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
]))
372 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
376 int options
= PCRE_MULTILINE
;
378 if (opt
->ignore_case
) {
379 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
))
380 p
->pcre1_tables
= pcre_maketables();
381 options
|= PCRE_CASELESS
;
383 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
))
384 options
|= PCRE_UTF8
;
386 p
->pcre1_regexp
= pcre_compile(p
->pattern
, options
, &error
, &erroffset
,
388 if (!p
->pcre1_regexp
)
389 compile_regexp_failed(p
, error
);
391 p
->pcre1_extra_info
= pcre_study(p
->pcre1_regexp
, GIT_PCRE_STUDY_JIT_COMPILE
, &error
);
392 if (!p
->pcre1_extra_info
&& error
)
395 #ifdef GIT_PCRE1_USE_JIT
396 pcre_config(PCRE_CONFIG_JIT
, &p
->pcre1_jit_on
);
398 fprintf(stderr
, "pcre1_jit_on=%d\n", p
->pcre1_jit_on
);
402 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
403 regmatch_t
*match
, int eflags
)
405 int ovector
[30], ret
, flags
= PCRE_NO_UTF8_CHECK
;
407 if (eflags
& REG_NOTBOL
)
408 flags
|= PCRE_NOTBOL
;
410 ret
= pcre_exec(p
->pcre1_regexp
, p
->pcre1_extra_info
, line
,
411 eol
- line
, 0, flags
, ovector
,
412 ARRAY_SIZE(ovector
));
414 if (ret
< 0 && ret
!= PCRE_ERROR_NOMATCH
)
415 die("pcre_exec failed with error code %d", ret
);
418 match
->rm_so
= ovector
[0];
419 match
->rm_eo
= ovector
[1];
425 static void free_pcre1_regexp(struct grep_pat
*p
)
427 pcre_free(p
->pcre1_regexp
);
428 #ifdef GIT_PCRE1_USE_JIT
430 pcre_free_study(p
->pcre1_extra_info
);
433 pcre_free(p
->pcre1_extra_info
);
434 pcre_free((void *)p
->pcre1_tables
);
436 #else /* !USE_LIBPCRE1 */
437 static void compile_pcre1_regexp(struct grep_pat
*p
, const struct grep_opt
*opt
)
439 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
442 static int pcre1match(struct grep_pat
*p
, const char *line
, const char *eol
,
443 regmatch_t
*match
, int eflags
)
448 static void free_pcre1_regexp(struct grep_pat
*p
)
451 #endif /* !USE_LIBPCRE1 */
454 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
457 PCRE2_UCHAR errbuf
[256];
458 PCRE2_SIZE erroffset
;
459 int options
= PCRE2_MULTILINE
;
460 const uint8_t *character_tables
= NULL
;
467 p
->pcre2_compile_context
= NULL
;
469 if (opt
->ignore_case
) {
470 if (!opt
->ignore_locale
&& has_non_ascii(p
->pattern
)) {
471 character_tables
= pcre2_maketables(NULL
);
472 p
->pcre2_compile_context
= pcre2_compile_context_create(NULL
);
473 pcre2_set_character_tables(p
->pcre2_compile_context
, character_tables
);
475 options
|= PCRE2_CASELESS
;
477 if (!opt
->ignore_locale
&& is_utf8_locale() && has_non_ascii(p
->pattern
) &&
478 !(!opt
->ignore_case
&& (p
->fixed
|| p
->is_fixed
)))
479 options
|= PCRE2_UTF
;
481 p
->pcre2_pattern
= pcre2_compile((PCRE2_SPTR
)p
->pattern
,
482 p
->patternlen
, options
, &error
, &erroffset
,
483 p
->pcre2_compile_context
);
485 if (p
->pcre2_pattern
) {
486 p
->pcre2_match_data
= pcre2_match_data_create_from_pattern(p
->pcre2_pattern
, NULL
);
487 if (!p
->pcre2_match_data
)
488 die("Couldn't allocate PCRE2 match data");
490 pcre2_get_error_message(error
, errbuf
, sizeof(errbuf
));
491 compile_regexp_failed(p
, (const char *)&errbuf
);
494 pcre2_config(PCRE2_CONFIG_JIT
, &p
->pcre2_jit_on
);
496 fprintf(stderr
, "pcre2_jit_on=%d\n", p
->pcre2_jit_on
);
497 if (p
->pcre2_jit_on
) {
498 jitret
= pcre2_jit_compile(p
->pcre2_pattern
, PCRE2_JIT_COMPLETE
);
500 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p
->pattern
, jitret
);
503 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
504 * tells us whether the library itself supports JIT,
505 * but to see whether we're going to be actually using
506 * JIT we need to extract PCRE2_INFO_JITSIZE from the
507 * pattern *after* we do pcre2_jit_compile() above.
509 * This is because if the pattern contains the
510 * (*NO_JIT) verb (see pcre2syntax(3))
511 * pcre2_jit_compile() will exit early with 0. If we
512 * then proceed to call pcre2_jit_match() further down
513 * the line instead of pcre2_match() we'll either
514 * segfault (pre PCRE 10.31) or run into a fatal error
517 patinforet
= pcre2_pattern_info(p
->pcre2_pattern
, PCRE2_INFO_JITSIZE
, &jitsizearg
);
519 BUG("pcre2_pattern_info() failed: %d", patinforet
);
520 if (jitsizearg
== 0) {
523 fprintf(stderr
, "pcre2_jit_on=%d: (*NO_JIT) in regex\n",
530 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
531 regmatch_t
*match
, int eflags
)
535 PCRE2_UCHAR errbuf
[256];
537 if (eflags
& REG_NOTBOL
)
538 flags
|= PCRE2_NOTBOL
;
541 ret
= pcre2_jit_match(p
->pcre2_pattern
, (unsigned char *)line
,
542 eol
- line
, 0, flags
, p
->pcre2_match_data
,
545 ret
= pcre2_match(p
->pcre2_pattern
, (unsigned char *)line
,
546 eol
- line
, 0, flags
, p
->pcre2_match_data
,
549 if (ret
< 0 && ret
!= PCRE2_ERROR_NOMATCH
) {
550 pcre2_get_error_message(ret
, errbuf
, sizeof(errbuf
));
551 die("%s failed with error code %d: %s",
552 (p
->pcre2_jit_on
? "pcre2_jit_match" : "pcre2_match"), ret
,
556 ovector
= pcre2_get_ovector_pointer(p
->pcre2_match_data
);
558 match
->rm_so
= (int)ovector
[0];
559 match
->rm_eo
= (int)ovector
[1];
565 static void free_pcre2_pattern(struct grep_pat
*p
)
567 pcre2_compile_context_free(p
->pcre2_compile_context
);
568 pcre2_code_free(p
->pcre2_pattern
);
569 pcre2_match_data_free(p
->pcre2_match_data
);
571 #else /* !USE_LIBPCRE2 */
572 static void compile_pcre2_pattern(struct grep_pat
*p
, const struct grep_opt
*opt
)
575 * Unreachable until USE_LIBPCRE2 becomes synonymous with
576 * USE_LIBPCRE. See the sibling comment in
577 * grep_set_pattern_type_option().
579 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
582 static int pcre2match(struct grep_pat
*p
, const char *line
, const char *eol
,
583 regmatch_t
*match
, int eflags
)
588 static void free_pcre2_pattern(struct grep_pat
*p
)
592 static void compile_fixed_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
594 struct strbuf sb
= STRBUF_INIT
;
598 basic_regex_quote_buf(&sb
, p
->pattern
);
599 if (opt
->ignore_case
)
600 regflags
|= REG_ICASE
;
601 err
= regcomp(&p
->regexp
, sb
.buf
, regflags
);
603 fprintf(stderr
, "fixed %s\n", sb
.buf
);
607 regerror(err
, &p
->regexp
, errbuf
, sizeof(errbuf
));
608 compile_regexp_failed(p
, errbuf
);
611 #endif /* !USE_LIBPCRE2 */
613 static void compile_regexp(struct grep_pat
*p
, struct grep_opt
*opt
)
616 int regflags
= REG_NEWLINE
;
618 p
->word_regexp
= opt
->word_regexp
;
619 p
->ignore_case
= opt
->ignore_case
;
620 p
->fixed
= opt
->fixed
;
622 if (memchr(p
->pattern
, 0, p
->patternlen
) && !opt
->pcre2
)
623 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
625 p
->is_fixed
= is_fixed(p
->pattern
, p
->patternlen
);
627 if (!p
->fixed
&& !p
->is_fixed
) {
628 const char *no_jit
= "(*NO_JIT)";
629 const int no_jit_len
= strlen(no_jit
);
630 if (starts_with(p
->pattern
, no_jit
) &&
631 is_fixed(p
->pattern
+ no_jit_len
,
632 p
->patternlen
- no_jit_len
))
636 if (p
->fixed
|| p
->is_fixed
) {
640 compile_pcre2_pattern(p
, opt
);
643 * E.g. t7811-grep-open.sh relies on the
644 * pattern being restored.
646 char *old_pattern
= p
->pattern
;
647 size_t old_patternlen
= p
->patternlen
;
648 struct strbuf sb
= STRBUF_INIT
;
651 * There is the PCRE2_LITERAL flag, but it's
652 * only in PCRE v2 10.30 and later. Needing to
653 * ifdef our way around that and dealing with
654 * it + PCRE2_MULTILINE being an error is more
655 * complex than just quoting this ourselves.
657 strbuf_add(&sb
, "\\Q", 2);
658 strbuf_add(&sb
, p
->pattern
, p
->patternlen
);
659 strbuf_add(&sb
, "\\E", 2);
662 p
->patternlen
= sb
.len
;
663 compile_pcre2_pattern(p
, opt
);
664 p
->pattern
= old_pattern
;
665 p
->patternlen
= old_patternlen
;
668 #else /* !USE_LIBPCRE2 */
669 compile_fixed_regexp(p
, opt
);
670 #endif /* !USE_LIBPCRE2 */
675 compile_pcre2_pattern(p
, opt
);
680 compile_pcre1_regexp(p
, opt
);
685 regflags
|= REG_ICASE
;
686 if (opt
->extended_regexp_option
)
687 regflags
|= REG_EXTENDED
;
688 err
= regcomp(&p
->regexp
, p
->pattern
, regflags
);
691 regerror(err
, &p
->regexp
, errbuf
, 1024);
692 compile_regexp_failed(p
, errbuf
);
696 static struct grep_expr
*compile_pattern_or(struct grep_pat
**);
697 static struct grep_expr
*compile_pattern_atom(struct grep_pat
**list
)
706 case GREP_PATTERN
: /* atom */
707 case GREP_PATTERN_HEAD
:
708 case GREP_PATTERN_BODY
:
709 x
= xcalloc(1, sizeof (struct grep_expr
));
710 x
->node
= GREP_NODE_ATOM
;
714 case GREP_OPEN_PAREN
:
716 x
= compile_pattern_or(list
);
717 if (!*list
|| (*list
)->token
!= GREP_CLOSE_PAREN
)
718 die("unmatched parenthesis");
719 *list
= (*list
)->next
;
726 static struct grep_expr
*compile_pattern_not(struct grep_pat
**list
)
737 die("--not not followed by pattern expression");
739 x
= xcalloc(1, sizeof (struct grep_expr
));
740 x
->node
= GREP_NODE_NOT
;
741 x
->u
.unary
= compile_pattern_not(list
);
743 die("--not followed by non pattern expression");
746 return compile_pattern_atom(list
);
750 static struct grep_expr
*compile_pattern_and(struct grep_pat
**list
)
753 struct grep_expr
*x
, *y
, *z
;
755 x
= compile_pattern_not(list
);
757 if (p
&& p
->token
== GREP_AND
) {
759 die("--and not followed by pattern expression");
761 y
= compile_pattern_and(list
);
763 die("--and not followed by pattern expression");
764 z
= xcalloc(1, sizeof (struct grep_expr
));
765 z
->node
= GREP_NODE_AND
;
766 z
->u
.binary
.left
= x
;
767 z
->u
.binary
.right
= y
;
773 static struct grep_expr
*compile_pattern_or(struct grep_pat
**list
)
776 struct grep_expr
*x
, *y
, *z
;
778 x
= compile_pattern_and(list
);
780 if (x
&& p
&& p
->token
!= GREP_CLOSE_PAREN
) {
781 y
= compile_pattern_or(list
);
783 die("not a pattern expression %s", p
->pattern
);
784 z
= xcalloc(1, sizeof (struct grep_expr
));
785 z
->node
= GREP_NODE_OR
;
786 z
->u
.binary
.left
= x
;
787 z
->u
.binary
.right
= y
;
793 static struct grep_expr
*compile_pattern_expr(struct grep_pat
**list
)
795 return compile_pattern_or(list
);
798 static void indent(int in
)
804 static void dump_grep_pat(struct grep_pat
*p
)
807 case GREP_AND
: fprintf(stderr
, "*and*"); break;
808 case GREP_OPEN_PAREN
: fprintf(stderr
, "*(*"); break;
809 case GREP_CLOSE_PAREN
: fprintf(stderr
, "*)*"); break;
810 case GREP_NOT
: fprintf(stderr
, "*not*"); break;
811 case GREP_OR
: fprintf(stderr
, "*or*"); break;
813 case GREP_PATTERN
: fprintf(stderr
, "pattern"); break;
814 case GREP_PATTERN_HEAD
: fprintf(stderr
, "pattern_head"); break;
815 case GREP_PATTERN_BODY
: fprintf(stderr
, "pattern_body"); break;
820 case GREP_PATTERN_HEAD
:
821 fprintf(stderr
, "<head %d>", p
->field
); break;
822 case GREP_PATTERN_BODY
:
823 fprintf(stderr
, "<body>"); break;
827 case GREP_PATTERN_HEAD
:
828 case GREP_PATTERN_BODY
:
830 fprintf(stderr
, "%.*s", (int)p
->patternlen
, p
->pattern
);
836 static void dump_grep_expression_1(struct grep_expr
*x
, int in
)
841 fprintf(stderr
, "true\n");
844 dump_grep_pat(x
->u
.atom
);
847 fprintf(stderr
, "(not\n");
848 dump_grep_expression_1(x
->u
.unary
, in
+1);
850 fprintf(stderr
, ")\n");
853 fprintf(stderr
, "(and\n");
854 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
855 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
857 fprintf(stderr
, ")\n");
860 fprintf(stderr
, "(or\n");
861 dump_grep_expression_1(x
->u
.binary
.left
, in
+1);
862 dump_grep_expression_1(x
->u
.binary
.right
, in
+1);
864 fprintf(stderr
, ")\n");
869 static void dump_grep_expression(struct grep_opt
*opt
)
871 struct grep_expr
*x
= opt
->pattern_expression
;
874 fprintf(stderr
, "[all-match]\n");
875 dump_grep_expression_1(x
, 0);
879 static struct grep_expr
*grep_true_expr(void)
881 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
882 z
->node
= GREP_NODE_TRUE
;
886 static struct grep_expr
*grep_or_expr(struct grep_expr
*left
, struct grep_expr
*right
)
888 struct grep_expr
*z
= xcalloc(1, sizeof(*z
));
889 z
->node
= GREP_NODE_OR
;
890 z
->u
.binary
.left
= left
;
891 z
->u
.binary
.right
= right
;
895 static struct grep_expr
*prep_header_patterns(struct grep_opt
*opt
)
898 struct grep_expr
*header_expr
;
899 struct grep_expr
*(header_group
[GREP_HEADER_FIELD_MAX
]);
900 enum grep_header_field fld
;
902 if (!opt
->header_list
)
905 for (p
= opt
->header_list
; p
; p
= p
->next
) {
906 if (p
->token
!= GREP_PATTERN_HEAD
)
907 BUG("a non-header pattern in grep header list.");
908 if (p
->field
< GREP_HEADER_FIELD_MIN
||
909 GREP_HEADER_FIELD_MAX
<= p
->field
)
910 BUG("unknown header field %d", p
->field
);
911 compile_regexp(p
, opt
);
914 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++)
915 header_group
[fld
] = NULL
;
917 for (p
= opt
->header_list
; p
; p
= p
->next
) {
919 struct grep_pat
*pp
= p
;
921 h
= compile_pattern_atom(&pp
);
922 if (!h
|| pp
!= p
->next
)
923 BUG("malformed header expr");
924 if (!header_group
[p
->field
]) {
925 header_group
[p
->field
] = h
;
928 header_group
[p
->field
] = grep_or_expr(h
, header_group
[p
->field
]);
933 for (fld
= 0; fld
< GREP_HEADER_FIELD_MAX
; fld
++) {
934 if (!header_group
[fld
])
937 header_expr
= grep_true_expr();
938 header_expr
= grep_or_expr(header_group
[fld
], header_expr
);
943 static struct grep_expr
*grep_splice_or(struct grep_expr
*x
, struct grep_expr
*y
)
945 struct grep_expr
*z
= x
;
948 assert(x
->node
== GREP_NODE_OR
);
949 if (x
->u
.binary
.right
&&
950 x
->u
.binary
.right
->node
== GREP_NODE_TRUE
) {
951 x
->u
.binary
.right
= y
;
954 x
= x
->u
.binary
.right
;
959 static void compile_grep_patterns_real(struct grep_opt
*opt
)
962 struct grep_expr
*header_expr
= prep_header_patterns(opt
);
964 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
966 case GREP_PATTERN
: /* atom */
967 case GREP_PATTERN_HEAD
:
968 case GREP_PATTERN_BODY
:
969 compile_regexp(p
, opt
);
977 if (opt
->all_match
|| header_expr
)
979 else if (!opt
->extended
&& !opt
->debug
)
982 p
= opt
->pattern_list
;
984 opt
->pattern_expression
= compile_pattern_expr(&p
);
986 die("incomplete pattern expression: %s", p
->pattern
);
991 if (!opt
->pattern_expression
)
992 opt
->pattern_expression
= header_expr
;
993 else if (opt
->all_match
)
994 opt
->pattern_expression
= grep_splice_or(header_expr
,
995 opt
->pattern_expression
);
997 opt
->pattern_expression
= grep_or_expr(opt
->pattern_expression
,
1002 void compile_grep_patterns(struct grep_opt
*opt
)
1004 compile_grep_patterns_real(opt
);
1006 dump_grep_expression(opt
);
1009 static void free_pattern_expr(struct grep_expr
*x
)
1012 case GREP_NODE_TRUE
:
1013 case GREP_NODE_ATOM
:
1016 free_pattern_expr(x
->u
.unary
);
1020 free_pattern_expr(x
->u
.binary
.left
);
1021 free_pattern_expr(x
->u
.binary
.right
);
1027 void free_grep_patterns(struct grep_opt
*opt
)
1029 struct grep_pat
*p
, *n
;
1031 for (p
= opt
->pattern_list
; p
; p
= n
) {
1034 case GREP_PATTERN
: /* atom */
1035 case GREP_PATTERN_HEAD
:
1036 case GREP_PATTERN_BODY
:
1037 if (p
->pcre1_regexp
)
1038 free_pcre1_regexp(p
);
1039 else if (p
->pcre2_pattern
)
1040 free_pcre2_pattern(p
);
1042 regfree(&p
->regexp
);
1053 free_pattern_expr(opt
->pattern_expression
);
1056 static char *end_of_line(char *cp
, unsigned long *left
)
1058 unsigned long l
= *left
;
1059 while (l
&& *cp
!= '\n') {
1067 static int word_char(char ch
)
1069 return isalnum(ch
) || ch
== '_';
1072 static void output_color(struct grep_opt
*opt
, const void *data
, size_t size
,
1075 if (want_color(opt
->color
) && color
&& color
[0]) {
1076 opt
->output(opt
, color
, strlen(color
));
1077 opt
->output(opt
, data
, size
);
1078 opt
->output(opt
, GIT_COLOR_RESET
, strlen(GIT_COLOR_RESET
));
1080 opt
->output(opt
, data
, size
);
1083 static void output_sep(struct grep_opt
*opt
, char sign
)
1085 if (opt
->null_following_name
)
1086 opt
->output(opt
, "\0", 1);
1088 output_color(opt
, &sign
, 1, opt
->colors
[GREP_COLOR_SEP
]);
1091 static void show_name(struct grep_opt
*opt
, const char *name
)
1093 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1094 opt
->output(opt
, opt
->null_following_name
? "\0" : "\n", 1);
1097 static int patmatch(struct grep_pat
*p
, char *line
, char *eol
,
1098 regmatch_t
*match
, int eflags
)
1102 if (p
->pcre1_regexp
)
1103 hit
= !pcre1match(p
, line
, eol
, match
, eflags
);
1104 else if (p
->pcre2_pattern
)
1105 hit
= !pcre2match(p
, line
, eol
, match
, eflags
);
1107 hit
= !regexec_buf(&p
->regexp
, line
, eol
- line
, 1, match
,
1113 static int strip_timestamp(char *bol
, char **eol_p
)
1118 while (bol
< --eol
) {
1132 } header_field
[] = {
1134 { "committer ", 10 },
1138 static int match_one_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1139 enum grep_context ctx
,
1140 regmatch_t
*pmatch
, int eflags
)
1144 const char *start
= bol
;
1146 if ((p
->token
!= GREP_PATTERN
) &&
1147 ((p
->token
== GREP_PATTERN_HEAD
) != (ctx
== GREP_CONTEXT_HEAD
)))
1150 if (p
->token
== GREP_PATTERN_HEAD
) {
1153 assert(p
->field
< ARRAY_SIZE(header_field
));
1154 field
= header_field
[p
->field
].field
;
1155 len
= header_field
[p
->field
].len
;
1156 if (strncmp(bol
, field
, len
))
1160 case GREP_HEADER_AUTHOR
:
1161 case GREP_HEADER_COMMITTER
:
1162 saved_ch
= strip_timestamp(bol
, &eol
);
1170 hit
= patmatch(p
, bol
, eol
, pmatch
, eflags
);
1172 if (hit
&& p
->word_regexp
) {
1173 if ((pmatch
[0].rm_so
< 0) ||
1174 (eol
- bol
) < pmatch
[0].rm_so
||
1175 (pmatch
[0].rm_eo
< 0) ||
1176 (eol
- bol
) < pmatch
[0].rm_eo
)
1177 die("regexp returned nonsense");
1179 /* Match beginning must be either beginning of the
1180 * line, or at word boundary (i.e. the last char must
1181 * not be a word char). Similarly, match end must be
1182 * either end of the line, or at word boundary
1183 * (i.e. the next char must not be a word char).
1185 if ( ((pmatch
[0].rm_so
== 0) ||
1186 !word_char(bol
[pmatch
[0].rm_so
-1])) &&
1187 ((pmatch
[0].rm_eo
== (eol
-bol
)) ||
1188 !word_char(bol
[pmatch
[0].rm_eo
])) )
1193 /* Words consist of at least one character. */
1194 if (pmatch
->rm_so
== pmatch
->rm_eo
)
1197 if (!hit
&& pmatch
[0].rm_so
+ bol
+ 1 < eol
) {
1198 /* There could be more than one match on the
1199 * line, and the first match might not be
1200 * strict word match. But later ones could be!
1201 * Forward to the next possible start, i.e. the
1202 * next position following a non-word char.
1204 bol
= pmatch
[0].rm_so
+ bol
+ 1;
1205 while (word_char(bol
[-1]) && bol
< eol
)
1207 eflags
|= REG_NOTBOL
;
1212 if (p
->token
== GREP_PATTERN_HEAD
&& saved_ch
)
1215 pmatch
[0].rm_so
+= bol
- start
;
1216 pmatch
[0].rm_eo
+= bol
- start
;
1221 static int match_expr_eval(struct grep_opt
*opt
, struct grep_expr
*x
, char *bol
,
1222 char *eol
, enum grep_context ctx
, ssize_t
*col
,
1223 ssize_t
*icol
, int collect_hits
)
1228 die("Not a valid grep expression");
1230 case GREP_NODE_TRUE
:
1233 case GREP_NODE_ATOM
:
1236 h
= match_one_pattern(x
->u
.atom
, bol
, eol
, ctx
,
1238 if (h
&& (*col
< 0 || tmp
.rm_so
< *col
))
1244 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1246 h
= !match_expr_eval(opt
, x
->u
.unary
, bol
, eol
, ctx
, icol
, col
,
1250 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1252 if (h
|| opt
->columnnum
) {
1254 * Don't short-circuit AND when given --column, since a
1255 * NOT earlier in the tree may turn this into an OR. In
1256 * this case, see the below comment.
1258 h
&= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
,
1263 if (!(collect_hits
|| opt
->columnnum
)) {
1265 * Don't short-circuit OR when given --column (or
1266 * collecting hits) to ensure we don't skip a later
1267 * child that would produce an earlier match.
1269 return (match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
,
1270 ctx
, col
, icol
, 0) ||
1271 match_expr_eval(opt
, x
->u
.binary
.right
, bol
,
1272 eol
, ctx
, col
, icol
, 0));
1274 h
= match_expr_eval(opt
, x
->u
.binary
.left
, bol
, eol
, ctx
, col
,
1277 x
->u
.binary
.left
->hit
|= h
;
1278 h
|= match_expr_eval(opt
, x
->u
.binary
.right
, bol
, eol
, ctx
, col
,
1279 icol
, collect_hits
);
1282 die("Unexpected node type (internal error) %d", x
->node
);
1289 static int match_expr(struct grep_opt
*opt
, char *bol
, char *eol
,
1290 enum grep_context ctx
, ssize_t
*col
,
1291 ssize_t
*icol
, int collect_hits
)
1293 struct grep_expr
*x
= opt
->pattern_expression
;
1294 return match_expr_eval(opt
, x
, bol
, eol
, ctx
, col
, icol
, collect_hits
);
1297 static int match_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1298 ssize_t
*col
, ssize_t
*icol
,
1299 enum grep_context ctx
, int collect_hits
)
1305 return match_expr(opt
, bol
, eol
, ctx
, col
, icol
,
1308 /* we do not call with collect_hits without being extended */
1309 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1311 if (match_one_pattern(p
, bol
, eol
, ctx
, &tmp
, 0)) {
1313 if (!opt
->columnnum
) {
1315 * Without --column, any single match on a line
1316 * is enough to know that it needs to be
1317 * printed. With --column, scan _all_ patterns
1318 * to find the earliest.
1322 if (*col
< 0 || tmp
.rm_so
< *col
)
1329 static int match_next_pattern(struct grep_pat
*p
, char *bol
, char *eol
,
1330 enum grep_context ctx
,
1331 regmatch_t
*pmatch
, int eflags
)
1335 if (!match_one_pattern(p
, bol
, eol
, ctx
, &match
, eflags
))
1337 if (match
.rm_so
< 0 || match
.rm_eo
< 0)
1339 if (pmatch
->rm_so
>= 0 && pmatch
->rm_eo
>= 0) {
1340 if (match
.rm_so
> pmatch
->rm_so
)
1342 if (match
.rm_so
== pmatch
->rm_so
&& match
.rm_eo
< pmatch
->rm_eo
)
1345 pmatch
->rm_so
= match
.rm_so
;
1346 pmatch
->rm_eo
= match
.rm_eo
;
1350 static int next_match(struct grep_opt
*opt
, char *bol
, char *eol
,
1351 enum grep_context ctx
, regmatch_t
*pmatch
, int eflags
)
1356 pmatch
->rm_so
= pmatch
->rm_eo
= -1;
1358 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1360 case GREP_PATTERN
: /* atom */
1361 case GREP_PATTERN_HEAD
:
1362 case GREP_PATTERN_BODY
:
1363 hit
|= match_next_pattern(p
, bol
, eol
, ctx
,
1374 static void show_line_header(struct grep_opt
*opt
, const char *name
,
1375 unsigned lno
, ssize_t cno
, char sign
)
1377 if (opt
->heading
&& opt
->last_shown
== 0) {
1378 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1379 opt
->output(opt
, "\n", 1);
1381 opt
->last_shown
= lno
;
1383 if (!opt
->heading
&& opt
->pathname
) {
1384 output_color(opt
, name
, strlen(name
), opt
->colors
[GREP_COLOR_FILENAME
]);
1385 output_sep(opt
, sign
);
1389 xsnprintf(buf
, sizeof(buf
), "%d", lno
);
1390 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_LINENO
]);
1391 output_sep(opt
, sign
);
1394 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1395 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1396 * being called with a context line.
1398 if (opt
->columnnum
&& cno
) {
1400 xsnprintf(buf
, sizeof(buf
), "%"PRIuMAX
, (uintmax_t)cno
);
1401 output_color(opt
, buf
, strlen(buf
), opt
->colors
[GREP_COLOR_COLUMNNO
]);
1402 output_sep(opt
, sign
);
1406 static void show_line(struct grep_opt
*opt
, char *bol
, char *eol
,
1407 const char *name
, unsigned lno
, ssize_t cno
, char sign
)
1409 int rest
= eol
- bol
;
1410 const char *match_color
= NULL
;
1411 const char *line_color
= NULL
;
1413 if (opt
->file_break
&& opt
->last_shown
== 0) {
1414 if (opt
->show_hunk_mark
)
1415 opt
->output(opt
, "\n", 1);
1416 } else if (opt
->pre_context
|| opt
->post_context
|| opt
->funcbody
) {
1417 if (opt
->last_shown
== 0) {
1418 if (opt
->show_hunk_mark
) {
1419 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1420 opt
->output(opt
, "\n", 1);
1422 } else if (lno
> opt
->last_shown
+ 1) {
1423 output_color(opt
, "--", 2, opt
->colors
[GREP_COLOR_SEP
]);
1424 opt
->output(opt
, "\n", 1);
1427 if (!opt
->only_matching
) {
1429 * In case the line we're being called with contains more than
1430 * one match, leave printing each header to the loop below.
1432 show_line_header(opt
, name
, lno
, cno
, sign
);
1434 if (opt
->color
|| opt
->only_matching
) {
1436 enum grep_context ctx
= GREP_CONTEXT_BODY
;
1442 match_color
= opt
->colors
[GREP_COLOR_MATCH_SELECTED
];
1444 match_color
= opt
->colors
[GREP_COLOR_MATCH_CONTEXT
];
1446 line_color
= opt
->colors
[GREP_COLOR_SELECTED
];
1447 else if (sign
== '-')
1448 line_color
= opt
->colors
[GREP_COLOR_CONTEXT
];
1449 else if (sign
== '=')
1450 line_color
= opt
->colors
[GREP_COLOR_FUNCTION
];
1453 while (next_match(opt
, bol
, eol
, ctx
, &match
, eflags
)) {
1454 if (match
.rm_so
== match
.rm_eo
)
1457 if (opt
->only_matching
)
1458 show_line_header(opt
, name
, lno
, cno
, sign
);
1460 output_color(opt
, bol
, match
.rm_so
, line_color
);
1461 output_color(opt
, bol
+ match
.rm_so
,
1462 match
.rm_eo
- match
.rm_so
, match_color
);
1463 if (opt
->only_matching
)
1464 opt
->output(opt
, "\n", 1);
1467 rest
-= match
.rm_eo
;
1468 eflags
= REG_NOTBOL
;
1472 if (!opt
->only_matching
) {
1473 output_color(opt
, bol
, rest
, line_color
);
1474 opt
->output(opt
, "\n", 1);
1481 * This lock protects access to the gitattributes machinery, which is
1484 pthread_mutex_t grep_attr_mutex
;
1486 static inline void grep_attr_lock(void)
1489 pthread_mutex_lock(&grep_attr_mutex
);
1492 static inline void grep_attr_unlock(void)
1495 pthread_mutex_unlock(&grep_attr_mutex
);
1499 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1501 pthread_mutex_t grep_read_mutex
;
1503 static int match_funcname(struct grep_opt
*opt
, struct grep_source
*gs
, char *bol
, char *eol
)
1505 xdemitconf_t
*xecfg
= opt
->priv
;
1506 if (xecfg
&& !xecfg
->find_func
) {
1507 grep_source_load_driver(gs
, opt
->repo
->index
);
1508 if (gs
->driver
->funcname
.pattern
) {
1509 const struct userdiff_funcname
*pe
= &gs
->driver
->funcname
;
1510 xdiff_set_find_func(xecfg
, pe
->pattern
, pe
->cflags
);
1512 xecfg
= opt
->priv
= NULL
;
1518 return xecfg
->find_func(bol
, eol
- bol
, buf
, 1,
1519 xecfg
->find_func_priv
) >= 0;
1524 if (isalpha(*bol
) || *bol
== '_' || *bol
== '$')
1529 static void show_funcname_line(struct grep_opt
*opt
, struct grep_source
*gs
,
1530 char *bol
, unsigned lno
)
1532 while (bol
> gs
->buf
) {
1535 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1539 if (lno
<= opt
->last_shown
)
1542 if (match_funcname(opt
, gs
, bol
, eol
)) {
1543 show_line(opt
, bol
, eol
, gs
->name
, lno
, 0, '=');
1549 static int is_empty_line(const char *bol
, const char *eol
);
1551 static void show_pre_context(struct grep_opt
*opt
, struct grep_source
*gs
,
1552 char *bol
, char *end
, unsigned lno
)
1554 unsigned cur
= lno
, from
= 1, funcname_lno
= 0, orig_from
;
1555 int funcname_needed
= !!opt
->funcname
, comment_needed
= 0;
1557 if (opt
->pre_context
< lno
)
1558 from
= lno
- opt
->pre_context
;
1559 if (from
<= opt
->last_shown
)
1560 from
= opt
->last_shown
+ 1;
1562 if (opt
->funcbody
) {
1563 if (match_funcname(opt
, gs
, bol
, end
))
1566 funcname_needed
= 1;
1567 from
= opt
->last_shown
+ 1;
1571 while (bol
> gs
->buf
&& cur
> from
) {
1572 char *next_bol
= bol
;
1575 while (bol
> gs
->buf
&& bol
[-1] != '\n')
1578 if (comment_needed
&& (is_empty_line(bol
, eol
) ||
1579 match_funcname(opt
, gs
, bol
, eol
))) {
1588 if (funcname_needed
&& match_funcname(opt
, gs
, bol
, eol
)) {
1590 funcname_needed
= 0;
1598 /* We need to look even further back to find a function signature. */
1599 if (opt
->funcname
&& funcname_needed
)
1600 show_funcname_line(opt
, gs
, bol
, cur
);
1604 char *eol
= bol
, sign
= (cur
== funcname_lno
) ? '=' : '-';
1606 while (*eol
!= '\n')
1608 show_line(opt
, bol
, eol
, gs
->name
, cur
, 0, sign
);
1614 static int should_lookahead(struct grep_opt
*opt
)
1619 return 0; /* punt for too complex stuff */
1622 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1623 if (p
->token
!= GREP_PATTERN
)
1624 return 0; /* punt for "header only" and stuff */
1629 static int look_ahead(struct grep_opt
*opt
,
1630 unsigned long *left_p
,
1634 unsigned lno
= *lno_p
;
1637 char *sp
, *last_bol
;
1638 regoff_t earliest
= -1;
1640 for (p
= opt
->pattern_list
; p
; p
= p
->next
) {
1644 hit
= patmatch(p
, bol
, bol
+ *left_p
, &m
, 0);
1645 if (!hit
|| m
.rm_so
< 0 || m
.rm_eo
< 0)
1647 if (earliest
< 0 || m
.rm_so
< earliest
)
1652 *bol_p
= bol
+ *left_p
;
1656 for (sp
= bol
+ earliest
; bol
< sp
&& sp
[-1] != '\n'; sp
--)
1657 ; /* find the beginning of the line */
1660 for (sp
= bol
; sp
< last_bol
; sp
++) {
1664 *left_p
-= last_bol
- bol
;
1670 static int fill_textconv_grep(struct repository
*r
,
1671 struct userdiff_driver
*driver
,
1672 struct grep_source
*gs
)
1674 struct diff_filespec
*df
;
1678 if (!driver
|| !driver
->textconv
)
1679 return grep_source_load(gs
);
1682 * The textconv interface is intimately tied to diff_filespecs, so we
1683 * have to pretend to be one. If we could unify the grep_source
1684 * and diff_filespec structs, this mess could just go away.
1686 df
= alloc_filespec(gs
->path
);
1688 case GREP_SOURCE_OID
:
1689 fill_filespec(df
, gs
->identifier
, 1, 0100644);
1691 case GREP_SOURCE_FILE
:
1692 fill_filespec(df
, &null_oid
, 0, 0100644);
1695 BUG("attempt to textconv something without a path?");
1699 * fill_textconv is not remotely thread-safe; it may load objects
1700 * behind the scenes, and it modifies the global diff tempfile
1704 size
= fill_textconv(r
, driver
, df
, &buf
);
1709 * The normal fill_textconv usage by the diff machinery would just keep
1710 * the textconv'd buf separate from the diff_filespec. But much of the
1711 * grep code passes around a grep_source and assumes that its "buf"
1712 * pointer is the beginning of the thing we are searching. So let's
1713 * install our textconv'd version into the grep_source, taking care not
1714 * to leak any existing buffer.
1716 grep_source_clear_data(gs
);
1723 static int is_empty_line(const char *bol
, const char *eol
)
1725 while (bol
< eol
&& isspace(*bol
))
1730 static int grep_source_1(struct grep_opt
*opt
, struct grep_source
*gs
, int collect_hits
)
1733 char *peek_bol
= NULL
;
1736 unsigned last_hit
= 0;
1737 int binary_match_only
= 0;
1739 int try_lookahead
= 0;
1740 int show_function
= 0;
1741 struct userdiff_driver
*textconv
= NULL
;
1742 enum grep_context ctx
= GREP_CONTEXT_HEAD
;
1745 if (!opt
->status_only
&& gs
->name
== NULL
)
1746 BUG("grep call which could print a name requires "
1747 "grep_source.name be non-NULL");
1750 opt
->output
= std_output
;
1752 if (opt
->pre_context
|| opt
->post_context
|| opt
->file_break
||
1754 /* Show hunk marks, except for the first file. */
1755 if (opt
->last_shown
)
1756 opt
->show_hunk_mark
= 1;
1758 * If we're using threads then we can't easily identify
1759 * the first file. Always put hunk marks in that case
1760 * and skip the very first one later in work_done().
1762 if (opt
->output
!= std_output
)
1763 opt
->show_hunk_mark
= 1;
1765 opt
->last_shown
= 0;
1767 if (opt
->allow_textconv
) {
1768 grep_source_load_driver(gs
, opt
->repo
->index
);
1770 * We might set up the shared textconv cache data here, which
1771 * is not thread-safe.
1774 textconv
= userdiff_get_textconv(opt
->repo
, gs
->driver
);
1779 * We know the result of a textconv is text, so we only have to care
1780 * about binary handling if we are not using it.
1783 switch (opt
->binary
) {
1784 case GREP_BINARY_DEFAULT
:
1785 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1786 binary_match_only
= 1;
1788 case GREP_BINARY_NOMATCH
:
1789 if (grep_source_is_binary(gs
, opt
->repo
->index
))
1790 return 0; /* Assume unmatch */
1792 case GREP_BINARY_TEXT
:
1795 BUG("unknown binary handling mode");
1799 memset(&xecfg
, 0, sizeof(xecfg
));
1802 try_lookahead
= should_lookahead(opt
);
1804 if (fill_textconv_grep(opt
->repo
, textconv
, gs
) < 0)
1813 ssize_t col
= -1, icol
= -1;
1816 * look_ahead() skips quickly to the line that possibly
1817 * has the next hit; don't call it if we need to do
1818 * something more than just skipping the current line
1819 * in response to an unmatch for the current line. E.g.
1820 * inside a post-context window, we will show the current
1821 * line as a context around the previous hit when it
1826 && (show_function
||
1827 lno
<= last_hit
+ opt
->post_context
))
1828 && look_ahead(opt
, &left
, &lno
, &bol
))
1830 eol
= end_of_line(bol
, &left
);
1834 if ((ctx
== GREP_CONTEXT_HEAD
) && (eol
== bol
))
1835 ctx
= GREP_CONTEXT_BODY
;
1837 hit
= match_line(opt
, bol
, eol
, &col
, &icol
, ctx
, collect_hits
);
1843 /* "grep -v -e foo -e bla" should list lines
1844 * that do not have either, so inversion should
1849 if (opt
->unmatch_name_only
) {
1856 if (opt
->status_only
)
1858 if (opt
->name_only
) {
1859 show_name(opt
, gs
->name
);
1864 if (binary_match_only
) {
1865 opt
->output(opt
, "Binary file ", 12);
1866 output_color(opt
, gs
->name
, strlen(gs
->name
),
1867 opt
->colors
[GREP_COLOR_FILENAME
]);
1868 opt
->output(opt
, " matches\n", 9);
1871 /* Hit at this line. If we haven't shown the
1872 * pre-context lines, we would need to show them.
1874 if (opt
->pre_context
|| opt
->funcbody
)
1875 show_pre_context(opt
, gs
, bol
, eol
, lno
);
1876 else if (opt
->funcname
)
1877 show_funcname_line(opt
, gs
, bol
, lno
);
1878 cno
= opt
->invert
? icol
: col
;
1881 * A negative cno indicates that there was no
1882 * match on the line. We are thus inverted and
1883 * being asked to show all lines that _don't_
1884 * match a given expression. Therefore, set cno
1885 * to 0 to suggest the whole line matches.
1889 show_line(opt
, bol
, eol
, gs
->name
, lno
, cno
+ 1, ':');
1895 if (show_function
&& (!peek_bol
|| peek_bol
< bol
)) {
1896 unsigned long peek_left
= left
;
1897 char *peek_eol
= eol
;
1900 * Trailing empty lines are not interesting.
1901 * Peek past them to see if they belong to the
1902 * body of the current function.
1905 while (is_empty_line(peek_bol
, peek_eol
)) {
1906 peek_bol
= peek_eol
+ 1;
1907 peek_eol
= end_of_line(peek_bol
, &peek_left
);
1910 if (match_funcname(opt
, gs
, peek_bol
, peek_eol
))
1913 if (show_function
||
1914 (last_hit
&& lno
<= last_hit
+ opt
->post_context
)) {
1915 /* If the last hit is within the post context,
1916 * we need to show this line.
1918 show_line(opt
, bol
, eol
, gs
->name
, lno
, col
+ 1, '-');
1932 if (opt
->status_only
)
1933 return opt
->unmatch_name_only
;
1934 if (opt
->unmatch_name_only
) {
1935 /* We did not see any hit, so we want to show this */
1936 show_name(opt
, gs
->name
);
1940 xdiff_clear_find_func(&xecfg
);
1944 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1945 * which feels mostly useless but sometimes useful. Maybe
1946 * make it another option? For now suppress them.
1948 if (opt
->count
&& count
) {
1950 if (opt
->pathname
) {
1951 output_color(opt
, gs
->name
, strlen(gs
->name
),
1952 opt
->colors
[GREP_COLOR_FILENAME
]);
1953 output_sep(opt
, ':');
1955 xsnprintf(buf
, sizeof(buf
), "%u\n", count
);
1956 opt
->output(opt
, buf
, strlen(buf
));
1962 static void clr_hit_marker(struct grep_expr
*x
)
1964 /* All-hit markers are meaningful only at the very top level
1969 if (x
->node
!= GREP_NODE_OR
)
1971 x
->u
.binary
.left
->hit
= 0;
1972 x
= x
->u
.binary
.right
;
1976 static int chk_hit_marker(struct grep_expr
*x
)
1978 /* Top level nodes have hit markers. See if they all are hits */
1980 if (x
->node
!= GREP_NODE_OR
)
1982 if (!x
->u
.binary
.left
->hit
)
1984 x
= x
->u
.binary
.right
;
1988 int grep_source(struct grep_opt
*opt
, struct grep_source
*gs
)
1991 * we do not have to do the two-pass grep when we do not check
1992 * buffer-wide "all-match".
1994 if (!opt
->all_match
)
1995 return grep_source_1(opt
, gs
, 0);
1997 /* Otherwise the toplevel "or" terms hit a bit differently.
1998 * We first clear hit markers from them.
2000 clr_hit_marker(opt
->pattern_expression
);
2001 grep_source_1(opt
, gs
, 1);
2003 if (!chk_hit_marker(opt
->pattern_expression
))
2006 return grep_source_1(opt
, gs
, 0);
2009 int grep_buffer(struct grep_opt
*opt
, char *buf
, unsigned long size
)
2011 struct grep_source gs
;
2014 grep_source_init(&gs
, GREP_SOURCE_BUF
, NULL
, NULL
, NULL
);
2018 r
= grep_source(opt
, &gs
);
2020 grep_source_clear(&gs
);
2024 void grep_source_init(struct grep_source
*gs
, enum grep_source_type type
,
2025 const char *name
, const char *path
,
2026 const void *identifier
)
2029 gs
->name
= xstrdup_or_null(name
);
2030 gs
->path
= xstrdup_or_null(path
);
2036 case GREP_SOURCE_FILE
:
2037 gs
->identifier
= xstrdup(identifier
);
2039 case GREP_SOURCE_OID
:
2040 gs
->identifier
= oiddup(identifier
);
2042 case GREP_SOURCE_BUF
:
2043 gs
->identifier
= NULL
;
2048 void grep_source_clear(struct grep_source
*gs
)
2050 FREE_AND_NULL(gs
->name
);
2051 FREE_AND_NULL(gs
->path
);
2052 FREE_AND_NULL(gs
->identifier
);
2053 grep_source_clear_data(gs
);
2056 void grep_source_clear_data(struct grep_source
*gs
)
2059 case GREP_SOURCE_FILE
:
2060 case GREP_SOURCE_OID
:
2061 FREE_AND_NULL(gs
->buf
);
2064 case GREP_SOURCE_BUF
:
2065 /* leave user-provided buf intact */
2070 static int grep_source_load_oid(struct grep_source
*gs
)
2072 enum object_type type
;
2075 gs
->buf
= read_object_file(gs
->identifier
, &type
, &gs
->size
);
2079 return error(_("'%s': unable to read %s"),
2081 oid_to_hex(gs
->identifier
));
2085 static int grep_source_load_file(struct grep_source
*gs
)
2087 const char *filename
= gs
->identifier
;
2093 if (lstat(filename
, &st
) < 0) {
2095 if (errno
!= ENOENT
)
2096 error_errno(_("failed to stat '%s'"), filename
);
2099 if (!S_ISREG(st
.st_mode
))
2101 size
= xsize_t(st
.st_size
);
2102 i
= open(filename
, O_RDONLY
);
2105 data
= xmallocz(size
);
2106 if (st
.st_size
!= read_in_full(i
, data
, size
)) {
2107 error_errno(_("'%s': short read"), filename
);
2119 static int grep_source_load(struct grep_source
*gs
)
2125 case GREP_SOURCE_FILE
:
2126 return grep_source_load_file(gs
);
2127 case GREP_SOURCE_OID
:
2128 return grep_source_load_oid(gs
);
2129 case GREP_SOURCE_BUF
:
2130 return gs
->buf
? 0 : -1;
2132 BUG("invalid grep_source type to load");
2135 void grep_source_load_driver(struct grep_source
*gs
,
2136 struct index_state
*istate
)
2143 gs
->driver
= userdiff_find_by_path(istate
, gs
->path
);
2145 gs
->driver
= userdiff_find_by_name("default");
2149 static int grep_source_is_binary(struct grep_source
*gs
,
2150 struct index_state
*istate
)
2152 grep_source_load_driver(gs
, istate
);
2153 if (gs
->driver
->binary
!= -1)
2154 return gs
->driver
->binary
;
2156 if (!grep_source_load(gs
))
2157 return buffer_is_binary(gs
->buf
, gs
->size
);