Mercurial > vim
annotate src/regexp.c @ 4687:79797c8534b5
Added tag v7-3-1090 for changeset 8db697ae406a
author | Bram Moolenaar <bram@vim.org> |
---|---|
date | Sat, 01 Jun 2013 23:02:55 +0200 |
parents | 2f51ee8825db |
children | 371cc0c44097 |
rev | line source |
---|---|
7 | 1 /* vi:set ts=8 sts=4 sw=4: |
2 * | |
3 * Handling of regular expressions: vim_regcomp(), vim_regexec(), vim_regsub() | |
4 * | |
5 * NOTICE: | |
6 * | |
7 * This is NOT the original regular expression code as written by Henry | |
8 * Spencer. This code has been modified specifically for use with the VIM | |
9 * editor, and should not be used separately from Vim. If you want a good | |
10 * regular expression library, get the original code. The copyright notice | |
11 * that follows is from the original. | |
12 * | |
13 * END NOTICE | |
14 * | |
15 * Copyright (c) 1986 by University of Toronto. | |
16 * Written by Henry Spencer. Not derived from licensed software. | |
17 * | |
18 * Permission is granted to anyone to use this software for any | |
19 * purpose on any computer system, and to redistribute it freely, | |
20 * subject to the following restrictions: | |
21 * | |
22 * 1. The author is not responsible for the consequences of use of | |
23 * this software, no matter how awful, even if they arise | |
24 * from defects in it. | |
25 * | |
26 * 2. The origin of this software must not be misrepresented, either | |
27 * by explicit claim or by omission. | |
28 * | |
29 * 3. Altered versions must be plainly marked as such, and must not | |
30 * be misrepresented as being the original software. | |
31 * | |
32 * Beware that some of this code is subtly aware of the way operator | |
33 * precedence is structured in regular expressions. Serious changes in | |
34 * regular-expression syntax might require a total rethink. | |
35 * | |
24 | 36 * Changes have been made by Tony Andrews, Olaf 'Rhialto' Seibert, Robert |
37 * Webb, Ciaran McCreesh and Bram Moolenaar. | |
7 | 38 * Named character class support added by Walter Briscoe (1998 Jul 01) |
39 */ | |
40 | |
4444 | 41 /* Uncomment the first if you do not want to see debugging logs or files |
42 * related to regular expressions, even when compiling with -DDEBUG. | |
43 * Uncomment the second to get the regexp debugging. */ | |
44 /* #undef DEBUG */ | |
45 /* #define DEBUG */ | |
46 | |
7 | 47 #include "vim.h" |
48 | |
4444 | 49 #ifdef DEBUG |
50 /* show/save debugging data when BT engine is used */ | |
51 # define BT_REGEXP_DUMP | |
52 /* save the debugging data to a file instead of displaying it */ | |
53 # define BT_REGEXP_LOG | |
4460 | 54 # define BT_REGEXP_DEBUG_LOG |
55 # define BT_REGEXP_DEBUG_LOG_NAME "bt_regexp_debug.log" | |
4444 | 56 #endif |
7 | 57 |
58 /* | |
59 * The "internal use only" fields in regexp.h are present to pass info from | |
60 * compile to execute that permits the execute phase to run lots faster on | |
61 * simple cases. They are: | |
62 * | |
63 * regstart char that must begin a match; NUL if none obvious; Can be a | |
64 * multi-byte character. | |
65 * reganch is the match anchored (at beginning-of-line only)? | |
66 * regmust string (pointer into program) that match must include, or NULL | |
67 * regmlen length of regmust string | |
68 * regflags RF_ values or'ed together | |
69 * | |
70 * Regstart and reganch permit very fast decisions on suitable starting points | |
71 * for a match, cutting down the work a lot. Regmust permits fast rejection | |
72 * of lines that cannot possibly match. The regmust tests are costly enough | |
73 * that vim_regcomp() supplies a regmust only if the r.e. contains something | |
74 * potentially expensive (at present, the only such thing detected is * or + | |
75 * at the start of the r.e., which can involve a lot of backup). Regmlen is | |
76 * supplied because the test in vim_regexec() needs it and vim_regcomp() is | |
77 * computing it anyway. | |
78 */ | |
79 | |
80 /* | |
81 * Structure for regexp "program". This is essentially a linear encoding | |
82 * of a nondeterministic finite-state machine (aka syntax charts or | |
83 * "railroad normal form" in parsing technology). Each node is an opcode | |
84 * plus a "next" pointer, possibly plus an operand. "Next" pointers of | |
85 * all nodes except BRANCH and BRACES_COMPLEX implement concatenation; a "next" | |
86 * pointer with a BRANCH on both ends of it is connecting two alternatives. | |
87 * (Here we have one of the subtle syntax dependencies: an individual BRANCH | |
88 * (as opposed to a collection of them) is never concatenated with anything | |
89 * because of operator precedence). The "next" pointer of a BRACES_COMPLEX | |
167 | 90 * node points to the node after the stuff to be repeated. |
91 * The operand of some types of node is a literal string; for others, it is a | |
92 * node leading into a sub-FSM. In particular, the operand of a BRANCH node | |
93 * is the first node of the branch. | |
94 * (NB this is *not* a tree structure: the tail of the branch connects to the | |
95 * thing following the set of BRANCHes.) | |
7 | 96 * |
97 * pattern is coded like: | |
98 * | |
99 * +-----------------+ | |
100 * | V | |
101 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END | |
102 * | ^ | ^ | |
103 * +------+ +----------+ | |
104 * | |
105 * | |
106 * +------------------+ | |
107 * V | | |
108 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END | |
109 * | | ^ ^ | |
110 * | +---------------+ | | |
111 * +---------------------------------------------+ | |
112 * | |
113 * | |
167 | 114 * +----------------------+ |
115 * V | | |
233 | 116 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END |
856 | 117 * | | ^ ^ |
118 * | +-----------+ | | |
179 | 119 * +--------------------------------------------------+ |
167 | 120 * |
121 * | |
7 | 122 * +-------------------------+ |
123 * V | | |
124 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END | |
125 * | | ^ | |
126 * | +----------------+ | |
127 * +-----------------------------------------------+ | |
128 * | |
129 * | |
130 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END | |
131 * | | ^ ^ | |
132 * | +----------------+ | | |
133 * +--------------------------------+ | |
134 * | |
135 * +---------+ | |
136 * | V | |
137 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END | |
138 * | | | | ^ ^ | |
139 * | | | +-----+ | | |
140 * | | +----------------+ | | |
141 * | +---------------------------+ | | |
142 * +------------------------------------------------------+ | |
143 * | |
1209 | 144 * They all start with a BRANCH for "\|" alternatives, even when there is only |
7 | 145 * one alternative. |
146 */ | |
147 | |
148 /* | |
149 * The opcodes are: | |
150 */ | |
151 | |
152 /* definition number opnd? meaning */ | |
153 #define END 0 /* End of program or NOMATCH operand. */ | |
154 #define BOL 1 /* Match "" at beginning of line. */ | |
155 #define EOL 2 /* Match "" at end of line. */ | |
156 #define BRANCH 3 /* node Match this alternative, or the | |
157 * next... */ | |
158 #define BACK 4 /* Match "", "next" ptr points backward. */ | |
159 #define EXACTLY 5 /* str Match this string. */ | |
160 #define NOTHING 6 /* Match empty string. */ | |
161 #define STAR 7 /* node Match this (simple) thing 0 or more | |
162 * times. */ | |
163 #define PLUS 8 /* node Match this (simple) thing 1 or more | |
164 * times. */ | |
165 #define MATCH 9 /* node match the operand zero-width */ | |
166 #define NOMATCH 10 /* node check for no match with operand */ | |
167 #define BEHIND 11 /* node look behind for a match with operand */ | |
168 #define NOBEHIND 12 /* node look behind for no match with operand */ | |
169 #define SUBPAT 13 /* node match the operand here */ | |
170 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and | |
171 * n times (\{m,n\}). */ | |
172 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */ | |
173 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */ | |
174 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE | |
175 * and BRACE_COMPLEX. */ | |
176 #define NEWL 18 /* Match line-break */ | |
177 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */ | |
178 | |
179 | |
180 /* character classes: 20-48 normal, 50-78 include a line-break */ | |
181 #define ADD_NL 30 | |
182 #define FIRST_NL ANY + ADD_NL | |
183 #define ANY 20 /* Match any one character. */ | |
184 #define ANYOF 21 /* str Match any character in this string. */ | |
185 #define ANYBUT 22 /* str Match any character not in this | |
186 * string. */ | |
187 #define IDENT 23 /* Match identifier char */ | |
188 #define SIDENT 24 /* Match identifier char but no digit */ | |
189 #define KWORD 25 /* Match keyword char */ | |
190 #define SKWORD 26 /* Match word char but no digit */ | |
191 #define FNAME 27 /* Match file name char */ | |
192 #define SFNAME 28 /* Match file name char but no digit */ | |
193 #define PRINT 29 /* Match printable char */ | |
194 #define SPRINT 30 /* Match printable char but no digit */ | |
195 #define WHITE 31 /* Match whitespace char */ | |
196 #define NWHITE 32 /* Match non-whitespace char */ | |
197 #define DIGIT 33 /* Match digit char */ | |
198 #define NDIGIT 34 /* Match non-digit char */ | |
199 #define HEX 35 /* Match hex char */ | |
200 #define NHEX 36 /* Match non-hex char */ | |
201 #define OCTAL 37 /* Match octal char */ | |
202 #define NOCTAL 38 /* Match non-octal char */ | |
203 #define WORD 39 /* Match word char */ | |
204 #define NWORD 40 /* Match non-word char */ | |
205 #define HEAD 41 /* Match head char */ | |
206 #define NHEAD 42 /* Match non-head char */ | |
207 #define ALPHA 43 /* Match alpha char */ | |
208 #define NALPHA 44 /* Match non-alpha char */ | |
209 #define LOWER 45 /* Match lowercase char */ | |
210 #define NLOWER 46 /* Match non-lowercase char */ | |
211 #define UPPER 47 /* Match uppercase char */ | |
212 #define NUPPER 48 /* Match non-uppercase char */ | |
213 #define LAST_NL NUPPER + ADD_NL | |
214 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL) | |
215 | |
216 #define MOPEN 80 /* -89 Mark this point in input as start of | |
217 * \( subexpr. MOPEN + 0 marks start of | |
218 * match. */ | |
219 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks | |
220 * end of match. */ | |
221 #define BACKREF 100 /* -109 node Match same string again \1-\9 */ | |
222 | |
223 #ifdef FEAT_SYN_HL | |
224 # define ZOPEN 110 /* -119 Mark this point in input as start of | |
225 * \z( subexpr. */ | |
226 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */ | |
227 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */ | |
228 #endif | |
229 | |
230 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */ | |
231 | |
232 #define NOPEN 150 /* Mark this point in input as start of | |
233 \%( subexpr. */ | |
234 #define NCLOSE 151 /* Analogous to NOPEN. */ | |
235 | |
236 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */ | |
237 #define RE_BOF 201 /* Match "" at beginning of file. */ | |
238 #define RE_EOF 202 /* Match "" at end of file. */ | |
239 #define CURSOR 203 /* Match location of cursor. */ | |
240 | |
241 #define RE_LNUM 204 /* nr cmp Match line number */ | |
242 #define RE_COL 205 /* nr cmp Match column number */ | |
243 #define RE_VCOL 206 /* nr cmp Match virtual column number */ | |
244 | |
639 | 245 #define RE_MARK 207 /* mark cmp Match mark position */ |
246 #define RE_VISUAL 208 /* Match Visual area */ | |
247 | |
7 | 248 /* |
249 * Magic characters have a special meaning, they don't match literally. | |
250 * Magic characters are negative. This separates them from literal characters | |
251 * (possibly multi-byte). Only ASCII characters can be Magic. | |
252 */ | |
253 #define Magic(x) ((int)(x) - 256) | |
254 #define un_Magic(x) ((x) + 256) | |
255 #define is_Magic(x) ((x) < 0) | |
256 | |
257 static int no_Magic __ARGS((int x)); | |
258 static int toggle_Magic __ARGS((int x)); | |
259 | |
260 static int | |
261 no_Magic(x) | |
262 int x; | |
263 { | |
264 if (is_Magic(x)) | |
265 return un_Magic(x); | |
266 return x; | |
267 } | |
268 | |
269 static int | |
270 toggle_Magic(x) | |
271 int x; | |
272 { | |
273 if (is_Magic(x)) | |
274 return un_Magic(x); | |
275 return Magic(x); | |
276 } | |
277 | |
278 /* | |
279 * The first byte of the regexp internal "program" is actually this magic | |
280 * number; the start node begins in the second byte. It's used to catch the | |
281 * most severe mutilation of the program by the caller. | |
282 */ | |
283 | |
284 #define REGMAGIC 0234 | |
285 | |
286 /* | |
287 * Opcode notes: | |
288 * | |
289 * BRANCH The set of branches constituting a single choice are hooked | |
290 * together with their "next" pointers, since precedence prevents | |
291 * anything being concatenated to any individual branch. The | |
292 * "next" pointer of the last BRANCH in a choice points to the | |
293 * thing following the whole choice. This is also where the | |
294 * final "next" pointer of each individual branch points; each | |
295 * branch starts with the operand node of a BRANCH node. | |
296 * | |
297 * BACK Normal "next" pointers all implicitly point forward; BACK | |
298 * exists to make loop structures possible. | |
299 * | |
300 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular | |
301 * BRANCH structures using BACK. Simple cases (one character | |
302 * per match) are implemented with STAR and PLUS for speed | |
303 * and to minimize recursive plunges. | |
304 * | |
305 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX | |
306 * node, and defines the min and max limits to be used for that | |
307 * node. | |
308 * | |
309 * MOPEN,MCLOSE ...are numbered at compile time. | |
310 * ZOPEN,ZCLOSE ...ditto | |
311 */ | |
312 | |
313 /* | |
314 * A node is one char of opcode followed by two chars of "next" pointer. | |
315 * "Next" pointers are stored as two 8-bit bytes, high order first. The | |
316 * value is a positive offset from the opcode of the node containing it. | |
317 * An operand, if any, simply follows the node. (Note that much of the | |
318 * code generation knows about this implicit relationship.) | |
319 * | |
320 * Using two bytes for the "next" pointer is vast overkill for most things, | |
321 * but allows patterns to get big without disasters. | |
322 */ | |
323 #define OP(p) ((int)*(p)) | |
324 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377)) | |
325 #define OPERAND(p) ((p) + 3) | |
326 /* Obtain an operand that was stored as four bytes, MSB first. */ | |
327 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \ | |
328 + ((long)(p)[5] << 8) + (long)(p)[6]) | |
329 /* Obtain a second operand stored as four bytes. */ | |
330 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4) | |
331 /* Obtain a second single-byte operand stored after a four bytes operand. */ | |
332 #define OPERAND_CMP(p) (p)[7] | |
333 | |
334 /* | |
335 * Utility definitions. | |
336 */ | |
337 #define UCHARAT(p) ((int)*(char_u *)(p)) | |
338 | |
339 /* Used for an error (down from) vim_regcomp(): give the error message, set | |
340 * rc_did_emsg and return NULL */ | |
653 | 341 #define EMSG_RET_NULL(m) return (EMSG(m), rc_did_emsg = TRUE, (void *)NULL) |
308 | 342 #define EMSG_RET_FAIL(m) return (EMSG(m), rc_did_emsg = TRUE, FAIL) |
4444 | 343 #define EMSG2_RET_NULL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, (void *)NULL) |
344 #define EMSG2_RET_FAIL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, FAIL) | |
345 #define EMSG_ONE_RET_NULL EMSG2_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL) | |
7 | 346 |
347 #define MAX_LIMIT (32767L << 16L) | |
348 | |
349 static int re_multi_type __ARGS((int)); | |
350 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n)); | |
351 static char_u *cstrchr __ARGS((char_u *, int)); | |
352 | |
4444 | 353 #ifdef BT_REGEXP_DUMP |
354 static void regdump __ARGS((char_u *, bt_regprog_T *)); | |
355 #endif | |
7 | 356 #ifdef DEBUG |
357 static char_u *regprop __ARGS((char_u *)); | |
358 #endif | |
359 | |
4444 | 360 static char_u e_missingbracket[] = N_("E769: Missing ] after %s["); |
361 static char_u e_unmatchedpp[] = N_("E53: Unmatched %s%%("); | |
362 static char_u e_unmatchedp[] = N_("E54: Unmatched %s("); | |
363 static char_u e_unmatchedpar[] = N_("E55: Unmatched %s)"); | |
364 | |
7 | 365 #define NOT_MULTI 0 |
366 #define MULTI_ONE 1 | |
367 #define MULTI_MULT 2 | |
368 /* | |
369 * Return NOT_MULTI if c is not a "multi" operator. | |
370 * Return MULTI_ONE if c is a single "multi" operator. | |
371 * Return MULTI_MULT if c is a multi "multi" operator. | |
372 */ | |
373 static int | |
374 re_multi_type(c) | |
375 int c; | |
376 { | |
377 if (c == Magic('@') || c == Magic('=') || c == Magic('?')) | |
378 return MULTI_ONE; | |
379 if (c == Magic('*') || c == Magic('+') || c == Magic('{')) | |
380 return MULTI_MULT; | |
381 return NOT_MULTI; | |
382 } | |
383 | |
384 /* | |
385 * Flags to be passed up and down. | |
386 */ | |
387 #define HASWIDTH 0x1 /* Known never to match null string. */ | |
388 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */ | |
389 #define SPSTART 0x4 /* Starts with * or +. */ | |
390 #define HASNL 0x8 /* Contains some \n. */ | |
391 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */ | |
392 #define WORST 0 /* Worst case. */ | |
393 | |
394 /* | |
395 * When regcode is set to this value, code is not emitted and size is computed | |
396 * instead. | |
397 */ | |
398 #define JUST_CALC_SIZE ((char_u *) -1) | |
399 | |
359 | 400 static char_u *reg_prev_sub = NULL; |
401 | |
7 | 402 /* |
403 * REGEXP_INRANGE contains all characters which are always special in a [] | |
404 * range after '\'. | |
405 * REGEXP_ABBR contains all characters which act as abbreviations after '\'. | |
406 * These are: | |
407 * \n - New line (NL). | |
408 * \r - Carriage Return (CR). | |
409 * \t - Tab (TAB). | |
410 * \e - Escape (ESC). | |
411 * \b - Backspace (Ctrl_H). | |
24 | 412 * \d - Character code in decimal, eg \d123 |
413 * \o - Character code in octal, eg \o80 | |
414 * \x - Character code in hex, eg \x4a | |
415 * \u - Multibyte character code, eg \u20ac | |
416 * \U - Long multibyte character code, eg \U12345678 | |
7 | 417 */ |
418 static char_u REGEXP_INRANGE[] = "]^-n\\"; | |
24 | 419 static char_u REGEXP_ABBR[] = "nrtebdoxuU"; |
7 | 420 |
421 static int backslash_trans __ARGS((int c)); | |
167 | 422 static int get_char_class __ARGS((char_u **pp)); |
423 static int get_equi_class __ARGS((char_u **pp)); | |
424 static void reg_equi_class __ARGS((int c)); | |
425 static int get_coll_element __ARGS((char_u **pp)); | |
7 | 426 static char_u *skip_anyof __ARGS((char_u *p)); |
427 static void init_class_tab __ARGS((void)); | |
428 | |
429 /* | |
430 * Translate '\x' to its control character, except "\n", which is Magic. | |
431 */ | |
432 static int | |
433 backslash_trans(c) | |
434 int c; | |
435 { | |
436 switch (c) | |
437 { | |
438 case 'r': return CAR; | |
439 case 't': return TAB; | |
440 case 'e': return ESC; | |
441 case 'b': return BS; | |
442 } | |
443 return c; | |
444 } | |
445 | |
446 /* | |
167 | 447 * Check for a character class name "[:name:]". "pp" points to the '['. |
7 | 448 * Returns one of the CLASS_ items. CLASS_NONE means that no item was |
449 * recognized. Otherwise "pp" is advanced to after the item. | |
450 */ | |
451 static int | |
167 | 452 get_char_class(pp) |
7 | 453 char_u **pp; |
454 { | |
455 static const char *(class_names[]) = | |
456 { | |
457 "alnum:]", | |
458 #define CLASS_ALNUM 0 | |
459 "alpha:]", | |
460 #define CLASS_ALPHA 1 | |
461 "blank:]", | |
462 #define CLASS_BLANK 2 | |
463 "cntrl:]", | |
464 #define CLASS_CNTRL 3 | |
465 "digit:]", | |
466 #define CLASS_DIGIT 4 | |
467 "graph:]", | |
468 #define CLASS_GRAPH 5 | |
469 "lower:]", | |
470 #define CLASS_LOWER 6 | |
471 "print:]", | |
472 #define CLASS_PRINT 7 | |
473 "punct:]", | |
474 #define CLASS_PUNCT 8 | |
475 "space:]", | |
476 #define CLASS_SPACE 9 | |
477 "upper:]", | |
478 #define CLASS_UPPER 10 | |
479 "xdigit:]", | |
480 #define CLASS_XDIGIT 11 | |
481 "tab:]", | |
482 #define CLASS_TAB 12 | |
483 "return:]", | |
484 #define CLASS_RETURN 13 | |
485 "backspace:]", | |
486 #define CLASS_BACKSPACE 14 | |
487 "escape:]", | |
488 #define CLASS_ESCAPE 15 | |
489 }; | |
490 #define CLASS_NONE 99 | |
491 int i; | |
492 | |
493 if ((*pp)[1] == ':') | |
494 { | |
1877 | 495 for (i = 0; i < (int)(sizeof(class_names) / sizeof(*class_names)); ++i) |
7 | 496 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0) |
497 { | |
498 *pp += STRLEN(class_names[i]) + 2; | |
499 return i; | |
500 } | |
501 } | |
502 return CLASS_NONE; | |
503 } | |
504 | |
505 /* | |
506 * Specific version of character class functions. | |
507 * Using a table to keep this fast. | |
508 */ | |
509 static short class_tab[256]; | |
510 | |
511 #define RI_DIGIT 0x01 | |
512 #define RI_HEX 0x02 | |
513 #define RI_OCTAL 0x04 | |
514 #define RI_WORD 0x08 | |
515 #define RI_HEAD 0x10 | |
516 #define RI_ALPHA 0x20 | |
517 #define RI_LOWER 0x40 | |
518 #define RI_UPPER 0x80 | |
519 #define RI_WHITE 0x100 | |
520 | |
521 static void | |
522 init_class_tab() | |
523 { | |
524 int i; | |
525 static int done = FALSE; | |
526 | |
527 if (done) | |
528 return; | |
529 | |
530 for (i = 0; i < 256; ++i) | |
531 { | |
532 if (i >= '0' && i <= '7') | |
533 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD; | |
534 else if (i >= '8' && i <= '9') | |
535 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD; | |
536 else if (i >= 'a' && i <= 'f') | |
537 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER; | |
538 #ifdef EBCDIC | |
539 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r') | |
540 || (i >= 's' && i <= 'z')) | |
541 #else | |
542 else if (i >= 'g' && i <= 'z') | |
543 #endif | |
544 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER; | |
545 else if (i >= 'A' && i <= 'F') | |
546 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER; | |
547 #ifdef EBCDIC | |
548 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R') | |
549 || (i >= 'S' && i <= 'Z')) | |
550 #else | |
551 else if (i >= 'G' && i <= 'Z') | |
552 #endif | |
553 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER; | |
554 else if (i == '_') | |
555 class_tab[i] = RI_WORD + RI_HEAD; | |
556 else | |
557 class_tab[i] = 0; | |
558 } | |
559 class_tab[' '] |= RI_WHITE; | |
560 class_tab['\t'] |= RI_WHITE; | |
561 done = TRUE; | |
562 } | |
563 | |
564 #ifdef FEAT_MBYTE | |
565 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT)) | |
566 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX)) | |
567 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL)) | |
568 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD)) | |
569 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD)) | |
570 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA)) | |
571 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER)) | |
572 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER)) | |
573 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE)) | |
574 #else | |
575 # define ri_digit(c) (class_tab[c] & RI_DIGIT) | |
576 # define ri_hex(c) (class_tab[c] & RI_HEX) | |
577 # define ri_octal(c) (class_tab[c] & RI_OCTAL) | |
578 # define ri_word(c) (class_tab[c] & RI_WORD) | |
579 # define ri_head(c) (class_tab[c] & RI_HEAD) | |
580 # define ri_alpha(c) (class_tab[c] & RI_ALPHA) | |
581 # define ri_lower(c) (class_tab[c] & RI_LOWER) | |
582 # define ri_upper(c) (class_tab[c] & RI_UPPER) | |
583 # define ri_white(c) (class_tab[c] & RI_WHITE) | |
584 #endif | |
585 | |
586 /* flags for regflags */ | |
587 #define RF_ICASE 1 /* ignore case */ | |
588 #define RF_NOICASE 2 /* don't ignore case */ | |
589 #define RF_HASNL 4 /* can match a NL */ | |
590 #define RF_ICOMBINE 8 /* ignore combining characters */ | |
591 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */ | |
592 | |
593 /* | |
594 * Global work variables for vim_regcomp(). | |
595 */ | |
596 | |
597 static char_u *regparse; /* Input-scan pointer. */ | |
598 static int prevchr_len; /* byte length of previous char */ | |
599 static int num_complex_braces; /* Complex \{...} count */ | |
600 static int regnpar; /* () count. */ | |
601 #ifdef FEAT_SYN_HL | |
602 static int regnzpar; /* \z() count. */ | |
603 static int re_has_z; /* \z item detected */ | |
604 #endif | |
605 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */ | |
606 static long regsize; /* Code size. */ | |
2010 | 607 static int reg_toolong; /* TRUE when offset out of range */ |
7 | 608 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */ |
609 static unsigned regflags; /* RF_ flags for prog */ | |
610 static long brace_min[10]; /* Minimums for complex brace repeats */ | |
611 static long brace_max[10]; /* Maximums for complex brace repeats */ | |
612 static int brace_count[10]; /* Current counts for complex brace repeats */ | |
613 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
614 static int had_eol; /* TRUE when EOL found by vim_regcomp() */ | |
615 #endif | |
616 static int one_exactly = FALSE; /* only do one char for EXACTLY */ | |
617 | |
618 static int reg_magic; /* magicness of the pattern: */ | |
619 #define MAGIC_NONE 1 /* "\V" very unmagic */ | |
620 #define MAGIC_OFF 2 /* "\M" or 'magic' off */ | |
621 #define MAGIC_ON 3 /* "\m" or 'magic' */ | |
622 #define MAGIC_ALL 4 /* "\v" very magic */ | |
623 | |
624 static int reg_string; /* matching with a string instead of a buffer | |
625 line */ | |
481 | 626 static int reg_strict; /* "[abc" is illegal */ |
7 | 627 |
628 /* | |
629 * META contains all characters that may be magic, except '^' and '$'. | |
630 */ | |
631 | |
632 #ifdef EBCDIC | |
633 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~"; | |
634 #else | |
635 /* META[] is used often enough to justify turning it into a table. */ | |
636 static char_u META_flags[] = { | |
637 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, | |
638 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, | |
639 /* % & ( ) * + . */ | |
640 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, | |
641 /* 1 2 3 4 5 6 7 8 9 < = > ? */ | |
642 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, | |
643 /* @ A C D F H I K L M O */ | |
644 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, | |
645 /* P S U V W X Z [ _ */ | |
646 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1, | |
647 /* a c d f h i k l m n o */ | |
648 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, | |
649 /* p s u v w x z { | ~ */ | |
650 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1 | |
651 }; | |
652 #endif | |
653 | |
4444 | 654 static int curchr; /* currently parsed character */ |
655 /* Previous character. Note: prevchr is sometimes -1 when we are not at the | |
656 * start, eg in /[ ^I]^ the pattern was never found even if it existed, | |
657 * because ^ was taken to be magic -- webb */ | |
658 static int prevchr; | |
659 static int prevprevchr; /* previous-previous character */ | |
660 static int nextchr; /* used for ungetchr() */ | |
7 | 661 |
662 /* arguments for reg() */ | |
663 #define REG_NOPAREN 0 /* toplevel reg() */ | |
664 #define REG_PAREN 1 /* \(\) */ | |
665 #define REG_ZPAREN 2 /* \z(\) */ | |
666 #define REG_NPAREN 3 /* \%(\) */ | |
667 | |
4679
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
668 typedef struct |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
669 { |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
670 char_u *regparse; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
671 int prevchr_len; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
672 int curchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
673 int prevchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
674 int prevprevchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
675 int nextchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
676 int at_start; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
677 int prev_at_start; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
678 int regnpar; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
679 } parse_state_T; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
680 |
7 | 681 /* |
682 * Forward declarations for vim_regcomp()'s friends. | |
683 */ | |
684 static void initchr __ARGS((char_u *)); | |
4679
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
685 static void save_parse_state __ARGS((parse_state_T *ps)); |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
686 static void restore_parse_state __ARGS((parse_state_T *ps)); |
7 | 687 static int getchr __ARGS((void)); |
688 static void skipchr_keepstart __ARGS((void)); | |
689 static int peekchr __ARGS((void)); | |
690 static void skipchr __ARGS((void)); | |
691 static void ungetchr __ARGS((void)); | |
24 | 692 static int gethexchrs __ARGS((int maxinputlen)); |
693 static int getoctchrs __ARGS((void)); | |
694 static int getdecchrs __ARGS((void)); | |
695 static int coll_get_char __ARGS((void)); | |
7 | 696 static void regcomp_start __ARGS((char_u *expr, int flags)); |
697 static char_u *reg __ARGS((int, int *)); | |
698 static char_u *regbranch __ARGS((int *flagp)); | |
699 static char_u *regconcat __ARGS((int *flagp)); | |
700 static char_u *regpiece __ARGS((int *)); | |
701 static char_u *regatom __ARGS((int *)); | |
702 static char_u *regnode __ARGS((int)); | |
714 | 703 #ifdef FEAT_MBYTE |
704 static int use_multibytecode __ARGS((int c)); | |
705 #endif | |
7 | 706 static int prog_magic_wrong __ARGS((void)); |
707 static char_u *regnext __ARGS((char_u *)); | |
708 static void regc __ARGS((int b)); | |
709 #ifdef FEAT_MBYTE | |
710 static void regmbc __ARGS((int c)); | |
2974 | 711 # define REGMBC(x) regmbc(x); |
712 # define CASEMBC(x) case x: | |
167 | 713 #else |
714 # define regmbc(c) regc(c) | |
2974 | 715 # define REGMBC(x) |
716 # define CASEMBC(x) | |
7 | 717 #endif |
718 static void reginsert __ARGS((int, char_u *)); | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
719 static void reginsert_nr __ARGS((int op, long val, char_u *opnd)); |
7 | 720 static void reginsert_limits __ARGS((int, long, long, char_u *)); |
721 static char_u *re_put_long __ARGS((char_u *pr, long_u val)); | |
722 static int read_limits __ARGS((long *, long *)); | |
723 static void regtail __ARGS((char_u *, char_u *)); | |
724 static void regoptail __ARGS((char_u *, char_u *)); | |
725 | |
4444 | 726 static regengine_T bt_regengine; |
727 static regengine_T nfa_regengine; | |
728 | |
7 | 729 /* |
730 * Return TRUE if compiled regular expression "prog" can match a line break. | |
731 */ | |
732 int | |
733 re_multiline(prog) | |
734 regprog_T *prog; | |
735 { | |
736 return (prog->regflags & RF_HASNL); | |
737 } | |
738 | |
739 /* | |
740 * Return TRUE if compiled regular expression "prog" looks before the start | |
741 * position (pattern contains "\@<=" or "\@<!"). | |
742 */ | |
743 int | |
744 re_lookbehind(prog) | |
745 regprog_T *prog; | |
746 { | |
747 return (prog->regflags & RF_LOOKBH); | |
748 } | |
749 | |
750 /* | |
167 | 751 * Check for an equivalence class name "[=a=]". "pp" points to the '['. |
752 * Returns a character representing the class. Zero means that no item was | |
753 * recognized. Otherwise "pp" is advanced to after the item. | |
754 */ | |
755 static int | |
756 get_equi_class(pp) | |
757 char_u **pp; | |
758 { | |
759 int c; | |
760 int l = 1; | |
761 char_u *p = *pp; | |
762 | |
763 if (p[1] == '=') | |
764 { | |
765 #ifdef FEAT_MBYTE | |
766 if (has_mbyte) | |
474 | 767 l = (*mb_ptr2len)(p + 2); |
167 | 768 #endif |
769 if (p[l + 2] == '=' && p[l + 3] == ']') | |
770 { | |
771 #ifdef FEAT_MBYTE | |
772 if (has_mbyte) | |
773 c = mb_ptr2char(p + 2); | |
774 else | |
775 #endif | |
776 c = p[2]; | |
777 *pp += l + 4; | |
778 return c; | |
779 } | |
780 } | |
781 return 0; | |
782 } | |
783 | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
784 #ifdef EBCDIC |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
785 /* |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
786 * 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
|
787 */ |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
788 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
|
789 "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
|
790 "C\x68", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
791 "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
|
792 "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
|
793 "N\x69", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
794 "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
|
795 "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
|
796 "Y\xBA", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
797 "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
|
798 "c\x48", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
799 "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
|
800 "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
|
801 "n\x49", |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
802 "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
|
803 "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
|
804 "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
|
805 }; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
806 #endif |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
807 |
167 | 808 /* |
809 * Produce the bytes for equivalence class "c". | |
810 * Currently only handles latin1, latin9 and utf-8. | |
4444 | 811 * NOTE: When changing this function, also change nfa_emit_equi_class() |
167 | 812 */ |
813 static void | |
814 reg_equi_class(c) | |
815 int c; | |
816 { | |
817 #ifdef FEAT_MBYTE | |
818 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0 | |
492 | 819 || STRCMP(p_enc, "iso-8859-15") == 0) |
167 | 820 #endif |
821 { | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
822 #ifdef EBCDIC |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
823 int i; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
824 |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
825 /* 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
|
826 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
|
827 { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
828 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
|
829 { |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
830 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
|
831 |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
832 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
|
833 regmbc(*p++); |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
834 return; |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
835 } |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
836 } |
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
837 #else |
167 | 838 switch (c) |
839 { | |
236 | 840 case 'A': case '\300': case '\301': case '\302': |
2974 | 841 CASEMBC(0x100) CASEMBC(0x102) CASEMBC(0x104) CASEMBC(0x1cd) |
842 CASEMBC(0x1de) CASEMBC(0x1e0) CASEMBC(0x1ea2) | |
236 | 843 case '\303': case '\304': case '\305': |
844 regmbc('A'); regmbc('\300'); regmbc('\301'); | |
845 regmbc('\302'); regmbc('\303'); regmbc('\304'); | |
846 regmbc('\305'); | |
2974 | 847 REGMBC(0x100) REGMBC(0x102) REGMBC(0x104) |
848 REGMBC(0x1cd) REGMBC(0x1de) REGMBC(0x1e0) | |
849 REGMBC(0x1ea2) | |
850 return; | |
851 case 'B': CASEMBC(0x1e02) CASEMBC(0x1e06) | |
852 regmbc('B'); REGMBC(0x1e02) REGMBC(0x1e06) | |
167 | 853 return; |
236 | 854 case 'C': case '\307': |
2974 | 855 CASEMBC(0x106) CASEMBC(0x108) CASEMBC(0x10a) CASEMBC(0x10c) |
236 | 856 regmbc('C'); regmbc('\307'); |
2974 | 857 REGMBC(0x106) REGMBC(0x108) REGMBC(0x10a) |
858 REGMBC(0x10c) | |
859 return; | |
860 case 'D': CASEMBC(0x10e) CASEMBC(0x110) CASEMBC(0x1e0a) | |
861 CASEMBC(0x1e0e) CASEMBC(0x1e10) | |
862 regmbc('D'); REGMBC(0x10e) REGMBC(0x110) | |
863 REGMBC(0x1e0a) REGMBC(0x1e0e) REGMBC(0x1e10) | |
167 | 864 return; |
236 | 865 case 'E': case '\310': case '\311': case '\312': case '\313': |
2974 | 866 CASEMBC(0x112) CASEMBC(0x114) CASEMBC(0x116) CASEMBC(0x118) |
867 CASEMBC(0x11a) CASEMBC(0x1eba) CASEMBC(0x1ebc) | |
236 | 868 regmbc('E'); regmbc('\310'); regmbc('\311'); |
869 regmbc('\312'); regmbc('\313'); | |
2974 | 870 REGMBC(0x112) REGMBC(0x114) REGMBC(0x116) |
871 REGMBC(0x118) REGMBC(0x11a) REGMBC(0x1eba) | |
872 REGMBC(0x1ebc) | |
873 return; | |
874 case 'F': CASEMBC(0x1e1e) | |
875 regmbc('F'); REGMBC(0x1e1e) | |
876 return; | |
877 case 'G': CASEMBC(0x11c) CASEMBC(0x11e) CASEMBC(0x120) | |
878 CASEMBC(0x122) CASEMBC(0x1e4) CASEMBC(0x1e6) CASEMBC(0x1f4) | |
879 CASEMBC(0x1e20) | |
880 regmbc('G'); REGMBC(0x11c) REGMBC(0x11e) | |
881 REGMBC(0x120) REGMBC(0x122) REGMBC(0x1e4) | |
882 REGMBC(0x1e6) REGMBC(0x1f4) REGMBC(0x1e20) | |
883 return; | |
884 case 'H': CASEMBC(0x124) CASEMBC(0x126) CASEMBC(0x1e22) | |
885 CASEMBC(0x1e26) CASEMBC(0x1e28) | |
886 regmbc('H'); REGMBC(0x124) REGMBC(0x126) | |
887 REGMBC(0x1e22) REGMBC(0x1e26) REGMBC(0x1e28) | |
167 | 888 return; |
236 | 889 case 'I': case '\314': case '\315': case '\316': case '\317': |
2974 | 890 CASEMBC(0x128) CASEMBC(0x12a) CASEMBC(0x12c) CASEMBC(0x12e) |
891 CASEMBC(0x130) CASEMBC(0x1cf) CASEMBC(0x1ec8) | |
236 | 892 regmbc('I'); regmbc('\314'); regmbc('\315'); |
893 regmbc('\316'); regmbc('\317'); | |
2974 | 894 REGMBC(0x128) REGMBC(0x12a) REGMBC(0x12c) |
895 REGMBC(0x12e) REGMBC(0x130) REGMBC(0x1cf) | |
896 REGMBC(0x1ec8) | |
897 return; | |
898 case 'J': CASEMBC(0x134) | |
899 regmbc('J'); REGMBC(0x134) | |
900 return; | |
901 case 'K': CASEMBC(0x136) CASEMBC(0x1e8) CASEMBC(0x1e30) | |
902 CASEMBC(0x1e34) | |
903 regmbc('K'); REGMBC(0x136) REGMBC(0x1e8) | |
904 REGMBC(0x1e30) REGMBC(0x1e34) | |
905 return; | |
906 case 'L': CASEMBC(0x139) CASEMBC(0x13b) CASEMBC(0x13d) | |
907 CASEMBC(0x13f) CASEMBC(0x141) CASEMBC(0x1e3a) | |
908 regmbc('L'); REGMBC(0x139) REGMBC(0x13b) | |
909 REGMBC(0x13d) REGMBC(0x13f) REGMBC(0x141) | |
910 REGMBC(0x1e3a) | |
911 return; | |
912 case 'M': CASEMBC(0x1e3e) CASEMBC(0x1e40) | |
913 regmbc('M'); REGMBC(0x1e3e) REGMBC(0x1e40) | |
167 | 914 return; |
236 | 915 case 'N': case '\321': |
2974 | 916 CASEMBC(0x143) CASEMBC(0x145) CASEMBC(0x147) CASEMBC(0x1e44) |
917 CASEMBC(0x1e48) | |
236 | 918 regmbc('N'); regmbc('\321'); |
2974 | 919 REGMBC(0x143) REGMBC(0x145) REGMBC(0x147) |
920 REGMBC(0x1e44) REGMBC(0x1e48) | |
167 | 921 return; |
236 | 922 case 'O': case '\322': case '\323': case '\324': case '\325': |
2974 | 923 case '\326': case '\330': |
924 CASEMBC(0x14c) CASEMBC(0x14e) CASEMBC(0x150) CASEMBC(0x1a0) | |
925 CASEMBC(0x1d1) CASEMBC(0x1ea) CASEMBC(0x1ec) CASEMBC(0x1ece) | |
236 | 926 regmbc('O'); regmbc('\322'); regmbc('\323'); |
927 regmbc('\324'); regmbc('\325'); regmbc('\326'); | |
2974 | 928 regmbc('\330'); |
929 REGMBC(0x14c) REGMBC(0x14e) REGMBC(0x150) | |
930 REGMBC(0x1a0) REGMBC(0x1d1) REGMBC(0x1ea) | |
931 REGMBC(0x1ec) REGMBC(0x1ece) | |
932 return; | |
933 case 'P': case 0x1e54: case 0x1e56: | |
934 regmbc('P'); REGMBC(0x1e54) REGMBC(0x1e56) | |
935 return; | |
936 case 'R': CASEMBC(0x154) CASEMBC(0x156) CASEMBC(0x158) | |
937 CASEMBC(0x1e58) CASEMBC(0x1e5e) | |
938 regmbc('R'); REGMBC(0x154) REGMBC(0x156) REGMBC(0x158) | |
939 REGMBC(0x1e58) REGMBC(0x1e5e) | |
940 return; | |
941 case 'S': CASEMBC(0x15a) CASEMBC(0x15c) CASEMBC(0x15e) | |
942 CASEMBC(0x160) CASEMBC(0x1e60) | |
943 regmbc('S'); REGMBC(0x15a) REGMBC(0x15c) | |
944 REGMBC(0x15e) REGMBC(0x160) REGMBC(0x1e60) | |
945 return; | |
946 case 'T': CASEMBC(0x162) CASEMBC(0x164) CASEMBC(0x166) | |
947 CASEMBC(0x1e6a) CASEMBC(0x1e6e) | |
948 regmbc('T'); REGMBC(0x162) REGMBC(0x164) | |
949 REGMBC(0x166) REGMBC(0x1e6a) REGMBC(0x1e6e) | |
167 | 950 return; |
236 | 951 case 'U': case '\331': case '\332': case '\333': case '\334': |
2974 | 952 CASEMBC(0x168) CASEMBC(0x16a) CASEMBC(0x16c) CASEMBC(0x16e) |
953 CASEMBC(0x170) CASEMBC(0x172) CASEMBC(0x1af) CASEMBC(0x1d3) | |
954 CASEMBC(0x1ee6) | |
236 | 955 regmbc('U'); regmbc('\331'); regmbc('\332'); |
956 regmbc('\333'); regmbc('\334'); | |
2974 | 957 REGMBC(0x168) REGMBC(0x16a) REGMBC(0x16c) |
958 REGMBC(0x16e) REGMBC(0x170) REGMBC(0x172) | |
959 REGMBC(0x1af) REGMBC(0x1d3) REGMBC(0x1ee6) | |
960 return; | |
961 case 'V': CASEMBC(0x1e7c) | |
962 regmbc('V'); REGMBC(0x1e7c) | |
963 return; | |
964 case 'W': CASEMBC(0x174) CASEMBC(0x1e80) CASEMBC(0x1e82) | |
965 CASEMBC(0x1e84) CASEMBC(0x1e86) | |
966 regmbc('W'); REGMBC(0x174) REGMBC(0x1e80) | |
967 REGMBC(0x1e82) REGMBC(0x1e84) REGMBC(0x1e86) | |
968 return; | |
969 case 'X': CASEMBC(0x1e8a) CASEMBC(0x1e8c) | |
970 regmbc('X'); REGMBC(0x1e8a) REGMBC(0x1e8c) | |
167 | 971 return; |
236 | 972 case 'Y': case '\335': |
2974 | 973 CASEMBC(0x176) CASEMBC(0x178) CASEMBC(0x1e8e) CASEMBC(0x1ef2) |
974 CASEMBC(0x1ef6) CASEMBC(0x1ef8) | |
236 | 975 regmbc('Y'); regmbc('\335'); |
2974 | 976 REGMBC(0x176) REGMBC(0x178) REGMBC(0x1e8e) |
977 REGMBC(0x1ef2) REGMBC(0x1ef6) REGMBC(0x1ef8) | |
978 return; | |
979 case 'Z': CASEMBC(0x179) CASEMBC(0x17b) CASEMBC(0x17d) | |
980 CASEMBC(0x1b5) CASEMBC(0x1e90) CASEMBC(0x1e94) | |
981 regmbc('Z'); REGMBC(0x179) REGMBC(0x17b) | |
982 REGMBC(0x17d) REGMBC(0x1b5) REGMBC(0x1e90) | |
983 REGMBC(0x1e94) | |
167 | 984 return; |
236 | 985 case 'a': case '\340': case '\341': case '\342': |
986 case '\343': case '\344': case '\345': | |
2974 | 987 CASEMBC(0x101) CASEMBC(0x103) CASEMBC(0x105) CASEMBC(0x1ce) |
988 CASEMBC(0x1df) CASEMBC(0x1e1) CASEMBC(0x1ea3) | |
236 | 989 regmbc('a'); regmbc('\340'); regmbc('\341'); |
990 regmbc('\342'); regmbc('\343'); regmbc('\344'); | |
991 regmbc('\345'); | |
2974 | 992 REGMBC(0x101) REGMBC(0x103) REGMBC(0x105) |
993 REGMBC(0x1ce) REGMBC(0x1df) REGMBC(0x1e1) | |
994 REGMBC(0x1ea3) | |
995 return; | |
996 case 'b': CASEMBC(0x1e03) CASEMBC(0x1e07) | |
997 regmbc('b'); REGMBC(0x1e03) REGMBC(0x1e07) | |
167 | 998 return; |
236 | 999 case 'c': case '\347': |
2974 | 1000 CASEMBC(0x107) CASEMBC(0x109) CASEMBC(0x10b) CASEMBC(0x10d) |
236 | 1001 regmbc('c'); regmbc('\347'); |
2974 | 1002 REGMBC(0x107) REGMBC(0x109) REGMBC(0x10b) |
1003 REGMBC(0x10d) | |
1004 return; | |
1005 case 'd': CASEMBC(0x10f) CASEMBC(0x111) CASEMBC(0x1d0b) | |
1006 CASEMBC(0x1e11) | |
1007 regmbc('d'); REGMBC(0x10f) REGMBC(0x111) | |
1008 REGMBC(0x1e0b) REGMBC(0x01e0f) REGMBC(0x1e11) | |
167 | 1009 return; |
236 | 1010 case 'e': case '\350': case '\351': case '\352': case '\353': |
2974 | 1011 CASEMBC(0x113) CASEMBC(0x115) CASEMBC(0x117) CASEMBC(0x119) |
1012 CASEMBC(0x11b) CASEMBC(0x1ebb) CASEMBC(0x1ebd) | |
236 | 1013 regmbc('e'); regmbc('\350'); regmbc('\351'); |
1014 regmbc('\352'); regmbc('\353'); | |
2974 | 1015 REGMBC(0x113) REGMBC(0x115) REGMBC(0x117) |
1016 REGMBC(0x119) REGMBC(0x11b) REGMBC(0x1ebb) | |
1017 REGMBC(0x1ebd) | |
1018 return; | |
1019 case 'f': CASEMBC(0x1e1f) | |
1020 regmbc('f'); REGMBC(0x1e1f) | |
1021 return; | |
1022 case 'g': CASEMBC(0x11d) CASEMBC(0x11f) CASEMBC(0x121) | |
1023 CASEMBC(0x123) CASEMBC(0x1e5) CASEMBC(0x1e7) CASEMBC(0x1f5) | |
1024 CASEMBC(0x1e21) | |
1025 regmbc('g'); REGMBC(0x11d) REGMBC(0x11f) | |
1026 REGMBC(0x121) REGMBC(0x123) REGMBC(0x1e5) | |
1027 REGMBC(0x1e7) REGMBC(0x1f5) REGMBC(0x1e21) | |
1028 return; | |
1029 case 'h': CASEMBC(0x125) CASEMBC(0x127) CASEMBC(0x1e23) | |
1030 CASEMBC(0x1e27) CASEMBC(0x1e29) CASEMBC(0x1e96) | |
1031 regmbc('h'); REGMBC(0x125) REGMBC(0x127) | |
1032 REGMBC(0x1e23) REGMBC(0x1e27) REGMBC(0x1e29) | |
1033 REGMBC(0x1e96) | |
167 | 1034 return; |
236 | 1035 case 'i': case '\354': case '\355': case '\356': case '\357': |
2974 | 1036 CASEMBC(0x129) CASEMBC(0x12b) CASEMBC(0x12d) CASEMBC(0x12f) |
1037 CASEMBC(0x1d0) CASEMBC(0x1ec9) | |
236 | 1038 regmbc('i'); regmbc('\354'); regmbc('\355'); |
1039 regmbc('\356'); regmbc('\357'); | |
2974 | 1040 REGMBC(0x129) REGMBC(0x12b) REGMBC(0x12d) |
1041 REGMBC(0x12f) REGMBC(0x1d0) REGMBC(0x1ec9) | |
1042 return; | |
1043 case 'j': CASEMBC(0x135) CASEMBC(0x1f0) | |
1044 regmbc('j'); REGMBC(0x135) REGMBC(0x1f0) | |
1045 return; | |
1046 case 'k': CASEMBC(0x137) CASEMBC(0x1e9) CASEMBC(0x1e31) | |
1047 CASEMBC(0x1e35) | |
1048 regmbc('k'); REGMBC(0x137) REGMBC(0x1e9) | |
1049 REGMBC(0x1e31) REGMBC(0x1e35) | |
1050 return; | |
1051 case 'l': CASEMBC(0x13a) CASEMBC(0x13c) CASEMBC(0x13e) | |
1052 CASEMBC(0x140) CASEMBC(0x142) CASEMBC(0x1e3b) | |
1053 regmbc('l'); REGMBC(0x13a) REGMBC(0x13c) | |
1054 REGMBC(0x13e) REGMBC(0x140) REGMBC(0x142) | |
1055 REGMBC(0x1e3b) | |
1056 return; | |
1057 case 'm': CASEMBC(0x1e3f) CASEMBC(0x1e41) | |
1058 regmbc('m'); REGMBC(0x1e3f) REGMBC(0x1e41) | |
167 | 1059 return; |
236 | 1060 case 'n': case '\361': |
2974 | 1061 CASEMBC(0x144) CASEMBC(0x146) CASEMBC(0x148) CASEMBC(0x149) |
1062 CASEMBC(0x1e45) CASEMBC(0x1e49) | |
236 | 1063 regmbc('n'); regmbc('\361'); |
2974 | 1064 REGMBC(0x144) REGMBC(0x146) REGMBC(0x148) |
1065 REGMBC(0x149) REGMBC(0x1e45) REGMBC(0x1e49) | |
167 | 1066 return; |
236 | 1067 case 'o': case '\362': case '\363': case '\364': case '\365': |
2974 | 1068 case '\366': case '\370': |
1069 CASEMBC(0x14d) CASEMBC(0x14f) CASEMBC(0x151) CASEMBC(0x1a1) | |
1070 CASEMBC(0x1d2) CASEMBC(0x1eb) CASEMBC(0x1ed) CASEMBC(0x1ecf) | |
236 | 1071 regmbc('o'); regmbc('\362'); regmbc('\363'); |
1072 regmbc('\364'); regmbc('\365'); regmbc('\366'); | |
2974 | 1073 regmbc('\370'); |
1074 REGMBC(0x14d) REGMBC(0x14f) REGMBC(0x151) | |
1075 REGMBC(0x1a1) REGMBC(0x1d2) REGMBC(0x1eb) | |
1076 REGMBC(0x1ed) REGMBC(0x1ecf) | |
1077 return; | |
1078 case 'p': CASEMBC(0x1e55) CASEMBC(0x1e57) | |
1079 regmbc('p'); REGMBC(0x1e55) REGMBC(0x1e57) | |
1080 return; | |
1081 case 'r': CASEMBC(0x155) CASEMBC(0x157) CASEMBC(0x159) | |
1082 CASEMBC(0x1e59) CASEMBC(0x1e5f) | |
1083 regmbc('r'); REGMBC(0x155) REGMBC(0x157) REGMBC(0x159) | |
1084 REGMBC(0x1e59) REGMBC(0x1e5f) | |
1085 return; | |
1086 case 's': CASEMBC(0x15b) CASEMBC(0x15d) CASEMBC(0x15f) | |
1087 CASEMBC(0x161) CASEMBC(0x1e61) | |
1088 regmbc('s'); REGMBC(0x15b) REGMBC(0x15d) | |
1089 REGMBC(0x15f) REGMBC(0x161) REGMBC(0x1e61) | |
1090 return; | |
1091 case 't': CASEMBC(0x163) CASEMBC(0x165) CASEMBC(0x167) | |
1092 CASEMBC(0x1e6b) CASEMBC(0x1e6f) CASEMBC(0x1e97) | |
1093 regmbc('t'); REGMBC(0x163) REGMBC(0x165) REGMBC(0x167) | |
1094 REGMBC(0x1e6b) REGMBC(0x1e6f) REGMBC(0x1e97) | |
167 | 1095 return; |
236 | 1096 case 'u': case '\371': case '\372': case '\373': case '\374': |
2974 | 1097 CASEMBC(0x169) CASEMBC(0x16b) CASEMBC(0x16d) CASEMBC(0x16f) |
1098 CASEMBC(0x171) CASEMBC(0x173) CASEMBC(0x1b0) CASEMBC(0x1d4) | |
1099 CASEMBC(0x1ee7) | |
236 | 1100 regmbc('u'); regmbc('\371'); regmbc('\372'); |
1101 regmbc('\373'); regmbc('\374'); | |
2974 | 1102 REGMBC(0x169) REGMBC(0x16b) REGMBC(0x16d) |
1103 REGMBC(0x16f) REGMBC(0x171) REGMBC(0x173) | |
1104 REGMBC(0x1b0) REGMBC(0x1d4) REGMBC(0x1ee7) | |
1105 return; | |
1106 case 'v': CASEMBC(0x1e7d) | |
1107 regmbc('v'); REGMBC(0x1e7d) | |
1108 return; | |
1109 case 'w': CASEMBC(0x175) CASEMBC(0x1e81) CASEMBC(0x1e83) | |
1110 CASEMBC(0x1e85) CASEMBC(0x1e87) CASEMBC(0x1e98) | |
1111 regmbc('w'); REGMBC(0x175) REGMBC(0x1e81) | |
1112 REGMBC(0x1e83) REGMBC(0x1e85) REGMBC(0x1e87) | |
1113 REGMBC(0x1e98) | |
1114 return; | |
1115 case 'x': CASEMBC(0x1e8b) CASEMBC(0x1e8d) | |
1116 regmbc('x'); REGMBC(0x1e8b) REGMBC(0x1e8d) | |
167 | 1117 return; |
236 | 1118 case 'y': case '\375': case '\377': |
2974 | 1119 CASEMBC(0x177) CASEMBC(0x1e8f) CASEMBC(0x1e99) |
1120 CASEMBC(0x1ef3) CASEMBC(0x1ef7) CASEMBC(0x1ef9) | |
236 | 1121 regmbc('y'); regmbc('\375'); regmbc('\377'); |
2974 | 1122 REGMBC(0x177) REGMBC(0x1e8f) REGMBC(0x1e99) |
1123 REGMBC(0x1ef3) REGMBC(0x1ef7) REGMBC(0x1ef9) | |
1124 return; | |
1125 case 'z': CASEMBC(0x17a) CASEMBC(0x17c) CASEMBC(0x17e) | |
1126 CASEMBC(0x1b6) CASEMBC(0x1e91) CASEMBC(0x1e95) | |
1127 regmbc('z'); REGMBC(0x17a) REGMBC(0x17c) | |
1128 REGMBC(0x17e) REGMBC(0x1b6) REGMBC(0x1e91) | |
1129 REGMBC(0x1e95) | |
167 | 1130 return; |
1131 } | |
2247
c40cd9aad546
Add patch to improve support of z/OS (OS/390). (Ralf Schandl)
Bram Moolenaar <bram@vim.org>
parents:
2173
diff
changeset
|
1132 #endif |
167 | 1133 } |
1134 regmbc(c); | |
1135 } | |
1136 | |
1137 /* | |
1138 * Check for a collating element "[.a.]". "pp" points to the '['. | |
1139 * Returns a character. Zero means that no item was recognized. Otherwise | |
1140 * "pp" is advanced to after the item. | |
1141 * Currently only single characters are recognized! | |
1142 */ | |
1143 static int | |
1144 get_coll_element(pp) | |
1145 char_u **pp; | |
1146 { | |
1147 int c; | |
1148 int l = 1; | |
1149 char_u *p = *pp; | |
1150 | |
1151 if (p[1] == '.') | |
1152 { | |
1153 #ifdef FEAT_MBYTE | |
1154 if (has_mbyte) | |
474 | 1155 l = (*mb_ptr2len)(p + 2); |
167 | 1156 #endif |
1157 if (p[l + 2] == '.' && p[l + 3] == ']') | |
1158 { | |
1159 #ifdef FEAT_MBYTE | |
1160 if (has_mbyte) | |
1161 c = mb_ptr2char(p + 2); | |
1162 else | |
1163 #endif | |
1164 c = p[2]; | |
1165 *pp += l + 4; | |
1166 return c; | |
1167 } | |
1168 } | |
1169 return 0; | |
1170 } | |
1171 | |
1172 | |
1173 /* | |
1174 * Skip over a "[]" range. | |
1175 * "p" must point to the character after the '['. | |
1176 * The returned pointer is on the matching ']', or the terminating NUL. | |
1177 */ | |
1178 static char_u * | |
1179 skip_anyof(p) | |
1180 char_u *p; | |
1181 { | |
1182 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
1183 int cpo_bsl; /* 'cpoptions' contains '\' flag */ | |
1184 #ifdef FEAT_MBYTE | |
1185 int l; | |
1186 #endif | |
1187 | |
410 | 1188 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; |
1189 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
167 | 1190 |
1191 if (*p == '^') /* Complement of range. */ | |
1192 ++p; | |
1193 if (*p == ']' || *p == '-') | |
1194 ++p; | |
1195 while (*p != NUL && *p != ']') | |
1196 { | |
1197 #ifdef FEAT_MBYTE | |
474 | 1198 if (has_mbyte && (l = (*mb_ptr2len)(p)) > 1) |
167 | 1199 p += l; |
1200 else | |
1201 #endif | |
1202 if (*p == '-') | |
1203 { | |
1204 ++p; | |
1205 if (*p != ']' && *p != NUL) | |
1206 mb_ptr_adv(p); | |
1207 } | |
1208 else if (*p == '\\' | |
1209 && !cpo_bsl | |
1210 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL | |
1211 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL))) | |
1212 p += 2; | |
1213 else if (*p == '[') | |
1214 { | |
1215 if (get_char_class(&p) == CLASS_NONE | |
1216 && get_equi_class(&p) == 0 | |
1217 && get_coll_element(&p) == 0) | |
1218 ++p; /* It was not a class name */ | |
1219 } | |
1220 else | |
1221 ++p; | |
1222 } | |
1223 | |
1224 return p; | |
1225 } | |
1226 | |
1227 /* | |
7 | 1228 * Skip past regular expression. |
153 | 1229 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc). |
7 | 1230 * Take care of characters with a backslash in front of it. |
1231 * Skip strings inside [ and ]. | |
1232 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the | |
1233 * expression and change "\?" to "?". If "*newp" is not NULL the expression | |
1234 * is changed in-place. | |
1235 */ | |
1236 char_u * | |
1237 skip_regexp(startp, dirc, magic, newp) | |
1238 char_u *startp; | |
1239 int dirc; | |
1240 int magic; | |
1241 char_u **newp; | |
1242 { | |
1243 int mymagic; | |
1244 char_u *p = startp; | |
1245 | |
1246 if (magic) | |
1247 mymagic = MAGIC_ON; | |
1248 else | |
1249 mymagic = MAGIC_OFF; | |
1250 | |
39 | 1251 for (; p[0] != NUL; mb_ptr_adv(p)) |
7 | 1252 { |
1253 if (p[0] == dirc) /* found end of regexp */ | |
1254 break; | |
1255 if ((p[0] == '[' && mymagic >= MAGIC_ON) | |
1256 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF)) | |
1257 { | |
1258 p = skip_anyof(p + 1); | |
1259 if (p[0] == NUL) | |
1260 break; | |
1261 } | |
1262 else if (p[0] == '\\' && p[1] != NUL) | |
1263 { | |
1264 if (dirc == '?' && newp != NULL && p[1] == '?') | |
1265 { | |
1266 /* change "\?" to "?", make a copy first. */ | |
1267 if (*newp == NULL) | |
1268 { | |
1269 *newp = vim_strsave(startp); | |
1270 if (*newp != NULL) | |
1271 p = *newp + (p - startp); | |
1272 } | |
1273 if (*newp != NULL) | |
1621 | 1274 STRMOVE(p, p + 1); |
7 | 1275 else |
1276 ++p; | |
1277 } | |
1278 else | |
1279 ++p; /* skip next character */ | |
1280 if (*p == 'v') | |
1281 mymagic = MAGIC_ALL; | |
1282 else if (*p == 'V') | |
1283 mymagic = MAGIC_NONE; | |
1284 } | |
1285 } | |
1286 return p; | |
1287 } | |
1288 | |
4444 | 1289 static regprog_T *bt_regcomp __ARGS((char_u *expr, int re_flags)); |
1290 | |
7 | 1291 /* |
4444 | 1292 * bt_regcomp() - compile a regular expression into internal code for the |
1293 * traditional back track matcher. | |
41 | 1294 * Returns the program in allocated space. Returns NULL for an error. |
7 | 1295 * |
1296 * We can't allocate space until we know how big the compiled form will be, | |
1297 * but we can't compile it (and thus know how big it is) until we've got a | |
1298 * place to put the code. So we cheat: we compile it twice, once with code | |
1299 * generation turned off and size counting turned on, and once "for real". | |
1300 * This also means that we don't allocate space until we are sure that the | |
1301 * thing really will compile successfully, and we never have to move the | |
1302 * code and thus invalidate pointers into it. (Note that it has to be in | |
1303 * one piece because vim_free() must be able to free it all.) | |
1304 * | |
1305 * Whether upper/lower case is to be ignored is decided when executing the | |
1306 * program, it does not matter here. | |
1307 * | |
1308 * Beware that the optimization-preparation code in here knows about some | |
1309 * of the structure of the compiled regexp. | |
1310 * "re_flags": RE_MAGIC and/or RE_STRING. | |
1311 */ | |
4444 | 1312 static regprog_T * |
1313 bt_regcomp(expr, re_flags) | |
7 | 1314 char_u *expr; |
1315 int re_flags; | |
1316 { | |
4444 | 1317 bt_regprog_T *r; |
7 | 1318 char_u *scan; |
1319 char_u *longest; | |
1320 int len; | |
1321 int flags; | |
1322 | |
1323 if (expr == NULL) | |
1324 EMSG_RET_NULL(_(e_null)); | |
1325 | |
1326 init_class_tab(); | |
1327 | |
1328 /* | |
1329 * First pass: determine size, legality. | |
1330 */ | |
1331 regcomp_start(expr, re_flags); | |
1332 regcode = JUST_CALC_SIZE; | |
1333 regc(REGMAGIC); | |
1334 if (reg(REG_NOPAREN, &flags) == NULL) | |
1335 return NULL; | |
1336 | |
1337 /* Small enough for pointer-storage convention? */ | |
1338 #ifdef SMALL_MALLOC /* 16 bit storage allocation */ | |
1339 if (regsize >= 65536L - 256L) | |
1340 EMSG_RET_NULL(_("E339: Pattern too long")); | |
1341 #endif | |
1342 | |
1343 /* Allocate space. */ | |
4444 | 1344 r = (bt_regprog_T *)lalloc(sizeof(bt_regprog_T) + regsize, TRUE); |
7 | 1345 if (r == NULL) |
1346 return NULL; | |
1347 | |
1348 /* | |
1349 * Second pass: emit code. | |
1350 */ | |
1351 regcomp_start(expr, re_flags); | |
1352 regcode = r->program; | |
1353 regc(REGMAGIC); | |
2010 | 1354 if (reg(REG_NOPAREN, &flags) == NULL || reg_toolong) |
7 | 1355 { |
1356 vim_free(r); | |
2010 | 1357 if (reg_toolong) |
1358 EMSG_RET_NULL(_("E339: Pattern too long")); | |
7 | 1359 return NULL; |
1360 } | |
1361 | |
1362 /* Dig out information for optimizations. */ | |
1363 r->regstart = NUL; /* Worst-case defaults. */ | |
1364 r->reganch = 0; | |
1365 r->regmust = NULL; | |
1366 r->regmlen = 0; | |
1367 r->regflags = regflags; | |
1368 if (flags & HASNL) | |
1369 r->regflags |= RF_HASNL; | |
1370 if (flags & HASLOOKBH) | |
1371 r->regflags |= RF_LOOKBH; | |
1372 #ifdef FEAT_SYN_HL | |
1373 /* Remember whether this pattern has any \z specials in it. */ | |
1374 r->reghasz = re_has_z; | |
1375 #endif | |
1376 scan = r->program + 1; /* First BRANCH. */ | |
1377 if (OP(regnext(scan)) == END) /* Only one top-level choice. */ | |
1378 { | |
1379 scan = OPERAND(scan); | |
1380 | |
1381 /* Starting-point info. */ | |
1382 if (OP(scan) == BOL || OP(scan) == RE_BOF) | |
1383 { | |
1384 r->reganch++; | |
1385 scan = regnext(scan); | |
1386 } | |
1387 | |
1388 if (OP(scan) == EXACTLY) | |
1389 { | |
1390 #ifdef FEAT_MBYTE | |
1391 if (has_mbyte) | |
1392 r->regstart = (*mb_ptr2char)(OPERAND(scan)); | |
1393 else | |
1394 #endif | |
1395 r->regstart = *OPERAND(scan); | |
1396 } | |
1397 else if ((OP(scan) == BOW | |
1398 || OP(scan) == EOW | |
1399 || OP(scan) == NOTHING | |
1400 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN | |
1401 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE) | |
1402 && OP(regnext(scan)) == EXACTLY) | |
1403 { | |
1404 #ifdef FEAT_MBYTE | |
1405 if (has_mbyte) | |
1406 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan))); | |
1407 else | |
1408 #endif | |
1409 r->regstart = *OPERAND(regnext(scan)); | |
1410 } | |
1411 | |
1412 /* | |
1413 * If there's something expensive in the r.e., find the longest | |
1414 * literal string that must appear and make it the regmust. Resolve | |
1415 * ties in favor of later strings, since the regstart check works | |
1416 * with the beginning of the r.e. and avoiding duplication | |
1417 * strengthens checking. Not a strong reason, but sufficient in the | |
1418 * absence of others. | |
1419 */ | |
1420 /* | |
1421 * When the r.e. starts with BOW, it is faster to look for a regmust | |
1422 * first. Used a lot for "#" and "*" commands. (Added by mool). | |
1423 */ | |
1424 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW) | |
1425 && !(flags & HASNL)) | |
1426 { | |
1427 longest = NULL; | |
1428 len = 0; | |
1429 for (; scan != NULL; scan = regnext(scan)) | |
1430 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len) | |
1431 { | |
1432 longest = OPERAND(scan); | |
1433 len = (int)STRLEN(OPERAND(scan)); | |
1434 } | |
1435 r->regmust = longest; | |
1436 r->regmlen = len; | |
1437 } | |
1438 } | |
4444 | 1439 #ifdef BT_REGEXP_DUMP |
7 | 1440 regdump(expr, r); |
1441 #endif | |
4444 | 1442 r->engine = &bt_regengine; |
1443 return (regprog_T *)r; | |
7 | 1444 } |
1445 | |
1446 /* | |
1447 * Setup to parse the regexp. Used once to get the length and once to do it. | |
1448 */ | |
1449 static void | |
1450 regcomp_start(expr, re_flags) | |
1451 char_u *expr; | |
1452 int re_flags; /* see vim_regcomp() */ | |
1453 { | |
1454 initchr(expr); | |
1455 if (re_flags & RE_MAGIC) | |
1456 reg_magic = MAGIC_ON; | |
1457 else | |
1458 reg_magic = MAGIC_OFF; | |
1459 reg_string = (re_flags & RE_STRING); | |
481 | 1460 reg_strict = (re_flags & RE_STRICT); |
7 | 1461 |
1462 num_complex_braces = 0; | |
1463 regnpar = 1; | |
1464 vim_memset(had_endbrace, 0, sizeof(had_endbrace)); | |
1465 #ifdef FEAT_SYN_HL | |
1466 regnzpar = 1; | |
1467 re_has_z = 0; | |
1468 #endif | |
1469 regsize = 0L; | |
2010 | 1470 reg_toolong = FALSE; |
7 | 1471 regflags = 0; |
1472 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1473 had_eol = FALSE; | |
1474 #endif | |
1475 } | |
1476 | |
1477 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1478 /* | |
1479 * Check if during the previous call to vim_regcomp the EOL item "$" has been | |
1480 * found. This is messy, but it works fine. | |
1481 */ | |
1482 int | |
1483 vim_regcomp_had_eol() | |
1484 { | |
1485 return had_eol; | |
1486 } | |
1487 #endif | |
1488 | |
1489 /* | |
4444 | 1490 * Parse regular expression, i.e. main body or parenthesized thing. |
7 | 1491 * |
1492 * Caller must absorb opening parenthesis. | |
1493 * | |
1494 * Combining parenthesis handling with the base level of regular expression | |
1495 * is a trifle forced, but the need to tie the tails of the branches to what | |
1496 * follows makes it hard to avoid. | |
1497 */ | |
1498 static char_u * | |
1499 reg(paren, flagp) | |
1500 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */ | |
1501 int *flagp; | |
1502 { | |
1503 char_u *ret; | |
1504 char_u *br; | |
1505 char_u *ender; | |
1506 int parno = 0; | |
1507 int flags; | |
1508 | |
1509 *flagp = HASWIDTH; /* Tentatively. */ | |
1510 | |
1511 #ifdef FEAT_SYN_HL | |
1512 if (paren == REG_ZPAREN) | |
1513 { | |
1514 /* Make a ZOPEN node. */ | |
1515 if (regnzpar >= NSUBEXP) | |
1516 EMSG_RET_NULL(_("E50: Too many \\z(")); | |
1517 parno = regnzpar; | |
1518 regnzpar++; | |
1519 ret = regnode(ZOPEN + parno); | |
1520 } | |
1521 else | |
1522 #endif | |
1523 if (paren == REG_PAREN) | |
1524 { | |
1525 /* Make a MOPEN node. */ | |
1526 if (regnpar >= NSUBEXP) | |
4444 | 1527 EMSG2_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL); |
7 | 1528 parno = regnpar; |
1529 ++regnpar; | |
1530 ret = regnode(MOPEN + parno); | |
1531 } | |
1532 else if (paren == REG_NPAREN) | |
1533 { | |
1534 /* Make a NOPEN node. */ | |
1535 ret = regnode(NOPEN); | |
1536 } | |
1537 else | |
1538 ret = NULL; | |
1539 | |
1540 /* Pick up the branches, linking them together. */ | |
1541 br = regbranch(&flags); | |
1542 if (br == NULL) | |
1543 return NULL; | |
1544 if (ret != NULL) | |
1545 regtail(ret, br); /* [MZ]OPEN -> first. */ | |
1546 else | |
1547 ret = br; | |
1548 /* If one of the branches can be zero-width, the whole thing can. | |
1549 * If one of the branches has * at start or matches a line-break, the | |
1550 * whole thing can. */ | |
1551 if (!(flags & HASWIDTH)) | |
1552 *flagp &= ~HASWIDTH; | |
1553 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH); | |
1554 while (peekchr() == Magic('|')) | |
1555 { | |
1556 skipchr(); | |
1557 br = regbranch(&flags); | |
2010 | 1558 if (br == NULL || reg_toolong) |
7 | 1559 return NULL; |
1560 regtail(ret, br); /* BRANCH -> BRANCH. */ | |
1561 if (!(flags & HASWIDTH)) | |
1562 *flagp &= ~HASWIDTH; | |
1563 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH); | |
1564 } | |
1565 | |
1566 /* Make a closing node, and hook it on the end. */ | |
1567 ender = regnode( | |
1568 #ifdef FEAT_SYN_HL | |
1569 paren == REG_ZPAREN ? ZCLOSE + parno : | |
1570 #endif | |
1571 paren == REG_PAREN ? MCLOSE + parno : | |
1572 paren == REG_NPAREN ? NCLOSE : END); | |
1573 regtail(ret, ender); | |
1574 | |
1575 /* Hook the tails of the branches to the closing node. */ | |
1576 for (br = ret; br != NULL; br = regnext(br)) | |
1577 regoptail(br, ender); | |
1578 | |
1579 /* Check for proper termination. */ | |
1580 if (paren != REG_NOPAREN && getchr() != Magic(')')) | |
1581 { | |
1582 #ifdef FEAT_SYN_HL | |
1583 if (paren == REG_ZPAREN) | |
308 | 1584 EMSG_RET_NULL(_("E52: Unmatched \\z(")); |
7 | 1585 else |
1586 #endif | |
1587 if (paren == REG_NPAREN) | |
4444 | 1588 EMSG2_RET_NULL(_(e_unmatchedpp), reg_magic == MAGIC_ALL); |
7 | 1589 else |
4444 | 1590 EMSG2_RET_NULL(_(e_unmatchedp), reg_magic == MAGIC_ALL); |
7 | 1591 } |
1592 else if (paren == REG_NOPAREN && peekchr() != NUL) | |
1593 { | |
1594 if (curchr == Magic(')')) | |
4444 | 1595 EMSG2_RET_NULL(_(e_unmatchedpar), reg_magic == MAGIC_ALL); |
7 | 1596 else |
308 | 1597 EMSG_RET_NULL(_(e_trailing)); /* "Can't happen". */ |
7 | 1598 /* NOTREACHED */ |
1599 } | |
1600 /* | |
1601 * Here we set the flag allowing back references to this set of | |
1602 * parentheses. | |
1603 */ | |
1604 if (paren == REG_PAREN) | |
1605 had_endbrace[parno] = TRUE; /* have seen the close paren */ | |
1606 return ret; | |
1607 } | |
1608 | |
1609 /* | |
4444 | 1610 * Parse one alternative of an | operator. |
7 | 1611 * Implements the & operator. |
1612 */ | |
1613 static char_u * | |
1614 regbranch(flagp) | |
1615 int *flagp; | |
1616 { | |
1617 char_u *ret; | |
1618 char_u *chain = NULL; | |
1619 char_u *latest; | |
1620 int flags; | |
1621 | |
1622 *flagp = WORST | HASNL; /* Tentatively. */ | |
1623 | |
1624 ret = regnode(BRANCH); | |
1625 for (;;) | |
1626 { | |
1627 latest = regconcat(&flags); | |
1628 if (latest == NULL) | |
1629 return NULL; | |
1630 /* If one of the branches has width, the whole thing has. If one of | |
1631 * the branches anchors at start-of-line, the whole thing does. | |
1632 * If one of the branches uses look-behind, the whole thing does. */ | |
1633 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH); | |
1634 /* If one of the branches doesn't match a line-break, the whole thing | |
1635 * doesn't. */ | |
1636 *flagp &= ~HASNL | (flags & HASNL); | |
1637 if (chain != NULL) | |
1638 regtail(chain, latest); | |
1639 if (peekchr() != Magic('&')) | |
1640 break; | |
1641 skipchr(); | |
1642 regtail(latest, regnode(END)); /* operand ends */ | |
2010 | 1643 if (reg_toolong) |
1644 break; | |
7 | 1645 reginsert(MATCH, latest); |
1646 chain = latest; | |
1647 } | |
1648 | |
1649 return ret; | |
1650 } | |
1651 | |
1652 /* | |
4444 | 1653 * Parse one alternative of an | or & operator. |
7 | 1654 * Implements the concatenation operator. |
1655 */ | |
1656 static char_u * | |
1657 regconcat(flagp) | |
1658 int *flagp; | |
1659 { | |
1660 char_u *first = NULL; | |
1661 char_u *chain = NULL; | |
1662 char_u *latest; | |
1663 int flags; | |
1664 int cont = TRUE; | |
1665 | |
1666 *flagp = WORST; /* Tentatively. */ | |
1667 | |
1668 while (cont) | |
1669 { | |
1670 switch (peekchr()) | |
1671 { | |
1672 case NUL: | |
1673 case Magic('|'): | |
1674 case Magic('&'): | |
1675 case Magic(')'): | |
1676 cont = FALSE; | |
1677 break; | |
1678 case Magic('Z'): | |
1679 #ifdef FEAT_MBYTE | |
1680 regflags |= RF_ICOMBINE; | |
1681 #endif | |
1682 skipchr_keepstart(); | |
1683 break; | |
1684 case Magic('c'): | |
1685 regflags |= RF_ICASE; | |
1686 skipchr_keepstart(); | |
1687 break; | |
1688 case Magic('C'): | |
1689 regflags |= RF_NOICASE; | |
1690 skipchr_keepstart(); | |
1691 break; | |
1692 case Magic('v'): | |
1693 reg_magic = MAGIC_ALL; | |
1694 skipchr_keepstart(); | |
1695 curchr = -1; | |
1696 break; | |
1697 case Magic('m'): | |
1698 reg_magic = MAGIC_ON; | |
1699 skipchr_keepstart(); | |
1700 curchr = -1; | |
1701 break; | |
1702 case Magic('M'): | |
1703 reg_magic = MAGIC_OFF; | |
1704 skipchr_keepstart(); | |
1705 curchr = -1; | |
1706 break; | |
1707 case Magic('V'): | |
1708 reg_magic = MAGIC_NONE; | |
1709 skipchr_keepstart(); | |
1710 curchr = -1; | |
1711 break; | |
1712 default: | |
1713 latest = regpiece(&flags); | |
2010 | 1714 if (latest == NULL || reg_toolong) |
7 | 1715 return NULL; |
1716 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH); | |
1717 if (chain == NULL) /* First piece. */ | |
1718 *flagp |= flags & SPSTART; | |
1719 else | |
1720 regtail(chain, latest); | |
1721 chain = latest; | |
1722 if (first == NULL) | |
1723 first = latest; | |
1724 break; | |
1725 } | |
1726 } | |
1727 if (first == NULL) /* Loop ran zero times. */ | |
1728 first = regnode(NOTHING); | |
1729 return first; | |
1730 } | |
1731 | |
1732 /* | |
4444 | 1733 * Parse something followed by possible [*+=]. |
7 | 1734 * |
1735 * Note that the branching code sequences used for = and the general cases | |
1736 * of * and + are somewhat optimized: they use the same NOTHING node as | |
1737 * both the endmarker for their branch list and the body of the last branch. | |
1738 * It might seem that this node could be dispensed with entirely, but the | |
1739 * endmarker role is not redundant. | |
1740 */ | |
1741 static char_u * | |
1742 regpiece(flagp) | |
1743 int *flagp; | |
1744 { | |
1745 char_u *ret; | |
1746 int op; | |
1747 char_u *next; | |
1748 int flags; | |
1749 long minval; | |
1750 long maxval; | |
1751 | |
1752 ret = regatom(&flags); | |
1753 if (ret == NULL) | |
1754 return NULL; | |
1755 | |
1756 op = peekchr(); | |
1757 if (re_multi_type(op) == NOT_MULTI) | |
1758 { | |
1759 *flagp = flags; | |
1760 return ret; | |
1761 } | |
1762 /* default flags */ | |
1763 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH))); | |
1764 | |
1765 skipchr(); | |
1766 switch (op) | |
1767 { | |
1768 case Magic('*'): | |
1769 if (flags & SIMPLE) | |
1770 reginsert(STAR, ret); | |
1771 else | |
1772 { | |
1773 /* Emit x* as (x&|), where & means "self". */ | |
1774 reginsert(BRANCH, ret); /* Either x */ | |
1775 regoptail(ret, regnode(BACK)); /* and loop */ | |
1776 regoptail(ret, ret); /* back */ | |
1777 regtail(ret, regnode(BRANCH)); /* or */ | |
1778 regtail(ret, regnode(NOTHING)); /* null. */ | |
1779 } | |
1780 break; | |
1781 | |
1782 case Magic('+'): | |
1783 if (flags & SIMPLE) | |
1784 reginsert(PLUS, ret); | |
1785 else | |
1786 { | |
1787 /* Emit x+ as x(&|), where & means "self". */ | |
1788 next = regnode(BRANCH); /* Either */ | |
1789 regtail(ret, next); | |
233 | 1790 regtail(regnode(BACK), ret); /* loop back */ |
7 | 1791 regtail(next, regnode(BRANCH)); /* or */ |
1792 regtail(ret, regnode(NOTHING)); /* null. */ | |
1793 } | |
1794 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH))); | |
1795 break; | |
1796 | |
1797 case Magic('@'): | |
1798 { | |
1799 int lop = END; | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1800 int nr; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1801 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1802 nr = getdecchrs(); |
7 | 1803 switch (no_Magic(getchr())) |
1804 { | |
1805 case '=': lop = MATCH; break; /* \@= */ | |
1806 case '!': lop = NOMATCH; break; /* \@! */ | |
1807 case '>': lop = SUBPAT; break; /* \@> */ | |
1808 case '<': switch (no_Magic(getchr())) | |
1809 { | |
1810 case '=': lop = BEHIND; break; /* \@<= */ | |
1811 case '!': lop = NOBEHIND; break; /* \@<! */ | |
1812 } | |
1813 } | |
1814 if (lop == END) | |
4444 | 1815 EMSG2_RET_NULL(_("E59: invalid character after %s@"), |
7 | 1816 reg_magic == MAGIC_ALL); |
1817 /* Look behind must match with behind_pos. */ | |
1818 if (lop == BEHIND || lop == NOBEHIND) | |
1819 { | |
1820 regtail(ret, regnode(BHPOS)); | |
1821 *flagp |= HASLOOKBH; | |
1822 } | |
1823 regtail(ret, regnode(END)); /* operand ends */ | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1824 if (lop == BEHIND || lop == NOBEHIND) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1825 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1826 if (nr < 0) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1827 nr = 0; /* no limit is same as zero limit */ |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1828 reginsert_nr(lop, nr, ret); |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1829 } |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1830 else |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
1831 reginsert(lop, ret); |
7 | 1832 break; |
1833 } | |
1834 | |
1835 case Magic('?'): | |
1836 case Magic('='): | |
1837 /* Emit x= as (x|) */ | |
1838 reginsert(BRANCH, ret); /* Either x */ | |
1839 regtail(ret, regnode(BRANCH)); /* or */ | |
1840 next = regnode(NOTHING); /* null. */ | |
1841 regtail(ret, next); | |
1842 regoptail(ret, next); | |
1843 break; | |
1844 | |
1845 case Magic('{'): | |
1846 if (!read_limits(&minval, &maxval)) | |
1847 return NULL; | |
1848 if (flags & SIMPLE) | |
1849 { | |
1850 reginsert(BRACE_SIMPLE, ret); | |
1851 reginsert_limits(BRACE_LIMITS, minval, maxval, ret); | |
1852 } | |
1853 else | |
1854 { | |
1855 if (num_complex_braces >= 10) | |
4444 | 1856 EMSG2_RET_NULL(_("E60: Too many complex %s{...}s"), |
7 | 1857 reg_magic == MAGIC_ALL); |
1858 reginsert(BRACE_COMPLEX + num_complex_braces, ret); | |
1859 regoptail(ret, regnode(BACK)); | |
1860 regoptail(ret, ret); | |
1861 reginsert_limits(BRACE_LIMITS, minval, maxval, ret); | |
1862 ++num_complex_braces; | |
1863 } | |
1864 if (minval > 0 && maxval > 0) | |
1865 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH))); | |
1866 break; | |
1867 } | |
1868 if (re_multi_type(peekchr()) != NOT_MULTI) | |
1869 { | |
1870 /* Can't have a multi follow a multi. */ | |
1871 if (peekchr() == Magic('*')) | |
1872 sprintf((char *)IObuff, _("E61: Nested %s*"), | |
1873 reg_magic >= MAGIC_ON ? "" : "\\"); | |
1874 else | |
1875 sprintf((char *)IObuff, _("E62: Nested %s%c"), | |
1876 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr())); | |
1877 EMSG_RET_NULL(IObuff); | |
1878 } | |
1879 | |
1880 return ret; | |
1881 } | |
1882 | |
4444 | 1883 /* When making changes to classchars also change nfa_classcodes. */ |
1884 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU"; | |
1885 static int classcodes[] = { | |
1886 ANY, IDENT, SIDENT, KWORD, SKWORD, | |
1887 FNAME, SFNAME, PRINT, SPRINT, | |
1888 WHITE, NWHITE, DIGIT, NDIGIT, | |
1889 HEX, NHEX, OCTAL, NOCTAL, | |
1890 WORD, NWORD, HEAD, NHEAD, | |
1891 ALPHA, NALPHA, LOWER, NLOWER, | |
1892 UPPER, NUPPER | |
1893 }; | |
1894 | |
7 | 1895 /* |
4444 | 1896 * Parse the lowest level. |
7 | 1897 * |
1898 * Optimization: gobbles an entire sequence of ordinary characters so that | |
1899 * it can turn them into a single node, which is smaller to store and | |
1900 * faster to run. Don't do this when one_exactly is set. | |
1901 */ | |
1902 static char_u * | |
1903 regatom(flagp) | |
1904 int *flagp; | |
1905 { | |
1906 char_u *ret; | |
1907 int flags; | |
1908 int cpo_lit; /* 'cpoptions' contains 'l' flag */ | |
167 | 1909 int cpo_bsl; /* 'cpoptions' contains '\' flag */ |
7 | 1910 int c; |
1911 char_u *p; | |
1912 int extra = 0; | |
1913 | |
1914 *flagp = WORST; /* Tentatively. */ | |
410 | 1915 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL; |
1916 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL; | |
7 | 1917 |
1918 c = getchr(); | |
1919 switch (c) | |
1920 { | |
1921 case Magic('^'): | |
1922 ret = regnode(BOL); | |
1923 break; | |
1924 | |
1925 case Magic('$'): | |
1926 ret = regnode(EOL); | |
1927 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1928 had_eol = TRUE; | |
1929 #endif | |
1930 break; | |
1931 | |
1932 case Magic('<'): | |
1933 ret = regnode(BOW); | |
1934 break; | |
1935 | |
1936 case Magic('>'): | |
1937 ret = regnode(EOW); | |
1938 break; | |
1939 | |
1940 case Magic('_'): | |
1941 c = no_Magic(getchr()); | |
1942 if (c == '^') /* "\_^" is start-of-line */ | |
1943 { | |
1944 ret = regnode(BOL); | |
1945 break; | |
1946 } | |
1947 if (c == '$') /* "\_$" is end-of-line */ | |
1948 { | |
1949 ret = regnode(EOL); | |
1950 #if defined(FEAT_SYN_HL) || defined(PROTO) | |
1951 had_eol = TRUE; | |
1952 #endif | |
1953 break; | |
1954 } | |
1955 | |
1956 extra = ADD_NL; | |
1957 *flagp |= HASNL; | |
1958 | |
1959 /* "\_[" is character range plus newline */ | |
1960 if (c == '[') | |
1961 goto collection; | |
1962 | |
1963 /* "\_x" is character class plus newline */ | |
1964 /*FALLTHROUGH*/ | |
1965 | |
1966 /* | |
1967 * Character classes. | |
1968 */ | |
1969 case Magic('.'): | |
1970 case Magic('i'): | |
1971 case Magic('I'): | |
1972 case Magic('k'): | |
1973 case Magic('K'): | |
1974 case Magic('f'): | |
1975 case Magic('F'): | |
1976 case Magic('p'): | |
1977 case Magic('P'): | |
1978 case Magic('s'): | |
1979 case Magic('S'): | |
1980 case Magic('d'): | |
1981 case Magic('D'): | |
1982 case Magic('x'): | |
1983 case Magic('X'): | |
1984 case Magic('o'): | |
1985 case Magic('O'): | |
1986 case Magic('w'): | |
1987 case Magic('W'): | |
1988 case Magic('h'): | |
1989 case Magic('H'): | |
1990 case Magic('a'): | |
1991 case Magic('A'): | |
1992 case Magic('l'): | |
1993 case Magic('L'): | |
1994 case Magic('u'): | |
1995 case Magic('U'): | |
1996 p = vim_strchr(classchars, no_Magic(c)); | |
1997 if (p == NULL) | |
1998 EMSG_RET_NULL(_("E63: invalid use of \\_")); | |
714 | 1999 #ifdef FEAT_MBYTE |
2000 /* When '.' is followed by a composing char ignore the dot, so that | |
2001 * the composing char is matched here. */ | |
2002 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr())) | |
2003 { | |
2004 c = getchr(); | |
2005 goto do_multibyte; | |
2006 } | |
2007 #endif | |
7 | 2008 ret = regnode(classcodes[p - classchars] + extra); |
2009 *flagp |= HASWIDTH | SIMPLE; | |
2010 break; | |
2011 | |
2012 case Magic('n'): | |
2013 if (reg_string) | |
2014 { | |
2015 /* In a string "\n" matches a newline character. */ | |
2016 ret = regnode(EXACTLY); | |
2017 regc(NL); | |
2018 regc(NUL); | |
2019 *flagp |= HASWIDTH | SIMPLE; | |
2020 } | |
2021 else | |
2022 { | |
2023 /* In buffer text "\n" matches the end of a line. */ | |
2024 ret = regnode(NEWL); | |
2025 *flagp |= HASWIDTH | HASNL; | |
2026 } | |
2027 break; | |
2028 | |
2029 case Magic('('): | |
2030 if (one_exactly) | |
2031 EMSG_ONE_RET_NULL; | |
2032 ret = reg(REG_PAREN, &flags); | |
2033 if (ret == NULL) | |
2034 return NULL; | |
2035 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH); | |
2036 break; | |
2037 | |
2038 case NUL: | |
2039 case Magic('|'): | |
2040 case Magic('&'): | |
2041 case Magic(')'): | |
1468 | 2042 if (one_exactly) |
2043 EMSG_ONE_RET_NULL; | |
7 | 2044 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */ |
2045 /* NOTREACHED */ | |
2046 | |
2047 case Magic('='): | |
2048 case Magic('?'): | |
2049 case Magic('+'): | |
2050 case Magic('@'): | |
2051 case Magic('{'): | |
2052 case Magic('*'): | |
2053 c = no_Magic(c); | |
2054 sprintf((char *)IObuff, _("E64: %s%c follows nothing"), | |
2055 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL) | |
2056 ? "" : "\\", c); | |
2057 EMSG_RET_NULL(IObuff); | |
2058 /* NOTREACHED */ | |
2059 | |
2060 case Magic('~'): /* previous substitute pattern */ | |
359 | 2061 if (reg_prev_sub != NULL) |
7 | 2062 { |
2063 char_u *lp; | |
2064 | |
2065 ret = regnode(EXACTLY); | |
2066 lp = reg_prev_sub; | |
2067 while (*lp != NUL) | |
2068 regc(*lp++); | |
2069 regc(NUL); | |
2070 if (*reg_prev_sub != NUL) | |
2071 { | |
2072 *flagp |= HASWIDTH; | |
2073 if ((lp - reg_prev_sub) == 1) | |
2074 *flagp |= SIMPLE; | |
2075 } | |
2076 } | |
2077 else | |
2078 EMSG_RET_NULL(_(e_nopresub)); | |
2079 break; | |
2080 | |
2081 case Magic('1'): | |
2082 case Magic('2'): | |
2083 case Magic('3'): | |
2084 case Magic('4'): | |
2085 case Magic('5'): | |
2086 case Magic('6'): | |
2087 case Magic('7'): | |
2088 case Magic('8'): | |
2089 case Magic('9'): | |
2090 { | |
2091 int refnum; | |
2092 | |
2093 refnum = c - Magic('0'); | |
2094 /* | |
2095 * Check if the back reference is legal. We must have seen the | |
2096 * close brace. | |
2097 * TODO: Should also check that we don't refer to something | |
2098 * that is repeated (+*=): what instance of the repetition | |
2099 * should we match? | |
2100 */ | |
2101 if (!had_endbrace[refnum]) | |
2102 { | |
2103 /* Trick: check if "@<=" or "@<!" follows, in which case | |
2104 * the \1 can appear before the referenced match. */ | |
2105 for (p = regparse; *p != NUL; ++p) | |
2106 if (p[0] == '@' && p[1] == '<' | |
2107 && (p[2] == '!' || p[2] == '=')) | |
2108 break; | |
2109 if (*p == NUL) | |
2110 EMSG_RET_NULL(_("E65: Illegal back reference")); | |
2111 } | |
2112 ret = regnode(BACKREF + refnum); | |
2113 } | |
2114 break; | |
2115 | |
2116 case Magic('z'): | |
2117 { | |
2118 c = no_Magic(getchr()); | |
2119 switch (c) | |
2120 { | |
741 | 2121 #ifdef FEAT_SYN_HL |
7 | 2122 case '(': if (reg_do_extmatch != REX_SET) |
2123 EMSG_RET_NULL(_("E66: \\z( not allowed here")); | |
2124 if (one_exactly) | |
2125 EMSG_ONE_RET_NULL; | |
2126 ret = reg(REG_ZPAREN, &flags); | |
2127 if (ret == NULL) | |
2128 return NULL; | |
2129 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH); | |
2130 re_has_z = REX_SET; | |
2131 break; | |
2132 | |
2133 case '1': | |
2134 case '2': | |
2135 case '3': | |
2136 case '4': | |
2137 case '5': | |
2138 case '6': | |
2139 case '7': | |
2140 case '8': | |
2141 case '9': if (reg_do_extmatch != REX_USE) | |
2142 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here")); | |
2143 ret = regnode(ZREF + c - '0'); | |
2144 re_has_z = REX_USE; | |
2145 break; | |
741 | 2146 #endif |
7 | 2147 |
2148 case 's': ret = regnode(MOPEN + 0); | |
2149 break; | |
2150 | |
2151 case 'e': ret = regnode(MCLOSE + 0); | |
2152 break; | |
2153 | |
2154 default: EMSG_RET_NULL(_("E68: Invalid character after \\z")); | |
2155 } | |
2156 } | |
2157 break; | |
2158 | |
2159 case Magic('%'): | |
2160 { | |
2161 c = no_Magic(getchr()); | |
2162 switch (c) | |
2163 { | |
2164 /* () without a back reference */ | |
2165 case '(': | |
2166 if (one_exactly) | |
2167 EMSG_ONE_RET_NULL; | |
2168 ret = reg(REG_NPAREN, &flags); | |
2169 if (ret == NULL) | |
2170 return NULL; | |
2171 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH); | |
2172 break; | |
2173 | |
2174 /* Catch \%^ and \%$ regardless of where they appear in the | |
2175 * pattern -- regardless of whether or not it makes sense. */ | |
2176 case '^': | |
2177 ret = regnode(RE_BOF); | |
2178 break; | |
2179 | |
2180 case '$': | |
2181 ret = regnode(RE_EOF); | |
2182 break; | |
2183 | |
2184 case '#': | |
2185 ret = regnode(CURSOR); | |
2186 break; | |
2187 | |
639 | 2188 case 'V': |
2189 ret = regnode(RE_VISUAL); | |
2190 break; | |
2191 | |
7 | 2192 /* \%[abc]: Emit as a list of branches, all ending at the last |
2193 * branch which matches nothing. */ | |
2194 case '[': | |
2195 if (one_exactly) /* doesn't nest */ | |
2196 EMSG_ONE_RET_NULL; | |
2197 { | |
2198 char_u *lastbranch; | |
2199 char_u *lastnode = NULL; | |
2200 char_u *br; | |
2201 | |
2202 ret = NULL; | |
2203 while ((c = getchr()) != ']') | |
2204 { | |
2205 if (c == NUL) | |
4444 | 2206 EMSG2_RET_NULL(_("E69: Missing ] after %s%%["), |
7 | 2207 reg_magic == MAGIC_ALL); |
2208 br = regnode(BRANCH); | |
2209 if (ret == NULL) | |
2210 ret = br; | |
2211 else | |
2212 regtail(lastnode, br); | |
2213 | |
2214 ungetchr(); | |
2215 one_exactly = TRUE; | |
2216 lastnode = regatom(flagp); | |
2217 one_exactly = FALSE; | |
2218 if (lastnode == NULL) | |
2219 return NULL; | |
2220 } | |
2221 if (ret == NULL) | |
4444 | 2222 EMSG2_RET_NULL(_("E70: Empty %s%%[]"), |
7 | 2223 reg_magic == MAGIC_ALL); |
2224 lastbranch = regnode(BRANCH); | |
2225 br = regnode(NOTHING); | |
2226 if (ret != JUST_CALC_SIZE) | |
2227 { | |
2228 regtail(lastnode, br); | |
2229 regtail(lastbranch, br); | |
2230 /* connect all branches to the NOTHING | |
2231 * branch at the end */ | |
2232 for (br = ret; br != lastnode; ) | |
2233 { | |
2234 if (OP(br) == BRANCH) | |
2235 { | |
2236 regtail(br, lastbranch); | |
2237 br = OPERAND(br); | |
2238 } | |
2239 else | |
2240 br = regnext(br); | |
2241 } | |
2242 } | |
1701 | 2243 *flagp &= ~(HASWIDTH | SIMPLE); |
7 | 2244 break; |
2245 } | |
2246 | |
24 | 2247 case 'd': /* %d123 decimal */ |
2248 case 'o': /* %o123 octal */ | |
2249 case 'x': /* %xab hex 2 */ | |
2250 case 'u': /* %uabcd hex 4 */ | |
2251 case 'U': /* %U1234abcd hex 8 */ | |
2252 { | |
2253 int i; | |
2254 | |
2255 switch (c) | |
2256 { | |
2257 case 'd': i = getdecchrs(); break; | |
2258 case 'o': i = getoctchrs(); break; | |
2259 case 'x': i = gethexchrs(2); break; | |
2260 case 'u': i = gethexchrs(4); break; | |
2261 case 'U': i = gethexchrs(8); break; | |
2262 default: i = -1; break; | |
2263 } | |
2264 | |
2265 if (i < 0) | |
4444 | 2266 EMSG2_RET_NULL( |
24 | 2267 _("E678: Invalid character after %s%%[dxouU]"), |
2268 reg_magic == MAGIC_ALL); | |
714 | 2269 #ifdef FEAT_MBYTE |
2270 if (use_multibytecode(i)) | |
2271 ret = regnode(MULTIBYTECODE); | |
2272 else | |
2273 #endif | |
2274 ret = regnode(EXACTLY); | |
24 | 2275 if (i == 0) |
2276 regc(0x0a); | |
2277 else | |
2278 #ifdef FEAT_MBYTE | |
2279 regmbc(i); | |
2280 #else | |
2281 regc(i); | |
2282 #endif | |
2283 regc(NUL); | |
2284 *flagp |= HASWIDTH; | |
2285 break; | |
2286 } | |
2287 | |
7 | 2288 default: |
639 | 2289 if (VIM_ISDIGIT(c) || c == '<' || c == '>' |
2290 || c == '\'') | |
7 | 2291 { |
2292 long_u n = 0; | |
2293 int cmp; | |
2294 | |
2295 cmp = c; | |
2296 if (cmp == '<' || cmp == '>') | |
2297 c = getchr(); | |
2298 while (VIM_ISDIGIT(c)) | |
2299 { | |
2300 n = n * 10 + (c - '0'); | |
2301 c = getchr(); | |
2302 } | |
639 | 2303 if (c == '\'' && n == 0) |
2304 { | |
2305 /* "\%'m", "\%<'m" and "\%>'m": Mark */ | |
2306 c = getchr(); | |
2307 ret = regnode(RE_MARK); | |
2308 if (ret == JUST_CALC_SIZE) | |
2309 regsize += 2; | |
2310 else | |
2311 { | |
2312 *regcode++ = c; | |
2313 *regcode++ = cmp; | |
2314 } | |
2315 break; | |
2316 } | |
2317 else if (c == 'l' || c == 'c' || c == 'v') | |
7 | 2318 { |
2319 if (c == 'l') | |
2320 ret = regnode(RE_LNUM); | |
2321 else if (c == 'c') | |
2322 ret = regnode(RE_COL); | |
2323 else | |
2324 ret = regnode(RE_VCOL); | |
2325 if (ret == JUST_CALC_SIZE) | |
2326 regsize += 5; | |
2327 else | |
2328 { | |
2329 /* put the number and the optional | |
2330 * comparator after the opcode */ | |
2331 regcode = re_put_long(regcode, n); | |
2332 *regcode++ = cmp; | |
2333 } | |
2334 break; | |
2335 } | |
2336 } | |
2337 | |
4444 | 2338 EMSG2_RET_NULL(_("E71: Invalid character after %s%%"), |
7 | 2339 reg_magic == MAGIC_ALL); |
2340 } | |
2341 } | |
2342 break; | |
2343 | |
2344 case Magic('['): | |
2345 collection: | |
2346 { | |
2347 char_u *lp; | |
2348 | |
2349 /* | |
2350 * If there is no matching ']', we assume the '[' is a normal | |
2351 * character. This makes 'incsearch' and ":help [" work. | |
2352 */ | |
2353 lp = skip_anyof(regparse); | |
2354 if (*lp == ']') /* there is a matching ']' */ | |
2355 { | |
2356 int startc = -1; /* > 0 when next '-' is a range */ | |
2357 int endc; | |
2358 | |
2359 /* | |
2360 * In a character class, different parsing rules apply. | |
2361 * Not even \ is special anymore, nothing is. | |
2362 */ | |
2363 if (*regparse == '^') /* Complement of range. */ | |
2364 { | |
2365 ret = regnode(ANYBUT + extra); | |
2366 regparse++; | |
2367 } | |
2368 else | |
2369 ret = regnode(ANYOF + extra); | |
2370 | |
2371 /* At the start ']' and '-' mean the literal character. */ | |
2372 if (*regparse == ']' || *regparse == '-') | |
167 | 2373 { |
2374 startc = *regparse; | |
7 | 2375 regc(*regparse++); |
167 | 2376 } |
7 | 2377 |
2378 while (*regparse != NUL && *regparse != ']') | |
2379 { | |
2380 if (*regparse == '-') | |
2381 { | |
2382 ++regparse; | |
2383 /* The '-' is not used for a range at the end and | |
2384 * after or before a '\n'. */ | |
2385 if (*regparse == ']' || *regparse == NUL | |
2386 || startc == -1 | |
2387 || (regparse[0] == '\\' && regparse[1] == 'n')) | |
2388 { | |
2389 regc('-'); | |
2390 startc = '-'; /* [--x] is a range */ | |
2391 } | |
2392 else | |
2393 { | |
167 | 2394 /* Also accept "a-[.z.]" */ |
2395 endc = 0; | |
2396 if (*regparse == '[') | |
2397 endc = get_coll_element(®parse); | |
2398 if (endc == 0) | |
2399 { | |
7 | 2400 #ifdef FEAT_MBYTE |
167 | 2401 if (has_mbyte) |
2402 endc = mb_ptr2char_adv(®parse); | |
2403 else | |
7 | 2404 #endif |
167 | 2405 endc = *regparse++; |
2406 } | |
24 | 2407 |
2408 /* Handle \o40, \x20 and \u20AC style sequences */ | |
167 | 2409 if (endc == '\\' && !cpo_lit && !cpo_bsl) |
24 | 2410 endc = coll_get_char(); |
2411 | |
7 | 2412 if (startc > endc) |
2413 EMSG_RET_NULL(_(e_invrange)); | |
2414 #ifdef FEAT_MBYTE | |
2415 if (has_mbyte && ((*mb_char2len)(startc) > 1 | |
2416 || (*mb_char2len)(endc) > 1)) | |
2417 { | |
2418 /* Limit to a range of 256 chars */ | |
2419 if (endc > startc + 256) | |
2420 EMSG_RET_NULL(_(e_invrange)); | |
2421 while (++startc <= endc) | |
2422 regmbc(startc); | |
2423 } | |
2424 else | |
2425 #endif | |
2426 { | |
2427 #ifdef EBCDIC | |
2428 int alpha_only = FALSE; | |
2429 | |
2430 /* for alphabetical range skip the gaps | |
2431 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */ | |
2432 if (isalpha(startc) && isalpha(endc)) | |
2433 alpha_only = TRUE; | |
2434 #endif | |
2435 while (++startc <= endc) | |
2436 #ifdef EBCDIC | |
2437 if (!alpha_only || isalpha(startc)) | |
2438 #endif | |
2439 regc(startc); | |
2440 } | |
2441 startc = -1; | |
2442 } | |
2443 } | |
2444 /* | |
2445 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim | |
2446 * accepts "\t", "\e", etc., but only when the 'l' flag in | |
2447 * 'cpoptions' is not included. | |
167 | 2448 * Posix doesn't recognize backslash at all. |
7 | 2449 */ |
2450 else if (*regparse == '\\' | |
167 | 2451 && !cpo_bsl |
7 | 2452 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL |
2453 || (!cpo_lit | |
2454 && vim_strchr(REGEXP_ABBR, | |
2455 regparse[1]) != NULL))) | |
2456 { | |
2457 regparse++; | |
2458 if (*regparse == 'n') | |
2459 { | |
2460 /* '\n' in range: also match NL */ | |
2461 if (ret != JUST_CALC_SIZE) | |
2462 { | |
4084 | 2463 /* Using \n inside [^] does not change what |
2464 * matches. "[^\n]" is the same as ".". */ | |
2465 if (*ret == ANYOF) | |
2466 { | |
7 | 2467 *ret = ANYOF + ADD_NL; |
4084 | 2468 *flagp |= HASNL; |
2469 } | |
7 | 2470 /* else: must have had a \n already */ |
2471 } | |
2472 regparse++; | |
2473 startc = -1; | |
2474 } | |
24 | 2475 else if (*regparse == 'd' |
2476 || *regparse == 'o' | |
2477 || *regparse == 'x' | |
2478 || *regparse == 'u' | |
2479 || *regparse == 'U') | |
2480 { | |
2481 startc = coll_get_char(); | |
2482 if (startc == 0) | |
2483 regc(0x0a); | |
2484 else | |
2485 #ifdef FEAT_MBYTE | |
2486 regmbc(startc); | |
2487 #else | |
2488 regc(startc); | |
2489 #endif | |
2490 } | |
7 | 2491 else |
2492 { | |
2493 startc = backslash_trans(*regparse++); | |
2494 regc(startc); | |
2495 } | |
2496 } | |
2497 else if (*regparse == '[') | |
2498 { | |
2499 int c_class; | |
2500 int cu; | |
2501 | |
167 | 2502 c_class = get_char_class(®parse); |
7 | 2503 startc = -1; |
2504 /* Characters assumed to be 8 bits! */ | |
2505 switch (c_class) | |
2506 { | |
2507 case CLASS_NONE: | |
167 | 2508 c_class = get_equi_class(®parse); |
2509 if (c_class != 0) | |
2510 { | |
2511 /* produce equivalence class */ | |
2512 reg_equi_class(c_class); | |
2513 } | |
2514 else if ((c_class = | |
2515 get_coll_element(®parse)) != 0) | |
2516 { | |
2517 /* produce a collating element */ | |
2518 regmbc(c_class); | |
2519 } | |
2520 else | |
2521 { | |
2522 /* literal '[', allow [[-x] as a range */ | |
2523 startc = *regparse++; | |
2524 regc(startc); | |
2525 } | |
7 | 2526 break; |
2527 case CLASS_ALNUM: | |
2528 for (cu = 1; cu <= 255; cu++) | |
2529 if (isalnum(cu)) | |
2530 regc(cu); | |
2531 break; | |
2532 case CLASS_ALPHA: | |
2533 for (cu = 1; cu <= 255; cu++) | |
2534 if (isalpha(cu)) | |
2535 regc(cu); | |
2536 break; | |
2537 case CLASS_BLANK: | |
2538 regc(' '); | |
2539 regc('\t'); | |
2540 break; | |
2541 case CLASS_CNTRL: | |
2542 for (cu = 1; cu <= 255; cu++) | |
2543 if (iscntrl(cu)) | |
2544 regc(cu); | |
2545 break; | |
2546 case CLASS_DIGIT: | |
2547 for (cu = 1; cu <= 255; cu++) | |
2548 if (VIM_ISDIGIT(cu)) | |
2549 regc(cu); | |
2550 break; | |
2551 case CLASS_GRAPH: | |
2552 for (cu = 1; cu <= 255; cu++) | |
2553 if (isgraph(cu)) | |
2554 regc(cu); | |
2555 break; | |
2556 case CLASS_LOWER: | |
2557 for (cu = 1; cu <= 255; cu++) | |
1347 | 2558 if (MB_ISLOWER(cu)) |
7 | 2559 regc(cu); |
2560 break; | |
2561 case CLASS_PRINT: | |
2562 for (cu = 1; cu <= 255; cu++) | |
2563 if (vim_isprintc(cu)) | |
2564 regc(cu); | |
2565 break; | |
2566 case CLASS_PUNCT: | |
2567 for (cu = 1; cu <= 255; cu++) | |
2568 if (ispunct(cu)) | |
2569 regc(cu); | |
2570 break; | |
2571 case CLASS_SPACE: | |
2572 for (cu = 9; cu <= 13; cu++) | |
2573 regc(cu); | |
2574 regc(' '); | |
2575 break; | |
2576 case CLASS_UPPER: | |
2577 for (cu = 1; cu <= 255; cu++) | |
1347 | 2578 if (MB_ISUPPER(cu)) |
7 | 2579 regc(cu); |
2580 break; | |
2581 case CLASS_XDIGIT: | |
2582 for (cu = 1; cu <= 255; cu++) | |
2583 if (vim_isxdigit(cu)) | |
2584 regc(cu); | |
2585 break; | |
2586 case CLASS_TAB: | |
2587 regc('\t'); | |
2588 break; | |
2589 case CLASS_RETURN: | |
2590 regc('\r'); | |
2591 break; | |
2592 case CLASS_BACKSPACE: | |
2593 regc('\b'); | |
2594 break; | |
2595 case CLASS_ESCAPE: | |
2596 regc('\033'); | |
2597 break; | |
2598 } | |
2599 } | |
2600 else | |
2601 { | |
2602 #ifdef FEAT_MBYTE | |
2603 if (has_mbyte) | |
2604 { | |
2605 int len; | |
2606 | |
2607 /* produce a multibyte character, including any | |
2608 * following composing characters */ | |
2609 startc = mb_ptr2char(regparse); | |
474 | 2610 len = (*mb_ptr2len)(regparse); |
7 | 2611 if (enc_utf8 && utf_char2len(startc) != len) |
2612 startc = -1; /* composing chars */ | |
2613 while (--len >= 0) | |
2614 regc(*regparse++); | |
2615 } | |
2616 else | |
2617 #endif | |
2618 { | |
2619 startc = *regparse++; | |
2620 regc(startc); | |
2621 } | |
2622 } | |
2623 } | |
2624 regc(NUL); | |
2625 prevchr_len = 1; /* last char was the ']' */ | |
2626 if (*regparse != ']') | |
2627 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */ | |
2628 skipchr(); /* let's be friends with the lexer again */ | |
2629 *flagp |= HASWIDTH | SIMPLE; | |
2630 break; | |
2631 } | |
481 | 2632 else if (reg_strict) |
4444 | 2633 EMSG2_RET_NULL(_(e_missingbracket), reg_magic > MAGIC_OFF); |
7 | 2634 } |
2635 /* FALLTHROUGH */ | |
2636 | |
2637 default: | |
2638 { | |
2639 int len; | |
2640 | |
2641 #ifdef FEAT_MBYTE | |
2642 /* A multi-byte character is handled as a separate atom if it's | |
714 | 2643 * before a multi and when it's a composing char. */ |
2644 if (use_multibytecode(c)) | |
7 | 2645 { |
714 | 2646 do_multibyte: |
7 | 2647 ret = regnode(MULTIBYTECODE); |
2648 regmbc(c); | |
2649 *flagp |= HASWIDTH | SIMPLE; | |
2650 break; | |
2651 } | |
2652 #endif | |
2653 | |
2654 ret = regnode(EXACTLY); | |
2655 | |
2656 /* | |
2657 * Append characters as long as: | |
2658 * - there is no following multi, we then need the character in | |
2659 * front of it as a single character operand | |
2660 * - not running into a Magic character | |
2661 * - "one_exactly" is not set | |
2662 * But always emit at least one character. Might be a Multi, | |
2663 * e.g., a "[" without matching "]". | |
2664 */ | |
2665 for (len = 0; c != NUL && (len == 0 | |
2666 || (re_multi_type(peekchr()) == NOT_MULTI | |
2667 && !one_exactly | |
2668 && !is_Magic(c))); ++len) | |
2669 { | |
2670 c = no_Magic(c); | |
2671 #ifdef FEAT_MBYTE | |
2672 if (has_mbyte) | |
2673 { | |
2674 regmbc(c); | |
2675 if (enc_utf8) | |
2676 { | |
2677 int l; | |
2678 | |
714 | 2679 /* Need to get composing character too. */ |
7 | 2680 for (;;) |
2681 { | |
714 | 2682 l = utf_ptr2len(regparse); |
2683 if (!UTF_COMPOSINGLIKE(regparse, regparse + l)) | |
7 | 2684 break; |
714 | 2685 regmbc(utf_ptr2char(regparse)); |
2686 skipchr(); | |
7 | 2687 } |
2688 } | |
2689 } | |
2690 else | |
2691 #endif | |
2692 regc(c); | |
2693 c = getchr(); | |
2694 } | |
2695 ungetchr(); | |
2696 | |
2697 regc(NUL); | |
2698 *flagp |= HASWIDTH; | |
2699 if (len == 1) | |
2700 *flagp |= SIMPLE; | |
2701 } | |
2702 break; | |
2703 } | |
2704 | |
2705 return ret; | |
2706 } | |
2707 | |
714 | 2708 #ifdef FEAT_MBYTE |
2709 /* | |
2710 * Return TRUE if MULTIBYTECODE should be used instead of EXACTLY for | |
2711 * character "c". | |
2712 */ | |
2713 static int | |
2714 use_multibytecode(c) | |
2715 int c; | |
2716 { | |
2717 return has_mbyte && (*mb_char2len)(c) > 1 | |
2718 && (re_multi_type(peekchr()) != NOT_MULTI | |
2719 || (enc_utf8 && utf_iscomposing(c))); | |
2720 } | |
2721 #endif | |
2722 | |
7 | 2723 /* |
4444 | 2724 * Emit a node. |
7 | 2725 * Return pointer to generated code. |
2726 */ | |
2727 static char_u * | |
2728 regnode(op) | |
2729 int op; | |
2730 { | |
2731 char_u *ret; | |
2732 | |
2733 ret = regcode; | |
2734 if (ret == JUST_CALC_SIZE) | |
2735 regsize += 3; | |
2736 else | |
2737 { | |
2738 *regcode++ = op; | |
2739 *regcode++ = NUL; /* Null "next" pointer. */ | |
2740 *regcode++ = NUL; | |
2741 } | |
2742 return ret; | |
2743 } | |
2744 | |
2745 /* | |
2746 * Emit (if appropriate) a byte of code | |
2747 */ | |
2748 static void | |
2749 regc(b) | |
2750 int b; | |
2751 { | |
2752 if (regcode == JUST_CALC_SIZE) | |
2753 regsize++; | |
2754 else | |
2755 *regcode++ = b; | |
2756 } | |
2757 | |
2758 #ifdef FEAT_MBYTE | |
2759 /* | |
2760 * Emit (if appropriate) a multi-byte character of code | |
2761 */ | |
2762 static void | |
2763 regmbc(c) | |
2764 int c; | |
2765 { | |
2974 | 2766 if (!has_mbyte && c > 0xff) |
2767 return; | |
7 | 2768 if (regcode == JUST_CALC_SIZE) |
2769 regsize += (*mb_char2len)(c); | |
2770 else | |
2771 regcode += (*mb_char2bytes)(c, regcode); | |
2772 } | |
2773 #endif | |
2774 | |
2775 /* | |
4444 | 2776 * Insert an operator in front of already-emitted operand |
7 | 2777 * |
2778 * Means relocating the operand. | |
2779 */ | |
2780 static void | |
2781 reginsert(op, opnd) | |
2782 int op; | |
2783 char_u *opnd; | |
2784 { | |
2785 char_u *src; | |
2786 char_u *dst; | |
2787 char_u *place; | |
2788 | |
2789 if (regcode == JUST_CALC_SIZE) | |
2790 { | |
2791 regsize += 3; | |
2792 return; | |
2793 } | |
2794 src = regcode; | |
2795 regcode += 3; | |
2796 dst = regcode; | |
2797 while (src > opnd) | |
2798 *--dst = *--src; | |
2799 | |
2800 place = opnd; /* Op node, where operand used to be. */ | |
2801 *place++ = op; | |
2802 *place++ = NUL; | |
2803 *place = NUL; | |
2804 } | |
2805 | |
2806 /* | |
4444 | 2807 * Insert an operator in front of already-emitted operand. |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2808 * Add a number to the operator. |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2809 */ |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2810 static void |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2811 reginsert_nr(op, val, opnd) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2812 int op; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2813 long val; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2814 char_u *opnd; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2815 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2816 char_u *src; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2817 char_u *dst; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2818 char_u *place; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2819 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2820 if (regcode == JUST_CALC_SIZE) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2821 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2822 regsize += 7; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2823 return; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2824 } |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2825 src = regcode; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2826 regcode += 7; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2827 dst = regcode; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2828 while (src > opnd) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2829 *--dst = *--src; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2830 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2831 place = opnd; /* Op node, where operand used to be. */ |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2832 *place++ = op; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2833 *place++ = NUL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2834 *place++ = NUL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2835 place = re_put_long(place, (long_u)val); |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2836 } |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2837 |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2838 /* |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
2839 * Insert an operator in front of already-emitted operand. |
7 | 2840 * The operator has the given limit values as operands. Also set next pointer. |
2841 * | |
2842 * Means relocating the operand. | |
2843 */ | |
2844 static void | |
2845 reginsert_limits(op, minval, maxval, opnd) | |
2846 int op; | |
2847 long minval; | |
2848 long maxval; | |
2849 char_u *opnd; | |
2850 { | |
2851 char_u *src; | |
2852 char_u *dst; | |
2853 char_u *place; | |
2854 | |
2855 if (regcode == JUST_CALC_SIZE) | |
2856 { | |
2857 regsize += 11; | |
2858 return; | |
2859 } | |
2860 src = regcode; | |
2861 regcode += 11; | |
2862 dst = regcode; | |
2863 while (src > opnd) | |
2864 *--dst = *--src; | |
2865 | |
2866 place = opnd; /* Op node, where operand used to be. */ | |
2867 *place++ = op; | |
2868 *place++ = NUL; | |
2869 *place++ = NUL; | |
2870 place = re_put_long(place, (long_u)minval); | |
2871 place = re_put_long(place, (long_u)maxval); | |
2872 regtail(opnd, place); | |
2873 } | |
2874 | |
2875 /* | |
2876 * Write a long as four bytes at "p" and return pointer to the next char. | |
2877 */ | |
2878 static char_u * | |
2879 re_put_long(p, val) | |
2880 char_u *p; | |
2881 long_u val; | |
2882 { | |
2883 *p++ = (char_u) ((val >> 24) & 0377); | |
2884 *p++ = (char_u) ((val >> 16) & 0377); | |
2885 *p++ = (char_u) ((val >> 8) & 0377); | |
2886 *p++ = (char_u) (val & 0377); | |
2887 return p; | |
2888 } | |
2889 | |
2890 /* | |
4444 | 2891 * Set the next-pointer at the end of a node chain. |
7 | 2892 */ |
2893 static void | |
2894 regtail(p, val) | |
2895 char_u *p; | |
2896 char_u *val; | |
2897 { | |
2898 char_u *scan; | |
2899 char_u *temp; | |
2900 int offset; | |
2901 | |
2902 if (p == JUST_CALC_SIZE) | |
2903 return; | |
2904 | |
2905 /* Find last node. */ | |
2906 scan = p; | |
2907 for (;;) | |
2908 { | |
2909 temp = regnext(scan); | |
2910 if (temp == NULL) | |
2911 break; | |
2912 scan = temp; | |
2913 } | |
2914 | |
233 | 2915 if (OP(scan) == BACK) |
7 | 2916 offset = (int)(scan - val); |
2917 else | |
2918 offset = (int)(val - scan); | |
2010 | 2919 /* When the offset uses more than 16 bits it can no longer fit in the two |
2974 | 2920 * bytes available. Use a global flag to avoid having to check return |
2010 | 2921 * values in too many places. */ |
2922 if (offset > 0xffff) | |
2923 reg_toolong = TRUE; | |
2924 else | |
2925 { | |
2926 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377); | |
2927 *(scan + 2) = (char_u) (offset & 0377); | |
2928 } | |
7 | 2929 } |
2930 | |
2931 /* | |
4444 | 2932 * Like regtail, on item after a BRANCH; nop if none. |
7 | 2933 */ |
2934 static void | |
2935 regoptail(p, val) | |
2936 char_u *p; | |
2937 char_u *val; | |
2938 { | |
2939 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */ | |
2940 if (p == NULL || p == JUST_CALC_SIZE | |
2941 || (OP(p) != BRANCH | |
2942 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9))) | |
2943 return; | |
2944 regtail(OPERAND(p), val); | |
2945 } | |
2946 | |
2947 /* | |
4444 | 2948 * Functions for getting characters from the regexp input. |
7 | 2949 */ |
2950 | |
2951 static int at_start; /* True when on the first character */ | |
2952 static int prev_at_start; /* True when on the second character */ | |
2953 | |
4444 | 2954 /* |
2955 * Start parsing at "str". | |
2956 */ | |
7 | 2957 static void |
2958 initchr(str) | |
2959 char_u *str; | |
2960 { | |
2961 regparse = str; | |
2962 prevchr_len = 0; | |
2963 curchr = prevprevchr = prevchr = nextchr = -1; | |
2964 at_start = TRUE; | |
2965 prev_at_start = FALSE; | |
2966 } | |
2967 | |
4444 | 2968 /* |
4679
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2969 * Save the current parse state, so that it can be restored and parsing |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2970 * starts in the same state again. |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2971 */ |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2972 static void |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2973 save_parse_state(ps) |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2974 parse_state_T *ps; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2975 { |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2976 ps->regparse = regparse; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2977 ps->prevchr_len = prevchr_len; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2978 ps->curchr = curchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2979 ps->prevchr = prevchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2980 ps->prevprevchr = prevprevchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2981 ps->nextchr = nextchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2982 ps->at_start = at_start; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2983 ps->prev_at_start = prev_at_start; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2984 ps->regnpar = regnpar; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2985 } |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2986 |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2987 /* |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2988 * Restore a previously saved parse state. |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2989 */ |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2990 static void |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2991 restore_parse_state(ps) |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2992 parse_state_T *ps; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2993 { |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2994 regparse = ps->regparse; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2995 prevchr_len = ps->prevchr_len; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2996 curchr = ps->curchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2997 prevchr = ps->prevchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2998 prevprevchr = ps->prevprevchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
2999 nextchr = ps->nextchr; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3000 at_start = ps->at_start; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3001 prev_at_start = ps->prev_at_start; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3002 regnpar = ps->regnpar; |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3003 } |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3004 |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3005 |
4d92b873acef
updated for version 7.3.1087
Bram Moolenaar <bram@vim.org>
parents:
4579
diff
changeset
|
3006 /* |
4444 | 3007 * Get the next character without advancing. |
3008 */ | |
7 | 3009 static int |
3010 peekchr() | |
3011 { | |
167 | 3012 static int after_slash = FALSE; |
3013 | |
7 | 3014 if (curchr == -1) |
3015 { | |
3016 switch (curchr = regparse[0]) | |
3017 { | |
3018 case '.': | |
3019 case '[': | |
3020 case '~': | |
3021 /* magic when 'magic' is on */ | |
3022 if (reg_magic >= MAGIC_ON) | |
3023 curchr = Magic(curchr); | |
3024 break; | |
3025 case '(': | |
3026 case ')': | |
3027 case '{': | |
3028 case '%': | |
3029 case '+': | |
3030 case '=': | |
3031 case '?': | |
3032 case '@': | |
3033 case '!': | |
3034 case '&': | |
3035 case '|': | |
3036 case '<': | |
3037 case '>': | |
3038 case '#': /* future ext. */ | |
3039 case '"': /* future ext. */ | |
3040 case '\'': /* future ext. */ | |
3041 case ',': /* future ext. */ | |
3042 case '-': /* future ext. */ | |
3043 case ':': /* future ext. */ | |
3044 case ';': /* future ext. */ | |
3045 case '`': /* future ext. */ | |
3046 case '/': /* Can't be used in / command */ | |
3047 /* magic only after "\v" */ | |
3048 if (reg_magic == MAGIC_ALL) | |
3049 curchr = Magic(curchr); | |
3050 break; | |
3051 case '*': | |
167 | 3052 /* * is not magic as the very first character, eg "?*ptr", when |
3053 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But | |
3054 * "\(\*" is not magic, thus must be magic if "after_slash" */ | |
3055 if (reg_magic >= MAGIC_ON | |
3056 && !at_start | |
3057 && !(prev_at_start && prevchr == Magic('^')) | |
3058 && (after_slash | |
3059 || (prevchr != Magic('(') | |
3060 && prevchr != Magic('&') | |
3061 && prevchr != Magic('|')))) | |
7 | 3062 curchr = Magic('*'); |
3063 break; | |
3064 case '^': | |
3065 /* '^' is only magic as the very first character and if it's after | |
3066 * "\(", "\|", "\&' or "\n" */ | |
3067 if (reg_magic >= MAGIC_OFF | |
3068 && (at_start | |
3069 || reg_magic == MAGIC_ALL | |
3070 || prevchr == Magic('(') | |
3071 || prevchr == Magic('|') | |
3072 || prevchr == Magic('&') | |
3073 || prevchr == Magic('n') | |
3074 || (no_Magic(prevchr) == '(' | |
3075 && prevprevchr == Magic('%')))) | |
3076 { | |
3077 curchr = Magic('^'); | |
3078 at_start = TRUE; | |
3079 prev_at_start = FALSE; | |
3080 } | |
3081 break; | |
3082 case '$': | |
3083 /* '$' is only magic as the very last char and if it's in front of | |
3084 * either "\|", "\)", "\&", or "\n" */ | |
3085 if (reg_magic >= MAGIC_OFF) | |
3086 { | |
3087 char_u *p = regparse + 1; | |
3088 | |
3089 /* ignore \c \C \m and \M after '$' */ | |
3090 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C' | |
3091 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z')) | |
3092 p += 2; | |
3093 if (p[0] == NUL | |
3094 || (p[0] == '\\' | |
3095 && (p[1] == '|' || p[1] == '&' || p[1] == ')' | |
3096 || p[1] == 'n')) | |
3097 || reg_magic == MAGIC_ALL) | |
3098 curchr = Magic('$'); | |
3099 } | |
3100 break; | |
3101 case '\\': | |
3102 { | |
3103 int c = regparse[1]; | |
3104 | |
3105 if (c == NUL) | |
3106 curchr = '\\'; /* trailing '\' */ | |
3107 else if ( | |
3108 #ifdef EBCDIC | |
3109 vim_strchr(META, c) | |
3110 #else | |
3111 c <= '~' && META_flags[c] | |
3112 #endif | |
3113 ) | |
3114 { | |
3115 /* | |
3116 * META contains everything that may be magic sometimes, | |
3117 * except ^ and $ ("\^" and "\$" are only magic after | |
3118 * "\v"). We now fetch the next character and toggle its | |
3119 * magicness. Therefore, \ is so meta-magic that it is | |
3120 * not in META. | |
3121 */ | |
3122 curchr = -1; | |
3123 prev_at_start = at_start; | |
3124 at_start = FALSE; /* be able to say "/\*ptr" */ | |
3125 ++regparse; | |
167 | 3126 ++after_slash; |
7 | 3127 peekchr(); |
3128 --regparse; | |
167 | 3129 --after_slash; |
7 | 3130 curchr = toggle_Magic(curchr); |
3131 } | |
3132 else if (vim_strchr(REGEXP_ABBR, c)) | |
3133 { | |
3134 /* | |
3135 * Handle abbreviations, like "\t" for TAB -- webb | |
3136 */ | |
3137 curchr = backslash_trans(c); | |
3138 } | |
3139 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^')) | |
3140 curchr = toggle_Magic(c); | |
3141 else | |
3142 { | |
3143 /* | |
3144 * Next character can never be (made) magic? | |
3145 * Then backslashing it won't do anything. | |
3146 */ | |
3147 #ifdef FEAT_MBYTE | |
3148 if (has_mbyte) | |
3149 curchr = (*mb_ptr2char)(regparse + 1); | |
3150 else | |
3151 #endif | |
3152 curchr = c; | |
3153 } | |
3154 break; | |
3155 } | |
3156 | |
3157 #ifdef FEAT_MBYTE | |
3158 default: | |
3159 if (has_mbyte) | |
3160 curchr = (*mb_ptr2char)(regparse); | |
3161 #endif | |
3162 } | |
3163 } | |
3164 | |
3165 return curchr; | |
3166 } | |
3167 | |
3168 /* | |
3169 * Eat one lexed character. Do this in a way that we can undo it. | |
3170 */ | |
3171 static void | |
3172 skipchr() | |
3173 { | |
3174 /* peekchr() eats a backslash, do the same here */ | |
3175 if (*regparse == '\\') | |
3176 prevchr_len = 1; | |
3177 else | |
3178 prevchr_len = 0; | |
3179 if (regparse[prevchr_len] != NUL) | |
3180 { | |
3181 #ifdef FEAT_MBYTE | |
714 | 3182 if (enc_utf8) |
1449 | 3183 /* exclude composing chars that mb_ptr2len does include */ |
3184 prevchr_len += utf_ptr2len(regparse + prevchr_len); | |
714 | 3185 else if (has_mbyte) |
474 | 3186 prevchr_len += (*mb_ptr2len)(regparse + prevchr_len); |
7 | 3187 else |
3188 #endif | |
3189 ++prevchr_len; | |
3190 } | |
3191 regparse += prevchr_len; | |
3192 prev_at_start = at_start; | |
3193 at_start = FALSE; | |
3194 prevprevchr = prevchr; | |
3195 prevchr = curchr; | |
3196 curchr = nextchr; /* use previously unget char, or -1 */ | |
3197 nextchr = -1; | |
3198 } | |
3199 | |
3200 /* | |
3201 * Skip a character while keeping the value of prev_at_start for at_start. | |
3202 * prevchr and prevprevchr are also kept. | |
3203 */ | |
3204 static void | |
3205 skipchr_keepstart() | |
3206 { | |
3207 int as = prev_at_start; | |
3208 int pr = prevchr; | |
3209 int prpr = prevprevchr; | |
3210 | |
3211 skipchr(); | |
3212 at_start = as; | |
3213 prevchr = pr; | |
3214 prevprevchr = prpr; | |
3215 } | |
3216 | |
4444 | 3217 /* |
3218 * Get the next character from the pattern. We know about magic and such, so | |
3219 * therefore we need a lexical analyzer. | |
3220 */ | |
7 | 3221 static int |
3222 getchr() | |
3223 { | |
3224 int chr = peekchr(); | |
3225 | |
3226 skipchr(); | |
3227 return chr; | |
3228 } | |
3229 | |
3230 /* | |
3231 * put character back. Works only once! | |
3232 */ | |
3233 static void | |
3234 ungetchr() | |
3235 { | |
3236 nextchr = curchr; | |
3237 curchr = prevchr; | |
3238 prevchr = prevprevchr; | |
3239 at_start = prev_at_start; | |
3240 prev_at_start = FALSE; | |
3241 | |
3242 /* Backup regparse, so that it's at the same position as before the | |
3243 * getchr(). */ | |
3244 regparse -= prevchr_len; | |
3245 } | |
3246 | |
3247 /* | |
29 | 3248 * Get and return the value of the hex string at the current position. |
3249 * Return -1 if there is no valid hex number. | |
3250 * The position is updated: | |
24 | 3251 * blahblah\%x20asdf |
856 | 3252 * before-^ ^-after |
24 | 3253 * The parameter controls the maximum number of input characters. This will be |
3254 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence. | |
3255 */ | |
3256 static int | |
3257 gethexchrs(maxinputlen) | |
3258 int maxinputlen; | |
3259 { | |
3260 int nr = 0; | |
3261 int c; | |
3262 int i; | |
3263 | |
3264 for (i = 0; i < maxinputlen; ++i) | |
3265 { | |
3266 c = regparse[0]; | |
3267 if (!vim_isxdigit(c)) | |
3268 break; | |
3269 nr <<= 4; | |
3270 nr |= hex2nr(c); | |
3271 ++regparse; | |
3272 } | |
3273 | |
3274 if (i == 0) | |
3275 return -1; | |
3276 return nr; | |
3277 } | |
3278 | |
3279 /* | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
3280 * Get and return the value of the decimal string immediately after the |
24 | 3281 * current position. Return -1 for invalid. Consumes all digits. |
3282 */ | |
3283 static int | |
3284 getdecchrs() | |
3285 { | |
3286 int nr = 0; | |
3287 int c; | |
3288 int i; | |
3289 | |
3290 for (i = 0; ; ++i) | |
3291 { | |
3292 c = regparse[0]; | |
3293 if (c < '0' || c > '9') | |
3294 break; | |
3295 nr *= 10; | |
3296 nr += c - '0'; | |
3297 ++regparse; | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
3298 curchr = -1; /* no longer valid */ |
24 | 3299 } |
3300 | |
3301 if (i == 0) | |
3302 return -1; | |
3303 return nr; | |
3304 } | |
3305 | |
3306 /* | |
3307 * get and return the value of the octal string immediately after the current | |
3308 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle | |
3309 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't | |
3310 * treat 8 or 9 as recognised characters. Position is updated: | |
3311 * blahblah\%o210asdf | |
856 | 3312 * before-^ ^-after |
24 | 3313 */ |
3314 static int | |
3315 getoctchrs() | |
3316 { | |
3317 int nr = 0; | |
3318 int c; | |
3319 int i; | |
3320 | |
3321 for (i = 0; i < 3 && nr < 040; ++i) | |
3322 { | |
3323 c = regparse[0]; | |
3324 if (c < '0' || c > '7') | |
3325 break; | |
3326 nr <<= 3; | |
3327 nr |= hex2nr(c); | |
3328 ++regparse; | |
3329 } | |
3330 | |
3331 if (i == 0) | |
3332 return -1; | |
3333 return nr; | |
3334 } | |
3335 | |
3336 /* | |
3337 * Get a number after a backslash that is inside []. | |
3338 * When nothing is recognized return a backslash. | |
3339 */ | |
3340 static int | |
3341 coll_get_char() | |
3342 { | |
3343 int nr = -1; | |
3344 | |
3345 switch (*regparse++) | |
3346 { | |
3347 case 'd': nr = getdecchrs(); break; | |
3348 case 'o': nr = getoctchrs(); break; | |
3349 case 'x': nr = gethexchrs(2); break; | |
3350 case 'u': nr = gethexchrs(4); break; | |
3351 case 'U': nr = gethexchrs(8); break; | |
3352 } | |
3353 if (nr < 0) | |
3354 { | |
3355 /* If getting the number fails be backwards compatible: the character | |
3356 * is a backslash. */ | |
3357 --regparse; | |
3358 nr = '\\'; | |
3359 } | |
3360 return nr; | |
3361 } | |
3362 | |
3363 /* | |
7 | 3364 * read_limits - Read two integers to be taken as a minimum and maximum. |
3365 * If the first character is '-', then the range is reversed. | |
3366 * Should end with 'end'. If minval is missing, zero is default, if maxval is | |
3367 * missing, a very big number is the default. | |
3368 */ | |
3369 static int | |
3370 read_limits(minval, maxval) | |
3371 long *minval; | |
3372 long *maxval; | |
3373 { | |
3374 int reverse = FALSE; | |
3375 char_u *first_char; | |
3376 long tmp; | |
3377 | |
3378 if (*regparse == '-') | |
3379 { | |
3380 /* Starts with '-', so reverse the range later */ | |
3381 regparse++; | |
3382 reverse = TRUE; | |
3383 } | |
3384 first_char = regparse; | |
3385 *minval = getdigits(®parse); | |
3386 if (*regparse == ',') /* There is a comma */ | |
3387 { | |
3388 if (vim_isdigit(*++regparse)) | |
3389 *maxval = getdigits(®parse); | |
3390 else | |
3391 *maxval = MAX_LIMIT; | |
3392 } | |
3393 else if (VIM_ISDIGIT(*first_char)) | |
3394 *maxval = *minval; /* It was \{n} or \{-n} */ | |
3395 else | |
3396 *maxval = MAX_LIMIT; /* It was \{} or \{-} */ | |
3397 if (*regparse == '\\') | |
3398 regparse++; /* Allow either \{...} or \{...\} */ | |
167 | 3399 if (*regparse != '}') |
7 | 3400 { |
3401 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"), | |
3402 reg_magic == MAGIC_ALL ? "" : "\\"); | |
3403 EMSG_RET_FAIL(IObuff); | |
3404 } | |
3405 | |
3406 /* | |
3407 * Reverse the range if there was a '-', or make sure it is in the right | |
3408 * order otherwise. | |
3409 */ | |
3410 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval)) | |
3411 { | |
3412 tmp = *minval; | |
3413 *minval = *maxval; | |
3414 *maxval = tmp; | |
3415 } | |
3416 skipchr(); /* let's be friends with the lexer again */ | |
3417 return OK; | |
3418 } | |
3419 | |
3420 /* | |
3421 * vim_regexec and friends | |
3422 */ | |
3423 | |
3424 /* | |
3425 * Global work variables for vim_regexec(). | |
3426 */ | |
3427 | |
3428 /* The current match-position is remembered with these variables: */ | |
3429 static linenr_T reglnum; /* line number, relative to first line */ | |
3430 static char_u *regline; /* start of current line */ | |
3431 static char_u *reginput; /* current input, points into "regline" */ | |
3432 | |
3433 static int need_clear_subexpr; /* subexpressions still need to be | |
3434 * cleared */ | |
3435 #ifdef FEAT_SYN_HL | |
3436 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions | |
3437 * still need to be cleared */ | |
3438 #endif | |
3439 | |
3440 /* | |
3441 * Structure used to save the current input state, when it needs to be | |
3442 * restored after trying a match. Used by reg_save() and reg_restore(). | |
233 | 3443 * Also stores the length of "backpos". |
7 | 3444 */ |
3445 typedef struct | |
3446 { | |
3447 union | |
3448 { | |
3449 char_u *ptr; /* reginput pointer, for single-line regexp */ | |
3450 lpos_T pos; /* reginput pos, for multi-line regexp */ | |
3451 } rs_u; | |
233 | 3452 int rs_len; |
7 | 3453 } regsave_T; |
3454 | |
3455 /* struct to save start/end pointer/position in for \(\) */ | |
3456 typedef struct | |
3457 { | |
3458 union | |
3459 { | |
3460 char_u *ptr; | |
3461 lpos_T pos; | |
3462 } se_u; | |
3463 } save_se_T; | |
3464 | |
1579 | 3465 /* used for BEHIND and NOBEHIND matching */ |
3466 typedef struct regbehind_S | |
3467 { | |
3468 regsave_T save_after; | |
3469 regsave_T save_behind; | |
1602 | 3470 int save_need_clear_subexpr; |
1579 | 3471 save_se_T save_start[NSUBEXP]; |
3472 save_se_T save_end[NSUBEXP]; | |
3473 } regbehind_T; | |
3474 | |
7 | 3475 static char_u *reg_getline __ARGS((linenr_T lnum)); |
4444 | 3476 static long bt_regexec_both __ARGS((char_u *line, colnr_T col, proftime_T *tm)); |
3477 static long regtry __ARGS((bt_regprog_T *prog, colnr_T col)); | |
7 | 3478 static void cleanup_subexpr __ARGS((void)); |
3479 #ifdef FEAT_SYN_HL | |
3480 static void cleanup_zsubexpr __ARGS((void)); | |
3481 #endif | |
1579 | 3482 static void save_subexpr __ARGS((regbehind_T *bp)); |
3483 static void restore_subexpr __ARGS((regbehind_T *bp)); | |
7 | 3484 static void reg_nextline __ARGS((void)); |
233 | 3485 static void reg_save __ARGS((regsave_T *save, garray_T *gap)); |
3486 static void reg_restore __ARGS((regsave_T *save, garray_T *gap)); | |
7 | 3487 static int reg_save_equal __ARGS((regsave_T *save)); |
3488 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp)); | |
3489 static void save_se_one __ARGS((save_se_T *savep, char_u **pp)); | |
3490 | |
3491 /* Save the sub-expressions before attempting a match. */ | |
3492 #define save_se(savep, posp, pp) \ | |
3493 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp)) | |
3494 | |
3495 /* After a failed match restore the sub-expressions. */ | |
3496 #define restore_se(savep, posp, pp) { \ | |
3497 if (REG_MULTI) \ | |
3498 *(posp) = (savep)->se_u.pos; \ | |
3499 else \ | |
3500 *(pp) = (savep)->se_u.ptr; } | |
3501 | |
3502 static int re_num_cmp __ARGS((long_u val, char_u *scan)); | |
180 | 3503 static int regmatch __ARGS((char_u *prog)); |
7 | 3504 static int regrepeat __ARGS((char_u *p, long maxcount)); |
3505 | |
3506 #ifdef DEBUG | |
3507 int regnarrate = 0; | |
3508 #endif | |
3509 | |
3510 /* | |
3511 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec(). | |
3512 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern | |
3513 * contains '\c' or '\C' the value is overruled. | |
3514 */ | |
3515 static int ireg_ic; | |
3516 | |
3517 #ifdef FEAT_MBYTE | |
3518 /* | |
3519 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag | |
3520 * in the regexp. Defaults to false, always. | |
3521 */ | |
3522 static int ireg_icombine; | |
3523 #endif | |
3524 | |
3525 /* | |
410 | 3526 * Copy of "rmm_maxcol": maximum column to search for a match. Zero when |
3527 * there is no maximum. | |
3528 */ | |
418 | 3529 static colnr_T ireg_maxcol; |
410 | 3530 |
3531 /* | |
7 | 3532 * Sometimes need to save a copy of a line. Since alloc()/free() is very |
3533 * slow, we keep one allocated piece of memory and only re-allocate it when | |
4444 | 3534 * it's too small. It's freed in bt_regexec_both() when finished. |
7 | 3535 */ |
1468 | 3536 static char_u *reg_tofree = NULL; |
7 | 3537 static unsigned reg_tofreelen; |
3538 | |
3539 /* | |
3540 * These variables are set when executing a regexp to speed up the execution. | |
1209 | 3541 * Which ones are set depends on whether a single-line or multi-line match is |
7 | 3542 * done: |
3543 * single-line multi-line | |
3544 * reg_match ®match_T NULL | |
3545 * reg_mmatch NULL ®mmatch_T | |
3546 * reg_startp reg_match->startp <invalid> | |
3547 * reg_endp reg_match->endp <invalid> | |
3548 * reg_startpos <invalid> reg_mmatch->startpos | |
3549 * reg_endpos <invalid> reg_mmatch->endpos | |
3550 * reg_win NULL window in which to search | |
4061 | 3551 * reg_buf curbuf buffer in which to search |
7 | 3552 * reg_firstlnum <invalid> first line in which to search |
3553 * reg_maxline 0 last line nr | |
3554 * reg_line_lbr FALSE or TRUE FALSE | |
3555 */ | |
3556 static regmatch_T *reg_match; | |
3557 static regmmatch_T *reg_mmatch; | |
3558 static char_u **reg_startp = NULL; | |
3559 static char_u **reg_endp = NULL; | |
3560 static lpos_T *reg_startpos = NULL; | |
3561 static lpos_T *reg_endpos = NULL; | |
3562 static win_T *reg_win; | |
3563 static buf_T *reg_buf; | |
3564 static linenr_T reg_firstlnum; | |
3565 static linenr_T reg_maxline; | |
3566 static int reg_line_lbr; /* "\n" in string is line break */ | |
3567 | |
270 | 3568 /* Values for rs_state in regitem_T. */ |
3569 typedef enum regstate_E | |
3570 { | |
3571 RS_NOPEN = 0 /* NOPEN and NCLOSE */ | |
3572 , RS_MOPEN /* MOPEN + [0-9] */ | |
3573 , RS_MCLOSE /* MCLOSE + [0-9] */ | |
3574 #ifdef FEAT_SYN_HL | |
3575 , RS_ZOPEN /* ZOPEN + [0-9] */ | |
3576 , RS_ZCLOSE /* ZCLOSE + [0-9] */ | |
3577 #endif | |
3578 , RS_BRANCH /* BRANCH */ | |
3579 , RS_BRCPLX_MORE /* BRACE_COMPLEX and trying one more match */ | |
3580 , RS_BRCPLX_LONG /* BRACE_COMPLEX and trying longest match */ | |
3581 , RS_BRCPLX_SHORT /* BRACE_COMPLEX and trying shortest match */ | |
3582 , RS_NOMATCH /* NOMATCH */ | |
3583 , RS_BEHIND1 /* BEHIND / NOBEHIND matching rest */ | |
3584 , RS_BEHIND2 /* BEHIND / NOBEHIND matching behind part */ | |
3585 , RS_STAR_LONG /* STAR/PLUS/BRACE_SIMPLE longest match */ | |
3586 , RS_STAR_SHORT /* STAR/PLUS/BRACE_SIMPLE shortest match */ | |
3587 } regstate_T; | |
3588 | |
3589 /* | |
3590 * When there are alternatives a regstate_T is put on the regstack to remember | |
3591 * what we are doing. | |
3592 * Before it may be another type of item, depending on rs_state, to remember | |
3593 * more things. | |
3594 */ | |
3595 typedef struct regitem_S | |
3596 { | |
3597 regstate_T rs_state; /* what we are doing, one of RS_ above */ | |
3598 char_u *rs_scan; /* current node in program */ | |
3599 union | |
3600 { | |
3601 save_se_T sesave; | |
3602 regsave_T regsave; | |
3603 } rs_un; /* room for saving reginput */ | |
1579 | 3604 short rs_no; /* submatch nr or BEHIND/NOBEHIND */ |
270 | 3605 } regitem_T; |
3606 | |
3607 static regitem_T *regstack_push __ARGS((regstate_T state, char_u *scan)); | |
3608 static void regstack_pop __ARGS((char_u **scan)); | |
3609 | |
3610 /* used for STAR, PLUS and BRACE_SIMPLE matching */ | |
3611 typedef struct regstar_S | |
3612 { | |
3613 int nextb; /* next byte */ | |
3614 int nextb_ic; /* next byte reverse case */ | |
3615 long count; | |
3616 long minval; | |
3617 long maxval; | |
3618 } regstar_T; | |
3619 | |
3620 /* used to store input position when a BACK was encountered, so that we now if | |
3621 * we made any progress since the last time. */ | |
3622 typedef struct backpos_S | |
3623 { | |
3624 char_u *bp_scan; /* "scan" where BACK was encountered */ | |
3625 regsave_T bp_pos; /* last input position */ | |
3626 } backpos_T; | |
3627 | |
3628 /* | |
1520 | 3629 * "regstack" and "backpos" are used by regmatch(). They are kept over calls |
3630 * to avoid invoking malloc() and free() often. | |
3631 * "regstack" is a stack with regitem_T items, sometimes preceded by regstar_T | |
3632 * or regbehind_T. | |
3633 * "backpos_T" is a table with backpos_T for BACK | |
270 | 3634 */ |
1520 | 3635 static garray_T regstack = {0, 0, 0, 0, NULL}; |
3636 static garray_T backpos = {0, 0, 0, 0, NULL}; | |
3637 | |
3638 /* | |
3639 * Both for regstack and backpos tables we use the following strategy of | |
3640 * allocation (to reduce malloc/free calls): | |
3641 * - Initial size is fairly small. | |
3642 * - When needed, the tables are grown bigger (8 times at first, double after | |
3643 * that). | |
3644 * - After executing the match we free the memory only if the array has grown. | |
3645 * Thus the memory is kept allocated when it's at the initial size. | |
3646 * This makes it fast while not keeping a lot of memory allocated. | |
3647 * A three times speed increase was observed when using many simple patterns. | |
3648 */ | |
3649 #define REGSTACK_INITIAL 2048 | |
3650 #define BACKPOS_INITIAL 64 | |
3651 | |
3652 #if defined(EXITFREE) || defined(PROTO) | |
3653 void | |
3654 free_regexp_stuff() | |
3655 { | |
3656 ga_clear(®stack); | |
3657 ga_clear(&backpos); | |
3658 vim_free(reg_tofree); | |
3659 vim_free(reg_prev_sub); | |
3660 } | |
3661 #endif | |
270 | 3662 |
7 | 3663 /* |
3664 * Get pointer to the line "lnum", which is relative to "reg_firstlnum". | |
3665 */ | |
3666 static char_u * | |
3667 reg_getline(lnum) | |
3668 linenr_T lnum; | |
3669 { | |
3670 /* when looking behind for a match/no-match lnum is negative. But we | |
3671 * can't go before line 1 */ | |
3672 if (reg_firstlnum + lnum < 1) | |
3673 return NULL; | |
482 | 3674 if (lnum > reg_maxline) |
481 | 3675 /* Must have matched the "\n" in the last line. */ |
3676 return (char_u *)""; | |
7 | 3677 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE); |
3678 } | |
3679 | |
3680 static regsave_T behind_pos; | |
3681 | |
3682 #ifdef FEAT_SYN_HL | |
3683 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */ | |
3684 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */ | |
3685 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */ | |
3686 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */ | |
3687 #endif | |
3688 | |
3689 /* TRUE if using multi-line regexp. */ | |
3690 #define REG_MULTI (reg_match == NULL) | |
3691 | |
4444 | 3692 static int bt_regexec __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); |
3693 | |
7 | 3694 /* |
3695 * Match a regexp against a string. | |
3696 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3697 * Uses curbuf for line count and 'iskeyword'. | |
3698 * | |
3699 * Return TRUE if there is a match, FALSE if not. | |
3700 */ | |
4444 | 3701 static int |
3702 bt_regexec(rmp, line, col) | |
7 | 3703 regmatch_T *rmp; |
3704 char_u *line; /* string to match against */ | |
3705 colnr_T col; /* column to start looking for match */ | |
3706 { | |
3707 reg_match = rmp; | |
3708 reg_mmatch = NULL; | |
3709 reg_maxline = 0; | |
3710 reg_line_lbr = FALSE; | |
4061 | 3711 reg_buf = curbuf; |
7 | 3712 reg_win = NULL; |
3713 ireg_ic = rmp->rm_ic; | |
3714 #ifdef FEAT_MBYTE | |
3715 ireg_icombine = FALSE; | |
3716 #endif | |
410 | 3717 ireg_maxcol = 0; |
4444 | 3718 return (bt_regexec_both(line, col, NULL) != 0); |
7 | 3719 } |
3720 | |
39 | 3721 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ |
3722 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
4444 | 3723 |
3724 static int bt_regexec_nl __ARGS((regmatch_T *rmp, char_u *line, colnr_T col)); | |
3725 | |
7 | 3726 /* |
3727 * Like vim_regexec(), but consider a "\n" in "line" to be a line break. | |
3728 */ | |
4444 | 3729 static int |
3730 bt_regexec_nl(rmp, line, col) | |
7 | 3731 regmatch_T *rmp; |
3732 char_u *line; /* string to match against */ | |
3733 colnr_T col; /* column to start looking for match */ | |
3734 { | |
3735 reg_match = rmp; | |
3736 reg_mmatch = NULL; | |
3737 reg_maxline = 0; | |
3738 reg_line_lbr = TRUE; | |
4061 | 3739 reg_buf = curbuf; |
7 | 3740 reg_win = NULL; |
3741 ireg_ic = rmp->rm_ic; | |
3742 #ifdef FEAT_MBYTE | |
3743 ireg_icombine = FALSE; | |
3744 #endif | |
410 | 3745 ireg_maxcol = 0; |
4444 | 3746 return (bt_regexec_both(line, col, NULL) != 0); |
7 | 3747 } |
3748 #endif | |
3749 | |
4444 | 3750 static long bt_regexec_multi __ARGS((regmmatch_T *rmp, win_T *win, buf_T *buf, linenr_T lnum, colnr_T col, proftime_T *tm)); |
3751 | |
7 | 3752 /* |
3753 * Match a regexp against multiple lines. | |
3754 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
3755 * Uses curbuf for line count and 'iskeyword'. | |
3756 * | |
3757 * Return zero if there is no match. Return number of lines contained in the | |
3758 * match otherwise. | |
3759 */ | |
4444 | 3760 static long |
3761 bt_regexec_multi(rmp, win, buf, lnum, col, tm) | |
7 | 3762 regmmatch_T *rmp; |
3763 win_T *win; /* window in which to search or NULL */ | |
3764 buf_T *buf; /* buffer in which to search */ | |
3765 linenr_T lnum; /* nr of line to start looking for match */ | |
3766 colnr_T col; /* column to start looking for match */ | |
1521 | 3767 proftime_T *tm; /* timeout limit or NULL */ |
7 | 3768 { |
3769 long r; | |
3770 | |
3771 reg_match = NULL; | |
3772 reg_mmatch = rmp; | |
3773 reg_buf = buf; | |
3774 reg_win = win; | |
3775 reg_firstlnum = lnum; | |
3776 reg_maxline = reg_buf->b_ml.ml_line_count - lnum; | |
3777 reg_line_lbr = FALSE; | |
3778 ireg_ic = rmp->rmm_ic; | |
3779 #ifdef FEAT_MBYTE | |
3780 ireg_icombine = FALSE; | |
3781 #endif | |
410 | 3782 ireg_maxcol = rmp->rmm_maxcol; |
7 | 3783 |
4444 | 3784 r = bt_regexec_both(NULL, col, tm); |
7 | 3785 |
3786 return r; | |
3787 } | |
3788 | |
3789 /* | |
3790 * Match a regexp against a string ("line" points to the string) or multiple | |
3791 * lines ("line" is NULL, use reg_getline()). | |
3792 */ | |
3793 static long | |
4444 | 3794 bt_regexec_both(line, col, tm) |
7 | 3795 char_u *line; |
3796 colnr_T col; /* column to start looking for match */ | |
1877 | 3797 proftime_T *tm UNUSED; /* timeout limit or NULL */ |
7 | 3798 { |
4444 | 3799 bt_regprog_T *prog; |
7 | 3800 char_u *s; |
189 | 3801 long retval = 0L; |
7 | 3802 |
1520 | 3803 /* Create "regstack" and "backpos" if they are not allocated yet. |
3804 * We allocate *_INITIAL amount of bytes first and then set the grow size | |
3805 * to much bigger value to avoid many malloc calls in case of deep regular | |
3806 * expressions. */ | |
3807 if (regstack.ga_data == NULL) | |
3808 { | |
3809 /* Use an item size of 1 byte, since we push different things | |
3810 * onto the regstack. */ | |
3811 ga_init2(®stack, 1, REGSTACK_INITIAL); | |
3812 ga_grow(®stack, REGSTACK_INITIAL); | |
3813 regstack.ga_growsize = REGSTACK_INITIAL * 8; | |
3814 } | |
3815 | |
3816 if (backpos.ga_data == NULL) | |
3817 { | |
3818 ga_init2(&backpos, sizeof(backpos_T), BACKPOS_INITIAL); | |
3819 ga_grow(&backpos, BACKPOS_INITIAL); | |
3820 backpos.ga_growsize = BACKPOS_INITIAL * 8; | |
3821 } | |
270 | 3822 |
7 | 3823 if (REG_MULTI) |
3824 { | |
4444 | 3825 prog = (bt_regprog_T *)reg_mmatch->regprog; |
7 | 3826 line = reg_getline((linenr_T)0); |
3827 reg_startpos = reg_mmatch->startpos; | |
3828 reg_endpos = reg_mmatch->endpos; | |
3829 } | |
3830 else | |
3831 { | |
4444 | 3832 prog = (bt_regprog_T *)reg_match->regprog; |
7 | 3833 reg_startp = reg_match->startp; |
3834 reg_endp = reg_match->endp; | |
3835 } | |
3836 | |
3837 /* Be paranoid... */ | |
3838 if (prog == NULL || line == NULL) | |
3839 { | |
3840 EMSG(_(e_null)); | |
3841 goto theend; | |
3842 } | |
3843 | |
3844 /* Check validity of program. */ | |
3845 if (prog_magic_wrong()) | |
3846 goto theend; | |
3847 | |
410 | 3848 /* If the start column is past the maximum column: no need to try. */ |
3849 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3850 goto theend; | |
3851 | |
7 | 3852 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */ |
3853 if (prog->regflags & RF_ICASE) | |
3854 ireg_ic = TRUE; | |
3855 else if (prog->regflags & RF_NOICASE) | |
3856 ireg_ic = FALSE; | |
3857 | |
3858 #ifdef FEAT_MBYTE | |
3859 /* If pattern contains "\Z" overrule value of ireg_icombine */ | |
3860 if (prog->regflags & RF_ICOMBINE) | |
3861 ireg_icombine = TRUE; | |
3862 #endif | |
3863 | |
3864 /* If there is a "must appear" string, look for it. */ | |
3865 if (prog->regmust != NULL) | |
3866 { | |
3867 int c; | |
3868 | |
3869 #ifdef FEAT_MBYTE | |
3870 if (has_mbyte) | |
3871 c = (*mb_ptr2char)(prog->regmust); | |
3872 else | |
3873 #endif | |
3874 c = *prog->regmust; | |
3875 s = line + col; | |
170 | 3876 |
3877 /* | |
3878 * This is used very often, esp. for ":global". Use three versions of | |
3879 * the loop to avoid overhead of conditions. | |
3880 */ | |
3881 if (!ireg_ic | |
3882 #ifdef FEAT_MBYTE | |
3883 && !has_mbyte | |
3884 #endif | |
3885 ) | |
3886 while ((s = vim_strbyte(s, c)) != NULL) | |
3887 { | |
3888 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3889 break; /* Found it. */ | |
3890 ++s; | |
3891 } | |
3892 #ifdef FEAT_MBYTE | |
3893 else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1)) | |
3894 while ((s = vim_strchr(s, c)) != NULL) | |
3895 { | |
3896 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3897 break; /* Found it. */ | |
3898 mb_ptr_adv(s); | |
3899 } | |
3900 #endif | |
3901 else | |
3902 while ((s = cstrchr(s, c)) != NULL) | |
3903 { | |
3904 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) | |
3905 break; /* Found it. */ | |
3906 mb_ptr_adv(s); | |
3907 } | |
7 | 3908 if (s == NULL) /* Not present. */ |
3909 goto theend; | |
3910 } | |
3911 | |
3912 regline = line; | |
3913 reglnum = 0; | |
2578 | 3914 reg_toolong = FALSE; |
7 | 3915 |
3916 /* Simplest case: Anchored match need be tried only once. */ | |
3917 if (prog->reganch) | |
3918 { | |
3919 int c; | |
3920 | |
3921 #ifdef FEAT_MBYTE | |
3922 if (has_mbyte) | |
3923 c = (*mb_ptr2char)(regline + col); | |
3924 else | |
3925 #endif | |
3926 c = regline[col]; | |
3927 if (prog->regstart == NUL | |
3928 || prog->regstart == c | |
3929 || (ireg_ic && (( | |
3930 #ifdef FEAT_MBYTE | |
3931 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c))) | |
3932 || (c < 255 && prog->regstart < 255 && | |
3933 #endif | |
1347 | 3934 MB_TOLOWER(prog->regstart) == MB_TOLOWER(c))))) |
7 | 3935 retval = regtry(prog, col); |
3936 else | |
3937 retval = 0; | |
3938 } | |
3939 else | |
3940 { | |
1521 | 3941 #ifdef FEAT_RELTIME |
3942 int tm_count = 0; | |
3943 #endif | |
7 | 3944 /* Messy cases: unanchored match. */ |
180 | 3945 while (!got_int) |
7 | 3946 { |
3947 if (prog->regstart != NUL) | |
3948 { | |
170 | 3949 /* Skip until the char we know it must start with. |
3950 * Used often, do some work to avoid call overhead. */ | |
3951 if (!ireg_ic | |
3952 #ifdef FEAT_MBYTE | |
3953 && !has_mbyte | |
3954 #endif | |
3955 ) | |
3956 s = vim_strbyte(regline + col, prog->regstart); | |
3957 else | |
3958 s = cstrchr(regline + col, prog->regstart); | |
7 | 3959 if (s == NULL) |
3960 { | |
3961 retval = 0; | |
3962 break; | |
3963 } | |
3964 col = (int)(s - regline); | |
3965 } | |
3966 | |
410 | 3967 /* Check for maximum column to try. */ |
3968 if (ireg_maxcol > 0 && col >= ireg_maxcol) | |
3969 { | |
3970 retval = 0; | |
3971 break; | |
3972 } | |
3973 | |
7 | 3974 retval = regtry(prog, col); |
3975 if (retval > 0) | |
3976 break; | |
3977 | |
3978 /* if not currently on the first line, get it again */ | |
3979 if (reglnum != 0) | |
3980 { | |
481 | 3981 reglnum = 0; |
7 | 3982 regline = reg_getline((linenr_T)0); |
3983 } | |
3984 if (regline[col] == NUL) | |
3985 break; | |
3986 #ifdef FEAT_MBYTE | |
3987 if (has_mbyte) | |
474 | 3988 col += (*mb_ptr2len)(regline + col); |
7 | 3989 else |
3990 #endif | |
3991 ++col; | |
1521 | 3992 #ifdef FEAT_RELTIME |
3993 /* Check for timeout once in a twenty times to avoid overhead. */ | |
3994 if (tm != NULL && ++tm_count == 20) | |
3995 { | |
3996 tm_count = 0; | |
3997 if (profile_passed_limit(tm)) | |
3998 break; | |
3999 } | |
4000 #endif | |
7 | 4001 } |
4002 } | |
4003 | |
4004 theend: | |
1520 | 4005 /* Free "reg_tofree" when it's a bit big. |
4006 * Free regstack and backpos if they are bigger than their initial size. */ | |
4007 if (reg_tofreelen > 400) | |
4008 { | |
4009 vim_free(reg_tofree); | |
4010 reg_tofree = NULL; | |
4011 } | |
4012 if (regstack.ga_maxlen > REGSTACK_INITIAL) | |
4013 ga_clear(®stack); | |
4014 if (backpos.ga_maxlen > BACKPOS_INITIAL) | |
4015 ga_clear(&backpos); | |
270 | 4016 |
7 | 4017 return retval; |
4018 } | |
4019 | |
4020 #ifdef FEAT_SYN_HL | |
4021 static reg_extmatch_T *make_extmatch __ARGS((void)); | |
4022 | |
4023 /* | |
4024 * Create a new extmatch and mark it as referenced once. | |
4025 */ | |
4026 static reg_extmatch_T * | |
4027 make_extmatch() | |
4028 { | |
4029 reg_extmatch_T *em; | |
4030 | |
4031 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T)); | |
4032 if (em != NULL) | |
4033 em->refcnt = 1; | |
4034 return em; | |
4035 } | |
4036 | |
4037 /* | |
4038 * Add a reference to an extmatch. | |
4039 */ | |
4040 reg_extmatch_T * | |
4041 ref_extmatch(em) | |
4042 reg_extmatch_T *em; | |
4043 { | |
4044 if (em != NULL) | |
4045 em->refcnt++; | |
4046 return em; | |
4047 } | |
4048 | |
4049 /* | |
4050 * Remove a reference to an extmatch. If there are no references left, free | |
4051 * the info. | |
4052 */ | |
4053 void | |
4054 unref_extmatch(em) | |
4055 reg_extmatch_T *em; | |
4056 { | |
4057 int i; | |
4058 | |
4059 if (em != NULL && --em->refcnt <= 0) | |
4060 { | |
4061 for (i = 0; i < NSUBEXP; ++i) | |
4062 vim_free(em->matches[i]); | |
4063 vim_free(em); | |
4064 } | |
4065 } | |
4066 #endif | |
4067 | |
4068 /* | |
4069 * regtry - try match of "prog" with at regline["col"]. | |
4070 * Returns 0 for failure, number of lines contained in the match otherwise. | |
4071 */ | |
4072 static long | |
4073 regtry(prog, col) | |
4444 | 4074 bt_regprog_T *prog; |
7 | 4075 colnr_T col; |
4076 { | |
4077 reginput = regline + col; | |
4078 need_clear_subexpr = TRUE; | |
4079 #ifdef FEAT_SYN_HL | |
4080 /* Clear the external match subpointers if necessary. */ | |
4081 if (prog->reghasz == REX_SET) | |
4082 need_clear_zsubexpr = TRUE; | |
4083 #endif | |
4084 | |
189 | 4085 if (regmatch(prog->program + 1) == 0) |
4086 return 0; | |
4087 | |
4088 cleanup_subexpr(); | |
4089 if (REG_MULTI) | |
7 | 4090 { |
189 | 4091 if (reg_startpos[0].lnum < 0) |
7 | 4092 { |
189 | 4093 reg_startpos[0].lnum = 0; |
4094 reg_startpos[0].col = col; | |
4095 } | |
4096 if (reg_endpos[0].lnum < 0) | |
4097 { | |
4098 reg_endpos[0].lnum = reglnum; | |
4099 reg_endpos[0].col = (int)(reginput - regline); | |
7 | 4100 } |
4101 else | |
189 | 4102 /* Use line number of "\ze". */ |
4103 reglnum = reg_endpos[0].lnum; | |
4104 } | |
4105 else | |
4106 { | |
4107 if (reg_startp[0] == NULL) | |
4108 reg_startp[0] = regline + col; | |
4109 if (reg_endp[0] == NULL) | |
4110 reg_endp[0] = reginput; | |
4111 } | |
7 | 4112 #ifdef FEAT_SYN_HL |
189 | 4113 /* Package any found \z(...\) matches for export. Default is none. */ |
4114 unref_extmatch(re_extmatch_out); | |
4115 re_extmatch_out = NULL; | |
4116 | |
4117 if (prog->reghasz == REX_SET) | |
4118 { | |
4119 int i; | |
4120 | |
4121 cleanup_zsubexpr(); | |
4122 re_extmatch_out = make_extmatch(); | |
4123 for (i = 0; i < NSUBEXP; i++) | |
7 | 4124 { |
189 | 4125 if (REG_MULTI) |
7 | 4126 { |
189 | 4127 /* Only accept single line matches. */ |
4128 if (reg_startzpos[i].lnum >= 0 | |
4129 && reg_endzpos[i].lnum == reg_startzpos[i].lnum) | |
4130 re_extmatch_out->matches[i] = | |
4131 vim_strnsave(reg_getline(reg_startzpos[i].lnum) | |
7 | 4132 + reg_startzpos[i].col, |
189 | 4133 reg_endzpos[i].col - reg_startzpos[i].col); |
4134 } | |
4135 else | |
4136 { | |
4137 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL) | |
4138 re_extmatch_out->matches[i] = | |
7 | 4139 vim_strnsave(reg_startzp[i], |
189 | 4140 (int)(reg_endzp[i] - reg_startzp[i])); |
7 | 4141 } |
4142 } | |
189 | 4143 } |
7 | 4144 #endif |
189 | 4145 return 1 + reglnum; |
7 | 4146 } |
4147 | |
4148 #ifdef FEAT_MBYTE | |
4149 static int reg_prev_class __ARGS((void)); | |
4150 | |
4151 /* | |
4152 * Get class of previous character. | |
4153 */ | |
4154 static int | |
4155 reg_prev_class() | |
4156 { | |
4157 if (reginput > regline) | |
4069 | 4158 return mb_get_class_buf(reginput - 1 |
4159 - (*mb_head_off)(regline, reginput - 1), reg_buf); | |
7 | 4160 return -1; |
4161 } | |
4162 | |
4163 #endif | |
39 | 4164 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput) |
7 | 4165 |
4166 /* | |
4167 * The arguments from BRACE_LIMITS are stored here. They are actually local | |
4168 * to regmatch(), but they are here to reduce the amount of stack space used | |
4169 * (it can be called recursively many times). | |
4170 */ | |
4171 static long bl_minval; | |
4172 static long bl_maxval; | |
4173 | |
4174 /* | |
4175 * regmatch - main matching routine | |
4176 * | |
180 | 4177 * Conceptually the strategy is simple: Check to see whether the current node |
4178 * matches, push an item onto the regstack and loop to see whether the rest | |
4179 * matches, and then act accordingly. In practice we make some effort to | |
4180 * avoid using the regstack, in particular by going through "ordinary" nodes | |
4181 * (that don't need to know whether the rest of the match failed) by a nested | |
4182 * loop. | |
7 | 4183 * |
4184 * Returns TRUE when there is a match. Leaves reginput and reglnum just after | |
4185 * the last matched character. | |
4186 * Returns FALSE when there is no match. Leaves reginput and reglnum in an | |
4187 * undefined state! | |
4188 */ | |
4189 static int | |
180 | 4190 regmatch(scan) |
7 | 4191 char_u *scan; /* Current node. */ |
4192 { | |
180 | 4193 char_u *next; /* Next node. */ |
4194 int op; | |
4195 int c; | |
4196 regitem_T *rp; | |
4197 int no; | |
4198 int status; /* one of the RA_ values: */ | |
4199 #define RA_FAIL 1 /* something failed, abort */ | |
4200 #define RA_CONT 2 /* continue in inner loop */ | |
4201 #define RA_BREAK 3 /* break inner loop */ | |
4202 #define RA_MATCH 4 /* successful match */ | |
4203 #define RA_NOMATCH 5 /* didn't match */ | |
270 | 4204 |
1520 | 4205 /* Make "regstack" and "backpos" empty. They are allocated and freed in |
4444 | 4206 * bt_regexec_both() to reduce malloc()/free() calls. */ |
270 | 4207 regstack.ga_len = 0; |
4208 backpos.ga_len = 0; | |
233 | 4209 |
180 | 4210 /* |
233 | 4211 * Repeat until "regstack" is empty. |
180 | 4212 */ |
4213 for (;;) | |
4214 { | |
4444 | 4215 /* Some patterns may cause a long time to match, even though they are not |
7 | 4216 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */ |
4217 fast_breakcheck(); | |
4218 | |
4219 #ifdef DEBUG | |
4220 if (scan != NULL && regnarrate) | |
4221 { | |
4444 | 4222 mch_errmsg((char *)regprop(scan)); |
7 | 4223 mch_errmsg("(\n"); |
4224 } | |
4225 #endif | |
180 | 4226 |
4227 /* | |
233 | 4228 * Repeat for items that can be matched sequentially, without using the |
180 | 4229 * regstack. |
4230 */ | |
4231 for (;;) | |
7 | 4232 { |
180 | 4233 if (got_int || scan == NULL) |
4234 { | |
4235 status = RA_FAIL; | |
4236 break; | |
4237 } | |
4238 status = RA_CONT; | |
4239 | |
7 | 4240 #ifdef DEBUG |
4241 if (regnarrate) | |
4242 { | |
4444 | 4243 mch_errmsg((char *)regprop(scan)); |
7 | 4244 mch_errmsg("...\n"); |
4245 # ifdef FEAT_SYN_HL | |
4246 if (re_extmatch_in != NULL) | |
4247 { | |
4248 int i; | |
4249 | |
4250 mch_errmsg(_("External submatches:\n")); | |
4251 for (i = 0; i < NSUBEXP; i++) | |
4252 { | |
4253 mch_errmsg(" \""); | |
4254 if (re_extmatch_in->matches[i] != NULL) | |
4444 | 4255 mch_errmsg((char *)re_extmatch_in->matches[i]); |
7 | 4256 mch_errmsg("\"\n"); |
4257 } | |
4258 } | |
4259 # endif | |
4260 } | |
4261 #endif | |
4262 next = regnext(scan); | |
4263 | |
4264 op = OP(scan); | |
4265 /* Check for character class with NL added. */ | |
1018 | 4266 if (!reg_line_lbr && WITH_NL(op) && REG_MULTI |
4267 && *reginput == NUL && reglnum <= reg_maxline) | |
7 | 4268 { |
4269 reg_nextline(); | |
4270 } | |
4271 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n') | |
4272 { | |
4273 ADVANCE_REGINPUT(); | |
4274 } | |
4275 else | |
4276 { | |
4277 if (WITH_NL(op)) | |
180 | 4278 op -= ADD_NL; |
7 | 4279 #ifdef FEAT_MBYTE |
4280 if (has_mbyte) | |
4281 c = (*mb_ptr2char)(reginput); | |
4282 else | |
4283 #endif | |
4284 c = *reginput; | |
4285 switch (op) | |
4286 { | |
4287 case BOL: | |
4288 if (reginput != regline) | |
180 | 4289 status = RA_NOMATCH; |
7 | 4290 break; |
4291 | |
4292 case EOL: | |
4293 if (c != NUL) | |
180 | 4294 status = RA_NOMATCH; |
7 | 4295 break; |
4296 | |
4297 case RE_BOF: | |
1458 | 4298 /* We're not at the beginning of the file when below the first |
4299 * line where we started, not at the start of the line or we | |
4300 * didn't start at the first line of the buffer. */ | |
7 | 4301 if (reglnum != 0 || reginput != regline |
1458 | 4302 || (REG_MULTI && reg_firstlnum > 1)) |
180 | 4303 status = RA_NOMATCH; |
7 | 4304 break; |
4305 | |
4306 case RE_EOF: | |
4307 if (reglnum != reg_maxline || c != NUL) | |
180 | 4308 status = RA_NOMATCH; |
7 | 4309 break; |
4310 | |
4311 case CURSOR: | |
4312 /* Check if the buffer is in a window and compare the | |
4313 * reg_win->w_cursor position to the match position. */ | |
4314 if (reg_win == NULL | |
4315 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum) | |
4316 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col)) | |
180 | 4317 status = RA_NOMATCH; |
7 | 4318 break; |
4319 | |
639 | 4320 case RE_MARK: |
4321 /* Compare the mark position to the match position. NOTE: Always | |
4322 * uses the current buffer. */ | |
4323 { | |
4324 int mark = OPERAND(scan)[0]; | |
4325 int cmp = OPERAND(scan)[1]; | |
4326 pos_T *pos; | |
4327 | |
4043 | 4328 pos = getmark_buf(reg_buf, mark, FALSE); |
1148 | 4329 if (pos == NULL /* mark doesn't exist */ |
639 | 4330 || pos->lnum <= 0 /* mark isn't set (in curbuf) */ |
4331 || (pos->lnum == reglnum + reg_firstlnum | |
4332 ? (pos->col == (colnr_T)(reginput - regline) | |
4333 ? (cmp == '<' || cmp == '>') | |
4334 : (pos->col < (colnr_T)(reginput - regline) | |
4335 ? cmp != '>' | |
4336 : cmp != '<')) | |
4337 : (pos->lnum < reglnum + reg_firstlnum | |
4338 ? cmp != '>' | |
4339 : cmp != '<'))) | |
4340 status = RA_NOMATCH; | |
4341 } | |
4342 break; | |
4343 | |
4344 case RE_VISUAL: | |
4345 #ifdef FEAT_VISUAL | |
4346 /* Check if the buffer is the current buffer. and whether the | |
4347 * position is inside the Visual area. */ | |
4348 if (reg_buf != curbuf || VIsual.lnum == 0) | |
4349 status = RA_NOMATCH; | |
4350 else | |
4351 { | |
4352 pos_T top, bot; | |
4353 linenr_T lnum; | |
4354 colnr_T col; | |
4355 win_T *wp = reg_win == NULL ? curwin : reg_win; | |
4356 int mode; | |
4357 | |
4358 if (VIsual_active) | |
4359 { | |
4360 if (lt(VIsual, wp->w_cursor)) | |
4361 { | |
4362 top = VIsual; | |
4363 bot = wp->w_cursor; | |
4364 } | |
4365 else | |
4366 { | |
4367 top = wp->w_cursor; | |
4368 bot = VIsual; | |
4369 } | |
4370 mode = VIsual_mode; | |
4371 } | |
4372 else | |
4373 { | |
692 | 4374 if (lt(curbuf->b_visual.vi_start, curbuf->b_visual.vi_end)) |
685 | 4375 { |
692 | 4376 top = curbuf->b_visual.vi_start; |
4377 bot = curbuf->b_visual.vi_end; | |
685 | 4378 } |
4379 else | |
4380 { | |
692 | 4381 top = curbuf->b_visual.vi_end; |
4382 bot = curbuf->b_visual.vi_start; | |
685 | 4383 } |
692 | 4384 mode = curbuf->b_visual.vi_mode; |
639 | 4385 } |
4386 lnum = reglnum + reg_firstlnum; | |
4387 col = (colnr_T)(reginput - regline); | |
4388 if (lnum < top.lnum || lnum > bot.lnum) | |
4389 status = RA_NOMATCH; | |
4390 else if (mode == 'v') | |
4391 { | |
4392 if ((lnum == top.lnum && col < top.col) | |
4393 || (lnum == bot.lnum | |
4394 && col >= bot.col + (*p_sel != 'e'))) | |
4395 status = RA_NOMATCH; | |
4396 } | |
4397 else if (mode == Ctrl_V) | |
4398 { | |
4399 colnr_T start, end; | |
4400 colnr_T start2, end2; | |
944 | 4401 colnr_T cols; |
639 | 4402 |
4403 getvvcol(wp, &top, &start, NULL, &end); | |
4404 getvvcol(wp, &bot, &start2, NULL, &end2); | |
4405 if (start2 < start) | |
4406 start = start2; | |
4407 if (end2 > end) | |
4408 end = end2; | |
4409 if (top.col == MAXCOL || bot.col == MAXCOL) | |
4410 end = MAXCOL; | |
944 | 4411 cols = win_linetabsize(wp, |
639 | 4412 regline, (colnr_T)(reginput - regline)); |
944 | 4413 if (cols < start || cols > end - (*p_sel == 'e')) |
639 | 4414 status = RA_NOMATCH; |
4415 } | |
4416 } | |
4417 #else | |
4418 status = RA_NOMATCH; | |
4419 #endif | |
4420 break; | |
4421 | |
7 | 4422 case RE_LNUM: |
4423 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum), | |
4424 scan)) | |
180 | 4425 status = RA_NOMATCH; |
7 | 4426 break; |
4427 | |
4428 case RE_COL: | |
4429 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan)) | |
180 | 4430 status = RA_NOMATCH; |
7 | 4431 break; |
4432 | |
4433 case RE_VCOL: | |
4434 if (!re_num_cmp((long_u)win_linetabsize( | |
4435 reg_win == NULL ? curwin : reg_win, | |
4436 regline, (colnr_T)(reginput - regline)) + 1, scan)) | |
180 | 4437 status = RA_NOMATCH; |
7 | 4438 break; |
4439 | |
4440 case BOW: /* \<word; reginput points to w */ | |
4441 if (c == NUL) /* Can't match at end of line */ | |
180 | 4442 status = RA_NOMATCH; |
7 | 4443 #ifdef FEAT_MBYTE |
180 | 4444 else if (has_mbyte) |
7 | 4445 { |
4446 int this_class; | |
4447 | |
4448 /* Get class of current and previous char (if it exists). */ | |
4069 | 4449 this_class = mb_get_class_buf(reginput, reg_buf); |
7 | 4450 if (this_class <= 1) |
180 | 4451 status = RA_NOMATCH; /* not on a word at all */ |
4452 else if (reg_prev_class() == this_class) | |
4453 status = RA_NOMATCH; /* previous char is in same word */ | |
7 | 4454 } |
4455 #endif | |
4456 else | |
4457 { | |
4061 | 4458 if (!vim_iswordc_buf(c, reg_buf) || (reginput > regline |
4459 && vim_iswordc_buf(reginput[-1], reg_buf))) | |
180 | 4460 status = RA_NOMATCH; |
7 | 4461 } |
4462 break; | |
4463 | |
4464 case EOW: /* word\>; reginput points after d */ | |
4465 if (reginput == regline) /* Can't match at start of line */ | |
180 | 4466 status = RA_NOMATCH; |
7 | 4467 #ifdef FEAT_MBYTE |
180 | 4468 else if (has_mbyte) |
7 | 4469 { |
4470 int this_class, prev_class; | |
4471 | |
4472 /* Get class of current and previous char (if it exists). */ | |
4069 | 4473 this_class = mb_get_class_buf(reginput, reg_buf); |
7 | 4474 prev_class = reg_prev_class(); |
180 | 4475 if (this_class == prev_class |
4476 || prev_class == 0 || prev_class == 1) | |
4477 status = RA_NOMATCH; | |
7 | 4478 } |
180 | 4479 #endif |
7 | 4480 else |
4481 { | |
4043 | 4482 if (!vim_iswordc_buf(reginput[-1], reg_buf) |
4483 || (reginput[0] != NUL && vim_iswordc_buf(c, reg_buf))) | |
180 | 4484 status = RA_NOMATCH; |
7 | 4485 } |
4486 break; /* Matched with EOW */ | |
4487 | |
4488 case ANY: | |
4084 | 4489 /* ANY does not match new lines. */ |
7 | 4490 if (c == NUL) |
180 | 4491 status = RA_NOMATCH; |
4492 else | |
4493 ADVANCE_REGINPUT(); | |
7 | 4494 break; |
4495 | |
4496 case IDENT: | |
4497 if (!vim_isIDc(c)) | |
180 | 4498 status = RA_NOMATCH; |
4499 else | |
4500 ADVANCE_REGINPUT(); | |
7 | 4501 break; |
4502 | |
4503 case SIDENT: | |
4504 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c)) | |
180 | 4505 status = RA_NOMATCH; |
4506 else | |
4507 ADVANCE_REGINPUT(); | |
7 | 4508 break; |
4509 | |
4510 case KWORD: | |
4069 | 4511 if (!vim_iswordp_buf(reginput, reg_buf)) |
180 | 4512 status = RA_NOMATCH; |
4513 else | |
4514 ADVANCE_REGINPUT(); | |
7 | 4515 break; |
4516 | |
4517 case SKWORD: | |
4069 | 4518 if (VIM_ISDIGIT(*reginput) || !vim_iswordp_buf(reginput, reg_buf)) |
180 | 4519 status = RA_NOMATCH; |
4520 else | |
4521 ADVANCE_REGINPUT(); | |
7 | 4522 break; |
4523 | |
4524 case FNAME: | |
4525 if (!vim_isfilec(c)) | |
180 | 4526 status = RA_NOMATCH; |
4527 else | |
4528 ADVANCE_REGINPUT(); | |
7 | 4529 break; |
4530 | |
4531 case SFNAME: | |
4532 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c)) | |
180 | 4533 status = RA_NOMATCH; |
4534 else | |
4535 ADVANCE_REGINPUT(); | |
7 | 4536 break; |
4537 | |
4538 case PRINT: | |
4539 if (ptr2cells(reginput) != 1) | |
180 | 4540 status = RA_NOMATCH; |
4541 else | |
4542 ADVANCE_REGINPUT(); | |
7 | 4543 break; |
4544 | |
4545 case SPRINT: | |
4546 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1) | |
180 | 4547 status = RA_NOMATCH; |
4548 else | |
4549 ADVANCE_REGINPUT(); | |
7 | 4550 break; |
4551 | |
4552 case WHITE: | |
4553 if (!vim_iswhite(c)) | |
180 | 4554 status = RA_NOMATCH; |
4555 else | |
4556 ADVANCE_REGINPUT(); | |
7 | 4557 break; |
4558 | |
4559 case NWHITE: | |
4560 if (c == NUL || vim_iswhite(c)) | |
180 | 4561 status = RA_NOMATCH; |
4562 else | |
4563 ADVANCE_REGINPUT(); | |
7 | 4564 break; |
4565 | |
4566 case DIGIT: | |
4567 if (!ri_digit(c)) | |
180 | 4568 status = RA_NOMATCH; |
4569 else | |
4570 ADVANCE_REGINPUT(); | |
7 | 4571 break; |
4572 | |
4573 case NDIGIT: | |
4574 if (c == NUL || ri_digit(c)) | |
180 | 4575 status = RA_NOMATCH; |
4576 else | |
4577 ADVANCE_REGINPUT(); | |
7 | 4578 break; |
4579 | |
4580 case HEX: | |
4581 if (!ri_hex(c)) | |
180 | 4582 status = RA_NOMATCH; |
4583 else | |
4584 ADVANCE_REGINPUT(); | |
7 | 4585 break; |
4586 | |
4587 case NHEX: | |
4588 if (c == NUL || ri_hex(c)) | |
180 | 4589 status = RA_NOMATCH; |
4590 else | |
4591 ADVANCE_REGINPUT(); | |
7 | 4592 break; |
4593 | |
4594 case OCTAL: | |
4595 if (!ri_octal(c)) | |
180 | 4596 status = RA_NOMATCH; |
4597 else | |
4598 ADVANCE_REGINPUT(); | |
7 | 4599 break; |
4600 | |
4601 case NOCTAL: | |
4602 if (c == NUL || ri_octal(c)) | |
180 | 4603 status = RA_NOMATCH; |
4604 else | |
4605 ADVANCE_REGINPUT(); | |
7 | 4606 break; |
4607 | |
4608 case WORD: | |
4609 if (!ri_word(c)) | |
180 | 4610 status = RA_NOMATCH; |
4611 else | |
4612 ADVANCE_REGINPUT(); | |
7 | 4613 break; |
4614 | |
4615 case NWORD: | |
4616 if (c == NUL || ri_word(c)) | |
180 | 4617 status = RA_NOMATCH; |
4618 else | |
4619 ADVANCE_REGINPUT(); | |
7 | 4620 break; |
4621 | |
4622 case HEAD: | |
4623 if (!ri_head(c)) | |
180 | 4624 status = RA_NOMATCH; |
4625 else | |
4626 ADVANCE_REGINPUT(); | |
7 | 4627 break; |
4628 | |
4629 case NHEAD: | |
4630 if (c == NUL || ri_head(c)) | |
180 | 4631 status = RA_NOMATCH; |
4632 else | |
4633 ADVANCE_REGINPUT(); | |
7 | 4634 break; |
4635 | |
4636 case ALPHA: | |
4637 if (!ri_alpha(c)) | |
180 | 4638 status = RA_NOMATCH; |
4639 else | |
4640 ADVANCE_REGINPUT(); | |
7 | 4641 break; |
4642 | |
4643 case NALPHA: | |
4644 if (c == NUL || ri_alpha(c)) | |
180 | 4645 status = RA_NOMATCH; |
4646 else | |
4647 ADVANCE_REGINPUT(); | |
7 | 4648 break; |
4649 | |
4650 case LOWER: | |
4651 if (!ri_lower(c)) | |
180 | 4652 status = RA_NOMATCH; |
4653 else | |
4654 ADVANCE_REGINPUT(); | |
7 | 4655 break; |
4656 | |
4657 case NLOWER: | |
4658 if (c == NUL || ri_lower(c)) | |
180 | 4659 status = RA_NOMATCH; |
4660 else | |
4661 ADVANCE_REGINPUT(); | |
7 | 4662 break; |
4663 | |
4664 case UPPER: | |
4665 if (!ri_upper(c)) | |
180 | 4666 status = RA_NOMATCH; |
4667 else | |
4668 ADVANCE_REGINPUT(); | |
7 | 4669 break; |
4670 | |
4671 case NUPPER: | |
4672 if (c == NUL || ri_upper(c)) | |
180 | 4673 status = RA_NOMATCH; |
4674 else | |
4675 ADVANCE_REGINPUT(); | |
7 | 4676 break; |
4677 | |
4678 case EXACTLY: | |
4679 { | |
4680 int len; | |
4681 char_u *opnd; | |
4682 | |
4683 opnd = OPERAND(scan); | |
4684 /* Inline the first byte, for speed. */ | |
4685 if (*opnd != *reginput | |
4686 && (!ireg_ic || ( | |
4687 #ifdef FEAT_MBYTE | |
4688 !enc_utf8 && | |
4689 #endif | |
1347 | 4690 MB_TOLOWER(*opnd) != MB_TOLOWER(*reginput)))) |
180 | 4691 status = RA_NOMATCH; |
4692 else if (*opnd == NUL) | |
7 | 4693 { |
4694 /* match empty string always works; happens when "~" is | |
4695 * empty. */ | |
4696 } | |
4697 else if (opnd[1] == NUL | |
4698 #ifdef FEAT_MBYTE | |
4699 && !(enc_utf8 && ireg_ic) | |
4700 #endif | |
4701 ) | |
4702 ++reginput; /* matched a single char */ | |
4703 else | |
4704 { | |
4705 len = (int)STRLEN(opnd); | |
4706 /* Need to match first byte again for multi-byte. */ | |
4707 if (cstrncmp(opnd, reginput, &len) != 0) | |
180 | 4708 status = RA_NOMATCH; |
7 | 4709 #ifdef FEAT_MBYTE |
4710 /* Check for following composing character. */ | |
180 | 4711 else if (enc_utf8 |
4712 && UTF_COMPOSINGLIKE(reginput, reginput + len)) | |
7 | 4713 { |
4714 /* raaron: This code makes a composing character get | |
4715 * ignored, which is the correct behavior (sometimes) | |
4716 * for voweled Hebrew texts. */ | |
4717 if (!ireg_icombine) | |
180 | 4718 status = RA_NOMATCH; |
7 | 4719 } |
180 | 4720 #endif |
7 | 4721 else |
180 | 4722 reginput += len; |
7 | 4723 } |
4724 } | |
4725 break; | |
4726 | |
4727 case ANYOF: | |
4728 case ANYBUT: | |
4729 if (c == NUL) | |
180 | 4730 status = RA_NOMATCH; |
4731 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF)) | |
4732 status = RA_NOMATCH; | |
4733 else | |
4734 ADVANCE_REGINPUT(); | |
7 | 4735 break; |
4736 | |
4737 #ifdef FEAT_MBYTE | |
4738 case MULTIBYTECODE: | |
4739 if (has_mbyte) | |
4740 { | |
4741 int i, len; | |
4742 char_u *opnd; | |
944 | 4743 int opndc = 0, inpc; |
7 | 4744 |
4745 opnd = OPERAND(scan); | |
4746 /* Safety check (just in case 'encoding' was changed since | |
4747 * compiling the program). */ | |
474 | 4748 if ((len = (*mb_ptr2len)(opnd)) < 2) |
180 | 4749 { |
4750 status = RA_NOMATCH; | |
4751 break; | |
4752 } | |
714 | 4753 if (enc_utf8) |
4754 opndc = mb_ptr2char(opnd); | |
4755 if (enc_utf8 && utf_iscomposing(opndc)) | |
4756 { | |
4757 /* When only a composing char is given match at any | |
4758 * position where that composing char appears. */ | |
4759 status = RA_NOMATCH; | |
4760 for (i = 0; reginput[i] != NUL; i += utf_char2len(inpc)) | |
180 | 4761 { |
714 | 4762 inpc = mb_ptr2char(reginput + i); |
4763 if (!utf_iscomposing(inpc)) | |
4764 { | |
4765 if (i > 0) | |
4766 break; | |
4767 } | |
4768 else if (opndc == inpc) | |
4769 { | |
4770 /* Include all following composing chars. */ | |
4771 len = i + mb_ptr2len(reginput + i); | |
4772 status = RA_MATCH; | |
4773 break; | |
4774 } | |
180 | 4775 } |
714 | 4776 } |
4777 else | |
4778 for (i = 0; i < len; ++i) | |
4779 if (opnd[i] != reginput[i]) | |
4780 { | |
4781 status = RA_NOMATCH; | |
4782 break; | |
4783 } | |
7 | 4784 reginput += len; |
4785 } | |
4786 else | |
180 | 4787 status = RA_NOMATCH; |
7 | 4788 break; |
4789 #endif | |
4790 | |
4791 case NOTHING: | |
4792 break; | |
4793 | |
4794 case BACK: | |
233 | 4795 { |
4796 int i; | |
4797 backpos_T *bp; | |
4798 | |
4799 /* | |
4800 * When we run into BACK we need to check if we don't keep | |
4801 * looping without matching any input. The second and later | |
4802 * times a BACK is encountered it fails if the input is still | |
4803 * at the same position as the previous time. | |
4804 * The positions are stored in "backpos" and found by the | |
4805 * current value of "scan", the position in the RE program. | |
4806 */ | |
4807 bp = (backpos_T *)backpos.ga_data; | |
4808 for (i = 0; i < backpos.ga_len; ++i) | |
4809 if (bp[i].bp_scan == scan) | |
4810 break; | |
4811 if (i == backpos.ga_len) | |
4812 { | |
4813 /* First time at this BACK, make room to store the pos. */ | |
4814 if (ga_grow(&backpos, 1) == FAIL) | |
4815 status = RA_FAIL; | |
4816 else | |
4817 { | |
4818 /* get "ga_data" again, it may have changed */ | |
4819 bp = (backpos_T *)backpos.ga_data; | |
4820 bp[i].bp_scan = scan; | |
4821 ++backpos.ga_len; | |
4822 } | |
4823 } | |
4824 else if (reg_save_equal(&bp[i].bp_pos)) | |
4825 /* Still at same position as last time, fail. */ | |
4826 status = RA_NOMATCH; | |
4827 | |
4828 if (status != RA_FAIL && status != RA_NOMATCH) | |
4829 reg_save(&bp[i].bp_pos, &backpos); | |
4830 } | |
179 | 4831 break; |
4832 | |
7 | 4833 case MOPEN + 0: /* Match start: \zs */ |
4834 case MOPEN + 1: /* \( */ | |
4835 case MOPEN + 2: | |
4836 case MOPEN + 3: | |
4837 case MOPEN + 4: | |
4838 case MOPEN + 5: | |
4839 case MOPEN + 6: | |
4840 case MOPEN + 7: | |
4841 case MOPEN + 8: | |
4842 case MOPEN + 9: | |
4843 { | |
4844 no = op - MOPEN; | |
4845 cleanup_subexpr(); | |
270 | 4846 rp = regstack_push(RS_MOPEN, scan); |
180 | 4847 if (rp == NULL) |
4848 status = RA_FAIL; | |
4849 else | |
4850 { | |
4851 rp->rs_no = no; | |
4852 save_se(&rp->rs_un.sesave, ®_startpos[no], | |
4853 ®_startp[no]); | |
4854 /* We simply continue and handle the result when done. */ | |
4855 } | |
7 | 4856 } |
180 | 4857 break; |
7 | 4858 |
4859 case NOPEN: /* \%( */ | |
4860 case NCLOSE: /* \) after \%( */ | |
270 | 4861 if (regstack_push(RS_NOPEN, scan) == NULL) |
180 | 4862 status = RA_FAIL; |
4863 /* We simply continue and handle the result when done. */ | |
4864 break; | |
7 | 4865 |
4866 #ifdef FEAT_SYN_HL | |
4867 case ZOPEN + 1: | |
4868 case ZOPEN + 2: | |
4869 case ZOPEN + 3: | |
4870 case ZOPEN + 4: | |
4871 case ZOPEN + 5: | |
4872 case ZOPEN + 6: | |
4873 case ZOPEN + 7: | |
4874 case ZOPEN + 8: | |
4875 case ZOPEN + 9: | |
4876 { | |
4877 no = op - ZOPEN; | |
4878 cleanup_zsubexpr(); | |
270 | 4879 rp = regstack_push(RS_ZOPEN, scan); |
180 | 4880 if (rp == NULL) |
4881 status = RA_FAIL; | |
4882 else | |
4883 { | |
4884 rp->rs_no = no; | |
4885 save_se(&rp->rs_un.sesave, ®_startzpos[no], | |
4886 ®_startzp[no]); | |
4887 /* We simply continue and handle the result when done. */ | |
4888 } | |
7 | 4889 } |
180 | 4890 break; |
7 | 4891 #endif |
4892 | |
4893 case MCLOSE + 0: /* Match end: \ze */ | |
4894 case MCLOSE + 1: /* \) */ | |
4895 case MCLOSE + 2: | |
4896 case MCLOSE + 3: | |
4897 case MCLOSE + 4: | |
4898 case MCLOSE + 5: | |
4899 case MCLOSE + 6: | |
4900 case MCLOSE + 7: | |
4901 case MCLOSE + 8: | |
4902 case MCLOSE + 9: | |
4903 { | |
4904 no = op - MCLOSE; | |
4905 cleanup_subexpr(); | |
270 | 4906 rp = regstack_push(RS_MCLOSE, scan); |
180 | 4907 if (rp == NULL) |
4908 status = RA_FAIL; | |
4909 else | |
4910 { | |
4911 rp->rs_no = no; | |
4912 save_se(&rp->rs_un.sesave, ®_endpos[no], ®_endp[no]); | |
4913 /* We simply continue and handle the result when done. */ | |
4914 } | |
7 | 4915 } |
180 | 4916 break; |
7 | 4917 |
4918 #ifdef FEAT_SYN_HL | |
4919 case ZCLOSE + 1: /* \) after \z( */ | |
4920 case ZCLOSE + 2: | |
4921 case ZCLOSE + 3: | |
4922 case ZCLOSE + 4: | |
4923 case ZCLOSE + 5: | |
4924 case ZCLOSE + 6: | |
4925 case ZCLOSE + 7: | |
4926 case ZCLOSE + 8: | |
4927 case ZCLOSE + 9: | |
4928 { | |
4929 no = op - ZCLOSE; | |
4930 cleanup_zsubexpr(); | |
270 | 4931 rp = regstack_push(RS_ZCLOSE, scan); |
180 | 4932 if (rp == NULL) |
4933 status = RA_FAIL; | |
4934 else | |
4935 { | |
4936 rp->rs_no = no; | |
4937 save_se(&rp->rs_un.sesave, ®_endzpos[no], | |
4938 ®_endzp[no]); | |
4939 /* We simply continue and handle the result when done. */ | |
4940 } | |
7 | 4941 } |
180 | 4942 break; |
7 | 4943 #endif |
4944 | |
4945 case BACKREF + 1: | |
4946 case BACKREF + 2: | |
4947 case BACKREF + 3: | |
4948 case BACKREF + 4: | |
4949 case BACKREF + 5: | |
4950 case BACKREF + 6: | |
4951 case BACKREF + 7: | |
4952 case BACKREF + 8: | |
4953 case BACKREF + 9: | |
4954 { | |
4955 int len; | |
4956 linenr_T clnum; | |
4957 colnr_T ccol; | |
4958 char_u *p; | |
4959 | |
4960 no = op - BACKREF; | |
4961 cleanup_subexpr(); | |
4962 if (!REG_MULTI) /* Single-line regexp */ | |
4963 { | |
1815 | 4964 if (reg_startp[no] == NULL || reg_endp[no] == NULL) |
7 | 4965 { |
4966 /* Backref was not set: Match an empty string. */ | |
4967 len = 0; | |
4968 } | |
4969 else | |
4970 { | |
4971 /* Compare current input with back-ref in the same | |
4972 * line. */ | |
4973 len = (int)(reg_endp[no] - reg_startp[no]); | |
4974 if (cstrncmp(reg_startp[no], reginput, &len) != 0) | |
180 | 4975 status = RA_NOMATCH; |
7 | 4976 } |
4977 } | |
4978 else /* Multi-line regexp */ | |
4979 { | |
1815 | 4980 if (reg_startpos[no].lnum < 0 || reg_endpos[no].lnum < 0) |
7 | 4981 { |
4982 /* Backref was not set: Match an empty string. */ | |
4983 len = 0; | |
4984 } | |
4985 else | |
4986 { | |
4987 if (reg_startpos[no].lnum == reglnum | |
4988 && reg_endpos[no].lnum == reglnum) | |
4989 { | |
4990 /* Compare back-ref within the current line. */ | |
4991 len = reg_endpos[no].col - reg_startpos[no].col; | |
4992 if (cstrncmp(regline + reg_startpos[no].col, | |
4993 reginput, &len) != 0) | |
180 | 4994 status = RA_NOMATCH; |
7 | 4995 } |
4996 else | |
4997 { | |
4998 /* Messy situation: Need to compare between two | |
4999 * lines. */ | |
5000 ccol = reg_startpos[no].col; | |
5001 clnum = reg_startpos[no].lnum; | |
5002 for (;;) | |
5003 { | |
5004 /* Since getting one line may invalidate | |
5005 * the other, need to make copy. Slow! */ | |
5006 if (regline != reg_tofree) | |
5007 { | |
5008 len = (int)STRLEN(regline); | |
5009 if (reg_tofree == NULL | |
5010 || len >= (int)reg_tofreelen) | |
5011 { | |
5012 len += 50; /* get some extra */ | |
5013 vim_free(reg_tofree); | |
5014 reg_tofree = alloc(len); | |
5015 if (reg_tofree == NULL) | |
180 | 5016 { |
5017 status = RA_FAIL; /* outof memory!*/ | |
5018 break; | |
5019 } | |
7 | 5020 reg_tofreelen = len; |
5021 } | |
5022 STRCPY(reg_tofree, regline); | |
5023 reginput = reg_tofree | |
5024 + (reginput - regline); | |
5025 regline = reg_tofree; | |
5026 } | |
5027 | |
5028 /* Get the line to compare with. */ | |
5029 p = reg_getline(clnum); | |
5030 if (clnum == reg_endpos[no].lnum) | |
5031 len = reg_endpos[no].col - ccol; | |
5032 else | |
5033 len = (int)STRLEN(p + ccol); | |
5034 | |
5035 if (cstrncmp(p + ccol, reginput, &len) != 0) | |
180 | 5036 { |
5037 status = RA_NOMATCH; /* doesn't match */ | |
5038 break; | |
5039 } | |
7 | 5040 if (clnum == reg_endpos[no].lnum) |
5041 break; /* match and at end! */ | |
481 | 5042 if (reglnum >= reg_maxline) |
180 | 5043 { |
5044 status = RA_NOMATCH; /* text too short */ | |
5045 break; | |
5046 } | |
7 | 5047 |
5048 /* Advance to next line. */ | |
5049 reg_nextline(); | |
5050 ++clnum; | |
5051 ccol = 0; | |
180 | 5052 if (got_int) |
5053 { | |
5054 status = RA_FAIL; | |
5055 break; | |
5056 } | |
7 | 5057 } |
5058 | |
5059 /* found a match! Note that regline may now point | |
5060 * to a copy of the line, that should not matter. */ | |
5061 } | |
5062 } | |
5063 } | |
5064 | |
5065 /* Matched the backref, skip over it. */ | |
5066 reginput += len; | |
5067 } | |
5068 break; | |
5069 | |
5070 #ifdef FEAT_SYN_HL | |
5071 case ZREF + 1: | |
5072 case ZREF + 2: | |
5073 case ZREF + 3: | |
5074 case ZREF + 4: | |
5075 case ZREF + 5: | |
5076 case ZREF + 6: | |
5077 case ZREF + 7: | |
5078 case ZREF + 8: | |
5079 case ZREF + 9: | |
5080 { | |
5081 int len; | |
5082 | |
5083 cleanup_zsubexpr(); | |
5084 no = op - ZREF; | |
5085 if (re_extmatch_in != NULL | |
5086 && re_extmatch_in->matches[no] != NULL) | |
5087 { | |
5088 len = (int)STRLEN(re_extmatch_in->matches[no]); | |
5089 if (cstrncmp(re_extmatch_in->matches[no], | |
5090 reginput, &len) != 0) | |
180 | 5091 status = RA_NOMATCH; |
5092 else | |
5093 reginput += len; | |
7 | 5094 } |
5095 else | |
5096 { | |
5097 /* Backref was not set: Match an empty string. */ | |
5098 } | |
5099 } | |
5100 break; | |
5101 #endif | |
5102 | |
5103 case BRANCH: | |
5104 { | |
5105 if (OP(next) != BRANCH) /* No choice. */ | |
5106 next = OPERAND(scan); /* Avoid recursion. */ | |
5107 else | |
5108 { | |
270 | 5109 rp = regstack_push(RS_BRANCH, scan); |
180 | 5110 if (rp == NULL) |
5111 status = RA_FAIL; | |
5112 else | |
5113 status = RA_BREAK; /* rest is below */ | |
7 | 5114 } |
5115 } | |
5116 break; | |
5117 | |
5118 case BRACE_LIMITS: | |
5119 { | |
5120 if (OP(next) == BRACE_SIMPLE) | |
5121 { | |
5122 bl_minval = OPERAND_MIN(scan); | |
5123 bl_maxval = OPERAND_MAX(scan); | |
5124 } | |
5125 else if (OP(next) >= BRACE_COMPLEX | |
5126 && OP(next) < BRACE_COMPLEX + 10) | |
5127 { | |
5128 no = OP(next) - BRACE_COMPLEX; | |
5129 brace_min[no] = OPERAND_MIN(scan); | |
5130 brace_max[no] = OPERAND_MAX(scan); | |
5131 brace_count[no] = 0; | |
5132 } | |
5133 else | |
5134 { | |
5135 EMSG(_(e_internal)); /* Shouldn't happen */ | |
180 | 5136 status = RA_FAIL; |
7 | 5137 } |
5138 } | |
5139 break; | |
5140 | |
5141 case BRACE_COMPLEX + 0: | |
5142 case BRACE_COMPLEX + 1: | |
5143 case BRACE_COMPLEX + 2: | |
5144 case BRACE_COMPLEX + 3: | |
5145 case BRACE_COMPLEX + 4: | |
5146 case BRACE_COMPLEX + 5: | |
5147 case BRACE_COMPLEX + 6: | |
5148 case BRACE_COMPLEX + 7: | |
5149 case BRACE_COMPLEX + 8: | |
5150 case BRACE_COMPLEX + 9: | |
5151 { | |
5152 no = op - BRACE_COMPLEX; | |
5153 ++brace_count[no]; | |
5154 | |
5155 /* If not matched enough times yet, try one more */ | |
5156 if (brace_count[no] <= (brace_min[no] <= brace_max[no] | |
180 | 5157 ? brace_min[no] : brace_max[no])) |
7 | 5158 { |
270 | 5159 rp = regstack_push(RS_BRCPLX_MORE, scan); |
180 | 5160 if (rp == NULL) |
5161 status = RA_FAIL; | |
5162 else | |
5163 { | |
5164 rp->rs_no = no; | |
233 | 5165 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5166 next = OPERAND(scan); |
5167 /* We continue and handle the result when done. */ | |
5168 } | |
5169 break; | |
7 | 5170 } |
5171 | |
5172 /* If matched enough times, may try matching some more */ | |
5173 if (brace_min[no] <= brace_max[no]) | |
5174 { | |
5175 /* Range is the normal way around, use longest match */ | |
5176 if (brace_count[no] <= brace_max[no]) | |
5177 { | |
270 | 5178 rp = regstack_push(RS_BRCPLX_LONG, scan); |
180 | 5179 if (rp == NULL) |
5180 status = RA_FAIL; | |
5181 else | |
5182 { | |
5183 rp->rs_no = no; | |
233 | 5184 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5185 next = OPERAND(scan); |
5186 /* We continue and handle the result when done. */ | |
5187 } | |
7 | 5188 } |
5189 } | |
5190 else | |
5191 { | |
5192 /* Range is backwards, use shortest match first */ | |
5193 if (brace_count[no] <= brace_min[no]) | |
5194 { | |
270 | 5195 rp = regstack_push(RS_BRCPLX_SHORT, scan); |
180 | 5196 if (rp == NULL) |
5197 status = RA_FAIL; | |
5198 else | |
5199 { | |
233 | 5200 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5201 /* We continue and handle the result when done. */ |
5202 } | |
7 | 5203 } |
5204 } | |
5205 } | |
5206 break; | |
5207 | |
5208 case BRACE_SIMPLE: | |
5209 case STAR: | |
5210 case PLUS: | |
5211 { | |
180 | 5212 regstar_T rst; |
7 | 5213 |
5214 /* | |
5215 * Lookahead to avoid useless match attempts when we know | |
5216 * what character comes next. | |
5217 */ | |
5218 if (OP(next) == EXACTLY) | |
5219 { | |
180 | 5220 rst.nextb = *OPERAND(next); |
7 | 5221 if (ireg_ic) |
5222 { | |
1347 | 5223 if (MB_ISUPPER(rst.nextb)) |
5224 rst.nextb_ic = MB_TOLOWER(rst.nextb); | |
7 | 5225 else |
1347 | 5226 rst.nextb_ic = MB_TOUPPER(rst.nextb); |
7 | 5227 } |
5228 else | |
180 | 5229 rst.nextb_ic = rst.nextb; |
7 | 5230 } |
5231 else | |
5232 { | |
180 | 5233 rst.nextb = NUL; |
5234 rst.nextb_ic = NUL; | |
7 | 5235 } |
5236 if (op != BRACE_SIMPLE) | |
5237 { | |
180 | 5238 rst.minval = (op == STAR) ? 0 : 1; |
5239 rst.maxval = MAX_LIMIT; | |
7 | 5240 } |
5241 else | |
5242 { | |
180 | 5243 rst.minval = bl_minval; |
5244 rst.maxval = bl_maxval; | |
7 | 5245 } |
5246 | |
5247 /* | |
5248 * When maxval > minval, try matching as much as possible, up | |
5249 * to maxval. When maxval < minval, try matching at least the | |
5250 * minimal number (since the range is backwards, that's also | |
5251 * maxval!). | |
5252 */ | |
180 | 5253 rst.count = regrepeat(OPERAND(scan), rst.maxval); |
7 | 5254 if (got_int) |
180 | 5255 { |
5256 status = RA_FAIL; | |
5257 break; | |
5258 } | |
5259 if (rst.minval <= rst.maxval | |
5260 ? rst.count >= rst.minval : rst.count >= rst.maxval) | |
7 | 5261 { |
180 | 5262 /* It could match. Prepare for trying to match what |
5263 * follows. The code is below. Parameters are stored in | |
5264 * a regstar_T on the regstack. */ | |
212 | 5265 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5266 { |
5267 EMSG(_(e_maxmempat)); | |
5268 status = RA_FAIL; | |
5269 } | |
5270 else if (ga_grow(®stack, sizeof(regstar_T)) == FAIL) | |
180 | 5271 status = RA_FAIL; |
5272 else | |
7 | 5273 { |
180 | 5274 regstack.ga_len += sizeof(regstar_T); |
270 | 5275 rp = regstack_push(rst.minval <= rst.maxval |
233 | 5276 ? RS_STAR_LONG : RS_STAR_SHORT, scan); |
180 | 5277 if (rp == NULL) |
5278 status = RA_FAIL; | |
5279 else | |
7 | 5280 { |
180 | 5281 *(((regstar_T *)rp) - 1) = rst; |
5282 status = RA_BREAK; /* skip the restore bits */ | |
7 | 5283 } |
5284 } | |
5285 } | |
5286 else | |
180 | 5287 status = RA_NOMATCH; |
5288 | |
7 | 5289 } |
180 | 5290 break; |
7 | 5291 |
5292 case NOMATCH: | |
5293 case MATCH: | |
5294 case SUBPAT: | |
270 | 5295 rp = regstack_push(RS_NOMATCH, scan); |
180 | 5296 if (rp == NULL) |
5297 status = RA_FAIL; | |
5298 else | |
7 | 5299 { |
180 | 5300 rp->rs_no = op; |
233 | 5301 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5302 next = OPERAND(scan); |
5303 /* We continue and handle the result when done. */ | |
7 | 5304 } |
5305 break; | |
5306 | |
5307 case BEHIND: | |
5308 case NOBEHIND: | |
180 | 5309 /* Need a bit of room to store extra positions. */ |
212 | 5310 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5311 { |
5312 EMSG(_(e_maxmempat)); | |
5313 status = RA_FAIL; | |
5314 } | |
5315 else if (ga_grow(®stack, sizeof(regbehind_T)) == FAIL) | |
180 | 5316 status = RA_FAIL; |
5317 else | |
7 | 5318 { |
180 | 5319 regstack.ga_len += sizeof(regbehind_T); |
270 | 5320 rp = regstack_push(RS_BEHIND1, scan); |
180 | 5321 if (rp == NULL) |
5322 status = RA_FAIL; | |
5323 else | |
7 | 5324 { |
1579 | 5325 /* Need to save the subexpr to be able to restore them |
5326 * when there is a match but we don't use it. */ | |
5327 save_subexpr(((regbehind_T *)rp) - 1); | |
5328 | |
180 | 5329 rp->rs_no = op; |
233 | 5330 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5331 /* First try if what follows matches. If it does then we |
5332 * check the behind match by looping. */ | |
7 | 5333 } |
5334 } | |
180 | 5335 break; |
7 | 5336 |
5337 case BHPOS: | |
5338 if (REG_MULTI) | |
5339 { | |
5340 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline) | |
5341 || behind_pos.rs_u.pos.lnum != reglnum) | |
180 | 5342 status = RA_NOMATCH; |
7 | 5343 } |
5344 else if (behind_pos.rs_u.ptr != reginput) | |
180 | 5345 status = RA_NOMATCH; |
7 | 5346 break; |
5347 | |
5348 case NEWL: | |
1018 | 5349 if ((c != NUL || !REG_MULTI || reglnum > reg_maxline |
5350 || reg_line_lbr) && (c != '\n' || !reg_line_lbr)) | |
180 | 5351 status = RA_NOMATCH; |
5352 else if (reg_line_lbr) | |
7 | 5353 ADVANCE_REGINPUT(); |
5354 else | |
5355 reg_nextline(); | |
5356 break; | |
5357 | |
5358 case END: | |
180 | 5359 status = RA_MATCH; /* Success! */ |
5360 break; | |
7 | 5361 |
5362 default: | |
5363 EMSG(_(e_re_corr)); | |
5364 #ifdef DEBUG | |
5365 printf("Illegal op code %d\n", op); | |
5366 #endif | |
180 | 5367 status = RA_FAIL; |
5368 break; | |
7 | 5369 } |
5370 } | |
5371 | |
180 | 5372 /* If we can't continue sequentially, break the inner loop. */ |
5373 if (status != RA_CONT) | |
5374 break; | |
5375 | |
5376 /* Continue in inner loop, advance to next item. */ | |
7 | 5377 scan = next; |
180 | 5378 |
5379 } /* end of inner loop */ | |
7 | 5380 |
5381 /* | |
180 | 5382 * If there is something on the regstack execute the code for the state. |
233 | 5383 * If the state is popped then loop and use the older state. |
7 | 5384 */ |
180 | 5385 while (regstack.ga_len > 0 && status != RA_FAIL) |
5386 { | |
5387 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1; | |
5388 switch (rp->rs_state) | |
5389 { | |
5390 case RS_NOPEN: | |
5391 /* Result is passed on as-is, simply pop the state. */ | |
270 | 5392 regstack_pop(&scan); |
180 | 5393 break; |
5394 | |
5395 case RS_MOPEN: | |
5396 /* Pop the state. Restore pointers when there is no match. */ | |
5397 if (status == RA_NOMATCH) | |
5398 restore_se(&rp->rs_un.sesave, ®_startpos[rp->rs_no], | |
5399 ®_startp[rp->rs_no]); | |
270 | 5400 regstack_pop(&scan); |
180 | 5401 break; |
5402 | |
5403 #ifdef FEAT_SYN_HL | |
5404 case RS_ZOPEN: | |
5405 /* Pop the state. Restore pointers when there is no match. */ | |
5406 if (status == RA_NOMATCH) | |
5407 restore_se(&rp->rs_un.sesave, ®_startzpos[rp->rs_no], | |
5408 ®_startzp[rp->rs_no]); | |
270 | 5409 regstack_pop(&scan); |
180 | 5410 break; |
5411 #endif | |
5412 | |
5413 case RS_MCLOSE: | |
5414 /* Pop the state. Restore pointers when there is no match. */ | |
5415 if (status == RA_NOMATCH) | |
5416 restore_se(&rp->rs_un.sesave, ®_endpos[rp->rs_no], | |
5417 ®_endp[rp->rs_no]); | |
270 | 5418 regstack_pop(&scan); |
180 | 5419 break; |
5420 | |
5421 #ifdef FEAT_SYN_HL | |
5422 case RS_ZCLOSE: | |
5423 /* Pop the state. Restore pointers when there is no match. */ | |
5424 if (status == RA_NOMATCH) | |
5425 restore_se(&rp->rs_un.sesave, ®_endzpos[rp->rs_no], | |
5426 ®_endzp[rp->rs_no]); | |
270 | 5427 regstack_pop(&scan); |
180 | 5428 break; |
5429 #endif | |
5430 | |
5431 case RS_BRANCH: | |
5432 if (status == RA_MATCH) | |
5433 /* this branch matched, use it */ | |
270 | 5434 regstack_pop(&scan); |
180 | 5435 else |
5436 { | |
5437 if (status != RA_BREAK) | |
5438 { | |
5439 /* After a non-matching branch: try next one. */ | |
233 | 5440 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5441 scan = rp->rs_scan; |
5442 } | |
5443 if (scan == NULL || OP(scan) != BRANCH) | |
5444 { | |
5445 /* no more branches, didn't find a match */ | |
5446 status = RA_NOMATCH; | |
270 | 5447 regstack_pop(&scan); |
180 | 5448 } |
5449 else | |
5450 { | |
5451 /* Prepare to try a branch. */ | |
5452 rp->rs_scan = regnext(scan); | |
233 | 5453 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5454 scan = OPERAND(scan); |
5455 } | |
5456 } | |
5457 break; | |
5458 | |
5459 case RS_BRCPLX_MORE: | |
5460 /* Pop the state. Restore pointers when there is no match. */ | |
5461 if (status == RA_NOMATCH) | |
5462 { | |
233 | 5463 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5464 --brace_count[rp->rs_no]; /* decrement match count */ |
5465 } | |
270 | 5466 regstack_pop(&scan); |
180 | 5467 break; |
5468 | |
5469 case RS_BRCPLX_LONG: | |
5470 /* Pop the state. Restore pointers when there is no match. */ | |
5471 if (status == RA_NOMATCH) | |
5472 { | |
5473 /* There was no match, but we did find enough matches. */ | |
233 | 5474 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5475 --brace_count[rp->rs_no]; |
5476 /* continue with the items after "\{}" */ | |
5477 status = RA_CONT; | |
5478 } | |
270 | 5479 regstack_pop(&scan); |
180 | 5480 if (status == RA_CONT) |
5481 scan = regnext(scan); | |
5482 break; | |
5483 | |
5484 case RS_BRCPLX_SHORT: | |
5485 /* Pop the state. Restore pointers when there is no match. */ | |
5486 if (status == RA_NOMATCH) | |
5487 /* There was no match, try to match one more item. */ | |
233 | 5488 reg_restore(&rp->rs_un.regsave, &backpos); |
270 | 5489 regstack_pop(&scan); |
180 | 5490 if (status == RA_NOMATCH) |
5491 { | |
5492 scan = OPERAND(scan); | |
5493 status = RA_CONT; | |
5494 } | |
5495 break; | |
5496 | |
5497 case RS_NOMATCH: | |
5498 /* Pop the state. If the operand matches for NOMATCH or | |
5499 * doesn't match for MATCH/SUBPAT, we fail. Otherwise backup, | |
5500 * except for SUBPAT, and continue with the next item. */ | |
5501 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH)) | |
5502 status = RA_NOMATCH; | |
5503 else | |
5504 { | |
5505 status = RA_CONT; | |
233 | 5506 if (rp->rs_no != SUBPAT) /* zero-width */ |
5507 reg_restore(&rp->rs_un.regsave, &backpos); | |
180 | 5508 } |
270 | 5509 regstack_pop(&scan); |
180 | 5510 if (status == RA_CONT) |
5511 scan = regnext(scan); | |
5512 break; | |
5513 | |
5514 case RS_BEHIND1: | |
5515 if (status == RA_NOMATCH) | |
5516 { | |
270 | 5517 regstack_pop(&scan); |
180 | 5518 regstack.ga_len -= sizeof(regbehind_T); |
5519 } | |
5520 else | |
5521 { | |
5522 /* The stuff after BEHIND/NOBEHIND matches. Now try if | |
5523 * the behind part does (not) match before the current | |
5524 * position in the input. This must be done at every | |
5525 * position in the input and checking if the match ends at | |
5526 * the current position. */ | |
5527 | |
5528 /* save the position after the found match for next */ | |
233 | 5529 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos); |
180 | 5530 |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5531 /* Start looking for a match with operand at the current |
1209 | 5532 * position. Go back one character until we find the |
180 | 5533 * result, hitting the start of the line or the previous |
5534 * line (for multi-line matching). | |
5535 * Set behind_pos to where the match should end, BHPOS | |
5536 * will match it. Save the current value. */ | |
5537 (((regbehind_T *)rp) - 1)->save_behind = behind_pos; | |
5538 behind_pos = rp->rs_un.regsave; | |
5539 | |
5540 rp->rs_state = RS_BEHIND2; | |
5541 | |
233 | 5542 reg_restore(&rp->rs_un.regsave, &backpos); |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5543 scan = OPERAND(rp->rs_scan) + 4; |
180 | 5544 } |
5545 break; | |
5546 | |
5547 case RS_BEHIND2: | |
5548 /* | |
5549 * Looping for BEHIND / NOBEHIND match. | |
5550 */ | |
5551 if (status == RA_MATCH && reg_save_equal(&behind_pos)) | |
5552 { | |
5553 /* found a match that ends where "next" started */ | |
5554 behind_pos = (((regbehind_T *)rp) - 1)->save_behind; | |
5555 if (rp->rs_no == BEHIND) | |
233 | 5556 reg_restore(&(((regbehind_T *)rp) - 1)->save_after, |
5557 &backpos); | |
180 | 5558 else |
1579 | 5559 { |
5560 /* But we didn't want a match. Need to restore the | |
5561 * subexpr, because what follows matched, so they have | |
5562 * been set. */ | |
180 | 5563 status = RA_NOMATCH; |
1579 | 5564 restore_subexpr(((regbehind_T *)rp) - 1); |
5565 } | |
270 | 5566 regstack_pop(&scan); |
180 | 5567 regstack.ga_len -= sizeof(regbehind_T); |
5568 } | |
5569 else | |
5570 { | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5571 long limit; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5572 |
1579 | 5573 /* No match or a match that doesn't end where we want it: Go |
5574 * back one character. May go to previous line once. */ | |
180 | 5575 no = OK; |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5576 limit = OPERAND_MIN(rp->rs_scan); |
180 | 5577 if (REG_MULTI) |
5578 { | |
4682
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5579 if (limit > 0 |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5580 && ((rp->rs_un.regsave.rs_u.pos.lnum |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5581 < behind_pos.rs_u.pos.lnum |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5582 ? (colnr_T)STRLEN(regline) |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5583 : behind_pos.rs_u.pos.col) |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5584 - rp->rs_un.regsave.rs_u.pos.col >= limit)) |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5585 no = FAIL; |
2f51ee8825db
updated for version 7.3.1088
Bram Moolenaar <bram@vim.org>
parents:
4679
diff
changeset
|
5586 else if (rp->rs_un.regsave.rs_u.pos.col == 0) |
180 | 5587 { |
5588 if (rp->rs_un.regsave.rs_u.pos.lnum | |
5589 < behind_pos.rs_u.pos.lnum | |
5590 || reg_getline( | |
5591 --rp->rs_un.regsave.rs_u.pos.lnum) | |
5592 == NULL) | |
5593 no = FAIL; | |
5594 else | |
5595 { | |
233 | 5596 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5597 rp->rs_un.regsave.rs_u.pos.col = |
5598 (colnr_T)STRLEN(regline); | |
5599 } | |
5600 } | |
5601 else | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5602 { |
4176 | 5603 #ifdef FEAT_MBYTE |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5604 if (has_mbyte) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5605 rp->rs_un.regsave.rs_u.pos.col -= |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5606 (*mb_head_off)(regline, regline |
4176 | 5607 + rp->rs_un.regsave.rs_u.pos.col - 1) + 1; |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5608 else |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5609 #endif |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5610 --rp->rs_un.regsave.rs_u.pos.col; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5611 } |
180 | 5612 } |
5613 else | |
5614 { | |
5615 if (rp->rs_un.regsave.rs_u.ptr == regline) | |
5616 no = FAIL; | |
5617 else | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5618 { |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5619 mb_ptr_back(regline, rp->rs_un.regsave.rs_u.ptr); |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5620 if (limit > 0 && (long)(behind_pos.rs_u.ptr |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5621 - rp->rs_un.regsave.rs_u.ptr) > limit) |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5622 no = FAIL; |
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5623 } |
180 | 5624 } |
5625 if (no == OK) | |
5626 { | |
5627 /* Advanced, prepare for finding match again. */ | |
233 | 5628 reg_restore(&rp->rs_un.regsave, &backpos); |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
5629 scan = OPERAND(rp->rs_scan) + 4; |
1579 | 5630 if (status == RA_MATCH) |
5631 { | |
5632 /* We did match, so subexpr may have been changed, | |
5633 * need to restore them for the next try. */ | |
5634 status = RA_NOMATCH; | |
5635 restore_subexpr(((regbehind_T *)rp) - 1); | |
5636 } | |
180 | 5637 } |
5638 else | |
5639 { | |
5640 /* Can't advance. For NOBEHIND that's a match. */ | |
5641 behind_pos = (((regbehind_T *)rp) - 1)->save_behind; | |
5642 if (rp->rs_no == NOBEHIND) | |
5643 { | |
233 | 5644 reg_restore(&(((regbehind_T *)rp) - 1)->save_after, |
5645 &backpos); | |
180 | 5646 status = RA_MATCH; |
5647 } | |
5648 else | |
1579 | 5649 { |
5650 /* We do want a proper match. Need to restore the | |
5651 * subexpr if we had a match, because they may have | |
5652 * been set. */ | |
5653 if (status == RA_MATCH) | |
5654 { | |
5655 status = RA_NOMATCH; | |
5656 restore_subexpr(((regbehind_T *)rp) - 1); | |
5657 } | |
5658 } | |
270 | 5659 regstack_pop(&scan); |
180 | 5660 regstack.ga_len -= sizeof(regbehind_T); |
5661 } | |
5662 } | |
5663 break; | |
5664 | |
5665 case RS_STAR_LONG: | |
5666 case RS_STAR_SHORT: | |
5667 { | |
5668 regstar_T *rst = ((regstar_T *)rp) - 1; | |
5669 | |
5670 if (status == RA_MATCH) | |
5671 { | |
270 | 5672 regstack_pop(&scan); |
180 | 5673 regstack.ga_len -= sizeof(regstar_T); |
5674 break; | |
5675 } | |
5676 | |
5677 /* Tried once already, restore input pointers. */ | |
5678 if (status != RA_BREAK) | |
233 | 5679 reg_restore(&rp->rs_un.regsave, &backpos); |
180 | 5680 |
5681 /* Repeat until we found a position where it could match. */ | |
5682 for (;;) | |
5683 { | |
5684 if (status != RA_BREAK) | |
5685 { | |
5686 /* Tried first position already, advance. */ | |
5687 if (rp->rs_state == RS_STAR_LONG) | |
5688 { | |
685 | 5689 /* Trying for longest match, but couldn't or |
5690 * didn't match -- back up one char. */ | |
180 | 5691 if (--rst->count < rst->minval) |
5692 break; | |
5693 if (reginput == regline) | |
5694 { | |
5695 /* backup to last char of previous line */ | |
5696 --reglnum; | |
5697 regline = reg_getline(reglnum); | |
5698 /* Just in case regrepeat() didn't count | |
5699 * right. */ | |
5700 if (regline == NULL) | |
5701 break; | |
5702 reginput = regline + STRLEN(regline); | |
5703 fast_breakcheck(); | |
5704 } | |
5705 else | |
5706 mb_ptr_back(regline, reginput); | |
5707 } | |
5708 else | |
5709 { | |
5710 /* Range is backwards, use shortest match first. | |
5711 * Careful: maxval and minval are exchanged! | |
5712 * Couldn't or didn't match: try advancing one | |
5713 * char. */ | |
5714 if (rst->count == rst->minval | |
5715 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0) | |
5716 break; | |
5717 ++rst->count; | |
5718 } | |
5719 if (got_int) | |
5720 break; | |
5721 } | |
5722 else | |
5723 status = RA_NOMATCH; | |
5724 | |
5725 /* If it could match, try it. */ | |
5726 if (rst->nextb == NUL || *reginput == rst->nextb | |
5727 || *reginput == rst->nextb_ic) | |
5728 { | |
233 | 5729 reg_save(&rp->rs_un.regsave, &backpos); |
180 | 5730 scan = regnext(rp->rs_scan); |
5731 status = RA_CONT; | |
5732 break; | |
5733 } | |
5734 } | |
5735 if (status != RA_CONT) | |
5736 { | |
5737 /* Failed. */ | |
270 | 5738 regstack_pop(&scan); |
180 | 5739 regstack.ga_len -= sizeof(regstar_T); |
5740 status = RA_NOMATCH; | |
5741 } | |
5742 } | |
5743 break; | |
5744 } | |
5745 | |
685 | 5746 /* If we want to continue the inner loop or didn't pop a state |
5747 * continue matching loop */ | |
180 | 5748 if (status == RA_CONT || rp == (regitem_T *) |
5749 ((char *)regstack.ga_data + regstack.ga_len) - 1) | |
5750 break; | |
5751 } | |
5752 | |
189 | 5753 /* May need to continue with the inner loop, starting at "scan". */ |
180 | 5754 if (status == RA_CONT) |
5755 continue; | |
5756 | |
5757 /* | |
5758 * If the regstack is empty or something failed we are done. | |
5759 */ | |
5760 if (regstack.ga_len == 0 || status == RA_FAIL) | |
5761 { | |
5762 if (scan == NULL) | |
5763 { | |
5764 /* | |
5765 * We get here only if there's trouble -- normally "case END" is | |
5766 * the terminating point. | |
5767 */ | |
5768 EMSG(_(e_re_corr)); | |
7 | 5769 #ifdef DEBUG |
180 | 5770 printf("Premature EOL\n"); |
7 | 5771 #endif |
180 | 5772 } |
189 | 5773 if (status == RA_FAIL) |
5774 got_int = TRUE; | |
180 | 5775 return (status == RA_MATCH); |
5776 } | |
5777 | |
5778 } /* End of loop until the regstack is empty. */ | |
5779 | |
5780 /* NOTREACHED */ | |
5781 } | |
5782 | |
5783 /* | |
5784 * Push an item onto the regstack. | |
5785 * Returns pointer to new item. Returns NULL when out of memory. | |
5786 */ | |
5787 static regitem_T * | |
270 | 5788 regstack_push(state, scan) |
180 | 5789 regstate_T state; |
5790 char_u *scan; | |
5791 { | |
5792 regitem_T *rp; | |
5793 | |
270 | 5794 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp) |
189 | 5795 { |
5796 EMSG(_(e_maxmempat)); | |
5797 return NULL; | |
5798 } | |
270 | 5799 if (ga_grow(®stack, sizeof(regitem_T)) == FAIL) |
180 | 5800 return NULL; |
5801 | |
270 | 5802 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len); |
180 | 5803 rp->rs_state = state; |
5804 rp->rs_scan = scan; | |
5805 | |
270 | 5806 regstack.ga_len += sizeof(regitem_T); |
180 | 5807 return rp; |
5808 } | |
5809 | |
5810 /* | |
5811 * Pop an item from the regstack. | |
5812 */ | |
5813 static void | |
270 | 5814 regstack_pop(scan) |
180 | 5815 char_u **scan; |
5816 { | |
5817 regitem_T *rp; | |
5818 | |
270 | 5819 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1; |
180 | 5820 *scan = rp->rs_scan; |
5821 | |
270 | 5822 regstack.ga_len -= sizeof(regitem_T); |
7 | 5823 } |
5824 | |
5825 /* | |
5826 * regrepeat - repeatedly match something simple, return how many. | |
5827 * Advances reginput (and reglnum) to just after the matched chars. | |
5828 */ | |
5829 static int | |
5830 regrepeat(p, maxcount) | |
5831 char_u *p; | |
5832 long maxcount; /* maximum number of matches allowed */ | |
5833 { | |
5834 long count = 0; | |
5835 char_u *scan; | |
5836 char_u *opnd; | |
5837 int mask; | |
5838 int testval = 0; | |
5839 | |
5840 scan = reginput; /* Make local copy of reginput for speed. */ | |
5841 opnd = OPERAND(p); | |
5842 switch (OP(p)) | |
5843 { | |
5844 case ANY: | |
5845 case ANY + ADD_NL: | |
5846 while (count < maxcount) | |
5847 { | |
5848 /* Matching anything means we continue until end-of-line (or | |
5849 * end-of-file for ANY + ADD_NL), only limited by maxcount. */ | |
5850 while (*scan != NUL && count < maxcount) | |
5851 { | |
5852 ++count; | |
39 | 5853 mb_ptr_adv(scan); |
7 | 5854 } |
1018 | 5855 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5856 || reg_line_lbr || count == maxcount) | |
7 | 5857 break; |
5858 ++count; /* count the line-break */ | |
5859 reg_nextline(); | |
5860 scan = reginput; | |
5861 if (got_int) | |
5862 break; | |
5863 } | |
5864 break; | |
5865 | |
5866 case IDENT: | |
5867 case IDENT + ADD_NL: | |
5868 testval = TRUE; | |
5869 /*FALLTHROUGH*/ | |
5870 case SIDENT: | |
5871 case SIDENT + ADD_NL: | |
5872 while (count < maxcount) | |
5873 { | |
4466 | 5874 if (vim_isIDc(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan))) |
7 | 5875 { |
39 | 5876 mb_ptr_adv(scan); |
7 | 5877 } |
5878 else if (*scan == NUL) | |
5879 { | |
1018 | 5880 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5881 || reg_line_lbr) | |
7 | 5882 break; |
5883 reg_nextline(); | |
5884 scan = reginput; | |
5885 if (got_int) | |
5886 break; | |
5887 } | |
5888 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5889 ++scan; | |
5890 else | |
5891 break; | |
5892 ++count; | |
5893 } | |
5894 break; | |
5895 | |
5896 case KWORD: | |
5897 case KWORD + ADD_NL: | |
5898 testval = TRUE; | |
5899 /*FALLTHROUGH*/ | |
5900 case SKWORD: | |
5901 case SKWORD + ADD_NL: | |
5902 while (count < maxcount) | |
5903 { | |
4069 | 5904 if (vim_iswordp_buf(scan, reg_buf) |
5905 && (testval || !VIM_ISDIGIT(*scan))) | |
7 | 5906 { |
39 | 5907 mb_ptr_adv(scan); |
7 | 5908 } |
5909 else if (*scan == NUL) | |
5910 { | |
1018 | 5911 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5912 || reg_line_lbr) | |
7 | 5913 break; |
5914 reg_nextline(); | |
5915 scan = reginput; | |
5916 if (got_int) | |
5917 break; | |
5918 } | |
5919 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5920 ++scan; | |
5921 else | |
5922 break; | |
5923 ++count; | |
5924 } | |
5925 break; | |
5926 | |
5927 case FNAME: | |
5928 case FNAME + ADD_NL: | |
5929 testval = TRUE; | |
5930 /*FALLTHROUGH*/ | |
5931 case SFNAME: | |
5932 case SFNAME + ADD_NL: | |
5933 while (count < maxcount) | |
5934 { | |
4466 | 5935 if (vim_isfilec(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan))) |
7 | 5936 { |
39 | 5937 mb_ptr_adv(scan); |
7 | 5938 } |
5939 else if (*scan == NUL) | |
5940 { | |
1018 | 5941 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5942 || reg_line_lbr) | |
7 | 5943 break; |
5944 reg_nextline(); | |
5945 scan = reginput; | |
5946 if (got_int) | |
5947 break; | |
5948 } | |
5949 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5950 ++scan; | |
5951 else | |
5952 break; | |
5953 ++count; | |
5954 } | |
5955 break; | |
5956 | |
5957 case PRINT: | |
5958 case PRINT + ADD_NL: | |
5959 testval = TRUE; | |
5960 /*FALLTHROUGH*/ | |
5961 case SPRINT: | |
5962 case SPRINT + ADD_NL: | |
5963 while (count < maxcount) | |
5964 { | |
5965 if (*scan == NUL) | |
5966 { | |
1018 | 5967 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5968 || reg_line_lbr) | |
7 | 5969 break; |
5970 reg_nextline(); | |
5971 scan = reginput; | |
5972 if (got_int) | |
5973 break; | |
5974 } | |
5975 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan))) | |
5976 { | |
39 | 5977 mb_ptr_adv(scan); |
7 | 5978 } |
5979 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
5980 ++scan; | |
5981 else | |
5982 break; | |
5983 ++count; | |
5984 } | |
5985 break; | |
5986 | |
5987 case WHITE: | |
5988 case WHITE + ADD_NL: | |
5989 testval = mask = RI_WHITE; | |
5990 do_class: | |
5991 while (count < maxcount) | |
5992 { | |
5993 #ifdef FEAT_MBYTE | |
5994 int l; | |
5995 #endif | |
5996 if (*scan == NUL) | |
5997 { | |
1018 | 5998 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
5999 || reg_line_lbr) | |
7 | 6000 break; |
6001 reg_nextline(); | |
6002 scan = reginput; | |
6003 if (got_int) | |
6004 break; | |
6005 } | |
6006 #ifdef FEAT_MBYTE | |
474 | 6007 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1) |
7 | 6008 { |
6009 if (testval != 0) | |
6010 break; | |
6011 scan += l; | |
6012 } | |
6013 #endif | |
6014 else if ((class_tab[*scan] & mask) == testval) | |
6015 ++scan; | |
6016 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
6017 ++scan; | |
6018 else | |
6019 break; | |
6020 ++count; | |
6021 } | |
6022 break; | |
6023 | |
6024 case NWHITE: | |
6025 case NWHITE + ADD_NL: | |
6026 mask = RI_WHITE; | |
6027 goto do_class; | |
6028 case DIGIT: | |
6029 case DIGIT + ADD_NL: | |
6030 testval = mask = RI_DIGIT; | |
6031 goto do_class; | |
6032 case NDIGIT: | |
6033 case NDIGIT + ADD_NL: | |
6034 mask = RI_DIGIT; | |
6035 goto do_class; | |
6036 case HEX: | |
6037 case HEX + ADD_NL: | |
6038 testval = mask = RI_HEX; | |
6039 goto do_class; | |
6040 case NHEX: | |
6041 case NHEX + ADD_NL: | |
6042 mask = RI_HEX; | |
6043 goto do_class; | |
6044 case OCTAL: | |
6045 case OCTAL + ADD_NL: | |
6046 testval = mask = RI_OCTAL; | |
6047 goto do_class; | |
6048 case NOCTAL: | |
6049 case NOCTAL + ADD_NL: | |
6050 mask = RI_OCTAL; | |
6051 goto do_class; | |
6052 case WORD: | |
6053 case WORD + ADD_NL: | |
6054 testval = mask = RI_WORD; | |
6055 goto do_class; | |
6056 case NWORD: | |
6057 case NWORD + ADD_NL: | |
6058 mask = RI_WORD; | |
6059 goto do_class; | |
6060 case HEAD: | |
6061 case HEAD + ADD_NL: | |
6062 testval = mask = RI_HEAD; | |
6063 goto do_class; | |
6064 case NHEAD: | |
6065 case NHEAD + ADD_NL: | |
6066 mask = RI_HEAD; | |
6067 goto do_class; | |
6068 case ALPHA: | |
6069 case ALPHA + ADD_NL: | |
6070 testval = mask = RI_ALPHA; | |
6071 goto do_class; | |
6072 case NALPHA: | |
6073 case NALPHA + ADD_NL: | |
6074 mask = RI_ALPHA; | |
6075 goto do_class; | |
6076 case LOWER: | |
6077 case LOWER + ADD_NL: | |
6078 testval = mask = RI_LOWER; | |
6079 goto do_class; | |
6080 case NLOWER: | |
6081 case NLOWER + ADD_NL: | |
6082 mask = RI_LOWER; | |
6083 goto do_class; | |
6084 case UPPER: | |
6085 case UPPER + ADD_NL: | |
6086 testval = mask = RI_UPPER; | |
6087 goto do_class; | |
6088 case NUPPER: | |
6089 case NUPPER + ADD_NL: | |
6090 mask = RI_UPPER; | |
6091 goto do_class; | |
6092 | |
6093 case EXACTLY: | |
6094 { | |
6095 int cu, cl; | |
6096 | |
6097 /* This doesn't do a multi-byte character, because a MULTIBYTECODE | |
1347 | 6098 * would have been used for it. It does handle single-byte |
6099 * characters, such as latin1. */ | |
7 | 6100 if (ireg_ic) |
6101 { | |
1347 | 6102 cu = MB_TOUPPER(*opnd); |
6103 cl = MB_TOLOWER(*opnd); | |
7 | 6104 while (count < maxcount && (*scan == cu || *scan == cl)) |
6105 { | |
6106 count++; | |
6107 scan++; | |
6108 } | |
6109 } | |
6110 else | |
6111 { | |
6112 cu = *opnd; | |
6113 while (count < maxcount && *scan == cu) | |
6114 { | |
6115 count++; | |
6116 scan++; | |
6117 } | |
6118 } | |
6119 break; | |
6120 } | |
6121 | |
6122 #ifdef FEAT_MBYTE | |
6123 case MULTIBYTECODE: | |
6124 { | |
6125 int i, len, cf = 0; | |
6126 | |
6127 /* Safety check (just in case 'encoding' was changed since | |
6128 * compiling the program). */ | |
474 | 6129 if ((len = (*mb_ptr2len)(opnd)) > 1) |
7 | 6130 { |
6131 if (ireg_ic && enc_utf8) | |
6132 cf = utf_fold(utf_ptr2char(opnd)); | |
6133 while (count < maxcount) | |
6134 { | |
6135 for (i = 0; i < len; ++i) | |
6136 if (opnd[i] != scan[i]) | |
6137 break; | |
6138 if (i < len && (!ireg_ic || !enc_utf8 | |
6139 || utf_fold(utf_ptr2char(scan)) != cf)) | |
6140 break; | |
6141 scan += len; | |
6142 ++count; | |
6143 } | |
6144 } | |
6145 } | |
6146 break; | |
6147 #endif | |
6148 | |
6149 case ANYOF: | |
6150 case ANYOF + ADD_NL: | |
6151 testval = TRUE; | |
6152 /*FALLTHROUGH*/ | |
6153 | |
6154 case ANYBUT: | |
6155 case ANYBUT + ADD_NL: | |
6156 while (count < maxcount) | |
6157 { | |
6158 #ifdef FEAT_MBYTE | |
6159 int len; | |
6160 #endif | |
6161 if (*scan == NUL) | |
6162 { | |
1018 | 6163 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline |
6164 || reg_line_lbr) | |
7 | 6165 break; |
6166 reg_nextline(); | |
6167 scan = reginput; | |
6168 if (got_int) | |
6169 break; | |
6170 } | |
6171 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p))) | |
6172 ++scan; | |
6173 #ifdef FEAT_MBYTE | |
474 | 6174 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1) |
7 | 6175 { |
6176 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval) | |
6177 break; | |
6178 scan += len; | |
6179 } | |
6180 #endif | |
6181 else | |
6182 { | |
6183 if ((cstrchr(opnd, *scan) == NULL) == testval) | |
6184 break; | |
6185 ++scan; | |
6186 } | |
6187 ++count; | |
6188 } | |
6189 break; | |
6190 | |
6191 case NEWL: | |
6192 while (count < maxcount | |
1018 | 6193 && ((*scan == NUL && reglnum <= reg_maxline && !reg_line_lbr |
6194 && REG_MULTI) || (*scan == '\n' && reg_line_lbr))) | |
7 | 6195 { |
6196 count++; | |
6197 if (reg_line_lbr) | |
6198 ADVANCE_REGINPUT(); | |
6199 else | |
6200 reg_nextline(); | |
6201 scan = reginput; | |
6202 if (got_int) | |
6203 break; | |
6204 } | |
6205 break; | |
6206 | |
6207 default: /* Oh dear. Called inappropriately. */ | |
6208 EMSG(_(e_re_corr)); | |
6209 #ifdef DEBUG | |
6210 printf("Called regrepeat with op code %d\n", OP(p)); | |
6211 #endif | |
6212 break; | |
6213 } | |
6214 | |
6215 reginput = scan; | |
6216 | |
6217 return (int)count; | |
6218 } | |
6219 | |
6220 /* | |
6221 * regnext - dig the "next" pointer out of a node | |
2010 | 6222 * Returns NULL when calculating size, when there is no next item and when |
6223 * there is an error. | |
7 | 6224 */ |
6225 static char_u * | |
6226 regnext(p) | |
6227 char_u *p; | |
6228 { | |
6229 int offset; | |
6230 | |
2010 | 6231 if (p == JUST_CALC_SIZE || reg_toolong) |
7 | 6232 return NULL; |
6233 | |
6234 offset = NEXT(p); | |
6235 if (offset == 0) | |
6236 return NULL; | |
6237 | |
233 | 6238 if (OP(p) == BACK) |
7 | 6239 return p - offset; |
6240 else | |
6241 return p + offset; | |
6242 } | |
6243 | |
6244 /* | |
6245 * Check the regexp program for its magic number. | |
6246 * Return TRUE if it's wrong. | |
6247 */ | |
6248 static int | |
6249 prog_magic_wrong() | |
6250 { | |
4444 | 6251 regprog_T *prog; |
6252 | |
6253 prog = REG_MULTI ? reg_mmatch->regprog : reg_match->regprog; | |
6254 if (prog->engine == &nfa_regengine) | |
6255 /* For NFA matcher we don't check the magic */ | |
6256 return FALSE; | |
6257 | |
6258 if (UCHARAT(((bt_regprog_T *)prog)->program) != REGMAGIC) | |
7 | 6259 { |
6260 EMSG(_(e_re_corr)); | |
6261 return TRUE; | |
6262 } | |
6263 return FALSE; | |
6264 } | |
6265 | |
6266 /* | |
6267 * Cleanup the subexpressions, if this wasn't done yet. | |
6268 * This construction is used to clear the subexpressions only when they are | |
6269 * used (to increase speed). | |
6270 */ | |
6271 static void | |
6272 cleanup_subexpr() | |
6273 { | |
6274 if (need_clear_subexpr) | |
6275 { | |
6276 if (REG_MULTI) | |
6277 { | |
6278 /* Use 0xff to set lnum to -1 */ | |
6279 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6280 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6281 } | |
6282 else | |
6283 { | |
6284 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP); | |
6285 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP); | |
6286 } | |
6287 need_clear_subexpr = FALSE; | |
6288 } | |
6289 } | |
6290 | |
6291 #ifdef FEAT_SYN_HL | |
6292 static void | |
6293 cleanup_zsubexpr() | |
6294 { | |
6295 if (need_clear_zsubexpr) | |
6296 { | |
6297 if (REG_MULTI) | |
6298 { | |
6299 /* Use 0xff to set lnum to -1 */ | |
6300 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6301 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP); | |
6302 } | |
6303 else | |
6304 { | |
6305 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP); | |
6306 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP); | |
6307 } | |
6308 need_clear_zsubexpr = FALSE; | |
6309 } | |
6310 } | |
6311 #endif | |
6312 | |
6313 /* | |
1579 | 6314 * Save the current subexpr to "bp", so that they can be restored |
6315 * later by restore_subexpr(). | |
6316 */ | |
6317 static void | |
6318 save_subexpr(bp) | |
6319 regbehind_T *bp; | |
6320 { | |
6321 int i; | |
6322 | |
1602 | 6323 /* When "need_clear_subexpr" is set we don't need to save the values, only |
6324 * remember that this flag needs to be set again when restoring. */ | |
6325 bp->save_need_clear_subexpr = need_clear_subexpr; | |
6326 if (!need_clear_subexpr) | |
1579 | 6327 { |
1602 | 6328 for (i = 0; i < NSUBEXP; ++i) |
1579 | 6329 { |
1602 | 6330 if (REG_MULTI) |
6331 { | |
6332 bp->save_start[i].se_u.pos = reg_startpos[i]; | |
6333 bp->save_end[i].se_u.pos = reg_endpos[i]; | |
6334 } | |
6335 else | |
6336 { | |
6337 bp->save_start[i].se_u.ptr = reg_startp[i]; | |
6338 bp->save_end[i].se_u.ptr = reg_endp[i]; | |
6339 } | |
1579 | 6340 } |
6341 } | |
6342 } | |
6343 | |
6344 /* | |
6345 * Restore the subexpr from "bp". | |
6346 */ | |
6347 static void | |
6348 restore_subexpr(bp) | |
6349 regbehind_T *bp; | |
6350 { | |
6351 int i; | |
6352 | |
1602 | 6353 /* Only need to restore saved values when they are not to be cleared. */ |
6354 need_clear_subexpr = bp->save_need_clear_subexpr; | |
6355 if (!need_clear_subexpr) | |
1579 | 6356 { |
1602 | 6357 for (i = 0; i < NSUBEXP; ++i) |
1579 | 6358 { |
1602 | 6359 if (REG_MULTI) |
6360 { | |
6361 reg_startpos[i] = bp->save_start[i].se_u.pos; | |
6362 reg_endpos[i] = bp->save_end[i].se_u.pos; | |
6363 } | |
6364 else | |
6365 { | |
6366 reg_startp[i] = bp->save_start[i].se_u.ptr; | |
6367 reg_endp[i] = bp->save_end[i].se_u.ptr; | |
6368 } | |
1579 | 6369 } |
6370 } | |
6371 } | |
6372 | |
6373 /* | |
7 | 6374 * Advance reglnum, regline and reginput to the next line. |
6375 */ | |
6376 static void | |
6377 reg_nextline() | |
6378 { | |
6379 regline = reg_getline(++reglnum); | |
6380 reginput = regline; | |
6381 fast_breakcheck(); | |
6382 } | |
6383 | |
6384 /* | |
6385 * Save the input line and position in a regsave_T. | |
6386 */ | |
6387 static void | |
233 | 6388 reg_save(save, gap) |
7 | 6389 regsave_T *save; |
233 | 6390 garray_T *gap; |
7 | 6391 { |
6392 if (REG_MULTI) | |
6393 { | |
6394 save->rs_u.pos.col = (colnr_T)(reginput - regline); | |
6395 save->rs_u.pos.lnum = reglnum; | |
6396 } | |
6397 else | |
6398 save->rs_u.ptr = reginput; | |
233 | 6399 save->rs_len = gap->ga_len; |
7 | 6400 } |
6401 | |
6402 /* | |
6403 * Restore the input line and position from a regsave_T. | |
6404 */ | |
6405 static void | |
233 | 6406 reg_restore(save, gap) |
7 | 6407 regsave_T *save; |
233 | 6408 garray_T *gap; |
7 | 6409 { |
6410 if (REG_MULTI) | |
6411 { | |
6412 if (reglnum != save->rs_u.pos.lnum) | |
6413 { | |
6414 /* only call reg_getline() when the line number changed to save | |
6415 * a bit of time */ | |
6416 reglnum = save->rs_u.pos.lnum; | |
6417 regline = reg_getline(reglnum); | |
6418 } | |
6419 reginput = regline + save->rs_u.pos.col; | |
6420 } | |
6421 else | |
6422 reginput = save->rs_u.ptr; | |
233 | 6423 gap->ga_len = save->rs_len; |
7 | 6424 } |
6425 | |
6426 /* | |
6427 * Return TRUE if current position is equal to saved position. | |
6428 */ | |
6429 static int | |
6430 reg_save_equal(save) | |
6431 regsave_T *save; | |
6432 { | |
6433 if (REG_MULTI) | |
6434 return reglnum == save->rs_u.pos.lnum | |
6435 && reginput == regline + save->rs_u.pos.col; | |
6436 return reginput == save->rs_u.ptr; | |
6437 } | |
6438 | |
6439 /* | |
6440 * Tentatively set the sub-expression start to the current position (after | |
6441 * calling regmatch() they will have changed). Need to save the existing | |
6442 * values for when there is no match. | |
6443 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()), | |
6444 * depending on REG_MULTI. | |
6445 */ | |
6446 static void | |
6447 save_se_multi(savep, posp) | |
6448 save_se_T *savep; | |
6449 lpos_T *posp; | |
6450 { | |
6451 savep->se_u.pos = *posp; | |
6452 posp->lnum = reglnum; | |
6453 posp->col = (colnr_T)(reginput - regline); | |
6454 } | |
6455 | |
6456 static void | |
6457 save_se_one(savep, pp) | |
6458 save_se_T *savep; | |
6459 char_u **pp; | |
6460 { | |
6461 savep->se_u.ptr = *pp; | |
6462 *pp = reginput; | |
6463 } | |
6464 | |
6465 /* | |
6466 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL. | |
6467 */ | |
6468 static int | |
6469 re_num_cmp(val, scan) | |
6470 long_u val; | |
6471 char_u *scan; | |
6472 { | |
6473 long_u n = OPERAND_MIN(scan); | |
6474 | |
6475 if (OPERAND_CMP(scan) == '>') | |
6476 return val > n; | |
6477 if (OPERAND_CMP(scan) == '<') | |
6478 return val < n; | |
6479 return val == n; | |
6480 } | |
6481 | |
6482 | |
4444 | 6483 #ifdef BT_REGEXP_DUMP |
7 | 6484 |
6485 /* | |
6486 * regdump - dump a regexp onto stdout in vaguely comprehensible form | |
6487 */ | |
6488 static void | |
6489 regdump(pattern, r) | |
6490 char_u *pattern; | |
4444 | 6491 bt_regprog_T *r; |
7 | 6492 { |
6493 char_u *s; | |
6494 int op = EXACTLY; /* Arbitrary non-END op. */ | |
6495 char_u *next; | |
6496 char_u *end = NULL; | |
4444 | 6497 FILE *f; |
6498 | |
6499 #ifdef BT_REGEXP_LOG | |
6500 f = fopen("bt_regexp_log.log", "a"); | |
6501 #else | |
6502 f = stdout; | |
6503 #endif | |
6504 if (f == NULL) | |
6505 return; | |
6506 fprintf(f, "-------------------------------------\n\r\nregcomp(%s):\r\n", pattern); | |
7 | 6507 |
6508 s = r->program + 1; | |
6509 /* | |
6510 * Loop until we find the END that isn't before a referred next (an END | |
6511 * can also appear in a NOMATCH operand). | |
6512 */ | |
6513 while (op != END || s <= end) | |
6514 { | |
6515 op = OP(s); | |
4444 | 6516 fprintf(f, "%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */ |
7 | 6517 next = regnext(s); |
6518 if (next == NULL) /* Next ptr. */ | |
4444 | 6519 fprintf(f, "(0)"); |
7 | 6520 else |
4444 | 6521 fprintf(f, "(%d)", (int)((s - r->program) + (next - s))); |
7 | 6522 if (end < next) |
6523 end = next; | |
6524 if (op == BRACE_LIMITS) | |
6525 { | |
6526 /* Two short ints */ | |
4444 | 6527 fprintf(f, " minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s)); |
7 | 6528 s += 8; |
6529 } | |
6530 s += 3; | |
6531 if (op == ANYOF || op == ANYOF + ADD_NL | |
6532 || op == ANYBUT || op == ANYBUT + ADD_NL | |
6533 || op == EXACTLY) | |
6534 { | |
6535 /* Literal string, where present. */ | |
4444 | 6536 fprintf(f, "\nxxxxxxxxx\n"); |
7 | 6537 while (*s != NUL) |
4444 | 6538 fprintf(f, "%c", *s++); |
6539 fprintf(f, "\nxxxxxxxxx\n"); | |
7 | 6540 s++; |
6541 } | |
4444 | 6542 fprintf(f, "\r\n"); |
7 | 6543 } |
6544 | |
6545 /* Header fields of interest. */ | |
6546 if (r->regstart != NUL) | |
4444 | 6547 fprintf(f, "start `%s' 0x%x; ", r->regstart < 256 |
7 | 6548 ? (char *)transchar(r->regstart) |
6549 : "multibyte", r->regstart); | |
6550 if (r->reganch) | |
4444 | 6551 fprintf(f, "anchored; "); |
7 | 6552 if (r->regmust != NULL) |
4444 | 6553 fprintf(f, "must have \"%s\"", r->regmust); |
6554 fprintf(f, "\r\n"); | |
6555 | |
6556 #ifdef BT_REGEXP_LOG | |
6557 fclose(f); | |
6558 #endif | |
7 | 6559 } |
4444 | 6560 #endif /* BT_REGEXP_DUMP */ |
6561 | |
6562 #ifdef DEBUG | |
7 | 6563 /* |
6564 * regprop - printable representation of opcode | |
6565 */ | |
6566 static char_u * | |
6567 regprop(op) | |
6568 char_u *op; | |
6569 { | |
4444 | 6570 char *p; |
6571 static char buf[50]; | |
6572 | |
6573 STRCPY(buf, ":"); | |
6574 | |
6575 switch ((int) OP(op)) | |
7 | 6576 { |
6577 case BOL: | |
6578 p = "BOL"; | |
6579 break; | |
6580 case EOL: | |
6581 p = "EOL"; | |
6582 break; | |
6583 case RE_BOF: | |
6584 p = "BOF"; | |
6585 break; | |
6586 case RE_EOF: | |
6587 p = "EOF"; | |
6588 break; | |
6589 case CURSOR: | |
6590 p = "CURSOR"; | |
6591 break; | |
639 | 6592 case RE_VISUAL: |
6593 p = "RE_VISUAL"; | |
6594 break; | |
7 | 6595 case RE_LNUM: |
6596 p = "RE_LNUM"; | |
6597 break; | |
639 | 6598 case RE_MARK: |
6599 p = "RE_MARK"; | |
6600 break; | |
7 | 6601 case RE_COL: |
6602 p = "RE_COL"; | |
6603 break; | |
6604 case RE_VCOL: | |
6605 p = "RE_VCOL"; | |
6606 break; | |
6607 case BOW: | |
6608 p = "BOW"; | |
6609 break; | |
6610 case EOW: | |
6611 p = "EOW"; | |
6612 break; | |
6613 case ANY: | |
6614 p = "ANY"; | |
6615 break; | |
6616 case ANY + ADD_NL: | |
6617 p = "ANY+NL"; | |
6618 break; | |
6619 case ANYOF: | |
6620 p = "ANYOF"; | |
6621 break; | |
6622 case ANYOF + ADD_NL: | |
6623 p = "ANYOF+NL"; | |
6624 break; | |
6625 case ANYBUT: | |
6626 p = "ANYBUT"; | |
6627 break; | |
6628 case ANYBUT + ADD_NL: | |
6629 p = "ANYBUT+NL"; | |
6630 break; | |
6631 case IDENT: | |
6632 p = "IDENT"; | |
6633 break; | |
6634 case IDENT + ADD_NL: | |
6635 p = "IDENT+NL"; | |
6636 break; | |
6637 case SIDENT: | |
6638 p = "SIDENT"; | |
6639 break; | |
6640 case SIDENT + ADD_NL: | |
6641 p = "SIDENT+NL"; | |
6642 break; | |
6643 case KWORD: | |
6644 p = "KWORD"; | |
6645 break; | |
6646 case KWORD + ADD_NL: | |
6647 p = "KWORD+NL"; | |
6648 break; | |
6649 case SKWORD: | |
6650 p = "SKWORD"; | |
6651 break; | |
6652 case SKWORD + ADD_NL: | |
6653 p = "SKWORD+NL"; | |
6654 break; | |
6655 case FNAME: | |
6656 p = "FNAME"; | |
6657 break; | |
6658 case FNAME + ADD_NL: | |
6659 p = "FNAME+NL"; | |
6660 break; | |
6661 case SFNAME: | |
6662 p = "SFNAME"; | |
6663 break; | |
6664 case SFNAME + ADD_NL: | |
6665 p = "SFNAME+NL"; | |
6666 break; | |
6667 case PRINT: | |
6668 p = "PRINT"; | |
6669 break; | |
6670 case PRINT + ADD_NL: | |
6671 p = "PRINT+NL"; | |
6672 break; | |
6673 case SPRINT: | |
6674 p = "SPRINT"; | |
6675 break; | |
6676 case SPRINT + ADD_NL: | |
6677 p = "SPRINT+NL"; | |
6678 break; | |
6679 case WHITE: | |
6680 p = "WHITE"; | |
6681 break; | |
6682 case WHITE + ADD_NL: | |
6683 p = "WHITE+NL"; | |
6684 break; | |
6685 case NWHITE: | |
6686 p = "NWHITE"; | |
6687 break; | |
6688 case NWHITE + ADD_NL: | |
6689 p = "NWHITE+NL"; | |
6690 break; | |
6691 case DIGIT: | |
6692 p = "DIGIT"; | |
6693 break; | |
6694 case DIGIT + ADD_NL: | |
6695 p = "DIGIT+NL"; | |
6696 break; | |
6697 case NDIGIT: | |
6698 p = "NDIGIT"; | |
6699 break; | |
6700 case NDIGIT + ADD_NL: | |
6701 p = "NDIGIT+NL"; | |
6702 break; | |
6703 case HEX: | |
6704 p = "HEX"; | |
6705 break; | |
6706 case HEX + ADD_NL: | |
6707 p = "HEX+NL"; | |
6708 break; | |
6709 case NHEX: | |
6710 p = "NHEX"; | |
6711 break; | |
6712 case NHEX + ADD_NL: | |
6713 p = "NHEX+NL"; | |
6714 break; | |
6715 case OCTAL: | |
6716 p = "OCTAL"; | |
6717 break; | |
6718 case OCTAL + ADD_NL: | |
6719 p = "OCTAL+NL"; | |
6720 break; | |
6721 case NOCTAL: | |
6722 p = "NOCTAL"; | |
6723 break; | |
6724 case NOCTAL + ADD_NL: | |
6725 p = "NOCTAL+NL"; | |
6726 break; | |
6727 case WORD: | |
6728 p = "WORD"; | |
6729 break; | |
6730 case WORD + ADD_NL: | |
6731 p = "WORD+NL"; | |
6732 break; | |
6733 case NWORD: | |
6734 p = "NWORD"; | |
6735 break; | |
6736 case NWORD + ADD_NL: | |
6737 p = "NWORD+NL"; | |
6738 break; | |
6739 case HEAD: | |
6740 p = "HEAD"; | |
6741 break; | |
6742 case HEAD + ADD_NL: | |
6743 p = "HEAD+NL"; | |
6744 break; | |
6745 case NHEAD: | |
6746 p = "NHEAD"; | |
6747 break; | |
6748 case NHEAD + ADD_NL: | |
6749 p = "NHEAD+NL"; | |
6750 break; | |
6751 case ALPHA: | |
6752 p = "ALPHA"; | |
6753 break; | |
6754 case ALPHA + ADD_NL: | |
6755 p = "ALPHA+NL"; | |
6756 break; | |
6757 case NALPHA: | |
6758 p = "NALPHA"; | |
6759 break; | |
6760 case NALPHA + ADD_NL: | |
6761 p = "NALPHA+NL"; | |
6762 break; | |
6763 case LOWER: | |
6764 p = "LOWER"; | |
6765 break; | |
6766 case LOWER + ADD_NL: | |
6767 p = "LOWER+NL"; | |
6768 break; | |
6769 case NLOWER: | |
6770 p = "NLOWER"; | |
6771 break; | |
6772 case NLOWER + ADD_NL: | |
6773 p = "NLOWER+NL"; | |
6774 break; | |
6775 case UPPER: | |
6776 p = "UPPER"; | |
6777 break; | |
6778 case UPPER + ADD_NL: | |
6779 p = "UPPER+NL"; | |
6780 break; | |
6781 case NUPPER: | |
6782 p = "NUPPER"; | |
6783 break; | |
6784 case NUPPER + ADD_NL: | |
6785 p = "NUPPER+NL"; | |
6786 break; | |
6787 case BRANCH: | |
6788 p = "BRANCH"; | |
6789 break; | |
6790 case EXACTLY: | |
6791 p = "EXACTLY"; | |
6792 break; | |
6793 case NOTHING: | |
6794 p = "NOTHING"; | |
6795 break; | |
6796 case BACK: | |
6797 p = "BACK"; | |
6798 break; | |
6799 case END: | |
6800 p = "END"; | |
6801 break; | |
6802 case MOPEN + 0: | |
6803 p = "MATCH START"; | |
6804 break; | |
6805 case MOPEN + 1: | |
6806 case MOPEN + 2: | |
6807 case MOPEN + 3: | |
6808 case MOPEN + 4: | |
6809 case MOPEN + 5: | |
6810 case MOPEN + 6: | |
6811 case MOPEN + 7: | |
6812 case MOPEN + 8: | |
6813 case MOPEN + 9: | |
6814 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN); | |
6815 p = NULL; | |
6816 break; | |
6817 case MCLOSE + 0: | |
6818 p = "MATCH END"; | |
6819 break; | |
6820 case MCLOSE + 1: | |
6821 case MCLOSE + 2: | |
6822 case MCLOSE + 3: | |
6823 case MCLOSE + 4: | |
6824 case MCLOSE + 5: | |
6825 case MCLOSE + 6: | |
6826 case MCLOSE + 7: | |
6827 case MCLOSE + 8: | |
6828 case MCLOSE + 9: | |
6829 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE); | |
6830 p = NULL; | |
6831 break; | |
6832 case BACKREF + 1: | |
6833 case BACKREF + 2: | |
6834 case BACKREF + 3: | |
6835 case BACKREF + 4: | |
6836 case BACKREF + 5: | |
6837 case BACKREF + 6: | |
6838 case BACKREF + 7: | |
6839 case BACKREF + 8: | |
6840 case BACKREF + 9: | |
6841 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF); | |
6842 p = NULL; | |
6843 break; | |
6844 case NOPEN: | |
6845 p = "NOPEN"; | |
6846 break; | |
6847 case NCLOSE: | |
6848 p = "NCLOSE"; | |
6849 break; | |
6850 #ifdef FEAT_SYN_HL | |
6851 case ZOPEN + 1: | |
6852 case ZOPEN + 2: | |
6853 case ZOPEN + 3: | |
6854 case ZOPEN + 4: | |
6855 case ZOPEN + 5: | |
6856 case ZOPEN + 6: | |
6857 case ZOPEN + 7: | |
6858 case ZOPEN + 8: | |
6859 case ZOPEN + 9: | |
6860 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN); | |
6861 p = NULL; | |
6862 break; | |
6863 case ZCLOSE + 1: | |
6864 case ZCLOSE + 2: | |
6865 case ZCLOSE + 3: | |
6866 case ZCLOSE + 4: | |
6867 case ZCLOSE + 5: | |
6868 case ZCLOSE + 6: | |
6869 case ZCLOSE + 7: | |
6870 case ZCLOSE + 8: | |
6871 case ZCLOSE + 9: | |
6872 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE); | |
6873 p = NULL; | |
6874 break; | |
6875 case ZREF + 1: | |
6876 case ZREF + 2: | |
6877 case ZREF + 3: | |
6878 case ZREF + 4: | |
6879 case ZREF + 5: | |
6880 case ZREF + 6: | |
6881 case ZREF + 7: | |
6882 case ZREF + 8: | |
6883 case ZREF + 9: | |
6884 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF); | |
6885 p = NULL; | |
6886 break; | |
6887 #endif | |
6888 case STAR: | |
6889 p = "STAR"; | |
6890 break; | |
6891 case PLUS: | |
6892 p = "PLUS"; | |
6893 break; | |
6894 case NOMATCH: | |
6895 p = "NOMATCH"; | |
6896 break; | |
6897 case MATCH: | |
6898 p = "MATCH"; | |
6899 break; | |
6900 case BEHIND: | |
6901 p = "BEHIND"; | |
6902 break; | |
6903 case NOBEHIND: | |
6904 p = "NOBEHIND"; | |
6905 break; | |
6906 case SUBPAT: | |
6907 p = "SUBPAT"; | |
6908 break; | |
6909 case BRACE_LIMITS: | |
6910 p = "BRACE_LIMITS"; | |
6911 break; | |
6912 case BRACE_SIMPLE: | |
6913 p = "BRACE_SIMPLE"; | |
6914 break; | |
6915 case BRACE_COMPLEX + 0: | |
6916 case BRACE_COMPLEX + 1: | |
6917 case BRACE_COMPLEX + 2: | |
6918 case BRACE_COMPLEX + 3: | |
6919 case BRACE_COMPLEX + 4: | |
6920 case BRACE_COMPLEX + 5: | |
6921 case BRACE_COMPLEX + 6: | |
6922 case BRACE_COMPLEX + 7: | |
6923 case BRACE_COMPLEX + 8: | |
6924 case BRACE_COMPLEX + 9: | |
6925 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX); | |
6926 p = NULL; | |
6927 break; | |
6928 #ifdef FEAT_MBYTE | |
6929 case MULTIBYTECODE: | |
6930 p = "MULTIBYTECODE"; | |
6931 break; | |
6932 #endif | |
6933 case NEWL: | |
6934 p = "NEWL"; | |
6935 break; | |
6936 default: | |
6937 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op)); | |
6938 p = NULL; | |
6939 break; | |
6940 } | |
6941 if (p != NULL) | |
4444 | 6942 STRCAT(buf, p); |
6943 return (char_u *)buf; | |
7 | 6944 } |
4444 | 6945 #endif /* DEBUG */ |
7 | 6946 |
6947 #ifdef FEAT_MBYTE | |
6948 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3)); | |
6949 | |
6950 typedef struct | |
6951 { | |
6952 int a, b, c; | |
6953 } decomp_T; | |
6954 | |
6955 | |
6956 /* 0xfb20 - 0xfb4f */ | |
297 | 6957 static decomp_T decomp_table[0xfb4f-0xfb20+1] = |
7 | 6958 { |
6959 {0x5e2,0,0}, /* 0xfb20 alt ayin */ | |
6960 {0x5d0,0,0}, /* 0xfb21 alt alef */ | |
6961 {0x5d3,0,0}, /* 0xfb22 alt dalet */ | |
6962 {0x5d4,0,0}, /* 0xfb23 alt he */ | |
6963 {0x5db,0,0}, /* 0xfb24 alt kaf */ | |
6964 {0x5dc,0,0}, /* 0xfb25 alt lamed */ | |
6965 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */ | |
6966 {0x5e8,0,0}, /* 0xfb27 alt resh */ | |
6967 {0x5ea,0,0}, /* 0xfb28 alt tav */ | |
6968 {'+', 0, 0}, /* 0xfb29 alt plus */ | |
6969 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */ | |
6970 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */ | |
6971 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */ | |
6972 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */ | |
6973 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */ | |
6974 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */ | |
6975 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */ | |
6976 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */ | |
6977 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */ | |
6978 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */ | |
6979 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */ | |
6980 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */ | |
6981 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */ | |
6982 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */ | |
6983 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */ | |
6984 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */ | |
6985 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */ | |
6986 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */ | |
6987 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */ | |
6988 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */ | |
6989 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */ | |
6990 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */ | |
6991 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */ | |
6992 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */ | |
6993 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */ | |
6994 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */ | |
6995 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */ | |
6996 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */ | |
6997 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */ | |
6998 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */ | |
6999 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */ | |
7000 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */ | |
7001 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */ | |
7002 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */ | |
7003 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */ | |
7004 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */ | |
7005 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */ | |
7006 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */ | |
7007 }; | |
7008 | |
7009 static void | |
7010 mb_decompose(c, c1, c2, c3) | |
7011 int c, *c1, *c2, *c3; | |
7012 { | |
7013 decomp_T d; | |
7014 | |
4505
d037b9cbdaaa
updated for version 7.3.1000
Bram Moolenaar <bram@vim.org>
parents:
4466
diff
changeset
|
7015 if (c >= 0xfb20 && c <= 0xfb4f) |
7 | 7016 { |
7017 d = decomp_table[c - 0xfb20]; | |
7018 *c1 = d.a; | |
7019 *c2 = d.b; | |
7020 *c3 = d.c; | |
7021 } | |
7022 else | |
7023 { | |
7024 *c1 = c; | |
7025 *c2 = *c3 = 0; | |
7026 } | |
7027 } | |
7028 #endif | |
7029 | |
7030 /* | |
7031 * Compare two strings, ignore case if ireg_ic set. | |
7032 * Return 0 if strings match, non-zero otherwise. | |
7033 * Correct the length "*n" when composing characters are ignored. | |
7034 */ | |
7035 static int | |
7036 cstrncmp(s1, s2, n) | |
7037 char_u *s1, *s2; | |
7038 int *n; | |
7039 { | |
7040 int result; | |
7041 | |
7042 if (!ireg_ic) | |
7043 result = STRNCMP(s1, s2, *n); | |
7044 else | |
7045 result = MB_STRNICMP(s1, s2, *n); | |
7046 | |
7047 #ifdef FEAT_MBYTE | |
7048 /* if it failed and it's utf8 and we want to combineignore: */ | |
7049 if (result != 0 && enc_utf8 && ireg_icombine) | |
7050 { | |
7051 char_u *str1, *str2; | |
7052 int c1, c2, c11, c12; | |
7053 int junk; | |
7054 | |
7055 /* we have to handle the strcmp ourselves, since it is necessary to | |
7056 * deal with the composing characters by ignoring them: */ | |
7057 str1 = s1; | |
7058 str2 = s2; | |
7059 c1 = c2 = 0; | |
507 | 7060 while ((int)(str1 - s1) < *n) |
7 | 7061 { |
7062 c1 = mb_ptr2char_adv(&str1); | |
7063 c2 = mb_ptr2char_adv(&str2); | |
7064 | |
7065 /* decompose the character if necessary, into 'base' characters | |
7066 * because I don't care about Arabic, I will hard-code the Hebrew | |
7067 * which I *do* care about! So sue me... */ | |
7068 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2))) | |
7069 { | |
7070 /* decomposition necessary? */ | |
7071 mb_decompose(c1, &c11, &junk, &junk); | |
7072 mb_decompose(c2, &c12, &junk, &junk); | |
7073 c1 = c11; | |
7074 c2 = c12; | |
7075 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12))) | |
7076 break; | |
7077 } | |
7078 } | |
7079 result = c2 - c1; | |
7080 if (result == 0) | |
7081 *n = (int)(str2 - s2); | |
7082 } | |
7083 #endif | |
7084 | |
7085 return result; | |
7086 } | |
7087 | |
7088 /* | |
7089 * cstrchr: This function is used a lot for simple searches, keep it fast! | |
7090 */ | |
7091 static char_u * | |
7092 cstrchr(s, c) | |
7093 char_u *s; | |
7094 int c; | |
7095 { | |
7096 char_u *p; | |
7097 int cc; | |
7098 | |
7099 if (!ireg_ic | |
7100 #ifdef FEAT_MBYTE | |
7101 || (!enc_utf8 && mb_char2len(c) > 1) | |
7102 #endif | |
7103 ) | |
7104 return vim_strchr(s, c); | |
7105 | |
7106 /* tolower() and toupper() can be slow, comparing twice should be a lot | |
7107 * faster (esp. when using MS Visual C++!). | |
7108 * For UTF-8 need to use folded case. */ | |
7109 #ifdef FEAT_MBYTE | |
7110 if (enc_utf8 && c > 0x80) | |
7111 cc = utf_fold(c); | |
7112 else | |
7113 #endif | |
1347 | 7114 if (MB_ISUPPER(c)) |
7115 cc = MB_TOLOWER(c); | |
7116 else if (MB_ISLOWER(c)) | |
7117 cc = MB_TOUPPER(c); | |
7 | 7118 else |
7119 return vim_strchr(s, c); | |
7120 | |
7121 #ifdef FEAT_MBYTE | |
7122 if (has_mbyte) | |
7123 { | |
474 | 7124 for (p = s; *p != NUL; p += (*mb_ptr2len)(p)) |
7 | 7125 { |
7126 if (enc_utf8 && c > 0x80) | |
7127 { | |
7128 if (utf_fold(utf_ptr2char(p)) == cc) | |
7129 return p; | |
7130 } | |
7131 else if (*p == c || *p == cc) | |
7132 return p; | |
7133 } | |
7134 } | |
7135 else | |
7136 #endif | |
7137 /* Faster version for when there are no multi-byte characters. */ | |
7138 for (p = s; *p != NUL; ++p) | |
7139 if (*p == c || *p == cc) | |
7140 return p; | |
7141 | |
7142 return NULL; | |
7143 } | |
7144 | |
7145 /*************************************************************** | |
7146 * regsub stuff * | |
7147 ***************************************************************/ | |
7148 | |
7149 /* This stuff below really confuses cc on an SGI -- webb */ | |
7150 #ifdef __sgi | |
7151 # undef __ARGS | |
7152 # define __ARGS(x) () | |
7153 #endif | |
7154 | |
7155 /* | |
7156 * We should define ftpr as a pointer to a function returning a pointer to | |
7157 * a function returning a pointer to a function ... | |
7158 * This is impossible, so we declare a pointer to a function returning a | |
7159 * pointer to a function returning void. This should work for all compilers. | |
7160 */ | |
772 | 7161 typedef void (*(*fptr_T) __ARGS((int *, int)))(); |
7162 | |
7163 static fptr_T do_upper __ARGS((int *, int)); | |
7164 static fptr_T do_Upper __ARGS((int *, int)); | |
7165 static fptr_T do_lower __ARGS((int *, int)); | |
7166 static fptr_T do_Lower __ARGS((int *, int)); | |
7 | 7167 |
7168 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash)); | |
7169 | |
772 | 7170 static fptr_T |
7 | 7171 do_upper(d, c) |
772 | 7172 int *d; |
7173 int c; | |
7 | 7174 { |
772 | 7175 *d = MB_TOUPPER(c); |
7176 | |
7177 return (fptr_T)NULL; | |
7 | 7178 } |
7179 | |
772 | 7180 static fptr_T |
7181 do_Upper(d, c) | |
7182 int *d; | |
7 | 7183 int c; |
7184 { | |
772 | 7185 *d = MB_TOUPPER(c); |
7186 | |
7187 return (fptr_T)do_Upper; | |
7188 } | |
7189 | |
7190 static fptr_T | |
7191 do_lower(d, c) | |
7192 int *d; | |
7193 int c; | |
7194 { | |
7195 *d = MB_TOLOWER(c); | |
7196 | |
7197 return (fptr_T)NULL; | |
7198 } | |
7199 | |
7200 static fptr_T | |
7201 do_Lower(d, c) | |
7202 int *d; | |
7203 int c; | |
7204 { | |
7205 *d = MB_TOLOWER(c); | |
7206 | |
7207 return (fptr_T)do_Lower; | |
7 | 7208 } |
7209 | |
7210 /* | |
7211 * regtilde(): Replace tildes in the pattern by the old pattern. | |
7212 * | |
7213 * Short explanation of the tilde: It stands for the previous replacement | |
7214 * pattern. If that previous pattern also contains a ~ we should go back a | |
7215 * step further... But we insert the previous pattern into the current one | |
7216 * and remember that. | |
772 | 7217 * This still does not handle the case where "magic" changes. So require the |
7218 * user to keep his hands off of "magic". | |
7 | 7219 * |
7220 * The tildes are parsed once before the first call to vim_regsub(). | |
7221 */ | |
7222 char_u * | |
7223 regtilde(source, magic) | |
7224 char_u *source; | |
7225 int magic; | |
7226 { | |
7227 char_u *newsub = source; | |
7228 char_u *tmpsub; | |
7229 char_u *p; | |
7230 int len; | |
7231 int prevlen; | |
7232 | |
7233 for (p = newsub; *p; ++p) | |
7234 { | |
7235 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic)) | |
7236 { | |
7237 if (reg_prev_sub != NULL) | |
7238 { | |
7239 /* length = len(newsub) - 1 + len(prev_sub) + 1 */ | |
7240 prevlen = (int)STRLEN(reg_prev_sub); | |
7241 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen)); | |
7242 if (tmpsub != NULL) | |
7243 { | |
7244 /* copy prefix */ | |
7245 len = (int)(p - newsub); /* not including ~ */ | |
7246 mch_memmove(tmpsub, newsub, (size_t)len); | |
1209 | 7247 /* interpret tilde */ |
7 | 7248 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen); |
7249 /* copy postfix */ | |
7250 if (!magic) | |
7251 ++p; /* back off \ */ | |
7252 STRCPY(tmpsub + len + prevlen, p + 1); | |
7253 | |
7254 if (newsub != source) /* already allocated newsub */ | |
7255 vim_free(newsub); | |
7256 newsub = tmpsub; | |
7257 p = newsub + len + prevlen; | |
7258 } | |
7259 } | |
7260 else if (magic) | |
1621 | 7261 STRMOVE(p, p + 1); /* remove '~' */ |
7 | 7262 else |
1621 | 7263 STRMOVE(p, p + 2); /* remove '\~' */ |
7 | 7264 --p; |
7265 } | |
7266 else | |
7267 { | |
7268 if (*p == '\\' && p[1]) /* skip escaped characters */ | |
7269 ++p; | |
7270 #ifdef FEAT_MBYTE | |
7271 if (has_mbyte) | |
474 | 7272 p += (*mb_ptr2len)(p) - 1; |
7 | 7273 #endif |
7274 } | |
7275 } | |
7276 | |
7277 vim_free(reg_prev_sub); | |
7278 if (newsub != source) /* newsub was allocated, just keep it */ | |
7279 reg_prev_sub = newsub; | |
7280 else /* no ~ found, need to save newsub */ | |
7281 reg_prev_sub = vim_strsave(newsub); | |
7282 return newsub; | |
7283 } | |
7284 | |
7285 #ifdef FEAT_EVAL | |
7286 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */ | |
7287 | |
7288 /* These pointers are used instead of reg_match and reg_mmatch for | |
7289 * reg_submatch(). Needed for when the substitution string is an expression | |
7290 * that contains a call to substitute() and submatch(). */ | |
7291 static regmatch_T *submatch_match; | |
7292 static regmmatch_T *submatch_mmatch; | |
2011 | 7293 static linenr_T submatch_firstlnum; |
7294 static linenr_T submatch_maxline; | |
2904 | 7295 static int submatch_line_lbr; |
7 | 7296 #endif |
7297 | |
7298 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO) | |
7299 /* | |
7300 * vim_regsub() - perform substitutions after a vim_regexec() or | |
7301 * vim_regexec_multi() match. | |
7302 * | |
7303 * If "copy" is TRUE really copy into "dest". | |
7304 * If "copy" is FALSE nothing is copied, this is just to find out the length | |
7305 * of the result. | |
7306 * | |
7307 * If "backslash" is TRUE, a backslash will be removed later, need to double | |
7308 * them to keep them, and insert a backslash before a CR to avoid it being | |
7309 * replaced with a line break later. | |
7310 * | |
7311 * Note: The matched text must not change between the call of | |
7312 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back | |
7313 * references invalid! | |
7314 * | |
7315 * Returns the size of the replacement, including terminating NUL. | |
7316 */ | |
7317 int | |
7318 vim_regsub(rmp, source, dest, copy, magic, backslash) | |
7319 regmatch_T *rmp; | |
7320 char_u *source; | |
7321 char_u *dest; | |
7322 int copy; | |
7323 int magic; | |
7324 int backslash; | |
7325 { | |
7326 reg_match = rmp; | |
7327 reg_mmatch = NULL; | |
7328 reg_maxline = 0; | |
4061 | 7329 reg_buf = curbuf; |
7 | 7330 return vim_regsub_both(source, dest, copy, magic, backslash); |
7331 } | |
7332 #endif | |
7333 | |
7334 int | |
7335 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash) | |
7336 regmmatch_T *rmp; | |
7337 linenr_T lnum; | |
7338 char_u *source; | |
7339 char_u *dest; | |
7340 int copy; | |
7341 int magic; | |
7342 int backslash; | |
7343 { | |
7344 reg_match = NULL; | |
7345 reg_mmatch = rmp; | |
7346 reg_buf = curbuf; /* always works on the current buffer! */ | |
7347 reg_firstlnum = lnum; | |
7348 reg_maxline = curbuf->b_ml.ml_line_count - lnum; | |
7349 return vim_regsub_both(source, dest, copy, magic, backslash); | |
7350 } | |
7351 | |
7352 static int | |
7353 vim_regsub_both(source, dest, copy, magic, backslash) | |
7354 char_u *source; | |
7355 char_u *dest; | |
7356 int copy; | |
7357 int magic; | |
7358 int backslash; | |
7359 { | |
7360 char_u *src; | |
7361 char_u *dst; | |
7362 char_u *s; | |
7363 int c; | |
772 | 7364 int cc; |
7 | 7365 int no = -1; |
4244 | 7366 fptr_T func_all = (fptr_T)NULL; |
7367 fptr_T func_one = (fptr_T)NULL; | |
7 | 7368 linenr_T clnum = 0; /* init for GCC */ |
7369 int len = 0; /* init for GCC */ | |
7370 #ifdef FEAT_EVAL | |
7371 static char_u *eval_result = NULL; | |
7372 #endif | |
7373 | |
7374 /* Be paranoid... */ | |
7375 if (source == NULL || dest == NULL) | |
7376 { | |
7377 EMSG(_(e_null)); | |
7378 return 0; | |
7379 } | |
7380 if (prog_magic_wrong()) | |
7381 return 0; | |
7382 src = source; | |
7383 dst = dest; | |
7384 | |
7385 /* | |
7386 * When the substitute part starts with "\=" evaluate it as an expression. | |
7387 */ | |
7388 if (source[0] == '\\' && source[1] == '=' | |
7389 #ifdef FEAT_EVAL | |
7390 && !can_f_submatch /* can't do this recursively */ | |
7391 #endif | |
7392 ) | |
7393 { | |
7394 #ifdef FEAT_EVAL | |
7395 /* To make sure that the length doesn't change between checking the | |
7396 * length and copying the string, and to speed up things, the | |
7397 * resulting string is saved from the call with "copy" == FALSE to the | |
7398 * call with "copy" == TRUE. */ | |
7399 if (copy) | |
7400 { | |
7401 if (eval_result != NULL) | |
7402 { | |
7403 STRCPY(dest, eval_result); | |
7404 dst += STRLEN(eval_result); | |
7405 vim_free(eval_result); | |
7406 eval_result = NULL; | |
7407 } | |
7408 } | |
7409 else | |
7410 { | |
7411 win_T *save_reg_win; | |
7412 int save_ireg_ic; | |
7413 | |
7414 vim_free(eval_result); | |
7415 | |
7416 /* The expression may contain substitute(), which calls us | |
7417 * recursively. Make sure submatch() gets the text from the first | |
7418 * level. Don't need to save "reg_buf", because | |
7419 * vim_regexec_multi() can't be called recursively. */ | |
7420 submatch_match = reg_match; | |
7421 submatch_mmatch = reg_mmatch; | |
2011 | 7422 submatch_firstlnum = reg_firstlnum; |
7423 submatch_maxline = reg_maxline; | |
2904 | 7424 submatch_line_lbr = reg_line_lbr; |
7 | 7425 save_reg_win = reg_win; |
7426 save_ireg_ic = ireg_ic; | |
7427 can_f_submatch = TRUE; | |
7428 | |
714 | 7429 eval_result = eval_to_string(source + 2, NULL, TRUE); |
7 | 7430 if (eval_result != NULL) |
7431 { | |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7432 int had_backslash = FALSE; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7433 |
39 | 7434 for (s = eval_result; *s != NUL; mb_ptr_adv(s)) |
7 | 7435 { |
2904 | 7436 /* Change NL to CR, so that it becomes a line break, |
7437 * unless called from vim_regexec_nl(). | |
7 | 7438 * Skip over a backslashed character. */ |
2904 | 7439 if (*s == NL && !submatch_line_lbr) |
7 | 7440 *s = CAR; |
7441 else if (*s == '\\' && s[1] != NUL) | |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7442 { |
7 | 7443 ++s; |
2173 | 7444 /* Change NL to CR here too, so that this works: |
7445 * :s/abc\\\ndef/\="aaa\\\nbbb"/ on text: | |
7446 * abc\ | |
7447 * def | |
2904 | 7448 * Not when called from vim_regexec_nl(). |
2173 | 7449 */ |
2904 | 7450 if (*s == NL && !submatch_line_lbr) |
2173 | 7451 *s = CAR; |
2125
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7452 had_backslash = TRUE; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7453 } |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7454 } |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7455 if (had_backslash && backslash) |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7456 { |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7457 /* Backslashes will be consumed, need to double them. */ |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7458 s = vim_strsave_escaped(eval_result, (char_u *)"\\"); |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7459 if (s != NULL) |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7460 { |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7461 vim_free(eval_result); |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7462 eval_result = s; |
b8744d1982d1
updated for version 7.2.407
Bram Moolenaar <bram@zimbu.org>
parents:
2012
diff
changeset
|
7463 } |
7 | 7464 } |
7465 | |
7466 dst += STRLEN(eval_result); | |
7467 } | |
7468 | |
7469 reg_match = submatch_match; | |
7470 reg_mmatch = submatch_mmatch; | |
2011 | 7471 reg_firstlnum = submatch_firstlnum; |
7472 reg_maxline = submatch_maxline; | |
2904 | 7473 reg_line_lbr = submatch_line_lbr; |
7 | 7474 reg_win = save_reg_win; |
7475 ireg_ic = save_ireg_ic; | |
7476 can_f_submatch = FALSE; | |
7477 } | |
7478 #endif | |
7479 } | |
7480 else | |
7481 while ((c = *src++) != NUL) | |
7482 { | |
7483 if (c == '&' && magic) | |
7484 no = 0; | |
7485 else if (c == '\\' && *src != NUL) | |
7486 { | |
7487 if (*src == '&' && !magic) | |
7488 { | |
7489 ++src; | |
7490 no = 0; | |
7491 } | |
7492 else if ('0' <= *src && *src <= '9') | |
7493 { | |
7494 no = *src++ - '0'; | |
7495 } | |
7496 else if (vim_strchr((char_u *)"uUlLeE", *src)) | |
7497 { | |
7498 switch (*src++) | |
7499 { | |
4244 | 7500 case 'u': func_one = (fptr_T)do_upper; |
7 | 7501 continue; |
4244 | 7502 case 'U': func_all = (fptr_T)do_Upper; |
7 | 7503 continue; |
4244 | 7504 case 'l': func_one = (fptr_T)do_lower; |
7 | 7505 continue; |
4244 | 7506 case 'L': func_all = (fptr_T)do_Lower; |
7 | 7507 continue; |
7508 case 'e': | |
4244 | 7509 case 'E': func_one = func_all = (fptr_T)NULL; |
7 | 7510 continue; |
7511 } | |
7512 } | |
7513 } | |
7514 if (no < 0) /* Ordinary character. */ | |
7515 { | |
798 | 7516 if (c == K_SPECIAL && src[0] != NUL && src[1] != NUL) |
7517 { | |
1209 | 7518 /* Copy a special key as-is. */ |
798 | 7519 if (copy) |
7520 { | |
7521 *dst++ = c; | |
7522 *dst++ = *src++; | |
7523 *dst++ = *src++; | |
7524 } | |
7525 else | |
7526 { | |
7527 dst += 3; | |
7528 src += 2; | |
7529 } | |
7530 continue; | |
7531 } | |
7532 | |
7 | 7533 if (c == '\\' && *src != NUL) |
7534 { | |
7535 /* Check for abbreviations -- webb */ | |
7536 switch (*src) | |
7537 { | |
7538 case 'r': c = CAR; ++src; break; | |
7539 case 'n': c = NL; ++src; break; | |
7540 case 't': c = TAB; ++src; break; | |
7541 /* Oh no! \e already has meaning in subst pat :-( */ | |
7542 /* case 'e': c = ESC; ++src; break; */ | |
7543 case 'b': c = Ctrl_H; ++src; break; | |
7544 | |
7545 /* If "backslash" is TRUE the backslash will be removed | |
7546 * later. Used to insert a literal CR. */ | |
7547 default: if (backslash) | |
7548 { | |
7549 if (copy) | |
7550 *dst = '\\'; | |
7551 ++dst; | |
7552 } | |
7553 c = *src++; | |
7554 } | |
7555 } | |
798 | 7556 #ifdef FEAT_MBYTE |
7557 else if (has_mbyte) | |
7558 c = mb_ptr2char(src - 1); | |
7559 #endif | |
7 | 7560 |
7561 /* Write to buffer, if copy is set. */ | |
4244 | 7562 if (func_one != (fptr_T)NULL) |
7563 /* Turbo C complains without the typecast */ | |
7564 func_one = (fptr_T)(func_one(&cc, c)); | |
7565 else if (func_all != (fptr_T)NULL) | |
7566 /* Turbo C complains without the typecast */ | |
7567 func_all = (fptr_T)(func_all(&cc, c)); | |
7568 else /* just copy */ | |
772 | 7569 cc = c; |
7570 | |
7571 #ifdef FEAT_MBYTE | |
7572 if (has_mbyte) | |
7 | 7573 { |
2307
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7574 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
|
7575 |
7 | 7576 if (copy) |
772 | 7577 mb_char2bytes(cc, dst); |
7578 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
|
7579 if (enc_utf8) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7580 { |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7581 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
|
7582 |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7583 /* 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
|
7584 * 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
|
7585 if (clen < totlen) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7586 { |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7587 if (copy) |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7588 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
|
7589 (size_t)(totlen - clen)); |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7590 dst += totlen - clen; |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7591 } |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7592 } |
81527f127fb1
Fix: Composing characters in :s substitute text were dropped.
Bram Moolenaar <bram@vim.org>
parents:
2247
diff
changeset
|
7593 src += totlen - 1; |
7 | 7594 } |
7595 else | |
7596 #endif | |
7597 if (copy) | |
772 | 7598 *dst = cc; |
7 | 7599 dst++; |
7600 } | |
7601 else | |
7602 { | |
7603 if (REG_MULTI) | |
7604 { | |
7605 clnum = reg_mmatch->startpos[no].lnum; | |
7606 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0) | |
7607 s = NULL; | |
7608 else | |
7609 { | |
7610 s = reg_getline(clnum) + reg_mmatch->startpos[no].col; | |
7611 if (reg_mmatch->endpos[no].lnum == clnum) | |
7612 len = reg_mmatch->endpos[no].col | |
7613 - reg_mmatch->startpos[no].col; | |
7614 else | |
7615 len = (int)STRLEN(s); | |
7616 } | |
7617 } | |
7618 else | |
7619 { | |
7620 s = reg_match->startp[no]; | |
7621 if (reg_match->endp[no] == NULL) | |
7622 s = NULL; | |
7623 else | |
7624 len = (int)(reg_match->endp[no] - s); | |
7625 } | |
7626 if (s != NULL) | |
7627 { | |
7628 for (;;) | |
7629 { | |
7630 if (len == 0) | |
7631 { | |
7632 if (REG_MULTI) | |
7633 { | |
7634 if (reg_mmatch->endpos[no].lnum == clnum) | |
7635 break; | |
7636 if (copy) | |
7637 *dst = CAR; | |
7638 ++dst; | |
7639 s = reg_getline(++clnum); | |
7640 if (reg_mmatch->endpos[no].lnum == clnum) | |
7641 len = reg_mmatch->endpos[no].col; | |
7642 else | |
7643 len = (int)STRLEN(s); | |
7644 } | |
7645 else | |
7646 break; | |
7647 } | |
7648 else if (*s == NUL) /* we hit NUL. */ | |
7649 { | |
7650 if (copy) | |
7651 EMSG(_(e_re_damg)); | |
7652 goto exit; | |
7653 } | |
7654 else | |
7655 { | |
7656 if (backslash && (*s == CAR || *s == '\\')) | |
7657 { | |
7658 /* | |
7659 * Insert a backslash in front of a CR, otherwise | |
7660 * it will be replaced by a line break. | |
7661 * Number of backslashes will be halved later, | |
7662 * double them here. | |
7663 */ | |
7664 if (copy) | |
7665 { | |
7666 dst[0] = '\\'; | |
7667 dst[1] = *s; | |
7668 } | |
7669 dst += 2; | |
7670 } | |
7671 else | |
7672 { | |
772 | 7673 #ifdef FEAT_MBYTE |
7674 if (has_mbyte) | |
7675 c = mb_ptr2char(s); | |
7676 else | |
7677 #endif | |
7678 c = *s; | |
7679 | |
4244 | 7680 if (func_one != (fptr_T)NULL) |
7681 /* Turbo C complains without the typecast */ | |
7682 func_one = (fptr_T)(func_one(&cc, c)); | |
7683 else if (func_all != (fptr_T)NULL) | |
7684 /* Turbo C complains without the typecast */ | |
7685 func_all = (fptr_T)(func_all(&cc, c)); | |
7686 else /* just copy */ | |
772 | 7687 cc = c; |
7688 | |
7689 #ifdef FEAT_MBYTE | |
7690 if (has_mbyte) | |
7 | 7691 { |
1332 | 7692 int l; |
7693 | |
7694 /* Copy composing characters separately, one | |
7695 * at a time. */ | |
7696 if (enc_utf8) | |
7697 l = utf_ptr2len(s) - 1; | |
7698 else | |
7699 l = mb_ptr2len(s) - 1; | |
772 | 7700 |
7701 s += l; | |
7702 len -= l; | |
7703 if (copy) | |
7704 mb_char2bytes(cc, dst); | |
7705 dst += mb_char2len(cc) - 1; | |
7 | 7706 } |
772 | 7707 else |
7708 #endif | |
7709 if (copy) | |
7710 *dst = cc; | |
7711 dst++; | |
7 | 7712 } |
772 | 7713 |
7 | 7714 ++s; |
7715 --len; | |
7716 } | |
7717 } | |
7718 } | |
7719 no = -1; | |
7720 } | |
7721 } | |
7722 if (copy) | |
7723 *dst = NUL; | |
7724 | |
7725 exit: | |
7726 return (int)((dst - dest) + 1); | |
7727 } | |
7728 | |
7729 #ifdef FEAT_EVAL | |
2012 | 7730 static char_u *reg_getline_submatch __ARGS((linenr_T lnum)); |
7731 | |
7 | 7732 /* |
2011 | 7733 * Call reg_getline() with the line numbers from the submatch. If a |
7734 * substitute() was used the reg_maxline and other values have been | |
7735 * overwritten. | |
7736 */ | |
7737 static char_u * | |
7738 reg_getline_submatch(lnum) | |
7739 linenr_T lnum; | |
7740 { | |
7741 char_u *s; | |
7742 linenr_T save_first = reg_firstlnum; | |
7743 linenr_T save_max = reg_maxline; | |
7744 | |
7745 reg_firstlnum = submatch_firstlnum; | |
7746 reg_maxline = submatch_maxline; | |
7747 | |
7748 s = reg_getline(lnum); | |
7749 | |
7750 reg_firstlnum = save_first; | |
7751 reg_maxline = save_max; | |
7752 return s; | |
7753 } | |
7754 | |
7755 /* | |
1209 | 7756 * Used for the submatch() function: get the string from the n'th submatch in |
7 | 7757 * allocated memory. |
7758 * Returns NULL when not in a ":s" command and for a non-existing submatch. | |
7759 */ | |
7760 char_u * | |
7761 reg_submatch(no) | |
7762 int no; | |
7763 { | |
7764 char_u *retval = NULL; | |
7765 char_u *s; | |
7766 int len; | |
7767 int round; | |
7768 linenr_T lnum; | |
7769 | |
840 | 7770 if (!can_f_submatch || no < 0) |
7 | 7771 return NULL; |
7772 | |
7773 if (submatch_match == NULL) | |
7774 { | |
7775 /* | |
7776 * First round: compute the length and allocate memory. | |
7777 * Second round: copy the text. | |
7778 */ | |
7779 for (round = 1; round <= 2; ++round) | |
7780 { | |
7781 lnum = submatch_mmatch->startpos[no].lnum; | |
7782 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0) | |
7783 return NULL; | |
7784 | |
2011 | 7785 s = reg_getline_submatch(lnum) + submatch_mmatch->startpos[no].col; |
7 | 7786 if (s == NULL) /* anti-crash check, cannot happen? */ |
7787 break; | |
7788 if (submatch_mmatch->endpos[no].lnum == lnum) | |
7789 { | |
7790 /* Within one line: take form start to end col. */ | |
7791 len = submatch_mmatch->endpos[no].col | |
7792 - submatch_mmatch->startpos[no].col; | |
7793 if (round == 2) | |
418 | 7794 vim_strncpy(retval, s, len); |
7 | 7795 ++len; |
7796 } | |
7797 else | |
7798 { | |
7799 /* Multiple lines: take start line from start col, middle | |
7800 * lines completely and end line up to end col. */ | |
7801 len = (int)STRLEN(s); | |
7802 if (round == 2) | |
7803 { | |
7804 STRCPY(retval, s); | |
7805 retval[len] = '\n'; | |
7806 } | |
7807 ++len; | |
7808 ++lnum; | |
7809 while (lnum < submatch_mmatch->endpos[no].lnum) | |
7810 { | |
2011 | 7811 s = reg_getline_submatch(lnum++); |
7 | 7812 if (round == 2) |
7813 STRCPY(retval + len, s); | |
7814 len += (int)STRLEN(s); | |
7815 if (round == 2) | |
7816 retval[len] = '\n'; | |
7817 ++len; | |
7818 } | |
7819 if (round == 2) | |
2011 | 7820 STRNCPY(retval + len, reg_getline_submatch(lnum), |
7 | 7821 submatch_mmatch->endpos[no].col); |
7822 len += submatch_mmatch->endpos[no].col; | |
7823 if (round == 2) | |
7824 retval[len] = NUL; | |
7825 ++len; | |
7826 } | |
7827 | |
840 | 7828 if (retval == NULL) |
7 | 7829 { |
7830 retval = lalloc((long_u)len, TRUE); | |
840 | 7831 if (retval == NULL) |
7 | 7832 return NULL; |
7833 } | |
7834 } | |
7835 } | |
7836 else | |
7837 { | |
1815 | 7838 s = submatch_match->startp[no]; |
7839 if (s == NULL || submatch_match->endp[no] == NULL) | |
7 | 7840 retval = NULL; |
7841 else | |
7842 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s)); | |
7843 } | |
7844 | |
7845 return retval; | |
7846 } | |
7847 #endif | |
4444 | 7848 |
7849 static regengine_T bt_regengine = | |
7850 { | |
7851 bt_regcomp, | |
7852 bt_regexec, | |
7853 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7854 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7855 bt_regexec_nl, | |
7856 #endif | |
7857 bt_regexec_multi | |
7858 #ifdef DEBUG | |
7859 ,(char_u *)"" | |
7860 #endif | |
7861 }; | |
7862 | |
7863 | |
7864 #include "regexp_nfa.c" | |
7865 | |
7866 static regengine_T nfa_regengine = | |
7867 { | |
7868 nfa_regcomp, | |
7869 nfa_regexec, | |
7870 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
7871 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
7872 nfa_regexec_nl, | |
7873 #endif | |
7874 nfa_regexec_multi | |
7875 #ifdef DEBUG | |
7876 ,(char_u *)"" | |
7877 #endif | |
7878 }; | |
7879 | |
7880 /* Which regexp engine to use? Needed for vim_regcomp(). | |
7881 * Must match with 'regexpengine'. */ | |
7882 static int regexp_engine = 0; | |
7883 #define AUTOMATIC_ENGINE 0 | |
7884 #define BACKTRACKING_ENGINE 1 | |
7885 #define NFA_ENGINE 2 | |
7886 #ifdef DEBUG | |
7887 static char_u regname[][30] = { | |
7888 "AUTOMATIC Regexp Engine", | |
4579
7a2be4a39423
updated for version 7.3.1037
Bram Moolenaar <bram@vim.org>
parents:
4505
diff
changeset
|
7889 "BACKTRACKING Regexp Engine", |
4444 | 7890 "NFA Regexp Engine" |
7891 }; | |
7892 #endif | |
7893 | |
7894 /* | |
7895 * Compile a regular expression into internal code. | |
7896 * Returns the program in allocated memory. Returns NULL for an error. | |
7897 */ | |
7898 regprog_T * | |
7899 vim_regcomp(expr_arg, re_flags) | |
7900 char_u *expr_arg; | |
7901 int re_flags; | |
7902 { | |
7903 regprog_T *prog = NULL; | |
7904 char_u *expr = expr_arg; | |
7905 | |
7906 syntax_error = FALSE; | |
7907 regexp_engine = p_re; | |
7908 | |
7909 /* Check for prefix "\%#=", that sets the regexp engine */ | |
7910 if (STRNCMP(expr, "\\%#=", 4) == 0) | |
7911 { | |
7912 int newengine = expr[4] - '0'; | |
7913 | |
7914 if (newengine == AUTOMATIC_ENGINE | |
7915 || newengine == BACKTRACKING_ENGINE | |
7916 || newengine == NFA_ENGINE) | |
7917 { | |
7918 regexp_engine = expr[4] - '0'; | |
7919 expr += 5; | |
7920 #ifdef DEBUG | |
7921 EMSG3("New regexp mode selected (%d): %s", regexp_engine, | |
7922 regname[newengine]); | |
7923 #endif | |
7924 } | |
7925 else | |
7926 { | |
7927 EMSG(_("E864: \\%#= can only be followed by 0, 1, or 2. The automatic engine will be used ")); | |
7928 regexp_engine = AUTOMATIC_ENGINE; | |
7929 } | |
7930 } | |
7931 #ifdef DEBUG | |
7932 bt_regengine.expr = expr; | |
7933 nfa_regengine.expr = expr; | |
7934 #endif | |
7935 | |
7936 /* | |
7937 * First try the NFA engine, unless backtracking was requested. | |
7938 */ | |
7939 if (regexp_engine != BACKTRACKING_ENGINE) | |
7940 prog = nfa_regengine.regcomp(expr, re_flags); | |
7941 else | |
7942 prog = bt_regengine.regcomp(expr, re_flags); | |
7943 | |
7944 if (prog == NULL) /* error compiling regexp with initial engine */ | |
7945 { | |
4460 | 7946 #ifdef BT_REGEXP_DEBUG_LOG |
4444 | 7947 if (regexp_engine != BACKTRACKING_ENGINE) /* debugging log for NFA */ |
7948 { | |
7949 FILE *f; | |
4460 | 7950 f = fopen(BT_REGEXP_DEBUG_LOG_NAME, "a"); |
4444 | 7951 if (f) |
7952 { | |
7953 if (!syntax_error) | |
7954 fprintf(f, "NFA engine could not handle \"%s\"\n", expr); | |
7955 else | |
7956 fprintf(f, "Syntax error in \"%s\"\n", expr); | |
7957 fclose(f); | |
7958 } | |
7959 else | |
4460 | 7960 EMSG2("(NFA) Could not open \"%s\" to write !!!", |
7961 BT_REGEXP_DEBUG_LOG_NAME); | |
4444 | 7962 /* |
7963 if (syntax_error) | |
7964 EMSG("NFA Regexp: Syntax Error !"); | |
7965 */ | |
7966 } | |
7967 #endif | |
7968 /* | |
7969 * If NFA engine failed, then revert to the backtracking engine. | |
7970 * Except when there was a syntax error, which was properly handled by | |
7971 * NFA engine. | |
7972 */ | |
7973 if (regexp_engine == AUTOMATIC_ENGINE) | |
7974 if (!syntax_error) | |
7975 prog = bt_regengine.regcomp(expr, re_flags); | |
7976 | |
7977 } /* endif prog==NULL */ | |
7978 | |
7979 | |
7980 return prog; | |
7981 } | |
7982 | |
7983 /* | |
7984 * Match a regexp against a string. | |
7985 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
7986 * Uses curbuf for line count and 'iskeyword'. | |
7987 * | |
7988 * Return TRUE if there is a match, FALSE if not. | |
7989 */ | |
7990 int | |
7991 vim_regexec(rmp, line, col) | |
7992 regmatch_T *rmp; | |
7993 char_u *line; /* string to match against */ | |
7994 colnr_T col; /* column to start looking for match */ | |
7995 { | |
7996 return rmp->regprog->engine->regexec(rmp, line, col); | |
7997 } | |
7998 | |
7999 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \ | |
8000 || defined(FIND_REPLACE_DIALOG) || defined(PROTO) | |
8001 /* | |
8002 * Like vim_regexec(), but consider a "\n" in "line" to be a line break. | |
8003 */ | |
8004 int | |
8005 vim_regexec_nl(rmp, line, col) | |
8006 regmatch_T *rmp; | |
8007 char_u *line; | |
8008 colnr_T col; | |
8009 { | |
8010 return rmp->regprog->engine->regexec_nl(rmp, line, col); | |
8011 } | |
8012 #endif | |
8013 | |
8014 /* | |
8015 * Match a regexp against multiple lines. | |
8016 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp(). | |
8017 * Uses curbuf for line count and 'iskeyword'. | |
8018 * | |
8019 * Return zero if there is no match. Return number of lines contained in the | |
8020 * match otherwise. | |
8021 */ | |
8022 long | |
8023 vim_regexec_multi(rmp, win, buf, lnum, col, tm) | |
8024 regmmatch_T *rmp; | |
8025 win_T *win; /* window in which to search or NULL */ | |
8026 buf_T *buf; /* buffer in which to search */ | |
8027 linenr_T lnum; /* nr of line to start looking for match */ | |
8028 colnr_T col; /* column to start looking for match */ | |
8029 proftime_T *tm; /* timeout limit or NULL */ | |
8030 { | |
8031 return rmp->regprog->engine->regexec_multi(rmp, win, buf, lnum, col, tm); | |
8032 } |