Merge branch 'yd/doc-merge-annotated-tag'
[git/gitweb.git] / pretty.c
blob41f04e669d3b9c38f49e0853053a6ba27b97357a
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, size_t 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? */
359 while (len) {
361 * RFC 2047, section 5 (3):
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.
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);
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;
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.
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 */
389 for (i = 0; i < chrlen; i++)
390 strbuf_addf(sb, encoded_fmt, p[i]);
391 line_len += encoded_len;
393 strbuf_addstr(sb, "?=");
396 void pp_user_info(const struct pretty_print_context *pp,
397 const char *what, struct strbuf *sb,
398 const char *line, const char *encoding)
400 struct strbuf name;
401 struct strbuf mail;
402 struct ident_split ident;
403 int linelen;
404 char *line_end, *date;
405 const char *mailbuf, *namebuf;
406 size_t namelen, maillen;
407 int max_length = 78; /* per rfc2822 */
408 unsigned long time;
409 int tz;
411 if (pp->fmt == CMIT_FMT_ONELINE)
412 return;
414 line_end = strchr(line, '\n');
415 if (!line_end) {
416 line_end = strchr(line, '\0');
417 if (!line_end)
418 return;
421 linelen = ++line_end - line;
422 if (split_ident_line(&ident, line, linelen))
423 return;
426 mailbuf = ident.mail_begin;
427 maillen = ident.mail_end - ident.mail_begin;
428 namebuf = ident.name_begin;
429 namelen = ident.name_end - ident.name_begin;
431 if (pp->mailmap)
432 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
434 strbuf_init(&mail, 0);
435 strbuf_init(&name, 0);
437 strbuf_add(&mail, mailbuf, maillen);
438 strbuf_add(&name, namebuf, namelen);
440 namelen = name.len + mail.len + 3; /* ' ' + '<' + '>' */
441 time = strtoul(ident.date_begin, &date, 10);
442 tz = strtol(date, NULL, 10);
444 if (pp->fmt == CMIT_FMT_EMAIL) {
445 strbuf_addstr(sb, "From: ");
446 if (needs_rfc2047_encoding(name.buf, name.len, RFC2047_ADDRESS)) {
447 add_rfc2047(sb, name.buf, name.len,
448 encoding, RFC2047_ADDRESS);
449 max_length = 76; /* per rfc2047 */
450 } else if (needs_rfc822_quoting(name.buf, name.len)) {
451 struct strbuf quoted = STRBUF_INIT;
452 add_rfc822_quoted(&quoted, name.buf, name.len);
453 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
454 -6, 1, max_length);
455 strbuf_release(&quoted);
456 } else {
457 strbuf_add_wrapped_bytes(sb, name.buf, name.len,
458 -6, 1, max_length);
460 if (namelen - name.len + last_line_length(sb) > max_length)
461 strbuf_addch(sb, '\n');
463 strbuf_addf(sb, " <%s>\n", mail.buf);
464 } else {
465 strbuf_addf(sb, "%s: %.*s%s <%s>\n", what,
466 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
467 " ", name.buf, mail.buf);
470 strbuf_release(&mail);
471 strbuf_release(&name);
473 switch (pp->fmt) {
474 case CMIT_FMT_MEDIUM:
475 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, pp->date_mode));
476 break;
477 case CMIT_FMT_EMAIL:
478 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
479 break;
480 case CMIT_FMT_FULLER:
481 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
482 break;
483 default:
484 /* notin' */
485 break;
489 static int is_empty_line(const char *line, int *len_p)
491 int len = *len_p;
492 while (len && isspace(line[len-1]))
493 len--;
494 *len_p = len;
495 return !len;
498 static const char *skip_empty_lines(const char *msg)
500 for (;;) {
501 int linelen = get_one_line(msg);
502 int ll = linelen;
503 if (!linelen)
504 break;
505 if (!is_empty_line(msg, &ll))
506 break;
507 msg += linelen;
509 return msg;
512 static void add_merge_info(const struct pretty_print_context *pp,
513 struct strbuf *sb, const struct commit *commit)
515 struct commit_list *parent = commit->parents;
517 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
518 !parent || !parent->next)
519 return;
521 strbuf_addstr(sb, "Merge:");
523 while (parent) {
524 struct commit *p = parent->item;
525 const char *hex = NULL;
526 if (pp->abbrev)
527 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
528 if (!hex)
529 hex = sha1_to_hex(p->object.sha1);
530 parent = parent->next;
532 strbuf_addf(sb, " %s", hex);
534 strbuf_addch(sb, '\n');
537 static char *get_header(const struct commit *commit, const char *msg,
538 const char *key)
540 int key_len = strlen(key);
541 const char *line = msg;
543 while (line) {
544 const char *eol = strchr(line, '\n'), *next;
546 if (line == eol)
547 return NULL;
548 if (!eol) {
549 warning("malformed commit (header is missing newline): %s",
550 sha1_to_hex(commit->object.sha1));
551 eol = line + strlen(line);
552 next = NULL;
553 } else
554 next = eol + 1;
555 if (eol - line > key_len &&
556 !strncmp(line, key, key_len) &&
557 line[key_len] == ' ') {
558 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
560 line = next;
562 return NULL;
565 static char *replace_encoding_header(char *buf, const char *encoding)
567 struct strbuf tmp = STRBUF_INIT;
568 size_t start, len;
569 char *cp = buf;
571 /* guess if there is an encoding header before a \n\n */
572 while (strncmp(cp, "encoding ", strlen("encoding "))) {
573 cp = strchr(cp, '\n');
574 if (!cp || *++cp == '\n')
575 return buf;
577 start = cp - buf;
578 cp = strchr(cp, '\n');
579 if (!cp)
580 return buf; /* should not happen but be defensive */
581 len = cp + 1 - (buf + start);
583 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
584 if (is_encoding_utf8(encoding)) {
585 /* we have re-coded to UTF-8; drop the header */
586 strbuf_remove(&tmp, start, len);
587 } else {
588 /* just replaces XXXX in 'encoding XXXX\n' */
589 strbuf_splice(&tmp, start + strlen("encoding "),
590 len - strlen("encoding \n"),
591 encoding, strlen(encoding));
593 return strbuf_detach(&tmp, NULL);
596 char *logmsg_reencode(const struct commit *commit,
597 const char *output_encoding)
599 static const char *utf8 = "UTF-8";
600 const char *use_encoding;
601 char *encoding;
602 char *msg = commit->buffer;
603 char *out;
605 if (!msg) {
606 enum object_type type;
607 unsigned long size;
609 msg = read_sha1_file(commit->object.sha1, &type, &size);
610 if (!msg)
611 die("Cannot read commit object %s",
612 sha1_to_hex(commit->object.sha1));
613 if (type != OBJ_COMMIT)
614 die("Expected commit for '%s', got %s",
615 sha1_to_hex(commit->object.sha1), typename(type));
618 if (!output_encoding || !*output_encoding)
619 return msg;
620 encoding = get_header(commit, msg, "encoding");
621 use_encoding = encoding ? encoding : utf8;
622 if (same_encoding(use_encoding, output_encoding)) {
624 * No encoding work to be done. If we have no encoding header
625 * at all, then there's nothing to do, and we can return the
626 * message verbatim (whether newly allocated or not).
628 if (!encoding)
629 return msg;
632 * Otherwise, we still want to munge the encoding header in the
633 * result, which will be done by modifying the buffer. If we
634 * are using a fresh copy, we can reuse it. But if we are using
635 * the cached copy from commit->buffer, we need to duplicate it
636 * to avoid munging commit->buffer.
638 out = msg;
639 if (out == commit->buffer)
640 out = xstrdup(out);
642 else {
644 * There's actual encoding work to do. Do the reencoding, which
645 * still leaves the header to be replaced in the next step. At
646 * this point, we are done with msg. If we allocated a fresh
647 * copy, we can free it.
649 out = reencode_string(msg, output_encoding, use_encoding);
650 if (out && msg != commit->buffer)
651 free(msg);
655 * This replacement actually consumes the buffer we hand it, so we do
656 * not have to worry about freeing the old "out" here.
658 if (out)
659 out = replace_encoding_header(out, output_encoding);
661 free(encoding);
663 * If the re-encoding failed, out might be NULL here; in that
664 * case we just return the commit message verbatim.
666 return out ? out : msg;
669 void logmsg_free(char *msg, const struct commit *commit)
671 if (msg != commit->buffer)
672 free(msg);
675 static int mailmap_name(const char **email, size_t *email_len,
676 const char **name, size_t *name_len)
678 static struct string_list *mail_map;
679 if (!mail_map) {
680 mail_map = xcalloc(1, sizeof(*mail_map));
681 read_mailmap(mail_map, NULL);
683 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
686 static size_t format_person_part(struct strbuf *sb, char part,
687 const char *msg, int len, enum date_mode dmode)
689 /* currently all placeholders have same length */
690 const int placeholder_len = 2;
691 int tz;
692 unsigned long date = 0;
693 struct ident_split s;
694 const char *name, *mail;
695 size_t maillen, namelen;
697 if (split_ident_line(&s, msg, len) < 0)
698 goto skip;
700 name = s.name_begin;
701 namelen = s.name_end - s.name_begin;
702 mail = s.mail_begin;
703 maillen = s.mail_end - s.mail_begin;
705 if (part == 'N' || part == 'E') /* mailmap lookup */
706 mailmap_name(&mail, &maillen, &name, &namelen);
707 if (part == 'n' || part == 'N') { /* name */
708 strbuf_add(sb, name, namelen);
709 return placeholder_len;
711 if (part == 'e' || part == 'E') { /* email */
712 strbuf_add(sb, mail, maillen);
713 return placeholder_len;
716 if (!s.date_begin)
717 goto skip;
719 date = strtoul(s.date_begin, NULL, 10);
721 if (part == 't') { /* date, UNIX timestamp */
722 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
723 return placeholder_len;
726 /* parse tz */
727 tz = strtoul(s.tz_begin + 1, NULL, 10);
728 if (*s.tz_begin == '-')
729 tz = -tz;
731 switch (part) {
732 case 'd': /* date */
733 strbuf_addstr(sb, show_date(date, tz, dmode));
734 return placeholder_len;
735 case 'D': /* date, RFC2822 style */
736 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
737 return placeholder_len;
738 case 'r': /* date, relative */
739 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
740 return placeholder_len;
741 case 'i': /* date, ISO 8601 */
742 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
743 return placeholder_len;
746 skip:
748 * reading from either a bogus commit, or a reflog entry with
749 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
750 * to compute a valid return value.
752 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
753 || part == 'D' || part == 'r' || part == 'i')
754 return placeholder_len;
756 return 0; /* unknown placeholder */
759 struct chunk {
760 size_t off;
761 size_t len;
764 struct format_commit_context {
765 const struct commit *commit;
766 const struct pretty_print_context *pretty_ctx;
767 unsigned commit_header_parsed:1;
768 unsigned commit_message_parsed:1;
769 unsigned commit_signature_parsed:1;
770 struct {
771 char *gpg_output;
772 char *gpg_status;
773 char good_bad;
774 char *signer;
775 char *key;
776 } signature;
777 char *message;
778 size_t width, indent1, indent2;
780 /* These offsets are relative to the start of the commit message. */
781 struct chunk author;
782 struct chunk committer;
783 struct chunk encoding;
784 size_t message_off;
785 size_t subject_off;
786 size_t body_off;
788 /* The following ones are relative to the result struct strbuf. */
789 struct chunk abbrev_commit_hash;
790 struct chunk abbrev_tree_hash;
791 struct chunk abbrev_parent_hashes;
792 size_t wrap_start;
795 static int add_again(struct strbuf *sb, struct chunk *chunk)
797 if (chunk->len) {
798 strbuf_adddup(sb, chunk->off, chunk->len);
799 return 1;
803 * We haven't seen this chunk before. Our caller is surely
804 * going to add it the hard way now. Remember the most likely
805 * start of the to-be-added chunk: the current end of the
806 * struct strbuf.
808 chunk->off = sb->len;
809 return 0;
812 static void parse_commit_header(struct format_commit_context *context)
814 const char *msg = context->message;
815 int i;
817 for (i = 0; msg[i]; i++) {
818 int eol;
819 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
820 ; /* do nothing */
822 if (i == eol) {
823 break;
824 } else if (!prefixcmp(msg + i, "author ")) {
825 context->author.off = i + 7;
826 context->author.len = eol - i - 7;
827 } else if (!prefixcmp(msg + i, "committer ")) {
828 context->committer.off = i + 10;
829 context->committer.len = eol - i - 10;
830 } else if (!prefixcmp(msg + i, "encoding ")) {
831 context->encoding.off = i + 9;
832 context->encoding.len = eol - i - 9;
834 i = eol;
836 context->message_off = i;
837 context->commit_header_parsed = 1;
840 static int istitlechar(char c)
842 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
843 (c >= '0' && c <= '9') || c == '.' || c == '_';
846 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
848 size_t trimlen;
849 size_t start_len = sb->len;
850 int space = 2;
852 for (; *msg && *msg != '\n'; msg++) {
853 if (istitlechar(*msg)) {
854 if (space == 1)
855 strbuf_addch(sb, '-');
856 space = 0;
857 strbuf_addch(sb, *msg);
858 if (*msg == '.')
859 while (*(msg+1) == '.')
860 msg++;
861 } else
862 space |= 1;
865 /* trim any trailing '.' or '-' characters */
866 trimlen = 0;
867 while (sb->len - trimlen > start_len &&
868 (sb->buf[sb->len - 1 - trimlen] == '.'
869 || sb->buf[sb->len - 1 - trimlen] == '-'))
870 trimlen++;
871 strbuf_remove(sb, sb->len - trimlen, trimlen);
874 const char *format_subject(struct strbuf *sb, const char *msg,
875 const char *line_separator)
877 int first = 1;
879 for (;;) {
880 const char *line = msg;
881 int linelen = get_one_line(line);
883 msg += linelen;
884 if (!linelen || is_empty_line(line, &linelen))
885 break;
887 if (!sb)
888 continue;
889 strbuf_grow(sb, linelen + 2);
890 if (!first)
891 strbuf_addstr(sb, line_separator);
892 strbuf_add(sb, line, linelen);
893 first = 0;
895 return msg;
898 static void parse_commit_message(struct format_commit_context *c)
900 const char *msg = c->message + c->message_off;
901 const char *start = c->message;
903 msg = skip_empty_lines(msg);
904 c->subject_off = msg - start;
906 msg = format_subject(NULL, msg, NULL);
907 msg = skip_empty_lines(msg);
908 c->body_off = msg - start;
910 c->commit_message_parsed = 1;
913 static void format_decoration(struct strbuf *sb, const struct commit *commit)
915 struct name_decoration *d;
916 const char *prefix = " (";
918 load_ref_decorations(DECORATE_SHORT_REFS);
919 d = lookup_decoration(&name_decoration, &commit->object);
920 while (d) {
921 strbuf_addstr(sb, prefix);
922 prefix = ", ";
923 strbuf_addstr(sb, d->name);
924 d = d->next;
926 if (prefix[0] == ',')
927 strbuf_addch(sb, ')');
930 static void strbuf_wrap(struct strbuf *sb, size_t pos,
931 size_t width, size_t indent1, size_t indent2)
933 struct strbuf tmp = STRBUF_INIT;
935 if (pos)
936 strbuf_add(&tmp, sb->buf, pos);
937 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
938 (int) indent1, (int) indent2, (int) width);
939 strbuf_swap(&tmp, sb);
940 strbuf_release(&tmp);
943 static void rewrap_message_tail(struct strbuf *sb,
944 struct format_commit_context *c,
945 size_t new_width, size_t new_indent1,
946 size_t new_indent2)
948 if (c->width == new_width && c->indent1 == new_indent1 &&
949 c->indent2 == new_indent2)
950 return;
951 if (c->wrap_start < sb->len)
952 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
953 c->wrap_start = sb->len;
954 c->width = new_width;
955 c->indent1 = new_indent1;
956 c->indent2 = new_indent2;
959 static struct {
960 char result;
961 const char *check;
962 } signature_check[] = {
963 { 'G', "\n[GNUPG:] GOODSIG " },
964 { 'B', "\n[GNUPG:] BADSIG " },
967 static void parse_signature_lines(struct format_commit_context *ctx)
969 const char *buf = ctx->signature.gpg_status;
970 int i;
972 for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
973 const char *found = strstr(buf, signature_check[i].check);
974 const char *next;
975 if (!found)
976 continue;
977 ctx->signature.good_bad = signature_check[i].result;
978 found += strlen(signature_check[i].check);
979 ctx->signature.key = xmemdupz(found, 16);
980 found += 17;
981 next = strchrnul(found, '\n');
982 ctx->signature.signer = xmemdupz(found, next - found);
983 break;
987 static void parse_commit_signature(struct format_commit_context *ctx)
989 struct strbuf payload = STRBUF_INIT;
990 struct strbuf signature = STRBUF_INIT;
991 struct strbuf gpg_output = STRBUF_INIT;
992 struct strbuf gpg_status = STRBUF_INIT;
993 int status;
995 ctx->commit_signature_parsed = 1;
997 if (parse_signed_commit(ctx->commit->object.sha1,
998 &payload, &signature) <= 0)
999 goto out;
1000 status = verify_signed_buffer(payload.buf, payload.len,
1001 signature.buf, signature.len,
1002 &gpg_output, &gpg_status);
1003 if (status && !gpg_output.len)
1004 goto out;
1005 ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
1006 ctx->signature.gpg_status = strbuf_detach(&gpg_status, NULL);
1007 parse_signature_lines(ctx);
1009 out:
1010 strbuf_release(&gpg_status);
1011 strbuf_release(&gpg_output);
1012 strbuf_release(&payload);
1013 strbuf_release(&signature);
1017 static int format_reflog_person(struct strbuf *sb,
1018 char part,
1019 struct reflog_walk_info *log,
1020 enum date_mode dmode)
1022 const char *ident;
1024 if (!log)
1025 return 2;
1027 ident = get_reflog_ident(log);
1028 if (!ident)
1029 return 2;
1031 return format_person_part(sb, part, ident, strlen(ident), dmode);
1034 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
1035 void *context)
1037 struct format_commit_context *c = context;
1038 const struct commit *commit = c->commit;
1039 const char *msg = c->message;
1040 struct commit_list *p;
1041 int h1, h2;
1043 /* these are independent of the commit */
1044 switch (placeholder[0]) {
1045 case 'C':
1046 if (placeholder[1] == '(') {
1047 const char *begin = placeholder + 2;
1048 const char *end = strchr(begin, ')');
1049 char color[COLOR_MAXLEN];
1051 if (!end)
1052 return 0;
1053 if (!prefixcmp(begin, "auto,")) {
1054 if (!want_color(c->pretty_ctx->color))
1055 return end - placeholder + 1;
1056 begin += 5;
1058 color_parse_mem(begin,
1059 end - begin,
1060 "--pretty format", color);
1061 strbuf_addstr(sb, color);
1062 return end - placeholder + 1;
1064 if (!prefixcmp(placeholder + 1, "red")) {
1065 strbuf_addstr(sb, GIT_COLOR_RED);
1066 return 4;
1067 } else if (!prefixcmp(placeholder + 1, "green")) {
1068 strbuf_addstr(sb, GIT_COLOR_GREEN);
1069 return 6;
1070 } else if (!prefixcmp(placeholder + 1, "blue")) {
1071 strbuf_addstr(sb, GIT_COLOR_BLUE);
1072 return 5;
1073 } else if (!prefixcmp(placeholder + 1, "reset")) {
1074 strbuf_addstr(sb, GIT_COLOR_RESET);
1075 return 6;
1076 } else
1077 return 0;
1078 case 'n': /* newline */
1079 strbuf_addch(sb, '\n');
1080 return 1;
1081 case 'x':
1082 /* %x00 == NUL, %x0a == LF, etc. */
1083 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1084 h1 <= 16 &&
1085 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1086 h2 <= 16) {
1087 strbuf_addch(sb, (h1<<4)|h2);
1088 return 3;
1089 } else
1090 return 0;
1091 case 'w':
1092 if (placeholder[1] == '(') {
1093 unsigned long width = 0, indent1 = 0, indent2 = 0;
1094 char *next;
1095 const char *start = placeholder + 2;
1096 const char *end = strchr(start, ')');
1097 if (!end)
1098 return 0;
1099 if (end > start) {
1100 width = strtoul(start, &next, 10);
1101 if (*next == ',') {
1102 indent1 = strtoul(next + 1, &next, 10);
1103 if (*next == ',') {
1104 indent2 = strtoul(next + 1,
1105 &next, 10);
1108 if (*next != ')')
1109 return 0;
1111 rewrap_message_tail(sb, c, width, indent1, indent2);
1112 return end - placeholder + 1;
1113 } else
1114 return 0;
1117 /* these depend on the commit */
1118 if (!commit->object.parsed)
1119 parse_object(commit->object.sha1);
1121 switch (placeholder[0]) {
1122 case 'H': /* commit hash */
1123 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1124 return 1;
1125 case 'h': /* abbreviated commit hash */
1126 if (add_again(sb, &c->abbrev_commit_hash))
1127 return 1;
1128 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1129 c->pretty_ctx->abbrev));
1130 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1131 return 1;
1132 case 'T': /* tree hash */
1133 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1134 return 1;
1135 case 't': /* abbreviated tree hash */
1136 if (add_again(sb, &c->abbrev_tree_hash))
1137 return 1;
1138 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1139 c->pretty_ctx->abbrev));
1140 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1141 return 1;
1142 case 'P': /* parent hashes */
1143 for (p = commit->parents; p; p = p->next) {
1144 if (p != commit->parents)
1145 strbuf_addch(sb, ' ');
1146 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1148 return 1;
1149 case 'p': /* abbreviated parent hashes */
1150 if (add_again(sb, &c->abbrev_parent_hashes))
1151 return 1;
1152 for (p = commit->parents; p; p = p->next) {
1153 if (p != commit->parents)
1154 strbuf_addch(sb, ' ');
1155 strbuf_addstr(sb, find_unique_abbrev(
1156 p->item->object.sha1,
1157 c->pretty_ctx->abbrev));
1159 c->abbrev_parent_hashes.len = sb->len -
1160 c->abbrev_parent_hashes.off;
1161 return 1;
1162 case 'm': /* left/right/bottom */
1163 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1164 return 1;
1165 case 'd':
1166 format_decoration(sb, commit);
1167 return 1;
1168 case 'g': /* reflog info */
1169 switch(placeholder[1]) {
1170 case 'd': /* reflog selector */
1171 case 'D':
1172 if (c->pretty_ctx->reflog_info)
1173 get_reflog_selector(sb,
1174 c->pretty_ctx->reflog_info,
1175 c->pretty_ctx->date_mode,
1176 c->pretty_ctx->date_mode_explicit,
1177 (placeholder[1] == 'd'));
1178 return 2;
1179 case 's': /* reflog message */
1180 if (c->pretty_ctx->reflog_info)
1181 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1182 return 2;
1183 case 'n':
1184 case 'N':
1185 case 'e':
1186 case 'E':
1187 return format_reflog_person(sb,
1188 placeholder[1],
1189 c->pretty_ctx->reflog_info,
1190 c->pretty_ctx->date_mode);
1192 return 0; /* unknown %g placeholder */
1193 case 'N':
1194 if (c->pretty_ctx->notes_message) {
1195 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1196 return 1;
1198 return 0;
1201 if (placeholder[0] == 'G') {
1202 if (!c->commit_signature_parsed)
1203 parse_commit_signature(c);
1204 switch (placeholder[1]) {
1205 case 'G':
1206 if (c->signature.gpg_output)
1207 strbuf_addstr(sb, c->signature.gpg_output);
1208 break;
1209 case '?':
1210 switch (c->signature.good_bad) {
1211 case 'G':
1212 case 'B':
1213 strbuf_addch(sb, c->signature.good_bad);
1215 break;
1216 case 'S':
1217 if (c->signature.signer)
1218 strbuf_addstr(sb, c->signature.signer);
1219 break;
1220 case 'K':
1221 if (c->signature.key)
1222 strbuf_addstr(sb, c->signature.key);
1223 break;
1225 return 2;
1229 /* For the rest we have to parse the commit header. */
1230 if (!c->commit_header_parsed)
1231 parse_commit_header(c);
1233 switch (placeholder[0]) {
1234 case 'a': /* author ... */
1235 return format_person_part(sb, placeholder[1],
1236 msg + c->author.off, c->author.len,
1237 c->pretty_ctx->date_mode);
1238 case 'c': /* committer ... */
1239 return format_person_part(sb, placeholder[1],
1240 msg + c->committer.off, c->committer.len,
1241 c->pretty_ctx->date_mode);
1242 case 'e': /* encoding */
1243 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1244 return 1;
1245 case 'B': /* raw body */
1246 /* message_off is always left at the initial newline */
1247 strbuf_addstr(sb, msg + c->message_off + 1);
1248 return 1;
1251 /* Now we need to parse the commit message. */
1252 if (!c->commit_message_parsed)
1253 parse_commit_message(c);
1255 switch (placeholder[0]) {
1256 case 's': /* subject */
1257 format_subject(sb, msg + c->subject_off, " ");
1258 return 1;
1259 case 'f': /* sanitized subject */
1260 format_sanitized_subject(sb, msg + c->subject_off);
1261 return 1;
1262 case 'b': /* body */
1263 strbuf_addstr(sb, msg + c->body_off);
1264 return 1;
1266 return 0; /* unknown placeholder */
1269 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1270 void *context)
1272 int consumed;
1273 size_t orig_len;
1274 enum {
1275 NO_MAGIC,
1276 ADD_LF_BEFORE_NON_EMPTY,
1277 DEL_LF_BEFORE_EMPTY,
1278 ADD_SP_BEFORE_NON_EMPTY
1279 } magic = NO_MAGIC;
1281 switch (placeholder[0]) {
1282 case '-':
1283 magic = DEL_LF_BEFORE_EMPTY;
1284 break;
1285 case '+':
1286 magic = ADD_LF_BEFORE_NON_EMPTY;
1287 break;
1288 case ' ':
1289 magic = ADD_SP_BEFORE_NON_EMPTY;
1290 break;
1291 default:
1292 break;
1294 if (magic != NO_MAGIC)
1295 placeholder++;
1297 orig_len = sb->len;
1298 consumed = format_commit_one(sb, placeholder, context);
1299 if (magic == NO_MAGIC)
1300 return consumed;
1302 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1303 while (sb->len && sb->buf[sb->len - 1] == '\n')
1304 strbuf_setlen(sb, sb->len - 1);
1305 } else if (orig_len != sb->len) {
1306 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1307 strbuf_insert(sb, orig_len, "\n", 1);
1308 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1309 strbuf_insert(sb, orig_len, " ", 1);
1311 return consumed + 1;
1314 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1315 void *context)
1317 struct userformat_want *w = context;
1319 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1320 placeholder++;
1322 switch (*placeholder) {
1323 case 'N':
1324 w->notes = 1;
1325 break;
1327 return 0;
1330 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1332 struct strbuf dummy = STRBUF_INIT;
1334 if (!fmt) {
1335 if (!user_format)
1336 return;
1337 fmt = user_format;
1339 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1340 strbuf_release(&dummy);
1343 void format_commit_message(const struct commit *commit,
1344 const char *format, struct strbuf *sb,
1345 const struct pretty_print_context *pretty_ctx)
1347 struct format_commit_context context;
1348 const char *output_enc = pretty_ctx->output_encoding;
1350 memset(&context, 0, sizeof(context));
1351 context.commit = commit;
1352 context.pretty_ctx = pretty_ctx;
1353 context.wrap_start = sb->len;
1354 context.message = logmsg_reencode(commit, output_enc);
1356 strbuf_expand(sb, format, format_commit_item, &context);
1357 rewrap_message_tail(sb, &context, 0, 0, 0);
1359 logmsg_free(context.message, commit);
1360 free(context.signature.gpg_output);
1361 free(context.signature.signer);
1364 static void pp_header(const struct pretty_print_context *pp,
1365 const char *encoding,
1366 const struct commit *commit,
1367 const char **msg_p,
1368 struct strbuf *sb)
1370 int parents_shown = 0;
1372 for (;;) {
1373 const char *line = *msg_p;
1374 int linelen = get_one_line(*msg_p);
1376 if (!linelen)
1377 return;
1378 *msg_p += linelen;
1380 if (linelen == 1)
1381 /* End of header */
1382 return;
1384 if (pp->fmt == CMIT_FMT_RAW) {
1385 strbuf_add(sb, line, linelen);
1386 continue;
1389 if (!prefixcmp(line, "parent ")) {
1390 if (linelen != 48)
1391 die("bad parent line in commit");
1392 continue;
1395 if (!parents_shown) {
1396 struct commit_list *parent;
1397 int num;
1398 for (parent = commit->parents, num = 0;
1399 parent;
1400 parent = parent->next, num++)
1402 /* with enough slop */
1403 strbuf_grow(sb, num * 50 + 20);
1404 add_merge_info(pp, sb, commit);
1405 parents_shown = 1;
1409 * MEDIUM == DEFAULT shows only author with dates.
1410 * FULL shows both authors but not dates.
1411 * FULLER shows both authors and dates.
1413 if (!prefixcmp(line, "author ")) {
1414 strbuf_grow(sb, linelen + 80);
1415 pp_user_info(pp, "Author", sb, line + 7, encoding);
1417 if (!prefixcmp(line, "committer ") &&
1418 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1419 strbuf_grow(sb, linelen + 80);
1420 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1425 void pp_title_line(const struct pretty_print_context *pp,
1426 const char **msg_p,
1427 struct strbuf *sb,
1428 const char *encoding,
1429 int need_8bit_cte)
1431 static const int max_length = 78; /* per rfc2047 */
1432 struct strbuf title;
1434 strbuf_init(&title, 80);
1435 *msg_p = format_subject(&title, *msg_p,
1436 pp->preserve_subject ? "\n" : " ");
1438 strbuf_grow(sb, title.len + 1024);
1439 if (pp->subject) {
1440 strbuf_addstr(sb, pp->subject);
1441 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1442 add_rfc2047(sb, title.buf, title.len,
1443 encoding, RFC2047_SUBJECT);
1444 else
1445 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1446 -last_line_length(sb), 1, max_length);
1447 } else {
1448 strbuf_addbuf(sb, &title);
1450 strbuf_addch(sb, '\n');
1452 if (need_8bit_cte > 0) {
1453 const char *header_fmt =
1454 "MIME-Version: 1.0\n"
1455 "Content-Type: text/plain; charset=%s\n"
1456 "Content-Transfer-Encoding: 8bit\n";
1457 strbuf_addf(sb, header_fmt, encoding);
1459 if (pp->after_subject) {
1460 strbuf_addstr(sb, pp->after_subject);
1462 if (pp->fmt == CMIT_FMT_EMAIL) {
1463 strbuf_addch(sb, '\n');
1465 strbuf_release(&title);
1468 void pp_remainder(const struct pretty_print_context *pp,
1469 const char **msg_p,
1470 struct strbuf *sb,
1471 int indent)
1473 int first = 1;
1474 for (;;) {
1475 const char *line = *msg_p;
1476 int linelen = get_one_line(line);
1477 *msg_p += linelen;
1479 if (!linelen)
1480 break;
1482 if (is_empty_line(line, &linelen)) {
1483 if (first)
1484 continue;
1485 if (pp->fmt == CMIT_FMT_SHORT)
1486 break;
1488 first = 0;
1490 strbuf_grow(sb, linelen + indent + 20);
1491 if (indent) {
1492 memset(sb->buf + sb->len, ' ', indent);
1493 strbuf_setlen(sb, sb->len + indent);
1495 strbuf_add(sb, line, linelen);
1496 strbuf_addch(sb, '\n');
1500 void pretty_print_commit(const struct pretty_print_context *pp,
1501 const struct commit *commit,
1502 struct strbuf *sb)
1504 unsigned long beginning_of_body;
1505 int indent = 4;
1506 const char *msg;
1507 char *reencoded;
1508 const char *encoding;
1509 int need_8bit_cte = pp->need_8bit_cte;
1511 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1512 format_commit_message(commit, user_format, sb, pp);
1513 return;
1516 encoding = get_log_output_encoding();
1517 msg = reencoded = logmsg_reencode(commit, encoding);
1519 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1520 indent = 0;
1523 * We need to check and emit Content-type: to mark it
1524 * as 8-bit if we haven't done so.
1526 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1527 int i, ch, in_body;
1529 for (in_body = i = 0; (ch = msg[i]); i++) {
1530 if (!in_body) {
1531 /* author could be non 7-bit ASCII but
1532 * the log may be so; skip over the
1533 * header part first.
1535 if (ch == '\n' && msg[i+1] == '\n')
1536 in_body = 1;
1538 else if (non_ascii(ch)) {
1539 need_8bit_cte = 1;
1540 break;
1545 pp_header(pp, encoding, commit, &msg, sb);
1546 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1547 strbuf_addch(sb, '\n');
1550 /* Skip excess blank lines at the beginning of body, if any... */
1551 msg = skip_empty_lines(msg);
1553 /* These formats treat the title line specially. */
1554 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1555 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1557 beginning_of_body = sb->len;
1558 if (pp->fmt != CMIT_FMT_ONELINE)
1559 pp_remainder(pp, &msg, sb, indent);
1560 strbuf_rtrim(sb);
1562 /* Make sure there is an EOLN for the non-oneline case */
1563 if (pp->fmt != CMIT_FMT_ONELINE)
1564 strbuf_addch(sb, '\n');
1567 * The caller may append additional body text in e-mail
1568 * format. Make sure we did not strip the blank line
1569 * between the header and the body.
1571 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1572 strbuf_addch(sb, '\n');
1574 logmsg_free(reencoded, commit);
1577 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1578 struct strbuf *sb)
1580 struct pretty_print_context pp = {0};
1581 pp.fmt = fmt;
1582 pretty_print_commit(&pp, commit, sb);