Andrew's git / gitweb.git / sha1_name.c
?
sha1_name.con commit t4020: don't use grep -a (53a5b44)
   1#include "cache.h"
   2#include "tag.h"
   3#include "commit.h"
   4#include "tree.h"
   5#include "blob.h"
   6#include "tree-walk.h"
   7#include "refs.h"
   8
   9static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
  10{
  11        struct alternate_object_database *alt;
  12        char hex[40];
  13        int found = 0;
  14        static struct alternate_object_database *fakeent;
  15
  16        if (!fakeent) {
  17                const char *objdir = get_object_directory();
  18                int objdir_len = strlen(objdir);
  19                int entlen = objdir_len + 43;
  20                fakeent = xmalloc(sizeof(*fakeent) + entlen);
  21                memcpy(fakeent->base, objdir, objdir_len);
  22                fakeent->name = fakeent->base + objdir_len + 1;
  23                fakeent->name[-1] = '/';
  24        }
  25        fakeent->next = alt_odb_list;
  26
  27        sprintf(hex, "%.2s", name);
  28        for (alt = fakeent; alt && found < 2; alt = alt->next) {
  29                struct dirent *de;
  30                DIR *dir;
  31                sprintf(alt->name, "%.2s/", name);
  32                dir = opendir(alt->base);
  33                if (!dir)
  34                        continue;
  35                while ((de = readdir(dir)) != NULL) {
  36                        if (strlen(de->d_name) != 38)
  37                                continue;
  38                        if (memcmp(de->d_name, name + 2, len - 2))
  39                                continue;
  40                        if (!found) {
  41                                memcpy(hex + 2, de->d_name, 38);
  42                                found++;
  43                        }
  44                        else if (memcmp(hex + 2, de->d_name, 38)) {
  45                                found = 2;
  46                                break;
  47                        }
  48                }
  49                closedir(dir);
  50        }
  51        if (found == 1)
  52                return get_sha1_hex(hex, sha1) == 0;
  53        return found;
  54}
  55
  56static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
  57{
  58        do {
  59                if (*a != *b)
  60                        return 0;
  61                a++;
  62                b++;
  63                len -= 2;
  64        } while (len > 1);
  65        if (len)
  66                if ((*a ^ *b) & 0xf0)
  67                        return 0;
  68        return 1;
  69}
  70
  71static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
  72{
  73        struct packed_git *p;
  74        const unsigned char *found_sha1 = NULL;
  75        int found = 0;
  76
  77        prepare_packed_git();
  78        for (p = packed_git; p && found < 2; p = p->next) {
  79                uint32_t num, last;
  80                uint32_t first = 0;
  81                open_pack_index(p);
  82                num = p->num_objects;
  83                last = num;
  84                while (first < last) {
  85                        uint32_t mid = (first + last) / 2;
  86                        const unsigned char *now;
  87                        int cmp;
  88
  89                        now = nth_packed_object_sha1(p, mid);
  90                        cmp = hashcmp(match, now);
  91                        if (!cmp) {
  92                                first = mid;
  93                                break;
  94                        }
  95                        if (cmp > 0) {
  96                                first = mid+1;
  97                                continue;
  98                        }
  99                        last = mid;
 100                }
 101                if (first < num) {
 102                        const unsigned char *now, *next;
 103                       now = nth_packed_object_sha1(p, first);
 104                        if (match_sha(len, match, now)) {
 105                                next = nth_packed_object_sha1(p, first+1);
 106                               if (!next|| !match_sha(len, match, next)) {
 107                                        /* unique within this pack */
 108                                        if (!found) {
 109                                                found_sha1 = now;
 110                                                found++;
 111                                        }
 112                                        else if (hashcmp(found_sha1, now)) {
 113                                                found = 2;
 114                                                break;
 115                                        }
 116                                }
 117                                else {
 118                                        /* not even unique within this pack */
 119                                        found = 2;
 120                                        break;
 121                                }
 122                        }
 123                }
 124        }
 125        if (found == 1)
 126                hashcpy(sha1, found_sha1);
 127        return found;
 128}
 129
 130#define SHORT_NAME_NOT_FOUND (-1)
 131#define SHORT_NAME_AMBIGUOUS (-2)
 132
 133static int find_unique_short_object(int len, char *canonical,
 134                                    unsigned char *res, unsigned char *sha1)
 135{
 136        int has_unpacked, has_packed;
 137        unsigned char unpacked_sha1[20], packed_sha1[20];
 138
 139        prepare_alt_odb();
 140        has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
 141        has_packed = find_short_packed_object(len, res, packed_sha1);
 142        if (!has_unpacked && !has_packed)
 143                return SHORT_NAME_NOT_FOUND;
 144        if (1 < has_unpacked || 1 < has_packed)
 145                return SHORT_NAME_AMBIGUOUS;
 146        if (has_unpacked != has_packed) {
 147                hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
 148                return 0;
 149        }
 150        /* Both have unique ones -- do they match? */
 151        if (hashcmp(packed_sha1, unpacked_sha1))
 152                return SHORT_NAME_AMBIGUOUS;
 153        hashcpy(sha1, packed_sha1);
 154        return 0;
 155}
 156
 157static int get_short_sha1(const char *name, int len, unsigned char *sha1,
 158                          int quietly)
 159{
 160        int i, status;
 161        char canonical[40];
 162        unsigned char res[20];
 163
 164        if (len < MINIMUM_ABBREV || len > 40)
 165                return -1;
 166        hashclr(res);
 167        memset(canonical, 'x', 40);
 168        for (i = 0; i < len ;i++) {
 169                unsigned char c = name[i];
 170                unsigned char val;
 171                if (c >= '0' && c <= '9')
 172                        val = c - '0';
 173                else if (c >= 'a' && c <= 'f')
 174                        val = c - 'a' + 10;
 175                else if (c >= 'A' && c <='F') {
 176                        val = c - 'A' + 10;
 177                        c -= 'A' - 'a';
 178                }
 179                else
 180                        return -1;
 181                canonical[i] = c;
 182                if (!(i & 1))
 183                        val <<= 4;
 184                res[i >> 1] |= val;
 185        }
 186
 187        status = find_unique_short_object(i, canonical, res, sha1);
 188        if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
 189                return error("short SHA1 %.*s is ambiguous.", len, canonical);
 190        return status;
 191}
 192
 193const char *find_unique_abbrev(const unsigned char *sha1, int len)
 194{
 195        int status, exists;
 196        static char hex[41];
 197
 198        exists = has_sha1_file(sha1);
 199        memcpy(hex, sha1_to_hex(sha1), 40);
 200        if (len == 40 || !len)
 201                return hex;
 202        while (len < 40) {
 203                unsigned char sha1_ret[20];
 204                status = get_short_sha1(hex, len, sha1_ret, 1);
 205                if (exists
 206                    ? !status
 207                    : status == SHORT_NAME_NOT_FOUND) {
 208                        hex[len] = 0;
 209                        return hex;
 210                }
 211                len++;
 212        }
 213        return hex;
 214}
 215
 216static int ambiguous_path(const char *path, int len)
 217{
 218        int slash = 1;
 219        int cnt;
 220
 221        for (cnt = 0; cnt < len; cnt++) {
 222                switch (*path++) {
 223                case '\0':
 224                        break;
 225                case '/':
 226                        if (slash)
 227                                break;
 228                        slash = 1;
 229                        continue;
 230                case '.':
 231                        continue;
 232                default:
 233                        slash = 0;
 234                        continue;
 235                }
 236                break;
 237        }
 238        return slash;
 239}
 240
 241int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
 242{
 243        const char **p, *r;
 244        int refs_found = 0;
 245
 246        *ref = NULL;
 247        for (p = ref_rev_parse_rules; *p; p++) {
 248                unsigned char sha1_from_ref[20];
 249                unsigned char *this_result;
 250
 251                this_result = refs_found ? sha1_from_ref : sha1;
 252                r = resolve_ref(mkpath(*p, len, str), this_result, 1, NULL);
 253                if (r) {
 254                        if (!refs_found++)
 255                                *ref = xstrdup(r);
 256                        if (!warn_ambiguous_refs)
 257                                break;
 258                }
 259        }
 260        return refs_found;
 261}
 262
 263int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
 264{
 265        const char **p;
 266        int logs_found = 0;
 267
 268        *log = NULL;
 269        for (p = ref_rev_parse_rules; *p; p++) {
 270                struct stat st;
 271                unsigned char hash[20];
 272                char path[PATH_MAX];
 273                const char *ref, *it;
 274
 275                strcpy(path, mkpath(*p, len, str));
 276                ref = resolve_ref(path, hash, 0, NULL);
 277                if (!ref)
 278                        continue;
 279                if (!stat(git_path("logs/%s", path), &st) &&
 280                    S_ISREG(st.st_mode))
 281                        it = path;
 282                else if (strcmp(ref, path) &&
 283                         !stat(git_path("logs/%s", ref), &st) &&
 284                         S_ISREG(st.st_mode))
 285                        it = ref;
 286                else
 287                        continue;
 288                if (!logs_found++) {
 289                        *log = xstrdup(it);
 290                        hashcpy(sha1, hash);
 291                }
 292                if (!warn_ambiguous_refs)
 293                        break;
 294        }
 295        return logs_found;
 296}
 297
 298static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
 299{
 300        static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
 301        char *real_ref = NULL;
 302        int refs_found = 0;
 303        int at, reflog_len;
 304
 305        if (len == 40 && !get_sha1_hex(str, sha1))
 306                return 0;
 307
 308        /* basic@{time or number} format to query ref-log */
 309        reflog_len = at = 0;
 310        if (str[len-1] == '}') {
 311                for (at = 0; at < len - 1; at++) {
 312                        if (str[at] == '@' && str[at+1] == '{') {
 313                                reflog_len = (len-1) - (at+2);
 314                                len = at;
 315                                break;
 316                        }
 317                }
 318        }
 319
 320        /* Accept only unambiguous ref paths. */
 321        if (len && ambiguous_path(str, len))
 322                return -1;
 323
 324        if (!len && reflog_len) {
 325                /* allow "@{...}" to mean the current branch reflog */
 326                refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
 327        } else if (reflog_len)
 328                refs_found = dwim_log(str, len, sha1, &real_ref);
 329        else
 330                refs_found = dwim_ref(str, len, sha1, &real_ref);
 331
 332        if (!refs_found)
 333                return -1;
 334
 335        if (warn_ambiguous_refs && refs_found > 1)
 336                fprintf(stderr, warning, len, str);
 337
 338        if (reflog_len) {
 339                int nth, i;
 340                unsigned long at_time;
 341                unsigned long co_time;
 342                int co_tz, co_cnt;
 343
 344                /* Is it asking for N-th entry, or approxidate? */
 345                for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
 346                        char ch = str[at+2+i];
 347                        if ('0' <= ch && ch <= '9')
 348                                nth = nth * 10 + ch - '0';
 349                        else
 350                                nth = -1;
 351                }
 352                if (0 <= nth)
 353                        at_time = 0;
 354                else
 355                        at_time = approxidate(str + at + 2);
 356                if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
 357                                &co_time, &co_tz, &co_cnt)) {
 358                        if (at_time)
 359                                fprintf(stderr,
 360                                        "warning: Log for '%.*s' only goes "
 361                                        "back to %s.\n", len, str,
 362                                        show_date(co_time, co_tz, DATE_RFC2822));
 363                        else
 364                                fprintf(stderr,
 365                                        "warning: Log for '%.*s' only has "
 366                                        "%d entries.\n", len, str, co_cnt);
 367                }
 368        }
 369
 370        free(real_ref);
 371        return 0;
 372}
 373
 374static int get_sha1_1(const char *name, int len, unsigned char *sha1);
 375
 376static int get_parent(const char *name, int len,
 377                      unsigned char *result, int idx)
 378{
 379        unsigned char sha1[20];
 380        int ret = get_sha1_1(name, len, sha1);
 381        struct commit *commit;
 382        struct commit_list *p;
 383
 384        if (ret)
 385                return ret;
 386        commit = lookup_commit_reference(sha1);
 387        if (!commit)
 388                return -1;
 389        if (parse_commit(commit))
 390                return -1;
 391        if (!idx) {
 392                hashcpy(result, commit->object.sha1);
 393                return 0;
 394        }
 395        p = commit->parents;
 396        while (p) {
 397                if (!--idx) {
 398                        hashcpy(result, p->item->object.sha1);
 399                        return 0;
 400                }
 401                p = p->next;
 402        }
 403        return -1;
 404}
 405
 406static int get_nth_ancestor(const char *name, int len,
 407                            unsigned char *result, int generation)
 408{
 409        unsigned char sha1[20];
 410        int ret = get_sha1_1(name, len, sha1);
 411        if (ret)
 412                return ret;
 413
 414        while (generation--) {
 415                struct commit *commit = lookup_commit_reference(sha1);
 416
 417                if (!commit || parse_commit(commit) || !commit->parents)
 418                        return -1;
 419                hashcpy(sha1, commit->parents->item->object.sha1);
 420        }
 421        hashcpy(result, sha1);
 422        return 0;
 423}
 424
 425struct object *peel_to_type(const char *name, int namelen,
 426                            struct object *o, enum object_type expected_type)
 427{
 428        if (name && !namelen)
 429                namelen = strlen(name);
 430        if (!o) {
 431                unsigned char sha1[20];
 432                if (get_sha1_1(name, namelen, sha1))
 433                        return NULL;
 434                o = parse_object(sha1);
 435        }
 436        while (1) {
 437                if (!o || (!o->parsed && !parse_object(o->sha1)))
 438                        return NULL;
 439                if (o->type == expected_type)
 440                        return o;
 441                if (o->type == OBJ_TAG)
 442                        o = ((struct tag*) o)->tagged;
 443                else if (o->type == OBJ_COMMIT)
 444                        o = &(((struct commit *) o)->tree->object);
 445                else {
 446                        if (name)
 447                                error("%.*s: expected %s type, but the object "
 448                                      "dereferences to %s type",
 449                                      namelen, name, typename(expected_type),
 450                                      typename(o->type));
 451                        return NULL;
 452                }
 453        }
 454}
 455
 456static int peel_onion(const char *name, int len, unsigned char *sha1)
 457{
 458        unsigned char outer[20];
 459        const char *sp;
 460        unsigned int expected_type = 0;
 461        struct object *o;
 462
 463        /*
 464         * "ref^{type}" dereferences ref repeatedly until you cannot
 465         * dereference anymore, or you get an object of given type,
 466         * whichever comes first.  "ref^{}" means just dereference
 467         * tags until you get a non-tag.  "ref^0" is a shorthand for
 468         * "ref^{commit}".  "commit^{tree}" could be used to find the
 469         * top-level tree of the given commit.
 470         */
 471        if (len < 4 || name[len-1] != '}')
 472                return -1;
 473
 474        for (sp = name + len - 1; name <= sp; sp--) {
 475                int ch = *sp;
 476                if (ch == '{' && name < sp && sp[-1] == '^')
 477                        break;
 478        }
 479        if (sp <= name)
 480                return -1;
 481
 482        sp++; /* beginning of type name, or closing brace for empty */
 483        if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
 484                expected_type = OBJ_COMMIT;
 485        else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
 486                expected_type = OBJ_TREE;
 487        else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
 488                expected_type = OBJ_BLOB;
 489        else if (sp[0] == '}')
 490                expected_type = OBJ_NONE;
 491        else
 492                return -1;
 493
 494        if (get_sha1_1(name, sp - name - 2, outer))
 495                return -1;
 496
 497        o = parse_object(outer);
 498        if (!o)
 499                return -1;
 500        if (!expected_type) {
 501                o = deref_tag(o, name, sp - name - 2);
 502                if (!o || (!o->parsed && !parse_object(o->sha1)))
 503                        return -1;
 504                hashcpy(sha1, o->sha1);
 505        }
 506        else {
 507                /*
 508                 * At this point, the syntax look correct, so
 509                 * if we do not get the needed object, we should
 510                 * barf.
 511                 */
 512                o = peel_to_type(name, len, o, expected_type);
 513                if (o) {
 514                        hashcpy(sha1, o->sha1);
 515                        return 0;
 516                }
 517                return -1;
 518        }
 519        return 0;
 520}
 521
 522static int get_describe_name(const char *name, int len, unsigned char *sha1)
 523{
 524        const char *cp;
 525
 526        for (cp = name + len - 1; name + 2 <= cp; cp--) {
 527                char ch = *cp;
 528                if (hexval(ch) & ~0377) {
 529                        /* We must be looking at g in "SOMETHING-g"
 530                         * for it to be describe output.
 531                         */
 532                        if (ch == 'g' && cp[-1] == '-') {
 533                                cp++;
 534                                len -= cp - name;
 535                                return get_short_sha1(cp, len, sha1, 1);
 536                        }
 537                }
 538        }
 539        return -1;
 540}
 541
 542static int get_sha1_1(const char *name, int len, unsigned char *sha1)
 543{
 544        int ret, has_suffix;
 545        const char *cp;
 546
 547        /* "name~3" is "name^^^",
 548         * "name~" and "name~0" are name -- not "name^0"!
 549         * "name^" is not "name^0"; it is "name^1".
 550         */
 551        has_suffix = 0;
 552        for (cp = name + len - 1; name <= cp; cp--) {
 553                int ch = *cp;
 554                if ('0' <= ch && ch <= '9')
 555                        continue;
 556                if (ch == '~' || ch == '^')
 557                        has_suffix = ch;
 558                break;
 559        }
 560
 561        if (has_suffix) {
 562                int num = 0;
 563                int len1 = cp - name;
 564                cp++;
 565                while (cp < name + len)
 566                        num = num * 10 + *cp++ - '0';
 567                if (has_suffix == '^') {
 568                        if (!num && len1 == len - 1)
 569                                num = 1;
 570                        return get_parent(name, len1, sha1, num);
 571                }
 572                /* else if (has_suffix == '~') -- goes without saying */
 573                return get_nth_ancestor(name, len1, sha1, num);
 574        }
 575
 576        ret = peel_onion(name, len, sha1);
 577        if (!ret)
 578                return 0;
 579
 580        ret = get_sha1_basic(name, len, sha1);
 581        if (!ret)
 582                return 0;
 583
 584        /* It could be describe output that is "SOMETHING-gXXXX" */
 585        ret = get_describe_name(name, len, sha1);
 586        if (!ret)
 587                return 0;
 588
 589        return get_short_sha1(name, len, sha1, 0);
 590}
 591
 592static int handle_one_ref(const char *path,
 593                const unsigned char *sha1, int flag, void *cb_data)
 594{
 595        struct commit_list **list = cb_data;
 596        struct object *object = parse_object(sha1);
 597        if (!object)
 598                return 0;
 599        if (object->type == OBJ_TAG) {
 600                object = deref_tag(object, path, strlen(path));
 601                if (!object)
 602                        return 0;
 603        }
 604        if (object->type != OBJ_COMMIT)
 605                return 0;
 606        insert_by_date((struct commit *)object, list);
 607        return 0;
 608}
 609
 610/*
 611 * This interprets names like ':/Initial revision of "git"' by searching
 612 * through history and returning the first commit whose message starts
 613 * with the given string.
 614 *
 615 * For future extension, ':/!' is reserved. If you want to match a message
 616 * beginning with a '!', you have to repeat the exclamation mark.
 617 */
 618
 619#define ONELINE_SEEN (1u<<20)
 620static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
 621{
 622        struct commit_list *list = NULL, *backup = NULL, *l;
 623        int retval = -1;
 624        char *temp_commit_buffer = NULL;
 625
 626        if (prefix[0] == '!') {
 627                if (prefix[1] != '!')
 628                        die ("Invalid search pattern: %s", prefix);
 629                prefix++;
 630        }
 631        for_each_ref(handle_one_ref, &list);
 632        for (l = list; l; l = l->next)
 633                commit_list_insert(l->item, &backup);
 634        while (list) {
 635                char *p;
 636                struct commit *commit;
 637                enum object_type type;
 638                unsigned long size;
 639
 640                commit = pop_most_recent_commit(&list, ONELINE_SEEN);
 641                if (!parse_object(commit->object.sha1))
 642                        continue;
 643                free(temp_commit_buffer);
 644                if (commit->buffer)
 645                        p = commit->buffer;
 646                else {
 647                        p = read_sha1_file(commit->object.sha1, &type, &size);
 648                        if (!p)
 649                                continue;
 650                        temp_commit_buffer = p;
 651                }
 652                if (!(p = strstr(p, "\n\n")))
 653                        continue;
 654                if (!prefixcmp(p + 2, prefix)) {
 655                        hashcpy(sha1, commit->object.sha1);
 656                        retval = 0;
 657                        break;
 658                }
 659        }
 660        free(temp_commit_buffer);
 661        free_commit_list(list);
 662        for (l = backup; l; l = l->next)
 663                clear_commit_marks(l->item, ONELINE_SEEN);
 664        return retval;
 665}
 666
 667/*
 668 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
 669 * notably "xyz^" for "parent of xyz"
 670 */
 671int get_sha1(const char *name, unsigned char *sha1)
 672{
 673        unsigned unused;
 674        return get_sha1_with_mode(name, sha1, &unused);
 675}
 676
 677int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
 678{
 679        int ret, bracket_depth;
 680        int namelen = strlen(name);
 681        const char *cp;
 682
 683        *mode = S_IFINVALID;
 684        ret = get_sha1_1(name, namelen, sha1);
 685        if (!ret)
 686                return ret;
 687        /* sha1:path --> object name of path in ent sha1
 688         * :path -> object name of path in index
 689         * :[0-3]:path -> object name of path in index at stage
 690         */
 691        if (name[0] == ':') {
 692                int stage = 0;
 693                struct cache_entry *ce;
 694                int pos;
 695                if (namelen > 2 && name[1] == '/')
 696                        return get_sha1_oneline(name + 2, sha1);
 697                if (namelen < 3 ||
 698                    name[2] != ':' ||
 699                    name[1] < '0' || '3' < name[1])
 700                        cp = name + 1;
 701                else {
 702                        stage = name[1] - '0';
 703                        cp = name + 3;
 704                }
 705                namelen = namelen - (cp - name);
 706                if (!active_cache)
 707                        read_cache();
 708                pos = cache_name_pos(cp, namelen);
 709                if (pos < 0)
 710                        pos = -pos - 1;
 711                while (pos < active_nr) {
 712                        ce = active_cache[pos];
 713                        if (ce_namelen(ce) != namelen ||
 714                            memcmp(ce->name, cp, namelen))
 715                                break;
 716                        if (ce_stage(ce) == stage) {
 717                                hashcpy(sha1, ce->sha1);
 718                                *mode = ce->ce_mode;
 719                                return 0;
 720                        }
 721                        pos++;
 722                }
 723                return -1;
 724        }
 725        for (cp = name, bracket_depth = 0; *cp; cp++) {
 726                if (*cp == '{')
 727                        bracket_depth++;
 728                else if (bracket_depth && *cp == '}')
 729                        bracket_depth--;
 730                else if (!bracket_depth && *cp == ':')
 731                        break;
 732        }
 733        if (*cp == ':') {
 734                unsigned char tree_sha1[20];
 735                if (!get_sha1_1(name, cp-name, tree_sha1))
 736                        return get_tree_entry(tree_sha1, cp+1, sha1,
 737                                              mode);
 738        }
 739        return ret;
 740}