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
|
|
41 #include "vim.h"
|
|
42
|
|
43 #undef DEBUG
|
|
44
|
|
45 /*
|
|
46 * The "internal use only" fields in regexp.h are present to pass info from
|
|
47 * compile to execute that permits the execute phase to run lots faster on
|
|
48 * simple cases. They are:
|
|
49 *
|
|
50 * regstart char that must begin a match; NUL if none obvious; Can be a
|
|
51 * multi-byte character.
|
|
52 * reganch is the match anchored (at beginning-of-line only)?
|
|
53 * regmust string (pointer into program) that match must include, or NULL
|
|
54 * regmlen length of regmust string
|
|
55 * regflags RF_ values or'ed together
|
|
56 *
|
|
57 * Regstart and reganch permit very fast decisions on suitable starting points
|
|
58 * for a match, cutting down the work a lot. Regmust permits fast rejection
|
|
59 * of lines that cannot possibly match. The regmust tests are costly enough
|
|
60 * that vim_regcomp() supplies a regmust only if the r.e. contains something
|
|
61 * potentially expensive (at present, the only such thing detected is * or +
|
|
62 * at the start of the r.e., which can involve a lot of backup). Regmlen is
|
|
63 * supplied because the test in vim_regexec() needs it and vim_regcomp() is
|
|
64 * computing it anyway.
|
|
65 */
|
|
66
|
|
67 /*
|
|
68 * Structure for regexp "program". This is essentially a linear encoding
|
|
69 * of a nondeterministic finite-state machine (aka syntax charts or
|
|
70 * "railroad normal form" in parsing technology). Each node is an opcode
|
|
71 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
|
|
72 * all nodes except BRANCH and BRACES_COMPLEX implement concatenation; a "next"
|
|
73 * pointer with a BRANCH on both ends of it is connecting two alternatives.
|
|
74 * (Here we have one of the subtle syntax dependencies: an individual BRANCH
|
|
75 * (as opposed to a collection of them) is never concatenated with anything
|
|
76 * because of operator precedence). The "next" pointer of a BRACES_COMPLEX
|
|
77 * node points to the node after the stuff to be repeated. The operand of some
|
|
78 * types of node is a literal string; for others, it is a node leading into a
|
|
79 * sub-FSM. In particular, the operand of a BRANCH node is the first node of
|
|
80 * the branch. (NB this is *not* a tree structure: the tail of the branch
|
|
81 * connects to the thing following the set of BRANCHes.)
|
|
82 *
|
|
83 * pattern is coded like:
|
|
84 *
|
|
85 * +-----------------+
|
|
86 * | V
|
|
87 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END
|
|
88 * | ^ | ^
|
|
89 * +------+ +----------+
|
|
90 *
|
|
91 *
|
|
92 * +------------------+
|
|
93 * V |
|
|
94 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END
|
|
95 * | | ^ ^
|
|
96 * | +---------------+ |
|
|
97 * +---------------------------------------------+
|
|
98 *
|
|
99 *
|
|
100 * +-------------------------+
|
|
101 * V |
|
|
102 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END
|
|
103 * | | ^
|
|
104 * | +----------------+
|
|
105 * +-----------------------------------------------+
|
|
106 *
|
|
107 *
|
|
108 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END
|
|
109 * | | ^ ^
|
|
110 * | +----------------+ |
|
|
111 * +--------------------------------+
|
|
112 *
|
|
113 * +---------+
|
|
114 * | V
|
|
115 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END
|
|
116 * | | | | ^ ^
|
|
117 * | | | +-----+ |
|
|
118 * | | +----------------+ |
|
|
119 * | +---------------------------+ |
|
|
120 * +------------------------------------------------------+
|
|
121 *
|
|
122 * They all start with a BRANCH for "\|" alternaties, even when there is only
|
|
123 * one alternative.
|
|
124 */
|
|
125
|
|
126 /*
|
|
127 * The opcodes are:
|
|
128 */
|
|
129
|
|
130 /* definition number opnd? meaning */
|
|
131 #define END 0 /* End of program or NOMATCH operand. */
|
|
132 #define BOL 1 /* Match "" at beginning of line. */
|
|
133 #define EOL 2 /* Match "" at end of line. */
|
|
134 #define BRANCH 3 /* node Match this alternative, or the
|
|
135 * next... */
|
|
136 #define BACK 4 /* Match "", "next" ptr points backward. */
|
|
137 #define EXACTLY 5 /* str Match this string. */
|
|
138 #define NOTHING 6 /* Match empty string. */
|
|
139 #define STAR 7 /* node Match this (simple) thing 0 or more
|
|
140 * times. */
|
|
141 #define PLUS 8 /* node Match this (simple) thing 1 or more
|
|
142 * times. */
|
|
143 #define MATCH 9 /* node match the operand zero-width */
|
|
144 #define NOMATCH 10 /* node check for no match with operand */
|
|
145 #define BEHIND 11 /* node look behind for a match with operand */
|
|
146 #define NOBEHIND 12 /* node look behind for no match with operand */
|
|
147 #define SUBPAT 13 /* node match the operand here */
|
|
148 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and
|
|
149 * n times (\{m,n\}). */
|
|
150 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */
|
|
151 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */
|
|
152 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE
|
|
153 * and BRACE_COMPLEX. */
|
|
154 #define NEWL 18 /* Match line-break */
|
|
155 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */
|
|
156
|
|
157
|
|
158 /* character classes: 20-48 normal, 50-78 include a line-break */
|
|
159 #define ADD_NL 30
|
|
160 #define FIRST_NL ANY + ADD_NL
|
|
161 #define ANY 20 /* Match any one character. */
|
|
162 #define ANYOF 21 /* str Match any character in this string. */
|
|
163 #define ANYBUT 22 /* str Match any character not in this
|
|
164 * string. */
|
|
165 #define IDENT 23 /* Match identifier char */
|
|
166 #define SIDENT 24 /* Match identifier char but no digit */
|
|
167 #define KWORD 25 /* Match keyword char */
|
|
168 #define SKWORD 26 /* Match word char but no digit */
|
|
169 #define FNAME 27 /* Match file name char */
|
|
170 #define SFNAME 28 /* Match file name char but no digit */
|
|
171 #define PRINT 29 /* Match printable char */
|
|
172 #define SPRINT 30 /* Match printable char but no digit */
|
|
173 #define WHITE 31 /* Match whitespace char */
|
|
174 #define NWHITE 32 /* Match non-whitespace char */
|
|
175 #define DIGIT 33 /* Match digit char */
|
|
176 #define NDIGIT 34 /* Match non-digit char */
|
|
177 #define HEX 35 /* Match hex char */
|
|
178 #define NHEX 36 /* Match non-hex char */
|
|
179 #define OCTAL 37 /* Match octal char */
|
|
180 #define NOCTAL 38 /* Match non-octal char */
|
|
181 #define WORD 39 /* Match word char */
|
|
182 #define NWORD 40 /* Match non-word char */
|
|
183 #define HEAD 41 /* Match head char */
|
|
184 #define NHEAD 42 /* Match non-head char */
|
|
185 #define ALPHA 43 /* Match alpha char */
|
|
186 #define NALPHA 44 /* Match non-alpha char */
|
|
187 #define LOWER 45 /* Match lowercase char */
|
|
188 #define NLOWER 46 /* Match non-lowercase char */
|
|
189 #define UPPER 47 /* Match uppercase char */
|
|
190 #define NUPPER 48 /* Match non-uppercase char */
|
|
191 #define LAST_NL NUPPER + ADD_NL
|
|
192 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL)
|
|
193
|
|
194 #define MOPEN 80 /* -89 Mark this point in input as start of
|
|
195 * \( subexpr. MOPEN + 0 marks start of
|
|
196 * match. */
|
|
197 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks
|
|
198 * end of match. */
|
|
199 #define BACKREF 100 /* -109 node Match same string again \1-\9 */
|
|
200
|
|
201 #ifdef FEAT_SYN_HL
|
|
202 # define ZOPEN 110 /* -119 Mark this point in input as start of
|
|
203 * \z( subexpr. */
|
|
204 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */
|
|
205 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */
|
|
206 #endif
|
|
207
|
|
208 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */
|
|
209
|
|
210 #define NOPEN 150 /* Mark this point in input as start of
|
|
211 \%( subexpr. */
|
|
212 #define NCLOSE 151 /* Analogous to NOPEN. */
|
|
213
|
|
214 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */
|
|
215 #define RE_BOF 201 /* Match "" at beginning of file. */
|
|
216 #define RE_EOF 202 /* Match "" at end of file. */
|
|
217 #define CURSOR 203 /* Match location of cursor. */
|
|
218
|
|
219 #define RE_LNUM 204 /* nr cmp Match line number */
|
|
220 #define RE_COL 205 /* nr cmp Match column number */
|
|
221 #define RE_VCOL 206 /* nr cmp Match virtual column number */
|
|
222
|
|
223 /*
|
|
224 * Magic characters have a special meaning, they don't match literally.
|
|
225 * Magic characters are negative. This separates them from literal characters
|
|
226 * (possibly multi-byte). Only ASCII characters can be Magic.
|
|
227 */
|
|
228 #define Magic(x) ((int)(x) - 256)
|
|
229 #define un_Magic(x) ((x) + 256)
|
|
230 #define is_Magic(x) ((x) < 0)
|
|
231
|
|
232 static int no_Magic __ARGS((int x));
|
|
233 static int toggle_Magic __ARGS((int x));
|
|
234
|
|
235 static int
|
|
236 no_Magic(x)
|
|
237 int x;
|
|
238 {
|
|
239 if (is_Magic(x))
|
|
240 return un_Magic(x);
|
|
241 return x;
|
|
242 }
|
|
243
|
|
244 static int
|
|
245 toggle_Magic(x)
|
|
246 int x;
|
|
247 {
|
|
248 if (is_Magic(x))
|
|
249 return un_Magic(x);
|
|
250 return Magic(x);
|
|
251 }
|
|
252
|
|
253 /*
|
|
254 * The first byte of the regexp internal "program" is actually this magic
|
|
255 * number; the start node begins in the second byte. It's used to catch the
|
|
256 * most severe mutilation of the program by the caller.
|
|
257 */
|
|
258
|
|
259 #define REGMAGIC 0234
|
|
260
|
|
261 /*
|
|
262 * Opcode notes:
|
|
263 *
|
|
264 * BRANCH The set of branches constituting a single choice are hooked
|
|
265 * together with their "next" pointers, since precedence prevents
|
|
266 * anything being concatenated to any individual branch. The
|
|
267 * "next" pointer of the last BRANCH in a choice points to the
|
|
268 * thing following the whole choice. This is also where the
|
|
269 * final "next" pointer of each individual branch points; each
|
|
270 * branch starts with the operand node of a BRANCH node.
|
|
271 *
|
|
272 * BACK Normal "next" pointers all implicitly point forward; BACK
|
|
273 * exists to make loop structures possible.
|
|
274 *
|
|
275 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular
|
|
276 * BRANCH structures using BACK. Simple cases (one character
|
|
277 * per match) are implemented with STAR and PLUS for speed
|
|
278 * and to minimize recursive plunges.
|
|
279 *
|
|
280 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX
|
|
281 * node, and defines the min and max limits to be used for that
|
|
282 * node.
|
|
283 *
|
|
284 * MOPEN,MCLOSE ...are numbered at compile time.
|
|
285 * ZOPEN,ZCLOSE ...ditto
|
|
286 */
|
|
287
|
|
288 /*
|
|
289 * A node is one char of opcode followed by two chars of "next" pointer.
|
|
290 * "Next" pointers are stored as two 8-bit bytes, high order first. The
|
|
291 * value is a positive offset from the opcode of the node containing it.
|
|
292 * An operand, if any, simply follows the node. (Note that much of the
|
|
293 * code generation knows about this implicit relationship.)
|
|
294 *
|
|
295 * Using two bytes for the "next" pointer is vast overkill for most things,
|
|
296 * but allows patterns to get big without disasters.
|
|
297 */
|
|
298 #define OP(p) ((int)*(p))
|
|
299 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377))
|
|
300 #define OPERAND(p) ((p) + 3)
|
|
301 /* Obtain an operand that was stored as four bytes, MSB first. */
|
|
302 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \
|
|
303 + ((long)(p)[5] << 8) + (long)(p)[6])
|
|
304 /* Obtain a second operand stored as four bytes. */
|
|
305 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4)
|
|
306 /* Obtain a second single-byte operand stored after a four bytes operand. */
|
|
307 #define OPERAND_CMP(p) (p)[7]
|
|
308
|
|
309 /*
|
|
310 * Utility definitions.
|
|
311 */
|
|
312 #define UCHARAT(p) ((int)*(char_u *)(p))
|
|
313
|
|
314 /* Used for an error (down from) vim_regcomp(): give the error message, set
|
|
315 * rc_did_emsg and return NULL */
|
|
316 #define EMSG_RET_NULL(m) { EMSG(m); rc_did_emsg = TRUE; return NULL; }
|
|
317 #define EMSG_M_RET_NULL(m, c) { EMSG2(m, c ? "" : "\\"); rc_did_emsg = TRUE; return NULL; }
|
|
318 #define EMSG_RET_FAIL(m) { EMSG(m); rc_did_emsg = TRUE; return FAIL; }
|
|
319 #define EMSG_ONE_RET_NULL EMSG_M_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL)
|
|
320
|
|
321 #define MAX_LIMIT (32767L << 16L)
|
|
322
|
|
323 static int re_multi_type __ARGS((int));
|
|
324 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n));
|
|
325 static char_u *cstrchr __ARGS((char_u *, int));
|
|
326
|
|
327 #ifdef DEBUG
|
|
328 static void regdump __ARGS((char_u *, regprog_T *));
|
|
329 static char_u *regprop __ARGS((char_u *));
|
|
330 #endif
|
|
331
|
|
332 #define NOT_MULTI 0
|
|
333 #define MULTI_ONE 1
|
|
334 #define MULTI_MULT 2
|
|
335 /*
|
|
336 * Return NOT_MULTI if c is not a "multi" operator.
|
|
337 * Return MULTI_ONE if c is a single "multi" operator.
|
|
338 * Return MULTI_MULT if c is a multi "multi" operator.
|
|
339 */
|
|
340 static int
|
|
341 re_multi_type(c)
|
|
342 int c;
|
|
343 {
|
|
344 if (c == Magic('@') || c == Magic('=') || c == Magic('?'))
|
|
345 return MULTI_ONE;
|
|
346 if (c == Magic('*') || c == Magic('+') || c == Magic('{'))
|
|
347 return MULTI_MULT;
|
|
348 return NOT_MULTI;
|
|
349 }
|
|
350
|
|
351 /*
|
|
352 * Flags to be passed up and down.
|
|
353 */
|
|
354 #define HASWIDTH 0x1 /* Known never to match null string. */
|
|
355 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
|
|
356 #define SPSTART 0x4 /* Starts with * or +. */
|
|
357 #define HASNL 0x8 /* Contains some \n. */
|
|
358 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */
|
|
359 #define WORST 0 /* Worst case. */
|
|
360
|
|
361 /*
|
|
362 * When regcode is set to this value, code is not emitted and size is computed
|
|
363 * instead.
|
|
364 */
|
|
365 #define JUST_CALC_SIZE ((char_u *) -1)
|
|
366
|
|
367 static char_u *reg_prev_sub;
|
|
368
|
|
369 /*
|
|
370 * REGEXP_INRANGE contains all characters which are always special in a []
|
|
371 * range after '\'.
|
|
372 * REGEXP_ABBR contains all characters which act as abbreviations after '\'.
|
|
373 * These are:
|
|
374 * \n - New line (NL).
|
|
375 * \r - Carriage Return (CR).
|
|
376 * \t - Tab (TAB).
|
|
377 * \e - Escape (ESC).
|
|
378 * \b - Backspace (Ctrl_H).
|
24
|
379 * \d - Character code in decimal, eg \d123
|
|
380 * \o - Character code in octal, eg \o80
|
|
381 * \x - Character code in hex, eg \x4a
|
|
382 * \u - Multibyte character code, eg \u20ac
|
|
383 * \U - Long multibyte character code, eg \U12345678
|
7
|
384 */
|
|
385 static char_u REGEXP_INRANGE[] = "]^-n\\";
|
24
|
386 static char_u REGEXP_ABBR[] = "nrtebdoxuU";
|
7
|
387
|
|
388 static int backslash_trans __ARGS((int c));
|
|
389 static int skip_class_name __ARGS((char_u **pp));
|
|
390 static char_u *skip_anyof __ARGS((char_u *p));
|
|
391 static void init_class_tab __ARGS((void));
|
|
392
|
|
393 /*
|
|
394 * Translate '\x' to its control character, except "\n", which is Magic.
|
|
395 */
|
|
396 static int
|
|
397 backslash_trans(c)
|
|
398 int c;
|
|
399 {
|
|
400 switch (c)
|
|
401 {
|
|
402 case 'r': return CAR;
|
|
403 case 't': return TAB;
|
|
404 case 'e': return ESC;
|
|
405 case 'b': return BS;
|
|
406 }
|
|
407 return c;
|
|
408 }
|
|
409
|
|
410 /*
|
|
411 * Check for a character class name. "pp" points to the '['.
|
|
412 * Returns one of the CLASS_ items. CLASS_NONE means that no item was
|
|
413 * recognized. Otherwise "pp" is advanced to after the item.
|
|
414 */
|
|
415 static int
|
|
416 skip_class_name(pp)
|
|
417 char_u **pp;
|
|
418 {
|
|
419 static const char *(class_names[]) =
|
|
420 {
|
|
421 "alnum:]",
|
|
422 #define CLASS_ALNUM 0
|
|
423 "alpha:]",
|
|
424 #define CLASS_ALPHA 1
|
|
425 "blank:]",
|
|
426 #define CLASS_BLANK 2
|
|
427 "cntrl:]",
|
|
428 #define CLASS_CNTRL 3
|
|
429 "digit:]",
|
|
430 #define CLASS_DIGIT 4
|
|
431 "graph:]",
|
|
432 #define CLASS_GRAPH 5
|
|
433 "lower:]",
|
|
434 #define CLASS_LOWER 6
|
|
435 "print:]",
|
|
436 #define CLASS_PRINT 7
|
|
437 "punct:]",
|
|
438 #define CLASS_PUNCT 8
|
|
439 "space:]",
|
|
440 #define CLASS_SPACE 9
|
|
441 "upper:]",
|
|
442 #define CLASS_UPPER 10
|
|
443 "xdigit:]",
|
|
444 #define CLASS_XDIGIT 11
|
|
445 "tab:]",
|
|
446 #define CLASS_TAB 12
|
|
447 "return:]",
|
|
448 #define CLASS_RETURN 13
|
|
449 "backspace:]",
|
|
450 #define CLASS_BACKSPACE 14
|
|
451 "escape:]",
|
|
452 #define CLASS_ESCAPE 15
|
|
453 };
|
|
454 #define CLASS_NONE 99
|
|
455 int i;
|
|
456
|
|
457 if ((*pp)[1] == ':')
|
|
458 {
|
|
459 for (i = 0; i < sizeof(class_names) / sizeof(*class_names); ++i)
|
|
460 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0)
|
|
461 {
|
|
462 *pp += STRLEN(class_names[i]) + 2;
|
|
463 return i;
|
|
464 }
|
|
465 }
|
|
466 return CLASS_NONE;
|
|
467 }
|
|
468
|
|
469 /*
|
|
470 * Skip over a "[]" range.
|
|
471 * "p" must point to the character after the '['.
|
|
472 * The returned pointer is on the matching ']', or the terminating NUL.
|
|
473 */
|
|
474 static char_u *
|
|
475 skip_anyof(p)
|
|
476 char_u *p;
|
|
477 {
|
|
478 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
|
479 #ifdef FEAT_MBYTE
|
|
480 int l;
|
|
481 #endif
|
|
482
|
|
483 cpo_lit = (!reg_syn && vim_strchr(p_cpo, CPO_LITERAL) != NULL);
|
|
484
|
|
485 if (*p == '^') /* Complement of range. */
|
|
486 ++p;
|
|
487 if (*p == ']' || *p == '-')
|
|
488 ++p;
|
|
489 while (*p != NUL && *p != ']')
|
|
490 {
|
|
491 #ifdef FEAT_MBYTE
|
|
492 if (has_mbyte && (l = (*mb_ptr2len_check)(p)) > 1)
|
|
493 p += l;
|
|
494 else
|
|
495 #endif
|
|
496 if (*p == '-')
|
|
497 {
|
|
498 ++p;
|
|
499 if (*p != ']' && *p != NUL)
|
39
|
500 mb_ptr_adv(p);
|
7
|
501 }
|
|
502 else if (*p == '\\'
|
|
503 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL
|
|
504 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL)))
|
|
505 p += 2;
|
|
506 else if (*p == '[')
|
|
507 {
|
|
508 if (skip_class_name(&p) == CLASS_NONE)
|
|
509 ++p; /* It was not a class name */
|
|
510 }
|
|
511 else
|
|
512 ++p;
|
|
513 }
|
|
514
|
|
515 return p;
|
|
516 }
|
|
517
|
|
518 /*
|
|
519 * Specific version of character class functions.
|
|
520 * Using a table to keep this fast.
|
|
521 */
|
|
522 static short class_tab[256];
|
|
523
|
|
524 #define RI_DIGIT 0x01
|
|
525 #define RI_HEX 0x02
|
|
526 #define RI_OCTAL 0x04
|
|
527 #define RI_WORD 0x08
|
|
528 #define RI_HEAD 0x10
|
|
529 #define RI_ALPHA 0x20
|
|
530 #define RI_LOWER 0x40
|
|
531 #define RI_UPPER 0x80
|
|
532 #define RI_WHITE 0x100
|
|
533
|
|
534 static void
|
|
535 init_class_tab()
|
|
536 {
|
|
537 int i;
|
|
538 static int done = FALSE;
|
|
539
|
|
540 if (done)
|
|
541 return;
|
|
542
|
|
543 for (i = 0; i < 256; ++i)
|
|
544 {
|
|
545 if (i >= '0' && i <= '7')
|
|
546 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD;
|
|
547 else if (i >= '8' && i <= '9')
|
|
548 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD;
|
|
549 else if (i >= 'a' && i <= 'f')
|
|
550 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
551 #ifdef EBCDIC
|
|
552 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r')
|
|
553 || (i >= 's' && i <= 'z'))
|
|
554 #else
|
|
555 else if (i >= 'g' && i <= 'z')
|
|
556 #endif
|
|
557 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
558 else if (i >= 'A' && i <= 'F')
|
|
559 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
560 #ifdef EBCDIC
|
|
561 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R')
|
|
562 || (i >= 'S' && i <= 'Z'))
|
|
563 #else
|
|
564 else if (i >= 'G' && i <= 'Z')
|
|
565 #endif
|
|
566 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
567 else if (i == '_')
|
|
568 class_tab[i] = RI_WORD + RI_HEAD;
|
|
569 else
|
|
570 class_tab[i] = 0;
|
|
571 }
|
|
572 class_tab[' '] |= RI_WHITE;
|
|
573 class_tab['\t'] |= RI_WHITE;
|
|
574 done = TRUE;
|
|
575 }
|
|
576
|
|
577 #ifdef FEAT_MBYTE
|
|
578 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT))
|
|
579 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX))
|
|
580 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL))
|
|
581 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD))
|
|
582 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD))
|
|
583 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA))
|
|
584 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER))
|
|
585 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER))
|
|
586 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE))
|
|
587 #else
|
|
588 # define ri_digit(c) (class_tab[c] & RI_DIGIT)
|
|
589 # define ri_hex(c) (class_tab[c] & RI_HEX)
|
|
590 # define ri_octal(c) (class_tab[c] & RI_OCTAL)
|
|
591 # define ri_word(c) (class_tab[c] & RI_WORD)
|
|
592 # define ri_head(c) (class_tab[c] & RI_HEAD)
|
|
593 # define ri_alpha(c) (class_tab[c] & RI_ALPHA)
|
|
594 # define ri_lower(c) (class_tab[c] & RI_LOWER)
|
|
595 # define ri_upper(c) (class_tab[c] & RI_UPPER)
|
|
596 # define ri_white(c) (class_tab[c] & RI_WHITE)
|
|
597 #endif
|
|
598
|
|
599 /* flags for regflags */
|
|
600 #define RF_ICASE 1 /* ignore case */
|
|
601 #define RF_NOICASE 2 /* don't ignore case */
|
|
602 #define RF_HASNL 4 /* can match a NL */
|
|
603 #define RF_ICOMBINE 8 /* ignore combining characters */
|
|
604 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */
|
|
605
|
|
606 /*
|
|
607 * Global work variables for vim_regcomp().
|
|
608 */
|
|
609
|
|
610 static char_u *regparse; /* Input-scan pointer. */
|
|
611 static int prevchr_len; /* byte length of previous char */
|
|
612 static int num_complex_braces; /* Complex \{...} count */
|
|
613 static int regnpar; /* () count. */
|
|
614 #ifdef FEAT_SYN_HL
|
|
615 static int regnzpar; /* \z() count. */
|
|
616 static int re_has_z; /* \z item detected */
|
|
617 #endif
|
|
618 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */
|
|
619 static long regsize; /* Code size. */
|
|
620 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */
|
|
621 static unsigned regflags; /* RF_ flags for prog */
|
|
622 static long brace_min[10]; /* Minimums for complex brace repeats */
|
|
623 static long brace_max[10]; /* Maximums for complex brace repeats */
|
|
624 static int brace_count[10]; /* Current counts for complex brace repeats */
|
|
625 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
626 static int had_eol; /* TRUE when EOL found by vim_regcomp() */
|
|
627 #endif
|
|
628 static int one_exactly = FALSE; /* only do one char for EXACTLY */
|
|
629
|
|
630 static int reg_magic; /* magicness of the pattern: */
|
|
631 #define MAGIC_NONE 1 /* "\V" very unmagic */
|
|
632 #define MAGIC_OFF 2 /* "\M" or 'magic' off */
|
|
633 #define MAGIC_ON 3 /* "\m" or 'magic' */
|
|
634 #define MAGIC_ALL 4 /* "\v" very magic */
|
|
635
|
|
636 static int reg_string; /* matching with a string instead of a buffer
|
|
637 line */
|
|
638
|
|
639 /*
|
|
640 * META contains all characters that may be magic, except '^' and '$'.
|
|
641 */
|
|
642
|
|
643 #ifdef EBCDIC
|
|
644 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~";
|
|
645 #else
|
|
646 /* META[] is used often enough to justify turning it into a table. */
|
|
647 static char_u META_flags[] = {
|
|
648 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
649 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
650 /* % & ( ) * + . */
|
|
651 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0,
|
|
652 /* 1 2 3 4 5 6 7 8 9 < = > ? */
|
|
653 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1,
|
|
654 /* @ A C D F H I K L M O */
|
|
655 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
|
|
656 /* P S U V W X Z [ _ */
|
|
657 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1,
|
|
658 /* a c d f h i k l m n o */
|
|
659 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
|
|
660 /* p s u v w x z { | ~ */
|
|
661 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1
|
|
662 };
|
|
663 #endif
|
|
664
|
|
665 static int curchr;
|
|
666
|
|
667 /* arguments for reg() */
|
|
668 #define REG_NOPAREN 0 /* toplevel reg() */
|
|
669 #define REG_PAREN 1 /* \(\) */
|
|
670 #define REG_ZPAREN 2 /* \z(\) */
|
|
671 #define REG_NPAREN 3 /* \%(\) */
|
|
672
|
|
673 /*
|
|
674 * Forward declarations for vim_regcomp()'s friends.
|
|
675 */
|
|
676 static void initchr __ARGS((char_u *));
|
|
677 static int getchr __ARGS((void));
|
|
678 static void skipchr_keepstart __ARGS((void));
|
|
679 static int peekchr __ARGS((void));
|
|
680 static void skipchr __ARGS((void));
|
|
681 static void ungetchr __ARGS((void));
|
24
|
682 static int gethexchrs __ARGS((int maxinputlen));
|
|
683 static int getoctchrs __ARGS((void));
|
|
684 static int getdecchrs __ARGS((void));
|
|
685 static int coll_get_char __ARGS((void));
|
7
|
686 static void regcomp_start __ARGS((char_u *expr, int flags));
|
|
687 static char_u *reg __ARGS((int, int *));
|
|
688 static char_u *regbranch __ARGS((int *flagp));
|
|
689 static char_u *regconcat __ARGS((int *flagp));
|
|
690 static char_u *regpiece __ARGS((int *));
|
|
691 static char_u *regatom __ARGS((int *));
|
|
692 static char_u *regnode __ARGS((int));
|
|
693 static int prog_magic_wrong __ARGS((void));
|
|
694 static char_u *regnext __ARGS((char_u *));
|
|
695 static void regc __ARGS((int b));
|
|
696 #ifdef FEAT_MBYTE
|
|
697 static void regmbc __ARGS((int c));
|
|
698 #endif
|
|
699 static void reginsert __ARGS((int, char_u *));
|
|
700 static void reginsert_limits __ARGS((int, long, long, char_u *));
|
|
701 static char_u *re_put_long __ARGS((char_u *pr, long_u val));
|
|
702 static int read_limits __ARGS((long *, long *));
|
|
703 static void regtail __ARGS((char_u *, char_u *));
|
|
704 static void regoptail __ARGS((char_u *, char_u *));
|
|
705
|
|
706 /*
|
|
707 * Return TRUE if compiled regular expression "prog" can match a line break.
|
|
708 */
|
|
709 int
|
|
710 re_multiline(prog)
|
|
711 regprog_T *prog;
|
|
712 {
|
|
713 return (prog->regflags & RF_HASNL);
|
|
714 }
|
|
715
|
|
716 /*
|
|
717 * Return TRUE if compiled regular expression "prog" looks before the start
|
|
718 * position (pattern contains "\@<=" or "\@<!").
|
|
719 */
|
|
720 int
|
|
721 re_lookbehind(prog)
|
|
722 regprog_T *prog;
|
|
723 {
|
|
724 return (prog->regflags & RF_LOOKBH);
|
|
725 }
|
|
726
|
|
727 /*
|
|
728 * Skip past regular expression.
|
41
|
729 * Stop at end of 'p' or where 'dirc' is found ('/', '?', etc).
|
7
|
730 * Take care of characters with a backslash in front of it.
|
|
731 * Skip strings inside [ and ].
|
|
732 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the
|
|
733 * expression and change "\?" to "?". If "*newp" is not NULL the expression
|
|
734 * is changed in-place.
|
|
735 */
|
|
736 char_u *
|
|
737 skip_regexp(startp, dirc, magic, newp)
|
|
738 char_u *startp;
|
|
739 int dirc;
|
|
740 int magic;
|
|
741 char_u **newp;
|
|
742 {
|
|
743 int mymagic;
|
|
744 char_u *p = startp;
|
|
745
|
|
746 if (magic)
|
|
747 mymagic = MAGIC_ON;
|
|
748 else
|
|
749 mymagic = MAGIC_OFF;
|
|
750
|
39
|
751 for (; p[0] != NUL; mb_ptr_adv(p))
|
7
|
752 {
|
|
753 if (p[0] == dirc) /* found end of regexp */
|
|
754 break;
|
|
755 if ((p[0] == '[' && mymagic >= MAGIC_ON)
|
|
756 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF))
|
|
757 {
|
|
758 p = skip_anyof(p + 1);
|
|
759 if (p[0] == NUL)
|
|
760 break;
|
|
761 }
|
|
762 else if (p[0] == '\\' && p[1] != NUL)
|
|
763 {
|
|
764 if (dirc == '?' && newp != NULL && p[1] == '?')
|
|
765 {
|
|
766 /* change "\?" to "?", make a copy first. */
|
|
767 if (*newp == NULL)
|
|
768 {
|
|
769 *newp = vim_strsave(startp);
|
|
770 if (*newp != NULL)
|
|
771 p = *newp + (p - startp);
|
|
772 }
|
|
773 if (*newp != NULL)
|
|
774 mch_memmove(p, p + 1, STRLEN(p));
|
|
775 else
|
|
776 ++p;
|
|
777 }
|
|
778 else
|
|
779 ++p; /* skip next character */
|
|
780 if (*p == 'v')
|
|
781 mymagic = MAGIC_ALL;
|
|
782 else if (*p == 'V')
|
|
783 mymagic = MAGIC_NONE;
|
|
784 }
|
|
785 }
|
|
786 return p;
|
|
787 }
|
|
788
|
|
789 /*
|
41
|
790 * vim_regcomp() - compile a regular expression into internal code
|
|
791 * Returns the program in allocated space. Returns NULL for an error.
|
7
|
792 *
|
|
793 * We can't allocate space until we know how big the compiled form will be,
|
|
794 * but we can't compile it (and thus know how big it is) until we've got a
|
|
795 * place to put the code. So we cheat: we compile it twice, once with code
|
|
796 * generation turned off and size counting turned on, and once "for real".
|
|
797 * This also means that we don't allocate space until we are sure that the
|
|
798 * thing really will compile successfully, and we never have to move the
|
|
799 * code and thus invalidate pointers into it. (Note that it has to be in
|
|
800 * one piece because vim_free() must be able to free it all.)
|
|
801 *
|
|
802 * Whether upper/lower case is to be ignored is decided when executing the
|
|
803 * program, it does not matter here.
|
|
804 *
|
|
805 * Beware that the optimization-preparation code in here knows about some
|
|
806 * of the structure of the compiled regexp.
|
|
807 * "re_flags": RE_MAGIC and/or RE_STRING.
|
|
808 */
|
|
809 regprog_T *
|
|
810 vim_regcomp(expr, re_flags)
|
|
811 char_u *expr;
|
|
812 int re_flags;
|
|
813 {
|
|
814 regprog_T *r;
|
|
815 char_u *scan;
|
|
816 char_u *longest;
|
|
817 int len;
|
|
818 int flags;
|
|
819
|
|
820 if (expr == NULL)
|
|
821 EMSG_RET_NULL(_(e_null));
|
|
822
|
|
823 init_class_tab();
|
|
824
|
|
825 /*
|
|
826 * First pass: determine size, legality.
|
|
827 */
|
|
828 regcomp_start(expr, re_flags);
|
|
829 regcode = JUST_CALC_SIZE;
|
|
830 regc(REGMAGIC);
|
|
831 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
832 return NULL;
|
|
833
|
|
834 /* Small enough for pointer-storage convention? */
|
|
835 #ifdef SMALL_MALLOC /* 16 bit storage allocation */
|
|
836 if (regsize >= 65536L - 256L)
|
|
837 EMSG_RET_NULL(_("E339: Pattern too long"));
|
|
838 #endif
|
|
839
|
|
840 /* Allocate space. */
|
|
841 r = (regprog_T *)lalloc(sizeof(regprog_T) + regsize, TRUE);
|
|
842 if (r == NULL)
|
|
843 return NULL;
|
|
844
|
|
845 /*
|
|
846 * Second pass: emit code.
|
|
847 */
|
|
848 regcomp_start(expr, re_flags);
|
|
849 regcode = r->program;
|
|
850 regc(REGMAGIC);
|
|
851 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
852 {
|
|
853 vim_free(r);
|
|
854 return NULL;
|
|
855 }
|
|
856
|
|
857 /* Dig out information for optimizations. */
|
|
858 r->regstart = NUL; /* Worst-case defaults. */
|
|
859 r->reganch = 0;
|
|
860 r->regmust = NULL;
|
|
861 r->regmlen = 0;
|
|
862 r->regflags = regflags;
|
|
863 if (flags & HASNL)
|
|
864 r->regflags |= RF_HASNL;
|
|
865 if (flags & HASLOOKBH)
|
|
866 r->regflags |= RF_LOOKBH;
|
|
867 #ifdef FEAT_SYN_HL
|
|
868 /* Remember whether this pattern has any \z specials in it. */
|
|
869 r->reghasz = re_has_z;
|
|
870 #endif
|
|
871 scan = r->program + 1; /* First BRANCH. */
|
|
872 if (OP(regnext(scan)) == END) /* Only one top-level choice. */
|
|
873 {
|
|
874 scan = OPERAND(scan);
|
|
875
|
|
876 /* Starting-point info. */
|
|
877 if (OP(scan) == BOL || OP(scan) == RE_BOF)
|
|
878 {
|
|
879 r->reganch++;
|
|
880 scan = regnext(scan);
|
|
881 }
|
|
882
|
|
883 if (OP(scan) == EXACTLY)
|
|
884 {
|
|
885 #ifdef FEAT_MBYTE
|
|
886 if (has_mbyte)
|
|
887 r->regstart = (*mb_ptr2char)(OPERAND(scan));
|
|
888 else
|
|
889 #endif
|
|
890 r->regstart = *OPERAND(scan);
|
|
891 }
|
|
892 else if ((OP(scan) == BOW
|
|
893 || OP(scan) == EOW
|
|
894 || OP(scan) == NOTHING
|
|
895 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
|
|
896 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
|
|
897 && OP(regnext(scan)) == EXACTLY)
|
|
898 {
|
|
899 #ifdef FEAT_MBYTE
|
|
900 if (has_mbyte)
|
|
901 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
|
|
902 else
|
|
903 #endif
|
|
904 r->regstart = *OPERAND(regnext(scan));
|
|
905 }
|
|
906
|
|
907 /*
|
|
908 * If there's something expensive in the r.e., find the longest
|
|
909 * literal string that must appear and make it the regmust. Resolve
|
|
910 * ties in favor of later strings, since the regstart check works
|
|
911 * with the beginning of the r.e. and avoiding duplication
|
|
912 * strengthens checking. Not a strong reason, but sufficient in the
|
|
913 * absence of others.
|
|
914 */
|
|
915 /*
|
|
916 * When the r.e. starts with BOW, it is faster to look for a regmust
|
|
917 * first. Used a lot for "#" and "*" commands. (Added by mool).
|
|
918 */
|
|
919 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
|
|
920 && !(flags & HASNL))
|
|
921 {
|
|
922 longest = NULL;
|
|
923 len = 0;
|
|
924 for (; scan != NULL; scan = regnext(scan))
|
|
925 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
|
|
926 {
|
|
927 longest = OPERAND(scan);
|
|
928 len = (int)STRLEN(OPERAND(scan));
|
|
929 }
|
|
930 r->regmust = longest;
|
|
931 r->regmlen = len;
|
|
932 }
|
|
933 }
|
|
934 #ifdef DEBUG
|
|
935 regdump(expr, r);
|
|
936 #endif
|
|
937 return r;
|
|
938 }
|
|
939
|
|
940 /*
|
|
941 * Setup to parse the regexp. Used once to get the length and once to do it.
|
|
942 */
|
|
943 static void
|
|
944 regcomp_start(expr, re_flags)
|
|
945 char_u *expr;
|
|
946 int re_flags; /* see vim_regcomp() */
|
|
947 {
|
|
948 initchr(expr);
|
|
949 if (re_flags & RE_MAGIC)
|
|
950 reg_magic = MAGIC_ON;
|
|
951 else
|
|
952 reg_magic = MAGIC_OFF;
|
|
953 reg_string = (re_flags & RE_STRING);
|
|
954
|
|
955 num_complex_braces = 0;
|
|
956 regnpar = 1;
|
|
957 vim_memset(had_endbrace, 0, sizeof(had_endbrace));
|
|
958 #ifdef FEAT_SYN_HL
|
|
959 regnzpar = 1;
|
|
960 re_has_z = 0;
|
|
961 #endif
|
|
962 regsize = 0L;
|
|
963 regflags = 0;
|
|
964 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
965 had_eol = FALSE;
|
|
966 #endif
|
|
967 }
|
|
968
|
|
969 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
970 /*
|
|
971 * Check if during the previous call to vim_regcomp the EOL item "$" has been
|
|
972 * found. This is messy, but it works fine.
|
|
973 */
|
|
974 int
|
|
975 vim_regcomp_had_eol()
|
|
976 {
|
|
977 return had_eol;
|
|
978 }
|
|
979 #endif
|
|
980
|
|
981 /*
|
|
982 * reg - regular expression, i.e. main body or parenthesized thing
|
|
983 *
|
|
984 * Caller must absorb opening parenthesis.
|
|
985 *
|
|
986 * Combining parenthesis handling with the base level of regular expression
|
|
987 * is a trifle forced, but the need to tie the tails of the branches to what
|
|
988 * follows makes it hard to avoid.
|
|
989 */
|
|
990 static char_u *
|
|
991 reg(paren, flagp)
|
|
992 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */
|
|
993 int *flagp;
|
|
994 {
|
|
995 char_u *ret;
|
|
996 char_u *br;
|
|
997 char_u *ender;
|
|
998 int parno = 0;
|
|
999 int flags;
|
|
1000
|
|
1001 *flagp = HASWIDTH; /* Tentatively. */
|
|
1002
|
|
1003 #ifdef FEAT_SYN_HL
|
|
1004 if (paren == REG_ZPAREN)
|
|
1005 {
|
|
1006 /* Make a ZOPEN node. */
|
|
1007 if (regnzpar >= NSUBEXP)
|
|
1008 EMSG_RET_NULL(_("E50: Too many \\z("));
|
|
1009 parno = regnzpar;
|
|
1010 regnzpar++;
|
|
1011 ret = regnode(ZOPEN + parno);
|
|
1012 }
|
|
1013 else
|
|
1014 #endif
|
|
1015 if (paren == REG_PAREN)
|
|
1016 {
|
|
1017 /* Make a MOPEN node. */
|
|
1018 if (regnpar >= NSUBEXP)
|
|
1019 EMSG_M_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL);
|
|
1020 parno = regnpar;
|
|
1021 ++regnpar;
|
|
1022 ret = regnode(MOPEN + parno);
|
|
1023 }
|
|
1024 else if (paren == REG_NPAREN)
|
|
1025 {
|
|
1026 /* Make a NOPEN node. */
|
|
1027 ret = regnode(NOPEN);
|
|
1028 }
|
|
1029 else
|
|
1030 ret = NULL;
|
|
1031
|
|
1032 /* Pick up the branches, linking them together. */
|
|
1033 br = regbranch(&flags);
|
|
1034 if (br == NULL)
|
|
1035 return NULL;
|
|
1036 if (ret != NULL)
|
|
1037 regtail(ret, br); /* [MZ]OPEN -> first. */
|
|
1038 else
|
|
1039 ret = br;
|
|
1040 /* If one of the branches can be zero-width, the whole thing can.
|
|
1041 * If one of the branches has * at start or matches a line-break, the
|
|
1042 * whole thing can. */
|
|
1043 if (!(flags & HASWIDTH))
|
|
1044 *flagp &= ~HASWIDTH;
|
|
1045 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1046 while (peekchr() == Magic('|'))
|
|
1047 {
|
|
1048 skipchr();
|
|
1049 br = regbranch(&flags);
|
|
1050 if (br == NULL)
|
|
1051 return NULL;
|
|
1052 regtail(ret, br); /* BRANCH -> BRANCH. */
|
|
1053 if (!(flags & HASWIDTH))
|
|
1054 *flagp &= ~HASWIDTH;
|
|
1055 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1056 }
|
|
1057
|
|
1058 /* Make a closing node, and hook it on the end. */
|
|
1059 ender = regnode(
|
|
1060 #ifdef FEAT_SYN_HL
|
|
1061 paren == REG_ZPAREN ? ZCLOSE + parno :
|
|
1062 #endif
|
|
1063 paren == REG_PAREN ? MCLOSE + parno :
|
|
1064 paren == REG_NPAREN ? NCLOSE : END);
|
|
1065 regtail(ret, ender);
|
|
1066
|
|
1067 /* Hook the tails of the branches to the closing node. */
|
|
1068 for (br = ret; br != NULL; br = regnext(br))
|
|
1069 regoptail(br, ender);
|
|
1070
|
|
1071 /* Check for proper termination. */
|
|
1072 if (paren != REG_NOPAREN && getchr() != Magic(')'))
|
|
1073 {
|
|
1074 #ifdef FEAT_SYN_HL
|
|
1075 if (paren == REG_ZPAREN)
|
|
1076 EMSG_RET_NULL(_("E52: Unmatched \\z("))
|
|
1077 else
|
|
1078 #endif
|
|
1079 if (paren == REG_NPAREN)
|
|
1080 EMSG_M_RET_NULL(_("E53: Unmatched %s%%("), reg_magic == MAGIC_ALL)
|
|
1081 else
|
|
1082 EMSG_M_RET_NULL(_("E54: Unmatched %s("), reg_magic == MAGIC_ALL)
|
|
1083 }
|
|
1084 else if (paren == REG_NOPAREN && peekchr() != NUL)
|
|
1085 {
|
|
1086 if (curchr == Magic(')'))
|
|
1087 EMSG_M_RET_NULL(_("E55: Unmatched %s)"), reg_magic == MAGIC_ALL)
|
|
1088 else
|
|
1089 EMSG_RET_NULL(_(e_trailing)) /* "Can't happen". */
|
|
1090 /* NOTREACHED */
|
|
1091 }
|
|
1092 /*
|
|
1093 * Here we set the flag allowing back references to this set of
|
|
1094 * parentheses.
|
|
1095 */
|
|
1096 if (paren == REG_PAREN)
|
|
1097 had_endbrace[parno] = TRUE; /* have seen the close paren */
|
|
1098 return ret;
|
|
1099 }
|
|
1100
|
|
1101 /*
|
|
1102 * regbranch - one alternative of an | operator
|
|
1103 *
|
|
1104 * Implements the & operator.
|
|
1105 */
|
|
1106 static char_u *
|
|
1107 regbranch(flagp)
|
|
1108 int *flagp;
|
|
1109 {
|
|
1110 char_u *ret;
|
|
1111 char_u *chain = NULL;
|
|
1112 char_u *latest;
|
|
1113 int flags;
|
|
1114
|
|
1115 *flagp = WORST | HASNL; /* Tentatively. */
|
|
1116
|
|
1117 ret = regnode(BRANCH);
|
|
1118 for (;;)
|
|
1119 {
|
|
1120 latest = regconcat(&flags);
|
|
1121 if (latest == NULL)
|
|
1122 return NULL;
|
|
1123 /* If one of the branches has width, the whole thing has. If one of
|
|
1124 * the branches anchors at start-of-line, the whole thing does.
|
|
1125 * If one of the branches uses look-behind, the whole thing does. */
|
|
1126 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
|
|
1127 /* If one of the branches doesn't match a line-break, the whole thing
|
|
1128 * doesn't. */
|
|
1129 *flagp &= ~HASNL | (flags & HASNL);
|
|
1130 if (chain != NULL)
|
|
1131 regtail(chain, latest);
|
|
1132 if (peekchr() != Magic('&'))
|
|
1133 break;
|
|
1134 skipchr();
|
|
1135 regtail(latest, regnode(END)); /* operand ends */
|
|
1136 reginsert(MATCH, latest);
|
|
1137 chain = latest;
|
|
1138 }
|
|
1139
|
|
1140 return ret;
|
|
1141 }
|
|
1142
|
|
1143 /*
|
|
1144 * regbranch - one alternative of an | or & operator
|
|
1145 *
|
|
1146 * Implements the concatenation operator.
|
|
1147 */
|
|
1148 static char_u *
|
|
1149 regconcat(flagp)
|
|
1150 int *flagp;
|
|
1151 {
|
|
1152 char_u *first = NULL;
|
|
1153 char_u *chain = NULL;
|
|
1154 char_u *latest;
|
|
1155 int flags;
|
|
1156 int cont = TRUE;
|
|
1157
|
|
1158 *flagp = WORST; /* Tentatively. */
|
|
1159
|
|
1160 while (cont)
|
|
1161 {
|
|
1162 switch (peekchr())
|
|
1163 {
|
|
1164 case NUL:
|
|
1165 case Magic('|'):
|
|
1166 case Magic('&'):
|
|
1167 case Magic(')'):
|
|
1168 cont = FALSE;
|
|
1169 break;
|
|
1170 case Magic('Z'):
|
|
1171 #ifdef FEAT_MBYTE
|
|
1172 regflags |= RF_ICOMBINE;
|
|
1173 #endif
|
|
1174 skipchr_keepstart();
|
|
1175 break;
|
|
1176 case Magic('c'):
|
|
1177 regflags |= RF_ICASE;
|
|
1178 skipchr_keepstart();
|
|
1179 break;
|
|
1180 case Magic('C'):
|
|
1181 regflags |= RF_NOICASE;
|
|
1182 skipchr_keepstart();
|
|
1183 break;
|
|
1184 case Magic('v'):
|
|
1185 reg_magic = MAGIC_ALL;
|
|
1186 skipchr_keepstart();
|
|
1187 curchr = -1;
|
|
1188 break;
|
|
1189 case Magic('m'):
|
|
1190 reg_magic = MAGIC_ON;
|
|
1191 skipchr_keepstart();
|
|
1192 curchr = -1;
|
|
1193 break;
|
|
1194 case Magic('M'):
|
|
1195 reg_magic = MAGIC_OFF;
|
|
1196 skipchr_keepstart();
|
|
1197 curchr = -1;
|
|
1198 break;
|
|
1199 case Magic('V'):
|
|
1200 reg_magic = MAGIC_NONE;
|
|
1201 skipchr_keepstart();
|
|
1202 curchr = -1;
|
|
1203 break;
|
|
1204 default:
|
|
1205 latest = regpiece(&flags);
|
|
1206 if (latest == NULL)
|
|
1207 return NULL;
|
|
1208 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
|
|
1209 if (chain == NULL) /* First piece. */
|
|
1210 *flagp |= flags & SPSTART;
|
|
1211 else
|
|
1212 regtail(chain, latest);
|
|
1213 chain = latest;
|
|
1214 if (first == NULL)
|
|
1215 first = latest;
|
|
1216 break;
|
|
1217 }
|
|
1218 }
|
|
1219 if (first == NULL) /* Loop ran zero times. */
|
|
1220 first = regnode(NOTHING);
|
|
1221 return first;
|
|
1222 }
|
|
1223
|
|
1224 /*
|
|
1225 * regpiece - something followed by possible [*+=]
|
|
1226 *
|
|
1227 * Note that the branching code sequences used for = and the general cases
|
|
1228 * of * and + are somewhat optimized: they use the same NOTHING node as
|
|
1229 * both the endmarker for their branch list and the body of the last branch.
|
|
1230 * It might seem that this node could be dispensed with entirely, but the
|
|
1231 * endmarker role is not redundant.
|
|
1232 */
|
|
1233 static char_u *
|
|
1234 regpiece(flagp)
|
|
1235 int *flagp;
|
|
1236 {
|
|
1237 char_u *ret;
|
|
1238 int op;
|
|
1239 char_u *next;
|
|
1240 int flags;
|
|
1241 long minval;
|
|
1242 long maxval;
|
|
1243
|
|
1244 ret = regatom(&flags);
|
|
1245 if (ret == NULL)
|
|
1246 return NULL;
|
|
1247
|
|
1248 op = peekchr();
|
|
1249 if (re_multi_type(op) == NOT_MULTI)
|
|
1250 {
|
|
1251 *flagp = flags;
|
|
1252 return ret;
|
|
1253 }
|
|
1254 if (!(flags & HASWIDTH) && re_multi_type(op) == MULTI_MULT)
|
|
1255 {
|
|
1256 if (op == Magic('*'))
|
|
1257 EMSG_M_RET_NULL(_("E56: %s* operand could be empty"),
|
|
1258 reg_magic >= MAGIC_ON);
|
|
1259 if (op == Magic('+'))
|
|
1260 EMSG_M_RET_NULL(_("E57: %s+ operand could be empty"),
|
|
1261 reg_magic == MAGIC_ALL);
|
|
1262 /* "\{}" is checked below, it's allowed when there is an upper limit */
|
|
1263 }
|
|
1264 /* default flags */
|
|
1265 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
|
|
1266
|
|
1267 skipchr();
|
|
1268 switch (op)
|
|
1269 {
|
|
1270 case Magic('*'):
|
|
1271 if (flags & SIMPLE)
|
|
1272 reginsert(STAR, ret);
|
|
1273 else
|
|
1274 {
|
|
1275 /* Emit x* as (x&|), where & means "self". */
|
|
1276 reginsert(BRANCH, ret); /* Either x */
|
|
1277 regoptail(ret, regnode(BACK)); /* and loop */
|
|
1278 regoptail(ret, ret); /* back */
|
|
1279 regtail(ret, regnode(BRANCH)); /* or */
|
|
1280 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1281 }
|
|
1282 break;
|
|
1283
|
|
1284 case Magic('+'):
|
|
1285 if (flags & SIMPLE)
|
|
1286 reginsert(PLUS, ret);
|
|
1287 else
|
|
1288 {
|
|
1289 /* Emit x+ as x(&|), where & means "self". */
|
|
1290 next = regnode(BRANCH); /* Either */
|
|
1291 regtail(ret, next);
|
|
1292 regtail(regnode(BACK), ret); /* loop back */
|
|
1293 regtail(next, regnode(BRANCH)); /* or */
|
|
1294 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1295 }
|
|
1296 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1297 break;
|
|
1298
|
|
1299 case Magic('@'):
|
|
1300 {
|
|
1301 int lop = END;
|
|
1302
|
|
1303 switch (no_Magic(getchr()))
|
|
1304 {
|
|
1305 case '=': lop = MATCH; break; /* \@= */
|
|
1306 case '!': lop = NOMATCH; break; /* \@! */
|
|
1307 case '>': lop = SUBPAT; break; /* \@> */
|
|
1308 case '<': switch (no_Magic(getchr()))
|
|
1309 {
|
|
1310 case '=': lop = BEHIND; break; /* \@<= */
|
|
1311 case '!': lop = NOBEHIND; break; /* \@<! */
|
|
1312 }
|
|
1313 }
|
|
1314 if (lop == END)
|
|
1315 EMSG_M_RET_NULL(_("E59: invalid character after %s@"),
|
|
1316 reg_magic == MAGIC_ALL);
|
|
1317 /* Look behind must match with behind_pos. */
|
|
1318 if (lop == BEHIND || lop == NOBEHIND)
|
|
1319 {
|
|
1320 regtail(ret, regnode(BHPOS));
|
|
1321 *flagp |= HASLOOKBH;
|
|
1322 }
|
|
1323 regtail(ret, regnode(END)); /* operand ends */
|
|
1324 reginsert(lop, ret);
|
|
1325 break;
|
|
1326 }
|
|
1327
|
|
1328 case Magic('?'):
|
|
1329 case Magic('='):
|
|
1330 /* Emit x= as (x|) */
|
|
1331 reginsert(BRANCH, ret); /* Either x */
|
|
1332 regtail(ret, regnode(BRANCH)); /* or */
|
|
1333 next = regnode(NOTHING); /* null. */
|
|
1334 regtail(ret, next);
|
|
1335 regoptail(ret, next);
|
|
1336 break;
|
|
1337
|
|
1338 case Magic('{'):
|
|
1339 if (!read_limits(&minval, &maxval))
|
|
1340 return NULL;
|
|
1341 if (!(flags & HASWIDTH) && (maxval > minval
|
|
1342 ? maxval >= MAX_LIMIT : minval >= MAX_LIMIT))
|
|
1343 EMSG_M_RET_NULL(_("E58: %s{ operand could be empty"),
|
|
1344 reg_magic == MAGIC_ALL);
|
|
1345 if (flags & SIMPLE)
|
|
1346 {
|
|
1347 reginsert(BRACE_SIMPLE, ret);
|
|
1348 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1349 }
|
|
1350 else
|
|
1351 {
|
|
1352 if (num_complex_braces >= 10)
|
|
1353 EMSG_M_RET_NULL(_("E60: Too many complex %s{...}s"),
|
|
1354 reg_magic == MAGIC_ALL);
|
|
1355 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
|
|
1356 regoptail(ret, regnode(BACK));
|
|
1357 regoptail(ret, ret);
|
|
1358 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1359 ++num_complex_braces;
|
|
1360 }
|
|
1361 if (minval > 0 && maxval > 0)
|
|
1362 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1363 break;
|
|
1364 }
|
|
1365 if (re_multi_type(peekchr()) != NOT_MULTI)
|
|
1366 {
|
|
1367 /* Can't have a multi follow a multi. */
|
|
1368 if (peekchr() == Magic('*'))
|
|
1369 sprintf((char *)IObuff, _("E61: Nested %s*"),
|
|
1370 reg_magic >= MAGIC_ON ? "" : "\\");
|
|
1371 else
|
|
1372 sprintf((char *)IObuff, _("E62: Nested %s%c"),
|
|
1373 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr()));
|
|
1374 EMSG_RET_NULL(IObuff);
|
|
1375 }
|
|
1376
|
|
1377 return ret;
|
|
1378 }
|
|
1379
|
|
1380 /*
|
|
1381 * regatom - the lowest level
|
|
1382 *
|
|
1383 * Optimization: gobbles an entire sequence of ordinary characters so that
|
|
1384 * it can turn them into a single node, which is smaller to store and
|
|
1385 * faster to run. Don't do this when one_exactly is set.
|
|
1386 */
|
|
1387 static char_u *
|
|
1388 regatom(flagp)
|
|
1389 int *flagp;
|
|
1390 {
|
|
1391 char_u *ret;
|
|
1392 int flags;
|
|
1393 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
|
1394 int c;
|
|
1395 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU";
|
|
1396 static int classcodes[] = {ANY, IDENT, SIDENT, KWORD, SKWORD,
|
|
1397 FNAME, SFNAME, PRINT, SPRINT,
|
|
1398 WHITE, NWHITE, DIGIT, NDIGIT,
|
|
1399 HEX, NHEX, OCTAL, NOCTAL,
|
|
1400 WORD, NWORD, HEAD, NHEAD,
|
|
1401 ALPHA, NALPHA, LOWER, NLOWER,
|
|
1402 UPPER, NUPPER
|
|
1403 };
|
|
1404 char_u *p;
|
|
1405 int extra = 0;
|
|
1406
|
|
1407 *flagp = WORST; /* Tentatively. */
|
|
1408 cpo_lit = (!reg_syn && vim_strchr(p_cpo, CPO_LITERAL) != NULL);
|
|
1409
|
|
1410 c = getchr();
|
|
1411 switch (c)
|
|
1412 {
|
|
1413 case Magic('^'):
|
|
1414 ret = regnode(BOL);
|
|
1415 break;
|
|
1416
|
|
1417 case Magic('$'):
|
|
1418 ret = regnode(EOL);
|
|
1419 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1420 had_eol = TRUE;
|
|
1421 #endif
|
|
1422 break;
|
|
1423
|
|
1424 case Magic('<'):
|
|
1425 ret = regnode(BOW);
|
|
1426 break;
|
|
1427
|
|
1428 case Magic('>'):
|
|
1429 ret = regnode(EOW);
|
|
1430 break;
|
|
1431
|
|
1432 case Magic('_'):
|
|
1433 c = no_Magic(getchr());
|
|
1434 if (c == '^') /* "\_^" is start-of-line */
|
|
1435 {
|
|
1436 ret = regnode(BOL);
|
|
1437 break;
|
|
1438 }
|
|
1439 if (c == '$') /* "\_$" is end-of-line */
|
|
1440 {
|
|
1441 ret = regnode(EOL);
|
|
1442 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1443 had_eol = TRUE;
|
|
1444 #endif
|
|
1445 break;
|
|
1446 }
|
|
1447
|
|
1448 extra = ADD_NL;
|
|
1449 *flagp |= HASNL;
|
|
1450
|
|
1451 /* "\_[" is character range plus newline */
|
|
1452 if (c == '[')
|
|
1453 goto collection;
|
|
1454
|
|
1455 /* "\_x" is character class plus newline */
|
|
1456 /*FALLTHROUGH*/
|
|
1457
|
|
1458 /*
|
|
1459 * Character classes.
|
|
1460 */
|
|
1461 case Magic('.'):
|
|
1462 case Magic('i'):
|
|
1463 case Magic('I'):
|
|
1464 case Magic('k'):
|
|
1465 case Magic('K'):
|
|
1466 case Magic('f'):
|
|
1467 case Magic('F'):
|
|
1468 case Magic('p'):
|
|
1469 case Magic('P'):
|
|
1470 case Magic('s'):
|
|
1471 case Magic('S'):
|
|
1472 case Magic('d'):
|
|
1473 case Magic('D'):
|
|
1474 case Magic('x'):
|
|
1475 case Magic('X'):
|
|
1476 case Magic('o'):
|
|
1477 case Magic('O'):
|
|
1478 case Magic('w'):
|
|
1479 case Magic('W'):
|
|
1480 case Magic('h'):
|
|
1481 case Magic('H'):
|
|
1482 case Magic('a'):
|
|
1483 case Magic('A'):
|
|
1484 case Magic('l'):
|
|
1485 case Magic('L'):
|
|
1486 case Magic('u'):
|
|
1487 case Magic('U'):
|
|
1488 p = vim_strchr(classchars, no_Magic(c));
|
|
1489 if (p == NULL)
|
|
1490 EMSG_RET_NULL(_("E63: invalid use of \\_"));
|
|
1491 ret = regnode(classcodes[p - classchars] + extra);
|
|
1492 *flagp |= HASWIDTH | SIMPLE;
|
|
1493 break;
|
|
1494
|
|
1495 case Magic('n'):
|
|
1496 if (reg_string)
|
|
1497 {
|
|
1498 /* In a string "\n" matches a newline character. */
|
|
1499 ret = regnode(EXACTLY);
|
|
1500 regc(NL);
|
|
1501 regc(NUL);
|
|
1502 *flagp |= HASWIDTH | SIMPLE;
|
|
1503 }
|
|
1504 else
|
|
1505 {
|
|
1506 /* In buffer text "\n" matches the end of a line. */
|
|
1507 ret = regnode(NEWL);
|
|
1508 *flagp |= HASWIDTH | HASNL;
|
|
1509 }
|
|
1510 break;
|
|
1511
|
|
1512 case Magic('('):
|
|
1513 if (one_exactly)
|
|
1514 EMSG_ONE_RET_NULL;
|
|
1515 ret = reg(REG_PAREN, &flags);
|
|
1516 if (ret == NULL)
|
|
1517 return NULL;
|
|
1518 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1519 break;
|
|
1520
|
|
1521 case NUL:
|
|
1522 case Magic('|'):
|
|
1523 case Magic('&'):
|
|
1524 case Magic(')'):
|
|
1525 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */
|
|
1526 /* NOTREACHED */
|
|
1527
|
|
1528 case Magic('='):
|
|
1529 case Magic('?'):
|
|
1530 case Magic('+'):
|
|
1531 case Magic('@'):
|
|
1532 case Magic('{'):
|
|
1533 case Magic('*'):
|
|
1534 c = no_Magic(c);
|
|
1535 sprintf((char *)IObuff, _("E64: %s%c follows nothing"),
|
|
1536 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL)
|
|
1537 ? "" : "\\", c);
|
|
1538 EMSG_RET_NULL(IObuff);
|
|
1539 /* NOTREACHED */
|
|
1540
|
|
1541 case Magic('~'): /* previous substitute pattern */
|
|
1542 if (reg_prev_sub)
|
|
1543 {
|
|
1544 char_u *lp;
|
|
1545
|
|
1546 ret = regnode(EXACTLY);
|
|
1547 lp = reg_prev_sub;
|
|
1548 while (*lp != NUL)
|
|
1549 regc(*lp++);
|
|
1550 regc(NUL);
|
|
1551 if (*reg_prev_sub != NUL)
|
|
1552 {
|
|
1553 *flagp |= HASWIDTH;
|
|
1554 if ((lp - reg_prev_sub) == 1)
|
|
1555 *flagp |= SIMPLE;
|
|
1556 }
|
|
1557 }
|
|
1558 else
|
|
1559 EMSG_RET_NULL(_(e_nopresub));
|
|
1560 break;
|
|
1561
|
|
1562 case Magic('1'):
|
|
1563 case Magic('2'):
|
|
1564 case Magic('3'):
|
|
1565 case Magic('4'):
|
|
1566 case Magic('5'):
|
|
1567 case Magic('6'):
|
|
1568 case Magic('7'):
|
|
1569 case Magic('8'):
|
|
1570 case Magic('9'):
|
|
1571 {
|
|
1572 int refnum;
|
|
1573
|
|
1574 refnum = c - Magic('0');
|
|
1575 /*
|
|
1576 * Check if the back reference is legal. We must have seen the
|
|
1577 * close brace.
|
|
1578 * TODO: Should also check that we don't refer to something
|
|
1579 * that is repeated (+*=): what instance of the repetition
|
|
1580 * should we match?
|
|
1581 */
|
|
1582 if (!had_endbrace[refnum])
|
|
1583 {
|
|
1584 /* Trick: check if "@<=" or "@<!" follows, in which case
|
|
1585 * the \1 can appear before the referenced match. */
|
|
1586 for (p = regparse; *p != NUL; ++p)
|
|
1587 if (p[0] == '@' && p[1] == '<'
|
|
1588 && (p[2] == '!' || p[2] == '='))
|
|
1589 break;
|
|
1590 if (*p == NUL)
|
|
1591 EMSG_RET_NULL(_("E65: Illegal back reference"));
|
|
1592 }
|
|
1593 ret = regnode(BACKREF + refnum);
|
|
1594 }
|
|
1595 break;
|
|
1596
|
|
1597 #ifdef FEAT_SYN_HL
|
|
1598 case Magic('z'):
|
|
1599 {
|
|
1600 c = no_Magic(getchr());
|
|
1601 switch (c)
|
|
1602 {
|
|
1603 case '(': if (reg_do_extmatch != REX_SET)
|
|
1604 EMSG_RET_NULL(_("E66: \\z( not allowed here"));
|
|
1605 if (one_exactly)
|
|
1606 EMSG_ONE_RET_NULL;
|
|
1607 ret = reg(REG_ZPAREN, &flags);
|
|
1608 if (ret == NULL)
|
|
1609 return NULL;
|
|
1610 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH);
|
|
1611 re_has_z = REX_SET;
|
|
1612 break;
|
|
1613
|
|
1614 case '1':
|
|
1615 case '2':
|
|
1616 case '3':
|
|
1617 case '4':
|
|
1618 case '5':
|
|
1619 case '6':
|
|
1620 case '7':
|
|
1621 case '8':
|
|
1622 case '9': if (reg_do_extmatch != REX_USE)
|
|
1623 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here"));
|
|
1624 ret = regnode(ZREF + c - '0');
|
|
1625 re_has_z = REX_USE;
|
|
1626 break;
|
|
1627
|
|
1628 case 's': ret = regnode(MOPEN + 0);
|
|
1629 break;
|
|
1630
|
|
1631 case 'e': ret = regnode(MCLOSE + 0);
|
|
1632 break;
|
|
1633
|
|
1634 default: EMSG_RET_NULL(_("E68: Invalid character after \\z"));
|
|
1635 }
|
|
1636 }
|
|
1637 break;
|
|
1638 #endif
|
|
1639
|
|
1640 case Magic('%'):
|
|
1641 {
|
|
1642 c = no_Magic(getchr());
|
|
1643 switch (c)
|
|
1644 {
|
|
1645 /* () without a back reference */
|
|
1646 case '(':
|
|
1647 if (one_exactly)
|
|
1648 EMSG_ONE_RET_NULL;
|
|
1649 ret = reg(REG_NPAREN, &flags);
|
|
1650 if (ret == NULL)
|
|
1651 return NULL;
|
|
1652 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1653 break;
|
|
1654
|
|
1655 /* Catch \%^ and \%$ regardless of where they appear in the
|
|
1656 * pattern -- regardless of whether or not it makes sense. */
|
|
1657 case '^':
|
|
1658 ret = regnode(RE_BOF);
|
|
1659 break;
|
|
1660
|
|
1661 case '$':
|
|
1662 ret = regnode(RE_EOF);
|
|
1663 break;
|
|
1664
|
|
1665 case '#':
|
|
1666 ret = regnode(CURSOR);
|
|
1667 break;
|
|
1668
|
|
1669 /* \%[abc]: Emit as a list of branches, all ending at the last
|
|
1670 * branch which matches nothing. */
|
|
1671 case '[':
|
|
1672 if (one_exactly) /* doesn't nest */
|
|
1673 EMSG_ONE_RET_NULL;
|
|
1674 {
|
|
1675 char_u *lastbranch;
|
|
1676 char_u *lastnode = NULL;
|
|
1677 char_u *br;
|
|
1678
|
|
1679 ret = NULL;
|
|
1680 while ((c = getchr()) != ']')
|
|
1681 {
|
|
1682 if (c == NUL)
|
|
1683 EMSG_M_RET_NULL(_("E69: Missing ] after %s%%["),
|
|
1684 reg_magic == MAGIC_ALL);
|
|
1685 br = regnode(BRANCH);
|
|
1686 if (ret == NULL)
|
|
1687 ret = br;
|
|
1688 else
|
|
1689 regtail(lastnode, br);
|
|
1690
|
|
1691 ungetchr();
|
|
1692 one_exactly = TRUE;
|
|
1693 lastnode = regatom(flagp);
|
|
1694 one_exactly = FALSE;
|
|
1695 if (lastnode == NULL)
|
|
1696 return NULL;
|
|
1697 }
|
|
1698 if (ret == NULL)
|
|
1699 EMSG_M_RET_NULL(_("E70: Empty %s%%[]"),
|
|
1700 reg_magic == MAGIC_ALL);
|
|
1701 lastbranch = regnode(BRANCH);
|
|
1702 br = regnode(NOTHING);
|
|
1703 if (ret != JUST_CALC_SIZE)
|
|
1704 {
|
|
1705 regtail(lastnode, br);
|
|
1706 regtail(lastbranch, br);
|
|
1707 /* connect all branches to the NOTHING
|
|
1708 * branch at the end */
|
|
1709 for (br = ret; br != lastnode; )
|
|
1710 {
|
|
1711 if (OP(br) == BRANCH)
|
|
1712 {
|
|
1713 regtail(br, lastbranch);
|
|
1714 br = OPERAND(br);
|
|
1715 }
|
|
1716 else
|
|
1717 br = regnext(br);
|
|
1718 }
|
|
1719 }
|
|
1720 *flagp &= ~HASWIDTH;
|
|
1721 break;
|
|
1722 }
|
|
1723
|
24
|
1724 case 'd': /* %d123 decimal */
|
|
1725 case 'o': /* %o123 octal */
|
|
1726 case 'x': /* %xab hex 2 */
|
|
1727 case 'u': /* %uabcd hex 4 */
|
|
1728 case 'U': /* %U1234abcd hex 8 */
|
|
1729 {
|
|
1730 int i;
|
|
1731
|
|
1732 switch (c)
|
|
1733 {
|
|
1734 case 'd': i = getdecchrs(); break;
|
|
1735 case 'o': i = getoctchrs(); break;
|
|
1736 case 'x': i = gethexchrs(2); break;
|
|
1737 case 'u': i = gethexchrs(4); break;
|
|
1738 case 'U': i = gethexchrs(8); break;
|
|
1739 default: i = -1; break;
|
|
1740 }
|
|
1741
|
|
1742 if (i < 0)
|
|
1743 EMSG_M_RET_NULL(
|
|
1744 _("E678: Invalid character after %s%%[dxouU]"),
|
|
1745 reg_magic == MAGIC_ALL);
|
|
1746 ret = regnode(EXACTLY);
|
|
1747 if (i == 0)
|
|
1748 regc(0x0a);
|
|
1749 else
|
|
1750 #ifdef FEAT_MBYTE
|
|
1751 regmbc(i);
|
|
1752 #else
|
|
1753 regc(i);
|
|
1754 #endif
|
|
1755 regc(NUL);
|
|
1756 *flagp |= HASWIDTH;
|
|
1757 break;
|
|
1758 }
|
|
1759
|
7
|
1760 default:
|
|
1761 if (VIM_ISDIGIT(c) || c == '<' || c == '>')
|
|
1762 {
|
|
1763 long_u n = 0;
|
|
1764 int cmp;
|
|
1765
|
|
1766 cmp = c;
|
|
1767 if (cmp == '<' || cmp == '>')
|
|
1768 c = getchr();
|
|
1769 while (VIM_ISDIGIT(c))
|
|
1770 {
|
|
1771 n = n * 10 + (c - '0');
|
|
1772 c = getchr();
|
|
1773 }
|
|
1774 if (c == 'l' || c == 'c' || c == 'v')
|
|
1775 {
|
|
1776 if (c == 'l')
|
|
1777 ret = regnode(RE_LNUM);
|
|
1778 else if (c == 'c')
|
|
1779 ret = regnode(RE_COL);
|
|
1780 else
|
|
1781 ret = regnode(RE_VCOL);
|
|
1782 if (ret == JUST_CALC_SIZE)
|
|
1783 regsize += 5;
|
|
1784 else
|
|
1785 {
|
|
1786 /* put the number and the optional
|
|
1787 * comparator after the opcode */
|
|
1788 regcode = re_put_long(regcode, n);
|
|
1789 *regcode++ = cmp;
|
|
1790 }
|
|
1791 break;
|
|
1792 }
|
|
1793 }
|
|
1794
|
|
1795 EMSG_M_RET_NULL(_("E71: Invalid character after %s%%"),
|
|
1796 reg_magic == MAGIC_ALL);
|
|
1797 }
|
|
1798 }
|
|
1799 break;
|
|
1800
|
|
1801 case Magic('['):
|
|
1802 collection:
|
|
1803 {
|
|
1804 char_u *lp;
|
|
1805
|
|
1806 /*
|
|
1807 * If there is no matching ']', we assume the '[' is a normal
|
|
1808 * character. This makes 'incsearch' and ":help [" work.
|
|
1809 */
|
|
1810 lp = skip_anyof(regparse);
|
|
1811 if (*lp == ']') /* there is a matching ']' */
|
|
1812 {
|
|
1813 int startc = -1; /* > 0 when next '-' is a range */
|
|
1814 int endc;
|
|
1815
|
|
1816 /*
|
|
1817 * In a character class, different parsing rules apply.
|
|
1818 * Not even \ is special anymore, nothing is.
|
|
1819 */
|
|
1820 if (*regparse == '^') /* Complement of range. */
|
|
1821 {
|
|
1822 ret = regnode(ANYBUT + extra);
|
|
1823 regparse++;
|
|
1824 }
|
|
1825 else
|
|
1826 ret = regnode(ANYOF + extra);
|
|
1827
|
|
1828 /* At the start ']' and '-' mean the literal character. */
|
|
1829 if (*regparse == ']' || *regparse == '-')
|
|
1830 regc(*regparse++);
|
|
1831
|
|
1832 while (*regparse != NUL && *regparse != ']')
|
|
1833 {
|
|
1834 if (*regparse == '-')
|
|
1835 {
|
|
1836 ++regparse;
|
|
1837 /* The '-' is not used for a range at the end and
|
|
1838 * after or before a '\n'. */
|
|
1839 if (*regparse == ']' || *regparse == NUL
|
|
1840 || startc == -1
|
|
1841 || (regparse[0] == '\\' && regparse[1] == 'n'))
|
|
1842 {
|
|
1843 regc('-');
|
|
1844 startc = '-'; /* [--x] is a range */
|
|
1845 }
|
|
1846 else
|
|
1847 {
|
|
1848 #ifdef FEAT_MBYTE
|
|
1849 if (has_mbyte)
|
|
1850 endc = mb_ptr2char_adv(®parse);
|
|
1851 else
|
|
1852 #endif
|
|
1853 endc = *regparse++;
|
24
|
1854
|
|
1855 /* Handle \o40, \x20 and \u20AC style sequences */
|
|
1856 if (endc == '\\' && !cpo_lit)
|
|
1857 endc = coll_get_char();
|
|
1858
|
7
|
1859 if (startc > endc)
|
|
1860 EMSG_RET_NULL(_(e_invrange));
|
|
1861 #ifdef FEAT_MBYTE
|
|
1862 if (has_mbyte && ((*mb_char2len)(startc) > 1
|
|
1863 || (*mb_char2len)(endc) > 1))
|
|
1864 {
|
|
1865 /* Limit to a range of 256 chars */
|
|
1866 if (endc > startc + 256)
|
|
1867 EMSG_RET_NULL(_(e_invrange));
|
|
1868 while (++startc <= endc)
|
|
1869 regmbc(startc);
|
|
1870 }
|
|
1871 else
|
|
1872 #endif
|
|
1873 {
|
|
1874 #ifdef EBCDIC
|
|
1875 int alpha_only = FALSE;
|
|
1876
|
|
1877 /* for alphabetical range skip the gaps
|
|
1878 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */
|
|
1879 if (isalpha(startc) && isalpha(endc))
|
|
1880 alpha_only = TRUE;
|
|
1881 #endif
|
|
1882 while (++startc <= endc)
|
|
1883 #ifdef EBCDIC
|
|
1884 if (!alpha_only || isalpha(startc))
|
|
1885 #endif
|
|
1886 regc(startc);
|
|
1887 }
|
|
1888 startc = -1;
|
|
1889 }
|
|
1890 }
|
|
1891 /*
|
|
1892 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
|
|
1893 * accepts "\t", "\e", etc., but only when the 'l' flag in
|
|
1894 * 'cpoptions' is not included.
|
|
1895 */
|
|
1896 else if (*regparse == '\\'
|
|
1897 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
|
|
1898 || (!cpo_lit
|
|
1899 && vim_strchr(REGEXP_ABBR,
|
|
1900 regparse[1]) != NULL)))
|
|
1901 {
|
|
1902 regparse++;
|
|
1903 if (*regparse == 'n')
|
|
1904 {
|
|
1905 /* '\n' in range: also match NL */
|
|
1906 if (ret != JUST_CALC_SIZE)
|
|
1907 {
|
|
1908 if (*ret == ANYBUT)
|
|
1909 *ret = ANYBUT + ADD_NL;
|
|
1910 else if (*ret == ANYOF)
|
|
1911 *ret = ANYOF + ADD_NL;
|
|
1912 /* else: must have had a \n already */
|
|
1913 }
|
|
1914 *flagp |= HASNL;
|
|
1915 regparse++;
|
|
1916 startc = -1;
|
|
1917 }
|
24
|
1918 else if (*regparse == 'd'
|
|
1919 || *regparse == 'o'
|
|
1920 || *regparse == 'x'
|
|
1921 || *regparse == 'u'
|
|
1922 || *regparse == 'U')
|
|
1923 {
|
|
1924 startc = coll_get_char();
|
|
1925 if (startc == 0)
|
|
1926 regc(0x0a);
|
|
1927 else
|
|
1928 #ifdef FEAT_MBYTE
|
|
1929 regmbc(startc);
|
|
1930 #else
|
|
1931 regc(startc);
|
|
1932 #endif
|
|
1933 }
|
7
|
1934 else
|
|
1935 {
|
|
1936 startc = backslash_trans(*regparse++);
|
|
1937 regc(startc);
|
|
1938 }
|
|
1939 }
|
|
1940 else if (*regparse == '[')
|
|
1941 {
|
|
1942 int c_class;
|
|
1943 int cu;
|
|
1944
|
|
1945 c_class = skip_class_name(®parse);
|
|
1946 startc = -1;
|
|
1947 /* Characters assumed to be 8 bits! */
|
|
1948 switch (c_class)
|
|
1949 {
|
|
1950 case CLASS_NONE:
|
|
1951 /* literal '[', allow [[-x] as a range */
|
|
1952 startc = *regparse++;
|
|
1953 regc(startc);
|
|
1954 break;
|
|
1955 case CLASS_ALNUM:
|
|
1956 for (cu = 1; cu <= 255; cu++)
|
|
1957 if (isalnum(cu))
|
|
1958 regc(cu);
|
|
1959 break;
|
|
1960 case CLASS_ALPHA:
|
|
1961 for (cu = 1; cu <= 255; cu++)
|
|
1962 if (isalpha(cu))
|
|
1963 regc(cu);
|
|
1964 break;
|
|
1965 case CLASS_BLANK:
|
|
1966 regc(' ');
|
|
1967 regc('\t');
|
|
1968 break;
|
|
1969 case CLASS_CNTRL:
|
|
1970 for (cu = 1; cu <= 255; cu++)
|
|
1971 if (iscntrl(cu))
|
|
1972 regc(cu);
|
|
1973 break;
|
|
1974 case CLASS_DIGIT:
|
|
1975 for (cu = 1; cu <= 255; cu++)
|
|
1976 if (VIM_ISDIGIT(cu))
|
|
1977 regc(cu);
|
|
1978 break;
|
|
1979 case CLASS_GRAPH:
|
|
1980 for (cu = 1; cu <= 255; cu++)
|
|
1981 if (isgraph(cu))
|
|
1982 regc(cu);
|
|
1983 break;
|
|
1984 case CLASS_LOWER:
|
|
1985 for (cu = 1; cu <= 255; cu++)
|
|
1986 if (islower(cu))
|
|
1987 regc(cu);
|
|
1988 break;
|
|
1989 case CLASS_PRINT:
|
|
1990 for (cu = 1; cu <= 255; cu++)
|
|
1991 if (vim_isprintc(cu))
|
|
1992 regc(cu);
|
|
1993 break;
|
|
1994 case CLASS_PUNCT:
|
|
1995 for (cu = 1; cu <= 255; cu++)
|
|
1996 if (ispunct(cu))
|
|
1997 regc(cu);
|
|
1998 break;
|
|
1999 case CLASS_SPACE:
|
|
2000 for (cu = 9; cu <= 13; cu++)
|
|
2001 regc(cu);
|
|
2002 regc(' ');
|
|
2003 break;
|
|
2004 case CLASS_UPPER:
|
|
2005 for (cu = 1; cu <= 255; cu++)
|
|
2006 if (isupper(cu))
|
|
2007 regc(cu);
|
|
2008 break;
|
|
2009 case CLASS_XDIGIT:
|
|
2010 for (cu = 1; cu <= 255; cu++)
|
|
2011 if (vim_isxdigit(cu))
|
|
2012 regc(cu);
|
|
2013 break;
|
|
2014 case CLASS_TAB:
|
|
2015 regc('\t');
|
|
2016 break;
|
|
2017 case CLASS_RETURN:
|
|
2018 regc('\r');
|
|
2019 break;
|
|
2020 case CLASS_BACKSPACE:
|
|
2021 regc('\b');
|
|
2022 break;
|
|
2023 case CLASS_ESCAPE:
|
|
2024 regc('\033');
|
|
2025 break;
|
|
2026 }
|
|
2027 }
|
|
2028 else
|
|
2029 {
|
|
2030 #ifdef FEAT_MBYTE
|
|
2031 if (has_mbyte)
|
|
2032 {
|
|
2033 int len;
|
|
2034
|
|
2035 /* produce a multibyte character, including any
|
|
2036 * following composing characters */
|
|
2037 startc = mb_ptr2char(regparse);
|
|
2038 len = (*mb_ptr2len_check)(regparse);
|
|
2039 if (enc_utf8 && utf_char2len(startc) != len)
|
|
2040 startc = -1; /* composing chars */
|
|
2041 while (--len >= 0)
|
|
2042 regc(*regparse++);
|
|
2043 }
|
|
2044 else
|
|
2045 #endif
|
|
2046 {
|
|
2047 startc = *regparse++;
|
|
2048 regc(startc);
|
|
2049 }
|
|
2050 }
|
|
2051 }
|
|
2052 regc(NUL);
|
|
2053 prevchr_len = 1; /* last char was the ']' */
|
|
2054 if (*regparse != ']')
|
|
2055 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */
|
|
2056 skipchr(); /* let's be friends with the lexer again */
|
|
2057 *flagp |= HASWIDTH | SIMPLE;
|
|
2058 break;
|
|
2059 }
|
|
2060 }
|
|
2061 /* FALLTHROUGH */
|
|
2062
|
|
2063 default:
|
|
2064 {
|
|
2065 int len;
|
|
2066
|
|
2067 #ifdef FEAT_MBYTE
|
|
2068 /* A multi-byte character is handled as a separate atom if it's
|
|
2069 * before a multi. */
|
|
2070 if (has_mbyte && (*mb_char2len)(c) > 1
|
|
2071 && re_multi_type(peekchr()) != NOT_MULTI)
|
|
2072 {
|
|
2073 ret = regnode(MULTIBYTECODE);
|
|
2074 regmbc(c);
|
|
2075 *flagp |= HASWIDTH | SIMPLE;
|
|
2076 break;
|
|
2077 }
|
|
2078 #endif
|
|
2079
|
|
2080 ret = regnode(EXACTLY);
|
|
2081
|
|
2082 /*
|
|
2083 * Append characters as long as:
|
|
2084 * - there is no following multi, we then need the character in
|
|
2085 * front of it as a single character operand
|
|
2086 * - not running into a Magic character
|
|
2087 * - "one_exactly" is not set
|
|
2088 * But always emit at least one character. Might be a Multi,
|
|
2089 * e.g., a "[" without matching "]".
|
|
2090 */
|
|
2091 for (len = 0; c != NUL && (len == 0
|
|
2092 || (re_multi_type(peekchr()) == NOT_MULTI
|
|
2093 && !one_exactly
|
|
2094 && !is_Magic(c))); ++len)
|
|
2095 {
|
|
2096 c = no_Magic(c);
|
|
2097 #ifdef FEAT_MBYTE
|
|
2098 if (has_mbyte)
|
|
2099 {
|
|
2100 regmbc(c);
|
|
2101 if (enc_utf8)
|
|
2102 {
|
|
2103 int off;
|
|
2104 int l;
|
|
2105
|
|
2106 /* Need to get composing character too, directly
|
|
2107 * access regparse for that, because skipchr() skips
|
|
2108 * over composing chars. */
|
|
2109 ungetchr();
|
|
2110 if (*regparse == '\\' && regparse[1] != NUL)
|
|
2111 off = 1;
|
|
2112 else
|
|
2113 off = 0;
|
|
2114 for (;;)
|
|
2115 {
|
|
2116 l = utf_ptr2len_check(regparse + off);
|
|
2117 if (!UTF_COMPOSINGLIKE(regparse + off,
|
|
2118 regparse + off + l))
|
|
2119 break;
|
|
2120 off += l;
|
|
2121 regmbc(utf_ptr2char(regparse + off));
|
|
2122 }
|
|
2123 skipchr();
|
|
2124 }
|
|
2125 }
|
|
2126 else
|
|
2127 #endif
|
|
2128 regc(c);
|
|
2129 c = getchr();
|
|
2130 }
|
|
2131 ungetchr();
|
|
2132
|
|
2133 regc(NUL);
|
|
2134 *flagp |= HASWIDTH;
|
|
2135 if (len == 1)
|
|
2136 *flagp |= SIMPLE;
|
|
2137 }
|
|
2138 break;
|
|
2139 }
|
|
2140
|
|
2141 return ret;
|
|
2142 }
|
|
2143
|
|
2144 /*
|
|
2145 * emit a node
|
|
2146 * Return pointer to generated code.
|
|
2147 */
|
|
2148 static char_u *
|
|
2149 regnode(op)
|
|
2150 int op;
|
|
2151 {
|
|
2152 char_u *ret;
|
|
2153
|
|
2154 ret = regcode;
|
|
2155 if (ret == JUST_CALC_SIZE)
|
|
2156 regsize += 3;
|
|
2157 else
|
|
2158 {
|
|
2159 *regcode++ = op;
|
|
2160 *regcode++ = NUL; /* Null "next" pointer. */
|
|
2161 *regcode++ = NUL;
|
|
2162 }
|
|
2163 return ret;
|
|
2164 }
|
|
2165
|
|
2166 /*
|
|
2167 * Emit (if appropriate) a byte of code
|
|
2168 */
|
|
2169 static void
|
|
2170 regc(b)
|
|
2171 int b;
|
|
2172 {
|
|
2173 if (regcode == JUST_CALC_SIZE)
|
|
2174 regsize++;
|
|
2175 else
|
|
2176 *regcode++ = b;
|
|
2177 }
|
|
2178
|
|
2179 #ifdef FEAT_MBYTE
|
|
2180 /*
|
|
2181 * Emit (if appropriate) a multi-byte character of code
|
|
2182 */
|
|
2183 static void
|
|
2184 regmbc(c)
|
|
2185 int c;
|
|
2186 {
|
|
2187 if (regcode == JUST_CALC_SIZE)
|
|
2188 regsize += (*mb_char2len)(c);
|
|
2189 else
|
|
2190 regcode += (*mb_char2bytes)(c, regcode);
|
|
2191 }
|
|
2192 #endif
|
|
2193
|
|
2194 /*
|
|
2195 * reginsert - insert an operator in front of already-emitted operand
|
|
2196 *
|
|
2197 * Means relocating the operand.
|
|
2198 */
|
|
2199 static void
|
|
2200 reginsert(op, opnd)
|
|
2201 int op;
|
|
2202 char_u *opnd;
|
|
2203 {
|
|
2204 char_u *src;
|
|
2205 char_u *dst;
|
|
2206 char_u *place;
|
|
2207
|
|
2208 if (regcode == JUST_CALC_SIZE)
|
|
2209 {
|
|
2210 regsize += 3;
|
|
2211 return;
|
|
2212 }
|
|
2213 src = regcode;
|
|
2214 regcode += 3;
|
|
2215 dst = regcode;
|
|
2216 while (src > opnd)
|
|
2217 *--dst = *--src;
|
|
2218
|
|
2219 place = opnd; /* Op node, where operand used to be. */
|
|
2220 *place++ = op;
|
|
2221 *place++ = NUL;
|
|
2222 *place = NUL;
|
|
2223 }
|
|
2224
|
|
2225 /*
|
|
2226 * reginsert_limits - insert an operator in front of already-emitted operand.
|
|
2227 * The operator has the given limit values as operands. Also set next pointer.
|
|
2228 *
|
|
2229 * Means relocating the operand.
|
|
2230 */
|
|
2231 static void
|
|
2232 reginsert_limits(op, minval, maxval, opnd)
|
|
2233 int op;
|
|
2234 long minval;
|
|
2235 long maxval;
|
|
2236 char_u *opnd;
|
|
2237 {
|
|
2238 char_u *src;
|
|
2239 char_u *dst;
|
|
2240 char_u *place;
|
|
2241
|
|
2242 if (regcode == JUST_CALC_SIZE)
|
|
2243 {
|
|
2244 regsize += 11;
|
|
2245 return;
|
|
2246 }
|
|
2247 src = regcode;
|
|
2248 regcode += 11;
|
|
2249 dst = regcode;
|
|
2250 while (src > opnd)
|
|
2251 *--dst = *--src;
|
|
2252
|
|
2253 place = opnd; /* Op node, where operand used to be. */
|
|
2254 *place++ = op;
|
|
2255 *place++ = NUL;
|
|
2256 *place++ = NUL;
|
|
2257 place = re_put_long(place, (long_u)minval);
|
|
2258 place = re_put_long(place, (long_u)maxval);
|
|
2259 regtail(opnd, place);
|
|
2260 }
|
|
2261
|
|
2262 /*
|
|
2263 * Write a long as four bytes at "p" and return pointer to the next char.
|
|
2264 */
|
|
2265 static char_u *
|
|
2266 re_put_long(p, val)
|
|
2267 char_u *p;
|
|
2268 long_u val;
|
|
2269 {
|
|
2270 *p++ = (char_u) ((val >> 24) & 0377);
|
|
2271 *p++ = (char_u) ((val >> 16) & 0377);
|
|
2272 *p++ = (char_u) ((val >> 8) & 0377);
|
|
2273 *p++ = (char_u) (val & 0377);
|
|
2274 return p;
|
|
2275 }
|
|
2276
|
|
2277 /*
|
|
2278 * regtail - set the next-pointer at the end of a node chain
|
|
2279 */
|
|
2280 static void
|
|
2281 regtail(p, val)
|
|
2282 char_u *p;
|
|
2283 char_u *val;
|
|
2284 {
|
|
2285 char_u *scan;
|
|
2286 char_u *temp;
|
|
2287 int offset;
|
|
2288
|
|
2289 if (p == JUST_CALC_SIZE)
|
|
2290 return;
|
|
2291
|
|
2292 /* Find last node. */
|
|
2293 scan = p;
|
|
2294 for (;;)
|
|
2295 {
|
|
2296 temp = regnext(scan);
|
|
2297 if (temp == NULL)
|
|
2298 break;
|
|
2299 scan = temp;
|
|
2300 }
|
|
2301
|
|
2302 if (OP(scan) == BACK)
|
|
2303 offset = (int)(scan - val);
|
|
2304 else
|
|
2305 offset = (int)(val - scan);
|
|
2306 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377);
|
|
2307 *(scan + 2) = (char_u) (offset & 0377);
|
|
2308 }
|
|
2309
|
|
2310 /*
|
|
2311 * regoptail - regtail on item after a BRANCH; nop if none
|
|
2312 */
|
|
2313 static void
|
|
2314 regoptail(p, val)
|
|
2315 char_u *p;
|
|
2316 char_u *val;
|
|
2317 {
|
|
2318 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */
|
|
2319 if (p == NULL || p == JUST_CALC_SIZE
|
|
2320 || (OP(p) != BRANCH
|
|
2321 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9)))
|
|
2322 return;
|
|
2323 regtail(OPERAND(p), val);
|
|
2324 }
|
|
2325
|
|
2326 /*
|
|
2327 * getchr() - get the next character from the pattern. We know about
|
|
2328 * magic and such, so therefore we need a lexical analyzer.
|
|
2329 */
|
|
2330
|
|
2331 /* static int curchr; */
|
|
2332 static int prevprevchr;
|
|
2333 static int prevchr;
|
|
2334 static int nextchr; /* used for ungetchr() */
|
|
2335 /*
|
|
2336 * Note: prevchr is sometimes -1 when we are not at the start,
|
|
2337 * eg in /[ ^I]^ the pattern was never found even if it existed, because ^ was
|
|
2338 * taken to be magic -- webb
|
|
2339 */
|
|
2340 static int at_start; /* True when on the first character */
|
|
2341 static int prev_at_start; /* True when on the second character */
|
|
2342
|
|
2343 static void
|
|
2344 initchr(str)
|
|
2345 char_u *str;
|
|
2346 {
|
|
2347 regparse = str;
|
|
2348 prevchr_len = 0;
|
|
2349 curchr = prevprevchr = prevchr = nextchr = -1;
|
|
2350 at_start = TRUE;
|
|
2351 prev_at_start = FALSE;
|
|
2352 }
|
|
2353
|
|
2354 static int
|
|
2355 peekchr()
|
|
2356 {
|
|
2357 if (curchr == -1)
|
|
2358 {
|
|
2359 switch (curchr = regparse[0])
|
|
2360 {
|
|
2361 case '.':
|
|
2362 case '[':
|
|
2363 case '~':
|
|
2364 /* magic when 'magic' is on */
|
|
2365 if (reg_magic >= MAGIC_ON)
|
|
2366 curchr = Magic(curchr);
|
|
2367 break;
|
|
2368 case '(':
|
|
2369 case ')':
|
|
2370 case '{':
|
|
2371 case '%':
|
|
2372 case '+':
|
|
2373 case '=':
|
|
2374 case '?':
|
|
2375 case '@':
|
|
2376 case '!':
|
|
2377 case '&':
|
|
2378 case '|':
|
|
2379 case '<':
|
|
2380 case '>':
|
|
2381 case '#': /* future ext. */
|
|
2382 case '"': /* future ext. */
|
|
2383 case '\'': /* future ext. */
|
|
2384 case ',': /* future ext. */
|
|
2385 case '-': /* future ext. */
|
|
2386 case ':': /* future ext. */
|
|
2387 case ';': /* future ext. */
|
|
2388 case '`': /* future ext. */
|
|
2389 case '/': /* Can't be used in / command */
|
|
2390 /* magic only after "\v" */
|
|
2391 if (reg_magic == MAGIC_ALL)
|
|
2392 curchr = Magic(curchr);
|
|
2393 break;
|
|
2394 case '*':
|
|
2395 /* * is not magic as the very first character, eg "?*ptr" and when
|
|
2396 * after '^', eg "/^*ptr" */
|
|
2397 if (reg_magic >= MAGIC_ON && !at_start
|
|
2398 && !(prev_at_start && prevchr == Magic('^')))
|
|
2399 curchr = Magic('*');
|
|
2400 break;
|
|
2401 case '^':
|
|
2402 /* '^' is only magic as the very first character and if it's after
|
|
2403 * "\(", "\|", "\&' or "\n" */
|
|
2404 if (reg_magic >= MAGIC_OFF
|
|
2405 && (at_start
|
|
2406 || reg_magic == MAGIC_ALL
|
|
2407 || prevchr == Magic('(')
|
|
2408 || prevchr == Magic('|')
|
|
2409 || prevchr == Magic('&')
|
|
2410 || prevchr == Magic('n')
|
|
2411 || (no_Magic(prevchr) == '('
|
|
2412 && prevprevchr == Magic('%'))))
|
|
2413 {
|
|
2414 curchr = Magic('^');
|
|
2415 at_start = TRUE;
|
|
2416 prev_at_start = FALSE;
|
|
2417 }
|
|
2418 break;
|
|
2419 case '$':
|
|
2420 /* '$' is only magic as the very last char and if it's in front of
|
|
2421 * either "\|", "\)", "\&", or "\n" */
|
|
2422 if (reg_magic >= MAGIC_OFF)
|
|
2423 {
|
|
2424 char_u *p = regparse + 1;
|
|
2425
|
|
2426 /* ignore \c \C \m and \M after '$' */
|
|
2427 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C'
|
|
2428 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z'))
|
|
2429 p += 2;
|
|
2430 if (p[0] == NUL
|
|
2431 || (p[0] == '\\'
|
|
2432 && (p[1] == '|' || p[1] == '&' || p[1] == ')'
|
|
2433 || p[1] == 'n'))
|
|
2434 || reg_magic == MAGIC_ALL)
|
|
2435 curchr = Magic('$');
|
|
2436 }
|
|
2437 break;
|
|
2438 case '\\':
|
|
2439 {
|
|
2440 int c = regparse[1];
|
|
2441
|
|
2442 if (c == NUL)
|
|
2443 curchr = '\\'; /* trailing '\' */
|
|
2444 else if (
|
|
2445 #ifdef EBCDIC
|
|
2446 vim_strchr(META, c)
|
|
2447 #else
|
|
2448 c <= '~' && META_flags[c]
|
|
2449 #endif
|
|
2450 )
|
|
2451 {
|
|
2452 /*
|
|
2453 * META contains everything that may be magic sometimes,
|
|
2454 * except ^ and $ ("\^" and "\$" are only magic after
|
|
2455 * "\v"). We now fetch the next character and toggle its
|
|
2456 * magicness. Therefore, \ is so meta-magic that it is
|
|
2457 * not in META.
|
|
2458 */
|
|
2459 curchr = -1;
|
|
2460 prev_at_start = at_start;
|
|
2461 at_start = FALSE; /* be able to say "/\*ptr" */
|
|
2462 ++regparse;
|
|
2463 peekchr();
|
|
2464 --regparse;
|
|
2465 curchr = toggle_Magic(curchr);
|
|
2466 }
|
|
2467 else if (vim_strchr(REGEXP_ABBR, c))
|
|
2468 {
|
|
2469 /*
|
|
2470 * Handle abbreviations, like "\t" for TAB -- webb
|
|
2471 */
|
|
2472 curchr = backslash_trans(c);
|
|
2473 }
|
|
2474 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^'))
|
|
2475 curchr = toggle_Magic(c);
|
|
2476 else
|
|
2477 {
|
|
2478 /*
|
|
2479 * Next character can never be (made) magic?
|
|
2480 * Then backslashing it won't do anything.
|
|
2481 */
|
|
2482 #ifdef FEAT_MBYTE
|
|
2483 if (has_mbyte)
|
|
2484 curchr = (*mb_ptr2char)(regparse + 1);
|
|
2485 else
|
|
2486 #endif
|
|
2487 curchr = c;
|
|
2488 }
|
|
2489 break;
|
|
2490 }
|
|
2491
|
|
2492 #ifdef FEAT_MBYTE
|
|
2493 default:
|
|
2494 if (has_mbyte)
|
|
2495 curchr = (*mb_ptr2char)(regparse);
|
|
2496 #endif
|
|
2497 }
|
|
2498 }
|
|
2499
|
|
2500 return curchr;
|
|
2501 }
|
|
2502
|
|
2503 /*
|
|
2504 * Eat one lexed character. Do this in a way that we can undo it.
|
|
2505 */
|
|
2506 static void
|
|
2507 skipchr()
|
|
2508 {
|
|
2509 /* peekchr() eats a backslash, do the same here */
|
|
2510 if (*regparse == '\\')
|
|
2511 prevchr_len = 1;
|
|
2512 else
|
|
2513 prevchr_len = 0;
|
|
2514 if (regparse[prevchr_len] != NUL)
|
|
2515 {
|
|
2516 #ifdef FEAT_MBYTE
|
|
2517 if (has_mbyte)
|
|
2518 prevchr_len += (*mb_ptr2len_check)(regparse + prevchr_len);
|
|
2519 else
|
|
2520 #endif
|
|
2521 ++prevchr_len;
|
|
2522 }
|
|
2523 regparse += prevchr_len;
|
|
2524 prev_at_start = at_start;
|
|
2525 at_start = FALSE;
|
|
2526 prevprevchr = prevchr;
|
|
2527 prevchr = curchr;
|
|
2528 curchr = nextchr; /* use previously unget char, or -1 */
|
|
2529 nextchr = -1;
|
|
2530 }
|
|
2531
|
|
2532 /*
|
|
2533 * Skip a character while keeping the value of prev_at_start for at_start.
|
|
2534 * prevchr and prevprevchr are also kept.
|
|
2535 */
|
|
2536 static void
|
|
2537 skipchr_keepstart()
|
|
2538 {
|
|
2539 int as = prev_at_start;
|
|
2540 int pr = prevchr;
|
|
2541 int prpr = prevprevchr;
|
|
2542
|
|
2543 skipchr();
|
|
2544 at_start = as;
|
|
2545 prevchr = pr;
|
|
2546 prevprevchr = prpr;
|
|
2547 }
|
|
2548
|
|
2549 static int
|
|
2550 getchr()
|
|
2551 {
|
|
2552 int chr = peekchr();
|
|
2553
|
|
2554 skipchr();
|
|
2555 return chr;
|
|
2556 }
|
|
2557
|
|
2558 /*
|
|
2559 * put character back. Works only once!
|
|
2560 */
|
|
2561 static void
|
|
2562 ungetchr()
|
|
2563 {
|
|
2564 nextchr = curchr;
|
|
2565 curchr = prevchr;
|
|
2566 prevchr = prevprevchr;
|
|
2567 at_start = prev_at_start;
|
|
2568 prev_at_start = FALSE;
|
|
2569
|
|
2570 /* Backup regparse, so that it's at the same position as before the
|
|
2571 * getchr(). */
|
|
2572 regparse -= prevchr_len;
|
|
2573 }
|
|
2574
|
|
2575 /*
|
29
|
2576 * Get and return the value of the hex string at the current position.
|
|
2577 * Return -1 if there is no valid hex number.
|
|
2578 * The position is updated:
|
24
|
2579 * blahblah\%x20asdf
|
|
2580 * before-^ ^-after
|
|
2581 * The parameter controls the maximum number of input characters. This will be
|
|
2582 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence.
|
|
2583 */
|
|
2584 static int
|
|
2585 gethexchrs(maxinputlen)
|
|
2586 int maxinputlen;
|
|
2587 {
|
|
2588 int nr = 0;
|
|
2589 int c;
|
|
2590 int i;
|
|
2591
|
|
2592 for (i = 0; i < maxinputlen; ++i)
|
|
2593 {
|
|
2594 c = regparse[0];
|
|
2595 if (!vim_isxdigit(c))
|
|
2596 break;
|
|
2597 nr <<= 4;
|
|
2598 nr |= hex2nr(c);
|
|
2599 ++regparse;
|
|
2600 }
|
|
2601
|
|
2602 if (i == 0)
|
|
2603 return -1;
|
|
2604 return nr;
|
|
2605 }
|
|
2606
|
|
2607 /*
|
|
2608 * get and return the value of the decimal string immediately after the
|
|
2609 * current position. Return -1 for invalid. Consumes all digits.
|
|
2610 */
|
|
2611 static int
|
|
2612 getdecchrs()
|
|
2613 {
|
|
2614 int nr = 0;
|
|
2615 int c;
|
|
2616 int i;
|
|
2617
|
|
2618 for (i = 0; ; ++i)
|
|
2619 {
|
|
2620 c = regparse[0];
|
|
2621 if (c < '0' || c > '9')
|
|
2622 break;
|
|
2623 nr *= 10;
|
|
2624 nr += c - '0';
|
|
2625 ++regparse;
|
|
2626 }
|
|
2627
|
|
2628 if (i == 0)
|
|
2629 return -1;
|
|
2630 return nr;
|
|
2631 }
|
|
2632
|
|
2633 /*
|
|
2634 * get and return the value of the octal string immediately after the current
|
|
2635 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle
|
|
2636 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't
|
|
2637 * treat 8 or 9 as recognised characters. Position is updated:
|
|
2638 * blahblah\%o210asdf
|
|
2639 * before-^ ^-after
|
|
2640 */
|
|
2641 static int
|
|
2642 getoctchrs()
|
|
2643 {
|
|
2644 int nr = 0;
|
|
2645 int c;
|
|
2646 int i;
|
|
2647
|
|
2648 for (i = 0; i < 3 && nr < 040; ++i)
|
|
2649 {
|
|
2650 c = regparse[0];
|
|
2651 if (c < '0' || c > '7')
|
|
2652 break;
|
|
2653 nr <<= 3;
|
|
2654 nr |= hex2nr(c);
|
|
2655 ++regparse;
|
|
2656 }
|
|
2657
|
|
2658 if (i == 0)
|
|
2659 return -1;
|
|
2660 return nr;
|
|
2661 }
|
|
2662
|
|
2663 /*
|
|
2664 * Get a number after a backslash that is inside [].
|
|
2665 * When nothing is recognized return a backslash.
|
|
2666 */
|
|
2667 static int
|
|
2668 coll_get_char()
|
|
2669 {
|
|
2670 int nr = -1;
|
|
2671
|
|
2672 switch (*regparse++)
|
|
2673 {
|
|
2674 case 'd': nr = getdecchrs(); break;
|
|
2675 case 'o': nr = getoctchrs(); break;
|
|
2676 case 'x': nr = gethexchrs(2); break;
|
|
2677 case 'u': nr = gethexchrs(4); break;
|
|
2678 case 'U': nr = gethexchrs(8); break;
|
|
2679 }
|
|
2680 if (nr < 0)
|
|
2681 {
|
|
2682 /* If getting the number fails be backwards compatible: the character
|
|
2683 * is a backslash. */
|
|
2684 --regparse;
|
|
2685 nr = '\\';
|
|
2686 }
|
|
2687 return nr;
|
|
2688 }
|
|
2689
|
|
2690 /*
|
7
|
2691 * read_limits - Read two integers to be taken as a minimum and maximum.
|
|
2692 * If the first character is '-', then the range is reversed.
|
|
2693 * Should end with 'end'. If minval is missing, zero is default, if maxval is
|
|
2694 * missing, a very big number is the default.
|
|
2695 */
|
|
2696 static int
|
|
2697 read_limits(minval, maxval)
|
|
2698 long *minval;
|
|
2699 long *maxval;
|
|
2700 {
|
|
2701 int reverse = FALSE;
|
|
2702 char_u *first_char;
|
|
2703 long tmp;
|
|
2704
|
|
2705 if (*regparse == '-')
|
|
2706 {
|
|
2707 /* Starts with '-', so reverse the range later */
|
|
2708 regparse++;
|
|
2709 reverse = TRUE;
|
|
2710 }
|
|
2711 first_char = regparse;
|
|
2712 *minval = getdigits(®parse);
|
|
2713 if (*regparse == ',') /* There is a comma */
|
|
2714 {
|
|
2715 if (vim_isdigit(*++regparse))
|
|
2716 *maxval = getdigits(®parse);
|
|
2717 else
|
|
2718 *maxval = MAX_LIMIT;
|
|
2719 }
|
|
2720 else if (VIM_ISDIGIT(*first_char))
|
|
2721 *maxval = *minval; /* It was \{n} or \{-n} */
|
|
2722 else
|
|
2723 *maxval = MAX_LIMIT; /* It was \{} or \{-} */
|
|
2724 if (*regparse == '\\')
|
|
2725 regparse++; /* Allow either \{...} or \{...\} */
|
|
2726 if (*regparse != '}' || (*maxval == 0 && *minval == 0))
|
|
2727 {
|
|
2728 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"),
|
|
2729 reg_magic == MAGIC_ALL ? "" : "\\");
|
|
2730 EMSG_RET_FAIL(IObuff);
|
|
2731 }
|
|
2732
|
|
2733 /*
|
|
2734 * Reverse the range if there was a '-', or make sure it is in the right
|
|
2735 * order otherwise.
|
|
2736 */
|
|
2737 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval))
|
|
2738 {
|
|
2739 tmp = *minval;
|
|
2740 *minval = *maxval;
|
|
2741 *maxval = tmp;
|
|
2742 }
|
|
2743 skipchr(); /* let's be friends with the lexer again */
|
|
2744 return OK;
|
|
2745 }
|
|
2746
|
|
2747 /*
|
|
2748 * vim_regexec and friends
|
|
2749 */
|
|
2750
|
|
2751 /*
|
|
2752 * Global work variables for vim_regexec().
|
|
2753 */
|
|
2754
|
|
2755 /* The current match-position is remembered with these variables: */
|
|
2756 static linenr_T reglnum; /* line number, relative to first line */
|
|
2757 static char_u *regline; /* start of current line */
|
|
2758 static char_u *reginput; /* current input, points into "regline" */
|
|
2759
|
|
2760 static int need_clear_subexpr; /* subexpressions still need to be
|
|
2761 * cleared */
|
|
2762 #ifdef FEAT_SYN_HL
|
|
2763 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions
|
|
2764 * still need to be cleared */
|
|
2765 #endif
|
|
2766
|
|
2767 static int out_of_stack; /* TRUE when ran out of stack space */
|
|
2768
|
|
2769 /*
|
|
2770 * Structure used to save the current input state, when it needs to be
|
|
2771 * restored after trying a match. Used by reg_save() and reg_restore().
|
|
2772 */
|
|
2773 typedef struct
|
|
2774 {
|
|
2775 union
|
|
2776 {
|
|
2777 char_u *ptr; /* reginput pointer, for single-line regexp */
|
|
2778 lpos_T pos; /* reginput pos, for multi-line regexp */
|
|
2779 } rs_u;
|
|
2780 } regsave_T;
|
|
2781
|
|
2782 /* struct to save start/end pointer/position in for \(\) */
|
|
2783 typedef struct
|
|
2784 {
|
|
2785 union
|
|
2786 {
|
|
2787 char_u *ptr;
|
|
2788 lpos_T pos;
|
|
2789 } se_u;
|
|
2790 } save_se_T;
|
|
2791
|
|
2792 static char_u *reg_getline __ARGS((linenr_T lnum));
|
|
2793 static long vim_regexec_both __ARGS((char_u *line, colnr_T col));
|
|
2794 static long regtry __ARGS((regprog_T *prog, colnr_T col));
|
|
2795 static void cleanup_subexpr __ARGS((void));
|
|
2796 #ifdef FEAT_SYN_HL
|
|
2797 static void cleanup_zsubexpr __ARGS((void));
|
|
2798 #endif
|
|
2799 static void reg_nextline __ARGS((void));
|
|
2800 static void reg_save __ARGS((regsave_T *save));
|
|
2801 static void reg_restore __ARGS((regsave_T *save));
|
|
2802 static int reg_save_equal __ARGS((regsave_T *save));
|
|
2803 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp));
|
|
2804 static void save_se_one __ARGS((save_se_T *savep, char_u **pp));
|
|
2805
|
|
2806 /* Save the sub-expressions before attempting a match. */
|
|
2807 #define save_se(savep, posp, pp) \
|
|
2808 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
|
|
2809
|
|
2810 /* After a failed match restore the sub-expressions. */
|
|
2811 #define restore_se(savep, posp, pp) { \
|
|
2812 if (REG_MULTI) \
|
|
2813 *(posp) = (savep)->se_u.pos; \
|
|
2814 else \
|
|
2815 *(pp) = (savep)->se_u.ptr; }
|
|
2816
|
|
2817 static int re_num_cmp __ARGS((long_u val, char_u *scan));
|
|
2818 static int regmatch __ARGS((char_u *prog));
|
|
2819 static int regrepeat __ARGS((char_u *p, long maxcount));
|
|
2820
|
|
2821 #ifdef DEBUG
|
|
2822 int regnarrate = 0;
|
|
2823 #endif
|
|
2824
|
|
2825 /*
|
|
2826 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec().
|
|
2827 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern
|
|
2828 * contains '\c' or '\C' the value is overruled.
|
|
2829 */
|
|
2830 static int ireg_ic;
|
|
2831
|
|
2832 #ifdef FEAT_MBYTE
|
|
2833 /*
|
|
2834 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag
|
|
2835 * in the regexp. Defaults to false, always.
|
|
2836 */
|
|
2837 static int ireg_icombine;
|
|
2838 #endif
|
|
2839
|
|
2840 /*
|
|
2841 * Sometimes need to save a copy of a line. Since alloc()/free() is very
|
|
2842 * slow, we keep one allocated piece of memory and only re-allocate it when
|
|
2843 * it's too small. It's freed in vim_regexec_both() when finished.
|
|
2844 */
|
|
2845 static char_u *reg_tofree;
|
|
2846 static unsigned reg_tofreelen;
|
|
2847
|
|
2848 /*
|
|
2849 * These variables are set when executing a regexp to speed up the execution.
|
|
2850 * Which ones are set depends on whethere a single-line or multi-line match is
|
|
2851 * done:
|
|
2852 * single-line multi-line
|
|
2853 * reg_match ®match_T NULL
|
|
2854 * reg_mmatch NULL ®mmatch_T
|
|
2855 * reg_startp reg_match->startp <invalid>
|
|
2856 * reg_endp reg_match->endp <invalid>
|
|
2857 * reg_startpos <invalid> reg_mmatch->startpos
|
|
2858 * reg_endpos <invalid> reg_mmatch->endpos
|
|
2859 * reg_win NULL window in which to search
|
|
2860 * reg_buf <invalid> buffer in which to search
|
|
2861 * reg_firstlnum <invalid> first line in which to search
|
|
2862 * reg_maxline 0 last line nr
|
|
2863 * reg_line_lbr FALSE or TRUE FALSE
|
|
2864 */
|
|
2865 static regmatch_T *reg_match;
|
|
2866 static regmmatch_T *reg_mmatch;
|
|
2867 static char_u **reg_startp = NULL;
|
|
2868 static char_u **reg_endp = NULL;
|
|
2869 static lpos_T *reg_startpos = NULL;
|
|
2870 static lpos_T *reg_endpos = NULL;
|
|
2871 static win_T *reg_win;
|
|
2872 static buf_T *reg_buf;
|
|
2873 static linenr_T reg_firstlnum;
|
|
2874 static linenr_T reg_maxline;
|
|
2875 static int reg_line_lbr; /* "\n" in string is line break */
|
|
2876
|
|
2877 /*
|
|
2878 * Get pointer to the line "lnum", which is relative to "reg_firstlnum".
|
|
2879 */
|
|
2880 static char_u *
|
|
2881 reg_getline(lnum)
|
|
2882 linenr_T lnum;
|
|
2883 {
|
|
2884 /* when looking behind for a match/no-match lnum is negative. But we
|
|
2885 * can't go before line 1 */
|
|
2886 if (reg_firstlnum + lnum < 1)
|
|
2887 return NULL;
|
|
2888 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE);
|
|
2889 }
|
|
2890
|
|
2891 static regsave_T behind_pos;
|
|
2892
|
|
2893 #ifdef FEAT_SYN_HL
|
|
2894 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */
|
|
2895 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */
|
|
2896 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */
|
|
2897 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */
|
|
2898 #endif
|
|
2899
|
|
2900 /* TRUE if using multi-line regexp. */
|
|
2901 #define REG_MULTI (reg_match == NULL)
|
|
2902
|
|
2903 /*
|
|
2904 * Match a regexp against a string.
|
|
2905 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
2906 * Uses curbuf for line count and 'iskeyword'.
|
|
2907 *
|
|
2908 * Return TRUE if there is a match, FALSE if not.
|
|
2909 */
|
|
2910 int
|
|
2911 vim_regexec(rmp, line, col)
|
|
2912 regmatch_T *rmp;
|
|
2913 char_u *line; /* string to match against */
|
|
2914 colnr_T col; /* column to start looking for match */
|
|
2915 {
|
|
2916 reg_match = rmp;
|
|
2917 reg_mmatch = NULL;
|
|
2918 reg_maxline = 0;
|
|
2919 reg_line_lbr = FALSE;
|
|
2920 reg_win = NULL;
|
|
2921 ireg_ic = rmp->rm_ic;
|
|
2922 #ifdef FEAT_MBYTE
|
|
2923 ireg_icombine = FALSE;
|
|
2924 #endif
|
|
2925 return (vim_regexec_both(line, col) != 0);
|
|
2926 }
|
|
2927
|
39
|
2928 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \
|
|
2929 || defined(FIND_REPLACE_DIALOG) || defined(PROTO)
|
7
|
2930 /*
|
|
2931 * Like vim_regexec(), but consider a "\n" in "line" to be a line break.
|
|
2932 */
|
|
2933 int
|
|
2934 vim_regexec_nl(rmp, line, col)
|
|
2935 regmatch_T *rmp;
|
|
2936 char_u *line; /* string to match against */
|
|
2937 colnr_T col; /* column to start looking for match */
|
|
2938 {
|
|
2939 reg_match = rmp;
|
|
2940 reg_mmatch = NULL;
|
|
2941 reg_maxline = 0;
|
|
2942 reg_line_lbr = TRUE;
|
|
2943 reg_win = NULL;
|
|
2944 ireg_ic = rmp->rm_ic;
|
|
2945 #ifdef FEAT_MBYTE
|
|
2946 ireg_icombine = FALSE;
|
|
2947 #endif
|
|
2948 return (vim_regexec_both(line, col) != 0);
|
|
2949 }
|
|
2950 #endif
|
|
2951
|
|
2952 /*
|
|
2953 * Match a regexp against multiple lines.
|
|
2954 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
2955 * Uses curbuf for line count and 'iskeyword'.
|
|
2956 *
|
|
2957 * Return zero if there is no match. Return number of lines contained in the
|
|
2958 * match otherwise.
|
|
2959 */
|
|
2960 long
|
|
2961 vim_regexec_multi(rmp, win, buf, lnum, col)
|
|
2962 regmmatch_T *rmp;
|
|
2963 win_T *win; /* window in which to search or NULL */
|
|
2964 buf_T *buf; /* buffer in which to search */
|
|
2965 linenr_T lnum; /* nr of line to start looking for match */
|
|
2966 colnr_T col; /* column to start looking for match */
|
|
2967 {
|
|
2968 long r;
|
|
2969 buf_T *save_curbuf = curbuf;
|
|
2970
|
|
2971 reg_match = NULL;
|
|
2972 reg_mmatch = rmp;
|
|
2973 reg_buf = buf;
|
|
2974 reg_win = win;
|
|
2975 reg_firstlnum = lnum;
|
|
2976 reg_maxline = reg_buf->b_ml.ml_line_count - lnum;
|
|
2977 reg_line_lbr = FALSE;
|
|
2978 ireg_ic = rmp->rmm_ic;
|
|
2979 #ifdef FEAT_MBYTE
|
|
2980 ireg_icombine = FALSE;
|
|
2981 #endif
|
|
2982
|
|
2983 /* Need to switch to buffer "buf" to make vim_iswordc() work. */
|
|
2984 curbuf = buf;
|
|
2985 r = vim_regexec_both(NULL, col);
|
|
2986 curbuf = save_curbuf;
|
|
2987
|
|
2988 return r;
|
|
2989 }
|
|
2990
|
|
2991 /*
|
|
2992 * Match a regexp against a string ("line" points to the string) or multiple
|
|
2993 * lines ("line" is NULL, use reg_getline()).
|
|
2994 */
|
|
2995 #ifdef HAVE_SETJMP_H
|
|
2996 static long
|
|
2997 vim_regexec_both(line_arg, col_arg)
|
|
2998 char_u *line_arg;
|
|
2999 colnr_T col_arg; /* column to start looking for match */
|
|
3000 #else
|
|
3001 static long
|
|
3002 vim_regexec_both(line, col)
|
|
3003 char_u *line;
|
|
3004 colnr_T col; /* column to start looking for match */
|
|
3005 #endif
|
|
3006 {
|
|
3007 regprog_T *prog;
|
|
3008 char_u *s;
|
|
3009 long retval;
|
|
3010 #ifdef HAVE_SETJMP_H
|
|
3011 char_u *line;
|
|
3012 colnr_T col;
|
|
3013 #endif
|
|
3014
|
|
3015 reg_tofree = NULL;
|
|
3016
|
|
3017 #ifdef HAVE_TRY_EXCEPT
|
|
3018 __try
|
|
3019 {
|
|
3020 #endif
|
|
3021
|
|
3022 #ifdef HAVE_SETJMP_H
|
|
3023 /*
|
|
3024 * Matching with a regexp may cause a very deep recursive call of
|
|
3025 * regmatch(). Vim will crash when running out of stack space. Catch
|
|
3026 * this here if the system supports it.
|
|
3027 */
|
|
3028 mch_startjmp();
|
|
3029 if (SETJMP(lc_jump_env) != 0)
|
|
3030 {
|
|
3031 mch_didjmp();
|
|
3032 # ifdef SIGHASARG
|
|
3033 if (lc_signal != SIGINT)
|
|
3034 # endif
|
|
3035 EMSG(_("E361: Crash intercepted; regexp too complex?"));
|
|
3036 retval = 0L;
|
|
3037 goto theend;
|
|
3038 }
|
|
3039
|
|
3040 /* Trick to avoid "might be clobbered by `longjmp'" warning from gcc. */
|
|
3041 line = line_arg;
|
|
3042 col = col_arg;
|
|
3043 #endif
|
|
3044 retval = 0L;
|
|
3045
|
|
3046 if (REG_MULTI)
|
|
3047 {
|
|
3048 prog = reg_mmatch->regprog;
|
|
3049 line = reg_getline((linenr_T)0);
|
|
3050 reg_startpos = reg_mmatch->startpos;
|
|
3051 reg_endpos = reg_mmatch->endpos;
|
|
3052 }
|
|
3053 else
|
|
3054 {
|
|
3055 prog = reg_match->regprog;
|
|
3056 reg_startp = reg_match->startp;
|
|
3057 reg_endp = reg_match->endp;
|
|
3058 }
|
|
3059
|
|
3060 /* Be paranoid... */
|
|
3061 if (prog == NULL || line == NULL)
|
|
3062 {
|
|
3063 EMSG(_(e_null));
|
|
3064 goto theend;
|
|
3065 }
|
|
3066
|
|
3067 /* Check validity of program. */
|
|
3068 if (prog_magic_wrong())
|
|
3069 goto theend;
|
|
3070
|
|
3071 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */
|
|
3072 if (prog->regflags & RF_ICASE)
|
|
3073 ireg_ic = TRUE;
|
|
3074 else if (prog->regflags & RF_NOICASE)
|
|
3075 ireg_ic = FALSE;
|
|
3076
|
|
3077 #ifdef FEAT_MBYTE
|
|
3078 /* If pattern contains "\Z" overrule value of ireg_icombine */
|
|
3079 if (prog->regflags & RF_ICOMBINE)
|
|
3080 ireg_icombine = TRUE;
|
|
3081 #endif
|
|
3082
|
|
3083 /* If there is a "must appear" string, look for it. */
|
|
3084 if (prog->regmust != NULL)
|
|
3085 {
|
|
3086 int c;
|
|
3087
|
|
3088 #ifdef FEAT_MBYTE
|
|
3089 if (has_mbyte)
|
|
3090 c = (*mb_ptr2char)(prog->regmust);
|
|
3091 else
|
|
3092 #endif
|
|
3093 c = *prog->regmust;
|
|
3094 s = line + col;
|
|
3095 while ((s = cstrchr(s, c)) != NULL)
|
|
3096 {
|
|
3097 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3098 break; /* Found it. */
|
39
|
3099 mb_ptr_adv(s);
|
7
|
3100 }
|
|
3101 if (s == NULL) /* Not present. */
|
|
3102 goto theend;
|
|
3103 }
|
|
3104
|
|
3105 regline = line;
|
|
3106 reglnum = 0;
|
|
3107 out_of_stack = FALSE;
|
|
3108
|
|
3109 /* Simplest case: Anchored match need be tried only once. */
|
|
3110 if (prog->reganch)
|
|
3111 {
|
|
3112 int c;
|
|
3113
|
|
3114 #ifdef FEAT_MBYTE
|
|
3115 if (has_mbyte)
|
|
3116 c = (*mb_ptr2char)(regline + col);
|
|
3117 else
|
|
3118 #endif
|
|
3119 c = regline[col];
|
|
3120 if (prog->regstart == NUL
|
|
3121 || prog->regstart == c
|
|
3122 || (ireg_ic && ((
|
|
3123 #ifdef FEAT_MBYTE
|
|
3124 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
|
|
3125 || (c < 255 && prog->regstart < 255 &&
|
|
3126 #endif
|
|
3127 TOLOWER_LOC(prog->regstart) == TOLOWER_LOC(c)))))
|
|
3128 retval = regtry(prog, col);
|
|
3129 else
|
|
3130 retval = 0;
|
|
3131 }
|
|
3132 else
|
|
3133 {
|
|
3134 /* Messy cases: unanchored match. */
|
|
3135 while (!got_int && !out_of_stack)
|
|
3136 {
|
|
3137 if (prog->regstart != NUL)
|
|
3138 {
|
|
3139 /* Skip until the char we know it must start with. */
|
|
3140 s = cstrchr(regline + col, prog->regstart);
|
|
3141 if (s == NULL)
|
|
3142 {
|
|
3143 retval = 0;
|
|
3144 break;
|
|
3145 }
|
|
3146 col = (int)(s - regline);
|
|
3147 }
|
|
3148
|
|
3149 retval = regtry(prog, col);
|
|
3150 if (retval > 0)
|
|
3151 break;
|
|
3152
|
|
3153 /* if not currently on the first line, get it again */
|
|
3154 if (reglnum != 0)
|
|
3155 {
|
|
3156 regline = reg_getline((linenr_T)0);
|
|
3157 reglnum = 0;
|
|
3158 }
|
|
3159 if (regline[col] == NUL)
|
|
3160 break;
|
|
3161 #ifdef FEAT_MBYTE
|
|
3162 if (has_mbyte)
|
|
3163 col += (*mb_ptr2len_check)(regline + col);
|
|
3164 else
|
|
3165 #endif
|
|
3166 ++col;
|
|
3167 }
|
|
3168 }
|
|
3169
|
|
3170 if (out_of_stack)
|
|
3171 EMSG(_("E363: pattern caused out-of-stack error"));
|
|
3172
|
|
3173 #ifdef HAVE_TRY_EXCEPT
|
|
3174 }
|
|
3175 __except(EXCEPTION_EXECUTE_HANDLER)
|
|
3176 {
|
|
3177 if (GetExceptionCode() == EXCEPTION_STACK_OVERFLOW)
|
|
3178 {
|
|
3179 RESETSTKOFLW();
|
|
3180 EMSG(_("E363: pattern caused out-of-stack error"));
|
|
3181 }
|
|
3182 else
|
|
3183 EMSG(_("E361: Crash intercepted; regexp too complex?"));
|
|
3184 retval = 0L;
|
|
3185 }
|
|
3186 #endif
|
|
3187
|
|
3188 theend:
|
|
3189 /* Didn't find a match. */
|
|
3190 vim_free(reg_tofree);
|
|
3191 #ifdef HAVE_SETJMP_H
|
|
3192 mch_endjmp();
|
|
3193 #endif
|
|
3194 return retval;
|
|
3195 }
|
|
3196
|
|
3197 #ifdef FEAT_SYN_HL
|
|
3198 static reg_extmatch_T *make_extmatch __ARGS((void));
|
|
3199
|
|
3200 /*
|
|
3201 * Create a new extmatch and mark it as referenced once.
|
|
3202 */
|
|
3203 static reg_extmatch_T *
|
|
3204 make_extmatch()
|
|
3205 {
|
|
3206 reg_extmatch_T *em;
|
|
3207
|
|
3208 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T));
|
|
3209 if (em != NULL)
|
|
3210 em->refcnt = 1;
|
|
3211 return em;
|
|
3212 }
|
|
3213
|
|
3214 /*
|
|
3215 * Add a reference to an extmatch.
|
|
3216 */
|
|
3217 reg_extmatch_T *
|
|
3218 ref_extmatch(em)
|
|
3219 reg_extmatch_T *em;
|
|
3220 {
|
|
3221 if (em != NULL)
|
|
3222 em->refcnt++;
|
|
3223 return em;
|
|
3224 }
|
|
3225
|
|
3226 /*
|
|
3227 * Remove a reference to an extmatch. If there are no references left, free
|
|
3228 * the info.
|
|
3229 */
|
|
3230 void
|
|
3231 unref_extmatch(em)
|
|
3232 reg_extmatch_T *em;
|
|
3233 {
|
|
3234 int i;
|
|
3235
|
|
3236 if (em != NULL && --em->refcnt <= 0)
|
|
3237 {
|
|
3238 for (i = 0; i < NSUBEXP; ++i)
|
|
3239 vim_free(em->matches[i]);
|
|
3240 vim_free(em);
|
|
3241 }
|
|
3242 }
|
|
3243 #endif
|
|
3244
|
|
3245 /*
|
|
3246 * regtry - try match of "prog" with at regline["col"].
|
|
3247 * Returns 0 for failure, number of lines contained in the match otherwise.
|
|
3248 */
|
|
3249 static long
|
|
3250 regtry(prog, col)
|
|
3251 regprog_T *prog;
|
|
3252 colnr_T col;
|
|
3253 {
|
|
3254 reginput = regline + col;
|
|
3255 need_clear_subexpr = TRUE;
|
|
3256 #ifdef FEAT_SYN_HL
|
|
3257 /* Clear the external match subpointers if necessary. */
|
|
3258 if (prog->reghasz == REX_SET)
|
|
3259 need_clear_zsubexpr = TRUE;
|
|
3260 #endif
|
|
3261
|
|
3262 if (regmatch(prog->program + 1))
|
|
3263 {
|
|
3264 cleanup_subexpr();
|
|
3265 if (REG_MULTI)
|
|
3266 {
|
|
3267 if (reg_startpos[0].lnum < 0)
|
|
3268 {
|
|
3269 reg_startpos[0].lnum = 0;
|
|
3270 reg_startpos[0].col = col;
|
|
3271 }
|
|
3272 if (reg_endpos[0].lnum < 0)
|
|
3273 {
|
|
3274 reg_endpos[0].lnum = reglnum;
|
|
3275 reg_endpos[0].col = (int)(reginput - regline);
|
|
3276 }
|
|
3277 else
|
|
3278 /* Use line number of "\ze". */
|
|
3279 reglnum = reg_endpos[0].lnum;
|
|
3280 }
|
|
3281 else
|
|
3282 {
|
|
3283 if (reg_startp[0] == NULL)
|
|
3284 reg_startp[0] = regline + col;
|
|
3285 if (reg_endp[0] == NULL)
|
|
3286 reg_endp[0] = reginput;
|
|
3287 }
|
|
3288 #ifdef FEAT_SYN_HL
|
|
3289 /* Package any found \z(...\) matches for export. Default is none. */
|
|
3290 unref_extmatch(re_extmatch_out);
|
|
3291 re_extmatch_out = NULL;
|
|
3292
|
|
3293 if (prog->reghasz == REX_SET)
|
|
3294 {
|
|
3295 int i;
|
|
3296
|
|
3297 cleanup_zsubexpr();
|
|
3298 re_extmatch_out = make_extmatch();
|
|
3299 for (i = 0; i < NSUBEXP; i++)
|
|
3300 {
|
|
3301 if (REG_MULTI)
|
|
3302 {
|
|
3303 /* Only accept single line matches. */
|
|
3304 if (reg_startzpos[i].lnum >= 0
|
|
3305 && reg_endzpos[i].lnum == reg_startzpos[i].lnum)
|
|
3306 re_extmatch_out->matches[i] =
|
|
3307 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
|
|
3308 + reg_startzpos[i].col,
|
|
3309 reg_endzpos[i].col - reg_startzpos[i].col);
|
|
3310 }
|
|
3311 else
|
|
3312 {
|
|
3313 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
|
|
3314 re_extmatch_out->matches[i] =
|
|
3315 vim_strnsave(reg_startzp[i],
|
|
3316 (int)(reg_endzp[i] - reg_startzp[i]));
|
|
3317 }
|
|
3318 }
|
|
3319 }
|
|
3320 #endif
|
|
3321 return 1 + reglnum;
|
|
3322 }
|
|
3323 return 0;
|
|
3324 }
|
|
3325
|
|
3326 #ifdef FEAT_MBYTE
|
|
3327 static int reg_prev_class __ARGS((void));
|
|
3328
|
|
3329 /*
|
|
3330 * Get class of previous character.
|
|
3331 */
|
|
3332 static int
|
|
3333 reg_prev_class()
|
|
3334 {
|
|
3335 if (reginput > regline)
|
|
3336 return mb_get_class(reginput - 1
|
|
3337 - (*mb_head_off)(regline, reginput - 1));
|
|
3338 return -1;
|
|
3339 }
|
|
3340
|
|
3341 #endif
|
39
|
3342 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput)
|
7
|
3343
|
|
3344 /*
|
|
3345 * The arguments from BRACE_LIMITS are stored here. They are actually local
|
|
3346 * to regmatch(), but they are here to reduce the amount of stack space used
|
|
3347 * (it can be called recursively many times).
|
|
3348 */
|
|
3349 static long bl_minval;
|
|
3350 static long bl_maxval;
|
|
3351
|
|
3352 /*
|
|
3353 * regmatch - main matching routine
|
|
3354 *
|
|
3355 * Conceptually the strategy is simple: Check to see whether the current
|
|
3356 * node matches, call self recursively to see whether the rest matches,
|
|
3357 * and then act accordingly. In practice we make some effort to avoid
|
|
3358 * recursion, in particular by going through "ordinary" nodes (that don't
|
|
3359 * need to know whether the rest of the match failed) by a loop instead of
|
|
3360 * by recursion.
|
|
3361 *
|
|
3362 * Returns TRUE when there is a match. Leaves reginput and reglnum just after
|
|
3363 * the last matched character.
|
|
3364 * Returns FALSE when there is no match. Leaves reginput and reglnum in an
|
|
3365 * undefined state!
|
|
3366 */
|
|
3367 static int
|
|
3368 regmatch(scan)
|
|
3369 char_u *scan; /* Current node. */
|
|
3370 {
|
|
3371 char_u *next; /* Next node. */
|
|
3372 int op;
|
|
3373 int c;
|
|
3374
|
|
3375 #ifdef HAVE_GETRLIMIT
|
|
3376 /* Check if we are running out of stack space. Could be caused by
|
|
3377 * recursively calling ourselves. */
|
|
3378 if (out_of_stack || mch_stackcheck((char *)&op) == FAIL)
|
|
3379 {
|
|
3380 out_of_stack = TRUE;
|
|
3381 return FALSE;
|
|
3382 }
|
|
3383 #endif
|
|
3384
|
|
3385 /* Some patterns my cause a long time to match, even though they are not
|
|
3386 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */
|
|
3387 fast_breakcheck();
|
|
3388
|
|
3389 #ifdef DEBUG
|
|
3390 if (scan != NULL && regnarrate)
|
|
3391 {
|
|
3392 mch_errmsg(regprop(scan));
|
|
3393 mch_errmsg("(\n");
|
|
3394 }
|
|
3395 #endif
|
|
3396 while (scan != NULL)
|
|
3397 {
|
|
3398 if (got_int || out_of_stack)
|
|
3399 return FALSE;
|
|
3400 #ifdef DEBUG
|
|
3401 if (regnarrate)
|
|
3402 {
|
|
3403 mch_errmsg(regprop(scan));
|
|
3404 mch_errmsg("...\n");
|
|
3405 # ifdef FEAT_SYN_HL
|
|
3406 if (re_extmatch_in != NULL)
|
|
3407 {
|
|
3408 int i;
|
|
3409
|
|
3410 mch_errmsg(_("External submatches:\n"));
|
|
3411 for (i = 0; i < NSUBEXP; i++)
|
|
3412 {
|
|
3413 mch_errmsg(" \"");
|
|
3414 if (re_extmatch_in->matches[i] != NULL)
|
|
3415 mch_errmsg(re_extmatch_in->matches[i]);
|
|
3416 mch_errmsg("\"\n");
|
|
3417 }
|
|
3418 }
|
|
3419 # endif
|
|
3420 }
|
|
3421 #endif
|
|
3422 next = regnext(scan);
|
|
3423
|
|
3424 op = OP(scan);
|
|
3425 /* Check for character class with NL added. */
|
|
3426 if (WITH_NL(op) && *reginput == NUL && reglnum < reg_maxline)
|
|
3427 {
|
|
3428 reg_nextline();
|
|
3429 }
|
|
3430 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n')
|
|
3431 {
|
|
3432 ADVANCE_REGINPUT();
|
|
3433 }
|
|
3434 else
|
|
3435 {
|
|
3436 if (WITH_NL(op))
|
|
3437 op -= ADD_NL;
|
|
3438 #ifdef FEAT_MBYTE
|
|
3439 if (has_mbyte)
|
|
3440 c = (*mb_ptr2char)(reginput);
|
|
3441 else
|
|
3442 #endif
|
|
3443 c = *reginput;
|
|
3444 switch (op)
|
|
3445 {
|
|
3446 case BOL:
|
|
3447 if (reginput != regline)
|
|
3448 return FALSE;
|
|
3449 break;
|
|
3450
|
|
3451 case EOL:
|
|
3452 if (c != NUL)
|
|
3453 return FALSE;
|
|
3454 break;
|
|
3455
|
|
3456 case RE_BOF:
|
|
3457 /* Passing -1 to the getline() function provided for the search
|
|
3458 * should always return NULL if the current line is the first
|
|
3459 * line of the file. */
|
|
3460 if (reglnum != 0 || reginput != regline
|
|
3461 || (REG_MULTI && reg_getline((linenr_T)-1) != NULL))
|
|
3462 return FALSE;
|
|
3463 break;
|
|
3464
|
|
3465 case RE_EOF:
|
|
3466 if (reglnum != reg_maxline || c != NUL)
|
|
3467 return FALSE;
|
|
3468 break;
|
|
3469
|
|
3470 case CURSOR:
|
|
3471 /* Check if the buffer is in a window and compare the
|
|
3472 * reg_win->w_cursor position to the match position. */
|
|
3473 if (reg_win == NULL
|
|
3474 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum)
|
|
3475 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col))
|
|
3476 return FALSE;
|
|
3477 break;
|
|
3478
|
|
3479 case RE_LNUM:
|
|
3480 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum),
|
|
3481 scan))
|
|
3482 return FALSE;
|
|
3483 break;
|
|
3484
|
|
3485 case RE_COL:
|
|
3486 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan))
|
|
3487 return FALSE;
|
|
3488 break;
|
|
3489
|
|
3490 case RE_VCOL:
|
|
3491 if (!re_num_cmp((long_u)win_linetabsize(
|
|
3492 reg_win == NULL ? curwin : reg_win,
|
|
3493 regline, (colnr_T)(reginput - regline)) + 1, scan))
|
|
3494 return FALSE;
|
|
3495 break;
|
|
3496
|
|
3497 case BOW: /* \<word; reginput points to w */
|
|
3498 if (c == NUL) /* Can't match at end of line */
|
|
3499 return FALSE;
|
|
3500 #ifdef FEAT_MBYTE
|
|
3501 if (has_mbyte)
|
|
3502 {
|
|
3503 int this_class;
|
|
3504
|
|
3505 /* Get class of current and previous char (if it exists). */
|
|
3506 this_class = mb_get_class(reginput);
|
|
3507 if (this_class <= 1)
|
|
3508 return FALSE; /* not on a word at all */
|
|
3509 if (reg_prev_class() == this_class)
|
|
3510 return FALSE; /* previous char is in same word */
|
|
3511 }
|
|
3512 #endif
|
|
3513 else
|
|
3514 {
|
|
3515 if (!vim_iswordc(c)
|
|
3516 || (reginput > regline && vim_iswordc(reginput[-1])))
|
|
3517 return FALSE;
|
|
3518 }
|
|
3519 break;
|
|
3520
|
|
3521 case EOW: /* word\>; reginput points after d */
|
|
3522 if (reginput == regline) /* Can't match at start of line */
|
|
3523 return FALSE;
|
|
3524 #ifdef FEAT_MBYTE
|
|
3525 if (has_mbyte)
|
|
3526 {
|
|
3527 int this_class, prev_class;
|
|
3528
|
|
3529 /* Get class of current and previous char (if it exists). */
|
|
3530 this_class = mb_get_class(reginput);
|
|
3531 prev_class = reg_prev_class();
|
|
3532 if (this_class == prev_class)
|
|
3533 return FALSE;
|
|
3534 if (prev_class == 0 || prev_class == 1)
|
|
3535 return FALSE;
|
|
3536 }
|
|
3537 else
|
|
3538 #endif
|
|
3539 {
|
|
3540 if (!vim_iswordc(reginput[-1]))
|
|
3541 return FALSE;
|
|
3542 if (reginput[0] != NUL && vim_iswordc(c))
|
|
3543 return FALSE;
|
|
3544 }
|
|
3545 break; /* Matched with EOW */
|
|
3546
|
|
3547 case ANY:
|
|
3548 if (c == NUL)
|
|
3549 return FALSE;
|
|
3550 ADVANCE_REGINPUT();
|
|
3551 break;
|
|
3552
|
|
3553 case IDENT:
|
|
3554 if (!vim_isIDc(c))
|
|
3555 return FALSE;
|
|
3556 ADVANCE_REGINPUT();
|
|
3557 break;
|
|
3558
|
|
3559 case SIDENT:
|
|
3560 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c))
|
|
3561 return FALSE;
|
|
3562 ADVANCE_REGINPUT();
|
|
3563 break;
|
|
3564
|
|
3565 case KWORD:
|
|
3566 if (!vim_iswordp(reginput))
|
|
3567 return FALSE;
|
|
3568 ADVANCE_REGINPUT();
|
|
3569 break;
|
|
3570
|
|
3571 case SKWORD:
|
|
3572 if (VIM_ISDIGIT(*reginput) || !vim_iswordp(reginput))
|
|
3573 return FALSE;
|
|
3574 ADVANCE_REGINPUT();
|
|
3575 break;
|
|
3576
|
|
3577 case FNAME:
|
|
3578 if (!vim_isfilec(c))
|
|
3579 return FALSE;
|
|
3580 ADVANCE_REGINPUT();
|
|
3581 break;
|
|
3582
|
|
3583 case SFNAME:
|
|
3584 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c))
|
|
3585 return FALSE;
|
|
3586 ADVANCE_REGINPUT();
|
|
3587 break;
|
|
3588
|
|
3589 case PRINT:
|
|
3590 if (ptr2cells(reginput) != 1)
|
|
3591 return FALSE;
|
|
3592 ADVANCE_REGINPUT();
|
|
3593 break;
|
|
3594
|
|
3595 case SPRINT:
|
|
3596 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1)
|
|
3597 return FALSE;
|
|
3598 ADVANCE_REGINPUT();
|
|
3599 break;
|
|
3600
|
|
3601 case WHITE:
|
|
3602 if (!vim_iswhite(c))
|
|
3603 return FALSE;
|
|
3604 ADVANCE_REGINPUT();
|
|
3605 break;
|
|
3606
|
|
3607 case NWHITE:
|
|
3608 if (c == NUL || vim_iswhite(c))
|
|
3609 return FALSE;
|
|
3610 ADVANCE_REGINPUT();
|
|
3611 break;
|
|
3612
|
|
3613 case DIGIT:
|
|
3614 if (!ri_digit(c))
|
|
3615 return FALSE;
|
|
3616 ADVANCE_REGINPUT();
|
|
3617 break;
|
|
3618
|
|
3619 case NDIGIT:
|
|
3620 if (c == NUL || ri_digit(c))
|
|
3621 return FALSE;
|
|
3622 ADVANCE_REGINPUT();
|
|
3623 break;
|
|
3624
|
|
3625 case HEX:
|
|
3626 if (!ri_hex(c))
|
|
3627 return FALSE;
|
|
3628 ADVANCE_REGINPUT();
|
|
3629 break;
|
|
3630
|
|
3631 case NHEX:
|
|
3632 if (c == NUL || ri_hex(c))
|
|
3633 return FALSE;
|
|
3634 ADVANCE_REGINPUT();
|
|
3635 break;
|
|
3636
|
|
3637 case OCTAL:
|
|
3638 if (!ri_octal(c))
|
|
3639 return FALSE;
|
|
3640 ADVANCE_REGINPUT();
|
|
3641 break;
|
|
3642
|
|
3643 case NOCTAL:
|
|
3644 if (c == NUL || ri_octal(c))
|
|
3645 return FALSE;
|
|
3646 ADVANCE_REGINPUT();
|
|
3647 break;
|
|
3648
|
|
3649 case WORD:
|
|
3650 if (!ri_word(c))
|
|
3651 return FALSE;
|
|
3652 ADVANCE_REGINPUT();
|
|
3653 break;
|
|
3654
|
|
3655 case NWORD:
|
|
3656 if (c == NUL || ri_word(c))
|
|
3657 return FALSE;
|
|
3658 ADVANCE_REGINPUT();
|
|
3659 break;
|
|
3660
|
|
3661 case HEAD:
|
|
3662 if (!ri_head(c))
|
|
3663 return FALSE;
|
|
3664 ADVANCE_REGINPUT();
|
|
3665 break;
|
|
3666
|
|
3667 case NHEAD:
|
|
3668 if (c == NUL || ri_head(c))
|
|
3669 return FALSE;
|
|
3670 ADVANCE_REGINPUT();
|
|
3671 break;
|
|
3672
|
|
3673 case ALPHA:
|
|
3674 if (!ri_alpha(c))
|
|
3675 return FALSE;
|
|
3676 ADVANCE_REGINPUT();
|
|
3677 break;
|
|
3678
|
|
3679 case NALPHA:
|
|
3680 if (c == NUL || ri_alpha(c))
|
|
3681 return FALSE;
|
|
3682 ADVANCE_REGINPUT();
|
|
3683 break;
|
|
3684
|
|
3685 case LOWER:
|
|
3686 if (!ri_lower(c))
|
|
3687 return FALSE;
|
|
3688 ADVANCE_REGINPUT();
|
|
3689 break;
|
|
3690
|
|
3691 case NLOWER:
|
|
3692 if (c == NUL || ri_lower(c))
|
|
3693 return FALSE;
|
|
3694 ADVANCE_REGINPUT();
|
|
3695 break;
|
|
3696
|
|
3697 case UPPER:
|
|
3698 if (!ri_upper(c))
|
|
3699 return FALSE;
|
|
3700 ADVANCE_REGINPUT();
|
|
3701 break;
|
|
3702
|
|
3703 case NUPPER:
|
|
3704 if (c == NUL || ri_upper(c))
|
|
3705 return FALSE;
|
|
3706 ADVANCE_REGINPUT();
|
|
3707 break;
|
|
3708
|
|
3709 case EXACTLY:
|
|
3710 {
|
|
3711 int len;
|
|
3712 char_u *opnd;
|
|
3713
|
|
3714 opnd = OPERAND(scan);
|
|
3715 /* Inline the first byte, for speed. */
|
|
3716 if (*opnd != *reginput
|
|
3717 && (!ireg_ic || (
|
|
3718 #ifdef FEAT_MBYTE
|
|
3719 !enc_utf8 &&
|
|
3720 #endif
|
|
3721 TOLOWER_LOC(*opnd) != TOLOWER_LOC(*reginput))))
|
|
3722 return FALSE;
|
|
3723 if (*opnd == NUL)
|
|
3724 {
|
|
3725 /* match empty string always works; happens when "~" is
|
|
3726 * empty. */
|
|
3727 }
|
|
3728 else if (opnd[1] == NUL
|
|
3729 #ifdef FEAT_MBYTE
|
|
3730 && !(enc_utf8 && ireg_ic)
|
|
3731 #endif
|
|
3732 )
|
|
3733 ++reginput; /* matched a single char */
|
|
3734 else
|
|
3735 {
|
|
3736 len = (int)STRLEN(opnd);
|
|
3737 /* Need to match first byte again for multi-byte. */
|
|
3738 if (cstrncmp(opnd, reginput, &len) != 0)
|
|
3739 return FALSE;
|
|
3740 #ifdef FEAT_MBYTE
|
|
3741 /* Check for following composing character. */
|
|
3742 if (enc_utf8 && UTF_COMPOSINGLIKE(reginput, reginput + len))
|
|
3743 {
|
|
3744 /* raaron: This code makes a composing character get
|
|
3745 * ignored, which is the correct behavior (sometimes)
|
|
3746 * for voweled Hebrew texts. */
|
|
3747 if (!ireg_icombine)
|
|
3748 return FALSE;
|
|
3749 }
|
|
3750 else
|
|
3751 #endif
|
|
3752 reginput += len;
|
|
3753 }
|
|
3754 }
|
|
3755 break;
|
|
3756
|
|
3757 case ANYOF:
|
|
3758 case ANYBUT:
|
|
3759 if (c == NUL)
|
|
3760 return FALSE;
|
|
3761 if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF))
|
|
3762 return FALSE;
|
|
3763 ADVANCE_REGINPUT();
|
|
3764 break;
|
|
3765
|
|
3766 #ifdef FEAT_MBYTE
|
|
3767 case MULTIBYTECODE:
|
|
3768 if (has_mbyte)
|
|
3769 {
|
|
3770 int i, len;
|
|
3771 char_u *opnd;
|
|
3772
|
|
3773 opnd = OPERAND(scan);
|
|
3774 /* Safety check (just in case 'encoding' was changed since
|
|
3775 * compiling the program). */
|
|
3776 if ((len = (*mb_ptr2len_check)(opnd)) < 2)
|
|
3777 return FALSE;
|
|
3778 for (i = 0; i < len; ++i)
|
|
3779 if (opnd[i] != reginput[i])
|
|
3780 return FALSE;
|
|
3781 reginput += len;
|
|
3782 }
|
|
3783 else
|
|
3784 return FALSE;
|
|
3785 break;
|
|
3786 #endif
|
|
3787
|
|
3788 case NOTHING:
|
|
3789 break;
|
|
3790
|
|
3791 case BACK:
|
|
3792 break;
|
|
3793
|
|
3794 case MOPEN + 0: /* Match start: \zs */
|
|
3795 case MOPEN + 1: /* \( */
|
|
3796 case MOPEN + 2:
|
|
3797 case MOPEN + 3:
|
|
3798 case MOPEN + 4:
|
|
3799 case MOPEN + 5:
|
|
3800 case MOPEN + 6:
|
|
3801 case MOPEN + 7:
|
|
3802 case MOPEN + 8:
|
|
3803 case MOPEN + 9:
|
|
3804 {
|
|
3805 int no;
|
|
3806 save_se_T save;
|
|
3807
|
|
3808 no = op - MOPEN;
|
|
3809 cleanup_subexpr();
|
|
3810 save_se(&save, ®_startpos[no], ®_startp[no]);
|
|
3811
|
|
3812 if (regmatch(next))
|
|
3813 return TRUE;
|
|
3814
|
|
3815 restore_se(&save, ®_startpos[no], ®_startp[no]);
|
|
3816 return FALSE;
|
|
3817 }
|
|
3818 /* break; Not Reached */
|
|
3819
|
|
3820 case NOPEN: /* \%( */
|
|
3821 case NCLOSE: /* \) after \%( */
|
|
3822 if (regmatch(next))
|
|
3823 return TRUE;
|
|
3824 return FALSE;
|
|
3825 /* break; Not Reached */
|
|
3826
|
|
3827 #ifdef FEAT_SYN_HL
|
|
3828 case ZOPEN + 1:
|
|
3829 case ZOPEN + 2:
|
|
3830 case ZOPEN + 3:
|
|
3831 case ZOPEN + 4:
|
|
3832 case ZOPEN + 5:
|
|
3833 case ZOPEN + 6:
|
|
3834 case ZOPEN + 7:
|
|
3835 case ZOPEN + 8:
|
|
3836 case ZOPEN + 9:
|
|
3837 {
|
|
3838 int no;
|
|
3839 save_se_T save;
|
|
3840
|
|
3841 no = op - ZOPEN;
|
|
3842 cleanup_zsubexpr();
|
|
3843 save_se(&save, ®_startzpos[no], ®_startzp[no]);
|
|
3844
|
|
3845 if (regmatch(next))
|
|
3846 return TRUE;
|
|
3847
|
|
3848 restore_se(&save, ®_startzpos[no], ®_startzp[no]);
|
|
3849 return FALSE;
|
|
3850 }
|
|
3851 /* break; Not Reached */
|
|
3852 #endif
|
|
3853
|
|
3854 case MCLOSE + 0: /* Match end: \ze */
|
|
3855 case MCLOSE + 1: /* \) */
|
|
3856 case MCLOSE + 2:
|
|
3857 case MCLOSE + 3:
|
|
3858 case MCLOSE + 4:
|
|
3859 case MCLOSE + 5:
|
|
3860 case MCLOSE + 6:
|
|
3861 case MCLOSE + 7:
|
|
3862 case MCLOSE + 8:
|
|
3863 case MCLOSE + 9:
|
|
3864 {
|
|
3865 int no;
|
|
3866 save_se_T save;
|
|
3867
|
|
3868 no = op - MCLOSE;
|
|
3869 cleanup_subexpr();
|
|
3870 save_se(&save, ®_endpos[no], ®_endp[no]);
|
|
3871
|
|
3872 if (regmatch(next))
|
|
3873 return TRUE;
|
|
3874
|
|
3875 restore_se(&save, ®_endpos[no], ®_endp[no]);
|
|
3876 return FALSE;
|
|
3877 }
|
|
3878 /* break; Not Reached */
|
|
3879
|
|
3880 #ifdef FEAT_SYN_HL
|
|
3881 case ZCLOSE + 1: /* \) after \z( */
|
|
3882 case ZCLOSE + 2:
|
|
3883 case ZCLOSE + 3:
|
|
3884 case ZCLOSE + 4:
|
|
3885 case ZCLOSE + 5:
|
|
3886 case ZCLOSE + 6:
|
|
3887 case ZCLOSE + 7:
|
|
3888 case ZCLOSE + 8:
|
|
3889 case ZCLOSE + 9:
|
|
3890 {
|
|
3891 int no;
|
|
3892 save_se_T save;
|
|
3893
|
|
3894 no = op - ZCLOSE;
|
|
3895 cleanup_zsubexpr();
|
|
3896 save_se(&save, ®_endzpos[no], ®_endzp[no]);
|
|
3897
|
|
3898 if (regmatch(next))
|
|
3899 return TRUE;
|
|
3900
|
|
3901 restore_se(&save, ®_endzpos[no], ®_endzp[no]);
|
|
3902 return FALSE;
|
|
3903 }
|
|
3904 /* break; Not Reached */
|
|
3905 #endif
|
|
3906
|
|
3907 case BACKREF + 1:
|
|
3908 case BACKREF + 2:
|
|
3909 case BACKREF + 3:
|
|
3910 case BACKREF + 4:
|
|
3911 case BACKREF + 5:
|
|
3912 case BACKREF + 6:
|
|
3913 case BACKREF + 7:
|
|
3914 case BACKREF + 8:
|
|
3915 case BACKREF + 9:
|
|
3916 {
|
|
3917 int no;
|
|
3918 int len;
|
|
3919 linenr_T clnum;
|
|
3920 colnr_T ccol;
|
|
3921 char_u *p;
|
|
3922
|
|
3923 no = op - BACKREF;
|
|
3924 cleanup_subexpr();
|
|
3925 if (!REG_MULTI) /* Single-line regexp */
|
|
3926 {
|
|
3927 if (reg_endp[no] == NULL)
|
|
3928 {
|
|
3929 /* Backref was not set: Match an empty string. */
|
|
3930 len = 0;
|
|
3931 }
|
|
3932 else
|
|
3933 {
|
|
3934 /* Compare current input with back-ref in the same
|
|
3935 * line. */
|
|
3936 len = (int)(reg_endp[no] - reg_startp[no]);
|
|
3937 if (cstrncmp(reg_startp[no], reginput, &len) != 0)
|
|
3938 return FALSE;
|
|
3939 }
|
|
3940 }
|
|
3941 else /* Multi-line regexp */
|
|
3942 {
|
|
3943 if (reg_endpos[no].lnum < 0)
|
|
3944 {
|
|
3945 /* Backref was not set: Match an empty string. */
|
|
3946 len = 0;
|
|
3947 }
|
|
3948 else
|
|
3949 {
|
|
3950 if (reg_startpos[no].lnum == reglnum
|
|
3951 && reg_endpos[no].lnum == reglnum)
|
|
3952 {
|
|
3953 /* Compare back-ref within the current line. */
|
|
3954 len = reg_endpos[no].col - reg_startpos[no].col;
|
|
3955 if (cstrncmp(regline + reg_startpos[no].col,
|
|
3956 reginput, &len) != 0)
|
|
3957 return FALSE;
|
|
3958 }
|
|
3959 else
|
|
3960 {
|
|
3961 /* Messy situation: Need to compare between two
|
|
3962 * lines. */
|
|
3963 ccol = reg_startpos[no].col;
|
|
3964 clnum = reg_startpos[no].lnum;
|
|
3965 for (;;)
|
|
3966 {
|
|
3967 /* Since getting one line may invalidate
|
|
3968 * the other, need to make copy. Slow! */
|
|
3969 if (regline != reg_tofree)
|
|
3970 {
|
|
3971 len = (int)STRLEN(regline);
|
|
3972 if (reg_tofree == NULL
|
|
3973 || len >= (int)reg_tofreelen)
|
|
3974 {
|
|
3975 len += 50; /* get some extra */
|
|
3976 vim_free(reg_tofree);
|
|
3977 reg_tofree = alloc(len);
|
|
3978 if (reg_tofree == NULL)
|
|
3979 return FALSE; /* out of memory! */
|
|
3980 reg_tofreelen = len;
|
|
3981 }
|
|
3982 STRCPY(reg_tofree, regline);
|
|
3983 reginput = reg_tofree
|
|
3984 + (reginput - regline);
|
|
3985 regline = reg_tofree;
|
|
3986 }
|
|
3987
|
|
3988 /* Get the line to compare with. */
|
|
3989 p = reg_getline(clnum);
|
|
3990 if (clnum == reg_endpos[no].lnum)
|
|
3991 len = reg_endpos[no].col - ccol;
|
|
3992 else
|
|
3993 len = (int)STRLEN(p + ccol);
|
|
3994
|
|
3995 if (cstrncmp(p + ccol, reginput, &len) != 0)
|
|
3996 return FALSE; /* doesn't match */
|
|
3997 if (clnum == reg_endpos[no].lnum)
|
|
3998 break; /* match and at end! */
|
|
3999 if (reglnum == reg_maxline)
|
|
4000 return FALSE; /* text too short */
|
|
4001
|
|
4002 /* Advance to next line. */
|
|
4003 reg_nextline();
|
|
4004 ++clnum;
|
|
4005 ccol = 0;
|
|
4006 if (got_int || out_of_stack)
|
|
4007 return FALSE;
|
|
4008 }
|
|
4009
|
|
4010 /* found a match! Note that regline may now point
|
|
4011 * to a copy of the line, that should not matter. */
|
|
4012 }
|
|
4013 }
|
|
4014 }
|
|
4015
|
|
4016 /* Matched the backref, skip over it. */
|
|
4017 reginput += len;
|
|
4018 }
|
|
4019 break;
|
|
4020
|
|
4021 #ifdef FEAT_SYN_HL
|
|
4022 case ZREF + 1:
|
|
4023 case ZREF + 2:
|
|
4024 case ZREF + 3:
|
|
4025 case ZREF + 4:
|
|
4026 case ZREF + 5:
|
|
4027 case ZREF + 6:
|
|
4028 case ZREF + 7:
|
|
4029 case ZREF + 8:
|
|
4030 case ZREF + 9:
|
|
4031 {
|
|
4032 int no;
|
|
4033 int len;
|
|
4034
|
|
4035 cleanup_zsubexpr();
|
|
4036 no = op - ZREF;
|
|
4037 if (re_extmatch_in != NULL
|
|
4038 && re_extmatch_in->matches[no] != NULL)
|
|
4039 {
|
|
4040 len = (int)STRLEN(re_extmatch_in->matches[no]);
|
|
4041 if (cstrncmp(re_extmatch_in->matches[no],
|
|
4042 reginput, &len) != 0)
|
|
4043 return FALSE;
|
|
4044 reginput += len;
|
|
4045 }
|
|
4046 else
|
|
4047 {
|
|
4048 /* Backref was not set: Match an empty string. */
|
|
4049 }
|
|
4050 }
|
|
4051 break;
|
|
4052 #endif
|
|
4053
|
|
4054 case BRANCH:
|
|
4055 {
|
|
4056 if (OP(next) != BRANCH) /* No choice. */
|
|
4057 next = OPERAND(scan); /* Avoid recursion. */
|
|
4058 else
|
|
4059 {
|
|
4060 regsave_T save;
|
|
4061
|
|
4062 do
|
|
4063 {
|
|
4064 reg_save(&save);
|
|
4065 if (regmatch(OPERAND(scan)))
|
|
4066 return TRUE;
|
|
4067 reg_restore(&save);
|
|
4068 scan = regnext(scan);
|
|
4069 } while (scan != NULL && OP(scan) == BRANCH);
|
|
4070 return FALSE;
|
|
4071 /* NOTREACHED */
|
|
4072 }
|
|
4073 }
|
|
4074 break;
|
|
4075
|
|
4076 case BRACE_LIMITS:
|
|
4077 {
|
|
4078 int no;
|
|
4079
|
|
4080 if (OP(next) == BRACE_SIMPLE)
|
|
4081 {
|
|
4082 bl_minval = OPERAND_MIN(scan);
|
|
4083 bl_maxval = OPERAND_MAX(scan);
|
|
4084 }
|
|
4085 else if (OP(next) >= BRACE_COMPLEX
|
|
4086 && OP(next) < BRACE_COMPLEX + 10)
|
|
4087 {
|
|
4088 no = OP(next) - BRACE_COMPLEX;
|
|
4089 brace_min[no] = OPERAND_MIN(scan);
|
|
4090 brace_max[no] = OPERAND_MAX(scan);
|
|
4091 brace_count[no] = 0;
|
|
4092 }
|
|
4093 else
|
|
4094 {
|
|
4095 EMSG(_(e_internal)); /* Shouldn't happen */
|
|
4096 return FALSE;
|
|
4097 }
|
|
4098 }
|
|
4099 break;
|
|
4100
|
|
4101 case BRACE_COMPLEX + 0:
|
|
4102 case BRACE_COMPLEX + 1:
|
|
4103 case BRACE_COMPLEX + 2:
|
|
4104 case BRACE_COMPLEX + 3:
|
|
4105 case BRACE_COMPLEX + 4:
|
|
4106 case BRACE_COMPLEX + 5:
|
|
4107 case BRACE_COMPLEX + 6:
|
|
4108 case BRACE_COMPLEX + 7:
|
|
4109 case BRACE_COMPLEX + 8:
|
|
4110 case BRACE_COMPLEX + 9:
|
|
4111 {
|
|
4112 int no;
|
|
4113 regsave_T save;
|
|
4114
|
|
4115 no = op - BRACE_COMPLEX;
|
|
4116 ++brace_count[no];
|
|
4117
|
|
4118 /* If not matched enough times yet, try one more */
|
|
4119 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
|
|
4120 ? brace_min[no] : brace_max[no]))
|
|
4121 {
|
|
4122 reg_save(&save);
|
|
4123 if (regmatch(OPERAND(scan)))
|
|
4124 return TRUE;
|
|
4125 reg_restore(&save);
|
|
4126 --brace_count[no]; /* failed, decrement match count */
|
|
4127 return FALSE;
|
|
4128 }
|
|
4129
|
|
4130 /* If matched enough times, may try matching some more */
|
|
4131 if (brace_min[no] <= brace_max[no])
|
|
4132 {
|
|
4133 /* Range is the normal way around, use longest match */
|
|
4134 if (brace_count[no] <= brace_max[no])
|
|
4135 {
|
|
4136 reg_save(&save);
|
|
4137 if (regmatch(OPERAND(scan)))
|
|
4138 return TRUE; /* matched some more times */
|
|
4139 reg_restore(&save);
|
|
4140 --brace_count[no]; /* matched just enough times */
|
|
4141 /* continue with the items after \{} */
|
|
4142 }
|
|
4143 }
|
|
4144 else
|
|
4145 {
|
|
4146 /* Range is backwards, use shortest match first */
|
|
4147 if (brace_count[no] <= brace_min[no])
|
|
4148 {
|
|
4149 reg_save(&save);
|
|
4150 if (regmatch(next))
|
|
4151 return TRUE;
|
|
4152 reg_restore(&save);
|
|
4153 next = OPERAND(scan);
|
|
4154 /* must try to match one more item */
|
|
4155 }
|
|
4156 }
|
|
4157 }
|
|
4158 break;
|
|
4159
|
|
4160 case BRACE_SIMPLE:
|
|
4161 case STAR:
|
|
4162 case PLUS:
|
|
4163 {
|
|
4164 int nextb; /* next byte */
|
|
4165 int nextb_ic; /* next byte reverse case */
|
|
4166 long count;
|
|
4167 regsave_T save;
|
|
4168 long minval;
|
|
4169 long maxval;
|
|
4170
|
|
4171 /*
|
|
4172 * Lookahead to avoid useless match attempts when we know
|
|
4173 * what character comes next.
|
|
4174 */
|
|
4175 if (OP(next) == EXACTLY)
|
|
4176 {
|
|
4177 nextb = *OPERAND(next);
|
|
4178 if (ireg_ic)
|
|
4179 {
|
|
4180 if (isupper(nextb))
|
|
4181 nextb_ic = TOLOWER_LOC(nextb);
|
|
4182 else
|
|
4183 nextb_ic = TOUPPER_LOC(nextb);
|
|
4184 }
|
|
4185 else
|
|
4186 nextb_ic = nextb;
|
|
4187 }
|
|
4188 else
|
|
4189 {
|
|
4190 nextb = NUL;
|
|
4191 nextb_ic = NUL;
|
|
4192 }
|
|
4193 if (op != BRACE_SIMPLE)
|
|
4194 {
|
|
4195 minval = (op == STAR) ? 0 : 1;
|
|
4196 maxval = MAX_LIMIT;
|
|
4197 }
|
|
4198 else
|
|
4199 {
|
|
4200 minval = bl_minval;
|
|
4201 maxval = bl_maxval;
|
|
4202 }
|
|
4203
|
|
4204 /*
|
|
4205 * When maxval > minval, try matching as much as possible, up
|
|
4206 * to maxval. When maxval < minval, try matching at least the
|
|
4207 * minimal number (since the range is backwards, that's also
|
|
4208 * maxval!).
|
|
4209 */
|
|
4210 count = regrepeat(OPERAND(scan), maxval);
|
|
4211 if (got_int)
|
|
4212 return FALSE;
|
|
4213 if (minval <= maxval)
|
|
4214 {
|
|
4215 /* Range is the normal way around, use longest match */
|
|
4216 while (count >= minval)
|
|
4217 {
|
|
4218 /* If it could match, try it. */
|
|
4219 if (nextb == NUL || *reginput == nextb
|
|
4220 || *reginput == nextb_ic)
|
|
4221 {
|
|
4222 reg_save(&save);
|
|
4223 if (regmatch(next))
|
|
4224 return TRUE;
|
|
4225 reg_restore(&save);
|
|
4226 }
|
|
4227 /* Couldn't or didn't match -- back up one char. */
|
|
4228 if (--count < minval)
|
|
4229 break;
|
|
4230 if (reginput == regline)
|
|
4231 {
|
|
4232 /* backup to last char of previous line */
|
|
4233 --reglnum;
|
|
4234 regline = reg_getline(reglnum);
|
|
4235 /* Just in case regrepeat() didn't count right. */
|
|
4236 if (regline == NULL)
|
|
4237 return FALSE;
|
|
4238 reginput = regline + STRLEN(regline);
|
|
4239 fast_breakcheck();
|
|
4240 if (got_int || out_of_stack)
|
|
4241 return FALSE;
|
|
4242 }
|
|
4243 else
|
39
|
4244 mb_ptr_back(regline, reginput);
|
7
|
4245 }
|
|
4246 }
|
|
4247 else
|
|
4248 {
|
|
4249 /* Range is backwards, use shortest match first.
|
|
4250 * Careful: maxval and minval are exchanged! */
|
|
4251 if (count < maxval)
|
|
4252 return FALSE;
|
|
4253 for (;;)
|
|
4254 {
|
|
4255 /* If it could work, try it. */
|
|
4256 if (nextb == NUL || *reginput == nextb
|
|
4257 || *reginput == nextb_ic)
|
|
4258 {
|
|
4259 reg_save(&save);
|
|
4260 if (regmatch(next))
|
|
4261 return TRUE;
|
|
4262 reg_restore(&save);
|
|
4263 }
|
|
4264 /* Couldn't or didn't match: try advancing one char. */
|
|
4265 if (count == minval
|
|
4266 || regrepeat(OPERAND(scan), 1L) == 0)
|
|
4267 break;
|
|
4268 ++count;
|
|
4269 if (got_int || out_of_stack)
|
|
4270 return FALSE;
|
|
4271 }
|
|
4272 }
|
|
4273 return FALSE;
|
|
4274 }
|
|
4275 /* break; Not Reached */
|
|
4276
|
|
4277 case NOMATCH:
|
|
4278 {
|
|
4279 regsave_T save;
|
|
4280
|
|
4281 /* If the operand matches, we fail. Otherwise backup and
|
|
4282 * continue with the next item. */
|
|
4283 reg_save(&save);
|
|
4284 if (regmatch(OPERAND(scan)))
|
|
4285 return FALSE;
|
|
4286 reg_restore(&save);
|
|
4287 }
|
|
4288 break;
|
|
4289
|
|
4290 case MATCH:
|
|
4291 case SUBPAT:
|
|
4292 {
|
|
4293 regsave_T save;
|
|
4294
|
|
4295 /* If the operand doesn't match, we fail. Otherwise backup
|
|
4296 * and continue with the next item. */
|
|
4297 reg_save(&save);
|
|
4298 if (!regmatch(OPERAND(scan)))
|
|
4299 return FALSE;
|
|
4300 if (op == MATCH) /* zero-width */
|
|
4301 reg_restore(&save);
|
|
4302 }
|
|
4303 break;
|
|
4304
|
|
4305 case BEHIND:
|
|
4306 case NOBEHIND:
|
|
4307 {
|
|
4308 regsave_T save_after, save_start;
|
|
4309 regsave_T save_behind_pos;
|
|
4310 int needmatch = (op == BEHIND);
|
|
4311
|
|
4312 /*
|
|
4313 * Look back in the input of the operand matches or not. This
|
|
4314 * must be done at every position in the input and checking if
|
|
4315 * the match ends at the current position.
|
|
4316 * First check if the next item matches, that's probably
|
|
4317 * faster.
|
|
4318 */
|
|
4319 reg_save(&save_start);
|
|
4320 if (regmatch(next))
|
|
4321 {
|
|
4322 /* save the position after the found match for next */
|
|
4323 reg_save(&save_after);
|
|
4324
|
|
4325 /* start looking for a match with operand at the current
|
|
4326 * postion. Go back one character until we find the
|
|
4327 * result, hitting the start of the line or the previous
|
|
4328 * line (for multi-line matching).
|
|
4329 * Set behind_pos to where the match should end, BHPOS
|
|
4330 * will match it. */
|
|
4331 save_behind_pos = behind_pos;
|
|
4332 behind_pos = save_start;
|
|
4333 for (;;)
|
|
4334 {
|
|
4335 reg_restore(&save_start);
|
|
4336 if (regmatch(OPERAND(scan))
|
|
4337 && reg_save_equal(&behind_pos))
|
|
4338 {
|
|
4339 behind_pos = save_behind_pos;
|
|
4340 /* found a match that ends where "next" started */
|
|
4341 if (needmatch)
|
|
4342 {
|
|
4343 reg_restore(&save_after);
|
|
4344 return TRUE;
|
|
4345 }
|
|
4346 return FALSE;
|
|
4347 }
|
|
4348 /*
|
|
4349 * No match: Go back one character. May go to
|
|
4350 * previous line once.
|
|
4351 */
|
|
4352 if (REG_MULTI)
|
|
4353 {
|
|
4354 if (save_start.rs_u.pos.col == 0)
|
|
4355 {
|
|
4356 if (save_start.rs_u.pos.lnum
|
|
4357 < behind_pos.rs_u.pos.lnum
|
|
4358 || reg_getline(
|
|
4359 --save_start.rs_u.pos.lnum) == NULL)
|
|
4360 break;
|
|
4361 reg_restore(&save_start);
|
|
4362 save_start.rs_u.pos.col =
|
|
4363 (colnr_T)STRLEN(regline);
|
|
4364 }
|
|
4365 else
|
|
4366 --save_start.rs_u.pos.col;
|
|
4367 }
|
|
4368 else
|
|
4369 {
|
|
4370 if (save_start.rs_u.ptr == regline)
|
|
4371 break;
|
|
4372 --save_start.rs_u.ptr;
|
|
4373 }
|
|
4374 }
|
|
4375
|
|
4376 /* NOBEHIND succeeds when no match was found */
|
|
4377 behind_pos = save_behind_pos;
|
|
4378 if (!needmatch)
|
|
4379 {
|
|
4380 reg_restore(&save_after);
|
|
4381 return TRUE;
|
|
4382 }
|
|
4383 }
|
|
4384 return FALSE;
|
|
4385 }
|
|
4386
|
|
4387 case BHPOS:
|
|
4388 if (REG_MULTI)
|
|
4389 {
|
|
4390 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline)
|
|
4391 || behind_pos.rs_u.pos.lnum != reglnum)
|
|
4392 return FALSE;
|
|
4393 }
|
|
4394 else if (behind_pos.rs_u.ptr != reginput)
|
|
4395 return FALSE;
|
|
4396 break;
|
|
4397
|
|
4398 case NEWL:
|
|
4399 if ((c != NUL || reglnum == reg_maxline)
|
|
4400 && (c != '\n' || !reg_line_lbr))
|
|
4401 return FALSE;
|
|
4402 if (reg_line_lbr)
|
|
4403 ADVANCE_REGINPUT();
|
|
4404 else
|
|
4405 reg_nextline();
|
|
4406 break;
|
|
4407
|
|
4408 case END:
|
|
4409 return TRUE; /* Success! */
|
|
4410
|
|
4411 default:
|
|
4412 EMSG(_(e_re_corr));
|
|
4413 #ifdef DEBUG
|
|
4414 printf("Illegal op code %d\n", op);
|
|
4415 #endif
|
|
4416 return FALSE;
|
|
4417 }
|
|
4418 }
|
|
4419
|
|
4420 scan = next;
|
|
4421 }
|
|
4422
|
|
4423 /*
|
|
4424 * We get here only if there's trouble -- normally "case END" is the
|
|
4425 * terminating point.
|
|
4426 */
|
|
4427 EMSG(_(e_re_corr));
|
|
4428 #ifdef DEBUG
|
|
4429 printf("Premature EOL\n");
|
|
4430 #endif
|
|
4431 return FALSE;
|
|
4432 }
|
|
4433
|
|
4434 /*
|
|
4435 * regrepeat - repeatedly match something simple, return how many.
|
|
4436 * Advances reginput (and reglnum) to just after the matched chars.
|
|
4437 */
|
|
4438 static int
|
|
4439 regrepeat(p, maxcount)
|
|
4440 char_u *p;
|
|
4441 long maxcount; /* maximum number of matches allowed */
|
|
4442 {
|
|
4443 long count = 0;
|
|
4444 char_u *scan;
|
|
4445 char_u *opnd;
|
|
4446 int mask;
|
|
4447 int testval = 0;
|
|
4448
|
|
4449 scan = reginput; /* Make local copy of reginput for speed. */
|
|
4450 opnd = OPERAND(p);
|
|
4451 switch (OP(p))
|
|
4452 {
|
|
4453 case ANY:
|
|
4454 case ANY + ADD_NL:
|
|
4455 while (count < maxcount)
|
|
4456 {
|
|
4457 /* Matching anything means we continue until end-of-line (or
|
|
4458 * end-of-file for ANY + ADD_NL), only limited by maxcount. */
|
|
4459 while (*scan != NUL && count < maxcount)
|
|
4460 {
|
|
4461 ++count;
|
39
|
4462 mb_ptr_adv(scan);
|
7
|
4463 }
|
|
4464 if (!WITH_NL(OP(p)) || reglnum == reg_maxline || count == maxcount)
|
|
4465 break;
|
|
4466 ++count; /* count the line-break */
|
|
4467 reg_nextline();
|
|
4468 scan = reginput;
|
|
4469 if (got_int)
|
|
4470 break;
|
|
4471 }
|
|
4472 break;
|
|
4473
|
|
4474 case IDENT:
|
|
4475 case IDENT + ADD_NL:
|
|
4476 testval = TRUE;
|
|
4477 /*FALLTHROUGH*/
|
|
4478 case SIDENT:
|
|
4479 case SIDENT + ADD_NL:
|
|
4480 while (count < maxcount)
|
|
4481 {
|
|
4482 if (vim_isIDc(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
4483 {
|
39
|
4484 mb_ptr_adv(scan);
|
7
|
4485 }
|
|
4486 else if (*scan == NUL)
|
|
4487 {
|
|
4488 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4489 break;
|
|
4490 reg_nextline();
|
|
4491 scan = reginput;
|
|
4492 if (got_int)
|
|
4493 break;
|
|
4494 }
|
|
4495 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4496 ++scan;
|
|
4497 else
|
|
4498 break;
|
|
4499 ++count;
|
|
4500 }
|
|
4501 break;
|
|
4502
|
|
4503 case KWORD:
|
|
4504 case KWORD + ADD_NL:
|
|
4505 testval = TRUE;
|
|
4506 /*FALLTHROUGH*/
|
|
4507 case SKWORD:
|
|
4508 case SKWORD + ADD_NL:
|
|
4509 while (count < maxcount)
|
|
4510 {
|
|
4511 if (vim_iswordp(scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
4512 {
|
39
|
4513 mb_ptr_adv(scan);
|
7
|
4514 }
|
|
4515 else if (*scan == NUL)
|
|
4516 {
|
|
4517 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4518 break;
|
|
4519 reg_nextline();
|
|
4520 scan = reginput;
|
|
4521 if (got_int)
|
|
4522 break;
|
|
4523 }
|
|
4524 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4525 ++scan;
|
|
4526 else
|
|
4527 break;
|
|
4528 ++count;
|
|
4529 }
|
|
4530 break;
|
|
4531
|
|
4532 case FNAME:
|
|
4533 case FNAME + ADD_NL:
|
|
4534 testval = TRUE;
|
|
4535 /*FALLTHROUGH*/
|
|
4536 case SFNAME:
|
|
4537 case SFNAME + ADD_NL:
|
|
4538 while (count < maxcount)
|
|
4539 {
|
|
4540 if (vim_isfilec(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
4541 {
|
39
|
4542 mb_ptr_adv(scan);
|
7
|
4543 }
|
|
4544 else if (*scan == NUL)
|
|
4545 {
|
|
4546 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4547 break;
|
|
4548 reg_nextline();
|
|
4549 scan = reginput;
|
|
4550 if (got_int)
|
|
4551 break;
|
|
4552 }
|
|
4553 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4554 ++scan;
|
|
4555 else
|
|
4556 break;
|
|
4557 ++count;
|
|
4558 }
|
|
4559 break;
|
|
4560
|
|
4561 case PRINT:
|
|
4562 case PRINT + ADD_NL:
|
|
4563 testval = TRUE;
|
|
4564 /*FALLTHROUGH*/
|
|
4565 case SPRINT:
|
|
4566 case SPRINT + ADD_NL:
|
|
4567 while (count < maxcount)
|
|
4568 {
|
|
4569 if (*scan == NUL)
|
|
4570 {
|
|
4571 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4572 break;
|
|
4573 reg_nextline();
|
|
4574 scan = reginput;
|
|
4575 if (got_int)
|
|
4576 break;
|
|
4577 }
|
|
4578 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan)))
|
|
4579 {
|
39
|
4580 mb_ptr_adv(scan);
|
7
|
4581 }
|
|
4582 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4583 ++scan;
|
|
4584 else
|
|
4585 break;
|
|
4586 ++count;
|
|
4587 }
|
|
4588 break;
|
|
4589
|
|
4590 case WHITE:
|
|
4591 case WHITE + ADD_NL:
|
|
4592 testval = mask = RI_WHITE;
|
|
4593 do_class:
|
|
4594 while (count < maxcount)
|
|
4595 {
|
|
4596 #ifdef FEAT_MBYTE
|
|
4597 int l;
|
|
4598 #endif
|
|
4599 if (*scan == NUL)
|
|
4600 {
|
|
4601 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4602 break;
|
|
4603 reg_nextline();
|
|
4604 scan = reginput;
|
|
4605 if (got_int)
|
|
4606 break;
|
|
4607 }
|
|
4608 #ifdef FEAT_MBYTE
|
|
4609 else if (has_mbyte && (l = (*mb_ptr2len_check)(scan)) > 1)
|
|
4610 {
|
|
4611 if (testval != 0)
|
|
4612 break;
|
|
4613 scan += l;
|
|
4614 }
|
|
4615 #endif
|
|
4616 else if ((class_tab[*scan] & mask) == testval)
|
|
4617 ++scan;
|
|
4618 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4619 ++scan;
|
|
4620 else
|
|
4621 break;
|
|
4622 ++count;
|
|
4623 }
|
|
4624 break;
|
|
4625
|
|
4626 case NWHITE:
|
|
4627 case NWHITE + ADD_NL:
|
|
4628 mask = RI_WHITE;
|
|
4629 goto do_class;
|
|
4630 case DIGIT:
|
|
4631 case DIGIT + ADD_NL:
|
|
4632 testval = mask = RI_DIGIT;
|
|
4633 goto do_class;
|
|
4634 case NDIGIT:
|
|
4635 case NDIGIT + ADD_NL:
|
|
4636 mask = RI_DIGIT;
|
|
4637 goto do_class;
|
|
4638 case HEX:
|
|
4639 case HEX + ADD_NL:
|
|
4640 testval = mask = RI_HEX;
|
|
4641 goto do_class;
|
|
4642 case NHEX:
|
|
4643 case NHEX + ADD_NL:
|
|
4644 mask = RI_HEX;
|
|
4645 goto do_class;
|
|
4646 case OCTAL:
|
|
4647 case OCTAL + ADD_NL:
|
|
4648 testval = mask = RI_OCTAL;
|
|
4649 goto do_class;
|
|
4650 case NOCTAL:
|
|
4651 case NOCTAL + ADD_NL:
|
|
4652 mask = RI_OCTAL;
|
|
4653 goto do_class;
|
|
4654 case WORD:
|
|
4655 case WORD + ADD_NL:
|
|
4656 testval = mask = RI_WORD;
|
|
4657 goto do_class;
|
|
4658 case NWORD:
|
|
4659 case NWORD + ADD_NL:
|
|
4660 mask = RI_WORD;
|
|
4661 goto do_class;
|
|
4662 case HEAD:
|
|
4663 case HEAD + ADD_NL:
|
|
4664 testval = mask = RI_HEAD;
|
|
4665 goto do_class;
|
|
4666 case NHEAD:
|
|
4667 case NHEAD + ADD_NL:
|
|
4668 mask = RI_HEAD;
|
|
4669 goto do_class;
|
|
4670 case ALPHA:
|
|
4671 case ALPHA + ADD_NL:
|
|
4672 testval = mask = RI_ALPHA;
|
|
4673 goto do_class;
|
|
4674 case NALPHA:
|
|
4675 case NALPHA + ADD_NL:
|
|
4676 mask = RI_ALPHA;
|
|
4677 goto do_class;
|
|
4678 case LOWER:
|
|
4679 case LOWER + ADD_NL:
|
|
4680 testval = mask = RI_LOWER;
|
|
4681 goto do_class;
|
|
4682 case NLOWER:
|
|
4683 case NLOWER + ADD_NL:
|
|
4684 mask = RI_LOWER;
|
|
4685 goto do_class;
|
|
4686 case UPPER:
|
|
4687 case UPPER + ADD_NL:
|
|
4688 testval = mask = RI_UPPER;
|
|
4689 goto do_class;
|
|
4690 case NUPPER:
|
|
4691 case NUPPER + ADD_NL:
|
|
4692 mask = RI_UPPER;
|
|
4693 goto do_class;
|
|
4694
|
|
4695 case EXACTLY:
|
|
4696 {
|
|
4697 int cu, cl;
|
|
4698
|
|
4699 /* This doesn't do a multi-byte character, because a MULTIBYTECODE
|
|
4700 * would have been used for it. */
|
|
4701 if (ireg_ic)
|
|
4702 {
|
|
4703 cu = TOUPPER_LOC(*opnd);
|
|
4704 cl = TOLOWER_LOC(*opnd);
|
|
4705 while (count < maxcount && (*scan == cu || *scan == cl))
|
|
4706 {
|
|
4707 count++;
|
|
4708 scan++;
|
|
4709 }
|
|
4710 }
|
|
4711 else
|
|
4712 {
|
|
4713 cu = *opnd;
|
|
4714 while (count < maxcount && *scan == cu)
|
|
4715 {
|
|
4716 count++;
|
|
4717 scan++;
|
|
4718 }
|
|
4719 }
|
|
4720 break;
|
|
4721 }
|
|
4722
|
|
4723 #ifdef FEAT_MBYTE
|
|
4724 case MULTIBYTECODE:
|
|
4725 {
|
|
4726 int i, len, cf = 0;
|
|
4727
|
|
4728 /* Safety check (just in case 'encoding' was changed since
|
|
4729 * compiling the program). */
|
|
4730 if ((len = (*mb_ptr2len_check)(opnd)) > 1)
|
|
4731 {
|
|
4732 if (ireg_ic && enc_utf8)
|
|
4733 cf = utf_fold(utf_ptr2char(opnd));
|
|
4734 while (count < maxcount)
|
|
4735 {
|
|
4736 for (i = 0; i < len; ++i)
|
|
4737 if (opnd[i] != scan[i])
|
|
4738 break;
|
|
4739 if (i < len && (!ireg_ic || !enc_utf8
|
|
4740 || utf_fold(utf_ptr2char(scan)) != cf))
|
|
4741 break;
|
|
4742 scan += len;
|
|
4743 ++count;
|
|
4744 }
|
|
4745 }
|
|
4746 }
|
|
4747 break;
|
|
4748 #endif
|
|
4749
|
|
4750 case ANYOF:
|
|
4751 case ANYOF + ADD_NL:
|
|
4752 testval = TRUE;
|
|
4753 /*FALLTHROUGH*/
|
|
4754
|
|
4755 case ANYBUT:
|
|
4756 case ANYBUT + ADD_NL:
|
|
4757 while (count < maxcount)
|
|
4758 {
|
|
4759 #ifdef FEAT_MBYTE
|
|
4760 int len;
|
|
4761 #endif
|
|
4762 if (*scan == NUL)
|
|
4763 {
|
|
4764 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4765 break;
|
|
4766 reg_nextline();
|
|
4767 scan = reginput;
|
|
4768 if (got_int)
|
|
4769 break;
|
|
4770 }
|
|
4771 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4772 ++scan;
|
|
4773 #ifdef FEAT_MBYTE
|
|
4774 else if (has_mbyte && (len = (*mb_ptr2len_check)(scan)) > 1)
|
|
4775 {
|
|
4776 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
|
|
4777 break;
|
|
4778 scan += len;
|
|
4779 }
|
|
4780 #endif
|
|
4781 else
|
|
4782 {
|
|
4783 if ((cstrchr(opnd, *scan) == NULL) == testval)
|
|
4784 break;
|
|
4785 ++scan;
|
|
4786 }
|
|
4787 ++count;
|
|
4788 }
|
|
4789 break;
|
|
4790
|
|
4791 case NEWL:
|
|
4792 while (count < maxcount
|
|
4793 && ((*scan == NUL && reglnum < reg_maxline)
|
|
4794 || (*scan == '\n' && reg_line_lbr)))
|
|
4795 {
|
|
4796 count++;
|
|
4797 if (reg_line_lbr)
|
|
4798 ADVANCE_REGINPUT();
|
|
4799 else
|
|
4800 reg_nextline();
|
|
4801 scan = reginput;
|
|
4802 if (got_int)
|
|
4803 break;
|
|
4804 }
|
|
4805 break;
|
|
4806
|
|
4807 default: /* Oh dear. Called inappropriately. */
|
|
4808 EMSG(_(e_re_corr));
|
|
4809 #ifdef DEBUG
|
|
4810 printf("Called regrepeat with op code %d\n", OP(p));
|
|
4811 #endif
|
|
4812 break;
|
|
4813 }
|
|
4814
|
|
4815 reginput = scan;
|
|
4816
|
|
4817 return (int)count;
|
|
4818 }
|
|
4819
|
|
4820 /*
|
|
4821 * regnext - dig the "next" pointer out of a node
|
|
4822 */
|
|
4823 static char_u *
|
|
4824 regnext(p)
|
|
4825 char_u *p;
|
|
4826 {
|
|
4827 int offset;
|
|
4828
|
|
4829 if (p == JUST_CALC_SIZE)
|
|
4830 return NULL;
|
|
4831
|
|
4832 offset = NEXT(p);
|
|
4833 if (offset == 0)
|
|
4834 return NULL;
|
|
4835
|
|
4836 if (OP(p) == BACK)
|
|
4837 return p - offset;
|
|
4838 else
|
|
4839 return p + offset;
|
|
4840 }
|
|
4841
|
|
4842 /*
|
|
4843 * Check the regexp program for its magic number.
|
|
4844 * Return TRUE if it's wrong.
|
|
4845 */
|
|
4846 static int
|
|
4847 prog_magic_wrong()
|
|
4848 {
|
|
4849 if (UCHARAT(REG_MULTI
|
|
4850 ? reg_mmatch->regprog->program
|
|
4851 : reg_match->regprog->program) != REGMAGIC)
|
|
4852 {
|
|
4853 EMSG(_(e_re_corr));
|
|
4854 return TRUE;
|
|
4855 }
|
|
4856 return FALSE;
|
|
4857 }
|
|
4858
|
|
4859 /*
|
|
4860 * Cleanup the subexpressions, if this wasn't done yet.
|
|
4861 * This construction is used to clear the subexpressions only when they are
|
|
4862 * used (to increase speed).
|
|
4863 */
|
|
4864 static void
|
|
4865 cleanup_subexpr()
|
|
4866 {
|
|
4867 if (need_clear_subexpr)
|
|
4868 {
|
|
4869 if (REG_MULTI)
|
|
4870 {
|
|
4871 /* Use 0xff to set lnum to -1 */
|
|
4872 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
4873 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
4874 }
|
|
4875 else
|
|
4876 {
|
|
4877 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP);
|
|
4878 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP);
|
|
4879 }
|
|
4880 need_clear_subexpr = FALSE;
|
|
4881 }
|
|
4882 }
|
|
4883
|
|
4884 #ifdef FEAT_SYN_HL
|
|
4885 static void
|
|
4886 cleanup_zsubexpr()
|
|
4887 {
|
|
4888 if (need_clear_zsubexpr)
|
|
4889 {
|
|
4890 if (REG_MULTI)
|
|
4891 {
|
|
4892 /* Use 0xff to set lnum to -1 */
|
|
4893 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
4894 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
4895 }
|
|
4896 else
|
|
4897 {
|
|
4898 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
4899 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
4900 }
|
|
4901 need_clear_zsubexpr = FALSE;
|
|
4902 }
|
|
4903 }
|
|
4904 #endif
|
|
4905
|
|
4906 /*
|
|
4907 * Advance reglnum, regline and reginput to the next line.
|
|
4908 */
|
|
4909 static void
|
|
4910 reg_nextline()
|
|
4911 {
|
|
4912 regline = reg_getline(++reglnum);
|
|
4913 reginput = regline;
|
|
4914 fast_breakcheck();
|
|
4915 }
|
|
4916
|
|
4917 /*
|
|
4918 * Save the input line and position in a regsave_T.
|
|
4919 */
|
|
4920 static void
|
|
4921 reg_save(save)
|
|
4922 regsave_T *save;
|
|
4923 {
|
|
4924 if (REG_MULTI)
|
|
4925 {
|
|
4926 save->rs_u.pos.col = (colnr_T)(reginput - regline);
|
|
4927 save->rs_u.pos.lnum = reglnum;
|
|
4928 }
|
|
4929 else
|
|
4930 save->rs_u.ptr = reginput;
|
|
4931 }
|
|
4932
|
|
4933 /*
|
|
4934 * Restore the input line and position from a regsave_T.
|
|
4935 */
|
|
4936 static void
|
|
4937 reg_restore(save)
|
|
4938 regsave_T *save;
|
|
4939 {
|
|
4940 if (REG_MULTI)
|
|
4941 {
|
|
4942 if (reglnum != save->rs_u.pos.lnum)
|
|
4943 {
|
|
4944 /* only call reg_getline() when the line number changed to save
|
|
4945 * a bit of time */
|
|
4946 reglnum = save->rs_u.pos.lnum;
|
|
4947 regline = reg_getline(reglnum);
|
|
4948 }
|
|
4949 reginput = regline + save->rs_u.pos.col;
|
|
4950 }
|
|
4951 else
|
|
4952 reginput = save->rs_u.ptr;
|
|
4953 }
|
|
4954
|
|
4955 /*
|
|
4956 * Return TRUE if current position is equal to saved position.
|
|
4957 */
|
|
4958 static int
|
|
4959 reg_save_equal(save)
|
|
4960 regsave_T *save;
|
|
4961 {
|
|
4962 if (REG_MULTI)
|
|
4963 return reglnum == save->rs_u.pos.lnum
|
|
4964 && reginput == regline + save->rs_u.pos.col;
|
|
4965 return reginput == save->rs_u.ptr;
|
|
4966 }
|
|
4967
|
|
4968 /*
|
|
4969 * Tentatively set the sub-expression start to the current position (after
|
|
4970 * calling regmatch() they will have changed). Need to save the existing
|
|
4971 * values for when there is no match.
|
|
4972 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
|
|
4973 * depending on REG_MULTI.
|
|
4974 */
|
|
4975 static void
|
|
4976 save_se_multi(savep, posp)
|
|
4977 save_se_T *savep;
|
|
4978 lpos_T *posp;
|
|
4979 {
|
|
4980 savep->se_u.pos = *posp;
|
|
4981 posp->lnum = reglnum;
|
|
4982 posp->col = (colnr_T)(reginput - regline);
|
|
4983 }
|
|
4984
|
|
4985 static void
|
|
4986 save_se_one(savep, pp)
|
|
4987 save_se_T *savep;
|
|
4988 char_u **pp;
|
|
4989 {
|
|
4990 savep->se_u.ptr = *pp;
|
|
4991 *pp = reginput;
|
|
4992 }
|
|
4993
|
|
4994 /*
|
|
4995 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
|
|
4996 */
|
|
4997 static int
|
|
4998 re_num_cmp(val, scan)
|
|
4999 long_u val;
|
|
5000 char_u *scan;
|
|
5001 {
|
|
5002 long_u n = OPERAND_MIN(scan);
|
|
5003
|
|
5004 if (OPERAND_CMP(scan) == '>')
|
|
5005 return val > n;
|
|
5006 if (OPERAND_CMP(scan) == '<')
|
|
5007 return val < n;
|
|
5008 return val == n;
|
|
5009 }
|
|
5010
|
|
5011
|
|
5012 #ifdef DEBUG
|
|
5013
|
|
5014 /*
|
|
5015 * regdump - dump a regexp onto stdout in vaguely comprehensible form
|
|
5016 */
|
|
5017 static void
|
|
5018 regdump(pattern, r)
|
|
5019 char_u *pattern;
|
|
5020 regprog_T *r;
|
|
5021 {
|
|
5022 char_u *s;
|
|
5023 int op = EXACTLY; /* Arbitrary non-END op. */
|
|
5024 char_u *next;
|
|
5025 char_u *end = NULL;
|
|
5026
|
|
5027 printf("\r\nregcomp(%s):\r\n", pattern);
|
|
5028
|
|
5029 s = r->program + 1;
|
|
5030 /*
|
|
5031 * Loop until we find the END that isn't before a referred next (an END
|
|
5032 * can also appear in a NOMATCH operand).
|
|
5033 */
|
|
5034 while (op != END || s <= end)
|
|
5035 {
|
|
5036 op = OP(s);
|
|
5037 printf("%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */
|
|
5038 next = regnext(s);
|
|
5039 if (next == NULL) /* Next ptr. */
|
|
5040 printf("(0)");
|
|
5041 else
|
|
5042 printf("(%d)", (int)((s - r->program) + (next - s)));
|
|
5043 if (end < next)
|
|
5044 end = next;
|
|
5045 if (op == BRACE_LIMITS)
|
|
5046 {
|
|
5047 /* Two short ints */
|
|
5048 printf(" minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
|
|
5049 s += 8;
|
|
5050 }
|
|
5051 s += 3;
|
|
5052 if (op == ANYOF || op == ANYOF + ADD_NL
|
|
5053 || op == ANYBUT || op == ANYBUT + ADD_NL
|
|
5054 || op == EXACTLY)
|
|
5055 {
|
|
5056 /* Literal string, where present. */
|
|
5057 while (*s != NUL)
|
|
5058 printf("%c", *s++);
|
|
5059 s++;
|
|
5060 }
|
|
5061 printf("\r\n");
|
|
5062 }
|
|
5063
|
|
5064 /* Header fields of interest. */
|
|
5065 if (r->regstart != NUL)
|
|
5066 printf("start `%s' 0x%x; ", r->regstart < 256
|
|
5067 ? (char *)transchar(r->regstart)
|
|
5068 : "multibyte", r->regstart);
|
|
5069 if (r->reganch)
|
|
5070 printf("anchored; ");
|
|
5071 if (r->regmust != NULL)
|
|
5072 printf("must have \"%s\"", r->regmust);
|
|
5073 printf("\r\n");
|
|
5074 }
|
|
5075
|
|
5076 /*
|
|
5077 * regprop - printable representation of opcode
|
|
5078 */
|
|
5079 static char_u *
|
|
5080 regprop(op)
|
|
5081 char_u *op;
|
|
5082 {
|
|
5083 char_u *p;
|
|
5084 static char_u buf[50];
|
|
5085
|
|
5086 (void) strcpy(buf, ":");
|
|
5087
|
|
5088 switch (OP(op))
|
|
5089 {
|
|
5090 case BOL:
|
|
5091 p = "BOL";
|
|
5092 break;
|
|
5093 case EOL:
|
|
5094 p = "EOL";
|
|
5095 break;
|
|
5096 case RE_BOF:
|
|
5097 p = "BOF";
|
|
5098 break;
|
|
5099 case RE_EOF:
|
|
5100 p = "EOF";
|
|
5101 break;
|
|
5102 case CURSOR:
|
|
5103 p = "CURSOR";
|
|
5104 break;
|
|
5105 case RE_LNUM:
|
|
5106 p = "RE_LNUM";
|
|
5107 break;
|
|
5108 case RE_COL:
|
|
5109 p = "RE_COL";
|
|
5110 break;
|
|
5111 case RE_VCOL:
|
|
5112 p = "RE_VCOL";
|
|
5113 break;
|
|
5114 case BOW:
|
|
5115 p = "BOW";
|
|
5116 break;
|
|
5117 case EOW:
|
|
5118 p = "EOW";
|
|
5119 break;
|
|
5120 case ANY:
|
|
5121 p = "ANY";
|
|
5122 break;
|
|
5123 case ANY + ADD_NL:
|
|
5124 p = "ANY+NL";
|
|
5125 break;
|
|
5126 case ANYOF:
|
|
5127 p = "ANYOF";
|
|
5128 break;
|
|
5129 case ANYOF + ADD_NL:
|
|
5130 p = "ANYOF+NL";
|
|
5131 break;
|
|
5132 case ANYBUT:
|
|
5133 p = "ANYBUT";
|
|
5134 break;
|
|
5135 case ANYBUT + ADD_NL:
|
|
5136 p = "ANYBUT+NL";
|
|
5137 break;
|
|
5138 case IDENT:
|
|
5139 p = "IDENT";
|
|
5140 break;
|
|
5141 case IDENT + ADD_NL:
|
|
5142 p = "IDENT+NL";
|
|
5143 break;
|
|
5144 case SIDENT:
|
|
5145 p = "SIDENT";
|
|
5146 break;
|
|
5147 case SIDENT + ADD_NL:
|
|
5148 p = "SIDENT+NL";
|
|
5149 break;
|
|
5150 case KWORD:
|
|
5151 p = "KWORD";
|
|
5152 break;
|
|
5153 case KWORD + ADD_NL:
|
|
5154 p = "KWORD+NL";
|
|
5155 break;
|
|
5156 case SKWORD:
|
|
5157 p = "SKWORD";
|
|
5158 break;
|
|
5159 case SKWORD + ADD_NL:
|
|
5160 p = "SKWORD+NL";
|
|
5161 break;
|
|
5162 case FNAME:
|
|
5163 p = "FNAME";
|
|
5164 break;
|
|
5165 case FNAME + ADD_NL:
|
|
5166 p = "FNAME+NL";
|
|
5167 break;
|
|
5168 case SFNAME:
|
|
5169 p = "SFNAME";
|
|
5170 break;
|
|
5171 case SFNAME + ADD_NL:
|
|
5172 p = "SFNAME+NL";
|
|
5173 break;
|
|
5174 case PRINT:
|
|
5175 p = "PRINT";
|
|
5176 break;
|
|
5177 case PRINT + ADD_NL:
|
|
5178 p = "PRINT+NL";
|
|
5179 break;
|
|
5180 case SPRINT:
|
|
5181 p = "SPRINT";
|
|
5182 break;
|
|
5183 case SPRINT + ADD_NL:
|
|
5184 p = "SPRINT+NL";
|
|
5185 break;
|
|
5186 case WHITE:
|
|
5187 p = "WHITE";
|
|
5188 break;
|
|
5189 case WHITE + ADD_NL:
|
|
5190 p = "WHITE+NL";
|
|
5191 break;
|
|
5192 case NWHITE:
|
|
5193 p = "NWHITE";
|
|
5194 break;
|
|
5195 case NWHITE + ADD_NL:
|
|
5196 p = "NWHITE+NL";
|
|
5197 break;
|
|
5198 case DIGIT:
|
|
5199 p = "DIGIT";
|
|
5200 break;
|
|
5201 case DIGIT + ADD_NL:
|
|
5202 p = "DIGIT+NL";
|
|
5203 break;
|
|
5204 case NDIGIT:
|
|
5205 p = "NDIGIT";
|
|
5206 break;
|
|
5207 case NDIGIT + ADD_NL:
|
|
5208 p = "NDIGIT+NL";
|
|
5209 break;
|
|
5210 case HEX:
|
|
5211 p = "HEX";
|
|
5212 break;
|
|
5213 case HEX + ADD_NL:
|
|
5214 p = "HEX+NL";
|
|
5215 break;
|
|
5216 case NHEX:
|
|
5217 p = "NHEX";
|
|
5218 break;
|
|
5219 case NHEX + ADD_NL:
|
|
5220 p = "NHEX+NL";
|
|
5221 break;
|
|
5222 case OCTAL:
|
|
5223 p = "OCTAL";
|
|
5224 break;
|
|
5225 case OCTAL + ADD_NL:
|
|
5226 p = "OCTAL+NL";
|
|
5227 break;
|
|
5228 case NOCTAL:
|
|
5229 p = "NOCTAL";
|
|
5230 break;
|
|
5231 case NOCTAL + ADD_NL:
|
|
5232 p = "NOCTAL+NL";
|
|
5233 break;
|
|
5234 case WORD:
|
|
5235 p = "WORD";
|
|
5236 break;
|
|
5237 case WORD + ADD_NL:
|
|
5238 p = "WORD+NL";
|
|
5239 break;
|
|
5240 case NWORD:
|
|
5241 p = "NWORD";
|
|
5242 break;
|
|
5243 case NWORD + ADD_NL:
|
|
5244 p = "NWORD+NL";
|
|
5245 break;
|
|
5246 case HEAD:
|
|
5247 p = "HEAD";
|
|
5248 break;
|
|
5249 case HEAD + ADD_NL:
|
|
5250 p = "HEAD+NL";
|
|
5251 break;
|
|
5252 case NHEAD:
|
|
5253 p = "NHEAD";
|
|
5254 break;
|
|
5255 case NHEAD + ADD_NL:
|
|
5256 p = "NHEAD+NL";
|
|
5257 break;
|
|
5258 case ALPHA:
|
|
5259 p = "ALPHA";
|
|
5260 break;
|
|
5261 case ALPHA + ADD_NL:
|
|
5262 p = "ALPHA+NL";
|
|
5263 break;
|
|
5264 case NALPHA:
|
|
5265 p = "NALPHA";
|
|
5266 break;
|
|
5267 case NALPHA + ADD_NL:
|
|
5268 p = "NALPHA+NL";
|
|
5269 break;
|
|
5270 case LOWER:
|
|
5271 p = "LOWER";
|
|
5272 break;
|
|
5273 case LOWER + ADD_NL:
|
|
5274 p = "LOWER+NL";
|
|
5275 break;
|
|
5276 case NLOWER:
|
|
5277 p = "NLOWER";
|
|
5278 break;
|
|
5279 case NLOWER + ADD_NL:
|
|
5280 p = "NLOWER+NL";
|
|
5281 break;
|
|
5282 case UPPER:
|
|
5283 p = "UPPER";
|
|
5284 break;
|
|
5285 case UPPER + ADD_NL:
|
|
5286 p = "UPPER+NL";
|
|
5287 break;
|
|
5288 case NUPPER:
|
|
5289 p = "NUPPER";
|
|
5290 break;
|
|
5291 case NUPPER + ADD_NL:
|
|
5292 p = "NUPPER+NL";
|
|
5293 break;
|
|
5294 case BRANCH:
|
|
5295 p = "BRANCH";
|
|
5296 break;
|
|
5297 case EXACTLY:
|
|
5298 p = "EXACTLY";
|
|
5299 break;
|
|
5300 case NOTHING:
|
|
5301 p = "NOTHING";
|
|
5302 break;
|
|
5303 case BACK:
|
|
5304 p = "BACK";
|
|
5305 break;
|
|
5306 case END:
|
|
5307 p = "END";
|
|
5308 break;
|
|
5309 case MOPEN + 0:
|
|
5310 p = "MATCH START";
|
|
5311 break;
|
|
5312 case MOPEN + 1:
|
|
5313 case MOPEN + 2:
|
|
5314 case MOPEN + 3:
|
|
5315 case MOPEN + 4:
|
|
5316 case MOPEN + 5:
|
|
5317 case MOPEN + 6:
|
|
5318 case MOPEN + 7:
|
|
5319 case MOPEN + 8:
|
|
5320 case MOPEN + 9:
|
|
5321 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
|
|
5322 p = NULL;
|
|
5323 break;
|
|
5324 case MCLOSE + 0:
|
|
5325 p = "MATCH END";
|
|
5326 break;
|
|
5327 case MCLOSE + 1:
|
|
5328 case MCLOSE + 2:
|
|
5329 case MCLOSE + 3:
|
|
5330 case MCLOSE + 4:
|
|
5331 case MCLOSE + 5:
|
|
5332 case MCLOSE + 6:
|
|
5333 case MCLOSE + 7:
|
|
5334 case MCLOSE + 8:
|
|
5335 case MCLOSE + 9:
|
|
5336 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
|
|
5337 p = NULL;
|
|
5338 break;
|
|
5339 case BACKREF + 1:
|
|
5340 case BACKREF + 2:
|
|
5341 case BACKREF + 3:
|
|
5342 case BACKREF + 4:
|
|
5343 case BACKREF + 5:
|
|
5344 case BACKREF + 6:
|
|
5345 case BACKREF + 7:
|
|
5346 case BACKREF + 8:
|
|
5347 case BACKREF + 9:
|
|
5348 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
|
|
5349 p = NULL;
|
|
5350 break;
|
|
5351 case NOPEN:
|
|
5352 p = "NOPEN";
|
|
5353 break;
|
|
5354 case NCLOSE:
|
|
5355 p = "NCLOSE";
|
|
5356 break;
|
|
5357 #ifdef FEAT_SYN_HL
|
|
5358 case ZOPEN + 1:
|
|
5359 case ZOPEN + 2:
|
|
5360 case ZOPEN + 3:
|
|
5361 case ZOPEN + 4:
|
|
5362 case ZOPEN + 5:
|
|
5363 case ZOPEN + 6:
|
|
5364 case ZOPEN + 7:
|
|
5365 case ZOPEN + 8:
|
|
5366 case ZOPEN + 9:
|
|
5367 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
|
|
5368 p = NULL;
|
|
5369 break;
|
|
5370 case ZCLOSE + 1:
|
|
5371 case ZCLOSE + 2:
|
|
5372 case ZCLOSE + 3:
|
|
5373 case ZCLOSE + 4:
|
|
5374 case ZCLOSE + 5:
|
|
5375 case ZCLOSE + 6:
|
|
5376 case ZCLOSE + 7:
|
|
5377 case ZCLOSE + 8:
|
|
5378 case ZCLOSE + 9:
|
|
5379 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
|
|
5380 p = NULL;
|
|
5381 break;
|
|
5382 case ZREF + 1:
|
|
5383 case ZREF + 2:
|
|
5384 case ZREF + 3:
|
|
5385 case ZREF + 4:
|
|
5386 case ZREF + 5:
|
|
5387 case ZREF + 6:
|
|
5388 case ZREF + 7:
|
|
5389 case ZREF + 8:
|
|
5390 case ZREF + 9:
|
|
5391 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
|
|
5392 p = NULL;
|
|
5393 break;
|
|
5394 #endif
|
|
5395 case STAR:
|
|
5396 p = "STAR";
|
|
5397 break;
|
|
5398 case PLUS:
|
|
5399 p = "PLUS";
|
|
5400 break;
|
|
5401 case NOMATCH:
|
|
5402 p = "NOMATCH";
|
|
5403 break;
|
|
5404 case MATCH:
|
|
5405 p = "MATCH";
|
|
5406 break;
|
|
5407 case BEHIND:
|
|
5408 p = "BEHIND";
|
|
5409 break;
|
|
5410 case NOBEHIND:
|
|
5411 p = "NOBEHIND";
|
|
5412 break;
|
|
5413 case SUBPAT:
|
|
5414 p = "SUBPAT";
|
|
5415 break;
|
|
5416 case BRACE_LIMITS:
|
|
5417 p = "BRACE_LIMITS";
|
|
5418 break;
|
|
5419 case BRACE_SIMPLE:
|
|
5420 p = "BRACE_SIMPLE";
|
|
5421 break;
|
|
5422 case BRACE_COMPLEX + 0:
|
|
5423 case BRACE_COMPLEX + 1:
|
|
5424 case BRACE_COMPLEX + 2:
|
|
5425 case BRACE_COMPLEX + 3:
|
|
5426 case BRACE_COMPLEX + 4:
|
|
5427 case BRACE_COMPLEX + 5:
|
|
5428 case BRACE_COMPLEX + 6:
|
|
5429 case BRACE_COMPLEX + 7:
|
|
5430 case BRACE_COMPLEX + 8:
|
|
5431 case BRACE_COMPLEX + 9:
|
|
5432 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
|
|
5433 p = NULL;
|
|
5434 break;
|
|
5435 #ifdef FEAT_MBYTE
|
|
5436 case MULTIBYTECODE:
|
|
5437 p = "MULTIBYTECODE";
|
|
5438 break;
|
|
5439 #endif
|
|
5440 case NEWL:
|
|
5441 p = "NEWL";
|
|
5442 break;
|
|
5443 default:
|
|
5444 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
|
|
5445 p = NULL;
|
|
5446 break;
|
|
5447 }
|
|
5448 if (p != NULL)
|
|
5449 (void) strcat(buf, p);
|
|
5450 return buf;
|
|
5451 }
|
|
5452 #endif
|
|
5453
|
|
5454 #ifdef FEAT_MBYTE
|
|
5455 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3));
|
|
5456
|
|
5457 typedef struct
|
|
5458 {
|
|
5459 int a, b, c;
|
|
5460 } decomp_T;
|
|
5461
|
|
5462
|
|
5463 /* 0xfb20 - 0xfb4f */
|
|
5464 decomp_T decomp_table[0xfb4f-0xfb20+1] =
|
|
5465 {
|
|
5466 {0x5e2,0,0}, /* 0xfb20 alt ayin */
|
|
5467 {0x5d0,0,0}, /* 0xfb21 alt alef */
|
|
5468 {0x5d3,0,0}, /* 0xfb22 alt dalet */
|
|
5469 {0x5d4,0,0}, /* 0xfb23 alt he */
|
|
5470 {0x5db,0,0}, /* 0xfb24 alt kaf */
|
|
5471 {0x5dc,0,0}, /* 0xfb25 alt lamed */
|
|
5472 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */
|
|
5473 {0x5e8,0,0}, /* 0xfb27 alt resh */
|
|
5474 {0x5ea,0,0}, /* 0xfb28 alt tav */
|
|
5475 {'+', 0, 0}, /* 0xfb29 alt plus */
|
|
5476 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */
|
|
5477 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */
|
|
5478 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */
|
|
5479 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */
|
|
5480 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */
|
|
5481 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */
|
|
5482 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */
|
|
5483 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */
|
|
5484 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */
|
|
5485 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */
|
|
5486 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */
|
|
5487 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */
|
|
5488 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */
|
|
5489 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */
|
|
5490 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */
|
|
5491 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */
|
|
5492 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */
|
|
5493 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */
|
|
5494 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */
|
|
5495 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */
|
|
5496 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */
|
|
5497 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */
|
|
5498 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */
|
|
5499 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */
|
|
5500 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */
|
|
5501 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */
|
|
5502 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */
|
|
5503 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */
|
|
5504 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */
|
|
5505 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */
|
|
5506 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */
|
|
5507 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */
|
|
5508 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */
|
|
5509 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */
|
|
5510 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */
|
|
5511 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */
|
|
5512 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */
|
|
5513 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */
|
|
5514 };
|
|
5515
|
|
5516 static void
|
|
5517 mb_decompose(c, c1, c2, c3)
|
|
5518 int c, *c1, *c2, *c3;
|
|
5519 {
|
|
5520 decomp_T d;
|
|
5521
|
|
5522 if (c >= 0x4b20 && c <= 0xfb4f)
|
|
5523 {
|
|
5524 d = decomp_table[c - 0xfb20];
|
|
5525 *c1 = d.a;
|
|
5526 *c2 = d.b;
|
|
5527 *c3 = d.c;
|
|
5528 }
|
|
5529 else
|
|
5530 {
|
|
5531 *c1 = c;
|
|
5532 *c2 = *c3 = 0;
|
|
5533 }
|
|
5534 }
|
|
5535 #endif
|
|
5536
|
|
5537 /*
|
|
5538 * Compare two strings, ignore case if ireg_ic set.
|
|
5539 * Return 0 if strings match, non-zero otherwise.
|
|
5540 * Correct the length "*n" when composing characters are ignored.
|
|
5541 */
|
|
5542 static int
|
|
5543 cstrncmp(s1, s2, n)
|
|
5544 char_u *s1, *s2;
|
|
5545 int *n;
|
|
5546 {
|
|
5547 int result;
|
|
5548
|
|
5549 if (!ireg_ic)
|
|
5550 result = STRNCMP(s1, s2, *n);
|
|
5551 else
|
|
5552 result = MB_STRNICMP(s1, s2, *n);
|
|
5553
|
|
5554 #ifdef FEAT_MBYTE
|
|
5555 /* if it failed and it's utf8 and we want to combineignore: */
|
|
5556 if (result != 0 && enc_utf8 && ireg_icombine)
|
|
5557 {
|
|
5558 char_u *str1, *str2;
|
|
5559 int c1, c2, c11, c12;
|
|
5560 int ix;
|
|
5561 int junk;
|
|
5562
|
|
5563 /* we have to handle the strcmp ourselves, since it is necessary to
|
|
5564 * deal with the composing characters by ignoring them: */
|
|
5565 str1 = s1;
|
|
5566 str2 = s2;
|
|
5567 c1 = c2 = 0;
|
|
5568 for (ix = 0; ix < *n; )
|
|
5569 {
|
|
5570 c1 = mb_ptr2char_adv(&str1);
|
|
5571 c2 = mb_ptr2char_adv(&str2);
|
|
5572 ix += utf_char2len(c1);
|
|
5573
|
|
5574 /* decompose the character if necessary, into 'base' characters
|
|
5575 * because I don't care about Arabic, I will hard-code the Hebrew
|
|
5576 * which I *do* care about! So sue me... */
|
|
5577 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2)))
|
|
5578 {
|
|
5579 /* decomposition necessary? */
|
|
5580 mb_decompose(c1, &c11, &junk, &junk);
|
|
5581 mb_decompose(c2, &c12, &junk, &junk);
|
|
5582 c1 = c11;
|
|
5583 c2 = c12;
|
|
5584 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12)))
|
|
5585 break;
|
|
5586 }
|
|
5587 }
|
|
5588 result = c2 - c1;
|
|
5589 if (result == 0)
|
|
5590 *n = (int)(str2 - s2);
|
|
5591 }
|
|
5592 #endif
|
|
5593
|
|
5594 return result;
|
|
5595 }
|
|
5596
|
|
5597 /*
|
|
5598 * cstrchr: This function is used a lot for simple searches, keep it fast!
|
|
5599 */
|
|
5600 static char_u *
|
|
5601 cstrchr(s, c)
|
|
5602 char_u *s;
|
|
5603 int c;
|
|
5604 {
|
|
5605 char_u *p;
|
|
5606 int cc;
|
|
5607
|
|
5608 if (!ireg_ic
|
|
5609 #ifdef FEAT_MBYTE
|
|
5610 || (!enc_utf8 && mb_char2len(c) > 1)
|
|
5611 #endif
|
|
5612 )
|
|
5613 return vim_strchr(s, c);
|
|
5614
|
|
5615 /* tolower() and toupper() can be slow, comparing twice should be a lot
|
|
5616 * faster (esp. when using MS Visual C++!).
|
|
5617 * For UTF-8 need to use folded case. */
|
|
5618 #ifdef FEAT_MBYTE
|
|
5619 if (enc_utf8 && c > 0x80)
|
|
5620 cc = utf_fold(c);
|
|
5621 else
|
|
5622 #endif
|
|
5623 if (isupper(c))
|
|
5624 cc = TOLOWER_LOC(c);
|
|
5625 else if (islower(c))
|
|
5626 cc = TOUPPER_LOC(c);
|
|
5627 else
|
|
5628 return vim_strchr(s, c);
|
|
5629
|
|
5630 #ifdef FEAT_MBYTE
|
|
5631 if (has_mbyte)
|
|
5632 {
|
|
5633 for (p = s; *p != NUL; p += (*mb_ptr2len_check)(p))
|
|
5634 {
|
|
5635 if (enc_utf8 && c > 0x80)
|
|
5636 {
|
|
5637 if (utf_fold(utf_ptr2char(p)) == cc)
|
|
5638 return p;
|
|
5639 }
|
|
5640 else if (*p == c || *p == cc)
|
|
5641 return p;
|
|
5642 }
|
|
5643 }
|
|
5644 else
|
|
5645 #endif
|
|
5646 /* Faster version for when there are no multi-byte characters. */
|
|
5647 for (p = s; *p != NUL; ++p)
|
|
5648 if (*p == c || *p == cc)
|
|
5649 return p;
|
|
5650
|
|
5651 return NULL;
|
|
5652 }
|
|
5653
|
|
5654 /***************************************************************
|
|
5655 * regsub stuff *
|
|
5656 ***************************************************************/
|
|
5657
|
|
5658 /* This stuff below really confuses cc on an SGI -- webb */
|
|
5659 #ifdef __sgi
|
|
5660 # undef __ARGS
|
|
5661 # define __ARGS(x) ()
|
|
5662 #endif
|
|
5663
|
|
5664 /*
|
|
5665 * We should define ftpr as a pointer to a function returning a pointer to
|
|
5666 * a function returning a pointer to a function ...
|
|
5667 * This is impossible, so we declare a pointer to a function returning a
|
|
5668 * pointer to a function returning void. This should work for all compilers.
|
|
5669 */
|
|
5670 typedef void (*(*fptr) __ARGS((char_u *, int)))();
|
|
5671
|
|
5672 static fptr do_upper __ARGS((char_u *, int));
|
|
5673 static fptr do_Upper __ARGS((char_u *, int));
|
|
5674 static fptr do_lower __ARGS((char_u *, int));
|
|
5675 static fptr do_Lower __ARGS((char_u *, int));
|
|
5676
|
|
5677 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash));
|
|
5678
|
|
5679 static fptr
|
|
5680 do_upper(d, c)
|
|
5681 char_u *d;
|
|
5682 int c;
|
|
5683 {
|
|
5684 *d = TOUPPER_LOC(c);
|
|
5685
|
|
5686 return (fptr)NULL;
|
|
5687 }
|
|
5688
|
|
5689 static fptr
|
|
5690 do_Upper(d, c)
|
|
5691 char_u *d;
|
|
5692 int c;
|
|
5693 {
|
|
5694 *d = TOUPPER_LOC(c);
|
|
5695
|
|
5696 return (fptr)do_Upper;
|
|
5697 }
|
|
5698
|
|
5699 static fptr
|
|
5700 do_lower(d, c)
|
|
5701 char_u *d;
|
|
5702 int c;
|
|
5703 {
|
|
5704 *d = TOLOWER_LOC(c);
|
|
5705
|
|
5706 return (fptr)NULL;
|
|
5707 }
|
|
5708
|
|
5709 static fptr
|
|
5710 do_Lower(d, c)
|
|
5711 char_u *d;
|
|
5712 int c;
|
|
5713 {
|
|
5714 *d = TOLOWER_LOC(c);
|
|
5715
|
|
5716 return (fptr)do_Lower;
|
|
5717 }
|
|
5718
|
|
5719 /*
|
|
5720 * regtilde(): Replace tildes in the pattern by the old pattern.
|
|
5721 *
|
|
5722 * Short explanation of the tilde: It stands for the previous replacement
|
|
5723 * pattern. If that previous pattern also contains a ~ we should go back a
|
|
5724 * step further... But we insert the previous pattern into the current one
|
|
5725 * and remember that.
|
|
5726 * This still does not handle the case where "magic" changes. TODO?
|
|
5727 *
|
|
5728 * The tildes are parsed once before the first call to vim_regsub().
|
|
5729 */
|
|
5730 char_u *
|
|
5731 regtilde(source, magic)
|
|
5732 char_u *source;
|
|
5733 int magic;
|
|
5734 {
|
|
5735 char_u *newsub = source;
|
|
5736 char_u *tmpsub;
|
|
5737 char_u *p;
|
|
5738 int len;
|
|
5739 int prevlen;
|
|
5740
|
|
5741 for (p = newsub; *p; ++p)
|
|
5742 {
|
|
5743 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic))
|
|
5744 {
|
|
5745 if (reg_prev_sub != NULL)
|
|
5746 {
|
|
5747 /* length = len(newsub) - 1 + len(prev_sub) + 1 */
|
|
5748 prevlen = (int)STRLEN(reg_prev_sub);
|
|
5749 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen));
|
|
5750 if (tmpsub != NULL)
|
|
5751 {
|
|
5752 /* copy prefix */
|
|
5753 len = (int)(p - newsub); /* not including ~ */
|
|
5754 mch_memmove(tmpsub, newsub, (size_t)len);
|
|
5755 /* interpretate tilde */
|
|
5756 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen);
|
|
5757 /* copy postfix */
|
|
5758 if (!magic)
|
|
5759 ++p; /* back off \ */
|
|
5760 STRCPY(tmpsub + len + prevlen, p + 1);
|
|
5761
|
|
5762 if (newsub != source) /* already allocated newsub */
|
|
5763 vim_free(newsub);
|
|
5764 newsub = tmpsub;
|
|
5765 p = newsub + len + prevlen;
|
|
5766 }
|
|
5767 }
|
|
5768 else if (magic)
|
|
5769 STRCPY(p, p + 1); /* remove '~' */
|
|
5770 else
|
|
5771 STRCPY(p, p + 2); /* remove '\~' */
|
|
5772 --p;
|
|
5773 }
|
|
5774 else
|
|
5775 {
|
|
5776 if (*p == '\\' && p[1]) /* skip escaped characters */
|
|
5777 ++p;
|
|
5778 #ifdef FEAT_MBYTE
|
|
5779 if (has_mbyte)
|
|
5780 p += (*mb_ptr2len_check)(p) - 1;
|
|
5781 #endif
|
|
5782 }
|
|
5783 }
|
|
5784
|
|
5785 vim_free(reg_prev_sub);
|
|
5786 if (newsub != source) /* newsub was allocated, just keep it */
|
|
5787 reg_prev_sub = newsub;
|
|
5788 else /* no ~ found, need to save newsub */
|
|
5789 reg_prev_sub = vim_strsave(newsub);
|
|
5790 return newsub;
|
|
5791 }
|
|
5792
|
|
5793 #ifdef FEAT_EVAL
|
|
5794 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */
|
|
5795
|
|
5796 /* These pointers are used instead of reg_match and reg_mmatch for
|
|
5797 * reg_submatch(). Needed for when the substitution string is an expression
|
|
5798 * that contains a call to substitute() and submatch(). */
|
|
5799 static regmatch_T *submatch_match;
|
|
5800 static regmmatch_T *submatch_mmatch;
|
|
5801 #endif
|
|
5802
|
|
5803 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO)
|
|
5804 /*
|
|
5805 * vim_regsub() - perform substitutions after a vim_regexec() or
|
|
5806 * vim_regexec_multi() match.
|
|
5807 *
|
|
5808 * If "copy" is TRUE really copy into "dest".
|
|
5809 * If "copy" is FALSE nothing is copied, this is just to find out the length
|
|
5810 * of the result.
|
|
5811 *
|
|
5812 * If "backslash" is TRUE, a backslash will be removed later, need to double
|
|
5813 * them to keep them, and insert a backslash before a CR to avoid it being
|
|
5814 * replaced with a line break later.
|
|
5815 *
|
|
5816 * Note: The matched text must not change between the call of
|
|
5817 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back
|
|
5818 * references invalid!
|
|
5819 *
|
|
5820 * Returns the size of the replacement, including terminating NUL.
|
|
5821 */
|
|
5822 int
|
|
5823 vim_regsub(rmp, source, dest, copy, magic, backslash)
|
|
5824 regmatch_T *rmp;
|
|
5825 char_u *source;
|
|
5826 char_u *dest;
|
|
5827 int copy;
|
|
5828 int magic;
|
|
5829 int backslash;
|
|
5830 {
|
|
5831 reg_match = rmp;
|
|
5832 reg_mmatch = NULL;
|
|
5833 reg_maxline = 0;
|
|
5834 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
5835 }
|
|
5836 #endif
|
|
5837
|
|
5838 int
|
|
5839 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash)
|
|
5840 regmmatch_T *rmp;
|
|
5841 linenr_T lnum;
|
|
5842 char_u *source;
|
|
5843 char_u *dest;
|
|
5844 int copy;
|
|
5845 int magic;
|
|
5846 int backslash;
|
|
5847 {
|
|
5848 reg_match = NULL;
|
|
5849 reg_mmatch = rmp;
|
|
5850 reg_buf = curbuf; /* always works on the current buffer! */
|
|
5851 reg_firstlnum = lnum;
|
|
5852 reg_maxline = curbuf->b_ml.ml_line_count - lnum;
|
|
5853 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
5854 }
|
|
5855
|
|
5856 static int
|
|
5857 vim_regsub_both(source, dest, copy, magic, backslash)
|
|
5858 char_u *source;
|
|
5859 char_u *dest;
|
|
5860 int copy;
|
|
5861 int magic;
|
|
5862 int backslash;
|
|
5863 {
|
|
5864 char_u *src;
|
|
5865 char_u *dst;
|
|
5866 char_u *s;
|
|
5867 int c;
|
|
5868 int no = -1;
|
|
5869 fptr func = (fptr)NULL;
|
|
5870 linenr_T clnum = 0; /* init for GCC */
|
|
5871 int len = 0; /* init for GCC */
|
|
5872 #ifdef FEAT_EVAL
|
|
5873 static char_u *eval_result = NULL;
|
|
5874 #endif
|
|
5875 #ifdef FEAT_MBYTE
|
|
5876 int l;
|
|
5877 #endif
|
|
5878
|
|
5879
|
|
5880 /* Be paranoid... */
|
|
5881 if (source == NULL || dest == NULL)
|
|
5882 {
|
|
5883 EMSG(_(e_null));
|
|
5884 return 0;
|
|
5885 }
|
|
5886 if (prog_magic_wrong())
|
|
5887 return 0;
|
|
5888 src = source;
|
|
5889 dst = dest;
|
|
5890
|
|
5891 /*
|
|
5892 * When the substitute part starts with "\=" evaluate it as an expression.
|
|
5893 */
|
|
5894 if (source[0] == '\\' && source[1] == '='
|
|
5895 #ifdef FEAT_EVAL
|
|
5896 && !can_f_submatch /* can't do this recursively */
|
|
5897 #endif
|
|
5898 )
|
|
5899 {
|
|
5900 #ifdef FEAT_EVAL
|
|
5901 /* To make sure that the length doesn't change between checking the
|
|
5902 * length and copying the string, and to speed up things, the
|
|
5903 * resulting string is saved from the call with "copy" == FALSE to the
|
|
5904 * call with "copy" == TRUE. */
|
|
5905 if (copy)
|
|
5906 {
|
|
5907 if (eval_result != NULL)
|
|
5908 {
|
|
5909 STRCPY(dest, eval_result);
|
|
5910 dst += STRLEN(eval_result);
|
|
5911 vim_free(eval_result);
|
|
5912 eval_result = NULL;
|
|
5913 }
|
|
5914 }
|
|
5915 else
|
|
5916 {
|
|
5917 linenr_T save_reg_maxline;
|
|
5918 win_T *save_reg_win;
|
|
5919 int save_ireg_ic;
|
|
5920
|
|
5921 vim_free(eval_result);
|
|
5922
|
|
5923 /* The expression may contain substitute(), which calls us
|
|
5924 * recursively. Make sure submatch() gets the text from the first
|
|
5925 * level. Don't need to save "reg_buf", because
|
|
5926 * vim_regexec_multi() can't be called recursively. */
|
|
5927 submatch_match = reg_match;
|
|
5928 submatch_mmatch = reg_mmatch;
|
|
5929 save_reg_maxline = reg_maxline;
|
|
5930 save_reg_win = reg_win;
|
|
5931 save_ireg_ic = ireg_ic;
|
|
5932 can_f_submatch = TRUE;
|
|
5933
|
|
5934 eval_result = eval_to_string(source + 2, NULL);
|
|
5935 if (eval_result != NULL)
|
|
5936 {
|
39
|
5937 for (s = eval_result; *s != NUL; mb_ptr_adv(s))
|
7
|
5938 {
|
|
5939 /* Change NL to CR, so that it becomes a line break.
|
|
5940 * Skip over a backslashed character. */
|
|
5941 if (*s == NL)
|
|
5942 *s = CAR;
|
|
5943 else if (*s == '\\' && s[1] != NUL)
|
|
5944 ++s;
|
|
5945 }
|
|
5946
|
|
5947 dst += STRLEN(eval_result);
|
|
5948 }
|
|
5949
|
|
5950 reg_match = submatch_match;
|
|
5951 reg_mmatch = submatch_mmatch;
|
|
5952 reg_maxline = save_reg_maxline;
|
|
5953 reg_win = save_reg_win;
|
|
5954 ireg_ic = save_ireg_ic;
|
|
5955 can_f_submatch = FALSE;
|
|
5956 }
|
|
5957 #endif
|
|
5958 }
|
|
5959 else
|
|
5960 while ((c = *src++) != NUL)
|
|
5961 {
|
|
5962 if (c == '&' && magic)
|
|
5963 no = 0;
|
|
5964 else if (c == '\\' && *src != NUL)
|
|
5965 {
|
|
5966 if (*src == '&' && !magic)
|
|
5967 {
|
|
5968 ++src;
|
|
5969 no = 0;
|
|
5970 }
|
|
5971 else if ('0' <= *src && *src <= '9')
|
|
5972 {
|
|
5973 no = *src++ - '0';
|
|
5974 }
|
|
5975 else if (vim_strchr((char_u *)"uUlLeE", *src))
|
|
5976 {
|
|
5977 switch (*src++)
|
|
5978 {
|
|
5979 case 'u': func = (fptr)do_upper;
|
|
5980 continue;
|
|
5981 case 'U': func = (fptr)do_Upper;
|
|
5982 continue;
|
|
5983 case 'l': func = (fptr)do_lower;
|
|
5984 continue;
|
|
5985 case 'L': func = (fptr)do_Lower;
|
|
5986 continue;
|
|
5987 case 'e':
|
|
5988 case 'E': func = (fptr)NULL;
|
|
5989 continue;
|
|
5990 }
|
|
5991 }
|
|
5992 }
|
|
5993 if (no < 0) /* Ordinary character. */
|
|
5994 {
|
|
5995 if (c == '\\' && *src != NUL)
|
|
5996 {
|
|
5997 /* Check for abbreviations -- webb */
|
|
5998 switch (*src)
|
|
5999 {
|
|
6000 case 'r': c = CAR; ++src; break;
|
|
6001 case 'n': c = NL; ++src; break;
|
|
6002 case 't': c = TAB; ++src; break;
|
|
6003 /* Oh no! \e already has meaning in subst pat :-( */
|
|
6004 /* case 'e': c = ESC; ++src; break; */
|
|
6005 case 'b': c = Ctrl_H; ++src; break;
|
|
6006
|
|
6007 /* If "backslash" is TRUE the backslash will be removed
|
|
6008 * later. Used to insert a literal CR. */
|
|
6009 default: if (backslash)
|
|
6010 {
|
|
6011 if (copy)
|
|
6012 *dst = '\\';
|
|
6013 ++dst;
|
|
6014 }
|
|
6015 c = *src++;
|
|
6016 }
|
|
6017 }
|
|
6018
|
|
6019 /* Write to buffer, if copy is set. */
|
|
6020 #ifdef FEAT_MBYTE
|
|
6021 if (has_mbyte && (l = (*mb_ptr2len_check)(src - 1)) > 1)
|
|
6022 {
|
|
6023 /* TODO: should use "func" here. */
|
|
6024 if (copy)
|
|
6025 mch_memmove(dst, src - 1, l);
|
|
6026 dst += l - 1;
|
|
6027 src += l - 1;
|
|
6028 }
|
|
6029 else
|
|
6030 {
|
|
6031 #endif
|
|
6032 if (copy)
|
|
6033 {
|
|
6034 if (func == (fptr)NULL) /* just copy */
|
|
6035 *dst = c;
|
|
6036 else /* change case */
|
|
6037 func = (fptr)(func(dst, c));
|
|
6038 /* Turbo C complains without the typecast */
|
|
6039 }
|
|
6040 #ifdef FEAT_MBYTE
|
|
6041 }
|
|
6042 #endif
|
|
6043 dst++;
|
|
6044 }
|
|
6045 else
|
|
6046 {
|
|
6047 if (REG_MULTI)
|
|
6048 {
|
|
6049 clnum = reg_mmatch->startpos[no].lnum;
|
|
6050 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0)
|
|
6051 s = NULL;
|
|
6052 else
|
|
6053 {
|
|
6054 s = reg_getline(clnum) + reg_mmatch->startpos[no].col;
|
|
6055 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6056 len = reg_mmatch->endpos[no].col
|
|
6057 - reg_mmatch->startpos[no].col;
|
|
6058 else
|
|
6059 len = (int)STRLEN(s);
|
|
6060 }
|
|
6061 }
|
|
6062 else
|
|
6063 {
|
|
6064 s = reg_match->startp[no];
|
|
6065 if (reg_match->endp[no] == NULL)
|
|
6066 s = NULL;
|
|
6067 else
|
|
6068 len = (int)(reg_match->endp[no] - s);
|
|
6069 }
|
|
6070 if (s != NULL)
|
|
6071 {
|
|
6072 for (;;)
|
|
6073 {
|
|
6074 if (len == 0)
|
|
6075 {
|
|
6076 if (REG_MULTI)
|
|
6077 {
|
|
6078 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6079 break;
|
|
6080 if (copy)
|
|
6081 *dst = CAR;
|
|
6082 ++dst;
|
|
6083 s = reg_getline(++clnum);
|
|
6084 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6085 len = reg_mmatch->endpos[no].col;
|
|
6086 else
|
|
6087 len = (int)STRLEN(s);
|
|
6088 }
|
|
6089 else
|
|
6090 break;
|
|
6091 }
|
|
6092 else if (*s == NUL) /* we hit NUL. */
|
|
6093 {
|
|
6094 if (copy)
|
|
6095 EMSG(_(e_re_damg));
|
|
6096 goto exit;
|
|
6097 }
|
|
6098 else
|
|
6099 {
|
|
6100 if (backslash && (*s == CAR || *s == '\\'))
|
|
6101 {
|
|
6102 /*
|
|
6103 * Insert a backslash in front of a CR, otherwise
|
|
6104 * it will be replaced by a line break.
|
|
6105 * Number of backslashes will be halved later,
|
|
6106 * double them here.
|
|
6107 */
|
|
6108 if (copy)
|
|
6109 {
|
|
6110 dst[0] = '\\';
|
|
6111 dst[1] = *s;
|
|
6112 }
|
|
6113 dst += 2;
|
|
6114 }
|
|
6115 #ifdef FEAT_MBYTE
|
|
6116 else if (has_mbyte && (l = (*mb_ptr2len_check)(s)) > 1)
|
|
6117 {
|
|
6118 /* TODO: should use "func" here. */
|
|
6119 if (copy)
|
|
6120 mch_memmove(dst, s, l);
|
|
6121 dst += l;
|
|
6122 s += l - 1;
|
|
6123 len -= l - 1;
|
|
6124 }
|
|
6125 #endif
|
|
6126 else
|
|
6127 {
|
|
6128 if (copy)
|
|
6129 {
|
|
6130 if (func == (fptr)NULL) /* just copy */
|
|
6131 *dst = *s;
|
|
6132 else /* change case */
|
|
6133 func = (fptr)(func(dst, *s));
|
|
6134 /* Turbo C complains without the typecast */
|
|
6135 }
|
|
6136 ++dst;
|
|
6137 }
|
|
6138 ++s;
|
|
6139 --len;
|
|
6140 }
|
|
6141 }
|
|
6142 }
|
|
6143 no = -1;
|
|
6144 }
|
|
6145 }
|
|
6146 if (copy)
|
|
6147 *dst = NUL;
|
|
6148
|
|
6149 exit:
|
|
6150 return (int)((dst - dest) + 1);
|
|
6151 }
|
|
6152
|
|
6153 #ifdef FEAT_EVAL
|
|
6154 /*
|
|
6155 * Used for the submatch() function: get the string from tne n'th submatch in
|
|
6156 * allocated memory.
|
|
6157 * Returns NULL when not in a ":s" command and for a non-existing submatch.
|
|
6158 */
|
|
6159 char_u *
|
|
6160 reg_submatch(no)
|
|
6161 int no;
|
|
6162 {
|
|
6163 char_u *retval = NULL;
|
|
6164 char_u *s;
|
|
6165 int len;
|
|
6166 int round;
|
|
6167 linenr_T lnum;
|
|
6168
|
|
6169 if (!can_f_submatch)
|
|
6170 return NULL;
|
|
6171
|
|
6172 if (submatch_match == NULL)
|
|
6173 {
|
|
6174 /*
|
|
6175 * First round: compute the length and allocate memory.
|
|
6176 * Second round: copy the text.
|
|
6177 */
|
|
6178 for (round = 1; round <= 2; ++round)
|
|
6179 {
|
|
6180 lnum = submatch_mmatch->startpos[no].lnum;
|
|
6181 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0)
|
|
6182 return NULL;
|
|
6183
|
|
6184 s = reg_getline(lnum) + submatch_mmatch->startpos[no].col;
|
|
6185 if (s == NULL) /* anti-crash check, cannot happen? */
|
|
6186 break;
|
|
6187 if (submatch_mmatch->endpos[no].lnum == lnum)
|
|
6188 {
|
|
6189 /* Within one line: take form start to end col. */
|
|
6190 len = submatch_mmatch->endpos[no].col
|
|
6191 - submatch_mmatch->startpos[no].col;
|
|
6192 if (round == 2)
|
|
6193 {
|
|
6194 STRNCPY(retval, s, len);
|
|
6195 retval[len] = NUL;
|
|
6196 }
|
|
6197 ++len;
|
|
6198 }
|
|
6199 else
|
|
6200 {
|
|
6201 /* Multiple lines: take start line from start col, middle
|
|
6202 * lines completely and end line up to end col. */
|
|
6203 len = (int)STRLEN(s);
|
|
6204 if (round == 2)
|
|
6205 {
|
|
6206 STRCPY(retval, s);
|
|
6207 retval[len] = '\n';
|
|
6208 }
|
|
6209 ++len;
|
|
6210 ++lnum;
|
|
6211 while (lnum < submatch_mmatch->endpos[no].lnum)
|
|
6212 {
|
|
6213 s = reg_getline(lnum++);
|
|
6214 if (round == 2)
|
|
6215 STRCPY(retval + len, s);
|
|
6216 len += (int)STRLEN(s);
|
|
6217 if (round == 2)
|
|
6218 retval[len] = '\n';
|
|
6219 ++len;
|
|
6220 }
|
|
6221 if (round == 2)
|
|
6222 STRNCPY(retval + len, reg_getline(lnum),
|
|
6223 submatch_mmatch->endpos[no].col);
|
|
6224 len += submatch_mmatch->endpos[no].col;
|
|
6225 if (round == 2)
|
|
6226 retval[len] = NUL;
|
|
6227 ++len;
|
|
6228 }
|
|
6229
|
|
6230 if (round == 1)
|
|
6231 {
|
|
6232 retval = lalloc((long_u)len, TRUE);
|
|
6233 if (s == NULL)
|
|
6234 return NULL;
|
|
6235 }
|
|
6236 }
|
|
6237 }
|
|
6238 else
|
|
6239 {
|
|
6240 if (submatch_match->endp[no] == NULL)
|
|
6241 retval = NULL;
|
|
6242 else
|
|
6243 {
|
|
6244 s = submatch_match->startp[no];
|
|
6245 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s));
|
|
6246 }
|
|
6247 }
|
|
6248
|
|
6249 return retval;
|
|
6250 }
|
|
6251 #endif
|