pack-bitmap: prepare to read lookup table extension
[git/debian.git] / pretty.c
blob6d819103fbf1db3a201f993e20140bfa288a0e12
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, void *cb)
48 struct cmt_fmt_map *commit_format = NULL;
49 const char *name;
50 const char *fmt;
51 int i;
53 if (!skip_prefix(var, "pretty.", &name))
54 return 0;
56 for (i = 0; i < builtin_formats_len; i++) {
57 if (!strcmp(commit_formats[i].name, name))
58 return 0;
61 for (i = builtin_formats_len; i < commit_formats_len; i++) {
62 if (!strcmp(commit_formats[i].name, name)) {
63 commit_format = &commit_formats[i];
64 break;
68 if (!commit_format) {
69 ALLOC_GROW(commit_formats, commit_formats_len+1,
70 commit_formats_alloc);
71 commit_format = &commit_formats[commit_formats_len];
72 memset(commit_format, 0, sizeof(*commit_format));
73 commit_formats_len++;
76 commit_format->name = xstrdup(name);
77 commit_format->format = CMIT_FMT_USERFORMAT;
78 if (git_config_string(&fmt, var, value))
79 return -1;
81 if (skip_prefix(fmt, "format:", &fmt))
82 commit_format->is_tformat = 0;
83 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
84 commit_format->is_tformat = 1;
85 else
86 commit_format->is_alias = 1;
87 commit_format->user_format = fmt;
89 return 0;
92 static void setup_commit_formats(void)
94 struct cmt_fmt_map builtin_formats[] = {
95 { "raw", CMIT_FMT_RAW, 0, 0 },
96 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
97 { "short", CMIT_FMT_SHORT, 0, 0 },
98 { "email", CMIT_FMT_EMAIL, 0, 0 },
99 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
100 { "fuller", CMIT_FMT_FULLER, 0, 8 },
101 { "full", CMIT_FMT_FULL, 0, 8 },
102 { "oneline", CMIT_FMT_ONELINE, 1, 0 },
103 { "reference", CMIT_FMT_USERFORMAT, 1, 0,
104 0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
106 * Please update $__git_log_pretty_formats in
107 * git-completion.bash when you add new formats.
110 commit_formats_len = ARRAY_SIZE(builtin_formats);
111 builtin_formats_len = commit_formats_len;
112 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
113 COPY_ARRAY(commit_formats, builtin_formats,
114 ARRAY_SIZE(builtin_formats));
116 git_config(git_pretty_formats_config, NULL);
119 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
120 const char *original,
121 int num_redirections)
123 struct cmt_fmt_map *found = NULL;
124 size_t found_match_len = 0;
125 int i;
127 if (num_redirections >= commit_formats_len)
128 die("invalid --pretty format: "
129 "'%s' references an alias which points to itself",
130 original);
132 for (i = 0; i < commit_formats_len; i++) {
133 size_t match_len;
135 if (!starts_with(commit_formats[i].name, sought))
136 continue;
138 match_len = strlen(commit_formats[i].name);
139 if (found == NULL || found_match_len > match_len) {
140 found = &commit_formats[i];
141 found_match_len = match_len;
145 if (found && found->is_alias) {
146 found = find_commit_format_recursive(found->user_format,
147 original,
148 num_redirections+1);
151 return found;
154 static struct cmt_fmt_map *find_commit_format(const char *sought)
156 if (!commit_formats)
157 setup_commit_formats();
159 return find_commit_format_recursive(sought, sought, 0);
162 void get_commit_format(const char *arg, struct rev_info *rev)
164 struct cmt_fmt_map *commit_format;
166 rev->use_terminator = 0;
167 if (!arg) {
168 rev->commit_format = CMIT_FMT_DEFAULT;
169 return;
171 if (skip_prefix(arg, "format:", &arg)) {
172 save_user_format(rev, arg, 0);
173 return;
176 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
177 save_user_format(rev, arg, 1);
178 return;
181 commit_format = find_commit_format(arg);
182 if (!commit_format)
183 die("invalid --pretty format: %s", arg);
185 rev->commit_format = commit_format->format;
186 rev->use_terminator = commit_format->is_tformat;
187 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
188 if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
189 rev->date_mode.type = commit_format->default_date_mode_type;
190 if (commit_format->format == CMIT_FMT_USERFORMAT) {
191 save_user_format(rev, commit_format->user_format,
192 commit_format->is_tformat);
197 * Generic support for pretty-printing the header
199 static int get_one_line(const char *msg)
201 int ret = 0;
203 for (;;) {
204 char c = *msg++;
205 if (!c)
206 break;
207 ret++;
208 if (c == '\n')
209 break;
211 return ret;
214 /* High bit set, or ISO-2022-INT */
215 static int non_ascii(int ch)
217 return !isascii(ch) || ch == '\033';
220 int has_non_ascii(const char *s)
222 int ch;
223 if (!s)
224 return 0;
225 while ((ch = *s++) != '\0') {
226 if (non_ascii(ch))
227 return 1;
229 return 0;
232 static int is_rfc822_special(char ch)
234 switch (ch) {
235 case '(':
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 return 1;
249 default:
250 return 0;
254 static int needs_rfc822_quoting(const char *s, int len)
256 int i;
257 for (i = 0; i < len; i++)
258 if (is_rfc822_special(s[i]))
259 return 1;
260 return 0;
263 static int last_line_length(struct strbuf *sb)
265 int i;
267 /* How many bytes are already used on the last line? */
268 for (i = sb->len - 1; i >= 0; i--)
269 if (sb->buf[i] == '\n')
270 break;
271 return sb->len - (i + 1);
274 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
276 int i;
278 /* just a guess, we may have to also backslash-quote */
279 strbuf_grow(out, len + 2);
281 strbuf_addch(out, '"');
282 for (i = 0; i < len; i++) {
283 switch (s[i]) {
284 case '"':
285 case '\\':
286 strbuf_addch(out, '\\');
287 /* fall through */
288 default:
289 strbuf_addch(out, s[i]);
292 strbuf_addch(out, '"');
295 enum rfc2047_type {
296 RFC2047_SUBJECT,
297 RFC2047_ADDRESS
300 static int is_rfc2047_special(char ch, enum rfc2047_type type)
303 * rfc2047, section 4.2:
305 * 8-bit values which correspond to printable ASCII characters other
306 * than "=", "?", and "_" (underscore), MAY be represented as those
307 * characters. (But see section 5 for restrictions.) In
308 * particular, SPACE and TAB MUST NOT be represented as themselves
309 * within encoded words.
313 * rule out non-ASCII characters and non-printable characters (the
314 * non-ASCII check should be redundant as isprint() is not localized
315 * and only knows about ASCII, but be defensive about that)
317 if (non_ascii(ch) || !isprint(ch))
318 return 1;
321 * rule out special printable characters (' ' should be the only
322 * whitespace character considered printable, but be defensive and use
323 * isspace())
325 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
326 return 1;
329 * rfc2047, section 5.3:
331 * As a replacement for a 'word' entity within a 'phrase', for example,
332 * one that precedes an address in a From, To, or Cc header. The ABNF
333 * definition for 'phrase' from RFC 822 thus becomes:
335 * phrase = 1*( encoded-word / word )
337 * In this case the set of characters that may be used in a "Q"-encoded
338 * 'encoded-word' is restricted to: <upper and lower case ASCII
339 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
340 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
341 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
342 * 'special' by 'linear-white-space'.
345 if (type != RFC2047_ADDRESS)
346 return 0;
348 /* '=' and '_' are special cases and have been checked above */
349 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
352 static int needs_rfc2047_encoding(const char *line, int len)
354 int i;
356 for (i = 0; i < len; i++) {
357 int ch = line[i];
358 if (non_ascii(ch) || ch == '\n')
359 return 1;
360 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
361 return 1;
364 return 0;
367 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
368 const char *encoding, enum rfc2047_type type)
370 static const int max_encoded_length = 76; /* per rfc2047 */
371 int i;
372 int line_len = last_line_length(sb);
374 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
375 strbuf_addf(sb, "=?%s?q?", encoding);
376 line_len += strlen(encoding) + 5; /* 5 for =??q? */
378 while (len) {
380 * RFC 2047, section 5 (3):
382 * Each 'encoded-word' MUST represent an integral number of
383 * characters. A multi-octet character may not be split across
384 * adjacent 'encoded- word's.
386 const unsigned char *p = (const unsigned char *)line;
387 int chrlen = mbs_chrlen(&line, &len, encoding);
388 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
390 /* "=%02X" * chrlen, or the byte itself */
391 const char *encoded_fmt = is_special ? "=%02X" : "%c";
392 int encoded_len = is_special ? 3 * chrlen : 1;
395 * According to RFC 2047, we could encode the special character
396 * ' ' (space) with '_' (underscore) for readability. But many
397 * programs do not understand this and just leave the
398 * underscore in place. Thus, we do nothing special here, which
399 * causes ' ' to be encoded as '=20', avoiding this problem.
402 if (line_len + encoded_len + 2 > max_encoded_length) {
403 /* It won't fit with trailing "?=" --- break the line */
404 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
405 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
408 for (i = 0; i < chrlen; i++)
409 strbuf_addf(sb, encoded_fmt, p[i]);
410 line_len += encoded_len;
412 strbuf_addstr(sb, "?=");
415 const char *show_ident_date(const struct ident_split *ident,
416 const struct date_mode *mode)
418 timestamp_t date = 0;
419 long tz = 0;
421 if (ident->date_begin && ident->date_end)
422 date = parse_timestamp(ident->date_begin, NULL, 10);
423 if (date_overflows(date))
424 date = 0;
425 else {
426 if (ident->tz_begin && ident->tz_end)
427 tz = strtol(ident->tz_begin, NULL, 10);
428 if (tz >= INT_MAX || tz <= INT_MIN)
429 tz = 0;
431 return show_date(date, tz, mode);
434 static inline void strbuf_add_with_color(struct strbuf *sb, const char *color,
435 const char *buf, size_t buflen)
437 strbuf_addstr(sb, color);
438 strbuf_add(sb, buf, buflen);
439 if (*color)
440 strbuf_addstr(sb, GIT_COLOR_RESET);
443 static void append_line_with_color(struct strbuf *sb, struct grep_opt *opt,
444 const char *line, size_t linelen,
445 int color, enum grep_context ctx,
446 enum grep_header_field field)
448 const char *buf, *eol, *line_color, *match_color;
449 regmatch_t match;
450 int eflags = 0;
452 buf = line;
453 eol = buf + linelen;
455 if (!opt || !want_color(color) || opt->invert)
456 goto end;
458 line_color = opt->colors[GREP_COLOR_SELECTED];
459 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
461 while (grep_next_match(opt, buf, eol, ctx, &match, field, eflags)) {
462 if (match.rm_so == match.rm_eo)
463 break;
465 strbuf_add_with_color(sb, line_color, buf, match.rm_so);
466 strbuf_add_with_color(sb, match_color, buf + match.rm_so,
467 match.rm_eo - match.rm_so);
468 buf += match.rm_eo;
469 eflags = REG_NOTBOL;
472 if (eflags)
473 strbuf_add_with_color(sb, line_color, buf, eol - buf);
474 else {
475 end:
476 strbuf_add(sb, buf, eol - buf);
480 void pp_user_info(struct pretty_print_context *pp,
481 const char *what, struct strbuf *sb,
482 const char *line, const char *encoding)
484 struct ident_split ident;
485 char *line_end;
486 const char *mailbuf, *namebuf;
487 size_t namelen, maillen;
488 int max_length = 78; /* per rfc2822 */
490 if (pp->fmt == CMIT_FMT_ONELINE)
491 return;
493 line_end = strchrnul(line, '\n');
494 if (split_ident_line(&ident, line, line_end - line))
495 return;
497 mailbuf = ident.mail_begin;
498 maillen = ident.mail_end - ident.mail_begin;
499 namebuf = ident.name_begin;
500 namelen = ident.name_end - ident.name_begin;
502 if (pp->mailmap)
503 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
505 if (cmit_fmt_is_mail(pp->fmt)) {
506 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
507 struct strbuf buf = STRBUF_INIT;
509 strbuf_addstr(&buf, "From: ");
510 strbuf_add(&buf, namebuf, namelen);
511 strbuf_addstr(&buf, " <");
512 strbuf_add(&buf, mailbuf, maillen);
513 strbuf_addstr(&buf, ">\n");
514 string_list_append(&pp->in_body_headers,
515 strbuf_detach(&buf, NULL));
517 mailbuf = pp->from_ident->mail_begin;
518 maillen = pp->from_ident->mail_end - mailbuf;
519 namebuf = pp->from_ident->name_begin;
520 namelen = pp->from_ident->name_end - namebuf;
523 strbuf_addstr(sb, "From: ");
524 if (pp->encode_email_headers &&
525 needs_rfc2047_encoding(namebuf, namelen)) {
526 add_rfc2047(sb, namebuf, namelen,
527 encoding, RFC2047_ADDRESS);
528 max_length = 76; /* per rfc2047 */
529 } else if (needs_rfc822_quoting(namebuf, namelen)) {
530 struct strbuf quoted = STRBUF_INIT;
531 add_rfc822_quoted(&quoted, namebuf, namelen);
532 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
533 -6, 1, max_length);
534 strbuf_release(&quoted);
535 } else {
536 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
537 -6, 1, max_length);
540 if (max_length <
541 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
542 strbuf_addch(sb, '\n');
543 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
544 } else {
545 struct strbuf id = STRBUF_INIT;
546 enum grep_header_field field = GREP_HEADER_FIELD_MAX;
547 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
549 if (!strcmp(what, "Author"))
550 field = GREP_HEADER_AUTHOR;
551 else if (!strcmp(what, "Commit"))
552 field = GREP_HEADER_COMMITTER;
554 strbuf_addf(sb, "%s: ", what);
555 if (pp->fmt == CMIT_FMT_FULLER)
556 strbuf_addchars(sb, ' ', 4);
558 strbuf_addf(&id, "%.*s <%.*s>", (int)namelen, namebuf,
559 (int)maillen, mailbuf);
561 append_line_with_color(sb, opt, id.buf, id.len, pp->color,
562 GREP_CONTEXT_HEAD, field);
563 strbuf_addch(sb, '\n');
564 strbuf_release(&id);
567 switch (pp->fmt) {
568 case CMIT_FMT_MEDIUM:
569 strbuf_addf(sb, "Date: %s\n",
570 show_ident_date(&ident, &pp->date_mode));
571 break;
572 case CMIT_FMT_EMAIL:
573 case CMIT_FMT_MBOXRD:
574 strbuf_addf(sb, "Date: %s\n",
575 show_ident_date(&ident, DATE_MODE(RFC2822)));
576 break;
577 case CMIT_FMT_FULLER:
578 strbuf_addf(sb, "%sDate: %s\n", what,
579 show_ident_date(&ident, &pp->date_mode));
580 break;
581 default:
582 /* notin' */
583 break;
587 static int is_blank_line(const char *line, int *len_p)
589 int len = *len_p;
590 while (len && isspace(line[len - 1]))
591 len--;
592 *len_p = len;
593 return !len;
596 const char *skip_blank_lines(const char *msg)
598 for (;;) {
599 int linelen = get_one_line(msg);
600 int ll = linelen;
601 if (!linelen)
602 break;
603 if (!is_blank_line(msg, &ll))
604 break;
605 msg += linelen;
607 return msg;
610 static void add_merge_info(const struct pretty_print_context *pp,
611 struct strbuf *sb, const struct commit *commit)
613 struct commit_list *parent = commit->parents;
615 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
616 !parent || !parent->next)
617 return;
619 strbuf_addstr(sb, "Merge:");
621 while (parent) {
622 struct object_id *oidp = &parent->item->object.oid;
623 strbuf_addch(sb, ' ');
624 if (pp->abbrev)
625 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
626 else
627 strbuf_addstr(sb, oid_to_hex(oidp));
628 parent = parent->next;
630 strbuf_addch(sb, '\n');
633 static char *get_header(const char *msg, const char *key)
635 size_t len;
636 const char *v = find_commit_header(msg, key, &len);
637 return v ? xmemdupz(v, len) : NULL;
640 static char *replace_encoding_header(char *buf, const char *encoding)
642 struct strbuf tmp = STRBUF_INIT;
643 size_t start, len;
644 char *cp = buf;
646 /* guess if there is an encoding header before a \n\n */
647 while (!starts_with(cp, "encoding ")) {
648 cp = strchr(cp, '\n');
649 if (!cp || *++cp == '\n')
650 return buf;
652 start = cp - buf;
653 cp = strchr(cp, '\n');
654 if (!cp)
655 return buf; /* should not happen but be defensive */
656 len = cp + 1 - (buf + start);
658 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
659 if (is_encoding_utf8(encoding)) {
660 /* we have re-coded to UTF-8; drop the header */
661 strbuf_remove(&tmp, start, len);
662 } else {
663 /* just replaces XXXX in 'encoding XXXX\n' */
664 strbuf_splice(&tmp, start + strlen("encoding "),
665 len - strlen("encoding \n"),
666 encoding, strlen(encoding));
668 return strbuf_detach(&tmp, NULL);
671 const char *repo_logmsg_reencode(struct repository *r,
672 const struct commit *commit,
673 char **commit_encoding,
674 const char *output_encoding)
676 static const char *utf8 = "UTF-8";
677 const char *use_encoding;
678 char *encoding;
679 const char *msg = repo_get_commit_buffer(r, commit, NULL);
680 char *out;
682 if (!output_encoding || !*output_encoding) {
683 if (commit_encoding)
684 *commit_encoding = get_header(msg, "encoding");
685 return msg;
687 encoding = get_header(msg, "encoding");
688 if (commit_encoding)
689 *commit_encoding = encoding;
690 use_encoding = encoding ? encoding : utf8;
691 if (same_encoding(use_encoding, output_encoding)) {
693 * No encoding work to be done. If we have no encoding header
694 * at all, then there's nothing to do, and we can return the
695 * message verbatim (whether newly allocated or not).
697 if (!encoding)
698 return msg;
701 * Otherwise, we still want to munge the encoding header in the
702 * result, which will be done by modifying the buffer. If we
703 * are using a fresh copy, we can reuse it. But if we are using
704 * the cached copy from get_commit_buffer, we need to duplicate it
705 * to avoid munging the cached copy.
707 if (msg == get_cached_commit_buffer(r, commit, NULL))
708 out = xstrdup(msg);
709 else
710 out = (char *)msg;
712 else {
714 * There's actual encoding work to do. Do the reencoding, which
715 * still leaves the header to be replaced in the next step. At
716 * this point, we are done with msg. If we allocated a fresh
717 * copy, we can free it.
719 out = reencode_string(msg, output_encoding, use_encoding);
720 if (out)
721 repo_unuse_commit_buffer(r, commit, msg);
725 * This replacement actually consumes the buffer we hand it, so we do
726 * not have to worry about freeing the old "out" here.
728 if (out)
729 out = replace_encoding_header(out, output_encoding);
731 if (!commit_encoding)
732 free(encoding);
734 * If the re-encoding failed, out might be NULL here; in that
735 * case we just return the commit message verbatim.
737 return out ? out : msg;
740 static int mailmap_name(const char **email, size_t *email_len,
741 const char **name, size_t *name_len)
743 static struct string_list *mail_map;
744 if (!mail_map) {
745 CALLOC_ARRAY(mail_map, 1);
746 read_mailmap(mail_map);
748 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
751 static size_t format_person_part(struct strbuf *sb, char part,
752 const char *msg, int len,
753 const struct date_mode *dmode)
755 /* currently all placeholders have same length */
756 const int placeholder_len = 2;
757 struct ident_split s;
758 const char *name, *mail;
759 size_t maillen, namelen;
761 if (split_ident_line(&s, msg, len) < 0)
762 goto skip;
764 name = s.name_begin;
765 namelen = s.name_end - s.name_begin;
766 mail = s.mail_begin;
767 maillen = s.mail_end - s.mail_begin;
769 if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
770 mailmap_name(&mail, &maillen, &name, &namelen);
771 if (part == 'n' || part == 'N') { /* name */
772 strbuf_add(sb, name, namelen);
773 return placeholder_len;
775 if (part == 'e' || part == 'E') { /* email */
776 strbuf_add(sb, mail, maillen);
777 return placeholder_len;
779 if (part == 'l' || part == 'L') { /* local-part */
780 const char *at = memchr(mail, '@', maillen);
781 if (at)
782 maillen = at - mail;
783 strbuf_add(sb, mail, maillen);
784 return placeholder_len;
787 if (!s.date_begin)
788 goto skip;
790 if (part == 't') { /* date, UNIX timestamp */
791 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
792 return placeholder_len;
795 switch (part) {
796 case 'd': /* date */
797 strbuf_addstr(sb, show_ident_date(&s, dmode));
798 return placeholder_len;
799 case 'D': /* date, RFC2822 style */
800 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
801 return placeholder_len;
802 case 'r': /* date, relative */
803 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
804 return placeholder_len;
805 case 'i': /* date, ISO 8601-like */
806 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
807 return placeholder_len;
808 case 'I': /* date, ISO 8601 strict */
809 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
810 return placeholder_len;
811 case 'h': /* date, human */
812 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(HUMAN)));
813 return placeholder_len;
814 case 's':
815 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
816 return placeholder_len;
819 skip:
821 * reading from either a bogus commit, or a reflog entry with
822 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
823 * to compute a valid return value.
825 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
826 || part == 'D' || part == 'r' || part == 'i')
827 return placeholder_len;
829 return 0; /* unknown placeholder */
832 struct chunk {
833 size_t off;
834 size_t len;
837 enum flush_type {
838 no_flush,
839 flush_right,
840 flush_left,
841 flush_left_and_steal,
842 flush_both
845 enum trunc_type {
846 trunc_none,
847 trunc_left,
848 trunc_middle,
849 trunc_right
852 struct format_commit_context {
853 struct repository *repository;
854 const struct commit *commit;
855 const struct pretty_print_context *pretty_ctx;
856 unsigned commit_header_parsed:1;
857 unsigned commit_message_parsed:1;
858 struct signature_check signature_check;
859 enum flush_type flush_type;
860 enum trunc_type truncate;
861 const char *message;
862 char *commit_encoding;
863 size_t width, indent1, indent2;
864 int auto_color;
865 int padding;
867 /* These offsets are relative to the start of the commit message. */
868 struct chunk author;
869 struct chunk committer;
870 size_t message_off;
871 size_t subject_off;
872 size_t body_off;
874 /* The following ones are relative to the result struct strbuf. */
875 size_t wrap_start;
878 static void parse_commit_header(struct format_commit_context *context)
880 const char *msg = context->message;
881 int i;
883 for (i = 0; msg[i]; i++) {
884 const char *name;
885 int eol;
886 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
887 ; /* do nothing */
889 if (i == eol) {
890 break;
891 } else if (skip_prefix(msg + i, "author ", &name)) {
892 context->author.off = name - msg;
893 context->author.len = msg + eol - name;
894 } else if (skip_prefix(msg + i, "committer ", &name)) {
895 context->committer.off = name - msg;
896 context->committer.len = msg + eol - name;
898 i = eol;
900 context->message_off = i;
901 context->commit_header_parsed = 1;
904 static int istitlechar(char c)
906 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
907 (c >= '0' && c <= '9') || c == '.' || c == '_';
910 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
912 size_t trimlen;
913 size_t start_len = sb->len;
914 int space = 2;
915 int i;
917 for (i = 0; i < len; i++) {
918 if (istitlechar(msg[i])) {
919 if (space == 1)
920 strbuf_addch(sb, '-');
921 space = 0;
922 strbuf_addch(sb, msg[i]);
923 if (msg[i] == '.')
924 while (msg[i+1] == '.')
925 i++;
926 } else
927 space |= 1;
930 /* trim any trailing '.' or '-' characters */
931 trimlen = 0;
932 while (sb->len - trimlen > start_len &&
933 (sb->buf[sb->len - 1 - trimlen] == '.'
934 || sb->buf[sb->len - 1 - trimlen] == '-'))
935 trimlen++;
936 strbuf_remove(sb, sb->len - trimlen, trimlen);
939 const char *format_subject(struct strbuf *sb, const char *msg,
940 const char *line_separator)
942 int first = 1;
944 for (;;) {
945 const char *line = msg;
946 int linelen = get_one_line(line);
948 msg += linelen;
949 if (!linelen || is_blank_line(line, &linelen))
950 break;
952 if (!sb)
953 continue;
954 strbuf_grow(sb, linelen + 2);
955 if (!first)
956 strbuf_addstr(sb, line_separator);
957 strbuf_add(sb, line, linelen);
958 first = 0;
960 return msg;
963 static void parse_commit_message(struct format_commit_context *c)
965 const char *msg = c->message + c->message_off;
966 const char *start = c->message;
968 msg = skip_blank_lines(msg);
969 c->subject_off = msg - start;
971 msg = format_subject(NULL, msg, NULL);
972 msg = skip_blank_lines(msg);
973 c->body_off = msg - start;
975 c->commit_message_parsed = 1;
978 static void strbuf_wrap(struct strbuf *sb, size_t pos,
979 size_t width, size_t indent1, size_t indent2)
981 struct strbuf tmp = STRBUF_INIT;
983 if (pos)
984 strbuf_add(&tmp, sb->buf, pos);
985 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
986 (int) indent1, (int) indent2, (int) width);
987 strbuf_swap(&tmp, sb);
988 strbuf_release(&tmp);
991 static void rewrap_message_tail(struct strbuf *sb,
992 struct format_commit_context *c,
993 size_t new_width, size_t new_indent1,
994 size_t new_indent2)
996 if (c->width == new_width && c->indent1 == new_indent1 &&
997 c->indent2 == new_indent2)
998 return;
999 if (c->wrap_start < sb->len)
1000 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
1001 c->wrap_start = sb->len;
1002 c->width = new_width;
1003 c->indent1 = new_indent1;
1004 c->indent2 = new_indent2;
1007 static int format_reflog_person(struct strbuf *sb,
1008 char part,
1009 struct reflog_walk_info *log,
1010 const struct date_mode *dmode)
1012 const char *ident;
1014 if (!log)
1015 return 2;
1017 ident = get_reflog_ident(log);
1018 if (!ident)
1019 return 2;
1021 return format_person_part(sb, part, ident, strlen(ident), dmode);
1024 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
1025 const char *placeholder,
1026 struct format_commit_context *c)
1028 const char *rest = placeholder;
1029 const char *basic_color = NULL;
1031 if (placeholder[1] == '(') {
1032 const char *begin = placeholder + 2;
1033 const char *end = strchr(begin, ')');
1034 char color[COLOR_MAXLEN];
1036 if (!end)
1037 return 0;
1039 if (skip_prefix(begin, "auto,", &begin)) {
1040 if (!want_color(c->pretty_ctx->color))
1041 return end - placeholder + 1;
1042 } else if (skip_prefix(begin, "always,", &begin)) {
1043 /* nothing to do; we do not respect want_color at all */
1044 } else {
1045 /* the default is the same as "auto" */
1046 if (!want_color(c->pretty_ctx->color))
1047 return end - placeholder + 1;
1050 if (color_parse_mem(begin, end - begin, color) < 0)
1051 die(_("unable to parse --pretty format"));
1052 strbuf_addstr(sb, color);
1053 return end - placeholder + 1;
1057 * We handle things like "%C(red)" above; for historical reasons, there
1058 * are a few colors that can be specified without parentheses (and
1059 * they cannot support things like "auto" or "always" at all).
1061 if (skip_prefix(placeholder + 1, "red", &rest))
1062 basic_color = GIT_COLOR_RED;
1063 else if (skip_prefix(placeholder + 1, "green", &rest))
1064 basic_color = GIT_COLOR_GREEN;
1065 else if (skip_prefix(placeholder + 1, "blue", &rest))
1066 basic_color = GIT_COLOR_BLUE;
1067 else if (skip_prefix(placeholder + 1, "reset", &rest))
1068 basic_color = GIT_COLOR_RESET;
1070 if (basic_color && want_color(c->pretty_ctx->color))
1071 strbuf_addstr(sb, basic_color);
1073 return rest - placeholder;
1076 static size_t parse_padding_placeholder(const char *placeholder,
1077 struct format_commit_context *c)
1079 const char *ch = placeholder;
1080 enum flush_type flush_type;
1081 int to_column = 0;
1083 switch (*ch++) {
1084 case '<':
1085 flush_type = flush_right;
1086 break;
1087 case '>':
1088 if (*ch == '<') {
1089 flush_type = flush_both;
1090 ch++;
1091 } else if (*ch == '>') {
1092 flush_type = flush_left_and_steal;
1093 ch++;
1094 } else
1095 flush_type = flush_left;
1096 break;
1097 default:
1098 return 0;
1101 /* the next value means "wide enough to that column" */
1102 if (*ch == '|') {
1103 to_column = 1;
1104 ch++;
1107 if (*ch == '(') {
1108 const char *start = ch + 1;
1109 const char *end = start + strcspn(start, ",)");
1110 char *next;
1111 int width;
1112 if (!end || end == start)
1113 return 0;
1114 width = strtol(start, &next, 10);
1115 if (next == start || width == 0)
1116 return 0;
1117 if (width < 0) {
1118 if (to_column)
1119 width += term_columns();
1120 if (width < 0)
1121 return 0;
1123 c->padding = to_column ? -width : width;
1124 c->flush_type = flush_type;
1126 if (*end == ',') {
1127 start = end + 1;
1128 end = strchr(start, ')');
1129 if (!end || end == start)
1130 return 0;
1131 if (starts_with(start, "trunc)"))
1132 c->truncate = trunc_right;
1133 else if (starts_with(start, "ltrunc)"))
1134 c->truncate = trunc_left;
1135 else if (starts_with(start, "mtrunc)"))
1136 c->truncate = trunc_middle;
1137 else
1138 return 0;
1139 } else
1140 c->truncate = trunc_none;
1142 return end - placeholder + 1;
1144 return 0;
1147 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1148 const char **end, const char **valuestart,
1149 size_t *valuelen)
1151 const char *p;
1153 if (!(skip_prefix(to_parse, candidate, &p)))
1154 return 0;
1155 if (valuestart) {
1156 if (*p == '=') {
1157 *valuestart = p + 1;
1158 *valuelen = strcspn(*valuestart, ",)");
1159 p = *valuestart + *valuelen;
1160 } else {
1161 if (*p != ',' && *p != ')')
1162 return 0;
1163 *valuestart = NULL;
1164 *valuelen = 0;
1167 if (*p == ',') {
1168 *end = p + 1;
1169 return 1;
1171 if (*p == ')') {
1172 *end = p;
1173 return 1;
1175 return 0;
1178 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1179 const char **end, int *val)
1181 const char *argval;
1182 char *strval;
1183 size_t arglen;
1184 int v;
1186 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1187 return 0;
1189 if (!argval) {
1190 *val = 1;
1191 return 1;
1194 strval = xstrndup(argval, arglen);
1195 v = git_parse_maybe_bool(strval);
1196 free(strval);
1198 if (v == -1)
1199 return 0;
1201 *val = v;
1203 return 1;
1206 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1208 const struct string_list *list = ud;
1209 const struct string_list_item *item;
1211 for_each_string_list_item (item, list) {
1212 if (key->len == (uintptr_t)item->util &&
1213 !strncasecmp(item->string, key->buf, key->len))
1214 return 1;
1216 return 0;
1219 int format_set_trailers_options(struct process_trailer_options *opts,
1220 struct string_list *filter_list,
1221 struct strbuf *sepbuf,
1222 struct strbuf *kvsepbuf,
1223 const char **arg,
1224 char **invalid_arg)
1226 for (;;) {
1227 const char *argval;
1228 size_t arglen;
1230 if (**arg == ')')
1231 break;
1233 if (match_placeholder_arg_value(*arg, "key", arg, &argval, &arglen)) {
1234 uintptr_t len = arglen;
1236 if (!argval)
1237 return -1;
1239 if (len && argval[len - 1] == ':')
1240 len--;
1241 string_list_append(filter_list, argval)->util = (char *)len;
1243 opts->filter = format_trailer_match_cb;
1244 opts->filter_data = filter_list;
1245 opts->only_trailers = 1;
1246 } else if (match_placeholder_arg_value(*arg, "separator", arg, &argval, &arglen)) {
1247 char *fmt;
1249 strbuf_reset(sepbuf);
1250 fmt = xstrndup(argval, arglen);
1251 strbuf_expand(sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1252 free(fmt);
1253 opts->separator = sepbuf;
1254 } else if (match_placeholder_arg_value(*arg, "key_value_separator", arg, &argval, &arglen)) {
1255 char *fmt;
1257 strbuf_reset(kvsepbuf);
1258 fmt = xstrndup(argval, arglen);
1259 strbuf_expand(kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1260 free(fmt);
1261 opts->key_value_separator = kvsepbuf;
1262 } else if (!match_placeholder_bool_arg(*arg, "only", arg, &opts->only_trailers) &&
1263 !match_placeholder_bool_arg(*arg, "unfold", arg, &opts->unfold) &&
1264 !match_placeholder_bool_arg(*arg, "keyonly", arg, &opts->key_only) &&
1265 !match_placeholder_bool_arg(*arg, "valueonly", arg, &opts->value_only)) {
1266 if (invalid_arg) {
1267 size_t len = strcspn(*arg, ",)");
1268 *invalid_arg = xstrndup(*arg, len);
1270 return -1;
1273 return 0;
1276 static size_t parse_describe_args(const char *start, struct strvec *args)
1278 struct {
1279 char *name;
1280 enum {
1281 DESCRIBE_ARG_BOOL,
1282 DESCRIBE_ARG_INTEGER,
1283 DESCRIBE_ARG_STRING,
1284 } type;
1285 } option[] = {
1286 { "tags", DESCRIBE_ARG_BOOL},
1287 { "abbrev", DESCRIBE_ARG_INTEGER },
1288 { "exclude", DESCRIBE_ARG_STRING },
1289 { "match", DESCRIBE_ARG_STRING },
1291 const char *arg = start;
1293 for (;;) {
1294 int found = 0;
1295 const char *argval;
1296 size_t arglen = 0;
1297 int optval = 0;
1298 int i;
1300 for (i = 0; !found && i < ARRAY_SIZE(option); i++) {
1301 switch (option[i].type) {
1302 case DESCRIBE_ARG_BOOL:
1303 if (match_placeholder_bool_arg(arg, option[i].name, &arg, &optval)) {
1304 if (optval)
1305 strvec_pushf(args, "--%s", option[i].name);
1306 else
1307 strvec_pushf(args, "--no-%s", option[i].name);
1308 found = 1;
1310 break;
1311 case DESCRIBE_ARG_INTEGER:
1312 if (match_placeholder_arg_value(arg, option[i].name, &arg,
1313 &argval, &arglen)) {
1314 char *endptr;
1315 if (!arglen)
1316 return 0;
1317 strtol(argval, &endptr, 10);
1318 if (endptr - argval != arglen)
1319 return 0;
1320 strvec_pushf(args, "--%s=%.*s", option[i].name, (int)arglen, argval);
1321 found = 1;
1323 break;
1324 case DESCRIBE_ARG_STRING:
1325 if (match_placeholder_arg_value(arg, option[i].name, &arg,
1326 &argval, &arglen)) {
1327 if (!arglen)
1328 return 0;
1329 strvec_pushf(args, "--%s=%.*s", option[i].name, (int)arglen, argval);
1330 found = 1;
1332 break;
1335 if (!found)
1336 break;
1339 return arg - start;
1342 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1343 const char *placeholder,
1344 void *context)
1346 struct format_commit_context *c = context;
1347 const struct commit *commit = c->commit;
1348 const char *msg = c->message;
1349 struct commit_list *p;
1350 const char *arg, *eol;
1351 size_t res;
1352 char **slot;
1354 /* these are independent of the commit */
1355 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1356 if (res)
1357 return res;
1359 switch (placeholder[0]) {
1360 case 'C':
1361 if (starts_with(placeholder + 1, "(auto)")) {
1362 c->auto_color = want_color(c->pretty_ctx->color);
1363 if (c->auto_color && sb->len)
1364 strbuf_addstr(sb, GIT_COLOR_RESET);
1365 return 7; /* consumed 7 bytes, "C(auto)" */
1366 } else {
1367 int ret = parse_color(sb, placeholder, c);
1368 if (ret)
1369 c->auto_color = 0;
1371 * Otherwise, we decided to treat %C<unknown>
1372 * as a literal string, and the previous
1373 * %C(auto) is still valid.
1375 return ret;
1377 case 'w':
1378 if (placeholder[1] == '(') {
1379 unsigned long width = 0, indent1 = 0, indent2 = 0;
1380 char *next;
1381 const char *start = placeholder + 2;
1382 const char *end = strchr(start, ')');
1383 if (!end)
1384 return 0;
1385 if (end > start) {
1386 width = strtoul(start, &next, 10);
1387 if (*next == ',') {
1388 indent1 = strtoul(next + 1, &next, 10);
1389 if (*next == ',') {
1390 indent2 = strtoul(next + 1,
1391 &next, 10);
1394 if (*next != ')')
1395 return 0;
1397 rewrap_message_tail(sb, c, width, indent1, indent2);
1398 return end - placeholder + 1;
1399 } else
1400 return 0;
1402 case '<':
1403 case '>':
1404 return parse_padding_placeholder(placeholder, c);
1407 if (skip_prefix(placeholder, "(describe", &arg)) {
1408 struct child_process cmd = CHILD_PROCESS_INIT;
1409 struct strbuf out = STRBUF_INIT;
1410 struct strbuf err = STRBUF_INIT;
1411 struct pretty_print_describe_status *describe_status;
1413 describe_status = c->pretty_ctx->describe_status;
1414 if (describe_status) {
1415 if (!describe_status->max_invocations)
1416 return 0;
1417 describe_status->max_invocations--;
1420 cmd.git_cmd = 1;
1421 strvec_push(&cmd.args, "describe");
1423 if (*arg == ':') {
1424 arg++;
1425 arg += parse_describe_args(arg, &cmd.args);
1428 if (*arg != ')') {
1429 child_process_clear(&cmd);
1430 return 0;
1433 strvec_push(&cmd.args, oid_to_hex(&commit->object.oid));
1434 pipe_command(&cmd, NULL, 0, &out, 0, &err, 0);
1435 strbuf_rtrim(&out);
1436 strbuf_addbuf(sb, &out);
1437 strbuf_release(&out);
1438 strbuf_release(&err);
1439 return arg - placeholder + 1;
1442 /* these depend on the commit */
1443 if (!commit->object.parsed)
1444 parse_object(the_repository, &commit->object.oid);
1446 switch (placeholder[0]) {
1447 case 'H': /* commit hash */
1448 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1449 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1450 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1451 return 1;
1452 case 'h': /* abbreviated commit hash */
1453 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1454 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1455 c->pretty_ctx->abbrev);
1456 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1457 return 1;
1458 case 'T': /* tree hash */
1459 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1460 return 1;
1461 case 't': /* abbreviated tree hash */
1462 strbuf_add_unique_abbrev(sb,
1463 get_commit_tree_oid(commit),
1464 c->pretty_ctx->abbrev);
1465 return 1;
1466 case 'P': /* parent hashes */
1467 for (p = commit->parents; p; p = p->next) {
1468 if (p != commit->parents)
1469 strbuf_addch(sb, ' ');
1470 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1472 return 1;
1473 case 'p': /* abbreviated parent hashes */
1474 for (p = commit->parents; p; p = p->next) {
1475 if (p != commit->parents)
1476 strbuf_addch(sb, ' ');
1477 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1478 c->pretty_ctx->abbrev);
1480 return 1;
1481 case 'm': /* left/right/bottom */
1482 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1483 return 1;
1484 case 'd':
1485 format_decorations(sb, commit, c->auto_color);
1486 return 1;
1487 case 'D':
1488 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1489 return 1;
1490 case 'S': /* tag/branch like --source */
1491 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1492 return 0;
1493 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1494 if (!(slot && *slot))
1495 return 0;
1496 strbuf_addstr(sb, *slot);
1497 return 1;
1498 case 'g': /* reflog info */
1499 switch(placeholder[1]) {
1500 case 'd': /* reflog selector */
1501 case 'D':
1502 if (c->pretty_ctx->reflog_info)
1503 get_reflog_selector(sb,
1504 c->pretty_ctx->reflog_info,
1505 &c->pretty_ctx->date_mode,
1506 c->pretty_ctx->date_mode_explicit,
1507 (placeholder[1] == 'd'));
1508 return 2;
1509 case 's': /* reflog message */
1510 if (c->pretty_ctx->reflog_info)
1511 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1512 return 2;
1513 case 'n':
1514 case 'N':
1515 case 'e':
1516 case 'E':
1517 return format_reflog_person(sb,
1518 placeholder[1],
1519 c->pretty_ctx->reflog_info,
1520 &c->pretty_ctx->date_mode);
1522 return 0; /* unknown %g placeholder */
1523 case 'N':
1524 if (c->pretty_ctx->notes_message) {
1525 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1526 return 1;
1528 return 0;
1531 if (placeholder[0] == 'G') {
1532 if (!c->signature_check.result)
1533 check_commit_signature(c->commit, &(c->signature_check));
1534 switch (placeholder[1]) {
1535 case 'G':
1536 if (c->signature_check.output)
1537 strbuf_addstr(sb, c->signature_check.output);
1538 break;
1539 case '?':
1540 switch (c->signature_check.result) {
1541 case 'G':
1542 switch (c->signature_check.trust_level) {
1543 case TRUST_UNDEFINED:
1544 case TRUST_NEVER:
1545 strbuf_addch(sb, 'U');
1546 break;
1547 default:
1548 strbuf_addch(sb, 'G');
1549 break;
1551 break;
1552 case 'B':
1553 case 'E':
1554 case 'N':
1555 case 'X':
1556 case 'Y':
1557 case 'R':
1558 strbuf_addch(sb, c->signature_check.result);
1560 break;
1561 case 'S':
1562 if (c->signature_check.signer)
1563 strbuf_addstr(sb, c->signature_check.signer);
1564 break;
1565 case 'K':
1566 if (c->signature_check.key)
1567 strbuf_addstr(sb, c->signature_check.key);
1568 break;
1569 case 'F':
1570 if (c->signature_check.fingerprint)
1571 strbuf_addstr(sb, c->signature_check.fingerprint);
1572 break;
1573 case 'P':
1574 if (c->signature_check.primary_key_fingerprint)
1575 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1576 break;
1577 case 'T':
1578 strbuf_addstr(sb, gpg_trust_level_to_str(c->signature_check.trust_level));
1579 break;
1580 default:
1581 return 0;
1583 return 2;
1586 /* For the rest we have to parse the commit header. */
1587 if (!c->commit_header_parsed) {
1588 msg = c->message =
1589 repo_logmsg_reencode(c->repository, commit,
1590 &c->commit_encoding, "UTF-8");
1591 parse_commit_header(c);
1594 switch (placeholder[0]) {
1595 case 'a': /* author ... */
1596 return format_person_part(sb, placeholder[1],
1597 msg + c->author.off, c->author.len,
1598 &c->pretty_ctx->date_mode);
1599 case 'c': /* committer ... */
1600 return format_person_part(sb, placeholder[1],
1601 msg + c->committer.off, c->committer.len,
1602 &c->pretty_ctx->date_mode);
1603 case 'e': /* encoding */
1604 if (c->commit_encoding)
1605 strbuf_addstr(sb, c->commit_encoding);
1606 return 1;
1607 case 'B': /* raw body */
1608 /* message_off is always left at the initial newline */
1609 strbuf_addstr(sb, msg + c->message_off + 1);
1610 return 1;
1613 /* Now we need to parse the commit message. */
1614 if (!c->commit_message_parsed)
1615 parse_commit_message(c);
1617 switch (placeholder[0]) {
1618 case 's': /* subject */
1619 format_subject(sb, msg + c->subject_off, " ");
1620 return 1;
1621 case 'f': /* sanitized subject */
1622 eol = strchrnul(msg + c->subject_off, '\n');
1623 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1624 return 1;
1625 case 'b': /* body */
1626 strbuf_addstr(sb, msg + c->body_off);
1627 return 1;
1630 if (skip_prefix(placeholder, "(trailers", &arg)) {
1631 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1632 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1633 struct strbuf sepbuf = STRBUF_INIT;
1634 struct strbuf kvsepbuf = STRBUF_INIT;
1635 size_t ret = 0;
1637 opts.no_divider = 1;
1639 if (*arg == ':') {
1640 arg++;
1641 if (format_set_trailers_options(&opts, &filter_list, &sepbuf, &kvsepbuf, &arg, NULL))
1642 goto trailer_out;
1644 if (*arg == ')') {
1645 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1646 ret = arg - placeholder + 1;
1648 trailer_out:
1649 string_list_clear(&filter_list, 0);
1650 strbuf_release(&sepbuf);
1651 return ret;
1654 return 0; /* unknown placeholder */
1657 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1658 const char *placeholder,
1659 struct format_commit_context *c)
1661 struct strbuf local_sb = STRBUF_INIT;
1662 int total_consumed = 0, len, padding = c->padding;
1663 if (padding < 0) {
1664 const char *start = strrchr(sb->buf, '\n');
1665 int occupied;
1666 if (!start)
1667 start = sb->buf;
1668 occupied = utf8_strnwidth(start, -1, 1);
1669 occupied += c->pretty_ctx->graph_width;
1670 padding = (-padding) - occupied;
1672 while (1) {
1673 int modifier = *placeholder == 'C';
1674 int consumed = format_commit_one(&local_sb, placeholder, c);
1675 total_consumed += consumed;
1677 if (!modifier)
1678 break;
1680 placeholder += consumed;
1681 if (*placeholder != '%')
1682 break;
1683 placeholder++;
1684 total_consumed++;
1686 len = utf8_strnwidth(local_sb.buf, -1, 1);
1688 if (c->flush_type == flush_left_and_steal) {
1689 const char *ch = sb->buf + sb->len - 1;
1690 while (len > padding && ch > sb->buf) {
1691 const char *p;
1692 if (*ch == ' ') {
1693 ch--;
1694 padding++;
1695 continue;
1697 /* check for trailing ansi sequences */
1698 if (*ch != 'm')
1699 break;
1700 p = ch - 1;
1701 while (ch - p < 10 && *p != '\033')
1702 p--;
1703 if (*p != '\033' ||
1704 ch + 1 - p != display_mode_esc_sequence_len(p))
1705 break;
1707 * got a good ansi sequence, put it back to
1708 * local_sb as we're cutting sb
1710 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1711 ch = p - 1;
1713 strbuf_setlen(sb, ch + 1 - sb->buf);
1714 c->flush_type = flush_left;
1717 if (len > padding) {
1718 switch (c->truncate) {
1719 case trunc_left:
1720 strbuf_utf8_replace(&local_sb,
1721 0, len - (padding - 2),
1722 "..");
1723 break;
1724 case trunc_middle:
1725 strbuf_utf8_replace(&local_sb,
1726 padding / 2 - 1,
1727 len - (padding - 2),
1728 "..");
1729 break;
1730 case trunc_right:
1731 strbuf_utf8_replace(&local_sb,
1732 padding - 2, len - (padding - 2),
1733 "..");
1734 break;
1735 case trunc_none:
1736 break;
1738 strbuf_addbuf(sb, &local_sb);
1739 } else {
1740 int sb_len = sb->len, offset = 0;
1741 if (c->flush_type == flush_left)
1742 offset = padding - len;
1743 else if (c->flush_type == flush_both)
1744 offset = (padding - len) / 2;
1746 * we calculate padding in columns, now
1747 * convert it back to chars
1749 padding = padding - len + local_sb.len;
1750 strbuf_addchars(sb, ' ', padding);
1751 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1752 local_sb.len);
1754 strbuf_release(&local_sb);
1755 c->flush_type = no_flush;
1756 return total_consumed;
1759 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1760 const char *placeholder,
1761 void *context)
1763 int consumed;
1764 size_t orig_len;
1765 enum {
1766 NO_MAGIC,
1767 ADD_LF_BEFORE_NON_EMPTY,
1768 DEL_LF_BEFORE_EMPTY,
1769 ADD_SP_BEFORE_NON_EMPTY
1770 } magic = NO_MAGIC;
1772 switch (placeholder[0]) {
1773 case '-':
1774 magic = DEL_LF_BEFORE_EMPTY;
1775 break;
1776 case '+':
1777 magic = ADD_LF_BEFORE_NON_EMPTY;
1778 break;
1779 case ' ':
1780 magic = ADD_SP_BEFORE_NON_EMPTY;
1781 break;
1782 default:
1783 break;
1785 if (magic != NO_MAGIC)
1786 placeholder++;
1788 orig_len = sb->len;
1789 if (((struct format_commit_context *)context)->flush_type != no_flush)
1790 consumed = format_and_pad_commit(sb, placeholder, context);
1791 else
1792 consumed = format_commit_one(sb, placeholder, context);
1793 if (magic == NO_MAGIC)
1794 return consumed;
1796 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1797 while (sb->len && sb->buf[sb->len - 1] == '\n')
1798 strbuf_setlen(sb, sb->len - 1);
1799 } else if (orig_len != sb->len) {
1800 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1801 strbuf_insertstr(sb, orig_len, "\n");
1802 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1803 strbuf_insertstr(sb, orig_len, " ");
1805 return consumed + 1;
1808 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1809 void *context)
1811 struct userformat_want *w = context;
1813 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1814 placeholder++;
1816 switch (*placeholder) {
1817 case 'N':
1818 w->notes = 1;
1819 break;
1820 case 'S':
1821 w->source = 1;
1822 break;
1823 case 'd':
1824 case 'D':
1825 w->decorate = 1;
1826 break;
1828 return 0;
1831 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1833 struct strbuf dummy = STRBUF_INIT;
1835 if (!fmt) {
1836 if (!user_format)
1837 return;
1838 fmt = user_format;
1840 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1841 strbuf_release(&dummy);
1844 void repo_format_commit_message(struct repository *r,
1845 const struct commit *commit,
1846 const char *format, struct strbuf *sb,
1847 const struct pretty_print_context *pretty_ctx)
1849 struct format_commit_context context = {
1850 .repository = r,
1851 .commit = commit,
1852 .pretty_ctx = pretty_ctx,
1853 .wrap_start = sb->len
1855 const char *output_enc = pretty_ctx->output_encoding;
1856 const char *utf8 = "UTF-8";
1858 strbuf_expand(sb, format, format_commit_item, &context);
1859 rewrap_message_tail(sb, &context, 0, 0, 0);
1862 * Convert output to an actual output encoding; note that
1863 * format_commit_item() will always use UTF-8, so we don't
1864 * have to bother if that's what the output wants.
1866 if (output_enc) {
1867 if (same_encoding(utf8, output_enc))
1868 output_enc = NULL;
1869 } else {
1870 if (context.commit_encoding &&
1871 !same_encoding(context.commit_encoding, utf8))
1872 output_enc = context.commit_encoding;
1875 if (output_enc) {
1876 size_t outsz;
1877 char *out = reencode_string_len(sb->buf, sb->len,
1878 output_enc, utf8, &outsz);
1879 if (out)
1880 strbuf_attach(sb, out, outsz, outsz + 1);
1883 free(context.commit_encoding);
1884 repo_unuse_commit_buffer(r, commit, context.message);
1887 static void pp_header(struct pretty_print_context *pp,
1888 const char *encoding,
1889 const struct commit *commit,
1890 const char **msg_p,
1891 struct strbuf *sb)
1893 int parents_shown = 0;
1895 for (;;) {
1896 const char *name, *line = *msg_p;
1897 int linelen = get_one_line(*msg_p);
1899 if (!linelen)
1900 return;
1901 *msg_p += linelen;
1903 if (linelen == 1)
1904 /* End of header */
1905 return;
1907 if (pp->fmt == CMIT_FMT_RAW) {
1908 strbuf_add(sb, line, linelen);
1909 continue;
1912 if (starts_with(line, "parent ")) {
1913 if (linelen != the_hash_algo->hexsz + 8)
1914 die("bad parent line in commit");
1915 continue;
1918 if (!parents_shown) {
1919 unsigned num = commit_list_count(commit->parents);
1920 /* with enough slop */
1921 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1922 add_merge_info(pp, sb, commit);
1923 parents_shown = 1;
1927 * MEDIUM == DEFAULT shows only author with dates.
1928 * FULL shows both authors but not dates.
1929 * FULLER shows both authors and dates.
1931 if (skip_prefix(line, "author ", &name)) {
1932 strbuf_grow(sb, linelen + 80);
1933 pp_user_info(pp, "Author", sb, name, encoding);
1935 if (skip_prefix(line, "committer ", &name) &&
1936 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1937 strbuf_grow(sb, linelen + 80);
1938 pp_user_info(pp, "Commit", sb, name, encoding);
1943 void pp_title_line(struct pretty_print_context *pp,
1944 const char **msg_p,
1945 struct strbuf *sb,
1946 const char *encoding,
1947 int need_8bit_cte)
1949 static const int max_length = 78; /* per rfc2047 */
1950 struct strbuf title;
1952 strbuf_init(&title, 80);
1953 *msg_p = format_subject(&title, *msg_p,
1954 pp->preserve_subject ? "\n" : " ");
1956 strbuf_grow(sb, title.len + 1024);
1957 if (pp->print_email_subject) {
1958 if (pp->rev)
1959 fmt_output_email_subject(sb, pp->rev);
1960 if (pp->encode_email_headers &&
1961 needs_rfc2047_encoding(title.buf, title.len))
1962 add_rfc2047(sb, title.buf, title.len,
1963 encoding, RFC2047_SUBJECT);
1964 else
1965 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1966 -last_line_length(sb), 1, max_length);
1967 } else {
1968 strbuf_addbuf(sb, &title);
1970 strbuf_addch(sb, '\n');
1972 if (need_8bit_cte == 0) {
1973 int i;
1974 for (i = 0; i < pp->in_body_headers.nr; i++) {
1975 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1976 need_8bit_cte = 1;
1977 break;
1982 if (need_8bit_cte > 0) {
1983 const char *header_fmt =
1984 "MIME-Version: 1.0\n"
1985 "Content-Type: text/plain; charset=%s\n"
1986 "Content-Transfer-Encoding: 8bit\n";
1987 strbuf_addf(sb, header_fmt, encoding);
1989 if (pp->after_subject) {
1990 strbuf_addstr(sb, pp->after_subject);
1992 if (cmit_fmt_is_mail(pp->fmt)) {
1993 strbuf_addch(sb, '\n');
1996 if (pp->in_body_headers.nr) {
1997 int i;
1998 for (i = 0; i < pp->in_body_headers.nr; i++) {
1999 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
2000 free(pp->in_body_headers.items[i].string);
2002 string_list_clear(&pp->in_body_headers, 0);
2003 strbuf_addch(sb, '\n');
2006 strbuf_release(&title);
2009 static int pp_utf8_width(const char *start, const char *end)
2011 int width = 0;
2012 size_t remain = end - start;
2014 while (remain) {
2015 int n = utf8_width(&start, &remain);
2016 if (n < 0 || !start)
2017 return -1;
2018 width += n;
2020 return width;
2023 static void strbuf_add_tabexpand(struct strbuf *sb, struct grep_opt *opt,
2024 int color, int tabwidth, const char *line,
2025 int linelen)
2027 const char *tab;
2029 while ((tab = memchr(line, '\t', linelen)) != NULL) {
2030 int width = pp_utf8_width(line, tab);
2033 * If it wasn't well-formed utf8, or it
2034 * had characters with badly defined
2035 * width (control characters etc), just
2036 * give up on trying to align things.
2038 if (width < 0)
2039 break;
2041 /* Output the data .. */
2042 append_line_with_color(sb, opt, line, tab - line, color,
2043 GREP_CONTEXT_BODY,
2044 GREP_HEADER_FIELD_MAX);
2046 /* .. and the de-tabified tab */
2047 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
2049 /* Skip over the printed part .. */
2050 linelen -= tab + 1 - line;
2051 line = tab + 1;
2055 * Print out everything after the last tab without
2056 * worrying about width - there's nothing more to
2057 * align.
2059 append_line_with_color(sb, opt, line, linelen, color, GREP_CONTEXT_BODY,
2060 GREP_HEADER_FIELD_MAX);
2064 * pp_handle_indent() prints out the intendation, and
2065 * the whole line (without the final newline), after
2066 * de-tabifying.
2068 static void pp_handle_indent(struct pretty_print_context *pp,
2069 struct strbuf *sb, int indent,
2070 const char *line, int linelen)
2072 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
2074 strbuf_addchars(sb, ' ', indent);
2075 if (pp->expand_tabs_in_log)
2076 strbuf_add_tabexpand(sb, opt, pp->color, pp->expand_tabs_in_log,
2077 line, linelen);
2078 else
2079 append_line_with_color(sb, opt, line, linelen, pp->color,
2080 GREP_CONTEXT_BODY,
2081 GREP_HEADER_FIELD_MAX);
2084 static int is_mboxrd_from(const char *line, int len)
2087 * a line matching /^From $/ here would only have len == 4
2088 * at this point because is_empty_line would've trimmed all
2089 * trailing space
2091 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
2094 void pp_remainder(struct pretty_print_context *pp,
2095 const char **msg_p,
2096 struct strbuf *sb,
2097 int indent)
2099 struct grep_opt *opt = pp->rev ? &pp->rev->grep_filter : NULL;
2100 int first = 1;
2102 for (;;) {
2103 const char *line = *msg_p;
2104 int linelen = get_one_line(line);
2105 *msg_p += linelen;
2107 if (!linelen)
2108 break;
2110 if (is_blank_line(line, &linelen)) {
2111 if (first)
2112 continue;
2113 if (pp->fmt == CMIT_FMT_SHORT)
2114 break;
2116 first = 0;
2118 strbuf_grow(sb, linelen + indent + 20);
2119 if (indent)
2120 pp_handle_indent(pp, sb, indent, line, linelen);
2121 else if (pp->expand_tabs_in_log)
2122 strbuf_add_tabexpand(sb, opt, pp->color,
2123 pp->expand_tabs_in_log, line,
2124 linelen);
2125 else {
2126 if (pp->fmt == CMIT_FMT_MBOXRD &&
2127 is_mboxrd_from(line, linelen))
2128 strbuf_addch(sb, '>');
2130 append_line_with_color(sb, opt, line, linelen,
2131 pp->color, GREP_CONTEXT_BODY,
2132 GREP_HEADER_FIELD_MAX);
2134 strbuf_addch(sb, '\n');
2138 void pretty_print_commit(struct pretty_print_context *pp,
2139 const struct commit *commit,
2140 struct strbuf *sb)
2142 unsigned long beginning_of_body;
2143 int indent = 4;
2144 const char *msg;
2145 const char *reencoded;
2146 const char *encoding;
2147 int need_8bit_cte = pp->need_8bit_cte;
2149 if (pp->fmt == CMIT_FMT_USERFORMAT) {
2150 format_commit_message(commit, user_format, sb, pp);
2151 return;
2154 encoding = get_log_output_encoding();
2155 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
2157 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2158 indent = 0;
2161 * We need to check and emit Content-type: to mark it
2162 * as 8-bit if we haven't done so.
2164 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
2165 int i, ch, in_body;
2167 for (in_body = i = 0; (ch = msg[i]); i++) {
2168 if (!in_body) {
2169 /* author could be non 7-bit ASCII but
2170 * the log may be so; skip over the
2171 * header part first.
2173 if (ch == '\n' && msg[i+1] == '\n')
2174 in_body = 1;
2176 else if (non_ascii(ch)) {
2177 need_8bit_cte = 1;
2178 break;
2183 pp_header(pp, encoding, commit, &msg, sb);
2184 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
2185 strbuf_addch(sb, '\n');
2188 /* Skip excess blank lines at the beginning of body, if any... */
2189 msg = skip_blank_lines(msg);
2191 /* These formats treat the title line specially. */
2192 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2193 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2195 beginning_of_body = sb->len;
2196 if (pp->fmt != CMIT_FMT_ONELINE)
2197 pp_remainder(pp, &msg, sb, indent);
2198 strbuf_rtrim(sb);
2200 /* Make sure there is an EOLN for the non-oneline case */
2201 if (pp->fmt != CMIT_FMT_ONELINE)
2202 strbuf_addch(sb, '\n');
2205 * The caller may append additional body text in e-mail
2206 * format. Make sure we did not strip the blank line
2207 * between the header and the body.
2209 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2210 strbuf_addch(sb, '\n');
2212 unuse_commit_buffer(commit, reencoded);
2215 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2216 struct strbuf *sb)
2218 struct pretty_print_context pp = {0};
2219 pp.fmt = fmt;
2220 pretty_print_commit(&pp, commit, sb);