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
|
167
|
77 * node points to the node after the stuff to be repeated.
|
|
78 * The operand of some types of node is a literal string; for others, it is a
|
|
79 * node leading into a sub-FSM. In particular, the operand of a BRANCH node
|
|
80 * is the first node of the branch.
|
|
81 * (NB this is *not* a tree structure: the tail of the branch connects to the
|
|
82 * thing following the set of BRANCHes.)
|
7
|
83 *
|
|
84 * pattern is coded like:
|
|
85 *
|
|
86 * +-----------------+
|
|
87 * | V
|
|
88 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END
|
|
89 * | ^ | ^
|
|
90 * +------+ +----------+
|
|
91 *
|
|
92 *
|
|
93 * +------------------+
|
|
94 * V |
|
|
95 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END
|
|
96 * | | ^ ^
|
|
97 * | +---------------+ |
|
|
98 * +---------------------------------------------+
|
|
99 *
|
|
100 *
|
167
|
101 * +----------------------+
|
|
102 * V |
|
233
|
103 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END
|
179
|
104 * | | ^ ^
|
|
105 * | +-----------+ |
|
|
106 * +--------------------------------------------------+
|
167
|
107 *
|
|
108 *
|
7
|
109 * +-------------------------+
|
|
110 * V |
|
|
111 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END
|
|
112 * | | ^
|
|
113 * | +----------------+
|
|
114 * +-----------------------------------------------+
|
|
115 *
|
|
116 *
|
|
117 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END
|
|
118 * | | ^ ^
|
|
119 * | +----------------+ |
|
|
120 * +--------------------------------+
|
|
121 *
|
|
122 * +---------+
|
|
123 * | V
|
|
124 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END
|
|
125 * | | | | ^ ^
|
|
126 * | | | +-----+ |
|
|
127 * | | +----------------+ |
|
|
128 * | +---------------------------+ |
|
|
129 * +------------------------------------------------------+
|
|
130 *
|
|
131 * They all start with a BRANCH for "\|" alternaties, even when there is only
|
|
132 * one alternative.
|
|
133 */
|
|
134
|
|
135 /*
|
|
136 * The opcodes are:
|
|
137 */
|
|
138
|
|
139 /* definition number opnd? meaning */
|
|
140 #define END 0 /* End of program or NOMATCH operand. */
|
|
141 #define BOL 1 /* Match "" at beginning of line. */
|
|
142 #define EOL 2 /* Match "" at end of line. */
|
|
143 #define BRANCH 3 /* node Match this alternative, or the
|
|
144 * next... */
|
|
145 #define BACK 4 /* Match "", "next" ptr points backward. */
|
|
146 #define EXACTLY 5 /* str Match this string. */
|
|
147 #define NOTHING 6 /* Match empty string. */
|
|
148 #define STAR 7 /* node Match this (simple) thing 0 or more
|
|
149 * times. */
|
|
150 #define PLUS 8 /* node Match this (simple) thing 1 or more
|
|
151 * times. */
|
|
152 #define MATCH 9 /* node match the operand zero-width */
|
|
153 #define NOMATCH 10 /* node check for no match with operand */
|
|
154 #define BEHIND 11 /* node look behind for a match with operand */
|
|
155 #define NOBEHIND 12 /* node look behind for no match with operand */
|
|
156 #define SUBPAT 13 /* node match the operand here */
|
|
157 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and
|
|
158 * n times (\{m,n\}). */
|
|
159 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */
|
|
160 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */
|
|
161 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE
|
|
162 * and BRACE_COMPLEX. */
|
|
163 #define NEWL 18 /* Match line-break */
|
|
164 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */
|
|
165
|
|
166
|
|
167 /* character classes: 20-48 normal, 50-78 include a line-break */
|
|
168 #define ADD_NL 30
|
|
169 #define FIRST_NL ANY + ADD_NL
|
|
170 #define ANY 20 /* Match any one character. */
|
|
171 #define ANYOF 21 /* str Match any character in this string. */
|
|
172 #define ANYBUT 22 /* str Match any character not in this
|
|
173 * string. */
|
|
174 #define IDENT 23 /* Match identifier char */
|
|
175 #define SIDENT 24 /* Match identifier char but no digit */
|
|
176 #define KWORD 25 /* Match keyword char */
|
|
177 #define SKWORD 26 /* Match word char but no digit */
|
|
178 #define FNAME 27 /* Match file name char */
|
|
179 #define SFNAME 28 /* Match file name char but no digit */
|
|
180 #define PRINT 29 /* Match printable char */
|
|
181 #define SPRINT 30 /* Match printable char but no digit */
|
|
182 #define WHITE 31 /* Match whitespace char */
|
|
183 #define NWHITE 32 /* Match non-whitespace char */
|
|
184 #define DIGIT 33 /* Match digit char */
|
|
185 #define NDIGIT 34 /* Match non-digit char */
|
|
186 #define HEX 35 /* Match hex char */
|
|
187 #define NHEX 36 /* Match non-hex char */
|
|
188 #define OCTAL 37 /* Match octal char */
|
|
189 #define NOCTAL 38 /* Match non-octal char */
|
|
190 #define WORD 39 /* Match word char */
|
|
191 #define NWORD 40 /* Match non-word char */
|
|
192 #define HEAD 41 /* Match head char */
|
|
193 #define NHEAD 42 /* Match non-head char */
|
|
194 #define ALPHA 43 /* Match alpha char */
|
|
195 #define NALPHA 44 /* Match non-alpha char */
|
|
196 #define LOWER 45 /* Match lowercase char */
|
|
197 #define NLOWER 46 /* Match non-lowercase char */
|
|
198 #define UPPER 47 /* Match uppercase char */
|
|
199 #define NUPPER 48 /* Match non-uppercase char */
|
|
200 #define LAST_NL NUPPER + ADD_NL
|
|
201 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL)
|
|
202
|
|
203 #define MOPEN 80 /* -89 Mark this point in input as start of
|
|
204 * \( subexpr. MOPEN + 0 marks start of
|
|
205 * match. */
|
|
206 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks
|
|
207 * end of match. */
|
|
208 #define BACKREF 100 /* -109 node Match same string again \1-\9 */
|
|
209
|
|
210 #ifdef FEAT_SYN_HL
|
|
211 # define ZOPEN 110 /* -119 Mark this point in input as start of
|
|
212 * \z( subexpr. */
|
|
213 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */
|
|
214 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */
|
|
215 #endif
|
|
216
|
|
217 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */
|
|
218
|
|
219 #define NOPEN 150 /* Mark this point in input as start of
|
|
220 \%( subexpr. */
|
|
221 #define NCLOSE 151 /* Analogous to NOPEN. */
|
|
222
|
|
223 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */
|
|
224 #define RE_BOF 201 /* Match "" at beginning of file. */
|
|
225 #define RE_EOF 202 /* Match "" at end of file. */
|
|
226 #define CURSOR 203 /* Match location of cursor. */
|
|
227
|
|
228 #define RE_LNUM 204 /* nr cmp Match line number */
|
|
229 #define RE_COL 205 /* nr cmp Match column number */
|
|
230 #define RE_VCOL 206 /* nr cmp Match virtual column number */
|
|
231
|
|
232 /*
|
|
233 * Magic characters have a special meaning, they don't match literally.
|
|
234 * Magic characters are negative. This separates them from literal characters
|
|
235 * (possibly multi-byte). Only ASCII characters can be Magic.
|
|
236 */
|
|
237 #define Magic(x) ((int)(x) - 256)
|
|
238 #define un_Magic(x) ((x) + 256)
|
|
239 #define is_Magic(x) ((x) < 0)
|
|
240
|
|
241 static int no_Magic __ARGS((int x));
|
|
242 static int toggle_Magic __ARGS((int x));
|
|
243
|
|
244 static int
|
|
245 no_Magic(x)
|
|
246 int x;
|
|
247 {
|
|
248 if (is_Magic(x))
|
|
249 return un_Magic(x);
|
|
250 return x;
|
|
251 }
|
|
252
|
|
253 static int
|
|
254 toggle_Magic(x)
|
|
255 int x;
|
|
256 {
|
|
257 if (is_Magic(x))
|
|
258 return un_Magic(x);
|
|
259 return Magic(x);
|
|
260 }
|
|
261
|
|
262 /*
|
|
263 * The first byte of the regexp internal "program" is actually this magic
|
|
264 * number; the start node begins in the second byte. It's used to catch the
|
|
265 * most severe mutilation of the program by the caller.
|
|
266 */
|
|
267
|
|
268 #define REGMAGIC 0234
|
|
269
|
|
270 /*
|
|
271 * Opcode notes:
|
|
272 *
|
|
273 * BRANCH The set of branches constituting a single choice are hooked
|
|
274 * together with their "next" pointers, since precedence prevents
|
|
275 * anything being concatenated to any individual branch. The
|
|
276 * "next" pointer of the last BRANCH in a choice points to the
|
|
277 * thing following the whole choice. This is also where the
|
|
278 * final "next" pointer of each individual branch points; each
|
|
279 * branch starts with the operand node of a BRANCH node.
|
|
280 *
|
|
281 * BACK Normal "next" pointers all implicitly point forward; BACK
|
|
282 * exists to make loop structures possible.
|
|
283 *
|
|
284 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular
|
|
285 * BRANCH structures using BACK. Simple cases (one character
|
|
286 * per match) are implemented with STAR and PLUS for speed
|
|
287 * and to minimize recursive plunges.
|
|
288 *
|
|
289 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX
|
|
290 * node, and defines the min and max limits to be used for that
|
|
291 * node.
|
|
292 *
|
|
293 * MOPEN,MCLOSE ...are numbered at compile time.
|
|
294 * ZOPEN,ZCLOSE ...ditto
|
|
295 */
|
|
296
|
|
297 /*
|
|
298 * A node is one char of opcode followed by two chars of "next" pointer.
|
|
299 * "Next" pointers are stored as two 8-bit bytes, high order first. The
|
|
300 * value is a positive offset from the opcode of the node containing it.
|
|
301 * An operand, if any, simply follows the node. (Note that much of the
|
|
302 * code generation knows about this implicit relationship.)
|
|
303 *
|
|
304 * Using two bytes for the "next" pointer is vast overkill for most things,
|
|
305 * but allows patterns to get big without disasters.
|
|
306 */
|
|
307 #define OP(p) ((int)*(p))
|
|
308 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377))
|
|
309 #define OPERAND(p) ((p) + 3)
|
|
310 /* Obtain an operand that was stored as four bytes, MSB first. */
|
|
311 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \
|
|
312 + ((long)(p)[5] << 8) + (long)(p)[6])
|
|
313 /* Obtain a second operand stored as four bytes. */
|
|
314 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4)
|
|
315 /* Obtain a second single-byte operand stored after a four bytes operand. */
|
|
316 #define OPERAND_CMP(p) (p)[7]
|
|
317
|
|
318 /*
|
|
319 * Utility definitions.
|
|
320 */
|
|
321 #define UCHARAT(p) ((int)*(char_u *)(p))
|
|
322
|
|
323 /* Used for an error (down from) vim_regcomp(): give the error message, set
|
|
324 * rc_did_emsg and return NULL */
|
308
|
325 #define EMSG_RET_NULL(m) return (EMSG(m), rc_did_emsg = TRUE, NULL)
|
|
326 #define EMSG_M_RET_NULL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, NULL)
|
|
327 #define EMSG_RET_FAIL(m) return (EMSG(m), rc_did_emsg = TRUE, FAIL)
|
7
|
328 #define EMSG_ONE_RET_NULL EMSG_M_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL)
|
|
329
|
|
330 #define MAX_LIMIT (32767L << 16L)
|
|
331
|
|
332 static int re_multi_type __ARGS((int));
|
|
333 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n));
|
|
334 static char_u *cstrchr __ARGS((char_u *, int));
|
|
335
|
|
336 #ifdef DEBUG
|
|
337 static void regdump __ARGS((char_u *, regprog_T *));
|
|
338 static char_u *regprop __ARGS((char_u *));
|
|
339 #endif
|
|
340
|
|
341 #define NOT_MULTI 0
|
|
342 #define MULTI_ONE 1
|
|
343 #define MULTI_MULT 2
|
|
344 /*
|
|
345 * Return NOT_MULTI if c is not a "multi" operator.
|
|
346 * Return MULTI_ONE if c is a single "multi" operator.
|
|
347 * Return MULTI_MULT if c is a multi "multi" operator.
|
|
348 */
|
|
349 static int
|
|
350 re_multi_type(c)
|
|
351 int c;
|
|
352 {
|
|
353 if (c == Magic('@') || c == Magic('=') || c == Magic('?'))
|
|
354 return MULTI_ONE;
|
|
355 if (c == Magic('*') || c == Magic('+') || c == Magic('{'))
|
|
356 return MULTI_MULT;
|
|
357 return NOT_MULTI;
|
|
358 }
|
|
359
|
|
360 /*
|
|
361 * Flags to be passed up and down.
|
|
362 */
|
|
363 #define HASWIDTH 0x1 /* Known never to match null string. */
|
|
364 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
|
|
365 #define SPSTART 0x4 /* Starts with * or +. */
|
|
366 #define HASNL 0x8 /* Contains some \n. */
|
|
367 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */
|
|
368 #define WORST 0 /* Worst case. */
|
|
369
|
|
370 /*
|
|
371 * When regcode is set to this value, code is not emitted and size is computed
|
|
372 * instead.
|
|
373 */
|
|
374 #define JUST_CALC_SIZE ((char_u *) -1)
|
|
375
|
359
|
376 static char_u *reg_prev_sub = NULL;
|
|
377
|
|
378 #if defined(EXITFREE) || defined(PROTO)
|
|
379 void
|
|
380 free_regexp_stuff()
|
|
381 {
|
|
382 vim_free(reg_prev_sub);
|
|
383 }
|
|
384 #endif
|
7
|
385
|
|
386 /*
|
|
387 * REGEXP_INRANGE contains all characters which are always special in a []
|
|
388 * range after '\'.
|
|
389 * REGEXP_ABBR contains all characters which act as abbreviations after '\'.
|
|
390 * These are:
|
|
391 * \n - New line (NL).
|
|
392 * \r - Carriage Return (CR).
|
|
393 * \t - Tab (TAB).
|
|
394 * \e - Escape (ESC).
|
|
395 * \b - Backspace (Ctrl_H).
|
24
|
396 * \d - Character code in decimal, eg \d123
|
|
397 * \o - Character code in octal, eg \o80
|
|
398 * \x - Character code in hex, eg \x4a
|
|
399 * \u - Multibyte character code, eg \u20ac
|
|
400 * \U - Long multibyte character code, eg \U12345678
|
7
|
401 */
|
|
402 static char_u REGEXP_INRANGE[] = "]^-n\\";
|
24
|
403 static char_u REGEXP_ABBR[] = "nrtebdoxuU";
|
7
|
404
|
|
405 static int backslash_trans __ARGS((int c));
|
167
|
406 static int get_char_class __ARGS((char_u **pp));
|
|
407 static int get_equi_class __ARGS((char_u **pp));
|
|
408 static void reg_equi_class __ARGS((int c));
|
|
409 static int get_coll_element __ARGS((char_u **pp));
|
7
|
410 static char_u *skip_anyof __ARGS((char_u *p));
|
|
411 static void init_class_tab __ARGS((void));
|
|
412
|
|
413 /*
|
|
414 * Translate '\x' to its control character, except "\n", which is Magic.
|
|
415 */
|
|
416 static int
|
|
417 backslash_trans(c)
|
|
418 int c;
|
|
419 {
|
|
420 switch (c)
|
|
421 {
|
|
422 case 'r': return CAR;
|
|
423 case 't': return TAB;
|
|
424 case 'e': return ESC;
|
|
425 case 'b': return BS;
|
|
426 }
|
|
427 return c;
|
|
428 }
|
|
429
|
|
430 /*
|
167
|
431 * Check for a character class name "[:name:]". "pp" points to the '['.
|
7
|
432 * Returns one of the CLASS_ items. CLASS_NONE means that no item was
|
|
433 * recognized. Otherwise "pp" is advanced to after the item.
|
|
434 */
|
|
435 static int
|
167
|
436 get_char_class(pp)
|
7
|
437 char_u **pp;
|
|
438 {
|
|
439 static const char *(class_names[]) =
|
|
440 {
|
|
441 "alnum:]",
|
|
442 #define CLASS_ALNUM 0
|
|
443 "alpha:]",
|
|
444 #define CLASS_ALPHA 1
|
|
445 "blank:]",
|
|
446 #define CLASS_BLANK 2
|
|
447 "cntrl:]",
|
|
448 #define CLASS_CNTRL 3
|
|
449 "digit:]",
|
|
450 #define CLASS_DIGIT 4
|
|
451 "graph:]",
|
|
452 #define CLASS_GRAPH 5
|
|
453 "lower:]",
|
|
454 #define CLASS_LOWER 6
|
|
455 "print:]",
|
|
456 #define CLASS_PRINT 7
|
|
457 "punct:]",
|
|
458 #define CLASS_PUNCT 8
|
|
459 "space:]",
|
|
460 #define CLASS_SPACE 9
|
|
461 "upper:]",
|
|
462 #define CLASS_UPPER 10
|
|
463 "xdigit:]",
|
|
464 #define CLASS_XDIGIT 11
|
|
465 "tab:]",
|
|
466 #define CLASS_TAB 12
|
|
467 "return:]",
|
|
468 #define CLASS_RETURN 13
|
|
469 "backspace:]",
|
|
470 #define CLASS_BACKSPACE 14
|
|
471 "escape:]",
|
|
472 #define CLASS_ESCAPE 15
|
|
473 };
|
|
474 #define CLASS_NONE 99
|
|
475 int i;
|
|
476
|
|
477 if ((*pp)[1] == ':')
|
|
478 {
|
|
479 for (i = 0; i < sizeof(class_names) / sizeof(*class_names); ++i)
|
|
480 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0)
|
|
481 {
|
|
482 *pp += STRLEN(class_names[i]) + 2;
|
|
483 return i;
|
|
484 }
|
|
485 }
|
|
486 return CLASS_NONE;
|
|
487 }
|
|
488
|
|
489 /*
|
|
490 * Specific version of character class functions.
|
|
491 * Using a table to keep this fast.
|
|
492 */
|
|
493 static short class_tab[256];
|
|
494
|
|
495 #define RI_DIGIT 0x01
|
|
496 #define RI_HEX 0x02
|
|
497 #define RI_OCTAL 0x04
|
|
498 #define RI_WORD 0x08
|
|
499 #define RI_HEAD 0x10
|
|
500 #define RI_ALPHA 0x20
|
|
501 #define RI_LOWER 0x40
|
|
502 #define RI_UPPER 0x80
|
|
503 #define RI_WHITE 0x100
|
|
504
|
|
505 static void
|
|
506 init_class_tab()
|
|
507 {
|
|
508 int i;
|
|
509 static int done = FALSE;
|
|
510
|
|
511 if (done)
|
|
512 return;
|
|
513
|
|
514 for (i = 0; i < 256; ++i)
|
|
515 {
|
|
516 if (i >= '0' && i <= '7')
|
|
517 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD;
|
|
518 else if (i >= '8' && i <= '9')
|
|
519 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD;
|
|
520 else if (i >= 'a' && i <= 'f')
|
|
521 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
522 #ifdef EBCDIC
|
|
523 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r')
|
|
524 || (i >= 's' && i <= 'z'))
|
|
525 #else
|
|
526 else if (i >= 'g' && i <= 'z')
|
|
527 #endif
|
|
528 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
529 else if (i >= 'A' && i <= 'F')
|
|
530 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
531 #ifdef EBCDIC
|
|
532 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R')
|
|
533 || (i >= 'S' && i <= 'Z'))
|
|
534 #else
|
|
535 else if (i >= 'G' && i <= 'Z')
|
|
536 #endif
|
|
537 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
538 else if (i == '_')
|
|
539 class_tab[i] = RI_WORD + RI_HEAD;
|
|
540 else
|
|
541 class_tab[i] = 0;
|
|
542 }
|
|
543 class_tab[' '] |= RI_WHITE;
|
|
544 class_tab['\t'] |= RI_WHITE;
|
|
545 done = TRUE;
|
|
546 }
|
|
547
|
|
548 #ifdef FEAT_MBYTE
|
|
549 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT))
|
|
550 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX))
|
|
551 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL))
|
|
552 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD))
|
|
553 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD))
|
|
554 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA))
|
|
555 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER))
|
|
556 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER))
|
|
557 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE))
|
|
558 #else
|
|
559 # define ri_digit(c) (class_tab[c] & RI_DIGIT)
|
|
560 # define ri_hex(c) (class_tab[c] & RI_HEX)
|
|
561 # define ri_octal(c) (class_tab[c] & RI_OCTAL)
|
|
562 # define ri_word(c) (class_tab[c] & RI_WORD)
|
|
563 # define ri_head(c) (class_tab[c] & RI_HEAD)
|
|
564 # define ri_alpha(c) (class_tab[c] & RI_ALPHA)
|
|
565 # define ri_lower(c) (class_tab[c] & RI_LOWER)
|
|
566 # define ri_upper(c) (class_tab[c] & RI_UPPER)
|
|
567 # define ri_white(c) (class_tab[c] & RI_WHITE)
|
|
568 #endif
|
|
569
|
|
570 /* flags for regflags */
|
|
571 #define RF_ICASE 1 /* ignore case */
|
|
572 #define RF_NOICASE 2 /* don't ignore case */
|
|
573 #define RF_HASNL 4 /* can match a NL */
|
|
574 #define RF_ICOMBINE 8 /* ignore combining characters */
|
|
575 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */
|
|
576
|
|
577 /*
|
|
578 * Global work variables for vim_regcomp().
|
|
579 */
|
|
580
|
|
581 static char_u *regparse; /* Input-scan pointer. */
|
|
582 static int prevchr_len; /* byte length of previous char */
|
|
583 static int num_complex_braces; /* Complex \{...} count */
|
|
584 static int regnpar; /* () count. */
|
|
585 #ifdef FEAT_SYN_HL
|
|
586 static int regnzpar; /* \z() count. */
|
|
587 static int re_has_z; /* \z item detected */
|
|
588 #endif
|
|
589 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */
|
|
590 static long regsize; /* Code size. */
|
|
591 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */
|
|
592 static unsigned regflags; /* RF_ flags for prog */
|
|
593 static long brace_min[10]; /* Minimums for complex brace repeats */
|
|
594 static long brace_max[10]; /* Maximums for complex brace repeats */
|
|
595 static int brace_count[10]; /* Current counts for complex brace repeats */
|
|
596 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
597 static int had_eol; /* TRUE when EOL found by vim_regcomp() */
|
|
598 #endif
|
|
599 static int one_exactly = FALSE; /* only do one char for EXACTLY */
|
|
600
|
|
601 static int reg_magic; /* magicness of the pattern: */
|
|
602 #define MAGIC_NONE 1 /* "\V" very unmagic */
|
|
603 #define MAGIC_OFF 2 /* "\M" or 'magic' off */
|
|
604 #define MAGIC_ON 3 /* "\m" or 'magic' */
|
|
605 #define MAGIC_ALL 4 /* "\v" very magic */
|
|
606
|
|
607 static int reg_string; /* matching with a string instead of a buffer
|
|
608 line */
|
481
|
609 static int reg_strict; /* "[abc" is illegal */
|
7
|
610
|
|
611 /*
|
|
612 * META contains all characters that may be magic, except '^' and '$'.
|
|
613 */
|
|
614
|
|
615 #ifdef EBCDIC
|
|
616 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~";
|
|
617 #else
|
|
618 /* META[] is used often enough to justify turning it into a table. */
|
|
619 static char_u META_flags[] = {
|
|
620 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
621 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
622 /* % & ( ) * + . */
|
|
623 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0,
|
|
624 /* 1 2 3 4 5 6 7 8 9 < = > ? */
|
|
625 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1,
|
|
626 /* @ A C D F H I K L M O */
|
|
627 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
|
|
628 /* P S U V W X Z [ _ */
|
|
629 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1,
|
|
630 /* a c d f h i k l m n o */
|
|
631 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
|
|
632 /* p s u v w x z { | ~ */
|
|
633 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1
|
|
634 };
|
|
635 #endif
|
|
636
|
|
637 static int curchr;
|
|
638
|
|
639 /* arguments for reg() */
|
|
640 #define REG_NOPAREN 0 /* toplevel reg() */
|
|
641 #define REG_PAREN 1 /* \(\) */
|
|
642 #define REG_ZPAREN 2 /* \z(\) */
|
|
643 #define REG_NPAREN 3 /* \%(\) */
|
|
644
|
|
645 /*
|
|
646 * Forward declarations for vim_regcomp()'s friends.
|
|
647 */
|
|
648 static void initchr __ARGS((char_u *));
|
|
649 static int getchr __ARGS((void));
|
|
650 static void skipchr_keepstart __ARGS((void));
|
|
651 static int peekchr __ARGS((void));
|
|
652 static void skipchr __ARGS((void));
|
|
653 static void ungetchr __ARGS((void));
|
24
|
654 static int gethexchrs __ARGS((int maxinputlen));
|
|
655 static int getoctchrs __ARGS((void));
|
|
656 static int getdecchrs __ARGS((void));
|
|
657 static int coll_get_char __ARGS((void));
|
7
|
658 static void regcomp_start __ARGS((char_u *expr, int flags));
|
|
659 static char_u *reg __ARGS((int, int *));
|
|
660 static char_u *regbranch __ARGS((int *flagp));
|
|
661 static char_u *regconcat __ARGS((int *flagp));
|
|
662 static char_u *regpiece __ARGS((int *));
|
|
663 static char_u *regatom __ARGS((int *));
|
|
664 static char_u *regnode __ARGS((int));
|
|
665 static int prog_magic_wrong __ARGS((void));
|
|
666 static char_u *regnext __ARGS((char_u *));
|
|
667 static void regc __ARGS((int b));
|
|
668 #ifdef FEAT_MBYTE
|
|
669 static void regmbc __ARGS((int c));
|
167
|
670 #else
|
|
671 # define regmbc(c) regc(c)
|
7
|
672 #endif
|
|
673 static void reginsert __ARGS((int, char_u *));
|
|
674 static void reginsert_limits __ARGS((int, long, long, char_u *));
|
|
675 static char_u *re_put_long __ARGS((char_u *pr, long_u val));
|
|
676 static int read_limits __ARGS((long *, long *));
|
|
677 static void regtail __ARGS((char_u *, char_u *));
|
|
678 static void regoptail __ARGS((char_u *, char_u *));
|
|
679
|
|
680 /*
|
|
681 * Return TRUE if compiled regular expression "prog" can match a line break.
|
|
682 */
|
|
683 int
|
|
684 re_multiline(prog)
|
|
685 regprog_T *prog;
|
|
686 {
|
|
687 return (prog->regflags & RF_HASNL);
|
|
688 }
|
|
689
|
|
690 /*
|
|
691 * Return TRUE if compiled regular expression "prog" looks before the start
|
|
692 * position (pattern contains "\@<=" or "\@<!").
|
|
693 */
|
|
694 int
|
|
695 re_lookbehind(prog)
|
|
696 regprog_T *prog;
|
|
697 {
|
|
698 return (prog->regflags & RF_LOOKBH);
|
|
699 }
|
|
700
|
|
701 /*
|
167
|
702 * Check for an equivalence class name "[=a=]". "pp" points to the '['.
|
|
703 * Returns a character representing the class. Zero means that no item was
|
|
704 * recognized. Otherwise "pp" is advanced to after the item.
|
|
705 */
|
|
706 static int
|
|
707 get_equi_class(pp)
|
|
708 char_u **pp;
|
|
709 {
|
|
710 int c;
|
|
711 int l = 1;
|
|
712 char_u *p = *pp;
|
|
713
|
|
714 if (p[1] == '=')
|
|
715 {
|
|
716 #ifdef FEAT_MBYTE
|
|
717 if (has_mbyte)
|
474
|
718 l = (*mb_ptr2len)(p + 2);
|
167
|
719 #endif
|
|
720 if (p[l + 2] == '=' && p[l + 3] == ']')
|
|
721 {
|
|
722 #ifdef FEAT_MBYTE
|
|
723 if (has_mbyte)
|
|
724 c = mb_ptr2char(p + 2);
|
|
725 else
|
|
726 #endif
|
|
727 c = p[2];
|
|
728 *pp += l + 4;
|
|
729 return c;
|
|
730 }
|
|
731 }
|
|
732 return 0;
|
|
733 }
|
|
734
|
|
735 /*
|
|
736 * Produce the bytes for equivalence class "c".
|
|
737 * Currently only handles latin1, latin9 and utf-8.
|
|
738 */
|
|
739 static void
|
|
740 reg_equi_class(c)
|
|
741 int c;
|
|
742 {
|
|
743 #ifdef FEAT_MBYTE
|
|
744 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0
|
492
|
745 || STRCMP(p_enc, "iso-8859-15") == 0)
|
167
|
746 #endif
|
|
747 {
|
|
748 switch (c)
|
|
749 {
|
236
|
750 case 'A': case '\300': case '\301': case '\302':
|
|
751 case '\303': case '\304': case '\305':
|
|
752 regmbc('A'); regmbc('\300'); regmbc('\301');
|
|
753 regmbc('\302'); regmbc('\303'); regmbc('\304');
|
|
754 regmbc('\305');
|
167
|
755 return;
|
236
|
756 case 'C': case '\307':
|
|
757 regmbc('C'); regmbc('\307');
|
167
|
758 return;
|
236
|
759 case 'E': case '\310': case '\311': case '\312': case '\313':
|
|
760 regmbc('E'); regmbc('\310'); regmbc('\311');
|
|
761 regmbc('\312'); regmbc('\313');
|
167
|
762 return;
|
236
|
763 case 'I': case '\314': case '\315': case '\316': case '\317':
|
|
764 regmbc('I'); regmbc('\314'); regmbc('\315');
|
|
765 regmbc('\316'); regmbc('\317');
|
167
|
766 return;
|
236
|
767 case 'N': case '\321':
|
|
768 regmbc('N'); regmbc('\321');
|
167
|
769 return;
|
236
|
770 case 'O': case '\322': case '\323': case '\324': case '\325':
|
|
771 case '\326':
|
|
772 regmbc('O'); regmbc('\322'); regmbc('\323');
|
|
773 regmbc('\324'); regmbc('\325'); regmbc('\326');
|
167
|
774 return;
|
236
|
775 case 'U': case '\331': case '\332': case '\333': case '\334':
|
|
776 regmbc('U'); regmbc('\331'); regmbc('\332');
|
|
777 regmbc('\333'); regmbc('\334');
|
167
|
778 return;
|
236
|
779 case 'Y': case '\335':
|
|
780 regmbc('Y'); regmbc('\335');
|
167
|
781 return;
|
236
|
782 case 'a': case '\340': case '\341': case '\342':
|
|
783 case '\343': case '\344': case '\345':
|
|
784 regmbc('a'); regmbc('\340'); regmbc('\341');
|
|
785 regmbc('\342'); regmbc('\343'); regmbc('\344');
|
|
786 regmbc('\345');
|
167
|
787 return;
|
236
|
788 case 'c': case '\347':
|
|
789 regmbc('c'); regmbc('\347');
|
167
|
790 return;
|
236
|
791 case 'e': case '\350': case '\351': case '\352': case '\353':
|
|
792 regmbc('e'); regmbc('\350'); regmbc('\351');
|
|
793 regmbc('\352'); regmbc('\353');
|
167
|
794 return;
|
236
|
795 case 'i': case '\354': case '\355': case '\356': case '\357':
|
|
796 regmbc('i'); regmbc('\354'); regmbc('\355');
|
|
797 regmbc('\356'); regmbc('\357');
|
167
|
798 return;
|
236
|
799 case 'n': case '\361':
|
|
800 regmbc('n'); regmbc('\361');
|
167
|
801 return;
|
236
|
802 case 'o': case '\362': case '\363': case '\364': case '\365':
|
|
803 case '\366':
|
|
804 regmbc('o'); regmbc('\362'); regmbc('\363');
|
|
805 regmbc('\364'); regmbc('\365'); regmbc('\366');
|
167
|
806 return;
|
236
|
807 case 'u': case '\371': case '\372': case '\373': case '\374':
|
|
808 regmbc('u'); regmbc('\371'); regmbc('\372');
|
|
809 regmbc('\373'); regmbc('\374');
|
167
|
810 return;
|
236
|
811 case 'y': case '\375': case '\377':
|
|
812 regmbc('y'); regmbc('\375'); regmbc('\377');
|
167
|
813 return;
|
|
814 }
|
|
815 }
|
|
816 regmbc(c);
|
|
817 }
|
|
818
|
|
819 /*
|
|
820 * Check for a collating element "[.a.]". "pp" points to the '['.
|
|
821 * Returns a character. Zero means that no item was recognized. Otherwise
|
|
822 * "pp" is advanced to after the item.
|
|
823 * Currently only single characters are recognized!
|
|
824 */
|
|
825 static int
|
|
826 get_coll_element(pp)
|
|
827 char_u **pp;
|
|
828 {
|
|
829 int c;
|
|
830 int l = 1;
|
|
831 char_u *p = *pp;
|
|
832
|
|
833 if (p[1] == '.')
|
|
834 {
|
|
835 #ifdef FEAT_MBYTE
|
|
836 if (has_mbyte)
|
474
|
837 l = (*mb_ptr2len)(p + 2);
|
167
|
838 #endif
|
|
839 if (p[l + 2] == '.' && p[l + 3] == ']')
|
|
840 {
|
|
841 #ifdef FEAT_MBYTE
|
|
842 if (has_mbyte)
|
|
843 c = mb_ptr2char(p + 2);
|
|
844 else
|
|
845 #endif
|
|
846 c = p[2];
|
|
847 *pp += l + 4;
|
|
848 return c;
|
|
849 }
|
|
850 }
|
|
851 return 0;
|
|
852 }
|
|
853
|
|
854
|
|
855 /*
|
|
856 * Skip over a "[]" range.
|
|
857 * "p" must point to the character after the '['.
|
|
858 * The returned pointer is on the matching ']', or the terminating NUL.
|
|
859 */
|
|
860 static char_u *
|
|
861 skip_anyof(p)
|
|
862 char_u *p;
|
|
863 {
|
|
864 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
|
865 int cpo_bsl; /* 'cpoptions' contains '\' flag */
|
|
866 #ifdef FEAT_MBYTE
|
|
867 int l;
|
|
868 #endif
|
|
869
|
410
|
870 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL;
|
|
871 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL;
|
167
|
872
|
|
873 if (*p == '^') /* Complement of range. */
|
|
874 ++p;
|
|
875 if (*p == ']' || *p == '-')
|
|
876 ++p;
|
|
877 while (*p != NUL && *p != ']')
|
|
878 {
|
|
879 #ifdef FEAT_MBYTE
|
474
|
880 if (has_mbyte && (l = (*mb_ptr2len)(p)) > 1)
|
167
|
881 p += l;
|
|
882 else
|
|
883 #endif
|
|
884 if (*p == '-')
|
|
885 {
|
|
886 ++p;
|
|
887 if (*p != ']' && *p != NUL)
|
|
888 mb_ptr_adv(p);
|
|
889 }
|
|
890 else if (*p == '\\'
|
|
891 && !cpo_bsl
|
|
892 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL
|
|
893 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL)))
|
|
894 p += 2;
|
|
895 else if (*p == '[')
|
|
896 {
|
|
897 if (get_char_class(&p) == CLASS_NONE
|
|
898 && get_equi_class(&p) == 0
|
|
899 && get_coll_element(&p) == 0)
|
|
900 ++p; /* It was not a class name */
|
|
901 }
|
|
902 else
|
|
903 ++p;
|
|
904 }
|
|
905
|
|
906 return p;
|
|
907 }
|
|
908
|
|
909 /*
|
7
|
910 * Skip past regular expression.
|
153
|
911 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc).
|
7
|
912 * Take care of characters with a backslash in front of it.
|
|
913 * Skip strings inside [ and ].
|
|
914 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the
|
|
915 * expression and change "\?" to "?". If "*newp" is not NULL the expression
|
|
916 * is changed in-place.
|
|
917 */
|
|
918 char_u *
|
|
919 skip_regexp(startp, dirc, magic, newp)
|
|
920 char_u *startp;
|
|
921 int dirc;
|
|
922 int magic;
|
|
923 char_u **newp;
|
|
924 {
|
|
925 int mymagic;
|
|
926 char_u *p = startp;
|
|
927
|
|
928 if (magic)
|
|
929 mymagic = MAGIC_ON;
|
|
930 else
|
|
931 mymagic = MAGIC_OFF;
|
|
932
|
39
|
933 for (; p[0] != NUL; mb_ptr_adv(p))
|
7
|
934 {
|
|
935 if (p[0] == dirc) /* found end of regexp */
|
|
936 break;
|
|
937 if ((p[0] == '[' && mymagic >= MAGIC_ON)
|
|
938 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF))
|
|
939 {
|
|
940 p = skip_anyof(p + 1);
|
|
941 if (p[0] == NUL)
|
|
942 break;
|
|
943 }
|
|
944 else if (p[0] == '\\' && p[1] != NUL)
|
|
945 {
|
|
946 if (dirc == '?' && newp != NULL && p[1] == '?')
|
|
947 {
|
|
948 /* change "\?" to "?", make a copy first. */
|
|
949 if (*newp == NULL)
|
|
950 {
|
|
951 *newp = vim_strsave(startp);
|
|
952 if (*newp != NULL)
|
|
953 p = *newp + (p - startp);
|
|
954 }
|
|
955 if (*newp != NULL)
|
|
956 mch_memmove(p, p + 1, STRLEN(p));
|
|
957 else
|
|
958 ++p;
|
|
959 }
|
|
960 else
|
|
961 ++p; /* skip next character */
|
|
962 if (*p == 'v')
|
|
963 mymagic = MAGIC_ALL;
|
|
964 else if (*p == 'V')
|
|
965 mymagic = MAGIC_NONE;
|
|
966 }
|
|
967 }
|
|
968 return p;
|
|
969 }
|
|
970
|
|
971 /*
|
41
|
972 * vim_regcomp() - compile a regular expression into internal code
|
|
973 * Returns the program in allocated space. Returns NULL for an error.
|
7
|
974 *
|
|
975 * We can't allocate space until we know how big the compiled form will be,
|
|
976 * but we can't compile it (and thus know how big it is) until we've got a
|
|
977 * place to put the code. So we cheat: we compile it twice, once with code
|
|
978 * generation turned off and size counting turned on, and once "for real".
|
|
979 * This also means that we don't allocate space until we are sure that the
|
|
980 * thing really will compile successfully, and we never have to move the
|
|
981 * code and thus invalidate pointers into it. (Note that it has to be in
|
|
982 * one piece because vim_free() must be able to free it all.)
|
|
983 *
|
|
984 * Whether upper/lower case is to be ignored is decided when executing the
|
|
985 * program, it does not matter here.
|
|
986 *
|
|
987 * Beware that the optimization-preparation code in here knows about some
|
|
988 * of the structure of the compiled regexp.
|
|
989 * "re_flags": RE_MAGIC and/or RE_STRING.
|
|
990 */
|
|
991 regprog_T *
|
|
992 vim_regcomp(expr, re_flags)
|
|
993 char_u *expr;
|
|
994 int re_flags;
|
|
995 {
|
|
996 regprog_T *r;
|
|
997 char_u *scan;
|
|
998 char_u *longest;
|
|
999 int len;
|
|
1000 int flags;
|
|
1001
|
|
1002 if (expr == NULL)
|
|
1003 EMSG_RET_NULL(_(e_null));
|
|
1004
|
|
1005 init_class_tab();
|
|
1006
|
|
1007 /*
|
|
1008 * First pass: determine size, legality.
|
|
1009 */
|
|
1010 regcomp_start(expr, re_flags);
|
|
1011 regcode = JUST_CALC_SIZE;
|
|
1012 regc(REGMAGIC);
|
|
1013 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
1014 return NULL;
|
|
1015
|
|
1016 /* Small enough for pointer-storage convention? */
|
|
1017 #ifdef SMALL_MALLOC /* 16 bit storage allocation */
|
|
1018 if (regsize >= 65536L - 256L)
|
|
1019 EMSG_RET_NULL(_("E339: Pattern too long"));
|
|
1020 #endif
|
|
1021
|
|
1022 /* Allocate space. */
|
|
1023 r = (regprog_T *)lalloc(sizeof(regprog_T) + regsize, TRUE);
|
|
1024 if (r == NULL)
|
|
1025 return NULL;
|
|
1026
|
|
1027 /*
|
|
1028 * Second pass: emit code.
|
|
1029 */
|
|
1030 regcomp_start(expr, re_flags);
|
|
1031 regcode = r->program;
|
|
1032 regc(REGMAGIC);
|
|
1033 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
1034 {
|
|
1035 vim_free(r);
|
|
1036 return NULL;
|
|
1037 }
|
|
1038
|
|
1039 /* Dig out information for optimizations. */
|
|
1040 r->regstart = NUL; /* Worst-case defaults. */
|
|
1041 r->reganch = 0;
|
|
1042 r->regmust = NULL;
|
|
1043 r->regmlen = 0;
|
|
1044 r->regflags = regflags;
|
|
1045 if (flags & HASNL)
|
|
1046 r->regflags |= RF_HASNL;
|
|
1047 if (flags & HASLOOKBH)
|
|
1048 r->regflags |= RF_LOOKBH;
|
|
1049 #ifdef FEAT_SYN_HL
|
|
1050 /* Remember whether this pattern has any \z specials in it. */
|
|
1051 r->reghasz = re_has_z;
|
|
1052 #endif
|
|
1053 scan = r->program + 1; /* First BRANCH. */
|
|
1054 if (OP(regnext(scan)) == END) /* Only one top-level choice. */
|
|
1055 {
|
|
1056 scan = OPERAND(scan);
|
|
1057
|
|
1058 /* Starting-point info. */
|
|
1059 if (OP(scan) == BOL || OP(scan) == RE_BOF)
|
|
1060 {
|
|
1061 r->reganch++;
|
|
1062 scan = regnext(scan);
|
|
1063 }
|
|
1064
|
|
1065 if (OP(scan) == EXACTLY)
|
|
1066 {
|
|
1067 #ifdef FEAT_MBYTE
|
|
1068 if (has_mbyte)
|
|
1069 r->regstart = (*mb_ptr2char)(OPERAND(scan));
|
|
1070 else
|
|
1071 #endif
|
|
1072 r->regstart = *OPERAND(scan);
|
|
1073 }
|
|
1074 else if ((OP(scan) == BOW
|
|
1075 || OP(scan) == EOW
|
|
1076 || OP(scan) == NOTHING
|
|
1077 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
|
|
1078 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
|
|
1079 && OP(regnext(scan)) == EXACTLY)
|
|
1080 {
|
|
1081 #ifdef FEAT_MBYTE
|
|
1082 if (has_mbyte)
|
|
1083 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
|
|
1084 else
|
|
1085 #endif
|
|
1086 r->regstart = *OPERAND(regnext(scan));
|
|
1087 }
|
|
1088
|
|
1089 /*
|
|
1090 * If there's something expensive in the r.e., find the longest
|
|
1091 * literal string that must appear and make it the regmust. Resolve
|
|
1092 * ties in favor of later strings, since the regstart check works
|
|
1093 * with the beginning of the r.e. and avoiding duplication
|
|
1094 * strengthens checking. Not a strong reason, but sufficient in the
|
|
1095 * absence of others.
|
|
1096 */
|
|
1097 /*
|
|
1098 * When the r.e. starts with BOW, it is faster to look for a regmust
|
|
1099 * first. Used a lot for "#" and "*" commands. (Added by mool).
|
|
1100 */
|
|
1101 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
|
|
1102 && !(flags & HASNL))
|
|
1103 {
|
|
1104 longest = NULL;
|
|
1105 len = 0;
|
|
1106 for (; scan != NULL; scan = regnext(scan))
|
|
1107 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
|
|
1108 {
|
|
1109 longest = OPERAND(scan);
|
|
1110 len = (int)STRLEN(OPERAND(scan));
|
|
1111 }
|
|
1112 r->regmust = longest;
|
|
1113 r->regmlen = len;
|
|
1114 }
|
|
1115 }
|
|
1116 #ifdef DEBUG
|
|
1117 regdump(expr, r);
|
|
1118 #endif
|
|
1119 return r;
|
|
1120 }
|
|
1121
|
|
1122 /*
|
|
1123 * Setup to parse the regexp. Used once to get the length and once to do it.
|
|
1124 */
|
|
1125 static void
|
|
1126 regcomp_start(expr, re_flags)
|
|
1127 char_u *expr;
|
|
1128 int re_flags; /* see vim_regcomp() */
|
|
1129 {
|
|
1130 initchr(expr);
|
|
1131 if (re_flags & RE_MAGIC)
|
|
1132 reg_magic = MAGIC_ON;
|
|
1133 else
|
|
1134 reg_magic = MAGIC_OFF;
|
|
1135 reg_string = (re_flags & RE_STRING);
|
481
|
1136 reg_strict = (re_flags & RE_STRICT);
|
7
|
1137
|
|
1138 num_complex_braces = 0;
|
|
1139 regnpar = 1;
|
|
1140 vim_memset(had_endbrace, 0, sizeof(had_endbrace));
|
|
1141 #ifdef FEAT_SYN_HL
|
|
1142 regnzpar = 1;
|
|
1143 re_has_z = 0;
|
|
1144 #endif
|
|
1145 regsize = 0L;
|
|
1146 regflags = 0;
|
|
1147 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1148 had_eol = FALSE;
|
|
1149 #endif
|
|
1150 }
|
|
1151
|
|
1152 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1153 /*
|
|
1154 * Check if during the previous call to vim_regcomp the EOL item "$" has been
|
|
1155 * found. This is messy, but it works fine.
|
|
1156 */
|
|
1157 int
|
|
1158 vim_regcomp_had_eol()
|
|
1159 {
|
|
1160 return had_eol;
|
|
1161 }
|
|
1162 #endif
|
|
1163
|
|
1164 /*
|
|
1165 * reg - regular expression, i.e. main body or parenthesized thing
|
|
1166 *
|
|
1167 * Caller must absorb opening parenthesis.
|
|
1168 *
|
|
1169 * Combining parenthesis handling with the base level of regular expression
|
|
1170 * is a trifle forced, but the need to tie the tails of the branches to what
|
|
1171 * follows makes it hard to avoid.
|
|
1172 */
|
|
1173 static char_u *
|
|
1174 reg(paren, flagp)
|
|
1175 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */
|
|
1176 int *flagp;
|
|
1177 {
|
|
1178 char_u *ret;
|
|
1179 char_u *br;
|
|
1180 char_u *ender;
|
|
1181 int parno = 0;
|
|
1182 int flags;
|
|
1183
|
|
1184 *flagp = HASWIDTH; /* Tentatively. */
|
|
1185
|
|
1186 #ifdef FEAT_SYN_HL
|
|
1187 if (paren == REG_ZPAREN)
|
|
1188 {
|
|
1189 /* Make a ZOPEN node. */
|
|
1190 if (regnzpar >= NSUBEXP)
|
|
1191 EMSG_RET_NULL(_("E50: Too many \\z("));
|
|
1192 parno = regnzpar;
|
|
1193 regnzpar++;
|
|
1194 ret = regnode(ZOPEN + parno);
|
|
1195 }
|
|
1196 else
|
|
1197 #endif
|
|
1198 if (paren == REG_PAREN)
|
|
1199 {
|
|
1200 /* Make a MOPEN node. */
|
|
1201 if (regnpar >= NSUBEXP)
|
|
1202 EMSG_M_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL);
|
|
1203 parno = regnpar;
|
|
1204 ++regnpar;
|
|
1205 ret = regnode(MOPEN + parno);
|
|
1206 }
|
|
1207 else if (paren == REG_NPAREN)
|
|
1208 {
|
|
1209 /* Make a NOPEN node. */
|
|
1210 ret = regnode(NOPEN);
|
|
1211 }
|
|
1212 else
|
|
1213 ret = NULL;
|
|
1214
|
|
1215 /* Pick up the branches, linking them together. */
|
|
1216 br = regbranch(&flags);
|
|
1217 if (br == NULL)
|
|
1218 return NULL;
|
|
1219 if (ret != NULL)
|
|
1220 regtail(ret, br); /* [MZ]OPEN -> first. */
|
|
1221 else
|
|
1222 ret = br;
|
|
1223 /* If one of the branches can be zero-width, the whole thing can.
|
|
1224 * If one of the branches has * at start or matches a line-break, the
|
|
1225 * whole thing can. */
|
|
1226 if (!(flags & HASWIDTH))
|
|
1227 *flagp &= ~HASWIDTH;
|
|
1228 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1229 while (peekchr() == Magic('|'))
|
|
1230 {
|
|
1231 skipchr();
|
|
1232 br = regbranch(&flags);
|
|
1233 if (br == NULL)
|
|
1234 return NULL;
|
|
1235 regtail(ret, br); /* BRANCH -> BRANCH. */
|
|
1236 if (!(flags & HASWIDTH))
|
|
1237 *flagp &= ~HASWIDTH;
|
|
1238 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1239 }
|
|
1240
|
|
1241 /* Make a closing node, and hook it on the end. */
|
|
1242 ender = regnode(
|
|
1243 #ifdef FEAT_SYN_HL
|
|
1244 paren == REG_ZPAREN ? ZCLOSE + parno :
|
|
1245 #endif
|
|
1246 paren == REG_PAREN ? MCLOSE + parno :
|
|
1247 paren == REG_NPAREN ? NCLOSE : END);
|
|
1248 regtail(ret, ender);
|
|
1249
|
|
1250 /* Hook the tails of the branches to the closing node. */
|
|
1251 for (br = ret; br != NULL; br = regnext(br))
|
|
1252 regoptail(br, ender);
|
|
1253
|
|
1254 /* Check for proper termination. */
|
|
1255 if (paren != REG_NOPAREN && getchr() != Magic(')'))
|
|
1256 {
|
|
1257 #ifdef FEAT_SYN_HL
|
|
1258 if (paren == REG_ZPAREN)
|
308
|
1259 EMSG_RET_NULL(_("E52: Unmatched \\z("));
|
7
|
1260 else
|
|
1261 #endif
|
|
1262 if (paren == REG_NPAREN)
|
308
|
1263 EMSG_M_RET_NULL(_("E53: Unmatched %s%%("), reg_magic == MAGIC_ALL);
|
7
|
1264 else
|
308
|
1265 EMSG_M_RET_NULL(_("E54: Unmatched %s("), reg_magic == MAGIC_ALL);
|
7
|
1266 }
|
|
1267 else if (paren == REG_NOPAREN && peekchr() != NUL)
|
|
1268 {
|
|
1269 if (curchr == Magic(')'))
|
308
|
1270 EMSG_M_RET_NULL(_("E55: Unmatched %s)"), reg_magic == MAGIC_ALL);
|
7
|
1271 else
|
308
|
1272 EMSG_RET_NULL(_(e_trailing)); /* "Can't happen". */
|
7
|
1273 /* NOTREACHED */
|
|
1274 }
|
|
1275 /*
|
|
1276 * Here we set the flag allowing back references to this set of
|
|
1277 * parentheses.
|
|
1278 */
|
|
1279 if (paren == REG_PAREN)
|
|
1280 had_endbrace[parno] = TRUE; /* have seen the close paren */
|
|
1281 return ret;
|
|
1282 }
|
|
1283
|
|
1284 /*
|
|
1285 * regbranch - one alternative of an | operator
|
|
1286 *
|
|
1287 * Implements the & operator.
|
|
1288 */
|
|
1289 static char_u *
|
|
1290 regbranch(flagp)
|
|
1291 int *flagp;
|
|
1292 {
|
|
1293 char_u *ret;
|
|
1294 char_u *chain = NULL;
|
|
1295 char_u *latest;
|
|
1296 int flags;
|
|
1297
|
|
1298 *flagp = WORST | HASNL; /* Tentatively. */
|
|
1299
|
|
1300 ret = regnode(BRANCH);
|
|
1301 for (;;)
|
|
1302 {
|
|
1303 latest = regconcat(&flags);
|
|
1304 if (latest == NULL)
|
|
1305 return NULL;
|
|
1306 /* If one of the branches has width, the whole thing has. If one of
|
|
1307 * the branches anchors at start-of-line, the whole thing does.
|
|
1308 * If one of the branches uses look-behind, the whole thing does. */
|
|
1309 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
|
|
1310 /* If one of the branches doesn't match a line-break, the whole thing
|
|
1311 * doesn't. */
|
|
1312 *flagp &= ~HASNL | (flags & HASNL);
|
|
1313 if (chain != NULL)
|
|
1314 regtail(chain, latest);
|
|
1315 if (peekchr() != Magic('&'))
|
|
1316 break;
|
|
1317 skipchr();
|
|
1318 regtail(latest, regnode(END)); /* operand ends */
|
|
1319 reginsert(MATCH, latest);
|
|
1320 chain = latest;
|
|
1321 }
|
|
1322
|
|
1323 return ret;
|
|
1324 }
|
|
1325
|
|
1326 /*
|
|
1327 * regbranch - one alternative of an | or & operator
|
|
1328 *
|
|
1329 * Implements the concatenation operator.
|
|
1330 */
|
|
1331 static char_u *
|
|
1332 regconcat(flagp)
|
|
1333 int *flagp;
|
|
1334 {
|
|
1335 char_u *first = NULL;
|
|
1336 char_u *chain = NULL;
|
|
1337 char_u *latest;
|
|
1338 int flags;
|
|
1339 int cont = TRUE;
|
|
1340
|
|
1341 *flagp = WORST; /* Tentatively. */
|
|
1342
|
|
1343 while (cont)
|
|
1344 {
|
|
1345 switch (peekchr())
|
|
1346 {
|
|
1347 case NUL:
|
|
1348 case Magic('|'):
|
|
1349 case Magic('&'):
|
|
1350 case Magic(')'):
|
|
1351 cont = FALSE;
|
|
1352 break;
|
|
1353 case Magic('Z'):
|
|
1354 #ifdef FEAT_MBYTE
|
|
1355 regflags |= RF_ICOMBINE;
|
|
1356 #endif
|
|
1357 skipchr_keepstart();
|
|
1358 break;
|
|
1359 case Magic('c'):
|
|
1360 regflags |= RF_ICASE;
|
|
1361 skipchr_keepstart();
|
|
1362 break;
|
|
1363 case Magic('C'):
|
|
1364 regflags |= RF_NOICASE;
|
|
1365 skipchr_keepstart();
|
|
1366 break;
|
|
1367 case Magic('v'):
|
|
1368 reg_magic = MAGIC_ALL;
|
|
1369 skipchr_keepstart();
|
|
1370 curchr = -1;
|
|
1371 break;
|
|
1372 case Magic('m'):
|
|
1373 reg_magic = MAGIC_ON;
|
|
1374 skipchr_keepstart();
|
|
1375 curchr = -1;
|
|
1376 break;
|
|
1377 case Magic('M'):
|
|
1378 reg_magic = MAGIC_OFF;
|
|
1379 skipchr_keepstart();
|
|
1380 curchr = -1;
|
|
1381 break;
|
|
1382 case Magic('V'):
|
|
1383 reg_magic = MAGIC_NONE;
|
|
1384 skipchr_keepstart();
|
|
1385 curchr = -1;
|
|
1386 break;
|
|
1387 default:
|
|
1388 latest = regpiece(&flags);
|
|
1389 if (latest == NULL)
|
|
1390 return NULL;
|
|
1391 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
|
|
1392 if (chain == NULL) /* First piece. */
|
|
1393 *flagp |= flags & SPSTART;
|
|
1394 else
|
|
1395 regtail(chain, latest);
|
|
1396 chain = latest;
|
|
1397 if (first == NULL)
|
|
1398 first = latest;
|
|
1399 break;
|
|
1400 }
|
|
1401 }
|
|
1402 if (first == NULL) /* Loop ran zero times. */
|
|
1403 first = regnode(NOTHING);
|
|
1404 return first;
|
|
1405 }
|
|
1406
|
|
1407 /*
|
|
1408 * regpiece - something followed by possible [*+=]
|
|
1409 *
|
|
1410 * Note that the branching code sequences used for = and the general cases
|
|
1411 * of * and + are somewhat optimized: they use the same NOTHING node as
|
|
1412 * both the endmarker for their branch list and the body of the last branch.
|
|
1413 * It might seem that this node could be dispensed with entirely, but the
|
|
1414 * endmarker role is not redundant.
|
|
1415 */
|
|
1416 static char_u *
|
|
1417 regpiece(flagp)
|
|
1418 int *flagp;
|
|
1419 {
|
|
1420 char_u *ret;
|
|
1421 int op;
|
|
1422 char_u *next;
|
|
1423 int flags;
|
|
1424 long minval;
|
|
1425 long maxval;
|
|
1426
|
|
1427 ret = regatom(&flags);
|
|
1428 if (ret == NULL)
|
|
1429 return NULL;
|
|
1430
|
|
1431 op = peekchr();
|
|
1432 if (re_multi_type(op) == NOT_MULTI)
|
|
1433 {
|
|
1434 *flagp = flags;
|
|
1435 return ret;
|
|
1436 }
|
|
1437 /* default flags */
|
|
1438 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
|
|
1439
|
|
1440 skipchr();
|
|
1441 switch (op)
|
|
1442 {
|
|
1443 case Magic('*'):
|
|
1444 if (flags & SIMPLE)
|
|
1445 reginsert(STAR, ret);
|
|
1446 else
|
|
1447 {
|
|
1448 /* Emit x* as (x&|), where & means "self". */
|
|
1449 reginsert(BRANCH, ret); /* Either x */
|
|
1450 regoptail(ret, regnode(BACK)); /* and loop */
|
|
1451 regoptail(ret, ret); /* back */
|
|
1452 regtail(ret, regnode(BRANCH)); /* or */
|
|
1453 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1454 }
|
|
1455 break;
|
|
1456
|
|
1457 case Magic('+'):
|
|
1458 if (flags & SIMPLE)
|
|
1459 reginsert(PLUS, ret);
|
|
1460 else
|
|
1461 {
|
|
1462 /* Emit x+ as x(&|), where & means "self". */
|
|
1463 next = regnode(BRANCH); /* Either */
|
|
1464 regtail(ret, next);
|
233
|
1465 regtail(regnode(BACK), ret); /* loop back */
|
7
|
1466 regtail(next, regnode(BRANCH)); /* or */
|
|
1467 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1468 }
|
|
1469 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1470 break;
|
|
1471
|
|
1472 case Magic('@'):
|
|
1473 {
|
|
1474 int lop = END;
|
|
1475
|
|
1476 switch (no_Magic(getchr()))
|
|
1477 {
|
|
1478 case '=': lop = MATCH; break; /* \@= */
|
|
1479 case '!': lop = NOMATCH; break; /* \@! */
|
|
1480 case '>': lop = SUBPAT; break; /* \@> */
|
|
1481 case '<': switch (no_Magic(getchr()))
|
|
1482 {
|
|
1483 case '=': lop = BEHIND; break; /* \@<= */
|
|
1484 case '!': lop = NOBEHIND; break; /* \@<! */
|
|
1485 }
|
|
1486 }
|
|
1487 if (lop == END)
|
|
1488 EMSG_M_RET_NULL(_("E59: invalid character after %s@"),
|
|
1489 reg_magic == MAGIC_ALL);
|
|
1490 /* Look behind must match with behind_pos. */
|
|
1491 if (lop == BEHIND || lop == NOBEHIND)
|
|
1492 {
|
|
1493 regtail(ret, regnode(BHPOS));
|
|
1494 *flagp |= HASLOOKBH;
|
|
1495 }
|
|
1496 regtail(ret, regnode(END)); /* operand ends */
|
|
1497 reginsert(lop, ret);
|
|
1498 break;
|
|
1499 }
|
|
1500
|
|
1501 case Magic('?'):
|
|
1502 case Magic('='):
|
|
1503 /* Emit x= as (x|) */
|
|
1504 reginsert(BRANCH, ret); /* Either x */
|
|
1505 regtail(ret, regnode(BRANCH)); /* or */
|
|
1506 next = regnode(NOTHING); /* null. */
|
|
1507 regtail(ret, next);
|
|
1508 regoptail(ret, next);
|
|
1509 break;
|
|
1510
|
|
1511 case Magic('{'):
|
|
1512 if (!read_limits(&minval, &maxval))
|
|
1513 return NULL;
|
|
1514 if (flags & SIMPLE)
|
|
1515 {
|
|
1516 reginsert(BRACE_SIMPLE, ret);
|
|
1517 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1518 }
|
|
1519 else
|
|
1520 {
|
|
1521 if (num_complex_braces >= 10)
|
|
1522 EMSG_M_RET_NULL(_("E60: Too many complex %s{...}s"),
|
|
1523 reg_magic == MAGIC_ALL);
|
|
1524 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
|
|
1525 regoptail(ret, regnode(BACK));
|
|
1526 regoptail(ret, ret);
|
|
1527 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1528 ++num_complex_braces;
|
|
1529 }
|
|
1530 if (minval > 0 && maxval > 0)
|
|
1531 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1532 break;
|
|
1533 }
|
|
1534 if (re_multi_type(peekchr()) != NOT_MULTI)
|
|
1535 {
|
|
1536 /* Can't have a multi follow a multi. */
|
|
1537 if (peekchr() == Magic('*'))
|
|
1538 sprintf((char *)IObuff, _("E61: Nested %s*"),
|
|
1539 reg_magic >= MAGIC_ON ? "" : "\\");
|
|
1540 else
|
|
1541 sprintf((char *)IObuff, _("E62: Nested %s%c"),
|
|
1542 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr()));
|
|
1543 EMSG_RET_NULL(IObuff);
|
|
1544 }
|
|
1545
|
|
1546 return ret;
|
|
1547 }
|
|
1548
|
|
1549 /*
|
|
1550 * regatom - the lowest level
|
|
1551 *
|
|
1552 * Optimization: gobbles an entire sequence of ordinary characters so that
|
|
1553 * it can turn them into a single node, which is smaller to store and
|
|
1554 * faster to run. Don't do this when one_exactly is set.
|
|
1555 */
|
|
1556 static char_u *
|
|
1557 regatom(flagp)
|
|
1558 int *flagp;
|
|
1559 {
|
|
1560 char_u *ret;
|
|
1561 int flags;
|
|
1562 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
167
|
1563 int cpo_bsl; /* 'cpoptions' contains '\' flag */
|
7
|
1564 int c;
|
|
1565 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU";
|
|
1566 static int classcodes[] = {ANY, IDENT, SIDENT, KWORD, SKWORD,
|
|
1567 FNAME, SFNAME, PRINT, SPRINT,
|
|
1568 WHITE, NWHITE, DIGIT, NDIGIT,
|
|
1569 HEX, NHEX, OCTAL, NOCTAL,
|
|
1570 WORD, NWORD, HEAD, NHEAD,
|
|
1571 ALPHA, NALPHA, LOWER, NLOWER,
|
|
1572 UPPER, NUPPER
|
|
1573 };
|
|
1574 char_u *p;
|
|
1575 int extra = 0;
|
|
1576
|
|
1577 *flagp = WORST; /* Tentatively. */
|
410
|
1578 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL;
|
|
1579 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL;
|
7
|
1580
|
|
1581 c = getchr();
|
|
1582 switch (c)
|
|
1583 {
|
|
1584 case Magic('^'):
|
|
1585 ret = regnode(BOL);
|
|
1586 break;
|
|
1587
|
|
1588 case Magic('$'):
|
|
1589 ret = regnode(EOL);
|
|
1590 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1591 had_eol = TRUE;
|
|
1592 #endif
|
|
1593 break;
|
|
1594
|
|
1595 case Magic('<'):
|
|
1596 ret = regnode(BOW);
|
|
1597 break;
|
|
1598
|
|
1599 case Magic('>'):
|
|
1600 ret = regnode(EOW);
|
|
1601 break;
|
|
1602
|
|
1603 case Magic('_'):
|
|
1604 c = no_Magic(getchr());
|
|
1605 if (c == '^') /* "\_^" is start-of-line */
|
|
1606 {
|
|
1607 ret = regnode(BOL);
|
|
1608 break;
|
|
1609 }
|
|
1610 if (c == '$') /* "\_$" is end-of-line */
|
|
1611 {
|
|
1612 ret = regnode(EOL);
|
|
1613 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1614 had_eol = TRUE;
|
|
1615 #endif
|
|
1616 break;
|
|
1617 }
|
|
1618
|
|
1619 extra = ADD_NL;
|
|
1620 *flagp |= HASNL;
|
|
1621
|
|
1622 /* "\_[" is character range plus newline */
|
|
1623 if (c == '[')
|
|
1624 goto collection;
|
|
1625
|
|
1626 /* "\_x" is character class plus newline */
|
|
1627 /*FALLTHROUGH*/
|
|
1628
|
|
1629 /*
|
|
1630 * Character classes.
|
|
1631 */
|
|
1632 case Magic('.'):
|
|
1633 case Magic('i'):
|
|
1634 case Magic('I'):
|
|
1635 case Magic('k'):
|
|
1636 case Magic('K'):
|
|
1637 case Magic('f'):
|
|
1638 case Magic('F'):
|
|
1639 case Magic('p'):
|
|
1640 case Magic('P'):
|
|
1641 case Magic('s'):
|
|
1642 case Magic('S'):
|
|
1643 case Magic('d'):
|
|
1644 case Magic('D'):
|
|
1645 case Magic('x'):
|
|
1646 case Magic('X'):
|
|
1647 case Magic('o'):
|
|
1648 case Magic('O'):
|
|
1649 case Magic('w'):
|
|
1650 case Magic('W'):
|
|
1651 case Magic('h'):
|
|
1652 case Magic('H'):
|
|
1653 case Magic('a'):
|
|
1654 case Magic('A'):
|
|
1655 case Magic('l'):
|
|
1656 case Magic('L'):
|
|
1657 case Magic('u'):
|
|
1658 case Magic('U'):
|
|
1659 p = vim_strchr(classchars, no_Magic(c));
|
|
1660 if (p == NULL)
|
|
1661 EMSG_RET_NULL(_("E63: invalid use of \\_"));
|
|
1662 ret = regnode(classcodes[p - classchars] + extra);
|
|
1663 *flagp |= HASWIDTH | SIMPLE;
|
|
1664 break;
|
|
1665
|
|
1666 case Magic('n'):
|
|
1667 if (reg_string)
|
|
1668 {
|
|
1669 /* In a string "\n" matches a newline character. */
|
|
1670 ret = regnode(EXACTLY);
|
|
1671 regc(NL);
|
|
1672 regc(NUL);
|
|
1673 *flagp |= HASWIDTH | SIMPLE;
|
|
1674 }
|
|
1675 else
|
|
1676 {
|
|
1677 /* In buffer text "\n" matches the end of a line. */
|
|
1678 ret = regnode(NEWL);
|
|
1679 *flagp |= HASWIDTH | HASNL;
|
|
1680 }
|
|
1681 break;
|
|
1682
|
|
1683 case Magic('('):
|
|
1684 if (one_exactly)
|
|
1685 EMSG_ONE_RET_NULL;
|
|
1686 ret = reg(REG_PAREN, &flags);
|
|
1687 if (ret == NULL)
|
|
1688 return NULL;
|
|
1689 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1690 break;
|
|
1691
|
|
1692 case NUL:
|
|
1693 case Magic('|'):
|
|
1694 case Magic('&'):
|
|
1695 case Magic(')'):
|
|
1696 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */
|
|
1697 /* NOTREACHED */
|
|
1698
|
|
1699 case Magic('='):
|
|
1700 case Magic('?'):
|
|
1701 case Magic('+'):
|
|
1702 case Magic('@'):
|
|
1703 case Magic('{'):
|
|
1704 case Magic('*'):
|
|
1705 c = no_Magic(c);
|
|
1706 sprintf((char *)IObuff, _("E64: %s%c follows nothing"),
|
|
1707 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL)
|
|
1708 ? "" : "\\", c);
|
|
1709 EMSG_RET_NULL(IObuff);
|
|
1710 /* NOTREACHED */
|
|
1711
|
|
1712 case Magic('~'): /* previous substitute pattern */
|
359
|
1713 if (reg_prev_sub != NULL)
|
7
|
1714 {
|
|
1715 char_u *lp;
|
|
1716
|
|
1717 ret = regnode(EXACTLY);
|
|
1718 lp = reg_prev_sub;
|
|
1719 while (*lp != NUL)
|
|
1720 regc(*lp++);
|
|
1721 regc(NUL);
|
|
1722 if (*reg_prev_sub != NUL)
|
|
1723 {
|
|
1724 *flagp |= HASWIDTH;
|
|
1725 if ((lp - reg_prev_sub) == 1)
|
|
1726 *flagp |= SIMPLE;
|
|
1727 }
|
|
1728 }
|
|
1729 else
|
|
1730 EMSG_RET_NULL(_(e_nopresub));
|
|
1731 break;
|
|
1732
|
|
1733 case Magic('1'):
|
|
1734 case Magic('2'):
|
|
1735 case Magic('3'):
|
|
1736 case Magic('4'):
|
|
1737 case Magic('5'):
|
|
1738 case Magic('6'):
|
|
1739 case Magic('7'):
|
|
1740 case Magic('8'):
|
|
1741 case Magic('9'):
|
|
1742 {
|
|
1743 int refnum;
|
|
1744
|
|
1745 refnum = c - Magic('0');
|
|
1746 /*
|
|
1747 * Check if the back reference is legal. We must have seen the
|
|
1748 * close brace.
|
|
1749 * TODO: Should also check that we don't refer to something
|
|
1750 * that is repeated (+*=): what instance of the repetition
|
|
1751 * should we match?
|
|
1752 */
|
|
1753 if (!had_endbrace[refnum])
|
|
1754 {
|
|
1755 /* Trick: check if "@<=" or "@<!" follows, in which case
|
|
1756 * the \1 can appear before the referenced match. */
|
|
1757 for (p = regparse; *p != NUL; ++p)
|
|
1758 if (p[0] == '@' && p[1] == '<'
|
|
1759 && (p[2] == '!' || p[2] == '='))
|
|
1760 break;
|
|
1761 if (*p == NUL)
|
|
1762 EMSG_RET_NULL(_("E65: Illegal back reference"));
|
|
1763 }
|
|
1764 ret = regnode(BACKREF + refnum);
|
|
1765 }
|
|
1766 break;
|
|
1767
|
|
1768 #ifdef FEAT_SYN_HL
|
|
1769 case Magic('z'):
|
|
1770 {
|
|
1771 c = no_Magic(getchr());
|
|
1772 switch (c)
|
|
1773 {
|
|
1774 case '(': if (reg_do_extmatch != REX_SET)
|
|
1775 EMSG_RET_NULL(_("E66: \\z( not allowed here"));
|
|
1776 if (one_exactly)
|
|
1777 EMSG_ONE_RET_NULL;
|
|
1778 ret = reg(REG_ZPAREN, &flags);
|
|
1779 if (ret == NULL)
|
|
1780 return NULL;
|
|
1781 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH);
|
|
1782 re_has_z = REX_SET;
|
|
1783 break;
|
|
1784
|
|
1785 case '1':
|
|
1786 case '2':
|
|
1787 case '3':
|
|
1788 case '4':
|
|
1789 case '5':
|
|
1790 case '6':
|
|
1791 case '7':
|
|
1792 case '8':
|
|
1793 case '9': if (reg_do_extmatch != REX_USE)
|
|
1794 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here"));
|
|
1795 ret = regnode(ZREF + c - '0');
|
|
1796 re_has_z = REX_USE;
|
|
1797 break;
|
|
1798
|
|
1799 case 's': ret = regnode(MOPEN + 0);
|
|
1800 break;
|
|
1801
|
|
1802 case 'e': ret = regnode(MCLOSE + 0);
|
|
1803 break;
|
|
1804
|
|
1805 default: EMSG_RET_NULL(_("E68: Invalid character after \\z"));
|
|
1806 }
|
|
1807 }
|
|
1808 break;
|
|
1809 #endif
|
|
1810
|
|
1811 case Magic('%'):
|
|
1812 {
|
|
1813 c = no_Magic(getchr());
|
|
1814 switch (c)
|
|
1815 {
|
|
1816 /* () without a back reference */
|
|
1817 case '(':
|
|
1818 if (one_exactly)
|
|
1819 EMSG_ONE_RET_NULL;
|
|
1820 ret = reg(REG_NPAREN, &flags);
|
|
1821 if (ret == NULL)
|
|
1822 return NULL;
|
|
1823 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1824 break;
|
|
1825
|
|
1826 /* Catch \%^ and \%$ regardless of where they appear in the
|
|
1827 * pattern -- regardless of whether or not it makes sense. */
|
|
1828 case '^':
|
|
1829 ret = regnode(RE_BOF);
|
|
1830 break;
|
|
1831
|
|
1832 case '$':
|
|
1833 ret = regnode(RE_EOF);
|
|
1834 break;
|
|
1835
|
|
1836 case '#':
|
|
1837 ret = regnode(CURSOR);
|
|
1838 break;
|
|
1839
|
|
1840 /* \%[abc]: Emit as a list of branches, all ending at the last
|
|
1841 * branch which matches nothing. */
|
|
1842 case '[':
|
|
1843 if (one_exactly) /* doesn't nest */
|
|
1844 EMSG_ONE_RET_NULL;
|
|
1845 {
|
|
1846 char_u *lastbranch;
|
|
1847 char_u *lastnode = NULL;
|
|
1848 char_u *br;
|
|
1849
|
|
1850 ret = NULL;
|
|
1851 while ((c = getchr()) != ']')
|
|
1852 {
|
|
1853 if (c == NUL)
|
|
1854 EMSG_M_RET_NULL(_("E69: Missing ] after %s%%["),
|
|
1855 reg_magic == MAGIC_ALL);
|
|
1856 br = regnode(BRANCH);
|
|
1857 if (ret == NULL)
|
|
1858 ret = br;
|
|
1859 else
|
|
1860 regtail(lastnode, br);
|
|
1861
|
|
1862 ungetchr();
|
|
1863 one_exactly = TRUE;
|
|
1864 lastnode = regatom(flagp);
|
|
1865 one_exactly = FALSE;
|
|
1866 if (lastnode == NULL)
|
|
1867 return NULL;
|
|
1868 }
|
|
1869 if (ret == NULL)
|
|
1870 EMSG_M_RET_NULL(_("E70: Empty %s%%[]"),
|
|
1871 reg_magic == MAGIC_ALL);
|
|
1872 lastbranch = regnode(BRANCH);
|
|
1873 br = regnode(NOTHING);
|
|
1874 if (ret != JUST_CALC_SIZE)
|
|
1875 {
|
|
1876 regtail(lastnode, br);
|
|
1877 regtail(lastbranch, br);
|
|
1878 /* connect all branches to the NOTHING
|
|
1879 * branch at the end */
|
|
1880 for (br = ret; br != lastnode; )
|
|
1881 {
|
|
1882 if (OP(br) == BRANCH)
|
|
1883 {
|
|
1884 regtail(br, lastbranch);
|
|
1885 br = OPERAND(br);
|
|
1886 }
|
|
1887 else
|
|
1888 br = regnext(br);
|
|
1889 }
|
|
1890 }
|
|
1891 *flagp &= ~HASWIDTH;
|
|
1892 break;
|
|
1893 }
|
|
1894
|
24
|
1895 case 'd': /* %d123 decimal */
|
|
1896 case 'o': /* %o123 octal */
|
|
1897 case 'x': /* %xab hex 2 */
|
|
1898 case 'u': /* %uabcd hex 4 */
|
|
1899 case 'U': /* %U1234abcd hex 8 */
|
|
1900 {
|
|
1901 int i;
|
|
1902
|
|
1903 switch (c)
|
|
1904 {
|
|
1905 case 'd': i = getdecchrs(); break;
|
|
1906 case 'o': i = getoctchrs(); break;
|
|
1907 case 'x': i = gethexchrs(2); break;
|
|
1908 case 'u': i = gethexchrs(4); break;
|
|
1909 case 'U': i = gethexchrs(8); break;
|
|
1910 default: i = -1; break;
|
|
1911 }
|
|
1912
|
|
1913 if (i < 0)
|
|
1914 EMSG_M_RET_NULL(
|
|
1915 _("E678: Invalid character after %s%%[dxouU]"),
|
|
1916 reg_magic == MAGIC_ALL);
|
|
1917 ret = regnode(EXACTLY);
|
|
1918 if (i == 0)
|
|
1919 regc(0x0a);
|
|
1920 else
|
|
1921 #ifdef FEAT_MBYTE
|
|
1922 regmbc(i);
|
|
1923 #else
|
|
1924 regc(i);
|
|
1925 #endif
|
|
1926 regc(NUL);
|
|
1927 *flagp |= HASWIDTH;
|
|
1928 break;
|
|
1929 }
|
|
1930
|
7
|
1931 default:
|
|
1932 if (VIM_ISDIGIT(c) || c == '<' || c == '>')
|
|
1933 {
|
|
1934 long_u n = 0;
|
|
1935 int cmp;
|
|
1936
|
|
1937 cmp = c;
|
|
1938 if (cmp == '<' || cmp == '>')
|
|
1939 c = getchr();
|
|
1940 while (VIM_ISDIGIT(c))
|
|
1941 {
|
|
1942 n = n * 10 + (c - '0');
|
|
1943 c = getchr();
|
|
1944 }
|
|
1945 if (c == 'l' || c == 'c' || c == 'v')
|
|
1946 {
|
|
1947 if (c == 'l')
|
|
1948 ret = regnode(RE_LNUM);
|
|
1949 else if (c == 'c')
|
|
1950 ret = regnode(RE_COL);
|
|
1951 else
|
|
1952 ret = regnode(RE_VCOL);
|
|
1953 if (ret == JUST_CALC_SIZE)
|
|
1954 regsize += 5;
|
|
1955 else
|
|
1956 {
|
|
1957 /* put the number and the optional
|
|
1958 * comparator after the opcode */
|
|
1959 regcode = re_put_long(regcode, n);
|
|
1960 *regcode++ = cmp;
|
|
1961 }
|
|
1962 break;
|
|
1963 }
|
|
1964 }
|
|
1965
|
|
1966 EMSG_M_RET_NULL(_("E71: Invalid character after %s%%"),
|
|
1967 reg_magic == MAGIC_ALL);
|
|
1968 }
|
|
1969 }
|
|
1970 break;
|
|
1971
|
|
1972 case Magic('['):
|
|
1973 collection:
|
|
1974 {
|
|
1975 char_u *lp;
|
|
1976
|
|
1977 /*
|
|
1978 * If there is no matching ']', we assume the '[' is a normal
|
|
1979 * character. This makes 'incsearch' and ":help [" work.
|
|
1980 */
|
|
1981 lp = skip_anyof(regparse);
|
|
1982 if (*lp == ']') /* there is a matching ']' */
|
|
1983 {
|
|
1984 int startc = -1; /* > 0 when next '-' is a range */
|
|
1985 int endc;
|
|
1986
|
|
1987 /*
|
|
1988 * In a character class, different parsing rules apply.
|
|
1989 * Not even \ is special anymore, nothing is.
|
|
1990 */
|
|
1991 if (*regparse == '^') /* Complement of range. */
|
|
1992 {
|
|
1993 ret = regnode(ANYBUT + extra);
|
|
1994 regparse++;
|
|
1995 }
|
|
1996 else
|
|
1997 ret = regnode(ANYOF + extra);
|
|
1998
|
|
1999 /* At the start ']' and '-' mean the literal character. */
|
|
2000 if (*regparse == ']' || *regparse == '-')
|
167
|
2001 {
|
|
2002 startc = *regparse;
|
7
|
2003 regc(*regparse++);
|
167
|
2004 }
|
7
|
2005
|
|
2006 while (*regparse != NUL && *regparse != ']')
|
|
2007 {
|
|
2008 if (*regparse == '-')
|
|
2009 {
|
|
2010 ++regparse;
|
|
2011 /* The '-' is not used for a range at the end and
|
|
2012 * after or before a '\n'. */
|
|
2013 if (*regparse == ']' || *regparse == NUL
|
|
2014 || startc == -1
|
|
2015 || (regparse[0] == '\\' && regparse[1] == 'n'))
|
|
2016 {
|
|
2017 regc('-');
|
|
2018 startc = '-'; /* [--x] is a range */
|
|
2019 }
|
|
2020 else
|
|
2021 {
|
167
|
2022 /* Also accept "a-[.z.]" */
|
|
2023 endc = 0;
|
|
2024 if (*regparse == '[')
|
|
2025 endc = get_coll_element(®parse);
|
|
2026 if (endc == 0)
|
|
2027 {
|
7
|
2028 #ifdef FEAT_MBYTE
|
167
|
2029 if (has_mbyte)
|
|
2030 endc = mb_ptr2char_adv(®parse);
|
|
2031 else
|
7
|
2032 #endif
|
167
|
2033 endc = *regparse++;
|
|
2034 }
|
24
|
2035
|
|
2036 /* Handle \o40, \x20 and \u20AC style sequences */
|
167
|
2037 if (endc == '\\' && !cpo_lit && !cpo_bsl)
|
24
|
2038 endc = coll_get_char();
|
|
2039
|
7
|
2040 if (startc > endc)
|
|
2041 EMSG_RET_NULL(_(e_invrange));
|
|
2042 #ifdef FEAT_MBYTE
|
|
2043 if (has_mbyte && ((*mb_char2len)(startc) > 1
|
|
2044 || (*mb_char2len)(endc) > 1))
|
|
2045 {
|
|
2046 /* Limit to a range of 256 chars */
|
|
2047 if (endc > startc + 256)
|
|
2048 EMSG_RET_NULL(_(e_invrange));
|
|
2049 while (++startc <= endc)
|
|
2050 regmbc(startc);
|
|
2051 }
|
|
2052 else
|
|
2053 #endif
|
|
2054 {
|
|
2055 #ifdef EBCDIC
|
|
2056 int alpha_only = FALSE;
|
|
2057
|
|
2058 /* for alphabetical range skip the gaps
|
|
2059 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */
|
|
2060 if (isalpha(startc) && isalpha(endc))
|
|
2061 alpha_only = TRUE;
|
|
2062 #endif
|
|
2063 while (++startc <= endc)
|
|
2064 #ifdef EBCDIC
|
|
2065 if (!alpha_only || isalpha(startc))
|
|
2066 #endif
|
|
2067 regc(startc);
|
|
2068 }
|
|
2069 startc = -1;
|
|
2070 }
|
|
2071 }
|
|
2072 /*
|
|
2073 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
|
|
2074 * accepts "\t", "\e", etc., but only when the 'l' flag in
|
|
2075 * 'cpoptions' is not included.
|
167
|
2076 * Posix doesn't recognize backslash at all.
|
7
|
2077 */
|
|
2078 else if (*regparse == '\\'
|
167
|
2079 && !cpo_bsl
|
7
|
2080 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
|
|
2081 || (!cpo_lit
|
|
2082 && vim_strchr(REGEXP_ABBR,
|
|
2083 regparse[1]) != NULL)))
|
|
2084 {
|
|
2085 regparse++;
|
|
2086 if (*regparse == 'n')
|
|
2087 {
|
|
2088 /* '\n' in range: also match NL */
|
|
2089 if (ret != JUST_CALC_SIZE)
|
|
2090 {
|
|
2091 if (*ret == ANYBUT)
|
|
2092 *ret = ANYBUT + ADD_NL;
|
|
2093 else if (*ret == ANYOF)
|
|
2094 *ret = ANYOF + ADD_NL;
|
|
2095 /* else: must have had a \n already */
|
|
2096 }
|
|
2097 *flagp |= HASNL;
|
|
2098 regparse++;
|
|
2099 startc = -1;
|
|
2100 }
|
24
|
2101 else if (*regparse == 'd'
|
|
2102 || *regparse == 'o'
|
|
2103 || *regparse == 'x'
|
|
2104 || *regparse == 'u'
|
|
2105 || *regparse == 'U')
|
|
2106 {
|
|
2107 startc = coll_get_char();
|
|
2108 if (startc == 0)
|
|
2109 regc(0x0a);
|
|
2110 else
|
|
2111 #ifdef FEAT_MBYTE
|
|
2112 regmbc(startc);
|
|
2113 #else
|
|
2114 regc(startc);
|
|
2115 #endif
|
|
2116 }
|
7
|
2117 else
|
|
2118 {
|
|
2119 startc = backslash_trans(*regparse++);
|
|
2120 regc(startc);
|
|
2121 }
|
|
2122 }
|
|
2123 else if (*regparse == '[')
|
|
2124 {
|
|
2125 int c_class;
|
|
2126 int cu;
|
|
2127
|
167
|
2128 c_class = get_char_class(®parse);
|
7
|
2129 startc = -1;
|
|
2130 /* Characters assumed to be 8 bits! */
|
|
2131 switch (c_class)
|
|
2132 {
|
|
2133 case CLASS_NONE:
|
167
|
2134 c_class = get_equi_class(®parse);
|
|
2135 if (c_class != 0)
|
|
2136 {
|
|
2137 /* produce equivalence class */
|
|
2138 reg_equi_class(c_class);
|
|
2139 }
|
|
2140 else if ((c_class =
|
|
2141 get_coll_element(®parse)) != 0)
|
|
2142 {
|
|
2143 /* produce a collating element */
|
|
2144 regmbc(c_class);
|
|
2145 }
|
|
2146 else
|
|
2147 {
|
|
2148 /* literal '[', allow [[-x] as a range */
|
|
2149 startc = *regparse++;
|
|
2150 regc(startc);
|
|
2151 }
|
7
|
2152 break;
|
|
2153 case CLASS_ALNUM:
|
|
2154 for (cu = 1; cu <= 255; cu++)
|
|
2155 if (isalnum(cu))
|
|
2156 regc(cu);
|
|
2157 break;
|
|
2158 case CLASS_ALPHA:
|
|
2159 for (cu = 1; cu <= 255; cu++)
|
|
2160 if (isalpha(cu))
|
|
2161 regc(cu);
|
|
2162 break;
|
|
2163 case CLASS_BLANK:
|
|
2164 regc(' ');
|
|
2165 regc('\t');
|
|
2166 break;
|
|
2167 case CLASS_CNTRL:
|
|
2168 for (cu = 1; cu <= 255; cu++)
|
|
2169 if (iscntrl(cu))
|
|
2170 regc(cu);
|
|
2171 break;
|
|
2172 case CLASS_DIGIT:
|
|
2173 for (cu = 1; cu <= 255; cu++)
|
|
2174 if (VIM_ISDIGIT(cu))
|
|
2175 regc(cu);
|
|
2176 break;
|
|
2177 case CLASS_GRAPH:
|
|
2178 for (cu = 1; cu <= 255; cu++)
|
|
2179 if (isgraph(cu))
|
|
2180 regc(cu);
|
|
2181 break;
|
|
2182 case CLASS_LOWER:
|
|
2183 for (cu = 1; cu <= 255; cu++)
|
|
2184 if (islower(cu))
|
|
2185 regc(cu);
|
|
2186 break;
|
|
2187 case CLASS_PRINT:
|
|
2188 for (cu = 1; cu <= 255; cu++)
|
|
2189 if (vim_isprintc(cu))
|
|
2190 regc(cu);
|
|
2191 break;
|
|
2192 case CLASS_PUNCT:
|
|
2193 for (cu = 1; cu <= 255; cu++)
|
|
2194 if (ispunct(cu))
|
|
2195 regc(cu);
|
|
2196 break;
|
|
2197 case CLASS_SPACE:
|
|
2198 for (cu = 9; cu <= 13; cu++)
|
|
2199 regc(cu);
|
|
2200 regc(' ');
|
|
2201 break;
|
|
2202 case CLASS_UPPER:
|
|
2203 for (cu = 1; cu <= 255; cu++)
|
|
2204 if (isupper(cu))
|
|
2205 regc(cu);
|
|
2206 break;
|
|
2207 case CLASS_XDIGIT:
|
|
2208 for (cu = 1; cu <= 255; cu++)
|
|
2209 if (vim_isxdigit(cu))
|
|
2210 regc(cu);
|
|
2211 break;
|
|
2212 case CLASS_TAB:
|
|
2213 regc('\t');
|
|
2214 break;
|
|
2215 case CLASS_RETURN:
|
|
2216 regc('\r');
|
|
2217 break;
|
|
2218 case CLASS_BACKSPACE:
|
|
2219 regc('\b');
|
|
2220 break;
|
|
2221 case CLASS_ESCAPE:
|
|
2222 regc('\033');
|
|
2223 break;
|
|
2224 }
|
|
2225 }
|
|
2226 else
|
|
2227 {
|
|
2228 #ifdef FEAT_MBYTE
|
|
2229 if (has_mbyte)
|
|
2230 {
|
|
2231 int len;
|
|
2232
|
|
2233 /* produce a multibyte character, including any
|
|
2234 * following composing characters */
|
|
2235 startc = mb_ptr2char(regparse);
|
474
|
2236 len = (*mb_ptr2len)(regparse);
|
7
|
2237 if (enc_utf8 && utf_char2len(startc) != len)
|
|
2238 startc = -1; /* composing chars */
|
|
2239 while (--len >= 0)
|
|
2240 regc(*regparse++);
|
|
2241 }
|
|
2242 else
|
|
2243 #endif
|
|
2244 {
|
|
2245 startc = *regparse++;
|
|
2246 regc(startc);
|
|
2247 }
|
|
2248 }
|
|
2249 }
|
|
2250 regc(NUL);
|
|
2251 prevchr_len = 1; /* last char was the ']' */
|
|
2252 if (*regparse != ']')
|
|
2253 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */
|
|
2254 skipchr(); /* let's be friends with the lexer again */
|
|
2255 *flagp |= HASWIDTH | SIMPLE;
|
|
2256 break;
|
|
2257 }
|
481
|
2258 else if (reg_strict)
|
|
2259 EMSG_M_RET_NULL(_("E769: Missing ] after %s["),
|
|
2260 reg_magic > MAGIC_OFF);
|
7
|
2261 }
|
|
2262 /* FALLTHROUGH */
|
|
2263
|
|
2264 default:
|
|
2265 {
|
|
2266 int len;
|
|
2267
|
|
2268 #ifdef FEAT_MBYTE
|
|
2269 /* A multi-byte character is handled as a separate atom if it's
|
|
2270 * before a multi. */
|
|
2271 if (has_mbyte && (*mb_char2len)(c) > 1
|
|
2272 && re_multi_type(peekchr()) != NOT_MULTI)
|
|
2273 {
|
|
2274 ret = regnode(MULTIBYTECODE);
|
|
2275 regmbc(c);
|
|
2276 *flagp |= HASWIDTH | SIMPLE;
|
|
2277 break;
|
|
2278 }
|
|
2279 #endif
|
|
2280
|
|
2281 ret = regnode(EXACTLY);
|
|
2282
|
|
2283 /*
|
|
2284 * Append characters as long as:
|
|
2285 * - there is no following multi, we then need the character in
|
|
2286 * front of it as a single character operand
|
|
2287 * - not running into a Magic character
|
|
2288 * - "one_exactly" is not set
|
|
2289 * But always emit at least one character. Might be a Multi,
|
|
2290 * e.g., a "[" without matching "]".
|
|
2291 */
|
|
2292 for (len = 0; c != NUL && (len == 0
|
|
2293 || (re_multi_type(peekchr()) == NOT_MULTI
|
|
2294 && !one_exactly
|
|
2295 && !is_Magic(c))); ++len)
|
|
2296 {
|
|
2297 c = no_Magic(c);
|
|
2298 #ifdef FEAT_MBYTE
|
|
2299 if (has_mbyte)
|
|
2300 {
|
|
2301 regmbc(c);
|
|
2302 if (enc_utf8)
|
|
2303 {
|
|
2304 int off;
|
|
2305 int l;
|
|
2306
|
|
2307 /* Need to get composing character too, directly
|
|
2308 * access regparse for that, because skipchr() skips
|
|
2309 * over composing chars. */
|
|
2310 ungetchr();
|
|
2311 if (*regparse == '\\' && regparse[1] != NUL)
|
|
2312 off = 1;
|
|
2313 else
|
|
2314 off = 0;
|
|
2315 for (;;)
|
|
2316 {
|
474
|
2317 l = utf_ptr2len(regparse + off);
|
7
|
2318 if (!UTF_COMPOSINGLIKE(regparse + off,
|
|
2319 regparse + off + l))
|
|
2320 break;
|
|
2321 off += l;
|
|
2322 regmbc(utf_ptr2char(regparse + off));
|
|
2323 }
|
|
2324 skipchr();
|
|
2325 }
|
|
2326 }
|
|
2327 else
|
|
2328 #endif
|
|
2329 regc(c);
|
|
2330 c = getchr();
|
|
2331 }
|
|
2332 ungetchr();
|
|
2333
|
|
2334 regc(NUL);
|
|
2335 *flagp |= HASWIDTH;
|
|
2336 if (len == 1)
|
|
2337 *flagp |= SIMPLE;
|
|
2338 }
|
|
2339 break;
|
|
2340 }
|
|
2341
|
|
2342 return ret;
|
|
2343 }
|
|
2344
|
|
2345 /*
|
|
2346 * emit a node
|
|
2347 * Return pointer to generated code.
|
|
2348 */
|
|
2349 static char_u *
|
|
2350 regnode(op)
|
|
2351 int op;
|
|
2352 {
|
|
2353 char_u *ret;
|
|
2354
|
|
2355 ret = regcode;
|
|
2356 if (ret == JUST_CALC_SIZE)
|
|
2357 regsize += 3;
|
|
2358 else
|
|
2359 {
|
|
2360 *regcode++ = op;
|
|
2361 *regcode++ = NUL; /* Null "next" pointer. */
|
|
2362 *regcode++ = NUL;
|
|
2363 }
|
|
2364 return ret;
|
|
2365 }
|
|
2366
|
|
2367 /*
|
|
2368 * Emit (if appropriate) a byte of code
|
|
2369 */
|
|
2370 static void
|
|
2371 regc(b)
|
|
2372 int b;
|
|
2373 {
|
|
2374 if (regcode == JUST_CALC_SIZE)
|
|
2375 regsize++;
|
|
2376 else
|
|
2377 *regcode++ = b;
|
|
2378 }
|
|
2379
|
|
2380 #ifdef FEAT_MBYTE
|
|
2381 /*
|
|
2382 * Emit (if appropriate) a multi-byte character of code
|
|
2383 */
|
|
2384 static void
|
|
2385 regmbc(c)
|
|
2386 int c;
|
|
2387 {
|
|
2388 if (regcode == JUST_CALC_SIZE)
|
|
2389 regsize += (*mb_char2len)(c);
|
|
2390 else
|
|
2391 regcode += (*mb_char2bytes)(c, regcode);
|
|
2392 }
|
|
2393 #endif
|
|
2394
|
|
2395 /*
|
|
2396 * reginsert - insert an operator in front of already-emitted operand
|
|
2397 *
|
|
2398 * Means relocating the operand.
|
|
2399 */
|
|
2400 static void
|
|
2401 reginsert(op, opnd)
|
|
2402 int op;
|
|
2403 char_u *opnd;
|
|
2404 {
|
|
2405 char_u *src;
|
|
2406 char_u *dst;
|
|
2407 char_u *place;
|
|
2408
|
|
2409 if (regcode == JUST_CALC_SIZE)
|
|
2410 {
|
|
2411 regsize += 3;
|
|
2412 return;
|
|
2413 }
|
|
2414 src = regcode;
|
|
2415 regcode += 3;
|
|
2416 dst = regcode;
|
|
2417 while (src > opnd)
|
|
2418 *--dst = *--src;
|
|
2419
|
|
2420 place = opnd; /* Op node, where operand used to be. */
|
|
2421 *place++ = op;
|
|
2422 *place++ = NUL;
|
|
2423 *place = NUL;
|
|
2424 }
|
|
2425
|
|
2426 /*
|
|
2427 * reginsert_limits - insert an operator in front of already-emitted operand.
|
|
2428 * The operator has the given limit values as operands. Also set next pointer.
|
|
2429 *
|
|
2430 * Means relocating the operand.
|
|
2431 */
|
|
2432 static void
|
|
2433 reginsert_limits(op, minval, maxval, opnd)
|
|
2434 int op;
|
|
2435 long minval;
|
|
2436 long maxval;
|
|
2437 char_u *opnd;
|
|
2438 {
|
|
2439 char_u *src;
|
|
2440 char_u *dst;
|
|
2441 char_u *place;
|
|
2442
|
|
2443 if (regcode == JUST_CALC_SIZE)
|
|
2444 {
|
|
2445 regsize += 11;
|
|
2446 return;
|
|
2447 }
|
|
2448 src = regcode;
|
|
2449 regcode += 11;
|
|
2450 dst = regcode;
|
|
2451 while (src > opnd)
|
|
2452 *--dst = *--src;
|
|
2453
|
|
2454 place = opnd; /* Op node, where operand used to be. */
|
|
2455 *place++ = op;
|
|
2456 *place++ = NUL;
|
|
2457 *place++ = NUL;
|
|
2458 place = re_put_long(place, (long_u)minval);
|
|
2459 place = re_put_long(place, (long_u)maxval);
|
|
2460 regtail(opnd, place);
|
|
2461 }
|
|
2462
|
|
2463 /*
|
|
2464 * Write a long as four bytes at "p" and return pointer to the next char.
|
|
2465 */
|
|
2466 static char_u *
|
|
2467 re_put_long(p, val)
|
|
2468 char_u *p;
|
|
2469 long_u val;
|
|
2470 {
|
|
2471 *p++ = (char_u) ((val >> 24) & 0377);
|
|
2472 *p++ = (char_u) ((val >> 16) & 0377);
|
|
2473 *p++ = (char_u) ((val >> 8) & 0377);
|
|
2474 *p++ = (char_u) (val & 0377);
|
|
2475 return p;
|
|
2476 }
|
|
2477
|
|
2478 /*
|
|
2479 * regtail - set the next-pointer at the end of a node chain
|
|
2480 */
|
|
2481 static void
|
|
2482 regtail(p, val)
|
|
2483 char_u *p;
|
|
2484 char_u *val;
|
|
2485 {
|
|
2486 char_u *scan;
|
|
2487 char_u *temp;
|
|
2488 int offset;
|
|
2489
|
|
2490 if (p == JUST_CALC_SIZE)
|
|
2491 return;
|
|
2492
|
|
2493 /* Find last node. */
|
|
2494 scan = p;
|
|
2495 for (;;)
|
|
2496 {
|
|
2497 temp = regnext(scan);
|
|
2498 if (temp == NULL)
|
|
2499 break;
|
|
2500 scan = temp;
|
|
2501 }
|
|
2502
|
233
|
2503 if (OP(scan) == BACK)
|
7
|
2504 offset = (int)(scan - val);
|
|
2505 else
|
|
2506 offset = (int)(val - scan);
|
|
2507 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377);
|
|
2508 *(scan + 2) = (char_u) (offset & 0377);
|
|
2509 }
|
|
2510
|
|
2511 /*
|
|
2512 * regoptail - regtail on item after a BRANCH; nop if none
|
|
2513 */
|
|
2514 static void
|
|
2515 regoptail(p, val)
|
|
2516 char_u *p;
|
|
2517 char_u *val;
|
|
2518 {
|
|
2519 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */
|
|
2520 if (p == NULL || p == JUST_CALC_SIZE
|
|
2521 || (OP(p) != BRANCH
|
|
2522 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9)))
|
|
2523 return;
|
|
2524 regtail(OPERAND(p), val);
|
|
2525 }
|
|
2526
|
|
2527 /*
|
|
2528 * getchr() - get the next character from the pattern. We know about
|
|
2529 * magic and such, so therefore we need a lexical analyzer.
|
|
2530 */
|
|
2531
|
|
2532 /* static int curchr; */
|
|
2533 static int prevprevchr;
|
|
2534 static int prevchr;
|
|
2535 static int nextchr; /* used for ungetchr() */
|
|
2536 /*
|
|
2537 * Note: prevchr is sometimes -1 when we are not at the start,
|
|
2538 * eg in /[ ^I]^ the pattern was never found even if it existed, because ^ was
|
|
2539 * taken to be magic -- webb
|
|
2540 */
|
|
2541 static int at_start; /* True when on the first character */
|
|
2542 static int prev_at_start; /* True when on the second character */
|
|
2543
|
|
2544 static void
|
|
2545 initchr(str)
|
|
2546 char_u *str;
|
|
2547 {
|
|
2548 regparse = str;
|
|
2549 prevchr_len = 0;
|
|
2550 curchr = prevprevchr = prevchr = nextchr = -1;
|
|
2551 at_start = TRUE;
|
|
2552 prev_at_start = FALSE;
|
|
2553 }
|
|
2554
|
|
2555 static int
|
|
2556 peekchr()
|
|
2557 {
|
167
|
2558 static int after_slash = FALSE;
|
|
2559
|
7
|
2560 if (curchr == -1)
|
|
2561 {
|
|
2562 switch (curchr = regparse[0])
|
|
2563 {
|
|
2564 case '.':
|
|
2565 case '[':
|
|
2566 case '~':
|
|
2567 /* magic when 'magic' is on */
|
|
2568 if (reg_magic >= MAGIC_ON)
|
|
2569 curchr = Magic(curchr);
|
|
2570 break;
|
|
2571 case '(':
|
|
2572 case ')':
|
|
2573 case '{':
|
|
2574 case '%':
|
|
2575 case '+':
|
|
2576 case '=':
|
|
2577 case '?':
|
|
2578 case '@':
|
|
2579 case '!':
|
|
2580 case '&':
|
|
2581 case '|':
|
|
2582 case '<':
|
|
2583 case '>':
|
|
2584 case '#': /* future ext. */
|
|
2585 case '"': /* future ext. */
|
|
2586 case '\'': /* future ext. */
|
|
2587 case ',': /* future ext. */
|
|
2588 case '-': /* future ext. */
|
|
2589 case ':': /* future ext. */
|
|
2590 case ';': /* future ext. */
|
|
2591 case '`': /* future ext. */
|
|
2592 case '/': /* Can't be used in / command */
|
|
2593 /* magic only after "\v" */
|
|
2594 if (reg_magic == MAGIC_ALL)
|
|
2595 curchr = Magic(curchr);
|
|
2596 break;
|
|
2597 case '*':
|
167
|
2598 /* * is not magic as the very first character, eg "?*ptr", when
|
|
2599 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But
|
|
2600 * "\(\*" is not magic, thus must be magic if "after_slash" */
|
|
2601 if (reg_magic >= MAGIC_ON
|
|
2602 && !at_start
|
|
2603 && !(prev_at_start && prevchr == Magic('^'))
|
|
2604 && (after_slash
|
|
2605 || (prevchr != Magic('(')
|
|
2606 && prevchr != Magic('&')
|
|
2607 && prevchr != Magic('|'))))
|
7
|
2608 curchr = Magic('*');
|
|
2609 break;
|
|
2610 case '^':
|
|
2611 /* '^' is only magic as the very first character and if it's after
|
|
2612 * "\(", "\|", "\&' or "\n" */
|
|
2613 if (reg_magic >= MAGIC_OFF
|
|
2614 && (at_start
|
|
2615 || reg_magic == MAGIC_ALL
|
|
2616 || prevchr == Magic('(')
|
|
2617 || prevchr == Magic('|')
|
|
2618 || prevchr == Magic('&')
|
|
2619 || prevchr == Magic('n')
|
|
2620 || (no_Magic(prevchr) == '('
|
|
2621 && prevprevchr == Magic('%'))))
|
|
2622 {
|
|
2623 curchr = Magic('^');
|
|
2624 at_start = TRUE;
|
|
2625 prev_at_start = FALSE;
|
|
2626 }
|
|
2627 break;
|
|
2628 case '$':
|
|
2629 /* '$' is only magic as the very last char and if it's in front of
|
|
2630 * either "\|", "\)", "\&", or "\n" */
|
|
2631 if (reg_magic >= MAGIC_OFF)
|
|
2632 {
|
|
2633 char_u *p = regparse + 1;
|
|
2634
|
|
2635 /* ignore \c \C \m and \M after '$' */
|
|
2636 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C'
|
|
2637 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z'))
|
|
2638 p += 2;
|
|
2639 if (p[0] == NUL
|
|
2640 || (p[0] == '\\'
|
|
2641 && (p[1] == '|' || p[1] == '&' || p[1] == ')'
|
|
2642 || p[1] == 'n'))
|
|
2643 || reg_magic == MAGIC_ALL)
|
|
2644 curchr = Magic('$');
|
|
2645 }
|
|
2646 break;
|
|
2647 case '\\':
|
|
2648 {
|
|
2649 int c = regparse[1];
|
|
2650
|
|
2651 if (c == NUL)
|
|
2652 curchr = '\\'; /* trailing '\' */
|
|
2653 else if (
|
|
2654 #ifdef EBCDIC
|
|
2655 vim_strchr(META, c)
|
|
2656 #else
|
|
2657 c <= '~' && META_flags[c]
|
|
2658 #endif
|
|
2659 )
|
|
2660 {
|
|
2661 /*
|
|
2662 * META contains everything that may be magic sometimes,
|
|
2663 * except ^ and $ ("\^" and "\$" are only magic after
|
|
2664 * "\v"). We now fetch the next character and toggle its
|
|
2665 * magicness. Therefore, \ is so meta-magic that it is
|
|
2666 * not in META.
|
|
2667 */
|
|
2668 curchr = -1;
|
|
2669 prev_at_start = at_start;
|
|
2670 at_start = FALSE; /* be able to say "/\*ptr" */
|
|
2671 ++regparse;
|
167
|
2672 ++after_slash;
|
7
|
2673 peekchr();
|
|
2674 --regparse;
|
167
|
2675 --after_slash;
|
7
|
2676 curchr = toggle_Magic(curchr);
|
|
2677 }
|
|
2678 else if (vim_strchr(REGEXP_ABBR, c))
|
|
2679 {
|
|
2680 /*
|
|
2681 * Handle abbreviations, like "\t" for TAB -- webb
|
|
2682 */
|
|
2683 curchr = backslash_trans(c);
|
|
2684 }
|
|
2685 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^'))
|
|
2686 curchr = toggle_Magic(c);
|
|
2687 else
|
|
2688 {
|
|
2689 /*
|
|
2690 * Next character can never be (made) magic?
|
|
2691 * Then backslashing it won't do anything.
|
|
2692 */
|
|
2693 #ifdef FEAT_MBYTE
|
|
2694 if (has_mbyte)
|
|
2695 curchr = (*mb_ptr2char)(regparse + 1);
|
|
2696 else
|
|
2697 #endif
|
|
2698 curchr = c;
|
|
2699 }
|
|
2700 break;
|
|
2701 }
|
|
2702
|
|
2703 #ifdef FEAT_MBYTE
|
|
2704 default:
|
|
2705 if (has_mbyte)
|
|
2706 curchr = (*mb_ptr2char)(regparse);
|
|
2707 #endif
|
|
2708 }
|
|
2709 }
|
|
2710
|
|
2711 return curchr;
|
|
2712 }
|
|
2713
|
|
2714 /*
|
|
2715 * Eat one lexed character. Do this in a way that we can undo it.
|
|
2716 */
|
|
2717 static void
|
|
2718 skipchr()
|
|
2719 {
|
|
2720 /* peekchr() eats a backslash, do the same here */
|
|
2721 if (*regparse == '\\')
|
|
2722 prevchr_len = 1;
|
|
2723 else
|
|
2724 prevchr_len = 0;
|
|
2725 if (regparse[prevchr_len] != NUL)
|
|
2726 {
|
|
2727 #ifdef FEAT_MBYTE
|
|
2728 if (has_mbyte)
|
474
|
2729 prevchr_len += (*mb_ptr2len)(regparse + prevchr_len);
|
7
|
2730 else
|
|
2731 #endif
|
|
2732 ++prevchr_len;
|
|
2733 }
|
|
2734 regparse += prevchr_len;
|
|
2735 prev_at_start = at_start;
|
|
2736 at_start = FALSE;
|
|
2737 prevprevchr = prevchr;
|
|
2738 prevchr = curchr;
|
|
2739 curchr = nextchr; /* use previously unget char, or -1 */
|
|
2740 nextchr = -1;
|
|
2741 }
|
|
2742
|
|
2743 /*
|
|
2744 * Skip a character while keeping the value of prev_at_start for at_start.
|
|
2745 * prevchr and prevprevchr are also kept.
|
|
2746 */
|
|
2747 static void
|
|
2748 skipchr_keepstart()
|
|
2749 {
|
|
2750 int as = prev_at_start;
|
|
2751 int pr = prevchr;
|
|
2752 int prpr = prevprevchr;
|
|
2753
|
|
2754 skipchr();
|
|
2755 at_start = as;
|
|
2756 prevchr = pr;
|
|
2757 prevprevchr = prpr;
|
|
2758 }
|
|
2759
|
|
2760 static int
|
|
2761 getchr()
|
|
2762 {
|
|
2763 int chr = peekchr();
|
|
2764
|
|
2765 skipchr();
|
|
2766 return chr;
|
|
2767 }
|
|
2768
|
|
2769 /*
|
|
2770 * put character back. Works only once!
|
|
2771 */
|
|
2772 static void
|
|
2773 ungetchr()
|
|
2774 {
|
|
2775 nextchr = curchr;
|
|
2776 curchr = prevchr;
|
|
2777 prevchr = prevprevchr;
|
|
2778 at_start = prev_at_start;
|
|
2779 prev_at_start = FALSE;
|
|
2780
|
|
2781 /* Backup regparse, so that it's at the same position as before the
|
|
2782 * getchr(). */
|
|
2783 regparse -= prevchr_len;
|
|
2784 }
|
|
2785
|
|
2786 /*
|
29
|
2787 * Get and return the value of the hex string at the current position.
|
|
2788 * Return -1 if there is no valid hex number.
|
|
2789 * The position is updated:
|
24
|
2790 * blahblah\%x20asdf
|
|
2791 * before-^ ^-after
|
|
2792 * The parameter controls the maximum number of input characters. This will be
|
|
2793 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence.
|
|
2794 */
|
|
2795 static int
|
|
2796 gethexchrs(maxinputlen)
|
|
2797 int maxinputlen;
|
|
2798 {
|
|
2799 int nr = 0;
|
|
2800 int c;
|
|
2801 int i;
|
|
2802
|
|
2803 for (i = 0; i < maxinputlen; ++i)
|
|
2804 {
|
|
2805 c = regparse[0];
|
|
2806 if (!vim_isxdigit(c))
|
|
2807 break;
|
|
2808 nr <<= 4;
|
|
2809 nr |= hex2nr(c);
|
|
2810 ++regparse;
|
|
2811 }
|
|
2812
|
|
2813 if (i == 0)
|
|
2814 return -1;
|
|
2815 return nr;
|
|
2816 }
|
|
2817
|
|
2818 /*
|
|
2819 * get and return the value of the decimal string immediately after the
|
|
2820 * current position. Return -1 for invalid. Consumes all digits.
|
|
2821 */
|
|
2822 static int
|
|
2823 getdecchrs()
|
|
2824 {
|
|
2825 int nr = 0;
|
|
2826 int c;
|
|
2827 int i;
|
|
2828
|
|
2829 for (i = 0; ; ++i)
|
|
2830 {
|
|
2831 c = regparse[0];
|
|
2832 if (c < '0' || c > '9')
|
|
2833 break;
|
|
2834 nr *= 10;
|
|
2835 nr += c - '0';
|
|
2836 ++regparse;
|
|
2837 }
|
|
2838
|
|
2839 if (i == 0)
|
|
2840 return -1;
|
|
2841 return nr;
|
|
2842 }
|
|
2843
|
|
2844 /*
|
|
2845 * get and return the value of the octal string immediately after the current
|
|
2846 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle
|
|
2847 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't
|
|
2848 * treat 8 or 9 as recognised characters. Position is updated:
|
|
2849 * blahblah\%o210asdf
|
|
2850 * before-^ ^-after
|
|
2851 */
|
|
2852 static int
|
|
2853 getoctchrs()
|
|
2854 {
|
|
2855 int nr = 0;
|
|
2856 int c;
|
|
2857 int i;
|
|
2858
|
|
2859 for (i = 0; i < 3 && nr < 040; ++i)
|
|
2860 {
|
|
2861 c = regparse[0];
|
|
2862 if (c < '0' || c > '7')
|
|
2863 break;
|
|
2864 nr <<= 3;
|
|
2865 nr |= hex2nr(c);
|
|
2866 ++regparse;
|
|
2867 }
|
|
2868
|
|
2869 if (i == 0)
|
|
2870 return -1;
|
|
2871 return nr;
|
|
2872 }
|
|
2873
|
|
2874 /*
|
|
2875 * Get a number after a backslash that is inside [].
|
|
2876 * When nothing is recognized return a backslash.
|
|
2877 */
|
|
2878 static int
|
|
2879 coll_get_char()
|
|
2880 {
|
|
2881 int nr = -1;
|
|
2882
|
|
2883 switch (*regparse++)
|
|
2884 {
|
|
2885 case 'd': nr = getdecchrs(); break;
|
|
2886 case 'o': nr = getoctchrs(); break;
|
|
2887 case 'x': nr = gethexchrs(2); break;
|
|
2888 case 'u': nr = gethexchrs(4); break;
|
|
2889 case 'U': nr = gethexchrs(8); break;
|
|
2890 }
|
|
2891 if (nr < 0)
|
|
2892 {
|
|
2893 /* If getting the number fails be backwards compatible: the character
|
|
2894 * is a backslash. */
|
|
2895 --regparse;
|
|
2896 nr = '\\';
|
|
2897 }
|
|
2898 return nr;
|
|
2899 }
|
|
2900
|
|
2901 /*
|
7
|
2902 * read_limits - Read two integers to be taken as a minimum and maximum.
|
|
2903 * If the first character is '-', then the range is reversed.
|
|
2904 * Should end with 'end'. If minval is missing, zero is default, if maxval is
|
|
2905 * missing, a very big number is the default.
|
|
2906 */
|
|
2907 static int
|
|
2908 read_limits(minval, maxval)
|
|
2909 long *minval;
|
|
2910 long *maxval;
|
|
2911 {
|
|
2912 int reverse = FALSE;
|
|
2913 char_u *first_char;
|
|
2914 long tmp;
|
|
2915
|
|
2916 if (*regparse == '-')
|
|
2917 {
|
|
2918 /* Starts with '-', so reverse the range later */
|
|
2919 regparse++;
|
|
2920 reverse = TRUE;
|
|
2921 }
|
|
2922 first_char = regparse;
|
|
2923 *minval = getdigits(®parse);
|
|
2924 if (*regparse == ',') /* There is a comma */
|
|
2925 {
|
|
2926 if (vim_isdigit(*++regparse))
|
|
2927 *maxval = getdigits(®parse);
|
|
2928 else
|
|
2929 *maxval = MAX_LIMIT;
|
|
2930 }
|
|
2931 else if (VIM_ISDIGIT(*first_char))
|
|
2932 *maxval = *minval; /* It was \{n} or \{-n} */
|
|
2933 else
|
|
2934 *maxval = MAX_LIMIT; /* It was \{} or \{-} */
|
|
2935 if (*regparse == '\\')
|
|
2936 regparse++; /* Allow either \{...} or \{...\} */
|
167
|
2937 if (*regparse != '}')
|
7
|
2938 {
|
|
2939 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"),
|
|
2940 reg_magic == MAGIC_ALL ? "" : "\\");
|
|
2941 EMSG_RET_FAIL(IObuff);
|
|
2942 }
|
|
2943
|
|
2944 /*
|
|
2945 * Reverse the range if there was a '-', or make sure it is in the right
|
|
2946 * order otherwise.
|
|
2947 */
|
|
2948 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval))
|
|
2949 {
|
|
2950 tmp = *minval;
|
|
2951 *minval = *maxval;
|
|
2952 *maxval = tmp;
|
|
2953 }
|
|
2954 skipchr(); /* let's be friends with the lexer again */
|
|
2955 return OK;
|
|
2956 }
|
|
2957
|
|
2958 /*
|
|
2959 * vim_regexec and friends
|
|
2960 */
|
|
2961
|
|
2962 /*
|
|
2963 * Global work variables for vim_regexec().
|
|
2964 */
|
|
2965
|
|
2966 /* The current match-position is remembered with these variables: */
|
|
2967 static linenr_T reglnum; /* line number, relative to first line */
|
|
2968 static char_u *regline; /* start of current line */
|
|
2969 static char_u *reginput; /* current input, points into "regline" */
|
|
2970
|
|
2971 static int need_clear_subexpr; /* subexpressions still need to be
|
|
2972 * cleared */
|
|
2973 #ifdef FEAT_SYN_HL
|
|
2974 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions
|
|
2975 * still need to be cleared */
|
|
2976 #endif
|
|
2977
|
|
2978 /*
|
|
2979 * Structure used to save the current input state, when it needs to be
|
|
2980 * restored after trying a match. Used by reg_save() and reg_restore().
|
233
|
2981 * Also stores the length of "backpos".
|
7
|
2982 */
|
|
2983 typedef struct
|
|
2984 {
|
|
2985 union
|
|
2986 {
|
|
2987 char_u *ptr; /* reginput pointer, for single-line regexp */
|
|
2988 lpos_T pos; /* reginput pos, for multi-line regexp */
|
|
2989 } rs_u;
|
233
|
2990 int rs_len;
|
7
|
2991 } regsave_T;
|
|
2992
|
|
2993 /* struct to save start/end pointer/position in for \(\) */
|
|
2994 typedef struct
|
|
2995 {
|
|
2996 union
|
|
2997 {
|
|
2998 char_u *ptr;
|
|
2999 lpos_T pos;
|
|
3000 } se_u;
|
|
3001 } save_se_T;
|
|
3002
|
|
3003 static char_u *reg_getline __ARGS((linenr_T lnum));
|
|
3004 static long vim_regexec_both __ARGS((char_u *line, colnr_T col));
|
|
3005 static long regtry __ARGS((regprog_T *prog, colnr_T col));
|
|
3006 static void cleanup_subexpr __ARGS((void));
|
|
3007 #ifdef FEAT_SYN_HL
|
|
3008 static void cleanup_zsubexpr __ARGS((void));
|
|
3009 #endif
|
|
3010 static void reg_nextline __ARGS((void));
|
233
|
3011 static void reg_save __ARGS((regsave_T *save, garray_T *gap));
|
|
3012 static void reg_restore __ARGS((regsave_T *save, garray_T *gap));
|
7
|
3013 static int reg_save_equal __ARGS((regsave_T *save));
|
|
3014 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp));
|
|
3015 static void save_se_one __ARGS((save_se_T *savep, char_u **pp));
|
|
3016
|
|
3017 /* Save the sub-expressions before attempting a match. */
|
|
3018 #define save_se(savep, posp, pp) \
|
|
3019 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
|
|
3020
|
|
3021 /* After a failed match restore the sub-expressions. */
|
|
3022 #define restore_se(savep, posp, pp) { \
|
|
3023 if (REG_MULTI) \
|
|
3024 *(posp) = (savep)->se_u.pos; \
|
|
3025 else \
|
|
3026 *(pp) = (savep)->se_u.ptr; }
|
|
3027
|
|
3028 static int re_num_cmp __ARGS((long_u val, char_u *scan));
|
180
|
3029 static int regmatch __ARGS((char_u *prog));
|
7
|
3030 static int regrepeat __ARGS((char_u *p, long maxcount));
|
|
3031
|
|
3032 #ifdef DEBUG
|
|
3033 int regnarrate = 0;
|
|
3034 #endif
|
|
3035
|
|
3036 /*
|
|
3037 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec().
|
|
3038 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern
|
|
3039 * contains '\c' or '\C' the value is overruled.
|
|
3040 */
|
|
3041 static int ireg_ic;
|
|
3042
|
|
3043 #ifdef FEAT_MBYTE
|
|
3044 /*
|
|
3045 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag
|
|
3046 * in the regexp. Defaults to false, always.
|
|
3047 */
|
|
3048 static int ireg_icombine;
|
|
3049 #endif
|
|
3050
|
|
3051 /*
|
410
|
3052 * Copy of "rmm_maxcol": maximum column to search for a match. Zero when
|
|
3053 * there is no maximum.
|
|
3054 */
|
418
|
3055 static colnr_T ireg_maxcol;
|
410
|
3056
|
|
3057 /*
|
7
|
3058 * Sometimes need to save a copy of a line. Since alloc()/free() is very
|
|
3059 * slow, we keep one allocated piece of memory and only re-allocate it when
|
|
3060 * it's too small. It's freed in vim_regexec_both() when finished.
|
|
3061 */
|
|
3062 static char_u *reg_tofree;
|
|
3063 static unsigned reg_tofreelen;
|
|
3064
|
|
3065 /*
|
|
3066 * These variables are set when executing a regexp to speed up the execution.
|
|
3067 * Which ones are set depends on whethere a single-line or multi-line match is
|
|
3068 * done:
|
|
3069 * single-line multi-line
|
|
3070 * reg_match ®match_T NULL
|
|
3071 * reg_mmatch NULL ®mmatch_T
|
|
3072 * reg_startp reg_match->startp <invalid>
|
|
3073 * reg_endp reg_match->endp <invalid>
|
|
3074 * reg_startpos <invalid> reg_mmatch->startpos
|
|
3075 * reg_endpos <invalid> reg_mmatch->endpos
|
|
3076 * reg_win NULL window in which to search
|
|
3077 * reg_buf <invalid> buffer in which to search
|
|
3078 * reg_firstlnum <invalid> first line in which to search
|
|
3079 * reg_maxline 0 last line nr
|
|
3080 * reg_line_lbr FALSE or TRUE FALSE
|
|
3081 */
|
|
3082 static regmatch_T *reg_match;
|
|
3083 static regmmatch_T *reg_mmatch;
|
|
3084 static char_u **reg_startp = NULL;
|
|
3085 static char_u **reg_endp = NULL;
|
|
3086 static lpos_T *reg_startpos = NULL;
|
|
3087 static lpos_T *reg_endpos = NULL;
|
|
3088 static win_T *reg_win;
|
|
3089 static buf_T *reg_buf;
|
|
3090 static linenr_T reg_firstlnum;
|
|
3091 static linenr_T reg_maxline;
|
|
3092 static int reg_line_lbr; /* "\n" in string is line break */
|
|
3093
|
270
|
3094 /* Values for rs_state in regitem_T. */
|
|
3095 typedef enum regstate_E
|
|
3096 {
|
|
3097 RS_NOPEN = 0 /* NOPEN and NCLOSE */
|
|
3098 , RS_MOPEN /* MOPEN + [0-9] */
|
|
3099 , RS_MCLOSE /* MCLOSE + [0-9] */
|
|
3100 #ifdef FEAT_SYN_HL
|
|
3101 , RS_ZOPEN /* ZOPEN + [0-9] */
|
|
3102 , RS_ZCLOSE /* ZCLOSE + [0-9] */
|
|
3103 #endif
|
|
3104 , RS_BRANCH /* BRANCH */
|
|
3105 , RS_BRCPLX_MORE /* BRACE_COMPLEX and trying one more match */
|
|
3106 , RS_BRCPLX_LONG /* BRACE_COMPLEX and trying longest match */
|
|
3107 , RS_BRCPLX_SHORT /* BRACE_COMPLEX and trying shortest match */
|
|
3108 , RS_NOMATCH /* NOMATCH */
|
|
3109 , RS_BEHIND1 /* BEHIND / NOBEHIND matching rest */
|
|
3110 , RS_BEHIND2 /* BEHIND / NOBEHIND matching behind part */
|
|
3111 , RS_STAR_LONG /* STAR/PLUS/BRACE_SIMPLE longest match */
|
|
3112 , RS_STAR_SHORT /* STAR/PLUS/BRACE_SIMPLE shortest match */
|
|
3113 } regstate_T;
|
|
3114
|
|
3115 /*
|
|
3116 * When there are alternatives a regstate_T is put on the regstack to remember
|
|
3117 * what we are doing.
|
|
3118 * Before it may be another type of item, depending on rs_state, to remember
|
|
3119 * more things.
|
|
3120 */
|
|
3121 typedef struct regitem_S
|
|
3122 {
|
|
3123 regstate_T rs_state; /* what we are doing, one of RS_ above */
|
|
3124 char_u *rs_scan; /* current node in program */
|
|
3125 union
|
|
3126 {
|
|
3127 save_se_T sesave;
|
|
3128 regsave_T regsave;
|
|
3129 } rs_un; /* room for saving reginput */
|
|
3130 short rs_no; /* submatch nr */
|
|
3131 } regitem_T;
|
|
3132
|
|
3133 static regitem_T *regstack_push __ARGS((regstate_T state, char_u *scan));
|
|
3134 static void regstack_pop __ARGS((char_u **scan));
|
|
3135
|
|
3136 /* used for BEHIND and NOBEHIND matching */
|
|
3137 typedef struct regbehind_S
|
|
3138 {
|
|
3139 regsave_T save_after;
|
|
3140 regsave_T save_behind;
|
|
3141 } regbehind_T;
|
|
3142
|
|
3143 /* used for STAR, PLUS and BRACE_SIMPLE matching */
|
|
3144 typedef struct regstar_S
|
|
3145 {
|
|
3146 int nextb; /* next byte */
|
|
3147 int nextb_ic; /* next byte reverse case */
|
|
3148 long count;
|
|
3149 long minval;
|
|
3150 long maxval;
|
|
3151 } regstar_T;
|
|
3152
|
|
3153 /* used to store input position when a BACK was encountered, so that we now if
|
|
3154 * we made any progress since the last time. */
|
|
3155 typedef struct backpos_S
|
|
3156 {
|
|
3157 char_u *bp_scan; /* "scan" where BACK was encountered */
|
|
3158 regsave_T bp_pos; /* last input position */
|
|
3159 } backpos_T;
|
|
3160
|
|
3161 /*
|
|
3162 * regstack and backpos are used by regmatch(). They are kept over calls to
|
|
3163 * avoid invoking malloc() and free() often.
|
|
3164 */
|
|
3165 static garray_T regstack; /* stack with regitem_T items, sometimes
|
|
3166 preceded by regstar_T or regbehind_T. */
|
|
3167 static garray_T backpos; /* table with backpos_T for BACK */
|
|
3168
|
7
|
3169 /*
|
|
3170 * Get pointer to the line "lnum", which is relative to "reg_firstlnum".
|
|
3171 */
|
|
3172 static char_u *
|
|
3173 reg_getline(lnum)
|
|
3174 linenr_T lnum;
|
|
3175 {
|
|
3176 /* when looking behind for a match/no-match lnum is negative. But we
|
|
3177 * can't go before line 1 */
|
|
3178 if (reg_firstlnum + lnum < 1)
|
|
3179 return NULL;
|
482
|
3180 if (lnum > reg_maxline)
|
481
|
3181 /* Must have matched the "\n" in the last line. */
|
|
3182 return (char_u *)"";
|
7
|
3183 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE);
|
|
3184 }
|
|
3185
|
|
3186 static regsave_T behind_pos;
|
|
3187
|
|
3188 #ifdef FEAT_SYN_HL
|
|
3189 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */
|
|
3190 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */
|
|
3191 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */
|
|
3192 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */
|
|
3193 #endif
|
|
3194
|
|
3195 /* TRUE if using multi-line regexp. */
|
|
3196 #define REG_MULTI (reg_match == NULL)
|
|
3197
|
|
3198 /*
|
|
3199 * Match a regexp against a string.
|
|
3200 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
3201 * Uses curbuf for line count and 'iskeyword'.
|
|
3202 *
|
|
3203 * Return TRUE if there is a match, FALSE if not.
|
|
3204 */
|
|
3205 int
|
|
3206 vim_regexec(rmp, line, col)
|
|
3207 regmatch_T *rmp;
|
|
3208 char_u *line; /* string to match against */
|
|
3209 colnr_T col; /* column to start looking for match */
|
|
3210 {
|
|
3211 reg_match = rmp;
|
|
3212 reg_mmatch = NULL;
|
|
3213 reg_maxline = 0;
|
|
3214 reg_line_lbr = FALSE;
|
|
3215 reg_win = NULL;
|
|
3216 ireg_ic = rmp->rm_ic;
|
|
3217 #ifdef FEAT_MBYTE
|
|
3218 ireg_icombine = FALSE;
|
|
3219 #endif
|
410
|
3220 ireg_maxcol = 0;
|
7
|
3221 return (vim_regexec_both(line, col) != 0);
|
|
3222 }
|
|
3223
|
39
|
3224 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \
|
|
3225 || defined(FIND_REPLACE_DIALOG) || defined(PROTO)
|
7
|
3226 /*
|
|
3227 * Like vim_regexec(), but consider a "\n" in "line" to be a line break.
|
|
3228 */
|
|
3229 int
|
|
3230 vim_regexec_nl(rmp, line, col)
|
|
3231 regmatch_T *rmp;
|
|
3232 char_u *line; /* string to match against */
|
|
3233 colnr_T col; /* column to start looking for match */
|
|
3234 {
|
|
3235 reg_match = rmp;
|
|
3236 reg_mmatch = NULL;
|
|
3237 reg_maxline = 0;
|
|
3238 reg_line_lbr = TRUE;
|
|
3239 reg_win = NULL;
|
|
3240 ireg_ic = rmp->rm_ic;
|
|
3241 #ifdef FEAT_MBYTE
|
|
3242 ireg_icombine = FALSE;
|
|
3243 #endif
|
410
|
3244 ireg_maxcol = 0;
|
7
|
3245 return (vim_regexec_both(line, col) != 0);
|
|
3246 }
|
|
3247 #endif
|
|
3248
|
|
3249 /*
|
|
3250 * Match a regexp against multiple lines.
|
|
3251 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
3252 * Uses curbuf for line count and 'iskeyword'.
|
|
3253 *
|
|
3254 * Return zero if there is no match. Return number of lines contained in the
|
|
3255 * match otherwise.
|
|
3256 */
|
|
3257 long
|
|
3258 vim_regexec_multi(rmp, win, buf, lnum, col)
|
|
3259 regmmatch_T *rmp;
|
|
3260 win_T *win; /* window in which to search or NULL */
|
|
3261 buf_T *buf; /* buffer in which to search */
|
|
3262 linenr_T lnum; /* nr of line to start looking for match */
|
|
3263 colnr_T col; /* column to start looking for match */
|
|
3264 {
|
|
3265 long r;
|
|
3266 buf_T *save_curbuf = curbuf;
|
|
3267
|
|
3268 reg_match = NULL;
|
|
3269 reg_mmatch = rmp;
|
|
3270 reg_buf = buf;
|
|
3271 reg_win = win;
|
|
3272 reg_firstlnum = lnum;
|
|
3273 reg_maxline = reg_buf->b_ml.ml_line_count - lnum;
|
|
3274 reg_line_lbr = FALSE;
|
|
3275 ireg_ic = rmp->rmm_ic;
|
|
3276 #ifdef FEAT_MBYTE
|
|
3277 ireg_icombine = FALSE;
|
|
3278 #endif
|
410
|
3279 ireg_maxcol = rmp->rmm_maxcol;
|
7
|
3280
|
|
3281 /* Need to switch to buffer "buf" to make vim_iswordc() work. */
|
|
3282 curbuf = buf;
|
|
3283 r = vim_regexec_both(NULL, col);
|
|
3284 curbuf = save_curbuf;
|
|
3285
|
|
3286 return r;
|
|
3287 }
|
|
3288
|
|
3289 /*
|
|
3290 * Match a regexp against a string ("line" points to the string) or multiple
|
|
3291 * lines ("line" is NULL, use reg_getline()).
|
|
3292 */
|
|
3293 static long
|
|
3294 vim_regexec_both(line, col)
|
|
3295 char_u *line;
|
|
3296 colnr_T col; /* column to start looking for match */
|
|
3297 {
|
|
3298 regprog_T *prog;
|
|
3299 char_u *s;
|
189
|
3300 long retval = 0L;
|
7
|
3301
|
|
3302 reg_tofree = NULL;
|
|
3303
|
270
|
3304 /* Init the regstack empty. Use an item size of 1 byte, since we push
|
|
3305 * different things onto it. Use a large grow size to avoid reallocating
|
|
3306 * it too often. */
|
|
3307 ga_init2(®stack, 1, 10000);
|
|
3308
|
|
3309 /* Init the backpos table empty. */
|
|
3310 ga_init2(&backpos, sizeof(backpos_T), 10);
|
|
3311
|
7
|
3312 if (REG_MULTI)
|
|
3313 {
|
|
3314 prog = reg_mmatch->regprog;
|
|
3315 line = reg_getline((linenr_T)0);
|
|
3316 reg_startpos = reg_mmatch->startpos;
|
|
3317 reg_endpos = reg_mmatch->endpos;
|
|
3318 }
|
|
3319 else
|
|
3320 {
|
|
3321 prog = reg_match->regprog;
|
|
3322 reg_startp = reg_match->startp;
|
|
3323 reg_endp = reg_match->endp;
|
|
3324 }
|
|
3325
|
|
3326 /* Be paranoid... */
|
|
3327 if (prog == NULL || line == NULL)
|
|
3328 {
|
|
3329 EMSG(_(e_null));
|
|
3330 goto theend;
|
|
3331 }
|
|
3332
|
|
3333 /* Check validity of program. */
|
|
3334 if (prog_magic_wrong())
|
|
3335 goto theend;
|
|
3336
|
410
|
3337 /* If the start column is past the maximum column: no need to try. */
|
|
3338 if (ireg_maxcol > 0 && col >= ireg_maxcol)
|
|
3339 goto theend;
|
|
3340
|
7
|
3341 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */
|
|
3342 if (prog->regflags & RF_ICASE)
|
|
3343 ireg_ic = TRUE;
|
|
3344 else if (prog->regflags & RF_NOICASE)
|
|
3345 ireg_ic = FALSE;
|
|
3346
|
|
3347 #ifdef FEAT_MBYTE
|
|
3348 /* If pattern contains "\Z" overrule value of ireg_icombine */
|
|
3349 if (prog->regflags & RF_ICOMBINE)
|
|
3350 ireg_icombine = TRUE;
|
|
3351 #endif
|
|
3352
|
|
3353 /* If there is a "must appear" string, look for it. */
|
|
3354 if (prog->regmust != NULL)
|
|
3355 {
|
|
3356 int c;
|
|
3357
|
|
3358 #ifdef FEAT_MBYTE
|
|
3359 if (has_mbyte)
|
|
3360 c = (*mb_ptr2char)(prog->regmust);
|
|
3361 else
|
|
3362 #endif
|
|
3363 c = *prog->regmust;
|
|
3364 s = line + col;
|
170
|
3365
|
|
3366 /*
|
|
3367 * This is used very often, esp. for ":global". Use three versions of
|
|
3368 * the loop to avoid overhead of conditions.
|
|
3369 */
|
|
3370 if (!ireg_ic
|
|
3371 #ifdef FEAT_MBYTE
|
|
3372 && !has_mbyte
|
|
3373 #endif
|
|
3374 )
|
|
3375 while ((s = vim_strbyte(s, c)) != NULL)
|
|
3376 {
|
|
3377 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3378 break; /* Found it. */
|
|
3379 ++s;
|
|
3380 }
|
|
3381 #ifdef FEAT_MBYTE
|
|
3382 else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1))
|
|
3383 while ((s = vim_strchr(s, c)) != NULL)
|
|
3384 {
|
|
3385 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3386 break; /* Found it. */
|
|
3387 mb_ptr_adv(s);
|
|
3388 }
|
|
3389 #endif
|
|
3390 else
|
|
3391 while ((s = cstrchr(s, c)) != NULL)
|
|
3392 {
|
|
3393 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3394 break; /* Found it. */
|
|
3395 mb_ptr_adv(s);
|
|
3396 }
|
7
|
3397 if (s == NULL) /* Not present. */
|
|
3398 goto theend;
|
|
3399 }
|
|
3400
|
|
3401 regline = line;
|
|
3402 reglnum = 0;
|
|
3403
|
|
3404 /* Simplest case: Anchored match need be tried only once. */
|
|
3405 if (prog->reganch)
|
|
3406 {
|
|
3407 int c;
|
|
3408
|
|
3409 #ifdef FEAT_MBYTE
|
|
3410 if (has_mbyte)
|
|
3411 c = (*mb_ptr2char)(regline + col);
|
|
3412 else
|
|
3413 #endif
|
|
3414 c = regline[col];
|
|
3415 if (prog->regstart == NUL
|
|
3416 || prog->regstart == c
|
|
3417 || (ireg_ic && ((
|
|
3418 #ifdef FEAT_MBYTE
|
|
3419 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
|
|
3420 || (c < 255 && prog->regstart < 255 &&
|
|
3421 #endif
|
|
3422 TOLOWER_LOC(prog->regstart) == TOLOWER_LOC(c)))))
|
|
3423 retval = regtry(prog, col);
|
|
3424 else
|
|
3425 retval = 0;
|
|
3426 }
|
|
3427 else
|
|
3428 {
|
|
3429 /* Messy cases: unanchored match. */
|
180
|
3430 while (!got_int)
|
7
|
3431 {
|
|
3432 if (prog->regstart != NUL)
|
|
3433 {
|
170
|
3434 /* Skip until the char we know it must start with.
|
|
3435 * Used often, do some work to avoid call overhead. */
|
|
3436 if (!ireg_ic
|
|
3437 #ifdef FEAT_MBYTE
|
|
3438 && !has_mbyte
|
|
3439 #endif
|
|
3440 )
|
|
3441 s = vim_strbyte(regline + col, prog->regstart);
|
|
3442 else
|
|
3443 s = cstrchr(regline + col, prog->regstart);
|
7
|
3444 if (s == NULL)
|
|
3445 {
|
|
3446 retval = 0;
|
|
3447 break;
|
|
3448 }
|
|
3449 col = (int)(s - regline);
|
|
3450 }
|
|
3451
|
410
|
3452 /* Check for maximum column to try. */
|
|
3453 if (ireg_maxcol > 0 && col >= ireg_maxcol)
|
|
3454 {
|
|
3455 retval = 0;
|
|
3456 break;
|
|
3457 }
|
|
3458
|
7
|
3459 retval = regtry(prog, col);
|
|
3460 if (retval > 0)
|
|
3461 break;
|
|
3462
|
|
3463 /* if not currently on the first line, get it again */
|
|
3464 if (reglnum != 0)
|
|
3465 {
|
481
|
3466 reglnum = 0;
|
7
|
3467 regline = reg_getline((linenr_T)0);
|
|
3468 }
|
|
3469 if (regline[col] == NUL)
|
|
3470 break;
|
|
3471 #ifdef FEAT_MBYTE
|
|
3472 if (has_mbyte)
|
474
|
3473 col += (*mb_ptr2len)(regline + col);
|
7
|
3474 else
|
|
3475 #endif
|
|
3476 ++col;
|
|
3477 }
|
|
3478 }
|
|
3479
|
|
3480 theend:
|
|
3481 vim_free(reg_tofree);
|
270
|
3482 ga_clear(®stack);
|
|
3483 ga_clear(&backpos);
|
|
3484
|
7
|
3485 return retval;
|
|
3486 }
|
|
3487
|
|
3488 #ifdef FEAT_SYN_HL
|
|
3489 static reg_extmatch_T *make_extmatch __ARGS((void));
|
|
3490
|
|
3491 /*
|
|
3492 * Create a new extmatch and mark it as referenced once.
|
|
3493 */
|
|
3494 static reg_extmatch_T *
|
|
3495 make_extmatch()
|
|
3496 {
|
|
3497 reg_extmatch_T *em;
|
|
3498
|
|
3499 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T));
|
|
3500 if (em != NULL)
|
|
3501 em->refcnt = 1;
|
|
3502 return em;
|
|
3503 }
|
|
3504
|
|
3505 /*
|
|
3506 * Add a reference to an extmatch.
|
|
3507 */
|
|
3508 reg_extmatch_T *
|
|
3509 ref_extmatch(em)
|
|
3510 reg_extmatch_T *em;
|
|
3511 {
|
|
3512 if (em != NULL)
|
|
3513 em->refcnt++;
|
|
3514 return em;
|
|
3515 }
|
|
3516
|
|
3517 /*
|
|
3518 * Remove a reference to an extmatch. If there are no references left, free
|
|
3519 * the info.
|
|
3520 */
|
|
3521 void
|
|
3522 unref_extmatch(em)
|
|
3523 reg_extmatch_T *em;
|
|
3524 {
|
|
3525 int i;
|
|
3526
|
|
3527 if (em != NULL && --em->refcnt <= 0)
|
|
3528 {
|
|
3529 for (i = 0; i < NSUBEXP; ++i)
|
|
3530 vim_free(em->matches[i]);
|
|
3531 vim_free(em);
|
|
3532 }
|
|
3533 }
|
|
3534 #endif
|
|
3535
|
|
3536 /*
|
|
3537 * regtry - try match of "prog" with at regline["col"].
|
|
3538 * Returns 0 for failure, number of lines contained in the match otherwise.
|
|
3539 */
|
|
3540 static long
|
|
3541 regtry(prog, col)
|
|
3542 regprog_T *prog;
|
|
3543 colnr_T col;
|
|
3544 {
|
|
3545 reginput = regline + col;
|
|
3546 need_clear_subexpr = TRUE;
|
|
3547 #ifdef FEAT_SYN_HL
|
|
3548 /* Clear the external match subpointers if necessary. */
|
|
3549 if (prog->reghasz == REX_SET)
|
|
3550 need_clear_zsubexpr = TRUE;
|
|
3551 #endif
|
|
3552
|
189
|
3553 if (regmatch(prog->program + 1) == 0)
|
|
3554 return 0;
|
|
3555
|
|
3556 cleanup_subexpr();
|
|
3557 if (REG_MULTI)
|
7
|
3558 {
|
189
|
3559 if (reg_startpos[0].lnum < 0)
|
7
|
3560 {
|
189
|
3561 reg_startpos[0].lnum = 0;
|
|
3562 reg_startpos[0].col = col;
|
|
3563 }
|
|
3564 if (reg_endpos[0].lnum < 0)
|
|
3565 {
|
|
3566 reg_endpos[0].lnum = reglnum;
|
|
3567 reg_endpos[0].col = (int)(reginput - regline);
|
7
|
3568 }
|
|
3569 else
|
189
|
3570 /* Use line number of "\ze". */
|
|
3571 reglnum = reg_endpos[0].lnum;
|
|
3572 }
|
|
3573 else
|
|
3574 {
|
|
3575 if (reg_startp[0] == NULL)
|
|
3576 reg_startp[0] = regline + col;
|
|
3577 if (reg_endp[0] == NULL)
|
|
3578 reg_endp[0] = reginput;
|
|
3579 }
|
7
|
3580 #ifdef FEAT_SYN_HL
|
189
|
3581 /* Package any found \z(...\) matches for export. Default is none. */
|
|
3582 unref_extmatch(re_extmatch_out);
|
|
3583 re_extmatch_out = NULL;
|
|
3584
|
|
3585 if (prog->reghasz == REX_SET)
|
|
3586 {
|
|
3587 int i;
|
|
3588
|
|
3589 cleanup_zsubexpr();
|
|
3590 re_extmatch_out = make_extmatch();
|
|
3591 for (i = 0; i < NSUBEXP; i++)
|
7
|
3592 {
|
189
|
3593 if (REG_MULTI)
|
7
|
3594 {
|
189
|
3595 /* Only accept single line matches. */
|
|
3596 if (reg_startzpos[i].lnum >= 0
|
|
3597 && reg_endzpos[i].lnum == reg_startzpos[i].lnum)
|
|
3598 re_extmatch_out->matches[i] =
|
|
3599 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
|
7
|
3600 + reg_startzpos[i].col,
|
189
|
3601 reg_endzpos[i].col - reg_startzpos[i].col);
|
|
3602 }
|
|
3603 else
|
|
3604 {
|
|
3605 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
|
|
3606 re_extmatch_out->matches[i] =
|
7
|
3607 vim_strnsave(reg_startzp[i],
|
189
|
3608 (int)(reg_endzp[i] - reg_startzp[i]));
|
7
|
3609 }
|
|
3610 }
|
189
|
3611 }
|
7
|
3612 #endif
|
189
|
3613 return 1 + reglnum;
|
7
|
3614 }
|
|
3615
|
|
3616 #ifdef FEAT_MBYTE
|
|
3617 static int reg_prev_class __ARGS((void));
|
|
3618
|
|
3619 /*
|
|
3620 * Get class of previous character.
|
|
3621 */
|
|
3622 static int
|
|
3623 reg_prev_class()
|
|
3624 {
|
|
3625 if (reginput > regline)
|
|
3626 return mb_get_class(reginput - 1
|
|
3627 - (*mb_head_off)(regline, reginput - 1));
|
|
3628 return -1;
|
|
3629 }
|
|
3630
|
|
3631 #endif
|
39
|
3632 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput)
|
7
|
3633
|
|
3634 /*
|
|
3635 * The arguments from BRACE_LIMITS are stored here. They are actually local
|
|
3636 * to regmatch(), but they are here to reduce the amount of stack space used
|
|
3637 * (it can be called recursively many times).
|
|
3638 */
|
|
3639 static long bl_minval;
|
|
3640 static long bl_maxval;
|
|
3641
|
|
3642 /*
|
|
3643 * regmatch - main matching routine
|
|
3644 *
|
180
|
3645 * Conceptually the strategy is simple: Check to see whether the current node
|
|
3646 * matches, push an item onto the regstack and loop to see whether the rest
|
|
3647 * matches, and then act accordingly. In practice we make some effort to
|
|
3648 * avoid using the regstack, in particular by going through "ordinary" nodes
|
|
3649 * (that don't need to know whether the rest of the match failed) by a nested
|
|
3650 * loop.
|
7
|
3651 *
|
|
3652 * Returns TRUE when there is a match. Leaves reginput and reglnum just after
|
|
3653 * the last matched character.
|
|
3654 * Returns FALSE when there is no match. Leaves reginput and reglnum in an
|
|
3655 * undefined state!
|
|
3656 */
|
|
3657 static int
|
180
|
3658 regmatch(scan)
|
7
|
3659 char_u *scan; /* Current node. */
|
|
3660 {
|
180
|
3661 char_u *next; /* Next node. */
|
|
3662 int op;
|
|
3663 int c;
|
|
3664 regitem_T *rp;
|
|
3665 int no;
|
|
3666 int status; /* one of the RA_ values: */
|
|
3667 #define RA_FAIL 1 /* something failed, abort */
|
|
3668 #define RA_CONT 2 /* continue in inner loop */
|
|
3669 #define RA_BREAK 3 /* break inner loop */
|
|
3670 #define RA_MATCH 4 /* successful match */
|
|
3671 #define RA_NOMATCH 5 /* didn't match */
|
270
|
3672
|
|
3673 /* Init the regstack and backpos table empty. They are initialized and
|
|
3674 * freed in vim_regexec_both() to reduce malloc()/free() calls. */
|
|
3675 regstack.ga_len = 0;
|
|
3676 backpos.ga_len = 0;
|
233
|
3677
|
180
|
3678 /*
|
233
|
3679 * Repeat until "regstack" is empty.
|
180
|
3680 */
|
|
3681 for (;;)
|
|
3682 {
|
7
|
3683 /* Some patterns my cause a long time to match, even though they are not
|
|
3684 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */
|
|
3685 fast_breakcheck();
|
|
3686
|
|
3687 #ifdef DEBUG
|
|
3688 if (scan != NULL && regnarrate)
|
|
3689 {
|
|
3690 mch_errmsg(regprop(scan));
|
|
3691 mch_errmsg("(\n");
|
|
3692 }
|
|
3693 #endif
|
180
|
3694
|
|
3695 /*
|
233
|
3696 * Repeat for items that can be matched sequentially, without using the
|
180
|
3697 * regstack.
|
|
3698 */
|
|
3699 for (;;)
|
7
|
3700 {
|
180
|
3701 if (got_int || scan == NULL)
|
|
3702 {
|
|
3703 status = RA_FAIL;
|
|
3704 break;
|
|
3705 }
|
|
3706 status = RA_CONT;
|
|
3707
|
7
|
3708 #ifdef DEBUG
|
|
3709 if (regnarrate)
|
|
3710 {
|
|
3711 mch_errmsg(regprop(scan));
|
|
3712 mch_errmsg("...\n");
|
|
3713 # ifdef FEAT_SYN_HL
|
|
3714 if (re_extmatch_in != NULL)
|
|
3715 {
|
|
3716 int i;
|
|
3717
|
|
3718 mch_errmsg(_("External submatches:\n"));
|
|
3719 for (i = 0; i < NSUBEXP; i++)
|
|
3720 {
|
|
3721 mch_errmsg(" \"");
|
|
3722 if (re_extmatch_in->matches[i] != NULL)
|
|
3723 mch_errmsg(re_extmatch_in->matches[i]);
|
|
3724 mch_errmsg("\"\n");
|
|
3725 }
|
|
3726 }
|
|
3727 # endif
|
|
3728 }
|
|
3729 #endif
|
|
3730 next = regnext(scan);
|
|
3731
|
|
3732 op = OP(scan);
|
|
3733 /* Check for character class with NL added. */
|
481
|
3734 if (!reg_line_lbr && WITH_NL(op) && *reginput == NUL
|
|
3735 && reglnum <= reg_maxline)
|
7
|
3736 {
|
|
3737 reg_nextline();
|
|
3738 }
|
|
3739 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n')
|
|
3740 {
|
|
3741 ADVANCE_REGINPUT();
|
|
3742 }
|
|
3743 else
|
|
3744 {
|
|
3745 if (WITH_NL(op))
|
180
|
3746 op -= ADD_NL;
|
7
|
3747 #ifdef FEAT_MBYTE
|
|
3748 if (has_mbyte)
|
|
3749 c = (*mb_ptr2char)(reginput);
|
|
3750 else
|
|
3751 #endif
|
|
3752 c = *reginput;
|
|
3753 switch (op)
|
|
3754 {
|
|
3755 case BOL:
|
|
3756 if (reginput != regline)
|
180
|
3757 status = RA_NOMATCH;
|
7
|
3758 break;
|
|
3759
|
|
3760 case EOL:
|
|
3761 if (c != NUL)
|
180
|
3762 status = RA_NOMATCH;
|
7
|
3763 break;
|
|
3764
|
|
3765 case RE_BOF:
|
|
3766 /* Passing -1 to the getline() function provided for the search
|
|
3767 * should always return NULL if the current line is the first
|
|
3768 * line of the file. */
|
|
3769 if (reglnum != 0 || reginput != regline
|
|
3770 || (REG_MULTI && reg_getline((linenr_T)-1) != NULL))
|
180
|
3771 status = RA_NOMATCH;
|
7
|
3772 break;
|
|
3773
|
|
3774 case RE_EOF:
|
|
3775 if (reglnum != reg_maxline || c != NUL)
|
180
|
3776 status = RA_NOMATCH;
|
7
|
3777 break;
|
|
3778
|
|
3779 case CURSOR:
|
|
3780 /* Check if the buffer is in a window and compare the
|
|
3781 * reg_win->w_cursor position to the match position. */
|
|
3782 if (reg_win == NULL
|
|
3783 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum)
|
|
3784 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col))
|
180
|
3785 status = RA_NOMATCH;
|
7
|
3786 break;
|
|
3787
|
|
3788 case RE_LNUM:
|
|
3789 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum),
|
|
3790 scan))
|
180
|
3791 status = RA_NOMATCH;
|
7
|
3792 break;
|
|
3793
|
|
3794 case RE_COL:
|
|
3795 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan))
|
180
|
3796 status = RA_NOMATCH;
|
7
|
3797 break;
|
|
3798
|
|
3799 case RE_VCOL:
|
|
3800 if (!re_num_cmp((long_u)win_linetabsize(
|
|
3801 reg_win == NULL ? curwin : reg_win,
|
|
3802 regline, (colnr_T)(reginput - regline)) + 1, scan))
|
180
|
3803 status = RA_NOMATCH;
|
7
|
3804 break;
|
|
3805
|
|
3806 case BOW: /* \<word; reginput points to w */
|
|
3807 if (c == NUL) /* Can't match at end of line */
|
180
|
3808 status = RA_NOMATCH;
|
7
|
3809 #ifdef FEAT_MBYTE
|
180
|
3810 else if (has_mbyte)
|
7
|
3811 {
|
|
3812 int this_class;
|
|
3813
|
|
3814 /* Get class of current and previous char (if it exists). */
|
|
3815 this_class = mb_get_class(reginput);
|
|
3816 if (this_class <= 1)
|
180
|
3817 status = RA_NOMATCH; /* not on a word at all */
|
|
3818 else if (reg_prev_class() == this_class)
|
|
3819 status = RA_NOMATCH; /* previous char is in same word */
|
7
|
3820 }
|
|
3821 #endif
|
|
3822 else
|
|
3823 {
|
|
3824 if (!vim_iswordc(c)
|
|
3825 || (reginput > regline && vim_iswordc(reginput[-1])))
|
180
|
3826 status = RA_NOMATCH;
|
7
|
3827 }
|
|
3828 break;
|
|
3829
|
|
3830 case EOW: /* word\>; reginput points after d */
|
|
3831 if (reginput == regline) /* Can't match at start of line */
|
180
|
3832 status = RA_NOMATCH;
|
7
|
3833 #ifdef FEAT_MBYTE
|
180
|
3834 else if (has_mbyte)
|
7
|
3835 {
|
|
3836 int this_class, prev_class;
|
|
3837
|
|
3838 /* Get class of current and previous char (if it exists). */
|
|
3839 this_class = mb_get_class(reginput);
|
|
3840 prev_class = reg_prev_class();
|
180
|
3841 if (this_class == prev_class
|
|
3842 || prev_class == 0 || prev_class == 1)
|
|
3843 status = RA_NOMATCH;
|
7
|
3844 }
|
180
|
3845 #endif
|
7
|
3846 else
|
|
3847 {
|
180
|
3848 if (!vim_iswordc(reginput[-1])
|
|
3849 || (reginput[0] != NUL && vim_iswordc(c)))
|
|
3850 status = RA_NOMATCH;
|
7
|
3851 }
|
|
3852 break; /* Matched with EOW */
|
|
3853
|
|
3854 case ANY:
|
|
3855 if (c == NUL)
|
180
|
3856 status = RA_NOMATCH;
|
|
3857 else
|
|
3858 ADVANCE_REGINPUT();
|
7
|
3859 break;
|
|
3860
|
|
3861 case IDENT:
|
|
3862 if (!vim_isIDc(c))
|
180
|
3863 status = RA_NOMATCH;
|
|
3864 else
|
|
3865 ADVANCE_REGINPUT();
|
7
|
3866 break;
|
|
3867
|
|
3868 case SIDENT:
|
|
3869 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c))
|
180
|
3870 status = RA_NOMATCH;
|
|
3871 else
|
|
3872 ADVANCE_REGINPUT();
|
7
|
3873 break;
|
|
3874
|
|
3875 case KWORD:
|
|
3876 if (!vim_iswordp(reginput))
|
180
|
3877 status = RA_NOMATCH;
|
|
3878 else
|
|
3879 ADVANCE_REGINPUT();
|
7
|
3880 break;
|
|
3881
|
|
3882 case SKWORD:
|
|
3883 if (VIM_ISDIGIT(*reginput) || !vim_iswordp(reginput))
|
180
|
3884 status = RA_NOMATCH;
|
|
3885 else
|
|
3886 ADVANCE_REGINPUT();
|
7
|
3887 break;
|
|
3888
|
|
3889 case FNAME:
|
|
3890 if (!vim_isfilec(c))
|
180
|
3891 status = RA_NOMATCH;
|
|
3892 else
|
|
3893 ADVANCE_REGINPUT();
|
7
|
3894 break;
|
|
3895
|
|
3896 case SFNAME:
|
|
3897 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c))
|
180
|
3898 status = RA_NOMATCH;
|
|
3899 else
|
|
3900 ADVANCE_REGINPUT();
|
7
|
3901 break;
|
|
3902
|
|
3903 case PRINT:
|
|
3904 if (ptr2cells(reginput) != 1)
|
180
|
3905 status = RA_NOMATCH;
|
|
3906 else
|
|
3907 ADVANCE_REGINPUT();
|
7
|
3908 break;
|
|
3909
|
|
3910 case SPRINT:
|
|
3911 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1)
|
180
|
3912 status = RA_NOMATCH;
|
|
3913 else
|
|
3914 ADVANCE_REGINPUT();
|
7
|
3915 break;
|
|
3916
|
|
3917 case WHITE:
|
|
3918 if (!vim_iswhite(c))
|
180
|
3919 status = RA_NOMATCH;
|
|
3920 else
|
|
3921 ADVANCE_REGINPUT();
|
7
|
3922 break;
|
|
3923
|
|
3924 case NWHITE:
|
|
3925 if (c == NUL || vim_iswhite(c))
|
180
|
3926 status = RA_NOMATCH;
|
|
3927 else
|
|
3928 ADVANCE_REGINPUT();
|
7
|
3929 break;
|
|
3930
|
|
3931 case DIGIT:
|
|
3932 if (!ri_digit(c))
|
180
|
3933 status = RA_NOMATCH;
|
|
3934 else
|
|
3935 ADVANCE_REGINPUT();
|
7
|
3936 break;
|
|
3937
|
|
3938 case NDIGIT:
|
|
3939 if (c == NUL || ri_digit(c))
|
180
|
3940 status = RA_NOMATCH;
|
|
3941 else
|
|
3942 ADVANCE_REGINPUT();
|
7
|
3943 break;
|
|
3944
|
|
3945 case HEX:
|
|
3946 if (!ri_hex(c))
|
180
|
3947 status = RA_NOMATCH;
|
|
3948 else
|
|
3949 ADVANCE_REGINPUT();
|
7
|
3950 break;
|
|
3951
|
|
3952 case NHEX:
|
|
3953 if (c == NUL || ri_hex(c))
|
180
|
3954 status = RA_NOMATCH;
|
|
3955 else
|
|
3956 ADVANCE_REGINPUT();
|
7
|
3957 break;
|
|
3958
|
|
3959 case OCTAL:
|
|
3960 if (!ri_octal(c))
|
180
|
3961 status = RA_NOMATCH;
|
|
3962 else
|
|
3963 ADVANCE_REGINPUT();
|
7
|
3964 break;
|
|
3965
|
|
3966 case NOCTAL:
|
|
3967 if (c == NUL || ri_octal(c))
|
180
|
3968 status = RA_NOMATCH;
|
|
3969 else
|
|
3970 ADVANCE_REGINPUT();
|
7
|
3971 break;
|
|
3972
|
|
3973 case WORD:
|
|
3974 if (!ri_word(c))
|
180
|
3975 status = RA_NOMATCH;
|
|
3976 else
|
|
3977 ADVANCE_REGINPUT();
|
7
|
3978 break;
|
|
3979
|
|
3980 case NWORD:
|
|
3981 if (c == NUL || ri_word(c))
|
180
|
3982 status = RA_NOMATCH;
|
|
3983 else
|
|
3984 ADVANCE_REGINPUT();
|
7
|
3985 break;
|
|
3986
|
|
3987 case HEAD:
|
|
3988 if (!ri_head(c))
|
180
|
3989 status = RA_NOMATCH;
|
|
3990 else
|
|
3991 ADVANCE_REGINPUT();
|
7
|
3992 break;
|
|
3993
|
|
3994 case NHEAD:
|
|
3995 if (c == NUL || ri_head(c))
|
180
|
3996 status = RA_NOMATCH;
|
|
3997 else
|
|
3998 ADVANCE_REGINPUT();
|
7
|
3999 break;
|
|
4000
|
|
4001 case ALPHA:
|
|
4002 if (!ri_alpha(c))
|
180
|
4003 status = RA_NOMATCH;
|
|
4004 else
|
|
4005 ADVANCE_REGINPUT();
|
7
|
4006 break;
|
|
4007
|
|
4008 case NALPHA:
|
|
4009 if (c == NUL || ri_alpha(c))
|
180
|
4010 status = RA_NOMATCH;
|
|
4011 else
|
|
4012 ADVANCE_REGINPUT();
|
7
|
4013 break;
|
|
4014
|
|
4015 case LOWER:
|
|
4016 if (!ri_lower(c))
|
180
|
4017 status = RA_NOMATCH;
|
|
4018 else
|
|
4019 ADVANCE_REGINPUT();
|
7
|
4020 break;
|
|
4021
|
|
4022 case NLOWER:
|
|
4023 if (c == NUL || ri_lower(c))
|
180
|
4024 status = RA_NOMATCH;
|
|
4025 else
|
|
4026 ADVANCE_REGINPUT();
|
7
|
4027 break;
|
|
4028
|
|
4029 case UPPER:
|
|
4030 if (!ri_upper(c))
|
180
|
4031 status = RA_NOMATCH;
|
|
4032 else
|
|
4033 ADVANCE_REGINPUT();
|
7
|
4034 break;
|
|
4035
|
|
4036 case NUPPER:
|
|
4037 if (c == NUL || ri_upper(c))
|
180
|
4038 status = RA_NOMATCH;
|
|
4039 else
|
|
4040 ADVANCE_REGINPUT();
|
7
|
4041 break;
|
|
4042
|
|
4043 case EXACTLY:
|
|
4044 {
|
|
4045 int len;
|
|
4046 char_u *opnd;
|
|
4047
|
|
4048 opnd = OPERAND(scan);
|
|
4049 /* Inline the first byte, for speed. */
|
|
4050 if (*opnd != *reginput
|
|
4051 && (!ireg_ic || (
|
|
4052 #ifdef FEAT_MBYTE
|
|
4053 !enc_utf8 &&
|
|
4054 #endif
|
|
4055 TOLOWER_LOC(*opnd) != TOLOWER_LOC(*reginput))))
|
180
|
4056 status = RA_NOMATCH;
|
|
4057 else if (*opnd == NUL)
|
7
|
4058 {
|
|
4059 /* match empty string always works; happens when "~" is
|
|
4060 * empty. */
|
|
4061 }
|
|
4062 else if (opnd[1] == NUL
|
|
4063 #ifdef FEAT_MBYTE
|
|
4064 && !(enc_utf8 && ireg_ic)
|
|
4065 #endif
|
|
4066 )
|
|
4067 ++reginput; /* matched a single char */
|
|
4068 else
|
|
4069 {
|
|
4070 len = (int)STRLEN(opnd);
|
|
4071 /* Need to match first byte again for multi-byte. */
|
|
4072 if (cstrncmp(opnd, reginput, &len) != 0)
|
180
|
4073 status = RA_NOMATCH;
|
7
|
4074 #ifdef FEAT_MBYTE
|
|
4075 /* Check for following composing character. */
|
180
|
4076 else if (enc_utf8
|
|
4077 && UTF_COMPOSINGLIKE(reginput, reginput + len))
|
7
|
4078 {
|
|
4079 /* raaron: This code makes a composing character get
|
|
4080 * ignored, which is the correct behavior (sometimes)
|
|
4081 * for voweled Hebrew texts. */
|
|
4082 if (!ireg_icombine)
|
180
|
4083 status = RA_NOMATCH;
|
7
|
4084 }
|
180
|
4085 #endif
|
7
|
4086 else
|
180
|
4087 reginput += len;
|
7
|
4088 }
|
|
4089 }
|
|
4090 break;
|
|
4091
|
|
4092 case ANYOF:
|
|
4093 case ANYBUT:
|
|
4094 if (c == NUL)
|
180
|
4095 status = RA_NOMATCH;
|
|
4096 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF))
|
|
4097 status = RA_NOMATCH;
|
|
4098 else
|
|
4099 ADVANCE_REGINPUT();
|
7
|
4100 break;
|
|
4101
|
|
4102 #ifdef FEAT_MBYTE
|
|
4103 case MULTIBYTECODE:
|
|
4104 if (has_mbyte)
|
|
4105 {
|
|
4106 int i, len;
|
|
4107 char_u *opnd;
|
|
4108
|
|
4109 opnd = OPERAND(scan);
|
|
4110 /* Safety check (just in case 'encoding' was changed since
|
|
4111 * compiling the program). */
|
474
|
4112 if ((len = (*mb_ptr2len)(opnd)) < 2)
|
180
|
4113 {
|
|
4114 status = RA_NOMATCH;
|
|
4115 break;
|
|
4116 }
|
7
|
4117 for (i = 0; i < len; ++i)
|
|
4118 if (opnd[i] != reginput[i])
|
180
|
4119 {
|
|
4120 status = RA_NOMATCH;
|
|
4121 break;
|
|
4122 }
|
7
|
4123 reginput += len;
|
|
4124 }
|
|
4125 else
|
180
|
4126 status = RA_NOMATCH;
|
7
|
4127 break;
|
|
4128 #endif
|
|
4129
|
|
4130 case NOTHING:
|
|
4131 break;
|
|
4132
|
|
4133 case BACK:
|
233
|
4134 {
|
|
4135 int i;
|
|
4136 backpos_T *bp;
|
|
4137
|
|
4138 /*
|
|
4139 * When we run into BACK we need to check if we don't keep
|
|
4140 * looping without matching any input. The second and later
|
|
4141 * times a BACK is encountered it fails if the input is still
|
|
4142 * at the same position as the previous time.
|
|
4143 * The positions are stored in "backpos" and found by the
|
|
4144 * current value of "scan", the position in the RE program.
|
|
4145 */
|
|
4146 bp = (backpos_T *)backpos.ga_data;
|
|
4147 for (i = 0; i < backpos.ga_len; ++i)
|
|
4148 if (bp[i].bp_scan == scan)
|
|
4149 break;
|
|
4150 if (i == backpos.ga_len)
|
|
4151 {
|
|
4152 /* First time at this BACK, make room to store the pos. */
|
|
4153 if (ga_grow(&backpos, 1) == FAIL)
|
|
4154 status = RA_FAIL;
|
|
4155 else
|
|
4156 {
|
|
4157 /* get "ga_data" again, it may have changed */
|
|
4158 bp = (backpos_T *)backpos.ga_data;
|
|
4159 bp[i].bp_scan = scan;
|
|
4160 ++backpos.ga_len;
|
|
4161 }
|
|
4162 }
|
|
4163 else if (reg_save_equal(&bp[i].bp_pos))
|
|
4164 /* Still at same position as last time, fail. */
|
|
4165 status = RA_NOMATCH;
|
|
4166
|
|
4167 if (status != RA_FAIL && status != RA_NOMATCH)
|
|
4168 reg_save(&bp[i].bp_pos, &backpos);
|
|
4169 }
|
179
|
4170 break;
|
|
4171
|
7
|
4172 case MOPEN + 0: /* Match start: \zs */
|
|
4173 case MOPEN + 1: /* \( */
|
|
4174 case MOPEN + 2:
|
|
4175 case MOPEN + 3:
|
|
4176 case MOPEN + 4:
|
|
4177 case MOPEN + 5:
|
|
4178 case MOPEN + 6:
|
|
4179 case MOPEN + 7:
|
|
4180 case MOPEN + 8:
|
|
4181 case MOPEN + 9:
|
|
4182 {
|
|
4183 no = op - MOPEN;
|
|
4184 cleanup_subexpr();
|
270
|
4185 rp = regstack_push(RS_MOPEN, scan);
|
180
|
4186 if (rp == NULL)
|
|
4187 status = RA_FAIL;
|
|
4188 else
|
|
4189 {
|
|
4190 rp->rs_no = no;
|
|
4191 save_se(&rp->rs_un.sesave, ®_startpos[no],
|
|
4192 ®_startp[no]);
|
|
4193 /* We simply continue and handle the result when done. */
|
|
4194 }
|
7
|
4195 }
|
180
|
4196 break;
|
7
|
4197
|
|
4198 case NOPEN: /* \%( */
|
|
4199 case NCLOSE: /* \) after \%( */
|
270
|
4200 if (regstack_push(RS_NOPEN, scan) == NULL)
|
180
|
4201 status = RA_FAIL;
|
|
4202 /* We simply continue and handle the result when done. */
|
|
4203 break;
|
7
|
4204
|
|
4205 #ifdef FEAT_SYN_HL
|
|
4206 case ZOPEN + 1:
|
|
4207 case ZOPEN + 2:
|
|
4208 case ZOPEN + 3:
|
|
4209 case ZOPEN + 4:
|
|
4210 case ZOPEN + 5:
|
|
4211 case ZOPEN + 6:
|
|
4212 case ZOPEN + 7:
|
|
4213 case ZOPEN + 8:
|
|
4214 case ZOPEN + 9:
|
|
4215 {
|
|
4216 no = op - ZOPEN;
|
|
4217 cleanup_zsubexpr();
|
270
|
4218 rp = regstack_push(RS_ZOPEN, scan);
|
180
|
4219 if (rp == NULL)
|
|
4220 status = RA_FAIL;
|
|
4221 else
|
|
4222 {
|
|
4223 rp->rs_no = no;
|
|
4224 save_se(&rp->rs_un.sesave, ®_startzpos[no],
|
|
4225 ®_startzp[no]);
|
|
4226 /* We simply continue and handle the result when done. */
|
|
4227 }
|
7
|
4228 }
|
180
|
4229 break;
|
7
|
4230 #endif
|
|
4231
|
|
4232 case MCLOSE + 0: /* Match end: \ze */
|
|
4233 case MCLOSE + 1: /* \) */
|
|
4234 case MCLOSE + 2:
|
|
4235 case MCLOSE + 3:
|
|
4236 case MCLOSE + 4:
|
|
4237 case MCLOSE + 5:
|
|
4238 case MCLOSE + 6:
|
|
4239 case MCLOSE + 7:
|
|
4240 case MCLOSE + 8:
|
|
4241 case MCLOSE + 9:
|
|
4242 {
|
|
4243 no = op - MCLOSE;
|
|
4244 cleanup_subexpr();
|
270
|
4245 rp = regstack_push(RS_MCLOSE, scan);
|
180
|
4246 if (rp == NULL)
|
|
4247 status = RA_FAIL;
|
|
4248 else
|
|
4249 {
|
|
4250 rp->rs_no = no;
|
|
4251 save_se(&rp->rs_un.sesave, ®_endpos[no], ®_endp[no]);
|
|
4252 /* We simply continue and handle the result when done. */
|
|
4253 }
|
7
|
4254 }
|
180
|
4255 break;
|
7
|
4256
|
|
4257 #ifdef FEAT_SYN_HL
|
|
4258 case ZCLOSE + 1: /* \) after \z( */
|
|
4259 case ZCLOSE + 2:
|
|
4260 case ZCLOSE + 3:
|
|
4261 case ZCLOSE + 4:
|
|
4262 case ZCLOSE + 5:
|
|
4263 case ZCLOSE + 6:
|
|
4264 case ZCLOSE + 7:
|
|
4265 case ZCLOSE + 8:
|
|
4266 case ZCLOSE + 9:
|
|
4267 {
|
|
4268 no = op - ZCLOSE;
|
|
4269 cleanup_zsubexpr();
|
270
|
4270 rp = regstack_push(RS_ZCLOSE, scan);
|
180
|
4271 if (rp == NULL)
|
|
4272 status = RA_FAIL;
|
|
4273 else
|
|
4274 {
|
|
4275 rp->rs_no = no;
|
|
4276 save_se(&rp->rs_un.sesave, ®_endzpos[no],
|
|
4277 ®_endzp[no]);
|
|
4278 /* We simply continue and handle the result when done. */
|
|
4279 }
|
7
|
4280 }
|
180
|
4281 break;
|
7
|
4282 #endif
|
|
4283
|
|
4284 case BACKREF + 1:
|
|
4285 case BACKREF + 2:
|
|
4286 case BACKREF + 3:
|
|
4287 case BACKREF + 4:
|
|
4288 case BACKREF + 5:
|
|
4289 case BACKREF + 6:
|
|
4290 case BACKREF + 7:
|
|
4291 case BACKREF + 8:
|
|
4292 case BACKREF + 9:
|
|
4293 {
|
|
4294 int len;
|
|
4295 linenr_T clnum;
|
|
4296 colnr_T ccol;
|
|
4297 char_u *p;
|
|
4298
|
|
4299 no = op - BACKREF;
|
|
4300 cleanup_subexpr();
|
|
4301 if (!REG_MULTI) /* Single-line regexp */
|
|
4302 {
|
|
4303 if (reg_endp[no] == NULL)
|
|
4304 {
|
|
4305 /* Backref was not set: Match an empty string. */
|
|
4306 len = 0;
|
|
4307 }
|
|
4308 else
|
|
4309 {
|
|
4310 /* Compare current input with back-ref in the same
|
|
4311 * line. */
|
|
4312 len = (int)(reg_endp[no] - reg_startp[no]);
|
|
4313 if (cstrncmp(reg_startp[no], reginput, &len) != 0)
|
180
|
4314 status = RA_NOMATCH;
|
7
|
4315 }
|
|
4316 }
|
|
4317 else /* Multi-line regexp */
|
|
4318 {
|
|
4319 if (reg_endpos[no].lnum < 0)
|
|
4320 {
|
|
4321 /* Backref was not set: Match an empty string. */
|
|
4322 len = 0;
|
|
4323 }
|
|
4324 else
|
|
4325 {
|
|
4326 if (reg_startpos[no].lnum == reglnum
|
|
4327 && reg_endpos[no].lnum == reglnum)
|
|
4328 {
|
|
4329 /* Compare back-ref within the current line. */
|
|
4330 len = reg_endpos[no].col - reg_startpos[no].col;
|
|
4331 if (cstrncmp(regline + reg_startpos[no].col,
|
|
4332 reginput, &len) != 0)
|
180
|
4333 status = RA_NOMATCH;
|
7
|
4334 }
|
|
4335 else
|
|
4336 {
|
|
4337 /* Messy situation: Need to compare between two
|
|
4338 * lines. */
|
|
4339 ccol = reg_startpos[no].col;
|
|
4340 clnum = reg_startpos[no].lnum;
|
|
4341 for (;;)
|
|
4342 {
|
|
4343 /* Since getting one line may invalidate
|
|
4344 * the other, need to make copy. Slow! */
|
|
4345 if (regline != reg_tofree)
|
|
4346 {
|
|
4347 len = (int)STRLEN(regline);
|
|
4348 if (reg_tofree == NULL
|
|
4349 || len >= (int)reg_tofreelen)
|
|
4350 {
|
|
4351 len += 50; /* get some extra */
|
|
4352 vim_free(reg_tofree);
|
|
4353 reg_tofree = alloc(len);
|
|
4354 if (reg_tofree == NULL)
|
180
|
4355 {
|
|
4356 status = RA_FAIL; /* outof memory!*/
|
|
4357 break;
|
|
4358 }
|
7
|
4359 reg_tofreelen = len;
|
|
4360 }
|
|
4361 STRCPY(reg_tofree, regline);
|
|
4362 reginput = reg_tofree
|
|
4363 + (reginput - regline);
|
|
4364 regline = reg_tofree;
|
|
4365 }
|
|
4366
|
|
4367 /* Get the line to compare with. */
|
|
4368 p = reg_getline(clnum);
|
|
4369 if (clnum == reg_endpos[no].lnum)
|
|
4370 len = reg_endpos[no].col - ccol;
|
|
4371 else
|
|
4372 len = (int)STRLEN(p + ccol);
|
|
4373
|
|
4374 if (cstrncmp(p + ccol, reginput, &len) != 0)
|
180
|
4375 {
|
|
4376 status = RA_NOMATCH; /* doesn't match */
|
|
4377 break;
|
|
4378 }
|
7
|
4379 if (clnum == reg_endpos[no].lnum)
|
|
4380 break; /* match and at end! */
|
481
|
4381 if (reglnum >= reg_maxline)
|
180
|
4382 {
|
|
4383 status = RA_NOMATCH; /* text too short */
|
|
4384 break;
|
|
4385 }
|
7
|
4386
|
|
4387 /* Advance to next line. */
|
|
4388 reg_nextline();
|
|
4389 ++clnum;
|
|
4390 ccol = 0;
|
180
|
4391 if (got_int)
|
|
4392 {
|
|
4393 status = RA_FAIL;
|
|
4394 break;
|
|
4395 }
|
7
|
4396 }
|
|
4397
|
|
4398 /* found a match! Note that regline may now point
|
|
4399 * to a copy of the line, that should not matter. */
|
|
4400 }
|
|
4401 }
|
|
4402 }
|
|
4403
|
|
4404 /* Matched the backref, skip over it. */
|
|
4405 reginput += len;
|
|
4406 }
|
|
4407 break;
|
|
4408
|
|
4409 #ifdef FEAT_SYN_HL
|
|
4410 case ZREF + 1:
|
|
4411 case ZREF + 2:
|
|
4412 case ZREF + 3:
|
|
4413 case ZREF + 4:
|
|
4414 case ZREF + 5:
|
|
4415 case ZREF + 6:
|
|
4416 case ZREF + 7:
|
|
4417 case ZREF + 8:
|
|
4418 case ZREF + 9:
|
|
4419 {
|
|
4420 int len;
|
|
4421
|
|
4422 cleanup_zsubexpr();
|
|
4423 no = op - ZREF;
|
|
4424 if (re_extmatch_in != NULL
|
|
4425 && re_extmatch_in->matches[no] != NULL)
|
|
4426 {
|
|
4427 len = (int)STRLEN(re_extmatch_in->matches[no]);
|
|
4428 if (cstrncmp(re_extmatch_in->matches[no],
|
|
4429 reginput, &len) != 0)
|
180
|
4430 status = RA_NOMATCH;
|
|
4431 else
|
|
4432 reginput += len;
|
7
|
4433 }
|
|
4434 else
|
|
4435 {
|
|
4436 /* Backref was not set: Match an empty string. */
|
|
4437 }
|
|
4438 }
|
|
4439 break;
|
|
4440 #endif
|
|
4441
|
|
4442 case BRANCH:
|
|
4443 {
|
|
4444 if (OP(next) != BRANCH) /* No choice. */
|
|
4445 next = OPERAND(scan); /* Avoid recursion. */
|
|
4446 else
|
|
4447 {
|
270
|
4448 rp = regstack_push(RS_BRANCH, scan);
|
180
|
4449 if (rp == NULL)
|
|
4450 status = RA_FAIL;
|
|
4451 else
|
|
4452 status = RA_BREAK; /* rest is below */
|
7
|
4453 }
|
|
4454 }
|
|
4455 break;
|
|
4456
|
|
4457 case BRACE_LIMITS:
|
|
4458 {
|
|
4459 if (OP(next) == BRACE_SIMPLE)
|
|
4460 {
|
|
4461 bl_minval = OPERAND_MIN(scan);
|
|
4462 bl_maxval = OPERAND_MAX(scan);
|
|
4463 }
|
|
4464 else if (OP(next) >= BRACE_COMPLEX
|
|
4465 && OP(next) < BRACE_COMPLEX + 10)
|
|
4466 {
|
|
4467 no = OP(next) - BRACE_COMPLEX;
|
|
4468 brace_min[no] = OPERAND_MIN(scan);
|
|
4469 brace_max[no] = OPERAND_MAX(scan);
|
|
4470 brace_count[no] = 0;
|
|
4471 }
|
|
4472 else
|
|
4473 {
|
|
4474 EMSG(_(e_internal)); /* Shouldn't happen */
|
180
|
4475 status = RA_FAIL;
|
7
|
4476 }
|
|
4477 }
|
|
4478 break;
|
|
4479
|
|
4480 case BRACE_COMPLEX + 0:
|
|
4481 case BRACE_COMPLEX + 1:
|
|
4482 case BRACE_COMPLEX + 2:
|
|
4483 case BRACE_COMPLEX + 3:
|
|
4484 case BRACE_COMPLEX + 4:
|
|
4485 case BRACE_COMPLEX + 5:
|
|
4486 case BRACE_COMPLEX + 6:
|
|
4487 case BRACE_COMPLEX + 7:
|
|
4488 case BRACE_COMPLEX + 8:
|
|
4489 case BRACE_COMPLEX + 9:
|
|
4490 {
|
|
4491 no = op - BRACE_COMPLEX;
|
|
4492 ++brace_count[no];
|
|
4493
|
|
4494 /* If not matched enough times yet, try one more */
|
|
4495 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
|
180
|
4496 ? brace_min[no] : brace_max[no]))
|
7
|
4497 {
|
270
|
4498 rp = regstack_push(RS_BRCPLX_MORE, scan);
|
180
|
4499 if (rp == NULL)
|
|
4500 status = RA_FAIL;
|
|
4501 else
|
|
4502 {
|
|
4503 rp->rs_no = no;
|
233
|
4504 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
4505 next = OPERAND(scan);
|
|
4506 /* We continue and handle the result when done. */
|
|
4507 }
|
|
4508 break;
|
7
|
4509 }
|
|
4510
|
|
4511 /* If matched enough times, may try matching some more */
|
|
4512 if (brace_min[no] <= brace_max[no])
|
|
4513 {
|
|
4514 /* Range is the normal way around, use longest match */
|
|
4515 if (brace_count[no] <= brace_max[no])
|
|
4516 {
|
270
|
4517 rp = regstack_push(RS_BRCPLX_LONG, scan);
|
180
|
4518 if (rp == NULL)
|
|
4519 status = RA_FAIL;
|
|
4520 else
|
|
4521 {
|
|
4522 rp->rs_no = no;
|
233
|
4523 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
4524 next = OPERAND(scan);
|
|
4525 /* We continue and handle the result when done. */
|
|
4526 }
|
7
|
4527 }
|
|
4528 }
|
|
4529 else
|
|
4530 {
|
|
4531 /* Range is backwards, use shortest match first */
|
|
4532 if (brace_count[no] <= brace_min[no])
|
|
4533 {
|
270
|
4534 rp = regstack_push(RS_BRCPLX_SHORT, scan);
|
180
|
4535 if (rp == NULL)
|
|
4536 status = RA_FAIL;
|
|
4537 else
|
|
4538 {
|
233
|
4539 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
4540 /* We continue and handle the result when done. */
|
|
4541 }
|
7
|
4542 }
|
|
4543 }
|
|
4544 }
|
|
4545 break;
|
|
4546
|
|
4547 case BRACE_SIMPLE:
|
|
4548 case STAR:
|
|
4549 case PLUS:
|
|
4550 {
|
180
|
4551 regstar_T rst;
|
7
|
4552
|
|
4553 /*
|
|
4554 * Lookahead to avoid useless match attempts when we know
|
|
4555 * what character comes next.
|
|
4556 */
|
|
4557 if (OP(next) == EXACTLY)
|
|
4558 {
|
180
|
4559 rst.nextb = *OPERAND(next);
|
7
|
4560 if (ireg_ic)
|
|
4561 {
|
180
|
4562 if (isupper(rst.nextb))
|
|
4563 rst.nextb_ic = TOLOWER_LOC(rst.nextb);
|
7
|
4564 else
|
180
|
4565 rst.nextb_ic = TOUPPER_LOC(rst.nextb);
|
7
|
4566 }
|
|
4567 else
|
180
|
4568 rst.nextb_ic = rst.nextb;
|
7
|
4569 }
|
|
4570 else
|
|
4571 {
|
180
|
4572 rst.nextb = NUL;
|
|
4573 rst.nextb_ic = NUL;
|
7
|
4574 }
|
|
4575 if (op != BRACE_SIMPLE)
|
|
4576 {
|
180
|
4577 rst.minval = (op == STAR) ? 0 : 1;
|
|
4578 rst.maxval = MAX_LIMIT;
|
7
|
4579 }
|
|
4580 else
|
|
4581 {
|
180
|
4582 rst.minval = bl_minval;
|
|
4583 rst.maxval = bl_maxval;
|
7
|
4584 }
|
|
4585
|
|
4586 /*
|
|
4587 * When maxval > minval, try matching as much as possible, up
|
|
4588 * to maxval. When maxval < minval, try matching at least the
|
|
4589 * minimal number (since the range is backwards, that's also
|
|
4590 * maxval!).
|
|
4591 */
|
180
|
4592 rst.count = regrepeat(OPERAND(scan), rst.maxval);
|
7
|
4593 if (got_int)
|
180
|
4594 {
|
|
4595 status = RA_FAIL;
|
|
4596 break;
|
|
4597 }
|
|
4598 if (rst.minval <= rst.maxval
|
|
4599 ? rst.count >= rst.minval : rst.count >= rst.maxval)
|
7
|
4600 {
|
180
|
4601 /* It could match. Prepare for trying to match what
|
|
4602 * follows. The code is below. Parameters are stored in
|
|
4603 * a regstar_T on the regstack. */
|
212
|
4604 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
|
189
|
4605 {
|
|
4606 EMSG(_(e_maxmempat));
|
|
4607 status = RA_FAIL;
|
|
4608 }
|
|
4609 else if (ga_grow(®stack, sizeof(regstar_T)) == FAIL)
|
180
|
4610 status = RA_FAIL;
|
|
4611 else
|
7
|
4612 {
|
180
|
4613 regstack.ga_len += sizeof(regstar_T);
|
270
|
4614 rp = regstack_push(rst.minval <= rst.maxval
|
233
|
4615 ? RS_STAR_LONG : RS_STAR_SHORT, scan);
|
180
|
4616 if (rp == NULL)
|
|
4617 status = RA_FAIL;
|
|
4618 else
|
7
|
4619 {
|
180
|
4620 *(((regstar_T *)rp) - 1) = rst;
|
|
4621 status = RA_BREAK; /* skip the restore bits */
|
7
|
4622 }
|
|
4623 }
|
|
4624 }
|
|
4625 else
|
180
|
4626 status = RA_NOMATCH;
|
|
4627
|
7
|
4628 }
|
180
|
4629 break;
|
7
|
4630
|
|
4631 case NOMATCH:
|
|
4632 case MATCH:
|
|
4633 case SUBPAT:
|
270
|
4634 rp = regstack_push(RS_NOMATCH, scan);
|
180
|
4635 if (rp == NULL)
|
|
4636 status = RA_FAIL;
|
|
4637 else
|
7
|
4638 {
|
180
|
4639 rp->rs_no = op;
|
233
|
4640 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
4641 next = OPERAND(scan);
|
|
4642 /* We continue and handle the result when done. */
|
7
|
4643 }
|
|
4644 break;
|
|
4645
|
|
4646 case BEHIND:
|
|
4647 case NOBEHIND:
|
180
|
4648 /* Need a bit of room to store extra positions. */
|
212
|
4649 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
|
189
|
4650 {
|
|
4651 EMSG(_(e_maxmempat));
|
|
4652 status = RA_FAIL;
|
|
4653 }
|
|
4654 else if (ga_grow(®stack, sizeof(regbehind_T)) == FAIL)
|
180
|
4655 status = RA_FAIL;
|
|
4656 else
|
7
|
4657 {
|
180
|
4658 regstack.ga_len += sizeof(regbehind_T);
|
270
|
4659 rp = regstack_push(RS_BEHIND1, scan);
|
180
|
4660 if (rp == NULL)
|
|
4661 status = RA_FAIL;
|
|
4662 else
|
7
|
4663 {
|
180
|
4664 rp->rs_no = op;
|
233
|
4665 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
4666 /* First try if what follows matches. If it does then we
|
|
4667 * check the behind match by looping. */
|
7
|
4668 }
|
|
4669 }
|
180
|
4670 break;
|
7
|
4671
|
|
4672 case BHPOS:
|
|
4673 if (REG_MULTI)
|
|
4674 {
|
|
4675 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline)
|
|
4676 || behind_pos.rs_u.pos.lnum != reglnum)
|
180
|
4677 status = RA_NOMATCH;
|
7
|
4678 }
|
|
4679 else if (behind_pos.rs_u.ptr != reginput)
|
180
|
4680 status = RA_NOMATCH;
|
7
|
4681 break;
|
|
4682
|
|
4683 case NEWL:
|
481
|
4684 if ((c != NUL || reglnum > reg_maxline || reg_line_lbr)
|
7
|
4685 && (c != '\n' || !reg_line_lbr))
|
180
|
4686 status = RA_NOMATCH;
|
|
4687 else if (reg_line_lbr)
|
7
|
4688 ADVANCE_REGINPUT();
|
|
4689 else
|
|
4690 reg_nextline();
|
|
4691 break;
|
|
4692
|
|
4693 case END:
|
180
|
4694 status = RA_MATCH; /* Success! */
|
|
4695 break;
|
7
|
4696
|
|
4697 default:
|
|
4698 EMSG(_(e_re_corr));
|
|
4699 #ifdef DEBUG
|
|
4700 printf("Illegal op code %d\n", op);
|
|
4701 #endif
|
180
|
4702 status = RA_FAIL;
|
|
4703 break;
|
7
|
4704 }
|
|
4705 }
|
|
4706
|
180
|
4707 /* If we can't continue sequentially, break the inner loop. */
|
|
4708 if (status != RA_CONT)
|
|
4709 break;
|
|
4710
|
|
4711 /* Continue in inner loop, advance to next item. */
|
7
|
4712 scan = next;
|
180
|
4713
|
|
4714 } /* end of inner loop */
|
7
|
4715
|
|
4716 /*
|
180
|
4717 * If there is something on the regstack execute the code for the state.
|
233
|
4718 * If the state is popped then loop and use the older state.
|
7
|
4719 */
|
180
|
4720 while (regstack.ga_len > 0 && status != RA_FAIL)
|
|
4721 {
|
|
4722 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
|
|
4723 switch (rp->rs_state)
|
|
4724 {
|
|
4725 case RS_NOPEN:
|
|
4726 /* Result is passed on as-is, simply pop the state. */
|
270
|
4727 regstack_pop(&scan);
|
180
|
4728 break;
|
|
4729
|
|
4730 case RS_MOPEN:
|
|
4731 /* Pop the state. Restore pointers when there is no match. */
|
|
4732 if (status == RA_NOMATCH)
|
|
4733 restore_se(&rp->rs_un.sesave, ®_startpos[rp->rs_no],
|
|
4734 ®_startp[rp->rs_no]);
|
270
|
4735 regstack_pop(&scan);
|
180
|
4736 break;
|
|
4737
|
|
4738 #ifdef FEAT_SYN_HL
|
|
4739 case RS_ZOPEN:
|
|
4740 /* Pop the state. Restore pointers when there is no match. */
|
|
4741 if (status == RA_NOMATCH)
|
|
4742 restore_se(&rp->rs_un.sesave, ®_startzpos[rp->rs_no],
|
|
4743 ®_startzp[rp->rs_no]);
|
270
|
4744 regstack_pop(&scan);
|
180
|
4745 break;
|
|
4746 #endif
|
|
4747
|
|
4748 case RS_MCLOSE:
|
|
4749 /* Pop the state. Restore pointers when there is no match. */
|
|
4750 if (status == RA_NOMATCH)
|
|
4751 restore_se(&rp->rs_un.sesave, ®_endpos[rp->rs_no],
|
|
4752 ®_endp[rp->rs_no]);
|
270
|
4753 regstack_pop(&scan);
|
180
|
4754 break;
|
|
4755
|
|
4756 #ifdef FEAT_SYN_HL
|
|
4757 case RS_ZCLOSE:
|
|
4758 /* Pop the state. Restore pointers when there is no match. */
|
|
4759 if (status == RA_NOMATCH)
|
|
4760 restore_se(&rp->rs_un.sesave, ®_endzpos[rp->rs_no],
|
|
4761 ®_endzp[rp->rs_no]);
|
270
|
4762 regstack_pop(&scan);
|
180
|
4763 break;
|
|
4764 #endif
|
|
4765
|
|
4766 case RS_BRANCH:
|
|
4767 if (status == RA_MATCH)
|
|
4768 /* this branch matched, use it */
|
270
|
4769 regstack_pop(&scan);
|
180
|
4770 else
|
|
4771 {
|
|
4772 if (status != RA_BREAK)
|
|
4773 {
|
|
4774 /* After a non-matching branch: try next one. */
|
233
|
4775 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4776 scan = rp->rs_scan;
|
|
4777 }
|
|
4778 if (scan == NULL || OP(scan) != BRANCH)
|
|
4779 {
|
|
4780 /* no more branches, didn't find a match */
|
|
4781 status = RA_NOMATCH;
|
270
|
4782 regstack_pop(&scan);
|
180
|
4783 }
|
|
4784 else
|
|
4785 {
|
|
4786 /* Prepare to try a branch. */
|
|
4787 rp->rs_scan = regnext(scan);
|
233
|
4788 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
4789 scan = OPERAND(scan);
|
|
4790 }
|
|
4791 }
|
|
4792 break;
|
|
4793
|
|
4794 case RS_BRCPLX_MORE:
|
|
4795 /* Pop the state. Restore pointers when there is no match. */
|
|
4796 if (status == RA_NOMATCH)
|
|
4797 {
|
233
|
4798 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4799 --brace_count[rp->rs_no]; /* decrement match count */
|
|
4800 }
|
270
|
4801 regstack_pop(&scan);
|
180
|
4802 break;
|
|
4803
|
|
4804 case RS_BRCPLX_LONG:
|
|
4805 /* Pop the state. Restore pointers when there is no match. */
|
|
4806 if (status == RA_NOMATCH)
|
|
4807 {
|
|
4808 /* There was no match, but we did find enough matches. */
|
233
|
4809 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4810 --brace_count[rp->rs_no];
|
|
4811 /* continue with the items after "\{}" */
|
|
4812 status = RA_CONT;
|
|
4813 }
|
270
|
4814 regstack_pop(&scan);
|
180
|
4815 if (status == RA_CONT)
|
|
4816 scan = regnext(scan);
|
|
4817 break;
|
|
4818
|
|
4819 case RS_BRCPLX_SHORT:
|
|
4820 /* Pop the state. Restore pointers when there is no match. */
|
|
4821 if (status == RA_NOMATCH)
|
|
4822 /* There was no match, try to match one more item. */
|
233
|
4823 reg_restore(&rp->rs_un.regsave, &backpos);
|
270
|
4824 regstack_pop(&scan);
|
180
|
4825 if (status == RA_NOMATCH)
|
|
4826 {
|
|
4827 scan = OPERAND(scan);
|
|
4828 status = RA_CONT;
|
|
4829 }
|
|
4830 break;
|
|
4831
|
|
4832 case RS_NOMATCH:
|
|
4833 /* Pop the state. If the operand matches for NOMATCH or
|
|
4834 * doesn't match for MATCH/SUBPAT, we fail. Otherwise backup,
|
|
4835 * except for SUBPAT, and continue with the next item. */
|
|
4836 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH))
|
|
4837 status = RA_NOMATCH;
|
|
4838 else
|
|
4839 {
|
|
4840 status = RA_CONT;
|
233
|
4841 if (rp->rs_no != SUBPAT) /* zero-width */
|
|
4842 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4843 }
|
270
|
4844 regstack_pop(&scan);
|
180
|
4845 if (status == RA_CONT)
|
|
4846 scan = regnext(scan);
|
|
4847 break;
|
|
4848
|
|
4849 case RS_BEHIND1:
|
|
4850 if (status == RA_NOMATCH)
|
|
4851 {
|
270
|
4852 regstack_pop(&scan);
|
180
|
4853 regstack.ga_len -= sizeof(regbehind_T);
|
|
4854 }
|
|
4855 else
|
|
4856 {
|
|
4857 /* The stuff after BEHIND/NOBEHIND matches. Now try if
|
|
4858 * the behind part does (not) match before the current
|
|
4859 * position in the input. This must be done at every
|
|
4860 * position in the input and checking if the match ends at
|
|
4861 * the current position. */
|
|
4862
|
|
4863 /* save the position after the found match for next */
|
233
|
4864 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos);
|
180
|
4865
|
|
4866 /* start looking for a match with operand at the current
|
|
4867 * postion. Go back one character until we find the
|
|
4868 * result, hitting the start of the line or the previous
|
|
4869 * line (for multi-line matching).
|
|
4870 * Set behind_pos to where the match should end, BHPOS
|
|
4871 * will match it. Save the current value. */
|
|
4872 (((regbehind_T *)rp) - 1)->save_behind = behind_pos;
|
|
4873 behind_pos = rp->rs_un.regsave;
|
|
4874
|
|
4875 rp->rs_state = RS_BEHIND2;
|
|
4876
|
233
|
4877 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4878 scan = OPERAND(rp->rs_scan);
|
|
4879 }
|
|
4880 break;
|
|
4881
|
|
4882 case RS_BEHIND2:
|
|
4883 /*
|
|
4884 * Looping for BEHIND / NOBEHIND match.
|
|
4885 */
|
|
4886 if (status == RA_MATCH && reg_save_equal(&behind_pos))
|
|
4887 {
|
|
4888 /* found a match that ends where "next" started */
|
|
4889 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
|
|
4890 if (rp->rs_no == BEHIND)
|
233
|
4891 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
|
|
4892 &backpos);
|
180
|
4893 else
|
|
4894 /* But we didn't want a match. */
|
|
4895 status = RA_NOMATCH;
|
270
|
4896 regstack_pop(&scan);
|
180
|
4897 regstack.ga_len -= sizeof(regbehind_T);
|
|
4898 }
|
|
4899 else
|
|
4900 {
|
|
4901 /* No match: Go back one character. May go to previous
|
|
4902 * line once. */
|
|
4903 no = OK;
|
|
4904 if (REG_MULTI)
|
|
4905 {
|
|
4906 if (rp->rs_un.regsave.rs_u.pos.col == 0)
|
|
4907 {
|
|
4908 if (rp->rs_un.regsave.rs_u.pos.lnum
|
|
4909 < behind_pos.rs_u.pos.lnum
|
|
4910 || reg_getline(
|
|
4911 --rp->rs_un.regsave.rs_u.pos.lnum)
|
|
4912 == NULL)
|
|
4913 no = FAIL;
|
|
4914 else
|
|
4915 {
|
233
|
4916 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4917 rp->rs_un.regsave.rs_u.pos.col =
|
|
4918 (colnr_T)STRLEN(regline);
|
|
4919 }
|
|
4920 }
|
|
4921 else
|
|
4922 --rp->rs_un.regsave.rs_u.pos.col;
|
|
4923 }
|
|
4924 else
|
|
4925 {
|
|
4926 if (rp->rs_un.regsave.rs_u.ptr == regline)
|
|
4927 no = FAIL;
|
|
4928 else
|
|
4929 --rp->rs_un.regsave.rs_u.ptr;
|
|
4930 }
|
|
4931 if (no == OK)
|
|
4932 {
|
|
4933 /* Advanced, prepare for finding match again. */
|
233
|
4934 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4935 scan = OPERAND(rp->rs_scan);
|
|
4936 }
|
|
4937 else
|
|
4938 {
|
|
4939 /* Can't advance. For NOBEHIND that's a match. */
|
|
4940 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
|
|
4941 if (rp->rs_no == NOBEHIND)
|
|
4942 {
|
233
|
4943 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
|
|
4944 &backpos);
|
180
|
4945 status = RA_MATCH;
|
|
4946 }
|
|
4947 else
|
|
4948 status = RA_NOMATCH;
|
270
|
4949 regstack_pop(&scan);
|
180
|
4950 regstack.ga_len -= sizeof(regbehind_T);
|
|
4951 }
|
|
4952 }
|
|
4953 break;
|
|
4954
|
|
4955 case RS_STAR_LONG:
|
|
4956 case RS_STAR_SHORT:
|
|
4957 {
|
|
4958 regstar_T *rst = ((regstar_T *)rp) - 1;
|
|
4959
|
|
4960 if (status == RA_MATCH)
|
|
4961 {
|
270
|
4962 regstack_pop(&scan);
|
180
|
4963 regstack.ga_len -= sizeof(regstar_T);
|
|
4964 break;
|
|
4965 }
|
|
4966
|
|
4967 /* Tried once already, restore input pointers. */
|
|
4968 if (status != RA_BREAK)
|
233
|
4969 reg_restore(&rp->rs_un.regsave, &backpos);
|
180
|
4970
|
|
4971 /* Repeat until we found a position where it could match. */
|
|
4972 for (;;)
|
|
4973 {
|
|
4974 if (status != RA_BREAK)
|
|
4975 {
|
|
4976 /* Tried first position already, advance. */
|
|
4977 if (rp->rs_state == RS_STAR_LONG)
|
|
4978 {
|
|
4979 /* Trying for longest matc, but couldn't or didn't
|
|
4980 * match -- back up one char. */
|
|
4981 if (--rst->count < rst->minval)
|
|
4982 break;
|
|
4983 if (reginput == regline)
|
|
4984 {
|
|
4985 /* backup to last char of previous line */
|
|
4986 --reglnum;
|
|
4987 regline = reg_getline(reglnum);
|
|
4988 /* Just in case regrepeat() didn't count
|
|
4989 * right. */
|
|
4990 if (regline == NULL)
|
|
4991 break;
|
|
4992 reginput = regline + STRLEN(regline);
|
|
4993 fast_breakcheck();
|
|
4994 }
|
|
4995 else
|
|
4996 mb_ptr_back(regline, reginput);
|
|
4997 }
|
|
4998 else
|
|
4999 {
|
|
5000 /* Range is backwards, use shortest match first.
|
|
5001 * Careful: maxval and minval are exchanged!
|
|
5002 * Couldn't or didn't match: try advancing one
|
|
5003 * char. */
|
|
5004 if (rst->count == rst->minval
|
|
5005 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0)
|
|
5006 break;
|
|
5007 ++rst->count;
|
|
5008 }
|
|
5009 if (got_int)
|
|
5010 break;
|
|
5011 }
|
|
5012 else
|
|
5013 status = RA_NOMATCH;
|
|
5014
|
|
5015 /* If it could match, try it. */
|
|
5016 if (rst->nextb == NUL || *reginput == rst->nextb
|
|
5017 || *reginput == rst->nextb_ic)
|
|
5018 {
|
233
|
5019 reg_save(&rp->rs_un.regsave, &backpos);
|
180
|
5020 scan = regnext(rp->rs_scan);
|
|
5021 status = RA_CONT;
|
|
5022 break;
|
|
5023 }
|
|
5024 }
|
|
5025 if (status != RA_CONT)
|
|
5026 {
|
|
5027 /* Failed. */
|
270
|
5028 regstack_pop(&scan);
|
180
|
5029 regstack.ga_len -= sizeof(regstar_T);
|
|
5030 status = RA_NOMATCH;
|
|
5031 }
|
|
5032 }
|
|
5033 break;
|
|
5034 }
|
|
5035
|
|
5036 /* If we want to continue the inner loop or didn't pop a state contine
|
|
5037 * matching loop */
|
|
5038 if (status == RA_CONT || rp == (regitem_T *)
|
|
5039 ((char *)regstack.ga_data + regstack.ga_len) - 1)
|
|
5040 break;
|
|
5041 }
|
|
5042
|
189
|
5043 /* May need to continue with the inner loop, starting at "scan". */
|
180
|
5044 if (status == RA_CONT)
|
|
5045 continue;
|
|
5046
|
|
5047 /*
|
|
5048 * If the regstack is empty or something failed we are done.
|
|
5049 */
|
|
5050 if (regstack.ga_len == 0 || status == RA_FAIL)
|
|
5051 {
|
|
5052 if (scan == NULL)
|
|
5053 {
|
|
5054 /*
|
|
5055 * We get here only if there's trouble -- normally "case END" is
|
|
5056 * the terminating point.
|
|
5057 */
|
|
5058 EMSG(_(e_re_corr));
|
7
|
5059 #ifdef DEBUG
|
180
|
5060 printf("Premature EOL\n");
|
7
|
5061 #endif
|
180
|
5062 }
|
189
|
5063 if (status == RA_FAIL)
|
|
5064 got_int = TRUE;
|
180
|
5065 return (status == RA_MATCH);
|
|
5066 }
|
|
5067
|
|
5068 } /* End of loop until the regstack is empty. */
|
|
5069
|
|
5070 /* NOTREACHED */
|
|
5071 }
|
|
5072
|
|
5073 /*
|
|
5074 * Push an item onto the regstack.
|
|
5075 * Returns pointer to new item. Returns NULL when out of memory.
|
|
5076 */
|
|
5077 static regitem_T *
|
270
|
5078 regstack_push(state, scan)
|
180
|
5079 regstate_T state;
|
|
5080 char_u *scan;
|
|
5081 {
|
|
5082 regitem_T *rp;
|
|
5083
|
270
|
5084 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
|
189
|
5085 {
|
|
5086 EMSG(_(e_maxmempat));
|
|
5087 return NULL;
|
|
5088 }
|
270
|
5089 if (ga_grow(®stack, sizeof(regitem_T)) == FAIL)
|
180
|
5090 return NULL;
|
|
5091
|
270
|
5092 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len);
|
180
|
5093 rp->rs_state = state;
|
|
5094 rp->rs_scan = scan;
|
|
5095
|
270
|
5096 regstack.ga_len += sizeof(regitem_T);
|
180
|
5097 return rp;
|
|
5098 }
|
|
5099
|
|
5100 /*
|
|
5101 * Pop an item from the regstack.
|
|
5102 */
|
|
5103 static void
|
270
|
5104 regstack_pop(scan)
|
180
|
5105 char_u **scan;
|
|
5106 {
|
|
5107 regitem_T *rp;
|
|
5108
|
270
|
5109 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
|
180
|
5110 *scan = rp->rs_scan;
|
|
5111
|
270
|
5112 regstack.ga_len -= sizeof(regitem_T);
|
7
|
5113 }
|
|
5114
|
|
5115 /*
|
|
5116 * regrepeat - repeatedly match something simple, return how many.
|
|
5117 * Advances reginput (and reglnum) to just after the matched chars.
|
|
5118 */
|
|
5119 static int
|
|
5120 regrepeat(p, maxcount)
|
|
5121 char_u *p;
|
|
5122 long maxcount; /* maximum number of matches allowed */
|
|
5123 {
|
|
5124 long count = 0;
|
|
5125 char_u *scan;
|
|
5126 char_u *opnd;
|
|
5127 int mask;
|
|
5128 int testval = 0;
|
|
5129
|
|
5130 scan = reginput; /* Make local copy of reginput for speed. */
|
|
5131 opnd = OPERAND(p);
|
|
5132 switch (OP(p))
|
|
5133 {
|
|
5134 case ANY:
|
|
5135 case ANY + ADD_NL:
|
|
5136 while (count < maxcount)
|
|
5137 {
|
|
5138 /* Matching anything means we continue until end-of-line (or
|
|
5139 * end-of-file for ANY + ADD_NL), only limited by maxcount. */
|
|
5140 while (*scan != NUL && count < maxcount)
|
|
5141 {
|
|
5142 ++count;
|
39
|
5143 mb_ptr_adv(scan);
|
7
|
5144 }
|
481
|
5145 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr
|
|
5146 || count == maxcount)
|
7
|
5147 break;
|
|
5148 ++count; /* count the line-break */
|
|
5149 reg_nextline();
|
|
5150 scan = reginput;
|
|
5151 if (got_int)
|
|
5152 break;
|
|
5153 }
|
|
5154 break;
|
|
5155
|
|
5156 case IDENT:
|
|
5157 case IDENT + ADD_NL:
|
|
5158 testval = TRUE;
|
|
5159 /*FALLTHROUGH*/
|
|
5160 case SIDENT:
|
|
5161 case SIDENT + ADD_NL:
|
|
5162 while (count < maxcount)
|
|
5163 {
|
|
5164 if (vim_isIDc(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
5165 {
|
39
|
5166 mb_ptr_adv(scan);
|
7
|
5167 }
|
|
5168 else if (*scan == NUL)
|
|
5169 {
|
481
|
5170 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr)
|
7
|
5171 break;
|
|
5172 reg_nextline();
|
|
5173 scan = reginput;
|
|
5174 if (got_int)
|
|
5175 break;
|
|
5176 }
|
|
5177 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5178 ++scan;
|
|
5179 else
|
|
5180 break;
|
|
5181 ++count;
|
|
5182 }
|
|
5183 break;
|
|
5184
|
|
5185 case KWORD:
|
|
5186 case KWORD + ADD_NL:
|
|
5187 testval = TRUE;
|
|
5188 /*FALLTHROUGH*/
|
|
5189 case SKWORD:
|
|
5190 case SKWORD + ADD_NL:
|
|
5191 while (count < maxcount)
|
|
5192 {
|
|
5193 if (vim_iswordp(scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
5194 {
|
39
|
5195 mb_ptr_adv(scan);
|
7
|
5196 }
|
|
5197 else if (*scan == NUL)
|
|
5198 {
|
481
|
5199 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr)
|
7
|
5200 break;
|
|
5201 reg_nextline();
|
|
5202 scan = reginput;
|
|
5203 if (got_int)
|
|
5204 break;
|
|
5205 }
|
|
5206 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5207 ++scan;
|
|
5208 else
|
|
5209 break;
|
|
5210 ++count;
|
|
5211 }
|
|
5212 break;
|
|
5213
|
|
5214 case FNAME:
|
|
5215 case FNAME + ADD_NL:
|
|
5216 testval = TRUE;
|
|
5217 /*FALLTHROUGH*/
|
|
5218 case SFNAME:
|
|
5219 case SFNAME + ADD_NL:
|
|
5220 while (count < maxcount)
|
|
5221 {
|
|
5222 if (vim_isfilec(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
5223 {
|
39
|
5224 mb_ptr_adv(scan);
|
7
|
5225 }
|
|
5226 else if (*scan == NUL)
|
|
5227 {
|
481
|
5228 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr)
|
7
|
5229 break;
|
|
5230 reg_nextline();
|
|
5231 scan = reginput;
|
|
5232 if (got_int)
|
|
5233 break;
|
|
5234 }
|
|
5235 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5236 ++scan;
|
|
5237 else
|
|
5238 break;
|
|
5239 ++count;
|
|
5240 }
|
|
5241 break;
|
|
5242
|
|
5243 case PRINT:
|
|
5244 case PRINT + ADD_NL:
|
|
5245 testval = TRUE;
|
|
5246 /*FALLTHROUGH*/
|
|
5247 case SPRINT:
|
|
5248 case SPRINT + ADD_NL:
|
|
5249 while (count < maxcount)
|
|
5250 {
|
|
5251 if (*scan == NUL)
|
|
5252 {
|
481
|
5253 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr)
|
7
|
5254 break;
|
|
5255 reg_nextline();
|
|
5256 scan = reginput;
|
|
5257 if (got_int)
|
|
5258 break;
|
|
5259 }
|
|
5260 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan)))
|
|
5261 {
|
39
|
5262 mb_ptr_adv(scan);
|
7
|
5263 }
|
|
5264 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5265 ++scan;
|
|
5266 else
|
|
5267 break;
|
|
5268 ++count;
|
|
5269 }
|
|
5270 break;
|
|
5271
|
|
5272 case WHITE:
|
|
5273 case WHITE + ADD_NL:
|
|
5274 testval = mask = RI_WHITE;
|
|
5275 do_class:
|
|
5276 while (count < maxcount)
|
|
5277 {
|
|
5278 #ifdef FEAT_MBYTE
|
|
5279 int l;
|
|
5280 #endif
|
|
5281 if (*scan == NUL)
|
|
5282 {
|
481
|
5283 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr)
|
7
|
5284 break;
|
|
5285 reg_nextline();
|
|
5286 scan = reginput;
|
|
5287 if (got_int)
|
|
5288 break;
|
|
5289 }
|
|
5290 #ifdef FEAT_MBYTE
|
474
|
5291 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1)
|
7
|
5292 {
|
|
5293 if (testval != 0)
|
|
5294 break;
|
|
5295 scan += l;
|
|
5296 }
|
|
5297 #endif
|
|
5298 else if ((class_tab[*scan] & mask) == testval)
|
|
5299 ++scan;
|
|
5300 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5301 ++scan;
|
|
5302 else
|
|
5303 break;
|
|
5304 ++count;
|
|
5305 }
|
|
5306 break;
|
|
5307
|
|
5308 case NWHITE:
|
|
5309 case NWHITE + ADD_NL:
|
|
5310 mask = RI_WHITE;
|
|
5311 goto do_class;
|
|
5312 case DIGIT:
|
|
5313 case DIGIT + ADD_NL:
|
|
5314 testval = mask = RI_DIGIT;
|
|
5315 goto do_class;
|
|
5316 case NDIGIT:
|
|
5317 case NDIGIT + ADD_NL:
|
|
5318 mask = RI_DIGIT;
|
|
5319 goto do_class;
|
|
5320 case HEX:
|
|
5321 case HEX + ADD_NL:
|
|
5322 testval = mask = RI_HEX;
|
|
5323 goto do_class;
|
|
5324 case NHEX:
|
|
5325 case NHEX + ADD_NL:
|
|
5326 mask = RI_HEX;
|
|
5327 goto do_class;
|
|
5328 case OCTAL:
|
|
5329 case OCTAL + ADD_NL:
|
|
5330 testval = mask = RI_OCTAL;
|
|
5331 goto do_class;
|
|
5332 case NOCTAL:
|
|
5333 case NOCTAL + ADD_NL:
|
|
5334 mask = RI_OCTAL;
|
|
5335 goto do_class;
|
|
5336 case WORD:
|
|
5337 case WORD + ADD_NL:
|
|
5338 testval = mask = RI_WORD;
|
|
5339 goto do_class;
|
|
5340 case NWORD:
|
|
5341 case NWORD + ADD_NL:
|
|
5342 mask = RI_WORD;
|
|
5343 goto do_class;
|
|
5344 case HEAD:
|
|
5345 case HEAD + ADD_NL:
|
|
5346 testval = mask = RI_HEAD;
|
|
5347 goto do_class;
|
|
5348 case NHEAD:
|
|
5349 case NHEAD + ADD_NL:
|
|
5350 mask = RI_HEAD;
|
|
5351 goto do_class;
|
|
5352 case ALPHA:
|
|
5353 case ALPHA + ADD_NL:
|
|
5354 testval = mask = RI_ALPHA;
|
|
5355 goto do_class;
|
|
5356 case NALPHA:
|
|
5357 case NALPHA + ADD_NL:
|
|
5358 mask = RI_ALPHA;
|
|
5359 goto do_class;
|
|
5360 case LOWER:
|
|
5361 case LOWER + ADD_NL:
|
|
5362 testval = mask = RI_LOWER;
|
|
5363 goto do_class;
|
|
5364 case NLOWER:
|
|
5365 case NLOWER + ADD_NL:
|
|
5366 mask = RI_LOWER;
|
|
5367 goto do_class;
|
|
5368 case UPPER:
|
|
5369 case UPPER + ADD_NL:
|
|
5370 testval = mask = RI_UPPER;
|
|
5371 goto do_class;
|
|
5372 case NUPPER:
|
|
5373 case NUPPER + ADD_NL:
|
|
5374 mask = RI_UPPER;
|
|
5375 goto do_class;
|
|
5376
|
|
5377 case EXACTLY:
|
|
5378 {
|
|
5379 int cu, cl;
|
|
5380
|
|
5381 /* This doesn't do a multi-byte character, because a MULTIBYTECODE
|
|
5382 * would have been used for it. */
|
|
5383 if (ireg_ic)
|
|
5384 {
|
|
5385 cu = TOUPPER_LOC(*opnd);
|
|
5386 cl = TOLOWER_LOC(*opnd);
|
|
5387 while (count < maxcount && (*scan == cu || *scan == cl))
|
|
5388 {
|
|
5389 count++;
|
|
5390 scan++;
|
|
5391 }
|
|
5392 }
|
|
5393 else
|
|
5394 {
|
|
5395 cu = *opnd;
|
|
5396 while (count < maxcount && *scan == cu)
|
|
5397 {
|
|
5398 count++;
|
|
5399 scan++;
|
|
5400 }
|
|
5401 }
|
|
5402 break;
|
|
5403 }
|
|
5404
|
|
5405 #ifdef FEAT_MBYTE
|
|
5406 case MULTIBYTECODE:
|
|
5407 {
|
|
5408 int i, len, cf = 0;
|
|
5409
|
|
5410 /* Safety check (just in case 'encoding' was changed since
|
|
5411 * compiling the program). */
|
474
|
5412 if ((len = (*mb_ptr2len)(opnd)) > 1)
|
7
|
5413 {
|
|
5414 if (ireg_ic && enc_utf8)
|
|
5415 cf = utf_fold(utf_ptr2char(opnd));
|
|
5416 while (count < maxcount)
|
|
5417 {
|
|
5418 for (i = 0; i < len; ++i)
|
|
5419 if (opnd[i] != scan[i])
|
|
5420 break;
|
|
5421 if (i < len && (!ireg_ic || !enc_utf8
|
|
5422 || utf_fold(utf_ptr2char(scan)) != cf))
|
|
5423 break;
|
|
5424 scan += len;
|
|
5425 ++count;
|
|
5426 }
|
|
5427 }
|
|
5428 }
|
|
5429 break;
|
|
5430 #endif
|
|
5431
|
|
5432 case ANYOF:
|
|
5433 case ANYOF + ADD_NL:
|
|
5434 testval = TRUE;
|
|
5435 /*FALLTHROUGH*/
|
|
5436
|
|
5437 case ANYBUT:
|
|
5438 case ANYBUT + ADD_NL:
|
|
5439 while (count < maxcount)
|
|
5440 {
|
|
5441 #ifdef FEAT_MBYTE
|
|
5442 int len;
|
|
5443 #endif
|
|
5444 if (*scan == NUL)
|
|
5445 {
|
481
|
5446 if (!WITH_NL(OP(p)) || reglnum > reg_maxline || reg_line_lbr)
|
7
|
5447 break;
|
|
5448 reg_nextline();
|
|
5449 scan = reginput;
|
|
5450 if (got_int)
|
|
5451 break;
|
|
5452 }
|
|
5453 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5454 ++scan;
|
|
5455 #ifdef FEAT_MBYTE
|
474
|
5456 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1)
|
7
|
5457 {
|
|
5458 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
|
|
5459 break;
|
|
5460 scan += len;
|
|
5461 }
|
|
5462 #endif
|
|
5463 else
|
|
5464 {
|
|
5465 if ((cstrchr(opnd, *scan) == NULL) == testval)
|
|
5466 break;
|
|
5467 ++scan;
|
|
5468 }
|
|
5469 ++count;
|
|
5470 }
|
|
5471 break;
|
|
5472
|
|
5473 case NEWL:
|
|
5474 while (count < maxcount
|
481
|
5475 && ((*scan == NUL && reglnum <= reg_maxline && !reg_line_lbr)
|
7
|
5476 || (*scan == '\n' && reg_line_lbr)))
|
|
5477 {
|
|
5478 count++;
|
|
5479 if (reg_line_lbr)
|
|
5480 ADVANCE_REGINPUT();
|
|
5481 else
|
|
5482 reg_nextline();
|
|
5483 scan = reginput;
|
|
5484 if (got_int)
|
|
5485 break;
|
|
5486 }
|
|
5487 break;
|
|
5488
|
|
5489 default: /* Oh dear. Called inappropriately. */
|
|
5490 EMSG(_(e_re_corr));
|
|
5491 #ifdef DEBUG
|
|
5492 printf("Called regrepeat with op code %d\n", OP(p));
|
|
5493 #endif
|
|
5494 break;
|
|
5495 }
|
|
5496
|
|
5497 reginput = scan;
|
|
5498
|
|
5499 return (int)count;
|
|
5500 }
|
|
5501
|
|
5502 /*
|
|
5503 * regnext - dig the "next" pointer out of a node
|
|
5504 */
|
|
5505 static char_u *
|
|
5506 regnext(p)
|
|
5507 char_u *p;
|
|
5508 {
|
|
5509 int offset;
|
|
5510
|
|
5511 if (p == JUST_CALC_SIZE)
|
|
5512 return NULL;
|
|
5513
|
|
5514 offset = NEXT(p);
|
|
5515 if (offset == 0)
|
|
5516 return NULL;
|
|
5517
|
233
|
5518 if (OP(p) == BACK)
|
7
|
5519 return p - offset;
|
|
5520 else
|
|
5521 return p + offset;
|
|
5522 }
|
|
5523
|
|
5524 /*
|
|
5525 * Check the regexp program for its magic number.
|
|
5526 * Return TRUE if it's wrong.
|
|
5527 */
|
|
5528 static int
|
|
5529 prog_magic_wrong()
|
|
5530 {
|
|
5531 if (UCHARAT(REG_MULTI
|
|
5532 ? reg_mmatch->regprog->program
|
|
5533 : reg_match->regprog->program) != REGMAGIC)
|
|
5534 {
|
|
5535 EMSG(_(e_re_corr));
|
|
5536 return TRUE;
|
|
5537 }
|
|
5538 return FALSE;
|
|
5539 }
|
|
5540
|
|
5541 /*
|
|
5542 * Cleanup the subexpressions, if this wasn't done yet.
|
|
5543 * This construction is used to clear the subexpressions only when they are
|
|
5544 * used (to increase speed).
|
|
5545 */
|
|
5546 static void
|
|
5547 cleanup_subexpr()
|
|
5548 {
|
|
5549 if (need_clear_subexpr)
|
|
5550 {
|
|
5551 if (REG_MULTI)
|
|
5552 {
|
|
5553 /* Use 0xff to set lnum to -1 */
|
|
5554 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5555 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5556 }
|
|
5557 else
|
|
5558 {
|
|
5559 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5560 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5561 }
|
|
5562 need_clear_subexpr = FALSE;
|
|
5563 }
|
|
5564 }
|
|
5565
|
|
5566 #ifdef FEAT_SYN_HL
|
|
5567 static void
|
|
5568 cleanup_zsubexpr()
|
|
5569 {
|
|
5570 if (need_clear_zsubexpr)
|
|
5571 {
|
|
5572 if (REG_MULTI)
|
|
5573 {
|
|
5574 /* Use 0xff to set lnum to -1 */
|
|
5575 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5576 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5577 }
|
|
5578 else
|
|
5579 {
|
|
5580 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5581 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5582 }
|
|
5583 need_clear_zsubexpr = FALSE;
|
|
5584 }
|
|
5585 }
|
|
5586 #endif
|
|
5587
|
|
5588 /*
|
|
5589 * Advance reglnum, regline and reginput to the next line.
|
|
5590 */
|
|
5591 static void
|
|
5592 reg_nextline()
|
|
5593 {
|
|
5594 regline = reg_getline(++reglnum);
|
|
5595 reginput = regline;
|
|
5596 fast_breakcheck();
|
|
5597 }
|
|
5598
|
|
5599 /*
|
|
5600 * Save the input line and position in a regsave_T.
|
|
5601 */
|
|
5602 static void
|
233
|
5603 reg_save(save, gap)
|
7
|
5604 regsave_T *save;
|
233
|
5605 garray_T *gap;
|
7
|
5606 {
|
|
5607 if (REG_MULTI)
|
|
5608 {
|
|
5609 save->rs_u.pos.col = (colnr_T)(reginput - regline);
|
|
5610 save->rs_u.pos.lnum = reglnum;
|
|
5611 }
|
|
5612 else
|
|
5613 save->rs_u.ptr = reginput;
|
233
|
5614 save->rs_len = gap->ga_len;
|
7
|
5615 }
|
|
5616
|
|
5617 /*
|
|
5618 * Restore the input line and position from a regsave_T.
|
|
5619 */
|
|
5620 static void
|
233
|
5621 reg_restore(save, gap)
|
7
|
5622 regsave_T *save;
|
233
|
5623 garray_T *gap;
|
7
|
5624 {
|
|
5625 if (REG_MULTI)
|
|
5626 {
|
|
5627 if (reglnum != save->rs_u.pos.lnum)
|
|
5628 {
|
|
5629 /* only call reg_getline() when the line number changed to save
|
|
5630 * a bit of time */
|
|
5631 reglnum = save->rs_u.pos.lnum;
|
|
5632 regline = reg_getline(reglnum);
|
|
5633 }
|
|
5634 reginput = regline + save->rs_u.pos.col;
|
|
5635 }
|
|
5636 else
|
|
5637 reginput = save->rs_u.ptr;
|
233
|
5638 gap->ga_len = save->rs_len;
|
7
|
5639 }
|
|
5640
|
|
5641 /*
|
|
5642 * Return TRUE if current position is equal to saved position.
|
|
5643 */
|
|
5644 static int
|
|
5645 reg_save_equal(save)
|
|
5646 regsave_T *save;
|
|
5647 {
|
|
5648 if (REG_MULTI)
|
|
5649 return reglnum == save->rs_u.pos.lnum
|
|
5650 && reginput == regline + save->rs_u.pos.col;
|
|
5651 return reginput == save->rs_u.ptr;
|
|
5652 }
|
|
5653
|
|
5654 /*
|
|
5655 * Tentatively set the sub-expression start to the current position (after
|
|
5656 * calling regmatch() they will have changed). Need to save the existing
|
|
5657 * values for when there is no match.
|
|
5658 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
|
|
5659 * depending on REG_MULTI.
|
|
5660 */
|
|
5661 static void
|
|
5662 save_se_multi(savep, posp)
|
|
5663 save_se_T *savep;
|
|
5664 lpos_T *posp;
|
|
5665 {
|
|
5666 savep->se_u.pos = *posp;
|
|
5667 posp->lnum = reglnum;
|
|
5668 posp->col = (colnr_T)(reginput - regline);
|
|
5669 }
|
|
5670
|
|
5671 static void
|
|
5672 save_se_one(savep, pp)
|
|
5673 save_se_T *savep;
|
|
5674 char_u **pp;
|
|
5675 {
|
|
5676 savep->se_u.ptr = *pp;
|
|
5677 *pp = reginput;
|
|
5678 }
|
|
5679
|
|
5680 /*
|
|
5681 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
|
|
5682 */
|
|
5683 static int
|
|
5684 re_num_cmp(val, scan)
|
|
5685 long_u val;
|
|
5686 char_u *scan;
|
|
5687 {
|
|
5688 long_u n = OPERAND_MIN(scan);
|
|
5689
|
|
5690 if (OPERAND_CMP(scan) == '>')
|
|
5691 return val > n;
|
|
5692 if (OPERAND_CMP(scan) == '<')
|
|
5693 return val < n;
|
|
5694 return val == n;
|
|
5695 }
|
|
5696
|
|
5697
|
|
5698 #ifdef DEBUG
|
|
5699
|
|
5700 /*
|
|
5701 * regdump - dump a regexp onto stdout in vaguely comprehensible form
|
|
5702 */
|
|
5703 static void
|
|
5704 regdump(pattern, r)
|
|
5705 char_u *pattern;
|
|
5706 regprog_T *r;
|
|
5707 {
|
|
5708 char_u *s;
|
|
5709 int op = EXACTLY; /* Arbitrary non-END op. */
|
|
5710 char_u *next;
|
|
5711 char_u *end = NULL;
|
|
5712
|
|
5713 printf("\r\nregcomp(%s):\r\n", pattern);
|
|
5714
|
|
5715 s = r->program + 1;
|
|
5716 /*
|
|
5717 * Loop until we find the END that isn't before a referred next (an END
|
|
5718 * can also appear in a NOMATCH operand).
|
|
5719 */
|
|
5720 while (op != END || s <= end)
|
|
5721 {
|
|
5722 op = OP(s);
|
|
5723 printf("%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */
|
|
5724 next = regnext(s);
|
|
5725 if (next == NULL) /* Next ptr. */
|
|
5726 printf("(0)");
|
|
5727 else
|
|
5728 printf("(%d)", (int)((s - r->program) + (next - s)));
|
|
5729 if (end < next)
|
|
5730 end = next;
|
|
5731 if (op == BRACE_LIMITS)
|
|
5732 {
|
|
5733 /* Two short ints */
|
|
5734 printf(" minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
|
|
5735 s += 8;
|
|
5736 }
|
|
5737 s += 3;
|
|
5738 if (op == ANYOF || op == ANYOF + ADD_NL
|
|
5739 || op == ANYBUT || op == ANYBUT + ADD_NL
|
|
5740 || op == EXACTLY)
|
|
5741 {
|
|
5742 /* Literal string, where present. */
|
|
5743 while (*s != NUL)
|
|
5744 printf("%c", *s++);
|
|
5745 s++;
|
|
5746 }
|
|
5747 printf("\r\n");
|
|
5748 }
|
|
5749
|
|
5750 /* Header fields of interest. */
|
|
5751 if (r->regstart != NUL)
|
|
5752 printf("start `%s' 0x%x; ", r->regstart < 256
|
|
5753 ? (char *)transchar(r->regstart)
|
|
5754 : "multibyte", r->regstart);
|
|
5755 if (r->reganch)
|
|
5756 printf("anchored; ");
|
|
5757 if (r->regmust != NULL)
|
|
5758 printf("must have \"%s\"", r->regmust);
|
|
5759 printf("\r\n");
|
|
5760 }
|
|
5761
|
|
5762 /*
|
|
5763 * regprop - printable representation of opcode
|
|
5764 */
|
|
5765 static char_u *
|
|
5766 regprop(op)
|
|
5767 char_u *op;
|
|
5768 {
|
|
5769 char_u *p;
|
|
5770 static char_u buf[50];
|
|
5771
|
|
5772 (void) strcpy(buf, ":");
|
|
5773
|
|
5774 switch (OP(op))
|
|
5775 {
|
|
5776 case BOL:
|
|
5777 p = "BOL";
|
|
5778 break;
|
|
5779 case EOL:
|
|
5780 p = "EOL";
|
|
5781 break;
|
|
5782 case RE_BOF:
|
|
5783 p = "BOF";
|
|
5784 break;
|
|
5785 case RE_EOF:
|
|
5786 p = "EOF";
|
|
5787 break;
|
|
5788 case CURSOR:
|
|
5789 p = "CURSOR";
|
|
5790 break;
|
|
5791 case RE_LNUM:
|
|
5792 p = "RE_LNUM";
|
|
5793 break;
|
|
5794 case RE_COL:
|
|
5795 p = "RE_COL";
|
|
5796 break;
|
|
5797 case RE_VCOL:
|
|
5798 p = "RE_VCOL";
|
|
5799 break;
|
|
5800 case BOW:
|
|
5801 p = "BOW";
|
|
5802 break;
|
|
5803 case EOW:
|
|
5804 p = "EOW";
|
|
5805 break;
|
|
5806 case ANY:
|
|
5807 p = "ANY";
|
|
5808 break;
|
|
5809 case ANY + ADD_NL:
|
|
5810 p = "ANY+NL";
|
|
5811 break;
|
|
5812 case ANYOF:
|
|
5813 p = "ANYOF";
|
|
5814 break;
|
|
5815 case ANYOF + ADD_NL:
|
|
5816 p = "ANYOF+NL";
|
|
5817 break;
|
|
5818 case ANYBUT:
|
|
5819 p = "ANYBUT";
|
|
5820 break;
|
|
5821 case ANYBUT + ADD_NL:
|
|
5822 p = "ANYBUT+NL";
|
|
5823 break;
|
|
5824 case IDENT:
|
|
5825 p = "IDENT";
|
|
5826 break;
|
|
5827 case IDENT + ADD_NL:
|
|
5828 p = "IDENT+NL";
|
|
5829 break;
|
|
5830 case SIDENT:
|
|
5831 p = "SIDENT";
|
|
5832 break;
|
|
5833 case SIDENT + ADD_NL:
|
|
5834 p = "SIDENT+NL";
|
|
5835 break;
|
|
5836 case KWORD:
|
|
5837 p = "KWORD";
|
|
5838 break;
|
|
5839 case KWORD + ADD_NL:
|
|
5840 p = "KWORD+NL";
|
|
5841 break;
|
|
5842 case SKWORD:
|
|
5843 p = "SKWORD";
|
|
5844 break;
|
|
5845 case SKWORD + ADD_NL:
|
|
5846 p = "SKWORD+NL";
|
|
5847 break;
|
|
5848 case FNAME:
|
|
5849 p = "FNAME";
|
|
5850 break;
|
|
5851 case FNAME + ADD_NL:
|
|
5852 p = "FNAME+NL";
|
|
5853 break;
|
|
5854 case SFNAME:
|
|
5855 p = "SFNAME";
|
|
5856 break;
|
|
5857 case SFNAME + ADD_NL:
|
|
5858 p = "SFNAME+NL";
|
|
5859 break;
|
|
5860 case PRINT:
|
|
5861 p = "PRINT";
|
|
5862 break;
|
|
5863 case PRINT + ADD_NL:
|
|
5864 p = "PRINT+NL";
|
|
5865 break;
|
|
5866 case SPRINT:
|
|
5867 p = "SPRINT";
|
|
5868 break;
|
|
5869 case SPRINT + ADD_NL:
|
|
5870 p = "SPRINT+NL";
|
|
5871 break;
|
|
5872 case WHITE:
|
|
5873 p = "WHITE";
|
|
5874 break;
|
|
5875 case WHITE + ADD_NL:
|
|
5876 p = "WHITE+NL";
|
|
5877 break;
|
|
5878 case NWHITE:
|
|
5879 p = "NWHITE";
|
|
5880 break;
|
|
5881 case NWHITE + ADD_NL:
|
|
5882 p = "NWHITE+NL";
|
|
5883 break;
|
|
5884 case DIGIT:
|
|
5885 p = "DIGIT";
|
|
5886 break;
|
|
5887 case DIGIT + ADD_NL:
|
|
5888 p = "DIGIT+NL";
|
|
5889 break;
|
|
5890 case NDIGIT:
|
|
5891 p = "NDIGIT";
|
|
5892 break;
|
|
5893 case NDIGIT + ADD_NL:
|
|
5894 p = "NDIGIT+NL";
|
|
5895 break;
|
|
5896 case HEX:
|
|
5897 p = "HEX";
|
|
5898 break;
|
|
5899 case HEX + ADD_NL:
|
|
5900 p = "HEX+NL";
|
|
5901 break;
|
|
5902 case NHEX:
|
|
5903 p = "NHEX";
|
|
5904 break;
|
|
5905 case NHEX + ADD_NL:
|
|
5906 p = "NHEX+NL";
|
|
5907 break;
|
|
5908 case OCTAL:
|
|
5909 p = "OCTAL";
|
|
5910 break;
|
|
5911 case OCTAL + ADD_NL:
|
|
5912 p = "OCTAL+NL";
|
|
5913 break;
|
|
5914 case NOCTAL:
|
|
5915 p = "NOCTAL";
|
|
5916 break;
|
|
5917 case NOCTAL + ADD_NL:
|
|
5918 p = "NOCTAL+NL";
|
|
5919 break;
|
|
5920 case WORD:
|
|
5921 p = "WORD";
|
|
5922 break;
|
|
5923 case WORD + ADD_NL:
|
|
5924 p = "WORD+NL";
|
|
5925 break;
|
|
5926 case NWORD:
|
|
5927 p = "NWORD";
|
|
5928 break;
|
|
5929 case NWORD + ADD_NL:
|
|
5930 p = "NWORD+NL";
|
|
5931 break;
|
|
5932 case HEAD:
|
|
5933 p = "HEAD";
|
|
5934 break;
|
|
5935 case HEAD + ADD_NL:
|
|
5936 p = "HEAD+NL";
|
|
5937 break;
|
|
5938 case NHEAD:
|
|
5939 p = "NHEAD";
|
|
5940 break;
|
|
5941 case NHEAD + ADD_NL:
|
|
5942 p = "NHEAD+NL";
|
|
5943 break;
|
|
5944 case ALPHA:
|
|
5945 p = "ALPHA";
|
|
5946 break;
|
|
5947 case ALPHA + ADD_NL:
|
|
5948 p = "ALPHA+NL";
|
|
5949 break;
|
|
5950 case NALPHA:
|
|
5951 p = "NALPHA";
|
|
5952 break;
|
|
5953 case NALPHA + ADD_NL:
|
|
5954 p = "NALPHA+NL";
|
|
5955 break;
|
|
5956 case LOWER:
|
|
5957 p = "LOWER";
|
|
5958 break;
|
|
5959 case LOWER + ADD_NL:
|
|
5960 p = "LOWER+NL";
|
|
5961 break;
|
|
5962 case NLOWER:
|
|
5963 p = "NLOWER";
|
|
5964 break;
|
|
5965 case NLOWER + ADD_NL:
|
|
5966 p = "NLOWER+NL";
|
|
5967 break;
|
|
5968 case UPPER:
|
|
5969 p = "UPPER";
|
|
5970 break;
|
|
5971 case UPPER + ADD_NL:
|
|
5972 p = "UPPER+NL";
|
|
5973 break;
|
|
5974 case NUPPER:
|
|
5975 p = "NUPPER";
|
|
5976 break;
|
|
5977 case NUPPER + ADD_NL:
|
|
5978 p = "NUPPER+NL";
|
|
5979 break;
|
|
5980 case BRANCH:
|
|
5981 p = "BRANCH";
|
|
5982 break;
|
|
5983 case EXACTLY:
|
|
5984 p = "EXACTLY";
|
|
5985 break;
|
|
5986 case NOTHING:
|
|
5987 p = "NOTHING";
|
|
5988 break;
|
|
5989 case BACK:
|
|
5990 p = "BACK";
|
|
5991 break;
|
|
5992 case END:
|
|
5993 p = "END";
|
|
5994 break;
|
|
5995 case MOPEN + 0:
|
|
5996 p = "MATCH START";
|
|
5997 break;
|
|
5998 case MOPEN + 1:
|
|
5999 case MOPEN + 2:
|
|
6000 case MOPEN + 3:
|
|
6001 case MOPEN + 4:
|
|
6002 case MOPEN + 5:
|
|
6003 case MOPEN + 6:
|
|
6004 case MOPEN + 7:
|
|
6005 case MOPEN + 8:
|
|
6006 case MOPEN + 9:
|
|
6007 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
|
|
6008 p = NULL;
|
|
6009 break;
|
|
6010 case MCLOSE + 0:
|
|
6011 p = "MATCH END";
|
|
6012 break;
|
|
6013 case MCLOSE + 1:
|
|
6014 case MCLOSE + 2:
|
|
6015 case MCLOSE + 3:
|
|
6016 case MCLOSE + 4:
|
|
6017 case MCLOSE + 5:
|
|
6018 case MCLOSE + 6:
|
|
6019 case MCLOSE + 7:
|
|
6020 case MCLOSE + 8:
|
|
6021 case MCLOSE + 9:
|
|
6022 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
|
|
6023 p = NULL;
|
|
6024 break;
|
|
6025 case BACKREF + 1:
|
|
6026 case BACKREF + 2:
|
|
6027 case BACKREF + 3:
|
|
6028 case BACKREF + 4:
|
|
6029 case BACKREF + 5:
|
|
6030 case BACKREF + 6:
|
|
6031 case BACKREF + 7:
|
|
6032 case BACKREF + 8:
|
|
6033 case BACKREF + 9:
|
|
6034 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
|
|
6035 p = NULL;
|
|
6036 break;
|
|
6037 case NOPEN:
|
|
6038 p = "NOPEN";
|
|
6039 break;
|
|
6040 case NCLOSE:
|
|
6041 p = "NCLOSE";
|
|
6042 break;
|
|
6043 #ifdef FEAT_SYN_HL
|
|
6044 case ZOPEN + 1:
|
|
6045 case ZOPEN + 2:
|
|
6046 case ZOPEN + 3:
|
|
6047 case ZOPEN + 4:
|
|
6048 case ZOPEN + 5:
|
|
6049 case ZOPEN + 6:
|
|
6050 case ZOPEN + 7:
|
|
6051 case ZOPEN + 8:
|
|
6052 case ZOPEN + 9:
|
|
6053 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
|
|
6054 p = NULL;
|
|
6055 break;
|
|
6056 case ZCLOSE + 1:
|
|
6057 case ZCLOSE + 2:
|
|
6058 case ZCLOSE + 3:
|
|
6059 case ZCLOSE + 4:
|
|
6060 case ZCLOSE + 5:
|
|
6061 case ZCLOSE + 6:
|
|
6062 case ZCLOSE + 7:
|
|
6063 case ZCLOSE + 8:
|
|
6064 case ZCLOSE + 9:
|
|
6065 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
|
|
6066 p = NULL;
|
|
6067 break;
|
|
6068 case ZREF + 1:
|
|
6069 case ZREF + 2:
|
|
6070 case ZREF + 3:
|
|
6071 case ZREF + 4:
|
|
6072 case ZREF + 5:
|
|
6073 case ZREF + 6:
|
|
6074 case ZREF + 7:
|
|
6075 case ZREF + 8:
|
|
6076 case ZREF + 9:
|
|
6077 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
|
|
6078 p = NULL;
|
|
6079 break;
|
|
6080 #endif
|
|
6081 case STAR:
|
|
6082 p = "STAR";
|
|
6083 break;
|
|
6084 case PLUS:
|
|
6085 p = "PLUS";
|
|
6086 break;
|
|
6087 case NOMATCH:
|
|
6088 p = "NOMATCH";
|
|
6089 break;
|
|
6090 case MATCH:
|
|
6091 p = "MATCH";
|
|
6092 break;
|
|
6093 case BEHIND:
|
|
6094 p = "BEHIND";
|
|
6095 break;
|
|
6096 case NOBEHIND:
|
|
6097 p = "NOBEHIND";
|
|
6098 break;
|
|
6099 case SUBPAT:
|
|
6100 p = "SUBPAT";
|
|
6101 break;
|
|
6102 case BRACE_LIMITS:
|
|
6103 p = "BRACE_LIMITS";
|
|
6104 break;
|
|
6105 case BRACE_SIMPLE:
|
|
6106 p = "BRACE_SIMPLE";
|
|
6107 break;
|
|
6108 case BRACE_COMPLEX + 0:
|
|
6109 case BRACE_COMPLEX + 1:
|
|
6110 case BRACE_COMPLEX + 2:
|
|
6111 case BRACE_COMPLEX + 3:
|
|
6112 case BRACE_COMPLEX + 4:
|
|
6113 case BRACE_COMPLEX + 5:
|
|
6114 case BRACE_COMPLEX + 6:
|
|
6115 case BRACE_COMPLEX + 7:
|
|
6116 case BRACE_COMPLEX + 8:
|
|
6117 case BRACE_COMPLEX + 9:
|
|
6118 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
|
|
6119 p = NULL;
|
|
6120 break;
|
|
6121 #ifdef FEAT_MBYTE
|
|
6122 case MULTIBYTECODE:
|
|
6123 p = "MULTIBYTECODE";
|
|
6124 break;
|
|
6125 #endif
|
|
6126 case NEWL:
|
|
6127 p = "NEWL";
|
|
6128 break;
|
|
6129 default:
|
|
6130 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
|
|
6131 p = NULL;
|
|
6132 break;
|
|
6133 }
|
|
6134 if (p != NULL)
|
|
6135 (void) strcat(buf, p);
|
|
6136 return buf;
|
|
6137 }
|
|
6138 #endif
|
|
6139
|
|
6140 #ifdef FEAT_MBYTE
|
|
6141 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3));
|
|
6142
|
|
6143 typedef struct
|
|
6144 {
|
|
6145 int a, b, c;
|
|
6146 } decomp_T;
|
|
6147
|
|
6148
|
|
6149 /* 0xfb20 - 0xfb4f */
|
297
|
6150 static decomp_T decomp_table[0xfb4f-0xfb20+1] =
|
7
|
6151 {
|
|
6152 {0x5e2,0,0}, /* 0xfb20 alt ayin */
|
|
6153 {0x5d0,0,0}, /* 0xfb21 alt alef */
|
|
6154 {0x5d3,0,0}, /* 0xfb22 alt dalet */
|
|
6155 {0x5d4,0,0}, /* 0xfb23 alt he */
|
|
6156 {0x5db,0,0}, /* 0xfb24 alt kaf */
|
|
6157 {0x5dc,0,0}, /* 0xfb25 alt lamed */
|
|
6158 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */
|
|
6159 {0x5e8,0,0}, /* 0xfb27 alt resh */
|
|
6160 {0x5ea,0,0}, /* 0xfb28 alt tav */
|
|
6161 {'+', 0, 0}, /* 0xfb29 alt plus */
|
|
6162 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */
|
|
6163 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */
|
|
6164 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */
|
|
6165 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */
|
|
6166 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */
|
|
6167 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */
|
|
6168 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */
|
|
6169 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */
|
|
6170 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */
|
|
6171 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */
|
|
6172 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */
|
|
6173 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */
|
|
6174 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */
|
|
6175 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */
|
|
6176 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */
|
|
6177 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */
|
|
6178 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */
|
|
6179 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */
|
|
6180 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */
|
|
6181 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */
|
|
6182 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */
|
|
6183 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */
|
|
6184 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */
|
|
6185 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */
|
|
6186 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */
|
|
6187 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */
|
|
6188 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */
|
|
6189 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */
|
|
6190 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */
|
|
6191 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */
|
|
6192 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */
|
|
6193 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */
|
|
6194 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */
|
|
6195 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */
|
|
6196 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */
|
|
6197 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */
|
|
6198 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */
|
|
6199 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */
|
|
6200 };
|
|
6201
|
|
6202 static void
|
|
6203 mb_decompose(c, c1, c2, c3)
|
|
6204 int c, *c1, *c2, *c3;
|
|
6205 {
|
|
6206 decomp_T d;
|
|
6207
|
|
6208 if (c >= 0x4b20 && c <= 0xfb4f)
|
|
6209 {
|
|
6210 d = decomp_table[c - 0xfb20];
|
|
6211 *c1 = d.a;
|
|
6212 *c2 = d.b;
|
|
6213 *c3 = d.c;
|
|
6214 }
|
|
6215 else
|
|
6216 {
|
|
6217 *c1 = c;
|
|
6218 *c2 = *c3 = 0;
|
|
6219 }
|
|
6220 }
|
|
6221 #endif
|
|
6222
|
|
6223 /*
|
|
6224 * Compare two strings, ignore case if ireg_ic set.
|
|
6225 * Return 0 if strings match, non-zero otherwise.
|
|
6226 * Correct the length "*n" when composing characters are ignored.
|
|
6227 */
|
|
6228 static int
|
|
6229 cstrncmp(s1, s2, n)
|
|
6230 char_u *s1, *s2;
|
|
6231 int *n;
|
|
6232 {
|
|
6233 int result;
|
|
6234
|
|
6235 if (!ireg_ic)
|
|
6236 result = STRNCMP(s1, s2, *n);
|
|
6237 else
|
|
6238 result = MB_STRNICMP(s1, s2, *n);
|
|
6239
|
|
6240 #ifdef FEAT_MBYTE
|
|
6241 /* if it failed and it's utf8 and we want to combineignore: */
|
|
6242 if (result != 0 && enc_utf8 && ireg_icombine)
|
|
6243 {
|
|
6244 char_u *str1, *str2;
|
|
6245 int c1, c2, c11, c12;
|
|
6246 int junk;
|
|
6247
|
|
6248 /* we have to handle the strcmp ourselves, since it is necessary to
|
|
6249 * deal with the composing characters by ignoring them: */
|
|
6250 str1 = s1;
|
|
6251 str2 = s2;
|
|
6252 c1 = c2 = 0;
|
507
|
6253 while ((int)(str1 - s1) < *n)
|
7
|
6254 {
|
|
6255 c1 = mb_ptr2char_adv(&str1);
|
|
6256 c2 = mb_ptr2char_adv(&str2);
|
|
6257
|
|
6258 /* decompose the character if necessary, into 'base' characters
|
|
6259 * because I don't care about Arabic, I will hard-code the Hebrew
|
|
6260 * which I *do* care about! So sue me... */
|
|
6261 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2)))
|
|
6262 {
|
|
6263 /* decomposition necessary? */
|
|
6264 mb_decompose(c1, &c11, &junk, &junk);
|
|
6265 mb_decompose(c2, &c12, &junk, &junk);
|
|
6266 c1 = c11;
|
|
6267 c2 = c12;
|
|
6268 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12)))
|
|
6269 break;
|
|
6270 }
|
|
6271 }
|
|
6272 result = c2 - c1;
|
|
6273 if (result == 0)
|
|
6274 *n = (int)(str2 - s2);
|
|
6275 }
|
|
6276 #endif
|
|
6277
|
|
6278 return result;
|
|
6279 }
|
|
6280
|
|
6281 /*
|
|
6282 * cstrchr: This function is used a lot for simple searches, keep it fast!
|
|
6283 */
|
|
6284 static char_u *
|
|
6285 cstrchr(s, c)
|
|
6286 char_u *s;
|
|
6287 int c;
|
|
6288 {
|
|
6289 char_u *p;
|
|
6290 int cc;
|
|
6291
|
|
6292 if (!ireg_ic
|
|
6293 #ifdef FEAT_MBYTE
|
|
6294 || (!enc_utf8 && mb_char2len(c) > 1)
|
|
6295 #endif
|
|
6296 )
|
|
6297 return vim_strchr(s, c);
|
|
6298
|
|
6299 /* tolower() and toupper() can be slow, comparing twice should be a lot
|
|
6300 * faster (esp. when using MS Visual C++!).
|
|
6301 * For UTF-8 need to use folded case. */
|
|
6302 #ifdef FEAT_MBYTE
|
|
6303 if (enc_utf8 && c > 0x80)
|
|
6304 cc = utf_fold(c);
|
|
6305 else
|
|
6306 #endif
|
|
6307 if (isupper(c))
|
|
6308 cc = TOLOWER_LOC(c);
|
|
6309 else if (islower(c))
|
|
6310 cc = TOUPPER_LOC(c);
|
|
6311 else
|
|
6312 return vim_strchr(s, c);
|
|
6313
|
|
6314 #ifdef FEAT_MBYTE
|
|
6315 if (has_mbyte)
|
|
6316 {
|
474
|
6317 for (p = s; *p != NUL; p += (*mb_ptr2len)(p))
|
7
|
6318 {
|
|
6319 if (enc_utf8 && c > 0x80)
|
|
6320 {
|
|
6321 if (utf_fold(utf_ptr2char(p)) == cc)
|
|
6322 return p;
|
|
6323 }
|
|
6324 else if (*p == c || *p == cc)
|
|
6325 return p;
|
|
6326 }
|
|
6327 }
|
|
6328 else
|
|
6329 #endif
|
|
6330 /* Faster version for when there are no multi-byte characters. */
|
|
6331 for (p = s; *p != NUL; ++p)
|
|
6332 if (*p == c || *p == cc)
|
|
6333 return p;
|
|
6334
|
|
6335 return NULL;
|
|
6336 }
|
|
6337
|
|
6338 /***************************************************************
|
|
6339 * regsub stuff *
|
|
6340 ***************************************************************/
|
|
6341
|
|
6342 /* This stuff below really confuses cc on an SGI -- webb */
|
|
6343 #ifdef __sgi
|
|
6344 # undef __ARGS
|
|
6345 # define __ARGS(x) ()
|
|
6346 #endif
|
|
6347
|
|
6348 /*
|
|
6349 * We should define ftpr as a pointer to a function returning a pointer to
|
|
6350 * a function returning a pointer to a function ...
|
|
6351 * This is impossible, so we declare a pointer to a function returning a
|
|
6352 * pointer to a function returning void. This should work for all compilers.
|
|
6353 */
|
|
6354 typedef void (*(*fptr) __ARGS((char_u *, int)))();
|
|
6355
|
|
6356 static fptr do_upper __ARGS((char_u *, int));
|
|
6357 static fptr do_Upper __ARGS((char_u *, int));
|
|
6358 static fptr do_lower __ARGS((char_u *, int));
|
|
6359 static fptr do_Lower __ARGS((char_u *, int));
|
|
6360
|
|
6361 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash));
|
|
6362
|
|
6363 static fptr
|
|
6364 do_upper(d, c)
|
|
6365 char_u *d;
|
|
6366 int c;
|
|
6367 {
|
|
6368 *d = TOUPPER_LOC(c);
|
|
6369
|
|
6370 return (fptr)NULL;
|
|
6371 }
|
|
6372
|
|
6373 static fptr
|
|
6374 do_Upper(d, c)
|
|
6375 char_u *d;
|
|
6376 int c;
|
|
6377 {
|
|
6378 *d = TOUPPER_LOC(c);
|
|
6379
|
|
6380 return (fptr)do_Upper;
|
|
6381 }
|
|
6382
|
|
6383 static fptr
|
|
6384 do_lower(d, c)
|
|
6385 char_u *d;
|
|
6386 int c;
|
|
6387 {
|
|
6388 *d = TOLOWER_LOC(c);
|
|
6389
|
|
6390 return (fptr)NULL;
|
|
6391 }
|
|
6392
|
|
6393 static fptr
|
|
6394 do_Lower(d, c)
|
|
6395 char_u *d;
|
|
6396 int c;
|
|
6397 {
|
|
6398 *d = TOLOWER_LOC(c);
|
|
6399
|
|
6400 return (fptr)do_Lower;
|
|
6401 }
|
|
6402
|
|
6403 /*
|
|
6404 * regtilde(): Replace tildes in the pattern by the old pattern.
|
|
6405 *
|
|
6406 * Short explanation of the tilde: It stands for the previous replacement
|
|
6407 * pattern. If that previous pattern also contains a ~ we should go back a
|
|
6408 * step further... But we insert the previous pattern into the current one
|
|
6409 * and remember that.
|
|
6410 * This still does not handle the case where "magic" changes. TODO?
|
|
6411 *
|
|
6412 * The tildes are parsed once before the first call to vim_regsub().
|
|
6413 */
|
|
6414 char_u *
|
|
6415 regtilde(source, magic)
|
|
6416 char_u *source;
|
|
6417 int magic;
|
|
6418 {
|
|
6419 char_u *newsub = source;
|
|
6420 char_u *tmpsub;
|
|
6421 char_u *p;
|
|
6422 int len;
|
|
6423 int prevlen;
|
|
6424
|
|
6425 for (p = newsub; *p; ++p)
|
|
6426 {
|
|
6427 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic))
|
|
6428 {
|
|
6429 if (reg_prev_sub != NULL)
|
|
6430 {
|
|
6431 /* length = len(newsub) - 1 + len(prev_sub) + 1 */
|
|
6432 prevlen = (int)STRLEN(reg_prev_sub);
|
|
6433 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen));
|
|
6434 if (tmpsub != NULL)
|
|
6435 {
|
|
6436 /* copy prefix */
|
|
6437 len = (int)(p - newsub); /* not including ~ */
|
|
6438 mch_memmove(tmpsub, newsub, (size_t)len);
|
|
6439 /* interpretate tilde */
|
|
6440 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen);
|
|
6441 /* copy postfix */
|
|
6442 if (!magic)
|
|
6443 ++p; /* back off \ */
|
|
6444 STRCPY(tmpsub + len + prevlen, p + 1);
|
|
6445
|
|
6446 if (newsub != source) /* already allocated newsub */
|
|
6447 vim_free(newsub);
|
|
6448 newsub = tmpsub;
|
|
6449 p = newsub + len + prevlen;
|
|
6450 }
|
|
6451 }
|
|
6452 else if (magic)
|
|
6453 STRCPY(p, p + 1); /* remove '~' */
|
|
6454 else
|
|
6455 STRCPY(p, p + 2); /* remove '\~' */
|
|
6456 --p;
|
|
6457 }
|
|
6458 else
|
|
6459 {
|
|
6460 if (*p == '\\' && p[1]) /* skip escaped characters */
|
|
6461 ++p;
|
|
6462 #ifdef FEAT_MBYTE
|
|
6463 if (has_mbyte)
|
474
|
6464 p += (*mb_ptr2len)(p) - 1;
|
7
|
6465 #endif
|
|
6466 }
|
|
6467 }
|
|
6468
|
|
6469 vim_free(reg_prev_sub);
|
|
6470 if (newsub != source) /* newsub was allocated, just keep it */
|
|
6471 reg_prev_sub = newsub;
|
|
6472 else /* no ~ found, need to save newsub */
|
|
6473 reg_prev_sub = vim_strsave(newsub);
|
|
6474 return newsub;
|
|
6475 }
|
|
6476
|
|
6477 #ifdef FEAT_EVAL
|
|
6478 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */
|
|
6479
|
|
6480 /* These pointers are used instead of reg_match and reg_mmatch for
|
|
6481 * reg_submatch(). Needed for when the substitution string is an expression
|
|
6482 * that contains a call to substitute() and submatch(). */
|
|
6483 static regmatch_T *submatch_match;
|
|
6484 static regmmatch_T *submatch_mmatch;
|
|
6485 #endif
|
|
6486
|
|
6487 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO)
|
|
6488 /*
|
|
6489 * vim_regsub() - perform substitutions after a vim_regexec() or
|
|
6490 * vim_regexec_multi() match.
|
|
6491 *
|
|
6492 * If "copy" is TRUE really copy into "dest".
|
|
6493 * If "copy" is FALSE nothing is copied, this is just to find out the length
|
|
6494 * of the result.
|
|
6495 *
|
|
6496 * If "backslash" is TRUE, a backslash will be removed later, need to double
|
|
6497 * them to keep them, and insert a backslash before a CR to avoid it being
|
|
6498 * replaced with a line break later.
|
|
6499 *
|
|
6500 * Note: The matched text must not change between the call of
|
|
6501 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back
|
|
6502 * references invalid!
|
|
6503 *
|
|
6504 * Returns the size of the replacement, including terminating NUL.
|
|
6505 */
|
|
6506 int
|
|
6507 vim_regsub(rmp, source, dest, copy, magic, backslash)
|
|
6508 regmatch_T *rmp;
|
|
6509 char_u *source;
|
|
6510 char_u *dest;
|
|
6511 int copy;
|
|
6512 int magic;
|
|
6513 int backslash;
|
|
6514 {
|
|
6515 reg_match = rmp;
|
|
6516 reg_mmatch = NULL;
|
|
6517 reg_maxline = 0;
|
|
6518 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
6519 }
|
|
6520 #endif
|
|
6521
|
|
6522 int
|
|
6523 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash)
|
|
6524 regmmatch_T *rmp;
|
|
6525 linenr_T lnum;
|
|
6526 char_u *source;
|
|
6527 char_u *dest;
|
|
6528 int copy;
|
|
6529 int magic;
|
|
6530 int backslash;
|
|
6531 {
|
|
6532 reg_match = NULL;
|
|
6533 reg_mmatch = rmp;
|
|
6534 reg_buf = curbuf; /* always works on the current buffer! */
|
|
6535 reg_firstlnum = lnum;
|
|
6536 reg_maxline = curbuf->b_ml.ml_line_count - lnum;
|
|
6537 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
6538 }
|
|
6539
|
|
6540 static int
|
|
6541 vim_regsub_both(source, dest, copy, magic, backslash)
|
|
6542 char_u *source;
|
|
6543 char_u *dest;
|
|
6544 int copy;
|
|
6545 int magic;
|
|
6546 int backslash;
|
|
6547 {
|
|
6548 char_u *src;
|
|
6549 char_u *dst;
|
|
6550 char_u *s;
|
|
6551 int c;
|
|
6552 int no = -1;
|
|
6553 fptr func = (fptr)NULL;
|
|
6554 linenr_T clnum = 0; /* init for GCC */
|
|
6555 int len = 0; /* init for GCC */
|
|
6556 #ifdef FEAT_EVAL
|
|
6557 static char_u *eval_result = NULL;
|
|
6558 #endif
|
|
6559 #ifdef FEAT_MBYTE
|
|
6560 int l;
|
|
6561 #endif
|
|
6562
|
|
6563
|
|
6564 /* Be paranoid... */
|
|
6565 if (source == NULL || dest == NULL)
|
|
6566 {
|
|
6567 EMSG(_(e_null));
|
|
6568 return 0;
|
|
6569 }
|
|
6570 if (prog_magic_wrong())
|
|
6571 return 0;
|
|
6572 src = source;
|
|
6573 dst = dest;
|
|
6574
|
|
6575 /*
|
|
6576 * When the substitute part starts with "\=" evaluate it as an expression.
|
|
6577 */
|
|
6578 if (source[0] == '\\' && source[1] == '='
|
|
6579 #ifdef FEAT_EVAL
|
|
6580 && !can_f_submatch /* can't do this recursively */
|
|
6581 #endif
|
|
6582 )
|
|
6583 {
|
|
6584 #ifdef FEAT_EVAL
|
|
6585 /* To make sure that the length doesn't change between checking the
|
|
6586 * length and copying the string, and to speed up things, the
|
|
6587 * resulting string is saved from the call with "copy" == FALSE to the
|
|
6588 * call with "copy" == TRUE. */
|
|
6589 if (copy)
|
|
6590 {
|
|
6591 if (eval_result != NULL)
|
|
6592 {
|
|
6593 STRCPY(dest, eval_result);
|
|
6594 dst += STRLEN(eval_result);
|
|
6595 vim_free(eval_result);
|
|
6596 eval_result = NULL;
|
|
6597 }
|
|
6598 }
|
|
6599 else
|
|
6600 {
|
|
6601 linenr_T save_reg_maxline;
|
|
6602 win_T *save_reg_win;
|
|
6603 int save_ireg_ic;
|
|
6604
|
|
6605 vim_free(eval_result);
|
|
6606
|
|
6607 /* The expression may contain substitute(), which calls us
|
|
6608 * recursively. Make sure submatch() gets the text from the first
|
|
6609 * level. Don't need to save "reg_buf", because
|
|
6610 * vim_regexec_multi() can't be called recursively. */
|
|
6611 submatch_match = reg_match;
|
|
6612 submatch_mmatch = reg_mmatch;
|
|
6613 save_reg_maxline = reg_maxline;
|
|
6614 save_reg_win = reg_win;
|
|
6615 save_ireg_ic = ireg_ic;
|
|
6616 can_f_submatch = TRUE;
|
|
6617
|
|
6618 eval_result = eval_to_string(source + 2, NULL);
|
|
6619 if (eval_result != NULL)
|
|
6620 {
|
39
|
6621 for (s = eval_result; *s != NUL; mb_ptr_adv(s))
|
7
|
6622 {
|
|
6623 /* Change NL to CR, so that it becomes a line break.
|
|
6624 * Skip over a backslashed character. */
|
|
6625 if (*s == NL)
|
|
6626 *s = CAR;
|
|
6627 else if (*s == '\\' && s[1] != NUL)
|
|
6628 ++s;
|
|
6629 }
|
|
6630
|
|
6631 dst += STRLEN(eval_result);
|
|
6632 }
|
|
6633
|
|
6634 reg_match = submatch_match;
|
|
6635 reg_mmatch = submatch_mmatch;
|
|
6636 reg_maxline = save_reg_maxline;
|
|
6637 reg_win = save_reg_win;
|
|
6638 ireg_ic = save_ireg_ic;
|
|
6639 can_f_submatch = FALSE;
|
|
6640 }
|
|
6641 #endif
|
|
6642 }
|
|
6643 else
|
|
6644 while ((c = *src++) != NUL)
|
|
6645 {
|
|
6646 if (c == '&' && magic)
|
|
6647 no = 0;
|
|
6648 else if (c == '\\' && *src != NUL)
|
|
6649 {
|
|
6650 if (*src == '&' && !magic)
|
|
6651 {
|
|
6652 ++src;
|
|
6653 no = 0;
|
|
6654 }
|
|
6655 else if ('0' <= *src && *src <= '9')
|
|
6656 {
|
|
6657 no = *src++ - '0';
|
|
6658 }
|
|
6659 else if (vim_strchr((char_u *)"uUlLeE", *src))
|
|
6660 {
|
|
6661 switch (*src++)
|
|
6662 {
|
|
6663 case 'u': func = (fptr)do_upper;
|
|
6664 continue;
|
|
6665 case 'U': func = (fptr)do_Upper;
|
|
6666 continue;
|
|
6667 case 'l': func = (fptr)do_lower;
|
|
6668 continue;
|
|
6669 case 'L': func = (fptr)do_Lower;
|
|
6670 continue;
|
|
6671 case 'e':
|
|
6672 case 'E': func = (fptr)NULL;
|
|
6673 continue;
|
|
6674 }
|
|
6675 }
|
|
6676 }
|
|
6677 if (no < 0) /* Ordinary character. */
|
|
6678 {
|
|
6679 if (c == '\\' && *src != NUL)
|
|
6680 {
|
|
6681 /* Check for abbreviations -- webb */
|
|
6682 switch (*src)
|
|
6683 {
|
|
6684 case 'r': c = CAR; ++src; break;
|
|
6685 case 'n': c = NL; ++src; break;
|
|
6686 case 't': c = TAB; ++src; break;
|
|
6687 /* Oh no! \e already has meaning in subst pat :-( */
|
|
6688 /* case 'e': c = ESC; ++src; break; */
|
|
6689 case 'b': c = Ctrl_H; ++src; break;
|
|
6690
|
|
6691 /* If "backslash" is TRUE the backslash will be removed
|
|
6692 * later. Used to insert a literal CR. */
|
|
6693 default: if (backslash)
|
|
6694 {
|
|
6695 if (copy)
|
|
6696 *dst = '\\';
|
|
6697 ++dst;
|
|
6698 }
|
|
6699 c = *src++;
|
|
6700 }
|
|
6701 }
|
|
6702
|
|
6703 /* Write to buffer, if copy is set. */
|
|
6704 #ifdef FEAT_MBYTE
|
474
|
6705 if (has_mbyte && (l = (*mb_ptr2len)(src - 1)) > 1)
|
7
|
6706 {
|
|
6707 /* TODO: should use "func" here. */
|
|
6708 if (copy)
|
|
6709 mch_memmove(dst, src - 1, l);
|
|
6710 dst += l - 1;
|
|
6711 src += l - 1;
|
|
6712 }
|
|
6713 else
|
|
6714 {
|
|
6715 #endif
|
|
6716 if (copy)
|
|
6717 {
|
|
6718 if (func == (fptr)NULL) /* just copy */
|
|
6719 *dst = c;
|
|
6720 else /* change case */
|
|
6721 func = (fptr)(func(dst, c));
|
|
6722 /* Turbo C complains without the typecast */
|
|
6723 }
|
|
6724 #ifdef FEAT_MBYTE
|
|
6725 }
|
|
6726 #endif
|
|
6727 dst++;
|
|
6728 }
|
|
6729 else
|
|
6730 {
|
|
6731 if (REG_MULTI)
|
|
6732 {
|
|
6733 clnum = reg_mmatch->startpos[no].lnum;
|
|
6734 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0)
|
|
6735 s = NULL;
|
|
6736 else
|
|
6737 {
|
|
6738 s = reg_getline(clnum) + reg_mmatch->startpos[no].col;
|
|
6739 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6740 len = reg_mmatch->endpos[no].col
|
|
6741 - reg_mmatch->startpos[no].col;
|
|
6742 else
|
|
6743 len = (int)STRLEN(s);
|
|
6744 }
|
|
6745 }
|
|
6746 else
|
|
6747 {
|
|
6748 s = reg_match->startp[no];
|
|
6749 if (reg_match->endp[no] == NULL)
|
|
6750 s = NULL;
|
|
6751 else
|
|
6752 len = (int)(reg_match->endp[no] - s);
|
|
6753 }
|
|
6754 if (s != NULL)
|
|
6755 {
|
|
6756 for (;;)
|
|
6757 {
|
|
6758 if (len == 0)
|
|
6759 {
|
|
6760 if (REG_MULTI)
|
|
6761 {
|
|
6762 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6763 break;
|
|
6764 if (copy)
|
|
6765 *dst = CAR;
|
|
6766 ++dst;
|
|
6767 s = reg_getline(++clnum);
|
|
6768 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6769 len = reg_mmatch->endpos[no].col;
|
|
6770 else
|
|
6771 len = (int)STRLEN(s);
|
|
6772 }
|
|
6773 else
|
|
6774 break;
|
|
6775 }
|
|
6776 else if (*s == NUL) /* we hit NUL. */
|
|
6777 {
|
|
6778 if (copy)
|
|
6779 EMSG(_(e_re_damg));
|
|
6780 goto exit;
|
|
6781 }
|
|
6782 else
|
|
6783 {
|
|
6784 if (backslash && (*s == CAR || *s == '\\'))
|
|
6785 {
|
|
6786 /*
|
|
6787 * Insert a backslash in front of a CR, otherwise
|
|
6788 * it will be replaced by a line break.
|
|
6789 * Number of backslashes will be halved later,
|
|
6790 * double them here.
|
|
6791 */
|
|
6792 if (copy)
|
|
6793 {
|
|
6794 dst[0] = '\\';
|
|
6795 dst[1] = *s;
|
|
6796 }
|
|
6797 dst += 2;
|
|
6798 }
|
|
6799 #ifdef FEAT_MBYTE
|
474
|
6800 else if (has_mbyte && (l = (*mb_ptr2len)(s)) > 1)
|
7
|
6801 {
|
|
6802 /* TODO: should use "func" here. */
|
|
6803 if (copy)
|
|
6804 mch_memmove(dst, s, l);
|
|
6805 dst += l;
|
|
6806 s += l - 1;
|
|
6807 len -= l - 1;
|
|
6808 }
|
|
6809 #endif
|
|
6810 else
|
|
6811 {
|
|
6812 if (copy)
|
|
6813 {
|
|
6814 if (func == (fptr)NULL) /* just copy */
|
|
6815 *dst = *s;
|
|
6816 else /* change case */
|
|
6817 func = (fptr)(func(dst, *s));
|
|
6818 /* Turbo C complains without the typecast */
|
|
6819 }
|
|
6820 ++dst;
|
|
6821 }
|
|
6822 ++s;
|
|
6823 --len;
|
|
6824 }
|
|
6825 }
|
|
6826 }
|
|
6827 no = -1;
|
|
6828 }
|
|
6829 }
|
|
6830 if (copy)
|
|
6831 *dst = NUL;
|
|
6832
|
|
6833 exit:
|
|
6834 return (int)((dst - dest) + 1);
|
|
6835 }
|
|
6836
|
|
6837 #ifdef FEAT_EVAL
|
|
6838 /*
|
|
6839 * Used for the submatch() function: get the string from tne n'th submatch in
|
|
6840 * allocated memory.
|
|
6841 * Returns NULL when not in a ":s" command and for a non-existing submatch.
|
|
6842 */
|
|
6843 char_u *
|
|
6844 reg_submatch(no)
|
|
6845 int no;
|
|
6846 {
|
|
6847 char_u *retval = NULL;
|
|
6848 char_u *s;
|
|
6849 int len;
|
|
6850 int round;
|
|
6851 linenr_T lnum;
|
|
6852
|
|
6853 if (!can_f_submatch)
|
|
6854 return NULL;
|
|
6855
|
|
6856 if (submatch_match == NULL)
|
|
6857 {
|
|
6858 /*
|
|
6859 * First round: compute the length and allocate memory.
|
|
6860 * Second round: copy the text.
|
|
6861 */
|
|
6862 for (round = 1; round <= 2; ++round)
|
|
6863 {
|
|
6864 lnum = submatch_mmatch->startpos[no].lnum;
|
|
6865 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0)
|
|
6866 return NULL;
|
|
6867
|
|
6868 s = reg_getline(lnum) + submatch_mmatch->startpos[no].col;
|
|
6869 if (s == NULL) /* anti-crash check, cannot happen? */
|
|
6870 break;
|
|
6871 if (submatch_mmatch->endpos[no].lnum == lnum)
|
|
6872 {
|
|
6873 /* Within one line: take form start to end col. */
|
|
6874 len = submatch_mmatch->endpos[no].col
|
|
6875 - submatch_mmatch->startpos[no].col;
|
|
6876 if (round == 2)
|
418
|
6877 vim_strncpy(retval, s, len);
|
7
|
6878 ++len;
|
|
6879 }
|
|
6880 else
|
|
6881 {
|
|
6882 /* Multiple lines: take start line from start col, middle
|
|
6883 * lines completely and end line up to end col. */
|
|
6884 len = (int)STRLEN(s);
|
|
6885 if (round == 2)
|
|
6886 {
|
|
6887 STRCPY(retval, s);
|
|
6888 retval[len] = '\n';
|
|
6889 }
|
|
6890 ++len;
|
|
6891 ++lnum;
|
|
6892 while (lnum < submatch_mmatch->endpos[no].lnum)
|
|
6893 {
|
|
6894 s = reg_getline(lnum++);
|
|
6895 if (round == 2)
|
|
6896 STRCPY(retval + len, s);
|
|
6897 len += (int)STRLEN(s);
|
|
6898 if (round == 2)
|
|
6899 retval[len] = '\n';
|
|
6900 ++len;
|
|
6901 }
|
|
6902 if (round == 2)
|
|
6903 STRNCPY(retval + len, reg_getline(lnum),
|
|
6904 submatch_mmatch->endpos[no].col);
|
|
6905 len += submatch_mmatch->endpos[no].col;
|
|
6906 if (round == 2)
|
|
6907 retval[len] = NUL;
|
|
6908 ++len;
|
|
6909 }
|
|
6910
|
|
6911 if (round == 1)
|
|
6912 {
|
|
6913 retval = lalloc((long_u)len, TRUE);
|
|
6914 if (s == NULL)
|
|
6915 return NULL;
|
|
6916 }
|
|
6917 }
|
|
6918 }
|
|
6919 else
|
|
6920 {
|
|
6921 if (submatch_match->endp[no] == NULL)
|
|
6922 retval = NULL;
|
|
6923 else
|
|
6924 {
|
|
6925 s = submatch_match->startp[no];
|
|
6926 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s));
|
|
6927 }
|
|
6928 }
|
|
6929
|
|
6930 return retval;
|
|
6931 }
|
|
6932 #endif
|