builtin/name-rev.c

   1 #include "builtin.h"
   2 #include "cache.h"
   3 #include "repository.h"
   4 #include "config.h"
   5 #include "commit.h"
   6 #include "tag.h"
   7 #include "refs.h"
   8 #include "parse-options.h"
   9 #include "prio-queue.h"
  10 #include "hash-lookup.h"
  11 #include "commit-slab.h"
  12 #include "commit-graph.h"
  13
  14 /*
  15  * One day.  See the 'name a rev shortly after epoch' test in t6120 when
  16  * changing this value
  17  */
  18 #define CUTOFF_DATE_SLOP 86400
  19
  20 struct rev_name {
  21         const char *tip_name;
  22         timestamp_t taggerdate;
  23         int generation;
  24         int distance;
  25         int from_tag;
  26 };
  27
  28 define_commit_slab(commit_rev_name, struct rev_name);
  29
  30 static timestamp_t generation_cutoff = GENERATION_NUMBER_INFINITY;
  31 static timestamp_t cutoff = TIME_MAX;
  32 static struct commit_rev_name rev_names;
  33
  34 /* Disable the cutoff checks entirely */
  35 static void disable_cutoff(void)
  36 {
  37         generation_cutoff = 0;
  38         cutoff = 0;
  39 }
  40
  41 /* Cutoff searching any commits older than this one */
  42 static void set_commit_cutoff(struct commit *commit)
  43 {
  44
  45         if (cutoff > commit->date)
  46                 cutoff = commit->date;
  47
  48         if (generation_cutoff) {
  49                 timestamp_t generation = commit_graph_generation(commit);
  50
  51                 if (generation_cutoff > generation)
  52                         generation_cutoff = generation;
  53         }
  54 }
  55
  56 /* adjust the commit date cutoff with a slop to allow for slightly incorrect
  57  * commit timestamps in case of clock skew.
  58  */
  59 static void adjust_cutoff_timestamp_for_slop(void)
  60 {
  61         if (cutoff) {
  62                 /* check for undeflow */
  63                 if (cutoff > TIME_MIN + CUTOFF_DATE_SLOP)
  64                         cutoff = cutoff - CUTOFF_DATE_SLOP;
  65                 else
  66                         cutoff = TIME_MIN;
  67         }
  68 }
  69
  70 /* Check if a commit is before the cutoff. Prioritize generation numbers
  71  * first, but use the commit timestamp if we lack generation data.
  72  */
  73 static int commit_is_before_cutoff(struct commit *commit)
  74 {
  75         if (generation_cutoff < GENERATION_NUMBER_INFINITY)
  76                 return generation_cutoff &&
  77                         commit_graph_generation(commit) < generation_cutoff;
  78
  79         return commit->date < cutoff;
  80 }
  81
  82 /* How many generations are maximally preferred over _one_ merge traversal? */
  83 #define MERGE_TRAVERSAL_WEIGHT 65535
  84
  85 static int is_valid_rev_name(const struct rev_name *name)
  86 {
  87         return name && name->tip_name;
  88 }
  89
  90 static struct rev_name *get_commit_rev_name(const struct commit *commit)
  91 {
  92         struct rev_name *name = commit_rev_name_peek(&rev_names, commit);
  93
  94         return is_valid_rev_name(name) ? name : NULL;
  95 }
  96
  97 static int effective_distance(int distance, int generation)
  98 {
  99         return distance + (generation > 0 ? MERGE_TRAVERSAL_WEIGHT : 0);
 100 }
 101
 102 static int is_better_name(struct rev_name *name,
 103                           timestamp_t taggerdate,
 104                           int generation,
 105                           int distance,
 106                           int from_tag)
 107 {
 108         int name_distance = effective_distance(name->distance, name->generation);
 109         int new_distance = effective_distance(distance, generation);
 110
 111         /* If both are tags, we prefer the nearer one. */
 112         if (from_tag && name->from_tag)
 113                 return name_distance > new_distance;
 114
 115         /* Favor a tag over a non-tag. */
 116         if (name->from_tag != from_tag)
 117                 return from_tag;
 118
 119         /*
 120          * We are now looking at two non-tags.  Tiebreak to favor
 121          * shorter hops.
 122          */
 123         if (name_distance != new_distance)
 124                 return name_distance > new_distance;
 125
 126         /* ... or tiebreak to favor older date */
 127         if (name->taggerdate != taggerdate)
 128                 return name->taggerdate > taggerdate;
 129
 130         /* keep the current one if we cannot decide */
 131         return 0;
 132 }
 133
 134 static struct rev_name *create_or_update_name(struct commit *commit,
 135                                               timestamp_t taggerdate,
 136                                               int generation, int distance,
 137                                               int from_tag)
 138 {
 139         struct rev_name *name = commit_rev_name_at(&rev_names, commit);
 140
 141         if (is_valid_rev_name(name) &&
 142             !is_better_name(name, taggerdate, generation, distance, from_tag))
 143                 return NULL;
 144
 145         name->taggerdate = taggerdate;
 146         name->generation = generation;
 147         name->distance = distance;
 148         name->from_tag = from_tag;
 149
 150         return name;
 151 }
 152
 153 static char *get_parent_name(const struct rev_name *name, int parent_number)
 154 {
 155         struct strbuf sb = STRBUF_INIT;
 156         size_t len;
 157
 158         strip_suffix(name->tip_name, "^0", &len);
 159         if (name->generation > 0) {
 160                 strbuf_grow(&sb, len +
 161                             1 + decimal_width(name->generation) +
 162                             1 + decimal_width(parent_number));
 163                 strbuf_addf(&sb, "%.*s~%d^%d", (int)len, name->tip_name,
 164                             name->generation, parent_number);
 165         } else {
 166                 strbuf_grow(&sb, len +
 167                             1 + decimal_width(parent_number));
 168                 strbuf_addf(&sb, "%.*s^%d", (int)len, name->tip_name,
 169                             parent_number);
 170         }
 171         return strbuf_detach(&sb, NULL);
 172 }
 173
 174 static void name_rev(struct commit *start_commit,
 175                 const char *tip_name, timestamp_t taggerdate,
 176                 int from_tag, int deref)
 177 {
 178         struct prio_queue queue;
 179         struct commit *commit;
 180         struct commit **parents_to_queue = NULL;
 181         size_t parents_to_queue_nr, parents_to_queue_alloc = 0;
 182         struct rev_name *start_name;
 183
 184         parse_commit(start_commit);
 185         if (commit_is_before_cutoff(start_commit))
 186                 return;
 187
 188         start_name = create_or_update_name(start_commit, taggerdate, 0, 0,
 189                                            from_tag);
 190         if (!start_name)
 191                 return;
 192         if (deref)
 193                 start_name->tip_name = xstrfmt("%s^0", tip_name);
 194         else
 195                 start_name->tip_name = xstrdup(tip_name);
 196
 197         memset(&queue, 0, sizeof(queue)); /* Use the prio_queue as LIFO */
 198         prio_queue_put(&queue, start_commit);
 199
 200         while ((commit = prio_queue_get(&queue))) {
 201                 struct rev_name *name = get_commit_rev_name(commit);
 202                 struct commit_list *parents;
 203                 int parent_number = 1;
 204
 205                 parents_to_queue_nr = 0;
 206
 207                 for (parents = commit->parents;
 208                                 parents;
 209                                 parents = parents->next, parent_number++) {
 210                         struct commit *parent = parents->item;
 211                         struct rev_name *parent_name;
 212                         int generation, distance;
 213
 214                         parse_commit(parent);
 215                         if (commit_is_before_cutoff(parent))
 216                                 continue;
 217
 218                         if (parent_number > 1) {
 219                                 generation = 0;
 220                                 distance = name->distance + MERGE_TRAVERSAL_WEIGHT;
 221                         } else {
 222                                 generation = name->generation + 1;
 223                                 distance = name->distance + 1;
 224                         }
 225
 226                         parent_name = create_or_update_name(parent, taggerdate,
 227                                                             generation,
 228                                                             distance, from_tag);
 229                         if (parent_name) {
 230                                 if (parent_number > 1)
 231                                         parent_name->tip_name =
 232                                                 get_parent_name(name,
 233                                                                 parent_number);
 234                                 else
 235                                         parent_name->tip_name = name->tip_name;
 236                                 ALLOC_GROW(parents_to_queue,
 237                                            parents_to_queue_nr + 1,
 238                                            parents_to_queue_alloc);
 239                                 parents_to_queue[parents_to_queue_nr] = parent;
 240                                 parents_to_queue_nr++;
 241                         }
 242                 }
 243
 244                 /* The first parent must come out first from the prio_queue */
 245                 while (parents_to_queue_nr)
 246                         prio_queue_put(&queue,
 247                                        parents_to_queue[--parents_to_queue_nr]);
 248         }
 249
 250         clear_prio_queue(&queue);
 251         free(parents_to_queue);
 252 }
 253
 254 static int subpath_matches(const char *path, const char *filter)
 255 {
 256         const char *subpath = path;
 257
 258         while (subpath) {
 259                 if (!wildmatch(filter, subpath, 0))
 260                         return subpath - path;
 261                 subpath = strchr(subpath, '/');
 262                 if (subpath)
 263                         subpath++;
 264         }
 265         return -1;
 266 }
 267
 268 struct name_ref_data {
 269         int tags_only;
 270         int name_only;
 271         struct string_list ref_filters;
 272         struct string_list exclude_filters;
 273 };
 274
 275 static struct tip_table {
 276         struct tip_table_entry {
 277                 struct object_id oid;
 278                 const char *refname;
 279                 struct commit *commit;
 280                 timestamp_t taggerdate;
 281                 unsigned int from_tag:1;
 282                 unsigned int deref:1;
 283         } *table;
 284         int nr;
 285         int alloc;
 286         int sorted;
 287 } tip_table;
 288
 289 static void add_to_tip_table(const struct object_id *oid, const char *refname,
 290                              int shorten_unambiguous, struct commit *commit,
 291                              timestamp_t taggerdate, int from_tag, int deref)
 292 {
 293         char *short_refname = NULL;
 294
 295         if (shorten_unambiguous)
 296                 short_refname = shorten_unambiguous_ref(refname, 0);
 297         else if (skip_prefix(refname, "refs/heads/", &refname))
 298                 ; /* refname already advanced */
 299         else
 300                 skip_prefix(refname, "refs/", &refname);
 301
 302         ALLOC_GROW(tip_table.table, tip_table.nr + 1, tip_table.alloc);
 303         oidcpy(&tip_table.table[tip_table.nr].oid, oid);
 304         tip_table.table[tip_table.nr].refname = short_refname ?
 305                 short_refname : xstrdup(refname);
 306         tip_table.table[tip_table.nr].commit = commit;
 307         tip_table.table[tip_table.nr].taggerdate = taggerdate;
 308         tip_table.table[tip_table.nr].from_tag = from_tag;
 309         tip_table.table[tip_table.nr].deref = deref;
 310         tip_table.nr++;
 311         tip_table.sorted = 0;
 312 }
 313
 314 static int tipcmp(const void *a_, const void *b_)
 315 {
 316         const struct tip_table_entry *a = a_, *b = b_;
 317         return oidcmp(&a->oid, &b->oid);
 318 }
 319
 320 static int cmp_by_tag_and_age(const void *a_, const void *b_)
 321 {
 322         const struct tip_table_entry *a = a_, *b = b_;
 323         int cmp;
 324
 325         /* Prefer tags. */
 326         cmp = b->from_tag - a->from_tag;
 327         if (cmp)
 328                 return cmp;
 329
 330         /* Older is better. */
 331         if (a->taggerdate < b->taggerdate)
 332                 return -1;
 333         return a->taggerdate != b->taggerdate;
 334 }
 335
 336 static int name_ref(const char *path, const struct object_id *oid,
 337                     int flags UNUSED, void *cb_data)
 338 {
 339         struct object *o = parse_object(the_repository, oid);
 340         struct name_ref_data *data = cb_data;
 341         int can_abbreviate_output = data->tags_only && data->name_only;
 342         int deref = 0;
 343         int from_tag = 0;
 344         struct commit *commit = NULL;
 345         timestamp_t taggerdate = TIME_MAX;
 346
 347         if (data->tags_only && !starts_with(path, "refs/tags/"))
 348                 return 0;
 349
 350         if (data->exclude_filters.nr) {
 351                 struct string_list_item *item;
 352
 353                 for_each_string_list_item(item, &data->exclude_filters) {
 354                         if (subpath_matches(path, item->string) >= 0)
 355                                 return 0;
 356                 }
 357         }
 358
 359         if (data->ref_filters.nr) {
 360                 struct string_list_item *item;
 361                 int matched = 0;
 362
 363                 /* See if any of the patterns match. */
 364                 for_each_string_list_item(item, &data->ref_filters) {
 365                         /*
 366                          * Check all patterns even after finding a match, so
 367                          * that we can see if a match with a subpath exists.
 368                          * When a user asked for 'refs/tags/v*' and 'v1.*',
 369                          * both of which match, the user is showing her
 370                          * willingness to accept a shortened output by having
 371                          * the 'v1.*' in the acceptable refnames, so we
 372                          * shouldn't stop when seeing 'refs/tags/v1.4' matches
 373                          * 'refs/tags/v*'.  We should show it as 'v1.4'.
 374                          */
 375                         switch (subpath_matches(path, item->string)) {
 376                         case -1: /* did not match */
 377                                 break;
 378                         case 0: /* matched fully */
 379                                 matched = 1;
 380                                 break;
 381                         default: /* matched subpath */
 382                                 matched = 1;
 383                                 can_abbreviate_output = 1;
 384                                 break;
 385                         }
 386                 }
 387
 388                 /* If none of the patterns matched, stop now */
 389                 if (!matched)
 390                         return 0;
 391         }
 392
 393         while (o && o->type == OBJ_TAG) {
 394                 struct tag *t = (struct tag *) o;
 395                 if (!t->tagged)
 396                         break; /* broken repository */
 397                 o = parse_object(the_repository, &t->tagged->oid);
 398                 deref = 1;
 399                 taggerdate = t->date;
 400         }
 401         if (o && o->type == OBJ_COMMIT) {
 402                 commit = (struct commit *)o;
 403                 from_tag = starts_with(path, "refs/tags/");
 404                 if (taggerdate == TIME_MAX)
 405                         taggerdate = commit->date;
 406         }
 407
 408         add_to_tip_table(oid, path, can_abbreviate_output, commit, taggerdate,
 409                          from_tag, deref);
 410         return 0;
 411 }
 412
 413 static void name_tips(void)
 414 {
 415         int i;
 416
 417         /*
 418          * Try to set better names first, so that worse ones spread
 419          * less.
 420          */
 421         QSORT(tip_table.table, tip_table.nr, cmp_by_tag_and_age);
 422         for (i = 0; i < tip_table.nr; i++) {
 423                 struct tip_table_entry *e = &tip_table.table[i];
 424                 if (e->commit) {
 425                         name_rev(e->commit, e->refname, e->taggerdate,
 426                                  e->from_tag, e->deref);
 427                 }
 428         }
 429 }
 430
 431 static const struct object_id *nth_tip_table_ent(size_t ix, const void *table_)
 432 {
 433         const struct tip_table_entry *table = table_;
 434         return &table[ix].oid;
 435 }
 436
 437 static const char *get_exact_ref_match(const struct object *o)
 438 {
 439         int found;
 440
 441         if (!tip_table.table || !tip_table.nr)
 442                 return NULL;
 443
 444         if (!tip_table.sorted) {
 445                 QSORT(tip_table.table, tip_table.nr, tipcmp);
 446                 tip_table.sorted = 1;
 447         }
 448
 449         found = oid_pos(&o->oid, tip_table.table, tip_table.nr,
 450                         nth_tip_table_ent);
 451         if (0 <= found)
 452                 return tip_table.table[found].refname;
 453         return NULL;
 454 }
 455
 456 /* may return a constant string or use "buf" as scratch space */
 457 static const char *get_rev_name(const struct object *o, struct strbuf *buf)
 458 {
 459         struct rev_name *n;
 460         const struct commit *c;
 461
 462         if (o->type != OBJ_COMMIT)
 463                 return get_exact_ref_match(o);
 464         c = (const struct commit *) o;
 465         n = get_commit_rev_name(c);
 466         if (!n)
 467                 return NULL;
 468
 469         if (!n->generation)
 470                 return n->tip_name;
 471         else {
 472                 strbuf_reset(buf);
 473                 strbuf_addstr(buf, n->tip_name);
 474                 strbuf_strip_suffix(buf, "^0");
 475                 strbuf_addf(buf, "~%d", n->generation);
 476                 return buf->buf;
 477         }
 478 }
 479
 480 static void show_name(const struct object *obj,
 481                       const char *caller_name,
 482                       int always, int allow_undefined, int name_only)
 483 {
 484         const char *name;
 485         const struct object_id *oid = &obj->oid;
 486         struct strbuf buf = STRBUF_INIT;
 487
 488         if (!name_only)
 489                 printf("%s ", caller_name ? caller_name : oid_to_hex(oid));
 490         name = get_rev_name(obj, &buf);
 491         if (name)
 492                 printf("%s\n", name);
 493         else if (allow_undefined)
 494                 printf("undefined\n");
 495         else if (always)
 496                 printf("%s\n", find_unique_abbrev(oid, DEFAULT_ABBREV));
 497         else
 498                 die("cannot describe '%s'", oid_to_hex(oid));
 499         strbuf_release(&buf);
 500 }
 501
 502 static char const * const name_rev_usage[] = {
 503         N_("git name-rev [<options>] <commit>..."),
 504         N_("git name-rev [<options>] --all"),
 505         N_("git name-rev [<options>] --annotate-stdin"),
 506         NULL
 507 };
 508
 509 static void name_rev_line(char *p, struct name_ref_data *data)
 510 {
 511         struct strbuf buf = STRBUF_INIT;
 512         int counter = 0;
 513         char *p_start;
 514         const unsigned hexsz = the_hash_algo->hexsz;
 515
 516         for (p_start = p; *p; p++) {
 517 #define ishex(x) (isdigit((x)) || ((x) >= 'a' && (x) <= 'f'))
 518                 if (!ishex(*p))
 519                         counter = 0;
 520                 else if (++counter == hexsz &&
 521                          !ishex(*(p+1))) {
 522                         struct object_id oid;
 523                         const char *name = NULL;
 524                         char c = *(p+1);
 525                         int p_len = p - p_start + 1;
 526
 527                         counter = 0;
 528
 529                         *(p+1) = 0;
 530                         if (!get_oid(p - (hexsz - 1), &oid)) {
 531                                 struct object *o =
 532                                         lookup_object(the_repository, &oid);
 533                                 if (o)
 534                                         name = get_rev_name(o, &buf);
 535                         }
 536                         *(p+1) = c;
 537
 538                         if (!name)
 539                                 continue;
 540
 541                         if (data->name_only)
 542                                 printf("%.*s%s", p_len - hexsz, p_start, name);
 543                         else
 544                                 printf("%.*s (%s)", p_len, p_start, name);
 545                         p_start = p + 1;
 546                 }
 547         }
 548
 549         /* flush */
 550         if (p_start != p)
 551                 fwrite(p_start, p - p_start, 1, stdout);
 552
 553         strbuf_release(&buf);
 554 }
 555
 556 int cmd_name_rev(int argc, const char **argv, const char *prefix)
 557 {
 558         struct object_array revs = OBJECT_ARRAY_INIT;
 559         int all = 0, annotate_stdin = 0, transform_stdin = 0, allow_undefined = 1, always = 0, peel_tag = 0;
 560         struct name_ref_data data = { 0, 0, STRING_LIST_INIT_NODUP, STRING_LIST_INIT_NODUP };
 561         struct option opts[] = {
 562                 OPT_BOOL(0, "name-only", &data.name_only, N_("print only ref-based names (no object names)")),
 563                 OPT_BOOL(0, "tags", &data.tags_only, N_("only use tags to name the commits")),
 564                 OPT_STRING_LIST(0, "refs", &data.ref_filters, N_("pattern"),
 565                                    N_("only use refs matching <pattern>")),
 566                 OPT_STRING_LIST(0, "exclude", &data.exclude_filters, N_("pattern"),
 567                                    N_("ignore refs matching <pattern>")),
 568                 OPT_GROUP(""),
 569                 OPT_BOOL(0, "all", &all, N_("list all commits reachable from all refs")),
 570                 OPT_BOOL(0, "stdin", &transform_stdin, N_("deprecated: use --annotate-stdin instead")),
 571                 OPT_BOOL(0, "annotate-stdin", &annotate_stdin, N_("annotate text from stdin")),
 572                 OPT_BOOL(0, "undefined", &allow_undefined, N_("allow to print `undefined` names (default)")),
 573                 OPT_BOOL(0, "always",     &always,
 574                            N_("show abbreviated commit object as fallback")),
 575                 {
 576                         /* A Hidden OPT_BOOL */
 577                         OPTION_SET_INT, 0, "peel-tag", &peel_tag, NULL,
 578                         N_("dereference tags in the input (internal use)"),
 579                         PARSE_OPT_NOARG | PARSE_OPT_HIDDEN, NULL, 1,
 580                 },
 581                 OPT_END(),
 582         };
 583
 584         init_commit_rev_name(&rev_names);
 585         git_config(git_default_config, NULL);
 586         argc = parse_options(argc, argv, prefix, opts, name_rev_usage, 0);
 587
 588         if (transform_stdin) {
 589                 warning("--stdin is deprecated. Please use --annotate-stdin instead, "
 590                                         "which is functionally equivalent.\n"
 591                                         "This option will be removed in a future release.");
 592                 annotate_stdin = 1;
 593         }
 594
 595         if (all + annotate_stdin + !!argc > 1) {
 596                 error("Specify either a list, or --all, not both!");
 597                 usage_with_options(name_rev_usage, opts);
 598         }
 599         if (all || annotate_stdin)
 600                 disable_cutoff();
 601
 602         for (; argc; argc--, argv++) {
 603                 struct object_id oid;
 604                 struct object *object;
 605                 struct commit *commit;
 606
 607                 if (get_oid(*argv, &oid)) {
 608                         fprintf(stderr, "Could not get sha1 for %s. Skipping.\n",
 609                                         *argv);
 610                         continue;
 611                 }
 612
 613                 commit = NULL;
 614                 object = parse_object(the_repository, &oid);
 615                 if (object) {
 616                         struct object *peeled = deref_tag(the_repository,
 617                                                           object, *argv, 0);
 618                         if (peeled && peeled->type == OBJ_COMMIT)
 619                                 commit = (struct commit *)peeled;
 620                 }
 621
 622                 if (!object) {
 623                         fprintf(stderr, "Could not get object for %s. Skipping.\n",
 624                                         *argv);
 625                         continue;
 626                 }
 627
 628                 if (commit)
 629                         set_commit_cutoff(commit);
 630
 631                 if (peel_tag) {
 632                         if (!commit) {
 633                                 fprintf(stderr, "Could not get commit for %s. Skipping.\n",
 634                                         *argv);
 635                                 continue;
 636                         }
 637                         object = (struct object *)commit;
 638                 }
 639                 add_object_array(object, *argv, &revs);
 640         }
 641
 642         adjust_cutoff_timestamp_for_slop();
 643
 644         for_each_ref(name_ref, &data);
 645         name_tips();
 646
 647         if (annotate_stdin) {
 648                 struct strbuf sb = STRBUF_INIT;
 649
 650                 while (strbuf_getline(&sb, stdin) != EOF) {
 651                         strbuf_addch(&sb, '\n');
 652                         name_rev_line(sb.buf, &data);
 653                 }
 654                 strbuf_release(&sb);
 655         } else if (all) {
 656                 int i, max;
 657
 658                 max = get_max_object_index();
 659                 for (i = 0; i < max; i++) {
 660                         struct object *obj = get_indexed_object(i);
 661                         if (!obj || obj->type != OBJ_COMMIT)
 662                                 continue;
 663                         show_name(obj, NULL,
 664                                   always, allow_undefined, data.name_only);
 665                 }
 666         } else {
 667                 int i;
 668                 for (i = 0; i < revs.nr; i++)
 669                         show_name(revs.objects[i].item, revs.objects[i].name,
 670                                   always, allow_undefined, data.name_only);
 671         }
 672
 673         UNLEAK(revs);
 674         return 0;
 675 }