clone: propagate object-format when cloning from void
[git/gitster.git] / pretty.c
blob05b557d09616de55ac4be5a03e1df02c6df65317
1 #include "cache.h"
2 #include "alloc.h"
3 #include "config.h"
4 #include "commit.h"
5 #include "hex.h"
6 #include "utf8.h"
7 #include "diff.h"
8 #include "revision.h"
9 #include "string-list.h"
10 #include "mailmap.h"
11 #include "log-tree.h"
12 #include "notes.h"
13 #include "color.h"
14 #include "reflog-walk.h"
15 #include "gpg-interface.h"
16 #include "trailer.h"
17 #include "run-command.h"
20 * The limit for formatting directives, which enable the caller to append
21 * arbitrarily many bytes to the formatted buffer. This includes padding
22 * and wrapping formatters.
24 #define FORMATTING_LIMIT (16 * 1024)
26 static char *user_format;
27 static struct cmt_fmt_map {
28 const char *name;
29 enum cmit_fmt format;
30 int is_tformat;
31 int expand_tabs_in_log;
32 int is_alias;
33 enum date_mode_type default_date_mode_type;
34 const char *user_format;
35 } *commit_formats;
36 static size_t builtin_formats_len;
37 static size_t commit_formats_len;
38 static size_t commit_formats_alloc;
39 static struct cmt_fmt_map *find_commit_format(const char *sought);
41 int commit_format_is_empty(enum cmit_fmt fmt)
43 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
46 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
48 free(user_format);
49 user_format = xstrdup(cp);
50 if (is_tformat)
51 rev->use_terminator = 1;
52 rev->commit_format = CMIT_FMT_USERFORMAT;
55 static int git_pretty_formats_config(const char *var, const char *value,
56 void *cb UNUSED)
58 struct cmt_fmt_map *commit_format = NULL;
59 const char *name;
60 const char *fmt;
61 int i;
63 if (!skip_prefix(var, "pretty.", &name))
64 return 0;
66 for (i = 0; i < builtin_formats_len; i++) {
67 if (!strcmp(commit_formats[i].name, name))
68 return 0;
71 for (i = builtin_formats_len; i < commit_formats_len; i++) {
72 if (!strcmp(commit_formats[i].name, name)) {
73 commit_format = &commit_formats[i];
74 break;
78 if (!commit_format) {
79 ALLOC_GROW(commit_formats, commit_formats_len+1,
80 commit_formats_alloc);
81 commit_format = &commit_formats[commit_formats_len];
82 memset(commit_format, 0, sizeof(*commit_format));
83 commit_formats_len++;
86 commit_format->name = xstrdup(name);
87 commit_format->format = CMIT_FMT_USERFORMAT;
88 if (git_config_string(&fmt, var, value))
89 return -1;
91 if (skip_prefix(fmt, "format:", &fmt))
92 commit_format->is_tformat = 0;
93 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
94 commit_format->is_tformat = 1;
95 else
96 commit_format->is_alias = 1;
97 commit_format->user_format = fmt;
99 return 0;
102 static void setup_commit_formats(void)
104 struct cmt_fmt_map builtin_formats[] = {
105 { "raw", CMIT_FMT_RAW, 0, 0 },
106 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
107 { "short", CMIT_FMT_SHORT, 0, 0 },
108 { "email", CMIT_FMT_EMAIL, 0, 0 },
109 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
110 { "fuller", CMIT_FMT_FULLER, 0, 8 },
111 { "full", CMIT_FMT_FULL, 0, 8 },
112 { "oneline", CMIT_FMT_ONELINE, 1, 0 },
113 { "reference", CMIT_FMT_USERFORMAT, 1, 0,
114 0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
116 * Please update $__git_log_pretty_formats in
117 * git-completion.bash when you add new formats.
120 commit_formats_len = ARRAY_SIZE(builtin_formats);
121 builtin_formats_len = commit_formats_len;
122 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
123 COPY_ARRAY(commit_formats, builtin_formats,
124 ARRAY_SIZE(builtin_formats));
126 git_config(git_pretty_formats_config, NULL);
129 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
130 const char *original,
131 int num_redirections)
133 struct cmt_fmt_map *found = NULL;
134 size_t found_match_len = 0;
135 int i;
137 if (num_redirections >= commit_formats_len)
138 die("invalid --pretty format: "
139 "'%s' references an alias which points to itself",
140 original);
142 for (i = 0; i < commit_formats_len; i++) {
143 size_t match_len;
145 if (!starts_with(commit_formats[i].name, sought))
146 continue;
148 match_len = strlen(commit_formats[i].name);
149 if (found == NULL || found_match_len > match_len) {
150 found = &commit_formats[i];
151 found_match_len = match_len;
155 if (found && found->is_alias) {
156 found = find_commit_format_recursive(found->user_format,
157 original,
158 num_redirections+1);
161 return found;
164 static struct cmt_fmt_map *find_commit_format(const char *sought)
166 if (!commit_formats)
167 setup_commit_formats();
169 return find_commit_format_recursive(sought, sought, 0);
172 void get_commit_format(const char *arg, struct rev_info *rev)
174 struct cmt_fmt_map *commit_format;
176 rev->use_terminator = 0;
177 if (!arg) {
178 rev->commit_format = CMIT_FMT_DEFAULT;
179 return;
181 if (skip_prefix(arg, "format:", &arg)) {
182 save_user_format(rev, arg, 0);
183 return;
186 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
187 save_user_format(rev, arg, 1);
188 return;
191 commit_format = find_commit_format(arg);
192 if (!commit_format)
193 die("invalid --pretty format: %s", arg);
195 rev->commit_format = commit_format->format;
196 rev->use_terminator = commit_format->is_tformat;
197 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
198 if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
199 rev->date_mode.type = commit_format->default_date_mode_type;
200 if (commit_format->format == CMIT_FMT_USERFORMAT) {
201 save_user_format(rev, commit_format->user_format,
202 commit_format->is_tformat);
207 * Generic support for pretty-printing the header
209 static int get_one_line(const char *msg)
211 int ret = 0;
213 for (;;) {
214 char c = *msg++;
215 if (!c)
216 break;
217 ret++;
218 if (c == '\n')
219 break;
221 return ret;
224 /* High bit set, or ISO-2022-INT */
225 static int non_ascii(int ch)
227 return !isascii(ch) || ch == '\033';
230 int has_non_ascii(const char *s)
232 int ch;
233 if (!s)
234 return 0;
235 while ((ch = *s++) != '\0') {
236 if (non_ascii(ch))
237 return 1;
239 return 0;
242 static int is_rfc822_special(char ch)
244 switch (ch) {
245 case '(':
246 case ')':
247 case '<':
248 case '>':
249 case '[':
250 case ']':
251 case ':':
252 case ';':
253 case '@':
254 case ',':
255 case '.':
256 case '"':
257 case '\\':
258 return 1;
259 default:
260 return 0;
264 static int needs_rfc822_quoting(const char *s, int len)
266 int i;
267 for (i = 0; i < len; i++)
268 if (is_rfc822_special(s[i]))
269 return 1;
270 return 0;
273 static int last_line_length(struct strbuf *sb)
275 int i;
277 /* How many bytes are already used on the last line? */
278 for (i = sb->len - 1; i >= 0; i--)
279 if (sb->buf[i] == '\n')
280 break;
281 return sb->len - (i + 1);
284 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
286 int i;
288 /* just a guess, we may have to also backslash-quote */
289 strbuf_grow(out, len + 2);
291 strbuf_addch(out, '"');
292 for (i = 0; i < len; i++) {
293 switch (s[i]) {
294 case '"':
295 case '\\':
296 strbuf_addch(out, '\\');
297 /* fall through */
298 default:
299 strbuf_addch(out, s[i]);
302 strbuf_addch(out, '"');
305 enum rfc2047_type {
306 RFC2047_SUBJECT,
307 RFC2047_ADDRESS
310 static int is_rfc2047_special(char ch, enum rfc2047_type type)
313 * rfc2047, section 4.2:
315 * 8-bit values which correspond to printable ASCII characters other
316 * than "=", "?", and "_" (underscore), MAY be represented as those
317 * characters. (But see section 5 for restrictions.) In
318 * particular, SPACE and TAB MUST NOT be represented as themselves
319 * within encoded words.
323 * rule out non-ASCII characters and non-printable characters (the
324 * non-ASCII check should be redundant as isprint() is not localized
325 * and only knows about ASCII, but be defensive about that)
327 if (non_ascii(ch) || !isprint(ch))
328 return 1;
331 * rule out special printable characters (' ' should be the only
332 * whitespace character considered printable, but be defensive and use
333 * isspace())
335 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
336 return 1;
339 * rfc2047, section 5.3:
341 * As a replacement for a 'word' entity within a 'phrase', for example,
342 * one that precedes an address in a From, To, or Cc header. The ABNF
343 * definition for 'phrase' from RFC 822 thus becomes:
345 * phrase = 1*( encoded-word / word )
347 * In this case the set of characters that may be used in a "Q"-encoded
348 * 'encoded-word' is restricted to: <upper and lower case ASCII
349 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
350 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
351 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
352 * 'special' by 'linear-white-space'.
355 if (type != RFC2047_ADDRESS)
356 return 0;
358 /* '=' and '_' are special cases and have been checked above */
359 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
362 static int needs_rfc2047_encoding(const char *line, int len)
364 int i;
366 for (i = 0; i < len; i++) {
367 int ch = line[i];
368 if (non_ascii(ch) || ch == '\n')
369 return 1;
370 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
371 return 1;
374 return 0;
377 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
378 const char *encoding, enum rfc2047_type type)
380 static const int max_encoded_length = 76; /* per rfc2047 */
381 int i;
382 int line_len = last_line_length(sb);
384 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
385 strbuf_addf(sb, "=?%s?q?", encoding);
386 line_len += strlen(encoding) + 5; /* 5 for =??q? */
388 while (len) {
390 * RFC 2047, section 5 (3):
392 * Each 'encoded-word' MUST represent an integral number of
393 * characters. A multi-octet character may not be split across
394 * adjacent 'encoded- word's.
396 const unsigned char *p = (const unsigned char *)line;
397 int chrlen = mbs_chrlen(&line, &len, encoding);
398 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
400 /* "=%02X" * chrlen, or the byte itself */
401 const char *encoded_fmt = is_special ? "=%02X" : "%c";
402 int encoded_len = is_special ? 3 * chrlen : 1;
405 * According to RFC 2047, we could encode the special character
406 * ' ' (space) with '_' (underscore) for readability. But many
407 * programs do not understand this and just leave the
408 * underscore in place. Thus, we do nothing special here, which
409 * causes ' ' to be encoded as '=20', avoiding this problem.
412 if (line_len + encoded_len + 2 > max_encoded_length) {
413 /* It won't fit with trailing "?=" --- break the line */
414 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
415 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
418 for (i = 0; i < chrlen; i++)
419 strbuf_addf(sb, encoded_fmt, p[i]);
420 line_len += encoded_len;
422 strbuf_addstr(sb, "?=");
425 const char *show_ident_date(const struct ident_split *ident,
426 const struct date_mode *mode)
428 timestamp_t date = 0;
429 long tz = 0;
431 if (ident->date_begin && ident->date_end)
432 date = parse_timestamp(ident->date_begin, NULL, 10);
433 if (date_overflows(date))
434 date = 0;
435 else {
436 if (ident->tz_begin && ident->tz_end)
437 tz = strtol(ident->tz_begin, NULL, 10);
438 if (tz >= INT_MAX || tz <= INT_MIN)
439 tz = 0;
441 return show_date(date, tz, mode);
444 static inline void strbuf_add_with_color(struct strbuf *sb, const char *color,
445 const char *buf, size_t buflen)
447 strbuf_addstr(sb, color);
448 strbuf_add(sb, buf, buflen);
449 if (*color)
450 strbuf_addstr(sb, GIT_COLOR_RESET);
453 static void append_line_with_color(struct strbuf *sb, struct grep_opt *opt,
454 const char *line, size_t linelen,
455 int color, enum grep_context ctx,
456 enum grep_header_field field)
458 const char *buf, *eol, *line_color, *match_color;
459 regmatch_t match;
460 int eflags = 0;
462 buf = line;
463 eol = buf + linelen;
465 if (!opt || !want_color(color) || opt->invert)
466 goto end;
468 line_color = opt->colors[GREP_COLOR_SELECTED];
469 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
471 while (grep_next_match(opt, buf, eol, ctx, &match, field, eflags)) {
472 if (match.rm_so == match.rm_eo)
473 break;
475 strbuf_add_with_color(sb, line_color, buf, match.rm_so);
476 strbuf_add_with_color(sb, match_color, buf + match.rm_so,
477 match.rm_eo - match.rm_so);
478 buf += match.rm_eo;
479 eflags = REG_NOTBOL;
482 if (eflags)
483 strbuf_add_with_color(sb, line_color, buf, eol - buf);
484 else {
485 end:
486 strbuf_add(sb, buf, eol - buf);
490 static int use_in_body_from(const struct pretty_print_context *pp,
491 const struct ident_split *ident)
493 if (pp->rev && pp->rev->force_in_body_from)
494 return 1;
495 if (ident_cmp(pp->from_ident, ident))
496 return 1;
497 return 0;
500 void pp_user_info(struct pretty_print_context *pp,
501 const char *what, struct strbuf *sb,
502 const char *line, const char *encoding)
504 struct ident_split ident;
505 char *line_end;
506 const char *mailbuf, *namebuf;
507 size_t namelen, maillen;
508 int max_length = 78; /* per rfc2822 */
510 if (pp->fmt == CMIT_FMT_ONELINE)
511 return;
513 line_end = strchrnul(line, '\n');
514 if (split_ident_line(&ident, line, line_end - line))
515 return;
517 mailbuf = ident.mail_begin;
518 maillen = ident.mail_end - ident.mail_begin;
519 namebuf = ident.name_begin;
520 namelen = ident.name_end - ident.name_begin;
522 if (pp->mailmap)
523 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
525 if (cmit_fmt_is_mail(pp->fmt)) {
526 if (pp->from_ident && use_in_body_from(pp, &ident)) {
527 struct strbuf buf = STRBUF_INIT;
529 strbuf_addstr(&buf, "From: ");
530 strbuf_add(&buf, namebuf, namelen);
531 strbuf_addstr(&buf, " <");
532 strbuf_add(&buf, mailbuf, maillen);
533 strbuf_addstr(&buf, ">\n");
534 string_list_append(&pp->in_body_headers,
535 strbuf_detach(&buf, NULL));
537 mailbuf = pp->from_ident->mail_begin;
538 maillen = pp->from_ident->mail_end - mailbuf;
539 namebuf = pp->from_ident->name_begin;
540 namelen = pp->from_ident->name_end - namebuf;
543 strbuf_addstr(sb, "From: ");
544 if (pp->encode_email_headers &&
545 needs_rfc2047_encoding(namebuf, namelen)) {
546 add_rfc2047(sb, namebuf, namelen,
547 encoding, RFC2047_ADDRESS);
548 max_length = 76; /* per rfc2047 */
549 } else if (needs_rfc822_quoting(namebuf, namelen)) {
550 struct strbuf quoted = STRBUF_INIT;
551 add_rfc822_quoted(&quoted, namebuf, namelen);
552 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
553 -6, 1, max_length);
554 strbuf_release(&quoted);
555 } else {
556 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
557 -6, 1, max_length);
560 if (max_length <
561 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
562 strbuf_addch(sb, '\n');
563 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
564 } else {
565 struct strbuf id = STRBUF_INIT;
566 enum grep_header_field field = GREP_HEADER_FIELD_MAX;
567 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
569 if (!strcmp(what, "Author"))
570 field = GREP_HEADER_AUTHOR;
571 else if (!strcmp(what, "Commit"))
572 field = GREP_HEADER_COMMITTER;
574 strbuf_addf(sb, "%s: ", what);
575 if (pp->fmt == CMIT_FMT_FULLER)
576 strbuf_addchars(sb, ' ', 4);
578 strbuf_addf(&id, "%.*s <%.*s>", (int)namelen, namebuf,
579 (int)maillen, mailbuf);
581 append_line_with_color(sb, opt, id.buf, id.len, pp->color,
582 GREP_CONTEXT_HEAD, field);
583 strbuf_addch(sb, '\n');
584 strbuf_release(&id);
587 switch (pp->fmt) {
588 case CMIT_FMT_MEDIUM:
589 strbuf_addf(sb, "Date: %s\n",
590 show_ident_date(&ident, &pp->date_mode));
591 break;
592 case CMIT_FMT_EMAIL:
593 case CMIT_FMT_MBOXRD:
594 strbuf_addf(sb, "Date: %s\n",
595 show_ident_date(&ident, DATE_MODE(RFC2822)));
596 break;
597 case CMIT_FMT_FULLER:
598 strbuf_addf(sb, "%sDate: %s\n", what,
599 show_ident_date(&ident, &pp->date_mode));
600 break;
601 default:
602 /* notin' */
603 break;
607 static int is_blank_line(const char *line, int *len_p)
609 int len = *len_p;
610 while (len && isspace(line[len - 1]))
611 len--;
612 *len_p = len;
613 return !len;
616 const char *skip_blank_lines(const char *msg)
618 for (;;) {
619 int linelen = get_one_line(msg);
620 int ll = linelen;
621 if (!linelen)
622 break;
623 if (!is_blank_line(msg, &ll))
624 break;
625 msg += linelen;
627 return msg;
630 static void add_merge_info(const struct pretty_print_context *pp,
631 struct strbuf *sb, const struct commit *commit)
633 struct commit_list *parent = commit->parents;
635 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
636 !parent || !parent->next)
637 return;
639 strbuf_addstr(sb, "Merge:");
641 while (parent) {
642 struct object_id *oidp = &parent->item->object.oid;
643 strbuf_addch(sb, ' ');
644 if (pp->abbrev)
645 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
646 else
647 strbuf_addstr(sb, oid_to_hex(oidp));
648 parent = parent->next;
650 strbuf_addch(sb, '\n');
653 static char *get_header(const char *msg, const char *key)
655 size_t len;
656 const char *v = find_commit_header(msg, key, &len);
657 return v ? xmemdupz(v, len) : NULL;
660 static char *replace_encoding_header(char *buf, const char *encoding)
662 struct strbuf tmp = STRBUF_INIT;
663 size_t start, len;
664 char *cp = buf;
666 /* guess if there is an encoding header before a \n\n */
667 while (!starts_with(cp, "encoding ")) {
668 cp = strchr(cp, '\n');
669 if (!cp || *++cp == '\n')
670 return buf;
672 start = cp - buf;
673 cp = strchr(cp, '\n');
674 if (!cp)
675 return buf; /* should not happen but be defensive */
676 len = cp + 1 - (buf + start);
678 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
679 if (is_encoding_utf8(encoding)) {
680 /* we have re-coded to UTF-8; drop the header */
681 strbuf_remove(&tmp, start, len);
682 } else {
683 /* just replaces XXXX in 'encoding XXXX\n' */
684 strbuf_splice(&tmp, start + strlen("encoding "),
685 len - strlen("encoding \n"),
686 encoding, strlen(encoding));
688 return strbuf_detach(&tmp, NULL);
691 const char *repo_logmsg_reencode(struct repository *r,
692 const struct commit *commit,
693 char **commit_encoding,
694 const char *output_encoding)
696 static const char *utf8 = "UTF-8";
697 const char *use_encoding;
698 char *encoding;
699 const char *msg = repo_get_commit_buffer(r, commit, NULL);
700 char *out;
702 if (!output_encoding || !*output_encoding) {
703 if (commit_encoding)
704 *commit_encoding = get_header(msg, "encoding");
705 return msg;
707 encoding = get_header(msg, "encoding");
708 if (commit_encoding)
709 *commit_encoding = encoding;
710 use_encoding = encoding ? encoding : utf8;
711 if (same_encoding(use_encoding, output_encoding)) {
713 * No encoding work to be done. If we have no encoding header
714 * at all, then there's nothing to do, and we can return the
715 * message verbatim (whether newly allocated or not).
717 if (!encoding)
718 return msg;
721 * Otherwise, we still want to munge the encoding header in the
722 * result, which will be done by modifying the buffer. If we
723 * are using a fresh copy, we can reuse it. But if we are using
724 * the cached copy from get_commit_buffer, we need to duplicate it
725 * to avoid munging the cached copy.
727 if (msg == get_cached_commit_buffer(r, commit, NULL))
728 out = xstrdup(msg);
729 else
730 out = (char *)msg;
732 else {
734 * There's actual encoding work to do. Do the reencoding, which
735 * still leaves the header to be replaced in the next step. At
736 * this point, we are done with msg. If we allocated a fresh
737 * copy, we can free it.
739 out = reencode_string(msg, output_encoding, use_encoding);
740 if (out)
741 repo_unuse_commit_buffer(r, commit, msg);
745 * This replacement actually consumes the buffer we hand it, so we do
746 * not have to worry about freeing the old "out" here.
748 if (out)
749 out = replace_encoding_header(out, output_encoding);
751 if (!commit_encoding)
752 free(encoding);
754 * If the re-encoding failed, out might be NULL here; in that
755 * case we just return the commit message verbatim.
757 return out ? out : msg;
760 static int mailmap_name(const char **email, size_t *email_len,
761 const char **name, size_t *name_len)
763 static struct string_list *mail_map;
764 if (!mail_map) {
765 CALLOC_ARRAY(mail_map, 1);
766 read_mailmap(mail_map);
768 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
771 static size_t format_person_part(struct strbuf *sb, char part,
772 const char *msg, int len,
773 const struct date_mode *dmode)
775 /* currently all placeholders have same length */
776 const int placeholder_len = 2;
777 struct ident_split s;
778 const char *name, *mail;
779 size_t maillen, namelen;
781 if (split_ident_line(&s, msg, len) < 0)
782 goto skip;
784 name = s.name_begin;
785 namelen = s.name_end - s.name_begin;
786 mail = s.mail_begin;
787 maillen = s.mail_end - s.mail_begin;
789 if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
790 mailmap_name(&mail, &maillen, &name, &namelen);
791 if (part == 'n' || part == 'N') { /* name */
792 strbuf_add(sb, name, namelen);
793 return placeholder_len;
795 if (part == 'e' || part == 'E') { /* email */
796 strbuf_add(sb, mail, maillen);
797 return placeholder_len;
799 if (part == 'l' || part == 'L') { /* local-part */
800 const char *at = memchr(mail, '@', maillen);
801 if (at)
802 maillen = at - mail;
803 strbuf_add(sb, mail, maillen);
804 return placeholder_len;
807 if (!s.date_begin)
808 goto skip;
810 if (part == 't') { /* date, UNIX timestamp */
811 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
812 return placeholder_len;
815 switch (part) {
816 case 'd': /* date */
817 strbuf_addstr(sb, show_ident_date(&s, dmode));
818 return placeholder_len;
819 case 'D': /* date, RFC2822 style */
820 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
821 return placeholder_len;
822 case 'r': /* date, relative */
823 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
824 return placeholder_len;
825 case 'i': /* date, ISO 8601-like */
826 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
827 return placeholder_len;
828 case 'I': /* date, ISO 8601 strict */
829 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
830 return placeholder_len;
831 case 'h': /* date, human */
832 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(HUMAN)));
833 return placeholder_len;
834 case 's':
835 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
836 return placeholder_len;
839 skip:
841 * reading from either a bogus commit, or a reflog entry with
842 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
843 * to compute a valid return value.
845 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
846 || part == 'D' || part == 'r' || part == 'i')
847 return placeholder_len;
849 return 0; /* unknown placeholder */
852 struct chunk {
853 size_t off;
854 size_t len;
857 enum flush_type {
858 no_flush,
859 flush_right,
860 flush_left,
861 flush_left_and_steal,
862 flush_both
865 enum trunc_type {
866 trunc_none,
867 trunc_left,
868 trunc_middle,
869 trunc_right
872 struct format_commit_context {
873 struct repository *repository;
874 const struct commit *commit;
875 const struct pretty_print_context *pretty_ctx;
876 unsigned commit_header_parsed:1;
877 unsigned commit_message_parsed:1;
878 struct signature_check signature_check;
879 enum flush_type flush_type;
880 enum trunc_type truncate;
881 const char *message;
882 char *commit_encoding;
883 size_t width, indent1, indent2;
884 int auto_color;
885 int padding;
887 /* These offsets are relative to the start of the commit message. */
888 struct chunk author;
889 struct chunk committer;
890 size_t message_off;
891 size_t subject_off;
892 size_t body_off;
894 /* The following ones are relative to the result struct strbuf. */
895 size_t wrap_start;
898 static void parse_commit_header(struct format_commit_context *context)
900 const char *msg = context->message;
901 int i;
903 for (i = 0; msg[i]; i++) {
904 const char *name;
905 int eol;
906 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
907 ; /* do nothing */
909 if (i == eol) {
910 break;
911 } else if (skip_prefix(msg + i, "author ", &name)) {
912 context->author.off = name - msg;
913 context->author.len = msg + eol - name;
914 } else if (skip_prefix(msg + i, "committer ", &name)) {
915 context->committer.off = name - msg;
916 context->committer.len = msg + eol - name;
918 i = eol;
920 context->message_off = i;
921 context->commit_header_parsed = 1;
924 static int istitlechar(char c)
926 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
927 (c >= '0' && c <= '9') || c == '.' || c == '_';
930 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
932 size_t trimlen;
933 size_t start_len = sb->len;
934 int space = 2;
935 int i;
937 for (i = 0; i < len; i++) {
938 if (istitlechar(msg[i])) {
939 if (space == 1)
940 strbuf_addch(sb, '-');
941 space = 0;
942 strbuf_addch(sb, msg[i]);
943 if (msg[i] == '.')
944 while (msg[i+1] == '.')
945 i++;
946 } else
947 space |= 1;
950 /* trim any trailing '.' or '-' characters */
951 trimlen = 0;
952 while (sb->len - trimlen > start_len &&
953 (sb->buf[sb->len - 1 - trimlen] == '.'
954 || sb->buf[sb->len - 1 - trimlen] == '-'))
955 trimlen++;
956 strbuf_remove(sb, sb->len - trimlen, trimlen);
959 const char *format_subject(struct strbuf *sb, const char *msg,
960 const char *line_separator)
962 int first = 1;
964 for (;;) {
965 const char *line = msg;
966 int linelen = get_one_line(line);
968 msg += linelen;
969 if (!linelen || is_blank_line(line, &linelen))
970 break;
972 if (!sb)
973 continue;
974 strbuf_grow(sb, linelen + 2);
975 if (!first)
976 strbuf_addstr(sb, line_separator);
977 strbuf_add(sb, line, linelen);
978 first = 0;
980 return msg;
983 static void parse_commit_message(struct format_commit_context *c)
985 const char *msg = c->message + c->message_off;
986 const char *start = c->message;
988 msg = skip_blank_lines(msg);
989 c->subject_off = msg - start;
991 msg = format_subject(NULL, msg, NULL);
992 msg = skip_blank_lines(msg);
993 c->body_off = msg - start;
995 c->commit_message_parsed = 1;
998 static void strbuf_wrap(struct strbuf *sb, size_t pos,
999 size_t width, size_t indent1, size_t indent2)
1001 struct strbuf tmp = STRBUF_INIT;
1003 if (pos)
1004 strbuf_add(&tmp, sb->buf, pos);
1005 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
1006 cast_size_t_to_int(indent1),
1007 cast_size_t_to_int(indent2),
1008 cast_size_t_to_int(width));
1009 strbuf_swap(&tmp, sb);
1010 strbuf_release(&tmp);
1013 static void rewrap_message_tail(struct strbuf *sb,
1014 struct format_commit_context *c,
1015 size_t new_width, size_t new_indent1,
1016 size_t new_indent2)
1018 if (c->width == new_width && c->indent1 == new_indent1 &&
1019 c->indent2 == new_indent2)
1020 return;
1021 if (c->wrap_start < sb->len)
1022 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
1023 c->wrap_start = sb->len;
1024 c->width = new_width;
1025 c->indent1 = new_indent1;
1026 c->indent2 = new_indent2;
1029 static int format_reflog_person(struct strbuf *sb,
1030 char part,
1031 struct reflog_walk_info *log,
1032 const struct date_mode *dmode)
1034 const char *ident;
1036 if (!log)
1037 return 2;
1039 ident = get_reflog_ident(log);
1040 if (!ident)
1041 return 2;
1043 return format_person_part(sb, part, ident, strlen(ident), dmode);
1046 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
1047 const char *placeholder,
1048 struct format_commit_context *c)
1050 const char *rest = placeholder;
1051 const char *basic_color = NULL;
1053 if (placeholder[1] == '(') {
1054 const char *begin = placeholder + 2;
1055 const char *end = strchr(begin, ')');
1056 char color[COLOR_MAXLEN];
1058 if (!end)
1059 return 0;
1061 if (skip_prefix(begin, "auto,", &begin)) {
1062 if (!want_color(c->pretty_ctx->color))
1063 return end - placeholder + 1;
1064 } else if (skip_prefix(begin, "always,", &begin)) {
1065 /* nothing to do; we do not respect want_color at all */
1066 } else {
1067 /* the default is the same as "auto" */
1068 if (!want_color(c->pretty_ctx->color))
1069 return end - placeholder + 1;
1072 if (color_parse_mem(begin, end - begin, color) < 0)
1073 die(_("unable to parse --pretty format"));
1074 strbuf_addstr(sb, color);
1075 return end - placeholder + 1;
1079 * We handle things like "%C(red)" above; for historical reasons, there
1080 * are a few colors that can be specified without parentheses (and
1081 * they cannot support things like "auto" or "always" at all).
1083 if (skip_prefix(placeholder + 1, "red", &rest))
1084 basic_color = GIT_COLOR_RED;
1085 else if (skip_prefix(placeholder + 1, "green", &rest))
1086 basic_color = GIT_COLOR_GREEN;
1087 else if (skip_prefix(placeholder + 1, "blue", &rest))
1088 basic_color = GIT_COLOR_BLUE;
1089 else if (skip_prefix(placeholder + 1, "reset", &rest))
1090 basic_color = GIT_COLOR_RESET;
1092 if (basic_color && want_color(c->pretty_ctx->color))
1093 strbuf_addstr(sb, basic_color);
1095 return rest - placeholder;
1098 static size_t parse_padding_placeholder(const char *placeholder,
1099 struct format_commit_context *c)
1101 const char *ch = placeholder;
1102 enum flush_type flush_type;
1103 int to_column = 0;
1105 switch (*ch++) {
1106 case '<':
1107 flush_type = flush_right;
1108 break;
1109 case '>':
1110 if (*ch == '<') {
1111 flush_type = flush_both;
1112 ch++;
1113 } else if (*ch == '>') {
1114 flush_type = flush_left_and_steal;
1115 ch++;
1116 } else
1117 flush_type = flush_left;
1118 break;
1119 default:
1120 return 0;
1123 /* the next value means "wide enough to that column" */
1124 if (*ch == '|') {
1125 to_column = 1;
1126 ch++;
1129 if (*ch == '(') {
1130 const char *start = ch + 1;
1131 const char *end = start + strcspn(start, ",)");
1132 char *next;
1133 int width;
1134 if (!*end || end == start)
1135 return 0;
1136 width = strtol(start, &next, 10);
1139 * We need to limit the amount of padding, or otherwise this
1140 * would allow the user to pad the buffer by arbitrarily many
1141 * bytes and thus cause resource exhaustion.
1143 if (width < -FORMATTING_LIMIT || width > FORMATTING_LIMIT)
1144 return 0;
1146 if (next == start || width == 0)
1147 return 0;
1148 if (width < 0) {
1149 if (to_column)
1150 width += term_columns();
1151 if (width < 0)
1152 return 0;
1154 c->padding = to_column ? -width : width;
1155 c->flush_type = flush_type;
1157 if (*end == ',') {
1158 start = end + 1;
1159 end = strchr(start, ')');
1160 if (!end || end == start)
1161 return 0;
1162 if (starts_with(start, "trunc)"))
1163 c->truncate = trunc_right;
1164 else if (starts_with(start, "ltrunc)"))
1165 c->truncate = trunc_left;
1166 else if (starts_with(start, "mtrunc)"))
1167 c->truncate = trunc_middle;
1168 else
1169 return 0;
1170 } else
1171 c->truncate = trunc_none;
1173 return end - placeholder + 1;
1175 return 0;
1178 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1179 const char **end, const char **valuestart,
1180 size_t *valuelen)
1182 const char *p;
1184 if (!(skip_prefix(to_parse, candidate, &p)))
1185 return 0;
1186 if (valuestart) {
1187 if (*p == '=') {
1188 *valuestart = p + 1;
1189 *valuelen = strcspn(*valuestart, ",)");
1190 p = *valuestart + *valuelen;
1191 } else {
1192 if (*p != ',' && *p != ')')
1193 return 0;
1194 *valuestart = NULL;
1195 *valuelen = 0;
1198 if (*p == ',') {
1199 *end = p + 1;
1200 return 1;
1202 if (*p == ')') {
1203 *end = p;
1204 return 1;
1206 return 0;
1209 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1210 const char **end, int *val)
1212 const char *argval;
1213 char *strval;
1214 size_t arglen;
1215 int v;
1217 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1218 return 0;
1220 if (!argval) {
1221 *val = 1;
1222 return 1;
1225 strval = xstrndup(argval, arglen);
1226 v = git_parse_maybe_bool(strval);
1227 free(strval);
1229 if (v == -1)
1230 return 0;
1232 *val = v;
1234 return 1;
1237 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1239 const struct string_list *list = ud;
1240 const struct string_list_item *item;
1242 for_each_string_list_item (item, list) {
1243 if (key->len == (uintptr_t)item->util &&
1244 !strncasecmp(item->string, key->buf, key->len))
1245 return 1;
1247 return 0;
1250 int format_set_trailers_options(struct process_trailer_options *opts,
1251 struct string_list *filter_list,
1252 struct strbuf *sepbuf,
1253 struct strbuf *kvsepbuf,
1254 const char **arg,
1255 char **invalid_arg)
1257 for (;;) {
1258 const char *argval;
1259 size_t arglen;
1261 if (**arg == ')')
1262 break;
1264 if (match_placeholder_arg_value(*arg, "key", arg, &argval, &arglen)) {
1265 uintptr_t len = arglen;
1267 if (!argval)
1268 return -1;
1270 if (len && argval[len - 1] == ':')
1271 len--;
1272 string_list_append(filter_list, argval)->util = (char *)len;
1274 opts->filter = format_trailer_match_cb;
1275 opts->filter_data = filter_list;
1276 opts->only_trailers = 1;
1277 } else if (match_placeholder_arg_value(*arg, "separator", arg, &argval, &arglen)) {
1278 char *fmt;
1280 strbuf_reset(sepbuf);
1281 fmt = xstrndup(argval, arglen);
1282 strbuf_expand(sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1283 free(fmt);
1284 opts->separator = sepbuf;
1285 } else if (match_placeholder_arg_value(*arg, "key_value_separator", arg, &argval, &arglen)) {
1286 char *fmt;
1288 strbuf_reset(kvsepbuf);
1289 fmt = xstrndup(argval, arglen);
1290 strbuf_expand(kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1291 free(fmt);
1292 opts->key_value_separator = kvsepbuf;
1293 } else if (!match_placeholder_bool_arg(*arg, "only", arg, &opts->only_trailers) &&
1294 !match_placeholder_bool_arg(*arg, "unfold", arg, &opts->unfold) &&
1295 !match_placeholder_bool_arg(*arg, "keyonly", arg, &opts->key_only) &&
1296 !match_placeholder_bool_arg(*arg, "valueonly", arg, &opts->value_only)) {
1297 if (invalid_arg) {
1298 size_t len = strcspn(*arg, ",)");
1299 *invalid_arg = xstrndup(*arg, len);
1301 return -1;
1304 return 0;
1307 static size_t parse_describe_args(const char *start, struct strvec *args)
1309 struct {
1310 char *name;
1311 enum {
1312 DESCRIBE_ARG_BOOL,
1313 DESCRIBE_ARG_INTEGER,
1314 DESCRIBE_ARG_STRING,
1315 } type;
1316 } option[] = {
1317 { "tags", DESCRIBE_ARG_BOOL},
1318 { "abbrev", DESCRIBE_ARG_INTEGER },
1319 { "exclude", DESCRIBE_ARG_STRING },
1320 { "match", DESCRIBE_ARG_STRING },
1322 const char *arg = start;
1324 for (;;) {
1325 int found = 0;
1326 const char *argval;
1327 size_t arglen = 0;
1328 int optval = 0;
1329 int i;
1331 for (i = 0; !found && i < ARRAY_SIZE(option); i++) {
1332 switch (option[i].type) {
1333 case DESCRIBE_ARG_BOOL:
1334 if (match_placeholder_bool_arg(arg, option[i].name, &arg, &optval)) {
1335 if (optval)
1336 strvec_pushf(args, "--%s", option[i].name);
1337 else
1338 strvec_pushf(args, "--no-%s", option[i].name);
1339 found = 1;
1341 break;
1342 case DESCRIBE_ARG_INTEGER:
1343 if (match_placeholder_arg_value(arg, option[i].name, &arg,
1344 &argval, &arglen)) {
1345 char *endptr;
1346 if (!arglen)
1347 return 0;
1348 strtol(argval, &endptr, 10);
1349 if (endptr - argval != arglen)
1350 return 0;
1351 strvec_pushf(args, "--%s=%.*s", option[i].name, (int)arglen, argval);
1352 found = 1;
1354 break;
1355 case DESCRIBE_ARG_STRING:
1356 if (match_placeholder_arg_value(arg, option[i].name, &arg,
1357 &argval, &arglen)) {
1358 if (!arglen)
1359 return 0;
1360 strvec_pushf(args, "--%s=%.*s", option[i].name, (int)arglen, argval);
1361 found = 1;
1363 break;
1366 if (!found)
1367 break;
1370 return arg - start;
1373 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1374 const char *placeholder,
1375 void *context)
1377 struct format_commit_context *c = context;
1378 const struct commit *commit = c->commit;
1379 const char *msg = c->message;
1380 struct commit_list *p;
1381 const char *arg, *eol;
1382 size_t res;
1383 char **slot;
1385 /* these are independent of the commit */
1386 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1387 if (res)
1388 return res;
1390 switch (placeholder[0]) {
1391 case 'C':
1392 if (starts_with(placeholder + 1, "(auto)")) {
1393 c->auto_color = want_color(c->pretty_ctx->color);
1394 if (c->auto_color && sb->len)
1395 strbuf_addstr(sb, GIT_COLOR_RESET);
1396 return 7; /* consumed 7 bytes, "C(auto)" */
1397 } else {
1398 int ret = parse_color(sb, placeholder, c);
1399 if (ret)
1400 c->auto_color = 0;
1402 * Otherwise, we decided to treat %C<unknown>
1403 * as a literal string, and the previous
1404 * %C(auto) is still valid.
1406 return ret;
1408 case 'w':
1409 if (placeholder[1] == '(') {
1410 unsigned long width = 0, indent1 = 0, indent2 = 0;
1411 char *next;
1412 const char *start = placeholder + 2;
1413 const char *end = strchr(start, ')');
1414 if (!end)
1415 return 0;
1416 if (end > start) {
1417 width = strtoul(start, &next, 10);
1418 if (*next == ',') {
1419 indent1 = strtoul(next + 1, &next, 10);
1420 if (*next == ',') {
1421 indent2 = strtoul(next + 1,
1422 &next, 10);
1425 if (*next != ')')
1426 return 0;
1430 * We need to limit the format here as it allows the
1431 * user to prepend arbitrarily many bytes to the buffer
1432 * when rewrapping.
1434 if (width > FORMATTING_LIMIT ||
1435 indent1 > FORMATTING_LIMIT ||
1436 indent2 > FORMATTING_LIMIT)
1437 return 0;
1438 rewrap_message_tail(sb, c, width, indent1, indent2);
1439 return end - placeholder + 1;
1440 } else
1441 return 0;
1443 case '<':
1444 case '>':
1445 return parse_padding_placeholder(placeholder, c);
1448 if (skip_prefix(placeholder, "(describe", &arg)) {
1449 struct child_process cmd = CHILD_PROCESS_INIT;
1450 struct strbuf out = STRBUF_INIT;
1451 struct strbuf err = STRBUF_INIT;
1452 struct pretty_print_describe_status *describe_status;
1454 describe_status = c->pretty_ctx->describe_status;
1455 if (describe_status) {
1456 if (!describe_status->max_invocations)
1457 return 0;
1458 describe_status->max_invocations--;
1461 cmd.git_cmd = 1;
1462 strvec_push(&cmd.args, "describe");
1464 if (*arg == ':') {
1465 arg++;
1466 arg += parse_describe_args(arg, &cmd.args);
1469 if (*arg != ')') {
1470 child_process_clear(&cmd);
1471 return 0;
1474 strvec_push(&cmd.args, oid_to_hex(&commit->object.oid));
1475 pipe_command(&cmd, NULL, 0, &out, 0, &err, 0);
1476 strbuf_rtrim(&out);
1477 strbuf_addbuf(sb, &out);
1478 strbuf_release(&out);
1479 strbuf_release(&err);
1480 return arg - placeholder + 1;
1483 /* these depend on the commit */
1484 if (!commit->object.parsed)
1485 parse_object(the_repository, &commit->object.oid);
1487 switch (placeholder[0]) {
1488 case 'H': /* commit hash */
1489 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1490 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1491 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1492 return 1;
1493 case 'h': /* abbreviated commit hash */
1494 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1495 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1496 c->pretty_ctx->abbrev);
1497 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1498 return 1;
1499 case 'T': /* tree hash */
1500 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1501 return 1;
1502 case 't': /* abbreviated tree hash */
1503 strbuf_add_unique_abbrev(sb,
1504 get_commit_tree_oid(commit),
1505 c->pretty_ctx->abbrev);
1506 return 1;
1507 case 'P': /* parent hashes */
1508 for (p = commit->parents; p; p = p->next) {
1509 if (p != commit->parents)
1510 strbuf_addch(sb, ' ');
1511 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1513 return 1;
1514 case 'p': /* abbreviated parent hashes */
1515 for (p = commit->parents; p; p = p->next) {
1516 if (p != commit->parents)
1517 strbuf_addch(sb, ' ');
1518 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1519 c->pretty_ctx->abbrev);
1521 return 1;
1522 case 'm': /* left/right/bottom */
1523 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1524 return 1;
1525 case 'd':
1526 format_decorations(sb, commit, c->auto_color);
1527 return 1;
1528 case 'D':
1529 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1530 return 1;
1531 case 'S': /* tag/branch like --source */
1532 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1533 return 0;
1534 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1535 if (!(slot && *slot))
1536 return 0;
1537 strbuf_addstr(sb, *slot);
1538 return 1;
1539 case 'g': /* reflog info */
1540 switch(placeholder[1]) {
1541 case 'd': /* reflog selector */
1542 case 'D':
1543 if (c->pretty_ctx->reflog_info)
1544 get_reflog_selector(sb,
1545 c->pretty_ctx->reflog_info,
1546 &c->pretty_ctx->date_mode,
1547 c->pretty_ctx->date_mode_explicit,
1548 (placeholder[1] == 'd'));
1549 return 2;
1550 case 's': /* reflog message */
1551 if (c->pretty_ctx->reflog_info)
1552 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1553 return 2;
1554 case 'n':
1555 case 'N':
1556 case 'e':
1557 case 'E':
1558 return format_reflog_person(sb,
1559 placeholder[1],
1560 c->pretty_ctx->reflog_info,
1561 &c->pretty_ctx->date_mode);
1563 return 0; /* unknown %g placeholder */
1564 case 'N':
1565 if (c->pretty_ctx->notes_message) {
1566 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1567 return 1;
1569 return 0;
1572 if (placeholder[0] == 'G') {
1573 if (!c->signature_check.result)
1574 check_commit_signature(c->commit, &(c->signature_check));
1575 switch (placeholder[1]) {
1576 case 'G':
1577 if (c->signature_check.output)
1578 strbuf_addstr(sb, c->signature_check.output);
1579 break;
1580 case '?':
1581 switch (c->signature_check.result) {
1582 case 'G':
1583 switch (c->signature_check.trust_level) {
1584 case TRUST_UNDEFINED:
1585 case TRUST_NEVER:
1586 strbuf_addch(sb, 'U');
1587 break;
1588 default:
1589 strbuf_addch(sb, 'G');
1590 break;
1592 break;
1593 case 'B':
1594 case 'E':
1595 case 'N':
1596 case 'X':
1597 case 'Y':
1598 case 'R':
1599 strbuf_addch(sb, c->signature_check.result);
1601 break;
1602 case 'S':
1603 if (c->signature_check.signer)
1604 strbuf_addstr(sb, c->signature_check.signer);
1605 break;
1606 case 'K':
1607 if (c->signature_check.key)
1608 strbuf_addstr(sb, c->signature_check.key);
1609 break;
1610 case 'F':
1611 if (c->signature_check.fingerprint)
1612 strbuf_addstr(sb, c->signature_check.fingerprint);
1613 break;
1614 case 'P':
1615 if (c->signature_check.primary_key_fingerprint)
1616 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1617 break;
1618 case 'T':
1619 strbuf_addstr(sb, gpg_trust_level_to_str(c->signature_check.trust_level));
1620 break;
1621 default:
1622 return 0;
1624 return 2;
1627 /* For the rest we have to parse the commit header. */
1628 if (!c->commit_header_parsed) {
1629 msg = c->message =
1630 repo_logmsg_reencode(c->repository, commit,
1631 &c->commit_encoding, "UTF-8");
1632 parse_commit_header(c);
1635 switch (placeholder[0]) {
1636 case 'a': /* author ... */
1637 return format_person_part(sb, placeholder[1],
1638 msg + c->author.off, c->author.len,
1639 &c->pretty_ctx->date_mode);
1640 case 'c': /* committer ... */
1641 return format_person_part(sb, placeholder[1],
1642 msg + c->committer.off, c->committer.len,
1643 &c->pretty_ctx->date_mode);
1644 case 'e': /* encoding */
1645 if (c->commit_encoding)
1646 strbuf_addstr(sb, c->commit_encoding);
1647 return 1;
1648 case 'B': /* raw body */
1649 /* message_off is always left at the initial newline */
1650 strbuf_addstr(sb, msg + c->message_off + 1);
1651 return 1;
1654 /* Now we need to parse the commit message. */
1655 if (!c->commit_message_parsed)
1656 parse_commit_message(c);
1658 switch (placeholder[0]) {
1659 case 's': /* subject */
1660 format_subject(sb, msg + c->subject_off, " ");
1661 return 1;
1662 case 'f': /* sanitized subject */
1663 eol = strchrnul(msg + c->subject_off, '\n');
1664 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1665 return 1;
1666 case 'b': /* body */
1667 strbuf_addstr(sb, msg + c->body_off);
1668 return 1;
1671 if (skip_prefix(placeholder, "(trailers", &arg)) {
1672 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1673 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1674 struct strbuf sepbuf = STRBUF_INIT;
1675 struct strbuf kvsepbuf = STRBUF_INIT;
1676 size_t ret = 0;
1678 opts.no_divider = 1;
1680 if (*arg == ':') {
1681 arg++;
1682 if (format_set_trailers_options(&opts, &filter_list, &sepbuf, &kvsepbuf, &arg, NULL))
1683 goto trailer_out;
1685 if (*arg == ')') {
1686 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1687 ret = arg - placeholder + 1;
1689 trailer_out:
1690 string_list_clear(&filter_list, 0);
1691 strbuf_release(&sepbuf);
1692 return ret;
1695 return 0; /* unknown placeholder */
1698 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1699 const char *placeholder,
1700 struct format_commit_context *c)
1702 struct strbuf local_sb = STRBUF_INIT;
1703 size_t total_consumed = 0;
1704 int len, padding = c->padding;
1706 if (padding < 0) {
1707 const char *start = strrchr(sb->buf, '\n');
1708 int occupied;
1709 if (!start)
1710 start = sb->buf;
1711 occupied = utf8_strnwidth(start, strlen(start), 1);
1712 occupied += c->pretty_ctx->graph_width;
1713 padding = (-padding) - occupied;
1715 while (1) {
1716 int modifier = *placeholder == 'C';
1717 size_t consumed = format_commit_one(&local_sb, placeholder, c);
1718 total_consumed += consumed;
1720 if (!modifier)
1721 break;
1723 placeholder += consumed;
1724 if (*placeholder != '%')
1725 break;
1726 placeholder++;
1727 total_consumed++;
1729 len = utf8_strnwidth(local_sb.buf, local_sb.len, 1);
1731 if (c->flush_type == flush_left_and_steal) {
1732 const char *ch = sb->buf + sb->len - 1;
1733 while (len > padding && ch > sb->buf) {
1734 const char *p;
1735 if (*ch == ' ') {
1736 ch--;
1737 padding++;
1738 continue;
1740 /* check for trailing ansi sequences */
1741 if (*ch != 'm')
1742 break;
1743 p = ch - 1;
1744 while (p > sb->buf && ch - p < 10 && *p != '\033')
1745 p--;
1746 if (*p != '\033' ||
1747 ch + 1 - p != display_mode_esc_sequence_len(p))
1748 break;
1750 * got a good ansi sequence, put it back to
1751 * local_sb as we're cutting sb
1753 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1754 ch = p - 1;
1756 strbuf_setlen(sb, ch + 1 - sb->buf);
1757 c->flush_type = flush_left;
1760 if (len > padding) {
1761 switch (c->truncate) {
1762 case trunc_left:
1763 strbuf_utf8_replace(&local_sb,
1764 0, len - (padding - 2),
1765 "..");
1766 break;
1767 case trunc_middle:
1768 strbuf_utf8_replace(&local_sb,
1769 padding / 2 - 1,
1770 len - (padding - 2),
1771 "..");
1772 break;
1773 case trunc_right:
1774 strbuf_utf8_replace(&local_sb,
1775 padding - 2, len - (padding - 2),
1776 "..");
1777 break;
1778 case trunc_none:
1779 break;
1781 strbuf_addbuf(sb, &local_sb);
1782 } else {
1783 size_t sb_len = sb->len, offset = 0;
1784 if (c->flush_type == flush_left)
1785 offset = padding - len;
1786 else if (c->flush_type == flush_both)
1787 offset = (padding - len) / 2;
1789 * we calculate padding in columns, now
1790 * convert it back to chars
1792 padding = padding - len + local_sb.len;
1793 strbuf_addchars(sb, ' ', padding);
1794 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1795 local_sb.len);
1797 strbuf_release(&local_sb);
1798 c->flush_type = no_flush;
1799 return total_consumed;
1802 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1803 const char *placeholder,
1804 void *context)
1806 size_t consumed, orig_len;
1807 enum {
1808 NO_MAGIC,
1809 ADD_LF_BEFORE_NON_EMPTY,
1810 DEL_LF_BEFORE_EMPTY,
1811 ADD_SP_BEFORE_NON_EMPTY
1812 } magic = NO_MAGIC;
1814 switch (placeholder[0]) {
1815 case '-':
1816 magic = DEL_LF_BEFORE_EMPTY;
1817 break;
1818 case '+':
1819 magic = ADD_LF_BEFORE_NON_EMPTY;
1820 break;
1821 case ' ':
1822 magic = ADD_SP_BEFORE_NON_EMPTY;
1823 break;
1824 default:
1825 break;
1827 if (magic != NO_MAGIC) {
1828 placeholder++;
1830 switch (placeholder[0]) {
1831 case 'w':
1833 * `%+w()` cannot ever expand to a non-empty string,
1834 * and it potentially changes the layout of preceding
1835 * contents. We're thus not able to handle the magic in
1836 * this combination and refuse the pattern.
1838 return 0;
1842 orig_len = sb->len;
1843 if (((struct format_commit_context *)context)->flush_type != no_flush)
1844 consumed = format_and_pad_commit(sb, placeholder, context);
1845 else
1846 consumed = format_commit_one(sb, placeholder, context);
1847 if (magic == NO_MAGIC)
1848 return consumed;
1850 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1851 while (sb->len && sb->buf[sb->len - 1] == '\n')
1852 strbuf_setlen(sb, sb->len - 1);
1853 } else if (orig_len != sb->len) {
1854 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1855 strbuf_insertstr(sb, orig_len, "\n");
1856 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1857 strbuf_insertstr(sb, orig_len, " ");
1859 return consumed + 1;
1862 static size_t userformat_want_item(struct strbuf *sb UNUSED,
1863 const char *placeholder,
1864 void *context)
1866 struct userformat_want *w = context;
1868 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1869 placeholder++;
1871 switch (*placeholder) {
1872 case 'N':
1873 w->notes = 1;
1874 break;
1875 case 'S':
1876 w->source = 1;
1877 break;
1878 case 'd':
1879 case 'D':
1880 w->decorate = 1;
1881 break;
1883 return 0;
1886 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1888 struct strbuf dummy = STRBUF_INIT;
1890 if (!fmt) {
1891 if (!user_format)
1892 return;
1893 fmt = user_format;
1895 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1896 strbuf_release(&dummy);
1899 void repo_format_commit_message(struct repository *r,
1900 const struct commit *commit,
1901 const char *format, struct strbuf *sb,
1902 const struct pretty_print_context *pretty_ctx)
1904 struct format_commit_context context = {
1905 .repository = r,
1906 .commit = commit,
1907 .pretty_ctx = pretty_ctx,
1908 .wrap_start = sb->len
1910 const char *output_enc = pretty_ctx->output_encoding;
1911 const char *utf8 = "UTF-8";
1913 strbuf_expand(sb, format, format_commit_item, &context);
1914 rewrap_message_tail(sb, &context, 0, 0, 0);
1917 * Convert output to an actual output encoding; note that
1918 * format_commit_item() will always use UTF-8, so we don't
1919 * have to bother if that's what the output wants.
1921 if (output_enc) {
1922 if (same_encoding(utf8, output_enc))
1923 output_enc = NULL;
1924 } else {
1925 if (context.commit_encoding &&
1926 !same_encoding(context.commit_encoding, utf8))
1927 output_enc = context.commit_encoding;
1930 if (output_enc) {
1931 size_t outsz;
1932 char *out = reencode_string_len(sb->buf, sb->len,
1933 output_enc, utf8, &outsz);
1934 if (out)
1935 strbuf_attach(sb, out, outsz, outsz + 1);
1938 free(context.commit_encoding);
1939 repo_unuse_commit_buffer(r, commit, context.message);
1942 static void pp_header(struct pretty_print_context *pp,
1943 const char *encoding,
1944 const struct commit *commit,
1945 const char **msg_p,
1946 struct strbuf *sb)
1948 int parents_shown = 0;
1950 for (;;) {
1951 const char *name, *line = *msg_p;
1952 int linelen = get_one_line(*msg_p);
1954 if (!linelen)
1955 return;
1956 *msg_p += linelen;
1958 if (linelen == 1)
1959 /* End of header */
1960 return;
1962 if (pp->fmt == CMIT_FMT_RAW) {
1963 strbuf_add(sb, line, linelen);
1964 continue;
1967 if (starts_with(line, "parent ")) {
1968 if (linelen != the_hash_algo->hexsz + 8)
1969 die("bad parent line in commit");
1970 continue;
1973 if (!parents_shown) {
1974 unsigned num = commit_list_count(commit->parents);
1975 /* with enough slop */
1976 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1977 add_merge_info(pp, sb, commit);
1978 parents_shown = 1;
1982 * MEDIUM == DEFAULT shows only author with dates.
1983 * FULL shows both authors but not dates.
1984 * FULLER shows both authors and dates.
1986 if (skip_prefix(line, "author ", &name)) {
1987 strbuf_grow(sb, linelen + 80);
1988 pp_user_info(pp, "Author", sb, name, encoding);
1990 if (skip_prefix(line, "committer ", &name) &&
1991 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1992 strbuf_grow(sb, linelen + 80);
1993 pp_user_info(pp, "Commit", sb, name, encoding);
1998 void pp_title_line(struct pretty_print_context *pp,
1999 const char **msg_p,
2000 struct strbuf *sb,
2001 const char *encoding,
2002 int need_8bit_cte)
2004 static const int max_length = 78; /* per rfc2047 */
2005 struct strbuf title;
2007 strbuf_init(&title, 80);
2008 *msg_p = format_subject(&title, *msg_p,
2009 pp->preserve_subject ? "\n" : " ");
2011 strbuf_grow(sb, title.len + 1024);
2012 if (pp->print_email_subject) {
2013 if (pp->rev)
2014 fmt_output_email_subject(sb, pp->rev);
2015 if (pp->encode_email_headers &&
2016 needs_rfc2047_encoding(title.buf, title.len))
2017 add_rfc2047(sb, title.buf, title.len,
2018 encoding, RFC2047_SUBJECT);
2019 else
2020 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
2021 -last_line_length(sb), 1, max_length);
2022 } else {
2023 strbuf_addbuf(sb, &title);
2025 strbuf_addch(sb, '\n');
2027 if (need_8bit_cte == 0) {
2028 int i;
2029 for (i = 0; i < pp->in_body_headers.nr; i++) {
2030 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
2031 need_8bit_cte = 1;
2032 break;
2037 if (need_8bit_cte > 0) {
2038 const char *header_fmt =
2039 "MIME-Version: 1.0\n"
2040 "Content-Type: text/plain; charset=%s\n"
2041 "Content-Transfer-Encoding: 8bit\n";
2042 strbuf_addf(sb, header_fmt, encoding);
2044 if (pp->after_subject) {
2045 strbuf_addstr(sb, pp->after_subject);
2047 if (cmit_fmt_is_mail(pp->fmt)) {
2048 strbuf_addch(sb, '\n');
2051 if (pp->in_body_headers.nr) {
2052 int i;
2053 for (i = 0; i < pp->in_body_headers.nr; i++) {
2054 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
2055 free(pp->in_body_headers.items[i].string);
2057 string_list_clear(&pp->in_body_headers, 0);
2058 strbuf_addch(sb, '\n');
2061 strbuf_release(&title);
2064 static int pp_utf8_width(const char *start, const char *end)
2066 int width = 0;
2067 size_t remain = end - start;
2069 while (remain) {
2070 int n = utf8_width(&start, &remain);
2071 if (n < 0 || !start)
2072 return -1;
2073 width += n;
2075 return width;
2078 static void strbuf_add_tabexpand(struct strbuf *sb, struct grep_opt *opt,
2079 int color, int tabwidth, const char *line,
2080 int linelen)
2082 const char *tab;
2084 while ((tab = memchr(line, '\t', linelen)) != NULL) {
2085 int width = pp_utf8_width(line, tab);
2088 * If it wasn't well-formed utf8, or it
2089 * had characters with badly defined
2090 * width (control characters etc), just
2091 * give up on trying to align things.
2093 if (width < 0)
2094 break;
2096 /* Output the data .. */
2097 append_line_with_color(sb, opt, line, tab - line, color,
2098 GREP_CONTEXT_BODY,
2099 GREP_HEADER_FIELD_MAX);
2101 /* .. and the de-tabified tab */
2102 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
2104 /* Skip over the printed part .. */
2105 linelen -= tab + 1 - line;
2106 line = tab + 1;
2110 * Print out everything after the last tab without
2111 * worrying about width - there's nothing more to
2112 * align.
2114 append_line_with_color(sb, opt, line, linelen, color, GREP_CONTEXT_BODY,
2115 GREP_HEADER_FIELD_MAX);
2119 * pp_handle_indent() prints out the intendation, and
2120 * the whole line (without the final newline), after
2121 * de-tabifying.
2123 static void pp_handle_indent(struct pretty_print_context *pp,
2124 struct strbuf *sb, int indent,
2125 const char *line, int linelen)
2127 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
2129 strbuf_addchars(sb, ' ', indent);
2130 if (pp->expand_tabs_in_log)
2131 strbuf_add_tabexpand(sb, opt, pp->color, pp->expand_tabs_in_log,
2132 line, linelen);
2133 else
2134 append_line_with_color(sb, opt, line, linelen, pp->color,
2135 GREP_CONTEXT_BODY,
2136 GREP_HEADER_FIELD_MAX);
2139 static int is_mboxrd_from(const char *line, int len)
2142 * a line matching /^From $/ here would only have len == 4
2143 * at this point because is_empty_line would've trimmed all
2144 * trailing space
2146 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
2149 void pp_remainder(struct pretty_print_context *pp,
2150 const char **msg_p,
2151 struct strbuf *sb,
2152 int indent)
2154 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
2155 int first = 1;
2157 for (;;) {
2158 const char *line = *msg_p;
2159 int linelen = get_one_line(line);
2160 *msg_p += linelen;
2162 if (!linelen)
2163 break;
2165 if (is_blank_line(line, &linelen)) {
2166 if (first)
2167 continue;
2168 if (pp->fmt == CMIT_FMT_SHORT)
2169 break;
2171 first = 0;
2173 strbuf_grow(sb, linelen + indent + 20);
2174 if (indent)
2175 pp_handle_indent(pp, sb, indent, line, linelen);
2176 else if (pp->expand_tabs_in_log)
2177 strbuf_add_tabexpand(sb, opt, pp->color,
2178 pp->expand_tabs_in_log, line,
2179 linelen);
2180 else {
2181 if (pp->fmt == CMIT_FMT_MBOXRD &&
2182 is_mboxrd_from(line, linelen))
2183 strbuf_addch(sb, '>');
2185 append_line_with_color(sb, opt, line, linelen,
2186 pp->color, GREP_CONTEXT_BODY,
2187 GREP_HEADER_FIELD_MAX);
2189 strbuf_addch(sb, '\n');
2193 void pretty_print_commit(struct pretty_print_context *pp,
2194 const struct commit *commit,
2195 struct strbuf *sb)
2197 unsigned long beginning_of_body;
2198 int indent = 4;
2199 const char *msg;
2200 const char *reencoded;
2201 const char *encoding;
2202 int need_8bit_cte = pp->need_8bit_cte;
2204 if (pp->fmt == CMIT_FMT_USERFORMAT) {
2205 format_commit_message(commit, user_format, sb, pp);
2206 return;
2209 encoding = get_log_output_encoding();
2210 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
2212 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2213 indent = 0;
2216 * We need to check and emit Content-type: to mark it
2217 * as 8-bit if we haven't done so.
2219 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
2220 int i, ch, in_body;
2222 for (in_body = i = 0; (ch = msg[i]); i++) {
2223 if (!in_body) {
2224 /* author could be non 7-bit ASCII but
2225 * the log may be so; skip over the
2226 * header part first.
2228 if (ch == '\n' && msg[i+1] == '\n')
2229 in_body = 1;
2231 else if (non_ascii(ch)) {
2232 need_8bit_cte = 1;
2233 break;
2238 pp_header(pp, encoding, commit, &msg, sb);
2239 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
2240 strbuf_addch(sb, '\n');
2243 /* Skip excess blank lines at the beginning of body, if any... */
2244 msg = skip_blank_lines(msg);
2246 /* These formats treat the title line specially. */
2247 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2248 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2250 beginning_of_body = sb->len;
2251 if (pp->fmt != CMIT_FMT_ONELINE)
2252 pp_remainder(pp, &msg, sb, indent);
2253 strbuf_rtrim(sb);
2255 /* Make sure there is an EOLN for the non-oneline case */
2256 if (pp->fmt != CMIT_FMT_ONELINE)
2257 strbuf_addch(sb, '\n');
2260 * The caller may append additional body text in e-mail
2261 * format. Make sure we did not strip the blank line
2262 * between the header and the body.
2264 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2265 strbuf_addch(sb, '\n');
2267 unuse_commit_buffer(commit, reencoded);
2270 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2271 struct strbuf *sb)
2273 struct pretty_print_context pp = {0};
2274 pp.fmt = fmt;
2275 pretty_print_commit(&pp, commit, sb);