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 const char *color_grep_slots
[] = {
17 [GREP_COLOR_CONTEXT
] = "context",
18 [GREP_COLOR_FILENAME
] = "filename",
19 [GREP_COLOR_FUNCTION
] = "function",
20 [GREP_COLOR_LINENO
] = "lineNumber",
21 [GREP_COLOR_MATCH_CONTEXT
] = "matchContext",
22 [GREP_COLOR_MATCH_SELECTED
] = "matchSelected",
23 [GREP_COLOR_SELECTED
] = "selected",
24 [GREP_COLOR_SEP
] = "separator",
27 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
29 fwrite(buf
, size
, 1, stdout
);
32 static void color_set(char *dst
, const char *color_bytes
)
34 xsnprintf(dst
, COLOR_MAXLEN
, "%s", color_bytes
);
38 * Initialize the grep_defaults template with hardcoded defaults.
39 * We could let the compiler do this, but without C99 initializers
40 * the code gets unwieldy and unreadable, so...
42 void init_grep_defaults(void)
44 struct grep_opt
*opt
= &grep_defaults
;
51 memset(opt
, 0, sizeof(*opt
));
55 opt
->pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
;
56 color_set(opt
->colors
[GREP_COLOR_CONTEXT
], "");
57 color_set(opt
->colors
[GREP_COLOR_FILENAME
], "");
58 color_set(opt
->colors
[GREP_COLOR_FUNCTION
], "");
59 color_set(opt
->colors
[GREP_COLOR_LINENO
], "");
60 color_set(opt
->colors
[GREP_COLOR_MATCH_CONTEXT
], GIT_COLOR_BOLD_RED
);
61 color_set(opt
->colors
[GREP_COLOR_MATCH_SELECTED
], GIT_COLOR_BOLD_RED
);
62 color_set(opt
->colors
[GREP_COLOR_SELECTED
], "");
63 color_set(opt
->colors
[GREP_COLOR_SEP
], GIT_COLOR_CYAN
);
65 opt
->output
= std_output
;
68 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
70 if (!strcmp(arg
, "default"))
71 return GREP_PATTERN_TYPE_UNSPECIFIED
;
72 else if (!strcmp(arg
, "basic"))
73 return GREP_PATTERN_TYPE_BRE
;
74 else if (!strcmp(arg
, "extended"))
75 return GREP_PATTERN_TYPE_ERE
;
76 else if (!strcmp(arg
, "fixed"))
77 return GREP_PATTERN_TYPE_FIXED
;
78 else if (!strcmp(arg
, "perl"))
79 return GREP_PATTERN_TYPE_PCRE
;
80 die("bad %s argument: %s", opt
, arg
);
84 * Read the configuration file once and store it in
85 * the grep_defaults template.
87 int grep_config(const char *var
, const char *value
, void *cb
)
89 struct grep_opt
*opt
= &grep_defaults
;
92 if (userdiff_config(var
, value
) < 0)
95 if (!strcmp(var
, "grep.extendedregexp")) {
96 opt
->extended_regexp_option
= git_config_bool(var
, value
);
100 if (!strcmp(var
, "grep.patterntype")) {
101 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
105 if (!strcmp(var
, "grep.linenumber")) {
106 opt
->linenum
= git_config_bool(var
, value
);
110 if (!strcmp(var
, "grep.fullname")) {
111 opt
->relative
= !git_config_bool(var
, value
);
115 if (!strcmp(var
, "color.grep"))
116 opt
->color
= git_config_colorbool(var
, value
);
117 if (!strcmp(var
, "color.grep.match")) {
118 if (grep_config("color.grep.matchcontext", value
, cb
) < 0)
120 if (grep_config("color.grep.matchselected", value
, cb
) < 0)
122 } else if (skip_prefix(var
, "color.grep.", &slot
)) {
123 int i
= LOOKUP_CONFIG(color_grep_slots
, slot
);
128 color
= opt
->colors
[i
];
130 return config_error_nonbool(var
);
131 return color_parse(value
, color
);
137 * Initialize one instance of grep_opt and copy the
138 * default values from the template we read the configuration
139 * information in an earlier call to git_config(grep_config).
141 void grep_init(struct grep_opt
*opt
, const char *prefix
)
143 struct grep_opt
*def
= &grep_defaults
;
146 memset(opt
, 0, sizeof(*opt
));
147 opt
->prefix
= prefix
;
148 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
149 opt
->pattern_tail
= &opt
->pattern_list
;
150 opt
->header_tail
= &opt
->header_list
;
152 opt
->color
= def
->color
;
153 opt
->extended_regexp_option
= def
->extended_regexp_option
;
154 opt
->pattern_type_option
= def
->pattern_type_option
;
155 opt
->linenum
= def
->linenum
;
156 opt
->max_depth
= def
->max_depth
;
157 opt
->pathname
= def
->pathname
;
158 opt
->relative
= def
->relative
;
159 opt
->output
= def
->output
;
161 for (i
= 0; i
< NR_GREP_COLORS
; i
++)
162 color_set(opt
->colors
[i
], def
->colors
[i
]);
165 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
168 * When committing to the pattern type by setting the relevant
169 * fields in grep_opt it's generally not necessary to zero out
170 * the fields we're not choosing, since they won't have been
171 * set by anything. The extended_regexp_option field is the
172 * only exception to this.
174 * This is because in the process of parsing grep.patternType
175 * & grep.extendedRegexp we set opt->pattern_type_option and
176 * opt->extended_regexp_option, respectively. We then
177 * internally use opt->extended_regexp_option to see if we're
178 * compiling an ERE. It must be unset if that's not actually
181 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
182 opt
->extended_regexp_option
)
183 opt
->extended_regexp_option
= 0;
185 switch (pattern_type
) {
186 case GREP_PATTERN_TYPE_UNSPECIFIED
:
189 case GREP_PATTERN_TYPE_BRE
:
192 case GREP_PATTERN_TYPE_ERE
:
193 opt
->extended_regexp_option
= 1;
196 case GREP_PATTERN_TYPE_FIXED
:
200 case GREP_PATTERN_TYPE_PCRE
:
205 * It's important that pcre1 always be assigned to
206 * even when there's no USE_LIBPCRE* defined. We still
207 * call the PCRE stub function, it just dies with
208 * "cannot use Perl-compatible regexes[...]".
216 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
218 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
219 grep_set_pattern_type_option(pattern_type
, opt
);
220 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
221 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
222 else if (opt
->extended_regexp_option
)
224 * This branch *must* happen after setting from the
225 * opt->pattern_type_option above, we don't want
226 * grep.extendedRegexp to override grep.patternType!
228 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
231 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
232 const char *origin
, int no
,
233 enum grep_pat_token t
,
234 enum grep_header_field field
)
236 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
237 p
->pattern
= xmemdupz(pat
, patlen
);
238 p
->patternlen
= patlen
;
246 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
253 case GREP_PATTERN
: /* atom */
254 case GREP_PATTERN_HEAD
:
255 case GREP_PATTERN_BODY
:
257 struct grep_pat
*new_pat
;
259 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
260 while (++len
<= p
->patternlen
) {
261 if (*(--cp
) == '\n') {
268 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
269 p
->no
, p
->token
, p
->field
);
270 new_pat
->next
= p
->next
;
272 *tail
= &new_pat
->next
;
275 p
->patternlen
-= len
;
283 void append_header_grep_pattern(struct grep_opt
*opt
,
284 enum grep_header_field field
, const char *pat
)
286 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
287 GREP_PATTERN_HEAD
, field
);
288 if (field
== GREP_HEADER_REFLOG
)
289 opt
->use_reflog_filter
= 1;
290 do_append_grep_pat(&opt
->header_tail
, p
);
293 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
294 const char *origin
, int no
, enum grep_pat_token t
)
296 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
299 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
300 const char *origin
, int no
, enum grep_pat_token t
)
302 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
303 do_append_grep_pat(&opt
->pattern_tail
, p
);
306 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
308 struct grep_pat
*pat
;
309 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
312 ret
->pattern_list
= NULL
;
313 ret
->pattern_tail
= &ret
->pattern_list
;
315 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
317 if(pat
->token
== GREP_PATTERN_HEAD
)
318 append_header_grep_pattern(ret
, pat
->field
,
321 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
322 pat
->origin
, pat
->no
, pat
->token
);
328 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
334 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
336 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
340 die("%s'%s': %s", where
, p
->pattern
, error
);
343 static int is_fixed(const char *s
, size_t len
)
347 for (i
= 0; i
< len
; i
++) {
348 if (is_regex_special(s
[i
]))
355 static int has_null(const char *s
, size_t len
)
358 * regcomp cannot accept patterns with NULs so when using it
359 * we consider any pattern containing a NUL fixed.
361 if (memchr(s
, 0, len
))
368 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
372 int options
= PCRE_MULTILINE
;
374 if (opt
->ignore_case
) {
375 if (has_non_ascii(p
->pattern
))
376 p
->pcre1_tables
= pcre_maketables();
377 options
|= PCRE_CASELESS
;
379 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
380 options
|= PCRE_UTF8
;
382 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
384 if (!p
->pcre1_regexp
)
385 compile_regexp_failed(p
, error
);
387 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, GIT_PCRE_STUDY_JIT_COMPILE
, &error
);
388 if (!p
->pcre1_extra_info
&& error
)
391 #ifdef GIT_PCRE1_USE_JIT
392 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
393 if (p
->pcre1_jit_on
== 1) {
394 p
->pcre1_jit_stack
= pcre_jit_stack_alloc(1, 1024 * 1024);
395 if (!p
->pcre1_jit_stack
)
396 die("Couldn't allocate PCRE JIT stack");
397 pcre_assign_jit_stack(p
->pcre1_extra_info
, NULL
, p
->pcre1_jit_stack
);
398 } else if (p
->pcre1_jit_on
!= 0) {
399 die("BUG: The pcre1_jit_on variable should be 0 or 1, not %d",
405 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
406 regmatch_t
*match
, int eflags
)
408 int ovector
[30], ret
, flags
= 0;
410 if (eflags
& REG_NOTBOL
)
411 flags
|= PCRE_NOTBOL
;
413 #ifdef GIT_PCRE1_USE_JIT
414 if (p
->pcre1_jit_on
) {
415 ret
= pcre_jit_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
416 eol
- line
, 0, flags
, ovector
,
417 ARRAY_SIZE(ovector
), p
->pcre1_jit_stack
);
421 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
422 eol
- line
, 0, flags
, ovector
,
423 ARRAY_SIZE(ovector
));
426 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
427 die("pcre_exec failed with error code %d", ret
);
430 match
->rm_so
= ovector
[0];
431 match
->rm_eo
= ovector
[1];
437 static void free_pcre1_regexp(struct grep_pat
*p
)
439 pcre_free(p
->pcre1_regexp
);
440 #ifdef GIT_PCRE1_USE_JIT
441 if (p
->pcre1_jit_on
) {
442 pcre_free_study(p
->pcre1_extra_info
);
443 pcre_jit_stack_free(p
->pcre1_jit_stack
);
447 pcre_free(p
->pcre1_extra_info
);
449 pcre_free((void *)p
->pcre1_tables
);
451 #else /* !USE_LIBPCRE1 */
452 static void compile_pcre1_regexp(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 pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
458 regmatch_t
*match
, int eflags
)
463 static void free_pcre1_regexp(struct grep_pat
*p
)
466 #endif /* !USE_LIBPCRE1 */
469 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
472 PCRE2_UCHAR errbuf
[256];
473 PCRE2_SIZE erroffset
;
474 int options
= PCRE2_MULTILINE
;
475 const uint8_t *character_tables
= NULL
;
482 p
->pcre2_compile_context
= NULL
;
484 if (opt
->ignore_case
) {
485 if (has_non_ascii(p
->pattern
)) {
486 character_tables
= pcre2_maketables(NULL
);
487 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
488 pcre2_set_character_tables(p
->pcre2_compile_context
, character_tables
);
490 options
|= PCRE2_CASELESS
;
492 if (is_utf8_locale() && has_non_ascii(p
->pattern
))
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
== 1) {
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) {
537 p
->pcre2_jit_stack
= pcre2_jit_stack_create(1, 1024 * 1024, NULL
);
538 if (!p
->pcre2_jit_stack
)
539 die("Couldn't allocate PCRE2 JIT stack");
540 p
->pcre2_match_context
= pcre2_match_context_create(NULL
);
541 if (!p
->pcre2_match_context
)
542 die("Couldn't allocate PCRE2 match context");
543 pcre2_jit_stack_assign(p
->pcre2_match_context
, NULL
, p
->pcre2_jit_stack
);
544 } else if (p
->pcre2_jit_on
!= 0) {
545 die("BUG: The pcre2_jit_on variable should be 0 or 1, not %d",
550 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
551 regmatch_t
*match
, int eflags
)
555 PCRE2_UCHAR errbuf
[256];
557 if (eflags
& REG_NOTBOL
)
558 flags
|= PCRE2_NOTBOL
;
561 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
562 eol
- line
, 0, flags
, p
->pcre2_match_data
,
565 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
566 eol
- line
, 0, flags
, p
->pcre2_match_data
,
569 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
570 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
571 die("%s failed with error code %d: %s",
572 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
576 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
578 match
->rm_so
= (int)ovector
[0];
579 match
->rm_eo
= (int)ovector
[1];
585 static void free_pcre2_pattern(struct grep_pat
*p
)
587 pcre2_compile_context_free(p
->pcre2_compile_context
);
588 pcre2_code_free(p
->pcre2_pattern
);
589 pcre2_match_data_free(p
->pcre2_match_data
);
590 pcre2_jit_stack_free(p
->pcre2_jit_stack
);
591 pcre2_match_context_free(p
->pcre2_match_context
);
593 #else /* !USE_LIBPCRE2 */
594 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
597 * Unreachable until USE_LIBPCRE2 becomes synonymous with
598 * USE_LIBPCRE. See the sibling comment in
599 * grep_set_pattern_type_option().
601 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
604 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
605 regmatch_t
*match
, int eflags
)
610 static void free_pcre2_pattern(struct grep_pat
*p
)
613 #endif /* !USE_LIBPCRE2 */
615 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
617 struct strbuf sb
= STRBUF_INIT
;
621 basic_regex_quote_buf(&sb
, p
->pattern
);
622 if (opt
->ignore_case
)
623 regflags
|= REG_ICASE
;
624 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
626 fprintf(stderr
, "fixed %s\n", sb
.buf
);
630 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
632 compile_regexp_failed(p
, errbuf
);
636 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
640 int regflags
= REG_NEWLINE
;
642 p
->word_regexp
= opt
->word_regexp
;
643 p
->ignore_case
= opt
->ignore_case
;
644 ascii_only
= !has_non_ascii(p
->pattern
);
647 * Even when -F (fixed) asks us to do a non-regexp search, we
648 * may not be able to correctly case-fold when -i
649 * (ignore-case) is asked (in which case, we'll synthesize a
650 * regexp to match the pattern that matches regexp special
651 * characters literally, while ignoring case differences). On
652 * the other hand, even without -F, if the pattern does not
653 * have any regexp special characters and there is no need for
654 * case-folding search, we can internally turn it into a
655 * simple string match using kws. p->fixed tells us if we
659 has_null(p
->pattern
, p
->patternlen
) ||
660 is_fixed(p
->pattern
, p
->patternlen
))
661 p
->fixed
= !p
->ignore_case
|| ascii_only
;
664 p
->kws
= kwsalloc(p
->ignore_case
? tolower_trans_tbl
: NULL
);
665 kwsincr(p
->kws
, p
->pattern
, p
->patternlen
);
668 } else if (opt
->fixed
) {
670 * We come here when the pattern has the non-ascii
671 * characters we cannot case-fold, and asked to
674 compile_fixed_regexp(p
, opt
);
679 compile_pcre2_pattern(p
, opt
);
684 compile_pcre1_regexp(p
, opt
);
689 regflags
|= REG_ICASE
;
690 if (opt
->extended_regexp_option
)
691 regflags
|= REG_EXTENDED
;
692 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
695 regerror(err
, &p
->regexp
, errbuf
, 1024);
697 compile_regexp_failed(p
, errbuf
);
701 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
702 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
711 case GREP_PATTERN
: /* atom */
712 case GREP_PATTERN_HEAD
:
713 case GREP_PATTERN_BODY
:
714 x
= xcalloc(1, sizeof (struct grep_expr
));
715 x
->node
= GREP_NODE_ATOM
;
719 case GREP_OPEN_PAREN
:
721 x
= compile_pattern_or(list
);
722 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
723 die("unmatched parenthesis");
724 *list
= (*list
)->next
;
731 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
742 die("--not not followed by pattern expression");
744 x
= xcalloc(1, sizeof (struct grep_expr
));
745 x
->node
= GREP_NODE_NOT
;
746 x
->u
.unary
= compile_pattern_not(list
);
748 die("--not followed by non pattern expression");
751 return compile_pattern_atom(list
);
755 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
758 struct grep_expr
*x
, *y
, *z
;
760 x
= compile_pattern_not(list
);
762 if (p
&& p
->token
== GREP_AND
) {
764 die("--and not followed by pattern expression");
766 y
= compile_pattern_and(list
);
768 die("--and not followed by pattern expression");
769 z
= xcalloc(1, sizeof (struct grep_expr
));
770 z
->node
= GREP_NODE_AND
;
771 z
->u
.binary
.left
= x
;
772 z
->u
.binary
.right
= y
;
778 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
781 struct grep_expr
*x
, *y
, *z
;
783 x
= compile_pattern_and(list
);
785 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
786 y
= compile_pattern_or(list
);
788 die("not a pattern expression %s", p
->pattern
);
789 z
= xcalloc(1, sizeof (struct grep_expr
));
790 z
->node
= GREP_NODE_OR
;
791 z
->u
.binary
.left
= x
;
792 z
->u
.binary
.right
= y
;
798 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
800 return compile_pattern_or(list
);
803 static void indent(int in
)
809 static void dump_grep_pat(struct grep_pat
*p
)
812 case GREP_AND
: fprintf(stderr
, "*and*"); break;
813 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
814 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
815 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
816 case GREP_OR
: fprintf(stderr
, "*or*"); break;
818 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
819 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
820 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
825 case GREP_PATTERN_HEAD
:
826 fprintf(stderr
, "<head %d>", p
->field
); break;
827 case GREP_PATTERN_BODY
:
828 fprintf(stderr
, "<body>"); break;
832 case GREP_PATTERN_HEAD
:
833 case GREP_PATTERN_BODY
:
835 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
841 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
846 fprintf(stderr
, "true\n");
849 dump_grep_pat(x
->u
.atom
);
852 fprintf(stderr
, "(not\n");
853 dump_grep_expression_1(x
->u
.unary
, in
+1);
855 fprintf(stderr
, ")\n");
858 fprintf(stderr
, "(and\n");
859 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
860 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
862 fprintf(stderr
, ")\n");
865 fprintf(stderr
, "(or\n");
866 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
867 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
869 fprintf(stderr
, ")\n");
874 static void dump_grep_expression(struct grep_opt
*opt
)
876 struct grep_expr
*x
= opt
->pattern_expression
;
879 fprintf(stderr
, "[all-match]\n");
880 dump_grep_expression_1(x
, 0);
884 static struct grep_expr
*grep_true_expr(void)
886 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
887 z
->node
= GREP_NODE_TRUE
;
891 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
893 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
894 z
->node
= GREP_NODE_OR
;
895 z
->u
.binary
.left
= left
;
896 z
->u
.binary
.right
= right
;
900 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
903 struct grep_expr
*header_expr
;
904 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
905 enum grep_header_field fld
;
907 if (!opt
->header_list
)
910 for (p
= opt
->header_list
; p
; p
= p
->next
) {
911 if (p
->token
!= GREP_PATTERN_HEAD
)
912 die("BUG: a non-header pattern in grep header list.");
913 if (p
->field
< GREP_HEADER_FIELD_MIN
||
914 GREP_HEADER_FIELD_MAX
<= p
->field
)
915 die("BUG: unknown header field %d", p
->field
);
916 compile_regexp(p
, opt
);
919 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
920 header_group
[fld
] = NULL
;
922 for (p
= opt
->header_list
; p
; p
= p
->next
) {
924 struct grep_pat
*pp
= p
;
926 h
= compile_pattern_atom(&pp
);
927 if (!h
|| pp
!= p
->next
)
928 die("BUG: malformed header expr");
929 if (!header_group
[p
->field
]) {
930 header_group
[p
->field
] = h
;
933 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
938 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
939 if (!header_group
[fld
])
942 header_expr
= grep_true_expr();
943 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
948 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
950 struct grep_expr
*z
= x
;
953 assert(x
->node
== GREP_NODE_OR
);
954 if (x
->u
.binary
.right
&&
955 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
956 x
->u
.binary
.right
= y
;
959 x
= x
->u
.binary
.right
;
964 static void compile_grep_patterns_real(struct grep_opt
*opt
)
967 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
969 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
971 case GREP_PATTERN
: /* atom */
972 case GREP_PATTERN_HEAD
:
973 case GREP_PATTERN_BODY
:
974 compile_regexp(p
, opt
);
982 if (opt
->all_match
|| header_expr
)
984 else if (!opt
->extended
&& !opt
->debug
)
987 p
= opt
->pattern_list
;
989 opt
->pattern_expression
= compile_pattern_expr(&p
);
991 die("incomplete pattern expression: %s", p
->pattern
);
996 if (!opt
->pattern_expression
)
997 opt
->pattern_expression
= header_expr
;
998 else if (opt
->all_match
)
999 opt
->pattern_expression
= grep_splice_or(header_expr
,
1000 opt
->pattern_expression
);
1002 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
1007 void compile_grep_patterns(struct grep_opt
*opt
)
1009 compile_grep_patterns_real(opt
);
1011 dump_grep_expression(opt
);
1014 static void free_pattern_expr(struct grep_expr
*x
)
1017 case GREP_NODE_TRUE
:
1018 case GREP_NODE_ATOM
:
1021 free_pattern_expr(x
->u
.unary
);
1025 free_pattern_expr(x
->u
.binary
.left
);
1026 free_pattern_expr(x
->u
.binary
.right
);
1032 void free_grep_patterns(struct grep_opt
*opt
)
1034 struct grep_pat
*p
, *n
;
1036 for (p
= opt
->pattern_list
; p
; p
= n
) {
1039 case GREP_PATTERN
: /* atom */
1040 case GREP_PATTERN_HEAD
:
1041 case GREP_PATTERN_BODY
:
1044 else if (p
->pcre1_regexp
)
1045 free_pcre1_regexp(p
);
1046 else if (p
->pcre2_pattern
)
1047 free_pcre2_pattern(p
);
1049 regfree(&p
->regexp
);
1060 free_pattern_expr(opt
->pattern_expression
);
1063 static char *end_of_line(char *cp
, unsigned long *left
)
1065 unsigned long l
= *left
;
1066 while (l
&& *cp
!= '\n') {
1074 static int word_char(char ch
)
1076 return isalnum(ch
) || ch
== '_';
1079 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1082 if (want_color(opt
->color
) && color
&& color
[0]) {
1083 opt
->output(opt
, color
, strlen(color
));
1084 opt
->output(opt
, data
, size
);
1085 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1087 opt
->output(opt
, data
, size
);
1090 static void output_sep(struct grep_opt
*opt
, char sign
)
1092 if (opt
->null_following_name
)
1093 opt
->output(opt
, "\0", 1);
1095 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
1098 static void show_name(struct grep_opt
*opt
, const char *name
)
1100 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1101 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1104 static int fixmatch(struct grep_pat
*p
, char *line
, char *eol
,
1107 struct kwsmatch kwsm
;
1108 size_t offset
= kwsexec(p
->kws
, line
, eol
- line
, &kwsm
);
1110 match
->rm_so
= match
->rm_eo
= -1;
1113 match
->rm_so
= offset
;
1114 match
->rm_eo
= match
->rm_so
+ kwsm
.size
[0];
1119 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1120 regmatch_t
*match
, int eflags
)
1125 hit
= !fixmatch(p
, line
, eol
, match
);
1126 else if (p
->pcre1_regexp
)
1127 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1128 else if (p
->pcre2_pattern
)
1129 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1131 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1137 static int strip_timestamp(char *bol
, char **eol_p
)
1142 while (bol
< --eol
) {
1156 } header_field
[] = {
1158 { "committer ", 10 },
1162 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1163 enum grep_context ctx
,
1164 regmatch_t
*pmatch
, int eflags
)
1168 const char *start
= bol
;
1170 if ((p
->token
!= GREP_PATTERN
) &&
1171 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1174 if (p
->token
== GREP_PATTERN_HEAD
) {
1177 assert(p
->field
< ARRAY_SIZE(header_field
));
1178 field
= header_field
[p
->field
].field
;
1179 len
= header_field
[p
->field
].len
;
1180 if (strncmp(bol
, field
, len
))
1184 case GREP_HEADER_AUTHOR
:
1185 case GREP_HEADER_COMMITTER
:
1186 saved_ch
= strip_timestamp(bol
, &eol
);
1194 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1196 if (hit
&& p
->word_regexp
) {
1197 if ((pmatch
[0].rm_so
< 0) ||
1198 (eol
- bol
) < pmatch
[0].rm_so
||
1199 (pmatch
[0].rm_eo
< 0) ||
1200 (eol
- bol
) < pmatch
[0].rm_eo
)
1201 die("regexp returned nonsense");
1203 /* Match beginning must be either beginning of the
1204 * line, or at word boundary (i.e. the last char must
1205 * not be a word char). Similarly, match end must be
1206 * either end of the line, or at word boundary
1207 * (i.e. the next char must not be a word char).
1209 if ( ((pmatch
[0].rm_so
== 0) ||
1210 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1211 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1212 !word_char(bol
[pmatch
[0].rm_eo
])) )
1217 /* Words consist of at least one character. */
1218 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1221 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1222 /* There could be more than one match on the
1223 * line, and the first match might not be
1224 * strict word match. But later ones could be!
1225 * Forward to the next possible start, i.e. the
1226 * next position following a non-word char.
1228 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1229 while (word_char(bol
[-1]) && bol
< eol
)
1231 eflags
|= REG_NOTBOL
;
1236 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1239 pmatch
[0].rm_so
+= bol
- start
;
1240 pmatch
[0].rm_eo
+= bol
- start
;
1245 static int match_expr_eval(struct grep_expr
*x
, char *bol
, char *eol
,
1246 enum grep_context ctx
, int collect_hits
)
1252 die("Not a valid grep expression");
1254 case GREP_NODE_TRUE
:
1257 case GREP_NODE_ATOM
:
1258 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
, &match
, 0);
1261 h
= !match_expr_eval(x
->u
.unary
, bol
, eol
, ctx
, 0);
1264 if (!match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0))
1266 h
= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 0);
1270 return (match_expr_eval(x
->u
.binary
.left
,
1271 bol
, eol
, ctx
, 0) ||
1272 match_expr_eval(x
->u
.binary
.right
,
1274 h
= match_expr_eval(x
->u
.binary
.left
, bol
, eol
, ctx
, 0);
1275 x
->u
.binary
.left
->hit
|= h
;
1276 h
|= match_expr_eval(x
->u
.binary
.right
, bol
, eol
, ctx
, 1);
1279 die("Unexpected node type (internal error) %d", x
->node
);
1286 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1287 enum grep_context ctx
, int collect_hits
)
1289 struct grep_expr
*x
= opt
->pattern_expression
;
1290 return match_expr_eval(x
, bol
, eol
, ctx
, collect_hits
);
1293 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1294 enum grep_context ctx
, int collect_hits
)
1300 return match_expr(opt
, bol
, eol
, ctx
, collect_hits
);
1302 /* we do not call with collect_hits without being extended */
1303 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1304 if (match_one_pattern(p
, bol
, eol
, ctx
, &match
, 0))
1310 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1311 enum grep_context ctx
,
1312 regmatch_t
*pmatch
, int eflags
)
1316 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1318 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1320 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1321 if (match
.rm_so
> pmatch
->rm_so
)
1323 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1326 pmatch
->rm_so
= match
.rm_so
;
1327 pmatch
->rm_eo
= match
.rm_eo
;
1331 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1332 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1337 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1339 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1341 case GREP_PATTERN
: /* atom */
1342 case GREP_PATTERN_HEAD
:
1343 case GREP_PATTERN_BODY
:
1344 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1355 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1356 const char *name
, unsigned lno
, char sign
)
1358 int rest
= eol
- bol
;
1359 const char *match_color
, *line_color
= NULL
;
1361 if (opt
->file_break
&& opt
->last_shown
== 0) {
1362 if (opt
->show_hunk_mark
)
1363 opt
->output(opt
, "\n", 1);
1364 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1365 if (opt
->last_shown
== 0) {
1366 if (opt
->show_hunk_mark
) {
1367 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1368 opt
->output(opt
, "\n", 1);
1370 } else if (lno
> opt
->last_shown
+ 1) {
1371 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1372 opt
->output(opt
, "\n", 1);
1375 if (opt
->heading
&& opt
->last_shown
== 0) {
1376 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1377 opt
->output(opt
, "\n", 1);
1379 opt
->last_shown
= lno
;
1381 if (!opt
->heading
&& opt
->pathname
) {
1382 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1383 output_sep(opt
, sign
);
1387 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1388 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1389 output_sep(opt
, sign
);
1393 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1398 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1400 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1402 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1403 else if (sign
== '-')
1404 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1405 else if (sign
== '=')
1406 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1408 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1409 if (match
.rm_so
== match
.rm_eo
)
1412 output_color(opt
, bol
, match
.rm_so
, line_color
);
1413 output_color(opt
, bol
+ match
.rm_so
,
1414 match
.rm_eo
- match
.rm_so
, match_color
);
1416 rest
-= match
.rm_eo
;
1417 eflags
= REG_NOTBOL
;
1421 output_color(opt
, bol
, rest
, line_color
);
1422 opt
->output(opt
, "\n", 1);
1429 * This lock protects access to the gitattributes machinery, which is
1432 pthread_mutex_t grep_attr_mutex
;
1434 static inline void grep_attr_lock(void)
1437 pthread_mutex_lock(&grep_attr_mutex
);
1440 static inline void grep_attr_unlock(void)
1443 pthread_mutex_unlock(&grep_attr_mutex
);
1447 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1449 pthread_mutex_t grep_read_mutex
;
1452 #define grep_attr_lock()
1453 #define grep_attr_unlock()
1456 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1458 xdemitconf_t
*xecfg
= opt
->priv
;
1459 if (xecfg
&& !xecfg
->find_func
) {
1460 grep_source_load_driver(gs
);
1461 if (gs
->driver
->funcname
.pattern
) {
1462 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1463 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1465 xecfg
= opt
->priv
= NULL
;
1471 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1472 xecfg
->find_func_priv
) >= 0;
1477 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1482 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1483 char *bol
, unsigned lno
)
1485 while (bol
> gs
->buf
) {
1488 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1492 if (lno
<= opt
->last_shown
)
1495 if (match_funcname(opt
, gs
, bol
, eol
)) {
1496 show_line(opt
, bol
, eol
, gs
->name
, lno
, '=');
1502 static int is_empty_line(const char *bol
, const char *eol
);
1504 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1505 char *bol
, char *end
, unsigned lno
)
1507 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1508 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1510 if (opt
->pre_context
< lno
)
1511 from
= lno
- opt
->pre_context
;
1512 if (from
<= opt
->last_shown
)
1513 from
= opt
->last_shown
+ 1;
1515 if (opt
->funcbody
) {
1516 if (match_funcname(opt
, gs
, bol
, end
))
1519 funcname_needed
= 1;
1520 from
= opt
->last_shown
+ 1;
1524 while (bol
> gs
->buf
&& cur
> from
) {
1525 char *next_bol
= bol
;
1528 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1531 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1532 match_funcname(opt
, gs
, bol
, eol
))) {
1541 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1543 funcname_needed
= 0;
1551 /* We need to look even further back to find a function signature. */
1552 if (opt
->funcname
&& funcname_needed
)
1553 show_funcname_line(opt
, gs
, bol
, cur
);
1557 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1559 while (*eol
!= '\n')
1561 show_line(opt
, bol
, eol
, gs
->name
, cur
, sign
);
1567 static int should_lookahead(struct grep_opt
*opt
)
1572 return 0; /* punt for too complex stuff */
1575 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1576 if (p
->token
!= GREP_PATTERN
)
1577 return 0; /* punt for "header only" and stuff */
1582 static int look_ahead(struct grep_opt
*opt
,
1583 unsigned long *left_p
,
1587 unsigned lno
= *lno_p
;
1590 char *sp
, *last_bol
;
1591 regoff_t earliest
= -1;
1593 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1597 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1598 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1600 if (earliest
< 0 || m
.rm_so
< earliest
)
1605 *bol_p
= bol
+ *left_p
;
1609 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1610 ; /* find the beginning of the line */
1613 for (sp
= bol
; sp
< last_bol
; sp
++) {
1617 *left_p
-= last_bol
- bol
;
1623 static int fill_textconv_grep(struct userdiff_driver
*driver
,
1624 struct grep_source
*gs
)
1626 struct diff_filespec
*df
;
1630 if (!driver
|| !driver
->textconv
)
1631 return grep_source_load(gs
);
1634 * The textconv interface is intimately tied to diff_filespecs, so we
1635 * have to pretend to be one. If we could unify the grep_source
1636 * and diff_filespec structs, this mess could just go away.
1638 df
= alloc_filespec(gs
->path
);
1640 case GREP_SOURCE_OID
:
1641 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1643 case GREP_SOURCE_FILE
:
1644 fill_filespec(df
, &null_oid
, 0, 0100644);
1647 die("BUG: attempt to textconv something without a path?");
1651 * fill_textconv is not remotely thread-safe; it may load objects
1652 * behind the scenes, and it modifies the global diff tempfile
1656 size
= fill_textconv(driver
, df
, &buf
);
1661 * The normal fill_textconv usage by the diff machinery would just keep
1662 * the textconv'd buf separate from the diff_filespec. But much of the
1663 * grep code passes around a grep_source and assumes that its "buf"
1664 * pointer is the beginning of the thing we are searching. So let's
1665 * install our textconv'd version into the grep_source, taking care not
1666 * to leak any existing buffer.
1668 grep_source_clear_data(gs
);
1675 static int is_empty_line(const char *bol
, const char *eol
)
1677 while (bol
< eol
&& isspace(*bol
))
1682 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1685 char *peek_bol
= NULL
;
1688 unsigned last_hit
= 0;
1689 int binary_match_only
= 0;
1691 int try_lookahead
= 0;
1692 int show_function
= 0;
1693 struct userdiff_driver
*textconv
= NULL
;
1694 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1698 opt
->output
= std_output
;
1700 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1702 /* Show hunk marks, except for the first file. */
1703 if (opt
->last_shown
)
1704 opt
->show_hunk_mark
= 1;
1706 * If we're using threads then we can't easily identify
1707 * the first file. Always put hunk marks in that case
1708 * and skip the very first one later in work_done().
1710 if (opt
->output
!= std_output
)
1711 opt
->show_hunk_mark
= 1;
1713 opt
->last_shown
= 0;
1715 if (opt
->allow_textconv
) {
1716 grep_source_load_driver(gs
);
1718 * We might set up the shared textconv cache data here, which
1719 * is not thread-safe.
1722 textconv
= userdiff_get_textconv(gs
->driver
);
1727 * We know the result of a textconv is text, so we only have to care
1728 * about binary handling if we are not using it.
1731 switch (opt
->binary
) {
1732 case GREP_BINARY_DEFAULT
:
1733 if (grep_source_is_binary(gs
))
1734 binary_match_only
= 1;
1736 case GREP_BINARY_NOMATCH
:
1737 if (grep_source_is_binary(gs
))
1738 return 0; /* Assume unmatch */
1740 case GREP_BINARY_TEXT
:
1743 die("BUG: unknown binary handling mode");
1747 memset(&xecfg
, 0, sizeof(xecfg
));
1750 try_lookahead
= should_lookahead(opt
);
1752 if (fill_textconv_grep(textconv
, gs
) < 0)
1762 * look_ahead() skips quickly to the line that possibly
1763 * has the next hit; don't call it if we need to do
1764 * something more than just skipping the current line
1765 * in response to an unmatch for the current line. E.g.
1766 * inside a post-context window, we will show the current
1767 * line as a context around the previous hit when it
1772 && (show_function
||
1773 lno
<= last_hit
+ opt
->post_context
))
1774 && look_ahead(opt
, &left
, &lno
, &bol
))
1776 eol
= end_of_line(bol
, &left
);
1780 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1781 ctx
= GREP_CONTEXT_BODY
;
1783 hit
= match_line(opt
, bol
, eol
, ctx
, collect_hits
);
1789 /* "grep -v -e foo -e bla" should list lines
1790 * that do not have either, so inversion should
1795 if (opt
->unmatch_name_only
) {
1802 if (opt
->status_only
)
1804 if (opt
->name_only
) {
1805 show_name(opt
, gs
->name
);
1810 if (binary_match_only
) {
1811 opt
->output(opt
, "Binary file ", 12);
1812 output_color(opt
, gs
->name
, strlen(gs
->name
),
1813 opt
->colors
[GREP_COLOR_FILENAME
]);
1814 opt
->output(opt
, " matches\n", 9);
1817 /* Hit at this line. If we haven't shown the
1818 * pre-context lines, we would need to show them.
1820 if (opt
->pre_context
|| opt
->funcbody
)
1821 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1822 else if (opt
->funcname
)
1823 show_funcname_line(opt
, gs
, bol
, lno
);
1824 show_line(opt
, bol
, eol
, gs
->name
, lno
, ':');
1830 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1831 unsigned long peek_left
= left
;
1832 char *peek_eol
= eol
;
1835 * Trailing empty lines are not interesting.
1836 * Peek past them to see if they belong to the
1837 * body of the current function.
1840 while (is_empty_line(peek_bol
, peek_eol
)) {
1841 peek_bol
= peek_eol
+ 1;
1842 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1845 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1848 if (show_function
||
1849 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1850 /* If the last hit is within the post context,
1851 * we need to show this line.
1853 show_line(opt
, bol
, eol
, gs
->name
, lno
, '-');
1867 if (opt
->status_only
)
1868 return opt
->unmatch_name_only
;
1869 if (opt
->unmatch_name_only
) {
1870 /* We did not see any hit, so we want to show this */
1871 show_name(opt
, gs
->name
);
1875 xdiff_clear_find_func(&xecfg
);
1879 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1880 * which feels mostly useless but sometimes useful. Maybe
1881 * make it another option? For now suppress them.
1883 if (opt
->count
&& count
) {
1885 if (opt
->pathname
) {
1886 output_color(opt
, gs
->name
, strlen(gs
->name
),
1887 opt
->colors
[GREP_COLOR_FILENAME
]);
1888 output_sep(opt
, ':');
1890 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1891 opt
->output(opt
, buf
, strlen(buf
));
1897 static void clr_hit_marker(struct grep_expr
*x
)
1899 /* All-hit markers are meaningful only at the very top level
1904 if (x
->node
!= GREP_NODE_OR
)
1906 x
->u
.binary
.left
->hit
= 0;
1907 x
= x
->u
.binary
.right
;
1911 static int chk_hit_marker(struct grep_expr
*x
)
1913 /* Top level nodes have hit markers. See if they all are hits */
1915 if (x
->node
!= GREP_NODE_OR
)
1917 if (!x
->u
.binary
.left
->hit
)
1919 x
= x
->u
.binary
.right
;
1923 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1926 * we do not have to do the two-pass grep when we do not check
1927 * buffer-wide "all-match".
1929 if (!opt
->all_match
)
1930 return grep_source_1(opt
, gs
, 0);
1932 /* Otherwise the toplevel "or" terms hit a bit differently.
1933 * We first clear hit markers from them.
1935 clr_hit_marker(opt
->pattern_expression
);
1936 grep_source_1(opt
, gs
, 1);
1938 if (!chk_hit_marker(opt
->pattern_expression
))
1941 return grep_source_1(opt
, gs
, 0);
1944 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
1946 struct grep_source gs
;
1949 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
1953 r
= grep_source(opt
, &gs
);
1955 grep_source_clear(&gs
);
1959 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
1960 const char *name
, const char *path
,
1961 const void *identifier
)
1964 gs
->name
= xstrdup_or_null(name
);
1965 gs
->path
= xstrdup_or_null(path
);
1971 case GREP_SOURCE_FILE
:
1972 gs
->identifier
= xstrdup(identifier
);
1974 case GREP_SOURCE_OID
:
1975 gs
->identifier
= oiddup(identifier
);
1977 case GREP_SOURCE_BUF
:
1978 gs
->identifier
= NULL
;
1983 void grep_source_clear(struct grep_source
*gs
)
1985 FREE_AND_NULL(gs
->name
);
1986 FREE_AND_NULL(gs
->path
);
1987 FREE_AND_NULL(gs
->identifier
);
1988 grep_source_clear_data(gs
);
1991 void grep_source_clear_data(struct grep_source
*gs
)
1994 case GREP_SOURCE_FILE
:
1995 case GREP_SOURCE_OID
:
1996 FREE_AND_NULL(gs
->buf
);
1999 case GREP_SOURCE_BUF
:
2000 /* leave user-provided buf intact */
2005 static int grep_source_load_oid(struct grep_source
*gs
)
2007 enum object_type type
;
2010 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2014 return error(_("'%s': unable to read %s"),
2016 oid_to_hex(gs
->identifier
));
2020 static int grep_source_load_file(struct grep_source
*gs
)
2022 const char *filename
= gs
->identifier
;
2028 if (lstat(filename
, &st
) < 0) {
2030 if (errno
!= ENOENT
)
2031 error_errno(_("failed to stat '%s'"), filename
);
2034 if (!S_ISREG(st
.st_mode
))
2036 size
= xsize_t(st
.st_size
);
2037 i
= open(filename
, O_RDONLY
);
2040 data
= xmallocz(size
);
2041 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2042 error_errno(_("'%s': short read"), filename
);
2054 static int grep_source_load(struct grep_source
*gs
)
2060 case GREP_SOURCE_FILE
:
2061 return grep_source_load_file(gs
);
2062 case GREP_SOURCE_OID
:
2063 return grep_source_load_oid(gs
);
2064 case GREP_SOURCE_BUF
:
2065 return gs
->buf
? 0 : -1;
2067 die("BUG: invalid grep_source type to load");
2070 void grep_source_load_driver(struct grep_source
*gs
)
2077 gs
->driver
= userdiff_find_by_path(gs
->path
);
2079 gs
->driver
= userdiff_find_by_name("default");
2083 static int grep_source_is_binary(struct grep_source
*gs
)
2085 grep_source_load_driver(gs
);
2086 if (gs
->driver
->binary
!= -1)
2087 return gs
->driver
->binary
;
2089 if (!grep_source_load(gs
))
2090 return buffer_is_binary(gs
->buf
, gs
->size
);