4 #include "object-store.h"
6 #include "xdiff-interface.h"
13 static int grep_source_load(struct grep_source *gs);
14 static int grep_source_is_binary(struct grep_source *gs,
15 struct index_state *istate);
17 static struct grep_opt grep_defaults;
19 static const char *color_grep_slots[] = {
20 [GREP_COLOR_CONTEXT] = "context",
21 [GREP_COLOR_FILENAME] = "filename",
22 [GREP_COLOR_FUNCTION] = "function",
23 [GREP_COLOR_LINENO] = "lineNumber",
24 [GREP_COLOR_COLUMNNO] = "column",
25 [GREP_COLOR_MATCH_CONTEXT] = "matchContext",
26 [GREP_COLOR_MATCH_SELECTED] = "matchSelected",
27 [GREP_COLOR_SELECTED] = "selected",
28 [GREP_COLOR_SEP] = "separator",
31 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
33 fwrite(buf, size, 1, stdout);
36 static void color_set(char *dst, const char *color_bytes)
38 xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes);
42 * Initialize the grep_defaults template with hardcoded defaults.
43 * We could let the compiler do this, but without C99 initializers
44 * the code gets unwieldy and unreadable, so...
46 void init_grep_defaults(struct repository *repo)
48 struct grep_opt *opt = &grep_defaults;
55 memset(opt, 0, sizeof(*opt));
60 opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
61 color_set(opt->colors[GREP_COLOR_CONTEXT], "");
62 color_set(opt->colors[GREP_COLOR_FILENAME], "");
63 color_set(opt->colors[GREP_COLOR_FUNCTION], "");
64 color_set(opt->colors[GREP_COLOR_LINENO], "");
65 color_set(opt->colors[GREP_COLOR_COLUMNNO], "");
66 color_set(opt->colors[GREP_COLOR_MATCH_CONTEXT], GIT_COLOR_BOLD_RED);
67 color_set(opt->colors[GREP_COLOR_MATCH_SELECTED], GIT_COLOR_BOLD_RED);
68 color_set(opt->colors[GREP_COLOR_SELECTED], "");
69 color_set(opt->colors[GREP_COLOR_SEP], GIT_COLOR_CYAN);
70 opt->only_matching = 0;
72 opt->output = std_output;
75 static int parse_pattern_type_arg(const char *opt, const char *arg)
77 if (!strcmp(arg, "default"))
78 return GREP_PATTERN_TYPE_UNSPECIFIED;
79 else if (!strcmp(arg, "basic"))
80 return GREP_PATTERN_TYPE_BRE;
81 else if (!strcmp(arg, "extended"))
82 return GREP_PATTERN_TYPE_ERE;
83 else if (!strcmp(arg, "fixed"))
84 return GREP_PATTERN_TYPE_FIXED;
85 else if (!strcmp(arg, "perl"))
86 return GREP_PATTERN_TYPE_PCRE;
87 die("bad %s argument: %s", opt, arg);
90 define_list_config_array_extra(color_grep_slots, {"match"});
93 * Read the configuration file once and store it in
94 * the grep_defaults template.
96 int grep_config(const char *var, const char *value, void *cb)
98 struct grep_opt *opt = &grep_defaults;
101 if (userdiff_config(var, value) < 0)
104 if (!strcmp(var, "grep.extendedregexp")) {
105 opt->extended_regexp_option = git_config_bool(var, value);
109 if (!strcmp(var, "grep.patterntype")) {
110 opt->pattern_type_option = parse_pattern_type_arg(var, value);
114 if (!strcmp(var, "grep.linenumber")) {
115 opt->linenum = git_config_bool(var, value);
118 if (!strcmp(var, "grep.column")) {
119 opt->columnnum = git_config_bool(var, value);
123 if (!strcmp(var, "grep.fullname")) {
124 opt->relative = !git_config_bool(var, value);
128 if (!strcmp(var, "color.grep"))
129 opt->color = git_config_colorbool(var, value);
130 if (!strcmp(var, "color.grep.match")) {
131 if (grep_config("color.grep.matchcontext", value, cb) < 0)
133 if (grep_config("color.grep.matchselected", value, cb) < 0)
135 } else if (skip_prefix(var, "color.grep.", &slot)) {
136 int i = LOOKUP_CONFIG(color_grep_slots, slot);
141 color = opt->colors[i];
143 return config_error_nonbool(var);
144 return color_parse(value, color);
150 * Initialize one instance of grep_opt and copy the
151 * default values from the template we read the configuration
152 * information in an earlier call to git_config(grep_config).
154 void grep_init(struct grep_opt *opt, struct repository *repo, const char *prefix)
156 struct grep_opt *def = &grep_defaults;
159 memset(opt, 0, sizeof(*opt));
161 opt->prefix = prefix;
162 opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
163 opt->pattern_tail = &opt->pattern_list;
164 opt->header_tail = &opt->header_list;
166 opt->only_matching = def->only_matching;
167 opt->color = def->color;
168 opt->extended_regexp_option = def->extended_regexp_option;
169 opt->pattern_type_option = def->pattern_type_option;
170 opt->linenum = def->linenum;
171 opt->columnnum = def->columnnum;
172 opt->max_depth = def->max_depth;
173 opt->pathname = def->pathname;
174 opt->relative = def->relative;
175 opt->output = def->output;
177 for (i = 0; i < NR_GREP_COLORS; i++)
178 color_set(opt->colors[i], def->colors[i]);
181 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
184 * When committing to the pattern type by setting the relevant
185 * fields in grep_opt it's generally not necessary to zero out
186 * the fields we're not choosing, since they won't have been
187 * set by anything. The extended_regexp_option field is the
188 * only exception to this.
190 * This is because in the process of parsing grep.patternType
191 * & grep.extendedRegexp we set opt->pattern_type_option and
192 * opt->extended_regexp_option, respectively. We then
193 * internally use opt->extended_regexp_option to see if we're
194 * compiling an ERE. It must be unset if that's not actually
197 if (pattern_type != GREP_PATTERN_TYPE_ERE &&
198 opt->extended_regexp_option)
199 opt->extended_regexp_option = 0;
201 switch (pattern_type) {
202 case GREP_PATTERN_TYPE_UNSPECIFIED:
205 case GREP_PATTERN_TYPE_BRE:
208 case GREP_PATTERN_TYPE_ERE:
209 opt->extended_regexp_option = 1;
212 case GREP_PATTERN_TYPE_FIXED:
216 case GREP_PATTERN_TYPE_PCRE:
221 * It's important that pcre1 always be assigned to
222 * even when there's no USE_LIBPCRE* defined. We still
223 * call the PCRE stub function, it just dies with
224 * "cannot use Perl-compatible regexes[...]".
232 void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
234 if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
235 grep_set_pattern_type_option(pattern_type, opt);
236 else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
237 grep_set_pattern_type_option(opt->pattern_type_option, opt);
238 else if (opt->extended_regexp_option)
240 * This branch *must* happen after setting from the
241 * opt->pattern_type_option above, we don't want
242 * grep.extendedRegexp to override grep.patternType!
244 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
247 static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
248 const char *origin, int no,
249 enum grep_pat_token t,
250 enum grep_header_field field)
252 struct grep_pat *p = xcalloc(1, sizeof(*p));
253 p->pattern = xmemdupz(pat, patlen);
254 p->patternlen = patlen;
262 static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
269 case GREP_PATTERN: /* atom */
270 case GREP_PATTERN_HEAD:
271 case GREP_PATTERN_BODY:
273 struct grep_pat *new_pat;
275 char *cp = p->pattern + p->patternlen, *nl = NULL;
276 while (++len <= p->patternlen) {
277 if (*(--cp) == '\n') {
284 new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
285 p->no, p->token, p->field);
286 new_pat->next = p->next;
288 *tail = &new_pat->next;
291 p->patternlen -= len;
299 void append_header_grep_pattern(struct grep_opt *opt,
300 enum grep_header_field field, const char *pat)
302 struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
303 GREP_PATTERN_HEAD, field);
304 if (field == GREP_HEADER_REFLOG)
305 opt->use_reflog_filter = 1;
306 do_append_grep_pat(&opt->header_tail, p);
309 void append_grep_pattern(struct grep_opt *opt, const char *pat,
310 const char *origin, int no, enum grep_pat_token t)
312 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
315 void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
316 const char *origin, int no, enum grep_pat_token t)
318 struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
319 do_append_grep_pat(&opt->pattern_tail, p);
322 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
324 struct grep_pat *pat;
325 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
328 ret->pattern_list = NULL;
329 ret->pattern_tail = &ret->pattern_list;
331 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
333 if(pat->token == GREP_PATTERN_HEAD)
334 append_header_grep_pattern(ret, pat->field,
337 append_grep_pat(ret, pat->pattern, pat->patternlen,
338 pat->origin, pat->no, pat->token);
344 static NORETURN void compile_regexp_failed(const struct grep_pat *p,
350 xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
352 xsnprintf(where, sizeof(where), "%s, ", p->origin);
356 die("%s'%s': %s", where, p->pattern, error);
359 static int is_fixed(const char *s, size_t len)
363 for (i = 0; i < len; i++) {
364 if (is_regex_special(s[i]))
372 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
376 int options = PCRE_MULTILINE;
378 if (opt->ignore_case) {
379 if (!opt->ignore_locale && has_non_ascii(p->pattern))
380 p->pcre1_tables = pcre_maketables();
381 options |= PCRE_CASELESS;
383 if (!opt->ignore_locale && is_utf8_locale() && has_non_ascii(p->pattern))
384 options |= PCRE_UTF8;
386 p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
388 if (!p->pcre1_regexp)
389 compile_regexp_failed(p, error);
391 p->pcre1_extra_info = pcre_study(p->pcre1_regexp, GIT_PCRE_STUDY_JIT_COMPILE, &error);
392 if (!p->pcre1_extra_info && error)
395 #ifdef GIT_PCRE1_USE_JIT
396 pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on);
397 if (p->pcre1_jit_on) {
398 p->pcre1_jit_stack = pcre_jit_stack_alloc(1, 1024 * 1024);
399 if (!p->pcre1_jit_stack)
400 die("Couldn't allocate PCRE JIT stack");
401 pcre_assign_jit_stack(p->pcre1_extra_info, NULL, p->pcre1_jit_stack);
406 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
407 regmatch_t *match, int eflags)
409 int ovector[30], ret, flags = 0;
411 if (eflags & REG_NOTBOL)
412 flags |= PCRE_NOTBOL;
414 #ifdef GIT_PCRE1_USE_JIT
415 if (p->pcre1_jit_on) {
416 ret = pcre_jit_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
417 eol - line, 0, flags, ovector,
418 ARRAY_SIZE(ovector), p->pcre1_jit_stack);
422 ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
423 eol - line, 0, flags, ovector,
424 ARRAY_SIZE(ovector));
427 if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
428 die("pcre_exec failed with error code %d", ret);
431 match->rm_so = ovector[0];
432 match->rm_eo = ovector[1];
438 static void free_pcre1_regexp(struct grep_pat *p)
440 pcre_free(p->pcre1_regexp);
441 #ifdef GIT_PCRE1_USE_JIT
442 if (p->pcre1_jit_on) {
443 pcre_free_study(p->pcre1_extra_info);
444 pcre_jit_stack_free(p->pcre1_jit_stack);
448 pcre_free(p->pcre1_extra_info);
450 pcre_free((void *)p->pcre1_tables);
452 #else /* !USE_LIBPCRE1 */
453 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
455 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
458 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
459 regmatch_t *match, int eflags)
464 static void free_pcre1_regexp(struct grep_pat *p)
467 #endif /* !USE_LIBPCRE1 */
470 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
473 PCRE2_UCHAR errbuf[256];
474 PCRE2_SIZE erroffset;
475 int options = PCRE2_MULTILINE;
476 const uint8_t *character_tables = NULL;
483 p->pcre2_compile_context = NULL;
485 if (opt->ignore_case) {
486 if (!opt->ignore_locale && has_non_ascii(p->pattern)) {
487 character_tables = pcre2_maketables(NULL);
488 p->pcre2_compile_context = pcre2_compile_context_create(NULL);
489 pcre2_set_character_tables(p->pcre2_compile_context, character_tables);
491 options |= PCRE2_CASELESS;
493 if (!opt->ignore_locale && is_utf8_locale() && has_non_ascii(p->pattern))
494 options |= PCRE2_UTF;
496 p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern,
497 p->patternlen, options, &error, &erroffset,
498 p->pcre2_compile_context);
500 if (p->pcre2_pattern) {
501 p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL);
502 if (!p->pcre2_match_data)
503 die("Couldn't allocate PCRE2 match data");
505 pcre2_get_error_message(error, errbuf, sizeof(errbuf));
506 compile_regexp_failed(p, (const char *)&errbuf);
509 pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on);
510 if (p->pcre2_jit_on) {
511 jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE);
513 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret);
516 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
517 * tells us whether the library itself supports JIT,
518 * but to see whether we're going to be actually using
519 * JIT we need to extract PCRE2_INFO_JITSIZE from the
520 * pattern *after* we do pcre2_jit_compile() above.
522 * This is because if the pattern contains the
523 * (*NO_JIT) verb (see pcre2syntax(3))
524 * pcre2_jit_compile() will exit early with 0. If we
525 * then proceed to call pcre2_jit_match() further down
526 * the line instead of pcre2_match() we'll either
527 * segfault (pre PCRE 10.31) or run into a fatal error
530 patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg);
532 BUG("pcre2_pattern_info() failed: %d", patinforet);
533 if (jitsizearg == 0) {
540 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
541 regmatch_t *match, int eflags)
545 PCRE2_UCHAR errbuf[256];
547 if (eflags & REG_NOTBOL)
548 flags |= PCRE2_NOTBOL;
551 ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line,
552 eol - line, 0, flags, p->pcre2_match_data,
555 ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line,
556 eol - line, 0, flags, p->pcre2_match_data,
559 if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) {
560 pcre2_get_error_message(ret, errbuf, sizeof(errbuf));
561 die("%s failed with error code %d: %s",
562 (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret,
566 ovector = pcre2_get_ovector_pointer(p->pcre2_match_data);
568 match->rm_so = (int)ovector[0];
569 match->rm_eo = (int)ovector[1];
575 static void free_pcre2_pattern(struct grep_pat *p)
577 pcre2_compile_context_free(p->pcre2_compile_context);
578 pcre2_code_free(p->pcre2_pattern);
579 pcre2_match_data_free(p->pcre2_match_data);
581 #else /* !USE_LIBPCRE2 */
582 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
585 * Unreachable until USE_LIBPCRE2 becomes synonymous with
586 * USE_LIBPCRE. See the sibling comment in
587 * grep_set_pattern_type_option().
589 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
592 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
593 regmatch_t *match, int eflags)
598 static void free_pcre2_pattern(struct grep_pat *p)
602 static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
604 struct strbuf sb = STRBUF_INIT;
608 basic_regex_quote_buf(&sb, p->pattern);
609 if (opt->ignore_case)
610 regflags |= REG_ICASE;
611 err = regcomp(&p->regexp, sb.buf, regflags);
613 fprintf(stderr, "fixed %s\n", sb.buf);
617 regerror(err, &p->regexp, errbuf, sizeof(errbuf));
618 compile_regexp_failed(p, errbuf);
621 #endif /* !USE_LIBPCRE2 */
623 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
626 int regflags = REG_NEWLINE;
629 p->word_regexp = opt->word_regexp;
630 p->ignore_case = opt->ignore_case;
631 p->fixed = opt->fixed;
633 if (memchr(p->pattern, 0, p->patternlen) && !opt->pcre2)
634 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
636 pat_is_fixed = is_fixed(p->pattern, p->patternlen);
637 if (opt->fixed || pat_is_fixed) {
641 compile_pcre2_pattern(p, opt);
644 * E.g. t7811-grep-open.sh relies on the
645 * pattern being restored.
647 char *old_pattern = p->pattern;
648 size_t old_patternlen = p->patternlen;
649 struct strbuf sb = STRBUF_INIT;
652 * There is the PCRE2_LITERAL flag, but it's
653 * only in PCRE v2 10.30 and later. Needing to
654 * ifdef our way around that and dealing with
655 * it + PCRE2_MULTILINE being an error is more
656 * complex than just quoting this ourselves.
658 strbuf_add(&sb, "\\Q", 2);
659 strbuf_add(&sb, p->pattern, p->patternlen);
660 strbuf_add(&sb, "\\E", 2);
663 p->patternlen = sb.len;
664 compile_pcre2_pattern(p, opt);
665 p->pattern = old_pattern;
666 p->patternlen = old_patternlen;
669 #else /* !USE_LIBPCRE2 */
670 compile_fixed_regexp(p, opt);
671 #endif /* !USE_LIBPCRE2 */
676 compile_pcre2_pattern(p, opt);
681 compile_pcre1_regexp(p, opt);
686 regflags |= REG_ICASE;
687 if (opt->extended_regexp_option)
688 regflags |= REG_EXTENDED;
689 err = regcomp(&p->regexp, p->pattern, regflags);
692 regerror(err, &p->regexp, errbuf, 1024);
693 compile_regexp_failed(p, errbuf);
697 static struct grep_expr *compile_pattern_or(struct grep_pat **);
698 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
707 case GREP_PATTERN: /* atom */
708 case GREP_PATTERN_HEAD:
709 case GREP_PATTERN_BODY:
710 x = xcalloc(1, sizeof (struct grep_expr));
711 x->node = GREP_NODE_ATOM;
715 case GREP_OPEN_PAREN:
717 x = compile_pattern_or(list);
718 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
719 die("unmatched parenthesis");
720 *list = (*list)->next;
727 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
738 die("--not not followed by pattern expression");
740 x = xcalloc(1, sizeof (struct grep_expr));
741 x->node = GREP_NODE_NOT;
742 x->u.unary = compile_pattern_not(list);
744 die("--not followed by non pattern expression");
747 return compile_pattern_atom(list);
751 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
754 struct grep_expr *x, *y, *z;
756 x = compile_pattern_not(list);
758 if (p && p->token == GREP_AND) {
760 die("--and not followed by pattern expression");
762 y = compile_pattern_and(list);
764 die("--and not followed by pattern expression");
765 z = xcalloc(1, sizeof (struct grep_expr));
766 z->node = GREP_NODE_AND;
767 z->u.binary.left = x;
768 z->u.binary.right = y;
774 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
777 struct grep_expr *x, *y, *z;
779 x = compile_pattern_and(list);
781 if (x && p && p->token != GREP_CLOSE_PAREN) {
782 y = compile_pattern_or(list);
784 die("not a pattern expression %s", p->pattern);
785 z = xcalloc(1, sizeof (struct grep_expr));
786 z->node = GREP_NODE_OR;
787 z->u.binary.left = x;
788 z->u.binary.right = y;
794 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
796 return compile_pattern_or(list);
799 static void indent(int in)
805 static void dump_grep_pat(struct grep_pat *p)
808 case GREP_AND: fprintf(stderr, "*and*"); break;
809 case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
810 case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
811 case GREP_NOT: fprintf(stderr, "*not*"); break;
812 case GREP_OR: fprintf(stderr, "*or*"); break;
814 case GREP_PATTERN: fprintf(stderr, "pattern"); break;
815 case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
816 case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
821 case GREP_PATTERN_HEAD:
822 fprintf(stderr, "<head %d>", p->field); break;
823 case GREP_PATTERN_BODY:
824 fprintf(stderr, "<body>"); break;
828 case GREP_PATTERN_HEAD:
829 case GREP_PATTERN_BODY:
831 fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
837 static void dump_grep_expression_1(struct grep_expr *x, int in)
842 fprintf(stderr, "true\n");
845 dump_grep_pat(x->u.atom);
848 fprintf(stderr, "(not\n");
849 dump_grep_expression_1(x->u.unary, in+1);
851 fprintf(stderr, ")\n");
854 fprintf(stderr, "(and\n");
855 dump_grep_expression_1(x->u.binary.left, in+1);
856 dump_grep_expression_1(x->u.binary.right, in+1);
858 fprintf(stderr, ")\n");
861 fprintf(stderr, "(or\n");
862 dump_grep_expression_1(x->u.binary.left, in+1);
863 dump_grep_expression_1(x->u.binary.right, in+1);
865 fprintf(stderr, ")\n");
870 static void dump_grep_expression(struct grep_opt *opt)
872 struct grep_expr *x = opt->pattern_expression;
875 fprintf(stderr, "[all-match]\n");
876 dump_grep_expression_1(x, 0);
880 static struct grep_expr *grep_true_expr(void)
882 struct grep_expr *z = xcalloc(1, sizeof(*z));
883 z->node = GREP_NODE_TRUE;
887 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
889 struct grep_expr *z = xcalloc(1, sizeof(*z));
890 z->node = GREP_NODE_OR;
891 z->u.binary.left = left;
892 z->u.binary.right = right;
896 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
899 struct grep_expr *header_expr;
900 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
901 enum grep_header_field fld;
903 if (!opt->header_list)
906 for (p = opt->header_list; p; p = p->next) {
907 if (p->token != GREP_PATTERN_HEAD)
908 BUG("a non-header pattern in grep header list.");
909 if (p->field < GREP_HEADER_FIELD_MIN ||
910 GREP_HEADER_FIELD_MAX <= p->field)
911 BUG("unknown header field %d", p->field);
912 compile_regexp(p, opt);
915 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
916 header_group[fld] = NULL;
918 for (p = opt->header_list; p; p = p->next) {
920 struct grep_pat *pp = p;
922 h = compile_pattern_atom(&pp);
923 if (!h || pp != p->next)
924 BUG("malformed header expr");
925 if (!header_group[p->field]) {
926 header_group[p->field] = h;
929 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
934 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
935 if (!header_group[fld])
938 header_expr = grep_true_expr();
939 header_expr = grep_or_expr(header_group[fld], header_expr);
944 static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
946 struct grep_expr *z = x;
949 assert(x->node == GREP_NODE_OR);
950 if (x->u.binary.right &&
951 x->u.binary.right->node == GREP_NODE_TRUE) {
952 x->u.binary.right = y;
955 x = x->u.binary.right;
960 static void compile_grep_patterns_real(struct grep_opt *opt)
963 struct grep_expr *header_expr = prep_header_patterns(opt);
965 for (p = opt->pattern_list; p; p = p->next) {
967 case GREP_PATTERN: /* atom */
968 case GREP_PATTERN_HEAD:
969 case GREP_PATTERN_BODY:
970 compile_regexp(p, opt);
978 if (opt->all_match || header_expr)
980 else if (!opt->extended && !opt->debug)
983 p = opt->pattern_list;
985 opt->pattern_expression = compile_pattern_expr(&p);
987 die("incomplete pattern expression: %s", p->pattern);
992 if (!opt->pattern_expression)
993 opt->pattern_expression = header_expr;
994 else if (opt->all_match)
995 opt->pattern_expression = grep_splice_or(header_expr,
996 opt->pattern_expression);
998 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
1003 void compile_grep_patterns(struct grep_opt *opt)
1005 compile_grep_patterns_real(opt);
1007 dump_grep_expression(opt);
1010 static void free_pattern_expr(struct grep_expr *x)
1013 case GREP_NODE_TRUE:
1014 case GREP_NODE_ATOM:
1017 free_pattern_expr(x->u.unary);
1021 free_pattern_expr(x->u.binary.left);
1022 free_pattern_expr(x->u.binary.right);
1028 void free_grep_patterns(struct grep_opt *opt)
1030 struct grep_pat *p, *n;
1032 for (p = opt->pattern_list; p; p = n) {
1035 case GREP_PATTERN: /* atom */
1036 case GREP_PATTERN_HEAD:
1037 case GREP_PATTERN_BODY:
1038 if (p->pcre1_regexp)
1039 free_pcre1_regexp(p);
1040 else if (p->pcre2_pattern)
1041 free_pcre2_pattern(p);
1043 regfree(&p->regexp);
1054 free_pattern_expr(opt->pattern_expression);
1057 static char *end_of_line(char *cp, unsigned long *left)
1059 unsigned long l = *left;
1060 while (l && *cp != '\n') {
1068 static int word_char(char ch)
1070 return isalnum(ch) || ch == '_';
1073 static void output_color(struct grep_opt *opt, const void *data, size_t size,
1076 if (want_color(opt->color) && color && color[0]) {
1077 opt->output(opt, color, strlen(color));
1078 opt->output(opt, data, size);
1079 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
1081 opt->output(opt, data, size);
1084 static void output_sep(struct grep_opt *opt, char sign)
1086 if (opt->null_following_name)
1087 opt->output(opt, "\0", 1);
1089 output_color(opt, &sign, 1, opt->colors[GREP_COLOR_SEP]);
1092 static void show_name(struct grep_opt *opt, const char *name)
1094 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1095 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
1098 static int patmatch(struct grep_pat *p, char *line, char *eol,
1099 regmatch_t *match, int eflags)
1103 if (p->pcre1_regexp)
1104 hit = !pcre1match(p, line, eol, match, eflags);
1105 else if (p->pcre2_pattern)
1106 hit = !pcre2match(p, line, eol, match, eflags);
1108 hit = !regexec_buf(&p->regexp, line, eol - line, 1, match,
1114 static int strip_timestamp(char *bol, char **eol_p)
1119 while (bol < --eol) {
1133 } header_field[] = {
1135 { "committer ", 10 },
1139 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
1140 enum grep_context ctx,
1141 regmatch_t *pmatch, int eflags)
1145 const char *start = bol;
1147 if ((p->token != GREP_PATTERN) &&
1148 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
1151 if (p->token == GREP_PATTERN_HEAD) {
1154 assert(p->field < ARRAY_SIZE(header_field));
1155 field = header_field[p->field].field;
1156 len = header_field[p->field].len;
1157 if (strncmp(bol, field, len))
1161 case GREP_HEADER_AUTHOR:
1162 case GREP_HEADER_COMMITTER:
1163 saved_ch = strip_timestamp(bol, &eol);
1171 hit = patmatch(p, bol, eol, pmatch, eflags);
1173 if (hit && p->word_regexp) {
1174 if ((pmatch[0].rm_so < 0) ||
1175 (eol - bol) < pmatch[0].rm_so ||
1176 (pmatch[0].rm_eo < 0) ||
1177 (eol - bol) < pmatch[0].rm_eo)
1178 die("regexp returned nonsense");
1180 /* Match beginning must be either beginning of the
1181 * line, or at word boundary (i.e. the last char must
1182 * not be a word char). Similarly, match end must be
1183 * either end of the line, or at word boundary
1184 * (i.e. the next char must not be a word char).
1186 if ( ((pmatch[0].rm_so == 0) ||
1187 !word_char(bol[pmatch[0].rm_so-1])) &&
1188 ((pmatch[0].rm_eo == (eol-bol)) ||
1189 !word_char(bol[pmatch[0].rm_eo])) )
1194 /* Words consist of at least one character. */
1195 if (pmatch->rm_so == pmatch->rm_eo)
1198 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
1199 /* There could be more than one match on the
1200 * line, and the first match might not be
1201 * strict word match. But later ones could be!
1202 * Forward to the next possible start, i.e. the
1203 * next position following a non-word char.
1205 bol = pmatch[0].rm_so + bol + 1;
1206 while (word_char(bol[-1]) && bol < eol)
1208 eflags |= REG_NOTBOL;
1213 if (p->token == GREP_PATTERN_HEAD && saved_ch)
1216 pmatch[0].rm_so += bol - start;
1217 pmatch[0].rm_eo += bol - start;
1222 static int match_expr_eval(struct grep_opt *opt, struct grep_expr *x, char *bol,
1223 char *eol, enum grep_context ctx, ssize_t *col,
1224 ssize_t *icol, int collect_hits)
1229 die("Not a valid grep expression");
1231 case GREP_NODE_TRUE:
1234 case GREP_NODE_ATOM:
1237 h = match_one_pattern(x->u.atom, bol, eol, ctx,
1239 if (h && (*col < 0 || tmp.rm_so < *col))
1245 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1247 h = !match_expr_eval(opt, x->u.unary, bol, eol, ctx, icol, col,
1251 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1253 if (h || opt->columnnum) {
1255 * Don't short-circuit AND when given --column, since a
1256 * NOT earlier in the tree may turn this into an OR. In
1257 * this case, see the below comment.
1259 h &= match_expr_eval(opt, x->u.binary.right, bol, eol,
1264 if (!(collect_hits || opt->columnnum)) {
1266 * Don't short-circuit OR when given --column (or
1267 * collecting hits) to ensure we don't skip a later
1268 * child that would produce an earlier match.
1270 return (match_expr_eval(opt, x->u.binary.left, bol, eol,
1271 ctx, col, icol, 0) ||
1272 match_expr_eval(opt, x->u.binary.right, bol,
1273 eol, ctx, col, icol, 0));
1275 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1278 x->u.binary.left->hit |= h;
1279 h |= match_expr_eval(opt, x->u.binary.right, bol, eol, ctx, col,
1280 icol, collect_hits);
1283 die("Unexpected node type (internal error) %d", x->node);
1290 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
1291 enum grep_context ctx, ssize_t *col,
1292 ssize_t *icol, int collect_hits)
1294 struct grep_expr *x = opt->pattern_expression;
1295 return match_expr_eval(opt, x, bol, eol, ctx, col, icol, collect_hits);
1298 static int match_line(struct grep_opt *opt, char *bol, char *eol,
1299 ssize_t *col, ssize_t *icol,
1300 enum grep_context ctx, int collect_hits)
1306 return match_expr(opt, bol, eol, ctx, col, icol,
1309 /* we do not call with collect_hits without being extended */
1310 for (p = opt->pattern_list; p; p = p->next) {
1312 if (match_one_pattern(p, bol, eol, ctx, &tmp, 0)) {
1314 if (!opt->columnnum) {
1316 * Without --column, any single match on a line
1317 * is enough to know that it needs to be
1318 * printed. With --column, scan _all_ patterns
1319 * to find the earliest.
1323 if (*col < 0 || tmp.rm_so < *col)
1330 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
1331 enum grep_context ctx,
1332 regmatch_t *pmatch, int eflags)
1336 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
1338 if (match.rm_so < 0 || match.rm_eo < 0)
1340 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
1341 if (match.rm_so > pmatch->rm_so)
1343 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
1346 pmatch->rm_so = match.rm_so;
1347 pmatch->rm_eo = match.rm_eo;
1351 static int next_match(struct grep_opt *opt, char *bol, char *eol,
1352 enum grep_context ctx, regmatch_t *pmatch, int eflags)
1357 pmatch->rm_so = pmatch->rm_eo = -1;
1359 for (p = opt->pattern_list; p; p = p->next) {
1361 case GREP_PATTERN: /* atom */
1362 case GREP_PATTERN_HEAD:
1363 case GREP_PATTERN_BODY:
1364 hit |= match_next_pattern(p, bol, eol, ctx,
1375 static void show_line_header(struct grep_opt *opt, const char *name,
1376 unsigned lno, ssize_t cno, char sign)
1378 if (opt->heading && opt->last_shown == 0) {
1379 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1380 opt->output(opt, "\n", 1);
1382 opt->last_shown = lno;
1384 if (!opt->heading && opt->pathname) {
1385 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1386 output_sep(opt, sign);
1390 xsnprintf(buf, sizeof(buf), "%d", lno);
1391 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_LINENO]);
1392 output_sep(opt, sign);
1395 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1396 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1397 * being called with a context line.
1399 if (opt->columnnum && cno) {
1401 xsnprintf(buf, sizeof(buf), "%"PRIuMAX, (uintmax_t)cno);
1402 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_COLUMNNO]);
1403 output_sep(opt, sign);
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 = NULL;
1412 const char *line_color = NULL;
1414 if (opt->file_break && opt->last_shown == 0) {
1415 if (opt->show_hunk_mark)
1416 opt->output(opt, "\n", 1);
1417 } else if (opt->pre_context || opt->post_context || opt->funcbody) {
1418 if (opt->last_shown == 0) {
1419 if (opt->show_hunk_mark) {
1420 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1421 opt->output(opt, "\n", 1);
1423 } else if (lno > opt->last_shown + 1) {
1424 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1425 opt->output(opt, "\n", 1);
1428 if (!opt->only_matching) {
1430 * In case the line we're being called with contains more than
1431 * one match, leave printing each header to the loop below.
1433 show_line_header(opt, name, lno, cno, sign);
1435 if (opt->color || opt->only_matching) {
1437 enum grep_context ctx = GREP_CONTEXT_BODY;
1443 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
1445 match_color = opt->colors[GREP_COLOR_MATCH_CONTEXT];
1447 line_color = opt->colors[GREP_COLOR_SELECTED];
1448 else if (sign == '-')
1449 line_color = opt->colors[GREP_COLOR_CONTEXT];
1450 else if (sign == '=')
1451 line_color = opt->colors[GREP_COLOR_FUNCTION];
1454 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
1455 if (match.rm_so == match.rm_eo)
1458 if (opt->only_matching)
1459 show_line_header(opt, name, lno, cno, sign);
1461 output_color(opt, bol, match.rm_so, line_color);
1462 output_color(opt, bol + match.rm_so,
1463 match.rm_eo - match.rm_so, match_color);
1464 if (opt->only_matching)
1465 opt->output(opt, "\n", 1);
1468 rest -= match.rm_eo;
1469 eflags = REG_NOTBOL;
1473 if (!opt->only_matching) {
1474 output_color(opt, bol, rest, line_color);
1475 opt->output(opt, "\n", 1);
1482 * This lock protects access to the gitattributes machinery, which is
1485 pthread_mutex_t grep_attr_mutex;
1487 static inline void grep_attr_lock(void)
1490 pthread_mutex_lock(&grep_attr_mutex);
1493 static inline void grep_attr_unlock(void)
1496 pthread_mutex_unlock(&grep_attr_mutex);
1500 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1502 pthread_mutex_t grep_read_mutex;
1504 static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
1506 xdemitconf_t *xecfg = opt->priv;
1507 if (xecfg && !xecfg->find_func) {
1508 grep_source_load_driver(gs, opt->repo->index);
1509 if (gs->driver->funcname.pattern) {
1510 const struct userdiff_funcname *pe = &gs->driver->funcname;
1511 xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
1513 xecfg = opt->priv = NULL;
1519 return xecfg->find_func(bol, eol - bol, buf, 1,
1520 xecfg->find_func_priv) >= 0;
1525 if (isalpha(*bol) || *bol == '_' || *bol == '$')
1530 static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
1531 char *bol, unsigned lno)
1533 while (bol > gs->buf) {
1536 while (bol > gs->buf && bol[-1] != '\n')
1540 if (lno <= opt->last_shown)
1543 if (match_funcname(opt, gs, bol, eol)) {
1544 show_line(opt, bol, eol, gs->name, lno, 0, '=');
1550 static int is_empty_line(const char *bol, const char *eol);
1552 static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
1553 char *bol, char *end, unsigned lno)
1555 unsigned cur = lno, from = 1, funcname_lno = 0, orig_from;
1556 int funcname_needed = !!opt->funcname, comment_needed = 0;
1558 if (opt->pre_context < lno)
1559 from = lno - opt->pre_context;
1560 if (from <= opt->last_shown)
1561 from = opt->last_shown + 1;
1563 if (opt->funcbody) {
1564 if (match_funcname(opt, gs, bol, end))
1567 funcname_needed = 1;
1568 from = opt->last_shown + 1;
1572 while (bol > gs->buf && cur > from) {
1573 char *next_bol = bol;
1576 while (bol > gs->buf && bol[-1] != '\n')
1579 if (comment_needed && (is_empty_line(bol, eol) ||
1580 match_funcname(opt, gs, bol, eol))) {
1589 if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
1591 funcname_needed = 0;
1599 /* We need to look even further back to find a function signature. */
1600 if (opt->funcname && funcname_needed)
1601 show_funcname_line(opt, gs, bol, cur);
1605 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
1607 while (*eol != '\n')
1609 show_line(opt, bol, eol, gs->name, cur, 0, sign);
1615 static int should_lookahead(struct grep_opt *opt)
1620 return 0; /* punt for too complex stuff */
1623 for (p = opt->pattern_list; p; p = p->next) {
1624 if (p->token != GREP_PATTERN)
1625 return 0; /* punt for "header only" and stuff */
1630 static int look_ahead(struct grep_opt *opt,
1631 unsigned long *left_p,
1635 unsigned lno = *lno_p;
1638 char *sp, *last_bol;
1639 regoff_t earliest = -1;
1641 for (p = opt->pattern_list; p; p = p->next) {
1645 hit = patmatch(p, bol, bol + *left_p, &m, 0);
1646 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
1648 if (earliest < 0 || m.rm_so < earliest)
1653 *bol_p = bol + *left_p;
1657 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
1658 ; /* find the beginning of the line */
1661 for (sp = bol; sp < last_bol; sp++) {
1665 *left_p -= last_bol - bol;
1671 static int fill_textconv_grep(struct repository *r,
1672 struct userdiff_driver *driver,
1673 struct grep_source *gs)
1675 struct diff_filespec *df;
1679 if (!driver || !driver->textconv)
1680 return grep_source_load(gs);
1683 * The textconv interface is intimately tied to diff_filespecs, so we
1684 * have to pretend to be one. If we could unify the grep_source
1685 * and diff_filespec structs, this mess could just go away.
1687 df = alloc_filespec(gs->path);
1689 case GREP_SOURCE_OID:
1690 fill_filespec(df, gs->identifier, 1, 0100644);
1692 case GREP_SOURCE_FILE:
1693 fill_filespec(df, &null_oid, 0, 0100644);
1696 BUG("attempt to textconv something without a path?");
1700 * fill_textconv is not remotely thread-safe; it may load objects
1701 * behind the scenes, and it modifies the global diff tempfile
1705 size = fill_textconv(r, driver, df, &buf);
1710 * The normal fill_textconv usage by the diff machinery would just keep
1711 * the textconv'd buf separate from the diff_filespec. But much of the
1712 * grep code passes around a grep_source and assumes that its "buf"
1713 * pointer is the beginning of the thing we are searching. So let's
1714 * install our textconv'd version into the grep_source, taking care not
1715 * to leak any existing buffer.
1717 grep_source_clear_data(gs);
1724 static int is_empty_line(const char *bol, const char *eol)
1726 while (bol < eol && isspace(*bol))
1731 static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
1734 char *peek_bol = NULL;
1737 unsigned last_hit = 0;
1738 int binary_match_only = 0;
1740 int try_lookahead = 0;
1741 int show_function = 0;
1742 struct userdiff_driver *textconv = NULL;
1743 enum grep_context ctx = GREP_CONTEXT_HEAD;
1746 if (!opt->status_only && gs->name == NULL)
1747 BUG("grep call which could print a name requires "
1748 "grep_source.name be non-NULL");
1751 opt->output = std_output;
1753 if (opt->pre_context || opt->post_context || opt->file_break ||
1755 /* Show hunk marks, except for the first file. */
1756 if (opt->last_shown)
1757 opt->show_hunk_mark = 1;
1759 * If we're using threads then we can't easily identify
1760 * the first file. Always put hunk marks in that case
1761 * and skip the very first one later in work_done().
1763 if (opt->output != std_output)
1764 opt->show_hunk_mark = 1;
1766 opt->last_shown = 0;
1768 if (opt->allow_textconv) {
1769 grep_source_load_driver(gs, opt->repo->index);
1771 * We might set up the shared textconv cache data here, which
1772 * is not thread-safe.
1775 textconv = userdiff_get_textconv(opt->repo, gs->driver);
1780 * We know the result of a textconv is text, so we only have to care
1781 * about binary handling if we are not using it.
1784 switch (opt->binary) {
1785 case GREP_BINARY_DEFAULT:
1786 if (grep_source_is_binary(gs, opt->repo->index))
1787 binary_match_only = 1;
1789 case GREP_BINARY_NOMATCH:
1790 if (grep_source_is_binary(gs, opt->repo->index))
1791 return 0; /* Assume unmatch */
1793 case GREP_BINARY_TEXT:
1796 BUG("unknown binary handling mode");
1800 memset(&xecfg, 0, sizeof(xecfg));
1803 try_lookahead = should_lookahead(opt);
1805 if (fill_textconv_grep(opt->repo, textconv, gs) < 0)
1814 ssize_t col = -1, icol = -1;
1817 * look_ahead() skips quickly to the line that possibly
1818 * has the next hit; don't call it if we need to do
1819 * something more than just skipping the current line
1820 * in response to an unmatch for the current line. E.g.
1821 * inside a post-context window, we will show the current
1822 * line as a context around the previous hit when it
1827 && (show_function ||
1828 lno <= last_hit + opt->post_context))
1829 && look_ahead(opt, &left, &lno, &bol))
1831 eol = end_of_line(bol, &left);
1835 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
1836 ctx = GREP_CONTEXT_BODY;
1838 hit = match_line(opt, bol, eol, &col, &icol, ctx, collect_hits);
1844 /* "grep -v -e foo -e bla" should list lines
1845 * that do not have either, so inversion should
1850 if (opt->unmatch_name_only) {
1857 if (opt->status_only)
1859 if (opt->name_only) {
1860 show_name(opt, gs->name);
1865 if (binary_match_only) {
1866 opt->output(opt, "Binary file ", 12);
1867 output_color(opt, gs->name, strlen(gs->name),
1868 opt->colors[GREP_COLOR_FILENAME]);
1869 opt->output(opt, " matches\n", 9);
1872 /* Hit at this line. If we haven't shown the
1873 * pre-context lines, we would need to show them.
1875 if (opt->pre_context || opt->funcbody)
1876 show_pre_context(opt, gs, bol, eol, lno);
1877 else if (opt->funcname)
1878 show_funcname_line(opt, gs, bol, lno);
1879 cno = opt->invert ? icol : col;
1882 * A negative cno indicates that there was no
1883 * match on the line. We are thus inverted and
1884 * being asked to show all lines that _don't_
1885 * match a given expression. Therefore, set cno
1886 * to 0 to suggest the whole line matches.
1890 show_line(opt, bol, eol, gs->name, lno, cno + 1, ':');
1896 if (show_function && (!peek_bol || peek_bol < bol)) {
1897 unsigned long peek_left = left;
1898 char *peek_eol = eol;
1901 * Trailing empty lines are not interesting.
1902 * Peek past them to see if they belong to the
1903 * body of the current function.
1906 while (is_empty_line(peek_bol, peek_eol)) {
1907 peek_bol = peek_eol + 1;
1908 peek_eol = end_of_line(peek_bol, &peek_left);
1911 if (match_funcname(opt, gs, peek_bol, peek_eol))
1914 if (show_function ||
1915 (last_hit && lno <= last_hit + opt->post_context)) {
1916 /* If the last hit is within the post context,
1917 * we need to show this line.
1919 show_line(opt, bol, eol, gs->name, lno, col + 1, '-');
1933 if (opt->status_only)
1934 return opt->unmatch_name_only;
1935 if (opt->unmatch_name_only) {
1936 /* We did not see any hit, so we want to show this */
1937 show_name(opt, gs->name);
1941 xdiff_clear_find_func(&xecfg);
1945 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1946 * which feels mostly useless but sometimes useful. Maybe
1947 * make it another option? For now suppress them.
1949 if (opt->count && count) {
1951 if (opt->pathname) {
1952 output_color(opt, gs->name, strlen(gs->name),
1953 opt->colors[GREP_COLOR_FILENAME]);
1954 output_sep(opt, ':');
1956 xsnprintf(buf, sizeof(buf), "%u\n", count);
1957 opt->output(opt, buf, strlen(buf));
1963 static void clr_hit_marker(struct grep_expr *x)
1965 /* All-hit markers are meaningful only at the very top level
1970 if (x->node != GREP_NODE_OR)
1972 x->u.binary.left->hit = 0;
1973 x = x->u.binary.right;
1977 static int chk_hit_marker(struct grep_expr *x)
1979 /* Top level nodes have hit markers. See if they all are hits */
1981 if (x->node != GREP_NODE_OR)
1983 if (!x->u.binary.left->hit)
1985 x = x->u.binary.right;
1989 int grep_source(struct grep_opt *opt, struct grep_source *gs)
1992 * we do not have to do the two-pass grep when we do not check
1993 * buffer-wide "all-match".
1995 if (!opt->all_match)
1996 return grep_source_1(opt, gs, 0);
1998 /* Otherwise the toplevel "or" terms hit a bit differently.
1999 * We first clear hit markers from them.
2001 clr_hit_marker(opt->pattern_expression);
2002 grep_source_1(opt, gs, 1);
2004 if (!chk_hit_marker(opt->pattern_expression))
2007 return grep_source_1(opt, gs, 0);
2010 int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
2012 struct grep_source gs;
2015 grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
2019 r = grep_source(opt, &gs);
2021 grep_source_clear(&gs);
2025 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
2026 const char *name, const char *path,
2027 const void *identifier)
2030 gs->name = xstrdup_or_null(name);
2031 gs->path = xstrdup_or_null(path);
2037 case GREP_SOURCE_FILE:
2038 gs->identifier = xstrdup(identifier);
2040 case GREP_SOURCE_OID:
2041 gs->identifier = oiddup(identifier);
2043 case GREP_SOURCE_BUF:
2044 gs->identifier = NULL;
2049 void grep_source_clear(struct grep_source *gs)
2051 FREE_AND_NULL(gs->name);
2052 FREE_AND_NULL(gs->path);
2053 FREE_AND_NULL(gs->identifier);
2054 grep_source_clear_data(gs);
2057 void grep_source_clear_data(struct grep_source *gs)
2060 case GREP_SOURCE_FILE:
2061 case GREP_SOURCE_OID:
2062 FREE_AND_NULL(gs->buf);
2065 case GREP_SOURCE_BUF:
2066 /* leave user-provided buf intact */
2071 static int grep_source_load_oid(struct grep_source *gs)
2073 enum object_type type;
2076 gs->buf = read_object_file(gs->identifier, &type, &gs->size);
2080 return error(_("'%s': unable to read %s"),
2082 oid_to_hex(gs->identifier));
2086 static int grep_source_load_file(struct grep_source *gs)
2088 const char *filename = gs->identifier;
2094 if (lstat(filename, &st) < 0) {
2096 if (errno != ENOENT)
2097 error_errno(_("failed to stat '%s'"), filename);
2100 if (!S_ISREG(st.st_mode))
2102 size = xsize_t(st.st_size);
2103 i = open(filename, O_RDONLY);
2106 data = xmallocz(size);
2107 if (st.st_size != read_in_full(i, data, size)) {
2108 error_errno(_("'%s': short read"), filename);
2120 static int grep_source_load(struct grep_source *gs)
2126 case GREP_SOURCE_FILE:
2127 return grep_source_load_file(gs);
2128 case GREP_SOURCE_OID:
2129 return grep_source_load_oid(gs);
2130 case GREP_SOURCE_BUF:
2131 return gs->buf ? 0 : -1;
2133 BUG("invalid grep_source type to load");
2136 void grep_source_load_driver(struct grep_source *gs,
2137 struct index_state *istate)
2144 gs->driver = userdiff_find_by_path(istate, gs->path);
2146 gs->driver = userdiff_find_by_name("default");
2150 static int grep_source_is_binary(struct grep_source *gs,
2151 struct index_state *istate)
2153 grep_source_load_driver(gs, istate);
2154 if (gs->driver->binary != -1)
2155 return gs->driver->binary;
2157 if (!grep_source_load(gs))
2158 return buffer_is_binary(gs->buf, gs->size);