git.oblomov.eu Git - git/blob - 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 "object-store.h"
   8 #include "hashmap.h"
   9 #include "progress.h"
  10
  11 /* Table of rename/copy destinations */
  12
  13 static struct diff_rename_dst {
  14         struct diff_filespec *two;
  15         struct diff_filepair *pair;
  16 } *rename_dst;
  17 static int rename_dst_nr, rename_dst_alloc;
  18
  19 static int find_rename_dst(struct diff_filespec *two)
  20 {
  21         int first, last;
  22
  23         first = 0;
  24         last = rename_dst_nr;
  25         while (last > first) {
  26                 int next = first + ((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 next;
  31                 if (cmp < 0) {
  32                         last = next;
  33                         continue;
  34                 }
  35                 first = next+1;
  36         }
  37         return -first - 1;
  38 }
  39
  40 static struct diff_rename_dst *locate_rename_dst(struct diff_filespec *two)
  41 {
  42         int ofs = find_rename_dst(two);
  43         return ofs < 0 ? NULL : &rename_dst[ofs];
  44 }
  45
  46 /*
  47  * Returns 0 on success, -1 if we found a duplicate.
  48  */
  49 static int add_rename_dst(struct diff_filespec *two)
  50 {
  51         int first = find_rename_dst(two);
  52
  53         if (first >= 0)
  54                 return -1;
  55         first = -first - 1;
  56
  57         /* insert to make it at "first" */
  58         ALLOC_GROW(rename_dst, rename_dst_nr + 1, rename_dst_alloc);
  59         rename_dst_nr++;
  60         if (first < rename_dst_nr)
  61                 MOVE_ARRAY(rename_dst + first + 1, rename_dst + first,
  62                            rename_dst_nr - first - 1);
  63         rename_dst[first].two = alloc_filespec(two->path);
  64         fill_filespec(rename_dst[first].two, &two->oid, two->oid_valid,
  65                       two->mode);
  66         rename_dst[first].pair = NULL;
  67         return 0;
  68 }
  69
  70 /* Table of rename/copy src files */
  71 static struct diff_rename_src {
  72         struct diff_filepair *p;
  73         unsigned short score; /* to remember the break score */
  74 } *rename_src;
  75 static int rename_src_nr, rename_src_alloc;
  76
  77 static struct diff_rename_src *register_rename_src(struct diff_filepair *p)
  78 {
  79         int first, last;
  80         struct diff_filespec *one = p->one;
  81         unsigned short score = p->score;
  82
  83         first = 0;
  84         last = rename_src_nr;
  85         while (last > first) {
  86                 int next = first + ((last - first) >> 1);
  87                 struct diff_rename_src *src = &(rename_src[next]);
  88                 int cmp = strcmp(one->path, src->p->one->path);
  89                 if (!cmp)
  90                         return src;
  91                 if (cmp < 0) {
  92                         last = next;
  93                         continue;
  94                 }
  95                 first = next+1;
  96         }
  97
  98         /* insert to make it at "first" */
  99         ALLOC_GROW(rename_src, rename_src_nr + 1, rename_src_alloc);
 100         rename_src_nr++;
 101         if (first < rename_src_nr)
 102                 MOVE_ARRAY(rename_src + first + 1, rename_src + first,
 103                            rename_src_nr - first - 1);
 104         rename_src[first].p = p;
 105         rename_src[first].score = score;
 106         return &(rename_src[first]);
 107 }
 108
 109 static int basename_same(struct diff_filespec *src, struct diff_filespec *dst)
 110 {
 111         int src_len = strlen(src->path), dst_len = strlen(dst->path);
 112         while (src_len && dst_len) {
 113                 char c1 = src->path[--src_len];
 114                 char c2 = dst->path[--dst_len];
 115                 if (c1 != c2)
 116                         return 0;
 117                 if (c1 == '/')
 118                         return 1;
 119         }
 120         return (!src_len || src->path[src_len - 1] == '/') &&
 121                 (!dst_len || dst->path[dst_len - 1] == '/');
 122 }
 123
 124 struct diff_score {
 125         int src; /* index in rename_src */
 126         int dst; /* index in rename_dst */
 127         unsigned short score;
 128         short name_score;
 129 };
 130
 131 static int estimate_similarity(struct repository *r,
 132                                struct diff_filespec *src,
 133                                struct diff_filespec *dst,
 134                                int minimum_score)
 135 {
 136         /* src points at a file that existed in the original tree (or
 137          * optionally a file in the destination tree) and dst points
 138          * at a newly created file.  They may be quite similar, in which
 139          * case we want to say src is renamed to dst or src is copied into
 140          * dst, and then some edit has been applied to dst.
 141          *
 142          * Compare them and return how similar they are, representing
 143          * the score as an integer between 0 and MAX_SCORE.
 144          *
 145          * When there is an exact match, it is considered a better
 146          * match than anything else; the destination does not even
 147          * call into this function in that case.
 148          */
 149         unsigned long max_size, delta_size, base_size, src_copied, literal_added;
 150         int score;
 151         struct diff_populate_filespec_options dpf_options = {
 152                 .check_size_only = 1
 153         };
 154
 155         /* We deal only with regular files.  Symlink renames are handled
 156          * only when they are exact matches --- in other words, no edits
 157          * after renaming.
 158          */
 159         if (!S_ISREG(src->mode) || !S_ISREG(dst->mode))
 160                 return 0;
 161
 162         /*
 163          * Need to check that source and destination sizes are
 164          * filled in before comparing them.
 165          *
 166          * If we already have "cnt_data" filled in, we know it's
 167          * all good (avoid checking the size for zero, as that
 168          * is a possible size - we really should have a flag to
 169          * say whether the size is valid or not!)
 170          */
 171         if (!src->cnt_data &&
 172             diff_populate_filespec(r, src, &dpf_options))
 173                 return 0;
 174         if (!dst->cnt_data &&
 175             diff_populate_filespec(r, dst, &dpf_options))
 176                 return 0;
 177
 178         max_size = ((src->size > dst->size) ? src->size : dst->size);
 179         base_size = ((src->size < dst->size) ? src->size : dst->size);
 180         delta_size = max_size - base_size;
 181
 182         /* We would not consider edits that change the file size so
 183          * drastically.  delta_size must be smaller than
 184          * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size).
 185          *
 186          * Note that base_size == 0 case is handled here already
 187          * and the final score computation below would not have a
 188          * divide-by-zero issue.
 189          */
 190         if (max_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE)
 191                 return 0;
 192
 193         if (!src->cnt_data && diff_populate_filespec(r, src, NULL))
 194                 return 0;
 195         if (!dst->cnt_data && diff_populate_filespec(r, dst, NULL))
 196                 return 0;
 197
 198         if (diffcore_count_changes(r, src, dst,
 199                                    &src->cnt_data, &dst->cnt_data,
 200                                    &src_copied, &literal_added))
 201                 return 0;
 202
 203         /* How similar are they?
 204          * what percentage of material in dst are from source?
 205          */
 206         if (!dst->size)
 207                 score = 0; /* should not happen */
 208         else
 209                 score = (int)(src_copied * MAX_SCORE / max_size);
 210         return score;
 211 }
 212
 213 static void record_rename_pair(int dst_index, int src_index, int score)
 214 {
 215         struct diff_filespec *src, *dst;
 216         struct diff_filepair *dp;
 217
 218         if (rename_dst[dst_index].pair)
 219                 die("internal error: dst already matched.");
 220
 221         src = rename_src[src_index].p->one;
 222         src->rename_used++;
 223         src->count++;
 224
 225         dst = rename_dst[dst_index].two;
 226         dst->count++;
 227
 228         dp = diff_queue(NULL, src, dst);
 229         dp->renamed_pair = 1;
 230         if (!strcmp(src->path, dst->path))
 231                 dp->score = rename_src[src_index].score;
 232         else
 233                 dp->score = score;
 234         rename_dst[dst_index].pair = dp;
 235 }
 236
 237 /*
 238  * We sort the rename similarity matrix with the score, in descending
 239  * order (the most similar first).
 240  */
 241 static int score_compare(const void *a_, const void *b_)
 242 {
 243         const struct diff_score *a = a_, *b = b_;
 244
 245         /* sink the unused ones to the bottom */
 246         if (a->dst < 0)
 247                 return (0 <= b->dst);
 248         else if (b->dst < 0)
 249                 return -1;
 250
 251         if (a->score == b->score)
 252                 return b->name_score - a->name_score;
 253
 254         return b->score - a->score;
 255 }
 256
 257 struct file_similarity {
 258         struct hashmap_entry entry;
 259         int index;
 260         struct diff_filespec *filespec;
 261 };
 262
 263 static unsigned int hash_filespec(struct repository *r,
 264                                   struct diff_filespec *filespec)
 265 {
 266         if (!filespec->oid_valid) {
 267                 if (diff_populate_filespec(r, filespec, NULL))
 268                         return 0;
 269                 hash_object_file(r->hash_algo, filespec->data, filespec->size,
 270                                  "blob", &filespec->oid);
 271         }
 272         return oidhash(&filespec->oid);
 273 }
 274
 275 static int find_identical_files(struct hashmap *srcs,
 276                                 int dst_index,
 277                                 struct diff_options *options)
 278 {
 279         int renames = 0;
 280         struct diff_filespec *target = rename_dst[dst_index].two;
 281         struct file_similarity *p, *best = NULL;
 282         int i = 100, best_score = -1;
 283         unsigned int hash = hash_filespec(options->repo, target);
 284
 285         /*
 286          * Find the best source match for specified destination.
 287          */
 288         p = hashmap_get_entry_from_hash(srcs, hash, NULL,
 289                                         struct file_similarity, entry);
 290         hashmap_for_each_entry_from(srcs, p, entry) {
 291                 int score;
 292                 struct diff_filespec *source = p->filespec;
 293
 294                 /* False hash collision? */
 295                 if (!oideq(&source->oid, &target->oid))
 296                         continue;
 297                 /* Non-regular files? If so, the modes must match! */
 298                 if (!S_ISREG(source->mode) || !S_ISREG(target->mode)) {
 299                         if (source->mode != target->mode)
 300                                 continue;
 301                 }
 302                 /* Give higher scores to sources that haven't been used already */
 303                 score = !source->rename_used;
 304                 if (source->rename_used && options->detect_rename != DIFF_DETECT_COPY)
 305                         continue;
 306                 score += basename_same(source, target);
 307                 if (score > best_score) {
 308                         best = p;
 309                         best_score = score;
 310                         if (score == 2)
 311                                 break;
 312                 }
 313
 314                 /* Too many identical alternatives? Pick one */
 315                 if (!--i)
 316                         break;
 317         }
 318         if (best) {
 319                 record_rename_pair(dst_index, best->index, MAX_SCORE);
 320                 renames++;
 321         }
 322         return renames;
 323 }
 324
 325 static void insert_file_table(struct repository *r,
 326                               struct hashmap *table, int index,
 327                               struct diff_filespec *filespec)
 328 {
 329         struct file_similarity *entry = xmalloc(sizeof(*entry));
 330
 331         entry->index = index;
 332         entry->filespec = filespec;
 333
 334         hashmap_entry_init(&entry->entry, hash_filespec(r, filespec));
 335         hashmap_add(table, &entry->entry);
 336 }
 337
 338 /*
 339  * Find exact renames first.
 340  *
 341  * The first round matches up the up-to-date entries,
 342  * and then during the second round we try to match
 343  * cache-dirty entries as well.
 344  */
 345 static int find_exact_renames(struct diff_options *options)
 346 {
 347         int i, renames = 0;
 348         struct hashmap file_table;
 349
 350         /* Add all sources to the hash table in reverse order, because
 351          * later on they will be retrieved in LIFO order.
 352          */
 353         hashmap_init(&file_table, NULL, NULL, rename_src_nr);
 354         for (i = rename_src_nr-1; i >= 0; i--)
 355                 insert_file_table(options->repo,
 356                                   &file_table, i,
 357                                   rename_src[i].p->one);
 358
 359         /* Walk the destinations and find best source match */
 360         for (i = 0; i < rename_dst_nr; i++)
 361                 renames += find_identical_files(&file_table, i, options);
 362
 363         /* Free the hash data structure and entries */
 364         hashmap_free_entries(&file_table, struct file_similarity, entry);
 365
 366         return renames;
 367 }
 368
 369 #define NUM_CANDIDATE_PER_DST 4
 370 static void record_if_better(struct diff_score m[], struct diff_score *o)
 371 {
 372         int i, worst;
 373
 374         /* find the worst one */
 375         worst = 0;
 376         for (i = 1; i < NUM_CANDIDATE_PER_DST; i++)
 377                 if (score_compare(&m[i], &m[worst]) > 0)
 378                         worst = i;
 379
 380         /* is it better than the worst one? */
 381         if (score_compare(&m[worst], o) > 0)
 382                 m[worst] = *o;
 383 }
 384
 385 /*
 386  * Returns:
 387  * 0 if we are under the limit;
 388  * 1 if we need to disable inexact rename detection;
 389  * 2 if we would be under the limit if we were given -C instead of -C -C.
 390  */
 391 static int too_many_rename_candidates(int num_create,
 392                                       struct diff_options *options)
 393 {
 394         int rename_limit = options->rename_limit;
 395         int num_src = rename_src_nr;
 396         int i;
 397
 398         options->needed_rename_limit = 0;
 399
 400         /*
 401          * This basically does a test for the rename matrix not
 402          * growing larger than a "rename_limit" square matrix, ie:
 403          *
 404          *    num_create * num_src > rename_limit * rename_limit
 405          */
 406         if (rename_limit <= 0)
 407                 rename_limit = 32767;
 408         if ((num_create <= rename_limit || num_src <= rename_limit) &&
 409             ((uint64_t)num_create * (uint64_t)num_src
 410              <= (uint64_t)rename_limit * (uint64_t)rename_limit))
 411                 return 0;
 412
 413         options->needed_rename_limit =
 414                 num_src > num_create ? num_src : num_create;
 415
 416         /* Are we running under -C -C? */
 417         if (!options->flags.find_copies_harder)
 418                 return 1;
 419
 420         /* Would we bust the limit if we were running under -C? */
 421         for (num_src = i = 0; i < rename_src_nr; i++) {
 422                 if (diff_unmodified_pair(rename_src[i].p))
 423                         continue;
 424                 num_src++;
 425         }
 426         if ((num_create <= rename_limit || num_src <= rename_limit) &&
 427             ((uint64_t)num_create * (uint64_t)num_src
 428              <= (uint64_t)rename_limit * (uint64_t)rename_limit))
 429                 return 2;
 430         return 1;
 431 }
 432
 433 static int find_renames(struct diff_score *mx, int dst_cnt, int minimum_score, int copies)
 434 {
 435         int count = 0, i;
 436
 437         for (i = 0; i < dst_cnt * NUM_CANDIDATE_PER_DST; i++) {
 438                 struct diff_rename_dst *dst;
 439
 440                 if ((mx[i].dst < 0) ||
 441                     (mx[i].score < minimum_score))
 442                         break; /* there is no more usable pair. */
 443                 dst = &rename_dst[mx[i].dst];
 444                 if (dst->pair)
 445                         continue; /* already done, either exact or fuzzy. */
 446                 if (!copies && rename_src[mx[i].src].p->one->rename_used)
 447                         continue;
 448                 record_rename_pair(mx[i].dst, mx[i].src, mx[i].score);
 449                 count++;
 450         }
 451         return count;
 452 }
 453
 454 void diffcore_rename(struct diff_options *options)
 455 {
 456         int detect_rename = options->detect_rename;
 457         int minimum_score = options->rename_score;
 458         struct diff_queue_struct *q = &diff_queued_diff;
 459         struct diff_queue_struct outq;
 460         struct diff_score *mx;
 461         int i, j, rename_count, skip_unmodified = 0;
 462         int num_create, dst_cnt;
 463         struct progress *progress = NULL;
 464
 465         if (!minimum_score)
 466                 minimum_score = DEFAULT_RENAME_SCORE;
 467
 468         for (i = 0; i < q->nr; i++) {
 469                 struct diff_filepair *p = q->queue[i];
 470                 if (!DIFF_FILE_VALID(p->one)) {
 471                         if (!DIFF_FILE_VALID(p->two))
 472                                 continue; /* unmerged */
 473                         else if (options->single_follow &&
 474                                  strcmp(options->single_follow, p->two->path))
 475                                 continue; /* not interested */
 476                         else if (!options->flags.rename_empty &&
 477                                  is_empty_blob_oid(&p->two->oid))
 478                                 continue;
 479                         else if (add_rename_dst(p->two) < 0) {
 480                                 warning("skipping rename detection, detected"
 481                                         " duplicate destination '%s'",
 482                                         p->two->path);
 483                                 goto cleanup;
 484                         }
 485                 }
 486                 else if (!options->flags.rename_empty &&
 487                          is_empty_blob_oid(&p->one->oid))
 488                         continue;
 489                 else if (!DIFF_PAIR_UNMERGED(p) && !DIFF_FILE_VALID(p->two)) {
 490                         /*
 491                          * If the source is a broken "delete", and
 492                          * they did not really want to get broken,
 493                          * that means the source actually stays.
 494                          * So we increment the "rename_used" score
 495                          * by one, to indicate ourselves as a user
 496                          */
 497                         if (p->broken_pair && !p->score)
 498                                 p->one->rename_used++;
 499                         register_rename_src(p);
 500                 }
 501                 else if (detect_rename == DIFF_DETECT_COPY) {
 502                         /*
 503                          * Increment the "rename_used" score by
 504                          * one, to indicate ourselves as a user.
 505                          */
 506                         p->one->rename_used++;
 507                         register_rename_src(p);
 508                 }
 509         }
 510         if (rename_dst_nr == 0 || rename_src_nr == 0)
 511                 goto cleanup; /* nothing to do */
 512
 513         /*
 514          * We really want to cull the candidates list early
 515          * with cheap tests in order to avoid doing deltas.
 516          */
 517         rename_count = find_exact_renames(options);
 518
 519         /* Did we only want exact renames? */
 520         if (minimum_score == MAX_SCORE)
 521                 goto cleanup;
 522
 523         /*
 524          * Calculate how many renames are left (but all the source
 525          * files still remain as options for rename/copies!)
 526          */
 527         num_create = (rename_dst_nr - rename_count);
 528
 529         /* All done? */
 530         if (!num_create)
 531                 goto cleanup;
 532
 533         switch (too_many_rename_candidates(num_create, options)) {
 534         case 1:
 535                 goto cleanup;
 536         case 2:
 537                 options->degraded_cc_to_c = 1;
 538                 skip_unmodified = 1;
 539                 break;
 540         default:
 541                 break;
 542         }
 543
 544         if (options->show_rename_progress) {
 545                 progress = start_delayed_progress(
 546                                 _("Performing inexact rename detection"),
 547                                 (uint64_t)rename_dst_nr * (uint64_t)rename_src_nr);
 548         }
 549
 550         mx = xcalloc(st_mult(NUM_CANDIDATE_PER_DST, num_create), sizeof(*mx));
 551         for (dst_cnt = i = 0; i < rename_dst_nr; i++) {
 552                 struct diff_filespec *two = rename_dst[i].two;
 553                 struct diff_score *m;
 554
 555                 if (rename_dst[i].pair)
 556                         continue; /* dealt with exact match already. */
 557
 558                 m = &mx[dst_cnt * NUM_CANDIDATE_PER_DST];
 559                 for (j = 0; j < NUM_CANDIDATE_PER_DST; j++)
 560                         m[j].dst = -1;
 561
 562                 for (j = 0; j < rename_src_nr; j++) {
 563                         struct diff_filespec *one = rename_src[j].p->one;
 564                         struct diff_score this_src;
 565
 566                         if (skip_unmodified &&
 567                             diff_unmodified_pair(rename_src[j].p))
 568                                 continue;
 569
 570                         this_src.score = estimate_similarity(options->repo,
 571                                                              one, two,
 572                                                              minimum_score);
 573                         this_src.name_score = basename_same(one, two);
 574                         this_src.dst = i;
 575                         this_src.src = j;
 576                         record_if_better(m, &this_src);
 577                         /*
 578                          * Once we run estimate_similarity,
 579                          * We do not need the text anymore.
 580                          */
 581                         diff_free_filespec_blob(one);
 582                         diff_free_filespec_blob(two);
 583                 }
 584                 dst_cnt++;
 585                 display_progress(progress, (uint64_t)(i+1)*(uint64_t)rename_src_nr);
 586         }
 587         stop_progress(&progress);
 588
 589         /* cost matrix sorted by most to least similar pair */
 590         STABLE_QSORT(mx, dst_cnt * NUM_CANDIDATE_PER_DST, score_compare);
 591
 592         rename_count += find_renames(mx, dst_cnt, minimum_score, 0);
 593         if (detect_rename == DIFF_DETECT_COPY)
 594                 rename_count += find_renames(mx, dst_cnt, minimum_score, 1);
 595         free(mx);
 596
 597  cleanup:
 598         /* At this point, we have found some renames and copies and they
 599          * are recorded in rename_dst.  The original list is still in *q.
 600          */
 601         DIFF_QUEUE_CLEAR(&outq);
 602         for (i = 0; i < q->nr; i++) {
 603                 struct diff_filepair *p = q->queue[i];
 604                 struct diff_filepair *pair_to_free = NULL;
 605
 606                 if (DIFF_PAIR_UNMERGED(p)) {
 607                         diff_q(&outq, p);
 608                 }
 609                 else if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
 610                         /*
 611                          * Creation
 612                          *
 613                          * We would output this create record if it has
 614                          * not been turned into a rename/copy already.
 615                          */
 616                         struct diff_rename_dst *dst = locate_rename_dst(p->two);
 617                         if (dst && dst->pair) {
 618                                 diff_q(&outq, dst->pair);
 619                                 pair_to_free = p;
 620                         }
 621                         else
 622                                 /* no matching rename/copy source, so
 623                                  * record this as a creation.
 624                                  */
 625                                 diff_q(&outq, p);
 626                 }
 627                 else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) {
 628                         /*
 629                          * Deletion
 630                          *
 631                          * We would output this delete record if:
 632                          *
 633                          * (1) this is a broken delete and the counterpart
 634                          *     broken create remains in the output; or
 635                          * (2) this is not a broken delete, and rename_dst
 636                          *     does not have a rename/copy to move p->one->path
 637                          *     out of existence.
 638                          *
 639                          * Otherwise, the counterpart broken create
 640                          * has been turned into a rename-edit; or
 641                          * delete did not have a matching create to
 642                          * begin with.
 643                          */
 644                         if (DIFF_PAIR_BROKEN(p)) {
 645                                 /* broken delete */
 646                                 struct diff_rename_dst *dst = locate_rename_dst(p->one);
 647                                 if (dst && dst->pair)
 648                                         /* counterpart is now rename/copy */
 649                                         pair_to_free = p;
 650                         }
 651                         else {
 652                                 if (p->one->rename_used)
 653                                         /* this path remains */
 654                                         pair_to_free = p;
 655                         }
 656
 657                         if (pair_to_free)
 658                                 ;
 659                         else
 660                                 diff_q(&outq, p);
 661                 }
 662                 else if (!diff_unmodified_pair(p))
 663                         /* all the usual ones need to be kept */
 664                         diff_q(&outq, p);
 665                 else
 666                         /* no need to keep unmodified pairs */
 667                         pair_to_free = p;
 668
 669                 if (pair_to_free)
 670                         diff_free_filepair(pair_to_free);
 671         }
 672         diff_debug_queue("done copying original", &outq);
 673
 674         free(q->queue);
 675         *q = outq;
 676         diff_debug_queue("done collapsing", q);
 677
 678         for (i = 0; i < rename_dst_nr; i++)
 679                 free_filespec(rename_dst[i].two);
 680
 681         FREE_AND_NULL(rename_dst);
 682         rename_dst_nr = rename_dst_alloc = 0;
 683         FREE_AND_NULL(rename_src);
 684         rename_src_nr = rename_src_alloc = 0;
 685         return;
 686 }