/src/git/diffcore-pickaxe.c
Line | Count | Source |
1 | | /* |
2 | | * Copyright (C) 2005 Junio C Hamano |
3 | | * Copyright (C) 2010 Google Inc. |
4 | | */ |
5 | | |
6 | | #define DISABLE_SIGN_COMPARE_WARNINGS |
7 | | |
8 | | #include "git-compat-util.h" |
9 | | #include "diff.h" |
10 | | #include "diffcore.h" |
11 | | #include "xdiff-interface.h" |
12 | | #include "kwset.h" |
13 | | #include "oidset.h" |
14 | | #include "pretty.h" |
15 | | #include "quote.h" |
16 | | |
17 | | typedef int (*pickaxe_fn)(mmfile_t *one, mmfile_t *two, |
18 | | struct diff_options *o, |
19 | | regex_t *regexp, kwset_t kws); |
20 | | |
21 | | struct diffgrep_cb { |
22 | | regex_t *regexp; |
23 | | int hit; |
24 | | }; |
25 | | |
26 | | static int diffgrep_consume(void *priv, char *line, unsigned long len) |
27 | 0 | { |
28 | 0 | struct diffgrep_cb *data = priv; |
29 | 0 | regmatch_t regmatch; |
30 | |
|
31 | 0 | if (line[0] != '+' && line[0] != '-') |
32 | 0 | return 0; |
33 | 0 | if (data->hit) |
34 | 0 | BUG("Already matched in diffgrep_consume! Broken xdiff_emit_line_fn?"); |
35 | 0 | if (!regexec_buf(data->regexp, line + 1, len - 1, 1, |
36 | 0 | ®match, 0)) { |
37 | 0 | data->hit = 1; |
38 | 0 | return 1; |
39 | 0 | } |
40 | 0 | return 0; |
41 | 0 | } |
42 | | |
43 | | static int diff_grep(mmfile_t *one, mmfile_t *two, |
44 | | struct diff_options *o, |
45 | | regex_t *regexp, kwset_t kws UNUSED) |
46 | 0 | { |
47 | 0 | struct diffgrep_cb ecbdata; |
48 | 0 | xpparam_t xpp; |
49 | 0 | xdemitconf_t xecfg; |
50 | 0 | int ret; |
51 | | |
52 | | /* |
53 | | * We have both sides; need to run textual diff and see if |
54 | | * the pattern appears on added/deleted lines. |
55 | | */ |
56 | 0 | memset(&xpp, 0, sizeof(xpp)); |
57 | 0 | memset(&xecfg, 0, sizeof(xecfg)); |
58 | 0 | ecbdata.regexp = regexp; |
59 | 0 | ecbdata.hit = 0; |
60 | 0 | xecfg.flags = XDL_EMIT_NO_HUNK_HDR; |
61 | 0 | xecfg.ctxlen = o->context; |
62 | 0 | xecfg.interhunkctxlen = o->interhunkcontext; |
63 | | |
64 | | /* |
65 | | * An xdiff error might be our "data->hit" from above. See the |
66 | | * comment for xdiff_emit_line_fn in xdiff-interface.h |
67 | | */ |
68 | 0 | ret = xdi_diff_outf(one, two, NULL, diffgrep_consume, |
69 | 0 | &ecbdata, &xpp, &xecfg); |
70 | 0 | if (ecbdata.hit) |
71 | 0 | return 1; |
72 | 0 | if (ret) |
73 | 0 | return ret; |
74 | 0 | return 0; |
75 | 0 | } |
76 | | |
77 | | static unsigned int contains(mmfile_t *mf, regex_t *regexp, kwset_t kws, |
78 | | unsigned int limit) |
79 | 0 | { |
80 | 0 | unsigned int cnt = 0; |
81 | 0 | unsigned long sz = mf->size; |
82 | 0 | const char *data = mf->ptr; |
83 | |
|
84 | 0 | if (regexp) { |
85 | 0 | regmatch_t regmatch; |
86 | 0 | int flags = 0; |
87 | |
|
88 | 0 | while (sz && |
89 | 0 | !regexec_buf(regexp, data, sz, 1, ®match, flags)) { |
90 | 0 | flags |= REG_NOTBOL; |
91 | 0 | data += regmatch.rm_eo; |
92 | 0 | sz -= regmatch.rm_eo; |
93 | 0 | if (sz && regmatch.rm_so == regmatch.rm_eo) { |
94 | 0 | data++; |
95 | 0 | sz--; |
96 | 0 | } |
97 | 0 | cnt++; |
98 | |
|
99 | 0 | if (limit && cnt == limit) |
100 | 0 | return cnt; |
101 | 0 | } |
102 | |
|
103 | 0 | } else { /* Classic exact string match */ |
104 | 0 | while (sz) { |
105 | 0 | struct kwsmatch kwsm; |
106 | 0 | size_t offset = kwsexec(kws, data, sz, &kwsm); |
107 | 0 | if (offset == -1) |
108 | 0 | break; |
109 | 0 | sz -= offset + kwsm.size[0]; |
110 | 0 | data += offset + kwsm.size[0]; |
111 | 0 | cnt++; |
112 | |
|
113 | 0 | if (limit && cnt == limit) |
114 | 0 | return cnt; |
115 | 0 | } |
116 | 0 | } |
117 | 0 | return cnt; |
118 | 0 | } |
119 | | |
120 | | static int has_changes(mmfile_t *one, mmfile_t *two, |
121 | | struct diff_options *o UNUSED, |
122 | | regex_t *regexp, kwset_t kws) |
123 | 0 | { |
124 | 0 | unsigned int c1 = one ? contains(one, regexp, kws, 0) : 0; |
125 | 0 | unsigned int c2 = two ? contains(two, regexp, kws, c1 + 1) : 0; |
126 | 0 | return c1 != c2; |
127 | 0 | } |
128 | | |
129 | | static int pickaxe_match(struct diff_filepair *p, struct diff_options *o, |
130 | | regex_t *regexp, kwset_t kws, pickaxe_fn fn) |
131 | 0 | { |
132 | 0 | struct userdiff_driver *textconv_one = NULL; |
133 | 0 | struct userdiff_driver *textconv_two = NULL; |
134 | 0 | mmfile_t mf1, mf2; |
135 | 0 | int ret; |
136 | | |
137 | | /* ignore unmerged */ |
138 | 0 | if (!DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) |
139 | 0 | return 0; |
140 | | |
141 | 0 | if (o->objfind) { |
142 | 0 | return (DIFF_FILE_VALID(p->one) && |
143 | 0 | oidset_contains(o->objfind, &p->one->oid)) || |
144 | 0 | (DIFF_FILE_VALID(p->two) && |
145 | 0 | oidset_contains(o->objfind, &p->two->oid)); |
146 | 0 | } |
147 | | |
148 | 0 | if (o->flags.allow_textconv) { |
149 | 0 | textconv_one = get_textconv(o->repo, p->one); |
150 | 0 | textconv_two = get_textconv(o->repo, p->two); |
151 | 0 | } |
152 | | |
153 | | /* |
154 | | * If we have an unmodified pair, we know that the count will be the |
155 | | * same and don't even have to load the blobs. Unless textconv is in |
156 | | * play, _and_ we are using two different textconv filters (e.g., |
157 | | * because a pair is an exact rename with different textconv attributes |
158 | | * for each side, which might generate different content). |
159 | | */ |
160 | 0 | if (textconv_one == textconv_two && diff_unmodified_pair(p)) |
161 | 0 | return 0; |
162 | | |
163 | 0 | if ((o->pickaxe_opts & DIFF_PICKAXE_KIND_G) && |
164 | 0 | !o->flags.text && |
165 | 0 | ((!textconv_one && diff_filespec_is_binary(o->repo, p->one)) || |
166 | 0 | (!textconv_two && diff_filespec_is_binary(o->repo, p->two)))) |
167 | 0 | return 0; |
168 | | |
169 | 0 | mf1.size = fill_textconv(o->repo, textconv_one, p->one, &mf1.ptr); |
170 | 0 | mf2.size = fill_textconv(o->repo, textconv_two, p->two, &mf2.ptr); |
171 | |
|
172 | 0 | ret = fn(&mf1, &mf2, o, regexp, kws); |
173 | |
|
174 | 0 | if (textconv_one) |
175 | 0 | free(mf1.ptr); |
176 | 0 | if (textconv_two) |
177 | 0 | free(mf2.ptr); |
178 | 0 | diff_free_filespec_data(p->one); |
179 | 0 | diff_free_filespec_data(p->two); |
180 | |
|
181 | 0 | return ret; |
182 | 0 | } |
183 | | |
184 | | static void pickaxe(struct diff_queue_struct *q, struct diff_options *o, |
185 | | regex_t *regexp, kwset_t kws, pickaxe_fn fn) |
186 | 0 | { |
187 | 0 | int i; |
188 | 0 | struct diff_queue_struct outq = DIFF_QUEUE_INIT; |
189 | |
|
190 | 0 | if (o->pickaxe_opts & DIFF_PICKAXE_ALL) { |
191 | | /* Showing the whole changeset if needle exists */ |
192 | 0 | for (i = 0; i < q->nr; i++) { |
193 | 0 | struct diff_filepair *p = q->queue[i]; |
194 | 0 | if (pickaxe_match(p, o, regexp, kws, fn)) |
195 | 0 | return; /* do not munge the queue */ |
196 | 0 | } |
197 | | |
198 | | /* |
199 | | * Otherwise we will clear the whole queue by copying |
200 | | * the empty outq at the end of this function, but |
201 | | * first clear the current entries in the queue. |
202 | | */ |
203 | 0 | for (i = 0; i < q->nr; i++) |
204 | 0 | diff_free_filepair(q->queue[i]); |
205 | 0 | } else { |
206 | | /* Showing only the filepairs that has the needle */ |
207 | 0 | for (i = 0; i < q->nr; i++) { |
208 | 0 | struct diff_filepair *p = q->queue[i]; |
209 | 0 | if (pickaxe_match(p, o, regexp, kws, fn)) |
210 | 0 | diff_q(&outq, p); |
211 | 0 | else |
212 | 0 | diff_free_filepair(p); |
213 | 0 | } |
214 | 0 | } |
215 | | |
216 | 0 | free(q->queue); |
217 | 0 | *q = outq; |
218 | 0 | } |
219 | | |
220 | | static void regcomp_or_die(regex_t *regex, const char *needle, int cflags) |
221 | 0 | { |
222 | 0 | int err = regcomp(regex, needle, cflags); |
223 | 0 | if (err) { |
224 | | /* The POSIX.2 people are surely sick */ |
225 | 0 | char errbuf[1024]; |
226 | 0 | regerror(err, regex, errbuf, 1024); |
227 | 0 | die("invalid regex: %s", errbuf); |
228 | 0 | } |
229 | 0 | } |
230 | | |
231 | | void diffcore_pickaxe(struct diff_options *o) |
232 | 0 | { |
233 | 0 | const char *needle = o->pickaxe; |
234 | 0 | int opts = o->pickaxe_opts; |
235 | 0 | regex_t regex, *regexp = NULL; |
236 | 0 | kwset_t kws = NULL; |
237 | 0 | pickaxe_fn fn; |
238 | |
|
239 | 0 | if (opts & ~DIFF_PICKAXE_KIND_OBJFIND && |
240 | 0 | (!needle || !*needle)) |
241 | 0 | BUG("should have needle under -G or -S"); |
242 | 0 | if (opts & (DIFF_PICKAXE_REGEX | DIFF_PICKAXE_KIND_G)) { |
243 | 0 | int cflags = REG_EXTENDED | REG_NEWLINE; |
244 | 0 | if (o->pickaxe_opts & DIFF_PICKAXE_IGNORE_CASE) |
245 | 0 | cflags |= REG_ICASE; |
246 | 0 | regcomp_or_die(®ex, needle, cflags); |
247 | 0 | regexp = ®ex; |
248 | |
|
249 | 0 | if (opts & DIFF_PICKAXE_KIND_G) |
250 | 0 | fn = diff_grep; |
251 | 0 | else if (opts & DIFF_PICKAXE_REGEX) |
252 | 0 | fn = has_changes; |
253 | 0 | else |
254 | | /* |
255 | | * We don't need to check the combination of |
256 | | * -G and --pickaxe-regex, by the time we get |
257 | | * here diff.c has already died if they're |
258 | | * combined. See the usage tests in |
259 | | * t4209-log-pickaxe.sh. |
260 | | */ |
261 | 0 | BUG("unreachable"); |
262 | 0 | } else if (opts & DIFF_PICKAXE_KIND_S) { |
263 | 0 | if (o->pickaxe_opts & DIFF_PICKAXE_IGNORE_CASE && |
264 | 0 | has_non_ascii(needle)) { |
265 | 0 | struct strbuf sb = STRBUF_INIT; |
266 | 0 | int cflags = REG_NEWLINE | REG_ICASE; |
267 | |
|
268 | 0 | basic_regex_quote_buf(&sb, needle); |
269 | 0 | regcomp_or_die(®ex, sb.buf, cflags); |
270 | 0 | strbuf_release(&sb); |
271 | 0 | regexp = ®ex; |
272 | 0 | } else { |
273 | 0 | kws = kwsalloc(o->pickaxe_opts & DIFF_PICKAXE_IGNORE_CASE |
274 | 0 | ? tolower_trans_tbl : NULL); |
275 | 0 | kwsincr(kws, needle, strlen(needle)); |
276 | 0 | kwsprep(kws); |
277 | 0 | } |
278 | 0 | fn = has_changes; |
279 | 0 | } else if (opts & DIFF_PICKAXE_KIND_OBJFIND) { |
280 | 0 | fn = NULL; |
281 | 0 | } else { |
282 | 0 | BUG("unknown pickaxe_opts flag"); |
283 | 0 | } |
284 | | |
285 | 0 | pickaxe(&diff_queued_diff, o, regexp, kws, fn); |
286 | |
|
287 | 0 | if (regexp) |
288 | 0 | regfree(regexp); |
289 | 0 | if (kws) |
290 | 0 | kwsfree(kws); |
291 | 0 | return; |
292 | 0 | } |