blob: 5d9450d8712b3f284ca6c23e341ce142e07a58a8 [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.
RestorerZ68ebcee2023-05-31 17:12:14 +01001379 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 '#':
Christian Brabandt360da402022-05-18 15:04:02 +01001506 if (regparse[0] == '=' && regparse[1] >= 48
1507 && regparse[1] <= 50)
1508 {
1509 // misplaced \%#=1
1510 semsg(_(e_atom_engine_must_be_at_start_of_pattern),
1511 regparse[1]);
1512 return FAIL;
1513 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001514 ret = regnode(CURSOR);
1515 break;
1516
1517 case 'V':
1518 ret = regnode(RE_VISUAL);
1519 break;
1520
1521 case 'C':
1522 ret = regnode(RE_COMPOSING);
1523 break;
1524
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001525 // \%[abc]: Emit as a list of branches, all ending at the last
1526 // branch which matches nothing.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001527 case '[':
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001528 if (one_exactly) // doesn't nest
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001529 EMSG_ONE_RET_NULL;
1530 {
1531 char_u *lastbranch;
1532 char_u *lastnode = NULL;
1533 char_u *br;
1534
1535 ret = NULL;
1536 while ((c = getchr()) != ']')
1537 {
1538 if (c == NUL)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001539 EMSG2_RET_NULL(_(e_missing_sb_after_str),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001540 reg_magic == MAGIC_ALL);
1541 br = regnode(BRANCH);
1542 if (ret == NULL)
1543 ret = br;
1544 else
1545 {
1546 regtail(lastnode, br);
1547 if (reg_toolong)
1548 return NULL;
1549 }
1550
1551 ungetchr();
1552 one_exactly = TRUE;
1553 lastnode = regatom(flagp);
1554 one_exactly = FALSE;
1555 if (lastnode == NULL)
1556 return NULL;
1557 }
1558 if (ret == NULL)
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001559 EMSG2_RET_NULL(_(e_empty_str_brackets),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001560 reg_magic == MAGIC_ALL);
1561 lastbranch = regnode(BRANCH);
1562 br = regnode(NOTHING);
1563 if (ret != JUST_CALC_SIZE)
1564 {
1565 regtail(lastnode, br);
1566 regtail(lastbranch, br);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001567 // connect all branches to the NOTHING
1568 // branch at the end
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001569 for (br = ret; br != lastnode; )
1570 {
1571 if (OP(br) == BRANCH)
1572 {
1573 regtail(br, lastbranch);
1574 if (reg_toolong)
1575 return NULL;
1576 br = OPERAND(br);
1577 }
1578 else
1579 br = regnext(br);
1580 }
1581 }
1582 *flagp &= ~(HASWIDTH | SIMPLE);
1583 break;
1584 }
1585
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001586 case 'd': // %d123 decimal
1587 case 'o': // %o123 octal
1588 case 'x': // %xab hex 2
1589 case 'u': // %uabcd hex 4
1590 case 'U': // %U1234abcd hex 8
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001591 {
1592 long i;
1593
1594 switch (c)
1595 {
1596 case 'd': i = getdecchrs(); break;
1597 case 'o': i = getoctchrs(); break;
1598 case 'x': i = gethexchrs(2); break;
1599 case 'u': i = gethexchrs(4); break;
1600 case 'U': i = gethexchrs(8); break;
1601 default: i = -1; break;
1602 }
1603
1604 if (i < 0 || i > INT_MAX)
1605 EMSG2_RET_NULL(
Bram Moolenaara6f79292022-01-04 21:30:47 +00001606 _(e_invalid_character_after_str_2),
1607 reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001608 if (use_multibytecode(i))
1609 ret = regnode(MULTIBYTECODE);
1610 else
1611 ret = regnode(EXACTLY);
1612 if (i == 0)
1613 regc(0x0a);
1614 else
1615 regmbc(i);
1616 regc(NUL);
1617 *flagp |= HASWIDTH;
1618 break;
1619 }
1620
1621 default:
1622 if (VIM_ISDIGIT(c) || c == '<' || c == '>'
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001623 || c == '\'' || c == '.')
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001624 {
1625 long_u n = 0;
1626 int cmp;
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001627 int cur = FALSE;
Bram Moolenaar72bb10d2022-04-05 14:00:32 +01001628 int got_digit = FALSE;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001629
1630 cmp = c;
1631 if (cmp == '<' || cmp == '>')
1632 c = getchr();
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001633 if (no_Magic(c) == '.')
1634 {
1635 cur = TRUE;
1636 c = getchr();
1637 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001638 while (VIM_ISDIGIT(c))
1639 {
Bram Moolenaar72bb10d2022-04-05 14:00:32 +01001640 got_digit = TRUE;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001641 n = n * 10 + (c - '0');
1642 c = getchr();
1643 }
Julio B46fa3c72024-03-28 10:23:37 +01001644 if (no_Magic(c) == '\'' && n == 0)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001645 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001646 // "\%'m", "\%<'m" and "\%>'m": Mark
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001647 c = getchr();
1648 ret = regnode(RE_MARK);
1649 if (ret == JUST_CALC_SIZE)
1650 regsize += 2;
1651 else
1652 {
1653 *regcode++ = c;
1654 *regcode++ = cmp;
1655 }
1656 break;
1657 }
Bram Moolenaar72bb10d2022-04-05 14:00:32 +01001658 else if ((c == 'l' || c == 'c' || c == 'v')
1659 && (cur || got_digit))
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001660 {
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001661 if (cur && n)
1662 {
Bram Moolenaar91ff3d42022-04-04 18:32:32 +01001663 semsg(_(e_regexp_number_after_dot_pos_search_chr),
1664 no_Magic(c));
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001665 rc_did_emsg = TRUE;
1666 return NULL;
1667 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001668 if (c == 'l')
1669 {
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001670 if (cur)
1671 n = curwin->w_cursor.lnum;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001672 ret = regnode(RE_LNUM);
1673 if (save_prev_at_start)
1674 at_start = TRUE;
1675 }
1676 else if (c == 'c')
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001677 {
1678 if (cur)
1679 {
1680 n = curwin->w_cursor.col;
1681 n++;
1682 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001683 ret = regnode(RE_COL);
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001684 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001685 else
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001686 {
1687 if (cur)
1688 {
1689 colnr_T vcol = 0;
1690
1691 getvvcol(curwin, &curwin->w_cursor,
1692 NULL, NULL, &vcol);
1693 ++vcol;
1694 n = vcol;
1695 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001696 ret = regnode(RE_VCOL);
Bram Moolenaar04db26b2021-07-05 20:15:23 +02001697 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001698 if (ret == JUST_CALC_SIZE)
1699 regsize += 5;
1700 else
1701 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001702 // put the number and the optional
1703 // comparator after the opcode
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001704 regcode = re_put_long(regcode, n);
1705 *regcode++ = cmp;
1706 }
1707 break;
1708 }
1709 }
1710
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001711 EMSG2_RET_NULL(_(e_invalid_character_after_str),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001712 reg_magic == MAGIC_ALL);
1713 }
1714 }
1715 break;
1716
1717 case Magic('['):
1718collection:
1719 {
1720 char_u *lp;
1721
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001722 // If there is no matching ']', we assume the '[' is a normal
1723 // character. This makes 'incsearch' and ":help [" work.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001724 lp = skip_anyof(regparse);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001725 if (*lp == ']') // there is a matching ']'
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001726 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001727 int startc = -1; // > 0 when next '-' is a range
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001728 int endc;
1729
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001730 // In a character class, different parsing rules apply.
1731 // Not even \ is special anymore, nothing is.
1732 if (*regparse == '^') // Complement of range.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001733 {
1734 ret = regnode(ANYBUT + extra);
1735 regparse++;
1736 }
1737 else
1738 ret = regnode(ANYOF + extra);
1739
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001740 // At the start ']' and '-' mean the literal character.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001741 if (*regparse == ']' || *regparse == '-')
1742 {
1743 startc = *regparse;
1744 regc(*regparse++);
1745 }
1746
1747 while (*regparse != NUL && *regparse != ']')
1748 {
1749 if (*regparse == '-')
1750 {
1751 ++regparse;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001752 // The '-' is not used for a range at the end and
1753 // after or before a '\n'.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001754 if (*regparse == ']' || *regparse == NUL
1755 || startc == -1
1756 || (regparse[0] == '\\' && regparse[1] == 'n'))
1757 {
1758 regc('-');
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001759 startc = '-'; // [--x] is a range
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001760 }
1761 else
1762 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001763 // Also accept "a-[.z.]"
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001764 endc = 0;
1765 if (*regparse == '[')
1766 endc = get_coll_element(&regparse);
1767 if (endc == 0)
1768 {
1769 if (has_mbyte)
1770 endc = mb_ptr2char_adv(&regparse);
1771 else
1772 endc = *regparse++;
1773 }
1774
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001775 // Handle \o40, \x20 and \u20AC style sequences
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001776 if (endc == '\\' && !reg_cpo_lit && !reg_cpo_bsl)
1777 endc = coll_get_char();
1778
1779 if (startc > endc)
Bram Moolenaar677658a2022-01-05 16:09:06 +00001780 EMSG_RET_NULL(_(e_reverse_range_in_character_class));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001781 if (has_mbyte && ((*mb_char2len)(startc) > 1
1782 || (*mb_char2len)(endc) > 1))
1783 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001784 // Limit to a range of 256 chars.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001785 if (endc > startc + 256)
Bram Moolenaar677658a2022-01-05 16:09:06 +00001786 EMSG_RET_NULL(_(e_range_too_large_in_character_class));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001787 while (++startc <= endc)
1788 regmbc(startc);
1789 }
1790 else
1791 {
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001792 while (++startc <= endc)
Bram Moolenaar424bcae2022-01-31 14:59:41 +00001793 regc(startc);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001794 }
1795 startc = -1;
1796 }
1797 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001798 // Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
1799 // accepts "\t", "\e", etc., but only when the 'l' flag in
1800 // 'cpoptions' is not included.
1801 // Posix doesn't recognize backslash at all.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001802 else if (*regparse == '\\'
1803 && !reg_cpo_bsl
1804 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
1805 || (!reg_cpo_lit
1806 && vim_strchr(REGEXP_ABBR,
1807 regparse[1]) != NULL)))
1808 {
1809 regparse++;
1810 if (*regparse == 'n')
1811 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001812 // '\n' in range: also match NL
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001813 if (ret != JUST_CALC_SIZE)
1814 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001815 // Using \n inside [^] does not change what
1816 // matches. "[^\n]" is the same as ".".
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001817 if (*ret == ANYOF)
1818 {
1819 *ret = ANYOF + ADD_NL;
1820 *flagp |= HASNL;
1821 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001822 // else: must have had a \n already
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001823 }
1824 regparse++;
1825 startc = -1;
1826 }
1827 else if (*regparse == 'd'
1828 || *regparse == 'o'
1829 || *regparse == 'x'
1830 || *regparse == 'u'
1831 || *regparse == 'U')
1832 {
1833 startc = coll_get_char();
1834 if (startc == 0)
1835 regc(0x0a);
1836 else
1837 regmbc(startc);
1838 }
1839 else
1840 {
1841 startc = backslash_trans(*regparse++);
1842 regc(startc);
1843 }
1844 }
1845 else if (*regparse == '[')
1846 {
1847 int c_class;
1848 int cu;
1849
1850 c_class = get_char_class(&regparse);
1851 startc = -1;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001852 // Characters assumed to be 8 bits!
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001853 switch (c_class)
1854 {
1855 case CLASS_NONE:
1856 c_class = get_equi_class(&regparse);
1857 if (c_class != 0)
1858 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001859 // produce equivalence class
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001860 reg_equi_class(c_class);
1861 }
1862 else if ((c_class =
1863 get_coll_element(&regparse)) != 0)
1864 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001865 // produce a collating element
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001866 regmbc(c_class);
1867 }
1868 else
1869 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001870 // literal '[', allow [[-x] as a range
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001871 startc = *regparse++;
1872 regc(startc);
1873 }
1874 break;
1875 case CLASS_ALNUM:
1876 for (cu = 1; cu < 128; cu++)
1877 if (isalnum(cu))
1878 regmbc(cu);
1879 break;
1880 case CLASS_ALPHA:
1881 for (cu = 1; cu < 128; cu++)
1882 if (isalpha(cu))
1883 regmbc(cu);
1884 break;
1885 case CLASS_BLANK:
1886 regc(' ');
1887 regc('\t');
1888 break;
1889 case CLASS_CNTRL:
1890 for (cu = 1; cu <= 127; cu++)
1891 if (iscntrl(cu))
1892 regmbc(cu);
1893 break;
1894 case CLASS_DIGIT:
1895 for (cu = 1; cu <= 127; cu++)
1896 if (VIM_ISDIGIT(cu))
1897 regmbc(cu);
1898 break;
1899 case CLASS_GRAPH:
1900 for (cu = 1; cu <= 127; cu++)
1901 if (isgraph(cu))
1902 regmbc(cu);
1903 break;
1904 case CLASS_LOWER:
1905 for (cu = 1; cu <= 255; cu++)
1906 if (MB_ISLOWER(cu) && cu != 170
1907 && cu != 186)
1908 regmbc(cu);
1909 break;
1910 case CLASS_PRINT:
1911 for (cu = 1; cu <= 255; cu++)
1912 if (vim_isprintc(cu))
1913 regmbc(cu);
1914 break;
1915 case CLASS_PUNCT:
1916 for (cu = 1; cu < 128; cu++)
1917 if (ispunct(cu))
1918 regmbc(cu);
1919 break;
1920 case CLASS_SPACE:
1921 for (cu = 9; cu <= 13; cu++)
1922 regc(cu);
1923 regc(' ');
1924 break;
1925 case CLASS_UPPER:
1926 for (cu = 1; cu <= 255; cu++)
1927 if (MB_ISUPPER(cu))
1928 regmbc(cu);
1929 break;
1930 case CLASS_XDIGIT:
1931 for (cu = 1; cu <= 255; cu++)
1932 if (vim_isxdigit(cu))
1933 regmbc(cu);
1934 break;
1935 case CLASS_TAB:
1936 regc('\t');
1937 break;
1938 case CLASS_RETURN:
1939 regc('\r');
1940 break;
1941 case CLASS_BACKSPACE:
1942 regc('\b');
1943 break;
1944 case CLASS_ESCAPE:
1945 regc('\033');
1946 break;
1947 case CLASS_IDENT:
1948 for (cu = 1; cu <= 255; cu++)
1949 if (vim_isIDc(cu))
1950 regmbc(cu);
1951 break;
1952 case CLASS_KEYWORD:
1953 for (cu = 1; cu <= 255; cu++)
1954 if (reg_iswordc(cu))
1955 regmbc(cu);
1956 break;
1957 case CLASS_FNAME:
1958 for (cu = 1; cu <= 255; cu++)
1959 if (vim_isfilec(cu))
1960 regmbc(cu);
1961 break;
1962 }
1963 }
1964 else
1965 {
1966 if (has_mbyte)
1967 {
1968 int len;
1969
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001970 // produce a multibyte character, including any
1971 // following composing characters
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001972 startc = mb_ptr2char(regparse);
1973 len = (*mb_ptr2len)(regparse);
1974 if (enc_utf8 && utf_char2len(startc) != len)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001975 startc = -1; // composing chars
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001976 while (--len >= 0)
1977 regc(*regparse++);
1978 }
1979 else
1980 {
1981 startc = *regparse++;
1982 regc(startc);
1983 }
1984 }
1985 }
1986 regc(NUL);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001987 prevchr_len = 1; // last char was the ']'
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001988 if (*regparse != ']')
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00001989 EMSG_RET_NULL(_(e_too_many_brackets)); // Cannot happen?
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001990 skipchr(); // let's be friends with the lexer again
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001991 *flagp |= HASWIDTH | SIMPLE;
1992 break;
1993 }
1994 else if (reg_strict)
Bram Moolenaar677658a2022-01-05 16:09:06 +00001995 EMSG2_RET_NULL(_(e_missing_rsb_after_str_lsb),
1996 reg_magic > MAGIC_OFF);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001997 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02001998 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02001999
2000 default:
2001 {
2002 int len;
2003
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002004 // A multi-byte character is handled as a separate atom if it's
2005 // before a multi and when it's a composing char.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002006 if (use_multibytecode(c))
2007 {
2008do_multibyte:
2009 ret = regnode(MULTIBYTECODE);
2010 regmbc(c);
2011 *flagp |= HASWIDTH | SIMPLE;
2012 break;
2013 }
2014
2015 ret = regnode(EXACTLY);
2016
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002017 // Append characters as long as:
2018 // - there is no following multi, we then need the character in
2019 // front of it as a single character operand
2020 // - not running into a Magic character
2021 // - "one_exactly" is not set
2022 // But always emit at least one character. Might be a Multi,
2023 // e.g., a "[" without matching "]".
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002024 for (len = 0; c != NUL && (len == 0
2025 || (re_multi_type(peekchr()) == NOT_MULTI
2026 && !one_exactly
2027 && !is_Magic(c))); ++len)
2028 {
2029 c = no_Magic(c);
2030 if (has_mbyte)
2031 {
2032 regmbc(c);
2033 if (enc_utf8)
2034 {
2035 int l;
2036
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002037 // Need to get composing character too.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002038 for (;;)
2039 {
2040 l = utf_ptr2len(regparse);
2041 if (!UTF_COMPOSINGLIKE(regparse, regparse + l))
2042 break;
2043 regmbc(utf_ptr2char(regparse));
2044 skipchr();
2045 }
2046 }
2047 }
2048 else
2049 regc(c);
2050 c = getchr();
2051 }
2052 ungetchr();
2053
2054 regc(NUL);
2055 *flagp |= HASWIDTH;
2056 if (len == 1)
2057 *flagp |= SIMPLE;
2058 }
2059 break;
2060 }
2061
2062 return ret;
2063}
2064
2065/*
2066 * Parse something followed by possible [*+=].
2067 *
2068 * Note that the branching code sequences used for = and the general cases
2069 * of * and + are somewhat optimized: they use the same NOTHING node as
2070 * both the endmarker for their branch list and the body of the last branch.
2071 * It might seem that this node could be dispensed with entirely, but the
2072 * endmarker role is not redundant.
2073 */
2074 static char_u *
2075regpiece(int *flagp)
2076{
2077 char_u *ret;
2078 int op;
2079 char_u *next;
2080 int flags;
2081 long minval;
2082 long maxval;
2083
2084 ret = regatom(&flags);
2085 if (ret == NULL)
2086 return NULL;
2087
2088 op = peekchr();
2089 if (re_multi_type(op) == NOT_MULTI)
2090 {
2091 *flagp = flags;
2092 return ret;
2093 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002094 // default flags
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002095 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
2096
2097 skipchr();
2098 switch (op)
2099 {
2100 case Magic('*'):
2101 if (flags & SIMPLE)
2102 reginsert(STAR, ret);
2103 else
2104 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002105 // Emit x* as (x&|), where & means "self".
2106 reginsert(BRANCH, ret); // Either x
2107 regoptail(ret, regnode(BACK)); // and loop
2108 regoptail(ret, ret); // back
2109 regtail(ret, regnode(BRANCH)); // or
2110 regtail(ret, regnode(NOTHING)); // null.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002111 }
2112 break;
2113
2114 case Magic('+'):
2115 if (flags & SIMPLE)
2116 reginsert(PLUS, ret);
2117 else
2118 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002119 // Emit x+ as x(&|), where & means "self".
2120 next = regnode(BRANCH); // Either
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002121 regtail(ret, next);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002122 regtail(regnode(BACK), ret); // loop back
2123 regtail(next, regnode(BRANCH)); // or
2124 regtail(ret, regnode(NOTHING)); // null.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002125 }
2126 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
2127 break;
2128
2129 case Magic('@'):
2130 {
2131 int lop = END;
2132 long nr;
2133
2134 nr = getdecchrs();
2135 switch (no_Magic(getchr()))
2136 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002137 case '=': lop = MATCH; break; // \@=
2138 case '!': lop = NOMATCH; break; // \@!
2139 case '>': lop = SUBPAT; break; // \@>
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002140 case '<': switch (no_Magic(getchr()))
2141 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002142 case '=': lop = BEHIND; break; // \@<=
2143 case '!': lop = NOBEHIND; break; // \@<!
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002144 }
2145 }
2146 if (lop == END)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002147 EMSG2_RET_NULL(_(e_invalid_character_after_str_at),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002148 reg_magic == MAGIC_ALL);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002149 // Look behind must match with behind_pos.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002150 if (lop == BEHIND || lop == NOBEHIND)
2151 {
2152 regtail(ret, regnode(BHPOS));
2153 *flagp |= HASLOOKBH;
2154 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002155 regtail(ret, regnode(END)); // operand ends
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002156 if (lop == BEHIND || lop == NOBEHIND)
2157 {
2158 if (nr < 0)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002159 nr = 0; // no limit is same as zero limit
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002160 reginsert_nr(lop, nr, ret);
2161 }
2162 else
2163 reginsert(lop, ret);
2164 break;
2165 }
2166
2167 case Magic('?'):
2168 case Magic('='):
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002169 // Emit x= as (x|)
2170 reginsert(BRANCH, ret); // Either x
2171 regtail(ret, regnode(BRANCH)); // or
2172 next = regnode(NOTHING); // null.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002173 regtail(ret, next);
2174 regoptail(ret, next);
2175 break;
2176
2177 case Magic('{'):
2178 if (!read_limits(&minval, &maxval))
2179 return NULL;
2180 if (flags & SIMPLE)
2181 {
2182 reginsert(BRACE_SIMPLE, ret);
2183 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
2184 }
2185 else
2186 {
2187 if (num_complex_braces >= 10)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002188 EMSG2_RET_NULL(_(e_too_many_complex_str_curly),
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002189 reg_magic == MAGIC_ALL);
2190 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
2191 regoptail(ret, regnode(BACK));
2192 regoptail(ret, ret);
2193 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
2194 ++num_complex_braces;
2195 }
2196 if (minval > 0 && maxval > 0)
2197 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
2198 break;
2199 }
2200 if (re_multi_type(peekchr()) != NOT_MULTI)
2201 {
2202 // Can't have a multi follow a multi.
2203 if (peekchr() == Magic('*'))
Bram Moolenaar12f3c1b2021-12-05 21:46:34 +00002204 EMSG2_RET_NULL(_(e_nested_str), reg_magic >= MAGIC_ON);
2205 EMSG3_RET_NULL(_(e_nested_str_chr), reg_magic == MAGIC_ALL,
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002206 no_Magic(peekchr()));
2207 }
2208
2209 return ret;
2210}
2211
2212/*
2213 * Parse one alternative of an | or & operator.
2214 * Implements the concatenation operator.
2215 */
2216 static char_u *
2217regconcat(int *flagp)
2218{
2219 char_u *first = NULL;
2220 char_u *chain = NULL;
2221 char_u *latest;
2222 int flags;
2223 int cont = TRUE;
2224
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002225 *flagp = WORST; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002226
2227 while (cont)
2228 {
2229 switch (peekchr())
2230 {
2231 case NUL:
2232 case Magic('|'):
2233 case Magic('&'):
2234 case Magic(')'):
2235 cont = FALSE;
2236 break;
2237 case Magic('Z'):
2238 regflags |= RF_ICOMBINE;
2239 skipchr_keepstart();
2240 break;
2241 case Magic('c'):
2242 regflags |= RF_ICASE;
2243 skipchr_keepstart();
2244 break;
2245 case Magic('C'):
2246 regflags |= RF_NOICASE;
2247 skipchr_keepstart();
2248 break;
2249 case Magic('v'):
2250 reg_magic = MAGIC_ALL;
2251 skipchr_keepstart();
2252 curchr = -1;
2253 break;
2254 case Magic('m'):
2255 reg_magic = MAGIC_ON;
2256 skipchr_keepstart();
2257 curchr = -1;
2258 break;
2259 case Magic('M'):
2260 reg_magic = MAGIC_OFF;
2261 skipchr_keepstart();
2262 curchr = -1;
2263 break;
2264 case Magic('V'):
2265 reg_magic = MAGIC_NONE;
2266 skipchr_keepstart();
2267 curchr = -1;
2268 break;
2269 default:
2270 latest = regpiece(&flags);
2271 if (latest == NULL || reg_toolong)
2272 return NULL;
2273 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002274 if (chain == NULL) // First piece.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002275 *flagp |= flags & SPSTART;
2276 else
2277 regtail(chain, latest);
2278 chain = latest;
2279 if (first == NULL)
2280 first = latest;
2281 break;
2282 }
2283 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002284 if (first == NULL) // Loop ran zero times.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002285 first = regnode(NOTHING);
2286 return first;
2287}
2288
2289/*
2290 * Parse one alternative of an | operator.
2291 * Implements the & operator.
2292 */
2293 static char_u *
2294regbranch(int *flagp)
2295{
2296 char_u *ret;
2297 char_u *chain = NULL;
2298 char_u *latest;
2299 int flags;
2300
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002301 *flagp = WORST | HASNL; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002302
2303 ret = regnode(BRANCH);
2304 for (;;)
2305 {
2306 latest = regconcat(&flags);
2307 if (latest == NULL)
2308 return NULL;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002309 // If one of the branches has width, the whole thing has. If one of
2310 // the branches anchors at start-of-line, the whole thing does.
2311 // If one of the branches uses look-behind, the whole thing does.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002312 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002313 // If one of the branches doesn't match a line-break, the whole thing
2314 // doesn't.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002315 *flagp &= ~HASNL | (flags & HASNL);
2316 if (chain != NULL)
2317 regtail(chain, latest);
2318 if (peekchr() != Magic('&'))
2319 break;
2320 skipchr();
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002321 regtail(latest, regnode(END)); // operand ends
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002322 if (reg_toolong)
2323 break;
2324 reginsert(MATCH, latest);
2325 chain = latest;
2326 }
2327
2328 return ret;
2329}
2330
2331/*
2332 * Parse regular expression, i.e. main body or parenthesized thing.
2333 *
2334 * Caller must absorb opening parenthesis.
2335 *
2336 * Combining parenthesis handling with the base level of regular expression
2337 * is a trifle forced, but the need to tie the tails of the branches to what
2338 * follows makes it hard to avoid.
2339 */
2340 static char_u *
2341reg(
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002342 int paren, // REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002343 int *flagp)
2344{
2345 char_u *ret;
2346 char_u *br;
2347 char_u *ender;
2348 int parno = 0;
2349 int flags;
2350
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002351 *flagp = HASWIDTH; // Tentatively.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002352
2353#ifdef FEAT_SYN_HL
2354 if (paren == REG_ZPAREN)
2355 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002356 // Make a ZOPEN node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002357 if (regnzpar >= NSUBEXP)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002358 EMSG_RET_NULL(_(e_too_many_z));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002359 parno = regnzpar;
2360 regnzpar++;
2361 ret = regnode(ZOPEN + parno);
2362 }
2363 else
2364#endif
2365 if (paren == REG_PAREN)
2366 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002367 // Make a MOPEN node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002368 if (regnpar >= NSUBEXP)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002369 EMSG2_RET_NULL(_(e_too_many_str_open), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002370 parno = regnpar;
2371 ++regnpar;
2372 ret = regnode(MOPEN + parno);
2373 }
2374 else if (paren == REG_NPAREN)
2375 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002376 // Make a NOPEN node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002377 ret = regnode(NOPEN);
2378 }
2379 else
2380 ret = NULL;
2381
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002382 // Pick up the branches, linking them together.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002383 br = regbranch(&flags);
2384 if (br == NULL)
2385 return NULL;
2386 if (ret != NULL)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002387 regtail(ret, br); // [MZ]OPEN -> first.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002388 else
2389 ret = br;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002390 // If one of the branches can be zero-width, the whole thing can.
2391 // If one of the branches has * at start or matches a line-break, the
2392 // whole thing can.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002393 if (!(flags & HASWIDTH))
2394 *flagp &= ~HASWIDTH;
2395 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
2396 while (peekchr() == Magic('|'))
2397 {
2398 skipchr();
2399 br = regbranch(&flags);
2400 if (br == NULL || reg_toolong)
2401 return NULL;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002402 regtail(ret, br); // BRANCH -> BRANCH.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002403 if (!(flags & HASWIDTH))
2404 *flagp &= ~HASWIDTH;
2405 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
2406 }
2407
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002408 // Make a closing node, and hook it on the end.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002409 ender = regnode(
2410#ifdef FEAT_SYN_HL
2411 paren == REG_ZPAREN ? ZCLOSE + parno :
2412#endif
2413 paren == REG_PAREN ? MCLOSE + parno :
2414 paren == REG_NPAREN ? NCLOSE : END);
2415 regtail(ret, ender);
2416
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002417 // Hook the tails of the branches to the closing node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002418 for (br = ret; br != NULL; br = regnext(br))
2419 regoptail(br, ender);
2420
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002421 // Check for proper termination.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002422 if (paren != REG_NOPAREN && getchr() != Magic(')'))
2423 {
2424#ifdef FEAT_SYN_HL
2425 if (paren == REG_ZPAREN)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002426 EMSG_RET_NULL(_(e_unmatched_z));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002427 else
2428#endif
2429 if (paren == REG_NPAREN)
Bram Moolenaard8e44472021-07-21 22:20:33 +02002430 EMSG2_RET_NULL(_(e_unmatched_str_percent_open), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002431 else
Bram Moolenaard8e44472021-07-21 22:20:33 +02002432 EMSG2_RET_NULL(_(e_unmatched_str_open), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002433 }
2434 else if (paren == REG_NOPAREN && peekchr() != NUL)
2435 {
2436 if (curchr == Magic(')'))
Bram Moolenaard8e44472021-07-21 22:20:33 +02002437 EMSG2_RET_NULL(_(e_unmatched_str_close), reg_magic == MAGIC_ALL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002438 else
Bram Moolenaar74409f62022-01-01 15:58:22 +00002439 EMSG_RET_NULL(_(e_trailing_characters)); // "Can't happen".
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002440 // NOTREACHED
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002441 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002442 // Here we set the flag allowing back references to this set of
2443 // parentheses.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002444 if (paren == REG_PAREN)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002445 had_endbrace[parno] = TRUE; // have seen the close paren
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002446 return ret;
2447}
2448
2449/*
2450 * bt_regcomp() - compile a regular expression into internal code for the
2451 * traditional back track matcher.
2452 * Returns the program in allocated space. Returns NULL for an error.
2453 *
2454 * We can't allocate space until we know how big the compiled form will be,
2455 * but we can't compile it (and thus know how big it is) until we've got a
2456 * place to put the code. So we cheat: we compile it twice, once with code
2457 * generation turned off and size counting turned on, and once "for real".
2458 * This also means that we don't allocate space until we are sure that the
2459 * thing really will compile successfully, and we never have to move the
2460 * code and thus invalidate pointers into it. (Note that it has to be in
2461 * one piece because vim_free() must be able to free it all.)
2462 *
2463 * Whether upper/lower case is to be ignored is decided when executing the
2464 * program, it does not matter here.
2465 *
2466 * Beware that the optimization-preparation code in here knows about some
2467 * of the structure of the compiled regexp.
2468 * "re_flags": RE_MAGIC and/or RE_STRING.
2469 */
2470 static regprog_T *
2471bt_regcomp(char_u *expr, int re_flags)
2472{
2473 bt_regprog_T *r;
2474 char_u *scan;
2475 char_u *longest;
2476 int len;
2477 int flags;
2478
2479 if (expr == NULL)
RestorerZ68ebcee2023-05-31 17:12:14 +01002480 IEMSG_RET_NULL(e_null_argument);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002481
2482 init_class_tab();
2483
2484 // First pass: determine size, legality.
2485 regcomp_start(expr, re_flags);
2486 regcode = JUST_CALC_SIZE;
2487 regc(REGMAGIC);
2488 if (reg(REG_NOPAREN, &flags) == NULL)
2489 return NULL;
2490
2491 // Allocate space.
2492 r = alloc(offsetof(bt_regprog_T, program) + regsize);
2493 if (r == NULL)
2494 return NULL;
2495 r->re_in_use = FALSE;
2496
2497 // Second pass: emit code.
2498 regcomp_start(expr, re_flags);
2499 regcode = r->program;
2500 regc(REGMAGIC);
2501 if (reg(REG_NOPAREN, &flags) == NULL || reg_toolong)
2502 {
2503 vim_free(r);
2504 if (reg_toolong)
Bram Moolenaareaaac012022-01-02 17:00:40 +00002505 EMSG_RET_NULL(_(e_pattern_too_long));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002506 return NULL;
2507 }
2508
2509 // Dig out information for optimizations.
2510 r->regstart = NUL; // Worst-case defaults.
2511 r->reganch = 0;
2512 r->regmust = NULL;
2513 r->regmlen = 0;
2514 r->regflags = regflags;
2515 if (flags & HASNL)
2516 r->regflags |= RF_HASNL;
2517 if (flags & HASLOOKBH)
2518 r->regflags |= RF_LOOKBH;
2519#ifdef FEAT_SYN_HL
2520 // Remember whether this pattern has any \z specials in it.
2521 r->reghasz = re_has_z;
2522#endif
2523 scan = r->program + 1; // First BRANCH.
2524 if (OP(regnext(scan)) == END) // Only one top-level choice.
2525 {
2526 scan = OPERAND(scan);
2527
2528 // Starting-point info.
2529 if (OP(scan) == BOL || OP(scan) == RE_BOF)
2530 {
2531 r->reganch++;
2532 scan = regnext(scan);
2533 }
2534
2535 if (OP(scan) == EXACTLY)
2536 {
2537 if (has_mbyte)
2538 r->regstart = (*mb_ptr2char)(OPERAND(scan));
2539 else
2540 r->regstart = *OPERAND(scan);
2541 }
2542 else if ((OP(scan) == BOW
2543 || OP(scan) == EOW
2544 || OP(scan) == NOTHING
2545 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
2546 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
2547 && OP(regnext(scan)) == EXACTLY)
2548 {
2549 if (has_mbyte)
2550 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
2551 else
2552 r->regstart = *OPERAND(regnext(scan));
2553 }
2554
2555 // If there's something expensive in the r.e., find the longest
2556 // literal string that must appear and make it the regmust. Resolve
2557 // ties in favor of later strings, since the regstart check works
2558 // with the beginning of the r.e. and avoiding duplication
2559 // strengthens checking. Not a strong reason, but sufficient in the
2560 // absence of others.
2561
2562 // When the r.e. starts with BOW, it is faster to look for a regmust
2563 // first. Used a lot for "#" and "*" commands. (Added by mool).
2564 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
2565 && !(flags & HASNL))
2566 {
2567 longest = NULL;
2568 len = 0;
2569 for (; scan != NULL; scan = regnext(scan))
2570 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
2571 {
2572 longest = OPERAND(scan);
2573 len = (int)STRLEN(OPERAND(scan));
2574 }
2575 r->regmust = longest;
2576 r->regmlen = len;
2577 }
2578 }
2579#ifdef BT_REGEXP_DUMP
2580 regdump(expr, r);
2581#endif
2582 r->engine = &bt_regengine;
2583 return (regprog_T *)r;
2584}
2585
2586#if defined(FEAT_SYN_HL) || defined(PROTO)
2587/*
2588 * Check if during the previous call to vim_regcomp the EOL item "$" has been
2589 * found. This is messy, but it works fine.
2590 */
2591 int
2592vim_regcomp_had_eol(void)
2593{
2594 return had_eol;
2595}
2596#endif
2597
2598/*
2599 * Get a number after a backslash that is inside [].
2600 * When nothing is recognized return a backslash.
2601 */
2602 static int
2603coll_get_char(void)
2604{
2605 long nr = -1;
2606
2607 switch (*regparse++)
2608 {
2609 case 'd': nr = getdecchrs(); break;
2610 case 'o': nr = getoctchrs(); break;
2611 case 'x': nr = gethexchrs(2); break;
2612 case 'u': nr = gethexchrs(4); break;
2613 case 'U': nr = gethexchrs(8); break;
2614 }
2615 if (nr < 0 || nr > INT_MAX)
2616 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002617 // If getting the number fails be backwards compatible: the character
2618 // is a backslash.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002619 --regparse;
2620 nr = '\\';
2621 }
2622 return nr;
2623}
2624
2625/*
2626 * Free a compiled regexp program, returned by bt_regcomp().
2627 */
2628 static void
2629bt_regfree(regprog_T *prog)
2630{
2631 vim_free(prog);
2632}
2633
2634#define ADVANCE_REGINPUT() MB_PTR_ADV(rex.input)
2635
2636/*
2637 * The arguments from BRACE_LIMITS are stored here. They are actually local
2638 * to regmatch(), but they are here to reduce the amount of stack space used
2639 * (it can be called recursively many times).
2640 */
2641static long bl_minval;
2642static long bl_maxval;
2643
2644/*
2645 * Save the input line and position in a regsave_T.
2646 */
2647 static void
2648reg_save(regsave_T *save, garray_T *gap)
2649{
2650 if (REG_MULTI)
2651 {
2652 save->rs_u.pos.col = (colnr_T)(rex.input - rex.line);
2653 save->rs_u.pos.lnum = rex.lnum;
2654 }
2655 else
2656 save->rs_u.ptr = rex.input;
2657 save->rs_len = gap->ga_len;
2658}
2659
2660/*
2661 * Restore the input line and position from a regsave_T.
2662 */
2663 static void
2664reg_restore(regsave_T *save, garray_T *gap)
2665{
2666 if (REG_MULTI)
2667 {
2668 if (rex.lnum != save->rs_u.pos.lnum)
2669 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002670 // only call reg_getline() when the line number changed to save
2671 // a bit of time
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002672 rex.lnum = save->rs_u.pos.lnum;
2673 rex.line = reg_getline(rex.lnum);
2674 }
2675 rex.input = rex.line + save->rs_u.pos.col;
2676 }
2677 else
2678 rex.input = save->rs_u.ptr;
2679 gap->ga_len = save->rs_len;
2680}
2681
2682/*
2683 * Return TRUE if current position is equal to saved position.
2684 */
2685 static int
2686reg_save_equal(regsave_T *save)
2687{
2688 if (REG_MULTI)
2689 return rex.lnum == save->rs_u.pos.lnum
2690 && rex.input == rex.line + save->rs_u.pos.col;
2691 return rex.input == save->rs_u.ptr;
2692}
2693
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002694// Save the sub-expressions before attempting a match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002695#define save_se(savep, posp, pp) \
2696 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
2697
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002698// After a failed match restore the sub-expressions.
Bram Moolenaarebfec1c2023-01-22 21:14:53 +00002699#define restore_se(savep, posp, pp) \
2700{ \
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002701 if (REG_MULTI) \
2702 *(posp) = (savep)->se_u.pos; \
2703 else \
Bram Moolenaarebfec1c2023-01-22 21:14:53 +00002704 *(pp) = (savep)->se_u.ptr; \
2705}
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002706
2707/*
2708 * Tentatively set the sub-expression start to the current position (after
2709 * calling regmatch() they will have changed). Need to save the existing
2710 * values for when there is no match.
2711 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
2712 * depending on REG_MULTI.
2713 */
2714 static void
2715save_se_multi(save_se_T *savep, lpos_T *posp)
2716{
2717 savep->se_u.pos = *posp;
2718 posp->lnum = rex.lnum;
2719 posp->col = (colnr_T)(rex.input - rex.line);
2720}
2721
2722 static void
2723save_se_one(save_se_T *savep, char_u **pp)
2724{
2725 savep->se_u.ptr = *pp;
2726 *pp = rex.input;
2727}
2728
2729/*
2730 * regrepeat - repeatedly match something simple, return how many.
2731 * Advances rex.input (and rex.lnum) to just after the matched chars.
2732 */
2733 static int
2734regrepeat(
2735 char_u *p,
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002736 long maxcount) // maximum number of matches allowed
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002737{
2738 long count = 0;
2739 char_u *scan;
2740 char_u *opnd;
2741 int mask;
2742 int testval = 0;
2743
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002744 scan = rex.input; // Make local copy of rex.input for speed.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002745 opnd = OPERAND(p);
2746 switch (OP(p))
2747 {
2748 case ANY:
2749 case ANY + ADD_NL:
2750 while (count < maxcount)
2751 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002752 // Matching anything means we continue until end-of-line (or
2753 // end-of-file for ANY + ADD_NL), only limited by maxcount.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002754 while (*scan != NUL && count < maxcount)
2755 {
2756 ++count;
2757 MB_PTR_ADV(scan);
2758 }
2759 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2760 || rex.reg_line_lbr || count == maxcount)
2761 break;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002762 ++count; // count the line-break
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002763 reg_nextline();
2764 scan = rex.input;
2765 if (got_int)
2766 break;
2767 }
2768 break;
2769
2770 case IDENT:
2771 case IDENT + ADD_NL:
2772 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002773 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002774 case SIDENT:
2775 case SIDENT + ADD_NL:
2776 while (count < maxcount)
2777 {
2778 if (vim_isIDc(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan)))
2779 {
2780 MB_PTR_ADV(scan);
2781 }
2782 else if (*scan == NUL)
2783 {
2784 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2785 || rex.reg_line_lbr)
2786 break;
2787 reg_nextline();
2788 scan = rex.input;
2789 if (got_int)
2790 break;
2791 }
2792 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2793 ++scan;
2794 else
2795 break;
2796 ++count;
2797 }
2798 break;
2799
2800 case KWORD:
2801 case KWORD + ADD_NL:
2802 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002803 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002804 case SKWORD:
2805 case SKWORD + ADD_NL:
2806 while (count < maxcount)
2807 {
2808 if (vim_iswordp_buf(scan, rex.reg_buf)
2809 && (testval || !VIM_ISDIGIT(*scan)))
2810 {
2811 MB_PTR_ADV(scan);
2812 }
2813 else if (*scan == NUL)
2814 {
2815 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2816 || rex.reg_line_lbr)
2817 break;
2818 reg_nextline();
2819 scan = rex.input;
2820 if (got_int)
2821 break;
2822 }
2823 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2824 ++scan;
2825 else
2826 break;
2827 ++count;
2828 }
2829 break;
2830
2831 case FNAME:
2832 case FNAME + ADD_NL:
2833 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002834 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002835 case SFNAME:
2836 case SFNAME + ADD_NL:
2837 while (count < maxcount)
2838 {
2839 if (vim_isfilec(PTR2CHAR(scan)) && (testval || !VIM_ISDIGIT(*scan)))
2840 {
2841 MB_PTR_ADV(scan);
2842 }
2843 else if (*scan == NUL)
2844 {
2845 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2846 || rex.reg_line_lbr)
2847 break;
2848 reg_nextline();
2849 scan = rex.input;
2850 if (got_int)
2851 break;
2852 }
2853 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2854 ++scan;
2855 else
2856 break;
2857 ++count;
2858 }
2859 break;
2860
2861 case PRINT:
2862 case PRINT + ADD_NL:
2863 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002864 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02002865 case SPRINT:
2866 case SPRINT + ADD_NL:
2867 while (count < maxcount)
2868 {
2869 if (*scan == NUL)
2870 {
2871 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2872 || rex.reg_line_lbr)
2873 break;
2874 reg_nextline();
2875 scan = rex.input;
2876 if (got_int)
2877 break;
2878 }
2879 else if (vim_isprintc(PTR2CHAR(scan)) == 1
2880 && (testval || !VIM_ISDIGIT(*scan)))
2881 {
2882 MB_PTR_ADV(scan);
2883 }
2884 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2885 ++scan;
2886 else
2887 break;
2888 ++count;
2889 }
2890 break;
2891
2892 case WHITE:
2893 case WHITE + ADD_NL:
2894 testval = mask = RI_WHITE;
2895do_class:
2896 while (count < maxcount)
2897 {
2898 int l;
2899
2900 if (*scan == NUL)
2901 {
2902 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
2903 || rex.reg_line_lbr)
2904 break;
2905 reg_nextline();
2906 scan = rex.input;
2907 if (got_int)
2908 break;
2909 }
2910 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1)
2911 {
2912 if (testval != 0)
2913 break;
2914 scan += l;
2915 }
2916 else if ((class_tab[*scan] & mask) == testval)
2917 ++scan;
2918 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
2919 ++scan;
2920 else
2921 break;
2922 ++count;
2923 }
2924 break;
2925
2926 case NWHITE:
2927 case NWHITE + ADD_NL:
2928 mask = RI_WHITE;
2929 goto do_class;
2930 case DIGIT:
2931 case DIGIT + ADD_NL:
2932 testval = mask = RI_DIGIT;
2933 goto do_class;
2934 case NDIGIT:
2935 case NDIGIT + ADD_NL:
2936 mask = RI_DIGIT;
2937 goto do_class;
2938 case HEX:
2939 case HEX + ADD_NL:
2940 testval = mask = RI_HEX;
2941 goto do_class;
2942 case NHEX:
2943 case NHEX + ADD_NL:
2944 mask = RI_HEX;
2945 goto do_class;
2946 case OCTAL:
2947 case OCTAL + ADD_NL:
2948 testval = mask = RI_OCTAL;
2949 goto do_class;
2950 case NOCTAL:
2951 case NOCTAL + ADD_NL:
2952 mask = RI_OCTAL;
2953 goto do_class;
2954 case WORD:
2955 case WORD + ADD_NL:
2956 testval = mask = RI_WORD;
2957 goto do_class;
2958 case NWORD:
2959 case NWORD + ADD_NL:
2960 mask = RI_WORD;
2961 goto do_class;
2962 case HEAD:
2963 case HEAD + ADD_NL:
2964 testval = mask = RI_HEAD;
2965 goto do_class;
2966 case NHEAD:
2967 case NHEAD + ADD_NL:
2968 mask = RI_HEAD;
2969 goto do_class;
2970 case ALPHA:
2971 case ALPHA + ADD_NL:
2972 testval = mask = RI_ALPHA;
2973 goto do_class;
2974 case NALPHA:
2975 case NALPHA + ADD_NL:
2976 mask = RI_ALPHA;
2977 goto do_class;
2978 case LOWER:
2979 case LOWER + ADD_NL:
2980 testval = mask = RI_LOWER;
2981 goto do_class;
2982 case NLOWER:
2983 case NLOWER + ADD_NL:
2984 mask = RI_LOWER;
2985 goto do_class;
2986 case UPPER:
2987 case UPPER + ADD_NL:
2988 testval = mask = RI_UPPER;
2989 goto do_class;
2990 case NUPPER:
2991 case NUPPER + ADD_NL:
2992 mask = RI_UPPER;
2993 goto do_class;
2994
2995 case EXACTLY:
2996 {
2997 int cu, cl;
2998
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02002999 // This doesn't do a multi-byte character, because a MULTIBYTECODE
3000 // would have been used for it. It does handle single-byte
3001 // characters, such as latin1.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003002 if (rex.reg_ic)
3003 {
3004 cu = MB_TOUPPER(*opnd);
3005 cl = MB_TOLOWER(*opnd);
3006 while (count < maxcount && (*scan == cu || *scan == cl))
3007 {
3008 count++;
3009 scan++;
3010 }
3011 }
3012 else
3013 {
3014 cu = *opnd;
3015 while (count < maxcount && *scan == cu)
3016 {
3017 count++;
3018 scan++;
3019 }
3020 }
3021 break;
3022 }
3023
3024 case MULTIBYTECODE:
3025 {
3026 int i, len, cf = 0;
3027
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003028 // Safety check (just in case 'encoding' was changed since
3029 // compiling the program).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003030 if ((len = (*mb_ptr2len)(opnd)) > 1)
3031 {
3032 if (rex.reg_ic && enc_utf8)
3033 cf = utf_fold(utf_ptr2char(opnd));
3034 while (count < maxcount && (*mb_ptr2len)(scan) >= len)
3035 {
3036 for (i = 0; i < len; ++i)
3037 if (opnd[i] != scan[i])
3038 break;
3039 if (i < len && (!rex.reg_ic || !enc_utf8
3040 || utf_fold(utf_ptr2char(scan)) != cf))
3041 break;
3042 scan += len;
3043 ++count;
3044 }
3045 }
3046 }
3047 break;
3048
3049 case ANYOF:
3050 case ANYOF + ADD_NL:
3051 testval = TRUE;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003052 // FALLTHROUGH
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003053
3054 case ANYBUT:
3055 case ANYBUT + ADD_NL:
3056 while (count < maxcount)
3057 {
3058 int len;
3059
3060 if (*scan == NUL)
3061 {
3062 if (!REG_MULTI || !WITH_NL(OP(p)) || rex.lnum > rex.reg_maxline
3063 || rex.reg_line_lbr)
3064 break;
3065 reg_nextline();
3066 scan = rex.input;
3067 if (got_int)
3068 break;
3069 }
3070 else if (rex.reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
3071 ++scan;
3072 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1)
3073 {
3074 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
3075 break;
3076 scan += len;
3077 }
3078 else
3079 {
3080 if ((cstrchr(opnd, *scan) == NULL) == testval)
3081 break;
3082 ++scan;
3083 }
3084 ++count;
3085 }
3086 break;
3087
3088 case NEWL:
3089 while (count < maxcount
3090 && ((*scan == NUL && rex.lnum <= rex.reg_maxline
3091 && !rex.reg_line_lbr && REG_MULTI)
3092 || (*scan == '\n' && rex.reg_line_lbr)))
3093 {
3094 count++;
3095 if (rex.reg_line_lbr)
3096 ADVANCE_REGINPUT();
3097 else
3098 reg_nextline();
3099 scan = rex.input;
3100 if (got_int)
3101 break;
3102 }
3103 break;
3104
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003105 default: // Oh dear. Called inappropriately.
RestorerZ68ebcee2023-05-31 17:12:14 +01003106 iemsg(e_corrupted_regexp_program);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003107#ifdef DEBUG
3108 printf("Called regrepeat with op code %d\n", OP(p));
3109#endif
3110 break;
3111 }
3112
3113 rex.input = scan;
3114
3115 return (int)count;
3116}
3117
3118/*
3119 * Push an item onto the regstack.
3120 * Returns pointer to new item. Returns NULL when out of memory.
3121 */
3122 static regitem_T *
3123regstack_push(regstate_T state, char_u *scan)
3124{
3125 regitem_T *rp;
3126
3127 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
3128 {
Bram Moolenaar74409f62022-01-01 15:58:22 +00003129 emsg(_(e_pattern_uses_more_memory_than_maxmempattern));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003130 return NULL;
3131 }
3132 if (ga_grow(&regstack, sizeof(regitem_T)) == FAIL)
3133 return NULL;
3134
3135 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len);
3136 rp->rs_state = state;
3137 rp->rs_scan = scan;
3138
3139 regstack.ga_len += sizeof(regitem_T);
3140 return rp;
3141}
3142
3143/*
3144 * Pop an item from the regstack.
3145 */
3146 static void
3147regstack_pop(char_u **scan)
3148{
3149 regitem_T *rp;
3150
3151 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
3152 *scan = rp->rs_scan;
3153
3154 regstack.ga_len -= sizeof(regitem_T);
3155}
3156
Bram Moolenaar616592e2022-06-17 15:17:10 +01003157#ifdef FEAT_RELTIME
3158/*
3159 * Check if the timer expired, return TRUE if so.
3160 */
3161 static int
3162bt_did_time_out(int *timed_out)
3163{
3164 if (*timeout_flag)
3165 {
3166 if (timed_out != NULL)
3167 {
Bram Moolenaar4c5678f2022-11-30 18:12:19 +00003168# ifdef FEAT_EVAL
Bram Moolenaar616592e2022-06-17 15:17:10 +01003169 if (!*timed_out)
3170 ch_log(NULL, "BT regexp timed out");
Bram Moolenaar509ce032022-06-20 11:23:01 +01003171# endif
Bram Moolenaar616592e2022-06-17 15:17:10 +01003172 *timed_out = TRUE;
3173 }
3174 return TRUE;
3175 }
3176 return FALSE;
3177}
3178#endif
3179
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003180/*
3181 * Save the current subexpr to "bp", so that they can be restored
3182 * later by restore_subexpr().
3183 */
3184 static void
3185save_subexpr(regbehind_T *bp)
3186{
3187 int i;
3188
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003189 // When "rex.need_clear_subexpr" is set we don't need to save the values,
3190 // only remember that this flag needs to be set again when restoring.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003191 bp->save_need_clear_subexpr = rex.need_clear_subexpr;
Yegappan Lakshmananf97a2952023-01-18 18:17:48 +00003192 if (rex.need_clear_subexpr)
3193 return;
3194
3195 for (i = 0; i < NSUBEXP; ++i)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003196 {
Yegappan Lakshmananf97a2952023-01-18 18:17:48 +00003197 if (REG_MULTI)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003198 {
Yegappan Lakshmananf97a2952023-01-18 18:17:48 +00003199 bp->save_start[i].se_u.pos = rex.reg_startpos[i];
3200 bp->save_end[i].se_u.pos = rex.reg_endpos[i];
3201 }
3202 else
3203 {
3204 bp->save_start[i].se_u.ptr = rex.reg_startp[i];
3205 bp->save_end[i].se_u.ptr = rex.reg_endp[i];
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003206 }
3207 }
3208}
3209
3210/*
3211 * Restore the subexpr from "bp".
3212 */
3213 static void
3214restore_subexpr(regbehind_T *bp)
3215{
3216 int i;
3217
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003218 // Only need to restore saved values when they are not to be cleared.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003219 rex.need_clear_subexpr = bp->save_need_clear_subexpr;
Yegappan Lakshmananf97a2952023-01-18 18:17:48 +00003220 if (rex.need_clear_subexpr)
3221 return;
3222
3223 for (i = 0; i < NSUBEXP; ++i)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003224 {
Yegappan Lakshmananf97a2952023-01-18 18:17:48 +00003225 if (REG_MULTI)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003226 {
Yegappan Lakshmananf97a2952023-01-18 18:17:48 +00003227 rex.reg_startpos[i] = bp->save_start[i].se_u.pos;
3228 rex.reg_endpos[i] = bp->save_end[i].se_u.pos;
3229 }
3230 else
3231 {
3232 rex.reg_startp[i] = bp->save_start[i].se_u.ptr;
3233 rex.reg_endp[i] = bp->save_end[i].se_u.ptr;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003234 }
3235 }
3236}
3237
3238/*
3239 * regmatch - main matching routine
3240 *
3241 * Conceptually the strategy is simple: Check to see whether the current node
3242 * matches, push an item onto the regstack and loop to see whether the rest
3243 * matches, and then act accordingly. In practice we make some effort to
3244 * avoid using the regstack, in particular by going through "ordinary" nodes
3245 * (that don't need to know whether the rest of the match failed) by a nested
3246 * loop.
3247 *
3248 * Returns TRUE when there is a match. Leaves rex.input and rex.lnum just after
3249 * the last matched character.
3250 * Returns FALSE when there is no match. Leaves rex.input and rex.lnum in an
3251 * undefined state!
3252 */
3253 static int
3254regmatch(
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003255 char_u *scan, // Current node.
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003256 int *timed_out UNUSED) // flag set on timeout or NULL
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003257{
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003258 char_u *next; // Next node.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003259 int op;
3260 int c;
3261 regitem_T *rp;
3262 int no;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003263 int status; // one of the RA_ values:
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003264
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003265 // Make "regstack" and "backpos" empty. They are allocated and freed in
3266 // bt_regexec_both() to reduce malloc()/free() calls.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003267 regstack.ga_len = 0;
3268 backpos.ga_len = 0;
3269
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003270 // Repeat until "regstack" is empty.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003271 for (;;)
3272 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003273 // Some patterns may take a long time to match, e.g., "\([a-z]\+\)\+Q".
3274 // Allow interrupting them with CTRL-C.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003275 fast_breakcheck();
3276
3277#ifdef DEBUG
3278 if (scan != NULL && regnarrate)
3279 {
3280 mch_errmsg((char *)regprop(scan));
3281 mch_errmsg("(\n");
3282 }
3283#endif
3284
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003285 // Repeat for items that can be matched sequentially, without using the
3286 // regstack.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003287 for (;;)
3288 {
3289 if (got_int || scan == NULL)
3290 {
3291 status = RA_FAIL;
3292 break;
3293 }
3294#ifdef FEAT_RELTIME
Bram Moolenaar616592e2022-06-17 15:17:10 +01003295 if (bt_did_time_out(timed_out))
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003296 {
Paul Ollis65745772022-06-05 16:55:54 +01003297 status = RA_FAIL;
3298 break;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003299 }
3300#endif
3301 status = RA_CONT;
3302
3303#ifdef DEBUG
3304 if (regnarrate)
3305 {
3306 mch_errmsg((char *)regprop(scan));
3307 mch_errmsg("...\n");
3308# ifdef FEAT_SYN_HL
3309 if (re_extmatch_in != NULL)
3310 {
3311 int i;
3312
3313 mch_errmsg(_("External submatches:\n"));
3314 for (i = 0; i < NSUBEXP; i++)
3315 {
3316 mch_errmsg(" \"");
3317 if (re_extmatch_in->matches[i] != NULL)
3318 mch_errmsg((char *)re_extmatch_in->matches[i]);
3319 mch_errmsg("\"\n");
3320 }
3321 }
3322# endif
3323 }
3324#endif
3325 next = regnext(scan);
3326
3327 op = OP(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003328 // Check for character class with NL added.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003329 if (!rex.reg_line_lbr && WITH_NL(op) && REG_MULTI
Paul Ollis65745772022-06-05 16:55:54 +01003330 && *rex.input == NUL && rex.lnum <= rex.reg_maxline)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003331 {
3332 reg_nextline();
3333 }
3334 else if (rex.reg_line_lbr && WITH_NL(op) && *rex.input == '\n')
3335 {
3336 ADVANCE_REGINPUT();
3337 }
3338 else
3339 {
3340 if (WITH_NL(op))
3341 op -= ADD_NL;
3342 if (has_mbyte)
3343 c = (*mb_ptr2char)(rex.input);
3344 else
3345 c = *rex.input;
3346 switch (op)
3347 {
3348 case BOL:
3349 if (rex.input != rex.line)
3350 status = RA_NOMATCH;
3351 break;
3352
3353 case EOL:
3354 if (c != NUL)
3355 status = RA_NOMATCH;
3356 break;
3357
3358 case RE_BOF:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003359 // We're not at the beginning of the file when below the first
3360 // line where we started, not at the start of the line or we
3361 // didn't start at the first line of the buffer.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003362 if (rex.lnum != 0 || rex.input != rex.line
3363 || (REG_MULTI && rex.reg_firstlnum > 1))
3364 status = RA_NOMATCH;
3365 break;
3366
3367 case RE_EOF:
3368 if (rex.lnum != rex.reg_maxline || c != NUL)
3369 status = RA_NOMATCH;
3370 break;
3371
3372 case CURSOR:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003373 // Check if the buffer is in a window and compare the
3374 // rex.reg_win->w_cursor position to the match position.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003375 if (rex.reg_win == NULL
3376 || (rex.lnum + rex.reg_firstlnum
3377 != rex.reg_win->w_cursor.lnum)
3378 || ((colnr_T)(rex.input - rex.line)
3379 != rex.reg_win->w_cursor.col))
3380 status = RA_NOMATCH;
3381 break;
3382
3383 case RE_MARK:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003384 // Compare the mark position to the match position.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003385 {
3386 int mark = OPERAND(scan)[0];
3387 int cmp = OPERAND(scan)[1];
3388 pos_T *pos;
Bram Moolenaarb55986c2022-03-29 13:24:58 +01003389 size_t col = REG_MULTI ? rex.input - rex.line : 0;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003390
3391 pos = getmark_buf(rex.reg_buf, mark, FALSE);
Bram Moolenaarb55986c2022-03-29 13:24:58 +01003392
3393 // Line may have been freed, get it again.
3394 if (REG_MULTI)
3395 {
3396 rex.line = reg_getline(rex.lnum);
3397 rex.input = rex.line + col;
3398 }
3399
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003400 if (pos == NULL // mark doesn't exist
Bram Moolenaar872bee52021-05-24 22:56:15 +02003401 || pos->lnum <= 0) // mark isn't set in reg_buf
3402 {
3403 status = RA_NOMATCH;
3404 }
3405 else
3406 {
3407 colnr_T pos_col = pos->lnum == rex.lnum + rex.reg_firstlnum
3408 && pos->col == MAXCOL
3409 ? (colnr_T)STRLEN(reg_getline(
3410 pos->lnum - rex.reg_firstlnum))
3411 : pos->col;
3412
3413 if ((pos->lnum == rex.lnum + rex.reg_firstlnum
3414 ? (pos_col == (colnr_T)(rex.input - rex.line)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003415 ? (cmp == '<' || cmp == '>')
Bram Moolenaar872bee52021-05-24 22:56:15 +02003416 : (pos_col < (colnr_T)(rex.input - rex.line)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003417 ? cmp != '>'
3418 : cmp != '<'))
3419 : (pos->lnum < rex.lnum + rex.reg_firstlnum
3420 ? cmp != '>'
3421 : cmp != '<')))
3422 status = RA_NOMATCH;
Bram Moolenaar872bee52021-05-24 22:56:15 +02003423 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003424 }
3425 break;
3426
3427 case RE_VISUAL:
3428 if (!reg_match_visual())
3429 status = RA_NOMATCH;
3430 break;
3431
3432 case RE_LNUM:
3433 if (!REG_MULTI || !re_num_cmp((long_u)(rex.lnum + rex.reg_firstlnum),
3434 scan))
3435 status = RA_NOMATCH;
3436 break;
3437
3438 case RE_COL:
3439 if (!re_num_cmp((long_u)(rex.input - rex.line) + 1, scan))
3440 status = RA_NOMATCH;
3441 break;
3442
3443 case RE_VCOL:
Bram Moolenaar13ed4942022-08-19 13:59:25 +01003444 {
3445 win_T *wp = rex.reg_win == NULL ? curwin : rex.reg_win;
Bram Moolenaar753aead2022-09-08 12:17:06 +01003446 linenr_T lnum = REG_MULTI ? rex.reg_firstlnum + rex.lnum : 1;
3447 long_u vcol;
Bram Moolenaar13ed4942022-08-19 13:59:25 +01003448
Bram Moolenaar753aead2022-09-08 12:17:06 +01003449 if (REG_MULTI && (lnum <= 0
3450 || lnum > wp->w_buffer->b_ml.ml_line_count))
3451 lnum = 1;
3452 vcol = (long_u)win_linetabsize(wp, lnum, rex.line,
Bram Moolenaar13ed4942022-08-19 13:59:25 +01003453 (colnr_T)(rex.input - rex.line));
3454 if (!re_num_cmp(vcol + 1, scan))
3455 status = RA_NOMATCH;
3456 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003457 break;
3458
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003459 case BOW: // \<word; rex.input points to w
3460 if (c == NUL) // Can't match at end of line
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003461 status = RA_NOMATCH;
3462 else if (has_mbyte)
3463 {
3464 int this_class;
3465
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003466 // Get class of current and previous char (if it exists).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003467 this_class = mb_get_class_buf(rex.input, rex.reg_buf);
3468 if (this_class <= 1)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003469 status = RA_NOMATCH; // not on a word at all
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003470 else if (reg_prev_class() == this_class)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003471 status = RA_NOMATCH; // previous char is in same word
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003472 }
3473 else
3474 {
3475 if (!vim_iswordc_buf(c, rex.reg_buf) || (rex.input > rex.line
3476 && vim_iswordc_buf(rex.input[-1], rex.reg_buf)))
3477 status = RA_NOMATCH;
3478 }
3479 break;
3480
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003481 case EOW: // word\>; rex.input points after d
3482 if (rex.input == rex.line) // Can't match at start of line
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003483 status = RA_NOMATCH;
3484 else if (has_mbyte)
3485 {
3486 int this_class, prev_class;
3487
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003488 // Get class of current and previous char (if it exists).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003489 this_class = mb_get_class_buf(rex.input, rex.reg_buf);
3490 prev_class = reg_prev_class();
3491 if (this_class == prev_class
3492 || prev_class == 0 || prev_class == 1)
3493 status = RA_NOMATCH;
3494 }
3495 else
3496 {
3497 if (!vim_iswordc_buf(rex.input[-1], rex.reg_buf)
3498 || (rex.input[0] != NUL
3499 && vim_iswordc_buf(c, rex.reg_buf)))
3500 status = RA_NOMATCH;
3501 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003502 break; // Matched with EOW
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003503
3504 case ANY:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003505 // ANY does not match new lines.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003506 if (c == NUL)
3507 status = RA_NOMATCH;
3508 else
3509 ADVANCE_REGINPUT();
3510 break;
3511
3512 case IDENT:
3513 if (!vim_isIDc(c))
3514 status = RA_NOMATCH;
3515 else
3516 ADVANCE_REGINPUT();
3517 break;
3518
3519 case SIDENT:
3520 if (VIM_ISDIGIT(*rex.input) || !vim_isIDc(c))
3521 status = RA_NOMATCH;
3522 else
3523 ADVANCE_REGINPUT();
3524 break;
3525
3526 case KWORD:
3527 if (!vim_iswordp_buf(rex.input, rex.reg_buf))
3528 status = RA_NOMATCH;
3529 else
3530 ADVANCE_REGINPUT();
3531 break;
3532
3533 case SKWORD:
3534 if (VIM_ISDIGIT(*rex.input)
3535 || !vim_iswordp_buf(rex.input, rex.reg_buf))
3536 status = RA_NOMATCH;
3537 else
3538 ADVANCE_REGINPUT();
3539 break;
3540
3541 case FNAME:
3542 if (!vim_isfilec(c))
3543 status = RA_NOMATCH;
3544 else
3545 ADVANCE_REGINPUT();
3546 break;
3547
3548 case SFNAME:
3549 if (VIM_ISDIGIT(*rex.input) || !vim_isfilec(c))
3550 status = RA_NOMATCH;
3551 else
3552 ADVANCE_REGINPUT();
3553 break;
3554
3555 case PRINT:
3556 if (!vim_isprintc(PTR2CHAR(rex.input)))
3557 status = RA_NOMATCH;
3558 else
3559 ADVANCE_REGINPUT();
3560 break;
3561
3562 case SPRINT:
3563 if (VIM_ISDIGIT(*rex.input) || !vim_isprintc(PTR2CHAR(rex.input)))
3564 status = RA_NOMATCH;
3565 else
3566 ADVANCE_REGINPUT();
3567 break;
3568
3569 case WHITE:
3570 if (!VIM_ISWHITE(c))
3571 status = RA_NOMATCH;
3572 else
3573 ADVANCE_REGINPUT();
3574 break;
3575
3576 case NWHITE:
3577 if (c == NUL || VIM_ISWHITE(c))
3578 status = RA_NOMATCH;
3579 else
3580 ADVANCE_REGINPUT();
3581 break;
3582
3583 case DIGIT:
3584 if (!ri_digit(c))
3585 status = RA_NOMATCH;
3586 else
3587 ADVANCE_REGINPUT();
3588 break;
3589
3590 case NDIGIT:
3591 if (c == NUL || ri_digit(c))
3592 status = RA_NOMATCH;
3593 else
3594 ADVANCE_REGINPUT();
3595 break;
3596
3597 case HEX:
3598 if (!ri_hex(c))
3599 status = RA_NOMATCH;
3600 else
3601 ADVANCE_REGINPUT();
3602 break;
3603
3604 case NHEX:
3605 if (c == NUL || ri_hex(c))
3606 status = RA_NOMATCH;
3607 else
3608 ADVANCE_REGINPUT();
3609 break;
3610
3611 case OCTAL:
3612 if (!ri_octal(c))
3613 status = RA_NOMATCH;
3614 else
3615 ADVANCE_REGINPUT();
3616 break;
3617
3618 case NOCTAL:
3619 if (c == NUL || ri_octal(c))
3620 status = RA_NOMATCH;
3621 else
3622 ADVANCE_REGINPUT();
3623 break;
3624
3625 case WORD:
3626 if (!ri_word(c))
3627 status = RA_NOMATCH;
3628 else
3629 ADVANCE_REGINPUT();
3630 break;
3631
3632 case NWORD:
3633 if (c == NUL || ri_word(c))
3634 status = RA_NOMATCH;
3635 else
3636 ADVANCE_REGINPUT();
3637 break;
3638
3639 case HEAD:
3640 if (!ri_head(c))
3641 status = RA_NOMATCH;
3642 else
3643 ADVANCE_REGINPUT();
3644 break;
3645
3646 case NHEAD:
3647 if (c == NUL || ri_head(c))
3648 status = RA_NOMATCH;
3649 else
3650 ADVANCE_REGINPUT();
3651 break;
3652
3653 case ALPHA:
3654 if (!ri_alpha(c))
3655 status = RA_NOMATCH;
3656 else
3657 ADVANCE_REGINPUT();
3658 break;
3659
3660 case NALPHA:
3661 if (c == NUL || ri_alpha(c))
3662 status = RA_NOMATCH;
3663 else
3664 ADVANCE_REGINPUT();
3665 break;
3666
3667 case LOWER:
3668 if (!ri_lower(c))
3669 status = RA_NOMATCH;
3670 else
3671 ADVANCE_REGINPUT();
3672 break;
3673
3674 case NLOWER:
3675 if (c == NUL || ri_lower(c))
3676 status = RA_NOMATCH;
3677 else
3678 ADVANCE_REGINPUT();
3679 break;
3680
3681 case UPPER:
3682 if (!ri_upper(c))
3683 status = RA_NOMATCH;
3684 else
3685 ADVANCE_REGINPUT();
3686 break;
3687
3688 case NUPPER:
3689 if (c == NUL || ri_upper(c))
3690 status = RA_NOMATCH;
3691 else
3692 ADVANCE_REGINPUT();
3693 break;
3694
3695 case EXACTLY:
3696 {
3697 int len;
3698 char_u *opnd;
3699
3700 opnd = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003701 // Inline the first byte, for speed.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003702 if (*opnd != *rex.input
3703 && (!rex.reg_ic
3704 || (!enc_utf8
3705 && MB_TOLOWER(*opnd) != MB_TOLOWER(*rex.input))))
3706 status = RA_NOMATCH;
3707 else if (*opnd == NUL)
3708 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003709 // match empty string always works; happens when "~" is
3710 // empty.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003711 }
3712 else
3713 {
3714 if (opnd[1] == NUL && !(enc_utf8 && rex.reg_ic))
3715 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003716 len = 1; // matched a single byte above
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003717 }
3718 else
3719 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003720 // Need to match first byte again for multi-byte.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003721 len = (int)STRLEN(opnd);
3722 if (cstrncmp(opnd, rex.input, &len) != 0)
3723 status = RA_NOMATCH;
3724 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003725 // Check for following composing character, unless %C
3726 // follows (skips over all composing chars).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003727 if (status != RA_NOMATCH
3728 && enc_utf8
3729 && UTF_COMPOSINGLIKE(rex.input, rex.input + len)
3730 && !rex.reg_icombine
3731 && OP(next) != RE_COMPOSING)
3732 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003733 // raaron: This code makes a composing character get
3734 // ignored, which is the correct behavior (sometimes)
3735 // for voweled Hebrew texts.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003736 status = RA_NOMATCH;
3737 }
3738 if (status != RA_NOMATCH)
3739 rex.input += len;
3740 }
3741 }
3742 break;
3743
3744 case ANYOF:
3745 case ANYBUT:
Christian Brabandtd2cc51f2024-01-04 22:54:08 +01003746 {
3747 char_u *q = OPERAND(scan);
3748
3749 if (c == NUL)
3750 status = RA_NOMATCH;
3751 else if ((cstrchr(q, c) == NULL) == (op == ANYOF))
3752 status = RA_NOMATCH;
3753 else
3754 {
3755 // Check following combining characters
3756 int len = 0;
3757 int i;
3758
3759 if (enc_utf8)
3760 len = utfc_ptr2len(q) - utf_ptr2len(q);
3761
3762 MB_CPTR_ADV(rex.input);
3763 MB_CPTR_ADV(q);
3764
3765 if (!enc_utf8 || len == 0)
3766 break;
3767
3768 for (i = 0; i < len; ++i)
3769 if (q[i] != rex.input[i])
3770 {
3771 status = RA_NOMATCH;
3772 break;
3773 }
3774 rex.input += len;
3775 }
3776 break;
3777 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003778
3779 case MULTIBYTECODE:
3780 if (has_mbyte)
3781 {
3782 int i, len;
3783 char_u *opnd;
3784 int opndc = 0, inpc;
3785
3786 opnd = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003787 // Safety check (just in case 'encoding' was changed since
3788 // compiling the program).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003789 if ((len = (*mb_ptr2len)(opnd)) < 2)
3790 {
3791 status = RA_NOMATCH;
3792 break;
3793 }
3794 if (enc_utf8)
3795 opndc = utf_ptr2char(opnd);
3796 if (enc_utf8 && utf_iscomposing(opndc))
3797 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003798 // When only a composing char is given match at any
3799 // position where that composing char appears.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003800 status = RA_NOMATCH;
3801 for (i = 0; rex.input[i] != NUL;
3802 i += utf_ptr2len(rex.input + i))
3803 {
3804 inpc = utf_ptr2char(rex.input + i);
3805 if (!utf_iscomposing(inpc))
3806 {
3807 if (i > 0)
3808 break;
3809 }
3810 else if (opndc == inpc)
3811 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003812 // Include all following composing chars.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003813 len = i + utfc_ptr2len(rex.input + i);
3814 status = RA_MATCH;
3815 break;
3816 }
3817 }
3818 }
3819 else
3820 for (i = 0; i < len; ++i)
3821 if (opnd[i] != rex.input[i])
3822 {
3823 status = RA_NOMATCH;
3824 break;
3825 }
3826 rex.input += len;
3827 }
3828 else
3829 status = RA_NOMATCH;
3830 break;
3831 case RE_COMPOSING:
3832 if (enc_utf8)
3833 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003834 // Skip composing characters.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003835 while (utf_iscomposing(utf_ptr2char(rex.input)))
3836 MB_CPTR_ADV(rex.input);
3837 }
3838 break;
3839
3840 case NOTHING:
3841 break;
3842
3843 case BACK:
3844 {
3845 int i;
3846 backpos_T *bp;
3847
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003848 // When we run into BACK we need to check if we don't keep
3849 // looping without matching any input. The second and later
3850 // times a BACK is encountered it fails if the input is still
3851 // at the same position as the previous time.
3852 // The positions are stored in "backpos" and found by the
3853 // current value of "scan", the position in the RE program.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003854 bp = (backpos_T *)backpos.ga_data;
3855 for (i = 0; i < backpos.ga_len; ++i)
3856 if (bp[i].bp_scan == scan)
3857 break;
3858 if (i == backpos.ga_len)
3859 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003860 // First time at this BACK, make room to store the pos.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003861 if (ga_grow(&backpos, 1) == FAIL)
3862 status = RA_FAIL;
3863 else
3864 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003865 // get "ga_data" again, it may have changed
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003866 bp = (backpos_T *)backpos.ga_data;
3867 bp[i].bp_scan = scan;
3868 ++backpos.ga_len;
3869 }
3870 }
3871 else if (reg_save_equal(&bp[i].bp_pos))
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003872 // Still at same position as last time, fail.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003873 status = RA_NOMATCH;
3874
3875 if (status != RA_FAIL && status != RA_NOMATCH)
3876 reg_save(&bp[i].bp_pos, &backpos);
3877 }
3878 break;
3879
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003880 case MOPEN + 0: // Match start: \zs
3881 case MOPEN + 1: // \(
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003882 case MOPEN + 2:
3883 case MOPEN + 3:
3884 case MOPEN + 4:
3885 case MOPEN + 5:
3886 case MOPEN + 6:
3887 case MOPEN + 7:
3888 case MOPEN + 8:
3889 case MOPEN + 9:
3890 {
3891 no = op - MOPEN;
3892 cleanup_subexpr();
3893 rp = regstack_push(RS_MOPEN, scan);
3894 if (rp == NULL)
3895 status = RA_FAIL;
3896 else
3897 {
3898 rp->rs_no = no;
3899 save_se(&rp->rs_un.sesave, &rex.reg_startpos[no],
3900 &rex.reg_startp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003901 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003902 }
3903 }
3904 break;
3905
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003906 case NOPEN: // \%(
3907 case NCLOSE: // \) after \%(
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003908 if (regstack_push(RS_NOPEN, scan) == NULL)
3909 status = RA_FAIL;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003910 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003911 break;
3912
3913#ifdef FEAT_SYN_HL
3914 case ZOPEN + 1:
3915 case ZOPEN + 2:
3916 case ZOPEN + 3:
3917 case ZOPEN + 4:
3918 case ZOPEN + 5:
3919 case ZOPEN + 6:
3920 case ZOPEN + 7:
3921 case ZOPEN + 8:
3922 case ZOPEN + 9:
3923 {
3924 no = op - ZOPEN;
3925 cleanup_zsubexpr();
3926 rp = regstack_push(RS_ZOPEN, scan);
3927 if (rp == NULL)
3928 status = RA_FAIL;
3929 else
3930 {
3931 rp->rs_no = no;
3932 save_se(&rp->rs_un.sesave, &reg_startzpos[no],
3933 &reg_startzp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003934 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003935 }
3936 }
3937 break;
3938#endif
3939
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003940 case MCLOSE + 0: // Match end: \ze
3941 case MCLOSE + 1: // \)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003942 case MCLOSE + 2:
3943 case MCLOSE + 3:
3944 case MCLOSE + 4:
3945 case MCLOSE + 5:
3946 case MCLOSE + 6:
3947 case MCLOSE + 7:
3948 case MCLOSE + 8:
3949 case MCLOSE + 9:
3950 {
3951 no = op - MCLOSE;
3952 cleanup_subexpr();
3953 rp = regstack_push(RS_MCLOSE, scan);
3954 if (rp == NULL)
3955 status = RA_FAIL;
3956 else
3957 {
3958 rp->rs_no = no;
3959 save_se(&rp->rs_un.sesave, &rex.reg_endpos[no],
3960 &rex.reg_endp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003961 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003962 }
3963 }
3964 break;
3965
3966#ifdef FEAT_SYN_HL
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003967 case ZCLOSE + 1: // \) after \z(
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003968 case ZCLOSE + 2:
3969 case ZCLOSE + 3:
3970 case ZCLOSE + 4:
3971 case ZCLOSE + 5:
3972 case ZCLOSE + 6:
3973 case ZCLOSE + 7:
3974 case ZCLOSE + 8:
3975 case ZCLOSE + 9:
3976 {
3977 no = op - ZCLOSE;
3978 cleanup_zsubexpr();
3979 rp = regstack_push(RS_ZCLOSE, scan);
3980 if (rp == NULL)
3981 status = RA_FAIL;
3982 else
3983 {
3984 rp->rs_no = no;
3985 save_se(&rp->rs_un.sesave, &reg_endzpos[no],
3986 &reg_endzp[no]);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02003987 // We simply continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02003988 }
3989 }
3990 break;
3991#endif
3992
3993 case BACKREF + 1:
3994 case BACKREF + 2:
3995 case BACKREF + 3:
3996 case BACKREF + 4:
3997 case BACKREF + 5:
3998 case BACKREF + 6:
3999 case BACKREF + 7:
4000 case BACKREF + 8:
4001 case BACKREF + 9:
4002 {
4003 int len;
4004
4005 no = op - BACKREF;
4006 cleanup_subexpr();
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004007 if (!REG_MULTI) // Single-line regexp
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004008 {
4009 if (rex.reg_startp[no] == NULL || rex.reg_endp[no] == NULL)
4010 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004011 // Backref was not set: Match an empty string.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004012 len = 0;
4013 }
4014 else
4015 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004016 // Compare current input with back-ref in the same
4017 // line.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004018 len = (int)(rex.reg_endp[no] - rex.reg_startp[no]);
4019 if (cstrncmp(rex.reg_startp[no], rex.input, &len) != 0)
4020 status = RA_NOMATCH;
4021 }
4022 }
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004023 else // Multi-line regexp
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004024 {
4025 if (rex.reg_startpos[no].lnum < 0
4026 || rex.reg_endpos[no].lnum < 0)
4027 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004028 // Backref was not set: Match an empty string.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004029 len = 0;
4030 }
4031 else
4032 {
4033 if (rex.reg_startpos[no].lnum == rex.lnum
4034 && rex.reg_endpos[no].lnum == rex.lnum)
4035 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004036 // Compare back-ref within the current line.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004037 len = rex.reg_endpos[no].col
4038 - rex.reg_startpos[no].col;
4039 if (cstrncmp(rex.line + rex.reg_startpos[no].col,
4040 rex.input, &len) != 0)
4041 status = RA_NOMATCH;
4042 }
4043 else
4044 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004045 // Messy situation: Need to compare between two
4046 // lines.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004047 int r = match_with_backref(
4048 rex.reg_startpos[no].lnum,
4049 rex.reg_startpos[no].col,
4050 rex.reg_endpos[no].lnum,
4051 rex.reg_endpos[no].col,
4052 &len);
4053
4054 if (r != RA_MATCH)
4055 status = r;
4056 }
4057 }
4058 }
4059
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004060 // Matched the backref, skip over it.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004061 rex.input += len;
4062 }
4063 break;
4064
4065#ifdef FEAT_SYN_HL
4066 case ZREF + 1:
4067 case ZREF + 2:
4068 case ZREF + 3:
4069 case ZREF + 4:
4070 case ZREF + 5:
4071 case ZREF + 6:
4072 case ZREF + 7:
4073 case ZREF + 8:
4074 case ZREF + 9:
4075 {
4076 int len;
4077
4078 cleanup_zsubexpr();
4079 no = op - ZREF;
4080 if (re_extmatch_in != NULL
4081 && re_extmatch_in->matches[no] != NULL)
4082 {
4083 len = (int)STRLEN(re_extmatch_in->matches[no]);
4084 if (cstrncmp(re_extmatch_in->matches[no],
4085 rex.input, &len) != 0)
4086 status = RA_NOMATCH;
4087 else
4088 rex.input += len;
4089 }
4090 else
4091 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004092 // Backref was not set: Match an empty string.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004093 }
4094 }
4095 break;
4096#endif
4097
4098 case BRANCH:
4099 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004100 if (OP(next) != BRANCH) // No choice.
4101 next = OPERAND(scan); // Avoid recursion.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004102 else
4103 {
4104 rp = regstack_push(RS_BRANCH, scan);
4105 if (rp == NULL)
4106 status = RA_FAIL;
4107 else
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004108 status = RA_BREAK; // rest is below
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004109 }
4110 }
4111 break;
4112
4113 case BRACE_LIMITS:
4114 {
4115 if (OP(next) == BRACE_SIMPLE)
4116 {
4117 bl_minval = OPERAND_MIN(scan);
4118 bl_maxval = OPERAND_MAX(scan);
4119 }
4120 else if (OP(next) >= BRACE_COMPLEX
4121 && OP(next) < BRACE_COMPLEX + 10)
4122 {
4123 no = OP(next) - BRACE_COMPLEX;
4124 brace_min[no] = OPERAND_MIN(scan);
4125 brace_max[no] = OPERAND_MAX(scan);
4126 brace_count[no] = 0;
4127 }
4128 else
4129 {
4130 internal_error("BRACE_LIMITS");
4131 status = RA_FAIL;
4132 }
4133 }
4134 break;
4135
4136 case BRACE_COMPLEX + 0:
4137 case BRACE_COMPLEX + 1:
4138 case BRACE_COMPLEX + 2:
4139 case BRACE_COMPLEX + 3:
4140 case BRACE_COMPLEX + 4:
4141 case BRACE_COMPLEX + 5:
4142 case BRACE_COMPLEX + 6:
4143 case BRACE_COMPLEX + 7:
4144 case BRACE_COMPLEX + 8:
4145 case BRACE_COMPLEX + 9:
4146 {
4147 no = op - BRACE_COMPLEX;
4148 ++brace_count[no];
4149
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004150 // If not matched enough times yet, try one more
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004151 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
4152 ? brace_min[no] : brace_max[no]))
4153 {
4154 rp = regstack_push(RS_BRCPLX_MORE, scan);
4155 if (rp == NULL)
4156 status = RA_FAIL;
4157 else
4158 {
4159 rp->rs_no = no;
4160 reg_save(&rp->rs_un.regsave, &backpos);
4161 next = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004162 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004163 }
4164 break;
4165 }
4166
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004167 // If matched enough times, may try matching some more
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004168 if (brace_min[no] <= brace_max[no])
4169 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004170 // Range is the normal way around, use longest match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004171 if (brace_count[no] <= brace_max[no])
4172 {
4173 rp = regstack_push(RS_BRCPLX_LONG, scan);
4174 if (rp == NULL)
4175 status = RA_FAIL;
4176 else
4177 {
4178 rp->rs_no = no;
4179 reg_save(&rp->rs_un.regsave, &backpos);
4180 next = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004181 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004182 }
4183 }
4184 }
4185 else
4186 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004187 // Range is backwards, use shortest match first
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004188 if (brace_count[no] <= brace_min[no])
4189 {
4190 rp = regstack_push(RS_BRCPLX_SHORT, scan);
4191 if (rp == NULL)
4192 status = RA_FAIL;
4193 else
4194 {
4195 reg_save(&rp->rs_un.regsave, &backpos);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004196 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004197 }
4198 }
4199 }
4200 }
4201 break;
4202
4203 case BRACE_SIMPLE:
4204 case STAR:
4205 case PLUS:
4206 {
4207 regstar_T rst;
4208
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004209 // Lookahead to avoid useless match attempts when we know
4210 // what character comes next.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004211 if (OP(next) == EXACTLY)
4212 {
4213 rst.nextb = *OPERAND(next);
4214 if (rex.reg_ic)
4215 {
4216 if (MB_ISUPPER(rst.nextb))
4217 rst.nextb_ic = MB_TOLOWER(rst.nextb);
4218 else
4219 rst.nextb_ic = MB_TOUPPER(rst.nextb);
4220 }
4221 else
4222 rst.nextb_ic = rst.nextb;
4223 }
4224 else
4225 {
4226 rst.nextb = NUL;
4227 rst.nextb_ic = NUL;
4228 }
4229 if (op != BRACE_SIMPLE)
4230 {
4231 rst.minval = (op == STAR) ? 0 : 1;
4232 rst.maxval = MAX_LIMIT;
4233 }
4234 else
4235 {
4236 rst.minval = bl_minval;
4237 rst.maxval = bl_maxval;
4238 }
4239
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004240 // When maxval > minval, try matching as much as possible, up
4241 // to maxval. When maxval < minval, try matching at least the
4242 // minimal number (since the range is backwards, that's also
4243 // maxval!).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004244 rst.count = regrepeat(OPERAND(scan), rst.maxval);
4245 if (got_int)
4246 {
4247 status = RA_FAIL;
4248 break;
4249 }
4250 if (rst.minval <= rst.maxval
4251 ? rst.count >= rst.minval : rst.count >= rst.maxval)
4252 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004253 // It could match. Prepare for trying to match what
4254 // follows. The code is below. Parameters are stored in
4255 // a regstar_T on the regstack.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004256 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
4257 {
Bram Moolenaar74409f62022-01-01 15:58:22 +00004258 emsg(_(e_pattern_uses_more_memory_than_maxmempattern));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004259 status = RA_FAIL;
4260 }
4261 else if (ga_grow(&regstack, sizeof(regstar_T)) == FAIL)
4262 status = RA_FAIL;
4263 else
4264 {
4265 regstack.ga_len += sizeof(regstar_T);
4266 rp = regstack_push(rst.minval <= rst.maxval
4267 ? RS_STAR_LONG : RS_STAR_SHORT, scan);
4268 if (rp == NULL)
4269 status = RA_FAIL;
4270 else
4271 {
4272 *(((regstar_T *)rp) - 1) = rst;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004273 status = RA_BREAK; // skip the restore bits
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004274 }
4275 }
4276 }
4277 else
4278 status = RA_NOMATCH;
4279
4280 }
4281 break;
4282
4283 case NOMATCH:
4284 case MATCH:
4285 case SUBPAT:
4286 rp = regstack_push(RS_NOMATCH, scan);
4287 if (rp == NULL)
4288 status = RA_FAIL;
4289 else
4290 {
4291 rp->rs_no = op;
4292 reg_save(&rp->rs_un.regsave, &backpos);
4293 next = OPERAND(scan);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004294 // We continue and handle the result when done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004295 }
4296 break;
4297
4298 case BEHIND:
4299 case NOBEHIND:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004300 // Need a bit of room to store extra positions.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004301 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
4302 {
Bram Moolenaar74409f62022-01-01 15:58:22 +00004303 emsg(_(e_pattern_uses_more_memory_than_maxmempattern));
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004304 status = RA_FAIL;
4305 }
4306 else if (ga_grow(&regstack, sizeof(regbehind_T)) == FAIL)
4307 status = RA_FAIL;
4308 else
4309 {
4310 regstack.ga_len += sizeof(regbehind_T);
4311 rp = regstack_push(RS_BEHIND1, scan);
4312 if (rp == NULL)
4313 status = RA_FAIL;
4314 else
4315 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004316 // Need to save the subexpr to be able to restore them
4317 // when there is a match but we don't use it.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004318 save_subexpr(((regbehind_T *)rp) - 1);
4319
4320 rp->rs_no = op;
4321 reg_save(&rp->rs_un.regsave, &backpos);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004322 // First try if what follows matches. If it does then we
4323 // check the behind match by looping.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004324 }
4325 }
4326 break;
4327
4328 case BHPOS:
4329 if (REG_MULTI)
4330 {
4331 if (behind_pos.rs_u.pos.col != (colnr_T)(rex.input - rex.line)
4332 || behind_pos.rs_u.pos.lnum != rex.lnum)
4333 status = RA_NOMATCH;
4334 }
4335 else if (behind_pos.rs_u.ptr != rex.input)
4336 status = RA_NOMATCH;
4337 break;
4338
4339 case NEWL:
4340 if ((c != NUL || !REG_MULTI || rex.lnum > rex.reg_maxline
4341 || rex.reg_line_lbr)
4342 && (c != '\n' || !rex.reg_line_lbr))
4343 status = RA_NOMATCH;
4344 else if (rex.reg_line_lbr)
4345 ADVANCE_REGINPUT();
4346 else
4347 reg_nextline();
4348 break;
4349
4350 case END:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004351 status = RA_MATCH; // Success!
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004352 break;
4353
4354 default:
RestorerZ68ebcee2023-05-31 17:12:14 +01004355 iemsg(e_corrupted_regexp_program);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004356#ifdef DEBUG
4357 printf("Illegal op code %d\n", op);
4358#endif
4359 status = RA_FAIL;
4360 break;
4361 }
4362 }
4363
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004364 // If we can't continue sequentially, break the inner loop.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004365 if (status != RA_CONT)
4366 break;
4367
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004368 // Continue in inner loop, advance to next item.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004369 scan = next;
4370
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004371 } // end of inner loop
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004372
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004373 // If there is something on the regstack execute the code for the state.
4374 // If the state is popped then loop and use the older state.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004375 while (regstack.ga_len > 0 && status != RA_FAIL)
4376 {
4377 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
4378 switch (rp->rs_state)
4379 {
4380 case RS_NOPEN:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004381 // Result is passed on as-is, simply pop the state.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004382 regstack_pop(&scan);
4383 break;
4384
4385 case RS_MOPEN:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004386 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004387 if (status == RA_NOMATCH)
4388 restore_se(&rp->rs_un.sesave, &rex.reg_startpos[rp->rs_no],
4389 &rex.reg_startp[rp->rs_no]);
4390 regstack_pop(&scan);
4391 break;
4392
4393#ifdef FEAT_SYN_HL
4394 case RS_ZOPEN:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004395 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004396 if (status == RA_NOMATCH)
4397 restore_se(&rp->rs_un.sesave, &reg_startzpos[rp->rs_no],
4398 &reg_startzp[rp->rs_no]);
4399 regstack_pop(&scan);
4400 break;
4401#endif
4402
4403 case RS_MCLOSE:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004404 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004405 if (status == RA_NOMATCH)
4406 restore_se(&rp->rs_un.sesave, &rex.reg_endpos[rp->rs_no],
4407 &rex.reg_endp[rp->rs_no]);
4408 regstack_pop(&scan);
4409 break;
4410
4411#ifdef FEAT_SYN_HL
4412 case RS_ZCLOSE:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004413 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004414 if (status == RA_NOMATCH)
4415 restore_se(&rp->rs_un.sesave, &reg_endzpos[rp->rs_no],
4416 &reg_endzp[rp->rs_no]);
4417 regstack_pop(&scan);
4418 break;
4419#endif
4420
4421 case RS_BRANCH:
4422 if (status == RA_MATCH)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004423 // this branch matched, use it
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004424 regstack_pop(&scan);
4425 else
4426 {
4427 if (status != RA_BREAK)
4428 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004429 // After a non-matching branch: try next one.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004430 reg_restore(&rp->rs_un.regsave, &backpos);
4431 scan = rp->rs_scan;
4432 }
4433 if (scan == NULL || OP(scan) != BRANCH)
4434 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004435 // no more branches, didn't find a match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004436 status = RA_NOMATCH;
4437 regstack_pop(&scan);
4438 }
4439 else
4440 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004441 // Prepare to try a branch.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004442 rp->rs_scan = regnext(scan);
4443 reg_save(&rp->rs_un.regsave, &backpos);
4444 scan = OPERAND(scan);
4445 }
4446 }
4447 break;
4448
4449 case RS_BRCPLX_MORE:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004450 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004451 if (status == RA_NOMATCH)
4452 {
4453 reg_restore(&rp->rs_un.regsave, &backpos);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004454 --brace_count[rp->rs_no]; // decrement match count
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004455 }
4456 regstack_pop(&scan);
4457 break;
4458
4459 case RS_BRCPLX_LONG:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004460 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004461 if (status == RA_NOMATCH)
4462 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004463 // There was no match, but we did find enough matches.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004464 reg_restore(&rp->rs_un.regsave, &backpos);
4465 --brace_count[rp->rs_no];
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004466 // continue with the items after "\{}"
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004467 status = RA_CONT;
4468 }
4469 regstack_pop(&scan);
4470 if (status == RA_CONT)
4471 scan = regnext(scan);
4472 break;
4473
4474 case RS_BRCPLX_SHORT:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004475 // Pop the state. Restore pointers when there is no match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004476 if (status == RA_NOMATCH)
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004477 // There was no match, try to match one more item.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004478 reg_restore(&rp->rs_un.regsave, &backpos);
4479 regstack_pop(&scan);
4480 if (status == RA_NOMATCH)
4481 {
4482 scan = OPERAND(scan);
4483 status = RA_CONT;
4484 }
4485 break;
4486
4487 case RS_NOMATCH:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004488 // Pop the state. If the operand matches for NOMATCH or
4489 // doesn't match for MATCH/SUBPAT, we fail. Otherwise backup,
4490 // except for SUBPAT, and continue with the next item.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004491 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH))
4492 status = RA_NOMATCH;
4493 else
4494 {
4495 status = RA_CONT;
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004496 if (rp->rs_no != SUBPAT) // zero-width
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004497 reg_restore(&rp->rs_un.regsave, &backpos);
4498 }
4499 regstack_pop(&scan);
4500 if (status == RA_CONT)
4501 scan = regnext(scan);
4502 break;
4503
4504 case RS_BEHIND1:
4505 if (status == RA_NOMATCH)
4506 {
4507 regstack_pop(&scan);
4508 regstack.ga_len -= sizeof(regbehind_T);
4509 }
4510 else
4511 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004512 // The stuff after BEHIND/NOBEHIND matches. Now try if
4513 // the behind part does (not) match before the current
4514 // position in the input. This must be done at every
4515 // position in the input and checking if the match ends at
4516 // the current position.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004517
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004518 // save the position after the found match for next
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004519 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos);
4520
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004521 // Start looking for a match with operand at the current
4522 // position. Go back one character until we find the
4523 // result, hitting the start of the line or the previous
4524 // line (for multi-line matching).
4525 // Set behind_pos to where the match should end, BHPOS
4526 // will match it. Save the current value.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004527 (((regbehind_T *)rp) - 1)->save_behind = behind_pos;
4528 behind_pos = rp->rs_un.regsave;
4529
4530 rp->rs_state = RS_BEHIND2;
4531
4532 reg_restore(&rp->rs_un.regsave, &backpos);
4533 scan = OPERAND(rp->rs_scan) + 4;
4534 }
4535 break;
4536
4537 case RS_BEHIND2:
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004538 // Looping for BEHIND / NOBEHIND match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004539 if (status == RA_MATCH && reg_save_equal(&behind_pos))
4540 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004541 // found a match that ends where "next" started
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004542 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
4543 if (rp->rs_no == BEHIND)
4544 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
4545 &backpos);
4546 else
4547 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004548 // But we didn't want a match. Need to restore the
4549 // subexpr, because what follows matched, so they have
4550 // been set.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004551 status = RA_NOMATCH;
4552 restore_subexpr(((regbehind_T *)rp) - 1);
4553 }
4554 regstack_pop(&scan);
4555 regstack.ga_len -= sizeof(regbehind_T);
4556 }
4557 else
4558 {
4559 long limit;
4560
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004561 // No match or a match that doesn't end where we want it: Go
4562 // back one character. May go to previous line once.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004563 no = OK;
4564 limit = OPERAND_MIN(rp->rs_scan);
4565 if (REG_MULTI)
4566 {
4567 if (limit > 0
4568 && ((rp->rs_un.regsave.rs_u.pos.lnum
4569 < behind_pos.rs_u.pos.lnum
4570 ? (colnr_T)STRLEN(rex.line)
4571 : behind_pos.rs_u.pos.col)
4572 - rp->rs_un.regsave.rs_u.pos.col >= limit))
4573 no = FAIL;
4574 else if (rp->rs_un.regsave.rs_u.pos.col == 0)
4575 {
4576 if (rp->rs_un.regsave.rs_u.pos.lnum
4577 < behind_pos.rs_u.pos.lnum
4578 || reg_getline(
4579 --rp->rs_un.regsave.rs_u.pos.lnum)
4580 == NULL)
4581 no = FAIL;
4582 else
4583 {
4584 reg_restore(&rp->rs_un.regsave, &backpos);
4585 rp->rs_un.regsave.rs_u.pos.col =
4586 (colnr_T)STRLEN(rex.line);
4587 }
4588 }
4589 else
4590 {
4591 if (has_mbyte)
4592 {
4593 char_u *line =
4594 reg_getline(rp->rs_un.regsave.rs_u.pos.lnum);
4595
4596 rp->rs_un.regsave.rs_u.pos.col -=
4597 (*mb_head_off)(line, line
4598 + rp->rs_un.regsave.rs_u.pos.col - 1) + 1;
4599 }
4600 else
4601 --rp->rs_un.regsave.rs_u.pos.col;
4602 }
4603 }
4604 else
4605 {
4606 if (rp->rs_un.regsave.rs_u.ptr == rex.line)
4607 no = FAIL;
4608 else
4609 {
4610 MB_PTR_BACK(rex.line, rp->rs_un.regsave.rs_u.ptr);
4611 if (limit > 0 && (long)(behind_pos.rs_u.ptr
4612 - rp->rs_un.regsave.rs_u.ptr) > limit)
4613 no = FAIL;
4614 }
4615 }
4616 if (no == OK)
4617 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004618 // Advanced, prepare for finding match again.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004619 reg_restore(&rp->rs_un.regsave, &backpos);
4620 scan = OPERAND(rp->rs_scan) + 4;
4621 if (status == RA_MATCH)
4622 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004623 // We did match, so subexpr may have been changed,
4624 // need to restore them for the next try.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004625 status = RA_NOMATCH;
4626 restore_subexpr(((regbehind_T *)rp) - 1);
4627 }
4628 }
4629 else
4630 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004631 // Can't advance. For NOBEHIND that's a match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004632 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
4633 if (rp->rs_no == NOBEHIND)
4634 {
4635 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
4636 &backpos);
4637 status = RA_MATCH;
4638 }
4639 else
4640 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004641 // We do want a proper match. Need to restore the
4642 // subexpr if we had a match, because they may have
4643 // been set.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004644 if (status == RA_MATCH)
4645 {
4646 status = RA_NOMATCH;
4647 restore_subexpr(((regbehind_T *)rp) - 1);
4648 }
4649 }
4650 regstack_pop(&scan);
4651 regstack.ga_len -= sizeof(regbehind_T);
4652 }
4653 }
4654 break;
4655
4656 case RS_STAR_LONG:
4657 case RS_STAR_SHORT:
4658 {
4659 regstar_T *rst = ((regstar_T *)rp) - 1;
4660
4661 if (status == RA_MATCH)
4662 {
4663 regstack_pop(&scan);
4664 regstack.ga_len -= sizeof(regstar_T);
4665 break;
4666 }
4667
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004668 // Tried once already, restore input pointers.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004669 if (status != RA_BREAK)
4670 reg_restore(&rp->rs_un.regsave, &backpos);
4671
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004672 // Repeat until we found a position where it could match.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004673 for (;;)
4674 {
4675 if (status != RA_BREAK)
4676 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004677 // Tried first position already, advance.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004678 if (rp->rs_state == RS_STAR_LONG)
4679 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004680 // Trying for longest match, but couldn't or
4681 // didn't match -- back up one char.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004682 if (--rst->count < rst->minval)
4683 break;
4684 if (rex.input == rex.line)
4685 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004686 // backup to last char of previous line
Bram Moolenaar6456fae2022-02-22 13:37:31 +00004687 if (rex.lnum == 0)
4688 {
4689 status = RA_NOMATCH;
4690 break;
4691 }
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004692 --rex.lnum;
4693 rex.line = reg_getline(rex.lnum);
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004694 // Just in case regrepeat() didn't count
4695 // right.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004696 if (rex.line == NULL)
4697 break;
4698 rex.input = rex.line + STRLEN(rex.line);
4699 fast_breakcheck();
4700 }
4701 else
4702 MB_PTR_BACK(rex.line, rex.input);
4703 }
4704 else
4705 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004706 // Range is backwards, use shortest match first.
4707 // Careful: maxval and minval are exchanged!
4708 // Couldn't or didn't match: try advancing one
4709 // char.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004710 if (rst->count == rst->minval
4711 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0)
4712 break;
4713 ++rst->count;
4714 }
4715 if (got_int)
4716 break;
4717 }
4718 else
4719 status = RA_NOMATCH;
4720
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004721 // If it could match, try it.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004722 if (rst->nextb == NUL || *rex.input == rst->nextb
4723 || *rex.input == rst->nextb_ic)
4724 {
4725 reg_save(&rp->rs_un.regsave, &backpos);
4726 scan = regnext(rp->rs_scan);
4727 status = RA_CONT;
4728 break;
4729 }
4730 }
4731 if (status != RA_CONT)
4732 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004733 // Failed.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004734 regstack_pop(&scan);
4735 regstack.ga_len -= sizeof(regstar_T);
4736 status = RA_NOMATCH;
4737 }
4738 }
4739 break;
4740 }
4741
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004742 // If we want to continue the inner loop or didn't pop a state
4743 // continue matching loop
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004744 if (status == RA_CONT || rp == (regitem_T *)
4745 ((char *)regstack.ga_data + regstack.ga_len) - 1)
4746 break;
Bram Moolenaar616592e2022-06-17 15:17:10 +01004747
4748#ifdef FEAT_RELTIME
4749 if (bt_did_time_out(timed_out))
4750 {
4751 status = RA_FAIL;
4752 break;
4753 }
4754#endif
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004755 }
4756
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004757 // May need to continue with the inner loop, starting at "scan".
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004758 if (status == RA_CONT)
4759 continue;
4760
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004761 // If the regstack is empty or something failed we are done.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004762 if (regstack.ga_len == 0 || status == RA_FAIL)
4763 {
4764 if (scan == NULL)
4765 {
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004766 // We get here only if there's trouble -- normally "case END" is
4767 // the terminating point.
RestorerZ68ebcee2023-05-31 17:12:14 +01004768 iemsg(e_corrupted_regexp_program);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004769#ifdef DEBUG
4770 printf("Premature EOL\n");
4771#endif
4772 }
4773 return (status == RA_MATCH);
4774 }
4775
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004776 } // End of loop until the regstack is empty.
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004777
Bram Moolenaar9490b9a2019-09-08 17:20:12 +02004778 // NOTREACHED
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004779}
4780
4781/*
4782 * regtry - try match of "prog" with at rex.line["col"].
4783 * Returns 0 for failure, number of lines contained in the match otherwise.
4784 */
4785 static long
4786regtry(
4787 bt_regprog_T *prog,
4788 colnr_T col,
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004789 int *timed_out) // flag set on timeout or NULL
4790{
4791 rex.input = rex.line + col;
4792 rex.need_clear_subexpr = TRUE;
4793#ifdef FEAT_SYN_HL
4794 // Clear the external match subpointers if necessary.
4795 rex.need_clear_zsubexpr = (prog->reghasz == REX_SET);
4796#endif
4797
Paul Ollis65745772022-06-05 16:55:54 +01004798 if (regmatch(prog->program + 1, timed_out) == 0)
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004799 return 0;
4800
4801 cleanup_subexpr();
4802 if (REG_MULTI)
4803 {
4804 if (rex.reg_startpos[0].lnum < 0)
4805 {
4806 rex.reg_startpos[0].lnum = 0;
4807 rex.reg_startpos[0].col = col;
4808 }
4809 if (rex.reg_endpos[0].lnum < 0)
4810 {
4811 rex.reg_endpos[0].lnum = rex.lnum;
4812 rex.reg_endpos[0].col = (int)(rex.input - rex.line);
4813 }
4814 else
4815 // Use line number of "\ze".
4816 rex.lnum = rex.reg_endpos[0].lnum;
4817 }
4818 else
4819 {
4820 if (rex.reg_startp[0] == NULL)
4821 rex.reg_startp[0] = rex.line + col;
4822 if (rex.reg_endp[0] == NULL)
4823 rex.reg_endp[0] = rex.input;
4824 }
4825#ifdef FEAT_SYN_HL
4826 // Package any found \z(...\) matches for export. Default is none.
4827 unref_extmatch(re_extmatch_out);
4828 re_extmatch_out = NULL;
4829
4830 if (prog->reghasz == REX_SET)
4831 {
4832 int i;
4833
4834 cleanup_zsubexpr();
4835 re_extmatch_out = make_extmatch();
Bram Moolenaar7c77b342019-12-22 19:40:40 +01004836 if (re_extmatch_out == NULL)
4837 return 0;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004838 for (i = 0; i < NSUBEXP; i++)
4839 {
4840 if (REG_MULTI)
4841 {
4842 // Only accept single line matches.
4843 if (reg_startzpos[i].lnum >= 0
4844 && reg_endzpos[i].lnum == reg_startzpos[i].lnum
4845 && reg_endzpos[i].col >= reg_startzpos[i].col)
4846 re_extmatch_out->matches[i] =
4847 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
4848 + reg_startzpos[i].col,
4849 reg_endzpos[i].col - reg_startzpos[i].col);
4850 }
4851 else
4852 {
4853 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
4854 re_extmatch_out->matches[i] =
4855 vim_strnsave(reg_startzp[i],
Bram Moolenaar71ccd032020-06-12 22:59:11 +02004856 reg_endzp[i] - reg_startzp[i]);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004857 }
4858 }
4859 }
4860#endif
4861 return 1 + rex.lnum;
4862}
4863
4864/*
4865 * Match a regexp against a string ("line" points to the string) or multiple
Bram Moolenaardf365142021-05-03 20:01:45 +02004866 * lines (if "line" is NULL, use reg_getline()).
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004867 * Returns 0 for failure, number of lines contained in the match otherwise.
4868 */
4869 static long
4870bt_regexec_both(
4871 char_u *line,
Bram Moolenaar01105b32022-11-26 11:47:10 +00004872 colnr_T startcol, // column to start looking for match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004873 int *timed_out) // flag set on timeout or NULL
4874{
4875 bt_regprog_T *prog;
4876 char_u *s;
Bram Moolenaar01105b32022-11-26 11:47:10 +00004877 colnr_T col = startcol;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004878 long retval = 0L;
4879
4880 // Create "regstack" and "backpos" if they are not allocated yet.
4881 // We allocate *_INITIAL amount of bytes first and then set the grow size
4882 // to much bigger value to avoid many malloc calls in case of deep regular
4883 // expressions.
4884 if (regstack.ga_data == NULL)
4885 {
4886 // Use an item size of 1 byte, since we push different things
4887 // onto the regstack.
4888 ga_init2(&regstack, 1, REGSTACK_INITIAL);
4889 (void)ga_grow(&regstack, REGSTACK_INITIAL);
4890 regstack.ga_growsize = REGSTACK_INITIAL * 8;
4891 }
4892
4893 if (backpos.ga_data == NULL)
4894 {
4895 ga_init2(&backpos, sizeof(backpos_T), BACKPOS_INITIAL);
4896 (void)ga_grow(&backpos, BACKPOS_INITIAL);
4897 backpos.ga_growsize = BACKPOS_INITIAL * 8;
4898 }
4899
4900 if (REG_MULTI)
4901 {
4902 prog = (bt_regprog_T *)rex.reg_mmatch->regprog;
4903 line = reg_getline((linenr_T)0);
4904 rex.reg_startpos = rex.reg_mmatch->startpos;
4905 rex.reg_endpos = rex.reg_mmatch->endpos;
4906 }
4907 else
4908 {
4909 prog = (bt_regprog_T *)rex.reg_match->regprog;
4910 rex.reg_startp = rex.reg_match->startp;
4911 rex.reg_endp = rex.reg_match->endp;
4912 }
4913
4914 // Be paranoid...
4915 if (prog == NULL || line == NULL)
4916 {
RestorerZ68ebcee2023-05-31 17:12:14 +01004917 iemsg(e_null_argument);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004918 goto theend;
4919 }
4920
4921 // Check validity of program.
4922 if (prog_magic_wrong())
4923 goto theend;
4924
4925 // If the start column is past the maximum column: no need to try.
4926 if (rex.reg_maxcol > 0 && col >= rex.reg_maxcol)
4927 goto theend;
4928
4929 // If pattern contains "\c" or "\C": overrule value of rex.reg_ic
4930 if (prog->regflags & RF_ICASE)
4931 rex.reg_ic = TRUE;
4932 else if (prog->regflags & RF_NOICASE)
4933 rex.reg_ic = FALSE;
4934
4935 // If pattern contains "\Z" overrule value of rex.reg_icombine
4936 if (prog->regflags & RF_ICOMBINE)
4937 rex.reg_icombine = TRUE;
4938
4939 // If there is a "must appear" string, look for it.
4940 if (prog->regmust != NULL)
4941 {
4942 int c;
4943
4944 if (has_mbyte)
4945 c = (*mb_ptr2char)(prog->regmust);
4946 else
4947 c = *prog->regmust;
4948 s = line + col;
4949
4950 // This is used very often, esp. for ":global". Use three versions of
4951 // the loop to avoid overhead of conditions.
4952 if (!rex.reg_ic && !has_mbyte)
4953 while ((s = vim_strbyte(s, c)) != NULL)
4954 {
4955 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
4956 break; // Found it.
4957 ++s;
4958 }
4959 else if (!rex.reg_ic || (!enc_utf8 && mb_char2len(c) > 1))
4960 while ((s = vim_strchr(s, c)) != NULL)
4961 {
4962 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
4963 break; // Found it.
4964 MB_PTR_ADV(s);
4965 }
4966 else
4967 while ((s = cstrchr(s, c)) != NULL)
4968 {
4969 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
4970 break; // Found it.
4971 MB_PTR_ADV(s);
4972 }
4973 if (s == NULL) // Not present.
4974 goto theend;
4975 }
4976
4977 rex.line = line;
4978 rex.lnum = 0;
4979 reg_toolong = FALSE;
4980
4981 // Simplest case: Anchored match need be tried only once.
4982 if (prog->reganch)
4983 {
4984 int c;
4985
4986 if (has_mbyte)
4987 c = (*mb_ptr2char)(rex.line + col);
4988 else
4989 c = rex.line[col];
4990 if (prog->regstart == NUL
4991 || prog->regstart == c
4992 || (rex.reg_ic
4993 && (((enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
4994 || (c < 255 && prog->regstart < 255 &&
4995 MB_TOLOWER(prog->regstart) == MB_TOLOWER(c)))))
Paul Ollis65745772022-06-05 16:55:54 +01004996 retval = regtry(prog, col, timed_out);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02004997 else
4998 retval = 0;
4999 }
5000 else
5001 {
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005002 // Messy cases: unanchored match.
5003 while (!got_int)
5004 {
5005 if (prog->regstart != NUL)
5006 {
5007 // Skip until the char we know it must start with.
5008 // Used often, do some work to avoid call overhead.
5009 if (!rex.reg_ic && !has_mbyte)
5010 s = vim_strbyte(rex.line + col, prog->regstart);
5011 else
5012 s = cstrchr(rex.line + col, prog->regstart);
5013 if (s == NULL)
5014 {
5015 retval = 0;
5016 break;
5017 }
5018 col = (int)(s - rex.line);
5019 }
5020
5021 // Check for maximum column to try.
5022 if (rex.reg_maxcol > 0 && col >= rex.reg_maxcol)
5023 {
5024 retval = 0;
5025 break;
5026 }
5027
Paul Ollis65745772022-06-05 16:55:54 +01005028 retval = regtry(prog, col, timed_out);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005029 if (retval > 0)
5030 break;
5031
5032 // if not currently on the first line, get it again
5033 if (rex.lnum != 0)
5034 {
5035 rex.lnum = 0;
5036 rex.line = reg_getline((linenr_T)0);
5037 }
5038 if (rex.line[col] == NUL)
5039 break;
5040 if (has_mbyte)
5041 col += (*mb_ptr2len)(rex.line + col);
5042 else
5043 ++col;
5044#ifdef FEAT_RELTIME
Bram Moolenaar616592e2022-06-17 15:17:10 +01005045 if (bt_did_time_out(timed_out))
Paul Ollis65745772022-06-05 16:55:54 +01005046 break;
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005047#endif
5048 }
5049 }
5050
5051theend:
5052 // Free "reg_tofree" when it's a bit big.
5053 // Free regstack and backpos if they are bigger than their initial size.
5054 if (reg_tofreelen > 400)
5055 VIM_CLEAR(reg_tofree);
5056 if (regstack.ga_maxlen > REGSTACK_INITIAL)
5057 ga_clear(&regstack);
5058 if (backpos.ga_maxlen > BACKPOS_INITIAL)
5059 ga_clear(&backpos);
5060
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005061 if (retval > 0)
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005062 {
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005063 // Make sure the end is never before the start. Can happen when \zs
5064 // and \ze are used.
5065 if (REG_MULTI)
5066 {
5067 lpos_T *start = &rex.reg_mmatch->startpos[0];
5068 lpos_T *end = &rex.reg_mmatch->endpos[0];
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005069
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005070 if (end->lnum < start->lnum
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005071 || (end->lnum == start->lnum && end->col < start->col))
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005072 rex.reg_mmatch->endpos[0] = rex.reg_mmatch->startpos[0];
Bram Moolenaar01105b32022-11-26 11:47:10 +00005073
5074 // startpos[0] may be set by "\zs", also return the column where
5075 // the whole pattern matched.
5076 rex.reg_mmatch->rmm_matchcol = col;
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005077 }
5078 else
5079 {
5080 if (rex.reg_match->endp[0] < rex.reg_match->startp[0])
5081 rex.reg_match->endp[0] = rex.reg_match->startp[0];
Bram Moolenaar01105b32022-11-26 11:47:10 +00005082
5083 // startpos[0] may be set by "\zs", also return the column where
5084 // the whole pattern matched.
5085 rex.reg_match->rm_matchcol = col;
Bram Moolenaara3d10a52020-12-21 18:24:00 +01005086 }
Bram Moolenaara7a691c2020-12-09 16:36:04 +01005087 }
5088
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005089 return retval;
5090}
5091
5092/*
5093 * Match a regexp against a string.
5094 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
5095 * Uses curbuf for line count and 'iskeyword'.
5096 * if "line_lbr" is TRUE consider a "\n" in "line" to be a line break.
5097 *
5098 * Returns 0 for failure, number of lines contained in the match otherwise.
5099 */
5100 static int
5101bt_regexec_nl(
5102 regmatch_T *rmp,
5103 char_u *line, // string to match against
5104 colnr_T col, // column to start looking for match
5105 int line_lbr)
5106{
5107 rex.reg_match = rmp;
5108 rex.reg_mmatch = NULL;
5109 rex.reg_maxline = 0;
5110 rex.reg_line_lbr = line_lbr;
5111 rex.reg_buf = curbuf;
5112 rex.reg_win = NULL;
5113 rex.reg_ic = rmp->rm_ic;
5114 rex.reg_icombine = FALSE;
5115 rex.reg_maxcol = 0;
5116
Paul Ollis65745772022-06-05 16:55:54 +01005117 return bt_regexec_both(line, col, NULL);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005118}
5119
5120/*
5121 * Match a regexp against multiple lines.
5122 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
5123 * Uses curbuf for line count and 'iskeyword'.
5124 *
5125 * Return zero if there is no match. Return number of lines contained in the
5126 * match otherwise.
5127 */
5128 static long
5129bt_regexec_multi(
5130 regmmatch_T *rmp,
5131 win_T *win, // window in which to search or NULL
5132 buf_T *buf, // buffer in which to search
5133 linenr_T lnum, // nr of line to start looking for match
5134 colnr_T col, // column to start looking for match
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005135 int *timed_out) // flag set on timeout or NULL
5136{
Bram Moolenaarf4140482020-02-15 23:06:45 +01005137 init_regexec_multi(rmp, win, buf, lnum);
Paul Ollis65745772022-06-05 16:55:54 +01005138 return bt_regexec_both(NULL, col, timed_out);
Bram Moolenaar6d7d7cf2019-09-07 23:16:33 +02005139}
5140
5141/*
5142 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
5143 */
5144 static int
5145re_num_cmp(long_u val, char_u *scan)
5146{
5147 long_u n = OPERAND_MIN(scan);
5148
5149 if (OPERAND_CMP(scan) == '>')
5150 return val > n;
5151 if (OPERAND_CMP(scan) == '<')
5152 return val < n;
5153 return val == n;
5154}
5155
5156#ifdef BT_REGEXP_DUMP
5157
5158/*
5159 * regdump - dump a regexp onto stdout in vaguely comprehensible form
5160 */
5161 static void
5162regdump(char_u *pattern, bt_regprog_T *r)
5163{
5164 char_u *s;
5165 int op = EXACTLY; // Arbitrary non-END op.
5166 char_u *next;
5167 char_u *end = NULL;
5168 FILE *f;
5169
5170#ifdef BT_REGEXP_LOG
5171 f = fopen("bt_regexp_log.log", "a");
5172#else
5173 f = stdout;
5174#endif
5175 if (f == NULL)
5176 return;
5177 fprintf(f, "-------------------------------------\n\r\nregcomp(%s):\r\n", pattern);
5178
5179 s = r->program + 1;
5180 // Loop until we find the END that isn't before a referred next (an END
5181 // can also appear in a NOMATCH operand).
5182 while (op != END || s <= end)
5183 {
5184 op = OP(s);
5185 fprintf(f, "%2d%s", (int)(s - r->program), regprop(s)); // Where, what.
5186 next = regnext(s);
5187 if (next == NULL) // Next ptr.
5188 fprintf(f, "(0)");
5189 else
5190 fprintf(f, "(%d)", (int)((s - r->program) + (next - s)));
5191 if (end < next)
5192 end = next;
5193 if (op == BRACE_LIMITS)
5194 {
5195 // Two ints
5196 fprintf(f, " minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
5197 s += 8;
5198 }
5199 else if (op == BEHIND || op == NOBEHIND)
5200 {
5201 // one int
5202 fprintf(f, " count %ld", OPERAND_MIN(s));
5203 s += 4;
5204 }
5205 else if (op == RE_LNUM || op == RE_COL || op == RE_VCOL)
5206 {
5207 // one int plus comparator
5208 fprintf(f, " count %ld", OPERAND_MIN(s));
5209 s += 5;
5210 }
5211 s += 3;
5212 if (op == ANYOF || op == ANYOF + ADD_NL
5213 || op == ANYBUT || op == ANYBUT + ADD_NL
5214 || op == EXACTLY)
5215 {
5216 // Literal string, where present.
5217 fprintf(f, "\nxxxxxxxxx\n");
5218 while (*s != NUL)
5219 fprintf(f, "%c", *s++);
5220 fprintf(f, "\nxxxxxxxxx\n");
5221 s++;
5222 }
5223 fprintf(f, "\r\n");
5224 }
5225
5226 // Header fields of interest.
5227 if (r->regstart != NUL)
5228 fprintf(f, "start `%s' 0x%x; ", r->regstart < 256
5229 ? (char *)transchar(r->regstart)
5230 : "multibyte", r->regstart);
5231 if (r->reganch)
5232 fprintf(f, "anchored; ");
5233 if (r->regmust != NULL)
5234 fprintf(f, "must have \"%s\"", r->regmust);
5235 fprintf(f, "\r\n");
5236
5237#ifdef BT_REGEXP_LOG
5238 fclose(f);
5239#endif
5240}
5241#endif // BT_REGEXP_DUMP
5242
5243#ifdef DEBUG
5244/*
5245 * regprop - printable representation of opcode
5246 */
5247 static char_u *
5248regprop(char_u *op)
5249{
5250 char *p;
5251 static char buf[50];
5252
5253 STRCPY(buf, ":");
5254
5255 switch ((int) OP(op))
5256 {
5257 case BOL:
5258 p = "BOL";
5259 break;
5260 case EOL:
5261 p = "EOL";
5262 break;
5263 case RE_BOF:
5264 p = "BOF";
5265 break;
5266 case RE_EOF:
5267 p = "EOF";
5268 break;
5269 case CURSOR:
5270 p = "CURSOR";
5271 break;
5272 case RE_VISUAL:
5273 p = "RE_VISUAL";
5274 break;
5275 case RE_LNUM:
5276 p = "RE_LNUM";
5277 break;
5278 case RE_MARK:
5279 p = "RE_MARK";
5280 break;
5281 case RE_COL:
5282 p = "RE_COL";
5283 break;
5284 case RE_VCOL:
5285 p = "RE_VCOL";
5286 break;
5287 case BOW:
5288 p = "BOW";
5289 break;
5290 case EOW:
5291 p = "EOW";
5292 break;
5293 case ANY:
5294 p = "ANY";
5295 break;
5296 case ANY + ADD_NL:
5297 p = "ANY+NL";
5298 break;
5299 case ANYOF:
5300 p = "ANYOF";
5301 break;
5302 case ANYOF + ADD_NL:
5303 p = "ANYOF+NL";
5304 break;
5305 case ANYBUT:
5306 p = "ANYBUT";
5307 break;
5308 case ANYBUT + ADD_NL:
5309 p = "ANYBUT+NL";
5310 break;
5311 case IDENT:
5312 p = "IDENT";
5313 break;
5314 case IDENT + ADD_NL:
5315 p = "IDENT+NL";
5316 break;
5317 case SIDENT:
5318 p = "SIDENT";
5319 break;
5320 case SIDENT + ADD_NL:
5321 p = "SIDENT+NL";
5322 break;
5323 case KWORD:
5324 p = "KWORD";
5325 break;
5326 case KWORD + ADD_NL:
5327 p = "KWORD+NL";
5328 break;
5329 case SKWORD:
5330 p = "SKWORD";
5331 break;
5332 case SKWORD + ADD_NL:
5333 p = "SKWORD+NL";
5334 break;
5335 case FNAME:
5336 p = "FNAME";
5337 break;
5338 case FNAME + ADD_NL:
5339 p = "FNAME+NL";
5340 break;
5341 case SFNAME:
5342 p = "SFNAME";
5343 break;
5344 case SFNAME + ADD_NL:
5345 p = "SFNAME+NL";
5346 break;
5347 case PRINT:
5348 p = "PRINT";
5349 break;
5350 case PRINT + ADD_NL:
5351 p = "PRINT+NL";
5352 break;
5353 case SPRINT:
5354 p = "SPRINT";
5355 break;
5356 case SPRINT + ADD_NL:
5357 p = "SPRINT+NL";
5358 break;
5359 case WHITE:
5360 p = "WHITE";
5361 break;
5362 case WHITE + ADD_NL:
5363 p = "WHITE+NL";
5364 break;
5365 case NWHITE:
5366 p = "NWHITE";
5367 break;
5368 case NWHITE + ADD_NL:
5369 p = "NWHITE+NL";
5370 break;
5371 case DIGIT:
5372 p = "DIGIT";
5373 break;
5374 case DIGIT + ADD_NL:
5375 p = "DIGIT+NL";
5376 break;
5377 case NDIGIT:
5378 p = "NDIGIT";
5379 break;
5380 case NDIGIT + ADD_NL:
5381 p = "NDIGIT+NL";
5382 break;
5383 case HEX:
5384 p = "HEX";
5385 break;
5386 case HEX + ADD_NL:
5387 p = "HEX+NL";
5388 break;
5389 case NHEX:
5390 p = "NHEX";
5391 break;
5392 case NHEX + ADD_NL:
5393 p = "NHEX+NL";
5394 break;
5395 case OCTAL:
5396 p = "OCTAL";
5397 break;
5398 case OCTAL + ADD_NL:
5399 p = "OCTAL+NL";
5400 break;
5401 case NOCTAL:
5402 p = "NOCTAL";
5403 break;
5404 case NOCTAL + ADD_NL:
5405 p = "NOCTAL+NL";
5406 break;
5407 case WORD:
5408 p = "WORD";
5409 break;
5410 case WORD + ADD_NL:
5411 p = "WORD+NL";
5412 break;
5413 case NWORD:
5414 p = "NWORD";
5415 break;
5416 case NWORD + ADD_NL:
5417 p = "NWORD+NL";
5418 break;
5419 case HEAD:
5420 p = "HEAD";
5421 break;
5422 case HEAD + ADD_NL:
5423 p = "HEAD+NL";
5424 break;
5425 case NHEAD:
5426 p = "NHEAD";
5427 break;
5428 case NHEAD + ADD_NL:
5429 p = "NHEAD+NL";
5430 break;
5431 case ALPHA:
5432 p = "ALPHA";
5433 break;
5434 case ALPHA + ADD_NL:
5435 p = "ALPHA+NL";
5436 break;
5437 case NALPHA:
5438 p = "NALPHA";
5439 break;
5440 case NALPHA + ADD_NL:
5441 p = "NALPHA+NL";
5442 break;
5443 case LOWER:
5444 p = "LOWER";
5445 break;
5446 case LOWER + ADD_NL:
5447 p = "LOWER+NL";
5448 break;
5449 case NLOWER:
5450 p = "NLOWER";
5451 break;
5452 case NLOWER + ADD_NL:
5453 p = "NLOWER+NL";
5454 break;
5455 case UPPER:
5456 p = "UPPER";
5457 break;
5458 case UPPER + ADD_NL:
5459 p = "UPPER+NL";
5460 break;
5461 case NUPPER:
5462 p = "NUPPER";
5463 break;
5464 case NUPPER + ADD_NL:
5465 p = "NUPPER+NL";
5466 break;
5467 case BRANCH:
5468 p = "BRANCH";
5469 break;
5470 case EXACTLY:
5471 p = "EXACTLY";
5472 break;
5473 case NOTHING:
5474 p = "NOTHING";
5475 break;
5476 case BACK:
5477 p = "BACK";
5478 break;
5479 case END:
5480 p = "END";
5481 break;
5482 case MOPEN + 0:
5483 p = "MATCH START";
5484 break;
5485 case MOPEN + 1:
5486 case MOPEN + 2:
5487 case MOPEN + 3:
5488 case MOPEN + 4:
5489 case MOPEN + 5:
5490 case MOPEN + 6:
5491 case MOPEN + 7:
5492 case MOPEN + 8:
5493 case MOPEN + 9:
5494 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
5495 p = NULL;
5496 break;
5497 case MCLOSE + 0:
5498 p = "MATCH END";
5499 break;
5500 case MCLOSE + 1:
5501 case MCLOSE + 2:
5502 case MCLOSE + 3:
5503 case MCLOSE + 4:
5504 case MCLOSE + 5:
5505 case MCLOSE + 6:
5506 case MCLOSE + 7:
5507 case MCLOSE + 8:
5508 case MCLOSE + 9:
5509 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
5510 p = NULL;
5511 break;
5512 case BACKREF + 1:
5513 case BACKREF + 2:
5514 case BACKREF + 3:
5515 case BACKREF + 4:
5516 case BACKREF + 5:
5517 case BACKREF + 6:
5518 case BACKREF + 7:
5519 case BACKREF + 8:
5520 case BACKREF + 9:
5521 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
5522 p = NULL;
5523 break;
5524 case NOPEN:
5525 p = "NOPEN";
5526 break;
5527 case NCLOSE:
5528 p = "NCLOSE";
5529 break;
5530#ifdef FEAT_SYN_HL
5531 case ZOPEN + 1:
5532 case ZOPEN + 2:
5533 case ZOPEN + 3:
5534 case ZOPEN + 4:
5535 case ZOPEN + 5:
5536 case ZOPEN + 6:
5537 case ZOPEN + 7:
5538 case ZOPEN + 8:
5539 case ZOPEN + 9:
5540 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
5541 p = NULL;
5542 break;
5543 case ZCLOSE + 1:
5544 case ZCLOSE + 2:
5545 case ZCLOSE + 3:
5546 case ZCLOSE + 4:
5547 case ZCLOSE + 5:
5548 case ZCLOSE + 6:
5549 case ZCLOSE + 7:
5550 case ZCLOSE + 8:
5551 case ZCLOSE + 9:
5552 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
5553 p = NULL;
5554 break;
5555 case ZREF + 1:
5556 case ZREF + 2:
5557 case ZREF + 3:
5558 case ZREF + 4:
5559 case ZREF + 5:
5560 case ZREF + 6:
5561 case ZREF + 7:
5562 case ZREF + 8:
5563 case ZREF + 9:
5564 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
5565 p = NULL;
5566 break;
5567#endif
5568 case STAR:
5569 p = "STAR";
5570 break;
5571 case PLUS:
5572 p = "PLUS";
5573 break;
5574 case NOMATCH:
5575 p = "NOMATCH";
5576 break;
5577 case MATCH:
5578 p = "MATCH";
5579 break;
5580 case BEHIND:
5581 p = "BEHIND";
5582 break;
5583 case NOBEHIND:
5584 p = "NOBEHIND";
5585 break;
5586 case SUBPAT:
5587 p = "SUBPAT";
5588 break;
5589 case BRACE_LIMITS:
5590 p = "BRACE_LIMITS";
5591 break;
5592 case BRACE_SIMPLE:
5593 p = "BRACE_SIMPLE";
5594 break;
5595 case BRACE_COMPLEX + 0:
5596 case BRACE_COMPLEX + 1:
5597 case BRACE_COMPLEX + 2:
5598 case BRACE_COMPLEX + 3:
5599 case BRACE_COMPLEX + 4:
5600 case BRACE_COMPLEX + 5:
5601 case BRACE_COMPLEX + 6:
5602 case BRACE_COMPLEX + 7:
5603 case BRACE_COMPLEX + 8:
5604 case BRACE_COMPLEX + 9:
5605 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
5606 p = NULL;
5607 break;
5608 case MULTIBYTECODE:
5609 p = "MULTIBYTECODE";
5610 break;
5611 case NEWL:
5612 p = "NEWL";
5613 break;
5614 default:
5615 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
5616 p = NULL;
5617 break;
5618 }
5619 if (p != NULL)
5620 STRCAT(buf, p);
5621 return (char_u *)buf;
5622}
5623#endif // DEBUG