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 struct grep_opt grep_defaults
= {
26 .pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
,
28 [GREP_COLOR_CONTEXT
] = "",
29 [GREP_COLOR_FILENAME
] = "",
30 [GREP_COLOR_FUNCTION
] = "",
31 [GREP_COLOR_LINENO
] = "",
32 [GREP_COLOR_COLUMNNO
] = "",
33 [GREP_COLOR_MATCH_CONTEXT
] = GIT_COLOR_BOLD_RED
,
34 [GREP_COLOR_MATCH_SELECTED
] = GIT_COLOR_BOLD_RED
,
35 [GREP_COLOR_SELECTED
] = "",
36 [GREP_COLOR_SEP
] = GIT_COLOR_CYAN
,
44 static pcre2_general_context
*pcre2_global_context
;
46 static void *pcre2_malloc(PCRE2_SIZE size
, MAYBE_UNUSED
void *memory_data
)
51 static void pcre2_free(void *pointer
, MAYBE_UNUSED
void *memory_data
)
57 static const char *color_grep_slots
[] = {
58 [GREP_COLOR_CONTEXT
] = "context",
59 [GREP_COLOR_FILENAME
] = "filename",
60 [GREP_COLOR_FUNCTION
] = "function",
61 [GREP_COLOR_LINENO
] = "lineNumber",
62 [GREP_COLOR_COLUMNNO
] = "column",
63 [GREP_COLOR_MATCH_CONTEXT
] = "matchContext",
64 [GREP_COLOR_MATCH_SELECTED
] = "matchSelected",
65 [GREP_COLOR_SELECTED
] = "selected",
66 [GREP_COLOR_SEP
] = "separator",
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
;
95 if (userdiff_config(var
, value
) < 0)
99 * The instance of grep_opt that we set up here is copied by
100 * grep_init() to be used by each individual invocation.
101 * When populating a new field of this structure here, be
102 * sure to think about ownership -- e.g., you might need to
103 * override the shallow copy in grep_init() with a deep copy.
106 if (!strcmp(var
, "grep.extendedregexp")) {
107 opt
->extended_regexp_option
= git_config_bool(var
, value
);
111 if (!strcmp(var
, "grep.patterntype")) {
112 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
116 if (!strcmp(var
, "grep.linenumber")) {
117 opt
->linenum
= git_config_bool(var
, value
);
120 if (!strcmp(var
, "grep.column")) {
121 opt
->columnnum
= git_config_bool(var
, value
);
125 if (!strcmp(var
, "grep.fullname")) {
126 opt
->relative
= !git_config_bool(var
, value
);
130 if (!strcmp(var
, "color.grep"))
131 opt
->color
= git_config_colorbool(var
, value
);
132 if (!strcmp(var
, "color.grep.match")) {
133 if (grep_config("color.grep.matchcontext", value
, cb
) < 0)
135 if (grep_config("color.grep.matchselected", value
, cb
) < 0)
137 } else if (skip_prefix(var
, "color.grep.", &slot
)) {
138 int i
= LOOKUP_CONFIG(color_grep_slots
, slot
);
143 color
= opt
->colors
[i
];
145 return config_error_nonbool(var
);
146 return color_parse(value
, color
);
152 * Initialize one instance of grep_opt and copy the
153 * default values from the template we read the configuration
154 * information in an earlier call to git_config(grep_config).
156 * If using PCRE, make sure that the library is configured
157 * to use the same allocator as Git (e.g. nedmalloc on Windows).
159 * Any allocated memory needs to be released in grep_destroy().
161 void grep_init(struct grep_opt
*opt
, struct repository
*repo
, const char *prefix
)
163 #if defined(USE_LIBPCRE2)
164 if (!pcre2_global_context
)
165 pcre2_global_context
= pcre2_general_context_create(
166 pcre2_malloc
, pcre2_free
, NULL
);
170 pcre_malloc
= malloc
;
174 *opt
= grep_defaults
;
177 opt
->prefix
= prefix
;
178 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
179 opt
->pattern_tail
= &opt
->pattern_list
;
180 opt
->header_tail
= &opt
->header_list
;
183 void grep_destroy(void)
186 pcre2_general_context_free(pcre2_global_context
);
190 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
193 * When committing to the pattern type by setting the relevant
194 * fields in grep_opt it's generally not necessary to zero out
195 * the fields we're not choosing, since they won't have been
196 * set by anything. The extended_regexp_option field is the
197 * only exception to this.
199 * This is because in the process of parsing grep.patternType
200 * & grep.extendedRegexp we set opt->pattern_type_option and
201 * opt->extended_regexp_option, respectively. We then
202 * internally use opt->extended_regexp_option to see if we're
203 * compiling an ERE. It must be unset if that's not actually
206 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
207 opt
->extended_regexp_option
)
208 opt
->extended_regexp_option
= 0;
210 switch (pattern_type
) {
211 case GREP_PATTERN_TYPE_UNSPECIFIED
:
214 case GREP_PATTERN_TYPE_BRE
:
217 case GREP_PATTERN_TYPE_ERE
:
218 opt
->extended_regexp_option
= 1;
221 case GREP_PATTERN_TYPE_FIXED
:
225 case GREP_PATTERN_TYPE_PCRE
:
230 * It's important that pcre1 always be assigned to
231 * even when there's no USE_LIBPCRE* defined. We still
232 * call the PCRE stub function, it just dies with
233 * "cannot use Perl-compatible regexes[...]".
241 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
243 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
244 grep_set_pattern_type_option(pattern_type
, opt
);
245 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
246 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
247 else if (opt
->extended_regexp_option
)
249 * This branch *must* happen after setting from the
250 * opt->pattern_type_option above, we don't want
251 * grep.extendedRegexp to override grep.patternType!
253 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
256 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
257 const char *origin
, int no
,
258 enum grep_pat_token t
,
259 enum grep_header_field field
)
261 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
262 p
->pattern
= xmemdupz(pat
, patlen
);
263 p
->patternlen
= patlen
;
271 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
278 case GREP_PATTERN
: /* atom */
279 case GREP_PATTERN_HEAD
:
280 case GREP_PATTERN_BODY
:
282 struct grep_pat
*new_pat
;
284 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
285 while (++len
<= p
->patternlen
) {
286 if (*(--cp
) == '\n') {
293 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
294 p
->no
, p
->token
, p
->field
);
295 new_pat
->next
= p
->next
;
297 *tail
= &new_pat
->next
;
300 p
->patternlen
-= len
;
308 void append_header_grep_pattern(struct grep_opt
*opt
,
309 enum grep_header_field field
, const char *pat
)
311 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
312 GREP_PATTERN_HEAD
, field
);
313 if (field
== GREP_HEADER_REFLOG
)
314 opt
->use_reflog_filter
= 1;
315 do_append_grep_pat(&opt
->header_tail
, p
);
318 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
319 const char *origin
, int no
, enum grep_pat_token t
)
321 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
324 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
325 const char *origin
, int no
, enum grep_pat_token t
)
327 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
328 do_append_grep_pat(&opt
->pattern_tail
, p
);
331 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
333 struct grep_pat
*pat
;
334 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
337 ret
->pattern_list
= NULL
;
338 ret
->pattern_tail
= &ret
->pattern_list
;
340 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
342 if(pat
->token
== GREP_PATTERN_HEAD
)
343 append_header_grep_pattern(ret
, pat
->field
,
346 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
347 pat
->origin
, pat
->no
, pat
->token
);
353 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
359 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
361 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
365 die("%s'%s': %s", where
, p
->pattern
, error
);
368 static int is_fixed(const char *s
, size_t len
)
372 for (i
= 0; i
< len
; i
++) {
373 if (is_regex_special(s
[i
]))
381 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
385 int options
= PCRE_MULTILINE
;
386 int study_options
= 0;
388 if (opt
->ignore_case
) {
389 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
))
390 p
->pcre1_tables
= pcre_maketables();
391 options
|= PCRE_CASELESS
;
393 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
))
394 options
|= PCRE_UTF8
;
396 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
398 if (!p
->pcre1_regexp
)
399 compile_regexp_failed(p
, error
);
401 #if defined(PCRE_CONFIG_JIT) && !defined(NO_LIBPCRE1_JIT)
402 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
405 study_options
= PCRE_STUDY_JIT_COMPILE
;
408 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, study_options
, &error
);
409 if (!p
->pcre1_extra_info
&& error
)
413 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
414 regmatch_t
*match
, int eflags
)
416 int ovector
[30], ret
, flags
= PCRE_NO_UTF8_CHECK
;
418 if (eflags
& REG_NOTBOL
)
419 flags
|= PCRE_NOTBOL
;
421 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
422 eol
- line
, 0, flags
, ovector
,
423 ARRAY_SIZE(ovector
));
425 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
426 die("pcre_exec failed with error code %d", ret
);
429 match
->rm_so
= ovector
[0];
430 match
->rm_eo
= ovector
[1];
436 static void free_pcre1_regexp(struct grep_pat
*p
)
438 pcre_free(p
->pcre1_regexp
);
439 #ifdef PCRE_CONFIG_JIT
441 pcre_free_study(p
->pcre1_extra_info
);
444 pcre_free(p
->pcre1_extra_info
);
445 pcre_free((void *)p
->pcre1_tables
);
447 #else /* !USE_LIBPCRE1 */
448 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
450 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
453 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
454 regmatch_t
*match
, int eflags
)
459 static void free_pcre1_regexp(struct grep_pat
*p
)
462 #endif /* !USE_LIBPCRE1 */
465 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
468 PCRE2_UCHAR errbuf
[256];
469 PCRE2_SIZE erroffset
;
470 int options
= PCRE2_MULTILINE
;
477 p
->pcre2_compile_context
= NULL
;
479 /* pcre2_global_context is initialized in append_grep_pattern */
480 if (opt
->ignore_case
) {
481 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
)) {
482 if (!pcre2_global_context
)
483 BUG("pcre2_global_context uninitialized");
484 p
->pcre2_tables
= pcre2_maketables(pcre2_global_context
);
485 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
486 pcre2_set_character_tables(p
->pcre2_compile_context
,
489 options
|= PCRE2_CASELESS
;
491 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
) &&
492 !(!opt
->ignore_case
&& (p
->fixed
|| p
->is_fixed
)))
493 options
|= PCRE2_UTF
;
495 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
496 p
->patternlen
, options
, &error
, &erroffset
,
497 p
->pcre2_compile_context
);
499 if (p
->pcre2_pattern
) {
500 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, NULL
);
501 if (!p
->pcre2_match_data
)
502 die("Couldn't allocate PCRE2 match data");
504 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
505 compile_regexp_failed(p
, (const char *)&errbuf
);
508 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
509 if (p
->pcre2_jit_on
) {
510 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
512 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
515 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
516 * tells us whether the library itself supports JIT,
517 * but to see whether we're going to be actually using
518 * JIT we need to extract PCRE2_INFO_JITSIZE from the
519 * pattern *after* we do pcre2_jit_compile() above.
521 * This is because if the pattern contains the
522 * (*NO_JIT) verb (see pcre2syntax(3))
523 * pcre2_jit_compile() will exit early with 0. If we
524 * then proceed to call pcre2_jit_match() further down
525 * the line instead of pcre2_match() we'll either
526 * segfault (pre PCRE 10.31) or run into a fatal error
529 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
531 BUG("pcre2_pattern_info() failed: %d", patinforet
);
532 if (jitsizearg
== 0) {
539 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
540 regmatch_t
*match
, int eflags
)
544 PCRE2_UCHAR errbuf
[256];
546 if (eflags
& REG_NOTBOL
)
547 flags
|= PCRE2_NOTBOL
;
550 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
551 eol
- line
, 0, flags
, p
->pcre2_match_data
,
554 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
555 eol
- line
, 0, flags
, p
->pcre2_match_data
,
558 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
559 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
560 die("%s failed with error code %d: %s",
561 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
565 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
567 match
->rm_so
= (int)ovector
[0];
568 match
->rm_eo
= (int)ovector
[1];
574 static void free_pcre2_pattern(struct grep_pat
*p
)
576 pcre2_compile_context_free(p
->pcre2_compile_context
);
577 pcre2_code_free(p
->pcre2_pattern
);
578 pcre2_match_data_free(p
->pcre2_match_data
);
579 free((void *)p
->pcre2_tables
);
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
);
615 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
616 compile_regexp_failed(p
, errbuf
);
619 #endif /* !USE_LIBPCRE2 */
621 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
624 int regflags
= REG_NEWLINE
;
626 p
->word_regexp
= opt
->word_regexp
;
627 p
->ignore_case
= opt
->ignore_case
;
628 p
->fixed
= opt
->fixed
;
630 if (memchr(p
->pattern
, 0, p
->patternlen
) && !opt
->pcre2
)
631 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
633 p
->is_fixed
= is_fixed(p
->pattern
, p
->patternlen
);
635 if (!p
->fixed
&& !p
->is_fixed
) {
636 const char *no_jit
= "(*NO_JIT)";
637 const int no_jit_len
= strlen(no_jit
);
638 if (starts_with(p
->pattern
, no_jit
) &&
639 is_fixed(p
->pattern
+ no_jit_len
,
640 p
->patternlen
- no_jit_len
))
644 if (p
->fixed
|| p
->is_fixed
) {
648 compile_pcre2_pattern(p
, opt
);
651 * E.g. t7811-grep-open.sh relies on the
652 * pattern being restored.
654 char *old_pattern
= p
->pattern
;
655 size_t old_patternlen
= p
->patternlen
;
656 struct strbuf sb
= STRBUF_INIT
;
659 * There is the PCRE2_LITERAL flag, but it's
660 * only in PCRE v2 10.30 and later. Needing to
661 * ifdef our way around that and dealing with
662 * it + PCRE2_MULTILINE being an error is more
663 * complex than just quoting this ourselves.
665 strbuf_add(&sb
, "\\Q", 2);
666 strbuf_add(&sb
, p
->pattern
, p
->patternlen
);
667 strbuf_add(&sb
, "\\E", 2);
670 p
->patternlen
= sb
.len
;
671 compile_pcre2_pattern(p
, opt
);
672 p
->pattern
= old_pattern
;
673 p
->patternlen
= old_patternlen
;
676 #else /* !USE_LIBPCRE2 */
677 compile_fixed_regexp(p
, opt
);
678 #endif /* !USE_LIBPCRE2 */
683 compile_pcre2_pattern(p
, opt
);
688 compile_pcre1_regexp(p
, opt
);
693 regflags
|= REG_ICASE
;
694 if (opt
->extended_regexp_option
)
695 regflags
|= REG_EXTENDED
;
696 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
699 regerror(err
, &p
->regexp
, errbuf
, 1024);
700 compile_regexp_failed(p
, errbuf
);
704 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
705 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
714 case GREP_PATTERN
: /* atom */
715 case GREP_PATTERN_HEAD
:
716 case GREP_PATTERN_BODY
:
717 x
= xcalloc(1, sizeof (struct grep_expr
));
718 x
->node
= GREP_NODE_ATOM
;
722 case GREP_OPEN_PAREN
:
724 x
= compile_pattern_or(list
);
725 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
726 die("unmatched parenthesis");
727 *list
= (*list
)->next
;
734 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
745 die("--not not followed by pattern expression");
747 x
= xcalloc(1, sizeof (struct grep_expr
));
748 x
->node
= GREP_NODE_NOT
;
749 x
->u
.unary
= compile_pattern_not(list
);
751 die("--not followed by non pattern expression");
754 return compile_pattern_atom(list
);
758 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
761 struct grep_expr
*x
, *y
, *z
;
763 x
= compile_pattern_not(list
);
765 if (p
&& p
->token
== GREP_AND
) {
767 die("--and not followed by pattern expression");
769 y
= compile_pattern_and(list
);
771 die("--and not followed by pattern expression");
772 z
= xcalloc(1, sizeof (struct grep_expr
));
773 z
->node
= GREP_NODE_AND
;
774 z
->u
.binary
.left
= x
;
775 z
->u
.binary
.right
= y
;
781 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
784 struct grep_expr
*x
, *y
, *z
;
786 x
= compile_pattern_and(list
);
788 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
789 y
= compile_pattern_or(list
);
791 die("not a pattern expression %s", p
->pattern
);
792 z
= xcalloc(1, sizeof (struct grep_expr
));
793 z
->node
= GREP_NODE_OR
;
794 z
->u
.binary
.left
= x
;
795 z
->u
.binary
.right
= y
;
801 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
803 return compile_pattern_or(list
);
806 static struct grep_expr
*grep_true_expr(void)
808 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
809 z
->node
= GREP_NODE_TRUE
;
813 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
815 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
816 z
->node
= GREP_NODE_OR
;
817 z
->u
.binary
.left
= left
;
818 z
->u
.binary
.right
= right
;
822 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
825 struct grep_expr
*header_expr
;
826 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
827 enum grep_header_field fld
;
829 if (!opt
->header_list
)
832 for (p
= opt
->header_list
; p
; p
= p
->next
) {
833 if (p
->token
!= GREP_PATTERN_HEAD
)
834 BUG("a non-header pattern in grep header list.");
835 if (p
->field
< GREP_HEADER_FIELD_MIN
||
836 GREP_HEADER_FIELD_MAX
<= p
->field
)
837 BUG("unknown header field %d", p
->field
);
838 compile_regexp(p
, opt
);
841 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
842 header_group
[fld
] = NULL
;
844 for (p
= opt
->header_list
; p
; p
= p
->next
) {
846 struct grep_pat
*pp
= p
;
848 h
= compile_pattern_atom(&pp
);
849 if (!h
|| pp
!= p
->next
)
850 BUG("malformed header expr");
851 if (!header_group
[p
->field
]) {
852 header_group
[p
->field
] = h
;
855 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
860 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
861 if (!header_group
[fld
])
864 header_expr
= grep_true_expr();
865 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
870 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
872 struct grep_expr
*z
= x
;
875 assert(x
->node
== GREP_NODE_OR
);
876 if (x
->u
.binary
.right
&&
877 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
878 x
->u
.binary
.right
= y
;
881 x
= x
->u
.binary
.right
;
886 void compile_grep_patterns(struct grep_opt
*opt
)
889 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
891 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
893 case GREP_PATTERN
: /* atom */
894 case GREP_PATTERN_HEAD
:
895 case GREP_PATTERN_BODY
:
896 compile_regexp(p
, opt
);
904 if (opt
->all_match
|| header_expr
)
906 else if (!opt
->extended
)
909 p
= opt
->pattern_list
;
911 opt
->pattern_expression
= compile_pattern_expr(&p
);
913 die("incomplete pattern expression: %s", p
->pattern
);
918 if (!opt
->pattern_expression
)
919 opt
->pattern_expression
= header_expr
;
920 else if (opt
->all_match
)
921 opt
->pattern_expression
= grep_splice_or(header_expr
,
922 opt
->pattern_expression
);
924 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
929 static void free_pattern_expr(struct grep_expr
*x
)
936 free_pattern_expr(x
->u
.unary
);
940 free_pattern_expr(x
->u
.binary
.left
);
941 free_pattern_expr(x
->u
.binary
.right
);
947 void free_grep_patterns(struct grep_opt
*opt
)
949 struct grep_pat
*p
, *n
;
951 for (p
= opt
->pattern_list
; p
; p
= n
) {
954 case GREP_PATTERN
: /* atom */
955 case GREP_PATTERN_HEAD
:
956 case GREP_PATTERN_BODY
:
958 free_pcre1_regexp(p
);
959 else if (p
->pcre2_pattern
)
960 free_pcre2_pattern(p
);
973 free_pattern_expr(opt
->pattern_expression
);
976 static char *end_of_line(char *cp
, unsigned long *left
)
978 unsigned long l
= *left
;
979 while (l
&& *cp
!= '\n') {
987 static int word_char(char ch
)
989 return isalnum(ch
) || ch
== '_';
992 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
995 if (want_color(opt
->color
) && color
&& color
[0]) {
996 opt
->output(opt
, color
, strlen(color
));
997 opt
->output(opt
, data
, size
);
998 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1000 opt
->output(opt
, data
, size
);
1003 static void output_sep(struct grep_opt
*opt
, char sign
)
1005 if (opt
->null_following_name
)
1006 opt
->output(opt
, "\0", 1);
1008 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
1011 static void show_name(struct grep_opt
*opt
, const char *name
)
1013 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1014 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1017 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1018 regmatch_t
*match
, int eflags
)
1022 if (p
->pcre1_regexp
)
1023 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1024 else if (p
->pcre2_pattern
)
1025 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1027 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1033 static int strip_timestamp(char *bol
, char **eol_p
)
1038 while (bol
< --eol
) {
1052 } header_field
[] = {
1054 { "committer ", 10 },
1058 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1059 enum grep_context ctx
,
1060 regmatch_t
*pmatch
, int eflags
)
1064 const char *start
= bol
;
1066 if ((p
->token
!= GREP_PATTERN
) &&
1067 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1070 if (p
->token
== GREP_PATTERN_HEAD
) {
1073 assert(p
->field
< ARRAY_SIZE(header_field
));
1074 field
= header_field
[p
->field
].field
;
1075 len
= header_field
[p
->field
].len
;
1076 if (strncmp(bol
, field
, len
))
1080 case GREP_HEADER_AUTHOR
:
1081 case GREP_HEADER_COMMITTER
:
1082 saved_ch
= strip_timestamp(bol
, &eol
);
1090 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1092 if (hit
&& p
->word_regexp
) {
1093 if ((pmatch
[0].rm_so
< 0) ||
1094 (eol
- bol
) < pmatch
[0].rm_so
||
1095 (pmatch
[0].rm_eo
< 0) ||
1096 (eol
- bol
) < pmatch
[0].rm_eo
)
1097 die("regexp returned nonsense");
1099 /* Match beginning must be either beginning of the
1100 * line, or at word boundary (i.e. the last char must
1101 * not be a word char). Similarly, match end must be
1102 * either end of the line, or at word boundary
1103 * (i.e. the next char must not be a word char).
1105 if ( ((pmatch
[0].rm_so
== 0) ||
1106 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1107 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1108 !word_char(bol
[pmatch
[0].rm_eo
])) )
1113 /* Words consist of at least one character. */
1114 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1117 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1118 /* There could be more than one match on the
1119 * line, and the first match might not be
1120 * strict word match. But later ones could be!
1121 * Forward to the next possible start, i.e. the
1122 * next position following a non-word char.
1124 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1125 while (word_char(bol
[-1]) && bol
< eol
)
1127 eflags
|= REG_NOTBOL
;
1132 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1135 pmatch
[0].rm_so
+= bol
- start
;
1136 pmatch
[0].rm_eo
+= bol
- start
;
1141 static int match_expr_eval(struct grep_opt
*opt
, struct grep_expr
*x
, char *bol
,
1142 char *eol
, enum grep_context ctx
, ssize_t
*col
,
1143 ssize_t
*icol
, int collect_hits
)
1148 die("Not a valid grep expression");
1150 case GREP_NODE_TRUE
:
1153 case GREP_NODE_ATOM
:
1156 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
,
1158 if (h
&& (*col
< 0 || tmp
.rm_so
< *col
))
1164 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1166 h
= !match_expr_eval(opt
, x
->u
.unary
, bol
, eol
, ctx
, icol
, col
,
1170 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1172 if (h
|| opt
->columnnum
) {
1174 * Don't short-circuit AND when given --column, since a
1175 * NOT earlier in the tree may turn this into an OR. In
1176 * this case, see the below comment.
1178 h
&= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
,
1183 if (!(collect_hits
|| opt
->columnnum
)) {
1185 * Don't short-circuit OR when given --column (or
1186 * collecting hits) to ensure we don't skip a later
1187 * child that would produce an earlier match.
1189 return (match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
,
1190 ctx
, col
, icol
, 0) ||
1191 match_expr_eval(opt
, x
->u
.binary
.right
, bol
,
1192 eol
, ctx
, col
, icol
, 0));
1194 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1197 x
->u
.binary
.left
->hit
|= h
;
1198 h
|= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
, ctx
, col
,
1199 icol
, collect_hits
);
1202 die("Unexpected node type (internal error) %d", x
->node
);
1209 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1210 enum grep_context ctx
, ssize_t
*col
,
1211 ssize_t
*icol
, int collect_hits
)
1213 struct grep_expr
*x
= opt
->pattern_expression
;
1214 return match_expr_eval(opt
, x
, bol
, eol
, ctx
, col
, icol
, collect_hits
);
1217 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1218 ssize_t
*col
, ssize_t
*icol
,
1219 enum grep_context ctx
, int collect_hits
)
1225 return match_expr(opt
, bol
, eol
, ctx
, col
, icol
,
1228 /* we do not call with collect_hits without being extended */
1229 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1231 if (match_one_pattern(p
, bol
, eol
, ctx
, &tmp
, 0)) {
1233 if (!opt
->columnnum
) {
1235 * Without --column, any single match on a line
1236 * is enough to know that it needs to be
1237 * printed. With --column, scan _all_ patterns
1238 * to find the earliest.
1242 if (*col
< 0 || tmp
.rm_so
< *col
)
1249 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1250 enum grep_context ctx
,
1251 regmatch_t
*pmatch
, int eflags
)
1255 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1257 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1259 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1260 if (match
.rm_so
> pmatch
->rm_so
)
1262 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1265 pmatch
->rm_so
= match
.rm_so
;
1266 pmatch
->rm_eo
= match
.rm_eo
;
1270 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1271 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1276 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1278 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1280 case GREP_PATTERN
: /* atom */
1281 case GREP_PATTERN_HEAD
:
1282 case GREP_PATTERN_BODY
:
1283 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1294 static void show_line_header(struct grep_opt
*opt
, const char *name
,
1295 unsigned lno
, ssize_t cno
, char sign
)
1297 if (opt
->heading
&& opt
->last_shown
== 0) {
1298 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1299 opt
->output(opt
, "\n", 1);
1301 opt
->last_shown
= lno
;
1303 if (!opt
->heading
&& opt
->pathname
) {
1304 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1305 output_sep(opt
, sign
);
1309 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1310 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1311 output_sep(opt
, sign
);
1314 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1315 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1316 * being called with a context line.
1318 if (opt
->columnnum
&& cno
) {
1320 xsnprintf(buf
, sizeof(buf
), "%"PRIuMAX
, (uintmax_t)cno
);
1321 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_COLUMNNO
]);
1322 output_sep(opt
, sign
);
1326 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1327 const char *name
, unsigned lno
, ssize_t cno
, char sign
)
1329 int rest
= eol
- bol
;
1330 const char *match_color
= NULL
;
1331 const char *line_color
= NULL
;
1333 if (opt
->file_break
&& opt
->last_shown
== 0) {
1334 if (opt
->show_hunk_mark
)
1335 opt
->output(opt
, "\n", 1);
1336 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1337 if (opt
->last_shown
== 0) {
1338 if (opt
->show_hunk_mark
) {
1339 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1340 opt
->output(opt
, "\n", 1);
1342 } else if (lno
> opt
->last_shown
+ 1) {
1343 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1344 opt
->output(opt
, "\n", 1);
1347 if (!opt
->only_matching
) {
1349 * In case the line we're being called with contains more than
1350 * one match, leave printing each header to the loop below.
1352 show_line_header(opt
, name
, lno
, cno
, sign
);
1354 if (opt
->color
|| opt
->only_matching
) {
1356 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1362 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1364 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1366 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1367 else if (sign
== '-')
1368 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1369 else if (sign
== '=')
1370 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1373 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1374 if (match
.rm_so
== match
.rm_eo
)
1377 if (opt
->only_matching
)
1378 show_line_header(opt
, name
, lno
, cno
, sign
);
1380 output_color(opt
, bol
, match
.rm_so
, line_color
);
1381 output_color(opt
, bol
+ match
.rm_so
,
1382 match
.rm_eo
- match
.rm_so
, match_color
);
1383 if (opt
->only_matching
)
1384 opt
->output(opt
, "\n", 1);
1387 rest
-= match
.rm_eo
;
1388 eflags
= REG_NOTBOL
;
1392 if (!opt
->only_matching
) {
1393 output_color(opt
, bol
, rest
, line_color
);
1394 opt
->output(opt
, "\n", 1);
1401 * This lock protects access to the gitattributes machinery, which is
1404 pthread_mutex_t grep_attr_mutex
;
1406 static inline void grep_attr_lock(void)
1409 pthread_mutex_lock(&grep_attr_mutex
);
1412 static inline void grep_attr_unlock(void)
1415 pthread_mutex_unlock(&grep_attr_mutex
);
1418 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1420 xdemitconf_t
*xecfg
= opt
->priv
;
1421 if (xecfg
&& !xecfg
->find_func
) {
1422 grep_source_load_driver(gs
, opt
->repo
->index
);
1423 if (gs
->driver
->funcname
.pattern
) {
1424 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1425 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1427 xecfg
= opt
->priv
= NULL
;
1433 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1434 xecfg
->find_func_priv
) >= 0;
1439 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1444 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1445 char *bol
, unsigned lno
)
1447 while (bol
> gs
->buf
) {
1450 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1454 if (lno
<= opt
->last_shown
)
1457 if (match_funcname(opt
, gs
, bol
, eol
)) {
1458 show_line(opt
, bol
, eol
, gs
->name
, lno
, 0, '=');
1464 static int is_empty_line(const char *bol
, const char *eol
);
1466 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1467 char *bol
, char *end
, unsigned lno
)
1469 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1470 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1472 if (opt
->pre_context
< lno
)
1473 from
= lno
- opt
->pre_context
;
1474 if (from
<= opt
->last_shown
)
1475 from
= opt
->last_shown
+ 1;
1477 if (opt
->funcbody
) {
1478 if (match_funcname(opt
, gs
, bol
, end
))
1481 funcname_needed
= 1;
1482 from
= opt
->last_shown
+ 1;
1486 while (bol
> gs
->buf
&& cur
> from
) {
1487 char *next_bol
= bol
;
1490 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1493 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1494 match_funcname(opt
, gs
, bol
, eol
))) {
1503 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1505 funcname_needed
= 0;
1513 /* We need to look even further back to find a function signature. */
1514 if (opt
->funcname
&& funcname_needed
)
1515 show_funcname_line(opt
, gs
, bol
, cur
);
1519 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1521 while (*eol
!= '\n')
1523 show_line(opt
, bol
, eol
, gs
->name
, cur
, 0, sign
);
1529 static int should_lookahead(struct grep_opt
*opt
)
1534 return 0; /* punt for too complex stuff */
1537 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1538 if (p
->token
!= GREP_PATTERN
)
1539 return 0; /* punt for "header only" and stuff */
1544 static int look_ahead(struct grep_opt
*opt
,
1545 unsigned long *left_p
,
1549 unsigned lno
= *lno_p
;
1552 char *sp
, *last_bol
;
1553 regoff_t earliest
= -1;
1555 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1559 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1560 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1562 if (earliest
< 0 || m
.rm_so
< earliest
)
1567 *bol_p
= bol
+ *left_p
;
1571 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1572 ; /* find the beginning of the line */
1575 for (sp
= bol
; sp
< last_bol
; sp
++) {
1579 *left_p
-= last_bol
- bol
;
1585 static int fill_textconv_grep(struct repository
*r
,
1586 struct userdiff_driver
*driver
,
1587 struct grep_source
*gs
)
1589 struct diff_filespec
*df
;
1593 if (!driver
|| !driver
->textconv
)
1594 return grep_source_load(gs
);
1597 * The textconv interface is intimately tied to diff_filespecs, so we
1598 * have to pretend to be one. If we could unify the grep_source
1599 * and diff_filespec structs, this mess could just go away.
1601 df
= alloc_filespec(gs
->path
);
1603 case GREP_SOURCE_OID
:
1604 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1606 case GREP_SOURCE_FILE
:
1607 fill_filespec(df
, &null_oid
, 0, 0100644);
1610 BUG("attempt to textconv something without a path?");
1614 * fill_textconv is not remotely thread-safe; it modifies the global
1615 * diff tempfile structure, writes to the_repo's odb and might
1616 * internally call thread-unsafe functions such as the
1617 * prepare_packed_git() lazy-initializator. Because of the last two, we
1618 * must ensure mutual exclusion between this call and the object reading
1619 * API, thus we use obj_read_lock() here.
1621 * TODO: allowing text conversion to run in parallel with object
1622 * reading operations might increase performance in the multithreaded
1623 * non-worktreee git-grep with --textconv.
1626 size
= fill_textconv(r
, driver
, df
, &buf
);
1631 * The normal fill_textconv usage by the diff machinery would just keep
1632 * the textconv'd buf separate from the diff_filespec. But much of the
1633 * grep code passes around a grep_source and assumes that its "buf"
1634 * pointer is the beginning of the thing we are searching. So let's
1635 * install our textconv'd version into the grep_source, taking care not
1636 * to leak any existing buffer.
1638 grep_source_clear_data(gs
);
1645 static int is_empty_line(const char *bol
, const char *eol
)
1647 while (bol
< eol
&& isspace(*bol
))
1652 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1655 char *peek_bol
= NULL
;
1658 unsigned last_hit
= 0;
1659 int binary_match_only
= 0;
1661 int try_lookahead
= 0;
1662 int show_function
= 0;
1663 struct userdiff_driver
*textconv
= NULL
;
1664 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1667 if (!opt
->status_only
&& gs
->name
== NULL
)
1668 BUG("grep call which could print a name requires "
1669 "grep_source.name be non-NULL");
1672 opt
->output
= std_output
;
1674 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1676 /* Show hunk marks, except for the first file. */
1677 if (opt
->last_shown
)
1678 opt
->show_hunk_mark
= 1;
1680 * If we're using threads then we can't easily identify
1681 * the first file. Always put hunk marks in that case
1682 * and skip the very first one later in work_done().
1684 if (opt
->output
!= std_output
)
1685 opt
->show_hunk_mark
= 1;
1687 opt
->last_shown
= 0;
1689 if (opt
->allow_textconv
) {
1690 grep_source_load_driver(gs
, opt
->repo
->index
);
1692 * We might set up the shared textconv cache data here, which
1693 * is not thread-safe. Also, get_oid_with_context() and
1694 * parse_object() might be internally called. As they are not
1695 * currently thread-safe and might be racy with object reading,
1696 * obj_read_lock() must be called.
1700 textconv
= userdiff_get_textconv(opt
->repo
, gs
->driver
);
1706 * We know the result of a textconv is text, so we only have to care
1707 * about binary handling if we are not using it.
1710 switch (opt
->binary
) {
1711 case GREP_BINARY_DEFAULT
:
1712 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1713 binary_match_only
= 1;
1715 case GREP_BINARY_NOMATCH
:
1716 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1717 return 0; /* Assume unmatch */
1719 case GREP_BINARY_TEXT
:
1722 BUG("unknown binary handling mode");
1726 memset(&xecfg
, 0, sizeof(xecfg
));
1729 try_lookahead
= should_lookahead(opt
);
1731 if (fill_textconv_grep(opt
->repo
, textconv
, gs
) < 0)
1740 ssize_t col
= -1, icol
= -1;
1743 * look_ahead() skips quickly to the line that possibly
1744 * has the next hit; don't call it if we need to do
1745 * something more than just skipping the current line
1746 * in response to an unmatch for the current line. E.g.
1747 * inside a post-context window, we will show the current
1748 * line as a context around the previous hit when it
1753 && (show_function
||
1754 lno
<= last_hit
+ opt
->post_context
))
1755 && look_ahead(opt
, &left
, &lno
, &bol
))
1757 eol
= end_of_line(bol
, &left
);
1761 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1762 ctx
= GREP_CONTEXT_BODY
;
1764 hit
= match_line(opt
, bol
, eol
, &col
, &icol
, ctx
, collect_hits
);
1770 /* "grep -v -e foo -e bla" should list lines
1771 * that do not have either, so inversion should
1776 if (opt
->unmatch_name_only
) {
1783 if (opt
->status_only
)
1785 if (opt
->name_only
) {
1786 show_name(opt
, gs
->name
);
1791 if (binary_match_only
) {
1792 opt
->output(opt
, "Binary file ", 12);
1793 output_color(opt
, gs
->name
, strlen(gs
->name
),
1794 opt
->colors
[GREP_COLOR_FILENAME
]);
1795 opt
->output(opt
, " matches\n", 9);
1798 /* Hit at this line. If we haven't shown the
1799 * pre-context lines, we would need to show them.
1801 if (opt
->pre_context
|| opt
->funcbody
)
1802 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1803 else if (opt
->funcname
)
1804 show_funcname_line(opt
, gs
, bol
, lno
);
1805 cno
= opt
->invert
? icol
: col
;
1808 * A negative cno indicates that there was no
1809 * match on the line. We are thus inverted and
1810 * being asked to show all lines that _don't_
1811 * match a given expression. Therefore, set cno
1812 * to 0 to suggest the whole line matches.
1816 show_line(opt
, bol
, eol
, gs
->name
, lno
, cno
+ 1, ':');
1822 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1823 unsigned long peek_left
= left
;
1824 char *peek_eol
= eol
;
1827 * Trailing empty lines are not interesting.
1828 * Peek past them to see if they belong to the
1829 * body of the current function.
1832 while (is_empty_line(peek_bol
, peek_eol
)) {
1833 peek_bol
= peek_eol
+ 1;
1834 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1837 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1840 if (show_function
||
1841 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1842 /* If the last hit is within the post context,
1843 * we need to show this line.
1845 show_line(opt
, bol
, eol
, gs
->name
, lno
, col
+ 1, '-');
1859 if (opt
->status_only
)
1860 return opt
->unmatch_name_only
;
1861 if (opt
->unmatch_name_only
) {
1862 /* We did not see any hit, so we want to show this */
1863 show_name(opt
, gs
->name
);
1867 xdiff_clear_find_func(&xecfg
);
1871 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1872 * which feels mostly useless but sometimes useful. Maybe
1873 * make it another option? For now suppress them.
1875 if (opt
->count
&& count
) {
1877 if (opt
->pathname
) {
1878 output_color(opt
, gs
->name
, strlen(gs
->name
),
1879 opt
->colors
[GREP_COLOR_FILENAME
]);
1880 output_sep(opt
, ':');
1882 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1883 opt
->output(opt
, buf
, strlen(buf
));
1889 static void clr_hit_marker(struct grep_expr
*x
)
1891 /* All-hit markers are meaningful only at the very top level
1896 if (x
->node
!= GREP_NODE_OR
)
1898 x
->u
.binary
.left
->hit
= 0;
1899 x
= x
->u
.binary
.right
;
1903 static int chk_hit_marker(struct grep_expr
*x
)
1905 /* Top level nodes have hit markers. See if they all are hits */
1907 if (x
->node
!= GREP_NODE_OR
)
1909 if (!x
->u
.binary
.left
->hit
)
1911 x
= x
->u
.binary
.right
;
1915 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1918 * we do not have to do the two-pass grep when we do not check
1919 * buffer-wide "all-match".
1921 if (!opt
->all_match
)
1922 return grep_source_1(opt
, gs
, 0);
1924 /* Otherwise the toplevel "or" terms hit a bit differently.
1925 * We first clear hit markers from them.
1927 clr_hit_marker(opt
->pattern_expression
);
1928 grep_source_1(opt
, gs
, 1);
1930 if (!chk_hit_marker(opt
->pattern_expression
))
1933 return grep_source_1(opt
, gs
, 0);
1936 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
1938 struct grep_source gs
;
1941 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
1945 r
= grep_source(opt
, &gs
);
1947 grep_source_clear(&gs
);
1951 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
1952 const char *name
, const char *path
,
1953 const void *identifier
)
1956 gs
->name
= xstrdup_or_null(name
);
1957 gs
->path
= xstrdup_or_null(path
);
1963 case GREP_SOURCE_FILE
:
1964 gs
->identifier
= xstrdup(identifier
);
1966 case GREP_SOURCE_OID
:
1967 gs
->identifier
= oiddup(identifier
);
1969 case GREP_SOURCE_BUF
:
1970 gs
->identifier
= NULL
;
1975 void grep_source_clear(struct grep_source
*gs
)
1977 FREE_AND_NULL(gs
->name
);
1978 FREE_AND_NULL(gs
->path
);
1979 FREE_AND_NULL(gs
->identifier
);
1980 grep_source_clear_data(gs
);
1983 void grep_source_clear_data(struct grep_source
*gs
)
1986 case GREP_SOURCE_FILE
:
1987 case GREP_SOURCE_OID
:
1988 FREE_AND_NULL(gs
->buf
);
1991 case GREP_SOURCE_BUF
:
1992 /* leave user-provided buf intact */
1997 static int grep_source_load_oid(struct grep_source
*gs
)
1999 enum object_type type
;
2001 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2003 return error(_("'%s': unable to read %s"),
2005 oid_to_hex(gs
->identifier
));
2009 static int grep_source_load_file(struct grep_source
*gs
)
2011 const char *filename
= gs
->identifier
;
2017 if (lstat(filename
, &st
) < 0) {
2019 if (errno
!= ENOENT
)
2020 error_errno(_("failed to stat '%s'"), filename
);
2023 if (!S_ISREG(st
.st_mode
))
2025 size
= xsize_t(st
.st_size
);
2026 i
= open(filename
, O_RDONLY
);
2029 data
= xmallocz(size
);
2030 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2031 error_errno(_("'%s': short read"), filename
);
2043 static int grep_source_load(struct grep_source
*gs
)
2049 case GREP_SOURCE_FILE
:
2050 return grep_source_load_file(gs
);
2051 case GREP_SOURCE_OID
:
2052 return grep_source_load_oid(gs
);
2053 case GREP_SOURCE_BUF
:
2054 return gs
->buf
? 0 : -1;
2056 BUG("invalid grep_source type to load");
2059 void grep_source_load_driver(struct grep_source
*gs
,
2060 struct index_state
*istate
)
2067 gs
->driver
= userdiff_find_by_path(istate
, gs
->path
);
2069 gs
->driver
= userdiff_find_by_name("default");
2073 static int grep_source_is_binary(struct grep_source
*gs
,
2074 struct index_state
*istate
)
2076 grep_source_load_driver(gs
, istate
);
2077 if (gs
->driver
->binary
!= -1)
2078 return gs
->driver
->binary
;
2080 if (!grep_source_load(gs
))
2081 return buffer_is_binary(gs
->buf
, gs
->size
);