?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Documentation/technical: describe multi-pack reverse indexes
[git/debian.git] / pretty.c
blobb4ff3f602f9b17068f74b79ad7611e806e0e6c01
1 #include "cache.h"
2 #include "config.h"
3 #include "commit.h"
4 #include "utf8.h"
5 #include "diff.h"
6 #include "revision.h"
7 #include "string-list.h"
8 #include "mailmap.h"
9 #include "log-tree.h"
10 #include "notes.h"
11 #include "color.h"
12 #include "reflog-walk.h"
13 #include "gpg-interface.h"
14 #include "trailer.h"
15
16 static char *user_format;
17 static struct cmt_fmt_map {
18 const char *name;
19 enum cmit_fmt format;
20 int is_tformat;
21 int expand_tabs_in_log;
22 int is_alias;
23 enum date_mode_type default_date_mode_type;
24 const char *user_format;
25 } *commit_formats;
26 static size_t builtin_formats_len;
27 static size_t commit_formats_len;
28 static size_t commit_formats_alloc;
29 static struct cmt_fmt_map *find_commit_format(const char *sought);
30
31 int commit_format_is_empty(enum cmit_fmt fmt)
32 {
33 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
34 }
35
36 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
37 {
38 free(user_format);
39 user_format = xstrdup(cp);
40 if (is_tformat)
41 rev->use_terminator = 1;
42 rev->commit_format = CMIT_FMT_USERFORMAT;
43 }
44
45 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
46 {
47 struct cmt_fmt_map *commit_format = NULL;
48 const char *name;
49 const char *fmt;
50 int i;
51
52 if (!skip_prefix(var, "pretty.", &name))
53 return 0;
54
55 for (i = 0; i < builtin_formats_len; i++) {
56 if (!strcmp(commit_formats[i].name, name))
57 return 0;
58 }
59
60 for (i = builtin_formats_len; i < commit_formats_len; i++) {
61 if (!strcmp(commit_formats[i].name, name)) {
62 commit_format = &commit_formats[i];
63 break;
64 }
65 }
66
67 if (!commit_format) {
68 ALLOC_GROW(commit_formats, commit_formats_len+1,
69 commit_formats_alloc);
70 commit_format = &commit_formats[commit_formats_len];
71 memset(commit_format, 0, sizeof(*commit_format));
72 commit_formats_len++;
73 }
74
75 commit_format->name = xstrdup(name);
76 commit_format->format = CMIT_FMT_USERFORMAT;
77 if (git_config_string(&fmt, var, value))
78 return -1;
79
80 if (skip_prefix(fmt, "format:", &fmt))
81 commit_format->is_tformat = 0;
82 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
83 commit_format->is_tformat = 1;
84 else
85 commit_format->is_alias = 1;
86 commit_format->user_format = fmt;
87
88 return 0;
89 }
90
91 static void setup_commit_formats(void)
92 {
93 struct cmt_fmt_map builtin_formats[] = {
94 { "raw", CMIT_FMT_RAW, 0, 0 },
95 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
96 { "short", CMIT_FMT_SHORT, 0, 0 },
97 { "email", CMIT_FMT_EMAIL, 0, 0 },
98 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
99 { "fuller", CMIT_FMT_FULLER, 0, 8 },
100 { "full", CMIT_FMT_FULL, 0, 8 },
101 { "oneline", CMIT_FMT_ONELINE, 1, 0 },
102 { "reference", CMIT_FMT_USERFORMAT, 1, 0,
103 0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
104 /*
105 * Please update $__git_log_pretty_formats in
106 * git-completion.bash when you add new formats.
107 */
108 };
109 commit_formats_len = ARRAY_SIZE(builtin_formats);
110 builtin_formats_len = commit_formats_len;
111 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
112 COPY_ARRAY(commit_formats, builtin_formats,
113 ARRAY_SIZE(builtin_formats));
114
115 git_config(git_pretty_formats_config, NULL);
116 }
117
118 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
119 const char *original,
120 int num_redirections)
121 {
122 struct cmt_fmt_map *found = NULL;
123 size_t found_match_len = 0;
124 int i;
125
126 if (num_redirections >= commit_formats_len)
127 die("invalid --pretty format: "
128 "'%s' references an alias which points to itself",
129 original);
130
131 for (i = 0; i < commit_formats_len; i++) {
132 size_t match_len;
133
134 if (!starts_with(commit_formats[i].name, sought))
135 continue;
136
137 match_len = strlen(commit_formats[i].name);
138 if (found == NULL || found_match_len > match_len) {
139 found = &commit_formats[i];
140 found_match_len = match_len;
141 }
142 }
143
144 if (found && found->is_alias) {
145 found = find_commit_format_recursive(found->user_format,
146 original,
147 num_redirections+1);
148 }
149
150 return found;
151 }
152
153 static struct cmt_fmt_map *find_commit_format(const char *sought)
154 {
155 if (!commit_formats)
156 setup_commit_formats();
157
158 return find_commit_format_recursive(sought, sought, 0);
159 }
160
161 void get_commit_format(const char *arg, struct rev_info *rev)
162 {
163 struct cmt_fmt_map *commit_format;
164
165 rev->use_terminator = 0;
166 if (!arg) {
167 rev->commit_format = CMIT_FMT_DEFAULT;
168 return;
169 }
170 if (skip_prefix(arg, "format:", &arg)) {
171 save_user_format(rev, arg, 0);
172 return;
173 }
174
175 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
176 save_user_format(rev, arg, 1);
177 return;
178 }
179
180 commit_format = find_commit_format(arg);
181 if (!commit_format)
182 die("invalid --pretty format: %s", arg);
183
184 rev->commit_format = commit_format->format;
185 rev->use_terminator = commit_format->is_tformat;
186 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
187 if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
188 rev->date_mode.type = commit_format->default_date_mode_type;
189 if (commit_format->format == CMIT_FMT_USERFORMAT) {
190 save_user_format(rev, commit_format->user_format,
191 commit_format->is_tformat);
192 }
193 }
194
195 /*
196 * Generic support for pretty-printing the header
197 */
198 static int get_one_line(const char *msg)
199 {
200 int ret = 0;
201
202 for (;;) {
203 char c = *msg++;
204 if (!c)
205 break;
206 ret++;
207 if (c == '\n')
208 break;
209 }
210 return ret;
211 }
212
213 /* High bit set, or ISO-2022-INT */
214 static int non_ascii(int ch)
215 {
216 return !isascii(ch) || ch == '\033';
217 }
218
219 int has_non_ascii(const char *s)
220 {
221 int ch;
222 if (!s)
223 return 0;
224 while ((ch = *s++) != '\0') {
225 if (non_ascii(ch))
226 return 1;
227 }
228 return 0;
229 }
230
231 static int is_rfc822_special(char ch)
232 {
233 switch (ch) {
234 case '(':
235 case ')':
236 case '<':
237 case '>':
238 case '[':
239 case ']':
240 case ':':
241 case ';':
242 case '@':
243 case ',':
244 case '.':
245 case '"':
246 case '\\':
247 return 1;
248 default:
249 return 0;
250 }
251 }
252
253 static int needs_rfc822_quoting(const char *s, int len)
254 {
255 int i;
256 for (i = 0; i < len; i++)
257 if (is_rfc822_special(s[i]))
258 return 1;
259 return 0;
260 }
261
262 static int last_line_length(struct strbuf *sb)
263 {
264 int i;
265
266 /* How many bytes are already used on the last line? */
267 for (i = sb->len - 1; i >= 0; i--)
268 if (sb->buf[i] == '\n')
269 break;
270 return sb->len - (i + 1);
271 }
272
273 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
274 {
275 int i;
276
277 /* just a guess, we may have to also backslash-quote */
278 strbuf_grow(out, len + 2);
279
280 strbuf_addch(out, '"');
281 for (i = 0; i < len; i++) {
282 switch (s[i]) {
283 case '"':
284 case '\\':
285 strbuf_addch(out, '\\');
286 /* fall through */
287 default:
288 strbuf_addch(out, s[i]);
289 }
290 }
291 strbuf_addch(out, '"');
292 }
293
294 enum rfc2047_type {
295 RFC2047_SUBJECT,
296 RFC2047_ADDRESS
297 };
298
299 static int is_rfc2047_special(char ch, enum rfc2047_type type)
300 {
301 /*
302 * rfc2047, section 4.2:
303 *
304 * 8-bit values which correspond to printable ASCII characters other
305 * than "=", "?", and "_" (underscore), MAY be represented as those
306 * characters. (But see section 5 for restrictions.) In
307 * particular, SPACE and TAB MUST NOT be represented as themselves
308 * within encoded words.
309 */
310
311 /*
312 * rule out non-ASCII characters and non-printable characters (the
313 * non-ASCII check should be redundant as isprint() is not localized
314 * and only knows about ASCII, but be defensive about that)
315 */
316 if (non_ascii(ch) || !isprint(ch))
317 return 1;
318
319 /*
320 * rule out special printable characters (' ' should be the only
321 * whitespace character considered printable, but be defensive and use
322 * isspace())
323 */
324 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
325 return 1;
326
327 /*
328 * rfc2047, section 5.3:
329 *
330 * As a replacement for a 'word' entity within a 'phrase', for example,
331 * one that precedes an address in a From, To, or Cc header. The ABNF
332 * definition for 'phrase' from RFC 822 thus becomes:
333 *
334 * phrase = 1*( encoded-word / word )
335 *
336 * In this case the set of characters that may be used in a "Q"-encoded
337 * 'encoded-word' is restricted to: <upper and lower case ASCII
338 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
339 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
340 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
341 * 'special' by 'linear-white-space'.
342 */
343
344 if (type != RFC2047_ADDRESS)
345 return 0;
346
347 /* '=' and '_' are special cases and have been checked above */
348 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
349 }
350
351 static int needs_rfc2047_encoding(const char *line, int len)
352 {
353 int i;
354
355 for (i = 0; i < len; i++) {
356 int ch = line[i];
357 if (non_ascii(ch) || ch == '\n')
358 return 1;
359 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
360 return 1;
361 }
362
363 return 0;
364 }
365
366 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
367 const char *encoding, enum rfc2047_type type)
368 {
369 static const int max_encoded_length = 76; /* per rfc2047 */
370 int i;
371 int line_len = last_line_length(sb);
372
373 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
374 strbuf_addf(sb, "=?%s?q?", encoding);
375 line_len += strlen(encoding) + 5; /* 5 for =??q? */
376
377 while (len) {
378 /*
379 * RFC 2047, section 5 (3):
380 *
381 * Each 'encoded-word' MUST represent an integral number of
382 * characters. A multi-octet character may not be split across
383 * adjacent 'encoded- word's.
384 */
385 const unsigned char *p = (const unsigned char *)line;
386 int chrlen = mbs_chrlen(&line, &len, encoding);
387 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
388
389 /* "=%02X" * chrlen, or the byte itself */
390 const char *encoded_fmt = is_special ? "=%02X" : "%c";
391 int encoded_len = is_special ? 3 * chrlen : 1;
392
393 /*
394 * According to RFC 2047, we could encode the special character
395 * ' ' (space) with '_' (underscore) for readability. But many
396 * programs do not understand this and just leave the
397 * underscore in place. Thus, we do nothing special here, which
398 * causes ' ' to be encoded as '=20', avoiding this problem.
399 */
400
401 if (line_len + encoded_len + 2 > max_encoded_length) {
402 /* It won't fit with trailing "?=" --- break the line */
403 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
404 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
405 }
406
407 for (i = 0; i < chrlen; i++)
408 strbuf_addf(sb, encoded_fmt, p[i]);
409 line_len += encoded_len;
410 }
411 strbuf_addstr(sb, "?=");
412 }
413
414 const char *show_ident_date(const struct ident_split *ident,
415 const struct date_mode *mode)
416 {
417 timestamp_t date = 0;
418 long tz = 0;
419
420 if (ident->date_begin && ident->date_end)
421 date = parse_timestamp(ident->date_begin, NULL, 10);
422 if (date_overflows(date))
423 date = 0;
424 else {
425 if (ident->tz_begin && ident->tz_end)
426 tz = strtol(ident->tz_begin, NULL, 10);
427 if (tz >= INT_MAX || tz <= INT_MIN)
428 tz = 0;
429 }
430 return show_date(date, tz, mode);
431 }
432
433 void pp_user_info(struct pretty_print_context *pp,
434 const char *what, struct strbuf *sb,
435 const char *line, const char *encoding)
436 {
437 struct ident_split ident;
438 char *line_end;
439 const char *mailbuf, *namebuf;
440 size_t namelen, maillen;
441 int max_length = 78; /* per rfc2822 */
442
443 if (pp->fmt == CMIT_FMT_ONELINE)
444 return;
445
446 line_end = strchrnul(line, '\n');
447 if (split_ident_line(&ident, line, line_end - line))
448 return;
449
450 mailbuf = ident.mail_begin;
451 maillen = ident.mail_end - ident.mail_begin;
452 namebuf = ident.name_begin;
453 namelen = ident.name_end - ident.name_begin;
454
455 if (pp->mailmap)
456 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
457
458 if (cmit_fmt_is_mail(pp->fmt)) {
459 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
460 struct strbuf buf = STRBUF_INIT;
461
462 strbuf_addstr(&buf, "From: ");
463 strbuf_add(&buf, namebuf, namelen);
464 strbuf_addstr(&buf, " <");
465 strbuf_add(&buf, mailbuf, maillen);
466 strbuf_addstr(&buf, ">\n");
467 string_list_append(&pp->in_body_headers,
468 strbuf_detach(&buf, NULL));
469
470 mailbuf = pp->from_ident->mail_begin;
471 maillen = pp->from_ident->mail_end - mailbuf;
472 namebuf = pp->from_ident->name_begin;
473 namelen = pp->from_ident->name_end - namebuf;
474 }
475
476 strbuf_addstr(sb, "From: ");
477 if (pp->encode_email_headers &&
478 needs_rfc2047_encoding(namebuf, namelen)) {
479 add_rfc2047(sb, namebuf, namelen,
480 encoding, RFC2047_ADDRESS);
481 max_length = 76; /* per rfc2047 */
482 } else if (needs_rfc822_quoting(namebuf, namelen)) {
483 struct strbuf quoted = STRBUF_INIT;
484 add_rfc822_quoted(&quoted, namebuf, namelen);
485 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
486 -6, 1, max_length);
487 strbuf_release(&quoted);
488 } else {
489 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
490 -6, 1, max_length);
491 }
492
493 if (max_length <
494 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
495 strbuf_addch(sb, '\n');
496 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
497 } else {
498 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
499 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
500 (int)namelen, namebuf, (int)maillen, mailbuf);
501 }
502
503 switch (pp->fmt) {
504 case CMIT_FMT_MEDIUM:
505 strbuf_addf(sb, "Date: %s\n",
506 show_ident_date(&ident, &pp->date_mode));
507 break;
508 case CMIT_FMT_EMAIL:
509 case CMIT_FMT_MBOXRD:
510 strbuf_addf(sb, "Date: %s\n",
511 show_ident_date(&ident, DATE_MODE(RFC2822)));
512 break;
513 case CMIT_FMT_FULLER:
514 strbuf_addf(sb, "%sDate: %s\n", what,
515 show_ident_date(&ident, &pp->date_mode));
516 break;
517 default:
518 /* notin' */
519 break;
520 }
521 }
522
523 static int is_blank_line(const char *line, int *len_p)
524 {
525 int len = *len_p;
526 while (len && isspace(line[len - 1]))
527 len--;
528 *len_p = len;
529 return !len;
530 }
531
532 const char *skip_blank_lines(const char *msg)
533 {
534 for (;;) {
535 int linelen = get_one_line(msg);
536 int ll = linelen;
537 if (!linelen)
538 break;
539 if (!is_blank_line(msg, &ll))
540 break;
541 msg += linelen;
542 }
543 return msg;
544 }
545
546 static void add_merge_info(const struct pretty_print_context *pp,
547 struct strbuf *sb, const struct commit *commit)
548 {
549 struct commit_list *parent = commit->parents;
550
551 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
552 !parent || !parent->next)
553 return;
554
555 strbuf_addstr(sb, "Merge:");
556
557 while (parent) {
558 struct object_id *oidp = &parent->item->object.oid;
559 strbuf_addch(sb, ' ');
560 if (pp->abbrev)
561 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
562 else
563 strbuf_addstr(sb, oid_to_hex(oidp));
564 parent = parent->next;
565 }
566 strbuf_addch(sb, '\n');
567 }
568
569 static char *get_header(const char *msg, const char *key)
570 {
571 size_t len;
572 const char *v = find_commit_header(msg, key, &len);
573 return v ? xmemdupz(v, len) : NULL;
574 }
575
576 static char *replace_encoding_header(char *buf, const char *encoding)
577 {
578 struct strbuf tmp = STRBUF_INIT;
579 size_t start, len;
580 char *cp = buf;
581
582 /* guess if there is an encoding header before a \n\n */
583 while (!starts_with(cp, "encoding ")) {
584 cp = strchr(cp, '\n');
585 if (!cp || *++cp == '\n')
586 return buf;
587 }
588 start = cp - buf;
589 cp = strchr(cp, '\n');
590 if (!cp)
591 return buf; /* should not happen but be defensive */
592 len = cp + 1 - (buf + start);
593
594 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
595 if (is_encoding_utf8(encoding)) {
596 /* we have re-coded to UTF-8; drop the header */
597 strbuf_remove(&tmp, start, len);
598 } else {
599 /* just replaces XXXX in 'encoding XXXX\n' */
600 strbuf_splice(&tmp, start + strlen("encoding "),
601 len - strlen("encoding \n"),
602 encoding, strlen(encoding));
603 }
604 return strbuf_detach(&tmp, NULL);
605 }
606
607 const char *repo_logmsg_reencode(struct repository *r,
608 const struct commit *commit,
609 char **commit_encoding,
610 const char *output_encoding)
611 {
612 static const char *utf8 = "UTF-8";
613 const char *use_encoding;
614 char *encoding;
615 const char *msg = repo_get_commit_buffer(r, commit, NULL);
616 char *out;
617
618 if (!output_encoding || !*output_encoding) {
619 if (commit_encoding)
620 *commit_encoding = get_header(msg, "encoding");
621 return msg;
622 }
623 encoding = get_header(msg, "encoding");
624 if (commit_encoding)
625 *commit_encoding = encoding;
626 use_encoding = encoding ? encoding : utf8;
627 if (same_encoding(use_encoding, output_encoding)) {
628 /*
629 * No encoding work to be done. If we have no encoding header
630 * at all, then there's nothing to do, and we can return the
631 * message verbatim (whether newly allocated or not).
632 */
633 if (!encoding)
634 return msg;
635
636 /*
637 * Otherwise, we still want to munge the encoding header in the
638 * result, which will be done by modifying the buffer. If we
639 * are using a fresh copy, we can reuse it. But if we are using
640 * the cached copy from get_commit_buffer, we need to duplicate it
641 * to avoid munging the cached copy.
642 */
643 if (msg == get_cached_commit_buffer(r, commit, NULL))
644 out = xstrdup(msg);
645 else
646 out = (char *)msg;
647 }
648 else {
649 /*
650 * There's actual encoding work to do. Do the reencoding, which
651 * still leaves the header to be replaced in the next step. At
652 * this point, we are done with msg. If we allocated a fresh
653 * copy, we can free it.
654 */
655 out = reencode_string(msg, output_encoding, use_encoding);
656 if (out)
657 repo_unuse_commit_buffer(r, commit, msg);
658 }
659
660 /*
661 * This replacement actually consumes the buffer we hand it, so we do
662 * not have to worry about freeing the old "out" here.
663 */
664 if (out)
665 out = replace_encoding_header(out, output_encoding);
666
667 if (!commit_encoding)
668 free(encoding);
669 /*
670 * If the re-encoding failed, out might be NULL here; in that
671 * case we just return the commit message verbatim.
672 */
673 return out ? out : msg;
674 }
675
676 static int mailmap_name(const char **email, size_t *email_len,
677 const char **name, size_t *name_len)
678 {
679 static struct string_list *mail_map;
680 if (!mail_map) {
681 mail_map = xcalloc(1, sizeof(*mail_map));
682 read_mailmap(mail_map);
683 }
684 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
685 }
686
687 static size_t format_person_part(struct strbuf *sb, char part,
688 const char *msg, int len,
689 const struct date_mode *dmode)
690 {
691 /* currently all placeholders have same length */
692 const int placeholder_len = 2;
693 struct ident_split s;
694 const char *name, *mail;
695 size_t maillen, namelen;
696
697 if (split_ident_line(&s, msg, len) < 0)
698 goto skip;
699
700 name = s.name_begin;
701 namelen = s.name_end - s.name_begin;
702 mail = s.mail_begin;
703 maillen = s.mail_end - s.mail_begin;
704
705 if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
706 mailmap_name(&mail, &maillen, &name, &namelen);
707 if (part == 'n' || part == 'N') { /* name */
708 strbuf_add(sb, name, namelen);
709 return placeholder_len;
710 }
711 if (part == 'e' || part == 'E') { /* email */
712 strbuf_add(sb, mail, maillen);
713 return placeholder_len;
714 }
715 if (part == 'l' || part == 'L') { /* local-part */
716 const char *at = memchr(mail, '@', maillen);
717 if (at)
718 maillen = at - mail;
719 strbuf_add(sb, mail, maillen);
720 return placeholder_len;
721 }
722
723 if (!s.date_begin)
724 goto skip;
725
726 if (part == 't') { /* date, UNIX timestamp */
727 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
728 return placeholder_len;
729 }
730
731 switch (part) {
732 case 'd': /* date */
733 strbuf_addstr(sb, show_ident_date(&s, dmode));
734 return placeholder_len;
735 case 'D': /* date, RFC2822 style */
736 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
737 return placeholder_len;
738 case 'r': /* date, relative */
739 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
740 return placeholder_len;
741 case 'i': /* date, ISO 8601-like */
742 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
743 return placeholder_len;
744 case 'I': /* date, ISO 8601 strict */
745 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
746 return placeholder_len;
747 case 's':
748 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
749 return placeholder_len;
750 }
751
752 skip:
753 /*
754 * reading from either a bogus commit, or a reflog entry with
755 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
756 * to compute a valid return value.
757 */
758 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
759 || part == 'D' || part == 'r' || part == 'i')
760 return placeholder_len;
761
762 return 0; /* unknown placeholder */
763 }
764
765 struct chunk {
766 size_t off;
767 size_t len;
768 };
769
770 enum flush_type {
771 no_flush,
772 flush_right,
773 flush_left,
774 flush_left_and_steal,
775 flush_both
776 };
777
778 enum trunc_type {
779 trunc_none,
780 trunc_left,
781 trunc_middle,
782 trunc_right
783 };
784
785 struct format_commit_context {
786 struct repository *repository;
787 const struct commit *commit;
788 const struct pretty_print_context *pretty_ctx;
789 unsigned commit_header_parsed:1;
790 unsigned commit_message_parsed:1;
791 struct signature_check signature_check;
792 enum flush_type flush_type;
793 enum trunc_type truncate;
794 const char *message;
795 char *commit_encoding;
796 size_t width, indent1, indent2;
797 int auto_color;
798 int padding;
799
800 /* These offsets are relative to the start of the commit message. */
801 struct chunk author;
802 struct chunk committer;
803 size_t message_off;
804 size_t subject_off;
805 size_t body_off;
806
807 /* The following ones are relative to the result struct strbuf. */
808 size_t wrap_start;
809 };
810
811 static void parse_commit_header(struct format_commit_context *context)
812 {
813 const char *msg = context->message;
814 int i;
815
816 for (i = 0; msg[i]; i++) {
817 const char *name;
818 int eol;
819 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
820 ; /* do nothing */
821
822 if (i == eol) {
823 break;
824 } else if (skip_prefix(msg + i, "author ", &name)) {
825 context->author.off = name - msg;
826 context->author.len = msg + eol - name;
827 } else if (skip_prefix(msg + i, "committer ", &name)) {
828 context->committer.off = name - msg;
829 context->committer.len = msg + eol - name;
830 }
831 i = eol;
832 }
833 context->message_off = i;
834 context->commit_header_parsed = 1;
835 }
836
837 static int istitlechar(char c)
838 {
839 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
840 (c >= '0' && c <= '9') || c == '.' || c == '_';
841 }
842
843 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
844 {
845 size_t trimlen;
846 size_t start_len = sb->len;
847 int space = 2;
848 int i;
849
850 for (i = 0; i < len; i++) {
851 if (istitlechar(msg[i])) {
852 if (space == 1)
853 strbuf_addch(sb, '-');
854 space = 0;
855 strbuf_addch(sb, msg[i]);
856 if (msg[i] == '.')
857 while (msg[i+1] == '.')
858 i++;
859 } else
860 space |= 1;
861 }
862
863 /* trim any trailing '.' or '-' characters */
864 trimlen = 0;
865 while (sb->len - trimlen > start_len &&
866 (sb->buf[sb->len - 1 - trimlen] == '.'
867 || sb->buf[sb->len - 1 - trimlen] == '-'))
868 trimlen++;
869 strbuf_remove(sb, sb->len - trimlen, trimlen);
870 }
871
872 const char *format_subject(struct strbuf *sb, const char *msg,
873 const char *line_separator)
874 {
875 int first = 1;
876
877 for (;;) {
878 const char *line = msg;
879 int linelen = get_one_line(line);
880
881 msg += linelen;
882 if (!linelen || is_blank_line(line, &linelen))
883 break;
884
885 if (!sb)
886 continue;
887 strbuf_grow(sb, linelen + 2);
888 if (!first)
889 strbuf_addstr(sb, line_separator);
890 strbuf_add(sb, line, linelen);
891 first = 0;
892 }
893 return msg;
894 }
895
896 static void parse_commit_message(struct format_commit_context *c)
897 {
898 const char *msg = c->message + c->message_off;
899 const char *start = c->message;
900
901 msg = skip_blank_lines(msg);
902 c->subject_off = msg - start;
903
904 msg = format_subject(NULL, msg, NULL);
905 msg = skip_blank_lines(msg);
906 c->body_off = msg - start;
907
908 c->commit_message_parsed = 1;
909 }
910
911 static void strbuf_wrap(struct strbuf *sb, size_t pos,
912 size_t width, size_t indent1, size_t indent2)
913 {
914 struct strbuf tmp = STRBUF_INIT;
915
916 if (pos)
917 strbuf_add(&tmp, sb->buf, pos);
918 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
919 (int) indent1, (int) indent2, (int) width);
920 strbuf_swap(&tmp, sb);
921 strbuf_release(&tmp);
922 }
923
924 static void rewrap_message_tail(struct strbuf *sb,
925 struct format_commit_context *c,
926 size_t new_width, size_t new_indent1,
927 size_t new_indent2)
928 {
929 if (c->width == new_width && c->indent1 == new_indent1 &&
930 c->indent2 == new_indent2)
931 return;
932 if (c->wrap_start < sb->len)
933 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
934 c->wrap_start = sb->len;
935 c->width = new_width;
936 c->indent1 = new_indent1;
937 c->indent2 = new_indent2;
938 }
939
940 static int format_reflog_person(struct strbuf *sb,
941 char part,
942 struct reflog_walk_info *log,
943 const struct date_mode *dmode)
944 {
945 const char *ident;
946
947 if (!log)
948 return 2;
949
950 ident = get_reflog_ident(log);
951 if (!ident)
952 return 2;
953
954 return format_person_part(sb, part, ident, strlen(ident), dmode);
955 }
956
957 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
958 const char *placeholder,
959 struct format_commit_context *c)
960 {
961 const char *rest = placeholder;
962 const char *basic_color = NULL;
963
964 if (placeholder[1] == '(') {
965 const char *begin = placeholder + 2;
966 const char *end = strchr(begin, ')');
967 char color[COLOR_MAXLEN];
968
969 if (!end)
970 return 0;
971
972 if (skip_prefix(begin, "auto,", &begin)) {
973 if (!want_color(c->pretty_ctx->color))
974 return end - placeholder + 1;
975 } else if (skip_prefix(begin, "always,", &begin)) {
976 /* nothing to do; we do not respect want_color at all */
977 } else {
978 /* the default is the same as "auto" */
979 if (!want_color(c->pretty_ctx->color))
980 return end - placeholder + 1;
981 }
982
983 if (color_parse_mem(begin, end - begin, color) < 0)
984 die(_("unable to parse --pretty format"));
985 strbuf_addstr(sb, color);
986 return end - placeholder + 1;
987 }
988
989 /*
990 * We handle things like "%C(red)" above; for historical reasons, there
991 * are a few colors that can be specified without parentheses (and
992 * they cannot support things like "auto" or "always" at all).
993 */
994 if (skip_prefix(placeholder + 1, "red", &rest))
995 basic_color = GIT_COLOR_RED;
996 else if (skip_prefix(placeholder + 1, "green", &rest))
997 basic_color = GIT_COLOR_GREEN;
998 else if (skip_prefix(placeholder + 1, "blue", &rest))
999 basic_color = GIT_COLOR_BLUE;
1000 else if (skip_prefix(placeholder + 1, "reset", &rest))
1001 basic_color = GIT_COLOR_RESET;
1002
1003 if (basic_color && want_color(c->pretty_ctx->color))
1004 strbuf_addstr(sb, basic_color);
1005
1006 return rest - placeholder;
1007 }
1008
1009 static size_t parse_padding_placeholder(const char *placeholder,
1010 struct format_commit_context *c)
1011 {
1012 const char *ch = placeholder;
1013 enum flush_type flush_type;
1014 int to_column = 0;
1015
1016 switch (*ch++) {
1017 case '<':
1018 flush_type = flush_right;
1019 break;
1020 case '>':
1021 if (*ch == '<') {
1022 flush_type = flush_both;
1023 ch++;
1024 } else if (*ch == '>') {
1025 flush_type = flush_left_and_steal;
1026 ch++;
1027 } else
1028 flush_type = flush_left;
1029 break;
1030 default:
1031 return 0;
1032 }
1033
1034 /* the next value means "wide enough to that column" */
1035 if (*ch == '|') {
1036 to_column = 1;
1037 ch++;
1038 }
1039
1040 if (*ch == '(') {
1041 const char *start = ch + 1;
1042 const char *end = start + strcspn(start, ",)");
1043 char *next;
1044 int width;
1045 if (!end || end == start)
1046 return 0;
1047 width = strtol(start, &next, 10);
1048 if (next == start || width == 0)
1049 return 0;
1050 if (width < 0) {
1051 if (to_column)
1052 width += term_columns();
1053 if (width < 0)
1054 return 0;
1055 }
1056 c->padding = to_column ? -width : width;
1057 c->flush_type = flush_type;
1058
1059 if (*end == ',') {
1060 start = end + 1;
1061 end = strchr(start, ')');
1062 if (!end || end == start)
1063 return 0;
1064 if (starts_with(start, "trunc)"))
1065 c->truncate = trunc_right;
1066 else if (starts_with(start, "ltrunc)"))
1067 c->truncate = trunc_left;
1068 else if (starts_with(start, "mtrunc)"))
1069 c->truncate = trunc_middle;
1070 else
1071 return 0;
1072 } else
1073 c->truncate = trunc_none;
1074
1075 return end - placeholder + 1;
1076 }
1077 return 0;
1078 }
1079
1080 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1081 const char **end, const char **valuestart,
1082 size_t *valuelen)
1083 {
1084 const char *p;
1085
1086 if (!(skip_prefix(to_parse, candidate, &p)))
1087 return 0;
1088 if (valuestart) {
1089 if (*p == '=') {
1090 *valuestart = p + 1;
1091 *valuelen = strcspn(*valuestart, ",)");
1092 p = *valuestart + *valuelen;
1093 } else {
1094 if (*p != ',' && *p != ')')
1095 return 0;
1096 *valuestart = NULL;
1097 *valuelen = 0;
1098 }
1099 }
1100 if (*p == ',') {
1101 *end = p + 1;
1102 return 1;
1103 }
1104 if (*p == ')') {
1105 *end = p;
1106 return 1;
1107 }
1108 return 0;
1109 }
1110
1111 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1112 const char **end, int *val)
1113 {
1114 const char *argval;
1115 char *strval;
1116 size_t arglen;
1117 int v;
1118
1119 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1120 return 0;
1121
1122 if (!argval) {
1123 *val = 1;
1124 return 1;
1125 }
1126
1127 strval = xstrndup(argval, arglen);
1128 v = git_parse_maybe_bool(strval);
1129 free(strval);
1130
1131 if (v == -1)
1132 return 0;
1133
1134 *val = v;
1135
1136 return 1;
1137 }
1138
1139 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1140 {
1141 const struct string_list *list = ud;
1142 const struct string_list_item *item;
1143
1144 for_each_string_list_item (item, list) {
1145 if (key->len == (uintptr_t)item->util &&
1146 !strncasecmp(item->string, key->buf, key->len))
1147 return 1;
1148 }
1149 return 0;
1150 }
1151
1152 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1153 const char *placeholder,
1154 void *context)
1155 {
1156 struct format_commit_context *c = context;
1157 const struct commit *commit = c->commit;
1158 const char *msg = c->message;
1159 struct commit_list *p;
1160 const char *arg, *eol;
1161 size_t res;
1162 char **slot;
1163
1164 /* these are independent of the commit */
1165 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1166 if (res)
1167 return res;
1168
1169 switch (placeholder[0]) {
1170 case 'C':
1171 if (starts_with(placeholder + 1, "(auto)")) {
1172 c->auto_color = want_color(c->pretty_ctx->color);
1173 if (c->auto_color && sb->len)
1174 strbuf_addstr(sb, GIT_COLOR_RESET);
1175 return 7; /* consumed 7 bytes, "C(auto)" */
1176 } else {
1177 int ret = parse_color(sb, placeholder, c);
1178 if (ret)
1179 c->auto_color = 0;
1180 /*
1181 * Otherwise, we decided to treat %C<unknown>
1182 * as a literal string, and the previous
1183 * %C(auto) is still valid.
1184 */
1185 return ret;
1186 }
1187 case 'w':
1188 if (placeholder[1] == '(') {
1189 unsigned long width = 0, indent1 = 0, indent2 = 0;
1190 char *next;
1191 const char *start = placeholder + 2;
1192 const char *end = strchr(start, ')');
1193 if (!end)
1194 return 0;
1195 if (end > start) {
1196 width = strtoul(start, &next, 10);
1197 if (*next == ',') {
1198 indent1 = strtoul(next + 1, &next, 10);
1199 if (*next == ',') {
1200 indent2 = strtoul(next + 1,
1201 &next, 10);
1202 }
1203 }
1204 if (*next != ')')
1205 return 0;
1206 }
1207 rewrap_message_tail(sb, c, width, indent1, indent2);
1208 return end - placeholder + 1;
1209 } else
1210 return 0;
1211
1212 case '<':
1213 case '>':
1214 return parse_padding_placeholder(placeholder, c);
1215 }
1216
1217 /* these depend on the commit */
1218 if (!commit->object.parsed)
1219 parse_object(the_repository, &commit->object.oid);
1220
1221 switch (placeholder[0]) {
1222 case 'H': /* commit hash */
1223 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1224 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1225 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1226 return 1;
1227 case 'h': /* abbreviated commit hash */
1228 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1229 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1230 c->pretty_ctx->abbrev);
1231 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1232 return 1;
1233 case 'T': /* tree hash */
1234 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1235 return 1;
1236 case 't': /* abbreviated tree hash */
1237 strbuf_add_unique_abbrev(sb,
1238 get_commit_tree_oid(commit),
1239 c->pretty_ctx->abbrev);
1240 return 1;
1241 case 'P': /* parent hashes */
1242 for (p = commit->parents; p; p = p->next) {
1243 if (p != commit->parents)
1244 strbuf_addch(sb, ' ');
1245 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1246 }
1247 return 1;
1248 case 'p': /* abbreviated parent hashes */
1249 for (p = commit->parents; p; p = p->next) {
1250 if (p != commit->parents)
1251 strbuf_addch(sb, ' ');
1252 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1253 c->pretty_ctx->abbrev);
1254 }
1255 return 1;
1256 case 'm': /* left/right/bottom */
1257 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1258 return 1;
1259 case 'd':
1260 format_decorations(sb, commit, c->auto_color);
1261 return 1;
1262 case 'D':
1263 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1264 return 1;
1265 case 'S': /* tag/branch like --source */
1266 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1267 return 0;
1268 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1269 if (!(slot && *slot))
1270 return 0;
1271 strbuf_addstr(sb, *slot);
1272 return 1;
1273 case 'g': /* reflog info */
1274 switch(placeholder[1]) {
1275 case 'd': /* reflog selector */
1276 case 'D':
1277 if (c->pretty_ctx->reflog_info)
1278 get_reflog_selector(sb,
1279 c->pretty_ctx->reflog_info,
1280 &c->pretty_ctx->date_mode,
1281 c->pretty_ctx->date_mode_explicit,
1282 (placeholder[1] == 'd'));
1283 return 2;
1284 case 's': /* reflog message */
1285 if (c->pretty_ctx->reflog_info)
1286 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1287 return 2;
1288 case 'n':
1289 case 'N':
1290 case 'e':
1291 case 'E':
1292 return format_reflog_person(sb,
1293 placeholder[1],
1294 c->pretty_ctx->reflog_info,
1295 &c->pretty_ctx->date_mode);
1296 }
1297 return 0; /* unknown %g placeholder */
1298 case 'N':
1299 if (c->pretty_ctx->notes_message) {
1300 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1301 return 1;
1302 }
1303 return 0;
1304 }
1305
1306 if (placeholder[0] == 'G') {
1307 if (!c->signature_check.result)
1308 check_commit_signature(c->commit, &(c->signature_check));
1309 switch (placeholder[1]) {
1310 case 'G':
1311 if (c->signature_check.gpg_output)
1312 strbuf_addstr(sb, c->signature_check.gpg_output);
1313 break;
1314 case '?':
1315 switch (c->signature_check.result) {
1316 case 'G':
1317 switch (c->signature_check.trust_level) {
1318 case TRUST_UNDEFINED:
1319 case TRUST_NEVER:
1320 strbuf_addch(sb, 'U');
1321 break;
1322 default:
1323 strbuf_addch(sb, 'G');
1324 break;
1325 }
1326 break;
1327 case 'B':
1328 case 'E':
1329 case 'N':
1330 case 'X':
1331 case 'Y':
1332 case 'R':
1333 strbuf_addch(sb, c->signature_check.result);
1334 }
1335 break;
1336 case 'S':
1337 if (c->signature_check.signer)
1338 strbuf_addstr(sb, c->signature_check.signer);
1339 break;
1340 case 'K':
1341 if (c->signature_check.key)
1342 strbuf_addstr(sb, c->signature_check.key);
1343 break;
1344 case 'F':
1345 if (c->signature_check.fingerprint)
1346 strbuf_addstr(sb, c->signature_check.fingerprint);
1347 break;
1348 case 'P':
1349 if (c->signature_check.primary_key_fingerprint)
1350 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1351 break;
1352 case 'T':
1353 switch (c->signature_check.trust_level) {
1354 case TRUST_UNDEFINED:
1355 strbuf_addstr(sb, "undefined");
1356 break;
1357 case TRUST_NEVER:
1358 strbuf_addstr(sb, "never");
1359 break;
1360 case TRUST_MARGINAL:
1361 strbuf_addstr(sb, "marginal");
1362 break;
1363 case TRUST_FULLY:
1364 strbuf_addstr(sb, "fully");
1365 break;
1366 case TRUST_ULTIMATE:
1367 strbuf_addstr(sb, "ultimate");
1368 break;
1369 }
1370 break;
1371 default:
1372 return 0;
1373 }
1374 return 2;
1375 }
1376
1377 /* For the rest we have to parse the commit header. */
1378 if (!c->commit_header_parsed) {
1379 msg = c->message =
1380 repo_logmsg_reencode(c->repository, commit,
1381 &c->commit_encoding, "UTF-8");
1382 parse_commit_header(c);
1383 }
1384
1385 switch (placeholder[0]) {
1386 case 'a': /* author ... */
1387 return format_person_part(sb, placeholder[1],
1388 msg + c->author.off, c->author.len,
1389 &c->pretty_ctx->date_mode);
1390 case 'c': /* committer ... */
1391 return format_person_part(sb, placeholder[1],
1392 msg + c->committer.off, c->committer.len,
1393 &c->pretty_ctx->date_mode);
1394 case 'e': /* encoding */
1395 if (c->commit_encoding)
1396 strbuf_addstr(sb, c->commit_encoding);
1397 return 1;
1398 case 'B': /* raw body */
1399 /* message_off is always left at the initial newline */
1400 strbuf_addstr(sb, msg + c->message_off + 1);
1401 return 1;
1402 }
1403
1404 /* Now we need to parse the commit message. */
1405 if (!c->commit_message_parsed)
1406 parse_commit_message(c);
1407
1408 switch (placeholder[0]) {
1409 case 's': /* subject */
1410 format_subject(sb, msg + c->subject_off, " ");
1411 return 1;
1412 case 'f': /* sanitized subject */
1413 eol = strchrnul(msg + c->subject_off, '\n');
1414 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1415 return 1;
1416 case 'b': /* body */
1417 strbuf_addstr(sb, msg + c->body_off);
1418 return 1;
1419 }
1420
1421 if (skip_prefix(placeholder, "(trailers", &arg)) {
1422 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1423 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1424 struct strbuf sepbuf = STRBUF_INIT;
1425 struct strbuf kvsepbuf = STRBUF_INIT;
1426 size_t ret = 0;
1427
1428 opts.no_divider = 1;
1429
1430 if (*arg == ':') {
1431 arg++;
1432 for (;;) {
1433 const char *argval;
1434 size_t arglen;
1435
1436 if (match_placeholder_arg_value(arg, "key", &arg, &argval, &arglen)) {
1437 uintptr_t len = arglen;
1438
1439 if (!argval)
1440 goto trailer_out;
1441
1442 if (len && argval[len - 1] == ':')
1443 len--;
1444 string_list_append(&filter_list, argval)->util = (char *)len;
1445
1446 opts.filter = format_trailer_match_cb;
1447 opts.filter_data = &filter_list;
1448 opts.only_trailers = 1;
1449 } else if (match_placeholder_arg_value(arg, "separator", &arg, &argval, &arglen)) {
1450 char *fmt;
1451
1452 strbuf_reset(&sepbuf);
1453 fmt = xstrndup(argval, arglen);
1454 strbuf_expand(&sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1455 free(fmt);
1456 opts.separator = &sepbuf;
1457 } else if (match_placeholder_arg_value(arg, "key_value_separator", &arg, &argval, &arglen)) {
1458 char *fmt;
1459
1460 strbuf_reset(&kvsepbuf);
1461 fmt = xstrndup(argval, arglen);
1462 strbuf_expand(&kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1463 free(fmt);
1464 opts.key_value_separator = &kvsepbuf;
1465 } else if (!match_placeholder_bool_arg(arg, "only", &arg, &opts.only_trailers) &&
1466 !match_placeholder_bool_arg(arg, "unfold", &arg, &opts.unfold) &&
1467 !match_placeholder_bool_arg(arg, "keyonly", &arg, &opts.key_only) &&
1468 !match_placeholder_bool_arg(arg, "valueonly", &arg, &opts.value_only))
1469 break;
1470 }
1471 }
1472 if (*arg == ')') {
1473 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1474 ret = arg - placeholder + 1;
1475 }
1476 trailer_out:
1477 string_list_clear(&filter_list, 0);
1478 strbuf_release(&sepbuf);
1479 return ret;
1480 }
1481
1482 return 0; /* unknown placeholder */
1483 }
1484
1485 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1486 const char *placeholder,
1487 struct format_commit_context *c)
1488 {
1489 struct strbuf local_sb = STRBUF_INIT;
1490 int total_consumed = 0, len, padding = c->padding;
1491 if (padding < 0) {
1492 const char *start = strrchr(sb->buf, '\n');
1493 int occupied;
1494 if (!start)
1495 start = sb->buf;
1496 occupied = utf8_strnwidth(start, -1, 1);
1497 occupied += c->pretty_ctx->graph_width;
1498 padding = (-padding) - occupied;
1499 }
1500 while (1) {
1501 int modifier = *placeholder == 'C';
1502 int consumed = format_commit_one(&local_sb, placeholder, c);
1503 total_consumed += consumed;
1504
1505 if (!modifier)
1506 break;
1507
1508 placeholder += consumed;
1509 if (*placeholder != '%')
1510 break;
1511 placeholder++;
1512 total_consumed++;
1513 }
1514 len = utf8_strnwidth(local_sb.buf, -1, 1);
1515
1516 if (c->flush_type == flush_left_and_steal) {
1517 const char *ch = sb->buf + sb->len - 1;
1518 while (len > padding && ch > sb->buf) {
1519 const char *p;
1520 if (*ch == ' ') {
1521 ch--;
1522 padding++;
1523 continue;
1524 }
1525 /* check for trailing ansi sequences */
1526 if (*ch != 'm')
1527 break;
1528 p = ch - 1;
1529 while (ch - p < 10 && *p != '\033')
1530 p--;
1531 if (*p != '\033' ||
1532 ch + 1 - p != display_mode_esc_sequence_len(p))
1533 break;
1534 /*
1535 * got a good ansi sequence, put it back to
1536 * local_sb as we're cutting sb
1537 */
1538 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1539 ch = p - 1;
1540 }
1541 strbuf_setlen(sb, ch + 1 - sb->buf);
1542 c->flush_type = flush_left;
1543 }
1544
1545 if (len > padding) {
1546 switch (c->truncate) {
1547 case trunc_left:
1548 strbuf_utf8_replace(&local_sb,
1549 0, len - (padding - 2),
1550 "..");
1551 break;
1552 case trunc_middle:
1553 strbuf_utf8_replace(&local_sb,
1554 padding / 2 - 1,
1555 len - (padding - 2),
1556 "..");
1557 break;
1558 case trunc_right:
1559 strbuf_utf8_replace(&local_sb,
1560 padding - 2, len - (padding - 2),
1561 "..");
1562 break;
1563 case trunc_none:
1564 break;
1565 }
1566 strbuf_addbuf(sb, &local_sb);
1567 } else {
1568 int sb_len = sb->len, offset = 0;
1569 if (c->flush_type == flush_left)
1570 offset = padding - len;
1571 else if (c->flush_type == flush_both)
1572 offset = (padding - len) / 2;
1573 /*
1574 * we calculate padding in columns, now
1575 * convert it back to chars
1576 */
1577 padding = padding - len + local_sb.len;
1578 strbuf_addchars(sb, ' ', padding);
1579 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1580 local_sb.len);
1581 }
1582 strbuf_release(&local_sb);
1583 c->flush_type = no_flush;
1584 return total_consumed;
1585 }
1586
1587 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1588 const char *placeholder,
1589 void *context)
1590 {
1591 int consumed;
1592 size_t orig_len;
1593 enum {
1594 NO_MAGIC,
1595 ADD_LF_BEFORE_NON_EMPTY,
1596 DEL_LF_BEFORE_EMPTY,
1597 ADD_SP_BEFORE_NON_EMPTY
1598 } magic = NO_MAGIC;
1599
1600 switch (placeholder[0]) {
1601 case '-':
1602 magic = DEL_LF_BEFORE_EMPTY;
1603 break;
1604 case '+':
1605 magic = ADD_LF_BEFORE_NON_EMPTY;
1606 break;
1607 case ' ':
1608 magic = ADD_SP_BEFORE_NON_EMPTY;
1609 break;
1610 default:
1611 break;
1612 }
1613 if (magic != NO_MAGIC)
1614 placeholder++;
1615
1616 orig_len = sb->len;
1617 if (((struct format_commit_context *)context)->flush_type != no_flush)
1618 consumed = format_and_pad_commit(sb, placeholder, context);
1619 else
1620 consumed = format_commit_one(sb, placeholder, context);
1621 if (magic == NO_MAGIC)
1622 return consumed;
1623
1624 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1625 while (sb->len && sb->buf[sb->len - 1] == '\n')
1626 strbuf_setlen(sb, sb->len - 1);
1627 } else if (orig_len != sb->len) {
1628 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1629 strbuf_insertstr(sb, orig_len, "\n");
1630 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1631 strbuf_insertstr(sb, orig_len, " ");
1632 }
1633 return consumed + 1;
1634 }
1635
1636 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1637 void *context)
1638 {
1639 struct userformat_want *w = context;
1640
1641 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1642 placeholder++;
1643
1644 switch (*placeholder) {
1645 case 'N':
1646 w->notes = 1;
1647 break;
1648 case 'S':
1649 w->source = 1;
1650 break;
1651 }
1652 return 0;
1653 }
1654
1655 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1656 {
1657 struct strbuf dummy = STRBUF_INIT;
1658
1659 if (!fmt) {
1660 if (!user_format)
1661 return;
1662 fmt = user_format;
1663 }
1664 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1665 strbuf_release(&dummy);
1666 }
1667
1668 void repo_format_commit_message(struct repository *r,
1669 const struct commit *commit,
1670 const char *format, struct strbuf *sb,
1671 const struct pretty_print_context *pretty_ctx)
1672 {
1673 struct format_commit_context context = {
1674 .repository = r,
1675 .commit = commit,
1676 .pretty_ctx = pretty_ctx,
1677 .wrap_start = sb->len
1678 };
1679 const char *output_enc = pretty_ctx->output_encoding;
1680 const char *utf8 = "UTF-8";
1681
1682 strbuf_expand(sb, format, format_commit_item, &context);
1683 rewrap_message_tail(sb, &context, 0, 0, 0);
1684
1685 /*
1686 * Convert output to an actual output encoding; note that
1687 * format_commit_item() will always use UTF-8, so we don't
1688 * have to bother if that's what the output wants.
1689 */
1690 if (output_enc) {
1691 if (same_encoding(utf8, output_enc))
1692 output_enc = NULL;
1693 } else {
1694 if (context.commit_encoding &&
1695 !same_encoding(context.commit_encoding, utf8))
1696 output_enc = context.commit_encoding;
1697 }
1698
1699 if (output_enc) {
1700 size_t outsz;
1701 char *out = reencode_string_len(sb->buf, sb->len,
1702 output_enc, utf8, &outsz);
1703 if (out)
1704 strbuf_attach(sb, out, outsz, outsz + 1);
1705 }
1706
1707 free(context.commit_encoding);
1708 repo_unuse_commit_buffer(r, commit, context.message);
1709 }
1710
1711 static void pp_header(struct pretty_print_context *pp,
1712 const char *encoding,
1713 const struct commit *commit,
1714 const char **msg_p,
1715 struct strbuf *sb)
1716 {
1717 int parents_shown = 0;
1718
1719 for (;;) {
1720 const char *name, *line = *msg_p;
1721 int linelen = get_one_line(*msg_p);
1722
1723 if (!linelen)
1724 return;
1725 *msg_p += linelen;
1726
1727 if (linelen == 1)
1728 /* End of header */
1729 return;
1730
1731 if (pp->fmt == CMIT_FMT_RAW) {
1732 strbuf_add(sb, line, linelen);
1733 continue;
1734 }
1735
1736 if (starts_with(line, "parent ")) {
1737 if (linelen != the_hash_algo->hexsz + 8)
1738 die("bad parent line in commit");
1739 continue;
1740 }
1741
1742 if (!parents_shown) {
1743 unsigned num = commit_list_count(commit->parents);
1744 /* with enough slop */
1745 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1746 add_merge_info(pp, sb, commit);
1747 parents_shown = 1;
1748 }
1749
1750 /*
1751 * MEDIUM == DEFAULT shows only author with dates.
1752 * FULL shows both authors but not dates.
1753 * FULLER shows both authors and dates.
1754 */
1755 if (skip_prefix(line, "author ", &name)) {
1756 strbuf_grow(sb, linelen + 80);
1757 pp_user_info(pp, "Author", sb, name, encoding);
1758 }
1759 if (skip_prefix(line, "committer ", &name) &&
1760 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1761 strbuf_grow(sb, linelen + 80);
1762 pp_user_info(pp, "Commit", sb, name, encoding);
1763 }
1764 }
1765 }
1766
1767 void pp_title_line(struct pretty_print_context *pp,
1768 const char **msg_p,
1769 struct strbuf *sb,
1770 const char *encoding,
1771 int need_8bit_cte)
1772 {
1773 static const int max_length = 78; /* per rfc2047 */
1774 struct strbuf title;
1775
1776 strbuf_init(&title, 80);
1777 *msg_p = format_subject(&title, *msg_p,
1778 pp->preserve_subject ? "\n" : " ");
1779
1780 strbuf_grow(sb, title.len + 1024);
1781 if (pp->print_email_subject) {
1782 if (pp->rev)
1783 fmt_output_email_subject(sb, pp->rev);
1784 if (pp->encode_email_headers &&
1785 needs_rfc2047_encoding(title.buf, title.len))
1786 add_rfc2047(sb, title.buf, title.len,
1787 encoding, RFC2047_SUBJECT);
1788 else
1789 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1790 -last_line_length(sb), 1, max_length);
1791 } else {
1792 strbuf_addbuf(sb, &title);
1793 }
1794 strbuf_addch(sb, '\n');
1795
1796 if (need_8bit_cte == 0) {
1797 int i;
1798 for (i = 0; i < pp->in_body_headers.nr; i++) {
1799 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1800 need_8bit_cte = 1;
1801 break;
1802 }
1803 }
1804 }
1805
1806 if (need_8bit_cte > 0) {
1807 const char *header_fmt =
1808 "MIME-Version: 1.0\n"
1809 "Content-Type: text/plain; charset=%s\n"
1810 "Content-Transfer-Encoding: 8bit\n";
1811 strbuf_addf(sb, header_fmt, encoding);
1812 }
1813 if (pp->after_subject) {
1814 strbuf_addstr(sb, pp->after_subject);
1815 }
1816 if (cmit_fmt_is_mail(pp->fmt)) {
1817 strbuf_addch(sb, '\n');
1818 }
1819
1820 if (pp->in_body_headers.nr) {
1821 int i;
1822 for (i = 0; i < pp->in_body_headers.nr; i++) {
1823 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1824 free(pp->in_body_headers.items[i].string);
1825 }
1826 string_list_clear(&pp->in_body_headers, 0);
1827 strbuf_addch(sb, '\n');
1828 }
1829
1830 strbuf_release(&title);
1831 }
1832
1833 static int pp_utf8_width(const char *start, const char *end)
1834 {
1835 int width = 0;
1836 size_t remain = end - start;
1837
1838 while (remain) {
1839 int n = utf8_width(&start, &remain);
1840 if (n < 0 || !start)
1841 return -1;
1842 width += n;
1843 }
1844 return width;
1845 }
1846
1847 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1848 const char *line, int linelen)
1849 {
1850 const char *tab;
1851
1852 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1853 int width = pp_utf8_width(line, tab);
1854
1855 /*
1856 * If it wasn't well-formed utf8, or it
1857 * had characters with badly defined
1858 * width (control characters etc), just
1859 * give up on trying to align things.
1860 */
1861 if (width < 0)
1862 break;
1863
1864 /* Output the data .. */
1865 strbuf_add(sb, line, tab - line);
1866
1867 /* .. and the de-tabified tab */
1868 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1869
1870 /* Skip over the printed part .. */
1871 linelen -= tab + 1 - line;
1872 line = tab + 1;
1873 }
1874
1875 /*
1876 * Print out everything after the last tab without
1877 * worrying about width - there's nothing more to
1878 * align.
1879 */
1880 strbuf_add(sb, line, linelen);
1881 }
1882
1883 /*
1884 * pp_handle_indent() prints out the intendation, and
1885 * the whole line (without the final newline), after
1886 * de-tabifying.
1887 */
1888 static void pp_handle_indent(struct pretty_print_context *pp,
1889 struct strbuf *sb, int indent,
1890 const char *line, int linelen)
1891 {
1892 strbuf_addchars(sb, ' ', indent);
1893 if (pp->expand_tabs_in_log)
1894 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1895 else
1896 strbuf_add(sb, line, linelen);
1897 }
1898
1899 static int is_mboxrd_from(const char *line, int len)
1900 {
1901 /*
1902 * a line matching /^From $/ here would only have len == 4
1903 * at this point because is_empty_line would've trimmed all
1904 * trailing space
1905 */
1906 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1907 }
1908
1909 void pp_remainder(struct pretty_print_context *pp,
1910 const char **msg_p,
1911 struct strbuf *sb,
1912 int indent)
1913 {
1914 int first = 1;
1915 for (;;) {
1916 const char *line = *msg_p;
1917 int linelen = get_one_line(line);
1918 *msg_p += linelen;
1919
1920 if (!linelen)
1921 break;
1922
1923 if (is_blank_line(line, &linelen)) {
1924 if (first)
1925 continue;
1926 if (pp->fmt == CMIT_FMT_SHORT)
1927 break;
1928 }
1929 first = 0;
1930
1931 strbuf_grow(sb, linelen + indent + 20);
1932 if (indent)
1933 pp_handle_indent(pp, sb, indent, line, linelen);
1934 else if (pp->expand_tabs_in_log)
1935 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1936 line, linelen);
1937 else {
1938 if (pp->fmt == CMIT_FMT_MBOXRD &&
1939 is_mboxrd_from(line, linelen))
1940 strbuf_addch(sb, '>');
1941
1942 strbuf_add(sb, line, linelen);
1943 }
1944 strbuf_addch(sb, '\n');
1945 }
1946 }
1947
1948 void pretty_print_commit(struct pretty_print_context *pp,
1949 const struct commit *commit,
1950 struct strbuf *sb)
1951 {
1952 unsigned long beginning_of_body;
1953 int indent = 4;
1954 const char *msg;
1955 const char *reencoded;
1956 const char *encoding;
1957 int need_8bit_cte = pp->need_8bit_cte;
1958
1959 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1960 format_commit_message(commit, user_format, sb, pp);
1961 return;
1962 }
1963
1964 encoding = get_log_output_encoding();
1965 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1966
1967 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1968 indent = 0;
1969
1970 /*
1971 * We need to check and emit Content-type: to mark it
1972 * as 8-bit if we haven't done so.
1973 */
1974 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1975 int i, ch, in_body;
1976
1977 for (in_body = i = 0; (ch = msg[i]); i++) {
1978 if (!in_body) {
1979 /* author could be non 7-bit ASCII but
1980 * the log may be so; skip over the
1981 * header part first.
1982 */
1983 if (ch == '\n' && msg[i+1] == '\n')
1984 in_body = 1;
1985 }
1986 else if (non_ascii(ch)) {
1987 need_8bit_cte = 1;
1988 break;
1989 }
1990 }
1991 }
1992
1993 pp_header(pp, encoding, commit, &msg, sb);
1994 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
1995 strbuf_addch(sb, '\n');
1996 }
1997
1998 /* Skip excess blank lines at the beginning of body, if any... */
1999 msg = skip_blank_lines(msg);
2000
2001 /* These formats treat the title line specially. */
2002 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2003 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2004
2005 beginning_of_body = sb->len;
2006 if (pp->fmt != CMIT_FMT_ONELINE)
2007 pp_remainder(pp, &msg, sb, indent);
2008 strbuf_rtrim(sb);
2009
2010 /* Make sure there is an EOLN for the non-oneline case */
2011 if (pp->fmt != CMIT_FMT_ONELINE)
2012 strbuf_addch(sb, '\n');
2013
2014 /*
2015 * The caller may append additional body text in e-mail
2016 * format. Make sure we did not strip the blank line
2017 * between the header and the body.
2018 */
2019 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2020 strbuf_addch(sb, '\n');
2021
2022 unuse_commit_buffer(commit, reencoded);
2023 }
2024
2025 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2026 struct strbuf *sb)
2027 {
2028 struct pretty_print_context pp = {0};
2029 pp.fmt = fmt;
2030 pretty_print_commit(&pp, commit, sb);
2031 }
git for Debian