packfile: ensure that enum object_type is defined
[git/gitster.git] / grep.c
blobcd7fc6f66cadfcb38ad73f171b97a419610aad37
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);
16 static struct grep_opt grep_defaults;
18 static const char *color_grep_slots[] = {
19 [GREP_COLOR_CONTEXT] = "context",
20 [GREP_COLOR_FILENAME] = "filename",
21 [GREP_COLOR_FUNCTION] = "function",
22 [GREP_COLOR_LINENO] = "lineNumber",
23 [GREP_COLOR_COLUMNNO] = "column",
24 [GREP_COLOR_MATCH_CONTEXT] = "matchContext",
25 [GREP_COLOR_MATCH_SELECTED] = "matchSelected",
26 [GREP_COLOR_SELECTED] = "selected",
27 [GREP_COLOR_SEP] = "separator",
30 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
32 fwrite(buf, size, 1, stdout);
35 static void color_set(char *dst, const char *color_bytes)
37 xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes);
41 * Initialize the grep_defaults template with hardcoded defaults.
42 * We could let the compiler do this, but without C99 initializers
43 * the code gets unwieldy and unreadable, so...
45 void init_grep_defaults(void)
47 struct grep_opt *opt = &grep_defaults;
48 static int run_once;
50 if (run_once)
51 return;
52 run_once++;
54 memset(opt, 0, sizeof(*opt));
55 opt->relative = 1;
56 opt->pathname = 1;
57 opt->max_depth = -1;
58 opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
59 color_set(opt->colors[GREP_COLOR_CONTEXT], "");
60 color_set(opt->colors[GREP_COLOR_FILENAME], "");
61 color_set(opt->colors[GREP_COLOR_FUNCTION], "");
62 color_set(opt->colors[GREP_COLOR_LINENO], "");
63 color_set(opt->colors[GREP_COLOR_COLUMNNO], "");
64 color_set(opt->colors[GREP_COLOR_MATCH_CONTEXT], GIT_COLOR_BOLD_RED);
65 color_set(opt->colors[GREP_COLOR_MATCH_SELECTED], GIT_COLOR_BOLD_RED);
66 color_set(opt->colors[GREP_COLOR_SELECTED], "");
67 color_set(opt->colors[GREP_COLOR_SEP], GIT_COLOR_CYAN);
68 opt->color = -1;
69 opt->output = std_output;
72 static int parse_pattern_type_arg(const char *opt, const char *arg)
74 if (!strcmp(arg, "default"))
75 return GREP_PATTERN_TYPE_UNSPECIFIED;
76 else if (!strcmp(arg, "basic"))
77 return GREP_PATTERN_TYPE_BRE;
78 else if (!strcmp(arg, "extended"))
79 return GREP_PATTERN_TYPE_ERE;
80 else if (!strcmp(arg, "fixed"))
81 return GREP_PATTERN_TYPE_FIXED;
82 else if (!strcmp(arg, "perl"))
83 return GREP_PATTERN_TYPE_PCRE;
84 die("bad %s argument: %s", opt, arg);
87 define_list_config_array_extra(color_grep_slots, {"match"});
90 * Read the configuration file once and store it in
91 * the grep_defaults template.
93 int grep_config(const char *var, const char *value, void *cb)
95 struct grep_opt *opt = &grep_defaults;
96 const char *slot;
98 if (userdiff_config(var, value) < 0)
99 return -1;
101 if (!strcmp(var, "grep.extendedregexp")) {
102 opt->extended_regexp_option = git_config_bool(var, value);
103 return 0;
106 if (!strcmp(var, "grep.patterntype")) {
107 opt->pattern_type_option = parse_pattern_type_arg(var, value);
108 return 0;
111 if (!strcmp(var, "grep.linenumber")) {
112 opt->linenum = git_config_bool(var, value);
113 return 0;
115 if (!strcmp(var, "grep.column")) {
116 opt->columnnum = git_config_bool(var, value);
117 return 0;
120 if (!strcmp(var, "grep.fullname")) {
121 opt->relative = !git_config_bool(var, value);
122 return 0;
125 if (!strcmp(var, "color.grep"))
126 opt->color = git_config_colorbool(var, value);
127 if (!strcmp(var, "color.grep.match")) {
128 if (grep_config("color.grep.matchcontext", value, cb) < 0)
129 return -1;
130 if (grep_config("color.grep.matchselected", value, cb) < 0)
131 return -1;
132 } else if (skip_prefix(var, "color.grep.", &slot)) {
133 int i = LOOKUP_CONFIG(color_grep_slots, slot);
134 char *color;
136 if (i < 0)
137 return -1;
138 color = opt->colors[i];
139 if (!value)
140 return config_error_nonbool(var);
141 return color_parse(value, color);
143 return 0;
147 * Initialize one instance of grep_opt and copy the
148 * default values from the template we read the configuration
149 * information in an earlier call to git_config(grep_config).
151 void grep_init(struct grep_opt *opt, const char *prefix)
153 struct grep_opt *def = &grep_defaults;
154 int i;
156 memset(opt, 0, sizeof(*opt));
157 opt->prefix = prefix;
158 opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
159 opt->pattern_tail = &opt->pattern_list;
160 opt->header_tail = &opt->header_list;
162 opt->color = def->color;
163 opt->extended_regexp_option = def->extended_regexp_option;
164 opt->pattern_type_option = def->pattern_type_option;
165 opt->linenum = def->linenum;
166 opt->columnnum = def->columnnum;
167 opt->max_depth = def->max_depth;
168 opt->pathname = def->pathname;
169 opt->relative = def->relative;
170 opt->output = def->output;
172 for (i = 0; i < NR_GREP_COLORS; i++)
173 color_set(opt->colors[i], def->colors[i]);
176 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
179 * When committing to the pattern type by setting the relevant
180 * fields in grep_opt it's generally not necessary to zero out
181 * the fields we're not choosing, since they won't have been
182 * set by anything. The extended_regexp_option field is the
183 * only exception to this.
185 * This is because in the process of parsing grep.patternType
186 * & grep.extendedRegexp we set opt->pattern_type_option and
187 * opt->extended_regexp_option, respectively. We then
188 * internally use opt->extended_regexp_option to see if we're
189 * compiling an ERE. It must be unset if that's not actually
190 * the case.
192 if (pattern_type != GREP_PATTERN_TYPE_ERE &&
193 opt->extended_regexp_option)
194 opt->extended_regexp_option = 0;
196 switch (pattern_type) {
197 case GREP_PATTERN_TYPE_UNSPECIFIED:
198 /* fall through */
200 case GREP_PATTERN_TYPE_BRE:
201 break;
203 case GREP_PATTERN_TYPE_ERE:
204 opt->extended_regexp_option = 1;
205 break;
207 case GREP_PATTERN_TYPE_FIXED:
208 opt->fixed = 1;
209 break;
211 case GREP_PATTERN_TYPE_PCRE:
212 #ifdef USE_LIBPCRE2
213 opt->pcre2 = 1;
214 #else
216 * It's important that pcre1 always be assigned to
217 * even when there's no USE_LIBPCRE* defined. We still
218 * call the PCRE stub function, it just dies with
219 * "cannot use Perl-compatible regexes[...]".
221 opt->pcre1 = 1;
222 #endif
223 break;
227 void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
229 if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
230 grep_set_pattern_type_option(pattern_type, opt);
231 else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
232 grep_set_pattern_type_option(opt->pattern_type_option, opt);
233 else if (opt->extended_regexp_option)
235 * This branch *must* happen after setting from the
236 * opt->pattern_type_option above, we don't want
237 * grep.extendedRegexp to override grep.patternType!
239 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
242 static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
243 const char *origin, int no,
244 enum grep_pat_token t,
245 enum grep_header_field field)
247 struct grep_pat *p = xcalloc(1, sizeof(*p));
248 p->pattern = xmemdupz(pat, patlen);
249 p->patternlen = patlen;
250 p->origin = origin;
251 p->no = no;
252 p->token = t;
253 p->field = field;
254 return p;
257 static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
259 **tail = p;
260 *tail = &p->next;
261 p->next = NULL;
263 switch (p->token) {
264 case GREP_PATTERN: /* atom */
265 case GREP_PATTERN_HEAD:
266 case GREP_PATTERN_BODY:
267 for (;;) {
268 struct grep_pat *new_pat;
269 size_t len = 0;
270 char *cp = p->pattern + p->patternlen, *nl = NULL;
271 while (++len <= p->patternlen) {
272 if (*(--cp) == '\n') {
273 nl = cp;
274 break;
277 if (!nl)
278 break;
279 new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
280 p->no, p->token, p->field);
281 new_pat->next = p->next;
282 if (!p->next)
283 *tail = &new_pat->next;
284 p->next = new_pat;
285 *nl = '\0';
286 p->patternlen -= len;
288 break;
289 default:
290 break;
294 void append_header_grep_pattern(struct grep_opt *opt,
295 enum grep_header_field field, const char *pat)
297 struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
298 GREP_PATTERN_HEAD, field);
299 if (field == GREP_HEADER_REFLOG)
300 opt->use_reflog_filter = 1;
301 do_append_grep_pat(&opt->header_tail, p);
304 void append_grep_pattern(struct grep_opt *opt, const char *pat,
305 const char *origin, int no, enum grep_pat_token t)
307 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
310 void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
311 const char *origin, int no, enum grep_pat_token t)
313 struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
314 do_append_grep_pat(&opt->pattern_tail, p);
317 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
319 struct grep_pat *pat;
320 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
321 *ret = *opt;
323 ret->pattern_list = NULL;
324 ret->pattern_tail = &ret->pattern_list;
326 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
328 if(pat->token == GREP_PATTERN_HEAD)
329 append_header_grep_pattern(ret, pat->field,
330 pat->pattern);
331 else
332 append_grep_pat(ret, pat->pattern, pat->patternlen,
333 pat->origin, pat->no, pat->token);
336 return ret;
339 static NORETURN void compile_regexp_failed(const struct grep_pat *p,
340 const char *error)
342 char where[1024];
344 if (p->no)
345 xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
346 else if (p->origin)
347 xsnprintf(where, sizeof(where), "%s, ", p->origin);
348 else
349 where[0] = 0;
351 die("%s'%s': %s", where, p->pattern, error);
354 static int is_fixed(const char *s, size_t len)
356 size_t i;
358 for (i = 0; i < len; i++) {
359 if (is_regex_special(s[i]))
360 return 0;
363 return 1;
366 static int has_null(const char *s, size_t len)
369 * regcomp cannot accept patterns with NULs so when using it
370 * we consider any pattern containing a NUL fixed.
372 if (memchr(s, 0, len))
373 return 1;
375 return 0;
378 #ifdef USE_LIBPCRE1
379 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
381 const char *error;
382 int erroffset;
383 int options = PCRE_MULTILINE;
385 if (opt->ignore_case) {
386 if (has_non_ascii(p->pattern))
387 p->pcre1_tables = pcre_maketables();
388 options |= PCRE_CASELESS;
390 if (is_utf8_locale() && has_non_ascii(p->pattern))
391 options |= PCRE_UTF8;
393 p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
394 p->pcre1_tables);
395 if (!p->pcre1_regexp)
396 compile_regexp_failed(p, error);
398 p->pcre1_extra_info = pcre_study(p->pcre1_regexp, GIT_PCRE_STUDY_JIT_COMPILE, &error);
399 if (!p->pcre1_extra_info && error)
400 die("%s", error);
402 #ifdef GIT_PCRE1_USE_JIT
403 pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on);
404 if (p->pcre1_jit_on == 1) {
405 p->pcre1_jit_stack = pcre_jit_stack_alloc(1, 1024 * 1024);
406 if (!p->pcre1_jit_stack)
407 die("Couldn't allocate PCRE JIT stack");
408 pcre_assign_jit_stack(p->pcre1_extra_info, NULL, p->pcre1_jit_stack);
409 } else if (p->pcre1_jit_on != 0) {
410 BUG("The pcre1_jit_on variable should be 0 or 1, not %d",
411 p->pcre1_jit_on);
413 #endif
416 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
417 regmatch_t *match, int eflags)
419 int ovector[30], ret, flags = 0;
421 if (eflags & REG_NOTBOL)
422 flags |= PCRE_NOTBOL;
424 #ifdef GIT_PCRE1_USE_JIT
425 if (p->pcre1_jit_on) {
426 ret = pcre_jit_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
427 eol - line, 0, flags, ovector,
428 ARRAY_SIZE(ovector), p->pcre1_jit_stack);
429 } else
430 #endif
432 ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
433 eol - line, 0, flags, ovector,
434 ARRAY_SIZE(ovector));
437 if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
438 die("pcre_exec failed with error code %d", ret);
439 if (ret > 0) {
440 ret = 0;
441 match->rm_so = ovector[0];
442 match->rm_eo = ovector[1];
445 return ret;
448 static void free_pcre1_regexp(struct grep_pat *p)
450 pcre_free(p->pcre1_regexp);
451 #ifdef GIT_PCRE1_USE_JIT
452 if (p->pcre1_jit_on) {
453 pcre_free_study(p->pcre1_extra_info);
454 pcre_jit_stack_free(p->pcre1_jit_stack);
455 } else
456 #endif
458 pcre_free(p->pcre1_extra_info);
460 pcre_free((void *)p->pcre1_tables);
462 #else /* !USE_LIBPCRE1 */
463 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
465 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
468 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
469 regmatch_t *match, int eflags)
471 return 1;
474 static void free_pcre1_regexp(struct grep_pat *p)
477 #endif /* !USE_LIBPCRE1 */
479 #ifdef USE_LIBPCRE2
480 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
482 int error;
483 PCRE2_UCHAR errbuf[256];
484 PCRE2_SIZE erroffset;
485 int options = PCRE2_MULTILINE;
486 const uint8_t *character_tables = NULL;
487 int jitret;
488 int patinforet;
489 size_t jitsizearg;
491 assert(opt->pcre2);
493 p->pcre2_compile_context = NULL;
495 if (opt->ignore_case) {
496 if (has_non_ascii(p->pattern)) {
497 character_tables = pcre2_maketables(NULL);
498 p->pcre2_compile_context = pcre2_compile_context_create(NULL);
499 pcre2_set_character_tables(p->pcre2_compile_context, character_tables);
501 options |= PCRE2_CASELESS;
503 if (is_utf8_locale() && has_non_ascii(p->pattern))
504 options |= PCRE2_UTF;
506 p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern,
507 p->patternlen, options, &error, &erroffset,
508 p->pcre2_compile_context);
510 if (p->pcre2_pattern) {
511 p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL);
512 if (!p->pcre2_match_data)
513 die("Couldn't allocate PCRE2 match data");
514 } else {
515 pcre2_get_error_message(error, errbuf, sizeof(errbuf));
516 compile_regexp_failed(p, (const char *)&errbuf);
519 pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on);
520 if (p->pcre2_jit_on == 1) {
521 jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE);
522 if (jitret)
523 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret);
526 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
527 * tells us whether the library itself supports JIT,
528 * but to see whether we're going to be actually using
529 * JIT we need to extract PCRE2_INFO_JITSIZE from the
530 * pattern *after* we do pcre2_jit_compile() above.
532 * This is because if the pattern contains the
533 * (*NO_JIT) verb (see pcre2syntax(3))
534 * pcre2_jit_compile() will exit early with 0. If we
535 * then proceed to call pcre2_jit_match() further down
536 * the line instead of pcre2_match() we'll either
537 * segfault (pre PCRE 10.31) or run into a fatal error
538 * (post PCRE2 10.31)
540 patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg);
541 if (patinforet)
542 BUG("pcre2_pattern_info() failed: %d", patinforet);
543 if (jitsizearg == 0) {
544 p->pcre2_jit_on = 0;
545 return;
548 p->pcre2_jit_stack = pcre2_jit_stack_create(1, 1024 * 1024, NULL);
549 if (!p->pcre2_jit_stack)
550 die("Couldn't allocate PCRE2 JIT stack");
551 p->pcre2_match_context = pcre2_match_context_create(NULL);
552 if (!p->pcre2_match_context)
553 die("Couldn't allocate PCRE2 match context");
554 pcre2_jit_stack_assign(p->pcre2_match_context, NULL, p->pcre2_jit_stack);
555 } else if (p->pcre2_jit_on != 0) {
556 BUG("The pcre2_jit_on variable should be 0 or 1, not %d",
557 p->pcre1_jit_on);
561 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
562 regmatch_t *match, int eflags)
564 int ret, flags = 0;
565 PCRE2_SIZE *ovector;
566 PCRE2_UCHAR errbuf[256];
568 if (eflags & REG_NOTBOL)
569 flags |= PCRE2_NOTBOL;
571 if (p->pcre2_jit_on)
572 ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line,
573 eol - line, 0, flags, p->pcre2_match_data,
574 NULL);
575 else
576 ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line,
577 eol - line, 0, flags, p->pcre2_match_data,
578 NULL);
580 if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) {
581 pcre2_get_error_message(ret, errbuf, sizeof(errbuf));
582 die("%s failed with error code %d: %s",
583 (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret,
584 errbuf);
586 if (ret > 0) {
587 ovector = pcre2_get_ovector_pointer(p->pcre2_match_data);
588 ret = 0;
589 match->rm_so = (int)ovector[0];
590 match->rm_eo = (int)ovector[1];
593 return ret;
596 static void free_pcre2_pattern(struct grep_pat *p)
598 pcre2_compile_context_free(p->pcre2_compile_context);
599 pcre2_code_free(p->pcre2_pattern);
600 pcre2_match_data_free(p->pcre2_match_data);
601 pcre2_jit_stack_free(p->pcre2_jit_stack);
602 pcre2_match_context_free(p->pcre2_match_context);
604 #else /* !USE_LIBPCRE2 */
605 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
608 * Unreachable until USE_LIBPCRE2 becomes synonymous with
609 * USE_LIBPCRE. See the sibling comment in
610 * grep_set_pattern_type_option().
612 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
615 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
616 regmatch_t *match, int eflags)
618 return 1;
621 static void free_pcre2_pattern(struct grep_pat *p)
624 #endif /* !USE_LIBPCRE2 */
626 static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
628 struct strbuf sb = STRBUF_INIT;
629 int err;
630 int regflags = 0;
632 basic_regex_quote_buf(&sb, p->pattern);
633 if (opt->ignore_case)
634 regflags |= REG_ICASE;
635 err = regcomp(&p->regexp, sb.buf, regflags);
636 if (opt->debug)
637 fprintf(stderr, "fixed %s\n", sb.buf);
638 strbuf_release(&sb);
639 if (err) {
640 char errbuf[1024];
641 regerror(err, &p->regexp, errbuf, sizeof(errbuf));
642 compile_regexp_failed(p, errbuf);
646 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
648 int ascii_only;
649 int err;
650 int regflags = REG_NEWLINE;
652 p->word_regexp = opt->word_regexp;
653 p->ignore_case = opt->ignore_case;
654 ascii_only = !has_non_ascii(p->pattern);
657 * Even when -F (fixed) asks us to do a non-regexp search, we
658 * may not be able to correctly case-fold when -i
659 * (ignore-case) is asked (in which case, we'll synthesize a
660 * regexp to match the pattern that matches regexp special
661 * characters literally, while ignoring case differences). On
662 * the other hand, even without -F, if the pattern does not
663 * have any regexp special characters and there is no need for
664 * case-folding search, we can internally turn it into a
665 * simple string match using kws. p->fixed tells us if we
666 * want to use kws.
668 if (opt->fixed ||
669 has_null(p->pattern, p->patternlen) ||
670 is_fixed(p->pattern, p->patternlen))
671 p->fixed = !p->ignore_case || ascii_only;
673 if (p->fixed) {
674 p->kws = kwsalloc(p->ignore_case ? tolower_trans_tbl : NULL);
675 kwsincr(p->kws, p->pattern, p->patternlen);
676 kwsprep(p->kws);
677 return;
678 } else if (opt->fixed) {
680 * We come here when the pattern has the non-ascii
681 * characters we cannot case-fold, and asked to
682 * ignore-case.
684 compile_fixed_regexp(p, opt);
685 return;
688 if (opt->pcre2) {
689 compile_pcre2_pattern(p, opt);
690 return;
693 if (opt->pcre1) {
694 compile_pcre1_regexp(p, opt);
695 return;
698 if (p->ignore_case)
699 regflags |= REG_ICASE;
700 if (opt->extended_regexp_option)
701 regflags |= REG_EXTENDED;
702 err = regcomp(&p->regexp, p->pattern, regflags);
703 if (err) {
704 char errbuf[1024];
705 regerror(err, &p->regexp, errbuf, 1024);
706 compile_regexp_failed(p, errbuf);
710 static struct grep_expr *compile_pattern_or(struct grep_pat **);
711 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
713 struct grep_pat *p;
714 struct grep_expr *x;
716 p = *list;
717 if (!p)
718 return NULL;
719 switch (p->token) {
720 case GREP_PATTERN: /* atom */
721 case GREP_PATTERN_HEAD:
722 case GREP_PATTERN_BODY:
723 x = xcalloc(1, sizeof (struct grep_expr));
724 x->node = GREP_NODE_ATOM;
725 x->u.atom = p;
726 *list = p->next;
727 return x;
728 case GREP_OPEN_PAREN:
729 *list = p->next;
730 x = compile_pattern_or(list);
731 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
732 die("unmatched parenthesis");
733 *list = (*list)->next;
734 return x;
735 default:
736 return NULL;
740 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
742 struct grep_pat *p;
743 struct grep_expr *x;
745 p = *list;
746 if (!p)
747 return NULL;
748 switch (p->token) {
749 case GREP_NOT:
750 if (!p->next)
751 die("--not not followed by pattern expression");
752 *list = p->next;
753 x = xcalloc(1, sizeof (struct grep_expr));
754 x->node = GREP_NODE_NOT;
755 x->u.unary = compile_pattern_not(list);
756 if (!x->u.unary)
757 die("--not followed by non pattern expression");
758 return x;
759 default:
760 return compile_pattern_atom(list);
764 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
766 struct grep_pat *p;
767 struct grep_expr *x, *y, *z;
769 x = compile_pattern_not(list);
770 p = *list;
771 if (p && p->token == GREP_AND) {
772 if (!p->next)
773 die("--and not followed by pattern expression");
774 *list = p->next;
775 y = compile_pattern_and(list);
776 if (!y)
777 die("--and not followed by pattern expression");
778 z = xcalloc(1, sizeof (struct grep_expr));
779 z->node = GREP_NODE_AND;
780 z->u.binary.left = x;
781 z->u.binary.right = y;
782 return z;
784 return x;
787 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
789 struct grep_pat *p;
790 struct grep_expr *x, *y, *z;
792 x = compile_pattern_and(list);
793 p = *list;
794 if (x && p && p->token != GREP_CLOSE_PAREN) {
795 y = compile_pattern_or(list);
796 if (!y)
797 die("not a pattern expression %s", p->pattern);
798 z = xcalloc(1, sizeof (struct grep_expr));
799 z->node = GREP_NODE_OR;
800 z->u.binary.left = x;
801 z->u.binary.right = y;
802 return z;
804 return x;
807 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
809 return compile_pattern_or(list);
812 static void indent(int in)
814 while (in-- > 0)
815 fputc(' ', stderr);
818 static void dump_grep_pat(struct grep_pat *p)
820 switch (p->token) {
821 case GREP_AND: fprintf(stderr, "*and*"); break;
822 case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
823 case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
824 case GREP_NOT: fprintf(stderr, "*not*"); break;
825 case GREP_OR: fprintf(stderr, "*or*"); break;
827 case GREP_PATTERN: fprintf(stderr, "pattern"); break;
828 case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
829 case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
832 switch (p->token) {
833 default: break;
834 case GREP_PATTERN_HEAD:
835 fprintf(stderr, "<head %d>", p->field); break;
836 case GREP_PATTERN_BODY:
837 fprintf(stderr, "<body>"); break;
839 switch (p->token) {
840 default: break;
841 case GREP_PATTERN_HEAD:
842 case GREP_PATTERN_BODY:
843 case GREP_PATTERN:
844 fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
845 break;
847 fputc('\n', stderr);
850 static void dump_grep_expression_1(struct grep_expr *x, int in)
852 indent(in);
853 switch (x->node) {
854 case GREP_NODE_TRUE:
855 fprintf(stderr, "true\n");
856 break;
857 case GREP_NODE_ATOM:
858 dump_grep_pat(x->u.atom);
859 break;
860 case GREP_NODE_NOT:
861 fprintf(stderr, "(not\n");
862 dump_grep_expression_1(x->u.unary, in+1);
863 indent(in);
864 fprintf(stderr, ")\n");
865 break;
866 case GREP_NODE_AND:
867 fprintf(stderr, "(and\n");
868 dump_grep_expression_1(x->u.binary.left, in+1);
869 dump_grep_expression_1(x->u.binary.right, in+1);
870 indent(in);
871 fprintf(stderr, ")\n");
872 break;
873 case GREP_NODE_OR:
874 fprintf(stderr, "(or\n");
875 dump_grep_expression_1(x->u.binary.left, in+1);
876 dump_grep_expression_1(x->u.binary.right, in+1);
877 indent(in);
878 fprintf(stderr, ")\n");
879 break;
883 static void dump_grep_expression(struct grep_opt *opt)
885 struct grep_expr *x = opt->pattern_expression;
887 if (opt->all_match)
888 fprintf(stderr, "[all-match]\n");
889 dump_grep_expression_1(x, 0);
890 fflush(NULL);
893 static struct grep_expr *grep_true_expr(void)
895 struct grep_expr *z = xcalloc(1, sizeof(*z));
896 z->node = GREP_NODE_TRUE;
897 return z;
900 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
902 struct grep_expr *z = xcalloc(1, sizeof(*z));
903 z->node = GREP_NODE_OR;
904 z->u.binary.left = left;
905 z->u.binary.right = right;
906 return z;
909 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
911 struct grep_pat *p;
912 struct grep_expr *header_expr;
913 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
914 enum grep_header_field fld;
916 if (!opt->header_list)
917 return NULL;
919 for (p = opt->header_list; p; p = p->next) {
920 if (p->token != GREP_PATTERN_HEAD)
921 BUG("a non-header pattern in grep header list.");
922 if (p->field < GREP_HEADER_FIELD_MIN ||
923 GREP_HEADER_FIELD_MAX <= p->field)
924 BUG("unknown header field %d", p->field);
925 compile_regexp(p, opt);
928 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
929 header_group[fld] = NULL;
931 for (p = opt->header_list; p; p = p->next) {
932 struct grep_expr *h;
933 struct grep_pat *pp = p;
935 h = compile_pattern_atom(&pp);
936 if (!h || pp != p->next)
937 BUG("malformed header expr");
938 if (!header_group[p->field]) {
939 header_group[p->field] = h;
940 continue;
942 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
945 header_expr = NULL;
947 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
948 if (!header_group[fld])
949 continue;
950 if (!header_expr)
951 header_expr = grep_true_expr();
952 header_expr = grep_or_expr(header_group[fld], header_expr);
954 return header_expr;
957 static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
959 struct grep_expr *z = x;
961 while (x) {
962 assert(x->node == GREP_NODE_OR);
963 if (x->u.binary.right &&
964 x->u.binary.right->node == GREP_NODE_TRUE) {
965 x->u.binary.right = y;
966 break;
968 x = x->u.binary.right;
970 return z;
973 static void compile_grep_patterns_real(struct grep_opt *opt)
975 struct grep_pat *p;
976 struct grep_expr *header_expr = prep_header_patterns(opt);
978 for (p = opt->pattern_list; p; p = p->next) {
979 switch (p->token) {
980 case GREP_PATTERN: /* atom */
981 case GREP_PATTERN_HEAD:
982 case GREP_PATTERN_BODY:
983 compile_regexp(p, opt);
984 break;
985 default:
986 opt->extended = 1;
987 break;
991 if (opt->all_match || header_expr)
992 opt->extended = 1;
993 else if (!opt->extended && !opt->debug)
994 return;
996 p = opt->pattern_list;
997 if (p)
998 opt->pattern_expression = compile_pattern_expr(&p);
999 if (p)
1000 die("incomplete pattern expression: %s", p->pattern);
1002 if (!header_expr)
1003 return;
1005 if (!opt->pattern_expression)
1006 opt->pattern_expression = header_expr;
1007 else if (opt->all_match)
1008 opt->pattern_expression = grep_splice_or(header_expr,
1009 opt->pattern_expression);
1010 else
1011 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
1012 header_expr);
1013 opt->all_match = 1;
1016 void compile_grep_patterns(struct grep_opt *opt)
1018 compile_grep_patterns_real(opt);
1019 if (opt->debug)
1020 dump_grep_expression(opt);
1023 static void free_pattern_expr(struct grep_expr *x)
1025 switch (x->node) {
1026 case GREP_NODE_TRUE:
1027 case GREP_NODE_ATOM:
1028 break;
1029 case GREP_NODE_NOT:
1030 free_pattern_expr(x->u.unary);
1031 break;
1032 case GREP_NODE_AND:
1033 case GREP_NODE_OR:
1034 free_pattern_expr(x->u.binary.left);
1035 free_pattern_expr(x->u.binary.right);
1036 break;
1038 free(x);
1041 void free_grep_patterns(struct grep_opt *opt)
1043 struct grep_pat *p, *n;
1045 for (p = opt->pattern_list; p; p = n) {
1046 n = p->next;
1047 switch (p->token) {
1048 case GREP_PATTERN: /* atom */
1049 case GREP_PATTERN_HEAD:
1050 case GREP_PATTERN_BODY:
1051 if (p->kws)
1052 kwsfree(p->kws);
1053 else if (p->pcre1_regexp)
1054 free_pcre1_regexp(p);
1055 else if (p->pcre2_pattern)
1056 free_pcre2_pattern(p);
1057 else
1058 regfree(&p->regexp);
1059 free(p->pattern);
1060 break;
1061 default:
1062 break;
1064 free(p);
1067 if (!opt->extended)
1068 return;
1069 free_pattern_expr(opt->pattern_expression);
1072 static char *end_of_line(char *cp, unsigned long *left)
1074 unsigned long l = *left;
1075 while (l && *cp != '\n') {
1076 l--;
1077 cp++;
1079 *left = l;
1080 return cp;
1083 static int word_char(char ch)
1085 return isalnum(ch) || ch == '_';
1088 static void output_color(struct grep_opt *opt, const void *data, size_t size,
1089 const char *color)
1091 if (want_color(opt->color) && color && color[0]) {
1092 opt->output(opt, color, strlen(color));
1093 opt->output(opt, data, size);
1094 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
1095 } else
1096 opt->output(opt, data, size);
1099 static void output_sep(struct grep_opt *opt, char sign)
1101 if (opt->null_following_name)
1102 opt->output(opt, "\0", 1);
1103 else
1104 output_color(opt, &sign, 1, opt->colors[GREP_COLOR_SEP]);
1107 static void show_name(struct grep_opt *opt, const char *name)
1109 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1110 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
1113 static int fixmatch(struct grep_pat *p, char *line, char *eol,
1114 regmatch_t *match)
1116 struct kwsmatch kwsm;
1117 size_t offset = kwsexec(p->kws, line, eol - line, &kwsm);
1118 if (offset == -1) {
1119 match->rm_so = match->rm_eo = -1;
1120 return REG_NOMATCH;
1121 } else {
1122 match->rm_so = offset;
1123 match->rm_eo = match->rm_so + kwsm.size[0];
1124 return 0;
1128 static int patmatch(struct grep_pat *p, char *line, char *eol,
1129 regmatch_t *match, int eflags)
1131 int hit;
1133 if (p->fixed)
1134 hit = !fixmatch(p, line, eol, match);
1135 else if (p->pcre1_regexp)
1136 hit = !pcre1match(p, line, eol, match, eflags);
1137 else if (p->pcre2_pattern)
1138 hit = !pcre2match(p, line, eol, match, eflags);
1139 else
1140 hit = !regexec_buf(&p->regexp, line, eol - line, 1, match,
1141 eflags);
1143 return hit;
1146 static int strip_timestamp(char *bol, char **eol_p)
1148 char *eol = *eol_p;
1149 int ch;
1151 while (bol < --eol) {
1152 if (*eol != '>')
1153 continue;
1154 *eol_p = ++eol;
1155 ch = *eol;
1156 *eol = '\0';
1157 return ch;
1159 return 0;
1162 static struct {
1163 const char *field;
1164 size_t len;
1165 } header_field[] = {
1166 { "author ", 7 },
1167 { "committer ", 10 },
1168 { "reflog ", 7 },
1171 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
1172 enum grep_context ctx,
1173 regmatch_t *pmatch, int eflags)
1175 int hit = 0;
1176 int saved_ch = 0;
1177 const char *start = bol;
1179 if ((p->token != GREP_PATTERN) &&
1180 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
1181 return 0;
1183 if (p->token == GREP_PATTERN_HEAD) {
1184 const char *field;
1185 size_t len;
1186 assert(p->field < ARRAY_SIZE(header_field));
1187 field = header_field[p->field].field;
1188 len = header_field[p->field].len;
1189 if (strncmp(bol, field, len))
1190 return 0;
1191 bol += len;
1192 switch (p->field) {
1193 case GREP_HEADER_AUTHOR:
1194 case GREP_HEADER_COMMITTER:
1195 saved_ch = strip_timestamp(bol, &eol);
1196 break;
1197 default:
1198 break;
1202 again:
1203 hit = patmatch(p, bol, eol, pmatch, eflags);
1205 if (hit && p->word_regexp) {
1206 if ((pmatch[0].rm_so < 0) ||
1207 (eol - bol) < pmatch[0].rm_so ||
1208 (pmatch[0].rm_eo < 0) ||
1209 (eol - bol) < pmatch[0].rm_eo)
1210 die("regexp returned nonsense");
1212 /* Match beginning must be either beginning of the
1213 * line, or at word boundary (i.e. the last char must
1214 * not be a word char). Similarly, match end must be
1215 * either end of the line, or at word boundary
1216 * (i.e. the next char must not be a word char).
1218 if ( ((pmatch[0].rm_so == 0) ||
1219 !word_char(bol[pmatch[0].rm_so-1])) &&
1220 ((pmatch[0].rm_eo == (eol-bol)) ||
1221 !word_char(bol[pmatch[0].rm_eo])) )
1223 else
1224 hit = 0;
1226 /* Words consist of at least one character. */
1227 if (pmatch->rm_so == pmatch->rm_eo)
1228 hit = 0;
1230 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
1231 /* There could be more than one match on the
1232 * line, and the first match might not be
1233 * strict word match. But later ones could be!
1234 * Forward to the next possible start, i.e. the
1235 * next position following a non-word char.
1237 bol = pmatch[0].rm_so + bol + 1;
1238 while (word_char(bol[-1]) && bol < eol)
1239 bol++;
1240 eflags |= REG_NOTBOL;
1241 if (bol < eol)
1242 goto again;
1245 if (p->token == GREP_PATTERN_HEAD && saved_ch)
1246 *eol = saved_ch;
1247 if (hit) {
1248 pmatch[0].rm_so += bol - start;
1249 pmatch[0].rm_eo += bol - start;
1251 return hit;
1254 static int match_expr_eval(struct grep_opt *opt, struct grep_expr *x, char *bol,
1255 char *eol, enum grep_context ctx, ssize_t *col,
1256 ssize_t *icol, int collect_hits)
1258 int h = 0;
1260 if (!x)
1261 die("Not a valid grep expression");
1262 switch (x->node) {
1263 case GREP_NODE_TRUE:
1264 h = 1;
1265 break;
1266 case GREP_NODE_ATOM:
1268 regmatch_t tmp;
1269 h = match_one_pattern(x->u.atom, bol, eol, ctx,
1270 &tmp, 0);
1271 if (h && (*col < 0 || tmp.rm_so < *col))
1272 *col = tmp.rm_so;
1274 break;
1275 case GREP_NODE_NOT:
1277 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1279 h = !match_expr_eval(opt, x->u.unary, bol, eol, ctx, icol, col,
1281 break;
1282 case GREP_NODE_AND:
1283 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1284 icol, 0);
1285 if (h || opt->columnnum) {
1287 * Don't short-circuit AND when given --column, since a
1288 * NOT earlier in the tree may turn this into an OR. In
1289 * this case, see the below comment.
1291 h &= match_expr_eval(opt, x->u.binary.right, bol, eol,
1292 ctx, col, icol, 0);
1294 break;
1295 case GREP_NODE_OR:
1296 if (!(collect_hits || opt->columnnum)) {
1298 * Don't short-circuit OR when given --column (or
1299 * collecting hits) to ensure we don't skip a later
1300 * child that would produce an earlier match.
1302 return (match_expr_eval(opt, x->u.binary.left, bol, eol,
1303 ctx, col, icol, 0) ||
1304 match_expr_eval(opt, x->u.binary.right, bol,
1305 eol, ctx, col, icol, 0));
1307 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1308 icol, 0);
1309 if (collect_hits)
1310 x->u.binary.left->hit |= h;
1311 h |= match_expr_eval(opt, x->u.binary.right, bol, eol, ctx, col,
1312 icol, collect_hits);
1313 break;
1314 default:
1315 die("Unexpected node type (internal error) %d", x->node);
1317 if (collect_hits)
1318 x->hit |= h;
1319 return h;
1322 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
1323 enum grep_context ctx, ssize_t *col,
1324 ssize_t *icol, int collect_hits)
1326 struct grep_expr *x = opt->pattern_expression;
1327 return match_expr_eval(opt, x, bol, eol, ctx, col, icol, collect_hits);
1330 static int match_line(struct grep_opt *opt, char *bol, char *eol,
1331 ssize_t *col, ssize_t *icol,
1332 enum grep_context ctx, int collect_hits)
1334 struct grep_pat *p;
1335 int hit = 0;
1337 if (opt->extended)
1338 return match_expr(opt, bol, eol, ctx, col, icol,
1339 collect_hits);
1341 /* we do not call with collect_hits without being extended */
1342 for (p = opt->pattern_list; p; p = p->next) {
1343 regmatch_t tmp;
1344 if (match_one_pattern(p, bol, eol, ctx, &tmp, 0)) {
1345 hit |= 1;
1346 if (!opt->columnnum) {
1348 * Without --column, any single match on a line
1349 * is enough to know that it needs to be
1350 * printed. With --column, scan _all_ patterns
1351 * to find the earliest.
1353 break;
1355 if (*col < 0 || tmp.rm_so < *col)
1356 *col = tmp.rm_so;
1359 return hit;
1362 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
1363 enum grep_context ctx,
1364 regmatch_t *pmatch, int eflags)
1366 regmatch_t match;
1368 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
1369 return 0;
1370 if (match.rm_so < 0 || match.rm_eo < 0)
1371 return 0;
1372 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
1373 if (match.rm_so > pmatch->rm_so)
1374 return 1;
1375 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
1376 return 1;
1378 pmatch->rm_so = match.rm_so;
1379 pmatch->rm_eo = match.rm_eo;
1380 return 1;
1383 static int next_match(struct grep_opt *opt, char *bol, char *eol,
1384 enum grep_context ctx, regmatch_t *pmatch, int eflags)
1386 struct grep_pat *p;
1387 int hit = 0;
1389 pmatch->rm_so = pmatch->rm_eo = -1;
1390 if (bol < eol) {
1391 for (p = opt->pattern_list; p; p = p->next) {
1392 switch (p->token) {
1393 case GREP_PATTERN: /* atom */
1394 case GREP_PATTERN_HEAD:
1395 case GREP_PATTERN_BODY:
1396 hit |= match_next_pattern(p, bol, eol, ctx,
1397 pmatch, eflags);
1398 break;
1399 default:
1400 break;
1404 return hit;
1407 static void show_line(struct grep_opt *opt, char *bol, char *eol,
1408 const char *name, unsigned lno, ssize_t cno, char sign)
1410 int rest = eol - bol;
1411 const char *match_color, *line_color = NULL;
1413 if (opt->file_break && opt->last_shown == 0) {
1414 if (opt->show_hunk_mark)
1415 opt->output(opt, "\n", 1);
1416 } else if (opt->pre_context || opt->post_context || opt->funcbody) {
1417 if (opt->last_shown == 0) {
1418 if (opt->show_hunk_mark) {
1419 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1420 opt->output(opt, "\n", 1);
1422 } else if (lno > opt->last_shown + 1) {
1423 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1424 opt->output(opt, "\n", 1);
1427 if (opt->heading && opt->last_shown == 0) {
1428 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1429 opt->output(opt, "\n", 1);
1431 opt->last_shown = lno;
1433 if (!opt->heading && opt->pathname) {
1434 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1435 output_sep(opt, sign);
1437 if (opt->linenum) {
1438 char buf[32];
1439 xsnprintf(buf, sizeof(buf), "%d", lno);
1440 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_LINENO]);
1441 output_sep(opt, sign);
1444 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1445 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1446 * being called with a context line.
1448 if (opt->columnnum && cno) {
1449 char buf[32];
1450 xsnprintf(buf, sizeof(buf), "%"PRIuMAX, (uintmax_t)cno);
1451 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_COLUMNNO]);
1452 output_sep(opt, sign);
1454 if (opt->color) {
1455 regmatch_t match;
1456 enum grep_context ctx = GREP_CONTEXT_BODY;
1457 int ch = *eol;
1458 int eflags = 0;
1460 if (sign == ':')
1461 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
1462 else
1463 match_color = opt->colors[GREP_COLOR_MATCH_CONTEXT];
1464 if (sign == ':')
1465 line_color = opt->colors[GREP_COLOR_SELECTED];
1466 else if (sign == '-')
1467 line_color = opt->colors[GREP_COLOR_CONTEXT];
1468 else if (sign == '=')
1469 line_color = opt->colors[GREP_COLOR_FUNCTION];
1470 *eol = '\0';
1471 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
1472 if (match.rm_so == match.rm_eo)
1473 break;
1475 output_color(opt, bol, match.rm_so, line_color);
1476 output_color(opt, bol + match.rm_so,
1477 match.rm_eo - match.rm_so, match_color);
1478 bol += match.rm_eo;
1479 rest -= match.rm_eo;
1480 eflags = REG_NOTBOL;
1482 *eol = ch;
1484 output_color(opt, bol, rest, line_color);
1485 opt->output(opt, "\n", 1);
1488 #ifndef NO_PTHREADS
1489 int grep_use_locks;
1492 * This lock protects access to the gitattributes machinery, which is
1493 * not thread-safe.
1495 pthread_mutex_t grep_attr_mutex;
1497 static inline void grep_attr_lock(void)
1499 if (grep_use_locks)
1500 pthread_mutex_lock(&grep_attr_mutex);
1503 static inline void grep_attr_unlock(void)
1505 if (grep_use_locks)
1506 pthread_mutex_unlock(&grep_attr_mutex);
1510 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1512 pthread_mutex_t grep_read_mutex;
1514 #else
1515 #define grep_attr_lock()
1516 #define grep_attr_unlock()
1517 #endif
1519 static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
1521 xdemitconf_t *xecfg = opt->priv;
1522 if (xecfg && !xecfg->find_func) {
1523 grep_source_load_driver(gs);
1524 if (gs->driver->funcname.pattern) {
1525 const struct userdiff_funcname *pe = &gs->driver->funcname;
1526 xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
1527 } else {
1528 xecfg = opt->priv = NULL;
1532 if (xecfg) {
1533 char buf[1];
1534 return xecfg->find_func(bol, eol - bol, buf, 1,
1535 xecfg->find_func_priv) >= 0;
1538 if (bol == eol)
1539 return 0;
1540 if (isalpha(*bol) || *bol == '_' || *bol == '$')
1541 return 1;
1542 return 0;
1545 static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
1546 char *bol, unsigned lno)
1548 while (bol > gs->buf) {
1549 char *eol = --bol;
1551 while (bol > gs->buf && bol[-1] != '\n')
1552 bol--;
1553 lno--;
1555 if (lno <= opt->last_shown)
1556 break;
1558 if (match_funcname(opt, gs, bol, eol)) {
1559 show_line(opt, bol, eol, gs->name, lno, 0, '=');
1560 break;
1565 static int is_empty_line(const char *bol, const char *eol);
1567 static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
1568 char *bol, char *end, unsigned lno)
1570 unsigned cur = lno, from = 1, funcname_lno = 0, orig_from;
1571 int funcname_needed = !!opt->funcname, comment_needed = 0;
1573 if (opt->pre_context < lno)
1574 from = lno - opt->pre_context;
1575 if (from <= opt->last_shown)
1576 from = opt->last_shown + 1;
1577 orig_from = from;
1578 if (opt->funcbody) {
1579 if (match_funcname(opt, gs, bol, end))
1580 comment_needed = 1;
1581 else
1582 funcname_needed = 1;
1583 from = opt->last_shown + 1;
1586 /* Rewind. */
1587 while (bol > gs->buf && cur > from) {
1588 char *next_bol = bol;
1589 char *eol = --bol;
1591 while (bol > gs->buf && bol[-1] != '\n')
1592 bol--;
1593 cur--;
1594 if (comment_needed && (is_empty_line(bol, eol) ||
1595 match_funcname(opt, gs, bol, eol))) {
1596 comment_needed = 0;
1597 from = orig_from;
1598 if (cur < from) {
1599 cur++;
1600 bol = next_bol;
1601 break;
1604 if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
1605 funcname_lno = cur;
1606 funcname_needed = 0;
1607 if (opt->funcbody)
1608 comment_needed = 1;
1609 else
1610 from = orig_from;
1614 /* We need to look even further back to find a function signature. */
1615 if (opt->funcname && funcname_needed)
1616 show_funcname_line(opt, gs, bol, cur);
1618 /* Back forward. */
1619 while (cur < lno) {
1620 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
1622 while (*eol != '\n')
1623 eol++;
1624 show_line(opt, bol, eol, gs->name, cur, 0, sign);
1625 bol = eol + 1;
1626 cur++;
1630 static int should_lookahead(struct grep_opt *opt)
1632 struct grep_pat *p;
1634 if (opt->extended)
1635 return 0; /* punt for too complex stuff */
1636 if (opt->invert)
1637 return 0;
1638 for (p = opt->pattern_list; p; p = p->next) {
1639 if (p->token != GREP_PATTERN)
1640 return 0; /* punt for "header only" and stuff */
1642 return 1;
1645 static int look_ahead(struct grep_opt *opt,
1646 unsigned long *left_p,
1647 unsigned *lno_p,
1648 char **bol_p)
1650 unsigned lno = *lno_p;
1651 char *bol = *bol_p;
1652 struct grep_pat *p;
1653 char *sp, *last_bol;
1654 regoff_t earliest = -1;
1656 for (p = opt->pattern_list; p; p = p->next) {
1657 int hit;
1658 regmatch_t m;
1660 hit = patmatch(p, bol, bol + *left_p, &m, 0);
1661 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
1662 continue;
1663 if (earliest < 0 || m.rm_so < earliest)
1664 earliest = m.rm_so;
1667 if (earliest < 0) {
1668 *bol_p = bol + *left_p;
1669 *left_p = 0;
1670 return 1;
1672 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
1673 ; /* find the beginning of the line */
1674 last_bol = sp;
1676 for (sp = bol; sp < last_bol; sp++) {
1677 if (*sp == '\n')
1678 lno++;
1680 *left_p -= last_bol - bol;
1681 *bol_p = last_bol;
1682 *lno_p = lno;
1683 return 0;
1686 static int fill_textconv_grep(struct userdiff_driver *driver,
1687 struct grep_source *gs)
1689 struct diff_filespec *df;
1690 char *buf;
1691 size_t size;
1693 if (!driver || !driver->textconv)
1694 return grep_source_load(gs);
1697 * The textconv interface is intimately tied to diff_filespecs, so we
1698 * have to pretend to be one. If we could unify the grep_source
1699 * and diff_filespec structs, this mess could just go away.
1701 df = alloc_filespec(gs->path);
1702 switch (gs->type) {
1703 case GREP_SOURCE_OID:
1704 fill_filespec(df, gs->identifier, 1, 0100644);
1705 break;
1706 case GREP_SOURCE_FILE:
1707 fill_filespec(df, &null_oid, 0, 0100644);
1708 break;
1709 default:
1710 BUG("attempt to textconv something without a path?");
1714 * fill_textconv is not remotely thread-safe; it may load objects
1715 * behind the scenes, and it modifies the global diff tempfile
1716 * structure.
1718 grep_read_lock();
1719 size = fill_textconv(driver, df, &buf);
1720 grep_read_unlock();
1721 free_filespec(df);
1724 * The normal fill_textconv usage by the diff machinery would just keep
1725 * the textconv'd buf separate from the diff_filespec. But much of the
1726 * grep code passes around a grep_source and assumes that its "buf"
1727 * pointer is the beginning of the thing we are searching. So let's
1728 * install our textconv'd version into the grep_source, taking care not
1729 * to leak any existing buffer.
1731 grep_source_clear_data(gs);
1732 gs->buf = buf;
1733 gs->size = size;
1735 return 0;
1738 static int is_empty_line(const char *bol, const char *eol)
1740 while (bol < eol && isspace(*bol))
1741 bol++;
1742 return bol == eol;
1745 static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
1747 char *bol;
1748 char *peek_bol = NULL;
1749 unsigned long left;
1750 unsigned lno = 1;
1751 unsigned last_hit = 0;
1752 int binary_match_only = 0;
1753 unsigned count = 0;
1754 int try_lookahead = 0;
1755 int show_function = 0;
1756 struct userdiff_driver *textconv = NULL;
1757 enum grep_context ctx = GREP_CONTEXT_HEAD;
1758 xdemitconf_t xecfg;
1760 if (!opt->output)
1761 opt->output = std_output;
1763 if (opt->pre_context || opt->post_context || opt->file_break ||
1764 opt->funcbody) {
1765 /* Show hunk marks, except for the first file. */
1766 if (opt->last_shown)
1767 opt->show_hunk_mark = 1;
1769 * If we're using threads then we can't easily identify
1770 * the first file. Always put hunk marks in that case
1771 * and skip the very first one later in work_done().
1773 if (opt->output != std_output)
1774 opt->show_hunk_mark = 1;
1776 opt->last_shown = 0;
1778 if (opt->allow_textconv) {
1779 grep_source_load_driver(gs);
1781 * We might set up the shared textconv cache data here, which
1782 * is not thread-safe.
1784 grep_attr_lock();
1785 textconv = userdiff_get_textconv(gs->driver);
1786 grep_attr_unlock();
1790 * We know the result of a textconv is text, so we only have to care
1791 * about binary handling if we are not using it.
1793 if (!textconv) {
1794 switch (opt->binary) {
1795 case GREP_BINARY_DEFAULT:
1796 if (grep_source_is_binary(gs))
1797 binary_match_only = 1;
1798 break;
1799 case GREP_BINARY_NOMATCH:
1800 if (grep_source_is_binary(gs))
1801 return 0; /* Assume unmatch */
1802 break;
1803 case GREP_BINARY_TEXT:
1804 break;
1805 default:
1806 BUG("unknown binary handling mode");
1810 memset(&xecfg, 0, sizeof(xecfg));
1811 opt->priv = &xecfg;
1813 try_lookahead = should_lookahead(opt);
1815 if (fill_textconv_grep(textconv, gs) < 0)
1816 return 0;
1818 bol = gs->buf;
1819 left = gs->size;
1820 while (left) {
1821 char *eol, ch;
1822 int hit;
1823 ssize_t cno;
1824 ssize_t col = -1, icol = -1;
1827 * look_ahead() skips quickly to the line that possibly
1828 * has the next hit; don't call it if we need to do
1829 * something more than just skipping the current line
1830 * in response to an unmatch for the current line. E.g.
1831 * inside a post-context window, we will show the current
1832 * line as a context around the previous hit when it
1833 * doesn't hit.
1835 if (try_lookahead
1836 && !(last_hit
1837 && (show_function ||
1838 lno <= last_hit + opt->post_context))
1839 && look_ahead(opt, &left, &lno, &bol))
1840 break;
1841 eol = end_of_line(bol, &left);
1842 ch = *eol;
1843 *eol = 0;
1845 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
1846 ctx = GREP_CONTEXT_BODY;
1848 hit = match_line(opt, bol, eol, &col, &icol, ctx, collect_hits);
1849 *eol = ch;
1851 if (collect_hits)
1852 goto next_line;
1854 /* "grep -v -e foo -e bla" should list lines
1855 * that do not have either, so inversion should
1856 * be done outside.
1858 if (opt->invert)
1859 hit = !hit;
1860 if (opt->unmatch_name_only) {
1861 if (hit)
1862 return 0;
1863 goto next_line;
1865 if (hit) {
1866 count++;
1867 if (opt->status_only)
1868 return 1;
1869 if (opt->name_only) {
1870 show_name(opt, gs->name);
1871 return 1;
1873 if (opt->count)
1874 goto next_line;
1875 if (binary_match_only) {
1876 opt->output(opt, "Binary file ", 12);
1877 output_color(opt, gs->name, strlen(gs->name),
1878 opt->colors[GREP_COLOR_FILENAME]);
1879 opt->output(opt, " matches\n", 9);
1880 return 1;
1882 /* Hit at this line. If we haven't shown the
1883 * pre-context lines, we would need to show them.
1885 if (opt->pre_context || opt->funcbody)
1886 show_pre_context(opt, gs, bol, eol, lno);
1887 else if (opt->funcname)
1888 show_funcname_line(opt, gs, bol, lno);
1889 cno = opt->invert ? icol : col;
1890 if (cno < 0) {
1892 * A negative cno indicates that there was no
1893 * match on the line. We are thus inverted and
1894 * being asked to show all lines that _don't_
1895 * match a given expression. Therefore, set cno
1896 * to 0 to suggest the whole line matches.
1898 cno = 0;
1900 show_line(opt, bol, eol, gs->name, lno, cno + 1, ':');
1901 last_hit = lno;
1902 if (opt->funcbody)
1903 show_function = 1;
1904 goto next_line;
1906 if (show_function && (!peek_bol || peek_bol < bol)) {
1907 unsigned long peek_left = left;
1908 char *peek_eol = eol;
1911 * Trailing empty lines are not interesting.
1912 * Peek past them to see if they belong to the
1913 * body of the current function.
1915 peek_bol = bol;
1916 while (is_empty_line(peek_bol, peek_eol)) {
1917 peek_bol = peek_eol + 1;
1918 peek_eol = end_of_line(peek_bol, &peek_left);
1921 if (match_funcname(opt, gs, peek_bol, peek_eol))
1922 show_function = 0;
1924 if (show_function ||
1925 (last_hit && lno <= last_hit + opt->post_context)) {
1926 /* If the last hit is within the post context,
1927 * we need to show this line.
1929 show_line(opt, bol, eol, gs->name, lno, col + 1, '-');
1932 next_line:
1933 bol = eol + 1;
1934 if (!left)
1935 break;
1936 left--;
1937 lno++;
1940 if (collect_hits)
1941 return 0;
1943 if (opt->status_only)
1944 return opt->unmatch_name_only;
1945 if (opt->unmatch_name_only) {
1946 /* We did not see any hit, so we want to show this */
1947 show_name(opt, gs->name);
1948 return 1;
1951 xdiff_clear_find_func(&xecfg);
1952 opt->priv = NULL;
1954 /* NEEDSWORK:
1955 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1956 * which feels mostly useless but sometimes useful. Maybe
1957 * make it another option? For now suppress them.
1959 if (opt->count && count) {
1960 char buf[32];
1961 if (opt->pathname) {
1962 output_color(opt, gs->name, strlen(gs->name),
1963 opt->colors[GREP_COLOR_FILENAME]);
1964 output_sep(opt, ':');
1966 xsnprintf(buf, sizeof(buf), "%u\n", count);
1967 opt->output(opt, buf, strlen(buf));
1968 return 1;
1970 return !!last_hit;
1973 static void clr_hit_marker(struct grep_expr *x)
1975 /* All-hit markers are meaningful only at the very top level
1976 * OR node.
1978 while (1) {
1979 x->hit = 0;
1980 if (x->node != GREP_NODE_OR)
1981 return;
1982 x->u.binary.left->hit = 0;
1983 x = x->u.binary.right;
1987 static int chk_hit_marker(struct grep_expr *x)
1989 /* Top level nodes have hit markers. See if they all are hits */
1990 while (1) {
1991 if (x->node != GREP_NODE_OR)
1992 return x->hit;
1993 if (!x->u.binary.left->hit)
1994 return 0;
1995 x = x->u.binary.right;
1999 int grep_source(struct grep_opt *opt, struct grep_source *gs)
2002 * we do not have to do the two-pass grep when we do not check
2003 * buffer-wide "all-match".
2005 if (!opt->all_match)
2006 return grep_source_1(opt, gs, 0);
2008 /* Otherwise the toplevel "or" terms hit a bit differently.
2009 * We first clear hit markers from them.
2011 clr_hit_marker(opt->pattern_expression);
2012 grep_source_1(opt, gs, 1);
2014 if (!chk_hit_marker(opt->pattern_expression))
2015 return 0;
2017 return grep_source_1(opt, gs, 0);
2020 int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
2022 struct grep_source gs;
2023 int r;
2025 grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
2026 gs.buf = buf;
2027 gs.size = size;
2029 r = grep_source(opt, &gs);
2031 grep_source_clear(&gs);
2032 return r;
2035 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
2036 const char *name, const char *path,
2037 const void *identifier)
2039 gs->type = type;
2040 gs->name = xstrdup_or_null(name);
2041 gs->path = xstrdup_or_null(path);
2042 gs->buf = NULL;
2043 gs->size = 0;
2044 gs->driver = NULL;
2046 switch (type) {
2047 case GREP_SOURCE_FILE:
2048 gs->identifier = xstrdup(identifier);
2049 break;
2050 case GREP_SOURCE_OID:
2051 gs->identifier = oiddup(identifier);
2052 break;
2053 case GREP_SOURCE_BUF:
2054 gs->identifier = NULL;
2055 break;
2059 void grep_source_clear(struct grep_source *gs)
2061 FREE_AND_NULL(gs->name);
2062 FREE_AND_NULL(gs->path);
2063 FREE_AND_NULL(gs->identifier);
2064 grep_source_clear_data(gs);
2067 void grep_source_clear_data(struct grep_source *gs)
2069 switch (gs->type) {
2070 case GREP_SOURCE_FILE:
2071 case GREP_SOURCE_OID:
2072 FREE_AND_NULL(gs->buf);
2073 gs->size = 0;
2074 break;
2075 case GREP_SOURCE_BUF:
2076 /* leave user-provided buf intact */
2077 break;
2081 static int grep_source_load_oid(struct grep_source *gs)
2083 enum object_type type;
2085 grep_read_lock();
2086 gs->buf = read_object_file(gs->identifier, &type, &gs->size);
2087 grep_read_unlock();
2089 if (!gs->buf)
2090 return error(_("'%s': unable to read %s"),
2091 gs->name,
2092 oid_to_hex(gs->identifier));
2093 return 0;
2096 static int grep_source_load_file(struct grep_source *gs)
2098 const char *filename = gs->identifier;
2099 struct stat st;
2100 char *data;
2101 size_t size;
2102 int i;
2104 if (lstat(filename, &st) < 0) {
2105 err_ret:
2106 if (errno != ENOENT)
2107 error_errno(_("failed to stat '%s'"), filename);
2108 return -1;
2110 if (!S_ISREG(st.st_mode))
2111 return -1;
2112 size = xsize_t(st.st_size);
2113 i = open(filename, O_RDONLY);
2114 if (i < 0)
2115 goto err_ret;
2116 data = xmallocz(size);
2117 if (st.st_size != read_in_full(i, data, size)) {
2118 error_errno(_("'%s': short read"), filename);
2119 close(i);
2120 free(data);
2121 return -1;
2123 close(i);
2125 gs->buf = data;
2126 gs->size = size;
2127 return 0;
2130 static int grep_source_load(struct grep_source *gs)
2132 if (gs->buf)
2133 return 0;
2135 switch (gs->type) {
2136 case GREP_SOURCE_FILE:
2137 return grep_source_load_file(gs);
2138 case GREP_SOURCE_OID:
2139 return grep_source_load_oid(gs);
2140 case GREP_SOURCE_BUF:
2141 return gs->buf ? 0 : -1;
2143 BUG("invalid grep_source type to load");
2146 void grep_source_load_driver(struct grep_source *gs)
2148 if (gs->driver)
2149 return;
2151 grep_attr_lock();
2152 if (gs->path)
2153 gs->driver = userdiff_find_by_path(gs->path);
2154 if (!gs->driver)
2155 gs->driver = userdiff_find_by_name("default");
2156 grep_attr_unlock();
2159 static int grep_source_is_binary(struct grep_source *gs)
2161 grep_source_load_driver(gs);
2162 if (gs->driver->binary != -1)
2163 return gs->driver->binary;
2165 if (!grep_source_load(gs))
2166 return buffer_is_binary(gs->buf, gs->size);
2168 return 0;