180 | 4339 status = RA_NOMATCH; |
7 | 4340 break; |
4341 | |
4342 case BOW: /* \<word; reginput points to w */ | |
4343 if (c == NUL) /* Can't match at end of line */ | |
180 | 4344 status = RA_NOMATCH; |
7 | 4345 #ifdef FEAT_MBYTE |
180 | 4346 else if (has_mbyte) |
7 | 4347 { |
4348 int this_class; | |
4349 | |
4350 /* Get class of current and previous char (if it exists). */ | |
4069 | 4351 this_class = mb_get_class_buf(reginput, reg_buf); |
7 | 4352 if (this_class <= 1) |
180 | 4353 status = RA_NOMATCH; /* not on a word at all */ |
4354 else if (reg_prev_class() == this_class) | |
4355 status = RA_NOMATCH; /* previous char is in same word */ | |
7 | 4356 } |
4357 #endif | |
4358 else | |
4359 { | |
4061 | 4360 if (!vim_iswordc_buf(c, reg_buf) || (reginput > regline |
4361 && vim_iswordc_buf(reginput[-1], reg_buf))) | |
180 | 4362 status = RA_NOMATCH; |
7 | 4363 } |
4364 break; | |
4365 | |
4366 case EOW: /* word\>; reginput points after d */ | |
4367 if (reginput == regline) /* Can't match at start of line */ | |
180 | 4368 status = RA_NOMATCH; |
7 | 4369 #ifdef FEAT_MBYTE |
180 | 4370 else if (has_mbyte) |
7 | 4371 { |
4372 int this_class, prev_class; | |
4373 | |
4374 /* Get class of current and previous char (if it exists). */ | |
4069 | 4375 this_class = mb_get_class_buf(reginput, reg_buf); |
7 | 4376 prev_class = reg_prev_class(); |
180 | 4377 if (this_class == prev_class |
4378 || prev_class == 0 || prev_class == 1) | |
4379 status = RA_NOMATCH; | |
7 | 4380 } |
180 | 4381 #endif |
7 | 4382 else |
4383 { | |
4043 | 4384 if (!vim_iswordc_buf(reginput[-1], reg_buf) |
4385 || (reginput[0] != NUL && vim_iswordc_buf(c, reg_buf))) | |
180 | 4386 status = RA_NOMATCH; |
7 | 4387 } |
4388 break; /* Matched with EOW */ | |
4389 | |
4390 case ANY: | |
4084 | 4391 /* ANY does not match new lines. */ |
7 | 4392 if (c == NUL) |
180 | 4393 status = RA_NOMATCH; |
4394 else | |
4395 ADVANCE_REGINPUT(); | |
7 | 4396 break; |
4397 | |
4398 case IDENT: | |
4399 if (!vim_isIDc(c)) | |
180 | 4400 status = RA_NOMATCH; |
4401 else | |
4402 ADVANCE_REGINPUT(); | |
7 | 4403 break; |
4404 | |
4405 case SIDENT: | |
4406 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c)) | |
180 | 4407 status = RA_NOMATCH; |
4408 else | |
4409 ADVANCE_REGINPUT(); | |
7 | 4410 break; |
4411 | |
4412 case KWORD: | |
4069 | 4413 if (!vim_iswordp_buf(reginput, reg_buf)) |
180 | 4414 status = RA_NOMATCH; |
4415 else | |
4416 ADVANCE_REGINPUT(); | |
7 | 4417 break; |
4418 | |
4419 case SKWORD: | |
4069 | 4420 if (VIM_ISDIGIT(*reginput) || !vim_iswordp_buf(reginput, reg_buf)) |
180 | 4421 status = RA_NOMATCH; |
4422 else | |
4423 ADVANCE_REGINPUT(); | |
7 | 4424 break; |
4425 | |
4426 case FNAME: | |
4427 if (!vim_isfilec(c)) | |
180 | 4428 status = RA_NOMATCH; |
4429 else | |
4430 ADVANCE_REGINPUT(); | |
7 | 4431 break; |
4432 | |
4433 case SFNAME: | |
4434 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c)) | |
180 | 4435 status = RA_NOMATCH; |
4436 else | |
4437 ADVANCE_REGINPUT(); | |
7 | 4438 break; |
4439 | |
4440 case PRINT: | |
4441 if (ptr2cells(reginput) != 1) | |
180 | 4442 status = RA_NOMATCH; |
4443 else | |
4444 ADVANCE_REGINPUT(); | |
7 | 4445 break; |
4446 | |
4447 case SPRINT: | |
4448 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1) | |
180 | 4449 status = RA_NOMATCH; |
4450 else | |
4451 ADVANCE_REGINPUT(); | |
7 | 4452 break; |
4453 | |
4454 case WHITE: | |
4455 if (!vim_iswhite(c)) | |
180 | 4456 status = RA_NOMATCH; |
4457 else | |
4458 ADVANCE_REGINPUT(); | |
7 | 4459 break; |
4460 | |
4461 case NWHITE: | |
4462 if (c == NUL || vim_iswhite(c)) | |
180 | 4463 status = RA_NOMATCH; |
4464 else | |
4465 ADVANCE_REGINPUT(); | |
7 | 4466 break; |
4467 | |
4468 case DIGIT: | |
4469 if (!ri_digit(c)) | |
180 | 4470 status = RA_NOMATCH; |
4471 else | |
4472 ADVANCE_REGINPUT(); | |
7 | 4473 break; |
4474 | |
4475 case NDIGIT: | |
4476 if (c == NUL || ri_digit(c)) | |
180 | 4477 status = RA_NOMATCH; |
4478 else | |
4479 ADVANCE_REGINPUT(); | |
7 | 4480 break; |
4481 | |
4482 case HEX: | |
4483 if (!ri_hex(c)) | |
180 | 4484 status = RA_NOMATCH; |
4485 else | |
4486 ADVANCE_REGINPUT(); | |
7 | 4487 break; |
4488 | |
4489 case NHEX: | |
4490 if (c == NUL || ri_hex(c)) | |
180 | 4491 status = RA_NOMATCH; |
4492 else | |
4493 ADVANCE_REGINPUT(); | |
7 | 4494 break; |
4495 | |
4496 case OCTAL: | |
4497 if (!ri_octal(c)) | |
180 | 4498 status = RA_NOMATCH; |
4499 else | |
4500 ADVANCE_REGINPUT(); | |
7 | 4501 break; |
4502 | |
4503 case NOCTAL: | |
4504 if (c == NUL || ri_octal(c)) | |
180 | 4505 status = RA_NOMATCH; |
4506 else | |
4507 ADVANCE_REGINPUT(); | |
7 | 4508 break; |
4509 | |
4510 case WORD: | |
4511 if (!ri_word(c)) | |
180 | 4512 status = RA_NOMATCH; |
4513 else | |
4514 ADVANCE_REGINPUT(); | |
7 | 4515 break; |
4516 | |
4517 case NWORD: | |
4518 if (c == NUL || ri_word(c)) | |
180 | 4519 status = RA_NOMATCH; |
4520 else | |
4521 ADVANCE_REGINPUT(); | |
7 | 4522 break; |
4523 | |
4524 case HEAD: | |
4525 if (!ri_head(c)) | |
180 | 4526 status = RA_NOMATCH; |
4527 else | |
4528 ADVANCE_REGINPUT(); | |
7 | 4529 break; |
4530 | |
4531 case NHEAD: | |
4532 if (c == NUL || ri_head(c)) | |
180 | 4533 status = RA_NOMATCH; |
4534 else | |
4535 ADVANCE_REGINPUT(); | |
7 | 4536 break; |
4537 | |
4538 case ALPHA: | |
4539 if (!ri_alpha(c)) | |
180 | 4540 status = RA_NOMATCH; |
4541 else | |
4542 ADVANCE_REGINPUT(); | |
7 | 4543 break; |
4544 | |
4545 case NALPHA: | |
4546 if (c == NUL || ri_alpha(c)) | |
180 | 4547 status = RA_NOMATCH; |
4548 else | |
4549 ADVANCE_REGINPUT(); | |
7 | 4550 break; |
4551 | |
4552 case LOWER: | |
4553 if (!ri_lower(c)) | |
180 | 4554 status = RA_NOMATCH; |
4555 else | |
4556 ADVANCE_REGINPUT(); | |
7 | 4557 break; |
4558 | |
4559 case NLOWER: | |
4560 if (c == NUL || ri_lower(c)) | |
180 | 4561 status = RA_NOMATCH; |
4562 else | |
4563 ADVANCE_REGINPUT(); | |
7 | 4564 break; |
4565 | |
4566 case UPPER: | |
4567 if (!ri_upper(c)) | |
180 | 4568 status = RA_NOMATCH; |
4569 else | |
4570 ADVANCE_REGINPUT(); | |
7 | 4571 break; |
4572 | |
4573 case NUPPER: | |
4574 if (c == NUL || ri_upper(c)) | |
180 | 4575 status = RA_NOMATCH; |
4576 else | |
4577 ADVANCE_REGINPUT(); | |
7 | 4578 break; |
4579 | |
4580 case EXACTLY: | |
4581 { | |
4582 int len; | |
4583 char_u *opnd; | |
4584 | |
4585 opnd = OPERAND(scan); | |
4586 /* Inline the first byte, for speed. */ | |
4587 if (*opnd != *reginput | |
4588 && (!ireg_ic || ( | |
4589 #ifdef FEAT_MBYTE | |
4590 !enc_utf8 && | |
4591 #endif | |
1347 | 4592 MB_TOLOWER(*opnd) != MB_TOLOWER(*reginput)))) |
180 | 4593 status = RA_NOMATCH; |
4594 else if (*opnd == NUL) | |
7 | 4595 { |
4596 /* match empty string always works; happens when "~" is | |
4597 * empty. */ | |
4598 } | |
4599 else if (opnd[1] == NUL | |
4600 #ifdef FEAT_MBYTE | |
4601 && !(enc_utf8 && ireg_ic) | |
4602 #endif | |
4603 ) | |
4604 ++reginput; /* matched a single char */ | |
4605 else | |
4606 { | |
4607 len = (int)STRLEN(opnd); | |
4608 /* Need to match first byte again for multi-byte. */ | |
4609 if (cstrncmp(opnd, reginput, &len) != 0) | |
180 | 4610 status = RA_NOMATCH; |
7 | 4611 #ifdef FEAT_MBYTE |
4612 /* Check for following composing character. */ | |
180 | 4613 else if (enc_utf8 |
4614 && UTF_COMPOSINGLIKE(reginput, reginput + len)) | |
7 | 4615 { |
4616 /* raaron: This code makes a composing character get | |
4617 * ignored, which is the correct behavior (sometimes) | |
4618 * for voweled Hebrew texts. */ | |
4619 if (!ireg_icombine) | |
180 | 4620 status = RA_NOMATCH; |
7 | 4621 } |
180 | 4622 #endif |
7 | 4623 else |
180 | 4624 reginput += len; |
7 | 4625 } |
4626 } | |
4627 break; | |
4628 | |
4629 case ANYOF: | |
4630 case ANYBUT: | |
4631 if (c == NUL) | |
180 | 4632 status = RA_NOMATCH; |
4633 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF)) | |
4634 status = RA_NOMATCH; | |
4635 else | |
4636 ADVANCE_REGINPUT(); | |
7 | 4637 break; |
4638 | |
4639 #ifdef FEAT_MBYTE | |
4640 case MULTIBYTECODE: | |
4641 if (has_mbyte) | |
4642 { | |
4643 int i, len; | |
4644 char_u *opnd; | |
944 | 4645 int opndc = 0, inpc; |
7 | 4646 |
4647 opnd = OPERAND(scan); | |
4648 /* Safety check (just in case 'encoding' was changed since | |
4649 * compiling the program). */ | |
474 | 4650 if ((len = (*mb_ptr2len)(opnd)) < 2) |
180 | 4651 { |
4652 status = RA_NOMATCH; | |
4653 break; | |
4654 } | |
714 | 4655 if (enc_utf8) |
4656 opndc = mb_ptr2char(opnd); | |
4657 if (enc_utf8 && utf_iscomposing(opndc)) | |
4658 { | |
4659 /* When only a composing char is given match at any | |
4660 * position where that composing char appears. */ | |
4661 status = RA_NOMATCH; | |
4662 for (i = 0; reginput[i] != NUL; i += utf_char2len(inpc)) | |
180 | 4663 { |
714 | 4664 inpc = mb_ptr2char(reginput + i); |
4665 if (!utf_iscomposing(inpc)) | |
4666 { | |
4667 if (i > 0) | |
4668 break; | |
4669 } | |
4670 else if (opndc == inpc) | |
4671 { | |
4672 /* Include all following composing chars. */ | |
4673 len = i + mb_ptr2len(reginput + i); | |
4674 status = RA_MATCH; | |
4675 break; | |
4676 } | |
180 | 4677 } |
714 | 4678 } |
4679 else | |
4680 for (i = 0; i < len; ++i) | |
4681 if (opnd[i] != reginput[i]) | |
4682 { | |
4683 status = RA_NOMATCH; | |
4684 break; | |
4685 } | |
7 | 4686 reginput += len; |
4687 } | |
4688 else | |
180 | 4689 status = RA_NOMATCH; |
7 | 4690 break; |
4691 #endif | |
4692 | |
4693 case NOTHING: | |
4694 break; | |
4695 | |
4696 case BACK: | |
233 | 4697 { |
4698 int i; | |
4699 backpos_T *bp; | |
4700 | |
4701 /* | |
4702 * When we run into BACK we need to check if we don't keep | |
4703 * looping without matching any input. The second and later | |
4704 * times a BACK is encountered it fails if the input is still | |
4705 * at the same position as the previous time. | |
4706 * The positions are stored in "backpos" and found by the | |
4707 * current value of "scan", the position in the RE program. | |
4708 */ | |
4709 bp = (backpos_T *)backpos.ga_data; | |
4710 for (i = 0; i < backpos.ga_len; ++i) | |
4711 if (bp[i].bp_scan == scan) | |
4712 break; | |
4713 if (i == backpos.ga_len) | |
4714 { | |
4715 /* First time at this BACK, make room to store the pos. */ | |
4716 if (ga_grow(&backpos, 1) == FAIL) | |
4717 status = RA_FAIL; | |
4718 else | |
4719 { | |
4720 /* get "ga_data" again, it may have changed */ | |
4721 bp = (backpos_T *)backpos.ga_data; | |
4722 bp[i].bp_scan = scan; | |
4723 ++backpos.ga_len; | |
4724 } | |
4725 } | |
4726 else if (reg_save_equal(&bp[i].bp_pos)) | |
4727 /* Still at same position as last time, fail. */ | |
4728 status = RA_NOMATCH; | |
4729 | |
4730 if (status != RA_FAIL && status != RA_NOMATCH) | |
4731 reg_save(&bp[i].bp_pos, &backpos); | |
4732 } | |
179 | 4733 break; |
4734 | |
7 | 4735 case MOPEN + 0: /* Match start: \zs */ |
4736 case MOPEN + 1: /* \( */ | |
4737 case MOPEN + 2: | |
4738 case MOPEN + 3: | |
4739 case MOPEN + 4: | |
4740 case MOPEN + 5: | |
4741 case MOPEN + 6: | |
4742 case MOPEN + 7: | |
4743 case MOPEN + 8: | |
4744 case MOPEN + 9: | |
4745 { | |
4746 no = op - MOPEN; | |
4747 cleanup_subexpr(); | |
270 | 4748 rp = regstack_push(RS_MOPEN, scan); |
180 | 4749 if (rp == NULL) |
4750 status = RA_FAIL; | |
4751 else | |
4752 { | |
4753 rp->rs_no = no; | |
4754 save_se(&rp->rs_un.sesave, ®_startpos[no], | |
4755 ®_startp[no]); | |
4756 /* We simply continue and handle the result when done. */ | |
4757 } | |
7 | 4758 } |
180 | 4759 break; |
7 | 4760 |
4761 case NOPEN: /* \%( */ | |
4762 case NCLOSE: /* \) after \%( */ | |
270 | 4763 if (regstack_push(RS_NOPEN, scan) == NULL) |
180 | 4764 status = RA_FAIL; |
4765 /* We simply continue and handle the result when done. */ | |
4766 break; | |
7 | 4767 |
4768 #ifdef FEAT_SYN_HL | |
4769 case ZOPEN + 1: | |
4770 case ZOPEN + 2: | |
4771 case ZOPEN + 3: | |
4772 case ZOPEN + 4: | |
4773 case ZOPEN + 5: | |
4774 case ZOPEN + 6: | |
4775 case ZOPEN + 7: | |
4776 case ZOPEN + 8: | |
4777 case ZOPEN + 9: | |
4778 { | |
4779 no = op - ZOPEN; | |
4780 cleanup_zsubexpr(); | |
270 | 4781 rp = regstack_push(RS_ZOPEN, scan); |
180 | 4782 if (rp == NULL) |
4783 status = RA_FAIL; | |
4784 else | |
4785 { | |
4786 rp->rs_no = no; | |
4787 save_se(&rp->rs_un.sesave, ®_startzpos[no], | |
4788 ®_startzp[no]); | |
4789 /* We simply continue and handle the result when done. */ | |
4790 } | |
7 | 4791 } |
180 | 4792 break; |
7 | 4793 #endif |
4794 | |
4795 case MCLOSE + 0: /* Match end: \ze */ | |
4796 case MCLOSE + 1: /* \) */ | |
4797 case MCLOSE + 2: | |
4798 case MCLOSE + 3: | |
4799 case MCLOSE + 4: | |
4800 case MCLOSE + 5: | |
4801 case MCLOSE + 6: | |
4802 case MCLOSE + 7: | |
4803 case MCLOSE + 8: | |
4804 case MCLOSE + 9: | |
4805 { | |
4806 no = op - MCLOSE; | |
4807 cleanup_subexpr(); | |
270 | 4808 rp = regstack_push(RS_MCLOSE, scan); |
180 | 4809 if (rp == NULL) |
4810 status = RA_FAIL; | |
4811 else | |
4812 { | |
4813 rp->rs_no = no; | |
4814 save_se(&rp->rs_un.sesave, ®_endpos[no], ®_endp[no]); | |
4815 /* We simply continue and handle the result when done. */ | |
4816 } | |
7 | 4817 } |
180 | 4818 break; |
7 | 4819 |
4820 #ifdef FEAT_SYN_HL | |
4821 case ZCLOSE + 1: /* \) after \z( */ | |
4822 case ZCLOSE + 2: | |
4823 case ZCLOSE + 3: | |
4824 case ZCLOSE + 4: | |
4825 case ZCLOSE + 5: | |
4826 case ZCLOSE + 6: | |
4827 case ZCLOSE + 7: | |
4828 case ZCLOSE + 8: | |
4829 case ZCLOSE + 9: | |
4830 { | |
4831 no = op - ZCLOSE; | |
4832 cleanup_zsubexpr(); | |
270 | 4833 rp = regstack_push(RS_ZCLOSE, scan); |
180 | 4834 if (rp == NULL) |
4835 status = RA_FAIL; | |
4836 else | |
4837 { | |
4838 rp->rs_no = no; | |
4839 save_se(&rp->rs_un.sesave, ®_endzpos[no], | |
4840 ®_endzp[no]); | |
4841 /* We simply continue and handle the result when done. */ | |
4842 } | |
7 | 4843 } |
180 | 4844 break; |
7 | 4845 #endif |
4846 | |
4847 case BACKREF + 1: | |
4848 case BACKREF + 2: | |
4849 case BACKREF + 3: | |
4850 case BACKREF + 4: | |
4851 case BACKREF + 5: | |
4852 case BACKREF + 6: | |
4853 case BACKREF + 7: | |
4854 case BACKREF + 8: | |
4855 case BACKREF + 9: | |
4856 { | |
4857 int len; | |
4858 linenr_T clnum; | |
4859 colnr_T ccol; | |
4860 char_u *p; | |
4861 | |
4862 no = op - BACKREF; | |
4863 cleanup_subexpr(); | |
4864 if (!REG_MULTI) /* Single-line regexp */ | |
4865 { | |
1815 | 4866 if (reg_startp[no] == NULL || reg_endp[no] == NULL) |
7 | 4867 { |
4868 /* Backref was not set: Match an empty string. */ | |
4869 len = 0; | |
4870 } | |
4871 else | |
4872 { | |
4873 /* Compare current input with back-ref in the same | |
4874 * line. */ | |
4875 len = (int)(reg_endp[no] - reg_startp[no]); | |
4876 if (cstrncmp(reg_startp[no], reginput, &len) != 0) | |
180 | 4877 status = RA_NOMATCH; |
7 | 4878 } |
4879 } | |
4880 else /* Multi-line regexp */ | |
4881 { | |
1815 | 4882 if (reg_startpos[no].lnum < 0 || reg_endpos[no].lnum < 0) |
7 | 4883 { |
4884 /* Backref was not set: Match an empty string. */ | |
4885 len = 0; | |
4886 } | |
4887 else | |
4888 { | |
4889 if (reg_startpos[no].lnum == reglnum | |
4890 && reg_endpos[no].lnum == reglnum) | |
4891 { | |
4892 /* Compare back-ref within the current line. */ | |
4893 len = reg_endpos[no].col - reg_startpos[no].col; | |
4894 if (cstrncmp(regline + reg_startpos[no].col, | |
4895 reginput, &len) != 0) | |
180 | 4896 status = RA_NOMATCH; |
7 | 4897 } |
4898 else | |
4899 { | |
4900 /* Messy situation: Need to compare between two | |
4901 * lines. */ | |
4902 ccol = reg_startpos[no].col; | |
4903 clnum = reg_startpos[no].lnum; | |
4904 for (;;) | |
4905 { | |
4906 /* Since getting one line may invalidate | |
4907 * the other, need to make copy. Slow! */ | |
4908 if (regline != reg_tofree) | |
4909 { | |
4910 len = (int)STRLEN(regline); | |
4911 if (reg_tofree == NULL | |
4912 || len >= (int)reg_tofreelen) | |
4913 { | |
4914 len += 50; /* get some extra */ | |
4915 vim_free(reg_tofree); | |
4916 reg_tofree = alloc(len); | |
4917 if (reg_tofree == NULL) | |
180 | 4918 { |
4919 status = RA_FAIL; /* outof memory!*/ | |
4920 break; | |
4921 } | |
7 | 4922 reg_tofreelen = len; |
4923 } | |
4924 STRCPY(reg_tofree, regline); | |
4925 reginput = reg_tofree | |
4926 + (reginput - regline); | |
4927 regline = reg_tofree; | |
4928 } | |
4929 | |
4930 /* Get the line to compare with. */ | |
4931 p = reg_getline(clnum); | |
4932 if (clnum == reg_endpos[no].lnum) | |
4933 len = reg_endpos[no].col - ccol; | |
4934 else | |
4935 len = (int)STRLEN(p + ccol); | |
4936 | |
4937 if (cstrncmp(p + ccol, reginput, &len) != 0) | |
180 | 4938 { |
4939 status = RA_NOMATCH; /* doesn't match */ | |
4940 break; | |
4941 } | |
7 | 4942 if (clnum == reg_endpos[no].lnum) |
4943 break; /* match and at end! */ | |
481 | 4944 if (reglnum >= reg_maxline) |
180 | 4945 { |
4946 status = RA_NOMATCH; /* text too short */ | |
4947 break; | |
4948 } | |
7 | 4949 |
4950 /* Advance to next line. */ | |
4951 reg_nextline(); | |
4952 ++clnum; | |
4953 ccol = 0; | |
180 | 4954 if (got_int) |
4955 { | |
4956 status = RA_FAIL; | |
4957 break; | |
4958 } | |
7 | 4959 } |
4960 | |
4961 /* found a match! Note that regline may now point | |
4962 * to a copy of the line, that should not matter. */ | |
4963 } | |
4964 } | |
4965 } | |
4966 | |
4967 /* Matched the backref, skip over it. */ | |
4968 reginput += len; | |
4969 } | |
4970 break; | |
4971 | |
4972 #ifdef FEAT_SYN_HL | |
4973 case ZREF + 1: | |
4974 case ZREF + 2: | |
4975 case ZREF + 3: | |
4976 case ZREF + 4: | |
4977 case ZREF + 5: | |
4978 case ZREF + 6: | |
4979 case ZREF + 7: | |
4980 case ZREF + 8: | |
4981 case ZREF + 9: | |
4982 { | |
4983 int len; | |
4984 | |
4985 cleanup_zsubexpr(); | |
4986 no = op - ZREF; | |
4987 if (re_extmatch_in != NULL | |
4988 && re_extmatch_in->matches[no] != NULL) | |
4989 { | |
4990 len = (int)STRLEN(re_extmatch_in->matches[no]); | |
4991 if (cstrncmp(re_extmatch_in->matches[no], | |
4992 reginput, &len) != 0) | |
180 | 4993 status = RA_NOMATCH; |
4994 else | |
4995 reginput += len; | |
7 | 4996 } |
4997 else | |
4998 { | |
4999 /* Backref was not set: Match an empty string. */ | |
5000 } | |
5001 } | |
5002 break; | |
5003 #endif | |
5004 | |
5005 case BRANCH: | |
5006 { | |
5007 if (OP(next) != BRANCH) /* No choice. */ | |
5008 next = OPERAND(scan); /* Avoid recursion. */ | |
5009 else | |
5010 { | |
270 | 5011 rp = regstack_push(RS_BRANCH, scan); |
180 | 5012 if (rp == NULL) |
5013 status = RA_FAIL; | |
5014 else | |
5015 status = RA_BREAK; /* rest is below */ | |
7 | 5016 } |
5017 } | |
5018 break; | |
5019 | |
5020 case BRACE_LIMITS: | |
5021 { | |
5022 if (OP(next) == BRACE_SIMPLE) | |
5023 { | |
5024 bl_minval = OPERAND_MIN(scan); | |
5025 bl_maxval = OPERAND_MAX(scan); | |
5026 } | |
5027 else if (OP(next) >= BRACE_COMPLEX | |
5028 && OP(next) < BRACE_COMPLEX + 10) | |
5029 { | |
5030 no = OP(next) - BRACE_COMPLEX; | |
5031 brace_min[no] = OPERAND_MIN(scan); | |
5032 brace_max[no] = OPERAND_MAX(scan); | |
5033 brace_count[no] = 0; | |
5034 } | |
5035 else | |
5036 { | |
5037 EMSG(_(e_internal)); /* Shouldn't happen */ | |
180 | 5038 status = RA_FAIL; |
7 | 5039 } |
5040 } | |
5041 break; | |
5042 | |
5043 case BRACE_COMPLEX + 0: | |
5044 case BRACE_COMPLEX + 1: | |
5045 case BRACE_COMPLEX + 2: | |
5046 case BRACE_COMPLEX + 3: | |
5047 case BRACE_COMPLEX + 4: | |
5048 case BRACE_COMPLEX + 5: | |
5049 case BRACE_COMPLEX + 6: | |
5050 case BRACE_COMPLEX + 7: | |
5051 case BRACE_COMPLEX + 8: | |
5052 case BRACE_COMPLEX + 9: | |
5053 { | |
5054 no = op - BRACE_COMPLEX; | |
5055 ++brace_count[no]; | |
5056 | |
5057 /* If not matched enough times yet, try one more */ | |
5058 if (brace_count[no] <= (brace_min[no] <= brace_max[no] | |
180 | 5059 ? brace_min[no] : brace_max[no])) |
7 | 5060 { |
270 | 5061 rp = regstack_push(RS_BRCPLX_MORE, scan); |
180 | 5062 if (rp == NULL) |
5063 status = RA_FAIL; | |
5064 else | |
5065 { | |
5066 rp->rs_no = no; | |
233 | 5067 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5068 next = OPERAND(scan); |
5069 /* We continue and handle the result when done. */ | |
5070 } | |
5071 break; | |
7 | 5072 } |
5073 | |
5074 /* If matched enough times, may try matching some more */ | |
5075 if (brace_min[no] <= brace_max[no]) | |
5076 { | |
5077 /* Range is the normal way around, use longest match */ | |
5078 if (brace_count[no] <= brace_max[no]) | |
5079 { | |
270 | 5080 rp = regstack_push(RS_BRCPLX_LONG, scan); |
180 | 5081 if (rp == NULL) |
5082 status = RA_FAIL; | |
5083 else | |
5084 { | |
5085 rp->rs_no = no; | |
233 | 5086 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5087 next = OPERAND(scan); |
5088 /* We continue and handle the result when done. */ | |
5089 } | |
7 | 5090 } |
5091 } | |
5092 else | |
5093 { | |
5094 /* Range is backwards, use shortest match first */ | |
5095 if (brace_count[no] <= brace_min[no]) | |
5096 { | |
270 | 5097 rp = regstack_push(RS_BRCPLX_SHORT, scan); |
180 | 5098 if (rp == NULL) |
5099 status = RA_FAIL; | |
5100 else | |
5101 { | |
233 | 5102 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5103 /* We continue and handle the result when done. */ |
5104 } | |
7 | 5105 } |
5106 } | |
5107 } | |
5108 break; | |
5109 | |
5110 case BRACE_SIMPLE: | |
5111 case STAR: | |
5112 case PLUS: | |
5113 { | |
180 | 5114 regstar_T rst; |
7 | 5115 |
5116 /* | |
5117 * Lookahead to avoid useless match attempts when we know | |
5118 * what character comes next. | |
5119 */ | |
5120 if (OP(next) == EXACTLY) | |
5121 { | |
180 | 5122 rst.nextb = *OPERAND(next); |
7 | 5123 if (ireg_ic) |
5124 { | |
1347 | 5125 if (MB_ISUPPER(rst.nextb)) |
5126 rst.nextb_ic = MB_TOLOWER(rst.nextb); | |
7 | 5127 else |
1347 | 5128 rst.nextb_ic = MB_TOUPPER(rst.nextb); |
7 | 5129 } |
5130 else | |
180 | 5131 rst.nextb_ic = rst.nextb; |
7 | 5132 } |
5133 else | |
5134 { | |
180 | 5135 rst.nextb = NUL; |
5136 rst.nextb_ic = NUL; | |
7 | 5137 } |
5138 if (op != BRACE_SIMPLE) | |
5139 { | |
180 | 5140 rst.minval = (op == STAR) ? 0 : 1; |
5141 rst.maxval = MAX_LIMIT; | |
7 | 5142 } |
5143 else | |
5144 { | |
180 | 5145 rst.minval = bl_minval; |
5146 rst.maxval = bl_maxval; | |
7 | 5147 } |
5148 | |
5149 /* | |
5150 * When maxval > minval, try matching as much as possible, up | |
5151 * to maxval. When maxval < minval, try matching at least the | |
5152 * minimal number (since the range is backwards, that's also | |
5153 * maxval!). | |
5154 */ | |
180 | 5155 rst.count = regrepeat(OPERAND(scan), rst.maxval); |
7 | 5156 if (got_int) |
180 | 5157 { |
5158 status = RA_FAIL; | |
5159 break; | |
5160 } | |
5161 if (rst.minval <= rst.maxval | |
5162 ? rst.count >= rst.minval : rst.count >= rst.maxval) | |
7 | 5163 { |
180 | 5164 /* It could match. Prepare for trying to match what |
5165 * follows. The code is below. Parameters are stored in | |
5166 * a regstar_T on the regstack. */ | |
212 | 5167 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5168 { |
5169 EMSG(_(e_maxmempat)); | |
5170 status = RA_FAIL; | |
5171 } | |
5172 else if (ga_grow(®stack, sizeof(regstar_T)) == FAIL) | |
180 | 5173 status = RA_FAIL; |
5174 else | |
7 | 5175 { |
180 | 5176 regstack.ga_len += sizeof(regstar_T); |
270 | 5177 rp = regstack_push(rst.minval <= rst.maxval |
233 | 5178 ? RS_STAR_LONG : RS_STAR_SHORT, scan); |
180 | 5179 if (rp == NULL) |
5180 status = RA_FAIL; | |
5181 else | |
7 | 5182 { |
180 | 5183 *(((regstar_T *)rp) - 1) = rst; |
5184 status = RA_BREAK; /* skip the restore bits */ | |
7 | 5185 } |
5186 } | |
5187 } | |
5188 else | |
180 | 5189 status = RA_NOMATCH; |
5190 | |
7 | 5191 } |
180 | 5192 break; |
7 | 5193 |
5194 case NOMATCH: | |
5195 case MATCH: | |
5196 case SUBPAT: | |
270 | 5197 rp = regstack_push(RS_NOMATCH, scan); |
180 | 5198 if (rp == NULL) |
5199 status = RA_FAIL; | |
5200 else | |
7 | 5201 { |
180 | 5202 rp->rs_no = op; |
233 | 5203 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5204 next = OPERAND(scan); |
5205 /* We continue and handle the result when done. */ | |
7 | 5206 } |
5207 break; | |
5208 | |
5209 case BEHIND: | |
5210 case NOBEHIND: | |
180 | 5211 /* Need a bit of room to store extra positions. */ |
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(regbehind_T)) == FAIL) | |
180 | 5218 status = RA_FAIL; |
5219 else | |
7 | 5220 { |
180 | 5221 regstack.ga_len += sizeof(regbehind_T); |
270 | 5222 rp = regstack_push(RS_BEHIND1, scan); |
180 | 5223 if (rp == NULL) |
5224 status = RA_FAIL; | |
5225 else | |
7 | 5226 { |
1579 | 5227 /* Need to save the subexpr to be able to restore them |
5228 * when there is a match but we don't use it. */ | |
5229 save_subexpr(((regbehind_T *)rp) - 1); | |
5230 | |
180 | 5231 rp->rs_no = op; |
233 | 5232 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5233 /* First try if what follows matches. If it does then we |
5234 * check the behind match by looping. */ | |
7 | 5235 } |
5236 } | |
180 | 5237 break; |
7 | 5238 |
5239 case BHPOS: | |
5240 if (REG_MULTI) | |
5241 { | |
5242 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline) | |
5243 || behind_pos.rs_u.pos.lnum != reglnum) | |
180 | 5244 status = RA_NOMATCH; |
7 | 5245 } |
5246 else if (behind_pos.rs_u.ptr != reginput) | |
180 | 5247 status = RA_NOMATCH; |
7 | 5248 break; |
5249 | |
5250 case NEWL: | |
1018 | 5251 if ((c != NUL || !REG_MULTI || reglnum > reg_maxline |
5252 || reg_line_lbr) && (c != '\n' || !reg_line_lbr)) | |
180 | 5253 status = RA_NOMATCH; |
5254 else if (reg_line_lbr) | |
7 | 5255 ADVANCE_REGINPUT(); |
5256 else | |
5257 reg_nextline(); | |
5258 break; | |
5259 | |
5260 case END: | |
180 | 5261 status = RA_MATCH; /* Success! */ |
5262 break; | |
7 | 5263 |
5264 default: | |
5265 EMSG(_(e_re_corr)); | |
5266 #ifdef DEBUG | |
5267 printf("Illegal op code %d\n", op); | |
5268 #endif | |
180 | 5269 status = RA_FAIL; |
5270 break; | |
7 | 5271 } |
5272 } | |
5273 | |
180 | 5274 /* If we can't continue sequentially, break the inner loop. */ |
5275 if (status != RA_CONT) | |
5276 break; | |
5277 | |
5278 /* Continue in inner loop, advance to next item. */ | |
7 | 5279 scan = next; |
180 | 5280 |
5281 } /* end of inner loop */ | |
7 | 5282 |
5283 /* | |
180 | 5284 * If there is something on the regstack execute the code for the state. |
233 | 5285 * If the state is popped then loop and use the older state. |
7 | 5286 */ |
180 | 5287 while (regstack.ga_len > 0 && status != RA_FAIL) |
5288 { | |
5289 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1; | |
5290 switch (rp->rs_state) | |
5291 { | |
5292 case RS_NOPEN: | |
5293 /* Result is passed on as-is, simply pop the state. */ | |
270 | 5294 regstack_pop(&scan); |
180 | 5295 break; |
5296 | |
5297 case RS_MOPEN: | |
5298 /* Pop the state. Restore pointers when there is no match. */ | |
5299 if (status == RA_NOMATCH) | |
5300 restore_se(&rp->rs_un.sesave, ®_startpos[rp->rs_no], | |
5301 ®_startp[rp->rs_no]); | |
270 | 5302 regstack_pop(&scan); |
180 | 5303 break; |
5304 | |
5305 #ifdef FEAT_SYN_HL | |
5306 case RS_ZOPEN: | |
5307 /* Pop the state. Restore pointers when there is no match. */ | |
5308 if (status == RA_NOMATCH) | |
5309 restore_se(&rp->rs_un.sesave, ®_startzpos[rp->rs_no], | |
5310 ®_startzp[rp->rs_no]); | |
270 | 5311 regstack_pop(&scan); |
180 | 5312 break; |
5313 #endif | |
5314 | |
5315 case RS_MCLOSE: | |
5316 /* Pop the state. Restore pointers when there is no match. */ | |
5317 if (status == RA_NOMATCH) | |
5318 restore_se(&rp->rs_un.sesave, ®_endpos[rp->rs_no], | |
5319 ®_endp[rp->rs_no]); | |
270 | 5320 regstack_pop(&scan); |
180 | 5321 break; |
5322 | |
5323 #ifdef FEAT_SYN_HL | |
5324 case RS_ZCLOSE: | |
5325 /* Pop the state. Restore pointers when there is no match. */ | |
5326 if (status == RA_NOMATCH) | |
5327 restore_se(&rp->rs_un.sesave, ®_endzpos[rp->rs_no], | |
5328 ®_endzp[rp->rs_no]); | |
270 | 5329 regstack_pop(&scan); |
180 | 5330 break; |
5331 #endif | |
5332 | |
5333 case RS_BRANCH: | |
5334 if (status == RA_MATCH) | |
5335 /* this branch matched, use it */ | |
270 | 5336 regstack_pop(&scan); |
180 | 5337 else |
5338 { | |
5339 if (status != RA_BREAK) | |
5340 { | |
5341 /* After a non-matching branch: try next one. */ | |
233 | 5342 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5343 scan = rp->rs_scan; |
5344 } | |
5345 if (scan == NULL || OP(scan) != BRANCH) | |
5346 { | |
5347 /* no more branches, didn't find a match */ | |
5348 status = RA_NOMATCH; | |
270 | 5349 regstack_pop(&scan); |
180 | 5350 } |
5351 else | |
5352 { | |
5353 /* Prepare to try a branch. */ | |
5354 rp->rs_scan = regnext(scan); | |
233 | 5355 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5356 scan = OPERAND(scan); |
5357 } | |
5358 } | |
5359 break; | |
5360 | |
5361 case RS_BRCPLX_MORE: | |
5362 /* Pop the state. Restore pointers when there is no match. */ | |
5363 if (status == RA_NOMATCH) | |
5364 { | |
233 | 5365 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5366 --brace_count[rp->rs_no]; /* decrement match count */ |
5367 } | |
270 | 5368 regstack_pop(&scan); |
180 | 5369 break; |
5370 | |
5371 case RS_BRCPLX_LONG: | |
5372 /* Pop the state. Restore pointers when there is no match. */ | |
5373 if (status == RA_NOMATCH) | |
5374 { | |
5375 /* There was no match, but we did find enough matches. */ | |
233 | 5376 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5377 --brace_count[rp->rs_no]; |
5378 /* continue with the items after "\{}" */ | |
5379 status = RA_CONT; | |
5380 } | |
270 | 5381 regstack_pop(&scan); |
180 | 5382 if (status == RA_CONT) |
5383 scan = regnext(scan); | |
5384 break; | |
5385 | |
5386 case RS_BRCPLX_SHORT: | |
5387 /* Pop the state. Restore pointers when there is no match. */ | |
5388 if (status == RA_NOMATCH) | |
5389 /* There was no match, try to match one more item. */ | |
233 | 5390 reg_restore(&rp->rs_un.regsave, &backpos); |
270 | 5391 regstack_pop(&scan); |
180 | 5392 if (status == RA_NOMATCH) |
5393 { | |
5394 scan = OPERAND(scan); | |
5395 status = RA_CONT; | |
5396 } | |
5397 break; | |
5398 | |
5399 case RS_NOMATCH: | |
5400 /* Pop the state. If the operand matches for NOMATCH or | |
5401 * doesn't match for MATCH/SUBPAT, we fail. Otherwise backup, | |
5402 * except for SUBPAT, and continue with the next item. */ | |
5403 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH)) | |
5404 status = RA_NOMATCH; | |
5405 else | |
5406 { | |
5407 status = RA_CONT; | |
233 | 5408 if (rp->rs_no != SUBPAT) /* zero-width */ |
5409 reg_restore(&rp->rs_un.regsave, &backpos); | |
180 | 5410 } |
270 | 5411 regstack_pop(&scan); |
180 | 5412 if (status == RA_CONT) |
5413 scan = regnext(scan); | |
5414 break; | |
5415 | |
5416 case RS_BEHIND1: | |
5417 if (status == RA_NOMATCH) | |
5418 { | |
270 | 5419 regstack_pop(&scan); |
180 | 5420 regstack.ga_len -= sizeof(regbehind_T); |
5421 } | |
5422 else | |
5423 { | |
5424 /* The stuff after BEHIND/NOBEHIND matches. Now try if | |
5425 * the behind part does (not) match before the current | |
5426 * position in the input. This must be done at every | |
5427 * position in the input and checking if the match ends at | |
5428 * the current position. */ | |
5429 | |
5430 /* save the position after the found match for next */ | |
233 | 5431 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos); |
180 | 5432 |
5433 /* start looking for a match with operand at the current | |
1209 | 5434 * position. Go back one character until we find the |
180 | 5435 * result, hitting the start of the line or the previous |
5436 * line (for multi-line matching). | |
5437 * Set behind_pos to where the match should end, BHPOS | |
5438 * will match it. Save the current value. */ | |
5439 (((regbehind_T *)rp) - 1)->save_behind = behind_pos; | |
5440 behind_pos = rp->rs_un.regsave; | |
5441 | |
5442 rp->rs_state = RS_BEHIND2; | |
5443 | |
233 | 5444 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5445 scan = OPERAND(rp->rs_scan); |
5446 } | |
5447 break; | |
5448 | |
5449 case RS_BEHIND2: | |
5450 /* | |
5451 * Looping for BEHIND / NOBEHIND match. | |
5452 */ | |
5453 if (status == RA_MATCH && reg_save_equal(&behind_pos)) | |
5454 { | |
5455 /* found a match that ends where "next" started */ | |
5456 behind_pos = (((regbehind_T *)rp) - 1)->save_behind; | |
5457 if (rp->rs_no == BEHIND) | |
233 | 5458 reg_restore(&(((regbehind_T *)rp) - 1)->save_after, |
5459 &backpos); | |
180 | 5460 else |
1579 | 5461 { |
5462 /* But we didn't want a match. Need to restore the | |
5463 * subexpr, because what follows matched, so they have | |
5464 * been set. */ | |
180 | 5465 status = RA_NOMATCH; |
1579 | 5466 restore_subexpr(((regbehind_T *)rp) - 1); |
5467 } | |
270 | 5468 regstack_pop(&scan); |
180 | 5469 regstack.ga_len -= sizeof(regbehind_T); |
5470 } | |
5471 else | |
5472 { | |
1579 | 5473 /* No match or a match that doesn't end where we want it: Go |
5474 * back one character. May go to previous line once. */ | |
180 | 5475 no = OK; |
5476 if (REG_MULTI) | |
5477 { | |
5478 if (rp->rs_un.regsave.rs_u.pos.col == 0) | |
5479 { | |
5480 if (rp->rs_un.regsave.rs_u.pos.lnum | |
5481 < behind_pos.rs_u.pos.lnum | |
5482 || reg_getline( | |
5483 --rp->rs_un.regsave.rs_u.pos.lnum) | |
5484 == NULL) | |
5485 no = FAIL; | |
5486 else | |
5487 { | |
233 | 5488 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5489 rp->rs_un.regsave.rs_u.pos.col = |
5490 (colnr_T)STRLEN(regline); | |
5491 } | |
5492 } | |
5493 else | |
4176 | 5494 #ifdef FEAT_MBYTE |
5495 if (has_mbyte) | |
5496 rp->rs_un.regsave.rs_u.pos.col -= | |
5497 (*mb_head_off)(regline, regline | |
5498 + rp->rs_un.regsave.rs_u.pos.col - 1) + 1; | |
5499 else | |
5500 #endif | |
180 | 5501 --rp->rs_un.regsave.rs_u.pos.col; |
5502 } | |
5503 else | |
5504 { | |
5505 if (rp->rs_un.regsave.rs_u.ptr == regline) | |
5506 no = FAIL; | |
5507 else | |
5508 --rp->rs_un.regsave.rs_u.ptr; | |
5509 } | |
5510 if (no == OK) | |
5511 { | |
5512 /* Advanced, prepare for finding match again. */ | |
233 | 5513 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5514 scan = OPERAND(rp->rs_scan); |
1579 | 5515 if (status == RA_MATCH) |
5516 { | |
5517 /* We did match, so subexpr may have been changed, | |
5518 * need to restore them for the next try. */ | |
5519 status = RA_NOMATCH; | |
5520 restore_subexpr(((regbehind_T *)rp) - 1); | |
5521 } | |
180 | 5522 } |
5523 else | |
5524 { | |
5525 /* Can't advance. For NOBEHIND that's a match. */ | |
5526 behind_pos = (((regbehind_T *)rp) - 1)->save_behind; | |
5527 if (rp->rs_no == NOBEHIND) | |
5528 { | |
233 | 5529 reg_restore(&(((regbehind_T *)rp) - 1)->save_after, |
5530 &backpos); | |
180 | 5531 status = RA_MATCH; |
5532 } | |
5533 else | |
1579 | 5534 { |
5535 /* We do want a proper match. Need to restore the | |
5536 * subexpr if we had a match, because they may have | |
5537 * been set. */ | |
5538 if (status == RA_MATCH) | |
5539 { | |
5540 status = RA_NOMATCH; | |
5541 restore_subexpr(((regbehind_T *)rp) - 1); | |
5542 } | |
5543 } | |
270 | 5544 regstack_pop(&scan); |
180 | 5545 regstack.ga_len -= sizeof(regbehind_T); |
5546 } | |
5547 } | |
5548 break; | |
5549 | |
5550 case RS_STAR_LONG: | |
5551 case RS_STAR_SHORT: | |
5552 { | |
5553 regstar_T *rst = ((regstar_T *)rp) - 1; | |
5554 | |
5555 if (status == RA_MATCH) | |
5556 { | |
270 | 5557 regstack_pop(&scan); |
180 | 5558 regstack.ga_len -= sizeof(regstar_T); |
5559 break; | |
5560 } | |
5561 | |
5562 /* Tried once already, restore input pointers. */ | |
5563 if (status != RA_BREAK) | |
233 | 5564 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5565 |
5566 /* Repeat until we found a position where it could match. */ | |
5567 for (;;) | |
5568 { | |
5569 if (status != RA_BREAK) | |
5570 { | |
5571 /* Tried first position already, advance. */ | |
5572 if (rp->rs_state == RS_STAR_LONG) | |
5573 { | |
685 | 5574 /* Trying for longest match, but couldn't or |
5575 * didn't match -- back up one char. */ | |
180 | 5576 if (--rst->count < rst->minval) |
5577 break; | |
5578 if (reginput == regline) | |
5579 { | |
5580 /* backup to last char of previous line */ | |
5581 --reglnum; | |
5582 regline = reg_getline(reglnum); | |
5583 /* Just in case regrepeat() didn't count | |
5584 * right. */ | |
5585 if (regline == NULL) | |
5586 break; | |
5587 reginput = regline + STRLEN(regline); | |
5588 fast_breakcheck(); | |
5589 } | |
5590 else | |
5591 mb_ptr_back(regline, reginput); | |
5592 } | |
5593 else | |
5594 { | |
5595 /* Range is backwards, use shortest match first. | |
5596 * Careful: maxval and minval are exchanged! | |
5597 * Couldn't or didn't match: try advancing one | |
5598 * char. */ | |
5599 if (rst->count == rst->minval | |
5600 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0) | |
5601 break; | |
5602 ++rst->count; | |
5603 } | |
5604 if (got_int) | |
5605 break; | |
5606 } | |
5607 else | |
5608 status = RA_NOMATCH; | |
5609 | |
5610 /* If it could match, try it. */ | |
5611 if (rst->nextb == NUL || *reginput == rst->nextb | |
5612 || *reginput == rst->nextb_ic) | |
5613 { | |
233 | 5614 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5615 scan = regnext(rp->rs_scan); |
5616 status = RA_CONT; | |
5617 break; | |
5618 } | |
5619 } | |
5620 if (status != RA_CONT) | |
5621 { | |
5622 /* Failed. */ | |
270 | 5623 regstack_pop(&scan); |
180 | 5624 regstack.ga_len -= sizeof(regstar_T); |
5625 status = RA_NOMATCH; | |
5626 } | |
5627 } | |
5628 break; | |
5629 } | |
5630 | |
685 | 5631 /* If we want to continue the inner loop or didn't pop a state |
5632 * continue matching loop */ | |
180 | 5633 if (status == RA_CONT || rp == (regitem_T *) |
5634 ((char *)regstack.ga_data + regstack.ga_len) - 1) | |
5635 break; | |
5636 } | |
5637 | |
189 | 5638 /* May need to continue with the inner loop, starting at "scan". */ |
180 | 5639 if (status == RA_CONT) |
5640 continue; | |
5641 | |
5642 /* | |
5643 * If the regstack is empty or something failed we are done. | |
5644 */ | |
5645 if (regstack.ga_len == 0 || status == RA_FAIL) | |
5646 { | |
5647 if (scan == NULL) | |
5648 { | |
5649 /* | |
5650 * We get here only if there's trouble -- normally "case END" is | |
5651 * the terminating point. | |
5652 */ | |
5653 EMSG(_(e_re_corr)); | |
7 | 5654 #ifdef DEBUG |
180 | 5655 printf("Premature EOL\n"); |
7 | 5656 #endif |
180 | 5657 } |
189 | 5658 if (status == RA_FAIL) |
5659 got_int = TRUE; | |
180 | 5660 return (status == RA_MATCH); |
5661 } | |
5662 | |
5663 } /* End of loop until the regstack is empty. */ | |
5664 | |
5665 /* NOTREACHED */ | |
5666 } | |
5667 | |
5668 /* | |
5669 * Push an item onto the regstack. | |
5670 * Returns pointer to new item. Returns NULL when out of memory. | |
5671 */ | |
5672 static regitem_T * | |
270 | 5673 regstack_push(state, scan) |
180 | 5674 regstate_T state; |
5675 char_u *scan; | |
5676 { | |
5677 regitem_T *rp; | |
5678 | |
270 | 5679 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5680 { |
5681 EMSG(_(e_maxmempat)); | |
5682 return NULL; | |
5683 } | |
270 | 5684 if (ga_grow(®stack, sizeof(regitem_T)) == FAIL) |
180 | 5685 return NULL; |
5686 | |
270 | 5687 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len); |
180 | 5688 rp->rs_state = state; |
5689 rp->rs_scan = scan; | |
5690 | |
270 | 5691 regstack.ga_len += sizeof(regitem_T); |
180 | 5692 return rp; |
5693 } | |
5694 | |
5695 /* | |
5696 * Pop an item from the regstack. | |
5697 */ | |
5698 static void | |
270 | 5699 regstack_pop(scan) |
180 | 5700 char_u **scan; |
5701 { | |
5702 regitem_T *rp; | |
5703 | |
270 | 5704 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1; |
180 | 5705 *scan = rp->rs_scan; |
5706 | |
270 | 5707 regstack.ga_len -= sizeof(regitem_T); |
7 | 5708 } |
5709 | |
5710 /* | |
5711 * regrepeat - repeatedly match something simple, return how many. | |
5712 * Advances reginput (and reglnum) to just after the matched chars. | |
5713 */ | |
5714 static int | |
5715 regrepeat(p, maxcount) | |
5716 char_u *p; | |
5717 long maxcount; /* maximum number of matches allowed */ | |
5718 { | |
5719 long count = 0; | |
5720 char_u *scan; | |
5721 char_u *opnd; | |
5722 int mask; | |
5723 int testval = 0; | |
5724 | |
5725 scan = reginput; /* Make local copy of reginput for speed. */ | |
5726 opnd = OPERAND(p); | |
5727 switch (OP(p)) | |
5728 { | |
5729 case ANY: | |
5730 case ANY + ADD_NL: | |
5731 while (count < maxcount) | |
5732 { | |
5733 /* Matching anything means we continue until end-of-line (or | |
5734 * end-of-file for ANY + ADD_NL), only limited by maxcount. */ | |
5735 while (*scan != NUL && count < maxcount) | |
5736 { | |
5737 ++count; | |
39 | 5738 mb_ptr_adv(scan); |
7 | 5739 } |
1018 | 5740 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5741 || reg_line_lbr || count == maxcount) | |
7 | 5742 break; |
5743 ++count; /* count the line-break */ | |
5744 reg_nextline(); | |
5745 scan = reginput; | |
5746 if (got_int) | |
5747 break; | |
5748 } | |
5749 break; | |
5750 | |
5751 case IDENT: | |
5752 case IDENT + ADD_NL: | |
5753 testval = TRUE; | |
5754 /*FALLTHROUGH*/ | |
5755 case SIDENT: | |
5756 case SIDENT + ADD_NL: | |
5757 while (count < maxcount) | |
5758 { | |
5759 if (vim_isIDc(*scan) && (testval || !VIM_ISDIGIT(*scan))) | |
5760 { | |
39 | 5761 mb_ptr_adv(scan); |
7 | 5762 } |
5763 else if (*scan == NUL) | |
5764 { | |
1018 | 5765 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5766 || reg_line_lbr) | |
7 | 5767 break; |
5768 reg_nextline(); | |
5769 scan = reginput; | |
5770 if (got_int) | |
5771 break; | |
5772 } | |
5773 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5774 ++scan; | |
5775 else | |
5776 break; | |
5777 ++count; | |
5778 } | |
5779 break; | |
5780 | |
5781 case KWORD: | |
5782 case KWORD + ADD_NL: | |
5783 testval = TRUE; | |
5784 /*FALLTHROUGH*/ | |
5785 case SKWORD: | |
5786 case SKWORD + ADD_NL: | |
5787 while (count < maxcount) | |
5788 { | |
4069 | 5789 if (vim_iswordp_buf(scan, reg_buf) |
5790 && (testval || !VIM_ISDIGIT(*scan))) | |
7 | 5791 { |
39 | 5792 mb_ptr_adv(scan); |
7 | 5793 } |
5794 else if (*scan == NUL) | |
5795 { | |
1018 | 5796 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5797 || reg_line_lbr) | |
7 | 5798 break; |
5799 reg_nextline(); | |
5800 scan = reginput; | |
5801 if (got_int) | |
5802 break; | |
5803 } | |
5804 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5805 ++scan; | |
5806 else | |
5807 break; | |
5808 ++count; | |
5809 } | |
5810 break; | |
5811 | |
5812 case FNAME: | |
5813 case FNAME + ADD_NL: | |
5814 testval = TRUE; | |
5815 /*FALLTHROUGH*/ | |
5816 case SFNAME: | |
5817 case SFNAME + ADD_NL: | |
5818 while (count < maxcount) | |
5819 { | |
5820 if (vim_isfilec(*scan) && (testval || !VIM_ISDIGIT(*scan))) | |
5821 { | |
39 | 5822 mb_ptr_adv(scan); |
7 | 5823 } |
5824 else if (*scan == NUL) | |
5825 { | |
1018 | 5826 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5827 || reg_line_lbr) | |
7 | 5828 break; |
5829 reg_nextline(); | |
5830 scan = reginput; | |
5831 if (got_int) | |
5832 break; | |
5833 } | |
5834 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5835 ++scan; | |
5836 else | |
5837 break; | |
5838 ++count; | |
5839 } | |
5840 break; | |
5841 | |
5842 case PRINT: | |
5843 case PRINT + ADD_NL: | |
5844 testval = TRUE; | |
5845 /*FALLTHROUGH*/ | |
5846 case SPRINT: | |
5847 case SPRINT + ADD_NL: | |
5848 while (count < maxcount) | |
5849 { | |
5850 if (*scan == NUL) | |
5851 { | |
1018 | 5852 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5853 || reg_line_lbr) | |
7 | 5854 break; |
5855 reg_nextline(); | |
5856 scan = reginput; | |
5857 if (got_int) | |
5858 break; | |
5859 } | |
5860 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan))) | |
5861 { | |
39 | 5862 mb_ptr_adv(scan); |
7 | 5863 } |
5864 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5865 ++scan; | |
5866 else | |
5867 break; | |
5868 ++count; | |
5869 } | |
5870 break; | |
5871 | |
5872 case WHITE: | |
5873 case WHITE + ADD_NL: | |
5874 testval = mask = RI_WHITE; | |
5875 do_class: | |
5876 while (count < maxcount) | |
5877 { | |
5878 #ifdef FEAT_MBYTE | |
5879 int l; | |
5880 #endif | |
5881 if (*scan == NUL) | |
5882 { | |
1018 | 5883 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5884 || reg_line_lbr) | |
7 | 5885 break; |
5886 reg_nextline(); | |
5887 scan = reginput; | |
5888 if (got_int) | |
5889 break; | |
5890 } | |
5891 #ifdef FEAT_MBYTE | |
474 | 5892 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1) |
7 | 5893 { |
5894 if (testval != 0) | |
5895 break; | |
5896 scan += l; | |
5897 } | |
5898 #endif | |
5899 else if ((class_tab[*scan] & mask) == testval) | |
5900 ++scan; | |
5901 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5902 ++scan; | |
5903 else | |
5904 break; | |
5905 ++count; | |
5906 } | |
5907 break; | |
5908 | |
5909 case NWHITE: | |
5910 case NWHITE + ADD_NL: | |
5911 mask = RI_WHITE; | |
5912 goto do_class; | |
5913 case DIGIT: | |
5914 case DIGIT + ADD_NL: | |
5915 testval = mask = RI_DIGIT; | |
5916 goto do_class; | |
5917 case NDIGIT: | |
5918 case NDIGIT + ADD_NL: | |
5919 mask = RI_DIGIT; | |
5920 goto do_class; | |
5921 case HEX: | |
5922 case HEX + ADD_NL: | |
5923 testval = mask = RI_HEX; | |
5924 goto do_class; | |
5925 case NHEX: | |
5926 case NHEX + ADD_NL: | |
5927 mask = RI_HEX; | |
5928 goto do_class; | |
5929 case OCTAL: | |
5930 case OCTAL + ADD_NL: | |
5931 testval = mask = RI_OCTAL; | |
5932 goto do_class; | |
5933 case NOCTAL: | |
5934 case NOCTAL + ADD_NL: | |
5935 mask = RI_OCTAL; | |
5936 goto do_class; | |
5937 case WORD: | |
5938 case WORD + ADD_NL: | |
5939 testval = mask = RI_WORD; | |
5940 goto do_class; | |
5941 case NWORD: | |
5942 case NWORD + ADD_NL: | |
5943 mask = RI_WORD; | |
5944 goto do_class; | |
5945 case HEAD: | |
5946 case HEAD + ADD_NL: | |
5947 testval = mask = RI_HEAD; | |
5948 goto do_class; | |
5949 case NHEAD: | |
5950 case NHEAD + ADD_NL: | |
5951 mask = RI_HEAD; | |
5952 goto do_class; | |
5953 case ALPHA: | |
5954 case ALPHA + ADD_NL: | |
5955 testval = mask = RI_ALPHA; | |
5956 goto do_class; | |
5957 case NALPHA: | |
5958 case NALPHA + ADD_NL: | |
5959 mask = RI_ALPHA; | |
5960 goto do_class; | |
5961 case LOWER: | |
5962 case LOWER + ADD_NL: | |
5963 testval = mask = RI_LOWER; | |
5964 goto do_class; | |
5965 case NLOWER: | |
5966 case NLOWER + ADD_NL: | |
5967 mask = RI_LOWER; | |
5968 goto do_class; | |
5969 case UPPER: | |
5970 case UPPER + ADD_NL: | |
5971 testval = mask = RI_UPPER; | |
5972 goto do_class; | |
5973 case NUPPER: | |
5974 case NUPPER + ADD_NL: | |
5975 mask = RI_UPPER; | |
5976 goto do_class; | |
5977 | |
5978 case EXACTLY: | |
5979 { | |
5980 int cu, cl; | |
5981 | |
5982 /* This doesn't do a multi-byte character, because a MULTIBYTECODE | |
1347 | 5983 * would have been used for it. It does handle single-byte |
5984 * characters, such as latin1. */ | |
7 | 5985 if (ireg_ic) |
5986 { | |
1347 | 5987 cu = MB_TOUPPER(*opnd); |
5988 cl = MB_TOLOWER(*opnd); | |
7 | 5989 while (count < maxcount && (*scan == cu || *scan == cl)) |
5990 { | |
5991 count++; | |
5992 scan++; | |
5993 } | |
5994 } | |
5995 else | |
5996 { | |
5997 cu = *opnd; | |
5998 while (count < maxcount && *scan == cu) | |
5999 { | |
6000 count++; | |
6001 scan++; | |
6002 } | |
6003 } | |
6004 break; | |
6005 } | |
6006 | |
6007 #ifdef FEAT_MBYTE | |
6008 case MULTIBYTECODE: | |
6009 { | |
6010 int i, len, cf = 0; | |
6011 | |
6012 /* Safety check (just in case 'encoding' was changed since | |
6013 * compiling the program). */ | |
474 | 6014 if ((len = (*mb_ptr2len)(opnd)) > 1) |
7 | 6015 { |
6016 if (ireg_ic && enc_utf8) | |
6017 cf = utf_fold(utf_ptr2char(opnd)); | |
6018 while (count < maxcount) | |
6019 { | |
6020 for (i = 0; i < len; ++i) | |
6021 if (opnd[i] != scan[i]) | |
6022 break; | |
6023 if (i < len && (!ireg_ic || !enc_utf8 | |
6024 || utf_fold(utf_ptr2char(scan)) != cf)) | |
6025 break; | |
6026 scan += len; | |
6027 ++count; | |
6028 } | |
6029 } | |
6030 } | |
6031 break; | |
6032 #endif | |
6033 | |
6034 case ANYOF: | |
6035 case ANYOF + ADD_NL: | |
6036 testval = TRUE; | |
6037 /*FALLTHROUGH*/ | |
6038 | |
6039 case ANYBUT: | |
6040 case ANYBUT + ADD_NL: | |
6041 while (count < maxcount) | |
6042 { | |
6043 #ifdef FEAT_MBYTE | |
6044 int len; | |
6045 #endif | |
6046 if (*scan == NUL) | |
6047 { | |
1018 | 6048 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
6049 || reg_line_lbr) | |
7 | 6050 break; |
6051 reg_nextline(); | |
6052 scan = reginput; | |
6053 if (got_int) | |
6054 break; | |
6055 } | |
6056 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
6057 ++scan; | |
6058 #ifdef FEAT_MBYTE | |
474 | 6059 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1) |
7 | 6060 { |
6061 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval) | |
6062 break; | |
6063 scan += len; | |
6064 } | |
6065 #endif | |
6066 else | |
6067 { | |
6068 if ((cstrchr(opnd, *scan) == NULL) == testval) | |
6069 break; | |
6070 ++scan; | |
6071 } | |
6072 ++count; | |
6073 } | |
6074 break; | |
6075 | |
6076 case NEWL: | |
6077 while (count < maxcount | |
1018 | 6078 && ((*scan == NUL && reglnum <= reg_maxline && !reg_line_lbr |
6079 && REG_MULTI) || (*scan == '\n' && reg_line_lbr))) | |
7 | 6080 { |
6081 count++; | |
6082 if (reg_line_lbr) | |
6083 ADVANCE_REGINPUT(); | |
6084 else | |
6085 reg_nextline(); | |
6086 scan = reginput; | |
6087 if (got_int) | |
6088 break; | |
6089 } | |
6090 break; | |
6091 | |
6092 default: /* Oh dear. Called inappropriately. */ | |
6093 EMSG(_(e_re_corr)); | |
6094 #ifdef DEBUG | |
6095 printf("Called regrepeat with op code %d\n", OP(p)); | |
6096 #endif | |
6097 break; | |
6098 } | |
6099 | |
6100 reginput = scan; | |
6101 | |
6102 return (int)count; | |
6103 } | |
6104 | |
6105 /* | |
6106 * regnext - dig the "next" pointer out of a node | |
2010 | 6107 * Returns NULL when calculating size, when there is no next item and when |
6108 * there is an error. | |
7 | 6109 */ |
6110 static char_u * | |
6111 regnext(p) | |
6112 char_u *p; | |
6113 { | |
6114 int offset; | |
6115 | |
2010 | 6116 if (p == JUST_CALC_SIZE || reg_toolong) |
7 | 6117 return NULL; |
6118 | |
6119 offset = NEXT(p); | |
6120 if (offset == 0) | |
6121 return NULL; | |
6122 | |
233 | 6123 if (OP(p) == BACK) |
7 | 6124 return p - offset; |
6125 else | |
6126 return p + offset; | |
6127 } | |
6128 | |
6129 /* | |
6130 * Check the regexp program for its magic number. | |
6131 * Return TRUE if it's wrong. | |
6132 */ | |
6133 static int | |
6134 prog_magic_wrong() | |
6135 { | |
4444 | 6136 regprog_T *prog; |
6137 | |
6138 prog = REG_MULTI ? reg_mmatch->regprog : reg_match->regprog; | |
6139 if (prog->engine == &nfa_regengine) | |
6140 /* For NFA matcher we don't check the magic */ | |
6141 return FALSE; | |
6142 | |
6143 if (UCHARAT(((bt_regprog_T *)prog)->program) != REGMAGIC) | |
7 | 6144 { |
6145 EMSG(_(e_re_corr)); | |
6146 return TRUE; | |
6147 } | |
6148 return FALSE; | |
6149 } | |
6150 | |
6151 /* | |
6152 * Cleanup the subexpressions, if this wasn't done yet. | |
6153 * This construction is used to clear the subexpressions only when they are | |
6154 * used (to increase speed). | |
6155 */ | |
6156 static void | |
6157 cleanup_subexpr() | |
6158 { | |
6159 if (need_clear_subexpr) | |
6160 { | |
6161 if (REG_MULTI) | |
6162 { | |
6163 /* Use 0xff to set lnum to -1 */ | |
6164 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6165 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6166 } | |
6167 else | |
6168 { | |
6169 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP); | |
6170 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP); | |
6171 } | |
6172 need_clear_subexpr = FALSE; | |
6173 } | |
6174 } | |
6175 | |
6176 #ifdef FEAT_SYN_HL | |
6177 static void | |
6178 cleanup_zsubexpr() | |
6179 { | |
6180 if (need_clear_zsubexpr) | |
6181 { | |
6182 if (REG_MULTI) | |
6183 { | |
6184 /* Use 0xff to set lnum to -1 */ | |
6185 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6186 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6187 } | |
6188 else | |
6189 { | |
6190 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP); | |
6191 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP); | |
6192 } | |
6193 need_clear_zsubexpr = FALSE; | |
6194 } | |
6195 } | |
6196 #endif | |
6197 | |
6198 /* | |
1579 | 6199 * Save the current subexpr to "bp", so that they can be restored |
6200 * later by restore_subexpr(). | |
6201 */ | |
6202 static void | |
6203 save_subexpr(bp) | |
6204 regbehind_T *bp; | |
6205 { | |
6206 int i; | |
6207 | |
1602 | 6208 /* When "need_clear_subexpr" is set we don't need to save the values, only |
6209 * remember that this flag needs to be set again when restoring. */ | |
6210 bp->save_need_clear_subexpr = need_clear_subexpr; | |
6211 if (!need_clear_subexpr) | |
1579 | 6212 { |
1602 | 6213 for (i = 0; i < NSUBEXP; ++i) |
1579 | 6214 { |
1602 | 6215 if (REG_MULTI) |
6216 { | |
6217 bp->save_start[i].se_u.pos = reg_startpos[i]; | |
6218 bp->save_end[i].se_u.pos = reg_endpos[i]; | |
6219 } | |
6220 else | |
6221 { | |
6222 bp->save_start[i].se_u.ptr = reg_startp[i]; | |
6223 bp->save_end[i].se_u.ptr = reg_endp[i]; | |
6224 } | |
1579 | 6225 } |
6226 } | |
6227 } | |
6228 | |
6229 /* | |
6230 * Restore the subexpr from "bp". | |
6231 */ | |
6232 static void | |
6233 restore_subexpr(bp) | |
6234 regbehind_T *bp; | |
6235 { | |
6236 int i; | |
6237 | |
1602 | 6238 /* Only need to restore saved values when they are not to be cleared. */ |
6239 need_clear_subexpr = bp->save_need_clear_subexpr; | |
6240 if (!need_clear_subexpr) | |
1579 | 6241 { |
1602 | 6242 for (i = 0; i < NSUBEXP; ++i) |
1579 | 6243 { |
1602 | 6244 if (REG_MULTI) |
6245 { | |
6246 reg_startpos[i] = bp->save_start[i].se_u.pos; | |
6247 reg_endpos[i] = bp->save_end[i].se_u.pos; | |
6248 } | |
6249 else | |
6250 { | |
6251 reg_startp[i] = bp->save_start[i].se_u.ptr; | |
6252 reg_endp[i] = bp->save_end[i].se_u.ptr; | |
6253 } | |
1579 | 6254 } |
6255 } | |
6256 } | |
6257 | |
6258 /* | |
7 | 6259 * Advance reglnum, regline and reginput to the next line. |
6260 */ | |
6261 static void | |
6262 reg_nextline() | |
6263 { | |
6264 regline = reg_getline(++reglnum); | |
6265 reginput = regline; | |
6266 fast_breakcheck(); | |
6267 } | |
6268 | |
6269 /* | |
6270 * Save the input line and position in a regsave_T. | |
6271 */ | |
6272 static void | |
233 | 6273 reg_save(save, gap) |
7 | 6274 regsave_T *save; |
233 | 6275 garray_T *gap; |
7 | 6276 { |
6277 if (REG_MULTI) | |
6278 { | |
6279 save->rs_u.pos.col = (colnr_T)(reginput - regline); | |
6280 save->rs_u.pos.lnum = reglnum; | |
6281 } | |
6282 else | |
6283 save->rs_u.ptr = reginput; | |
233 | 6284 save->rs_len = gap->ga_len; |
7 | 6285 } |
6286 | |
6287 /* | |
6288 * Restore the input line and position from a regsave_T. | |
6289 */ | |
6290 static void | |
233 | 6291 reg_restore(save, gap) |
7 | 6292 regsave_T *save; |
233 | 6293 garray_T *gap; |
7 | 6294 { |
6295 if (REG_MULTI) | |
6296 { | |
6297 if (reglnum != save->rs_u.pos.lnum) | |
6298 { | |
6299 /* only call reg_getline() when the line number changed to save | |
6300 * a bit of time */ | |
6301 reglnum = save->rs_u.pos.lnum; | |
6302 regline = reg_getline(reglnum); | |
6303 } | |
6304 reginput = regline + save->rs_u.pos.col; | |
6305 } | |
6306 else | |
6307 reginput = save->rs_u.ptr; | |
233 | 6308 gap->ga_len = save->rs_len; |
7 | 6309 } |
6310 | |
6311 /* | |
6312 * Return TRUE if current position is equal to saved position. | |
6313 */ | |
6314 static int | |
6315 reg_save_equal(save) | |
6316 regsave_T *save; | |
6317 { | |
6318 if (REG_MULTI) | |
6319 return reglnum == save->rs_u.pos.lnum | |
6320 && reginput == regline + save->rs_u.pos.col; | |
6321 return reginput == save->rs_u.ptr; | |
6322 } | |
6323 | |
6324 /* | |
6325 * Tentatively set the sub-expression start to the current position (after | |
6326 * calling regmatch() they will have changed). Need to save the existing | |
6327 * values for when there is no match. | |
6328 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()), | |
6329 * depending on REG_MULTI. | |
6330 */ | |
6331 static void | |
6332 save_se_multi(savep, posp) | |
6333 save_se_T *savep; | |
6334 lpos_T *posp; | |
6335 { | |
6336 savep->se_u.pos = *posp; | |
6337 posp->lnum = reglnum; | |
6338 posp->col = (colnr_T)(reginput - regline); | |
6339 } | |
6340 | |
6341 static void | |
6342 save_se_one(savep, pp) | |
6343 save_se_T *savep; | |
6344 char_u **pp; | |
6345 { | |
6346 savep->se_u.ptr = *pp; | |
6347 *pp = reginput; | |
6348 } | |
6349 | |
6350 /* | |
6351 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL. | |
6352 */ | |
6353 static int | |
6354 re_num_cmp(val, scan) | |
6355 long_u val; | |
6356 char_u *scan; | |
6357 { | |
6358 long_u n = OPERAND_MIN(scan); | |
6359 | |
6360 if (OPERAND_CMP(scan) == '>') | |
6361 return val > n; | |
6362 if (OPERAND_CMP(scan) == '<') | |
6363 return val < n; | |
6364 return val == n; | |
6365 } | |
6366 | |
6367 | |
4444 | 6368 #ifdef BT_REGEXP_DUMP |
7 | 6369 |
6370 /* | |
6371 * regdump - dump a regexp onto stdout in vaguely comprehensible form | |
6372 */ | |
6373 static void | |
6374 regdump(pattern, r) | |
6375 char_u *pattern; | |
4444 | 6376 bt_regprog_T *r; |
7 | 6377 { |
6378 char_u *s; | |
6379 int op = EXACTLY; /* Arbitrary non-END op. */ | |
6380 char_u *next; | |
6381 char_u *end = NULL; | |
4444 | 6382 FILE *f; |
6383 | |
6384 #ifdef BT_REGEXP_LOG | |
6385 f = fopen("bt_regexp_log.log", "a"); | |
6386 #else | |
6387 f = stdout; | |
6388 #endif | |
6389 if (f == NULL) | |
6390 return; | |
6391 fprintf(f, "-------------------------------------\n\r\nregcomp(%s):\r\n", pattern); | |
7 | 6392 |
6393 s = r->program + 1; | |
6394 /* | |
6395 * Loop until we find the END that isn't before a referred next (an END | |
6396 * can also appear in a NOMATCH operand). | |
6397 */ | |
6398 while (op != END || s <= end) | |
6399 { | |
6400 op = OP(s); | |
4444 | 6401 fprintf(f, "%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */ |
7 | 6402 next = regnext(s); |
6403 if (next == NULL) /* Next ptr. */ | |
4444 | 6404 fprintf(f, "(0)"); |
7 | 6405 else |
4444 | 6406 fprintf(f, "(%d)", (int)((s - r->program) + (next - s))); |
7 | 6407 if (end < next) |
6408 end = next; | |
6409 if (op == BRACE_LIMITS) | |
6410 { | |
6411 /* Two short ints */ | |
4444 | 6412 fprintf(f, " minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s)); |
7 | 6413 s += 8; |
6414 } | |
6415 s += 3; | |
6416 if (op == ANYOF || op == ANYOF + ADD_NL | |
6417 || op == ANYBUT || op == ANYBUT + ADD_NL | |
6418 || op == EXACTLY) | |
6419 { | |
6420 /* Literal string, where present. */ | |
4444 | 6421 fprintf(f, "\nxxxxxxxxx\n"); |
7 | 6422 while (*s != NUL) |
4444 | 6423 fprintf(f, "%c", *s++); |
6424 fprintf(f, "\nxxxxxxxxx\n"); | |
7 | 6425 s++; |
6426 } | |
4444 | 6427 fprintf(f, "\r\n"); |
7 | 6428 } |
6429 | |
6430 /* Header fields of interest. */ | |
6431 if (r->regstart != NUL) | |
4444 | 6432 fprintf(f, "start `%s' 0x%x; ", r->regstart < 256 |
7 | 6433 ? (char *)transchar(r->regstart) |
6434 : "multibyte", r->regstart); | |
6435 if (r->reganch) | |
4444 | 6436 fprintf(f, "anchored; "); |
7 | 6437 if (r->regmust != NULL) |
4444 | 6438 fprintf(f, "must have \"%s\"", r->regmust); |
6439 fprintf(f, "\r\n"); | |
6440 | |
6441 #ifdef BT_REGEXP_LOG | |
6442 fclose(f); | |
6443 #endif | |
7 | 6444 } |
4444 | 6445 #endif /* BT_REGEXP_DUMP */ |
6446 | |
6447 #ifdef DEBUG | |
7 | 6448 /* |
6449 * regprop - printable representation of opcode | |
6450 */ | |
6451 static char_u * | |
6452 regprop(op) | |
6453 char_u *op; | |
6454 { | |
4444 | 6455 char *p; |
6456 static char buf[50]; | |
6457 | |
6458 STRCPY(buf, ":"); | |
6459 | |
6460 switch ((int) OP(op)) | |
7 | 6461 { |
6462 case BOL: | |
6463 p = "BOL"; | |
6464 break; | |
6465 case EOL: | |
6466 p = "EOL"; | |
6467 break; | |
6468 case RE_BOF: | |
6469 p = "BOF"; | |
6470 break; | |
6471 case RE_EOF: | |
6472 p = "EOF"; | |
6473 break; | |
6474 case CURSOR: | |
6475 p = "CURSOR"; | |
6476 break; | |
639 | 6477 case RE_VISUAL: |
6478 p = "RE_VISUAL"; | |
6479 break; | |
7 | 6480 case RE_LNUM: |
6481 p = "RE_LNUM"; | |
6482 break; | |
639 | 6483 case RE_MARK: |
6484 p = "RE_MARK"; | |
6485 break; | |
7 | 6486 case RE_COL: |
6487 p = "RE_COL"; | |
6488 break; | |
6489 case RE_VCOL: | |
6490 p = "RE_VCOL"; | |
6491 break; | |
6492 case BOW: | |
6493 p = "BOW"; | |
6494 break; | |
6495 case EOW: | |
6496 p = "EOW"; | |
6497 break; | |
6498 case ANY: | |
6499 p = "ANY"; | |
6500 break; | |
6501 case ANY + ADD_NL: | |
6502 p = "ANY+NL"; | |
6503 break; | |
6504 case ANYOF: | |
6505 p = "ANYOF"; | |
6506 break; | |
6507 case ANYOF + ADD_NL: | |
6508 p = "ANYOF+NL"; | |
6509 break; | |
6510 case ANYBUT: | |
6511 p = "ANYBUT"; | |
6512 break; | |
6513 case ANYBUT + ADD_NL: | |
6514 p = "ANYBUT+NL"; | |
6515 break; | |
6516 case IDENT: | |
6517 p = "IDENT"; | |
6518 break; | |
6519 case IDENT + ADD_NL: | |
6520 p = "IDENT+NL"; | |
6521 break; | |
6522 case SIDENT: | |
6523 p = "SIDENT"; | |
6524 break; | |
6525 case SIDENT + ADD_NL: | |
6526 p = "SIDENT+NL"; | |
6527 break; | |
6528 case KWORD: | |
6529 p = "KWORD"; | |
6530 break; | |
6531 case KWORD + ADD_NL: | |
6532 p = "KWORD+NL"; | |
6533 break; | |
6534 case SKWORD: | |
6535 p = "SKWORD"; | |
6536 break; | |
6537 case SKWORD + ADD_NL: | |
6538 p = "SKWORD+NL"; | |
6539 break; | |
6540 case FNAME: | |
6541 p = "FNAME"; | |
6542 break; | |
6543 case FNAME + ADD_NL: | |
6544 p = "FNAME+NL"; | |
6545 break; | |
6546 case SFNAME: | |
6547 p = "SFNAME"; | |
6548 break; | |
6549 case SFNAME + ADD_NL: | |
6550 p = "SFNAME+NL"; | |
6551 break; | |
6552 case PRINT: | |
6553 p = "PRINT"; | |
6554 break; | |
6555 case PRINT + ADD_NL: | |
6556 p = "PRINT+NL"; | |
6557 break; | |
6558 case SPRINT: | |
6559 p = "SPRINT"; | |
6560 break; | |
6561 case SPRINT + ADD_NL: | |
6562 p = "SPRINT+NL"; | |
6563 break; | |
6564 case WHITE: | |
6565 p = "WHITE"; | |
6566 break; | |
6567 case WHITE + ADD_NL: | |
6568 p = "WHITE+NL"; | |
6569 break; | |
6570 case NWHITE: | |
6571 p = "NWHITE"; | |
6572 break; | |
6573 case NWHITE + ADD_NL: | |
6574 p = "NWHITE+NL"; | |
6575 break; | |
6576 case DIGIT: | |
6577 p = "DIGIT"; | |
6578 break; | |
6579 case DIGIT + ADD_NL: | |
6580 p = "DIGIT+NL"; | |
6581 break; | |
6582 case NDIGIT: | |
6583 p = "NDIGIT"; | |
6584 break; | |
6585 case NDIGIT + ADD_NL: | |
6586 p = "NDIGIT+NL"; | |
6587 break; | |
6588 case HEX: | |
6589 p = "HEX"; | |
6590 break; | |
6591 case HEX + ADD_NL: | |
6592 p = "HEX+NL"; | |
6593 break; | |
6594 case NHEX: | |
6595 p = "NHEX"; | |
6596 break; | |
6597 case NHEX + ADD_NL: | |
6598 p = "NHEX+NL"; | |
6599 break; | |
6600 case OCTAL: | |
6601 p = "OCTAL"; | |
6602 break; | |
6603 case OCTAL + ADD_NL: | |
6604 p = "OCTAL+NL"; | |
6605 break; | |
6606 case NOCTAL: | |
6607 p = "NOCTAL"; | |
6608 break; | |
6609 case NOCTAL + ADD_NL: | |
6610 p = "NOCTAL+NL"; | |
6611 break; | |
6612 case WORD: | |
6613 p = "WORD"; | |
6614 break; | |
6615 case WORD + ADD_NL: | |
6616 p = "WORD+NL"; | |
6617 break; | |
6618 case NWORD: | |
6619 p = "NWORD"; | |
6620 break; | |
6621 case NWORD + ADD_NL: | |
6622 p = "NWORD+NL"; | |
6623 break; | |
6624 case HEAD: | |
6625 p = "HEAD"; | |
6626 break; | |
6627 case HEAD + ADD_NL: | |
6628 p = "HEAD+NL"; | |
6629 break; | |
6630 case NHEAD: | |
6631 p = "NHEAD"; | |
6632 break; | |
6633 case NHEAD + ADD_NL: | |
6634 p = "NHEAD+NL"; | |
6635 break; | |
6636 case ALPHA: | |
6637 p = "ALPHA"; | |
6638 break; | |
6639 case ALPHA + ADD_NL: | |
6640 p = "ALPHA+NL"; | |
6641 break; | |
6642 case NALPHA: | |
6643 p = "NALPHA"; | |
6644 break; | |
6645 case NALPHA + ADD_NL: | |
6646 p = "NALPHA+NL"; | |
6647 break; | |
6648 case LOWER: | |
6649 p = "LOWER"; | |
6650 break; | |
6651 case LOWER + ADD_NL: | |
6652 p = "LOWER+NL"; | |
6653 break; | |
6654 case NLOWER: | |
6655 p = "NLOWER"; | |
6656 break; | |
6657 case NLOWER + ADD_NL: | |
6658 p = "NLOWER+NL"; | |
6659 break; | |
6660 case UPPER: | |
6661 p = "UPPER"; | |
6662 break; | |
6663 case UPPER + ADD_NL: | |
6664 p = "UPPER+NL"; | |
6665 break; | |
6666 case NUPPER: | |
6667 p = "NUPPER"; | |
6668 break; | |
6669 case NUPPER + ADD_NL: | |
6670 p = "NUPPER+NL"; | |
6671 break; | |
6672 case BRANCH: | |
6673 p = "BRANCH"; | |
6674 break; | |
6675 case EXACTLY: | |
6676 p = "EXACTLY"; | |
6677 break; | |
6678 case NOTHING: | |
6679 p = "NOTHING"; | |
6680 break; | |
6681 case BACK: | |
6682 p = "BACK"; | |
6683 break; | |
6684 case END: | |
6685 p = "END"; | |
6686 break; | |
6687 case MOPEN + 0: | |
6688 p = "MATCH START"; | |
6689 break; | |
6690 case MOPEN + 1: | |
6691 case MOPEN + 2: | |
6692 case MOPEN + 3: | |
6693 case MOPEN + 4: | |
6694 case MOPEN + 5: | |
6695 case MOPEN + 6: | |
6696 case MOPEN + 7: | |
6697 case MOPEN + 8: | |
6698 case MOPEN + 9: | |
6699 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN); | |
6700 p = NULL; | |
6701 break; | |
6702 case MCLOSE + 0: | |
6703 p = "MATCH END"; | |
6704 break; | |
6705 case MCLOSE + 1: | |
6706 case MCLOSE + 2: | |
6707 case MCLOSE + 3: | |
6708 case MCLOSE + 4: | |
6709 case MCLOSE + 5: | |
6710 case MCLOSE + 6: | |
6711 case MCLOSE + 7: | |
6712 case MCLOSE + 8: | |
6713 case MCLOSE + 9: | |
6714 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE); | |
6715 p = NULL; | |
6716 break; | |
6717 case BACKREF + 1: | |
6718 case BACKREF + 2: | |
6719 case BACKREF + 3: | |
6720 case BACKREF + 4: | |
6721 case BACKREF + 5: | |
6722 case BACKREF + 6: | |
6723 case BACKREF + 7: | |
6724 case BACKREF + 8: | |
6725 case BACKREF + 9: | |
6726 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF); | |
6727 p = NULL; | |
6728 break; | |
6729 case NOPEN: | |
6730 p = "NOPEN"; | |
6731 break; | |
6732 case NCLOSE: | |
6733 p = "NCLOSE"; | |
6734 break; | |
6735 #ifdef FEAT_SYN_HL | |
6736 case ZOPEN + 1: | |
6737 case ZOPEN + 2: | |
6738 case ZOPEN + 3: | |
6739 case ZOPEN + 4: | |
6740 case ZOPEN + 5: | |
6741 case ZOPEN + 6: | |
6742 case ZOPEN + 7: | |
6743 case ZOPEN + 8: | |
6744 case ZOPEN + 9: | |
6745 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN); | |
6746 p = NULL; | |
6747 break; | |
6748 case ZCLOSE + 1: | |
6749 case ZCLOSE + 2: | |
6750 case ZCLOSE + 3: | |
6751 case ZCLOSE + 4: | |
6752 case ZCLOSE + 5: | |
6753 case ZCLOSE + 6: | |
6754 case ZCLOSE + 7: | |
6755 case ZCLOSE + 8: | |
6756 case ZCLOSE + 9: | |
6757 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE); | |
6758 p = NULL; | |
6759 break; | |
6760 case ZREF + 1: | |
6761 case ZREF + 2: | |
6762 case ZREF + 3: | |
6763 case ZREF + 4: | |
6764 case ZREF + 5: | |
6765 case ZREF + 6: | |
6766 case ZREF + 7: | |
6767 case ZREF + 8: | |
6768 case ZREF + 9: | |
6769 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF); | |
6770 p = NULL; | |
6771 break; | |
6772 #endif | |
6773 case STAR: | |
6774 p = "STAR"; | |
6775 break; | |
6776 case PLUS: | |
6777 p = "PLUS"; | |
6778 break; | |
6779 case NOMATCH: | |
6780 p = "NOMATCH"; | |
6781 break; | |
6782 case MATCH: | |
6783 p = "MATCH"; | |
6784 break; | |
6785 case BEHIND: | |
6786 p = "BEHIND"; | |
6787 break; | |
6788 case NOBEHIND: | |
6789 p = "NOBEHIND"; | |
6790 break; | |
6791 case SUBPAT: | |
6792 p = "SUBPAT"; | |
6793 break; | |
6794 case BRACE_LIMITS: | |
6795 p = "BRACE_LIMITS"; | |
6796 break; | |
6797 case BRACE_SIMPLE: | |
6798 p = "BRACE_SIMPLE"; | |
6799 break; | |
6800 case BRACE_COMPLEX + 0: | |
6801 case BRACE_COMPLEX + 1: | |
6802 case BRACE_COMPLEX + 2: | |
6803 case BRACE_COMPLEX + 3: | |
6804 case BRACE_COMPLEX + 4: | |
6805 case BRACE_COMPLEX + 5: | |
6806 case BRACE_COMPLEX + 6: | |
6807 case BRACE_COMPLEX + 7: | |
6808 case BRACE_COMPLEX + 8: | |
6809 case BRACE_COMPLEX + 9: | |
6810 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX); | |
6811 p = NULL; | |
6812 break; | |
6813 #ifdef FEAT_MBYTE | |
6814 case MULTIBYTECODE: | |
6815 p = "MULTIBYTECODE"; | |
6816 break; | |
6817 #endif | |
6818 case NEWL: | |
6819 p = "NEWL"; | |
6820 break; | |
6821 default: | |
6822 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op)); | |
6823 p = NULL; | |
6824 break; | |
6825 } | |
6826 if (p != NULL) | |
4444 | 6827 STRCAT(buf, p); |
6828 return (char_u *)buf; | |
7 | 6829 } |
4444 | 6830 #endif /* DEBUG */ |
7 | 6831 |
6832 #ifdef FEAT_MBYTE | |
6833 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3)); | |
6834 | |
6835 typedef struct | |
6836 { | |
6837 int a, b, c; | |
6838 } decomp_T; | |
6839 | |
6840 | |
6841 /* 0xfb20 - 0xfb4f */ | |
297 | 6842 static decomp_T decomp_table[0xfb4f-0xfb20+1] = |
7 | 6843 { |
6844 {0x5e2,0,0}, /* 0xfb20 alt ayin */ | |
6845 {0x5d0,0,0}, /* 0xfb21 alt alef */ | |
6846 {0x5d3,0,0}, /* 0xfb22 alt dalet */ | |
6847 {0x5d4,0,0}, /* 0xfb23 alt he */ | |
6848 {0x5db,0,0}, /* 0xfb24 alt kaf */ | |
6849 {0x5dc,0,0}, /* 0xfb25 alt lamed */ | |
6850 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */ | |
6851 {0x5e8,0,0}, /* 0xfb27 alt resh */ | |
6852 {0x5ea,0,0}, /* 0xfb28 alt tav */ | |
6853 {'+', 0, 0}, /* 0xfb29 alt plus */ | |
6854 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */ | |
6855 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */ | |
6856 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */ | |
6857 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */ | |
6858 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */ | |
6859 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */ | |
6860 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */ | |
6861 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */ | |
6862 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */ | |
6863 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */ | |
6864 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */ | |
6865 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */ | |
6866 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */ | |
6867 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */ | |
6868 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */ | |
6869 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */ | |
6870 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */ | |
6871 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */ | |
6872 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */ | |
6873 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */ | |
6874 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */ | |
6875 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */ | |
6876 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */ | |
6877 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */ | |
6878 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */ | |
6879 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */ | |
6880 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */ | |
6881 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */ | |
6882 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */ | |
6883 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */ | |
6884 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */ | |
6885 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */ | |
6886 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */ | |
6887 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */ | |
6888 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */ | |
6889 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */ | |
6890 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */ | |
6891 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */ | |
6892 }; | |
6893 | |
6894 static void | |
6895 mb_decompose(c, c1, c2, c3) | |
6896 int c, *c1, *c2, *c3; | |
6897 { | |
6898 decomp_T d; | |
6899 | |
6900 if (c >= 0x4b20 && c <= 0xfb4f) | |
6901 { | |
6902 d = decomp_table[c - 0xfb20]; | |
6903 *c1 = d.a; | |
6904 *c2 = d.b; | |
6905 *c3 = d.c; | |
6906 } | |
6907 else | |
6908 { | |
6909 *c1 = c; | |
6910 *c2 = *c3 = 0; | |
6911 } | |
6912 } | |
6913 #endif | |
6914 | |
6915 /* | |
6916 * Compare two strings, ignore case if ireg_ic set. | |
6917 * Return 0 if strings match, non-zero otherwise. | |
6918 * Correct the length "*n" when composing characters are ignored. | |
6919 */ | |
6920 static int | |
6921 cstrncmp(s1, s2, n) | |
6922 char_u *s1, *s2; | |
6923 int *n; | |
6924 { | |
6925 int result; | |
6926 | |
6927 if (!ireg_ic) | |
6928 result = STRNCMP(s1, s2, *n); | |
6929 else | |
6930 result = MB_STRNICMP(s1, s2, *n); | |
6931 | |
6932 #ifdef FEAT_MBYTE | |
6933 /* if it failed and it's utf8 and we want to combineignore: */ | |
6934 if (result != 0 && enc_utf8 && ireg_icombine) | |
6935 { | |
6936 char_u *str1, *str2; | |
6937 int c1, c2, c11, c12; | |
6938 int junk; | |
6939 | |
6940 /* we have to handle the strcmp ourselves, since it is necessary to | |
6941 * deal with the composing characters by ignoring them: */ | |
6942 str1 = s1; | |
6943 str2 = s2; | |
6944 c1 = c2 = 0; | |
507 | 6945 while ((int)(str1 - s1) < *n) |
7 | 6946 { |
6947 c1 = mb_ptr2char_adv(&str1); | |
6948 c2 = mb_ptr2char_adv(&str2); | |
6949 | |
6950 /* decompose the character if necessary, into 'base' characters | |
6951 * because I don't care about Arabic, I will hard-code the Hebrew | |
6952 * which I *do* care about! So sue me... */ | |
6953 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2))) | |
6954 { | |
6955 /* decomposition necessary? */ | |
6956 mb_decompose(c1, &c11, &junk, &junk); | |
6957 mb_decompose(c2, &c12, &junk, &junk); | |
6958 c1 = c11; | |
6959 c2 = c12; | |
6960 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12))) | |
6961 break; | |
6962 } | |
6963 } | |
6964 result = c2 - c1; | |
6965 if (result == 0) | |
6966 *n = (int)(str2 - s2); | |
6967 } | |
6968 #endif | |
6969 | |
6970 return result; | |
6971 } | |
6972 | |
6973 /* | |
6974 * cstrchr: This function is used a lot for simple searches, keep it fast! | |
6975 */ | |
6976 static char_u * | |
6977 cstrchr(s, c) | |
6978 char_u *s; | |
6979 int c; | |
6980 { | |
6981 char_u *p; | |
6982 int cc; | |
6983 | |
6984 if (!ireg_ic | |
6985 #ifdef FEAT_MBYTE | |
6986 || (!enc_utf8 && mb_char2len(c) > 1) | |
6987 #endif | |
6988 ) | |
6989 return vim_strchr(s, c); | |
6990 | |
6991 /* tolower() and toupper() can be slow, comparing twice should be a lot | |
6992 * faster (esp. when using MS Visual C++!). | |
6993 * For UTF-8 need to use folded case. */ | |
6994 #ifdef FEAT_MBYTE | |
6995 if (enc_utf8 && c > 0x80) | |
6996 cc = utf_fold(c); | |
6997 else | |
6998 #endif | |
1347 | 6999 if (MB_ISUPPER(c)) |
7000 cc = MB_TOLOWER(c); | |
7001 else if (MB_ISLOWER(c)) | |
7002 cc = MB_TOUPPER(c); | |
7 | 7003 else |
7004 return vim_strchr(s, c); | |
7005 | |
7006 #ifdef FEAT_MBYTE | |
7007 if (has_mbyte) | |
7008 { | |
474 | 7009 for (p = s; *p != NUL; p += (*mb_ptr2len)(p)) |
7 | 7010 { |
7011 if (enc_utf8 && c > 0x80) | |
7012 { | |
7013 if (utf_fold(utf_ptr2char(p)) == cc) | |
7014 return p; | |
7015 } | |
7016 else if (*p == c || *p == cc) | |
7017 return p; | |
7018 } | |
7019 } | |
7020 else | |
7021 #endif | |
7022 /* Faster version for when there are no multi-byte characters. */ | |
7023 for (p = s; *p != NUL; ++p) | |
7024 if (*p == c || *p == cc) | |
7025 return p; | |
7026 | |
7027 return NULL; | |
7028 } | |
7029 | |
7030 /*************************************************************** | |
7031 * regsub stuff * | |
7032 ***************************************************************/ | |
7033 | |
7034 /* This stuff below really confuses cc on an SGI -- webb */ | |
7035 #ifdef __sgi | |
7036 # undef __ARGS | |
7037 # define __ARGS(x) () | |
7038 #endif | |
7039 | |
7040 /* | |
7041 * We should define ftpr as a pointer to a function returning a pointer to | |
7042 * a function returning a pointer to a function ... | |
7043 * This is impossible, so we declare a pointer to a function returning a | |
7044 * pointer to a function returning void. This should work for all compilers. | |
7045 */ | |
772 | 7046 typedef void (*(*fptr_T) __ARGS((int *, int)))(); |
7047 | |
7048 static fptr_T do_upper __ARGS((int *, int)); | |
7049 static fptr_T do_Upper __ARGS((int *, int)); | |
7050 static fptr_T do_lower __ARGS((int *, int)); | |
7051 static fptr_T do_Lower __ARGS((int *, int)); | |
7 | 7052 |
7053 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash)); | |
7054 | |
772 | 7055 static fptr_T |
7 | 7056 do_upper(d, c) |
772 | 7057 int *d; |
7058 int c; | |
7 | 7059 { |
772 | 7060 *d = MB_TOUPPER(c); |
7061 | |
7062 return (fptr_T)NULL; | |
7 | 7063 } |
7064 | |
772 | 7065 static fptr_T |
7066 do_Upper(d, c) | |
7067 int *d; | |
7 | 7068 int c; |
7069 { | |
772 | 7070 *d = MB_TOUPPER(c); |
7071 | |
7072 return (fptr_T)do_Upper; | |
7073 } | |
7074 | |
7075 static fptr_T | |
7076 do_lower(d, c) | |
7077 int *d; | |
7078 int c; | |
7079 { | |
7080 *d = MB_TOLOWER(c); | |
7081 | |
7082 return (fptr_T)NULL; | |
7083 } | |
7084 | |
7085 static fptr_T | |
7086 do_Lower(d, c) | |
7087 int *d; | |
7088 int c; | |
7089 { | |
7090 *d = MB_TOLOWER(c); | |
7091 | |
7092 return (fptr_T)do_Lower; | |
7 | 7093 } |
7094 | |
7095 /* | |
7096 * regtilde(): Replace tildes in the pattern by the old pattern. | |
7097 * | |
7098 * Short explanation of the tilde: It stands for the previous replacement | |
7099 * pattern. If that previous pattern also contains a ~ we should go back a | |
7100 * step further... But we insert the previous pattern into the current one | |
7101 * and remember that. | |
772 | 7102 * This still does not handle the case where "magic" changes. So require the |
7103 * user to keep his hands off of "magic". | |
7 | 7104 * |
7105 * The tildes are parsed once before the first call to vim_regsub(). | |
7106 */ | |
7107 char_u * | |
7108 regtilde(source, magic) | |
7109 char_u *source; | |
7110 int magic; | |
7111 { | |
7112 char_u *newsub = source; | |
7113 char_u *tmpsub; | |
7114 char_u *p; | |
7115 int len; | |
7116 int prevlen; | |
7117 | |
7118 for (p = newsub; *p; ++p) | |
7119 { | |
7120 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic)) | |
7121 { | |
7122 if (reg_prev_sub != NULL) | |
7123 { | |
7124 /* length = len(newsub) - 1 + len(prev_sub) + 1 */ | |
7125 prevlen = (int)STRLEN(reg_prev_sub); | |
7126 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen)); | |
7127 if (tmpsub != NULL) | |
7128 { | |
7129 /* copy prefix */ | |
7130 len = (int)(p - newsub); /* not including ~ */ | |
7131 mch_memmove(tmpsub, newsub, (size_t)len); | |
1209 | 7132 /* interpret tilde */ |
7 | 7133 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen); |
7134 /* copy postfix */ | |
7135 if (!magic) | |
7136 ++p; /* back off \ */ | |
7137 STRCPY(tmpsub + len + prevlen, p + 1); | |
7138 | |
7139 if (newsub != source) /* already allocated newsub */ | |
7140 vim_free(newsub); | |
7141 newsub = tmpsub; | |
7142 p = newsub + len + prevlen; | |
7143 } | |
7144 } | |
7145 else if (magic) | |
1621 | 7146 STRMOVE(p, p + 1); /* remove '~' */ |
7 | 7147 else |
1621 | 7148 STRMOVE(p, p + 2); /* remove '\~' */ |
7 | 7149 --p; |
7150 } | |
7151 else | |
7152 { | |
7153 if (*p == '\\' && p[1]) /* skip escaped characters */ | |
7154 ++p; | |
7155 #ifdef FEAT_MBYTE | |
7156 if (has_mbyte) | |
474 | 7157 p += (*mb_ptr2len)(p) - 1; |
7 | 7158 #endif |
7159 } | |
7160 } | |
7161 | |
7162 vim_free(reg_prev_sub); | |
7163 if (newsub != source) /* newsub was allocated, just keep it */ | |
7164 reg_prev_sub = newsub; | |
7165 else /* no ~ found, need to save newsub */ | |
7166 reg_prev_sub = vim_strsave(newsub); | |
7167 return newsub; | |
7168 } | |
7169 | |
7170 #ifdef FEAT_EVAL | |
7171 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */ | |
7172 | |
7173 /* These pointers are used instead of reg_match and reg_mmatch for | |
7174 * reg_submatch(). Needed for when the substitution string is an expression | |
7175 * that contains a call to substitute() and submatch(). */ | |
7176 static regmatch_T *submatch_match; | |
7177 static regmmatch_T *submatch_mmatch; | |
2011 | 7178 static linenr_T submatch_firstlnum; |
7179 static linenr_T submatch_maxline; | |
2904 | 7180 static int submatch_line_lbr; |
7 | 7181 #endif |
7182 | |
7183 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO) | |
7184 /* | |
7185 * vim_regsub() - perform substitutions after a vim_regexec() or | |
7186 * vim_regexec_multi() match. | |
7187 * | |
7188 * If "copy" is TRUE really copy into "dest". | |
7189 * If "copy" is FALSE nothing is copied, this is just to find out the length | |
7190 * of the result. | |
7191 * | |
7192 * If "backslash" is TRUE, a backslash will be removed later, need to double | |
7193 * them to keep them, and insert a backslash before a CR to avoid it being | |
7194 * replaced with a line break later. | |
7195 * | |
7196 * Note: The matched text must not change between the call of | |
7197 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back | |
7198 * references invalid! | |
7199 * | |
7200 * Returns the size of the replacement, including terminating NUL. | |
7201 */ | |
7202 int | |
7203 vim_regsub(rmp, source, dest, copy, magic, backslash) | |
7204 regmatch_T *rmp; | |
7205 char_u *source; | |
7206 char_u *dest; | |
7207 int copy; | |
7208 int magic; | |
7209 int backslash; | |
7210 { | |
7211 reg_match = rmp; | |
7212 reg_mmatch = NULL; | |
7213 reg_maxline = 0; | |
4061 | 7214 reg_buf = curbuf; |
7 | 7215 return vim_regsub_both(source, dest, copy, magic, backslash); |
7216 } | |
7217 #endif | |
7218 | |
7219 int | |
7220 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash) | |
7221 regmmatch_T *rmp; | |
7222 linenr_T lnum; | |
7223 char_u *source; | |
7224 char_u *dest; | |
7225 int copy; | |
7226 int magic; | |
7227 int backslash; | |
7228 { | |
7229 reg_match = NULL; | |
7230 reg_mmatch = rmp; | |
7231 reg_buf = curbuf; /* always works on the current buffer! */ | |
7232 reg_firstlnum = lnum; | |
7233 reg_maxline = curbuf->b_ml.ml_line_count - lnum; | |
7234 return vim_regsub_both(source, dest, copy, magic, backslash); | |
7235 } | |
7236 | |
7237 static int | |
7238 vim_regsub_both(source, dest, copy, magic, backslash) | |
7239 char_u *source; | |
7240 char_u *dest; | |
7241 int copy; | |
7242 int magic; | |
7243 int backslash; | |
7244 { | |
7245 char_u *src; | |
7246 char_u *dst; | |
7247 char_u *s; | |
7248 int c; | |
772 | 7249 int cc; |
7 | 7250 int no = -1; |
4244 | 7251 fptr_T func_all = (fptr_T)NULL; |
7252 fptr_T func_one = (fptr_T)NULL; | |
7 | 7253 linenr_T clnum = 0; /* init for GCC */ |
7254 int len = 0; /* init for GCC */ | |
7255 #ifdef FEAT_EVAL | |
7256 static char_u *eval_result = NULL; | |
7257 #endif | |
7258 | |
7259 /* Be paranoid... */ | |
7260 if (source == NULL || dest == NULL) | |
7261 { | |
7262 EMSG(_(e_null)); | |
7263 return 0; | |
7264 } | |
7265 if (prog_magic_wrong()) | |
7266 return 0; | |
7267 src = source; | |
7268 dst = dest; | |
7269 | |
7270 /* | |
7271 * When the substitute part starts with "\=" evaluate it as an expression. | |
7272 */ | |
7273 if (source[0] == '\\' && source[1] == '=' | |
7274 #ifdef FEAT_EVAL | |
7275 && !can_f_submatch /* can't do this recursively */ | |
7276 #endif | |
7277 ) | |
7278 { | |
7279 #ifdef FEAT_EVAL | |
7280 /* To make sure that the length doesn't change between checking the | |
7281 * length and copying the string, and to speed up things, the | |
7282 * resulting string is saved from the call with "copy" == FALSE to the | |
7283 * call with "copy" == TRUE. */ | |
7284 if (copy) | |
7285 { | |
7286 if (eval_result != NULL) | |
7287 { | |
7288 STRCPY(dest, eval_result); | |
7289 dst += STRLEN(eval_result); | |
7290 vim_free(eval_result); | |
7291 eval_result = NULL; | |
7292 } | |
7293 } | |
7294 else | |
7295 { | |
7296 win_T *save_reg_win; | |
7297 int save_ireg_ic; | |
7298 | |
7299 vim_free(eval_result); | |
7300 | |
7301 /* The expression may contain substitute(), which calls us | |
7302 * recursively. Make sure submatch() gets the text from the first | |
7303 * level. Don't need to save "reg_buf", because | |
7304 * vim_regexec_multi() can't be called recursively. */ | |
7305 submatch_match = reg_match; | |
7306 submatch_mmatch = reg_mmatch; | |
2011 | 7307 submatch_firstlnum = reg_firstlnum; |
7308 submatch_maxline = reg_maxline; | |
2904 | 7309 submatch_line_lbr = reg_line_lbr; |
7 | 7310 save_reg_win = reg_win; |
7311 save_ireg_ic = ireg_ic; | |
7312 can_f_submatch = TRUE; | |
7313 | |
714 | 7314 eval_result = eval_to_string(source + 2, NULL, TRUE); |
7 | 7315 if (eval_result != NULL) |
7316 { | |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7317 int had_backslash = FALSE; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7318 |
39 | 7319 for (s = eval_result; *s != NUL; mb_ptr_adv(s)) |
7 | 7320 { |
2904 | 7321 /* Change NL to CR, so that it becomes a line break, |
7322 * unless called from vim_regexec_nl(). | |
7 | 7323 * Skip over a backslashed character. */ |
2904 | 7324 if (*s == NL && !submatch_line_lbr) |
7 | 7325 *s = CAR; |
7326 else if (*s == '\\' && s[1] != NUL) | |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7327 { |
7 | 7328 ++s; |
2173 | 7329 /* Change NL to CR here too, so that this works: |
7330 * :s/abc\\\ndef/\="aaa\\\nbbb"/ on text: | |
7331 * abc\ | |
7332 * def | |
2904 | 7333 * Not when called from vim_regexec_nl(). |
2173 | 7334 */ |
2904 | 7335 if (*s == NL && !submatch_line_lbr) |
2173 | 7336 *s = CAR; |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7337 had_backslash = TRUE; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7338 } |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7339 } |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7340 if (had_backslash && backslash) |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7341 { |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7342 /* Backslashes will be consumed, need to double them. */ |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7343 s = vim_strsave_escaped(eval_result, (char_u *)"\\"); |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7344 if (s != NULL) |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7345 { |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7346 vim_free(eval_result); |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7347 eval_result = s; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7348 } |
7 | 7349 } |
7350 | |
7351 dst += STRLEN(eval_result); | |
7352 } | |
7353 | |
7354 reg_match = submatch_match; | |
7355 reg_mmatch = submatch_mmatch; | |
2011 | 7356 reg_firstlnum = submatch_firstlnum; |
7357 reg_maxline = submatch_maxline; | |
2904 | 7358 reg_line_lbr = submatch_line_lbr; |
7 | 7359 reg_win = save_reg_win; |
7360 ireg_ic = save_ireg_ic; | |
7361 can_f_submatch = FALSE; | |
7362 } | |
7363 #endif | |
7364 } | |
7365 else | |
7366 while ((c = *src++) != NUL) | |
7367 { | |
7368 if (c == '&' && magic) | |
7369 no = 0; | |
7370 else if (c == '\\' && *src != NUL) | |
7371 { | |
7372 if (*src == '&' && !magic) | |
7373 { | |
7374 ++src; | |
7375 no = 0; | |
7376 } | |
7377 else if ('0' <= *src && *src <= '9') | |
7378 { | |
7379 no = *src++ - '0'; | |
7380 } | |
7381 else if (vim_strchr((char_u *)"uUlLeE", *src)) | |
7382 { | |
7383 switch (*src++) | |
7384 { | |
4244 | 7385 case 'u': func_one = (fptr_T)do_upper; |
7 | 7386 continue; |
4244 | 7387 case 'U': func_all = (fptr_T)do_Upper; |
7 | 7388 continue; |
4244 | 7389 case 'l': func_one = (fptr_T)do_lower; |
7 | 7390 continue; |
4244 | 7391 case 'L': func_all = (fptr_T)do_Lower; |
7 | 7392 continue; |
7393 case 'e': | |
4244 | 7394 case 'E': func_one = func_all = (fptr_T)NULL; |
7 | 7395 continue; |
7396 } | |
7397 } | |
7398 } | |
7399 if (no < 0) /* Ordinary character. */ | |
7400 { | |
798 | 7401 if (c == K_SPECIAL && src[0] != NUL && src[1] != NUL) |
7402 { | |
1209 | 7403 /* Copy a special key as-is. */ |
798 | 7404 if (copy) |
7405 { | |
7406 *dst++ = c; | |
7407 *dst++ = *src++; | |
7408 *dst++ = *src++; | |
7409 } | |
7410 else | |
7411 { | |
7412 dst += 3; | |
7413 src += 2; | |
7414 } | |
7415 continue; | |
7416 } | |
7417 | |
7 | 7418 if (c == '\\' && *src != NUL) |
7419 { | |
7420 /* Check for abbreviations -- webb */ | |
7421 switch (*src) | |
7422 { | |
7423 case 'r': c = CAR; ++src; break; | |
7424 case 'n': c = NL; ++src; break; | |
7425 case 't': c = TAB; ++src; break; | |
7426 /* Oh no! \e already has meaning in subst pat :-( */ | |
7427 /* case 'e': c = ESC; ++src; break; */ | |
7428 case 'b': c = Ctrl_H; ++src; break; | |
7429 | |
7430 /* If "backslash" is TRUE the backslash will be removed | |
7431 * later. Used to insert a literal CR. */ | |
7432 default: if (backslash) | |
7433 { | |
7434 if (copy) | |
7435 *dst = '\\'; | |
7436 ++dst; | |
7437 } | |
7438 c = *src++; | |
7439 } | |
7440 } | |
798 | 7441 #ifdef FEAT_MBYTE |
7442 else if (has_mbyte) | |
7443 c = mb_ptr2char(src - 1); | |
7444 #endif | |
7 | 7445 |
7446 /* Write to buffer, if copy is set. */ | |
4244 | 7447 if (func_one != (fptr_T)NULL) |
7448 /* Turbo C complains without the typecast */ | |
7449 func_one = (fptr_T)(func_one(&cc, c)); | |
7450 else if (func_all != (fptr_T)NULL) | |
7451 /* Turbo C complains without the typecast */ | |
7452 func_all = (fptr_T)(func_all(&cc, c)); | |
7453 else /* just copy */ | |
772 | 7454 cc = c; |
7455 | |
7456 #ifdef FEAT_MBYTE | |
7457 if (has_mbyte) | |
7 | 7458 { |
2307
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7459 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
|
7460 |
7 | 7461 if (copy) |
772 | 7462 mb_char2bytes(cc, dst); |
7463 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
|
7464 if (enc_utf8) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7465 { |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7466 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
|
7467 |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7468 /* 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
|
7469 * 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
|
7470 if (clen < totlen) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7471 { |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7472 if (copy) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7473 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
|
7474 (size_t)(totlen - clen)); |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7475 dst += totlen - clen; |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7476 } |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7477 } |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7478 src += totlen - 1; |
7 | 7479 } |
7480 else | |
7481 #endif | |
7482 if (copy) | |
772 | 7483 *dst = cc; |
7 | 7484 dst++; |
7485 } | |
7486 else | |
7487 { | |
7488 if (REG_MULTI) | |
7489 { | |
7490 clnum = reg_mmatch->startpos[no].lnum; | |
7491 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0) | |
7492 s = NULL; | |
7493 else | |
7494 { | |
7495 s = reg_getline(clnum) + reg_mmatch->startpos[no].col; | |
7496 if (reg_mmatch->endpos[no].lnum == clnum) | |
7497 len = reg_mmatch->endpos[no].col | |
7498 - reg_mmatch->startpos[no].col; | |
7499 else | |
7500 len = (int)STRLEN(s); | |
7501 } | |
7502 } | |
7503 else | |
7504 { | |
7505 s = reg_match->startp[no]; | |
7506 if (reg_match->endp[no] == NULL) | |
7507 s = NULL; | |
7508 else | |
7509 len = (int)(reg_match->endp[no] - s); | |
7510 } | |
7511 if (s != NULL) | |
7512 { | |
7513 for (;;) | |
7514 { | |
7515 if (len == 0) | |
7516 { | |
7517 if (REG_MULTI) | |
7518 { | |
7519 if (reg_mmatch->endpos[no].lnum == clnum) | |
7520 break; | |
7521 if (copy) | |
7522 *dst = CAR; | |
7523 ++dst; | |
7524 s = reg_getline(++clnum); | |
7525 if (reg_mmatch->endpos[no].lnum == clnum) | |
7526 len = reg_mmatch->endpos[no].col; | |
7527 else | |
7528 len = (int)STRLEN(s); | |
7529 } | |
7530 else | |
7531 break; | |
7532 } | |
7533 else if (*s == NUL) /* we hit NUL. */ | |
7534 { | |
7535 if (copy) | |
7536 EMSG(_(e_re_damg)); | |
7537 goto exit; | |
7538 } | |
7539 else | |
7540 { | |
7541 if (backslash && (*s == CAR || *s == '\\')) | |
7542 { | |
7543 /* | |
7544 * Insert a backslash in front of a CR, otherwise | |
7545 * it will be replaced by a line break. | |
7546 * Number of backslashes will be halved later, | |
7547 * double them here. | |
7548 */ | |
7549 if (copy) | |
7550 { | |
7551 dst[0] = '\\'; | |
7552 dst[1] = *s; | |
7553 } | |
7554 dst += 2; | |
7555 } | |
7556 else | |
7557 { | |
772 | 7558 #ifdef FEAT_MBYTE |
7559 if (has_mbyte) | |
7560 c = mb_ptr2char(s); | |
7561 else | |
7562 #endif | |
7563 c = *s; | |
7564 | |
4244 | 7565 if (func_one != (fptr_T)NULL) |
7566 /* Turbo C complains without the typecast */ | |
7567 func_one = (fptr_T)(func_one(&cc, c)); | |
7568 else if (func_all != (fptr_T)NULL) | |
7569 /* Turbo C complains without the typecast */ | |
7570 func_all = (fptr_T)(func_all(&cc, c)); | |
7571 else /* just copy */ | |
772 | 7572 cc = c; |
7573 | |
7574 #ifdef FEAT_MBYTE | |
7575 if (has_mbyte) | |
7 | 7576 { |
1332 | 7577 int l; |
7578 | |
7579 /* Copy composing characters separately, one | |
7580 * at a time. */ | |
7581 if (enc_utf8) | |
7582 l = utf_ptr2len(s) - 1; | |
7583 else | |
7584 l = mb_ptr2len(s) - 1; | |
772 | 7585 |
7586 s += l; | |
7587 len -= l; | |
7588 if (copy) | |
7589 mb_char2bytes(cc, dst); | |
7590 dst += mb_char2len(cc) - 1; | |
7 | 7591 } |
772 | 7592 else |
7593 #endif | |
7594 if (copy) | |
7595 *dst = cc; | |
7596 dst++; | |
7 | 7597 } |
772 | 7598 |
7 | 7599 ++s; |
7600 --len; | |
7601 } | |
7602 } | |
7603 } | |
7604 no = -1; | |
7605 } | |
7606 } | |
7607 if (copy) | |
7608 *dst = NUL; | |
7609 | |
7610 exit: | |
7611 return (int)((dst - dest) + 1); | |
7612 } | |
7613 | |
7614 #ifdef FEAT_EVAL | |
2012 | 7615 static char_u *reg_getline_submatch __ARGS((linenr_T lnum)); |
7616 | |
7 | 7617 /* |
2011 | 7618 * Call reg_getline() with the line numbers from the submatch. If a |
7619 * substitute() was used the reg_maxline and other values have been | |
7620 * overwritten. | |
7621 */ | |
7622 static char_u * | |
7623 reg_getline_submatch(lnum) | |
7624 linenr_T lnum; | |
7625 { | |
7626 char_u *s; | |
7627 linenr_T save_first = reg_firstlnum; | |
7628 linenr_T save_max = reg_maxline; | |
7629 | |
7630 reg_firstlnum = submatch_firstlnum; | |
7631 reg_maxline = submatch_maxline; | |
7632 | |
7633 s = reg_getline(lnum); | |
7634 | |
7635 reg_firstlnum = save_first; | |
7636 reg_maxline = save_max; | |
7637 return s; | |
7638 } | |
7639 | |
7640 /* | |
1209 | 7641 * Used for the submatch() function: get the string from the n'th submatch in |
7 | 7642 * allocated memory. |
7643 * Returns NULL when not in a ":s" command and for a non-existing submatch. | |
7644 */ | |
7645 char_u * | |
7646 reg_submatch(no) | |
7647 int no; | |
7648 { | |
7649 char_u *retval = NULL; | |
7650 char_u *s; | |
7651 int len; | |
7652 int round; | |
7653 linenr_T lnum; | |
7654 | |
840 | 7655 if (!can_f_submatch || no < 0) |
7 | 7656 return NULL; |
7657 | |
7658 if (submatch_match == NULL) | |
7659 { | |
7660 /* | |
7661 * First round: compute the length and allocate memory. | |
7662 * Second round: copy the text. | |
7663 */ | |
7664 for (round = 1; round <= 2; ++round) | |
7665 { | |
7666 lnum = submatch_mmatch->startpos[no].lnum; | |
7667 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0) | |
7668 return NULL; | |
7669 | |
2011 | 7670 s = reg_getline_submatch(lnum) + submatch_mmatch->startpos[no].col; |
7 | 7671 if (s == NULL) /* anti-crash check, cannot happen? */ |
7672 break; | |
7673 if (submatch_mmatch->endpos[no].lnum == lnum) | |
7674 { | |
7675 /* Within one line: take form start to end col. */ | |
7676 len = submatch_mmatch->endpos[no].col | |
7677 - submatch_mmatch->startpos[no].col; | |
7678 if (round == 2) | |
418 | 7679 vim_strncpy(retval, s, len); |
7 | 7680 ++len; |
7681 } | |
7682 else | |
7683 { | |
7684 /* Multiple lines: take start line from start col, middle | |
7685 * lines completely and end line up to end col. */ | |
7686 len = (int)STRLEN(s); | |
7687 if (round == 2) | |
7688 { | |
7689 STRCPY(retval, s); | |
7690 retval[len] = '\n'; | |
7691 } | |
7692 ++len; | |
7693 ++lnum; | |
7694 while (lnum < submatch_mmatch->endpos[no].lnum) | |
7695 { | |
2011 | 7696 s = reg_getline_submatch(lnum++); |
7 | 7697 if (round == 2) |
7698 STRCPY(retval + len, s); | |
7699 len += (int)STRLEN(s); | |
7700 if (round == 2) | |
7701 retval[len] = '\n'; | |
7702 ++len; | |
7703 } | |
7704 if (round == 2) | |
2011 | 7705 STRNCPY(retval + len, reg_getline_submatch(lnum), |
7 | 7706 submatch_mmatch->endpos[no].col); |
7707 len += submatch_mmatch->endpos[no].col; | |
7708 if (round == 2) | |
7709 retval[len] = NUL; | |
7710 ++len; | |
7711 } | |
7712 | |
840 | 7713 if (retval == NULL) |
7 | 7714 { |
7715 retval = lalloc((long_u)len, TRUE); | |
840 | 7716 if (retval == NULL) |
7 | 7717 return NULL; |
7718 } | |
7719 } | |
7720 } | |
7721 else | |
7722 { | |
1815 | 7723 s = submatch_match->startp[no]; |
7724 if (s == NULL || submatch_match->endp[no] == NULL) | |
7 | 7725 retval = NULL; |
7726 else | |
7727 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s)); | |
7728 } | |
7729 | |
7730 return retval; | |
7731 } | |
7732 #endif | |
4444 | 7733 |
7734 static regengine_T bt_regengine = | |
7735 { | |
7736 bt_regcomp, | |
7737 bt_regexec, | |
7738 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7739 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7740 bt_regexec_nl, | |
7741 #endif | |
7742 bt_regexec_multi | |
7743 #ifdef DEBUG | |
7744 ,(char_u *)"" | |
7745 #endif | |
7746 }; | |
7747 | |
7748 | |
7749 #include "regexp_nfa.c" | |
7750 | |
7751 static regengine_T nfa_regengine = | |
7752 { | |
7753 nfa_regcomp, | |
7754 nfa_regexec, | |
7755 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7756 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7757 nfa_regexec_nl, | |
7758 #endif | |
7759 nfa_regexec_multi | |
7760 #ifdef DEBUG | |
7761 ,(char_u *)"" | |
7762 #endif | |
7763 }; | |
7764 | |
7765 /* Which regexp engine to use? Needed for vim_regcomp(). | |
7766 * Must match with 'regexpengine'. */ | |
7767 static int regexp_engine = 0; | |
7768 #define AUTOMATIC_ENGINE 0 | |
7769 #define BACKTRACKING_ENGINE 1 | |
7770 #define NFA_ENGINE 2 | |
7771 #ifdef DEBUG | |
7772 static char_u regname[][30] = { | |
7773 "AUTOMATIC Regexp Engine", | |
7774 "BACKTACKING Regexp Engine", | |
7775 "NFA Regexp Engine" | |
7776 }; | |
7777 #endif | |
7778 | |
7779 /* | |
7780 * Compile a regular expression into internal code. | |
7781 * Returns the program in allocated memory. Returns NULL for an error. | |
7782 */ | |
7783 regprog_T * | |
7784 vim_regcomp(expr_arg, re_flags) | |
7785 char_u *expr_arg; | |
7786 int re_flags; | |
7787 { | |
7788 regprog_T *prog = NULL; | |
7789 char_u *expr = expr_arg; | |
7790 | |
7791 syntax_error = FALSE; | |
7792 regexp_engine = p_re; | |
7793 | |
7794 /* Check for prefix "\%#=", that sets the regexp engine */ | |
7795 if (STRNCMP(expr, "\\%#=", 4) == 0) | |
7796 { | |
7797 int newengine = expr[4] - '0'; | |
7798 | |
7799 if (newengine == AUTOMATIC_ENGINE | |
7800 || newengine == BACKTRACKING_ENGINE | |
7801 || newengine == NFA_ENGINE) | |
7802 { | |
7803 regexp_engine = expr[4] - '0'; | |
7804 expr += 5; | |
7805 #ifdef DEBUG | |
7806 EMSG3("New regexp mode selected (%d): %s", regexp_engine, | |
7807 regname[newengine]); | |
7808 #endif | |
7809 } | |
7810 else | |
7811 { | |
7812 EMSG(_("E864: \\%#= can only be followed by 0, 1, or 2. The automatic engine will be used ")); | |
7813 regexp_engine = AUTOMATIC_ENGINE; | |
7814 } | |
7815 } | |
7816 #ifdef DEBUG | |
7817 bt_regengine.expr = expr; | |
7818 nfa_regengine.expr = expr; | |
7819 #endif | |
7820 | |
7821 /* | |
7822 * First try the NFA engine, unless backtracking was requested. | |
7823 */ | |
7824 if (regexp_engine != BACKTRACKING_ENGINE) | |
7825 prog = nfa_regengine.regcomp(expr, re_flags); | |
7826 else | |
7827 prog = bt_regengine.regcomp(expr, re_flags); | |
7828 | |
7829 if (prog == NULL) /* error compiling regexp with initial engine */ | |
7830 { | |
7831 #ifdef DEBUG | |
7832 if (regexp_engine != BACKTRACKING_ENGINE) /* debugging log for NFA */ | |
7833 { | |
7834 FILE *f; | |
7835 f = fopen("debug.log", "a"); | |
7836 if (f) | |
7837 { | |
7838 if (!syntax_error) | |
7839 fprintf(f, "NFA engine could not handle \"%s\"\n", expr); | |
7840 else | |
7841 fprintf(f, "Syntax error in \"%s\"\n", expr); | |
7842 fclose(f); | |
7843 } | |
7844 else | |
7845 EMSG("(NFA) Could not open \"debug.log\" to write !!!"); | |
7846 /* | |
7847 if (syntax_error) | |
7848 EMSG("NFA Regexp: Syntax Error !"); | |
7849 */ | |
7850 } | |
7851 #endif | |
7852 /* | |
7853 * If NFA engine failed, then revert to the backtracking engine. | |
7854 * Except when there was a syntax error, which was properly handled by | |
7855 * NFA engine. | |
7856 */ | |
7857 if (regexp_engine == AUTOMATIC_ENGINE) | |
7858 if (!syntax_error) | |
7859 prog = bt_regengine.regcomp(expr, re_flags); | |
7860 | |
7861 } /* endif prog==NULL */ | |
7862 | |
7863 | |
7864 return prog; | |
7865 } | |
7866 | |
7867 /* | |
7868 * Match a regexp against a string. | |
7869 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
7870 * Uses curbuf for line count and 'iskeyword'. | |
7871 * | |
7872 * Return TRUE if there is a match, FALSE if not. | |
7873 */ | |
7874 int | |
7875 vim_regexec(rmp, line, col) | |
7876 regmatch_T *rmp; | |
7877 char_u *line; /* string to match against */ | |
7878 colnr_T col; /* column to start looking for match */ | |
7879 { | |
7880 return rmp->regprog->engine->regexec(rmp, line, col); | |
7881 } | |
7882 | |
7883 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7884 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7885 /* | |
7886 * Like vim_regexec(), but consider a "\n" in "line" to be a line break. | |
7887 */ | |
7888 int | |
7889 vim_regexec_nl(rmp, line, col) | |
7890 regmatch_T *rmp; | |
7891 char_u *line; | |
7892 colnr_T col; | |
7893 { | |
7894 return rmp->regprog->engine->regexec_nl(rmp, line, col); | |
7895 } | |
7896 #endif | |
7897 | |
7898 /* | |
7899 * Match a regexp against multiple lines. | |
7900 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
7901 * Uses curbuf for line count and 'iskeyword'. | |
7902 * | |
7903 * Return zero if there is no match. Return number of lines contained in the | |
7904 * match otherwise. | |
7905 */ | |
7906 long | |
7907 vim_regexec_multi(rmp, win, buf, lnum, col, tm) | |
7908 regmmatch_T *rmp; | |
7909 win_T *win; /* window in which to search or NULL */ | |
7910 buf_T *buf; /* buffer in which to search */ | |
7911 linenr_T lnum; /* nr of line to start looking for match */ | |
7912 colnr_T col; /* column to start looking for match */ | |
7913 proftime_T *tm; /* timeout limit or NULL */ | |
7914 { | |
7915 return rmp->regprog->engine->regexec_multi(rmp, win, buf, lnum, col, tm); | |
7916 } |