4 #include "object-store.h"
6 #include "xdiff-interface.h"
13 static int grep_source_load(struct grep_source
*gs
);
14 static int grep_source_is_binary(struct grep_source
*gs
,
15 struct index_state
*istate
);
17 static struct grep_opt grep_defaults
;
19 static const char *color_grep_slots
[] = {
20 [GREP_COLOR_CONTEXT
] = "context",
21 [GREP_COLOR_FILENAME
] = "filename",
22 [GREP_COLOR_FUNCTION
] = "function",
23 [GREP_COLOR_LINENO
] = "lineNumber",
24 [GREP_COLOR_COLUMNNO
] = "column",
25 [GREP_COLOR_MATCH_CONTEXT
] = "matchContext",
26 [GREP_COLOR_MATCH_SELECTED
] = "matchSelected",
27 [GREP_COLOR_SELECTED
] = "selected",
28 [GREP_COLOR_SEP
] = "separator",
31 static void std_output(struct grep_opt
*opt
, const void *buf
, size_t size
)
33 fwrite(buf
, size
, 1, stdout
);
36 static void color_set(char *dst
, const char *color_bytes
)
38 xsnprintf(dst
, COLOR_MAXLEN
, "%s", color_bytes
);
42 * Initialize the grep_defaults template with hardcoded defaults.
43 * We could let the compiler do this, but without C99 initializers
44 * the code gets unwieldy and unreadable, so...
46 void init_grep_defaults(struct repository
*repo
)
48 struct grep_opt
*opt
= &grep_defaults
;
55 memset(opt
, 0, sizeof(*opt
));
60 opt
->pattern_type_option
= GREP_PATTERN_TYPE_UNSPECIFIED
;
61 color_set(opt
->colors
[GREP_COLOR_CONTEXT
], "");
62 color_set(opt
->colors
[GREP_COLOR_FILENAME
], "");
63 color_set(opt
->colors
[GREP_COLOR_FUNCTION
], "");
64 color_set(opt
->colors
[GREP_COLOR_LINENO
], "");
65 color_set(opt
->colors
[GREP_COLOR_COLUMNNO
], "");
66 color_set(opt
->colors
[GREP_COLOR_MATCH_CONTEXT
], GIT_COLOR_BOLD_RED
);
67 color_set(opt
->colors
[GREP_COLOR_MATCH_SELECTED
], GIT_COLOR_BOLD_RED
);
68 color_set(opt
->colors
[GREP_COLOR_SELECTED
], "");
69 color_set(opt
->colors
[GREP_COLOR_SEP
], GIT_COLOR_CYAN
);
70 opt
->only_matching
= 0;
72 opt
->output
= std_output
;
75 static int parse_pattern_type_arg(const char *opt
, const char *arg
)
77 if (!strcmp(arg
, "default"))
78 return GREP_PATTERN_TYPE_UNSPECIFIED
;
79 else if (!strcmp(arg
, "basic"))
80 return GREP_PATTERN_TYPE_BRE
;
81 else if (!strcmp(arg
, "extended"))
82 return GREP_PATTERN_TYPE_ERE
;
83 else if (!strcmp(arg
, "fixed"))
84 return GREP_PATTERN_TYPE_FIXED
;
85 else if (!strcmp(arg
, "perl"))
86 return GREP_PATTERN_TYPE_PCRE
;
87 die("bad %s argument: %s", opt
, arg
);
90 define_list_config_array_extra(color_grep_slots
, {"match"});
93 * Read the configuration file once and store it in
94 * the grep_defaults template.
96 int grep_config(const char *var
, const char *value
, void *cb
)
98 struct grep_opt
*opt
= &grep_defaults
;
101 if (userdiff_config(var
, value
) < 0)
104 if (!strcmp(var
, "grep.extendedregexp")) {
105 opt
->extended_regexp_option
= git_config_bool(var
, value
);
109 if (!strcmp(var
, "grep.patterntype")) {
110 opt
->pattern_type_option
= parse_pattern_type_arg(var
, value
);
114 if (!strcmp(var
, "grep.linenumber")) {
115 opt
->linenum
= git_config_bool(var
, value
);
118 if (!strcmp(var
, "grep.column")) {
119 opt
->columnnum
= git_config_bool(var
, value
);
123 if (!strcmp(var
, "grep.fullname")) {
124 opt
->relative
= !git_config_bool(var
, value
);
128 if (!strcmp(var
, "color.grep"))
129 opt
->color
= git_config_colorbool(var
, value
);
130 if (!strcmp(var
, "color.grep.match")) {
131 if (grep_config("color.grep.matchcontext", value
, cb
) < 0)
133 if (grep_config("color.grep.matchselected", value
, cb
) < 0)
135 } else if (skip_prefix(var
, "color.grep.", &slot
)) {
136 int i
= LOOKUP_CONFIG(color_grep_slots
, slot
);
141 color
= opt
->colors
[i
];
143 return config_error_nonbool(var
);
144 return color_parse(value
, color
);
150 * Initialize one instance of grep_opt and copy the
151 * default values from the template we read the configuration
152 * information in an earlier call to git_config(grep_config).
154 void grep_init(struct grep_opt
*opt
, struct repository
*repo
, const char *prefix
)
156 struct grep_opt
*def
= &grep_defaults
;
159 memset(opt
, 0, sizeof(*opt
));
161 opt
->prefix
= prefix
;
162 opt
->prefix_length
= (prefix
&& *prefix
) ? strlen(prefix
) : 0;
163 opt
->pattern_tail
= &opt
->pattern_list
;
164 opt
->header_tail
= &opt
->header_list
;
166 opt
->only_matching
= def
->only_matching
;
167 opt
->color
= def
->color
;
168 opt
->extended_regexp_option
= def
->extended_regexp_option
;
169 opt
->pattern_type_option
= def
->pattern_type_option
;
170 opt
->linenum
= def
->linenum
;
171 opt
->columnnum
= def
->columnnum
;
172 opt
->max_depth
= def
->max_depth
;
173 opt
->pathname
= def
->pathname
;
174 opt
->relative
= def
->relative
;
175 opt
->output
= def
->output
;
177 for (i
= 0; i
< NR_GREP_COLORS
; i
++)
178 color_set(opt
->colors
[i
], def
->colors
[i
]);
181 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
184 * When committing to the pattern type by setting the relevant
185 * fields in grep_opt it's generally not necessary to zero out
186 * the fields we're not choosing, since they won't have been
187 * set by anything. The extended_regexp_option field is the
188 * only exception to this.
190 * This is because in the process of parsing grep.patternType
191 * & grep.extendedRegexp we set opt->pattern_type_option and
192 * opt->extended_regexp_option, respectively. We then
193 * internally use opt->extended_regexp_option to see if we're
194 * compiling an ERE. It must be unset if that's not actually
197 if (pattern_type
!= GREP_PATTERN_TYPE_ERE
&&
198 opt
->extended_regexp_option
)
199 opt
->extended_regexp_option
= 0;
201 switch (pattern_type
) {
202 case GREP_PATTERN_TYPE_UNSPECIFIED
:
205 case GREP_PATTERN_TYPE_BRE
:
208 case GREP_PATTERN_TYPE_ERE
:
209 opt
->extended_regexp_option
= 1;
212 case GREP_PATTERN_TYPE_FIXED
:
216 case GREP_PATTERN_TYPE_PCRE
:
221 * It's important that pcre1 always be assigned to
222 * even when there's no USE_LIBPCRE* defined. We still
223 * call the PCRE stub function, it just dies with
224 * "cannot use Perl-compatible regexes[...]".
232 void grep_commit_pattern_type(enum grep_pattern_type pattern_type
, struct grep_opt
*opt
)
234 if (pattern_type
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
235 grep_set_pattern_type_option(pattern_type
, opt
);
236 else if (opt
->pattern_type_option
!= GREP_PATTERN_TYPE_UNSPECIFIED
)
237 grep_set_pattern_type_option(opt
->pattern_type_option
, opt
);
238 else if (opt
->extended_regexp_option
)
240 * This branch *must* happen after setting from the
241 * opt->pattern_type_option above, we don't want
242 * grep.extendedRegexp to override grep.patternType!
244 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE
, opt
);
247 static struct grep_pat
*create_grep_pat(const char *pat
, size_t patlen
,
248 const char *origin
, int no
,
249 enum grep_pat_token t
,
250 enum grep_header_field field
)
252 struct grep_pat
*p
= xcalloc(1, sizeof(*p
));
253 p
->pattern
= xmemdupz(pat
, patlen
);
254 p
->patternlen
= patlen
;
262 static void do_append_grep_pat(struct grep_pat
***tail
, struct grep_pat
*p
)
269 case GREP_PATTERN
: /* atom */
270 case GREP_PATTERN_HEAD
:
271 case GREP_PATTERN_BODY
:
273 struct grep_pat
*new_pat
;
275 char *cp
= p
->pattern
+ p
->patternlen
, *nl
= NULL
;
276 while (++len
<= p
->patternlen
) {
277 if (*(--cp
) == '\n') {
284 new_pat
= create_grep_pat(nl
+ 1, len
- 1, p
->origin
,
285 p
->no
, p
->token
, p
->field
);
286 new_pat
->next
= p
->next
;
288 *tail
= &new_pat
->next
;
291 p
->patternlen
-= len
;
299 void append_header_grep_pattern(struct grep_opt
*opt
,
300 enum grep_header_field field
, const char *pat
)
302 struct grep_pat
*p
= create_grep_pat(pat
, strlen(pat
), "header", 0,
303 GREP_PATTERN_HEAD
, field
);
304 if (field
== GREP_HEADER_REFLOG
)
305 opt
->use_reflog_filter
= 1;
306 do_append_grep_pat(&opt
->header_tail
, p
);
309 void append_grep_pattern(struct grep_opt
*opt
, const char *pat
,
310 const char *origin
, int no
, enum grep_pat_token t
)
312 append_grep_pat(opt
, pat
, strlen(pat
), origin
, no
, t
);
315 void append_grep_pat(struct grep_opt
*opt
, const char *pat
, size_t patlen
,
316 const char *origin
, int no
, enum grep_pat_token t
)
318 struct grep_pat
*p
= create_grep_pat(pat
, patlen
, origin
, no
, t
, 0);
319 do_append_grep_pat(&opt
->pattern_tail
, p
);
322 struct grep_opt
*grep_opt_dup(const struct grep_opt
*opt
)
324 struct grep_pat
*pat
;
325 struct grep_opt
*ret
= xmalloc(sizeof(struct grep_opt
));
328 ret
->pattern_list
= NULL
;
329 ret
->pattern_tail
= &ret
->pattern_list
;
331 for(pat
= opt
->pattern_list
; pat
!= NULL
; pat
= pat
->next
)
333 if(pat
->token
== GREP_PATTERN_HEAD
)
334 append_header_grep_pattern(ret
, pat
->field
,
337 append_grep_pat(ret
, pat
->pattern
, pat
->patternlen
,
338 pat
->origin
, pat
->no
, pat
->token
);
344 static NORETURN
void compile_regexp_failed(const struct grep_pat
*p
,
350 xsnprintf(where
, sizeof(where
), "In '%s' at %d, ", p
->origin
, p
->no
);
352 xsnprintf(where
, sizeof(where
), "%s, ", p
->origin
);
356 die("%s'%s': %s", where
, p
->pattern
, error
);
359 static int is_fixed(const char *s
, size_t len
)
363 for (i
= 0; i
< len
; i
++) {
364 if (is_regex_special(s
[i
]))
371 static int has_null(const char *s
, size_t len
)
374 * regcomp cannot accept patterns with NULs so when using it
375 * we consider any pattern containing a NUL fixed.
377 if (memchr(s
, 0, len
))
384 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
388 int options
= PCRE_MULTILINE
;
390 if (opt
->ignore_case
) {
391 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
))
392 p
->pcre1_tables
= pcre_maketables();
393 options
|= PCRE_CASELESS
;
395 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
))
396 options
|= PCRE_UTF8
;
398 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
400 if (!p
->pcre1_regexp
)
401 compile_regexp_failed(p
, error
);
403 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, GIT_PCRE_STUDY_JIT_COMPILE
, &error
);
404 if (!p
->pcre1_extra_info
&& error
)
407 #ifdef GIT_PCRE1_USE_JIT
408 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
409 if (p
->pcre1_jit_on
== 1) {
410 p
->pcre1_jit_stack
= pcre_jit_stack_alloc(1, 1024 * 1024);
411 if (!p
->pcre1_jit_stack
)
412 die("Couldn't allocate PCRE JIT stack");
413 pcre_assign_jit_stack(p
->pcre1_extra_info
, NULL
, p
->pcre1_jit_stack
);
414 } else if (p
->pcre1_jit_on
!= 0) {
415 BUG("The pcre1_jit_on variable should be 0 or 1, not %d",
421 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
422 regmatch_t
*match
, int eflags
)
424 int ovector
[30], ret
, flags
= 0;
426 if (eflags
& REG_NOTBOL
)
427 flags
|= PCRE_NOTBOL
;
429 #ifdef GIT_PCRE1_USE_JIT
430 if (p
->pcre1_jit_on
) {
431 ret
= pcre_jit_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
432 eol
- line
, 0, flags
, ovector
,
433 ARRAY_SIZE(ovector
), p
->pcre1_jit_stack
);
437 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
438 eol
- line
, 0, flags
, ovector
,
439 ARRAY_SIZE(ovector
));
442 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
443 die("pcre_exec failed with error code %d", ret
);
446 match
->rm_so
= ovector
[0];
447 match
->rm_eo
= ovector
[1];
453 static void free_pcre1_regexp(struct grep_pat
*p
)
455 pcre_free(p
->pcre1_regexp
);
456 #ifdef GIT_PCRE1_USE_JIT
457 if (p
->pcre1_jit_on
) {
458 pcre_free_study(p
->pcre1_extra_info
);
459 pcre_jit_stack_free(p
->pcre1_jit_stack
);
463 pcre_free(p
->pcre1_extra_info
);
465 pcre_free((void *)p
->pcre1_tables
);
467 #else /* !USE_LIBPCRE1 */
468 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
470 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
473 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
474 regmatch_t
*match
, int eflags
)
479 static void free_pcre1_regexp(struct grep_pat
*p
)
482 #endif /* !USE_LIBPCRE1 */
485 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
488 PCRE2_UCHAR errbuf
[256];
489 PCRE2_SIZE erroffset
;
490 int options
= PCRE2_MULTILINE
;
491 const uint8_t *character_tables
= NULL
;
498 p
->pcre2_compile_context
= NULL
;
500 if (opt
->ignore_case
) {
501 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
)) {
502 character_tables
= pcre2_maketables(NULL
);
503 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
504 pcre2_set_character_tables(p
->pcre2_compile_context
, character_tables
);
506 options
|= PCRE2_CASELESS
;
508 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
))
509 options
|= PCRE2_UTF
;
511 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
512 p
->patternlen
, options
, &error
, &erroffset
,
513 p
->pcre2_compile_context
);
515 if (p
->pcre2_pattern
) {
516 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, NULL
);
517 if (!p
->pcre2_match_data
)
518 die("Couldn't allocate PCRE2 match data");
520 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
521 compile_regexp_failed(p
, (const char *)&errbuf
);
524 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
525 if (p
->pcre2_jit_on
== 1) {
526 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
528 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
531 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
532 * tells us whether the library itself supports JIT,
533 * but to see whether we're going to be actually using
534 * JIT we need to extract PCRE2_INFO_JITSIZE from the
535 * pattern *after* we do pcre2_jit_compile() above.
537 * This is because if the pattern contains the
538 * (*NO_JIT) verb (see pcre2syntax(3))
539 * pcre2_jit_compile() will exit early with 0. If we
540 * then proceed to call pcre2_jit_match() further down
541 * the line instead of pcre2_match() we'll either
542 * segfault (pre PCRE 10.31) or run into a fatal error
545 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
547 BUG("pcre2_pattern_info() failed: %d", patinforet
);
548 if (jitsizearg
== 0) {
553 p
->pcre2_jit_stack
= pcre2_jit_stack_create(1, 1024 * 1024, NULL
);
554 if (!p
->pcre2_jit_stack
)
555 die("Couldn't allocate PCRE2 JIT stack");
556 p
->pcre2_match_context
= pcre2_match_context_create(NULL
);
557 if (!p
->pcre2_match_context
)
558 die("Couldn't allocate PCRE2 match context");
559 pcre2_jit_stack_assign(p
->pcre2_match_context
, NULL
, p
->pcre2_jit_stack
);
560 } else if (p
->pcre2_jit_on
!= 0) {
561 BUG("The pcre2_jit_on variable should be 0 or 1, not %d",
566 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
567 regmatch_t
*match
, int eflags
)
571 PCRE2_UCHAR errbuf
[256];
573 if (eflags
& REG_NOTBOL
)
574 flags
|= PCRE2_NOTBOL
;
577 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
578 eol
- line
, 0, flags
, p
->pcre2_match_data
,
581 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
582 eol
- line
, 0, flags
, p
->pcre2_match_data
,
585 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
586 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
587 die("%s failed with error code %d: %s",
588 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
592 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
594 match
->rm_so
= (int)ovector
[0];
595 match
->rm_eo
= (int)ovector
[1];
601 static void free_pcre2_pattern(struct grep_pat
*p
)
603 pcre2_compile_context_free(p
->pcre2_compile_context
);
604 pcre2_code_free(p
->pcre2_pattern
);
605 pcre2_match_data_free(p
->pcre2_match_data
);
606 pcre2_jit_stack_free(p
->pcre2_jit_stack
);
607 pcre2_match_context_free(p
->pcre2_match_context
);
609 #else /* !USE_LIBPCRE2 */
610 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
613 * Unreachable until USE_LIBPCRE2 becomes synonymous with
614 * USE_LIBPCRE. See the sibling comment in
615 * grep_set_pattern_type_option().
617 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
620 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
621 regmatch_t
*match
, int eflags
)
626 static void free_pcre2_pattern(struct grep_pat
*p
)
629 #endif /* !USE_LIBPCRE2 */
631 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
633 struct strbuf sb
= STRBUF_INIT
;
637 basic_regex_quote_buf(&sb
, p
->pattern
);
638 if (opt
->ignore_case
)
639 regflags
|= REG_ICASE
;
640 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
642 fprintf(stderr
, "fixed %s\n", sb
.buf
);
646 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
647 compile_regexp_failed(p
, errbuf
);
651 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
654 int regflags
= REG_NEWLINE
;
656 p
->word_regexp
= opt
->word_regexp
;
657 p
->ignore_case
= opt
->ignore_case
;
660 * Even when -F (fixed) asks us to do a non-regexp search, we
661 * may not be able to correctly case-fold when -i
662 * (ignore-case) is asked (in which case, we'll synthesize a
663 * regexp to match the pattern that matches regexp special
664 * characters literally, while ignoring case differences). On
665 * the other hand, even without -F, if the pattern does not
666 * have any regexp special characters and there is no need for
667 * case-folding search, we can internally turn it into a
668 * simple string match using kws. p->fixed tells us if we
672 has_null(p
->pattern
, p
->patternlen
) ||
673 is_fixed(p
->pattern
, p
->patternlen
))
674 p
->fixed
= !p
->ignore_case
|| !has_non_ascii(p
->pattern
);
677 p
->kws
= kwsalloc(p
->ignore_case
? tolower_trans_tbl
: NULL
);
678 kwsincr(p
->kws
, p
->pattern
, p
->patternlen
);
681 } else if (opt
->fixed
) {
683 * We come here when the pattern has the non-ascii
684 * characters we cannot case-fold, and asked to
687 compile_fixed_regexp(p
, opt
);
692 compile_pcre2_pattern(p
, opt
);
697 compile_pcre1_regexp(p
, opt
);
702 regflags
|= REG_ICASE
;
703 if (opt
->extended_regexp_option
)
704 regflags
|= REG_EXTENDED
;
705 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
708 regerror(err
, &p
->regexp
, errbuf
, 1024);
709 compile_regexp_failed(p
, errbuf
);
713 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
714 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
723 case GREP_PATTERN
: /* atom */
724 case GREP_PATTERN_HEAD
:
725 case GREP_PATTERN_BODY
:
726 x
= xcalloc(1, sizeof (struct grep_expr
));
727 x
->node
= GREP_NODE_ATOM
;
731 case GREP_OPEN_PAREN
:
733 x
= compile_pattern_or(list
);
734 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
735 die("unmatched parenthesis");
736 *list
= (*list
)->next
;
743 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
754 die("--not not followed by pattern expression");
756 x
= xcalloc(1, sizeof (struct grep_expr
));
757 x
->node
= GREP_NODE_NOT
;
758 x
->u
.unary
= compile_pattern_not(list
);
760 die("--not followed by non pattern expression");
763 return compile_pattern_atom(list
);
767 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
770 struct grep_expr
*x
, *y
, *z
;
772 x
= compile_pattern_not(list
);
774 if (p
&& p
->token
== GREP_AND
) {
776 die("--and not followed by pattern expression");
778 y
= compile_pattern_and(list
);
780 die("--and not followed by pattern expression");
781 z
= xcalloc(1, sizeof (struct grep_expr
));
782 z
->node
= GREP_NODE_AND
;
783 z
->u
.binary
.left
= x
;
784 z
->u
.binary
.right
= y
;
790 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
793 struct grep_expr
*x
, *y
, *z
;
795 x
= compile_pattern_and(list
);
797 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
798 y
= compile_pattern_or(list
);
800 die("not a pattern expression %s", p
->pattern
);
801 z
= xcalloc(1, sizeof (struct grep_expr
));
802 z
->node
= GREP_NODE_OR
;
803 z
->u
.binary
.left
= x
;
804 z
->u
.binary
.right
= y
;
810 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
812 return compile_pattern_or(list
);
815 static void indent(int in
)
821 static void dump_grep_pat(struct grep_pat
*p
)
824 case GREP_AND
: fprintf(stderr
, "*and*"); break;
825 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
826 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
827 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
828 case GREP_OR
: fprintf(stderr
, "*or*"); break;
830 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
831 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
832 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
837 case GREP_PATTERN_HEAD
:
838 fprintf(stderr
, "<head %d>", p
->field
); break;
839 case GREP_PATTERN_BODY
:
840 fprintf(stderr
, "<body>"); break;
844 case GREP_PATTERN_HEAD
:
845 case GREP_PATTERN_BODY
:
847 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
853 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
858 fprintf(stderr
, "true\n");
861 dump_grep_pat(x
->u
.atom
);
864 fprintf(stderr
, "(not\n");
865 dump_grep_expression_1(x
->u
.unary
, in
+1);
867 fprintf(stderr
, ")\n");
870 fprintf(stderr
, "(and\n");
871 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
872 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
874 fprintf(stderr
, ")\n");
877 fprintf(stderr
, "(or\n");
878 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
879 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
881 fprintf(stderr
, ")\n");
886 static void dump_grep_expression(struct grep_opt
*opt
)
888 struct grep_expr
*x
= opt
->pattern_expression
;
891 fprintf(stderr
, "[all-match]\n");
892 dump_grep_expression_1(x
, 0);
896 static struct grep_expr
*grep_true_expr(void)
898 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
899 z
->node
= GREP_NODE_TRUE
;
903 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
905 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
906 z
->node
= GREP_NODE_OR
;
907 z
->u
.binary
.left
= left
;
908 z
->u
.binary
.right
= right
;
912 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
915 struct grep_expr
*header_expr
;
916 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
917 enum grep_header_field fld
;
919 if (!opt
->header_list
)
922 for (p
= opt
->header_list
; p
; p
= p
->next
) {
923 if (p
->token
!= GREP_PATTERN_HEAD
)
924 BUG("a non-header pattern in grep header list.");
925 if (p
->field
< GREP_HEADER_FIELD_MIN
||
926 GREP_HEADER_FIELD_MAX
<= p
->field
)
927 BUG("unknown header field %d", p
->field
);
928 compile_regexp(p
, opt
);
931 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
932 header_group
[fld
] = NULL
;
934 for (p
= opt
->header_list
; p
; p
= p
->next
) {
936 struct grep_pat
*pp
= p
;
938 h
= compile_pattern_atom(&pp
);
939 if (!h
|| pp
!= p
->next
)
940 BUG("malformed header expr");
941 if (!header_group
[p
->field
]) {
942 header_group
[p
->field
] = h
;
945 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
950 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
951 if (!header_group
[fld
])
954 header_expr
= grep_true_expr();
955 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
960 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
962 struct grep_expr
*z
= x
;
965 assert(x
->node
== GREP_NODE_OR
);
966 if (x
->u
.binary
.right
&&
967 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
968 x
->u
.binary
.right
= y
;
971 x
= x
->u
.binary
.right
;
976 static void compile_grep_patterns_real(struct grep_opt
*opt
)
979 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
981 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
983 case GREP_PATTERN
: /* atom */
984 case GREP_PATTERN_HEAD
:
985 case GREP_PATTERN_BODY
:
986 compile_regexp(p
, opt
);
994 if (opt
->all_match
|| header_expr
)
996 else if (!opt
->extended
&& !opt
->debug
)
999 p
= opt
->pattern_list
;
1001 opt
->pattern_expression
= compile_pattern_expr(&p
);
1003 die("incomplete pattern expression: %s", p
->pattern
);
1008 if (!opt
->pattern_expression
)
1009 opt
->pattern_expression
= header_expr
;
1010 else if (opt
->all_match
)
1011 opt
->pattern_expression
= grep_splice_or(header_expr
,
1012 opt
->pattern_expression
);
1014 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
1019 void compile_grep_patterns(struct grep_opt
*opt
)
1021 compile_grep_patterns_real(opt
);
1023 dump_grep_expression(opt
);
1026 static void free_pattern_expr(struct grep_expr
*x
)
1029 case GREP_NODE_TRUE
:
1030 case GREP_NODE_ATOM
:
1033 free_pattern_expr(x
->u
.unary
);
1037 free_pattern_expr(x
->u
.binary
.left
);
1038 free_pattern_expr(x
->u
.binary
.right
);
1044 void free_grep_patterns(struct grep_opt
*opt
)
1046 struct grep_pat
*p
, *n
;
1048 for (p
= opt
->pattern_list
; p
; p
= n
) {
1051 case GREP_PATTERN
: /* atom */
1052 case GREP_PATTERN_HEAD
:
1053 case GREP_PATTERN_BODY
:
1056 else if (p
->pcre1_regexp
)
1057 free_pcre1_regexp(p
);
1058 else if (p
->pcre2_pattern
)
1059 free_pcre2_pattern(p
);
1061 regfree(&p
->regexp
);
1072 free_pattern_expr(opt
->pattern_expression
);
1075 static char *end_of_line(char *cp
, unsigned long *left
)
1077 unsigned long l
= *left
;
1078 while (l
&& *cp
!= '\n') {
1086 static int word_char(char ch
)
1088 return isalnum(ch
) || ch
== '_';
1091 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1094 if (want_color(opt
->color
) && color
&& color
[0]) {
1095 opt
->output(opt
, color
, strlen(color
));
1096 opt
->output(opt
, data
, size
);
1097 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1099 opt
->output(opt
, data
, size
);
1102 static void output_sep(struct grep_opt
*opt
, char sign
)
1104 if (opt
->null_following_name
)
1105 opt
->output(opt
, "\0", 1);
1107 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
1110 static void show_name(struct grep_opt
*opt
, const char *name
)
1112 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1113 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1116 static int fixmatch(struct grep_pat
*p
, char *line
, char *eol
,
1119 struct kwsmatch kwsm
;
1120 size_t offset
= kwsexec(p
->kws
, line
, eol
- line
, &kwsm
);
1122 match
->rm_so
= match
->rm_eo
= -1;
1125 match
->rm_so
= offset
;
1126 match
->rm_eo
= match
->rm_so
+ kwsm
.size
[0];
1131 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1132 regmatch_t
*match
, int eflags
)
1137 hit
= !fixmatch(p
, line
, eol
, match
);
1138 else if (p
->pcre1_regexp
)
1139 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1140 else if (p
->pcre2_pattern
)
1141 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1143 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1149 static int strip_timestamp(char *bol
, char **eol_p
)
1154 while (bol
< --eol
) {
1168 } header_field
[] = {
1170 { "committer ", 10 },
1174 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1175 enum grep_context ctx
,
1176 regmatch_t
*pmatch
, int eflags
)
1180 const char *start
= bol
;
1182 if ((p
->token
!= GREP_PATTERN
) &&
1183 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1186 if (p
->token
== GREP_PATTERN_HEAD
) {
1189 assert(p
->field
< ARRAY_SIZE(header_field
));
1190 field
= header_field
[p
->field
].field
;
1191 len
= header_field
[p
->field
].len
;
1192 if (strncmp(bol
, field
, len
))
1196 case GREP_HEADER_AUTHOR
:
1197 case GREP_HEADER_COMMITTER
:
1198 saved_ch
= strip_timestamp(bol
, &eol
);
1206 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1208 if (hit
&& p
->word_regexp
) {
1209 if ((pmatch
[0].rm_so
< 0) ||
1210 (eol
- bol
) < pmatch
[0].rm_so
||
1211 (pmatch
[0].rm_eo
< 0) ||
1212 (eol
- bol
) < pmatch
[0].rm_eo
)
1213 die("regexp returned nonsense");
1215 /* Match beginning must be either beginning of the
1216 * line, or at word boundary (i.e. the last char must
1217 * not be a word char). Similarly, match end must be
1218 * either end of the line, or at word boundary
1219 * (i.e. the next char must not be a word char).
1221 if ( ((pmatch
[0].rm_so
== 0) ||
1222 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1223 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1224 !word_char(bol
[pmatch
[0].rm_eo
])) )
1229 /* Words consist of at least one character. */
1230 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1233 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1234 /* There could be more than one match on the
1235 * line, and the first match might not be
1236 * strict word match. But later ones could be!
1237 * Forward to the next possible start, i.e. the
1238 * next position following a non-word char.
1240 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1241 while (word_char(bol
[-1]) && bol
< eol
)
1243 eflags
|= REG_NOTBOL
;
1248 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1251 pmatch
[0].rm_so
+= bol
- start
;
1252 pmatch
[0].rm_eo
+= bol
- start
;
1257 static int match_expr_eval(struct grep_opt
*opt
, struct grep_expr
*x
, char *bol
,
1258 char *eol
, enum grep_context ctx
, ssize_t
*col
,
1259 ssize_t
*icol
, int collect_hits
)
1264 die("Not a valid grep expression");
1266 case GREP_NODE_TRUE
:
1269 case GREP_NODE_ATOM
:
1272 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
,
1274 if (h
&& (*col
< 0 || tmp
.rm_so
< *col
))
1280 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1282 h
= !match_expr_eval(opt
, x
->u
.unary
, bol
, eol
, ctx
, icol
, col
,
1286 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1288 if (h
|| opt
->columnnum
) {
1290 * Don't short-circuit AND when given --column, since a
1291 * NOT earlier in the tree may turn this into an OR. In
1292 * this case, see the below comment.
1294 h
&= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
,
1299 if (!(collect_hits
|| opt
->columnnum
)) {
1301 * Don't short-circuit OR when given --column (or
1302 * collecting hits) to ensure we don't skip a later
1303 * child that would produce an earlier match.
1305 return (match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
,
1306 ctx
, col
, icol
, 0) ||
1307 match_expr_eval(opt
, x
->u
.binary
.right
, bol
,
1308 eol
, ctx
, col
, icol
, 0));
1310 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1313 x
->u
.binary
.left
->hit
|= h
;
1314 h
|= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
, ctx
, col
,
1315 icol
, collect_hits
);
1318 die("Unexpected node type (internal error) %d", x
->node
);
1325 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1326 enum grep_context ctx
, ssize_t
*col
,
1327 ssize_t
*icol
, int collect_hits
)
1329 struct grep_expr
*x
= opt
->pattern_expression
;
1330 return match_expr_eval(opt
, x
, bol
, eol
, ctx
, col
, icol
, collect_hits
);
1333 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1334 ssize_t
*col
, ssize_t
*icol
,
1335 enum grep_context ctx
, int collect_hits
)
1341 return match_expr(opt
, bol
, eol
, ctx
, col
, icol
,
1344 /* we do not call with collect_hits without being extended */
1345 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1347 if (match_one_pattern(p
, bol
, eol
, ctx
, &tmp
, 0)) {
1349 if (!opt
->columnnum
) {
1351 * Without --column, any single match on a line
1352 * is enough to know that it needs to be
1353 * printed. With --column, scan _all_ patterns
1354 * to find the earliest.
1358 if (*col
< 0 || tmp
.rm_so
< *col
)
1365 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1366 enum grep_context ctx
,
1367 regmatch_t
*pmatch
, int eflags
)
1371 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1373 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1375 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1376 if (match
.rm_so
> pmatch
->rm_so
)
1378 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1381 pmatch
->rm_so
= match
.rm_so
;
1382 pmatch
->rm_eo
= match
.rm_eo
;
1386 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1387 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1392 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1394 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1396 case GREP_PATTERN
: /* atom */
1397 case GREP_PATTERN_HEAD
:
1398 case GREP_PATTERN_BODY
:
1399 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1410 static void show_line_header(struct grep_opt
*opt
, const char *name
,
1411 unsigned lno
, ssize_t cno
, char sign
)
1413 if (opt
->heading
&& opt
->last_shown
== 0) {
1414 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1415 opt
->output(opt
, "\n", 1);
1417 opt
->last_shown
= lno
;
1419 if (!opt
->heading
&& opt
->pathname
) {
1420 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1421 output_sep(opt
, sign
);
1425 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1426 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1427 output_sep(opt
, sign
);
1430 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1431 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1432 * being called with a context line.
1434 if (opt
->columnnum
&& cno
) {
1436 xsnprintf(buf
, sizeof(buf
), "%"PRIuMAX
, (uintmax_t)cno
);
1437 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_COLUMNNO
]);
1438 output_sep(opt
, sign
);
1442 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1443 const char *name
, unsigned lno
, ssize_t cno
, char sign
)
1445 int rest
= eol
- bol
;
1446 const char *match_color
= NULL
;
1447 const char *line_color
= NULL
;
1449 if (opt
->file_break
&& opt
->last_shown
== 0) {
1450 if (opt
->show_hunk_mark
)
1451 opt
->output(opt
, "\n", 1);
1452 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1453 if (opt
->last_shown
== 0) {
1454 if (opt
->show_hunk_mark
) {
1455 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1456 opt
->output(opt
, "\n", 1);
1458 } else if (lno
> opt
->last_shown
+ 1) {
1459 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1460 opt
->output(opt
, "\n", 1);
1463 if (!opt
->only_matching
) {
1465 * In case the line we're being called with contains more than
1466 * one match, leave printing each header to the loop below.
1468 show_line_header(opt
, name
, lno
, cno
, sign
);
1470 if (opt
->color
|| opt
->only_matching
) {
1472 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1478 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1480 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1482 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1483 else if (sign
== '-')
1484 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1485 else if (sign
== '=')
1486 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1489 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1490 if (match
.rm_so
== match
.rm_eo
)
1493 if (opt
->only_matching
)
1494 show_line_header(opt
, name
, lno
, cno
, sign
);
1496 output_color(opt
, bol
, match
.rm_so
, line_color
);
1497 output_color(opt
, bol
+ match
.rm_so
,
1498 match
.rm_eo
- match
.rm_so
, match_color
);
1499 if (opt
->only_matching
)
1500 opt
->output(opt
, "\n", 1);
1503 rest
-= match
.rm_eo
;
1504 eflags
= REG_NOTBOL
;
1508 if (!opt
->only_matching
) {
1509 output_color(opt
, bol
, rest
, line_color
);
1510 opt
->output(opt
, "\n", 1);
1517 * This lock protects access to the gitattributes machinery, which is
1520 pthread_mutex_t grep_attr_mutex
;
1522 static inline void grep_attr_lock(void)
1525 pthread_mutex_lock(&grep_attr_mutex
);
1528 static inline void grep_attr_unlock(void)
1531 pthread_mutex_unlock(&grep_attr_mutex
);
1535 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1537 pthread_mutex_t grep_read_mutex
;
1539 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1541 xdemitconf_t
*xecfg
= opt
->priv
;
1542 if (xecfg
&& !xecfg
->find_func
) {
1543 grep_source_load_driver(gs
, opt
->repo
->index
);
1544 if (gs
->driver
->funcname
.pattern
) {
1545 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1546 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1548 xecfg
= opt
->priv
= NULL
;
1554 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1555 xecfg
->find_func_priv
) >= 0;
1560 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1565 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1566 char *bol
, unsigned lno
)
1568 while (bol
> gs
->buf
) {
1571 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1575 if (lno
<= opt
->last_shown
)
1578 if (match_funcname(opt
, gs
, bol
, eol
)) {
1579 show_line(opt
, bol
, eol
, gs
->name
, lno
, 0, '=');
1585 static int is_empty_line(const char *bol
, const char *eol
);
1587 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1588 char *bol
, char *end
, unsigned lno
)
1590 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1591 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1593 if (opt
->pre_context
< lno
)
1594 from
= lno
- opt
->pre_context
;
1595 if (from
<= opt
->last_shown
)
1596 from
= opt
->last_shown
+ 1;
1598 if (opt
->funcbody
) {
1599 if (match_funcname(opt
, gs
, bol
, end
))
1602 funcname_needed
= 1;
1603 from
= opt
->last_shown
+ 1;
1607 while (bol
> gs
->buf
&& cur
> from
) {
1608 char *next_bol
= bol
;
1611 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1614 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1615 match_funcname(opt
, gs
, bol
, eol
))) {
1624 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1626 funcname_needed
= 0;
1634 /* We need to look even further back to find a function signature. */
1635 if (opt
->funcname
&& funcname_needed
)
1636 show_funcname_line(opt
, gs
, bol
, cur
);
1640 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1642 while (*eol
!= '\n')
1644 show_line(opt
, bol
, eol
, gs
->name
, cur
, 0, sign
);
1650 static int should_lookahead(struct grep_opt
*opt
)
1655 return 0; /* punt for too complex stuff */
1658 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1659 if (p
->token
!= GREP_PATTERN
)
1660 return 0; /* punt for "header only" and stuff */
1665 static int look_ahead(struct grep_opt
*opt
,
1666 unsigned long *left_p
,
1670 unsigned lno
= *lno_p
;
1673 char *sp
, *last_bol
;
1674 regoff_t earliest
= -1;
1676 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1680 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1681 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1683 if (earliest
< 0 || m
.rm_so
< earliest
)
1688 *bol_p
= bol
+ *left_p
;
1692 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1693 ; /* find the beginning of the line */
1696 for (sp
= bol
; sp
< last_bol
; sp
++) {
1700 *left_p
-= last_bol
- bol
;
1706 static int fill_textconv_grep(struct repository
*r
,
1707 struct userdiff_driver
*driver
,
1708 struct grep_source
*gs
)
1710 struct diff_filespec
*df
;
1714 if (!driver
|| !driver
->textconv
)
1715 return grep_source_load(gs
);
1718 * The textconv interface is intimately tied to diff_filespecs, so we
1719 * have to pretend to be one. If we could unify the grep_source
1720 * and diff_filespec structs, this mess could just go away.
1722 df
= alloc_filespec(gs
->path
);
1724 case GREP_SOURCE_OID
:
1725 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1727 case GREP_SOURCE_FILE
:
1728 fill_filespec(df
, &null_oid
, 0, 0100644);
1731 BUG("attempt to textconv something without a path?");
1735 * fill_textconv is not remotely thread-safe; it may load objects
1736 * behind the scenes, and it modifies the global diff tempfile
1740 size
= fill_textconv(r
, driver
, df
, &buf
);
1745 * The normal fill_textconv usage by the diff machinery would just keep
1746 * the textconv'd buf separate from the diff_filespec. But much of the
1747 * grep code passes around a grep_source and assumes that its "buf"
1748 * pointer is the beginning of the thing we are searching. So let's
1749 * install our textconv'd version into the grep_source, taking care not
1750 * to leak any existing buffer.
1752 grep_source_clear_data(gs
);
1759 static int is_empty_line(const char *bol
, const char *eol
)
1761 while (bol
< eol
&& isspace(*bol
))
1766 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1769 char *peek_bol
= NULL
;
1772 unsigned last_hit
= 0;
1773 int binary_match_only
= 0;
1775 int try_lookahead
= 0;
1776 int show_function
= 0;
1777 struct userdiff_driver
*textconv
= NULL
;
1778 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1781 if (!opt
->status_only
&& gs
->name
== NULL
)
1782 BUG("grep call which could print a name requires "
1783 "grep_source.name be non-NULL");
1786 opt
->output
= std_output
;
1788 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1790 /* Show hunk marks, except for the first file. */
1791 if (opt
->last_shown
)
1792 opt
->show_hunk_mark
= 1;
1794 * If we're using threads then we can't easily identify
1795 * the first file. Always put hunk marks in that case
1796 * and skip the very first one later in work_done().
1798 if (opt
->output
!= std_output
)
1799 opt
->show_hunk_mark
= 1;
1801 opt
->last_shown
= 0;
1803 if (opt
->allow_textconv
) {
1804 grep_source_load_driver(gs
, opt
->repo
->index
);
1806 * We might set up the shared textconv cache data here, which
1807 * is not thread-safe.
1810 textconv
= userdiff_get_textconv(opt
->repo
, gs
->driver
);
1815 * We know the result of a textconv is text, so we only have to care
1816 * about binary handling if we are not using it.
1819 switch (opt
->binary
) {
1820 case GREP_BINARY_DEFAULT
:
1821 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1822 binary_match_only
= 1;
1824 case GREP_BINARY_NOMATCH
:
1825 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1826 return 0; /* Assume unmatch */
1828 case GREP_BINARY_TEXT
:
1831 BUG("unknown binary handling mode");
1835 memset(&xecfg
, 0, sizeof(xecfg
));
1838 try_lookahead
= should_lookahead(opt
);
1840 if (fill_textconv_grep(opt
->repo
, textconv
, gs
) < 0)
1849 ssize_t col
= -1, icol
= -1;
1852 * look_ahead() skips quickly to the line that possibly
1853 * has the next hit; don't call it if we need to do
1854 * something more than just skipping the current line
1855 * in response to an unmatch for the current line. E.g.
1856 * inside a post-context window, we will show the current
1857 * line as a context around the previous hit when it
1862 && (show_function
||
1863 lno
<= last_hit
+ opt
->post_context
))
1864 && look_ahead(opt
, &left
, &lno
, &bol
))
1866 eol
= end_of_line(bol
, &left
);
1870 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1871 ctx
= GREP_CONTEXT_BODY
;
1873 hit
= match_line(opt
, bol
, eol
, &col
, &icol
, ctx
, collect_hits
);
1879 /* "grep -v -e foo -e bla" should list lines
1880 * that do not have either, so inversion should
1885 if (opt
->unmatch_name_only
) {
1892 if (opt
->status_only
)
1894 if (opt
->name_only
) {
1895 show_name(opt
, gs
->name
);
1900 if (binary_match_only
) {
1901 opt
->output(opt
, "Binary file ", 12);
1902 output_color(opt
, gs
->name
, strlen(gs
->name
),
1903 opt
->colors
[GREP_COLOR_FILENAME
]);
1904 opt
->output(opt
, " matches\n", 9);
1907 /* Hit at this line. If we haven't shown the
1908 * pre-context lines, we would need to show them.
1910 if (opt
->pre_context
|| opt
->funcbody
)
1911 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1912 else if (opt
->funcname
)
1913 show_funcname_line(opt
, gs
, bol
, lno
);
1914 cno
= opt
->invert
? icol
: col
;
1917 * A negative cno indicates that there was no
1918 * match on the line. We are thus inverted and
1919 * being asked to show all lines that _don't_
1920 * match a given expression. Therefore, set cno
1921 * to 0 to suggest the whole line matches.
1925 show_line(opt
, bol
, eol
, gs
->name
, lno
, cno
+ 1, ':');
1931 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1932 unsigned long peek_left
= left
;
1933 char *peek_eol
= eol
;
1936 * Trailing empty lines are not interesting.
1937 * Peek past them to see if they belong to the
1938 * body of the current function.
1941 while (is_empty_line(peek_bol
, peek_eol
)) {
1942 peek_bol
= peek_eol
+ 1;
1943 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1946 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1949 if (show_function
||
1950 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1951 /* If the last hit is within the post context,
1952 * we need to show this line.
1954 show_line(opt
, bol
, eol
, gs
->name
, lno
, col
+ 1, '-');
1968 if (opt
->status_only
)
1969 return opt
->unmatch_name_only
;
1970 if (opt
->unmatch_name_only
) {
1971 /* We did not see any hit, so we want to show this */
1972 show_name(opt
, gs
->name
);
1976 xdiff_clear_find_func(&xecfg
);
1980 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1981 * which feels mostly useless but sometimes useful. Maybe
1982 * make it another option? For now suppress them.
1984 if (opt
->count
&& count
) {
1986 if (opt
->pathname
) {
1987 output_color(opt
, gs
->name
, strlen(gs
->name
),
1988 opt
->colors
[GREP_COLOR_FILENAME
]);
1989 output_sep(opt
, ':');
1991 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1992 opt
->output(opt
, buf
, strlen(buf
));
1998 static void clr_hit_marker(struct grep_expr
*x
)
2000 /* All-hit markers are meaningful only at the very top level
2005 if (x
->node
!= GREP_NODE_OR
)
2007 x
->u
.binary
.left
->hit
= 0;
2008 x
= x
->u
.binary
.right
;
2012 static int chk_hit_marker(struct grep_expr
*x
)
2014 /* Top level nodes have hit markers. See if they all are hits */
2016 if (x
->node
!= GREP_NODE_OR
)
2018 if (!x
->u
.binary
.left
->hit
)
2020 x
= x
->u
.binary
.right
;
2024 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
2027 * we do not have to do the two-pass grep when we do not check
2028 * buffer-wide "all-match".
2030 if (!opt
->all_match
)
2031 return grep_source_1(opt
, gs
, 0);
2033 /* Otherwise the toplevel "or" terms hit a bit differently.
2034 * We first clear hit markers from them.
2036 clr_hit_marker(opt
->pattern_expression
);
2037 grep_source_1(opt
, gs
, 1);
2039 if (!chk_hit_marker(opt
->pattern_expression
))
2042 return grep_source_1(opt
, gs
, 0);
2045 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
2047 struct grep_source gs
;
2050 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
2054 r
= grep_source(opt
, &gs
);
2056 grep_source_clear(&gs
);
2060 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
2061 const char *name
, const char *path
,
2062 const void *identifier
)
2065 gs
->name
= xstrdup_or_null(name
);
2066 gs
->path
= xstrdup_or_null(path
);
2072 case GREP_SOURCE_FILE
:
2073 gs
->identifier
= xstrdup(identifier
);
2075 case GREP_SOURCE_OID
:
2076 gs
->identifier
= oiddup(identifier
);
2078 case GREP_SOURCE_BUF
:
2079 gs
->identifier
= NULL
;
2084 void grep_source_clear(struct grep_source
*gs
)
2086 FREE_AND_NULL(gs
->name
);
2087 FREE_AND_NULL(gs
->path
);
2088 FREE_AND_NULL(gs
->identifier
);
2089 grep_source_clear_data(gs
);
2092 void grep_source_clear_data(struct grep_source
*gs
)
2095 case GREP_SOURCE_FILE
:
2096 case GREP_SOURCE_OID
:
2097 FREE_AND_NULL(gs
->buf
);
2100 case GREP_SOURCE_BUF
:
2101 /* leave user-provided buf intact */
2106 static int grep_source_load_oid(struct grep_source
*gs
)
2108 enum object_type type
;
2111 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2115 return error(_("'%s': unable to read %s"),
2117 oid_to_hex(gs
->identifier
));
2121 static int grep_source_load_file(struct grep_source
*gs
)
2123 const char *filename
= gs
->identifier
;
2129 if (lstat(filename
, &st
) < 0) {
2131 if (errno
!= ENOENT
)
2132 error_errno(_("failed to stat '%s'"), filename
);
2135 if (!S_ISREG(st
.st_mode
))
2137 size
= xsize_t(st
.st_size
);
2138 i
= open(filename
, O_RDONLY
);
2141 data
= xmallocz(size
);
2142 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2143 error_errno(_("'%s': short read"), filename
);
2155 static int grep_source_load(struct grep_source
*gs
)
2161 case GREP_SOURCE_FILE
:
2162 return grep_source_load_file(gs
);
2163 case GREP_SOURCE_OID
:
2164 return grep_source_load_oid(gs
);
2165 case GREP_SOURCE_BUF
:
2166 return gs
->buf
? 0 : -1;
2168 BUG("invalid grep_source type to load");
2171 void grep_source_load_driver(struct grep_source
*gs
,
2172 struct index_state
*istate
)
2179 gs
->driver
= userdiff_find_by_path(istate
, gs
->path
);
2181 gs
->driver
= userdiff_find_by_name("default");
2185 static int grep_source_is_binary(struct grep_source
*gs
,
2186 struct index_state
*istate
)
2188 grep_source_load_driver(gs
, istate
);
2189 if (gs
->driver
->binary
!= -1)
2190 return gs
->driver
->binary
;
2192 if (!grep_source_load(gs
))
2193 return buffer_is_binary(gs
->buf
, gs
->size
);