ref-filter: use pretty.c logic for trailers
[git] / 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 "object-store.h"
7 #include "repository.h"
8 #include "commit.h"
9 #include "remote.h"
10 #include "color.h"
11 #include "tag.h"
12 #include "quote.h"
13 #include "ref-filter.h"
14 #include "revision.h"
15 #include "utf8.h"
16 #include "git-compat-util.h"
17 #include "version.h"
18 #include "trailer.h"
19 #include "wt-status.h"
20 #include "commit-slab.h"
21 #include "commit-graph.h"
22 #include "commit-reach.h"
23 #include "worktree.h"
24 #include "hashmap.h"
25 #include "strvec.h"
26
27 static struct ref_msg {
28         const char *gone;
29         const char *ahead;
30         const char *behind;
31         const char *ahead_behind;
32 } msgs = {
33          /* Untranslated plumbing messages: */
34         "gone",
35         "ahead %d",
36         "behind %d",
37         "ahead %d, behind %d"
38 };
39
40 void setup_ref_filter_porcelain_msg(void)
41 {
42         msgs.gone = _("gone");
43         msgs.ahead = _("ahead %d");
44         msgs.behind = _("behind %d");
45         msgs.ahead_behind = _("ahead %d, behind %d");
46 }
47
48 typedef enum { FIELD_STR, FIELD_ULONG, FIELD_TIME } cmp_type;
49 typedef enum { COMPARE_EQUAL, COMPARE_UNEQUAL, COMPARE_NONE } cmp_status;
50 typedef enum { SOURCE_NONE = 0, SOURCE_OBJ, SOURCE_OTHER } info_source;
51
52 struct align {
53         align_type position;
54         unsigned int width;
55 };
56
57 struct if_then_else {
58         cmp_status cmp_status;
59         const char *str;
60         unsigned int then_atom_seen : 1,
61                 else_atom_seen : 1,
62                 condition_satisfied : 1;
63 };
64
65 struct refname_atom {
66         enum { R_NORMAL, R_SHORT, R_LSTRIP, R_RSTRIP } option;
67         int lstrip, rstrip;
68 };
69
70 static struct ref_trailer_buf {
71         struct string_list filter_list;
72         struct strbuf sepbuf;
73         struct strbuf kvsepbuf;
74 } ref_trailer_buf = {STRING_LIST_INIT_NODUP, STRBUF_INIT, STRBUF_INIT};
75
76 static struct expand_data {
77         struct object_id oid;
78         enum object_type type;
79         unsigned long size;
80         off_t disk_size;
81         struct object_id delta_base_oid;
82         void *content;
83
84         struct object_info info;
85 } oi, oi_deref;
86
87 struct ref_to_worktree_entry {
88         struct hashmap_entry ent;
89         struct worktree *wt; /* key is wt->head_ref */
90 };
91
92 static int ref_to_worktree_map_cmpfnc(const void *unused_lookupdata,
93                                       const struct hashmap_entry *eptr,
94                                       const struct hashmap_entry *kptr,
95                                       const void *keydata_aka_refname)
96 {
97         const struct ref_to_worktree_entry *e, *k;
98
99         e = container_of(eptr, const struct ref_to_worktree_entry, ent);
100         k = container_of(kptr, const struct ref_to_worktree_entry, ent);
101
102         return strcmp(e->wt->head_ref,
103                 keydata_aka_refname ? keydata_aka_refname : k->wt->head_ref);
104 }
105
106 static struct ref_to_worktree_map {
107         struct hashmap map;
108         struct worktree **worktrees;
109 } ref_to_worktree_map;
110
111 /*
112  * An atom is a valid field atom listed below, possibly prefixed with
113  * a "*" to denote deref_tag().
114  *
115  * We parse given format string and sort specifiers, and make a list
116  * of properties that we need to extract out of objects.  ref_array_item
117  * structure will hold an array of values extracted that can be
118  * indexed with the "atom number", which is an index into this
119  * array.
120  */
121 static struct used_atom {
122         const char *name;
123         cmp_type type;
124         info_source source;
125         union {
126                 char color[COLOR_MAXLEN];
127                 struct align align;
128                 struct {
129                         enum {
130                                 RR_REF, RR_TRACK, RR_TRACKSHORT, RR_REMOTE_NAME, RR_REMOTE_REF
131                         } option;
132                         struct refname_atom refname;
133                         unsigned int nobracket : 1, push : 1, push_remote : 1;
134                 } remote_ref;
135                 struct {
136                         enum { C_BARE, C_BODY, C_BODY_DEP, C_LENGTH, C_LINES,
137                                C_SIG, C_SUB, C_SUB_SANITIZE, C_TRAILERS } option;
138                         struct process_trailer_options trailer_opts;
139                         unsigned int nlines;
140                 } contents;
141                 struct {
142                         cmp_status cmp_status;
143                         const char *str;
144                 } if_then_else;
145                 struct {
146                         enum { O_FULL, O_LENGTH, O_SHORT } option;
147                         unsigned int length;
148                 } oid;
149                 struct email_option {
150                         enum { EO_RAW, EO_TRIM, EO_LOCALPART } option;
151                 } email_option;
152                 struct refname_atom refname;
153                 char *head;
154         } u;
155 } *used_atom;
156 static int used_atom_cnt, need_tagged, need_symref;
157
158 /*
159  * Expand string, append it to strbuf *sb, then return error code ret.
160  * Allow to save few lines of code.
161  */
162 static int strbuf_addf_ret(struct strbuf *sb, int ret, const char *fmt, ...)
163 {
164         va_list ap;
165         va_start(ap, fmt);
166         strbuf_vaddf(sb, fmt, ap);
167         va_end(ap);
168         return ret;
169 }
170
171 static int color_atom_parser(const struct ref_format *format, struct used_atom *atom,
172                              const char *color_value, struct strbuf *err)
173 {
174         if (!color_value)
175                 return strbuf_addf_ret(err, -1, _("expected format: %%(color:<color>)"));
176         if (color_parse(color_value, atom->u.color) < 0)
177                 return strbuf_addf_ret(err, -1, _("unrecognized color: %%(color:%s)"),
178                                        color_value);
179         /*
180          * We check this after we've parsed the color, which lets us complain
181          * about syntactically bogus color names even if they won't be used.
182          */
183         if (!want_color(format->use_color))
184                 color_parse("", atom->u.color);
185         return 0;
186 }
187
188 static int refname_atom_parser_internal(struct refname_atom *atom, const char *arg,
189                                          const char *name, struct strbuf *err)
190 {
191         if (!arg)
192                 atom->option = R_NORMAL;
193         else if (!strcmp(arg, "short"))
194                 atom->option = R_SHORT;
195         else if (skip_prefix(arg, "lstrip=", &arg) ||
196                  skip_prefix(arg, "strip=", &arg)) {
197                 atom->option = R_LSTRIP;
198                 if (strtol_i(arg, 10, &atom->lstrip))
199                         return strbuf_addf_ret(err, -1, _("Integer value expected refname:lstrip=%s"), arg);
200         } else if (skip_prefix(arg, "rstrip=", &arg)) {
201                 atom->option = R_RSTRIP;
202                 if (strtol_i(arg, 10, &atom->rstrip))
203                         return strbuf_addf_ret(err, -1, _("Integer value expected refname:rstrip=%s"), arg);
204         } else
205                 return strbuf_addf_ret(err, -1, _("unrecognized %%(%s) argument: %s"), name, arg);
206         return 0;
207 }
208
209 static int remote_ref_atom_parser(const struct ref_format *format, struct used_atom *atom,
210                                   const char *arg, struct strbuf *err)
211 {
212         struct string_list params = STRING_LIST_INIT_DUP;
213         int i;
214
215         if (!strcmp(atom->name, "push") || starts_with(atom->name, "push:"))
216                 atom->u.remote_ref.push = 1;
217
218         if (!arg) {
219                 atom->u.remote_ref.option = RR_REF;
220                 return refname_atom_parser_internal(&atom->u.remote_ref.refname,
221                                                     arg, atom->name, err);
222         }
223
224         atom->u.remote_ref.nobracket = 0;
225         string_list_split(&params, arg, ',', -1);
226
227         for (i = 0; i < params.nr; i++) {
228                 const char *s = params.items[i].string;
229
230                 if (!strcmp(s, "track"))
231                         atom->u.remote_ref.option = RR_TRACK;
232                 else if (!strcmp(s, "trackshort"))
233                         atom->u.remote_ref.option = RR_TRACKSHORT;
234                 else if (!strcmp(s, "nobracket"))
235                         atom->u.remote_ref.nobracket = 1;
236                 else if (!strcmp(s, "remotename")) {
237                         atom->u.remote_ref.option = RR_REMOTE_NAME;
238                         atom->u.remote_ref.push_remote = 1;
239                 } else if (!strcmp(s, "remoteref")) {
240                         atom->u.remote_ref.option = RR_REMOTE_REF;
241                         atom->u.remote_ref.push_remote = 1;
242                 } else {
243                         atom->u.remote_ref.option = RR_REF;
244                         if (refname_atom_parser_internal(&atom->u.remote_ref.refname,
245                                                          arg, atom->name, err)) {
246                                 string_list_clear(&params, 0);
247                                 return -1;
248                         }
249                 }
250         }
251
252         string_list_clear(&params, 0);
253         return 0;
254 }
255
256 static int objecttype_atom_parser(const struct ref_format *format, struct used_atom *atom,
257                                   const char *arg, struct strbuf *err)
258 {
259         if (arg)
260                 return strbuf_addf_ret(err, -1, _("%%(objecttype) does not take arguments"));
261         if (*atom->name == '*')
262                 oi_deref.info.typep = &oi_deref.type;
263         else
264                 oi.info.typep = &oi.type;
265         return 0;
266 }
267
268 static int objectsize_atom_parser(const struct ref_format *format, struct used_atom *atom,
269                                   const char *arg, struct strbuf *err)
270 {
271         if (!arg) {
272                 if (*atom->name == '*')
273                         oi_deref.info.sizep = &oi_deref.size;
274                 else
275                         oi.info.sizep = &oi.size;
276         } else if (!strcmp(arg, "disk")) {
277                 if (*atom->name == '*')
278                         oi_deref.info.disk_sizep = &oi_deref.disk_size;
279                 else
280                         oi.info.disk_sizep = &oi.disk_size;
281         } else
282                 return strbuf_addf_ret(err, -1, _("unrecognized %%(objectsize) argument: %s"), arg);
283         return 0;
284 }
285
286 static int deltabase_atom_parser(const struct ref_format *format, struct used_atom *atom,
287                                  const char *arg, struct strbuf *err)
288 {
289         if (arg)
290                 return strbuf_addf_ret(err, -1, _("%%(deltabase) does not take arguments"));
291         if (*atom->name == '*')
292                 oi_deref.info.delta_base_oid = &oi_deref.delta_base_oid;
293         else
294                 oi.info.delta_base_oid = &oi.delta_base_oid;
295         return 0;
296 }
297
298 static int body_atom_parser(const struct ref_format *format, struct used_atom *atom,
299                             const char *arg, struct strbuf *err)
300 {
301         if (arg)
302                 return strbuf_addf_ret(err, -1, _("%%(body) does not take arguments"));
303         atom->u.contents.option = C_BODY_DEP;
304         return 0;
305 }
306
307 static int subject_atom_parser(const struct ref_format *format, struct used_atom *atom,
308                                const char *arg, struct strbuf *err)
309 {
310         if (!arg)
311                 atom->u.contents.option = C_SUB;
312         else if (!strcmp(arg, "sanitize"))
313                 atom->u.contents.option = C_SUB_SANITIZE;
314         else
315                 return strbuf_addf_ret(err, -1, _("unrecognized %%(subject) argument: %s"), arg);
316         return 0;
317 }
318
319 static int trailers_atom_parser(const struct ref_format *format, struct used_atom *atom,
320                                 const char *arg, struct strbuf *err)
321 {
322         atom->u.contents.trailer_opts.no_divider = 1;
323
324         if (arg) {
325                 const char *argbuf = xstrfmt("%s)", arg);
326                 char *invalid_arg = NULL;
327
328                 if (format_set_trailers_options(&atom->u.contents.trailer_opts,
329                     &ref_trailer_buf.filter_list,
330                     &ref_trailer_buf.sepbuf,
331                     &ref_trailer_buf.kvsepbuf,
332                     &argbuf, &invalid_arg)) {
333                         if (!invalid_arg)
334                                 strbuf_addf(err, _("expected %%(trailers:key=<value>)"));
335                         else
336                                 strbuf_addf(err, _("unknown %%(trailers) argument: %s"), invalid_arg);
337                         free((char *)invalid_arg);
338                         return -1;
339                 }
340         }
341         atom->u.contents.option = C_TRAILERS;
342         return 0;
343 }
344
345 static int contents_atom_parser(const struct ref_format *format, struct used_atom *atom,
346                                 const char *arg, struct strbuf *err)
347 {
348         if (!arg)
349                 atom->u.contents.option = C_BARE;
350         else if (!strcmp(arg, "body"))
351                 atom->u.contents.option = C_BODY;
352         else if (!strcmp(arg, "size"))
353                 atom->u.contents.option = C_LENGTH;
354         else if (!strcmp(arg, "signature"))
355                 atom->u.contents.option = C_SIG;
356         else if (!strcmp(arg, "subject"))
357                 atom->u.contents.option = C_SUB;
358         else if (!strcmp(arg, "trailers")) {
359                 if (trailers_atom_parser(format, atom, NULL, err))
360                         return -1;
361         } else if (skip_prefix(arg, "trailers:", &arg)) {
362                 if (trailers_atom_parser(format, atom, arg, err))
363                         return -1;
364         } else if (skip_prefix(arg, "lines=", &arg)) {
365                 atom->u.contents.option = C_LINES;
366                 if (strtoul_ui(arg, 10, &atom->u.contents.nlines))
367                         return strbuf_addf_ret(err, -1, _("positive value expected contents:lines=%s"), arg);
368         } else
369                 return strbuf_addf_ret(err, -1, _("unrecognized %%(contents) argument: %s"), arg);
370         return 0;
371 }
372
373 static int oid_atom_parser(const struct ref_format *format, struct used_atom *atom,
374                            const char *arg, struct strbuf *err)
375 {
376         if (!arg)
377                 atom->u.oid.option = O_FULL;
378         else if (!strcmp(arg, "short"))
379                 atom->u.oid.option = O_SHORT;
380         else if (skip_prefix(arg, "short=", &arg)) {
381                 atom->u.oid.option = O_LENGTH;
382                 if (strtoul_ui(arg, 10, &atom->u.oid.length) ||
383                     atom->u.oid.length == 0)
384                         return strbuf_addf_ret(err, -1, _("positive value expected '%s' in %%(%s)"), arg, atom->name);
385                 if (atom->u.oid.length < MINIMUM_ABBREV)
386                         atom->u.oid.length = MINIMUM_ABBREV;
387         } else
388                 return strbuf_addf_ret(err, -1, _("unrecognized argument '%s' in %%(%s)"), arg, atom->name);
389         return 0;
390 }
391
392 static int person_email_atom_parser(const struct ref_format *format, struct used_atom *atom,
393                                     const char *arg, struct strbuf *err)
394 {
395         if (!arg)
396                 atom->u.email_option.option = EO_RAW;
397         else if (!strcmp(arg, "trim"))
398                 atom->u.email_option.option = EO_TRIM;
399         else if (!strcmp(arg, "localpart"))
400                 atom->u.email_option.option = EO_LOCALPART;
401         else
402                 return strbuf_addf_ret(err, -1, _("unrecognized email option: %s"), arg);
403         return 0;
404 }
405
406 static int refname_atom_parser(const struct ref_format *format, struct used_atom *atom,
407                                const char *arg, struct strbuf *err)
408 {
409         return refname_atom_parser_internal(&atom->u.refname, arg, atom->name, err);
410 }
411
412 static align_type parse_align_position(const char *s)
413 {
414         if (!strcmp(s, "right"))
415                 return ALIGN_RIGHT;
416         else if (!strcmp(s, "middle"))
417                 return ALIGN_MIDDLE;
418         else if (!strcmp(s, "left"))
419                 return ALIGN_LEFT;
420         return -1;
421 }
422
423 static int align_atom_parser(const struct ref_format *format, struct used_atom *atom,
424                              const char *arg, struct strbuf *err)
425 {
426         struct align *align = &atom->u.align;
427         struct string_list params = STRING_LIST_INIT_DUP;
428         int i;
429         unsigned int width = ~0U;
430
431         if (!arg)
432                 return strbuf_addf_ret(err, -1, _("expected format: %%(align:<width>,<position>)"));
433
434         align->position = ALIGN_LEFT;
435
436         string_list_split(&params, arg, ',', -1);
437         for (i = 0; i < params.nr; i++) {
438                 const char *s = params.items[i].string;
439                 int position;
440
441                 if (skip_prefix(s, "position=", &s)) {
442                         position = parse_align_position(s);
443                         if (position < 0) {
444                                 strbuf_addf(err, _("unrecognized position:%s"), s);
445                                 string_list_clear(&params, 0);
446                                 return -1;
447                         }
448                         align->position = position;
449                 } else if (skip_prefix(s, "width=", &s)) {
450                         if (strtoul_ui(s, 10, &width)) {
451                                 strbuf_addf(err, _("unrecognized width:%s"), s);
452                                 string_list_clear(&params, 0);
453                                 return -1;
454                         }
455                 } else if (!strtoul_ui(s, 10, &width))
456                         ;
457                 else if ((position = parse_align_position(s)) >= 0)
458                         align->position = position;
459                 else {
460                         strbuf_addf(err, _("unrecognized %%(align) argument: %s"), s);
461                         string_list_clear(&params, 0);
462                         return -1;
463                 }
464         }
465
466         if (width == ~0U) {
467                 string_list_clear(&params, 0);
468                 return strbuf_addf_ret(err, -1, _("positive width expected with the %%(align) atom"));
469         }
470         align->width = width;
471         string_list_clear(&params, 0);
472         return 0;
473 }
474
475 static int if_atom_parser(const struct ref_format *format, struct used_atom *atom,
476                           const char *arg, struct strbuf *err)
477 {
478         if (!arg) {
479                 atom->u.if_then_else.cmp_status = COMPARE_NONE;
480                 return 0;
481         } else if (skip_prefix(arg, "equals=", &atom->u.if_then_else.str)) {
482                 atom->u.if_then_else.cmp_status = COMPARE_EQUAL;
483         } else if (skip_prefix(arg, "notequals=", &atom->u.if_then_else.str)) {
484                 atom->u.if_then_else.cmp_status = COMPARE_UNEQUAL;
485         } else
486                 return strbuf_addf_ret(err, -1, _("unrecognized %%(if) argument: %s"), arg);
487         return 0;
488 }
489
490 static int head_atom_parser(const struct ref_format *format, struct used_atom *atom,
491                             const char *arg, struct strbuf *unused_err)
492 {
493         atom->u.head = resolve_refdup("HEAD", RESOLVE_REF_READING, NULL, NULL);
494         return 0;
495 }
496
497 static struct {
498         const char *name;
499         info_source source;
500         cmp_type cmp_type;
501         int (*parser)(const struct ref_format *format, struct used_atom *atom,
502                       const char *arg, struct strbuf *err);
503 } valid_atom[] = {
504         { "refname", SOURCE_NONE, FIELD_STR, refname_atom_parser },
505         { "objecttype", SOURCE_OTHER, FIELD_STR, objecttype_atom_parser },
506         { "objectsize", SOURCE_OTHER, FIELD_ULONG, objectsize_atom_parser },
507         { "objectname", SOURCE_OTHER, FIELD_STR, oid_atom_parser },
508         { "deltabase", SOURCE_OTHER, FIELD_STR, deltabase_atom_parser },
509         { "tree", SOURCE_OBJ, FIELD_STR, oid_atom_parser },
510         { "parent", SOURCE_OBJ, FIELD_STR, oid_atom_parser },
511         { "numparent", SOURCE_OBJ, FIELD_ULONG },
512         { "object", SOURCE_OBJ },
513         { "type", SOURCE_OBJ },
514         { "tag", SOURCE_OBJ },
515         { "author", SOURCE_OBJ },
516         { "authorname", SOURCE_OBJ },
517         { "authoremail", SOURCE_OBJ, FIELD_STR, person_email_atom_parser },
518         { "authordate", SOURCE_OBJ, FIELD_TIME },
519         { "committer", SOURCE_OBJ },
520         { "committername", SOURCE_OBJ },
521         { "committeremail", SOURCE_OBJ, FIELD_STR, person_email_atom_parser },
522         { "committerdate", SOURCE_OBJ, FIELD_TIME },
523         { "tagger", SOURCE_OBJ },
524         { "taggername", SOURCE_OBJ },
525         { "taggeremail", SOURCE_OBJ, FIELD_STR, person_email_atom_parser },
526         { "taggerdate", SOURCE_OBJ, FIELD_TIME },
527         { "creator", SOURCE_OBJ },
528         { "creatordate", SOURCE_OBJ, FIELD_TIME },
529         { "subject", SOURCE_OBJ, FIELD_STR, subject_atom_parser },
530         { "body", SOURCE_OBJ, FIELD_STR, body_atom_parser },
531         { "trailers", SOURCE_OBJ, FIELD_STR, trailers_atom_parser },
532         { "contents", SOURCE_OBJ, FIELD_STR, contents_atom_parser },
533         { "upstream", SOURCE_NONE, FIELD_STR, remote_ref_atom_parser },
534         { "push", SOURCE_NONE, FIELD_STR, remote_ref_atom_parser },
535         { "symref", SOURCE_NONE, FIELD_STR, refname_atom_parser },
536         { "flag", SOURCE_NONE },
537         { "HEAD", SOURCE_NONE, FIELD_STR, head_atom_parser },
538         { "color", SOURCE_NONE, FIELD_STR, color_atom_parser },
539         { "worktreepath", SOURCE_NONE },
540         { "align", SOURCE_NONE, FIELD_STR, align_atom_parser },
541         { "end", SOURCE_NONE },
542         { "if", SOURCE_NONE, FIELD_STR, if_atom_parser },
543         { "then", SOURCE_NONE },
544         { "else", SOURCE_NONE },
545         /*
546          * Please update $__git_ref_fieldlist in git-completion.bash
547          * when you add new atoms
548          */
549 };
550
551 #define REF_FORMATTING_STATE_INIT  { 0, NULL }
552
553 struct ref_formatting_stack {
554         struct ref_formatting_stack *prev;
555         struct strbuf output;
556         void (*at_end)(struct ref_formatting_stack **stack);
557         void *at_end_data;
558 };
559
560 struct ref_formatting_state {
561         int quote_style;
562         struct ref_formatting_stack *stack;
563 };
564
565 struct atom_value {
566         const char *s;
567         int (*handler)(struct atom_value *atomv, struct ref_formatting_state *state,
568                        struct strbuf *err);
569         uintmax_t value; /* used for sorting when not FIELD_STR */
570         struct used_atom *atom;
571 };
572
573 /*
574  * Used to parse format string and sort specifiers
575  */
576 static int parse_ref_filter_atom(const struct ref_format *format,
577                                  const char *atom, const char *ep,
578                                  struct strbuf *err)
579 {
580         const char *sp;
581         const char *arg;
582         int i, at, atom_len;
583
584         sp = atom;
585         if (*sp == '*' && sp < ep)
586                 sp++; /* deref */
587         if (ep <= sp)
588                 return strbuf_addf_ret(err, -1, _("malformed field name: %.*s"),
589                                        (int)(ep-atom), atom);
590
591         /* Do we have the atom already used elsewhere? */
592         for (i = 0; i < used_atom_cnt; i++) {
593                 int len = strlen(used_atom[i].name);
594                 if (len == ep - atom && !memcmp(used_atom[i].name, atom, len))
595                         return i;
596         }
597
598         /*
599          * If the atom name has a colon, strip it and everything after
600          * it off - it specifies the format for this entry, and
601          * shouldn't be used for checking against the valid_atom
602          * table.
603          */
604         arg = memchr(sp, ':', ep - sp);
605         atom_len = (arg ? arg : ep) - sp;
606
607         /* Is the atom a valid one? */
608         for (i = 0; i < ARRAY_SIZE(valid_atom); i++) {
609                 int len = strlen(valid_atom[i].name);
610                 if (len == atom_len && !memcmp(valid_atom[i].name, sp, len))
611                         break;
612         }
613
614         if (ARRAY_SIZE(valid_atom) <= i)
615                 return strbuf_addf_ret(err, -1, _("unknown field name: %.*s"),
616                                        (int)(ep-atom), atom);
617         if (valid_atom[i].source != SOURCE_NONE && !have_git_dir())
618                 return strbuf_addf_ret(err, -1,
619                                        _("not a git repository, but the field '%.*s' requires access to object data"),
620                                        (int)(ep-atom), atom);
621
622         /* Add it in, including the deref prefix */
623         at = used_atom_cnt;
624         used_atom_cnt++;
625         REALLOC_ARRAY(used_atom, used_atom_cnt);
626         used_atom[at].name = xmemdupz(atom, ep - atom);
627         used_atom[at].type = valid_atom[i].cmp_type;
628         used_atom[at].source = valid_atom[i].source;
629         if (used_atom[at].source == SOURCE_OBJ) {
630                 if (*atom == '*')
631                         oi_deref.info.contentp = &oi_deref.content;
632                 else
633                         oi.info.contentp = &oi.content;
634         }
635         if (arg) {
636                 arg = used_atom[at].name + (arg - atom) + 1;
637                 if (!*arg) {
638                         /*
639                          * Treat empty sub-arguments list as NULL (i.e.,
640                          * "%(atom:)" is equivalent to "%(atom)").
641                          */
642                         arg = NULL;
643                 }
644         }
645         memset(&used_atom[at].u, 0, sizeof(used_atom[at].u));
646         if (valid_atom[i].parser && valid_atom[i].parser(format, &used_atom[at], arg, err))
647                 return -1;
648         if (*atom == '*')
649                 need_tagged = 1;
650         if (!strcmp(valid_atom[i].name, "symref"))
651                 need_symref = 1;
652         return at;
653 }
654
655 static void quote_formatting(struct strbuf *s, const char *str, int quote_style)
656 {
657         switch (quote_style) {
658         case QUOTE_NONE:
659                 strbuf_addstr(s, str);
660                 break;
661         case QUOTE_SHELL:
662                 sq_quote_buf(s, str);
663                 break;
664         case QUOTE_PERL:
665                 perl_quote_buf(s, str);
666                 break;
667         case QUOTE_PYTHON:
668                 python_quote_buf(s, str);
669                 break;
670         case QUOTE_TCL:
671                 tcl_quote_buf(s, str);
672                 break;
673         }
674 }
675
676 static int append_atom(struct atom_value *v, struct ref_formatting_state *state,
677                        struct strbuf *unused_err)
678 {
679         /*
680          * Quote formatting is only done when the stack has a single
681          * element. Otherwise quote formatting is done on the
682          * element's entire output strbuf when the %(end) atom is
683          * encountered.
684          */
685         if (!state->stack->prev)
686                 quote_formatting(&state->stack->output, v->s, state->quote_style);
687         else
688                 strbuf_addstr(&state->stack->output, v->s);
689         return 0;
690 }
691
692 static void push_stack_element(struct ref_formatting_stack **stack)
693 {
694         struct ref_formatting_stack *s = xcalloc(1, sizeof(struct ref_formatting_stack));
695
696         strbuf_init(&s->output, 0);
697         s->prev = *stack;
698         *stack = s;
699 }
700
701 static void pop_stack_element(struct ref_formatting_stack **stack)
702 {
703         struct ref_formatting_stack *current = *stack;
704         struct ref_formatting_stack *prev = current->prev;
705
706         if (prev)
707                 strbuf_addbuf(&prev->output, &current->output);
708         strbuf_release(&current->output);
709         free(current);
710         *stack = prev;
711 }
712
713 static void end_align_handler(struct ref_formatting_stack **stack)
714 {
715         struct ref_formatting_stack *cur = *stack;
716         struct align *align = (struct align *)cur->at_end_data;
717         struct strbuf s = STRBUF_INIT;
718
719         strbuf_utf8_align(&s, align->position, align->width, cur->output.buf);
720         strbuf_swap(&cur->output, &s);
721         strbuf_release(&s);
722 }
723
724 static int align_atom_handler(struct atom_value *atomv, struct ref_formatting_state *state,
725                               struct strbuf *unused_err)
726 {
727         struct ref_formatting_stack *new_stack;
728
729         push_stack_element(&state->stack);
730         new_stack = state->stack;
731         new_stack->at_end = end_align_handler;
732         new_stack->at_end_data = &atomv->atom->u.align;
733         return 0;
734 }
735
736 static void if_then_else_handler(struct ref_formatting_stack **stack)
737 {
738         struct ref_formatting_stack *cur = *stack;
739         struct ref_formatting_stack *prev = cur->prev;
740         struct if_then_else *if_then_else = (struct if_then_else *)cur->at_end_data;
741
742         if (!if_then_else->then_atom_seen)
743                 die(_("format: %%(if) atom used without a %%(then) atom"));
744
745         if (if_then_else->else_atom_seen) {
746                 /*
747                  * There is an %(else) atom: we need to drop one state from the
748                  * stack, either the %(else) branch if the condition is satisfied, or
749                  * the %(then) branch if it isn't.
750                  */
751                 if (if_then_else->condition_satisfied) {
752                         strbuf_reset(&cur->output);
753                         pop_stack_element(&cur);
754                 } else {
755                         strbuf_swap(&cur->output, &prev->output);
756                         strbuf_reset(&cur->output);
757                         pop_stack_element(&cur);
758                 }
759         } else if (!if_then_else->condition_satisfied) {
760                 /*
761                  * No %(else) atom: just drop the %(then) branch if the
762                  * condition is not satisfied.
763                  */
764                 strbuf_reset(&cur->output);
765         }
766
767         *stack = cur;
768         free(if_then_else);
769 }
770
771 static int if_atom_handler(struct atom_value *atomv, struct ref_formatting_state *state,
772                            struct strbuf *unused_err)
773 {
774         struct ref_formatting_stack *new_stack;
775         struct if_then_else *if_then_else = xcalloc(sizeof(struct if_then_else), 1);
776
777         if_then_else->str = atomv->atom->u.if_then_else.str;
778         if_then_else->cmp_status = atomv->atom->u.if_then_else.cmp_status;
779
780         push_stack_element(&state->stack);
781         new_stack = state->stack;
782         new_stack->at_end = if_then_else_handler;
783         new_stack->at_end_data = if_then_else;
784         return 0;
785 }
786
787 static int is_empty(const char *s)
788 {
789         while (*s != '\0') {
790                 if (!isspace(*s))
791                         return 0;
792                 s++;
793         }
794         return 1;
795 }
796
797 static int then_atom_handler(struct atom_value *atomv, struct ref_formatting_state *state,
798                              struct strbuf *err)
799 {
800         struct ref_formatting_stack *cur = state->stack;
801         struct if_then_else *if_then_else = NULL;
802
803         if (cur->at_end == if_then_else_handler)
804                 if_then_else = (struct if_then_else *)cur->at_end_data;
805         if (!if_then_else)
806                 return strbuf_addf_ret(err, -1, _("format: %%(then) atom used without an %%(if) atom"));
807         if (if_then_else->then_atom_seen)
808                 return strbuf_addf_ret(err, -1, _("format: %%(then) atom used more than once"));
809         if (if_then_else->else_atom_seen)
810                 return strbuf_addf_ret(err, -1, _("format: %%(then) atom used after %%(else)"));
811         if_then_else->then_atom_seen = 1;
812         /*
813          * If the 'equals' or 'notequals' attribute is used then
814          * perform the required comparison. If not, only non-empty
815          * strings satisfy the 'if' condition.
816          */
817         if (if_then_else->cmp_status == COMPARE_EQUAL) {
818                 if (!strcmp(if_then_else->str, cur->output.buf))
819                         if_then_else->condition_satisfied = 1;
820         } else if (if_then_else->cmp_status == COMPARE_UNEQUAL) {
821                 if (strcmp(if_then_else->str, cur->output.buf))
822                         if_then_else->condition_satisfied = 1;
823         } else if (cur->output.len && !is_empty(cur->output.buf))
824                 if_then_else->condition_satisfied = 1;
825         strbuf_reset(&cur->output);
826         return 0;
827 }
828
829 static int else_atom_handler(struct atom_value *atomv, struct ref_formatting_state *state,
830                              struct strbuf *err)
831 {
832         struct ref_formatting_stack *prev = state->stack;
833         struct if_then_else *if_then_else = NULL;
834
835         if (prev->at_end == if_then_else_handler)
836                 if_then_else = (struct if_then_else *)prev->at_end_data;
837         if (!if_then_else)
838                 return strbuf_addf_ret(err, -1, _("format: %%(else) atom used without an %%(if) atom"));
839         if (!if_then_else->then_atom_seen)
840                 return strbuf_addf_ret(err, -1, _("format: %%(else) atom used without a %%(then) atom"));
841         if (if_then_else->else_atom_seen)
842                 return strbuf_addf_ret(err, -1, _("format: %%(else) atom used more than once"));
843         if_then_else->else_atom_seen = 1;
844         push_stack_element(&state->stack);
845         state->stack->at_end_data = prev->at_end_data;
846         state->stack->at_end = prev->at_end;
847         return 0;
848 }
849
850 static int end_atom_handler(struct atom_value *atomv, struct ref_formatting_state *state,
851                             struct strbuf *err)
852 {
853         struct ref_formatting_stack *current = state->stack;
854         struct strbuf s = STRBUF_INIT;
855
856         if (!current->at_end)
857                 return strbuf_addf_ret(err, -1, _("format: %%(end) atom used without corresponding atom"));
858         current->at_end(&state->stack);
859
860         /*  Stack may have been popped within at_end(), hence reset the current pointer */
861         current = state->stack;
862
863         /*
864          * Perform quote formatting when the stack element is that of
865          * a supporting atom. If nested then perform quote formatting
866          * only on the topmost supporting atom.
867          */
868         if (!current->prev->prev) {
869                 quote_formatting(&s, current->output.buf, state->quote_style);
870                 strbuf_swap(&current->output, &s);
871         }
872         strbuf_release(&s);
873         pop_stack_element(&state->stack);
874         return 0;
875 }
876
877 /*
878  * In a format string, find the next occurrence of %(atom).
879  */
880 static const char *find_next(const char *cp)
881 {
882         while (*cp) {
883                 if (*cp == '%') {
884                         /*
885                          * %( is the start of an atom;
886                          * %% is a quoted per-cent.
887                          */
888                         if (cp[1] == '(')
889                                 return cp;
890                         else if (cp[1] == '%')
891                                 cp++; /* skip over two % */
892                         /* otherwise this is a singleton, literal % */
893                 }
894                 cp++;
895         }
896         return NULL;
897 }
898
899 /*
900  * Make sure the format string is well formed, and parse out
901  * the used atoms.
902  */
903 int verify_ref_format(struct ref_format *format)
904 {
905         const char *cp, *sp;
906
907         format->need_color_reset_at_eol = 0;
908         for (cp = format->format; *cp && (sp = find_next(cp)); ) {
909                 struct strbuf err = STRBUF_INIT;
910                 const char *color, *ep = strchr(sp, ')');
911                 int at;
912
913                 if (!ep)
914                         return error(_("malformed format string %s"), sp);
915                 /* sp points at "%(" and ep points at the closing ")" */
916                 at = parse_ref_filter_atom(format, sp + 2, ep, &err);
917                 if (at < 0)
918                         die("%s", err.buf);
919                 cp = ep + 1;
920
921                 if (skip_prefix(used_atom[at].name, "color:", &color))
922                         format->need_color_reset_at_eol = !!strcmp(color, "reset");
923                 strbuf_release(&err);
924         }
925         if (format->need_color_reset_at_eol && !want_color(format->use_color))
926                 format->need_color_reset_at_eol = 0;
927         return 0;
928 }
929
930 static const char *do_grab_oid(const char *field, const struct object_id *oid,
931                                struct used_atom *atom)
932 {
933         switch (atom->u.oid.option) {
934         case O_FULL:
935                 return oid_to_hex(oid);
936         case O_LENGTH:
937                 return find_unique_abbrev(oid, atom->u.oid.length);
938         case O_SHORT:
939                 return find_unique_abbrev(oid, DEFAULT_ABBREV);
940         default:
941                 BUG("unknown %%(%s) option", field);
942         }
943 }
944
945 static int grab_oid(const char *name, const char *field, const struct object_id *oid,
946                     struct atom_value *v, struct used_atom *atom)
947 {
948         if (starts_with(name, field)) {
949                 v->s = xstrdup(do_grab_oid(field, oid, atom));
950                 return 1;
951         }
952         return 0;
953 }
954
955 /* See grab_values */
956 static void grab_common_values(struct atom_value *val, int deref, struct expand_data *oi)
957 {
958         int i;
959
960         for (i = 0; i < used_atom_cnt; i++) {
961                 const char *name = used_atom[i].name;
962                 struct atom_value *v = &val[i];
963                 if (!!deref != (*name == '*'))
964                         continue;
965                 if (deref)
966                         name++;
967                 if (!strcmp(name, "objecttype"))
968                         v->s = xstrdup(type_name(oi->type));
969                 else if (!strcmp(name, "objectsize:disk")) {
970                         v->value = oi->disk_size;
971                         v->s = xstrfmt("%"PRIuMAX, (uintmax_t)oi->disk_size);
972                 } else if (!strcmp(name, "objectsize")) {
973                         v->value = oi->size;
974                         v->s = xstrfmt("%"PRIuMAX , (uintmax_t)oi->size);
975                 } else if (!strcmp(name, "deltabase"))
976                         v->s = xstrdup(oid_to_hex(&oi->delta_base_oid));
977                 else if (deref)
978                         grab_oid(name, "objectname", &oi->oid, v, &used_atom[i]);
979         }
980 }
981
982 /* See grab_values */
983 static void grab_tag_values(struct atom_value *val, int deref, struct object *obj)
984 {
985         int i;
986         struct tag *tag = (struct tag *) obj;
987
988         for (i = 0; i < used_atom_cnt; i++) {
989                 const char *name = used_atom[i].name;
990                 struct atom_value *v = &val[i];
991                 if (!!deref != (*name == '*'))
992                         continue;
993                 if (deref)
994                         name++;
995                 if (!strcmp(name, "tag"))
996                         v->s = xstrdup(tag->tag);
997                 else if (!strcmp(name, "type") && tag->tagged)
998                         v->s = xstrdup(type_name(tag->tagged->type));
999                 else if (!strcmp(name, "object") && tag->tagged)
1000                         v->s = xstrdup(oid_to_hex(&tag->tagged->oid));
1001         }
1002 }
1003
1004 /* See grab_values */
1005 static void grab_commit_values(struct atom_value *val, int deref, struct object *obj)
1006 {
1007         int i;
1008         struct commit *commit = (struct commit *) obj;
1009
1010         for (i = 0; i < used_atom_cnt; i++) {
1011                 const char *name = used_atom[i].name;
1012                 struct atom_value *v = &val[i];
1013                 if (!!deref != (*name == '*'))
1014                         continue;
1015                 if (deref)
1016                         name++;
1017                 if (grab_oid(name, "tree", get_commit_tree_oid(commit), v, &used_atom[i]))
1018                         continue;
1019                 if (!strcmp(name, "numparent")) {
1020                         v->value = commit_list_count(commit->parents);
1021                         v->s = xstrfmt("%lu", (unsigned long)v->value);
1022                 }
1023                 else if (starts_with(name, "parent")) {
1024                         struct commit_list *parents;
1025                         struct strbuf s = STRBUF_INIT;
1026                         for (parents = commit->parents; parents; parents = parents->next) {
1027                                 struct object_id *oid = &parents->item->object.oid;
1028                                 if (parents != commit->parents)
1029                                         strbuf_addch(&s, ' ');
1030                                 strbuf_addstr(&s, do_grab_oid("parent", oid, &used_atom[i]));
1031                         }
1032                         v->s = strbuf_detach(&s, NULL);
1033                 }
1034         }
1035 }
1036
1037 static const char *find_wholine(const char *who, int wholen, const char *buf)
1038 {
1039         const char *eol;
1040         while (*buf) {
1041                 if (!strncmp(buf, who, wholen) &&
1042                     buf[wholen] == ' ')
1043                         return buf + wholen + 1;
1044                 eol = strchr(buf, '\n');
1045                 if (!eol)
1046                         return "";
1047                 eol++;
1048                 if (*eol == '\n')
1049                         return ""; /* end of header */
1050                 buf = eol;
1051         }
1052         return "";
1053 }
1054
1055 static const char *copy_line(const char *buf)
1056 {
1057         const char *eol = strchrnul(buf, '\n');
1058         return xmemdupz(buf, eol - buf);
1059 }
1060
1061 static const char *copy_name(const char *buf)
1062 {
1063         const char *cp;
1064         for (cp = buf; *cp && *cp != '\n'; cp++) {
1065                 if (!strncmp(cp, " <", 2))
1066                         return xmemdupz(buf, cp - buf);
1067         }
1068         return xstrdup("");
1069 }
1070
1071 static const char *copy_email(const char *buf, struct used_atom *atom)
1072 {
1073         const char *email = strchr(buf, '<');
1074         const char *eoemail;
1075         if (!email)
1076                 return xstrdup("");
1077         switch (atom->u.email_option.option) {
1078         case EO_RAW:
1079                 eoemail = strchr(email, '>');
1080                 if (eoemail)
1081                         eoemail++;
1082                 break;
1083         case EO_TRIM:
1084                 email++;
1085                 eoemail = strchr(email, '>');
1086                 break;
1087         case EO_LOCALPART:
1088                 email++;
1089                 eoemail = strchr(email, '@');
1090                 if (!eoemail)
1091                         eoemail = strchr(email, '>');
1092                 break;
1093         default:
1094                 BUG("unknown email option");
1095         }
1096
1097         if (!eoemail)
1098                 return xstrdup("");
1099         return xmemdupz(email, eoemail - email);
1100 }
1101
1102 static char *copy_subject(const char *buf, unsigned long len)
1103 {
1104         struct strbuf sb = STRBUF_INIT;
1105         int i;
1106
1107         for (i = 0; i < len; i++) {
1108                 if (buf[i] == '\r' && i + 1 < len && buf[i + 1] == '\n')
1109                         continue; /* ignore CR in CRLF */
1110
1111                 if (buf[i] == '\n')
1112                         strbuf_addch(&sb, ' ');
1113                 else
1114                         strbuf_addch(&sb, buf[i]);
1115         }
1116         return strbuf_detach(&sb, NULL);
1117 }
1118
1119 static void grab_date(const char *buf, struct atom_value *v, const char *atomname)
1120 {
1121         const char *eoemail = strstr(buf, "> ");
1122         char *zone;
1123         timestamp_t timestamp;
1124         long tz;
1125         struct date_mode date_mode = { DATE_NORMAL };
1126         const char *formatp;
1127
1128         /*
1129          * We got here because atomname ends in "date" or "date<something>";
1130          * it's not possible that <something> is not ":<format>" because
1131          * parse_ref_filter_atom() wouldn't have allowed it, so we can assume that no
1132          * ":" means no format is specified, and use the default.
1133          */
1134         formatp = strchr(atomname, ':');
1135         if (formatp != NULL) {
1136                 formatp++;
1137                 parse_date_format(formatp, &date_mode);
1138         }
1139
1140         if (!eoemail)
1141                 goto bad;
1142         timestamp = parse_timestamp(eoemail + 2, &zone, 10);
1143         if (timestamp == TIME_MAX)
1144                 goto bad;
1145         tz = strtol(zone, NULL, 10);
1146         if ((tz == LONG_MIN || tz == LONG_MAX) && errno == ERANGE)
1147                 goto bad;
1148         v->s = xstrdup(show_date(timestamp, tz, &date_mode));
1149         v->value = timestamp;
1150         return;
1151  bad:
1152         v->s = xstrdup("");
1153         v->value = 0;
1154 }
1155
1156 /* See grab_values */
1157 static void grab_person(const char *who, struct atom_value *val, int deref, void *buf)
1158 {
1159         int i;
1160         int wholen = strlen(who);
1161         const char *wholine = NULL;
1162
1163         for (i = 0; i < used_atom_cnt; i++) {
1164                 const char *name = used_atom[i].name;
1165                 struct atom_value *v = &val[i];
1166                 if (!!deref != (*name == '*'))
1167                         continue;
1168                 if (deref)
1169                         name++;
1170                 if (strncmp(who, name, wholen))
1171                         continue;
1172                 if (name[wholen] != 0 &&
1173                     strcmp(name + wholen, "name") &&
1174                     !starts_with(name + wholen, "email") &&
1175                     !starts_with(name + wholen, "date"))
1176                         continue;
1177                 if (!wholine)
1178                         wholine = find_wholine(who, wholen, buf);
1179                 if (!wholine)
1180                         return; /* no point looking for it */
1181                 if (name[wholen] == 0)
1182                         v->s = copy_line(wholine);
1183                 else if (!strcmp(name + wholen, "name"))
1184                         v->s = copy_name(wholine);
1185                 else if (starts_with(name + wholen, "email"))
1186                         v->s = copy_email(wholine, &used_atom[i]);
1187                 else if (starts_with(name + wholen, "date"))
1188                         grab_date(wholine, v, name);
1189         }
1190
1191         /*
1192          * For a tag or a commit object, if "creator" or "creatordate" is
1193          * requested, do something special.
1194          */
1195         if (strcmp(who, "tagger") && strcmp(who, "committer"))
1196                 return; /* "author" for commit object is not wanted */
1197         if (!wholine)
1198                 wholine = find_wholine(who, wholen, buf);
1199         if (!wholine)
1200                 return;
1201         for (i = 0; i < used_atom_cnt; i++) {
1202                 const char *name = used_atom[i].name;
1203                 struct atom_value *v = &val[i];
1204                 if (!!deref != (*name == '*'))
1205                         continue;
1206                 if (deref)
1207                         name++;
1208
1209                 if (starts_with(name, "creatordate"))
1210                         grab_date(wholine, v, name);
1211                 else if (!strcmp(name, "creator"))
1212                         v->s = copy_line(wholine);
1213         }
1214 }
1215
1216 static void find_subpos(const char *buf,
1217                         const char **sub, unsigned long *sublen,
1218                         const char **body, unsigned long *bodylen,
1219                         unsigned long *nonsiglen,
1220                         const char **sig, unsigned long *siglen)
1221 {
1222         const char *eol;
1223         /* skip past header until we hit empty line */
1224         while (*buf && *buf != '\n') {
1225                 eol = strchrnul(buf, '\n');
1226                 if (*eol)
1227                         eol++;
1228                 buf = eol;
1229         }
1230         /* skip any empty lines */
1231         while (*buf == '\n')
1232                 buf++;
1233
1234         /* parse signature first; we might not even have a subject line */
1235         *sig = buf + parse_signature(buf, strlen(buf));
1236         *siglen = strlen(*sig);
1237
1238         /* subject is first non-empty line */
1239         *sub = buf;
1240         /* subject goes to first empty line before signature begins */
1241         if ((eol = strstr(*sub, "\n\n"))) {
1242                 eol = eol < *sig ? eol : *sig;
1243         /* check if message uses CRLF */
1244         } else if (! (eol = strstr(*sub, "\r\n\r\n"))) {
1245                 /* treat whole message as subject */
1246                 eol = strrchr(*sub, '\0');
1247         }
1248         buf = eol;
1249         *sublen = buf - *sub;
1250         /* drop trailing newline, if present */
1251         while (*sublen && ((*sub)[*sublen - 1] == '\n' ||
1252                            (*sub)[*sublen - 1] == '\r'))
1253                 *sublen -= 1;
1254
1255         /* skip any empty lines */
1256         while (*buf == '\n' || *buf == '\r')
1257                 buf++;
1258         *body = buf;
1259         *bodylen = strlen(buf);
1260         *nonsiglen = *sig - buf;
1261 }
1262
1263 /*
1264  * If 'lines' is greater than 0, append that many lines from the given
1265  * 'buf' of length 'size' to the given strbuf.
1266  */
1267 static void append_lines(struct strbuf *out, const char *buf, unsigned long size, int lines)
1268 {
1269         int i;
1270         const char *sp, *eol;
1271         size_t len;
1272
1273         sp = buf;
1274
1275         for (i = 0; i < lines && sp < buf + size; i++) {
1276                 if (i)
1277                         strbuf_addstr(out, "\n    ");
1278                 eol = memchr(sp, '\n', size - (sp - buf));
1279                 len = eol ? eol - sp : size - (sp - buf);
1280                 strbuf_add(out, sp, len);
1281                 if (!eol)
1282                         break;
1283                 sp = eol + 1;
1284         }
1285 }
1286
1287 /* See grab_values */
1288 static void grab_sub_body_contents(struct atom_value *val, int deref, void *buf)
1289 {
1290         int i;
1291         const char *subpos = NULL, *bodypos = NULL, *sigpos = NULL;
1292         unsigned long sublen = 0, bodylen = 0, nonsiglen = 0, siglen = 0;
1293
1294         for (i = 0; i < used_atom_cnt; i++) {
1295                 struct used_atom *atom = &used_atom[i];
1296                 const char *name = atom->name;
1297                 struct atom_value *v = &val[i];
1298                 if (!!deref != (*name == '*'))
1299                         continue;
1300                 if (deref)
1301                         name++;
1302                 if (strcmp(name, "body") &&
1303                     !starts_with(name, "subject") &&
1304                     !starts_with(name, "trailers") &&
1305                     !starts_with(name, "contents"))
1306                         continue;
1307                 if (!subpos)
1308                         find_subpos(buf,
1309                                     &subpos, &sublen,
1310                                     &bodypos, &bodylen, &nonsiglen,
1311                                     &sigpos, &siglen);
1312
1313                 if (atom->u.contents.option == C_SUB)
1314                         v->s = copy_subject(subpos, sublen);
1315                 else if (atom->u.contents.option == C_SUB_SANITIZE) {
1316                         struct strbuf sb = STRBUF_INIT;
1317                         format_sanitized_subject(&sb, subpos, sublen);
1318                         v->s = strbuf_detach(&sb, NULL);
1319                 } else if (atom->u.contents.option == C_BODY_DEP)
1320                         v->s = xmemdupz(bodypos, bodylen);
1321                 else if (atom->u.contents.option == C_LENGTH)
1322                         v->s = xstrfmt("%"PRIuMAX, (uintmax_t)strlen(subpos));
1323                 else if (atom->u.contents.option == C_BODY)
1324                         v->s = xmemdupz(bodypos, nonsiglen);
1325                 else if (atom->u.contents.option == C_SIG)
1326                         v->s = xmemdupz(sigpos, siglen);
1327                 else if (atom->u.contents.option == C_LINES) {
1328                         struct strbuf s = STRBUF_INIT;
1329                         const char *contents_end = bodylen + bodypos - siglen;
1330
1331                         /*  Size is the length of the message after removing the signature */
1332                         append_lines(&s, subpos, contents_end - subpos, atom->u.contents.nlines);
1333                         v->s = strbuf_detach(&s, NULL);
1334                 } else if (atom->u.contents.option == C_TRAILERS) {
1335                         struct strbuf s = STRBUF_INIT;
1336
1337                         /* Format the trailer info according to the trailer_opts given */
1338                         format_trailers_from_commit(&s, subpos, &atom->u.contents.trailer_opts);
1339
1340                         v->s = strbuf_detach(&s, NULL);
1341                 } else if (atom->u.contents.option == C_BARE)
1342                         v->s = xstrdup(subpos);
1343         }
1344 }
1345
1346 /*
1347  * We want to have empty print-string for field requests
1348  * that do not apply (e.g. "authordate" for a tag object)
1349  */
1350 static void fill_missing_values(struct atom_value *val)
1351 {
1352         int i;
1353         for (i = 0; i < used_atom_cnt; i++) {
1354                 struct atom_value *v = &val[i];
1355                 if (v->s == NULL)
1356                         v->s = xstrdup("");
1357         }
1358 }
1359
1360 /*
1361  * val is a list of atom_value to hold returned values.  Extract
1362  * the values for atoms in used_atom array out of (obj, buf, sz).
1363  * when deref is false, (obj, buf, sz) is the object that is
1364  * pointed at by the ref itself; otherwise it is the object the
1365  * ref (which is a tag) refers to.
1366  */
1367 static void grab_values(struct atom_value *val, int deref, struct object *obj, void *buf)
1368 {
1369         switch (obj->type) {
1370         case OBJ_TAG:
1371                 grab_tag_values(val, deref, obj);
1372                 grab_sub_body_contents(val, deref, buf);
1373                 grab_person("tagger", val, deref, buf);
1374                 break;
1375         case OBJ_COMMIT:
1376                 grab_commit_values(val, deref, obj);
1377                 grab_sub_body_contents(val, deref, buf);
1378                 grab_person("author", val, deref, buf);
1379                 grab_person("committer", val, deref, buf);
1380                 break;
1381         case OBJ_TREE:
1382                 /* grab_tree_values(val, deref, obj, buf, sz); */
1383                 break;
1384         case OBJ_BLOB:
1385                 /* grab_blob_values(val, deref, obj, buf, sz); */
1386                 break;
1387         default:
1388                 die("Eh?  Object of type %d?", obj->type);
1389         }
1390 }
1391
1392 static inline char *copy_advance(char *dst, const char *src)
1393 {
1394         while (*src)
1395                 *dst++ = *src++;
1396         return dst;
1397 }
1398
1399 static const char *lstrip_ref_components(const char *refname, int len)
1400 {
1401         long remaining = len;
1402         const char *start = xstrdup(refname);
1403         const char *to_free = start;
1404
1405         if (len < 0) {
1406                 int i;
1407                 const char *p = refname;
1408
1409                 /* Find total no of '/' separated path-components */
1410                 for (i = 0; p[i]; p[i] == '/' ? i++ : *p++)
1411                         ;
1412                 /*
1413                  * The number of components we need to strip is now
1414                  * the total minus the components to be left (Plus one
1415                  * because we count the number of '/', but the number
1416                  * of components is one more than the no of '/').
1417                  */
1418                 remaining = i + len + 1;
1419         }
1420
1421         while (remaining > 0) {
1422                 switch (*start++) {
1423                 case '\0':
1424                         free((char *)to_free);
1425                         return xstrdup("");
1426                 case '/':
1427                         remaining--;
1428                         break;
1429                 }
1430         }
1431
1432         start = xstrdup(start);
1433         free((char *)to_free);
1434         return start;
1435 }
1436
1437 static const char *rstrip_ref_components(const char *refname, int len)
1438 {
1439         long remaining = len;
1440         const char *start = xstrdup(refname);
1441         const char *to_free = start;
1442
1443         if (len < 0) {
1444                 int i;
1445                 const char *p = refname;
1446
1447                 /* Find total no of '/' separated path-components */
1448                 for (i = 0; p[i]; p[i] == '/' ? i++ : *p++)
1449                         ;
1450                 /*
1451                  * The number of components we need to strip is now
1452                  * the total minus the components to be left (Plus one
1453                  * because we count the number of '/', but the number
1454                  * of components is one more than the no of '/').
1455                  */
1456                 remaining = i + len + 1;
1457         }
1458
1459         while (remaining-- > 0) {
1460                 char *p = strrchr(start, '/');
1461                 if (p == NULL) {
1462                         free((char *)to_free);
1463                         return xstrdup("");
1464                 } else
1465                         p[0] = '\0';
1466         }
1467         return start;
1468 }
1469
1470 static const char *show_ref(struct refname_atom *atom, const char *refname)
1471 {
1472         if (atom->option == R_SHORT)
1473                 return shorten_unambiguous_ref(refname, warn_ambiguous_refs);
1474         else if (atom->option == R_LSTRIP)
1475                 return lstrip_ref_components(refname, atom->lstrip);
1476         else if (atom->option == R_RSTRIP)
1477                 return rstrip_ref_components(refname, atom->rstrip);
1478         else
1479                 return xstrdup(refname);
1480 }
1481
1482 static void fill_remote_ref_details(struct used_atom *atom, const char *refname,
1483                                     struct branch *branch, const char **s)
1484 {
1485         int num_ours, num_theirs;
1486         if (atom->u.remote_ref.option == RR_REF)
1487                 *s = show_ref(&atom->u.remote_ref.refname, refname);
1488         else if (atom->u.remote_ref.option == RR_TRACK) {
1489                 if (stat_tracking_info(branch, &num_ours, &num_theirs,
1490                                        NULL, atom->u.remote_ref.push,
1491                                        AHEAD_BEHIND_FULL) < 0) {
1492                         *s = xstrdup(msgs.gone);
1493                 } else if (!num_ours && !num_theirs)
1494                         *s = xstrdup("");
1495                 else if (!num_ours)
1496                         *s = xstrfmt(msgs.behind, num_theirs);
1497                 else if (!num_theirs)
1498                         *s = xstrfmt(msgs.ahead, num_ours);
1499                 else
1500                         *s = xstrfmt(msgs.ahead_behind,
1501                                      num_ours, num_theirs);
1502                 if (!atom->u.remote_ref.nobracket && *s[0]) {
1503                         const char *to_free = *s;
1504                         *s = xstrfmt("[%s]", *s);
1505                         free((void *)to_free);
1506                 }
1507         } else if (atom->u.remote_ref.option == RR_TRACKSHORT) {
1508                 if (stat_tracking_info(branch, &num_ours, &num_theirs,
1509                                        NULL, atom->u.remote_ref.push,
1510                                        AHEAD_BEHIND_FULL) < 0) {
1511                         *s = xstrdup("");
1512                         return;
1513                 }
1514                 if (!num_ours && !num_theirs)
1515                         *s = xstrdup("=");
1516                 else if (!num_ours)
1517                         *s = xstrdup("<");
1518                 else if (!num_theirs)
1519                         *s = xstrdup(">");
1520                 else
1521                         *s = xstrdup("<>");
1522         } else if (atom->u.remote_ref.option == RR_REMOTE_NAME) {
1523                 int explicit;
1524                 const char *remote = atom->u.remote_ref.push ?
1525                         pushremote_for_branch(branch, &explicit) :
1526                         remote_for_branch(branch, &explicit);
1527                 *s = xstrdup(explicit ? remote : "");
1528         } else if (atom->u.remote_ref.option == RR_REMOTE_REF) {
1529                 const char *merge;
1530
1531                 merge = remote_ref_for_branch(branch, atom->u.remote_ref.push);
1532                 *s = xstrdup(merge ? merge : "");
1533         } else
1534                 BUG("unhandled RR_* enum");
1535 }
1536
1537 char *get_head_description(void)
1538 {
1539         struct strbuf desc = STRBUF_INIT;
1540         struct wt_status_state state;
1541         memset(&state, 0, sizeof(state));
1542         wt_status_get_state(the_repository, &state, 1);
1543         if (state.rebase_in_progress ||
1544             state.rebase_interactive_in_progress) {
1545                 if (state.branch)
1546                         strbuf_addf(&desc, _("(no branch, rebasing %s)"),
1547                                     state.branch);
1548                 else
1549                         strbuf_addf(&desc, _("(no branch, rebasing detached HEAD %s)"),
1550                                     state.detached_from);
1551         } else if (state.bisect_in_progress)
1552                 strbuf_addf(&desc, _("(no branch, bisect started on %s)"),
1553                             state.branch);
1554         else if (state.detached_from) {
1555                 if (state.detached_at)
1556                         strbuf_addf(&desc, _("(HEAD detached at %s)"),
1557                                 state.detached_from);
1558                 else
1559                         strbuf_addf(&desc, _("(HEAD detached from %s)"),
1560                                 state.detached_from);
1561         } else
1562                 strbuf_addstr(&desc, _("(no branch)"));
1563
1564         return strbuf_detach(&desc, NULL);
1565 }
1566
1567 static const char *get_symref(struct used_atom *atom, struct ref_array_item *ref)
1568 {
1569         if (!ref->symref)
1570                 return xstrdup("");
1571         else
1572                 return show_ref(&atom->u.refname, ref->symref);
1573 }
1574
1575 static const char *get_refname(struct used_atom *atom, struct ref_array_item *ref)
1576 {
1577         if (ref->kind & FILTER_REFS_DETACHED_HEAD)
1578                 return get_head_description();
1579         return show_ref(&atom->u.refname, ref->refname);
1580 }
1581
1582 static int get_object(struct ref_array_item *ref, int deref, struct object **obj,
1583                       struct expand_data *oi, struct strbuf *err)
1584 {
1585         /* parse_object_buffer() will set eaten to 0 if free() will be needed */
1586         int eaten = 1;
1587         if (oi->info.contentp) {
1588                 /* We need to know that to use parse_object_buffer properly */
1589                 oi->info.sizep = &oi->size;
1590                 oi->info.typep = &oi->type;
1591         }
1592         if (oid_object_info_extended(the_repository, &oi->oid, &oi->info,
1593                                      OBJECT_INFO_LOOKUP_REPLACE))
1594                 return strbuf_addf_ret(err, -1, _("missing object %s for %s"),
1595                                        oid_to_hex(&oi->oid), ref->refname);
1596         if (oi->info.disk_sizep && oi->disk_size < 0)
1597                 BUG("Object size is less than zero.");
1598
1599         if (oi->info.contentp) {
1600                 *obj = parse_object_buffer(the_repository, &oi->oid, oi->type, oi->size, oi->content, &eaten);
1601                 if (!obj) {
1602                         if (!eaten)
1603                                 free(oi->content);
1604                         return strbuf_addf_ret(err, -1, _("parse_object_buffer failed on %s for %s"),
1605                                                oid_to_hex(&oi->oid), ref->refname);
1606                 }
1607                 grab_values(ref->value, deref, *obj, oi->content);
1608         }
1609
1610         grab_common_values(ref->value, deref, oi);
1611         if (!eaten)
1612                 free(oi->content);
1613         return 0;
1614 }
1615
1616 static void populate_worktree_map(struct hashmap *map, struct worktree **worktrees)
1617 {
1618         int i;
1619
1620         for (i = 0; worktrees[i]; i++) {
1621                 if (worktrees[i]->head_ref) {
1622                         struct ref_to_worktree_entry *entry;
1623                         entry = xmalloc(sizeof(*entry));
1624                         entry->wt = worktrees[i];
1625                         hashmap_entry_init(&entry->ent,
1626                                         strhash(worktrees[i]->head_ref));
1627
1628                         hashmap_add(map, &entry->ent);
1629                 }
1630         }
1631 }
1632
1633 static void lazy_init_worktree_map(void)
1634 {
1635         if (ref_to_worktree_map.worktrees)
1636                 return;
1637
1638         ref_to_worktree_map.worktrees = get_worktrees();
1639         hashmap_init(&(ref_to_worktree_map.map), ref_to_worktree_map_cmpfnc, NULL, 0);
1640         populate_worktree_map(&(ref_to_worktree_map.map), ref_to_worktree_map.worktrees);
1641 }
1642
1643 static char *get_worktree_path(const struct used_atom *atom, const struct ref_array_item *ref)
1644 {
1645         struct hashmap_entry entry, *e;
1646         struct ref_to_worktree_entry *lookup_result;
1647
1648         lazy_init_worktree_map();
1649
1650         hashmap_entry_init(&entry, strhash(ref->refname));
1651         e = hashmap_get(&(ref_to_worktree_map.map), &entry, ref->refname);
1652
1653         if (!e)
1654                 return xstrdup("");
1655
1656         lookup_result = container_of(e, struct ref_to_worktree_entry, ent);
1657
1658         return xstrdup(lookup_result->wt->path);
1659 }
1660
1661 /*
1662  * Parse the object referred by ref, and grab needed value.
1663  */
1664 static int populate_value(struct ref_array_item *ref, struct strbuf *err)
1665 {
1666         struct object *obj;
1667         int i;
1668         struct object_info empty = OBJECT_INFO_INIT;
1669
1670         ref->value = xcalloc(used_atom_cnt, sizeof(struct atom_value));
1671
1672         if (need_symref && (ref->flag & REF_ISSYMREF) && !ref->symref) {
1673                 ref->symref = resolve_refdup(ref->refname, RESOLVE_REF_READING,
1674                                              NULL, NULL);
1675                 if (!ref->symref)
1676                         ref->symref = xstrdup("");
1677         }
1678
1679         /* Fill in specials first */
1680         for (i = 0; i < used_atom_cnt; i++) {
1681                 struct used_atom *atom = &used_atom[i];
1682                 const char *name = used_atom[i].name;
1683                 struct atom_value *v = &ref->value[i];
1684                 int deref = 0;
1685                 const char *refname;
1686                 struct branch *branch = NULL;
1687
1688                 v->handler = append_atom;
1689                 v->atom = atom;
1690
1691                 if (*name == '*') {
1692                         deref = 1;
1693                         name++;
1694                 }
1695
1696                 if (starts_with(name, "refname"))
1697                         refname = get_refname(atom, ref);
1698                 else if (!strcmp(name, "worktreepath")) {
1699                         if (ref->kind == FILTER_REFS_BRANCHES)
1700                                 v->s = get_worktree_path(atom, ref);
1701                         else
1702                                 v->s = xstrdup("");
1703                         continue;
1704                 }
1705                 else if (starts_with(name, "symref"))
1706                         refname = get_symref(atom, ref);
1707                 else if (starts_with(name, "upstream")) {
1708                         const char *branch_name;
1709                         /* only local branches may have an upstream */
1710                         if (!skip_prefix(ref->refname, "refs/heads/",
1711                                          &branch_name)) {
1712                                 v->s = xstrdup("");
1713                                 continue;
1714                         }
1715                         branch = branch_get(branch_name);
1716
1717                         refname = branch_get_upstream(branch, NULL);
1718                         if (refname)
1719                                 fill_remote_ref_details(atom, refname, branch, &v->s);
1720                         else
1721                                 v->s = xstrdup("");
1722                         continue;
1723                 } else if (atom->u.remote_ref.push) {
1724                         const char *branch_name;
1725                         v->s = xstrdup("");
1726                         if (!skip_prefix(ref->refname, "refs/heads/",
1727                                          &branch_name))
1728                                 continue;
1729                         branch = branch_get(branch_name);
1730
1731                         if (atom->u.remote_ref.push_remote)
1732                                 refname = NULL;
1733                         else {
1734                                 refname = branch_get_push(branch, NULL);
1735                                 if (!refname)
1736                                         continue;
1737                         }
1738                         /* We will definitely re-init v->s on the next line. */
1739                         free((char *)v->s);
1740                         fill_remote_ref_details(atom, refname, branch, &v->s);
1741                         continue;
1742                 } else if (starts_with(name, "color:")) {
1743                         v->s = xstrdup(atom->u.color);
1744                         continue;
1745                 } else if (!strcmp(name, "flag")) {
1746                         char buf[256], *cp = buf;
1747                         if (ref->flag & REF_ISSYMREF)
1748                                 cp = copy_advance(cp, ",symref");
1749                         if (ref->flag & REF_ISPACKED)
1750                                 cp = copy_advance(cp, ",packed");
1751                         if (cp == buf)
1752                                 v->s = xstrdup("");
1753                         else {
1754                                 *cp = '\0';
1755                                 v->s = xstrdup(buf + 1);
1756                         }
1757                         continue;
1758                 } else if (!deref && grab_oid(name, "objectname", &ref->objectname, v, atom)) {
1759                         continue;
1760                 } else if (!strcmp(name, "HEAD")) {
1761                         if (atom->u.head && !strcmp(ref->refname, atom->u.head))
1762                                 v->s = xstrdup("*");
1763                         else
1764                                 v->s = xstrdup(" ");
1765                         continue;
1766                 } else if (starts_with(name, "align")) {
1767                         v->handler = align_atom_handler;
1768                         v->s = xstrdup("");
1769                         continue;
1770                 } else if (!strcmp(name, "end")) {
1771                         v->handler = end_atom_handler;
1772                         v->s = xstrdup("");
1773                         continue;
1774                 } else if (starts_with(name, "if")) {
1775                         const char *s;
1776                         if (skip_prefix(name, "if:", &s))
1777                                 v->s = xstrdup(s);
1778                         else
1779                                 v->s = xstrdup("");
1780                         v->handler = if_atom_handler;
1781                         continue;
1782                 } else if (!strcmp(name, "then")) {
1783                         v->handler = then_atom_handler;
1784                         v->s = xstrdup("");
1785                         continue;
1786                 } else if (!strcmp(name, "else")) {
1787                         v->handler = else_atom_handler;
1788                         v->s = xstrdup("");
1789                         continue;
1790                 } else
1791                         continue;
1792
1793                 if (!deref)
1794                         v->s = xstrdup(refname);
1795                 else
1796                         v->s = xstrfmt("%s^{}", refname);
1797                 free((char *)refname);
1798         }
1799
1800         for (i = 0; i < used_atom_cnt; i++) {
1801                 struct atom_value *v = &ref->value[i];
1802                 if (v->s == NULL && used_atom[i].source == SOURCE_NONE)
1803                         return strbuf_addf_ret(err, -1, _("missing object %s for %s"),
1804                                                oid_to_hex(&ref->objectname), ref->refname);
1805         }
1806
1807         if (need_tagged)
1808                 oi.info.contentp = &oi.content;
1809         if (!memcmp(&oi.info, &empty, sizeof(empty)) &&
1810             !memcmp(&oi_deref.info, &empty, sizeof(empty)))
1811                 return 0;
1812
1813
1814         oi.oid = ref->objectname;
1815         if (get_object(ref, 0, &obj, &oi, err))
1816                 return -1;
1817
1818         /*
1819          * If there is no atom that wants to know about tagged
1820          * object, we are done.
1821          */
1822         if (!need_tagged || (obj->type != OBJ_TAG))
1823                 return 0;
1824
1825         /*
1826          * If it is a tag object, see if we use a value that derefs
1827          * the object, and if we do grab the object it refers to.
1828          */
1829         oi_deref.oid = *get_tagged_oid((struct tag *)obj);
1830
1831         /*
1832          * NEEDSWORK: This derefs tag only once, which
1833          * is good to deal with chains of trust, but
1834          * is not consistent with what deref_tag() does
1835          * which peels the onion to the core.
1836          */
1837         return get_object(ref, 1, &obj, &oi_deref, err);
1838 }
1839
1840 /*
1841  * Given a ref, return the value for the atom.  This lazily gets value
1842  * out of the object by calling populate value.
1843  */
1844 static int get_ref_atom_value(struct ref_array_item *ref, int atom,
1845                               struct atom_value **v, struct strbuf *err)
1846 {
1847         if (!ref->value) {
1848                 if (populate_value(ref, err))
1849                         return -1;
1850                 fill_missing_values(ref->value);
1851         }
1852         *v = &ref->value[atom];
1853         return 0;
1854 }
1855
1856 /*
1857  * Return 1 if the refname matches one of the patterns, otherwise 0.
1858  * A pattern can be a literal prefix (e.g. a refname "refs/heads/master"
1859  * matches a pattern "refs/heads/mas") or a wildcard (e.g. the same ref
1860  * matches "refs/heads/mas*", too).
1861  */
1862 static int match_pattern(const struct ref_filter *filter, const char *refname)
1863 {
1864         const char **patterns = filter->name_patterns;
1865         unsigned flags = 0;
1866
1867         if (filter->ignore_case)
1868                 flags |= WM_CASEFOLD;
1869
1870         /*
1871          * When no '--format' option is given we need to skip the prefix
1872          * for matching refs of tags and branches.
1873          */
1874         (void)(skip_prefix(refname, "refs/tags/", &refname) ||
1875                skip_prefix(refname, "refs/heads/", &refname) ||
1876                skip_prefix(refname, "refs/remotes/", &refname) ||
1877                skip_prefix(refname, "refs/", &refname));
1878
1879         for (; *patterns; patterns++) {
1880                 if (!wildmatch(*patterns, refname, flags))
1881                         return 1;
1882         }
1883         return 0;
1884 }
1885
1886 /*
1887  * Return 1 if the refname matches one of the patterns, otherwise 0.
1888  * A pattern can be path prefix (e.g. a refname "refs/heads/master"
1889  * matches a pattern "refs/heads/" but not "refs/heads/m") or a
1890  * wildcard (e.g. the same ref matches "refs/heads/m*", too).
1891  */
1892 static int match_name_as_path(const struct ref_filter *filter, const char *refname)
1893 {
1894         const char **pattern = filter->name_patterns;
1895         int namelen = strlen(refname);
1896         unsigned flags = WM_PATHNAME;
1897
1898         if (filter->ignore_case)
1899                 flags |= WM_CASEFOLD;
1900
1901         for (; *pattern; pattern++) {
1902                 const char *p = *pattern;
1903                 int plen = strlen(p);
1904
1905                 if ((plen <= namelen) &&
1906                     !strncmp(refname, p, plen) &&
1907                     (refname[plen] == '\0' ||
1908                      refname[plen] == '/' ||
1909                      p[plen-1] == '/'))
1910                         return 1;
1911                 if (!wildmatch(p, refname, flags))
1912                         return 1;
1913         }
1914         return 0;
1915 }
1916
1917 /* Return 1 if the refname matches one of the patterns, otherwise 0. */
1918 static int filter_pattern_match(struct ref_filter *filter, const char *refname)
1919 {
1920         if (!*filter->name_patterns)
1921                 return 1; /* No pattern always matches */
1922         if (filter->match_as_path)
1923                 return match_name_as_path(filter, refname);
1924         return match_pattern(filter, refname);
1925 }
1926
1927 static int qsort_strcmp(const void *va, const void *vb)
1928 {
1929         const char *a = *(const char **)va;
1930         const char *b = *(const char **)vb;
1931
1932         return strcmp(a, b);
1933 }
1934
1935 static void find_longest_prefixes_1(struct string_list *out,
1936                                   struct strbuf *prefix,
1937                                   const char **patterns, size_t nr)
1938 {
1939         size_t i;
1940
1941         for (i = 0; i < nr; i++) {
1942                 char c = patterns[i][prefix->len];
1943                 if (!c || is_glob_special(c)) {
1944                         string_list_append(out, prefix->buf);
1945                         return;
1946                 }
1947         }
1948
1949         i = 0;
1950         while (i < nr) {
1951                 size_t end;
1952
1953                 /*
1954                 * Set "end" to the index of the element _after_ the last one
1955                 * in our group.
1956                 */
1957                 for (end = i + 1; end < nr; end++) {
1958                         if (patterns[i][prefix->len] != patterns[end][prefix->len])
1959                                 break;
1960                 }
1961
1962                 strbuf_addch(prefix, patterns[i][prefix->len]);
1963                 find_longest_prefixes_1(out, prefix, patterns + i, end - i);
1964                 strbuf_setlen(prefix, prefix->len - 1);
1965
1966                 i = end;
1967         }
1968 }
1969
1970 static void find_longest_prefixes(struct string_list *out,
1971                                   const char **patterns)
1972 {
1973         struct strvec sorted = STRVEC_INIT;
1974         struct strbuf prefix = STRBUF_INIT;
1975
1976         strvec_pushv(&sorted, patterns);
1977         QSORT(sorted.v, sorted.nr, qsort_strcmp);
1978
1979         find_longest_prefixes_1(out, &prefix, sorted.v, sorted.nr);
1980
1981         strvec_clear(&sorted);
1982         strbuf_release(&prefix);
1983 }
1984
1985 /*
1986  * This is the same as for_each_fullref_in(), but it tries to iterate
1987  * only over the patterns we'll care about. Note that it _doesn't_ do a full
1988  * pattern match, so the callback still has to match each ref individually.
1989  */
1990 static int for_each_fullref_in_pattern(struct ref_filter *filter,
1991                                        each_ref_fn cb,
1992                                        void *cb_data,
1993                                        int broken)
1994 {
1995         struct string_list prefixes = STRING_LIST_INIT_DUP;
1996         struct string_list_item *prefix;
1997         int ret;
1998
1999         if (!filter->match_as_path) {
2000                 /*
2001                  * in this case, the patterns are applied after
2002                  * prefixes like "refs/heads/" etc. are stripped off,
2003                  * so we have to look at everything:
2004                  */
2005                 return for_each_fullref_in("", cb, cb_data, broken);
2006         }
2007
2008         if (filter->ignore_case) {
2009                 /*
2010                  * we can't handle case-insensitive comparisons,
2011                  * so just return everything and let the caller
2012                  * sort it out.
2013                  */
2014                 return for_each_fullref_in("", cb, cb_data, broken);
2015         }
2016
2017         if (!filter->name_patterns[0]) {
2018                 /* no patterns; we have to look at everything */
2019                 return for_each_fullref_in("", cb, cb_data, broken);
2020         }
2021
2022         find_longest_prefixes(&prefixes, filter->name_patterns);
2023
2024         for_each_string_list_item(prefix, &prefixes) {
2025                 ret = for_each_fullref_in(prefix->string, cb, cb_data, broken);
2026                 if (ret)
2027                         break;
2028         }
2029
2030         string_list_clear(&prefixes, 0);
2031         return ret;
2032 }
2033
2034 /*
2035  * Given a ref (oid, refname), check if the ref belongs to the array
2036  * of oids. If the given ref is a tag, check if the given tag points
2037  * at one of the oids in the given oid array.
2038  * NEEDSWORK:
2039  * 1. Only a single level of indirection is obtained, we might want to
2040  * change this to account for multiple levels (e.g. annotated tags
2041  * pointing to annotated tags pointing to a commit.)
2042  * 2. As the refs are cached we might know what refname peels to without
2043  * the need to parse the object via parse_object(). peel_ref() might be a
2044  * more efficient alternative to obtain the pointee.
2045  */
2046 static const struct object_id *match_points_at(struct oid_array *points_at,
2047                                                const struct object_id *oid,
2048                                                const char *refname)
2049 {
2050         const struct object_id *tagged_oid = NULL;
2051         struct object *obj;
2052
2053         if (oid_array_lookup(points_at, oid) >= 0)
2054                 return oid;
2055         obj = parse_object(the_repository, oid);
2056         if (!obj)
2057                 die(_("malformed object at '%s'"), refname);
2058         if (obj->type == OBJ_TAG)
2059                 tagged_oid = get_tagged_oid((struct tag *)obj);
2060         if (tagged_oid && oid_array_lookup(points_at, tagged_oid) >= 0)
2061                 return tagged_oid;
2062         return NULL;
2063 }
2064
2065 /*
2066  * Allocate space for a new ref_array_item and copy the name and oid to it.
2067  *
2068  * Callers can then fill in other struct members at their leisure.
2069  */
2070 static struct ref_array_item *new_ref_array_item(const char *refname,
2071                                                  const struct object_id *oid)
2072 {
2073         struct ref_array_item *ref;
2074
2075         FLEX_ALLOC_STR(ref, refname, refname);
2076         oidcpy(&ref->objectname, oid);
2077
2078         return ref;
2079 }
2080
2081 struct ref_array_item *ref_array_push(struct ref_array *array,
2082                                       const char *refname,
2083                                       const struct object_id *oid)
2084 {
2085         struct ref_array_item *ref = new_ref_array_item(refname, oid);
2086
2087         ALLOC_GROW(array->items, array->nr + 1, array->alloc);
2088         array->items[array->nr++] = ref;
2089
2090         return ref;
2091 }
2092
2093 static int ref_kind_from_refname(const char *refname)
2094 {
2095         unsigned int i;
2096
2097         static struct {
2098                 const char *prefix;
2099                 unsigned int kind;
2100         } ref_kind[] = {
2101                 { "refs/heads/" , FILTER_REFS_BRANCHES },
2102                 { "refs/remotes/" , FILTER_REFS_REMOTES },
2103                 { "refs/tags/", FILTER_REFS_TAGS}
2104         };
2105
2106         if (!strcmp(refname, "HEAD"))
2107                 return FILTER_REFS_DETACHED_HEAD;
2108
2109         for (i = 0; i < ARRAY_SIZE(ref_kind); i++) {
2110                 if (starts_with(refname, ref_kind[i].prefix))
2111                         return ref_kind[i].kind;
2112         }
2113
2114         return FILTER_REFS_OTHERS;
2115 }
2116
2117 static int filter_ref_kind(struct ref_filter *filter, const char *refname)
2118 {
2119         if (filter->kind == FILTER_REFS_BRANCHES ||
2120             filter->kind == FILTER_REFS_REMOTES ||
2121             filter->kind == FILTER_REFS_TAGS)
2122                 return filter->kind;
2123         return ref_kind_from_refname(refname);
2124 }
2125
2126 struct ref_filter_cbdata {
2127         struct ref_array *array;
2128         struct ref_filter *filter;
2129         struct contains_cache contains_cache;
2130         struct contains_cache no_contains_cache;
2131 };
2132
2133 /*
2134  * A call-back given to for_each_ref().  Filter refs and keep them for
2135  * later object processing.
2136  */
2137 static int ref_filter_handler(const char *refname, const struct object_id *oid, int flag, void *cb_data)
2138 {
2139         struct ref_filter_cbdata *ref_cbdata = cb_data;
2140         struct ref_filter *filter = ref_cbdata->filter;
2141         struct ref_array_item *ref;
2142         struct commit *commit = NULL;
2143         unsigned int kind;
2144
2145         if (flag & REF_BAD_NAME) {
2146                 warning(_("ignoring ref with broken name %s"), refname);
2147                 return 0;
2148         }
2149
2150         if (flag & REF_ISBROKEN) {
2151                 warning(_("ignoring broken ref %s"), refname);
2152                 return 0;
2153         }
2154
2155         /* Obtain the current ref kind from filter_ref_kind() and ignore unwanted refs. */
2156         kind = filter_ref_kind(filter, refname);
2157         if (!(kind & filter->kind))
2158                 return 0;
2159
2160         if (!filter_pattern_match(filter, refname))
2161                 return 0;
2162
2163         if (filter->points_at.nr && !match_points_at(&filter->points_at, oid, refname))
2164                 return 0;
2165
2166         /*
2167          * A merge filter is applied on refs pointing to commits. Hence
2168          * obtain the commit using the 'oid' available and discard all
2169          * non-commits early. The actual filtering is done later.
2170          */
2171         if (filter->reachable_from || filter->unreachable_from ||
2172             filter->with_commit || filter->no_commit || filter->verbose) {
2173                 commit = lookup_commit_reference_gently(the_repository, oid, 1);
2174                 if (!commit)
2175                         return 0;
2176                 /* We perform the filtering for the '--contains' option... */
2177                 if (filter->with_commit &&
2178                     !commit_contains(filter, commit, filter->with_commit, &ref_cbdata->contains_cache))
2179                         return 0;
2180                 /* ...or for the `--no-contains' option */
2181                 if (filter->no_commit &&
2182                     commit_contains(filter, commit, filter->no_commit, &ref_cbdata->no_contains_cache))
2183                         return 0;
2184         }
2185
2186         /*
2187          * We do not open the object yet; sort may only need refname
2188          * to do its job and the resulting list may yet to be pruned
2189          * by maxcount logic.
2190          */
2191         ref = ref_array_push(ref_cbdata->array, refname, oid);
2192         ref->commit = commit;
2193         ref->flag = flag;
2194         ref->kind = kind;
2195
2196         return 0;
2197 }
2198
2199 /*  Free memory allocated for a ref_array_item */
2200 static void free_array_item(struct ref_array_item *item)
2201 {
2202         free((char *)item->symref);
2203         if (item->value) {
2204                 int i;
2205                 for (i = 0; i < used_atom_cnt; i++)
2206                         free((char *)item->value[i].s);
2207                 free(item->value);
2208         }
2209         free(item);
2210 }
2211
2212 /* Free all memory allocated for ref_array */
2213 void ref_array_clear(struct ref_array *array)
2214 {
2215         int i;
2216
2217         for (i = 0; i < array->nr; i++)
2218                 free_array_item(array->items[i]);
2219         FREE_AND_NULL(array->items);
2220         array->nr = array->alloc = 0;
2221
2222         for (i = 0; i < used_atom_cnt; i++)
2223                 free((char *)used_atom[i].name);
2224         FREE_AND_NULL(used_atom);
2225         used_atom_cnt = 0;
2226
2227         if (ref_to_worktree_map.worktrees) {
2228                 hashmap_clear_and_free(&(ref_to_worktree_map.map),
2229                                         struct ref_to_worktree_entry, ent);
2230                 free_worktrees(ref_to_worktree_map.worktrees);
2231                 ref_to_worktree_map.worktrees = NULL;
2232         }
2233 }
2234
2235 #define EXCLUDE_REACHED 0
2236 #define INCLUDE_REACHED 1
2237 static void reach_filter(struct ref_array *array,
2238                          struct commit_list *check_reachable,
2239                          int include_reached)
2240 {
2241         struct rev_info revs;
2242         int i, old_nr;
2243         struct commit **to_clear;
2244         struct commit_list *cr;
2245
2246         if (!check_reachable)
2247                 return;
2248
2249         to_clear = xcalloc(sizeof(struct commit *), array->nr);
2250
2251         repo_init_revisions(the_repository, &revs, NULL);
2252
2253         for (i = 0; i < array->nr; i++) {
2254                 struct ref_array_item *item = array->items[i];
2255                 add_pending_object(&revs, &item->commit->object, item->refname);
2256                 to_clear[i] = item->commit;
2257         }
2258
2259         for (cr = check_reachable; cr; cr = cr->next) {
2260                 struct commit *merge_commit = cr->item;
2261                 merge_commit->object.flags |= UNINTERESTING;
2262                 add_pending_object(&revs, &merge_commit->object, "");
2263         }
2264
2265         revs.limited = 1;
2266         if (prepare_revision_walk(&revs))
2267                 die(_("revision walk setup failed"));
2268
2269         old_nr = array->nr;
2270         array->nr = 0;
2271
2272         for (i = 0; i < old_nr; i++) {
2273                 struct ref_array_item *item = array->items[i];
2274                 struct commit *commit = item->commit;
2275
2276                 int is_merged = !!(commit->object.flags & UNINTERESTING);
2277
2278                 if (is_merged == include_reached)
2279                         array->items[array->nr++] = array->items[i];
2280                 else
2281                         free_array_item(item);
2282         }
2283
2284         clear_commit_marks_many(old_nr, to_clear, ALL_REV_FLAGS);
2285
2286         while (check_reachable) {
2287                 struct commit *merge_commit = pop_commit(&check_reachable);
2288                 clear_commit_marks(merge_commit, ALL_REV_FLAGS);
2289         }
2290
2291         free(to_clear);
2292 }
2293
2294 /*
2295  * API for filtering a set of refs. Based on the type of refs the user
2296  * has requested, we iterate through those refs and apply filters
2297  * as per the given ref_filter structure and finally store the
2298  * filtered refs in the ref_array structure.
2299  */
2300 int filter_refs(struct ref_array *array, struct ref_filter *filter, unsigned int type)
2301 {
2302         struct ref_filter_cbdata ref_cbdata;
2303         int ret = 0;
2304         unsigned int broken = 0;
2305
2306         ref_cbdata.array = array;
2307         ref_cbdata.filter = filter;
2308
2309         if (type & FILTER_REFS_INCLUDE_BROKEN)
2310                 broken = 1;
2311         filter->kind = type & FILTER_REFS_KIND_MASK;
2312
2313         init_contains_cache(&ref_cbdata.contains_cache);
2314         init_contains_cache(&ref_cbdata.no_contains_cache);
2315
2316         /*  Simple per-ref filtering */
2317         if (!filter->kind)
2318                 die("filter_refs: invalid type");
2319         else {
2320                 /*
2321                  * For common cases where we need only branches or remotes or tags,
2322                  * we only iterate through those refs. If a mix of refs is needed,
2323                  * we iterate over all refs and filter out required refs with the help
2324                  * of filter_ref_kind().
2325                  */
2326                 if (filter->kind == FILTER_REFS_BRANCHES)
2327                         ret = for_each_fullref_in("refs/heads/", ref_filter_handler, &ref_cbdata, broken);
2328                 else if (filter->kind == FILTER_REFS_REMOTES)
2329                         ret = for_each_fullref_in("refs/remotes/", ref_filter_handler, &ref_cbdata, broken);
2330                 else if (filter->kind == FILTER_REFS_TAGS)
2331                         ret = for_each_fullref_in("refs/tags/", ref_filter_handler, &ref_cbdata, broken);
2332                 else if (filter->kind & FILTER_REFS_ALL)
2333                         ret = for_each_fullref_in_pattern(filter, ref_filter_handler, &ref_cbdata, broken);
2334                 if (!ret && (filter->kind & FILTER_REFS_DETACHED_HEAD))
2335                         head_ref(ref_filter_handler, &ref_cbdata);
2336         }
2337
2338         clear_contains_cache(&ref_cbdata.contains_cache);
2339         clear_contains_cache(&ref_cbdata.no_contains_cache);
2340
2341         /*  Filters that need revision walking */
2342         reach_filter(array, filter->reachable_from, INCLUDE_REACHED);
2343         reach_filter(array, filter->unreachable_from, EXCLUDE_REACHED);
2344
2345         return ret;
2346 }
2347
2348 static int compare_detached_head(struct ref_array_item *a, struct ref_array_item *b)
2349 {
2350         if (!(a->kind ^ b->kind))
2351                 BUG("ref_kind_from_refname() should only mark one ref as HEAD");
2352         if (a->kind & FILTER_REFS_DETACHED_HEAD)
2353                 return -1;
2354         else if (b->kind & FILTER_REFS_DETACHED_HEAD)
2355                 return 1;
2356         BUG("should have died in the xor check above");
2357         return 0;
2358 }
2359
2360 static int cmp_ref_sorting(struct ref_sorting *s, struct ref_array_item *a, struct ref_array_item *b)
2361 {
2362         struct atom_value *va, *vb;
2363         int cmp;
2364         int cmp_detached_head = 0;
2365         cmp_type cmp_type = used_atom[s->atom].type;
2366         struct strbuf err = STRBUF_INIT;
2367
2368         if (get_ref_atom_value(a, s->atom, &va, &err))
2369                 die("%s", err.buf);
2370         if (get_ref_atom_value(b, s->atom, &vb, &err))
2371                 die("%s", err.buf);
2372         strbuf_release(&err);
2373         if (s->sort_flags & REF_SORTING_DETACHED_HEAD_FIRST &&
2374             ((a->kind | b->kind) & FILTER_REFS_DETACHED_HEAD)) {
2375                 cmp = compare_detached_head(a, b);
2376                 cmp_detached_head = 1;
2377         } else if (s->sort_flags & REF_SORTING_VERSION) {
2378                 cmp = versioncmp(va->s, vb->s);
2379         } else if (cmp_type == FIELD_STR) {
2380                 int (*cmp_fn)(const char *, const char *);
2381                 cmp_fn = s->sort_flags & REF_SORTING_ICASE
2382                         ? strcasecmp : strcmp;
2383                 cmp = cmp_fn(va->s, vb->s);
2384         } else {
2385                 if (va->value < vb->value)
2386                         cmp = -1;
2387                 else if (va->value == vb->value)
2388                         cmp = 0;
2389                 else
2390                         cmp = 1;
2391         }
2392
2393         return (s->sort_flags & REF_SORTING_REVERSE && !cmp_detached_head)
2394                 ? -cmp : cmp;
2395 }
2396
2397 static int compare_refs(const void *a_, const void *b_, void *ref_sorting)
2398 {
2399         struct ref_array_item *a = *((struct ref_array_item **)a_);
2400         struct ref_array_item *b = *((struct ref_array_item **)b_);
2401         struct ref_sorting *s;
2402
2403         for (s = ref_sorting; s; s = s->next) {
2404                 int cmp = cmp_ref_sorting(s, a, b);
2405                 if (cmp)
2406                         return cmp;
2407         }
2408         s = ref_sorting;
2409         return s && s->sort_flags & REF_SORTING_ICASE ?
2410                 strcasecmp(a->refname, b->refname) :
2411                 strcmp(a->refname, b->refname);
2412 }
2413
2414 void ref_sorting_set_sort_flags_all(struct ref_sorting *sorting,
2415                                     unsigned int mask, int on)
2416 {
2417         for (; sorting; sorting = sorting->next) {
2418                 if (on)
2419                         sorting->sort_flags |= mask;
2420                 else
2421                         sorting->sort_flags &= ~mask;
2422         }
2423 }
2424
2425 void ref_array_sort(struct ref_sorting *sorting, struct ref_array *array)
2426 {
2427         QSORT_S(array->items, array->nr, compare_refs, sorting);
2428 }
2429
2430 static void append_literal(const char *cp, const char *ep, struct ref_formatting_state *state)
2431 {
2432         struct strbuf *s = &state->stack->output;
2433
2434         while (*cp && (!ep || cp < ep)) {
2435                 if (*cp == '%') {
2436                         if (cp[1] == '%')
2437                                 cp++;
2438                         else {
2439                                 int ch = hex2chr(cp + 1);
2440                                 if (0 <= ch) {
2441                                         strbuf_addch(s, ch);
2442                                         cp += 3;
2443                                         continue;
2444                                 }
2445                         }
2446                 }
2447                 strbuf_addch(s, *cp);
2448                 cp++;
2449         }
2450 }
2451
2452 int format_ref_array_item(struct ref_array_item *info,
2453                            const struct ref_format *format,
2454                            struct strbuf *final_buf,
2455                            struct strbuf *error_buf)
2456 {
2457         const char *cp, *sp, *ep;
2458         struct ref_formatting_state state = REF_FORMATTING_STATE_INIT;
2459
2460         state.quote_style = format->quote_style;
2461         push_stack_element(&state.stack);
2462
2463         for (cp = format->format; *cp && (sp = find_next(cp)); cp = ep + 1) {
2464                 struct atom_value *atomv;
2465                 int pos;
2466
2467                 ep = strchr(sp, ')');
2468                 if (cp < sp)
2469                         append_literal(cp, sp, &state);
2470                 pos = parse_ref_filter_atom(format, sp + 2, ep, error_buf);
2471                 if (pos < 0 || get_ref_atom_value(info, pos, &atomv, error_buf) ||
2472                     atomv->handler(atomv, &state, error_buf)) {
2473                         pop_stack_element(&state.stack);
2474                         return -1;
2475                 }
2476         }
2477         if (*cp) {
2478                 sp = cp + strlen(cp);
2479                 append_literal(cp, sp, &state);
2480         }
2481         if (format->need_color_reset_at_eol) {
2482                 struct atom_value resetv;
2483                 resetv.s = GIT_COLOR_RESET;
2484                 if (append_atom(&resetv, &state, error_buf)) {
2485                         pop_stack_element(&state.stack);
2486                         return -1;
2487                 }
2488         }
2489         if (state.stack->prev) {
2490                 pop_stack_element(&state.stack);
2491                 return strbuf_addf_ret(error_buf, -1, _("format: %%(end) atom missing"));
2492         }
2493         strbuf_addbuf(final_buf, &state.stack->output);
2494         pop_stack_element(&state.stack);
2495         return 0;
2496 }
2497
2498 void show_ref_array_item(struct ref_array_item *info,
2499                          const struct ref_format *format)
2500 {
2501         struct strbuf final_buf = STRBUF_INIT;
2502         struct strbuf error_buf = STRBUF_INIT;
2503
2504         if (format_ref_array_item(info, format, &final_buf, &error_buf))
2505                 die("%s", error_buf.buf);
2506         fwrite(final_buf.buf, 1, final_buf.len, stdout);
2507         strbuf_release(&error_buf);
2508         strbuf_release(&final_buf);
2509         putchar('\n');
2510 }
2511
2512 void pretty_print_ref(const char *name, const struct object_id *oid,
2513                       const struct ref_format *format)
2514 {
2515         struct ref_array_item *ref_item;
2516         ref_item = new_ref_array_item(name, oid);
2517         ref_item->kind = ref_kind_from_refname(name);
2518         show_ref_array_item(ref_item, format);
2519         free_array_item(ref_item);
2520 }
2521
2522 static int parse_sorting_atom(const char *atom)
2523 {
2524         /*
2525          * This parses an atom using a dummy ref_format, since we don't
2526          * actually care about the formatting details.
2527          */
2528         struct ref_format dummy = REF_FORMAT_INIT;
2529         const char *end = atom + strlen(atom);
2530         struct strbuf err = STRBUF_INIT;
2531         int res = parse_ref_filter_atom(&dummy, atom, end, &err);
2532         if (res < 0)
2533                 die("%s", err.buf);
2534         strbuf_release(&err);
2535         return res;
2536 }
2537
2538 /*  If no sorting option is given, use refname to sort as default */
2539 struct ref_sorting *ref_default_sorting(void)
2540 {
2541         static const char cstr_name[] = "refname";
2542
2543         struct ref_sorting *sorting = xcalloc(1, sizeof(*sorting));
2544
2545         sorting->next = NULL;
2546         sorting->atom = parse_sorting_atom(cstr_name);
2547         return sorting;
2548 }
2549
2550 void parse_ref_sorting(struct ref_sorting **sorting_tail, const char *arg)
2551 {
2552         struct ref_sorting *s;
2553
2554         s = xcalloc(1, sizeof(*s));
2555         s->next = *sorting_tail;
2556         *sorting_tail = s;
2557
2558         if (*arg == '-') {
2559                 s->sort_flags |= REF_SORTING_REVERSE;
2560                 arg++;
2561         }
2562         if (skip_prefix(arg, "version:", &arg) ||
2563             skip_prefix(arg, "v:", &arg))
2564                 s->sort_flags |= REF_SORTING_VERSION;
2565         s->atom = parse_sorting_atom(arg);
2566 }
2567
2568 int parse_opt_ref_sorting(const struct option *opt, const char *arg, int unset)
2569 {
2570         /*
2571          * NEEDSWORK: We should probably clear the list in this case, but we've
2572          * already munged the global used_atoms list, which would need to be
2573          * undone.
2574          */
2575         BUG_ON_OPT_NEG(unset);
2576
2577         parse_ref_sorting(opt->value, arg);
2578         return 0;
2579 }
2580
2581 int parse_opt_merge_filter(const struct option *opt, const char *arg, int unset)
2582 {
2583         struct ref_filter *rf = opt->value;
2584         struct object_id oid;
2585         struct commit *merge_commit;
2586
2587         BUG_ON_OPT_NEG(unset);
2588
2589         if (get_oid(arg, &oid))
2590                 die(_("malformed object name %s"), arg);
2591
2592         merge_commit = lookup_commit_reference_gently(the_repository, &oid, 0);
2593
2594         if (!merge_commit)
2595                 return error(_("option `%s' must point to a commit"), opt->long_name);
2596
2597         if (starts_with(opt->long_name, "no"))
2598                 commit_list_insert(merge_commit, &rf->unreachable_from);
2599         else
2600                 commit_list_insert(merge_commit, &rf->reachable_from);
2601
2602         return 0;
2603 }