?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'jk/daemon-tolower'
[git.git] / pretty.c
blob4f512876ceb20f538256bac30dd98288abe2d682
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 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
28 {
29 free(user_format);
30 user_format = xstrdup(cp);
31 if (is_tformat)
32 rev->use_terminator = 1;
33 rev->commit_format = CMIT_FMT_USERFORMAT;
34 }
35
36 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
37 {
38 struct cmt_fmt_map *commit_format = NULL;
39 const char *name;
40 const char *fmt;
41 int i;
42
43 if (!starts_with(var, "pretty."))
44 return 0;
45
46 name = var + strlen("pretty.");
47 for (i = 0; i < builtin_formats_len; i++) {
48 if (!strcmp(commit_formats[i].name, name))
49 return 0;
50 }
51
52 for (i = builtin_formats_len; i < commit_formats_len; i++) {
53 if (!strcmp(commit_formats[i].name, name)) {
54 commit_format = &commit_formats[i];
55 break;
56 }
57 }
58
59 if (!commit_format) {
60 ALLOC_GROW(commit_formats, commit_formats_len+1,
61 commit_formats_alloc);
62 commit_format = &commit_formats[commit_formats_len];
63 memset(commit_format, 0, sizeof(*commit_format));
64 commit_formats_len++;
65 }
66
67 commit_format->name = xstrdup(name);
68 commit_format->format = CMIT_FMT_USERFORMAT;
69 git_config_string(&fmt, var, value);
70 if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
71 commit_format->is_tformat = fmt[0] == 't';
72 fmt = strchr(fmt, ':') + 1;
73 } else if (strchr(fmt, '%'))
74 commit_format->is_tformat = 1;
75 else
76 commit_format->is_alias = 1;
77 commit_format->user_format = fmt;
78
79 return 0;
80 }
81
82 static void setup_commit_formats(void)
83 {
84 struct cmt_fmt_map builtin_formats[] = {
85 { "raw", CMIT_FMT_RAW, 0 },
86 { "medium", CMIT_FMT_MEDIUM, 0 },
87 { "short", CMIT_FMT_SHORT, 0 },
88 { "email", CMIT_FMT_EMAIL, 0 },
89 { "fuller", CMIT_FMT_FULLER, 0 },
90 { "full", CMIT_FMT_FULL, 0 },
91 { "oneline", CMIT_FMT_ONELINE, 1 }
92 };
93 commit_formats_len = ARRAY_SIZE(builtin_formats);
94 builtin_formats_len = commit_formats_len;
95 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
96 memcpy(commit_formats, builtin_formats,
97 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
98
99 git_config(git_pretty_formats_config, NULL);
100 }
101
102 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
103 const char *original,
104 int num_redirections)
105 {
106 struct cmt_fmt_map *found = NULL;
107 size_t found_match_len = 0;
108 int i;
109
110 if (num_redirections >= commit_formats_len)
111 die("invalid --pretty format: "
112 "'%s' references an alias which points to itself",
113 original);
114
115 for (i = 0; i < commit_formats_len; i++) {
116 size_t match_len;
117
118 if (!starts_with(commit_formats[i].name, sought))
119 continue;
120
121 match_len = strlen(commit_formats[i].name);
122 if (found == NULL || found_match_len > match_len) {
123 found = &commit_formats[i];
124 found_match_len = match_len;
125 }
126 }
127
128 if (found && found->is_alias) {
129 found = find_commit_format_recursive(found->user_format,
130 original,
131 num_redirections+1);
132 }
133
134 return found;
135 }
136
137 static struct cmt_fmt_map *find_commit_format(const char *sought)
138 {
139 if (!commit_formats)
140 setup_commit_formats();
141
142 return find_commit_format_recursive(sought, sought, 0);
143 }
144
145 void get_commit_format(const char *arg, struct rev_info *rev)
146 {
147 struct cmt_fmt_map *commit_format;
148
149 rev->use_terminator = 0;
150 if (!arg || !*arg) {
151 rev->commit_format = CMIT_FMT_DEFAULT;
152 return;
153 }
154 if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
155 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
156 return;
157 }
158
159 if (strchr(arg, '%')) {
160 save_user_format(rev, arg, 1);
161 return;
162 }
163
164 commit_format = find_commit_format(arg);
165 if (!commit_format)
166 die("invalid --pretty format: %s", arg);
167
168 rev->commit_format = commit_format->format;
169 rev->use_terminator = commit_format->is_tformat;
170 if (commit_format->format == CMIT_FMT_USERFORMAT) {
171 save_user_format(rev, commit_format->user_format,
172 commit_format->is_tformat);
173 }
174 }
175
176 /*
177 * Generic support for pretty-printing the header
178 */
179 static int get_one_line(const char *msg)
180 {
181 int ret = 0;
182
183 for (;;) {
184 char c = *msg++;
185 if (!c)
186 break;
187 ret++;
188 if (c == '\n')
189 break;
190 }
191 return ret;
192 }
193
194 /* High bit set, or ISO-2022-INT */
195 static int non_ascii(int ch)
196 {
197 return !isascii(ch) || ch == '\033';
198 }
199
200 int has_non_ascii(const char *s)
201 {
202 int ch;
203 if (!s)
204 return 0;
205 while ((ch = *s++) != '\0') {
206 if (non_ascii(ch))
207 return 1;
208 }
209 return 0;
210 }
211
212 static int is_rfc822_special(char ch)
213 {
214 switch (ch) {
215 case '(':
216 case ')':
217 case '<':
218 case '>':
219 case '[':
220 case ']':
221 case ':':
222 case ';':
223 case '@':
224 case ',':
225 case '.':
226 case '"':
227 case '\\':
228 return 1;
229 default:
230 return 0;
231 }
232 }
233
234 static int needs_rfc822_quoting(const char *s, int len)
235 {
236 int i;
237 for (i = 0; i < len; i++)
238 if (is_rfc822_special(s[i]))
239 return 1;
240 return 0;
241 }
242
243 static int last_line_length(struct strbuf *sb)
244 {
245 int i;
246
247 /* How many bytes are already used on the last line? */
248 for (i = sb->len - 1; i >= 0; i--)
249 if (sb->buf[i] == '\n')
250 break;
251 return sb->len - (i + 1);
252 }
253
254 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
255 {
256 int i;
257
258 /* just a guess, we may have to also backslash-quote */
259 strbuf_grow(out, len + 2);
260
261 strbuf_addch(out, '"');
262 for (i = 0; i < len; i++) {
263 switch (s[i]) {
264 case '"':
265 case '\\':
266 strbuf_addch(out, '\\');
267 /* fall through */
268 default:
269 strbuf_addch(out, s[i]);
270 }
271 }
272 strbuf_addch(out, '"');
273 }
274
275 enum rfc2047_type {
276 RFC2047_SUBJECT,
277 RFC2047_ADDRESS,
278 };
279
280 static int is_rfc2047_special(char ch, enum rfc2047_type type)
281 {
282 /*
283 * rfc2047, section 4.2:
284 *
285 * 8-bit values which correspond to printable ASCII characters other
286 * than "=", "?", and "_" (underscore), MAY be represented as those
287 * characters. (But see section 5 for restrictions.) In
288 * particular, SPACE and TAB MUST NOT be represented as themselves
289 * within encoded words.
290 */
291
292 /*
293 * rule out non-ASCII characters and non-printable characters (the
294 * non-ASCII check should be redundant as isprint() is not localized
295 * and only knows about ASCII, but be defensive about that)
296 */
297 if (non_ascii(ch) || !isprint(ch))
298 return 1;
299
300 /*
301 * rule out special printable characters (' ' should be the only
302 * whitespace character considered printable, but be defensive and use
303 * isspace())
304 */
305 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
306 return 1;
307
308 /*
309 * rfc2047, section 5.3:
310 *
311 * As a replacement for a 'word' entity within a 'phrase', for example,
312 * one that precedes an address in a From, To, or Cc header. The ABNF
313 * definition for 'phrase' from RFC 822 thus becomes:
314 *
315 * phrase = 1*( encoded-word / word )
316 *
317 * In this case the set of characters that may be used in a "Q"-encoded
318 * 'encoded-word' is restricted to: <upper and lower case ASCII
319 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
320 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
321 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
322 * 'special' by 'linear-white-space'.
323 */
324
325 if (type != RFC2047_ADDRESS)
326 return 0;
327
328 /* '=' and '_' are special cases and have been checked above */
329 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
330 }
331
332 static int needs_rfc2047_encoding(const char *line, int len,
333 enum rfc2047_type type)
334 {
335 int i;
336
337 for (i = 0; i < len; i++) {
338 int ch = line[i];
339 if (non_ascii(ch) || ch == '\n')
340 return 1;
341 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
342 return 1;
343 }
344
345 return 0;
346 }
347
348 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
349 const char *encoding, enum rfc2047_type type)
350 {
351 static const int max_encoded_length = 76; /* per rfc2047 */
352 int i;
353 int line_len = last_line_length(sb);
354
355 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
356 strbuf_addf(sb, "=?%s?q?", encoding);
357 line_len += strlen(encoding) + 5; /* 5 for =??q? */
358
359 while (len) {
360 /*
361 * RFC 2047, section 5 (3):
362 *
363 * Each 'encoded-word' MUST represent an integral number of
364 * characters. A multi-octet character may not be split across
365 * adjacent 'encoded- word's.
366 */
367 const unsigned char *p = (const unsigned char *)line;
368 int chrlen = mbs_chrlen(&line, &len, encoding);
369 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
370
371 /* "=%02X" * chrlen, or the byte itself */
372 const char *encoded_fmt = is_special ? "=%02X" : "%c";
373 int encoded_len = is_special ? 3 * chrlen : 1;
374
375 /*
376 * According to RFC 2047, we could encode the special character
377 * ' ' (space) with '_' (underscore) for readability. But many
378 * programs do not understand this and just leave the
379 * underscore in place. Thus, we do nothing special here, which
380 * causes ' ' to be encoded as '=20', avoiding this problem.
381 */
382
383 if (line_len + encoded_len + 2 > max_encoded_length) {
384 /* It won't fit with trailing "?=" --- break the line */
385 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
386 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
387 }
388
389 for (i = 0; i < chrlen; i++)
390 strbuf_addf(sb, encoded_fmt, p[i]);
391 line_len += encoded_len;
392 }
393 strbuf_addstr(sb, "?=");
394 }
395
396 const char *show_ident_date(const struct ident_split *ident,
397 enum date_mode mode)
398 {
399 unsigned long date = 0;
400 long tz = 0;
401
402 if (ident->date_begin && ident->date_end)
403 date = strtoul(ident->date_begin, NULL, 10);
404 if (date_overflows(date))
405 date = 0;
406 else {
407 if (ident->tz_begin && ident->tz_end)
408 tz = strtol(ident->tz_begin, NULL, 10);
409 if (tz >= INT_MAX || tz <= INT_MIN)
410 tz = 0;
411 }
412 return show_date(date, tz, mode);
413 }
414
415 void pp_user_info(struct pretty_print_context *pp,
416 const char *what, struct strbuf *sb,
417 const char *line, const char *encoding)
418 {
419 struct ident_split ident;
420 char *line_end;
421 const char *mailbuf, *namebuf;
422 size_t namelen, maillen;
423 int max_length = 78; /* per rfc2822 */
424
425 if (pp->fmt == CMIT_FMT_ONELINE)
426 return;
427
428 line_end = strchrnul(line, '\n');
429 if (split_ident_line(&ident, line, line_end - line))
430 return;
431
432 mailbuf = ident.mail_begin;
433 maillen = ident.mail_end - ident.mail_begin;
434 namebuf = ident.name_begin;
435 namelen = ident.name_end - ident.name_begin;
436
437 if (pp->mailmap)
438 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
439
440 if (pp->fmt == CMIT_FMT_EMAIL) {
441 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
442 struct strbuf buf = STRBUF_INIT;
443
444 strbuf_addstr(&buf, "From: ");
445 strbuf_add(&buf, namebuf, namelen);
446 strbuf_addstr(&buf, " <");
447 strbuf_add(&buf, mailbuf, maillen);
448 strbuf_addstr(&buf, ">\n");
449 string_list_append(&pp->in_body_headers,
450 strbuf_detach(&buf, NULL));
451
452 mailbuf = pp->from_ident->mail_begin;
453 maillen = pp->from_ident->mail_end - mailbuf;
454 namebuf = pp->from_ident->name_begin;
455 namelen = pp->from_ident->name_end - namebuf;
456 }
457
458 strbuf_addstr(sb, "From: ");
459 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
460 add_rfc2047(sb, namebuf, namelen,
461 encoding, RFC2047_ADDRESS);
462 max_length = 76; /* per rfc2047 */
463 } else if (needs_rfc822_quoting(namebuf, namelen)) {
464 struct strbuf quoted = STRBUF_INIT;
465 add_rfc822_quoted(&quoted, namebuf, namelen);
466 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
467 -6, 1, max_length);
468 strbuf_release(&quoted);
469 } else {
470 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
471 -6, 1, max_length);
472 }
473
474 if (max_length <
475 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
476 strbuf_addch(sb, '\n');
477 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
478 } else {
479 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
480 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
481 (int)namelen, namebuf, (int)maillen, mailbuf);
482 }
483
484 switch (pp->fmt) {
485 case CMIT_FMT_MEDIUM:
486 strbuf_addf(sb, "Date: %s\n",
487 show_ident_date(&ident, pp->date_mode));
488 break;
489 case CMIT_FMT_EMAIL:
490 strbuf_addf(sb, "Date: %s\n",
491 show_ident_date(&ident, DATE_RFC2822));
492 break;
493 case CMIT_FMT_FULLER:
494 strbuf_addf(sb, "%sDate: %s\n", what,
495 show_ident_date(&ident, pp->date_mode));
496 break;
497 default:
498 /* notin' */
499 break;
500 }
501 }
502
503 static int is_empty_line(const char *line, int *len_p)
504 {
505 int len = *len_p;
506 while (len && isspace(line[len - 1]))
507 len--;
508 *len_p = len;
509 return !len;
510 }
511
512 static const char *skip_empty_lines(const char *msg)
513 {
514 for (;;) {
515 int linelen = get_one_line(msg);
516 int ll = linelen;
517 if (!linelen)
518 break;
519 if (!is_empty_line(msg, &ll))
520 break;
521 msg += linelen;
522 }
523 return msg;
524 }
525
526 static void add_merge_info(const struct pretty_print_context *pp,
527 struct strbuf *sb, const struct commit *commit)
528 {
529 struct commit_list *parent = commit->parents;
530
531 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
532 !parent || !parent->next)
533 return;
534
535 strbuf_addstr(sb, "Merge:");
536
537 while (parent) {
538 struct commit *p = parent->item;
539 const char *hex = NULL;
540 if (pp->abbrev)
541 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
542 if (!hex)
543 hex = sha1_to_hex(p->object.sha1);
544 parent = parent->next;
545
546 strbuf_addf(sb, " %s", hex);
547 }
548 strbuf_addch(sb, '\n');
549 }
550
551 static char *get_header(const struct commit *commit, const char *msg,
552 const char *key)
553 {
554 int key_len = strlen(key);
555 const char *line = msg;
556
557 while (line) {
558 const char *eol = strchrnul(line, '\n'), *next;
559
560 if (line == eol)
561 return NULL;
562 if (!*eol) {
563 warning("malformed commit (header is missing newline): %s",
564 sha1_to_hex(commit->object.sha1));
565 next = NULL;
566 } else
567 next = eol + 1;
568 if (eol - line > key_len &&
569 !strncmp(line, key, key_len) &&
570 line[key_len] == ' ') {
571 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
572 }
573 line = next;
574 }
575 return NULL;
576 }
577
578 static char *replace_encoding_header(char *buf, const char *encoding)
579 {
580 struct strbuf tmp = STRBUF_INIT;
581 size_t start, len;
582 char *cp = buf;
583
584 /* guess if there is an encoding header before a \n\n */
585 while (strncmp(cp, "encoding ", strlen("encoding "))) {
586 cp = strchr(cp, '\n');
587 if (!cp || *++cp == '\n')
588 return buf;
589 }
590 start = cp - buf;
591 cp = strchr(cp, '\n');
592 if (!cp)
593 return buf; /* should not happen but be defensive */
594 len = cp + 1 - (buf + start);
595
596 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
597 if (is_encoding_utf8(encoding)) {
598 /* we have re-coded to UTF-8; drop the header */
599 strbuf_remove(&tmp, start, len);
600 } else {
601 /* just replaces XXXX in 'encoding XXXX\n' */
602 strbuf_splice(&tmp, start + strlen("encoding "),
603 len - strlen("encoding \n"),
604 encoding, strlen(encoding));
605 }
606 return strbuf_detach(&tmp, NULL);
607 }
608
609 char *logmsg_reencode(const struct commit *commit,
610 char **commit_encoding,
611 const char *output_encoding)
612 {
613 static const char *utf8 = "UTF-8";
614 const char *use_encoding;
615 char *encoding;
616 char *msg = commit->buffer;
617 char *out;
618
619 if (!msg) {
620 enum object_type type;
621 unsigned long size;
622
623 msg = read_sha1_file(commit->object.sha1, &type, &size);
624 if (!msg)
625 die("Cannot read commit object %s",
626 sha1_to_hex(commit->object.sha1));
627 if (type != OBJ_COMMIT)
628 die("Expected commit for '%s', got %s",
629 sha1_to_hex(commit->object.sha1), typename(type));
630 }
631
632 if (!output_encoding || !*output_encoding) {
633 if (commit_encoding)
634 *commit_encoding =
635 get_header(commit, msg, "encoding");
636 return msg;
637 }
638 encoding = get_header(commit, msg, "encoding");
639 if (commit_encoding)
640 *commit_encoding = encoding;
641 use_encoding = encoding ? encoding : utf8;
642 if (same_encoding(use_encoding, output_encoding)) {
643 /*
644 * No encoding work to be done. If we have no encoding header
645 * at all, then there's nothing to do, and we can return the
646 * message verbatim (whether newly allocated or not).
647 */
648 if (!encoding)
649 return msg;
650
651 /*
652 * Otherwise, we still want to munge the encoding header in the
653 * result, which will be done by modifying the buffer. If we
654 * are using a fresh copy, we can reuse it. But if we are using
655 * the cached copy from commit->buffer, we need to duplicate it
656 * to avoid munging commit->buffer.
657 */
658 out = msg;
659 if (out == commit->buffer)
660 out = xstrdup(out);
661 }
662 else {
663 /*
664 * There's actual encoding work to do. Do the reencoding, which
665 * still leaves the header to be replaced in the next step. At
666 * this point, we are done with msg. If we allocated a fresh
667 * copy, we can free it.
668 */
669 out = reencode_string(msg, output_encoding, use_encoding);
670 if (out && msg != commit->buffer)
671 free(msg);
672 }
673
674 /*
675 * This replacement actually consumes the buffer we hand it, so we do
676 * not have to worry about freeing the old "out" here.
677 */
678 if (out)
679 out = replace_encoding_header(out, output_encoding);
680
681 if (!commit_encoding)
682 free(encoding);
683 /*
684 * If the re-encoding failed, out might be NULL here; in that
685 * case we just return the commit message verbatim.
686 */
687 return out ? out : msg;
688 }
689
690 void logmsg_free(char *msg, const struct commit *commit)
691 {
692 if (msg != commit->buffer)
693 free(msg);
694 }
695
696 static int mailmap_name(const char **email, size_t *email_len,
697 const char **name, size_t *name_len)
698 {
699 static struct string_list *mail_map;
700 if (!mail_map) {
701 mail_map = xcalloc(1, sizeof(*mail_map));
702 read_mailmap(mail_map, NULL);
703 }
704 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
705 }
706
707 static size_t format_person_part(struct strbuf *sb, char part,
708 const char *msg, int len, enum date_mode dmode)
709 {
710 /* currently all placeholders have same length */
711 const int placeholder_len = 2;
712 struct ident_split s;
713 const char *name, *mail;
714 size_t maillen, namelen;
715
716 if (split_ident_line(&s, msg, len) < 0)
717 goto skip;
718
719 name = s.name_begin;
720 namelen = s.name_end - s.name_begin;
721 mail = s.mail_begin;
722 maillen = s.mail_end - s.mail_begin;
723
724 if (part == 'N' || part == 'E') /* mailmap lookup */
725 mailmap_name(&mail, &maillen, &name, &namelen);
726 if (part == 'n' || part == 'N') { /* name */
727 strbuf_add(sb, name, namelen);
728 return placeholder_len;
729 }
730 if (part == 'e' || part == 'E') { /* email */
731 strbuf_add(sb, mail, maillen);
732 return placeholder_len;
733 }
734
735 if (!s.date_begin)
736 goto skip;
737
738 if (part == 't') { /* date, UNIX timestamp */
739 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
740 return placeholder_len;
741 }
742
743 switch (part) {
744 case 'd': /* date */
745 strbuf_addstr(sb, show_ident_date(&s, dmode));
746 return placeholder_len;
747 case 'D': /* date, RFC2822 style */
748 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
749 return placeholder_len;
750 case 'r': /* date, relative */
751 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
752 return placeholder_len;
753 case 'i': /* date, ISO 8601 */
754 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
755 return placeholder_len;
756 }
757
758 skip:
759 /*
760 * reading from either a bogus commit, or a reflog entry with
761 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
762 * to compute a valid return value.
763 */
764 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
765 || part == 'D' || part == 'r' || part == 'i')
766 return placeholder_len;
767
768 return 0; /* unknown placeholder */
769 }
770
771 struct chunk {
772 size_t off;
773 size_t len;
774 };
775
776 enum flush_type {
777 no_flush,
778 flush_right,
779 flush_left,
780 flush_left_and_steal,
781 flush_both
782 };
783
784 enum trunc_type {
785 trunc_none,
786 trunc_left,
787 trunc_middle,
788 trunc_right
789 };
790
791 struct format_commit_context {
792 const struct commit *commit;
793 const struct pretty_print_context *pretty_ctx;
794 unsigned commit_header_parsed:1;
795 unsigned commit_message_parsed:1;
796 struct signature_check signature_check;
797 enum flush_type flush_type;
798 enum trunc_type truncate;
799 char *message;
800 char *commit_encoding;
801 size_t width, indent1, indent2;
802 int auto_color;
803 int padding;
804
805 /* These offsets are relative to the start of the commit message. */
806 struct chunk author;
807 struct chunk committer;
808 size_t message_off;
809 size_t subject_off;
810 size_t body_off;
811
812 /* The following ones are relative to the result struct strbuf. */
813 struct chunk abbrev_commit_hash;
814 struct chunk abbrev_tree_hash;
815 struct chunk abbrev_parent_hashes;
816 size_t wrap_start;
817 };
818
819 static int add_again(struct strbuf *sb, struct chunk *chunk)
820 {
821 if (chunk->len) {
822 strbuf_adddup(sb, chunk->off, chunk->len);
823 return 1;
824 }
825
826 /*
827 * We haven't seen this chunk before. Our caller is surely
828 * going to add it the hard way now. Remember the most likely
829 * start of the to-be-added chunk: the current end of the
830 * struct strbuf.
831 */
832 chunk->off = sb->len;
833 return 0;
834 }
835
836 static void parse_commit_header(struct format_commit_context *context)
837 {
838 const char *msg = context->message;
839 int i;
840
841 for (i = 0; msg[i]; i++) {
842 int eol;
843 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
844 ; /* do nothing */
845
846 if (i == eol) {
847 break;
848 } else if (starts_with(msg + i, "author ")) {
849 context->author.off = i + 7;
850 context->author.len = eol - i - 7;
851 } else if (starts_with(msg + i, "committer ")) {
852 context->committer.off = i + 10;
853 context->committer.len = eol - i - 10;
854 }
855 i = eol;
856 }
857 context->message_off = i;
858 context->commit_header_parsed = 1;
859 }
860
861 static int istitlechar(char c)
862 {
863 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
864 (c >= '0' && c <= '9') || c == '.' || c == '_';
865 }
866
867 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
868 {
869 size_t trimlen;
870 size_t start_len = sb->len;
871 int space = 2;
872
873 for (; *msg && *msg != '\n'; msg++) {
874 if (istitlechar(*msg)) {
875 if (space == 1)
876 strbuf_addch(sb, '-');
877 space = 0;
878 strbuf_addch(sb, *msg);
879 if (*msg == '.')
880 while (*(msg+1) == '.')
881 msg++;
882 } else
883 space |= 1;
884 }
885
886 /* trim any trailing '.' or '-' characters */
887 trimlen = 0;
888 while (sb->len - trimlen > start_len &&
889 (sb->buf[sb->len - 1 - trimlen] == '.'
890 || sb->buf[sb->len - 1 - trimlen] == '-'))
891 trimlen++;
892 strbuf_remove(sb, sb->len - trimlen, trimlen);
893 }
894
895 const char *format_subject(struct strbuf *sb, const char *msg,
896 const char *line_separator)
897 {
898 int first = 1;
899
900 for (;;) {
901 const char *line = msg;
902 int linelen = get_one_line(line);
903
904 msg += linelen;
905 if (!linelen || is_empty_line(line, &linelen))
906 break;
907
908 if (!sb)
909 continue;
910 strbuf_grow(sb, linelen + 2);
911 if (!first)
912 strbuf_addstr(sb, line_separator);
913 strbuf_add(sb, line, linelen);
914 first = 0;
915 }
916 return msg;
917 }
918
919 static void parse_commit_message(struct format_commit_context *c)
920 {
921 const char *msg = c->message + c->message_off;
922 const char *start = c->message;
923
924 msg = skip_empty_lines(msg);
925 c->subject_off = msg - start;
926
927 msg = format_subject(NULL, msg, NULL);
928 msg = skip_empty_lines(msg);
929 c->body_off = msg - start;
930
931 c->commit_message_parsed = 1;
932 }
933
934 static void strbuf_wrap(struct strbuf *sb, size_t pos,
935 size_t width, size_t indent1, size_t indent2)
936 {
937 struct strbuf tmp = STRBUF_INIT;
938
939 if (pos)
940 strbuf_add(&tmp, sb->buf, pos);
941 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
942 (int) indent1, (int) indent2, (int) width);
943 strbuf_swap(&tmp, sb);
944 strbuf_release(&tmp);
945 }
946
947 static void rewrap_message_tail(struct strbuf *sb,
948 struct format_commit_context *c,
949 size_t new_width, size_t new_indent1,
950 size_t new_indent2)
951 {
952 if (c->width == new_width && c->indent1 == new_indent1 &&
953 c->indent2 == new_indent2)
954 return;
955 if (c->wrap_start < sb->len)
956 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
957 c->wrap_start = sb->len;
958 c->width = new_width;
959 c->indent1 = new_indent1;
960 c->indent2 = new_indent2;
961 }
962
963 static int format_reflog_person(struct strbuf *sb,
964 char part,
965 struct reflog_walk_info *log,
966 enum date_mode dmode)
967 {
968 const char *ident;
969
970 if (!log)
971 return 2;
972
973 ident = get_reflog_ident(log);
974 if (!ident)
975 return 2;
976
977 return format_person_part(sb, part, ident, strlen(ident), dmode);
978 }
979
980 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
981 const char *placeholder,
982 struct format_commit_context *c)
983 {
984 if (placeholder[1] == '(') {
985 const char *begin = placeholder + 2;
986 const char *end = strchr(begin, ')');
987 char color[COLOR_MAXLEN];
988
989 if (!end)
990 return 0;
991 if (starts_with(begin, "auto,")) {
992 if (!want_color(c->pretty_ctx->color))
993 return end - placeholder + 1;
994 begin += 5;
995 }
996 color_parse_mem(begin,
997 end - begin,
998 "--pretty format", color);
999 strbuf_addstr(sb, color);
1000 return end - placeholder + 1;
1001 }
1002 if (starts_with(placeholder + 1, "red")) {
1003 strbuf_addstr(sb, GIT_COLOR_RED);
1004 return 4;
1005 } else if (starts_with(placeholder + 1, "green")) {
1006 strbuf_addstr(sb, GIT_COLOR_GREEN);
1007 return 6;
1008 } else if (starts_with(placeholder + 1, "blue")) {
1009 strbuf_addstr(sb, GIT_COLOR_BLUE);
1010 return 5;
1011 } else if (starts_with(placeholder + 1, "reset")) {
1012 strbuf_addstr(sb, GIT_COLOR_RESET);
1013 return 6;
1014 } else
1015 return 0;
1016 }
1017
1018 static size_t parse_padding_placeholder(struct strbuf *sb,
1019 const char *placeholder,
1020 struct format_commit_context *c)
1021 {
1022 const char *ch = placeholder;
1023 enum flush_type flush_type;
1024 int to_column = 0;
1025
1026 switch (*ch++) {
1027 case '<':
1028 flush_type = flush_right;
1029 break;
1030 case '>':
1031 if (*ch == '<') {
1032 flush_type = flush_both;
1033 ch++;
1034 } else if (*ch == '>') {
1035 flush_type = flush_left_and_steal;
1036 ch++;
1037 } else
1038 flush_type = flush_left;
1039 break;
1040 default:
1041 return 0;
1042 }
1043
1044 /* the next value means "wide enough to that column" */
1045 if (*ch == '|') {
1046 to_column = 1;
1047 ch++;
1048 }
1049
1050 if (*ch == '(') {
1051 const char *start = ch + 1;
1052 const char *end = start + strcspn(start, ",)");
1053 char *next;
1054 int width;
1055 if (!end || end == start)
1056 return 0;
1057 width = strtoul(start, &next, 10);
1058 if (next == start || width == 0)
1059 return 0;
1060 c->padding = to_column ? -width : width;
1061 c->flush_type = flush_type;
1062
1063 if (*end == ',') {
1064 start = end + 1;
1065 end = strchr(start, ')');
1066 if (!end || end == start)
1067 return 0;
1068 if (starts_with(start, "trunc)"))
1069 c->truncate = trunc_right;
1070 else if (starts_with(start, "ltrunc)"))
1071 c->truncate = trunc_left;
1072 else if (starts_with(start, "mtrunc)"))
1073 c->truncate = trunc_middle;
1074 else
1075 return 0;
1076 } else
1077 c->truncate = trunc_none;
1078
1079 return end - placeholder + 1;
1080 }
1081 return 0;
1082 }
1083
1084 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1085 const char *placeholder,
1086 void *context)
1087 {
1088 struct format_commit_context *c = context;
1089 const struct commit *commit = c->commit;
1090 const char *msg = c->message;
1091 struct commit_list *p;
1092 int h1, h2;
1093
1094 /* these are independent of the commit */
1095 switch (placeholder[0]) {
1096 case 'C':
1097 if (starts_with(placeholder + 1, "(auto)")) {
1098 c->auto_color = 1;
1099 return 7; /* consumed 7 bytes, "C(auto)" */
1100 } else {
1101 int ret = parse_color(sb, placeholder, c);
1102 if (ret)
1103 c->auto_color = 0;
1104 /*
1105 * Otherwise, we decided to treat %C<unknown>
1106 * as a literal string, and the previous
1107 * %C(auto) is still valid.
1108 */
1109 return ret;
1110 }
1111 case 'n': /* newline */
1112 strbuf_addch(sb, '\n');
1113 return 1;
1114 case 'x':
1115 /* %x00 == NUL, %x0a == LF, etc. */
1116 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1117 h1 <= 16 &&
1118 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1119 h2 <= 16) {
1120 strbuf_addch(sb, (h1<<4)|h2);
1121 return 3;
1122 } else
1123 return 0;
1124 case 'w':
1125 if (placeholder[1] == '(') {
1126 unsigned long width = 0, indent1 = 0, indent2 = 0;
1127 char *next;
1128 const char *start = placeholder + 2;
1129 const char *end = strchr(start, ')');
1130 if (!end)
1131 return 0;
1132 if (end > start) {
1133 width = strtoul(start, &next, 10);
1134 if (*next == ',') {
1135 indent1 = strtoul(next + 1, &next, 10);
1136 if (*next == ',') {
1137 indent2 = strtoul(next + 1,
1138 &next, 10);
1139 }
1140 }
1141 if (*next != ')')
1142 return 0;
1143 }
1144 rewrap_message_tail(sb, c, width, indent1, indent2);
1145 return end - placeholder + 1;
1146 } else
1147 return 0;
1148
1149 case '<':
1150 case '>':
1151 return parse_padding_placeholder(sb, placeholder, c);
1152 }
1153
1154 /* these depend on the commit */
1155 if (!commit->object.parsed)
1156 parse_object(commit->object.sha1);
1157
1158 switch (placeholder[0]) {
1159 case 'H': /* commit hash */
1160 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1161 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1162 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1163 return 1;
1164 case 'h': /* abbreviated commit hash */
1165 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1166 if (add_again(sb, &c->abbrev_commit_hash)) {
1167 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1168 return 1;
1169 }
1170 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1171 c->pretty_ctx->abbrev));
1172 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1173 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1174 return 1;
1175 case 'T': /* tree hash */
1176 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1177 return 1;
1178 case 't': /* abbreviated tree hash */
1179 if (add_again(sb, &c->abbrev_tree_hash))
1180 return 1;
1181 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1182 c->pretty_ctx->abbrev));
1183 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1184 return 1;
1185 case 'P': /* parent hashes */
1186 for (p = commit->parents; p; p = p->next) {
1187 if (p != commit->parents)
1188 strbuf_addch(sb, ' ');
1189 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1190 }
1191 return 1;
1192 case 'p': /* abbreviated parent hashes */
1193 if (add_again(sb, &c->abbrev_parent_hashes))
1194 return 1;
1195 for (p = commit->parents; p; p = p->next) {
1196 if (p != commit->parents)
1197 strbuf_addch(sb, ' ');
1198 strbuf_addstr(sb, find_unique_abbrev(
1199 p->item->object.sha1,
1200 c->pretty_ctx->abbrev));
1201 }
1202 c->abbrev_parent_hashes.len = sb->len -
1203 c->abbrev_parent_hashes.off;
1204 return 1;
1205 case 'm': /* left/right/bottom */
1206 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1207 return 1;
1208 case 'd':
1209 load_ref_decorations(DECORATE_SHORT_REFS);
1210 format_decorations(sb, commit, c->auto_color);
1211 return 1;
1212 case 'g': /* reflog info */
1213 switch(placeholder[1]) {
1214 case 'd': /* reflog selector */
1215 case 'D':
1216 if (c->pretty_ctx->reflog_info)
1217 get_reflog_selector(sb,
1218 c->pretty_ctx->reflog_info,
1219 c->pretty_ctx->date_mode,
1220 c->pretty_ctx->date_mode_explicit,
1221 (placeholder[1] == 'd'));
1222 return 2;
1223 case 's': /* reflog message */
1224 if (c->pretty_ctx->reflog_info)
1225 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1226 return 2;
1227 case 'n':
1228 case 'N':
1229 case 'e':
1230 case 'E':
1231 return format_reflog_person(sb,
1232 placeholder[1],
1233 c->pretty_ctx->reflog_info,
1234 c->pretty_ctx->date_mode);
1235 }
1236 return 0; /* unknown %g placeholder */
1237 case 'N':
1238 if (c->pretty_ctx->notes_message) {
1239 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1240 return 1;
1241 }
1242 return 0;
1243 }
1244
1245 if (placeholder[0] == 'G') {
1246 if (!c->signature_check.result)
1247 check_commit_signature(c->commit, &(c->signature_check));
1248 switch (placeholder[1]) {
1249 case 'G':
1250 if (c->signature_check.gpg_output)
1251 strbuf_addstr(sb, c->signature_check.gpg_output);
1252 break;
1253 case '?':
1254 switch (c->signature_check.result) {
1255 case 'G':
1256 case 'B':
1257 case 'U':
1258 case 'N':
1259 strbuf_addch(sb, c->signature_check.result);
1260 }
1261 break;
1262 case 'S':
1263 if (c->signature_check.signer)
1264 strbuf_addstr(sb, c->signature_check.signer);
1265 break;
1266 case 'K':
1267 if (c->signature_check.key)
1268 strbuf_addstr(sb, c->signature_check.key);
1269 break;
1270 }
1271 return 2;
1272 }
1273
1274
1275 /* For the rest we have to parse the commit header. */
1276 if (!c->commit_header_parsed)
1277 parse_commit_header(c);
1278
1279 switch (placeholder[0]) {
1280 case 'a': /* author ... */
1281 return format_person_part(sb, placeholder[1],
1282 msg + c->author.off, c->author.len,
1283 c->pretty_ctx->date_mode);
1284 case 'c': /* committer ... */
1285 return format_person_part(sb, placeholder[1],
1286 msg + c->committer.off, c->committer.len,
1287 c->pretty_ctx->date_mode);
1288 case 'e': /* encoding */
1289 if (c->commit_encoding)
1290 strbuf_addstr(sb, c->commit_encoding);
1291 return 1;
1292 case 'B': /* raw body */
1293 /* message_off is always left at the initial newline */
1294 strbuf_addstr(sb, msg + c->message_off + 1);
1295 return 1;
1296 }
1297
1298 /* Now we need to parse the commit message. */
1299 if (!c->commit_message_parsed)
1300 parse_commit_message(c);
1301
1302 switch (placeholder[0]) {
1303 case 's': /* subject */
1304 format_subject(sb, msg + c->subject_off, " ");
1305 return 1;
1306 case 'f': /* sanitized subject */
1307 format_sanitized_subject(sb, msg + c->subject_off);
1308 return 1;
1309 case 'b': /* body */
1310 strbuf_addstr(sb, msg + c->body_off);
1311 return 1;
1312 }
1313 return 0; /* unknown placeholder */
1314 }
1315
1316 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1317 const char *placeholder,
1318 struct format_commit_context *c)
1319 {
1320 struct strbuf local_sb = STRBUF_INIT;
1321 int total_consumed = 0, len, padding = c->padding;
1322 if (padding < 0) {
1323 const char *start = strrchr(sb->buf, '\n');
1324 int occupied;
1325 if (!start)
1326 start = sb->buf;
1327 occupied = utf8_strnwidth(start, -1, 1);
1328 padding = (-padding) - occupied;
1329 }
1330 while (1) {
1331 int modifier = *placeholder == 'C';
1332 int consumed = format_commit_one(&local_sb, placeholder, c);
1333 total_consumed += consumed;
1334
1335 if (!modifier)
1336 break;
1337
1338 placeholder += consumed;
1339 if (*placeholder != '%')
1340 break;
1341 placeholder++;
1342 total_consumed++;
1343 }
1344 len = utf8_strnwidth(local_sb.buf, -1, 1);
1345
1346 if (c->flush_type == flush_left_and_steal) {
1347 const char *ch = sb->buf + sb->len - 1;
1348 while (len > padding && ch > sb->buf) {
1349 const char *p;
1350 if (*ch == ' ') {
1351 ch--;
1352 padding++;
1353 continue;
1354 }
1355 /* check for trailing ansi sequences */
1356 if (*ch != 'm')
1357 break;
1358 p = ch - 1;
1359 while (ch - p < 10 && *p != '\033')
1360 p--;
1361 if (*p != '\033' ||
1362 ch + 1 - p != display_mode_esc_sequence_len(p))
1363 break;
1364 /*
1365 * got a good ansi sequence, put it back to
1366 * local_sb as we're cutting sb
1367 */
1368 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1369 ch = p - 1;
1370 }
1371 strbuf_setlen(sb, ch + 1 - sb->buf);
1372 c->flush_type = flush_left;
1373 }
1374
1375 if (len > padding) {
1376 switch (c->truncate) {
1377 case trunc_left:
1378 strbuf_utf8_replace(&local_sb,
1379 0, len - (padding - 2),
1380 "..");
1381 break;
1382 case trunc_middle:
1383 strbuf_utf8_replace(&local_sb,
1384 padding / 2 - 1,
1385 len - (padding - 2),
1386 "..");
1387 break;
1388 case trunc_right:
1389 strbuf_utf8_replace(&local_sb,
1390 padding - 2, len - (padding - 2),
1391 "..");
1392 break;
1393 case trunc_none:
1394 break;
1395 }
1396 strbuf_addstr(sb, local_sb.buf);
1397 } else {
1398 int sb_len = sb->len, offset = 0;
1399 if (c->flush_type == flush_left)
1400 offset = padding - len;
1401 else if (c->flush_type == flush_both)
1402 offset = (padding - len) / 2;
1403 /*
1404 * we calculate padding in columns, now
1405 * convert it back to chars
1406 */
1407 padding = padding - len + local_sb.len;
1408 strbuf_grow(sb, padding);
1409 strbuf_setlen(sb, sb_len + padding);
1410 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1411 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1412 local_sb.len);
1413 }
1414 strbuf_release(&local_sb);
1415 c->flush_type = no_flush;
1416 return total_consumed;
1417 }
1418
1419 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1420 const char *placeholder,
1421 void *context)
1422 {
1423 int consumed;
1424 size_t orig_len;
1425 enum {
1426 NO_MAGIC,
1427 ADD_LF_BEFORE_NON_EMPTY,
1428 DEL_LF_BEFORE_EMPTY,
1429 ADD_SP_BEFORE_NON_EMPTY
1430 } magic = NO_MAGIC;
1431
1432 switch (placeholder[0]) {
1433 case '-':
1434 magic = DEL_LF_BEFORE_EMPTY;
1435 break;
1436 case '+':
1437 magic = ADD_LF_BEFORE_NON_EMPTY;
1438 break;
1439 case ' ':
1440 magic = ADD_SP_BEFORE_NON_EMPTY;
1441 break;
1442 default:
1443 break;
1444 }
1445 if (magic != NO_MAGIC)
1446 placeholder++;
1447
1448 orig_len = sb->len;
1449 if (((struct format_commit_context *)context)->flush_type != no_flush)
1450 consumed = format_and_pad_commit(sb, placeholder, context);
1451 else
1452 consumed = format_commit_one(sb, placeholder, context);
1453 if (magic == NO_MAGIC)
1454 return consumed;
1455
1456 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1457 while (sb->len && sb->buf[sb->len - 1] == '\n')
1458 strbuf_setlen(sb, sb->len - 1);
1459 } else if (orig_len != sb->len) {
1460 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1461 strbuf_insert(sb, orig_len, "\n", 1);
1462 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1463 strbuf_insert(sb, orig_len, " ", 1);
1464 }
1465 return consumed + 1;
1466 }
1467
1468 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1469 void *context)
1470 {
1471 struct userformat_want *w = context;
1472
1473 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1474 placeholder++;
1475
1476 switch (*placeholder) {
1477 case 'N':
1478 w->notes = 1;
1479 break;
1480 }
1481 return 0;
1482 }
1483
1484 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1485 {
1486 struct strbuf dummy = STRBUF_INIT;
1487
1488 if (!fmt) {
1489 if (!user_format)
1490 return;
1491 fmt = user_format;
1492 }
1493 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1494 strbuf_release(&dummy);
1495 }
1496
1497 void format_commit_message(const struct commit *commit,
1498 const char *format, struct strbuf *sb,
1499 const struct pretty_print_context *pretty_ctx)
1500 {
1501 struct format_commit_context context;
1502 const char *output_enc = pretty_ctx->output_encoding;
1503 const char *utf8 = "UTF-8";
1504
1505 memset(&context, 0, sizeof(context));
1506 context.commit = commit;
1507 context.pretty_ctx = pretty_ctx;
1508 context.wrap_start = sb->len;
1509 /*
1510 * convert a commit message to UTF-8 first
1511 * as far as 'format_commit_item' assumes it in UTF-8
1512 */
1513 context.message = logmsg_reencode(commit,
1514 &context.commit_encoding,
1515 utf8);
1516
1517 strbuf_expand(sb, format, format_commit_item, &context);
1518 rewrap_message_tail(sb, &context, 0, 0, 0);
1519
1520 /* then convert a commit message to an actual output encoding */
1521 if (output_enc) {
1522 if (same_encoding(utf8, output_enc))
1523 output_enc = NULL;
1524 } else {
1525 if (context.commit_encoding &&
1526 !same_encoding(context.commit_encoding, utf8))
1527 output_enc = context.commit_encoding;
1528 }
1529
1530 if (output_enc) {
1531 int outsz;
1532 char *out = reencode_string_len(sb->buf, sb->len,
1533 output_enc, utf8, &outsz);
1534 if (out)
1535 strbuf_attach(sb, out, outsz, outsz + 1);
1536 }
1537
1538 free(context.commit_encoding);
1539 logmsg_free(context.message, commit);
1540 free(context.signature_check.gpg_output);
1541 free(context.signature_check.signer);
1542 }
1543
1544 static void pp_header(struct pretty_print_context *pp,
1545 const char *encoding,
1546 const struct commit *commit,
1547 const char **msg_p,
1548 struct strbuf *sb)
1549 {
1550 int parents_shown = 0;
1551
1552 for (;;) {
1553 const char *line = *msg_p;
1554 int linelen = get_one_line(*msg_p);
1555
1556 if (!linelen)
1557 return;
1558 *msg_p += linelen;
1559
1560 if (linelen == 1)
1561 /* End of header */
1562 return;
1563
1564 if (pp->fmt == CMIT_FMT_RAW) {
1565 strbuf_add(sb, line, linelen);
1566 continue;
1567 }
1568
1569 if (starts_with(line, "parent ")) {
1570 if (linelen != 48)
1571 die("bad parent line in commit");
1572 continue;
1573 }
1574
1575 if (!parents_shown) {
1576 struct commit_list *parent;
1577 int num;
1578 for (parent = commit->parents, num = 0;
1579 parent;
1580 parent = parent->next, num++)
1581 ;
1582 /* with enough slop */
1583 strbuf_grow(sb, num * 50 + 20);
1584 add_merge_info(pp, sb, commit);
1585 parents_shown = 1;
1586 }
1587
1588 /*
1589 * MEDIUM == DEFAULT shows only author with dates.
1590 * FULL shows both authors but not dates.
1591 * FULLER shows both authors and dates.
1592 */
1593 if (starts_with(line, "author ")) {
1594 strbuf_grow(sb, linelen + 80);
1595 pp_user_info(pp, "Author", sb, line + 7, encoding);
1596 }
1597 if (starts_with(line, "committer ") &&
1598 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1599 strbuf_grow(sb, linelen + 80);
1600 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1601 }
1602 }
1603 }
1604
1605 void pp_title_line(struct pretty_print_context *pp,
1606 const char **msg_p,
1607 struct strbuf *sb,
1608 const char *encoding,
1609 int need_8bit_cte)
1610 {
1611 static const int max_length = 78; /* per rfc2047 */
1612 struct strbuf title;
1613
1614 strbuf_init(&title, 80);
1615 *msg_p = format_subject(&title, *msg_p,
1616 pp->preserve_subject ? "\n" : " ");
1617
1618 strbuf_grow(sb, title.len + 1024);
1619 if (pp->subject) {
1620 strbuf_addstr(sb, pp->subject);
1621 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1622 add_rfc2047(sb, title.buf, title.len,
1623 encoding, RFC2047_SUBJECT);
1624 else
1625 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1626 -last_line_length(sb), 1, max_length);
1627 } else {
1628 strbuf_addbuf(sb, &title);
1629 }
1630 strbuf_addch(sb, '\n');
1631
1632 if (need_8bit_cte == 0) {
1633 int i;
1634 for (i = 0; i < pp->in_body_headers.nr; i++) {
1635 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1636 need_8bit_cte = 1;
1637 break;
1638 }
1639 }
1640 }
1641
1642 if (need_8bit_cte > 0) {
1643 const char *header_fmt =
1644 "MIME-Version: 1.0\n"
1645 "Content-Type: text/plain; charset=%s\n"
1646 "Content-Transfer-Encoding: 8bit\n";
1647 strbuf_addf(sb, header_fmt, encoding);
1648 }
1649 if (pp->after_subject) {
1650 strbuf_addstr(sb, pp->after_subject);
1651 }
1652 if (pp->fmt == CMIT_FMT_EMAIL) {
1653 strbuf_addch(sb, '\n');
1654 }
1655
1656 if (pp->in_body_headers.nr) {
1657 int i;
1658 for (i = 0; i < pp->in_body_headers.nr; i++) {
1659 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1660 free(pp->in_body_headers.items[i].string);
1661 }
1662 string_list_clear(&pp->in_body_headers, 0);
1663 strbuf_addch(sb, '\n');
1664 }
1665
1666 strbuf_release(&title);
1667 }
1668
1669 void pp_remainder(struct pretty_print_context *pp,
1670 const char **msg_p,
1671 struct strbuf *sb,
1672 int indent)
1673 {
1674 int first = 1;
1675 for (;;) {
1676 const char *line = *msg_p;
1677 int linelen = get_one_line(line);
1678 *msg_p += linelen;
1679
1680 if (!linelen)
1681 break;
1682
1683 if (is_empty_line(line, &linelen)) {
1684 if (first)
1685 continue;
1686 if (pp->fmt == CMIT_FMT_SHORT)
1687 break;
1688 }
1689 first = 0;
1690
1691 strbuf_grow(sb, linelen + indent + 20);
1692 if (indent) {
1693 memset(sb->buf + sb->len, ' ', indent);
1694 strbuf_setlen(sb, sb->len + indent);
1695 }
1696 strbuf_add(sb, line, linelen);
1697 strbuf_addch(sb, '\n');
1698 }
1699 }
1700
1701 void pretty_print_commit(struct pretty_print_context *pp,
1702 const struct commit *commit,
1703 struct strbuf *sb)
1704 {
1705 unsigned long beginning_of_body;
1706 int indent = 4;
1707 const char *msg;
1708 char *reencoded;
1709 const char *encoding;
1710 int need_8bit_cte = pp->need_8bit_cte;
1711
1712 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1713 format_commit_message(commit, user_format, sb, pp);
1714 return;
1715 }
1716
1717 encoding = get_log_output_encoding();
1718 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1719
1720 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1721 indent = 0;
1722
1723 /*
1724 * We need to check and emit Content-type: to mark it
1725 * as 8-bit if we haven't done so.
1726 */
1727 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1728 int i, ch, in_body;
1729
1730 for (in_body = i = 0; (ch = msg[i]); i++) {
1731 if (!in_body) {
1732 /* author could be non 7-bit ASCII but
1733 * the log may be so; skip over the
1734 * header part first.
1735 */
1736 if (ch == '\n' && msg[i+1] == '\n')
1737 in_body = 1;
1738 }
1739 else if (non_ascii(ch)) {
1740 need_8bit_cte = 1;
1741 break;
1742 }
1743 }
1744 }
1745
1746 pp_header(pp, encoding, commit, &msg, sb);
1747 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1748 strbuf_addch(sb, '\n');
1749 }
1750
1751 /* Skip excess blank lines at the beginning of body, if any... */
1752 msg = skip_empty_lines(msg);
1753
1754 /* These formats treat the title line specially. */
1755 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1756 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1757
1758 beginning_of_body = sb->len;
1759 if (pp->fmt != CMIT_FMT_ONELINE)
1760 pp_remainder(pp, &msg, sb, indent);
1761 strbuf_rtrim(sb);
1762
1763 /* Make sure there is an EOLN for the non-oneline case */
1764 if (pp->fmt != CMIT_FMT_ONELINE)
1765 strbuf_addch(sb, '\n');
1766
1767 /*
1768 * The caller may append additional body text in e-mail
1769 * format. Make sure we did not strip the blank line
1770 * between the header and the body.
1771 */
1772 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1773 strbuf_addch(sb, '\n');
1774
1775 logmsg_free(reencoded, commit);
1776 }
1777
1778 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1779 struct strbuf *sb)
1780 {
1781 struct pretty_print_context pp = {0};
1782 pp.fmt = fmt;
1783 pretty_print_commit(&pp, commit, sb);
1784 }
The Best VCS Around