hashfile: allow skipping the hash function
[git/debian.git] / pretty.c
blob6cb363ae1c9e90e47e2002fb7e89a1f03a12f23c
1 #include "cache.h"
2 #include "config.h"
3 #include "commit.h"
4 #include "utf8.h"
5 #include "diff.h"
6 #include "revision.h"
7 #include "string-list.h"
8 #include "mailmap.h"
9 #include "log-tree.h"
10 #include "notes.h"
11 #include "color.h"
12 #include "reflog-walk.h"
13 #include "gpg-interface.h"
14 #include "trailer.h"
15 #include "run-command.h"
17 static char *user_format;
18 static struct cmt_fmt_map {
19 const char *name;
20 enum cmit_fmt format;
21 int is_tformat;
22 int expand_tabs_in_log;
23 int is_alias;
24 enum date_mode_type default_date_mode_type;
25 const char *user_format;
26 } *commit_formats;
27 static size_t builtin_formats_len;
28 static size_t commit_formats_len;
29 static size_t commit_formats_alloc;
30 static struct cmt_fmt_map *find_commit_format(const char *sought);
32 int commit_format_is_empty(enum cmit_fmt fmt)
34 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
37 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
39 free(user_format);
40 user_format = xstrdup(cp);
41 if (is_tformat)
42 rev->use_terminator = 1;
43 rev->commit_format = CMIT_FMT_USERFORMAT;
46 static int git_pretty_formats_config(const char *var, const char *value,
47 void *cb UNUSED)
49 struct cmt_fmt_map *commit_format = NULL;
50 const char *name;
51 const char *fmt;
52 int i;
54 if (!skip_prefix(var, "pretty.", &name))
55 return 0;
57 for (i = 0; i < builtin_formats_len; i++) {
58 if (!strcmp(commit_formats[i].name, name))
59 return 0;
62 for (i = builtin_formats_len; i < commit_formats_len; i++) {
63 if (!strcmp(commit_formats[i].name, name)) {
64 commit_format = &commit_formats[i];
65 break;
69 if (!commit_format) {
70 ALLOC_GROW(commit_formats, commit_formats_len+1,
71 commit_formats_alloc);
72 commit_format = &commit_formats[commit_formats_len];
73 memset(commit_format, 0, sizeof(*commit_format));
74 commit_formats_len++;
77 commit_format->name = xstrdup(name);
78 commit_format->format = CMIT_FMT_USERFORMAT;
79 if (git_config_string(&fmt, var, value))
80 return -1;
82 if (skip_prefix(fmt, "format:", &fmt))
83 commit_format->is_tformat = 0;
84 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
85 commit_format->is_tformat = 1;
86 else
87 commit_format->is_alias = 1;
88 commit_format->user_format = fmt;
90 return 0;
93 static void setup_commit_formats(void)
95 struct cmt_fmt_map builtin_formats[] = {
96 { "raw", CMIT_FMT_RAW, 0, 0 },
97 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
98 { "short", CMIT_FMT_SHORT, 0, 0 },
99 { "email", CMIT_FMT_EMAIL, 0, 0 },
100 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
101 { "fuller", CMIT_FMT_FULLER, 0, 8 },
102 { "full", CMIT_FMT_FULL, 0, 8 },
103 { "oneline", CMIT_FMT_ONELINE, 1, 0 },
104 { "reference", CMIT_FMT_USERFORMAT, 1, 0,
105 0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
107 * Please update $__git_log_pretty_formats in
108 * git-completion.bash when you add new formats.
111 commit_formats_len = ARRAY_SIZE(builtin_formats);
112 builtin_formats_len = commit_formats_len;
113 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
114 COPY_ARRAY(commit_formats, builtin_formats,
115 ARRAY_SIZE(builtin_formats));
117 git_config(git_pretty_formats_config, NULL);
120 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
121 const char *original,
122 int num_redirections)
124 struct cmt_fmt_map *found = NULL;
125 size_t found_match_len = 0;
126 int i;
128 if (num_redirections >= commit_formats_len)
129 die("invalid --pretty format: "
130 "'%s' references an alias which points to itself",
131 original);
133 for (i = 0; i < commit_formats_len; i++) {
134 size_t match_len;
136 if (!starts_with(commit_formats[i].name, sought))
137 continue;
139 match_len = strlen(commit_formats[i].name);
140 if (found == NULL || found_match_len > match_len) {
141 found = &commit_formats[i];
142 found_match_len = match_len;
146 if (found && found->is_alias) {
147 found = find_commit_format_recursive(found->user_format,
148 original,
149 num_redirections+1);
152 return found;
155 static struct cmt_fmt_map *find_commit_format(const char *sought)
157 if (!commit_formats)
158 setup_commit_formats();
160 return find_commit_format_recursive(sought, sought, 0);
163 void get_commit_format(const char *arg, struct rev_info *rev)
165 struct cmt_fmt_map *commit_format;
167 rev->use_terminator = 0;
168 if (!arg) {
169 rev->commit_format = CMIT_FMT_DEFAULT;
170 return;
172 if (skip_prefix(arg, "format:", &arg)) {
173 save_user_format(rev, arg, 0);
174 return;
177 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
178 save_user_format(rev, arg, 1);
179 return;
182 commit_format = find_commit_format(arg);
183 if (!commit_format)
184 die("invalid --pretty format: %s", arg);
186 rev->commit_format = commit_format->format;
187 rev->use_terminator = commit_format->is_tformat;
188 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
189 if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
190 rev->date_mode.type = commit_format->default_date_mode_type;
191 if (commit_format->format == CMIT_FMT_USERFORMAT) {
192 save_user_format(rev, commit_format->user_format,
193 commit_format->is_tformat);
198 * Generic support for pretty-printing the header
200 static int get_one_line(const char *msg)
202 int ret = 0;
204 for (;;) {
205 char c = *msg++;
206 if (!c)
207 break;
208 ret++;
209 if (c == '\n')
210 break;
212 return ret;
215 /* High bit set, or ISO-2022-INT */
216 static int non_ascii(int ch)
218 return !isascii(ch) || ch == '\033';
221 int has_non_ascii(const char *s)
223 int ch;
224 if (!s)
225 return 0;
226 while ((ch = *s++) != '\0') {
227 if (non_ascii(ch))
228 return 1;
230 return 0;
233 static int is_rfc822_special(char ch)
235 switch (ch) {
236 case '(':
237 case ')':
238 case '<':
239 case '>':
240 case '[':
241 case ']':
242 case ':':
243 case ';':
244 case '@':
245 case ',':
246 case '.':
247 case '"':
248 case '\\':
249 return 1;
250 default:
251 return 0;
255 static int needs_rfc822_quoting(const char *s, int len)
257 int i;
258 for (i = 0; i < len; i++)
259 if (is_rfc822_special(s[i]))
260 return 1;
261 return 0;
264 static int last_line_length(struct strbuf *sb)
266 int i;
268 /* How many bytes are already used on the last line? */
269 for (i = sb->len - 1; i >= 0; i--)
270 if (sb->buf[i] == '\n')
271 break;
272 return sb->len - (i + 1);
275 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
277 int i;
279 /* just a guess, we may have to also backslash-quote */
280 strbuf_grow(out, len + 2);
282 strbuf_addch(out, '"');
283 for (i = 0; i < len; i++) {
284 switch (s[i]) {
285 case '"':
286 case '\\':
287 strbuf_addch(out, '\\');
288 /* fall through */
289 default:
290 strbuf_addch(out, s[i]);
293 strbuf_addch(out, '"');
296 enum rfc2047_type {
297 RFC2047_SUBJECT,
298 RFC2047_ADDRESS
301 static int is_rfc2047_special(char ch, enum rfc2047_type type)
304 * rfc2047, section 4.2:
306 * 8-bit values which correspond to printable ASCII characters other
307 * than "=", "?", and "_" (underscore), MAY be represented as those
308 * characters. (But see section 5 for restrictions.) In
309 * particular, SPACE and TAB MUST NOT be represented as themselves
310 * within encoded words.
314 * rule out non-ASCII characters and non-printable characters (the
315 * non-ASCII check should be redundant as isprint() is not localized
316 * and only knows about ASCII, but be defensive about that)
318 if (non_ascii(ch) || !isprint(ch))
319 return 1;
322 * rule out special printable characters (' ' should be the only
323 * whitespace character considered printable, but be defensive and use
324 * isspace())
326 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
327 return 1;
330 * rfc2047, section 5.3:
332 * As a replacement for a 'word' entity within a 'phrase', for example,
333 * one that precedes an address in a From, To, or Cc header. The ABNF
334 * definition for 'phrase' from RFC 822 thus becomes:
336 * phrase = 1*( encoded-word / word )
338 * In this case the set of characters that may be used in a "Q"-encoded
339 * 'encoded-word' is restricted to: <upper and lower case ASCII
340 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
341 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
342 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
343 * 'special' by 'linear-white-space'.
346 if (type != RFC2047_ADDRESS)
347 return 0;
349 /* '=' and '_' are special cases and have been checked above */
350 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
353 static int needs_rfc2047_encoding(const char *line, int len)
355 int i;
357 for (i = 0; i < len; i++) {
358 int ch = line[i];
359 if (non_ascii(ch) || ch == '\n')
360 return 1;
361 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
362 return 1;
365 return 0;
368 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
369 const char *encoding, enum rfc2047_type type)
371 static const int max_encoded_length = 76; /* per rfc2047 */
372 int i;
373 int line_len = last_line_length(sb);
375 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
376 strbuf_addf(sb, "=?%s?q?", encoding);
377 line_len += strlen(encoding) + 5; /* 5 for =??q? */
379 while (len) {
381 * RFC 2047, section 5 (3):
383 * Each 'encoded-word' MUST represent an integral number of
384 * characters. A multi-octet character may not be split across
385 * adjacent 'encoded- word's.
387 const unsigned char *p = (const unsigned char *)line;
388 int chrlen = mbs_chrlen(&line, &len, encoding);
389 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
391 /* "=%02X" * chrlen, or the byte itself */
392 const char *encoded_fmt = is_special ? "=%02X" : "%c";
393 int encoded_len = is_special ? 3 * chrlen : 1;
396 * According to RFC 2047, we could encode the special character
397 * ' ' (space) with '_' (underscore) for readability. But many
398 * programs do not understand this and just leave the
399 * underscore in place. Thus, we do nothing special here, which
400 * causes ' ' to be encoded as '=20', avoiding this problem.
403 if (line_len + encoded_len + 2 > max_encoded_length) {
404 /* It won't fit with trailing "?=" --- break the line */
405 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
406 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
409 for (i = 0; i < chrlen; i++)
410 strbuf_addf(sb, encoded_fmt, p[i]);
411 line_len += encoded_len;
413 strbuf_addstr(sb, "?=");
416 const char *show_ident_date(const struct ident_split *ident,
417 const struct date_mode *mode)
419 timestamp_t date = 0;
420 long tz = 0;
422 if (ident->date_begin && ident->date_end)
423 date = parse_timestamp(ident->date_begin, NULL, 10);
424 if (date_overflows(date))
425 date = 0;
426 else {
427 if (ident->tz_begin && ident->tz_end)
428 tz = strtol(ident->tz_begin, NULL, 10);
429 if (tz >= INT_MAX || tz <= INT_MIN)
430 tz = 0;
432 return show_date(date, tz, mode);
435 static inline void strbuf_add_with_color(struct strbuf *sb, const char *color,
436 const char *buf, size_t buflen)
438 strbuf_addstr(sb, color);
439 strbuf_add(sb, buf, buflen);
440 if (*color)
441 strbuf_addstr(sb, GIT_COLOR_RESET);
444 static void append_line_with_color(struct strbuf *sb, struct grep_opt *opt,
445 const char *line, size_t linelen,
446 int color, enum grep_context ctx,
447 enum grep_header_field field)
449 const char *buf, *eol, *line_color, *match_color;
450 regmatch_t match;
451 int eflags = 0;
453 buf = line;
454 eol = buf + linelen;
456 if (!opt || !want_color(color) || opt->invert)
457 goto end;
459 line_color = opt->colors[GREP_COLOR_SELECTED];
460 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
462 while (grep_next_match(opt, buf, eol, ctx, &match, field, eflags)) {
463 if (match.rm_so == match.rm_eo)
464 break;
466 strbuf_add_with_color(sb, line_color, buf, match.rm_so);
467 strbuf_add_with_color(sb, match_color, buf + match.rm_so,
468 match.rm_eo - match.rm_so);
469 buf += match.rm_eo;
470 eflags = REG_NOTBOL;
473 if (eflags)
474 strbuf_add_with_color(sb, line_color, buf, eol - buf);
475 else {
476 end:
477 strbuf_add(sb, buf, eol - buf);
481 static int use_in_body_from(const struct pretty_print_context *pp,
482 const struct ident_split *ident)
484 if (pp->rev && pp->rev->force_in_body_from)
485 return 1;
486 if (ident_cmp(pp->from_ident, ident))
487 return 1;
488 return 0;
491 void pp_user_info(struct pretty_print_context *pp,
492 const char *what, struct strbuf *sb,
493 const char *line, const char *encoding)
495 struct ident_split ident;
496 char *line_end;
497 const char *mailbuf, *namebuf;
498 size_t namelen, maillen;
499 int max_length = 78; /* per rfc2822 */
501 if (pp->fmt == CMIT_FMT_ONELINE)
502 return;
504 line_end = strchrnul(line, '\n');
505 if (split_ident_line(&ident, line, line_end - line))
506 return;
508 mailbuf = ident.mail_begin;
509 maillen = ident.mail_end - ident.mail_begin;
510 namebuf = ident.name_begin;
511 namelen = ident.name_end - ident.name_begin;
513 if (pp->mailmap)
514 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
516 if (cmit_fmt_is_mail(pp->fmt)) {
517 if (pp->from_ident && use_in_body_from(pp, &ident)) {
518 struct strbuf buf = STRBUF_INIT;
520 strbuf_addstr(&buf, "From: ");
521 strbuf_add(&buf, namebuf, namelen);
522 strbuf_addstr(&buf, " <");
523 strbuf_add(&buf, mailbuf, maillen);
524 strbuf_addstr(&buf, ">\n");
525 string_list_append(&pp->in_body_headers,
526 strbuf_detach(&buf, NULL));
528 mailbuf = pp->from_ident->mail_begin;
529 maillen = pp->from_ident->mail_end - mailbuf;
530 namebuf = pp->from_ident->name_begin;
531 namelen = pp->from_ident->name_end - namebuf;
534 strbuf_addstr(sb, "From: ");
535 if (pp->encode_email_headers &&
536 needs_rfc2047_encoding(namebuf, namelen)) {
537 add_rfc2047(sb, namebuf, namelen,
538 encoding, RFC2047_ADDRESS);
539 max_length = 76; /* per rfc2047 */
540 } else if (needs_rfc822_quoting(namebuf, namelen)) {
541 struct strbuf quoted = STRBUF_INIT;
542 add_rfc822_quoted(&quoted, namebuf, namelen);
543 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
544 -6, 1, max_length);
545 strbuf_release(&quoted);
546 } else {
547 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
548 -6, 1, max_length);
551 if (max_length <
552 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
553 strbuf_addch(sb, '\n');
554 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
555 } else {
556 struct strbuf id = STRBUF_INIT;
557 enum grep_header_field field = GREP_HEADER_FIELD_MAX;
558 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
560 if (!strcmp(what, "Author"))
561 field = GREP_HEADER_AUTHOR;
562 else if (!strcmp(what, "Commit"))
563 field = GREP_HEADER_COMMITTER;
565 strbuf_addf(sb, "%s: ", what);
566 if (pp->fmt == CMIT_FMT_FULLER)
567 strbuf_addchars(sb, ' ', 4);
569 strbuf_addf(&id, "%.*s <%.*s>", (int)namelen, namebuf,
570 (int)maillen, mailbuf);
572 append_line_with_color(sb, opt, id.buf, id.len, pp->color,
573 GREP_CONTEXT_HEAD, field);
574 strbuf_addch(sb, '\n');
575 strbuf_release(&id);
578 switch (pp->fmt) {
579 case CMIT_FMT_MEDIUM:
580 strbuf_addf(sb, "Date: %s\n",
581 show_ident_date(&ident, &pp->date_mode));
582 break;
583 case CMIT_FMT_EMAIL:
584 case CMIT_FMT_MBOXRD:
585 strbuf_addf(sb, "Date: %s\n",
586 show_ident_date(&ident, DATE_MODE(RFC2822)));
587 break;
588 case CMIT_FMT_FULLER:
589 strbuf_addf(sb, "%sDate: %s\n", what,
590 show_ident_date(&ident, &pp->date_mode));
591 break;
592 default:
593 /* notin' */
594 break;
598 static int is_blank_line(const char *line, int *len_p)
600 int len = *len_p;
601 while (len && isspace(line[len - 1]))
602 len--;
603 *len_p = len;
604 return !len;
607 const char *skip_blank_lines(const char *msg)
609 for (;;) {
610 int linelen = get_one_line(msg);
611 int ll = linelen;
612 if (!linelen)
613 break;
614 if (!is_blank_line(msg, &ll))
615 break;
616 msg += linelen;
618 return msg;
621 static void add_merge_info(const struct pretty_print_context *pp,
622 struct strbuf *sb, const struct commit *commit)
624 struct commit_list *parent = commit->parents;
626 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
627 !parent || !parent->next)
628 return;
630 strbuf_addstr(sb, "Merge:");
632 while (parent) {
633 struct object_id *oidp = &parent->item->object.oid;
634 strbuf_addch(sb, ' ');
635 if (pp->abbrev)
636 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
637 else
638 strbuf_addstr(sb, oid_to_hex(oidp));
639 parent = parent->next;
641 strbuf_addch(sb, '\n');
644 static char *get_header(const char *msg, const char *key)
646 size_t len;
647 const char *v = find_commit_header(msg, key, &len);
648 return v ? xmemdupz(v, len) : NULL;
651 static char *replace_encoding_header(char *buf, const char *encoding)
653 struct strbuf tmp = STRBUF_INIT;
654 size_t start, len;
655 char *cp = buf;
657 /* guess if there is an encoding header before a \n\n */
658 while (!starts_with(cp, "encoding ")) {
659 cp = strchr(cp, '\n');
660 if (!cp || *++cp == '\n')
661 return buf;
663 start = cp - buf;
664 cp = strchr(cp, '\n');
665 if (!cp)
666 return buf; /* should not happen but be defensive */
667 len = cp + 1 - (buf + start);
669 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
670 if (is_encoding_utf8(encoding)) {
671 /* we have re-coded to UTF-8; drop the header */
672 strbuf_remove(&tmp, start, len);
673 } else {
674 /* just replaces XXXX in 'encoding XXXX\n' */
675 strbuf_splice(&tmp, start + strlen("encoding "),
676 len - strlen("encoding \n"),
677 encoding, strlen(encoding));
679 return strbuf_detach(&tmp, NULL);
682 const char *repo_logmsg_reencode(struct repository *r,
683 const struct commit *commit,
684 char **commit_encoding,
685 const char *output_encoding)
687 static const char *utf8 = "UTF-8";
688 const char *use_encoding;
689 char *encoding;
690 const char *msg = repo_get_commit_buffer(r, commit, NULL);
691 char *out;
693 if (!output_encoding || !*output_encoding) {
694 if (commit_encoding)
695 *commit_encoding = get_header(msg, "encoding");
696 return msg;
698 encoding = get_header(msg, "encoding");
699 if (commit_encoding)
700 *commit_encoding = encoding;
701 use_encoding = encoding ? encoding : utf8;
702 if (same_encoding(use_encoding, output_encoding)) {
704 * No encoding work to be done. If we have no encoding header
705 * at all, then there's nothing to do, and we can return the
706 * message verbatim (whether newly allocated or not).
708 if (!encoding)
709 return msg;
712 * Otherwise, we still want to munge the encoding header in the
713 * result, which will be done by modifying the buffer. If we
714 * are using a fresh copy, we can reuse it. But if we are using
715 * the cached copy from get_commit_buffer, we need to duplicate it
716 * to avoid munging the cached copy.
718 if (msg == get_cached_commit_buffer(r, commit, NULL))
719 out = xstrdup(msg);
720 else
721 out = (char *)msg;
723 else {
725 * There's actual encoding work to do. Do the reencoding, which
726 * still leaves the header to be replaced in the next step. At
727 * this point, we are done with msg. If we allocated a fresh
728 * copy, we can free it.
730 out = reencode_string(msg, output_encoding, use_encoding);
731 if (out)
732 repo_unuse_commit_buffer(r, commit, msg);
736 * This replacement actually consumes the buffer we hand it, so we do
737 * not have to worry about freeing the old "out" here.
739 if (out)
740 out = replace_encoding_header(out, output_encoding);
742 if (!commit_encoding)
743 free(encoding);
745 * If the re-encoding failed, out might be NULL here; in that
746 * case we just return the commit message verbatim.
748 return out ? out : msg;
751 static int mailmap_name(const char **email, size_t *email_len,
752 const char **name, size_t *name_len)
754 static struct string_list *mail_map;
755 if (!mail_map) {
756 CALLOC_ARRAY(mail_map, 1);
757 read_mailmap(mail_map);
759 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
762 static size_t format_person_part(struct strbuf *sb, char part,
763 const char *msg, int len,
764 const struct date_mode *dmode)
766 /* currently all placeholders have same length */
767 const int placeholder_len = 2;
768 struct ident_split s;
769 const char *name, *mail;
770 size_t maillen, namelen;
772 if (split_ident_line(&s, msg, len) < 0)
773 goto skip;
775 name = s.name_begin;
776 namelen = s.name_end - s.name_begin;
777 mail = s.mail_begin;
778 maillen = s.mail_end - s.mail_begin;
780 if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
781 mailmap_name(&mail, &maillen, &name, &namelen);
782 if (part == 'n' || part == 'N') { /* name */
783 strbuf_add(sb, name, namelen);
784 return placeholder_len;
786 if (part == 'e' || part == 'E') { /* email */
787 strbuf_add(sb, mail, maillen);
788 return placeholder_len;
790 if (part == 'l' || part == 'L') { /* local-part */
791 const char *at = memchr(mail, '@', maillen);
792 if (at)
793 maillen = at - mail;
794 strbuf_add(sb, mail, maillen);
795 return placeholder_len;
798 if (!s.date_begin)
799 goto skip;
801 if (part == 't') { /* date, UNIX timestamp */
802 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
803 return placeholder_len;
806 switch (part) {
807 case 'd': /* date */
808 strbuf_addstr(sb, show_ident_date(&s, dmode));
809 return placeholder_len;
810 case 'D': /* date, RFC2822 style */
811 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
812 return placeholder_len;
813 case 'r': /* date, relative */
814 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
815 return placeholder_len;
816 case 'i': /* date, ISO 8601-like */
817 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
818 return placeholder_len;
819 case 'I': /* date, ISO 8601 strict */
820 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
821 return placeholder_len;
822 case 'h': /* date, human */
823 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(HUMAN)));
824 return placeholder_len;
825 case 's':
826 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
827 return placeholder_len;
830 skip:
832 * reading from either a bogus commit, or a reflog entry with
833 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
834 * to compute a valid return value.
836 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
837 || part == 'D' || part == 'r' || part == 'i')
838 return placeholder_len;
840 return 0; /* unknown placeholder */
843 struct chunk {
844 size_t off;
845 size_t len;
848 enum flush_type {
849 no_flush,
850 flush_right,
851 flush_left,
852 flush_left_and_steal,
853 flush_both
856 enum trunc_type {
857 trunc_none,
858 trunc_left,
859 trunc_middle,
860 trunc_right
863 struct format_commit_context {
864 struct repository *repository;
865 const struct commit *commit;
866 const struct pretty_print_context *pretty_ctx;
867 unsigned commit_header_parsed:1;
868 unsigned commit_message_parsed:1;
869 struct signature_check signature_check;
870 enum flush_type flush_type;
871 enum trunc_type truncate;
872 const char *message;
873 char *commit_encoding;
874 size_t width, indent1, indent2;
875 int auto_color;
876 int padding;
878 /* These offsets are relative to the start of the commit message. */
879 struct chunk author;
880 struct chunk committer;
881 size_t message_off;
882 size_t subject_off;
883 size_t body_off;
885 /* The following ones are relative to the result struct strbuf. */
886 size_t wrap_start;
889 static void parse_commit_header(struct format_commit_context *context)
891 const char *msg = context->message;
892 int i;
894 for (i = 0; msg[i]; i++) {
895 const char *name;
896 int eol;
897 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
898 ; /* do nothing */
900 if (i == eol) {
901 break;
902 } else if (skip_prefix(msg + i, "author ", &name)) {
903 context->author.off = name - msg;
904 context->author.len = msg + eol - name;
905 } else if (skip_prefix(msg + i, "committer ", &name)) {
906 context->committer.off = name - msg;
907 context->committer.len = msg + eol - name;
909 i = eol;
911 context->message_off = i;
912 context->commit_header_parsed = 1;
915 static int istitlechar(char c)
917 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
918 (c >= '0' && c <= '9') || c == '.' || c == '_';
921 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
923 size_t trimlen;
924 size_t start_len = sb->len;
925 int space = 2;
926 int i;
928 for (i = 0; i < len; i++) {
929 if (istitlechar(msg[i])) {
930 if (space == 1)
931 strbuf_addch(sb, '-');
932 space = 0;
933 strbuf_addch(sb, msg[i]);
934 if (msg[i] == '.')
935 while (msg[i+1] == '.')
936 i++;
937 } else
938 space |= 1;
941 /* trim any trailing '.' or '-' characters */
942 trimlen = 0;
943 while (sb->len - trimlen > start_len &&
944 (sb->buf[sb->len - 1 - trimlen] == '.'
945 || sb->buf[sb->len - 1 - trimlen] == '-'))
946 trimlen++;
947 strbuf_remove(sb, sb->len - trimlen, trimlen);
950 const char *format_subject(struct strbuf *sb, const char *msg,
951 const char *line_separator)
953 int first = 1;
955 for (;;) {
956 const char *line = msg;
957 int linelen = get_one_line(line);
959 msg += linelen;
960 if (!linelen || is_blank_line(line, &linelen))
961 break;
963 if (!sb)
964 continue;
965 strbuf_grow(sb, linelen + 2);
966 if (!first)
967 strbuf_addstr(sb, line_separator);
968 strbuf_add(sb, line, linelen);
969 first = 0;
971 return msg;
974 static void parse_commit_message(struct format_commit_context *c)
976 const char *msg = c->message + c->message_off;
977 const char *start = c->message;
979 msg = skip_blank_lines(msg);
980 c->subject_off = msg - start;
982 msg = format_subject(NULL, msg, NULL);
983 msg = skip_blank_lines(msg);
984 c->body_off = msg - start;
986 c->commit_message_parsed = 1;
989 static void strbuf_wrap(struct strbuf *sb, size_t pos,
990 size_t width, size_t indent1, size_t indent2)
992 struct strbuf tmp = STRBUF_INIT;
994 if (pos)
995 strbuf_add(&tmp, sb->buf, pos);
996 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
997 (int) indent1, (int) indent2, (int) width);
998 strbuf_swap(&tmp, sb);
999 strbuf_release(&tmp);
1002 static void rewrap_message_tail(struct strbuf *sb,
1003 struct format_commit_context *c,
1004 size_t new_width, size_t new_indent1,
1005 size_t new_indent2)
1007 if (c->width == new_width && c->indent1 == new_indent1 &&
1008 c->indent2 == new_indent2)
1009 return;
1010 if (c->wrap_start < sb->len)
1011 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
1012 c->wrap_start = sb->len;
1013 c->width = new_width;
1014 c->indent1 = new_indent1;
1015 c->indent2 = new_indent2;
1018 static int format_reflog_person(struct strbuf *sb,
1019 char part,
1020 struct reflog_walk_info *log,
1021 const struct date_mode *dmode)
1023 const char *ident;
1025 if (!log)
1026 return 2;
1028 ident = get_reflog_ident(log);
1029 if (!ident)
1030 return 2;
1032 return format_person_part(sb, part, ident, strlen(ident), dmode);
1035 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
1036 const char *placeholder,
1037 struct format_commit_context *c)
1039 const char *rest = placeholder;
1040 const char *basic_color = NULL;
1042 if (placeholder[1] == '(') {
1043 const char *begin = placeholder + 2;
1044 const char *end = strchr(begin, ')');
1045 char color[COLOR_MAXLEN];
1047 if (!end)
1048 return 0;
1050 if (skip_prefix(begin, "auto,", &begin)) {
1051 if (!want_color(c->pretty_ctx->color))
1052 return end - placeholder + 1;
1053 } else if (skip_prefix(begin, "always,", &begin)) {
1054 /* nothing to do; we do not respect want_color at all */
1055 } else {
1056 /* the default is the same as "auto" */
1057 if (!want_color(c->pretty_ctx->color))
1058 return end - placeholder + 1;
1061 if (color_parse_mem(begin, end - begin, color) < 0)
1062 die(_("unable to parse --pretty format"));
1063 strbuf_addstr(sb, color);
1064 return end - placeholder + 1;
1068 * We handle things like "%C(red)" above; for historical reasons, there
1069 * are a few colors that can be specified without parentheses (and
1070 * they cannot support things like "auto" or "always" at all).
1072 if (skip_prefix(placeholder + 1, "red", &rest))
1073 basic_color = GIT_COLOR_RED;
1074 else if (skip_prefix(placeholder + 1, "green", &rest))
1075 basic_color = GIT_COLOR_GREEN;
1076 else if (skip_prefix(placeholder + 1, "blue", &rest))
1077 basic_color = GIT_COLOR_BLUE;
1078 else if (skip_prefix(placeholder + 1, "reset", &rest))
1079 basic_color = GIT_COLOR_RESET;
1081 if (basic_color && want_color(c->pretty_ctx->color))
1082 strbuf_addstr(sb, basic_color);
1084 return rest - placeholder;
1087 static size_t parse_padding_placeholder(const char *placeholder,
1088 struct format_commit_context *c)
1090 const char *ch = placeholder;
1091 enum flush_type flush_type;
1092 int to_column = 0;
1094 switch (*ch++) {
1095 case '<':
1096 flush_type = flush_right;
1097 break;
1098 case '>':
1099 if (*ch == '<') {
1100 flush_type = flush_both;
1101 ch++;
1102 } else if (*ch == '>') {
1103 flush_type = flush_left_and_steal;
1104 ch++;
1105 } else
1106 flush_type = flush_left;
1107 break;
1108 default:
1109 return 0;
1112 /* the next value means "wide enough to that column" */
1113 if (*ch == '|') {
1114 to_column = 1;
1115 ch++;
1118 if (*ch == '(') {
1119 const char *start = ch + 1;
1120 const char *end = start + strcspn(start, ",)");
1121 char *next;
1122 int width;
1123 if (!end || end == start)
1124 return 0;
1125 width = strtol(start, &next, 10);
1126 if (next == start || width == 0)
1127 return 0;
1128 if (width < 0) {
1129 if (to_column)
1130 width += term_columns();
1131 if (width < 0)
1132 return 0;
1134 c->padding = to_column ? -width : width;
1135 c->flush_type = flush_type;
1137 if (*end == ',') {
1138 start = end + 1;
1139 end = strchr(start, ')');
1140 if (!end || end == start)
1141 return 0;
1142 if (starts_with(start, "trunc)"))
1143 c->truncate = trunc_right;
1144 else if (starts_with(start, "ltrunc)"))
1145 c->truncate = trunc_left;
1146 else if (starts_with(start, "mtrunc)"))
1147 c->truncate = trunc_middle;
1148 else
1149 return 0;
1150 } else
1151 c->truncate = trunc_none;
1153 return end - placeholder + 1;
1155 return 0;
1158 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1159 const char **end, const char **valuestart,
1160 size_t *valuelen)
1162 const char *p;
1164 if (!(skip_prefix(to_parse, candidate, &p)))
1165 return 0;
1166 if (valuestart) {
1167 if (*p == '=') {
1168 *valuestart = p + 1;
1169 *valuelen = strcspn(*valuestart, ",)");
1170 p = *valuestart + *valuelen;
1171 } else {
1172 if (*p != ',' && *p != ')')
1173 return 0;
1174 *valuestart = NULL;
1175 *valuelen = 0;
1178 if (*p == ',') {
1179 *end = p + 1;
1180 return 1;
1182 if (*p == ')') {
1183 *end = p;
1184 return 1;
1186 return 0;
1189 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1190 const char **end, int *val)
1192 const char *argval;
1193 char *strval;
1194 size_t arglen;
1195 int v;
1197 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1198 return 0;
1200 if (!argval) {
1201 *val = 1;
1202 return 1;
1205 strval = xstrndup(argval, arglen);
1206 v = git_parse_maybe_bool(strval);
1207 free(strval);
1209 if (v == -1)
1210 return 0;
1212 *val = v;
1214 return 1;
1217 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1219 const struct string_list *list = ud;
1220 const struct string_list_item *item;
1222 for_each_string_list_item (item, list) {
1223 if (key->len == (uintptr_t)item->util &&
1224 !strncasecmp(item->string, key->buf, key->len))
1225 return 1;
1227 return 0;
1230 int format_set_trailers_options(struct process_trailer_options *opts,
1231 struct string_list *filter_list,
1232 struct strbuf *sepbuf,
1233 struct strbuf *kvsepbuf,
1234 const char **arg,
1235 char **invalid_arg)
1237 for (;;) {
1238 const char *argval;
1239 size_t arglen;
1241 if (**arg == ')')
1242 break;
1244 if (match_placeholder_arg_value(*arg, "key", arg, &argval, &arglen)) {
1245 uintptr_t len = arglen;
1247 if (!argval)
1248 return -1;
1250 if (len && argval[len - 1] == ':')
1251 len--;
1252 string_list_append(filter_list, argval)->util = (char *)len;
1254 opts->filter = format_trailer_match_cb;
1255 opts->filter_data = filter_list;
1256 opts->only_trailers = 1;
1257 } else if (match_placeholder_arg_value(*arg, "separator", arg, &argval, &arglen)) {
1258 char *fmt;
1260 strbuf_reset(sepbuf);
1261 fmt = xstrndup(argval, arglen);
1262 strbuf_expand(sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1263 free(fmt);
1264 opts->separator = sepbuf;
1265 } else if (match_placeholder_arg_value(*arg, "key_value_separator", arg, &argval, &arglen)) {
1266 char *fmt;
1268 strbuf_reset(kvsepbuf);
1269 fmt = xstrndup(argval, arglen);
1270 strbuf_expand(kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1271 free(fmt);
1272 opts->key_value_separator = kvsepbuf;
1273 } else if (!match_placeholder_bool_arg(*arg, "only", arg, &opts->only_trailers) &&
1274 !match_placeholder_bool_arg(*arg, "unfold", arg, &opts->unfold) &&
1275 !match_placeholder_bool_arg(*arg, "keyonly", arg, &opts->key_only) &&
1276 !match_placeholder_bool_arg(*arg, "valueonly", arg, &opts->value_only)) {
1277 if (invalid_arg) {
1278 size_t len = strcspn(*arg, ",)");
1279 *invalid_arg = xstrndup(*arg, len);
1281 return -1;
1284 return 0;
1287 static size_t parse_describe_args(const char *start, struct strvec *args)
1289 struct {
1290 char *name;
1291 enum {
1292 DESCRIBE_ARG_BOOL,
1293 DESCRIBE_ARG_INTEGER,
1294 DESCRIBE_ARG_STRING,
1295 } type;
1296 } option[] = {
1297 { "tags", DESCRIBE_ARG_BOOL},
1298 { "abbrev", DESCRIBE_ARG_INTEGER },
1299 { "exclude", DESCRIBE_ARG_STRING },
1300 { "match", DESCRIBE_ARG_STRING },
1302 const char *arg = start;
1304 for (;;) {
1305 int found = 0;
1306 const char *argval;
1307 size_t arglen = 0;
1308 int optval = 0;
1309 int i;
1311 for (i = 0; !found && i < ARRAY_SIZE(option); i++) {
1312 switch (option[i].type) {
1313 case DESCRIBE_ARG_BOOL:
1314 if (match_placeholder_bool_arg(arg, option[i].name, &arg, &optval)) {
1315 if (optval)
1316 strvec_pushf(args, "--%s", option[i].name);
1317 else
1318 strvec_pushf(args, "--no-%s", option[i].name);
1319 found = 1;
1321 break;
1322 case DESCRIBE_ARG_INTEGER:
1323 if (match_placeholder_arg_value(arg, option[i].name, &arg,
1324 &argval, &arglen)) {
1325 char *endptr;
1326 if (!arglen)
1327 return 0;
1328 strtol(argval, &endptr, 10);
1329 if (endptr - argval != arglen)
1330 return 0;
1331 strvec_pushf(args, "--%s=%.*s", option[i].name, (int)arglen, argval);
1332 found = 1;
1334 break;
1335 case DESCRIBE_ARG_STRING:
1336 if (match_placeholder_arg_value(arg, option[i].name, &arg,
1337 &argval, &arglen)) {
1338 if (!arglen)
1339 return 0;
1340 strvec_pushf(args, "--%s=%.*s", option[i].name, (int)arglen, argval);
1341 found = 1;
1343 break;
1346 if (!found)
1347 break;
1350 return arg - start;
1353 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1354 const char *placeholder,
1355 void *context)
1357 struct format_commit_context *c = context;
1358 const struct commit *commit = c->commit;
1359 const char *msg = c->message;
1360 struct commit_list *p;
1361 const char *arg, *eol;
1362 size_t res;
1363 char **slot;
1365 /* these are independent of the commit */
1366 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1367 if (res)
1368 return res;
1370 switch (placeholder[0]) {
1371 case 'C':
1372 if (starts_with(placeholder + 1, "(auto)")) {
1373 c->auto_color = want_color(c->pretty_ctx->color);
1374 if (c->auto_color && sb->len)
1375 strbuf_addstr(sb, GIT_COLOR_RESET);
1376 return 7; /* consumed 7 bytes, "C(auto)" */
1377 } else {
1378 int ret = parse_color(sb, placeholder, c);
1379 if (ret)
1380 c->auto_color = 0;
1382 * Otherwise, we decided to treat %C<unknown>
1383 * as a literal string, and the previous
1384 * %C(auto) is still valid.
1386 return ret;
1388 case 'w':
1389 if (placeholder[1] == '(') {
1390 unsigned long width = 0, indent1 = 0, indent2 = 0;
1391 char *next;
1392 const char *start = placeholder + 2;
1393 const char *end = strchr(start, ')');
1394 if (!end)
1395 return 0;
1396 if (end > start) {
1397 width = strtoul(start, &next, 10);
1398 if (*next == ',') {
1399 indent1 = strtoul(next + 1, &next, 10);
1400 if (*next == ',') {
1401 indent2 = strtoul(next + 1,
1402 &next, 10);
1405 if (*next != ')')
1406 return 0;
1408 rewrap_message_tail(sb, c, width, indent1, indent2);
1409 return end - placeholder + 1;
1410 } else
1411 return 0;
1413 case '<':
1414 case '>':
1415 return parse_padding_placeholder(placeholder, c);
1418 if (skip_prefix(placeholder, "(describe", &arg)) {
1419 struct child_process cmd = CHILD_PROCESS_INIT;
1420 struct strbuf out = STRBUF_INIT;
1421 struct strbuf err = STRBUF_INIT;
1422 struct pretty_print_describe_status *describe_status;
1424 describe_status = c->pretty_ctx->describe_status;
1425 if (describe_status) {
1426 if (!describe_status->max_invocations)
1427 return 0;
1428 describe_status->max_invocations--;
1431 cmd.git_cmd = 1;
1432 strvec_push(&cmd.args, "describe");
1434 if (*arg == ':') {
1435 arg++;
1436 arg += parse_describe_args(arg, &cmd.args);
1439 if (*arg != ')') {
1440 child_process_clear(&cmd);
1441 return 0;
1444 strvec_push(&cmd.args, oid_to_hex(&commit->object.oid));
1445 pipe_command(&cmd, NULL, 0, &out, 0, &err, 0);
1446 strbuf_rtrim(&out);
1447 strbuf_addbuf(sb, &out);
1448 strbuf_release(&out);
1449 strbuf_release(&err);
1450 return arg - placeholder + 1;
1453 /* these depend on the commit */
1454 if (!commit->object.parsed)
1455 parse_object(the_repository, &commit->object.oid);
1457 switch (placeholder[0]) {
1458 case 'H': /* commit hash */
1459 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1460 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1461 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1462 return 1;
1463 case 'h': /* abbreviated commit hash */
1464 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1465 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1466 c->pretty_ctx->abbrev);
1467 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1468 return 1;
1469 case 'T': /* tree hash */
1470 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1471 return 1;
1472 case 't': /* abbreviated tree hash */
1473 strbuf_add_unique_abbrev(sb,
1474 get_commit_tree_oid(commit),
1475 c->pretty_ctx->abbrev);
1476 return 1;
1477 case 'P': /* parent hashes */
1478 for (p = commit->parents; p; p = p->next) {
1479 if (p != commit->parents)
1480 strbuf_addch(sb, ' ');
1481 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1483 return 1;
1484 case 'p': /* abbreviated parent hashes */
1485 for (p = commit->parents; p; p = p->next) {
1486 if (p != commit->parents)
1487 strbuf_addch(sb, ' ');
1488 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1489 c->pretty_ctx->abbrev);
1491 return 1;
1492 case 'm': /* left/right/bottom */
1493 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1494 return 1;
1495 case 'd':
1496 format_decorations(sb, commit, c->auto_color);
1497 return 1;
1498 case 'D':
1499 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1500 return 1;
1501 case 'S': /* tag/branch like --source */
1502 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1503 return 0;
1504 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1505 if (!(slot && *slot))
1506 return 0;
1507 strbuf_addstr(sb, *slot);
1508 return 1;
1509 case 'g': /* reflog info */
1510 switch(placeholder[1]) {
1511 case 'd': /* reflog selector */
1512 case 'D':
1513 if (c->pretty_ctx->reflog_info)
1514 get_reflog_selector(sb,
1515 c->pretty_ctx->reflog_info,
1516 &c->pretty_ctx->date_mode,
1517 c->pretty_ctx->date_mode_explicit,
1518 (placeholder[1] == 'd'));
1519 return 2;
1520 case 's': /* reflog message */
1521 if (c->pretty_ctx->reflog_info)
1522 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1523 return 2;
1524 case 'n':
1525 case 'N':
1526 case 'e':
1527 case 'E':
1528 return format_reflog_person(sb,
1529 placeholder[1],
1530 c->pretty_ctx->reflog_info,
1531 &c->pretty_ctx->date_mode);
1533 return 0; /* unknown %g placeholder */
1534 case 'N':
1535 if (c->pretty_ctx->notes_message) {
1536 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1537 return 1;
1539 return 0;
1542 if (placeholder[0] == 'G') {
1543 if (!c->signature_check.result)
1544 check_commit_signature(c->commit, &(c->signature_check));
1545 switch (placeholder[1]) {
1546 case 'G':
1547 if (c->signature_check.output)
1548 strbuf_addstr(sb, c->signature_check.output);
1549 break;
1550 case '?':
1551 switch (c->signature_check.result) {
1552 case 'G':
1553 switch (c->signature_check.trust_level) {
1554 case TRUST_UNDEFINED:
1555 case TRUST_NEVER:
1556 strbuf_addch(sb, 'U');
1557 break;
1558 default:
1559 strbuf_addch(sb, 'G');
1560 break;
1562 break;
1563 case 'B':
1564 case 'E':
1565 case 'N':
1566 case 'X':
1567 case 'Y':
1568 case 'R':
1569 strbuf_addch(sb, c->signature_check.result);
1571 break;
1572 case 'S':
1573 if (c->signature_check.signer)
1574 strbuf_addstr(sb, c->signature_check.signer);
1575 break;
1576 case 'K':
1577 if (c->signature_check.key)
1578 strbuf_addstr(sb, c->signature_check.key);
1579 break;
1580 case 'F':
1581 if (c->signature_check.fingerprint)
1582 strbuf_addstr(sb, c->signature_check.fingerprint);
1583 break;
1584 case 'P':
1585 if (c->signature_check.primary_key_fingerprint)
1586 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1587 break;
1588 case 'T':
1589 strbuf_addstr(sb, gpg_trust_level_to_str(c->signature_check.trust_level));
1590 break;
1591 default:
1592 return 0;
1594 return 2;
1597 /* For the rest we have to parse the commit header. */
1598 if (!c->commit_header_parsed) {
1599 msg = c->message =
1600 repo_logmsg_reencode(c->repository, commit,
1601 &c->commit_encoding, "UTF-8");
1602 parse_commit_header(c);
1605 switch (placeholder[0]) {
1606 case 'a': /* author ... */
1607 return format_person_part(sb, placeholder[1],
1608 msg + c->author.off, c->author.len,
1609 &c->pretty_ctx->date_mode);
1610 case 'c': /* committer ... */
1611 return format_person_part(sb, placeholder[1],
1612 msg + c->committer.off, c->committer.len,
1613 &c->pretty_ctx->date_mode);
1614 case 'e': /* encoding */
1615 if (c->commit_encoding)
1616 strbuf_addstr(sb, c->commit_encoding);
1617 return 1;
1618 case 'B': /* raw body */
1619 /* message_off is always left at the initial newline */
1620 strbuf_addstr(sb, msg + c->message_off + 1);
1621 return 1;
1624 /* Now we need to parse the commit message. */
1625 if (!c->commit_message_parsed)
1626 parse_commit_message(c);
1628 switch (placeholder[0]) {
1629 case 's': /* subject */
1630 format_subject(sb, msg + c->subject_off, " ");
1631 return 1;
1632 case 'f': /* sanitized subject */
1633 eol = strchrnul(msg + c->subject_off, '\n');
1634 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1635 return 1;
1636 case 'b': /* body */
1637 strbuf_addstr(sb, msg + c->body_off);
1638 return 1;
1641 if (skip_prefix(placeholder, "(trailers", &arg)) {
1642 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1643 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1644 struct strbuf sepbuf = STRBUF_INIT;
1645 struct strbuf kvsepbuf = STRBUF_INIT;
1646 size_t ret = 0;
1648 opts.no_divider = 1;
1650 if (*arg == ':') {
1651 arg++;
1652 if (format_set_trailers_options(&opts, &filter_list, &sepbuf, &kvsepbuf, &arg, NULL))
1653 goto trailer_out;
1655 if (*arg == ')') {
1656 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1657 ret = arg - placeholder + 1;
1659 trailer_out:
1660 string_list_clear(&filter_list, 0);
1661 strbuf_release(&sepbuf);
1662 return ret;
1665 return 0; /* unknown placeholder */
1668 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1669 const char *placeholder,
1670 struct format_commit_context *c)
1672 struct strbuf local_sb = STRBUF_INIT;
1673 int total_consumed = 0, len, padding = c->padding;
1674 if (padding < 0) {
1675 const char *start = strrchr(sb->buf, '\n');
1676 int occupied;
1677 if (!start)
1678 start = sb->buf;
1679 occupied = utf8_strnwidth(start, -1, 1);
1680 occupied += c->pretty_ctx->graph_width;
1681 padding = (-padding) - occupied;
1683 while (1) {
1684 int modifier = *placeholder == 'C';
1685 int consumed = format_commit_one(&local_sb, placeholder, c);
1686 total_consumed += consumed;
1688 if (!modifier)
1689 break;
1691 placeholder += consumed;
1692 if (*placeholder != '%')
1693 break;
1694 placeholder++;
1695 total_consumed++;
1697 len = utf8_strnwidth(local_sb.buf, -1, 1);
1699 if (c->flush_type == flush_left_and_steal) {
1700 const char *ch = sb->buf + sb->len - 1;
1701 while (len > padding && ch > sb->buf) {
1702 const char *p;
1703 if (*ch == ' ') {
1704 ch--;
1705 padding++;
1706 continue;
1708 /* check for trailing ansi sequences */
1709 if (*ch != 'm')
1710 break;
1711 p = ch - 1;
1712 while (ch - p < 10 && *p != '\033')
1713 p--;
1714 if (*p != '\033' ||
1715 ch + 1 - p != display_mode_esc_sequence_len(p))
1716 break;
1718 * got a good ansi sequence, put it back to
1719 * local_sb as we're cutting sb
1721 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1722 ch = p - 1;
1724 strbuf_setlen(sb, ch + 1 - sb->buf);
1725 c->flush_type = flush_left;
1728 if (len > padding) {
1729 switch (c->truncate) {
1730 case trunc_left:
1731 strbuf_utf8_replace(&local_sb,
1732 0, len - (padding - 2),
1733 "..");
1734 break;
1735 case trunc_middle:
1736 strbuf_utf8_replace(&local_sb,
1737 padding / 2 - 1,
1738 len - (padding - 2),
1739 "..");
1740 break;
1741 case trunc_right:
1742 strbuf_utf8_replace(&local_sb,
1743 padding - 2, len - (padding - 2),
1744 "..");
1745 break;
1746 case trunc_none:
1747 break;
1749 strbuf_addbuf(sb, &local_sb);
1750 } else {
1751 int sb_len = sb->len, offset = 0;
1752 if (c->flush_type == flush_left)
1753 offset = padding - len;
1754 else if (c->flush_type == flush_both)
1755 offset = (padding - len) / 2;
1757 * we calculate padding in columns, now
1758 * convert it back to chars
1760 padding = padding - len + local_sb.len;
1761 strbuf_addchars(sb, ' ', padding);
1762 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1763 local_sb.len);
1765 strbuf_release(&local_sb);
1766 c->flush_type = no_flush;
1767 return total_consumed;
1770 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1771 const char *placeholder,
1772 void *context)
1774 int consumed;
1775 size_t orig_len;
1776 enum {
1777 NO_MAGIC,
1778 ADD_LF_BEFORE_NON_EMPTY,
1779 DEL_LF_BEFORE_EMPTY,
1780 ADD_SP_BEFORE_NON_EMPTY
1781 } magic = NO_MAGIC;
1783 switch (placeholder[0]) {
1784 case '-':
1785 magic = DEL_LF_BEFORE_EMPTY;
1786 break;
1787 case '+':
1788 magic = ADD_LF_BEFORE_NON_EMPTY;
1789 break;
1790 case ' ':
1791 magic = ADD_SP_BEFORE_NON_EMPTY;
1792 break;
1793 default:
1794 break;
1796 if (magic != NO_MAGIC)
1797 placeholder++;
1799 orig_len = sb->len;
1800 if (((struct format_commit_context *)context)->flush_type != no_flush)
1801 consumed = format_and_pad_commit(sb, placeholder, context);
1802 else
1803 consumed = format_commit_one(sb, placeholder, context);
1804 if (magic == NO_MAGIC)
1805 return consumed;
1807 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1808 while (sb->len && sb->buf[sb->len - 1] == '\n')
1809 strbuf_setlen(sb, sb->len - 1);
1810 } else if (orig_len != sb->len) {
1811 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1812 strbuf_insertstr(sb, orig_len, "\n");
1813 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1814 strbuf_insertstr(sb, orig_len, " ");
1816 return consumed + 1;
1819 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1820 void *context)
1822 struct userformat_want *w = context;
1824 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1825 placeholder++;
1827 switch (*placeholder) {
1828 case 'N':
1829 w->notes = 1;
1830 break;
1831 case 'S':
1832 w->source = 1;
1833 break;
1834 case 'd':
1835 case 'D':
1836 w->decorate = 1;
1837 break;
1839 return 0;
1842 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1844 struct strbuf dummy = STRBUF_INIT;
1846 if (!fmt) {
1847 if (!user_format)
1848 return;
1849 fmt = user_format;
1851 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1852 strbuf_release(&dummy);
1855 void repo_format_commit_message(struct repository *r,
1856 const struct commit *commit,
1857 const char *format, struct strbuf *sb,
1858 const struct pretty_print_context *pretty_ctx)
1860 struct format_commit_context context = {
1861 .repository = r,
1862 .commit = commit,
1863 .pretty_ctx = pretty_ctx,
1864 .wrap_start = sb->len
1866 const char *output_enc = pretty_ctx->output_encoding;
1867 const char *utf8 = "UTF-8";
1869 strbuf_expand(sb, format, format_commit_item, &context);
1870 rewrap_message_tail(sb, &context, 0, 0, 0);
1873 * Convert output to an actual output encoding; note that
1874 * format_commit_item() will always use UTF-8, so we don't
1875 * have to bother if that's what the output wants.
1877 if (output_enc) {
1878 if (same_encoding(utf8, output_enc))
1879 output_enc = NULL;
1880 } else {
1881 if (context.commit_encoding &&
1882 !same_encoding(context.commit_encoding, utf8))
1883 output_enc = context.commit_encoding;
1886 if (output_enc) {
1887 size_t outsz;
1888 char *out = reencode_string_len(sb->buf, sb->len,
1889 output_enc, utf8, &outsz);
1890 if (out)
1891 strbuf_attach(sb, out, outsz, outsz + 1);
1894 free(context.commit_encoding);
1895 repo_unuse_commit_buffer(r, commit, context.message);
1898 static void pp_header(struct pretty_print_context *pp,
1899 const char *encoding,
1900 const struct commit *commit,
1901 const char **msg_p,
1902 struct strbuf *sb)
1904 int parents_shown = 0;
1906 for (;;) {
1907 const char *name, *line = *msg_p;
1908 int linelen = get_one_line(*msg_p);
1910 if (!linelen)
1911 return;
1912 *msg_p += linelen;
1914 if (linelen == 1)
1915 /* End of header */
1916 return;
1918 if (pp->fmt == CMIT_FMT_RAW) {
1919 strbuf_add(sb, line, linelen);
1920 continue;
1923 if (starts_with(line, "parent ")) {
1924 if (linelen != the_hash_algo->hexsz + 8)
1925 die("bad parent line in commit");
1926 continue;
1929 if (!parents_shown) {
1930 unsigned num = commit_list_count(commit->parents);
1931 /* with enough slop */
1932 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1933 add_merge_info(pp, sb, commit);
1934 parents_shown = 1;
1938 * MEDIUM == DEFAULT shows only author with dates.
1939 * FULL shows both authors but not dates.
1940 * FULLER shows both authors and dates.
1942 if (skip_prefix(line, "author ", &name)) {
1943 strbuf_grow(sb, linelen + 80);
1944 pp_user_info(pp, "Author", sb, name, encoding);
1946 if (skip_prefix(line, "committer ", &name) &&
1947 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1948 strbuf_grow(sb, linelen + 80);
1949 pp_user_info(pp, "Commit", sb, name, encoding);
1954 void pp_title_line(struct pretty_print_context *pp,
1955 const char **msg_p,
1956 struct strbuf *sb,
1957 const char *encoding,
1958 int need_8bit_cte)
1960 static const int max_length = 78; /* per rfc2047 */
1961 struct strbuf title;
1963 strbuf_init(&title, 80);
1964 *msg_p = format_subject(&title, *msg_p,
1965 pp->preserve_subject ? "\n" : " ");
1967 strbuf_grow(sb, title.len + 1024);
1968 if (pp->print_email_subject) {
1969 if (pp->rev)
1970 fmt_output_email_subject(sb, pp->rev);
1971 if (pp->encode_email_headers &&
1972 needs_rfc2047_encoding(title.buf, title.len))
1973 add_rfc2047(sb, title.buf, title.len,
1974 encoding, RFC2047_SUBJECT);
1975 else
1976 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1977 -last_line_length(sb), 1, max_length);
1978 } else {
1979 strbuf_addbuf(sb, &title);
1981 strbuf_addch(sb, '\n');
1983 if (need_8bit_cte == 0) {
1984 int i;
1985 for (i = 0; i < pp->in_body_headers.nr; i++) {
1986 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1987 need_8bit_cte = 1;
1988 break;
1993 if (need_8bit_cte > 0) {
1994 const char *header_fmt =
1995 "MIME-Version: 1.0\n"
1996 "Content-Type: text/plain; charset=%s\n"
1997 "Content-Transfer-Encoding: 8bit\n";
1998 strbuf_addf(sb, header_fmt, encoding);
2000 if (pp->after_subject) {
2001 strbuf_addstr(sb, pp->after_subject);
2003 if (cmit_fmt_is_mail(pp->fmt)) {
2004 strbuf_addch(sb, '\n');
2007 if (pp->in_body_headers.nr) {
2008 int i;
2009 for (i = 0; i < pp->in_body_headers.nr; i++) {
2010 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
2011 free(pp->in_body_headers.items[i].string);
2013 string_list_clear(&pp->in_body_headers, 0);
2014 strbuf_addch(sb, '\n');
2017 strbuf_release(&title);
2020 static int pp_utf8_width(const char *start, const char *end)
2022 int width = 0;
2023 size_t remain = end - start;
2025 while (remain) {
2026 int n = utf8_width(&start, &remain);
2027 if (n < 0 || !start)
2028 return -1;
2029 width += n;
2031 return width;
2034 static void strbuf_add_tabexpand(struct strbuf *sb, struct grep_opt *opt,
2035 int color, int tabwidth, const char *line,
2036 int linelen)
2038 const char *tab;
2040 while ((tab = memchr(line, '\t', linelen)) != NULL) {
2041 int width = pp_utf8_width(line, tab);
2044 * If it wasn't well-formed utf8, or it
2045 * had characters with badly defined
2046 * width (control characters etc), just
2047 * give up on trying to align things.
2049 if (width < 0)
2050 break;
2052 /* Output the data .. */
2053 append_line_with_color(sb, opt, line, tab - line, color,
2054 GREP_CONTEXT_BODY,
2055 GREP_HEADER_FIELD_MAX);
2057 /* .. and the de-tabified tab */
2058 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
2060 /* Skip over the printed part .. */
2061 linelen -= tab + 1 - line;
2062 line = tab + 1;
2066 * Print out everything after the last tab without
2067 * worrying about width - there's nothing more to
2068 * align.
2070 append_line_with_color(sb, opt, line, linelen, color, GREP_CONTEXT_BODY,
2071 GREP_HEADER_FIELD_MAX);
2075 * pp_handle_indent() prints out the intendation, and
2076 * the whole line (without the final newline), after
2077 * de-tabifying.
2079 static void pp_handle_indent(struct pretty_print_context *pp,
2080 struct strbuf *sb, int indent,
2081 const char *line, int linelen)
2083 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
2085 strbuf_addchars(sb, ' ', indent);
2086 if (pp->expand_tabs_in_log)
2087 strbuf_add_tabexpand(sb, opt, pp->color, pp->expand_tabs_in_log,
2088 line, linelen);
2089 else
2090 append_line_with_color(sb, opt, line, linelen, pp->color,
2091 GREP_CONTEXT_BODY,
2092 GREP_HEADER_FIELD_MAX);
2095 static int is_mboxrd_from(const char *line, int len)
2098 * a line matching /^From $/ here would only have len == 4
2099 * at this point because is_empty_line would've trimmed all
2100 * trailing space
2102 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
2105 void pp_remainder(struct pretty_print_context *pp,
2106 const char **msg_p,
2107 struct strbuf *sb,
2108 int indent)
2110 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
2111 int first = 1;
2113 for (;;) {
2114 const char *line = *msg_p;
2115 int linelen = get_one_line(line);
2116 *msg_p += linelen;
2118 if (!linelen)
2119 break;
2121 if (is_blank_line(line, &linelen)) {
2122 if (first)
2123 continue;
2124 if (pp->fmt == CMIT_FMT_SHORT)
2125 break;
2127 first = 0;
2129 strbuf_grow(sb, linelen + indent + 20);
2130 if (indent)
2131 pp_handle_indent(pp, sb, indent, line, linelen);
2132 else if (pp->expand_tabs_in_log)
2133 strbuf_add_tabexpand(sb, opt, pp->color,
2134 pp->expand_tabs_in_log, line,
2135 linelen);
2136 else {
2137 if (pp->fmt == CMIT_FMT_MBOXRD &&
2138 is_mboxrd_from(line, linelen))
2139 strbuf_addch(sb, '>');
2141 append_line_with_color(sb, opt, line, linelen,
2142 pp->color, GREP_CONTEXT_BODY,
2143 GREP_HEADER_FIELD_MAX);
2145 strbuf_addch(sb, '\n');
2149 void pretty_print_commit(struct pretty_print_context *pp,
2150 const struct commit *commit,
2151 struct strbuf *sb)
2153 unsigned long beginning_of_body;
2154 int indent = 4;
2155 const char *msg;
2156 const char *reencoded;
2157 const char *encoding;
2158 int need_8bit_cte = pp->need_8bit_cte;
2160 if (pp->fmt == CMIT_FMT_USERFORMAT) {
2161 format_commit_message(commit, user_format, sb, pp);
2162 return;
2165 encoding = get_log_output_encoding();
2166 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
2168 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2169 indent = 0;
2172 * We need to check and emit Content-type: to mark it
2173 * as 8-bit if we haven't done so.
2175 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
2176 int i, ch, in_body;
2178 for (in_body = i = 0; (ch = msg[i]); i++) {
2179 if (!in_body) {
2180 /* author could be non 7-bit ASCII but
2181 * the log may be so; skip over the
2182 * header part first.
2184 if (ch == '\n' && msg[i+1] == '\n')
2185 in_body = 1;
2187 else if (non_ascii(ch)) {
2188 need_8bit_cte = 1;
2189 break;
2194 pp_header(pp, encoding, commit, &msg, sb);
2195 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
2196 strbuf_addch(sb, '\n');
2199 /* Skip excess blank lines at the beginning of body, if any... */
2200 msg = skip_blank_lines(msg);
2202 /* These formats treat the title line specially. */
2203 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2204 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2206 beginning_of_body = sb->len;
2207 if (pp->fmt != CMIT_FMT_ONELINE)
2208 pp_remainder(pp, &msg, sb, indent);
2209 strbuf_rtrim(sb);
2211 /* Make sure there is an EOLN for the non-oneline case */
2212 if (pp->fmt != CMIT_FMT_ONELINE)
2213 strbuf_addch(sb, '\n');
2216 * The caller may append additional body text in e-mail
2217 * format. Make sure we did not strip the blank line
2218 * between the header and the body.
2220 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2221 strbuf_addch(sb, '\n');
2223 unuse_commit_buffer(commit, reencoded);
2226 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2227 struct strbuf *sb)
2229 struct pretty_print_context pp = {0};
2230 pp.fmt = fmt;
2231 pretty_print_commit(&pp, commit, sb);