]>
Commit | Line | Data |
---|---|---|
427dcb4b | 1 | /* |
95acf11a | 2 | * |
427dcb4b JH |
3 | * Copyright (C) 2005 Junio C Hamano |
4 | */ | |
fc7bd51b | 5 | #include "git-compat-util.h" |
427dcb4b JH |
6 | #include "diff.h" |
7 | #include "diffcore.h" | |
a034e910 | 8 | #include "object-store-ll.h" |
f79d9c58 | 9 | #include "hashmap.h" |
fc7bd51b EN |
10 | #include "mem-pool.h" |
11 | #include "oid-array.h" | |
3ac942d4 | 12 | #include "progress.h" |
95acf11a | 13 | #include "promisor-remote.h" |
fc7bd51b | 14 | #include "string-list.h" |
9db2ac56 | 15 | #include "strmap.h" |
fc7bd51b | 16 | #include "trace2.h" |
427dcb4b | 17 | |
25d5ea41 JH |
18 | /* Table of rename/copy destinations */ |
19 | ||
20 | static struct diff_rename_dst { | |
9db2ac56 EN |
21 | struct diff_filepair *p; |
22 | struct diff_filespec *filespec_to_free; | |
23 | int is_rename; /* false -> just a create; true -> rename or copy */ | |
25d5ea41 JH |
24 | } *rename_dst; |
25 | static int rename_dst_nr, rename_dst_alloc; | |
9db2ac56 EN |
26 | /* Mapping from break source pathname to break destination index */ |
27 | static struct strintmap *break_idx = NULL; | |
427dcb4b | 28 | |
9db2ac56 | 29 | static struct diff_rename_dst *locate_rename_dst(struct diff_filepair *p) |
f98c2f7e | 30 | { |
9db2ac56 EN |
31 | /* Lookup by p->ONE->path */ |
32 | int idx = break_idx ? strintmap_get(break_idx, p->one->path) : -1; | |
33 | return (idx == -1) ? NULL : &rename_dst[idx]; | |
f98c2f7e JK |
34 | } |
35 | ||
36 | /* | |
37 | * Returns 0 on success, -1 if we found a duplicate. | |
38 | */ | |
9db2ac56 | 39 | static int add_rename_dst(struct diff_filepair *p) |
f98c2f7e | 40 | { |
337ce247 | 41 | ALLOC_GROW(rename_dst, rename_dst_nr + 1, rename_dst_alloc); |
9db2ac56 EN |
42 | rename_dst[rename_dst_nr].p = p; |
43 | rename_dst[rename_dst_nr].filespec_to_free = NULL; | |
44 | rename_dst[rename_dst_nr].is_rename = 0; | |
25d5ea41 | 45 | rename_dst_nr++; |
f98c2f7e | 46 | return 0; |
427dcb4b JH |
47 | } |
48 | ||
15d061b4 | 49 | /* Table of rename/copy src files */ |
25d5ea41 | 50 | static struct diff_rename_src { |
e88d6bc6 | 51 | struct diff_filepair *p; |
fc580719 | 52 | unsigned short score; /* to remember the break score */ |
25d5ea41 JH |
53 | } *rename_src; |
54 | static int rename_src_nr, rename_src_alloc; | |
427dcb4b | 55 | |
b970b4ef | 56 | static void register_rename_src(struct diff_filepair *p) |
25d5ea41 | 57 | { |
9db2ac56 EN |
58 | if (p->broken_pair) { |
59 | if (!break_idx) { | |
60 | break_idx = xmalloc(sizeof(*break_idx)); | |
61bf4490 | 61 | strintmap_init_with_options(break_idx, -1, NULL, 0); |
9db2ac56 EN |
62 | } |
63 | strintmap_set(break_idx, p->one->path, rename_dst_nr); | |
64 | } | |
65 | ||
337ce247 | 66 | ALLOC_GROW(rename_src, rename_src_nr + 1, rename_src_alloc); |
b970b4ef EN |
67 | rename_src[rename_src_nr].p = p; |
68 | rename_src[rename_src_nr].score = p->score; | |
25d5ea41 | 69 | rename_src_nr++; |
427dcb4b JH |
70 | } |
71 | ||
0ce39643 JS |
72 | static int basename_same(struct diff_filespec *src, struct diff_filespec *dst) |
73 | { | |
74 | int src_len = strlen(src->path), dst_len = strlen(dst->path); | |
75 | while (src_len && dst_len) { | |
76 | char c1 = src->path[--src_len]; | |
77 | char c2 = dst->path[--dst_len]; | |
78 | if (c1 != c2) | |
79 | return 0; | |
80 | if (c1 == '/') | |
81 | return 1; | |
82 | } | |
83 | return (!src_len || src->path[src_len - 1] == '/') && | |
84 | (!dst_len || dst->path[dst_len - 1] == '/'); | |
85 | } | |
86 | ||
427dcb4b | 87 | struct diff_score { |
25d5ea41 JH |
88 | int src; /* index in rename_src */ |
89 | int dst; /* index in rename_dst */ | |
6d24ad97 JH |
90 | unsigned short score; |
91 | short name_score; | |
427dcb4b JH |
92 | }; |
93 | ||
1aedd03a | 94 | struct inexact_prefetch_options { |
95acf11a JT |
95 | struct repository *repo; |
96 | int skip_unmodified; | |
97 | }; | |
1aedd03a | 98 | static void inexact_prefetch(void *prefetch_options) |
95acf11a | 99 | { |
1aedd03a | 100 | struct inexact_prefetch_options *options = prefetch_options; |
95acf11a JT |
101 | int i; |
102 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
103 | ||
104 | for (i = 0; i < rename_dst_nr; i++) { | |
9db2ac56 | 105 | if (rename_dst[i].p->renamed_pair) |
95acf11a JT |
106 | /* |
107 | * The loop in diffcore_rename() will not need these | |
108 | * blobs, so skip prefetching. | |
109 | */ | |
110 | continue; /* already found exact match */ | |
111 | diff_add_if_missing(options->repo, &to_fetch, | |
9db2ac56 | 112 | rename_dst[i].p->two); |
95acf11a JT |
113 | } |
114 | for (i = 0; i < rename_src_nr; i++) { | |
115 | if (options->skip_unmodified && | |
116 | diff_unmodified_pair(rename_src[i].p)) | |
117 | /* | |
118 | * The loop in diffcore_rename() will not need these | |
119 | * blobs, so skip prefetching. | |
120 | */ | |
121 | continue; | |
122 | diff_add_if_missing(options->repo, &to_fetch, | |
123 | rename_src[i].p->one); | |
124 | } | |
125 | promisor_remote_get_direct(options->repo, to_fetch.oid, to_fetch.nr); | |
126 | oid_array_clear(&to_fetch); | |
127 | } | |
128 | ||
b78ea5fc NTND |
129 | static int estimate_similarity(struct repository *r, |
130 | struct diff_filespec *src, | |
427dcb4b | 131 | struct diff_filespec *dst, |
95acf11a | 132 | int minimum_score, |
d331dd3b | 133 | struct diff_populate_filespec_options *dpf_opt) |
427dcb4b JH |
134 | { |
135 | /* src points at a file that existed in the original tree (or | |
136 | * optionally a file in the destination tree) and dst points | |
137 | * at a newly created file. They may be quite similar, in which | |
138 | * case we want to say src is renamed to dst or src is copied into | |
139 | * dst, and then some edit has been applied to dst. | |
140 | * | |
141 | * Compare them and return how similar they are, representing | |
f0c6b2a2 JH |
142 | * the score as an integer between 0 and MAX_SCORE. |
143 | * | |
144 | * When there is an exact match, it is considered a better | |
145 | * match than anything else; the destination does not even | |
146 | * call into this function in that case. | |
427dcb4b | 147 | */ |
90bd932c | 148 | unsigned long max_size, delta_size, base_size, src_copied, literal_added; |
427dcb4b JH |
149 | int score; |
150 | ||
60896c7b JH |
151 | /* We deal only with regular files. Symlink renames are handled |
152 | * only when they are exact matches --- in other words, no edits | |
153 | * after renaming. | |
154 | */ | |
155 | if (!S_ISREG(src->mode) || !S_ISREG(dst->mode)) | |
156 | return 0; | |
157 | ||
81ac051d LT |
158 | /* |
159 | * Need to check that source and destination sizes are | |
160 | * filled in before comparing them. | |
161 | * | |
162 | * If we already have "cnt_data" filled in, we know it's | |
163 | * all good (avoid checking the size for zero, as that | |
164 | * is a possible size - we really should have a flag to | |
165 | * say whether the size is valid or not!) | |
166 | */ | |
d331dd3b EN |
167 | dpf_opt->check_size_only = 1; |
168 | ||
8e5dd3d6 | 169 | if (!src->cnt_data && |
d331dd3b | 170 | diff_populate_filespec(r, src, dpf_opt)) |
81ac051d | 171 | return 0; |
8e5dd3d6 | 172 | if (!dst->cnt_data && |
d331dd3b | 173 | diff_populate_filespec(r, dst, dpf_opt)) |
81ac051d LT |
174 | return 0; |
175 | ||
90bd932c | 176 | max_size = ((src->size > dst->size) ? src->size : dst->size); |
58b103f5 | 177 | base_size = ((src->size < dst->size) ? src->size : dst->size); |
90bd932c | 178 | delta_size = max_size - base_size; |
427dcb4b | 179 | |
58b103f5 JH |
180 | /* We would not consider edits that change the file size so |
181 | * drastically. delta_size must be smaller than | |
cd1870ed | 182 | * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size). |
f0c6b2a2 | 183 | * |
58b103f5 JH |
184 | * Note that base_size == 0 case is handled here already |
185 | * and the final score computation below would not have a | |
186 | * divide-by-zero issue. | |
427dcb4b | 187 | */ |
3a4d6769 | 188 | if (max_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE) |
427dcb4b JH |
189 | return 0; |
190 | ||
d331dd3b | 191 | dpf_opt->check_size_only = 0; |
95acf11a | 192 | |
d331dd3b | 193 | if (!src->cnt_data && diff_populate_filespec(r, src, dpf_opt)) |
885c716f | 194 | return 0; |
d331dd3b | 195 | if (!dst->cnt_data && diff_populate_filespec(r, dst, dpf_opt)) |
885c716f BS |
196 | return 0; |
197 | ||
b78ea5fc | 198 | if (diffcore_count_changes(r, src, dst, |
c06c7966 | 199 | &src->cnt_data, &dst->cnt_data, |
65416758 | 200 | &src_copied, &literal_added)) |
85976974 | 201 | return 0; |
355e76a4 | 202 | |
1706306a JH |
203 | /* How similar are they? |
204 | * what percentage of material in dst are from source? | |
427dcb4b | 205 | */ |
90bd932c | 206 | if (!dst->size) |
1706306a | 207 | score = 0; /* should not happen */ |
cfc0aef1 | 208 | else |
dc49cd76 | 209 | score = (int)(src_copied * MAX_SCORE / max_size); |
427dcb4b JH |
210 | return score; |
211 | } | |
212 | ||
5098bafb | 213 | static void record_rename_pair(int dst_index, int src_index, int score) |
427dcb4b | 214 | { |
9db2ac56 EN |
215 | struct diff_filepair *src = rename_src[src_index].p; |
216 | struct diff_filepair *dst = rename_dst[dst_index].p; | |
f7c1512a | 217 | |
9db2ac56 | 218 | if (dst->renamed_pair) |
25d5ea41 | 219 | die("internal error: dst already matched."); |
427dcb4b | 220 | |
9db2ac56 EN |
221 | src->one->rename_used++; |
222 | src->one->count++; | |
427dcb4b | 223 | |
9db2ac56 EN |
224 | rename_dst[dst_index].filespec_to_free = dst->one; |
225 | rename_dst[dst_index].is_rename = 1; | |
427dcb4b | 226 | |
9db2ac56 EN |
227 | dst->one = src->one; |
228 | dst->renamed_pair = 1; | |
229 | if (!strcmp(dst->one->path, dst->two->path)) | |
230 | dst->score = rename_src[src_index].score; | |
fc580719 | 231 | else |
9db2ac56 | 232 | dst->score = score; |
427dcb4b JH |
233 | } |
234 | ||
235 | /* | |
236 | * We sort the rename similarity matrix with the score, in descending | |
15d061b4 | 237 | * order (the most similar first). |
427dcb4b JH |
238 | */ |
239 | static int score_compare(const void *a_, const void *b_) | |
240 | { | |
241 | const struct diff_score *a = a_, *b = b_; | |
cfc0aef1 | 242 | |
6d24ad97 JH |
243 | /* sink the unused ones to the bottom */ |
244 | if (a->dst < 0) | |
245 | return (0 <= b->dst); | |
246 | else if (b->dst < 0) | |
247 | return -1; | |
248 | ||
cfc0aef1 RS |
249 | if (a->score == b->score) |
250 | return b->name_score - a->name_score; | |
251 | ||
427dcb4b JH |
252 | return b->score - a->score; |
253 | } | |
254 | ||
9027f53c | 255 | struct file_similarity { |
f79d9c58 | 256 | struct hashmap_entry entry; |
7c85f8ac | 257 | int index; |
9027f53c | 258 | struct diff_filespec *filespec; |
9027f53c LT |
259 | }; |
260 | ||
b78ea5fc NTND |
261 | static unsigned int hash_filespec(struct repository *r, |
262 | struct diff_filespec *filespec) | |
48f6407f | 263 | { |
41c9560e | 264 | if (!filespec->oid_valid) { |
1c37e86a | 265 | if (diff_populate_filespec(r, filespec, NULL)) |
48f6407f | 266 | return 0; |
2dcde20e | 267 | hash_object_file(r->hash_algo, filespec->data, filespec->size, |
44439c1c | 268 | OBJ_BLOB, &filespec->oid); |
48f6407f | 269 | } |
d40abc8e | 270 | return oidhash(&filespec->oid); |
48f6407f KB |
271 | } |
272 | ||
f79d9c58 | 273 | static int find_identical_files(struct hashmap *srcs, |
7c85f8ac | 274 | int dst_index, |
11f944dd | 275 | struct diff_options *options) |
9027f53c LT |
276 | { |
277 | int renames = 0; | |
9db2ac56 | 278 | struct diff_filespec *target = rename_dst[dst_index].p->two; |
ab73a9d1 | 279 | struct file_similarity *p, *best = NULL; |
48f6407f | 280 | int i = 100, best_score = -1; |
f0e63c41 | 281 | unsigned int hash = hash_filespec(options->repo, target); |
48f6407f KB |
282 | |
283 | /* | |
7c85f8ac | 284 | * Find the best source match for specified destination. |
48f6407f | 285 | */ |
f0e63c41 EW |
286 | p = hashmap_get_entry_from_hash(srcs, hash, NULL, |
287 | struct file_similarity, entry); | |
23dee69f | 288 | hashmap_for_each_entry_from(srcs, p, entry) { |
48f6407f KB |
289 | int score; |
290 | struct diff_filespec *source = p->filespec; | |
291 | ||
292 | /* False hash collision? */ | |
9001dc2a | 293 | if (!oideq(&source->oid, &target->oid)) |
48f6407f KB |
294 | continue; |
295 | /* Non-regular files? If so, the modes must match! */ | |
296 | if (!S_ISREG(source->mode) || !S_ISREG(target->mode)) { | |
297 | if (source->mode != target->mode) | |
0940e5f2 | 298 | continue; |
9027f53c | 299 | } |
48f6407f KB |
300 | /* Give higher scores to sources that haven't been used already */ |
301 | score = !source->rename_used; | |
302 | if (source->rename_used && options->detect_rename != DIFF_DETECT_COPY) | |
303 | continue; | |
304 | score += basename_same(source, target); | |
305 | if (score > best_score) { | |
306 | best = p; | |
307 | best_score = score; | |
308 | if (score == 2) | |
309 | break; | |
9027f53c | 310 | } |
48f6407f KB |
311 | |
312 | /* Too many identical alternatives? Pick one */ | |
313 | if (!--i) | |
314 | break; | |
315 | } | |
316 | if (best) { | |
7c85f8ac | 317 | record_rename_pair(dst_index, best->index, MAX_SCORE); |
48f6407f KB |
318 | renames++; |
319 | } | |
9027f53c LT |
320 | return renames; |
321 | } | |
322 | ||
b78ea5fc | 323 | static void insert_file_table(struct repository *r, |
fa0e936f | 324 | struct mem_pool *pool, |
b78ea5fc NTND |
325 | struct hashmap *table, int index, |
326 | struct diff_filespec *filespec) | |
9027f53c | 327 | { |
fa0e936f | 328 | struct file_similarity *entry = mem_pool_alloc(pool, sizeof(*entry)); |
9027f53c | 329 | |
9027f53c LT |
330 | entry->index = index; |
331 | entry->filespec = filespec; | |
9027f53c | 332 | |
d22245a2 | 333 | hashmap_entry_init(&entry->entry, hash_filespec(r, filespec)); |
b94e5c1d | 334 | hashmap_add(table, &entry->entry); |
9027f53c LT |
335 | } |
336 | ||
cb1491b6 LT |
337 | /* |
338 | * Find exact renames first. | |
339 | * | |
340 | * The first round matches up the up-to-date entries, | |
341 | * and then during the second round we try to match | |
342 | * cache-dirty entries as well. | |
cb1491b6 | 343 | */ |
fa0e936f EN |
344 | static int find_exact_renames(struct diff_options *options, |
345 | struct mem_pool *pool) | |
cb1491b6 | 346 | { |
7c85f8ac | 347 | int i, renames = 0; |
f79d9c58 | 348 | struct hashmap file_table; |
cb1491b6 | 349 | |
ca4e3ca0 SG |
350 | /* Add all sources to the hash table in reverse order, because |
351 | * later on they will be retrieved in LIFO order. | |
352 | */ | |
7663cdc8 | 353 | hashmap_init(&file_table, NULL, NULL, rename_src_nr); |
ca4e3ca0 | 354 | for (i = rename_src_nr-1; i >= 0; i--) |
fa0e936f | 355 | insert_file_table(options->repo, pool, |
b78ea5fc NTND |
356 | &file_table, i, |
357 | rename_src[i].p->one); | |
9027f53c | 358 | |
7c85f8ac | 359 | /* Walk the destinations and find best source match */ |
9027f53c | 360 | for (i = 0; i < rename_dst_nr; i++) |
7c85f8ac | 361 | renames += find_identical_files(&file_table, i, options); |
9027f53c | 362 | |
fa0e936f EN |
363 | /* Free the hash data structure (entries will be freed with the pool) */ |
364 | hashmap_clear(&file_table); | |
9027f53c | 365 | |
7c85f8ac | 366 | return renames; |
cb1491b6 LT |
367 | } |
368 | ||
bde8b9f3 EN |
369 | struct dir_rename_info { |
370 | struct strintmap idx_map; | |
371 | struct strmap dir_rename_guess; | |
372 | struct strmap *dir_rename_count; | |
a49b55d5 | 373 | struct strintmap *relevant_source_dirs; |
bde8b9f3 EN |
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 | ||
81afdf7a EN |
391 | static const char *get_highest_rename_path(struct strintmap *counts) |
392 | { | |
393 | int highest_count = 0; | |
394 | const char *highest_destination_dir = NULL; | |
395 | struct hashmap_iter iter; | |
396 | struct strmap_entry *entry; | |
397 | ||
398 | strintmap_for_each_entry(counts, &iter, entry) { | |
399 | const char *destination_dir = entry->key; | |
400 | intptr_t count = (intptr_t)entry->value; | |
401 | if (count > highest_count) { | |
402 | highest_count = count; | |
403 | highest_destination_dir = destination_dir; | |
404 | } | |
405 | } | |
406 | return highest_destination_dir; | |
407 | } | |
408 | ||
0491d392 EN |
409 | static char *UNKNOWN_DIR = "/"; /* placeholder -- short, illegal directory */ |
410 | ||
411 | static int dir_rename_already_determinable(struct strintmap *counts) | |
412 | { | |
413 | struct hashmap_iter iter; | |
414 | struct strmap_entry *entry; | |
415 | int first = 0, second = 0, unknown = 0; | |
416 | strintmap_for_each_entry(counts, &iter, entry) { | |
417 | const char *destination_dir = entry->key; | |
418 | intptr_t count = (intptr_t)entry->value; | |
419 | if (!strcmp(destination_dir, UNKNOWN_DIR)) { | |
420 | unknown = count; | |
421 | } else if (count >= first) { | |
422 | second = first; | |
423 | first = count; | |
424 | } else if (count >= second) { | |
425 | second = count; | |
426 | } | |
427 | } | |
428 | return first > second + unknown; | |
429 | } | |
430 | ||
b6e3d274 | 431 | static void increment_count(struct dir_rename_info *info, |
0c4fd732 EN |
432 | char *old_dir, |
433 | char *new_dir) | |
434 | { | |
435 | struct strintmap *counts; | |
436 | struct strmap_entry *e; | |
437 | ||
438 | /* Get the {new_dirs -> counts} mapping using old_dir */ | |
b6e3d274 | 439 | e = strmap_get_entry(info->dir_rename_count, old_dir); |
0c4fd732 EN |
440 | if (e) { |
441 | counts = e->value; | |
442 | } else { | |
443 | counts = xmalloc(sizeof(*counts)); | |
444 | strintmap_init_with_options(counts, 0, NULL, 1); | |
b6e3d274 | 445 | strmap_put(info->dir_rename_count, old_dir, counts); |
0c4fd732 EN |
446 | } |
447 | ||
448 | /* Increment the count for new_dir */ | |
449 | strintmap_incr(counts, new_dir, 1); | |
450 | } | |
451 | ||
b6e3d274 | 452 | static void update_dir_rename_counts(struct dir_rename_info *info, |
a49b55d5 | 453 | struct strintmap *dirs_removed, |
0c4fd732 EN |
454 | const char *oldname, |
455 | const char *newname) | |
456 | { | |
4e3250b7 AH |
457 | char *old_dir; |
458 | char *new_dir; | |
459 | const char new_dir_first_char = newname[0]; | |
0c4fd732 EN |
460 | int first_time_in_loop = 1; |
461 | ||
1ad69eb0 EN |
462 | if (!info->setup) |
463 | /* | |
464 | * info->setup is 0 here in two cases: (1) all auxiliary | |
465 | * vars (like dirs_removed) were NULL so | |
466 | * initialize_dir_rename_info() returned early, or (2) | |
467 | * either break detection or copy detection are active so | |
468 | * that we never called initialize_dir_rename_info(). In | |
469 | * the former case, we don't have enough info to know if | |
470 | * directories were renamed (because dirs_removed lets us | |
471 | * know about a necessary prerequisite, namely if they were | |
472 | * removed), and in the latter, we don't care about | |
473 | * directory renames or find_basename_matches. | |
474 | * | |
475 | * This matters because both basename and inexact matching | |
476 | * will also call update_dir_rename_counts(). In either of | |
477 | * the above two cases info->dir_rename_counts will not | |
478 | * have been properly initialized which prevents us from | |
479 | * updating it, but in these two cases we don't care about | |
480 | * dir_rename_counts anyway, so we can just exit early. | |
481 | */ | |
482 | return; | |
483 | ||
4e3250b7 AH |
484 | |
485 | old_dir = xstrdup(oldname); | |
486 | new_dir = xstrdup(newname); | |
487 | ||
0c4fd732 | 488 | while (1) { |
e54385b9 EN |
489 | int drd_flag = NOT_RELEVANT; |
490 | ||
333899e1 | 491 | /* Get old_dir, skip if its directory isn't relevant. */ |
0c4fd732 | 492 | dirname_munge(old_dir); |
333899e1 | 493 | if (info->relevant_source_dirs && |
a49b55d5 | 494 | !strintmap_contains(info->relevant_source_dirs, old_dir)) |
333899e1 EN |
495 | break; |
496 | ||
497 | /* Get new_dir */ | |
0c4fd732 EN |
498 | dirname_munge(new_dir); |
499 | ||
500 | /* | |
501 | * When renaming | |
502 | * "a/b/c/d/e/foo.c" -> "a/b/some/thing/else/e/foo.c" | |
503 | * then this suggests that both | |
504 | * a/b/c/d/e/ => a/b/some/thing/else/e/ | |
505 | * a/b/c/d/ => a/b/some/thing/else/ | |
506 | * so we want to increment counters for both. We do NOT, | |
507 | * however, also want to suggest that there was the following | |
508 | * rename: | |
509 | * a/b/c/ => a/b/some/thing/ | |
510 | * so we need to quit at that point. | |
511 | * | |
512 | * Note the when first_time_in_loop, we only strip off the | |
513 | * basename, and we don't care if that's different. | |
514 | */ | |
515 | if (!first_time_in_loop) { | |
516 | char *old_sub_dir = strchr(old_dir, '\0')+1; | |
517 | char *new_sub_dir = strchr(new_dir, '\0')+1; | |
518 | if (!*new_dir) { | |
519 | /* | |
520 | * Special case when renaming to root directory, | |
521 | * i.e. when new_dir == "". In this case, we had | |
522 | * something like | |
523 | * a/b/subdir => subdir | |
524 | * and so dirname_munge() sets things up so that | |
525 | * old_dir = "a/b\0subdir\0" | |
526 | * new_dir = "\0ubdir\0" | |
527 | * We didn't have a '/' to overwrite a '\0' onto | |
528 | * in new_dir, so we have to compare differently. | |
529 | */ | |
530 | if (new_dir_first_char != old_sub_dir[0] || | |
531 | strcmp(old_sub_dir+1, new_sub_dir)) | |
532 | break; | |
533 | } else { | |
534 | if (strcmp(old_sub_dir, new_sub_dir)) | |
535 | break; | |
536 | } | |
537 | } | |
538 | ||
e54385b9 EN |
539 | /* |
540 | * Above we suggested that we'd keep recording renames for | |
541 | * all ancestor directories where the trailing directories | |
542 | * matched, i.e. for | |
543 | * "a/b/c/d/e/foo.c" -> "a/b/some/thing/else/e/foo.c" | |
544 | * we'd increment rename counts for each of | |
545 | * a/b/c/d/e/ => a/b/some/thing/else/e/ | |
546 | * a/b/c/d/ => a/b/some/thing/else/ | |
547 | * However, we only need the rename counts for directories | |
548 | * in dirs_removed whose value is RELEVANT_FOR_SELF. | |
549 | * However, we add one special case of also recording it for | |
550 | * first_time_in_loop because find_basename_matches() can | |
551 | * use that as a hint to find a good pairing. | |
552 | */ | |
553 | if (dirs_removed) | |
554 | drd_flag = strintmap_get(dirs_removed, old_dir); | |
555 | if (drd_flag == RELEVANT_FOR_SELF || first_time_in_loop) | |
b6e3d274 | 556 | increment_count(info, old_dir, new_dir); |
0c4fd732 | 557 | |
e54385b9 EN |
558 | first_time_in_loop = 0; |
559 | if (drd_flag == NOT_RELEVANT) | |
560 | break; | |
0c4fd732 EN |
561 | /* If we hit toplevel directory ("") for old or new dir, quit */ |
562 | if (!*old_dir || !*new_dir) | |
563 | break; | |
0c4fd732 EN |
564 | } |
565 | ||
566 | /* Free resources we don't need anymore */ | |
567 | free(old_dir); | |
568 | free(new_dir); | |
569 | } | |
570 | ||
1ad69eb0 | 571 | static void initialize_dir_rename_info(struct dir_rename_info *info, |
a49b55d5 EN |
572 | struct strintmap *relevant_sources, |
573 | struct strintmap *dirs_removed, | |
25e65b6d EN |
574 | struct strmap *dir_rename_count, |
575 | struct strmap *cached_pairs) | |
0c4fd732 | 576 | { |
81afdf7a EN |
577 | struct hashmap_iter iter; |
578 | struct strmap_entry *entry; | |
0c4fd732 EN |
579 | int i; |
580 | ||
e4fd06e7 | 581 | if (!dirs_removed && !relevant_sources) { |
1ad69eb0 EN |
582 | info->setup = 0; |
583 | return; | |
0c4fd732 | 584 | } |
ae8cf74d EN |
585 | info->setup = 1; |
586 | ||
1ad69eb0 EN |
587 | info->dir_rename_count = dir_rename_count; |
588 | if (!info->dir_rename_count) { | |
589 | info->dir_rename_count = xmalloc(sizeof(*dir_rename_count)); | |
590 | strmap_init(info->dir_rename_count); | |
591 | } | |
ae8cf74d EN |
592 | strintmap_init_with_options(&info->idx_map, -1, NULL, 0); |
593 | strmap_init_with_options(&info->dir_rename_guess, NULL, 0); | |
ae8cf74d | 594 | |
333899e1 | 595 | /* Setup info->relevant_source_dirs */ |
e4fd06e7 EN |
596 | info->relevant_source_dirs = NULL; |
597 | if (dirs_removed || !relevant_sources) { | |
598 | info->relevant_source_dirs = dirs_removed; /* might be NULL */ | |
599 | } else { | |
600 | info->relevant_source_dirs = xmalloc(sizeof(struct strintmap)); | |
a49b55d5 EN |
601 | strintmap_init(info->relevant_source_dirs, 0 /* unused */); |
602 | strintmap_for_each_entry(relevant_sources, &iter, entry) { | |
e4fd06e7 EN |
603 | char *dirname = get_dirname(entry->key); |
604 | if (!dirs_removed || | |
a49b55d5 EN |
605 | strintmap_contains(dirs_removed, dirname)) |
606 | strintmap_set(info->relevant_source_dirs, | |
607 | dirname, 0 /* value irrelevant */); | |
e4fd06e7 EN |
608 | free(dirname); |
609 | } | |
610 | } | |
333899e1 | 611 | |
ae8cf74d | 612 | /* |
1ad69eb0 EN |
613 | * Loop setting up both info->idx_map, and doing setup of |
614 | * info->dir_rename_count. | |
ae8cf74d EN |
615 | */ |
616 | for (i = 0; i < rename_dst_nr; ++i) { | |
617 | /* | |
618 | * For non-renamed files, make idx_map contain mapping of | |
619 | * filename -> index (index within rename_dst, that is) | |
620 | */ | |
621 | if (!rename_dst[i].is_rename) { | |
622 | char *filename = rename_dst[i].p->two->path; | |
623 | strintmap_set(&info->idx_map, filename, i); | |
1ad69eb0 | 624 | continue; |
ae8cf74d | 625 | } |
1ad69eb0 EN |
626 | |
627 | /* | |
628 | * For everything else (i.e. renamed files), make | |
629 | * dir_rename_count contain a map of a map: | |
630 | * old_directory -> {new_directory -> count} | |
631 | * In other words, for every pair look at the directories for | |
632 | * the old filename and the new filename and count how many | |
633 | * times that pairing occurs. | |
634 | */ | |
635 | update_dir_rename_counts(info, dirs_removed, | |
636 | rename_dst[i].p->one->path, | |
637 | rename_dst[i].p->two->path); | |
ae8cf74d | 638 | } |
81afdf7a | 639 | |
25e65b6d EN |
640 | /* Add cached_pairs to counts */ |
641 | strmap_for_each_entry(cached_pairs, &iter, entry) { | |
642 | const char *old_name = entry->key; | |
643 | const char *new_name = entry->value; | |
644 | if (!new_name) | |
645 | /* known delete; ignore it */ | |
646 | continue; | |
647 | ||
648 | update_dir_rename_counts(info, dirs_removed, old_name, new_name); | |
649 | } | |
650 | ||
81afdf7a EN |
651 | /* |
652 | * Now we collapse | |
653 | * dir_rename_count: old_directory -> {new_directory -> count} | |
654 | * down to | |
655 | * dir_rename_guess: old_directory -> best_new_directory | |
656 | * where best_new_directory is the one with the highest count. | |
657 | */ | |
658 | strmap_for_each_entry(info->dir_rename_count, &iter, entry) { | |
659 | /* entry->key is source_dir */ | |
660 | struct strintmap *counts = entry->value; | |
661 | char *best_newdir; | |
662 | ||
663 | best_newdir = xstrdup(get_highest_rename_path(counts)); | |
664 | strmap_put(&info->dir_rename_guess, entry->key, | |
665 | best_newdir); | |
666 | } | |
ae8cf74d EN |
667 | } |
668 | ||
cd52e005 EN |
669 | void partial_clear_dir_rename_count(struct strmap *dir_rename_count) |
670 | { | |
671 | struct hashmap_iter iter; | |
672 | struct strmap_entry *entry; | |
673 | ||
674 | strmap_for_each_entry(dir_rename_count, &iter, entry) { | |
675 | struct strintmap *counts = entry->value; | |
676 | strintmap_clear(counts); | |
677 | } | |
678 | strmap_partial_clear(dir_rename_count, 1); | |
679 | } | |
680 | ||
b1473019 | 681 | static void cleanup_dir_rename_info(struct dir_rename_info *info, |
a49b55d5 | 682 | struct strintmap *dirs_removed, |
b1473019 | 683 | int keep_dir_rename_count) |
ae8cf74d | 684 | { |
b1473019 EN |
685 | struct hashmap_iter iter; |
686 | struct strmap_entry *entry; | |
687 | struct string_list to_remove = STRING_LIST_INIT_NODUP; | |
688 | int i; | |
689 | ||
ae8cf74d EN |
690 | if (!info->setup) |
691 | return; | |
692 | ||
693 | /* idx_map */ | |
694 | strintmap_clear(&info->idx_map); | |
695 | ||
696 | /* dir_rename_guess */ | |
697 | strmap_clear(&info->dir_rename_guess, 1); | |
698 | ||
e4fd06e7 EN |
699 | /* relevant_source_dirs */ |
700 | if (info->relevant_source_dirs && | |
701 | info->relevant_source_dirs != dirs_removed) { | |
a49b55d5 | 702 | strintmap_clear(info->relevant_source_dirs); |
e4fd06e7 EN |
703 | FREE_AND_NULL(info->relevant_source_dirs); |
704 | } | |
705 | ||
b6e3d274 | 706 | /* dir_rename_count */ |
b1473019 EN |
707 | if (!keep_dir_rename_count) { |
708 | partial_clear_dir_rename_count(info->dir_rename_count); | |
709 | strmap_clear(info->dir_rename_count, 1); | |
710 | FREE_AND_NULL(info->dir_rename_count); | |
711 | return; | |
712 | } | |
713 | ||
714 | /* | |
715 | * Although dir_rename_count was passed in | |
716 | * diffcore_rename_extended() and we want to keep it around and | |
bf238b71 EN |
717 | * return it to that caller, we first want to remove any counts in |
718 | * the maps associated with UNKNOWN_DIR entries and any data | |
b1473019 EN |
719 | * associated with directories that weren't renamed. |
720 | */ | |
721 | strmap_for_each_entry(info->dir_rename_count, &iter, entry) { | |
722 | const char *source_dir = entry->key; | |
723 | struct strintmap *counts = entry->value; | |
724 | ||
fb52938e | 725 | if (!strintmap_get(dirs_removed, source_dir)) { |
b1473019 EN |
726 | string_list_append(&to_remove, source_dir); |
727 | strintmap_clear(counts); | |
728 | continue; | |
729 | } | |
bf238b71 EN |
730 | |
731 | if (strintmap_contains(counts, UNKNOWN_DIR)) | |
732 | strintmap_remove(counts, UNKNOWN_DIR); | |
b1473019 EN |
733 | } |
734 | for (i = 0; i < to_remove.nr; ++i) | |
735 | strmap_remove(info->dir_rename_count, | |
736 | to_remove.items[i].string, 1); | |
737 | string_list_clear(&to_remove, 0); | |
ae8cf74d EN |
738 | } |
739 | ||
a35df337 EN |
740 | static const char *get_basename(const char *filename) |
741 | { | |
742 | /* | |
743 | * gitbasename() has to worry about special drives, multiple | |
744 | * directory separator characters, trailing slashes, NULL or | |
745 | * empty strings, etc. We only work on filenames as stored in | |
746 | * git, and thus get to ignore all those complications. | |
747 | */ | |
748 | const char *base = strrchr(filename, '/'); | |
749 | return base ? base + 1 : filename; | |
750 | } | |
751 | ||
bde8b9f3 | 752 | static int idx_possible_rename(char *filename, struct dir_rename_info *info) |
37a25143 | 753 | { |
bde8b9f3 EN |
754 | /* |
755 | * Our comparison of files with the same basename (see | |
756 | * find_basename_matches() below), is only helpful when after exact | |
757 | * rename detection we have exactly one file with a given basename | |
758 | * among the rename sources and also only exactly one file with | |
759 | * that basename among the rename destinations. When we have | |
760 | * multiple files with the same basename in either set, we do not | |
761 | * know which to compare against. However, there are some | |
762 | * filenames that occur in large numbers (particularly | |
763 | * build-related filenames such as 'Makefile', '.gitignore', or | |
764 | * 'build.gradle' that potentially exist within every single | |
765 | * subdirectory), and for performance we want to be able to quickly | |
766 | * find renames for these files too. | |
767 | * | |
768 | * The reason basename comparisons are a useful heuristic was that it | |
769 | * is common for people to move files across directories while keeping | |
770 | * their filename the same. If we had a way of determining or even | |
771 | * making a good educated guess about which directory these non-unique | |
772 | * basename files had moved the file to, we could check it. | |
773 | * Luckily... | |
774 | * | |
775 | * When an entire directory is in fact renamed, we have two factors | |
776 | * helping us out: | |
777 | * (a) the original directory disappeared giving us a hint | |
778 | * about when we can apply an extra heuristic. | |
779 | * (a) we often have several files within that directory and | |
780 | * subdirectories that are renamed without changes | |
781 | * So, rules for a heuristic: | |
782 | * (0) If there basename matches are non-unique (the condition under | |
783 | * which this function is called) AND | |
784 | * (1) the directory in which the file was found has disappeared | |
785 | * (i.e. dirs_removed is non-NULL and has a relevant entry) THEN | |
786 | * (2) use exact renames of files within the directory to determine | |
787 | * where the directory is likely to have been renamed to. IF | |
788 | * there is at least one exact rename from within that | |
789 | * directory, we can proceed. | |
790 | * (3) If there are multiple places the directory could have been | |
791 | * renamed to based on exact renames, ignore all but one of them. | |
792 | * Just use the destination with the most renames going to it. | |
793 | * (4) Check if applying that directory rename to the original file | |
794 | * would result in a destination filename that is in the | |
795 | * potential rename set. If so, return the index of the | |
796 | * destination file (the index within rename_dst). | |
797 | * (5) Compare the original file and returned destination for | |
798 | * similarity, and if they are sufficiently similar, record the | |
799 | * rename. | |
800 | * | |
801 | * This function, idx_possible_rename(), is only responsible for (4). | |
81afdf7a EN |
802 | * The conditions/steps in (1)-(3) are handled via setting up |
803 | * dir_rename_count and dir_rename_guess in | |
804 | * initialize_dir_rename_info(). Steps (0) and (5) are handled by | |
805 | * the caller of this function. | |
bde8b9f3 EN |
806 | */ |
807 | char *old_dir, *new_dir; | |
808 | struct strbuf new_path = STRBUF_INIT; | |
809 | int idx; | |
810 | ||
811 | if (!info->setup) | |
812 | return -1; | |
813 | ||
814 | old_dir = get_dirname(filename); | |
815 | new_dir = strmap_get(&info->dir_rename_guess, old_dir); | |
816 | free(old_dir); | |
817 | if (!new_dir) | |
818 | return -1; | |
819 | ||
820 | strbuf_addstr(&new_path, new_dir); | |
821 | strbuf_addch(&new_path, '/'); | |
822 | strbuf_addstr(&new_path, get_basename(filename)); | |
823 | ||
824 | idx = strintmap_get(&info->idx_map, new_path.buf); | |
825 | strbuf_release(&new_path); | |
826 | return idx; | |
37a25143 EN |
827 | } |
828 | ||
1aedd03a EN |
829 | struct basename_prefetch_options { |
830 | struct repository *repo; | |
831 | struct strintmap *relevant_sources; | |
832 | struct strintmap *sources; | |
833 | struct strintmap *dests; | |
834 | struct dir_rename_info *info; | |
835 | }; | |
836 | static void basename_prefetch(void *prefetch_options) | |
837 | { | |
838 | struct basename_prefetch_options *options = prefetch_options; | |
839 | struct strintmap *relevant_sources = options->relevant_sources; | |
840 | struct strintmap *sources = options->sources; | |
841 | struct strintmap *dests = options->dests; | |
842 | struct dir_rename_info *info = options->info; | |
843 | int i; | |
844 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
845 | ||
846 | /* | |
847 | * TODO: The following loops mirror the code/logic from | |
848 | * find_basename_matches(), though not quite exactly. Maybe | |
849 | * abstract the iteration logic out somehow? | |
850 | */ | |
851 | for (i = 0; i < rename_src_nr; ++i) { | |
852 | char *filename = rename_src[i].p->one->path; | |
853 | const char *base = NULL; | |
854 | intptr_t src_index; | |
855 | intptr_t dst_index; | |
856 | ||
857 | /* Skip irrelevant sources */ | |
858 | if (relevant_sources && | |
859 | !strintmap_contains(relevant_sources, filename)) | |
860 | continue; | |
861 | ||
862 | /* | |
863 | * If the basename is unique among remaining sources, then | |
864 | * src_index will equal 'i' and we can attempt to match it | |
865 | * to a unique basename in the destinations. Otherwise, | |
866 | * use directory rename heuristics, if possible. | |
867 | */ | |
868 | base = get_basename(filename); | |
869 | src_index = strintmap_get(sources, base); | |
870 | assert(src_index == -1 || src_index == i); | |
871 | ||
872 | if (strintmap_contains(dests, base)) { | |
873 | struct diff_filespec *one, *two; | |
874 | ||
875 | /* Find a matching destination, if possible */ | |
876 | dst_index = strintmap_get(dests, base); | |
877 | if (src_index == -1 || dst_index == -1) { | |
878 | src_index = i; | |
879 | dst_index = idx_possible_rename(filename, info); | |
880 | } | |
881 | if (dst_index == -1) | |
882 | continue; | |
883 | ||
884 | /* Ignore this dest if already used in a rename */ | |
885 | if (rename_dst[dst_index].is_rename) | |
886 | continue; /* already used previously */ | |
887 | ||
888 | one = rename_src[src_index].p->one; | |
889 | two = rename_dst[dst_index].p->two; | |
890 | ||
891 | /* Add the pairs */ | |
892 | diff_add_if_missing(options->repo, &to_fetch, two); | |
893 | diff_add_if_missing(options->repo, &to_fetch, one); | |
894 | } | |
895 | } | |
896 | ||
897 | promisor_remote_get_direct(options->repo, to_fetch.oid, to_fetch.nr); | |
898 | oid_array_clear(&to_fetch); | |
899 | } | |
900 | ||
a35df337 | 901 | static int find_basename_matches(struct diff_options *options, |
bde8b9f3 | 902 | int minimum_score, |
1ad69eb0 | 903 | struct dir_rename_info *info, |
a49b55d5 EN |
904 | struct strintmap *relevant_sources, |
905 | struct strintmap *dirs_removed) | |
a35df337 | 906 | { |
da09f651 EN |
907 | /* |
908 | * When I checked in early 2020, over 76% of file renames in linux | |
909 | * just moved files to a different directory but kept the same | |
910 | * basename. gcc did that with over 64% of renames, gecko did it | |
911 | * with over 79%, and WebKit did it with over 89%. | |
912 | * | |
913 | * Therefore we can bypass the normal exhaustive NxM matrix | |
914 | * comparison of similarities between all potential rename sources | |
915 | * and destinations by instead using file basename as a hint (i.e. | |
916 | * the portion of the filename after the last '/'), checking for | |
917 | * similarity between files with the same basename, and if we find | |
918 | * a pair that are sufficiently similar, record the rename pair and | |
919 | * exclude those two from the NxM matrix. | |
920 | * | |
921 | * This *might* cause us to find a less than optimal pairing (if | |
922 | * there is another file that we are even more similar to but has a | |
923 | * different basename). Given the huge performance advantage | |
924 | * basename matching provides, and given the frequency with which | |
925 | * people use the same basename in real world projects, that's a | |
926 | * trade-off we are willing to accept when doing just rename | |
927 | * detection. | |
928 | * | |
929 | * If someone wants copy detection that implies they are willing to | |
930 | * spend more cycles to find similarities between files, so it may | |
931 | * be less likely that this heuristic is wanted. If someone is | |
932 | * doing break detection, that means they do not want filename | |
933 | * similarity to imply any form of content similiarity, and thus | |
934 | * this heuristic would definitely be incompatible. | |
935 | */ | |
936 | ||
937 | int i, renames = 0; | |
a35df337 EN |
938 | struct strintmap sources; |
939 | struct strintmap dests; | |
d331dd3b EN |
940 | struct diff_populate_filespec_options dpf_options = { |
941 | .check_binary = 0, | |
942 | .missing_object_cb = NULL, | |
943 | .missing_object_data = NULL | |
944 | }; | |
1aedd03a | 945 | struct basename_prefetch_options prefetch_options = { |
d331dd3b | 946 | .repo = options->repo, |
1aedd03a EN |
947 | .relevant_sources = relevant_sources, |
948 | .sources = &sources, | |
949 | .dests = &dests, | |
950 | .info = info | |
d331dd3b | 951 | }; |
a35df337 EN |
952 | |
953 | /* | |
954 | * Create maps of basename -> fullname(s) for remaining sources and | |
955 | * dests. | |
956 | */ | |
957 | strintmap_init_with_options(&sources, -1, NULL, 0); | |
958 | strintmap_init_with_options(&dests, -1, NULL, 0); | |
959 | for (i = 0; i < rename_src_nr; ++i) { | |
960 | char *filename = rename_src[i].p->one->path; | |
961 | const char *base; | |
962 | ||
963 | /* exact renames removed in remove_unneeded_paths_from_src() */ | |
964 | assert(!rename_src[i].p->one->rename_used); | |
965 | ||
966 | /* Record index within rename_src (i) if basename is unique */ | |
967 | base = get_basename(filename); | |
968 | if (strintmap_contains(&sources, base)) | |
969 | strintmap_set(&sources, base, -1); | |
970 | else | |
971 | strintmap_set(&sources, base, i); | |
972 | } | |
973 | for (i = 0; i < rename_dst_nr; ++i) { | |
974 | char *filename = rename_dst[i].p->two->path; | |
975 | const char *base; | |
976 | ||
977 | if (rename_dst[i].is_rename) | |
978 | continue; /* involved in exact match already. */ | |
979 | ||
980 | /* Record index within rename_dst (i) if basename is unique */ | |
981 | base = get_basename(filename); | |
982 | if (strintmap_contains(&dests, base)) | |
983 | strintmap_set(&dests, base, -1); | |
984 | else | |
985 | strintmap_set(&dests, base, i); | |
986 | } | |
987 | ||
a5183d76 | 988 | if (options->repo == the_repository && repo_has_promisor_remote(the_repository)) { |
1aedd03a | 989 | dpf_options.missing_object_cb = basename_prefetch; |
d331dd3b EN |
990 | dpf_options.missing_object_data = &prefetch_options; |
991 | } | |
992 | ||
da09f651 | 993 | /* Now look for basename matchups and do similarity estimation */ |
37a25143 EN |
994 | for (i = 0; i < rename_src_nr; ++i) { |
995 | char *filename = rename_src[i].p->one->path; | |
996 | const char *base = NULL; | |
997 | intptr_t src_index; | |
da09f651 | 998 | intptr_t dst_index; |
da09f651 | 999 | |
e4fd06e7 EN |
1000 | /* Skip irrelevant sources */ |
1001 | if (relevant_sources && | |
a49b55d5 | 1002 | !strintmap_contains(relevant_sources, filename)) |
e4fd06e7 EN |
1003 | continue; |
1004 | ||
37a25143 EN |
1005 | /* |
1006 | * If the basename is unique among remaining sources, then | |
1007 | * src_index will equal 'i' and we can attempt to match it | |
1008 | * to a unique basename in the destinations. Otherwise, | |
1009 | * use directory rename heuristics, if possible. | |
1010 | */ | |
1011 | base = get_basename(filename); | |
1012 | src_index = strintmap_get(&sources, base); | |
1013 | assert(src_index == -1 || src_index == i); | |
1014 | ||
1015 | if (strintmap_contains(&dests, base)) { | |
da09f651 EN |
1016 | struct diff_filespec *one, *two; |
1017 | int score; | |
1018 | ||
37a25143 EN |
1019 | /* Find a matching destination, if possible */ |
1020 | dst_index = strintmap_get(&dests, base); | |
1021 | if (src_index == -1 || dst_index == -1) { | |
1022 | src_index = i; | |
bde8b9f3 | 1023 | dst_index = idx_possible_rename(filename, info); |
37a25143 EN |
1024 | } |
1025 | if (dst_index == -1) | |
1026 | continue; | |
1027 | ||
1028 | /* Ignore this dest if already used in a rename */ | |
1029 | if (rename_dst[dst_index].is_rename) | |
1030 | continue; /* already used previously */ | |
1031 | ||
da09f651 EN |
1032 | /* Estimate the similarity */ |
1033 | one = rename_src[src_index].p->one; | |
1034 | two = rename_dst[dst_index].p->two; | |
1035 | score = estimate_similarity(options->repo, one, two, | |
d331dd3b | 1036 | minimum_score, &dpf_options); |
da09f651 EN |
1037 | |
1038 | /* If sufficiently similar, record as rename pair */ | |
1039 | if (score < minimum_score) | |
1040 | continue; | |
1041 | record_rename_pair(dst_index, src_index, score); | |
1042 | renames++; | |
1ad69eb0 EN |
1043 | update_dir_rename_counts(info, dirs_removed, |
1044 | one->path, two->path); | |
da09f651 EN |
1045 | |
1046 | /* | |
1047 | * Found a rename so don't need text anymore; if we | |
1048 | * didn't find a rename, the filespec_blob would get | |
1049 | * re-used when doing the matrix of comparisons. | |
1050 | */ | |
1051 | diff_free_filespec_blob(one); | |
1052 | diff_free_filespec_blob(two); | |
1053 | } | |
1054 | } | |
a35df337 EN |
1055 | |
1056 | strintmap_clear(&sources); | |
1057 | strintmap_clear(&dests); | |
1058 | ||
da09f651 | 1059 | return renames; |
a35df337 EN |
1060 | } |
1061 | ||
6d24ad97 JH |
1062 | #define NUM_CANDIDATE_PER_DST 4 |
1063 | static void record_if_better(struct diff_score m[], struct diff_score *o) | |
1064 | { | |
1065 | int i, worst; | |
1066 | ||
1067 | /* find the worst one */ | |
1068 | worst = 0; | |
1069 | for (i = 1; i < NUM_CANDIDATE_PER_DST; i++) | |
1070 | if (score_compare(&m[i], &m[worst]) > 0) | |
1071 | worst = i; | |
1072 | ||
1073 | /* is it better than the worst one? */ | |
1074 | if (score_compare(&m[worst], o) > 0) | |
1075 | m[worst] = *o; | |
1076 | } | |
1077 | ||
f31027c9 JH |
1078 | /* |
1079 | * Returns: | |
1080 | * 0 if we are under the limit; | |
1081 | * 1 if we need to disable inexact rename detection; | |
1082 | * 2 if we would be under the limit if we were given -C instead of -C -C. | |
1083 | */ | |
00b8cccd | 1084 | static int too_many_rename_candidates(int num_destinations, int num_sources, |
9d8a5a50 JH |
1085 | struct diff_options *options) |
1086 | { | |
1087 | int rename_limit = options->rename_limit; | |
00b8cccd | 1088 | int i, limited_sources; |
9d8a5a50 JH |
1089 | |
1090 | options->needed_rename_limit = 0; | |
1091 | ||
1092 | /* | |
1093 | * This basically does a test for the rename matrix not | |
1094 | * growing larger than a "rename_limit" square matrix, ie: | |
1095 | * | |
00b8cccd | 1096 | * num_destinations * num_sources > rename_limit * rename_limit |
ad8a1be5 EN |
1097 | * |
1098 | * We use st_mult() to check overflow conditions; in the | |
1099 | * exceptional circumstance that size_t isn't large enough to hold | |
1100 | * the multiplication, the system won't be able to allocate enough | |
1101 | * memory for the matrix anyway. | |
9d8a5a50 | 1102 | */ |
89973554 | 1103 | if (rename_limit <= 0) |
9dd29dbe | 1104 | return 0; /* treat as unlimited */ |
ad8a1be5 EN |
1105 | if (st_mult(num_destinations, num_sources) |
1106 | <= st_mult(rename_limit, rename_limit)) | |
9d8a5a50 JH |
1107 | return 0; |
1108 | ||
1109 | options->needed_rename_limit = | |
00b8cccd | 1110 | num_sources > num_destinations ? num_sources : num_destinations; |
f31027c9 JH |
1111 | |
1112 | /* Are we running under -C -C? */ | |
0d1e0e78 | 1113 | if (!options->flags.find_copies_harder) |
f31027c9 JH |
1114 | return 1; |
1115 | ||
1116 | /* Would we bust the limit if we were running under -C? */ | |
00b8cccd | 1117 | for (limited_sources = i = 0; i < num_sources; i++) { |
f31027c9 JH |
1118 | if (diff_unmodified_pair(rename_src[i].p)) |
1119 | continue; | |
00b8cccd | 1120 | limited_sources++; |
f31027c9 | 1121 | } |
ad8a1be5 EN |
1122 | if (st_mult(num_destinations, limited_sources) |
1123 | <= st_mult(rename_limit, rename_limit)) | |
f31027c9 | 1124 | return 2; |
9d8a5a50 JH |
1125 | return 1; |
1126 | } | |
1127 | ||
1ad69eb0 EN |
1128 | static int find_renames(struct diff_score *mx, |
1129 | int dst_cnt, | |
1130 | int minimum_score, | |
1131 | int copies, | |
1132 | struct dir_rename_info *info, | |
a49b55d5 | 1133 | struct strintmap *dirs_removed) |
0940e5f2 LT |
1134 | { |
1135 | int count = 0, i; | |
1136 | ||
1137 | for (i = 0; i < dst_cnt * NUM_CANDIDATE_PER_DST; i++) { | |
1138 | struct diff_rename_dst *dst; | |
1139 | ||
1140 | if ((mx[i].dst < 0) || | |
1141 | (mx[i].score < minimum_score)) | |
1142 | break; /* there is no more usable pair. */ | |
1143 | dst = &rename_dst[mx[i].dst]; | |
9db2ac56 | 1144 | if (dst->is_rename) |
0940e5f2 | 1145 | continue; /* already done, either exact or fuzzy. */ |
e88d6bc6 | 1146 | if (!copies && rename_src[mx[i].src].p->one->rename_used) |
0940e5f2 LT |
1147 | continue; |
1148 | record_rename_pair(mx[i].dst, mx[i].src, mx[i].score); | |
1149 | count++; | |
1ad69eb0 EN |
1150 | update_dir_rename_counts(info, dirs_removed, |
1151 | rename_src[mx[i].src].p->one->path, | |
1152 | rename_dst[mx[i].dst].p->two->path); | |
0940e5f2 LT |
1153 | } |
1154 | return count; | |
1155 | } | |
1156 | ||
9799889f | 1157 | static void remove_unneeded_paths_from_src(int detecting_copies, |
a49b55d5 | 1158 | struct strintmap *interesting) |
829514c5 EN |
1159 | { |
1160 | int i, new_num_src; | |
1161 | ||
9799889f | 1162 | if (detecting_copies && !interesting) |
829514c5 EN |
1163 | return; /* nothing to remove */ |
1164 | if (break_idx) | |
1165 | return; /* culling incompatible with break detection */ | |
1166 | ||
1167 | /* | |
1168 | * Note on reasons why we cull unneeded sources but not destinations: | |
1169 | * 1) Pairings are stored in rename_dst (not rename_src), which we | |
1170 | * need to keep around. So, we just can't cull rename_dst even | |
1171 | * if we wanted to. But doing so wouldn't help because... | |
1172 | * | |
1173 | * 2) There is a matrix pairwise comparison that follows the | |
1174 | * "Performing inexact rename detection" progress message. | |
1175 | * Iterating over the destinations is done in the outer loop, | |
1176 | * hence we only iterate over each of those once and we can | |
1177 | * easily skip the outer loop early if the destination isn't | |
1178 | * relevant. That's only one check per destination path to | |
1179 | * skip. | |
1180 | * | |
1181 | * By contrast, the sources are iterated in the inner loop; if | |
1182 | * we check whether a source can be skipped, then we'll be | |
1183 | * checking it N separate times, once for each destination. | |
1184 | * We don't want to have to iterate over known-not-needed | |
1185 | * sources N times each, so avoid that by removing the sources | |
1186 | * from rename_src here. | |
1187 | */ | |
1188 | for (i = 0, new_num_src = 0; i < rename_src_nr; i++) { | |
9799889f EN |
1189 | struct diff_filespec *one = rename_src[i].p->one; |
1190 | ||
829514c5 EN |
1191 | /* |
1192 | * renames are stored in rename_dst, so if a rename has | |
1193 | * already been detected using this source, we can just | |
1194 | * remove the source knowing rename_dst has its info. | |
1195 | */ | |
9799889f EN |
1196 | if (!detecting_copies && one->rename_used) |
1197 | continue; | |
1198 | ||
1199 | /* If we don't care about the source path, skip it */ | |
a49b55d5 | 1200 | if (interesting && !strintmap_contains(interesting, one->path)) |
829514c5 EN |
1201 | continue; |
1202 | ||
1203 | if (new_num_src < i) | |
1204 | memcpy(&rename_src[new_num_src], &rename_src[i], | |
1205 | sizeof(struct diff_rename_src)); | |
1206 | new_num_src++; | |
1207 | } | |
1208 | ||
1209 | rename_src_nr = new_num_src; | |
1210 | } | |
1211 | ||
ae1db7b3 | 1212 | static void handle_early_known_dir_renames(struct dir_rename_info *info, |
a49b55d5 EN |
1213 | struct strintmap *relevant_sources, |
1214 | struct strintmap *dirs_removed) | |
ae1db7b3 EN |
1215 | { |
1216 | /* | |
0491d392 EN |
1217 | * Directory renames are determined via an aggregate of all renames |
1218 | * under them and using a "majority wins" rule. The fact that | |
1219 | * "majority wins", though, means we don't need all the renames | |
1220 | * under the given directory, we only need enough to ensure we have | |
1221 | * a majority. | |
ae1db7b3 | 1222 | */ |
0491d392 | 1223 | |
9bd34213 | 1224 | int i, new_num_src; |
0491d392 EN |
1225 | struct hashmap_iter iter; |
1226 | struct strmap_entry *entry; | |
1227 | ||
1228 | if (!dirs_removed || !relevant_sources) | |
1229 | return; /* nothing to cull */ | |
1230 | if (break_idx) | |
1231 | return; /* culling incompatbile with break detection */ | |
1232 | ||
1233 | /* | |
bf238b71 EN |
1234 | * Supplement dir_rename_count with number of potential renames, |
1235 | * marking all potential rename sources as mapping to UNKNOWN_DIR. | |
0491d392 | 1236 | */ |
bf238b71 EN |
1237 | for (i = 0; i < rename_src_nr; i++) { |
1238 | char *old_dir; | |
1239 | struct diff_filespec *one = rename_src[i].p->one; | |
1240 | ||
1241 | /* | |
1242 | * sources that are part of a rename will have already been | |
1243 | * removed by a prior call to remove_unneeded_paths_from_src() | |
1244 | */ | |
1245 | assert(!one->rename_used); | |
1246 | ||
1247 | old_dir = get_dirname(one->path); | |
1248 | while (*old_dir != '\0' && | |
1249 | NOT_RELEVANT != strintmap_get(dirs_removed, old_dir)) { | |
1250 | char *freeme = old_dir; | |
1251 | ||
1252 | increment_count(info, old_dir, UNKNOWN_DIR); | |
1253 | old_dir = get_dirname(old_dir); | |
1254 | ||
1255 | /* Free resources we don't need anymore */ | |
1256 | free(freeme); | |
1257 | } | |
1258 | /* | |
1259 | * old_dir and new_dir free'd in increment_count, but | |
1260 | * get_dirname() gives us a new pointer we need to free for | |
1261 | * old_dir. Also, if the loop runs 0 times we need old_dir | |
1262 | * to be freed. | |
1263 | */ | |
1264 | free(old_dir); | |
1265 | } | |
0491d392 EN |
1266 | |
1267 | /* | |
1268 | * For any directory which we need a potential rename detected for | |
1269 | * (i.e. those marked as RELEVANT_FOR_SELF in dirs_removed), check | |
1270 | * whether we have enough renames to satisfy the "majority rules" | |
1271 | * requirement such that detecting any more renames of files under | |
1272 | * it won't change the result. For any such directory, mark that | |
1273 | * we no longer need to detect a rename for it. However, since we | |
1274 | * might need to still detect renames for an ancestor of that | |
1275 | * directory, use RELEVANT_FOR_ANCESTOR. | |
1276 | */ | |
1277 | strmap_for_each_entry(info->dir_rename_count, &iter, entry) { | |
1278 | /* entry->key is source_dir */ | |
1279 | struct strintmap *counts = entry->value; | |
1280 | ||
1281 | if (strintmap_get(dirs_removed, entry->key) == | |
1282 | RELEVANT_FOR_SELF && | |
1283 | dir_rename_already_determinable(counts)) { | |
1284 | strintmap_set(dirs_removed, entry->key, | |
1285 | RELEVANT_FOR_ANCESTOR); | |
1286 | } | |
1287 | } | |
9bd34213 EN |
1288 | |
1289 | for (i = 0, new_num_src = 0; i < rename_src_nr; i++) { | |
1290 | struct diff_filespec *one = rename_src[i].p->one; | |
1291 | int val; | |
1292 | ||
1293 | val = strintmap_get(relevant_sources, one->path); | |
1294 | ||
1295 | /* | |
1296 | * sources that were not found in relevant_sources should | |
1297 | * have already been removed by a prior call to | |
1298 | * remove_unneeded_paths_from_src() | |
1299 | */ | |
1300 | assert(val != -1); | |
1301 | ||
1302 | if (val == RELEVANT_LOCATION) { | |
1303 | int removable = 1; | |
1304 | char *dir = get_dirname(one->path); | |
1305 | while (1) { | |
1306 | char *freeme = dir; | |
1307 | int res = strintmap_get(dirs_removed, dir); | |
1308 | ||
1309 | /* Quit if not found or irrelevant */ | |
1310 | if (res == NOT_RELEVANT) | |
1311 | break; | |
1312 | /* If RELEVANT_FOR_SELF, can't remove */ | |
1313 | if (res == RELEVANT_FOR_SELF) { | |
1314 | removable = 0; | |
1315 | break; | |
1316 | } | |
1317 | /* Else continue searching upwards */ | |
1318 | assert(res == RELEVANT_FOR_ANCESTOR); | |
1319 | dir = get_dirname(dir); | |
1320 | free(freeme); | |
1321 | } | |
1322 | free(dir); | |
1323 | if (removable) { | |
1324 | strintmap_set(relevant_sources, one->path, | |
1325 | RELEVANT_NO_MORE); | |
1326 | continue; | |
1327 | } | |
1328 | } | |
1329 | ||
1330 | if (new_num_src < i) | |
1331 | memcpy(&rename_src[new_num_src], &rename_src[i], | |
1332 | sizeof(struct diff_rename_src)); | |
1333 | new_num_src++; | |
1334 | } | |
1335 | ||
1336 | rename_src_nr = new_num_src; | |
ae1db7b3 EN |
1337 | } |
1338 | ||
a8791ef6 EN |
1339 | static void free_filespec_data(struct diff_filespec *spec) |
1340 | { | |
1341 | if (!--spec->count) | |
1342 | diff_free_filespec_data(spec); | |
1343 | } | |
1344 | ||
a8791ef6 EN |
1345 | static void pool_free_filespec(struct mem_pool *pool, |
1346 | struct diff_filespec *spec) | |
1347 | { | |
1348 | if (!pool) { | |
1349 | free_filespec(spec); | |
1350 | return; | |
1351 | } | |
1352 | ||
1353 | /* | |
1354 | * Similar to free_filespec(), but only frees the data. The spec | |
1355 | * itself was allocated in the pool and should not be individually | |
1356 | * freed. | |
1357 | */ | |
1358 | free_filespec_data(spec); | |
1359 | } | |
1360 | ||
a8791ef6 EN |
1361 | void pool_diff_free_filepair(struct mem_pool *pool, |
1362 | struct diff_filepair *p) | |
1363 | { | |
1364 | if (!pool) { | |
1365 | diff_free_filepair(p); | |
1366 | return; | |
1367 | } | |
1368 | ||
1369 | /* | |
1370 | * Similar to diff_free_filepair() but only frees the data from the | |
1371 | * filespecs; not the filespecs or the filepair which were | |
1372 | * allocated from the pool. | |
1373 | */ | |
1374 | free_filespec_data(p->one); | |
1375 | free_filespec_data(p->two); | |
1376 | } | |
1377 | ||
0c4fd732 | 1378 | void diffcore_rename_extended(struct diff_options *options, |
f239fff4 | 1379 | struct mem_pool *pool, |
a49b55d5 EN |
1380 | struct strintmap *relevant_sources, |
1381 | struct strintmap *dirs_removed, | |
25e65b6d EN |
1382 | struct strmap *dir_rename_count, |
1383 | struct strmap *cached_pairs) | |
427dcb4b | 1384 | { |
8082d8d3 JH |
1385 | int detect_rename = options->detect_rename; |
1386 | int minimum_score = options->rename_score; | |
38c6f780 | 1387 | struct diff_queue_struct *q = &diff_queued_diff; |
5098bafb | 1388 | struct diff_queue_struct outq; |
427dcb4b | 1389 | struct diff_score *mx; |
f31027c9 | 1390 | int i, j, rename_count, skip_unmodified = 0; |
26a66a6b | 1391 | int num_destinations, dst_cnt; |
f15eb7c1 | 1392 | int num_sources, want_copies; |
3ac942d4 | 1393 | struct progress *progress = NULL; |
fa0e936f | 1394 | struct mem_pool local_pool; |
bde8b9f3 | 1395 | struct dir_rename_info info; |
d331dd3b EN |
1396 | struct diff_populate_filespec_options dpf_options = { |
1397 | .check_binary = 0, | |
1398 | .missing_object_cb = NULL, | |
1399 | .missing_object_data = NULL | |
1400 | }; | |
1aedd03a | 1401 | struct inexact_prefetch_options prefetch_options = { |
d331dd3b EN |
1402 | .repo = options->repo |
1403 | }; | |
427dcb4b | 1404 | |
557ac035 | 1405 | trace2_region_enter("diff", "setup", options->repo); |
bde8b9f3 | 1406 | info.setup = 0; |
0c4fd732 | 1407 | assert(!dir_rename_count || strmap_empty(dir_rename_count)); |
f15eb7c1 | 1408 | want_copies = (detect_rename == DIFF_DETECT_COPY); |
1ad69eb0 EN |
1409 | if (dirs_removed && (break_idx || want_copies)) |
1410 | BUG("dirs_removed incompatible with break/copy detection"); | |
9799889f EN |
1411 | if (break_idx && relevant_sources) |
1412 | BUG("break detection incompatible with source specification"); | |
26dee0ad | 1413 | if (!minimum_score) |
f345b0a0 | 1414 | minimum_score = DEFAULT_RENAME_SCORE; |
427dcb4b | 1415 | |
427dcb4b | 1416 | for (i = 0; i < q->nr; i++) { |
52e95789 | 1417 | struct diff_filepair *p = q->queue[i]; |
2f3f8b21 | 1418 | if (!DIFF_FILE_VALID(p->one)) { |
81e50eab | 1419 | if (!DIFF_FILE_VALID(p->two)) |
60896c7b | 1420 | continue; /* unmerged */ |
2f3f8b21 JH |
1421 | else if (options->single_follow && |
1422 | strcmp(options->single_follow, p->two->path)) | |
1423 | continue; /* not interested */ | |
0d1e0e78 | 1424 | else if (!options->flags.rename_empty && |
02491b67 | 1425 | is_empty_blob_oid(&p->two->oid)) |
90d43b07 | 1426 | continue; |
9db2ac56 | 1427 | else if (add_rename_dst(p) < 0) { |
4d6be03b JK |
1428 | warning("skipping rename detection, detected" |
1429 | " duplicate destination '%s'", | |
1430 | p->two->path); | |
1431 | goto cleanup; | |
1432 | } | |
2f3f8b21 | 1433 | } |
0d1e0e78 | 1434 | else if (!options->flags.rename_empty && |
02491b67 | 1435 | is_empty_blob_oid(&p->one->oid)) |
90d43b07 | 1436 | continue; |
d7c9bf22 | 1437 | else if (!DIFF_PAIR_UNMERGED(p) && !DIFF_FILE_VALID(p->two)) { |
64479711 LT |
1438 | /* |
1439 | * If the source is a broken "delete", and | |
2210100a JH |
1440 | * they did not really want to get broken, |
1441 | * that means the source actually stays. | |
64479711 LT |
1442 | * So we increment the "rename_used" score |
1443 | * by one, to indicate ourselves as a user | |
1444 | */ | |
1445 | if (p->broken_pair && !p->score) | |
1446 | p->one->rename_used++; | |
e88d6bc6 | 1447 | register_rename_src(p); |
64479711 | 1448 | } |
f15eb7c1 | 1449 | else if (want_copies) { |
64479711 LT |
1450 | /* |
1451 | * Increment the "rename_used" score by | |
1452 | * one, to indicate ourselves as a user. | |
2210100a | 1453 | */ |
64479711 | 1454 | p->one->rename_used++; |
e88d6bc6 | 1455 | register_rename_src(p); |
2210100a | 1456 | } |
427dcb4b | 1457 | } |
557ac035 | 1458 | trace2_region_leave("diff", "setup", options->repo); |
0024a549 | 1459 | if (rename_dst_nr == 0 || rename_src_nr == 0) |
427dcb4b JH |
1460 | goto cleanup; /* nothing to do */ |
1461 | ||
557ac035 | 1462 | trace2_region_enter("diff", "exact renames", options->repo); |
fa0e936f | 1463 | mem_pool_init(&local_pool, 32*1024); |
17559a64 LT |
1464 | /* |
1465 | * We really want to cull the candidates list early | |
1466 | * with cheap tests in order to avoid doing deltas. | |
1467 | */ | |
fa0e936f EN |
1468 | rename_count = find_exact_renames(options, &local_pool); |
1469 | /* | |
1470 | * Discard local_pool immediately instead of at "cleanup:" in order | |
1471 | * to reduce maximum memory usage; inexact rename detection uses up | |
1472 | * a fair amount of memory, and mem_pools can too. | |
1473 | */ | |
1474 | mem_pool_discard(&local_pool, 0); | |
557ac035 | 1475 | trace2_region_leave("diff", "exact renames", options->repo); |
17559a64 | 1476 | |
42899ac8 LT |
1477 | /* Did we only want exact renames? */ |
1478 | if (minimum_score == MAX_SCORE) | |
1479 | goto cleanup; | |
1480 | ||
f15eb7c1 | 1481 | num_sources = rename_src_nr; |
42899ac8 | 1482 | |
bd24aa2f EN |
1483 | if (want_copies || break_idx) { |
1484 | /* | |
1485 | * Cull sources: | |
1486 | * - remove ones corresponding to exact renames | |
9799889f | 1487 | * - remove ones not found in relevant_sources |
bd24aa2f EN |
1488 | */ |
1489 | trace2_region_enter("diff", "cull after exact", options->repo); | |
9799889f | 1490 | remove_unneeded_paths_from_src(want_copies, relevant_sources); |
bd24aa2f EN |
1491 | trace2_region_leave("diff", "cull after exact", options->repo); |
1492 | } else { | |
1493 | /* Determine minimum score to match basenames */ | |
1494 | double factor = 0.5; | |
1495 | char *basename_factor = getenv("GIT_BASENAME_FACTOR"); | |
1496 | int min_basename_score; | |
1497 | ||
1498 | if (basename_factor) | |
1499 | factor = strtol(basename_factor, NULL, 10)/100.0; | |
1500 | assert(factor >= 0.0 && factor <= 1.0); | |
1501 | min_basename_score = minimum_score + | |
1502 | (int)(factor * (MAX_SCORE - minimum_score)); | |
1503 | ||
1504 | /* | |
1505 | * Cull sources: | |
1506 | * - remove ones involved in renames (found via exact match) | |
1507 | */ | |
1508 | trace2_region_enter("diff", "cull after exact", options->repo); | |
9799889f | 1509 | remove_unneeded_paths_from_src(want_copies, NULL); |
bd24aa2f EN |
1510 | trace2_region_leave("diff", "cull after exact", options->repo); |
1511 | ||
ae8cf74d EN |
1512 | /* Preparation for basename-driven matching. */ |
1513 | trace2_region_enter("diff", "dir rename setup", options->repo); | |
e4fd06e7 | 1514 | initialize_dir_rename_info(&info, relevant_sources, |
25e65b6d EN |
1515 | dirs_removed, dir_rename_count, |
1516 | cached_pairs); | |
ae8cf74d EN |
1517 | trace2_region_leave("diff", "dir rename setup", options->repo); |
1518 | ||
bd24aa2f EN |
1519 | /* Utilize file basenames to quickly find renames. */ |
1520 | trace2_region_enter("diff", "basename matches", options->repo); | |
1521 | rename_count += find_basename_matches(options, | |
bde8b9f3 | 1522 | min_basename_score, |
e4fd06e7 EN |
1523 | &info, |
1524 | relevant_sources, | |
1525 | dirs_removed); | |
bd24aa2f EN |
1526 | trace2_region_leave("diff", "basename matches", options->repo); |
1527 | ||
1528 | /* | |
1529 | * Cull sources, again: | |
1530 | * - remove ones involved in renames (found via basenames) | |
9799889f | 1531 | * - remove ones not found in relevant_sources |
ae1db7b3 EN |
1532 | * and |
1533 | * - remove ones in relevant_sources which are needed only | |
1534 | * for directory renames IF no ancestory directory | |
1535 | * actually needs to know any more individual path | |
1536 | * renames under them | |
bd24aa2f EN |
1537 | */ |
1538 | trace2_region_enter("diff", "cull basename", options->repo); | |
9799889f | 1539 | remove_unneeded_paths_from_src(want_copies, relevant_sources); |
ae1db7b3 EN |
1540 | handle_early_known_dir_renames(&info, relevant_sources, |
1541 | dirs_removed); | |
bd24aa2f EN |
1542 | trace2_region_leave("diff", "cull basename", options->repo); |
1543 | } | |
1544 | ||
1545 | /* Calculate how many rename destinations are left */ | |
1546 | num_destinations = (rename_dst_nr - rename_count); | |
1547 | num_sources = rename_src_nr; /* rename_src_nr reflects lower number */ | |
1548 | ||
42899ac8 | 1549 | /* All done? */ |
f15eb7c1 | 1550 | if (!num_destinations || !num_sources) |
42899ac8 LT |
1551 | goto cleanup; |
1552 | ||
f15eb7c1 | 1553 | switch (too_many_rename_candidates(num_destinations, num_sources, |
00b8cccd | 1554 | options)) { |
f31027c9 | 1555 | case 1: |
0024a549 | 1556 | goto cleanup; |
f31027c9 JH |
1557 | case 2: |
1558 | options->degraded_cc_to_c = 1; | |
1559 | skip_unmodified = 1; | |
1560 | break; | |
1561 | default: | |
1562 | break; | |
1563 | } | |
0024a549 | 1564 | |
557ac035 | 1565 | trace2_region_enter("diff", "inexact renames", options->repo); |
3ac942d4 | 1566 | if (options->show_rename_progress) { |
8aade107 | 1567 | progress = start_delayed_progress( |
754dbc43 | 1568 | _("Performing inexact rename detection"), |
f15eb7c1 | 1569 | (uint64_t)num_destinations * (uint64_t)num_sources); |
3ac942d4 JK |
1570 | } |
1571 | ||
d331dd3b EN |
1572 | /* Finish setting up dpf_options */ |
1573 | prefetch_options.skip_unmodified = skip_unmodified; | |
a5183d76 | 1574 | if (options->repo == the_repository && repo_has_promisor_remote(the_repository)) { |
1aedd03a | 1575 | dpf_options.missing_object_cb = inexact_prefetch; |
d331dd3b EN |
1576 | dpf_options.missing_object_data = &prefetch_options; |
1577 | } | |
1578 | ||
ca56dadb | 1579 | CALLOC_ARRAY(mx, st_mult(NUM_CANDIDATE_PER_DST, num_destinations)); |
25d5ea41 | 1580 | for (dst_cnt = i = 0; i < rename_dst_nr; i++) { |
9db2ac56 | 1581 | struct diff_filespec *two = rename_dst[i].p->two; |
6d24ad97 JH |
1582 | struct diff_score *m; |
1583 | ||
9db2ac56 | 1584 | if (rename_dst[i].is_rename) |
bd24aa2f | 1585 | continue; /* exact or basename match already handled */ |
6d24ad97 JH |
1586 | |
1587 | m = &mx[dst_cnt * NUM_CANDIDATE_PER_DST]; | |
1588 | for (j = 0; j < NUM_CANDIDATE_PER_DST; j++) | |
1589 | m[j].dst = -1; | |
1590 | ||
25d5ea41 | 1591 | for (j = 0; j < rename_src_nr; j++) { |
e88d6bc6 | 1592 | struct diff_filespec *one = rename_src[j].p->one; |
6d24ad97 | 1593 | struct diff_score this_src; |
f31027c9 | 1594 | |
829514c5 | 1595 | assert(!one->rename_used || want_copies || break_idx); |
f15eb7c1 | 1596 | |
f31027c9 JH |
1597 | if (skip_unmodified && |
1598 | diff_unmodified_pair(rename_src[j].p)) | |
1599 | continue; | |
1600 | ||
b78ea5fc NTND |
1601 | this_src.score = estimate_similarity(options->repo, |
1602 | one, two, | |
95acf11a | 1603 | minimum_score, |
d331dd3b | 1604 | &dpf_options); |
6d24ad97 JH |
1605 | this_src.name_score = basename_same(one, two); |
1606 | this_src.dst = i; | |
1607 | this_src.src = j; | |
1608 | record_if_better(m, &this_src); | |
809809bb JH |
1609 | /* |
1610 | * Once we run estimate_similarity, | |
1611 | * We do not need the text anymore. | |
1612 | */ | |
8ae92e63 | 1613 | diff_free_filespec_blob(one); |
809809bb | 1614 | diff_free_filespec_blob(two); |
427dcb4b JH |
1615 | } |
1616 | dst_cnt++; | |
81c4bf02 | 1617 | display_progress(progress, |
f15eb7c1 | 1618 | (uint64_t)dst_cnt * (uint64_t)num_sources); |
427dcb4b | 1619 | } |
3ac942d4 | 1620 | stop_progress(&progress); |
6d24ad97 | 1621 | |
427dcb4b | 1622 | /* cost matrix sorted by most to least similar pair */ |
2049b8dc | 1623 | STABLE_QSORT(mx, dst_cnt * NUM_CANDIDATE_PER_DST, score_compare); |
6d24ad97 | 1624 | |
1ad69eb0 EN |
1625 | rename_count += find_renames(mx, dst_cnt, minimum_score, 0, |
1626 | &info, dirs_removed); | |
f15eb7c1 | 1627 | if (want_copies) |
1ad69eb0 EN |
1628 | rename_count += find_renames(mx, dst_cnt, minimum_score, 1, |
1629 | &info, dirs_removed); | |
427dcb4b | 1630 | free(mx); |
557ac035 | 1631 | trace2_region_leave("diff", "inexact renames", options->repo); |
427dcb4b | 1632 | |
15d061b4 | 1633 | cleanup: |
427dcb4b | 1634 | /* At this point, we have found some renames and copies and they |
5098bafb | 1635 | * are recorded in rename_dst. The original list is still in *q. |
427dcb4b | 1636 | */ |
557ac035 | 1637 | trace2_region_enter("diff", "write back to queue", options->repo); |
9ca5df90 | 1638 | DIFF_QUEUE_CLEAR(&outq); |
427dcb4b | 1639 | for (i = 0; i < q->nr; i++) { |
25d5ea41 | 1640 | struct diff_filepair *p = q->queue[i]; |
25d5ea41 JH |
1641 | struct diff_filepair *pair_to_free = NULL; |
1642 | ||
d7c9bf22 MZ |
1643 | if (DIFF_PAIR_UNMERGED(p)) { |
1644 | diff_q(&outq, p); | |
1645 | } | |
1646 | else if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) { | |
9db2ac56 EN |
1647 | /* Creation */ |
1648 | diff_q(&outq, p); | |
427dcb4b | 1649 | } |
2cd68882 JH |
1650 | else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) { |
1651 | /* | |
1652 | * Deletion | |
1653 | * | |
f345b0a0 JH |
1654 | * We would output this delete record if: |
1655 | * | |
1656 | * (1) this is a broken delete and the counterpart | |
1657 | * broken create remains in the output; or | |
5098bafb JH |
1658 | * (2) this is not a broken delete, and rename_dst |
1659 | * does not have a rename/copy to move p->one->path | |
1660 | * out of existence. | |
f345b0a0 JH |
1661 | * |
1662 | * Otherwise, the counterpart broken create | |
1663 | * has been turned into a rename-edit; or | |
1664 | * delete did not have a matching create to | |
1665 | * begin with. | |
2cd68882 | 1666 | */ |
f345b0a0 JH |
1667 | if (DIFF_PAIR_BROKEN(p)) { |
1668 | /* broken delete */ | |
9db2ac56 EN |
1669 | struct diff_rename_dst *dst = locate_rename_dst(p); |
1670 | if (!dst) | |
1671 | BUG("tracking failed somehow; failed to find associated dst for broken pair"); | |
1672 | if (dst->is_rename) | |
f345b0a0 JH |
1673 | /* counterpart is now rename/copy */ |
1674 | pair_to_free = p; | |
1675 | } | |
1676 | else { | |
64479711 | 1677 | if (p->one->rename_used) |
f345b0a0 JH |
1678 | /* this path remains */ |
1679 | pair_to_free = p; | |
1680 | } | |
2cd68882 | 1681 | |
9db2ac56 | 1682 | if (!pair_to_free) |
2cd68882 JH |
1683 | diff_q(&outq, p); |
1684 | } | |
25d5ea41 | 1685 | else if (!diff_unmodified_pair(p)) |
15d061b4 | 1686 | /* all the usual ones need to be kept */ |
25d5ea41 | 1687 | diff_q(&outq, p); |
15d061b4 | 1688 | else |
356da0f9 | 1689 | /* no need to keep unmodified pairs */ |
15d061b4 JH |
1690 | pair_to_free = p; |
1691 | ||
226406f6 | 1692 | if (pair_to_free) |
f239fff4 | 1693 | pool_diff_free_filepair(pool, pair_to_free); |
427dcb4b | 1694 | } |
25d5ea41 | 1695 | diff_debug_queue("done copying original", &outq); |
427dcb4b | 1696 | |
25d5ea41 JH |
1697 | free(q->queue); |
1698 | *q = outq; | |
1699 | diff_debug_queue("done collapsing", q); | |
427dcb4b | 1700 | |
9fb88419 | 1701 | for (i = 0; i < rename_dst_nr; i++) |
9db2ac56 | 1702 | if (rename_dst[i].filespec_to_free) |
f239fff4 | 1703 | pool_free_filespec(pool, rename_dst[i].filespec_to_free); |
5098bafb | 1704 | |
b1473019 | 1705 | cleanup_dir_rename_info(&info, dirs_removed, dir_rename_count != NULL); |
6a83d902 | 1706 | FREE_AND_NULL(rename_dst); |
25d5ea41 | 1707 | rename_dst_nr = rename_dst_alloc = 0; |
6a83d902 | 1708 | FREE_AND_NULL(rename_src); |
25d5ea41 | 1709 | rename_src_nr = rename_src_alloc = 0; |
9db2ac56 EN |
1710 | if (break_idx) { |
1711 | strintmap_clear(break_idx); | |
1712 | FREE_AND_NULL(break_idx); | |
1713 | } | |
557ac035 | 1714 | trace2_region_leave("diff", "write back to queue", options->repo); |
427dcb4b JH |
1715 | return; |
1716 | } | |
0c4fd732 EN |
1717 | |
1718 | void diffcore_rename(struct diff_options *options) | |
1719 | { | |
f239fff4 | 1720 | diffcore_rename_extended(options, NULL, NULL, NULL, NULL, NULL); |
0c4fd732 | 1721 | } |