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
   8 /* Table of rename/copy destinations */
   9
  10 static struct diff_rename_dst {
  11         struct diff_filespec *two;
  12         struct diff_filepair *pair;
  13 } *rename_dst;
  14 static int rename_dst_nr, rename_dst_alloc;
  15
  16 static struct diff_rename_dst *locate_rename_dst(struct diff_filespec *two,
  17                                                  int insert_ok)
  18 {
  19         int first, last;
  20
  21         first = 0;
  22         last = rename_dst_nr;
  23         while (last > first) {
  24                 int next = (last + first) >> 1;
  25                 struct diff_rename_dst *dst = &(rename_dst[next]);
  26                 int cmp = strcmp(two->path, dst->two->path);
  27                 if (!cmp)
  28                         return dst;
  29                 if (cmp < 0) {
  30                         last = next;
  31                         continue;
  32                 }
  33                 first = next+1;
  34         }
  35         /* not found */
  36         if (!insert_ok)
  37                 return NULL;
  38         /* insert to make it at "first" */
  39         if (rename_dst_alloc <= rename_dst_nr) {
  40                 rename_dst_alloc = alloc_nr(rename_dst_alloc);
  41                 rename_dst = xrealloc(rename_dst,
  42                                       rename_dst_alloc * sizeof(*rename_dst));
  43         }
  44         rename_dst_nr++;
  45         if (first < rename_dst_nr)
  46                 memmove(rename_dst + first + 1, rename_dst + first,
  47                         (rename_dst_nr - first - 1) * sizeof(*rename_dst));
  48         rename_dst[first].two = alloc_filespec(two->path);
  49         fill_filespec(rename_dst[first].two, two->sha1, two->mode);
  50         rename_dst[first].pair = NULL;
  51         return &(rename_dst[first]);
  52 }
  53
  54 /* Table of rename/copy src files */
  55 static struct diff_rename_src {
  56         struct diff_filespec *one;
  57         unsigned short score; /* to remember the break score */
  58         unsigned src_path_left : 1;
  59 } *rename_src;
  60 static int rename_src_nr, rename_src_alloc;
  61
  62 static struct diff_rename_src *register_rename_src(struct diff_filespec *one,
  63                                                    int src_path_left,
  64                                                    unsigned short score)
  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].score = score;
  95         rename_src[first].src_path_left = src_path_left;
  96         return &(rename_src[first]);
  97 }
  98
  99 static int is_exact_match(struct diff_filespec *src,
 100                           struct diff_filespec *dst,
 101                           int contents_too)
 102 {
 103         if (src->sha1_valid && dst->sha1_valid &&
 104             !hashcmp(src->sha1, dst->sha1))
 105                 return 1;
 106         if (!contents_too)
 107                 return 0;
 108         if (diff_populate_filespec(src, 1) || diff_populate_filespec(dst, 1))
 109                 return 0;
 110         if (src->size != dst->size)
 111                 return 0;
 112         if (src->sha1_valid && dst->sha1_valid)
 113             return !hashcmp(src->sha1, dst->sha1);
 114         if (diff_populate_filespec(src, 0) || diff_populate_filespec(dst, 0))
 115                 return 0;
 116         if (src->size == dst->size &&
 117             !memcmp(src->data, dst->data, src->size))
 118                 return 1;
 119         return 0;
 120 }
 121
 122 struct diff_score {
 123         int src; /* index in rename_src */
 124         int dst; /* index in rename_dst */
 125         int score;
 126 };
 127
 128 static int estimate_similarity(struct diff_filespec *src,
 129                                struct diff_filespec *dst,
 130                                int minimum_score)
 131 {
 132         /* src points at a file that existed in the original tree (or
 133          * optionally a file in the destination tree) and dst points
 134          * at a newly created file.  They may be quite similar, in which
 135          * case we want to say src is renamed to dst or src is copied into
 136          * dst, and then some edit has been applied to dst.
 137          *
 138          * Compare them and return how similar they are, representing
 139          * the score as an integer between 0 and MAX_SCORE.
 140          *
 141          * When there is an exact match, it is considered a better
 142          * match than anything else; the destination does not even
 143          * call into this function in that case.
 144          */
 145         unsigned long max_size, delta_size, base_size, src_copied, literal_added;
 146         unsigned long delta_limit;
 147         int score;
 148
 149         /* We deal only with regular files.  Symlink renames are handled
 150          * only when they are exact matches --- in other words, no edits
 151          * after renaming.
 152          */
 153         if (!S_ISREG(src->mode) || !S_ISREG(dst->mode))
 154                 return 0;
 155
 156         max_size = ((src->size > dst->size) ? src->size : dst->size);
 157         base_size = ((src->size < dst->size) ? src->size : dst->size);
 158         delta_size = max_size - base_size;
 159
 160         /* We would not consider edits that change the file size so
 161          * drastically.  delta_size must be smaller than
 162          * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size).
 163          *
 164          * Note that base_size == 0 case is handled here already
 165          * and the final score computation below would not have a
 166          * divide-by-zero issue.
 167          */
 168         if (base_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE)
 169                 return 0;
 170
 171         if (diff_populate_filespec(src, 0) || diff_populate_filespec(dst, 0))
 172                 return 0; /* error but caught downstream */
 173
 174
 175         delta_limit = (unsigned long)
 176                 (base_size * (MAX_SCORE-minimum_score) / MAX_SCORE);
 177         if (diffcore_count_changes(src->data, src->size,
 178                                    dst->data, dst->size,
 179                                    &src->cnt_data, &dst->cnt_data,
 180                                    delta_limit,
 181                                    &src_copied, &literal_added))
 182                 return 0;
 183
 184         /* How similar are they?
 185          * what percentage of material in dst are from source?
 186          */
 187         if (!dst->size)
 188                 score = 0; /* should not happen */
 189         else
 190                 score = (int)(src_copied * MAX_SCORE / max_size);
 191         return score;
 192 }
 193
 194 static void record_rename_pair(int dst_index, int src_index, int score)
 195 {
 196         struct diff_filespec *one, *two, *src, *dst;
 197         struct diff_filepair *dp;
 198
 199         if (rename_dst[dst_index].pair)
 200                 die("internal error: dst already matched.");
 201
 202         src = rename_src[src_index].one;
 203         one = alloc_filespec(src->path);
 204         fill_filespec(one, src->sha1, src->mode);
 205
 206         dst = rename_dst[dst_index].two;
 207         two = alloc_filespec(dst->path);
 208         fill_filespec(two, dst->sha1, dst->mode);
 209
 210         dp = diff_queue(NULL, one, two);
 211         dp->renamed_pair = 1;
 212         if (!strcmp(src->path, dst->path))
 213                 dp->score = rename_src[src_index].score;
 214         else
 215                 dp->score = score;
 216         dp->source_stays = rename_src[src_index].src_path_left;
 217         rename_dst[dst_index].pair = dp;
 218 }
 219
 220 /*
 221  * We sort the rename similarity matrix with the score, in descending
 222  * order (the most similar first).
 223  */
 224 static int score_compare(const void *a_, const void *b_)
 225 {
 226         const struct diff_score *a = a_, *b = b_;
 227         return b->score - a->score;
 228 }
 229
 230 static int compute_stays(struct diff_queue_struct *q,
 231                          struct diff_filespec *one)
 232 {
 233         int i;
 234         for (i = 0; i < q->nr; i++) {
 235                 struct diff_filepair *p = q->queue[i];
 236                 if (strcmp(one->path, p->two->path))
 237                         continue;
 238                 if (DIFF_PAIR_RENAME(p)) {
 239                         return 0; /* something else is renamed into this */
 240                 }
 241         }
 242         return 1;
 243 }
 244
 245 void diffcore_rename(struct diff_options *options)
 246 {
 247         int detect_rename = options->detect_rename;
 248         int minimum_score = options->rename_score;
 249         int rename_limit = options->rename_limit;
 250         struct diff_queue_struct *q = &diff_queued_diff;
 251         struct diff_queue_struct outq;
 252         struct diff_score *mx;
 253         int i, j, rename_count, contents_too;
 254         int num_create, num_src, dst_cnt;
 255
 256         if (!minimum_score)
 257                 minimum_score = DEFAULT_RENAME_SCORE;
 258         rename_count = 0;
 259
 260         for (i = 0; i < q->nr; i++) {
 261                 struct diff_filepair *p = q->queue[i];
 262                 if (!DIFF_FILE_VALID(p->one)) {
 263                         if (!DIFF_FILE_VALID(p->two))
 264                                 continue; /* unmerged */
 265                         else if (options->single_follow &&
 266                                  strcmp(options->single_follow, p->two->path))
 267                                 continue; /* not interested */
 268                         else
 269                                 locate_rename_dst(p->two, 1);
 270                 }
 271                 else if (!DIFF_FILE_VALID(p->two)) {
 272                         /* If the source is a broken "delete", and
 273                          * they did not really want to get broken,
 274                          * that means the source actually stays.
 275                          */
 276                         int stays = (p->broken_pair && !p->score);
 277                         register_rename_src(p->one, stays, p->score);
 278                 }
 279                 else if (detect_rename == DIFF_DETECT_COPY)
 280                         register_rename_src(p->one, 1, p->score);
 281         }
 282         if (rename_dst_nr == 0 || rename_src_nr == 0 ||
 283             (0 < rename_limit && rename_limit < rename_dst_nr))
 284                 goto cleanup; /* nothing to do */
 285
 286         /* We really want to cull the candidates list early
 287          * with cheap tests in order to avoid doing deltas.
 288          * The first round matches up the up-to-date entries,
 289          * and then during the second round we try to match
 290          * cache-dirty entries as well.
 291          */
 292         for (contents_too = 0; contents_too < 2; contents_too++) {
 293                 for (i = 0; i < rename_dst_nr; i++) {
 294                         struct diff_filespec *two = rename_dst[i].two;
 295                         if (rename_dst[i].pair)
 296                                 continue; /* dealt with an earlier round */
 297                         for (j = 0; j < rename_src_nr; j++) {
 298                                 struct diff_filespec *one = rename_src[j].one;
 299                                 if (!is_exact_match(one, two, contents_too))
 300                                         continue;
 301                                 record_rename_pair(i, j, (int)MAX_SCORE);
 302                                 rename_count++;
 303                                 break; /* we are done with this entry */
 304                         }
 305                 }
 306         }
 307
 308         /* Have we run out the created file pool?  If so we can avoid
 309          * doing the delta matrix altogether.
 310          */
 311         if (rename_count == rename_dst_nr)
 312                 goto cleanup;
 313
 314         if (minimum_score == MAX_SCORE)
 315                 goto cleanup;
 316
 317         num_create = (rename_dst_nr - rename_count);
 318         num_src = rename_src_nr;
 319         mx = xmalloc(sizeof(*mx) * num_create * num_src);
 320         for (dst_cnt = i = 0; i < rename_dst_nr; i++) {
 321                 int base = dst_cnt * num_src;
 322                 struct diff_filespec *two = rename_dst[i].two;
 323                 if (rename_dst[i].pair)
 324                         continue; /* dealt with exact match already. */
 325                 for (j = 0; j < rename_src_nr; j++) {
 326                         struct diff_filespec *one = rename_src[j].one;
 327                         struct diff_score *m = &mx[base+j];
 328                         m->src = j;
 329                         m->dst = i;
 330                         m->score = estimate_similarity(one, two,
 331                                                        minimum_score);
 332                         diff_free_filespec_data(one);
 333                 }
 334                 /* We do not need the text anymore */
 335                 diff_free_filespec_data(two);
 336                 dst_cnt++;
 337         }
 338         /* cost matrix sorted by most to least similar pair */
 339         qsort(mx, num_create * num_src, sizeof(*mx), score_compare);
 340         for (i = 0; i < num_create * num_src; i++) {
 341                 struct diff_rename_dst *dst = &rename_dst[mx[i].dst];
 342                 if (dst->pair)
 343                         continue; /* already done, either exact or fuzzy. */
 344                 if (mx[i].score < minimum_score)
 345                         break; /* there is no more usable pair. */
 346                 record_rename_pair(mx[i].dst, mx[i].src, mx[i].score);
 347                 rename_count++;
 348         }
 349         free(mx);
 350
 351  cleanup:
 352         /* At this point, we have found some renames and copies and they
 353          * are recorded in rename_dst.  The original list is still in *q.
 354          */
 355         outq.queue = NULL;
 356         outq.nr = outq.alloc = 0;
 357         for (i = 0; i < q->nr; i++) {
 358                 struct diff_filepair *p = q->queue[i];
 359                 struct diff_filepair *pair_to_free = NULL;
 360
 361                 if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
 362                         /*
 363                          * Creation
 364                          *
 365                          * We would output this create record if it has
 366                          * not been turned into a rename/copy already.
 367                          */
 368                         struct diff_rename_dst *dst =
 369                                 locate_rename_dst(p->two, 0);
 370                         if (dst && dst->pair) {
 371                                 diff_q(&outq, dst->pair);
 372                                 pair_to_free = p;
 373                         }
 374                         else
 375                                 /* no matching rename/copy source, so
 376                                  * record this as a creation.
 377                                  */
 378                                 diff_q(&outq, p);
 379                 }
 380                 else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) {
 381                         /*
 382                          * Deletion
 383                          *
 384                          * We would output this delete record if:
 385                          *
 386                          * (1) this is a broken delete and the counterpart
 387                          *     broken create remains in the output; or
 388                          * (2) this is not a broken delete, and rename_dst
 389                          *     does not have a rename/copy to move p->one->path
 390                          *     out of existence.
 391                          *
 392                          * Otherwise, the counterpart broken create
 393                          * has been turned into a rename-edit; or
 394                          * delete did not have a matching create to
 395                          * begin with.
 396                          */
 397                         if (DIFF_PAIR_BROKEN(p)) {
 398                                 /* broken delete */
 399                                 struct diff_rename_dst *dst =
 400                                         locate_rename_dst(p->one, 0);
 401                                 if (dst && dst->pair)
 402                                         /* counterpart is now rename/copy */
 403                                         pair_to_free = p;
 404                         }
 405                         else {
 406                                 for (j = 0; j < rename_dst_nr; j++) {
 407                                         if (!rename_dst[j].pair)
 408                                                 continue;
 409                                         if (strcmp(rename_dst[j].pair->
 410                                                    one->path,
 411                                                    p->one->path))
 412                                                 continue;
 413                                         break;
 414                                 }
 415                                 if (j < rename_dst_nr)
 416                                         /* this path remains */
 417                                         pair_to_free = p;
 418                         }
 419
 420                         if (pair_to_free)
 421                                 ;
 422                         else
 423                                 diff_q(&outq, p);
 424                 }
 425                 else if (!diff_unmodified_pair(p))
 426                         /* all the usual ones need to be kept */
 427                         diff_q(&outq, p);
 428                 else
 429                         /* no need to keep unmodified pairs */
 430                         pair_to_free = p;
 431
 432                 if (pair_to_free)
 433                         diff_free_filepair(pair_to_free);
 434         }
 435         diff_debug_queue("done copying original", &outq);
 436
 437         free(q->queue);
 438         *q = outq;
 439         diff_debug_queue("done collapsing", q);
 440
 441         /* We need to see which rename source really stays here;
 442          * earlier we only checked if the path is left in the result,
 443          * but even if a path remains in the result, if that is coming
 444          * from copying something else on top of it, then the original
 445          * source is lost and does not stay.
 446          */
 447         for (i = 0; i < q->nr; i++) {
 448                 struct diff_filepair *p = q->queue[i];
 449                 if (DIFF_PAIR_RENAME(p) && p->source_stays) {
 450                         /* If one appears as the target of a rename-copy,
 451                          * then mark p->source_stays = 0; otherwise
 452                          * leave it as is.
 453                          */
 454                         p->source_stays = compute_stays(q, p->one);
 455                 }
 456         }
 457
 458         for (i = 0; i < rename_dst_nr; i++) {
 459                 diff_free_filespec_data(rename_dst[i].two);
 460                 free(rename_dst[i].two);
 461         }
 462
 463         free(rename_dst);
 464         rename_dst = NULL;
 465         rename_dst_nr = rename_dst_alloc = 0;
 466         free(rename_src);
 467         rename_src = NULL;
 468         rename_src_nr = rename_src_alloc = 0;
 469         return;
 470 }