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 (prefixcmp(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 (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
71 commit_format->is_tformat = fmt[0] == 't';
72 fmt = strchr(fmt, ':') + 1;
73 } else if (strchr(fmt, '%'))
74 commit_format->is_tformat = 1;
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 (prefixcmp(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 (!prefixcmp(arg, "format:") || !prefixcmp(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 void pp_user_info(const struct pretty_print_context *pp,
397 const char *what, struct strbuf *sb,
398 const char *line, const char *encoding)
402 struct ident_split ident;
404 char *line_end, *date;
405 const char *mailbuf, *namebuf;
406 size_t namelen, maillen;
407 int max_length = 78; /* per rfc2822 */
411 if (pp->fmt == CMIT_FMT_ONELINE)
414 line_end = strchr(line, '\n');
416 line_end = strchr(line, '\0');
421 linelen = ++line_end - line;
422 if (split_ident_line(&ident, line, linelen))
426 mailbuf = ident.mail_begin;
427 maillen = ident.mail_end - ident.mail_begin;
428 namebuf = ident.name_begin;
429 namelen = ident.name_end - ident.name_begin;
432 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
434 strbuf_init(&mail, 0);
435 strbuf_init(&name, 0);
437 strbuf_add(&mail, mailbuf, maillen);
438 strbuf_add(&name, namebuf, namelen);
440 namelen = name.len + mail.len + 3; /* ' ' + '<' + '>' */
441 time = strtoul(ident.date_begin, &date, 10);
442 tz = strtol(date, NULL, 10);
444 if (pp->fmt == CMIT_FMT_EMAIL) {
445 strbuf_addstr(sb, "From: ");
446 if (needs_rfc2047_encoding(name.buf, name.len, RFC2047_ADDRESS)) {
447 add_rfc2047(sb, name.buf, name.len,
448 encoding, RFC2047_ADDRESS);
449 max_length = 76; /* per rfc2047 */
450 } else if (needs_rfc822_quoting(name.buf, name.len)) {
451 struct strbuf quoted = STRBUF_INIT;
452 add_rfc822_quoted("ed, name.buf, name.len);
453 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
455 strbuf_release("ed);
457 strbuf_add_wrapped_bytes(sb, name.buf, name.len,
460 if (namelen - name.len + last_line_length(sb) > max_length)
461 strbuf_addch(sb, '\n');
463 strbuf_addf(sb, " <%s>\n", mail.buf);
465 strbuf_addf(sb, "%s: %.*s%s <%s>\n", what,
466 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
467 " ", name.buf, mail.buf);
470 strbuf_release(&mail);
471 strbuf_release(&name);
474 case CMIT_FMT_MEDIUM:
475 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, pp->date_mode));
478 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
480 case CMIT_FMT_FULLER:
481 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
489 static int is_empty_line(const char *line, int *len_p)
492 while (len && isspace(line[len-1]))
498 static const char *skip_empty_lines(const char *msg)
501 int linelen = get_one_line(msg);
505 if (!is_empty_line(msg, &ll))
512 static void add_merge_info(const struct pretty_print_context *pp,
513 struct strbuf *sb, const struct commit *commit)
515 struct commit_list *parent = commit->parents;
517 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
518 !parent || !parent->next)
521 strbuf_addstr(sb, "Merge:");
524 struct commit *p = parent->item;
525 const char *hex = NULL;
527 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
529 hex = sha1_to_hex(p->object.sha1);
530 parent = parent->next;
532 strbuf_addf(sb, " %s", hex);
534 strbuf_addch(sb, '\n');
537 static char *get_header(const struct commit *commit, const char *msg,
540 int key_len = strlen(key);
541 const char *line = msg;
544 const char *eol = strchr(line, '\n'), *next;
549 warning("malformed commit (header is missing newline): %s",
550 sha1_to_hex(commit->object.sha1));
551 eol = line + strlen(line);
555 if (eol - line > key_len &&
556 !strncmp(line, key, key_len) &&
557 line[key_len] == ' ') {
558 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
565 static char *replace_encoding_header(char *buf, const char *encoding)
567 struct strbuf tmp = STRBUF_INIT;
571 /* guess if there is an encoding header before a \n\n */
572 while (strncmp(cp, "encoding ", strlen("encoding "))) {
573 cp = strchr(cp, '\n');
574 if (!cp || *++cp == '\n')
578 cp = strchr(cp, '\n');
580 return buf; /* should not happen but be defensive */
581 len = cp + 1 - (buf + start);
583 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
584 if (is_encoding_utf8(encoding)) {
585 /* we have re-coded to UTF-8; drop the header */
586 strbuf_remove(&tmp, start, len);
588 /* just replaces XXXX in 'encoding XXXX\n' */
589 strbuf_splice(&tmp, start + strlen("encoding "),
590 len - strlen("encoding \n"),
591 encoding, strlen(encoding));
593 return strbuf_detach(&tmp, NULL);
596 char *logmsg_reencode(const struct commit *commit,
597 char **commit_encoding,
598 const char *output_encoding)
600 static const char *utf8 = "UTF-8";
601 const char *use_encoding;
603 char *msg = commit->buffer;
607 enum object_type type;
610 msg = read_sha1_file(commit->object.sha1, &type, &size);
612 die("Cannot read commit object %s",
613 sha1_to_hex(commit->object.sha1));
614 if (type != OBJ_COMMIT)
615 die("Expected commit for '%s', got %s",
616 sha1_to_hex(commit->object.sha1), typename(type));
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 commit->buffer, we need to duplicate it
643 * to avoid munging commit->buffer.
646 if (out == commit->buffer)
651 * There's actual encoding work to do. Do the reencoding, which
652 * still leaves the header to be replaced in the next step. At
653 * this point, we are done with msg. If we allocated a fresh
654 * copy, we can free it.
656 out = reencode_string(msg, output_encoding, use_encoding);
657 if (out && msg != commit->buffer)
662 * This replacement actually consumes the buffer we hand it, so we do
663 * not have to worry about freeing the old "out" here.
666 out = replace_encoding_header(out, output_encoding);
668 if (!commit_encoding)
671 * If the re-encoding failed, out might be NULL here; in that
672 * case we just return the commit message verbatim.
674 return out ? out : msg;
677 void logmsg_free(char *msg, const struct commit *commit)
679 if (msg != commit->buffer)
683 static int mailmap_name(const char **email, size_t *email_len,
684 const char **name, size_t *name_len)
686 static struct string_list *mail_map;
688 mail_map = xcalloc(1, sizeof(*mail_map));
689 read_mailmap(mail_map, NULL);
691 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
694 static size_t format_person_part(struct strbuf *sb, char part,
695 const char *msg, int len, enum date_mode dmode)
697 /* currently all placeholders have same length */
698 const int placeholder_len = 2;
700 unsigned long date = 0;
701 struct ident_split s;
702 const char *name, *mail;
703 size_t maillen, namelen;
705 if (split_ident_line(&s, msg, len) < 0)
709 namelen = s.name_end - s.name_begin;
711 maillen = s.mail_end - s.mail_begin;
713 if (part == 'N' || part == 'E') /* mailmap lookup */
714 mailmap_name(&mail, &maillen, &name, &namelen);
715 if (part == 'n' || part == 'N') { /* name */
716 strbuf_add(sb, name, namelen);
717 return placeholder_len;
719 if (part == 'e' || part == 'E') { /* email */
720 strbuf_add(sb, mail, maillen);
721 return placeholder_len;
727 date = strtoul(s.date_begin, NULL, 10);
729 if (part == 't') { /* date, UNIX timestamp */
730 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
731 return placeholder_len;
735 tz = strtoul(s.tz_begin + 1, NULL, 10);
736 if (*s.tz_begin == '-')
741 strbuf_addstr(sb, show_date(date, tz, dmode));
742 return placeholder_len;
743 case 'D': /* date, RFC2822 style */
744 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
745 return placeholder_len;
746 case 'r': /* date, relative */
747 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
748 return placeholder_len;
749 case 'i': /* date, ISO 8601 */
750 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
751 return placeholder_len;
756 * reading from either a bogus commit, or a reflog entry with
757 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
758 * to compute a valid return value.
760 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
761 || part == 'D' || part == 'r' || part == 'i')
762 return placeholder_len;
764 return 0; /* unknown placeholder */
776 flush_left_and_steal,
787 struct format_commit_context {
788 const struct commit *commit;
789 const struct pretty_print_context *pretty_ctx;
790 unsigned commit_header_parsed:1;
791 unsigned commit_message_parsed:1;
792 struct signature_check signature_check;
793 enum flush_type flush_type;
794 enum trunc_type truncate;
796 char *commit_encoding;
797 size_t width, indent1, indent2;
801 /* These offsets are relative to the start of the commit message. */
803 struct chunk committer;
808 /* The following ones are relative to the result struct strbuf. */
809 struct chunk abbrev_commit_hash;
810 struct chunk abbrev_tree_hash;
811 struct chunk abbrev_parent_hashes;
815 static int add_again(struct strbuf *sb, struct chunk *chunk)
818 strbuf_adddup(sb, chunk->off, chunk->len);
823 * We haven't seen this chunk before. Our caller is surely
824 * going to add it the hard way now. Remember the most likely
825 * start of the to-be-added chunk: the current end of the
828 chunk->off = sb->len;
832 static void parse_commit_header(struct format_commit_context *context)
834 const char *msg = context->message;
837 for (i = 0; msg[i]; i++) {
839 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
844 } else if (!prefixcmp(msg + i, "author ")) {
845 context->author.off = i + 7;
846 context->author.len = eol - i - 7;
847 } else if (!prefixcmp(msg + i, "committer ")) {
848 context->committer.off = i + 10;
849 context->committer.len = eol - i - 10;
853 context->message_off = i;
854 context->commit_header_parsed = 1;
857 static int istitlechar(char c)
859 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
860 (c >= '0' && c <= '9') || c == '.' || c == '_';
863 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
866 size_t start_len = sb->len;
869 for (; *msg && *msg != '\n'; msg++) {
870 if (istitlechar(*msg)) {
872 strbuf_addch(sb, '-');
874 strbuf_addch(sb, *msg);
876 while (*(msg+1) == '.')
882 /* trim any trailing '.' or '-' characters */
884 while (sb->len - trimlen > start_len &&
885 (sb->buf[sb->len - 1 - trimlen] == '.'
886 || sb->buf[sb->len - 1 - trimlen] == '-'))
888 strbuf_remove(sb, sb->len - trimlen, trimlen);
891 const char *format_subject(struct strbuf *sb, const char *msg,
892 const char *line_separator)
897 const char *line = msg;
898 int linelen = get_one_line(line);
901 if (!linelen || is_empty_line(line, &linelen))
906 strbuf_grow(sb, linelen + 2);
908 strbuf_addstr(sb, line_separator);
909 strbuf_add(sb, line, linelen);
915 static void parse_commit_message(struct format_commit_context *c)
917 const char *msg = c->message + c->message_off;
918 const char *start = c->message;
920 msg = skip_empty_lines(msg);
921 c->subject_off = msg - start;
923 msg = format_subject(NULL, msg, NULL);
924 msg = skip_empty_lines(msg);
925 c->body_off = msg - start;
927 c->commit_message_parsed = 1;
930 static void strbuf_wrap(struct strbuf *sb, size_t pos,
931 size_t width, size_t indent1, size_t indent2)
933 struct strbuf tmp = STRBUF_INIT;
936 strbuf_add(&tmp, sb->buf, pos);
937 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
938 (int) indent1, (int) indent2, (int) width);
939 strbuf_swap(&tmp, sb);
940 strbuf_release(&tmp);
943 static void rewrap_message_tail(struct strbuf *sb,
944 struct format_commit_context *c,
945 size_t new_width, size_t new_indent1,
948 if (c->width == new_width && c->indent1 == new_indent1 &&
949 c->indent2 == new_indent2)
951 if (c->wrap_start < sb->len)
952 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
953 c->wrap_start = sb->len;
954 c->width = new_width;
955 c->indent1 = new_indent1;
956 c->indent2 = new_indent2;
959 static int format_reflog_person(struct strbuf *sb,
961 struct reflog_walk_info *log,
962 enum date_mode dmode)
969 ident = get_reflog_ident(log);
973 return format_person_part(sb, part, ident, strlen(ident), dmode);
976 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
977 const char *placeholder,
978 struct format_commit_context *c)
980 if (placeholder[1] == '(') {
981 const char *begin = placeholder + 2;
982 const char *end = strchr(begin, ')');
983 char color[COLOR_MAXLEN];
987 if (!prefixcmp(begin, "auto,")) {
988 if (!want_color(c->pretty_ctx->color))
989 return end - placeholder + 1;
992 color_parse_mem(begin,
994 "--pretty format", color);
995 strbuf_addstr(sb, color);
996 return end - placeholder + 1;
998 if (!prefixcmp(placeholder + 1, "red")) {
999 strbuf_addstr(sb, GIT_COLOR_RED);
1001 } else if (!prefixcmp(placeholder + 1, "green")) {
1002 strbuf_addstr(sb, GIT_COLOR_GREEN);
1004 } else if (!prefixcmp(placeholder + 1, "blue")) {
1005 strbuf_addstr(sb, GIT_COLOR_BLUE);
1007 } else if (!prefixcmp(placeholder + 1, "reset")) {
1008 strbuf_addstr(sb, GIT_COLOR_RESET);
1014 static size_t parse_padding_placeholder(struct strbuf *sb,
1015 const char *placeholder,
1016 struct format_commit_context *c)
1018 const char *ch = placeholder;
1019 enum flush_type flush_type;
1024 flush_type = flush_right;
1028 flush_type = flush_both;
1030 } else if (*ch == '>') {
1031 flush_type = flush_left_and_steal;
1034 flush_type = flush_left;
1040 /* the next value means "wide enough to that column" */
1047 const char *start = ch + 1;
1048 const char *end = start + strcspn(start, ",)");
1051 if (!end || end == start)
1053 width = strtoul(start, &next, 10);
1054 if (next == start || width == 0)
1056 c->padding = to_column ? -width : width;
1057 c->flush_type = flush_type;
1061 end = strchr(start, ')');
1062 if (!end || end == start)
1064 if (!prefixcmp(start, "trunc)"))
1065 c->truncate = trunc_right;
1066 else if (!prefixcmp(start, "ltrunc)"))
1067 c->truncate = trunc_left;
1068 else if (!prefixcmp(start, "mtrunc)"))
1069 c->truncate = trunc_middle;
1073 c->truncate = trunc_none;
1075 return end - placeholder + 1;
1080 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1081 const char *placeholder,
1084 struct format_commit_context *c = context;
1085 const struct commit *commit = c->commit;
1086 const char *msg = c->message;
1087 struct commit_list *p;
1090 /* these are independent of the commit */
1091 switch (placeholder[0]) {
1093 if (!prefixcmp(placeholder + 1, "(auto)")) {
1095 return 7; /* consumed 7 bytes, "C(auto)" */
1097 int ret = parse_color(sb, placeholder, c);
1101 * Otherwise, we decided to treat %C<unknown>
1102 * as a literal string, and the previous
1103 * %C(auto) is still valid.
1107 case 'n': /* newline */
1108 strbuf_addch(sb, '\n');
1111 /* %x00 == NUL, %x0a == LF, etc. */
1112 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1114 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1116 strbuf_addch(sb, (h1<<4)|h2);
1121 if (placeholder[1] == '(') {
1122 unsigned long width = 0, indent1 = 0, indent2 = 0;
1124 const char *start = placeholder + 2;
1125 const char *end = strchr(start, ')');
1129 width = strtoul(start, &next, 10);
1131 indent1 = strtoul(next + 1, &next, 10);
1133 indent2 = strtoul(next + 1,
1140 rewrap_message_tail(sb, c, width, indent1, indent2);
1141 return end - placeholder + 1;
1147 return parse_padding_placeholder(sb, placeholder, c);
1150 /* these depend on the commit */
1151 if (!commit->object.parsed)
1152 parse_object(commit->object.sha1);
1154 switch (placeholder[0]) {
1155 case 'H': /* commit hash */
1156 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1157 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1158 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1160 case 'h': /* abbreviated commit hash */
1161 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1162 if (add_again(sb, &c->abbrev_commit_hash)) {
1163 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1166 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1167 c->pretty_ctx->abbrev));
1168 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1169 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1171 case 'T': /* tree hash */
1172 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1174 case 't': /* abbreviated tree hash */
1175 if (add_again(sb, &c->abbrev_tree_hash))
1177 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1178 c->pretty_ctx->abbrev));
1179 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1181 case 'P': /* parent hashes */
1182 for (p = commit->parents; p; p = p->next) {
1183 if (p != commit->parents)
1184 strbuf_addch(sb, ' ');
1185 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1188 case 'p': /* abbreviated parent hashes */
1189 if (add_again(sb, &c->abbrev_parent_hashes))
1191 for (p = commit->parents; p; p = p->next) {
1192 if (p != commit->parents)
1193 strbuf_addch(sb, ' ');
1194 strbuf_addstr(sb, find_unique_abbrev(
1195 p->item->object.sha1,
1196 c->pretty_ctx->abbrev));
1198 c->abbrev_parent_hashes.len = sb->len -
1199 c->abbrev_parent_hashes.off;
1201 case 'm': /* left/right/bottom */
1202 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1205 load_ref_decorations(DECORATE_SHORT_REFS);
1206 format_decorations(sb, commit, c->auto_color);
1208 case 'g': /* reflog info */
1209 switch(placeholder[1]) {
1210 case 'd': /* reflog selector */
1212 if (c->pretty_ctx->reflog_info)
1213 get_reflog_selector(sb,
1214 c->pretty_ctx->reflog_info,
1215 c->pretty_ctx->date_mode,
1216 c->pretty_ctx->date_mode_explicit,
1217 (placeholder[1] == 'd'));
1219 case 's': /* reflog message */
1220 if (c->pretty_ctx->reflog_info)
1221 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1227 return format_reflog_person(sb,
1229 c->pretty_ctx->reflog_info,
1230 c->pretty_ctx->date_mode);
1232 return 0; /* unknown %g placeholder */
1234 if (c->pretty_ctx->notes_message) {
1235 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1241 if (placeholder[0] == 'G') {
1242 if (!c->signature_check.result)
1243 check_commit_signature(c->commit, &(c->signature_check));
1244 switch (placeholder[1]) {
1246 if (c->signature_check.gpg_output)
1247 strbuf_addstr(sb, c->signature_check.gpg_output);
1250 switch (c->signature_check.result) {
1255 strbuf_addch(sb, c->signature_check.result);
1259 if (c->signature_check.signer)
1260 strbuf_addstr(sb, c->signature_check.signer);
1263 if (c->signature_check.key)
1264 strbuf_addstr(sb, c->signature_check.key);
1271 /* For the rest we have to parse the commit header. */
1272 if (!c->commit_header_parsed)
1273 parse_commit_header(c);
1275 switch (placeholder[0]) {
1276 case 'a': /* author ... */
1277 return format_person_part(sb, placeholder[1],
1278 msg + c->author.off, c->author.len,
1279 c->pretty_ctx->date_mode);
1280 case 'c': /* committer ... */
1281 return format_person_part(sb, placeholder[1],
1282 msg + c->committer.off, c->committer.len,
1283 c->pretty_ctx->date_mode);
1284 case 'e': /* encoding */
1285 if (c->commit_encoding)
1286 strbuf_addstr(sb, c->commit_encoding);
1288 case 'B': /* raw body */
1289 /* message_off is always left at the initial newline */
1290 strbuf_addstr(sb, msg + c->message_off + 1);
1294 /* Now we need to parse the commit message. */
1295 if (!c->commit_message_parsed)
1296 parse_commit_message(c);
1298 switch (placeholder[0]) {
1299 case 's': /* subject */
1300 format_subject(sb, msg + c->subject_off, " ");
1302 case 'f': /* sanitized subject */
1303 format_sanitized_subject(sb, msg + c->subject_off);
1305 case 'b': /* body */
1306 strbuf_addstr(sb, msg + c->body_off);
1309 return 0; /* unknown placeholder */
1312 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1313 const char *placeholder,
1314 struct format_commit_context *c)
1316 struct strbuf local_sb = STRBUF_INIT;
1317 int total_consumed = 0, len, padding = c->padding;
1319 const char *start = strrchr(sb->buf, '\n');
1323 occupied = utf8_strnwidth(start, -1, 1);
1324 padding = (-padding) - occupied;
1327 int modifier = *placeholder == 'C';
1328 int consumed = format_commit_one(&local_sb, placeholder, c);
1329 total_consumed += consumed;
1334 placeholder += consumed;
1335 if (*placeholder != '%')
1340 len = utf8_strnwidth(local_sb.buf, -1, 1);
1342 if (c->flush_type == flush_left_and_steal) {
1343 const char *ch = sb->buf + sb->len - 1;
1344 while (len > padding && ch > sb->buf) {
1351 /* check for trailing ansi sequences */
1355 while (ch - p < 10 && *p != '\033')
1358 ch + 1 - p != display_mode_esc_sequence_len(p))
1361 * got a good ansi sequence, put it back to
1362 * local_sb as we're cutting sb
1364 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1367 strbuf_setlen(sb, ch + 1 - sb->buf);
1368 c->flush_type = flush_left;
1371 if (len > padding) {
1372 switch (c->truncate) {
1374 strbuf_utf8_replace(&local_sb,
1375 0, len - (padding - 2),
1379 strbuf_utf8_replace(&local_sb,
1381 len - (padding - 2),
1385 strbuf_utf8_replace(&local_sb,
1386 padding - 2, len - (padding - 2),
1392 strbuf_addstr(sb, local_sb.buf);
1394 int sb_len = sb->len, offset = 0;
1395 if (c->flush_type == flush_left)
1396 offset = padding - len;
1397 else if (c->flush_type == flush_both)
1398 offset = (padding - len) / 2;
1400 * we calculate padding in columns, now
1401 * convert it back to chars
1403 padding = padding - len + local_sb.len;
1404 strbuf_grow(sb, padding);
1405 strbuf_setlen(sb, sb_len + padding);
1406 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1407 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1410 strbuf_release(&local_sb);
1411 c->flush_type = no_flush;
1412 return total_consumed;
1415 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1416 const char *placeholder,
1423 ADD_LF_BEFORE_NON_EMPTY,
1424 DEL_LF_BEFORE_EMPTY,
1425 ADD_SP_BEFORE_NON_EMPTY
1428 switch (placeholder[0]) {
1430 magic = DEL_LF_BEFORE_EMPTY;
1433 magic = ADD_LF_BEFORE_NON_EMPTY;
1436 magic = ADD_SP_BEFORE_NON_EMPTY;
1441 if (magic != NO_MAGIC)
1445 if (((struct format_commit_context *)context)->flush_type != no_flush)
1446 consumed = format_and_pad_commit(sb, placeholder, context);
1448 consumed = format_commit_one(sb, placeholder, context);
1449 if (magic == NO_MAGIC)
1452 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1453 while (sb->len && sb->buf[sb->len - 1] == '\n')
1454 strbuf_setlen(sb, sb->len - 1);
1455 } else if (orig_len != sb->len) {
1456 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1457 strbuf_insert(sb, orig_len, "\n", 1);
1458 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1459 strbuf_insert(sb, orig_len, " ", 1);
1461 return consumed + 1;
1464 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1467 struct userformat_want *w = context;
1469 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1472 switch (*placeholder) {
1480 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1482 struct strbuf dummy = STRBUF_INIT;
1489 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1490 strbuf_release(&dummy);
1493 void format_commit_message(const struct commit *commit,
1494 const char *format, struct strbuf *sb,
1495 const struct pretty_print_context *pretty_ctx)
1497 struct format_commit_context context;
1498 const char *output_enc = pretty_ctx->output_encoding;
1499 const char *utf8 = "UTF-8";
1501 memset(&context, 0, sizeof(context));
1502 context.commit = commit;
1503 context.pretty_ctx = pretty_ctx;
1504 context.wrap_start = sb->len;
1505 context.message = logmsg_reencode(commit,
1506 &context.commit_encoding,
1509 strbuf_expand(sb, format, format_commit_item, &context);
1510 rewrap_message_tail(sb, &context, 0, 0, 0);
1513 if (same_encoding(utf8, output_enc))
1516 if (context.commit_encoding &&
1517 !same_encoding(context.commit_encoding, utf8))
1518 output_enc = context.commit_encoding;
1523 char *out = reencode_string_len(sb->buf, sb->len,
1524 output_enc, utf8, &outsz);
1526 strbuf_attach(sb, out, outsz, outsz + 1);
1529 free(context.commit_encoding);
1530 logmsg_free(context.message, commit);
1531 free(context.signature_check.gpg_output);
1532 free(context.signature_check.signer);
1535 static void pp_header(const struct pretty_print_context *pp,
1536 const char *encoding,
1537 const struct commit *commit,
1541 int parents_shown = 0;
1544 const char *line = *msg_p;
1545 int linelen = get_one_line(*msg_p);
1555 if (pp->fmt == CMIT_FMT_RAW) {
1556 strbuf_add(sb, line, linelen);
1560 if (!prefixcmp(line, "parent ")) {
1562 die("bad parent line in commit");
1566 if (!parents_shown) {
1567 struct commit_list *parent;
1569 for (parent = commit->parents, num = 0;
1571 parent = parent->next, num++)
1573 /* with enough slop */
1574 strbuf_grow(sb, num * 50 + 20);
1575 add_merge_info(pp, sb, commit);
1580 * MEDIUM == DEFAULT shows only author with dates.
1581 * FULL shows both authors but not dates.
1582 * FULLER shows both authors and dates.
1584 if (!prefixcmp(line, "author ")) {
1585 strbuf_grow(sb, linelen + 80);
1586 pp_user_info(pp, "Author", sb, line + 7, encoding);
1588 if (!prefixcmp(line, "committer ") &&
1589 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1590 strbuf_grow(sb, linelen + 80);
1591 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1596 void pp_title_line(const struct pretty_print_context *pp,
1599 const char *encoding,
1602 static const int max_length = 78; /* per rfc2047 */
1603 struct strbuf title;
1605 strbuf_init(&title, 80);
1606 *msg_p = format_subject(&title, *msg_p,
1607 pp->preserve_subject ? "\n" : " ");
1609 strbuf_grow(sb, title.len + 1024);
1611 strbuf_addstr(sb, pp->subject);
1612 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1613 add_rfc2047(sb, title.buf, title.len,
1614 encoding, RFC2047_SUBJECT);
1616 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1617 -last_line_length(sb), 1, max_length);
1619 strbuf_addbuf(sb, &title);
1621 strbuf_addch(sb, '\n');
1623 if (need_8bit_cte > 0) {
1624 const char *header_fmt =
1625 "MIME-Version: 1.0\n"
1626 "Content-Type: text/plain; charset=%s\n"
1627 "Content-Transfer-Encoding: 8bit\n";
1628 strbuf_addf(sb, header_fmt, encoding);
1630 if (pp->after_subject) {
1631 strbuf_addstr(sb, pp->after_subject);
1633 if (pp->fmt == CMIT_FMT_EMAIL) {
1634 strbuf_addch(sb, '\n');
1636 strbuf_release(&title);
1639 void pp_remainder(const struct pretty_print_context *pp,
1646 const char *line = *msg_p;
1647 int linelen = get_one_line(line);
1653 if (is_empty_line(line, &linelen)) {
1656 if (pp->fmt == CMIT_FMT_SHORT)
1661 strbuf_grow(sb, linelen + indent + 20);
1663 memset(sb->buf + sb->len, ' ', indent);
1664 strbuf_setlen(sb, sb->len + indent);
1666 strbuf_add(sb, line, linelen);
1667 strbuf_addch(sb, '\n');
1671 void pretty_print_commit(const struct pretty_print_context *pp,
1672 const struct commit *commit,
1675 unsigned long beginning_of_body;
1679 const char *encoding;
1680 int need_8bit_cte = pp->need_8bit_cte;
1682 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1683 format_commit_message(commit, user_format, sb, pp);
1687 encoding = get_log_output_encoding();
1688 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1690 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1694 * We need to check and emit Content-type: to mark it
1695 * as 8-bit if we haven't done so.
1697 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1700 for (in_body = i = 0; (ch = msg[i]); i++) {
1702 /* author could be non 7-bit ASCII but
1703 * the log may be so; skip over the
1704 * header part first.
1706 if (ch == '\n' && msg[i+1] == '\n')
1709 else if (non_ascii(ch)) {
1716 pp_header(pp, encoding, commit, &msg, sb);
1717 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1718 strbuf_addch(sb, '\n');
1721 /* Skip excess blank lines at the beginning of body, if any... */
1722 msg = skip_empty_lines(msg);
1724 /* These formats treat the title line specially. */
1725 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1726 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1728 beginning_of_body = sb->len;
1729 if (pp->fmt != CMIT_FMT_ONELINE)
1730 pp_remainder(pp, &msg, sb, indent);
1733 /* Make sure there is an EOLN for the non-oneline case */
1734 if (pp->fmt != CMIT_FMT_ONELINE)
1735 strbuf_addch(sb, '\n');
1738 * The caller may append additional body text in e-mail
1739 * format. Make sure we did not strip the blank line
1740 * between the header and the body.
1742 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1743 strbuf_addch(sb, '\n');
1745 logmsg_free(reencoded, commit);
1748 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1751 struct pretty_print_context pp = {0};
1753 pretty_print_commit(&pp, commit, sb);