?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
reachable: use revision machinery's --indexed-objects code
[git.git] / pretty.c
blob5fd9de6ce11da25b8ddf8ece9f718023394624e6
1 #include "cache.h"
2 #include "commit.h"
3 #include "utf8.h"
4 #include "diff.h"
5 #include "revision.h"
6 #include "string-list.h"
7 #include "mailmap.h"
8 #include "log-tree.h"
9 #include "notes.h"
10 #include "color.h"
11 #include "reflog-walk.h"
12 #include "gpg-interface.h"
13
14 static char *user_format;
15 static struct cmt_fmt_map {
16 const char *name;
17 enum cmit_fmt format;
18 int is_tformat;
19 int is_alias;
20 const char *user_format;
21 } *commit_formats;
22 static size_t builtin_formats_len;
23 static size_t commit_formats_len;
24 static size_t commit_formats_alloc;
25 static struct cmt_fmt_map *find_commit_format(const char *sought);
26
27 int commit_format_is_empty(enum cmit_fmt fmt)
28 {
29 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
30 }
31
32 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
33 {
34 free(user_format);
35 user_format = xstrdup(cp);
36 if (is_tformat)
37 rev->use_terminator = 1;
38 rev->commit_format = CMIT_FMT_USERFORMAT;
39 }
40
41 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
42 {
43 struct cmt_fmt_map *commit_format = NULL;
44 const char *name;
45 const char *fmt;
46 int i;
47
48 if (!skip_prefix(var, "pretty.", &name))
49 return 0;
50
51 for (i = 0; i < builtin_formats_len; i++) {
52 if (!strcmp(commit_formats[i].name, name))
53 return 0;
54 }
55
56 for (i = builtin_formats_len; i < commit_formats_len; i++) {
57 if (!strcmp(commit_formats[i].name, name)) {
58 commit_format = &commit_formats[i];
59 break;
60 }
61 }
62
63 if (!commit_format) {
64 ALLOC_GROW(commit_formats, commit_formats_len+1,
65 commit_formats_alloc);
66 commit_format = &commit_formats[commit_formats_len];
67 memset(commit_format, 0, sizeof(*commit_format));
68 commit_formats_len++;
69 }
70
71 commit_format->name = xstrdup(name);
72 commit_format->format = CMIT_FMT_USERFORMAT;
73 if (git_config_string(&fmt, var, value))
74 return -1;
75
76 if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
77 commit_format->is_tformat = fmt[0] == 't';
78 fmt = strchr(fmt, ':') + 1;
79 } else if (strchr(fmt, '%'))
80 commit_format->is_tformat = 1;
81 else
82 commit_format->is_alias = 1;
83 commit_format->user_format = fmt;
84
85 return 0;
86 }
87
88 static void setup_commit_formats(void)
89 {
90 struct cmt_fmt_map builtin_formats[] = {
91 { "raw", CMIT_FMT_RAW, 0 },
92 { "medium", CMIT_FMT_MEDIUM, 0 },
93 { "short", CMIT_FMT_SHORT, 0 },
94 { "email", CMIT_FMT_EMAIL, 0 },
95 { "fuller", CMIT_FMT_FULLER, 0 },
96 { "full", CMIT_FMT_FULL, 0 },
97 { "oneline", CMIT_FMT_ONELINE, 1 }
98 };
99 commit_formats_len = ARRAY_SIZE(builtin_formats);
100 builtin_formats_len = commit_formats_len;
101 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
102 memcpy(commit_formats, builtin_formats,
103 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
104
105 git_config(git_pretty_formats_config, NULL);
106 }
107
108 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
109 const char *original,
110 int num_redirections)
111 {
112 struct cmt_fmt_map *found = NULL;
113 size_t found_match_len = 0;
114 int i;
115
116 if (num_redirections >= commit_formats_len)
117 die("invalid --pretty format: "
118 "'%s' references an alias which points to itself",
119 original);
120
121 for (i = 0; i < commit_formats_len; i++) {
122 size_t match_len;
123
124 if (!starts_with(commit_formats[i].name, sought))
125 continue;
126
127 match_len = strlen(commit_formats[i].name);
128 if (found == NULL || found_match_len > match_len) {
129 found = &commit_formats[i];
130 found_match_len = match_len;
131 }
132 }
133
134 if (found && found->is_alias) {
135 found = find_commit_format_recursive(found->user_format,
136 original,
137 num_redirections+1);
138 }
139
140 return found;
141 }
142
143 static struct cmt_fmt_map *find_commit_format(const char *sought)
144 {
145 if (!commit_formats)
146 setup_commit_formats();
147
148 return find_commit_format_recursive(sought, sought, 0);
149 }
150
151 void get_commit_format(const char *arg, struct rev_info *rev)
152 {
153 struct cmt_fmt_map *commit_format;
154
155 rev->use_terminator = 0;
156 if (!arg) {
157 rev->commit_format = CMIT_FMT_DEFAULT;
158 return;
159 }
160 if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
161 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
162 return;
163 }
164
165 if (!*arg || strchr(arg, '%')) {
166 save_user_format(rev, arg, 1);
167 return;
168 }
169
170 commit_format = find_commit_format(arg);
171 if (!commit_format)
172 die("invalid --pretty format: %s", arg);
173
174 rev->commit_format = commit_format->format;
175 rev->use_terminator = commit_format->is_tformat;
176 if (commit_format->format == CMIT_FMT_USERFORMAT) {
177 save_user_format(rev, commit_format->user_format,
178 commit_format->is_tformat);
179 }
180 }
181
182 /*
183 * Generic support for pretty-printing the header
184 */
185 static int get_one_line(const char *msg)
186 {
187 int ret = 0;
188
189 for (;;) {
190 char c = *msg++;
191 if (!c)
192 break;
193 ret++;
194 if (c == '\n')
195 break;
196 }
197 return ret;
198 }
199
200 /* High bit set, or ISO-2022-INT */
201 static int non_ascii(int ch)
202 {
203 return !isascii(ch) || ch == '\033';
204 }
205
206 int has_non_ascii(const char *s)
207 {
208 int ch;
209 if (!s)
210 return 0;
211 while ((ch = *s++) != '\0') {
212 if (non_ascii(ch))
213 return 1;
214 }
215 return 0;
216 }
217
218 static int is_rfc822_special(char ch)
219 {
220 switch (ch) {
221 case '(':
222 case ')':
223 case '<':
224 case '>':
225 case '[':
226 case ']':
227 case ':':
228 case ';':
229 case '@':
230 case ',':
231 case '.':
232 case '"':
233 case '\\':
234 return 1;
235 default:
236 return 0;
237 }
238 }
239
240 static int needs_rfc822_quoting(const char *s, int len)
241 {
242 int i;
243 for (i = 0; i < len; i++)
244 if (is_rfc822_special(s[i]))
245 return 1;
246 return 0;
247 }
248
249 static int last_line_length(struct strbuf *sb)
250 {
251 int i;
252
253 /* How many bytes are already used on the last line? */
254 for (i = sb->len - 1; i >= 0; i--)
255 if (sb->buf[i] == '\n')
256 break;
257 return sb->len - (i + 1);
258 }
259
260 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
261 {
262 int i;
263
264 /* just a guess, we may have to also backslash-quote */
265 strbuf_grow(out, len + 2);
266
267 strbuf_addch(out, '"');
268 for (i = 0; i < len; i++) {
269 switch (s[i]) {
270 case '"':
271 case '\\':
272 strbuf_addch(out, '\\');
273 /* fall through */
274 default:
275 strbuf_addch(out, s[i]);
276 }
277 }
278 strbuf_addch(out, '"');
279 }
280
281 enum rfc2047_type {
282 RFC2047_SUBJECT,
283 RFC2047_ADDRESS
284 };
285
286 static int is_rfc2047_special(char ch, enum rfc2047_type type)
287 {
288 /*
289 * rfc2047, section 4.2:
290 *
291 * 8-bit values which correspond to printable ASCII characters other
292 * than "=", "?", and "_" (underscore), MAY be represented as those
293 * characters. (But see section 5 for restrictions.) In
294 * particular, SPACE and TAB MUST NOT be represented as themselves
295 * within encoded words.
296 */
297
298 /*
299 * rule out non-ASCII characters and non-printable characters (the
300 * non-ASCII check should be redundant as isprint() is not localized
301 * and only knows about ASCII, but be defensive about that)
302 */
303 if (non_ascii(ch) || !isprint(ch))
304 return 1;
305
306 /*
307 * rule out special printable characters (' ' should be the only
308 * whitespace character considered printable, but be defensive and use
309 * isspace())
310 */
311 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
312 return 1;
313
314 /*
315 * rfc2047, section 5.3:
316 *
317 * As a replacement for a 'word' entity within a 'phrase', for example,
318 * one that precedes an address in a From, To, or Cc header. The ABNF
319 * definition for 'phrase' from RFC 822 thus becomes:
320 *
321 * phrase = 1*( encoded-word / word )
322 *
323 * In this case the set of characters that may be used in a "Q"-encoded
324 * 'encoded-word' is restricted to: <upper and lower case ASCII
325 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
326 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
327 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
328 * 'special' by 'linear-white-space'.
329 */
330
331 if (type != RFC2047_ADDRESS)
332 return 0;
333
334 /* '=' and '_' are special cases and have been checked above */
335 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
336 }
337
338 static int needs_rfc2047_encoding(const char *line, int len,
339 enum rfc2047_type type)
340 {
341 int i;
342
343 for (i = 0; i < len; i++) {
344 int ch = line[i];
345 if (non_ascii(ch) || ch == '\n')
346 return 1;
347 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
348 return 1;
349 }
350
351 return 0;
352 }
353
354 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
355 const char *encoding, enum rfc2047_type type)
356 {
357 static const int max_encoded_length = 76; /* per rfc2047 */
358 int i;
359 int line_len = last_line_length(sb);
360
361 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
362 strbuf_addf(sb, "=?%s?q?", encoding);
363 line_len += strlen(encoding) + 5; /* 5 for =??q? */
364
365 while (len) {
366 /*
367 * RFC 2047, section 5 (3):
368 *
369 * Each 'encoded-word' MUST represent an integral number of
370 * characters. A multi-octet character may not be split across
371 * adjacent 'encoded- word's.
372 */
373 const unsigned char *p = (const unsigned char *)line;
374 int chrlen = mbs_chrlen(&line, &len, encoding);
375 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
376
377 /* "=%02X" * chrlen, or the byte itself */
378 const char *encoded_fmt = is_special ? "=%02X" : "%c";
379 int encoded_len = is_special ? 3 * chrlen : 1;
380
381 /*
382 * According to RFC 2047, we could encode the special character
383 * ' ' (space) with '_' (underscore) for readability. But many
384 * programs do not understand this and just leave the
385 * underscore in place. Thus, we do nothing special here, which
386 * causes ' ' to be encoded as '=20', avoiding this problem.
387 */
388
389 if (line_len + encoded_len + 2 > max_encoded_length) {
390 /* It won't fit with trailing "?=" --- break the line */
391 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
392 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
393 }
394
395 for (i = 0; i < chrlen; i++)
396 strbuf_addf(sb, encoded_fmt, p[i]);
397 line_len += encoded_len;
398 }
399 strbuf_addstr(sb, "?=");
400 }
401
402 const char *show_ident_date(const struct ident_split *ident,
403 enum date_mode mode)
404 {
405 unsigned long date = 0;
406 long tz = 0;
407
408 if (ident->date_begin && ident->date_end)
409 date = strtoul(ident->date_begin, NULL, 10);
410 if (date_overflows(date))
411 date = 0;
412 else {
413 if (ident->tz_begin && ident->tz_end)
414 tz = strtol(ident->tz_begin, NULL, 10);
415 if (tz >= INT_MAX || tz <= INT_MIN)
416 tz = 0;
417 }
418 return show_date(date, tz, mode);
419 }
420
421 void pp_user_info(struct pretty_print_context *pp,
422 const char *what, struct strbuf *sb,
423 const char *line, const char *encoding)
424 {
425 struct ident_split ident;
426 char *line_end;
427 const char *mailbuf, *namebuf;
428 size_t namelen, maillen;
429 int max_length = 78; /* per rfc2822 */
430
431 if (pp->fmt == CMIT_FMT_ONELINE)
432 return;
433
434 line_end = strchrnul(line, '\n');
435 if (split_ident_line(&ident, line, line_end - line))
436 return;
437
438 mailbuf = ident.mail_begin;
439 maillen = ident.mail_end - ident.mail_begin;
440 namebuf = ident.name_begin;
441 namelen = ident.name_end - ident.name_begin;
442
443 if (pp->mailmap)
444 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
445
446 if (pp->fmt == CMIT_FMT_EMAIL) {
447 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
448 struct strbuf buf = STRBUF_INIT;
449
450 strbuf_addstr(&buf, "From: ");
451 strbuf_add(&buf, namebuf, namelen);
452 strbuf_addstr(&buf, " <");
453 strbuf_add(&buf, mailbuf, maillen);
454 strbuf_addstr(&buf, ">\n");
455 string_list_append(&pp->in_body_headers,
456 strbuf_detach(&buf, NULL));
457
458 mailbuf = pp->from_ident->mail_begin;
459 maillen = pp->from_ident->mail_end - mailbuf;
460 namebuf = pp->from_ident->name_begin;
461 namelen = pp->from_ident->name_end - namebuf;
462 }
463
464 strbuf_addstr(sb, "From: ");
465 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
466 add_rfc2047(sb, namebuf, namelen,
467 encoding, RFC2047_ADDRESS);
468 max_length = 76; /* per rfc2047 */
469 } else if (needs_rfc822_quoting(namebuf, namelen)) {
470 struct strbuf quoted = STRBUF_INIT;
471 add_rfc822_quoted(&quoted, namebuf, namelen);
472 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
473 -6, 1, max_length);
474 strbuf_release(&quoted);
475 } else {
476 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
477 -6, 1, max_length);
478 }
479
480 if (max_length <
481 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
482 strbuf_addch(sb, '\n');
483 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
484 } else {
485 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
486 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
487 (int)namelen, namebuf, (int)maillen, mailbuf);
488 }
489
490 switch (pp->fmt) {
491 case CMIT_FMT_MEDIUM:
492 strbuf_addf(sb, "Date: %s\n",
493 show_ident_date(&ident, pp->date_mode));
494 break;
495 case CMIT_FMT_EMAIL:
496 strbuf_addf(sb, "Date: %s\n",
497 show_ident_date(&ident, DATE_RFC2822));
498 break;
499 case CMIT_FMT_FULLER:
500 strbuf_addf(sb, "%sDate: %s\n", what,
501 show_ident_date(&ident, pp->date_mode));
502 break;
503 default:
504 /* notin' */
505 break;
506 }
507 }
508
509 static int is_empty_line(const char *line, int *len_p)
510 {
511 int len = *len_p;
512 while (len && isspace(line[len - 1]))
513 len--;
514 *len_p = len;
515 return !len;
516 }
517
518 static const char *skip_empty_lines(const char *msg)
519 {
520 for (;;) {
521 int linelen = get_one_line(msg);
522 int ll = linelen;
523 if (!linelen)
524 break;
525 if (!is_empty_line(msg, &ll))
526 break;
527 msg += linelen;
528 }
529 return msg;
530 }
531
532 static void add_merge_info(const struct pretty_print_context *pp,
533 struct strbuf *sb, const struct commit *commit)
534 {
535 struct commit_list *parent = commit->parents;
536
537 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
538 !parent || !parent->next)
539 return;
540
541 strbuf_addstr(sb, "Merge:");
542
543 while (parent) {
544 struct commit *p = parent->item;
545 const char *hex = NULL;
546 if (pp->abbrev)
547 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
548 if (!hex)
549 hex = sha1_to_hex(p->object.sha1);
550 parent = parent->next;
551
552 strbuf_addf(sb, " %s", hex);
553 }
554 strbuf_addch(sb, '\n');
555 }
556
557 static char *get_header(const char *msg, const char *key)
558 {
559 size_t len;
560 const char *v = find_commit_header(msg, key, &len);
561 return v ? xmemdupz(v, len) : NULL;
562 }
563
564 static char *replace_encoding_header(char *buf, const char *encoding)
565 {
566 struct strbuf tmp = STRBUF_INIT;
567 size_t start, len;
568 char *cp = buf;
569
570 /* guess if there is an encoding header before a \n\n */
571 while (strncmp(cp, "encoding ", strlen("encoding "))) {
572 cp = strchr(cp, '\n');
573 if (!cp || *++cp == '\n')
574 return buf;
575 }
576 start = cp - buf;
577 cp = strchr(cp, '\n');
578 if (!cp)
579 return buf; /* should not happen but be defensive */
580 len = cp + 1 - (buf + start);
581
582 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
583 if (is_encoding_utf8(encoding)) {
584 /* we have re-coded to UTF-8; drop the header */
585 strbuf_remove(&tmp, start, len);
586 } else {
587 /* just replaces XXXX in 'encoding XXXX\n' */
588 strbuf_splice(&tmp, start + strlen("encoding "),
589 len - strlen("encoding \n"),
590 encoding, strlen(encoding));
591 }
592 return strbuf_detach(&tmp, NULL);
593 }
594
595 const char *logmsg_reencode(const struct commit *commit,
596 char **commit_encoding,
597 const char *output_encoding)
598 {
599 static const char *utf8 = "UTF-8";
600 const char *use_encoding;
601 char *encoding;
602 const char *msg = get_commit_buffer(commit, NULL);
603 char *out;
604
605 if (!output_encoding || !*output_encoding) {
606 if (commit_encoding)
607 *commit_encoding = get_header(msg, "encoding");
608 return msg;
609 }
610 encoding = get_header(msg, "encoding");
611 if (commit_encoding)
612 *commit_encoding = encoding;
613 use_encoding = encoding ? encoding : utf8;
614 if (same_encoding(use_encoding, output_encoding)) {
615 /*
616 * No encoding work to be done. If we have no encoding header
617 * at all, then there's nothing to do, and we can return the
618 * message verbatim (whether newly allocated or not).
619 */
620 if (!encoding)
621 return msg;
622
623 /*
624 * Otherwise, we still want to munge the encoding header in the
625 * result, which will be done by modifying the buffer. If we
626 * are using a fresh copy, we can reuse it. But if we are using
627 * the cached copy from get_commit_buffer, we need to duplicate it
628 * to avoid munging the cached copy.
629 */
630 if (msg == get_cached_commit_buffer(commit, NULL))
631 out = xstrdup(msg);
632 else
633 out = (char *)msg;
634 }
635 else {
636 /*
637 * There's actual encoding work to do. Do the reencoding, which
638 * still leaves the header to be replaced in the next step. At
639 * this point, we are done with msg. If we allocated a fresh
640 * copy, we can free it.
641 */
642 out = reencode_string(msg, output_encoding, use_encoding);
643 if (out)
644 unuse_commit_buffer(commit, msg);
645 }
646
647 /*
648 * This replacement actually consumes the buffer we hand it, so we do
649 * not have to worry about freeing the old "out" here.
650 */
651 if (out)
652 out = replace_encoding_header(out, output_encoding);
653
654 if (!commit_encoding)
655 free(encoding);
656 /*
657 * If the re-encoding failed, out might be NULL here; in that
658 * case we just return the commit message verbatim.
659 */
660 return out ? out : msg;
661 }
662
663 static int mailmap_name(const char **email, size_t *email_len,
664 const char **name, size_t *name_len)
665 {
666 static struct string_list *mail_map;
667 if (!mail_map) {
668 mail_map = xcalloc(1, sizeof(*mail_map));
669 read_mailmap(mail_map, NULL);
670 }
671 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
672 }
673
674 static size_t format_person_part(struct strbuf *sb, char part,
675 const char *msg, int len, enum date_mode dmode)
676 {
677 /* currently all placeholders have same length */
678 const int placeholder_len = 2;
679 struct ident_split s;
680 const char *name, *mail;
681 size_t maillen, namelen;
682
683 if (split_ident_line(&s, msg, len) < 0)
684 goto skip;
685
686 name = s.name_begin;
687 namelen = s.name_end - s.name_begin;
688 mail = s.mail_begin;
689 maillen = s.mail_end - s.mail_begin;
690
691 if (part == 'N' || part == 'E') /* mailmap lookup */
692 mailmap_name(&mail, &maillen, &name, &namelen);
693 if (part == 'n' || part == 'N') { /* name */
694 strbuf_add(sb, name, namelen);
695 return placeholder_len;
696 }
697 if (part == 'e' || part == 'E') { /* email */
698 strbuf_add(sb, mail, maillen);
699 return placeholder_len;
700 }
701
702 if (!s.date_begin)
703 goto skip;
704
705 if (part == 't') { /* date, UNIX timestamp */
706 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
707 return placeholder_len;
708 }
709
710 switch (part) {
711 case 'd': /* date */
712 strbuf_addstr(sb, show_ident_date(&s, dmode));
713 return placeholder_len;
714 case 'D': /* date, RFC2822 style */
715 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
716 return placeholder_len;
717 case 'r': /* date, relative */
718 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
719 return placeholder_len;
720 case 'i': /* date, ISO 8601-like */
721 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
722 return placeholder_len;
723 case 'I': /* date, ISO 8601 strict */
724 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601_STRICT));
725 return placeholder_len;
726 }
727
728 skip:
729 /*
730 * reading from either a bogus commit, or a reflog entry with
731 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
732 * to compute a valid return value.
733 */
734 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
735 || part == 'D' || part == 'r' || part == 'i')
736 return placeholder_len;
737
738 return 0; /* unknown placeholder */
739 }
740
741 struct chunk {
742 size_t off;
743 size_t len;
744 };
745
746 enum flush_type {
747 no_flush,
748 flush_right,
749 flush_left,
750 flush_left_and_steal,
751 flush_both
752 };
753
754 enum trunc_type {
755 trunc_none,
756 trunc_left,
757 trunc_middle,
758 trunc_right
759 };
760
761 struct format_commit_context {
762 const struct commit *commit;
763 const struct pretty_print_context *pretty_ctx;
764 unsigned commit_header_parsed:1;
765 unsigned commit_message_parsed:1;
766 struct signature_check signature_check;
767 enum flush_type flush_type;
768 enum trunc_type truncate;
769 const char *message;
770 char *commit_encoding;
771 size_t width, indent1, indent2;
772 int auto_color;
773 int padding;
774
775 /* These offsets are relative to the start of the commit message. */
776 struct chunk author;
777 struct chunk committer;
778 size_t message_off;
779 size_t subject_off;
780 size_t body_off;
781
782 /* The following ones are relative to the result struct strbuf. */
783 struct chunk abbrev_commit_hash;
784 struct chunk abbrev_tree_hash;
785 struct chunk abbrev_parent_hashes;
786 size_t wrap_start;
787 };
788
789 static int add_again(struct strbuf *sb, struct chunk *chunk)
790 {
791 if (chunk->len) {
792 strbuf_adddup(sb, chunk->off, chunk->len);
793 return 1;
794 }
795
796 /*
797 * We haven't seen this chunk before. Our caller is surely
798 * going to add it the hard way now. Remember the most likely
799 * start of the to-be-added chunk: the current end of the
800 * struct strbuf.
801 */
802 chunk->off = sb->len;
803 return 0;
804 }
805
806 static void parse_commit_header(struct format_commit_context *context)
807 {
808 const char *msg = context->message;
809 int i;
810
811 for (i = 0; msg[i]; i++) {
812 int eol;
813 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
814 ; /* do nothing */
815
816 if (i == eol) {
817 break;
818 } else if (starts_with(msg + i, "author ")) {
819 context->author.off = i + 7;
820 context->author.len = eol - i - 7;
821 } else if (starts_with(msg + i, "committer ")) {
822 context->committer.off = i + 10;
823 context->committer.len = eol - i - 10;
824 }
825 i = eol;
826 }
827 context->message_off = i;
828 context->commit_header_parsed = 1;
829 }
830
831 static int istitlechar(char c)
832 {
833 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
834 (c >= '0' && c <= '9') || c == '.' || c == '_';
835 }
836
837 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
838 {
839 size_t trimlen;
840 size_t start_len = sb->len;
841 int space = 2;
842
843 for (; *msg && *msg != '\n'; msg++) {
844 if (istitlechar(*msg)) {
845 if (space == 1)
846 strbuf_addch(sb, '-');
847 space = 0;
848 strbuf_addch(sb, *msg);
849 if (*msg == '.')
850 while (*(msg+1) == '.')
851 msg++;
852 } else
853 space |= 1;
854 }
855
856 /* trim any trailing '.' or '-' characters */
857 trimlen = 0;
858 while (sb->len - trimlen > start_len &&
859 (sb->buf[sb->len - 1 - trimlen] == '.'
860 || sb->buf[sb->len - 1 - trimlen] == '-'))
861 trimlen++;
862 strbuf_remove(sb, sb->len - trimlen, trimlen);
863 }
864
865 const char *format_subject(struct strbuf *sb, const char *msg,
866 const char *line_separator)
867 {
868 int first = 1;
869
870 for (;;) {
871 const char *line = msg;
872 int linelen = get_one_line(line);
873
874 msg += linelen;
875 if (!linelen || is_empty_line(line, &linelen))
876 break;
877
878 if (!sb)
879 continue;
880 strbuf_grow(sb, linelen + 2);
881 if (!first)
882 strbuf_addstr(sb, line_separator);
883 strbuf_add(sb, line, linelen);
884 first = 0;
885 }
886 return msg;
887 }
888
889 static void parse_commit_message(struct format_commit_context *c)
890 {
891 const char *msg = c->message + c->message_off;
892 const char *start = c->message;
893
894 msg = skip_empty_lines(msg);
895 c->subject_off = msg - start;
896
897 msg = format_subject(NULL, msg, NULL);
898 msg = skip_empty_lines(msg);
899 c->body_off = msg - start;
900
901 c->commit_message_parsed = 1;
902 }
903
904 static void strbuf_wrap(struct strbuf *sb, size_t pos,
905 size_t width, size_t indent1, size_t indent2)
906 {
907 struct strbuf tmp = STRBUF_INIT;
908
909 if (pos)
910 strbuf_add(&tmp, sb->buf, pos);
911 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
912 (int) indent1, (int) indent2, (int) width);
913 strbuf_swap(&tmp, sb);
914 strbuf_release(&tmp);
915 }
916
917 static void rewrap_message_tail(struct strbuf *sb,
918 struct format_commit_context *c,
919 size_t new_width, size_t new_indent1,
920 size_t new_indent2)
921 {
922 if (c->width == new_width && c->indent1 == new_indent1 &&
923 c->indent2 == new_indent2)
924 return;
925 if (c->wrap_start < sb->len)
926 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
927 c->wrap_start = sb->len;
928 c->width = new_width;
929 c->indent1 = new_indent1;
930 c->indent2 = new_indent2;
931 }
932
933 static int format_reflog_person(struct strbuf *sb,
934 char part,
935 struct reflog_walk_info *log,
936 enum date_mode dmode)
937 {
938 const char *ident;
939
940 if (!log)
941 return 2;
942
943 ident = get_reflog_ident(log);
944 if (!ident)
945 return 2;
946
947 return format_person_part(sb, part, ident, strlen(ident), dmode);
948 }
949
950 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
951 const char *placeholder,
952 struct format_commit_context *c)
953 {
954 if (placeholder[1] == '(') {
955 const char *begin = placeholder + 2;
956 const char *end = strchr(begin, ')');
957 char color[COLOR_MAXLEN];
958
959 if (!end)
960 return 0;
961 if (starts_with(begin, "auto,")) {
962 if (!want_color(c->pretty_ctx->color))
963 return end - placeholder + 1;
964 begin += 5;
965 }
966 color_parse_mem(begin,
967 end - begin,
968 "--pretty format", color);
969 strbuf_addstr(sb, color);
970 return end - placeholder + 1;
971 }
972 if (starts_with(placeholder + 1, "red")) {
973 strbuf_addstr(sb, GIT_COLOR_RED);
974 return 4;
975 } else if (starts_with(placeholder + 1, "green")) {
976 strbuf_addstr(sb, GIT_COLOR_GREEN);
977 return 6;
978 } else if (starts_with(placeholder + 1, "blue")) {
979 strbuf_addstr(sb, GIT_COLOR_BLUE);
980 return 5;
981 } else if (starts_with(placeholder + 1, "reset")) {
982 strbuf_addstr(sb, GIT_COLOR_RESET);
983 return 6;
984 } else
985 return 0;
986 }
987
988 static size_t parse_padding_placeholder(struct strbuf *sb,
989 const char *placeholder,
990 struct format_commit_context *c)
991 {
992 const char *ch = placeholder;
993 enum flush_type flush_type;
994 int to_column = 0;
995
996 switch (*ch++) {
997 case '<':
998 flush_type = flush_right;
999 break;
1000 case '>':
1001 if (*ch == '<') {
1002 flush_type = flush_both;
1003 ch++;
1004 } else if (*ch == '>') {
1005 flush_type = flush_left_and_steal;
1006 ch++;
1007 } else
1008 flush_type = flush_left;
1009 break;
1010 default:
1011 return 0;
1012 }
1013
1014 /* the next value means "wide enough to that column" */
1015 if (*ch == '|') {
1016 to_column = 1;
1017 ch++;
1018 }
1019
1020 if (*ch == '(') {
1021 const char *start = ch + 1;
1022 const char *end = start + strcspn(start, ",)");
1023 char *next;
1024 int width;
1025 if (!end || end == start)
1026 return 0;
1027 width = strtoul(start, &next, 10);
1028 if (next == start || width == 0)
1029 return 0;
1030 c->padding = to_column ? -width : width;
1031 c->flush_type = flush_type;
1032
1033 if (*end == ',') {
1034 start = end + 1;
1035 end = strchr(start, ')');
1036 if (!end || end == start)
1037 return 0;
1038 if (starts_with(start, "trunc)"))
1039 c->truncate = trunc_right;
1040 else if (starts_with(start, "ltrunc)"))
1041 c->truncate = trunc_left;
1042 else if (starts_with(start, "mtrunc)"))
1043 c->truncate = trunc_middle;
1044 else
1045 return 0;
1046 } else
1047 c->truncate = trunc_none;
1048
1049 return end - placeholder + 1;
1050 }
1051 return 0;
1052 }
1053
1054 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1055 const char *placeholder,
1056 void *context)
1057 {
1058 struct format_commit_context *c = context;
1059 const struct commit *commit = c->commit;
1060 const char *msg = c->message;
1061 struct commit_list *p;
1062 int h1, h2;
1063
1064 /* these are independent of the commit */
1065 switch (placeholder[0]) {
1066 case 'C':
1067 if (starts_with(placeholder + 1, "(auto)")) {
1068 c->auto_color = 1;
1069 return 7; /* consumed 7 bytes, "C(auto)" */
1070 } else {
1071 int ret = parse_color(sb, placeholder, c);
1072 if (ret)
1073 c->auto_color = 0;
1074 /*
1075 * Otherwise, we decided to treat %C<unknown>
1076 * as a literal string, and the previous
1077 * %C(auto) is still valid.
1078 */
1079 return ret;
1080 }
1081 case 'n': /* newline */
1082 strbuf_addch(sb, '\n');
1083 return 1;
1084 case 'x':
1085 /* %x00 == NUL, %x0a == LF, etc. */
1086 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1087 h1 <= 16 &&
1088 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1089 h2 <= 16) {
1090 strbuf_addch(sb, (h1<<4)|h2);
1091 return 3;
1092 } else
1093 return 0;
1094 case 'w':
1095 if (placeholder[1] == '(') {
1096 unsigned long width = 0, indent1 = 0, indent2 = 0;
1097 char *next;
1098 const char *start = placeholder + 2;
1099 const char *end = strchr(start, ')');
1100 if (!end)
1101 return 0;
1102 if (end > start) {
1103 width = strtoul(start, &next, 10);
1104 if (*next == ',') {
1105 indent1 = strtoul(next + 1, &next, 10);
1106 if (*next == ',') {
1107 indent2 = strtoul(next + 1,
1108 &next, 10);
1109 }
1110 }
1111 if (*next != ')')
1112 return 0;
1113 }
1114 rewrap_message_tail(sb, c, width, indent1, indent2);
1115 return end - placeholder + 1;
1116 } else
1117 return 0;
1118
1119 case '<':
1120 case '>':
1121 return parse_padding_placeholder(sb, placeholder, c);
1122 }
1123
1124 /* these depend on the commit */
1125 if (!commit->object.parsed)
1126 parse_object(commit->object.sha1);
1127
1128 switch (placeholder[0]) {
1129 case 'H': /* commit hash */
1130 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1131 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1132 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1133 return 1;
1134 case 'h': /* abbreviated commit hash */
1135 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1136 if (add_again(sb, &c->abbrev_commit_hash)) {
1137 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1138 return 1;
1139 }
1140 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1141 c->pretty_ctx->abbrev));
1142 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1143 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1144 return 1;
1145 case 'T': /* tree hash */
1146 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1147 return 1;
1148 case 't': /* abbreviated tree hash */
1149 if (add_again(sb, &c->abbrev_tree_hash))
1150 return 1;
1151 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1152 c->pretty_ctx->abbrev));
1153 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1154 return 1;
1155 case 'P': /* parent hashes */
1156 for (p = commit->parents; p; p = p->next) {
1157 if (p != commit->parents)
1158 strbuf_addch(sb, ' ');
1159 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1160 }
1161 return 1;
1162 case 'p': /* abbreviated parent hashes */
1163 if (add_again(sb, &c->abbrev_parent_hashes))
1164 return 1;
1165 for (p = commit->parents; p; p = p->next) {
1166 if (p != commit->parents)
1167 strbuf_addch(sb, ' ');
1168 strbuf_addstr(sb, find_unique_abbrev(
1169 p->item->object.sha1,
1170 c->pretty_ctx->abbrev));
1171 }
1172 c->abbrev_parent_hashes.len = sb->len -
1173 c->abbrev_parent_hashes.off;
1174 return 1;
1175 case 'm': /* left/right/bottom */
1176 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1177 return 1;
1178 case 'd':
1179 load_ref_decorations(DECORATE_SHORT_REFS);
1180 format_decorations(sb, commit, c->auto_color);
1181 return 1;
1182 case 'D':
1183 load_ref_decorations(DECORATE_SHORT_REFS);
1184 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1185 return 1;
1186 case 'g': /* reflog info */
1187 switch(placeholder[1]) {
1188 case 'd': /* reflog selector */
1189 case 'D':
1190 if (c->pretty_ctx->reflog_info)
1191 get_reflog_selector(sb,
1192 c->pretty_ctx->reflog_info,
1193 c->pretty_ctx->date_mode,
1194 c->pretty_ctx->date_mode_explicit,
1195 (placeholder[1] == 'd'));
1196 return 2;
1197 case 's': /* reflog message */
1198 if (c->pretty_ctx->reflog_info)
1199 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1200 return 2;
1201 case 'n':
1202 case 'N':
1203 case 'e':
1204 case 'E':
1205 return format_reflog_person(sb,
1206 placeholder[1],
1207 c->pretty_ctx->reflog_info,
1208 c->pretty_ctx->date_mode);
1209 }
1210 return 0; /* unknown %g placeholder */
1211 case 'N':
1212 if (c->pretty_ctx->notes_message) {
1213 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1214 return 1;
1215 }
1216 return 0;
1217 }
1218
1219 if (placeholder[0] == 'G') {
1220 if (!c->signature_check.result)
1221 check_commit_signature(c->commit, &(c->signature_check));
1222 switch (placeholder[1]) {
1223 case 'G':
1224 if (c->signature_check.gpg_output)
1225 strbuf_addstr(sb, c->signature_check.gpg_output);
1226 break;
1227 case '?':
1228 switch (c->signature_check.result) {
1229 case 'G':
1230 case 'B':
1231 case 'U':
1232 case 'N':
1233 strbuf_addch(sb, c->signature_check.result);
1234 }
1235 break;
1236 case 'S':
1237 if (c->signature_check.signer)
1238 strbuf_addstr(sb, c->signature_check.signer);
1239 break;
1240 case 'K':
1241 if (c->signature_check.key)
1242 strbuf_addstr(sb, c->signature_check.key);
1243 break;
1244 default:
1245 return 0;
1246 }
1247 return 2;
1248 }
1249
1250
1251 /* For the rest we have to parse the commit header. */
1252 if (!c->commit_header_parsed)
1253 parse_commit_header(c);
1254
1255 switch (placeholder[0]) {
1256 case 'a': /* author ... */
1257 return format_person_part(sb, placeholder[1],
1258 msg + c->author.off, c->author.len,
1259 c->pretty_ctx->date_mode);
1260 case 'c': /* committer ... */
1261 return format_person_part(sb, placeholder[1],
1262 msg + c->committer.off, c->committer.len,
1263 c->pretty_ctx->date_mode);
1264 case 'e': /* encoding */
1265 if (c->commit_encoding)
1266 strbuf_addstr(sb, c->commit_encoding);
1267 return 1;
1268 case 'B': /* raw body */
1269 /* message_off is always left at the initial newline */
1270 strbuf_addstr(sb, msg + c->message_off + 1);
1271 return 1;
1272 }
1273
1274 /* Now we need to parse the commit message. */
1275 if (!c->commit_message_parsed)
1276 parse_commit_message(c);
1277
1278 switch (placeholder[0]) {
1279 case 's': /* subject */
1280 format_subject(sb, msg + c->subject_off, " ");
1281 return 1;
1282 case 'f': /* sanitized subject */
1283 format_sanitized_subject(sb, msg + c->subject_off);
1284 return 1;
1285 case 'b': /* body */
1286 strbuf_addstr(sb, msg + c->body_off);
1287 return 1;
1288 }
1289 return 0; /* unknown placeholder */
1290 }
1291
1292 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1293 const char *placeholder,
1294 struct format_commit_context *c)
1295 {
1296 struct strbuf local_sb = STRBUF_INIT;
1297 int total_consumed = 0, len, padding = c->padding;
1298 if (padding < 0) {
1299 const char *start = strrchr(sb->buf, '\n');
1300 int occupied;
1301 if (!start)
1302 start = sb->buf;
1303 occupied = utf8_strnwidth(start, -1, 1);
1304 padding = (-padding) - occupied;
1305 }
1306 while (1) {
1307 int modifier = *placeholder == 'C';
1308 int consumed = format_commit_one(&local_sb, placeholder, c);
1309 total_consumed += consumed;
1310
1311 if (!modifier)
1312 break;
1313
1314 placeholder += consumed;
1315 if (*placeholder != '%')
1316 break;
1317 placeholder++;
1318 total_consumed++;
1319 }
1320 len = utf8_strnwidth(local_sb.buf, -1, 1);
1321
1322 if (c->flush_type == flush_left_and_steal) {
1323 const char *ch = sb->buf + sb->len - 1;
1324 while (len > padding && ch > sb->buf) {
1325 const char *p;
1326 if (*ch == ' ') {
1327 ch--;
1328 padding++;
1329 continue;
1330 }
1331 /* check for trailing ansi sequences */
1332 if (*ch != 'm')
1333 break;
1334 p = ch - 1;
1335 while (ch - p < 10 && *p != '\033')
1336 p--;
1337 if (*p != '\033' ||
1338 ch + 1 - p != display_mode_esc_sequence_len(p))
1339 break;
1340 /*
1341 * got a good ansi sequence, put it back to
1342 * local_sb as we're cutting sb
1343 */
1344 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1345 ch = p - 1;
1346 }
1347 strbuf_setlen(sb, ch + 1 - sb->buf);
1348 c->flush_type = flush_left;
1349 }
1350
1351 if (len > padding) {
1352 switch (c->truncate) {
1353 case trunc_left:
1354 strbuf_utf8_replace(&local_sb,
1355 0, len - (padding - 2),
1356 "..");
1357 break;
1358 case trunc_middle:
1359 strbuf_utf8_replace(&local_sb,
1360 padding / 2 - 1,
1361 len - (padding - 2),
1362 "..");
1363 break;
1364 case trunc_right:
1365 strbuf_utf8_replace(&local_sb,
1366 padding - 2, len - (padding - 2),
1367 "..");
1368 break;
1369 case trunc_none:
1370 break;
1371 }
1372 strbuf_addbuf(sb, &local_sb);
1373 } else {
1374 int sb_len = sb->len, offset = 0;
1375 if (c->flush_type == flush_left)
1376 offset = padding - len;
1377 else if (c->flush_type == flush_both)
1378 offset = (padding - len) / 2;
1379 /*
1380 * we calculate padding in columns, now
1381 * convert it back to chars
1382 */
1383 padding = padding - len + local_sb.len;
1384 strbuf_addchars(sb, ' ', padding);
1385 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1386 local_sb.len);
1387 }
1388 strbuf_release(&local_sb);
1389 c->flush_type = no_flush;
1390 return total_consumed;
1391 }
1392
1393 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1394 const char *placeholder,
1395 void *context)
1396 {
1397 int consumed;
1398 size_t orig_len;
1399 enum {
1400 NO_MAGIC,
1401 ADD_LF_BEFORE_NON_EMPTY,
1402 DEL_LF_BEFORE_EMPTY,
1403 ADD_SP_BEFORE_NON_EMPTY
1404 } magic = NO_MAGIC;
1405
1406 switch (placeholder[0]) {
1407 case '-':
1408 magic = DEL_LF_BEFORE_EMPTY;
1409 break;
1410 case '+':
1411 magic = ADD_LF_BEFORE_NON_EMPTY;
1412 break;
1413 case ' ':
1414 magic = ADD_SP_BEFORE_NON_EMPTY;
1415 break;
1416 default:
1417 break;
1418 }
1419 if (magic != NO_MAGIC)
1420 placeholder++;
1421
1422 orig_len = sb->len;
1423 if (((struct format_commit_context *)context)->flush_type != no_flush)
1424 consumed = format_and_pad_commit(sb, placeholder, context);
1425 else
1426 consumed = format_commit_one(sb, placeholder, context);
1427 if (magic == NO_MAGIC)
1428 return consumed;
1429
1430 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1431 while (sb->len && sb->buf[sb->len - 1] == '\n')
1432 strbuf_setlen(sb, sb->len - 1);
1433 } else if (orig_len != sb->len) {
1434 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1435 strbuf_insert(sb, orig_len, "\n", 1);
1436 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1437 strbuf_insert(sb, orig_len, " ", 1);
1438 }
1439 return consumed + 1;
1440 }
1441
1442 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1443 void *context)
1444 {
1445 struct userformat_want *w = context;
1446
1447 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1448 placeholder++;
1449
1450 switch (*placeholder) {
1451 case 'N':
1452 w->notes = 1;
1453 break;
1454 }
1455 return 0;
1456 }
1457
1458 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1459 {
1460 struct strbuf dummy = STRBUF_INIT;
1461
1462 if (!fmt) {
1463 if (!user_format)
1464 return;
1465 fmt = user_format;
1466 }
1467 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1468 strbuf_release(&dummy);
1469 }
1470
1471 void format_commit_message(const struct commit *commit,
1472 const char *format, struct strbuf *sb,
1473 const struct pretty_print_context *pretty_ctx)
1474 {
1475 struct format_commit_context context;
1476 const char *output_enc = pretty_ctx->output_encoding;
1477 const char *utf8 = "UTF-8";
1478
1479 memset(&context, 0, sizeof(context));
1480 context.commit = commit;
1481 context.pretty_ctx = pretty_ctx;
1482 context.wrap_start = sb->len;
1483 /*
1484 * convert a commit message to UTF-8 first
1485 * as far as 'format_commit_item' assumes it in UTF-8
1486 */
1487 context.message = logmsg_reencode(commit,
1488 &context.commit_encoding,
1489 utf8);
1490
1491 strbuf_expand(sb, format, format_commit_item, &context);
1492 rewrap_message_tail(sb, &context, 0, 0, 0);
1493
1494 /* then convert a commit message to an actual output encoding */
1495 if (output_enc) {
1496 if (same_encoding(utf8, output_enc))
1497 output_enc = NULL;
1498 } else {
1499 if (context.commit_encoding &&
1500 !same_encoding(context.commit_encoding, utf8))
1501 output_enc = context.commit_encoding;
1502 }
1503
1504 if (output_enc) {
1505 int outsz;
1506 char *out = reencode_string_len(sb->buf, sb->len,
1507 output_enc, utf8, &outsz);
1508 if (out)
1509 strbuf_attach(sb, out, outsz, outsz + 1);
1510 }
1511
1512 free(context.commit_encoding);
1513 unuse_commit_buffer(commit, context.message);
1514 }
1515
1516 static void pp_header(struct pretty_print_context *pp,
1517 const char *encoding,
1518 const struct commit *commit,
1519 const char **msg_p,
1520 struct strbuf *sb)
1521 {
1522 int parents_shown = 0;
1523
1524 for (;;) {
1525 const char *line = *msg_p;
1526 int linelen = get_one_line(*msg_p);
1527
1528 if (!linelen)
1529 return;
1530 *msg_p += linelen;
1531
1532 if (linelen == 1)
1533 /* End of header */
1534 return;
1535
1536 if (pp->fmt == CMIT_FMT_RAW) {
1537 strbuf_add(sb, line, linelen);
1538 continue;
1539 }
1540
1541 if (starts_with(line, "parent ")) {
1542 if (linelen != 48)
1543 die("bad parent line in commit");
1544 continue;
1545 }
1546
1547 if (!parents_shown) {
1548 unsigned num = commit_list_count(commit->parents);
1549 /* with enough slop */
1550 strbuf_grow(sb, num * 50 + 20);
1551 add_merge_info(pp, sb, commit);
1552 parents_shown = 1;
1553 }
1554
1555 /*
1556 * MEDIUM == DEFAULT shows only author with dates.
1557 * FULL shows both authors but not dates.
1558 * FULLER shows both authors and dates.
1559 */
1560 if (starts_with(line, "author ")) {
1561 strbuf_grow(sb, linelen + 80);
1562 pp_user_info(pp, "Author", sb, line + 7, encoding);
1563 }
1564 if (starts_with(line, "committer ") &&
1565 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1566 strbuf_grow(sb, linelen + 80);
1567 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1568 }
1569 }
1570 }
1571
1572 void pp_title_line(struct pretty_print_context *pp,
1573 const char **msg_p,
1574 struct strbuf *sb,
1575 const char *encoding,
1576 int need_8bit_cte)
1577 {
1578 static const int max_length = 78; /* per rfc2047 */
1579 struct strbuf title;
1580
1581 strbuf_init(&title, 80);
1582 *msg_p = format_subject(&title, *msg_p,
1583 pp->preserve_subject ? "\n" : " ");
1584
1585 strbuf_grow(sb, title.len + 1024);
1586 if (pp->subject) {
1587 strbuf_addstr(sb, pp->subject);
1588 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1589 add_rfc2047(sb, title.buf, title.len,
1590 encoding, RFC2047_SUBJECT);
1591 else
1592 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1593 -last_line_length(sb), 1, max_length);
1594 } else {
1595 strbuf_addbuf(sb, &title);
1596 }
1597 strbuf_addch(sb, '\n');
1598
1599 if (need_8bit_cte == 0) {
1600 int i;
1601 for (i = 0; i < pp->in_body_headers.nr; i++) {
1602 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1603 need_8bit_cte = 1;
1604 break;
1605 }
1606 }
1607 }
1608
1609 if (need_8bit_cte > 0) {
1610 const char *header_fmt =
1611 "MIME-Version: 1.0\n"
1612 "Content-Type: text/plain; charset=%s\n"
1613 "Content-Transfer-Encoding: 8bit\n";
1614 strbuf_addf(sb, header_fmt, encoding);
1615 }
1616 if (pp->after_subject) {
1617 strbuf_addstr(sb, pp->after_subject);
1618 }
1619 if (pp->fmt == CMIT_FMT_EMAIL) {
1620 strbuf_addch(sb, '\n');
1621 }
1622
1623 if (pp->in_body_headers.nr) {
1624 int i;
1625 for (i = 0; i < pp->in_body_headers.nr; i++) {
1626 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1627 free(pp->in_body_headers.items[i].string);
1628 }
1629 string_list_clear(&pp->in_body_headers, 0);
1630 strbuf_addch(sb, '\n');
1631 }
1632
1633 strbuf_release(&title);
1634 }
1635
1636 void pp_remainder(struct pretty_print_context *pp,
1637 const char **msg_p,
1638 struct strbuf *sb,
1639 int indent)
1640 {
1641 int first = 1;
1642 for (;;) {
1643 const char *line = *msg_p;
1644 int linelen = get_one_line(line);
1645 *msg_p += linelen;
1646
1647 if (!linelen)
1648 break;
1649
1650 if (is_empty_line(line, &linelen)) {
1651 if (first)
1652 continue;
1653 if (pp->fmt == CMIT_FMT_SHORT)
1654 break;
1655 }
1656 first = 0;
1657
1658 strbuf_grow(sb, linelen + indent + 20);
1659 if (indent)
1660 strbuf_addchars(sb, ' ', indent);
1661 strbuf_add(sb, line, linelen);
1662 strbuf_addch(sb, '\n');
1663 }
1664 }
1665
1666 void pretty_print_commit(struct pretty_print_context *pp,
1667 const struct commit *commit,
1668 struct strbuf *sb)
1669 {
1670 unsigned long beginning_of_body;
1671 int indent = 4;
1672 const char *msg;
1673 const char *reencoded;
1674 const char *encoding;
1675 int need_8bit_cte = pp->need_8bit_cte;
1676
1677 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1678 format_commit_message(commit, user_format, sb, pp);
1679 return;
1680 }
1681
1682 encoding = get_log_output_encoding();
1683 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1684
1685 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1686 indent = 0;
1687
1688 /*
1689 * We need to check and emit Content-type: to mark it
1690 * as 8-bit if we haven't done so.
1691 */
1692 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1693 int i, ch, in_body;
1694
1695 for (in_body = i = 0; (ch = msg[i]); i++) {
1696 if (!in_body) {
1697 /* author could be non 7-bit ASCII but
1698 * the log may be so; skip over the
1699 * header part first.
1700 */
1701 if (ch == '\n' && msg[i+1] == '\n')
1702 in_body = 1;
1703 }
1704 else if (non_ascii(ch)) {
1705 need_8bit_cte = 1;
1706 break;
1707 }
1708 }
1709 }
1710
1711 pp_header(pp, encoding, commit, &msg, sb);
1712 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1713 strbuf_addch(sb, '\n');
1714 }
1715
1716 /* Skip excess blank lines at the beginning of body, if any... */
1717 msg = skip_empty_lines(msg);
1718
1719 /* These formats treat the title line specially. */
1720 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1721 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1722
1723 beginning_of_body = sb->len;
1724 if (pp->fmt != CMIT_FMT_ONELINE)
1725 pp_remainder(pp, &msg, sb, indent);
1726 strbuf_rtrim(sb);
1727
1728 /* Make sure there is an EOLN for the non-oneline case */
1729 if (pp->fmt != CMIT_FMT_ONELINE)
1730 strbuf_addch(sb, '\n');
1731
1732 /*
1733 * The caller may append additional body text in e-mail
1734 * format. Make sure we did not strip the blank line
1735 * between the header and the body.
1736 */
1737 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1738 strbuf_addch(sb, '\n');
1739
1740 unuse_commit_buffer(commit, reencoded);
1741 }
1742
1743 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1744 struct strbuf *sb)
1745 {
1746 struct pretty_print_context pp = {0};
1747 pp.fmt = fmt;
1748 pretty_print_commit(&pp, commit, sb);
1749 }
The Best VCS Around