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 static int basename_same(struct diff_filespec *src, struct diff_filespec *dst)
 123 {
 124         int src_len = strlen(src->path), dst_len = strlen(dst->path);
 125         while (src_len && dst_len) {
 126                 char c1 = src->path[--src_len];
 127                 char c2 = dst->path[--dst_len];
 128                 if (c1 != c2)
 129                         return 0;
 130                 if (c1 == '/')
 131                         return 1;
 132         }
 133         return (!src_len || src->path[src_len - 1] == '/') &&
 134                 (!dst_len || dst->path[dst_len - 1] == '/');
 135 }
 136
 137 struct diff_score {
 138         int src; /* index in rename_src */
 139         int dst; /* index in rename_dst */
 140         int score;
 141 };
 142
 143 static int estimate_similarity(struct diff_filespec *src,
 144                                struct diff_filespec *dst,
 145                                int minimum_score)
 146 {
 147         /* src points at a file that existed in the original tree (or
 148          * optionally a file in the destination tree) and dst points
 149          * at a newly created file.  They may be quite similar, in which
 150          * case we want to say src is renamed to dst or src is copied into
 151          * dst, and then some edit has been applied to dst.
 152          *
 153          * Compare them and return how similar they are, representing
 154          * the score as an integer between 0 and MAX_SCORE.
 155          *
 156          * When there is an exact match, it is considered a better
 157          * match than anything else; the destination does not even
 158          * call into this function in that case.
 159          */
 160         unsigned long max_size, delta_size, base_size, src_copied, literal_added;
 161         unsigned long delta_limit;
 162         int score;
 163
 164         /* We deal only with regular files.  Symlink renames are handled
 165          * only when they are exact matches --- in other words, no edits
 166          * after renaming.
 167          */
 168         if (!S_ISREG(src->mode) || !S_ISREG(dst->mode))
 169                 return 0;
 170
 171         max_size = ((src->size > dst->size) ? src->size : dst->size);
 172         base_size = ((src->size < dst->size) ? src->size : dst->size);
 173         delta_size = max_size - base_size;
 174
 175         /* We would not consider edits that change the file size so
 176          * drastically.  delta_size must be smaller than
 177          * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size).
 178          *
 179          * Note that base_size == 0 case is handled here already
 180          * and the final score computation below would not have a
 181          * divide-by-zero issue.
 182          */
 183         if (base_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE)
 184                 return 0;
 185
 186         if (diff_populate_filespec(src, 0) || diff_populate_filespec(dst, 0))
 187                 return 0; /* error but caught downstream */
 188
 189
 190         delta_limit = (unsigned long)
 191                 (base_size * (MAX_SCORE-minimum_score) / MAX_SCORE);
 192         if (diffcore_count_changes(src->data, src->size,
 193                                    dst->data, dst->size,
 194                                    &src->cnt_data, &dst->cnt_data,
 195                                    delta_limit,
 196                                    &src_copied, &literal_added))
 197                 return 0;
 198
 199         /* How similar are they?
 200          * what percentage of material in dst are from source?
 201          */
 202         if (!dst->size)
 203                 score = 0; /* should not happen */
 204         else {
 205                 score = (int)(src_copied * MAX_SCORE / max_size);
 206                 if (basename_same(src, dst))
 207                         score++;
 208         }
 209         return score;
 210 }
 211
 212 static void record_rename_pair(int dst_index, int src_index, int score)
 213 {
 214         struct diff_filespec *one, *two, *src, *dst;
 215         struct diff_filepair *dp;
 216
 217         if (rename_dst[dst_index].pair)
 218                 die("internal error: dst already matched.");
 219
 220         src = rename_src[src_index].one;
 221         one = alloc_filespec(src->path);
 222         fill_filespec(one, src->sha1, src->mode);
 223
 224         dst = rename_dst[dst_index].two;
 225         two = alloc_filespec(dst->path);
 226         fill_filespec(two, dst->sha1, dst->mode);
 227
 228         dp = diff_queue(NULL, one, two);
 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         dp->source_stays = rename_src[src_index].src_path_left;
 235         rename_dst[dst_index].pair = dp;
 236 }
 237
 238 /*
 239  * We sort the rename similarity matrix with the score, in descending
 240  * order (the most similar first).
 241  */
 242 static int score_compare(const void *a_, const void *b_)
 243 {
 244         const struct diff_score *a = a_, *b = b_;
 245         return b->score - a->score;
 246 }
 247
 248 static int compute_stays(struct diff_queue_struct *q,
 249                          struct diff_filespec *one)
 250 {
 251         int i;
 252         for (i = 0; i < q->nr; i++) {
 253                 struct diff_filepair *p = q->queue[i];
 254                 if (strcmp(one->path, p->two->path))
 255                         continue;
 256                 if (DIFF_PAIR_RENAME(p)) {
 257                         return 0; /* something else is renamed into this */
 258                 }
 259         }
 260         return 1;
 261 }
 262
 263 void diffcore_rename(struct diff_options *options)
 264 {
 265         int detect_rename = options->detect_rename;
 266         int minimum_score = options->rename_score;
 267         int rename_limit = options->rename_limit;
 268         struct diff_queue_struct *q = &diff_queued_diff;
 269         struct diff_queue_struct outq;
 270         struct diff_score *mx;
 271         int i, j, rename_count, contents_too;
 272         int num_create, num_src, dst_cnt;
 273
 274         if (!minimum_score)
 275                 minimum_score = DEFAULT_RENAME_SCORE;
 276         rename_count = 0;
 277
 278         for (i = 0; i < q->nr; i++) {
 279                 struct diff_filepair *p = q->queue[i];
 280                 if (!DIFF_FILE_VALID(p->one)) {
 281                         if (!DIFF_FILE_VALID(p->two))
 282                                 continue; /* unmerged */
 283                         else if (options->single_follow &&
 284                                  strcmp(options->single_follow, p->two->path))
 285                                 continue; /* not interested */
 286                         else
 287                                 locate_rename_dst(p->two, 1);
 288                 }
 289                 else if (!DIFF_FILE_VALID(p->two)) {
 290                         /* If the source is a broken "delete", and
 291                          * they did not really want to get broken,
 292                          * that means the source actually stays.
 293                          */
 294                         int stays = (p->broken_pair && !p->score);
 295                         register_rename_src(p->one, stays, p->score);
 296                 }
 297                 else if (detect_rename == DIFF_DETECT_COPY)
 298                         register_rename_src(p->one, 1, p->score);
 299         }
 300         if (rename_dst_nr == 0 || rename_src_nr == 0 ||
 301             (0 < rename_limit && rename_limit < rename_dst_nr))
 302                 goto cleanup; /* nothing to do */
 303
 304         /* We really want to cull the candidates list early
 305          * with cheap tests in order to avoid doing deltas.
 306          * The first round matches up the up-to-date entries,
 307          * and then during the second round we try to match
 308          * cache-dirty entries as well.
 309          */
 310         for (contents_too = 0; contents_too < 2; contents_too++) {
 311                 for (i = 0; i < rename_dst_nr; i++) {
 312                         struct diff_filespec *two = rename_dst[i].two;
 313                         if (rename_dst[i].pair)
 314                                 continue; /* dealt with an earlier round */
 315                         for (j = 0; j < rename_src_nr; j++) {
 316                                 int k;
 317                                 struct diff_filespec *one = rename_src[j].one;
 318                                 if (!is_exact_match(one, two, contents_too))
 319                                         continue;
 320
 321                                 /* see if there is a basename match, too */
 322                                 for (k = j; k < rename_src_nr; k++) {
 323                                         one = rename_src[k].one;
 324                                         if (basename_same(one, two) &&
 325                                                 is_exact_match(one, two,
 326                                                         contents_too)) {
 327                                                 j = k;
 328                                                 break;
 329                                         }
 330                                 }
 331
 332                                 record_rename_pair(i, j, (int)MAX_SCORE);
 333                                 rename_count++;
 334                                 break; /* we are done with this entry */
 335                         }
 336                 }
 337         }
 338
 339         /* Have we run out the created file pool?  If so we can avoid
 340          * doing the delta matrix altogether.
 341          */
 342         if (rename_count == rename_dst_nr)
 343                 goto cleanup;
 344
 345         if (minimum_score == MAX_SCORE)
 346                 goto cleanup;
 347
 348         num_create = (rename_dst_nr - rename_count);
 349         num_src = rename_src_nr;
 350         mx = xmalloc(sizeof(*mx) * num_create * num_src);
 351         for (dst_cnt = i = 0; i < rename_dst_nr; i++) {
 352                 int base = dst_cnt * num_src;
 353                 struct diff_filespec *two = rename_dst[i].two;
 354                 if (rename_dst[i].pair)
 355                         continue; /* dealt with exact match already. */
 356                 for (j = 0; j < rename_src_nr; j++) {
 357                         struct diff_filespec *one = rename_src[j].one;
 358                         struct diff_score *m = &mx[base+j];
 359                         m->src = j;
 360                         m->dst = i;
 361                         m->score = estimate_similarity(one, two,
 362                                                        minimum_score);
 363                         diff_free_filespec_data(one);
 364                 }
 365                 /* We do not need the text anymore */
 366                 diff_free_filespec_data(two);
 367                 dst_cnt++;
 368         }
 369         /* cost matrix sorted by most to least similar pair */
 370         qsort(mx, num_create * num_src, sizeof(*mx), score_compare);
 371         for (i = 0; i < num_create * num_src; i++) {
 372                 struct diff_rename_dst *dst = &rename_dst[mx[i].dst];
 373                 if (dst->pair)
 374                         continue; /* already done, either exact or fuzzy. */
 375                 if (mx[i].score < minimum_score)
 376                         break; /* there is no more usable pair. */
 377                 record_rename_pair(mx[i].dst, mx[i].src, mx[i].score);
 378                 rename_count++;
 379         }
 380         free(mx);
 381
 382  cleanup:
 383         /* At this point, we have found some renames and copies and they
 384          * are recorded in rename_dst.  The original list is still in *q.
 385          */
 386         outq.queue = NULL;
 387         outq.nr = outq.alloc = 0;
 388         for (i = 0; i < q->nr; i++) {
 389                 struct diff_filepair *p = q->queue[i];
 390                 struct diff_filepair *pair_to_free = NULL;
 391
 392                 if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
 393                         /*
 394                          * Creation
 395                          *
 396                          * We would output this create record if it has
 397                          * not been turned into a rename/copy already.
 398                          */
 399                         struct diff_rename_dst *dst =
 400                                 locate_rename_dst(p->two, 0);
 401                         if (dst && dst->pair) {
 402                                 diff_q(&outq, dst->pair);
 403                                 pair_to_free = p;
 404                         }
 405                         else
 406                                 /* no matching rename/copy source, so
 407                                  * record this as a creation.
 408                                  */
 409                                 diff_q(&outq, p);
 410                 }
 411                 else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) {
 412                         /*
 413                          * Deletion
 414                          *
 415                          * We would output this delete record if:
 416                          *
 417                          * (1) this is a broken delete and the counterpart
 418                          *     broken create remains in the output; or
 419                          * (2) this is not a broken delete, and rename_dst
 420                          *     does not have a rename/copy to move p->one->path
 421                          *     out of existence.
 422                          *
 423                          * Otherwise, the counterpart broken create
 424                          * has been turned into a rename-edit; or
 425                          * delete did not have a matching create to
 426                          * begin with.
 427                          */
 428                         if (DIFF_PAIR_BROKEN(p)) {
 429                                 /* broken delete */
 430                                 struct diff_rename_dst *dst =
 431                                         locate_rename_dst(p->one, 0);
 432                                 if (dst && dst->pair)
 433                                         /* counterpart is now rename/copy */
 434                                         pair_to_free = p;
 435                         }
 436                         else {
 437                                 for (j = 0; j < rename_dst_nr; j++) {
 438                                         if (!rename_dst[j].pair)
 439                                                 continue;
 440                                         if (strcmp(rename_dst[j].pair->
 441                                                    one->path,
 442                                                    p->one->path))
 443                                                 continue;
 444                                         break;
 445                                 }
 446                                 if (j < rename_dst_nr)
 447                                         /* this path remains */
 448                                         pair_to_free = p;
 449                         }
 450
 451                         if (pair_to_free)
 452                                 ;
 453                         else
 454                                 diff_q(&outq, p);
 455                 }
 456                 else if (!diff_unmodified_pair(p))
 457                         /* all the usual ones need to be kept */
 458                         diff_q(&outq, p);
 459                 else
 460                         /* no need to keep unmodified pairs */
 461                         pair_to_free = p;
 462
 463                 if (pair_to_free)
 464                         diff_free_filepair(pair_to_free);
 465         }
 466         diff_debug_queue("done copying original", &outq);
 467
 468         free(q->queue);
 469         *q = outq;
 470         diff_debug_queue("done collapsing", q);
 471
 472         /* We need to see which rename source really stays here;
 473          * earlier we only checked if the path is left in the result,
 474          * but even if a path remains in the result, if that is coming
 475          * from copying something else on top of it, then the original
 476          * source is lost and does not stay.
 477          */
 478         for (i = 0; i < q->nr; i++) {
 479                 struct diff_filepair *p = q->queue[i];
 480                 if (DIFF_PAIR_RENAME(p) && p->source_stays) {
 481                         /* If one appears as the target of a rename-copy,
 482                          * then mark p->source_stays = 0; otherwise
 483                          * leave it as is.
 484                          */
 485                         p->source_stays = compute_stays(q, p->one);
 486                 }
 487         }
 488
 489         for (i = 0; i < rename_dst_nr; i++) {
 490                 diff_free_filespec_data(rename_dst[i].two);
 491                 free(rename_dst[i].two);
 492         }
 493
 494         free(rename_dst);
 495         rename_dst = NULL;
 496         rename_dst_nr = rename_dst_alloc = 0;
 497         free(rename_src);
 498         rename_src = NULL;
 499         rename_src_nr = rename_src_alloc = 0;
 500         return;
 501 }