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 void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
19 fwrite(buf
, size
, 1, stdout
);
22 static const char *color_grep_slots
[] = {
23 [GREP_COLOR_CONTEXT
] = "context",
24 [GREP_COLOR_FILENAME
] = "filename",
25 [GREP_COLOR_FUNCTION
] = "function",
26 [GREP_COLOR_LINENO
] = "lineNumber",
27 [GREP_COLOR_COLUMNNO
] = "column",
28 [GREP_COLOR_MATCH_CONTEXT
] = "matchContext",
29 [GREP_COLOR_MATCH_SELECTED
] = "matchSelected",
30 [GREP_COLOR_SELECTED
] = "selected",
31 [GREP_COLOR_SEP
] = "separator",
34 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
36 if (!strcmp(arg
, "default"))
37 return GREP_PATTERN_TYPE_UNSPECIFIED
;
38 else if (!strcmp(arg
, "basic"))
39 return GREP_PATTERN_TYPE_BRE
;
40 else if (!strcmp(arg
, "extended"))
41 return GREP_PATTERN_TYPE_ERE
;
42 else if (!strcmp(arg
, "fixed"))
43 return GREP_PATTERN_TYPE_FIXED
;
44 else if (!strcmp(arg
, "perl"))
45 return GREP_PATTERN_TYPE_PCRE
;
46 die("bad %s argument: %s", opt
, arg
);
49 define_list_config_array_extra(color_grep_slots
, {"match"});
52 * Read the configuration file once and store it in
53 * the grep_defaults template.
55 int grep_config(const char *var
, const char *value
, void *cb
)
57 struct grep_opt
*opt
= cb
;
60 if (userdiff_config(var
, value
) < 0)
63 if (!strcmp(var
, "grep.extendedregexp")) {
64 opt
->extended_regexp_option
= git_config_bool(var
, value
);
68 if (!strcmp(var
, "grep.patterntype")) {
69 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
73 if (!strcmp(var
, "grep.linenumber")) {
74 opt
->linenum
= git_config_bool(var
, value
);
77 if (!strcmp(var
, "grep.column")) {
78 opt
->columnnum
= git_config_bool(var
, value
);
82 if (!strcmp(var
, "grep.fullname")) {
83 opt
->relative
= !git_config_bool(var
, value
);
87 if (!strcmp(var
, "color.grep"))
88 opt
->color
= git_config_colorbool(var
, value
);
89 if (!strcmp(var
, "color.grep.match")) {
90 if (grep_config("color.grep.matchcontext", value
, cb
) < 0)
92 if (grep_config("color.grep.matchselected", value
, cb
) < 0)
94 } else if (skip_prefix(var
, "color.grep.", &slot
)) {
95 int i
= LOOKUP_CONFIG(color_grep_slots
, slot
);
100 color
= opt
->colors
[i
];
102 return config_error_nonbool(var
);
103 return color_parse(value
, color
);
108 void grep_init(struct grep_opt
*opt
, struct repository
*repo
)
110 struct grep_opt blank
= GREP_OPT_INIT
;
111 memcpy(opt
, &blank
, sizeof(*opt
));
114 opt
->pattern_tail
= &opt
->pattern_list
;
115 opt
->header_tail
= &opt
->header_list
;
118 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
121 * When committing to the pattern type by setting the relevant
122 * fields in grep_opt it's generally not necessary to zero out
123 * the fields we're not choosing, since they won't have been
124 * set by anything. The extended_regexp_option field is the
125 * only exception to this.
127 * This is because in the process of parsing grep.patternType
128 * & grep.extendedRegexp we set opt->pattern_type_option and
129 * opt->extended_regexp_option, respectively. We then
130 * internally use opt->extended_regexp_option to see if we're
131 * compiling an ERE. It must be unset if that's not actually
134 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
135 opt
->extended_regexp_option
)
136 opt
->extended_regexp_option
= 0;
138 switch (pattern_type
) {
139 case GREP_PATTERN_TYPE_UNSPECIFIED
:
142 case GREP_PATTERN_TYPE_BRE
:
145 case GREP_PATTERN_TYPE_ERE
:
146 opt
->extended_regexp_option
= 1;
149 case GREP_PATTERN_TYPE_FIXED
:
153 case GREP_PATTERN_TYPE_PCRE
:
159 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
161 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
162 grep_set_pattern_type_option(pattern_type
, opt
);
163 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
164 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
165 else if (opt
->extended_regexp_option
)
167 * This branch *must* happen after setting from the
168 * opt->pattern_type_option above, we don't want
169 * grep.extendedRegexp to override grep.patternType!
171 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
174 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
175 const char *origin
, int no
,
176 enum grep_pat_token t
,
177 enum grep_header_field field
)
179 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
180 p
->pattern
= xmemdupz(pat
, patlen
);
181 p
->patternlen
= patlen
;
189 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
196 case GREP_PATTERN
: /* atom */
197 case GREP_PATTERN_HEAD
:
198 case GREP_PATTERN_BODY
:
200 struct grep_pat
*new_pat
;
202 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
203 while (++len
<= p
->patternlen
) {
204 if (*(--cp
) == '\n') {
211 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
212 p
->no
, p
->token
, p
->field
);
213 new_pat
->next
= p
->next
;
215 *tail
= &new_pat
->next
;
218 p
->patternlen
-= len
;
226 void append_header_grep_pattern(struct grep_opt
*opt
,
227 enum grep_header_field field
, const char *pat
)
229 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
230 GREP_PATTERN_HEAD
, field
);
231 if (field
== GREP_HEADER_REFLOG
)
232 opt
->use_reflog_filter
= 1;
233 do_append_grep_pat(&opt
->header_tail
, p
);
236 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
237 const char *origin
, int no
, enum grep_pat_token t
)
239 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
242 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
243 const char *origin
, int no
, enum grep_pat_token t
)
245 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
246 do_append_grep_pat(&opt
->pattern_tail
, p
);
249 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
251 struct grep_pat
*pat
;
252 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
255 ret
->pattern_list
= NULL
;
256 ret
->pattern_tail
= &ret
->pattern_list
;
258 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
260 if(pat
->token
== GREP_PATTERN_HEAD
)
261 append_header_grep_pattern(ret
, pat
->field
,
264 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
265 pat
->origin
, pat
->no
, pat
->token
);
271 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
277 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
279 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
283 die("%s'%s': %s", where
, p
->pattern
, error
);
286 static int is_fixed(const char *s
, size_t len
)
290 for (i
= 0; i
< len
; i
++) {
291 if (is_regex_special(s
[i
]))
299 #define GREP_PCRE2_DEBUG_MALLOC 0
301 static void *pcre2_malloc(PCRE2_SIZE size
, MAYBE_UNUSED
void *memory_data
)
303 void *pointer
= malloc(size
);
304 #if GREP_PCRE2_DEBUG_MALLOC
305 static int count
= 1;
306 fprintf(stderr
, "PCRE2:%p -> #%02d: alloc(%lu)\n", pointer
, count
++, size
);
311 static void pcre2_free(void *pointer
, MAYBE_UNUSED
void *memory_data
)
313 #if GREP_PCRE2_DEBUG_MALLOC
314 static int count
= 1;
316 fprintf(stderr
, "PCRE2:%p -> #%02d: free()\n", pointer
, count
++);
321 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
324 PCRE2_UCHAR errbuf
[256];
325 PCRE2_SIZE erroffset
;
326 int options
= PCRE2_MULTILINE
;
330 int literal
= !opt
->ignore_case
&& (p
->fixed
|| p
->is_fixed
);
333 * Call pcre2_general_context_create() before calling any
334 * other pcre2_*(). It sets up our malloc()/free() functions
335 * with which everything else is allocated.
337 p
->pcre2_general_context
= pcre2_general_context_create(
338 pcre2_malloc
, pcre2_free
, NULL
);
339 if (!p
->pcre2_general_context
)
340 die("Couldn't allocate PCRE2 general context");
342 if (opt
->ignore_case
) {
343 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
)) {
344 p
->pcre2_tables
= pcre2_maketables(p
->pcre2_general_context
);
345 p
->pcre2_compile_context
= pcre2_compile_context_create(p
->pcre2_general_context
);
346 pcre2_set_character_tables(p
->pcre2_compile_context
,
349 options
|= PCRE2_CASELESS
;
351 if (!opt
->ignore_locale
&& is_utf8_locale() && !literal
)
352 options
|= (PCRE2_UTF
| PCRE2_MATCH_INVALID_UTF
);
354 #ifdef GIT_PCRE2_VERSION_10_36_OR_HIGHER
355 /* Work around https://bugs.exim.org/show_bug.cgi?id=2642 fixed in 10.36 */
356 if (PCRE2_MATCH_INVALID_UTF
&& options
& (PCRE2_UTF
| PCRE2_CASELESS
))
357 options
|= PCRE2_NO_START_OPTIMIZE
;
360 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
361 p
->patternlen
, options
, &error
, &erroffset
,
362 p
->pcre2_compile_context
);
364 if (p
->pcre2_pattern
) {
365 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, p
->pcre2_general_context
);
366 if (!p
->pcre2_match_data
)
367 die("Couldn't allocate PCRE2 match data");
369 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
370 compile_regexp_failed(p
, (const char *)&errbuf
);
373 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
374 if (p
->pcre2_jit_on
) {
375 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
377 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
380 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
381 * tells us whether the library itself supports JIT,
382 * but to see whether we're going to be actually using
383 * JIT we need to extract PCRE2_INFO_JITSIZE from the
384 * pattern *after* we do pcre2_jit_compile() above.
386 * This is because if the pattern contains the
387 * (*NO_JIT) verb (see pcre2syntax(3))
388 * pcre2_jit_compile() will exit early with 0. If we
389 * then proceed to call pcre2_jit_match() further down
390 * the line instead of pcre2_match() we'll either
391 * segfault (pre PCRE 10.31) or run into a fatal error
394 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
396 BUG("pcre2_pattern_info() failed: %d", patinforet
);
397 if (jitsizearg
== 0) {
404 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
405 regmatch_t
*match
, int eflags
)
409 PCRE2_UCHAR errbuf
[256];
411 if (eflags
& REG_NOTBOL
)
412 flags
|= PCRE2_NOTBOL
;
415 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
416 eol
- line
, 0, flags
, p
->pcre2_match_data
,
419 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
420 eol
- line
, 0, flags
, p
->pcre2_match_data
,
423 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
424 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
425 die("%s failed with error code %d: %s",
426 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
430 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
432 match
->rm_so
= (int)ovector
[0];
433 match
->rm_eo
= (int)ovector
[1];
439 static void free_pcre2_pattern(struct grep_pat
*p
)
441 pcre2_compile_context_free(p
->pcre2_compile_context
);
442 pcre2_code_free(p
->pcre2_pattern
);
443 pcre2_match_data_free(p
->pcre2_match_data
);
444 #ifdef GIT_PCRE2_VERSION_10_34_OR_HIGHER
445 pcre2_maketables_free(p
->pcre2_general_context
, p
->pcre2_tables
);
447 free((void *)p
->pcre2_tables
);
449 pcre2_general_context_free(p
->pcre2_general_context
);
451 #else /* !USE_LIBPCRE2 */
452 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
454 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
457 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
458 regmatch_t
*match
, int eflags
)
463 static void free_pcre2_pattern(struct grep_pat
*p
)
467 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
469 struct strbuf sb
= STRBUF_INIT
;
473 basic_regex_quote_buf(&sb
, p
->pattern
);
474 if (opt
->ignore_case
)
475 regflags
|= REG_ICASE
;
476 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
480 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
481 compile_regexp_failed(p
, errbuf
);
484 #endif /* !USE_LIBPCRE2 */
486 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
489 int regflags
= REG_NEWLINE
;
491 p
->word_regexp
= opt
->word_regexp
;
492 p
->ignore_case
= opt
->ignore_case
;
493 p
->fixed
= opt
->fixed
;
495 if (memchr(p
->pattern
, 0, p
->patternlen
) && !opt
->pcre2
)
496 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
498 p
->is_fixed
= is_fixed(p
->pattern
, p
->patternlen
);
500 if (!p
->fixed
&& !p
->is_fixed
) {
501 const char *no_jit
= "(*NO_JIT)";
502 const int no_jit_len
= strlen(no_jit
);
503 if (starts_with(p
->pattern
, no_jit
) &&
504 is_fixed(p
->pattern
+ no_jit_len
,
505 p
->patternlen
- no_jit_len
))
509 if (p
->fixed
|| p
->is_fixed
) {
512 compile_pcre2_pattern(p
, opt
);
515 * E.g. t7811-grep-open.sh relies on the
516 * pattern being restored.
518 char *old_pattern
= p
->pattern
;
519 size_t old_patternlen
= p
->patternlen
;
520 struct strbuf sb
= STRBUF_INIT
;
523 * There is the PCRE2_LITERAL flag, but it's
524 * only in PCRE v2 10.30 and later. Needing to
525 * ifdef our way around that and dealing with
526 * it + PCRE2_MULTILINE being an error is more
527 * complex than just quoting this ourselves.
529 strbuf_add(&sb
, "\\Q", 2);
530 strbuf_add(&sb
, p
->pattern
, p
->patternlen
);
531 strbuf_add(&sb
, "\\E", 2);
534 p
->patternlen
= sb
.len
;
535 compile_pcre2_pattern(p
, opt
);
536 p
->pattern
= old_pattern
;
537 p
->patternlen
= old_patternlen
;
540 #else /* !USE_LIBPCRE2 */
541 compile_fixed_regexp(p
, opt
);
542 #endif /* !USE_LIBPCRE2 */
547 compile_pcre2_pattern(p
, opt
);
552 regflags
|= REG_ICASE
;
553 if (opt
->extended_regexp_option
)
554 regflags
|= REG_EXTENDED
;
555 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
558 regerror(err
, &p
->regexp
, errbuf
, 1024);
559 compile_regexp_failed(p
, errbuf
);
563 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
564 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
573 case GREP_PATTERN
: /* atom */
574 case GREP_PATTERN_HEAD
:
575 case GREP_PATTERN_BODY
:
577 x
->node
= GREP_NODE_ATOM
;
581 case GREP_OPEN_PAREN
:
583 x
= compile_pattern_or(list
);
584 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
585 die("unmatched parenthesis");
586 *list
= (*list
)->next
;
593 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
604 die("--not not followed by pattern expression");
607 x
->node
= GREP_NODE_NOT
;
608 x
->u
.unary
= compile_pattern_not(list
);
610 die("--not followed by non pattern expression");
613 return compile_pattern_atom(list
);
617 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
620 struct grep_expr
*x
, *y
, *z
;
622 x
= compile_pattern_not(list
);
624 if (p
&& p
->token
== GREP_AND
) {
626 die("--and not preceded by pattern expression");
628 die("--and not followed by pattern expression");
630 y
= compile_pattern_and(list
);
632 die("--and not followed by pattern expression");
634 z
->node
= GREP_NODE_AND
;
635 z
->u
.binary
.left
= x
;
636 z
->u
.binary
.right
= y
;
642 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
645 struct grep_expr
*x
, *y
, *z
;
647 x
= compile_pattern_and(list
);
649 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
650 y
= compile_pattern_or(list
);
652 die("not a pattern expression %s", p
->pattern
);
654 z
->node
= GREP_NODE_OR
;
655 z
->u
.binary
.left
= x
;
656 z
->u
.binary
.right
= y
;
662 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
664 return compile_pattern_or(list
);
667 static struct grep_expr
*grep_not_expr(struct grep_expr
*expr
)
669 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
670 z
->node
= GREP_NODE_NOT
;
675 static struct grep_expr
*grep_true_expr(void)
677 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
678 z
->node
= GREP_NODE_TRUE
;
682 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
684 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
685 z
->node
= GREP_NODE_OR
;
686 z
->u
.binary
.left
= left
;
687 z
->u
.binary
.right
= right
;
691 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
694 struct grep_expr
*header_expr
;
695 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
696 enum grep_header_field fld
;
698 if (!opt
->header_list
)
701 for (p
= opt
->header_list
; p
; p
= p
->next
) {
702 if (p
->token
!= GREP_PATTERN_HEAD
)
703 BUG("a non-header pattern in grep header list.");
704 if (p
->field
< GREP_HEADER_FIELD_MIN
||
705 GREP_HEADER_FIELD_MAX
<= p
->field
)
706 BUG("unknown header field %d", p
->field
);
707 compile_regexp(p
, opt
);
710 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
711 header_group
[fld
] = NULL
;
713 for (p
= opt
->header_list
; p
; p
= p
->next
) {
715 struct grep_pat
*pp
= p
;
717 h
= compile_pattern_atom(&pp
);
718 if (!h
|| pp
!= p
->next
)
719 BUG("malformed header expr");
720 if (!header_group
[p
->field
]) {
721 header_group
[p
->field
] = h
;
724 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
729 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
730 if (!header_group
[fld
])
733 header_expr
= grep_true_expr();
734 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
739 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
741 struct grep_expr
*z
= x
;
744 assert(x
->node
== GREP_NODE_OR
);
745 if (x
->u
.binary
.right
&&
746 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
747 x
->u
.binary
.right
= y
;
750 x
= x
->u
.binary
.right
;
755 void compile_grep_patterns(struct grep_opt
*opt
)
758 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
760 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
762 case GREP_PATTERN
: /* atom */
763 case GREP_PATTERN_HEAD
:
764 case GREP_PATTERN_BODY
:
765 compile_regexp(p
, opt
);
773 if (opt
->all_match
|| opt
->no_body_match
|| header_expr
)
775 else if (!opt
->extended
)
778 p
= opt
->pattern_list
;
780 opt
->pattern_expression
= compile_pattern_expr(&p
);
782 die("incomplete pattern expression: %s", p
->pattern
);
784 if (opt
->no_body_match
&& opt
->pattern_expression
)
785 opt
->pattern_expression
= grep_not_expr(opt
->pattern_expression
);
790 if (!opt
->pattern_expression
)
791 opt
->pattern_expression
= header_expr
;
792 else if (opt
->all_match
)
793 opt
->pattern_expression
= grep_splice_or(header_expr
,
794 opt
->pattern_expression
);
796 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
801 static void free_pattern_expr(struct grep_expr
*x
)
808 free_pattern_expr(x
->u
.unary
);
812 free_pattern_expr(x
->u
.binary
.left
);
813 free_pattern_expr(x
->u
.binary
.right
);
819 void free_grep_patterns(struct grep_opt
*opt
)
821 struct grep_pat
*p
, *n
;
823 for (p
= opt
->pattern_list
; p
; p
= n
) {
826 case GREP_PATTERN
: /* atom */
827 case GREP_PATTERN_HEAD
:
828 case GREP_PATTERN_BODY
:
829 if (p
->pcre2_pattern
)
830 free_pcre2_pattern(p
);
843 free_pattern_expr(opt
->pattern_expression
);
846 static const char *end_of_line(const char *cp
, unsigned long *left
)
848 unsigned long l
= *left
;
849 while (l
&& *cp
!= '\n') {
857 static int word_char(char ch
)
859 return isalnum(ch
) || ch
== '_';
862 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
865 if (want_color(opt
->color
) && color
&& color
[0]) {
866 opt
->output(opt
, color
, strlen(color
));
867 opt
->output(opt
, data
, size
);
868 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
870 opt
->output(opt
, data
, size
);
873 static void output_sep(struct grep_opt
*opt
, char sign
)
875 if (opt
->null_following_name
)
876 opt
->output(opt
, "\0", 1);
878 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
881 static void show_name(struct grep_opt
*opt
, const char *name
)
883 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
884 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
887 static int patmatch(struct grep_pat
*p
,
888 const char *line
, const char *eol
,
889 regmatch_t
*match
, int eflags
)
893 if (p
->pcre2_pattern
)
894 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
896 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
902 static void strip_timestamp(const char *bol
, const char **eol_p
)
904 const char *eol
= *eol_p
;
906 while (bol
< --eol
) {
919 { "committer ", 10 },
923 static int headerless_match_one_pattern(struct grep_pat
*p
,
924 const char *bol
, const char *eol
,
925 enum grep_context ctx
,
926 regmatch_t
*pmatch
, int eflags
)
929 const char *start
= bol
;
931 if ((p
->token
!= GREP_PATTERN
) &&
932 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
936 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
938 if (hit
&& p
->word_regexp
) {
939 if ((pmatch
[0].rm_so
< 0) ||
940 (eol
- bol
) < pmatch
[0].rm_so
||
941 (pmatch
[0].rm_eo
< 0) ||
942 (eol
- bol
) < pmatch
[0].rm_eo
)
943 die("regexp returned nonsense");
945 /* Match beginning must be either beginning of the
946 * line, or at word boundary (i.e. the last char must
947 * not be a word char). Similarly, match end must be
948 * either end of the line, or at word boundary
949 * (i.e. the next char must not be a word char).
951 if ( ((pmatch
[0].rm_so
== 0) ||
952 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
953 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
954 !word_char(bol
[pmatch
[0].rm_eo
])) )
959 /* Words consist of at least one character. */
960 if (pmatch
->rm_so
== pmatch
->rm_eo
)
963 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
964 /* There could be more than one match on the
965 * line, and the first match might not be
966 * strict word match. But later ones could be!
967 * Forward to the next possible start, i.e. the
968 * next position following a non-word char.
970 bol
= pmatch
[0].rm_so
+ bol
+ 1;
971 while (word_char(bol
[-1]) && bol
< eol
)
973 eflags
|= REG_NOTBOL
;
979 pmatch
[0].rm_so
+= bol
- start
;
980 pmatch
[0].rm_eo
+= bol
- start
;
985 static int match_one_pattern(struct grep_pat
*p
,
986 const char *bol
, const char *eol
,
987 enum grep_context ctx
, regmatch_t
*pmatch
,
993 if (p
->token
== GREP_PATTERN_HEAD
) {
994 assert(p
->field
< ARRAY_SIZE(header_field
));
995 field
= header_field
[p
->field
].field
;
996 len
= header_field
[p
->field
].len
;
997 if (strncmp(bol
, field
, len
))
1002 case GREP_HEADER_AUTHOR
:
1003 case GREP_HEADER_COMMITTER
:
1004 strip_timestamp(bol
, &eol
);
1011 return headerless_match_one_pattern(p
, bol
, eol
, ctx
, pmatch
, eflags
);
1015 static int match_expr_eval(struct grep_opt
*opt
, struct grep_expr
*x
,
1016 const char *bol
, const char *eol
,
1017 enum grep_context ctx
, ssize_t
*col
,
1018 ssize_t
*icol
, int collect_hits
)
1023 die("Not a valid grep expression");
1025 case GREP_NODE_TRUE
:
1028 case GREP_NODE_ATOM
:
1031 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
,
1033 if (h
&& (*col
< 0 || tmp
.rm_so
< *col
))
1036 if (x
->u
.atom
->token
== GREP_PATTERN_BODY
)
1041 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1043 h
= !match_expr_eval(opt
, x
->u
.unary
, bol
, eol
, ctx
, icol
, col
,
1047 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1049 if (h
|| opt
->columnnum
) {
1051 * Don't short-circuit AND when given --column, since a
1052 * NOT earlier in the tree may turn this into an OR. In
1053 * this case, see the below comment.
1055 h
&= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
,
1060 if (!(collect_hits
|| opt
->columnnum
)) {
1062 * Don't short-circuit OR when given --column (or
1063 * collecting hits) to ensure we don't skip a later
1064 * child that would produce an earlier match.
1066 return (match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
,
1067 ctx
, col
, icol
, 0) ||
1068 match_expr_eval(opt
, x
->u
.binary
.right
, bol
,
1069 eol
, ctx
, col
, icol
, 0));
1071 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1074 x
->u
.binary
.left
->hit
|= h
;
1075 h
|= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
, ctx
, col
,
1076 icol
, collect_hits
);
1079 die("Unexpected node type (internal error) %d", x
->node
);
1086 static int match_expr(struct grep_opt
*opt
,
1087 const char *bol
, const char *eol
,
1088 enum grep_context ctx
, ssize_t
*col
,
1089 ssize_t
*icol
, int collect_hits
)
1091 struct grep_expr
*x
= opt
->pattern_expression
;
1092 return match_expr_eval(opt
, x
, bol
, eol
, ctx
, col
, icol
, collect_hits
);
1095 static int match_line(struct grep_opt
*opt
,
1096 const char *bol
, const char *eol
,
1097 ssize_t
*col
, ssize_t
*icol
,
1098 enum grep_context ctx
, int collect_hits
)
1104 return match_expr(opt
, bol
, eol
, ctx
, col
, icol
,
1107 /* we do not call with collect_hits without being extended */
1108 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1110 if (match_one_pattern(p
, bol
, eol
, ctx
, &tmp
, 0)) {
1112 if (!opt
->columnnum
) {
1114 * Without --column, any single match on a line
1115 * is enough to know that it needs to be
1116 * printed. With --column, scan _all_ patterns
1117 * to find the earliest.
1121 if (*col
< 0 || tmp
.rm_so
< *col
)
1128 static int match_next_pattern(struct grep_pat
*p
,
1129 const char *bol
, const char *eol
,
1130 enum grep_context ctx
,
1131 regmatch_t
*pmatch
, int eflags
)
1135 if (!headerless_match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1137 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1139 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1140 if (match
.rm_so
> pmatch
->rm_so
)
1142 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1145 pmatch
->rm_so
= match
.rm_so
;
1146 pmatch
->rm_eo
= match
.rm_eo
;
1150 int grep_next_match(struct grep_opt
*opt
,
1151 const char *bol
, const char *eol
,
1152 enum grep_context ctx
, regmatch_t
*pmatch
,
1153 enum grep_header_field field
, int eflags
)
1158 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1160 for (p
= ((ctx
== GREP_CONTEXT_HEAD
)
1161 ? opt
->header_list
: opt
->pattern_list
);
1164 case GREP_PATTERN_HEAD
:
1165 if ((field
!= GREP_HEADER_FIELD_MAX
) &&
1166 (p
->field
!= field
))
1169 case GREP_PATTERN
: /* atom */
1170 case GREP_PATTERN_BODY
:
1171 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1182 static void show_line_header(struct grep_opt
*opt
, const char *name
,
1183 unsigned lno
, ssize_t cno
, char sign
)
1185 if (opt
->heading
&& opt
->last_shown
== 0) {
1186 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1187 opt
->output(opt
, "\n", 1);
1189 opt
->last_shown
= lno
;
1191 if (!opt
->heading
&& opt
->pathname
) {
1192 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1193 output_sep(opt
, sign
);
1197 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1198 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1199 output_sep(opt
, sign
);
1202 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1203 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1204 * being called with a context line.
1206 if (opt
->columnnum
&& cno
) {
1208 xsnprintf(buf
, sizeof(buf
), "%"PRIuMAX
, (uintmax_t)cno
);
1209 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_COLUMNNO
]);
1210 output_sep(opt
, sign
);
1214 static void show_line(struct grep_opt
*opt
,
1215 const char *bol
, const char *eol
,
1216 const char *name
, unsigned lno
, ssize_t cno
, char sign
)
1218 int rest
= eol
- bol
;
1219 const char *match_color
= NULL
;
1220 const char *line_color
= NULL
;
1222 if (opt
->file_break
&& opt
->last_shown
== 0) {
1223 if (opt
->show_hunk_mark
)
1224 opt
->output(opt
, "\n", 1);
1225 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1226 if (opt
->last_shown
== 0) {
1227 if (opt
->show_hunk_mark
) {
1228 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1229 opt
->output(opt
, "\n", 1);
1231 } else if (lno
> opt
->last_shown
+ 1) {
1232 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1233 opt
->output(opt
, "\n", 1);
1236 if (!opt
->only_matching
) {
1238 * In case the line we're being called with contains more than
1239 * one match, leave printing each header to the loop below.
1241 show_line_header(opt
, name
, lno
, cno
, sign
);
1243 if (opt
->color
|| opt
->only_matching
) {
1245 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1250 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1252 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1254 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1255 else if (sign
== '-')
1256 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1257 else if (sign
== '=')
1258 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1260 while (grep_next_match(opt
, bol
, eol
, ctx
, &match
,
1261 GREP_HEADER_FIELD_MAX
, eflags
)) {
1262 if (match
.rm_so
== match
.rm_eo
)
1265 if (opt
->only_matching
)
1266 show_line_header(opt
, name
, lno
, cno
, sign
);
1268 output_color(opt
, bol
, match
.rm_so
, line_color
);
1269 output_color(opt
, bol
+ match
.rm_so
,
1270 match
.rm_eo
- match
.rm_so
, match_color
);
1271 if (opt
->only_matching
)
1272 opt
->output(opt
, "\n", 1);
1275 rest
-= match
.rm_eo
;
1276 eflags
= REG_NOTBOL
;
1279 if (!opt
->only_matching
) {
1280 output_color(opt
, bol
, rest
, line_color
);
1281 opt
->output(opt
, "\n", 1);
1288 * This lock protects access to the gitattributes machinery, which is
1291 pthread_mutex_t grep_attr_mutex
;
1293 static inline void grep_attr_lock(void)
1296 pthread_mutex_lock(&grep_attr_mutex
);
1299 static inline void grep_attr_unlock(void)
1302 pthread_mutex_unlock(&grep_attr_mutex
);
1305 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
,
1306 const char *bol
, const char *eol
)
1308 xdemitconf_t
*xecfg
= opt
->priv
;
1309 if (xecfg
&& !xecfg
->find_func
) {
1310 grep_source_load_driver(gs
, opt
->repo
->index
);
1311 if (gs
->driver
->funcname
.pattern
) {
1312 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1313 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1315 xecfg
= opt
->priv
= NULL
;
1321 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1322 xecfg
->find_func_priv
) >= 0;
1327 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1332 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1333 const char *bol
, unsigned lno
)
1335 while (bol
> gs
->buf
) {
1336 const char *eol
= --bol
;
1338 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1342 if (lno
<= opt
->last_shown
)
1345 if (match_funcname(opt
, gs
, bol
, eol
)) {
1346 show_line(opt
, bol
, eol
, gs
->name
, lno
, 0, '=');
1352 static int is_empty_line(const char *bol
, const char *eol
);
1354 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1355 const char *bol
, const char *end
, unsigned lno
)
1357 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1358 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1360 if (opt
->pre_context
< lno
)
1361 from
= lno
- opt
->pre_context
;
1362 if (from
<= opt
->last_shown
)
1363 from
= opt
->last_shown
+ 1;
1365 if (opt
->funcbody
) {
1366 if (match_funcname(opt
, gs
, bol
, end
))
1369 funcname_needed
= 1;
1370 from
= opt
->last_shown
+ 1;
1374 while (bol
> gs
->buf
&& cur
> from
) {
1375 const char *next_bol
= bol
;
1376 const char *eol
= --bol
;
1378 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1381 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1382 match_funcname(opt
, gs
, bol
, eol
))) {
1391 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1393 funcname_needed
= 0;
1401 /* We need to look even further back to find a function signature. */
1402 if (opt
->funcname
&& funcname_needed
)
1403 show_funcname_line(opt
, gs
, bol
, cur
);
1407 const char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1409 while (*eol
!= '\n')
1411 show_line(opt
, bol
, eol
, gs
->name
, cur
, 0, sign
);
1417 static int should_lookahead(struct grep_opt
*opt
)
1422 return 0; /* punt for too complex stuff */
1425 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1426 if (p
->token
!= GREP_PATTERN
)
1427 return 0; /* punt for "header only" and stuff */
1432 static int look_ahead(struct grep_opt
*opt
,
1433 unsigned long *left_p
,
1437 unsigned lno
= *lno_p
;
1438 const char *bol
= *bol_p
;
1440 const char *sp
, *last_bol
;
1441 regoff_t earliest
= -1;
1443 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1447 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1448 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1450 if (earliest
< 0 || m
.rm_so
< earliest
)
1455 *bol_p
= bol
+ *left_p
;
1459 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1460 ; /* find the beginning of the line */
1463 for (sp
= bol
; sp
< last_bol
; sp
++) {
1467 *left_p
-= last_bol
- bol
;
1473 static int fill_textconv_grep(struct repository
*r
,
1474 struct userdiff_driver
*driver
,
1475 struct grep_source
*gs
)
1477 struct diff_filespec
*df
;
1481 if (!driver
|| !driver
->textconv
)
1482 return grep_source_load(gs
);
1485 * The textconv interface is intimately tied to diff_filespecs, so we
1486 * have to pretend to be one. If we could unify the grep_source
1487 * and diff_filespec structs, this mess could just go away.
1489 df
= alloc_filespec(gs
->path
);
1491 case GREP_SOURCE_OID
:
1492 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1494 case GREP_SOURCE_FILE
:
1495 fill_filespec(df
, null_oid(), 0, 0100644);
1498 BUG("attempt to textconv something without a path?");
1502 * fill_textconv is not remotely thread-safe; it modifies the global
1503 * diff tempfile structure, writes to the_repo's odb and might
1504 * internally call thread-unsafe functions such as the
1505 * prepare_packed_git() lazy-initializator. Because of the last two, we
1506 * must ensure mutual exclusion between this call and the object reading
1507 * API, thus we use obj_read_lock() here.
1509 * TODO: allowing text conversion to run in parallel with object
1510 * reading operations might increase performance in the multithreaded
1511 * non-worktreee git-grep with --textconv.
1514 size
= fill_textconv(r
, driver
, df
, &buf
);
1519 * The normal fill_textconv usage by the diff machinery would just keep
1520 * the textconv'd buf separate from the diff_filespec. But much of the
1521 * grep code passes around a grep_source and assumes that its "buf"
1522 * pointer is the beginning of the thing we are searching. So let's
1523 * install our textconv'd version into the grep_source, taking care not
1524 * to leak any existing buffer.
1526 grep_source_clear_data(gs
);
1533 static int is_empty_line(const char *bol
, const char *eol
)
1535 while (bol
< eol
&& isspace(*bol
))
1540 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1543 const char *peek_bol
= NULL
;
1546 unsigned last_hit
= 0;
1547 int binary_match_only
= 0;
1549 int try_lookahead
= 0;
1550 int show_function
= 0;
1551 struct userdiff_driver
*textconv
= NULL
;
1552 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1555 if (!opt
->status_only
&& gs
->name
== NULL
)
1556 BUG("grep call which could print a name requires "
1557 "grep_source.name be non-NULL");
1560 opt
->output
= std_output
;
1562 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1564 /* Show hunk marks, except for the first file. */
1565 if (opt
->last_shown
)
1566 opt
->show_hunk_mark
= 1;
1568 * If we're using threads then we can't easily identify
1569 * the first file. Always put hunk marks in that case
1570 * and skip the very first one later in work_done().
1572 if (opt
->output
!= std_output
)
1573 opt
->show_hunk_mark
= 1;
1575 opt
->last_shown
= 0;
1577 if (opt
->allow_textconv
) {
1578 grep_source_load_driver(gs
, opt
->repo
->index
);
1580 * We might set up the shared textconv cache data here, which
1581 * is not thread-safe. Also, get_oid_with_context() and
1582 * parse_object() might be internally called. As they are not
1583 * currently thread-safe and might be racy with object reading,
1584 * obj_read_lock() must be called.
1588 textconv
= userdiff_get_textconv(opt
->repo
, gs
->driver
);
1594 * We know the result of a textconv is text, so we only have to care
1595 * about binary handling if we are not using it.
1598 switch (opt
->binary
) {
1599 case GREP_BINARY_DEFAULT
:
1600 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1601 binary_match_only
= 1;
1603 case GREP_BINARY_NOMATCH
:
1604 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1605 return 0; /* Assume unmatch */
1607 case GREP_BINARY_TEXT
:
1610 BUG("unknown binary handling mode");
1614 memset(&xecfg
, 0, sizeof(xecfg
));
1617 try_lookahead
= should_lookahead(opt
);
1619 if (fill_textconv_grep(opt
->repo
, textconv
, gs
) < 0)
1628 ssize_t col
= -1, icol
= -1;
1631 * look_ahead() skips quickly to the line that possibly
1632 * has the next hit; don't call it if we need to do
1633 * something more than just skipping the current line
1634 * in response to an unmatch for the current line. E.g.
1635 * inside a post-context window, we will show the current
1636 * line as a context around the previous hit when it
1641 && (show_function
||
1642 lno
<= last_hit
+ opt
->post_context
))
1643 && look_ahead(opt
, &left
, &lno
, &bol
))
1645 eol
= end_of_line(bol
, &left
);
1647 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1648 ctx
= GREP_CONTEXT_BODY
;
1650 hit
= match_line(opt
, bol
, eol
, &col
, &icol
, ctx
, collect_hits
);
1655 /* "grep -v -e foo -e bla" should list lines
1656 * that do not have either, so inversion should
1661 if (opt
->unmatch_name_only
) {
1668 if (opt
->status_only
)
1670 if (opt
->name_only
) {
1671 show_name(opt
, gs
->name
);
1676 if (binary_match_only
) {
1677 opt
->output(opt
, "Binary file ", 12);
1678 output_color(opt
, gs
->name
, strlen(gs
->name
),
1679 opt
->colors
[GREP_COLOR_FILENAME
]);
1680 opt
->output(opt
, " matches\n", 9);
1683 /* Hit at this line. If we haven't shown the
1684 * pre-context lines, we would need to show them.
1686 if (opt
->pre_context
|| opt
->funcbody
)
1687 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1688 else if (opt
->funcname
)
1689 show_funcname_line(opt
, gs
, bol
, lno
);
1690 cno
= opt
->invert
? icol
: col
;
1693 * A negative cno indicates that there was no
1694 * match on the line. We are thus inverted and
1695 * being asked to show all lines that _don't_
1696 * match a given expression. Therefore, set cno
1697 * to 0 to suggest the whole line matches.
1701 show_line(opt
, bol
, eol
, gs
->name
, lno
, cno
+ 1, ':');
1707 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1708 unsigned long peek_left
= left
;
1709 const char *peek_eol
= eol
;
1712 * Trailing empty lines are not interesting.
1713 * Peek past them to see if they belong to the
1714 * body of the current function.
1717 while (is_empty_line(peek_bol
, peek_eol
)) {
1718 peek_bol
= peek_eol
+ 1;
1719 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1722 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1725 if (show_function
||
1726 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1727 /* If the last hit is within the post context,
1728 * we need to show this line.
1730 show_line(opt
, bol
, eol
, gs
->name
, lno
, col
+ 1, '-');
1744 if (opt
->status_only
)
1745 return opt
->unmatch_name_only
;
1746 if (opt
->unmatch_name_only
) {
1747 /* We did not see any hit, so we want to show this */
1748 show_name(opt
, gs
->name
);
1752 xdiff_clear_find_func(&xecfg
);
1756 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1757 * which feels mostly useless but sometimes useful. Maybe
1758 * make it another option? For now suppress them.
1760 if (opt
->count
&& count
) {
1762 if (opt
->pathname
) {
1763 output_color(opt
, gs
->name
, strlen(gs
->name
),
1764 opt
->colors
[GREP_COLOR_FILENAME
]);
1765 output_sep(opt
, ':');
1767 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1768 opt
->output(opt
, buf
, strlen(buf
));
1774 static void clr_hit_marker(struct grep_expr
*x
)
1776 /* All-hit markers are meaningful only at the very top level
1781 if (x
->node
!= GREP_NODE_OR
)
1783 x
->u
.binary
.left
->hit
= 0;
1784 x
= x
->u
.binary
.right
;
1788 static int chk_hit_marker(struct grep_expr
*x
)
1790 /* Top level nodes have hit markers. See if they all are hits */
1792 if (x
->node
!= GREP_NODE_OR
)
1794 if (!x
->u
.binary
.left
->hit
)
1796 x
= x
->u
.binary
.right
;
1800 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1803 * we do not have to do the two-pass grep when we do not check
1804 * buffer-wide "all-match".
1806 if (!opt
->all_match
&& !opt
->no_body_match
)
1807 return grep_source_1(opt
, gs
, 0);
1809 /* Otherwise the toplevel "or" terms hit a bit differently.
1810 * We first clear hit markers from them.
1812 clr_hit_marker(opt
->pattern_expression
);
1814 grep_source_1(opt
, gs
, 1);
1816 if (opt
->all_match
&& !chk_hit_marker(opt
->pattern_expression
))
1818 if (opt
->no_body_match
&& opt
->body_hit
)
1821 return grep_source_1(opt
, gs
, 0);
1824 static void grep_source_init_buf(struct grep_source
*gs
,
1828 gs
->type
= GREP_SOURCE_BUF
;
1834 gs
->identifier
= NULL
;
1837 int grep_buffer(struct grep_opt
*opt
, const char *buf
, unsigned long size
)
1839 struct grep_source gs
;
1842 grep_source_init_buf(&gs
, buf
, size
);
1844 r
= grep_source(opt
, &gs
);
1846 grep_source_clear(&gs
);
1850 void grep_source_init_file(struct grep_source
*gs
, const char *name
,
1853 gs
->type
= GREP_SOURCE_FILE
;
1854 gs
->name
= xstrdup_or_null(name
);
1855 gs
->path
= xstrdup_or_null(path
);
1859 gs
->identifier
= xstrdup(path
);
1862 void grep_source_init_oid(struct grep_source
*gs
, const char *name
,
1863 const char *path
, const struct object_id
*oid
,
1864 struct repository
*repo
)
1866 gs
->type
= GREP_SOURCE_OID
;
1867 gs
->name
= xstrdup_or_null(name
);
1868 gs
->path
= xstrdup_or_null(path
);
1872 gs
->identifier
= oiddup(oid
);
1876 void grep_source_clear(struct grep_source
*gs
)
1878 FREE_AND_NULL(gs
->name
);
1879 FREE_AND_NULL(gs
->path
);
1880 FREE_AND_NULL(gs
->identifier
);
1881 grep_source_clear_data(gs
);
1884 void grep_source_clear_data(struct grep_source
*gs
)
1887 case GREP_SOURCE_FILE
:
1888 case GREP_SOURCE_OID
:
1889 /* these types own the buffer */
1890 free((char *)gs
->buf
);
1894 case GREP_SOURCE_BUF
:
1895 /* leave user-provided buf intact */
1900 static int grep_source_load_oid(struct grep_source
*gs
)
1902 enum object_type type
;
1904 gs
->buf
= repo_read_object_file(gs
->repo
, gs
->identifier
, &type
,
1907 return error(_("'%s': unable to read %s"),
1909 oid_to_hex(gs
->identifier
));
1913 static int grep_source_load_file(struct grep_source
*gs
)
1915 const char *filename
= gs
->identifier
;
1921 if (lstat(filename
, &st
) < 0) {
1923 if (errno
!= ENOENT
)
1924 error_errno(_("failed to stat '%s'"), filename
);
1927 if (!S_ISREG(st
.st_mode
))
1929 size
= xsize_t(st
.st_size
);
1930 i
= open(filename
, O_RDONLY
);
1933 data
= xmallocz(size
);
1934 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
1935 error_errno(_("'%s': short read"), filename
);
1947 static int grep_source_load(struct grep_source
*gs
)
1953 case GREP_SOURCE_FILE
:
1954 return grep_source_load_file(gs
);
1955 case GREP_SOURCE_OID
:
1956 return grep_source_load_oid(gs
);
1957 case GREP_SOURCE_BUF
:
1958 return gs
->buf
? 0 : -1;
1960 BUG("invalid grep_source type to load");
1963 void grep_source_load_driver(struct grep_source
*gs
,
1964 struct index_state
*istate
)
1971 gs
->driver
= userdiff_find_by_path(istate
, gs
->path
);
1973 gs
->driver
= userdiff_find_by_name("default");
1977 static int grep_source_is_binary(struct grep_source
*gs
,
1978 struct index_state
*istate
)
1980 grep_source_load_driver(gs
, istate
);
1981 if (gs
->driver
->binary
!= -1)
1982 return gs
->driver
->binary
;
1984 if (!grep_source_load(gs
))
1985 return buffer_is_binary(gs
->buf
, gs
->size
);