]>
Commit | Line | Data |
---|---|---|
d9c66f0b JS |
1 | #include "cache.h" |
2 | #include "range-diff.h" | |
3 | #include "string-list.h" | |
4 | #include "run-command.h" | |
5 | #include "argv-array.h" | |
6 | #include "hashmap.h" | |
7 | #include "xdiff-interface.h" | |
8 | #include "linear-assignment.h" | |
c8c5e43a | 9 | #include "diffcore.h" |
eb0be38c JS |
10 | #include "commit.h" |
11 | #include "pretty.h" | |
d9c66f0b JS |
12 | |
13 | struct patch_util { | |
14 | /* For the search for an exact match */ | |
15 | struct hashmap_entry e; | |
16 | const char *diff, *patch; | |
17 | ||
9dc46e02 | 18 | int i, shown; |
d9c66f0b JS |
19 | int diffsize; |
20 | size_t diff_offset; | |
21 | /* the index of the matching item in the other branch, or -1 */ | |
22 | int matching; | |
23 | struct object_id oid; | |
24 | }; | |
25 | ||
26 | /* | |
27 | * Reads the patches into a string list, with the `util` field being populated | |
28 | * as struct object_id (will need to be free()d). | |
29 | */ | |
30 | static int read_patches(const char *range, struct string_list *list) | |
31 | { | |
32 | struct child_process cp = CHILD_PROCESS_INIT; | |
33 | FILE *in; | |
34 | struct strbuf buf = STRBUF_INIT, line = STRBUF_INIT; | |
35 | struct patch_util *util = NULL; | |
36 | int in_header = 1; | |
37 | ||
38 | argv_array_pushl(&cp.args, "log", "--no-color", "-p", "--no-merges", | |
39 | "--reverse", "--date-order", "--decorate=no", | |
40 | "--no-abbrev-commit", range, | |
41 | NULL); | |
42 | cp.out = -1; | |
43 | cp.no_stdin = 1; | |
44 | cp.git_cmd = 1; | |
45 | ||
46 | if (start_command(&cp)) | |
47 | return error_errno(_("could not start `log`")); | |
48 | in = fdopen(cp.out, "r"); | |
49 | if (!in) { | |
50 | error_errno(_("could not read `log` output")); | |
51 | finish_command(&cp); | |
52 | return -1; | |
53 | } | |
54 | ||
55 | while (strbuf_getline(&line, in) != EOF) { | |
56 | const char *p; | |
57 | ||
58 | if (skip_prefix(line.buf, "commit ", &p)) { | |
59 | if (util) { | |
60 | string_list_append(list, buf.buf)->util = util; | |
61 | strbuf_reset(&buf); | |
62 | } | |
63 | util = xcalloc(sizeof(*util), 1); | |
64 | if (get_oid(p, &util->oid)) { | |
65 | error(_("could not parse commit '%s'"), p); | |
66 | free(util); | |
67 | string_list_clear(list, 1); | |
68 | strbuf_release(&buf); | |
69 | strbuf_release(&line); | |
70 | fclose(in); | |
71 | finish_command(&cp); | |
72 | return -1; | |
73 | } | |
74 | util->matching = -1; | |
75 | in_header = 1; | |
76 | continue; | |
77 | } | |
78 | ||
79 | if (starts_with(line.buf, "diff --git")) { | |
80 | in_header = 0; | |
81 | strbuf_addch(&buf, '\n'); | |
82 | if (!util->diff_offset) | |
83 | util->diff_offset = buf.len; | |
84 | strbuf_addbuf(&buf, &line); | |
85 | } else if (in_header) { | |
86 | if (starts_with(line.buf, "Author: ")) { | |
87 | strbuf_addbuf(&buf, &line); | |
88 | strbuf_addstr(&buf, "\n\n"); | |
89 | } else if (starts_with(line.buf, " ")) { | |
a142f978 | 90 | strbuf_rtrim(&line); |
d9c66f0b JS |
91 | strbuf_addbuf(&buf, &line); |
92 | strbuf_addch(&buf, '\n'); | |
93 | } | |
94 | continue; | |
95 | } else if (starts_with(line.buf, "@@ ")) | |
96 | strbuf_addstr(&buf, "@@"); | |
97 | else if (!line.buf[0] || starts_with(line.buf, "index ")) | |
98 | /* | |
99 | * A completely blank (not ' \n', which is context) | |
100 | * line is not valid in a diff. We skip it | |
101 | * silently, because this neatly handles the blank | |
102 | * separator line between commits in git-log | |
103 | * output. | |
104 | * | |
105 | * We also want to ignore the diff's `index` lines | |
106 | * because they contain exact blob hashes in which | |
107 | * we are not interested. | |
108 | */ | |
109 | continue; | |
110 | else | |
111 | strbuf_addbuf(&buf, &line); | |
112 | ||
113 | strbuf_addch(&buf, '\n'); | |
114 | util->diffsize++; | |
115 | } | |
116 | fclose(in); | |
117 | strbuf_release(&line); | |
118 | ||
119 | if (util) | |
120 | string_list_append(list, buf.buf)->util = util; | |
121 | strbuf_release(&buf); | |
122 | ||
123 | if (finish_command(&cp)) | |
124 | return -1; | |
125 | ||
126 | return 0; | |
127 | } | |
128 | ||
129 | static int patch_util_cmp(const void *dummy, const struct patch_util *a, | |
130 | const struct patch_util *b, const char *keydata) | |
131 | { | |
132 | return strcmp(a->diff, keydata ? keydata : b->diff); | |
133 | } | |
134 | ||
135 | static void find_exact_matches(struct string_list *a, struct string_list *b) | |
136 | { | |
137 | struct hashmap map; | |
138 | int i; | |
139 | ||
140 | hashmap_init(&map, (hashmap_cmp_fn)patch_util_cmp, NULL, 0); | |
141 | ||
142 | /* First, add the patches of a to a hash map */ | |
143 | for (i = 0; i < a->nr; i++) { | |
144 | struct patch_util *util = a->items[i].util; | |
145 | ||
146 | util->i = i; | |
147 | util->patch = a->items[i].string; | |
148 | util->diff = util->patch + util->diff_offset; | |
149 | hashmap_entry_init(util, strhash(util->diff)); | |
150 | hashmap_add(&map, util); | |
151 | } | |
152 | ||
153 | /* Now try to find exact matches in b */ | |
154 | for (i = 0; i < b->nr; i++) { | |
155 | struct patch_util *util = b->items[i].util, *other; | |
156 | ||
157 | util->i = i; | |
158 | util->patch = b->items[i].string; | |
159 | util->diff = util->patch + util->diff_offset; | |
160 | hashmap_entry_init(util, strhash(util->diff)); | |
161 | other = hashmap_remove(&map, util, NULL); | |
162 | if (other) { | |
163 | if (other->matching >= 0) | |
164 | BUG("already assigned!"); | |
165 | ||
166 | other->matching = i; | |
167 | util->matching = other->i; | |
168 | } | |
169 | } | |
170 | ||
171 | hashmap_free(&map, 0); | |
172 | } | |
173 | ||
174 | static void diffsize_consume(void *data, char *line, unsigned long len) | |
175 | { | |
176 | (*(int *)data)++; | |
177 | } | |
178 | ||
179 | static int diffsize(const char *a, const char *b) | |
180 | { | |
181 | xpparam_t pp = { 0 }; | |
182 | xdemitconf_t cfg = { 0 }; | |
183 | mmfile_t mf1, mf2; | |
184 | int count = 0; | |
185 | ||
186 | mf1.ptr = (char *)a; | |
187 | mf1.size = strlen(a); | |
188 | mf2.ptr = (char *)b; | |
189 | mf2.size = strlen(b); | |
190 | ||
191 | cfg.ctxlen = 3; | |
192 | if (!xdi_diff_outf(&mf1, &mf2, diffsize_consume, &count, &pp, &cfg)) | |
193 | return count; | |
194 | ||
195 | error(_("failed to generate diff")); | |
196 | return COST_MAX; | |
197 | } | |
198 | ||
199 | static void get_correspondences(struct string_list *a, struct string_list *b, | |
200 | int creation_factor) | |
201 | { | |
202 | int n = a->nr + b->nr; | |
203 | int *cost, c, *a2b, *b2a; | |
204 | int i, j; | |
205 | ||
206 | ALLOC_ARRAY(cost, st_mult(n, n)); | |
207 | ALLOC_ARRAY(a2b, n); | |
208 | ALLOC_ARRAY(b2a, n); | |
209 | ||
210 | for (i = 0; i < a->nr; i++) { | |
211 | struct patch_util *a_util = a->items[i].util; | |
212 | ||
213 | for (j = 0; j < b->nr; j++) { | |
214 | struct patch_util *b_util = b->items[j].util; | |
215 | ||
216 | if (a_util->matching == j) | |
217 | c = 0; | |
218 | else if (a_util->matching < 0 && b_util->matching < 0) | |
219 | c = diffsize(a_util->diff, b_util->diff); | |
220 | else | |
221 | c = COST_MAX; | |
222 | cost[i + n * j] = c; | |
223 | } | |
224 | ||
225 | c = a_util->matching < 0 ? | |
226 | a_util->diffsize * creation_factor / 100 : COST_MAX; | |
227 | for (j = b->nr; j < n; j++) | |
228 | cost[i + n * j] = c; | |
229 | } | |
230 | ||
231 | for (j = 0; j < b->nr; j++) { | |
232 | struct patch_util *util = b->items[j].util; | |
233 | ||
234 | c = util->matching < 0 ? | |
235 | util->diffsize * creation_factor / 100 : COST_MAX; | |
236 | for (i = a->nr; i < n; i++) | |
237 | cost[i + n * j] = c; | |
238 | } | |
239 | ||
240 | for (i = a->nr; i < n; i++) | |
241 | for (j = b->nr; j < n; j++) | |
242 | cost[i + n * j] = 0; | |
243 | ||
244 | compute_assignment(n, n, cost, a2b, b2a); | |
245 | ||
246 | for (i = 0; i < a->nr; i++) | |
247 | if (a2b[i] >= 0 && a2b[i] < b->nr) { | |
248 | struct patch_util *a_util = a->items[i].util; | |
249 | struct patch_util *b_util = b->items[a2b[i]].util; | |
250 | ||
251 | a_util->matching = a2b[i]; | |
252 | b_util->matching = i; | |
253 | } | |
254 | ||
255 | free(cost); | |
256 | free(a2b); | |
257 | free(b2a); | |
258 | } | |
259 | ||
eb0be38c JS |
260 | static void output_pair_header(struct strbuf *buf, |
261 | struct strbuf *dashes, | |
262 | struct patch_util *a_util, | |
263 | struct patch_util *b_util) | |
d9c66f0b | 264 | { |
eb0be38c JS |
265 | struct object_id *oid = a_util ? &a_util->oid : &b_util->oid; |
266 | struct commit *commit; | |
267 | ||
268 | if (!dashes->len) | |
269 | strbuf_addchars(dashes, '-', | |
270 | strlen(find_unique_abbrev(oid, | |
271 | DEFAULT_ABBREV))); | |
272 | ||
273 | strbuf_reset(buf); | |
274 | if (!a_util) | |
275 | strbuf_addf(buf, "-: %s ", dashes->buf); | |
276 | else | |
277 | strbuf_addf(buf, "%d: %s ", a_util->i + 1, | |
278 | find_unique_abbrev(&a_util->oid, DEFAULT_ABBREV)); | |
279 | ||
280 | if (!a_util) | |
281 | strbuf_addch(buf, '>'); | |
282 | else if (!b_util) | |
283 | strbuf_addch(buf, '<'); | |
284 | else if (strcmp(a_util->patch, b_util->patch)) | |
285 | strbuf_addch(buf, '!'); | |
286 | else | |
287 | strbuf_addch(buf, '='); | |
288 | ||
289 | if (!b_util) | |
290 | strbuf_addf(buf, " -: %s", dashes->buf); | |
291 | else | |
292 | strbuf_addf(buf, " %d: %s", b_util->i + 1, | |
293 | find_unique_abbrev(&b_util->oid, DEFAULT_ABBREV)); | |
294 | ||
295 | commit = lookup_commit_reference(the_repository, oid); | |
296 | if (commit) { | |
297 | strbuf_addch(buf, ' '); | |
298 | pp_commit_easy(CMIT_FMT_ONELINE, commit, buf); | |
299 | } | |
300 | strbuf_addch(buf, '\n'); | |
301 | ||
302 | fwrite(buf->buf, buf->len, 1, stdout); | |
d9c66f0b JS |
303 | } |
304 | ||
c8c5e43a JS |
305 | static struct diff_filespec *get_filespec(const char *name, const char *p) |
306 | { | |
307 | struct diff_filespec *spec = alloc_filespec(name); | |
308 | ||
309 | fill_filespec(spec, &null_oid, 0, 0644); | |
310 | spec->data = (char *)p; | |
311 | spec->size = strlen(p); | |
312 | spec->should_munmap = 0; | |
313 | spec->is_stdin = 1; | |
314 | ||
315 | return spec; | |
316 | } | |
317 | ||
318 | static void patch_diff(const char *a, const char *b, | |
319 | struct diff_options *diffopt) | |
320 | { | |
321 | diff_queue(&diff_queued_diff, | |
322 | get_filespec("a", a), get_filespec("b", b)); | |
323 | ||
324 | diffcore_std(diffopt); | |
325 | diff_flush(diffopt); | |
326 | } | |
327 | ||
328 | static void output(struct string_list *a, struct string_list *b, | |
329 | struct diff_options *diffopt) | |
d9c66f0b | 330 | { |
eb0be38c | 331 | struct strbuf buf = STRBUF_INIT, dashes = STRBUF_INIT; |
9dc46e02 JS |
332 | int i = 0, j = 0; |
333 | ||
334 | /* | |
335 | * We assume the user is really more interested in the second argument | |
336 | * ("newer" version). To that end, we print the output in the order of | |
337 | * the RHS (the `b` parameter). To put the LHS (the `a` parameter) | |
338 | * commits that are no longer in the RHS into a good place, we place | |
339 | * them once we have shown all of their predecessors in the LHS. | |
340 | */ | |
341 | ||
342 | while (i < a->nr || j < b->nr) { | |
343 | struct patch_util *a_util, *b_util; | |
344 | a_util = i < a->nr ? a->items[i].util : NULL; | |
345 | b_util = j < b->nr ? b->items[j].util : NULL; | |
346 | ||
347 | /* Skip all the already-shown commits from the LHS. */ | |
348 | while (i < a->nr && a_util->shown) | |
349 | a_util = ++i < a->nr ? a->items[i].util : NULL; | |
350 | ||
351 | /* Show unmatched LHS commit whose predecessors were shown. */ | |
352 | if (i < a->nr && a_util->matching < 0) { | |
eb0be38c | 353 | output_pair_header(&buf, &dashes, a_util, NULL); |
9dc46e02 JS |
354 | i++; |
355 | continue; | |
356 | } | |
d9c66f0b | 357 | |
9dc46e02 JS |
358 | /* Show unmatched RHS commits. */ |
359 | while (j < b->nr && b_util->matching < 0) { | |
eb0be38c | 360 | output_pair_header(&buf, &dashes, NULL, b_util); |
9dc46e02 | 361 | b_util = ++j < b->nr ? b->items[j].util : NULL; |
d9c66f0b | 362 | } |
d9c66f0b | 363 | |
9dc46e02 JS |
364 | /* Show matching LHS/RHS pair. */ |
365 | if (j < b->nr) { | |
366 | a_util = a->items[b_util->matching].util; | |
eb0be38c | 367 | output_pair_header(&buf, &dashes, a_util, b_util); |
c8c5e43a JS |
368 | if (!(diffopt->output_format & DIFF_FORMAT_NO_OUTPUT)) |
369 | patch_diff(a->items[b_util->matching].string, | |
370 | b->items[j].string, diffopt); | |
9dc46e02 JS |
371 | a_util->shown = 1; |
372 | j++; | |
373 | } | |
d9c66f0b | 374 | } |
eb0be38c JS |
375 | strbuf_release(&buf); |
376 | strbuf_release(&dashes); | |
d9c66f0b JS |
377 | } |
378 | ||
379 | int show_range_diff(const char *range1, const char *range2, | |
c8c5e43a | 380 | int creation_factor, struct diff_options *diffopt) |
d9c66f0b JS |
381 | { |
382 | int res = 0; | |
383 | ||
384 | struct string_list branch1 = STRING_LIST_INIT_DUP; | |
385 | struct string_list branch2 = STRING_LIST_INIT_DUP; | |
386 | ||
387 | if (read_patches(range1, &branch1)) | |
388 | res = error(_("could not parse log for '%s'"), range1); | |
389 | if (!res && read_patches(range2, &branch2)) | |
390 | res = error(_("could not parse log for '%s'"), range2); | |
391 | ||
392 | if (!res) { | |
393 | find_exact_matches(&branch1, &branch2); | |
394 | get_correspondences(&branch1, &branch2, creation_factor); | |
c8c5e43a | 395 | output(&branch1, &branch2, diffopt); |
d9c66f0b JS |
396 | } |
397 | ||
398 | string_list_clear(&branch1, 1); | |
399 | string_list_clear(&branch2, 1); | |
400 | ||
401 | return res; | |
402 | } |