pathspec: support :(glob) syntax
:(glob)path differs from plain pathspec that it uses wildmatch with WM_PATHNAME while the other uses fnmatch without FNM_PATHNAME. The difference lies in how '*' (and '**') is processed. With the introduction of :(glob) and :(literal) and their global options --[no]glob-pathspecs, the user can: - make everything literal by default via --noglob-pathspecs --literal-pathspecs cannot be used for this purpose as it disables _all_ pathspec magic. - individually turn on globbing with :(glob) - make everything globbing by default via --glob-pathspecs - individually turn off globbing with :(literal) The implication behind this is, there is no way to gain the default matching behavior (i.e. fnmatch without FNM_PATHNAME). You either get new globbing or literal. The old fnmatch behavior is considered deprecated and discouraged to use. Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Nguyễn Thái Ngọc Duy committed
Jul 14, 2013 at 15:36 UTC
bd30c2e48432c692f9e77d3529c9cf25117066bb
12 files changed
+198
-31
Documentation/git.txt
+19
@@ -454,6 +454,17 @@ help ...`.
454
This is equivalent to setting the `GIT_LITERAL_PATHSPECS` environment
455
variable to `1`.
456
457
+--glob-pathspecs:
458
+ Add "glob" magic to all pathspec. This is equivalent to setting
459
+ the `GIT_GLOB_PATHSPECS` environment variable to `1`. Disabling
460
+ globbing on individual pathspecs can be done using pathspec
461
+ magic ":(literal)"
462
+
463
+--noglob-pathspecs:
464
+ Add "literal" magic to all pathspec. This is equivalent to setting
465
+ the `GIT_NOGLOB_PATHSPECS` environment variable to `1`. Enabling
466
+ globbing on individual pathspecs can be done using pathspec
467
+ magic ":(glob)"
468
469
GIT COMMANDS
470
------------
@@ -860,6 +871,14 @@ GIT_LITERAL_PATHSPECS::
871
literal paths to Git (e.g., paths previously given to you by
872
`git ls-tree`, `--raw` diff output, etc).
873
874
+GIT_GLOB_PATHSPECS::
875
+ Setting this variable to `1` will cause Git to treat all
876
+ pathspecs as glob patterns (aka "glob" magic).
877
+
878
+GIT_NOGLOB_PATHSPECS::
879
+ Setting this variable to `1` will cause Git to treat all
880
+ pathspecs as literal (aka "literal" magic).
881
+
882
883
Discussion[[Discussion]]
884
------------------------
Documentation/glossary-content.txt
+29
@@ -333,6 +333,35 @@ top `/`;;
333
literal;;
334
Wildcards in the pattern such as `*` or `?` are treated
335
as literal characters.
336
+
337
+glob;;
338
+ Git treats the pattern as a shell glob suitable for
339
+ consumption by fnmatch(3) with the FNM_PATHNAME flag:
340
+ wildcards in the pattern will not match a / in the pathname.
341
+ For example, "Documentation/{asterisk}.html" matches
342
+ "Documentation/git.html" but not "Documentation/ppc/ppc.html"
343
+ or "tools/perf/Documentation/perf.html".
344
++
345
+Two consecutive asterisks ("`**`") in patterns matched against
346
+full pathname may have special meaning:
347
+
348
+ - A leading "`**`" followed by a slash means match in all
349
+ directories. For example, "`**/foo`" matches file or directory
350
+ "`foo`" anywhere, the same as pattern "`foo`". "**/foo/bar"
351
+ matches file or directory "`bar`" anywhere that is directly
352
+ under directory "`foo`".
353
+
354
+ - A trailing "/**" matches everything inside. For example,
355
+ "abc/**" matches all files inside directory "abc", relative
356
+ to the location of the `.gitignore` file, with infinite depth.
357
+
358
+ - A slash followed by two consecutive asterisks then a slash
359
+ matches zero or more directories. For example, "`a/**/b`"
360
+ matches "`a/b`", "`a/x/b`", "`a/x/y/b`" and so on.
361
+
362
+ - Other consecutive asterisks are considered invalid.
363
++
364
+Glob magic is incompatible with literal magic.
365
--
366
+
367
Currently only the slash `/` is recognized as the "magic signature",
builtin/add.c
+7
-2
@@ -541,11 +541,16 @@ int cmd_add(int argc, const char **argv, const char *prefix)
541
/*
542
* file_exists() assumes exact match
543
*/
544
- GUARD_PATHSPEC(&pathspec, PATHSPEC_FROMTOP | PATHSPEC_LITERAL);
544
+ GUARD_PATHSPEC(&pathspec,
545
+ PATHSPEC_FROMTOP |
546
+ PATHSPEC_LITERAL |
547
+ PATHSPEC_GLOB);
548
549
for (i = 0; i < pathspec.nr; i++) {
550
const char *path = pathspec.items[i].match;
548
- if (!seen[i] && !file_exists(path)) {
551
+ if (!seen[i] &&
552
+ ((pathspec.items[i].magic & PATHSPEC_GLOB) ||
553
+ !file_exists(path))) {
554
if (ignore_missing) {
555
int dtype = DT_UNKNOWN;
556
if (is_excluded(&dir, path, &dtype))
builtin/ls-tree.c
+1
-1
@@ -173,7 +173,7 @@ int cmd_ls_tree(int argc, const char **argv, const char *prefix)
173
* cannot be lifted until it is converted to use
174
* match_pathspec_depth() or tree_entry_interesting()
175
*/
176
- parse_pathspec(&pathspec, 0,
176
+ parse_pathspec(&pathspec, PATHSPEC_GLOB,
177
PATHSPEC_PREFER_CWD,
178
prefix, argv + 1);
179
for (i = 0; i < pathspec.nr; i++)
cache.h
+2
@@ -367,6 +367,8 @@ static inline enum object_type object_type(unsigned int mode)
367
#define GIT_NOTES_REWRITE_REF_ENVIRONMENT "GIT_NOTES_REWRITE_REF"
368
#define GIT_NOTES_REWRITE_MODE_ENVIRONMENT "GIT_NOTES_REWRITE_MODE"
369
#define GIT_LITERAL_PATHSPECS_ENVIRONMENT "GIT_LITERAL_PATHSPECS"
370
+#define GIT_GLOB_PATHSPECS_ENVIRONMENT "GIT_GLOB_PATHSPECS"
371
+#define GIT_NOGLOB_PATHSPECS_ENVIRONMENT "GIT_NOGLOB_PATHSPECS"
372
373
/*
374
* This environment variable is expected to contain a boolean indicating
dir.c
+16
-12
@@ -52,26 +52,28 @@ int fnmatch_icase(const char *pattern, const char *string, int flags)
52
return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
53
}
54
55
-inline int git_fnmatch(const char *pattern, const char *string,
56
- int flags, int prefix)
55
+inline int git_fnmatch(const struct pathspec_item *item,
56
+ const char *pattern, const char *string,
57
+ int prefix)
58
{
58
- int fnm_flags = 0;
59
- if (flags & GFNM_PATHNAME)
60
- fnm_flags |= FNM_PATHNAME;
59
if (prefix > 0) {
60
if (strncmp(pattern, string, prefix))
61
return FNM_NOMATCH;
62
pattern += prefix;
63
string += prefix;
64
}
67
- if (flags & GFNM_ONESTAR) {
65
+ if (item->flags & PATHSPEC_ONESTAR) {
66
int pattern_len = strlen(++pattern);
67
int string_len = strlen(string);
68
return string_len < pattern_len ||
69
strcmp(pattern,
70
string + string_len - pattern_len);
71
}
74
- return fnmatch(pattern, string, fnm_flags);
72
+ if (item->magic & PATHSPEC_GLOB)
73
+ return wildmatch(pattern, string, WM_PATHNAME, NULL);
74
+ else
75
+ /* wildmatch has not learned no FNM_PATHNAME mode yet */
76
+ return fnmatch(pattern, string, 0);
77
}
78
79
static int fnmatch_icase_mem(const char *pattern, int patternlen,
@@ -111,7 +113,8 @@ static size_t common_prefix_len(const struct pathspec *pathspec)
113
GUARD_PATHSPEC(pathspec,
114
PATHSPEC_FROMTOP |
115
PATHSPEC_MAXDEPTH |
114
- PATHSPEC_LITERAL);
116
+ PATHSPEC_LITERAL |
117
+ PATHSPEC_GLOB);
118
119
for (n = 0; n < pathspec->nr; n++) {
120
size_t i = 0, len = 0;
@@ -206,8 +209,7 @@ static int match_pathspec_item(const struct pathspec_item *item, int prefix,
209
}
210
211
if (item->nowildcard_len < item->len &&
209
- !git_fnmatch(match, name,
210
- item->flags & PATHSPEC_ONESTAR ? GFNM_ONESTAR : 0,
212
+ !git_fnmatch(item, match, name,
213
item->nowildcard_len - prefix))
214
return MATCHED_FNMATCH;
215
@@ -238,7 +240,8 @@ int match_pathspec_depth(const struct pathspec *ps,
240
GUARD_PATHSPEC(ps,
241
PATHSPEC_FROMTOP |
242
PATHSPEC_MAXDEPTH |
241
- PATHSPEC_LITERAL);
243
+ PATHSPEC_LITERAL |
244
+ PATHSPEC_GLOB);
245
246
if (!ps->nr) {
247
if (!ps->recursive ||
@@ -1297,7 +1300,8 @@ int read_directory(struct dir_struct *dir, const char *path, int len, const stru
1300
GUARD_PATHSPEC(pathspec,
1301
PATHSPEC_FROMTOP |
1302
PATHSPEC_MAXDEPTH |
1300
- PATHSPEC_LITERAL);
1303
+ PATHSPEC_LITERAL |
1304
+ PATHSPEC_GLOB);
1305
1306
if (has_symlink_leading_path(path, len))
1307
return dir->nr;
dir.h
+4
-5
@@ -199,10 +199,9 @@ extern int fnmatch_icase(const char *pattern, const char *string, int flags);
199
/*
200
* The prefix part of pattern must not contains wildcards.
201
*/
202
-#define GFNM_PATHNAME 1 /* similar to FNM_PATHNAME */
203
-#define GFNM_ONESTAR 2 /* there is only _one_ wildcard, a star */
204
-
205
-extern int git_fnmatch(const char *pattern, const char *string,
206
- int flags, int prefix);
202
+struct pathspec_item;
203
+extern int git_fnmatch(const struct pathspec_item *item,
204
+ const char *pattern, const char *string,
205
+ int prefix);
206
207
#endif
git.c
+8
@@ -147,6 +147,14 @@ static int handle_options(const char ***argv, int *argc, int *envchanged)
147
setenv(GIT_LITERAL_PATHSPECS_ENVIRONMENT, "0", 1);
148
if (envchanged)
149
*envchanged = 1;
150
+ } else if (!strcmp(cmd, "--glob-pathspecs")) {
151
+ setenv(GIT_GLOB_PATHSPECS_ENVIRONMENT, "1", 1);
152
+ if (envchanged)
153
+ *envchanged = 1;
154
+ } else if (!strcmp(cmd, "--noglob-pathspecs")) {
155
+ setenv(GIT_NOGLOB_PATHSPECS_ENVIRONMENT, "1", 1);
156
+ if (envchanged)
157
+ *envchanged = 1;
158
} else if (!strcmp(cmd, "--shallow-file")) {
159
(*argv)++;
160
(*argc)--;
pathspec.c
+42
-5
@@ -57,7 +57,6 @@ char *find_pathspecs_matching_against_index(const struct pathspec *pathspec)
57
*
58
* Possible future magic semantics include stuff like:
59
*
60
- * { PATHSPEC_NOGLOB, '!', "noglob" },
60
* { PATHSPEC_ICASE, '\0', "icase" },
61
* { PATHSPEC_RECURSIVE, '*', "recursive" },
62
* { PATHSPEC_REGEXP, '\0', "regexp" },
@@ -71,6 +70,7 @@ static struct pathspec_magic {
70
} pathspec_magic[] = {
71
{ PATHSPEC_FROMTOP, '/', "top" },
72
{ PATHSPEC_LITERAL, 0, "literal" },
73
+ { PATHSPEC_GLOB, '\0', "glob" },
74
};
75
76
/*
@@ -93,6 +93,8 @@ static unsigned prefix_pathspec(struct pathspec_item *item,
93
const char *elt)
94
{
95
static int literal_global = -1;
96
+ static int glob_global = -1;
97
+ static int noglob_global = -1;
98
unsigned magic = 0, short_magic = 0, global_magic = 0;
99
const char *copyfrom = elt, *long_magic_end = NULL;
100
char *match;
@@ -103,6 +105,22 @@ static unsigned prefix_pathspec(struct pathspec_item *item,
105
if (literal_global)
106
global_magic |= PATHSPEC_LITERAL;
107
108
+ if (glob_global < 0)
109
+ glob_global = git_env_bool(GIT_GLOB_PATHSPECS_ENVIRONMENT, 0);
110
+ if (glob_global)
111
+ global_magic |= PATHSPEC_GLOB;
112
+
113
+ if (noglob_global < 0)
114
+ noglob_global = git_env_bool(GIT_NOGLOB_PATHSPECS_ENVIRONMENT, 0);
115
+
116
+ if (glob_global && noglob_global)
117
+ die(_("global 'glob' and 'noglob' pathspec settings are incompatible"));
118
+
119
+ if ((global_magic & PATHSPEC_LITERAL) &&
120
+ (global_magic & ~PATHSPEC_LITERAL))
121
+ die(_("global 'literal' pathspec setting is incompatible "
122
+ "with all other global pathspec settings"));
123
+
124
if (elt[0] != ':' || literal_global) {
125
; /* nothing to do */
126
} else if (elt[1] == '(') {
@@ -167,12 +185,24 @@ static unsigned prefix_pathspec(struct pathspec_item *item,
185
186
magic |= short_magic;
187
*p_short_magic = short_magic;
188
+
189
+ /* --noglob-pathspec adds :(literal) _unless_ :(glob) is specifed */
190
+ if (noglob_global && !(magic & PATHSPEC_GLOB))
191
+ global_magic |= PATHSPEC_LITERAL;
192
+
193
+ /* --glob-pathspec is overriden by :(literal) */
194
+ if ((global_magic & PATHSPEC_GLOB) && (magic & PATHSPEC_LITERAL))
195
+ global_magic &= ~PATHSPEC_GLOB;
196
+
197
magic |= global_magic;
198
199
if (pathspec_prefix >= 0 &&
200
(prefixlen || (prefix && *prefix)))
201
die("BUG: 'prefix' magic is supposed to be used at worktree's root");
202
203
+ if ((magic & PATHSPEC_LITERAL) && (magic & PATHSPEC_GLOB))
204
+ die(_("%s: 'literal' and 'glob' are incompatible"), elt);
205
+
206
if (pathspec_prefix >= 0) {
207
match = xstrdup(copyfrom);
208
prefixlen = pathspec_prefix;
@@ -248,10 +278,17 @@ static unsigned prefix_pathspec(struct pathspec_item *item,
278
item->nowildcard_len = prefixlen;
279
}
280
item->flags = 0;
251
- if (item->nowildcard_len < item->len &&
252
- item->match[item->nowildcard_len] == '*' &&
253
- no_wildcard(item->match + item->nowildcard_len + 1))
254
- item->flags |= PATHSPEC_ONESTAR;
281
+ if (magic & PATHSPEC_GLOB) {
282
+ /*
283
+ * FIXME: should we enable ONESTAR in _GLOB for
284
+ * pattern "* * / * . c"?
285
+ */
286
+ } else {
287
+ if (item->nowildcard_len < item->len &&
288
+ item->match[item->nowildcard_len] == '*' &&
289
+ no_wildcard(item->match + item->nowildcard_len + 1))
290
+ item->flags |= PATHSPEC_ONESTAR;
291
+ }
292
293
/* sanity checks, pathspec matchers assume these are sane */
294
assert(item->nowildcard_len <= item->len &&
pathspec.h
+3
-1
@@ -5,10 +5,12 @@
5
#define PATHSPEC_FROMTOP (1<<0)
6
#define PATHSPEC_MAXDEPTH (1<<1)
7
#define PATHSPEC_LITERAL (1<<2)
8
+#define PATHSPEC_GLOB (1<<3)
9
#define PATHSPEC_ALL_MAGIC \
10
(PATHSPEC_FROMTOP | \
11
PATHSPEC_MAXDEPTH | \
11
- PATHSPEC_LITERAL)
12
+ PATHSPEC_LITERAL | \
13
+ PATHSPEC_GLOB)
14
15
#define PATHSPEC_ONESTAR 1 /* the pathspec pattern sastisfies GFNM_ONESTAR */
16
t/t6130-pathspec-noglob.sh
+63
@@ -32,6 +32,16 @@ test_expect_success 'star pathspec globs' '
32
test_cmp expect actual
33
'
34
35
+test_expect_success 'star pathspec globs' '
36
+ cat >expect <<-\EOF &&
37
+ bracket
38
+ star
39
+ vanilla
40
+ EOF
41
+ git log --format=%s -- ":(glob)f*" >actual &&
42
+ test_cmp expect actual
43
+'
44
+
45
test_expect_success 'bracket pathspec globs and matches literal brackets' '
46
cat >expect <<-\EOF &&
47
bracket
@@ -41,6 +51,15 @@ test_expect_success 'bracket pathspec globs and matches literal brackets' '
51
test_cmp expect actual
52
'
53
54
+test_expect_success 'bracket pathspec globs and matches literal brackets' '
55
+ cat >expect <<-\EOF &&
56
+ bracket
57
+ vanilla
58
+ EOF
59
+ git log --format=%s -- ":(glob)f[o][o]" >actual &&
60
+ test_cmp expect actual
61
+'
62
+
63
test_expect_success 'no-glob option matches literally (vanilla)' '
64
echo vanilla >expect &&
65
git --literal-pathspecs log --format=%s -- foo >actual &&
@@ -89,4 +108,48 @@ test_expect_success 'no-glob environment variable works' '
108
test_cmp expect actual
109
'
110
111
+test_expect_success 'setup xxx/bar' '
112
+ mkdir xxx &&
113
+ test_commit xxx xxx/bar
114
+'
115
+
116
+test_expect_success '**/ works with :(glob)' '
117
+ cat >expect <<-\EOF &&
118
+ xxx
119
+ unrelated
120
+ EOF
121
+ git log --format=%s -- ":(glob)**/bar" >actual &&
122
+ test_cmp expect actual
123
+'
124
+
125
+test_expect_success '**/ does not work with --noglob-pathspecs' '
126
+ : >expect &&
127
+ git --noglob-pathspecs log --format=%s -- "**/bar" >actual &&
128
+ test_cmp expect actual
129
+'
130
+
131
+test_expect_success '**/ works with :(glob) and --noglob-pathspecs' '
132
+ cat >expect <<-\EOF &&
133
+ xxx
134
+ unrelated
135
+ EOF
136
+ git --noglob-pathspecs log --format=%s -- ":(glob)**/bar" >actual &&
137
+ test_cmp expect actual
138
+'
139
+
140
+test_expect_success '**/ works with --glob-pathspecs' '
141
+ cat >expect <<-\EOF &&
142
+ xxx
143
+ unrelated
144
+ EOF
145
+ git --glob-pathspecs log --format=%s -- "**/bar" >actual &&
146
+ test_cmp expect actual
147
+'
148
+
149
+test_expect_success '**/ does not work with :(literal) and --glob-pathspecs' '
150
+ : >expect &&
151
+ git --glob-pathspecs log --format=%s -- ":(literal)**/bar" >actual &&
152
+ test_cmp expect actual
153
+'
154
+
155
test_done
tree-walk.c
+4
-5
@@ -639,7 +639,8 @@ enum interesting tree_entry_interesting(const struct name_entry *entry,
639
GUARD_PATHSPEC(ps,
640
PATHSPEC_FROMTOP |
641
PATHSPEC_MAXDEPTH |
642
- PATHSPEC_LITERAL);
642
+ PATHSPEC_LITERAL |
643
+ PATHSPEC_GLOB);
644
645
if (!ps->nr) {
646
if (!ps->recursive ||
@@ -685,8 +686,7 @@ enum interesting tree_entry_interesting(const struct name_entry *entry,
686
return entry_interesting;
687
688
if (item->nowildcard_len < item->len) {
688
- if (!git_fnmatch(match + baselen, entry->path,
689
- item->flags & PATHSPEC_ONESTAR ? GFNM_ONESTAR : 0,
689
+ if (!git_fnmatch(item, match + baselen, entry->path,
690
item->nowildcard_len - baselen))
691
return entry_interesting;
692
@@ -727,8 +727,7 @@ match_wildcards:
727
728
strbuf_add(base, entry->path, pathlen);
729
730
- if (!git_fnmatch(match, base->buf + base_offset,
731
- item->flags & PATHSPEC_ONESTAR ? GFNM_ONESTAR : 0,
730
+ if (!git_fnmatch(item, match, base->buf + base_offset,
731
item->nowildcard_len)) {
732
strbuf_setlen(base, base_offset + baselen);
733
return entry_interesting;