Add new git-rm command with documentation
[git.git] / diffcore-rename.c
1 /*
2  * Copyright (C) 2005 Junio C Hamano
3  */
4 #include "cache.h"
5 #include "diff.h"
6 #include "diffcore.h"
7 #include "delta.h"
8 #include "count-delta.h"
9
10 /* Table of rename/copy destinations */
11
12 static struct diff_rename_dst {
13         struct diff_filespec *two;
14         struct diff_filepair *pair;
15 } *rename_dst;
16 static int rename_dst_nr, rename_dst_alloc;
17
18 static struct diff_rename_dst *locate_rename_dst(struct diff_filespec *two,
19                                                  int insert_ok)
20 {
21         int first, last;
22
23         first = 0;
24         last = rename_dst_nr;
25         while (last > first) {
26                 int next = (last + first) >> 1;
27                 struct diff_rename_dst *dst = &(rename_dst[next]);
28                 int cmp = strcmp(two->path, dst->two->path);
29                 if (!cmp)
30                         return dst;
31                 if (cmp < 0) {
32                         last = next;
33                         continue;
34                 }
35                 first = next+1;
36         }
37         /* not found */
38         if (!insert_ok)
39                 return NULL;
40         /* insert to make it at "first" */
41         if (rename_dst_alloc <= rename_dst_nr) {
42                 rename_dst_alloc = alloc_nr(rename_dst_alloc);
43                 rename_dst = xrealloc(rename_dst,
44                                       rename_dst_alloc * sizeof(*rename_dst));
45         }
46         rename_dst_nr++;
47         if (first < rename_dst_nr)
48                 memmove(rename_dst + first + 1, rename_dst + first,
49                         (rename_dst_nr - first - 1) * sizeof(*rename_dst));
50         rename_dst[first].two = alloc_filespec(two->path);
51         fill_filespec(rename_dst[first].two, two->sha1, two->mode);
52         rename_dst[first].pair = NULL;
53         return &(rename_dst[first]);
54 }
55
56 /* Table of rename/copy src files */
57 static struct diff_rename_src {
58         struct diff_filespec *one;
59         unsigned src_path_left : 1;
60 } *rename_src;
61 static int rename_src_nr, rename_src_alloc;
62
63 static struct diff_rename_src *register_rename_src(struct diff_filespec *one,
64                                                    int src_path_left)
65 {
66         int first, last;
67
68         first = 0;
69         last = rename_src_nr;
70         while (last > first) {
71                 int next = (last + first) >> 1;
72                 struct diff_rename_src *src = &(rename_src[next]);
73                 int cmp = strcmp(one->path, src->one->path);
74                 if (!cmp)
75                         return src;
76                 if (cmp < 0) {
77                         last = next;
78                         continue;
79                 }
80                 first = next+1;
81         }
82
83         /* insert to make it at "first" */
84         if (rename_src_alloc <= rename_src_nr) {
85                 rename_src_alloc = alloc_nr(rename_src_alloc);
86                 rename_src = xrealloc(rename_src,
87                                       rename_src_alloc * sizeof(*rename_src));
88         }
89         rename_src_nr++;
90         if (first < rename_src_nr)
91                 memmove(rename_src + first + 1, rename_src + first,
92                         (rename_src_nr - first - 1) * sizeof(*rename_src));
93         rename_src[first].one = one;
94         rename_src[first].src_path_left = src_path_left;
95         return &(rename_src[first]);
96 }
97
98 static int is_exact_match(struct diff_filespec *src, struct diff_filespec *dst)
99 {
100         if (src->sha1_valid && dst->sha1_valid &&
101             !memcmp(src->sha1, dst->sha1, 20))
102                 return 1;
103         if (diff_populate_filespec(src, 1) || diff_populate_filespec(dst, 1))
104                 return 0;
105         if (src->size != dst->size)
106                 return 0;
107         if (diff_populate_filespec(src, 0) || diff_populate_filespec(dst, 0))
108                 return 0;
109         if (src->size == dst->size &&
110             !memcmp(src->data, dst->data, src->size))
111                 return 1;
112         return 0;
113 }
114
115 struct diff_score {
116         int src; /* index in rename_src */
117         int dst; /* index in rename_dst */
118         int score;
119 };
120
121 static int estimate_similarity(struct diff_filespec *src,
122                                struct diff_filespec *dst,
123                                int minimum_score)
124 {
125         /* src points at a file that existed in the original tree (or
126          * optionally a file in the destination tree) and dst points
127          * at a newly created file.  They may be quite similar, in which
128          * case we want to say src is renamed to dst or src is copied into
129          * dst, and then some edit has been applied to dst.
130          *
131          * Compare them and return how similar they are, representing
132          * the score as an integer between 0 and MAX_SCORE.
133          *
134          * When there is an exact match, it is considered a better
135          * match than anything else; the destination does not even
136          * call into this function in that case.
137          */
138         void *delta;
139         unsigned long delta_size, base_size, src_copied, literal_added;
140         unsigned long delta_limit;
141         int score;
142
143         /* We deal only with regular files.  Symlink renames are handled
144          * only when they are exact matches --- in other words, no edits
145          * after renaming.
146          */
147         if (!S_ISREG(src->mode) || !S_ISREG(dst->mode))
148                 return 0;
149
150         delta_size = ((src->size < dst->size) ?
151                       (dst->size - src->size) : (src->size - dst->size));
152         base_size = ((src->size < dst->size) ? src->size : dst->size);
153
154         /* We would not consider edits that change the file size so
155          * drastically.  delta_size must be smaller than
156          * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size).
157          *
158          * Note that base_size == 0 case is handled here already
159          * and the final score computation below would not have a
160          * divide-by-zero issue.
161          */
162         if (base_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE)
163                 return 0;
164
165         if (diff_populate_filespec(src, 0) || diff_populate_filespec(dst, 0))
166                 return 0; /* error but caught downstream */
167
168         delta_limit = base_size * (MAX_SCORE-minimum_score) / MAX_SCORE;
169         delta = diff_delta(src->data, src->size,
170                            dst->data, dst->size,
171                            &delta_size, delta_limit);
172         if (!delta)
173                 /* If delta_limit is exceeded, we have too much differences */
174                 return 0;
175
176         /* A delta that has a lot of literal additions would have
177          * big delta_size no matter what else it does.
178          */
179         if (base_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE)
180                 return 0;
181
182         /* Estimate the edit size by interpreting delta. */
183         if (count_delta(delta, delta_size, &src_copied, &literal_added)) {
184                 free(delta);
185                 return 0;
186         }
187         free(delta);
188
189         /* Extent of damage */
190         if (src->size + literal_added < src_copied)
191                 delta_size = 0;
192         else
193                 delta_size = (src->size - src_copied) + literal_added;
194
195         /*
196          * Now we will give some score to it.  100% edit gets 0 points
197          * and 0% edit gets MAX_SCORE points.
198          */
199         score = MAX_SCORE - (MAX_SCORE * delta_size / base_size); 
200         if (score < 0) return 0;
201         if (MAX_SCORE < score) return MAX_SCORE;
202         return score;
203 }
204
205 static void record_rename_pair(int dst_index, int src_index, int score)
206 {
207         struct diff_filespec *one, *two, *src, *dst;
208         struct diff_filepair *dp;
209
210         if (rename_dst[dst_index].pair)
211                 die("internal error: dst already matched.");
212
213         src = rename_src[src_index].one;
214         one = alloc_filespec(src->path);
215         fill_filespec(one, src->sha1, src->mode);
216
217         dst = rename_dst[dst_index].two;
218         two = alloc_filespec(dst->path);
219         fill_filespec(two, dst->sha1, dst->mode);
220
221         dp = diff_queue(NULL, one, two);
222         dp->score = score;
223         dp->source_stays = rename_src[src_index].src_path_left;
224         rename_dst[dst_index].pair = dp;
225 }
226
227 /*
228  * We sort the rename similarity matrix with the score, in descending
229  * order (the most similar first).
230  */
231 static int score_compare(const void *a_, const void *b_)
232 {
233         const struct diff_score *a = a_, *b = b_;
234         return b->score - a->score;
235 }
236
237 static int compute_stays(struct diff_queue_struct *q,
238                          struct diff_filespec *one)
239 {
240         int i;
241         for (i = 0; i < q->nr; i++) {
242                 struct diff_filepair *p = q->queue[i];
243                 if (strcmp(one->path, p->two->path))
244                         continue;
245                 if (DIFF_PAIR_RENAME(p)) {
246                         return 0; /* something else is renamed into this */
247                 }
248         }
249         return 1;
250 }
251
252 void diffcore_rename(struct diff_options *options)
253 {
254         int detect_rename = options->detect_rename;
255         int minimum_score = options->rename_score;
256         int rename_limit = options->rename_limit;
257         struct diff_queue_struct *q = &diff_queued_diff;
258         struct diff_queue_struct outq;
259         struct diff_score *mx;
260         int i, j, rename_count;
261         int num_create, num_src, dst_cnt;
262
263         if (!minimum_score)
264                 minimum_score = DEFAULT_RENAME_SCORE;
265         rename_count = 0;
266
267         for (i = 0; i < q->nr; i++) {
268                 struct diff_filepair *p = q->queue[i];
269                 if (!DIFF_FILE_VALID(p->one))
270                         if (!DIFF_FILE_VALID(p->two))
271                                 continue; /* unmerged */
272                         else
273                                 locate_rename_dst(p->two, 1);
274                 else if (!DIFF_FILE_VALID(p->two)) {
275                         /* If the source is a broken "delete", and
276                          * they did not really want to get broken,
277                          * that means the source actually stays.
278                          */
279                         int stays = (p->broken_pair && !p->score);
280                         register_rename_src(p->one, stays);
281                 }
282                 else if (detect_rename == DIFF_DETECT_COPY)
283                         register_rename_src(p->one, 1);
284         }
285         if (rename_dst_nr == 0 || rename_src_nr == 0 ||
286             (0 < rename_limit && rename_limit < rename_dst_nr))
287                 goto cleanup; /* nothing to do */
288
289         /* We really want to cull the candidates list early
290          * with cheap tests in order to avoid doing deltas.
291          */
292         for (i = 0; i < rename_dst_nr; i++) {
293                 struct diff_filespec *two = rename_dst[i].two;
294                 for (j = 0; j < rename_src_nr; j++) {
295                         struct diff_filespec *one = rename_src[j].one;
296                         if (!is_exact_match(one, two))
297                                 continue;
298                         record_rename_pair(i, j, MAX_SCORE);
299                         rename_count++;
300                         break; /* we are done with this entry */
301                 }
302         }
303
304         /* Have we run out the created file pool?  If so we can avoid
305          * doing the delta matrix altogether.
306          */
307         if (rename_count == rename_dst_nr)
308                 goto cleanup;
309
310         if (minimum_score == MAX_SCORE)
311                 goto cleanup;
312
313         num_create = (rename_dst_nr - rename_count);
314         num_src = rename_src_nr;
315         mx = xmalloc(sizeof(*mx) * num_create * num_src);
316         for (dst_cnt = i = 0; i < rename_dst_nr; i++) {
317                 int base = dst_cnt * num_src;
318                 struct diff_filespec *two = rename_dst[i].two;
319                 if (rename_dst[i].pair)
320                         continue; /* dealt with exact match already. */
321                 for (j = 0; j < rename_src_nr; j++) {
322                         struct diff_filespec *one = rename_src[j].one;
323                         struct diff_score *m = &mx[base+j];
324                         m->src = j;
325                         m->dst = i;
326                         m->score = estimate_similarity(one, two,
327                                                        minimum_score);
328                 }
329                 dst_cnt++;
330         }
331         /* cost matrix sorted by most to least similar pair */
332         qsort(mx, num_create * num_src, sizeof(*mx), score_compare);
333         for (i = 0; i < num_create * num_src; i++) {
334                 struct diff_rename_dst *dst = &rename_dst[mx[i].dst];
335                 if (dst->pair)
336                         continue; /* already done, either exact or fuzzy. */
337                 if (mx[i].score < minimum_score)
338                         break; /* there is no more usable pair. */
339                 record_rename_pair(mx[i].dst, mx[i].src, mx[i].score);
340                 rename_count++;
341         }
342         free(mx);
343
344  cleanup:
345         /* At this point, we have found some renames and copies and they
346          * are recorded in rename_dst.  The original list is still in *q.
347          */
348         outq.queue = NULL;
349         outq.nr = outq.alloc = 0;
350         for (i = 0; i < q->nr; i++) {
351                 struct diff_filepair *p = q->queue[i];
352                 struct diff_filepair *pair_to_free = NULL;
353
354                 if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
355                         /*
356                          * Creation
357                          *
358                          * We would output this create record if it has
359                          * not been turned into a rename/copy already.
360                          */
361                         struct diff_rename_dst *dst =
362                                 locate_rename_dst(p->two, 0);
363                         if (dst && dst->pair) {
364                                 diff_q(&outq, dst->pair);
365                                 pair_to_free = p;
366                         }
367                         else
368                                 /* no matching rename/copy source, so
369                                  * record this as a creation.
370                                  */
371                                 diff_q(&outq, p);
372                 }
373                 else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) {
374                         /*
375                          * Deletion
376                          *
377                          * We would output this delete record if:
378                          *
379                          * (1) this is a broken delete and the counterpart
380                          *     broken create remains in the output; or
381                          * (2) this is not a broken delete, and rename_dst
382                          *     does not have a rename/copy to move p->one->path
383                          *     out of existence.
384                          *
385                          * Otherwise, the counterpart broken create
386                          * has been turned into a rename-edit; or
387                          * delete did not have a matching create to
388                          * begin with.
389                          */
390                         if (DIFF_PAIR_BROKEN(p)) {
391                                 /* broken delete */
392                                 struct diff_rename_dst *dst =
393                                         locate_rename_dst(p->one, 0);
394                                 if (dst && dst->pair)
395                                         /* counterpart is now rename/copy */
396                                         pair_to_free = p;
397                         }
398                         else {
399                                 for (j = 0; j < rename_dst_nr; j++) {
400                                         if (!rename_dst[j].pair)
401                                                 continue;
402                                         if (strcmp(rename_dst[j].pair->
403                                                    one->path,
404                                                    p->one->path))
405                                                 continue;
406                                         break;
407                                 }
408                                 if (j < rename_dst_nr)
409                                         /* this path remains */
410                                         pair_to_free = p;
411                         }
412
413                         if (pair_to_free)
414                                 ;
415                         else
416                                 diff_q(&outq, p);
417                 }
418                 else if (!diff_unmodified_pair(p))
419                         /* all the usual ones need to be kept */
420                         diff_q(&outq, p);
421                 else
422                         /* no need to keep unmodified pairs */
423                         pair_to_free = p;
424
425                 if (pair_to_free)
426                         diff_free_filepair(pair_to_free);
427         }
428         diff_debug_queue("done copying original", &outq);
429
430         free(q->queue);
431         *q = outq;
432         diff_debug_queue("done collapsing", q);
433
434         /* We need to see which rename source really stays here;
435          * earlier we only checked if the path is left in the result,
436          * but even if a path remains in the result, if that is coming
437          * from copying something else on top of it, then the original
438          * source is lost and does not stay.
439          */
440         for (i = 0; i < q->nr; i++) {
441                 struct diff_filepair *p = q->queue[i];
442                 if (DIFF_PAIR_RENAME(p) && p->source_stays) {
443                         /* If one appears as the target of a rename-copy,
444                          * then mark p->source_stays = 0; otherwise
445                          * leave it as is.
446                          */
447                         p->source_stays = compute_stays(q, p->one);
448                 }
449         }
450
451         for (i = 0; i < rename_dst_nr; i++) {
452                 diff_free_filespec_data(rename_dst[i].two);
453                 free(rename_dst[i].two);
454         }
455
456         free(rename_dst);
457         rename_dst = NULL;
458         rename_dst_nr = rename_dst_alloc = 0;
459         free(rename_src);
460         rename_src = NULL;
461         rename_src_nr = rename_src_alloc = 0;
462         return;
463 }