?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'jk/prompt-stash-could-be-packed' into jch
[git/jrn.git] / pretty.c
blob44b9f643725c68325a9caf92dff819e22e38e3ad
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 struct commit *commit, const char *msg,
558 const char *key)
559 {
560 int key_len = strlen(key);
561 const char *line = msg;
562
563 while (line) {
564 const char *eol = strchrnul(line, '\n'), *next;
565
566 if (line == eol)
567 return NULL;
568 if (!*eol) {
569 warning("malformed commit (header is missing newline): %s",
570 sha1_to_hex(commit->object.sha1));
571 next = NULL;
572 } else
573 next = eol + 1;
574 if (eol - line > key_len &&
575 !strncmp(line, key, key_len) &&
576 line[key_len] == ' ') {
577 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
578 }
579 line = next;
580 }
581 return NULL;
582 }
583
584 static char *replace_encoding_header(char *buf, const char *encoding)
585 {
586 struct strbuf tmp = STRBUF_INIT;
587 size_t start, len;
588 char *cp = buf;
589
590 /* guess if there is an encoding header before a \n\n */
591 while (strncmp(cp, "encoding ", strlen("encoding "))) {
592 cp = strchr(cp, '\n');
593 if (!cp || *++cp == '\n')
594 return buf;
595 }
596 start = cp - buf;
597 cp = strchr(cp, '\n');
598 if (!cp)
599 return buf; /* should not happen but be defensive */
600 len = cp + 1 - (buf + start);
601
602 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
603 if (is_encoding_utf8(encoding)) {
604 /* we have re-coded to UTF-8; drop the header */
605 strbuf_remove(&tmp, start, len);
606 } else {
607 /* just replaces XXXX in 'encoding XXXX\n' */
608 strbuf_splice(&tmp, start + strlen("encoding "),
609 len - strlen("encoding \n"),
610 encoding, strlen(encoding));
611 }
612 return strbuf_detach(&tmp, NULL);
613 }
614
615 const char *logmsg_reencode(const struct commit *commit,
616 char **commit_encoding,
617 const char *output_encoding)
618 {
619 static const char *utf8 = "UTF-8";
620 const char *use_encoding;
621 char *encoding;
622 const char *msg = get_commit_buffer(commit, NULL);
623 char *out;
624
625 if (!output_encoding || !*output_encoding) {
626 if (commit_encoding)
627 *commit_encoding =
628 get_header(commit, msg, "encoding");
629 return msg;
630 }
631 encoding = get_header(commit, msg, "encoding");
632 if (commit_encoding)
633 *commit_encoding = encoding;
634 use_encoding = encoding ? encoding : utf8;
635 if (same_encoding(use_encoding, output_encoding)) {
636 /*
637 * No encoding work to be done. If we have no encoding header
638 * at all, then there's nothing to do, and we can return the
639 * message verbatim (whether newly allocated or not).
640 */
641 if (!encoding)
642 return msg;
643
644 /*
645 * Otherwise, we still want to munge the encoding header in the
646 * result, which will be done by modifying the buffer. If we
647 * are using a fresh copy, we can reuse it. But if we are using
648 * the cached copy from get_commit_buffer, we need to duplicate it
649 * to avoid munging the cached copy.
650 */
651 if (msg == get_cached_commit_buffer(commit, NULL))
652 out = xstrdup(msg);
653 else
654 out = (char *)msg;
655 }
656 else {
657 /*
658 * There's actual encoding work to do. Do the reencoding, which
659 * still leaves the header to be replaced in the next step. At
660 * this point, we are done with msg. If we allocated a fresh
661 * copy, we can free it.
662 */
663 out = reencode_string(msg, output_encoding, use_encoding);
664 if (out)
665 unuse_commit_buffer(commit, msg);
666 }
667
668 /*
669 * This replacement actually consumes the buffer we hand it, so we do
670 * not have to worry about freeing the old "out" here.
671 */
672 if (out)
673 out = replace_encoding_header(out, output_encoding);
674
675 if (!commit_encoding)
676 free(encoding);
677 /*
678 * If the re-encoding failed, out might be NULL here; in that
679 * case we just return the commit message verbatim.
680 */
681 return out ? out : msg;
682 }
683
684 static int mailmap_name(const char **email, size_t *email_len,
685 const char **name, size_t *name_len)
686 {
687 static struct string_list *mail_map;
688 if (!mail_map) {
689 mail_map = xcalloc(1, sizeof(*mail_map));
690 read_mailmap(mail_map, NULL);
691 }
692 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
693 }
694
695 static size_t format_person_part(struct strbuf *sb, char part,
696 const char *msg, int len, enum date_mode dmode)
697 {
698 /* currently all placeholders have same length */
699 const int placeholder_len = 2;
700 struct ident_split s;
701 const char *name, *mail;
702 size_t maillen, namelen;
703
704 if (split_ident_line(&s, msg, len) < 0)
705 goto skip;
706
707 name = s.name_begin;
708 namelen = s.name_end - s.name_begin;
709 mail = s.mail_begin;
710 maillen = s.mail_end - s.mail_begin;
711
712 if (part == 'N' || part == 'E') /* mailmap lookup */
713 mailmap_name(&mail, &maillen, &name, &namelen);
714 if (part == 'n' || part == 'N') { /* name */
715 strbuf_add(sb, name, namelen);
716 return placeholder_len;
717 }
718 if (part == 'e' || part == 'E') { /* email */
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_RFC2822));
737 return placeholder_len;
738 case 'r': /* date, relative */
739 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
740 return placeholder_len;
741 case 'i': /* date, ISO 8601 */
742 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
743 return placeholder_len;
744 }
745
746 skip:
747 /*
748 * reading from either a bogus commit, or a reflog entry with
749 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
750 * to compute a valid return value.
751 */
752 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
753 || part == 'D' || part == 'r' || part == 'i')
754 return placeholder_len;
755
756 return 0; /* unknown placeholder */
757 }
758
759 struct chunk {
760 size_t off;
761 size_t len;
762 };
763
764 enum flush_type {
765 no_flush,
766 flush_right,
767 flush_left,
768 flush_left_and_steal,
769 flush_both
770 };
771
772 enum trunc_type {
773 trunc_none,
774 trunc_left,
775 trunc_middle,
776 trunc_right
777 };
778
779 struct format_commit_context {
780 const struct commit *commit;
781 const struct pretty_print_context *pretty_ctx;
782 unsigned commit_header_parsed:1;
783 unsigned commit_message_parsed:1;
784 struct signature_check signature_check;
785 enum flush_type flush_type;
786 enum trunc_type truncate;
787 const char *message;
788 char *commit_encoding;
789 size_t width, indent1, indent2;
790 int auto_color;
791 int padding;
792
793 /* These offsets are relative to the start of the commit message. */
794 struct chunk author;
795 struct chunk committer;
796 size_t message_off;
797 size_t subject_off;
798 size_t body_off;
799
800 /* The following ones are relative to the result struct strbuf. */
801 struct chunk abbrev_commit_hash;
802 struct chunk abbrev_tree_hash;
803 struct chunk abbrev_parent_hashes;
804 size_t wrap_start;
805 };
806
807 static int add_again(struct strbuf *sb, struct chunk *chunk)
808 {
809 if (chunk->len) {
810 strbuf_adddup(sb, chunk->off, chunk->len);
811 return 1;
812 }
813
814 /*
815 * We haven't seen this chunk before. Our caller is surely
816 * going to add it the hard way now. Remember the most likely
817 * start of the to-be-added chunk: the current end of the
818 * struct strbuf.
819 */
820 chunk->off = sb->len;
821 return 0;
822 }
823
824 static void parse_commit_header(struct format_commit_context *context)
825 {
826 const char *msg = context->message;
827 int i;
828
829 for (i = 0; msg[i]; i++) {
830 int eol;
831 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
832 ; /* do nothing */
833
834 if (i == eol) {
835 break;
836 } else if (starts_with(msg + i, "author ")) {
837 context->author.off = i + 7;
838 context->author.len = eol - i - 7;
839 } else if (starts_with(msg + i, "committer ")) {
840 context->committer.off = i + 10;
841 context->committer.len = eol - i - 10;
842 }
843 i = eol;
844 }
845 context->message_off = i;
846 context->commit_header_parsed = 1;
847 }
848
849 static int istitlechar(char c)
850 {
851 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
852 (c >= '0' && c <= '9') || c == '.' || c == '_';
853 }
854
855 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
856 {
857 size_t trimlen;
858 size_t start_len = sb->len;
859 int space = 2;
860
861 for (; *msg && *msg != '\n'; msg++) {
862 if (istitlechar(*msg)) {
863 if (space == 1)
864 strbuf_addch(sb, '-');
865 space = 0;
866 strbuf_addch(sb, *msg);
867 if (*msg == '.')
868 while (*(msg+1) == '.')
869 msg++;
870 } else
871 space |= 1;
872 }
873
874 /* trim any trailing '.' or '-' characters */
875 trimlen = 0;
876 while (sb->len - trimlen > start_len &&
877 (sb->buf[sb->len - 1 - trimlen] == '.'
878 || sb->buf[sb->len - 1 - trimlen] == '-'))
879 trimlen++;
880 strbuf_remove(sb, sb->len - trimlen, trimlen);
881 }
882
883 const char *format_subject(struct strbuf *sb, const char *msg,
884 const char *line_separator)
885 {
886 int first = 1;
887
888 for (;;) {
889 const char *line = msg;
890 int linelen = get_one_line(line);
891
892 msg += linelen;
893 if (!linelen || is_empty_line(line, &linelen))
894 break;
895
896 if (!sb)
897 continue;
898 strbuf_grow(sb, linelen + 2);
899 if (!first)
900 strbuf_addstr(sb, line_separator);
901 strbuf_add(sb, line, linelen);
902 first = 0;
903 }
904 return msg;
905 }
906
907 static void parse_commit_message(struct format_commit_context *c)
908 {
909 const char *msg = c->message + c->message_off;
910 const char *start = c->message;
911
912 msg = skip_empty_lines(msg);
913 c->subject_off = msg - start;
914
915 msg = format_subject(NULL, msg, NULL);
916 msg = skip_empty_lines(msg);
917 c->body_off = msg - start;
918
919 c->commit_message_parsed = 1;
920 }
921
922 static void strbuf_wrap(struct strbuf *sb, size_t pos,
923 size_t width, size_t indent1, size_t indent2)
924 {
925 struct strbuf tmp = STRBUF_INIT;
926
927 if (pos)
928 strbuf_add(&tmp, sb->buf, pos);
929 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
930 (int) indent1, (int) indent2, (int) width);
931 strbuf_swap(&tmp, sb);
932 strbuf_release(&tmp);
933 }
934
935 static void rewrap_message_tail(struct strbuf *sb,
936 struct format_commit_context *c,
937 size_t new_width, size_t new_indent1,
938 size_t new_indent2)
939 {
940 if (c->width == new_width && c->indent1 == new_indent1 &&
941 c->indent2 == new_indent2)
942 return;
943 if (c->wrap_start < sb->len)
944 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
945 c->wrap_start = sb->len;
946 c->width = new_width;
947 c->indent1 = new_indent1;
948 c->indent2 = new_indent2;
949 }
950
951 static int format_reflog_person(struct strbuf *sb,
952 char part,
953 struct reflog_walk_info *log,
954 enum date_mode dmode)
955 {
956 const char *ident;
957
958 if (!log)
959 return 2;
960
961 ident = get_reflog_ident(log);
962 if (!ident)
963 return 2;
964
965 return format_person_part(sb, part, ident, strlen(ident), dmode);
966 }
967
968 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
969 const char *placeholder,
970 struct format_commit_context *c)
971 {
972 if (placeholder[1] == '(') {
973 const char *begin = placeholder + 2;
974 const char *end = strchr(begin, ')');
975 char color[COLOR_MAXLEN];
976
977 if (!end)
978 return 0;
979 if (starts_with(begin, "auto,")) {
980 if (!want_color(c->pretty_ctx->color))
981 return end - placeholder + 1;
982 begin += 5;
983 }
984 color_parse_mem(begin,
985 end - begin,
986 "--pretty format", color);
987 strbuf_addstr(sb, color);
988 return end - placeholder + 1;
989 }
990 if (starts_with(placeholder + 1, "red")) {
991 strbuf_addstr(sb, GIT_COLOR_RED);
992 return 4;
993 } else if (starts_with(placeholder + 1, "green")) {
994 strbuf_addstr(sb, GIT_COLOR_GREEN);
995 return 6;
996 } else if (starts_with(placeholder + 1, "blue")) {
997 strbuf_addstr(sb, GIT_COLOR_BLUE);
998 return 5;
999 } else if (starts_with(placeholder + 1, "reset")) {
1000 strbuf_addstr(sb, GIT_COLOR_RESET);
1001 return 6;
1002 } else
1003 return 0;
1004 }
1005
1006 static size_t parse_padding_placeholder(struct strbuf *sb,
1007 const char *placeholder,
1008 struct format_commit_context *c)
1009 {
1010 const char *ch = placeholder;
1011 enum flush_type flush_type;
1012 int to_column = 0;
1013
1014 switch (*ch++) {
1015 case '<':
1016 flush_type = flush_right;
1017 break;
1018 case '>':
1019 if (*ch == '<') {
1020 flush_type = flush_both;
1021 ch++;
1022 } else if (*ch == '>') {
1023 flush_type = flush_left_and_steal;
1024 ch++;
1025 } else
1026 flush_type = flush_left;
1027 break;
1028 default:
1029 return 0;
1030 }
1031
1032 /* the next value means "wide enough to that column" */
1033 if (*ch == '|') {
1034 to_column = 1;
1035 ch++;
1036 }
1037
1038 if (*ch == '(') {
1039 const char *start = ch + 1;
1040 const char *end = start + strcspn(start, ",)");
1041 char *next;
1042 int width;
1043 if (!end || end == start)
1044 return 0;
1045 width = strtoul(start, &next, 10);
1046 if (next == start || width == 0)
1047 return 0;
1048 c->padding = to_column ? -width : width;
1049 c->flush_type = flush_type;
1050
1051 if (*end == ',') {
1052 start = end + 1;
1053 end = strchr(start, ')');
1054 if (!end || end == start)
1055 return 0;
1056 if (starts_with(start, "trunc)"))
1057 c->truncate = trunc_right;
1058 else if (starts_with(start, "ltrunc)"))
1059 c->truncate = trunc_left;
1060 else if (starts_with(start, "mtrunc)"))
1061 c->truncate = trunc_middle;
1062 else
1063 return 0;
1064 } else
1065 c->truncate = trunc_none;
1066
1067 return end - placeholder + 1;
1068 }
1069 return 0;
1070 }
1071
1072 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1073 const char *placeholder,
1074 void *context)
1075 {
1076 struct format_commit_context *c = context;
1077 const struct commit *commit = c->commit;
1078 const char *msg = c->message;
1079 struct commit_list *p;
1080 int h1, h2;
1081
1082 /* these are independent of the commit */
1083 switch (placeholder[0]) {
1084 case 'C':
1085 if (starts_with(placeholder + 1, "(auto)")) {
1086 c->auto_color = 1;
1087 return 7; /* consumed 7 bytes, "C(auto)" */
1088 } else {
1089 int ret = parse_color(sb, placeholder, c);
1090 if (ret)
1091 c->auto_color = 0;
1092 /*
1093 * Otherwise, we decided to treat %C<unknown>
1094 * as a literal string, and the previous
1095 * %C(auto) is still valid.
1096 */
1097 return ret;
1098 }
1099 case 'n': /* newline */
1100 strbuf_addch(sb, '\n');
1101 return 1;
1102 case 'x':
1103 /* %x00 == NUL, %x0a == LF, etc. */
1104 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1105 h1 <= 16 &&
1106 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1107 h2 <= 16) {
1108 strbuf_addch(sb, (h1<<4)|h2);
1109 return 3;
1110 } else
1111 return 0;
1112 case 'w':
1113 if (placeholder[1] == '(') {
1114 unsigned long width = 0, indent1 = 0, indent2 = 0;
1115 char *next;
1116 const char *start = placeholder + 2;
1117 const char *end = strchr(start, ')');
1118 if (!end)
1119 return 0;
1120 if (end > start) {
1121 width = strtoul(start, &next, 10);
1122 if (*next == ',') {
1123 indent1 = strtoul(next + 1, &next, 10);
1124 if (*next == ',') {
1125 indent2 = strtoul(next + 1,
1126 &next, 10);
1127 }
1128 }
1129 if (*next != ')')
1130 return 0;
1131 }
1132 rewrap_message_tail(sb, c, width, indent1, indent2);
1133 return end - placeholder + 1;
1134 } else
1135 return 0;
1136
1137 case '<':
1138 case '>':
1139 return parse_padding_placeholder(sb, placeholder, c);
1140 }
1141
1142 /* these depend on the commit */
1143 if (!commit->object.parsed)
1144 parse_object(commit->object.sha1);
1145
1146 switch (placeholder[0]) {
1147 case 'H': /* commit hash */
1148 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1149 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1150 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1151 return 1;
1152 case 'h': /* abbreviated commit hash */
1153 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1154 if (add_again(sb, &c->abbrev_commit_hash)) {
1155 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1156 return 1;
1157 }
1158 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1159 c->pretty_ctx->abbrev));
1160 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1161 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1162 return 1;
1163 case 'T': /* tree hash */
1164 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1165 return 1;
1166 case 't': /* abbreviated tree hash */
1167 if (add_again(sb, &c->abbrev_tree_hash))
1168 return 1;
1169 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1170 c->pretty_ctx->abbrev));
1171 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1172 return 1;
1173 case 'P': /* parent hashes */
1174 for (p = commit->parents; p; p = p->next) {
1175 if (p != commit->parents)
1176 strbuf_addch(sb, ' ');
1177 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1178 }
1179 return 1;
1180 case 'p': /* abbreviated parent hashes */
1181 if (add_again(sb, &c->abbrev_parent_hashes))
1182 return 1;
1183 for (p = commit->parents; p; p = p->next) {
1184 if (p != commit->parents)
1185 strbuf_addch(sb, ' ');
1186 strbuf_addstr(sb, find_unique_abbrev(
1187 p->item->object.sha1,
1188 c->pretty_ctx->abbrev));
1189 }
1190 c->abbrev_parent_hashes.len = sb->len -
1191 c->abbrev_parent_hashes.off;
1192 return 1;
1193 case 'm': /* left/right/bottom */
1194 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1195 return 1;
1196 case 'd':
1197 load_ref_decorations(DECORATE_SHORT_REFS);
1198 format_decorations(sb, commit, c->auto_color);
1199 return 1;
1200 case 'g': /* reflog info */
1201 switch(placeholder[1]) {
1202 case 'd': /* reflog selector */
1203 case 'D':
1204 if (c->pretty_ctx->reflog_info)
1205 get_reflog_selector(sb,
1206 c->pretty_ctx->reflog_info,
1207 c->pretty_ctx->date_mode,
1208 c->pretty_ctx->date_mode_explicit,
1209 (placeholder[1] == 'd'));
1210 return 2;
1211 case 's': /* reflog message */
1212 if (c->pretty_ctx->reflog_info)
1213 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1214 return 2;
1215 case 'n':
1216 case 'N':
1217 case 'e':
1218 case 'E':
1219 return format_reflog_person(sb,
1220 placeholder[1],
1221 c->pretty_ctx->reflog_info,
1222 c->pretty_ctx->date_mode);
1223 }
1224 return 0; /* unknown %g placeholder */
1225 case 'N':
1226 if (c->pretty_ctx->notes_message) {
1227 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1228 return 1;
1229 }
1230 return 0;
1231 }
1232
1233 if (placeholder[0] == 'G') {
1234 if (!c->signature_check.result)
1235 check_commit_signature(c->commit, &(c->signature_check));
1236 switch (placeholder[1]) {
1237 case 'G':
1238 if (c->signature_check.gpg_output)
1239 strbuf_addstr(sb, c->signature_check.gpg_output);
1240 break;
1241 case '?':
1242 switch (c->signature_check.result) {
1243 case 'G':
1244 case 'B':
1245 case 'U':
1246 case 'N':
1247 strbuf_addch(sb, c->signature_check.result);
1248 }
1249 break;
1250 case 'S':
1251 if (c->signature_check.signer)
1252 strbuf_addstr(sb, c->signature_check.signer);
1253 break;
1254 case 'K':
1255 if (c->signature_check.key)
1256 strbuf_addstr(sb, c->signature_check.key);
1257 break;
1258 default:
1259 return 0;
1260 }
1261 return 2;
1262 }
1263
1264
1265 /* For the rest we have to parse the commit header. */
1266 if (!c->commit_header_parsed)
1267 parse_commit_header(c);
1268
1269 switch (placeholder[0]) {
1270 case 'a': /* author ... */
1271 return format_person_part(sb, placeholder[1],
1272 msg + c->author.off, c->author.len,
1273 c->pretty_ctx->date_mode);
1274 case 'c': /* committer ... */
1275 return format_person_part(sb, placeholder[1],
1276 msg + c->committer.off, c->committer.len,
1277 c->pretty_ctx->date_mode);
1278 case 'e': /* encoding */
1279 if (c->commit_encoding)
1280 strbuf_addstr(sb, c->commit_encoding);
1281 return 1;
1282 case 'B': /* raw body */
1283 /* message_off is always left at the initial newline */
1284 strbuf_addstr(sb, msg + c->message_off + 1);
1285 return 1;
1286 }
1287
1288 /* Now we need to parse the commit message. */
1289 if (!c->commit_message_parsed)
1290 parse_commit_message(c);
1291
1292 switch (placeholder[0]) {
1293 case 's': /* subject */
1294 format_subject(sb, msg + c->subject_off, " ");
1295 return 1;
1296 case 'f': /* sanitized subject */
1297 format_sanitized_subject(sb, msg + c->subject_off);
1298 return 1;
1299 case 'b': /* body */
1300 strbuf_addstr(sb, msg + c->body_off);
1301 return 1;
1302 }
1303 return 0; /* unknown placeholder */
1304 }
1305
1306 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1307 const char *placeholder,
1308 struct format_commit_context *c)
1309 {
1310 struct strbuf local_sb = STRBUF_INIT;
1311 int total_consumed = 0, len, padding = c->padding;
1312 if (padding < 0) {
1313 const char *start = strrchr(sb->buf, '\n');
1314 int occupied;
1315 if (!start)
1316 start = sb->buf;
1317 occupied = utf8_strnwidth(start, -1, 1);
1318 padding = (-padding) - occupied;
1319 }
1320 while (1) {
1321 int modifier = *placeholder == 'C';
1322 int consumed = format_commit_one(&local_sb, placeholder, c);
1323 total_consumed += consumed;
1324
1325 if (!modifier)
1326 break;
1327
1328 placeholder += consumed;
1329 if (*placeholder != '%')
1330 break;
1331 placeholder++;
1332 total_consumed++;
1333 }
1334 len = utf8_strnwidth(local_sb.buf, -1, 1);
1335
1336 if (c->flush_type == flush_left_and_steal) {
1337 const char *ch = sb->buf + sb->len - 1;
1338 while (len > padding && ch > sb->buf) {
1339 const char *p;
1340 if (*ch == ' ') {
1341 ch--;
1342 padding++;
1343 continue;
1344 }
1345 /* check for trailing ansi sequences */
1346 if (*ch != 'm')
1347 break;
1348 p = ch - 1;
1349 while (ch - p < 10 && *p != '\033')
1350 p--;
1351 if (*p != '\033' ||
1352 ch + 1 - p != display_mode_esc_sequence_len(p))
1353 break;
1354 /*
1355 * got a good ansi sequence, put it back to
1356 * local_sb as we're cutting sb
1357 */
1358 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1359 ch = p - 1;
1360 }
1361 strbuf_setlen(sb, ch + 1 - sb->buf);
1362 c->flush_type = flush_left;
1363 }
1364
1365 if (len > padding) {
1366 switch (c->truncate) {
1367 case trunc_left:
1368 strbuf_utf8_replace(&local_sb,
1369 0, len - (padding - 2),
1370 "..");
1371 break;
1372 case trunc_middle:
1373 strbuf_utf8_replace(&local_sb,
1374 padding / 2 - 1,
1375 len - (padding - 2),
1376 "..");
1377 break;
1378 case trunc_right:
1379 strbuf_utf8_replace(&local_sb,
1380 padding - 2, len - (padding - 2),
1381 "..");
1382 break;
1383 case trunc_none:
1384 break;
1385 }
1386 strbuf_addbuf(sb, &local_sb);
1387 } else {
1388 int sb_len = sb->len, offset = 0;
1389 if (c->flush_type == flush_left)
1390 offset = padding - len;
1391 else if (c->flush_type == flush_both)
1392 offset = (padding - len) / 2;
1393 /*
1394 * we calculate padding in columns, now
1395 * convert it back to chars
1396 */
1397 padding = padding - len + local_sb.len;
1398 strbuf_grow(sb, padding);
1399 strbuf_setlen(sb, sb_len + padding);
1400 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1401 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1402 local_sb.len);
1403 }
1404 strbuf_release(&local_sb);
1405 c->flush_type = no_flush;
1406 return total_consumed;
1407 }
1408
1409 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1410 const char *placeholder,
1411 void *context)
1412 {
1413 int consumed;
1414 size_t orig_len;
1415 enum {
1416 NO_MAGIC,
1417 ADD_LF_BEFORE_NON_EMPTY,
1418 DEL_LF_BEFORE_EMPTY,
1419 ADD_SP_BEFORE_NON_EMPTY
1420 } magic = NO_MAGIC;
1421
1422 switch (placeholder[0]) {
1423 case '-':
1424 magic = DEL_LF_BEFORE_EMPTY;
1425 break;
1426 case '+':
1427 magic = ADD_LF_BEFORE_NON_EMPTY;
1428 break;
1429 case ' ':
1430 magic = ADD_SP_BEFORE_NON_EMPTY;
1431 break;
1432 default:
1433 break;
1434 }
1435 if (magic != NO_MAGIC)
1436 placeholder++;
1437
1438 orig_len = sb->len;
1439 if (((struct format_commit_context *)context)->flush_type != no_flush)
1440 consumed = format_and_pad_commit(sb, placeholder, context);
1441 else
1442 consumed = format_commit_one(sb, placeholder, context);
1443 if (magic == NO_MAGIC)
1444 return consumed;
1445
1446 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1447 while (sb->len && sb->buf[sb->len - 1] == '\n')
1448 strbuf_setlen(sb, sb->len - 1);
1449 } else if (orig_len != sb->len) {
1450 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1451 strbuf_insert(sb, orig_len, "\n", 1);
1452 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1453 strbuf_insert(sb, orig_len, " ", 1);
1454 }
1455 return consumed + 1;
1456 }
1457
1458 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1459 void *context)
1460 {
1461 struct userformat_want *w = context;
1462
1463 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1464 placeholder++;
1465
1466 switch (*placeholder) {
1467 case 'N':
1468 w->notes = 1;
1469 break;
1470 }
1471 return 0;
1472 }
1473
1474 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1475 {
1476 struct strbuf dummy = STRBUF_INIT;
1477
1478 if (!fmt) {
1479 if (!user_format)
1480 return;
1481 fmt = user_format;
1482 }
1483 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1484 strbuf_release(&dummy);
1485 }
1486
1487 void format_commit_message(const struct commit *commit,
1488 const char *format, struct strbuf *sb,
1489 const struct pretty_print_context *pretty_ctx)
1490 {
1491 struct format_commit_context context;
1492 const char *output_enc = pretty_ctx->output_encoding;
1493 const char *utf8 = "UTF-8";
1494
1495 memset(&context, 0, sizeof(context));
1496 context.commit = commit;
1497 context.pretty_ctx = pretty_ctx;
1498 context.wrap_start = sb->len;
1499 /*
1500 * convert a commit message to UTF-8 first
1501 * as far as 'format_commit_item' assumes it in UTF-8
1502 */
1503 context.message = logmsg_reencode(commit,
1504 &context.commit_encoding,
1505 utf8);
1506
1507 strbuf_expand(sb, format, format_commit_item, &context);
1508 rewrap_message_tail(sb, &context, 0, 0, 0);
1509
1510 /* then convert a commit message to an actual output encoding */
1511 if (output_enc) {
1512 if (same_encoding(utf8, output_enc))
1513 output_enc = NULL;
1514 } else {
1515 if (context.commit_encoding &&
1516 !same_encoding(context.commit_encoding, utf8))
1517 output_enc = context.commit_encoding;
1518 }
1519
1520 if (output_enc) {
1521 int outsz;
1522 char *out = reencode_string_len(sb->buf, sb->len,
1523 output_enc, utf8, &outsz);
1524 if (out)
1525 strbuf_attach(sb, out, outsz, outsz + 1);
1526 }
1527
1528 free(context.commit_encoding);
1529 unuse_commit_buffer(commit, context.message);
1530 }
1531
1532 static void pp_header(struct pretty_print_context *pp,
1533 const char *encoding,
1534 const struct commit *commit,
1535 const char **msg_p,
1536 struct strbuf *sb)
1537 {
1538 int parents_shown = 0;
1539
1540 for (;;) {
1541 const char *line = *msg_p;
1542 int linelen = get_one_line(*msg_p);
1543
1544 if (!linelen)
1545 return;
1546 *msg_p += linelen;
1547
1548 if (linelen == 1)
1549 /* End of header */
1550 return;
1551
1552 if (pp->fmt == CMIT_FMT_RAW) {
1553 strbuf_add(sb, line, linelen);
1554 continue;
1555 }
1556
1557 if (starts_with(line, "parent ")) {
1558 if (linelen != 48)
1559 die("bad parent line in commit");
1560 continue;
1561 }
1562
1563 if (!parents_shown) {
1564 unsigned num = commit_list_count(commit->parents);
1565 /* with enough slop */
1566 strbuf_grow(sb, num * 50 + 20);
1567 add_merge_info(pp, sb, commit);
1568 parents_shown = 1;
1569 }
1570
1571 /*
1572 * MEDIUM == DEFAULT shows only author with dates.
1573 * FULL shows both authors but not dates.
1574 * FULLER shows both authors and dates.
1575 */
1576 if (starts_with(line, "author ")) {
1577 strbuf_grow(sb, linelen + 80);
1578 pp_user_info(pp, "Author", sb, line + 7, encoding);
1579 }
1580 if (starts_with(line, "committer ") &&
1581 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1582 strbuf_grow(sb, linelen + 80);
1583 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1584 }
1585 }
1586 }
1587
1588 void pp_title_line(struct pretty_print_context *pp,
1589 const char **msg_p,
1590 struct strbuf *sb,
1591 const char *encoding,
1592 int need_8bit_cte)
1593 {
1594 static const int max_length = 78; /* per rfc2047 */
1595 struct strbuf title;
1596
1597 strbuf_init(&title, 80);
1598 *msg_p = format_subject(&title, *msg_p,
1599 pp->preserve_subject ? "\n" : " ");
1600
1601 strbuf_grow(sb, title.len + 1024);
1602 if (pp->subject) {
1603 strbuf_addstr(sb, pp->subject);
1604 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1605 add_rfc2047(sb, title.buf, title.len,
1606 encoding, RFC2047_SUBJECT);
1607 else
1608 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1609 -last_line_length(sb), 1, max_length);
1610 } else {
1611 strbuf_addbuf(sb, &title);
1612 }
1613 strbuf_addch(sb, '\n');
1614
1615 if (need_8bit_cte == 0) {
1616 int i;
1617 for (i = 0; i < pp->in_body_headers.nr; i++) {
1618 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1619 need_8bit_cte = 1;
1620 break;
1621 }
1622 }
1623 }
1624
1625 if (need_8bit_cte > 0) {
1626 const char *header_fmt =
1627 "MIME-Version: 1.0\n"
1628 "Content-Type: text/plain; charset=%s\n"
1629 "Content-Transfer-Encoding: 8bit\n";
1630 strbuf_addf(sb, header_fmt, encoding);
1631 }
1632 if (pp->after_subject) {
1633 strbuf_addstr(sb, pp->after_subject);
1634 }
1635 if (pp->fmt == CMIT_FMT_EMAIL) {
1636 strbuf_addch(sb, '\n');
1637 }
1638
1639 if (pp->in_body_headers.nr) {
1640 int i;
1641 for (i = 0; i < pp->in_body_headers.nr; i++) {
1642 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1643 free(pp->in_body_headers.items[i].string);
1644 }
1645 string_list_clear(&pp->in_body_headers, 0);
1646 strbuf_addch(sb, '\n');
1647 }
1648
1649 strbuf_release(&title);
1650 }
1651
1652 void pp_remainder(struct pretty_print_context *pp,
1653 const char **msg_p,
1654 struct strbuf *sb,
1655 int indent)
1656 {
1657 int first = 1;
1658 for (;;) {
1659 const char *line = *msg_p;
1660 int linelen = get_one_line(line);
1661 *msg_p += linelen;
1662
1663 if (!linelen)
1664 break;
1665
1666 if (is_empty_line(line, &linelen)) {
1667 if (first)
1668 continue;
1669 if (pp->fmt == CMIT_FMT_SHORT)
1670 break;
1671 }
1672 first = 0;
1673
1674 strbuf_grow(sb, linelen + indent + 20);
1675 if (indent) {
1676 memset(sb->buf + sb->len, ' ', indent);
1677 strbuf_setlen(sb, sb->len + indent);
1678 }
1679 strbuf_add(sb, line, linelen);
1680 strbuf_addch(sb, '\n');
1681 }
1682 }
1683
1684 void pretty_print_commit(struct pretty_print_context *pp,
1685 const struct commit *commit,
1686 struct strbuf *sb)
1687 {
1688 unsigned long beginning_of_body;
1689 int indent = 4;
1690 const char *msg;
1691 const char *reencoded;
1692 const char *encoding;
1693 int need_8bit_cte = pp->need_8bit_cte;
1694
1695 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1696 format_commit_message(commit, user_format, sb, pp);
1697 return;
1698 }
1699
1700 encoding = get_log_output_encoding();
1701 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1702
1703 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1704 indent = 0;
1705
1706 /*
1707 * We need to check and emit Content-type: to mark it
1708 * as 8-bit if we haven't done so.
1709 */
1710 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1711 int i, ch, in_body;
1712
1713 for (in_body = i = 0; (ch = msg[i]); i++) {
1714 if (!in_body) {
1715 /* author could be non 7-bit ASCII but
1716 * the log may be so; skip over the
1717 * header part first.
1718 */
1719 if (ch == '\n' && msg[i+1] == '\n')
1720 in_body = 1;
1721 }
1722 else if (non_ascii(ch)) {
1723 need_8bit_cte = 1;
1724 break;
1725 }
1726 }
1727 }
1728
1729 pp_header(pp, encoding, commit, &msg, sb);
1730 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1731 strbuf_addch(sb, '\n');
1732 }
1733
1734 /* Skip excess blank lines at the beginning of body, if any... */
1735 msg = skip_empty_lines(msg);
1736
1737 /* These formats treat the title line specially. */
1738 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1739 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1740
1741 beginning_of_body = sb->len;
1742 if (pp->fmt != CMIT_FMT_ONELINE)
1743 pp_remainder(pp, &msg, sb, indent);
1744 strbuf_rtrim(sb);
1745
1746 /* Make sure there is an EOLN for the non-oneline case */
1747 if (pp->fmt != CMIT_FMT_ONELINE)
1748 strbuf_addch(sb, '\n');
1749
1750 /*
1751 * The caller may append additional body text in e-mail
1752 * format. Make sure we did not strip the blank line
1753 * between the header and the body.
1754 */
1755 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1756 strbuf_addch(sb, '\n');
1757
1758 unuse_commit_buffer(commit, reencoded);
1759 }
1760
1761 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1762 struct strbuf *sb)
1763 {
1764 struct pretty_print_context pp = {0};
1765 pp.fmt = fmt;
1766 pretty_print_commit(&pp, commit, sb);
1767 }
Jonathan Nieder’s miscellaneous git branches