6 #include "string-list.h"
11 #include "reflog-walk.h"
12 #include "gpg-interface.h"
14 static char *user_format;
15 static struct cmt_fmt_map {
20 const char *user_format;
22 static size_t builtin_formats_len;
23 static size_t commit_formats_len;
24 static size_t commit_formats_alloc;
25 static struct cmt_fmt_map *find_commit_format(const char *sought);
27 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
30 user_format = xstrdup(cp);
32 rev->use_terminator = 1;
33 rev->commit_format = CMIT_FMT_USERFORMAT;
36 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
38 struct cmt_fmt_map *commit_format = NULL;
43 if (!starts_with(var, "pretty."))
46 name = var + strlen("pretty.");
47 for (i = 0; i < builtin_formats_len; i++) {
48 if (!strcmp(commit_formats[i].name, name))
52 for (i = builtin_formats_len; i < commit_formats_len; i++) {
53 if (!strcmp(commit_formats[i].name, name)) {
54 commit_format = &commit_formats[i];
60 ALLOC_GROW(commit_formats, commit_formats_len+1,
61 commit_formats_alloc);
62 commit_format = &commit_formats[commit_formats_len];
63 memset(commit_format, 0, sizeof(*commit_format));
67 commit_format->name = xstrdup(name);
68 commit_format->format = CMIT_FMT_USERFORMAT;
69 git_config_string(&fmt, var, value);
70 if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
71 commit_format->is_tformat = fmt[0] == 't';
72 fmt = strchr(fmt, ':') + 1;
73 } else if (strchr(fmt, '%'))
74 commit_format->is_tformat = 1;
76 commit_format->is_alias = 1;
77 commit_format->user_format = fmt;
82 static void setup_commit_formats(void)
84 struct cmt_fmt_map builtin_formats[] = {
85 { "raw", CMIT_FMT_RAW, 0 },
86 { "medium", CMIT_FMT_MEDIUM, 0 },
87 { "short", CMIT_FMT_SHORT, 0 },
88 { "email", CMIT_FMT_EMAIL, 0 },
89 { "fuller", CMIT_FMT_FULLER, 0 },
90 { "full", CMIT_FMT_FULL, 0 },
91 { "oneline", CMIT_FMT_ONELINE, 1 }
93 commit_formats_len = ARRAY_SIZE(builtin_formats);
94 builtin_formats_len = commit_formats_len;
95 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
96 memcpy(commit_formats, builtin_formats,
97 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
99 git_config(git_pretty_formats_config, NULL);
102 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
103 const char *original,
104 int num_redirections)
106 struct cmt_fmt_map *found = NULL;
107 size_t found_match_len = 0;
110 if (num_redirections >= commit_formats_len)
111 die("invalid --pretty format: "
112 "'%s' references an alias which points to itself",
115 for (i = 0; i < commit_formats_len; i++) {
118 if (!starts_with(commit_formats[i].name, sought))
121 match_len = strlen(commit_formats[i].name);
122 if (found == NULL || found_match_len > match_len) {
123 found = &commit_formats[i];
124 found_match_len = match_len;
128 if (found && found->is_alias) {
129 found = find_commit_format_recursive(found->user_format,
137 static struct cmt_fmt_map *find_commit_format(const char *sought)
140 setup_commit_formats();
142 return find_commit_format_recursive(sought, sought, 0);
145 void get_commit_format(const char *arg, struct rev_info *rev)
147 struct cmt_fmt_map *commit_format;
149 rev->use_terminator = 0;
151 rev->commit_format = CMIT_FMT_DEFAULT;
154 if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
155 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
159 if (strchr(arg, '%')) {
160 save_user_format(rev, arg, 1);
164 commit_format = find_commit_format(arg);
166 die("invalid --pretty format: %s", arg);
168 rev->commit_format = commit_format->format;
169 rev->use_terminator = commit_format->is_tformat;
170 if (commit_format->format == CMIT_FMT_USERFORMAT) {
171 save_user_format(rev, commit_format->user_format,
172 commit_format->is_tformat);
177 * Generic support for pretty-printing the header
179 static int get_one_line(const char *msg)
194 /* High bit set, or ISO-2022-INT */
195 static int non_ascii(int ch)
197 return !isascii(ch) || ch == '\033';
200 int has_non_ascii(const char *s)
205 while ((ch = *s++) != '\0') {
212 static int is_rfc822_special(char ch)
234 static int needs_rfc822_quoting(const char *s, int len)
237 for (i = 0; i < len; i++)
238 if (is_rfc822_special(s[i]))
243 static int last_line_length(struct strbuf *sb)
247 /* How many bytes are already used on the last line? */
248 for (i = sb->len - 1; i >= 0; i--)
249 if (sb->buf[i] == '\n')
251 return sb->len - (i + 1);
254 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
258 /* just a guess, we may have to also backslash-quote */
259 strbuf_grow(out, len + 2);
261 strbuf_addch(out, '"');
262 for (i = 0; i < len; i++) {
266 strbuf_addch(out, '\\');
269 strbuf_addch(out, s[i]);
272 strbuf_addch(out, '"');
280 static int is_rfc2047_special(char ch, enum rfc2047_type type)
283 * rfc2047, section 4.2:
285 * 8-bit values which correspond to printable ASCII characters other
286 * than "=", "?", and "_" (underscore), MAY be represented as those
287 * characters. (But see section 5 for restrictions.) In
288 * particular, SPACE and TAB MUST NOT be represented as themselves
289 * within encoded words.
293 * rule out non-ASCII characters and non-printable characters (the
294 * non-ASCII check should be redundant as isprint() is not localized
295 * and only knows about ASCII, but be defensive about that)
297 if (non_ascii(ch) || !isprint(ch))
301 * rule out special printable characters (' ' should be the only
302 * whitespace character considered printable, but be defensive and use
305 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
309 * rfc2047, section 5.3:
311 * As a replacement for a 'word' entity within a 'phrase', for example,
312 * one that precedes an address in a From, To, or Cc header. The ABNF
313 * definition for 'phrase' from RFC 822 thus becomes:
315 * phrase = 1*( encoded-word / word )
317 * In this case the set of characters that may be used in a "Q"-encoded
318 * 'encoded-word' is restricted to: <upper and lower case ASCII
319 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
320 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
321 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
322 * 'special' by 'linear-white-space'.
325 if (type != RFC2047_ADDRESS)
328 /* '=' and '_' are special cases and have been checked above */
329 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
332 static int needs_rfc2047_encoding(const char *line, int len,
333 enum rfc2047_type type)
337 for (i = 0; i < len; i++) {
339 if (non_ascii(ch) || ch == '\n')
341 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
348 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
349 const char *encoding, enum rfc2047_type type)
351 static const int max_encoded_length = 76; /* per rfc2047 */
353 int line_len = last_line_length(sb);
355 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
356 strbuf_addf(sb, "=?%s?q?", encoding);
357 line_len += strlen(encoding) + 5; /* 5 for =??q? */
361 * RFC 2047, section 5 (3):
363 * Each 'encoded-word' MUST represent an integral number of
364 * characters. A multi-octet character may not be split across
365 * adjacent 'encoded- word's.
367 const unsigned char *p = (const unsigned char *)line;
368 int chrlen = mbs_chrlen(&line, &len, encoding);
369 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
371 /* "=%02X" * chrlen, or the byte itself */
372 const char *encoded_fmt = is_special ? "=%02X" : "%c";
373 int encoded_len = is_special ? 3 * chrlen : 1;
376 * According to RFC 2047, we could encode the special character
377 * ' ' (space) with '_' (underscore) for readability. But many
378 * programs do not understand this and just leave the
379 * underscore in place. Thus, we do nothing special here, which
380 * causes ' ' to be encoded as '=20', avoiding this problem.
383 if (line_len + encoded_len + 2 > max_encoded_length) {
384 /* It won't fit with trailing "?=" --- break the line */
385 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
386 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
389 for (i = 0; i < chrlen; i++)
390 strbuf_addf(sb, encoded_fmt, p[i]);
391 line_len += encoded_len;
393 strbuf_addstr(sb, "?=");
396 static const char *show_ident_date(const struct ident_split *ident,
399 unsigned long date = 0;
402 if (ident->date_begin && ident->date_end)
403 date = strtoul(ident->date_begin, NULL, 10);
404 if (date_overflows(date))
407 if (ident->tz_begin && ident->tz_end)
408 tz = strtol(ident->tz_begin, NULL, 10);
409 if (tz >= INT_MAX || tz <= INT_MIN)
412 return show_date(date, tz, mode);
415 void pp_user_info(struct pretty_print_context *pp,
416 const char *what, struct strbuf *sb,
417 const char *line, const char *encoding)
419 struct ident_split ident;
421 const char *mailbuf, *namebuf;
422 size_t namelen, maillen;
423 int max_length = 78; /* per rfc2822 */
425 if (pp->fmt == CMIT_FMT_ONELINE)
428 line_end = strchrnul(line, '\n');
429 if (split_ident_line(&ident, line, line_end - line))
432 mailbuf = ident.mail_begin;
433 maillen = ident.mail_end - ident.mail_begin;
434 namebuf = ident.name_begin;
435 namelen = ident.name_end - ident.name_begin;
438 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
440 if (pp->fmt == CMIT_FMT_EMAIL) {
441 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
442 struct strbuf buf = STRBUF_INIT;
444 strbuf_addstr(&buf, "From: ");
445 strbuf_add(&buf, namebuf, namelen);
446 strbuf_addstr(&buf, " <");
447 strbuf_add(&buf, mailbuf, maillen);
448 strbuf_addstr(&buf, ">\n");
449 string_list_append(&pp->in_body_headers,
450 strbuf_detach(&buf, NULL));
452 mailbuf = pp->from_ident->mail_begin;
453 maillen = pp->from_ident->mail_end - mailbuf;
454 namebuf = pp->from_ident->name_begin;
455 namelen = pp->from_ident->name_end - namebuf;
458 strbuf_addstr(sb, "From: ");
459 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
460 add_rfc2047(sb, namebuf, namelen,
461 encoding, RFC2047_ADDRESS);
462 max_length = 76; /* per rfc2047 */
463 } else if (needs_rfc822_quoting(namebuf, namelen)) {
464 struct strbuf quoted = STRBUF_INIT;
465 add_rfc822_quoted("ed, namebuf, namelen);
466 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
468 strbuf_release("ed);
470 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
475 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
476 strbuf_addch(sb, '\n');
477 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
479 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
480 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
481 (int)namelen, namebuf, (int)maillen, mailbuf);
485 case CMIT_FMT_MEDIUM:
486 strbuf_addf(sb, "Date: %s\n",
487 show_ident_date(&ident, pp->date_mode));
490 strbuf_addf(sb, "Date: %s\n",
491 show_ident_date(&ident, DATE_RFC2822));
493 case CMIT_FMT_FULLER:
494 strbuf_addf(sb, "%sDate: %s\n", what,
495 show_ident_date(&ident, pp->date_mode));
503 static int is_empty_line(const char *line, int *len_p)
506 while (len && isspace(line[len - 1]))
512 static const char *skip_empty_lines(const char *msg)
515 int linelen = get_one_line(msg);
519 if (!is_empty_line(msg, &ll))
526 static void add_merge_info(const struct pretty_print_context *pp,
527 struct strbuf *sb, const struct commit *commit)
529 struct commit_list *parent = commit->parents;
531 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
532 !parent || !parent->next)
535 strbuf_addstr(sb, "Merge:");
538 struct commit *p = parent->item;
539 const char *hex = NULL;
541 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
543 hex = sha1_to_hex(p->object.sha1);
544 parent = parent->next;
546 strbuf_addf(sb, " %s", hex);
548 strbuf_addch(sb, '\n');
551 static char *get_header(const struct commit *commit, const char *msg,
554 int key_len = strlen(key);
555 const char *line = msg;
558 const char *eol = strchrnul(line, '\n'), *next;
563 warning("malformed commit (header is missing newline): %s",
564 sha1_to_hex(commit->object.sha1));
568 if (eol - line > key_len &&
569 !strncmp(line, key, key_len) &&
570 line[key_len] == ' ') {
571 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
578 static char *replace_encoding_header(char *buf, const char *encoding)
580 struct strbuf tmp = STRBUF_INIT;
584 /* guess if there is an encoding header before a \n\n */
585 while (strncmp(cp, "encoding ", strlen("encoding "))) {
586 cp = strchr(cp, '\n');
587 if (!cp || *++cp == '\n')
591 cp = strchr(cp, '\n');
593 return buf; /* should not happen but be defensive */
594 len = cp + 1 - (buf + start);
596 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
597 if (is_encoding_utf8(encoding)) {
598 /* we have re-coded to UTF-8; drop the header */
599 strbuf_remove(&tmp, start, len);
601 /* just replaces XXXX in 'encoding XXXX\n' */
602 strbuf_splice(&tmp, start + strlen("encoding "),
603 len - strlen("encoding \n"),
604 encoding, strlen(encoding));
606 return strbuf_detach(&tmp, NULL);
609 const char *logmsg_reencode(const struct commit *commit,
610 char **commit_encoding,
611 const char *output_encoding)
613 static const char *utf8 = "UTF-8";
614 const char *use_encoding;
616 const char *msg = get_commit_buffer(commit, NULL);
619 if (!output_encoding || !*output_encoding) {
622 get_header(commit, msg, "encoding");
625 encoding = get_header(commit, msg, "encoding");
627 *commit_encoding = encoding;
628 use_encoding = encoding ? encoding : utf8;
629 if (same_encoding(use_encoding, output_encoding)) {
631 * No encoding work to be done. If we have no encoding header
632 * at all, then there's nothing to do, and we can return the
633 * message verbatim (whether newly allocated or not).
639 * Otherwise, we still want to munge the encoding header in the
640 * result, which will be done by modifying the buffer. If we
641 * are using a fresh copy, we can reuse it. But if we are using
642 * the cached copy from get_commit_buffer, we need to duplicate it
643 * to avoid munging the cached copy.
645 if (msg == get_cached_commit_buffer(commit, NULL))
652 * There's actual encoding work to do. Do the reencoding, which
653 * still leaves the header to be replaced in the next step. At
654 * this point, we are done with msg. If we allocated a fresh
655 * copy, we can free it.
657 out = reencode_string(msg, output_encoding, use_encoding);
659 unuse_commit_buffer(commit, msg);
663 * This replacement actually consumes the buffer we hand it, so we do
664 * not have to worry about freeing the old "out" here.
667 out = replace_encoding_header(out, output_encoding);
669 if (!commit_encoding)
672 * If the re-encoding failed, out might be NULL here; in that
673 * case we just return the commit message verbatim.
675 return out ? out : msg;
678 static int mailmap_name(const char **email, size_t *email_len,
679 const char **name, size_t *name_len)
681 static struct string_list *mail_map;
683 mail_map = xcalloc(1, sizeof(*mail_map));
684 read_mailmap(mail_map, NULL);
686 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
689 static size_t format_person_part(struct strbuf *sb, char part,
690 const char *msg, int len, enum date_mode dmode)
692 /* currently all placeholders have same length */
693 const int placeholder_len = 2;
694 struct ident_split s;
695 const char *name, *mail;
696 size_t maillen, namelen;
698 if (split_ident_line(&s, msg, len) < 0)
702 namelen = s.name_end - s.name_begin;
704 maillen = s.mail_end - s.mail_begin;
706 if (part == 'N' || part == 'E') /* mailmap lookup */
707 mailmap_name(&mail, &maillen, &name, &namelen);
708 if (part == 'n' || part == 'N') { /* name */
709 strbuf_add(sb, name, namelen);
710 return placeholder_len;
712 if (part == 'e' || part == 'E') { /* email */
713 strbuf_add(sb, mail, maillen);
714 return placeholder_len;
720 if (part == 't') { /* date, UNIX timestamp */
721 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
722 return placeholder_len;
727 strbuf_addstr(sb, show_ident_date(&s, dmode));
728 return placeholder_len;
729 case 'D': /* date, RFC2822 style */
730 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
731 return placeholder_len;
732 case 'r': /* date, relative */
733 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
734 return placeholder_len;
735 case 'i': /* date, ISO 8601 */
736 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
737 return placeholder_len;
742 * reading from either a bogus commit, or a reflog entry with
743 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
744 * to compute a valid return value.
746 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
747 || part == 'D' || part == 'r' || part == 'i')
748 return placeholder_len;
750 return 0; /* unknown placeholder */
762 flush_left_and_steal,
773 struct format_commit_context {
774 const struct commit *commit;
775 const struct pretty_print_context *pretty_ctx;
776 unsigned commit_header_parsed:1;
777 unsigned commit_message_parsed:1;
778 struct signature_check signature_check;
779 enum flush_type flush_type;
780 enum trunc_type truncate;
782 char *commit_encoding;
783 size_t width, indent1, indent2;
787 /* These offsets are relative to the start of the commit message. */
789 struct chunk committer;
794 /* The following ones are relative to the result struct strbuf. */
795 struct chunk abbrev_commit_hash;
796 struct chunk abbrev_tree_hash;
797 struct chunk abbrev_parent_hashes;
801 static int add_again(struct strbuf *sb, struct chunk *chunk)
804 strbuf_adddup(sb, chunk->off, chunk->len);
809 * We haven't seen this chunk before. Our caller is surely
810 * going to add it the hard way now. Remember the most likely
811 * start of the to-be-added chunk: the current end of the
814 chunk->off = sb->len;
818 static void parse_commit_header(struct format_commit_context *context)
820 const char *msg = context->message;
823 for (i = 0; msg[i]; i++) {
825 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
830 } else if (starts_with(msg + i, "author ")) {
831 context->author.off = i + 7;
832 context->author.len = eol - i - 7;
833 } else if (starts_with(msg + i, "committer ")) {
834 context->committer.off = i + 10;
835 context->committer.len = eol - i - 10;
839 context->message_off = i;
840 context->commit_header_parsed = 1;
843 static int istitlechar(char c)
845 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
846 (c >= '0' && c <= '9') || c == '.' || c == '_';
849 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
852 size_t start_len = sb->len;
855 for (; *msg && *msg != '\n'; msg++) {
856 if (istitlechar(*msg)) {
858 strbuf_addch(sb, '-');
860 strbuf_addch(sb, *msg);
862 while (*(msg+1) == '.')
868 /* trim any trailing '.' or '-' characters */
870 while (sb->len - trimlen > start_len &&
871 (sb->buf[sb->len - 1 - trimlen] == '.'
872 || sb->buf[sb->len - 1 - trimlen] == '-'))
874 strbuf_remove(sb, sb->len - trimlen, trimlen);
877 const char *format_subject(struct strbuf *sb, const char *msg,
878 const char *line_separator)
883 const char *line = msg;
884 int linelen = get_one_line(line);
887 if (!linelen || is_empty_line(line, &linelen))
892 strbuf_grow(sb, linelen + 2);
894 strbuf_addstr(sb, line_separator);
895 strbuf_add(sb, line, linelen);
901 static void parse_commit_message(struct format_commit_context *c)
903 const char *msg = c->message + c->message_off;
904 const char *start = c->message;
906 msg = skip_empty_lines(msg);
907 c->subject_off = msg - start;
909 msg = format_subject(NULL, msg, NULL);
910 msg = skip_empty_lines(msg);
911 c->body_off = msg - start;
913 c->commit_message_parsed = 1;
916 static void strbuf_wrap(struct strbuf *sb, size_t pos,
917 size_t width, size_t indent1, size_t indent2)
919 struct strbuf tmp = STRBUF_INIT;
922 strbuf_add(&tmp, sb->buf, pos);
923 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
924 (int) indent1, (int) indent2, (int) width);
925 strbuf_swap(&tmp, sb);
926 strbuf_release(&tmp);
929 static void rewrap_message_tail(struct strbuf *sb,
930 struct format_commit_context *c,
931 size_t new_width, size_t new_indent1,
934 if (c->width == new_width && c->indent1 == new_indent1 &&
935 c->indent2 == new_indent2)
937 if (c->wrap_start < sb->len)
938 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
939 c->wrap_start = sb->len;
940 c->width = new_width;
941 c->indent1 = new_indent1;
942 c->indent2 = new_indent2;
945 static int format_reflog_person(struct strbuf *sb,
947 struct reflog_walk_info *log,
948 enum date_mode dmode)
955 ident = get_reflog_ident(log);
959 return format_person_part(sb, part, ident, strlen(ident), dmode);
962 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
963 const char *placeholder,
964 struct format_commit_context *c)
966 if (placeholder[1] == '(') {
967 const char *begin = placeholder + 2;
968 const char *end = strchr(begin, ')');
969 char color[COLOR_MAXLEN];
973 if (starts_with(begin, "auto,")) {
974 if (!want_color(c->pretty_ctx->color))
975 return end - placeholder + 1;
978 color_parse_mem(begin,
980 "--pretty format", color);
981 strbuf_addstr(sb, color);
982 return end - placeholder + 1;
984 if (starts_with(placeholder + 1, "red")) {
985 strbuf_addstr(sb, GIT_COLOR_RED);
987 } else if (starts_with(placeholder + 1, "green")) {
988 strbuf_addstr(sb, GIT_COLOR_GREEN);
990 } else if (starts_with(placeholder + 1, "blue")) {
991 strbuf_addstr(sb, GIT_COLOR_BLUE);
993 } else if (starts_with(placeholder + 1, "reset")) {
994 strbuf_addstr(sb, GIT_COLOR_RESET);
1000 static size_t parse_padding_placeholder(struct strbuf *sb,
1001 const char *placeholder,
1002 struct format_commit_context *c)
1004 const char *ch = placeholder;
1005 enum flush_type flush_type;
1010 flush_type = flush_right;
1014 flush_type = flush_both;
1016 } else if (*ch == '>') {
1017 flush_type = flush_left_and_steal;
1020 flush_type = flush_left;
1026 /* the next value means "wide enough to that column" */
1033 const char *start = ch + 1;
1034 const char *end = start + strcspn(start, ",)");
1037 if (!end || end == start)
1039 width = strtoul(start, &next, 10);
1040 if (next == start || width == 0)
1042 c->padding = to_column ? -width : width;
1043 c->flush_type = flush_type;
1047 end = strchr(start, ')');
1048 if (!end || end == start)
1050 if (starts_with(start, "trunc)"))
1051 c->truncate = trunc_right;
1052 else if (starts_with(start, "ltrunc)"))
1053 c->truncate = trunc_left;
1054 else if (starts_with(start, "mtrunc)"))
1055 c->truncate = trunc_middle;
1059 c->truncate = trunc_none;
1061 return end - placeholder + 1;
1066 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1067 const char *placeholder,
1070 struct format_commit_context *c = context;
1071 const struct commit *commit = c->commit;
1072 const char *msg = c->message;
1073 struct commit_list *p;
1076 /* these are independent of the commit */
1077 switch (placeholder[0]) {
1079 if (starts_with(placeholder + 1, "(auto)")) {
1081 return 7; /* consumed 7 bytes, "C(auto)" */
1083 int ret = parse_color(sb, placeholder, c);
1087 * Otherwise, we decided to treat %C<unknown>
1088 * as a literal string, and the previous
1089 * %C(auto) is still valid.
1093 case 'n': /* newline */
1094 strbuf_addch(sb, '\n');
1097 /* %x00 == NUL, %x0a == LF, etc. */
1098 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1100 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1102 strbuf_addch(sb, (h1<<4)|h2);
1107 if (placeholder[1] == '(') {
1108 unsigned long width = 0, indent1 = 0, indent2 = 0;
1110 const char *start = placeholder + 2;
1111 const char *end = strchr(start, ')');
1115 width = strtoul(start, &next, 10);
1117 indent1 = strtoul(next + 1, &next, 10);
1119 indent2 = strtoul(next + 1,
1126 rewrap_message_tail(sb, c, width, indent1, indent2);
1127 return end - placeholder + 1;
1133 return parse_padding_placeholder(sb, placeholder, c);
1136 /* these depend on the commit */
1137 if (!commit->object.parsed)
1138 parse_object(commit->object.sha1);
1140 switch (placeholder[0]) {
1141 case 'H': /* commit hash */
1142 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1143 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1144 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1146 case 'h': /* abbreviated commit hash */
1147 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1148 if (add_again(sb, &c->abbrev_commit_hash)) {
1149 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1152 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1153 c->pretty_ctx->abbrev));
1154 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1155 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1157 case 'T': /* tree hash */
1158 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1160 case 't': /* abbreviated tree hash */
1161 if (add_again(sb, &c->abbrev_tree_hash))
1163 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1164 c->pretty_ctx->abbrev));
1165 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1167 case 'P': /* parent hashes */
1168 for (p = commit->parents; p; p = p->next) {
1169 if (p != commit->parents)
1170 strbuf_addch(sb, ' ');
1171 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1174 case 'p': /* abbreviated parent hashes */
1175 if (add_again(sb, &c->abbrev_parent_hashes))
1177 for (p = commit->parents; p; p = p->next) {
1178 if (p != commit->parents)
1179 strbuf_addch(sb, ' ');
1180 strbuf_addstr(sb, find_unique_abbrev(
1181 p->item->object.sha1,
1182 c->pretty_ctx->abbrev));
1184 c->abbrev_parent_hashes.len = sb->len -
1185 c->abbrev_parent_hashes.off;
1187 case 'm': /* left/right/bottom */
1188 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1191 load_ref_decorations(DECORATE_SHORT_REFS);
1192 format_decorations(sb, commit, c->auto_color);
1194 case 'g': /* reflog info */
1195 switch(placeholder[1]) {
1196 case 'd': /* reflog selector */
1198 if (c->pretty_ctx->reflog_info)
1199 get_reflog_selector(sb,
1200 c->pretty_ctx->reflog_info,
1201 c->pretty_ctx->date_mode,
1202 c->pretty_ctx->date_mode_explicit,
1203 (placeholder[1] == 'd'));
1205 case 's': /* reflog message */
1206 if (c->pretty_ctx->reflog_info)
1207 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1213 return format_reflog_person(sb,
1215 c->pretty_ctx->reflog_info,
1216 c->pretty_ctx->date_mode);
1218 return 0; /* unknown %g placeholder */
1220 if (c->pretty_ctx->notes_message) {
1221 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1227 if (placeholder[0] == 'G') {
1228 if (!c->signature_check.result)
1229 check_commit_signature(c->commit, &(c->signature_check));
1230 switch (placeholder[1]) {
1232 if (c->signature_check.gpg_output)
1233 strbuf_addstr(sb, c->signature_check.gpg_output);
1236 switch (c->signature_check.result) {
1241 strbuf_addch(sb, c->signature_check.result);
1245 if (c->signature_check.signer)
1246 strbuf_addstr(sb, c->signature_check.signer);
1249 if (c->signature_check.key)
1250 strbuf_addstr(sb, c->signature_check.key);
1259 /* For the rest we have to parse the commit header. */
1260 if (!c->commit_header_parsed)
1261 parse_commit_header(c);
1263 switch (placeholder[0]) {
1264 case 'a': /* author ... */
1265 return format_person_part(sb, placeholder[1],
1266 msg + c->author.off, c->author.len,
1267 c->pretty_ctx->date_mode);
1268 case 'c': /* committer ... */
1269 return format_person_part(sb, placeholder[1],
1270 msg + c->committer.off, c->committer.len,
1271 c->pretty_ctx->date_mode);
1272 case 'e': /* encoding */
1273 if (c->commit_encoding)
1274 strbuf_addstr(sb, c->commit_encoding);
1276 case 'B': /* raw body */
1277 /* message_off is always left at the initial newline */
1278 strbuf_addstr(sb, msg + c->message_off + 1);
1282 /* Now we need to parse the commit message. */
1283 if (!c->commit_message_parsed)
1284 parse_commit_message(c);
1286 switch (placeholder[0]) {
1287 case 's': /* subject */
1288 format_subject(sb, msg + c->subject_off, " ");
1290 case 'f': /* sanitized subject */
1291 format_sanitized_subject(sb, msg + c->subject_off);
1293 case 'b': /* body */
1294 strbuf_addstr(sb, msg + c->body_off);
1297 return 0; /* unknown placeholder */
1300 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1301 const char *placeholder,
1302 struct format_commit_context *c)
1304 struct strbuf local_sb = STRBUF_INIT;
1305 int total_consumed = 0, len, padding = c->padding;
1307 const char *start = strrchr(sb->buf, '\n');
1311 occupied = utf8_strnwidth(start, -1, 1);
1312 padding = (-padding) - occupied;
1315 int modifier = *placeholder == 'C';
1316 int consumed = format_commit_one(&local_sb, placeholder, c);
1317 total_consumed += consumed;
1322 placeholder += consumed;
1323 if (*placeholder != '%')
1328 len = utf8_strnwidth(local_sb.buf, -1, 1);
1330 if (c->flush_type == flush_left_and_steal) {
1331 const char *ch = sb->buf + sb->len - 1;
1332 while (len > padding && ch > sb->buf) {
1339 /* check for trailing ansi sequences */
1343 while (ch - p < 10 && *p != '\033')
1346 ch + 1 - p != display_mode_esc_sequence_len(p))
1349 * got a good ansi sequence, put it back to
1350 * local_sb as we're cutting sb
1352 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1355 strbuf_setlen(sb, ch + 1 - sb->buf);
1356 c->flush_type = flush_left;
1359 if (len > padding) {
1360 switch (c->truncate) {
1362 strbuf_utf8_replace(&local_sb,
1363 0, len - (padding - 2),
1367 strbuf_utf8_replace(&local_sb,
1369 len - (padding - 2),
1373 strbuf_utf8_replace(&local_sb,
1374 padding - 2, len - (padding - 2),
1380 strbuf_addstr(sb, local_sb.buf);
1382 int sb_len = sb->len, offset = 0;
1383 if (c->flush_type == flush_left)
1384 offset = padding - len;
1385 else if (c->flush_type == flush_both)
1386 offset = (padding - len) / 2;
1388 * we calculate padding in columns, now
1389 * convert it back to chars
1391 padding = padding - len + local_sb.len;
1392 strbuf_grow(sb, padding);
1393 strbuf_setlen(sb, sb_len + padding);
1394 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1395 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1398 strbuf_release(&local_sb);
1399 c->flush_type = no_flush;
1400 return total_consumed;
1403 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1404 const char *placeholder,
1411 ADD_LF_BEFORE_NON_EMPTY,
1412 DEL_LF_BEFORE_EMPTY,
1413 ADD_SP_BEFORE_NON_EMPTY
1416 switch (placeholder[0]) {
1418 magic = DEL_LF_BEFORE_EMPTY;
1421 magic = ADD_LF_BEFORE_NON_EMPTY;
1424 magic = ADD_SP_BEFORE_NON_EMPTY;
1429 if (magic != NO_MAGIC)
1433 if (((struct format_commit_context *)context)->flush_type != no_flush)
1434 consumed = format_and_pad_commit(sb, placeholder, context);
1436 consumed = format_commit_one(sb, placeholder, context);
1437 if (magic == NO_MAGIC)
1440 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1441 while (sb->len && sb->buf[sb->len - 1] == '\n')
1442 strbuf_setlen(sb, sb->len - 1);
1443 } else if (orig_len != sb->len) {
1444 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1445 strbuf_insert(sb, orig_len, "\n", 1);
1446 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1447 strbuf_insert(sb, orig_len, " ", 1);
1449 return consumed + 1;
1452 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1455 struct userformat_want *w = context;
1457 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1460 switch (*placeholder) {
1468 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1470 struct strbuf dummy = STRBUF_INIT;
1477 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1478 strbuf_release(&dummy);
1481 void format_commit_message(const struct commit *commit,
1482 const char *format, struct strbuf *sb,
1483 const struct pretty_print_context *pretty_ctx)
1485 struct format_commit_context context;
1486 const char *output_enc = pretty_ctx->output_encoding;
1487 const char *utf8 = "UTF-8";
1489 memset(&context, 0, sizeof(context));
1490 context.commit = commit;
1491 context.pretty_ctx = pretty_ctx;
1492 context.wrap_start = sb->len;
1494 * convert a commit message to UTF-8 first
1495 * as far as 'format_commit_item' assumes it in UTF-8
1497 context.message = logmsg_reencode(commit,
1498 &context.commit_encoding,
1501 strbuf_expand(sb, format, format_commit_item, &context);
1502 rewrap_message_tail(sb, &context, 0, 0, 0);
1504 /* then convert a commit message to an actual output encoding */
1506 if (same_encoding(utf8, output_enc))
1509 if (context.commit_encoding &&
1510 !same_encoding(context.commit_encoding, utf8))
1511 output_enc = context.commit_encoding;
1516 char *out = reencode_string_len(sb->buf, sb->len,
1517 output_enc, utf8, &outsz);
1519 strbuf_attach(sb, out, outsz, outsz + 1);
1522 free(context.commit_encoding);
1523 unuse_commit_buffer(commit, context.message);
1524 free(context.signature_check.gpg_output);
1525 free(context.signature_check.signer);
1528 static void pp_header(struct pretty_print_context *pp,
1529 const char *encoding,
1530 const struct commit *commit,
1534 int parents_shown = 0;
1537 const char *line = *msg_p;
1538 int linelen = get_one_line(*msg_p);
1548 if (pp->fmt == CMIT_FMT_RAW) {
1549 strbuf_add(sb, line, linelen);
1553 if (starts_with(line, "parent ")) {
1555 die("bad parent line in commit");
1559 if (!parents_shown) {
1560 struct commit_list *parent;
1562 for (parent = commit->parents, num = 0;
1564 parent = parent->next, num++)
1566 /* with enough slop */
1567 strbuf_grow(sb, num * 50 + 20);
1568 add_merge_info(pp, sb, commit);
1573 * MEDIUM == DEFAULT shows only author with dates.
1574 * FULL shows both authors but not dates.
1575 * FULLER shows both authors and dates.
1577 if (starts_with(line, "author ")) {
1578 strbuf_grow(sb, linelen + 80);
1579 pp_user_info(pp, "Author", sb, line + 7, encoding);
1581 if (starts_with(line, "committer ") &&
1582 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1583 strbuf_grow(sb, linelen + 80);
1584 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1589 void pp_title_line(struct pretty_print_context *pp,
1592 const char *encoding,
1595 static const int max_length = 78; /* per rfc2047 */
1596 struct strbuf title;
1598 strbuf_init(&title, 80);
1599 *msg_p = format_subject(&title, *msg_p,
1600 pp->preserve_subject ? "\n" : " ");
1602 strbuf_grow(sb, title.len + 1024);
1604 strbuf_addstr(sb, pp->subject);
1605 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1606 add_rfc2047(sb, title.buf, title.len,
1607 encoding, RFC2047_SUBJECT);
1609 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1610 -last_line_length(sb), 1, max_length);
1612 strbuf_addbuf(sb, &title);
1614 strbuf_addch(sb, '\n');
1616 if (need_8bit_cte == 0) {
1618 for (i = 0; i < pp->in_body_headers.nr; i++) {
1619 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1626 if (need_8bit_cte > 0) {
1627 const char *header_fmt =
1628 "MIME-Version: 1.0\n"
1629 "Content-Type: text/plain; charset=%s\n"
1630 "Content-Transfer-Encoding: 8bit\n";
1631 strbuf_addf(sb, header_fmt, encoding);
1633 if (pp->after_subject) {
1634 strbuf_addstr(sb, pp->after_subject);
1636 if (pp->fmt == CMIT_FMT_EMAIL) {
1637 strbuf_addch(sb, '\n');
1640 if (pp->in_body_headers.nr) {
1642 for (i = 0; i < pp->in_body_headers.nr; i++) {
1643 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1644 free(pp->in_body_headers.items[i].string);
1646 string_list_clear(&pp->in_body_headers, 0);
1647 strbuf_addch(sb, '\n');
1650 strbuf_release(&title);
1653 void pp_remainder(struct pretty_print_context *pp,
1660 const char *line = *msg_p;
1661 int linelen = get_one_line(line);
1667 if (is_empty_line(line, &linelen)) {
1670 if (pp->fmt == CMIT_FMT_SHORT)
1675 strbuf_grow(sb, linelen + indent + 20);
1677 memset(sb->buf + sb->len, ' ', indent);
1678 strbuf_setlen(sb, sb->len + indent);
1680 strbuf_add(sb, line, linelen);
1681 strbuf_addch(sb, '\n');
1685 void pretty_print_commit(struct pretty_print_context *pp,
1686 const struct commit *commit,
1689 unsigned long beginning_of_body;
1692 const char *reencoded;
1693 const char *encoding;
1694 int need_8bit_cte = pp->need_8bit_cte;
1696 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1697 format_commit_message(commit, user_format, sb, pp);
1701 encoding = get_log_output_encoding();
1702 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1704 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1708 * We need to check and emit Content-type: to mark it
1709 * as 8-bit if we haven't done so.
1711 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1714 for (in_body = i = 0; (ch = msg[i]); i++) {
1716 /* author could be non 7-bit ASCII but
1717 * the log may be so; skip over the
1718 * header part first.
1720 if (ch == '\n' && msg[i+1] == '\n')
1723 else if (non_ascii(ch)) {
1730 pp_header(pp, encoding, commit, &msg, sb);
1731 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1732 strbuf_addch(sb, '\n');
1735 /* Skip excess blank lines at the beginning of body, if any... */
1736 msg = skip_empty_lines(msg);
1738 /* These formats treat the title line specially. */
1739 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1740 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1742 beginning_of_body = sb->len;
1743 if (pp->fmt != CMIT_FMT_ONELINE)
1744 pp_remainder(pp, &msg, sb, indent);
1747 /* Make sure there is an EOLN for the non-oneline case */
1748 if (pp->fmt != CMIT_FMT_ONELINE)
1749 strbuf_addch(sb, '\n');
1752 * The caller may append additional body text in e-mail
1753 * format. Make sure we did not strip the blank line
1754 * between the header and the body.
1756 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1757 strbuf_addch(sb, '\n');
1759 unuse_commit_buffer(commit, reencoded);
1762 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1765 struct pretty_print_context pp = {0};
1767 pretty_print_commit(&pp, commit, sb);