The seventh batch
[git.git] / grep.c
blob13232a904aca4906f28aae2e04165d1fda47b922
1 #include "cache.h"
2 #include "config.h"
3 #include "grep.h"
4 #include "object-store.h"
5 #include "userdiff.h"
6 #include "xdiff-interface.h"
7 #include "diff.h"
8 #include "diffcore.h"
9 #include "commit.h"
10 #include "quote.h"
11 #include "help.h"
13 static int grep_source_load(struct grep_source *gs);
14 static int grep_source_is_binary(struct grep_source *gs,
15 struct index_state *istate);
17 static struct grep_opt grep_defaults;
19 #ifdef USE_LIBPCRE2
20 static pcre2_general_context *pcre2_global_context;
22 static void *pcre2_malloc(PCRE2_SIZE size, MAYBE_UNUSED void *memory_data)
24 return malloc(size);
27 static void pcre2_free(void *pointer, MAYBE_UNUSED void *memory_data)
29 free(pointer);
31 #endif
33 static const char *color_grep_slots[] = {
34 [GREP_COLOR_CONTEXT] = "context",
35 [GREP_COLOR_FILENAME] = "filename",
36 [GREP_COLOR_FUNCTION] = "function",
37 [GREP_COLOR_LINENO] = "lineNumber",
38 [GREP_COLOR_COLUMNNO] = "column",
39 [GREP_COLOR_MATCH_CONTEXT] = "matchContext",
40 [GREP_COLOR_MATCH_SELECTED] = "matchSelected",
41 [GREP_COLOR_SELECTED] = "selected",
42 [GREP_COLOR_SEP] = "separator",
45 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
47 fwrite(buf, size, 1, stdout);
50 static void color_set(char *dst, const char *color_bytes)
52 xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes);
56 * Initialize the grep_defaults template with hardcoded defaults.
57 * We could let the compiler do this, but without C99 initializers
58 * the code gets unwieldy and unreadable, so...
60 void init_grep_defaults(struct repository *repo)
62 struct grep_opt *opt = &grep_defaults;
63 static int run_once;
65 if (run_once)
66 return;
67 run_once++;
69 memset(opt, 0, sizeof(*opt));
70 opt->repo = repo;
71 opt->relative = 1;
72 opt->pathname = 1;
73 opt->max_depth = -1;
74 opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
75 color_set(opt->colors[GREP_COLOR_CONTEXT], "");
76 color_set(opt->colors[GREP_COLOR_FILENAME], "");
77 color_set(opt->colors[GREP_COLOR_FUNCTION], "");
78 color_set(opt->colors[GREP_COLOR_LINENO], "");
79 color_set(opt->colors[GREP_COLOR_COLUMNNO], "");
80 color_set(opt->colors[GREP_COLOR_MATCH_CONTEXT], GIT_COLOR_BOLD_RED);
81 color_set(opt->colors[GREP_COLOR_MATCH_SELECTED], GIT_COLOR_BOLD_RED);
82 color_set(opt->colors[GREP_COLOR_SELECTED], "");
83 color_set(opt->colors[GREP_COLOR_SEP], GIT_COLOR_CYAN);
84 opt->only_matching = 0;
85 opt->color = -1;
86 opt->output = std_output;
89 static int parse_pattern_type_arg(const char *opt, const char *arg)
91 if (!strcmp(arg, "default"))
92 return GREP_PATTERN_TYPE_UNSPECIFIED;
93 else if (!strcmp(arg, "basic"))
94 return GREP_PATTERN_TYPE_BRE;
95 else if (!strcmp(arg, "extended"))
96 return GREP_PATTERN_TYPE_ERE;
97 else if (!strcmp(arg, "fixed"))
98 return GREP_PATTERN_TYPE_FIXED;
99 else if (!strcmp(arg, "perl"))
100 return GREP_PATTERN_TYPE_PCRE;
101 die("bad %s argument: %s", opt, arg);
104 define_list_config_array_extra(color_grep_slots, {"match"});
107 * Read the configuration file once and store it in
108 * the grep_defaults template.
110 int grep_config(const char *var, const char *value, void *cb)
112 struct grep_opt *opt = &grep_defaults;
113 const char *slot;
115 if (userdiff_config(var, value) < 0)
116 return -1;
118 if (!strcmp(var, "grep.extendedregexp")) {
119 opt->extended_regexp_option = git_config_bool(var, value);
120 return 0;
123 if (!strcmp(var, "grep.patterntype")) {
124 opt->pattern_type_option = parse_pattern_type_arg(var, value);
125 return 0;
128 if (!strcmp(var, "grep.linenumber")) {
129 opt->linenum = git_config_bool(var, value);
130 return 0;
132 if (!strcmp(var, "grep.column")) {
133 opt->columnnum = git_config_bool(var, value);
134 return 0;
137 if (!strcmp(var, "grep.fullname")) {
138 opt->relative = !git_config_bool(var, value);
139 return 0;
142 if (!strcmp(var, "color.grep"))
143 opt->color = git_config_colorbool(var, value);
144 if (!strcmp(var, "color.grep.match")) {
145 if (grep_config("color.grep.matchcontext", value, cb) < 0)
146 return -1;
147 if (grep_config("color.grep.matchselected", value, cb) < 0)
148 return -1;
149 } else if (skip_prefix(var, "color.grep.", &slot)) {
150 int i = LOOKUP_CONFIG(color_grep_slots, slot);
151 char *color;
153 if (i < 0)
154 return -1;
155 color = opt->colors[i];
156 if (!value)
157 return config_error_nonbool(var);
158 return color_parse(value, color);
160 return 0;
164 * Initialize one instance of grep_opt and copy the
165 * default values from the template we read the configuration
166 * information in an earlier call to git_config(grep_config).
168 * If using PCRE, make sure that the library is configured
169 * to use the same allocator as Git (e.g. nedmalloc on Windows).
171 * Any allocated memory needs to be released in grep_destroy().
173 void grep_init(struct grep_opt *opt, struct repository *repo, const char *prefix)
175 struct grep_opt *def = &grep_defaults;
176 int i;
178 #if defined(USE_LIBPCRE2)
179 if (!pcre2_global_context)
180 pcre2_global_context = pcre2_general_context_create(
181 pcre2_malloc, pcre2_free, NULL);
182 #endif
184 #ifdef USE_LIBPCRE1
185 pcre_malloc = malloc;
186 pcre_free = free;
187 #endif
189 memset(opt, 0, sizeof(*opt));
190 opt->repo = repo;
191 opt->prefix = prefix;
192 opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
193 opt->pattern_tail = &opt->pattern_list;
194 opt->header_tail = &opt->header_list;
196 opt->only_matching = def->only_matching;
197 opt->color = def->color;
198 opt->extended_regexp_option = def->extended_regexp_option;
199 opt->pattern_type_option = def->pattern_type_option;
200 opt->linenum = def->linenum;
201 opt->columnnum = def->columnnum;
202 opt->max_depth = def->max_depth;
203 opt->pathname = def->pathname;
204 opt->relative = def->relative;
205 opt->output = def->output;
207 for (i = 0; i < NR_GREP_COLORS; i++)
208 color_set(opt->colors[i], def->colors[i]);
211 void grep_destroy(void)
213 #ifdef USE_LIBPCRE2
214 pcre2_general_context_free(pcre2_global_context);
215 #endif
218 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
221 * When committing to the pattern type by setting the relevant
222 * fields in grep_opt it's generally not necessary to zero out
223 * the fields we're not choosing, since they won't have been
224 * set by anything. The extended_regexp_option field is the
225 * only exception to this.
227 * This is because in the process of parsing grep.patternType
228 * & grep.extendedRegexp we set opt->pattern_type_option and
229 * opt->extended_regexp_option, respectively. We then
230 * internally use opt->extended_regexp_option to see if we're
231 * compiling an ERE. It must be unset if that's not actually
232 * the case.
234 if (pattern_type != GREP_PATTERN_TYPE_ERE &&
235 opt->extended_regexp_option)
236 opt->extended_regexp_option = 0;
238 switch (pattern_type) {
239 case GREP_PATTERN_TYPE_UNSPECIFIED:
240 /* fall through */
242 case GREP_PATTERN_TYPE_BRE:
243 break;
245 case GREP_PATTERN_TYPE_ERE:
246 opt->extended_regexp_option = 1;
247 break;
249 case GREP_PATTERN_TYPE_FIXED:
250 opt->fixed = 1;
251 break;
253 case GREP_PATTERN_TYPE_PCRE:
254 #ifdef USE_LIBPCRE2
255 opt->pcre2 = 1;
256 #else
258 * It's important that pcre1 always be assigned to
259 * even when there's no USE_LIBPCRE* defined. We still
260 * call the PCRE stub function, it just dies with
261 * "cannot use Perl-compatible regexes[...]".
263 opt->pcre1 = 1;
264 #endif
265 break;
269 void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
271 if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
272 grep_set_pattern_type_option(pattern_type, opt);
273 else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
274 grep_set_pattern_type_option(opt->pattern_type_option, opt);
275 else if (opt->extended_regexp_option)
277 * This branch *must* happen after setting from the
278 * opt->pattern_type_option above, we don't want
279 * grep.extendedRegexp to override grep.patternType!
281 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
284 static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
285 const char *origin, int no,
286 enum grep_pat_token t,
287 enum grep_header_field field)
289 struct grep_pat *p = xcalloc(1, sizeof(*p));
290 p->pattern = xmemdupz(pat, patlen);
291 p->patternlen = patlen;
292 p->origin = origin;
293 p->no = no;
294 p->token = t;
295 p->field = field;
296 return p;
299 static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
301 **tail = p;
302 *tail = &p->next;
303 p->next = NULL;
305 switch (p->token) {
306 case GREP_PATTERN: /* atom */
307 case GREP_PATTERN_HEAD:
308 case GREP_PATTERN_BODY:
309 for (;;) {
310 struct grep_pat *new_pat;
311 size_t len = 0;
312 char *cp = p->pattern + p->patternlen, *nl = NULL;
313 while (++len <= p->patternlen) {
314 if (*(--cp) == '\n') {
315 nl = cp;
316 break;
319 if (!nl)
320 break;
321 new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
322 p->no, p->token, p->field);
323 new_pat->next = p->next;
324 if (!p->next)
325 *tail = &new_pat->next;
326 p->next = new_pat;
327 *nl = '\0';
328 p->patternlen -= len;
330 break;
331 default:
332 break;
336 void append_header_grep_pattern(struct grep_opt *opt,
337 enum grep_header_field field, const char *pat)
339 struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
340 GREP_PATTERN_HEAD, field);
341 if (field == GREP_HEADER_REFLOG)
342 opt->use_reflog_filter = 1;
343 do_append_grep_pat(&opt->header_tail, p);
346 void append_grep_pattern(struct grep_opt *opt, const char *pat,
347 const char *origin, int no, enum grep_pat_token t)
349 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
352 void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
353 const char *origin, int no, enum grep_pat_token t)
355 struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
356 do_append_grep_pat(&opt->pattern_tail, p);
359 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
361 struct grep_pat *pat;
362 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
363 *ret = *opt;
365 ret->pattern_list = NULL;
366 ret->pattern_tail = &ret->pattern_list;
368 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
370 if(pat->token == GREP_PATTERN_HEAD)
371 append_header_grep_pattern(ret, pat->field,
372 pat->pattern);
373 else
374 append_grep_pat(ret, pat->pattern, pat->patternlen,
375 pat->origin, pat->no, pat->token);
378 return ret;
381 static NORETURN void compile_regexp_failed(const struct grep_pat *p,
382 const char *error)
384 char where[1024];
386 if (p->no)
387 xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
388 else if (p->origin)
389 xsnprintf(where, sizeof(where), "%s, ", p->origin);
390 else
391 where[0] = 0;
393 die("%s'%s': %s", where, p->pattern, error);
396 static int is_fixed(const char *s, size_t len)
398 size_t i;
400 for (i = 0; i < len; i++) {
401 if (is_regex_special(s[i]))
402 return 0;
405 return 1;
408 #ifdef USE_LIBPCRE1
409 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
411 const char *error;
412 int erroffset;
413 int options = PCRE_MULTILINE;
414 int study_options = 0;
416 if (opt->ignore_case) {
417 if (!opt->ignore_locale && has_non_ascii(p->pattern))
418 p->pcre1_tables = pcre_maketables();
419 options |= PCRE_CASELESS;
421 if (!opt->ignore_locale && is_utf8_locale() && has_non_ascii(p->pattern))
422 options |= PCRE_UTF8;
424 p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
425 p->pcre1_tables);
426 if (!p->pcre1_regexp)
427 compile_regexp_failed(p, error);
429 #if defined(PCRE_CONFIG_JIT) && !defined(NO_LIBPCRE1_JIT)
430 pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on);
431 if (opt->debug)
432 fprintf(stderr, "pcre1_jit_on=%d\n", p->pcre1_jit_on);
434 if (p->pcre1_jit_on)
435 study_options = PCRE_STUDY_JIT_COMPILE;
436 #endif
438 p->pcre1_extra_info = pcre_study(p->pcre1_regexp, study_options, &error);
439 if (!p->pcre1_extra_info && error)
440 die("%s", error);
443 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
444 regmatch_t *match, int eflags)
446 int ovector[30], ret, flags = PCRE_NO_UTF8_CHECK;
448 if (eflags & REG_NOTBOL)
449 flags |= PCRE_NOTBOL;
451 ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
452 eol - line, 0, flags, ovector,
453 ARRAY_SIZE(ovector));
455 if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
456 die("pcre_exec failed with error code %d", ret);
457 if (ret > 0) {
458 ret = 0;
459 match->rm_so = ovector[0];
460 match->rm_eo = ovector[1];
463 return ret;
466 static void free_pcre1_regexp(struct grep_pat *p)
468 pcre_free(p->pcre1_regexp);
469 #ifdef PCRE_CONFIG_JIT
470 if (p->pcre1_jit_on)
471 pcre_free_study(p->pcre1_extra_info);
472 else
473 #endif
474 pcre_free(p->pcre1_extra_info);
475 pcre_free((void *)p->pcre1_tables);
477 #else /* !USE_LIBPCRE1 */
478 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
480 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
483 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
484 regmatch_t *match, int eflags)
486 return 1;
489 static void free_pcre1_regexp(struct grep_pat *p)
492 #endif /* !USE_LIBPCRE1 */
494 #ifdef USE_LIBPCRE2
495 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
497 int error;
498 PCRE2_UCHAR errbuf[256];
499 PCRE2_SIZE erroffset;
500 int options = PCRE2_MULTILINE;
501 int jitret;
502 int patinforet;
503 size_t jitsizearg;
505 assert(opt->pcre2);
507 p->pcre2_compile_context = NULL;
509 /* pcre2_global_context is initialized in append_grep_pattern */
510 if (opt->ignore_case) {
511 if (!opt->ignore_locale && has_non_ascii(p->pattern)) {
512 if (!pcre2_global_context)
513 BUG("pcre2_global_context uninitialized");
514 p->pcre2_tables = pcre2_maketables(pcre2_global_context);
515 p->pcre2_compile_context = pcre2_compile_context_create(NULL);
516 pcre2_set_character_tables(p->pcre2_compile_context,
517 p->pcre2_tables);
519 options |= PCRE2_CASELESS;
521 if (!opt->ignore_locale && is_utf8_locale() && has_non_ascii(p->pattern) &&
522 !(!opt->ignore_case && (p->fixed || p->is_fixed)))
523 options |= PCRE2_UTF;
525 p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern,
526 p->patternlen, options, &error, &erroffset,
527 p->pcre2_compile_context);
529 if (p->pcre2_pattern) {
530 p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL);
531 if (!p->pcre2_match_data)
532 die("Couldn't allocate PCRE2 match data");
533 } else {
534 pcre2_get_error_message(error, errbuf, sizeof(errbuf));
535 compile_regexp_failed(p, (const char *)&errbuf);
538 pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on);
539 if (opt->debug)
540 fprintf(stderr, "pcre2_jit_on=%d\n", p->pcre2_jit_on);
541 if (p->pcre2_jit_on) {
542 jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE);
543 if (jitret)
544 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret);
547 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
548 * tells us whether the library itself supports JIT,
549 * but to see whether we're going to be actually using
550 * JIT we need to extract PCRE2_INFO_JITSIZE from the
551 * pattern *after* we do pcre2_jit_compile() above.
553 * This is because if the pattern contains the
554 * (*NO_JIT) verb (see pcre2syntax(3))
555 * pcre2_jit_compile() will exit early with 0. If we
556 * then proceed to call pcre2_jit_match() further down
557 * the line instead of pcre2_match() we'll either
558 * segfault (pre PCRE 10.31) or run into a fatal error
559 * (post PCRE2 10.31)
561 patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg);
562 if (patinforet)
563 BUG("pcre2_pattern_info() failed: %d", patinforet);
564 if (jitsizearg == 0) {
565 p->pcre2_jit_on = 0;
566 if (opt->debug)
567 fprintf(stderr, "pcre2_jit_on=%d: (*NO_JIT) in regex\n",
568 p->pcre2_jit_on);
569 return;
574 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
575 regmatch_t *match, int eflags)
577 int ret, flags = 0;
578 PCRE2_SIZE *ovector;
579 PCRE2_UCHAR errbuf[256];
581 if (eflags & REG_NOTBOL)
582 flags |= PCRE2_NOTBOL;
584 if (p->pcre2_jit_on)
585 ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line,
586 eol - line, 0, flags, p->pcre2_match_data,
587 NULL);
588 else
589 ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line,
590 eol - line, 0, flags, p->pcre2_match_data,
591 NULL);
593 if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) {
594 pcre2_get_error_message(ret, errbuf, sizeof(errbuf));
595 die("%s failed with error code %d: %s",
596 (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret,
597 errbuf);
599 if (ret > 0) {
600 ovector = pcre2_get_ovector_pointer(p->pcre2_match_data);
601 ret = 0;
602 match->rm_so = (int)ovector[0];
603 match->rm_eo = (int)ovector[1];
606 return ret;
609 static void free_pcre2_pattern(struct grep_pat *p)
611 pcre2_compile_context_free(p->pcre2_compile_context);
612 pcre2_code_free(p->pcre2_pattern);
613 pcre2_match_data_free(p->pcre2_match_data);
614 free((void *)p->pcre2_tables);
616 #else /* !USE_LIBPCRE2 */
617 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
620 * Unreachable until USE_LIBPCRE2 becomes synonymous with
621 * USE_LIBPCRE. See the sibling comment in
622 * grep_set_pattern_type_option().
624 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
627 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
628 regmatch_t *match, int eflags)
630 return 1;
633 static void free_pcre2_pattern(struct grep_pat *p)
637 static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
639 struct strbuf sb = STRBUF_INIT;
640 int err;
641 int regflags = 0;
643 basic_regex_quote_buf(&sb, p->pattern);
644 if (opt->ignore_case)
645 regflags |= REG_ICASE;
646 err = regcomp(&p->regexp, sb.buf, regflags);
647 if (opt->debug)
648 fprintf(stderr, "fixed %s\n", sb.buf);
649 strbuf_release(&sb);
650 if (err) {
651 char errbuf[1024];
652 regerror(err, &p->regexp, errbuf, sizeof(errbuf));
653 compile_regexp_failed(p, errbuf);
656 #endif /* !USE_LIBPCRE2 */
658 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
660 int err;
661 int regflags = REG_NEWLINE;
663 p->word_regexp = opt->word_regexp;
664 p->ignore_case = opt->ignore_case;
665 p->fixed = opt->fixed;
667 if (memchr(p->pattern, 0, p->patternlen) && !opt->pcre2)
668 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
670 p->is_fixed = is_fixed(p->pattern, p->patternlen);
671 #ifdef USE_LIBPCRE2
672 if (!p->fixed && !p->is_fixed) {
673 const char *no_jit = "(*NO_JIT)";
674 const int no_jit_len = strlen(no_jit);
675 if (starts_with(p->pattern, no_jit) &&
676 is_fixed(p->pattern + no_jit_len,
677 p->patternlen - no_jit_len))
678 p->is_fixed = 1;
680 #endif
681 if (p->fixed || p->is_fixed) {
682 #ifdef USE_LIBPCRE2
683 opt->pcre2 = 1;
684 if (p->is_fixed) {
685 compile_pcre2_pattern(p, opt);
686 } else {
688 * E.g. t7811-grep-open.sh relies on the
689 * pattern being restored.
691 char *old_pattern = p->pattern;
692 size_t old_patternlen = p->patternlen;
693 struct strbuf sb = STRBUF_INIT;
696 * There is the PCRE2_LITERAL flag, but it's
697 * only in PCRE v2 10.30 and later. Needing to
698 * ifdef our way around that and dealing with
699 * it + PCRE2_MULTILINE being an error is more
700 * complex than just quoting this ourselves.
702 strbuf_add(&sb, "\\Q", 2);
703 strbuf_add(&sb, p->pattern, p->patternlen);
704 strbuf_add(&sb, "\\E", 2);
706 p->pattern = sb.buf;
707 p->patternlen = sb.len;
708 compile_pcre2_pattern(p, opt);
709 p->pattern = old_pattern;
710 p->patternlen = old_patternlen;
711 strbuf_release(&sb);
713 #else /* !USE_LIBPCRE2 */
714 compile_fixed_regexp(p, opt);
715 #endif /* !USE_LIBPCRE2 */
716 return;
719 if (opt->pcre2) {
720 compile_pcre2_pattern(p, opt);
721 return;
724 if (opt->pcre1) {
725 compile_pcre1_regexp(p, opt);
726 return;
729 if (p->ignore_case)
730 regflags |= REG_ICASE;
731 if (opt->extended_regexp_option)
732 regflags |= REG_EXTENDED;
733 err = regcomp(&p->regexp, p->pattern, regflags);
734 if (err) {
735 char errbuf[1024];
736 regerror(err, &p->regexp, errbuf, 1024);
737 compile_regexp_failed(p, errbuf);
741 static struct grep_expr *compile_pattern_or(struct grep_pat **);
742 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
744 struct grep_pat *p;
745 struct grep_expr *x;
747 p = *list;
748 if (!p)
749 return NULL;
750 switch (p->token) {
751 case GREP_PATTERN: /* atom */
752 case GREP_PATTERN_HEAD:
753 case GREP_PATTERN_BODY:
754 x = xcalloc(1, sizeof (struct grep_expr));
755 x->node = GREP_NODE_ATOM;
756 x->u.atom = p;
757 *list = p->next;
758 return x;
759 case GREP_OPEN_PAREN:
760 *list = p->next;
761 x = compile_pattern_or(list);
762 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
763 die("unmatched parenthesis");
764 *list = (*list)->next;
765 return x;
766 default:
767 return NULL;
771 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
773 struct grep_pat *p;
774 struct grep_expr *x;
776 p = *list;
777 if (!p)
778 return NULL;
779 switch (p->token) {
780 case GREP_NOT:
781 if (!p->next)
782 die("--not not followed by pattern expression");
783 *list = p->next;
784 x = xcalloc(1, sizeof (struct grep_expr));
785 x->node = GREP_NODE_NOT;
786 x->u.unary = compile_pattern_not(list);
787 if (!x->u.unary)
788 die("--not followed by non pattern expression");
789 return x;
790 default:
791 return compile_pattern_atom(list);
795 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
797 struct grep_pat *p;
798 struct grep_expr *x, *y, *z;
800 x = compile_pattern_not(list);
801 p = *list;
802 if (p && p->token == GREP_AND) {
803 if (!p->next)
804 die("--and not followed by pattern expression");
805 *list = p->next;
806 y = compile_pattern_and(list);
807 if (!y)
808 die("--and not followed by pattern expression");
809 z = xcalloc(1, sizeof (struct grep_expr));
810 z->node = GREP_NODE_AND;
811 z->u.binary.left = x;
812 z->u.binary.right = y;
813 return z;
815 return x;
818 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
820 struct grep_pat *p;
821 struct grep_expr *x, *y, *z;
823 x = compile_pattern_and(list);
824 p = *list;
825 if (x && p && p->token != GREP_CLOSE_PAREN) {
826 y = compile_pattern_or(list);
827 if (!y)
828 die("not a pattern expression %s", p->pattern);
829 z = xcalloc(1, sizeof (struct grep_expr));
830 z->node = GREP_NODE_OR;
831 z->u.binary.left = x;
832 z->u.binary.right = y;
833 return z;
835 return x;
838 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
840 return compile_pattern_or(list);
843 static void indent(int in)
845 while (in-- > 0)
846 fputc(' ', stderr);
849 static void dump_grep_pat(struct grep_pat *p)
851 switch (p->token) {
852 case GREP_AND: fprintf(stderr, "*and*"); break;
853 case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
854 case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
855 case GREP_NOT: fprintf(stderr, "*not*"); break;
856 case GREP_OR: fprintf(stderr, "*or*"); break;
858 case GREP_PATTERN: fprintf(stderr, "pattern"); break;
859 case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
860 case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
863 switch (p->token) {
864 default: break;
865 case GREP_PATTERN_HEAD:
866 fprintf(stderr, "<head %d>", p->field); break;
867 case GREP_PATTERN_BODY:
868 fprintf(stderr, "<body>"); break;
870 switch (p->token) {
871 default: break;
872 case GREP_PATTERN_HEAD:
873 case GREP_PATTERN_BODY:
874 case GREP_PATTERN:
875 fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
876 break;
878 fputc('\n', stderr);
881 static void dump_grep_expression_1(struct grep_expr *x, int in)
883 indent(in);
884 switch (x->node) {
885 case GREP_NODE_TRUE:
886 fprintf(stderr, "true\n");
887 break;
888 case GREP_NODE_ATOM:
889 dump_grep_pat(x->u.atom);
890 break;
891 case GREP_NODE_NOT:
892 fprintf(stderr, "(not\n");
893 dump_grep_expression_1(x->u.unary, in+1);
894 indent(in);
895 fprintf(stderr, ")\n");
896 break;
897 case GREP_NODE_AND:
898 fprintf(stderr, "(and\n");
899 dump_grep_expression_1(x->u.binary.left, in+1);
900 dump_grep_expression_1(x->u.binary.right, in+1);
901 indent(in);
902 fprintf(stderr, ")\n");
903 break;
904 case GREP_NODE_OR:
905 fprintf(stderr, "(or\n");
906 dump_grep_expression_1(x->u.binary.left, in+1);
907 dump_grep_expression_1(x->u.binary.right, in+1);
908 indent(in);
909 fprintf(stderr, ")\n");
910 break;
914 static void dump_grep_expression(struct grep_opt *opt)
916 struct grep_expr *x = opt->pattern_expression;
918 if (opt->all_match)
919 fprintf(stderr, "[all-match]\n");
920 dump_grep_expression_1(x, 0);
921 fflush(NULL);
924 static struct grep_expr *grep_true_expr(void)
926 struct grep_expr *z = xcalloc(1, sizeof(*z));
927 z->node = GREP_NODE_TRUE;
928 return z;
931 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
933 struct grep_expr *z = xcalloc(1, sizeof(*z));
934 z->node = GREP_NODE_OR;
935 z->u.binary.left = left;
936 z->u.binary.right = right;
937 return z;
940 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
942 struct grep_pat *p;
943 struct grep_expr *header_expr;
944 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
945 enum grep_header_field fld;
947 if (!opt->header_list)
948 return NULL;
950 for (p = opt->header_list; p; p = p->next) {
951 if (p->token != GREP_PATTERN_HEAD)
952 BUG("a non-header pattern in grep header list.");
953 if (p->field < GREP_HEADER_FIELD_MIN ||
954 GREP_HEADER_FIELD_MAX <= p->field)
955 BUG("unknown header field %d", p->field);
956 compile_regexp(p, opt);
959 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
960 header_group[fld] = NULL;
962 for (p = opt->header_list; p; p = p->next) {
963 struct grep_expr *h;
964 struct grep_pat *pp = p;
966 h = compile_pattern_atom(&pp);
967 if (!h || pp != p->next)
968 BUG("malformed header expr");
969 if (!header_group[p->field]) {
970 header_group[p->field] = h;
971 continue;
973 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
976 header_expr = NULL;
978 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
979 if (!header_group[fld])
980 continue;
981 if (!header_expr)
982 header_expr = grep_true_expr();
983 header_expr = grep_or_expr(header_group[fld], header_expr);
985 return header_expr;
988 static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
990 struct grep_expr *z = x;
992 while (x) {
993 assert(x->node == GREP_NODE_OR);
994 if (x->u.binary.right &&
995 x->u.binary.right->node == GREP_NODE_TRUE) {
996 x->u.binary.right = y;
997 break;
999 x = x->u.binary.right;
1001 return z;
1004 static void compile_grep_patterns_real(struct grep_opt *opt)
1006 struct grep_pat *p;
1007 struct grep_expr *header_expr = prep_header_patterns(opt);
1009 for (p = opt->pattern_list; p; p = p->next) {
1010 switch (p->token) {
1011 case GREP_PATTERN: /* atom */
1012 case GREP_PATTERN_HEAD:
1013 case GREP_PATTERN_BODY:
1014 compile_regexp(p, opt);
1015 break;
1016 default:
1017 opt->extended = 1;
1018 break;
1022 if (opt->all_match || header_expr)
1023 opt->extended = 1;
1024 else if (!opt->extended && !opt->debug)
1025 return;
1027 p = opt->pattern_list;
1028 if (p)
1029 opt->pattern_expression = compile_pattern_expr(&p);
1030 if (p)
1031 die("incomplete pattern expression: %s", p->pattern);
1033 if (!header_expr)
1034 return;
1036 if (!opt->pattern_expression)
1037 opt->pattern_expression = header_expr;
1038 else if (opt->all_match)
1039 opt->pattern_expression = grep_splice_or(header_expr,
1040 opt->pattern_expression);
1041 else
1042 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
1043 header_expr);
1044 opt->all_match = 1;
1047 void compile_grep_patterns(struct grep_opt *opt)
1049 compile_grep_patterns_real(opt);
1050 if (opt->debug)
1051 dump_grep_expression(opt);
1054 static void free_pattern_expr(struct grep_expr *x)
1056 switch (x->node) {
1057 case GREP_NODE_TRUE:
1058 case GREP_NODE_ATOM:
1059 break;
1060 case GREP_NODE_NOT:
1061 free_pattern_expr(x->u.unary);
1062 break;
1063 case GREP_NODE_AND:
1064 case GREP_NODE_OR:
1065 free_pattern_expr(x->u.binary.left);
1066 free_pattern_expr(x->u.binary.right);
1067 break;
1069 free(x);
1072 void free_grep_patterns(struct grep_opt *opt)
1074 struct grep_pat *p, *n;
1076 for (p = opt->pattern_list; p; p = n) {
1077 n = p->next;
1078 switch (p->token) {
1079 case GREP_PATTERN: /* atom */
1080 case GREP_PATTERN_HEAD:
1081 case GREP_PATTERN_BODY:
1082 if (p->pcre1_regexp)
1083 free_pcre1_regexp(p);
1084 else if (p->pcre2_pattern)
1085 free_pcre2_pattern(p);
1086 else
1087 regfree(&p->regexp);
1088 free(p->pattern);
1089 break;
1090 default:
1091 break;
1093 free(p);
1096 if (!opt->extended)
1097 return;
1098 free_pattern_expr(opt->pattern_expression);
1101 static char *end_of_line(char *cp, unsigned long *left)
1103 unsigned long l = *left;
1104 while (l && *cp != '\n') {
1105 l--;
1106 cp++;
1108 *left = l;
1109 return cp;
1112 static int word_char(char ch)
1114 return isalnum(ch) || ch == '_';
1117 static void output_color(struct grep_opt *opt, const void *data, size_t size,
1118 const char *color)
1120 if (want_color(opt->color) && color && color[0]) {
1121 opt->output(opt, color, strlen(color));
1122 opt->output(opt, data, size);
1123 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
1124 } else
1125 opt->output(opt, data, size);
1128 static void output_sep(struct grep_opt *opt, char sign)
1130 if (opt->null_following_name)
1131 opt->output(opt, "\0", 1);
1132 else
1133 output_color(opt, &sign, 1, opt->colors[GREP_COLOR_SEP]);
1136 static void show_name(struct grep_opt *opt, const char *name)
1138 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1139 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
1142 static int patmatch(struct grep_pat *p, char *line, char *eol,
1143 regmatch_t *match, int eflags)
1145 int hit;
1147 if (p->pcre1_regexp)
1148 hit = !pcre1match(p, line, eol, match, eflags);
1149 else if (p->pcre2_pattern)
1150 hit = !pcre2match(p, line, eol, match, eflags);
1151 else
1152 hit = !regexec_buf(&p->regexp, line, eol - line, 1, match,
1153 eflags);
1155 return hit;
1158 static int strip_timestamp(char *bol, char **eol_p)
1160 char *eol = *eol_p;
1161 int ch;
1163 while (bol < --eol) {
1164 if (*eol != '>')
1165 continue;
1166 *eol_p = ++eol;
1167 ch = *eol;
1168 *eol = '\0';
1169 return ch;
1171 return 0;
1174 static struct {
1175 const char *field;
1176 size_t len;
1177 } header_field[] = {
1178 { "author ", 7 },
1179 { "committer ", 10 },
1180 { "reflog ", 7 },
1183 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
1184 enum grep_context ctx,
1185 regmatch_t *pmatch, int eflags)
1187 int hit = 0;
1188 int saved_ch = 0;
1189 const char *start = bol;
1191 if ((p->token != GREP_PATTERN) &&
1192 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
1193 return 0;
1195 if (p->token == GREP_PATTERN_HEAD) {
1196 const char *field;
1197 size_t len;
1198 assert(p->field < ARRAY_SIZE(header_field));
1199 field = header_field[p->field].field;
1200 len = header_field[p->field].len;
1201 if (strncmp(bol, field, len))
1202 return 0;
1203 bol += len;
1204 switch (p->field) {
1205 case GREP_HEADER_AUTHOR:
1206 case GREP_HEADER_COMMITTER:
1207 saved_ch = strip_timestamp(bol, &eol);
1208 break;
1209 default:
1210 break;
1214 again:
1215 hit = patmatch(p, bol, eol, pmatch, eflags);
1217 if (hit && p->word_regexp) {
1218 if ((pmatch[0].rm_so < 0) ||
1219 (eol - bol) < pmatch[0].rm_so ||
1220 (pmatch[0].rm_eo < 0) ||
1221 (eol - bol) < pmatch[0].rm_eo)
1222 die("regexp returned nonsense");
1224 /* Match beginning must be either beginning of the
1225 * line, or at word boundary (i.e. the last char must
1226 * not be a word char). Similarly, match end must be
1227 * either end of the line, or at word boundary
1228 * (i.e. the next char must not be a word char).
1230 if ( ((pmatch[0].rm_so == 0) ||
1231 !word_char(bol[pmatch[0].rm_so-1])) &&
1232 ((pmatch[0].rm_eo == (eol-bol)) ||
1233 !word_char(bol[pmatch[0].rm_eo])) )
1235 else
1236 hit = 0;
1238 /* Words consist of at least one character. */
1239 if (pmatch->rm_so == pmatch->rm_eo)
1240 hit = 0;
1242 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
1243 /* There could be more than one match on the
1244 * line, and the first match might not be
1245 * strict word match. But later ones could be!
1246 * Forward to the next possible start, i.e. the
1247 * next position following a non-word char.
1249 bol = pmatch[0].rm_so + bol + 1;
1250 while (word_char(bol[-1]) && bol < eol)
1251 bol++;
1252 eflags |= REG_NOTBOL;
1253 if (bol < eol)
1254 goto again;
1257 if (p->token == GREP_PATTERN_HEAD && saved_ch)
1258 *eol = saved_ch;
1259 if (hit) {
1260 pmatch[0].rm_so += bol - start;
1261 pmatch[0].rm_eo += bol - start;
1263 return hit;
1266 static int match_expr_eval(struct grep_opt *opt, struct grep_expr *x, char *bol,
1267 char *eol, enum grep_context ctx, ssize_t *col,
1268 ssize_t *icol, int collect_hits)
1270 int h = 0;
1272 if (!x)
1273 die("Not a valid grep expression");
1274 switch (x->node) {
1275 case GREP_NODE_TRUE:
1276 h = 1;
1277 break;
1278 case GREP_NODE_ATOM:
1280 regmatch_t tmp;
1281 h = match_one_pattern(x->u.atom, bol, eol, ctx,
1282 &tmp, 0);
1283 if (h && (*col < 0 || tmp.rm_so < *col))
1284 *col = tmp.rm_so;
1286 break;
1287 case GREP_NODE_NOT:
1289 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1291 h = !match_expr_eval(opt, x->u.unary, bol, eol, ctx, icol, col,
1293 break;
1294 case GREP_NODE_AND:
1295 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1296 icol, 0);
1297 if (h || opt->columnnum) {
1299 * Don't short-circuit AND when given --column, since a
1300 * NOT earlier in the tree may turn this into an OR. In
1301 * this case, see the below comment.
1303 h &= match_expr_eval(opt, x->u.binary.right, bol, eol,
1304 ctx, col, icol, 0);
1306 break;
1307 case GREP_NODE_OR:
1308 if (!(collect_hits || opt->columnnum)) {
1310 * Don't short-circuit OR when given --column (or
1311 * collecting hits) to ensure we don't skip a later
1312 * child that would produce an earlier match.
1314 return (match_expr_eval(opt, x->u.binary.left, bol, eol,
1315 ctx, col, icol, 0) ||
1316 match_expr_eval(opt, x->u.binary.right, bol,
1317 eol, ctx, col, icol, 0));
1319 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1320 icol, 0);
1321 if (collect_hits)
1322 x->u.binary.left->hit |= h;
1323 h |= match_expr_eval(opt, x->u.binary.right, bol, eol, ctx, col,
1324 icol, collect_hits);
1325 break;
1326 default:
1327 die("Unexpected node type (internal error) %d", x->node);
1329 if (collect_hits)
1330 x->hit |= h;
1331 return h;
1334 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
1335 enum grep_context ctx, ssize_t *col,
1336 ssize_t *icol, int collect_hits)
1338 struct grep_expr *x = opt->pattern_expression;
1339 return match_expr_eval(opt, x, bol, eol, ctx, col, icol, collect_hits);
1342 static int match_line(struct grep_opt *opt, char *bol, char *eol,
1343 ssize_t *col, ssize_t *icol,
1344 enum grep_context ctx, int collect_hits)
1346 struct grep_pat *p;
1347 int hit = 0;
1349 if (opt->extended)
1350 return match_expr(opt, bol, eol, ctx, col, icol,
1351 collect_hits);
1353 /* we do not call with collect_hits without being extended */
1354 for (p = opt->pattern_list; p; p = p->next) {
1355 regmatch_t tmp;
1356 if (match_one_pattern(p, bol, eol, ctx, &tmp, 0)) {
1357 hit |= 1;
1358 if (!opt->columnnum) {
1360 * Without --column, any single match on a line
1361 * is enough to know that it needs to be
1362 * printed. With --column, scan _all_ patterns
1363 * to find the earliest.
1365 break;
1367 if (*col < 0 || tmp.rm_so < *col)
1368 *col = tmp.rm_so;
1371 return hit;
1374 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
1375 enum grep_context ctx,
1376 regmatch_t *pmatch, int eflags)
1378 regmatch_t match;
1380 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
1381 return 0;
1382 if (match.rm_so < 0 || match.rm_eo < 0)
1383 return 0;
1384 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
1385 if (match.rm_so > pmatch->rm_so)
1386 return 1;
1387 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
1388 return 1;
1390 pmatch->rm_so = match.rm_so;
1391 pmatch->rm_eo = match.rm_eo;
1392 return 1;
1395 static int next_match(struct grep_opt *opt, char *bol, char *eol,
1396 enum grep_context ctx, regmatch_t *pmatch, int eflags)
1398 struct grep_pat *p;
1399 int hit = 0;
1401 pmatch->rm_so = pmatch->rm_eo = -1;
1402 if (bol < eol) {
1403 for (p = opt->pattern_list; p; p = p->next) {
1404 switch (p->token) {
1405 case GREP_PATTERN: /* atom */
1406 case GREP_PATTERN_HEAD:
1407 case GREP_PATTERN_BODY:
1408 hit |= match_next_pattern(p, bol, eol, ctx,
1409 pmatch, eflags);
1410 break;
1411 default:
1412 break;
1416 return hit;
1419 static void show_line_header(struct grep_opt *opt, const char *name,
1420 unsigned lno, ssize_t cno, char sign)
1422 if (opt->heading && opt->last_shown == 0) {
1423 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1424 opt->output(opt, "\n", 1);
1426 opt->last_shown = lno;
1428 if (!opt->heading && opt->pathname) {
1429 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1430 output_sep(opt, sign);
1432 if (opt->linenum) {
1433 char buf[32];
1434 xsnprintf(buf, sizeof(buf), "%d", lno);
1435 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_LINENO]);
1436 output_sep(opt, sign);
1439 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1440 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1441 * being called with a context line.
1443 if (opt->columnnum && cno) {
1444 char buf[32];
1445 xsnprintf(buf, sizeof(buf), "%"PRIuMAX, (uintmax_t)cno);
1446 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_COLUMNNO]);
1447 output_sep(opt, sign);
1451 static void show_line(struct grep_opt *opt, char *bol, char *eol,
1452 const char *name, unsigned lno, ssize_t cno, char sign)
1454 int rest = eol - bol;
1455 const char *match_color = NULL;
1456 const char *line_color = NULL;
1458 if (opt->file_break && opt->last_shown == 0) {
1459 if (opt->show_hunk_mark)
1460 opt->output(opt, "\n", 1);
1461 } else if (opt->pre_context || opt->post_context || opt->funcbody) {
1462 if (opt->last_shown == 0) {
1463 if (opt->show_hunk_mark) {
1464 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1465 opt->output(opt, "\n", 1);
1467 } else if (lno > opt->last_shown + 1) {
1468 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1469 opt->output(opt, "\n", 1);
1472 if (!opt->only_matching) {
1474 * In case the line we're being called with contains more than
1475 * one match, leave printing each header to the loop below.
1477 show_line_header(opt, name, lno, cno, sign);
1479 if (opt->color || opt->only_matching) {
1480 regmatch_t match;
1481 enum grep_context ctx = GREP_CONTEXT_BODY;
1482 int ch = *eol;
1483 int eflags = 0;
1485 if (opt->color) {
1486 if (sign == ':')
1487 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
1488 else
1489 match_color = opt->colors[GREP_COLOR_MATCH_CONTEXT];
1490 if (sign == ':')
1491 line_color = opt->colors[GREP_COLOR_SELECTED];
1492 else if (sign == '-')
1493 line_color = opt->colors[GREP_COLOR_CONTEXT];
1494 else if (sign == '=')
1495 line_color = opt->colors[GREP_COLOR_FUNCTION];
1497 *eol = '\0';
1498 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
1499 if (match.rm_so == match.rm_eo)
1500 break;
1502 if (opt->only_matching)
1503 show_line_header(opt, name, lno, cno, sign);
1504 else
1505 output_color(opt, bol, match.rm_so, line_color);
1506 output_color(opt, bol + match.rm_so,
1507 match.rm_eo - match.rm_so, match_color);
1508 if (opt->only_matching)
1509 opt->output(opt, "\n", 1);
1510 bol += match.rm_eo;
1511 cno += match.rm_eo;
1512 rest -= match.rm_eo;
1513 eflags = REG_NOTBOL;
1515 *eol = ch;
1517 if (!opt->only_matching) {
1518 output_color(opt, bol, rest, line_color);
1519 opt->output(opt, "\n", 1);
1523 int grep_use_locks;
1526 * This lock protects access to the gitattributes machinery, which is
1527 * not thread-safe.
1529 pthread_mutex_t grep_attr_mutex;
1531 static inline void grep_attr_lock(void)
1533 if (grep_use_locks)
1534 pthread_mutex_lock(&grep_attr_mutex);
1537 static inline void grep_attr_unlock(void)
1539 if (grep_use_locks)
1540 pthread_mutex_unlock(&grep_attr_mutex);
1543 static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
1545 xdemitconf_t *xecfg = opt->priv;
1546 if (xecfg && !xecfg->find_func) {
1547 grep_source_load_driver(gs, opt->repo->index);
1548 if (gs->driver->funcname.pattern) {
1549 const struct userdiff_funcname *pe = &gs->driver->funcname;
1550 xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
1551 } else {
1552 xecfg = opt->priv = NULL;
1556 if (xecfg) {
1557 char buf[1];
1558 return xecfg->find_func(bol, eol - bol, buf, 1,
1559 xecfg->find_func_priv) >= 0;
1562 if (bol == eol)
1563 return 0;
1564 if (isalpha(*bol) || *bol == '_' || *bol == '$')
1565 return 1;
1566 return 0;
1569 static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
1570 char *bol, unsigned lno)
1572 while (bol > gs->buf) {
1573 char *eol = --bol;
1575 while (bol > gs->buf && bol[-1] != '\n')
1576 bol--;
1577 lno--;
1579 if (lno <= opt->last_shown)
1580 break;
1582 if (match_funcname(opt, gs, bol, eol)) {
1583 show_line(opt, bol, eol, gs->name, lno, 0, '=');
1584 break;
1589 static int is_empty_line(const char *bol, const char *eol);
1591 static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
1592 char *bol, char *end, unsigned lno)
1594 unsigned cur = lno, from = 1, funcname_lno = 0, orig_from;
1595 int funcname_needed = !!opt->funcname, comment_needed = 0;
1597 if (opt->pre_context < lno)
1598 from = lno - opt->pre_context;
1599 if (from <= opt->last_shown)
1600 from = opt->last_shown + 1;
1601 orig_from = from;
1602 if (opt->funcbody) {
1603 if (match_funcname(opt, gs, bol, end))
1604 comment_needed = 1;
1605 else
1606 funcname_needed = 1;
1607 from = opt->last_shown + 1;
1610 /* Rewind. */
1611 while (bol > gs->buf && cur > from) {
1612 char *next_bol = bol;
1613 char *eol = --bol;
1615 while (bol > gs->buf && bol[-1] != '\n')
1616 bol--;
1617 cur--;
1618 if (comment_needed && (is_empty_line(bol, eol) ||
1619 match_funcname(opt, gs, bol, eol))) {
1620 comment_needed = 0;
1621 from = orig_from;
1622 if (cur < from) {
1623 cur++;
1624 bol = next_bol;
1625 break;
1628 if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
1629 funcname_lno = cur;
1630 funcname_needed = 0;
1631 if (opt->funcbody)
1632 comment_needed = 1;
1633 else
1634 from = orig_from;
1638 /* We need to look even further back to find a function signature. */
1639 if (opt->funcname && funcname_needed)
1640 show_funcname_line(opt, gs, bol, cur);
1642 /* Back forward. */
1643 while (cur < lno) {
1644 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
1646 while (*eol != '\n')
1647 eol++;
1648 show_line(opt, bol, eol, gs->name, cur, 0, sign);
1649 bol = eol + 1;
1650 cur++;
1654 static int should_lookahead(struct grep_opt *opt)
1656 struct grep_pat *p;
1658 if (opt->extended)
1659 return 0; /* punt for too complex stuff */
1660 if (opt->invert)
1661 return 0;
1662 for (p = opt->pattern_list; p; p = p->next) {
1663 if (p->token != GREP_PATTERN)
1664 return 0; /* punt for "header only" and stuff */
1666 return 1;
1669 static int look_ahead(struct grep_opt *opt,
1670 unsigned long *left_p,
1671 unsigned *lno_p,
1672 char **bol_p)
1674 unsigned lno = *lno_p;
1675 char *bol = *bol_p;
1676 struct grep_pat *p;
1677 char *sp, *last_bol;
1678 regoff_t earliest = -1;
1680 for (p = opt->pattern_list; p; p = p->next) {
1681 int hit;
1682 regmatch_t m;
1684 hit = patmatch(p, bol, bol + *left_p, &m, 0);
1685 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
1686 continue;
1687 if (earliest < 0 || m.rm_so < earliest)
1688 earliest = m.rm_so;
1691 if (earliest < 0) {
1692 *bol_p = bol + *left_p;
1693 *left_p = 0;
1694 return 1;
1696 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
1697 ; /* find the beginning of the line */
1698 last_bol = sp;
1700 for (sp = bol; sp < last_bol; sp++) {
1701 if (*sp == '\n')
1702 lno++;
1704 *left_p -= last_bol - bol;
1705 *bol_p = last_bol;
1706 *lno_p = lno;
1707 return 0;
1710 static int fill_textconv_grep(struct repository *r,
1711 struct userdiff_driver *driver,
1712 struct grep_source *gs)
1714 struct diff_filespec *df;
1715 char *buf;
1716 size_t size;
1718 if (!driver || !driver->textconv)
1719 return grep_source_load(gs);
1722 * The textconv interface is intimately tied to diff_filespecs, so we
1723 * have to pretend to be one. If we could unify the grep_source
1724 * and diff_filespec structs, this mess could just go away.
1726 df = alloc_filespec(gs->path);
1727 switch (gs->type) {
1728 case GREP_SOURCE_OID:
1729 fill_filespec(df, gs->identifier, 1, 0100644);
1730 break;
1731 case GREP_SOURCE_FILE:
1732 fill_filespec(df, &null_oid, 0, 0100644);
1733 break;
1734 default:
1735 BUG("attempt to textconv something without a path?");
1739 * fill_textconv is not remotely thread-safe; it modifies the global
1740 * diff tempfile structure, writes to the_repo's odb and might
1741 * internally call thread-unsafe functions such as the
1742 * prepare_packed_git() lazy-initializator. Because of the last two, we
1743 * must ensure mutual exclusion between this call and the object reading
1744 * API, thus we use obj_read_lock() here.
1746 * TODO: allowing text conversion to run in parallel with object
1747 * reading operations might increase performance in the multithreaded
1748 * non-worktreee git-grep with --textconv.
1750 obj_read_lock();
1751 size = fill_textconv(r, driver, df, &buf);
1752 obj_read_unlock();
1753 free_filespec(df);
1756 * The normal fill_textconv usage by the diff machinery would just keep
1757 * the textconv'd buf separate from the diff_filespec. But much of the
1758 * grep code passes around a grep_source and assumes that its "buf"
1759 * pointer is the beginning of the thing we are searching. So let's
1760 * install our textconv'd version into the grep_source, taking care not
1761 * to leak any existing buffer.
1763 grep_source_clear_data(gs);
1764 gs->buf = buf;
1765 gs->size = size;
1767 return 0;
1770 static int is_empty_line(const char *bol, const char *eol)
1772 while (bol < eol && isspace(*bol))
1773 bol++;
1774 return bol == eol;
1777 static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
1779 char *bol;
1780 char *peek_bol = NULL;
1781 unsigned long left;
1782 unsigned lno = 1;
1783 unsigned last_hit = 0;
1784 int binary_match_only = 0;
1785 unsigned count = 0;
1786 int try_lookahead = 0;
1787 int show_function = 0;
1788 struct userdiff_driver *textconv = NULL;
1789 enum grep_context ctx = GREP_CONTEXT_HEAD;
1790 xdemitconf_t xecfg;
1792 if (!opt->status_only && gs->name == NULL)
1793 BUG("grep call which could print a name requires "
1794 "grep_source.name be non-NULL");
1796 if (!opt->output)
1797 opt->output = std_output;
1799 if (opt->pre_context || opt->post_context || opt->file_break ||
1800 opt->funcbody) {
1801 /* Show hunk marks, except for the first file. */
1802 if (opt->last_shown)
1803 opt->show_hunk_mark = 1;
1805 * If we're using threads then we can't easily identify
1806 * the first file. Always put hunk marks in that case
1807 * and skip the very first one later in work_done().
1809 if (opt->output != std_output)
1810 opt->show_hunk_mark = 1;
1812 opt->last_shown = 0;
1814 if (opt->allow_textconv) {
1815 grep_source_load_driver(gs, opt->repo->index);
1817 * We might set up the shared textconv cache data here, which
1818 * is not thread-safe. Also, get_oid_with_context() and
1819 * parse_object() might be internally called. As they are not
1820 * currenty thread-safe and might be racy with object reading,
1821 * obj_read_lock() must be called.
1823 grep_attr_lock();
1824 obj_read_lock();
1825 textconv = userdiff_get_textconv(opt->repo, gs->driver);
1826 obj_read_unlock();
1827 grep_attr_unlock();
1831 * We know the result of a textconv is text, so we only have to care
1832 * about binary handling if we are not using it.
1834 if (!textconv) {
1835 switch (opt->binary) {
1836 case GREP_BINARY_DEFAULT:
1837 if (grep_source_is_binary(gs, opt->repo->index))
1838 binary_match_only = 1;
1839 break;
1840 case GREP_BINARY_NOMATCH:
1841 if (grep_source_is_binary(gs, opt->repo->index))
1842 return 0; /* Assume unmatch */
1843 break;
1844 case GREP_BINARY_TEXT:
1845 break;
1846 default:
1847 BUG("unknown binary handling mode");
1851 memset(&xecfg, 0, sizeof(xecfg));
1852 opt->priv = &xecfg;
1854 try_lookahead = should_lookahead(opt);
1856 if (fill_textconv_grep(opt->repo, textconv, gs) < 0)
1857 return 0;
1859 bol = gs->buf;
1860 left = gs->size;
1861 while (left) {
1862 char *eol, ch;
1863 int hit;
1864 ssize_t cno;
1865 ssize_t col = -1, icol = -1;
1868 * look_ahead() skips quickly to the line that possibly
1869 * has the next hit; don't call it if we need to do
1870 * something more than just skipping the current line
1871 * in response to an unmatch for the current line. E.g.
1872 * inside a post-context window, we will show the current
1873 * line as a context around the previous hit when it
1874 * doesn't hit.
1876 if (try_lookahead
1877 && !(last_hit
1878 && (show_function ||
1879 lno <= last_hit + opt->post_context))
1880 && look_ahead(opt, &left, &lno, &bol))
1881 break;
1882 eol = end_of_line(bol, &left);
1883 ch = *eol;
1884 *eol = 0;
1886 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
1887 ctx = GREP_CONTEXT_BODY;
1889 hit = match_line(opt, bol, eol, &col, &icol, ctx, collect_hits);
1890 *eol = ch;
1892 if (collect_hits)
1893 goto next_line;
1895 /* "grep -v -e foo -e bla" should list lines
1896 * that do not have either, so inversion should
1897 * be done outside.
1899 if (opt->invert)
1900 hit = !hit;
1901 if (opt->unmatch_name_only) {
1902 if (hit)
1903 return 0;
1904 goto next_line;
1906 if (hit) {
1907 count++;
1908 if (opt->status_only)
1909 return 1;
1910 if (opt->name_only) {
1911 show_name(opt, gs->name);
1912 return 1;
1914 if (opt->count)
1915 goto next_line;
1916 if (binary_match_only) {
1917 opt->output(opt, "Binary file ", 12);
1918 output_color(opt, gs->name, strlen(gs->name),
1919 opt->colors[GREP_COLOR_FILENAME]);
1920 opt->output(opt, " matches\n", 9);
1921 return 1;
1923 /* Hit at this line. If we haven't shown the
1924 * pre-context lines, we would need to show them.
1926 if (opt->pre_context || opt->funcbody)
1927 show_pre_context(opt, gs, bol, eol, lno);
1928 else if (opt->funcname)
1929 show_funcname_line(opt, gs, bol, lno);
1930 cno = opt->invert ? icol : col;
1931 if (cno < 0) {
1933 * A negative cno indicates that there was no
1934 * match on the line. We are thus inverted and
1935 * being asked to show all lines that _don't_
1936 * match a given expression. Therefore, set cno
1937 * to 0 to suggest the whole line matches.
1939 cno = 0;
1941 show_line(opt, bol, eol, gs->name, lno, cno + 1, ':');
1942 last_hit = lno;
1943 if (opt->funcbody)
1944 show_function = 1;
1945 goto next_line;
1947 if (show_function && (!peek_bol || peek_bol < bol)) {
1948 unsigned long peek_left = left;
1949 char *peek_eol = eol;
1952 * Trailing empty lines are not interesting.
1953 * Peek past them to see if they belong to the
1954 * body of the current function.
1956 peek_bol = bol;
1957 while (is_empty_line(peek_bol, peek_eol)) {
1958 peek_bol = peek_eol + 1;
1959 peek_eol = end_of_line(peek_bol, &peek_left);
1962 if (match_funcname(opt, gs, peek_bol, peek_eol))
1963 show_function = 0;
1965 if (show_function ||
1966 (last_hit && lno <= last_hit + opt->post_context)) {
1967 /* If the last hit is within the post context,
1968 * we need to show this line.
1970 show_line(opt, bol, eol, gs->name, lno, col + 1, '-');
1973 next_line:
1974 bol = eol + 1;
1975 if (!left)
1976 break;
1977 left--;
1978 lno++;
1981 if (collect_hits)
1982 return 0;
1984 if (opt->status_only)
1985 return opt->unmatch_name_only;
1986 if (opt->unmatch_name_only) {
1987 /* We did not see any hit, so we want to show this */
1988 show_name(opt, gs->name);
1989 return 1;
1992 xdiff_clear_find_func(&xecfg);
1993 opt->priv = NULL;
1995 /* NEEDSWORK:
1996 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1997 * which feels mostly useless but sometimes useful. Maybe
1998 * make it another option? For now suppress them.
2000 if (opt->count && count) {
2001 char buf[32];
2002 if (opt->pathname) {
2003 output_color(opt, gs->name, strlen(gs->name),
2004 opt->colors[GREP_COLOR_FILENAME]);
2005 output_sep(opt, ':');
2007 xsnprintf(buf, sizeof(buf), "%u\n", count);
2008 opt->output(opt, buf, strlen(buf));
2009 return 1;
2011 return !!last_hit;
2014 static void clr_hit_marker(struct grep_expr *x)
2016 /* All-hit markers are meaningful only at the very top level
2017 * OR node.
2019 while (1) {
2020 x->hit = 0;
2021 if (x->node != GREP_NODE_OR)
2022 return;
2023 x->u.binary.left->hit = 0;
2024 x = x->u.binary.right;
2028 static int chk_hit_marker(struct grep_expr *x)
2030 /* Top level nodes have hit markers. See if they all are hits */
2031 while (1) {
2032 if (x->node != GREP_NODE_OR)
2033 return x->hit;
2034 if (!x->u.binary.left->hit)
2035 return 0;
2036 x = x->u.binary.right;
2040 int grep_source(struct grep_opt *opt, struct grep_source *gs)
2043 * we do not have to do the two-pass grep when we do not check
2044 * buffer-wide "all-match".
2046 if (!opt->all_match)
2047 return grep_source_1(opt, gs, 0);
2049 /* Otherwise the toplevel "or" terms hit a bit differently.
2050 * We first clear hit markers from them.
2052 clr_hit_marker(opt->pattern_expression);
2053 grep_source_1(opt, gs, 1);
2055 if (!chk_hit_marker(opt->pattern_expression))
2056 return 0;
2058 return grep_source_1(opt, gs, 0);
2061 int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
2063 struct grep_source gs;
2064 int r;
2066 grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
2067 gs.buf = buf;
2068 gs.size = size;
2070 r = grep_source(opt, &gs);
2072 grep_source_clear(&gs);
2073 return r;
2076 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
2077 const char *name, const char *path,
2078 const void *identifier)
2080 gs->type = type;
2081 gs->name = xstrdup_or_null(name);
2082 gs->path = xstrdup_or_null(path);
2083 gs->buf = NULL;
2084 gs->size = 0;
2085 gs->driver = NULL;
2087 switch (type) {
2088 case GREP_SOURCE_FILE:
2089 gs->identifier = xstrdup(identifier);
2090 break;
2091 case GREP_SOURCE_OID:
2092 gs->identifier = oiddup(identifier);
2093 break;
2094 case GREP_SOURCE_BUF:
2095 gs->identifier = NULL;
2096 break;
2100 void grep_source_clear(struct grep_source *gs)
2102 FREE_AND_NULL(gs->name);
2103 FREE_AND_NULL(gs->path);
2104 FREE_AND_NULL(gs->identifier);
2105 grep_source_clear_data(gs);
2108 void grep_source_clear_data(struct grep_source *gs)
2110 switch (gs->type) {
2111 case GREP_SOURCE_FILE:
2112 case GREP_SOURCE_OID:
2113 FREE_AND_NULL(gs->buf);
2114 gs->size = 0;
2115 break;
2116 case GREP_SOURCE_BUF:
2117 /* leave user-provided buf intact */
2118 break;
2122 static int grep_source_load_oid(struct grep_source *gs)
2124 enum object_type type;
2126 gs->buf = read_object_file(gs->identifier, &type, &gs->size);
2127 if (!gs->buf)
2128 return error(_("'%s': unable to read %s"),
2129 gs->name,
2130 oid_to_hex(gs->identifier));
2131 return 0;
2134 static int grep_source_load_file(struct grep_source *gs)
2136 const char *filename = gs->identifier;
2137 struct stat st;
2138 char *data;
2139 size_t size;
2140 int i;
2142 if (lstat(filename, &st) < 0) {
2143 err_ret:
2144 if (errno != ENOENT)
2145 error_errno(_("failed to stat '%s'"), filename);
2146 return -1;
2148 if (!S_ISREG(st.st_mode))
2149 return -1;
2150 size = xsize_t(st.st_size);
2151 i = open(filename, O_RDONLY);
2152 if (i < 0)
2153 goto err_ret;
2154 data = xmallocz(size);
2155 if (st.st_size != read_in_full(i, data, size)) {
2156 error_errno(_("'%s': short read"), filename);
2157 close(i);
2158 free(data);
2159 return -1;
2161 close(i);
2163 gs->buf = data;
2164 gs->size = size;
2165 return 0;
2168 static int grep_source_load(struct grep_source *gs)
2170 if (gs->buf)
2171 return 0;
2173 switch (gs->type) {
2174 case GREP_SOURCE_FILE:
2175 return grep_source_load_file(gs);
2176 case GREP_SOURCE_OID:
2177 return grep_source_load_oid(gs);
2178 case GREP_SOURCE_BUF:
2179 return gs->buf ? 0 : -1;
2181 BUG("invalid grep_source type to load");
2184 void grep_source_load_driver(struct grep_source *gs,
2185 struct index_state *istate)
2187 if (gs->driver)
2188 return;
2190 grep_attr_lock();
2191 if (gs->path)
2192 gs->driver = userdiff_find_by_path(istate, gs->path);
2193 if (!gs->driver)
2194 gs->driver = userdiff_find_by_name("default");
2195 grep_attr_unlock();
2198 static int grep_source_is_binary(struct grep_source *gs,
2199 struct index_state *istate)
2201 grep_source_load_driver(gs, istate);
2202 if (gs->driver->binary != -1)
2203 return gs->driver->binary;
2205 if (!grep_source_load(gs))
2206 return buffer_is_binary(gs->buf, gs->size);
2208 return 0;