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 |
|
|
103 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END
|
|
104 * | | ^ ^
|
|
105 * | +----------+ |
|
|
106 * +-------------------------------------------------+
|
|
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 */
|
|
325 #define EMSG_RET_NULL(m) { EMSG(m); rc_did_emsg = TRUE; return NULL; }
|
|
326 #define EMSG_M_RET_NULL(m, c) { EMSG2(m, c ? "" : "\\"); rc_did_emsg = TRUE; return NULL; }
|
|
327 #define EMSG_RET_FAIL(m) { EMSG(m); rc_did_emsg = TRUE; return FAIL; }
|
|
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
|
|
376 static char_u *reg_prev_sub;
|
|
377
|
|
378 /*
|
|
379 * REGEXP_INRANGE contains all characters which are always special in a []
|
|
380 * range after '\'.
|
|
381 * REGEXP_ABBR contains all characters which act as abbreviations after '\'.
|
|
382 * These are:
|
|
383 * \n - New line (NL).
|
|
384 * \r - Carriage Return (CR).
|
|
385 * \t - Tab (TAB).
|
|
386 * \e - Escape (ESC).
|
|
387 * \b - Backspace (Ctrl_H).
|
24
|
388 * \d - Character code in decimal, eg \d123
|
|
389 * \o - Character code in octal, eg \o80
|
|
390 * \x - Character code in hex, eg \x4a
|
|
391 * \u - Multibyte character code, eg \u20ac
|
|
392 * \U - Long multibyte character code, eg \U12345678
|
7
|
393 */
|
|
394 static char_u REGEXP_INRANGE[] = "]^-n\\";
|
24
|
395 static char_u REGEXP_ABBR[] = "nrtebdoxuU";
|
7
|
396
|
|
397 static int backslash_trans __ARGS((int c));
|
167
|
398 static int get_char_class __ARGS((char_u **pp));
|
|
399 static int get_equi_class __ARGS((char_u **pp));
|
|
400 static void reg_equi_class __ARGS((int c));
|
|
401 static int get_coll_element __ARGS((char_u **pp));
|
7
|
402 static char_u *skip_anyof __ARGS((char_u *p));
|
|
403 static void init_class_tab __ARGS((void));
|
|
404
|
|
405 /*
|
|
406 * Translate '\x' to its control character, except "\n", which is Magic.
|
|
407 */
|
|
408 static int
|
|
409 backslash_trans(c)
|
|
410 int c;
|
|
411 {
|
|
412 switch (c)
|
|
413 {
|
|
414 case 'r': return CAR;
|
|
415 case 't': return TAB;
|
|
416 case 'e': return ESC;
|
|
417 case 'b': return BS;
|
|
418 }
|
|
419 return c;
|
|
420 }
|
|
421
|
|
422 /*
|
167
|
423 * Check for a character class name "[:name:]". "pp" points to the '['.
|
7
|
424 * Returns one of the CLASS_ items. CLASS_NONE means that no item was
|
|
425 * recognized. Otherwise "pp" is advanced to after the item.
|
|
426 */
|
|
427 static int
|
167
|
428 get_char_class(pp)
|
7
|
429 char_u **pp;
|
|
430 {
|
|
431 static const char *(class_names[]) =
|
|
432 {
|
|
433 "alnum:]",
|
|
434 #define CLASS_ALNUM 0
|
|
435 "alpha:]",
|
|
436 #define CLASS_ALPHA 1
|
|
437 "blank:]",
|
|
438 #define CLASS_BLANK 2
|
|
439 "cntrl:]",
|
|
440 #define CLASS_CNTRL 3
|
|
441 "digit:]",
|
|
442 #define CLASS_DIGIT 4
|
|
443 "graph:]",
|
|
444 #define CLASS_GRAPH 5
|
|
445 "lower:]",
|
|
446 #define CLASS_LOWER 6
|
|
447 "print:]",
|
|
448 #define CLASS_PRINT 7
|
|
449 "punct:]",
|
|
450 #define CLASS_PUNCT 8
|
|
451 "space:]",
|
|
452 #define CLASS_SPACE 9
|
|
453 "upper:]",
|
|
454 #define CLASS_UPPER 10
|
|
455 "xdigit:]",
|
|
456 #define CLASS_XDIGIT 11
|
|
457 "tab:]",
|
|
458 #define CLASS_TAB 12
|
|
459 "return:]",
|
|
460 #define CLASS_RETURN 13
|
|
461 "backspace:]",
|
|
462 #define CLASS_BACKSPACE 14
|
|
463 "escape:]",
|
|
464 #define CLASS_ESCAPE 15
|
|
465 };
|
|
466 #define CLASS_NONE 99
|
|
467 int i;
|
|
468
|
|
469 if ((*pp)[1] == ':')
|
|
470 {
|
|
471 for (i = 0; i < sizeof(class_names) / sizeof(*class_names); ++i)
|
|
472 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0)
|
|
473 {
|
|
474 *pp += STRLEN(class_names[i]) + 2;
|
|
475 return i;
|
|
476 }
|
|
477 }
|
|
478 return CLASS_NONE;
|
|
479 }
|
|
480
|
|
481 /*
|
|
482 * Specific version of character class functions.
|
|
483 * Using a table to keep this fast.
|
|
484 */
|
|
485 static short class_tab[256];
|
|
486
|
|
487 #define RI_DIGIT 0x01
|
|
488 #define RI_HEX 0x02
|
|
489 #define RI_OCTAL 0x04
|
|
490 #define RI_WORD 0x08
|
|
491 #define RI_HEAD 0x10
|
|
492 #define RI_ALPHA 0x20
|
|
493 #define RI_LOWER 0x40
|
|
494 #define RI_UPPER 0x80
|
|
495 #define RI_WHITE 0x100
|
|
496
|
|
497 static void
|
|
498 init_class_tab()
|
|
499 {
|
|
500 int i;
|
|
501 static int done = FALSE;
|
|
502
|
|
503 if (done)
|
|
504 return;
|
|
505
|
|
506 for (i = 0; i < 256; ++i)
|
|
507 {
|
|
508 if (i >= '0' && i <= '7')
|
|
509 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD;
|
|
510 else if (i >= '8' && i <= '9')
|
|
511 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD;
|
|
512 else if (i >= 'a' && i <= 'f')
|
|
513 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
514 #ifdef EBCDIC
|
|
515 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r')
|
|
516 || (i >= 's' && i <= 'z'))
|
|
517 #else
|
|
518 else if (i >= 'g' && i <= 'z')
|
|
519 #endif
|
|
520 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
521 else if (i >= 'A' && i <= 'F')
|
|
522 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
523 #ifdef EBCDIC
|
|
524 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R')
|
|
525 || (i >= 'S' && i <= 'Z'))
|
|
526 #else
|
|
527 else if (i >= 'G' && i <= 'Z')
|
|
528 #endif
|
|
529 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
530 else if (i == '_')
|
|
531 class_tab[i] = RI_WORD + RI_HEAD;
|
|
532 else
|
|
533 class_tab[i] = 0;
|
|
534 }
|
|
535 class_tab[' '] |= RI_WHITE;
|
|
536 class_tab['\t'] |= RI_WHITE;
|
|
537 done = TRUE;
|
|
538 }
|
|
539
|
|
540 #ifdef FEAT_MBYTE
|
|
541 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT))
|
|
542 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX))
|
|
543 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL))
|
|
544 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD))
|
|
545 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD))
|
|
546 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA))
|
|
547 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER))
|
|
548 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER))
|
|
549 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE))
|
|
550 #else
|
|
551 # define ri_digit(c) (class_tab[c] & RI_DIGIT)
|
|
552 # define ri_hex(c) (class_tab[c] & RI_HEX)
|
|
553 # define ri_octal(c) (class_tab[c] & RI_OCTAL)
|
|
554 # define ri_word(c) (class_tab[c] & RI_WORD)
|
|
555 # define ri_head(c) (class_tab[c] & RI_HEAD)
|
|
556 # define ri_alpha(c) (class_tab[c] & RI_ALPHA)
|
|
557 # define ri_lower(c) (class_tab[c] & RI_LOWER)
|
|
558 # define ri_upper(c) (class_tab[c] & RI_UPPER)
|
|
559 # define ri_white(c) (class_tab[c] & RI_WHITE)
|
|
560 #endif
|
|
561
|
|
562 /* flags for regflags */
|
|
563 #define RF_ICASE 1 /* ignore case */
|
|
564 #define RF_NOICASE 2 /* don't ignore case */
|
|
565 #define RF_HASNL 4 /* can match a NL */
|
|
566 #define RF_ICOMBINE 8 /* ignore combining characters */
|
|
567 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */
|
|
568
|
|
569 /*
|
|
570 * Global work variables for vim_regcomp().
|
|
571 */
|
|
572
|
|
573 static char_u *regparse; /* Input-scan pointer. */
|
|
574 static int prevchr_len; /* byte length of previous char */
|
|
575 static int num_complex_braces; /* Complex \{...} count */
|
|
576 static int regnpar; /* () count. */
|
|
577 #ifdef FEAT_SYN_HL
|
|
578 static int regnzpar; /* \z() count. */
|
|
579 static int re_has_z; /* \z item detected */
|
|
580 #endif
|
|
581 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */
|
|
582 static long regsize; /* Code size. */
|
|
583 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */
|
|
584 static unsigned regflags; /* RF_ flags for prog */
|
|
585 static long brace_min[10]; /* Minimums for complex brace repeats */
|
|
586 static long brace_max[10]; /* Maximums for complex brace repeats */
|
|
587 static int brace_count[10]; /* Current counts for complex brace repeats */
|
|
588 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
589 static int had_eol; /* TRUE when EOL found by vim_regcomp() */
|
|
590 #endif
|
|
591 static int one_exactly = FALSE; /* only do one char for EXACTLY */
|
|
592
|
|
593 static int reg_magic; /* magicness of the pattern: */
|
|
594 #define MAGIC_NONE 1 /* "\V" very unmagic */
|
|
595 #define MAGIC_OFF 2 /* "\M" or 'magic' off */
|
|
596 #define MAGIC_ON 3 /* "\m" or 'magic' */
|
|
597 #define MAGIC_ALL 4 /* "\v" very magic */
|
|
598
|
|
599 static int reg_string; /* matching with a string instead of a buffer
|
|
600 line */
|
|
601
|
|
602 /*
|
|
603 * META contains all characters that may be magic, except '^' and '$'.
|
|
604 */
|
|
605
|
|
606 #ifdef EBCDIC
|
|
607 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~";
|
|
608 #else
|
|
609 /* META[] is used often enough to justify turning it into a table. */
|
|
610 static char_u META_flags[] = {
|
|
611 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
612 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
613 /* % & ( ) * + . */
|
|
614 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0,
|
|
615 /* 1 2 3 4 5 6 7 8 9 < = > ? */
|
|
616 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1,
|
|
617 /* @ A C D F H I K L M O */
|
|
618 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
|
|
619 /* P S U V W X Z [ _ */
|
|
620 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1,
|
|
621 /* a c d f h i k l m n o */
|
|
622 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
|
|
623 /* p s u v w x z { | ~ */
|
|
624 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1
|
|
625 };
|
|
626 #endif
|
|
627
|
|
628 static int curchr;
|
|
629
|
|
630 /* arguments for reg() */
|
|
631 #define REG_NOPAREN 0 /* toplevel reg() */
|
|
632 #define REG_PAREN 1 /* \(\) */
|
|
633 #define REG_ZPAREN 2 /* \z(\) */
|
|
634 #define REG_NPAREN 3 /* \%(\) */
|
|
635
|
|
636 /*
|
|
637 * Forward declarations for vim_regcomp()'s friends.
|
|
638 */
|
|
639 static void initchr __ARGS((char_u *));
|
|
640 static int getchr __ARGS((void));
|
|
641 static void skipchr_keepstart __ARGS((void));
|
|
642 static int peekchr __ARGS((void));
|
|
643 static void skipchr __ARGS((void));
|
|
644 static void ungetchr __ARGS((void));
|
24
|
645 static int gethexchrs __ARGS((int maxinputlen));
|
|
646 static int getoctchrs __ARGS((void));
|
|
647 static int getdecchrs __ARGS((void));
|
|
648 static int coll_get_char __ARGS((void));
|
7
|
649 static void regcomp_start __ARGS((char_u *expr, int flags));
|
|
650 static char_u *reg __ARGS((int, int *));
|
|
651 static char_u *regbranch __ARGS((int *flagp));
|
|
652 static char_u *regconcat __ARGS((int *flagp));
|
|
653 static char_u *regpiece __ARGS((int *));
|
|
654 static char_u *regatom __ARGS((int *));
|
|
655 static char_u *regnode __ARGS((int));
|
|
656 static int prog_magic_wrong __ARGS((void));
|
|
657 static char_u *regnext __ARGS((char_u *));
|
|
658 static void regc __ARGS((int b));
|
|
659 #ifdef FEAT_MBYTE
|
|
660 static void regmbc __ARGS((int c));
|
167
|
661 #else
|
|
662 # define regmbc(c) regc(c)
|
7
|
663 #endif
|
|
664 static void reginsert __ARGS((int, char_u *));
|
|
665 static void reginsert_limits __ARGS((int, long, long, char_u *));
|
|
666 static char_u *re_put_long __ARGS((char_u *pr, long_u val));
|
|
667 static int read_limits __ARGS((long *, long *));
|
|
668 static void regtail __ARGS((char_u *, char_u *));
|
|
669 static void regoptail __ARGS((char_u *, char_u *));
|
|
670
|
|
671 /*
|
|
672 * Return TRUE if compiled regular expression "prog" can match a line break.
|
|
673 */
|
|
674 int
|
|
675 re_multiline(prog)
|
|
676 regprog_T *prog;
|
|
677 {
|
|
678 return (prog->regflags & RF_HASNL);
|
|
679 }
|
|
680
|
|
681 /*
|
|
682 * Return TRUE if compiled regular expression "prog" looks before the start
|
|
683 * position (pattern contains "\@<=" or "\@<!").
|
|
684 */
|
|
685 int
|
|
686 re_lookbehind(prog)
|
|
687 regprog_T *prog;
|
|
688 {
|
|
689 return (prog->regflags & RF_LOOKBH);
|
|
690 }
|
|
691
|
|
692 /*
|
167
|
693 * Check for an equivalence class name "[=a=]". "pp" points to the '['.
|
|
694 * Returns a character representing the class. Zero means that no item was
|
|
695 * recognized. Otherwise "pp" is advanced to after the item.
|
|
696 */
|
|
697 static int
|
|
698 get_equi_class(pp)
|
|
699 char_u **pp;
|
|
700 {
|
|
701 int c;
|
|
702 int l = 1;
|
|
703 char_u *p = *pp;
|
|
704
|
|
705 if (p[1] == '=')
|
|
706 {
|
|
707 #ifdef FEAT_MBYTE
|
|
708 if (has_mbyte)
|
|
709 l = mb_ptr2len_check(p + 2);
|
|
710 #endif
|
|
711 if (p[l + 2] == '=' && p[l + 3] == ']')
|
|
712 {
|
|
713 #ifdef FEAT_MBYTE
|
|
714 if (has_mbyte)
|
|
715 c = mb_ptr2char(p + 2);
|
|
716 else
|
|
717 #endif
|
|
718 c = p[2];
|
|
719 *pp += l + 4;
|
|
720 return c;
|
|
721 }
|
|
722 }
|
|
723 return 0;
|
|
724 }
|
|
725
|
|
726 /*
|
|
727 * Produce the bytes for equivalence class "c".
|
|
728 * Currently only handles latin1, latin9 and utf-8.
|
|
729 */
|
|
730 static void
|
|
731 reg_equi_class(c)
|
|
732 int c;
|
|
733 {
|
|
734 #ifdef FEAT_MBYTE
|
|
735 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0
|
|
736 || STRCMP(p_enc, "latin9") == 0)
|
|
737 #endif
|
|
738 {
|
|
739 switch (c)
|
|
740 {
|
|
741 case 'A': case 'À': case 'Á': case 'Â':
|
|
742 case 'Ã': case 'Ä': case 'Å':
|
|
743 regmbc('A'); regmbc('À'); regmbc('Á'); regmbc('Â');
|
|
744 regmbc('Ã'); regmbc('Ä'); regmbc('Å');
|
|
745 return;
|
|
746 case 'C': case 'Ç':
|
|
747 regmbc('C'); regmbc('Ç');
|
|
748 return;
|
|
749 case 'E': case 'È': case 'É': case 'Ê': case 'Ë':
|
|
750 regmbc('E'); regmbc('È'); regmbc('É'); regmbc('Ê');
|
|
751 regmbc('Ë');
|
|
752 return;
|
|
753 case 'I': case 'Ì': case 'Í': case 'Î': case 'Ï':
|
|
754 regmbc('I'); regmbc('Ì'); regmbc('Í'); regmbc('Î');
|
|
755 regmbc('Ï');
|
|
756 return;
|
|
757 case 'N': case 'Ñ':
|
|
758 regmbc('N'); regmbc('Ñ');
|
|
759 return;
|
|
760 case 'O': case 'Ò': case 'Ó': case 'Ô': case 'Õ': case 'Ö':
|
|
761 regmbc('O'); regmbc('Ò'); regmbc('Ó'); regmbc('Ô');
|
|
762 regmbc('Õ'); regmbc('Ö');
|
|
763 return;
|
|
764 case 'U': case 'Ù': case 'Ú': case 'Û': case 'Ü':
|
|
765 regmbc('U'); regmbc('Ù'); regmbc('Ú'); regmbc('Û');
|
|
766 regmbc('Ü');
|
|
767 return;
|
|
768 case 'Y': case 'Ý':
|
|
769 regmbc('Y'); regmbc('Ý');
|
|
770 return;
|
|
771 case 'a': case 'à': case 'á': case 'â':
|
|
772 case 'ã': case 'ä': case 'å':
|
|
773 regmbc('a'); regmbc('à'); regmbc('á'); regmbc('â');
|
|
774 regmbc('ã'); regmbc('ä'); regmbc('å');
|
|
775 return;
|
|
776 case 'c': case 'ç':
|
|
777 regmbc('c'); regmbc('ç');
|
|
778 return;
|
|
779 case 'e': case 'è': case 'é': case 'ê': case 'ë':
|
|
780 regmbc('e'); regmbc('è'); regmbc('é'); regmbc('ê');
|
|
781 regmbc('ë');
|
|
782 return;
|
|
783 case 'i': case 'ì': case 'í': case 'î': case 'ï':
|
|
784 regmbc('i'); regmbc('ì'); regmbc('í'); regmbc('î');
|
|
785 regmbc('ï');
|
|
786 return;
|
|
787 case 'n': case 'ñ':
|
|
788 regmbc('n'); regmbc('ñ');
|
|
789 return;
|
|
790 case 'o': case 'ò': case 'ó': case 'ô': case 'õ': case 'ö':
|
|
791 regmbc('o'); regmbc('ò'); regmbc('ó'); regmbc('ô');
|
|
792 regmbc('õ'); regmbc('ö');
|
|
793 return;
|
|
794 case 'u': case 'ù': case 'ú': case 'û': case 'ü':
|
|
795 regmbc('u'); regmbc('ù'); regmbc('ú'); regmbc('û');
|
|
796 regmbc('ü');
|
|
797 return;
|
|
798 case 'y': case 'ý': case 'ÿ':
|
|
799 regmbc('y'); regmbc('ý'); regmbc('ÿ');
|
|
800 return;
|
|
801 }
|
|
802 }
|
|
803 regmbc(c);
|
|
804 }
|
|
805
|
|
806 /*
|
|
807 * Check for a collating element "[.a.]". "pp" points to the '['.
|
|
808 * Returns a character. Zero means that no item was recognized. Otherwise
|
|
809 * "pp" is advanced to after the item.
|
|
810 * Currently only single characters are recognized!
|
|
811 */
|
|
812 static int
|
|
813 get_coll_element(pp)
|
|
814 char_u **pp;
|
|
815 {
|
|
816 int c;
|
|
817 int l = 1;
|
|
818 char_u *p = *pp;
|
|
819
|
|
820 if (p[1] == '.')
|
|
821 {
|
|
822 #ifdef FEAT_MBYTE
|
|
823 if (has_mbyte)
|
|
824 l = mb_ptr2len_check(p + 2);
|
|
825 #endif
|
|
826 if (p[l + 2] == '.' && p[l + 3] == ']')
|
|
827 {
|
|
828 #ifdef FEAT_MBYTE
|
|
829 if (has_mbyte)
|
|
830 c = mb_ptr2char(p + 2);
|
|
831 else
|
|
832 #endif
|
|
833 c = p[2];
|
|
834 *pp += l + 4;
|
|
835 return c;
|
|
836 }
|
|
837 }
|
|
838 return 0;
|
|
839 }
|
|
840
|
|
841
|
|
842 /*
|
|
843 * Skip over a "[]" range.
|
|
844 * "p" must point to the character after the '['.
|
|
845 * The returned pointer is on the matching ']', or the terminating NUL.
|
|
846 */
|
|
847 static char_u *
|
|
848 skip_anyof(p)
|
|
849 char_u *p;
|
|
850 {
|
|
851 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
|
852 int cpo_bsl; /* 'cpoptions' contains '\' flag */
|
|
853 #ifdef FEAT_MBYTE
|
|
854 int l;
|
|
855 #endif
|
|
856
|
|
857 cpo_lit = (!reg_syn && vim_strchr(p_cpo, CPO_LITERAL) != NULL);
|
|
858 cpo_bsl = (!reg_syn && vim_strchr(p_cpo, CPO_BACKSL) != NULL);
|
|
859
|
|
860 if (*p == '^') /* Complement of range. */
|
|
861 ++p;
|
|
862 if (*p == ']' || *p == '-')
|
|
863 ++p;
|
|
864 while (*p != NUL && *p != ']')
|
|
865 {
|
|
866 #ifdef FEAT_MBYTE
|
|
867 if (has_mbyte && (l = (*mb_ptr2len_check)(p)) > 1)
|
|
868 p += l;
|
|
869 else
|
|
870 #endif
|
|
871 if (*p == '-')
|
|
872 {
|
|
873 ++p;
|
|
874 if (*p != ']' && *p != NUL)
|
|
875 mb_ptr_adv(p);
|
|
876 }
|
|
877 else if (*p == '\\'
|
|
878 && !cpo_bsl
|
|
879 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL
|
|
880 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL)))
|
|
881 p += 2;
|
|
882 else if (*p == '[')
|
|
883 {
|
|
884 if (get_char_class(&p) == CLASS_NONE
|
|
885 && get_equi_class(&p) == 0
|
|
886 && get_coll_element(&p) == 0)
|
|
887 ++p; /* It was not a class name */
|
|
888 }
|
|
889 else
|
|
890 ++p;
|
|
891 }
|
|
892
|
|
893 return p;
|
|
894 }
|
|
895
|
|
896 /*
|
7
|
897 * Skip past regular expression.
|
153
|
898 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc).
|
7
|
899 * Take care of characters with a backslash in front of it.
|
|
900 * Skip strings inside [ and ].
|
|
901 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the
|
|
902 * expression and change "\?" to "?". If "*newp" is not NULL the expression
|
|
903 * is changed in-place.
|
|
904 */
|
|
905 char_u *
|
|
906 skip_regexp(startp, dirc, magic, newp)
|
|
907 char_u *startp;
|
|
908 int dirc;
|
|
909 int magic;
|
|
910 char_u **newp;
|
|
911 {
|
|
912 int mymagic;
|
|
913 char_u *p = startp;
|
|
914
|
|
915 if (magic)
|
|
916 mymagic = MAGIC_ON;
|
|
917 else
|
|
918 mymagic = MAGIC_OFF;
|
|
919
|
39
|
920 for (; p[0] != NUL; mb_ptr_adv(p))
|
7
|
921 {
|
|
922 if (p[0] == dirc) /* found end of regexp */
|
|
923 break;
|
|
924 if ((p[0] == '[' && mymagic >= MAGIC_ON)
|
|
925 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF))
|
|
926 {
|
|
927 p = skip_anyof(p + 1);
|
|
928 if (p[0] == NUL)
|
|
929 break;
|
|
930 }
|
|
931 else if (p[0] == '\\' && p[1] != NUL)
|
|
932 {
|
|
933 if (dirc == '?' && newp != NULL && p[1] == '?')
|
|
934 {
|
|
935 /* change "\?" to "?", make a copy first. */
|
|
936 if (*newp == NULL)
|
|
937 {
|
|
938 *newp = vim_strsave(startp);
|
|
939 if (*newp != NULL)
|
|
940 p = *newp + (p - startp);
|
|
941 }
|
|
942 if (*newp != NULL)
|
|
943 mch_memmove(p, p + 1, STRLEN(p));
|
|
944 else
|
|
945 ++p;
|
|
946 }
|
|
947 else
|
|
948 ++p; /* skip next character */
|
|
949 if (*p == 'v')
|
|
950 mymagic = MAGIC_ALL;
|
|
951 else if (*p == 'V')
|
|
952 mymagic = MAGIC_NONE;
|
|
953 }
|
|
954 }
|
|
955 return p;
|
|
956 }
|
|
957
|
|
958 /*
|
41
|
959 * vim_regcomp() - compile a regular expression into internal code
|
|
960 * Returns the program in allocated space. Returns NULL for an error.
|
7
|
961 *
|
|
962 * We can't allocate space until we know how big the compiled form will be,
|
|
963 * but we can't compile it (and thus know how big it is) until we've got a
|
|
964 * place to put the code. So we cheat: we compile it twice, once with code
|
|
965 * generation turned off and size counting turned on, and once "for real".
|
|
966 * This also means that we don't allocate space until we are sure that the
|
|
967 * thing really will compile successfully, and we never have to move the
|
|
968 * code and thus invalidate pointers into it. (Note that it has to be in
|
|
969 * one piece because vim_free() must be able to free it all.)
|
|
970 *
|
|
971 * Whether upper/lower case is to be ignored is decided when executing the
|
|
972 * program, it does not matter here.
|
|
973 *
|
|
974 * Beware that the optimization-preparation code in here knows about some
|
|
975 * of the structure of the compiled regexp.
|
|
976 * "re_flags": RE_MAGIC and/or RE_STRING.
|
|
977 */
|
|
978 regprog_T *
|
|
979 vim_regcomp(expr, re_flags)
|
|
980 char_u *expr;
|
|
981 int re_flags;
|
|
982 {
|
|
983 regprog_T *r;
|
|
984 char_u *scan;
|
|
985 char_u *longest;
|
|
986 int len;
|
|
987 int flags;
|
|
988
|
|
989 if (expr == NULL)
|
|
990 EMSG_RET_NULL(_(e_null));
|
|
991
|
|
992 init_class_tab();
|
|
993
|
|
994 /*
|
|
995 * First pass: determine size, legality.
|
|
996 */
|
|
997 regcomp_start(expr, re_flags);
|
|
998 regcode = JUST_CALC_SIZE;
|
|
999 regc(REGMAGIC);
|
|
1000 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
1001 return NULL;
|
|
1002
|
|
1003 /* Small enough for pointer-storage convention? */
|
|
1004 #ifdef SMALL_MALLOC /* 16 bit storage allocation */
|
|
1005 if (regsize >= 65536L - 256L)
|
|
1006 EMSG_RET_NULL(_("E339: Pattern too long"));
|
|
1007 #endif
|
|
1008
|
|
1009 /* Allocate space. */
|
|
1010 r = (regprog_T *)lalloc(sizeof(regprog_T) + regsize, TRUE);
|
|
1011 if (r == NULL)
|
|
1012 return NULL;
|
|
1013
|
|
1014 /*
|
|
1015 * Second pass: emit code.
|
|
1016 */
|
|
1017 regcomp_start(expr, re_flags);
|
|
1018 regcode = r->program;
|
|
1019 regc(REGMAGIC);
|
|
1020 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
1021 {
|
|
1022 vim_free(r);
|
|
1023 return NULL;
|
|
1024 }
|
|
1025
|
|
1026 /* Dig out information for optimizations. */
|
|
1027 r->regstart = NUL; /* Worst-case defaults. */
|
|
1028 r->reganch = 0;
|
|
1029 r->regmust = NULL;
|
|
1030 r->regmlen = 0;
|
|
1031 r->regflags = regflags;
|
|
1032 if (flags & HASNL)
|
|
1033 r->regflags |= RF_HASNL;
|
|
1034 if (flags & HASLOOKBH)
|
|
1035 r->regflags |= RF_LOOKBH;
|
|
1036 #ifdef FEAT_SYN_HL
|
|
1037 /* Remember whether this pattern has any \z specials in it. */
|
|
1038 r->reghasz = re_has_z;
|
|
1039 #endif
|
|
1040 scan = r->program + 1; /* First BRANCH. */
|
|
1041 if (OP(regnext(scan)) == END) /* Only one top-level choice. */
|
|
1042 {
|
|
1043 scan = OPERAND(scan);
|
|
1044
|
|
1045 /* Starting-point info. */
|
|
1046 if (OP(scan) == BOL || OP(scan) == RE_BOF)
|
|
1047 {
|
|
1048 r->reganch++;
|
|
1049 scan = regnext(scan);
|
|
1050 }
|
|
1051
|
|
1052 if (OP(scan) == EXACTLY)
|
|
1053 {
|
|
1054 #ifdef FEAT_MBYTE
|
|
1055 if (has_mbyte)
|
|
1056 r->regstart = (*mb_ptr2char)(OPERAND(scan));
|
|
1057 else
|
|
1058 #endif
|
|
1059 r->regstart = *OPERAND(scan);
|
|
1060 }
|
|
1061 else if ((OP(scan) == BOW
|
|
1062 || OP(scan) == EOW
|
|
1063 || OP(scan) == NOTHING
|
|
1064 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
|
|
1065 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
|
|
1066 && OP(regnext(scan)) == EXACTLY)
|
|
1067 {
|
|
1068 #ifdef FEAT_MBYTE
|
|
1069 if (has_mbyte)
|
|
1070 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
|
|
1071 else
|
|
1072 #endif
|
|
1073 r->regstart = *OPERAND(regnext(scan));
|
|
1074 }
|
|
1075
|
|
1076 /*
|
|
1077 * If there's something expensive in the r.e., find the longest
|
|
1078 * literal string that must appear and make it the regmust. Resolve
|
|
1079 * ties in favor of later strings, since the regstart check works
|
|
1080 * with the beginning of the r.e. and avoiding duplication
|
|
1081 * strengthens checking. Not a strong reason, but sufficient in the
|
|
1082 * absence of others.
|
|
1083 */
|
|
1084 /*
|
|
1085 * When the r.e. starts with BOW, it is faster to look for a regmust
|
|
1086 * first. Used a lot for "#" and "*" commands. (Added by mool).
|
|
1087 */
|
|
1088 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
|
|
1089 && !(flags & HASNL))
|
|
1090 {
|
|
1091 longest = NULL;
|
|
1092 len = 0;
|
|
1093 for (; scan != NULL; scan = regnext(scan))
|
|
1094 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
|
|
1095 {
|
|
1096 longest = OPERAND(scan);
|
|
1097 len = (int)STRLEN(OPERAND(scan));
|
|
1098 }
|
|
1099 r->regmust = longest;
|
|
1100 r->regmlen = len;
|
|
1101 }
|
|
1102 }
|
|
1103 #ifdef DEBUG
|
|
1104 regdump(expr, r);
|
|
1105 #endif
|
|
1106 return r;
|
|
1107 }
|
|
1108
|
|
1109 /*
|
|
1110 * Setup to parse the regexp. Used once to get the length and once to do it.
|
|
1111 */
|
|
1112 static void
|
|
1113 regcomp_start(expr, re_flags)
|
|
1114 char_u *expr;
|
|
1115 int re_flags; /* see vim_regcomp() */
|
|
1116 {
|
|
1117 initchr(expr);
|
|
1118 if (re_flags & RE_MAGIC)
|
|
1119 reg_magic = MAGIC_ON;
|
|
1120 else
|
|
1121 reg_magic = MAGIC_OFF;
|
|
1122 reg_string = (re_flags & RE_STRING);
|
|
1123
|
|
1124 num_complex_braces = 0;
|
|
1125 regnpar = 1;
|
|
1126 vim_memset(had_endbrace, 0, sizeof(had_endbrace));
|
|
1127 #ifdef FEAT_SYN_HL
|
|
1128 regnzpar = 1;
|
|
1129 re_has_z = 0;
|
|
1130 #endif
|
|
1131 regsize = 0L;
|
|
1132 regflags = 0;
|
|
1133 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1134 had_eol = FALSE;
|
|
1135 #endif
|
|
1136 }
|
|
1137
|
|
1138 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1139 /*
|
|
1140 * Check if during the previous call to vim_regcomp the EOL item "$" has been
|
|
1141 * found. This is messy, but it works fine.
|
|
1142 */
|
|
1143 int
|
|
1144 vim_regcomp_had_eol()
|
|
1145 {
|
|
1146 return had_eol;
|
|
1147 }
|
|
1148 #endif
|
|
1149
|
|
1150 /*
|
|
1151 * reg - regular expression, i.e. main body or parenthesized thing
|
|
1152 *
|
|
1153 * Caller must absorb opening parenthesis.
|
|
1154 *
|
|
1155 * Combining parenthesis handling with the base level of regular expression
|
|
1156 * is a trifle forced, but the need to tie the tails of the branches to what
|
|
1157 * follows makes it hard to avoid.
|
|
1158 */
|
|
1159 static char_u *
|
|
1160 reg(paren, flagp)
|
|
1161 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */
|
|
1162 int *flagp;
|
|
1163 {
|
|
1164 char_u *ret;
|
|
1165 char_u *br;
|
|
1166 char_u *ender;
|
|
1167 int parno = 0;
|
|
1168 int flags;
|
|
1169
|
|
1170 *flagp = HASWIDTH; /* Tentatively. */
|
|
1171
|
|
1172 #ifdef FEAT_SYN_HL
|
|
1173 if (paren == REG_ZPAREN)
|
|
1174 {
|
|
1175 /* Make a ZOPEN node. */
|
|
1176 if (regnzpar >= NSUBEXP)
|
|
1177 EMSG_RET_NULL(_("E50: Too many \\z("));
|
|
1178 parno = regnzpar;
|
|
1179 regnzpar++;
|
|
1180 ret = regnode(ZOPEN + parno);
|
|
1181 }
|
|
1182 else
|
|
1183 #endif
|
|
1184 if (paren == REG_PAREN)
|
|
1185 {
|
|
1186 /* Make a MOPEN node. */
|
|
1187 if (regnpar >= NSUBEXP)
|
|
1188 EMSG_M_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL);
|
|
1189 parno = regnpar;
|
|
1190 ++regnpar;
|
|
1191 ret = regnode(MOPEN + parno);
|
|
1192 }
|
|
1193 else if (paren == REG_NPAREN)
|
|
1194 {
|
|
1195 /* Make a NOPEN node. */
|
|
1196 ret = regnode(NOPEN);
|
|
1197 }
|
|
1198 else
|
|
1199 ret = NULL;
|
|
1200
|
|
1201 /* Pick up the branches, linking them together. */
|
|
1202 br = regbranch(&flags);
|
|
1203 if (br == NULL)
|
|
1204 return NULL;
|
|
1205 if (ret != NULL)
|
|
1206 regtail(ret, br); /* [MZ]OPEN -> first. */
|
|
1207 else
|
|
1208 ret = br;
|
|
1209 /* If one of the branches can be zero-width, the whole thing can.
|
|
1210 * If one of the branches has * at start or matches a line-break, the
|
|
1211 * whole thing can. */
|
|
1212 if (!(flags & HASWIDTH))
|
|
1213 *flagp &= ~HASWIDTH;
|
|
1214 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1215 while (peekchr() == Magic('|'))
|
|
1216 {
|
|
1217 skipchr();
|
|
1218 br = regbranch(&flags);
|
|
1219 if (br == NULL)
|
|
1220 return NULL;
|
|
1221 regtail(ret, br); /* BRANCH -> BRANCH. */
|
|
1222 if (!(flags & HASWIDTH))
|
|
1223 *flagp &= ~HASWIDTH;
|
|
1224 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1225 }
|
|
1226
|
|
1227 /* Make a closing node, and hook it on the end. */
|
|
1228 ender = regnode(
|
|
1229 #ifdef FEAT_SYN_HL
|
|
1230 paren == REG_ZPAREN ? ZCLOSE + parno :
|
|
1231 #endif
|
|
1232 paren == REG_PAREN ? MCLOSE + parno :
|
|
1233 paren == REG_NPAREN ? NCLOSE : END);
|
|
1234 regtail(ret, ender);
|
|
1235
|
|
1236 /* Hook the tails of the branches to the closing node. */
|
|
1237 for (br = ret; br != NULL; br = regnext(br))
|
|
1238 regoptail(br, ender);
|
|
1239
|
|
1240 /* Check for proper termination. */
|
|
1241 if (paren != REG_NOPAREN && getchr() != Magic(')'))
|
|
1242 {
|
|
1243 #ifdef FEAT_SYN_HL
|
|
1244 if (paren == REG_ZPAREN)
|
|
1245 EMSG_RET_NULL(_("E52: Unmatched \\z("))
|
|
1246 else
|
|
1247 #endif
|
|
1248 if (paren == REG_NPAREN)
|
|
1249 EMSG_M_RET_NULL(_("E53: Unmatched %s%%("), reg_magic == MAGIC_ALL)
|
|
1250 else
|
|
1251 EMSG_M_RET_NULL(_("E54: Unmatched %s("), reg_magic == MAGIC_ALL)
|
|
1252 }
|
|
1253 else if (paren == REG_NOPAREN && peekchr() != NUL)
|
|
1254 {
|
|
1255 if (curchr == Magic(')'))
|
|
1256 EMSG_M_RET_NULL(_("E55: Unmatched %s)"), reg_magic == MAGIC_ALL)
|
|
1257 else
|
|
1258 EMSG_RET_NULL(_(e_trailing)) /* "Can't happen". */
|
|
1259 /* NOTREACHED */
|
|
1260 }
|
|
1261 /*
|
|
1262 * Here we set the flag allowing back references to this set of
|
|
1263 * parentheses.
|
|
1264 */
|
|
1265 if (paren == REG_PAREN)
|
|
1266 had_endbrace[parno] = TRUE; /* have seen the close paren */
|
|
1267 return ret;
|
|
1268 }
|
|
1269
|
|
1270 /*
|
|
1271 * regbranch - one alternative of an | operator
|
|
1272 *
|
|
1273 * Implements the & operator.
|
|
1274 */
|
|
1275 static char_u *
|
|
1276 regbranch(flagp)
|
|
1277 int *flagp;
|
|
1278 {
|
|
1279 char_u *ret;
|
|
1280 char_u *chain = NULL;
|
|
1281 char_u *latest;
|
|
1282 int flags;
|
|
1283
|
|
1284 *flagp = WORST | HASNL; /* Tentatively. */
|
|
1285
|
|
1286 ret = regnode(BRANCH);
|
|
1287 for (;;)
|
|
1288 {
|
|
1289 latest = regconcat(&flags);
|
|
1290 if (latest == NULL)
|
|
1291 return NULL;
|
|
1292 /* If one of the branches has width, the whole thing has. If one of
|
|
1293 * the branches anchors at start-of-line, the whole thing does.
|
|
1294 * If one of the branches uses look-behind, the whole thing does. */
|
|
1295 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
|
|
1296 /* If one of the branches doesn't match a line-break, the whole thing
|
|
1297 * doesn't. */
|
|
1298 *flagp &= ~HASNL | (flags & HASNL);
|
|
1299 if (chain != NULL)
|
|
1300 regtail(chain, latest);
|
|
1301 if (peekchr() != Magic('&'))
|
|
1302 break;
|
|
1303 skipchr();
|
|
1304 regtail(latest, regnode(END)); /* operand ends */
|
|
1305 reginsert(MATCH, latest);
|
|
1306 chain = latest;
|
|
1307 }
|
|
1308
|
|
1309 return ret;
|
|
1310 }
|
|
1311
|
|
1312 /*
|
|
1313 * regbranch - one alternative of an | or & operator
|
|
1314 *
|
|
1315 * Implements the concatenation operator.
|
|
1316 */
|
|
1317 static char_u *
|
|
1318 regconcat(flagp)
|
|
1319 int *flagp;
|
|
1320 {
|
|
1321 char_u *first = NULL;
|
|
1322 char_u *chain = NULL;
|
|
1323 char_u *latest;
|
|
1324 int flags;
|
|
1325 int cont = TRUE;
|
|
1326
|
|
1327 *flagp = WORST; /* Tentatively. */
|
|
1328
|
|
1329 while (cont)
|
|
1330 {
|
|
1331 switch (peekchr())
|
|
1332 {
|
|
1333 case NUL:
|
|
1334 case Magic('|'):
|
|
1335 case Magic('&'):
|
|
1336 case Magic(')'):
|
|
1337 cont = FALSE;
|
|
1338 break;
|
|
1339 case Magic('Z'):
|
|
1340 #ifdef FEAT_MBYTE
|
|
1341 regflags |= RF_ICOMBINE;
|
|
1342 #endif
|
|
1343 skipchr_keepstart();
|
|
1344 break;
|
|
1345 case Magic('c'):
|
|
1346 regflags |= RF_ICASE;
|
|
1347 skipchr_keepstart();
|
|
1348 break;
|
|
1349 case Magic('C'):
|
|
1350 regflags |= RF_NOICASE;
|
|
1351 skipchr_keepstart();
|
|
1352 break;
|
|
1353 case Magic('v'):
|
|
1354 reg_magic = MAGIC_ALL;
|
|
1355 skipchr_keepstart();
|
|
1356 curchr = -1;
|
|
1357 break;
|
|
1358 case Magic('m'):
|
|
1359 reg_magic = MAGIC_ON;
|
|
1360 skipchr_keepstart();
|
|
1361 curchr = -1;
|
|
1362 break;
|
|
1363 case Magic('M'):
|
|
1364 reg_magic = MAGIC_OFF;
|
|
1365 skipchr_keepstart();
|
|
1366 curchr = -1;
|
|
1367 break;
|
|
1368 case Magic('V'):
|
|
1369 reg_magic = MAGIC_NONE;
|
|
1370 skipchr_keepstart();
|
|
1371 curchr = -1;
|
|
1372 break;
|
|
1373 default:
|
|
1374 latest = regpiece(&flags);
|
|
1375 if (latest == NULL)
|
|
1376 return NULL;
|
|
1377 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
|
|
1378 if (chain == NULL) /* First piece. */
|
|
1379 *flagp |= flags & SPSTART;
|
|
1380 else
|
|
1381 regtail(chain, latest);
|
|
1382 chain = latest;
|
|
1383 if (first == NULL)
|
|
1384 first = latest;
|
|
1385 break;
|
|
1386 }
|
|
1387 }
|
|
1388 if (first == NULL) /* Loop ran zero times. */
|
|
1389 first = regnode(NOTHING);
|
|
1390 return first;
|
|
1391 }
|
|
1392
|
|
1393 /*
|
|
1394 * regpiece - something followed by possible [*+=]
|
|
1395 *
|
|
1396 * Note that the branching code sequences used for = and the general cases
|
|
1397 * of * and + are somewhat optimized: they use the same NOTHING node as
|
|
1398 * both the endmarker for their branch list and the body of the last branch.
|
|
1399 * It might seem that this node could be dispensed with entirely, but the
|
|
1400 * endmarker role is not redundant.
|
|
1401 */
|
|
1402 static char_u *
|
|
1403 regpiece(flagp)
|
|
1404 int *flagp;
|
|
1405 {
|
|
1406 char_u *ret;
|
|
1407 int op;
|
|
1408 char_u *next;
|
|
1409 int flags;
|
|
1410 long minval;
|
|
1411 long maxval;
|
|
1412
|
|
1413 ret = regatom(&flags);
|
|
1414 if (ret == NULL)
|
|
1415 return NULL;
|
|
1416
|
|
1417 op = peekchr();
|
|
1418 if (re_multi_type(op) == NOT_MULTI)
|
|
1419 {
|
|
1420 *flagp = flags;
|
|
1421 return ret;
|
|
1422 }
|
|
1423 /* default flags */
|
|
1424 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
|
|
1425
|
|
1426 skipchr();
|
|
1427 switch (op)
|
|
1428 {
|
|
1429 case Magic('*'):
|
|
1430 if (flags & SIMPLE)
|
|
1431 reginsert(STAR, ret);
|
|
1432 else
|
|
1433 {
|
|
1434 /* Emit x* as (x&|), where & means "self". */
|
|
1435 reginsert(BRANCH, ret); /* Either x */
|
|
1436 regoptail(ret, regnode(BACK)); /* and loop */
|
|
1437 regoptail(ret, ret); /* back */
|
|
1438 regtail(ret, regnode(BRANCH)); /* or */
|
|
1439 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1440 }
|
|
1441 break;
|
|
1442
|
|
1443 case Magic('+'):
|
|
1444 if (flags & SIMPLE)
|
|
1445 reginsert(PLUS, ret);
|
|
1446 else
|
|
1447 {
|
|
1448 /* Emit x+ as x(&|), where & means "self". */
|
|
1449 next = regnode(BRANCH); /* Either */
|
|
1450 regtail(ret, next);
|
|
1451 regtail(regnode(BACK), ret); /* loop back */
|
|
1452 regtail(next, regnode(BRANCH)); /* or */
|
|
1453 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1454 }
|
|
1455 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1456 break;
|
|
1457
|
|
1458 case Magic('@'):
|
|
1459 {
|
|
1460 int lop = END;
|
|
1461
|
|
1462 switch (no_Magic(getchr()))
|
|
1463 {
|
|
1464 case '=': lop = MATCH; break; /* \@= */
|
|
1465 case '!': lop = NOMATCH; break; /* \@! */
|
|
1466 case '>': lop = SUBPAT; break; /* \@> */
|
|
1467 case '<': switch (no_Magic(getchr()))
|
|
1468 {
|
|
1469 case '=': lop = BEHIND; break; /* \@<= */
|
|
1470 case '!': lop = NOBEHIND; break; /* \@<! */
|
|
1471 }
|
|
1472 }
|
|
1473 if (lop == END)
|
|
1474 EMSG_M_RET_NULL(_("E59: invalid character after %s@"),
|
|
1475 reg_magic == MAGIC_ALL);
|
|
1476 /* Look behind must match with behind_pos. */
|
|
1477 if (lop == BEHIND || lop == NOBEHIND)
|
|
1478 {
|
|
1479 regtail(ret, regnode(BHPOS));
|
|
1480 *flagp |= HASLOOKBH;
|
|
1481 }
|
|
1482 regtail(ret, regnode(END)); /* operand ends */
|
|
1483 reginsert(lop, ret);
|
|
1484 break;
|
|
1485 }
|
|
1486
|
|
1487 case Magic('?'):
|
|
1488 case Magic('='):
|
|
1489 /* Emit x= as (x|) */
|
|
1490 reginsert(BRANCH, ret); /* Either x */
|
|
1491 regtail(ret, regnode(BRANCH)); /* or */
|
|
1492 next = regnode(NOTHING); /* null. */
|
|
1493 regtail(ret, next);
|
|
1494 regoptail(ret, next);
|
|
1495 break;
|
|
1496
|
|
1497 case Magic('{'):
|
|
1498 if (!read_limits(&minval, &maxval))
|
|
1499 return NULL;
|
|
1500 if (flags & SIMPLE)
|
|
1501 {
|
|
1502 reginsert(BRACE_SIMPLE, ret);
|
|
1503 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1504 }
|
|
1505 else
|
|
1506 {
|
|
1507 if (num_complex_braces >= 10)
|
|
1508 EMSG_M_RET_NULL(_("E60: Too many complex %s{...}s"),
|
|
1509 reg_magic == MAGIC_ALL);
|
|
1510 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
|
|
1511 regoptail(ret, regnode(BACK));
|
|
1512 regoptail(ret, ret);
|
|
1513 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1514 ++num_complex_braces;
|
|
1515 }
|
|
1516 if (minval > 0 && maxval > 0)
|
|
1517 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1518 break;
|
|
1519 }
|
|
1520 if (re_multi_type(peekchr()) != NOT_MULTI)
|
|
1521 {
|
|
1522 /* Can't have a multi follow a multi. */
|
|
1523 if (peekchr() == Magic('*'))
|
|
1524 sprintf((char *)IObuff, _("E61: Nested %s*"),
|
|
1525 reg_magic >= MAGIC_ON ? "" : "\\");
|
|
1526 else
|
|
1527 sprintf((char *)IObuff, _("E62: Nested %s%c"),
|
|
1528 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr()));
|
|
1529 EMSG_RET_NULL(IObuff);
|
|
1530 }
|
|
1531
|
|
1532 return ret;
|
|
1533 }
|
|
1534
|
|
1535 /*
|
|
1536 * regatom - the lowest level
|
|
1537 *
|
|
1538 * Optimization: gobbles an entire sequence of ordinary characters so that
|
|
1539 * it can turn them into a single node, which is smaller to store and
|
|
1540 * faster to run. Don't do this when one_exactly is set.
|
|
1541 */
|
|
1542 static char_u *
|
|
1543 regatom(flagp)
|
|
1544 int *flagp;
|
|
1545 {
|
|
1546 char_u *ret;
|
|
1547 int flags;
|
|
1548 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
167
|
1549 int cpo_bsl; /* 'cpoptions' contains '\' flag */
|
7
|
1550 int c;
|
|
1551 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU";
|
|
1552 static int classcodes[] = {ANY, IDENT, SIDENT, KWORD, SKWORD,
|
|
1553 FNAME, SFNAME, PRINT, SPRINT,
|
|
1554 WHITE, NWHITE, DIGIT, NDIGIT,
|
|
1555 HEX, NHEX, OCTAL, NOCTAL,
|
|
1556 WORD, NWORD, HEAD, NHEAD,
|
|
1557 ALPHA, NALPHA, LOWER, NLOWER,
|
|
1558 UPPER, NUPPER
|
|
1559 };
|
|
1560 char_u *p;
|
|
1561 int extra = 0;
|
|
1562
|
|
1563 *flagp = WORST; /* Tentatively. */
|
|
1564 cpo_lit = (!reg_syn && vim_strchr(p_cpo, CPO_LITERAL) != NULL);
|
167
|
1565 cpo_bsl = (!reg_syn && vim_strchr(p_cpo, CPO_BACKSL) != NULL);
|
7
|
1566
|
|
1567 c = getchr();
|
|
1568 switch (c)
|
|
1569 {
|
|
1570 case Magic('^'):
|
|
1571 ret = regnode(BOL);
|
|
1572 break;
|
|
1573
|
|
1574 case Magic('$'):
|
|
1575 ret = regnode(EOL);
|
|
1576 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1577 had_eol = TRUE;
|
|
1578 #endif
|
|
1579 break;
|
|
1580
|
|
1581 case Magic('<'):
|
|
1582 ret = regnode(BOW);
|
|
1583 break;
|
|
1584
|
|
1585 case Magic('>'):
|
|
1586 ret = regnode(EOW);
|
|
1587 break;
|
|
1588
|
|
1589 case Magic('_'):
|
|
1590 c = no_Magic(getchr());
|
|
1591 if (c == '^') /* "\_^" is start-of-line */
|
|
1592 {
|
|
1593 ret = regnode(BOL);
|
|
1594 break;
|
|
1595 }
|
|
1596 if (c == '$') /* "\_$" is end-of-line */
|
|
1597 {
|
|
1598 ret = regnode(EOL);
|
|
1599 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1600 had_eol = TRUE;
|
|
1601 #endif
|
|
1602 break;
|
|
1603 }
|
|
1604
|
|
1605 extra = ADD_NL;
|
|
1606 *flagp |= HASNL;
|
|
1607
|
|
1608 /* "\_[" is character range plus newline */
|
|
1609 if (c == '[')
|
|
1610 goto collection;
|
|
1611
|
|
1612 /* "\_x" is character class plus newline */
|
|
1613 /*FALLTHROUGH*/
|
|
1614
|
|
1615 /*
|
|
1616 * Character classes.
|
|
1617 */
|
|
1618 case Magic('.'):
|
|
1619 case Magic('i'):
|
|
1620 case Magic('I'):
|
|
1621 case Magic('k'):
|
|
1622 case Magic('K'):
|
|
1623 case Magic('f'):
|
|
1624 case Magic('F'):
|
|
1625 case Magic('p'):
|
|
1626 case Magic('P'):
|
|
1627 case Magic('s'):
|
|
1628 case Magic('S'):
|
|
1629 case Magic('d'):
|
|
1630 case Magic('D'):
|
|
1631 case Magic('x'):
|
|
1632 case Magic('X'):
|
|
1633 case Magic('o'):
|
|
1634 case Magic('O'):
|
|
1635 case Magic('w'):
|
|
1636 case Magic('W'):
|
|
1637 case Magic('h'):
|
|
1638 case Magic('H'):
|
|
1639 case Magic('a'):
|
|
1640 case Magic('A'):
|
|
1641 case Magic('l'):
|
|
1642 case Magic('L'):
|
|
1643 case Magic('u'):
|
|
1644 case Magic('U'):
|
|
1645 p = vim_strchr(classchars, no_Magic(c));
|
|
1646 if (p == NULL)
|
|
1647 EMSG_RET_NULL(_("E63: invalid use of \\_"));
|
|
1648 ret = regnode(classcodes[p - classchars] + extra);
|
|
1649 *flagp |= HASWIDTH | SIMPLE;
|
|
1650 break;
|
|
1651
|
|
1652 case Magic('n'):
|
|
1653 if (reg_string)
|
|
1654 {
|
|
1655 /* In a string "\n" matches a newline character. */
|
|
1656 ret = regnode(EXACTLY);
|
|
1657 regc(NL);
|
|
1658 regc(NUL);
|
|
1659 *flagp |= HASWIDTH | SIMPLE;
|
|
1660 }
|
|
1661 else
|
|
1662 {
|
|
1663 /* In buffer text "\n" matches the end of a line. */
|
|
1664 ret = regnode(NEWL);
|
|
1665 *flagp |= HASWIDTH | HASNL;
|
|
1666 }
|
|
1667 break;
|
|
1668
|
|
1669 case Magic('('):
|
|
1670 if (one_exactly)
|
|
1671 EMSG_ONE_RET_NULL;
|
|
1672 ret = reg(REG_PAREN, &flags);
|
|
1673 if (ret == NULL)
|
|
1674 return NULL;
|
|
1675 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1676 break;
|
|
1677
|
|
1678 case NUL:
|
|
1679 case Magic('|'):
|
|
1680 case Magic('&'):
|
|
1681 case Magic(')'):
|
|
1682 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */
|
|
1683 /* NOTREACHED */
|
|
1684
|
|
1685 case Magic('='):
|
|
1686 case Magic('?'):
|
|
1687 case Magic('+'):
|
|
1688 case Magic('@'):
|
|
1689 case Magic('{'):
|
|
1690 case Magic('*'):
|
|
1691 c = no_Magic(c);
|
|
1692 sprintf((char *)IObuff, _("E64: %s%c follows nothing"),
|
|
1693 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL)
|
|
1694 ? "" : "\\", c);
|
|
1695 EMSG_RET_NULL(IObuff);
|
|
1696 /* NOTREACHED */
|
|
1697
|
|
1698 case Magic('~'): /* previous substitute pattern */
|
|
1699 if (reg_prev_sub)
|
|
1700 {
|
|
1701 char_u *lp;
|
|
1702
|
|
1703 ret = regnode(EXACTLY);
|
|
1704 lp = reg_prev_sub;
|
|
1705 while (*lp != NUL)
|
|
1706 regc(*lp++);
|
|
1707 regc(NUL);
|
|
1708 if (*reg_prev_sub != NUL)
|
|
1709 {
|
|
1710 *flagp |= HASWIDTH;
|
|
1711 if ((lp - reg_prev_sub) == 1)
|
|
1712 *flagp |= SIMPLE;
|
|
1713 }
|
|
1714 }
|
|
1715 else
|
|
1716 EMSG_RET_NULL(_(e_nopresub));
|
|
1717 break;
|
|
1718
|
|
1719 case Magic('1'):
|
|
1720 case Magic('2'):
|
|
1721 case Magic('3'):
|
|
1722 case Magic('4'):
|
|
1723 case Magic('5'):
|
|
1724 case Magic('6'):
|
|
1725 case Magic('7'):
|
|
1726 case Magic('8'):
|
|
1727 case Magic('9'):
|
|
1728 {
|
|
1729 int refnum;
|
|
1730
|
|
1731 refnum = c - Magic('0');
|
|
1732 /*
|
|
1733 * Check if the back reference is legal. We must have seen the
|
|
1734 * close brace.
|
|
1735 * TODO: Should also check that we don't refer to something
|
|
1736 * that is repeated (+*=): what instance of the repetition
|
|
1737 * should we match?
|
|
1738 */
|
|
1739 if (!had_endbrace[refnum])
|
|
1740 {
|
|
1741 /* Trick: check if "@<=" or "@<!" follows, in which case
|
|
1742 * the \1 can appear before the referenced match. */
|
|
1743 for (p = regparse; *p != NUL; ++p)
|
|
1744 if (p[0] == '@' && p[1] == '<'
|
|
1745 && (p[2] == '!' || p[2] == '='))
|
|
1746 break;
|
|
1747 if (*p == NUL)
|
|
1748 EMSG_RET_NULL(_("E65: Illegal back reference"));
|
|
1749 }
|
|
1750 ret = regnode(BACKREF + refnum);
|
|
1751 }
|
|
1752 break;
|
|
1753
|
|
1754 #ifdef FEAT_SYN_HL
|
|
1755 case Magic('z'):
|
|
1756 {
|
|
1757 c = no_Magic(getchr());
|
|
1758 switch (c)
|
|
1759 {
|
|
1760 case '(': if (reg_do_extmatch != REX_SET)
|
|
1761 EMSG_RET_NULL(_("E66: \\z( not allowed here"));
|
|
1762 if (one_exactly)
|
|
1763 EMSG_ONE_RET_NULL;
|
|
1764 ret = reg(REG_ZPAREN, &flags);
|
|
1765 if (ret == NULL)
|
|
1766 return NULL;
|
|
1767 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH);
|
|
1768 re_has_z = REX_SET;
|
|
1769 break;
|
|
1770
|
|
1771 case '1':
|
|
1772 case '2':
|
|
1773 case '3':
|
|
1774 case '4':
|
|
1775 case '5':
|
|
1776 case '6':
|
|
1777 case '7':
|
|
1778 case '8':
|
|
1779 case '9': if (reg_do_extmatch != REX_USE)
|
|
1780 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here"));
|
|
1781 ret = regnode(ZREF + c - '0');
|
|
1782 re_has_z = REX_USE;
|
|
1783 break;
|
|
1784
|
|
1785 case 's': ret = regnode(MOPEN + 0);
|
|
1786 break;
|
|
1787
|
|
1788 case 'e': ret = regnode(MCLOSE + 0);
|
|
1789 break;
|
|
1790
|
|
1791 default: EMSG_RET_NULL(_("E68: Invalid character after \\z"));
|
|
1792 }
|
|
1793 }
|
|
1794 break;
|
|
1795 #endif
|
|
1796
|
|
1797 case Magic('%'):
|
|
1798 {
|
|
1799 c = no_Magic(getchr());
|
|
1800 switch (c)
|
|
1801 {
|
|
1802 /* () without a back reference */
|
|
1803 case '(':
|
|
1804 if (one_exactly)
|
|
1805 EMSG_ONE_RET_NULL;
|
|
1806 ret = reg(REG_NPAREN, &flags);
|
|
1807 if (ret == NULL)
|
|
1808 return NULL;
|
|
1809 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1810 break;
|
|
1811
|
|
1812 /* Catch \%^ and \%$ regardless of where they appear in the
|
|
1813 * pattern -- regardless of whether or not it makes sense. */
|
|
1814 case '^':
|
|
1815 ret = regnode(RE_BOF);
|
|
1816 break;
|
|
1817
|
|
1818 case '$':
|
|
1819 ret = regnode(RE_EOF);
|
|
1820 break;
|
|
1821
|
|
1822 case '#':
|
|
1823 ret = regnode(CURSOR);
|
|
1824 break;
|
|
1825
|
|
1826 /* \%[abc]: Emit as a list of branches, all ending at the last
|
|
1827 * branch which matches nothing. */
|
|
1828 case '[':
|
|
1829 if (one_exactly) /* doesn't nest */
|
|
1830 EMSG_ONE_RET_NULL;
|
|
1831 {
|
|
1832 char_u *lastbranch;
|
|
1833 char_u *lastnode = NULL;
|
|
1834 char_u *br;
|
|
1835
|
|
1836 ret = NULL;
|
|
1837 while ((c = getchr()) != ']')
|
|
1838 {
|
|
1839 if (c == NUL)
|
|
1840 EMSG_M_RET_NULL(_("E69: Missing ] after %s%%["),
|
|
1841 reg_magic == MAGIC_ALL);
|
|
1842 br = regnode(BRANCH);
|
|
1843 if (ret == NULL)
|
|
1844 ret = br;
|
|
1845 else
|
|
1846 regtail(lastnode, br);
|
|
1847
|
|
1848 ungetchr();
|
|
1849 one_exactly = TRUE;
|
|
1850 lastnode = regatom(flagp);
|
|
1851 one_exactly = FALSE;
|
|
1852 if (lastnode == NULL)
|
|
1853 return NULL;
|
|
1854 }
|
|
1855 if (ret == NULL)
|
|
1856 EMSG_M_RET_NULL(_("E70: Empty %s%%[]"),
|
|
1857 reg_magic == MAGIC_ALL);
|
|
1858 lastbranch = regnode(BRANCH);
|
|
1859 br = regnode(NOTHING);
|
|
1860 if (ret != JUST_CALC_SIZE)
|
|
1861 {
|
|
1862 regtail(lastnode, br);
|
|
1863 regtail(lastbranch, br);
|
|
1864 /* connect all branches to the NOTHING
|
|
1865 * branch at the end */
|
|
1866 for (br = ret; br != lastnode; )
|
|
1867 {
|
|
1868 if (OP(br) == BRANCH)
|
|
1869 {
|
|
1870 regtail(br, lastbranch);
|
|
1871 br = OPERAND(br);
|
|
1872 }
|
|
1873 else
|
|
1874 br = regnext(br);
|
|
1875 }
|
|
1876 }
|
|
1877 *flagp &= ~HASWIDTH;
|
|
1878 break;
|
|
1879 }
|
|
1880
|
24
|
1881 case 'd': /* %d123 decimal */
|
|
1882 case 'o': /* %o123 octal */
|
|
1883 case 'x': /* %xab hex 2 */
|
|
1884 case 'u': /* %uabcd hex 4 */
|
|
1885 case 'U': /* %U1234abcd hex 8 */
|
|
1886 {
|
|
1887 int i;
|
|
1888
|
|
1889 switch (c)
|
|
1890 {
|
|
1891 case 'd': i = getdecchrs(); break;
|
|
1892 case 'o': i = getoctchrs(); break;
|
|
1893 case 'x': i = gethexchrs(2); break;
|
|
1894 case 'u': i = gethexchrs(4); break;
|
|
1895 case 'U': i = gethexchrs(8); break;
|
|
1896 default: i = -1; break;
|
|
1897 }
|
|
1898
|
|
1899 if (i < 0)
|
|
1900 EMSG_M_RET_NULL(
|
|
1901 _("E678: Invalid character after %s%%[dxouU]"),
|
|
1902 reg_magic == MAGIC_ALL);
|
|
1903 ret = regnode(EXACTLY);
|
|
1904 if (i == 0)
|
|
1905 regc(0x0a);
|
|
1906 else
|
|
1907 #ifdef FEAT_MBYTE
|
|
1908 regmbc(i);
|
|
1909 #else
|
|
1910 regc(i);
|
|
1911 #endif
|
|
1912 regc(NUL);
|
|
1913 *flagp |= HASWIDTH;
|
|
1914 break;
|
|
1915 }
|
|
1916
|
7
|
1917 default:
|
|
1918 if (VIM_ISDIGIT(c) || c == '<' || c == '>')
|
|
1919 {
|
|
1920 long_u n = 0;
|
|
1921 int cmp;
|
|
1922
|
|
1923 cmp = c;
|
|
1924 if (cmp == '<' || cmp == '>')
|
|
1925 c = getchr();
|
|
1926 while (VIM_ISDIGIT(c))
|
|
1927 {
|
|
1928 n = n * 10 + (c - '0');
|
|
1929 c = getchr();
|
|
1930 }
|
|
1931 if (c == 'l' || c == 'c' || c == 'v')
|
|
1932 {
|
|
1933 if (c == 'l')
|
|
1934 ret = regnode(RE_LNUM);
|
|
1935 else if (c == 'c')
|
|
1936 ret = regnode(RE_COL);
|
|
1937 else
|
|
1938 ret = regnode(RE_VCOL);
|
|
1939 if (ret == JUST_CALC_SIZE)
|
|
1940 regsize += 5;
|
|
1941 else
|
|
1942 {
|
|
1943 /* put the number and the optional
|
|
1944 * comparator after the opcode */
|
|
1945 regcode = re_put_long(regcode, n);
|
|
1946 *regcode++ = cmp;
|
|
1947 }
|
|
1948 break;
|
|
1949 }
|
|
1950 }
|
|
1951
|
|
1952 EMSG_M_RET_NULL(_("E71: Invalid character after %s%%"),
|
|
1953 reg_magic == MAGIC_ALL);
|
|
1954 }
|
|
1955 }
|
|
1956 break;
|
|
1957
|
|
1958 case Magic('['):
|
|
1959 collection:
|
|
1960 {
|
|
1961 char_u *lp;
|
|
1962
|
|
1963 /*
|
|
1964 * If there is no matching ']', we assume the '[' is a normal
|
|
1965 * character. This makes 'incsearch' and ":help [" work.
|
|
1966 */
|
|
1967 lp = skip_anyof(regparse);
|
|
1968 if (*lp == ']') /* there is a matching ']' */
|
|
1969 {
|
|
1970 int startc = -1; /* > 0 when next '-' is a range */
|
|
1971 int endc;
|
|
1972
|
|
1973 /*
|
|
1974 * In a character class, different parsing rules apply.
|
|
1975 * Not even \ is special anymore, nothing is.
|
|
1976 */
|
|
1977 if (*regparse == '^') /* Complement of range. */
|
|
1978 {
|
|
1979 ret = regnode(ANYBUT + extra);
|
|
1980 regparse++;
|
|
1981 }
|
|
1982 else
|
|
1983 ret = regnode(ANYOF + extra);
|
|
1984
|
|
1985 /* At the start ']' and '-' mean the literal character. */
|
|
1986 if (*regparse == ']' || *regparse == '-')
|
167
|
1987 {
|
|
1988 startc = *regparse;
|
7
|
1989 regc(*regparse++);
|
167
|
1990 }
|
7
|
1991
|
|
1992 while (*regparse != NUL && *regparse != ']')
|
|
1993 {
|
|
1994 if (*regparse == '-')
|
|
1995 {
|
|
1996 ++regparse;
|
|
1997 /* The '-' is not used for a range at the end and
|
|
1998 * after or before a '\n'. */
|
|
1999 if (*regparse == ']' || *regparse == NUL
|
|
2000 || startc == -1
|
|
2001 || (regparse[0] == '\\' && regparse[1] == 'n'))
|
|
2002 {
|
|
2003 regc('-');
|
|
2004 startc = '-'; /* [--x] is a range */
|
|
2005 }
|
|
2006 else
|
|
2007 {
|
167
|
2008 /* Also accept "a-[.z.]" */
|
|
2009 endc = 0;
|
|
2010 if (*regparse == '[')
|
|
2011 endc = get_coll_element(®parse);
|
|
2012 if (endc == 0)
|
|
2013 {
|
7
|
2014 #ifdef FEAT_MBYTE
|
167
|
2015 if (has_mbyte)
|
|
2016 endc = mb_ptr2char_adv(®parse);
|
|
2017 else
|
7
|
2018 #endif
|
167
|
2019 endc = *regparse++;
|
|
2020 }
|
24
|
2021
|
|
2022 /* Handle \o40, \x20 and \u20AC style sequences */
|
167
|
2023 if (endc == '\\' && !cpo_lit && !cpo_bsl)
|
24
|
2024 endc = coll_get_char();
|
|
2025
|
7
|
2026 if (startc > endc)
|
|
2027 EMSG_RET_NULL(_(e_invrange));
|
|
2028 #ifdef FEAT_MBYTE
|
|
2029 if (has_mbyte && ((*mb_char2len)(startc) > 1
|
|
2030 || (*mb_char2len)(endc) > 1))
|
|
2031 {
|
|
2032 /* Limit to a range of 256 chars */
|
|
2033 if (endc > startc + 256)
|
|
2034 EMSG_RET_NULL(_(e_invrange));
|
|
2035 while (++startc <= endc)
|
|
2036 regmbc(startc);
|
|
2037 }
|
|
2038 else
|
|
2039 #endif
|
|
2040 {
|
|
2041 #ifdef EBCDIC
|
|
2042 int alpha_only = FALSE;
|
|
2043
|
|
2044 /* for alphabetical range skip the gaps
|
|
2045 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */
|
|
2046 if (isalpha(startc) && isalpha(endc))
|
|
2047 alpha_only = TRUE;
|
|
2048 #endif
|
|
2049 while (++startc <= endc)
|
|
2050 #ifdef EBCDIC
|
|
2051 if (!alpha_only || isalpha(startc))
|
|
2052 #endif
|
|
2053 regc(startc);
|
|
2054 }
|
|
2055 startc = -1;
|
|
2056 }
|
|
2057 }
|
|
2058 /*
|
|
2059 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
|
|
2060 * accepts "\t", "\e", etc., but only when the 'l' flag in
|
|
2061 * 'cpoptions' is not included.
|
167
|
2062 * Posix doesn't recognize backslash at all.
|
7
|
2063 */
|
|
2064 else if (*regparse == '\\'
|
167
|
2065 && !cpo_bsl
|
7
|
2066 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
|
|
2067 || (!cpo_lit
|
|
2068 && vim_strchr(REGEXP_ABBR,
|
|
2069 regparse[1]) != NULL)))
|
|
2070 {
|
|
2071 regparse++;
|
|
2072 if (*regparse == 'n')
|
|
2073 {
|
|
2074 /* '\n' in range: also match NL */
|
|
2075 if (ret != JUST_CALC_SIZE)
|
|
2076 {
|
|
2077 if (*ret == ANYBUT)
|
|
2078 *ret = ANYBUT + ADD_NL;
|
|
2079 else if (*ret == ANYOF)
|
|
2080 *ret = ANYOF + ADD_NL;
|
|
2081 /* else: must have had a \n already */
|
|
2082 }
|
|
2083 *flagp |= HASNL;
|
|
2084 regparse++;
|
|
2085 startc = -1;
|
|
2086 }
|
24
|
2087 else if (*regparse == 'd'
|
|
2088 || *regparse == 'o'
|
|
2089 || *regparse == 'x'
|
|
2090 || *regparse == 'u'
|
|
2091 || *regparse == 'U')
|
|
2092 {
|
|
2093 startc = coll_get_char();
|
|
2094 if (startc == 0)
|
|
2095 regc(0x0a);
|
|
2096 else
|
|
2097 #ifdef FEAT_MBYTE
|
|
2098 regmbc(startc);
|
|
2099 #else
|
|
2100 regc(startc);
|
|
2101 #endif
|
|
2102 }
|
7
|
2103 else
|
|
2104 {
|
|
2105 startc = backslash_trans(*regparse++);
|
|
2106 regc(startc);
|
|
2107 }
|
|
2108 }
|
|
2109 else if (*regparse == '[')
|
|
2110 {
|
|
2111 int c_class;
|
|
2112 int cu;
|
|
2113
|
167
|
2114 c_class = get_char_class(®parse);
|
7
|
2115 startc = -1;
|
|
2116 /* Characters assumed to be 8 bits! */
|
|
2117 switch (c_class)
|
|
2118 {
|
|
2119 case CLASS_NONE:
|
167
|
2120 c_class = get_equi_class(®parse);
|
|
2121 if (c_class != 0)
|
|
2122 {
|
|
2123 /* produce equivalence class */
|
|
2124 reg_equi_class(c_class);
|
|
2125 }
|
|
2126 else if ((c_class =
|
|
2127 get_coll_element(®parse)) != 0)
|
|
2128 {
|
|
2129 /* produce a collating element */
|
|
2130 regmbc(c_class);
|
|
2131 }
|
|
2132 else
|
|
2133 {
|
|
2134 /* literal '[', allow [[-x] as a range */
|
|
2135 startc = *regparse++;
|
|
2136 regc(startc);
|
|
2137 }
|
7
|
2138 break;
|
|
2139 case CLASS_ALNUM:
|
|
2140 for (cu = 1; cu <= 255; cu++)
|
|
2141 if (isalnum(cu))
|
|
2142 regc(cu);
|
|
2143 break;
|
|
2144 case CLASS_ALPHA:
|
|
2145 for (cu = 1; cu <= 255; cu++)
|
|
2146 if (isalpha(cu))
|
|
2147 regc(cu);
|
|
2148 break;
|
|
2149 case CLASS_BLANK:
|
|
2150 regc(' ');
|
|
2151 regc('\t');
|
|
2152 break;
|
|
2153 case CLASS_CNTRL:
|
|
2154 for (cu = 1; cu <= 255; cu++)
|
|
2155 if (iscntrl(cu))
|
|
2156 regc(cu);
|
|
2157 break;
|
|
2158 case CLASS_DIGIT:
|
|
2159 for (cu = 1; cu <= 255; cu++)
|
|
2160 if (VIM_ISDIGIT(cu))
|
|
2161 regc(cu);
|
|
2162 break;
|
|
2163 case CLASS_GRAPH:
|
|
2164 for (cu = 1; cu <= 255; cu++)
|
|
2165 if (isgraph(cu))
|
|
2166 regc(cu);
|
|
2167 break;
|
|
2168 case CLASS_LOWER:
|
|
2169 for (cu = 1; cu <= 255; cu++)
|
|
2170 if (islower(cu))
|
|
2171 regc(cu);
|
|
2172 break;
|
|
2173 case CLASS_PRINT:
|
|
2174 for (cu = 1; cu <= 255; cu++)
|
|
2175 if (vim_isprintc(cu))
|
|
2176 regc(cu);
|
|
2177 break;
|
|
2178 case CLASS_PUNCT:
|
|
2179 for (cu = 1; cu <= 255; cu++)
|
|
2180 if (ispunct(cu))
|
|
2181 regc(cu);
|
|
2182 break;
|
|
2183 case CLASS_SPACE:
|
|
2184 for (cu = 9; cu <= 13; cu++)
|
|
2185 regc(cu);
|
|
2186 regc(' ');
|
|
2187 break;
|
|
2188 case CLASS_UPPER:
|
|
2189 for (cu = 1; cu <= 255; cu++)
|
|
2190 if (isupper(cu))
|
|
2191 regc(cu);
|
|
2192 break;
|
|
2193 case CLASS_XDIGIT:
|
|
2194 for (cu = 1; cu <= 255; cu++)
|
|
2195 if (vim_isxdigit(cu))
|
|
2196 regc(cu);
|
|
2197 break;
|
|
2198 case CLASS_TAB:
|
|
2199 regc('\t');
|
|
2200 break;
|
|
2201 case CLASS_RETURN:
|
|
2202 regc('\r');
|
|
2203 break;
|
|
2204 case CLASS_BACKSPACE:
|
|
2205 regc('\b');
|
|
2206 break;
|
|
2207 case CLASS_ESCAPE:
|
|
2208 regc('\033');
|
|
2209 break;
|
|
2210 }
|
|
2211 }
|
|
2212 else
|
|
2213 {
|
|
2214 #ifdef FEAT_MBYTE
|
|
2215 if (has_mbyte)
|
|
2216 {
|
|
2217 int len;
|
|
2218
|
|
2219 /* produce a multibyte character, including any
|
|
2220 * following composing characters */
|
|
2221 startc = mb_ptr2char(regparse);
|
|
2222 len = (*mb_ptr2len_check)(regparse);
|
|
2223 if (enc_utf8 && utf_char2len(startc) != len)
|
|
2224 startc = -1; /* composing chars */
|
|
2225 while (--len >= 0)
|
|
2226 regc(*regparse++);
|
|
2227 }
|
|
2228 else
|
|
2229 #endif
|
|
2230 {
|
|
2231 startc = *regparse++;
|
|
2232 regc(startc);
|
|
2233 }
|
|
2234 }
|
|
2235 }
|
|
2236 regc(NUL);
|
|
2237 prevchr_len = 1; /* last char was the ']' */
|
|
2238 if (*regparse != ']')
|
|
2239 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */
|
|
2240 skipchr(); /* let's be friends with the lexer again */
|
|
2241 *flagp |= HASWIDTH | SIMPLE;
|
|
2242 break;
|
|
2243 }
|
|
2244 }
|
|
2245 /* FALLTHROUGH */
|
|
2246
|
|
2247 default:
|
|
2248 {
|
|
2249 int len;
|
|
2250
|
|
2251 #ifdef FEAT_MBYTE
|
|
2252 /* A multi-byte character is handled as a separate atom if it's
|
|
2253 * before a multi. */
|
|
2254 if (has_mbyte && (*mb_char2len)(c) > 1
|
|
2255 && re_multi_type(peekchr()) != NOT_MULTI)
|
|
2256 {
|
|
2257 ret = regnode(MULTIBYTECODE);
|
|
2258 regmbc(c);
|
|
2259 *flagp |= HASWIDTH | SIMPLE;
|
|
2260 break;
|
|
2261 }
|
|
2262 #endif
|
|
2263
|
|
2264 ret = regnode(EXACTLY);
|
|
2265
|
|
2266 /*
|
|
2267 * Append characters as long as:
|
|
2268 * - there is no following multi, we then need the character in
|
|
2269 * front of it as a single character operand
|
|
2270 * - not running into a Magic character
|
|
2271 * - "one_exactly" is not set
|
|
2272 * But always emit at least one character. Might be a Multi,
|
|
2273 * e.g., a "[" without matching "]".
|
|
2274 */
|
|
2275 for (len = 0; c != NUL && (len == 0
|
|
2276 || (re_multi_type(peekchr()) == NOT_MULTI
|
|
2277 && !one_exactly
|
|
2278 && !is_Magic(c))); ++len)
|
|
2279 {
|
|
2280 c = no_Magic(c);
|
|
2281 #ifdef FEAT_MBYTE
|
|
2282 if (has_mbyte)
|
|
2283 {
|
|
2284 regmbc(c);
|
|
2285 if (enc_utf8)
|
|
2286 {
|
|
2287 int off;
|
|
2288 int l;
|
|
2289
|
|
2290 /* Need to get composing character too, directly
|
|
2291 * access regparse for that, because skipchr() skips
|
|
2292 * over composing chars. */
|
|
2293 ungetchr();
|
|
2294 if (*regparse == '\\' && regparse[1] != NUL)
|
|
2295 off = 1;
|
|
2296 else
|
|
2297 off = 0;
|
|
2298 for (;;)
|
|
2299 {
|
|
2300 l = utf_ptr2len_check(regparse + off);
|
|
2301 if (!UTF_COMPOSINGLIKE(regparse + off,
|
|
2302 regparse + off + l))
|
|
2303 break;
|
|
2304 off += l;
|
|
2305 regmbc(utf_ptr2char(regparse + off));
|
|
2306 }
|
|
2307 skipchr();
|
|
2308 }
|
|
2309 }
|
|
2310 else
|
|
2311 #endif
|
|
2312 regc(c);
|
|
2313 c = getchr();
|
|
2314 }
|
|
2315 ungetchr();
|
|
2316
|
|
2317 regc(NUL);
|
|
2318 *flagp |= HASWIDTH;
|
|
2319 if (len == 1)
|
|
2320 *flagp |= SIMPLE;
|
|
2321 }
|
|
2322 break;
|
|
2323 }
|
|
2324
|
|
2325 return ret;
|
|
2326 }
|
|
2327
|
|
2328 /*
|
|
2329 * emit a node
|
|
2330 * Return pointer to generated code.
|
|
2331 */
|
|
2332 static char_u *
|
|
2333 regnode(op)
|
|
2334 int op;
|
|
2335 {
|
|
2336 char_u *ret;
|
|
2337
|
|
2338 ret = regcode;
|
|
2339 if (ret == JUST_CALC_SIZE)
|
|
2340 regsize += 3;
|
|
2341 else
|
|
2342 {
|
|
2343 *regcode++ = op;
|
|
2344 *regcode++ = NUL; /* Null "next" pointer. */
|
|
2345 *regcode++ = NUL;
|
|
2346 }
|
|
2347 return ret;
|
|
2348 }
|
|
2349
|
|
2350 /*
|
|
2351 * Emit (if appropriate) a byte of code
|
|
2352 */
|
|
2353 static void
|
|
2354 regc(b)
|
|
2355 int b;
|
|
2356 {
|
|
2357 if (regcode == JUST_CALC_SIZE)
|
|
2358 regsize++;
|
|
2359 else
|
|
2360 *regcode++ = b;
|
|
2361 }
|
|
2362
|
|
2363 #ifdef FEAT_MBYTE
|
|
2364 /*
|
|
2365 * Emit (if appropriate) a multi-byte character of code
|
|
2366 */
|
|
2367 static void
|
|
2368 regmbc(c)
|
|
2369 int c;
|
|
2370 {
|
|
2371 if (regcode == JUST_CALC_SIZE)
|
|
2372 regsize += (*mb_char2len)(c);
|
|
2373 else
|
|
2374 regcode += (*mb_char2bytes)(c, regcode);
|
|
2375 }
|
|
2376 #endif
|
|
2377
|
|
2378 /*
|
|
2379 * reginsert - insert an operator in front of already-emitted operand
|
|
2380 *
|
|
2381 * Means relocating the operand.
|
|
2382 */
|
|
2383 static void
|
|
2384 reginsert(op, opnd)
|
|
2385 int op;
|
|
2386 char_u *opnd;
|
|
2387 {
|
|
2388 char_u *src;
|
|
2389 char_u *dst;
|
|
2390 char_u *place;
|
|
2391
|
|
2392 if (regcode == JUST_CALC_SIZE)
|
|
2393 {
|
|
2394 regsize += 3;
|
|
2395 return;
|
|
2396 }
|
|
2397 src = regcode;
|
|
2398 regcode += 3;
|
|
2399 dst = regcode;
|
|
2400 while (src > opnd)
|
|
2401 *--dst = *--src;
|
|
2402
|
|
2403 place = opnd; /* Op node, where operand used to be. */
|
|
2404 *place++ = op;
|
|
2405 *place++ = NUL;
|
|
2406 *place = NUL;
|
|
2407 }
|
|
2408
|
|
2409 /*
|
|
2410 * reginsert_limits - insert an operator in front of already-emitted operand.
|
|
2411 * The operator has the given limit values as operands. Also set next pointer.
|
|
2412 *
|
|
2413 * Means relocating the operand.
|
|
2414 */
|
|
2415 static void
|
|
2416 reginsert_limits(op, minval, maxval, opnd)
|
|
2417 int op;
|
|
2418 long minval;
|
|
2419 long maxval;
|
|
2420 char_u *opnd;
|
|
2421 {
|
|
2422 char_u *src;
|
|
2423 char_u *dst;
|
|
2424 char_u *place;
|
|
2425
|
|
2426 if (regcode == JUST_CALC_SIZE)
|
|
2427 {
|
|
2428 regsize += 11;
|
|
2429 return;
|
|
2430 }
|
|
2431 src = regcode;
|
|
2432 regcode += 11;
|
|
2433 dst = regcode;
|
|
2434 while (src > opnd)
|
|
2435 *--dst = *--src;
|
|
2436
|
|
2437 place = opnd; /* Op node, where operand used to be. */
|
|
2438 *place++ = op;
|
|
2439 *place++ = NUL;
|
|
2440 *place++ = NUL;
|
|
2441 place = re_put_long(place, (long_u)minval);
|
|
2442 place = re_put_long(place, (long_u)maxval);
|
|
2443 regtail(opnd, place);
|
|
2444 }
|
|
2445
|
|
2446 /*
|
|
2447 * Write a long as four bytes at "p" and return pointer to the next char.
|
|
2448 */
|
|
2449 static char_u *
|
|
2450 re_put_long(p, val)
|
|
2451 char_u *p;
|
|
2452 long_u val;
|
|
2453 {
|
|
2454 *p++ = (char_u) ((val >> 24) & 0377);
|
|
2455 *p++ = (char_u) ((val >> 16) & 0377);
|
|
2456 *p++ = (char_u) ((val >> 8) & 0377);
|
|
2457 *p++ = (char_u) (val & 0377);
|
|
2458 return p;
|
|
2459 }
|
|
2460
|
|
2461 /*
|
|
2462 * regtail - set the next-pointer at the end of a node chain
|
|
2463 */
|
|
2464 static void
|
|
2465 regtail(p, val)
|
|
2466 char_u *p;
|
|
2467 char_u *val;
|
|
2468 {
|
|
2469 char_u *scan;
|
|
2470 char_u *temp;
|
|
2471 int offset;
|
|
2472
|
|
2473 if (p == JUST_CALC_SIZE)
|
|
2474 return;
|
|
2475
|
|
2476 /* Find last node. */
|
|
2477 scan = p;
|
|
2478 for (;;)
|
|
2479 {
|
|
2480 temp = regnext(scan);
|
|
2481 if (temp == NULL)
|
|
2482 break;
|
|
2483 scan = temp;
|
|
2484 }
|
|
2485
|
|
2486 if (OP(scan) == BACK)
|
|
2487 offset = (int)(scan - val);
|
|
2488 else
|
|
2489 offset = (int)(val - scan);
|
|
2490 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377);
|
|
2491 *(scan + 2) = (char_u) (offset & 0377);
|
|
2492 }
|
|
2493
|
|
2494 /*
|
|
2495 * regoptail - regtail on item after a BRANCH; nop if none
|
|
2496 */
|
|
2497 static void
|
|
2498 regoptail(p, val)
|
|
2499 char_u *p;
|
|
2500 char_u *val;
|
|
2501 {
|
|
2502 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */
|
|
2503 if (p == NULL || p == JUST_CALC_SIZE
|
|
2504 || (OP(p) != BRANCH
|
|
2505 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9)))
|
|
2506 return;
|
|
2507 regtail(OPERAND(p), val);
|
|
2508 }
|
|
2509
|
|
2510 /*
|
|
2511 * getchr() - get the next character from the pattern. We know about
|
|
2512 * magic and such, so therefore we need a lexical analyzer.
|
|
2513 */
|
|
2514
|
|
2515 /* static int curchr; */
|
|
2516 static int prevprevchr;
|
|
2517 static int prevchr;
|
|
2518 static int nextchr; /* used for ungetchr() */
|
|
2519 /*
|
|
2520 * Note: prevchr is sometimes -1 when we are not at the start,
|
|
2521 * eg in /[ ^I]^ the pattern was never found even if it existed, because ^ was
|
|
2522 * taken to be magic -- webb
|
|
2523 */
|
|
2524 static int at_start; /* True when on the first character */
|
|
2525 static int prev_at_start; /* True when on the second character */
|
|
2526
|
|
2527 static void
|
|
2528 initchr(str)
|
|
2529 char_u *str;
|
|
2530 {
|
|
2531 regparse = str;
|
|
2532 prevchr_len = 0;
|
|
2533 curchr = prevprevchr = prevchr = nextchr = -1;
|
|
2534 at_start = TRUE;
|
|
2535 prev_at_start = FALSE;
|
|
2536 }
|
|
2537
|
|
2538 static int
|
|
2539 peekchr()
|
|
2540 {
|
167
|
2541 static int after_slash = FALSE;
|
|
2542
|
7
|
2543 if (curchr == -1)
|
|
2544 {
|
|
2545 switch (curchr = regparse[0])
|
|
2546 {
|
|
2547 case '.':
|
|
2548 case '[':
|
|
2549 case '~':
|
|
2550 /* magic when 'magic' is on */
|
|
2551 if (reg_magic >= MAGIC_ON)
|
|
2552 curchr = Magic(curchr);
|
|
2553 break;
|
|
2554 case '(':
|
|
2555 case ')':
|
|
2556 case '{':
|
|
2557 case '%':
|
|
2558 case '+':
|
|
2559 case '=':
|
|
2560 case '?':
|
|
2561 case '@':
|
|
2562 case '!':
|
|
2563 case '&':
|
|
2564 case '|':
|
|
2565 case '<':
|
|
2566 case '>':
|
|
2567 case '#': /* future ext. */
|
|
2568 case '"': /* future ext. */
|
|
2569 case '\'': /* future ext. */
|
|
2570 case ',': /* future ext. */
|
|
2571 case '-': /* future ext. */
|
|
2572 case ':': /* future ext. */
|
|
2573 case ';': /* future ext. */
|
|
2574 case '`': /* future ext. */
|
|
2575 case '/': /* Can't be used in / command */
|
|
2576 /* magic only after "\v" */
|
|
2577 if (reg_magic == MAGIC_ALL)
|
|
2578 curchr = Magic(curchr);
|
|
2579 break;
|
|
2580 case '*':
|
167
|
2581 /* * is not magic as the very first character, eg "?*ptr", when
|
|
2582 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But
|
|
2583 * "\(\*" is not magic, thus must be magic if "after_slash" */
|
|
2584 if (reg_magic >= MAGIC_ON
|
|
2585 && !at_start
|
|
2586 && !(prev_at_start && prevchr == Magic('^'))
|
|
2587 && (after_slash
|
|
2588 || (prevchr != Magic('(')
|
|
2589 && prevchr != Magic('&')
|
|
2590 && prevchr != Magic('|'))))
|
7
|
2591 curchr = Magic('*');
|
|
2592 break;
|
|
2593 case '^':
|
|
2594 /* '^' is only magic as the very first character and if it's after
|
|
2595 * "\(", "\|", "\&' or "\n" */
|
|
2596 if (reg_magic >= MAGIC_OFF
|
|
2597 && (at_start
|
|
2598 || reg_magic == MAGIC_ALL
|
|
2599 || prevchr == Magic('(')
|
|
2600 || prevchr == Magic('|')
|
|
2601 || prevchr == Magic('&')
|
|
2602 || prevchr == Magic('n')
|
|
2603 || (no_Magic(prevchr) == '('
|
|
2604 && prevprevchr == Magic('%'))))
|
|
2605 {
|
|
2606 curchr = Magic('^');
|
|
2607 at_start = TRUE;
|
|
2608 prev_at_start = FALSE;
|
|
2609 }
|
|
2610 break;
|
|
2611 case '$':
|
|
2612 /* '$' is only magic as the very last char and if it's in front of
|
|
2613 * either "\|", "\)", "\&", or "\n" */
|
|
2614 if (reg_magic >= MAGIC_OFF)
|
|
2615 {
|
|
2616 char_u *p = regparse + 1;
|
|
2617
|
|
2618 /* ignore \c \C \m and \M after '$' */
|
|
2619 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C'
|
|
2620 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z'))
|
|
2621 p += 2;
|
|
2622 if (p[0] == NUL
|
|
2623 || (p[0] == '\\'
|
|
2624 && (p[1] == '|' || p[1] == '&' || p[1] == ')'
|
|
2625 || p[1] == 'n'))
|
|
2626 || reg_magic == MAGIC_ALL)
|
|
2627 curchr = Magic('$');
|
|
2628 }
|
|
2629 break;
|
|
2630 case '\\':
|
|
2631 {
|
|
2632 int c = regparse[1];
|
|
2633
|
|
2634 if (c == NUL)
|
|
2635 curchr = '\\'; /* trailing '\' */
|
|
2636 else if (
|
|
2637 #ifdef EBCDIC
|
|
2638 vim_strchr(META, c)
|
|
2639 #else
|
|
2640 c <= '~' && META_flags[c]
|
|
2641 #endif
|
|
2642 )
|
|
2643 {
|
|
2644 /*
|
|
2645 * META contains everything that may be magic sometimes,
|
|
2646 * except ^ and $ ("\^" and "\$" are only magic after
|
|
2647 * "\v"). We now fetch the next character and toggle its
|
|
2648 * magicness. Therefore, \ is so meta-magic that it is
|
|
2649 * not in META.
|
|
2650 */
|
|
2651 curchr = -1;
|
|
2652 prev_at_start = at_start;
|
|
2653 at_start = FALSE; /* be able to say "/\*ptr" */
|
|
2654 ++regparse;
|
167
|
2655 ++after_slash;
|
7
|
2656 peekchr();
|
|
2657 --regparse;
|
167
|
2658 --after_slash;
|
7
|
2659 curchr = toggle_Magic(curchr);
|
|
2660 }
|
|
2661 else if (vim_strchr(REGEXP_ABBR, c))
|
|
2662 {
|
|
2663 /*
|
|
2664 * Handle abbreviations, like "\t" for TAB -- webb
|
|
2665 */
|
|
2666 curchr = backslash_trans(c);
|
|
2667 }
|
|
2668 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^'))
|
|
2669 curchr = toggle_Magic(c);
|
|
2670 else
|
|
2671 {
|
|
2672 /*
|
|
2673 * Next character can never be (made) magic?
|
|
2674 * Then backslashing it won't do anything.
|
|
2675 */
|
|
2676 #ifdef FEAT_MBYTE
|
|
2677 if (has_mbyte)
|
|
2678 curchr = (*mb_ptr2char)(regparse + 1);
|
|
2679 else
|
|
2680 #endif
|
|
2681 curchr = c;
|
|
2682 }
|
|
2683 break;
|
|
2684 }
|
|
2685
|
|
2686 #ifdef FEAT_MBYTE
|
|
2687 default:
|
|
2688 if (has_mbyte)
|
|
2689 curchr = (*mb_ptr2char)(regparse);
|
|
2690 #endif
|
|
2691 }
|
|
2692 }
|
|
2693
|
|
2694 return curchr;
|
|
2695 }
|
|
2696
|
|
2697 /*
|
|
2698 * Eat one lexed character. Do this in a way that we can undo it.
|
|
2699 */
|
|
2700 static void
|
|
2701 skipchr()
|
|
2702 {
|
|
2703 /* peekchr() eats a backslash, do the same here */
|
|
2704 if (*regparse == '\\')
|
|
2705 prevchr_len = 1;
|
|
2706 else
|
|
2707 prevchr_len = 0;
|
|
2708 if (regparse[prevchr_len] != NUL)
|
|
2709 {
|
|
2710 #ifdef FEAT_MBYTE
|
|
2711 if (has_mbyte)
|
|
2712 prevchr_len += (*mb_ptr2len_check)(regparse + prevchr_len);
|
|
2713 else
|
|
2714 #endif
|
|
2715 ++prevchr_len;
|
|
2716 }
|
|
2717 regparse += prevchr_len;
|
|
2718 prev_at_start = at_start;
|
|
2719 at_start = FALSE;
|
|
2720 prevprevchr = prevchr;
|
|
2721 prevchr = curchr;
|
|
2722 curchr = nextchr; /* use previously unget char, or -1 */
|
|
2723 nextchr = -1;
|
|
2724 }
|
|
2725
|
|
2726 /*
|
|
2727 * Skip a character while keeping the value of prev_at_start for at_start.
|
|
2728 * prevchr and prevprevchr are also kept.
|
|
2729 */
|
|
2730 static void
|
|
2731 skipchr_keepstart()
|
|
2732 {
|
|
2733 int as = prev_at_start;
|
|
2734 int pr = prevchr;
|
|
2735 int prpr = prevprevchr;
|
|
2736
|
|
2737 skipchr();
|
|
2738 at_start = as;
|
|
2739 prevchr = pr;
|
|
2740 prevprevchr = prpr;
|
|
2741 }
|
|
2742
|
|
2743 static int
|
|
2744 getchr()
|
|
2745 {
|
|
2746 int chr = peekchr();
|
|
2747
|
|
2748 skipchr();
|
|
2749 return chr;
|
|
2750 }
|
|
2751
|
|
2752 /*
|
|
2753 * put character back. Works only once!
|
|
2754 */
|
|
2755 static void
|
|
2756 ungetchr()
|
|
2757 {
|
|
2758 nextchr = curchr;
|
|
2759 curchr = prevchr;
|
|
2760 prevchr = prevprevchr;
|
|
2761 at_start = prev_at_start;
|
|
2762 prev_at_start = FALSE;
|
|
2763
|
|
2764 /* Backup regparse, so that it's at the same position as before the
|
|
2765 * getchr(). */
|
|
2766 regparse -= prevchr_len;
|
|
2767 }
|
|
2768
|
|
2769 /*
|
29
|
2770 * Get and return the value of the hex string at the current position.
|
|
2771 * Return -1 if there is no valid hex number.
|
|
2772 * The position is updated:
|
24
|
2773 * blahblah\%x20asdf
|
|
2774 * before-^ ^-after
|
|
2775 * The parameter controls the maximum number of input characters. This will be
|
|
2776 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence.
|
|
2777 */
|
|
2778 static int
|
|
2779 gethexchrs(maxinputlen)
|
|
2780 int maxinputlen;
|
|
2781 {
|
|
2782 int nr = 0;
|
|
2783 int c;
|
|
2784 int i;
|
|
2785
|
|
2786 for (i = 0; i < maxinputlen; ++i)
|
|
2787 {
|
|
2788 c = regparse[0];
|
|
2789 if (!vim_isxdigit(c))
|
|
2790 break;
|
|
2791 nr <<= 4;
|
|
2792 nr |= hex2nr(c);
|
|
2793 ++regparse;
|
|
2794 }
|
|
2795
|
|
2796 if (i == 0)
|
|
2797 return -1;
|
|
2798 return nr;
|
|
2799 }
|
|
2800
|
|
2801 /*
|
|
2802 * get and return the value of the decimal string immediately after the
|
|
2803 * current position. Return -1 for invalid. Consumes all digits.
|
|
2804 */
|
|
2805 static int
|
|
2806 getdecchrs()
|
|
2807 {
|
|
2808 int nr = 0;
|
|
2809 int c;
|
|
2810 int i;
|
|
2811
|
|
2812 for (i = 0; ; ++i)
|
|
2813 {
|
|
2814 c = regparse[0];
|
|
2815 if (c < '0' || c > '9')
|
|
2816 break;
|
|
2817 nr *= 10;
|
|
2818 nr += c - '0';
|
|
2819 ++regparse;
|
|
2820 }
|
|
2821
|
|
2822 if (i == 0)
|
|
2823 return -1;
|
|
2824 return nr;
|
|
2825 }
|
|
2826
|
|
2827 /*
|
|
2828 * get and return the value of the octal string immediately after the current
|
|
2829 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle
|
|
2830 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't
|
|
2831 * treat 8 or 9 as recognised characters. Position is updated:
|
|
2832 * blahblah\%o210asdf
|
|
2833 * before-^ ^-after
|
|
2834 */
|
|
2835 static int
|
|
2836 getoctchrs()
|
|
2837 {
|
|
2838 int nr = 0;
|
|
2839 int c;
|
|
2840 int i;
|
|
2841
|
|
2842 for (i = 0; i < 3 && nr < 040; ++i)
|
|
2843 {
|
|
2844 c = regparse[0];
|
|
2845 if (c < '0' || c > '7')
|
|
2846 break;
|
|
2847 nr <<= 3;
|
|
2848 nr |= hex2nr(c);
|
|
2849 ++regparse;
|
|
2850 }
|
|
2851
|
|
2852 if (i == 0)
|
|
2853 return -1;
|
|
2854 return nr;
|
|
2855 }
|
|
2856
|
|
2857 /*
|
|
2858 * Get a number after a backslash that is inside [].
|
|
2859 * When nothing is recognized return a backslash.
|
|
2860 */
|
|
2861 static int
|
|
2862 coll_get_char()
|
|
2863 {
|
|
2864 int nr = -1;
|
|
2865
|
|
2866 switch (*regparse++)
|
|
2867 {
|
|
2868 case 'd': nr = getdecchrs(); break;
|
|
2869 case 'o': nr = getoctchrs(); break;
|
|
2870 case 'x': nr = gethexchrs(2); break;
|
|
2871 case 'u': nr = gethexchrs(4); break;
|
|
2872 case 'U': nr = gethexchrs(8); break;
|
|
2873 }
|
|
2874 if (nr < 0)
|
|
2875 {
|
|
2876 /* If getting the number fails be backwards compatible: the character
|
|
2877 * is a backslash. */
|
|
2878 --regparse;
|
|
2879 nr = '\\';
|
|
2880 }
|
|
2881 return nr;
|
|
2882 }
|
|
2883
|
|
2884 /*
|
7
|
2885 * read_limits - Read two integers to be taken as a minimum and maximum.
|
|
2886 * If the first character is '-', then the range is reversed.
|
|
2887 * Should end with 'end'. If minval is missing, zero is default, if maxval is
|
|
2888 * missing, a very big number is the default.
|
|
2889 */
|
|
2890 static int
|
|
2891 read_limits(minval, maxval)
|
|
2892 long *minval;
|
|
2893 long *maxval;
|
|
2894 {
|
|
2895 int reverse = FALSE;
|
|
2896 char_u *first_char;
|
|
2897 long tmp;
|
|
2898
|
|
2899 if (*regparse == '-')
|
|
2900 {
|
|
2901 /* Starts with '-', so reverse the range later */
|
|
2902 regparse++;
|
|
2903 reverse = TRUE;
|
|
2904 }
|
|
2905 first_char = regparse;
|
|
2906 *minval = getdigits(®parse);
|
|
2907 if (*regparse == ',') /* There is a comma */
|
|
2908 {
|
|
2909 if (vim_isdigit(*++regparse))
|
|
2910 *maxval = getdigits(®parse);
|
|
2911 else
|
|
2912 *maxval = MAX_LIMIT;
|
|
2913 }
|
|
2914 else if (VIM_ISDIGIT(*first_char))
|
|
2915 *maxval = *minval; /* It was \{n} or \{-n} */
|
|
2916 else
|
|
2917 *maxval = MAX_LIMIT; /* It was \{} or \{-} */
|
|
2918 if (*regparse == '\\')
|
|
2919 regparse++; /* Allow either \{...} or \{...\} */
|
167
|
2920 if (*regparse != '}')
|
7
|
2921 {
|
|
2922 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"),
|
|
2923 reg_magic == MAGIC_ALL ? "" : "\\");
|
|
2924 EMSG_RET_FAIL(IObuff);
|
|
2925 }
|
|
2926
|
|
2927 /*
|
|
2928 * Reverse the range if there was a '-', or make sure it is in the right
|
|
2929 * order otherwise.
|
|
2930 */
|
|
2931 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval))
|
|
2932 {
|
|
2933 tmp = *minval;
|
|
2934 *minval = *maxval;
|
|
2935 *maxval = tmp;
|
|
2936 }
|
|
2937 skipchr(); /* let's be friends with the lexer again */
|
|
2938 return OK;
|
|
2939 }
|
|
2940
|
|
2941 /*
|
|
2942 * vim_regexec and friends
|
|
2943 */
|
|
2944
|
|
2945 /*
|
|
2946 * Global work variables for vim_regexec().
|
|
2947 */
|
|
2948
|
|
2949 /* The current match-position is remembered with these variables: */
|
|
2950 static linenr_T reglnum; /* line number, relative to first line */
|
|
2951 static char_u *regline; /* start of current line */
|
|
2952 static char_u *reginput; /* current input, points into "regline" */
|
|
2953
|
|
2954 static int need_clear_subexpr; /* subexpressions still need to be
|
|
2955 * cleared */
|
|
2956 #ifdef FEAT_SYN_HL
|
|
2957 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions
|
|
2958 * still need to be cleared */
|
|
2959 #endif
|
|
2960
|
|
2961 static int out_of_stack; /* TRUE when ran out of stack space */
|
|
2962
|
|
2963 /*
|
|
2964 * Structure used to save the current input state, when it needs to be
|
|
2965 * restored after trying a match. Used by reg_save() and reg_restore().
|
|
2966 */
|
|
2967 typedef struct
|
|
2968 {
|
|
2969 union
|
|
2970 {
|
|
2971 char_u *ptr; /* reginput pointer, for single-line regexp */
|
|
2972 lpos_T pos; /* reginput pos, for multi-line regexp */
|
|
2973 } rs_u;
|
|
2974 } regsave_T;
|
|
2975
|
|
2976 /* struct to save start/end pointer/position in for \(\) */
|
|
2977 typedef struct
|
|
2978 {
|
|
2979 union
|
|
2980 {
|
|
2981 char_u *ptr;
|
|
2982 lpos_T pos;
|
|
2983 } se_u;
|
|
2984 } save_se_T;
|
|
2985
|
|
2986 static char_u *reg_getline __ARGS((linenr_T lnum));
|
|
2987 static long vim_regexec_both __ARGS((char_u *line, colnr_T col));
|
|
2988 static long regtry __ARGS((regprog_T *prog, colnr_T col));
|
|
2989 static void cleanup_subexpr __ARGS((void));
|
|
2990 #ifdef FEAT_SYN_HL
|
|
2991 static void cleanup_zsubexpr __ARGS((void));
|
|
2992 #endif
|
|
2993 static void reg_nextline __ARGS((void));
|
|
2994 static void reg_save __ARGS((regsave_T *save));
|
|
2995 static void reg_restore __ARGS((regsave_T *save));
|
|
2996 static int reg_save_equal __ARGS((regsave_T *save));
|
|
2997 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp));
|
|
2998 static void save_se_one __ARGS((save_se_T *savep, char_u **pp));
|
|
2999
|
|
3000 /* Save the sub-expressions before attempting a match. */
|
|
3001 #define save_se(savep, posp, pp) \
|
|
3002 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
|
|
3003
|
|
3004 /* After a failed match restore the sub-expressions. */
|
|
3005 #define restore_se(savep, posp, pp) { \
|
|
3006 if (REG_MULTI) \
|
|
3007 *(posp) = (savep)->se_u.pos; \
|
|
3008 else \
|
|
3009 *(pp) = (savep)->se_u.ptr; }
|
|
3010
|
|
3011 static int re_num_cmp __ARGS((long_u val, char_u *scan));
|
167
|
3012 static int regmatch __ARGS((char_u *prog, regsave_T *startp));
|
7
|
3013 static int regrepeat __ARGS((char_u *p, long maxcount));
|
|
3014
|
|
3015 #ifdef DEBUG
|
|
3016 int regnarrate = 0;
|
|
3017 #endif
|
|
3018
|
|
3019 /*
|
|
3020 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec().
|
|
3021 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern
|
|
3022 * contains '\c' or '\C' the value is overruled.
|
|
3023 */
|
|
3024 static int ireg_ic;
|
|
3025
|
|
3026 #ifdef FEAT_MBYTE
|
|
3027 /*
|
|
3028 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag
|
|
3029 * in the regexp. Defaults to false, always.
|
|
3030 */
|
|
3031 static int ireg_icombine;
|
|
3032 #endif
|
|
3033
|
|
3034 /*
|
|
3035 * Sometimes need to save a copy of a line. Since alloc()/free() is very
|
|
3036 * slow, we keep one allocated piece of memory and only re-allocate it when
|
|
3037 * it's too small. It's freed in vim_regexec_both() when finished.
|
|
3038 */
|
|
3039 static char_u *reg_tofree;
|
|
3040 static unsigned reg_tofreelen;
|
|
3041
|
|
3042 /*
|
|
3043 * These variables are set when executing a regexp to speed up the execution.
|
|
3044 * Which ones are set depends on whethere a single-line or multi-line match is
|
|
3045 * done:
|
|
3046 * single-line multi-line
|
|
3047 * reg_match ®match_T NULL
|
|
3048 * reg_mmatch NULL ®mmatch_T
|
|
3049 * reg_startp reg_match->startp <invalid>
|
|
3050 * reg_endp reg_match->endp <invalid>
|
|
3051 * reg_startpos <invalid> reg_mmatch->startpos
|
|
3052 * reg_endpos <invalid> reg_mmatch->endpos
|
|
3053 * reg_win NULL window in which to search
|
|
3054 * reg_buf <invalid> buffer in which to search
|
|
3055 * reg_firstlnum <invalid> first line in which to search
|
|
3056 * reg_maxline 0 last line nr
|
|
3057 * reg_line_lbr FALSE or TRUE FALSE
|
|
3058 */
|
|
3059 static regmatch_T *reg_match;
|
|
3060 static regmmatch_T *reg_mmatch;
|
|
3061 static char_u **reg_startp = NULL;
|
|
3062 static char_u **reg_endp = NULL;
|
|
3063 static lpos_T *reg_startpos = NULL;
|
|
3064 static lpos_T *reg_endpos = NULL;
|
|
3065 static win_T *reg_win;
|
|
3066 static buf_T *reg_buf;
|
|
3067 static linenr_T reg_firstlnum;
|
|
3068 static linenr_T reg_maxline;
|
|
3069 static int reg_line_lbr; /* "\n" in string is line break */
|
|
3070
|
|
3071 /*
|
|
3072 * Get pointer to the line "lnum", which is relative to "reg_firstlnum".
|
|
3073 */
|
|
3074 static char_u *
|
|
3075 reg_getline(lnum)
|
|
3076 linenr_T lnum;
|
|
3077 {
|
|
3078 /* when looking behind for a match/no-match lnum is negative. But we
|
|
3079 * can't go before line 1 */
|
|
3080 if (reg_firstlnum + lnum < 1)
|
|
3081 return NULL;
|
|
3082 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE);
|
|
3083 }
|
|
3084
|
|
3085 static regsave_T behind_pos;
|
|
3086
|
|
3087 #ifdef FEAT_SYN_HL
|
|
3088 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */
|
|
3089 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */
|
|
3090 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */
|
|
3091 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */
|
|
3092 #endif
|
|
3093
|
|
3094 /* TRUE if using multi-line regexp. */
|
|
3095 #define REG_MULTI (reg_match == NULL)
|
|
3096
|
|
3097 /*
|
|
3098 * Match a regexp against a string.
|
|
3099 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
3100 * Uses curbuf for line count and 'iskeyword'.
|
|
3101 *
|
|
3102 * Return TRUE if there is a match, FALSE if not.
|
|
3103 */
|
|
3104 int
|
|
3105 vim_regexec(rmp, line, col)
|
|
3106 regmatch_T *rmp;
|
|
3107 char_u *line; /* string to match against */
|
|
3108 colnr_T col; /* column to start looking for match */
|
|
3109 {
|
|
3110 reg_match = rmp;
|
|
3111 reg_mmatch = NULL;
|
|
3112 reg_maxline = 0;
|
|
3113 reg_line_lbr = FALSE;
|
|
3114 reg_win = NULL;
|
|
3115 ireg_ic = rmp->rm_ic;
|
|
3116 #ifdef FEAT_MBYTE
|
|
3117 ireg_icombine = FALSE;
|
|
3118 #endif
|
|
3119 return (vim_regexec_both(line, col) != 0);
|
|
3120 }
|
|
3121
|
39
|
3122 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \
|
|
3123 || defined(FIND_REPLACE_DIALOG) || defined(PROTO)
|
7
|
3124 /*
|
|
3125 * Like vim_regexec(), but consider a "\n" in "line" to be a line break.
|
|
3126 */
|
|
3127 int
|
|
3128 vim_regexec_nl(rmp, line, col)
|
|
3129 regmatch_T *rmp;
|
|
3130 char_u *line; /* string to match against */
|
|
3131 colnr_T col; /* column to start looking for match */
|
|
3132 {
|
|
3133 reg_match = rmp;
|
|
3134 reg_mmatch = NULL;
|
|
3135 reg_maxline = 0;
|
|
3136 reg_line_lbr = TRUE;
|
|
3137 reg_win = NULL;
|
|
3138 ireg_ic = rmp->rm_ic;
|
|
3139 #ifdef FEAT_MBYTE
|
|
3140 ireg_icombine = FALSE;
|
|
3141 #endif
|
|
3142 return (vim_regexec_both(line, col) != 0);
|
|
3143 }
|
|
3144 #endif
|
|
3145
|
|
3146 /*
|
|
3147 * Match a regexp against multiple lines.
|
|
3148 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
3149 * Uses curbuf for line count and 'iskeyword'.
|
|
3150 *
|
|
3151 * Return zero if there is no match. Return number of lines contained in the
|
|
3152 * match otherwise.
|
|
3153 */
|
|
3154 long
|
|
3155 vim_regexec_multi(rmp, win, buf, lnum, col)
|
|
3156 regmmatch_T *rmp;
|
|
3157 win_T *win; /* window in which to search or NULL */
|
|
3158 buf_T *buf; /* buffer in which to search */
|
|
3159 linenr_T lnum; /* nr of line to start looking for match */
|
|
3160 colnr_T col; /* column to start looking for match */
|
|
3161 {
|
|
3162 long r;
|
|
3163 buf_T *save_curbuf = curbuf;
|
|
3164
|
|
3165 reg_match = NULL;
|
|
3166 reg_mmatch = rmp;
|
|
3167 reg_buf = buf;
|
|
3168 reg_win = win;
|
|
3169 reg_firstlnum = lnum;
|
|
3170 reg_maxline = reg_buf->b_ml.ml_line_count - lnum;
|
|
3171 reg_line_lbr = FALSE;
|
|
3172 ireg_ic = rmp->rmm_ic;
|
|
3173 #ifdef FEAT_MBYTE
|
|
3174 ireg_icombine = FALSE;
|
|
3175 #endif
|
|
3176
|
|
3177 /* Need to switch to buffer "buf" to make vim_iswordc() work. */
|
|
3178 curbuf = buf;
|
|
3179 r = vim_regexec_both(NULL, col);
|
|
3180 curbuf = save_curbuf;
|
|
3181
|
|
3182 return r;
|
|
3183 }
|
|
3184
|
|
3185 /*
|
|
3186 * Match a regexp against a string ("line" points to the string) or multiple
|
|
3187 * lines ("line" is NULL, use reg_getline()).
|
|
3188 */
|
|
3189 #ifdef HAVE_SETJMP_H
|
|
3190 static long
|
|
3191 vim_regexec_both(line_arg, col_arg)
|
|
3192 char_u *line_arg;
|
|
3193 colnr_T col_arg; /* column to start looking for match */
|
|
3194 #else
|
|
3195 static long
|
|
3196 vim_regexec_both(line, col)
|
|
3197 char_u *line;
|
|
3198 colnr_T col; /* column to start looking for match */
|
|
3199 #endif
|
|
3200 {
|
|
3201 regprog_T *prog;
|
|
3202 char_u *s;
|
|
3203 long retval;
|
|
3204 #ifdef HAVE_SETJMP_H
|
|
3205 char_u *line;
|
|
3206 colnr_T col;
|
153
|
3207 int did_mch_startjmp = FALSE;
|
7
|
3208 #endif
|
|
3209
|
|
3210 reg_tofree = NULL;
|
|
3211
|
|
3212 #ifdef HAVE_SETJMP_H
|
|
3213 /* Trick to avoid "might be clobbered by `longjmp'" warning from gcc. */
|
|
3214 line = line_arg;
|
|
3215 col = col_arg;
|
|
3216 #endif
|
|
3217 retval = 0L;
|
|
3218
|
|
3219 if (REG_MULTI)
|
|
3220 {
|
|
3221 prog = reg_mmatch->regprog;
|
|
3222 line = reg_getline((linenr_T)0);
|
|
3223 reg_startpos = reg_mmatch->startpos;
|
|
3224 reg_endpos = reg_mmatch->endpos;
|
|
3225 }
|
|
3226 else
|
|
3227 {
|
|
3228 prog = reg_match->regprog;
|
|
3229 reg_startp = reg_match->startp;
|
|
3230 reg_endp = reg_match->endp;
|
|
3231 }
|
|
3232
|
|
3233 /* Be paranoid... */
|
|
3234 if (prog == NULL || line == NULL)
|
|
3235 {
|
|
3236 EMSG(_(e_null));
|
|
3237 goto theend;
|
|
3238 }
|
|
3239
|
|
3240 /* Check validity of program. */
|
|
3241 if (prog_magic_wrong())
|
|
3242 goto theend;
|
|
3243
|
|
3244 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */
|
|
3245 if (prog->regflags & RF_ICASE)
|
|
3246 ireg_ic = TRUE;
|
|
3247 else if (prog->regflags & RF_NOICASE)
|
|
3248 ireg_ic = FALSE;
|
|
3249
|
|
3250 #ifdef FEAT_MBYTE
|
|
3251 /* If pattern contains "\Z" overrule value of ireg_icombine */
|
|
3252 if (prog->regflags & RF_ICOMBINE)
|
|
3253 ireg_icombine = TRUE;
|
|
3254 #endif
|
|
3255
|
|
3256 /* If there is a "must appear" string, look for it. */
|
|
3257 if (prog->regmust != NULL)
|
|
3258 {
|
|
3259 int c;
|
|
3260
|
|
3261 #ifdef FEAT_MBYTE
|
|
3262 if (has_mbyte)
|
|
3263 c = (*mb_ptr2char)(prog->regmust);
|
|
3264 else
|
|
3265 #endif
|
|
3266 c = *prog->regmust;
|
|
3267 s = line + col;
|
|
3268 while ((s = cstrchr(s, c)) != NULL)
|
|
3269 {
|
|
3270 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3271 break; /* Found it. */
|
39
|
3272 mb_ptr_adv(s);
|
7
|
3273 }
|
|
3274 if (s == NULL) /* Not present. */
|
|
3275 goto theend;
|
|
3276 }
|
|
3277
|
153
|
3278 #ifdef HAVE_TRY_EXCEPT
|
|
3279 __try
|
|
3280 {
|
|
3281 #endif
|
|
3282
|
|
3283 #ifdef HAVE_SETJMP_H
|
|
3284 /*
|
|
3285 * Matching with a regexp may cause a very deep recursive call of
|
|
3286 * regmatch(). Vim will crash when running out of stack space. Catch
|
|
3287 * this here if the system supports it.
|
|
3288 * It's a bit slow, do it after the check for "regmust".
|
|
3289 * Don't do it if the caller already set it up.
|
|
3290 */
|
|
3291 if (!lc_active)
|
|
3292 {
|
|
3293 did_mch_startjmp = TRUE;
|
|
3294 mch_startjmp();
|
|
3295 if (SETJMP(lc_jump_env) != 0)
|
|
3296 {
|
|
3297 mch_didjmp();
|
|
3298 # ifdef SIGHASARG
|
|
3299 if (lc_signal != SIGINT)
|
|
3300 # endif
|
|
3301 EMSG(_(e_complex));
|
|
3302 retval = 0L;
|
|
3303 goto inner_end;
|
|
3304 }
|
|
3305 }
|
|
3306 #endif
|
|
3307
|
7
|
3308 regline = line;
|
|
3309 reglnum = 0;
|
|
3310 out_of_stack = FALSE;
|
|
3311
|
|
3312 /* Simplest case: Anchored match need be tried only once. */
|
|
3313 if (prog->reganch)
|
|
3314 {
|
|
3315 int c;
|
|
3316
|
|
3317 #ifdef FEAT_MBYTE
|
|
3318 if (has_mbyte)
|
|
3319 c = (*mb_ptr2char)(regline + col);
|
|
3320 else
|
|
3321 #endif
|
|
3322 c = regline[col];
|
|
3323 if (prog->regstart == NUL
|
|
3324 || prog->regstart == c
|
|
3325 || (ireg_ic && ((
|
|
3326 #ifdef FEAT_MBYTE
|
|
3327 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
|
|
3328 || (c < 255 && prog->regstart < 255 &&
|
|
3329 #endif
|
|
3330 TOLOWER_LOC(prog->regstart) == TOLOWER_LOC(c)))))
|
|
3331 retval = regtry(prog, col);
|
|
3332 else
|
|
3333 retval = 0;
|
|
3334 }
|
|
3335 else
|
|
3336 {
|
|
3337 /* Messy cases: unanchored match. */
|
|
3338 while (!got_int && !out_of_stack)
|
|
3339 {
|
|
3340 if (prog->regstart != NUL)
|
|
3341 {
|
|
3342 /* Skip until the char we know it must start with. */
|
|
3343 s = cstrchr(regline + col, prog->regstart);
|
|
3344 if (s == NULL)
|
|
3345 {
|
|
3346 retval = 0;
|
|
3347 break;
|
|
3348 }
|
|
3349 col = (int)(s - regline);
|
|
3350 }
|
|
3351
|
|
3352 retval = regtry(prog, col);
|
|
3353 if (retval > 0)
|
|
3354 break;
|
|
3355
|
|
3356 /* if not currently on the first line, get it again */
|
|
3357 if (reglnum != 0)
|
|
3358 {
|
|
3359 regline = reg_getline((linenr_T)0);
|
|
3360 reglnum = 0;
|
|
3361 }
|
|
3362 if (regline[col] == NUL)
|
|
3363 break;
|
|
3364 #ifdef FEAT_MBYTE
|
|
3365 if (has_mbyte)
|
|
3366 col += (*mb_ptr2len_check)(regline + col);
|
|
3367 else
|
|
3368 #endif
|
|
3369 ++col;
|
|
3370 }
|
|
3371 }
|
|
3372
|
|
3373 if (out_of_stack)
|
153
|
3374 EMSG(_(e_outofstack));
|
|
3375
|
|
3376 #ifdef HAVE_SETJMP_H
|
|
3377 inner_end:
|
|
3378 ;
|
|
3379 #endif
|
7
|
3380 #ifdef HAVE_TRY_EXCEPT
|
|
3381 }
|
|
3382 __except(EXCEPTION_EXECUTE_HANDLER)
|
|
3383 {
|
|
3384 if (GetExceptionCode() == EXCEPTION_STACK_OVERFLOW)
|
|
3385 {
|
|
3386 RESETSTKOFLW();
|
153
|
3387 EMSG(_(e_outofstack));
|
7
|
3388 }
|
|
3389 else
|
153
|
3390 EMSG(_(e_complex));
|
7
|
3391 retval = 0L;
|
|
3392 }
|
|
3393 #endif
|
153
|
3394 #ifdef HAVE_SETJMP_H
|
|
3395 if (did_mch_startjmp)
|
|
3396 mch_endjmp();
|
|
3397 #endif
|
7
|
3398
|
|
3399 theend:
|
|
3400 /* Didn't find a match. */
|
|
3401 vim_free(reg_tofree);
|
|
3402 return retval;
|
|
3403 }
|
|
3404
|
|
3405 #ifdef FEAT_SYN_HL
|
|
3406 static reg_extmatch_T *make_extmatch __ARGS((void));
|
|
3407
|
|
3408 /*
|
|
3409 * Create a new extmatch and mark it as referenced once.
|
|
3410 */
|
|
3411 static reg_extmatch_T *
|
|
3412 make_extmatch()
|
|
3413 {
|
|
3414 reg_extmatch_T *em;
|
|
3415
|
|
3416 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T));
|
|
3417 if (em != NULL)
|
|
3418 em->refcnt = 1;
|
|
3419 return em;
|
|
3420 }
|
|
3421
|
|
3422 /*
|
|
3423 * Add a reference to an extmatch.
|
|
3424 */
|
|
3425 reg_extmatch_T *
|
|
3426 ref_extmatch(em)
|
|
3427 reg_extmatch_T *em;
|
|
3428 {
|
|
3429 if (em != NULL)
|
|
3430 em->refcnt++;
|
|
3431 return em;
|
|
3432 }
|
|
3433
|
|
3434 /*
|
|
3435 * Remove a reference to an extmatch. If there are no references left, free
|
|
3436 * the info.
|
|
3437 */
|
|
3438 void
|
|
3439 unref_extmatch(em)
|
|
3440 reg_extmatch_T *em;
|
|
3441 {
|
|
3442 int i;
|
|
3443
|
|
3444 if (em != NULL && --em->refcnt <= 0)
|
|
3445 {
|
|
3446 for (i = 0; i < NSUBEXP; ++i)
|
|
3447 vim_free(em->matches[i]);
|
|
3448 vim_free(em);
|
|
3449 }
|
|
3450 }
|
|
3451 #endif
|
|
3452
|
|
3453 /*
|
|
3454 * regtry - try match of "prog" with at regline["col"].
|
|
3455 * Returns 0 for failure, number of lines contained in the match otherwise.
|
|
3456 */
|
|
3457 static long
|
|
3458 regtry(prog, col)
|
|
3459 regprog_T *prog;
|
|
3460 colnr_T col;
|
|
3461 {
|
|
3462 reginput = regline + col;
|
|
3463 need_clear_subexpr = TRUE;
|
|
3464 #ifdef FEAT_SYN_HL
|
|
3465 /* Clear the external match subpointers if necessary. */
|
|
3466 if (prog->reghasz == REX_SET)
|
|
3467 need_clear_zsubexpr = TRUE;
|
|
3468 #endif
|
|
3469
|
167
|
3470 if (regmatch(prog->program + 1, NULL))
|
7
|
3471 {
|
|
3472 cleanup_subexpr();
|
|
3473 if (REG_MULTI)
|
|
3474 {
|
|
3475 if (reg_startpos[0].lnum < 0)
|
|
3476 {
|
|
3477 reg_startpos[0].lnum = 0;
|
|
3478 reg_startpos[0].col = col;
|
|
3479 }
|
|
3480 if (reg_endpos[0].lnum < 0)
|
|
3481 {
|
|
3482 reg_endpos[0].lnum = reglnum;
|
|
3483 reg_endpos[0].col = (int)(reginput - regline);
|
|
3484 }
|
|
3485 else
|
|
3486 /* Use line number of "\ze". */
|
|
3487 reglnum = reg_endpos[0].lnum;
|
|
3488 }
|
|
3489 else
|
|
3490 {
|
|
3491 if (reg_startp[0] == NULL)
|
|
3492 reg_startp[0] = regline + col;
|
|
3493 if (reg_endp[0] == NULL)
|
|
3494 reg_endp[0] = reginput;
|
|
3495 }
|
|
3496 #ifdef FEAT_SYN_HL
|
|
3497 /* Package any found \z(...\) matches for export. Default is none. */
|
|
3498 unref_extmatch(re_extmatch_out);
|
|
3499 re_extmatch_out = NULL;
|
|
3500
|
|
3501 if (prog->reghasz == REX_SET)
|
|
3502 {
|
|
3503 int i;
|
|
3504
|
|
3505 cleanup_zsubexpr();
|
|
3506 re_extmatch_out = make_extmatch();
|
|
3507 for (i = 0; i < NSUBEXP; i++)
|
|
3508 {
|
|
3509 if (REG_MULTI)
|
|
3510 {
|
|
3511 /* Only accept single line matches. */
|
|
3512 if (reg_startzpos[i].lnum >= 0
|
|
3513 && reg_endzpos[i].lnum == reg_startzpos[i].lnum)
|
|
3514 re_extmatch_out->matches[i] =
|
|
3515 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
|
|
3516 + reg_startzpos[i].col,
|
|
3517 reg_endzpos[i].col - reg_startzpos[i].col);
|
|
3518 }
|
|
3519 else
|
|
3520 {
|
|
3521 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
|
|
3522 re_extmatch_out->matches[i] =
|
|
3523 vim_strnsave(reg_startzp[i],
|
|
3524 (int)(reg_endzp[i] - reg_startzp[i]));
|
|
3525 }
|
|
3526 }
|
|
3527 }
|
|
3528 #endif
|
|
3529 return 1 + reglnum;
|
|
3530 }
|
|
3531 return 0;
|
|
3532 }
|
|
3533
|
|
3534 #ifdef FEAT_MBYTE
|
|
3535 static int reg_prev_class __ARGS((void));
|
|
3536
|
|
3537 /*
|
|
3538 * Get class of previous character.
|
|
3539 */
|
|
3540 static int
|
|
3541 reg_prev_class()
|
|
3542 {
|
|
3543 if (reginput > regline)
|
|
3544 return mb_get_class(reginput - 1
|
|
3545 - (*mb_head_off)(regline, reginput - 1));
|
|
3546 return -1;
|
|
3547 }
|
|
3548
|
|
3549 #endif
|
39
|
3550 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput)
|
7
|
3551
|
|
3552 /*
|
|
3553 * The arguments from BRACE_LIMITS are stored here. They are actually local
|
|
3554 * to regmatch(), but they are here to reduce the amount of stack space used
|
|
3555 * (it can be called recursively many times).
|
|
3556 */
|
|
3557 static long bl_minval;
|
|
3558 static long bl_maxval;
|
|
3559
|
|
3560 /*
|
|
3561 * regmatch - main matching routine
|
|
3562 *
|
|
3563 * Conceptually the strategy is simple: Check to see whether the current
|
|
3564 * node matches, call self recursively to see whether the rest matches,
|
|
3565 * and then act accordingly. In practice we make some effort to avoid
|
|
3566 * recursion, in particular by going through "ordinary" nodes (that don't
|
|
3567 * need to know whether the rest of the match failed) by a loop instead of
|
|
3568 * by recursion.
|
|
3569 *
|
|
3570 * Returns TRUE when there is a match. Leaves reginput and reglnum just after
|
|
3571 * the last matched character.
|
|
3572 * Returns FALSE when there is no match. Leaves reginput and reglnum in an
|
|
3573 * undefined state!
|
|
3574 */
|
|
3575 static int
|
167
|
3576 regmatch(scan, startp)
|
7
|
3577 char_u *scan; /* Current node. */
|
167
|
3578 regsave_T *startp; /* start position for BACK */
|
7
|
3579 {
|
|
3580 char_u *next; /* Next node. */
|
|
3581 int op;
|
|
3582 int c;
|
|
3583
|
|
3584 #ifdef HAVE_GETRLIMIT
|
|
3585 /* Check if we are running out of stack space. Could be caused by
|
|
3586 * recursively calling ourselves. */
|
|
3587 if (out_of_stack || mch_stackcheck((char *)&op) == FAIL)
|
|
3588 {
|
|
3589 out_of_stack = TRUE;
|
|
3590 return FALSE;
|
|
3591 }
|
|
3592 #endif
|
|
3593
|
|
3594 /* Some patterns my cause a long time to match, even though they are not
|
|
3595 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */
|
|
3596 fast_breakcheck();
|
|
3597
|
|
3598 #ifdef DEBUG
|
|
3599 if (scan != NULL && regnarrate)
|
|
3600 {
|
|
3601 mch_errmsg(regprop(scan));
|
|
3602 mch_errmsg("(\n");
|
|
3603 }
|
|
3604 #endif
|
|
3605 while (scan != NULL)
|
|
3606 {
|
|
3607 if (got_int || out_of_stack)
|
|
3608 return FALSE;
|
|
3609 #ifdef DEBUG
|
|
3610 if (regnarrate)
|
|
3611 {
|
|
3612 mch_errmsg(regprop(scan));
|
|
3613 mch_errmsg("...\n");
|
|
3614 # ifdef FEAT_SYN_HL
|
|
3615 if (re_extmatch_in != NULL)
|
|
3616 {
|
|
3617 int i;
|
|
3618
|
|
3619 mch_errmsg(_("External submatches:\n"));
|
|
3620 for (i = 0; i < NSUBEXP; i++)
|
|
3621 {
|
|
3622 mch_errmsg(" \"");
|
|
3623 if (re_extmatch_in->matches[i] != NULL)
|
|
3624 mch_errmsg(re_extmatch_in->matches[i]);
|
|
3625 mch_errmsg("\"\n");
|
|
3626 }
|
|
3627 }
|
|
3628 # endif
|
|
3629 }
|
|
3630 #endif
|
|
3631 next = regnext(scan);
|
|
3632
|
|
3633 op = OP(scan);
|
|
3634 /* Check for character class with NL added. */
|
|
3635 if (WITH_NL(op) && *reginput == NUL && reglnum < reg_maxline)
|
|
3636 {
|
|
3637 reg_nextline();
|
|
3638 }
|
|
3639 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n')
|
|
3640 {
|
|
3641 ADVANCE_REGINPUT();
|
|
3642 }
|
|
3643 else
|
|
3644 {
|
|
3645 if (WITH_NL(op))
|
|
3646 op -= ADD_NL;
|
|
3647 #ifdef FEAT_MBYTE
|
|
3648 if (has_mbyte)
|
|
3649 c = (*mb_ptr2char)(reginput);
|
|
3650 else
|
|
3651 #endif
|
|
3652 c = *reginput;
|
|
3653 switch (op)
|
|
3654 {
|
|
3655 case BOL:
|
|
3656 if (reginput != regline)
|
|
3657 return FALSE;
|
|
3658 break;
|
|
3659
|
|
3660 case EOL:
|
|
3661 if (c != NUL)
|
|
3662 return FALSE;
|
|
3663 break;
|
|
3664
|
|
3665 case RE_BOF:
|
|
3666 /* Passing -1 to the getline() function provided for the search
|
|
3667 * should always return NULL if the current line is the first
|
|
3668 * line of the file. */
|
|
3669 if (reglnum != 0 || reginput != regline
|
|
3670 || (REG_MULTI && reg_getline((linenr_T)-1) != NULL))
|
|
3671 return FALSE;
|
|
3672 break;
|
|
3673
|
|
3674 case RE_EOF:
|
|
3675 if (reglnum != reg_maxline || c != NUL)
|
|
3676 return FALSE;
|
|
3677 break;
|
|
3678
|
|
3679 case CURSOR:
|
|
3680 /* Check if the buffer is in a window and compare the
|
|
3681 * reg_win->w_cursor position to the match position. */
|
|
3682 if (reg_win == NULL
|
|
3683 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum)
|
|
3684 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col))
|
|
3685 return FALSE;
|
|
3686 break;
|
|
3687
|
|
3688 case RE_LNUM:
|
|
3689 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum),
|
|
3690 scan))
|
|
3691 return FALSE;
|
|
3692 break;
|
|
3693
|
|
3694 case RE_COL:
|
|
3695 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan))
|
|
3696 return FALSE;
|
|
3697 break;
|
|
3698
|
|
3699 case RE_VCOL:
|
|
3700 if (!re_num_cmp((long_u)win_linetabsize(
|
|
3701 reg_win == NULL ? curwin : reg_win,
|
|
3702 regline, (colnr_T)(reginput - regline)) + 1, scan))
|
|
3703 return FALSE;
|
|
3704 break;
|
|
3705
|
|
3706 case BOW: /* \<word; reginput points to w */
|
|
3707 if (c == NUL) /* Can't match at end of line */
|
|
3708 return FALSE;
|
|
3709 #ifdef FEAT_MBYTE
|
|
3710 if (has_mbyte)
|
|
3711 {
|
|
3712 int this_class;
|
|
3713
|
|
3714 /* Get class of current and previous char (if it exists). */
|
|
3715 this_class = mb_get_class(reginput);
|
|
3716 if (this_class <= 1)
|
|
3717 return FALSE; /* not on a word at all */
|
|
3718 if (reg_prev_class() == this_class)
|
|
3719 return FALSE; /* previous char is in same word */
|
|
3720 }
|
|
3721 #endif
|
|
3722 else
|
|
3723 {
|
|
3724 if (!vim_iswordc(c)
|
|
3725 || (reginput > regline && vim_iswordc(reginput[-1])))
|
|
3726 return FALSE;
|
|
3727 }
|
|
3728 break;
|
|
3729
|
|
3730 case EOW: /* word\>; reginput points after d */
|
|
3731 if (reginput == regline) /* Can't match at start of line */
|
|
3732 return FALSE;
|
|
3733 #ifdef FEAT_MBYTE
|
|
3734 if (has_mbyte)
|
|
3735 {
|
|
3736 int this_class, prev_class;
|
|
3737
|
|
3738 /* Get class of current and previous char (if it exists). */
|
|
3739 this_class = mb_get_class(reginput);
|
|
3740 prev_class = reg_prev_class();
|
|
3741 if (this_class == prev_class)
|
|
3742 return FALSE;
|
|
3743 if (prev_class == 0 || prev_class == 1)
|
|
3744 return FALSE;
|
|
3745 }
|
|
3746 else
|
|
3747 #endif
|
|
3748 {
|
|
3749 if (!vim_iswordc(reginput[-1]))
|
|
3750 return FALSE;
|
|
3751 if (reginput[0] != NUL && vim_iswordc(c))
|
|
3752 return FALSE;
|
|
3753 }
|
|
3754 break; /* Matched with EOW */
|
|
3755
|
|
3756 case ANY:
|
|
3757 if (c == NUL)
|
|
3758 return FALSE;
|
|
3759 ADVANCE_REGINPUT();
|
|
3760 break;
|
|
3761
|
|
3762 case IDENT:
|
|
3763 if (!vim_isIDc(c))
|
|
3764 return FALSE;
|
|
3765 ADVANCE_REGINPUT();
|
|
3766 break;
|
|
3767
|
|
3768 case SIDENT:
|
|
3769 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c))
|
|
3770 return FALSE;
|
|
3771 ADVANCE_REGINPUT();
|
|
3772 break;
|
|
3773
|
|
3774 case KWORD:
|
|
3775 if (!vim_iswordp(reginput))
|
|
3776 return FALSE;
|
|
3777 ADVANCE_REGINPUT();
|
|
3778 break;
|
|
3779
|
|
3780 case SKWORD:
|
|
3781 if (VIM_ISDIGIT(*reginput) || !vim_iswordp(reginput))
|
|
3782 return FALSE;
|
|
3783 ADVANCE_REGINPUT();
|
|
3784 break;
|
|
3785
|
|
3786 case FNAME:
|
|
3787 if (!vim_isfilec(c))
|
|
3788 return FALSE;
|
|
3789 ADVANCE_REGINPUT();
|
|
3790 break;
|
|
3791
|
|
3792 case SFNAME:
|
|
3793 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c))
|
|
3794 return FALSE;
|
|
3795 ADVANCE_REGINPUT();
|
|
3796 break;
|
|
3797
|
|
3798 case PRINT:
|
|
3799 if (ptr2cells(reginput) != 1)
|
|
3800 return FALSE;
|
|
3801 ADVANCE_REGINPUT();
|
|
3802 break;
|
|
3803
|
|
3804 case SPRINT:
|
|
3805 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1)
|
|
3806 return FALSE;
|
|
3807 ADVANCE_REGINPUT();
|
|
3808 break;
|
|
3809
|
|
3810 case WHITE:
|
|
3811 if (!vim_iswhite(c))
|
|
3812 return FALSE;
|
|
3813 ADVANCE_REGINPUT();
|
|
3814 break;
|
|
3815
|
|
3816 case NWHITE:
|
|
3817 if (c == NUL || vim_iswhite(c))
|
|
3818 return FALSE;
|
|
3819 ADVANCE_REGINPUT();
|
|
3820 break;
|
|
3821
|
|
3822 case DIGIT:
|
|
3823 if (!ri_digit(c))
|
|
3824 return FALSE;
|
|
3825 ADVANCE_REGINPUT();
|
|
3826 break;
|
|
3827
|
|
3828 case NDIGIT:
|
|
3829 if (c == NUL || ri_digit(c))
|
|
3830 return FALSE;
|
|
3831 ADVANCE_REGINPUT();
|
|
3832 break;
|
|
3833
|
|
3834 case HEX:
|
|
3835 if (!ri_hex(c))
|
|
3836 return FALSE;
|
|
3837 ADVANCE_REGINPUT();
|
|
3838 break;
|
|
3839
|
|
3840 case NHEX:
|
|
3841 if (c == NUL || ri_hex(c))
|
|
3842 return FALSE;
|
|
3843 ADVANCE_REGINPUT();
|
|
3844 break;
|
|
3845
|
|
3846 case OCTAL:
|
|
3847 if (!ri_octal(c))
|
|
3848 return FALSE;
|
|
3849 ADVANCE_REGINPUT();
|
|
3850 break;
|
|
3851
|
|
3852 case NOCTAL:
|
|
3853 if (c == NUL || ri_octal(c))
|
|
3854 return FALSE;
|
|
3855 ADVANCE_REGINPUT();
|
|
3856 break;
|
|
3857
|
|
3858 case WORD:
|
|
3859 if (!ri_word(c))
|
|
3860 return FALSE;
|
|
3861 ADVANCE_REGINPUT();
|
|
3862 break;
|
|
3863
|
|
3864 case NWORD:
|
|
3865 if (c == NUL || ri_word(c))
|
|
3866 return FALSE;
|
|
3867 ADVANCE_REGINPUT();
|
|
3868 break;
|
|
3869
|
|
3870 case HEAD:
|
|
3871 if (!ri_head(c))
|
|
3872 return FALSE;
|
|
3873 ADVANCE_REGINPUT();
|
|
3874 break;
|
|
3875
|
|
3876 case NHEAD:
|
|
3877 if (c == NUL || ri_head(c))
|
|
3878 return FALSE;
|
|
3879 ADVANCE_REGINPUT();
|
|
3880 break;
|
|
3881
|
|
3882 case ALPHA:
|
|
3883 if (!ri_alpha(c))
|
|
3884 return FALSE;
|
|
3885 ADVANCE_REGINPUT();
|
|
3886 break;
|
|
3887
|
|
3888 case NALPHA:
|
|
3889 if (c == NUL || ri_alpha(c))
|
|
3890 return FALSE;
|
|
3891 ADVANCE_REGINPUT();
|
|
3892 break;
|
|
3893
|
|
3894 case LOWER:
|
|
3895 if (!ri_lower(c))
|
|
3896 return FALSE;
|
|
3897 ADVANCE_REGINPUT();
|
|
3898 break;
|
|
3899
|
|
3900 case NLOWER:
|
|
3901 if (c == NUL || ri_lower(c))
|
|
3902 return FALSE;
|
|
3903 ADVANCE_REGINPUT();
|
|
3904 break;
|
|
3905
|
|
3906 case UPPER:
|
|
3907 if (!ri_upper(c))
|
|
3908 return FALSE;
|
|
3909 ADVANCE_REGINPUT();
|
|
3910 break;
|
|
3911
|
|
3912 case NUPPER:
|
|
3913 if (c == NUL || ri_upper(c))
|
|
3914 return FALSE;
|
|
3915 ADVANCE_REGINPUT();
|
|
3916 break;
|
|
3917
|
|
3918 case EXACTLY:
|
|
3919 {
|
|
3920 int len;
|
|
3921 char_u *opnd;
|
|
3922
|
|
3923 opnd = OPERAND(scan);
|
|
3924 /* Inline the first byte, for speed. */
|
|
3925 if (*opnd != *reginput
|
|
3926 && (!ireg_ic || (
|
|
3927 #ifdef FEAT_MBYTE
|
|
3928 !enc_utf8 &&
|
|
3929 #endif
|
|
3930 TOLOWER_LOC(*opnd) != TOLOWER_LOC(*reginput))))
|
|
3931 return FALSE;
|
|
3932 if (*opnd == NUL)
|
|
3933 {
|
|
3934 /* match empty string always works; happens when "~" is
|
|
3935 * empty. */
|
|
3936 }
|
|
3937 else if (opnd[1] == NUL
|
|
3938 #ifdef FEAT_MBYTE
|
|
3939 && !(enc_utf8 && ireg_ic)
|
|
3940 #endif
|
|
3941 )
|
|
3942 ++reginput; /* matched a single char */
|
|
3943 else
|
|
3944 {
|
|
3945 len = (int)STRLEN(opnd);
|
|
3946 /* Need to match first byte again for multi-byte. */
|
|
3947 if (cstrncmp(opnd, reginput, &len) != 0)
|
|
3948 return FALSE;
|
|
3949 #ifdef FEAT_MBYTE
|
|
3950 /* Check for following composing character. */
|
|
3951 if (enc_utf8 && UTF_COMPOSINGLIKE(reginput, reginput + len))
|
|
3952 {
|
|
3953 /* raaron: This code makes a composing character get
|
|
3954 * ignored, which is the correct behavior (sometimes)
|
|
3955 * for voweled Hebrew texts. */
|
|
3956 if (!ireg_icombine)
|
|
3957 return FALSE;
|
|
3958 }
|
|
3959 else
|
|
3960 #endif
|
|
3961 reginput += len;
|
|
3962 }
|
|
3963 }
|
|
3964 break;
|
|
3965
|
|
3966 case ANYOF:
|
|
3967 case ANYBUT:
|
|
3968 if (c == NUL)
|
|
3969 return FALSE;
|
|
3970 if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF))
|
|
3971 return FALSE;
|
|
3972 ADVANCE_REGINPUT();
|
|
3973 break;
|
|
3974
|
|
3975 #ifdef FEAT_MBYTE
|
|
3976 case MULTIBYTECODE:
|
|
3977 if (has_mbyte)
|
|
3978 {
|
|
3979 int i, len;
|
|
3980 char_u *opnd;
|
|
3981
|
|
3982 opnd = OPERAND(scan);
|
|
3983 /* Safety check (just in case 'encoding' was changed since
|
|
3984 * compiling the program). */
|
|
3985 if ((len = (*mb_ptr2len_check)(opnd)) < 2)
|
|
3986 return FALSE;
|
|
3987 for (i = 0; i < len; ++i)
|
|
3988 if (opnd[i] != reginput[i])
|
|
3989 return FALSE;
|
|
3990 reginput += len;
|
|
3991 }
|
|
3992 else
|
|
3993 return FALSE;
|
|
3994 break;
|
|
3995 #endif
|
|
3996
|
|
3997 case NOTHING:
|
|
3998 break;
|
|
3999
|
|
4000 case BACK:
|
167
|
4001 /* When we run into BACK without matching something non-empty, we
|
|
4002 * fail. */
|
|
4003 if (startp != NULL && reg_save_equal(startp))
|
|
4004 return FALSE;
|
7
|
4005 break;
|
|
4006
|
|
4007 case MOPEN + 0: /* Match start: \zs */
|
|
4008 case MOPEN + 1: /* \( */
|
|
4009 case MOPEN + 2:
|
|
4010 case MOPEN + 3:
|
|
4011 case MOPEN + 4:
|
|
4012 case MOPEN + 5:
|
|
4013 case MOPEN + 6:
|
|
4014 case MOPEN + 7:
|
|
4015 case MOPEN + 8:
|
|
4016 case MOPEN + 9:
|
|
4017 {
|
|
4018 int no;
|
|
4019 save_se_T save;
|
|
4020
|
|
4021 no = op - MOPEN;
|
|
4022 cleanup_subexpr();
|
|
4023 save_se(&save, ®_startpos[no], ®_startp[no]);
|
|
4024
|
167
|
4025 if (regmatch(next, startp))
|
7
|
4026 return TRUE;
|
|
4027
|
|
4028 restore_se(&save, ®_startpos[no], ®_startp[no]);
|
|
4029 return FALSE;
|
|
4030 }
|
|
4031 /* break; Not Reached */
|
|
4032
|
|
4033 case NOPEN: /* \%( */
|
|
4034 case NCLOSE: /* \) after \%( */
|
167
|
4035 if (regmatch(next, startp))
|
7
|
4036 return TRUE;
|
|
4037 return FALSE;
|
|
4038 /* break; Not Reached */
|
|
4039
|
|
4040 #ifdef FEAT_SYN_HL
|
|
4041 case ZOPEN + 1:
|
|
4042 case ZOPEN + 2:
|
|
4043 case ZOPEN + 3:
|
|
4044 case ZOPEN + 4:
|
|
4045 case ZOPEN + 5:
|
|
4046 case ZOPEN + 6:
|
|
4047 case ZOPEN + 7:
|
|
4048 case ZOPEN + 8:
|
|
4049 case ZOPEN + 9:
|
|
4050 {
|
|
4051 int no;
|
|
4052 save_se_T save;
|
|
4053
|
|
4054 no = op - ZOPEN;
|
|
4055 cleanup_zsubexpr();
|
|
4056 save_se(&save, ®_startzpos[no], ®_startzp[no]);
|
|
4057
|
167
|
4058 if (regmatch(next, startp))
|
7
|
4059 return TRUE;
|
|
4060
|
|
4061 restore_se(&save, ®_startzpos[no], ®_startzp[no]);
|
|
4062 return FALSE;
|
|
4063 }
|
|
4064 /* break; Not Reached */
|
|
4065 #endif
|
|
4066
|
|
4067 case MCLOSE + 0: /* Match end: \ze */
|
|
4068 case MCLOSE + 1: /* \) */
|
|
4069 case MCLOSE + 2:
|
|
4070 case MCLOSE + 3:
|
|
4071 case MCLOSE + 4:
|
|
4072 case MCLOSE + 5:
|
|
4073 case MCLOSE + 6:
|
|
4074 case MCLOSE + 7:
|
|
4075 case MCLOSE + 8:
|
|
4076 case MCLOSE + 9:
|
|
4077 {
|
|
4078 int no;
|
|
4079 save_se_T save;
|
|
4080
|
|
4081 no = op - MCLOSE;
|
|
4082 cleanup_subexpr();
|
|
4083 save_se(&save, ®_endpos[no], ®_endp[no]);
|
|
4084
|
167
|
4085 if (regmatch(next, startp))
|
7
|
4086 return TRUE;
|
|
4087
|
|
4088 restore_se(&save, ®_endpos[no], ®_endp[no]);
|
|
4089 return FALSE;
|
|
4090 }
|
|
4091 /* break; Not Reached */
|
|
4092
|
|
4093 #ifdef FEAT_SYN_HL
|
|
4094 case ZCLOSE + 1: /* \) after \z( */
|
|
4095 case ZCLOSE + 2:
|
|
4096 case ZCLOSE + 3:
|
|
4097 case ZCLOSE + 4:
|
|
4098 case ZCLOSE + 5:
|
|
4099 case ZCLOSE + 6:
|
|
4100 case ZCLOSE + 7:
|
|
4101 case ZCLOSE + 8:
|
|
4102 case ZCLOSE + 9:
|
|
4103 {
|
|
4104 int no;
|
|
4105 save_se_T save;
|
|
4106
|
|
4107 no = op - ZCLOSE;
|
|
4108 cleanup_zsubexpr();
|
|
4109 save_se(&save, ®_endzpos[no], ®_endzp[no]);
|
|
4110
|
167
|
4111 if (regmatch(next, startp))
|
7
|
4112 return TRUE;
|
|
4113
|
|
4114 restore_se(&save, ®_endzpos[no], ®_endzp[no]);
|
|
4115 return FALSE;
|
|
4116 }
|
|
4117 /* break; Not Reached */
|
|
4118 #endif
|
|
4119
|
|
4120 case BACKREF + 1:
|
|
4121 case BACKREF + 2:
|
|
4122 case BACKREF + 3:
|
|
4123 case BACKREF + 4:
|
|
4124 case BACKREF + 5:
|
|
4125 case BACKREF + 6:
|
|
4126 case BACKREF + 7:
|
|
4127 case BACKREF + 8:
|
|
4128 case BACKREF + 9:
|
|
4129 {
|
|
4130 int no;
|
|
4131 int len;
|
|
4132 linenr_T clnum;
|
|
4133 colnr_T ccol;
|
|
4134 char_u *p;
|
|
4135
|
|
4136 no = op - BACKREF;
|
|
4137 cleanup_subexpr();
|
|
4138 if (!REG_MULTI) /* Single-line regexp */
|
|
4139 {
|
|
4140 if (reg_endp[no] == NULL)
|
|
4141 {
|
|
4142 /* Backref was not set: Match an empty string. */
|
|
4143 len = 0;
|
|
4144 }
|
|
4145 else
|
|
4146 {
|
|
4147 /* Compare current input with back-ref in the same
|
|
4148 * line. */
|
|
4149 len = (int)(reg_endp[no] - reg_startp[no]);
|
|
4150 if (cstrncmp(reg_startp[no], reginput, &len) != 0)
|
|
4151 return FALSE;
|
|
4152 }
|
|
4153 }
|
|
4154 else /* Multi-line regexp */
|
|
4155 {
|
|
4156 if (reg_endpos[no].lnum < 0)
|
|
4157 {
|
|
4158 /* Backref was not set: Match an empty string. */
|
|
4159 len = 0;
|
|
4160 }
|
|
4161 else
|
|
4162 {
|
|
4163 if (reg_startpos[no].lnum == reglnum
|
|
4164 && reg_endpos[no].lnum == reglnum)
|
|
4165 {
|
|
4166 /* Compare back-ref within the current line. */
|
|
4167 len = reg_endpos[no].col - reg_startpos[no].col;
|
|
4168 if (cstrncmp(regline + reg_startpos[no].col,
|
|
4169 reginput, &len) != 0)
|
|
4170 return FALSE;
|
|
4171 }
|
|
4172 else
|
|
4173 {
|
|
4174 /* Messy situation: Need to compare between two
|
|
4175 * lines. */
|
|
4176 ccol = reg_startpos[no].col;
|
|
4177 clnum = reg_startpos[no].lnum;
|
|
4178 for (;;)
|
|
4179 {
|
|
4180 /* Since getting one line may invalidate
|
|
4181 * the other, need to make copy. Slow! */
|
|
4182 if (regline != reg_tofree)
|
|
4183 {
|
|
4184 len = (int)STRLEN(regline);
|
|
4185 if (reg_tofree == NULL
|
|
4186 || len >= (int)reg_tofreelen)
|
|
4187 {
|
|
4188 len += 50; /* get some extra */
|
|
4189 vim_free(reg_tofree);
|
|
4190 reg_tofree = alloc(len);
|
|
4191 if (reg_tofree == NULL)
|
|
4192 return FALSE; /* out of memory! */
|
|
4193 reg_tofreelen = len;
|
|
4194 }
|
|
4195 STRCPY(reg_tofree, regline);
|
|
4196 reginput = reg_tofree
|
|
4197 + (reginput - regline);
|
|
4198 regline = reg_tofree;
|
|
4199 }
|
|
4200
|
|
4201 /* Get the line to compare with. */
|
|
4202 p = reg_getline(clnum);
|
|
4203 if (clnum == reg_endpos[no].lnum)
|
|
4204 len = reg_endpos[no].col - ccol;
|
|
4205 else
|
|
4206 len = (int)STRLEN(p + ccol);
|
|
4207
|
|
4208 if (cstrncmp(p + ccol, reginput, &len) != 0)
|
|
4209 return FALSE; /* doesn't match */
|
|
4210 if (clnum == reg_endpos[no].lnum)
|
|
4211 break; /* match and at end! */
|
|
4212 if (reglnum == reg_maxline)
|
|
4213 return FALSE; /* text too short */
|
|
4214
|
|
4215 /* Advance to next line. */
|
|
4216 reg_nextline();
|
|
4217 ++clnum;
|
|
4218 ccol = 0;
|
|
4219 if (got_int || out_of_stack)
|
|
4220 return FALSE;
|
|
4221 }
|
|
4222
|
|
4223 /* found a match! Note that regline may now point
|
|
4224 * to a copy of the line, that should not matter. */
|
|
4225 }
|
|
4226 }
|
|
4227 }
|
|
4228
|
|
4229 /* Matched the backref, skip over it. */
|
|
4230 reginput += len;
|
|
4231 }
|
|
4232 break;
|
|
4233
|
|
4234 #ifdef FEAT_SYN_HL
|
|
4235 case ZREF + 1:
|
|
4236 case ZREF + 2:
|
|
4237 case ZREF + 3:
|
|
4238 case ZREF + 4:
|
|
4239 case ZREF + 5:
|
|
4240 case ZREF + 6:
|
|
4241 case ZREF + 7:
|
|
4242 case ZREF + 8:
|
|
4243 case ZREF + 9:
|
|
4244 {
|
|
4245 int no;
|
|
4246 int len;
|
|
4247
|
|
4248 cleanup_zsubexpr();
|
|
4249 no = op - ZREF;
|
|
4250 if (re_extmatch_in != NULL
|
|
4251 && re_extmatch_in->matches[no] != NULL)
|
|
4252 {
|
|
4253 len = (int)STRLEN(re_extmatch_in->matches[no]);
|
|
4254 if (cstrncmp(re_extmatch_in->matches[no],
|
|
4255 reginput, &len) != 0)
|
|
4256 return FALSE;
|
|
4257 reginput += len;
|
|
4258 }
|
|
4259 else
|
|
4260 {
|
|
4261 /* Backref was not set: Match an empty string. */
|
|
4262 }
|
|
4263 }
|
|
4264 break;
|
|
4265 #endif
|
|
4266
|
|
4267 case BRANCH:
|
|
4268 {
|
|
4269 if (OP(next) != BRANCH) /* No choice. */
|
|
4270 next = OPERAND(scan); /* Avoid recursion. */
|
|
4271 else
|
|
4272 {
|
|
4273 regsave_T save;
|
|
4274
|
|
4275 do
|
|
4276 {
|
|
4277 reg_save(&save);
|
167
|
4278 if (regmatch(OPERAND(scan), &save))
|
7
|
4279 return TRUE;
|
|
4280 reg_restore(&save);
|
|
4281 scan = regnext(scan);
|
|
4282 } while (scan != NULL && OP(scan) == BRANCH);
|
|
4283 return FALSE;
|
|
4284 /* NOTREACHED */
|
|
4285 }
|
|
4286 }
|
|
4287 break;
|
|
4288
|
|
4289 case BRACE_LIMITS:
|
|
4290 {
|
|
4291 int no;
|
|
4292
|
|
4293 if (OP(next) == BRACE_SIMPLE)
|
|
4294 {
|
|
4295 bl_minval = OPERAND_MIN(scan);
|
|
4296 bl_maxval = OPERAND_MAX(scan);
|
|
4297 }
|
|
4298 else if (OP(next) >= BRACE_COMPLEX
|
|
4299 && OP(next) < BRACE_COMPLEX + 10)
|
|
4300 {
|
|
4301 no = OP(next) - BRACE_COMPLEX;
|
|
4302 brace_min[no] = OPERAND_MIN(scan);
|
|
4303 brace_max[no] = OPERAND_MAX(scan);
|
|
4304 brace_count[no] = 0;
|
|
4305 }
|
|
4306 else
|
|
4307 {
|
|
4308 EMSG(_(e_internal)); /* Shouldn't happen */
|
|
4309 return FALSE;
|
|
4310 }
|
|
4311 }
|
|
4312 break;
|
|
4313
|
|
4314 case BRACE_COMPLEX + 0:
|
|
4315 case BRACE_COMPLEX + 1:
|
|
4316 case BRACE_COMPLEX + 2:
|
|
4317 case BRACE_COMPLEX + 3:
|
|
4318 case BRACE_COMPLEX + 4:
|
|
4319 case BRACE_COMPLEX + 5:
|
|
4320 case BRACE_COMPLEX + 6:
|
|
4321 case BRACE_COMPLEX + 7:
|
|
4322 case BRACE_COMPLEX + 8:
|
|
4323 case BRACE_COMPLEX + 9:
|
|
4324 {
|
|
4325 int no;
|
|
4326 regsave_T save;
|
|
4327
|
|
4328 no = op - BRACE_COMPLEX;
|
|
4329 ++brace_count[no];
|
|
4330
|
|
4331 /* If not matched enough times yet, try one more */
|
|
4332 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
|
|
4333 ? brace_min[no] : brace_max[no]))
|
|
4334 {
|
|
4335 reg_save(&save);
|
167
|
4336 if (regmatch(OPERAND(scan), &save))
|
7
|
4337 return TRUE;
|
|
4338 reg_restore(&save);
|
|
4339 --brace_count[no]; /* failed, decrement match count */
|
|
4340 return FALSE;
|
|
4341 }
|
|
4342
|
|
4343 /* If matched enough times, may try matching some more */
|
|
4344 if (brace_min[no] <= brace_max[no])
|
|
4345 {
|
|
4346 /* Range is the normal way around, use longest match */
|
|
4347 if (brace_count[no] <= brace_max[no])
|
|
4348 {
|
|
4349 reg_save(&save);
|
167
|
4350 if (regmatch(OPERAND(scan), &save))
|
7
|
4351 return TRUE; /* matched some more times */
|
|
4352 reg_restore(&save);
|
|
4353 --brace_count[no]; /* matched just enough times */
|
167
|
4354 /* { continue with the items after \{} */
|
7
|
4355 }
|
|
4356 }
|
|
4357 else
|
|
4358 {
|
|
4359 /* Range is backwards, use shortest match first */
|
|
4360 if (brace_count[no] <= brace_min[no])
|
|
4361 {
|
|
4362 reg_save(&save);
|
167
|
4363 if (regmatch(next, &save))
|
7
|
4364 return TRUE;
|
|
4365 reg_restore(&save);
|
|
4366 next = OPERAND(scan);
|
|
4367 /* must try to match one more item */
|
|
4368 }
|
|
4369 }
|
|
4370 }
|
|
4371 break;
|
|
4372
|
|
4373 case BRACE_SIMPLE:
|
|
4374 case STAR:
|
|
4375 case PLUS:
|
|
4376 {
|
|
4377 int nextb; /* next byte */
|
|
4378 int nextb_ic; /* next byte reverse case */
|
|
4379 long count;
|
|
4380 regsave_T save;
|
|
4381 long minval;
|
|
4382 long maxval;
|
|
4383
|
|
4384 /*
|
|
4385 * Lookahead to avoid useless match attempts when we know
|
|
4386 * what character comes next.
|
|
4387 */
|
|
4388 if (OP(next) == EXACTLY)
|
|
4389 {
|
|
4390 nextb = *OPERAND(next);
|
|
4391 if (ireg_ic)
|
|
4392 {
|
|
4393 if (isupper(nextb))
|
|
4394 nextb_ic = TOLOWER_LOC(nextb);
|
|
4395 else
|
|
4396 nextb_ic = TOUPPER_LOC(nextb);
|
|
4397 }
|
|
4398 else
|
|
4399 nextb_ic = nextb;
|
|
4400 }
|
|
4401 else
|
|
4402 {
|
|
4403 nextb = NUL;
|
|
4404 nextb_ic = NUL;
|
|
4405 }
|
|
4406 if (op != BRACE_SIMPLE)
|
|
4407 {
|
|
4408 minval = (op == STAR) ? 0 : 1;
|
|
4409 maxval = MAX_LIMIT;
|
|
4410 }
|
|
4411 else
|
|
4412 {
|
|
4413 minval = bl_minval;
|
|
4414 maxval = bl_maxval;
|
|
4415 }
|
|
4416
|
|
4417 /*
|
|
4418 * When maxval > minval, try matching as much as possible, up
|
|
4419 * to maxval. When maxval < minval, try matching at least the
|
|
4420 * minimal number (since the range is backwards, that's also
|
|
4421 * maxval!).
|
|
4422 */
|
|
4423 count = regrepeat(OPERAND(scan), maxval);
|
|
4424 if (got_int)
|
|
4425 return FALSE;
|
|
4426 if (minval <= maxval)
|
|
4427 {
|
|
4428 /* Range is the normal way around, use longest match */
|
|
4429 while (count >= minval)
|
|
4430 {
|
|
4431 /* If it could match, try it. */
|
|
4432 if (nextb == NUL || *reginput == nextb
|
|
4433 || *reginput == nextb_ic)
|
|
4434 {
|
|
4435 reg_save(&save);
|
167
|
4436 if (regmatch(next, startp))
|
7
|
4437 return TRUE;
|
|
4438 reg_restore(&save);
|
|
4439 }
|
|
4440 /* Couldn't or didn't match -- back up one char. */
|
|
4441 if (--count < minval)
|
|
4442 break;
|
|
4443 if (reginput == regline)
|
|
4444 {
|
|
4445 /* backup to last char of previous line */
|
|
4446 --reglnum;
|
|
4447 regline = reg_getline(reglnum);
|
|
4448 /* Just in case regrepeat() didn't count right. */
|
|
4449 if (regline == NULL)
|
|
4450 return FALSE;
|
|
4451 reginput = regline + STRLEN(regline);
|
|
4452 fast_breakcheck();
|
|
4453 if (got_int || out_of_stack)
|
|
4454 return FALSE;
|
|
4455 }
|
|
4456 else
|
39
|
4457 mb_ptr_back(regline, reginput);
|
7
|
4458 }
|
|
4459 }
|
|
4460 else
|
|
4461 {
|
|
4462 /* Range is backwards, use shortest match first.
|
|
4463 * Careful: maxval and minval are exchanged! */
|
|
4464 if (count < maxval)
|
|
4465 return FALSE;
|
|
4466 for (;;)
|
|
4467 {
|
|
4468 /* If it could work, try it. */
|
|
4469 if (nextb == NUL || *reginput == nextb
|
|
4470 || *reginput == nextb_ic)
|
|
4471 {
|
|
4472 reg_save(&save);
|
167
|
4473 if (regmatch(next, &save))
|
7
|
4474 return TRUE;
|
|
4475 reg_restore(&save);
|
|
4476 }
|
|
4477 /* Couldn't or didn't match: try advancing one char. */
|
|
4478 if (count == minval
|
|
4479 || regrepeat(OPERAND(scan), 1L) == 0)
|
|
4480 break;
|
|
4481 ++count;
|
|
4482 if (got_int || out_of_stack)
|
|
4483 return FALSE;
|
|
4484 }
|
|
4485 }
|
|
4486 return FALSE;
|
|
4487 }
|
|
4488 /* break; Not Reached */
|
|
4489
|
|
4490 case NOMATCH:
|
|
4491 {
|
|
4492 regsave_T save;
|
|
4493
|
|
4494 /* If the operand matches, we fail. Otherwise backup and
|
|
4495 * continue with the next item. */
|
|
4496 reg_save(&save);
|
167
|
4497 if (regmatch(OPERAND(scan), startp))
|
7
|
4498 return FALSE;
|
|
4499 reg_restore(&save);
|
|
4500 }
|
|
4501 break;
|
|
4502
|
|
4503 case MATCH:
|
|
4504 case SUBPAT:
|
|
4505 {
|
|
4506 regsave_T save;
|
|
4507
|
|
4508 /* If the operand doesn't match, we fail. Otherwise backup
|
|
4509 * and continue with the next item. */
|
|
4510 reg_save(&save);
|
167
|
4511 if (!regmatch(OPERAND(scan), startp))
|
7
|
4512 return FALSE;
|
|
4513 if (op == MATCH) /* zero-width */
|
|
4514 reg_restore(&save);
|
|
4515 }
|
|
4516 break;
|
|
4517
|
|
4518 case BEHIND:
|
|
4519 case NOBEHIND:
|
|
4520 {
|
|
4521 regsave_T save_after, save_start;
|
|
4522 regsave_T save_behind_pos;
|
|
4523 int needmatch = (op == BEHIND);
|
|
4524
|
|
4525 /*
|
|
4526 * Look back in the input of the operand matches or not. This
|
|
4527 * must be done at every position in the input and checking if
|
|
4528 * the match ends at the current position.
|
|
4529 * First check if the next item matches, that's probably
|
|
4530 * faster.
|
|
4531 */
|
|
4532 reg_save(&save_start);
|
167
|
4533 if (regmatch(next, startp))
|
7
|
4534 {
|
|
4535 /* save the position after the found match for next */
|
|
4536 reg_save(&save_after);
|
|
4537
|
|
4538 /* start looking for a match with operand at the current
|
|
4539 * postion. Go back one character until we find the
|
|
4540 * result, hitting the start of the line or the previous
|
|
4541 * line (for multi-line matching).
|
|
4542 * Set behind_pos to where the match should end, BHPOS
|
|
4543 * will match it. */
|
|
4544 save_behind_pos = behind_pos;
|
|
4545 behind_pos = save_start;
|
|
4546 for (;;)
|
|
4547 {
|
|
4548 reg_restore(&save_start);
|
167
|
4549 if (regmatch(OPERAND(scan), startp)
|
7
|
4550 && reg_save_equal(&behind_pos))
|
|
4551 {
|
|
4552 behind_pos = save_behind_pos;
|
|
4553 /* found a match that ends where "next" started */
|
|
4554 if (needmatch)
|
|
4555 {
|
|
4556 reg_restore(&save_after);
|
|
4557 return TRUE;
|
|
4558 }
|
|
4559 return FALSE;
|
|
4560 }
|
|
4561 /*
|
|
4562 * No match: Go back one character. May go to
|
|
4563 * previous line once.
|
|
4564 */
|
|
4565 if (REG_MULTI)
|
|
4566 {
|
|
4567 if (save_start.rs_u.pos.col == 0)
|
|
4568 {
|
|
4569 if (save_start.rs_u.pos.lnum
|
|
4570 < behind_pos.rs_u.pos.lnum
|
|
4571 || reg_getline(
|
|
4572 --save_start.rs_u.pos.lnum) == NULL)
|
|
4573 break;
|
|
4574 reg_restore(&save_start);
|
|
4575 save_start.rs_u.pos.col =
|
|
4576 (colnr_T)STRLEN(regline);
|
|
4577 }
|
|
4578 else
|
|
4579 --save_start.rs_u.pos.col;
|
|
4580 }
|
|
4581 else
|
|
4582 {
|
|
4583 if (save_start.rs_u.ptr == regline)
|
|
4584 break;
|
|
4585 --save_start.rs_u.ptr;
|
|
4586 }
|
|
4587 }
|
|
4588
|
|
4589 /* NOBEHIND succeeds when no match was found */
|
|
4590 behind_pos = save_behind_pos;
|
|
4591 if (!needmatch)
|
|
4592 {
|
|
4593 reg_restore(&save_after);
|
|
4594 return TRUE;
|
|
4595 }
|
|
4596 }
|
|
4597 return FALSE;
|
|
4598 }
|
|
4599
|
|
4600 case BHPOS:
|
|
4601 if (REG_MULTI)
|
|
4602 {
|
|
4603 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline)
|
|
4604 || behind_pos.rs_u.pos.lnum != reglnum)
|
|
4605 return FALSE;
|
|
4606 }
|
|
4607 else if (behind_pos.rs_u.ptr != reginput)
|
|
4608 return FALSE;
|
|
4609 break;
|
|
4610
|
|
4611 case NEWL:
|
|
4612 if ((c != NUL || reglnum == reg_maxline)
|
|
4613 && (c != '\n' || !reg_line_lbr))
|
|
4614 return FALSE;
|
|
4615 if (reg_line_lbr)
|
|
4616 ADVANCE_REGINPUT();
|
|
4617 else
|
|
4618 reg_nextline();
|
|
4619 break;
|
|
4620
|
|
4621 case END:
|
|
4622 return TRUE; /* Success! */
|
|
4623
|
|
4624 default:
|
|
4625 EMSG(_(e_re_corr));
|
|
4626 #ifdef DEBUG
|
|
4627 printf("Illegal op code %d\n", op);
|
|
4628 #endif
|
|
4629 return FALSE;
|
|
4630 }
|
|
4631 }
|
|
4632
|
|
4633 scan = next;
|
|
4634 }
|
|
4635
|
|
4636 /*
|
|
4637 * We get here only if there's trouble -- normally "case END" is the
|
|
4638 * terminating point.
|
|
4639 */
|
|
4640 EMSG(_(e_re_corr));
|
|
4641 #ifdef DEBUG
|
|
4642 printf("Premature EOL\n");
|
|
4643 #endif
|
|
4644 return FALSE;
|
|
4645 }
|
|
4646
|
|
4647 /*
|
|
4648 * regrepeat - repeatedly match something simple, return how many.
|
|
4649 * Advances reginput (and reglnum) to just after the matched chars.
|
|
4650 */
|
|
4651 static int
|
|
4652 regrepeat(p, maxcount)
|
|
4653 char_u *p;
|
|
4654 long maxcount; /* maximum number of matches allowed */
|
|
4655 {
|
|
4656 long count = 0;
|
|
4657 char_u *scan;
|
|
4658 char_u *opnd;
|
|
4659 int mask;
|
|
4660 int testval = 0;
|
|
4661
|
|
4662 scan = reginput; /* Make local copy of reginput for speed. */
|
|
4663 opnd = OPERAND(p);
|
|
4664 switch (OP(p))
|
|
4665 {
|
|
4666 case ANY:
|
|
4667 case ANY + ADD_NL:
|
|
4668 while (count < maxcount)
|
|
4669 {
|
|
4670 /* Matching anything means we continue until end-of-line (or
|
|
4671 * end-of-file for ANY + ADD_NL), only limited by maxcount. */
|
|
4672 while (*scan != NUL && count < maxcount)
|
|
4673 {
|
|
4674 ++count;
|
39
|
4675 mb_ptr_adv(scan);
|
7
|
4676 }
|
|
4677 if (!WITH_NL(OP(p)) || reglnum == reg_maxline || count == maxcount)
|
|
4678 break;
|
|
4679 ++count; /* count the line-break */
|
|
4680 reg_nextline();
|
|
4681 scan = reginput;
|
|
4682 if (got_int)
|
|
4683 break;
|
|
4684 }
|
|
4685 break;
|
|
4686
|
|
4687 case IDENT:
|
|
4688 case IDENT + ADD_NL:
|
|
4689 testval = TRUE;
|
|
4690 /*FALLTHROUGH*/
|
|
4691 case SIDENT:
|
|
4692 case SIDENT + ADD_NL:
|
|
4693 while (count < maxcount)
|
|
4694 {
|
|
4695 if (vim_isIDc(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
4696 {
|
39
|
4697 mb_ptr_adv(scan);
|
7
|
4698 }
|
|
4699 else if (*scan == NUL)
|
|
4700 {
|
|
4701 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4702 break;
|
|
4703 reg_nextline();
|
|
4704 scan = reginput;
|
|
4705 if (got_int)
|
|
4706 break;
|
|
4707 }
|
|
4708 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4709 ++scan;
|
|
4710 else
|
|
4711 break;
|
|
4712 ++count;
|
|
4713 }
|
|
4714 break;
|
|
4715
|
|
4716 case KWORD:
|
|
4717 case KWORD + ADD_NL:
|
|
4718 testval = TRUE;
|
|
4719 /*FALLTHROUGH*/
|
|
4720 case SKWORD:
|
|
4721 case SKWORD + ADD_NL:
|
|
4722 while (count < maxcount)
|
|
4723 {
|
|
4724 if (vim_iswordp(scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
4725 {
|
39
|
4726 mb_ptr_adv(scan);
|
7
|
4727 }
|
|
4728 else if (*scan == NUL)
|
|
4729 {
|
|
4730 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4731 break;
|
|
4732 reg_nextline();
|
|
4733 scan = reginput;
|
|
4734 if (got_int)
|
|
4735 break;
|
|
4736 }
|
|
4737 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4738 ++scan;
|
|
4739 else
|
|
4740 break;
|
|
4741 ++count;
|
|
4742 }
|
|
4743 break;
|
|
4744
|
|
4745 case FNAME:
|
|
4746 case FNAME + ADD_NL:
|
|
4747 testval = TRUE;
|
|
4748 /*FALLTHROUGH*/
|
|
4749 case SFNAME:
|
|
4750 case SFNAME + ADD_NL:
|
|
4751 while (count < maxcount)
|
|
4752 {
|
|
4753 if (vim_isfilec(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
4754 {
|
39
|
4755 mb_ptr_adv(scan);
|
7
|
4756 }
|
|
4757 else if (*scan == NUL)
|
|
4758 {
|
|
4759 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4760 break;
|
|
4761 reg_nextline();
|
|
4762 scan = reginput;
|
|
4763 if (got_int)
|
|
4764 break;
|
|
4765 }
|
|
4766 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4767 ++scan;
|
|
4768 else
|
|
4769 break;
|
|
4770 ++count;
|
|
4771 }
|
|
4772 break;
|
|
4773
|
|
4774 case PRINT:
|
|
4775 case PRINT + ADD_NL:
|
|
4776 testval = TRUE;
|
|
4777 /*FALLTHROUGH*/
|
|
4778 case SPRINT:
|
|
4779 case SPRINT + ADD_NL:
|
|
4780 while (count < maxcount)
|
|
4781 {
|
|
4782 if (*scan == NUL)
|
|
4783 {
|
|
4784 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4785 break;
|
|
4786 reg_nextline();
|
|
4787 scan = reginput;
|
|
4788 if (got_int)
|
|
4789 break;
|
|
4790 }
|
|
4791 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan)))
|
|
4792 {
|
39
|
4793 mb_ptr_adv(scan);
|
7
|
4794 }
|
|
4795 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4796 ++scan;
|
|
4797 else
|
|
4798 break;
|
|
4799 ++count;
|
|
4800 }
|
|
4801 break;
|
|
4802
|
|
4803 case WHITE:
|
|
4804 case WHITE + ADD_NL:
|
|
4805 testval = mask = RI_WHITE;
|
|
4806 do_class:
|
|
4807 while (count < maxcount)
|
|
4808 {
|
|
4809 #ifdef FEAT_MBYTE
|
|
4810 int l;
|
|
4811 #endif
|
|
4812 if (*scan == NUL)
|
|
4813 {
|
|
4814 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4815 break;
|
|
4816 reg_nextline();
|
|
4817 scan = reginput;
|
|
4818 if (got_int)
|
|
4819 break;
|
|
4820 }
|
|
4821 #ifdef FEAT_MBYTE
|
|
4822 else if (has_mbyte && (l = (*mb_ptr2len_check)(scan)) > 1)
|
|
4823 {
|
|
4824 if (testval != 0)
|
|
4825 break;
|
|
4826 scan += l;
|
|
4827 }
|
|
4828 #endif
|
|
4829 else if ((class_tab[*scan] & mask) == testval)
|
|
4830 ++scan;
|
|
4831 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4832 ++scan;
|
|
4833 else
|
|
4834 break;
|
|
4835 ++count;
|
|
4836 }
|
|
4837 break;
|
|
4838
|
|
4839 case NWHITE:
|
|
4840 case NWHITE + ADD_NL:
|
|
4841 mask = RI_WHITE;
|
|
4842 goto do_class;
|
|
4843 case DIGIT:
|
|
4844 case DIGIT + ADD_NL:
|
|
4845 testval = mask = RI_DIGIT;
|
|
4846 goto do_class;
|
|
4847 case NDIGIT:
|
|
4848 case NDIGIT + ADD_NL:
|
|
4849 mask = RI_DIGIT;
|
|
4850 goto do_class;
|
|
4851 case HEX:
|
|
4852 case HEX + ADD_NL:
|
|
4853 testval = mask = RI_HEX;
|
|
4854 goto do_class;
|
|
4855 case NHEX:
|
|
4856 case NHEX + ADD_NL:
|
|
4857 mask = RI_HEX;
|
|
4858 goto do_class;
|
|
4859 case OCTAL:
|
|
4860 case OCTAL + ADD_NL:
|
|
4861 testval = mask = RI_OCTAL;
|
|
4862 goto do_class;
|
|
4863 case NOCTAL:
|
|
4864 case NOCTAL + ADD_NL:
|
|
4865 mask = RI_OCTAL;
|
|
4866 goto do_class;
|
|
4867 case WORD:
|
|
4868 case WORD + ADD_NL:
|
|
4869 testval = mask = RI_WORD;
|
|
4870 goto do_class;
|
|
4871 case NWORD:
|
|
4872 case NWORD + ADD_NL:
|
|
4873 mask = RI_WORD;
|
|
4874 goto do_class;
|
|
4875 case HEAD:
|
|
4876 case HEAD + ADD_NL:
|
|
4877 testval = mask = RI_HEAD;
|
|
4878 goto do_class;
|
|
4879 case NHEAD:
|
|
4880 case NHEAD + ADD_NL:
|
|
4881 mask = RI_HEAD;
|
|
4882 goto do_class;
|
|
4883 case ALPHA:
|
|
4884 case ALPHA + ADD_NL:
|
|
4885 testval = mask = RI_ALPHA;
|
|
4886 goto do_class;
|
|
4887 case NALPHA:
|
|
4888 case NALPHA + ADD_NL:
|
|
4889 mask = RI_ALPHA;
|
|
4890 goto do_class;
|
|
4891 case LOWER:
|
|
4892 case LOWER + ADD_NL:
|
|
4893 testval = mask = RI_LOWER;
|
|
4894 goto do_class;
|
|
4895 case NLOWER:
|
|
4896 case NLOWER + ADD_NL:
|
|
4897 mask = RI_LOWER;
|
|
4898 goto do_class;
|
|
4899 case UPPER:
|
|
4900 case UPPER + ADD_NL:
|
|
4901 testval = mask = RI_UPPER;
|
|
4902 goto do_class;
|
|
4903 case NUPPER:
|
|
4904 case NUPPER + ADD_NL:
|
|
4905 mask = RI_UPPER;
|
|
4906 goto do_class;
|
|
4907
|
|
4908 case EXACTLY:
|
|
4909 {
|
|
4910 int cu, cl;
|
|
4911
|
|
4912 /* This doesn't do a multi-byte character, because a MULTIBYTECODE
|
|
4913 * would have been used for it. */
|
|
4914 if (ireg_ic)
|
|
4915 {
|
|
4916 cu = TOUPPER_LOC(*opnd);
|
|
4917 cl = TOLOWER_LOC(*opnd);
|
|
4918 while (count < maxcount && (*scan == cu || *scan == cl))
|
|
4919 {
|
|
4920 count++;
|
|
4921 scan++;
|
|
4922 }
|
|
4923 }
|
|
4924 else
|
|
4925 {
|
|
4926 cu = *opnd;
|
|
4927 while (count < maxcount && *scan == cu)
|
|
4928 {
|
|
4929 count++;
|
|
4930 scan++;
|
|
4931 }
|
|
4932 }
|
|
4933 break;
|
|
4934 }
|
|
4935
|
|
4936 #ifdef FEAT_MBYTE
|
|
4937 case MULTIBYTECODE:
|
|
4938 {
|
|
4939 int i, len, cf = 0;
|
|
4940
|
|
4941 /* Safety check (just in case 'encoding' was changed since
|
|
4942 * compiling the program). */
|
|
4943 if ((len = (*mb_ptr2len_check)(opnd)) > 1)
|
|
4944 {
|
|
4945 if (ireg_ic && enc_utf8)
|
|
4946 cf = utf_fold(utf_ptr2char(opnd));
|
|
4947 while (count < maxcount)
|
|
4948 {
|
|
4949 for (i = 0; i < len; ++i)
|
|
4950 if (opnd[i] != scan[i])
|
|
4951 break;
|
|
4952 if (i < len && (!ireg_ic || !enc_utf8
|
|
4953 || utf_fold(utf_ptr2char(scan)) != cf))
|
|
4954 break;
|
|
4955 scan += len;
|
|
4956 ++count;
|
|
4957 }
|
|
4958 }
|
|
4959 }
|
|
4960 break;
|
|
4961 #endif
|
|
4962
|
|
4963 case ANYOF:
|
|
4964 case ANYOF + ADD_NL:
|
|
4965 testval = TRUE;
|
|
4966 /*FALLTHROUGH*/
|
|
4967
|
|
4968 case ANYBUT:
|
|
4969 case ANYBUT + ADD_NL:
|
|
4970 while (count < maxcount)
|
|
4971 {
|
|
4972 #ifdef FEAT_MBYTE
|
|
4973 int len;
|
|
4974 #endif
|
|
4975 if (*scan == NUL)
|
|
4976 {
|
|
4977 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
4978 break;
|
|
4979 reg_nextline();
|
|
4980 scan = reginput;
|
|
4981 if (got_int)
|
|
4982 break;
|
|
4983 }
|
|
4984 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
4985 ++scan;
|
|
4986 #ifdef FEAT_MBYTE
|
|
4987 else if (has_mbyte && (len = (*mb_ptr2len_check)(scan)) > 1)
|
|
4988 {
|
|
4989 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
|
|
4990 break;
|
|
4991 scan += len;
|
|
4992 }
|
|
4993 #endif
|
|
4994 else
|
|
4995 {
|
|
4996 if ((cstrchr(opnd, *scan) == NULL) == testval)
|
|
4997 break;
|
|
4998 ++scan;
|
|
4999 }
|
|
5000 ++count;
|
|
5001 }
|
|
5002 break;
|
|
5003
|
|
5004 case NEWL:
|
|
5005 while (count < maxcount
|
|
5006 && ((*scan == NUL && reglnum < reg_maxline)
|
|
5007 || (*scan == '\n' && reg_line_lbr)))
|
|
5008 {
|
|
5009 count++;
|
|
5010 if (reg_line_lbr)
|
|
5011 ADVANCE_REGINPUT();
|
|
5012 else
|
|
5013 reg_nextline();
|
|
5014 scan = reginput;
|
|
5015 if (got_int)
|
|
5016 break;
|
|
5017 }
|
|
5018 break;
|
|
5019
|
|
5020 default: /* Oh dear. Called inappropriately. */
|
|
5021 EMSG(_(e_re_corr));
|
|
5022 #ifdef DEBUG
|
|
5023 printf("Called regrepeat with op code %d\n", OP(p));
|
|
5024 #endif
|
|
5025 break;
|
|
5026 }
|
|
5027
|
|
5028 reginput = scan;
|
|
5029
|
|
5030 return (int)count;
|
|
5031 }
|
|
5032
|
|
5033 /*
|
|
5034 * regnext - dig the "next" pointer out of a node
|
|
5035 */
|
|
5036 static char_u *
|
|
5037 regnext(p)
|
|
5038 char_u *p;
|
|
5039 {
|
|
5040 int offset;
|
|
5041
|
|
5042 if (p == JUST_CALC_SIZE)
|
|
5043 return NULL;
|
|
5044
|
|
5045 offset = NEXT(p);
|
|
5046 if (offset == 0)
|
|
5047 return NULL;
|
|
5048
|
|
5049 if (OP(p) == BACK)
|
|
5050 return p - offset;
|
|
5051 else
|
|
5052 return p + offset;
|
|
5053 }
|
|
5054
|
|
5055 /*
|
|
5056 * Check the regexp program for its magic number.
|
|
5057 * Return TRUE if it's wrong.
|
|
5058 */
|
|
5059 static int
|
|
5060 prog_magic_wrong()
|
|
5061 {
|
|
5062 if (UCHARAT(REG_MULTI
|
|
5063 ? reg_mmatch->regprog->program
|
|
5064 : reg_match->regprog->program) != REGMAGIC)
|
|
5065 {
|
|
5066 EMSG(_(e_re_corr));
|
|
5067 return TRUE;
|
|
5068 }
|
|
5069 return FALSE;
|
|
5070 }
|
|
5071
|
|
5072 /*
|
|
5073 * Cleanup the subexpressions, if this wasn't done yet.
|
|
5074 * This construction is used to clear the subexpressions only when they are
|
|
5075 * used (to increase speed).
|
|
5076 */
|
|
5077 static void
|
|
5078 cleanup_subexpr()
|
|
5079 {
|
|
5080 if (need_clear_subexpr)
|
|
5081 {
|
|
5082 if (REG_MULTI)
|
|
5083 {
|
|
5084 /* Use 0xff to set lnum to -1 */
|
|
5085 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5086 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5087 }
|
|
5088 else
|
|
5089 {
|
|
5090 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5091 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5092 }
|
|
5093 need_clear_subexpr = FALSE;
|
|
5094 }
|
|
5095 }
|
|
5096
|
|
5097 #ifdef FEAT_SYN_HL
|
|
5098 static void
|
|
5099 cleanup_zsubexpr()
|
|
5100 {
|
|
5101 if (need_clear_zsubexpr)
|
|
5102 {
|
|
5103 if (REG_MULTI)
|
|
5104 {
|
|
5105 /* Use 0xff to set lnum to -1 */
|
|
5106 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5107 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5108 }
|
|
5109 else
|
|
5110 {
|
|
5111 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5112 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5113 }
|
|
5114 need_clear_zsubexpr = FALSE;
|
|
5115 }
|
|
5116 }
|
|
5117 #endif
|
|
5118
|
|
5119 /*
|
|
5120 * Advance reglnum, regline and reginput to the next line.
|
|
5121 */
|
|
5122 static void
|
|
5123 reg_nextline()
|
|
5124 {
|
|
5125 regline = reg_getline(++reglnum);
|
|
5126 reginput = regline;
|
|
5127 fast_breakcheck();
|
|
5128 }
|
|
5129
|
|
5130 /*
|
|
5131 * Save the input line and position in a regsave_T.
|
|
5132 */
|
|
5133 static void
|
|
5134 reg_save(save)
|
|
5135 regsave_T *save;
|
|
5136 {
|
|
5137 if (REG_MULTI)
|
|
5138 {
|
|
5139 save->rs_u.pos.col = (colnr_T)(reginput - regline);
|
|
5140 save->rs_u.pos.lnum = reglnum;
|
|
5141 }
|
|
5142 else
|
|
5143 save->rs_u.ptr = reginput;
|
|
5144 }
|
|
5145
|
|
5146 /*
|
|
5147 * Restore the input line and position from a regsave_T.
|
|
5148 */
|
|
5149 static void
|
|
5150 reg_restore(save)
|
|
5151 regsave_T *save;
|
|
5152 {
|
|
5153 if (REG_MULTI)
|
|
5154 {
|
|
5155 if (reglnum != save->rs_u.pos.lnum)
|
|
5156 {
|
|
5157 /* only call reg_getline() when the line number changed to save
|
|
5158 * a bit of time */
|
|
5159 reglnum = save->rs_u.pos.lnum;
|
|
5160 regline = reg_getline(reglnum);
|
|
5161 }
|
|
5162 reginput = regline + save->rs_u.pos.col;
|
|
5163 }
|
|
5164 else
|
|
5165 reginput = save->rs_u.ptr;
|
|
5166 }
|
|
5167
|
|
5168 /*
|
|
5169 * Return TRUE if current position is equal to saved position.
|
|
5170 */
|
|
5171 static int
|
|
5172 reg_save_equal(save)
|
|
5173 regsave_T *save;
|
|
5174 {
|
|
5175 if (REG_MULTI)
|
|
5176 return reglnum == save->rs_u.pos.lnum
|
|
5177 && reginput == regline + save->rs_u.pos.col;
|
|
5178 return reginput == save->rs_u.ptr;
|
|
5179 }
|
|
5180
|
|
5181 /*
|
|
5182 * Tentatively set the sub-expression start to the current position (after
|
|
5183 * calling regmatch() they will have changed). Need to save the existing
|
|
5184 * values for when there is no match.
|
|
5185 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
|
|
5186 * depending on REG_MULTI.
|
|
5187 */
|
|
5188 static void
|
|
5189 save_se_multi(savep, posp)
|
|
5190 save_se_T *savep;
|
|
5191 lpos_T *posp;
|
|
5192 {
|
|
5193 savep->se_u.pos = *posp;
|
|
5194 posp->lnum = reglnum;
|
|
5195 posp->col = (colnr_T)(reginput - regline);
|
|
5196 }
|
|
5197
|
|
5198 static void
|
|
5199 save_se_one(savep, pp)
|
|
5200 save_se_T *savep;
|
|
5201 char_u **pp;
|
|
5202 {
|
|
5203 savep->se_u.ptr = *pp;
|
|
5204 *pp = reginput;
|
|
5205 }
|
|
5206
|
|
5207 /*
|
|
5208 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
|
|
5209 */
|
|
5210 static int
|
|
5211 re_num_cmp(val, scan)
|
|
5212 long_u val;
|
|
5213 char_u *scan;
|
|
5214 {
|
|
5215 long_u n = OPERAND_MIN(scan);
|
|
5216
|
|
5217 if (OPERAND_CMP(scan) == '>')
|
|
5218 return val > n;
|
|
5219 if (OPERAND_CMP(scan) == '<')
|
|
5220 return val < n;
|
|
5221 return val == n;
|
|
5222 }
|
|
5223
|
|
5224
|
|
5225 #ifdef DEBUG
|
|
5226
|
|
5227 /*
|
|
5228 * regdump - dump a regexp onto stdout in vaguely comprehensible form
|
|
5229 */
|
|
5230 static void
|
|
5231 regdump(pattern, r)
|
|
5232 char_u *pattern;
|
|
5233 regprog_T *r;
|
|
5234 {
|
|
5235 char_u *s;
|
|
5236 int op = EXACTLY; /* Arbitrary non-END op. */
|
|
5237 char_u *next;
|
|
5238 char_u *end = NULL;
|
|
5239
|
|
5240 printf("\r\nregcomp(%s):\r\n", pattern);
|
|
5241
|
|
5242 s = r->program + 1;
|
|
5243 /*
|
|
5244 * Loop until we find the END that isn't before a referred next (an END
|
|
5245 * can also appear in a NOMATCH operand).
|
|
5246 */
|
|
5247 while (op != END || s <= end)
|
|
5248 {
|
|
5249 op = OP(s);
|
|
5250 printf("%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */
|
|
5251 next = regnext(s);
|
|
5252 if (next == NULL) /* Next ptr. */
|
|
5253 printf("(0)");
|
|
5254 else
|
|
5255 printf("(%d)", (int)((s - r->program) + (next - s)));
|
|
5256 if (end < next)
|
|
5257 end = next;
|
|
5258 if (op == BRACE_LIMITS)
|
|
5259 {
|
|
5260 /* Two short ints */
|
|
5261 printf(" minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
|
|
5262 s += 8;
|
|
5263 }
|
|
5264 s += 3;
|
|
5265 if (op == ANYOF || op == ANYOF + ADD_NL
|
|
5266 || op == ANYBUT || op == ANYBUT + ADD_NL
|
|
5267 || op == EXACTLY)
|
|
5268 {
|
|
5269 /* Literal string, where present. */
|
|
5270 while (*s != NUL)
|
|
5271 printf("%c", *s++);
|
|
5272 s++;
|
|
5273 }
|
|
5274 printf("\r\n");
|
|
5275 }
|
|
5276
|
|
5277 /* Header fields of interest. */
|
|
5278 if (r->regstart != NUL)
|
|
5279 printf("start `%s' 0x%x; ", r->regstart < 256
|
|
5280 ? (char *)transchar(r->regstart)
|
|
5281 : "multibyte", r->regstart);
|
|
5282 if (r->reganch)
|
|
5283 printf("anchored; ");
|
|
5284 if (r->regmust != NULL)
|
|
5285 printf("must have \"%s\"", r->regmust);
|
|
5286 printf("\r\n");
|
|
5287 }
|
|
5288
|
|
5289 /*
|
|
5290 * regprop - printable representation of opcode
|
|
5291 */
|
|
5292 static char_u *
|
|
5293 regprop(op)
|
|
5294 char_u *op;
|
|
5295 {
|
|
5296 char_u *p;
|
|
5297 static char_u buf[50];
|
|
5298
|
|
5299 (void) strcpy(buf, ":");
|
|
5300
|
|
5301 switch (OP(op))
|
|
5302 {
|
|
5303 case BOL:
|
|
5304 p = "BOL";
|
|
5305 break;
|
|
5306 case EOL:
|
|
5307 p = "EOL";
|
|
5308 break;
|
|
5309 case RE_BOF:
|
|
5310 p = "BOF";
|
|
5311 break;
|
|
5312 case RE_EOF:
|
|
5313 p = "EOF";
|
|
5314 break;
|
|
5315 case CURSOR:
|
|
5316 p = "CURSOR";
|
|
5317 break;
|
|
5318 case RE_LNUM:
|
|
5319 p = "RE_LNUM";
|
|
5320 break;
|
|
5321 case RE_COL:
|
|
5322 p = "RE_COL";
|
|
5323 break;
|
|
5324 case RE_VCOL:
|
|
5325 p = "RE_VCOL";
|
|
5326 break;
|
|
5327 case BOW:
|
|
5328 p = "BOW";
|
|
5329 break;
|
|
5330 case EOW:
|
|
5331 p = "EOW";
|
|
5332 break;
|
|
5333 case ANY:
|
|
5334 p = "ANY";
|
|
5335 break;
|
|
5336 case ANY + ADD_NL:
|
|
5337 p = "ANY+NL";
|
|
5338 break;
|
|
5339 case ANYOF:
|
|
5340 p = "ANYOF";
|
|
5341 break;
|
|
5342 case ANYOF + ADD_NL:
|
|
5343 p = "ANYOF+NL";
|
|
5344 break;
|
|
5345 case ANYBUT:
|
|
5346 p = "ANYBUT";
|
|
5347 break;
|
|
5348 case ANYBUT + ADD_NL:
|
|
5349 p = "ANYBUT+NL";
|
|
5350 break;
|
|
5351 case IDENT:
|
|
5352 p = "IDENT";
|
|
5353 break;
|
|
5354 case IDENT + ADD_NL:
|
|
5355 p = "IDENT+NL";
|
|
5356 break;
|
|
5357 case SIDENT:
|
|
5358 p = "SIDENT";
|
|
5359 break;
|
|
5360 case SIDENT + ADD_NL:
|
|
5361 p = "SIDENT+NL";
|
|
5362 break;
|
|
5363 case KWORD:
|
|
5364 p = "KWORD";
|
|
5365 break;
|
|
5366 case KWORD + ADD_NL:
|
|
5367 p = "KWORD+NL";
|
|
5368 break;
|
|
5369 case SKWORD:
|
|
5370 p = "SKWORD";
|
|
5371 break;
|
|
5372 case SKWORD + ADD_NL:
|
|
5373 p = "SKWORD+NL";
|
|
5374 break;
|
|
5375 case FNAME:
|
|
5376 p = "FNAME";
|
|
5377 break;
|
|
5378 case FNAME + ADD_NL:
|
|
5379 p = "FNAME+NL";
|
|
5380 break;
|
|
5381 case SFNAME:
|
|
5382 p = "SFNAME";
|
|
5383 break;
|
|
5384 case SFNAME + ADD_NL:
|
|
5385 p = "SFNAME+NL";
|
|
5386 break;
|
|
5387 case PRINT:
|
|
5388 p = "PRINT";
|
|
5389 break;
|
|
5390 case PRINT + ADD_NL:
|
|
5391 p = "PRINT+NL";
|
|
5392 break;
|
|
5393 case SPRINT:
|
|
5394 p = "SPRINT";
|
|
5395 break;
|
|
5396 case SPRINT + ADD_NL:
|
|
5397 p = "SPRINT+NL";
|
|
5398 break;
|
|
5399 case WHITE:
|
|
5400 p = "WHITE";
|
|
5401 break;
|
|
5402 case WHITE + ADD_NL:
|
|
5403 p = "WHITE+NL";
|
|
5404 break;
|
|
5405 case NWHITE:
|
|
5406 p = "NWHITE";
|
|
5407 break;
|
|
5408 case NWHITE + ADD_NL:
|
|
5409 p = "NWHITE+NL";
|
|
5410 break;
|
|
5411 case DIGIT:
|
|
5412 p = "DIGIT";
|
|
5413 break;
|
|
5414 case DIGIT + ADD_NL:
|
|
5415 p = "DIGIT+NL";
|
|
5416 break;
|
|
5417 case NDIGIT:
|
|
5418 p = "NDIGIT";
|
|
5419 break;
|
|
5420 case NDIGIT + ADD_NL:
|
|
5421 p = "NDIGIT+NL";
|
|
5422 break;
|
|
5423 case HEX:
|
|
5424 p = "HEX";
|
|
5425 break;
|
|
5426 case HEX + ADD_NL:
|
|
5427 p = "HEX+NL";
|
|
5428 break;
|
|
5429 case NHEX:
|
|
5430 p = "NHEX";
|
|
5431 break;
|
|
5432 case NHEX + ADD_NL:
|
|
5433 p = "NHEX+NL";
|
|
5434 break;
|
|
5435 case OCTAL:
|
|
5436 p = "OCTAL";
|
|
5437 break;
|
|
5438 case OCTAL + ADD_NL:
|
|
5439 p = "OCTAL+NL";
|
|
5440 break;
|
|
5441 case NOCTAL:
|
|
5442 p = "NOCTAL";
|
|
5443 break;
|
|
5444 case NOCTAL + ADD_NL:
|
|
5445 p = "NOCTAL+NL";
|
|
5446 break;
|
|
5447 case WORD:
|
|
5448 p = "WORD";
|
|
5449 break;
|
|
5450 case WORD + ADD_NL:
|
|
5451 p = "WORD+NL";
|
|
5452 break;
|
|
5453 case NWORD:
|
|
5454 p = "NWORD";
|
|
5455 break;
|
|
5456 case NWORD + ADD_NL:
|
|
5457 p = "NWORD+NL";
|
|
5458 break;
|
|
5459 case HEAD:
|
|
5460 p = "HEAD";
|
|
5461 break;
|
|
5462 case HEAD + ADD_NL:
|
|
5463 p = "HEAD+NL";
|
|
5464 break;
|
|
5465 case NHEAD:
|
|
5466 p = "NHEAD";
|
|
5467 break;
|
|
5468 case NHEAD + ADD_NL:
|
|
5469 p = "NHEAD+NL";
|
|
5470 break;
|
|
5471 case ALPHA:
|
|
5472 p = "ALPHA";
|
|
5473 break;
|
|
5474 case ALPHA + ADD_NL:
|
|
5475 p = "ALPHA+NL";
|
|
5476 break;
|
|
5477 case NALPHA:
|
|
5478 p = "NALPHA";
|
|
5479 break;
|
|
5480 case NALPHA + ADD_NL:
|
|
5481 p = "NALPHA+NL";
|
|
5482 break;
|
|
5483 case LOWER:
|
|
5484 p = "LOWER";
|
|
5485 break;
|
|
5486 case LOWER + ADD_NL:
|
|
5487 p = "LOWER+NL";
|
|
5488 break;
|
|
5489 case NLOWER:
|
|
5490 p = "NLOWER";
|
|
5491 break;
|
|
5492 case NLOWER + ADD_NL:
|
|
5493 p = "NLOWER+NL";
|
|
5494 break;
|
|
5495 case UPPER:
|
|
5496 p = "UPPER";
|
|
5497 break;
|
|
5498 case UPPER + ADD_NL:
|
|
5499 p = "UPPER+NL";
|
|
5500 break;
|
|
5501 case NUPPER:
|
|
5502 p = "NUPPER";
|
|
5503 break;
|
|
5504 case NUPPER + ADD_NL:
|
|
5505 p = "NUPPER+NL";
|
|
5506 break;
|
|
5507 case BRANCH:
|
|
5508 p = "BRANCH";
|
|
5509 break;
|
|
5510 case EXACTLY:
|
|
5511 p = "EXACTLY";
|
|
5512 break;
|
|
5513 case NOTHING:
|
|
5514 p = "NOTHING";
|
|
5515 break;
|
|
5516 case BACK:
|
|
5517 p = "BACK";
|
|
5518 break;
|
|
5519 case END:
|
|
5520 p = "END";
|
|
5521 break;
|
|
5522 case MOPEN + 0:
|
|
5523 p = "MATCH START";
|
|
5524 break;
|
|
5525 case MOPEN + 1:
|
|
5526 case MOPEN + 2:
|
|
5527 case MOPEN + 3:
|
|
5528 case MOPEN + 4:
|
|
5529 case MOPEN + 5:
|
|
5530 case MOPEN + 6:
|
|
5531 case MOPEN + 7:
|
|
5532 case MOPEN + 8:
|
|
5533 case MOPEN + 9:
|
|
5534 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
|
|
5535 p = NULL;
|
|
5536 break;
|
|
5537 case MCLOSE + 0:
|
|
5538 p = "MATCH END";
|
|
5539 break;
|
|
5540 case MCLOSE + 1:
|
|
5541 case MCLOSE + 2:
|
|
5542 case MCLOSE + 3:
|
|
5543 case MCLOSE + 4:
|
|
5544 case MCLOSE + 5:
|
|
5545 case MCLOSE + 6:
|
|
5546 case MCLOSE + 7:
|
|
5547 case MCLOSE + 8:
|
|
5548 case MCLOSE + 9:
|
|
5549 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
|
|
5550 p = NULL;
|
|
5551 break;
|
|
5552 case BACKREF + 1:
|
|
5553 case BACKREF + 2:
|
|
5554 case BACKREF + 3:
|
|
5555 case BACKREF + 4:
|
|
5556 case BACKREF + 5:
|
|
5557 case BACKREF + 6:
|
|
5558 case BACKREF + 7:
|
|
5559 case BACKREF + 8:
|
|
5560 case BACKREF + 9:
|
|
5561 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
|
|
5562 p = NULL;
|
|
5563 break;
|
|
5564 case NOPEN:
|
|
5565 p = "NOPEN";
|
|
5566 break;
|
|
5567 case NCLOSE:
|
|
5568 p = "NCLOSE";
|
|
5569 break;
|
|
5570 #ifdef FEAT_SYN_HL
|
|
5571 case ZOPEN + 1:
|
|
5572 case ZOPEN + 2:
|
|
5573 case ZOPEN + 3:
|
|
5574 case ZOPEN + 4:
|
|
5575 case ZOPEN + 5:
|
|
5576 case ZOPEN + 6:
|
|
5577 case ZOPEN + 7:
|
|
5578 case ZOPEN + 8:
|
|
5579 case ZOPEN + 9:
|
|
5580 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
|
|
5581 p = NULL;
|
|
5582 break;
|
|
5583 case ZCLOSE + 1:
|
|
5584 case ZCLOSE + 2:
|
|
5585 case ZCLOSE + 3:
|
|
5586 case ZCLOSE + 4:
|
|
5587 case ZCLOSE + 5:
|
|
5588 case ZCLOSE + 6:
|
|
5589 case ZCLOSE + 7:
|
|
5590 case ZCLOSE + 8:
|
|
5591 case ZCLOSE + 9:
|
|
5592 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
|
|
5593 p = NULL;
|
|
5594 break;
|
|
5595 case ZREF + 1:
|
|
5596 case ZREF + 2:
|
|
5597 case ZREF + 3:
|
|
5598 case ZREF + 4:
|
|
5599 case ZREF + 5:
|
|
5600 case ZREF + 6:
|
|
5601 case ZREF + 7:
|
|
5602 case ZREF + 8:
|
|
5603 case ZREF + 9:
|
|
5604 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
|
|
5605 p = NULL;
|
|
5606 break;
|
|
5607 #endif
|
|
5608 case STAR:
|
|
5609 p = "STAR";
|
|
5610 break;
|
|
5611 case PLUS:
|
|
5612 p = "PLUS";
|
|
5613 break;
|
|
5614 case NOMATCH:
|
|
5615 p = "NOMATCH";
|
|
5616 break;
|
|
5617 case MATCH:
|
|
5618 p = "MATCH";
|
|
5619 break;
|
|
5620 case BEHIND:
|
|
5621 p = "BEHIND";
|
|
5622 break;
|
|
5623 case NOBEHIND:
|
|
5624 p = "NOBEHIND";
|
|
5625 break;
|
|
5626 case SUBPAT:
|
|
5627 p = "SUBPAT";
|
|
5628 break;
|
|
5629 case BRACE_LIMITS:
|
|
5630 p = "BRACE_LIMITS";
|
|
5631 break;
|
|
5632 case BRACE_SIMPLE:
|
|
5633 p = "BRACE_SIMPLE";
|
|
5634 break;
|
|
5635 case BRACE_COMPLEX + 0:
|
|
5636 case BRACE_COMPLEX + 1:
|
|
5637 case BRACE_COMPLEX + 2:
|
|
5638 case BRACE_COMPLEX + 3:
|
|
5639 case BRACE_COMPLEX + 4:
|
|
5640 case BRACE_COMPLEX + 5:
|
|
5641 case BRACE_COMPLEX + 6:
|
|
5642 case BRACE_COMPLEX + 7:
|
|
5643 case BRACE_COMPLEX + 8:
|
|
5644 case BRACE_COMPLEX + 9:
|
|
5645 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
|
|
5646 p = NULL;
|
|
5647 break;
|
|
5648 #ifdef FEAT_MBYTE
|
|
5649 case MULTIBYTECODE:
|
|
5650 p = "MULTIBYTECODE";
|
|
5651 break;
|
|
5652 #endif
|
|
5653 case NEWL:
|
|
5654 p = "NEWL";
|
|
5655 break;
|
|
5656 default:
|
|
5657 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
|
|
5658 p = NULL;
|
|
5659 break;
|
|
5660 }
|
|
5661 if (p != NULL)
|
|
5662 (void) strcat(buf, p);
|
|
5663 return buf;
|
|
5664 }
|
|
5665 #endif
|
|
5666
|
|
5667 #ifdef FEAT_MBYTE
|
|
5668 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3));
|
|
5669
|
|
5670 typedef struct
|
|
5671 {
|
|
5672 int a, b, c;
|
|
5673 } decomp_T;
|
|
5674
|
|
5675
|
|
5676 /* 0xfb20 - 0xfb4f */
|
|
5677 decomp_T decomp_table[0xfb4f-0xfb20+1] =
|
|
5678 {
|
|
5679 {0x5e2,0,0}, /* 0xfb20 alt ayin */
|
|
5680 {0x5d0,0,0}, /* 0xfb21 alt alef */
|
|
5681 {0x5d3,0,0}, /* 0xfb22 alt dalet */
|
|
5682 {0x5d4,0,0}, /* 0xfb23 alt he */
|
|
5683 {0x5db,0,0}, /* 0xfb24 alt kaf */
|
|
5684 {0x5dc,0,0}, /* 0xfb25 alt lamed */
|
|
5685 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */
|
|
5686 {0x5e8,0,0}, /* 0xfb27 alt resh */
|
|
5687 {0x5ea,0,0}, /* 0xfb28 alt tav */
|
|
5688 {'+', 0, 0}, /* 0xfb29 alt plus */
|
|
5689 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */
|
|
5690 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */
|
|
5691 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */
|
|
5692 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */
|
|
5693 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */
|
|
5694 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */
|
|
5695 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */
|
|
5696 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */
|
|
5697 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */
|
|
5698 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */
|
|
5699 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */
|
|
5700 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */
|
|
5701 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */
|
|
5702 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */
|
|
5703 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */
|
|
5704 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */
|
|
5705 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */
|
|
5706 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */
|
|
5707 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */
|
|
5708 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */
|
|
5709 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */
|
|
5710 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */
|
|
5711 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */
|
|
5712 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */
|
|
5713 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */
|
|
5714 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */
|
|
5715 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */
|
|
5716 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */
|
|
5717 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */
|
|
5718 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */
|
|
5719 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */
|
|
5720 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */
|
|
5721 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */
|
|
5722 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */
|
|
5723 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */
|
|
5724 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */
|
|
5725 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */
|
|
5726 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */
|
|
5727 };
|
|
5728
|
|
5729 static void
|
|
5730 mb_decompose(c, c1, c2, c3)
|
|
5731 int c, *c1, *c2, *c3;
|
|
5732 {
|
|
5733 decomp_T d;
|
|
5734
|
|
5735 if (c >= 0x4b20 && c <= 0xfb4f)
|
|
5736 {
|
|
5737 d = decomp_table[c - 0xfb20];
|
|
5738 *c1 = d.a;
|
|
5739 *c2 = d.b;
|
|
5740 *c3 = d.c;
|
|
5741 }
|
|
5742 else
|
|
5743 {
|
|
5744 *c1 = c;
|
|
5745 *c2 = *c3 = 0;
|
|
5746 }
|
|
5747 }
|
|
5748 #endif
|
|
5749
|
|
5750 /*
|
|
5751 * Compare two strings, ignore case if ireg_ic set.
|
|
5752 * Return 0 if strings match, non-zero otherwise.
|
|
5753 * Correct the length "*n" when composing characters are ignored.
|
|
5754 */
|
|
5755 static int
|
|
5756 cstrncmp(s1, s2, n)
|
|
5757 char_u *s1, *s2;
|
|
5758 int *n;
|
|
5759 {
|
|
5760 int result;
|
|
5761
|
|
5762 if (!ireg_ic)
|
|
5763 result = STRNCMP(s1, s2, *n);
|
|
5764 else
|
|
5765 result = MB_STRNICMP(s1, s2, *n);
|
|
5766
|
|
5767 #ifdef FEAT_MBYTE
|
|
5768 /* if it failed and it's utf8 and we want to combineignore: */
|
|
5769 if (result != 0 && enc_utf8 && ireg_icombine)
|
|
5770 {
|
|
5771 char_u *str1, *str2;
|
|
5772 int c1, c2, c11, c12;
|
|
5773 int ix;
|
|
5774 int junk;
|
|
5775
|
|
5776 /* we have to handle the strcmp ourselves, since it is necessary to
|
|
5777 * deal with the composing characters by ignoring them: */
|
|
5778 str1 = s1;
|
|
5779 str2 = s2;
|
|
5780 c1 = c2 = 0;
|
|
5781 for (ix = 0; ix < *n; )
|
|
5782 {
|
|
5783 c1 = mb_ptr2char_adv(&str1);
|
|
5784 c2 = mb_ptr2char_adv(&str2);
|
|
5785 ix += utf_char2len(c1);
|
|
5786
|
|
5787 /* decompose the character if necessary, into 'base' characters
|
|
5788 * because I don't care about Arabic, I will hard-code the Hebrew
|
|
5789 * which I *do* care about! So sue me... */
|
|
5790 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2)))
|
|
5791 {
|
|
5792 /* decomposition necessary? */
|
|
5793 mb_decompose(c1, &c11, &junk, &junk);
|
|
5794 mb_decompose(c2, &c12, &junk, &junk);
|
|
5795 c1 = c11;
|
|
5796 c2 = c12;
|
|
5797 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12)))
|
|
5798 break;
|
|
5799 }
|
|
5800 }
|
|
5801 result = c2 - c1;
|
|
5802 if (result == 0)
|
|
5803 *n = (int)(str2 - s2);
|
|
5804 }
|
|
5805 #endif
|
|
5806
|
|
5807 return result;
|
|
5808 }
|
|
5809
|
|
5810 /*
|
|
5811 * cstrchr: This function is used a lot for simple searches, keep it fast!
|
|
5812 */
|
|
5813 static char_u *
|
|
5814 cstrchr(s, c)
|
|
5815 char_u *s;
|
|
5816 int c;
|
|
5817 {
|
|
5818 char_u *p;
|
|
5819 int cc;
|
|
5820
|
|
5821 if (!ireg_ic
|
|
5822 #ifdef FEAT_MBYTE
|
|
5823 || (!enc_utf8 && mb_char2len(c) > 1)
|
|
5824 #endif
|
|
5825 )
|
|
5826 return vim_strchr(s, c);
|
|
5827
|
|
5828 /* tolower() and toupper() can be slow, comparing twice should be a lot
|
|
5829 * faster (esp. when using MS Visual C++!).
|
|
5830 * For UTF-8 need to use folded case. */
|
|
5831 #ifdef FEAT_MBYTE
|
|
5832 if (enc_utf8 && c > 0x80)
|
|
5833 cc = utf_fold(c);
|
|
5834 else
|
|
5835 #endif
|
|
5836 if (isupper(c))
|
|
5837 cc = TOLOWER_LOC(c);
|
|
5838 else if (islower(c))
|
|
5839 cc = TOUPPER_LOC(c);
|
|
5840 else
|
|
5841 return vim_strchr(s, c);
|
|
5842
|
|
5843 #ifdef FEAT_MBYTE
|
|
5844 if (has_mbyte)
|
|
5845 {
|
|
5846 for (p = s; *p != NUL; p += (*mb_ptr2len_check)(p))
|
|
5847 {
|
|
5848 if (enc_utf8 && c > 0x80)
|
|
5849 {
|
|
5850 if (utf_fold(utf_ptr2char(p)) == cc)
|
|
5851 return p;
|
|
5852 }
|
|
5853 else if (*p == c || *p == cc)
|
|
5854 return p;
|
|
5855 }
|
|
5856 }
|
|
5857 else
|
|
5858 #endif
|
|
5859 /* Faster version for when there are no multi-byte characters. */
|
|
5860 for (p = s; *p != NUL; ++p)
|
|
5861 if (*p == c || *p == cc)
|
|
5862 return p;
|
|
5863
|
|
5864 return NULL;
|
|
5865 }
|
|
5866
|
|
5867 /***************************************************************
|
|
5868 * regsub stuff *
|
|
5869 ***************************************************************/
|
|
5870
|
|
5871 /* This stuff below really confuses cc on an SGI -- webb */
|
|
5872 #ifdef __sgi
|
|
5873 # undef __ARGS
|
|
5874 # define __ARGS(x) ()
|
|
5875 #endif
|
|
5876
|
|
5877 /*
|
|
5878 * We should define ftpr as a pointer to a function returning a pointer to
|
|
5879 * a function returning a pointer to a function ...
|
|
5880 * This is impossible, so we declare a pointer to a function returning a
|
|
5881 * pointer to a function returning void. This should work for all compilers.
|
|
5882 */
|
|
5883 typedef void (*(*fptr) __ARGS((char_u *, int)))();
|
|
5884
|
|
5885 static fptr do_upper __ARGS((char_u *, int));
|
|
5886 static fptr do_Upper __ARGS((char_u *, int));
|
|
5887 static fptr do_lower __ARGS((char_u *, int));
|
|
5888 static fptr do_Lower __ARGS((char_u *, int));
|
|
5889
|
|
5890 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash));
|
|
5891
|
|
5892 static fptr
|
|
5893 do_upper(d, c)
|
|
5894 char_u *d;
|
|
5895 int c;
|
|
5896 {
|
|
5897 *d = TOUPPER_LOC(c);
|
|
5898
|
|
5899 return (fptr)NULL;
|
|
5900 }
|
|
5901
|
|
5902 static fptr
|
|
5903 do_Upper(d, c)
|
|
5904 char_u *d;
|
|
5905 int c;
|
|
5906 {
|
|
5907 *d = TOUPPER_LOC(c);
|
|
5908
|
|
5909 return (fptr)do_Upper;
|
|
5910 }
|
|
5911
|
|
5912 static fptr
|
|
5913 do_lower(d, c)
|
|
5914 char_u *d;
|
|
5915 int c;
|
|
5916 {
|
|
5917 *d = TOLOWER_LOC(c);
|
|
5918
|
|
5919 return (fptr)NULL;
|
|
5920 }
|
|
5921
|
|
5922 static fptr
|
|
5923 do_Lower(d, c)
|
|
5924 char_u *d;
|
|
5925 int c;
|
|
5926 {
|
|
5927 *d = TOLOWER_LOC(c);
|
|
5928
|
|
5929 return (fptr)do_Lower;
|
|
5930 }
|
|
5931
|
|
5932 /*
|
|
5933 * regtilde(): Replace tildes in the pattern by the old pattern.
|
|
5934 *
|
|
5935 * Short explanation of the tilde: It stands for the previous replacement
|
|
5936 * pattern. If that previous pattern also contains a ~ we should go back a
|
|
5937 * step further... But we insert the previous pattern into the current one
|
|
5938 * and remember that.
|
|
5939 * This still does not handle the case where "magic" changes. TODO?
|
|
5940 *
|
|
5941 * The tildes are parsed once before the first call to vim_regsub().
|
|
5942 */
|
|
5943 char_u *
|
|
5944 regtilde(source, magic)
|
|
5945 char_u *source;
|
|
5946 int magic;
|
|
5947 {
|
|
5948 char_u *newsub = source;
|
|
5949 char_u *tmpsub;
|
|
5950 char_u *p;
|
|
5951 int len;
|
|
5952 int prevlen;
|
|
5953
|
|
5954 for (p = newsub; *p; ++p)
|
|
5955 {
|
|
5956 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic))
|
|
5957 {
|
|
5958 if (reg_prev_sub != NULL)
|
|
5959 {
|
|
5960 /* length = len(newsub) - 1 + len(prev_sub) + 1 */
|
|
5961 prevlen = (int)STRLEN(reg_prev_sub);
|
|
5962 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen));
|
|
5963 if (tmpsub != NULL)
|
|
5964 {
|
|
5965 /* copy prefix */
|
|
5966 len = (int)(p - newsub); /* not including ~ */
|
|
5967 mch_memmove(tmpsub, newsub, (size_t)len);
|
|
5968 /* interpretate tilde */
|
|
5969 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen);
|
|
5970 /* copy postfix */
|
|
5971 if (!magic)
|
|
5972 ++p; /* back off \ */
|
|
5973 STRCPY(tmpsub + len + prevlen, p + 1);
|
|
5974
|
|
5975 if (newsub != source) /* already allocated newsub */
|
|
5976 vim_free(newsub);
|
|
5977 newsub = tmpsub;
|
|
5978 p = newsub + len + prevlen;
|
|
5979 }
|
|
5980 }
|
|
5981 else if (magic)
|
|
5982 STRCPY(p, p + 1); /* remove '~' */
|
|
5983 else
|
|
5984 STRCPY(p, p + 2); /* remove '\~' */
|
|
5985 --p;
|
|
5986 }
|
|
5987 else
|
|
5988 {
|
|
5989 if (*p == '\\' && p[1]) /* skip escaped characters */
|
|
5990 ++p;
|
|
5991 #ifdef FEAT_MBYTE
|
|
5992 if (has_mbyte)
|
|
5993 p += (*mb_ptr2len_check)(p) - 1;
|
|
5994 #endif
|
|
5995 }
|
|
5996 }
|
|
5997
|
|
5998 vim_free(reg_prev_sub);
|
|
5999 if (newsub != source) /* newsub was allocated, just keep it */
|
|
6000 reg_prev_sub = newsub;
|
|
6001 else /* no ~ found, need to save newsub */
|
|
6002 reg_prev_sub = vim_strsave(newsub);
|
|
6003 return newsub;
|
|
6004 }
|
|
6005
|
|
6006 #ifdef FEAT_EVAL
|
|
6007 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */
|
|
6008
|
|
6009 /* These pointers are used instead of reg_match and reg_mmatch for
|
|
6010 * reg_submatch(). Needed for when the substitution string is an expression
|
|
6011 * that contains a call to substitute() and submatch(). */
|
|
6012 static regmatch_T *submatch_match;
|
|
6013 static regmmatch_T *submatch_mmatch;
|
|
6014 #endif
|
|
6015
|
|
6016 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO)
|
|
6017 /*
|
|
6018 * vim_regsub() - perform substitutions after a vim_regexec() or
|
|
6019 * vim_regexec_multi() match.
|
|
6020 *
|
|
6021 * If "copy" is TRUE really copy into "dest".
|
|
6022 * If "copy" is FALSE nothing is copied, this is just to find out the length
|
|
6023 * of the result.
|
|
6024 *
|
|
6025 * If "backslash" is TRUE, a backslash will be removed later, need to double
|
|
6026 * them to keep them, and insert a backslash before a CR to avoid it being
|
|
6027 * replaced with a line break later.
|
|
6028 *
|
|
6029 * Note: The matched text must not change between the call of
|
|
6030 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back
|
|
6031 * references invalid!
|
|
6032 *
|
|
6033 * Returns the size of the replacement, including terminating NUL.
|
|
6034 */
|
|
6035 int
|
|
6036 vim_regsub(rmp, source, dest, copy, magic, backslash)
|
|
6037 regmatch_T *rmp;
|
|
6038 char_u *source;
|
|
6039 char_u *dest;
|
|
6040 int copy;
|
|
6041 int magic;
|
|
6042 int backslash;
|
|
6043 {
|
|
6044 reg_match = rmp;
|
|
6045 reg_mmatch = NULL;
|
|
6046 reg_maxline = 0;
|
|
6047 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
6048 }
|
|
6049 #endif
|
|
6050
|
|
6051 int
|
|
6052 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash)
|
|
6053 regmmatch_T *rmp;
|
|
6054 linenr_T lnum;
|
|
6055 char_u *source;
|
|
6056 char_u *dest;
|
|
6057 int copy;
|
|
6058 int magic;
|
|
6059 int backslash;
|
|
6060 {
|
|
6061 reg_match = NULL;
|
|
6062 reg_mmatch = rmp;
|
|
6063 reg_buf = curbuf; /* always works on the current buffer! */
|
|
6064 reg_firstlnum = lnum;
|
|
6065 reg_maxline = curbuf->b_ml.ml_line_count - lnum;
|
|
6066 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
6067 }
|
|
6068
|
|
6069 static int
|
|
6070 vim_regsub_both(source, dest, copy, magic, backslash)
|
|
6071 char_u *source;
|
|
6072 char_u *dest;
|
|
6073 int copy;
|
|
6074 int magic;
|
|
6075 int backslash;
|
|
6076 {
|
|
6077 char_u *src;
|
|
6078 char_u *dst;
|
|
6079 char_u *s;
|
|
6080 int c;
|
|
6081 int no = -1;
|
|
6082 fptr func = (fptr)NULL;
|
|
6083 linenr_T clnum = 0; /* init for GCC */
|
|
6084 int len = 0; /* init for GCC */
|
|
6085 #ifdef FEAT_EVAL
|
|
6086 static char_u *eval_result = NULL;
|
|
6087 #endif
|
|
6088 #ifdef FEAT_MBYTE
|
|
6089 int l;
|
|
6090 #endif
|
|
6091
|
|
6092
|
|
6093 /* Be paranoid... */
|
|
6094 if (source == NULL || dest == NULL)
|
|
6095 {
|
|
6096 EMSG(_(e_null));
|
|
6097 return 0;
|
|
6098 }
|
|
6099 if (prog_magic_wrong())
|
|
6100 return 0;
|
|
6101 src = source;
|
|
6102 dst = dest;
|
|
6103
|
|
6104 /*
|
|
6105 * When the substitute part starts with "\=" evaluate it as an expression.
|
|
6106 */
|
|
6107 if (source[0] == '\\' && source[1] == '='
|
|
6108 #ifdef FEAT_EVAL
|
|
6109 && !can_f_submatch /* can't do this recursively */
|
|
6110 #endif
|
|
6111 )
|
|
6112 {
|
|
6113 #ifdef FEAT_EVAL
|
|
6114 /* To make sure that the length doesn't change between checking the
|
|
6115 * length and copying the string, and to speed up things, the
|
|
6116 * resulting string is saved from the call with "copy" == FALSE to the
|
|
6117 * call with "copy" == TRUE. */
|
|
6118 if (copy)
|
|
6119 {
|
|
6120 if (eval_result != NULL)
|
|
6121 {
|
|
6122 STRCPY(dest, eval_result);
|
|
6123 dst += STRLEN(eval_result);
|
|
6124 vim_free(eval_result);
|
|
6125 eval_result = NULL;
|
|
6126 }
|
|
6127 }
|
|
6128 else
|
|
6129 {
|
|
6130 linenr_T save_reg_maxline;
|
|
6131 win_T *save_reg_win;
|
|
6132 int save_ireg_ic;
|
|
6133
|
|
6134 vim_free(eval_result);
|
|
6135
|
|
6136 /* The expression may contain substitute(), which calls us
|
|
6137 * recursively. Make sure submatch() gets the text from the first
|
|
6138 * level. Don't need to save "reg_buf", because
|
|
6139 * vim_regexec_multi() can't be called recursively. */
|
|
6140 submatch_match = reg_match;
|
|
6141 submatch_mmatch = reg_mmatch;
|
|
6142 save_reg_maxline = reg_maxline;
|
|
6143 save_reg_win = reg_win;
|
|
6144 save_ireg_ic = ireg_ic;
|
|
6145 can_f_submatch = TRUE;
|
|
6146
|
|
6147 eval_result = eval_to_string(source + 2, NULL);
|
|
6148 if (eval_result != NULL)
|
|
6149 {
|
39
|
6150 for (s = eval_result; *s != NUL; mb_ptr_adv(s))
|
7
|
6151 {
|
|
6152 /* Change NL to CR, so that it becomes a line break.
|
|
6153 * Skip over a backslashed character. */
|
|
6154 if (*s == NL)
|
|
6155 *s = CAR;
|
|
6156 else if (*s == '\\' && s[1] != NUL)
|
|
6157 ++s;
|
|
6158 }
|
|
6159
|
|
6160 dst += STRLEN(eval_result);
|
|
6161 }
|
|
6162
|
|
6163 reg_match = submatch_match;
|
|
6164 reg_mmatch = submatch_mmatch;
|
|
6165 reg_maxline = save_reg_maxline;
|
|
6166 reg_win = save_reg_win;
|
|
6167 ireg_ic = save_ireg_ic;
|
|
6168 can_f_submatch = FALSE;
|
|
6169 }
|
|
6170 #endif
|
|
6171 }
|
|
6172 else
|
|
6173 while ((c = *src++) != NUL)
|
|
6174 {
|
|
6175 if (c == '&' && magic)
|
|
6176 no = 0;
|
|
6177 else if (c == '\\' && *src != NUL)
|
|
6178 {
|
|
6179 if (*src == '&' && !magic)
|
|
6180 {
|
|
6181 ++src;
|
|
6182 no = 0;
|
|
6183 }
|
|
6184 else if ('0' <= *src && *src <= '9')
|
|
6185 {
|
|
6186 no = *src++ - '0';
|
|
6187 }
|
|
6188 else if (vim_strchr((char_u *)"uUlLeE", *src))
|
|
6189 {
|
|
6190 switch (*src++)
|
|
6191 {
|
|
6192 case 'u': func = (fptr)do_upper;
|
|
6193 continue;
|
|
6194 case 'U': func = (fptr)do_Upper;
|
|
6195 continue;
|
|
6196 case 'l': func = (fptr)do_lower;
|
|
6197 continue;
|
|
6198 case 'L': func = (fptr)do_Lower;
|
|
6199 continue;
|
|
6200 case 'e':
|
|
6201 case 'E': func = (fptr)NULL;
|
|
6202 continue;
|
|
6203 }
|
|
6204 }
|
|
6205 }
|
|
6206 if (no < 0) /* Ordinary character. */
|
|
6207 {
|
|
6208 if (c == '\\' && *src != NUL)
|
|
6209 {
|
|
6210 /* Check for abbreviations -- webb */
|
|
6211 switch (*src)
|
|
6212 {
|
|
6213 case 'r': c = CAR; ++src; break;
|
|
6214 case 'n': c = NL; ++src; break;
|
|
6215 case 't': c = TAB; ++src; break;
|
|
6216 /* Oh no! \e already has meaning in subst pat :-( */
|
|
6217 /* case 'e': c = ESC; ++src; break; */
|
|
6218 case 'b': c = Ctrl_H; ++src; break;
|
|
6219
|
|
6220 /* If "backslash" is TRUE the backslash will be removed
|
|
6221 * later. Used to insert a literal CR. */
|
|
6222 default: if (backslash)
|
|
6223 {
|
|
6224 if (copy)
|
|
6225 *dst = '\\';
|
|
6226 ++dst;
|
|
6227 }
|
|
6228 c = *src++;
|
|
6229 }
|
|
6230 }
|
|
6231
|
|
6232 /* Write to buffer, if copy is set. */
|
|
6233 #ifdef FEAT_MBYTE
|
|
6234 if (has_mbyte && (l = (*mb_ptr2len_check)(src - 1)) > 1)
|
|
6235 {
|
|
6236 /* TODO: should use "func" here. */
|
|
6237 if (copy)
|
|
6238 mch_memmove(dst, src - 1, l);
|
|
6239 dst += l - 1;
|
|
6240 src += l - 1;
|
|
6241 }
|
|
6242 else
|
|
6243 {
|
|
6244 #endif
|
|
6245 if (copy)
|
|
6246 {
|
|
6247 if (func == (fptr)NULL) /* just copy */
|
|
6248 *dst = c;
|
|
6249 else /* change case */
|
|
6250 func = (fptr)(func(dst, c));
|
|
6251 /* Turbo C complains without the typecast */
|
|
6252 }
|
|
6253 #ifdef FEAT_MBYTE
|
|
6254 }
|
|
6255 #endif
|
|
6256 dst++;
|
|
6257 }
|
|
6258 else
|
|
6259 {
|
|
6260 if (REG_MULTI)
|
|
6261 {
|
|
6262 clnum = reg_mmatch->startpos[no].lnum;
|
|
6263 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0)
|
|
6264 s = NULL;
|
|
6265 else
|
|
6266 {
|
|
6267 s = reg_getline(clnum) + reg_mmatch->startpos[no].col;
|
|
6268 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6269 len = reg_mmatch->endpos[no].col
|
|
6270 - reg_mmatch->startpos[no].col;
|
|
6271 else
|
|
6272 len = (int)STRLEN(s);
|
|
6273 }
|
|
6274 }
|
|
6275 else
|
|
6276 {
|
|
6277 s = reg_match->startp[no];
|
|
6278 if (reg_match->endp[no] == NULL)
|
|
6279 s = NULL;
|
|
6280 else
|
|
6281 len = (int)(reg_match->endp[no] - s);
|
|
6282 }
|
|
6283 if (s != NULL)
|
|
6284 {
|
|
6285 for (;;)
|
|
6286 {
|
|
6287 if (len == 0)
|
|
6288 {
|
|
6289 if (REG_MULTI)
|
|
6290 {
|
|
6291 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6292 break;
|
|
6293 if (copy)
|
|
6294 *dst = CAR;
|
|
6295 ++dst;
|
|
6296 s = reg_getline(++clnum);
|
|
6297 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6298 len = reg_mmatch->endpos[no].col;
|
|
6299 else
|
|
6300 len = (int)STRLEN(s);
|
|
6301 }
|
|
6302 else
|
|
6303 break;
|
|
6304 }
|
|
6305 else if (*s == NUL) /* we hit NUL. */
|
|
6306 {
|
|
6307 if (copy)
|
|
6308 EMSG(_(e_re_damg));
|
|
6309 goto exit;
|
|
6310 }
|
|
6311 else
|
|
6312 {
|
|
6313 if (backslash && (*s == CAR || *s == '\\'))
|
|
6314 {
|
|
6315 /*
|
|
6316 * Insert a backslash in front of a CR, otherwise
|
|
6317 * it will be replaced by a line break.
|
|
6318 * Number of backslashes will be halved later,
|
|
6319 * double them here.
|
|
6320 */
|
|
6321 if (copy)
|
|
6322 {
|
|
6323 dst[0] = '\\';
|
|
6324 dst[1] = *s;
|
|
6325 }
|
|
6326 dst += 2;
|
|
6327 }
|
|
6328 #ifdef FEAT_MBYTE
|
|
6329 else if (has_mbyte && (l = (*mb_ptr2len_check)(s)) > 1)
|
|
6330 {
|
|
6331 /* TODO: should use "func" here. */
|
|
6332 if (copy)
|
|
6333 mch_memmove(dst, s, l);
|
|
6334 dst += l;
|
|
6335 s += l - 1;
|
|
6336 len -= l - 1;
|
|
6337 }
|
|
6338 #endif
|
|
6339 else
|
|
6340 {
|
|
6341 if (copy)
|
|
6342 {
|
|
6343 if (func == (fptr)NULL) /* just copy */
|
|
6344 *dst = *s;
|
|
6345 else /* change case */
|
|
6346 func = (fptr)(func(dst, *s));
|
|
6347 /* Turbo C complains without the typecast */
|
|
6348 }
|
|
6349 ++dst;
|
|
6350 }
|
|
6351 ++s;
|
|
6352 --len;
|
|
6353 }
|
|
6354 }
|
|
6355 }
|
|
6356 no = -1;
|
|
6357 }
|
|
6358 }
|
|
6359 if (copy)
|
|
6360 *dst = NUL;
|
|
6361
|
|
6362 exit:
|
|
6363 return (int)((dst - dest) + 1);
|
|
6364 }
|
|
6365
|
|
6366 #ifdef FEAT_EVAL
|
|
6367 /*
|
|
6368 * Used for the submatch() function: get the string from tne n'th submatch in
|
|
6369 * allocated memory.
|
|
6370 * Returns NULL when not in a ":s" command and for a non-existing submatch.
|
|
6371 */
|
|
6372 char_u *
|
|
6373 reg_submatch(no)
|
|
6374 int no;
|
|
6375 {
|
|
6376 char_u *retval = NULL;
|
|
6377 char_u *s;
|
|
6378 int len;
|
|
6379 int round;
|
|
6380 linenr_T lnum;
|
|
6381
|
|
6382 if (!can_f_submatch)
|
|
6383 return NULL;
|
|
6384
|
|
6385 if (submatch_match == NULL)
|
|
6386 {
|
|
6387 /*
|
|
6388 * First round: compute the length and allocate memory.
|
|
6389 * Second round: copy the text.
|
|
6390 */
|
|
6391 for (round = 1; round <= 2; ++round)
|
|
6392 {
|
|
6393 lnum = submatch_mmatch->startpos[no].lnum;
|
|
6394 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0)
|
|
6395 return NULL;
|
|
6396
|
|
6397 s = reg_getline(lnum) + submatch_mmatch->startpos[no].col;
|
|
6398 if (s == NULL) /* anti-crash check, cannot happen? */
|
|
6399 break;
|
|
6400 if (submatch_mmatch->endpos[no].lnum == lnum)
|
|
6401 {
|
|
6402 /* Within one line: take form start to end col. */
|
|
6403 len = submatch_mmatch->endpos[no].col
|
|
6404 - submatch_mmatch->startpos[no].col;
|
|
6405 if (round == 2)
|
|
6406 {
|
|
6407 STRNCPY(retval, s, len);
|
|
6408 retval[len] = NUL;
|
|
6409 }
|
|
6410 ++len;
|
|
6411 }
|
|
6412 else
|
|
6413 {
|
|
6414 /* Multiple lines: take start line from start col, middle
|
|
6415 * lines completely and end line up to end col. */
|
|
6416 len = (int)STRLEN(s);
|
|
6417 if (round == 2)
|
|
6418 {
|
|
6419 STRCPY(retval, s);
|
|
6420 retval[len] = '\n';
|
|
6421 }
|
|
6422 ++len;
|
|
6423 ++lnum;
|
|
6424 while (lnum < submatch_mmatch->endpos[no].lnum)
|
|
6425 {
|
|
6426 s = reg_getline(lnum++);
|
|
6427 if (round == 2)
|
|
6428 STRCPY(retval + len, s);
|
|
6429 len += (int)STRLEN(s);
|
|
6430 if (round == 2)
|
|
6431 retval[len] = '\n';
|
|
6432 ++len;
|
|
6433 }
|
|
6434 if (round == 2)
|
|
6435 STRNCPY(retval + len, reg_getline(lnum),
|
|
6436 submatch_mmatch->endpos[no].col);
|
|
6437 len += submatch_mmatch->endpos[no].col;
|
|
6438 if (round == 2)
|
|
6439 retval[len] = NUL;
|
|
6440 ++len;
|
|
6441 }
|
|
6442
|
|
6443 if (round == 1)
|
|
6444 {
|
|
6445 retval = lalloc((long_u)len, TRUE);
|
|
6446 if (s == NULL)
|
|
6447 return NULL;
|
|
6448 }
|
|
6449 }
|
|
6450 }
|
|
6451 else
|
|
6452 {
|
|
6453 if (submatch_match->endp[no] == NULL)
|
|
6454 retval = NULL;
|
|
6455 else
|
|
6456 {
|
|
6457 s = submatch_match->startp[no];
|
|
6458 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s));
|
|
6459 }
|
|
6460 }
|
|
6461
|
|
6462 return retval;
|
|
6463 }
|
|
6464 #endif
|