Merge git://bogomips.org/git-svn
[git/jnareb-git.git] / pretty.c
blob01795de68704f9690aea2b9c36e557b31a151ca6
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"
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);
27 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
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;
36 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
38 struct cmt_fmt_map *commit_format = NULL;
39 const char *name;
40 const char *fmt;
41 int i;
43 if (prefixcmp(var, "pretty."))
44 return 0;
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;
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;
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++;
67 commit_format->name = xstrdup(name);
68 commit_format->format = CMIT_FMT_USERFORMAT;
69 git_config_string(&fmt, var, value);
70 if (!prefixcmp(fmt, "format:") || !prefixcmp(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;
79 return 0;
82 static void setup_commit_formats(void)
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 }
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));
99 git_config(git_pretty_formats_config, NULL);
102 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
103 const char *original,
104 int num_redirections)
106 struct cmt_fmt_map *found = NULL;
107 size_t found_match_len = 0;
108 int i;
110 if (num_redirections >= commit_formats_len)
111 die("invalid --pretty format: "
112 "'%s' references an alias which points to itself",
113 original);
115 for (i = 0; i < commit_formats_len; i++) {
116 size_t match_len;
118 if (prefixcmp(commit_formats[i].name, sought))
119 continue;
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;
128 if (found && found->is_alias) {
129 found = find_commit_format_recursive(found->user_format,
130 original,
131 num_redirections+1);
134 return found;
137 static struct cmt_fmt_map *find_commit_format(const char *sought)
139 if (!commit_formats)
140 setup_commit_formats();
142 return find_commit_format_recursive(sought, sought, 0);
145 void get_commit_format(const char *arg, struct rev_info *rev)
147 struct cmt_fmt_map *commit_format;
149 rev->use_terminator = 0;
150 if (!arg || !*arg) {
151 rev->commit_format = CMIT_FMT_DEFAULT;
152 return;
154 if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
155 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
156 return;
159 if (strchr(arg, '%')) {
160 save_user_format(rev, arg, 1);
161 return;
164 commit_format = find_commit_format(arg);
165 if (!commit_format)
166 die("invalid --pretty format: %s", arg);
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);
177 * Generic support for pretty-printing the header
179 static int get_one_line(const char *msg)
181 int ret = 0;
183 for (;;) {
184 char c = *msg++;
185 if (!c)
186 break;
187 ret++;
188 if (c == '\n')
189 break;
191 return ret;
194 /* High bit set, or ISO-2022-INT */
195 static int non_ascii(int ch)
197 return !isascii(ch) || ch == '\033';
200 int has_non_ascii(const char *s)
202 int ch;
203 if (!s)
204 return 0;
205 while ((ch = *s++) != '\0') {
206 if (non_ascii(ch))
207 return 1;
209 return 0;
212 static int is_rfc822_special(char ch)
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;
234 static int needs_rfc822_quoting(const char *s, int len)
236 int i;
237 for (i = 0; i < len; i++)
238 if (is_rfc822_special(s[i]))
239 return 1;
240 return 0;
243 static int last_line_length(struct strbuf *sb)
245 int i;
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);
254 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
256 int i;
258 /* just a guess, we may have to also backslash-quote */
259 strbuf_grow(out, len + 2);
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]);
272 strbuf_addch(out, '"');
275 enum rfc2047_type {
276 RFC2047_SUBJECT,
277 RFC2047_ADDRESS,
280 static int is_rfc2047_special(char ch, enum rfc2047_type type)
283 * rfc2047, section 4.2:
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.
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)
297 if (non_ascii(ch) || !isprint(ch))
298 return 1;
301 * rule out special printable characters (' ' should be the only
302 * whitespace character considered printable, but be defensive and use
303 * isspace())
305 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
306 return 1;
309 * rfc2047, section 5.3:
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:
315 * phrase = 1*( encoded-word / word )
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'.
325 if (type != RFC2047_ADDRESS)
326 return 0;
328 /* '=' and '_' are special cases and have been checked above */
329 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
332 static int needs_rfc2047_encoding(const char *line, int len,
333 enum rfc2047_type type)
335 int i;
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;
345 return 0;
348 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
349 const char *encoding, enum rfc2047_type type)
351 static const int max_encoded_length = 76; /* per rfc2047 */
352 int i;
353 int line_len = last_line_length(sb);
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 for (i = 0; i < len; i++) {
359 unsigned ch = line[i] & 0xFF;
360 int is_special = is_rfc2047_special(ch, type);
363 * According to RFC 2047, we could encode the special character
364 * ' ' (space) with '_' (underscore) for readability. But many
365 * programs do not understand this and just leave the
366 * underscore in place. Thus, we do nothing special here, which
367 * causes ' ' to be encoded as '=20', avoiding this problem.
370 if (line_len + 2 + (is_special ? 3 : 1) > max_encoded_length) {
371 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
372 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
375 if (is_special) {
376 strbuf_addf(sb, "=%02X", ch);
377 line_len += 3;
378 } else {
379 strbuf_addch(sb, ch);
380 line_len++;
383 strbuf_addstr(sb, "?=");
386 void pp_user_info(const struct pretty_print_context *pp,
387 const char *what, struct strbuf *sb,
388 const char *line, const char *encoding)
390 int max_length = 78; /* per rfc2822 */
391 char *date;
392 int namelen;
393 unsigned long time;
394 int tz;
396 if (pp->fmt == CMIT_FMT_ONELINE)
397 return;
398 date = strchr(line, '>');
399 if (!date)
400 return;
401 namelen = ++date - line;
402 time = strtoul(date, &date, 10);
403 tz = strtol(date, NULL, 10);
405 if (pp->fmt == CMIT_FMT_EMAIL) {
406 char *name_tail = strchr(line, '<');
407 int display_name_length;
408 if (!name_tail)
409 return;
410 while (line < name_tail && isspace(name_tail[-1]))
411 name_tail--;
412 display_name_length = name_tail - line;
413 strbuf_addstr(sb, "From: ");
414 if (needs_rfc2047_encoding(line, display_name_length, RFC2047_ADDRESS)) {
415 add_rfc2047(sb, line, display_name_length,
416 encoding, RFC2047_ADDRESS);
417 max_length = 76; /* per rfc2047 */
418 } else if (needs_rfc822_quoting(line, display_name_length)) {
419 struct strbuf quoted = STRBUF_INIT;
420 add_rfc822_quoted(&quoted, line, display_name_length);
421 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
422 -6, 1, max_length);
423 strbuf_release(&quoted);
424 } else {
425 strbuf_add_wrapped_bytes(sb, line, display_name_length,
426 -6, 1, max_length);
428 if (namelen - display_name_length + last_line_length(sb) > max_length) {
429 strbuf_addch(sb, '\n');
430 if (!isspace(name_tail[0]))
431 strbuf_addch(sb, ' ');
433 strbuf_add(sb, name_tail, namelen - display_name_length);
434 strbuf_addch(sb, '\n');
435 } else {
436 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
437 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
438 " ", namelen, line);
440 switch (pp->fmt) {
441 case CMIT_FMT_MEDIUM:
442 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, pp->date_mode));
443 break;
444 case CMIT_FMT_EMAIL:
445 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
446 break;
447 case CMIT_FMT_FULLER:
448 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
449 break;
450 default:
451 /* notin' */
452 break;
456 static int is_empty_line(const char *line, int *len_p)
458 int len = *len_p;
459 while (len && isspace(line[len-1]))
460 len--;
461 *len_p = len;
462 return !len;
465 static const char *skip_empty_lines(const char *msg)
467 for (;;) {
468 int linelen = get_one_line(msg);
469 int ll = linelen;
470 if (!linelen)
471 break;
472 if (!is_empty_line(msg, &ll))
473 break;
474 msg += linelen;
476 return msg;
479 static void add_merge_info(const struct pretty_print_context *pp,
480 struct strbuf *sb, const struct commit *commit)
482 struct commit_list *parent = commit->parents;
484 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
485 !parent || !parent->next)
486 return;
488 strbuf_addstr(sb, "Merge:");
490 while (parent) {
491 struct commit *p = parent->item;
492 const char *hex = NULL;
493 if (pp->abbrev)
494 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
495 if (!hex)
496 hex = sha1_to_hex(p->object.sha1);
497 parent = parent->next;
499 strbuf_addf(sb, " %s", hex);
501 strbuf_addch(sb, '\n');
504 static char *get_header(const struct commit *commit, const char *key)
506 int key_len = strlen(key);
507 const char *line = commit->buffer;
509 while (line) {
510 const char *eol = strchr(line, '\n'), *next;
512 if (line == eol)
513 return NULL;
514 if (!eol) {
515 warning("malformed commit (header is missing newline): %s",
516 sha1_to_hex(commit->object.sha1));
517 eol = line + strlen(line);
518 next = NULL;
519 } else
520 next = eol + 1;
521 if (eol - line > key_len &&
522 !strncmp(line, key, key_len) &&
523 line[key_len] == ' ') {
524 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
526 line = next;
528 return NULL;
531 static char *replace_encoding_header(char *buf, const char *encoding)
533 struct strbuf tmp = STRBUF_INIT;
534 size_t start, len;
535 char *cp = buf;
537 /* guess if there is an encoding header before a \n\n */
538 while (strncmp(cp, "encoding ", strlen("encoding "))) {
539 cp = strchr(cp, '\n');
540 if (!cp || *++cp == '\n')
541 return buf;
543 start = cp - buf;
544 cp = strchr(cp, '\n');
545 if (!cp)
546 return buf; /* should not happen but be defensive */
547 len = cp + 1 - (buf + start);
549 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
550 if (is_encoding_utf8(encoding)) {
551 /* we have re-coded to UTF-8; drop the header */
552 strbuf_remove(&tmp, start, len);
553 } else {
554 /* just replaces XXXX in 'encoding XXXX\n' */
555 strbuf_splice(&tmp, start + strlen("encoding "),
556 len - strlen("encoding \n"),
557 encoding, strlen(encoding));
559 return strbuf_detach(&tmp, NULL);
562 char *logmsg_reencode(const struct commit *commit,
563 const char *output_encoding)
565 static const char *utf8 = "UTF-8";
566 const char *use_encoding;
567 char *encoding;
568 char *out;
570 if (!output_encoding || !*output_encoding)
571 return NULL;
572 encoding = get_header(commit, "encoding");
573 use_encoding = encoding ? encoding : utf8;
574 if (same_encoding(use_encoding, output_encoding))
575 if (encoding) /* we'll strip encoding header later */
576 out = xstrdup(commit->buffer);
577 else
578 return NULL; /* nothing to do */
579 else
580 out = reencode_string(commit->buffer,
581 output_encoding, use_encoding);
582 if (out)
583 out = replace_encoding_header(out, output_encoding);
585 free(encoding);
586 return out;
589 static int mailmap_name(char *email, int email_len, char *name, int name_len)
591 static struct string_list *mail_map;
592 if (!mail_map) {
593 mail_map = xcalloc(1, sizeof(*mail_map));
594 read_mailmap(mail_map, NULL);
596 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
599 static size_t format_person_part(struct strbuf *sb, char part,
600 const char *msg, int len, enum date_mode dmode)
602 /* currently all placeholders have same length */
603 const int placeholder_len = 2;
604 int tz;
605 unsigned long date = 0;
606 char person_name[1024];
607 char person_mail[1024];
608 struct ident_split s;
609 const char *name_start, *name_end, *mail_start, *mail_end;
611 if (split_ident_line(&s, msg, len) < 0)
612 goto skip;
614 name_start = s.name_begin;
615 name_end = s.name_end;
616 mail_start = s.mail_begin;
617 mail_end = s.mail_end;
619 if (part == 'N' || part == 'E') { /* mailmap lookup */
620 snprintf(person_name, sizeof(person_name), "%.*s",
621 (int)(name_end - name_start), name_start);
622 snprintf(person_mail, sizeof(person_mail), "%.*s",
623 (int)(mail_end - mail_start), mail_start);
624 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
625 name_start = person_name;
626 name_end = name_start + strlen(person_name);
627 mail_start = person_mail;
628 mail_end = mail_start + strlen(person_mail);
630 if (part == 'n' || part == 'N') { /* name */
631 strbuf_add(sb, name_start, name_end-name_start);
632 return placeholder_len;
634 if (part == 'e' || part == 'E') { /* email */
635 strbuf_add(sb, mail_start, mail_end-mail_start);
636 return placeholder_len;
639 if (!s.date_begin)
640 goto skip;
642 date = strtoul(s.date_begin, NULL, 10);
644 if (part == 't') { /* date, UNIX timestamp */
645 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
646 return placeholder_len;
649 /* parse tz */
650 tz = strtoul(s.tz_begin + 1, NULL, 10);
651 if (*s.tz_begin == '-')
652 tz = -tz;
654 switch (part) {
655 case 'd': /* date */
656 strbuf_addstr(sb, show_date(date, tz, dmode));
657 return placeholder_len;
658 case 'D': /* date, RFC2822 style */
659 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
660 return placeholder_len;
661 case 'r': /* date, relative */
662 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
663 return placeholder_len;
664 case 'i': /* date, ISO 8601 */
665 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
666 return placeholder_len;
669 skip:
671 * reading from either a bogus commit, or a reflog entry with
672 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
673 * to compute a valid return value.
675 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
676 || part == 'D' || part == 'r' || part == 'i')
677 return placeholder_len;
679 return 0; /* unknown placeholder */
682 struct chunk {
683 size_t off;
684 size_t len;
687 struct format_commit_context {
688 const struct commit *commit;
689 const struct pretty_print_context *pretty_ctx;
690 unsigned commit_header_parsed:1;
691 unsigned commit_message_parsed:1;
692 unsigned commit_signature_parsed:1;
693 struct {
694 char *gpg_output;
695 char good_bad;
696 char *signer;
697 } signature;
698 char *message;
699 size_t width, indent1, indent2;
701 /* These offsets are relative to the start of the commit message. */
702 struct chunk author;
703 struct chunk committer;
704 struct chunk encoding;
705 size_t message_off;
706 size_t subject_off;
707 size_t body_off;
709 /* The following ones are relative to the result struct strbuf. */
710 struct chunk abbrev_commit_hash;
711 struct chunk abbrev_tree_hash;
712 struct chunk abbrev_parent_hashes;
713 size_t wrap_start;
716 static int add_again(struct strbuf *sb, struct chunk *chunk)
718 if (chunk->len) {
719 strbuf_adddup(sb, chunk->off, chunk->len);
720 return 1;
724 * We haven't seen this chunk before. Our caller is surely
725 * going to add it the hard way now. Remember the most likely
726 * start of the to-be-added chunk: the current end of the
727 * struct strbuf.
729 chunk->off = sb->len;
730 return 0;
733 static void parse_commit_header(struct format_commit_context *context)
735 const char *msg = context->message;
736 int i;
738 for (i = 0; msg[i]; i++) {
739 int eol;
740 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
741 ; /* do nothing */
743 if (i == eol) {
744 break;
745 } else if (!prefixcmp(msg + i, "author ")) {
746 context->author.off = i + 7;
747 context->author.len = eol - i - 7;
748 } else if (!prefixcmp(msg + i, "committer ")) {
749 context->committer.off = i + 10;
750 context->committer.len = eol - i - 10;
751 } else if (!prefixcmp(msg + i, "encoding ")) {
752 context->encoding.off = i + 9;
753 context->encoding.len = eol - i - 9;
755 i = eol;
757 context->message_off = i;
758 context->commit_header_parsed = 1;
761 static int istitlechar(char c)
763 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
764 (c >= '0' && c <= '9') || c == '.' || c == '_';
767 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
769 size_t trimlen;
770 size_t start_len = sb->len;
771 int space = 2;
773 for (; *msg && *msg != '\n'; msg++) {
774 if (istitlechar(*msg)) {
775 if (space == 1)
776 strbuf_addch(sb, '-');
777 space = 0;
778 strbuf_addch(sb, *msg);
779 if (*msg == '.')
780 while (*(msg+1) == '.')
781 msg++;
782 } else
783 space |= 1;
786 /* trim any trailing '.' or '-' characters */
787 trimlen = 0;
788 while (sb->len - trimlen > start_len &&
789 (sb->buf[sb->len - 1 - trimlen] == '.'
790 || sb->buf[sb->len - 1 - trimlen] == '-'))
791 trimlen++;
792 strbuf_remove(sb, sb->len - trimlen, trimlen);
795 const char *format_subject(struct strbuf *sb, const char *msg,
796 const char *line_separator)
798 int first = 1;
800 for (;;) {
801 const char *line = msg;
802 int linelen = get_one_line(line);
804 msg += linelen;
805 if (!linelen || is_empty_line(line, &linelen))
806 break;
808 if (!sb)
809 continue;
810 strbuf_grow(sb, linelen + 2);
811 if (!first)
812 strbuf_addstr(sb, line_separator);
813 strbuf_add(sb, line, linelen);
814 first = 0;
816 return msg;
819 static void parse_commit_message(struct format_commit_context *c)
821 const char *msg = c->message + c->message_off;
822 const char *start = c->message;
824 msg = skip_empty_lines(msg);
825 c->subject_off = msg - start;
827 msg = format_subject(NULL, msg, NULL);
828 msg = skip_empty_lines(msg);
829 c->body_off = msg - start;
831 c->commit_message_parsed = 1;
834 static void format_decoration(struct strbuf *sb, const struct commit *commit)
836 struct name_decoration *d;
837 const char *prefix = " (";
839 load_ref_decorations(DECORATE_SHORT_REFS);
840 d = lookup_decoration(&name_decoration, &commit->object);
841 while (d) {
842 strbuf_addstr(sb, prefix);
843 prefix = ", ";
844 strbuf_addstr(sb, d->name);
845 d = d->next;
847 if (prefix[0] == ',')
848 strbuf_addch(sb, ')');
851 static void strbuf_wrap(struct strbuf *sb, size_t pos,
852 size_t width, size_t indent1, size_t indent2)
854 struct strbuf tmp = STRBUF_INIT;
856 if (pos)
857 strbuf_add(&tmp, sb->buf, pos);
858 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
859 (int) indent1, (int) indent2, (int) width);
860 strbuf_swap(&tmp, sb);
861 strbuf_release(&tmp);
864 static void rewrap_message_tail(struct strbuf *sb,
865 struct format_commit_context *c,
866 size_t new_width, size_t new_indent1,
867 size_t new_indent2)
869 if (c->width == new_width && c->indent1 == new_indent1 &&
870 c->indent2 == new_indent2)
871 return;
872 if (c->wrap_start < sb->len)
873 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
874 c->wrap_start = sb->len;
875 c->width = new_width;
876 c->indent1 = new_indent1;
877 c->indent2 = new_indent2;
880 static struct {
881 char result;
882 const char *check;
883 } signature_check[] = {
884 { 'G', ": Good signature from " },
885 { 'B', ": BAD signature from " },
888 static void parse_signature_lines(struct format_commit_context *ctx)
890 const char *buf = ctx->signature.gpg_output;
891 int i;
893 for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
894 const char *found = strstr(buf, signature_check[i].check);
895 const char *next;
896 if (!found)
897 continue;
898 ctx->signature.good_bad = signature_check[i].result;
899 found += strlen(signature_check[i].check);
900 next = strchrnul(found, '\n');
901 ctx->signature.signer = xmemdupz(found, next - found);
902 break;
906 static void parse_commit_signature(struct format_commit_context *ctx)
908 struct strbuf payload = STRBUF_INIT;
909 struct strbuf signature = STRBUF_INIT;
910 struct strbuf gpg_output = STRBUF_INIT;
911 int status;
913 ctx->commit_signature_parsed = 1;
915 if (parse_signed_commit(ctx->commit->object.sha1,
916 &payload, &signature) <= 0)
917 goto out;
918 status = verify_signed_buffer(payload.buf, payload.len,
919 signature.buf, signature.len,
920 &gpg_output);
921 if (status && !gpg_output.len)
922 goto out;
923 ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
924 parse_signature_lines(ctx);
926 out:
927 strbuf_release(&gpg_output);
928 strbuf_release(&payload);
929 strbuf_release(&signature);
933 static int format_reflog_person(struct strbuf *sb,
934 char part,
935 struct reflog_walk_info *log,
936 enum date_mode dmode)
938 const char *ident;
940 if (!log)
941 return 2;
943 ident = get_reflog_ident(log);
944 if (!ident)
945 return 2;
947 return format_person_part(sb, part, ident, strlen(ident), dmode);
950 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
951 void *context)
953 struct format_commit_context *c = context;
954 const struct commit *commit = c->commit;
955 const char *msg = c->message;
956 struct commit_list *p;
957 int h1, h2;
959 /* these are independent of the commit */
960 switch (placeholder[0]) {
961 case 'C':
962 if (placeholder[1] == '(') {
963 const char *begin = placeholder + 2;
964 const char *end = strchr(begin, ')');
965 char color[COLOR_MAXLEN];
967 if (!end)
968 return 0;
969 if (!prefixcmp(begin, "auto,")) {
970 if (!want_color(c->pretty_ctx->color))
971 return end - placeholder + 1;
972 begin += 5;
974 color_parse_mem(begin,
975 end - begin,
976 "--pretty format", color);
977 strbuf_addstr(sb, color);
978 return end - placeholder + 1;
980 if (!prefixcmp(placeholder + 1, "red")) {
981 strbuf_addstr(sb, GIT_COLOR_RED);
982 return 4;
983 } else if (!prefixcmp(placeholder + 1, "green")) {
984 strbuf_addstr(sb, GIT_COLOR_GREEN);
985 return 6;
986 } else if (!prefixcmp(placeholder + 1, "blue")) {
987 strbuf_addstr(sb, GIT_COLOR_BLUE);
988 return 5;
989 } else if (!prefixcmp(placeholder + 1, "reset")) {
990 strbuf_addstr(sb, GIT_COLOR_RESET);
991 return 6;
992 } else
993 return 0;
994 case 'n': /* newline */
995 strbuf_addch(sb, '\n');
996 return 1;
997 case 'x':
998 /* %x00 == NUL, %x0a == LF, etc. */
999 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1000 h1 <= 16 &&
1001 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1002 h2 <= 16) {
1003 strbuf_addch(sb, (h1<<4)|h2);
1004 return 3;
1005 } else
1006 return 0;
1007 case 'w':
1008 if (placeholder[1] == '(') {
1009 unsigned long width = 0, indent1 = 0, indent2 = 0;
1010 char *next;
1011 const char *start = placeholder + 2;
1012 const char *end = strchr(start, ')');
1013 if (!end)
1014 return 0;
1015 if (end > start) {
1016 width = strtoul(start, &next, 10);
1017 if (*next == ',') {
1018 indent1 = strtoul(next + 1, &next, 10);
1019 if (*next == ',') {
1020 indent2 = strtoul(next + 1,
1021 &next, 10);
1024 if (*next != ')')
1025 return 0;
1027 rewrap_message_tail(sb, c, width, indent1, indent2);
1028 return end - placeholder + 1;
1029 } else
1030 return 0;
1033 /* these depend on the commit */
1034 if (!commit->object.parsed)
1035 parse_object(commit->object.sha1);
1037 switch (placeholder[0]) {
1038 case 'H': /* commit hash */
1039 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1040 return 1;
1041 case 'h': /* abbreviated commit hash */
1042 if (add_again(sb, &c->abbrev_commit_hash))
1043 return 1;
1044 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1045 c->pretty_ctx->abbrev));
1046 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1047 return 1;
1048 case 'T': /* tree hash */
1049 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1050 return 1;
1051 case 't': /* abbreviated tree hash */
1052 if (add_again(sb, &c->abbrev_tree_hash))
1053 return 1;
1054 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1055 c->pretty_ctx->abbrev));
1056 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1057 return 1;
1058 case 'P': /* parent hashes */
1059 for (p = commit->parents; p; p = p->next) {
1060 if (p != commit->parents)
1061 strbuf_addch(sb, ' ');
1062 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1064 return 1;
1065 case 'p': /* abbreviated parent hashes */
1066 if (add_again(sb, &c->abbrev_parent_hashes))
1067 return 1;
1068 for (p = commit->parents; p; p = p->next) {
1069 if (p != commit->parents)
1070 strbuf_addch(sb, ' ');
1071 strbuf_addstr(sb, find_unique_abbrev(
1072 p->item->object.sha1,
1073 c->pretty_ctx->abbrev));
1075 c->abbrev_parent_hashes.len = sb->len -
1076 c->abbrev_parent_hashes.off;
1077 return 1;
1078 case 'm': /* left/right/bottom */
1079 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1080 return 1;
1081 case 'd':
1082 format_decoration(sb, commit);
1083 return 1;
1084 case 'g': /* reflog info */
1085 switch(placeholder[1]) {
1086 case 'd': /* reflog selector */
1087 case 'D':
1088 if (c->pretty_ctx->reflog_info)
1089 get_reflog_selector(sb,
1090 c->pretty_ctx->reflog_info,
1091 c->pretty_ctx->date_mode,
1092 c->pretty_ctx->date_mode_explicit,
1093 (placeholder[1] == 'd'));
1094 return 2;
1095 case 's': /* reflog message */
1096 if (c->pretty_ctx->reflog_info)
1097 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1098 return 2;
1099 case 'n':
1100 case 'N':
1101 case 'e':
1102 case 'E':
1103 return format_reflog_person(sb,
1104 placeholder[1],
1105 c->pretty_ctx->reflog_info,
1106 c->pretty_ctx->date_mode);
1108 return 0; /* unknown %g placeholder */
1109 case 'N':
1110 if (c->pretty_ctx->notes_message) {
1111 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1112 return 1;
1114 return 0;
1117 if (placeholder[0] == 'G') {
1118 if (!c->commit_signature_parsed)
1119 parse_commit_signature(c);
1120 switch (placeholder[1]) {
1121 case 'G':
1122 if (c->signature.gpg_output)
1123 strbuf_addstr(sb, c->signature.gpg_output);
1124 break;
1125 case '?':
1126 switch (c->signature.good_bad) {
1127 case 'G':
1128 case 'B':
1129 strbuf_addch(sb, c->signature.good_bad);
1131 break;
1132 case 'S':
1133 if (c->signature.signer)
1134 strbuf_addstr(sb, c->signature.signer);
1135 break;
1137 return 2;
1141 /* For the rest we have to parse the commit header. */
1142 if (!c->commit_header_parsed)
1143 parse_commit_header(c);
1145 switch (placeholder[0]) {
1146 case 'a': /* author ... */
1147 return format_person_part(sb, placeholder[1],
1148 msg + c->author.off, c->author.len,
1149 c->pretty_ctx->date_mode);
1150 case 'c': /* committer ... */
1151 return format_person_part(sb, placeholder[1],
1152 msg + c->committer.off, c->committer.len,
1153 c->pretty_ctx->date_mode);
1154 case 'e': /* encoding */
1155 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1156 return 1;
1157 case 'B': /* raw body */
1158 /* message_off is always left at the initial newline */
1159 strbuf_addstr(sb, msg + c->message_off + 1);
1160 return 1;
1163 /* Now we need to parse the commit message. */
1164 if (!c->commit_message_parsed)
1165 parse_commit_message(c);
1167 switch (placeholder[0]) {
1168 case 's': /* subject */
1169 format_subject(sb, msg + c->subject_off, " ");
1170 return 1;
1171 case 'f': /* sanitized subject */
1172 format_sanitized_subject(sb, msg + c->subject_off);
1173 return 1;
1174 case 'b': /* body */
1175 strbuf_addstr(sb, msg + c->body_off);
1176 return 1;
1178 return 0; /* unknown placeholder */
1181 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1182 void *context)
1184 int consumed;
1185 size_t orig_len;
1186 enum {
1187 NO_MAGIC,
1188 ADD_LF_BEFORE_NON_EMPTY,
1189 DEL_LF_BEFORE_EMPTY,
1190 ADD_SP_BEFORE_NON_EMPTY
1191 } magic = NO_MAGIC;
1193 switch (placeholder[0]) {
1194 case '-':
1195 magic = DEL_LF_BEFORE_EMPTY;
1196 break;
1197 case '+':
1198 magic = ADD_LF_BEFORE_NON_EMPTY;
1199 break;
1200 case ' ':
1201 magic = ADD_SP_BEFORE_NON_EMPTY;
1202 break;
1203 default:
1204 break;
1206 if (magic != NO_MAGIC)
1207 placeholder++;
1209 orig_len = sb->len;
1210 consumed = format_commit_one(sb, placeholder, context);
1211 if (magic == NO_MAGIC)
1212 return consumed;
1214 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1215 while (sb->len && sb->buf[sb->len - 1] == '\n')
1216 strbuf_setlen(sb, sb->len - 1);
1217 } else if (orig_len != sb->len) {
1218 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1219 strbuf_insert(sb, orig_len, "\n", 1);
1220 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1221 strbuf_insert(sb, orig_len, " ", 1);
1223 return consumed + 1;
1226 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1227 void *context)
1229 struct userformat_want *w = context;
1231 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1232 placeholder++;
1234 switch (*placeholder) {
1235 case 'N':
1236 w->notes = 1;
1237 break;
1239 return 0;
1242 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1244 struct strbuf dummy = STRBUF_INIT;
1246 if (!fmt) {
1247 if (!user_format)
1248 return;
1249 fmt = user_format;
1251 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1252 strbuf_release(&dummy);
1255 void format_commit_message(const struct commit *commit,
1256 const char *format, struct strbuf *sb,
1257 const struct pretty_print_context *pretty_ctx)
1259 struct format_commit_context context;
1260 const char *output_enc = pretty_ctx->output_encoding;
1262 memset(&context, 0, sizeof(context));
1263 context.commit = commit;
1264 context.pretty_ctx = pretty_ctx;
1265 context.wrap_start = sb->len;
1266 context.message = logmsg_reencode(commit, output_enc);
1267 if (!context.message)
1268 context.message = commit->buffer;
1270 strbuf_expand(sb, format, format_commit_item, &context);
1271 rewrap_message_tail(sb, &context, 0, 0, 0);
1273 if (context.message != commit->buffer)
1274 free(context.message);
1275 free(context.signature.gpg_output);
1276 free(context.signature.signer);
1279 static void pp_header(const struct pretty_print_context *pp,
1280 const char *encoding,
1281 const struct commit *commit,
1282 const char **msg_p,
1283 struct strbuf *sb)
1285 int parents_shown = 0;
1287 for (;;) {
1288 const char *line = *msg_p;
1289 int linelen = get_one_line(*msg_p);
1291 if (!linelen)
1292 return;
1293 *msg_p += linelen;
1295 if (linelen == 1)
1296 /* End of header */
1297 return;
1299 if (pp->fmt == CMIT_FMT_RAW) {
1300 strbuf_add(sb, line, linelen);
1301 continue;
1304 if (!prefixcmp(line, "parent ")) {
1305 if (linelen != 48)
1306 die("bad parent line in commit");
1307 continue;
1310 if (!parents_shown) {
1311 struct commit_list *parent;
1312 int num;
1313 for (parent = commit->parents, num = 0;
1314 parent;
1315 parent = parent->next, num++)
1317 /* with enough slop */
1318 strbuf_grow(sb, num * 50 + 20);
1319 add_merge_info(pp, sb, commit);
1320 parents_shown = 1;
1324 * MEDIUM == DEFAULT shows only author with dates.
1325 * FULL shows both authors but not dates.
1326 * FULLER shows both authors and dates.
1328 if (!prefixcmp(line, "author ")) {
1329 strbuf_grow(sb, linelen + 80);
1330 pp_user_info(pp, "Author", sb, line + 7, encoding);
1332 if (!prefixcmp(line, "committer ") &&
1333 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1334 strbuf_grow(sb, linelen + 80);
1335 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1340 void pp_title_line(const struct pretty_print_context *pp,
1341 const char **msg_p,
1342 struct strbuf *sb,
1343 const char *encoding,
1344 int need_8bit_cte)
1346 static const int max_length = 78; /* per rfc2047 */
1347 struct strbuf title;
1349 strbuf_init(&title, 80);
1350 *msg_p = format_subject(&title, *msg_p,
1351 pp->preserve_subject ? "\n" : " ");
1353 strbuf_grow(sb, title.len + 1024);
1354 if (pp->subject) {
1355 strbuf_addstr(sb, pp->subject);
1356 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1357 add_rfc2047(sb, title.buf, title.len,
1358 encoding, RFC2047_SUBJECT);
1359 else
1360 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1361 -last_line_length(sb), 1, max_length);
1362 } else {
1363 strbuf_addbuf(sb, &title);
1365 strbuf_addch(sb, '\n');
1367 if (need_8bit_cte > 0) {
1368 const char *header_fmt =
1369 "MIME-Version: 1.0\n"
1370 "Content-Type: text/plain; charset=%s\n"
1371 "Content-Transfer-Encoding: 8bit\n";
1372 strbuf_addf(sb, header_fmt, encoding);
1374 if (pp->after_subject) {
1375 strbuf_addstr(sb, pp->after_subject);
1377 if (pp->fmt == CMIT_FMT_EMAIL) {
1378 strbuf_addch(sb, '\n');
1380 strbuf_release(&title);
1383 void pp_remainder(const struct pretty_print_context *pp,
1384 const char **msg_p,
1385 struct strbuf *sb,
1386 int indent)
1388 int first = 1;
1389 for (;;) {
1390 const char *line = *msg_p;
1391 int linelen = get_one_line(line);
1392 *msg_p += linelen;
1394 if (!linelen)
1395 break;
1397 if (is_empty_line(line, &linelen)) {
1398 if (first)
1399 continue;
1400 if (pp->fmt == CMIT_FMT_SHORT)
1401 break;
1403 first = 0;
1405 strbuf_grow(sb, linelen + indent + 20);
1406 if (indent) {
1407 memset(sb->buf + sb->len, ' ', indent);
1408 strbuf_setlen(sb, sb->len + indent);
1410 strbuf_add(sb, line, linelen);
1411 strbuf_addch(sb, '\n');
1415 void pretty_print_commit(const struct pretty_print_context *pp,
1416 const struct commit *commit,
1417 struct strbuf *sb)
1419 unsigned long beginning_of_body;
1420 int indent = 4;
1421 const char *msg = commit->buffer;
1422 char *reencoded;
1423 const char *encoding;
1424 int need_8bit_cte = pp->need_8bit_cte;
1426 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1427 format_commit_message(commit, user_format, sb, pp);
1428 return;
1431 encoding = get_log_output_encoding();
1432 reencoded = logmsg_reencode(commit, encoding);
1433 if (reencoded) {
1434 msg = reencoded;
1437 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1438 indent = 0;
1441 * We need to check and emit Content-type: to mark it
1442 * as 8-bit if we haven't done so.
1444 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1445 int i, ch, in_body;
1447 for (in_body = i = 0; (ch = msg[i]); i++) {
1448 if (!in_body) {
1449 /* author could be non 7-bit ASCII but
1450 * the log may be so; skip over the
1451 * header part first.
1453 if (ch == '\n' && msg[i+1] == '\n')
1454 in_body = 1;
1456 else if (non_ascii(ch)) {
1457 need_8bit_cte = 1;
1458 break;
1463 pp_header(pp, encoding, commit, &msg, sb);
1464 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1465 strbuf_addch(sb, '\n');
1468 /* Skip excess blank lines at the beginning of body, if any... */
1469 msg = skip_empty_lines(msg);
1471 /* These formats treat the title line specially. */
1472 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1473 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1475 beginning_of_body = sb->len;
1476 if (pp->fmt != CMIT_FMT_ONELINE)
1477 pp_remainder(pp, &msg, sb, indent);
1478 strbuf_rtrim(sb);
1480 /* Make sure there is an EOLN for the non-oneline case */
1481 if (pp->fmt != CMIT_FMT_ONELINE)
1482 strbuf_addch(sb, '\n');
1485 * The caller may append additional body text in e-mail
1486 * format. Make sure we did not strip the blank line
1487 * between the header and the body.
1489 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1490 strbuf_addch(sb, '\n');
1492 free(reencoded);
1495 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1496 struct strbuf *sb)
1498 struct pretty_print_context pp = {0};
1499 pp.fmt = fmt;
1500 pretty_print_commit(&pp, commit, sb);