]>
Commit | Line | Data |
---|---|---|
427dcb4b | 1 | /* |
95acf11a | 2 | * |
427dcb4b JH |
3 | * Copyright (C) 2005 Junio C Hamano |
4 | */ | |
5 | #include "cache.h" | |
6 | #include "diff.h" | |
7 | #include "diffcore.h" | |
cbd53a21 | 8 | #include "object-store.h" |
f79d9c58 | 9 | #include "hashmap.h" |
3ac942d4 | 10 | #include "progress.h" |
95acf11a | 11 | #include "promisor-remote.h" |
9db2ac56 | 12 | #include "strmap.h" |
427dcb4b | 13 | |
25d5ea41 JH |
14 | /* Table of rename/copy destinations */ |
15 | ||
16 | static struct diff_rename_dst { | |
9db2ac56 EN |
17 | struct diff_filepair *p; |
18 | struct diff_filespec *filespec_to_free; | |
19 | int is_rename; /* false -> just a create; true -> rename or copy */ | |
25d5ea41 JH |
20 | } *rename_dst; |
21 | static int rename_dst_nr, rename_dst_alloc; | |
9db2ac56 EN |
22 | /* Mapping from break source pathname to break destination index */ |
23 | static struct strintmap *break_idx = NULL; | |
427dcb4b | 24 | |
9db2ac56 | 25 | static struct diff_rename_dst *locate_rename_dst(struct diff_filepair *p) |
f98c2f7e | 26 | { |
9db2ac56 EN |
27 | /* Lookup by p->ONE->path */ |
28 | int idx = break_idx ? strintmap_get(break_idx, p->one->path) : -1; | |
29 | return (idx == -1) ? NULL : &rename_dst[idx]; | |
f98c2f7e JK |
30 | } |
31 | ||
32 | /* | |
33 | * Returns 0 on success, -1 if we found a duplicate. | |
34 | */ | |
9db2ac56 | 35 | static int add_rename_dst(struct diff_filepair *p) |
f98c2f7e | 36 | { |
337ce247 | 37 | ALLOC_GROW(rename_dst, rename_dst_nr + 1, rename_dst_alloc); |
9db2ac56 EN |
38 | rename_dst[rename_dst_nr].p = p; |
39 | rename_dst[rename_dst_nr].filespec_to_free = NULL; | |
40 | rename_dst[rename_dst_nr].is_rename = 0; | |
25d5ea41 | 41 | rename_dst_nr++; |
f98c2f7e | 42 | return 0; |
427dcb4b JH |
43 | } |
44 | ||
15d061b4 | 45 | /* Table of rename/copy src files */ |
25d5ea41 | 46 | static struct diff_rename_src { |
e88d6bc6 | 47 | struct diff_filepair *p; |
fc580719 | 48 | unsigned short score; /* to remember the break score */ |
25d5ea41 JH |
49 | } *rename_src; |
50 | static int rename_src_nr, rename_src_alloc; | |
427dcb4b | 51 | |
b970b4ef | 52 | static void register_rename_src(struct diff_filepair *p) |
25d5ea41 | 53 | { |
9db2ac56 EN |
54 | if (p->broken_pair) { |
55 | if (!break_idx) { | |
56 | break_idx = xmalloc(sizeof(*break_idx)); | |
57 | strintmap_init(break_idx, -1); | |
58 | } | |
59 | strintmap_set(break_idx, p->one->path, rename_dst_nr); | |
60 | } | |
61 | ||
337ce247 | 62 | ALLOC_GROW(rename_src, rename_src_nr + 1, rename_src_alloc); |
b970b4ef EN |
63 | rename_src[rename_src_nr].p = p; |
64 | rename_src[rename_src_nr].score = p->score; | |
25d5ea41 | 65 | rename_src_nr++; |
427dcb4b JH |
66 | } |
67 | ||
0ce39643 JS |
68 | static int basename_same(struct diff_filespec *src, struct diff_filespec *dst) |
69 | { | |
70 | int src_len = strlen(src->path), dst_len = strlen(dst->path); | |
71 | while (src_len && dst_len) { | |
72 | char c1 = src->path[--src_len]; | |
73 | char c2 = dst->path[--dst_len]; | |
74 | if (c1 != c2) | |
75 | return 0; | |
76 | if (c1 == '/') | |
77 | return 1; | |
78 | } | |
79 | return (!src_len || src->path[src_len - 1] == '/') && | |
80 | (!dst_len || dst->path[dst_len - 1] == '/'); | |
81 | } | |
82 | ||
427dcb4b | 83 | struct diff_score { |
25d5ea41 JH |
84 | int src; /* index in rename_src */ |
85 | int dst; /* index in rename_dst */ | |
6d24ad97 JH |
86 | unsigned short score; |
87 | short name_score; | |
427dcb4b JH |
88 | }; |
89 | ||
95acf11a JT |
90 | struct prefetch_options { |
91 | struct repository *repo; | |
92 | int skip_unmodified; | |
93 | }; | |
94 | static void prefetch(void *prefetch_options) | |
95 | { | |
96 | struct prefetch_options *options = prefetch_options; | |
97 | int i; | |
98 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
99 | ||
100 | for (i = 0; i < rename_dst_nr; i++) { | |
9db2ac56 | 101 | if (rename_dst[i].p->renamed_pair) |
95acf11a JT |
102 | /* |
103 | * The loop in diffcore_rename() will not need these | |
104 | * blobs, so skip prefetching. | |
105 | */ | |
106 | continue; /* already found exact match */ | |
107 | diff_add_if_missing(options->repo, &to_fetch, | |
9db2ac56 | 108 | rename_dst[i].p->two); |
95acf11a JT |
109 | } |
110 | for (i = 0; i < rename_src_nr; i++) { | |
111 | if (options->skip_unmodified && | |
112 | diff_unmodified_pair(rename_src[i].p)) | |
113 | /* | |
114 | * The loop in diffcore_rename() will not need these | |
115 | * blobs, so skip prefetching. | |
116 | */ | |
117 | continue; | |
118 | diff_add_if_missing(options->repo, &to_fetch, | |
119 | rename_src[i].p->one); | |
120 | } | |
121 | promisor_remote_get_direct(options->repo, to_fetch.oid, to_fetch.nr); | |
122 | oid_array_clear(&to_fetch); | |
123 | } | |
124 | ||
b78ea5fc NTND |
125 | static int estimate_similarity(struct repository *r, |
126 | struct diff_filespec *src, | |
427dcb4b | 127 | struct diff_filespec *dst, |
95acf11a JT |
128 | int minimum_score, |
129 | int skip_unmodified) | |
427dcb4b JH |
130 | { |
131 | /* src points at a file that existed in the original tree (or | |
132 | * optionally a file in the destination tree) and dst points | |
133 | * at a newly created file. They may be quite similar, in which | |
134 | * case we want to say src is renamed to dst or src is copied into | |
135 | * dst, and then some edit has been applied to dst. | |
136 | * | |
137 | * Compare them and return how similar they are, representing | |
f0c6b2a2 JH |
138 | * the score as an integer between 0 and MAX_SCORE. |
139 | * | |
140 | * When there is an exact match, it is considered a better | |
141 | * match than anything else; the destination does not even | |
142 | * call into this function in that case. | |
427dcb4b | 143 | */ |
90bd932c | 144 | unsigned long max_size, delta_size, base_size, src_copied, literal_added; |
427dcb4b | 145 | int score; |
1c37e86a JT |
146 | struct diff_populate_filespec_options dpf_options = { |
147 | .check_size_only = 1 | |
148 | }; | |
95acf11a JT |
149 | struct prefetch_options prefetch_options = {r, skip_unmodified}; |
150 | ||
151 | if (r == the_repository && has_promisor_remote()) { | |
152 | dpf_options.missing_object_cb = prefetch; | |
153 | dpf_options.missing_object_data = &prefetch_options; | |
154 | } | |
427dcb4b | 155 | |
60896c7b JH |
156 | /* We deal only with regular files. Symlink renames are handled |
157 | * only when they are exact matches --- in other words, no edits | |
158 | * after renaming. | |
159 | */ | |
160 | if (!S_ISREG(src->mode) || !S_ISREG(dst->mode)) | |
161 | return 0; | |
162 | ||
81ac051d LT |
163 | /* |
164 | * Need to check that source and destination sizes are | |
165 | * filled in before comparing them. | |
166 | * | |
167 | * If we already have "cnt_data" filled in, we know it's | |
168 | * all good (avoid checking the size for zero, as that | |
169 | * is a possible size - we really should have a flag to | |
170 | * say whether the size is valid or not!) | |
171 | */ | |
8e5dd3d6 | 172 | if (!src->cnt_data && |
1c37e86a | 173 | diff_populate_filespec(r, src, &dpf_options)) |
81ac051d | 174 | return 0; |
8e5dd3d6 | 175 | if (!dst->cnt_data && |
1c37e86a | 176 | diff_populate_filespec(r, dst, &dpf_options)) |
81ac051d LT |
177 | return 0; |
178 | ||
90bd932c | 179 | max_size = ((src->size > dst->size) ? src->size : dst->size); |
58b103f5 | 180 | base_size = ((src->size < dst->size) ? src->size : dst->size); |
90bd932c | 181 | delta_size = max_size - base_size; |
427dcb4b | 182 | |
58b103f5 JH |
183 | /* We would not consider edits that change the file size so |
184 | * drastically. delta_size must be smaller than | |
cd1870ed | 185 | * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size). |
f0c6b2a2 | 186 | * |
58b103f5 JH |
187 | * Note that base_size == 0 case is handled here already |
188 | * and the final score computation below would not have a | |
189 | * divide-by-zero issue. | |
427dcb4b | 190 | */ |
3a4d6769 | 191 | if (max_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE) |
427dcb4b JH |
192 | return 0; |
193 | ||
95acf11a JT |
194 | dpf_options.check_size_only = 0; |
195 | ||
196 | if (!src->cnt_data && diff_populate_filespec(r, src, &dpf_options)) | |
885c716f | 197 | return 0; |
95acf11a | 198 | if (!dst->cnt_data && diff_populate_filespec(r, dst, &dpf_options)) |
885c716f BS |
199 | return 0; |
200 | ||
b78ea5fc | 201 | if (diffcore_count_changes(r, src, dst, |
c06c7966 | 202 | &src->cnt_data, &dst->cnt_data, |
65416758 | 203 | &src_copied, &literal_added)) |
85976974 | 204 | return 0; |
355e76a4 | 205 | |
1706306a JH |
206 | /* How similar are they? |
207 | * what percentage of material in dst are from source? | |
427dcb4b | 208 | */ |
90bd932c | 209 | if (!dst->size) |
1706306a | 210 | score = 0; /* should not happen */ |
cfc0aef1 | 211 | else |
dc49cd76 | 212 | score = (int)(src_copied * MAX_SCORE / max_size); |
427dcb4b JH |
213 | return score; |
214 | } | |
215 | ||
5098bafb | 216 | static void record_rename_pair(int dst_index, int src_index, int score) |
427dcb4b | 217 | { |
9db2ac56 EN |
218 | struct diff_filepair *src = rename_src[src_index].p; |
219 | struct diff_filepair *dst = rename_dst[dst_index].p; | |
f7c1512a | 220 | |
9db2ac56 | 221 | if (dst->renamed_pair) |
25d5ea41 | 222 | die("internal error: dst already matched."); |
427dcb4b | 223 | |
9db2ac56 EN |
224 | src->one->rename_used++; |
225 | src->one->count++; | |
427dcb4b | 226 | |
9db2ac56 EN |
227 | rename_dst[dst_index].filespec_to_free = dst->one; |
228 | rename_dst[dst_index].is_rename = 1; | |
427dcb4b | 229 | |
9db2ac56 EN |
230 | dst->one = src->one; |
231 | dst->renamed_pair = 1; | |
232 | if (!strcmp(dst->one->path, dst->two->path)) | |
233 | dst->score = rename_src[src_index].score; | |
fc580719 | 234 | else |
9db2ac56 | 235 | dst->score = score; |
427dcb4b JH |
236 | } |
237 | ||
238 | /* | |
239 | * We sort the rename similarity matrix with the score, in descending | |
15d061b4 | 240 | * order (the most similar first). |
427dcb4b JH |
241 | */ |
242 | static int score_compare(const void *a_, const void *b_) | |
243 | { | |
244 | const struct diff_score *a = a_, *b = b_; | |
cfc0aef1 | 245 | |
6d24ad97 JH |
246 | /* sink the unused ones to the bottom */ |
247 | if (a->dst < 0) | |
248 | return (0 <= b->dst); | |
249 | else if (b->dst < 0) | |
250 | return -1; | |
251 | ||
cfc0aef1 RS |
252 | if (a->score == b->score) |
253 | return b->name_score - a->name_score; | |
254 | ||
427dcb4b JH |
255 | return b->score - a->score; |
256 | } | |
257 | ||
9027f53c | 258 | struct file_similarity { |
f79d9c58 | 259 | struct hashmap_entry entry; |
7c85f8ac | 260 | int index; |
9027f53c | 261 | struct diff_filespec *filespec; |
9027f53c LT |
262 | }; |
263 | ||
b78ea5fc NTND |
264 | static unsigned int hash_filespec(struct repository *r, |
265 | struct diff_filespec *filespec) | |
48f6407f | 266 | { |
41c9560e | 267 | if (!filespec->oid_valid) { |
1c37e86a | 268 | if (diff_populate_filespec(r, filespec, NULL)) |
48f6407f | 269 | return 0; |
2dcde20e MT |
270 | hash_object_file(r->hash_algo, filespec->data, filespec->size, |
271 | "blob", &filespec->oid); | |
48f6407f | 272 | } |
d40abc8e | 273 | return oidhash(&filespec->oid); |
48f6407f KB |
274 | } |
275 | ||
f79d9c58 | 276 | static int find_identical_files(struct hashmap *srcs, |
7c85f8ac | 277 | int dst_index, |
11f944dd | 278 | struct diff_options *options) |
9027f53c LT |
279 | { |
280 | int renames = 0; | |
9db2ac56 | 281 | struct diff_filespec *target = rename_dst[dst_index].p->two; |
ab73a9d1 | 282 | struct file_similarity *p, *best = NULL; |
48f6407f | 283 | int i = 100, best_score = -1; |
f0e63c41 | 284 | unsigned int hash = hash_filespec(options->repo, target); |
48f6407f KB |
285 | |
286 | /* | |
7c85f8ac | 287 | * Find the best source match for specified destination. |
48f6407f | 288 | */ |
f0e63c41 EW |
289 | p = hashmap_get_entry_from_hash(srcs, hash, NULL, |
290 | struct file_similarity, entry); | |
23dee69f | 291 | hashmap_for_each_entry_from(srcs, p, entry) { |
48f6407f KB |
292 | int score; |
293 | struct diff_filespec *source = p->filespec; | |
294 | ||
295 | /* False hash collision? */ | |
9001dc2a | 296 | if (!oideq(&source->oid, &target->oid)) |
48f6407f KB |
297 | continue; |
298 | /* Non-regular files? If so, the modes must match! */ | |
299 | if (!S_ISREG(source->mode) || !S_ISREG(target->mode)) { | |
300 | if (source->mode != target->mode) | |
0940e5f2 | 301 | continue; |
9027f53c | 302 | } |
48f6407f KB |
303 | /* Give higher scores to sources that haven't been used already */ |
304 | score = !source->rename_used; | |
305 | if (source->rename_used && options->detect_rename != DIFF_DETECT_COPY) | |
306 | continue; | |
307 | score += basename_same(source, target); | |
308 | if (score > best_score) { | |
309 | best = p; | |
310 | best_score = score; | |
311 | if (score == 2) | |
312 | break; | |
9027f53c | 313 | } |
48f6407f KB |
314 | |
315 | /* Too many identical alternatives? Pick one */ | |
316 | if (!--i) | |
317 | break; | |
318 | } | |
319 | if (best) { | |
7c85f8ac | 320 | record_rename_pair(dst_index, best->index, MAX_SCORE); |
48f6407f KB |
321 | renames++; |
322 | } | |
9027f53c LT |
323 | return renames; |
324 | } | |
325 | ||
b78ea5fc NTND |
326 | static void insert_file_table(struct repository *r, |
327 | struct hashmap *table, int index, | |
328 | struct diff_filespec *filespec) | |
9027f53c | 329 | { |
9027f53c LT |
330 | struct file_similarity *entry = xmalloc(sizeof(*entry)); |
331 | ||
9027f53c LT |
332 | entry->index = index; |
333 | entry->filespec = filespec; | |
9027f53c | 334 | |
d22245a2 | 335 | hashmap_entry_init(&entry->entry, hash_filespec(r, filespec)); |
b94e5c1d | 336 | hashmap_add(table, &entry->entry); |
9027f53c LT |
337 | } |
338 | ||
cb1491b6 LT |
339 | /* |
340 | * Find exact renames first. | |
341 | * | |
342 | * The first round matches up the up-to-date entries, | |
343 | * and then during the second round we try to match | |
344 | * cache-dirty entries as well. | |
cb1491b6 | 345 | */ |
11f944dd | 346 | static int find_exact_renames(struct diff_options *options) |
cb1491b6 | 347 | { |
7c85f8ac | 348 | int i, renames = 0; |
f79d9c58 | 349 | struct hashmap file_table; |
cb1491b6 | 350 | |
ca4e3ca0 SG |
351 | /* Add all sources to the hash table in reverse order, because |
352 | * later on they will be retrieved in LIFO order. | |
353 | */ | |
7663cdc8 | 354 | hashmap_init(&file_table, NULL, NULL, rename_src_nr); |
ca4e3ca0 | 355 | for (i = rename_src_nr-1; i >= 0; i--) |
b78ea5fc NTND |
356 | insert_file_table(options->repo, |
357 | &file_table, i, | |
358 | rename_src[i].p->one); | |
9027f53c | 359 | |
7c85f8ac | 360 | /* Walk the destinations and find best source match */ |
9027f53c | 361 | for (i = 0; i < rename_dst_nr; i++) |
7c85f8ac | 362 | renames += find_identical_files(&file_table, i, options); |
9027f53c | 363 | |
f79d9c58 | 364 | /* Free the hash data structure and entries */ |
6da1a258 | 365 | hashmap_clear_and_free(&file_table, struct file_similarity, entry); |
9027f53c | 366 | |
7c85f8ac | 367 | return renames; |
cb1491b6 LT |
368 | } |
369 | ||
bde8b9f3 EN |
370 | struct dir_rename_info { |
371 | struct strintmap idx_map; | |
372 | struct strmap dir_rename_guess; | |
373 | struct strmap *dir_rename_count; | |
374 | unsigned setup; | |
375 | }; | |
376 | ||
377 | static char *get_dirname(const char *filename) | |
378 | { | |
379 | char *slash = strrchr(filename, '/'); | |
380 | return slash ? xstrndup(filename, slash - filename) : xstrdup(""); | |
381 | } | |
382 | ||
0c4fd732 EN |
383 | static void dirname_munge(char *filename) |
384 | { | |
385 | char *slash = strrchr(filename, '/'); | |
386 | if (!slash) | |
387 | slash = filename; | |
388 | *slash = '\0'; | |
389 | } | |
390 | ||
391 | static void increment_count(struct strmap *dir_rename_count, | |
392 | char *old_dir, | |
393 | char *new_dir) | |
394 | { | |
395 | struct strintmap *counts; | |
396 | struct strmap_entry *e; | |
397 | ||
398 | /* Get the {new_dirs -> counts} mapping using old_dir */ | |
399 | e = strmap_get_entry(dir_rename_count, old_dir); | |
400 | if (e) { | |
401 | counts = e->value; | |
402 | } else { | |
403 | counts = xmalloc(sizeof(*counts)); | |
404 | strintmap_init_with_options(counts, 0, NULL, 1); | |
405 | strmap_put(dir_rename_count, old_dir, counts); | |
406 | } | |
407 | ||
408 | /* Increment the count for new_dir */ | |
409 | strintmap_incr(counts, new_dir, 1); | |
410 | } | |
411 | ||
412 | static void update_dir_rename_counts(struct strmap *dir_rename_count, | |
413 | struct strset *dirs_removed, | |
414 | const char *oldname, | |
415 | const char *newname) | |
416 | { | |
417 | char *old_dir = xstrdup(oldname); | |
418 | char *new_dir = xstrdup(newname); | |
419 | char new_dir_first_char = new_dir[0]; | |
420 | int first_time_in_loop = 1; | |
421 | ||
422 | while (1) { | |
423 | dirname_munge(old_dir); | |
424 | dirname_munge(new_dir); | |
425 | ||
426 | /* | |
427 | * When renaming | |
428 | * "a/b/c/d/e/foo.c" -> "a/b/some/thing/else/e/foo.c" | |
429 | * then this suggests that both | |
430 | * a/b/c/d/e/ => a/b/some/thing/else/e/ | |
431 | * a/b/c/d/ => a/b/some/thing/else/ | |
432 | * so we want to increment counters for both. We do NOT, | |
433 | * however, also want to suggest that there was the following | |
434 | * rename: | |
435 | * a/b/c/ => a/b/some/thing/ | |
436 | * so we need to quit at that point. | |
437 | * | |
438 | * Note the when first_time_in_loop, we only strip off the | |
439 | * basename, and we don't care if that's different. | |
440 | */ | |
441 | if (!first_time_in_loop) { | |
442 | char *old_sub_dir = strchr(old_dir, '\0')+1; | |
443 | char *new_sub_dir = strchr(new_dir, '\0')+1; | |
444 | if (!*new_dir) { | |
445 | /* | |
446 | * Special case when renaming to root directory, | |
447 | * i.e. when new_dir == "". In this case, we had | |
448 | * something like | |
449 | * a/b/subdir => subdir | |
450 | * and so dirname_munge() sets things up so that | |
451 | * old_dir = "a/b\0subdir\0" | |
452 | * new_dir = "\0ubdir\0" | |
453 | * We didn't have a '/' to overwrite a '\0' onto | |
454 | * in new_dir, so we have to compare differently. | |
455 | */ | |
456 | if (new_dir_first_char != old_sub_dir[0] || | |
457 | strcmp(old_sub_dir+1, new_sub_dir)) | |
458 | break; | |
459 | } else { | |
460 | if (strcmp(old_sub_dir, new_sub_dir)) | |
461 | break; | |
462 | } | |
463 | } | |
464 | ||
465 | if (strset_contains(dirs_removed, old_dir)) | |
466 | increment_count(dir_rename_count, old_dir, new_dir); | |
467 | else | |
468 | break; | |
469 | ||
470 | /* If we hit toplevel directory ("") for old or new dir, quit */ | |
471 | if (!*old_dir || !*new_dir) | |
472 | break; | |
473 | ||
474 | first_time_in_loop = 0; | |
475 | } | |
476 | ||
477 | /* Free resources we don't need anymore */ | |
478 | free(old_dir); | |
479 | free(new_dir); | |
480 | } | |
481 | ||
482 | static void compute_dir_rename_counts(struct strmap *dir_rename_count, | |
483 | struct strset *dirs_removed) | |
484 | { | |
485 | int i; | |
486 | ||
487 | /* Set up dir_rename_count */ | |
488 | for (i = 0; i < rename_dst_nr; ++i) { | |
489 | /* File not part of directory rename counts if not a rename */ | |
490 | if (!rename_dst[i].is_rename) | |
491 | continue; | |
492 | ||
493 | /* | |
494 | * Make dir_rename_count contain a map of a map: | |
495 | * old_directory -> {new_directory -> count} | |
496 | * In other words, for every pair look at the directories for | |
497 | * the old filename and the new filename and count how many | |
498 | * times that pairing occurs. | |
499 | */ | |
500 | update_dir_rename_counts(dir_rename_count, dirs_removed, | |
501 | rename_dst[i].p->one->path, | |
502 | rename_dst[i].p->two->path); | |
503 | } | |
504 | } | |
505 | ||
ae8cf74d EN |
506 | static void initialize_dir_rename_info(struct dir_rename_info *info) |
507 | { | |
508 | int i; | |
509 | ||
510 | info->setup = 1; | |
511 | ||
512 | strintmap_init_with_options(&info->idx_map, -1, NULL, 0); | |
513 | strmap_init_with_options(&info->dir_rename_guess, NULL, 0); | |
514 | info->dir_rename_count = NULL; | |
515 | ||
516 | /* | |
517 | * Loop setting up both info->idx_map. | |
518 | */ | |
519 | for (i = 0; i < rename_dst_nr; ++i) { | |
520 | /* | |
521 | * For non-renamed files, make idx_map contain mapping of | |
522 | * filename -> index (index within rename_dst, that is) | |
523 | */ | |
524 | if (!rename_dst[i].is_rename) { | |
525 | char *filename = rename_dst[i].p->two->path; | |
526 | strintmap_set(&info->idx_map, filename, i); | |
527 | } | |
528 | } | |
529 | } | |
530 | ||
cd52e005 EN |
531 | void partial_clear_dir_rename_count(struct strmap *dir_rename_count) |
532 | { | |
533 | struct hashmap_iter iter; | |
534 | struct strmap_entry *entry; | |
535 | ||
536 | strmap_for_each_entry(dir_rename_count, &iter, entry) { | |
537 | struct strintmap *counts = entry->value; | |
538 | strintmap_clear(counts); | |
539 | } | |
540 | strmap_partial_clear(dir_rename_count, 1); | |
541 | } | |
542 | ||
ae8cf74d EN |
543 | static void cleanup_dir_rename_info(struct dir_rename_info *info) |
544 | { | |
545 | if (!info->setup) | |
546 | return; | |
547 | ||
548 | /* idx_map */ | |
549 | strintmap_clear(&info->idx_map); | |
550 | ||
551 | /* dir_rename_guess */ | |
552 | strmap_clear(&info->dir_rename_guess, 1); | |
553 | ||
554 | /* Nothing to do for dir_rename_count, yet */ | |
555 | } | |
556 | ||
a35df337 EN |
557 | static const char *get_basename(const char *filename) |
558 | { | |
559 | /* | |
560 | * gitbasename() has to worry about special drives, multiple | |
561 | * directory separator characters, trailing slashes, NULL or | |
562 | * empty strings, etc. We only work on filenames as stored in | |
563 | * git, and thus get to ignore all those complications. | |
564 | */ | |
565 | const char *base = strrchr(filename, '/'); | |
566 | return base ? base + 1 : filename; | |
567 | } | |
568 | ||
bde8b9f3 | 569 | static int idx_possible_rename(char *filename, struct dir_rename_info *info) |
37a25143 | 570 | { |
bde8b9f3 EN |
571 | /* |
572 | * Our comparison of files with the same basename (see | |
573 | * find_basename_matches() below), is only helpful when after exact | |
574 | * rename detection we have exactly one file with a given basename | |
575 | * among the rename sources and also only exactly one file with | |
576 | * that basename among the rename destinations. When we have | |
577 | * multiple files with the same basename in either set, we do not | |
578 | * know which to compare against. However, there are some | |
579 | * filenames that occur in large numbers (particularly | |
580 | * build-related filenames such as 'Makefile', '.gitignore', or | |
581 | * 'build.gradle' that potentially exist within every single | |
582 | * subdirectory), and for performance we want to be able to quickly | |
583 | * find renames for these files too. | |
584 | * | |
585 | * The reason basename comparisons are a useful heuristic was that it | |
586 | * is common for people to move files across directories while keeping | |
587 | * their filename the same. If we had a way of determining or even | |
588 | * making a good educated guess about which directory these non-unique | |
589 | * basename files had moved the file to, we could check it. | |
590 | * Luckily... | |
591 | * | |
592 | * When an entire directory is in fact renamed, we have two factors | |
593 | * helping us out: | |
594 | * (a) the original directory disappeared giving us a hint | |
595 | * about when we can apply an extra heuristic. | |
596 | * (a) we often have several files within that directory and | |
597 | * subdirectories that are renamed without changes | |
598 | * So, rules for a heuristic: | |
599 | * (0) If there basename matches are non-unique (the condition under | |
600 | * which this function is called) AND | |
601 | * (1) the directory in which the file was found has disappeared | |
602 | * (i.e. dirs_removed is non-NULL and has a relevant entry) THEN | |
603 | * (2) use exact renames of files within the directory to determine | |
604 | * where the directory is likely to have been renamed to. IF | |
605 | * there is at least one exact rename from within that | |
606 | * directory, we can proceed. | |
607 | * (3) If there are multiple places the directory could have been | |
608 | * renamed to based on exact renames, ignore all but one of them. | |
609 | * Just use the destination with the most renames going to it. | |
610 | * (4) Check if applying that directory rename to the original file | |
611 | * would result in a destination filename that is in the | |
612 | * potential rename set. If so, return the index of the | |
613 | * destination file (the index within rename_dst). | |
614 | * (5) Compare the original file and returned destination for | |
615 | * similarity, and if they are sufficiently similar, record the | |
616 | * rename. | |
617 | * | |
618 | * This function, idx_possible_rename(), is only responsible for (4). | |
619 | * The conditions/steps in (1)-(3) will be handled via setting up | |
620 | * dir_rename_count and dir_rename_guess in a future | |
621 | * initialize_dir_rename_info() function. Steps (0) and (5) are | |
622 | * handled by the caller of this function. | |
623 | */ | |
624 | char *old_dir, *new_dir; | |
625 | struct strbuf new_path = STRBUF_INIT; | |
626 | int idx; | |
627 | ||
628 | if (!info->setup) | |
629 | return -1; | |
630 | ||
631 | old_dir = get_dirname(filename); | |
632 | new_dir = strmap_get(&info->dir_rename_guess, old_dir); | |
633 | free(old_dir); | |
634 | if (!new_dir) | |
635 | return -1; | |
636 | ||
637 | strbuf_addstr(&new_path, new_dir); | |
638 | strbuf_addch(&new_path, '/'); | |
639 | strbuf_addstr(&new_path, get_basename(filename)); | |
640 | ||
641 | idx = strintmap_get(&info->idx_map, new_path.buf); | |
642 | strbuf_release(&new_path); | |
643 | return idx; | |
37a25143 EN |
644 | } |
645 | ||
a35df337 | 646 | static int find_basename_matches(struct diff_options *options, |
bde8b9f3 EN |
647 | int minimum_score, |
648 | struct dir_rename_info *info) | |
a35df337 | 649 | { |
da09f651 EN |
650 | /* |
651 | * When I checked in early 2020, over 76% of file renames in linux | |
652 | * just moved files to a different directory but kept the same | |
653 | * basename. gcc did that with over 64% of renames, gecko did it | |
654 | * with over 79%, and WebKit did it with over 89%. | |
655 | * | |
656 | * Therefore we can bypass the normal exhaustive NxM matrix | |
657 | * comparison of similarities between all potential rename sources | |
658 | * and destinations by instead using file basename as a hint (i.e. | |
659 | * the portion of the filename after the last '/'), checking for | |
660 | * similarity between files with the same basename, and if we find | |
661 | * a pair that are sufficiently similar, record the rename pair and | |
662 | * exclude those two from the NxM matrix. | |
663 | * | |
664 | * This *might* cause us to find a less than optimal pairing (if | |
665 | * there is another file that we are even more similar to but has a | |
666 | * different basename). Given the huge performance advantage | |
667 | * basename matching provides, and given the frequency with which | |
668 | * people use the same basename in real world projects, that's a | |
669 | * trade-off we are willing to accept when doing just rename | |
670 | * detection. | |
671 | * | |
672 | * If someone wants copy detection that implies they are willing to | |
673 | * spend more cycles to find similarities between files, so it may | |
674 | * be less likely that this heuristic is wanted. If someone is | |
675 | * doing break detection, that means they do not want filename | |
676 | * similarity to imply any form of content similiarity, and thus | |
677 | * this heuristic would definitely be incompatible. | |
678 | */ | |
679 | ||
680 | int i, renames = 0; | |
a35df337 EN |
681 | struct strintmap sources; |
682 | struct strintmap dests; | |
da09f651 EN |
683 | |
684 | /* | |
685 | * The prefeteching stuff wants to know if it can skip prefetching | |
686 | * blobs that are unmodified...and will then do a little extra work | |
687 | * to verify that the oids are indeed different before prefetching. | |
688 | * Unmodified blobs are only relevant when doing copy detection; | |
689 | * when limiting to rename detection, diffcore_rename[_extended]() | |
690 | * will never be called with unmodified source paths fed to us, so | |
691 | * the extra work necessary to check if rename_src entries are | |
692 | * unmodified would be a small waste. | |
693 | */ | |
694 | int skip_unmodified = 0; | |
a35df337 EN |
695 | |
696 | /* | |
697 | * Create maps of basename -> fullname(s) for remaining sources and | |
698 | * dests. | |
699 | */ | |
700 | strintmap_init_with_options(&sources, -1, NULL, 0); | |
701 | strintmap_init_with_options(&dests, -1, NULL, 0); | |
702 | for (i = 0; i < rename_src_nr; ++i) { | |
703 | char *filename = rename_src[i].p->one->path; | |
704 | const char *base; | |
705 | ||
706 | /* exact renames removed in remove_unneeded_paths_from_src() */ | |
707 | assert(!rename_src[i].p->one->rename_used); | |
708 | ||
709 | /* Record index within rename_src (i) if basename is unique */ | |
710 | base = get_basename(filename); | |
711 | if (strintmap_contains(&sources, base)) | |
712 | strintmap_set(&sources, base, -1); | |
713 | else | |
714 | strintmap_set(&sources, base, i); | |
715 | } | |
716 | for (i = 0; i < rename_dst_nr; ++i) { | |
717 | char *filename = rename_dst[i].p->two->path; | |
718 | const char *base; | |
719 | ||
720 | if (rename_dst[i].is_rename) | |
721 | continue; /* involved in exact match already. */ | |
722 | ||
723 | /* Record index within rename_dst (i) if basename is unique */ | |
724 | base = get_basename(filename); | |
725 | if (strintmap_contains(&dests, base)) | |
726 | strintmap_set(&dests, base, -1); | |
727 | else | |
728 | strintmap_set(&dests, base, i); | |
729 | } | |
730 | ||
da09f651 | 731 | /* Now look for basename matchups and do similarity estimation */ |
37a25143 EN |
732 | for (i = 0; i < rename_src_nr; ++i) { |
733 | char *filename = rename_src[i].p->one->path; | |
734 | const char *base = NULL; | |
735 | intptr_t src_index; | |
da09f651 | 736 | intptr_t dst_index; |
da09f651 | 737 | |
37a25143 EN |
738 | /* |
739 | * If the basename is unique among remaining sources, then | |
740 | * src_index will equal 'i' and we can attempt to match it | |
741 | * to a unique basename in the destinations. Otherwise, | |
742 | * use directory rename heuristics, if possible. | |
743 | */ | |
744 | base = get_basename(filename); | |
745 | src_index = strintmap_get(&sources, base); | |
746 | assert(src_index == -1 || src_index == i); | |
747 | ||
748 | if (strintmap_contains(&dests, base)) { | |
da09f651 EN |
749 | struct diff_filespec *one, *two; |
750 | int score; | |
751 | ||
37a25143 EN |
752 | /* Find a matching destination, if possible */ |
753 | dst_index = strintmap_get(&dests, base); | |
754 | if (src_index == -1 || dst_index == -1) { | |
755 | src_index = i; | |
bde8b9f3 | 756 | dst_index = idx_possible_rename(filename, info); |
37a25143 EN |
757 | } |
758 | if (dst_index == -1) | |
759 | continue; | |
760 | ||
761 | /* Ignore this dest if already used in a rename */ | |
762 | if (rename_dst[dst_index].is_rename) | |
763 | continue; /* already used previously */ | |
764 | ||
da09f651 EN |
765 | /* Estimate the similarity */ |
766 | one = rename_src[src_index].p->one; | |
767 | two = rename_dst[dst_index].p->two; | |
768 | score = estimate_similarity(options->repo, one, two, | |
769 | minimum_score, skip_unmodified); | |
770 | ||
771 | /* If sufficiently similar, record as rename pair */ | |
772 | if (score < minimum_score) | |
773 | continue; | |
774 | record_rename_pair(dst_index, src_index, score); | |
775 | renames++; | |
776 | ||
777 | /* | |
778 | * Found a rename so don't need text anymore; if we | |
779 | * didn't find a rename, the filespec_blob would get | |
780 | * re-used when doing the matrix of comparisons. | |
781 | */ | |
782 | diff_free_filespec_blob(one); | |
783 | diff_free_filespec_blob(two); | |
784 | } | |
785 | } | |
a35df337 EN |
786 | |
787 | strintmap_clear(&sources); | |
788 | strintmap_clear(&dests); | |
789 | ||
da09f651 | 790 | return renames; |
a35df337 EN |
791 | } |
792 | ||
6d24ad97 JH |
793 | #define NUM_CANDIDATE_PER_DST 4 |
794 | static void record_if_better(struct diff_score m[], struct diff_score *o) | |
795 | { | |
796 | int i, worst; | |
797 | ||
798 | /* find the worst one */ | |
799 | worst = 0; | |
800 | for (i = 1; i < NUM_CANDIDATE_PER_DST; i++) | |
801 | if (score_compare(&m[i], &m[worst]) > 0) | |
802 | worst = i; | |
803 | ||
804 | /* is it better than the worst one? */ | |
805 | if (score_compare(&m[worst], o) > 0) | |
806 | m[worst] = *o; | |
807 | } | |
808 | ||
f31027c9 JH |
809 | /* |
810 | * Returns: | |
811 | * 0 if we are under the limit; | |
812 | * 1 if we need to disable inexact rename detection; | |
813 | * 2 if we would be under the limit if we were given -C instead of -C -C. | |
814 | */ | |
00b8cccd | 815 | static int too_many_rename_candidates(int num_destinations, int num_sources, |
9d8a5a50 JH |
816 | struct diff_options *options) |
817 | { | |
818 | int rename_limit = options->rename_limit; | |
00b8cccd | 819 | int i, limited_sources; |
9d8a5a50 JH |
820 | |
821 | options->needed_rename_limit = 0; | |
822 | ||
823 | /* | |
824 | * This basically does a test for the rename matrix not | |
825 | * growing larger than a "rename_limit" square matrix, ie: | |
826 | * | |
00b8cccd | 827 | * num_destinations * num_sources > rename_limit * rename_limit |
ad8a1be5 EN |
828 | * |
829 | * We use st_mult() to check overflow conditions; in the | |
830 | * exceptional circumstance that size_t isn't large enough to hold | |
831 | * the multiplication, the system won't be able to allocate enough | |
832 | * memory for the matrix anyway. | |
9d8a5a50 | 833 | */ |
89973554 JT |
834 | if (rename_limit <= 0) |
835 | rename_limit = 32767; | |
ad8a1be5 EN |
836 | if (st_mult(num_destinations, num_sources) |
837 | <= st_mult(rename_limit, rename_limit)) | |
9d8a5a50 JH |
838 | return 0; |
839 | ||
840 | options->needed_rename_limit = | |
00b8cccd | 841 | num_sources > num_destinations ? num_sources : num_destinations; |
f31027c9 JH |
842 | |
843 | /* Are we running under -C -C? */ | |
0d1e0e78 | 844 | if (!options->flags.find_copies_harder) |
f31027c9 JH |
845 | return 1; |
846 | ||
847 | /* Would we bust the limit if we were running under -C? */ | |
00b8cccd | 848 | for (limited_sources = i = 0; i < num_sources; i++) { |
f31027c9 JH |
849 | if (diff_unmodified_pair(rename_src[i].p)) |
850 | continue; | |
00b8cccd | 851 | limited_sources++; |
f31027c9 | 852 | } |
ad8a1be5 EN |
853 | if (st_mult(num_destinations, limited_sources) |
854 | <= st_mult(rename_limit, rename_limit)) | |
f31027c9 | 855 | return 2; |
9d8a5a50 JH |
856 | return 1; |
857 | } | |
858 | ||
0940e5f2 LT |
859 | static int find_renames(struct diff_score *mx, int dst_cnt, int minimum_score, int copies) |
860 | { | |
861 | int count = 0, i; | |
862 | ||
863 | for (i = 0; i < dst_cnt * NUM_CANDIDATE_PER_DST; i++) { | |
864 | struct diff_rename_dst *dst; | |
865 | ||
866 | if ((mx[i].dst < 0) || | |
867 | (mx[i].score < minimum_score)) | |
868 | break; /* there is no more usable pair. */ | |
869 | dst = &rename_dst[mx[i].dst]; | |
9db2ac56 | 870 | if (dst->is_rename) |
0940e5f2 | 871 | continue; /* already done, either exact or fuzzy. */ |
e88d6bc6 | 872 | if (!copies && rename_src[mx[i].src].p->one->rename_used) |
0940e5f2 LT |
873 | continue; |
874 | record_rename_pair(mx[i].dst, mx[i].src, mx[i].score); | |
875 | count++; | |
876 | } | |
877 | return count; | |
878 | } | |
879 | ||
829514c5 EN |
880 | static void remove_unneeded_paths_from_src(int detecting_copies) |
881 | { | |
882 | int i, new_num_src; | |
883 | ||
884 | if (detecting_copies) | |
885 | return; /* nothing to remove */ | |
886 | if (break_idx) | |
887 | return; /* culling incompatible with break detection */ | |
888 | ||
889 | /* | |
890 | * Note on reasons why we cull unneeded sources but not destinations: | |
891 | * 1) Pairings are stored in rename_dst (not rename_src), which we | |
892 | * need to keep around. So, we just can't cull rename_dst even | |
893 | * if we wanted to. But doing so wouldn't help because... | |
894 | * | |
895 | * 2) There is a matrix pairwise comparison that follows the | |
896 | * "Performing inexact rename detection" progress message. | |
897 | * Iterating over the destinations is done in the outer loop, | |
898 | * hence we only iterate over each of those once and we can | |
899 | * easily skip the outer loop early if the destination isn't | |
900 | * relevant. That's only one check per destination path to | |
901 | * skip. | |
902 | * | |
903 | * By contrast, the sources are iterated in the inner loop; if | |
904 | * we check whether a source can be skipped, then we'll be | |
905 | * checking it N separate times, once for each destination. | |
906 | * We don't want to have to iterate over known-not-needed | |
907 | * sources N times each, so avoid that by removing the sources | |
908 | * from rename_src here. | |
909 | */ | |
910 | for (i = 0, new_num_src = 0; i < rename_src_nr; i++) { | |
911 | /* | |
912 | * renames are stored in rename_dst, so if a rename has | |
913 | * already been detected using this source, we can just | |
914 | * remove the source knowing rename_dst has its info. | |
915 | */ | |
916 | if (rename_src[i].p->one->rename_used) | |
917 | continue; | |
918 | ||
919 | if (new_num_src < i) | |
920 | memcpy(&rename_src[new_num_src], &rename_src[i], | |
921 | sizeof(struct diff_rename_src)); | |
922 | new_num_src++; | |
923 | } | |
924 | ||
925 | rename_src_nr = new_num_src; | |
926 | } | |
927 | ||
0c4fd732 EN |
928 | void diffcore_rename_extended(struct diff_options *options, |
929 | struct strset *dirs_removed, | |
930 | struct strmap *dir_rename_count) | |
427dcb4b | 931 | { |
8082d8d3 JH |
932 | int detect_rename = options->detect_rename; |
933 | int minimum_score = options->rename_score; | |
38c6f780 | 934 | struct diff_queue_struct *q = &diff_queued_diff; |
5098bafb | 935 | struct diff_queue_struct outq; |
427dcb4b | 936 | struct diff_score *mx; |
f31027c9 | 937 | int i, j, rename_count, skip_unmodified = 0; |
26a66a6b | 938 | int num_destinations, dst_cnt; |
f15eb7c1 | 939 | int num_sources, want_copies; |
3ac942d4 | 940 | struct progress *progress = NULL; |
bde8b9f3 | 941 | struct dir_rename_info info; |
427dcb4b | 942 | |
557ac035 | 943 | trace2_region_enter("diff", "setup", options->repo); |
bde8b9f3 | 944 | info.setup = 0; |
0c4fd732 | 945 | assert(!dir_rename_count || strmap_empty(dir_rename_count)); |
f15eb7c1 | 946 | want_copies = (detect_rename == DIFF_DETECT_COPY); |
26dee0ad | 947 | if (!minimum_score) |
f345b0a0 | 948 | minimum_score = DEFAULT_RENAME_SCORE; |
427dcb4b | 949 | |
427dcb4b | 950 | for (i = 0; i < q->nr; i++) { |
52e95789 | 951 | struct diff_filepair *p = q->queue[i]; |
2f3f8b21 | 952 | if (!DIFF_FILE_VALID(p->one)) { |
81e50eab | 953 | if (!DIFF_FILE_VALID(p->two)) |
60896c7b | 954 | continue; /* unmerged */ |
2f3f8b21 JH |
955 | else if (options->single_follow && |
956 | strcmp(options->single_follow, p->two->path)) | |
957 | continue; /* not interested */ | |
0d1e0e78 | 958 | else if (!options->flags.rename_empty && |
02491b67 | 959 | is_empty_blob_oid(&p->two->oid)) |
90d43b07 | 960 | continue; |
9db2ac56 | 961 | else if (add_rename_dst(p) < 0) { |
4d6be03b JK |
962 | warning("skipping rename detection, detected" |
963 | " duplicate destination '%s'", | |
964 | p->two->path); | |
965 | goto cleanup; | |
966 | } | |
2f3f8b21 | 967 | } |
0d1e0e78 | 968 | else if (!options->flags.rename_empty && |
02491b67 | 969 | is_empty_blob_oid(&p->one->oid)) |
90d43b07 | 970 | continue; |
d7c9bf22 | 971 | else if (!DIFF_PAIR_UNMERGED(p) && !DIFF_FILE_VALID(p->two)) { |
64479711 LT |
972 | /* |
973 | * If the source is a broken "delete", and | |
2210100a JH |
974 | * they did not really want to get broken, |
975 | * that means the source actually stays. | |
64479711 LT |
976 | * So we increment the "rename_used" score |
977 | * by one, to indicate ourselves as a user | |
978 | */ | |
979 | if (p->broken_pair && !p->score) | |
980 | p->one->rename_used++; | |
e88d6bc6 | 981 | register_rename_src(p); |
64479711 | 982 | } |
f15eb7c1 | 983 | else if (want_copies) { |
64479711 LT |
984 | /* |
985 | * Increment the "rename_used" score by | |
986 | * one, to indicate ourselves as a user. | |
2210100a | 987 | */ |
64479711 | 988 | p->one->rename_used++; |
e88d6bc6 | 989 | register_rename_src(p); |
2210100a | 990 | } |
427dcb4b | 991 | } |
557ac035 | 992 | trace2_region_leave("diff", "setup", options->repo); |
0024a549 | 993 | if (rename_dst_nr == 0 || rename_src_nr == 0) |
427dcb4b JH |
994 | goto cleanup; /* nothing to do */ |
995 | ||
557ac035 | 996 | trace2_region_enter("diff", "exact renames", options->repo); |
17559a64 LT |
997 | /* |
998 | * We really want to cull the candidates list early | |
999 | * with cheap tests in order to avoid doing deltas. | |
1000 | */ | |
11f944dd | 1001 | rename_count = find_exact_renames(options); |
557ac035 | 1002 | trace2_region_leave("diff", "exact renames", options->repo); |
17559a64 | 1003 | |
42899ac8 LT |
1004 | /* Did we only want exact renames? */ |
1005 | if (minimum_score == MAX_SCORE) | |
1006 | goto cleanup; | |
1007 | ||
f15eb7c1 | 1008 | num_sources = rename_src_nr; |
42899ac8 | 1009 | |
bd24aa2f EN |
1010 | if (want_copies || break_idx) { |
1011 | /* | |
1012 | * Cull sources: | |
1013 | * - remove ones corresponding to exact renames | |
1014 | */ | |
1015 | trace2_region_enter("diff", "cull after exact", options->repo); | |
1016 | remove_unneeded_paths_from_src(want_copies); | |
1017 | trace2_region_leave("diff", "cull after exact", options->repo); | |
1018 | } else { | |
1019 | /* Determine minimum score to match basenames */ | |
1020 | double factor = 0.5; | |
1021 | char *basename_factor = getenv("GIT_BASENAME_FACTOR"); | |
1022 | int min_basename_score; | |
1023 | ||
1024 | if (basename_factor) | |
1025 | factor = strtol(basename_factor, NULL, 10)/100.0; | |
1026 | assert(factor >= 0.0 && factor <= 1.0); | |
1027 | min_basename_score = minimum_score + | |
1028 | (int)(factor * (MAX_SCORE - minimum_score)); | |
1029 | ||
1030 | /* | |
1031 | * Cull sources: | |
1032 | * - remove ones involved in renames (found via exact match) | |
1033 | */ | |
1034 | trace2_region_enter("diff", "cull after exact", options->repo); | |
1035 | remove_unneeded_paths_from_src(want_copies); | |
1036 | trace2_region_leave("diff", "cull after exact", options->repo); | |
1037 | ||
ae8cf74d EN |
1038 | /* Preparation for basename-driven matching. */ |
1039 | trace2_region_enter("diff", "dir rename setup", options->repo); | |
1040 | initialize_dir_rename_info(&info); | |
1041 | trace2_region_leave("diff", "dir rename setup", options->repo); | |
1042 | ||
bd24aa2f EN |
1043 | /* Utilize file basenames to quickly find renames. */ |
1044 | trace2_region_enter("diff", "basename matches", options->repo); | |
1045 | rename_count += find_basename_matches(options, | |
bde8b9f3 EN |
1046 | min_basename_score, |
1047 | &info); | |
bd24aa2f EN |
1048 | trace2_region_leave("diff", "basename matches", options->repo); |
1049 | ||
1050 | /* | |
1051 | * Cull sources, again: | |
1052 | * - remove ones involved in renames (found via basenames) | |
1053 | */ | |
1054 | trace2_region_enter("diff", "cull basename", options->repo); | |
1055 | remove_unneeded_paths_from_src(want_copies); | |
1056 | trace2_region_leave("diff", "cull basename", options->repo); | |
1057 | } | |
1058 | ||
1059 | /* Calculate how many rename destinations are left */ | |
1060 | num_destinations = (rename_dst_nr - rename_count); | |
1061 | num_sources = rename_src_nr; /* rename_src_nr reflects lower number */ | |
1062 | ||
42899ac8 | 1063 | /* All done? */ |
f15eb7c1 | 1064 | if (!num_destinations || !num_sources) |
42899ac8 LT |
1065 | goto cleanup; |
1066 | ||
f15eb7c1 | 1067 | switch (too_many_rename_candidates(num_destinations, num_sources, |
00b8cccd | 1068 | options)) { |
f31027c9 | 1069 | case 1: |
0024a549 | 1070 | goto cleanup; |
f31027c9 JH |
1071 | case 2: |
1072 | options->degraded_cc_to_c = 1; | |
1073 | skip_unmodified = 1; | |
1074 | break; | |
1075 | default: | |
1076 | break; | |
1077 | } | |
0024a549 | 1078 | |
557ac035 | 1079 | trace2_region_enter("diff", "inexact renames", options->repo); |
3ac942d4 | 1080 | if (options->show_rename_progress) { |
8aade107 | 1081 | progress = start_delayed_progress( |
754dbc43 | 1082 | _("Performing inexact rename detection"), |
f15eb7c1 | 1083 | (uint64_t)num_destinations * (uint64_t)num_sources); |
3ac942d4 JK |
1084 | } |
1085 | ||
26a66a6b EN |
1086 | mx = xcalloc(st_mult(NUM_CANDIDATE_PER_DST, num_destinations), |
1087 | sizeof(*mx)); | |
25d5ea41 | 1088 | for (dst_cnt = i = 0; i < rename_dst_nr; i++) { |
9db2ac56 | 1089 | struct diff_filespec *two = rename_dst[i].p->two; |
6d24ad97 JH |
1090 | struct diff_score *m; |
1091 | ||
9db2ac56 | 1092 | if (rename_dst[i].is_rename) |
bd24aa2f | 1093 | continue; /* exact or basename match already handled */ |
6d24ad97 JH |
1094 | |
1095 | m = &mx[dst_cnt * NUM_CANDIDATE_PER_DST]; | |
1096 | for (j = 0; j < NUM_CANDIDATE_PER_DST; j++) | |
1097 | m[j].dst = -1; | |
1098 | ||
25d5ea41 | 1099 | for (j = 0; j < rename_src_nr; j++) { |
e88d6bc6 | 1100 | struct diff_filespec *one = rename_src[j].p->one; |
6d24ad97 | 1101 | struct diff_score this_src; |
f31027c9 | 1102 | |
829514c5 | 1103 | assert(!one->rename_used || want_copies || break_idx); |
f15eb7c1 | 1104 | |
f31027c9 JH |
1105 | if (skip_unmodified && |
1106 | diff_unmodified_pair(rename_src[j].p)) | |
1107 | continue; | |
1108 | ||
b78ea5fc NTND |
1109 | this_src.score = estimate_similarity(options->repo, |
1110 | one, two, | |
95acf11a JT |
1111 | minimum_score, |
1112 | skip_unmodified); | |
6d24ad97 JH |
1113 | this_src.name_score = basename_same(one, two); |
1114 | this_src.dst = i; | |
1115 | this_src.src = j; | |
1116 | record_if_better(m, &this_src); | |
809809bb JH |
1117 | /* |
1118 | * Once we run estimate_similarity, | |
1119 | * We do not need the text anymore. | |
1120 | */ | |
8ae92e63 | 1121 | diff_free_filespec_blob(one); |
809809bb | 1122 | diff_free_filespec_blob(two); |
427dcb4b JH |
1123 | } |
1124 | dst_cnt++; | |
81c4bf02 | 1125 | display_progress(progress, |
f15eb7c1 | 1126 | (uint64_t)dst_cnt * (uint64_t)num_sources); |
427dcb4b | 1127 | } |
3ac942d4 | 1128 | stop_progress(&progress); |
6d24ad97 | 1129 | |
427dcb4b | 1130 | /* cost matrix sorted by most to least similar pair */ |
2049b8dc | 1131 | STABLE_QSORT(mx, dst_cnt * NUM_CANDIDATE_PER_DST, score_compare); |
6d24ad97 | 1132 | |
0940e5f2 | 1133 | rename_count += find_renames(mx, dst_cnt, minimum_score, 0); |
f15eb7c1 | 1134 | if (want_copies) |
0940e5f2 | 1135 | rename_count += find_renames(mx, dst_cnt, minimum_score, 1); |
427dcb4b | 1136 | free(mx); |
557ac035 | 1137 | trace2_region_leave("diff", "inexact renames", options->repo); |
427dcb4b | 1138 | |
15d061b4 | 1139 | cleanup: |
0c4fd732 EN |
1140 | /* |
1141 | * Now that renames have been computed, compute dir_rename_count */ | |
1142 | if (dirs_removed && dir_rename_count) | |
1143 | compute_dir_rename_counts(dir_rename_count, dirs_removed); | |
1144 | ||
427dcb4b | 1145 | /* At this point, we have found some renames and copies and they |
5098bafb | 1146 | * are recorded in rename_dst. The original list is still in *q. |
427dcb4b | 1147 | */ |
557ac035 | 1148 | trace2_region_enter("diff", "write back to queue", options->repo); |
9ca5df90 | 1149 | DIFF_QUEUE_CLEAR(&outq); |
427dcb4b | 1150 | for (i = 0; i < q->nr; i++) { |
25d5ea41 | 1151 | struct diff_filepair *p = q->queue[i]; |
25d5ea41 JH |
1152 | struct diff_filepair *pair_to_free = NULL; |
1153 | ||
d7c9bf22 MZ |
1154 | if (DIFF_PAIR_UNMERGED(p)) { |
1155 | diff_q(&outq, p); | |
1156 | } | |
1157 | else if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) { | |
9db2ac56 EN |
1158 | /* Creation */ |
1159 | diff_q(&outq, p); | |
427dcb4b | 1160 | } |
2cd68882 JH |
1161 | else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) { |
1162 | /* | |
1163 | * Deletion | |
1164 | * | |
f345b0a0 JH |
1165 | * We would output this delete record if: |
1166 | * | |
1167 | * (1) this is a broken delete and the counterpart | |
1168 | * broken create remains in the output; or | |
5098bafb JH |
1169 | * (2) this is not a broken delete, and rename_dst |
1170 | * does not have a rename/copy to move p->one->path | |
1171 | * out of existence. | |
f345b0a0 JH |
1172 | * |
1173 | * Otherwise, the counterpart broken create | |
1174 | * has been turned into a rename-edit; or | |
1175 | * delete did not have a matching create to | |
1176 | * begin with. | |
2cd68882 | 1177 | */ |
f345b0a0 JH |
1178 | if (DIFF_PAIR_BROKEN(p)) { |
1179 | /* broken delete */ | |
9db2ac56 EN |
1180 | struct diff_rename_dst *dst = locate_rename_dst(p); |
1181 | if (!dst) | |
1182 | BUG("tracking failed somehow; failed to find associated dst for broken pair"); | |
1183 | if (dst->is_rename) | |
f345b0a0 JH |
1184 | /* counterpart is now rename/copy */ |
1185 | pair_to_free = p; | |
1186 | } | |
1187 | else { | |
64479711 | 1188 | if (p->one->rename_used) |
f345b0a0 JH |
1189 | /* this path remains */ |
1190 | pair_to_free = p; | |
1191 | } | |
2cd68882 | 1192 | |
9db2ac56 | 1193 | if (!pair_to_free) |
2cd68882 JH |
1194 | diff_q(&outq, p); |
1195 | } | |
25d5ea41 | 1196 | else if (!diff_unmodified_pair(p)) |
15d061b4 | 1197 | /* all the usual ones need to be kept */ |
25d5ea41 | 1198 | diff_q(&outq, p); |
15d061b4 | 1199 | else |
9db2ac56 | 1200 | /* no need to keep unmodified pairs; FIXME: remove earlier? */ |
15d061b4 JH |
1201 | pair_to_free = p; |
1202 | ||
226406f6 JH |
1203 | if (pair_to_free) |
1204 | diff_free_filepair(pair_to_free); | |
427dcb4b | 1205 | } |
25d5ea41 | 1206 | diff_debug_queue("done copying original", &outq); |
427dcb4b | 1207 | |
25d5ea41 JH |
1208 | free(q->queue); |
1209 | *q = outq; | |
1210 | diff_debug_queue("done collapsing", q); | |
427dcb4b | 1211 | |
9fb88419 | 1212 | for (i = 0; i < rename_dst_nr; i++) |
9db2ac56 EN |
1213 | if (rename_dst[i].filespec_to_free) |
1214 | free_filespec(rename_dst[i].filespec_to_free); | |
5098bafb | 1215 | |
ae8cf74d | 1216 | cleanup_dir_rename_info(&info); |
6a83d902 | 1217 | FREE_AND_NULL(rename_dst); |
25d5ea41 | 1218 | rename_dst_nr = rename_dst_alloc = 0; |
6a83d902 | 1219 | FREE_AND_NULL(rename_src); |
25d5ea41 | 1220 | rename_src_nr = rename_src_alloc = 0; |
9db2ac56 EN |
1221 | if (break_idx) { |
1222 | strintmap_clear(break_idx); | |
1223 | FREE_AND_NULL(break_idx); | |
1224 | } | |
557ac035 | 1225 | trace2_region_leave("diff", "write back to queue", options->repo); |
427dcb4b JH |
1226 | return; |
1227 | } | |
0c4fd732 EN |
1228 | |
1229 | void diffcore_rename(struct diff_options *options) | |
1230 | { | |
1231 | diffcore_rename_extended(options, NULL, NULL); | |
1232 | } |