ref-filter.c

   1 #include "builtin.h"
   2 #include "cache.h"
   3 #include "parse-options.h"
   4 #include "refs.h"
   5 #include "wildmatch.h"
   6 #include "commit.h"
   7 #include "remote.h"
   8 #include "color.h"
   9 #include "tag.h"
  10 #include "quote.h"
  11 #include "ref-filter.h"
  12 #include "revision.h"
  13
  14 typedef enum { FIELD_STR, FIELD_ULONG, FIELD_TIME } cmp_type;
  15
  16 static struct {
  17         const char *name;
  18         cmp_type cmp_type;
  19 } valid_atom[] = {
  20         { "refname" },
  21         { "objecttype" },
  22         { "objectsize", FIELD_ULONG },
  23         { "objectname" },
  24         { "tree" },
  25         { "parent" },
  26         { "numparent", FIELD_ULONG },
  27         { "object" },
  28         { "type" },
  29         { "tag" },
  30         { "author" },
  31         { "authorname" },
  32         { "authoremail" },
  33         { "authordate", FIELD_TIME },
  34         { "committer" },
  35         { "committername" },
  36         { "committeremail" },
  37         { "committerdate", FIELD_TIME },
  38         { "tagger" },
  39         { "taggername" },
  40         { "taggeremail" },
  41         { "taggerdate", FIELD_TIME },
  42         { "creator" },
  43         { "creatordate", FIELD_TIME },
  44         { "subject" },
  45         { "body" },
  46         { "contents" },
  47         { "contents:subject" },
  48         { "contents:body" },
  49         { "contents:signature" },
  50         { "upstream" },
  51         { "push" },
  52         { "symref" },
  53         { "flag" },
  54         { "HEAD" },
  55         { "color" },
  56 };
  57
  58 /*
  59  * An atom is a valid field atom listed above, possibly prefixed with
  60  * a "*" to denote deref_tag().
  61  *
  62  * We parse given format string and sort specifiers, and make a list
  63  * of properties that we need to extract out of objects.  ref_array_item
  64  * structure will hold an array of values extracted that can be
  65  * indexed with the "atom number", which is an index into this
  66  * array.
  67  */
  68 static const char **used_atom;
  69 static cmp_type *used_atom_type;
  70 static int used_atom_cnt, need_tagged, need_symref;
  71 static int need_color_reset_at_eol;
  72
  73 /*
  74  * Used to parse format string and sort specifiers
  75  */
  76 int parse_ref_filter_atom(const char *atom, const char *ep)
  77 {
  78         const char *sp;
  79         int i, at;
  80
  81         sp = atom;
  82         if (*sp == '*' && sp < ep)
  83                 sp++; /* deref */
  84         if (ep <= sp)
  85                 die("malformed field name: %.*s", (int)(ep-atom), atom);
  86
  87         /* Do we have the atom already used elsewhere? */
  88         for (i = 0; i < used_atom_cnt; i++) {
  89                 int len = strlen(used_atom[i]);
  90                 if (len == ep - atom && !memcmp(used_atom[i], atom, len))
  91                         return i;
  92         }
  93
  94         /* Is the atom a valid one? */
  95         for (i = 0; i < ARRAY_SIZE(valid_atom); i++) {
  96                 int len = strlen(valid_atom[i].name);
  97                 /*
  98                  * If the atom name has a colon, strip it and everything after
  99                  * it off - it specifies the format for this entry, and
 100                  * shouldn't be used for checking against the valid_atom
 101                  * table.
 102                  */
 103                 const char *formatp = strchr(sp, ':');
 104                 if (!formatp || ep < formatp)
 105                         formatp = ep;
 106                 if (len == formatp - sp && !memcmp(valid_atom[i].name, sp, len))
 107                         break;
 108         }
 109
 110         if (ARRAY_SIZE(valid_atom) <= i)
 111                 die("unknown field name: %.*s", (int)(ep-atom), atom);
 112
 113         /* Add it in, including the deref prefix */
 114         at = used_atom_cnt;
 115         used_atom_cnt++;
 116         REALLOC_ARRAY(used_atom, used_atom_cnt);
 117         REALLOC_ARRAY(used_atom_type, used_atom_cnt);
 118         used_atom[at] = xmemdupz(atom, ep - atom);
 119         used_atom_type[at] = valid_atom[i].cmp_type;
 120         if (*atom == '*')
 121                 need_tagged = 1;
 122         if (!strcmp(used_atom[at], "symref"))
 123                 need_symref = 1;
 124         return at;
 125 }
 126
 127 /*
 128  * In a format string, find the next occurrence of %(atom).
 129  */
 130 static const char *find_next(const char *cp)
 131 {
 132         while (*cp) {
 133                 if (*cp == '%') {
 134                         /*
 135                          * %( is the start of an atom;
 136                          * %% is a quoted per-cent.
 137                          */
 138                         if (cp[1] == '(')
 139                                 return cp;
 140                         else if (cp[1] == '%')
 141                                 cp++; /* skip over two % */
 142                         /* otherwise this is a singleton, literal % */
 143                 }
 144                 cp++;
 145         }
 146         return NULL;
 147 }
 148
 149 /*
 150  * Make sure the format string is well formed, and parse out
 151  * the used atoms.
 152  */
 153 int verify_ref_format(const char *format)
 154 {
 155         const char *cp, *sp;
 156
 157         need_color_reset_at_eol = 0;
 158         for (cp = format; *cp && (sp = find_next(cp)); ) {
 159                 const char *color, *ep = strchr(sp, ')');
 160                 int at;
 161
 162                 if (!ep)
 163                         return error("malformed format string %s", sp);
 164                 /* sp points at "%(" and ep points at the closing ")" */
 165                 at = parse_ref_filter_atom(sp + 2, ep);
 166                 cp = ep + 1;
 167
 168                 if (skip_prefix(used_atom[at], "color:", &color))
 169                         need_color_reset_at_eol = !!strcmp(color, "reset");
 170         }
 171         return 0;
 172 }
 173
 174 /*
 175  * Given an object name, read the object data and size, and return a
 176  * "struct object".  If the object data we are returning is also borrowed
 177  * by the "struct object" representation, set *eaten as well---it is a
 178  * signal from parse_object_buffer to us not to free the buffer.
 179  */
 180 static void *get_obj(const unsigned char *sha1, struct object **obj, unsigned long *sz, int *eaten)
 181 {
 182         enum object_type type;
 183         void *buf = read_sha1_file(sha1, &type, sz);
 184
 185         if (buf)
 186                 *obj = parse_object_buffer(sha1, type, *sz, buf, eaten);
 187         else
 188                 *obj = NULL;
 189         return buf;
 190 }
 191
 192 static int grab_objectname(const char *name, const unsigned char *sha1,
 193                             struct atom_value *v)
 194 {
 195         if (!strcmp(name, "objectname")) {
 196                 char *s = xmalloc(41);
 197                 strcpy(s, sha1_to_hex(sha1));
 198                 v->s = s;
 199                 return 1;
 200         }
 201         if (!strcmp(name, "objectname:short")) {
 202                 v->s = xstrdup(find_unique_abbrev(sha1, DEFAULT_ABBREV));
 203                 return 1;
 204         }
 205         return 0;
 206 }
 207
 208 /* See grab_values */
 209 static void grab_common_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
 210 {
 211         int i;
 212
 213         for (i = 0; i < used_atom_cnt; i++) {
 214                 const char *name = used_atom[i];
 215                 struct atom_value *v = &val[i];
 216                 if (!!deref != (*name == '*'))
 217                         continue;
 218                 if (deref)
 219                         name++;
 220                 if (!strcmp(name, "objecttype"))
 221                         v->s = typename(obj->type);
 222                 else if (!strcmp(name, "objectsize")) {
 223                         char *s = xmalloc(40);
 224                         sprintf(s, "%lu", sz);
 225                         v->ul = sz;
 226                         v->s = s;
 227                 }
 228                 else if (deref)
 229                         grab_objectname(name, obj->sha1, v);
 230         }
 231 }
 232
 233 /* See grab_values */
 234 static void grab_tag_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
 235 {
 236         int i;
 237         struct tag *tag = (struct tag *) obj;
 238
 239         for (i = 0; i < used_atom_cnt; i++) {
 240                 const char *name = used_atom[i];
 241                 struct atom_value *v = &val[i];
 242                 if (!!deref != (*name == '*'))
 243                         continue;
 244                 if (deref)
 245                         name++;
 246                 if (!strcmp(name, "tag"))
 247                         v->s = tag->tag;
 248                 else if (!strcmp(name, "type") && tag->tagged)
 249                         v->s = typename(tag->tagged->type);
 250                 else if (!strcmp(name, "object") && tag->tagged) {
 251                         char *s = xmalloc(41);
 252                         strcpy(s, sha1_to_hex(tag->tagged->sha1));
 253                         v->s = s;
 254                 }
 255         }
 256 }
 257
 258 /* See grab_values */
 259 static void grab_commit_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
 260 {
 261         int i;
 262         struct commit *commit = (struct commit *) obj;
 263
 264         for (i = 0; i < used_atom_cnt; i++) {
 265                 const char *name = used_atom[i];
 266                 struct atom_value *v = &val[i];
 267                 if (!!deref != (*name == '*'))
 268                         continue;
 269                 if (deref)
 270                         name++;
 271                 if (!strcmp(name, "tree")) {
 272                         char *s = xmalloc(41);
 273                         strcpy(s, sha1_to_hex(commit->tree->object.sha1));
 274                         v->s = s;
 275                 }
 276                 if (!strcmp(name, "numparent")) {
 277                         char *s = xmalloc(40);
 278                         v->ul = commit_list_count(commit->parents);
 279                         sprintf(s, "%lu", v->ul);
 280                         v->s = s;
 281                 }
 282                 else if (!strcmp(name, "parent")) {
 283                         int num = commit_list_count(commit->parents);
 284                         int i;
 285                         struct commit_list *parents;
 286                         char *s = xmalloc(41 * num + 1);
 287                         v->s = s;
 288                         for (i = 0, parents = commit->parents;
 289                              parents;
 290                              parents = parents->next, i = i + 41) {
 291                                 struct commit *parent = parents->item;
 292                                 strcpy(s+i, sha1_to_hex(parent->object.sha1));
 293                                 if (parents->next)
 294                                         s[i+40] = ' ';
 295                         }
 296                         if (!i)
 297                                 *s = '\0';
 298                 }
 299         }
 300 }
 301
 302 static const char *find_wholine(const char *who, int wholen, const char *buf, unsigned long sz)
 303 {
 304         const char *eol;
 305         while (*buf) {
 306                 if (!strncmp(buf, who, wholen) &&
 307                     buf[wholen] == ' ')
 308                         return buf + wholen + 1;
 309                 eol = strchr(buf, '\n');
 310                 if (!eol)
 311                         return "";
 312                 eol++;
 313                 if (*eol == '\n')
 314                         return ""; /* end of header */
 315                 buf = eol;
 316         }
 317         return "";
 318 }
 319
 320 static const char *copy_line(const char *buf)
 321 {
 322         const char *eol = strchrnul(buf, '\n');
 323         return xmemdupz(buf, eol - buf);
 324 }
 325
 326 static const char *copy_name(const char *buf)
 327 {
 328         const char *cp;
 329         for (cp = buf; *cp && *cp != '\n'; cp++) {
 330                 if (!strncmp(cp, " <", 2))
 331                         return xmemdupz(buf, cp - buf);
 332         }
 333         return "";
 334 }
 335
 336 static const char *copy_email(const char *buf)
 337 {
 338         const char *email = strchr(buf, '<');
 339         const char *eoemail;
 340         if (!email)
 341                 return "";
 342         eoemail = strchr(email, '>');
 343         if (!eoemail)
 344                 return "";
 345         return xmemdupz(email, eoemail + 1 - email);
 346 }
 347
 348 static char *copy_subject(const char *buf, unsigned long len)
 349 {
 350         char *r = xmemdupz(buf, len);
 351         int i;
 352
 353         for (i = 0; i < len; i++)
 354                 if (r[i] == '\n')
 355                         r[i] = ' ';
 356
 357         return r;
 358 }
 359
 360 static void grab_date(const char *buf, struct atom_value *v, const char *atomname)
 361 {
 362         const char *eoemail = strstr(buf, "> ");
 363         char *zone;
 364         unsigned long timestamp;
 365         long tz;
 366         enum date_mode date_mode = DATE_NORMAL;
 367         const char *formatp;
 368
 369         /*
 370          * We got here because atomname ends in "date" or "date<something>";
 371          * it's not possible that <something> is not ":<format>" because
 372          * parse_ref_filter_atom() wouldn't have allowed it, so we can assume that no
 373          * ":" means no format is specified, and use the default.
 374          */
 375         formatp = strchr(atomname, ':');
 376         if (formatp != NULL) {
 377                 formatp++;
 378                 date_mode = parse_date_format(formatp);
 379         }
 380
 381         if (!eoemail)
 382                 goto bad;
 383         timestamp = strtoul(eoemail + 2, &zone, 10);
 384         if (timestamp == ULONG_MAX)
 385                 goto bad;
 386         tz = strtol(zone, NULL, 10);
 387         if ((tz == LONG_MIN || tz == LONG_MAX) && errno == ERANGE)
 388                 goto bad;
 389         v->s = xstrdup(show_date(timestamp, tz, date_mode));
 390         v->ul = timestamp;
 391         return;
 392  bad:
 393         v->s = "";
 394         v->ul = 0;
 395 }
 396
 397 /* See grab_values */
 398 static void grab_person(const char *who, struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
 399 {
 400         int i;
 401         int wholen = strlen(who);
 402         const char *wholine = NULL;
 403
 404         for (i = 0; i < used_atom_cnt; i++) {
 405                 const char *name = used_atom[i];
 406                 struct atom_value *v = &val[i];
 407                 if (!!deref != (*name == '*'))
 408                         continue;
 409                 if (deref)
 410                         name++;
 411                 if (strncmp(who, name, wholen))
 412                         continue;
 413                 if (name[wholen] != 0 &&
 414                     strcmp(name + wholen, "name") &&
 415                     strcmp(name + wholen, "email") &&
 416                     !starts_with(name + wholen, "date"))
 417                         continue;
 418                 if (!wholine)
 419                         wholine = find_wholine(who, wholen, buf, sz);
 420                 if (!wholine)
 421                         return; /* no point looking for it */
 422                 if (name[wholen] == 0)
 423                         v->s = copy_line(wholine);
 424                 else if (!strcmp(name + wholen, "name"))
 425                         v->s = copy_name(wholine);
 426                 else if (!strcmp(name + wholen, "email"))
 427                         v->s = copy_email(wholine);
 428                 else if (starts_with(name + wholen, "date"))
 429                         grab_date(wholine, v, name);
 430         }
 431
 432         /*
 433          * For a tag or a commit object, if "creator" or "creatordate" is
 434          * requested, do something special.
 435          */
 436         if (strcmp(who, "tagger") && strcmp(who, "committer"))
 437                 return; /* "author" for commit object is not wanted */
 438         if (!wholine)
 439                 wholine = find_wholine(who, wholen, buf, sz);
 440         if (!wholine)
 441                 return;
 442         for (i = 0; i < used_atom_cnt; i++) {
 443                 const char *name = used_atom[i];
 444                 struct atom_value *v = &val[i];
 445                 if (!!deref != (*name == '*'))
 446                         continue;
 447                 if (deref)
 448                         name++;
 449
 450                 if (starts_with(name, "creatordate"))
 451                         grab_date(wholine, v, name);
 452                 else if (!strcmp(name, "creator"))
 453                         v->s = copy_line(wholine);
 454         }
 455 }
 456
 457 static void find_subpos(const char *buf, unsigned long sz,
 458                         const char **sub, unsigned long *sublen,
 459                         const char **body, unsigned long *bodylen,
 460                         unsigned long *nonsiglen,
 461                         const char **sig, unsigned long *siglen)
 462 {
 463         const char *eol;
 464         /* skip past header until we hit empty line */
 465         while (*buf && *buf != '\n') {
 466                 eol = strchrnul(buf, '\n');
 467                 if (*eol)
 468                         eol++;
 469                 buf = eol;
 470         }
 471         /* skip any empty lines */
 472         while (*buf == '\n')
 473                 buf++;
 474
 475         /* parse signature first; we might not even have a subject line */
 476         *sig = buf + parse_signature(buf, strlen(buf));
 477         *siglen = strlen(*sig);
 478
 479         /* subject is first non-empty line */
 480         *sub = buf;
 481         /* subject goes to first empty line */
 482         while (buf < *sig && *buf && *buf != '\n') {
 483                 eol = strchrnul(buf, '\n');
 484                 if (*eol)
 485                         eol++;
 486                 buf = eol;
 487         }
 488         *sublen = buf - *sub;
 489         /* drop trailing newline, if present */
 490         if (*sublen && (*sub)[*sublen - 1] == '\n')
 491                 *sublen -= 1;
 492
 493         /* skip any empty lines */
 494         while (*buf == '\n')
 495                 buf++;
 496         *body = buf;
 497         *bodylen = strlen(buf);
 498         *nonsiglen = *sig - buf;
 499 }
 500
 501 /* See grab_values */
 502 static void grab_sub_body_contents(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
 503 {
 504         int i;
 505         const char *subpos = NULL, *bodypos = NULL, *sigpos = NULL;
 506         unsigned long sublen = 0, bodylen = 0, nonsiglen = 0, siglen = 0;
 507
 508         for (i = 0; i < used_atom_cnt; i++) {
 509                 const char *name = used_atom[i];
 510                 struct atom_value *v = &val[i];
 511                 if (!!deref != (*name == '*'))
 512                         continue;
 513                 if (deref)
 514                         name++;
 515                 if (strcmp(name, "subject") &&
 516                     strcmp(name, "body") &&
 517                     strcmp(name, "contents") &&
 518                     strcmp(name, "contents:subject") &&
 519                     strcmp(name, "contents:body") &&
 520                     strcmp(name, "contents:signature"))
 521                         continue;
 522                 if (!subpos)
 523                         find_subpos(buf, sz,
 524                                     &subpos, &sublen,
 525                                     &bodypos, &bodylen, &nonsiglen,
 526                                     &sigpos, &siglen);
 527
 528                 if (!strcmp(name, "subject"))
 529                         v->s = copy_subject(subpos, sublen);
 530                 else if (!strcmp(name, "contents:subject"))
 531                         v->s = copy_subject(subpos, sublen);
 532                 else if (!strcmp(name, "body"))
 533                         v->s = xmemdupz(bodypos, bodylen);
 534                 else if (!strcmp(name, "contents:body"))
 535                         v->s = xmemdupz(bodypos, nonsiglen);
 536                 else if (!strcmp(name, "contents:signature"))
 537                         v->s = xmemdupz(sigpos, siglen);
 538                 else if (!strcmp(name, "contents"))
 539                         v->s = xstrdup(subpos);
 540         }
 541 }
 542
 543 /*
 544  * We want to have empty print-string for field requests
 545  * that do not apply (e.g. "authordate" for a tag object)
 546  */
 547 static void fill_missing_values(struct atom_value *val)
 548 {
 549         int i;
 550         for (i = 0; i < used_atom_cnt; i++) {
 551                 struct atom_value *v = &val[i];
 552                 if (v->s == NULL)
 553                         v->s = "";
 554         }
 555 }
 556
 557 /*
 558  * val is a list of atom_value to hold returned values.  Extract
 559  * the values for atoms in used_atom array out of (obj, buf, sz).
 560  * when deref is false, (obj, buf, sz) is the object that is
 561  * pointed at by the ref itself; otherwise it is the object the
 562  * ref (which is a tag) refers to.
 563  */
 564 static void grab_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
 565 {
 566         grab_common_values(val, deref, obj, buf, sz);
 567         switch (obj->type) {
 568         case OBJ_TAG:
 569                 grab_tag_values(val, deref, obj, buf, sz);
 570                 grab_sub_body_contents(val, deref, obj, buf, sz);
 571                 grab_person("tagger", val, deref, obj, buf, sz);
 572                 break;
 573         case OBJ_COMMIT:
 574                 grab_commit_values(val, deref, obj, buf, sz);
 575                 grab_sub_body_contents(val, deref, obj, buf, sz);
 576                 grab_person("author", val, deref, obj, buf, sz);
 577                 grab_person("committer", val, deref, obj, buf, sz);
 578                 break;
 579         case OBJ_TREE:
 580                 /* grab_tree_values(val, deref, obj, buf, sz); */
 581                 break;
 582         case OBJ_BLOB:
 583                 /* grab_blob_values(val, deref, obj, buf, sz); */
 584                 break;
 585         default:
 586                 die("Eh?  Object of type %d?", obj->type);
 587         }
 588 }
 589
 590 static inline char *copy_advance(char *dst, const char *src)
 591 {
 592         while (*src)
 593                 *dst++ = *src++;
 594         return dst;
 595 }
 596
 597 /*
 598  * Parse the object referred by ref, and grab needed value.
 599  */
 600 static void populate_value(struct ref_array_item *ref)
 601 {
 602         void *buf;
 603         struct object *obj;
 604         int eaten, i;
 605         unsigned long size;
 606         const unsigned char *tagged;
 607
 608         ref->value = xcalloc(used_atom_cnt, sizeof(struct atom_value));
 609
 610         if (need_symref && (ref->flag & REF_ISSYMREF) && !ref->symref) {
 611                 unsigned char unused1[20];
 612                 ref->symref = resolve_refdup(ref->refname, RESOLVE_REF_READING,
 613                                              unused1, NULL);
 614                 if (!ref->symref)
 615                         ref->symref = "";
 616         }
 617
 618         /* Fill in specials first */
 619         for (i = 0; i < used_atom_cnt; i++) {
 620                 const char *name = used_atom[i];
 621                 struct atom_value *v = &ref->value[i];
 622                 int deref = 0;
 623                 const char *refname;
 624                 const char *formatp;
 625                 struct branch *branch = NULL;
 626
 627                 if (*name == '*') {
 628                         deref = 1;
 629                         name++;
 630                 }
 631
 632                 if (starts_with(name, "refname"))
 633                         refname = ref->refname;
 634                 else if (starts_with(name, "symref"))
 635                         refname = ref->symref ? ref->symref : "";
 636                 else if (starts_with(name, "upstream")) {
 637                         const char *branch_name;
 638                         /* only local branches may have an upstream */
 639                         if (!skip_prefix(ref->refname, "refs/heads/",
 640                                          &branch_name))
 641                                 continue;
 642                         branch = branch_get(branch_name);
 643
 644                         refname = branch_get_upstream(branch, NULL);
 645                         if (!refname)
 646                                 continue;
 647                 } else if (starts_with(name, "push")) {
 648                         const char *branch_name;
 649                         if (!skip_prefix(ref->refname, "refs/heads/",
 650                                          &branch_name))
 651                                 continue;
 652                         branch = branch_get(branch_name);
 653
 654                         refname = branch_get_push(branch, NULL);
 655                         if (!refname)
 656                                 continue;
 657                 } else if (starts_with(name, "color:")) {
 658                         char color[COLOR_MAXLEN] = "";
 659
 660                         if (color_parse(name + 6, color) < 0)
 661                                 die(_("unable to parse format"));
 662                         v->s = xstrdup(color);
 663                         continue;
 664                 } else if (!strcmp(name, "flag")) {
 665                         char buf[256], *cp = buf;
 666                         if (ref->flag & REF_ISSYMREF)
 667                                 cp = copy_advance(cp, ",symref");
 668                         if (ref->flag & REF_ISPACKED)
 669                                 cp = copy_advance(cp, ",packed");
 670                         if (cp == buf)
 671                                 v->s = "";
 672                         else {
 673                                 *cp = '\0';
 674                                 v->s = xstrdup(buf + 1);
 675                         }
 676                         continue;
 677                 } else if (!deref && grab_objectname(name, ref->objectname, v)) {
 678                         continue;
 679                 } else if (!strcmp(name, "HEAD")) {
 680                         const char *head;
 681                         unsigned char sha1[20];
 682
 683                         head = resolve_ref_unsafe("HEAD", RESOLVE_REF_READING,
 684                                                   sha1, NULL);
 685                         if (!strcmp(ref->refname, head))
 686                                 v->s = "*";
 687                         else
 688                                 v->s = " ";
 689                         continue;
 690                 } else
 691                         continue;
 692
 693                 formatp = strchr(name, ':');
 694                 if (formatp) {
 695                         int num_ours, num_theirs;
 696
 697                         formatp++;
 698                         if (!strcmp(formatp, "short"))
 699                                 refname = shorten_unambiguous_ref(refname,
 700                                                       warn_ambiguous_refs);
 701                         else if (!strcmp(formatp, "track") &&
 702                                  (starts_with(name, "upstream") ||
 703                                   starts_with(name, "push"))) {
 704                                 char buf[40];
 705
 706                                 if (stat_tracking_info(branch, &num_ours,
 707                                                        &num_theirs, NULL))
 708                                         continue;
 709
 710                                 if (!num_ours && !num_theirs)
 711                                         v->s = "";
 712                                 else if (!num_ours) {
 713                                         sprintf(buf, "[behind %d]", num_theirs);
 714                                         v->s = xstrdup(buf);
 715                                 } else if (!num_theirs) {
 716                                         sprintf(buf, "[ahead %d]", num_ours);
 717                                         v->s = xstrdup(buf);
 718                                 } else {
 719                                         sprintf(buf, "[ahead %d, behind %d]",
 720                                                 num_ours, num_theirs);
 721                                         v->s = xstrdup(buf);
 722                                 }
 723                                 continue;
 724                         } else if (!strcmp(formatp, "trackshort") &&
 725                                    (starts_with(name, "upstream") ||
 726                                     starts_with(name, "push"))) {
 727                                 assert(branch);
 728
 729                                 if (stat_tracking_info(branch, &num_ours,
 730                                                         &num_theirs, NULL))
 731                                         continue;
 732
 733                                 if (!num_ours && !num_theirs)
 734                                         v->s = "=";
 735                                 else if (!num_ours)
 736                                         v->s = "<";
 737                                 else if (!num_theirs)
 738                                         v->s = ">";
 739                                 else
 740                                         v->s = "<>";
 741                                 continue;
 742                         } else
 743                                 die("unknown %.*s format %s",
 744                                     (int)(formatp - name), name, formatp);
 745                 }
 746
 747                 if (!deref)
 748                         v->s = refname;
 749                 else {
 750                         int len = strlen(refname);
 751                         char *s = xmalloc(len + 4);
 752                         sprintf(s, "%s^{}", refname);
 753                         v->s = s;
 754                 }
 755         }
 756
 757         for (i = 0; i < used_atom_cnt; i++) {
 758                 struct atom_value *v = &ref->value[i];
 759                 if (v->s == NULL)
 760                         goto need_obj;
 761         }
 762         return;
 763
 764  need_obj:
 765         buf = get_obj(ref->objectname, &obj, &size, &eaten);
 766         if (!buf)
 767                 die("missing object %s for %s",
 768                     sha1_to_hex(ref->objectname), ref->refname);
 769         if (!obj)
 770                 die("parse_object_buffer failed on %s for %s",
 771                     sha1_to_hex(ref->objectname), ref->refname);
 772
 773         grab_values(ref->value, 0, obj, buf, size);
 774         if (!eaten)
 775                 free(buf);
 776
 777         /*
 778          * If there is no atom that wants to know about tagged
 779          * object, we are done.
 780          */
 781         if (!need_tagged || (obj->type != OBJ_TAG))
 782                 return;
 783
 784         /*
 785          * If it is a tag object, see if we use a value that derefs
 786          * the object, and if we do grab the object it refers to.
 787          */
 788         tagged = ((struct tag *)obj)->tagged->sha1;
 789
 790         /*
 791          * NEEDSWORK: This derefs tag only once, which
 792          * is good to deal with chains of trust, but
 793          * is not consistent with what deref_tag() does
 794          * which peels the onion to the core.
 795          */
 796         buf = get_obj(tagged, &obj, &size, &eaten);
 797         if (!buf)
 798                 die("missing object %s for %s",
 799                     sha1_to_hex(tagged), ref->refname);
 800         if (!obj)
 801                 die("parse_object_buffer failed on %s for %s",
 802                     sha1_to_hex(tagged), ref->refname);
 803         grab_values(ref->value, 1, obj, buf, size);
 804         if (!eaten)
 805                 free(buf);
 806 }
 807
 808 /*
 809  * Given a ref, return the value for the atom.  This lazily gets value
 810  * out of the object by calling populate value.
 811  */
 812 static void get_ref_atom_value(struct ref_array_item *ref, int atom, struct atom_value **v)
 813 {
 814         if (!ref->value) {
 815                 populate_value(ref);
 816                 fill_missing_values(ref->value);
 817         }
 818         *v = &ref->value[atom];
 819 }
 820
 821 enum contains_result {
 822         CONTAINS_UNKNOWN = -1,
 823         CONTAINS_NO = 0,
 824         CONTAINS_YES = 1
 825 };
 826
 827 /*
 828  * Mimicking the real stack, this stack lives on the heap, avoiding stack
 829  * overflows.
 830  *
 831  * At each recursion step, the stack items points to the commits whose
 832  * ancestors are to be inspected.
 833  */
 834 struct contains_stack {
 835         int nr, alloc;
 836         struct contains_stack_entry {
 837                 struct commit *commit;
 838                 struct commit_list *parents;
 839         } *contains_stack;
 840 };
 841
 842 static int in_commit_list(const struct commit_list *want, struct commit *c)
 843 {
 844         for (; want; want = want->next)
 845                 if (!hashcmp(want->item->object.sha1, c->object.sha1))
 846                         return 1;
 847         return 0;
 848 }
 849
 850 /*
 851  * Test whether the candidate or one of its parents is contained in the list.
 852  * Do not recurse to find out, though, but return -1 if inconclusive.
 853  */
 854 static enum contains_result contains_test(struct commit *candidate,
 855                             const struct commit_list *want)
 856 {
 857         /* was it previously marked as containing a want commit? */
 858         if (candidate->object.flags & TMP_MARK)
 859                 return 1;
 860         /* or marked as not possibly containing a want commit? */
 861         if (candidate->object.flags & UNINTERESTING)
 862                 return 0;
 863         /* or are we it? */
 864         if (in_commit_list(want, candidate)) {
 865                 candidate->object.flags |= TMP_MARK;
 866                 return 1;
 867         }
 868
 869         if (parse_commit(candidate) < 0)
 870                 return 0;
 871
 872         return -1;
 873 }
 874
 875 static void push_to_contains_stack(struct commit *candidate, struct contains_stack *contains_stack)
 876 {
 877         ALLOC_GROW(contains_stack->contains_stack, contains_stack->nr + 1, contains_stack->alloc);
 878         contains_stack->contains_stack[contains_stack->nr].commit = candidate;
 879         contains_stack->contains_stack[contains_stack->nr++].parents = candidate->parents;
 880 }
 881
 882 static enum contains_result contains_tag_algo(struct commit *candidate,
 883                 const struct commit_list *want)
 884 {
 885         struct contains_stack contains_stack = { 0, 0, NULL };
 886         int result = contains_test(candidate, want);
 887
 888         if (result != CONTAINS_UNKNOWN)
 889                 return result;
 890
 891         push_to_contains_stack(candidate, &contains_stack);
 892         while (contains_stack.nr) {
 893                 struct contains_stack_entry *entry = &contains_stack.contains_stack[contains_stack.nr - 1];
 894                 struct commit *commit = entry->commit;
 895                 struct commit_list *parents = entry->parents;
 896
 897                 if (!parents) {
 898                         commit->object.flags |= UNINTERESTING;
 899                         contains_stack.nr--;
 900                 }
 901                 /*
 902                  * If we just popped the stack, parents->item has been marked,
 903                  * therefore contains_test will return a meaningful 0 or 1.
 904                  */
 905                 else switch (contains_test(parents->item, want)) {
 906                 case CONTAINS_YES:
 907                         commit->object.flags |= TMP_MARK;
 908                         contains_stack.nr--;
 909                         break;
 910                 case CONTAINS_NO:
 911                         entry->parents = parents->next;
 912                         break;
 913                 case CONTAINS_UNKNOWN:
 914                         push_to_contains_stack(parents->item, &contains_stack);
 915                         break;
 916                 }
 917         }
 918         free(contains_stack.contains_stack);
 919         return contains_test(candidate, want);
 920 }
 921
 922 static int commit_contains(struct ref_filter *filter, struct commit *commit)
 923 {
 924         if (filter->with_commit_tag_algo)
 925                 return contains_tag_algo(commit, filter->with_commit);
 926         return is_descendant_of(commit, filter->with_commit);
 927 }
 928
 929 /*
 930  * Return 1 if the refname matches one of the patterns, otherwise 0.
 931  * A pattern can be path prefix (e.g. a refname "refs/heads/master"
 932  * matches a pattern "refs/heads/") or a wildcard (e.g. the same ref
 933  * matches "refs/heads/m*",too).
 934  */
 935 static int match_name_as_path(const char **pattern, const char *refname)
 936 {
 937         int namelen = strlen(refname);
 938         for (; *pattern; pattern++) {
 939                 const char *p = *pattern;
 940                 int plen = strlen(p);
 941
 942                 if ((plen <= namelen) &&
 943                     !strncmp(refname, p, plen) &&
 944                     (refname[plen] == '\0' ||
 945                      refname[plen] == '/' ||
 946                      p[plen-1] == '/'))
 947                         return 1;
 948                 if (!wildmatch(p, refname, WM_PATHNAME, NULL))
 949                         return 1;
 950         }
 951         return 0;
 952 }
 953
 954 /*
 955  * Given a ref (sha1, refname), check if the ref belongs to the array
 956  * of sha1s. If the given ref is a tag, check if the given tag points
 957  * at one of the sha1s in the given sha1 array.
 958  * the given sha1_array.
 959  * NEEDSWORK:
 960  * 1. Only a single level of inderection is obtained, we might want to
 961  * change this to account for multiple levels (e.g. annotated tags
 962  * pointing to annotated tags pointing to a commit.)
 963  * 2. As the refs are cached we might know what refname peels to without
 964  * the need to parse the object via parse_object(). peel_ref() might be a
 965  * more efficient alternative to obtain the pointee.
 966  */
 967 static const unsigned char *match_points_at(struct sha1_array *points_at,
 968                                             const unsigned char *sha1,
 969                                             const char *refname)
 970 {
 971         const unsigned char *tagged_sha1 = NULL;
 972         struct object *obj;
 973
 974         if (sha1_array_lookup(points_at, sha1) >= 0)
 975                 return sha1;
 976         obj = parse_object(sha1);
 977         if (!obj)
 978                 die(_("malformed object at '%s'"), refname);
 979         if (obj->type == OBJ_TAG)
 980                 tagged_sha1 = ((struct tag *)obj)->tagged->sha1;
 981         if (tagged_sha1 && sha1_array_lookup(points_at, tagged_sha1) >= 0)
 982                 return tagged_sha1;
 983         return NULL;
 984 }
 985
 986 /* Allocate space for a new ref_array_item and copy the objectname and flag to it */
 987 static struct ref_array_item *new_ref_array_item(const char *refname,
 988                                                  const unsigned char *objectname,
 989                                                  int flag)
 990 {
 991         size_t len = strlen(refname);
 992         struct ref_array_item *ref = xcalloc(1, sizeof(struct ref_array_item) + len + 1);
 993         memcpy(ref->refname, refname, len);
 994         ref->refname[len] = '\0';
 995         hashcpy(ref->objectname, objectname);
 996         ref->flag = flag;
 997
 998         return ref;
 999 }
1000
1001 /*
1002  * A call-back given to for_each_ref().  Filter refs and keep them for
1003  * later object processing.
1004  */
1005 static int ref_filter_handler(const char *refname, const struct object_id *oid, int flag, void *cb_data)
1006 {
1007         struct ref_filter_cbdata *ref_cbdata = cb_data;
1008         struct ref_filter *filter = ref_cbdata->filter;
1009         struct ref_array_item *ref;
1010         struct commit *commit = NULL;
1011
1012         if (flag & REF_BAD_NAME) {
1013                 warning("ignoring ref with broken name %s", refname);
1014                 return 0;
1015         }
1016
1017         if (*filter->name_patterns && !match_name_as_path(filter->name_patterns, refname))
1018                 return 0;
1019
1020         if (filter->points_at.nr && !match_points_at(&filter->points_at, oid->hash, refname))
1021                 return 0;
1022
1023         /*
1024          * A merge filter is applied on refs pointing to commits. Hence
1025          * obtain the commit using the 'oid' available and discard all
1026          * non-commits early. The actual filtering is done later.
1027          */
1028         if (filter->merge_commit || filter->with_commit) {
1029                 commit = lookup_commit_reference_gently(oid->hash, 1);
1030                 if (!commit)
1031                         return 0;
1032                 /* We perform the filtering for the '--contains' option */
1033                 if (filter->with_commit &&
1034                     !commit_contains(filter, commit))
1035                         return 0;
1036         }
1037
1038         /*
1039          * We do not open the object yet; sort may only need refname
1040          * to do its job and the resulting list may yet to be pruned
1041          * by maxcount logic.
1042          */
1043         ref = new_ref_array_item(refname, oid->hash, flag);
1044         ref->commit = commit;
1045
1046         REALLOC_ARRAY(ref_cbdata->array->items, ref_cbdata->array->nr + 1);
1047         ref_cbdata->array->items[ref_cbdata->array->nr++] = ref;
1048         return 0;
1049 }
1050
1051 /*  Free memory allocated for a ref_array_item */
1052 static void free_array_item(struct ref_array_item *item)
1053 {
1054         free((char *)item->symref);
1055         free(item);
1056 }
1057
1058 /* Free all memory allocated for ref_array */
1059 void ref_array_clear(struct ref_array *array)
1060 {
1061         int i;
1062
1063         for (i = 0; i < array->nr; i++)
1064                 free_array_item(array->items[i]);
1065         free(array->items);
1066         array->items = NULL;
1067         array->nr = array->alloc = 0;
1068 }
1069
1070 static void do_merge_filter(struct ref_filter_cbdata *ref_cbdata)
1071 {
1072         struct rev_info revs;
1073         int i, old_nr;
1074         struct ref_filter *filter = ref_cbdata->filter;
1075         struct ref_array *array = ref_cbdata->array;
1076         struct commit **to_clear = xcalloc(sizeof(struct commit *), array->nr);
1077
1078         init_revisions(&revs, NULL);
1079
1080         for (i = 0; i < array->nr; i++) {
1081                 struct ref_array_item *item = array->items[i];
1082                 add_pending_object(&revs, &item->commit->object, item->refname);
1083                 to_clear[i] = item->commit;
1084         }
1085
1086         filter->merge_commit->object.flags |= UNINTERESTING;
1087         add_pending_object(&revs, &filter->merge_commit->object, "");
1088
1089         revs.limited = 1;
1090         if (prepare_revision_walk(&revs))
1091                 die(_("revision walk setup failed"));
1092
1093         old_nr = array->nr;
1094         array->nr = 0;
1095
1096         for (i = 0; i < old_nr; i++) {
1097                 struct ref_array_item *item = array->items[i];
1098                 struct commit *commit = item->commit;
1099
1100                 int is_merged = !!(commit->object.flags & UNINTERESTING);
1101
1102                 if (is_merged == (filter->merge == REF_FILTER_MERGED_INCLUDE))
1103                         array->items[array->nr++] = array->items[i];
1104                 else
1105                         free_array_item(item);
1106         }
1107
1108         for (i = 0; i < old_nr; i++)
1109                 clear_commit_marks(to_clear[i], ALL_REV_FLAGS);
1110         clear_commit_marks(filter->merge_commit, ALL_REV_FLAGS);
1111         free(to_clear);
1112 }
1113
1114 /*
1115  * API for filtering a set of refs. Based on the type of refs the user
1116  * has requested, we iterate through those refs and apply filters
1117  * as per the given ref_filter structure and finally store the
1118  * filtered refs in the ref_array structure.
1119  */
1120 int filter_refs(struct ref_array *array, struct ref_filter *filter, unsigned int type)
1121 {
1122         struct ref_filter_cbdata ref_cbdata;
1123         int ret = 0;
1124
1125         ref_cbdata.array = array;
1126         ref_cbdata.filter = filter;
1127
1128         /*  Simple per-ref filtering */
1129         if (type & (FILTER_REFS_ALL | FILTER_REFS_INCLUDE_BROKEN))
1130                 ret = for_each_rawref(ref_filter_handler, &ref_cbdata);
1131         else if (type & FILTER_REFS_ALL)
1132                 ret = for_each_ref(ref_filter_handler, &ref_cbdata);
1133         else if (type)
1134                 die("filter_refs: invalid type");
1135
1136         /*  Filters that need revision walking */
1137         if (filter->merge_commit)
1138                 do_merge_filter(&ref_cbdata);
1139
1140         return ret;
1141 }
1142
1143 static int cmp_ref_sorting(struct ref_sorting *s, struct ref_array_item *a, struct ref_array_item *b)
1144 {
1145         struct atom_value *va, *vb;
1146         int cmp;
1147         cmp_type cmp_type = used_atom_type[s->atom];
1148
1149         get_ref_atom_value(a, s->atom, &va);
1150         get_ref_atom_value(b, s->atom, &vb);
1151         switch (cmp_type) {
1152         case FIELD_STR:
1153                 cmp = strcmp(va->s, vb->s);
1154                 break;
1155         default:
1156                 if (va->ul < vb->ul)
1157                         cmp = -1;
1158                 else if (va->ul == vb->ul)
1159                         cmp = 0;
1160                 else
1161                         cmp = 1;
1162                 break;
1163         }
1164         return (s->reverse) ? -cmp : cmp;
1165 }
1166
1167 static struct ref_sorting *ref_sorting;
1168 static int compare_refs(const void *a_, const void *b_)
1169 {
1170         struct ref_array_item *a = *((struct ref_array_item **)a_);
1171         struct ref_array_item *b = *((struct ref_array_item **)b_);
1172         struct ref_sorting *s;
1173
1174         for (s = ref_sorting; s; s = s->next) {
1175                 int cmp = cmp_ref_sorting(s, a, b);
1176                 if (cmp)
1177                         return cmp;
1178         }
1179         return 0;
1180 }
1181
1182 void ref_array_sort(struct ref_sorting *sorting, struct ref_array *array)
1183 {
1184         ref_sorting = sorting;
1185         qsort(array->items, array->nr, sizeof(struct ref_array_item *), compare_refs);
1186 }
1187
1188 static void print_value(struct atom_value *v, int quote_style)
1189 {
1190         struct strbuf sb = STRBUF_INIT;
1191         switch (quote_style) {
1192         case QUOTE_NONE:
1193                 fputs(v->s, stdout);
1194                 break;
1195         case QUOTE_SHELL:
1196                 sq_quote_buf(&sb, v->s);
1197                 break;
1198         case QUOTE_PERL:
1199                 perl_quote_buf(&sb, v->s);
1200                 break;
1201         case QUOTE_PYTHON:
1202                 python_quote_buf(&sb, v->s);
1203                 break;
1204         case QUOTE_TCL:
1205                 tcl_quote_buf(&sb, v->s);
1206                 break;
1207         }
1208         if (quote_style != QUOTE_NONE) {
1209                 fputs(sb.buf, stdout);
1210                 strbuf_release(&sb);
1211         }
1212 }
1213
1214 static int hex1(char ch)
1215 {
1216         if ('0' <= ch && ch <= '9')
1217                 return ch - '0';
1218         else if ('a' <= ch && ch <= 'f')
1219                 return ch - 'a' + 10;
1220         else if ('A' <= ch && ch <= 'F')
1221                 return ch - 'A' + 10;
1222         return -1;
1223 }
1224 static int hex2(const char *cp)
1225 {
1226         if (cp[0] && cp[1])
1227                 return (hex1(cp[0]) << 4) | hex1(cp[1]);
1228         else
1229                 return -1;
1230 }
1231
1232 static void emit(const char *cp, const char *ep)
1233 {
1234         while (*cp && (!ep || cp < ep)) {
1235                 if (*cp == '%') {
1236                         if (cp[1] == '%')
1237                                 cp++;
1238                         else {
1239                                 int ch = hex2(cp + 1);
1240                                 if (0 <= ch) {
1241                                         putchar(ch);
1242                                         cp += 3;
1243                                         continue;
1244                                 }
1245                         }
1246                 }
1247                 putchar(*cp);
1248                 cp++;
1249         }
1250 }
1251
1252 void show_ref_array_item(struct ref_array_item *info, const char *format, int quote_style)
1253 {
1254         const char *cp, *sp, *ep;
1255
1256         for (cp = format; *cp && (sp = find_next(cp)); cp = ep + 1) {
1257                 struct atom_value *atomv;
1258
1259                 ep = strchr(sp, ')');
1260                 if (cp < sp)
1261                         emit(cp, sp);
1262                 get_ref_atom_value(info, parse_ref_filter_atom(sp + 2, ep), &atomv);
1263                 print_value(atomv, quote_style);
1264         }
1265         if (*cp) {
1266                 sp = cp + strlen(cp);
1267                 emit(cp, sp);
1268         }
1269         if (need_color_reset_at_eol) {
1270                 struct atom_value resetv;
1271                 char color[COLOR_MAXLEN] = "";
1272
1273                 if (color_parse("reset", color) < 0)
1274                         die("BUG: couldn't parse 'reset' as a color");
1275                 resetv.s = color;
1276                 print_value(&resetv, quote_style);
1277         }
1278         putchar('\n');
1279 }
1280
1281 /*  If no sorting option is given, use refname to sort as default */
1282 struct ref_sorting *ref_default_sorting(void)
1283 {
1284         static const char cstr_name[] = "refname";
1285
1286         struct ref_sorting *sorting = xcalloc(1, sizeof(*sorting));
1287
1288         sorting->next = NULL;
1289         sorting->atom = parse_ref_filter_atom(cstr_name, cstr_name + strlen(cstr_name));
1290         return sorting;
1291 }
1292
1293 int parse_opt_ref_sorting(const struct option *opt, const char *arg, int unset)
1294 {
1295         struct ref_sorting **sorting_tail = opt->value;
1296         struct ref_sorting *s;
1297         int len;
1298
1299         if (!arg) /* should --no-sort void the list ? */
1300                 return -1;
1301
1302         s = xcalloc(1, sizeof(*s));
1303         s->next = *sorting_tail;
1304         *sorting_tail = s;
1305
1306         if (*arg == '-') {
1307                 s->reverse = 1;
1308                 arg++;
1309         }
1310         len = strlen(arg);
1311         s->atom = parse_ref_filter_atom(arg, arg+len);
1312         return 0;
1313 }
1314
1315 int parse_opt_merge_filter(const struct option *opt, const char *arg, int unset)
1316 {
1317         struct ref_filter *rf = opt->value;
1318         unsigned char sha1[20];
1319
1320         rf->merge = starts_with(opt->long_name, "no")
1321                 ? REF_FILTER_MERGED_OMIT
1322                 : REF_FILTER_MERGED_INCLUDE;
1323
1324         if (get_sha1(arg, sha1))
1325                 die(_("malformed object name %s"), arg);
1326
1327         rf->merge_commit = lookup_commit_reference_gently(sha1, 0);
1328         if (!rf->merge_commit)
1329                 return opterror(opt, "must point to a commit", 0);
1330
1331         return 0;
1332 }