Merge branch 'cc/tests-without-assuming-ref-files-backend'
[alt-git.git] / grep.c
blob7c1b8e2e8ba38f27ee4fe16e39cfe319ab89265f
1 #include "cache.h"
2 #include "config.h"
3 #include "grep.h"
4 #include "userdiff.h"
5 #include "xdiff-interface.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "commit.h"
9 #include "quote.h"
10 #include "help.h"
12 static int grep_source_load(struct grep_source *gs);
13 static int grep_source_is_binary(struct grep_source *gs);
15 static struct grep_opt grep_defaults;
17 static const char *color_grep_slots[] = {
18 [GREP_COLOR_CONTEXT] = "context",
19 [GREP_COLOR_FILENAME] = "filename",
20 [GREP_COLOR_FUNCTION] = "function",
21 [GREP_COLOR_LINENO] = "lineNumber",
22 [GREP_COLOR_MATCH_CONTEXT] = "matchContext",
23 [GREP_COLOR_MATCH_SELECTED] = "matchSelected",
24 [GREP_COLOR_SELECTED] = "selected",
25 [GREP_COLOR_SEP] = "separator",
28 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
30 fwrite(buf, size, 1, stdout);
33 static void color_set(char *dst, const char *color_bytes)
35 xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes);
39 * Initialize the grep_defaults template with hardcoded defaults.
40 * We could let the compiler do this, but without C99 initializers
41 * the code gets unwieldy and unreadable, so...
43 void init_grep_defaults(void)
45 struct grep_opt *opt = &grep_defaults;
46 static int run_once;
48 if (run_once)
49 return;
50 run_once++;
52 memset(opt, 0, sizeof(*opt));
53 opt->relative = 1;
54 opt->pathname = 1;
55 opt->max_depth = -1;
56 opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
57 color_set(opt->colors[GREP_COLOR_CONTEXT], "");
58 color_set(opt->colors[GREP_COLOR_FILENAME], "");
59 color_set(opt->colors[GREP_COLOR_FUNCTION], "");
60 color_set(opt->colors[GREP_COLOR_LINENO], "");
61 color_set(opt->colors[GREP_COLOR_MATCH_CONTEXT], GIT_COLOR_BOLD_RED);
62 color_set(opt->colors[GREP_COLOR_MATCH_SELECTED], GIT_COLOR_BOLD_RED);
63 color_set(opt->colors[GREP_COLOR_SELECTED], "");
64 color_set(opt->colors[GREP_COLOR_SEP], GIT_COLOR_CYAN);
65 opt->color = -1;
66 opt->output = std_output;
69 static int parse_pattern_type_arg(const char *opt, const char *arg)
71 if (!strcmp(arg, "default"))
72 return GREP_PATTERN_TYPE_UNSPECIFIED;
73 else if (!strcmp(arg, "basic"))
74 return GREP_PATTERN_TYPE_BRE;
75 else if (!strcmp(arg, "extended"))
76 return GREP_PATTERN_TYPE_ERE;
77 else if (!strcmp(arg, "fixed"))
78 return GREP_PATTERN_TYPE_FIXED;
79 else if (!strcmp(arg, "perl"))
80 return GREP_PATTERN_TYPE_PCRE;
81 die("bad %s argument: %s", opt, arg);
84 define_list_config_array_extra(color_grep_slots, {"match"});
87 * Read the configuration file once and store it in
88 * the grep_defaults template.
90 int grep_config(const char *var, const char *value, void *cb)
92 struct grep_opt *opt = &grep_defaults;
93 const char *slot;
95 if (userdiff_config(var, value) < 0)
96 return -1;
98 if (!strcmp(var, "grep.extendedregexp")) {
99 opt->extended_regexp_option = git_config_bool(var, value);
100 return 0;
103 if (!strcmp(var, "grep.patterntype")) {
104 opt->pattern_type_option = parse_pattern_type_arg(var, value);
105 return 0;
108 if (!strcmp(var, "grep.linenumber")) {
109 opt->linenum = git_config_bool(var, value);
110 return 0;
113 if (!strcmp(var, "grep.fullname")) {
114 opt->relative = !git_config_bool(var, value);
115 return 0;
118 if (!strcmp(var, "color.grep"))
119 opt->color = git_config_colorbool(var, value);
120 if (!strcmp(var, "color.grep.match")) {
121 if (grep_config("color.grep.matchcontext", value, cb) < 0)
122 return -1;
123 if (grep_config("color.grep.matchselected", value, cb) < 0)
124 return -1;
125 } else if (skip_prefix(var, "color.grep.", &slot)) {
126 int i = LOOKUP_CONFIG(color_grep_slots, slot);
127 char *color;
129 if (i < 0)
130 return -1;
131 color = opt->colors[i];
132 if (!value)
133 return config_error_nonbool(var);
134 return color_parse(value, color);
136 return 0;
140 * Initialize one instance of grep_opt and copy the
141 * default values from the template we read the configuration
142 * information in an earlier call to git_config(grep_config).
144 void grep_init(struct grep_opt *opt, const char *prefix)
146 struct grep_opt *def = &grep_defaults;
147 int i;
149 memset(opt, 0, sizeof(*opt));
150 opt->prefix = prefix;
151 opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
152 opt->pattern_tail = &opt->pattern_list;
153 opt->header_tail = &opt->header_list;
155 opt->color = def->color;
156 opt->extended_regexp_option = def->extended_regexp_option;
157 opt->pattern_type_option = def->pattern_type_option;
158 opt->linenum = def->linenum;
159 opt->max_depth = def->max_depth;
160 opt->pathname = def->pathname;
161 opt->relative = def->relative;
162 opt->output = def->output;
164 for (i = 0; i < NR_GREP_COLORS; i++)
165 color_set(opt->colors[i], def->colors[i]);
168 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
171 * When committing to the pattern type by setting the relevant
172 * fields in grep_opt it's generally not necessary to zero out
173 * the fields we're not choosing, since they won't have been
174 * set by anything. The extended_regexp_option field is the
175 * only exception to this.
177 * This is because in the process of parsing grep.patternType
178 * & grep.extendedRegexp we set opt->pattern_type_option and
179 * opt->extended_regexp_option, respectively. We then
180 * internally use opt->extended_regexp_option to see if we're
181 * compiling an ERE. It must be unset if that's not actually
182 * the case.
184 if (pattern_type != GREP_PATTERN_TYPE_ERE &&
185 opt->extended_regexp_option)
186 opt->extended_regexp_option = 0;
188 switch (pattern_type) {
189 case GREP_PATTERN_TYPE_UNSPECIFIED:
190 /* fall through */
192 case GREP_PATTERN_TYPE_BRE:
193 break;
195 case GREP_PATTERN_TYPE_ERE:
196 opt->extended_regexp_option = 1;
197 break;
199 case GREP_PATTERN_TYPE_FIXED:
200 opt->fixed = 1;
201 break;
203 case GREP_PATTERN_TYPE_PCRE:
204 #ifdef USE_LIBPCRE2
205 opt->pcre2 = 1;
206 #else
208 * It's important that pcre1 always be assigned to
209 * even when there's no USE_LIBPCRE* defined. We still
210 * call the PCRE stub function, it just dies with
211 * "cannot use Perl-compatible regexes[...]".
213 opt->pcre1 = 1;
214 #endif
215 break;
219 void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
221 if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
222 grep_set_pattern_type_option(pattern_type, opt);
223 else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
224 grep_set_pattern_type_option(opt->pattern_type_option, opt);
225 else if (opt->extended_regexp_option)
227 * This branch *must* happen after setting from the
228 * opt->pattern_type_option above, we don't want
229 * grep.extendedRegexp to override grep.patternType!
231 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
234 static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
235 const char *origin, int no,
236 enum grep_pat_token t,
237 enum grep_header_field field)
239 struct grep_pat *p = xcalloc(1, sizeof(*p));
240 p->pattern = xmemdupz(pat, patlen);
241 p->patternlen = patlen;
242 p->origin = origin;
243 p->no = no;
244 p->token = t;
245 p->field = field;
246 return p;
249 static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
251 **tail = p;
252 *tail = &p->next;
253 p->next = NULL;
255 switch (p->token) {
256 case GREP_PATTERN: /* atom */
257 case GREP_PATTERN_HEAD:
258 case GREP_PATTERN_BODY:
259 for (;;) {
260 struct grep_pat *new_pat;
261 size_t len = 0;
262 char *cp = p->pattern + p->patternlen, *nl = NULL;
263 while (++len <= p->patternlen) {
264 if (*(--cp) == '\n') {
265 nl = cp;
266 break;
269 if (!nl)
270 break;
271 new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
272 p->no, p->token, p->field);
273 new_pat->next = p->next;
274 if (!p->next)
275 *tail = &new_pat->next;
276 p->next = new_pat;
277 *nl = '\0';
278 p->patternlen -= len;
280 break;
281 default:
282 break;
286 void append_header_grep_pattern(struct grep_opt *opt,
287 enum grep_header_field field, const char *pat)
289 struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
290 GREP_PATTERN_HEAD, field);
291 if (field == GREP_HEADER_REFLOG)
292 opt->use_reflog_filter = 1;
293 do_append_grep_pat(&opt->header_tail, p);
296 void append_grep_pattern(struct grep_opt *opt, const char *pat,
297 const char *origin, int no, enum grep_pat_token t)
299 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
302 void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
303 const char *origin, int no, enum grep_pat_token t)
305 struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
306 do_append_grep_pat(&opt->pattern_tail, p);
309 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
311 struct grep_pat *pat;
312 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
313 *ret = *opt;
315 ret->pattern_list = NULL;
316 ret->pattern_tail = &ret->pattern_list;
318 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
320 if(pat->token == GREP_PATTERN_HEAD)
321 append_header_grep_pattern(ret, pat->field,
322 pat->pattern);
323 else
324 append_grep_pat(ret, pat->pattern, pat->patternlen,
325 pat->origin, pat->no, pat->token);
328 return ret;
331 static NORETURN void compile_regexp_failed(const struct grep_pat *p,
332 const char *error)
334 char where[1024];
336 if (p->no)
337 xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
338 else if (p->origin)
339 xsnprintf(where, sizeof(where), "%s, ", p->origin);
340 else
341 where[0] = 0;
343 die("%s'%s': %s", where, p->pattern, error);
346 static int is_fixed(const char *s, size_t len)
348 size_t i;
350 for (i = 0; i < len; i++) {
351 if (is_regex_special(s[i]))
352 return 0;
355 return 1;
358 static int has_null(const char *s, size_t len)
361 * regcomp cannot accept patterns with NULs so when using it
362 * we consider any pattern containing a NUL fixed.
364 if (memchr(s, 0, len))
365 return 1;
367 return 0;
370 #ifdef USE_LIBPCRE1
371 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
373 const char *error;
374 int erroffset;
375 int options = PCRE_MULTILINE;
377 if (opt->ignore_case) {
378 if (has_non_ascii(p->pattern))
379 p->pcre1_tables = pcre_maketables();
380 options |= PCRE_CASELESS;
382 if (is_utf8_locale() && has_non_ascii(p->pattern))
383 options |= PCRE_UTF8;
385 p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
386 p->pcre1_tables);
387 if (!p->pcre1_regexp)
388 compile_regexp_failed(p, error);
390 p->pcre1_extra_info = pcre_study(p->pcre1_regexp, GIT_PCRE_STUDY_JIT_COMPILE, &error);
391 if (!p->pcre1_extra_info && error)
392 die("%s", error);
394 #ifdef GIT_PCRE1_USE_JIT
395 pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on);
396 if (p->pcre1_jit_on == 1) {
397 p->pcre1_jit_stack = pcre_jit_stack_alloc(1, 1024 * 1024);
398 if (!p->pcre1_jit_stack)
399 die("Couldn't allocate PCRE JIT stack");
400 pcre_assign_jit_stack(p->pcre1_extra_info, NULL, p->pcre1_jit_stack);
401 } else if (p->pcre1_jit_on != 0) {
402 BUG("The pcre1_jit_on variable should be 0 or 1, not %d",
403 p->pcre1_jit_on);
405 #endif
408 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
409 regmatch_t *match, int eflags)
411 int ovector[30], ret, flags = 0;
413 if (eflags & REG_NOTBOL)
414 flags |= PCRE_NOTBOL;
416 #ifdef GIT_PCRE1_USE_JIT
417 if (p->pcre1_jit_on) {
418 ret = pcre_jit_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
419 eol - line, 0, flags, ovector,
420 ARRAY_SIZE(ovector), p->pcre1_jit_stack);
421 } else
422 #endif
424 ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
425 eol - line, 0, flags, ovector,
426 ARRAY_SIZE(ovector));
429 if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
430 die("pcre_exec failed with error code %d", ret);
431 if (ret > 0) {
432 ret = 0;
433 match->rm_so = ovector[0];
434 match->rm_eo = ovector[1];
437 return ret;
440 static void free_pcre1_regexp(struct grep_pat *p)
442 pcre_free(p->pcre1_regexp);
443 #ifdef GIT_PCRE1_USE_JIT
444 if (p->pcre1_jit_on) {
445 pcre_free_study(p->pcre1_extra_info);
446 pcre_jit_stack_free(p->pcre1_jit_stack);
447 } else
448 #endif
450 pcre_free(p->pcre1_extra_info);
452 pcre_free((void *)p->pcre1_tables);
454 #else /* !USE_LIBPCRE1 */
455 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
457 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
460 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
461 regmatch_t *match, int eflags)
463 return 1;
466 static void free_pcre1_regexp(struct grep_pat *p)
469 #endif /* !USE_LIBPCRE1 */
471 #ifdef USE_LIBPCRE2
472 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
474 int error;
475 PCRE2_UCHAR errbuf[256];
476 PCRE2_SIZE erroffset;
477 int options = PCRE2_MULTILINE;
478 const uint8_t *character_tables = NULL;
479 int jitret;
480 int patinforet;
481 size_t jitsizearg;
483 assert(opt->pcre2);
485 p->pcre2_compile_context = NULL;
487 if (opt->ignore_case) {
488 if (has_non_ascii(p->pattern)) {
489 character_tables = pcre2_maketables(NULL);
490 p->pcre2_compile_context = pcre2_compile_context_create(NULL);
491 pcre2_set_character_tables(p->pcre2_compile_context, character_tables);
493 options |= PCRE2_CASELESS;
495 if (is_utf8_locale() && has_non_ascii(p->pattern))
496 options |= PCRE2_UTF;
498 p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern,
499 p->patternlen, options, &error, &erroffset,
500 p->pcre2_compile_context);
502 if (p->pcre2_pattern) {
503 p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL);
504 if (!p->pcre2_match_data)
505 die("Couldn't allocate PCRE2 match data");
506 } else {
507 pcre2_get_error_message(error, errbuf, sizeof(errbuf));
508 compile_regexp_failed(p, (const char *)&errbuf);
511 pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on);
512 if (p->pcre2_jit_on == 1) {
513 jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE);
514 if (jitret)
515 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret);
518 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
519 * tells us whether the library itself supports JIT,
520 * but to see whether we're going to be actually using
521 * JIT we need to extract PCRE2_INFO_JITSIZE from the
522 * pattern *after* we do pcre2_jit_compile() above.
524 * This is because if the pattern contains the
525 * (*NO_JIT) verb (see pcre2syntax(3))
526 * pcre2_jit_compile() will exit early with 0. If we
527 * then proceed to call pcre2_jit_match() further down
528 * the line instead of pcre2_match() we'll either
529 * segfault (pre PCRE 10.31) or run into a fatal error
530 * (post PCRE2 10.31)
532 patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg);
533 if (patinforet)
534 BUG("pcre2_pattern_info() failed: %d", patinforet);
535 if (jitsizearg == 0) {
536 p->pcre2_jit_on = 0;
537 return;
540 p->pcre2_jit_stack = pcre2_jit_stack_create(1, 1024 * 1024, NULL);
541 if (!p->pcre2_jit_stack)
542 die("Couldn't allocate PCRE2 JIT stack");
543 p->pcre2_match_context = pcre2_match_context_create(NULL);
544 if (!p->pcre2_match_context)
545 die("Couldn't allocate PCRE2 match context");
546 pcre2_jit_stack_assign(p->pcre2_match_context, NULL, p->pcre2_jit_stack);
547 } else if (p->pcre2_jit_on != 0) {
548 BUG("The pcre2_jit_on variable should be 0 or 1, not %d",
549 p->pcre1_jit_on);
553 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
554 regmatch_t *match, int eflags)
556 int ret, flags = 0;
557 PCRE2_SIZE *ovector;
558 PCRE2_UCHAR errbuf[256];
560 if (eflags & REG_NOTBOL)
561 flags |= PCRE2_NOTBOL;
563 if (p->pcre2_jit_on)
564 ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line,
565 eol - line, 0, flags, p->pcre2_match_data,
566 NULL);
567 else
568 ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line,
569 eol - line, 0, flags, p->pcre2_match_data,
570 NULL);
572 if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) {
573 pcre2_get_error_message(ret, errbuf, sizeof(errbuf));
574 die("%s failed with error code %d: %s",
575 (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret,
576 errbuf);
578 if (ret > 0) {
579 ovector = pcre2_get_ovector_pointer(p->pcre2_match_data);
580 ret = 0;
581 match->rm_so = (int)ovector[0];
582 match->rm_eo = (int)ovector[1];
585 return ret;
588 static void free_pcre2_pattern(struct grep_pat *p)
590 pcre2_compile_context_free(p->pcre2_compile_context);
591 pcre2_code_free(p->pcre2_pattern);
592 pcre2_match_data_free(p->pcre2_match_data);
593 pcre2_jit_stack_free(p->pcre2_jit_stack);
594 pcre2_match_context_free(p->pcre2_match_context);
596 #else /* !USE_LIBPCRE2 */
597 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
600 * Unreachable until USE_LIBPCRE2 becomes synonymous with
601 * USE_LIBPCRE. See the sibling comment in
602 * grep_set_pattern_type_option().
604 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
607 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
608 regmatch_t *match, int eflags)
610 return 1;
613 static void free_pcre2_pattern(struct grep_pat *p)
616 #endif /* !USE_LIBPCRE2 */
618 static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
620 struct strbuf sb = STRBUF_INIT;
621 int err;
622 int regflags = 0;
624 basic_regex_quote_buf(&sb, p->pattern);
625 if (opt->ignore_case)
626 regflags |= REG_ICASE;
627 err = regcomp(&p->regexp, sb.buf, regflags);
628 if (opt->debug)
629 fprintf(stderr, "fixed %s\n", sb.buf);
630 strbuf_release(&sb);
631 if (err) {
632 char errbuf[1024];
633 regerror(err, &p->regexp, errbuf, sizeof(errbuf));
634 compile_regexp_failed(p, errbuf);
638 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
640 int ascii_only;
641 int err;
642 int regflags = REG_NEWLINE;
644 p->word_regexp = opt->word_regexp;
645 p->ignore_case = opt->ignore_case;
646 ascii_only = !has_non_ascii(p->pattern);
649 * Even when -F (fixed) asks us to do a non-regexp search, we
650 * may not be able to correctly case-fold when -i
651 * (ignore-case) is asked (in which case, we'll synthesize a
652 * regexp to match the pattern that matches regexp special
653 * characters literally, while ignoring case differences). On
654 * the other hand, even without -F, if the pattern does not
655 * have any regexp special characters and there is no need for
656 * case-folding search, we can internally turn it into a
657 * simple string match using kws. p->fixed tells us if we
658 * want to use kws.
660 if (opt->fixed ||
661 has_null(p->pattern, p->patternlen) ||
662 is_fixed(p->pattern, p->patternlen))
663 p->fixed = !p->ignore_case || ascii_only;
665 if (p->fixed) {
666 p->kws = kwsalloc(p->ignore_case ? tolower_trans_tbl : NULL);
667 kwsincr(p->kws, p->pattern, p->patternlen);
668 kwsprep(p->kws);
669 return;
670 } else if (opt->fixed) {
672 * We come here when the pattern has the non-ascii
673 * characters we cannot case-fold, and asked to
674 * ignore-case.
676 compile_fixed_regexp(p, opt);
677 return;
680 if (opt->pcre2) {
681 compile_pcre2_pattern(p, opt);
682 return;
685 if (opt->pcre1) {
686 compile_pcre1_regexp(p, opt);
687 return;
690 if (p->ignore_case)
691 regflags |= REG_ICASE;
692 if (opt->extended_regexp_option)
693 regflags |= REG_EXTENDED;
694 err = regcomp(&p->regexp, p->pattern, regflags);
695 if (err) {
696 char errbuf[1024];
697 regerror(err, &p->regexp, errbuf, 1024);
698 compile_regexp_failed(p, errbuf);
702 static struct grep_expr *compile_pattern_or(struct grep_pat **);
703 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
705 struct grep_pat *p;
706 struct grep_expr *x;
708 p = *list;
709 if (!p)
710 return NULL;
711 switch (p->token) {
712 case GREP_PATTERN: /* atom */
713 case GREP_PATTERN_HEAD:
714 case GREP_PATTERN_BODY:
715 x = xcalloc(1, sizeof (struct grep_expr));
716 x->node = GREP_NODE_ATOM;
717 x->u.atom = p;
718 *list = p->next;
719 return x;
720 case GREP_OPEN_PAREN:
721 *list = p->next;
722 x = compile_pattern_or(list);
723 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
724 die("unmatched parenthesis");
725 *list = (*list)->next;
726 return x;
727 default:
728 return NULL;
732 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
734 struct grep_pat *p;
735 struct grep_expr *x;
737 p = *list;
738 if (!p)
739 return NULL;
740 switch (p->token) {
741 case GREP_NOT:
742 if (!p->next)
743 die("--not not followed by pattern expression");
744 *list = p->next;
745 x = xcalloc(1, sizeof (struct grep_expr));
746 x->node = GREP_NODE_NOT;
747 x->u.unary = compile_pattern_not(list);
748 if (!x->u.unary)
749 die("--not followed by non pattern expression");
750 return x;
751 default:
752 return compile_pattern_atom(list);
756 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
758 struct grep_pat *p;
759 struct grep_expr *x, *y, *z;
761 x = compile_pattern_not(list);
762 p = *list;
763 if (p && p->token == GREP_AND) {
764 if (!p->next)
765 die("--and not followed by pattern expression");
766 *list = p->next;
767 y = compile_pattern_and(list);
768 if (!y)
769 die("--and not followed by pattern expression");
770 z = xcalloc(1, sizeof (struct grep_expr));
771 z->node = GREP_NODE_AND;
772 z->u.binary.left = x;
773 z->u.binary.right = y;
774 return z;
776 return x;
779 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
781 struct grep_pat *p;
782 struct grep_expr *x, *y, *z;
784 x = compile_pattern_and(list);
785 p = *list;
786 if (x && p && p->token != GREP_CLOSE_PAREN) {
787 y = compile_pattern_or(list);
788 if (!y)
789 die("not a pattern expression %s", p->pattern);
790 z = xcalloc(1, sizeof (struct grep_expr));
791 z->node = GREP_NODE_OR;
792 z->u.binary.left = x;
793 z->u.binary.right = y;
794 return z;
796 return x;
799 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
801 return compile_pattern_or(list);
804 static void indent(int in)
806 while (in-- > 0)
807 fputc(' ', stderr);
810 static void dump_grep_pat(struct grep_pat *p)
812 switch (p->token) {
813 case GREP_AND: fprintf(stderr, "*and*"); break;
814 case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
815 case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
816 case GREP_NOT: fprintf(stderr, "*not*"); break;
817 case GREP_OR: fprintf(stderr, "*or*"); break;
819 case GREP_PATTERN: fprintf(stderr, "pattern"); break;
820 case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
821 case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
824 switch (p->token) {
825 default: break;
826 case GREP_PATTERN_HEAD:
827 fprintf(stderr, "<head %d>", p->field); break;
828 case GREP_PATTERN_BODY:
829 fprintf(stderr, "<body>"); break;
831 switch (p->token) {
832 default: break;
833 case GREP_PATTERN_HEAD:
834 case GREP_PATTERN_BODY:
835 case GREP_PATTERN:
836 fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
837 break;
839 fputc('\n', stderr);
842 static void dump_grep_expression_1(struct grep_expr *x, int in)
844 indent(in);
845 switch (x->node) {
846 case GREP_NODE_TRUE:
847 fprintf(stderr, "true\n");
848 break;
849 case GREP_NODE_ATOM:
850 dump_grep_pat(x->u.atom);
851 break;
852 case GREP_NODE_NOT:
853 fprintf(stderr, "(not\n");
854 dump_grep_expression_1(x->u.unary, in+1);
855 indent(in);
856 fprintf(stderr, ")\n");
857 break;
858 case GREP_NODE_AND:
859 fprintf(stderr, "(and\n");
860 dump_grep_expression_1(x->u.binary.left, in+1);
861 dump_grep_expression_1(x->u.binary.right, in+1);
862 indent(in);
863 fprintf(stderr, ")\n");
864 break;
865 case GREP_NODE_OR:
866 fprintf(stderr, "(or\n");
867 dump_grep_expression_1(x->u.binary.left, in+1);
868 dump_grep_expression_1(x->u.binary.right, in+1);
869 indent(in);
870 fprintf(stderr, ")\n");
871 break;
875 static void dump_grep_expression(struct grep_opt *opt)
877 struct grep_expr *x = opt->pattern_expression;
879 if (opt->all_match)
880 fprintf(stderr, "[all-match]\n");
881 dump_grep_expression_1(x, 0);
882 fflush(NULL);
885 static struct grep_expr *grep_true_expr(void)
887 struct grep_expr *z = xcalloc(1, sizeof(*z));
888 z->node = GREP_NODE_TRUE;
889 return z;
892 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
894 struct grep_expr *z = xcalloc(1, sizeof(*z));
895 z->node = GREP_NODE_OR;
896 z->u.binary.left = left;
897 z->u.binary.right = right;
898 return z;
901 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
903 struct grep_pat *p;
904 struct grep_expr *header_expr;
905 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
906 enum grep_header_field fld;
908 if (!opt->header_list)
909 return NULL;
911 for (p = opt->header_list; p; p = p->next) {
912 if (p->token != GREP_PATTERN_HEAD)
913 BUG("a non-header pattern in grep header list.");
914 if (p->field < GREP_HEADER_FIELD_MIN ||
915 GREP_HEADER_FIELD_MAX <= p->field)
916 BUG("unknown header field %d", p->field);
917 compile_regexp(p, opt);
920 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
921 header_group[fld] = NULL;
923 for (p = opt->header_list; p; p = p->next) {
924 struct grep_expr *h;
925 struct grep_pat *pp = p;
927 h = compile_pattern_atom(&pp);
928 if (!h || pp != p->next)
929 BUG("malformed header expr");
930 if (!header_group[p->field]) {
931 header_group[p->field] = h;
932 continue;
934 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
937 header_expr = NULL;
939 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
940 if (!header_group[fld])
941 continue;
942 if (!header_expr)
943 header_expr = grep_true_expr();
944 header_expr = grep_or_expr(header_group[fld], header_expr);
946 return header_expr;
949 static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
951 struct grep_expr *z = x;
953 while (x) {
954 assert(x->node == GREP_NODE_OR);
955 if (x->u.binary.right &&
956 x->u.binary.right->node == GREP_NODE_TRUE) {
957 x->u.binary.right = y;
958 break;
960 x = x->u.binary.right;
962 return z;
965 static void compile_grep_patterns_real(struct grep_opt *opt)
967 struct grep_pat *p;
968 struct grep_expr *header_expr = prep_header_patterns(opt);
970 for (p = opt->pattern_list; p; p = p->next) {
971 switch (p->token) {
972 case GREP_PATTERN: /* atom */
973 case GREP_PATTERN_HEAD:
974 case GREP_PATTERN_BODY:
975 compile_regexp(p, opt);
976 break;
977 default:
978 opt->extended = 1;
979 break;
983 if (opt->all_match || header_expr)
984 opt->extended = 1;
985 else if (!opt->extended && !opt->debug)
986 return;
988 p = opt->pattern_list;
989 if (p)
990 opt->pattern_expression = compile_pattern_expr(&p);
991 if (p)
992 die("incomplete pattern expression: %s", p->pattern);
994 if (!header_expr)
995 return;
997 if (!opt->pattern_expression)
998 opt->pattern_expression = header_expr;
999 else if (opt->all_match)
1000 opt->pattern_expression = grep_splice_or(header_expr,
1001 opt->pattern_expression);
1002 else
1003 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
1004 header_expr);
1005 opt->all_match = 1;
1008 void compile_grep_patterns(struct grep_opt *opt)
1010 compile_grep_patterns_real(opt);
1011 if (opt->debug)
1012 dump_grep_expression(opt);
1015 static void free_pattern_expr(struct grep_expr *x)
1017 switch (x->node) {
1018 case GREP_NODE_TRUE:
1019 case GREP_NODE_ATOM:
1020 break;
1021 case GREP_NODE_NOT:
1022 free_pattern_expr(x->u.unary);
1023 break;
1024 case GREP_NODE_AND:
1025 case GREP_NODE_OR:
1026 free_pattern_expr(x->u.binary.left);
1027 free_pattern_expr(x->u.binary.right);
1028 break;
1030 free(x);
1033 void free_grep_patterns(struct grep_opt *opt)
1035 struct grep_pat *p, *n;
1037 for (p = opt->pattern_list; p; p = n) {
1038 n = p->next;
1039 switch (p->token) {
1040 case GREP_PATTERN: /* atom */
1041 case GREP_PATTERN_HEAD:
1042 case GREP_PATTERN_BODY:
1043 if (p->kws)
1044 kwsfree(p->kws);
1045 else if (p->pcre1_regexp)
1046 free_pcre1_regexp(p);
1047 else if (p->pcre2_pattern)
1048 free_pcre2_pattern(p);
1049 else
1050 regfree(&p->regexp);
1051 free(p->pattern);
1052 break;
1053 default:
1054 break;
1056 free(p);
1059 if (!opt->extended)
1060 return;
1061 free_pattern_expr(opt->pattern_expression);
1064 static char *end_of_line(char *cp, unsigned long *left)
1066 unsigned long l = *left;
1067 while (l && *cp != '\n') {
1068 l--;
1069 cp++;
1071 *left = l;
1072 return cp;
1075 static int word_char(char ch)
1077 return isalnum(ch) || ch == '_';
1080 static void output_color(struct grep_opt *opt, const void *data, size_t size,
1081 const char *color)
1083 if (want_color(opt->color) && color && color[0]) {
1084 opt->output(opt, color, strlen(color));
1085 opt->output(opt, data, size);
1086 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
1087 } else
1088 opt->output(opt, data, size);
1091 static void output_sep(struct grep_opt *opt, char sign)
1093 if (opt->null_following_name)
1094 opt->output(opt, "\0", 1);
1095 else
1096 output_color(opt, &sign, 1, opt->colors[GREP_COLOR_SEP]);
1099 static void show_name(struct grep_opt *opt, const char *name)
1101 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1102 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
1105 static int fixmatch(struct grep_pat *p, char *line, char *eol,
1106 regmatch_t *match)
1108 struct kwsmatch kwsm;
1109 size_t offset = kwsexec(p->kws, line, eol - line, &kwsm);
1110 if (offset == -1) {
1111 match->rm_so = match->rm_eo = -1;
1112 return REG_NOMATCH;
1113 } else {
1114 match->rm_so = offset;
1115 match->rm_eo = match->rm_so + kwsm.size[0];
1116 return 0;
1120 static int patmatch(struct grep_pat *p, char *line, char *eol,
1121 regmatch_t *match, int eflags)
1123 int hit;
1125 if (p->fixed)
1126 hit = !fixmatch(p, line, eol, match);
1127 else if (p->pcre1_regexp)
1128 hit = !pcre1match(p, line, eol, match, eflags);
1129 else if (p->pcre2_pattern)
1130 hit = !pcre2match(p, line, eol, match, eflags);
1131 else
1132 hit = !regexec_buf(&p->regexp, line, eol - line, 1, match,
1133 eflags);
1135 return hit;
1138 static int strip_timestamp(char *bol, char **eol_p)
1140 char *eol = *eol_p;
1141 int ch;
1143 while (bol < --eol) {
1144 if (*eol != '>')
1145 continue;
1146 *eol_p = ++eol;
1147 ch = *eol;
1148 *eol = '\0';
1149 return ch;
1151 return 0;
1154 static struct {
1155 const char *field;
1156 size_t len;
1157 } header_field[] = {
1158 { "author ", 7 },
1159 { "committer ", 10 },
1160 { "reflog ", 7 },
1163 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
1164 enum grep_context ctx,
1165 regmatch_t *pmatch, int eflags)
1167 int hit = 0;
1168 int saved_ch = 0;
1169 const char *start = bol;
1171 if ((p->token != GREP_PATTERN) &&
1172 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
1173 return 0;
1175 if (p->token == GREP_PATTERN_HEAD) {
1176 const char *field;
1177 size_t len;
1178 assert(p->field < ARRAY_SIZE(header_field));
1179 field = header_field[p->field].field;
1180 len = header_field[p->field].len;
1181 if (strncmp(bol, field, len))
1182 return 0;
1183 bol += len;
1184 switch (p->field) {
1185 case GREP_HEADER_AUTHOR:
1186 case GREP_HEADER_COMMITTER:
1187 saved_ch = strip_timestamp(bol, &eol);
1188 break;
1189 default:
1190 break;
1194 again:
1195 hit = patmatch(p, bol, eol, pmatch, eflags);
1197 if (hit && p->word_regexp) {
1198 if ((pmatch[0].rm_so < 0) ||
1199 (eol - bol) < pmatch[0].rm_so ||
1200 (pmatch[0].rm_eo < 0) ||
1201 (eol - bol) < pmatch[0].rm_eo)
1202 die("regexp returned nonsense");
1204 /* Match beginning must be either beginning of the
1205 * line, or at word boundary (i.e. the last char must
1206 * not be a word char). Similarly, match end must be
1207 * either end of the line, or at word boundary
1208 * (i.e. the next char must not be a word char).
1210 if ( ((pmatch[0].rm_so == 0) ||
1211 !word_char(bol[pmatch[0].rm_so-1])) &&
1212 ((pmatch[0].rm_eo == (eol-bol)) ||
1213 !word_char(bol[pmatch[0].rm_eo])) )
1215 else
1216 hit = 0;
1218 /* Words consist of at least one character. */
1219 if (pmatch->rm_so == pmatch->rm_eo)
1220 hit = 0;
1222 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
1223 /* There could be more than one match on the
1224 * line, and the first match might not be
1225 * strict word match. But later ones could be!
1226 * Forward to the next possible start, i.e. the
1227 * next position following a non-word char.
1229 bol = pmatch[0].rm_so + bol + 1;
1230 while (word_char(bol[-1]) && bol < eol)
1231 bol++;
1232 eflags |= REG_NOTBOL;
1233 if (bol < eol)
1234 goto again;
1237 if (p->token == GREP_PATTERN_HEAD && saved_ch)
1238 *eol = saved_ch;
1239 if (hit) {
1240 pmatch[0].rm_so += bol - start;
1241 pmatch[0].rm_eo += bol - start;
1243 return hit;
1246 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
1247 enum grep_context ctx, int collect_hits)
1249 int h = 0;
1250 regmatch_t match;
1252 if (!x)
1253 die("Not a valid grep expression");
1254 switch (x->node) {
1255 case GREP_NODE_TRUE:
1256 h = 1;
1257 break;
1258 case GREP_NODE_ATOM:
1259 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
1260 break;
1261 case GREP_NODE_NOT:
1262 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
1263 break;
1264 case GREP_NODE_AND:
1265 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
1266 return 0;
1267 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
1268 break;
1269 case GREP_NODE_OR:
1270 if (!collect_hits)
1271 return (match_expr_eval(x->u.binary.left,
1272 bol, eol, ctx, 0) ||
1273 match_expr_eval(x->u.binary.right,
1274 bol, eol, ctx, 0));
1275 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
1276 x->u.binary.left->hit |= h;
1277 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
1278 break;
1279 default:
1280 die("Unexpected node type (internal error) %d", x->node);
1282 if (collect_hits)
1283 x->hit |= h;
1284 return h;
1287 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
1288 enum grep_context ctx, int collect_hits)
1290 struct grep_expr *x = opt->pattern_expression;
1291 return match_expr_eval(x, bol, eol, ctx, collect_hits);
1294 static int match_line(struct grep_opt *opt, char *bol, char *eol,
1295 enum grep_context ctx, int collect_hits)
1297 struct grep_pat *p;
1298 regmatch_t match;
1300 if (opt->extended)
1301 return match_expr(opt, bol, eol, ctx, collect_hits);
1303 /* we do not call with collect_hits without being extended */
1304 for (p = opt->pattern_list; p; p = p->next) {
1305 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
1306 return 1;
1308 return 0;
1311 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
1312 enum grep_context ctx,
1313 regmatch_t *pmatch, int eflags)
1315 regmatch_t match;
1317 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
1318 return 0;
1319 if (match.rm_so < 0 || match.rm_eo < 0)
1320 return 0;
1321 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
1322 if (match.rm_so > pmatch->rm_so)
1323 return 1;
1324 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
1325 return 1;
1327 pmatch->rm_so = match.rm_so;
1328 pmatch->rm_eo = match.rm_eo;
1329 return 1;
1332 static int next_match(struct grep_opt *opt, char *bol, char *eol,
1333 enum grep_context ctx, regmatch_t *pmatch, int eflags)
1335 struct grep_pat *p;
1336 int hit = 0;
1338 pmatch->rm_so = pmatch->rm_eo = -1;
1339 if (bol < eol) {
1340 for (p = opt->pattern_list; p; p = p->next) {
1341 switch (p->token) {
1342 case GREP_PATTERN: /* atom */
1343 case GREP_PATTERN_HEAD:
1344 case GREP_PATTERN_BODY:
1345 hit |= match_next_pattern(p, bol, eol, ctx,
1346 pmatch, eflags);
1347 break;
1348 default:
1349 break;
1353 return hit;
1356 static void show_line(struct grep_opt *opt, char *bol, char *eol,
1357 const char *name, unsigned lno, char sign)
1359 int rest = eol - bol;
1360 const char *match_color, *line_color = NULL;
1362 if (opt->file_break && opt->last_shown == 0) {
1363 if (opt->show_hunk_mark)
1364 opt->output(opt, "\n", 1);
1365 } else if (opt->pre_context || opt->post_context || opt->funcbody) {
1366 if (opt->last_shown == 0) {
1367 if (opt->show_hunk_mark) {
1368 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1369 opt->output(opt, "\n", 1);
1371 } else if (lno > opt->last_shown + 1) {
1372 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1373 opt->output(opt, "\n", 1);
1376 if (opt->heading && opt->last_shown == 0) {
1377 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1378 opt->output(opt, "\n", 1);
1380 opt->last_shown = lno;
1382 if (!opt->heading && opt->pathname) {
1383 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1384 output_sep(opt, sign);
1386 if (opt->linenum) {
1387 char buf[32];
1388 xsnprintf(buf, sizeof(buf), "%d", lno);
1389 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_LINENO]);
1390 output_sep(opt, sign);
1392 if (opt->color) {
1393 regmatch_t match;
1394 enum grep_context ctx = GREP_CONTEXT_BODY;
1395 int ch = *eol;
1396 int eflags = 0;
1398 if (sign == ':')
1399 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
1400 else
1401 match_color = opt->colors[GREP_COLOR_MATCH_CONTEXT];
1402 if (sign == ':')
1403 line_color = opt->colors[GREP_COLOR_SELECTED];
1404 else if (sign == '-')
1405 line_color = opt->colors[GREP_COLOR_CONTEXT];
1406 else if (sign == '=')
1407 line_color = opt->colors[GREP_COLOR_FUNCTION];
1408 *eol = '\0';
1409 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
1410 if (match.rm_so == match.rm_eo)
1411 break;
1413 output_color(opt, bol, match.rm_so, line_color);
1414 output_color(opt, bol + match.rm_so,
1415 match.rm_eo - match.rm_so, match_color);
1416 bol += match.rm_eo;
1417 rest -= match.rm_eo;
1418 eflags = REG_NOTBOL;
1420 *eol = ch;
1422 output_color(opt, bol, rest, line_color);
1423 opt->output(opt, "\n", 1);
1426 #ifndef NO_PTHREADS
1427 int grep_use_locks;
1430 * This lock protects access to the gitattributes machinery, which is
1431 * not thread-safe.
1433 pthread_mutex_t grep_attr_mutex;
1435 static inline void grep_attr_lock(void)
1437 if (grep_use_locks)
1438 pthread_mutex_lock(&grep_attr_mutex);
1441 static inline void grep_attr_unlock(void)
1443 if (grep_use_locks)
1444 pthread_mutex_unlock(&grep_attr_mutex);
1448 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1450 pthread_mutex_t grep_read_mutex;
1452 #else
1453 #define grep_attr_lock()
1454 #define grep_attr_unlock()
1455 #endif
1457 static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
1459 xdemitconf_t *xecfg = opt->priv;
1460 if (xecfg && !xecfg->find_func) {
1461 grep_source_load_driver(gs);
1462 if (gs->driver->funcname.pattern) {
1463 const struct userdiff_funcname *pe = &gs->driver->funcname;
1464 xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
1465 } else {
1466 xecfg = opt->priv = NULL;
1470 if (xecfg) {
1471 char buf[1];
1472 return xecfg->find_func(bol, eol - bol, buf, 1,
1473 xecfg->find_func_priv) >= 0;
1476 if (bol == eol)
1477 return 0;
1478 if (isalpha(*bol) || *bol == '_' || *bol == '$')
1479 return 1;
1480 return 0;
1483 static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
1484 char *bol, unsigned lno)
1486 while (bol > gs->buf) {
1487 char *eol = --bol;
1489 while (bol > gs->buf && bol[-1] != '\n')
1490 bol--;
1491 lno--;
1493 if (lno <= opt->last_shown)
1494 break;
1496 if (match_funcname(opt, gs, bol, eol)) {
1497 show_line(opt, bol, eol, gs->name, lno, '=');
1498 break;
1503 static int is_empty_line(const char *bol, const char *eol);
1505 static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
1506 char *bol, char *end, unsigned lno)
1508 unsigned cur = lno, from = 1, funcname_lno = 0, orig_from;
1509 int funcname_needed = !!opt->funcname, comment_needed = 0;
1511 if (opt->pre_context < lno)
1512 from = lno - opt->pre_context;
1513 if (from <= opt->last_shown)
1514 from = opt->last_shown + 1;
1515 orig_from = from;
1516 if (opt->funcbody) {
1517 if (match_funcname(opt, gs, bol, end))
1518 comment_needed = 1;
1519 else
1520 funcname_needed = 1;
1521 from = opt->last_shown + 1;
1524 /* Rewind. */
1525 while (bol > gs->buf && cur > from) {
1526 char *next_bol = bol;
1527 char *eol = --bol;
1529 while (bol > gs->buf && bol[-1] != '\n')
1530 bol--;
1531 cur--;
1532 if (comment_needed && (is_empty_line(bol, eol) ||
1533 match_funcname(opt, gs, bol, eol))) {
1534 comment_needed = 0;
1535 from = orig_from;
1536 if (cur < from) {
1537 cur++;
1538 bol = next_bol;
1539 break;
1542 if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
1543 funcname_lno = cur;
1544 funcname_needed = 0;
1545 if (opt->funcbody)
1546 comment_needed = 1;
1547 else
1548 from = orig_from;
1552 /* We need to look even further back to find a function signature. */
1553 if (opt->funcname && funcname_needed)
1554 show_funcname_line(opt, gs, bol, cur);
1556 /* Back forward. */
1557 while (cur < lno) {
1558 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
1560 while (*eol != '\n')
1561 eol++;
1562 show_line(opt, bol, eol, gs->name, cur, sign);
1563 bol = eol + 1;
1564 cur++;
1568 static int should_lookahead(struct grep_opt *opt)
1570 struct grep_pat *p;
1572 if (opt->extended)
1573 return 0; /* punt for too complex stuff */
1574 if (opt->invert)
1575 return 0;
1576 for (p = opt->pattern_list; p; p = p->next) {
1577 if (p->token != GREP_PATTERN)
1578 return 0; /* punt for "header only" and stuff */
1580 return 1;
1583 static int look_ahead(struct grep_opt *opt,
1584 unsigned long *left_p,
1585 unsigned *lno_p,
1586 char **bol_p)
1588 unsigned lno = *lno_p;
1589 char *bol = *bol_p;
1590 struct grep_pat *p;
1591 char *sp, *last_bol;
1592 regoff_t earliest = -1;
1594 for (p = opt->pattern_list; p; p = p->next) {
1595 int hit;
1596 regmatch_t m;
1598 hit = patmatch(p, bol, bol + *left_p, &m, 0);
1599 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
1600 continue;
1601 if (earliest < 0 || m.rm_so < earliest)
1602 earliest = m.rm_so;
1605 if (earliest < 0) {
1606 *bol_p = bol + *left_p;
1607 *left_p = 0;
1608 return 1;
1610 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
1611 ; /* find the beginning of the line */
1612 last_bol = sp;
1614 for (sp = bol; sp < last_bol; sp++) {
1615 if (*sp == '\n')
1616 lno++;
1618 *left_p -= last_bol - bol;
1619 *bol_p = last_bol;
1620 *lno_p = lno;
1621 return 0;
1624 static int fill_textconv_grep(struct userdiff_driver *driver,
1625 struct grep_source *gs)
1627 struct diff_filespec *df;
1628 char *buf;
1629 size_t size;
1631 if (!driver || !driver->textconv)
1632 return grep_source_load(gs);
1635 * The textconv interface is intimately tied to diff_filespecs, so we
1636 * have to pretend to be one. If we could unify the grep_source
1637 * and diff_filespec structs, this mess could just go away.
1639 df = alloc_filespec(gs->path);
1640 switch (gs->type) {
1641 case GREP_SOURCE_OID:
1642 fill_filespec(df, gs->identifier, 1, 0100644);
1643 break;
1644 case GREP_SOURCE_FILE:
1645 fill_filespec(df, &null_oid, 0, 0100644);
1646 break;
1647 default:
1648 BUG("attempt to textconv something without a path?");
1652 * fill_textconv is not remotely thread-safe; it may load objects
1653 * behind the scenes, and it modifies the global diff tempfile
1654 * structure.
1656 grep_read_lock();
1657 size = fill_textconv(driver, df, &buf);
1658 grep_read_unlock();
1659 free_filespec(df);
1662 * The normal fill_textconv usage by the diff machinery would just keep
1663 * the textconv'd buf separate from the diff_filespec. But much of the
1664 * grep code passes around a grep_source and assumes that its "buf"
1665 * pointer is the beginning of the thing we are searching. So let's
1666 * install our textconv'd version into the grep_source, taking care not
1667 * to leak any existing buffer.
1669 grep_source_clear_data(gs);
1670 gs->buf = buf;
1671 gs->size = size;
1673 return 0;
1676 static int is_empty_line(const char *bol, const char *eol)
1678 while (bol < eol && isspace(*bol))
1679 bol++;
1680 return bol == eol;
1683 static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
1685 char *bol;
1686 char *peek_bol = NULL;
1687 unsigned long left;
1688 unsigned lno = 1;
1689 unsigned last_hit = 0;
1690 int binary_match_only = 0;
1691 unsigned count = 0;
1692 int try_lookahead = 0;
1693 int show_function = 0;
1694 struct userdiff_driver *textconv = NULL;
1695 enum grep_context ctx = GREP_CONTEXT_HEAD;
1696 xdemitconf_t xecfg;
1698 if (!opt->output)
1699 opt->output = std_output;
1701 if (opt->pre_context || opt->post_context || opt->file_break ||
1702 opt->funcbody) {
1703 /* Show hunk marks, except for the first file. */
1704 if (opt->last_shown)
1705 opt->show_hunk_mark = 1;
1707 * If we're using threads then we can't easily identify
1708 * the first file. Always put hunk marks in that case
1709 * and skip the very first one later in work_done().
1711 if (opt->output != std_output)
1712 opt->show_hunk_mark = 1;
1714 opt->last_shown = 0;
1716 if (opt->allow_textconv) {
1717 grep_source_load_driver(gs);
1719 * We might set up the shared textconv cache data here, which
1720 * is not thread-safe.
1722 grep_attr_lock();
1723 textconv = userdiff_get_textconv(gs->driver);
1724 grep_attr_unlock();
1728 * We know the result of a textconv is text, so we only have to care
1729 * about binary handling if we are not using it.
1731 if (!textconv) {
1732 switch (opt->binary) {
1733 case GREP_BINARY_DEFAULT:
1734 if (grep_source_is_binary(gs))
1735 binary_match_only = 1;
1736 break;
1737 case GREP_BINARY_NOMATCH:
1738 if (grep_source_is_binary(gs))
1739 return 0; /* Assume unmatch */
1740 break;
1741 case GREP_BINARY_TEXT:
1742 break;
1743 default:
1744 BUG("unknown binary handling mode");
1748 memset(&xecfg, 0, sizeof(xecfg));
1749 opt->priv = &xecfg;
1751 try_lookahead = should_lookahead(opt);
1753 if (fill_textconv_grep(textconv, gs) < 0)
1754 return 0;
1756 bol = gs->buf;
1757 left = gs->size;
1758 while (left) {
1759 char *eol, ch;
1760 int hit;
1763 * look_ahead() skips quickly to the line that possibly
1764 * has the next hit; don't call it if we need to do
1765 * something more than just skipping the current line
1766 * in response to an unmatch for the current line. E.g.
1767 * inside a post-context window, we will show the current
1768 * line as a context around the previous hit when it
1769 * doesn't hit.
1771 if (try_lookahead
1772 && !(last_hit
1773 && (show_function ||
1774 lno <= last_hit + opt->post_context))
1775 && look_ahead(opt, &left, &lno, &bol))
1776 break;
1777 eol = end_of_line(bol, &left);
1778 ch = *eol;
1779 *eol = 0;
1781 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
1782 ctx = GREP_CONTEXT_BODY;
1784 hit = match_line(opt, bol, eol, ctx, collect_hits);
1785 *eol = ch;
1787 if (collect_hits)
1788 goto next_line;
1790 /* "grep -v -e foo -e bla" should list lines
1791 * that do not have either, so inversion should
1792 * be done outside.
1794 if (opt->invert)
1795 hit = !hit;
1796 if (opt->unmatch_name_only) {
1797 if (hit)
1798 return 0;
1799 goto next_line;
1801 if (hit) {
1802 count++;
1803 if (opt->status_only)
1804 return 1;
1805 if (opt->name_only) {
1806 show_name(opt, gs->name);
1807 return 1;
1809 if (opt->count)
1810 goto next_line;
1811 if (binary_match_only) {
1812 opt->output(opt, "Binary file ", 12);
1813 output_color(opt, gs->name, strlen(gs->name),
1814 opt->colors[GREP_COLOR_FILENAME]);
1815 opt->output(opt, " matches\n", 9);
1816 return 1;
1818 /* Hit at this line. If we haven't shown the
1819 * pre-context lines, we would need to show them.
1821 if (opt->pre_context || opt->funcbody)
1822 show_pre_context(opt, gs, bol, eol, lno);
1823 else if (opt->funcname)
1824 show_funcname_line(opt, gs, bol, lno);
1825 show_line(opt, bol, eol, gs->name, lno, ':');
1826 last_hit = lno;
1827 if (opt->funcbody)
1828 show_function = 1;
1829 goto next_line;
1831 if (show_function && (!peek_bol || peek_bol < bol)) {
1832 unsigned long peek_left = left;
1833 char *peek_eol = eol;
1836 * Trailing empty lines are not interesting.
1837 * Peek past them to see if they belong to the
1838 * body of the current function.
1840 peek_bol = bol;
1841 while (is_empty_line(peek_bol, peek_eol)) {
1842 peek_bol = peek_eol + 1;
1843 peek_eol = end_of_line(peek_bol, &peek_left);
1846 if (match_funcname(opt, gs, peek_bol, peek_eol))
1847 show_function = 0;
1849 if (show_function ||
1850 (last_hit && lno <= last_hit + opt->post_context)) {
1851 /* If the last hit is within the post context,
1852 * we need to show this line.
1854 show_line(opt, bol, eol, gs->name, lno, '-');
1857 next_line:
1858 bol = eol + 1;
1859 if (!left)
1860 break;
1861 left--;
1862 lno++;
1865 if (collect_hits)
1866 return 0;
1868 if (opt->status_only)
1869 return opt->unmatch_name_only;
1870 if (opt->unmatch_name_only) {
1871 /* We did not see any hit, so we want to show this */
1872 show_name(opt, gs->name);
1873 return 1;
1876 xdiff_clear_find_func(&xecfg);
1877 opt->priv = NULL;
1879 /* NEEDSWORK:
1880 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1881 * which feels mostly useless but sometimes useful. Maybe
1882 * make it another option? For now suppress them.
1884 if (opt->count && count) {
1885 char buf[32];
1886 if (opt->pathname) {
1887 output_color(opt, gs->name, strlen(gs->name),
1888 opt->colors[GREP_COLOR_FILENAME]);
1889 output_sep(opt, ':');
1891 xsnprintf(buf, sizeof(buf), "%u\n", count);
1892 opt->output(opt, buf, strlen(buf));
1893 return 1;
1895 return !!last_hit;
1898 static void clr_hit_marker(struct grep_expr *x)
1900 /* All-hit markers are meaningful only at the very top level
1901 * OR node.
1903 while (1) {
1904 x->hit = 0;
1905 if (x->node != GREP_NODE_OR)
1906 return;
1907 x->u.binary.left->hit = 0;
1908 x = x->u.binary.right;
1912 static int chk_hit_marker(struct grep_expr *x)
1914 /* Top level nodes have hit markers. See if they all are hits */
1915 while (1) {
1916 if (x->node != GREP_NODE_OR)
1917 return x->hit;
1918 if (!x->u.binary.left->hit)
1919 return 0;
1920 x = x->u.binary.right;
1924 int grep_source(struct grep_opt *opt, struct grep_source *gs)
1927 * we do not have to do the two-pass grep when we do not check
1928 * buffer-wide "all-match".
1930 if (!opt->all_match)
1931 return grep_source_1(opt, gs, 0);
1933 /* Otherwise the toplevel "or" terms hit a bit differently.
1934 * We first clear hit markers from them.
1936 clr_hit_marker(opt->pattern_expression);
1937 grep_source_1(opt, gs, 1);
1939 if (!chk_hit_marker(opt->pattern_expression))
1940 return 0;
1942 return grep_source_1(opt, gs, 0);
1945 int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
1947 struct grep_source gs;
1948 int r;
1950 grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
1951 gs.buf = buf;
1952 gs.size = size;
1954 r = grep_source(opt, &gs);
1956 grep_source_clear(&gs);
1957 return r;
1960 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
1961 const char *name, const char *path,
1962 const void *identifier)
1964 gs->type = type;
1965 gs->name = xstrdup_or_null(name);
1966 gs->path = xstrdup_or_null(path);
1967 gs->buf = NULL;
1968 gs->size = 0;
1969 gs->driver = NULL;
1971 switch (type) {
1972 case GREP_SOURCE_FILE:
1973 gs->identifier = xstrdup(identifier);
1974 break;
1975 case GREP_SOURCE_OID:
1976 gs->identifier = oiddup(identifier);
1977 break;
1978 case GREP_SOURCE_BUF:
1979 gs->identifier = NULL;
1980 break;
1984 void grep_source_clear(struct grep_source *gs)
1986 FREE_AND_NULL(gs->name);
1987 FREE_AND_NULL(gs->path);
1988 FREE_AND_NULL(gs->identifier);
1989 grep_source_clear_data(gs);
1992 void grep_source_clear_data(struct grep_source *gs)
1994 switch (gs->type) {
1995 case GREP_SOURCE_FILE:
1996 case GREP_SOURCE_OID:
1997 FREE_AND_NULL(gs->buf);
1998 gs->size = 0;
1999 break;
2000 case GREP_SOURCE_BUF:
2001 /* leave user-provided buf intact */
2002 break;
2006 static int grep_source_load_oid(struct grep_source *gs)
2008 enum object_type type;
2010 grep_read_lock();
2011 gs->buf = read_object_file(gs->identifier, &type, &gs->size);
2012 grep_read_unlock();
2014 if (!gs->buf)
2015 return error(_("'%s': unable to read %s"),
2016 gs->name,
2017 oid_to_hex(gs->identifier));
2018 return 0;
2021 static int grep_source_load_file(struct grep_source *gs)
2023 const char *filename = gs->identifier;
2024 struct stat st;
2025 char *data;
2026 size_t size;
2027 int i;
2029 if (lstat(filename, &st) < 0) {
2030 err_ret:
2031 if (errno != ENOENT)
2032 error_errno(_("failed to stat '%s'"), filename);
2033 return -1;
2035 if (!S_ISREG(st.st_mode))
2036 return -1;
2037 size = xsize_t(st.st_size);
2038 i = open(filename, O_RDONLY);
2039 if (i < 0)
2040 goto err_ret;
2041 data = xmallocz(size);
2042 if (st.st_size != read_in_full(i, data, size)) {
2043 error_errno(_("'%s': short read"), filename);
2044 close(i);
2045 free(data);
2046 return -1;
2048 close(i);
2050 gs->buf = data;
2051 gs->size = size;
2052 return 0;
2055 static int grep_source_load(struct grep_source *gs)
2057 if (gs->buf)
2058 return 0;
2060 switch (gs->type) {
2061 case GREP_SOURCE_FILE:
2062 return grep_source_load_file(gs);
2063 case GREP_SOURCE_OID:
2064 return grep_source_load_oid(gs);
2065 case GREP_SOURCE_BUF:
2066 return gs->buf ? 0 : -1;
2068 BUG("invalid grep_source type to load");
2071 void grep_source_load_driver(struct grep_source *gs)
2073 if (gs->driver)
2074 return;
2076 grep_attr_lock();
2077 if (gs->path)
2078 gs->driver = userdiff_find_by_path(gs->path);
2079 if (!gs->driver)
2080 gs->driver = userdiff_find_by_name("default");
2081 grep_attr_unlock();
2084 static int grep_source_is_binary(struct grep_source *gs)
2086 grep_source_load_driver(gs);
2087 if (gs->driver->binary != -1)
2088 return gs->driver->binary;
2090 if (!grep_source_load(gs))
2091 return buffer_is_binary(gs->buf, gs->size);
2093 return 0;