5 #include "xdiff-interface.h"
11 static int grep_source_load(struct grep_source
*gs
);
12 static int grep_source_is_binary(struct grep_source
*gs
);
14 static struct grep_opt grep_defaults
;
16 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
18 fwrite(buf
, size
, 1, stdout
);
22 * Initialize the grep_defaults template with hardcoded defaults.
23 * We could let the compiler do this, but without C99 initializers
24 * the code gets unwieldy and unreadable, so...
26 void init_grep_defaults(void)
28 struct grep_opt
*opt
= &grep_defaults
;
35 memset(opt
, 0, sizeof(*opt
));
39 opt
->pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
;
40 color_set(opt
->color_context
, "");
41 color_set(opt
->color_filename
, "");
42 color_set(opt
->color_function
, "");
43 color_set(opt
->color_lineno
, "");
44 color_set(opt
->color_match_context
, GIT_COLOR_BOLD_RED
);
45 color_set(opt
->color_match_selected
, GIT_COLOR_BOLD_RED
);
46 color_set(opt
->color_selected
, "");
47 color_set(opt
->color_sep
, GIT_COLOR_CYAN
);
49 opt
->output
= std_output
;
52 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
54 if (!strcmp(arg
, "default"))
55 return GREP_PATTERN_TYPE_UNSPECIFIED
;
56 else if (!strcmp(arg
, "basic"))
57 return GREP_PATTERN_TYPE_BRE
;
58 else if (!strcmp(arg
, "extended"))
59 return GREP_PATTERN_TYPE_ERE
;
60 else if (!strcmp(arg
, "fixed"))
61 return GREP_PATTERN_TYPE_FIXED
;
62 else if (!strcmp(arg
, "perl"))
63 return GREP_PATTERN_TYPE_PCRE
;
64 die("bad %s argument: %s", opt
, arg
);
68 * Read the configuration file once and store it in
69 * the grep_defaults template.
71 int grep_config(const char *var
, const char *value
, void *cb
)
73 struct grep_opt
*opt
= &grep_defaults
;
76 if (userdiff_config(var
, value
) < 0)
79 if (!strcmp(var
, "grep.extendedregexp")) {
80 opt
->extended_regexp_option
= git_config_bool(var
, value
);
84 if (!strcmp(var
, "grep.patterntype")) {
85 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
89 if (!strcmp(var
, "grep.linenumber")) {
90 opt
->linenum
= git_config_bool(var
, value
);
94 if (!strcmp(var
, "grep.fullname")) {
95 opt
->relative
= !git_config_bool(var
, value
);
99 if (!strcmp(var
, "color.grep"))
100 opt
->color
= git_config_colorbool(var
, value
);
101 else if (!strcmp(var
, "color.grep.context"))
102 color
= opt
->color_context
;
103 else if (!strcmp(var
, "color.grep.filename"))
104 color
= opt
->color_filename
;
105 else if (!strcmp(var
, "color.grep.function"))
106 color
= opt
->color_function
;
107 else if (!strcmp(var
, "color.grep.linenumber"))
108 color
= opt
->color_lineno
;
109 else if (!strcmp(var
, "color.grep.matchcontext"))
110 color
= opt
->color_match_context
;
111 else if (!strcmp(var
, "color.grep.matchselected"))
112 color
= opt
->color_match_selected
;
113 else if (!strcmp(var
, "color.grep.selected"))
114 color
= opt
->color_selected
;
115 else if (!strcmp(var
, "color.grep.separator"))
116 color
= opt
->color_sep
;
117 else if (!strcmp(var
, "color.grep.match")) {
120 return config_error_nonbool(var
);
121 rc
|= color_parse(value
, opt
->color_match_context
);
122 rc
|= color_parse(value
, opt
->color_match_selected
);
128 return config_error_nonbool(var
);
129 return color_parse(value
, color
);
135 * Initialize one instance of grep_opt and copy the
136 * default values from the template we read the configuration
137 * information in an earlier call to git_config(grep_config).
139 void grep_init(struct grep_opt
*opt
, const char *prefix
)
141 struct grep_opt
*def
= &grep_defaults
;
143 memset(opt
, 0, sizeof(*opt
));
144 opt
->prefix
= prefix
;
145 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
146 opt
->pattern_tail
= &opt
->pattern_list
;
147 opt
->header_tail
= &opt
->header_list
;
149 opt
->color
= def
->color
;
150 opt
->extended_regexp_option
= def
->extended_regexp_option
;
151 opt
->pattern_type_option
= def
->pattern_type_option
;
152 opt
->linenum
= def
->linenum
;
153 opt
->max_depth
= def
->max_depth
;
154 opt
->pathname
= def
->pathname
;
155 opt
->relative
= def
->relative
;
156 opt
->output
= def
->output
;
158 color_set(opt
->color_context
, def
->color_context
);
159 color_set(opt
->color_filename
, def
->color_filename
);
160 color_set(opt
->color_function
, def
->color_function
);
161 color_set(opt
->color_lineno
, def
->color_lineno
);
162 color_set(opt
->color_match_context
, def
->color_match_context
);
163 color_set(opt
->color_match_selected
, def
->color_match_selected
);
164 color_set(opt
->color_selected
, def
->color_selected
);
165 color_set(opt
->color_sep
, def
->color_sep
);
168 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
171 * When committing to the pattern type by setting the relevant
172 * fields in grep_opt it's generally not necessary to zero out
173 * the fields we're not choosing, since they won't have been
174 * set by anything. The extended_regexp_option field is the
175 * only exception to this.
177 * This is because in the process of parsing grep.patternType
178 * & grep.extendedRegexp we set opt->pattern_type_option and
179 * opt->extended_regexp_option, respectively. We then
180 * internally use opt->extended_regexp_option to see if we're
181 * compiling an ERE. It must be unset if that's not actually
184 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
185 opt
->extended_regexp_option
)
186 opt
->extended_regexp_option
= 0;
188 switch (pattern_type
) {
189 case GREP_PATTERN_TYPE_UNSPECIFIED
:
192 case GREP_PATTERN_TYPE_BRE
:
195 case GREP_PATTERN_TYPE_ERE
:
196 opt
->extended_regexp_option
= 1;
199 case GREP_PATTERN_TYPE_FIXED
:
203 case GREP_PATTERN_TYPE_PCRE
:
208 * It's important that pcre1 always be assigned to
209 * even when there's no USE_LIBPCRE* defined. We still
210 * call the PCRE stub function, it just dies with
211 * "cannot use Perl-compatible regexes[...]".
219 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
221 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
222 grep_set_pattern_type_option(pattern_type
, opt
);
223 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
224 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
225 else if (opt
->extended_regexp_option
)
227 * This branch *must* happen after setting from the
228 * opt->pattern_type_option above, we don't want
229 * grep.extendedRegexp to override grep.patternType!
231 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
234 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
235 const char *origin
, int no
,
236 enum grep_pat_token t
,
237 enum grep_header_field field
)
239 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
240 p
->pattern
= xmemdupz(pat
, patlen
);
241 p
->patternlen
= patlen
;
249 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
256 case GREP_PATTERN
: /* atom */
257 case GREP_PATTERN_HEAD
:
258 case GREP_PATTERN_BODY
:
260 struct grep_pat
*new_pat
;
262 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
263 while (++len
<= p
->patternlen
) {
264 if (*(--cp
) == '\n') {
271 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
272 p
->no
, p
->token
, p
->field
);
273 new_pat
->next
= p
->next
;
275 *tail
= &new_pat
->next
;
278 p
->patternlen
-= len
;
286 void append_header_grep_pattern(struct grep_opt
*opt
,
287 enum grep_header_field field
, const char *pat
)
289 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
290 GREP_PATTERN_HEAD
, field
);
291 if (field
== GREP_HEADER_REFLOG
)
292 opt
->use_reflog_filter
= 1;
293 do_append_grep_pat(&opt
->header_tail
, p
);
296 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
297 const char *origin
, int no
, enum grep_pat_token t
)
299 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
302 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
303 const char *origin
, int no
, enum grep_pat_token t
)
305 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
306 do_append_grep_pat(&opt
->pattern_tail
, p
);
309 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
311 struct grep_pat
*pat
;
312 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
315 ret
->pattern_list
= NULL
;
316 ret
->pattern_tail
= &ret
->pattern_list
;
318 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
320 if(pat
->token
== GREP_PATTERN_HEAD
)
321 append_header_grep_pattern(ret
, pat
->field
,
324 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
325 pat
->origin
, pat
->no
, pat
->token
);
331 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
337 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
339 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
343 die("%s'%s': %s", where
, p
->pattern
, error
);
346 static int is_fixed(const char *s
, size_t len
)
350 for (i
= 0; i
< len
; i
++) {
351 if (is_regex_special(s
[i
]))
358 static int has_null(const char *s
, size_t len
)
361 * regcomp cannot accept patterns with NULs so when using it
362 * we consider any pattern containing a NUL fixed.
364 if (memchr(s
, 0, len
))
371 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
375 int options
= PCRE_MULTILINE
;
377 if (opt
->ignore_case
) {
378 if (has_non_ascii(p
->pattern
))
379 p
->pcre1_tables
= pcre_maketables();
380 options
|= PCRE_CASELESS
;
382 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
383 options
|= PCRE_UTF8
;
385 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
387 if (!p
->pcre1_regexp
)
388 compile_regexp_failed(p
, error
);
390 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, PCRE_STUDY_JIT_COMPILE
, &error
);
391 if (!p
->pcre1_extra_info
&& error
)
394 #ifdef GIT_PCRE1_USE_JIT
395 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
396 if (p
->pcre1_jit_on
== 1) {
397 p
->pcre1_jit_stack
= pcre_jit_stack_alloc(1, 1024 * 1024);
398 if (!p
->pcre1_jit_stack
)
399 die("Couldn't allocate PCRE JIT stack");
400 pcre_assign_jit_stack(p
->pcre1_extra_info
, NULL
, p
->pcre1_jit_stack
);
401 } else if (p
->pcre1_jit_on
!= 0) {
402 die("BUG: The pcre1_jit_on variable should be 0 or 1, not %d",
408 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
409 regmatch_t
*match
, int eflags
)
411 int ovector
[30], ret
, flags
= 0;
413 if (eflags
& REG_NOTBOL
)
414 flags
|= PCRE_NOTBOL
;
416 #ifdef GIT_PCRE1_USE_JIT
417 if (p
->pcre1_jit_on
) {
418 ret
= pcre_jit_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
419 eol
- line
, 0, flags
, ovector
,
420 ARRAY_SIZE(ovector
), p
->pcre1_jit_stack
);
424 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
425 eol
- line
, 0, flags
, ovector
,
426 ARRAY_SIZE(ovector
));
429 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
430 die("pcre_exec failed with error code %d", ret
);
433 match
->rm_so
= ovector
[0];
434 match
->rm_eo
= ovector
[1];
440 static void free_pcre1_regexp(struct grep_pat
*p
)
442 pcre_free(p
->pcre1_regexp
);
443 #ifdef GIT_PCRE1_USE_JIT
444 if (p
->pcre1_jit_on
) {
445 pcre_free_study(p
->pcre1_extra_info
);
446 pcre_jit_stack_free(p
->pcre1_jit_stack
);
450 pcre_free(p
->pcre1_extra_info
);
452 pcre_free((void *)p
->pcre1_tables
);
454 #else /* !USE_LIBPCRE1 */
455 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
457 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
460 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
461 regmatch_t
*match
, int eflags
)
466 static void free_pcre1_regexp(struct grep_pat
*p
)
469 #endif /* !USE_LIBPCRE1 */
472 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
475 PCRE2_UCHAR errbuf
[256];
476 PCRE2_SIZE erroffset
;
477 int options
= PCRE2_MULTILINE
;
478 const uint8_t *character_tables
= NULL
;
483 p
->pcre2_compile_context
= NULL
;
485 if (opt
->ignore_case
) {
486 if (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 (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
== 1) {
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
);
514 p
->pcre2_jit_stack
= pcre2_jit_stack_create(1, 1024 * 1024, NULL
);
515 if (!p
->pcre2_jit_stack
)
516 die("Couldn't allocate PCRE2 JIT stack");
517 p
->pcre2_match_context
= pcre2_match_context_create(NULL
);
518 if (!p
->pcre2_match_context
)
519 die("Couldn't allocate PCRE2 match context");
520 pcre2_jit_stack_assign(p
->pcre2_match_context
, NULL
, p
->pcre2_jit_stack
);
521 } else if (p
->pcre2_jit_on
!= 0) {
522 die("BUG: The pcre2_jit_on variable should be 0 or 1, not %d",
527 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
528 regmatch_t
*match
, int eflags
)
532 PCRE2_UCHAR errbuf
[256];
534 if (eflags
& REG_NOTBOL
)
535 flags
|= PCRE2_NOTBOL
;
538 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
539 eol
- line
, 0, flags
, p
->pcre2_match_data
,
542 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
543 eol
- line
, 0, flags
, p
->pcre2_match_data
,
546 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
547 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
548 die("%s failed with error code %d: %s",
549 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
553 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
555 match
->rm_so
= (int)ovector
[0];
556 match
->rm_eo
= (int)ovector
[1];
562 static void free_pcre2_pattern(struct grep_pat
*p
)
564 pcre2_compile_context_free(p
->pcre2_compile_context
);
565 pcre2_code_free(p
->pcre2_pattern
);
566 pcre2_match_data_free(p
->pcre2_match_data
);
567 pcre2_jit_stack_free(p
->pcre2_jit_stack
);
568 pcre2_match_context_free(p
->pcre2_match_context
);
570 #else /* !USE_LIBPCRE2 */
571 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
574 * Unreachable until USE_LIBPCRE2 becomes synonymous with
575 * USE_LIBPCRE. See the sibling comment in
576 * grep_set_pattern_type_option().
578 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
581 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
582 regmatch_t
*match
, int eflags
)
587 static void free_pcre2_pattern(struct grep_pat
*p
)
590 #endif /* !USE_LIBPCRE2 */
592 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
594 struct strbuf sb
= STRBUF_INIT
;
598 basic_regex_quote_buf(&sb
, p
->pattern
);
599 if (opt
->ignore_case
)
600 regflags
|= REG_ICASE
;
601 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
603 fprintf(stderr
, "fixed %s\n", sb
.buf
);
607 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
609 compile_regexp_failed(p
, errbuf
);
613 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
617 int regflags
= REG_NEWLINE
;
619 p
->word_regexp
= opt
->word_regexp
;
620 p
->ignore_case
= opt
->ignore_case
;
621 ascii_only
= !has_non_ascii(p
->pattern
);
624 * Even when -F (fixed) asks us to do a non-regexp search, we
625 * may not be able to correctly case-fold when -i
626 * (ignore-case) is asked (in which case, we'll synthesize a
627 * regexp to match the pattern that matches regexp special
628 * characters literally, while ignoring case differences). On
629 * the other hand, even without -F, if the pattern does not
630 * have any regexp special characters and there is no need for
631 * case-folding search, we can internally turn it into a
632 * simple string match using kws. p->fixed tells us if we
636 has_null(p
->pattern
, p
->patternlen
) ||
637 is_fixed(p
->pattern
, p
->patternlen
))
638 p
->fixed
= !p
->ignore_case
|| ascii_only
;
641 p
->kws
= kwsalloc(p
->ignore_case
? tolower_trans_tbl
: NULL
);
642 kwsincr(p
->kws
, p
->pattern
, p
->patternlen
);
645 } else if (opt
->fixed
) {
647 * We come here when the pattern has the non-ascii
648 * characters we cannot case-fold, and asked to
651 compile_fixed_regexp(p
, opt
);
656 compile_pcre2_pattern(p
, opt
);
661 compile_pcre1_regexp(p
, opt
);
666 regflags
|= REG_ICASE
;
667 if (opt
->extended_regexp_option
)
668 regflags
|= REG_EXTENDED
;
669 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
672 regerror(err
, &p
->regexp
, errbuf
, 1024);
674 compile_regexp_failed(p
, errbuf
);
678 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
679 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
688 case GREP_PATTERN
: /* atom */
689 case GREP_PATTERN_HEAD
:
690 case GREP_PATTERN_BODY
:
691 x
= xcalloc(1, sizeof (struct grep_expr
));
692 x
->node
= GREP_NODE_ATOM
;
696 case GREP_OPEN_PAREN
:
698 x
= compile_pattern_or(list
);
699 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
700 die("unmatched parenthesis");
701 *list
= (*list
)->next
;
708 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
719 die("--not not followed by pattern expression");
721 x
= xcalloc(1, sizeof (struct grep_expr
));
722 x
->node
= GREP_NODE_NOT
;
723 x
->u
.unary
= compile_pattern_not(list
);
725 die("--not followed by non pattern expression");
728 return compile_pattern_atom(list
);
732 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
735 struct grep_expr
*x
, *y
, *z
;
737 x
= compile_pattern_not(list
);
739 if (p
&& p
->token
== GREP_AND
) {
741 die("--and not followed by pattern expression");
743 y
= compile_pattern_and(list
);
745 die("--and not followed by pattern expression");
746 z
= xcalloc(1, sizeof (struct grep_expr
));
747 z
->node
= GREP_NODE_AND
;
748 z
->u
.binary
.left
= x
;
749 z
->u
.binary
.right
= y
;
755 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
758 struct grep_expr
*x
, *y
, *z
;
760 x
= compile_pattern_and(list
);
762 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
763 y
= compile_pattern_or(list
);
765 die("not a pattern expression %s", p
->pattern
);
766 z
= xcalloc(1, sizeof (struct grep_expr
));
767 z
->node
= GREP_NODE_OR
;
768 z
->u
.binary
.left
= x
;
769 z
->u
.binary
.right
= y
;
775 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
777 return compile_pattern_or(list
);
780 static void indent(int in
)
786 static void dump_grep_pat(struct grep_pat
*p
)
789 case GREP_AND
: fprintf(stderr
, "*and*"); break;
790 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
791 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
792 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
793 case GREP_OR
: fprintf(stderr
, "*or*"); break;
795 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
796 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
797 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
802 case GREP_PATTERN_HEAD
:
803 fprintf(stderr
, "<head %d>", p
->field
); break;
804 case GREP_PATTERN_BODY
:
805 fprintf(stderr
, "<body>"); break;
809 case GREP_PATTERN_HEAD
:
810 case GREP_PATTERN_BODY
:
812 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
818 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
823 fprintf(stderr
, "true\n");
826 dump_grep_pat(x
->u
.atom
);
829 fprintf(stderr
, "(not\n");
830 dump_grep_expression_1(x
->u
.unary
, in
+1);
832 fprintf(stderr
, ")\n");
835 fprintf(stderr
, "(and\n");
836 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
837 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
839 fprintf(stderr
, ")\n");
842 fprintf(stderr
, "(or\n");
843 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
844 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
846 fprintf(stderr
, ")\n");
851 static void dump_grep_expression(struct grep_opt
*opt
)
853 struct grep_expr
*x
= opt
->pattern_expression
;
856 fprintf(stderr
, "[all-match]\n");
857 dump_grep_expression_1(x
, 0);
861 static struct grep_expr
*grep_true_expr(void)
863 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
864 z
->node
= GREP_NODE_TRUE
;
868 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
870 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
871 z
->node
= GREP_NODE_OR
;
872 z
->u
.binary
.left
= left
;
873 z
->u
.binary
.right
= right
;
877 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
880 struct grep_expr
*header_expr
;
881 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
882 enum grep_header_field fld
;
884 if (!opt
->header_list
)
887 for (p
= opt
->header_list
; p
; p
= p
->next
) {
888 if (p
->token
!= GREP_PATTERN_HEAD
)
889 die("BUG: a non-header pattern in grep header list.");
890 if (p
->field
< GREP_HEADER_FIELD_MIN
||
891 GREP_HEADER_FIELD_MAX
<= p
->field
)
892 die("BUG: unknown header field %d", p
->field
);
893 compile_regexp(p
, opt
);
896 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
897 header_group
[fld
] = NULL
;
899 for (p
= opt
->header_list
; p
; p
= p
->next
) {
901 struct grep_pat
*pp
= p
;
903 h
= compile_pattern_atom(&pp
);
904 if (!h
|| pp
!= p
->next
)
905 die("BUG: malformed header expr");
906 if (!header_group
[p
->field
]) {
907 header_group
[p
->field
] = h
;
910 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
915 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
916 if (!header_group
[fld
])
919 header_expr
= grep_true_expr();
920 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
925 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
927 struct grep_expr
*z
= x
;
930 assert(x
->node
== GREP_NODE_OR
);
931 if (x
->u
.binary
.right
&&
932 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
933 x
->u
.binary
.right
= y
;
936 x
= x
->u
.binary
.right
;
941 static void compile_grep_patterns_real(struct grep_opt
*opt
)
944 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
946 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
948 case GREP_PATTERN
: /* atom */
949 case GREP_PATTERN_HEAD
:
950 case GREP_PATTERN_BODY
:
951 compile_regexp(p
, opt
);
959 if (opt
->all_match
|| header_expr
)
961 else if (!opt
->extended
&& !opt
->debug
)
964 p
= opt
->pattern_list
;
966 opt
->pattern_expression
= compile_pattern_expr(&p
);
968 die("incomplete pattern expression: %s", p
->pattern
);
973 if (!opt
->pattern_expression
)
974 opt
->pattern_expression
= header_expr
;
975 else if (opt
->all_match
)
976 opt
->pattern_expression
= grep_splice_or(header_expr
,
977 opt
->pattern_expression
);
979 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
984 void compile_grep_patterns(struct grep_opt
*opt
)
986 compile_grep_patterns_real(opt
);
988 dump_grep_expression(opt
);
991 static void free_pattern_expr(struct grep_expr
*x
)
998 free_pattern_expr(x
->u
.unary
);
1002 free_pattern_expr(x
->u
.binary
.left
);
1003 free_pattern_expr(x
->u
.binary
.right
);
1009 void free_grep_patterns(struct grep_opt
*opt
)
1011 struct grep_pat
*p
, *n
;
1013 for (p
= opt
->pattern_list
; p
; p
= n
) {
1016 case GREP_PATTERN
: /* atom */
1017 case GREP_PATTERN_HEAD
:
1018 case GREP_PATTERN_BODY
:
1021 else if (p
->pcre1_regexp
)
1022 free_pcre1_regexp(p
);
1023 else if (p
->pcre2_pattern
)
1024 free_pcre2_pattern(p
);
1026 regfree(&p
->regexp
);
1037 free_pattern_expr(opt
->pattern_expression
);
1040 static char *end_of_line(char *cp
, unsigned long *left
)
1042 unsigned long l
= *left
;
1043 while (l
&& *cp
!= '\n') {
1051 static int word_char(char ch
)
1053 return isalnum(ch
) || ch
== '_';
1056 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1059 if (want_color(opt
->color
) && color
&& color
[0]) {
1060 opt
->output(opt
, color
, strlen(color
));
1061 opt
->output(opt
, data
, size
);
1062 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1064 opt
->output(opt
, data
, size
);
1067 static void output_sep(struct grep_opt
*opt
, char sign
)
1069 if (opt
->null_following_name
)
1070 opt
->output(opt
, "\0", 1);
1072 output_color(opt
, &sign
, 1, opt
->color_sep
);
1075 static void show_name(struct grep_opt
*opt
, const char *name
)
1077 output_color(opt
, name
, strlen(name
), opt
->color_filename
);
1078 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1081 static int fixmatch(struct grep_pat
*p
, char *line
, char *eol
,
1084 struct kwsmatch kwsm
;
1085 size_t offset
= kwsexec(p
->kws
, line
, eol
- line
, &kwsm
);
1087 match
->rm_so
= match
->rm_eo
= -1;
1090 match
->rm_so
= offset
;
1091 match
->rm_eo
= match
->rm_so
+ kwsm
.size
[0];
1096 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1097 regmatch_t
*match
, int eflags
)
1102 hit
= !fixmatch(p
, line
, eol
, match
);
1103 else 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_expr
*x
, char *bol
, char *eol
,
1223 enum grep_context ctx
, int collect_hits
)
1229 die("Not a valid grep expression");
1231 case GREP_NODE_TRUE
:
1234 case GREP_NODE_ATOM
:
1235 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
, &match
, 0);
1238 h
= !match_expr_eval(x
->u
.unary
, bol
, eol
, ctx
, 0);
1241 if (!match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0))
1243 h
= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 0);
1247 return (match_expr_eval(x
->u
.binary
.left
,
1248 bol
, eol
, ctx
, 0) ||
1249 match_expr_eval(x
->u
.binary
.right
,
1251 h
= match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0);
1252 x
->u
.binary
.left
->hit
|= h
;
1253 h
|= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 1);
1256 die("Unexpected node type (internal error) %d", x
->node
);
1263 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1264 enum grep_context ctx
, int collect_hits
)
1266 struct grep_expr
*x
= opt
->pattern_expression
;
1267 return match_expr_eval(x
, bol
, eol
, ctx
, collect_hits
);
1270 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1271 enum grep_context ctx
, int collect_hits
)
1277 return match_expr(opt
, bol
, eol
, ctx
, collect_hits
);
1279 /* we do not call with collect_hits without being extended */
1280 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1281 if (match_one_pattern(p
, bol
, eol
, ctx
, &match
, 0))
1287 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1288 enum grep_context ctx
,
1289 regmatch_t
*pmatch
, int eflags
)
1293 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1295 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1297 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1298 if (match
.rm_so
> pmatch
->rm_so
)
1300 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1303 pmatch
->rm_so
= match
.rm_so
;
1304 pmatch
->rm_eo
= match
.rm_eo
;
1308 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1309 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1314 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1316 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1318 case GREP_PATTERN
: /* atom */
1319 case GREP_PATTERN_HEAD
:
1320 case GREP_PATTERN_BODY
:
1321 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1332 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1333 const char *name
, unsigned lno
, char sign
)
1335 int rest
= eol
- bol
;
1336 const char *match_color
, *line_color
= NULL
;
1338 if (opt
->file_break
&& opt
->last_shown
== 0) {
1339 if (opt
->show_hunk_mark
)
1340 opt
->output(opt
, "\n", 1);
1341 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1342 if (opt
->last_shown
== 0) {
1343 if (opt
->show_hunk_mark
) {
1344 output_color(opt
, "--", 2, opt
->color_sep
);
1345 opt
->output(opt
, "\n", 1);
1347 } else if (lno
> opt
->last_shown
+ 1) {
1348 output_color(opt
, "--", 2, opt
->color_sep
);
1349 opt
->output(opt
, "\n", 1);
1352 if (opt
->heading
&& opt
->last_shown
== 0) {
1353 output_color(opt
, name
, strlen(name
), opt
->color_filename
);
1354 opt
->output(opt
, "\n", 1);
1356 opt
->last_shown
= lno
;
1358 if (!opt
->heading
&& opt
->pathname
) {
1359 output_color(opt
, name
, strlen(name
), opt
->color_filename
);
1360 output_sep(opt
, sign
);
1364 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1365 output_color(opt
, buf
, strlen(buf
), opt
->color_lineno
);
1366 output_sep(opt
, sign
);
1370 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1375 match_color
= opt
->color_match_selected
;
1377 match_color
= opt
->color_match_context
;
1379 line_color
= opt
->color_selected
;
1380 else if (sign
== '-')
1381 line_color
= opt
->color_context
;
1382 else if (sign
== '=')
1383 line_color
= opt
->color_function
;
1385 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1386 if (match
.rm_so
== match
.rm_eo
)
1389 output_color(opt
, bol
, match
.rm_so
, line_color
);
1390 output_color(opt
, bol
+ match
.rm_so
,
1391 match
.rm_eo
- match
.rm_so
, match_color
);
1393 rest
-= match
.rm_eo
;
1394 eflags
= REG_NOTBOL
;
1398 output_color(opt
, bol
, rest
, line_color
);
1399 opt
->output(opt
, "\n", 1);
1406 * This lock protects access to the gitattributes machinery, which is
1409 pthread_mutex_t grep_attr_mutex
;
1411 static inline void grep_attr_lock(void)
1414 pthread_mutex_lock(&grep_attr_mutex
);
1417 static inline void grep_attr_unlock(void)
1420 pthread_mutex_unlock(&grep_attr_mutex
);
1424 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1426 pthread_mutex_t grep_read_mutex
;
1429 #define grep_attr_lock()
1430 #define grep_attr_unlock()
1433 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1435 xdemitconf_t
*xecfg
= opt
->priv
;
1436 if (xecfg
&& !xecfg
->find_func
) {
1437 grep_source_load_driver(gs
);
1438 if (gs
->driver
->funcname
.pattern
) {
1439 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1440 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1442 xecfg
= opt
->priv
= NULL
;
1448 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1449 xecfg
->find_func_priv
) >= 0;
1454 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1459 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1460 char *bol
, unsigned lno
)
1462 while (bol
> gs
->buf
) {
1465 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1469 if (lno
<= opt
->last_shown
)
1472 if (match_funcname(opt
, gs
, bol
, eol
)) {
1473 show_line(opt
, bol
, eol
, gs
->name
, lno
, '=');
1479 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1480 char *bol
, char *end
, unsigned lno
)
1482 unsigned cur
= lno
, from
= 1, funcname_lno
= 0;
1483 int funcname_needed
= !!opt
->funcname
;
1485 if (opt
->funcbody
&& !match_funcname(opt
, gs
, bol
, end
))
1486 funcname_needed
= 2;
1488 if (opt
->pre_context
< lno
)
1489 from
= lno
- opt
->pre_context
;
1490 if (from
<= opt
->last_shown
)
1491 from
= opt
->last_shown
+ 1;
1494 while (bol
> gs
->buf
&&
1495 cur
> (funcname_needed
== 2 ? opt
->last_shown
+ 1 : from
)) {
1498 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1501 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1503 funcname_needed
= 0;
1507 /* We need to look even further back to find a function signature. */
1508 if (opt
->funcname
&& funcname_needed
)
1509 show_funcname_line(opt
, gs
, bol
, cur
);
1513 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1515 while (*eol
!= '\n')
1517 show_line(opt
, bol
, eol
, gs
->name
, cur
, sign
);
1523 static int should_lookahead(struct grep_opt
*opt
)
1528 return 0; /* punt for too complex stuff */
1531 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1532 if (p
->token
!= GREP_PATTERN
)
1533 return 0; /* punt for "header only" and stuff */
1538 static int look_ahead(struct grep_opt
*opt
,
1539 unsigned long *left_p
,
1543 unsigned lno
= *lno_p
;
1546 char *sp
, *last_bol
;
1547 regoff_t earliest
= -1;
1549 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1553 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1554 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1556 if (earliest
< 0 || m
.rm_so
< earliest
)
1561 *bol_p
= bol
+ *left_p
;
1565 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1566 ; /* find the beginning of the line */
1569 for (sp
= bol
; sp
< last_bol
; sp
++) {
1573 *left_p
-= last_bol
- bol
;
1579 static int fill_textconv_grep(struct userdiff_driver
*driver
,
1580 struct grep_source
*gs
)
1582 struct diff_filespec
*df
;
1586 if (!driver
|| !driver
->textconv
)
1587 return grep_source_load(gs
);
1590 * The textconv interface is intimately tied to diff_filespecs, so we
1591 * have to pretend to be one. If we could unify the grep_source
1592 * and diff_filespec structs, this mess could just go away.
1594 df
= alloc_filespec(gs
->path
);
1596 case GREP_SOURCE_OID
:
1597 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1599 case GREP_SOURCE_FILE
:
1600 fill_filespec(df
, &null_oid
, 0, 0100644);
1603 die("BUG: attempt to textconv something without a path?");
1607 * fill_textconv is not remotely thread-safe; it may load objects
1608 * behind the scenes, and it modifies the global diff tempfile
1612 size
= fill_textconv(driver
, df
, &buf
);
1617 * The normal fill_textconv usage by the diff machinery would just keep
1618 * the textconv'd buf separate from the diff_filespec. But much of the
1619 * grep code passes around a grep_source and assumes that its "buf"
1620 * pointer is the beginning of the thing we are searching. So let's
1621 * install our textconv'd version into the grep_source, taking care not
1622 * to leak any existing buffer.
1624 grep_source_clear_data(gs
);
1631 static int is_empty_line(const char *bol
, const char *eol
)
1633 while (bol
< eol
&& isspace(*bol
))
1638 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1641 char *peek_bol
= NULL
;
1644 unsigned last_hit
= 0;
1645 int binary_match_only
= 0;
1647 int try_lookahead
= 0;
1648 int show_function
= 0;
1649 struct userdiff_driver
*textconv
= NULL
;
1650 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1654 opt
->output
= std_output
;
1656 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1658 /* Show hunk marks, except for the first file. */
1659 if (opt
->last_shown
)
1660 opt
->show_hunk_mark
= 1;
1662 * If we're using threads then we can't easily identify
1663 * the first file. Always put hunk marks in that case
1664 * and skip the very first one later in work_done().
1666 if (opt
->output
!= std_output
)
1667 opt
->show_hunk_mark
= 1;
1669 opt
->last_shown
= 0;
1671 if (opt
->allow_textconv
) {
1672 grep_source_load_driver(gs
);
1674 * We might set up the shared textconv cache data here, which
1675 * is not thread-safe.
1678 textconv
= userdiff_get_textconv(gs
->driver
);
1683 * We know the result of a textconv is text, so we only have to care
1684 * about binary handling if we are not using it.
1687 switch (opt
->binary
) {
1688 case GREP_BINARY_DEFAULT
:
1689 if (grep_source_is_binary(gs
))
1690 binary_match_only
= 1;
1692 case GREP_BINARY_NOMATCH
:
1693 if (grep_source_is_binary(gs
))
1694 return 0; /* Assume unmatch */
1696 case GREP_BINARY_TEXT
:
1699 die("BUG: unknown binary handling mode");
1703 memset(&xecfg
, 0, sizeof(xecfg
));
1706 try_lookahead
= should_lookahead(opt
);
1708 if (fill_textconv_grep(textconv
, gs
) < 0)
1718 * look_ahead() skips quickly to the line that possibly
1719 * has the next hit; don't call it if we need to do
1720 * something more than just skipping the current line
1721 * in response to an unmatch for the current line. E.g.
1722 * inside a post-context window, we will show the current
1723 * line as a context around the previous hit when it
1728 && (show_function
||
1729 lno
<= last_hit
+ opt
->post_context
))
1730 && look_ahead(opt
, &left
, &lno
, &bol
))
1732 eol
= end_of_line(bol
, &left
);
1736 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1737 ctx
= GREP_CONTEXT_BODY
;
1739 hit
= match_line(opt
, bol
, eol
, ctx
, collect_hits
);
1745 /* "grep -v -e foo -e bla" should list lines
1746 * that do not have either, so inversion should
1751 if (opt
->unmatch_name_only
) {
1758 if (opt
->status_only
)
1760 if (opt
->name_only
) {
1761 show_name(opt
, gs
->name
);
1766 if (binary_match_only
) {
1767 opt
->output(opt
, "Binary file ", 12);
1768 output_color(opt
, gs
->name
, strlen(gs
->name
),
1769 opt
->color_filename
);
1770 opt
->output(opt
, " matches\n", 9);
1773 /* Hit at this line. If we haven't shown the
1774 * pre-context lines, we would need to show them.
1776 if (opt
->pre_context
|| opt
->funcbody
)
1777 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1778 else if (opt
->funcname
)
1779 show_funcname_line(opt
, gs
, bol
, lno
);
1780 show_line(opt
, bol
, eol
, gs
->name
, lno
, ':');
1786 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1787 unsigned long peek_left
= left
;
1788 char *peek_eol
= eol
;
1791 * Trailing empty lines are not interesting.
1792 * Peek past them to see if they belong to the
1793 * body of the current function.
1796 while (is_empty_line(peek_bol
, peek_eol
)) {
1797 peek_bol
= peek_eol
+ 1;
1798 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1801 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1804 if (show_function
||
1805 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1806 /* If the last hit is within the post context,
1807 * we need to show this line.
1809 show_line(opt
, bol
, eol
, gs
->name
, lno
, '-');
1823 if (opt
->status_only
)
1824 return opt
->unmatch_name_only
;
1825 if (opt
->unmatch_name_only
) {
1826 /* We did not see any hit, so we want to show this */
1827 show_name(opt
, gs
->name
);
1831 xdiff_clear_find_func(&xecfg
);
1835 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1836 * which feels mostly useless but sometimes useful. Maybe
1837 * make it another option? For now suppress them.
1839 if (opt
->count
&& count
) {
1841 if (opt
->pathname
) {
1842 output_color(opt
, gs
->name
, strlen(gs
->name
),
1843 opt
->color_filename
);
1844 output_sep(opt
, ':');
1846 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1847 opt
->output(opt
, buf
, strlen(buf
));
1853 static void clr_hit_marker(struct grep_expr
*x
)
1855 /* All-hit markers are meaningful only at the very top level
1860 if (x
->node
!= GREP_NODE_OR
)
1862 x
->u
.binary
.left
->hit
= 0;
1863 x
= x
->u
.binary
.right
;
1867 static int chk_hit_marker(struct grep_expr
*x
)
1869 /* Top level nodes have hit markers. See if they all are hits */
1871 if (x
->node
!= GREP_NODE_OR
)
1873 if (!x
->u
.binary
.left
->hit
)
1875 x
= x
->u
.binary
.right
;
1879 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1882 * we do not have to do the two-pass grep when we do not check
1883 * buffer-wide "all-match".
1885 if (!opt
->all_match
)
1886 return grep_source_1(opt
, gs
, 0);
1888 /* Otherwise the toplevel "or" terms hit a bit differently.
1889 * We first clear hit markers from them.
1891 clr_hit_marker(opt
->pattern_expression
);
1892 grep_source_1(opt
, gs
, 1);
1894 if (!chk_hit_marker(opt
->pattern_expression
))
1897 return grep_source_1(opt
, gs
, 0);
1900 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
1902 struct grep_source gs
;
1905 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
1909 r
= grep_source(opt
, &gs
);
1911 grep_source_clear(&gs
);
1915 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
1916 const char *name
, const char *path
,
1917 const void *identifier
)
1920 gs
->name
= xstrdup_or_null(name
);
1921 gs
->path
= xstrdup_or_null(path
);
1927 case GREP_SOURCE_FILE
:
1928 gs
->identifier
= xstrdup(identifier
);
1930 case GREP_SOURCE_SUBMODULE
:
1932 gs
->identifier
= NULL
;
1937 * If the identifier is non-NULL (in the submodule case) it
1938 * will be a SHA1 that needs to be copied.
1940 case GREP_SOURCE_OID
:
1941 gs
->identifier
= oiddup(identifier
);
1943 case GREP_SOURCE_BUF
:
1944 gs
->identifier
= NULL
;
1949 void grep_source_clear(struct grep_source
*gs
)
1951 FREE_AND_NULL(gs
->name
);
1952 FREE_AND_NULL(gs
->path
);
1953 FREE_AND_NULL(gs
->identifier
);
1954 grep_source_clear_data(gs
);
1957 void grep_source_clear_data(struct grep_source
*gs
)
1960 case GREP_SOURCE_FILE
:
1961 case GREP_SOURCE_OID
:
1962 case GREP_SOURCE_SUBMODULE
:
1963 FREE_AND_NULL(gs
->buf
);
1966 case GREP_SOURCE_BUF
:
1967 /* leave user-provided buf intact */
1972 static int grep_source_load_oid(struct grep_source
*gs
)
1974 enum object_type type
;
1977 gs
->buf
= read_sha1_file(gs
->identifier
, &type
, &gs
->size
);
1981 return error(_("'%s': unable to read %s"),
1983 oid_to_hex(gs
->identifier
));
1987 static int grep_source_load_file(struct grep_source
*gs
)
1989 const char *filename
= gs
->identifier
;
1995 if (lstat(filename
, &st
) < 0) {
1997 if (errno
!= ENOENT
)
1998 error_errno(_("failed to stat '%s'"), filename
);
2001 if (!S_ISREG(st
.st_mode
))
2003 size
= xsize_t(st
.st_size
);
2004 i
= open(filename
, O_RDONLY
);
2007 data
= xmallocz(size
);
2008 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2009 error_errno(_("'%s': short read"), filename
);
2021 static int grep_source_load(struct grep_source
*gs
)
2027 case GREP_SOURCE_FILE
:
2028 return grep_source_load_file(gs
);
2029 case GREP_SOURCE_OID
:
2030 return grep_source_load_oid(gs
);
2031 case GREP_SOURCE_BUF
:
2032 return gs
->buf
? 0 : -1;
2033 case GREP_SOURCE_SUBMODULE
:
2036 die("BUG: invalid grep_source type to load");
2039 void grep_source_load_driver(struct grep_source
*gs
)
2046 gs
->driver
= userdiff_find_by_path(gs
->path
);
2048 gs
->driver
= userdiff_find_by_name("default");
2052 static int grep_source_is_binary(struct grep_source
*gs
)
2054 grep_source_load_driver(gs
);
2055 if (gs
->driver
->binary
!= -1)
2056 return gs
->driver
->binary
;
2058 if (!grep_source_load(gs
))
2059 return buffer_is_binary(gs
->buf
, gs
->size
);