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 |
|
179
|
103 * <aa>\+ BRANCH <aa> --> BRANCH --> BACKP BRANCH --> NOTHING --> END
|
|
104 * | | ^ ^
|
|
105 * | +-----------+ |
|
|
106 * +--------------------------------------------------+
|
167
|
107 *
|
|
108 *
|
7
|
109 * +-------------------------+
|
|
110 * V |
|
|
111 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END
|
|
112 * | | ^
|
|
113 * | +----------------+
|
|
114 * +-----------------------------------------------+
|
|
115 *
|
|
116 *
|
|
117 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END
|
|
118 * | | ^ ^
|
|
119 * | +----------------+ |
|
|
120 * +--------------------------------+
|
|
121 *
|
|
122 * +---------+
|
|
123 * | V
|
|
124 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END
|
|
125 * | | | | ^ ^
|
|
126 * | | | +-----+ |
|
|
127 * | | +----------------+ |
|
|
128 * | +---------------------------+ |
|
|
129 * +------------------------------------------------------+
|
|
130 *
|
|
131 * They all start with a BRANCH for "\|" alternaties, even when there is only
|
|
132 * one alternative.
|
|
133 */
|
|
134
|
|
135 /*
|
|
136 * The opcodes are:
|
|
137 */
|
|
138
|
|
139 /* definition number opnd? meaning */
|
|
140 #define END 0 /* End of program or NOMATCH operand. */
|
|
141 #define BOL 1 /* Match "" at beginning of line. */
|
|
142 #define EOL 2 /* Match "" at end of line. */
|
|
143 #define BRANCH 3 /* node Match this alternative, or the
|
|
144 * next... */
|
|
145 #define BACK 4 /* Match "", "next" ptr points backward. */
|
|
146 #define EXACTLY 5 /* str Match this string. */
|
|
147 #define NOTHING 6 /* Match empty string. */
|
|
148 #define STAR 7 /* node Match this (simple) thing 0 or more
|
|
149 * times. */
|
|
150 #define PLUS 8 /* node Match this (simple) thing 1 or more
|
|
151 * times. */
|
|
152 #define MATCH 9 /* node match the operand zero-width */
|
|
153 #define NOMATCH 10 /* node check for no match with operand */
|
|
154 #define BEHIND 11 /* node look behind for a match with operand */
|
|
155 #define NOBEHIND 12 /* node look behind for no match with operand */
|
|
156 #define SUBPAT 13 /* node match the operand here */
|
|
157 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and
|
|
158 * n times (\{m,n\}). */
|
|
159 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */
|
|
160 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */
|
|
161 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE
|
|
162 * and BRACE_COMPLEX. */
|
|
163 #define NEWL 18 /* Match line-break */
|
|
164 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */
|
|
165
|
|
166
|
|
167 /* character classes: 20-48 normal, 50-78 include a line-break */
|
|
168 #define ADD_NL 30
|
|
169 #define FIRST_NL ANY + ADD_NL
|
|
170 #define ANY 20 /* Match any one character. */
|
|
171 #define ANYOF 21 /* str Match any character in this string. */
|
|
172 #define ANYBUT 22 /* str Match any character not in this
|
|
173 * string. */
|
|
174 #define IDENT 23 /* Match identifier char */
|
|
175 #define SIDENT 24 /* Match identifier char but no digit */
|
|
176 #define KWORD 25 /* Match keyword char */
|
|
177 #define SKWORD 26 /* Match word char but no digit */
|
|
178 #define FNAME 27 /* Match file name char */
|
|
179 #define SFNAME 28 /* Match file name char but no digit */
|
|
180 #define PRINT 29 /* Match printable char */
|
|
181 #define SPRINT 30 /* Match printable char but no digit */
|
|
182 #define WHITE 31 /* Match whitespace char */
|
|
183 #define NWHITE 32 /* Match non-whitespace char */
|
|
184 #define DIGIT 33 /* Match digit char */
|
|
185 #define NDIGIT 34 /* Match non-digit char */
|
|
186 #define HEX 35 /* Match hex char */
|
|
187 #define NHEX 36 /* Match non-hex char */
|
|
188 #define OCTAL 37 /* Match octal char */
|
|
189 #define NOCTAL 38 /* Match non-octal char */
|
|
190 #define WORD 39 /* Match word char */
|
|
191 #define NWORD 40 /* Match non-word char */
|
|
192 #define HEAD 41 /* Match head char */
|
|
193 #define NHEAD 42 /* Match non-head char */
|
|
194 #define ALPHA 43 /* Match alpha char */
|
|
195 #define NALPHA 44 /* Match non-alpha char */
|
|
196 #define LOWER 45 /* Match lowercase char */
|
|
197 #define NLOWER 46 /* Match non-lowercase char */
|
|
198 #define UPPER 47 /* Match uppercase char */
|
|
199 #define NUPPER 48 /* Match non-uppercase char */
|
|
200 #define LAST_NL NUPPER + ADD_NL
|
|
201 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL)
|
|
202
|
|
203 #define MOPEN 80 /* -89 Mark this point in input as start of
|
|
204 * \( subexpr. MOPEN + 0 marks start of
|
|
205 * match. */
|
|
206 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks
|
|
207 * end of match. */
|
|
208 #define BACKREF 100 /* -109 node Match same string again \1-\9 */
|
|
209
|
|
210 #ifdef FEAT_SYN_HL
|
|
211 # define ZOPEN 110 /* -119 Mark this point in input as start of
|
|
212 * \z( subexpr. */
|
|
213 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */
|
|
214 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */
|
|
215 #endif
|
|
216
|
|
217 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */
|
|
218
|
|
219 #define NOPEN 150 /* Mark this point in input as start of
|
|
220 \%( subexpr. */
|
|
221 #define NCLOSE 151 /* Analogous to NOPEN. */
|
|
222
|
|
223 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */
|
|
224 #define RE_BOF 201 /* Match "" at beginning of file. */
|
|
225 #define RE_EOF 202 /* Match "" at end of file. */
|
|
226 #define CURSOR 203 /* Match location of cursor. */
|
|
227
|
|
228 #define RE_LNUM 204 /* nr cmp Match line number */
|
|
229 #define RE_COL 205 /* nr cmp Match column number */
|
|
230 #define RE_VCOL 206 /* nr cmp Match virtual column number */
|
|
231
|
179
|
232 #define BACKP 207 /* Like BACK but for \+ */
|
|
233
|
7
|
234 /*
|
|
235 * Magic characters have a special meaning, they don't match literally.
|
|
236 * Magic characters are negative. This separates them from literal characters
|
|
237 * (possibly multi-byte). Only ASCII characters can be Magic.
|
|
238 */
|
|
239 #define Magic(x) ((int)(x) - 256)
|
|
240 #define un_Magic(x) ((x) + 256)
|
|
241 #define is_Magic(x) ((x) < 0)
|
|
242
|
|
243 static int no_Magic __ARGS((int x));
|
|
244 static int toggle_Magic __ARGS((int x));
|
|
245
|
|
246 static int
|
|
247 no_Magic(x)
|
|
248 int x;
|
|
249 {
|
|
250 if (is_Magic(x))
|
|
251 return un_Magic(x);
|
|
252 return x;
|
|
253 }
|
|
254
|
|
255 static int
|
|
256 toggle_Magic(x)
|
|
257 int x;
|
|
258 {
|
|
259 if (is_Magic(x))
|
|
260 return un_Magic(x);
|
|
261 return Magic(x);
|
|
262 }
|
|
263
|
|
264 /*
|
|
265 * The first byte of the regexp internal "program" is actually this magic
|
|
266 * number; the start node begins in the second byte. It's used to catch the
|
|
267 * most severe mutilation of the program by the caller.
|
|
268 */
|
|
269
|
|
270 #define REGMAGIC 0234
|
|
271
|
|
272 /*
|
|
273 * Opcode notes:
|
|
274 *
|
|
275 * BRANCH The set of branches constituting a single choice are hooked
|
|
276 * together with their "next" pointers, since precedence prevents
|
|
277 * anything being concatenated to any individual branch. The
|
|
278 * "next" pointer of the last BRANCH in a choice points to the
|
|
279 * thing following the whole choice. This is also where the
|
|
280 * final "next" pointer of each individual branch points; each
|
|
281 * branch starts with the operand node of a BRANCH node.
|
|
282 *
|
|
283 * BACK Normal "next" pointers all implicitly point forward; BACK
|
|
284 * exists to make loop structures possible.
|
|
285 *
|
179
|
286 * BACKP Like BACK, but used for \+. Doesn't check for an empty match.
|
|
287 *
|
7
|
288 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular
|
|
289 * BRANCH structures using BACK. Simple cases (one character
|
|
290 * per match) are implemented with STAR and PLUS for speed
|
|
291 * and to minimize recursive plunges.
|
|
292 *
|
|
293 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX
|
|
294 * node, and defines the min and max limits to be used for that
|
|
295 * node.
|
|
296 *
|
|
297 * MOPEN,MCLOSE ...are numbered at compile time.
|
|
298 * ZOPEN,ZCLOSE ...ditto
|
|
299 */
|
|
300
|
|
301 /*
|
|
302 * A node is one char of opcode followed by two chars of "next" pointer.
|
|
303 * "Next" pointers are stored as two 8-bit bytes, high order first. The
|
|
304 * value is a positive offset from the opcode of the node containing it.
|
|
305 * An operand, if any, simply follows the node. (Note that much of the
|
|
306 * code generation knows about this implicit relationship.)
|
|
307 *
|
|
308 * Using two bytes for the "next" pointer is vast overkill for most things,
|
|
309 * but allows patterns to get big without disasters.
|
|
310 */
|
|
311 #define OP(p) ((int)*(p))
|
|
312 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377))
|
|
313 #define OPERAND(p) ((p) + 3)
|
|
314 /* Obtain an operand that was stored as four bytes, MSB first. */
|
|
315 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \
|
|
316 + ((long)(p)[5] << 8) + (long)(p)[6])
|
|
317 /* Obtain a second operand stored as four bytes. */
|
|
318 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4)
|
|
319 /* Obtain a second single-byte operand stored after a four bytes operand. */
|
|
320 #define OPERAND_CMP(p) (p)[7]
|
|
321
|
|
322 /*
|
|
323 * Utility definitions.
|
|
324 */
|
|
325 #define UCHARAT(p) ((int)*(char_u *)(p))
|
|
326
|
|
327 /* Used for an error (down from) vim_regcomp(): give the error message, set
|
|
328 * rc_did_emsg and return NULL */
|
|
329 #define EMSG_RET_NULL(m) { EMSG(m); rc_did_emsg = TRUE; return NULL; }
|
|
330 #define EMSG_M_RET_NULL(m, c) { EMSG2(m, c ? "" : "\\"); rc_did_emsg = TRUE; return NULL; }
|
|
331 #define EMSG_RET_FAIL(m) { EMSG(m); rc_did_emsg = TRUE; return FAIL; }
|
|
332 #define EMSG_ONE_RET_NULL EMSG_M_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL)
|
|
333
|
|
334 #define MAX_LIMIT (32767L << 16L)
|
|
335
|
|
336 static int re_multi_type __ARGS((int));
|
|
337 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n));
|
|
338 static char_u *cstrchr __ARGS((char_u *, int));
|
|
339
|
|
340 #ifdef DEBUG
|
|
341 static void regdump __ARGS((char_u *, regprog_T *));
|
|
342 static char_u *regprop __ARGS((char_u *));
|
|
343 #endif
|
|
344
|
|
345 #define NOT_MULTI 0
|
|
346 #define MULTI_ONE 1
|
|
347 #define MULTI_MULT 2
|
|
348 /*
|
|
349 * Return NOT_MULTI if c is not a "multi" operator.
|
|
350 * Return MULTI_ONE if c is a single "multi" operator.
|
|
351 * Return MULTI_MULT if c is a multi "multi" operator.
|
|
352 */
|
|
353 static int
|
|
354 re_multi_type(c)
|
|
355 int c;
|
|
356 {
|
|
357 if (c == Magic('@') || c == Magic('=') || c == Magic('?'))
|
|
358 return MULTI_ONE;
|
|
359 if (c == Magic('*') || c == Magic('+') || c == Magic('{'))
|
|
360 return MULTI_MULT;
|
|
361 return NOT_MULTI;
|
|
362 }
|
|
363
|
|
364 /*
|
|
365 * Flags to be passed up and down.
|
|
366 */
|
|
367 #define HASWIDTH 0x1 /* Known never to match null string. */
|
|
368 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
|
|
369 #define SPSTART 0x4 /* Starts with * or +. */
|
|
370 #define HASNL 0x8 /* Contains some \n. */
|
|
371 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */
|
|
372 #define WORST 0 /* Worst case. */
|
|
373
|
|
374 /*
|
|
375 * When regcode is set to this value, code is not emitted and size is computed
|
|
376 * instead.
|
|
377 */
|
|
378 #define JUST_CALC_SIZE ((char_u *) -1)
|
|
379
|
|
380 static char_u *reg_prev_sub;
|
|
381
|
|
382 /*
|
|
383 * REGEXP_INRANGE contains all characters which are always special in a []
|
|
384 * range after '\'.
|
|
385 * REGEXP_ABBR contains all characters which act as abbreviations after '\'.
|
|
386 * These are:
|
|
387 * \n - New line (NL).
|
|
388 * \r - Carriage Return (CR).
|
|
389 * \t - Tab (TAB).
|
|
390 * \e - Escape (ESC).
|
|
391 * \b - Backspace (Ctrl_H).
|
24
|
392 * \d - Character code in decimal, eg \d123
|
|
393 * \o - Character code in octal, eg \o80
|
|
394 * \x - Character code in hex, eg \x4a
|
|
395 * \u - Multibyte character code, eg \u20ac
|
|
396 * \U - Long multibyte character code, eg \U12345678
|
7
|
397 */
|
|
398 static char_u REGEXP_INRANGE[] = "]^-n\\";
|
24
|
399 static char_u REGEXP_ABBR[] = "nrtebdoxuU";
|
7
|
400
|
|
401 static int backslash_trans __ARGS((int c));
|
167
|
402 static int get_char_class __ARGS((char_u **pp));
|
|
403 static int get_equi_class __ARGS((char_u **pp));
|
|
404 static void reg_equi_class __ARGS((int c));
|
|
405 static int get_coll_element __ARGS((char_u **pp));
|
7
|
406 static char_u *skip_anyof __ARGS((char_u *p));
|
|
407 static void init_class_tab __ARGS((void));
|
|
408
|
|
409 /*
|
|
410 * Translate '\x' to its control character, except "\n", which is Magic.
|
|
411 */
|
|
412 static int
|
|
413 backslash_trans(c)
|
|
414 int c;
|
|
415 {
|
|
416 switch (c)
|
|
417 {
|
|
418 case 'r': return CAR;
|
|
419 case 't': return TAB;
|
|
420 case 'e': return ESC;
|
|
421 case 'b': return BS;
|
|
422 }
|
|
423 return c;
|
|
424 }
|
|
425
|
|
426 /*
|
167
|
427 * Check for a character class name "[:name:]". "pp" points to the '['.
|
7
|
428 * Returns one of the CLASS_ items. CLASS_NONE means that no item was
|
|
429 * recognized. Otherwise "pp" is advanced to after the item.
|
|
430 */
|
|
431 static int
|
167
|
432 get_char_class(pp)
|
7
|
433 char_u **pp;
|
|
434 {
|
|
435 static const char *(class_names[]) =
|
|
436 {
|
|
437 "alnum:]",
|
|
438 #define CLASS_ALNUM 0
|
|
439 "alpha:]",
|
|
440 #define CLASS_ALPHA 1
|
|
441 "blank:]",
|
|
442 #define CLASS_BLANK 2
|
|
443 "cntrl:]",
|
|
444 #define CLASS_CNTRL 3
|
|
445 "digit:]",
|
|
446 #define CLASS_DIGIT 4
|
|
447 "graph:]",
|
|
448 #define CLASS_GRAPH 5
|
|
449 "lower:]",
|
|
450 #define CLASS_LOWER 6
|
|
451 "print:]",
|
|
452 #define CLASS_PRINT 7
|
|
453 "punct:]",
|
|
454 #define CLASS_PUNCT 8
|
|
455 "space:]",
|
|
456 #define CLASS_SPACE 9
|
|
457 "upper:]",
|
|
458 #define CLASS_UPPER 10
|
|
459 "xdigit:]",
|
|
460 #define CLASS_XDIGIT 11
|
|
461 "tab:]",
|
|
462 #define CLASS_TAB 12
|
|
463 "return:]",
|
|
464 #define CLASS_RETURN 13
|
|
465 "backspace:]",
|
|
466 #define CLASS_BACKSPACE 14
|
|
467 "escape:]",
|
|
468 #define CLASS_ESCAPE 15
|
|
469 };
|
|
470 #define CLASS_NONE 99
|
|
471 int i;
|
|
472
|
|
473 if ((*pp)[1] == ':')
|
|
474 {
|
|
475 for (i = 0; i < sizeof(class_names) / sizeof(*class_names); ++i)
|
|
476 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0)
|
|
477 {
|
|
478 *pp += STRLEN(class_names[i]) + 2;
|
|
479 return i;
|
|
480 }
|
|
481 }
|
|
482 return CLASS_NONE;
|
|
483 }
|
|
484
|
|
485 /*
|
|
486 * Specific version of character class functions.
|
|
487 * Using a table to keep this fast.
|
|
488 */
|
|
489 static short class_tab[256];
|
|
490
|
|
491 #define RI_DIGIT 0x01
|
|
492 #define RI_HEX 0x02
|
|
493 #define RI_OCTAL 0x04
|
|
494 #define RI_WORD 0x08
|
|
495 #define RI_HEAD 0x10
|
|
496 #define RI_ALPHA 0x20
|
|
497 #define RI_LOWER 0x40
|
|
498 #define RI_UPPER 0x80
|
|
499 #define RI_WHITE 0x100
|
|
500
|
|
501 static void
|
|
502 init_class_tab()
|
|
503 {
|
|
504 int i;
|
|
505 static int done = FALSE;
|
|
506
|
|
507 if (done)
|
|
508 return;
|
|
509
|
|
510 for (i = 0; i < 256; ++i)
|
|
511 {
|
|
512 if (i >= '0' && i <= '7')
|
|
513 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD;
|
|
514 else if (i >= '8' && i <= '9')
|
|
515 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD;
|
|
516 else if (i >= 'a' && i <= 'f')
|
|
517 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
518 #ifdef EBCDIC
|
|
519 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r')
|
|
520 || (i >= 's' && i <= 'z'))
|
|
521 #else
|
|
522 else if (i >= 'g' && i <= 'z')
|
|
523 #endif
|
|
524 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
|
|
525 else if (i >= 'A' && i <= 'F')
|
|
526 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
527 #ifdef EBCDIC
|
|
528 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R')
|
|
529 || (i >= 'S' && i <= 'Z'))
|
|
530 #else
|
|
531 else if (i >= 'G' && i <= 'Z')
|
|
532 #endif
|
|
533 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
|
|
534 else if (i == '_')
|
|
535 class_tab[i] = RI_WORD + RI_HEAD;
|
|
536 else
|
|
537 class_tab[i] = 0;
|
|
538 }
|
|
539 class_tab[' '] |= RI_WHITE;
|
|
540 class_tab['\t'] |= RI_WHITE;
|
|
541 done = TRUE;
|
|
542 }
|
|
543
|
|
544 #ifdef FEAT_MBYTE
|
|
545 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT))
|
|
546 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX))
|
|
547 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL))
|
|
548 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD))
|
|
549 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD))
|
|
550 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA))
|
|
551 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER))
|
|
552 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER))
|
|
553 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE))
|
|
554 #else
|
|
555 # define ri_digit(c) (class_tab[c] & RI_DIGIT)
|
|
556 # define ri_hex(c) (class_tab[c] & RI_HEX)
|
|
557 # define ri_octal(c) (class_tab[c] & RI_OCTAL)
|
|
558 # define ri_word(c) (class_tab[c] & RI_WORD)
|
|
559 # define ri_head(c) (class_tab[c] & RI_HEAD)
|
|
560 # define ri_alpha(c) (class_tab[c] & RI_ALPHA)
|
|
561 # define ri_lower(c) (class_tab[c] & RI_LOWER)
|
|
562 # define ri_upper(c) (class_tab[c] & RI_UPPER)
|
|
563 # define ri_white(c) (class_tab[c] & RI_WHITE)
|
|
564 #endif
|
|
565
|
|
566 /* flags for regflags */
|
|
567 #define RF_ICASE 1 /* ignore case */
|
|
568 #define RF_NOICASE 2 /* don't ignore case */
|
|
569 #define RF_HASNL 4 /* can match a NL */
|
|
570 #define RF_ICOMBINE 8 /* ignore combining characters */
|
|
571 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */
|
|
572
|
|
573 /*
|
|
574 * Global work variables for vim_regcomp().
|
|
575 */
|
|
576
|
|
577 static char_u *regparse; /* Input-scan pointer. */
|
|
578 static int prevchr_len; /* byte length of previous char */
|
|
579 static int num_complex_braces; /* Complex \{...} count */
|
|
580 static int regnpar; /* () count. */
|
|
581 #ifdef FEAT_SYN_HL
|
|
582 static int regnzpar; /* \z() count. */
|
|
583 static int re_has_z; /* \z item detected */
|
|
584 #endif
|
|
585 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */
|
|
586 static long regsize; /* Code size. */
|
|
587 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */
|
|
588 static unsigned regflags; /* RF_ flags for prog */
|
|
589 static long brace_min[10]; /* Minimums for complex brace repeats */
|
|
590 static long brace_max[10]; /* Maximums for complex brace repeats */
|
|
591 static int brace_count[10]; /* Current counts for complex brace repeats */
|
|
592 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
593 static int had_eol; /* TRUE when EOL found by vim_regcomp() */
|
|
594 #endif
|
|
595 static int one_exactly = FALSE; /* only do one char for EXACTLY */
|
|
596
|
|
597 static int reg_magic; /* magicness of the pattern: */
|
|
598 #define MAGIC_NONE 1 /* "\V" very unmagic */
|
|
599 #define MAGIC_OFF 2 /* "\M" or 'magic' off */
|
|
600 #define MAGIC_ON 3 /* "\m" or 'magic' */
|
|
601 #define MAGIC_ALL 4 /* "\v" very magic */
|
|
602
|
|
603 static int reg_string; /* matching with a string instead of a buffer
|
|
604 line */
|
|
605
|
|
606 /*
|
|
607 * META contains all characters that may be magic, except '^' and '$'.
|
|
608 */
|
|
609
|
|
610 #ifdef EBCDIC
|
|
611 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~";
|
|
612 #else
|
|
613 /* META[] is used often enough to justify turning it into a table. */
|
|
614 static char_u META_flags[] = {
|
|
615 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
616 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
617 /* % & ( ) * + . */
|
|
618 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0,
|
|
619 /* 1 2 3 4 5 6 7 8 9 < = > ? */
|
|
620 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1,
|
|
621 /* @ A C D F H I K L M O */
|
|
622 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
|
|
623 /* P S U V W X Z [ _ */
|
|
624 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1,
|
|
625 /* a c d f h i k l m n o */
|
|
626 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
|
|
627 /* p s u v w x z { | ~ */
|
|
628 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1
|
|
629 };
|
|
630 #endif
|
|
631
|
|
632 static int curchr;
|
|
633
|
|
634 /* arguments for reg() */
|
|
635 #define REG_NOPAREN 0 /* toplevel reg() */
|
|
636 #define REG_PAREN 1 /* \(\) */
|
|
637 #define REG_ZPAREN 2 /* \z(\) */
|
|
638 #define REG_NPAREN 3 /* \%(\) */
|
|
639
|
|
640 /*
|
|
641 * Forward declarations for vim_regcomp()'s friends.
|
|
642 */
|
|
643 static void initchr __ARGS((char_u *));
|
|
644 static int getchr __ARGS((void));
|
|
645 static void skipchr_keepstart __ARGS((void));
|
|
646 static int peekchr __ARGS((void));
|
|
647 static void skipchr __ARGS((void));
|
|
648 static void ungetchr __ARGS((void));
|
24
|
649 static int gethexchrs __ARGS((int maxinputlen));
|
|
650 static int getoctchrs __ARGS((void));
|
|
651 static int getdecchrs __ARGS((void));
|
|
652 static int coll_get_char __ARGS((void));
|
7
|
653 static void regcomp_start __ARGS((char_u *expr, int flags));
|
|
654 static char_u *reg __ARGS((int, int *));
|
|
655 static char_u *regbranch __ARGS((int *flagp));
|
|
656 static char_u *regconcat __ARGS((int *flagp));
|
|
657 static char_u *regpiece __ARGS((int *));
|
|
658 static char_u *regatom __ARGS((int *));
|
|
659 static char_u *regnode __ARGS((int));
|
|
660 static int prog_magic_wrong __ARGS((void));
|
|
661 static char_u *regnext __ARGS((char_u *));
|
|
662 static void regc __ARGS((int b));
|
|
663 #ifdef FEAT_MBYTE
|
|
664 static void regmbc __ARGS((int c));
|
167
|
665 #else
|
|
666 # define regmbc(c) regc(c)
|
7
|
667 #endif
|
|
668 static void reginsert __ARGS((int, char_u *));
|
|
669 static void reginsert_limits __ARGS((int, long, long, char_u *));
|
|
670 static char_u *re_put_long __ARGS((char_u *pr, long_u val));
|
|
671 static int read_limits __ARGS((long *, long *));
|
|
672 static void regtail __ARGS((char_u *, char_u *));
|
|
673 static void regoptail __ARGS((char_u *, char_u *));
|
|
674
|
|
675 /*
|
|
676 * Return TRUE if compiled regular expression "prog" can match a line break.
|
|
677 */
|
|
678 int
|
|
679 re_multiline(prog)
|
|
680 regprog_T *prog;
|
|
681 {
|
|
682 return (prog->regflags & RF_HASNL);
|
|
683 }
|
|
684
|
|
685 /*
|
|
686 * Return TRUE if compiled regular expression "prog" looks before the start
|
|
687 * position (pattern contains "\@<=" or "\@<!").
|
|
688 */
|
|
689 int
|
|
690 re_lookbehind(prog)
|
|
691 regprog_T *prog;
|
|
692 {
|
|
693 return (prog->regflags & RF_LOOKBH);
|
|
694 }
|
|
695
|
|
696 /*
|
167
|
697 * Check for an equivalence class name "[=a=]". "pp" points to the '['.
|
|
698 * Returns a character representing the class. Zero means that no item was
|
|
699 * recognized. Otherwise "pp" is advanced to after the item.
|
|
700 */
|
|
701 static int
|
|
702 get_equi_class(pp)
|
|
703 char_u **pp;
|
|
704 {
|
|
705 int c;
|
|
706 int l = 1;
|
|
707 char_u *p = *pp;
|
|
708
|
|
709 if (p[1] == '=')
|
|
710 {
|
|
711 #ifdef FEAT_MBYTE
|
|
712 if (has_mbyte)
|
|
713 l = mb_ptr2len_check(p + 2);
|
|
714 #endif
|
|
715 if (p[l + 2] == '=' && p[l + 3] == ']')
|
|
716 {
|
|
717 #ifdef FEAT_MBYTE
|
|
718 if (has_mbyte)
|
|
719 c = mb_ptr2char(p + 2);
|
|
720 else
|
|
721 #endif
|
|
722 c = p[2];
|
|
723 *pp += l + 4;
|
|
724 return c;
|
|
725 }
|
|
726 }
|
|
727 return 0;
|
|
728 }
|
|
729
|
|
730 /*
|
|
731 * Produce the bytes for equivalence class "c".
|
|
732 * Currently only handles latin1, latin9 and utf-8.
|
|
733 */
|
|
734 static void
|
|
735 reg_equi_class(c)
|
|
736 int c;
|
|
737 {
|
|
738 #ifdef FEAT_MBYTE
|
|
739 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0
|
|
740 || STRCMP(p_enc, "latin9") == 0)
|
|
741 #endif
|
|
742 {
|
|
743 switch (c)
|
|
744 {
|
|
745 case 'A': case 'À': case 'Á': case 'Â':
|
|
746 case 'Ã': case 'Ä': case 'Å':
|
|
747 regmbc('A'); regmbc('À'); regmbc('Á'); regmbc('Â');
|
|
748 regmbc('Ã'); regmbc('Ä'); regmbc('Å');
|
|
749 return;
|
|
750 case 'C': case 'Ç':
|
|
751 regmbc('C'); regmbc('Ç');
|
|
752 return;
|
|
753 case 'E': case 'È': case 'É': case 'Ê': case 'Ë':
|
|
754 regmbc('E'); regmbc('È'); regmbc('É'); regmbc('Ê');
|
|
755 regmbc('Ë');
|
|
756 return;
|
|
757 case 'I': case 'Ì': case 'Í': case 'Î': case 'Ï':
|
|
758 regmbc('I'); regmbc('Ì'); regmbc('Í'); regmbc('Î');
|
|
759 regmbc('Ï');
|
|
760 return;
|
|
761 case 'N': case 'Ñ':
|
|
762 regmbc('N'); regmbc('Ñ');
|
|
763 return;
|
|
764 case 'O': case 'Ò': case 'Ó': case 'Ô': case 'Õ': case 'Ö':
|
|
765 regmbc('O'); regmbc('Ò'); regmbc('Ó'); regmbc('Ô');
|
|
766 regmbc('Õ'); regmbc('Ö');
|
|
767 return;
|
|
768 case 'U': case 'Ù': case 'Ú': case 'Û': case 'Ü':
|
|
769 regmbc('U'); regmbc('Ù'); regmbc('Ú'); regmbc('Û');
|
|
770 regmbc('Ü');
|
|
771 return;
|
|
772 case 'Y': case 'Ý':
|
|
773 regmbc('Y'); regmbc('Ý');
|
|
774 return;
|
|
775 case 'a': case 'à': case 'á': case 'â':
|
|
776 case 'ã': case 'ä': case 'å':
|
|
777 regmbc('a'); regmbc('à'); regmbc('á'); regmbc('â');
|
|
778 regmbc('ã'); regmbc('ä'); regmbc('å');
|
|
779 return;
|
|
780 case 'c': case 'ç':
|
|
781 regmbc('c'); regmbc('ç');
|
|
782 return;
|
|
783 case 'e': case 'è': case 'é': case 'ê': case 'ë':
|
|
784 regmbc('e'); regmbc('è'); regmbc('é'); regmbc('ê');
|
|
785 regmbc('ë');
|
|
786 return;
|
|
787 case 'i': case 'ì': case 'í': case 'î': case 'ï':
|
|
788 regmbc('i'); regmbc('ì'); regmbc('í'); regmbc('î');
|
|
789 regmbc('ï');
|
|
790 return;
|
|
791 case 'n': case 'ñ':
|
|
792 regmbc('n'); regmbc('ñ');
|
|
793 return;
|
|
794 case 'o': case 'ò': case 'ó': case 'ô': case 'õ': case 'ö':
|
|
795 regmbc('o'); regmbc('ò'); regmbc('ó'); regmbc('ô');
|
|
796 regmbc('õ'); regmbc('ö');
|
|
797 return;
|
|
798 case 'u': case 'ù': case 'ú': case 'û': case 'ü':
|
|
799 regmbc('u'); regmbc('ù'); regmbc('ú'); regmbc('û');
|
|
800 regmbc('ü');
|
|
801 return;
|
|
802 case 'y': case 'ý': case 'ÿ':
|
|
803 regmbc('y'); regmbc('ý'); regmbc('ÿ');
|
|
804 return;
|
|
805 }
|
|
806 }
|
|
807 regmbc(c);
|
|
808 }
|
|
809
|
|
810 /*
|
|
811 * Check for a collating element "[.a.]". "pp" points to the '['.
|
|
812 * Returns a character. Zero means that no item was recognized. Otherwise
|
|
813 * "pp" is advanced to after the item.
|
|
814 * Currently only single characters are recognized!
|
|
815 */
|
|
816 static int
|
|
817 get_coll_element(pp)
|
|
818 char_u **pp;
|
|
819 {
|
|
820 int c;
|
|
821 int l = 1;
|
|
822 char_u *p = *pp;
|
|
823
|
|
824 if (p[1] == '.')
|
|
825 {
|
|
826 #ifdef FEAT_MBYTE
|
|
827 if (has_mbyte)
|
|
828 l = mb_ptr2len_check(p + 2);
|
|
829 #endif
|
|
830 if (p[l + 2] == '.' && p[l + 3] == ']')
|
|
831 {
|
|
832 #ifdef FEAT_MBYTE
|
|
833 if (has_mbyte)
|
|
834 c = mb_ptr2char(p + 2);
|
|
835 else
|
|
836 #endif
|
|
837 c = p[2];
|
|
838 *pp += l + 4;
|
|
839 return c;
|
|
840 }
|
|
841 }
|
|
842 return 0;
|
|
843 }
|
|
844
|
|
845
|
|
846 /*
|
|
847 * Skip over a "[]" range.
|
|
848 * "p" must point to the character after the '['.
|
|
849 * The returned pointer is on the matching ']', or the terminating NUL.
|
|
850 */
|
|
851 static char_u *
|
|
852 skip_anyof(p)
|
|
853 char_u *p;
|
|
854 {
|
|
855 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
|
856 int cpo_bsl; /* 'cpoptions' contains '\' flag */
|
|
857 #ifdef FEAT_MBYTE
|
|
858 int l;
|
|
859 #endif
|
|
860
|
|
861 cpo_lit = (!reg_syn && vim_strchr(p_cpo, CPO_LITERAL) != NULL);
|
|
862 cpo_bsl = (!reg_syn && vim_strchr(p_cpo, CPO_BACKSL) != NULL);
|
|
863
|
|
864 if (*p == '^') /* Complement of range. */
|
|
865 ++p;
|
|
866 if (*p == ']' || *p == '-')
|
|
867 ++p;
|
|
868 while (*p != NUL && *p != ']')
|
|
869 {
|
|
870 #ifdef FEAT_MBYTE
|
|
871 if (has_mbyte && (l = (*mb_ptr2len_check)(p)) > 1)
|
|
872 p += l;
|
|
873 else
|
|
874 #endif
|
|
875 if (*p == '-')
|
|
876 {
|
|
877 ++p;
|
|
878 if (*p != ']' && *p != NUL)
|
|
879 mb_ptr_adv(p);
|
|
880 }
|
|
881 else if (*p == '\\'
|
|
882 && !cpo_bsl
|
|
883 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL
|
|
884 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL)))
|
|
885 p += 2;
|
|
886 else if (*p == '[')
|
|
887 {
|
|
888 if (get_char_class(&p) == CLASS_NONE
|
|
889 && get_equi_class(&p) == 0
|
|
890 && get_coll_element(&p) == 0)
|
|
891 ++p; /* It was not a class name */
|
|
892 }
|
|
893 else
|
|
894 ++p;
|
|
895 }
|
|
896
|
|
897 return p;
|
|
898 }
|
|
899
|
|
900 /*
|
7
|
901 * Skip past regular expression.
|
153
|
902 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc).
|
7
|
903 * Take care of characters with a backslash in front of it.
|
|
904 * Skip strings inside [ and ].
|
|
905 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the
|
|
906 * expression and change "\?" to "?". If "*newp" is not NULL the expression
|
|
907 * is changed in-place.
|
|
908 */
|
|
909 char_u *
|
|
910 skip_regexp(startp, dirc, magic, newp)
|
|
911 char_u *startp;
|
|
912 int dirc;
|
|
913 int magic;
|
|
914 char_u **newp;
|
|
915 {
|
|
916 int mymagic;
|
|
917 char_u *p = startp;
|
|
918
|
|
919 if (magic)
|
|
920 mymagic = MAGIC_ON;
|
|
921 else
|
|
922 mymagic = MAGIC_OFF;
|
|
923
|
39
|
924 for (; p[0] != NUL; mb_ptr_adv(p))
|
7
|
925 {
|
|
926 if (p[0] == dirc) /* found end of regexp */
|
|
927 break;
|
|
928 if ((p[0] == '[' && mymagic >= MAGIC_ON)
|
|
929 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF))
|
|
930 {
|
|
931 p = skip_anyof(p + 1);
|
|
932 if (p[0] == NUL)
|
|
933 break;
|
|
934 }
|
|
935 else if (p[0] == '\\' && p[1] != NUL)
|
|
936 {
|
|
937 if (dirc == '?' && newp != NULL && p[1] == '?')
|
|
938 {
|
|
939 /* change "\?" to "?", make a copy first. */
|
|
940 if (*newp == NULL)
|
|
941 {
|
|
942 *newp = vim_strsave(startp);
|
|
943 if (*newp != NULL)
|
|
944 p = *newp + (p - startp);
|
|
945 }
|
|
946 if (*newp != NULL)
|
|
947 mch_memmove(p, p + 1, STRLEN(p));
|
|
948 else
|
|
949 ++p;
|
|
950 }
|
|
951 else
|
|
952 ++p; /* skip next character */
|
|
953 if (*p == 'v')
|
|
954 mymagic = MAGIC_ALL;
|
|
955 else if (*p == 'V')
|
|
956 mymagic = MAGIC_NONE;
|
|
957 }
|
|
958 }
|
|
959 return p;
|
|
960 }
|
|
961
|
|
962 /*
|
41
|
963 * vim_regcomp() - compile a regular expression into internal code
|
|
964 * Returns the program in allocated space. Returns NULL for an error.
|
7
|
965 *
|
|
966 * We can't allocate space until we know how big the compiled form will be,
|
|
967 * but we can't compile it (and thus know how big it is) until we've got a
|
|
968 * place to put the code. So we cheat: we compile it twice, once with code
|
|
969 * generation turned off and size counting turned on, and once "for real".
|
|
970 * This also means that we don't allocate space until we are sure that the
|
|
971 * thing really will compile successfully, and we never have to move the
|
|
972 * code and thus invalidate pointers into it. (Note that it has to be in
|
|
973 * one piece because vim_free() must be able to free it all.)
|
|
974 *
|
|
975 * Whether upper/lower case is to be ignored is decided when executing the
|
|
976 * program, it does not matter here.
|
|
977 *
|
|
978 * Beware that the optimization-preparation code in here knows about some
|
|
979 * of the structure of the compiled regexp.
|
|
980 * "re_flags": RE_MAGIC and/or RE_STRING.
|
|
981 */
|
|
982 regprog_T *
|
|
983 vim_regcomp(expr, re_flags)
|
|
984 char_u *expr;
|
|
985 int re_flags;
|
|
986 {
|
|
987 regprog_T *r;
|
|
988 char_u *scan;
|
|
989 char_u *longest;
|
|
990 int len;
|
|
991 int flags;
|
|
992
|
|
993 if (expr == NULL)
|
|
994 EMSG_RET_NULL(_(e_null));
|
|
995
|
|
996 init_class_tab();
|
|
997
|
|
998 /*
|
|
999 * First pass: determine size, legality.
|
|
1000 */
|
|
1001 regcomp_start(expr, re_flags);
|
|
1002 regcode = JUST_CALC_SIZE;
|
|
1003 regc(REGMAGIC);
|
|
1004 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
1005 return NULL;
|
|
1006
|
|
1007 /* Small enough for pointer-storage convention? */
|
|
1008 #ifdef SMALL_MALLOC /* 16 bit storage allocation */
|
|
1009 if (regsize >= 65536L - 256L)
|
|
1010 EMSG_RET_NULL(_("E339: Pattern too long"));
|
|
1011 #endif
|
|
1012
|
|
1013 /* Allocate space. */
|
|
1014 r = (regprog_T *)lalloc(sizeof(regprog_T) + regsize, TRUE);
|
|
1015 if (r == NULL)
|
|
1016 return NULL;
|
|
1017
|
|
1018 /*
|
|
1019 * Second pass: emit code.
|
|
1020 */
|
|
1021 regcomp_start(expr, re_flags);
|
|
1022 regcode = r->program;
|
|
1023 regc(REGMAGIC);
|
|
1024 if (reg(REG_NOPAREN, &flags) == NULL)
|
|
1025 {
|
|
1026 vim_free(r);
|
|
1027 return NULL;
|
|
1028 }
|
|
1029
|
|
1030 /* Dig out information for optimizations. */
|
|
1031 r->regstart = NUL; /* Worst-case defaults. */
|
|
1032 r->reganch = 0;
|
|
1033 r->regmust = NULL;
|
|
1034 r->regmlen = 0;
|
|
1035 r->regflags = regflags;
|
|
1036 if (flags & HASNL)
|
|
1037 r->regflags |= RF_HASNL;
|
|
1038 if (flags & HASLOOKBH)
|
|
1039 r->regflags |= RF_LOOKBH;
|
|
1040 #ifdef FEAT_SYN_HL
|
|
1041 /* Remember whether this pattern has any \z specials in it. */
|
|
1042 r->reghasz = re_has_z;
|
|
1043 #endif
|
|
1044 scan = r->program + 1; /* First BRANCH. */
|
|
1045 if (OP(regnext(scan)) == END) /* Only one top-level choice. */
|
|
1046 {
|
|
1047 scan = OPERAND(scan);
|
|
1048
|
|
1049 /* Starting-point info. */
|
|
1050 if (OP(scan) == BOL || OP(scan) == RE_BOF)
|
|
1051 {
|
|
1052 r->reganch++;
|
|
1053 scan = regnext(scan);
|
|
1054 }
|
|
1055
|
|
1056 if (OP(scan) == EXACTLY)
|
|
1057 {
|
|
1058 #ifdef FEAT_MBYTE
|
|
1059 if (has_mbyte)
|
|
1060 r->regstart = (*mb_ptr2char)(OPERAND(scan));
|
|
1061 else
|
|
1062 #endif
|
|
1063 r->regstart = *OPERAND(scan);
|
|
1064 }
|
|
1065 else if ((OP(scan) == BOW
|
|
1066 || OP(scan) == EOW
|
|
1067 || OP(scan) == NOTHING
|
|
1068 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
|
|
1069 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
|
|
1070 && OP(regnext(scan)) == EXACTLY)
|
|
1071 {
|
|
1072 #ifdef FEAT_MBYTE
|
|
1073 if (has_mbyte)
|
|
1074 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
|
|
1075 else
|
|
1076 #endif
|
|
1077 r->regstart = *OPERAND(regnext(scan));
|
|
1078 }
|
|
1079
|
|
1080 /*
|
|
1081 * If there's something expensive in the r.e., find the longest
|
|
1082 * literal string that must appear and make it the regmust. Resolve
|
|
1083 * ties in favor of later strings, since the regstart check works
|
|
1084 * with the beginning of the r.e. and avoiding duplication
|
|
1085 * strengthens checking. Not a strong reason, but sufficient in the
|
|
1086 * absence of others.
|
|
1087 */
|
|
1088 /*
|
|
1089 * When the r.e. starts with BOW, it is faster to look for a regmust
|
|
1090 * first. Used a lot for "#" and "*" commands. (Added by mool).
|
|
1091 */
|
|
1092 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
|
|
1093 && !(flags & HASNL))
|
|
1094 {
|
|
1095 longest = NULL;
|
|
1096 len = 0;
|
|
1097 for (; scan != NULL; scan = regnext(scan))
|
|
1098 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
|
|
1099 {
|
|
1100 longest = OPERAND(scan);
|
|
1101 len = (int)STRLEN(OPERAND(scan));
|
|
1102 }
|
|
1103 r->regmust = longest;
|
|
1104 r->regmlen = len;
|
|
1105 }
|
|
1106 }
|
|
1107 #ifdef DEBUG
|
|
1108 regdump(expr, r);
|
|
1109 #endif
|
|
1110 return r;
|
|
1111 }
|
|
1112
|
|
1113 /*
|
|
1114 * Setup to parse the regexp. Used once to get the length and once to do it.
|
|
1115 */
|
|
1116 static void
|
|
1117 regcomp_start(expr, re_flags)
|
|
1118 char_u *expr;
|
|
1119 int re_flags; /* see vim_regcomp() */
|
|
1120 {
|
|
1121 initchr(expr);
|
|
1122 if (re_flags & RE_MAGIC)
|
|
1123 reg_magic = MAGIC_ON;
|
|
1124 else
|
|
1125 reg_magic = MAGIC_OFF;
|
|
1126 reg_string = (re_flags & RE_STRING);
|
|
1127
|
|
1128 num_complex_braces = 0;
|
|
1129 regnpar = 1;
|
|
1130 vim_memset(had_endbrace, 0, sizeof(had_endbrace));
|
|
1131 #ifdef FEAT_SYN_HL
|
|
1132 regnzpar = 1;
|
|
1133 re_has_z = 0;
|
|
1134 #endif
|
|
1135 regsize = 0L;
|
|
1136 regflags = 0;
|
|
1137 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1138 had_eol = FALSE;
|
|
1139 #endif
|
|
1140 }
|
|
1141
|
|
1142 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1143 /*
|
|
1144 * Check if during the previous call to vim_regcomp the EOL item "$" has been
|
|
1145 * found. This is messy, but it works fine.
|
|
1146 */
|
|
1147 int
|
|
1148 vim_regcomp_had_eol()
|
|
1149 {
|
|
1150 return had_eol;
|
|
1151 }
|
|
1152 #endif
|
|
1153
|
|
1154 /*
|
|
1155 * reg - regular expression, i.e. main body or parenthesized thing
|
|
1156 *
|
|
1157 * Caller must absorb opening parenthesis.
|
|
1158 *
|
|
1159 * Combining parenthesis handling with the base level of regular expression
|
|
1160 * is a trifle forced, but the need to tie the tails of the branches to what
|
|
1161 * follows makes it hard to avoid.
|
|
1162 */
|
|
1163 static char_u *
|
|
1164 reg(paren, flagp)
|
|
1165 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */
|
|
1166 int *flagp;
|
|
1167 {
|
|
1168 char_u *ret;
|
|
1169 char_u *br;
|
|
1170 char_u *ender;
|
|
1171 int parno = 0;
|
|
1172 int flags;
|
|
1173
|
|
1174 *flagp = HASWIDTH; /* Tentatively. */
|
|
1175
|
|
1176 #ifdef FEAT_SYN_HL
|
|
1177 if (paren == REG_ZPAREN)
|
|
1178 {
|
|
1179 /* Make a ZOPEN node. */
|
|
1180 if (regnzpar >= NSUBEXP)
|
|
1181 EMSG_RET_NULL(_("E50: Too many \\z("));
|
|
1182 parno = regnzpar;
|
|
1183 regnzpar++;
|
|
1184 ret = regnode(ZOPEN + parno);
|
|
1185 }
|
|
1186 else
|
|
1187 #endif
|
|
1188 if (paren == REG_PAREN)
|
|
1189 {
|
|
1190 /* Make a MOPEN node. */
|
|
1191 if (regnpar >= NSUBEXP)
|
|
1192 EMSG_M_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL);
|
|
1193 parno = regnpar;
|
|
1194 ++regnpar;
|
|
1195 ret = regnode(MOPEN + parno);
|
|
1196 }
|
|
1197 else if (paren == REG_NPAREN)
|
|
1198 {
|
|
1199 /* Make a NOPEN node. */
|
|
1200 ret = regnode(NOPEN);
|
|
1201 }
|
|
1202 else
|
|
1203 ret = NULL;
|
|
1204
|
|
1205 /* Pick up the branches, linking them together. */
|
|
1206 br = regbranch(&flags);
|
|
1207 if (br == NULL)
|
|
1208 return NULL;
|
|
1209 if (ret != NULL)
|
|
1210 regtail(ret, br); /* [MZ]OPEN -> first. */
|
|
1211 else
|
|
1212 ret = br;
|
|
1213 /* If one of the branches can be zero-width, the whole thing can.
|
|
1214 * If one of the branches has * at start or matches a line-break, the
|
|
1215 * whole thing can. */
|
|
1216 if (!(flags & HASWIDTH))
|
|
1217 *flagp &= ~HASWIDTH;
|
|
1218 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1219 while (peekchr() == Magic('|'))
|
|
1220 {
|
|
1221 skipchr();
|
|
1222 br = regbranch(&flags);
|
|
1223 if (br == NULL)
|
|
1224 return NULL;
|
|
1225 regtail(ret, br); /* BRANCH -> BRANCH. */
|
|
1226 if (!(flags & HASWIDTH))
|
|
1227 *flagp &= ~HASWIDTH;
|
|
1228 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
|
|
1229 }
|
|
1230
|
|
1231 /* Make a closing node, and hook it on the end. */
|
|
1232 ender = regnode(
|
|
1233 #ifdef FEAT_SYN_HL
|
|
1234 paren == REG_ZPAREN ? ZCLOSE + parno :
|
|
1235 #endif
|
|
1236 paren == REG_PAREN ? MCLOSE + parno :
|
|
1237 paren == REG_NPAREN ? NCLOSE : END);
|
|
1238 regtail(ret, ender);
|
|
1239
|
|
1240 /* Hook the tails of the branches to the closing node. */
|
|
1241 for (br = ret; br != NULL; br = regnext(br))
|
|
1242 regoptail(br, ender);
|
|
1243
|
|
1244 /* Check for proper termination. */
|
|
1245 if (paren != REG_NOPAREN && getchr() != Magic(')'))
|
|
1246 {
|
|
1247 #ifdef FEAT_SYN_HL
|
|
1248 if (paren == REG_ZPAREN)
|
|
1249 EMSG_RET_NULL(_("E52: Unmatched \\z("))
|
|
1250 else
|
|
1251 #endif
|
|
1252 if (paren == REG_NPAREN)
|
|
1253 EMSG_M_RET_NULL(_("E53: Unmatched %s%%("), reg_magic == MAGIC_ALL)
|
|
1254 else
|
|
1255 EMSG_M_RET_NULL(_("E54: Unmatched %s("), reg_magic == MAGIC_ALL)
|
|
1256 }
|
|
1257 else if (paren == REG_NOPAREN && peekchr() != NUL)
|
|
1258 {
|
|
1259 if (curchr == Magic(')'))
|
|
1260 EMSG_M_RET_NULL(_("E55: Unmatched %s)"), reg_magic == MAGIC_ALL)
|
|
1261 else
|
|
1262 EMSG_RET_NULL(_(e_trailing)) /* "Can't happen". */
|
|
1263 /* NOTREACHED */
|
|
1264 }
|
|
1265 /*
|
|
1266 * Here we set the flag allowing back references to this set of
|
|
1267 * parentheses.
|
|
1268 */
|
|
1269 if (paren == REG_PAREN)
|
|
1270 had_endbrace[parno] = TRUE; /* have seen the close paren */
|
|
1271 return ret;
|
|
1272 }
|
|
1273
|
|
1274 /*
|
|
1275 * regbranch - one alternative of an | operator
|
|
1276 *
|
|
1277 * Implements the & operator.
|
|
1278 */
|
|
1279 static char_u *
|
|
1280 regbranch(flagp)
|
|
1281 int *flagp;
|
|
1282 {
|
|
1283 char_u *ret;
|
|
1284 char_u *chain = NULL;
|
|
1285 char_u *latest;
|
|
1286 int flags;
|
|
1287
|
|
1288 *flagp = WORST | HASNL; /* Tentatively. */
|
|
1289
|
|
1290 ret = regnode(BRANCH);
|
|
1291 for (;;)
|
|
1292 {
|
|
1293 latest = regconcat(&flags);
|
|
1294 if (latest == NULL)
|
|
1295 return NULL;
|
|
1296 /* If one of the branches has width, the whole thing has. If one of
|
|
1297 * the branches anchors at start-of-line, the whole thing does.
|
|
1298 * If one of the branches uses look-behind, the whole thing does. */
|
|
1299 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
|
|
1300 /* If one of the branches doesn't match a line-break, the whole thing
|
|
1301 * doesn't. */
|
|
1302 *flagp &= ~HASNL | (flags & HASNL);
|
|
1303 if (chain != NULL)
|
|
1304 regtail(chain, latest);
|
|
1305 if (peekchr() != Magic('&'))
|
|
1306 break;
|
|
1307 skipchr();
|
|
1308 regtail(latest, regnode(END)); /* operand ends */
|
|
1309 reginsert(MATCH, latest);
|
|
1310 chain = latest;
|
|
1311 }
|
|
1312
|
|
1313 return ret;
|
|
1314 }
|
|
1315
|
|
1316 /*
|
|
1317 * regbranch - one alternative of an | or & operator
|
|
1318 *
|
|
1319 * Implements the concatenation operator.
|
|
1320 */
|
|
1321 static char_u *
|
|
1322 regconcat(flagp)
|
|
1323 int *flagp;
|
|
1324 {
|
|
1325 char_u *first = NULL;
|
|
1326 char_u *chain = NULL;
|
|
1327 char_u *latest;
|
|
1328 int flags;
|
|
1329 int cont = TRUE;
|
|
1330
|
|
1331 *flagp = WORST; /* Tentatively. */
|
|
1332
|
|
1333 while (cont)
|
|
1334 {
|
|
1335 switch (peekchr())
|
|
1336 {
|
|
1337 case NUL:
|
|
1338 case Magic('|'):
|
|
1339 case Magic('&'):
|
|
1340 case Magic(')'):
|
|
1341 cont = FALSE;
|
|
1342 break;
|
|
1343 case Magic('Z'):
|
|
1344 #ifdef FEAT_MBYTE
|
|
1345 regflags |= RF_ICOMBINE;
|
|
1346 #endif
|
|
1347 skipchr_keepstart();
|
|
1348 break;
|
|
1349 case Magic('c'):
|
|
1350 regflags |= RF_ICASE;
|
|
1351 skipchr_keepstart();
|
|
1352 break;
|
|
1353 case Magic('C'):
|
|
1354 regflags |= RF_NOICASE;
|
|
1355 skipchr_keepstart();
|
|
1356 break;
|
|
1357 case Magic('v'):
|
|
1358 reg_magic = MAGIC_ALL;
|
|
1359 skipchr_keepstart();
|
|
1360 curchr = -1;
|
|
1361 break;
|
|
1362 case Magic('m'):
|
|
1363 reg_magic = MAGIC_ON;
|
|
1364 skipchr_keepstart();
|
|
1365 curchr = -1;
|
|
1366 break;
|
|
1367 case Magic('M'):
|
|
1368 reg_magic = MAGIC_OFF;
|
|
1369 skipchr_keepstart();
|
|
1370 curchr = -1;
|
|
1371 break;
|
|
1372 case Magic('V'):
|
|
1373 reg_magic = MAGIC_NONE;
|
|
1374 skipchr_keepstart();
|
|
1375 curchr = -1;
|
|
1376 break;
|
|
1377 default:
|
|
1378 latest = regpiece(&flags);
|
|
1379 if (latest == NULL)
|
|
1380 return NULL;
|
|
1381 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
|
|
1382 if (chain == NULL) /* First piece. */
|
|
1383 *flagp |= flags & SPSTART;
|
|
1384 else
|
|
1385 regtail(chain, latest);
|
|
1386 chain = latest;
|
|
1387 if (first == NULL)
|
|
1388 first = latest;
|
|
1389 break;
|
|
1390 }
|
|
1391 }
|
|
1392 if (first == NULL) /* Loop ran zero times. */
|
|
1393 first = regnode(NOTHING);
|
|
1394 return first;
|
|
1395 }
|
|
1396
|
|
1397 /*
|
|
1398 * regpiece - something followed by possible [*+=]
|
|
1399 *
|
|
1400 * Note that the branching code sequences used for = and the general cases
|
|
1401 * of * and + are somewhat optimized: they use the same NOTHING node as
|
|
1402 * both the endmarker for their branch list and the body of the last branch.
|
|
1403 * It might seem that this node could be dispensed with entirely, but the
|
|
1404 * endmarker role is not redundant.
|
|
1405 */
|
|
1406 static char_u *
|
|
1407 regpiece(flagp)
|
|
1408 int *flagp;
|
|
1409 {
|
|
1410 char_u *ret;
|
|
1411 int op;
|
|
1412 char_u *next;
|
|
1413 int flags;
|
|
1414 long minval;
|
|
1415 long maxval;
|
|
1416
|
|
1417 ret = regatom(&flags);
|
|
1418 if (ret == NULL)
|
|
1419 return NULL;
|
|
1420
|
|
1421 op = peekchr();
|
|
1422 if (re_multi_type(op) == NOT_MULTI)
|
|
1423 {
|
|
1424 *flagp = flags;
|
|
1425 return ret;
|
|
1426 }
|
|
1427 /* default flags */
|
|
1428 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
|
|
1429
|
|
1430 skipchr();
|
|
1431 switch (op)
|
|
1432 {
|
|
1433 case Magic('*'):
|
|
1434 if (flags & SIMPLE)
|
|
1435 reginsert(STAR, ret);
|
|
1436 else
|
|
1437 {
|
|
1438 /* Emit x* as (x&|), where & means "self". */
|
|
1439 reginsert(BRANCH, ret); /* Either x */
|
|
1440 regoptail(ret, regnode(BACK)); /* and loop */
|
|
1441 regoptail(ret, ret); /* back */
|
|
1442 regtail(ret, regnode(BRANCH)); /* or */
|
|
1443 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1444 }
|
|
1445 break;
|
|
1446
|
|
1447 case Magic('+'):
|
|
1448 if (flags & SIMPLE)
|
|
1449 reginsert(PLUS, ret);
|
|
1450 else
|
|
1451 {
|
|
1452 /* Emit x+ as x(&|), where & means "self". */
|
|
1453 next = regnode(BRANCH); /* Either */
|
|
1454 regtail(ret, next);
|
179
|
1455 regtail(regnode(BACKP), ret); /* loop back */
|
7
|
1456 regtail(next, regnode(BRANCH)); /* or */
|
|
1457 regtail(ret, regnode(NOTHING)); /* null. */
|
|
1458 }
|
|
1459 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1460 break;
|
|
1461
|
|
1462 case Magic('@'):
|
|
1463 {
|
|
1464 int lop = END;
|
|
1465
|
|
1466 switch (no_Magic(getchr()))
|
|
1467 {
|
|
1468 case '=': lop = MATCH; break; /* \@= */
|
|
1469 case '!': lop = NOMATCH; break; /* \@! */
|
|
1470 case '>': lop = SUBPAT; break; /* \@> */
|
|
1471 case '<': switch (no_Magic(getchr()))
|
|
1472 {
|
|
1473 case '=': lop = BEHIND; break; /* \@<= */
|
|
1474 case '!': lop = NOBEHIND; break; /* \@<! */
|
|
1475 }
|
|
1476 }
|
|
1477 if (lop == END)
|
|
1478 EMSG_M_RET_NULL(_("E59: invalid character after %s@"),
|
|
1479 reg_magic == MAGIC_ALL);
|
|
1480 /* Look behind must match with behind_pos. */
|
|
1481 if (lop == BEHIND || lop == NOBEHIND)
|
|
1482 {
|
|
1483 regtail(ret, regnode(BHPOS));
|
|
1484 *flagp |= HASLOOKBH;
|
|
1485 }
|
|
1486 regtail(ret, regnode(END)); /* operand ends */
|
|
1487 reginsert(lop, ret);
|
|
1488 break;
|
|
1489 }
|
|
1490
|
|
1491 case Magic('?'):
|
|
1492 case Magic('='):
|
|
1493 /* Emit x= as (x|) */
|
|
1494 reginsert(BRANCH, ret); /* Either x */
|
|
1495 regtail(ret, regnode(BRANCH)); /* or */
|
|
1496 next = regnode(NOTHING); /* null. */
|
|
1497 regtail(ret, next);
|
|
1498 regoptail(ret, next);
|
|
1499 break;
|
|
1500
|
|
1501 case Magic('{'):
|
|
1502 if (!read_limits(&minval, &maxval))
|
|
1503 return NULL;
|
|
1504 if (flags & SIMPLE)
|
|
1505 {
|
|
1506 reginsert(BRACE_SIMPLE, ret);
|
|
1507 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1508 }
|
|
1509 else
|
|
1510 {
|
|
1511 if (num_complex_braces >= 10)
|
|
1512 EMSG_M_RET_NULL(_("E60: Too many complex %s{...}s"),
|
|
1513 reg_magic == MAGIC_ALL);
|
|
1514 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
|
|
1515 regoptail(ret, regnode(BACK));
|
|
1516 regoptail(ret, ret);
|
|
1517 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
|
|
1518 ++num_complex_braces;
|
|
1519 }
|
|
1520 if (minval > 0 && maxval > 0)
|
|
1521 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
|
|
1522 break;
|
|
1523 }
|
|
1524 if (re_multi_type(peekchr()) != NOT_MULTI)
|
|
1525 {
|
|
1526 /* Can't have a multi follow a multi. */
|
|
1527 if (peekchr() == Magic('*'))
|
|
1528 sprintf((char *)IObuff, _("E61: Nested %s*"),
|
|
1529 reg_magic >= MAGIC_ON ? "" : "\\");
|
|
1530 else
|
|
1531 sprintf((char *)IObuff, _("E62: Nested %s%c"),
|
|
1532 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr()));
|
|
1533 EMSG_RET_NULL(IObuff);
|
|
1534 }
|
|
1535
|
|
1536 return ret;
|
|
1537 }
|
|
1538
|
|
1539 /*
|
|
1540 * regatom - the lowest level
|
|
1541 *
|
|
1542 * Optimization: gobbles an entire sequence of ordinary characters so that
|
|
1543 * it can turn them into a single node, which is smaller to store and
|
|
1544 * faster to run. Don't do this when one_exactly is set.
|
|
1545 */
|
|
1546 static char_u *
|
|
1547 regatom(flagp)
|
|
1548 int *flagp;
|
|
1549 {
|
|
1550 char_u *ret;
|
|
1551 int flags;
|
|
1552 int cpo_lit; /* 'cpoptions' contains 'l' flag */
|
167
|
1553 int cpo_bsl; /* 'cpoptions' contains '\' flag */
|
7
|
1554 int c;
|
|
1555 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU";
|
|
1556 static int classcodes[] = {ANY, IDENT, SIDENT, KWORD, SKWORD,
|
|
1557 FNAME, SFNAME, PRINT, SPRINT,
|
|
1558 WHITE, NWHITE, DIGIT, NDIGIT,
|
|
1559 HEX, NHEX, OCTAL, NOCTAL,
|
|
1560 WORD, NWORD, HEAD, NHEAD,
|
|
1561 ALPHA, NALPHA, LOWER, NLOWER,
|
|
1562 UPPER, NUPPER
|
|
1563 };
|
|
1564 char_u *p;
|
|
1565 int extra = 0;
|
|
1566
|
|
1567 *flagp = WORST; /* Tentatively. */
|
|
1568 cpo_lit = (!reg_syn && vim_strchr(p_cpo, CPO_LITERAL) != NULL);
|
167
|
1569 cpo_bsl = (!reg_syn && vim_strchr(p_cpo, CPO_BACKSL) != NULL);
|
7
|
1570
|
|
1571 c = getchr();
|
|
1572 switch (c)
|
|
1573 {
|
|
1574 case Magic('^'):
|
|
1575 ret = regnode(BOL);
|
|
1576 break;
|
|
1577
|
|
1578 case Magic('$'):
|
|
1579 ret = regnode(EOL);
|
|
1580 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1581 had_eol = TRUE;
|
|
1582 #endif
|
|
1583 break;
|
|
1584
|
|
1585 case Magic('<'):
|
|
1586 ret = regnode(BOW);
|
|
1587 break;
|
|
1588
|
|
1589 case Magic('>'):
|
|
1590 ret = regnode(EOW);
|
|
1591 break;
|
|
1592
|
|
1593 case Magic('_'):
|
|
1594 c = no_Magic(getchr());
|
|
1595 if (c == '^') /* "\_^" is start-of-line */
|
|
1596 {
|
|
1597 ret = regnode(BOL);
|
|
1598 break;
|
|
1599 }
|
|
1600 if (c == '$') /* "\_$" is end-of-line */
|
|
1601 {
|
|
1602 ret = regnode(EOL);
|
|
1603 #if defined(FEAT_SYN_HL) || defined(PROTO)
|
|
1604 had_eol = TRUE;
|
|
1605 #endif
|
|
1606 break;
|
|
1607 }
|
|
1608
|
|
1609 extra = ADD_NL;
|
|
1610 *flagp |= HASNL;
|
|
1611
|
|
1612 /* "\_[" is character range plus newline */
|
|
1613 if (c == '[')
|
|
1614 goto collection;
|
|
1615
|
|
1616 /* "\_x" is character class plus newline */
|
|
1617 /*FALLTHROUGH*/
|
|
1618
|
|
1619 /*
|
|
1620 * Character classes.
|
|
1621 */
|
|
1622 case Magic('.'):
|
|
1623 case Magic('i'):
|
|
1624 case Magic('I'):
|
|
1625 case Magic('k'):
|
|
1626 case Magic('K'):
|
|
1627 case Magic('f'):
|
|
1628 case Magic('F'):
|
|
1629 case Magic('p'):
|
|
1630 case Magic('P'):
|
|
1631 case Magic('s'):
|
|
1632 case Magic('S'):
|
|
1633 case Magic('d'):
|
|
1634 case Magic('D'):
|
|
1635 case Magic('x'):
|
|
1636 case Magic('X'):
|
|
1637 case Magic('o'):
|
|
1638 case Magic('O'):
|
|
1639 case Magic('w'):
|
|
1640 case Magic('W'):
|
|
1641 case Magic('h'):
|
|
1642 case Magic('H'):
|
|
1643 case Magic('a'):
|
|
1644 case Magic('A'):
|
|
1645 case Magic('l'):
|
|
1646 case Magic('L'):
|
|
1647 case Magic('u'):
|
|
1648 case Magic('U'):
|
|
1649 p = vim_strchr(classchars, no_Magic(c));
|
|
1650 if (p == NULL)
|
|
1651 EMSG_RET_NULL(_("E63: invalid use of \\_"));
|
|
1652 ret = regnode(classcodes[p - classchars] + extra);
|
|
1653 *flagp |= HASWIDTH | SIMPLE;
|
|
1654 break;
|
|
1655
|
|
1656 case Magic('n'):
|
|
1657 if (reg_string)
|
|
1658 {
|
|
1659 /* In a string "\n" matches a newline character. */
|
|
1660 ret = regnode(EXACTLY);
|
|
1661 regc(NL);
|
|
1662 regc(NUL);
|
|
1663 *flagp |= HASWIDTH | SIMPLE;
|
|
1664 }
|
|
1665 else
|
|
1666 {
|
|
1667 /* In buffer text "\n" matches the end of a line. */
|
|
1668 ret = regnode(NEWL);
|
|
1669 *flagp |= HASWIDTH | HASNL;
|
|
1670 }
|
|
1671 break;
|
|
1672
|
|
1673 case Magic('('):
|
|
1674 if (one_exactly)
|
|
1675 EMSG_ONE_RET_NULL;
|
|
1676 ret = reg(REG_PAREN, &flags);
|
|
1677 if (ret == NULL)
|
|
1678 return NULL;
|
|
1679 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1680 break;
|
|
1681
|
|
1682 case NUL:
|
|
1683 case Magic('|'):
|
|
1684 case Magic('&'):
|
|
1685 case Magic(')'):
|
|
1686 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */
|
|
1687 /* NOTREACHED */
|
|
1688
|
|
1689 case Magic('='):
|
|
1690 case Magic('?'):
|
|
1691 case Magic('+'):
|
|
1692 case Magic('@'):
|
|
1693 case Magic('{'):
|
|
1694 case Magic('*'):
|
|
1695 c = no_Magic(c);
|
|
1696 sprintf((char *)IObuff, _("E64: %s%c follows nothing"),
|
|
1697 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL)
|
|
1698 ? "" : "\\", c);
|
|
1699 EMSG_RET_NULL(IObuff);
|
|
1700 /* NOTREACHED */
|
|
1701
|
|
1702 case Magic('~'): /* previous substitute pattern */
|
|
1703 if (reg_prev_sub)
|
|
1704 {
|
|
1705 char_u *lp;
|
|
1706
|
|
1707 ret = regnode(EXACTLY);
|
|
1708 lp = reg_prev_sub;
|
|
1709 while (*lp != NUL)
|
|
1710 regc(*lp++);
|
|
1711 regc(NUL);
|
|
1712 if (*reg_prev_sub != NUL)
|
|
1713 {
|
|
1714 *flagp |= HASWIDTH;
|
|
1715 if ((lp - reg_prev_sub) == 1)
|
|
1716 *flagp |= SIMPLE;
|
|
1717 }
|
|
1718 }
|
|
1719 else
|
|
1720 EMSG_RET_NULL(_(e_nopresub));
|
|
1721 break;
|
|
1722
|
|
1723 case Magic('1'):
|
|
1724 case Magic('2'):
|
|
1725 case Magic('3'):
|
|
1726 case Magic('4'):
|
|
1727 case Magic('5'):
|
|
1728 case Magic('6'):
|
|
1729 case Magic('7'):
|
|
1730 case Magic('8'):
|
|
1731 case Magic('9'):
|
|
1732 {
|
|
1733 int refnum;
|
|
1734
|
|
1735 refnum = c - Magic('0');
|
|
1736 /*
|
|
1737 * Check if the back reference is legal. We must have seen the
|
|
1738 * close brace.
|
|
1739 * TODO: Should also check that we don't refer to something
|
|
1740 * that is repeated (+*=): what instance of the repetition
|
|
1741 * should we match?
|
|
1742 */
|
|
1743 if (!had_endbrace[refnum])
|
|
1744 {
|
|
1745 /* Trick: check if "@<=" or "@<!" follows, in which case
|
|
1746 * the \1 can appear before the referenced match. */
|
|
1747 for (p = regparse; *p != NUL; ++p)
|
|
1748 if (p[0] == '@' && p[1] == '<'
|
|
1749 && (p[2] == '!' || p[2] == '='))
|
|
1750 break;
|
|
1751 if (*p == NUL)
|
|
1752 EMSG_RET_NULL(_("E65: Illegal back reference"));
|
|
1753 }
|
|
1754 ret = regnode(BACKREF + refnum);
|
|
1755 }
|
|
1756 break;
|
|
1757
|
|
1758 #ifdef FEAT_SYN_HL
|
|
1759 case Magic('z'):
|
|
1760 {
|
|
1761 c = no_Magic(getchr());
|
|
1762 switch (c)
|
|
1763 {
|
|
1764 case '(': if (reg_do_extmatch != REX_SET)
|
|
1765 EMSG_RET_NULL(_("E66: \\z( not allowed here"));
|
|
1766 if (one_exactly)
|
|
1767 EMSG_ONE_RET_NULL;
|
|
1768 ret = reg(REG_ZPAREN, &flags);
|
|
1769 if (ret == NULL)
|
|
1770 return NULL;
|
|
1771 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH);
|
|
1772 re_has_z = REX_SET;
|
|
1773 break;
|
|
1774
|
|
1775 case '1':
|
|
1776 case '2':
|
|
1777 case '3':
|
|
1778 case '4':
|
|
1779 case '5':
|
|
1780 case '6':
|
|
1781 case '7':
|
|
1782 case '8':
|
|
1783 case '9': if (reg_do_extmatch != REX_USE)
|
|
1784 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here"));
|
|
1785 ret = regnode(ZREF + c - '0');
|
|
1786 re_has_z = REX_USE;
|
|
1787 break;
|
|
1788
|
|
1789 case 's': ret = regnode(MOPEN + 0);
|
|
1790 break;
|
|
1791
|
|
1792 case 'e': ret = regnode(MCLOSE + 0);
|
|
1793 break;
|
|
1794
|
|
1795 default: EMSG_RET_NULL(_("E68: Invalid character after \\z"));
|
|
1796 }
|
|
1797 }
|
|
1798 break;
|
|
1799 #endif
|
|
1800
|
|
1801 case Magic('%'):
|
|
1802 {
|
|
1803 c = no_Magic(getchr());
|
|
1804 switch (c)
|
|
1805 {
|
|
1806 /* () without a back reference */
|
|
1807 case '(':
|
|
1808 if (one_exactly)
|
|
1809 EMSG_ONE_RET_NULL;
|
|
1810 ret = reg(REG_NPAREN, &flags);
|
|
1811 if (ret == NULL)
|
|
1812 return NULL;
|
|
1813 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
|
|
1814 break;
|
|
1815
|
|
1816 /* Catch \%^ and \%$ regardless of where they appear in the
|
|
1817 * pattern -- regardless of whether or not it makes sense. */
|
|
1818 case '^':
|
|
1819 ret = regnode(RE_BOF);
|
|
1820 break;
|
|
1821
|
|
1822 case '$':
|
|
1823 ret = regnode(RE_EOF);
|
|
1824 break;
|
|
1825
|
|
1826 case '#':
|
|
1827 ret = regnode(CURSOR);
|
|
1828 break;
|
|
1829
|
|
1830 /* \%[abc]: Emit as a list of branches, all ending at the last
|
|
1831 * branch which matches nothing. */
|
|
1832 case '[':
|
|
1833 if (one_exactly) /* doesn't nest */
|
|
1834 EMSG_ONE_RET_NULL;
|
|
1835 {
|
|
1836 char_u *lastbranch;
|
|
1837 char_u *lastnode = NULL;
|
|
1838 char_u *br;
|
|
1839
|
|
1840 ret = NULL;
|
|
1841 while ((c = getchr()) != ']')
|
|
1842 {
|
|
1843 if (c == NUL)
|
|
1844 EMSG_M_RET_NULL(_("E69: Missing ] after %s%%["),
|
|
1845 reg_magic == MAGIC_ALL);
|
|
1846 br = regnode(BRANCH);
|
|
1847 if (ret == NULL)
|
|
1848 ret = br;
|
|
1849 else
|
|
1850 regtail(lastnode, br);
|
|
1851
|
|
1852 ungetchr();
|
|
1853 one_exactly = TRUE;
|
|
1854 lastnode = regatom(flagp);
|
|
1855 one_exactly = FALSE;
|
|
1856 if (lastnode == NULL)
|
|
1857 return NULL;
|
|
1858 }
|
|
1859 if (ret == NULL)
|
|
1860 EMSG_M_RET_NULL(_("E70: Empty %s%%[]"),
|
|
1861 reg_magic == MAGIC_ALL);
|
|
1862 lastbranch = regnode(BRANCH);
|
|
1863 br = regnode(NOTHING);
|
|
1864 if (ret != JUST_CALC_SIZE)
|
|
1865 {
|
|
1866 regtail(lastnode, br);
|
|
1867 regtail(lastbranch, br);
|
|
1868 /* connect all branches to the NOTHING
|
|
1869 * branch at the end */
|
|
1870 for (br = ret; br != lastnode; )
|
|
1871 {
|
|
1872 if (OP(br) == BRANCH)
|
|
1873 {
|
|
1874 regtail(br, lastbranch);
|
|
1875 br = OPERAND(br);
|
|
1876 }
|
|
1877 else
|
|
1878 br = regnext(br);
|
|
1879 }
|
|
1880 }
|
|
1881 *flagp &= ~HASWIDTH;
|
|
1882 break;
|
|
1883 }
|
|
1884
|
24
|
1885 case 'd': /* %d123 decimal */
|
|
1886 case 'o': /* %o123 octal */
|
|
1887 case 'x': /* %xab hex 2 */
|
|
1888 case 'u': /* %uabcd hex 4 */
|
|
1889 case 'U': /* %U1234abcd hex 8 */
|
|
1890 {
|
|
1891 int i;
|
|
1892
|
|
1893 switch (c)
|
|
1894 {
|
|
1895 case 'd': i = getdecchrs(); break;
|
|
1896 case 'o': i = getoctchrs(); break;
|
|
1897 case 'x': i = gethexchrs(2); break;
|
|
1898 case 'u': i = gethexchrs(4); break;
|
|
1899 case 'U': i = gethexchrs(8); break;
|
|
1900 default: i = -1; break;
|
|
1901 }
|
|
1902
|
|
1903 if (i < 0)
|
|
1904 EMSG_M_RET_NULL(
|
|
1905 _("E678: Invalid character after %s%%[dxouU]"),
|
|
1906 reg_magic == MAGIC_ALL);
|
|
1907 ret = regnode(EXACTLY);
|
|
1908 if (i == 0)
|
|
1909 regc(0x0a);
|
|
1910 else
|
|
1911 #ifdef FEAT_MBYTE
|
|
1912 regmbc(i);
|
|
1913 #else
|
|
1914 regc(i);
|
|
1915 #endif
|
|
1916 regc(NUL);
|
|
1917 *flagp |= HASWIDTH;
|
|
1918 break;
|
|
1919 }
|
|
1920
|
7
|
1921 default:
|
|
1922 if (VIM_ISDIGIT(c) || c == '<' || c == '>')
|
|
1923 {
|
|
1924 long_u n = 0;
|
|
1925 int cmp;
|
|
1926
|
|
1927 cmp = c;
|
|
1928 if (cmp == '<' || cmp == '>')
|
|
1929 c = getchr();
|
|
1930 while (VIM_ISDIGIT(c))
|
|
1931 {
|
|
1932 n = n * 10 + (c - '0');
|
|
1933 c = getchr();
|
|
1934 }
|
|
1935 if (c == 'l' || c == 'c' || c == 'v')
|
|
1936 {
|
|
1937 if (c == 'l')
|
|
1938 ret = regnode(RE_LNUM);
|
|
1939 else if (c == 'c')
|
|
1940 ret = regnode(RE_COL);
|
|
1941 else
|
|
1942 ret = regnode(RE_VCOL);
|
|
1943 if (ret == JUST_CALC_SIZE)
|
|
1944 regsize += 5;
|
|
1945 else
|
|
1946 {
|
|
1947 /* put the number and the optional
|
|
1948 * comparator after the opcode */
|
|
1949 regcode = re_put_long(regcode, n);
|
|
1950 *regcode++ = cmp;
|
|
1951 }
|
|
1952 break;
|
|
1953 }
|
|
1954 }
|
|
1955
|
|
1956 EMSG_M_RET_NULL(_("E71: Invalid character after %s%%"),
|
|
1957 reg_magic == MAGIC_ALL);
|
|
1958 }
|
|
1959 }
|
|
1960 break;
|
|
1961
|
|
1962 case Magic('['):
|
|
1963 collection:
|
|
1964 {
|
|
1965 char_u *lp;
|
|
1966
|
|
1967 /*
|
|
1968 * If there is no matching ']', we assume the '[' is a normal
|
|
1969 * character. This makes 'incsearch' and ":help [" work.
|
|
1970 */
|
|
1971 lp = skip_anyof(regparse);
|
|
1972 if (*lp == ']') /* there is a matching ']' */
|
|
1973 {
|
|
1974 int startc = -1; /* > 0 when next '-' is a range */
|
|
1975 int endc;
|
|
1976
|
|
1977 /*
|
|
1978 * In a character class, different parsing rules apply.
|
|
1979 * Not even \ is special anymore, nothing is.
|
|
1980 */
|
|
1981 if (*regparse == '^') /* Complement of range. */
|
|
1982 {
|
|
1983 ret = regnode(ANYBUT + extra);
|
|
1984 regparse++;
|
|
1985 }
|
|
1986 else
|
|
1987 ret = regnode(ANYOF + extra);
|
|
1988
|
|
1989 /* At the start ']' and '-' mean the literal character. */
|
|
1990 if (*regparse == ']' || *regparse == '-')
|
167
|
1991 {
|
|
1992 startc = *regparse;
|
7
|
1993 regc(*regparse++);
|
167
|
1994 }
|
7
|
1995
|
|
1996 while (*regparse != NUL && *regparse != ']')
|
|
1997 {
|
|
1998 if (*regparse == '-')
|
|
1999 {
|
|
2000 ++regparse;
|
|
2001 /* The '-' is not used for a range at the end and
|
|
2002 * after or before a '\n'. */
|
|
2003 if (*regparse == ']' || *regparse == NUL
|
|
2004 || startc == -1
|
|
2005 || (regparse[0] == '\\' && regparse[1] == 'n'))
|
|
2006 {
|
|
2007 regc('-');
|
|
2008 startc = '-'; /* [--x] is a range */
|
|
2009 }
|
|
2010 else
|
|
2011 {
|
167
|
2012 /* Also accept "a-[.z.]" */
|
|
2013 endc = 0;
|
|
2014 if (*regparse == '[')
|
|
2015 endc = get_coll_element(®parse);
|
|
2016 if (endc == 0)
|
|
2017 {
|
7
|
2018 #ifdef FEAT_MBYTE
|
167
|
2019 if (has_mbyte)
|
|
2020 endc = mb_ptr2char_adv(®parse);
|
|
2021 else
|
7
|
2022 #endif
|
167
|
2023 endc = *regparse++;
|
|
2024 }
|
24
|
2025
|
|
2026 /* Handle \o40, \x20 and \u20AC style sequences */
|
167
|
2027 if (endc == '\\' && !cpo_lit && !cpo_bsl)
|
24
|
2028 endc = coll_get_char();
|
|
2029
|
7
|
2030 if (startc > endc)
|
|
2031 EMSG_RET_NULL(_(e_invrange));
|
|
2032 #ifdef FEAT_MBYTE
|
|
2033 if (has_mbyte && ((*mb_char2len)(startc) > 1
|
|
2034 || (*mb_char2len)(endc) > 1))
|
|
2035 {
|
|
2036 /* Limit to a range of 256 chars */
|
|
2037 if (endc > startc + 256)
|
|
2038 EMSG_RET_NULL(_(e_invrange));
|
|
2039 while (++startc <= endc)
|
|
2040 regmbc(startc);
|
|
2041 }
|
|
2042 else
|
|
2043 #endif
|
|
2044 {
|
|
2045 #ifdef EBCDIC
|
|
2046 int alpha_only = FALSE;
|
|
2047
|
|
2048 /* for alphabetical range skip the gaps
|
|
2049 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */
|
|
2050 if (isalpha(startc) && isalpha(endc))
|
|
2051 alpha_only = TRUE;
|
|
2052 #endif
|
|
2053 while (++startc <= endc)
|
|
2054 #ifdef EBCDIC
|
|
2055 if (!alpha_only || isalpha(startc))
|
|
2056 #endif
|
|
2057 regc(startc);
|
|
2058 }
|
|
2059 startc = -1;
|
|
2060 }
|
|
2061 }
|
|
2062 /*
|
|
2063 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
|
|
2064 * accepts "\t", "\e", etc., but only when the 'l' flag in
|
|
2065 * 'cpoptions' is not included.
|
167
|
2066 * Posix doesn't recognize backslash at all.
|
7
|
2067 */
|
|
2068 else if (*regparse == '\\'
|
167
|
2069 && !cpo_bsl
|
7
|
2070 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
|
|
2071 || (!cpo_lit
|
|
2072 && vim_strchr(REGEXP_ABBR,
|
|
2073 regparse[1]) != NULL)))
|
|
2074 {
|
|
2075 regparse++;
|
|
2076 if (*regparse == 'n')
|
|
2077 {
|
|
2078 /* '\n' in range: also match NL */
|
|
2079 if (ret != JUST_CALC_SIZE)
|
|
2080 {
|
|
2081 if (*ret == ANYBUT)
|
|
2082 *ret = ANYBUT + ADD_NL;
|
|
2083 else if (*ret == ANYOF)
|
|
2084 *ret = ANYOF + ADD_NL;
|
|
2085 /* else: must have had a \n already */
|
|
2086 }
|
|
2087 *flagp |= HASNL;
|
|
2088 regparse++;
|
|
2089 startc = -1;
|
|
2090 }
|
24
|
2091 else if (*regparse == 'd'
|
|
2092 || *regparse == 'o'
|
|
2093 || *regparse == 'x'
|
|
2094 || *regparse == 'u'
|
|
2095 || *regparse == 'U')
|
|
2096 {
|
|
2097 startc = coll_get_char();
|
|
2098 if (startc == 0)
|
|
2099 regc(0x0a);
|
|
2100 else
|
|
2101 #ifdef FEAT_MBYTE
|
|
2102 regmbc(startc);
|
|
2103 #else
|
|
2104 regc(startc);
|
|
2105 #endif
|
|
2106 }
|
7
|
2107 else
|
|
2108 {
|
|
2109 startc = backslash_trans(*regparse++);
|
|
2110 regc(startc);
|
|
2111 }
|
|
2112 }
|
|
2113 else if (*regparse == '[')
|
|
2114 {
|
|
2115 int c_class;
|
|
2116 int cu;
|
|
2117
|
167
|
2118 c_class = get_char_class(®parse);
|
7
|
2119 startc = -1;
|
|
2120 /* Characters assumed to be 8 bits! */
|
|
2121 switch (c_class)
|
|
2122 {
|
|
2123 case CLASS_NONE:
|
167
|
2124 c_class = get_equi_class(®parse);
|
|
2125 if (c_class != 0)
|
|
2126 {
|
|
2127 /* produce equivalence class */
|
|
2128 reg_equi_class(c_class);
|
|
2129 }
|
|
2130 else if ((c_class =
|
|
2131 get_coll_element(®parse)) != 0)
|
|
2132 {
|
|
2133 /* produce a collating element */
|
|
2134 regmbc(c_class);
|
|
2135 }
|
|
2136 else
|
|
2137 {
|
|
2138 /* literal '[', allow [[-x] as a range */
|
|
2139 startc = *regparse++;
|
|
2140 regc(startc);
|
|
2141 }
|
7
|
2142 break;
|
|
2143 case CLASS_ALNUM:
|
|
2144 for (cu = 1; cu <= 255; cu++)
|
|
2145 if (isalnum(cu))
|
|
2146 regc(cu);
|
|
2147 break;
|
|
2148 case CLASS_ALPHA:
|
|
2149 for (cu = 1; cu <= 255; cu++)
|
|
2150 if (isalpha(cu))
|
|
2151 regc(cu);
|
|
2152 break;
|
|
2153 case CLASS_BLANK:
|
|
2154 regc(' ');
|
|
2155 regc('\t');
|
|
2156 break;
|
|
2157 case CLASS_CNTRL:
|
|
2158 for (cu = 1; cu <= 255; cu++)
|
|
2159 if (iscntrl(cu))
|
|
2160 regc(cu);
|
|
2161 break;
|
|
2162 case CLASS_DIGIT:
|
|
2163 for (cu = 1; cu <= 255; cu++)
|
|
2164 if (VIM_ISDIGIT(cu))
|
|
2165 regc(cu);
|
|
2166 break;
|
|
2167 case CLASS_GRAPH:
|
|
2168 for (cu = 1; cu <= 255; cu++)
|
|
2169 if (isgraph(cu))
|
|
2170 regc(cu);
|
|
2171 break;
|
|
2172 case CLASS_LOWER:
|
|
2173 for (cu = 1; cu <= 255; cu++)
|
|
2174 if (islower(cu))
|
|
2175 regc(cu);
|
|
2176 break;
|
|
2177 case CLASS_PRINT:
|
|
2178 for (cu = 1; cu <= 255; cu++)
|
|
2179 if (vim_isprintc(cu))
|
|
2180 regc(cu);
|
|
2181 break;
|
|
2182 case CLASS_PUNCT:
|
|
2183 for (cu = 1; cu <= 255; cu++)
|
|
2184 if (ispunct(cu))
|
|
2185 regc(cu);
|
|
2186 break;
|
|
2187 case CLASS_SPACE:
|
|
2188 for (cu = 9; cu <= 13; cu++)
|
|
2189 regc(cu);
|
|
2190 regc(' ');
|
|
2191 break;
|
|
2192 case CLASS_UPPER:
|
|
2193 for (cu = 1; cu <= 255; cu++)
|
|
2194 if (isupper(cu))
|
|
2195 regc(cu);
|
|
2196 break;
|
|
2197 case CLASS_XDIGIT:
|
|
2198 for (cu = 1; cu <= 255; cu++)
|
|
2199 if (vim_isxdigit(cu))
|
|
2200 regc(cu);
|
|
2201 break;
|
|
2202 case CLASS_TAB:
|
|
2203 regc('\t');
|
|
2204 break;
|
|
2205 case CLASS_RETURN:
|
|
2206 regc('\r');
|
|
2207 break;
|
|
2208 case CLASS_BACKSPACE:
|
|
2209 regc('\b');
|
|
2210 break;
|
|
2211 case CLASS_ESCAPE:
|
|
2212 regc('\033');
|
|
2213 break;
|
|
2214 }
|
|
2215 }
|
|
2216 else
|
|
2217 {
|
|
2218 #ifdef FEAT_MBYTE
|
|
2219 if (has_mbyte)
|
|
2220 {
|
|
2221 int len;
|
|
2222
|
|
2223 /* produce a multibyte character, including any
|
|
2224 * following composing characters */
|
|
2225 startc = mb_ptr2char(regparse);
|
|
2226 len = (*mb_ptr2len_check)(regparse);
|
|
2227 if (enc_utf8 && utf_char2len(startc) != len)
|
|
2228 startc = -1; /* composing chars */
|
|
2229 while (--len >= 0)
|
|
2230 regc(*regparse++);
|
|
2231 }
|
|
2232 else
|
|
2233 #endif
|
|
2234 {
|
|
2235 startc = *regparse++;
|
|
2236 regc(startc);
|
|
2237 }
|
|
2238 }
|
|
2239 }
|
|
2240 regc(NUL);
|
|
2241 prevchr_len = 1; /* last char was the ']' */
|
|
2242 if (*regparse != ']')
|
|
2243 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */
|
|
2244 skipchr(); /* let's be friends with the lexer again */
|
|
2245 *flagp |= HASWIDTH | SIMPLE;
|
|
2246 break;
|
|
2247 }
|
|
2248 }
|
|
2249 /* FALLTHROUGH */
|
|
2250
|
|
2251 default:
|
|
2252 {
|
|
2253 int len;
|
|
2254
|
|
2255 #ifdef FEAT_MBYTE
|
|
2256 /* A multi-byte character is handled as a separate atom if it's
|
|
2257 * before a multi. */
|
|
2258 if (has_mbyte && (*mb_char2len)(c) > 1
|
|
2259 && re_multi_type(peekchr()) != NOT_MULTI)
|
|
2260 {
|
|
2261 ret = regnode(MULTIBYTECODE);
|
|
2262 regmbc(c);
|
|
2263 *flagp |= HASWIDTH | SIMPLE;
|
|
2264 break;
|
|
2265 }
|
|
2266 #endif
|
|
2267
|
|
2268 ret = regnode(EXACTLY);
|
|
2269
|
|
2270 /*
|
|
2271 * Append characters as long as:
|
|
2272 * - there is no following multi, we then need the character in
|
|
2273 * front of it as a single character operand
|
|
2274 * - not running into a Magic character
|
|
2275 * - "one_exactly" is not set
|
|
2276 * But always emit at least one character. Might be a Multi,
|
|
2277 * e.g., a "[" without matching "]".
|
|
2278 */
|
|
2279 for (len = 0; c != NUL && (len == 0
|
|
2280 || (re_multi_type(peekchr()) == NOT_MULTI
|
|
2281 && !one_exactly
|
|
2282 && !is_Magic(c))); ++len)
|
|
2283 {
|
|
2284 c = no_Magic(c);
|
|
2285 #ifdef FEAT_MBYTE
|
|
2286 if (has_mbyte)
|
|
2287 {
|
|
2288 regmbc(c);
|
|
2289 if (enc_utf8)
|
|
2290 {
|
|
2291 int off;
|
|
2292 int l;
|
|
2293
|
|
2294 /* Need to get composing character too, directly
|
|
2295 * access regparse for that, because skipchr() skips
|
|
2296 * over composing chars. */
|
|
2297 ungetchr();
|
|
2298 if (*regparse == '\\' && regparse[1] != NUL)
|
|
2299 off = 1;
|
|
2300 else
|
|
2301 off = 0;
|
|
2302 for (;;)
|
|
2303 {
|
|
2304 l = utf_ptr2len_check(regparse + off);
|
|
2305 if (!UTF_COMPOSINGLIKE(regparse + off,
|
|
2306 regparse + off + l))
|
|
2307 break;
|
|
2308 off += l;
|
|
2309 regmbc(utf_ptr2char(regparse + off));
|
|
2310 }
|
|
2311 skipchr();
|
|
2312 }
|
|
2313 }
|
|
2314 else
|
|
2315 #endif
|
|
2316 regc(c);
|
|
2317 c = getchr();
|
|
2318 }
|
|
2319 ungetchr();
|
|
2320
|
|
2321 regc(NUL);
|
|
2322 *flagp |= HASWIDTH;
|
|
2323 if (len == 1)
|
|
2324 *flagp |= SIMPLE;
|
|
2325 }
|
|
2326 break;
|
|
2327 }
|
|
2328
|
|
2329 return ret;
|
|
2330 }
|
|
2331
|
|
2332 /*
|
|
2333 * emit a node
|
|
2334 * Return pointer to generated code.
|
|
2335 */
|
|
2336 static char_u *
|
|
2337 regnode(op)
|
|
2338 int op;
|
|
2339 {
|
|
2340 char_u *ret;
|
|
2341
|
|
2342 ret = regcode;
|
|
2343 if (ret == JUST_CALC_SIZE)
|
|
2344 regsize += 3;
|
|
2345 else
|
|
2346 {
|
|
2347 *regcode++ = op;
|
|
2348 *regcode++ = NUL; /* Null "next" pointer. */
|
|
2349 *regcode++ = NUL;
|
|
2350 }
|
|
2351 return ret;
|
|
2352 }
|
|
2353
|
|
2354 /*
|
|
2355 * Emit (if appropriate) a byte of code
|
|
2356 */
|
|
2357 static void
|
|
2358 regc(b)
|
|
2359 int b;
|
|
2360 {
|
|
2361 if (regcode == JUST_CALC_SIZE)
|
|
2362 regsize++;
|
|
2363 else
|
|
2364 *regcode++ = b;
|
|
2365 }
|
|
2366
|
|
2367 #ifdef FEAT_MBYTE
|
|
2368 /*
|
|
2369 * Emit (if appropriate) a multi-byte character of code
|
|
2370 */
|
|
2371 static void
|
|
2372 regmbc(c)
|
|
2373 int c;
|
|
2374 {
|
|
2375 if (regcode == JUST_CALC_SIZE)
|
|
2376 regsize += (*mb_char2len)(c);
|
|
2377 else
|
|
2378 regcode += (*mb_char2bytes)(c, regcode);
|
|
2379 }
|
|
2380 #endif
|
|
2381
|
|
2382 /*
|
|
2383 * reginsert - insert an operator in front of already-emitted operand
|
|
2384 *
|
|
2385 * Means relocating the operand.
|
|
2386 */
|
|
2387 static void
|
|
2388 reginsert(op, opnd)
|
|
2389 int op;
|
|
2390 char_u *opnd;
|
|
2391 {
|
|
2392 char_u *src;
|
|
2393 char_u *dst;
|
|
2394 char_u *place;
|
|
2395
|
|
2396 if (regcode == JUST_CALC_SIZE)
|
|
2397 {
|
|
2398 regsize += 3;
|
|
2399 return;
|
|
2400 }
|
|
2401 src = regcode;
|
|
2402 regcode += 3;
|
|
2403 dst = regcode;
|
|
2404 while (src > opnd)
|
|
2405 *--dst = *--src;
|
|
2406
|
|
2407 place = opnd; /* Op node, where operand used to be. */
|
|
2408 *place++ = op;
|
|
2409 *place++ = NUL;
|
|
2410 *place = NUL;
|
|
2411 }
|
|
2412
|
|
2413 /*
|
|
2414 * reginsert_limits - insert an operator in front of already-emitted operand.
|
|
2415 * The operator has the given limit values as operands. Also set next pointer.
|
|
2416 *
|
|
2417 * Means relocating the operand.
|
|
2418 */
|
|
2419 static void
|
|
2420 reginsert_limits(op, minval, maxval, opnd)
|
|
2421 int op;
|
|
2422 long minval;
|
|
2423 long maxval;
|
|
2424 char_u *opnd;
|
|
2425 {
|
|
2426 char_u *src;
|
|
2427 char_u *dst;
|
|
2428 char_u *place;
|
|
2429
|
|
2430 if (regcode == JUST_CALC_SIZE)
|
|
2431 {
|
|
2432 regsize += 11;
|
|
2433 return;
|
|
2434 }
|
|
2435 src = regcode;
|
|
2436 regcode += 11;
|
|
2437 dst = regcode;
|
|
2438 while (src > opnd)
|
|
2439 *--dst = *--src;
|
|
2440
|
|
2441 place = opnd; /* Op node, where operand used to be. */
|
|
2442 *place++ = op;
|
|
2443 *place++ = NUL;
|
|
2444 *place++ = NUL;
|
|
2445 place = re_put_long(place, (long_u)minval);
|
|
2446 place = re_put_long(place, (long_u)maxval);
|
|
2447 regtail(opnd, place);
|
|
2448 }
|
|
2449
|
|
2450 /*
|
|
2451 * Write a long as four bytes at "p" and return pointer to the next char.
|
|
2452 */
|
|
2453 static char_u *
|
|
2454 re_put_long(p, val)
|
|
2455 char_u *p;
|
|
2456 long_u val;
|
|
2457 {
|
|
2458 *p++ = (char_u) ((val >> 24) & 0377);
|
|
2459 *p++ = (char_u) ((val >> 16) & 0377);
|
|
2460 *p++ = (char_u) ((val >> 8) & 0377);
|
|
2461 *p++ = (char_u) (val & 0377);
|
|
2462 return p;
|
|
2463 }
|
|
2464
|
|
2465 /*
|
|
2466 * regtail - set the next-pointer at the end of a node chain
|
|
2467 */
|
|
2468 static void
|
|
2469 regtail(p, val)
|
|
2470 char_u *p;
|
|
2471 char_u *val;
|
|
2472 {
|
|
2473 char_u *scan;
|
|
2474 char_u *temp;
|
|
2475 int offset;
|
|
2476
|
|
2477 if (p == JUST_CALC_SIZE)
|
|
2478 return;
|
|
2479
|
|
2480 /* Find last node. */
|
|
2481 scan = p;
|
|
2482 for (;;)
|
|
2483 {
|
|
2484 temp = regnext(scan);
|
|
2485 if (temp == NULL)
|
|
2486 break;
|
|
2487 scan = temp;
|
|
2488 }
|
|
2489
|
179
|
2490 if (OP(scan) == BACK || OP(scan) == BACKP)
|
7
|
2491 offset = (int)(scan - val);
|
|
2492 else
|
|
2493 offset = (int)(val - scan);
|
|
2494 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377);
|
|
2495 *(scan + 2) = (char_u) (offset & 0377);
|
|
2496 }
|
|
2497
|
|
2498 /*
|
|
2499 * regoptail - regtail on item after a BRANCH; nop if none
|
|
2500 */
|
|
2501 static void
|
|
2502 regoptail(p, val)
|
|
2503 char_u *p;
|
|
2504 char_u *val;
|
|
2505 {
|
|
2506 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */
|
|
2507 if (p == NULL || p == JUST_CALC_SIZE
|
|
2508 || (OP(p) != BRANCH
|
|
2509 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9)))
|
|
2510 return;
|
|
2511 regtail(OPERAND(p), val);
|
|
2512 }
|
|
2513
|
|
2514 /*
|
|
2515 * getchr() - get the next character from the pattern. We know about
|
|
2516 * magic and such, so therefore we need a lexical analyzer.
|
|
2517 */
|
|
2518
|
|
2519 /* static int curchr; */
|
|
2520 static int prevprevchr;
|
|
2521 static int prevchr;
|
|
2522 static int nextchr; /* used for ungetchr() */
|
|
2523 /*
|
|
2524 * Note: prevchr is sometimes -1 when we are not at the start,
|
|
2525 * eg in /[ ^I]^ the pattern was never found even if it existed, because ^ was
|
|
2526 * taken to be magic -- webb
|
|
2527 */
|
|
2528 static int at_start; /* True when on the first character */
|
|
2529 static int prev_at_start; /* True when on the second character */
|
|
2530
|
|
2531 static void
|
|
2532 initchr(str)
|
|
2533 char_u *str;
|
|
2534 {
|
|
2535 regparse = str;
|
|
2536 prevchr_len = 0;
|
|
2537 curchr = prevprevchr = prevchr = nextchr = -1;
|
|
2538 at_start = TRUE;
|
|
2539 prev_at_start = FALSE;
|
|
2540 }
|
|
2541
|
|
2542 static int
|
|
2543 peekchr()
|
|
2544 {
|
167
|
2545 static int after_slash = FALSE;
|
|
2546
|
7
|
2547 if (curchr == -1)
|
|
2548 {
|
|
2549 switch (curchr = regparse[0])
|
|
2550 {
|
|
2551 case '.':
|
|
2552 case '[':
|
|
2553 case '~':
|
|
2554 /* magic when 'magic' is on */
|
|
2555 if (reg_magic >= MAGIC_ON)
|
|
2556 curchr = Magic(curchr);
|
|
2557 break;
|
|
2558 case '(':
|
|
2559 case ')':
|
|
2560 case '{':
|
|
2561 case '%':
|
|
2562 case '+':
|
|
2563 case '=':
|
|
2564 case '?':
|
|
2565 case '@':
|
|
2566 case '!':
|
|
2567 case '&':
|
|
2568 case '|':
|
|
2569 case '<':
|
|
2570 case '>':
|
|
2571 case '#': /* future ext. */
|
|
2572 case '"': /* future ext. */
|
|
2573 case '\'': /* future ext. */
|
|
2574 case ',': /* future ext. */
|
|
2575 case '-': /* future ext. */
|
|
2576 case ':': /* future ext. */
|
|
2577 case ';': /* future ext. */
|
|
2578 case '`': /* future ext. */
|
|
2579 case '/': /* Can't be used in / command */
|
|
2580 /* magic only after "\v" */
|
|
2581 if (reg_magic == MAGIC_ALL)
|
|
2582 curchr = Magic(curchr);
|
|
2583 break;
|
|
2584 case '*':
|
167
|
2585 /* * is not magic as the very first character, eg "?*ptr", when
|
|
2586 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But
|
|
2587 * "\(\*" is not magic, thus must be magic if "after_slash" */
|
|
2588 if (reg_magic >= MAGIC_ON
|
|
2589 && !at_start
|
|
2590 && !(prev_at_start && prevchr == Magic('^'))
|
|
2591 && (after_slash
|
|
2592 || (prevchr != Magic('(')
|
|
2593 && prevchr != Magic('&')
|
|
2594 && prevchr != Magic('|'))))
|
7
|
2595 curchr = Magic('*');
|
|
2596 break;
|
|
2597 case '^':
|
|
2598 /* '^' is only magic as the very first character and if it's after
|
|
2599 * "\(", "\|", "\&' or "\n" */
|
|
2600 if (reg_magic >= MAGIC_OFF
|
|
2601 && (at_start
|
|
2602 || reg_magic == MAGIC_ALL
|
|
2603 || prevchr == Magic('(')
|
|
2604 || prevchr == Magic('|')
|
|
2605 || prevchr == Magic('&')
|
|
2606 || prevchr == Magic('n')
|
|
2607 || (no_Magic(prevchr) == '('
|
|
2608 && prevprevchr == Magic('%'))))
|
|
2609 {
|
|
2610 curchr = Magic('^');
|
|
2611 at_start = TRUE;
|
|
2612 prev_at_start = FALSE;
|
|
2613 }
|
|
2614 break;
|
|
2615 case '$':
|
|
2616 /* '$' is only magic as the very last char and if it's in front of
|
|
2617 * either "\|", "\)", "\&", or "\n" */
|
|
2618 if (reg_magic >= MAGIC_OFF)
|
|
2619 {
|
|
2620 char_u *p = regparse + 1;
|
|
2621
|
|
2622 /* ignore \c \C \m and \M after '$' */
|
|
2623 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C'
|
|
2624 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z'))
|
|
2625 p += 2;
|
|
2626 if (p[0] == NUL
|
|
2627 || (p[0] == '\\'
|
|
2628 && (p[1] == '|' || p[1] == '&' || p[1] == ')'
|
|
2629 || p[1] == 'n'))
|
|
2630 || reg_magic == MAGIC_ALL)
|
|
2631 curchr = Magic('$');
|
|
2632 }
|
|
2633 break;
|
|
2634 case '\\':
|
|
2635 {
|
|
2636 int c = regparse[1];
|
|
2637
|
|
2638 if (c == NUL)
|
|
2639 curchr = '\\'; /* trailing '\' */
|
|
2640 else if (
|
|
2641 #ifdef EBCDIC
|
|
2642 vim_strchr(META, c)
|
|
2643 #else
|
|
2644 c <= '~' && META_flags[c]
|
|
2645 #endif
|
|
2646 )
|
|
2647 {
|
|
2648 /*
|
|
2649 * META contains everything that may be magic sometimes,
|
|
2650 * except ^ and $ ("\^" and "\$" are only magic after
|
|
2651 * "\v"). We now fetch the next character and toggle its
|
|
2652 * magicness. Therefore, \ is so meta-magic that it is
|
|
2653 * not in META.
|
|
2654 */
|
|
2655 curchr = -1;
|
|
2656 prev_at_start = at_start;
|
|
2657 at_start = FALSE; /* be able to say "/\*ptr" */
|
|
2658 ++regparse;
|
167
|
2659 ++after_slash;
|
7
|
2660 peekchr();
|
|
2661 --regparse;
|
167
|
2662 --after_slash;
|
7
|
2663 curchr = toggle_Magic(curchr);
|
|
2664 }
|
|
2665 else if (vim_strchr(REGEXP_ABBR, c))
|
|
2666 {
|
|
2667 /*
|
|
2668 * Handle abbreviations, like "\t" for TAB -- webb
|
|
2669 */
|
|
2670 curchr = backslash_trans(c);
|
|
2671 }
|
|
2672 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^'))
|
|
2673 curchr = toggle_Magic(c);
|
|
2674 else
|
|
2675 {
|
|
2676 /*
|
|
2677 * Next character can never be (made) magic?
|
|
2678 * Then backslashing it won't do anything.
|
|
2679 */
|
|
2680 #ifdef FEAT_MBYTE
|
|
2681 if (has_mbyte)
|
|
2682 curchr = (*mb_ptr2char)(regparse + 1);
|
|
2683 else
|
|
2684 #endif
|
|
2685 curchr = c;
|
|
2686 }
|
|
2687 break;
|
|
2688 }
|
|
2689
|
|
2690 #ifdef FEAT_MBYTE
|
|
2691 default:
|
|
2692 if (has_mbyte)
|
|
2693 curchr = (*mb_ptr2char)(regparse);
|
|
2694 #endif
|
|
2695 }
|
|
2696 }
|
|
2697
|
|
2698 return curchr;
|
|
2699 }
|
|
2700
|
|
2701 /*
|
|
2702 * Eat one lexed character. Do this in a way that we can undo it.
|
|
2703 */
|
|
2704 static void
|
|
2705 skipchr()
|
|
2706 {
|
|
2707 /* peekchr() eats a backslash, do the same here */
|
|
2708 if (*regparse == '\\')
|
|
2709 prevchr_len = 1;
|
|
2710 else
|
|
2711 prevchr_len = 0;
|
|
2712 if (regparse[prevchr_len] != NUL)
|
|
2713 {
|
|
2714 #ifdef FEAT_MBYTE
|
|
2715 if (has_mbyte)
|
|
2716 prevchr_len += (*mb_ptr2len_check)(regparse + prevchr_len);
|
|
2717 else
|
|
2718 #endif
|
|
2719 ++prevchr_len;
|
|
2720 }
|
|
2721 regparse += prevchr_len;
|
|
2722 prev_at_start = at_start;
|
|
2723 at_start = FALSE;
|
|
2724 prevprevchr = prevchr;
|
|
2725 prevchr = curchr;
|
|
2726 curchr = nextchr; /* use previously unget char, or -1 */
|
|
2727 nextchr = -1;
|
|
2728 }
|
|
2729
|
|
2730 /*
|
|
2731 * Skip a character while keeping the value of prev_at_start for at_start.
|
|
2732 * prevchr and prevprevchr are also kept.
|
|
2733 */
|
|
2734 static void
|
|
2735 skipchr_keepstart()
|
|
2736 {
|
|
2737 int as = prev_at_start;
|
|
2738 int pr = prevchr;
|
|
2739 int prpr = prevprevchr;
|
|
2740
|
|
2741 skipchr();
|
|
2742 at_start = as;
|
|
2743 prevchr = pr;
|
|
2744 prevprevchr = prpr;
|
|
2745 }
|
|
2746
|
|
2747 static int
|
|
2748 getchr()
|
|
2749 {
|
|
2750 int chr = peekchr();
|
|
2751
|
|
2752 skipchr();
|
|
2753 return chr;
|
|
2754 }
|
|
2755
|
|
2756 /*
|
|
2757 * put character back. Works only once!
|
|
2758 */
|
|
2759 static void
|
|
2760 ungetchr()
|
|
2761 {
|
|
2762 nextchr = curchr;
|
|
2763 curchr = prevchr;
|
|
2764 prevchr = prevprevchr;
|
|
2765 at_start = prev_at_start;
|
|
2766 prev_at_start = FALSE;
|
|
2767
|
|
2768 /* Backup regparse, so that it's at the same position as before the
|
|
2769 * getchr(). */
|
|
2770 regparse -= prevchr_len;
|
|
2771 }
|
|
2772
|
|
2773 /*
|
29
|
2774 * Get and return the value of the hex string at the current position.
|
|
2775 * Return -1 if there is no valid hex number.
|
|
2776 * The position is updated:
|
24
|
2777 * blahblah\%x20asdf
|
|
2778 * before-^ ^-after
|
|
2779 * The parameter controls the maximum number of input characters. This will be
|
|
2780 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence.
|
|
2781 */
|
|
2782 static int
|
|
2783 gethexchrs(maxinputlen)
|
|
2784 int maxinputlen;
|
|
2785 {
|
|
2786 int nr = 0;
|
|
2787 int c;
|
|
2788 int i;
|
|
2789
|
|
2790 for (i = 0; i < maxinputlen; ++i)
|
|
2791 {
|
|
2792 c = regparse[0];
|
|
2793 if (!vim_isxdigit(c))
|
|
2794 break;
|
|
2795 nr <<= 4;
|
|
2796 nr |= hex2nr(c);
|
|
2797 ++regparse;
|
|
2798 }
|
|
2799
|
|
2800 if (i == 0)
|
|
2801 return -1;
|
|
2802 return nr;
|
|
2803 }
|
|
2804
|
|
2805 /*
|
|
2806 * get and return the value of the decimal string immediately after the
|
|
2807 * current position. Return -1 for invalid. Consumes all digits.
|
|
2808 */
|
|
2809 static int
|
|
2810 getdecchrs()
|
|
2811 {
|
|
2812 int nr = 0;
|
|
2813 int c;
|
|
2814 int i;
|
|
2815
|
|
2816 for (i = 0; ; ++i)
|
|
2817 {
|
|
2818 c = regparse[0];
|
|
2819 if (c < '0' || c > '9')
|
|
2820 break;
|
|
2821 nr *= 10;
|
|
2822 nr += c - '0';
|
|
2823 ++regparse;
|
|
2824 }
|
|
2825
|
|
2826 if (i == 0)
|
|
2827 return -1;
|
|
2828 return nr;
|
|
2829 }
|
|
2830
|
|
2831 /*
|
|
2832 * get and return the value of the octal string immediately after the current
|
|
2833 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle
|
|
2834 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't
|
|
2835 * treat 8 or 9 as recognised characters. Position is updated:
|
|
2836 * blahblah\%o210asdf
|
|
2837 * before-^ ^-after
|
|
2838 */
|
|
2839 static int
|
|
2840 getoctchrs()
|
|
2841 {
|
|
2842 int nr = 0;
|
|
2843 int c;
|
|
2844 int i;
|
|
2845
|
|
2846 for (i = 0; i < 3 && nr < 040; ++i)
|
|
2847 {
|
|
2848 c = regparse[0];
|
|
2849 if (c < '0' || c > '7')
|
|
2850 break;
|
|
2851 nr <<= 3;
|
|
2852 nr |= hex2nr(c);
|
|
2853 ++regparse;
|
|
2854 }
|
|
2855
|
|
2856 if (i == 0)
|
|
2857 return -1;
|
|
2858 return nr;
|
|
2859 }
|
|
2860
|
|
2861 /*
|
|
2862 * Get a number after a backslash that is inside [].
|
|
2863 * When nothing is recognized return a backslash.
|
|
2864 */
|
|
2865 static int
|
|
2866 coll_get_char()
|
|
2867 {
|
|
2868 int nr = -1;
|
|
2869
|
|
2870 switch (*regparse++)
|
|
2871 {
|
|
2872 case 'd': nr = getdecchrs(); break;
|
|
2873 case 'o': nr = getoctchrs(); break;
|
|
2874 case 'x': nr = gethexchrs(2); break;
|
|
2875 case 'u': nr = gethexchrs(4); break;
|
|
2876 case 'U': nr = gethexchrs(8); break;
|
|
2877 }
|
|
2878 if (nr < 0)
|
|
2879 {
|
|
2880 /* If getting the number fails be backwards compatible: the character
|
|
2881 * is a backslash. */
|
|
2882 --regparse;
|
|
2883 nr = '\\';
|
|
2884 }
|
|
2885 return nr;
|
|
2886 }
|
|
2887
|
|
2888 /*
|
7
|
2889 * read_limits - Read two integers to be taken as a minimum and maximum.
|
|
2890 * If the first character is '-', then the range is reversed.
|
|
2891 * Should end with 'end'. If minval is missing, zero is default, if maxval is
|
|
2892 * missing, a very big number is the default.
|
|
2893 */
|
|
2894 static int
|
|
2895 read_limits(minval, maxval)
|
|
2896 long *minval;
|
|
2897 long *maxval;
|
|
2898 {
|
|
2899 int reverse = FALSE;
|
|
2900 char_u *first_char;
|
|
2901 long tmp;
|
|
2902
|
|
2903 if (*regparse == '-')
|
|
2904 {
|
|
2905 /* Starts with '-', so reverse the range later */
|
|
2906 regparse++;
|
|
2907 reverse = TRUE;
|
|
2908 }
|
|
2909 first_char = regparse;
|
|
2910 *minval = getdigits(®parse);
|
|
2911 if (*regparse == ',') /* There is a comma */
|
|
2912 {
|
|
2913 if (vim_isdigit(*++regparse))
|
|
2914 *maxval = getdigits(®parse);
|
|
2915 else
|
|
2916 *maxval = MAX_LIMIT;
|
|
2917 }
|
|
2918 else if (VIM_ISDIGIT(*first_char))
|
|
2919 *maxval = *minval; /* It was \{n} or \{-n} */
|
|
2920 else
|
|
2921 *maxval = MAX_LIMIT; /* It was \{} or \{-} */
|
|
2922 if (*regparse == '\\')
|
|
2923 regparse++; /* Allow either \{...} or \{...\} */
|
167
|
2924 if (*regparse != '}')
|
7
|
2925 {
|
|
2926 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"),
|
|
2927 reg_magic == MAGIC_ALL ? "" : "\\");
|
|
2928 EMSG_RET_FAIL(IObuff);
|
|
2929 }
|
|
2930
|
|
2931 /*
|
|
2932 * Reverse the range if there was a '-', or make sure it is in the right
|
|
2933 * order otherwise.
|
|
2934 */
|
|
2935 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval))
|
|
2936 {
|
|
2937 tmp = *minval;
|
|
2938 *minval = *maxval;
|
|
2939 *maxval = tmp;
|
|
2940 }
|
|
2941 skipchr(); /* let's be friends with the lexer again */
|
|
2942 return OK;
|
|
2943 }
|
|
2944
|
|
2945 /*
|
|
2946 * vim_regexec and friends
|
|
2947 */
|
|
2948
|
|
2949 /*
|
|
2950 * Global work variables for vim_regexec().
|
|
2951 */
|
|
2952
|
|
2953 /* The current match-position is remembered with these variables: */
|
|
2954 static linenr_T reglnum; /* line number, relative to first line */
|
|
2955 static char_u *regline; /* start of current line */
|
|
2956 static char_u *reginput; /* current input, points into "regline" */
|
|
2957
|
|
2958 static int need_clear_subexpr; /* subexpressions still need to be
|
|
2959 * cleared */
|
|
2960 #ifdef FEAT_SYN_HL
|
|
2961 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions
|
|
2962 * still need to be cleared */
|
|
2963 #endif
|
|
2964
|
|
2965 /*
|
|
2966 * Structure used to save the current input state, when it needs to be
|
|
2967 * restored after trying a match. Used by reg_save() and reg_restore().
|
|
2968 */
|
|
2969 typedef struct
|
|
2970 {
|
|
2971 union
|
|
2972 {
|
|
2973 char_u *ptr; /* reginput pointer, for single-line regexp */
|
|
2974 lpos_T pos; /* reginput pos, for multi-line regexp */
|
|
2975 } rs_u;
|
|
2976 } regsave_T;
|
|
2977
|
|
2978 /* struct to save start/end pointer/position in for \(\) */
|
|
2979 typedef struct
|
|
2980 {
|
|
2981 union
|
|
2982 {
|
|
2983 char_u *ptr;
|
|
2984 lpos_T pos;
|
|
2985 } se_u;
|
|
2986 } save_se_T;
|
|
2987
|
|
2988 static char_u *reg_getline __ARGS((linenr_T lnum));
|
|
2989 static long vim_regexec_both __ARGS((char_u *line, colnr_T col));
|
|
2990 static long regtry __ARGS((regprog_T *prog, colnr_T col));
|
|
2991 static void cleanup_subexpr __ARGS((void));
|
|
2992 #ifdef FEAT_SYN_HL
|
|
2993 static void cleanup_zsubexpr __ARGS((void));
|
|
2994 #endif
|
|
2995 static void reg_nextline __ARGS((void));
|
|
2996 static void reg_save __ARGS((regsave_T *save));
|
|
2997 static void reg_restore __ARGS((regsave_T *save));
|
|
2998 static int reg_save_equal __ARGS((regsave_T *save));
|
|
2999 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp));
|
|
3000 static void save_se_one __ARGS((save_se_T *savep, char_u **pp));
|
|
3001
|
|
3002 /* Save the sub-expressions before attempting a match. */
|
|
3003 #define save_se(savep, posp, pp) \
|
|
3004 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
|
|
3005
|
|
3006 /* After a failed match restore the sub-expressions. */
|
|
3007 #define restore_se(savep, posp, pp) { \
|
|
3008 if (REG_MULTI) \
|
|
3009 *(posp) = (savep)->se_u.pos; \
|
|
3010 else \
|
|
3011 *(pp) = (savep)->se_u.ptr; }
|
|
3012
|
|
3013 static int re_num_cmp __ARGS((long_u val, char_u *scan));
|
180
|
3014 static int regmatch __ARGS((char_u *prog));
|
7
|
3015 static int regrepeat __ARGS((char_u *p, long maxcount));
|
|
3016
|
|
3017 #ifdef DEBUG
|
|
3018 int regnarrate = 0;
|
|
3019 #endif
|
|
3020
|
|
3021 /*
|
|
3022 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec().
|
|
3023 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern
|
|
3024 * contains '\c' or '\C' the value is overruled.
|
|
3025 */
|
|
3026 static int ireg_ic;
|
|
3027
|
|
3028 #ifdef FEAT_MBYTE
|
|
3029 /*
|
|
3030 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag
|
|
3031 * in the regexp. Defaults to false, always.
|
|
3032 */
|
|
3033 static int ireg_icombine;
|
|
3034 #endif
|
|
3035
|
|
3036 /*
|
|
3037 * Sometimes need to save a copy of a line. Since alloc()/free() is very
|
|
3038 * slow, we keep one allocated piece of memory and only re-allocate it when
|
|
3039 * it's too small. It's freed in vim_regexec_both() when finished.
|
|
3040 */
|
|
3041 static char_u *reg_tofree;
|
|
3042 static unsigned reg_tofreelen;
|
|
3043
|
|
3044 /*
|
|
3045 * These variables are set when executing a regexp to speed up the execution.
|
|
3046 * Which ones are set depends on whethere a single-line or multi-line match is
|
|
3047 * done:
|
|
3048 * single-line multi-line
|
|
3049 * reg_match ®match_T NULL
|
|
3050 * reg_mmatch NULL ®mmatch_T
|
|
3051 * reg_startp reg_match->startp <invalid>
|
|
3052 * reg_endp reg_match->endp <invalid>
|
|
3053 * reg_startpos <invalid> reg_mmatch->startpos
|
|
3054 * reg_endpos <invalid> reg_mmatch->endpos
|
|
3055 * reg_win NULL window in which to search
|
|
3056 * reg_buf <invalid> buffer in which to search
|
|
3057 * reg_firstlnum <invalid> first line in which to search
|
|
3058 * reg_maxline 0 last line nr
|
|
3059 * reg_line_lbr FALSE or TRUE FALSE
|
|
3060 */
|
|
3061 static regmatch_T *reg_match;
|
|
3062 static regmmatch_T *reg_mmatch;
|
|
3063 static char_u **reg_startp = NULL;
|
|
3064 static char_u **reg_endp = NULL;
|
|
3065 static lpos_T *reg_startpos = NULL;
|
|
3066 static lpos_T *reg_endpos = NULL;
|
|
3067 static win_T *reg_win;
|
|
3068 static buf_T *reg_buf;
|
|
3069 static linenr_T reg_firstlnum;
|
|
3070 static linenr_T reg_maxline;
|
|
3071 static int reg_line_lbr; /* "\n" in string is line break */
|
|
3072
|
|
3073 /*
|
|
3074 * Get pointer to the line "lnum", which is relative to "reg_firstlnum".
|
|
3075 */
|
|
3076 static char_u *
|
|
3077 reg_getline(lnum)
|
|
3078 linenr_T lnum;
|
|
3079 {
|
|
3080 /* when looking behind for a match/no-match lnum is negative. But we
|
|
3081 * can't go before line 1 */
|
|
3082 if (reg_firstlnum + lnum < 1)
|
|
3083 return NULL;
|
|
3084 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE);
|
|
3085 }
|
|
3086
|
|
3087 static regsave_T behind_pos;
|
|
3088
|
|
3089 #ifdef FEAT_SYN_HL
|
|
3090 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */
|
|
3091 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */
|
|
3092 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */
|
|
3093 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */
|
|
3094 #endif
|
|
3095
|
|
3096 /* TRUE if using multi-line regexp. */
|
|
3097 #define REG_MULTI (reg_match == NULL)
|
|
3098
|
|
3099 /*
|
|
3100 * Match a regexp against a string.
|
|
3101 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
3102 * Uses curbuf for line count and 'iskeyword'.
|
|
3103 *
|
|
3104 * Return TRUE if there is a match, FALSE if not.
|
|
3105 */
|
|
3106 int
|
|
3107 vim_regexec(rmp, line, col)
|
|
3108 regmatch_T *rmp;
|
|
3109 char_u *line; /* string to match against */
|
|
3110 colnr_T col; /* column to start looking for match */
|
|
3111 {
|
|
3112 reg_match = rmp;
|
|
3113 reg_mmatch = NULL;
|
|
3114 reg_maxline = 0;
|
|
3115 reg_line_lbr = FALSE;
|
|
3116 reg_win = NULL;
|
|
3117 ireg_ic = rmp->rm_ic;
|
|
3118 #ifdef FEAT_MBYTE
|
|
3119 ireg_icombine = FALSE;
|
|
3120 #endif
|
|
3121 return (vim_regexec_both(line, col) != 0);
|
|
3122 }
|
|
3123
|
39
|
3124 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \
|
|
3125 || defined(FIND_REPLACE_DIALOG) || defined(PROTO)
|
7
|
3126 /*
|
|
3127 * Like vim_regexec(), but consider a "\n" in "line" to be a line break.
|
|
3128 */
|
|
3129 int
|
|
3130 vim_regexec_nl(rmp, line, col)
|
|
3131 regmatch_T *rmp;
|
|
3132 char_u *line; /* string to match against */
|
|
3133 colnr_T col; /* column to start looking for match */
|
|
3134 {
|
|
3135 reg_match = rmp;
|
|
3136 reg_mmatch = NULL;
|
|
3137 reg_maxline = 0;
|
|
3138 reg_line_lbr = TRUE;
|
|
3139 reg_win = NULL;
|
|
3140 ireg_ic = rmp->rm_ic;
|
|
3141 #ifdef FEAT_MBYTE
|
|
3142 ireg_icombine = FALSE;
|
|
3143 #endif
|
|
3144 return (vim_regexec_both(line, col) != 0);
|
|
3145 }
|
|
3146 #endif
|
|
3147
|
|
3148 /*
|
|
3149 * Match a regexp against multiple lines.
|
|
3150 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
|
|
3151 * Uses curbuf for line count and 'iskeyword'.
|
|
3152 *
|
|
3153 * Return zero if there is no match. Return number of lines contained in the
|
|
3154 * match otherwise.
|
|
3155 */
|
|
3156 long
|
|
3157 vim_regexec_multi(rmp, win, buf, lnum, col)
|
|
3158 regmmatch_T *rmp;
|
|
3159 win_T *win; /* window in which to search or NULL */
|
|
3160 buf_T *buf; /* buffer in which to search */
|
|
3161 linenr_T lnum; /* nr of line to start looking for match */
|
|
3162 colnr_T col; /* column to start looking for match */
|
|
3163 {
|
|
3164 long r;
|
|
3165 buf_T *save_curbuf = curbuf;
|
|
3166
|
|
3167 reg_match = NULL;
|
|
3168 reg_mmatch = rmp;
|
|
3169 reg_buf = buf;
|
|
3170 reg_win = win;
|
|
3171 reg_firstlnum = lnum;
|
|
3172 reg_maxline = reg_buf->b_ml.ml_line_count - lnum;
|
|
3173 reg_line_lbr = FALSE;
|
|
3174 ireg_ic = rmp->rmm_ic;
|
|
3175 #ifdef FEAT_MBYTE
|
|
3176 ireg_icombine = FALSE;
|
|
3177 #endif
|
|
3178
|
|
3179 /* Need to switch to buffer "buf" to make vim_iswordc() work. */
|
|
3180 curbuf = buf;
|
|
3181 r = vim_regexec_both(NULL, col);
|
|
3182 curbuf = save_curbuf;
|
|
3183
|
|
3184 return r;
|
|
3185 }
|
|
3186
|
|
3187 /*
|
|
3188 * Match a regexp against a string ("line" points to the string) or multiple
|
|
3189 * lines ("line" is NULL, use reg_getline()).
|
|
3190 */
|
|
3191 static long
|
|
3192 vim_regexec_both(line, col)
|
|
3193 char_u *line;
|
|
3194 colnr_T col; /* column to start looking for match */
|
|
3195 {
|
|
3196 regprog_T *prog;
|
|
3197 char_u *s;
|
189
|
3198 long retval = 0L;
|
7
|
3199
|
|
3200 reg_tofree = NULL;
|
|
3201
|
|
3202 if (REG_MULTI)
|
|
3203 {
|
|
3204 prog = reg_mmatch->regprog;
|
|
3205 line = reg_getline((linenr_T)0);
|
|
3206 reg_startpos = reg_mmatch->startpos;
|
|
3207 reg_endpos = reg_mmatch->endpos;
|
|
3208 }
|
|
3209 else
|
|
3210 {
|
|
3211 prog = reg_match->regprog;
|
|
3212 reg_startp = reg_match->startp;
|
|
3213 reg_endp = reg_match->endp;
|
|
3214 }
|
|
3215
|
|
3216 /* Be paranoid... */
|
|
3217 if (prog == NULL || line == NULL)
|
|
3218 {
|
|
3219 EMSG(_(e_null));
|
|
3220 goto theend;
|
|
3221 }
|
|
3222
|
|
3223 /* Check validity of program. */
|
|
3224 if (prog_magic_wrong())
|
|
3225 goto theend;
|
|
3226
|
|
3227 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */
|
|
3228 if (prog->regflags & RF_ICASE)
|
|
3229 ireg_ic = TRUE;
|
|
3230 else if (prog->regflags & RF_NOICASE)
|
|
3231 ireg_ic = FALSE;
|
|
3232
|
|
3233 #ifdef FEAT_MBYTE
|
|
3234 /* If pattern contains "\Z" overrule value of ireg_icombine */
|
|
3235 if (prog->regflags & RF_ICOMBINE)
|
|
3236 ireg_icombine = TRUE;
|
|
3237 #endif
|
|
3238
|
|
3239 /* If there is a "must appear" string, look for it. */
|
|
3240 if (prog->regmust != NULL)
|
|
3241 {
|
|
3242 int c;
|
|
3243
|
|
3244 #ifdef FEAT_MBYTE
|
|
3245 if (has_mbyte)
|
|
3246 c = (*mb_ptr2char)(prog->regmust);
|
|
3247 else
|
|
3248 #endif
|
|
3249 c = *prog->regmust;
|
|
3250 s = line + col;
|
170
|
3251
|
|
3252 /*
|
|
3253 * This is used very often, esp. for ":global". Use three versions of
|
|
3254 * the loop to avoid overhead of conditions.
|
|
3255 */
|
|
3256 if (!ireg_ic
|
|
3257 #ifdef FEAT_MBYTE
|
|
3258 && !has_mbyte
|
|
3259 #endif
|
|
3260 )
|
|
3261 while ((s = vim_strbyte(s, c)) != NULL)
|
|
3262 {
|
|
3263 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3264 break; /* Found it. */
|
|
3265 ++s;
|
|
3266 }
|
|
3267 #ifdef FEAT_MBYTE
|
|
3268 else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1))
|
|
3269 while ((s = vim_strchr(s, c)) != NULL)
|
|
3270 {
|
|
3271 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3272 break; /* Found it. */
|
|
3273 mb_ptr_adv(s);
|
|
3274 }
|
|
3275 #endif
|
|
3276 else
|
|
3277 while ((s = cstrchr(s, c)) != NULL)
|
|
3278 {
|
|
3279 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
|
|
3280 break; /* Found it. */
|
|
3281 mb_ptr_adv(s);
|
|
3282 }
|
7
|
3283 if (s == NULL) /* Not present. */
|
|
3284 goto theend;
|
|
3285 }
|
|
3286
|
|
3287 regline = line;
|
|
3288 reglnum = 0;
|
|
3289
|
|
3290 /* Simplest case: Anchored match need be tried only once. */
|
|
3291 if (prog->reganch)
|
|
3292 {
|
|
3293 int c;
|
|
3294
|
|
3295 #ifdef FEAT_MBYTE
|
|
3296 if (has_mbyte)
|
|
3297 c = (*mb_ptr2char)(regline + col);
|
|
3298 else
|
|
3299 #endif
|
|
3300 c = regline[col];
|
|
3301 if (prog->regstart == NUL
|
|
3302 || prog->regstart == c
|
|
3303 || (ireg_ic && ((
|
|
3304 #ifdef FEAT_MBYTE
|
|
3305 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
|
|
3306 || (c < 255 && prog->regstart < 255 &&
|
|
3307 #endif
|
|
3308 TOLOWER_LOC(prog->regstart) == TOLOWER_LOC(c)))))
|
|
3309 retval = regtry(prog, col);
|
|
3310 else
|
|
3311 retval = 0;
|
|
3312 }
|
|
3313 else
|
|
3314 {
|
|
3315 /* Messy cases: unanchored match. */
|
180
|
3316 while (!got_int)
|
7
|
3317 {
|
|
3318 if (prog->regstart != NUL)
|
|
3319 {
|
170
|
3320 /* Skip until the char we know it must start with.
|
|
3321 * Used often, do some work to avoid call overhead. */
|
|
3322 if (!ireg_ic
|
|
3323 #ifdef FEAT_MBYTE
|
|
3324 && !has_mbyte
|
|
3325 #endif
|
|
3326 )
|
|
3327 s = vim_strbyte(regline + col, prog->regstart);
|
|
3328 else
|
|
3329 s = cstrchr(regline + col, prog->regstart);
|
7
|
3330 if (s == NULL)
|
|
3331 {
|
|
3332 retval = 0;
|
|
3333 break;
|
|
3334 }
|
|
3335 col = (int)(s - regline);
|
|
3336 }
|
|
3337
|
|
3338 retval = regtry(prog, col);
|
|
3339 if (retval > 0)
|
|
3340 break;
|
|
3341
|
|
3342 /* if not currently on the first line, get it again */
|
|
3343 if (reglnum != 0)
|
|
3344 {
|
|
3345 regline = reg_getline((linenr_T)0);
|
|
3346 reglnum = 0;
|
|
3347 }
|
|
3348 if (regline[col] == NUL)
|
|
3349 break;
|
|
3350 #ifdef FEAT_MBYTE
|
|
3351 if (has_mbyte)
|
|
3352 col += (*mb_ptr2len_check)(regline + col);
|
|
3353 else
|
|
3354 #endif
|
|
3355 ++col;
|
|
3356 }
|
|
3357 }
|
|
3358
|
|
3359 theend:
|
|
3360 /* Didn't find a match. */
|
|
3361 vim_free(reg_tofree);
|
|
3362 return retval;
|
|
3363 }
|
|
3364
|
|
3365 #ifdef FEAT_SYN_HL
|
|
3366 static reg_extmatch_T *make_extmatch __ARGS((void));
|
|
3367
|
|
3368 /*
|
|
3369 * Create a new extmatch and mark it as referenced once.
|
|
3370 */
|
|
3371 static reg_extmatch_T *
|
|
3372 make_extmatch()
|
|
3373 {
|
|
3374 reg_extmatch_T *em;
|
|
3375
|
|
3376 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T));
|
|
3377 if (em != NULL)
|
|
3378 em->refcnt = 1;
|
|
3379 return em;
|
|
3380 }
|
|
3381
|
|
3382 /*
|
|
3383 * Add a reference to an extmatch.
|
|
3384 */
|
|
3385 reg_extmatch_T *
|
|
3386 ref_extmatch(em)
|
|
3387 reg_extmatch_T *em;
|
|
3388 {
|
|
3389 if (em != NULL)
|
|
3390 em->refcnt++;
|
|
3391 return em;
|
|
3392 }
|
|
3393
|
|
3394 /*
|
|
3395 * Remove a reference to an extmatch. If there are no references left, free
|
|
3396 * the info.
|
|
3397 */
|
|
3398 void
|
|
3399 unref_extmatch(em)
|
|
3400 reg_extmatch_T *em;
|
|
3401 {
|
|
3402 int i;
|
|
3403
|
|
3404 if (em != NULL && --em->refcnt <= 0)
|
|
3405 {
|
|
3406 for (i = 0; i < NSUBEXP; ++i)
|
|
3407 vim_free(em->matches[i]);
|
|
3408 vim_free(em);
|
|
3409 }
|
|
3410 }
|
|
3411 #endif
|
|
3412
|
|
3413 /*
|
|
3414 * regtry - try match of "prog" with at regline["col"].
|
|
3415 * Returns 0 for failure, number of lines contained in the match otherwise.
|
|
3416 */
|
|
3417 static long
|
|
3418 regtry(prog, col)
|
|
3419 regprog_T *prog;
|
|
3420 colnr_T col;
|
|
3421 {
|
|
3422 reginput = regline + col;
|
|
3423 need_clear_subexpr = TRUE;
|
|
3424 #ifdef FEAT_SYN_HL
|
|
3425 /* Clear the external match subpointers if necessary. */
|
|
3426 if (prog->reghasz == REX_SET)
|
|
3427 need_clear_zsubexpr = TRUE;
|
|
3428 #endif
|
|
3429
|
189
|
3430 if (regmatch(prog->program + 1) == 0)
|
|
3431 return 0;
|
|
3432
|
|
3433 cleanup_subexpr();
|
|
3434 if (REG_MULTI)
|
7
|
3435 {
|
189
|
3436 if (reg_startpos[0].lnum < 0)
|
7
|
3437 {
|
189
|
3438 reg_startpos[0].lnum = 0;
|
|
3439 reg_startpos[0].col = col;
|
|
3440 }
|
|
3441 if (reg_endpos[0].lnum < 0)
|
|
3442 {
|
|
3443 reg_endpos[0].lnum = reglnum;
|
|
3444 reg_endpos[0].col = (int)(reginput - regline);
|
7
|
3445 }
|
|
3446 else
|
189
|
3447 /* Use line number of "\ze". */
|
|
3448 reglnum = reg_endpos[0].lnum;
|
|
3449 }
|
|
3450 else
|
|
3451 {
|
|
3452 if (reg_startp[0] == NULL)
|
|
3453 reg_startp[0] = regline + col;
|
|
3454 if (reg_endp[0] == NULL)
|
|
3455 reg_endp[0] = reginput;
|
|
3456 }
|
7
|
3457 #ifdef FEAT_SYN_HL
|
189
|
3458 /* Package any found \z(...\) matches for export. Default is none. */
|
|
3459 unref_extmatch(re_extmatch_out);
|
|
3460 re_extmatch_out = NULL;
|
|
3461
|
|
3462 if (prog->reghasz == REX_SET)
|
|
3463 {
|
|
3464 int i;
|
|
3465
|
|
3466 cleanup_zsubexpr();
|
|
3467 re_extmatch_out = make_extmatch();
|
|
3468 for (i = 0; i < NSUBEXP; i++)
|
7
|
3469 {
|
189
|
3470 if (REG_MULTI)
|
7
|
3471 {
|
189
|
3472 /* Only accept single line matches. */
|
|
3473 if (reg_startzpos[i].lnum >= 0
|
|
3474 && reg_endzpos[i].lnum == reg_startzpos[i].lnum)
|
|
3475 re_extmatch_out->matches[i] =
|
|
3476 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
|
7
|
3477 + reg_startzpos[i].col,
|
189
|
3478 reg_endzpos[i].col - reg_startzpos[i].col);
|
|
3479 }
|
|
3480 else
|
|
3481 {
|
|
3482 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
|
|
3483 re_extmatch_out->matches[i] =
|
7
|
3484 vim_strnsave(reg_startzp[i],
|
189
|
3485 (int)(reg_endzp[i] - reg_startzp[i]));
|
7
|
3486 }
|
|
3487 }
|
189
|
3488 }
|
7
|
3489 #endif
|
189
|
3490 return 1 + reglnum;
|
7
|
3491 }
|
|
3492
|
|
3493 #ifdef FEAT_MBYTE
|
|
3494 static int reg_prev_class __ARGS((void));
|
|
3495
|
|
3496 /*
|
|
3497 * Get class of previous character.
|
|
3498 */
|
|
3499 static int
|
|
3500 reg_prev_class()
|
|
3501 {
|
|
3502 if (reginput > regline)
|
|
3503 return mb_get_class(reginput - 1
|
|
3504 - (*mb_head_off)(regline, reginput - 1));
|
|
3505 return -1;
|
|
3506 }
|
|
3507
|
|
3508 #endif
|
39
|
3509 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput)
|
7
|
3510
|
|
3511 /*
|
|
3512 * The arguments from BRACE_LIMITS are stored here. They are actually local
|
|
3513 * to regmatch(), but they are here to reduce the amount of stack space used
|
|
3514 * (it can be called recursively many times).
|
|
3515 */
|
|
3516 static long bl_minval;
|
|
3517 static long bl_maxval;
|
|
3518
|
195
|
3519 /* Values for rs_state in regitem_T. */
|
180
|
3520 typedef enum regstate_E
|
|
3521 {
|
|
3522 RS_NOPEN = 0 /* NOPEN and NCLOSE */
|
|
3523 , RS_MOPEN /* MOPEN + [0-9] */
|
|
3524 , RS_MCLOSE /* MCLOSE + [0-9] */
|
|
3525 #ifdef FEAT_SYN_HL
|
|
3526 , RS_ZOPEN /* ZOPEN + [0-9] */
|
|
3527 , RS_ZCLOSE /* ZCLOSE + [0-9] */
|
|
3528 #endif
|
|
3529 , RS_BRANCH /* BRANCH */
|
|
3530 , RS_BRCPLX_MORE /* BRACE_COMPLEX and trying one more match */
|
|
3531 , RS_BRCPLX_LONG /* BRACE_COMPLEX and trying longest match */
|
|
3532 , RS_BRCPLX_SHORT /* BRACE_COMPLEX and trying shortest match */
|
|
3533 , RS_NOMATCH /* NOMATCH */
|
|
3534 , RS_BEHIND1 /* BEHIND / NOBEHIND matching rest */
|
|
3535 , RS_BEHIND2 /* BEHIND / NOBEHIND matching behind part */
|
|
3536 , RS_STAR_LONG /* STAR/PLUS/BRACE_SIMPLE longest match */
|
|
3537 , RS_STAR_SHORT /* STAR/PLUS/BRACE_SIMPLE shortest match */
|
|
3538 } regstate_T;
|
|
3539
|
|
3540 /*
|
|
3541 * When there are alternatives a regstate_T is put on the regstack to remember
|
|
3542 * what we are doing.
|
|
3543 * Before it may be another type of item, depending on rs_state, to remember
|
|
3544 * more things.
|
|
3545 */
|
|
3546 typedef struct regitem_S
|
|
3547 {
|
195
|
3548 regstate_T rs_state; /* what we are doing, one of RS_ above */
|
180
|
3549 char_u *rs_scan; /* current node in program */
|
|
3550 long rs_startp; /* start position for BACK (offset) */
|
|
3551 union
|
|
3552 {
|
|
3553 save_se_T sesave;
|
|
3554 regsave_T regsave;
|
|
3555 } rs_un; /* room for saving reginput */
|
|
3556 short rs_no; /* submatch nr */
|
|
3557 } regitem_T;
|
|
3558
|
|
3559 static regitem_T *regstack_push __ARGS((garray_T *regstack, regstate_T state, char_u *scan, long startp));
|
|
3560 static void regstack_pop __ARGS((garray_T *regstack, char_u **scan, long *startp));
|
|
3561
|
|
3562 /* used for BEHIND and NOBEHIND matching */
|
|
3563 typedef struct regbehind_S
|
|
3564 {
|
|
3565 regsave_T save_after;
|
|
3566 regsave_T save_behind;
|
|
3567 } regbehind_T;
|
|
3568
|
|
3569 /* used for STAR, PLUS and BRACE_SIMPLE matching */
|
|
3570 typedef struct regstar_S
|
|
3571 {
|
|
3572 int nextb; /* next byte */
|
|
3573 int nextb_ic; /* next byte reverse case */
|
|
3574 long count;
|
|
3575 long minval;
|
|
3576 long maxval;
|
|
3577 } regstar_T;
|
|
3578
|
7
|
3579 /*
|
|
3580 * regmatch - main matching routine
|
|
3581 *
|
180
|
3582 * Conceptually the strategy is simple: Check to see whether the current node
|
|
3583 * matches, push an item onto the regstack and loop to see whether the rest
|
|
3584 * matches, and then act accordingly. In practice we make some effort to
|
|
3585 * avoid using the regstack, in particular by going through "ordinary" nodes
|
|
3586 * (that don't need to know whether the rest of the match failed) by a nested
|
|
3587 * loop.
|
7
|
3588 *
|
|
3589 * Returns TRUE when there is a match. Leaves reginput and reglnum just after
|
|
3590 * the last matched character.
|
|
3591 * Returns FALSE when there is no match. Leaves reginput and reglnum in an
|
|
3592 * undefined state!
|
|
3593 */
|
|
3594 static int
|
180
|
3595 regmatch(scan)
|
7
|
3596 char_u *scan; /* Current node. */
|
|
3597 {
|
180
|
3598 char_u *next; /* Next node. */
|
|
3599 int op;
|
|
3600 int c;
|
|
3601 garray_T regstack;
|
|
3602 regitem_T *rp;
|
|
3603 int no;
|
|
3604 int status; /* one of the RA_ values: */
|
|
3605 #define RA_FAIL 1 /* something failed, abort */
|
|
3606 #define RA_CONT 2 /* continue in inner loop */
|
|
3607 #define RA_BREAK 3 /* break inner loop */
|
|
3608 #define RA_MATCH 4 /* successful match */
|
|
3609 #define RA_NOMATCH 5 /* didn't match */
|
|
3610 long startp = 0; /* start position for BACK, offset to
|
|
3611 regstack.ga_data */
|
|
3612 #define STARTP2REGS(startp) (regsave_T *)(((char *)regstack.ga_data) + startp)
|
|
3613 #define REGS2STARTP(p) (long)((char *)p - (char *)regstack.ga_data)
|
|
3614
|
|
3615 /* Init the regstack empty. Use an item size of 1 byte, since we push
|
|
3616 * different things onto it. Use a large grow size to avoid reallocating
|
|
3617 * it too often. */
|
|
3618 ga_init2(®stack, 1, 10000);
|
|
3619
|
|
3620 /*
|
|
3621 * Repeat until the stack is empty.
|
|
3622 */
|
|
3623 for (;;)
|
|
3624 {
|
7
|
3625 /* Some patterns my cause a long time to match, even though they are not
|
|
3626 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */
|
|
3627 fast_breakcheck();
|
|
3628
|
|
3629 #ifdef DEBUG
|
|
3630 if (scan != NULL && regnarrate)
|
|
3631 {
|
|
3632 mch_errmsg(regprop(scan));
|
|
3633 mch_errmsg("(\n");
|
|
3634 }
|
|
3635 #endif
|
180
|
3636
|
|
3637 /*
|
|
3638 * Repeat for items that can be matched sequential, without using the
|
|
3639 * regstack.
|
|
3640 */
|
|
3641 for (;;)
|
7
|
3642 {
|
180
|
3643 if (got_int || scan == NULL)
|
|
3644 {
|
|
3645 status = RA_FAIL;
|
|
3646 break;
|
|
3647 }
|
|
3648 status = RA_CONT;
|
|
3649
|
7
|
3650 #ifdef DEBUG
|
|
3651 if (regnarrate)
|
|
3652 {
|
|
3653 mch_errmsg(regprop(scan));
|
|
3654 mch_errmsg("...\n");
|
|
3655 # ifdef FEAT_SYN_HL
|
|
3656 if (re_extmatch_in != NULL)
|
|
3657 {
|
|
3658 int i;
|
|
3659
|
|
3660 mch_errmsg(_("External submatches:\n"));
|
|
3661 for (i = 0; i < NSUBEXP; i++)
|
|
3662 {
|
|
3663 mch_errmsg(" \"");
|
|
3664 if (re_extmatch_in->matches[i] != NULL)
|
|
3665 mch_errmsg(re_extmatch_in->matches[i]);
|
|
3666 mch_errmsg("\"\n");
|
|
3667 }
|
|
3668 }
|
|
3669 # endif
|
|
3670 }
|
|
3671 #endif
|
|
3672 next = regnext(scan);
|
|
3673
|
|
3674 op = OP(scan);
|
|
3675 /* Check for character class with NL added. */
|
|
3676 if (WITH_NL(op) && *reginput == NUL && reglnum < reg_maxline)
|
|
3677 {
|
|
3678 reg_nextline();
|
|
3679 }
|
|
3680 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n')
|
|
3681 {
|
|
3682 ADVANCE_REGINPUT();
|
|
3683 }
|
|
3684 else
|
|
3685 {
|
|
3686 if (WITH_NL(op))
|
180
|
3687 op -= ADD_NL;
|
7
|
3688 #ifdef FEAT_MBYTE
|
|
3689 if (has_mbyte)
|
|
3690 c = (*mb_ptr2char)(reginput);
|
|
3691 else
|
|
3692 #endif
|
|
3693 c = *reginput;
|
|
3694 switch (op)
|
|
3695 {
|
|
3696 case BOL:
|
|
3697 if (reginput != regline)
|
180
|
3698 status = RA_NOMATCH;
|
7
|
3699 break;
|
|
3700
|
|
3701 case EOL:
|
|
3702 if (c != NUL)
|
180
|
3703 status = RA_NOMATCH;
|
7
|
3704 break;
|
|
3705
|
|
3706 case RE_BOF:
|
|
3707 /* Passing -1 to the getline() function provided for the search
|
|
3708 * should always return NULL if the current line is the first
|
|
3709 * line of the file. */
|
|
3710 if (reglnum != 0 || reginput != regline
|
|
3711 || (REG_MULTI && reg_getline((linenr_T)-1) != NULL))
|
180
|
3712 status = RA_NOMATCH;
|
7
|
3713 break;
|
|
3714
|
|
3715 case RE_EOF:
|
|
3716 if (reglnum != reg_maxline || c != NUL)
|
180
|
3717 status = RA_NOMATCH;
|
7
|
3718 break;
|
|
3719
|
|
3720 case CURSOR:
|
|
3721 /* Check if the buffer is in a window and compare the
|
|
3722 * reg_win->w_cursor position to the match position. */
|
|
3723 if (reg_win == NULL
|
|
3724 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum)
|
|
3725 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col))
|
180
|
3726 status = RA_NOMATCH;
|
7
|
3727 break;
|
|
3728
|
|
3729 case RE_LNUM:
|
|
3730 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum),
|
|
3731 scan))
|
180
|
3732 status = RA_NOMATCH;
|
7
|
3733 break;
|
|
3734
|
|
3735 case RE_COL:
|
|
3736 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan))
|
180
|
3737 status = RA_NOMATCH;
|
7
|
3738 break;
|
|
3739
|
|
3740 case RE_VCOL:
|
|
3741 if (!re_num_cmp((long_u)win_linetabsize(
|
|
3742 reg_win == NULL ? curwin : reg_win,
|
|
3743 regline, (colnr_T)(reginput - regline)) + 1, scan))
|
180
|
3744 status = RA_NOMATCH;
|
7
|
3745 break;
|
|
3746
|
|
3747 case BOW: /* \<word; reginput points to w */
|
|
3748 if (c == NUL) /* Can't match at end of line */
|
180
|
3749 status = RA_NOMATCH;
|
7
|
3750 #ifdef FEAT_MBYTE
|
180
|
3751 else if (has_mbyte)
|
7
|
3752 {
|
|
3753 int this_class;
|
|
3754
|
|
3755 /* Get class of current and previous char (if it exists). */
|
|
3756 this_class = mb_get_class(reginput);
|
|
3757 if (this_class <= 1)
|
180
|
3758 status = RA_NOMATCH; /* not on a word at all */
|
|
3759 else if (reg_prev_class() == this_class)
|
|
3760 status = RA_NOMATCH; /* previous char is in same word */
|
7
|
3761 }
|
|
3762 #endif
|
|
3763 else
|
|
3764 {
|
|
3765 if (!vim_iswordc(c)
|
|
3766 || (reginput > regline && vim_iswordc(reginput[-1])))
|
180
|
3767 status = RA_NOMATCH;
|
7
|
3768 }
|
|
3769 break;
|
|
3770
|
|
3771 case EOW: /* word\>; reginput points after d */
|
|
3772 if (reginput == regline) /* Can't match at start of line */
|
180
|
3773 status = RA_NOMATCH;
|
7
|
3774 #ifdef FEAT_MBYTE
|
180
|
3775 else if (has_mbyte)
|
7
|
3776 {
|
|
3777 int this_class, prev_class;
|
|
3778
|
|
3779 /* Get class of current and previous char (if it exists). */
|
|
3780 this_class = mb_get_class(reginput);
|
|
3781 prev_class = reg_prev_class();
|
180
|
3782 if (this_class == prev_class
|
|
3783 || prev_class == 0 || prev_class == 1)
|
|
3784 status = RA_NOMATCH;
|
7
|
3785 }
|
180
|
3786 #endif
|
7
|
3787 else
|
|
3788 {
|
180
|
3789 if (!vim_iswordc(reginput[-1])
|
|
3790 || (reginput[0] != NUL && vim_iswordc(c)))
|
|
3791 status = RA_NOMATCH;
|
7
|
3792 }
|
|
3793 break; /* Matched with EOW */
|
|
3794
|
|
3795 case ANY:
|
|
3796 if (c == NUL)
|
180
|
3797 status = RA_NOMATCH;
|
|
3798 else
|
|
3799 ADVANCE_REGINPUT();
|
7
|
3800 break;
|
|
3801
|
|
3802 case IDENT:
|
|
3803 if (!vim_isIDc(c))
|
180
|
3804 status = RA_NOMATCH;
|
|
3805 else
|
|
3806 ADVANCE_REGINPUT();
|
7
|
3807 break;
|
|
3808
|
|
3809 case SIDENT:
|
|
3810 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c))
|
180
|
3811 status = RA_NOMATCH;
|
|
3812 else
|
|
3813 ADVANCE_REGINPUT();
|
7
|
3814 break;
|
|
3815
|
|
3816 case KWORD:
|
|
3817 if (!vim_iswordp(reginput))
|
180
|
3818 status = RA_NOMATCH;
|
|
3819 else
|
|
3820 ADVANCE_REGINPUT();
|
7
|
3821 break;
|
|
3822
|
|
3823 case SKWORD:
|
|
3824 if (VIM_ISDIGIT(*reginput) || !vim_iswordp(reginput))
|
180
|
3825 status = RA_NOMATCH;
|
|
3826 else
|
|
3827 ADVANCE_REGINPUT();
|
7
|
3828 break;
|
|
3829
|
|
3830 case FNAME:
|
|
3831 if (!vim_isfilec(c))
|
180
|
3832 status = RA_NOMATCH;
|
|
3833 else
|
|
3834 ADVANCE_REGINPUT();
|
7
|
3835 break;
|
|
3836
|
|
3837 case SFNAME:
|
|
3838 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c))
|
180
|
3839 status = RA_NOMATCH;
|
|
3840 else
|
|
3841 ADVANCE_REGINPUT();
|
7
|
3842 break;
|
|
3843
|
|
3844 case PRINT:
|
|
3845 if (ptr2cells(reginput) != 1)
|
180
|
3846 status = RA_NOMATCH;
|
|
3847 else
|
|
3848 ADVANCE_REGINPUT();
|
7
|
3849 break;
|
|
3850
|
|
3851 case SPRINT:
|
|
3852 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1)
|
180
|
3853 status = RA_NOMATCH;
|
|
3854 else
|
|
3855 ADVANCE_REGINPUT();
|
7
|
3856 break;
|
|
3857
|
|
3858 case WHITE:
|
|
3859 if (!vim_iswhite(c))
|
180
|
3860 status = RA_NOMATCH;
|
|
3861 else
|
|
3862 ADVANCE_REGINPUT();
|
7
|
3863 break;
|
|
3864
|
|
3865 case NWHITE:
|
|
3866 if (c == NUL || vim_iswhite(c))
|
180
|
3867 status = RA_NOMATCH;
|
|
3868 else
|
|
3869 ADVANCE_REGINPUT();
|
7
|
3870 break;
|
|
3871
|
|
3872 case DIGIT:
|
|
3873 if (!ri_digit(c))
|
180
|
3874 status = RA_NOMATCH;
|
|
3875 else
|
|
3876 ADVANCE_REGINPUT();
|
7
|
3877 break;
|
|
3878
|
|
3879 case NDIGIT:
|
|
3880 if (c == NUL || ri_digit(c))
|
180
|
3881 status = RA_NOMATCH;
|
|
3882 else
|
|
3883 ADVANCE_REGINPUT();
|
7
|
3884 break;
|
|
3885
|
|
3886 case HEX:
|
|
3887 if (!ri_hex(c))
|
180
|
3888 status = RA_NOMATCH;
|
|
3889 else
|
|
3890 ADVANCE_REGINPUT();
|
7
|
3891 break;
|
|
3892
|
|
3893 case NHEX:
|
|
3894 if (c == NUL || ri_hex(c))
|
180
|
3895 status = RA_NOMATCH;
|
|
3896 else
|
|
3897 ADVANCE_REGINPUT();
|
7
|
3898 break;
|
|
3899
|
|
3900 case OCTAL:
|
|
3901 if (!ri_octal(c))
|
180
|
3902 status = RA_NOMATCH;
|
|
3903 else
|
|
3904 ADVANCE_REGINPUT();
|
7
|
3905 break;
|
|
3906
|
|
3907 case NOCTAL:
|
|
3908 if (c == NUL || ri_octal(c))
|
180
|
3909 status = RA_NOMATCH;
|
|
3910 else
|
|
3911 ADVANCE_REGINPUT();
|
7
|
3912 break;
|
|
3913
|
|
3914 case WORD:
|
|
3915 if (!ri_word(c))
|
180
|
3916 status = RA_NOMATCH;
|
|
3917 else
|
|
3918 ADVANCE_REGINPUT();
|
7
|
3919 break;
|
|
3920
|
|
3921 case NWORD:
|
|
3922 if (c == NUL || ri_word(c))
|
180
|
3923 status = RA_NOMATCH;
|
|
3924 else
|
|
3925 ADVANCE_REGINPUT();
|
7
|
3926 break;
|
|
3927
|
|
3928 case HEAD:
|
|
3929 if (!ri_head(c))
|
180
|
3930 status = RA_NOMATCH;
|
|
3931 else
|
|
3932 ADVANCE_REGINPUT();
|
7
|
3933 break;
|
|
3934
|
|
3935 case NHEAD:
|
|
3936 if (c == NUL || ri_head(c))
|
180
|
3937 status = RA_NOMATCH;
|
|
3938 else
|
|
3939 ADVANCE_REGINPUT();
|
7
|
3940 break;
|
|
3941
|
|
3942 case ALPHA:
|
|
3943 if (!ri_alpha(c))
|
180
|
3944 status = RA_NOMATCH;
|
|
3945 else
|
|
3946 ADVANCE_REGINPUT();
|
7
|
3947 break;
|
|
3948
|
|
3949 case NALPHA:
|
|
3950 if (c == NUL || ri_alpha(c))
|
180
|
3951 status = RA_NOMATCH;
|
|
3952 else
|
|
3953 ADVANCE_REGINPUT();
|
7
|
3954 break;
|
|
3955
|
|
3956 case LOWER:
|
|
3957 if (!ri_lower(c))
|
180
|
3958 status = RA_NOMATCH;
|
|
3959 else
|
|
3960 ADVANCE_REGINPUT();
|
7
|
3961 break;
|
|
3962
|
|
3963 case NLOWER:
|
|
3964 if (c == NUL || ri_lower(c))
|
180
|
3965 status = RA_NOMATCH;
|
|
3966 else
|
|
3967 ADVANCE_REGINPUT();
|
7
|
3968 break;
|
|
3969
|
|
3970 case UPPER:
|
|
3971 if (!ri_upper(c))
|
180
|
3972 status = RA_NOMATCH;
|
|
3973 else
|
|
3974 ADVANCE_REGINPUT();
|
7
|
3975 break;
|
|
3976
|
|
3977 case NUPPER:
|
|
3978 if (c == NUL || ri_upper(c))
|
180
|
3979 status = RA_NOMATCH;
|
|
3980 else
|
|
3981 ADVANCE_REGINPUT();
|
7
|
3982 break;
|
|
3983
|
|
3984 case EXACTLY:
|
|
3985 {
|
|
3986 int len;
|
|
3987 char_u *opnd;
|
|
3988
|
|
3989 opnd = OPERAND(scan);
|
|
3990 /* Inline the first byte, for speed. */
|
|
3991 if (*opnd != *reginput
|
|
3992 && (!ireg_ic || (
|
|
3993 #ifdef FEAT_MBYTE
|
|
3994 !enc_utf8 &&
|
|
3995 #endif
|
|
3996 TOLOWER_LOC(*opnd) != TOLOWER_LOC(*reginput))))
|
180
|
3997 status = RA_NOMATCH;
|
|
3998 else if (*opnd == NUL)
|
7
|
3999 {
|
|
4000 /* match empty string always works; happens when "~" is
|
|
4001 * empty. */
|
|
4002 }
|
|
4003 else if (opnd[1] == NUL
|
|
4004 #ifdef FEAT_MBYTE
|
|
4005 && !(enc_utf8 && ireg_ic)
|
|
4006 #endif
|
|
4007 )
|
|
4008 ++reginput; /* matched a single char */
|
|
4009 else
|
|
4010 {
|
|
4011 len = (int)STRLEN(opnd);
|
|
4012 /* Need to match first byte again for multi-byte. */
|
|
4013 if (cstrncmp(opnd, reginput, &len) != 0)
|
180
|
4014 status = RA_NOMATCH;
|
7
|
4015 #ifdef FEAT_MBYTE
|
|
4016 /* Check for following composing character. */
|
180
|
4017 else if (enc_utf8
|
|
4018 && UTF_COMPOSINGLIKE(reginput, reginput + len))
|
7
|
4019 {
|
|
4020 /* raaron: This code makes a composing character get
|
|
4021 * ignored, which is the correct behavior (sometimes)
|
|
4022 * for voweled Hebrew texts. */
|
|
4023 if (!ireg_icombine)
|
180
|
4024 status = RA_NOMATCH;
|
7
|
4025 }
|
180
|
4026 #endif
|
7
|
4027 else
|
180
|
4028 reginput += len;
|
7
|
4029 }
|
|
4030 }
|
|
4031 break;
|
|
4032
|
|
4033 case ANYOF:
|
|
4034 case ANYBUT:
|
|
4035 if (c == NUL)
|
180
|
4036 status = RA_NOMATCH;
|
|
4037 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF))
|
|
4038 status = RA_NOMATCH;
|
|
4039 else
|
|
4040 ADVANCE_REGINPUT();
|
7
|
4041 break;
|
|
4042
|
|
4043 #ifdef FEAT_MBYTE
|
|
4044 case MULTIBYTECODE:
|
|
4045 if (has_mbyte)
|
|
4046 {
|
|
4047 int i, len;
|
|
4048 char_u *opnd;
|
|
4049
|
|
4050 opnd = OPERAND(scan);
|
|
4051 /* Safety check (just in case 'encoding' was changed since
|
|
4052 * compiling the program). */
|
|
4053 if ((len = (*mb_ptr2len_check)(opnd)) < 2)
|
180
|
4054 {
|
|
4055 status = RA_NOMATCH;
|
|
4056 break;
|
|
4057 }
|
7
|
4058 for (i = 0; i < len; ++i)
|
|
4059 if (opnd[i] != reginput[i])
|
180
|
4060 {
|
|
4061 status = RA_NOMATCH;
|
|
4062 break;
|
|
4063 }
|
7
|
4064 reginput += len;
|
|
4065 }
|
|
4066 else
|
180
|
4067 status = RA_NOMATCH;
|
7
|
4068 break;
|
|
4069 #endif
|
|
4070
|
|
4071 case NOTHING:
|
|
4072 break;
|
|
4073
|
|
4074 case BACK:
|
167
|
4075 /* When we run into BACK without matching something non-empty, we
|
|
4076 * fail. */
|
180
|
4077 if (startp != 0 && reg_save_equal(STARTP2REGS(startp)))
|
|
4078 status = RA_NOMATCH;
|
7
|
4079 break;
|
|
4080
|
179
|
4081 case BACKP:
|
|
4082 break;
|
|
4083
|
7
|
4084 case MOPEN + 0: /* Match start: \zs */
|
|
4085 case MOPEN + 1: /* \( */
|
|
4086 case MOPEN + 2:
|
|
4087 case MOPEN + 3:
|
|
4088 case MOPEN + 4:
|
|
4089 case MOPEN + 5:
|
|
4090 case MOPEN + 6:
|
|
4091 case MOPEN + 7:
|
|
4092 case MOPEN + 8:
|
|
4093 case MOPEN + 9:
|
|
4094 {
|
|
4095 no = op - MOPEN;
|
|
4096 cleanup_subexpr();
|
180
|
4097 rp = regstack_push(®stack, RS_MOPEN, scan, startp);
|
|
4098 if (rp == NULL)
|
|
4099 status = RA_FAIL;
|
|
4100 else
|
|
4101 {
|
|
4102 rp->rs_no = no;
|
|
4103 save_se(&rp->rs_un.sesave, ®_startpos[no],
|
|
4104 ®_startp[no]);
|
|
4105 /* We simply continue and handle the result when done. */
|
|
4106 }
|
7
|
4107 }
|
180
|
4108 break;
|
7
|
4109
|
|
4110 case NOPEN: /* \%( */
|
|
4111 case NCLOSE: /* \) after \%( */
|
180
|
4112 if (regstack_push(®stack, RS_NOPEN, scan, startp) == NULL)
|
|
4113 status = RA_FAIL;
|
|
4114 /* We simply continue and handle the result when done. */
|
|
4115 break;
|
7
|
4116
|
|
4117 #ifdef FEAT_SYN_HL
|
|
4118 case ZOPEN + 1:
|
|
4119 case ZOPEN + 2:
|
|
4120 case ZOPEN + 3:
|
|
4121 case ZOPEN + 4:
|
|
4122 case ZOPEN + 5:
|
|
4123 case ZOPEN + 6:
|
|
4124 case ZOPEN + 7:
|
|
4125 case ZOPEN + 8:
|
|
4126 case ZOPEN + 9:
|
|
4127 {
|
|
4128 no = op - ZOPEN;
|
|
4129 cleanup_zsubexpr();
|
180
|
4130 rp = regstack_push(®stack, RS_ZOPEN, scan, startp);
|
|
4131 if (rp == NULL)
|
|
4132 status = RA_FAIL;
|
|
4133 else
|
|
4134 {
|
|
4135 rp->rs_no = no;
|
|
4136 save_se(&rp->rs_un.sesave, ®_startzpos[no],
|
|
4137 ®_startzp[no]);
|
|
4138 /* We simply continue and handle the result when done. */
|
|
4139 }
|
7
|
4140 }
|
180
|
4141 break;
|
7
|
4142 #endif
|
|
4143
|
|
4144 case MCLOSE + 0: /* Match end: \ze */
|
|
4145 case MCLOSE + 1: /* \) */
|
|
4146 case MCLOSE + 2:
|
|
4147 case MCLOSE + 3:
|
|
4148 case MCLOSE + 4:
|
|
4149 case MCLOSE + 5:
|
|
4150 case MCLOSE + 6:
|
|
4151 case MCLOSE + 7:
|
|
4152 case MCLOSE + 8:
|
|
4153 case MCLOSE + 9:
|
|
4154 {
|
|
4155 no = op - MCLOSE;
|
|
4156 cleanup_subexpr();
|
180
|
4157 rp = regstack_push(®stack, RS_MCLOSE, scan, startp);
|
|
4158 if (rp == NULL)
|
|
4159 status = RA_FAIL;
|
|
4160 else
|
|
4161 {
|
|
4162 rp->rs_no = no;
|
|
4163 save_se(&rp->rs_un.sesave, ®_endpos[no], ®_endp[no]);
|
|
4164 /* We simply continue and handle the result when done. */
|
|
4165 }
|
7
|
4166 }
|
180
|
4167 break;
|
7
|
4168
|
|
4169 #ifdef FEAT_SYN_HL
|
|
4170 case ZCLOSE + 1: /* \) after \z( */
|
|
4171 case ZCLOSE + 2:
|
|
4172 case ZCLOSE + 3:
|
|
4173 case ZCLOSE + 4:
|
|
4174 case ZCLOSE + 5:
|
|
4175 case ZCLOSE + 6:
|
|
4176 case ZCLOSE + 7:
|
|
4177 case ZCLOSE + 8:
|
|
4178 case ZCLOSE + 9:
|
|
4179 {
|
|
4180 no = op - ZCLOSE;
|
|
4181 cleanup_zsubexpr();
|
180
|
4182 rp = regstack_push(®stack, RS_ZCLOSE, scan, startp);
|
|
4183 if (rp == NULL)
|
|
4184 status = RA_FAIL;
|
|
4185 else
|
|
4186 {
|
|
4187 rp->rs_no = no;
|
|
4188 save_se(&rp->rs_un.sesave, ®_endzpos[no],
|
|
4189 ®_endzp[no]);
|
|
4190 /* We simply continue and handle the result when done. */
|
|
4191 }
|
7
|
4192 }
|
180
|
4193 break;
|
7
|
4194 #endif
|
|
4195
|
|
4196 case BACKREF + 1:
|
|
4197 case BACKREF + 2:
|
|
4198 case BACKREF + 3:
|
|
4199 case BACKREF + 4:
|
|
4200 case BACKREF + 5:
|
|
4201 case BACKREF + 6:
|
|
4202 case BACKREF + 7:
|
|
4203 case BACKREF + 8:
|
|
4204 case BACKREF + 9:
|
|
4205 {
|
|
4206 int len;
|
|
4207 linenr_T clnum;
|
|
4208 colnr_T ccol;
|
|
4209 char_u *p;
|
|
4210
|
|
4211 no = op - BACKREF;
|
|
4212 cleanup_subexpr();
|
|
4213 if (!REG_MULTI) /* Single-line regexp */
|
|
4214 {
|
|
4215 if (reg_endp[no] == NULL)
|
|
4216 {
|
|
4217 /* Backref was not set: Match an empty string. */
|
|
4218 len = 0;
|
|
4219 }
|
|
4220 else
|
|
4221 {
|
|
4222 /* Compare current input with back-ref in the same
|
|
4223 * line. */
|
|
4224 len = (int)(reg_endp[no] - reg_startp[no]);
|
|
4225 if (cstrncmp(reg_startp[no], reginput, &len) != 0)
|
180
|
4226 status = RA_NOMATCH;
|
7
|
4227 }
|
|
4228 }
|
|
4229 else /* Multi-line regexp */
|
|
4230 {
|
|
4231 if (reg_endpos[no].lnum < 0)
|
|
4232 {
|
|
4233 /* Backref was not set: Match an empty string. */
|
|
4234 len = 0;
|
|
4235 }
|
|
4236 else
|
|
4237 {
|
|
4238 if (reg_startpos[no].lnum == reglnum
|
|
4239 && reg_endpos[no].lnum == reglnum)
|
|
4240 {
|
|
4241 /* Compare back-ref within the current line. */
|
|
4242 len = reg_endpos[no].col - reg_startpos[no].col;
|
|
4243 if (cstrncmp(regline + reg_startpos[no].col,
|
|
4244 reginput, &len) != 0)
|
180
|
4245 status = RA_NOMATCH;
|
7
|
4246 }
|
|
4247 else
|
|
4248 {
|
|
4249 /* Messy situation: Need to compare between two
|
|
4250 * lines. */
|
|
4251 ccol = reg_startpos[no].col;
|
|
4252 clnum = reg_startpos[no].lnum;
|
|
4253 for (;;)
|
|
4254 {
|
|
4255 /* Since getting one line may invalidate
|
|
4256 * the other, need to make copy. Slow! */
|
|
4257 if (regline != reg_tofree)
|
|
4258 {
|
|
4259 len = (int)STRLEN(regline);
|
|
4260 if (reg_tofree == NULL
|
|
4261 || len >= (int)reg_tofreelen)
|
|
4262 {
|
|
4263 len += 50; /* get some extra */
|
|
4264 vim_free(reg_tofree);
|
|
4265 reg_tofree = alloc(len);
|
|
4266 if (reg_tofree == NULL)
|
180
|
4267 {
|
|
4268 status = RA_FAIL; /* outof memory!*/
|
|
4269 break;
|
|
4270 }
|
7
|
4271 reg_tofreelen = len;
|
|
4272 }
|
|
4273 STRCPY(reg_tofree, regline);
|
|
4274 reginput = reg_tofree
|
|
4275 + (reginput - regline);
|
|
4276 regline = reg_tofree;
|
|
4277 }
|
|
4278
|
|
4279 /* Get the line to compare with. */
|
|
4280 p = reg_getline(clnum);
|
|
4281 if (clnum == reg_endpos[no].lnum)
|
|
4282 len = reg_endpos[no].col - ccol;
|
|
4283 else
|
|
4284 len = (int)STRLEN(p + ccol);
|
|
4285
|
|
4286 if (cstrncmp(p + ccol, reginput, &len) != 0)
|
180
|
4287 {
|
|
4288 status = RA_NOMATCH; /* doesn't match */
|
|
4289 break;
|
|
4290 }
|
7
|
4291 if (clnum == reg_endpos[no].lnum)
|
|
4292 break; /* match and at end! */
|
|
4293 if (reglnum == reg_maxline)
|
180
|
4294 {
|
|
4295 status = RA_NOMATCH; /* text too short */
|
|
4296 break;
|
|
4297 }
|
7
|
4298
|
|
4299 /* Advance to next line. */
|
|
4300 reg_nextline();
|
|
4301 ++clnum;
|
|
4302 ccol = 0;
|
180
|
4303 if (got_int)
|
|
4304 {
|
|
4305 status = RA_FAIL;
|
|
4306 break;
|
|
4307 }
|
7
|
4308 }
|
|
4309
|
|
4310 /* found a match! Note that regline may now point
|
|
4311 * to a copy of the line, that should not matter. */
|
|
4312 }
|
|
4313 }
|
|
4314 }
|
|
4315
|
|
4316 /* Matched the backref, skip over it. */
|
|
4317 reginput += len;
|
|
4318 }
|
|
4319 break;
|
|
4320
|
|
4321 #ifdef FEAT_SYN_HL
|
|
4322 case ZREF + 1:
|
|
4323 case ZREF + 2:
|
|
4324 case ZREF + 3:
|
|
4325 case ZREF + 4:
|
|
4326 case ZREF + 5:
|
|
4327 case ZREF + 6:
|
|
4328 case ZREF + 7:
|
|
4329 case ZREF + 8:
|
|
4330 case ZREF + 9:
|
|
4331 {
|
|
4332 int len;
|
|
4333
|
|
4334 cleanup_zsubexpr();
|
|
4335 no = op - ZREF;
|
|
4336 if (re_extmatch_in != NULL
|
|
4337 && re_extmatch_in->matches[no] != NULL)
|
|
4338 {
|
|
4339 len = (int)STRLEN(re_extmatch_in->matches[no]);
|
|
4340 if (cstrncmp(re_extmatch_in->matches[no],
|
|
4341 reginput, &len) != 0)
|
180
|
4342 status = RA_NOMATCH;
|
|
4343 else
|
|
4344 reginput += len;
|
7
|
4345 }
|
|
4346 else
|
|
4347 {
|
|
4348 /* Backref was not set: Match an empty string. */
|
|
4349 }
|
|
4350 }
|
|
4351 break;
|
|
4352 #endif
|
|
4353
|
|
4354 case BRANCH:
|
|
4355 {
|
|
4356 if (OP(next) != BRANCH) /* No choice. */
|
|
4357 next = OPERAND(scan); /* Avoid recursion. */
|
180
|
4358 else if (startp != 0 && OP(OPERAND(scan)) == BACKP
|
|
4359 && reg_save_equal(STARTP2REGS(startp)))
|
179
|
4360 /* \+ with something empty before it */
|
180
|
4361 status = RA_NOMATCH;
|
7
|
4362 else
|
|
4363 {
|
180
|
4364 rp = regstack_push(®stack, RS_BRANCH, scan, startp);
|
|
4365 if (rp == NULL)
|
|
4366 status = RA_FAIL;
|
|
4367 else
|
|
4368 status = RA_BREAK; /* rest is below */
|
7
|
4369 }
|
|
4370 }
|
|
4371 break;
|
|
4372
|
|
4373 case BRACE_LIMITS:
|
|
4374 {
|
|
4375 if (OP(next) == BRACE_SIMPLE)
|
|
4376 {
|
|
4377 bl_minval = OPERAND_MIN(scan);
|
|
4378 bl_maxval = OPERAND_MAX(scan);
|
|
4379 }
|
|
4380 else if (OP(next) >= BRACE_COMPLEX
|
|
4381 && OP(next) < BRACE_COMPLEX + 10)
|
|
4382 {
|
|
4383 no = OP(next) - BRACE_COMPLEX;
|
|
4384 brace_min[no] = OPERAND_MIN(scan);
|
|
4385 brace_max[no] = OPERAND_MAX(scan);
|
|
4386 brace_count[no] = 0;
|
|
4387 }
|
|
4388 else
|
|
4389 {
|
|
4390 EMSG(_(e_internal)); /* Shouldn't happen */
|
180
|
4391 status = RA_FAIL;
|
7
|
4392 }
|
|
4393 }
|
|
4394 break;
|
|
4395
|
|
4396 case BRACE_COMPLEX + 0:
|
|
4397 case BRACE_COMPLEX + 1:
|
|
4398 case BRACE_COMPLEX + 2:
|
|
4399 case BRACE_COMPLEX + 3:
|
|
4400 case BRACE_COMPLEX + 4:
|
|
4401 case BRACE_COMPLEX + 5:
|
|
4402 case BRACE_COMPLEX + 6:
|
|
4403 case BRACE_COMPLEX + 7:
|
|
4404 case BRACE_COMPLEX + 8:
|
|
4405 case BRACE_COMPLEX + 9:
|
|
4406 {
|
|
4407 no = op - BRACE_COMPLEX;
|
|
4408 ++brace_count[no];
|
|
4409
|
|
4410 /* If not matched enough times yet, try one more */
|
|
4411 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
|
180
|
4412 ? brace_min[no] : brace_max[no]))
|
7
|
4413 {
|
180
|
4414 rp = regstack_push(®stack, RS_BRCPLX_MORE, scan, startp);
|
|
4415 if (rp == NULL)
|
|
4416 status = RA_FAIL;
|
|
4417 else
|
|
4418 {
|
|
4419 rp->rs_no = no;
|
|
4420 reg_save(&rp->rs_un.regsave);
|
|
4421 startp = REGS2STARTP(&rp->rs_un.regsave);
|
|
4422 next = OPERAND(scan);
|
|
4423 /* We continue and handle the result when done. */
|
|
4424 }
|
|
4425 break;
|
7
|
4426 }
|
|
4427
|
|
4428 /* If matched enough times, may try matching some more */
|
|
4429 if (brace_min[no] <= brace_max[no])
|
|
4430 {
|
|
4431 /* Range is the normal way around, use longest match */
|
|
4432 if (brace_count[no] <= brace_max[no])
|
|
4433 {
|
180
|
4434 rp = regstack_push(®stack, RS_BRCPLX_LONG,
|
|
4435 scan, startp);
|
|
4436 if (rp == NULL)
|
|
4437 status = RA_FAIL;
|
|
4438 else
|
|
4439 {
|
|
4440 rp->rs_no = no;
|
|
4441 reg_save(&rp->rs_un.regsave);
|
|
4442 startp = REGS2STARTP(&rp->rs_un.regsave);
|
|
4443 next = OPERAND(scan);
|
|
4444 /* We continue and handle the result when done. */
|
|
4445 }
|
7
|
4446 }
|
|
4447 }
|
|
4448 else
|
|
4449 {
|
|
4450 /* Range is backwards, use shortest match first */
|
|
4451 if (brace_count[no] <= brace_min[no])
|
|
4452 {
|
180
|
4453 rp = regstack_push(®stack, RS_BRCPLX_SHORT,
|
|
4454 scan, startp);
|
|
4455 if (rp == NULL)
|
|
4456 status = RA_FAIL;
|
|
4457 else
|
|
4458 {
|
|
4459 reg_save(&rp->rs_un.regsave);
|
|
4460 startp = REGS2STARTP(&rp->rs_un.regsave);
|
|
4461 /* We continue and handle the result when done. */
|
|
4462 }
|
7
|
4463 }
|
|
4464 }
|
|
4465 }
|
|
4466 break;
|
|
4467
|
|
4468 case BRACE_SIMPLE:
|
|
4469 case STAR:
|
|
4470 case PLUS:
|
|
4471 {
|
180
|
4472 regstar_T rst;
|
7
|
4473
|
|
4474 /*
|
|
4475 * Lookahead to avoid useless match attempts when we know
|
|
4476 * what character comes next.
|
|
4477 */
|
|
4478 if (OP(next) == EXACTLY)
|
|
4479 {
|
180
|
4480 rst.nextb = *OPERAND(next);
|
7
|
4481 if (ireg_ic)
|
|
4482 {
|
180
|
4483 if (isupper(rst.nextb))
|
|
4484 rst.nextb_ic = TOLOWER_LOC(rst.nextb);
|
7
|
4485 else
|
180
|
4486 rst.nextb_ic = TOUPPER_LOC(rst.nextb);
|
7
|
4487 }
|
|
4488 else
|
180
|
4489 rst.nextb_ic = rst.nextb;
|
7
|
4490 }
|
|
4491 else
|
|
4492 {
|
180
|
4493 rst.nextb = NUL;
|
|
4494 rst.nextb_ic = NUL;
|
7
|
4495 }
|
|
4496 if (op != BRACE_SIMPLE)
|
|
4497 {
|
180
|
4498 rst.minval = (op == STAR) ? 0 : 1;
|
|
4499 rst.maxval = MAX_LIMIT;
|
7
|
4500 }
|
|
4501 else
|
|
4502 {
|
180
|
4503 rst.minval = bl_minval;
|
|
4504 rst.maxval = bl_maxval;
|
7
|
4505 }
|
|
4506
|
|
4507 /*
|
|
4508 * When maxval > minval, try matching as much as possible, up
|
|
4509 * to maxval. When maxval < minval, try matching at least the
|
|
4510 * minimal number (since the range is backwards, that's also
|
|
4511 * maxval!).
|
|
4512 */
|
180
|
4513 rst.count = regrepeat(OPERAND(scan), rst.maxval);
|
7
|
4514 if (got_int)
|
180
|
4515 {
|
|
4516 status = RA_FAIL;
|
|
4517 break;
|
|
4518 }
|
|
4519 if (rst.minval <= rst.maxval
|
|
4520 ? rst.count >= rst.minval : rst.count >= rst.maxval)
|
7
|
4521 {
|
180
|
4522 /* It could match. Prepare for trying to match what
|
|
4523 * follows. The code is below. Parameters are stored in
|
|
4524 * a regstar_T on the regstack. */
|
212
|
4525 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
|
189
|
4526 {
|
|
4527 EMSG(_(e_maxmempat));
|
|
4528 status = RA_FAIL;
|
|
4529 }
|
|
4530 else if (ga_grow(®stack, sizeof(regstar_T)) == FAIL)
|
180
|
4531 status = RA_FAIL;
|
|
4532 else
|
7
|
4533 {
|
180
|
4534 regstack.ga_len += sizeof(regstar_T);
|
|
4535 rp = regstack_push(®stack, rst.minval <= rst.maxval
|
|
4536 ? RS_STAR_LONG : RS_STAR_SHORT, scan, startp);
|
|
4537 if (rp == NULL)
|
|
4538 status = RA_FAIL;
|
|
4539 else
|
7
|
4540 {
|
180
|
4541 *(((regstar_T *)rp) - 1) = rst;
|
|
4542 status = RA_BREAK; /* skip the restore bits */
|
7
|
4543 }
|
|
4544 }
|
|
4545 }
|
|
4546 else
|
180
|
4547 status = RA_NOMATCH;
|
|
4548
|
7
|
4549 }
|
180
|
4550 break;
|
7
|
4551
|
|
4552 case NOMATCH:
|
|
4553 case MATCH:
|
|
4554 case SUBPAT:
|
180
|
4555 rp = regstack_push(®stack, RS_NOMATCH, scan, startp);
|
|
4556 if (rp == NULL)
|
|
4557 status = RA_FAIL;
|
|
4558 else
|
7
|
4559 {
|
180
|
4560 rp->rs_no = op;
|
|
4561 reg_save(&rp->rs_un.regsave);
|
|
4562 next = OPERAND(scan);
|
|
4563 /* We continue and handle the result when done. */
|
7
|
4564 }
|
|
4565 break;
|
|
4566
|
|
4567 case BEHIND:
|
|
4568 case NOBEHIND:
|
180
|
4569 /* Need a bit of room to store extra positions. */
|
212
|
4570 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
|
189
|
4571 {
|
|
4572 EMSG(_(e_maxmempat));
|
|
4573 status = RA_FAIL;
|
|
4574 }
|
|
4575 else if (ga_grow(®stack, sizeof(regbehind_T)) == FAIL)
|
180
|
4576 status = RA_FAIL;
|
|
4577 else
|
7
|
4578 {
|
180
|
4579 regstack.ga_len += sizeof(regbehind_T);
|
|
4580 rp = regstack_push(®stack, RS_BEHIND1, scan, startp);
|
|
4581 if (rp == NULL)
|
|
4582 status = RA_FAIL;
|
|
4583 else
|
7
|
4584 {
|
180
|
4585 rp->rs_no = op;
|
|
4586 reg_save(&rp->rs_un.regsave);
|
|
4587 /* First try if what follows matches. If it does then we
|
|
4588 * check the behind match by looping. */
|
7
|
4589 }
|
|
4590 }
|
180
|
4591 break;
|
7
|
4592
|
|
4593 case BHPOS:
|
|
4594 if (REG_MULTI)
|
|
4595 {
|
|
4596 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline)
|
|
4597 || behind_pos.rs_u.pos.lnum != reglnum)
|
180
|
4598 status = RA_NOMATCH;
|
7
|
4599 }
|
|
4600 else if (behind_pos.rs_u.ptr != reginput)
|
180
|
4601 status = RA_NOMATCH;
|
7
|
4602 break;
|
|
4603
|
|
4604 case NEWL:
|
|
4605 if ((c != NUL || reglnum == reg_maxline)
|
|
4606 && (c != '\n' || !reg_line_lbr))
|
180
|
4607 status = RA_NOMATCH;
|
|
4608 else if (reg_line_lbr)
|
7
|
4609 ADVANCE_REGINPUT();
|
|
4610 else
|
|
4611 reg_nextline();
|
|
4612 break;
|
|
4613
|
|
4614 case END:
|
180
|
4615 status = RA_MATCH; /* Success! */
|
|
4616 break;
|
7
|
4617
|
|
4618 default:
|
|
4619 EMSG(_(e_re_corr));
|
|
4620 #ifdef DEBUG
|
|
4621 printf("Illegal op code %d\n", op);
|
|
4622 #endif
|
180
|
4623 status = RA_FAIL;
|
|
4624 break;
|
7
|
4625 }
|
|
4626 }
|
|
4627
|
180
|
4628 /* If we can't continue sequentially, break the inner loop. */
|
|
4629 if (status != RA_CONT)
|
|
4630 break;
|
|
4631
|
|
4632 /* Continue in inner loop, advance to next item. */
|
7
|
4633 scan = next;
|
180
|
4634
|
|
4635 } /* end of inner loop */
|
7
|
4636
|
|
4637 /*
|
180
|
4638 * If there is something on the regstack execute the code for the state.
|
|
4639 * If the state is popped then loop.
|
7
|
4640 */
|
180
|
4641 while (regstack.ga_len > 0 && status != RA_FAIL)
|
|
4642 {
|
|
4643 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
|
|
4644 switch (rp->rs_state)
|
|
4645 {
|
|
4646 case RS_NOPEN:
|
|
4647 /* Result is passed on as-is, simply pop the state. */
|
|
4648 regstack_pop(®stack, &scan, &startp);
|
|
4649 break;
|
|
4650
|
|
4651 case RS_MOPEN:
|
|
4652 /* Pop the state. Restore pointers when there is no match. */
|
|
4653 if (status == RA_NOMATCH)
|
|
4654 restore_se(&rp->rs_un.sesave, ®_startpos[rp->rs_no],
|
|
4655 ®_startp[rp->rs_no]);
|
|
4656 regstack_pop(®stack, &scan, &startp);
|
|
4657 break;
|
|
4658
|
|
4659 #ifdef FEAT_SYN_HL
|
|
4660 case RS_ZOPEN:
|
|
4661 /* Pop the state. Restore pointers when there is no match. */
|
|
4662 if (status == RA_NOMATCH)
|
|
4663 restore_se(&rp->rs_un.sesave, ®_startzpos[rp->rs_no],
|
|
4664 ®_startzp[rp->rs_no]);
|
|
4665 regstack_pop(®stack, &scan, &startp);
|
|
4666 break;
|
|
4667 #endif
|
|
4668
|
|
4669 case RS_MCLOSE:
|
|
4670 /* Pop the state. Restore pointers when there is no match. */
|
|
4671 if (status == RA_NOMATCH)
|
|
4672 restore_se(&rp->rs_un.sesave, ®_endpos[rp->rs_no],
|
|
4673 ®_endp[rp->rs_no]);
|
|
4674 regstack_pop(®stack, &scan, &startp);
|
|
4675 break;
|
|
4676
|
|
4677 #ifdef FEAT_SYN_HL
|
|
4678 case RS_ZCLOSE:
|
|
4679 /* Pop the state. Restore pointers when there is no match. */
|
|
4680 if (status == RA_NOMATCH)
|
|
4681 restore_se(&rp->rs_un.sesave, ®_endzpos[rp->rs_no],
|
|
4682 ®_endzp[rp->rs_no]);
|
|
4683 regstack_pop(®stack, &scan, &startp);
|
|
4684 break;
|
|
4685 #endif
|
|
4686
|
|
4687 case RS_BRANCH:
|
|
4688 if (status == RA_MATCH)
|
|
4689 /* this branch matched, use it */
|
|
4690 regstack_pop(®stack, &scan, &startp);
|
|
4691 else
|
|
4692 {
|
|
4693 if (status != RA_BREAK)
|
|
4694 {
|
|
4695 /* After a non-matching branch: try next one. */
|
|
4696 reg_restore(&rp->rs_un.regsave);
|
|
4697 scan = rp->rs_scan;
|
|
4698 }
|
|
4699 if (scan == NULL || OP(scan) != BRANCH)
|
|
4700 {
|
|
4701 /* no more branches, didn't find a match */
|
|
4702 status = RA_NOMATCH;
|
|
4703 regstack_pop(®stack, &scan, &startp);
|
|
4704 }
|
|
4705 else
|
|
4706 {
|
|
4707 /* Prepare to try a branch. */
|
|
4708 rp->rs_scan = regnext(scan);
|
|
4709 reg_save(&rp->rs_un.regsave);
|
|
4710 startp = REGS2STARTP(&rp->rs_un.regsave);
|
|
4711 scan = OPERAND(scan);
|
|
4712 }
|
|
4713 }
|
|
4714 break;
|
|
4715
|
|
4716 case RS_BRCPLX_MORE:
|
|
4717 /* Pop the state. Restore pointers when there is no match. */
|
|
4718 if (status == RA_NOMATCH)
|
|
4719 {
|
|
4720 reg_restore(&rp->rs_un.regsave);
|
|
4721 --brace_count[rp->rs_no]; /* decrement match count */
|
|
4722 }
|
|
4723 regstack_pop(®stack, &scan, &startp);
|
|
4724 break;
|
|
4725
|
|
4726 case RS_BRCPLX_LONG:
|
|
4727 /* Pop the state. Restore pointers when there is no match. */
|
|
4728 if (status == RA_NOMATCH)
|
|
4729 {
|
|
4730 /* There was no match, but we did find enough matches. */
|
|
4731 reg_restore(&rp->rs_un.regsave);
|
|
4732 --brace_count[rp->rs_no];
|
|
4733 /* continue with the items after "\{}" */
|
|
4734 status = RA_CONT;
|
|
4735 }
|
|
4736 regstack_pop(®stack, &scan, &startp);
|
|
4737 if (status == RA_CONT)
|
|
4738 scan = regnext(scan);
|
|
4739 break;
|
|
4740
|
|
4741 case RS_BRCPLX_SHORT:
|
|
4742 /* Pop the state. Restore pointers when there is no match. */
|
|
4743 if (status == RA_NOMATCH)
|
|
4744 /* There was no match, try to match one more item. */
|
|
4745 reg_restore(&rp->rs_un.regsave);
|
|
4746 regstack_pop(®stack, &scan, &startp);
|
|
4747 if (status == RA_NOMATCH)
|
|
4748 {
|
|
4749 scan = OPERAND(scan);
|
|
4750 status = RA_CONT;
|
|
4751 }
|
|
4752 break;
|
|
4753
|
|
4754 case RS_NOMATCH:
|
|
4755 /* Pop the state. If the operand matches for NOMATCH or
|
|
4756 * doesn't match for MATCH/SUBPAT, we fail. Otherwise backup,
|
|
4757 * except for SUBPAT, and continue with the next item. */
|
|
4758 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH))
|
|
4759 status = RA_NOMATCH;
|
|
4760 else
|
|
4761 {
|
|
4762 status = RA_CONT;
|
|
4763 if (rp->rs_no != SUBPAT)
|
|
4764 reg_restore(&rp->rs_un.regsave); /* zero-width */
|
|
4765 }
|
|
4766 regstack_pop(®stack, &scan, &startp);
|
|
4767 if (status == RA_CONT)
|
|
4768 scan = regnext(scan);
|
|
4769 break;
|
|
4770
|
|
4771 case RS_BEHIND1:
|
|
4772 if (status == RA_NOMATCH)
|
|
4773 {
|
|
4774 regstack_pop(®stack, &scan, &startp);
|
|
4775 regstack.ga_len -= sizeof(regbehind_T);
|
|
4776 }
|
|
4777 else
|
|
4778 {
|
|
4779 /* The stuff after BEHIND/NOBEHIND matches. Now try if
|
|
4780 * the behind part does (not) match before the current
|
|
4781 * position in the input. This must be done at every
|
|
4782 * position in the input and checking if the match ends at
|
|
4783 * the current position. */
|
|
4784
|
|
4785 /* save the position after the found match for next */
|
|
4786 reg_save(&(((regbehind_T *)rp) - 1)->save_after);
|
|
4787
|
|
4788 /* start looking for a match with operand at the current
|
|
4789 * postion. Go back one character until we find the
|
|
4790 * result, hitting the start of the line or the previous
|
|
4791 * line (for multi-line matching).
|
|
4792 * Set behind_pos to where the match should end, BHPOS
|
|
4793 * will match it. Save the current value. */
|
|
4794 (((regbehind_T *)rp) - 1)->save_behind = behind_pos;
|
|
4795 behind_pos = rp->rs_un.regsave;
|
|
4796
|
|
4797 rp->rs_state = RS_BEHIND2;
|
|
4798
|
|
4799 reg_restore(&rp->rs_un.regsave);
|
|
4800 scan = OPERAND(rp->rs_scan);
|
|
4801 }
|
|
4802 break;
|
|
4803
|
|
4804 case RS_BEHIND2:
|
|
4805 /*
|
|
4806 * Looping for BEHIND / NOBEHIND match.
|
|
4807 */
|
|
4808 if (status == RA_MATCH && reg_save_equal(&behind_pos))
|
|
4809 {
|
|
4810 /* found a match that ends where "next" started */
|
|
4811 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
|
|
4812 if (rp->rs_no == BEHIND)
|
|
4813 reg_restore(&(((regbehind_T *)rp) - 1)->save_after);
|
|
4814 else
|
|
4815 /* But we didn't want a match. */
|
|
4816 status = RA_NOMATCH;
|
|
4817 regstack_pop(®stack, &scan, &startp);
|
|
4818 regstack.ga_len -= sizeof(regbehind_T);
|
|
4819 }
|
|
4820 else
|
|
4821 {
|
|
4822 /* No match: Go back one character. May go to previous
|
|
4823 * line once. */
|
|
4824 no = OK;
|
|
4825 if (REG_MULTI)
|
|
4826 {
|
|
4827 if (rp->rs_un.regsave.rs_u.pos.col == 0)
|
|
4828 {
|
|
4829 if (rp->rs_un.regsave.rs_u.pos.lnum
|
|
4830 < behind_pos.rs_u.pos.lnum
|
|
4831 || reg_getline(
|
|
4832 --rp->rs_un.regsave.rs_u.pos.lnum)
|
|
4833 == NULL)
|
|
4834 no = FAIL;
|
|
4835 else
|
|
4836 {
|
|
4837 reg_restore(&rp->rs_un.regsave);
|
|
4838 rp->rs_un.regsave.rs_u.pos.col =
|
|
4839 (colnr_T)STRLEN(regline);
|
|
4840 }
|
|
4841 }
|
|
4842 else
|
|
4843 --rp->rs_un.regsave.rs_u.pos.col;
|
|
4844 }
|
|
4845 else
|
|
4846 {
|
|
4847 if (rp->rs_un.regsave.rs_u.ptr == regline)
|
|
4848 no = FAIL;
|
|
4849 else
|
|
4850 --rp->rs_un.regsave.rs_u.ptr;
|
|
4851 }
|
|
4852 if (no == OK)
|
|
4853 {
|
|
4854 /* Advanced, prepare for finding match again. */
|
|
4855 reg_restore(&rp->rs_un.regsave);
|
|
4856 scan = OPERAND(rp->rs_scan);
|
|
4857 }
|
|
4858 else
|
|
4859 {
|
|
4860 /* Can't advance. For NOBEHIND that's a match. */
|
|
4861 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
|
|
4862 if (rp->rs_no == NOBEHIND)
|
|
4863 {
|
|
4864 reg_restore(&(((regbehind_T *)rp) - 1)->save_after);
|
|
4865 status = RA_MATCH;
|
|
4866 }
|
|
4867 else
|
|
4868 status = RA_NOMATCH;
|
|
4869 regstack_pop(®stack, &scan, &startp);
|
|
4870 regstack.ga_len -= sizeof(regbehind_T);
|
|
4871 }
|
|
4872 }
|
|
4873 break;
|
|
4874
|
|
4875 case RS_STAR_LONG:
|
|
4876 case RS_STAR_SHORT:
|
|
4877 {
|
|
4878 regstar_T *rst = ((regstar_T *)rp) - 1;
|
|
4879
|
|
4880 if (status == RA_MATCH)
|
|
4881 {
|
|
4882 regstack_pop(®stack, &scan, &startp);
|
|
4883 regstack.ga_len -= sizeof(regstar_T);
|
|
4884 break;
|
|
4885 }
|
|
4886
|
|
4887 /* Tried once already, restore input pointers. */
|
|
4888 if (status != RA_BREAK)
|
|
4889 reg_restore(&rp->rs_un.regsave);
|
|
4890
|
|
4891 /* Repeat until we found a position where it could match. */
|
|
4892 for (;;)
|
|
4893 {
|
|
4894 if (status != RA_BREAK)
|
|
4895 {
|
|
4896 /* Tried first position already, advance. */
|
|
4897 if (rp->rs_state == RS_STAR_LONG)
|
|
4898 {
|
|
4899 /* Trying for longest matc, but couldn't or didn't
|
|
4900 * match -- back up one char. */
|
|
4901 if (--rst->count < rst->minval)
|
|
4902 break;
|
|
4903 if (reginput == regline)
|
|
4904 {
|
|
4905 /* backup to last char of previous line */
|
|
4906 --reglnum;
|
|
4907 regline = reg_getline(reglnum);
|
|
4908 /* Just in case regrepeat() didn't count
|
|
4909 * right. */
|
|
4910 if (regline == NULL)
|
|
4911 break;
|
|
4912 reginput = regline + STRLEN(regline);
|
|
4913 fast_breakcheck();
|
|
4914 }
|
|
4915 else
|
|
4916 mb_ptr_back(regline, reginput);
|
|
4917 }
|
|
4918 else
|
|
4919 {
|
|
4920 /* Range is backwards, use shortest match first.
|
|
4921 * Careful: maxval and minval are exchanged!
|
|
4922 * Couldn't or didn't match: try advancing one
|
|
4923 * char. */
|
|
4924 if (rst->count == rst->minval
|
|
4925 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0)
|
|
4926 break;
|
|
4927 ++rst->count;
|
|
4928 }
|
|
4929 if (got_int)
|
|
4930 break;
|
|
4931 }
|
|
4932 else
|
|
4933 status = RA_NOMATCH;
|
|
4934
|
|
4935 /* If it could match, try it. */
|
|
4936 if (rst->nextb == NUL || *reginput == rst->nextb
|
|
4937 || *reginput == rst->nextb_ic)
|
|
4938 {
|
|
4939 reg_save(&rp->rs_un.regsave);
|
|
4940 scan = regnext(rp->rs_scan);
|
|
4941 status = RA_CONT;
|
|
4942 break;
|
|
4943 }
|
|
4944 }
|
|
4945 if (status != RA_CONT)
|
|
4946 {
|
|
4947 /* Failed. */
|
|
4948 regstack_pop(®stack, &scan, &startp);
|
|
4949 regstack.ga_len -= sizeof(regstar_T);
|
|
4950 status = RA_NOMATCH;
|
|
4951 }
|
|
4952 }
|
|
4953 break;
|
|
4954 }
|
|
4955
|
|
4956 /* If we want to continue the inner loop or didn't pop a state contine
|
|
4957 * matching loop */
|
|
4958 if (status == RA_CONT || rp == (regitem_T *)
|
|
4959 ((char *)regstack.ga_data + regstack.ga_len) - 1)
|
|
4960 break;
|
|
4961 }
|
|
4962
|
189
|
4963 /* May need to continue with the inner loop, starting at "scan". */
|
180
|
4964 if (status == RA_CONT)
|
|
4965 continue;
|
|
4966
|
|
4967 /*
|
|
4968 * If the regstack is empty or something failed we are done.
|
|
4969 */
|
|
4970 if (regstack.ga_len == 0 || status == RA_FAIL)
|
|
4971 {
|
|
4972 ga_clear(®stack);
|
|
4973 if (scan == NULL)
|
|
4974 {
|
|
4975 /*
|
|
4976 * We get here only if there's trouble -- normally "case END" is
|
|
4977 * the terminating point.
|
|
4978 */
|
|
4979 EMSG(_(e_re_corr));
|
7
|
4980 #ifdef DEBUG
|
180
|
4981 printf("Premature EOL\n");
|
7
|
4982 #endif
|
180
|
4983 }
|
189
|
4984 if (status == RA_FAIL)
|
|
4985 got_int = TRUE;
|
180
|
4986 return (status == RA_MATCH);
|
|
4987 }
|
|
4988
|
|
4989 } /* End of loop until the regstack is empty. */
|
|
4990
|
|
4991 /* NOTREACHED */
|
|
4992 }
|
|
4993
|
|
4994 /*
|
|
4995 * Push an item onto the regstack.
|
|
4996 * Returns pointer to new item. Returns NULL when out of memory.
|
|
4997 */
|
|
4998 static regitem_T *
|
|
4999 regstack_push(regstack, state, scan, startp)
|
|
5000 garray_T *regstack;
|
|
5001 regstate_T state;
|
|
5002 char_u *scan;
|
|
5003 long startp;
|
|
5004 {
|
|
5005 regitem_T *rp;
|
|
5006
|
212
|
5007 if ((long)((unsigned)regstack->ga_len >> 10) >= p_mmp)
|
189
|
5008 {
|
|
5009 EMSG(_(e_maxmempat));
|
|
5010 return NULL;
|
|
5011 }
|
180
|
5012 if (ga_grow(regstack, sizeof(regitem_T)) == FAIL)
|
|
5013 return NULL;
|
|
5014
|
|
5015 rp = (regitem_T *)((char *)regstack->ga_data + regstack->ga_len);
|
|
5016 rp->rs_state = state;
|
|
5017 rp->rs_scan = scan;
|
|
5018 rp->rs_startp = startp;
|
|
5019
|
|
5020 regstack->ga_len += sizeof(regitem_T);
|
|
5021 return rp;
|
|
5022 }
|
|
5023
|
|
5024 /*
|
|
5025 * Pop an item from the regstack.
|
|
5026 */
|
|
5027 static void
|
|
5028 regstack_pop(regstack, scan, startp)
|
|
5029 garray_T *regstack;
|
|
5030 char_u **scan;
|
|
5031 long *startp;
|
|
5032 {
|
|
5033 regitem_T *rp;
|
|
5034
|
|
5035 rp = (regitem_T *)((char *)regstack->ga_data + regstack->ga_len) - 1;
|
|
5036 *scan = rp->rs_scan;
|
|
5037 *startp = rp->rs_startp;
|
|
5038
|
|
5039 regstack->ga_len -= sizeof(regitem_T);
|
7
|
5040 }
|
|
5041
|
|
5042 /*
|
|
5043 * regrepeat - repeatedly match something simple, return how many.
|
|
5044 * Advances reginput (and reglnum) to just after the matched chars.
|
|
5045 */
|
|
5046 static int
|
|
5047 regrepeat(p, maxcount)
|
|
5048 char_u *p;
|
|
5049 long maxcount; /* maximum number of matches allowed */
|
|
5050 {
|
|
5051 long count = 0;
|
|
5052 char_u *scan;
|
|
5053 char_u *opnd;
|
|
5054 int mask;
|
|
5055 int testval = 0;
|
|
5056
|
|
5057 scan = reginput; /* Make local copy of reginput for speed. */
|
|
5058 opnd = OPERAND(p);
|
|
5059 switch (OP(p))
|
|
5060 {
|
|
5061 case ANY:
|
|
5062 case ANY + ADD_NL:
|
|
5063 while (count < maxcount)
|
|
5064 {
|
|
5065 /* Matching anything means we continue until end-of-line (or
|
|
5066 * end-of-file for ANY + ADD_NL), only limited by maxcount. */
|
|
5067 while (*scan != NUL && count < maxcount)
|
|
5068 {
|
|
5069 ++count;
|
39
|
5070 mb_ptr_adv(scan);
|
7
|
5071 }
|
|
5072 if (!WITH_NL(OP(p)) || reglnum == reg_maxline || count == maxcount)
|
|
5073 break;
|
|
5074 ++count; /* count the line-break */
|
|
5075 reg_nextline();
|
|
5076 scan = reginput;
|
|
5077 if (got_int)
|
|
5078 break;
|
|
5079 }
|
|
5080 break;
|
|
5081
|
|
5082 case IDENT:
|
|
5083 case IDENT + ADD_NL:
|
|
5084 testval = TRUE;
|
|
5085 /*FALLTHROUGH*/
|
|
5086 case SIDENT:
|
|
5087 case SIDENT + ADD_NL:
|
|
5088 while (count < maxcount)
|
|
5089 {
|
|
5090 if (vim_isIDc(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
5091 {
|
39
|
5092 mb_ptr_adv(scan);
|
7
|
5093 }
|
|
5094 else if (*scan == NUL)
|
|
5095 {
|
|
5096 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
5097 break;
|
|
5098 reg_nextline();
|
|
5099 scan = reginput;
|
|
5100 if (got_int)
|
|
5101 break;
|
|
5102 }
|
|
5103 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5104 ++scan;
|
|
5105 else
|
|
5106 break;
|
|
5107 ++count;
|
|
5108 }
|
|
5109 break;
|
|
5110
|
|
5111 case KWORD:
|
|
5112 case KWORD + ADD_NL:
|
|
5113 testval = TRUE;
|
|
5114 /*FALLTHROUGH*/
|
|
5115 case SKWORD:
|
|
5116 case SKWORD + ADD_NL:
|
|
5117 while (count < maxcount)
|
|
5118 {
|
|
5119 if (vim_iswordp(scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
5120 {
|
39
|
5121 mb_ptr_adv(scan);
|
7
|
5122 }
|
|
5123 else if (*scan == NUL)
|
|
5124 {
|
|
5125 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
5126 break;
|
|
5127 reg_nextline();
|
|
5128 scan = reginput;
|
|
5129 if (got_int)
|
|
5130 break;
|
|
5131 }
|
|
5132 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5133 ++scan;
|
|
5134 else
|
|
5135 break;
|
|
5136 ++count;
|
|
5137 }
|
|
5138 break;
|
|
5139
|
|
5140 case FNAME:
|
|
5141 case FNAME + ADD_NL:
|
|
5142 testval = TRUE;
|
|
5143 /*FALLTHROUGH*/
|
|
5144 case SFNAME:
|
|
5145 case SFNAME + ADD_NL:
|
|
5146 while (count < maxcount)
|
|
5147 {
|
|
5148 if (vim_isfilec(*scan) && (testval || !VIM_ISDIGIT(*scan)))
|
|
5149 {
|
39
|
5150 mb_ptr_adv(scan);
|
7
|
5151 }
|
|
5152 else if (*scan == NUL)
|
|
5153 {
|
|
5154 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
5155 break;
|
|
5156 reg_nextline();
|
|
5157 scan = reginput;
|
|
5158 if (got_int)
|
|
5159 break;
|
|
5160 }
|
|
5161 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5162 ++scan;
|
|
5163 else
|
|
5164 break;
|
|
5165 ++count;
|
|
5166 }
|
|
5167 break;
|
|
5168
|
|
5169 case PRINT:
|
|
5170 case PRINT + ADD_NL:
|
|
5171 testval = TRUE;
|
|
5172 /*FALLTHROUGH*/
|
|
5173 case SPRINT:
|
|
5174 case SPRINT + ADD_NL:
|
|
5175 while (count < maxcount)
|
|
5176 {
|
|
5177 if (*scan == NUL)
|
|
5178 {
|
|
5179 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
5180 break;
|
|
5181 reg_nextline();
|
|
5182 scan = reginput;
|
|
5183 if (got_int)
|
|
5184 break;
|
|
5185 }
|
|
5186 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan)))
|
|
5187 {
|
39
|
5188 mb_ptr_adv(scan);
|
7
|
5189 }
|
|
5190 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5191 ++scan;
|
|
5192 else
|
|
5193 break;
|
|
5194 ++count;
|
|
5195 }
|
|
5196 break;
|
|
5197
|
|
5198 case WHITE:
|
|
5199 case WHITE + ADD_NL:
|
|
5200 testval = mask = RI_WHITE;
|
|
5201 do_class:
|
|
5202 while (count < maxcount)
|
|
5203 {
|
|
5204 #ifdef FEAT_MBYTE
|
|
5205 int l;
|
|
5206 #endif
|
|
5207 if (*scan == NUL)
|
|
5208 {
|
|
5209 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
5210 break;
|
|
5211 reg_nextline();
|
|
5212 scan = reginput;
|
|
5213 if (got_int)
|
|
5214 break;
|
|
5215 }
|
|
5216 #ifdef FEAT_MBYTE
|
|
5217 else if (has_mbyte && (l = (*mb_ptr2len_check)(scan)) > 1)
|
|
5218 {
|
|
5219 if (testval != 0)
|
|
5220 break;
|
|
5221 scan += l;
|
|
5222 }
|
|
5223 #endif
|
|
5224 else if ((class_tab[*scan] & mask) == testval)
|
|
5225 ++scan;
|
|
5226 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5227 ++scan;
|
|
5228 else
|
|
5229 break;
|
|
5230 ++count;
|
|
5231 }
|
|
5232 break;
|
|
5233
|
|
5234 case NWHITE:
|
|
5235 case NWHITE + ADD_NL:
|
|
5236 mask = RI_WHITE;
|
|
5237 goto do_class;
|
|
5238 case DIGIT:
|
|
5239 case DIGIT + ADD_NL:
|
|
5240 testval = mask = RI_DIGIT;
|
|
5241 goto do_class;
|
|
5242 case NDIGIT:
|
|
5243 case NDIGIT + ADD_NL:
|
|
5244 mask = RI_DIGIT;
|
|
5245 goto do_class;
|
|
5246 case HEX:
|
|
5247 case HEX + ADD_NL:
|
|
5248 testval = mask = RI_HEX;
|
|
5249 goto do_class;
|
|
5250 case NHEX:
|
|
5251 case NHEX + ADD_NL:
|
|
5252 mask = RI_HEX;
|
|
5253 goto do_class;
|
|
5254 case OCTAL:
|
|
5255 case OCTAL + ADD_NL:
|
|
5256 testval = mask = RI_OCTAL;
|
|
5257 goto do_class;
|
|
5258 case NOCTAL:
|
|
5259 case NOCTAL + ADD_NL:
|
|
5260 mask = RI_OCTAL;
|
|
5261 goto do_class;
|
|
5262 case WORD:
|
|
5263 case WORD + ADD_NL:
|
|
5264 testval = mask = RI_WORD;
|
|
5265 goto do_class;
|
|
5266 case NWORD:
|
|
5267 case NWORD + ADD_NL:
|
|
5268 mask = RI_WORD;
|
|
5269 goto do_class;
|
|
5270 case HEAD:
|
|
5271 case HEAD + ADD_NL:
|
|
5272 testval = mask = RI_HEAD;
|
|
5273 goto do_class;
|
|
5274 case NHEAD:
|
|
5275 case NHEAD + ADD_NL:
|
|
5276 mask = RI_HEAD;
|
|
5277 goto do_class;
|
|
5278 case ALPHA:
|
|
5279 case ALPHA + ADD_NL:
|
|
5280 testval = mask = RI_ALPHA;
|
|
5281 goto do_class;
|
|
5282 case NALPHA:
|
|
5283 case NALPHA + ADD_NL:
|
|
5284 mask = RI_ALPHA;
|
|
5285 goto do_class;
|
|
5286 case LOWER:
|
|
5287 case LOWER + ADD_NL:
|
|
5288 testval = mask = RI_LOWER;
|
|
5289 goto do_class;
|
|
5290 case NLOWER:
|
|
5291 case NLOWER + ADD_NL:
|
|
5292 mask = RI_LOWER;
|
|
5293 goto do_class;
|
|
5294 case UPPER:
|
|
5295 case UPPER + ADD_NL:
|
|
5296 testval = mask = RI_UPPER;
|
|
5297 goto do_class;
|
|
5298 case NUPPER:
|
|
5299 case NUPPER + ADD_NL:
|
|
5300 mask = RI_UPPER;
|
|
5301 goto do_class;
|
|
5302
|
|
5303 case EXACTLY:
|
|
5304 {
|
|
5305 int cu, cl;
|
|
5306
|
|
5307 /* This doesn't do a multi-byte character, because a MULTIBYTECODE
|
|
5308 * would have been used for it. */
|
|
5309 if (ireg_ic)
|
|
5310 {
|
|
5311 cu = TOUPPER_LOC(*opnd);
|
|
5312 cl = TOLOWER_LOC(*opnd);
|
|
5313 while (count < maxcount && (*scan == cu || *scan == cl))
|
|
5314 {
|
|
5315 count++;
|
|
5316 scan++;
|
|
5317 }
|
|
5318 }
|
|
5319 else
|
|
5320 {
|
|
5321 cu = *opnd;
|
|
5322 while (count < maxcount && *scan == cu)
|
|
5323 {
|
|
5324 count++;
|
|
5325 scan++;
|
|
5326 }
|
|
5327 }
|
|
5328 break;
|
|
5329 }
|
|
5330
|
|
5331 #ifdef FEAT_MBYTE
|
|
5332 case MULTIBYTECODE:
|
|
5333 {
|
|
5334 int i, len, cf = 0;
|
|
5335
|
|
5336 /* Safety check (just in case 'encoding' was changed since
|
|
5337 * compiling the program). */
|
|
5338 if ((len = (*mb_ptr2len_check)(opnd)) > 1)
|
|
5339 {
|
|
5340 if (ireg_ic && enc_utf8)
|
|
5341 cf = utf_fold(utf_ptr2char(opnd));
|
|
5342 while (count < maxcount)
|
|
5343 {
|
|
5344 for (i = 0; i < len; ++i)
|
|
5345 if (opnd[i] != scan[i])
|
|
5346 break;
|
|
5347 if (i < len && (!ireg_ic || !enc_utf8
|
|
5348 || utf_fold(utf_ptr2char(scan)) != cf))
|
|
5349 break;
|
|
5350 scan += len;
|
|
5351 ++count;
|
|
5352 }
|
|
5353 }
|
|
5354 }
|
|
5355 break;
|
|
5356 #endif
|
|
5357
|
|
5358 case ANYOF:
|
|
5359 case ANYOF + ADD_NL:
|
|
5360 testval = TRUE;
|
|
5361 /*FALLTHROUGH*/
|
|
5362
|
|
5363 case ANYBUT:
|
|
5364 case ANYBUT + ADD_NL:
|
|
5365 while (count < maxcount)
|
|
5366 {
|
|
5367 #ifdef FEAT_MBYTE
|
|
5368 int len;
|
|
5369 #endif
|
|
5370 if (*scan == NUL)
|
|
5371 {
|
|
5372 if (!WITH_NL(OP(p)) || reglnum == reg_maxline)
|
|
5373 break;
|
|
5374 reg_nextline();
|
|
5375 scan = reginput;
|
|
5376 if (got_int)
|
|
5377 break;
|
|
5378 }
|
|
5379 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
|
|
5380 ++scan;
|
|
5381 #ifdef FEAT_MBYTE
|
|
5382 else if (has_mbyte && (len = (*mb_ptr2len_check)(scan)) > 1)
|
|
5383 {
|
|
5384 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
|
|
5385 break;
|
|
5386 scan += len;
|
|
5387 }
|
|
5388 #endif
|
|
5389 else
|
|
5390 {
|
|
5391 if ((cstrchr(opnd, *scan) == NULL) == testval)
|
|
5392 break;
|
|
5393 ++scan;
|
|
5394 }
|
|
5395 ++count;
|
|
5396 }
|
|
5397 break;
|
|
5398
|
|
5399 case NEWL:
|
|
5400 while (count < maxcount
|
|
5401 && ((*scan == NUL && reglnum < reg_maxline)
|
|
5402 || (*scan == '\n' && reg_line_lbr)))
|
|
5403 {
|
|
5404 count++;
|
|
5405 if (reg_line_lbr)
|
|
5406 ADVANCE_REGINPUT();
|
|
5407 else
|
|
5408 reg_nextline();
|
|
5409 scan = reginput;
|
|
5410 if (got_int)
|
|
5411 break;
|
|
5412 }
|
|
5413 break;
|
|
5414
|
|
5415 default: /* Oh dear. Called inappropriately. */
|
|
5416 EMSG(_(e_re_corr));
|
|
5417 #ifdef DEBUG
|
|
5418 printf("Called regrepeat with op code %d\n", OP(p));
|
|
5419 #endif
|
|
5420 break;
|
|
5421 }
|
|
5422
|
|
5423 reginput = scan;
|
|
5424
|
|
5425 return (int)count;
|
|
5426 }
|
|
5427
|
|
5428 /*
|
|
5429 * regnext - dig the "next" pointer out of a node
|
|
5430 */
|
|
5431 static char_u *
|
|
5432 regnext(p)
|
|
5433 char_u *p;
|
|
5434 {
|
|
5435 int offset;
|
|
5436
|
|
5437 if (p == JUST_CALC_SIZE)
|
|
5438 return NULL;
|
|
5439
|
|
5440 offset = NEXT(p);
|
|
5441 if (offset == 0)
|
|
5442 return NULL;
|
|
5443
|
179
|
5444 if (OP(p) == BACK || OP(p) == BACKP)
|
7
|
5445 return p - offset;
|
|
5446 else
|
|
5447 return p + offset;
|
|
5448 }
|
|
5449
|
|
5450 /*
|
|
5451 * Check the regexp program for its magic number.
|
|
5452 * Return TRUE if it's wrong.
|
|
5453 */
|
|
5454 static int
|
|
5455 prog_magic_wrong()
|
|
5456 {
|
|
5457 if (UCHARAT(REG_MULTI
|
|
5458 ? reg_mmatch->regprog->program
|
|
5459 : reg_match->regprog->program) != REGMAGIC)
|
|
5460 {
|
|
5461 EMSG(_(e_re_corr));
|
|
5462 return TRUE;
|
|
5463 }
|
|
5464 return FALSE;
|
|
5465 }
|
|
5466
|
|
5467 /*
|
|
5468 * Cleanup the subexpressions, if this wasn't done yet.
|
|
5469 * This construction is used to clear the subexpressions only when they are
|
|
5470 * used (to increase speed).
|
|
5471 */
|
|
5472 static void
|
|
5473 cleanup_subexpr()
|
|
5474 {
|
|
5475 if (need_clear_subexpr)
|
|
5476 {
|
|
5477 if (REG_MULTI)
|
|
5478 {
|
|
5479 /* Use 0xff to set lnum to -1 */
|
|
5480 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5481 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5482 }
|
|
5483 else
|
|
5484 {
|
|
5485 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5486 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5487 }
|
|
5488 need_clear_subexpr = FALSE;
|
|
5489 }
|
|
5490 }
|
|
5491
|
|
5492 #ifdef FEAT_SYN_HL
|
|
5493 static void
|
|
5494 cleanup_zsubexpr()
|
|
5495 {
|
|
5496 if (need_clear_zsubexpr)
|
|
5497 {
|
|
5498 if (REG_MULTI)
|
|
5499 {
|
|
5500 /* Use 0xff to set lnum to -1 */
|
|
5501 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5502 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
|
|
5503 }
|
|
5504 else
|
|
5505 {
|
|
5506 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5507 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP);
|
|
5508 }
|
|
5509 need_clear_zsubexpr = FALSE;
|
|
5510 }
|
|
5511 }
|
|
5512 #endif
|
|
5513
|
|
5514 /*
|
|
5515 * Advance reglnum, regline and reginput to the next line.
|
|
5516 */
|
|
5517 static void
|
|
5518 reg_nextline()
|
|
5519 {
|
|
5520 regline = reg_getline(++reglnum);
|
|
5521 reginput = regline;
|
|
5522 fast_breakcheck();
|
|
5523 }
|
|
5524
|
|
5525 /*
|
|
5526 * Save the input line and position in a regsave_T.
|
|
5527 */
|
|
5528 static void
|
|
5529 reg_save(save)
|
|
5530 regsave_T *save;
|
|
5531 {
|
|
5532 if (REG_MULTI)
|
|
5533 {
|
|
5534 save->rs_u.pos.col = (colnr_T)(reginput - regline);
|
|
5535 save->rs_u.pos.lnum = reglnum;
|
|
5536 }
|
|
5537 else
|
|
5538 save->rs_u.ptr = reginput;
|
|
5539 }
|
|
5540
|
|
5541 /*
|
|
5542 * Restore the input line and position from a regsave_T.
|
|
5543 */
|
|
5544 static void
|
|
5545 reg_restore(save)
|
|
5546 regsave_T *save;
|
|
5547 {
|
|
5548 if (REG_MULTI)
|
|
5549 {
|
|
5550 if (reglnum != save->rs_u.pos.lnum)
|
|
5551 {
|
|
5552 /* only call reg_getline() when the line number changed to save
|
|
5553 * a bit of time */
|
|
5554 reglnum = save->rs_u.pos.lnum;
|
|
5555 regline = reg_getline(reglnum);
|
|
5556 }
|
|
5557 reginput = regline + save->rs_u.pos.col;
|
|
5558 }
|
|
5559 else
|
|
5560 reginput = save->rs_u.ptr;
|
|
5561 }
|
|
5562
|
|
5563 /*
|
|
5564 * Return TRUE if current position is equal to saved position.
|
|
5565 */
|
|
5566 static int
|
|
5567 reg_save_equal(save)
|
|
5568 regsave_T *save;
|
|
5569 {
|
|
5570 if (REG_MULTI)
|
|
5571 return reglnum == save->rs_u.pos.lnum
|
|
5572 && reginput == regline + save->rs_u.pos.col;
|
|
5573 return reginput == save->rs_u.ptr;
|
|
5574 }
|
|
5575
|
|
5576 /*
|
|
5577 * Tentatively set the sub-expression start to the current position (after
|
|
5578 * calling regmatch() they will have changed). Need to save the existing
|
|
5579 * values for when there is no match.
|
|
5580 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
|
|
5581 * depending on REG_MULTI.
|
|
5582 */
|
|
5583 static void
|
|
5584 save_se_multi(savep, posp)
|
|
5585 save_se_T *savep;
|
|
5586 lpos_T *posp;
|
|
5587 {
|
|
5588 savep->se_u.pos = *posp;
|
|
5589 posp->lnum = reglnum;
|
|
5590 posp->col = (colnr_T)(reginput - regline);
|
|
5591 }
|
|
5592
|
|
5593 static void
|
|
5594 save_se_one(savep, pp)
|
|
5595 save_se_T *savep;
|
|
5596 char_u **pp;
|
|
5597 {
|
|
5598 savep->se_u.ptr = *pp;
|
|
5599 *pp = reginput;
|
|
5600 }
|
|
5601
|
|
5602 /*
|
|
5603 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
|
|
5604 */
|
|
5605 static int
|
|
5606 re_num_cmp(val, scan)
|
|
5607 long_u val;
|
|
5608 char_u *scan;
|
|
5609 {
|
|
5610 long_u n = OPERAND_MIN(scan);
|
|
5611
|
|
5612 if (OPERAND_CMP(scan) == '>')
|
|
5613 return val > n;
|
|
5614 if (OPERAND_CMP(scan) == '<')
|
|
5615 return val < n;
|
|
5616 return val == n;
|
|
5617 }
|
|
5618
|
|
5619
|
|
5620 #ifdef DEBUG
|
|
5621
|
|
5622 /*
|
|
5623 * regdump - dump a regexp onto stdout in vaguely comprehensible form
|
|
5624 */
|
|
5625 static void
|
|
5626 regdump(pattern, r)
|
|
5627 char_u *pattern;
|
|
5628 regprog_T *r;
|
|
5629 {
|
|
5630 char_u *s;
|
|
5631 int op = EXACTLY; /* Arbitrary non-END op. */
|
|
5632 char_u *next;
|
|
5633 char_u *end = NULL;
|
|
5634
|
|
5635 printf("\r\nregcomp(%s):\r\n", pattern);
|
|
5636
|
|
5637 s = r->program + 1;
|
|
5638 /*
|
|
5639 * Loop until we find the END that isn't before a referred next (an END
|
|
5640 * can also appear in a NOMATCH operand).
|
|
5641 */
|
|
5642 while (op != END || s <= end)
|
|
5643 {
|
|
5644 op = OP(s);
|
|
5645 printf("%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */
|
|
5646 next = regnext(s);
|
|
5647 if (next == NULL) /* Next ptr. */
|
|
5648 printf("(0)");
|
|
5649 else
|
|
5650 printf("(%d)", (int)((s - r->program) + (next - s)));
|
|
5651 if (end < next)
|
|
5652 end = next;
|
|
5653 if (op == BRACE_LIMITS)
|
|
5654 {
|
|
5655 /* Two short ints */
|
|
5656 printf(" minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
|
|
5657 s += 8;
|
|
5658 }
|
|
5659 s += 3;
|
|
5660 if (op == ANYOF || op == ANYOF + ADD_NL
|
|
5661 || op == ANYBUT || op == ANYBUT + ADD_NL
|
|
5662 || op == EXACTLY)
|
|
5663 {
|
|
5664 /* Literal string, where present. */
|
|
5665 while (*s != NUL)
|
|
5666 printf("%c", *s++);
|
|
5667 s++;
|
|
5668 }
|
|
5669 printf("\r\n");
|
|
5670 }
|
|
5671
|
|
5672 /* Header fields of interest. */
|
|
5673 if (r->regstart != NUL)
|
|
5674 printf("start `%s' 0x%x; ", r->regstart < 256
|
|
5675 ? (char *)transchar(r->regstart)
|
|
5676 : "multibyte", r->regstart);
|
|
5677 if (r->reganch)
|
|
5678 printf("anchored; ");
|
|
5679 if (r->regmust != NULL)
|
|
5680 printf("must have \"%s\"", r->regmust);
|
|
5681 printf("\r\n");
|
|
5682 }
|
|
5683
|
|
5684 /*
|
|
5685 * regprop - printable representation of opcode
|
|
5686 */
|
|
5687 static char_u *
|
|
5688 regprop(op)
|
|
5689 char_u *op;
|
|
5690 {
|
|
5691 char_u *p;
|
|
5692 static char_u buf[50];
|
|
5693
|
|
5694 (void) strcpy(buf, ":");
|
|
5695
|
|
5696 switch (OP(op))
|
|
5697 {
|
|
5698 case BOL:
|
|
5699 p = "BOL";
|
|
5700 break;
|
|
5701 case EOL:
|
|
5702 p = "EOL";
|
|
5703 break;
|
|
5704 case RE_BOF:
|
|
5705 p = "BOF";
|
|
5706 break;
|
|
5707 case RE_EOF:
|
|
5708 p = "EOF";
|
|
5709 break;
|
|
5710 case CURSOR:
|
|
5711 p = "CURSOR";
|
|
5712 break;
|
|
5713 case RE_LNUM:
|
|
5714 p = "RE_LNUM";
|
|
5715 break;
|
|
5716 case RE_COL:
|
|
5717 p = "RE_COL";
|
|
5718 break;
|
|
5719 case RE_VCOL:
|
|
5720 p = "RE_VCOL";
|
|
5721 break;
|
|
5722 case BOW:
|
|
5723 p = "BOW";
|
|
5724 break;
|
|
5725 case EOW:
|
|
5726 p = "EOW";
|
|
5727 break;
|
|
5728 case ANY:
|
|
5729 p = "ANY";
|
|
5730 break;
|
|
5731 case ANY + ADD_NL:
|
|
5732 p = "ANY+NL";
|
|
5733 break;
|
|
5734 case ANYOF:
|
|
5735 p = "ANYOF";
|
|
5736 break;
|
|
5737 case ANYOF + ADD_NL:
|
|
5738 p = "ANYOF+NL";
|
|
5739 break;
|
|
5740 case ANYBUT:
|
|
5741 p = "ANYBUT";
|
|
5742 break;
|
|
5743 case ANYBUT + ADD_NL:
|
|
5744 p = "ANYBUT+NL";
|
|
5745 break;
|
|
5746 case IDENT:
|
|
5747 p = "IDENT";
|
|
5748 break;
|
|
5749 case IDENT + ADD_NL:
|
|
5750 p = "IDENT+NL";
|
|
5751 break;
|
|
5752 case SIDENT:
|
|
5753 p = "SIDENT";
|
|
5754 break;
|
|
5755 case SIDENT + ADD_NL:
|
|
5756 p = "SIDENT+NL";
|
|
5757 break;
|
|
5758 case KWORD:
|
|
5759 p = "KWORD";
|
|
5760 break;
|
|
5761 case KWORD + ADD_NL:
|
|
5762 p = "KWORD+NL";
|
|
5763 break;
|
|
5764 case SKWORD:
|
|
5765 p = "SKWORD";
|
|
5766 break;
|
|
5767 case SKWORD + ADD_NL:
|
|
5768 p = "SKWORD+NL";
|
|
5769 break;
|
|
5770 case FNAME:
|
|
5771 p = "FNAME";
|
|
5772 break;
|
|
5773 case FNAME + ADD_NL:
|
|
5774 p = "FNAME+NL";
|
|
5775 break;
|
|
5776 case SFNAME:
|
|
5777 p = "SFNAME";
|
|
5778 break;
|
|
5779 case SFNAME + ADD_NL:
|
|
5780 p = "SFNAME+NL";
|
|
5781 break;
|
|
5782 case PRINT:
|
|
5783 p = "PRINT";
|
|
5784 break;
|
|
5785 case PRINT + ADD_NL:
|
|
5786 p = "PRINT+NL";
|
|
5787 break;
|
|
5788 case SPRINT:
|
|
5789 p = "SPRINT";
|
|
5790 break;
|
|
5791 case SPRINT + ADD_NL:
|
|
5792 p = "SPRINT+NL";
|
|
5793 break;
|
|
5794 case WHITE:
|
|
5795 p = "WHITE";
|
|
5796 break;
|
|
5797 case WHITE + ADD_NL:
|
|
5798 p = "WHITE+NL";
|
|
5799 break;
|
|
5800 case NWHITE:
|
|
5801 p = "NWHITE";
|
|
5802 break;
|
|
5803 case NWHITE + ADD_NL:
|
|
5804 p = "NWHITE+NL";
|
|
5805 break;
|
|
5806 case DIGIT:
|
|
5807 p = "DIGIT";
|
|
5808 break;
|
|
5809 case DIGIT + ADD_NL:
|
|
5810 p = "DIGIT+NL";
|
|
5811 break;
|
|
5812 case NDIGIT:
|
|
5813 p = "NDIGIT";
|
|
5814 break;
|
|
5815 case NDIGIT + ADD_NL:
|
|
5816 p = "NDIGIT+NL";
|
|
5817 break;
|
|
5818 case HEX:
|
|
5819 p = "HEX";
|
|
5820 break;
|
|
5821 case HEX + ADD_NL:
|
|
5822 p = "HEX+NL";
|
|
5823 break;
|
|
5824 case NHEX:
|
|
5825 p = "NHEX";
|
|
5826 break;
|
|
5827 case NHEX + ADD_NL:
|
|
5828 p = "NHEX+NL";
|
|
5829 break;
|
|
5830 case OCTAL:
|
|
5831 p = "OCTAL";
|
|
5832 break;
|
|
5833 case OCTAL + ADD_NL:
|
|
5834 p = "OCTAL+NL";
|
|
5835 break;
|
|
5836 case NOCTAL:
|
|
5837 p = "NOCTAL";
|
|
5838 break;
|
|
5839 case NOCTAL + ADD_NL:
|
|
5840 p = "NOCTAL+NL";
|
|
5841 break;
|
|
5842 case WORD:
|
|
5843 p = "WORD";
|
|
5844 break;
|
|
5845 case WORD + ADD_NL:
|
|
5846 p = "WORD+NL";
|
|
5847 break;
|
|
5848 case NWORD:
|
|
5849 p = "NWORD";
|
|
5850 break;
|
|
5851 case NWORD + ADD_NL:
|
|
5852 p = "NWORD+NL";
|
|
5853 break;
|
|
5854 case HEAD:
|
|
5855 p = "HEAD";
|
|
5856 break;
|
|
5857 case HEAD + ADD_NL:
|
|
5858 p = "HEAD+NL";
|
|
5859 break;
|
|
5860 case NHEAD:
|
|
5861 p = "NHEAD";
|
|
5862 break;
|
|
5863 case NHEAD + ADD_NL:
|
|
5864 p = "NHEAD+NL";
|
|
5865 break;
|
|
5866 case ALPHA:
|
|
5867 p = "ALPHA";
|
|
5868 break;
|
|
5869 case ALPHA + ADD_NL:
|
|
5870 p = "ALPHA+NL";
|
|
5871 break;
|
|
5872 case NALPHA:
|
|
5873 p = "NALPHA";
|
|
5874 break;
|
|
5875 case NALPHA + ADD_NL:
|
|
5876 p = "NALPHA+NL";
|
|
5877 break;
|
|
5878 case LOWER:
|
|
5879 p = "LOWER";
|
|
5880 break;
|
|
5881 case LOWER + ADD_NL:
|
|
5882 p = "LOWER+NL";
|
|
5883 break;
|
|
5884 case NLOWER:
|
|
5885 p = "NLOWER";
|
|
5886 break;
|
|
5887 case NLOWER + ADD_NL:
|
|
5888 p = "NLOWER+NL";
|
|
5889 break;
|
|
5890 case UPPER:
|
|
5891 p = "UPPER";
|
|
5892 break;
|
|
5893 case UPPER + ADD_NL:
|
|
5894 p = "UPPER+NL";
|
|
5895 break;
|
|
5896 case NUPPER:
|
|
5897 p = "NUPPER";
|
|
5898 break;
|
|
5899 case NUPPER + ADD_NL:
|
|
5900 p = "NUPPER+NL";
|
|
5901 break;
|
|
5902 case BRANCH:
|
|
5903 p = "BRANCH";
|
|
5904 break;
|
|
5905 case EXACTLY:
|
|
5906 p = "EXACTLY";
|
|
5907 break;
|
|
5908 case NOTHING:
|
|
5909 p = "NOTHING";
|
|
5910 break;
|
|
5911 case BACK:
|
|
5912 p = "BACK";
|
|
5913 break;
|
179
|
5914 case BACKP:
|
|
5915 p = "BACKP";
|
|
5916 break;
|
7
|
5917 case END:
|
|
5918 p = "END";
|
|
5919 break;
|
|
5920 case MOPEN + 0:
|
|
5921 p = "MATCH START";
|
|
5922 break;
|
|
5923 case MOPEN + 1:
|
|
5924 case MOPEN + 2:
|
|
5925 case MOPEN + 3:
|
|
5926 case MOPEN + 4:
|
|
5927 case MOPEN + 5:
|
|
5928 case MOPEN + 6:
|
|
5929 case MOPEN + 7:
|
|
5930 case MOPEN + 8:
|
|
5931 case MOPEN + 9:
|
|
5932 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
|
|
5933 p = NULL;
|
|
5934 break;
|
|
5935 case MCLOSE + 0:
|
|
5936 p = "MATCH END";
|
|
5937 break;
|
|
5938 case MCLOSE + 1:
|
|
5939 case MCLOSE + 2:
|
|
5940 case MCLOSE + 3:
|
|
5941 case MCLOSE + 4:
|
|
5942 case MCLOSE + 5:
|
|
5943 case MCLOSE + 6:
|
|
5944 case MCLOSE + 7:
|
|
5945 case MCLOSE + 8:
|
|
5946 case MCLOSE + 9:
|
|
5947 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
|
|
5948 p = NULL;
|
|
5949 break;
|
|
5950 case BACKREF + 1:
|
|
5951 case BACKREF + 2:
|
|
5952 case BACKREF + 3:
|
|
5953 case BACKREF + 4:
|
|
5954 case BACKREF + 5:
|
|
5955 case BACKREF + 6:
|
|
5956 case BACKREF + 7:
|
|
5957 case BACKREF + 8:
|
|
5958 case BACKREF + 9:
|
|
5959 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
|
|
5960 p = NULL;
|
|
5961 break;
|
|
5962 case NOPEN:
|
|
5963 p = "NOPEN";
|
|
5964 break;
|
|
5965 case NCLOSE:
|
|
5966 p = "NCLOSE";
|
|
5967 break;
|
|
5968 #ifdef FEAT_SYN_HL
|
|
5969 case ZOPEN + 1:
|
|
5970 case ZOPEN + 2:
|
|
5971 case ZOPEN + 3:
|
|
5972 case ZOPEN + 4:
|
|
5973 case ZOPEN + 5:
|
|
5974 case ZOPEN + 6:
|
|
5975 case ZOPEN + 7:
|
|
5976 case ZOPEN + 8:
|
|
5977 case ZOPEN + 9:
|
|
5978 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
|
|
5979 p = NULL;
|
|
5980 break;
|
|
5981 case ZCLOSE + 1:
|
|
5982 case ZCLOSE + 2:
|
|
5983 case ZCLOSE + 3:
|
|
5984 case ZCLOSE + 4:
|
|
5985 case ZCLOSE + 5:
|
|
5986 case ZCLOSE + 6:
|
|
5987 case ZCLOSE + 7:
|
|
5988 case ZCLOSE + 8:
|
|
5989 case ZCLOSE + 9:
|
|
5990 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
|
|
5991 p = NULL;
|
|
5992 break;
|
|
5993 case ZREF + 1:
|
|
5994 case ZREF + 2:
|
|
5995 case ZREF + 3:
|
|
5996 case ZREF + 4:
|
|
5997 case ZREF + 5:
|
|
5998 case ZREF + 6:
|
|
5999 case ZREF + 7:
|
|
6000 case ZREF + 8:
|
|
6001 case ZREF + 9:
|
|
6002 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
|
|
6003 p = NULL;
|
|
6004 break;
|
|
6005 #endif
|
|
6006 case STAR:
|
|
6007 p = "STAR";
|
|
6008 break;
|
|
6009 case PLUS:
|
|
6010 p = "PLUS";
|
|
6011 break;
|
|
6012 case NOMATCH:
|
|
6013 p = "NOMATCH";
|
|
6014 break;
|
|
6015 case MATCH:
|
|
6016 p = "MATCH";
|
|
6017 break;
|
|
6018 case BEHIND:
|
|
6019 p = "BEHIND";
|
|
6020 break;
|
|
6021 case NOBEHIND:
|
|
6022 p = "NOBEHIND";
|
|
6023 break;
|
|
6024 case SUBPAT:
|
|
6025 p = "SUBPAT";
|
|
6026 break;
|
|
6027 case BRACE_LIMITS:
|
|
6028 p = "BRACE_LIMITS";
|
|
6029 break;
|
|
6030 case BRACE_SIMPLE:
|
|
6031 p = "BRACE_SIMPLE";
|
|
6032 break;
|
|
6033 case BRACE_COMPLEX + 0:
|
|
6034 case BRACE_COMPLEX + 1:
|
|
6035 case BRACE_COMPLEX + 2:
|
|
6036 case BRACE_COMPLEX + 3:
|
|
6037 case BRACE_COMPLEX + 4:
|
|
6038 case BRACE_COMPLEX + 5:
|
|
6039 case BRACE_COMPLEX + 6:
|
|
6040 case BRACE_COMPLEX + 7:
|
|
6041 case BRACE_COMPLEX + 8:
|
|
6042 case BRACE_COMPLEX + 9:
|
|
6043 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
|
|
6044 p = NULL;
|
|
6045 break;
|
|
6046 #ifdef FEAT_MBYTE
|
|
6047 case MULTIBYTECODE:
|
|
6048 p = "MULTIBYTECODE";
|
|
6049 break;
|
|
6050 #endif
|
|
6051 case NEWL:
|
|
6052 p = "NEWL";
|
|
6053 break;
|
|
6054 default:
|
|
6055 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
|
|
6056 p = NULL;
|
|
6057 break;
|
|
6058 }
|
|
6059 if (p != NULL)
|
|
6060 (void) strcat(buf, p);
|
|
6061 return buf;
|
|
6062 }
|
|
6063 #endif
|
|
6064
|
|
6065 #ifdef FEAT_MBYTE
|
|
6066 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3));
|
|
6067
|
|
6068 typedef struct
|
|
6069 {
|
|
6070 int a, b, c;
|
|
6071 } decomp_T;
|
|
6072
|
|
6073
|
|
6074 /* 0xfb20 - 0xfb4f */
|
|
6075 decomp_T decomp_table[0xfb4f-0xfb20+1] =
|
|
6076 {
|
|
6077 {0x5e2,0,0}, /* 0xfb20 alt ayin */
|
|
6078 {0x5d0,0,0}, /* 0xfb21 alt alef */
|
|
6079 {0x5d3,0,0}, /* 0xfb22 alt dalet */
|
|
6080 {0x5d4,0,0}, /* 0xfb23 alt he */
|
|
6081 {0x5db,0,0}, /* 0xfb24 alt kaf */
|
|
6082 {0x5dc,0,0}, /* 0xfb25 alt lamed */
|
|
6083 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */
|
|
6084 {0x5e8,0,0}, /* 0xfb27 alt resh */
|
|
6085 {0x5ea,0,0}, /* 0xfb28 alt tav */
|
|
6086 {'+', 0, 0}, /* 0xfb29 alt plus */
|
|
6087 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */
|
|
6088 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */
|
|
6089 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */
|
|
6090 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */
|
|
6091 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */
|
|
6092 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */
|
|
6093 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */
|
|
6094 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */
|
|
6095 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */
|
|
6096 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */
|
|
6097 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */
|
|
6098 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */
|
|
6099 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */
|
|
6100 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */
|
|
6101 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */
|
|
6102 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */
|
|
6103 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */
|
|
6104 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */
|
|
6105 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */
|
|
6106 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */
|
|
6107 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */
|
|
6108 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */
|
|
6109 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */
|
|
6110 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */
|
|
6111 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */
|
|
6112 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */
|
|
6113 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */
|
|
6114 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */
|
|
6115 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */
|
|
6116 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */
|
|
6117 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */
|
|
6118 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */
|
|
6119 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */
|
|
6120 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */
|
|
6121 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */
|
|
6122 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */
|
|
6123 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */
|
|
6124 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */
|
|
6125 };
|
|
6126
|
|
6127 static void
|
|
6128 mb_decompose(c, c1, c2, c3)
|
|
6129 int c, *c1, *c2, *c3;
|
|
6130 {
|
|
6131 decomp_T d;
|
|
6132
|
|
6133 if (c >= 0x4b20 && c <= 0xfb4f)
|
|
6134 {
|
|
6135 d = decomp_table[c - 0xfb20];
|
|
6136 *c1 = d.a;
|
|
6137 *c2 = d.b;
|
|
6138 *c3 = d.c;
|
|
6139 }
|
|
6140 else
|
|
6141 {
|
|
6142 *c1 = c;
|
|
6143 *c2 = *c3 = 0;
|
|
6144 }
|
|
6145 }
|
|
6146 #endif
|
|
6147
|
|
6148 /*
|
|
6149 * Compare two strings, ignore case if ireg_ic set.
|
|
6150 * Return 0 if strings match, non-zero otherwise.
|
|
6151 * Correct the length "*n" when composing characters are ignored.
|
|
6152 */
|
|
6153 static int
|
|
6154 cstrncmp(s1, s2, n)
|
|
6155 char_u *s1, *s2;
|
|
6156 int *n;
|
|
6157 {
|
|
6158 int result;
|
|
6159
|
|
6160 if (!ireg_ic)
|
|
6161 result = STRNCMP(s1, s2, *n);
|
|
6162 else
|
|
6163 result = MB_STRNICMP(s1, s2, *n);
|
|
6164
|
|
6165 #ifdef FEAT_MBYTE
|
|
6166 /* if it failed and it's utf8 and we want to combineignore: */
|
|
6167 if (result != 0 && enc_utf8 && ireg_icombine)
|
|
6168 {
|
|
6169 char_u *str1, *str2;
|
|
6170 int c1, c2, c11, c12;
|
|
6171 int ix;
|
|
6172 int junk;
|
|
6173
|
|
6174 /* we have to handle the strcmp ourselves, since it is necessary to
|
|
6175 * deal with the composing characters by ignoring them: */
|
|
6176 str1 = s1;
|
|
6177 str2 = s2;
|
|
6178 c1 = c2 = 0;
|
|
6179 for (ix = 0; ix < *n; )
|
|
6180 {
|
|
6181 c1 = mb_ptr2char_adv(&str1);
|
|
6182 c2 = mb_ptr2char_adv(&str2);
|
|
6183 ix += utf_char2len(c1);
|
|
6184
|
|
6185 /* decompose the character if necessary, into 'base' characters
|
|
6186 * because I don't care about Arabic, I will hard-code the Hebrew
|
|
6187 * which I *do* care about! So sue me... */
|
|
6188 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2)))
|
|
6189 {
|
|
6190 /* decomposition necessary? */
|
|
6191 mb_decompose(c1, &c11, &junk, &junk);
|
|
6192 mb_decompose(c2, &c12, &junk, &junk);
|
|
6193 c1 = c11;
|
|
6194 c2 = c12;
|
|
6195 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12)))
|
|
6196 break;
|
|
6197 }
|
|
6198 }
|
|
6199 result = c2 - c1;
|
|
6200 if (result == 0)
|
|
6201 *n = (int)(str2 - s2);
|
|
6202 }
|
|
6203 #endif
|
|
6204
|
|
6205 return result;
|
|
6206 }
|
|
6207
|
|
6208 /*
|
|
6209 * cstrchr: This function is used a lot for simple searches, keep it fast!
|
|
6210 */
|
|
6211 static char_u *
|
|
6212 cstrchr(s, c)
|
|
6213 char_u *s;
|
|
6214 int c;
|
|
6215 {
|
|
6216 char_u *p;
|
|
6217 int cc;
|
|
6218
|
|
6219 if (!ireg_ic
|
|
6220 #ifdef FEAT_MBYTE
|
|
6221 || (!enc_utf8 && mb_char2len(c) > 1)
|
|
6222 #endif
|
|
6223 )
|
|
6224 return vim_strchr(s, c);
|
|
6225
|
|
6226 /* tolower() and toupper() can be slow, comparing twice should be a lot
|
|
6227 * faster (esp. when using MS Visual C++!).
|
|
6228 * For UTF-8 need to use folded case. */
|
|
6229 #ifdef FEAT_MBYTE
|
|
6230 if (enc_utf8 && c > 0x80)
|
|
6231 cc = utf_fold(c);
|
|
6232 else
|
|
6233 #endif
|
|
6234 if (isupper(c))
|
|
6235 cc = TOLOWER_LOC(c);
|
|
6236 else if (islower(c))
|
|
6237 cc = TOUPPER_LOC(c);
|
|
6238 else
|
|
6239 return vim_strchr(s, c);
|
|
6240
|
|
6241 #ifdef FEAT_MBYTE
|
|
6242 if (has_mbyte)
|
|
6243 {
|
|
6244 for (p = s; *p != NUL; p += (*mb_ptr2len_check)(p))
|
|
6245 {
|
|
6246 if (enc_utf8 && c > 0x80)
|
|
6247 {
|
|
6248 if (utf_fold(utf_ptr2char(p)) == cc)
|
|
6249 return p;
|
|
6250 }
|
|
6251 else if (*p == c || *p == cc)
|
|
6252 return p;
|
|
6253 }
|
|
6254 }
|
|
6255 else
|
|
6256 #endif
|
|
6257 /* Faster version for when there are no multi-byte characters. */
|
|
6258 for (p = s; *p != NUL; ++p)
|
|
6259 if (*p == c || *p == cc)
|
|
6260 return p;
|
|
6261
|
|
6262 return NULL;
|
|
6263 }
|
|
6264
|
|
6265 /***************************************************************
|
|
6266 * regsub stuff *
|
|
6267 ***************************************************************/
|
|
6268
|
|
6269 /* This stuff below really confuses cc on an SGI -- webb */
|
|
6270 #ifdef __sgi
|
|
6271 # undef __ARGS
|
|
6272 # define __ARGS(x) ()
|
|
6273 #endif
|
|
6274
|
|
6275 /*
|
|
6276 * We should define ftpr as a pointer to a function returning a pointer to
|
|
6277 * a function returning a pointer to a function ...
|
|
6278 * This is impossible, so we declare a pointer to a function returning a
|
|
6279 * pointer to a function returning void. This should work for all compilers.
|
|
6280 */
|
|
6281 typedef void (*(*fptr) __ARGS((char_u *, int)))();
|
|
6282
|
|
6283 static fptr do_upper __ARGS((char_u *, int));
|
|
6284 static fptr do_Upper __ARGS((char_u *, int));
|
|
6285 static fptr do_lower __ARGS((char_u *, int));
|
|
6286 static fptr do_Lower __ARGS((char_u *, int));
|
|
6287
|
|
6288 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash));
|
|
6289
|
|
6290 static fptr
|
|
6291 do_upper(d, c)
|
|
6292 char_u *d;
|
|
6293 int c;
|
|
6294 {
|
|
6295 *d = TOUPPER_LOC(c);
|
|
6296
|
|
6297 return (fptr)NULL;
|
|
6298 }
|
|
6299
|
|
6300 static fptr
|
|
6301 do_Upper(d, c)
|
|
6302 char_u *d;
|
|
6303 int c;
|
|
6304 {
|
|
6305 *d = TOUPPER_LOC(c);
|
|
6306
|
|
6307 return (fptr)do_Upper;
|
|
6308 }
|
|
6309
|
|
6310 static fptr
|
|
6311 do_lower(d, c)
|
|
6312 char_u *d;
|
|
6313 int c;
|
|
6314 {
|
|
6315 *d = TOLOWER_LOC(c);
|
|
6316
|
|
6317 return (fptr)NULL;
|
|
6318 }
|
|
6319
|
|
6320 static fptr
|
|
6321 do_Lower(d, c)
|
|
6322 char_u *d;
|
|
6323 int c;
|
|
6324 {
|
|
6325 *d = TOLOWER_LOC(c);
|
|
6326
|
|
6327 return (fptr)do_Lower;
|
|
6328 }
|
|
6329
|
|
6330 /*
|
|
6331 * regtilde(): Replace tildes in the pattern by the old pattern.
|
|
6332 *
|
|
6333 * Short explanation of the tilde: It stands for the previous replacement
|
|
6334 * pattern. If that previous pattern also contains a ~ we should go back a
|
|
6335 * step further... But we insert the previous pattern into the current one
|
|
6336 * and remember that.
|
|
6337 * This still does not handle the case where "magic" changes. TODO?
|
|
6338 *
|
|
6339 * The tildes are parsed once before the first call to vim_regsub().
|
|
6340 */
|
|
6341 char_u *
|
|
6342 regtilde(source, magic)
|
|
6343 char_u *source;
|
|
6344 int magic;
|
|
6345 {
|
|
6346 char_u *newsub = source;
|
|
6347 char_u *tmpsub;
|
|
6348 char_u *p;
|
|
6349 int len;
|
|
6350 int prevlen;
|
|
6351
|
|
6352 for (p = newsub; *p; ++p)
|
|
6353 {
|
|
6354 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic))
|
|
6355 {
|
|
6356 if (reg_prev_sub != NULL)
|
|
6357 {
|
|
6358 /* length = len(newsub) - 1 + len(prev_sub) + 1 */
|
|
6359 prevlen = (int)STRLEN(reg_prev_sub);
|
|
6360 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen));
|
|
6361 if (tmpsub != NULL)
|
|
6362 {
|
|
6363 /* copy prefix */
|
|
6364 len = (int)(p - newsub); /* not including ~ */
|
|
6365 mch_memmove(tmpsub, newsub, (size_t)len);
|
|
6366 /* interpretate tilde */
|
|
6367 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen);
|
|
6368 /* copy postfix */
|
|
6369 if (!magic)
|
|
6370 ++p; /* back off \ */
|
|
6371 STRCPY(tmpsub + len + prevlen, p + 1);
|
|
6372
|
|
6373 if (newsub != source) /* already allocated newsub */
|
|
6374 vim_free(newsub);
|
|
6375 newsub = tmpsub;
|
|
6376 p = newsub + len + prevlen;
|
|
6377 }
|
|
6378 }
|
|
6379 else if (magic)
|
|
6380 STRCPY(p, p + 1); /* remove '~' */
|
|
6381 else
|
|
6382 STRCPY(p, p + 2); /* remove '\~' */
|
|
6383 --p;
|
|
6384 }
|
|
6385 else
|
|
6386 {
|
|
6387 if (*p == '\\' && p[1]) /* skip escaped characters */
|
|
6388 ++p;
|
|
6389 #ifdef FEAT_MBYTE
|
|
6390 if (has_mbyte)
|
|
6391 p += (*mb_ptr2len_check)(p) - 1;
|
|
6392 #endif
|
|
6393 }
|
|
6394 }
|
|
6395
|
|
6396 vim_free(reg_prev_sub);
|
|
6397 if (newsub != source) /* newsub was allocated, just keep it */
|
|
6398 reg_prev_sub = newsub;
|
|
6399 else /* no ~ found, need to save newsub */
|
|
6400 reg_prev_sub = vim_strsave(newsub);
|
|
6401 return newsub;
|
|
6402 }
|
|
6403
|
|
6404 #ifdef FEAT_EVAL
|
|
6405 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */
|
|
6406
|
|
6407 /* These pointers are used instead of reg_match and reg_mmatch for
|
|
6408 * reg_submatch(). Needed for when the substitution string is an expression
|
|
6409 * that contains a call to substitute() and submatch(). */
|
|
6410 static regmatch_T *submatch_match;
|
|
6411 static regmmatch_T *submatch_mmatch;
|
|
6412 #endif
|
|
6413
|
|
6414 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO)
|
|
6415 /*
|
|
6416 * vim_regsub() - perform substitutions after a vim_regexec() or
|
|
6417 * vim_regexec_multi() match.
|
|
6418 *
|
|
6419 * If "copy" is TRUE really copy into "dest".
|
|
6420 * If "copy" is FALSE nothing is copied, this is just to find out the length
|
|
6421 * of the result.
|
|
6422 *
|
|
6423 * If "backslash" is TRUE, a backslash will be removed later, need to double
|
|
6424 * them to keep them, and insert a backslash before a CR to avoid it being
|
|
6425 * replaced with a line break later.
|
|
6426 *
|
|
6427 * Note: The matched text must not change between the call of
|
|
6428 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back
|
|
6429 * references invalid!
|
|
6430 *
|
|
6431 * Returns the size of the replacement, including terminating NUL.
|
|
6432 */
|
|
6433 int
|
|
6434 vim_regsub(rmp, source, dest, copy, magic, backslash)
|
|
6435 regmatch_T *rmp;
|
|
6436 char_u *source;
|
|
6437 char_u *dest;
|
|
6438 int copy;
|
|
6439 int magic;
|
|
6440 int backslash;
|
|
6441 {
|
|
6442 reg_match = rmp;
|
|
6443 reg_mmatch = NULL;
|
|
6444 reg_maxline = 0;
|
|
6445 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
6446 }
|
|
6447 #endif
|
|
6448
|
|
6449 int
|
|
6450 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash)
|
|
6451 regmmatch_T *rmp;
|
|
6452 linenr_T lnum;
|
|
6453 char_u *source;
|
|
6454 char_u *dest;
|
|
6455 int copy;
|
|
6456 int magic;
|
|
6457 int backslash;
|
|
6458 {
|
|
6459 reg_match = NULL;
|
|
6460 reg_mmatch = rmp;
|
|
6461 reg_buf = curbuf; /* always works on the current buffer! */
|
|
6462 reg_firstlnum = lnum;
|
|
6463 reg_maxline = curbuf->b_ml.ml_line_count - lnum;
|
|
6464 return vim_regsub_both(source, dest, copy, magic, backslash);
|
|
6465 }
|
|
6466
|
|
6467 static int
|
|
6468 vim_regsub_both(source, dest, copy, magic, backslash)
|
|
6469 char_u *source;
|
|
6470 char_u *dest;
|
|
6471 int copy;
|
|
6472 int magic;
|
|
6473 int backslash;
|
|
6474 {
|
|
6475 char_u *src;
|
|
6476 char_u *dst;
|
|
6477 char_u *s;
|
|
6478 int c;
|
|
6479 int no = -1;
|
|
6480 fptr func = (fptr)NULL;
|
|
6481 linenr_T clnum = 0; /* init for GCC */
|
|
6482 int len = 0; /* init for GCC */
|
|
6483 #ifdef FEAT_EVAL
|
|
6484 static char_u *eval_result = NULL;
|
|
6485 #endif
|
|
6486 #ifdef FEAT_MBYTE
|
|
6487 int l;
|
|
6488 #endif
|
|
6489
|
|
6490
|
|
6491 /* Be paranoid... */
|
|
6492 if (source == NULL || dest == NULL)
|
|
6493 {
|
|
6494 EMSG(_(e_null));
|
|
6495 return 0;
|
|
6496 }
|
|
6497 if (prog_magic_wrong())
|
|
6498 return 0;
|
|
6499 src = source;
|
|
6500 dst = dest;
|
|
6501
|
|
6502 /*
|
|
6503 * When the substitute part starts with "\=" evaluate it as an expression.
|
|
6504 */
|
|
6505 if (source[0] == '\\' && source[1] == '='
|
|
6506 #ifdef FEAT_EVAL
|
|
6507 && !can_f_submatch /* can't do this recursively */
|
|
6508 #endif
|
|
6509 )
|
|
6510 {
|
|
6511 #ifdef FEAT_EVAL
|
|
6512 /* To make sure that the length doesn't change between checking the
|
|
6513 * length and copying the string, and to speed up things, the
|
|
6514 * resulting string is saved from the call with "copy" == FALSE to the
|
|
6515 * call with "copy" == TRUE. */
|
|
6516 if (copy)
|
|
6517 {
|
|
6518 if (eval_result != NULL)
|
|
6519 {
|
|
6520 STRCPY(dest, eval_result);
|
|
6521 dst += STRLEN(eval_result);
|
|
6522 vim_free(eval_result);
|
|
6523 eval_result = NULL;
|
|
6524 }
|
|
6525 }
|
|
6526 else
|
|
6527 {
|
|
6528 linenr_T save_reg_maxline;
|
|
6529 win_T *save_reg_win;
|
|
6530 int save_ireg_ic;
|
|
6531
|
|
6532 vim_free(eval_result);
|
|
6533
|
|
6534 /* The expression may contain substitute(), which calls us
|
|
6535 * recursively. Make sure submatch() gets the text from the first
|
|
6536 * level. Don't need to save "reg_buf", because
|
|
6537 * vim_regexec_multi() can't be called recursively. */
|
|
6538 submatch_match = reg_match;
|
|
6539 submatch_mmatch = reg_mmatch;
|
|
6540 save_reg_maxline = reg_maxline;
|
|
6541 save_reg_win = reg_win;
|
|
6542 save_ireg_ic = ireg_ic;
|
|
6543 can_f_submatch = TRUE;
|
|
6544
|
|
6545 eval_result = eval_to_string(source + 2, NULL);
|
|
6546 if (eval_result != NULL)
|
|
6547 {
|
39
|
6548 for (s = eval_result; *s != NUL; mb_ptr_adv(s))
|
7
|
6549 {
|
|
6550 /* Change NL to CR, so that it becomes a line break.
|
|
6551 * Skip over a backslashed character. */
|
|
6552 if (*s == NL)
|
|
6553 *s = CAR;
|
|
6554 else if (*s == '\\' && s[1] != NUL)
|
|
6555 ++s;
|
|
6556 }
|
|
6557
|
|
6558 dst += STRLEN(eval_result);
|
|
6559 }
|
|
6560
|
|
6561 reg_match = submatch_match;
|
|
6562 reg_mmatch = submatch_mmatch;
|
|
6563 reg_maxline = save_reg_maxline;
|
|
6564 reg_win = save_reg_win;
|
|
6565 ireg_ic = save_ireg_ic;
|
|
6566 can_f_submatch = FALSE;
|
|
6567 }
|
|
6568 #endif
|
|
6569 }
|
|
6570 else
|
|
6571 while ((c = *src++) != NUL)
|
|
6572 {
|
|
6573 if (c == '&' && magic)
|
|
6574 no = 0;
|
|
6575 else if (c == '\\' && *src != NUL)
|
|
6576 {
|
|
6577 if (*src == '&' && !magic)
|
|
6578 {
|
|
6579 ++src;
|
|
6580 no = 0;
|
|
6581 }
|
|
6582 else if ('0' <= *src && *src <= '9')
|
|
6583 {
|
|
6584 no = *src++ - '0';
|
|
6585 }
|
|
6586 else if (vim_strchr((char_u *)"uUlLeE", *src))
|
|
6587 {
|
|
6588 switch (*src++)
|
|
6589 {
|
|
6590 case 'u': func = (fptr)do_upper;
|
|
6591 continue;
|
|
6592 case 'U': func = (fptr)do_Upper;
|
|
6593 continue;
|
|
6594 case 'l': func = (fptr)do_lower;
|
|
6595 continue;
|
|
6596 case 'L': func = (fptr)do_Lower;
|
|
6597 continue;
|
|
6598 case 'e':
|
|
6599 case 'E': func = (fptr)NULL;
|
|
6600 continue;
|
|
6601 }
|
|
6602 }
|
|
6603 }
|
|
6604 if (no < 0) /* Ordinary character. */
|
|
6605 {
|
|
6606 if (c == '\\' && *src != NUL)
|
|
6607 {
|
|
6608 /* Check for abbreviations -- webb */
|
|
6609 switch (*src)
|
|
6610 {
|
|
6611 case 'r': c = CAR; ++src; break;
|
|
6612 case 'n': c = NL; ++src; break;
|
|
6613 case 't': c = TAB; ++src; break;
|
|
6614 /* Oh no! \e already has meaning in subst pat :-( */
|
|
6615 /* case 'e': c = ESC; ++src; break; */
|
|
6616 case 'b': c = Ctrl_H; ++src; break;
|
|
6617
|
|
6618 /* If "backslash" is TRUE the backslash will be removed
|
|
6619 * later. Used to insert a literal CR. */
|
|
6620 default: if (backslash)
|
|
6621 {
|
|
6622 if (copy)
|
|
6623 *dst = '\\';
|
|
6624 ++dst;
|
|
6625 }
|
|
6626 c = *src++;
|
|
6627 }
|
|
6628 }
|
|
6629
|
|
6630 /* Write to buffer, if copy is set. */
|
|
6631 #ifdef FEAT_MBYTE
|
|
6632 if (has_mbyte && (l = (*mb_ptr2len_check)(src - 1)) > 1)
|
|
6633 {
|
|
6634 /* TODO: should use "func" here. */
|
|
6635 if (copy)
|
|
6636 mch_memmove(dst, src - 1, l);
|
|
6637 dst += l - 1;
|
|
6638 src += l - 1;
|
|
6639 }
|
|
6640 else
|
|
6641 {
|
|
6642 #endif
|
|
6643 if (copy)
|
|
6644 {
|
|
6645 if (func == (fptr)NULL) /* just copy */
|
|
6646 *dst = c;
|
|
6647 else /* change case */
|
|
6648 func = (fptr)(func(dst, c));
|
|
6649 /* Turbo C complains without the typecast */
|
|
6650 }
|
|
6651 #ifdef FEAT_MBYTE
|
|
6652 }
|
|
6653 #endif
|
|
6654 dst++;
|
|
6655 }
|
|
6656 else
|
|
6657 {
|
|
6658 if (REG_MULTI)
|
|
6659 {
|
|
6660 clnum = reg_mmatch->startpos[no].lnum;
|
|
6661 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0)
|
|
6662 s = NULL;
|
|
6663 else
|
|
6664 {
|
|
6665 s = reg_getline(clnum) + reg_mmatch->startpos[no].col;
|
|
6666 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6667 len = reg_mmatch->endpos[no].col
|
|
6668 - reg_mmatch->startpos[no].col;
|
|
6669 else
|
|
6670 len = (int)STRLEN(s);
|
|
6671 }
|
|
6672 }
|
|
6673 else
|
|
6674 {
|
|
6675 s = reg_match->startp[no];
|
|
6676 if (reg_match->endp[no] == NULL)
|
|
6677 s = NULL;
|
|
6678 else
|
|
6679 len = (int)(reg_match->endp[no] - s);
|
|
6680 }
|
|
6681 if (s != NULL)
|
|
6682 {
|
|
6683 for (;;)
|
|
6684 {
|
|
6685 if (len == 0)
|
|
6686 {
|
|
6687 if (REG_MULTI)
|
|
6688 {
|
|
6689 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6690 break;
|
|
6691 if (copy)
|
|
6692 *dst = CAR;
|
|
6693 ++dst;
|
|
6694 s = reg_getline(++clnum);
|
|
6695 if (reg_mmatch->endpos[no].lnum == clnum)
|
|
6696 len = reg_mmatch->endpos[no].col;
|
|
6697 else
|
|
6698 len = (int)STRLEN(s);
|
|
6699 }
|
|
6700 else
|
|
6701 break;
|
|
6702 }
|
|
6703 else if (*s == NUL) /* we hit NUL. */
|
|
6704 {
|
|
6705 if (copy)
|
|
6706 EMSG(_(e_re_damg));
|
|
6707 goto exit;
|
|
6708 }
|
|
6709 else
|
|
6710 {
|
|
6711 if (backslash && (*s == CAR || *s == '\\'))
|
|
6712 {
|
|
6713 /*
|
|
6714 * Insert a backslash in front of a CR, otherwise
|
|
6715 * it will be replaced by a line break.
|
|
6716 * Number of backslashes will be halved later,
|
|
6717 * double them here.
|
|
6718 */
|
|
6719 if (copy)
|
|
6720 {
|
|
6721 dst[0] = '\\';
|
|
6722 dst[1] = *s;
|
|
6723 }
|
|
6724 dst += 2;
|
|
6725 }
|
|
6726 #ifdef FEAT_MBYTE
|
|
6727 else if (has_mbyte && (l = (*mb_ptr2len_check)(s)) > 1)
|
|
6728 {
|
|
6729 /* TODO: should use "func" here. */
|
|
6730 if (copy)
|
|
6731 mch_memmove(dst, s, l);
|
|
6732 dst += l;
|
|
6733 s += l - 1;
|
|
6734 len -= l - 1;
|
|
6735 }
|
|
6736 #endif
|
|
6737 else
|
|
6738 {
|
|
6739 if (copy)
|
|
6740 {
|
|
6741 if (func == (fptr)NULL) /* just copy */
|
|
6742 *dst = *s;
|
|
6743 else /* change case */
|
|
6744 func = (fptr)(func(dst, *s));
|
|
6745 /* Turbo C complains without the typecast */
|
|
6746 }
|
|
6747 ++dst;
|
|
6748 }
|
|
6749 ++s;
|
|
6750 --len;
|
|
6751 }
|
|
6752 }
|
|
6753 }
|
|
6754 no = -1;
|
|
6755 }
|
|
6756 }
|
|
6757 if (copy)
|
|
6758 *dst = NUL;
|
|
6759
|
|
6760 exit:
|
|
6761 return (int)((dst - dest) + 1);
|
|
6762 }
|
|
6763
|
|
6764 #ifdef FEAT_EVAL
|
|
6765 /*
|
|
6766 * Used for the submatch() function: get the string from tne n'th submatch in
|
|
6767 * allocated memory.
|
|
6768 * Returns NULL when not in a ":s" command and for a non-existing submatch.
|
|
6769 */
|
|
6770 char_u *
|
|
6771 reg_submatch(no)
|
|
6772 int no;
|
|
6773 {
|
|
6774 char_u *retval = NULL;
|
|
6775 char_u *s;
|
|
6776 int len;
|
|
6777 int round;
|
|
6778 linenr_T lnum;
|
|
6779
|
|
6780 if (!can_f_submatch)
|
|
6781 return NULL;
|
|
6782
|
|
6783 if (submatch_match == NULL)
|
|
6784 {
|
|
6785 /*
|
|
6786 * First round: compute the length and allocate memory.
|
|
6787 * Second round: copy the text.
|
|
6788 */
|
|
6789 for (round = 1; round <= 2; ++round)
|
|
6790 {
|
|
6791 lnum = submatch_mmatch->startpos[no].lnum;
|
|
6792 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0)
|
|
6793 return NULL;
|
|
6794
|
|
6795 s = reg_getline(lnum) + submatch_mmatch->startpos[no].col;
|
|
6796 if (s == NULL) /* anti-crash check, cannot happen? */
|
|
6797 break;
|
|
6798 if (submatch_mmatch->endpos[no].lnum == lnum)
|
|
6799 {
|
|
6800 /* Within one line: take form start to end col. */
|
|
6801 len = submatch_mmatch->endpos[no].col
|
|
6802 - submatch_mmatch->startpos[no].col;
|
|
6803 if (round == 2)
|
|
6804 {
|
|
6805 STRNCPY(retval, s, len);
|
|
6806 retval[len] = NUL;
|
|
6807 }
|
|
6808 ++len;
|
|
6809 }
|
|
6810 else
|
|
6811 {
|
|
6812 /* Multiple lines: take start line from start col, middle
|
|
6813 * lines completely and end line up to end col. */
|
|
6814 len = (int)STRLEN(s);
|
|
6815 if (round == 2)
|
|
6816 {
|
|
6817 STRCPY(retval, s);
|
|
6818 retval[len] = '\n';
|
|
6819 }
|
|
6820 ++len;
|
|
6821 ++lnum;
|
|
6822 while (lnum < submatch_mmatch->endpos[no].lnum)
|
|
6823 {
|
|
6824 s = reg_getline(lnum++);
|
|
6825 if (round == 2)
|
|
6826 STRCPY(retval + len, s);
|
|
6827 len += (int)STRLEN(s);
|
|
6828 if (round == 2)
|
|
6829 retval[len] = '\n';
|
|
6830 ++len;
|
|
6831 }
|
|
6832 if (round == 2)
|
|
6833 STRNCPY(retval + len, reg_getline(lnum),
|
|
6834 submatch_mmatch->endpos[no].col);
|
|
6835 len += submatch_mmatch->endpos[no].col;
|
|
6836 if (round == 2)
|
|
6837 retval[len] = NUL;
|
|
6838 ++len;
|
|
6839 }
|
|
6840
|
|
6841 if (round == 1)
|
|
6842 {
|
|
6843 retval = lalloc((long_u)len, TRUE);
|
|
6844 if (s == NULL)
|
|
6845 return NULL;
|
|
6846 }
|
|
6847 }
|
|
6848 }
|
|
6849 else
|
|
6850 {
|
|
6851 if (submatch_match->endp[no] == NULL)
|
|
6852 retval = NULL;
|
|
6853 else
|
|
6854 {
|
|
6855 s = submatch_match->startp[no];
|
|
6856 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s));
|
|
6857 }
|
|
6858 }
|
|
6859
|
|
6860 return retval;
|
|
6861 }
|
|
6862 #endif
|