?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
am: support --patch-format=mboxrd
[git/git-svn.git] / pretty.c
blob6abd8a12151635206d587f078015ad8b88b80b6f
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 expand_tabs_in_log;
20 int is_alias;
21 const char *user_format;
22 } *commit_formats;
23 static size_t builtin_formats_len;
24 static size_t commit_formats_len;
25 static size_t commit_formats_alloc;
26 static struct cmt_fmt_map *find_commit_format(const char *sought);
27
28 int commit_format_is_empty(enum cmit_fmt fmt)
29 {
30 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
31 }
32
33 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
34 {
35 free(user_format);
36 user_format = xstrdup(cp);
37 if (is_tformat)
38 rev->use_terminator = 1;
39 rev->commit_format = CMIT_FMT_USERFORMAT;
40 }
41
42 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
43 {
44 struct cmt_fmt_map *commit_format = NULL;
45 const char *name;
46 const char *fmt;
47 int i;
48
49 if (!skip_prefix(var, "pretty.", &name))
50 return 0;
51
52 for (i = 0; i < builtin_formats_len; i++) {
53 if (!strcmp(commit_formats[i].name, name))
54 return 0;
55 }
56
57 for (i = builtin_formats_len; i < commit_formats_len; i++) {
58 if (!strcmp(commit_formats[i].name, name)) {
59 commit_format = &commit_formats[i];
60 break;
61 }
62 }
63
64 if (!commit_format) {
65 ALLOC_GROW(commit_formats, commit_formats_len+1,
66 commit_formats_alloc);
67 commit_format = &commit_formats[commit_formats_len];
68 memset(commit_format, 0, sizeof(*commit_format));
69 commit_formats_len++;
70 }
71
72 commit_format->name = xstrdup(name);
73 commit_format->format = CMIT_FMT_USERFORMAT;
74 if (git_config_string(&fmt, var, value))
75 return -1;
76
77 if (skip_prefix(fmt, "format:", &fmt))
78 commit_format->is_tformat = 0;
79 else if (skip_prefix(fmt, "tformat:", &fmt) || 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, 0 },
92 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
93 { "short", CMIT_FMT_SHORT, 0, 0 },
94 { "email", CMIT_FMT_EMAIL, 0, 0 },
95 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
96 { "fuller", CMIT_FMT_FULLER, 0, 8 },
97 { "full", CMIT_FMT_FULL, 0, 8 },
98 { "oneline", CMIT_FMT_ONELINE, 1, 0 }
99 };
100 commit_formats_len = ARRAY_SIZE(builtin_formats);
101 builtin_formats_len = commit_formats_len;
102 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
103 memcpy(commit_formats, builtin_formats,
104 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
105
106 git_config(git_pretty_formats_config, NULL);
107 }
108
109 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
110 const char *original,
111 int num_redirections)
112 {
113 struct cmt_fmt_map *found = NULL;
114 size_t found_match_len = 0;
115 int i;
116
117 if (num_redirections >= commit_formats_len)
118 die("invalid --pretty format: "
119 "'%s' references an alias which points to itself",
120 original);
121
122 for (i = 0; i < commit_formats_len; i++) {
123 size_t match_len;
124
125 if (!starts_with(commit_formats[i].name, sought))
126 continue;
127
128 match_len = strlen(commit_formats[i].name);
129 if (found == NULL || found_match_len > match_len) {
130 found = &commit_formats[i];
131 found_match_len = match_len;
132 }
133 }
134
135 if (found && found->is_alias) {
136 found = find_commit_format_recursive(found->user_format,
137 original,
138 num_redirections+1);
139 }
140
141 return found;
142 }
143
144 static struct cmt_fmt_map *find_commit_format(const char *sought)
145 {
146 if (!commit_formats)
147 setup_commit_formats();
148
149 return find_commit_format_recursive(sought, sought, 0);
150 }
151
152 void get_commit_format(const char *arg, struct rev_info *rev)
153 {
154 struct cmt_fmt_map *commit_format;
155
156 rev->use_terminator = 0;
157 if (!arg) {
158 rev->commit_format = CMIT_FMT_DEFAULT;
159 return;
160 }
161 if (skip_prefix(arg, "format:", &arg)) {
162 save_user_format(rev, arg, 0);
163 return;
164 }
165
166 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
167 save_user_format(rev, arg, 1);
168 return;
169 }
170
171 commit_format = find_commit_format(arg);
172 if (!commit_format)
173 die("invalid --pretty format: %s", arg);
174
175 rev->commit_format = commit_format->format;
176 rev->use_terminator = commit_format->is_tformat;
177 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
178 if (commit_format->format == CMIT_FMT_USERFORMAT) {
179 save_user_format(rev, commit_format->user_format,
180 commit_format->is_tformat);
181 }
182 }
183
184 /*
185 * Generic support for pretty-printing the header
186 */
187 static int get_one_line(const char *msg)
188 {
189 int ret = 0;
190
191 for (;;) {
192 char c = *msg++;
193 if (!c)
194 break;
195 ret++;
196 if (c == '\n')
197 break;
198 }
199 return ret;
200 }
201
202 /* High bit set, or ISO-2022-INT */
203 static int non_ascii(int ch)
204 {
205 return !isascii(ch) || ch == '\033';
206 }
207
208 int has_non_ascii(const char *s)
209 {
210 int ch;
211 if (!s)
212 return 0;
213 while ((ch = *s++) != '\0') {
214 if (non_ascii(ch))
215 return 1;
216 }
217 return 0;
218 }
219
220 static int is_rfc822_special(char ch)
221 {
222 switch (ch) {
223 case '(':
224 case ')':
225 case '<':
226 case '>':
227 case '[':
228 case ']':
229 case ':':
230 case ';':
231 case '@':
232 case ',':
233 case '.':
234 case '"':
235 case '\\':
236 return 1;
237 default:
238 return 0;
239 }
240 }
241
242 static int needs_rfc822_quoting(const char *s, int len)
243 {
244 int i;
245 for (i = 0; i < len; i++)
246 if (is_rfc822_special(s[i]))
247 return 1;
248 return 0;
249 }
250
251 static int last_line_length(struct strbuf *sb)
252 {
253 int i;
254
255 /* How many bytes are already used on the last line? */
256 for (i = sb->len - 1; i >= 0; i--)
257 if (sb->buf[i] == '\n')
258 break;
259 return sb->len - (i + 1);
260 }
261
262 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
263 {
264 int i;
265
266 /* just a guess, we may have to also backslash-quote */
267 strbuf_grow(out, len + 2);
268
269 strbuf_addch(out, '"');
270 for (i = 0; i < len; i++) {
271 switch (s[i]) {
272 case '"':
273 case '\\':
274 strbuf_addch(out, '\\');
275 /* fall through */
276 default:
277 strbuf_addch(out, s[i]);
278 }
279 }
280 strbuf_addch(out, '"');
281 }
282
283 enum rfc2047_type {
284 RFC2047_SUBJECT,
285 RFC2047_ADDRESS
286 };
287
288 static int is_rfc2047_special(char ch, enum rfc2047_type type)
289 {
290 /*
291 * rfc2047, section 4.2:
292 *
293 * 8-bit values which correspond to printable ASCII characters other
294 * than "=", "?", and "_" (underscore), MAY be represented as those
295 * characters. (But see section 5 for restrictions.) In
296 * particular, SPACE and TAB MUST NOT be represented as themselves
297 * within encoded words.
298 */
299
300 /*
301 * rule out non-ASCII characters and non-printable characters (the
302 * non-ASCII check should be redundant as isprint() is not localized
303 * and only knows about ASCII, but be defensive about that)
304 */
305 if (non_ascii(ch) || !isprint(ch))
306 return 1;
307
308 /*
309 * rule out special printable characters (' ' should be the only
310 * whitespace character considered printable, but be defensive and use
311 * isspace())
312 */
313 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
314 return 1;
315
316 /*
317 * rfc2047, section 5.3:
318 *
319 * As a replacement for a 'word' entity within a 'phrase', for example,
320 * one that precedes an address in a From, To, or Cc header. The ABNF
321 * definition for 'phrase' from RFC 822 thus becomes:
322 *
323 * phrase = 1*( encoded-word / word )
324 *
325 * In this case the set of characters that may be used in a "Q"-encoded
326 * 'encoded-word' is restricted to: <upper and lower case ASCII
327 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
328 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
329 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
330 * 'special' by 'linear-white-space'.
331 */
332
333 if (type != RFC2047_ADDRESS)
334 return 0;
335
336 /* '=' and '_' are special cases and have been checked above */
337 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
338 }
339
340 static int needs_rfc2047_encoding(const char *line, int len,
341 enum rfc2047_type type)
342 {
343 int i;
344
345 for (i = 0; i < len; i++) {
346 int ch = line[i];
347 if (non_ascii(ch) || ch == '\n')
348 return 1;
349 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
350 return 1;
351 }
352
353 return 0;
354 }
355
356 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
357 const char *encoding, enum rfc2047_type type)
358 {
359 static const int max_encoded_length = 76; /* per rfc2047 */
360 int i;
361 int line_len = last_line_length(sb);
362
363 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
364 strbuf_addf(sb, "=?%s?q?", encoding);
365 line_len += strlen(encoding) + 5; /* 5 for =??q? */
366
367 while (len) {
368 /*
369 * RFC 2047, section 5 (3):
370 *
371 * Each 'encoded-word' MUST represent an integral number of
372 * characters. A multi-octet character may not be split across
373 * adjacent 'encoded- word's.
374 */
375 const unsigned char *p = (const unsigned char *)line;
376 int chrlen = mbs_chrlen(&line, &len, encoding);
377 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
378
379 /* "=%02X" * chrlen, or the byte itself */
380 const char *encoded_fmt = is_special ? "=%02X" : "%c";
381 int encoded_len = is_special ? 3 * chrlen : 1;
382
383 /*
384 * According to RFC 2047, we could encode the special character
385 * ' ' (space) with '_' (underscore) for readability. But many
386 * programs do not understand this and just leave the
387 * underscore in place. Thus, we do nothing special here, which
388 * causes ' ' to be encoded as '=20', avoiding this problem.
389 */
390
391 if (line_len + encoded_len + 2 > max_encoded_length) {
392 /* It won't fit with trailing "?=" --- break the line */
393 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
394 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
395 }
396
397 for (i = 0; i < chrlen; i++)
398 strbuf_addf(sb, encoded_fmt, p[i]);
399 line_len += encoded_len;
400 }
401 strbuf_addstr(sb, "?=");
402 }
403
404 const char *show_ident_date(const struct ident_split *ident,
405 const struct date_mode *mode)
406 {
407 unsigned long date = 0;
408 long tz = 0;
409
410 if (ident->date_begin && ident->date_end)
411 date = strtoul(ident->date_begin, NULL, 10);
412 if (date_overflows(date))
413 date = 0;
414 else {
415 if (ident->tz_begin && ident->tz_end)
416 tz = strtol(ident->tz_begin, NULL, 10);
417 if (tz >= INT_MAX || tz <= INT_MIN)
418 tz = 0;
419 }
420 return show_date(date, tz, mode);
421 }
422
423 void pp_user_info(struct pretty_print_context *pp,
424 const char *what, struct strbuf *sb,
425 const char *line, const char *encoding)
426 {
427 struct ident_split ident;
428 char *line_end;
429 const char *mailbuf, *namebuf;
430 size_t namelen, maillen;
431 int max_length = 78; /* per rfc2822 */
432
433 if (pp->fmt == CMIT_FMT_ONELINE)
434 return;
435
436 line_end = strchrnul(line, '\n');
437 if (split_ident_line(&ident, line, line_end - line))
438 return;
439
440 mailbuf = ident.mail_begin;
441 maillen = ident.mail_end - ident.mail_begin;
442 namebuf = ident.name_begin;
443 namelen = ident.name_end - ident.name_begin;
444
445 if (pp->mailmap)
446 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
447
448 if (cmit_fmt_is_mail(pp->fmt)) {
449 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
450 struct strbuf buf = STRBUF_INIT;
451
452 strbuf_addstr(&buf, "From: ");
453 strbuf_add(&buf, namebuf, namelen);
454 strbuf_addstr(&buf, " <");
455 strbuf_add(&buf, mailbuf, maillen);
456 strbuf_addstr(&buf, ">\n");
457 string_list_append(&pp->in_body_headers,
458 strbuf_detach(&buf, NULL));
459
460 mailbuf = pp->from_ident->mail_begin;
461 maillen = pp->from_ident->mail_end - mailbuf;
462 namebuf = pp->from_ident->name_begin;
463 namelen = pp->from_ident->name_end - namebuf;
464 }
465
466 strbuf_addstr(sb, "From: ");
467 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
468 add_rfc2047(sb, namebuf, namelen,
469 encoding, RFC2047_ADDRESS);
470 max_length = 76; /* per rfc2047 */
471 } else if (needs_rfc822_quoting(namebuf, namelen)) {
472 struct strbuf quoted = STRBUF_INIT;
473 add_rfc822_quoted(&quoted, namebuf, namelen);
474 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
475 -6, 1, max_length);
476 strbuf_release(&quoted);
477 } else {
478 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
479 -6, 1, max_length);
480 }
481
482 if (max_length <
483 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
484 strbuf_addch(sb, '\n');
485 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
486 } else {
487 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
488 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
489 (int)namelen, namebuf, (int)maillen, mailbuf);
490 }
491
492 switch (pp->fmt) {
493 case CMIT_FMT_MEDIUM:
494 strbuf_addf(sb, "Date: %s\n",
495 show_ident_date(&ident, &pp->date_mode));
496 break;
497 case CMIT_FMT_EMAIL:
498 case CMIT_FMT_MBOXRD:
499 strbuf_addf(sb, "Date: %s\n",
500 show_ident_date(&ident, DATE_MODE(RFC2822)));
501 break;
502 case CMIT_FMT_FULLER:
503 strbuf_addf(sb, "%sDate: %s\n", what,
504 show_ident_date(&ident, &pp->date_mode));
505 break;
506 default:
507 /* notin' */
508 break;
509 }
510 }
511
512 static int is_empty_line(const char *line, int *len_p)
513 {
514 int len = *len_p;
515 while (len && isspace(line[len - 1]))
516 len--;
517 *len_p = len;
518 return !len;
519 }
520
521 static const char *skip_empty_lines(const char *msg)
522 {
523 for (;;) {
524 int linelen = get_one_line(msg);
525 int ll = linelen;
526 if (!linelen)
527 break;
528 if (!is_empty_line(msg, &ll))
529 break;
530 msg += linelen;
531 }
532 return msg;
533 }
534
535 static void add_merge_info(const struct pretty_print_context *pp,
536 struct strbuf *sb, const struct commit *commit)
537 {
538 struct commit_list *parent = commit->parents;
539
540 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
541 !parent || !parent->next)
542 return;
543
544 strbuf_addstr(sb, "Merge:");
545
546 while (parent) {
547 struct commit *p = parent->item;
548 const char *hex = NULL;
549 if (pp->abbrev)
550 hex = find_unique_abbrev(p->object.oid.hash, pp->abbrev);
551 if (!hex)
552 hex = oid_to_hex(&p->object.oid);
553 parent = parent->next;
554
555 strbuf_addf(sb, " %s", hex);
556 }
557 strbuf_addch(sb, '\n');
558 }
559
560 static char *get_header(const char *msg, const char *key)
561 {
562 size_t len;
563 const char *v = find_commit_header(msg, key, &len);
564 return v ? xmemdupz(v, len) : NULL;
565 }
566
567 static char *replace_encoding_header(char *buf, const char *encoding)
568 {
569 struct strbuf tmp = STRBUF_INIT;
570 size_t start, len;
571 char *cp = buf;
572
573 /* guess if there is an encoding header before a \n\n */
574 while (!starts_with(cp, "encoding ")) {
575 cp = strchr(cp, '\n');
576 if (!cp || *++cp == '\n')
577 return buf;
578 }
579 start = cp - buf;
580 cp = strchr(cp, '\n');
581 if (!cp)
582 return buf; /* should not happen but be defensive */
583 len = cp + 1 - (buf + start);
584
585 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
586 if (is_encoding_utf8(encoding)) {
587 /* we have re-coded to UTF-8; drop the header */
588 strbuf_remove(&tmp, start, len);
589 } else {
590 /* just replaces XXXX in 'encoding XXXX\n' */
591 strbuf_splice(&tmp, start + strlen("encoding "),
592 len - strlen("encoding \n"),
593 encoding, strlen(encoding));
594 }
595 return strbuf_detach(&tmp, NULL);
596 }
597
598 const char *logmsg_reencode(const struct commit *commit,
599 char **commit_encoding,
600 const char *output_encoding)
601 {
602 static const char *utf8 = "UTF-8";
603 const char *use_encoding;
604 char *encoding;
605 const char *msg = get_commit_buffer(commit, NULL);
606 char *out;
607
608 if (!output_encoding || !*output_encoding) {
609 if (commit_encoding)
610 *commit_encoding = get_header(msg, "encoding");
611 return msg;
612 }
613 encoding = get_header(msg, "encoding");
614 if (commit_encoding)
615 *commit_encoding = encoding;
616 use_encoding = encoding ? encoding : utf8;
617 if (same_encoding(use_encoding, output_encoding)) {
618 /*
619 * No encoding work to be done. If we have no encoding header
620 * at all, then there's nothing to do, and we can return the
621 * message verbatim (whether newly allocated or not).
622 */
623 if (!encoding)
624 return msg;
625
626 /*
627 * Otherwise, we still want to munge the encoding header in the
628 * result, which will be done by modifying the buffer. If we
629 * are using a fresh copy, we can reuse it. But if we are using
630 * the cached copy from get_commit_buffer, we need to duplicate it
631 * to avoid munging the cached copy.
632 */
633 if (msg == get_cached_commit_buffer(commit, NULL))
634 out = xstrdup(msg);
635 else
636 out = (char *)msg;
637 }
638 else {
639 /*
640 * There's actual encoding work to do. Do the reencoding, which
641 * still leaves the header to be replaced in the next step. At
642 * this point, we are done with msg. If we allocated a fresh
643 * copy, we can free it.
644 */
645 out = reencode_string(msg, output_encoding, use_encoding);
646 if (out)
647 unuse_commit_buffer(commit, msg);
648 }
649
650 /*
651 * This replacement actually consumes the buffer we hand it, so we do
652 * not have to worry about freeing the old "out" here.
653 */
654 if (out)
655 out = replace_encoding_header(out, output_encoding);
656
657 if (!commit_encoding)
658 free(encoding);
659 /*
660 * If the re-encoding failed, out might be NULL here; in that
661 * case we just return the commit message verbatim.
662 */
663 return out ? out : msg;
664 }
665
666 static int mailmap_name(const char **email, size_t *email_len,
667 const char **name, size_t *name_len)
668 {
669 static struct string_list *mail_map;
670 if (!mail_map) {
671 mail_map = xcalloc(1, sizeof(*mail_map));
672 read_mailmap(mail_map, NULL);
673 }
674 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
675 }
676
677 static size_t format_person_part(struct strbuf *sb, char part,
678 const char *msg, int len,
679 const struct date_mode *dmode)
680 {
681 /* currently all placeholders have same length */
682 const int placeholder_len = 2;
683 struct ident_split s;
684 const char *name, *mail;
685 size_t maillen, namelen;
686
687 if (split_ident_line(&s, msg, len) < 0)
688 goto skip;
689
690 name = s.name_begin;
691 namelen = s.name_end - s.name_begin;
692 mail = s.mail_begin;
693 maillen = s.mail_end - s.mail_begin;
694
695 if (part == 'N' || part == 'E') /* mailmap lookup */
696 mailmap_name(&mail, &maillen, &name, &namelen);
697 if (part == 'n' || part == 'N') { /* name */
698 strbuf_add(sb, name, namelen);
699 return placeholder_len;
700 }
701 if (part == 'e' || part == 'E') { /* email */
702 strbuf_add(sb, mail, maillen);
703 return placeholder_len;
704 }
705
706 if (!s.date_begin)
707 goto skip;
708
709 if (part == 't') { /* date, UNIX timestamp */
710 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
711 return placeholder_len;
712 }
713
714 switch (part) {
715 case 'd': /* date */
716 strbuf_addstr(sb, show_ident_date(&s, dmode));
717 return placeholder_len;
718 case 'D': /* date, RFC2822 style */
719 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
720 return placeholder_len;
721 case 'r': /* date, relative */
722 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
723 return placeholder_len;
724 case 'i': /* date, ISO 8601-like */
725 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
726 return placeholder_len;
727 case 'I': /* date, ISO 8601 strict */
728 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
729 return placeholder_len;
730 }
731
732 skip:
733 /*
734 * reading from either a bogus commit, or a reflog entry with
735 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
736 * to compute a valid return value.
737 */
738 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
739 || part == 'D' || part == 'r' || part == 'i')
740 return placeholder_len;
741
742 return 0; /* unknown placeholder */
743 }
744
745 struct chunk {
746 size_t off;
747 size_t len;
748 };
749
750 enum flush_type {
751 no_flush,
752 flush_right,
753 flush_left,
754 flush_left_and_steal,
755 flush_both
756 };
757
758 enum trunc_type {
759 trunc_none,
760 trunc_left,
761 trunc_middle,
762 trunc_right
763 };
764
765 struct format_commit_context {
766 const struct commit *commit;
767 const struct pretty_print_context *pretty_ctx;
768 unsigned commit_header_parsed:1;
769 unsigned commit_message_parsed:1;
770 struct signature_check signature_check;
771 enum flush_type flush_type;
772 enum trunc_type truncate;
773 const char *message;
774 char *commit_encoding;
775 size_t width, indent1, indent2;
776 int auto_color;
777 int padding;
778
779 /* These offsets are relative to the start of the commit message. */
780 struct chunk author;
781 struct chunk committer;
782 size_t message_off;
783 size_t subject_off;
784 size_t body_off;
785
786 /* The following ones are relative to the result struct strbuf. */
787 struct chunk abbrev_commit_hash;
788 struct chunk abbrev_tree_hash;
789 struct chunk abbrev_parent_hashes;
790 size_t wrap_start;
791 };
792
793 static int add_again(struct strbuf *sb, struct chunk *chunk)
794 {
795 if (chunk->len) {
796 strbuf_adddup(sb, chunk->off, chunk->len);
797 return 1;
798 }
799
800 /*
801 * We haven't seen this chunk before. Our caller is surely
802 * going to add it the hard way now. Remember the most likely
803 * start of the to-be-added chunk: the current end of the
804 * struct strbuf.
805 */
806 chunk->off = sb->len;
807 return 0;
808 }
809
810 static void parse_commit_header(struct format_commit_context *context)
811 {
812 const char *msg = context->message;
813 int i;
814
815 for (i = 0; msg[i]; i++) {
816 const char *name;
817 int eol;
818 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
819 ; /* do nothing */
820
821 if (i == eol) {
822 break;
823 } else if (skip_prefix(msg + i, "author ", &name)) {
824 context->author.off = name - msg;
825 context->author.len = msg + eol - name;
826 } else if (skip_prefix(msg + i, "committer ", &name)) {
827 context->committer.off = name - msg;
828 context->committer.len = msg + eol - name;
829 }
830 i = eol;
831 }
832 context->message_off = i;
833 context->commit_header_parsed = 1;
834 }
835
836 static int istitlechar(char c)
837 {
838 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
839 (c >= '0' && c <= '9') || c == '.' || c == '_';
840 }
841
842 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
843 {
844 size_t trimlen;
845 size_t start_len = sb->len;
846 int space = 2;
847
848 for (; *msg && *msg != '\n'; msg++) {
849 if (istitlechar(*msg)) {
850 if (space == 1)
851 strbuf_addch(sb, '-');
852 space = 0;
853 strbuf_addch(sb, *msg);
854 if (*msg == '.')
855 while (*(msg+1) == '.')
856 msg++;
857 } else
858 space |= 1;
859 }
860
861 /* trim any trailing '.' or '-' characters */
862 trimlen = 0;
863 while (sb->len - trimlen > start_len &&
864 (sb->buf[sb->len - 1 - trimlen] == '.'
865 || sb->buf[sb->len - 1 - trimlen] == '-'))
866 trimlen++;
867 strbuf_remove(sb, sb->len - trimlen, trimlen);
868 }
869
870 const char *format_subject(struct strbuf *sb, const char *msg,
871 const char *line_separator)
872 {
873 int first = 1;
874
875 for (;;) {
876 const char *line = msg;
877 int linelen = get_one_line(line);
878
879 msg += linelen;
880 if (!linelen || is_empty_line(line, &linelen))
881 break;
882
883 if (!sb)
884 continue;
885 strbuf_grow(sb, linelen + 2);
886 if (!first)
887 strbuf_addstr(sb, line_separator);
888 strbuf_add(sb, line, linelen);
889 first = 0;
890 }
891 return msg;
892 }
893
894 static void parse_commit_message(struct format_commit_context *c)
895 {
896 const char *msg = c->message + c->message_off;
897 const char *start = c->message;
898
899 msg = skip_empty_lines(msg);
900 c->subject_off = msg - start;
901
902 msg = format_subject(NULL, msg, NULL);
903 msg = skip_empty_lines(msg);
904 c->body_off = msg - start;
905
906 c->commit_message_parsed = 1;
907 }
908
909 static void strbuf_wrap(struct strbuf *sb, size_t pos,
910 size_t width, size_t indent1, size_t indent2)
911 {
912 struct strbuf tmp = STRBUF_INIT;
913
914 if (pos)
915 strbuf_add(&tmp, sb->buf, pos);
916 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
917 (int) indent1, (int) indent2, (int) width);
918 strbuf_swap(&tmp, sb);
919 strbuf_release(&tmp);
920 }
921
922 static void rewrap_message_tail(struct strbuf *sb,
923 struct format_commit_context *c,
924 size_t new_width, size_t new_indent1,
925 size_t new_indent2)
926 {
927 if (c->width == new_width && c->indent1 == new_indent1 &&
928 c->indent2 == new_indent2)
929 return;
930 if (c->wrap_start < sb->len)
931 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
932 c->wrap_start = sb->len;
933 c->width = new_width;
934 c->indent1 = new_indent1;
935 c->indent2 = new_indent2;
936 }
937
938 static int format_reflog_person(struct strbuf *sb,
939 char part,
940 struct reflog_walk_info *log,
941 const struct date_mode *dmode)
942 {
943 const char *ident;
944
945 if (!log)
946 return 2;
947
948 ident = get_reflog_ident(log);
949 if (!ident)
950 return 2;
951
952 return format_person_part(sb, part, ident, strlen(ident), dmode);
953 }
954
955 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
956 const char *placeholder,
957 struct format_commit_context *c)
958 {
959 const char *rest = placeholder;
960
961 if (placeholder[1] == '(') {
962 const char *begin = placeholder + 2;
963 const char *end = strchr(begin, ')');
964 char color[COLOR_MAXLEN];
965
966 if (!end)
967 return 0;
968 if (skip_prefix(begin, "auto,", &begin)) {
969 if (!want_color(c->pretty_ctx->color))
970 return end - placeholder + 1;
971 }
972 if (color_parse_mem(begin, end - begin, color) < 0)
973 die(_("unable to parse --pretty format"));
974 strbuf_addstr(sb, color);
975 return end - placeholder + 1;
976 }
977 if (skip_prefix(placeholder + 1, "red", &rest))
978 strbuf_addstr(sb, GIT_COLOR_RED);
979 else if (skip_prefix(placeholder + 1, "green", &rest))
980 strbuf_addstr(sb, GIT_COLOR_GREEN);
981 else if (skip_prefix(placeholder + 1, "blue", &rest))
982 strbuf_addstr(sb, GIT_COLOR_BLUE);
983 else if (skip_prefix(placeholder + 1, "reset", &rest))
984 strbuf_addstr(sb, GIT_COLOR_RESET);
985 return rest - placeholder;
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.oid.hash);
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, oid_to_hex(&commit->object.oid));
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.oid.hash,
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, oid_to_hex(&commit->tree->object.oid));
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.oid.hash,
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, oid_to_hex(&p->item->object.oid));
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.oid.hash,
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 *name, *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 (skip_prefix(line, "author ", &name)) {
1561 strbuf_grow(sb, linelen + 80);
1562 pp_user_info(pp, "Author", sb, name, encoding);
1563 }
1564 if (skip_prefix(line, "committer ", &name) &&
1565 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1566 strbuf_grow(sb, linelen + 80);
1567 pp_user_info(pp, "Commit", sb, name, 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 (cmit_fmt_is_mail(pp->fmt)) {
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 static int pp_utf8_width(const char *start, const char *end)
1637 {
1638 int width = 0;
1639 size_t remain = end - start;
1640
1641 while (remain) {
1642 int n = utf8_width(&start, &remain);
1643 if (n < 0 || !start)
1644 return -1;
1645 width += n;
1646 }
1647 return width;
1648 }
1649
1650 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1651 const char *line, int linelen)
1652 {
1653 const char *tab;
1654
1655 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1656 int width = pp_utf8_width(line, tab);
1657
1658 /*
1659 * If it wasn't well-formed utf8, or it
1660 * had characters with badly defined
1661 * width (control characters etc), just
1662 * give up on trying to align things.
1663 */
1664 if (width < 0)
1665 break;
1666
1667 /* Output the data .. */
1668 strbuf_add(sb, line, tab - line);
1669
1670 /* .. and the de-tabified tab */
1671 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1672
1673 /* Skip over the printed part .. */
1674 linelen -= tab + 1 - line;
1675 line = tab + 1;
1676 }
1677
1678 /*
1679 * Print out everything after the last tab without
1680 * worrying about width - there's nothing more to
1681 * align.
1682 */
1683 strbuf_add(sb, line, linelen);
1684 }
1685
1686 /*
1687 * pp_handle_indent() prints out the intendation, and
1688 * the whole line (without the final newline), after
1689 * de-tabifying.
1690 */
1691 static void pp_handle_indent(struct pretty_print_context *pp,
1692 struct strbuf *sb, int indent,
1693 const char *line, int linelen)
1694 {
1695 strbuf_addchars(sb, ' ', indent);
1696 if (pp->expand_tabs_in_log)
1697 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1698 else
1699 strbuf_add(sb, line, linelen);
1700 }
1701
1702 static int is_mboxrd_from(const char *line, int len)
1703 {
1704 /*
1705 * a line matching /^From $/ here would only have len == 4
1706 * at this point because is_empty_line would've trimmed all
1707 * trailing space
1708 */
1709 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1710 }
1711
1712 void pp_remainder(struct pretty_print_context *pp,
1713 const char **msg_p,
1714 struct strbuf *sb,
1715 int indent)
1716 {
1717 int first = 1;
1718 for (;;) {
1719 const char *line = *msg_p;
1720 int linelen = get_one_line(line);
1721 *msg_p += linelen;
1722
1723 if (!linelen)
1724 break;
1725
1726 if (is_empty_line(line, &linelen)) {
1727 if (first)
1728 continue;
1729 if (pp->fmt == CMIT_FMT_SHORT)
1730 break;
1731 }
1732 first = 0;
1733
1734 strbuf_grow(sb, linelen + indent + 20);
1735 if (indent)
1736 pp_handle_indent(pp, sb, indent, line, linelen);
1737 else if (pp->expand_tabs_in_log)
1738 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1739 line, linelen);
1740 else {
1741 if (pp->fmt == CMIT_FMT_MBOXRD &&
1742 is_mboxrd_from(line, linelen))
1743 strbuf_addch(sb, '>');
1744
1745 strbuf_add(sb, line, linelen);
1746 }
1747 strbuf_addch(sb, '\n');
1748 }
1749 }
1750
1751 void pretty_print_commit(struct pretty_print_context *pp,
1752 const struct commit *commit,
1753 struct strbuf *sb)
1754 {
1755 unsigned long beginning_of_body;
1756 int indent = 4;
1757 const char *msg;
1758 const char *reencoded;
1759 const char *encoding;
1760 int need_8bit_cte = pp->need_8bit_cte;
1761
1762 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1763 format_commit_message(commit, user_format, sb, pp);
1764 return;
1765 }
1766
1767 encoding = get_log_output_encoding();
1768 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1769
1770 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1771 indent = 0;
1772
1773 /*
1774 * We need to check and emit Content-type: to mark it
1775 * as 8-bit if we haven't done so.
1776 */
1777 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1778 int i, ch, in_body;
1779
1780 for (in_body = i = 0; (ch = msg[i]); i++) {
1781 if (!in_body) {
1782 /* author could be non 7-bit ASCII but
1783 * the log may be so; skip over the
1784 * header part first.
1785 */
1786 if (ch == '\n' && msg[i+1] == '\n')
1787 in_body = 1;
1788 }
1789 else if (non_ascii(ch)) {
1790 need_8bit_cte = 1;
1791 break;
1792 }
1793 }
1794 }
1795
1796 pp_header(pp, encoding, commit, &msg, sb);
1797 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1798 strbuf_addch(sb, '\n');
1799 }
1800
1801 /* Skip excess blank lines at the beginning of body, if any... */
1802 msg = skip_empty_lines(msg);
1803
1804 /* These formats treat the title line specially. */
1805 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1806 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1807
1808 beginning_of_body = sb->len;
1809 if (pp->fmt != CMIT_FMT_ONELINE)
1810 pp_remainder(pp, &msg, sb, indent);
1811 strbuf_rtrim(sb);
1812
1813 /* Make sure there is an EOLN for the non-oneline case */
1814 if (pp->fmt != CMIT_FMT_ONELINE)
1815 strbuf_addch(sb, '\n');
1816
1817 /*
1818 * The caller may append additional body text in e-mail
1819 * format. Make sure we did not strip the blank line
1820 * between the header and the body.
1821 */
1822 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1823 strbuf_addch(sb, '\n');
1824
1825 unuse_commit_buffer(commit, reencoded);
1826 }
1827
1828 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1829 struct strbuf *sb)
1830 {
1831 struct pretty_print_context pp = {0};
1832 pp.fmt = fmt;
1833 pretty_print_commit(&pp, commit, sb);
1834 }
Eric Wong's git-svn branch