blob: 4082f59d35cc31dd9968152af42c05a933ee999e [file] [log] [blame]
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001/* vi:set ts=8 sts=4 sw=4 noet:
2 *
3 * Backtracking regular expression implementation.
4 *
5 * This file is included in "regexp.c".
6 *
7 * NOTICE:
8 *
9 * This is NOT the original regular expression code as written by Henry
10 * Spencer. This code has been modified specifically for use with the VIM
11 * editor, and should not be used separately from Vim. If you want a good
12 * regular expression library, get the original code. The copyright notice
13 * that follows is from the original.
14 *
15 * END NOTICE
16 *
17 * Copyright (c) 1986 by University of Toronto.
18 * Written by Henry Spencer. Not derived from licensed software.
19 *
20 * Permission is granted to anyone to use this software for any
21 * purpose on any computer system, and to redistribute it freely,
22 * subject to the following restrictions:
23 *
24 * 1. The author is not responsible for the consequences of use of
25 * this software, no matter how awful, even if they arise
26 * from defects in it.
27 *
28 * 2. The origin of this software must not be misrepresented, either
29 * by explicit claim or by omission.
30 *
31 * 3. Altered versions must be plainly marked as such, and must not
32 * be misrepresented as being the original software.
33 *
34 * Beware that some of this code is subtly aware of the way operator
35 * precedence is structured in regular expressions. Serious changes in
36 * regular-expression syntax might require a total rethink.
37 *
38 * Changes have been made by Tony Andrews, Olaf 'Rhialto' Seibert, Robert
39 * Webb, Ciaran McCreesh and Bram Moolenaar.
40 * Named character class support added by Walter Briscoe (1998 Jul 01)
41 */
42
43/*
44 * The "internal use only" fields in regexp.h are present to pass info from
45 * compile to execute that permits the execute phase to run lots faster on
46 * simple cases. They are:
47 *
48 * regstart char that must begin a match; NUL if none obvious; Can be a
49 * multi-byte character.
50 * reganch is the match anchored (at beginning-of-line only)?
51 * regmust string (pointer into program) that match must include, or NULL
52 * regmlen length of regmust string
53 * regflags RF_ values or'ed together
54 *
55 * Regstart and reganch permit very fast decisions on suitable starting points
56 * for a match, cutting down the work a lot. Regmust permits fast rejection
57 * of lines that cannot possibly match. The regmust tests are costly enough
58 * that vim_regcomp() supplies a regmust only if the r.e. contains something
59 * potentially expensive (at present, the only such thing detected is * or +
60 * at the start of the r.e., which can involve a lot of backup). Regmlen is
61 * supplied because the test in vim_regexec() needs it and vim_regcomp() is
62 * computing it anyway.
63 */
64
65/*
66 * Structure for regexp "program". This is essentially a linear encoding
67 * of a nondeterministic finite-state machine (aka syntax charts or
68 * "railroad normal form" in parsing technology). Each node is an opcode
69 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
70 * all nodes except BRANCH and BRACES_COMPLEX implement concatenation; a "next"
71 * pointer with a BRANCH on both ends of it is connecting two alternatives.
72 * (Here we have one of the subtle syntax dependencies: an individual BRANCH
73 * (as opposed to a collection of them) is never concatenated with anything
74 * because of operator precedence). The "next" pointer of a BRACES_COMPLEX
75 * node points to the node after the stuff to be repeated.
76 * The operand of some types of node is a literal string; for others, it is a
77 * node leading into a sub-FSM. In particular, the operand of a BRANCH node
78 * is the first node of the branch.
79 * (NB this is *not* a tree structure: the tail of the branch connects to the
80 * thing following the set of BRANCHes.)
81 *
82 * pattern is coded like:
83 *
84 * +-----------------+
85 * | V
86 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END
87 * | ^ | ^
88 * +------+ +----------+
89 *
90 *
91 * +------------------+
92 * V |
93 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END
94 * | | ^ ^
95 * | +---------------+ |
96 * +---------------------------------------------+
97 *
98 *
99 * +----------------------+
100 * V |
101 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END
102 * | | ^ ^
103 * | +-----------+ |
104 * +--------------------------------------------------+
105 *
106 *
107 * +-------------------------+
108 * V |
109 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END
110 * | | ^
111 * | +----------------+
112 * +-----------------------------------------------+
113 *
114 *
115 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END
116 * | | ^ ^
117 * | +----------------+ |
118 * +--------------------------------+
119 *
120 * +---------+
121 * | V
122 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END
123 * | | | | ^ ^
124 * | | | +-----+ |
125 * | | +----------------+ |
126 * | +---------------------------+ |
127 * +------------------------------------------------------+
128 *
129 * They all start with a BRANCH for "\|" alternatives, even when there is only
130 * one alternative.
131 */
132
133/*
134 * The opcodes are:
135 */
136
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200137// definition number opnd? meaning
138#define END 0 // End of program or NOMATCH operand.
139#define BOL 1 // Match "" at beginning of line.
140#define EOL 2 // Match "" at end of line.
141#define BRANCH 3 // node Match this alternative, or the
142 // next...
143#define BACK 4 // Match "", "next" ptr points backward.
144#define EXACTLY 5 // str Match this string.
145#define NOTHING 6 // Match empty string.
146#define STAR 7 // node Match this (simple) thing 0 or more
147 // times.
148#define PLUS 8 // node Match this (simple) thing 1 or more
149 // times.
150#define MATCH 9 // node match the operand zero-width
151#define NOMATCH 10 // node check for no match with operand
152#define BEHIND 11 // node look behind for a match with operand
153#define NOBEHIND 12 // node look behind for no match with operand
154#define SUBPAT 13 // node match the operand here
155#define BRACE_SIMPLE 14 // node Match this (simple) thing between m and
156 // n times (\{m,n\}).
157#define BOW 15 // Match "" after [^a-zA-Z0-9_]
158#define EOW 16 // Match "" at [^a-zA-Z0-9_]
159#define BRACE_LIMITS 17 // nr nr define the min & max for BRACE_SIMPLE
160 // and BRACE_COMPLEX.
161#define NEWL 18 // Match line-break
162#define BHPOS 19 // End position for BEHIND or NOBEHIND
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200163
164
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200165// character classes: 20-48 normal, 50-78 include a line-break
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200166#define ADD_NL 30
167#define FIRST_NL ANY + ADD_NL
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200168#define ANY 20 // Match any one character.
169#define ANYOF 21 // str Match any character in this string.
170#define ANYBUT 22 // str Match any character not in this
171 // string.
172#define IDENT 23 // Match identifier char
173#define SIDENT 24 // Match identifier char but no digit
174#define KWORD 25 // Match keyword char
175#define SKWORD 26 // Match word char but no digit
176#define FNAME 27 // Match file name char
177#define SFNAME 28 // Match file name char but no digit
178#define PRINT 29 // Match printable char
179#define SPRINT 30 // Match printable char but no digit
180#define WHITE 31 // Match whitespace char
181#define NWHITE 32 // Match non-whitespace char
182#define DIGIT 33 // Match digit char
183#define NDIGIT 34 // Match non-digit char
184#define HEX 35 // Match hex char
185#define NHEX 36 // Match non-hex char
186#define OCTAL 37 // Match octal char
187#define NOCTAL 38 // Match non-octal char
188#define WORD 39 // Match word char
189#define NWORD 40 // Match non-word char
190#define HEAD 41 // Match head char
191#define NHEAD 42 // Match non-head char
192#define ALPHA 43 // Match alpha char
193#define NALPHA 44 // Match non-alpha char
194#define LOWER 45 // Match lowercase char
195#define NLOWER 46 // Match non-lowercase char
196#define UPPER 47 // Match uppercase char
197#define NUPPER 48 // Match non-uppercase char
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200198#define LAST_NL NUPPER + ADD_NL
199#define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL)
200
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200201#define MOPEN 80 // -89 Mark this point in input as start of
202 // \( subexpr. MOPEN + 0 marks start of
203 // match.
204#define MCLOSE 90 // -99 Analogous to MOPEN. MCLOSE + 0 marks
205 // end of match.
206#define BACKREF 100 // -109 node Match same string again \1-\9
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200207
208#ifdef FEAT_SYN_HL
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200209# define ZOPEN 110 // -119 Mark this point in input as start of
210 // \z( subexpr.
211# define ZCLOSE 120 // -129 Analogous to ZOPEN.
212# define ZREF 130 // -139 node Match external submatch \z1-\z9
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200213#endif
214
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200215#define BRACE_COMPLEX 140 // -149 node Match nodes between m & n times
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200216
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200217#define NOPEN 150 // Mark this point in input as start of
218 // \%( subexpr.
219#define NCLOSE 151 // Analogous to NOPEN.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200220
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200221#define MULTIBYTECODE 200 // mbc Match one multi-byte character
222#define RE_BOF 201 // Match "" at beginning of file.
223#define RE_EOF 202 // Match "" at end of file.
224#define CURSOR 203 // Match location of cursor.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200225
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200226#define RE_LNUM 204 // nr cmp Match line number
227#define RE_COL 205 // nr cmp Match column number
228#define RE_VCOL 206 // nr cmp Match virtual column number
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200229
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200230#define RE_MARK 207 // mark cmp Match mark position
231#define RE_VISUAL 208 // Match Visual area
232#define RE_COMPOSING 209 // any composing characters
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200233
234/*
235 * Flags to be passed up and down.
236 */
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200237#define HASWIDTH 0x1 // Known never to match null string.
238#define SIMPLE 0x2 // Simple enough to be STAR/PLUS operand.
239#define SPSTART 0x4 // Starts with * or +.
240#define HASNL 0x8 // Contains some \n.
241#define HASLOOKBH 0x10 // Contains "\@<=" or "\@<!".
242#define WORST 0 // Worst case.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200243
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200244static int num_complex_braces; // Complex \{...} count
245static char_u *regcode; // Code-emit pointer, or JUST_CALC_SIZE
246static long regsize; // Code size.
247static int reg_toolong; // TRUE when offset out of range
248static char_u had_endbrace[NSUBEXP]; // flags, TRUE if end of () found
249static long brace_min[10]; // Minimums for complex brace repeats
250static long brace_max[10]; // Maximums for complex brace repeats
251static int brace_count[10]; // Current counts for complex brace repeats
252static int one_exactly = FALSE; // only do one char for EXACTLY
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200253
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200254// When making changes to classchars also change nfa_classcodes.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200255static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU";
256static int classcodes[] = {
257 ANY, IDENT, SIDENT, KWORD, SKWORD,
258 FNAME, SFNAME, PRINT, SPRINT,
259 WHITE, NWHITE, DIGIT, NDIGIT,
260 HEX, NHEX, OCTAL, NOCTAL,
261 WORD, NWORD, HEAD, NHEAD,
262 ALPHA, NALPHA, LOWER, NLOWER,
263 UPPER, NUPPER
264};
265
266/*
267 * When regcode is set to this value, code is not emitted and size is computed
268 * instead.
269 */
270#define JUST_CALC_SIZE ((char_u *) -1)
271
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200272// Values for rs_state in regitem_T.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200273typedef enum regstate_E
274{
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200275 RS_NOPEN = 0 // NOPEN and NCLOSE
276 , RS_MOPEN // MOPEN + [0-9]
277 , RS_MCLOSE // MCLOSE + [0-9]
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200278#ifdef FEAT_SYN_HL
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200279 , RS_ZOPEN // ZOPEN + [0-9]
280 , RS_ZCLOSE // ZCLOSE + [0-9]
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200281#endif
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200282 , RS_BRANCH // BRANCH
283 , RS_BRCPLX_MORE // BRACE_COMPLEX and trying one more match
284 , RS_BRCPLX_LONG // BRACE_COMPLEX and trying longest match
285 , RS_BRCPLX_SHORT // BRACE_COMPLEX and trying shortest match
286 , RS_NOMATCH // NOMATCH
287 , RS_BEHIND1 // BEHIND / NOBEHIND matching rest
288 , RS_BEHIND2 // BEHIND / NOBEHIND matching behind part
289 , RS_STAR_LONG // STAR/PLUS/BRACE_SIMPLE longest match
290 , RS_STAR_SHORT // STAR/PLUS/BRACE_SIMPLE shortest match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200291} regstate_T;
292
293/*
294 * Structure used to save the current input state, when it needs to be
295 * restored after trying a match. Used by reg_save() and reg_restore().
296 * Also stores the length of "backpos".
297 */
298typedef struct
299{
300 union
301 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200302 char_u *ptr; // rex.input pointer, for single-line regexp
303 lpos_T pos; // rex.input pos, for multi-line regexp
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200304 } rs_u;
305 int rs_len;
306} regsave_T;
307
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200308// struct to save start/end pointer/position in for \(\)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200309typedef struct
310{
311 union
312 {
313 char_u *ptr;
314 lpos_T pos;
315 } se_u;
316} save_se_T;
317
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200318// used for BEHIND and NOBEHIND matching
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200319typedef struct regbehind_S
320{
321 regsave_T save_after;
322 regsave_T save_behind;
323 int save_need_clear_subexpr;
324 save_se_T save_start[NSUBEXP];
325 save_se_T save_end[NSUBEXP];
326} regbehind_T;
327
328/*
329 * When there are alternatives a regstate_T is put on the regstack to remember
330 * what we are doing.
331 * Before it may be another type of item, depending on rs_state, to remember
332 * more things.
333 */
334typedef struct regitem_S
335{
336 regstate_T rs_state; // what we are doing, one of RS_ above
337 short rs_no; // submatch nr or BEHIND/NOBEHIND
338 char_u *rs_scan; // current node in program
339 union
340 {
341 save_se_T sesave;
342 regsave_T regsave;
343 } rs_un; // room for saving rex.input
344} regitem_T;
345
346
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200347// used for STAR, PLUS and BRACE_SIMPLE matching
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200348typedef struct regstar_S
349{
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200350 int nextb; // next byte
351 int nextb_ic; // next byte reverse case
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200352 long count;
353 long minval;
354 long maxval;
355} regstar_T;
356
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200357// used to store input position when a BACK was encountered, so that we now if
358// we made any progress since the last time.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200359typedef struct backpos_S
360{
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200361 char_u *bp_scan; // "scan" where BACK was encountered
362 regsave_T bp_pos; // last input position
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200363} backpos_T;
364
365/*
366 * "regstack" and "backpos" are used by regmatch(). They are kept over calls
367 * to avoid invoking malloc() and free() often.
368 * "regstack" is a stack with regitem_T items, sometimes preceded by regstar_T
369 * or regbehind_T.
370 * "backpos_T" is a table with backpos_T for BACK
371 */
372static garray_T regstack = {0, 0, 0, 0, NULL};
373static garray_T backpos = {0, 0, 0, 0, NULL};
374
375static regsave_T behind_pos;
376
377/*
378 * Both for regstack and backpos tables we use the following strategy of
379 * allocation (to reduce malloc/free calls):
380 * - Initial size is fairly small.
381 * - When needed, the tables are grown bigger (8 times at first, double after
382 * that).
383 * - After executing the match we free the memory only if the array has grown.
384 * Thus the memory is kept allocated when it's at the initial size.
385 * This makes it fast while not keeping a lot of memory allocated.
386 * A three times speed increase was observed when using many simple patterns.
387 */
388#define REGSTACK_INITIAL 2048
389#define BACKPOS_INITIAL 64
390
391/*
392 * Opcode notes:
393 *
394 * BRANCH The set of branches constituting a single choice are hooked
395 * together with their "next" pointers, since precedence prevents
396 * anything being concatenated to any individual branch. The
397 * "next" pointer of the last BRANCH in a choice points to the
398 * thing following the whole choice. This is also where the
399 * final "next" pointer of each individual branch points; each
400 * branch starts with the operand node of a BRANCH node.
401 *
402 * BACK Normal "next" pointers all implicitly point forward; BACK
403 * exists to make loop structures possible.
404 *
405 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular
406 * BRANCH structures using BACK. Simple cases (one character
407 * per match) are implemented with STAR and PLUS for speed
408 * and to minimize recursive plunges.
409 *
410 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX
411 * node, and defines the min and max limits to be used for that
412 * node.
413 *
414 * MOPEN,MCLOSE ...are numbered at compile time.
415 * ZOPEN,ZCLOSE ...ditto
416 */
417
418/*
419 * A node is one char of opcode followed by two chars of "next" pointer.
420 * "Next" pointers are stored as two 8-bit bytes, high order first. The
421 * value is a positive offset from the opcode of the node containing it.
422 * An operand, if any, simply follows the node. (Note that much of the
423 * code generation knows about this implicit relationship.)
424 *
425 * Using two bytes for the "next" pointer is vast overkill for most things,
426 * but allows patterns to get big without disasters.
427 */
428#define OP(p) ((int)*(p))
429#define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377))
430#define OPERAND(p) ((p) + 3)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200431// Obtain an operand that was stored as four bytes, MSB first.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200432#define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \
433 + ((long)(p)[5] << 8) + (long)(p)[6])
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200434// Obtain a second operand stored as four bytes.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200435#define OPERAND_MAX(p) OPERAND_MIN((p) + 4)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200436// Obtain a second single-byte operand stored after a four bytes operand.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200437#define OPERAND_CMP(p) (p)[7]
438
439static char_u *reg(int paren, int *flagp);
440
441#ifdef BT_REGEXP_DUMP
442static void regdump(char_u *, bt_regprog_T *);
443#endif
444
445static int re_num_cmp(long_u val, char_u *scan);
446
447#ifdef DEBUG
448static char_u *regprop(char_u *);
449
450static int regnarrate = 0;
451#endif
452
453
454/*
455 * Setup to parse the regexp. Used once to get the length and once to do it.
456 */
457 static void
458regcomp_start(
459 char_u *expr,
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200460 int re_flags) // see vim_regcomp()
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200461{
462 initchr(expr);
463 if (re_flags & RE_MAGIC)
464 reg_magic = MAGIC_ON;
465 else
466 reg_magic = MAGIC_OFF;
467 reg_string = (re_flags & RE_STRING);
468 reg_strict = (re_flags & RE_STRICT);
469 get_cpo_flags();
470
471 num_complex_braces = 0;
472 regnpar = 1;
Bram Moolenaara80faa82020-04-12 19:37:17 +0200473 CLEAR_FIELD(had_endbrace);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200474#ifdef FEAT_SYN_HL
475 regnzpar = 1;
476 re_has_z = 0;
477#endif
478 regsize = 0L;
479 reg_toolong = FALSE;
480 regflags = 0;
481#if defined(FEAT_SYN_HL) || defined(PROTO)
482 had_eol = FALSE;
483#endif
484}
485
486/*
487 * Return TRUE if MULTIBYTECODE should be used instead of EXACTLY for
488 * character "c".
489 */
490 static int
491use_multibytecode(int c)
492{
493 return has_mbyte && (*mb_char2len)(c) > 1
494 && (re_multi_type(peekchr()) != NOT_MULTI
495 || (enc_utf8 && utf_iscomposing(c)));
496}
497
498/*
499 * Emit (if appropriate) a byte of code
500 */
501 static void
502regc(int b)
503{
504 if (regcode == JUST_CALC_SIZE)
505 regsize++;
506 else
507 *regcode++ = b;
508}
509
510/*
511 * Emit (if appropriate) a multi-byte character of code
512 */
513 static void
514regmbc(int c)
515{
516 if (!has_mbyte && c > 0xff)
517 return;
518 if (regcode == JUST_CALC_SIZE)
519 regsize += (*mb_char2len)(c);
520 else
521 regcode += (*mb_char2bytes)(c, regcode);
522}
523
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200524
525/*
526 * Produce the bytes for equivalence class "c".
527 * Currently only handles latin1, latin9 and utf-8.
528 * NOTE: When changing this function, also change nfa_emit_equi_class()
529 */
530 static void
531reg_equi_class(int c)
532{
533 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0
534 || STRCMP(p_enc, "iso-8859-15") == 0)
535 {
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200536 switch (c)
537 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +0200538 // Do not use '\300' style, it results in a negative number.
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200539 case 'A': case 0xc0: case 0xc1: case 0xc2: case 0xc3: case 0xc4:
540 case 0xc5: case 0x100: case 0x102: case 0x104: case 0x1cd:
541 case 0x1de: case 0x1e0: case 0x1fa: case 0x202: case 0x226:
542 case 0x23a: case 0x1e00: case 0x1ea0: case 0x1ea2: case 0x1ea4:
543 case 0x1ea6: case 0x1ea8: case 0x1eaa: case 0x1eac: case 0x1eae:
544 case 0x1eb0: case 0x1eb2: case 0x1eb4: case 0x1eb6:
545 regmbc('A'); regmbc(0xc0); regmbc(0xc1); regmbc(0xc2);
546 regmbc(0xc3); regmbc(0xc4); regmbc(0xc5);
547 regmbc(0x100); regmbc(0x102); regmbc(0x104);
548 regmbc(0x1cd); regmbc(0x1de); regmbc(0x1e0);
549 regmbc(0x1fa); regmbc(0x202); regmbc(0x226);
550 regmbc(0x23a); regmbc(0x1e00); regmbc(0x1ea0);
551 regmbc(0x1ea2); regmbc(0x1ea4); regmbc(0x1ea6);
552 regmbc(0x1ea8); regmbc(0x1eaa); regmbc(0x1eac);
553 regmbc(0x1eae); regmbc(0x1eb0); regmbc(0x1eb2);
554 regmbc(0x1eb4); regmbc(0x1eb6);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200555 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200556 case 'B': case 0x181: case 0x243: case 0x1e02:
557 case 0x1e04: case 0x1e06:
558 regmbc('B');
559 regmbc(0x181); regmbc(0x243); regmbc(0x1e02);
560 regmbc(0x1e04); regmbc(0x1e06);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200561 return;
562 case 'C': case 0xc7:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200563 case 0x106: case 0x108: case 0x10a: case 0x10c: case 0x187:
564 case 0x23b: case 0x1e08: case 0xa792:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200565 regmbc('C'); regmbc(0xc7);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200566 regmbc(0x106); regmbc(0x108); regmbc(0x10a);
567 regmbc(0x10c); regmbc(0x187); regmbc(0x23b);
568 regmbc(0x1e08); regmbc(0xa792);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200569 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200570 case 'D': case 0x10e: case 0x110: case 0x18a:
571 case 0x1e0a: case 0x1e0c: case 0x1e0e: case 0x1e10:
572 case 0x1e12:
573 regmbc('D'); regmbc(0x10e); regmbc(0x110);
574 regmbc(0x18a); regmbc(0x1e0a); regmbc(0x1e0c);
575 regmbc(0x1e0e); regmbc(0x1e10); regmbc(0x1e12);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200576 return;
577 case 'E': case 0xc8: case 0xc9: case 0xca: case 0xcb:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200578 case 0x112: case 0x114: case 0x116: case 0x118: case 0x11a:
579 case 0x204: case 0x206: case 0x228: case 0x246: case 0x1e14:
580 case 0x1e16: case 0x1e18: case 0x1e1a: case 0x1e1c:
581 case 0x1eb8: case 0x1eba: case 0x1ebc: case 0x1ebe:
582 case 0x1ec0: case 0x1ec2: case 0x1ec4: case 0x1ec6:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200583 regmbc('E'); regmbc(0xc8); regmbc(0xc9);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200584 regmbc(0xca); regmbc(0xcb); regmbc(0x112);
585 regmbc(0x114); regmbc(0x116); regmbc(0x118);
586 regmbc(0x11a); regmbc(0x204); regmbc(0x206);
587 regmbc(0x228); regmbc(0x246); regmbc(0x1e14);
588 regmbc(0x1e16); regmbc(0x1e18); regmbc(0x1e1a);
589 regmbc(0x1e1c); regmbc(0x1eb8); regmbc(0x1eba);
590 regmbc(0x1ebc); regmbc(0x1ebe); regmbc(0x1ec0);
591 regmbc(0x1ec2); regmbc(0x1ec4); regmbc(0x1ec6);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200592 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200593 case 'F': case 0x191: case 0x1e1e: case 0xa798:
594 regmbc('F'); regmbc(0x191); regmbc(0x1e1e);
595 regmbc(0xa798);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200596 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200597 case 'G': case 0x11c: case 0x11e: case 0x120:
598 case 0x122: case 0x193: case 0x1e4: case 0x1e6:
599 case 0x1f4: case 0x1e20: case 0xa7a0:
600 regmbc('G'); regmbc(0x11c); regmbc(0x11e);
601 regmbc(0x120); regmbc(0x122); regmbc(0x193);
602 regmbc(0x1e4); regmbc(0x1e6); regmbc(0x1f4);
603 regmbc(0x1e20); regmbc(0xa7a0);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200604 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200605 case 'H': case 0x124: case 0x126: case 0x21e:
606 case 0x1e22: case 0x1e24: case 0x1e26:
607 case 0x1e28: case 0x1e2a: case 0x2c67:
608 regmbc('H'); regmbc(0x124); regmbc(0x126);
609 regmbc(0x21e); regmbc(0x1e22); regmbc(0x1e24);
610 regmbc(0x1e26); regmbc(0x1e28); regmbc(0x1e2a);
611 regmbc(0x2c67);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200612 return;
613 case 'I': case 0xcc: case 0xcd: case 0xce: case 0xcf:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200614 case 0x128: case 0x12a: case 0x12c: case 0x12e:
615 case 0x130: case 0x197: case 0x1cf: case 0x208:
616 case 0x20a: case 0x1e2c: case 0x1e2e: case 0x1ec8:
617 case 0x1eca:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200618 regmbc('I'); regmbc(0xcc); regmbc(0xcd);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200619 regmbc(0xce); regmbc(0xcf); regmbc(0x128);
620 regmbc(0x12a); regmbc(0x12c); regmbc(0x12e);
621 regmbc(0x130); regmbc(0x197); regmbc(0x1cf);
622 regmbc(0x208); regmbc(0x20a); regmbc(0x1e2c);
623 regmbc(0x1e2e); regmbc(0x1ec8); regmbc(0x1eca);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200624 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200625 case 'J': case 0x134: case 0x248:
626 regmbc('J'); regmbc(0x134); regmbc(0x248);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200627 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200628 case 'K': case 0x136: case 0x198: case 0x1e8: case 0x1e30:
629 case 0x1e32: case 0x1e34: case 0x2c69: case 0xa740:
630 regmbc('K'); regmbc(0x136); regmbc(0x198);
631 regmbc(0x1e8); regmbc(0x1e30); regmbc(0x1e32);
632 regmbc(0x1e34); regmbc(0x2c69); regmbc(0xa740);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200633 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200634 case 'L': case 0x139: case 0x13b: case 0x13d: case 0x13f:
635 case 0x141: case 0x23d: case 0x1e36: case 0x1e38:
636 case 0x1e3a: case 0x1e3c: case 0x2c60:
637 regmbc('L'); regmbc(0x139); regmbc(0x13b);
638 regmbc(0x13d); regmbc(0x13f); regmbc(0x141);
639 regmbc(0x23d); regmbc(0x1e36); regmbc(0x1e38);
640 regmbc(0x1e3a); regmbc(0x1e3c); regmbc(0x2c60);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200641 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200642 case 'M': case 0x1e3e: case 0x1e40: case 0x1e42:
643 regmbc('M'); regmbc(0x1e3e); regmbc(0x1e40);
644 regmbc(0x1e42);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200645 return;
646 case 'N': case 0xd1:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200647 case 0x143: case 0x145: case 0x147: case 0x1f8:
648 case 0x1e44: case 0x1e46: case 0x1e48: case 0x1e4a:
649 case 0xa7a4:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200650 regmbc('N'); regmbc(0xd1);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200651 regmbc(0x143); regmbc(0x145); regmbc(0x147);
652 regmbc(0x1f8); regmbc(0x1e44); regmbc(0x1e46);
653 regmbc(0x1e48); regmbc(0x1e4a); regmbc(0xa7a4);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200654 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200655 case 'O': case 0xd2: case 0xd3: case 0xd4: case 0xd5: case 0xd6:
656 case 0xd8: case 0x14c: case 0x14e: case 0x150: case 0x19f:
657 case 0x1a0: case 0x1d1: case 0x1ea: case 0x1ec: case 0x1fe:
658 case 0x20c: case 0x20e: case 0x22a: case 0x22c: case 0x22e:
659 case 0x230: case 0x1e4c: case 0x1e4e: case 0x1e50: case 0x1e52:
660 case 0x1ecc: case 0x1ece: case 0x1ed0: case 0x1ed2: case 0x1ed4:
661 case 0x1ed6: case 0x1ed8: case 0x1eda: case 0x1edc: case 0x1ede:
662 case 0x1ee0: case 0x1ee2:
663 regmbc('O'); regmbc(0xd2); regmbc(0xd3); regmbc(0xd4);
664 regmbc(0xd5); regmbc(0xd6); regmbc(0xd8);
665 regmbc(0x14c); regmbc(0x14e); regmbc(0x150);
666 regmbc(0x19f); regmbc(0x1a0); regmbc(0x1d1);
667 regmbc(0x1ea); regmbc(0x1ec); regmbc(0x1fe);
668 regmbc(0x20c); regmbc(0x20e); regmbc(0x22a);
669 regmbc(0x22c); regmbc(0x22e); regmbc(0x230);
670 regmbc(0x1e4c); regmbc(0x1e4e); regmbc(0x1e50);
671 regmbc(0x1e52); regmbc(0x1ecc); regmbc(0x1ece);
672 regmbc(0x1ed0); regmbc(0x1ed2); regmbc(0x1ed4);
673 regmbc(0x1ed6); regmbc(0x1ed8); regmbc(0x1eda);
674 regmbc(0x1edc); regmbc(0x1ede); regmbc(0x1ee0);
675 regmbc(0x1ee2);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200676 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200677 case 'P': case 0x1a4: case 0x1e54: case 0x1e56: case 0x2c63:
678 regmbc('P'); regmbc(0x1a4); regmbc(0x1e54);
679 regmbc(0x1e56); regmbc(0x2c63);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200680 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200681 case 'Q': case 0x24a:
682 regmbc('Q'); regmbc(0x24a);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200683 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200684 case 'R': case 0x154: case 0x156: case 0x158: case 0x210:
685 case 0x212: case 0x24c: case 0x1e58: case 0x1e5a:
686 case 0x1e5c: case 0x1e5e: case 0x2c64: case 0xa7a6:
687 regmbc('R'); regmbc(0x154); regmbc(0x156);
688 regmbc(0x210); regmbc(0x212); regmbc(0x158);
689 regmbc(0x24c); regmbc(0x1e58); regmbc(0x1e5a);
690 regmbc(0x1e5c); regmbc(0x1e5e); regmbc(0x2c64);
691 regmbc(0xa7a6);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200692 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200693 case 'S': case 0x15a: case 0x15c: case 0x15e: case 0x160:
694 case 0x218: case 0x1e60: case 0x1e62: case 0x1e64:
695 case 0x1e66: case 0x1e68: case 0x2c7e: case 0xa7a8:
696 regmbc('S'); regmbc(0x15a); regmbc(0x15c);
697 regmbc(0x15e); regmbc(0x160); regmbc(0x218);
698 regmbc(0x1e60); regmbc(0x1e62); regmbc(0x1e64);
699 regmbc(0x1e66); regmbc(0x1e68); regmbc(0x2c7e);
700 regmbc(0xa7a8);
701 return;
702 case 'T': case 0x162: case 0x164: case 0x166: case 0x1ac:
703 case 0x1ae: case 0x21a: case 0x23e: case 0x1e6a: case 0x1e6c:
704 case 0x1e6e: case 0x1e70:
705 regmbc('T'); regmbc(0x162); regmbc(0x164);
706 regmbc(0x166); regmbc(0x1ac); regmbc(0x23e);
707 regmbc(0x1ae); regmbc(0x21a); regmbc(0x1e6a);
708 regmbc(0x1e6c); regmbc(0x1e6e); regmbc(0x1e70);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200709 return;
710 case 'U': case 0xd9: case 0xda: case 0xdb: case 0xdc:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200711 case 0x168: case 0x16a: case 0x16c: case 0x16e:
712 case 0x170: case 0x172: case 0x1af: case 0x1d3:
713 case 0x1d5: case 0x1d7: case 0x1d9: case 0x1db:
714 case 0x214: case 0x216: case 0x244: case 0x1e72:
715 case 0x1e74: case 0x1e76: case 0x1e78: case 0x1e7a:
716 case 0x1ee4: case 0x1ee6: case 0x1ee8: case 0x1eea:
717 case 0x1eec: case 0x1eee: case 0x1ef0:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200718 regmbc('U'); regmbc(0xd9); regmbc(0xda);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200719 regmbc(0xdb); regmbc(0xdc); regmbc(0x168);
720 regmbc(0x16a); regmbc(0x16c); regmbc(0x16e);
721 regmbc(0x170); regmbc(0x172); regmbc(0x1af);
722 regmbc(0x1d3); regmbc(0x1d5); regmbc(0x1d7);
723 regmbc(0x1d9); regmbc(0x1db); regmbc(0x214);
724 regmbc(0x216); regmbc(0x244); regmbc(0x1e72);
725 regmbc(0x1e74); regmbc(0x1e76); regmbc(0x1e78);
726 regmbc(0x1e7a); regmbc(0x1ee4); regmbc(0x1ee6);
727 regmbc(0x1ee8); regmbc(0x1eea); regmbc(0x1eec);
728 regmbc(0x1eee); regmbc(0x1ef0);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200729 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200730 case 'V': case 0x1b2: case 0x1e7c: case 0x1e7e:
731 regmbc('V'); regmbc(0x1b2); regmbc(0x1e7c);
732 regmbc(0x1e7e);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200733 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200734 case 'W': case 0x174: case 0x1e80: case 0x1e82:
735 case 0x1e84: case 0x1e86: case 0x1e88:
736 regmbc('W'); regmbc(0x174); regmbc(0x1e80);
737 regmbc(0x1e82); regmbc(0x1e84); regmbc(0x1e86);
738 regmbc(0x1e88);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200739 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200740 case 'X': case 0x1e8a: case 0x1e8c:
741 regmbc('X'); regmbc(0x1e8a); regmbc(0x1e8c);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200742 return;
743 case 'Y': case 0xdd:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200744 case 0x176: case 0x178: case 0x1b3: case 0x232: case 0x24e:
745 case 0x1e8e: case 0x1ef2: case 0x1ef6: case 0x1ef4: case 0x1ef8:
746 regmbc('Y'); regmbc(0xdd); regmbc(0x176);
747 regmbc(0x178); regmbc(0x1b3); regmbc(0x232);
748 regmbc(0x24e); regmbc(0x1e8e); regmbc(0x1ef2);
749 regmbc(0x1ef4); regmbc(0x1ef6); regmbc(0x1ef8);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200750 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200751 case 'Z': case 0x179: case 0x17b: case 0x17d: case 0x1b5:
752 case 0x1e90: case 0x1e92: case 0x1e94: case 0x2c6b:
753 regmbc('Z'); regmbc(0x179); regmbc(0x17b);
754 regmbc(0x17d); regmbc(0x1b5); regmbc(0x1e90);
755 regmbc(0x1e92); regmbc(0x1e94); regmbc(0x2c6b);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200756 return;
757 case 'a': case 0xe0: case 0xe1: case 0xe2:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200758 case 0xe3: case 0xe4: case 0xe5: case 0x101: case 0x103:
759 case 0x105: case 0x1ce: case 0x1df: case 0x1e1: case 0x1fb:
760 case 0x201: case 0x203: case 0x227: case 0x1d8f: case 0x1e01:
761 case 0x1e9a: case 0x1ea1: case 0x1ea3: case 0x1ea5:
762 case 0x1ea7: case 0x1ea9: case 0x1eab: case 0x1ead:
763 case 0x1eaf: case 0x1eb1: case 0x1eb3: case 0x1eb5:
764 case 0x1eb7: case 0x2c65:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200765 regmbc('a'); regmbc(0xe0); regmbc(0xe1);
766 regmbc(0xe2); regmbc(0xe3); regmbc(0xe4);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200767 regmbc(0xe5); regmbc(0x101); regmbc(0x103);
768 regmbc(0x105); regmbc(0x1ce); regmbc(0x1df);
769 regmbc(0x1e1); regmbc(0x1fb); regmbc(0x201);
770 regmbc(0x203); regmbc(0x227); regmbc(0x1d8f);
771 regmbc(0x1e01); regmbc(0x1e9a); regmbc(0x1ea1);
772 regmbc(0x1ea3); regmbc(0x1ea5); regmbc(0x1ea7);
773 regmbc(0x1ea9); regmbc(0x1eab); regmbc(0x1ead);
774 regmbc(0x1eaf); regmbc(0x1eb1); regmbc(0x1eb3);
775 regmbc(0x1eb5); regmbc(0x1eb7); regmbc(0x2c65);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200776 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200777 case 'b': case 0x180: case 0x253: case 0x1d6c: case 0x1d80:
778 case 0x1e03: case 0x1e05: case 0x1e07:
779 regmbc('b');
780 regmbc(0x180); regmbc(0x253); regmbc(0x1d6c);
781 regmbc(0x1d80); regmbc(0x1e03); regmbc(0x1e05);
782 regmbc(0x1e07);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200783 return;
784 case 'c': case 0xe7:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200785 case 0x107: case 0x109: case 0x10b: case 0x10d: case 0x188:
786 case 0x23c: case 0x1e09: case 0xa793: case 0xa794:
787 regmbc('c'); regmbc(0xe7); regmbc(0x107);
788 regmbc(0x109); regmbc(0x10b); regmbc(0x10d);
789 regmbc(0x188); regmbc(0x23c); regmbc(0x1e09);
790 regmbc(0xa793); regmbc(0xa794);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200791 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200792 case 'd': case 0x10f: case 0x111: case 0x257: case 0x1d6d:
793 case 0x1d81: case 0x1d91: case 0x1e0b: case 0x1e0d:
794 case 0x1e0f: case 0x1e11: case 0x1e13:
795 regmbc('d'); regmbc(0x10f); regmbc(0x111);
796 regmbc(0x257); regmbc(0x1d6d); regmbc(0x1d81);
797 regmbc(0x1d91); regmbc(0x1e0b); regmbc(0x1e0d);
798 regmbc(0x1e0f); regmbc(0x1e11); regmbc(0x1e13);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200799 return;
800 case 'e': case 0xe8: case 0xe9: case 0xea: case 0xeb:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200801 case 0x113: case 0x115: case 0x117: case 0x119:
802 case 0x11b: case 0x205: case 0x207: case 0x229:
803 case 0x247: case 0x1d92: case 0x1e15: case 0x1e17:
804 case 0x1e19: case 0x1e1b: case 0x1eb9: case 0x1ebb:
805 case 0x1e1d: case 0x1ebd: case 0x1ebf: case 0x1ec1:
806 case 0x1ec3: case 0x1ec5: case 0x1ec7:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200807 regmbc('e'); regmbc(0xe8); regmbc(0xe9);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200808 regmbc(0xea); regmbc(0xeb); regmbc(0x113);
809 regmbc(0x115); regmbc(0x117); regmbc(0x119);
810 regmbc(0x11b); regmbc(0x205); regmbc(0x207);
811 regmbc(0x229); regmbc(0x247); regmbc(0x1d92);
812 regmbc(0x1e15); regmbc(0x1e17); regmbc(0x1e19);
813 regmbc(0x1e1b); regmbc(0x1e1d); regmbc(0x1eb9);
814 regmbc(0x1ebb); regmbc(0x1ebd); regmbc(0x1ebf);
815 regmbc(0x1ec1); regmbc(0x1ec3); regmbc(0x1ec5);
816 regmbc(0x1ec7);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200817 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200818 case 'f': case 0x192: case 0x1d6e: case 0x1d82:
819 case 0x1e1f: case 0xa799:
820 regmbc('f'); regmbc(0x192); regmbc(0x1d6e);
821 regmbc(0x1d82); regmbc(0x1e1f); regmbc(0xa799);
822 return;
823 case 'g': case 0x11d: case 0x11f: case 0x121: case 0x123:
824 case 0x1e5: case 0x1e7: case 0x260: case 0x1f5: case 0x1d83:
825 case 0x1e21: case 0xa7a1:
826 regmbc('g'); regmbc(0x11d); regmbc(0x11f);
827 regmbc(0x121); regmbc(0x123); regmbc(0x1e5);
828 regmbc(0x1e7); regmbc(0x1f5); regmbc(0x260);
829 regmbc(0x1d83); regmbc(0x1e21); regmbc(0xa7a1);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200830 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200831 case 'h': case 0x125: case 0x127: case 0x21f: case 0x1e23:
832 case 0x1e25: case 0x1e27: case 0x1e29: case 0x1e2b:
833 case 0x1e96: case 0x2c68: case 0xa795:
834 regmbc('h'); regmbc(0x125); regmbc(0x127);
835 regmbc(0x21f); regmbc(0x1e23); regmbc(0x1e25);
836 regmbc(0x1e27); regmbc(0x1e29); regmbc(0x1e2b);
837 regmbc(0x1e96); regmbc(0x2c68); regmbc(0xa795);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200838 return;
839 case 'i': case 0xec: case 0xed: case 0xee: case 0xef:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200840 case 0x129: case 0x12b: case 0x12d: case 0x12f:
841 case 0x1d0: case 0x209: case 0x20b: case 0x268:
842 case 0x1d96: case 0x1e2d: case 0x1e2f: case 0x1ec9:
843 case 0x1ecb:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200844 regmbc('i'); regmbc(0xec); regmbc(0xed);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200845 regmbc(0xee); regmbc(0xef); regmbc(0x129);
846 regmbc(0x12b); regmbc(0x12d); regmbc(0x12f);
847 regmbc(0x1d0); regmbc(0x209); regmbc(0x20b);
848 regmbc(0x268); regmbc(0x1d96); regmbc(0x1e2d);
849 regmbc(0x1e2f); regmbc(0x1ec9); regmbc(0x1ecb);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200850 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200851 case 'j': case 0x135: case 0x1f0: case 0x249:
852 regmbc('j'); regmbc(0x135); regmbc(0x1f0);
853 regmbc(0x249);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200854 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200855 case 'k': case 0x137: case 0x199: case 0x1e9:
856 case 0x1d84: case 0x1e31: case 0x1e33: case 0x1e35:
857 case 0x2c6a: case 0xa741:
858 regmbc('k'); regmbc(0x137); regmbc(0x199);
859 regmbc(0x1e9); regmbc(0x1d84); regmbc(0x1e31);
860 regmbc(0x1e33); regmbc(0x1e35); regmbc(0x2c6a);
861 regmbc(0xa741);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200862 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200863 case 'l': case 0x13a: case 0x13c: case 0x13e:
864 case 0x140: case 0x142: case 0x19a: case 0x1e37:
865 case 0x1e39: case 0x1e3b: case 0x1e3d: case 0x2c61:
866 regmbc('l'); regmbc(0x13a); regmbc(0x13c);
867 regmbc(0x13e); regmbc(0x140); regmbc(0x142);
868 regmbc(0x19a); regmbc(0x1e37); regmbc(0x1e39);
869 regmbc(0x1e3b); regmbc(0x1e3d); regmbc(0x2c61);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200870 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200871 case 'm': case 0x1d6f: case 0x1e3f: case 0x1e41: case 0x1e43:
872 regmbc('m'); regmbc(0x1d6f); regmbc(0x1e3f);
873 regmbc(0x1e41); regmbc(0x1e43);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200874 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200875 case 'n': case 0xf1: case 0x144: case 0x146: case 0x148:
876 case 0x149: case 0x1f9: case 0x1d70: case 0x1d87:
877 case 0x1e45: case 0x1e47: case 0x1e49: case 0x1e4b:
878 case 0xa7a5:
879 regmbc('n'); regmbc(0xf1); regmbc(0x144);
880 regmbc(0x146); regmbc(0x148); regmbc(0x149);
881 regmbc(0x1f9); regmbc(0x1d70); regmbc(0x1d87);
882 regmbc(0x1e45); regmbc(0x1e47); regmbc(0x1e49);
883 regmbc(0x1e4b); regmbc(0xa7a5);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200884 return;
885 case 'o': case 0xf2: case 0xf3: case 0xf4: case 0xf5:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200886 case 0xf6: case 0xf8: case 0x14d: case 0x14f: case 0x151:
887 case 0x1a1: case 0x1d2: case 0x1eb: case 0x1ed: case 0x1ff:
888 case 0x20d: case 0x20f: case 0x22b: case 0x22d: case 0x22f:
889 case 0x231: case 0x275: case 0x1e4d: case 0x1e4f:
890 case 0x1e51: case 0x1e53: case 0x1ecd: case 0x1ecf:
891 case 0x1ed1: case 0x1ed3: case 0x1ed5: case 0x1ed7:
892 case 0x1ed9: case 0x1edb: case 0x1edd: case 0x1edf:
893 case 0x1ee1: case 0x1ee3:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200894 regmbc('o'); regmbc(0xf2); regmbc(0xf3);
895 regmbc(0xf4); regmbc(0xf5); regmbc(0xf6);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200896 regmbc(0xf8); regmbc(0x14d); regmbc(0x14f);
897 regmbc(0x151); regmbc(0x1a1); regmbc(0x1d2);
898 regmbc(0x1eb); regmbc(0x1ed); regmbc(0x1ff);
899 regmbc(0x20d); regmbc(0x20f); regmbc(0x22b);
900 regmbc(0x22d); regmbc(0x22f); regmbc(0x231);
901 regmbc(0x275); regmbc(0x1e4d); regmbc(0x1e4f);
902 regmbc(0x1e51); regmbc(0x1e53); regmbc(0x1ecd);
903 regmbc(0x1ecf); regmbc(0x1ed1); regmbc(0x1ed3);
904 regmbc(0x1ed5); regmbc(0x1ed7); regmbc(0x1ed9);
905 regmbc(0x1edb); regmbc(0x1edd); regmbc(0x1edf);
906 regmbc(0x1ee1); regmbc(0x1ee3);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200907 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200908 case 'p': case 0x1a5: case 0x1d71: case 0x1d88: case 0x1d7d:
909 case 0x1e55: case 0x1e57:
910 regmbc('p'); regmbc(0x1a5); regmbc(0x1d71);
911 regmbc(0x1d7d); regmbc(0x1d88); regmbc(0x1e55);
912 regmbc(0x1e57);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200913 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200914 case 'q': case 0x24b: case 0x2a0:
915 regmbc('q'); regmbc(0x24b); regmbc(0x2a0);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200916 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200917 case 'r': case 0x155: case 0x157: case 0x159: case 0x211:
918 case 0x213: case 0x24d: case 0x27d: case 0x1d72: case 0x1d73:
919 case 0x1d89: case 0x1e59: case 0x1e5b: case 0x1e5d: case 0x1e5f:
920 case 0xa7a7:
921 regmbc('r'); regmbc(0x155); regmbc(0x157);
922 regmbc(0x159); regmbc(0x211); regmbc(0x213);
923 regmbc(0x24d); regmbc(0x1d72); regmbc(0x1d73);
924 regmbc(0x1d89); regmbc(0x1e59); regmbc(0x27d);
925 regmbc(0x1e5b); regmbc(0x1e5d); regmbc(0x1e5f);
926 regmbc(0xa7a7);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200927 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200928 case 's': case 0x15b: case 0x15d: case 0x15f: case 0x161:
929 case 0x1e61: case 0x219: case 0x23f: case 0x1d74: case 0x1d8a:
930 case 0x1e63: case 0x1e65: case 0x1e67: case 0x1e69: case 0xa7a9:
931 regmbc('s'); regmbc(0x15b); regmbc(0x15d);
932 regmbc(0x15f); regmbc(0x161); regmbc(0x23f);
933 regmbc(0x219); regmbc(0x1d74); regmbc(0x1d8a);
934 regmbc(0x1e61); regmbc(0x1e63); regmbc(0x1e65);
935 regmbc(0x1e67); regmbc(0x1e69); regmbc(0xa7a9);
936 return;
937 case 't': case 0x163: case 0x165: case 0x167: case 0x1ab:
938 case 0x1ad: case 0x21b: case 0x288: case 0x1d75: case 0x1e6b:
939 case 0x1e6d: case 0x1e6f: case 0x1e71: case 0x1e97: case 0x2c66:
940 regmbc('t'); regmbc(0x163); regmbc(0x165);
941 regmbc(0x167); regmbc(0x1ab); regmbc(0x21b);
942 regmbc(0x1ad); regmbc(0x288); regmbc(0x1d75);
943 regmbc(0x1e6b); regmbc(0x1e6d); regmbc(0x1e6f);
944 regmbc(0x1e71); regmbc(0x1e97); regmbc(0x2c66);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200945 return;
946 case 'u': case 0xf9: case 0xfa: case 0xfb: case 0xfc:
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200947 case 0x169: case 0x16b: case 0x16d: case 0x16f:
948 case 0x171: case 0x173: case 0x1b0: case 0x1d4:
949 case 0x1d6: case 0x1d8: case 0x1da: case 0x1dc:
950 case 0x215: case 0x217: case 0x289: case 0x1e73:
951 case 0x1d7e: case 0x1d99: case 0x1e75: case 0x1e77:
952 case 0x1e79: case 0x1e7b: case 0x1ee5: case 0x1ee7:
953 case 0x1ee9: case 0x1eeb: case 0x1eed: case 0x1eef:
954 case 0x1ef1:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200955 regmbc('u'); regmbc(0xf9); regmbc(0xfa);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200956 regmbc(0xfb); regmbc(0xfc); regmbc(0x169);
957 regmbc(0x16b); regmbc(0x16d); regmbc(0x16f);
958 regmbc(0x171); regmbc(0x173); regmbc(0x1d6);
959 regmbc(0x1d8); regmbc(0x1da); regmbc(0x1dc);
960 regmbc(0x215); regmbc(0x217); regmbc(0x1b0);
961 regmbc(0x1d4); regmbc(0x289); regmbc(0x1d7e);
962 regmbc(0x1d99); regmbc(0x1e73); regmbc(0x1e75);
963 regmbc(0x1e77); regmbc(0x1e79); regmbc(0x1e7b);
964 regmbc(0x1ee5); regmbc(0x1ee7); regmbc(0x1ee9);
965 regmbc(0x1eeb); regmbc(0x1eed); regmbc(0x1eef);
966 regmbc(0x1ef1);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200967 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200968 case 'v': case 0x28b: case 0x1d8c: case 0x1e7d: case 0x1e7f:
969 regmbc('v'); regmbc(0x28b); regmbc(0x1d8c);
970 regmbc(0x1e7d); regmbc(0x1e7f);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200971 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200972 case 'w': case 0x175: case 0x1e81: case 0x1e83:
973 case 0x1e85: case 0x1e87: case 0x1e89: case 0x1e98:
974 regmbc('w'); regmbc(0x175); regmbc(0x1e81);
975 regmbc(0x1e83); regmbc(0x1e85); regmbc(0x1e87);
976 regmbc(0x1e89); regmbc(0x1e98);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200977 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200978 case 'x': case 0x1e8b: case 0x1e8d:
979 regmbc('x'); regmbc(0x1e8b); regmbc(0x1e8d);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200980 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200981 case 'y': case 0xfd: case 0xff: case 0x177: case 0x1b4:
982 case 0x233: case 0x24f: case 0x1e8f: case 0x1e99: case 0x1ef3:
983 case 0x1ef5: case 0x1ef7: case 0x1ef9:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200984 regmbc('y'); regmbc(0xfd); regmbc(0xff);
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200985 regmbc(0x177); regmbc(0x1b4); regmbc(0x233);
986 regmbc(0x24f); regmbc(0x1e8f); regmbc(0x1e99);
987 regmbc(0x1ef3); regmbc(0x1ef5); regmbc(0x1ef7);
988 regmbc(0x1ef9);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200989 return;
Bram Moolenaar0b94e292021-04-05 13:59:53 +0200990 case 'z': case 0x17a: case 0x17c: case 0x17e: case 0x1b6:
991 case 0x1d76: case 0x1d8e: case 0x1e91: case 0x1e93:
992 case 0x1e95: case 0x2c6c:
993 regmbc('z'); regmbc(0x17a); regmbc(0x17c);
994 regmbc(0x17e); regmbc(0x1b6); regmbc(0x1d76);
995 regmbc(0x1d8e); regmbc(0x1e91); regmbc(0x1e93);
996 regmbc(0x1e95); regmbc(0x2c6c);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200997 return;
998 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +0200999 }
1000 regmbc(c);
1001}
1002
1003/*
1004 * Emit a node.
1005 * Return pointer to generated code.
1006 */
1007 static char_u *
1008regnode(int op)
1009{
1010 char_u *ret;
1011
1012 ret = regcode;
1013 if (ret == JUST_CALC_SIZE)
1014 regsize += 3;
1015 else
1016 {
1017 *regcode++ = op;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001018 *regcode++ = NUL; // Null "next" pointer.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001019 *regcode++ = NUL;
1020 }
1021 return ret;
1022}
1023
1024/*
1025 * Write a long as four bytes at "p" and return pointer to the next char.
1026 */
1027 static char_u *
1028re_put_long(char_u *p, long_u val)
1029{
1030 *p++ = (char_u) ((val >> 24) & 0377);
1031 *p++ = (char_u) ((val >> 16) & 0377);
1032 *p++ = (char_u) ((val >> 8) & 0377);
1033 *p++ = (char_u) (val & 0377);
1034 return p;
1035}
1036
1037/*
1038 * regnext - dig the "next" pointer out of a node
1039 * Returns NULL when calculating size, when there is no next item and when
1040 * there is an error.
1041 */
1042 static char_u *
1043regnext(char_u *p)
1044{
1045 int offset;
1046
1047 if (p == JUST_CALC_SIZE || reg_toolong)
1048 return NULL;
1049
1050 offset = NEXT(p);
1051 if (offset == 0)
1052 return NULL;
1053
1054 if (OP(p) == BACK)
1055 return p - offset;
1056 else
1057 return p + offset;
1058}
1059
1060/*
1061 * Set the next-pointer at the end of a node chain.
1062 */
1063 static void
1064regtail(char_u *p, char_u *val)
1065{
1066 char_u *scan;
1067 char_u *temp;
1068 int offset;
1069
1070 if (p == JUST_CALC_SIZE)
1071 return;
1072
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001073 // Find last node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001074 scan = p;
1075 for (;;)
1076 {
1077 temp = regnext(scan);
1078 if (temp == NULL)
1079 break;
1080 scan = temp;
1081 }
1082
1083 if (OP(scan) == BACK)
1084 offset = (int)(scan - val);
1085 else
1086 offset = (int)(val - scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001087 // When the offset uses more than 16 bits it can no longer fit in the two
1088 // bytes available. Use a global flag to avoid having to check return
1089 // values in too many places.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001090 if (offset > 0xffff)
1091 reg_toolong = TRUE;
1092 else
1093 {
1094 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377);
1095 *(scan + 2) = (char_u) (offset & 0377);
1096 }
1097}
1098
1099/*
1100 * Like regtail, on item after a BRANCH; nop if none.
1101 */
1102 static void
1103regoptail(char_u *p, char_u *val)
1104{
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001105 // When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless"
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001106 if (p == NULL || p == JUST_CALC_SIZE
1107 || (OP(p) != BRANCH
1108 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9)))
1109 return;
1110 regtail(OPERAND(p), val);
1111}
1112
1113/*
1114 * Insert an operator in front of already-emitted operand
1115 *
1116 * Means relocating the operand.
1117 */
1118 static void
1119reginsert(int op, char_u *opnd)
1120{
1121 char_u *src;
1122 char_u *dst;
1123 char_u *place;
1124
1125 if (regcode == JUST_CALC_SIZE)
1126 {
1127 regsize += 3;
1128 return;
1129 }
1130 src = regcode;
1131 regcode += 3;
1132 dst = regcode;
1133 while (src > opnd)
1134 *--dst = *--src;
1135
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001136 place = opnd; // Op node, where operand used to be.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001137 *place++ = op;
1138 *place++ = NUL;
1139 *place = NUL;
1140}
1141
1142/*
1143 * Insert an operator in front of already-emitted operand.
1144 * Add a number to the operator.
1145 */
1146 static void
1147reginsert_nr(int op, long val, char_u *opnd)
1148{
1149 char_u *src;
1150 char_u *dst;
1151 char_u *place;
1152
1153 if (regcode == JUST_CALC_SIZE)
1154 {
1155 regsize += 7;
1156 return;
1157 }
1158 src = regcode;
1159 regcode += 7;
1160 dst = regcode;
1161 while (src > opnd)
1162 *--dst = *--src;
1163
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001164 place = opnd; // Op node, where operand used to be.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001165 *place++ = op;
1166 *place++ = NUL;
1167 *place++ = NUL;
1168 re_put_long(place, (long_u)val);
1169}
1170
1171/*
1172 * Insert an operator in front of already-emitted operand.
1173 * The operator has the given limit values as operands. Also set next pointer.
1174 *
1175 * Means relocating the operand.
1176 */
1177 static void
1178reginsert_limits(
1179 int op,
1180 long minval,
1181 long maxval,
1182 char_u *opnd)
1183{
1184 char_u *src;
1185 char_u *dst;
1186 char_u *place;
1187
1188 if (regcode == JUST_CALC_SIZE)
1189 {
1190 regsize += 11;
1191 return;
1192 }
1193 src = regcode;
1194 regcode += 11;
1195 dst = regcode;
1196 while (src > opnd)
1197 *--dst = *--src;
1198
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001199 place = opnd; // Op node, where operand used to be.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001200 *place++ = op;
1201 *place++ = NUL;
1202 *place++ = NUL;
1203 place = re_put_long(place, (long_u)minval);
1204 place = re_put_long(place, (long_u)maxval);
1205 regtail(opnd, place);
1206}
1207
1208/*
1209 * Return TRUE if the back reference is legal. We must have seen the close
1210 * brace.
1211 * TODO: Should also check that we don't refer to something that is repeated
1212 * (+*=): what instance of the repetition should we match?
1213 */
1214 static int
1215seen_endbrace(int refnum)
1216{
1217 if (!had_endbrace[refnum])
1218 {
1219 char_u *p;
1220
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001221 // Trick: check if "@<=" or "@<!" follows, in which case
1222 // the \1 can appear before the referenced match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001223 for (p = regparse; *p != NUL; ++p)
1224 if (p[0] == '@' && p[1] == '<' && (p[2] == '!' || p[2] == '='))
1225 break;
1226 if (*p == NUL)
1227 {
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001228 emsg(_(e_illegal_back_reference));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001229 rc_did_emsg = TRUE;
1230 return FALSE;
1231 }
1232 }
1233 return TRUE;
1234}
1235
1236/*
1237 * Parse the lowest level.
1238 *
1239 * Optimization: gobbles an entire sequence of ordinary characters so that
1240 * it can turn them into a single node, which is smaller to store and
1241 * faster to run. Don't do this when one_exactly is set.
1242 */
1243 static char_u *
1244regatom(int *flagp)
1245{
1246 char_u *ret;
1247 int flags;
1248 int c;
1249 char_u *p;
1250 int extra = 0;
1251 int save_prev_at_start = prev_at_start;
1252
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001253 *flagp = WORST; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001254
1255 c = getchr();
1256 switch (c)
1257 {
1258 case Magic('^'):
1259 ret = regnode(BOL);
1260 break;
1261
1262 case Magic('$'):
1263 ret = regnode(EOL);
1264#if defined(FEAT_SYN_HL) || defined(PROTO)
1265 had_eol = TRUE;
1266#endif
1267 break;
1268
1269 case Magic('<'):
1270 ret = regnode(BOW);
1271 break;
1272
1273 case Magic('>'):
1274 ret = regnode(EOW);
1275 break;
1276
1277 case Magic('_'):
1278 c = no_Magic(getchr());
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001279 if (c == '^') // "\_^" is start-of-line
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001280 {
1281 ret = regnode(BOL);
1282 break;
1283 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001284 if (c == '$') // "\_$" is end-of-line
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001285 {
1286 ret = regnode(EOL);
1287#if defined(FEAT_SYN_HL) || defined(PROTO)
1288 had_eol = TRUE;
1289#endif
1290 break;
1291 }
1292
1293 extra = ADD_NL;
1294 *flagp |= HASNL;
1295
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001296 // "\_[" is character range plus newline
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001297 if (c == '[')
1298 goto collection;
1299
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001300 // "\_x" is character class plus newline
1301 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001302
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001303 // Character classes.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001304 case Magic('.'):
1305 case Magic('i'):
1306 case Magic('I'):
1307 case Magic('k'):
1308 case Magic('K'):
1309 case Magic('f'):
1310 case Magic('F'):
1311 case Magic('p'):
1312 case Magic('P'):
1313 case Magic('s'):
1314 case Magic('S'):
1315 case Magic('d'):
1316 case Magic('D'):
1317 case Magic('x'):
1318 case Magic('X'):
1319 case Magic('o'):
1320 case Magic('O'):
1321 case Magic('w'):
1322 case Magic('W'):
1323 case Magic('h'):
1324 case Magic('H'):
1325 case Magic('a'):
1326 case Magic('A'):
1327 case Magic('l'):
1328 case Magic('L'):
1329 case Magic('u'):
1330 case Magic('U'):
1331 p = vim_strchr(classchars, no_Magic(c));
1332 if (p == NULL)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001333 EMSG_RET_NULL(_(e_invalid_use_of_underscore));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001334
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001335 // When '.' is followed by a composing char ignore the dot, so that
1336 // the composing char is matched here.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001337 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr()))
1338 {
1339 c = getchr();
1340 goto do_multibyte;
1341 }
1342 ret = regnode(classcodes[p - classchars] + extra);
1343 *flagp |= HASWIDTH | SIMPLE;
1344 break;
1345
1346 case Magic('n'):
1347 if (reg_string)
1348 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001349 // In a string "\n" matches a newline character.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001350 ret = regnode(EXACTLY);
1351 regc(NL);
1352 regc(NUL);
1353 *flagp |= HASWIDTH | SIMPLE;
1354 }
1355 else
1356 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001357 // In buffer text "\n" matches the end of a line.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001358 ret = regnode(NEWL);
1359 *flagp |= HASWIDTH | HASNL;
1360 }
1361 break;
1362
1363 case Magic('('):
1364 if (one_exactly)
1365 EMSG_ONE_RET_NULL;
1366 ret = reg(REG_PAREN, &flags);
1367 if (ret == NULL)
1368 return NULL;
1369 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
1370 break;
1371
1372 case NUL:
1373 case Magic('|'):
1374 case Magic('&'):
1375 case Magic(')'):
1376 if (one_exactly)
1377 EMSG_ONE_RET_NULL;
Bram Moolenaard0819d12021-12-31 23:15:53 +00001378 // Supposed to be caught earlier.
1379 IEMSG_RET_NULL(_(e_internal_error_in_regexp));
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001380 // NOTREACHED
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001381
1382 case Magic('='):
1383 case Magic('?'):
1384 case Magic('+'):
1385 case Magic('@'):
1386 case Magic('{'):
1387 case Magic('*'):
1388 c = no_Magic(c);
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001389 EMSG3_RET_NULL(_(e_str_chr_follows_nothing),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001390 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL), c);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001391 // NOTREACHED
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001392
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001393 case Magic('~'): // previous substitute pattern
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001394 if (reg_prev_sub != NULL)
1395 {
1396 char_u *lp;
1397
1398 ret = regnode(EXACTLY);
1399 lp = reg_prev_sub;
1400 while (*lp != NUL)
1401 regc(*lp++);
1402 regc(NUL);
1403 if (*reg_prev_sub != NUL)
1404 {
1405 *flagp |= HASWIDTH;
1406 if ((lp - reg_prev_sub) == 1)
1407 *flagp |= SIMPLE;
1408 }
1409 }
1410 else
Bram Moolenaare29a27f2021-07-20 21:07:36 +02001411 EMSG_RET_NULL(_(e_no_previous_substitute_regular_expression));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001412 break;
1413
1414 case Magic('1'):
1415 case Magic('2'):
1416 case Magic('3'):
1417 case Magic('4'):
1418 case Magic('5'):
1419 case Magic('6'):
1420 case Magic('7'):
1421 case Magic('8'):
1422 case Magic('9'):
1423 {
1424 int refnum;
1425
1426 refnum = c - Magic('0');
1427 if (!seen_endbrace(refnum))
1428 return NULL;
1429 ret = regnode(BACKREF + refnum);
1430 }
1431 break;
1432
1433 case Magic('z'):
1434 {
1435 c = no_Magic(getchr());
1436 switch (c)
1437 {
1438#ifdef FEAT_SYN_HL
1439 case '(': if ((reg_do_extmatch & REX_SET) == 0)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001440 EMSG_RET_NULL(_(e_z_not_allowed_here));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001441 if (one_exactly)
1442 EMSG_ONE_RET_NULL;
1443 ret = reg(REG_ZPAREN, &flags);
1444 if (ret == NULL)
1445 return NULL;
1446 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH);
1447 re_has_z = REX_SET;
1448 break;
1449
1450 case '1':
1451 case '2':
1452 case '3':
1453 case '4':
1454 case '5':
1455 case '6':
1456 case '7':
1457 case '8':
1458 case '9': if ((reg_do_extmatch & REX_USE) == 0)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001459 EMSG_RET_NULL(_(e_z1_z9_not_allowed_here));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001460 ret = regnode(ZREF + c - '0');
1461 re_has_z = REX_USE;
1462 break;
1463#endif
1464
1465 case 's': ret = regnode(MOPEN + 0);
1466 if (re_mult_next("\\zs") == FAIL)
1467 return NULL;
1468 break;
1469
1470 case 'e': ret = regnode(MCLOSE + 0);
1471 if (re_mult_next("\\ze") == FAIL)
1472 return NULL;
1473 break;
1474
Bram Moolenaarb2810f12022-01-08 21:38:52 +00001475 default: EMSG_RET_NULL(_(e_invalid_character_after_bsl_z));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001476 }
1477 }
1478 break;
1479
1480 case Magic('%'):
1481 {
1482 c = no_Magic(getchr());
1483 switch (c)
1484 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001485 // () without a back reference
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001486 case '(':
1487 if (one_exactly)
1488 EMSG_ONE_RET_NULL;
1489 ret = reg(REG_NPAREN, &flags);
1490 if (ret == NULL)
1491 return NULL;
1492 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
1493 break;
1494
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001495 // Catch \%^ and \%$ regardless of where they appear in the
1496 // pattern -- regardless of whether or not it makes sense.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001497 case '^':
1498 ret = regnode(RE_BOF);
1499 break;
1500
1501 case '$':
1502 ret = regnode(RE_EOF);
1503 break;
1504
1505 case '#':
1506 ret = regnode(CURSOR);
1507 break;
1508
1509 case 'V':
1510 ret = regnode(RE_VISUAL);
1511 break;
1512
1513 case 'C':
1514 ret = regnode(RE_COMPOSING);
1515 break;
1516
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001517 // \%[abc]: Emit as a list of branches, all ending at the last
1518 // branch which matches nothing.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001519 case '[':
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001520 if (one_exactly) // doesn't nest
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001521 EMSG_ONE_RET_NULL;
1522 {
1523 char_u *lastbranch;
1524 char_u *lastnode = NULL;
1525 char_u *br;
1526
1527 ret = NULL;
1528 while ((c = getchr()) != ']')
1529 {
1530 if (c == NUL)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001531 EMSG2_RET_NULL(_(e_missing_sb_after_str),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001532 reg_magic == MAGIC_ALL);
1533 br = regnode(BRANCH);
1534 if (ret == NULL)
1535 ret = br;
1536 else
1537 {
1538 regtail(lastnode, br);
1539 if (reg_toolong)
1540 return NULL;
1541 }
1542
1543 ungetchr();
1544 one_exactly = TRUE;
1545 lastnode = regatom(flagp);
1546 one_exactly = FALSE;
1547 if (lastnode == NULL)
1548 return NULL;
1549 }
1550 if (ret == NULL)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001551 EMSG2_RET_NULL(_(e_empty_str_brackets),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001552 reg_magic == MAGIC_ALL);
1553 lastbranch = regnode(BRANCH);
1554 br = regnode(NOTHING);
1555 if (ret != JUST_CALC_SIZE)
1556 {
1557 regtail(lastnode, br);
1558 regtail(lastbranch, br);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001559 // connect all branches to the NOTHING
1560 // branch at the end
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001561 for (br = ret; br != lastnode; )
1562 {
1563 if (OP(br) == BRANCH)
1564 {
1565 regtail(br, lastbranch);
1566 if (reg_toolong)
1567 return NULL;
1568 br = OPERAND(br);
1569 }
1570 else
1571 br = regnext(br);
1572 }
1573 }
1574 *flagp &= ~(HASWIDTH | SIMPLE);
1575 break;
1576 }
1577
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001578 case 'd': // %d123 decimal
1579 case 'o': // %o123 octal
1580 case 'x': // %xab hex 2
1581 case 'u': // %uabcd hex 4
1582 case 'U': // %U1234abcd hex 8
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001583 {
1584 long i;
1585
1586 switch (c)
1587 {
1588 case 'd': i = getdecchrs(); break;
1589 case 'o': i = getoctchrs(); break;
1590 case 'x': i = gethexchrs(2); break;
1591 case 'u': i = gethexchrs(4); break;
1592 case 'U': i = gethexchrs(8); break;
1593 default: i = -1; break;
1594 }
1595
1596 if (i < 0 || i > INT_MAX)
1597 EMSG2_RET_NULL(
Bram Moolenaara6f79292022-01-04 21:30:47 +00001598 _(e_invalid_character_after_str_2),
1599 reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001600 if (use_multibytecode(i))
1601 ret = regnode(MULTIBYTECODE);
1602 else
1603 ret = regnode(EXACTLY);
1604 if (i == 0)
1605 regc(0x0a);
1606 else
1607 regmbc(i);
1608 regc(NUL);
1609 *flagp |= HASWIDTH;
1610 break;
1611 }
1612
1613 default:
1614 if (VIM_ISDIGIT(c) || c == '<' || c == '>'
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001615 || c == '\'' || c == '.')
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001616 {
1617 long_u n = 0;
1618 int cmp;
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001619 int cur = FALSE;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001620
1621 cmp = c;
1622 if (cmp == '<' || cmp == '>')
1623 c = getchr();
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001624 if (no_Magic(c) == '.')
1625 {
1626 cur = TRUE;
1627 c = getchr();
1628 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001629 while (VIM_ISDIGIT(c))
1630 {
1631 n = n * 10 + (c - '0');
1632 c = getchr();
1633 }
1634 if (c == '\'' && n == 0)
1635 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001636 // "\%'m", "\%<'m" and "\%>'m": Mark
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001637 c = getchr();
1638 ret = regnode(RE_MARK);
1639 if (ret == JUST_CALC_SIZE)
1640 regsize += 2;
1641 else
1642 {
1643 *regcode++ = c;
1644 *regcode++ = cmp;
1645 }
1646 break;
1647 }
1648 else if (c == 'l' || c == 'c' || c == 'v')
1649 {
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001650 if (cur && n)
1651 {
1652 semsg(_(e_regexp_number_after_dot_pos_search), no_Magic(c));
1653 rc_did_emsg = TRUE;
1654 return NULL;
1655 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001656 if (c == 'l')
1657 {
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001658 if (cur)
1659 n = curwin->w_cursor.lnum;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001660 ret = regnode(RE_LNUM);
1661 if (save_prev_at_start)
1662 at_start = TRUE;
1663 }
1664 else if (c == 'c')
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001665 {
1666 if (cur)
1667 {
1668 n = curwin->w_cursor.col;
1669 n++;
1670 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001671 ret = regnode(RE_COL);
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001672 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001673 else
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001674 {
1675 if (cur)
1676 {
1677 colnr_T vcol = 0;
1678
1679 getvvcol(curwin, &curwin->w_cursor,
1680 NULL, NULL, &vcol);
1681 ++vcol;
1682 n = vcol;
1683 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001684 ret = regnode(RE_VCOL);
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001685 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001686 if (ret == JUST_CALC_SIZE)
1687 regsize += 5;
1688 else
1689 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001690 // put the number and the optional
1691 // comparator after the opcode
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001692 regcode = re_put_long(regcode, n);
1693 *regcode++ = cmp;
1694 }
1695 break;
1696 }
1697 }
1698
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001699 EMSG2_RET_NULL(_(e_invalid_character_after_str),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001700 reg_magic == MAGIC_ALL);
1701 }
1702 }
1703 break;
1704
1705 case Magic('['):
1706collection:
1707 {
1708 char_u *lp;
1709
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001710 // If there is no matching ']', we assume the '[' is a normal
1711 // character. This makes 'incsearch' and ":help [" work.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001712 lp = skip_anyof(regparse);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001713 if (*lp == ']') // there is a matching ']'
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001714 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001715 int startc = -1; // > 0 when next '-' is a range
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001716 int endc;
1717
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001718 // In a character class, different parsing rules apply.
1719 // Not even \ is special anymore, nothing is.
1720 if (*regparse == '^') // Complement of range.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001721 {
1722 ret = regnode(ANYBUT + extra);
1723 regparse++;
1724 }
1725 else
1726 ret = regnode(ANYOF + extra);
1727
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001728 // At the start ']' and '-' mean the literal character.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001729 if (*regparse == ']' || *regparse == '-')
1730 {
1731 startc = *regparse;
1732 regc(*regparse++);
1733 }
1734
1735 while (*regparse != NUL && *regparse != ']')
1736 {
1737 if (*regparse == '-')
1738 {
1739 ++regparse;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001740 // The '-' is not used for a range at the end and
1741 // after or before a '\n'.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001742 if (*regparse == ']' || *regparse == NUL
1743 || startc == -1
1744 || (regparse[0] == '\\' && regparse[1] == 'n'))
1745 {
1746 regc('-');
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001747 startc = '-'; // [--x] is a range
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001748 }
1749 else
1750 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001751 // Also accept "a-[.z.]"
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001752 endc = 0;
1753 if (*regparse == '[')
1754 endc = get_coll_element(&regparse);
1755 if (endc == 0)
1756 {
1757 if (has_mbyte)
1758 endc = mb_ptr2char_adv(&regparse);
1759 else
1760 endc = *regparse++;
1761 }
1762
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001763 // Handle \o40, \x20 and \u20AC style sequences
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001764 if (endc == '\\' && !reg_cpo_lit && !reg_cpo_bsl)
1765 endc = coll_get_char();
1766
1767 if (startc > endc)
Bram Moolenaar677658a2022-01-05 16:09:06 +00001768 EMSG_RET_NULL(_(e_reverse_range_in_character_class));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001769 if (has_mbyte && ((*mb_char2len)(startc) > 1
1770 || (*mb_char2len)(endc) > 1))
1771 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001772 // Limit to a range of 256 chars.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001773 if (endc > startc + 256)
Bram Moolenaar677658a2022-01-05 16:09:06 +00001774 EMSG_RET_NULL(_(e_range_too_large_in_character_class));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001775 while (++startc <= endc)
1776 regmbc(startc);
1777 }
1778 else
1779 {
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001780 while (++startc <= endc)
Bram Moolenaar424bcae2022-01-31 14:59:41 +00001781 regc(startc);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001782 }
1783 startc = -1;
1784 }
1785 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001786 // Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
1787 // accepts "\t", "\e", etc., but only when the 'l' flag in
1788 // 'cpoptions' is not included.
1789 // Posix doesn't recognize backslash at all.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001790 else if (*regparse == '\\'
1791 && !reg_cpo_bsl
1792 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
1793 || (!reg_cpo_lit
1794 && vim_strchr(REGEXP_ABBR,
1795 regparse[1]) != NULL)))
1796 {
1797 regparse++;
1798 if (*regparse == 'n')
1799 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001800 // '\n' in range: also match NL
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001801 if (ret != JUST_CALC_SIZE)
1802 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001803 // Using \n inside [^] does not change what
1804 // matches. "[^\n]" is the same as ".".
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001805 if (*ret == ANYOF)
1806 {
1807 *ret = ANYOF + ADD_NL;
1808 *flagp |= HASNL;
1809 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001810 // else: must have had a \n already
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001811 }
1812 regparse++;
1813 startc = -1;
1814 }
1815 else if (*regparse == 'd'
1816 || *regparse == 'o'
1817 || *regparse == 'x'
1818 || *regparse == 'u'
1819 || *regparse == 'U')
1820 {
1821 startc = coll_get_char();
1822 if (startc == 0)
1823 regc(0x0a);
1824 else
1825 regmbc(startc);
1826 }
1827 else
1828 {
1829 startc = backslash_trans(*regparse++);
1830 regc(startc);
1831 }
1832 }
1833 else if (*regparse == '[')
1834 {
1835 int c_class;
1836 int cu;
1837
1838 c_class = get_char_class(&regparse);
1839 startc = -1;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001840 // Characters assumed to be 8 bits!
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001841 switch (c_class)
1842 {
1843 case CLASS_NONE:
1844 c_class = get_equi_class(&regparse);
1845 if (c_class != 0)
1846 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001847 // produce equivalence class
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001848 reg_equi_class(c_class);
1849 }
1850 else if ((c_class =
1851 get_coll_element(&regparse)) != 0)
1852 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001853 // produce a collating element
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001854 regmbc(c_class);
1855 }
1856 else
1857 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001858 // literal '[', allow [[-x] as a range
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001859 startc = *regparse++;
1860 regc(startc);
1861 }
1862 break;
1863 case CLASS_ALNUM:
1864 for (cu = 1; cu < 128; cu++)
1865 if (isalnum(cu))
1866 regmbc(cu);
1867 break;
1868 case CLASS_ALPHA:
1869 for (cu = 1; cu < 128; cu++)
1870 if (isalpha(cu))
1871 regmbc(cu);
1872 break;
1873 case CLASS_BLANK:
1874 regc(' ');
1875 regc('\t');
1876 break;
1877 case CLASS_CNTRL:
1878 for (cu = 1; cu <= 127; cu++)
1879 if (iscntrl(cu))
1880 regmbc(cu);
1881 break;
1882 case CLASS_DIGIT:
1883 for (cu = 1; cu <= 127; cu++)
1884 if (VIM_ISDIGIT(cu))
1885 regmbc(cu);
1886 break;
1887 case CLASS_GRAPH:
1888 for (cu = 1; cu <= 127; cu++)
1889 if (isgraph(cu))
1890 regmbc(cu);
1891 break;
1892 case CLASS_LOWER:
1893 for (cu = 1; cu <= 255; cu++)
1894 if (MB_ISLOWER(cu) && cu != 170
1895 && cu != 186)
1896 regmbc(cu);
1897 break;
1898 case CLASS_PRINT:
1899 for (cu = 1; cu <= 255; cu++)
1900 if (vim_isprintc(cu))
1901 regmbc(cu);
1902 break;
1903 case CLASS_PUNCT:
1904 for (cu = 1; cu < 128; cu++)
1905 if (ispunct(cu))
1906 regmbc(cu);
1907 break;
1908 case CLASS_SPACE:
1909 for (cu = 9; cu <= 13; cu++)
1910 regc(cu);
1911 regc(' ');
1912 break;
1913 case CLASS_UPPER:
1914 for (cu = 1; cu <= 255; cu++)
1915 if (MB_ISUPPER(cu))
1916 regmbc(cu);
1917 break;
1918 case CLASS_XDIGIT:
1919 for (cu = 1; cu <= 255; cu++)
1920 if (vim_isxdigit(cu))
1921 regmbc(cu);
1922 break;
1923 case CLASS_TAB:
1924 regc('\t');
1925 break;
1926 case CLASS_RETURN:
1927 regc('\r');
1928 break;
1929 case CLASS_BACKSPACE:
1930 regc('\b');
1931 break;
1932 case CLASS_ESCAPE:
1933 regc('\033');
1934 break;
1935 case CLASS_IDENT:
1936 for (cu = 1; cu <= 255; cu++)
1937 if (vim_isIDc(cu))
1938 regmbc(cu);
1939 break;
1940 case CLASS_KEYWORD:
1941 for (cu = 1; cu <= 255; cu++)
1942 if (reg_iswordc(cu))
1943 regmbc(cu);
1944 break;
1945 case CLASS_FNAME:
1946 for (cu = 1; cu <= 255; cu++)
1947 if (vim_isfilec(cu))
1948 regmbc(cu);
1949 break;
1950 }
1951 }
1952 else
1953 {
1954 if (has_mbyte)
1955 {
1956 int len;
1957
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001958 // produce a multibyte character, including any
1959 // following composing characters
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001960 startc = mb_ptr2char(regparse);
1961 len = (*mb_ptr2len)(regparse);
1962 if (enc_utf8 && utf_char2len(startc) != len)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001963 startc = -1; // composing chars
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001964 while (--len >= 0)
1965 regc(*regparse++);
1966 }
1967 else
1968 {
1969 startc = *regparse++;
1970 regc(startc);
1971 }
1972 }
1973 }
1974 regc(NUL);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001975 prevchr_len = 1; // last char was the ']'
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001976 if (*regparse != ']')
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001977 EMSG_RET_NULL(_(e_too_many_brackets)); // Cannot happen?
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001978 skipchr(); // let's be friends with the lexer again
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001979 *flagp |= HASWIDTH | SIMPLE;
1980 break;
1981 }
1982 else if (reg_strict)
Bram Moolenaar677658a2022-01-05 16:09:06 +00001983 EMSG2_RET_NULL(_(e_missing_rsb_after_str_lsb),
1984 reg_magic > MAGIC_OFF);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001985 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001986 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001987
1988 default:
1989 {
1990 int len;
1991
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001992 // A multi-byte character is handled as a separate atom if it's
1993 // before a multi and when it's a composing char.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001994 if (use_multibytecode(c))
1995 {
1996do_multibyte:
1997 ret = regnode(MULTIBYTECODE);
1998 regmbc(c);
1999 *flagp |= HASWIDTH | SIMPLE;
2000 break;
2001 }
2002
2003 ret = regnode(EXACTLY);
2004
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002005 // Append characters as long as:
2006 // - there is no following multi, we then need the character in
2007 // front of it as a single character operand
2008 // - not running into a Magic character
2009 // - "one_exactly" is not set
2010 // But always emit at least one character. Might be a Multi,
2011 // e.g., a "[" without matching "]".
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002012 for (len = 0; c != NUL && (len == 0
2013 || (re_multi_type(peekchr()) == NOT_MULTI
2014 && !one_exactly
2015 && !is_Magic(c))); ++len)
2016 {
2017 c = no_Magic(c);
2018 if (has_mbyte)
2019 {
2020 regmbc(c);
2021 if (enc_utf8)
2022 {
2023 int l;
2024
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002025 // Need to get composing character too.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002026 for (;;)
2027 {
2028 l = utf_ptr2len(regparse);
2029 if (!UTF_COMPOSINGLIKE(regparse, regparse + l))
2030 break;
2031 regmbc(utf_ptr2char(regparse));
2032 skipchr();
2033 }
2034 }
2035 }
2036 else
2037 regc(c);
2038 c = getchr();
2039 }
2040 ungetchr();
2041
2042 regc(NUL);
2043 *flagp |= HASWIDTH;
2044 if (len == 1)
2045 *flagp |= SIMPLE;
2046 }
2047 break;
2048 }
2049
2050 return ret;
2051}
2052
2053/*
2054 * Parse something followed by possible [*+=].
2055 *
2056 * Note that the branching code sequences used for = and the general cases
2057 * of * and + are somewhat optimized: they use the same NOTHING node as
2058 * both the endmarker for their branch list and the body of the last branch.
2059 * It might seem that this node could be dispensed with entirely, but the
2060 * endmarker role is not redundant.
2061 */
2062 static char_u *
2063regpiece(int *flagp)
2064{
2065 char_u *ret;
2066 int op;
2067 char_u *next;
2068 int flags;
2069 long minval;
2070 long maxval;
2071
2072 ret = regatom(&flags);
2073 if (ret == NULL)
2074 return NULL;
2075
2076 op = peekchr();
2077 if (re_multi_type(op) == NOT_MULTI)
2078 {
2079 *flagp = flags;
2080 return ret;
2081 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002082 // default flags
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002083 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
2084
2085 skipchr();
2086 switch (op)
2087 {
2088 case Magic('*'):
2089 if (flags & SIMPLE)
2090 reginsert(STAR, ret);
2091 else
2092 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002093 // Emit x* as (x&|), where & means "self".
2094 reginsert(BRANCH, ret); // Either x
2095 regoptail(ret, regnode(BACK)); // and loop
2096 regoptail(ret, ret); // back
2097 regtail(ret, regnode(BRANCH)); // or
2098 regtail(ret, regnode(NOTHING)); // null.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002099 }
2100 break;
2101
2102 case Magic('+'):
2103 if (flags & SIMPLE)
2104 reginsert(PLUS, ret);
2105 else
2106 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002107 // Emit x+ as x(&|), where & means "self".
2108 next = regnode(BRANCH); // Either
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002109 regtail(ret, next);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002110 regtail(regnode(BACK), ret); // loop back
2111 regtail(next, regnode(BRANCH)); // or
2112 regtail(ret, regnode(NOTHING)); // null.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002113 }
2114 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
2115 break;
2116
2117 case Magic('@'):
2118 {
2119 int lop = END;
2120 long nr;
2121
2122 nr = getdecchrs();
2123 switch (no_Magic(getchr()))
2124 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002125 case '=': lop = MATCH; break; // \@=
2126 case '!': lop = NOMATCH; break; // \@!
2127 case '>': lop = SUBPAT; break; // \@>
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002128 case '<': switch (no_Magic(getchr()))
2129 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002130 case '=': lop = BEHIND; break; // \@<=
2131 case '!': lop = NOBEHIND; break; // \@<!
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002132 }
2133 }
2134 if (lop == END)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002135 EMSG2_RET_NULL(_(e_invalid_character_after_str_at),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002136 reg_magic == MAGIC_ALL);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002137 // Look behind must match with behind_pos.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002138 if (lop == BEHIND || lop == NOBEHIND)
2139 {
2140 regtail(ret, regnode(BHPOS));
2141 *flagp |= HASLOOKBH;
2142 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002143 regtail(ret, regnode(END)); // operand ends
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002144 if (lop == BEHIND || lop == NOBEHIND)
2145 {
2146 if (nr < 0)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002147 nr = 0; // no limit is same as zero limit
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002148 reginsert_nr(lop, nr, ret);
2149 }
2150 else
2151 reginsert(lop, ret);
2152 break;
2153 }
2154
2155 case Magic('?'):
2156 case Magic('='):
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002157 // Emit x= as (x|)
2158 reginsert(BRANCH, ret); // Either x
2159 regtail(ret, regnode(BRANCH)); // or
2160 next = regnode(NOTHING); // null.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002161 regtail(ret, next);
2162 regoptail(ret, next);
2163 break;
2164
2165 case Magic('{'):
2166 if (!read_limits(&minval, &maxval))
2167 return NULL;
2168 if (flags & SIMPLE)
2169 {
2170 reginsert(BRACE_SIMPLE, ret);
2171 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
2172 }
2173 else
2174 {
2175 if (num_complex_braces >= 10)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002176 EMSG2_RET_NULL(_(e_too_many_complex_str_curly),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002177 reg_magic == MAGIC_ALL);
2178 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
2179 regoptail(ret, regnode(BACK));
2180 regoptail(ret, ret);
2181 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
2182 ++num_complex_braces;
2183 }
2184 if (minval > 0 && maxval > 0)
2185 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
2186 break;
2187 }
2188 if (re_multi_type(peekchr()) != NOT_MULTI)
2189 {
2190 // Can't have a multi follow a multi.
2191 if (peekchr() == Magic('*'))
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00002192 EMSG2_RET_NULL(_(e_nested_str), reg_magic >= MAGIC_ON);
2193 EMSG3_RET_NULL(_(e_nested_str_chr), reg_magic == MAGIC_ALL,
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002194 no_Magic(peekchr()));
2195 }
2196
2197 return ret;
2198}
2199
2200/*
2201 * Parse one alternative of an | or & operator.
2202 * Implements the concatenation operator.
2203 */
2204 static char_u *
2205regconcat(int *flagp)
2206{
2207 char_u *first = NULL;
2208 char_u *chain = NULL;
2209 char_u *latest;
2210 int flags;
2211 int cont = TRUE;
2212
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002213 *flagp = WORST; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002214
2215 while (cont)
2216 {
2217 switch (peekchr())
2218 {
2219 case NUL:
2220 case Magic('|'):
2221 case Magic('&'):
2222 case Magic(')'):
2223 cont = FALSE;
2224 break;
2225 case Magic('Z'):
2226 regflags |= RF_ICOMBINE;
2227 skipchr_keepstart();
2228 break;
2229 case Magic('c'):
2230 regflags |= RF_ICASE;
2231 skipchr_keepstart();
2232 break;
2233 case Magic('C'):
2234 regflags |= RF_NOICASE;
2235 skipchr_keepstart();
2236 break;
2237 case Magic('v'):
2238 reg_magic = MAGIC_ALL;
2239 skipchr_keepstart();
2240 curchr = -1;
2241 break;
2242 case Magic('m'):
2243 reg_magic = MAGIC_ON;
2244 skipchr_keepstart();
2245 curchr = -1;
2246 break;
2247 case Magic('M'):
2248 reg_magic = MAGIC_OFF;
2249 skipchr_keepstart();
2250 curchr = -1;
2251 break;
2252 case Magic('V'):
2253 reg_magic = MAGIC_NONE;
2254 skipchr_keepstart();
2255 curchr = -1;
2256 break;
2257 default:
2258 latest = regpiece(&flags);
2259 if (latest == NULL || reg_toolong)
2260 return NULL;
2261 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002262 if (chain == NULL) // First piece.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002263 *flagp |= flags & SPSTART;
2264 else
2265 regtail(chain, latest);
2266 chain = latest;
2267 if (first == NULL)
2268 first = latest;
2269 break;
2270 }
2271 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002272 if (first == NULL) // Loop ran zero times.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002273 first = regnode(NOTHING);
2274 return first;
2275}
2276
2277/*
2278 * Parse one alternative of an | operator.
2279 * Implements the & operator.
2280 */
2281 static char_u *
2282regbranch(int *flagp)
2283{
2284 char_u *ret;
2285 char_u *chain = NULL;
2286 char_u *latest;
2287 int flags;
2288
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002289 *flagp = WORST | HASNL; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002290
2291 ret = regnode(BRANCH);
2292 for (;;)
2293 {
2294 latest = regconcat(&flags);
2295 if (latest == NULL)
2296 return NULL;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002297 // If one of the branches has width, the whole thing has. If one of
2298 // the branches anchors at start-of-line, the whole thing does.
2299 // If one of the branches uses look-behind, the whole thing does.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002300 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002301 // If one of the branches doesn't match a line-break, the whole thing
2302 // doesn't.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002303 *flagp &= ~HASNL | (flags & HASNL);
2304 if (chain != NULL)
2305 regtail(chain, latest);
2306 if (peekchr() != Magic('&'))
2307 break;
2308 skipchr();
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002309 regtail(latest, regnode(END)); // operand ends
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002310 if (reg_toolong)
2311 break;
2312 reginsert(MATCH, latest);
2313 chain = latest;
2314 }
2315
2316 return ret;
2317}
2318
2319/*
2320 * Parse regular expression, i.e. main body or parenthesized thing.
2321 *
2322 * Caller must absorb opening parenthesis.
2323 *
2324 * Combining parenthesis handling with the base level of regular expression
2325 * is a trifle forced, but the need to tie the tails of the branches to what
2326 * follows makes it hard to avoid.
2327 */
2328 static char_u *
2329reg(
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002330 int paren, // REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002331 int *flagp)
2332{
2333 char_u *ret;
2334 char_u *br;
2335 char_u *ender;
2336 int parno = 0;
2337 int flags;
2338
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002339 *flagp = HASWIDTH; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002340
2341#ifdef FEAT_SYN_HL
2342 if (paren == REG_ZPAREN)
2343 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002344 // Make a ZOPEN node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002345 if (regnzpar >= NSUBEXP)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002346 EMSG_RET_NULL(_(e_too_many_z));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002347 parno = regnzpar;
2348 regnzpar++;
2349 ret = regnode(ZOPEN + parno);
2350 }
2351 else
2352#endif
2353 if (paren == REG_PAREN)
2354 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002355 // Make a MOPEN node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002356 if (regnpar >= NSUBEXP)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002357 EMSG2_RET_NULL(_(e_too_many_str_open), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002358 parno = regnpar;
2359 ++regnpar;
2360 ret = regnode(MOPEN + parno);
2361 }
2362 else if (paren == REG_NPAREN)
2363 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002364 // Make a NOPEN node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002365 ret = regnode(NOPEN);
2366 }
2367 else
2368 ret = NULL;
2369
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002370 // Pick up the branches, linking them together.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002371 br = regbranch(&flags);
2372 if (br == NULL)
2373 return NULL;
2374 if (ret != NULL)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002375 regtail(ret, br); // [MZ]OPEN -> first.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002376 else
2377 ret = br;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002378 // If one of the branches can be zero-width, the whole thing can.
2379 // If one of the branches has * at start or matches a line-break, the
2380 // whole thing can.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002381 if (!(flags & HASWIDTH))
2382 *flagp &= ~HASWIDTH;
2383 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
2384 while (peekchr() == Magic('|'))
2385 {
2386 skipchr();
2387 br = regbranch(&flags);
2388 if (br == NULL || reg_toolong)
2389 return NULL;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002390 regtail(ret, br); // BRANCH -> BRANCH.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002391 if (!(flags & HASWIDTH))
2392 *flagp &= ~HASWIDTH;
2393 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
2394 }
2395
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002396 // Make a closing node, and hook it on the end.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002397 ender = regnode(
2398#ifdef FEAT_SYN_HL
2399 paren == REG_ZPAREN ? ZCLOSE + parno :
2400#endif
2401 paren == REG_PAREN ? MCLOSE + parno :
2402 paren == REG_NPAREN ? NCLOSE : END);
2403 regtail(ret, ender);
2404
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002405 // Hook the tails of the branches to the closing node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002406 for (br = ret; br != NULL; br = regnext(br))
2407 regoptail(br, ender);
2408
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002409 // Check for proper termination.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002410 if (paren != REG_NOPAREN && getchr() != Magic(')'))
2411 {
2412#ifdef FEAT_SYN_HL
2413 if (paren == REG_ZPAREN)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002414 EMSG_RET_NULL(_(e_unmatched_z));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002415 else
2416#endif
2417 if (paren == REG_NPAREN)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002418 EMSG2_RET_NULL(_(e_unmatched_str_percent_open), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002419 else
Bram Moolenaard8e44472021-07-21 22:20:33 +02002420 EMSG2_RET_NULL(_(e_unmatched_str_open), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002421 }
2422 else if (paren == REG_NOPAREN && peekchr() != NUL)
2423 {
2424 if (curchr == Magic(')'))
Bram Moolenaard8e44472021-07-21 22:20:33 +02002425 EMSG2_RET_NULL(_(e_unmatched_str_close), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002426 else
Bram Moolenaar74409f62022-01-01 15:58:22 +00002427 EMSG_RET_NULL(_(e_trailing_characters)); // "Can't happen".
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002428 // NOTREACHED
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002429 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002430 // Here we set the flag allowing back references to this set of
2431 // parentheses.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002432 if (paren == REG_PAREN)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002433 had_endbrace[parno] = TRUE; // have seen the close paren
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002434 return ret;
2435}
2436
2437/*
2438 * bt_regcomp() - compile a regular expression into internal code for the
2439 * traditional back track matcher.
2440 * Returns the program in allocated space. Returns NULL for an error.
2441 *
2442 * We can't allocate space until we know how big the compiled form will be,
2443 * but we can't compile it (and thus know how big it is) until we've got a
2444 * place to put the code. So we cheat: we compile it twice, once with code
2445 * generation turned off and size counting turned on, and once "for real".
2446 * This also means that we don't allocate space until we are sure that the
2447 * thing really will compile successfully, and we never have to move the
2448 * code and thus invalidate pointers into it. (Note that it has to be in
2449 * one piece because vim_free() must be able to free it all.)
2450 *
2451 * Whether upper/lower case is to be ignored is decided when executing the
2452 * program, it does not matter here.
2453 *
2454 * Beware that the optimization-preparation code in here knows about some
2455 * of the structure of the compiled regexp.
2456 * "re_flags": RE_MAGIC and/or RE_STRING.
2457 */
2458 static regprog_T *
2459bt_regcomp(char_u *expr, int re_flags)
2460{
2461 bt_regprog_T *r;
2462 char_u *scan;
2463 char_u *longest;
2464 int len;
2465 int flags;
2466
2467 if (expr == NULL)
Bram Moolenaare29a27f2021-07-20 21:07:36 +02002468 IEMSG_RET_NULL(_(e_null_argument));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002469
2470 init_class_tab();
2471
2472 // First pass: determine size, legality.
2473 regcomp_start(expr, re_flags);
2474 regcode = JUST_CALC_SIZE;
2475 regc(REGMAGIC);
2476 if (reg(REG_NOPAREN, &flags) == NULL)
2477 return NULL;
2478
2479 // Allocate space.
2480 r = alloc(offsetof(bt_regprog_T, program) + regsize);
2481 if (r == NULL)
2482 return NULL;
2483 r->re_in_use = FALSE;
2484
2485 // Second pass: emit code.
2486 regcomp_start(expr, re_flags);
2487 regcode = r->program;
2488 regc(REGMAGIC);
2489 if (reg(REG_NOPAREN, &flags) == NULL || reg_toolong)
2490 {
2491 vim_free(r);
2492 if (reg_toolong)
Bram Moolenaareaaac012022-01-02 17:00:40 +00002493 EMSG_RET_NULL(_(e_pattern_too_long));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002494 return NULL;
2495 }
2496
2497 // Dig out information for optimizations.
2498 r->regstart = NUL; // Worst-case defaults.
2499 r->reganch = 0;
2500 r->regmust = NULL;
2501 r->regmlen = 0;
2502 r->regflags = regflags;
2503 if (flags & HASNL)
2504 r->regflags |= RF_HASNL;
2505 if (flags & HASLOOKBH)
2506 r->regflags |= RF_LOOKBH;
2507#ifdef FEAT_SYN_HL
2508 // Remember whether this pattern has any \z specials in it.
2509 r->reghasz = re_has_z;
2510#endif
2511 scan = r->program + 1; // First BRANCH.
2512 if (OP(regnext(scan)) == END) // Only one top-level choice.
2513 {
2514 scan = OPERAND(scan);
2515
2516 // Starting-point info.
2517 if (OP(scan) == BOL || OP(scan) == RE_BOF)
2518 {
2519 r->reganch++;
2520 scan = regnext(scan);
2521 }
2522
2523 if (OP(scan) == EXACTLY)
2524 {
2525 if (has_mbyte)
2526 r->regstart = (*mb_ptr2char)(OPERAND(scan));
2527 else
2528 r->regstart = *OPERAND(scan);
2529 }
2530 else if ((OP(scan) == BOW
2531 || OP(scan) == EOW
2532 || OP(scan) == NOTHING
2533 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
2534 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
2535 && OP(regnext(scan)) == EXACTLY)
2536 {
2537 if (has_mbyte)
2538 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
2539 else
2540 r->regstart = *OPERAND(regnext(scan));
2541 }
2542
2543 // If there's something expensive in the r.e., find the longest
2544 // literal string that must appear and make it the regmust. Resolve
2545 // ties in favor of later strings, since the regstart check works
2546 // with the beginning of the r.e. and avoiding duplication
2547 // strengthens checking. Not a strong reason, but sufficient in the
2548 // absence of others.
2549
2550 // When the r.e. starts with BOW, it is faster to look for a regmust
2551 // first. Used a lot for "#" and "*" commands. (Added by mool).
2552 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
2553 && !(flags & HASNL))
2554 {
2555 longest = NULL;
2556 len = 0;
2557 for (; scan != NULL; scan = regnext(scan))
2558 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
2559 {
2560 longest = OPERAND(scan);
2561 len = (int)STRLEN(OPERAND(scan));
2562 }
2563 r->regmust = longest;
2564 r->regmlen = len;
2565 }
2566 }
2567#ifdef BT_REGEXP_DUMP
2568 regdump(expr, r);
2569#endif
2570 r->engine = &bt_regengine;
2571 return (regprog_T *)r;
2572}
2573
2574#if defined(FEAT_SYN_HL) || defined(PROTO)
2575/*
2576 * Check if during the previous call to vim_regcomp the EOL item "$" has been
2577 * found. This is messy, but it works fine.
2578 */
2579 int
2580vim_regcomp_had_eol(void)
2581{
2582 return had_eol;
2583}
2584#endif
2585
2586/*
2587 * Get a number after a backslash that is inside [].
2588 * When nothing is recognized return a backslash.
2589 */
2590 static int
2591coll_get_char(void)
2592{
2593 long nr = -1;
2594
2595 switch (*regparse++)
2596 {
2597 case 'd': nr = getdecchrs(); break;
2598 case 'o': nr = getoctchrs(); break;
2599 case 'x': nr = gethexchrs(2); break;
2600 case 'u': nr = gethexchrs(4); break;
2601 case 'U': nr = gethexchrs(8); break;
2602 }
2603 if (nr < 0 || nr > INT_MAX)
2604 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002605 // If getting the number fails be backwards compatible: the character
2606 // is a backslash.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002607 --regparse;
2608 nr = '\\';
2609 }
2610 return nr;
2611}
2612
2613/*
2614 * Free a compiled regexp program, returned by bt_regcomp().
2615 */
2616 static void
2617bt_regfree(regprog_T *prog)
2618{
2619 vim_free(prog);
2620}
2621
2622#define ADVANCE_REGINPUT() MB_PTR_ADV(rex.input)
2623
2624/*
2625 * The arguments from BRACE_LIMITS are stored here. They are actually local
2626 * to regmatch(), but they are here to reduce the amount of stack space used
2627 * (it can be called recursively many times).
2628 */
2629static long bl_minval;
2630static long bl_maxval;
2631
2632/*
2633 * Save the input line and position in a regsave_T.
2634 */
2635 static void
2636reg_save(regsave_T *save, garray_T *gap)
2637{
2638 if (REG_MULTI)
2639 {
2640 save->rs_u.pos.col = (colnr_T)(rex.input - rex.line);
2641 save->rs_u.pos.lnum = rex.lnum;
2642 }
2643 else
2644 save->rs_u.ptr = rex.input;
2645 save->rs_len = gap->ga_len;
2646}
2647
2648/*
2649 * Restore the input line and position from a regsave_T.
2650 */
2651 static void
2652reg_restore(regsave_T *save, garray_T *gap)
2653{
2654 if (REG_MULTI)
2655 {
2656 if (rex.lnum != save->rs_u.pos.lnum)
2657 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002658 // only call reg_getline() when the line number changed to save
2659 // a bit of time
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002660 rex.lnum = save->rs_u.pos.lnum;
2661 rex.line = reg_getline(rex.lnum);
2662 }
2663 rex.input = rex.line + save->rs_u.pos.col;
2664 }
2665 else
2666 rex.input = save->rs_u.ptr;
2667 gap->ga_len = save->rs_len;
2668}
2669
2670/*
2671 * Return TRUE if current position is equal to saved position.
2672 */
2673 static int
2674reg_save_equal(regsave_T *save)
2675{
2676 if (REG_MULTI)
2677 return rex.lnum == save->rs_u.pos.lnum
2678 && rex.input == rex.line + save->rs_u.pos.col;
2679 return rex.input == save->rs_u.ptr;
2680}
2681
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002682// Save the sub-expressions before attempting a match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002683#define save_se(savep, posp, pp) \
2684 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
2685
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002686// After a failed match restore the sub-expressions.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002687#define restore_se(savep, posp, pp) { \
2688 if (REG_MULTI) \
2689 *(posp) = (savep)->se_u.pos; \
2690 else \
2691 *(pp) = (savep)->se_u.ptr; }
2692
2693/*
2694 * Tentatively set the sub-expression start to the current position (after
2695 * calling regmatch() they will have changed). Need to save the existing
2696 * values for when there is no match.
2697 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
2698 * depending on REG_MULTI.
2699 */
2700 static void
2701save_se_multi(save_se_T *savep, lpos_T *posp)
2702{
2703 savep->se_u.pos = *posp;
2704 posp->lnum = rex.lnum;
2705 posp->col = (colnr_T)(rex.input - rex.line);
2706}
2707
2708 static void
2709save_se_one(save_se_T *savep, char_u **pp)
2710{
2711 savep->se_u.ptr = *pp;
2712 *pp = rex.input;
2713}
2714
2715/*
2716 * regrepeat - repeatedly match something simple, return how many.
2717 * Advances rex.input (and rex.lnum) to just after the matched chars.
2718 */
2719 static int
2720regrepeat(
2721 char_u *p,
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002722 long maxcount) // maximum number of matches allowed
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002723{
2724 long count = 0;
2725 char_u *scan;
2726 char_u *opnd;
2727 int mask;
2728 int testval = 0;
2729
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002730 scan = rex.input; // Make local copy of rex.input for speed.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002731 opnd = OPERAND(p);
2732 switch (OP(p))
2733 {
2734 case ANY:
2735 case ANY + ADD_NL:
2736 while (count < maxcount)
2737 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002738 // Matching anything means we continue until end-of-line (or
2739 // end-of-file for ANY + ADD_NL), only limited by maxcount.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002740 while (*scan != NUL && count < maxcount)
2741 {
2742 ++count;
2743 MB_PTR_ADV(scan);
2744 }
2745 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2746 || rex.reg_line_lbr || count == maxcount)
2747 break;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002748 ++count; // count the line-break
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002749 reg_nextline();
2750 scan = rex.input;
2751 if (got_int)
2752 break;
2753 }
2754 break;
2755
2756 case IDENT:
2757 case IDENT + ADD_NL:
2758 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002759 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002760 case SIDENT:
2761 case SIDENT + ADD_NL:
2762 while (count < maxcount)
2763 {
2764 if (vim_isIDc(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan)))
2765 {
2766 MB_PTR_ADV(scan);
2767 }
2768 else if (*scan == NUL)
2769 {
2770 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2771 || rex.reg_line_lbr)
2772 break;
2773 reg_nextline();
2774 scan = rex.input;
2775 if (got_int)
2776 break;
2777 }
2778 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2779 ++scan;
2780 else
2781 break;
2782 ++count;
2783 }
2784 break;
2785
2786 case KWORD:
2787 case KWORD + ADD_NL:
2788 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002789 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002790 case SKWORD:
2791 case SKWORD + ADD_NL:
2792 while (count < maxcount)
2793 {
2794 if (vim_iswordp_buf(scan, rex.reg_buf)
2795 && (testval || !VIM_ISDIGIT(*scan)))
2796 {
2797 MB_PTR_ADV(scan);
2798 }
2799 else if (*scan == NUL)
2800 {
2801 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2802 || rex.reg_line_lbr)
2803 break;
2804 reg_nextline();
2805 scan = rex.input;
2806 if (got_int)
2807 break;
2808 }
2809 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2810 ++scan;
2811 else
2812 break;
2813 ++count;
2814 }
2815 break;
2816
2817 case FNAME:
2818 case FNAME + ADD_NL:
2819 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002820 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002821 case SFNAME:
2822 case SFNAME + ADD_NL:
2823 while (count < maxcount)
2824 {
2825 if (vim_isfilec(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan)))
2826 {
2827 MB_PTR_ADV(scan);
2828 }
2829 else if (*scan == NUL)
2830 {
2831 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2832 || rex.reg_line_lbr)
2833 break;
2834 reg_nextline();
2835 scan = rex.input;
2836 if (got_int)
2837 break;
2838 }
2839 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2840 ++scan;
2841 else
2842 break;
2843 ++count;
2844 }
2845 break;
2846
2847 case PRINT:
2848 case PRINT + ADD_NL:
2849 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002850 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002851 case SPRINT:
2852 case SPRINT + ADD_NL:
2853 while (count < maxcount)
2854 {
2855 if (*scan == NUL)
2856 {
2857 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2858 || rex.reg_line_lbr)
2859 break;
2860 reg_nextline();
2861 scan = rex.input;
2862 if (got_int)
2863 break;
2864 }
2865 else if (vim_isprintc(PTR2CHAR(scan)) == 1
2866 && (testval || !VIM_ISDIGIT(*scan)))
2867 {
2868 MB_PTR_ADV(scan);
2869 }
2870 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2871 ++scan;
2872 else
2873 break;
2874 ++count;
2875 }
2876 break;
2877
2878 case WHITE:
2879 case WHITE + ADD_NL:
2880 testval = mask = RI_WHITE;
2881do_class:
2882 while (count < maxcount)
2883 {
2884 int l;
2885
2886 if (*scan == NUL)
2887 {
2888 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2889 || rex.reg_line_lbr)
2890 break;
2891 reg_nextline();
2892 scan = rex.input;
2893 if (got_int)
2894 break;
2895 }
2896 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1)
2897 {
2898 if (testval != 0)
2899 break;
2900 scan += l;
2901 }
2902 else if ((class_tab[*scan] & mask) == testval)
2903 ++scan;
2904 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2905 ++scan;
2906 else
2907 break;
2908 ++count;
2909 }
2910 break;
2911
2912 case NWHITE:
2913 case NWHITE + ADD_NL:
2914 mask = RI_WHITE;
2915 goto do_class;
2916 case DIGIT:
2917 case DIGIT + ADD_NL:
2918 testval = mask = RI_DIGIT;
2919 goto do_class;
2920 case NDIGIT:
2921 case NDIGIT + ADD_NL:
2922 mask = RI_DIGIT;
2923 goto do_class;
2924 case HEX:
2925 case HEX + ADD_NL:
2926 testval = mask = RI_HEX;
2927 goto do_class;
2928 case NHEX:
2929 case NHEX + ADD_NL:
2930 mask = RI_HEX;
2931 goto do_class;
2932 case OCTAL:
2933 case OCTAL + ADD_NL:
2934 testval = mask = RI_OCTAL;
2935 goto do_class;
2936 case NOCTAL:
2937 case NOCTAL + ADD_NL:
2938 mask = RI_OCTAL;
2939 goto do_class;
2940 case WORD:
2941 case WORD + ADD_NL:
2942 testval = mask = RI_WORD;
2943 goto do_class;
2944 case NWORD:
2945 case NWORD + ADD_NL:
2946 mask = RI_WORD;
2947 goto do_class;
2948 case HEAD:
2949 case HEAD + ADD_NL:
2950 testval = mask = RI_HEAD;
2951 goto do_class;
2952 case NHEAD:
2953 case NHEAD + ADD_NL:
2954 mask = RI_HEAD;
2955 goto do_class;
2956 case ALPHA:
2957 case ALPHA + ADD_NL:
2958 testval = mask = RI_ALPHA;
2959 goto do_class;
2960 case NALPHA:
2961 case NALPHA + ADD_NL:
2962 mask = RI_ALPHA;
2963 goto do_class;
2964 case LOWER:
2965 case LOWER + ADD_NL:
2966 testval = mask = RI_LOWER;
2967 goto do_class;
2968 case NLOWER:
2969 case NLOWER + ADD_NL:
2970 mask = RI_LOWER;
2971 goto do_class;
2972 case UPPER:
2973 case UPPER + ADD_NL:
2974 testval = mask = RI_UPPER;
2975 goto do_class;
2976 case NUPPER:
2977 case NUPPER + ADD_NL:
2978 mask = RI_UPPER;
2979 goto do_class;
2980
2981 case EXACTLY:
2982 {
2983 int cu, cl;
2984
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002985 // This doesn't do a multi-byte character, because a MULTIBYTECODE
2986 // would have been used for it. It does handle single-byte
2987 // characters, such as latin1.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002988 if (rex.reg_ic)
2989 {
2990 cu = MB_TOUPPER(*opnd);
2991 cl = MB_TOLOWER(*opnd);
2992 while (count < maxcount && (*scan == cu || *scan == cl))
2993 {
2994 count++;
2995 scan++;
2996 }
2997 }
2998 else
2999 {
3000 cu = *opnd;
3001 while (count < maxcount && *scan == cu)
3002 {
3003 count++;
3004 scan++;
3005 }
3006 }
3007 break;
3008 }
3009
3010 case MULTIBYTECODE:
3011 {
3012 int i, len, cf = 0;
3013
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003014 // Safety check (just in case 'encoding' was changed since
3015 // compiling the program).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003016 if ((len = (*mb_ptr2len)(opnd)) > 1)
3017 {
3018 if (rex.reg_ic && enc_utf8)
3019 cf = utf_fold(utf_ptr2char(opnd));
3020 while (count < maxcount && (*mb_ptr2len)(scan) >= len)
3021 {
3022 for (i = 0; i < len; ++i)
3023 if (opnd[i] != scan[i])
3024 break;
3025 if (i < len && (!rex.reg_ic || !enc_utf8
3026 || utf_fold(utf_ptr2char(scan)) != cf))
3027 break;
3028 scan += len;
3029 ++count;
3030 }
3031 }
3032 }
3033 break;
3034
3035 case ANYOF:
3036 case ANYOF + ADD_NL:
3037 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003038 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003039
3040 case ANYBUT:
3041 case ANYBUT + ADD_NL:
3042 while (count < maxcount)
3043 {
3044 int len;
3045
3046 if (*scan == NUL)
3047 {
3048 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
3049 || rex.reg_line_lbr)
3050 break;
3051 reg_nextline();
3052 scan = rex.input;
3053 if (got_int)
3054 break;
3055 }
3056 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
3057 ++scan;
3058 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1)
3059 {
3060 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
3061 break;
3062 scan += len;
3063 }
3064 else
3065 {
3066 if ((cstrchr(opnd, *scan) == NULL) == testval)
3067 break;
3068 ++scan;
3069 }
3070 ++count;
3071 }
3072 break;
3073
3074 case NEWL:
3075 while (count < maxcount
3076 && ((*scan == NUL && rex.lnum <= rex.reg_maxline
3077 && !rex.reg_line_lbr && REG_MULTI)
3078 || (*scan == '\n' && rex.reg_line_lbr)))
3079 {
3080 count++;
3081 if (rex.reg_line_lbr)
3082 ADVANCE_REGINPUT();
3083 else
3084 reg_nextline();
3085 scan = rex.input;
3086 if (got_int)
3087 break;
3088 }
3089 break;
3090
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003091 default: // Oh dear. Called inappropriately.
Bram Moolenaare29a27f2021-07-20 21:07:36 +02003092 iemsg(_(e_corrupted_regexp_program));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003093#ifdef DEBUG
3094 printf("Called regrepeat with op code %d\n", OP(p));
3095#endif
3096 break;
3097 }
3098
3099 rex.input = scan;
3100
3101 return (int)count;
3102}
3103
3104/*
3105 * Push an item onto the regstack.
3106 * Returns pointer to new item. Returns NULL when out of memory.
3107 */
3108 static regitem_T *
3109regstack_push(regstate_T state, char_u *scan)
3110{
3111 regitem_T *rp;
3112
3113 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
3114 {
Bram Moolenaar74409f62022-01-01 15:58:22 +00003115 emsg(_(e_pattern_uses_more_memory_than_maxmempattern));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003116 return NULL;
3117 }
3118 if (ga_grow(&regstack, sizeof(regitem_T)) == FAIL)
3119 return NULL;
3120
3121 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len);
3122 rp->rs_state = state;
3123 rp->rs_scan = scan;
3124
3125 regstack.ga_len += sizeof(regitem_T);
3126 return rp;
3127}
3128
3129/*
3130 * Pop an item from the regstack.
3131 */
3132 static void
3133regstack_pop(char_u **scan)
3134{
3135 regitem_T *rp;
3136
3137 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
3138 *scan = rp->rs_scan;
3139
3140 regstack.ga_len -= sizeof(regitem_T);
3141}
3142
3143/*
3144 * Save the current subexpr to "bp", so that they can be restored
3145 * later by restore_subexpr().
3146 */
3147 static void
3148save_subexpr(regbehind_T *bp)
3149{
3150 int i;
3151
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003152 // When "rex.need_clear_subexpr" is set we don't need to save the values,
3153 // only remember that this flag needs to be set again when restoring.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003154 bp->save_need_clear_subexpr = rex.need_clear_subexpr;
3155 if (!rex.need_clear_subexpr)
3156 {
3157 for (i = 0; i < NSUBEXP; ++i)
3158 {
3159 if (REG_MULTI)
3160 {
3161 bp->save_start[i].se_u.pos = rex.reg_startpos[i];
3162 bp->save_end[i].se_u.pos = rex.reg_endpos[i];
3163 }
3164 else
3165 {
3166 bp->save_start[i].se_u.ptr = rex.reg_startp[i];
3167 bp->save_end[i].se_u.ptr = rex.reg_endp[i];
3168 }
3169 }
3170 }
3171}
3172
3173/*
3174 * Restore the subexpr from "bp".
3175 */
3176 static void
3177restore_subexpr(regbehind_T *bp)
3178{
3179 int i;
3180
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003181 // Only need to restore saved values when they are not to be cleared.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003182 rex.need_clear_subexpr = bp->save_need_clear_subexpr;
3183 if (!rex.need_clear_subexpr)
3184 {
3185 for (i = 0; i < NSUBEXP; ++i)
3186 {
3187 if (REG_MULTI)
3188 {
3189 rex.reg_startpos[i] = bp->save_start[i].se_u.pos;
3190 rex.reg_endpos[i] = bp->save_end[i].se_u.pos;
3191 }
3192 else
3193 {
3194 rex.reg_startp[i] = bp->save_start[i].se_u.ptr;
3195 rex.reg_endp[i] = bp->save_end[i].se_u.ptr;
3196 }
3197 }
3198 }
3199}
3200
3201/*
3202 * regmatch - main matching routine
3203 *
3204 * Conceptually the strategy is simple: Check to see whether the current node
3205 * matches, push an item onto the regstack and loop to see whether the rest
3206 * matches, and then act accordingly. In practice we make some effort to
3207 * avoid using the regstack, in particular by going through "ordinary" nodes
3208 * (that don't need to know whether the rest of the match failed) by a nested
3209 * loop.
3210 *
3211 * Returns TRUE when there is a match. Leaves rex.input and rex.lnum just after
3212 * the last matched character.
3213 * Returns FALSE when there is no match. Leaves rex.input and rex.lnum in an
3214 * undefined state!
3215 */
3216 static int
3217regmatch(
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003218 char_u *scan, // Current node.
3219 proftime_T *tm UNUSED, // timeout limit or NULL
3220 int *timed_out UNUSED) // flag set on timeout or NULL
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003221{
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003222 char_u *next; // Next node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003223 int op;
3224 int c;
3225 regitem_T *rp;
3226 int no;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003227 int status; // one of the RA_ values:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003228#ifdef FEAT_RELTIME
3229 int tm_count = 0;
3230#endif
3231
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003232 // Make "regstack" and "backpos" empty. They are allocated and freed in
3233 // bt_regexec_both() to reduce malloc()/free() calls.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003234 regstack.ga_len = 0;
3235 backpos.ga_len = 0;
3236
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003237 // Repeat until "regstack" is empty.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003238 for (;;)
3239 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003240 // Some patterns may take a long time to match, e.g., "\([a-z]\+\)\+Q".
3241 // Allow interrupting them with CTRL-C.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003242 fast_breakcheck();
3243
3244#ifdef DEBUG
3245 if (scan != NULL && regnarrate)
3246 {
3247 mch_errmsg((char *)regprop(scan));
3248 mch_errmsg("(\n");
3249 }
3250#endif
3251
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003252 // Repeat for items that can be matched sequentially, without using the
3253 // regstack.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003254 for (;;)
3255 {
3256 if (got_int || scan == NULL)
3257 {
3258 status = RA_FAIL;
3259 break;
3260 }
3261#ifdef FEAT_RELTIME
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003262 // Check for timeout once in a 100 times to avoid overhead.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003263 if (tm != NULL && ++tm_count == 100)
3264 {
3265 tm_count = 0;
3266 if (profile_passed_limit(tm))
3267 {
3268 if (timed_out != NULL)
3269 *timed_out = TRUE;
3270 status = RA_FAIL;
3271 break;
3272 }
3273 }
3274#endif
3275 status = RA_CONT;
3276
3277#ifdef DEBUG
3278 if (regnarrate)
3279 {
3280 mch_errmsg((char *)regprop(scan));
3281 mch_errmsg("...\n");
3282# ifdef FEAT_SYN_HL
3283 if (re_extmatch_in != NULL)
3284 {
3285 int i;
3286
3287 mch_errmsg(_("External submatches:\n"));
3288 for (i = 0; i < NSUBEXP; i++)
3289 {
3290 mch_errmsg(" \"");
3291 if (re_extmatch_in->matches[i] != NULL)
3292 mch_errmsg((char *)re_extmatch_in->matches[i]);
3293 mch_errmsg("\"\n");
3294 }
3295 }
3296# endif
3297 }
3298#endif
3299 next = regnext(scan);
3300
3301 op = OP(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003302 // Check for character class with NL added.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003303 if (!rex.reg_line_lbr && WITH_NL(op) && REG_MULTI
3304 && *rex.input == NUL && rex.lnum <= rex.reg_maxline)
3305 {
3306 reg_nextline();
3307 }
3308 else if (rex.reg_line_lbr && WITH_NL(op) && *rex.input == '\n')
3309 {
3310 ADVANCE_REGINPUT();
3311 }
3312 else
3313 {
3314 if (WITH_NL(op))
3315 op -= ADD_NL;
3316 if (has_mbyte)
3317 c = (*mb_ptr2char)(rex.input);
3318 else
3319 c = *rex.input;
3320 switch (op)
3321 {
3322 case BOL:
3323 if (rex.input != rex.line)
3324 status = RA_NOMATCH;
3325 break;
3326
3327 case EOL:
3328 if (c != NUL)
3329 status = RA_NOMATCH;
3330 break;
3331
3332 case RE_BOF:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003333 // We're not at the beginning of the file when below the first
3334 // line where we started, not at the start of the line or we
3335 // didn't start at the first line of the buffer.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003336 if (rex.lnum != 0 || rex.input != rex.line
3337 || (REG_MULTI && rex.reg_firstlnum > 1))
3338 status = RA_NOMATCH;
3339 break;
3340
3341 case RE_EOF:
3342 if (rex.lnum != rex.reg_maxline || c != NUL)
3343 status = RA_NOMATCH;
3344 break;
3345
3346 case CURSOR:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003347 // Check if the buffer is in a window and compare the
3348 // rex.reg_win->w_cursor position to the match position.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003349 if (rex.reg_win == NULL
3350 || (rex.lnum + rex.reg_firstlnum
3351 != rex.reg_win->w_cursor.lnum)
3352 || ((colnr_T)(rex.input - rex.line)
3353 != rex.reg_win->w_cursor.col))
3354 status = RA_NOMATCH;
3355 break;
3356
3357 case RE_MARK:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003358 // Compare the mark position to the match position.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003359 {
3360 int mark = OPERAND(scan)[0];
3361 int cmp = OPERAND(scan)[1];
3362 pos_T *pos;
3363
3364 pos = getmark_buf(rex.reg_buf, mark, FALSE);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003365 if (pos == NULL // mark doesn't exist
Bram Moolenaar872bee52021-05-24 22:56:15 +02003366 || pos->lnum <= 0) // mark isn't set in reg_buf
3367 {
3368 status = RA_NOMATCH;
3369 }
3370 else
3371 {
3372 colnr_T pos_col = pos->lnum == rex.lnum + rex.reg_firstlnum
3373 && pos->col == MAXCOL
3374 ? (colnr_T)STRLEN(reg_getline(
3375 pos->lnum - rex.reg_firstlnum))
3376 : pos->col;
3377
3378 if ((pos->lnum == rex.lnum + rex.reg_firstlnum
3379 ? (pos_col == (colnr_T)(rex.input - rex.line)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003380 ? (cmp == '<' || cmp == '>')
Bram Moolenaar872bee52021-05-24 22:56:15 +02003381 : (pos_col < (colnr_T)(rex.input - rex.line)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003382 ? cmp != '>'
3383 : cmp != '<'))
3384 : (pos->lnum < rex.lnum + rex.reg_firstlnum
3385 ? cmp != '>'
3386 : cmp != '<')))
3387 status = RA_NOMATCH;
Bram Moolenaar872bee52021-05-24 22:56:15 +02003388 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003389 }
3390 break;
3391
3392 case RE_VISUAL:
3393 if (!reg_match_visual())
3394 status = RA_NOMATCH;
3395 break;
3396
3397 case RE_LNUM:
3398 if (!REG_MULTI || !re_num_cmp((long_u)(rex.lnum + rex.reg_firstlnum),
3399 scan))
3400 status = RA_NOMATCH;
3401 break;
3402
3403 case RE_COL:
3404 if (!re_num_cmp((long_u)(rex.input - rex.line) + 1, scan))
3405 status = RA_NOMATCH;
3406 break;
3407
3408 case RE_VCOL:
3409 if (!re_num_cmp((long_u)win_linetabsize(
3410 rex.reg_win == NULL ? curwin : rex.reg_win,
3411 rex.line, (colnr_T)(rex.input - rex.line)) + 1, scan))
3412 status = RA_NOMATCH;
3413 break;
3414
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003415 case BOW: // \<word; rex.input points to w
3416 if (c == NUL) // Can't match at end of line
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003417 status = RA_NOMATCH;
3418 else if (has_mbyte)
3419 {
3420 int this_class;
3421
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003422 // Get class of current and previous char (if it exists).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003423 this_class = mb_get_class_buf(rex.input, rex.reg_buf);
3424 if (this_class <= 1)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003425 status = RA_NOMATCH; // not on a word at all
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003426 else if (reg_prev_class() == this_class)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003427 status = RA_NOMATCH; // previous char is in same word
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003428 }
3429 else
3430 {
3431 if (!vim_iswordc_buf(c, rex.reg_buf) || (rex.input > rex.line
3432 && vim_iswordc_buf(rex.input[-1], rex.reg_buf)))
3433 status = RA_NOMATCH;
3434 }
3435 break;
3436
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003437 case EOW: // word\>; rex.input points after d
3438 if (rex.input == rex.line) // Can't match at start of line
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003439 status = RA_NOMATCH;
3440 else if (has_mbyte)
3441 {
3442 int this_class, prev_class;
3443
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003444 // Get class of current and previous char (if it exists).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003445 this_class = mb_get_class_buf(rex.input, rex.reg_buf);
3446 prev_class = reg_prev_class();
3447 if (this_class == prev_class
3448 || prev_class == 0 || prev_class == 1)
3449 status = RA_NOMATCH;
3450 }
3451 else
3452 {
3453 if (!vim_iswordc_buf(rex.input[-1], rex.reg_buf)
3454 || (rex.input[0] != NUL
3455 && vim_iswordc_buf(c, rex.reg_buf)))
3456 status = RA_NOMATCH;
3457 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003458 break; // Matched with EOW
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003459
3460 case ANY:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003461 // ANY does not match new lines.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003462 if (c == NUL)
3463 status = RA_NOMATCH;
3464 else
3465 ADVANCE_REGINPUT();
3466 break;
3467
3468 case IDENT:
3469 if (!vim_isIDc(c))
3470 status = RA_NOMATCH;
3471 else
3472 ADVANCE_REGINPUT();
3473 break;
3474
3475 case SIDENT:
3476 if (VIM_ISDIGIT(*rex.input) || !vim_isIDc(c))
3477 status = RA_NOMATCH;
3478 else
3479 ADVANCE_REGINPUT();
3480 break;
3481
3482 case KWORD:
3483 if (!vim_iswordp_buf(rex.input, rex.reg_buf))
3484 status = RA_NOMATCH;
3485 else
3486 ADVANCE_REGINPUT();
3487 break;
3488
3489 case SKWORD:
3490 if (VIM_ISDIGIT(*rex.input)
3491 || !vim_iswordp_buf(rex.input, rex.reg_buf))
3492 status = RA_NOMATCH;
3493 else
3494 ADVANCE_REGINPUT();
3495 break;
3496
3497 case FNAME:
3498 if (!vim_isfilec(c))
3499 status = RA_NOMATCH;
3500 else
3501 ADVANCE_REGINPUT();
3502 break;
3503
3504 case SFNAME:
3505 if (VIM_ISDIGIT(*rex.input) || !vim_isfilec(c))
3506 status = RA_NOMATCH;
3507 else
3508 ADVANCE_REGINPUT();
3509 break;
3510
3511 case PRINT:
3512 if (!vim_isprintc(PTR2CHAR(rex.input)))
3513 status = RA_NOMATCH;
3514 else
3515 ADVANCE_REGINPUT();
3516 break;
3517
3518 case SPRINT:
3519 if (VIM_ISDIGIT(*rex.input) || !vim_isprintc(PTR2CHAR(rex.input)))
3520 status = RA_NOMATCH;
3521 else
3522 ADVANCE_REGINPUT();
3523 break;
3524
3525 case WHITE:
3526 if (!VIM_ISWHITE(c))
3527 status = RA_NOMATCH;
3528 else
3529 ADVANCE_REGINPUT();
3530 break;
3531
3532 case NWHITE:
3533 if (c == NUL || VIM_ISWHITE(c))
3534 status = RA_NOMATCH;
3535 else
3536 ADVANCE_REGINPUT();
3537 break;
3538
3539 case DIGIT:
3540 if (!ri_digit(c))
3541 status = RA_NOMATCH;
3542 else
3543 ADVANCE_REGINPUT();
3544 break;
3545
3546 case NDIGIT:
3547 if (c == NUL || ri_digit(c))
3548 status = RA_NOMATCH;
3549 else
3550 ADVANCE_REGINPUT();
3551 break;
3552
3553 case HEX:
3554 if (!ri_hex(c))
3555 status = RA_NOMATCH;
3556 else
3557 ADVANCE_REGINPUT();
3558 break;
3559
3560 case NHEX:
3561 if (c == NUL || ri_hex(c))
3562 status = RA_NOMATCH;
3563 else
3564 ADVANCE_REGINPUT();
3565 break;
3566
3567 case OCTAL:
3568 if (!ri_octal(c))
3569 status = RA_NOMATCH;
3570 else
3571 ADVANCE_REGINPUT();
3572 break;
3573
3574 case NOCTAL:
3575 if (c == NUL || ri_octal(c))
3576 status = RA_NOMATCH;
3577 else
3578 ADVANCE_REGINPUT();
3579 break;
3580
3581 case WORD:
3582 if (!ri_word(c))
3583 status = RA_NOMATCH;
3584 else
3585 ADVANCE_REGINPUT();
3586 break;
3587
3588 case NWORD:
3589 if (c == NUL || ri_word(c))
3590 status = RA_NOMATCH;
3591 else
3592 ADVANCE_REGINPUT();
3593 break;
3594
3595 case HEAD:
3596 if (!ri_head(c))
3597 status = RA_NOMATCH;
3598 else
3599 ADVANCE_REGINPUT();
3600 break;
3601
3602 case NHEAD:
3603 if (c == NUL || ri_head(c))
3604 status = RA_NOMATCH;
3605 else
3606 ADVANCE_REGINPUT();
3607 break;
3608
3609 case ALPHA:
3610 if (!ri_alpha(c))
3611 status = RA_NOMATCH;
3612 else
3613 ADVANCE_REGINPUT();
3614 break;
3615
3616 case NALPHA:
3617 if (c == NUL || ri_alpha(c))
3618 status = RA_NOMATCH;
3619 else
3620 ADVANCE_REGINPUT();
3621 break;
3622
3623 case LOWER:
3624 if (!ri_lower(c))
3625 status = RA_NOMATCH;
3626 else
3627 ADVANCE_REGINPUT();
3628 break;
3629
3630 case NLOWER:
3631 if (c == NUL || ri_lower(c))
3632 status = RA_NOMATCH;
3633 else
3634 ADVANCE_REGINPUT();
3635 break;
3636
3637 case UPPER:
3638 if (!ri_upper(c))
3639 status = RA_NOMATCH;
3640 else
3641 ADVANCE_REGINPUT();
3642 break;
3643
3644 case NUPPER:
3645 if (c == NUL || ri_upper(c))
3646 status = RA_NOMATCH;
3647 else
3648 ADVANCE_REGINPUT();
3649 break;
3650
3651 case EXACTLY:
3652 {
3653 int len;
3654 char_u *opnd;
3655
3656 opnd = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003657 // Inline the first byte, for speed.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003658 if (*opnd != *rex.input
3659 && (!rex.reg_ic
3660 || (!enc_utf8
3661 && MB_TOLOWER(*opnd) != MB_TOLOWER(*rex.input))))
3662 status = RA_NOMATCH;
3663 else if (*opnd == NUL)
3664 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003665 // match empty string always works; happens when "~" is
3666 // empty.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003667 }
3668 else
3669 {
3670 if (opnd[1] == NUL && !(enc_utf8 && rex.reg_ic))
3671 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003672 len = 1; // matched a single byte above
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003673 }
3674 else
3675 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003676 // Need to match first byte again for multi-byte.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003677 len = (int)STRLEN(opnd);
3678 if (cstrncmp(opnd, rex.input, &len) != 0)
3679 status = RA_NOMATCH;
3680 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003681 // Check for following composing character, unless %C
3682 // follows (skips over all composing chars).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003683 if (status != RA_NOMATCH
3684 && enc_utf8
3685 && UTF_COMPOSINGLIKE(rex.input, rex.input + len)
3686 && !rex.reg_icombine
3687 && OP(next) != RE_COMPOSING)
3688 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003689 // raaron: This code makes a composing character get
3690 // ignored, which is the correct behavior (sometimes)
3691 // for voweled Hebrew texts.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003692 status = RA_NOMATCH;
3693 }
3694 if (status != RA_NOMATCH)
3695 rex.input += len;
3696 }
3697 }
3698 break;
3699
3700 case ANYOF:
3701 case ANYBUT:
3702 if (c == NUL)
3703 status = RA_NOMATCH;
3704 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF))
3705 status = RA_NOMATCH;
3706 else
3707 ADVANCE_REGINPUT();
3708 break;
3709
3710 case MULTIBYTECODE:
3711 if (has_mbyte)
3712 {
3713 int i, len;
3714 char_u *opnd;
3715 int opndc = 0, inpc;
3716
3717 opnd = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003718 // Safety check (just in case 'encoding' was changed since
3719 // compiling the program).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003720 if ((len = (*mb_ptr2len)(opnd)) < 2)
3721 {
3722 status = RA_NOMATCH;
3723 break;
3724 }
3725 if (enc_utf8)
3726 opndc = utf_ptr2char(opnd);
3727 if (enc_utf8 && utf_iscomposing(opndc))
3728 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003729 // When only a composing char is given match at any
3730 // position where that composing char appears.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003731 status = RA_NOMATCH;
3732 for (i = 0; rex.input[i] != NUL;
3733 i += utf_ptr2len(rex.input + i))
3734 {
3735 inpc = utf_ptr2char(rex.input + i);
3736 if (!utf_iscomposing(inpc))
3737 {
3738 if (i > 0)
3739 break;
3740 }
3741 else if (opndc == inpc)
3742 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003743 // Include all following composing chars.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003744 len = i + utfc_ptr2len(rex.input + i);
3745 status = RA_MATCH;
3746 break;
3747 }
3748 }
3749 }
3750 else
3751 for (i = 0; i < len; ++i)
3752 if (opnd[i] != rex.input[i])
3753 {
3754 status = RA_NOMATCH;
3755 break;
3756 }
3757 rex.input += len;
3758 }
3759 else
3760 status = RA_NOMATCH;
3761 break;
3762 case RE_COMPOSING:
3763 if (enc_utf8)
3764 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003765 // Skip composing characters.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003766 while (utf_iscomposing(utf_ptr2char(rex.input)))
3767 MB_CPTR_ADV(rex.input);
3768 }
3769 break;
3770
3771 case NOTHING:
3772 break;
3773
3774 case BACK:
3775 {
3776 int i;
3777 backpos_T *bp;
3778
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003779 // When we run into BACK we need to check if we don't keep
3780 // looping without matching any input. The second and later
3781 // times a BACK is encountered it fails if the input is still
3782 // at the same position as the previous time.
3783 // The positions are stored in "backpos" and found by the
3784 // current value of "scan", the position in the RE program.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003785 bp = (backpos_T *)backpos.ga_data;
3786 for (i = 0; i < backpos.ga_len; ++i)
3787 if (bp[i].bp_scan == scan)
3788 break;
3789 if (i == backpos.ga_len)
3790 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003791 // First time at this BACK, make room to store the pos.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003792 if (ga_grow(&backpos, 1) == FAIL)
3793 status = RA_FAIL;
3794 else
3795 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003796 // get "ga_data" again, it may have changed
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003797 bp = (backpos_T *)backpos.ga_data;
3798 bp[i].bp_scan = scan;
3799 ++backpos.ga_len;
3800 }
3801 }
3802 else if (reg_save_equal(&bp[i].bp_pos))
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003803 // Still at same position as last time, fail.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003804 status = RA_NOMATCH;
3805
3806 if (status != RA_FAIL && status != RA_NOMATCH)
3807 reg_save(&bp[i].bp_pos, &backpos);
3808 }
3809 break;
3810
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003811 case MOPEN + 0: // Match start: \zs
3812 case MOPEN + 1: // \(
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003813 case MOPEN + 2:
3814 case MOPEN + 3:
3815 case MOPEN + 4:
3816 case MOPEN + 5:
3817 case MOPEN + 6:
3818 case MOPEN + 7:
3819 case MOPEN + 8:
3820 case MOPEN + 9:
3821 {
3822 no = op - MOPEN;
3823 cleanup_subexpr();
3824 rp = regstack_push(RS_MOPEN, scan);
3825 if (rp == NULL)
3826 status = RA_FAIL;
3827 else
3828 {
3829 rp->rs_no = no;
3830 save_se(&rp->rs_un.sesave, &rex.reg_startpos[no],
3831 &rex.reg_startp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003832 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003833 }
3834 }
3835 break;
3836
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003837 case NOPEN: // \%(
3838 case NCLOSE: // \) after \%(
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003839 if (regstack_push(RS_NOPEN, scan) == NULL)
3840 status = RA_FAIL;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003841 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003842 break;
3843
3844#ifdef FEAT_SYN_HL
3845 case ZOPEN + 1:
3846 case ZOPEN + 2:
3847 case ZOPEN + 3:
3848 case ZOPEN + 4:
3849 case ZOPEN + 5:
3850 case ZOPEN + 6:
3851 case ZOPEN + 7:
3852 case ZOPEN + 8:
3853 case ZOPEN + 9:
3854 {
3855 no = op - ZOPEN;
3856 cleanup_zsubexpr();
3857 rp = regstack_push(RS_ZOPEN, scan);
3858 if (rp == NULL)
3859 status = RA_FAIL;
3860 else
3861 {
3862 rp->rs_no = no;
3863 save_se(&rp->rs_un.sesave, &reg_startzpos[no],
3864 &reg_startzp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003865 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003866 }
3867 }
3868 break;
3869#endif
3870
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003871 case MCLOSE + 0: // Match end: \ze
3872 case MCLOSE + 1: // \)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003873 case MCLOSE + 2:
3874 case MCLOSE + 3:
3875 case MCLOSE + 4:
3876 case MCLOSE + 5:
3877 case MCLOSE + 6:
3878 case MCLOSE + 7:
3879 case MCLOSE + 8:
3880 case MCLOSE + 9:
3881 {
3882 no = op - MCLOSE;
3883 cleanup_subexpr();
3884 rp = regstack_push(RS_MCLOSE, scan);
3885 if (rp == NULL)
3886 status = RA_FAIL;
3887 else
3888 {
3889 rp->rs_no = no;
3890 save_se(&rp->rs_un.sesave, &rex.reg_endpos[no],
3891 &rex.reg_endp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003892 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003893 }
3894 }
3895 break;
3896
3897#ifdef FEAT_SYN_HL
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003898 case ZCLOSE + 1: // \) after \z(
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003899 case ZCLOSE + 2:
3900 case ZCLOSE + 3:
3901 case ZCLOSE + 4:
3902 case ZCLOSE + 5:
3903 case ZCLOSE + 6:
3904 case ZCLOSE + 7:
3905 case ZCLOSE + 8:
3906 case ZCLOSE + 9:
3907 {
3908 no = op - ZCLOSE;
3909 cleanup_zsubexpr();
3910 rp = regstack_push(RS_ZCLOSE, scan);
3911 if (rp == NULL)
3912 status = RA_FAIL;
3913 else
3914 {
3915 rp->rs_no = no;
3916 save_se(&rp->rs_un.sesave, &reg_endzpos[no],
3917 &reg_endzp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003918 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003919 }
3920 }
3921 break;
3922#endif
3923
3924 case BACKREF + 1:
3925 case BACKREF + 2:
3926 case BACKREF + 3:
3927 case BACKREF + 4:
3928 case BACKREF + 5:
3929 case BACKREF + 6:
3930 case BACKREF + 7:
3931 case BACKREF + 8:
3932 case BACKREF + 9:
3933 {
3934 int len;
3935
3936 no = op - BACKREF;
3937 cleanup_subexpr();
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003938 if (!REG_MULTI) // Single-line regexp
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003939 {
3940 if (rex.reg_startp[no] == NULL || rex.reg_endp[no] == NULL)
3941 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003942 // Backref was not set: Match an empty string.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003943 len = 0;
3944 }
3945 else
3946 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003947 // Compare current input with back-ref in the same
3948 // line.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003949 len = (int)(rex.reg_endp[no] - rex.reg_startp[no]);
3950 if (cstrncmp(rex.reg_startp[no], rex.input, &len) != 0)
3951 status = RA_NOMATCH;
3952 }
3953 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003954 else // Multi-line regexp
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003955 {
3956 if (rex.reg_startpos[no].lnum < 0
3957 || rex.reg_endpos[no].lnum < 0)
3958 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003959 // Backref was not set: Match an empty string.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003960 len = 0;
3961 }
3962 else
3963 {
3964 if (rex.reg_startpos[no].lnum == rex.lnum
3965 && rex.reg_endpos[no].lnum == rex.lnum)
3966 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003967 // Compare back-ref within the current line.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003968 len = rex.reg_endpos[no].col
3969 - rex.reg_startpos[no].col;
3970 if (cstrncmp(rex.line + rex.reg_startpos[no].col,
3971 rex.input, &len) != 0)
3972 status = RA_NOMATCH;
3973 }
3974 else
3975 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003976 // Messy situation: Need to compare between two
3977 // lines.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003978 int r = match_with_backref(
3979 rex.reg_startpos[no].lnum,
3980 rex.reg_startpos[no].col,
3981 rex.reg_endpos[no].lnum,
3982 rex.reg_endpos[no].col,
3983 &len);
3984
3985 if (r != RA_MATCH)
3986 status = r;
3987 }
3988 }
3989 }
3990
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003991 // Matched the backref, skip over it.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003992 rex.input += len;
3993 }
3994 break;
3995
3996#ifdef FEAT_SYN_HL
3997 case ZREF + 1:
3998 case ZREF + 2:
3999 case ZREF + 3:
4000 case ZREF + 4:
4001 case ZREF + 5:
4002 case ZREF + 6:
4003 case ZREF + 7:
4004 case ZREF + 8:
4005 case ZREF + 9:
4006 {
4007 int len;
4008
4009 cleanup_zsubexpr();
4010 no = op - ZREF;
4011 if (re_extmatch_in != NULL
4012 && re_extmatch_in->matches[no] != NULL)
4013 {
4014 len = (int)STRLEN(re_extmatch_in->matches[no]);
4015 if (cstrncmp(re_extmatch_in->matches[no],
4016 rex.input, &len) != 0)
4017 status = RA_NOMATCH;
4018 else
4019 rex.input += len;
4020 }
4021 else
4022 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004023 // Backref was not set: Match an empty string.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004024 }
4025 }
4026 break;
4027#endif
4028
4029 case BRANCH:
4030 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004031 if (OP(next) != BRANCH) // No choice.
4032 next = OPERAND(scan); // Avoid recursion.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004033 else
4034 {
4035 rp = regstack_push(RS_BRANCH, scan);
4036 if (rp == NULL)
4037 status = RA_FAIL;
4038 else
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004039 status = RA_BREAK; // rest is below
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004040 }
4041 }
4042 break;
4043
4044 case BRACE_LIMITS:
4045 {
4046 if (OP(next) == BRACE_SIMPLE)
4047 {
4048 bl_minval = OPERAND_MIN(scan);
4049 bl_maxval = OPERAND_MAX(scan);
4050 }
4051 else if (OP(next) >= BRACE_COMPLEX
4052 && OP(next) < BRACE_COMPLEX + 10)
4053 {
4054 no = OP(next) - BRACE_COMPLEX;
4055 brace_min[no] = OPERAND_MIN(scan);
4056 brace_max[no] = OPERAND_MAX(scan);
4057 brace_count[no] = 0;
4058 }
4059 else
4060 {
4061 internal_error("BRACE_LIMITS");
4062 status = RA_FAIL;
4063 }
4064 }
4065 break;
4066
4067 case BRACE_COMPLEX + 0:
4068 case BRACE_COMPLEX + 1:
4069 case BRACE_COMPLEX + 2:
4070 case BRACE_COMPLEX + 3:
4071 case BRACE_COMPLEX + 4:
4072 case BRACE_COMPLEX + 5:
4073 case BRACE_COMPLEX + 6:
4074 case BRACE_COMPLEX + 7:
4075 case BRACE_COMPLEX + 8:
4076 case BRACE_COMPLEX + 9:
4077 {
4078 no = op - BRACE_COMPLEX;
4079 ++brace_count[no];
4080
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004081 // If not matched enough times yet, try one more
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004082 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
4083 ? brace_min[no] : brace_max[no]))
4084 {
4085 rp = regstack_push(RS_BRCPLX_MORE, scan);
4086 if (rp == NULL)
4087 status = RA_FAIL;
4088 else
4089 {
4090 rp->rs_no = no;
4091 reg_save(&rp->rs_un.regsave, &backpos);
4092 next = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004093 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004094 }
4095 break;
4096 }
4097
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004098 // If matched enough times, may try matching some more
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004099 if (brace_min[no] <= brace_max[no])
4100 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004101 // Range is the normal way around, use longest match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004102 if (brace_count[no] <= brace_max[no])
4103 {
4104 rp = regstack_push(RS_BRCPLX_LONG, scan);
4105 if (rp == NULL)
4106 status = RA_FAIL;
4107 else
4108 {
4109 rp->rs_no = no;
4110 reg_save(&rp->rs_un.regsave, &backpos);
4111 next = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004112 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004113 }
4114 }
4115 }
4116 else
4117 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004118 // Range is backwards, use shortest match first
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004119 if (brace_count[no] <= brace_min[no])
4120 {
4121 rp = regstack_push(RS_BRCPLX_SHORT, scan);
4122 if (rp == NULL)
4123 status = RA_FAIL;
4124 else
4125 {
4126 reg_save(&rp->rs_un.regsave, &backpos);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004127 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004128 }
4129 }
4130 }
4131 }
4132 break;
4133
4134 case BRACE_SIMPLE:
4135 case STAR:
4136 case PLUS:
4137 {
4138 regstar_T rst;
4139
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004140 // Lookahead to avoid useless match attempts when we know
4141 // what character comes next.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004142 if (OP(next) == EXACTLY)
4143 {
4144 rst.nextb = *OPERAND(next);
4145 if (rex.reg_ic)
4146 {
4147 if (MB_ISUPPER(rst.nextb))
4148 rst.nextb_ic = MB_TOLOWER(rst.nextb);
4149 else
4150 rst.nextb_ic = MB_TOUPPER(rst.nextb);
4151 }
4152 else
4153 rst.nextb_ic = rst.nextb;
4154 }
4155 else
4156 {
4157 rst.nextb = NUL;
4158 rst.nextb_ic = NUL;
4159 }
4160 if (op != BRACE_SIMPLE)
4161 {
4162 rst.minval = (op == STAR) ? 0 : 1;
4163 rst.maxval = MAX_LIMIT;
4164 }
4165 else
4166 {
4167 rst.minval = bl_minval;
4168 rst.maxval = bl_maxval;
4169 }
4170
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004171 // When maxval > minval, try matching as much as possible, up
4172 // to maxval. When maxval < minval, try matching at least the
4173 // minimal number (since the range is backwards, that's also
4174 // maxval!).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004175 rst.count = regrepeat(OPERAND(scan), rst.maxval);
4176 if (got_int)
4177 {
4178 status = RA_FAIL;
4179 break;
4180 }
4181 if (rst.minval <= rst.maxval
4182 ? rst.count >= rst.minval : rst.count >= rst.maxval)
4183 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004184 // It could match. Prepare for trying to match what
4185 // follows. The code is below. Parameters are stored in
4186 // a regstar_T on the regstack.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004187 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
4188 {
Bram Moolenaar74409f62022-01-01 15:58:22 +00004189 emsg(_(e_pattern_uses_more_memory_than_maxmempattern));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004190 status = RA_FAIL;
4191 }
4192 else if (ga_grow(&regstack, sizeof(regstar_T)) == FAIL)
4193 status = RA_FAIL;
4194 else
4195 {
4196 regstack.ga_len += sizeof(regstar_T);
4197 rp = regstack_push(rst.minval <= rst.maxval
4198 ? RS_STAR_LONG : RS_STAR_SHORT, scan);
4199 if (rp == NULL)
4200 status = RA_FAIL;
4201 else
4202 {
4203 *(((regstar_T *)rp) - 1) = rst;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004204 status = RA_BREAK; // skip the restore bits
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004205 }
4206 }
4207 }
4208 else
4209 status = RA_NOMATCH;
4210
4211 }
4212 break;
4213
4214 case NOMATCH:
4215 case MATCH:
4216 case SUBPAT:
4217 rp = regstack_push(RS_NOMATCH, scan);
4218 if (rp == NULL)
4219 status = RA_FAIL;
4220 else
4221 {
4222 rp->rs_no = op;
4223 reg_save(&rp->rs_un.regsave, &backpos);
4224 next = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004225 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004226 }
4227 break;
4228
4229 case BEHIND:
4230 case NOBEHIND:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004231 // Need a bit of room to store extra positions.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004232 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
4233 {
Bram Moolenaar74409f62022-01-01 15:58:22 +00004234 emsg(_(e_pattern_uses_more_memory_than_maxmempattern));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004235 status = RA_FAIL;
4236 }
4237 else if (ga_grow(&regstack, sizeof(regbehind_T)) == FAIL)
4238 status = RA_FAIL;
4239 else
4240 {
4241 regstack.ga_len += sizeof(regbehind_T);
4242 rp = regstack_push(RS_BEHIND1, scan);
4243 if (rp == NULL)
4244 status = RA_FAIL;
4245 else
4246 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004247 // Need to save the subexpr to be able to restore them
4248 // when there is a match but we don't use it.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004249 save_subexpr(((regbehind_T *)rp) - 1);
4250
4251 rp->rs_no = op;
4252 reg_save(&rp->rs_un.regsave, &backpos);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004253 // First try if what follows matches. If it does then we
4254 // check the behind match by looping.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004255 }
4256 }
4257 break;
4258
4259 case BHPOS:
4260 if (REG_MULTI)
4261 {
4262 if (behind_pos.rs_u.pos.col != (colnr_T)(rex.input - rex.line)
4263 || behind_pos.rs_u.pos.lnum != rex.lnum)
4264 status = RA_NOMATCH;
4265 }
4266 else if (behind_pos.rs_u.ptr != rex.input)
4267 status = RA_NOMATCH;
4268 break;
4269
4270 case NEWL:
4271 if ((c != NUL || !REG_MULTI || rex.lnum > rex.reg_maxline
4272 || rex.reg_line_lbr)
4273 && (c != '\n' || !rex.reg_line_lbr))
4274 status = RA_NOMATCH;
4275 else if (rex.reg_line_lbr)
4276 ADVANCE_REGINPUT();
4277 else
4278 reg_nextline();
4279 break;
4280
4281 case END:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004282 status = RA_MATCH; // Success!
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004283 break;
4284
4285 default:
Bram Moolenaare29a27f2021-07-20 21:07:36 +02004286 iemsg(_(e_corrupted_regexp_program));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004287#ifdef DEBUG
4288 printf("Illegal op code %d\n", op);
4289#endif
4290 status = RA_FAIL;
4291 break;
4292 }
4293 }
4294
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004295 // If we can't continue sequentially, break the inner loop.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004296 if (status != RA_CONT)
4297 break;
4298
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004299 // Continue in inner loop, advance to next item.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004300 scan = next;
4301
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004302 } // end of inner loop
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004303
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004304 // If there is something on the regstack execute the code for the state.
4305 // If the state is popped then loop and use the older state.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004306 while (regstack.ga_len > 0 && status != RA_FAIL)
4307 {
4308 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
4309 switch (rp->rs_state)
4310 {
4311 case RS_NOPEN:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004312 // Result is passed on as-is, simply pop the state.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004313 regstack_pop(&scan);
4314 break;
4315
4316 case RS_MOPEN:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004317 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004318 if (status == RA_NOMATCH)
4319 restore_se(&rp->rs_un.sesave, &rex.reg_startpos[rp->rs_no],
4320 &rex.reg_startp[rp->rs_no]);
4321 regstack_pop(&scan);
4322 break;
4323
4324#ifdef FEAT_SYN_HL
4325 case RS_ZOPEN:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004326 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004327 if (status == RA_NOMATCH)
4328 restore_se(&rp->rs_un.sesave, &reg_startzpos[rp->rs_no],
4329 &reg_startzp[rp->rs_no]);
4330 regstack_pop(&scan);
4331 break;
4332#endif
4333
4334 case RS_MCLOSE:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004335 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004336 if (status == RA_NOMATCH)
4337 restore_se(&rp->rs_un.sesave, &rex.reg_endpos[rp->rs_no],
4338 &rex.reg_endp[rp->rs_no]);
4339 regstack_pop(&scan);
4340 break;
4341
4342#ifdef FEAT_SYN_HL
4343 case RS_ZCLOSE:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004344 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004345 if (status == RA_NOMATCH)
4346 restore_se(&rp->rs_un.sesave, &reg_endzpos[rp->rs_no],
4347 &reg_endzp[rp->rs_no]);
4348 regstack_pop(&scan);
4349 break;
4350#endif
4351
4352 case RS_BRANCH:
4353 if (status == RA_MATCH)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004354 // this branch matched, use it
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004355 regstack_pop(&scan);
4356 else
4357 {
4358 if (status != RA_BREAK)
4359 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004360 // After a non-matching branch: try next one.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004361 reg_restore(&rp->rs_un.regsave, &backpos);
4362 scan = rp->rs_scan;
4363 }
4364 if (scan == NULL || OP(scan) != BRANCH)
4365 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004366 // no more branches, didn't find a match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004367 status = RA_NOMATCH;
4368 regstack_pop(&scan);
4369 }
4370 else
4371 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004372 // Prepare to try a branch.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004373 rp->rs_scan = regnext(scan);
4374 reg_save(&rp->rs_un.regsave, &backpos);
4375 scan = OPERAND(scan);
4376 }
4377 }
4378 break;
4379
4380 case RS_BRCPLX_MORE:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004381 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004382 if (status == RA_NOMATCH)
4383 {
4384 reg_restore(&rp->rs_un.regsave, &backpos);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004385 --brace_count[rp->rs_no]; // decrement match count
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004386 }
4387 regstack_pop(&scan);
4388 break;
4389
4390 case RS_BRCPLX_LONG:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004391 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004392 if (status == RA_NOMATCH)
4393 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004394 // There was no match, but we did find enough matches.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004395 reg_restore(&rp->rs_un.regsave, &backpos);
4396 --brace_count[rp->rs_no];
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004397 // continue with the items after "\{}"
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004398 status = RA_CONT;
4399 }
4400 regstack_pop(&scan);
4401 if (status == RA_CONT)
4402 scan = regnext(scan);
4403 break;
4404
4405 case RS_BRCPLX_SHORT:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004406 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004407 if (status == RA_NOMATCH)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004408 // There was no match, try to match one more item.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004409 reg_restore(&rp->rs_un.regsave, &backpos);
4410 regstack_pop(&scan);
4411 if (status == RA_NOMATCH)
4412 {
4413 scan = OPERAND(scan);
4414 status = RA_CONT;
4415 }
4416 break;
4417
4418 case RS_NOMATCH:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004419 // Pop the state. If the operand matches for NOMATCH or
4420 // doesn't match for MATCH/SUBPAT, we fail. Otherwise backup,
4421 // except for SUBPAT, and continue with the next item.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004422 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH))
4423 status = RA_NOMATCH;
4424 else
4425 {
4426 status = RA_CONT;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004427 if (rp->rs_no != SUBPAT) // zero-width
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004428 reg_restore(&rp->rs_un.regsave, &backpos);
4429 }
4430 regstack_pop(&scan);
4431 if (status == RA_CONT)
4432 scan = regnext(scan);
4433 break;
4434
4435 case RS_BEHIND1:
4436 if (status == RA_NOMATCH)
4437 {
4438 regstack_pop(&scan);
4439 regstack.ga_len -= sizeof(regbehind_T);
4440 }
4441 else
4442 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004443 // The stuff after BEHIND/NOBEHIND matches. Now try if
4444 // the behind part does (not) match before the current
4445 // position in the input. This must be done at every
4446 // position in the input and checking if the match ends at
4447 // the current position.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004448
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004449 // save the position after the found match for next
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004450 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos);
4451
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004452 // Start looking for a match with operand at the current
4453 // position. Go back one character until we find the
4454 // result, hitting the start of the line or the previous
4455 // line (for multi-line matching).
4456 // Set behind_pos to where the match should end, BHPOS
4457 // will match it. Save the current value.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004458 (((regbehind_T *)rp) - 1)->save_behind = behind_pos;
4459 behind_pos = rp->rs_un.regsave;
4460
4461 rp->rs_state = RS_BEHIND2;
4462
4463 reg_restore(&rp->rs_un.regsave, &backpos);
4464 scan = OPERAND(rp->rs_scan) + 4;
4465 }
4466 break;
4467
4468 case RS_BEHIND2:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004469 // Looping for BEHIND / NOBEHIND match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004470 if (status == RA_MATCH && reg_save_equal(&behind_pos))
4471 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004472 // found a match that ends where "next" started
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004473 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
4474 if (rp->rs_no == BEHIND)
4475 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
4476 &backpos);
4477 else
4478 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004479 // But we didn't want a match. Need to restore the
4480 // subexpr, because what follows matched, so they have
4481 // been set.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004482 status = RA_NOMATCH;
4483 restore_subexpr(((regbehind_T *)rp) - 1);
4484 }
4485 regstack_pop(&scan);
4486 regstack.ga_len -= sizeof(regbehind_T);
4487 }
4488 else
4489 {
4490 long limit;
4491
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004492 // No match or a match that doesn't end where we want it: Go
4493 // back one character. May go to previous line once.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004494 no = OK;
4495 limit = OPERAND_MIN(rp->rs_scan);
4496 if (REG_MULTI)
4497 {
4498 if (limit > 0
4499 && ((rp->rs_un.regsave.rs_u.pos.lnum
4500 < behind_pos.rs_u.pos.lnum
4501 ? (colnr_T)STRLEN(rex.line)
4502 : behind_pos.rs_u.pos.col)
4503 - rp->rs_un.regsave.rs_u.pos.col >= limit))
4504 no = FAIL;
4505 else if (rp->rs_un.regsave.rs_u.pos.col == 0)
4506 {
4507 if (rp->rs_un.regsave.rs_u.pos.lnum
4508 < behind_pos.rs_u.pos.lnum
4509 || reg_getline(
4510 --rp->rs_un.regsave.rs_u.pos.lnum)
4511 == NULL)
4512 no = FAIL;
4513 else
4514 {
4515 reg_restore(&rp->rs_un.regsave, &backpos);
4516 rp->rs_un.regsave.rs_u.pos.col =
4517 (colnr_T)STRLEN(rex.line);
4518 }
4519 }
4520 else
4521 {
4522 if (has_mbyte)
4523 {
4524 char_u *line =
4525 reg_getline(rp->rs_un.regsave.rs_u.pos.lnum);
4526
4527 rp->rs_un.regsave.rs_u.pos.col -=
4528 (*mb_head_off)(line, line
4529 + rp->rs_un.regsave.rs_u.pos.col - 1) + 1;
4530 }
4531 else
4532 --rp->rs_un.regsave.rs_u.pos.col;
4533 }
4534 }
4535 else
4536 {
4537 if (rp->rs_un.regsave.rs_u.ptr == rex.line)
4538 no = FAIL;
4539 else
4540 {
4541 MB_PTR_BACK(rex.line, rp->rs_un.regsave.rs_u.ptr);
4542 if (limit > 0 && (long)(behind_pos.rs_u.ptr
4543 - rp->rs_un.regsave.rs_u.ptr) > limit)
4544 no = FAIL;
4545 }
4546 }
4547 if (no == OK)
4548 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004549 // Advanced, prepare for finding match again.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004550 reg_restore(&rp->rs_un.regsave, &backpos);
4551 scan = OPERAND(rp->rs_scan) + 4;
4552 if (status == RA_MATCH)
4553 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004554 // We did match, so subexpr may have been changed,
4555 // need to restore them for the next try.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004556 status = RA_NOMATCH;
4557 restore_subexpr(((regbehind_T *)rp) - 1);
4558 }
4559 }
4560 else
4561 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004562 // Can't advance. For NOBEHIND that's a match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004563 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
4564 if (rp->rs_no == NOBEHIND)
4565 {
4566 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
4567 &backpos);
4568 status = RA_MATCH;
4569 }
4570 else
4571 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004572 // We do want a proper match. Need to restore the
4573 // subexpr if we had a match, because they may have
4574 // been set.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004575 if (status == RA_MATCH)
4576 {
4577 status = RA_NOMATCH;
4578 restore_subexpr(((regbehind_T *)rp) - 1);
4579 }
4580 }
4581 regstack_pop(&scan);
4582 regstack.ga_len -= sizeof(regbehind_T);
4583 }
4584 }
4585 break;
4586
4587 case RS_STAR_LONG:
4588 case RS_STAR_SHORT:
4589 {
4590 regstar_T *rst = ((regstar_T *)rp) - 1;
4591
4592 if (status == RA_MATCH)
4593 {
4594 regstack_pop(&scan);
4595 regstack.ga_len -= sizeof(regstar_T);
4596 break;
4597 }
4598
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004599 // Tried once already, restore input pointers.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004600 if (status != RA_BREAK)
4601 reg_restore(&rp->rs_un.regsave, &backpos);
4602
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004603 // Repeat until we found a position where it could match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004604 for (;;)
4605 {
4606 if (status != RA_BREAK)
4607 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004608 // Tried first position already, advance.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004609 if (rp->rs_state == RS_STAR_LONG)
4610 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004611 // Trying for longest match, but couldn't or
4612 // didn't match -- back up one char.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004613 if (--rst->count < rst->minval)
4614 break;
4615 if (rex.input == rex.line)
4616 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004617 // backup to last char of previous line
Bram Moolenaar6456fae2022-02-22 13:37:31 +00004618 if (rex.lnum == 0)
4619 {
4620 status = RA_NOMATCH;
4621 break;
4622 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004623 --rex.lnum;
4624 rex.line = reg_getline(rex.lnum);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004625 // Just in case regrepeat() didn't count
4626 // right.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004627 if (rex.line == NULL)
4628 break;
4629 rex.input = rex.line + STRLEN(rex.line);
4630 fast_breakcheck();
4631 }
4632 else
4633 MB_PTR_BACK(rex.line, rex.input);
4634 }
4635 else
4636 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004637 // Range is backwards, use shortest match first.
4638 // Careful: maxval and minval are exchanged!
4639 // Couldn't or didn't match: try advancing one
4640 // char.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004641 if (rst->count == rst->minval
4642 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0)
4643 break;
4644 ++rst->count;
4645 }
4646 if (got_int)
4647 break;
4648 }
4649 else
4650 status = RA_NOMATCH;
4651
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004652 // If it could match, try it.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004653 if (rst->nextb == NUL || *rex.input == rst->nextb
4654 || *rex.input == rst->nextb_ic)
4655 {
4656 reg_save(&rp->rs_un.regsave, &backpos);
4657 scan = regnext(rp->rs_scan);
4658 status = RA_CONT;
4659 break;
4660 }
4661 }
4662 if (status != RA_CONT)
4663 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004664 // Failed.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004665 regstack_pop(&scan);
4666 regstack.ga_len -= sizeof(regstar_T);
4667 status = RA_NOMATCH;
4668 }
4669 }
4670 break;
4671 }
4672
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004673 // If we want to continue the inner loop or didn't pop a state
4674 // continue matching loop
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004675 if (status == RA_CONT || rp == (regitem_T *)
4676 ((char *)regstack.ga_data + regstack.ga_len) - 1)
4677 break;
4678 }
4679
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004680 // May need to continue with the inner loop, starting at "scan".
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004681 if (status == RA_CONT)
4682 continue;
4683
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004684 // If the regstack is empty or something failed we are done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004685 if (regstack.ga_len == 0 || status == RA_FAIL)
4686 {
4687 if (scan == NULL)
4688 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004689 // We get here only if there's trouble -- normally "case END" is
4690 // the terminating point.
Bram Moolenaare29a27f2021-07-20 21:07:36 +02004691 iemsg(_(e_corrupted_regexp_program));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004692#ifdef DEBUG
4693 printf("Premature EOL\n");
4694#endif
4695 }
4696 return (status == RA_MATCH);
4697 }
4698
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004699 } // End of loop until the regstack is empty.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004700
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004701 // NOTREACHED
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004702}
4703
4704/*
4705 * regtry - try match of "prog" with at rex.line["col"].
4706 * Returns 0 for failure, number of lines contained in the match otherwise.
4707 */
4708 static long
4709regtry(
4710 bt_regprog_T *prog,
4711 colnr_T col,
4712 proftime_T *tm, // timeout limit or NULL
4713 int *timed_out) // flag set on timeout or NULL
4714{
4715 rex.input = rex.line + col;
4716 rex.need_clear_subexpr = TRUE;
4717#ifdef FEAT_SYN_HL
4718 // Clear the external match subpointers if necessary.
4719 rex.need_clear_zsubexpr = (prog->reghasz == REX_SET);
4720#endif
4721
4722 if (regmatch(prog->program + 1, tm, timed_out) == 0)
4723 return 0;
4724
4725 cleanup_subexpr();
4726 if (REG_MULTI)
4727 {
4728 if (rex.reg_startpos[0].lnum < 0)
4729 {
4730 rex.reg_startpos[0].lnum = 0;
4731 rex.reg_startpos[0].col = col;
4732 }
4733 if (rex.reg_endpos[0].lnum < 0)
4734 {
4735 rex.reg_endpos[0].lnum = rex.lnum;
4736 rex.reg_endpos[0].col = (int)(rex.input - rex.line);
4737 }
4738 else
4739 // Use line number of "\ze".
4740 rex.lnum = rex.reg_endpos[0].lnum;
4741 }
4742 else
4743 {
4744 if (rex.reg_startp[0] == NULL)
4745 rex.reg_startp[0] = rex.line + col;
4746 if (rex.reg_endp[0] == NULL)
4747 rex.reg_endp[0] = rex.input;
4748 }
4749#ifdef FEAT_SYN_HL
4750 // Package any found \z(...\) matches for export. Default is none.
4751 unref_extmatch(re_extmatch_out);
4752 re_extmatch_out = NULL;
4753
4754 if (prog->reghasz == REX_SET)
4755 {
4756 int i;
4757
4758 cleanup_zsubexpr();
4759 re_extmatch_out = make_extmatch();
Bram Moolenaar7c77b342019-12-22 19:40:40 +01004760 if (re_extmatch_out == NULL)
4761 return 0;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004762 for (i = 0; i < NSUBEXP; i++)
4763 {
4764 if (REG_MULTI)
4765 {
4766 // Only accept single line matches.
4767 if (reg_startzpos[i].lnum >= 0
4768 && reg_endzpos[i].lnum == reg_startzpos[i].lnum
4769 && reg_endzpos[i].col >= reg_startzpos[i].col)
4770 re_extmatch_out->matches[i] =
4771 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
4772 + reg_startzpos[i].col,
4773 reg_endzpos[i].col - reg_startzpos[i].col);
4774 }
4775 else
4776 {
4777 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
4778 re_extmatch_out->matches[i] =
4779 vim_strnsave(reg_startzp[i],
Bram Moolenaar71ccd032020-06-12 22:59:11 +02004780 reg_endzp[i] - reg_startzp[i]);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004781 }
4782 }
4783 }
4784#endif
4785 return 1 + rex.lnum;
4786}
4787
4788/*
4789 * Match a regexp against a string ("line" points to the string) or multiple
Bram Moolenaardf365142021-05-03 20:01:45 +02004790 * lines (if "line" is NULL, use reg_getline()).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004791 * Returns 0 for failure, number of lines contained in the match otherwise.
4792 */
4793 static long
4794bt_regexec_both(
4795 char_u *line,
4796 colnr_T col, // column to start looking for match
4797 proftime_T *tm, // timeout limit or NULL
4798 int *timed_out) // flag set on timeout or NULL
4799{
4800 bt_regprog_T *prog;
4801 char_u *s;
4802 long retval = 0L;
4803
4804 // Create "regstack" and "backpos" if they are not allocated yet.
4805 // We allocate *_INITIAL amount of bytes first and then set the grow size
4806 // to much bigger value to avoid many malloc calls in case of deep regular
4807 // expressions.
4808 if (regstack.ga_data == NULL)
4809 {
4810 // Use an item size of 1 byte, since we push different things
4811 // onto the regstack.
4812 ga_init2(&regstack, 1, REGSTACK_INITIAL);
4813 (void)ga_grow(&regstack, REGSTACK_INITIAL);
4814 regstack.ga_growsize = REGSTACK_INITIAL * 8;
4815 }
4816
4817 if (backpos.ga_data == NULL)
4818 {
4819 ga_init2(&backpos, sizeof(backpos_T), BACKPOS_INITIAL);
4820 (void)ga_grow(&backpos, BACKPOS_INITIAL);
4821 backpos.ga_growsize = BACKPOS_INITIAL * 8;
4822 }
4823
4824 if (REG_MULTI)
4825 {
4826 prog = (bt_regprog_T *)rex.reg_mmatch->regprog;
4827 line = reg_getline((linenr_T)0);
4828 rex.reg_startpos = rex.reg_mmatch->startpos;
4829 rex.reg_endpos = rex.reg_mmatch->endpos;
4830 }
4831 else
4832 {
4833 prog = (bt_regprog_T *)rex.reg_match->regprog;
4834 rex.reg_startp = rex.reg_match->startp;
4835 rex.reg_endp = rex.reg_match->endp;
4836 }
4837
4838 // Be paranoid...
4839 if (prog == NULL || line == NULL)
4840 {
Bram Moolenaare29a27f2021-07-20 21:07:36 +02004841 iemsg(_(e_null_argument));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004842 goto theend;
4843 }
4844
4845 // Check validity of program.
4846 if (prog_magic_wrong())
4847 goto theend;
4848
4849 // If the start column is past the maximum column: no need to try.
4850 if (rex.reg_maxcol > 0 && col >= rex.reg_maxcol)
4851 goto theend;
4852
4853 // If pattern contains "\c" or "\C": overrule value of rex.reg_ic
4854 if (prog->regflags & RF_ICASE)
4855 rex.reg_ic = TRUE;
4856 else if (prog->regflags & RF_NOICASE)
4857 rex.reg_ic = FALSE;
4858
4859 // If pattern contains "\Z" overrule value of rex.reg_icombine
4860 if (prog->regflags & RF_ICOMBINE)
4861 rex.reg_icombine = TRUE;
4862
4863 // If there is a "must appear" string, look for it.
4864 if (prog->regmust != NULL)
4865 {
4866 int c;
4867
4868 if (has_mbyte)
4869 c = (*mb_ptr2char)(prog->regmust);
4870 else
4871 c = *prog->regmust;
4872 s = line + col;
4873
4874 // This is used very often, esp. for ":global". Use three versions of
4875 // the loop to avoid overhead of conditions.
4876 if (!rex.reg_ic && !has_mbyte)
4877 while ((s = vim_strbyte(s, c)) != NULL)
4878 {
4879 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
4880 break; // Found it.
4881 ++s;
4882 }
4883 else if (!rex.reg_ic || (!enc_utf8 && mb_char2len(c) > 1))
4884 while ((s = vim_strchr(s, c)) != NULL)
4885 {
4886 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
4887 break; // Found it.
4888 MB_PTR_ADV(s);
4889 }
4890 else
4891 while ((s = cstrchr(s, c)) != NULL)
4892 {
4893 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
4894 break; // Found it.
4895 MB_PTR_ADV(s);
4896 }
4897 if (s == NULL) // Not present.
4898 goto theend;
4899 }
4900
4901 rex.line = line;
4902 rex.lnum = 0;
4903 reg_toolong = FALSE;
4904
4905 // Simplest case: Anchored match need be tried only once.
4906 if (prog->reganch)
4907 {
4908 int c;
4909
4910 if (has_mbyte)
4911 c = (*mb_ptr2char)(rex.line + col);
4912 else
4913 c = rex.line[col];
4914 if (prog->regstart == NUL
4915 || prog->regstart == c
4916 || (rex.reg_ic
4917 && (((enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
4918 || (c < 255 && prog->regstart < 255 &&
4919 MB_TOLOWER(prog->regstart) == MB_TOLOWER(c)))))
4920 retval = regtry(prog, col, tm, timed_out);
4921 else
4922 retval = 0;
4923 }
4924 else
4925 {
4926#ifdef FEAT_RELTIME
4927 int tm_count = 0;
4928#endif
4929 // Messy cases: unanchored match.
4930 while (!got_int)
4931 {
4932 if (prog->regstart != NUL)
4933 {
4934 // Skip until the char we know it must start with.
4935 // Used often, do some work to avoid call overhead.
4936 if (!rex.reg_ic && !has_mbyte)
4937 s = vim_strbyte(rex.line + col, prog->regstart);
4938 else
4939 s = cstrchr(rex.line + col, prog->regstart);
4940 if (s == NULL)
4941 {
4942 retval = 0;
4943 break;
4944 }
4945 col = (int)(s - rex.line);
4946 }
4947
4948 // Check for maximum column to try.
4949 if (rex.reg_maxcol > 0 && col >= rex.reg_maxcol)
4950 {
4951 retval = 0;
4952 break;
4953 }
4954
4955 retval = regtry(prog, col, tm, timed_out);
4956 if (retval > 0)
4957 break;
4958
4959 // if not currently on the first line, get it again
4960 if (rex.lnum != 0)
4961 {
4962 rex.lnum = 0;
4963 rex.line = reg_getline((linenr_T)0);
4964 }
4965 if (rex.line[col] == NUL)
4966 break;
4967 if (has_mbyte)
4968 col += (*mb_ptr2len)(rex.line + col);
4969 else
4970 ++col;
4971#ifdef FEAT_RELTIME
4972 // Check for timeout once in a twenty times to avoid overhead.
4973 if (tm != NULL && ++tm_count == 20)
4974 {
4975 tm_count = 0;
4976 if (profile_passed_limit(tm))
4977 {
4978 if (timed_out != NULL)
4979 *timed_out = TRUE;
4980 break;
4981 }
4982 }
4983#endif
4984 }
4985 }
4986
4987theend:
4988 // Free "reg_tofree" when it's a bit big.
4989 // Free regstack and backpos if they are bigger than their initial size.
4990 if (reg_tofreelen > 400)
4991 VIM_CLEAR(reg_tofree);
4992 if (regstack.ga_maxlen > REGSTACK_INITIAL)
4993 ga_clear(&regstack);
4994 if (backpos.ga_maxlen > BACKPOS_INITIAL)
4995 ga_clear(&backpos);
4996
Bram Moolenaara3d10a52020-12-21 18:24:00 +01004997 if (retval > 0)
Bram Moolenaara7a691c2020-12-09 16:36:04 +01004998 {
Bram Moolenaara3d10a52020-12-21 18:24:00 +01004999 // Make sure the end is never before the start. Can happen when \zs
5000 // and \ze are used.
5001 if (REG_MULTI)
5002 {
5003 lpos_T *start = &rex.reg_mmatch->startpos[0];
5004 lpos_T *end = &rex.reg_mmatch->endpos[0];
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005005
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005006 if (end->lnum < start->lnum
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005007 || (end->lnum == start->lnum && end->col < start->col))
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005008 rex.reg_mmatch->endpos[0] = rex.reg_mmatch->startpos[0];
5009 }
5010 else
5011 {
5012 if (rex.reg_match->endp[0] < rex.reg_match->startp[0])
5013 rex.reg_match->endp[0] = rex.reg_match->startp[0];
5014 }
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005015 }
5016
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005017 return retval;
5018}
5019
5020/*
5021 * Match a regexp against a string.
5022 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
5023 * Uses curbuf for line count and 'iskeyword'.
5024 * if "line_lbr" is TRUE consider a "\n" in "line" to be a line break.
5025 *
5026 * Returns 0 for failure, number of lines contained in the match otherwise.
5027 */
5028 static int
5029bt_regexec_nl(
5030 regmatch_T *rmp,
5031 char_u *line, // string to match against
5032 colnr_T col, // column to start looking for match
5033 int line_lbr)
5034{
5035 rex.reg_match = rmp;
5036 rex.reg_mmatch = NULL;
5037 rex.reg_maxline = 0;
5038 rex.reg_line_lbr = line_lbr;
5039 rex.reg_buf = curbuf;
5040 rex.reg_win = NULL;
5041 rex.reg_ic = rmp->rm_ic;
5042 rex.reg_icombine = FALSE;
5043 rex.reg_maxcol = 0;
5044
5045 return bt_regexec_both(line, col, NULL, NULL);
5046}
5047
5048/*
5049 * Match a regexp against multiple lines.
5050 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
5051 * Uses curbuf for line count and 'iskeyword'.
5052 *
5053 * Return zero if there is no match. Return number of lines contained in the
5054 * match otherwise.
5055 */
5056 static long
5057bt_regexec_multi(
5058 regmmatch_T *rmp,
5059 win_T *win, // window in which to search or NULL
5060 buf_T *buf, // buffer in which to search
5061 linenr_T lnum, // nr of line to start looking for match
5062 colnr_T col, // column to start looking for match
5063 proftime_T *tm, // timeout limit or NULL
5064 int *timed_out) // flag set on timeout or NULL
5065{
Bram Moolenaarf4140482020-02-15 23:06:45 +01005066 init_regexec_multi(rmp, win, buf, lnum);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005067 return bt_regexec_both(NULL, col, tm, timed_out);
5068}
5069
5070/*
5071 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
5072 */
5073 static int
5074re_num_cmp(long_u val, char_u *scan)
5075{
5076 long_u n = OPERAND_MIN(scan);
5077
5078 if (OPERAND_CMP(scan) == '>')
5079 return val > n;
5080 if (OPERAND_CMP(scan) == '<')
5081 return val < n;
5082 return val == n;
5083}
5084
5085#ifdef BT_REGEXP_DUMP
5086
5087/*
5088 * regdump - dump a regexp onto stdout in vaguely comprehensible form
5089 */
5090 static void
5091regdump(char_u *pattern, bt_regprog_T *r)
5092{
5093 char_u *s;
5094 int op = EXACTLY; // Arbitrary non-END op.
5095 char_u *next;
5096 char_u *end = NULL;
5097 FILE *f;
5098
5099#ifdef BT_REGEXP_LOG
5100 f = fopen("bt_regexp_log.log", "a");
5101#else
5102 f = stdout;
5103#endif
5104 if (f == NULL)
5105 return;
5106 fprintf(f, "-------------------------------------\n\r\nregcomp(%s):\r\n", pattern);
5107
5108 s = r->program + 1;
5109 // Loop until we find the END that isn't before a referred next (an END
5110 // can also appear in a NOMATCH operand).
5111 while (op != END || s <= end)
5112 {
5113 op = OP(s);
5114 fprintf(f, "%2d%s", (int)(s - r->program), regprop(s)); // Where, what.
5115 next = regnext(s);
5116 if (next == NULL) // Next ptr.
5117 fprintf(f, "(0)");
5118 else
5119 fprintf(f, "(%d)", (int)((s - r->program) + (next - s)));
5120 if (end < next)
5121 end = next;
5122 if (op == BRACE_LIMITS)
5123 {
5124 // Two ints
5125 fprintf(f, " minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
5126 s += 8;
5127 }
5128 else if (op == BEHIND || op == NOBEHIND)
5129 {
5130 // one int
5131 fprintf(f, " count %ld", OPERAND_MIN(s));
5132 s += 4;
5133 }
5134 else if (op == RE_LNUM || op == RE_COL || op == RE_VCOL)
5135 {
5136 // one int plus comparator
5137 fprintf(f, " count %ld", OPERAND_MIN(s));
5138 s += 5;
5139 }
5140 s += 3;
5141 if (op == ANYOF || op == ANYOF + ADD_NL
5142 || op == ANYBUT || op == ANYBUT + ADD_NL
5143 || op == EXACTLY)
5144 {
5145 // Literal string, where present.
5146 fprintf(f, "\nxxxxxxxxx\n");
5147 while (*s != NUL)
5148 fprintf(f, "%c", *s++);
5149 fprintf(f, "\nxxxxxxxxx\n");
5150 s++;
5151 }
5152 fprintf(f, "\r\n");
5153 }
5154
5155 // Header fields of interest.
5156 if (r->regstart != NUL)
5157 fprintf(f, "start `%s' 0x%x; ", r->regstart < 256
5158 ? (char *)transchar(r->regstart)
5159 : "multibyte", r->regstart);
5160 if (r->reganch)
5161 fprintf(f, "anchored; ");
5162 if (r->regmust != NULL)
5163 fprintf(f, "must have \"%s\"", r->regmust);
5164 fprintf(f, "\r\n");
5165
5166#ifdef BT_REGEXP_LOG
5167 fclose(f);
5168#endif
5169}
5170#endif // BT_REGEXP_DUMP
5171
5172#ifdef DEBUG
5173/*
5174 * regprop - printable representation of opcode
5175 */
5176 static char_u *
5177regprop(char_u *op)
5178{
5179 char *p;
5180 static char buf[50];
5181
5182 STRCPY(buf, ":");
5183
5184 switch ((int) OP(op))
5185 {
5186 case BOL:
5187 p = "BOL";
5188 break;
5189 case EOL:
5190 p = "EOL";
5191 break;
5192 case RE_BOF:
5193 p = "BOF";
5194 break;
5195 case RE_EOF:
5196 p = "EOF";
5197 break;
5198 case CURSOR:
5199 p = "CURSOR";
5200 break;
5201 case RE_VISUAL:
5202 p = "RE_VISUAL";
5203 break;
5204 case RE_LNUM:
5205 p = "RE_LNUM";
5206 break;
5207 case RE_MARK:
5208 p = "RE_MARK";
5209 break;
5210 case RE_COL:
5211 p = "RE_COL";
5212 break;
5213 case RE_VCOL:
5214 p = "RE_VCOL";
5215 break;
5216 case BOW:
5217 p = "BOW";
5218 break;
5219 case EOW:
5220 p = "EOW";
5221 break;
5222 case ANY:
5223 p = "ANY";
5224 break;
5225 case ANY + ADD_NL:
5226 p = "ANY+NL";
5227 break;
5228 case ANYOF:
5229 p = "ANYOF";
5230 break;
5231 case ANYOF + ADD_NL:
5232 p = "ANYOF+NL";
5233 break;
5234 case ANYBUT:
5235 p = "ANYBUT";
5236 break;
5237 case ANYBUT + ADD_NL:
5238 p = "ANYBUT+NL";
5239 break;
5240 case IDENT:
5241 p = "IDENT";
5242 break;
5243 case IDENT + ADD_NL:
5244 p = "IDENT+NL";
5245 break;
5246 case SIDENT:
5247 p = "SIDENT";
5248 break;
5249 case SIDENT + ADD_NL:
5250 p = "SIDENT+NL";
5251 break;
5252 case KWORD:
5253 p = "KWORD";
5254 break;
5255 case KWORD + ADD_NL:
5256 p = "KWORD+NL";
5257 break;
5258 case SKWORD:
5259 p = "SKWORD";
5260 break;
5261 case SKWORD + ADD_NL:
5262 p = "SKWORD+NL";
5263 break;
5264 case FNAME:
5265 p = "FNAME";
5266 break;
5267 case FNAME + ADD_NL:
5268 p = "FNAME+NL";
5269 break;
5270 case SFNAME:
5271 p = "SFNAME";
5272 break;
5273 case SFNAME + ADD_NL:
5274 p = "SFNAME+NL";
5275 break;
5276 case PRINT:
5277 p = "PRINT";
5278 break;
5279 case PRINT + ADD_NL:
5280 p = "PRINT+NL";
5281 break;
5282 case SPRINT:
5283 p = "SPRINT";
5284 break;
5285 case SPRINT + ADD_NL:
5286 p = "SPRINT+NL";
5287 break;
5288 case WHITE:
5289 p = "WHITE";
5290 break;
5291 case WHITE + ADD_NL:
5292 p = "WHITE+NL";
5293 break;
5294 case NWHITE:
5295 p = "NWHITE";
5296 break;
5297 case NWHITE + ADD_NL:
5298 p = "NWHITE+NL";
5299 break;
5300 case DIGIT:
5301 p = "DIGIT";
5302 break;
5303 case DIGIT + ADD_NL:
5304 p = "DIGIT+NL";
5305 break;
5306 case NDIGIT:
5307 p = "NDIGIT";
5308 break;
5309 case NDIGIT + ADD_NL:
5310 p = "NDIGIT+NL";
5311 break;
5312 case HEX:
5313 p = "HEX";
5314 break;
5315 case HEX + ADD_NL:
5316 p = "HEX+NL";
5317 break;
5318 case NHEX:
5319 p = "NHEX";
5320 break;
5321 case NHEX + ADD_NL:
5322 p = "NHEX+NL";
5323 break;
5324 case OCTAL:
5325 p = "OCTAL";
5326 break;
5327 case OCTAL + ADD_NL:
5328 p = "OCTAL+NL";
5329 break;
5330 case NOCTAL:
5331 p = "NOCTAL";
5332 break;
5333 case NOCTAL + ADD_NL:
5334 p = "NOCTAL+NL";
5335 break;
5336 case WORD:
5337 p = "WORD";
5338 break;
5339 case WORD + ADD_NL:
5340 p = "WORD+NL";
5341 break;
5342 case NWORD:
5343 p = "NWORD";
5344 break;
5345 case NWORD + ADD_NL:
5346 p = "NWORD+NL";
5347 break;
5348 case HEAD:
5349 p = "HEAD";
5350 break;
5351 case HEAD + ADD_NL:
5352 p = "HEAD+NL";
5353 break;
5354 case NHEAD:
5355 p = "NHEAD";
5356 break;
5357 case NHEAD + ADD_NL:
5358 p = "NHEAD+NL";
5359 break;
5360 case ALPHA:
5361 p = "ALPHA";
5362 break;
5363 case ALPHA + ADD_NL:
5364 p = "ALPHA+NL";
5365 break;
5366 case NALPHA:
5367 p = "NALPHA";
5368 break;
5369 case NALPHA + ADD_NL:
5370 p = "NALPHA+NL";
5371 break;
5372 case LOWER:
5373 p = "LOWER";
5374 break;
5375 case LOWER + ADD_NL:
5376 p = "LOWER+NL";
5377 break;
5378 case NLOWER:
5379 p = "NLOWER";
5380 break;
5381 case NLOWER + ADD_NL:
5382 p = "NLOWER+NL";
5383 break;
5384 case UPPER:
5385 p = "UPPER";
5386 break;
5387 case UPPER + ADD_NL:
5388 p = "UPPER+NL";
5389 break;
5390 case NUPPER:
5391 p = "NUPPER";
5392 break;
5393 case NUPPER + ADD_NL:
5394 p = "NUPPER+NL";
5395 break;
5396 case BRANCH:
5397 p = "BRANCH";
5398 break;
5399 case EXACTLY:
5400 p = "EXACTLY";
5401 break;
5402 case NOTHING:
5403 p = "NOTHING";
5404 break;
5405 case BACK:
5406 p = "BACK";
5407 break;
5408 case END:
5409 p = "END";
5410 break;
5411 case MOPEN + 0:
5412 p = "MATCH START";
5413 break;
5414 case MOPEN + 1:
5415 case MOPEN + 2:
5416 case MOPEN + 3:
5417 case MOPEN + 4:
5418 case MOPEN + 5:
5419 case MOPEN + 6:
5420 case MOPEN + 7:
5421 case MOPEN + 8:
5422 case MOPEN + 9:
5423 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
5424 p = NULL;
5425 break;
5426 case MCLOSE + 0:
5427 p = "MATCH END";
5428 break;
5429 case MCLOSE + 1:
5430 case MCLOSE + 2:
5431 case MCLOSE + 3:
5432 case MCLOSE + 4:
5433 case MCLOSE + 5:
5434 case MCLOSE + 6:
5435 case MCLOSE + 7:
5436 case MCLOSE + 8:
5437 case MCLOSE + 9:
5438 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
5439 p = NULL;
5440 break;
5441 case BACKREF + 1:
5442 case BACKREF + 2:
5443 case BACKREF + 3:
5444 case BACKREF + 4:
5445 case BACKREF + 5:
5446 case BACKREF + 6:
5447 case BACKREF + 7:
5448 case BACKREF + 8:
5449 case BACKREF + 9:
5450 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
5451 p = NULL;
5452 break;
5453 case NOPEN:
5454 p = "NOPEN";
5455 break;
5456 case NCLOSE:
5457 p = "NCLOSE";
5458 break;
5459#ifdef FEAT_SYN_HL
5460 case ZOPEN + 1:
5461 case ZOPEN + 2:
5462 case ZOPEN + 3:
5463 case ZOPEN + 4:
5464 case ZOPEN + 5:
5465 case ZOPEN + 6:
5466 case ZOPEN + 7:
5467 case ZOPEN + 8:
5468 case ZOPEN + 9:
5469 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
5470 p = NULL;
5471 break;
5472 case ZCLOSE + 1:
5473 case ZCLOSE + 2:
5474 case ZCLOSE + 3:
5475 case ZCLOSE + 4:
5476 case ZCLOSE + 5:
5477 case ZCLOSE + 6:
5478 case ZCLOSE + 7:
5479 case ZCLOSE + 8:
5480 case ZCLOSE + 9:
5481 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
5482 p = NULL;
5483 break;
5484 case ZREF + 1:
5485 case ZREF + 2:
5486 case ZREF + 3:
5487 case ZREF + 4:
5488 case ZREF + 5:
5489 case ZREF + 6:
5490 case ZREF + 7:
5491 case ZREF + 8:
5492 case ZREF + 9:
5493 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
5494 p = NULL;
5495 break;
5496#endif
5497 case STAR:
5498 p = "STAR";
5499 break;
5500 case PLUS:
5501 p = "PLUS";
5502 break;
5503 case NOMATCH:
5504 p = "NOMATCH";
5505 break;
5506 case MATCH:
5507 p = "MATCH";
5508 break;
5509 case BEHIND:
5510 p = "BEHIND";
5511 break;
5512 case NOBEHIND:
5513 p = "NOBEHIND";
5514 break;
5515 case SUBPAT:
5516 p = "SUBPAT";
5517 break;
5518 case BRACE_LIMITS:
5519 p = "BRACE_LIMITS";
5520 break;
5521 case BRACE_SIMPLE:
5522 p = "BRACE_SIMPLE";
5523 break;
5524 case BRACE_COMPLEX + 0:
5525 case BRACE_COMPLEX + 1:
5526 case BRACE_COMPLEX + 2:
5527 case BRACE_COMPLEX + 3:
5528 case BRACE_COMPLEX + 4:
5529 case BRACE_COMPLEX + 5:
5530 case BRACE_COMPLEX + 6:
5531 case BRACE_COMPLEX + 7:
5532 case BRACE_COMPLEX + 8:
5533 case BRACE_COMPLEX + 9:
5534 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
5535 p = NULL;
5536 break;
5537 case MULTIBYTECODE:
5538 p = "MULTIBYTECODE";
5539 break;
5540 case NEWL:
5541 p = "NEWL";
5542 break;
5543 default:
5544 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
5545 p = NULL;
5546 break;
5547 }
5548 if (p != NULL)
5549 STRCAT(buf, p);
5550 return (char_u *)buf;
5551}
5552#endif // DEBUG