Merge branch 'jx/utf8-printf-width'
[alt-git.git] / pretty.c
blobeae57ad9d7f3b06a5a76f9d934825f9824def477
1 #include "cache.h"
2 #include "commit.h"
3 #include "utf8.h"
4 #include "diff.h"
5 #include "revision.h"
6 #include "string-list.h"
7 #include "mailmap.h"
8 #include "log-tree.h"
9 #include "notes.h"
10 #include "color.h"
11 #include "reflog-walk.h"
12 #include "gpg-interface.h"
14 static char *user_format;
15 static struct cmt_fmt_map {
16 const char *name;
17 enum cmit_fmt format;
18 int is_tformat;
19 int is_alias;
20 const char *user_format;
21 } *commit_formats;
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)
29 free(user_format);
30 user_format = xstrdup(cp);
31 if (is_tformat)
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;
39 const char *name;
40 const char *fmt;
41 int i;
43 if (prefixcmp(var, "pretty."))
44 return 0;
46 name = var + strlen("pretty.");
47 for (i = 0; i < builtin_formats_len; i++) {
48 if (!strcmp(commit_formats[i].name, name))
49 return 0;
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];
55 break;
59 if (!commit_format) {
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));
64 commit_formats_len++;
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;
75 else
76 commit_format->is_alias = 1;
77 commit_format->user_format = fmt;
79 return 0;
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;
108 int i;
110 if (num_redirections >= commit_formats_len)
111 die("invalid --pretty format: "
112 "'%s' references an alias which points to itself",
113 original);
115 for (i = 0; i < commit_formats_len; i++) {
116 size_t match_len;
118 if (prefixcmp(commit_formats[i].name, sought))
119 continue;
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,
130 original,
131 num_redirections+1);
134 return found;
137 static struct cmt_fmt_map *find_commit_format(const char *sought)
139 if (!commit_formats)
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;
150 if (!arg || !*arg) {
151 rev->commit_format = CMIT_FMT_DEFAULT;
152 return;
154 if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
155 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
156 return;
159 if (strchr(arg, '%')) {
160 save_user_format(rev, arg, 1);
161 return;
164 commit_format = find_commit_format(arg);
165 if (!commit_format)
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)
181 int ret = 0;
183 for (;;) {
184 char c = *msg++;
185 if (!c)
186 break;
187 ret++;
188 if (c == '\n')
189 break;
191 return ret;
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)
202 int ch;
203 if (!s)
204 return 0;
205 while ((ch = *s++) != '\0') {
206 if (non_ascii(ch))
207 return 1;
209 return 0;
212 static int is_rfc822_special(char ch)
214 switch (ch) {
215 case '(':
216 case ')':
217 case '<':
218 case '>':
219 case '[':
220 case ']':
221 case ':':
222 case ';':
223 case '@':
224 case ',':
225 case '.':
226 case '"':
227 case '\\':
228 return 1;
229 default:
230 return 0;
234 static int needs_rfc822_quoting(const char *s, int len)
236 int i;
237 for (i = 0; i < len; i++)
238 if (is_rfc822_special(s[i]))
239 return 1;
240 return 0;
243 static int last_line_length(struct strbuf *sb)
245 int i;
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')
250 break;
251 return sb->len - (i + 1);
254 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
256 int i;
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++) {
263 switch (s[i]) {
264 case '"':
265 case '\\':
266 strbuf_addch(out, '\\');
267 /* fall through */
268 default:
269 strbuf_addch(out, s[i]);
272 strbuf_addch(out, '"');
275 enum rfc2047_type {
276 RFC2047_SUBJECT,
277 RFC2047_ADDRESS,
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))
298 return 1;
301 * rule out special printable characters (' ' should be the only
302 * whitespace character considered printable, but be defensive and use
303 * isspace())
305 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
306 return 1;
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)
326 return 0;
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)
335 int i;
337 for (i = 0; i < len; i++) {
338 int ch = line[i];
339 if (non_ascii(ch) || ch == '\n')
340 return 1;
341 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
342 return 1;
345 return 0;
348 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
349 const char *encoding, enum rfc2047_type type)
351 static const int max_encoded_length = 76; /* per rfc2047 */
352 int i;
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? */
358 for (i = 0; i < len; i++) {
359 unsigned ch = line[i] & 0xFF;
360 int is_special = is_rfc2047_special(ch, type);
363 * According to RFC 2047, we could encode the special character
364 * ' ' (space) with '_' (underscore) for readability. But many
365 * programs do not understand this and just leave the
366 * underscore in place. Thus, we do nothing special here, which
367 * causes ' ' to be encoded as '=20', avoiding this problem.
370 if (line_len + 2 + (is_special ? 3 : 1) > max_encoded_length) {
371 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
372 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
375 if (is_special) {
376 strbuf_addf(sb, "=%02X", ch);
377 line_len += 3;
378 } else {
379 strbuf_addch(sb, ch);
380 line_len++;
383 strbuf_addstr(sb, "?=");
386 void pp_user_info(const struct pretty_print_context *pp,
387 const char *what, struct strbuf *sb,
388 const char *line, const char *encoding)
390 struct strbuf name;
391 struct strbuf mail;
392 struct ident_split ident;
393 int linelen;
394 char *line_end, *date;
395 const char *mailbuf, *namebuf;
396 size_t namelen, maillen;
397 int max_length = 78; /* per rfc2822 */
398 unsigned long time;
399 int tz;
401 if (pp->fmt == CMIT_FMT_ONELINE)
402 return;
404 line_end = strchr(line, '\n');
405 if (!line_end) {
406 line_end = strchr(line, '\0');
407 if (!line_end)
408 return;
411 linelen = ++line_end - line;
412 if (split_ident_line(&ident, line, linelen))
413 return;
416 mailbuf = ident.mail_begin;
417 maillen = ident.mail_end - ident.mail_begin;
418 namebuf = ident.name_begin;
419 namelen = ident.name_end - ident.name_begin;
421 if (pp->mailmap)
422 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
424 strbuf_init(&mail, 0);
425 strbuf_init(&name, 0);
427 strbuf_add(&mail, mailbuf, maillen);
428 strbuf_add(&name, namebuf, namelen);
430 namelen = name.len + mail.len + 3; /* ' ' + '<' + '>' */
431 time = strtoul(ident.date_begin, &date, 10);
432 tz = strtol(date, NULL, 10);
434 if (pp->fmt == CMIT_FMT_EMAIL) {
435 strbuf_addstr(sb, "From: ");
436 if (needs_rfc2047_encoding(name.buf, name.len, RFC2047_ADDRESS)) {
437 add_rfc2047(sb, name.buf, name.len,
438 encoding, RFC2047_ADDRESS);
439 max_length = 76; /* per rfc2047 */
440 } else if (needs_rfc822_quoting(name.buf, name.len)) {
441 struct strbuf quoted = STRBUF_INIT;
442 add_rfc822_quoted(&quoted, name.buf, name.len);
443 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
444 -6, 1, max_length);
445 strbuf_release(&quoted);
446 } else {
447 strbuf_add_wrapped_bytes(sb, name.buf, name.len,
448 -6, 1, max_length);
450 if (namelen - name.len + last_line_length(sb) > max_length)
451 strbuf_addch(sb, '\n');
453 strbuf_addf(sb, " <%s>\n", mail.buf);
454 } else {
455 strbuf_addf(sb, "%s: %.*s%s <%s>\n", what,
456 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
457 " ", name.buf, mail.buf);
460 strbuf_release(&mail);
461 strbuf_release(&name);
463 switch (pp->fmt) {
464 case CMIT_FMT_MEDIUM:
465 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, pp->date_mode));
466 break;
467 case CMIT_FMT_EMAIL:
468 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
469 break;
470 case CMIT_FMT_FULLER:
471 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
472 break;
473 default:
474 /* notin' */
475 break;
479 static int is_empty_line(const char *line, int *len_p)
481 int len = *len_p;
482 while (len && isspace(line[len-1]))
483 len--;
484 *len_p = len;
485 return !len;
488 static const char *skip_empty_lines(const char *msg)
490 for (;;) {
491 int linelen = get_one_line(msg);
492 int ll = linelen;
493 if (!linelen)
494 break;
495 if (!is_empty_line(msg, &ll))
496 break;
497 msg += linelen;
499 return msg;
502 static void add_merge_info(const struct pretty_print_context *pp,
503 struct strbuf *sb, const struct commit *commit)
505 struct commit_list *parent = commit->parents;
507 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
508 !parent || !parent->next)
509 return;
511 strbuf_addstr(sb, "Merge:");
513 while (parent) {
514 struct commit *p = parent->item;
515 const char *hex = NULL;
516 if (pp->abbrev)
517 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
518 if (!hex)
519 hex = sha1_to_hex(p->object.sha1);
520 parent = parent->next;
522 strbuf_addf(sb, " %s", hex);
524 strbuf_addch(sb, '\n');
527 static char *get_header(const struct commit *commit, const char *msg,
528 const char *key)
530 int key_len = strlen(key);
531 const char *line = msg;
533 while (line) {
534 const char *eol = strchr(line, '\n'), *next;
536 if (line == eol)
537 return NULL;
538 if (!eol) {
539 warning("malformed commit (header is missing newline): %s",
540 sha1_to_hex(commit->object.sha1));
541 eol = line + strlen(line);
542 next = NULL;
543 } else
544 next = eol + 1;
545 if (eol - line > key_len &&
546 !strncmp(line, key, key_len) &&
547 line[key_len] == ' ') {
548 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
550 line = next;
552 return NULL;
555 static char *replace_encoding_header(char *buf, const char *encoding)
557 struct strbuf tmp = STRBUF_INIT;
558 size_t start, len;
559 char *cp = buf;
561 /* guess if there is an encoding header before a \n\n */
562 while (strncmp(cp, "encoding ", strlen("encoding "))) {
563 cp = strchr(cp, '\n');
564 if (!cp || *++cp == '\n')
565 return buf;
567 start = cp - buf;
568 cp = strchr(cp, '\n');
569 if (!cp)
570 return buf; /* should not happen but be defensive */
571 len = cp + 1 - (buf + start);
573 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
574 if (is_encoding_utf8(encoding)) {
575 /* we have re-coded to UTF-8; drop the header */
576 strbuf_remove(&tmp, start, len);
577 } else {
578 /* just replaces XXXX in 'encoding XXXX\n' */
579 strbuf_splice(&tmp, start + strlen("encoding "),
580 len - strlen("encoding \n"),
581 encoding, strlen(encoding));
583 return strbuf_detach(&tmp, NULL);
586 char *logmsg_reencode(const struct commit *commit,
587 const char *output_encoding)
589 static const char *utf8 = "UTF-8";
590 const char *use_encoding;
591 char *encoding;
592 char *msg = commit->buffer;
593 char *out;
595 if (!msg) {
596 enum object_type type;
597 unsigned long size;
599 msg = read_sha1_file(commit->object.sha1, &type, &size);
600 if (!msg)
601 die("Cannot read commit object %s",
602 sha1_to_hex(commit->object.sha1));
603 if (type != OBJ_COMMIT)
604 die("Expected commit for '%s', got %s",
605 sha1_to_hex(commit->object.sha1), typename(type));
608 if (!output_encoding || !*output_encoding)
609 return msg;
610 encoding = get_header(commit, msg, "encoding");
611 use_encoding = encoding ? encoding : utf8;
612 if (same_encoding(use_encoding, output_encoding)) {
614 * No encoding work to be done. If we have no encoding header
615 * at all, then there's nothing to do, and we can return the
616 * message verbatim (whether newly allocated or not).
618 if (!encoding)
619 return msg;
622 * Otherwise, we still want to munge the encoding header in the
623 * result, which will be done by modifying the buffer. If we
624 * are using a fresh copy, we can reuse it. But if we are using
625 * the cached copy from commit->buffer, we need to duplicate it
626 * to avoid munging commit->buffer.
628 out = msg;
629 if (out == commit->buffer)
630 out = xstrdup(out);
632 else {
634 * There's actual encoding work to do. Do the reencoding, which
635 * still leaves the header to be replaced in the next step. At
636 * this point, we are done with msg. If we allocated a fresh
637 * copy, we can free it.
639 out = reencode_string(msg, output_encoding, use_encoding);
640 if (out && msg != commit->buffer)
641 free(msg);
645 * This replacement actually consumes the buffer we hand it, so we do
646 * not have to worry about freeing the old "out" here.
648 if (out)
649 out = replace_encoding_header(out, output_encoding);
651 free(encoding);
653 * If the re-encoding failed, out might be NULL here; in that
654 * case we just return the commit message verbatim.
656 return out ? out : msg;
659 void logmsg_free(char *msg, const struct commit *commit)
661 if (msg != commit->buffer)
662 free(msg);
665 static int mailmap_name(const char **email, size_t *email_len,
666 const char **name, size_t *name_len)
668 static struct string_list *mail_map;
669 if (!mail_map) {
670 mail_map = xcalloc(1, sizeof(*mail_map));
671 read_mailmap(mail_map, NULL);
673 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
676 static size_t format_person_part(struct strbuf *sb, char part,
677 const char *msg, int len, enum date_mode dmode)
679 /* currently all placeholders have same length */
680 const int placeholder_len = 2;
681 int tz;
682 unsigned long date = 0;
683 struct ident_split s;
684 const char *name, *mail;
685 size_t maillen, namelen;
687 if (split_ident_line(&s, msg, len) < 0)
688 goto skip;
690 name = s.name_begin;
691 namelen = s.name_end - s.name_begin;
692 mail = s.mail_begin;
693 maillen = s.mail_end - s.mail_begin;
695 if (part == 'N' || part == 'E') /* mailmap lookup */
696 mailmap_name(&mail, &maillen, &name, &namelen);
697 if (part == 'n' || part == 'N') { /* name */
698 strbuf_add(sb, name, namelen);
699 return placeholder_len;
701 if (part == 'e' || part == 'E') { /* email */
702 strbuf_add(sb, mail, maillen);
703 return placeholder_len;
706 if (!s.date_begin)
707 goto skip;
709 date = strtoul(s.date_begin, NULL, 10);
711 if (part == 't') { /* date, UNIX timestamp */
712 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
713 return placeholder_len;
716 /* parse tz */
717 tz = strtoul(s.tz_begin + 1, NULL, 10);
718 if (*s.tz_begin == '-')
719 tz = -tz;
721 switch (part) {
722 case 'd': /* date */
723 strbuf_addstr(sb, show_date(date, tz, dmode));
724 return placeholder_len;
725 case 'D': /* date, RFC2822 style */
726 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
727 return placeholder_len;
728 case 'r': /* date, relative */
729 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
730 return placeholder_len;
731 case 'i': /* date, ISO 8601 */
732 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
733 return placeholder_len;
736 skip:
738 * reading from either a bogus commit, or a reflog entry with
739 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
740 * to compute a valid return value.
742 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
743 || part == 'D' || part == 'r' || part == 'i')
744 return placeholder_len;
746 return 0; /* unknown placeholder */
749 struct chunk {
750 size_t off;
751 size_t len;
754 struct format_commit_context {
755 const struct commit *commit;
756 const struct pretty_print_context *pretty_ctx;
757 unsigned commit_header_parsed:1;
758 unsigned commit_message_parsed:1;
759 unsigned commit_signature_parsed:1;
760 struct {
761 char *gpg_output;
762 char good_bad;
763 char *signer;
764 } signature;
765 char *message;
766 size_t width, indent1, indent2;
768 /* These offsets are relative to the start of the commit message. */
769 struct chunk author;
770 struct chunk committer;
771 struct chunk encoding;
772 size_t message_off;
773 size_t subject_off;
774 size_t body_off;
776 /* The following ones are relative to the result struct strbuf. */
777 struct chunk abbrev_commit_hash;
778 struct chunk abbrev_tree_hash;
779 struct chunk abbrev_parent_hashes;
780 size_t wrap_start;
783 static int add_again(struct strbuf *sb, struct chunk *chunk)
785 if (chunk->len) {
786 strbuf_adddup(sb, chunk->off, chunk->len);
787 return 1;
791 * We haven't seen this chunk before. Our caller is surely
792 * going to add it the hard way now. Remember the most likely
793 * start of the to-be-added chunk: the current end of the
794 * struct strbuf.
796 chunk->off = sb->len;
797 return 0;
800 static void parse_commit_header(struct format_commit_context *context)
802 const char *msg = context->message;
803 int i;
805 for (i = 0; msg[i]; i++) {
806 int eol;
807 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
808 ; /* do nothing */
810 if (i == eol) {
811 break;
812 } else if (!prefixcmp(msg + i, "author ")) {
813 context->author.off = i + 7;
814 context->author.len = eol - i - 7;
815 } else if (!prefixcmp(msg + i, "committer ")) {
816 context->committer.off = i + 10;
817 context->committer.len = eol - i - 10;
818 } else if (!prefixcmp(msg + i, "encoding ")) {
819 context->encoding.off = i + 9;
820 context->encoding.len = eol - i - 9;
822 i = eol;
824 context->message_off = i;
825 context->commit_header_parsed = 1;
828 static int istitlechar(char c)
830 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
831 (c >= '0' && c <= '9') || c == '.' || c == '_';
834 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
836 size_t trimlen;
837 size_t start_len = sb->len;
838 int space = 2;
840 for (; *msg && *msg != '\n'; msg++) {
841 if (istitlechar(*msg)) {
842 if (space == 1)
843 strbuf_addch(sb, '-');
844 space = 0;
845 strbuf_addch(sb, *msg);
846 if (*msg == '.')
847 while (*(msg+1) == '.')
848 msg++;
849 } else
850 space |= 1;
853 /* trim any trailing '.' or '-' characters */
854 trimlen = 0;
855 while (sb->len - trimlen > start_len &&
856 (sb->buf[sb->len - 1 - trimlen] == '.'
857 || sb->buf[sb->len - 1 - trimlen] == '-'))
858 trimlen++;
859 strbuf_remove(sb, sb->len - trimlen, trimlen);
862 const char *format_subject(struct strbuf *sb, const char *msg,
863 const char *line_separator)
865 int first = 1;
867 for (;;) {
868 const char *line = msg;
869 int linelen = get_one_line(line);
871 msg += linelen;
872 if (!linelen || is_empty_line(line, &linelen))
873 break;
875 if (!sb)
876 continue;
877 strbuf_grow(sb, linelen + 2);
878 if (!first)
879 strbuf_addstr(sb, line_separator);
880 strbuf_add(sb, line, linelen);
881 first = 0;
883 return msg;
886 static void parse_commit_message(struct format_commit_context *c)
888 const char *msg = c->message + c->message_off;
889 const char *start = c->message;
891 msg = skip_empty_lines(msg);
892 c->subject_off = msg - start;
894 msg = format_subject(NULL, msg, NULL);
895 msg = skip_empty_lines(msg);
896 c->body_off = msg - start;
898 c->commit_message_parsed = 1;
901 static void format_decoration(struct strbuf *sb, const struct commit *commit)
903 struct name_decoration *d;
904 const char *prefix = " (";
906 load_ref_decorations(DECORATE_SHORT_REFS);
907 d = lookup_decoration(&name_decoration, &commit->object);
908 while (d) {
909 strbuf_addstr(sb, prefix);
910 prefix = ", ";
911 strbuf_addstr(sb, d->name);
912 d = d->next;
914 if (prefix[0] == ',')
915 strbuf_addch(sb, ')');
918 static void strbuf_wrap(struct strbuf *sb, size_t pos,
919 size_t width, size_t indent1, size_t indent2)
921 struct strbuf tmp = STRBUF_INIT;
923 if (pos)
924 strbuf_add(&tmp, sb->buf, pos);
925 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
926 (int) indent1, (int) indent2, (int) width);
927 strbuf_swap(&tmp, sb);
928 strbuf_release(&tmp);
931 static void rewrap_message_tail(struct strbuf *sb,
932 struct format_commit_context *c,
933 size_t new_width, size_t new_indent1,
934 size_t new_indent2)
936 if (c->width == new_width && c->indent1 == new_indent1 &&
937 c->indent2 == new_indent2)
938 return;
939 if (c->wrap_start < sb->len)
940 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
941 c->wrap_start = sb->len;
942 c->width = new_width;
943 c->indent1 = new_indent1;
944 c->indent2 = new_indent2;
947 static struct {
948 char result;
949 const char *check;
950 } signature_check[] = {
951 { 'G', ": Good signature from " },
952 { 'B', ": BAD signature from " },
955 static void parse_signature_lines(struct format_commit_context *ctx)
957 const char *buf = ctx->signature.gpg_output;
958 int i;
960 for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
961 const char *found = strstr(buf, signature_check[i].check);
962 const char *next;
963 if (!found)
964 continue;
965 ctx->signature.good_bad = signature_check[i].result;
966 found += strlen(signature_check[i].check);
967 next = strchrnul(found, '\n');
968 ctx->signature.signer = xmemdupz(found, next - found);
969 break;
973 static void parse_commit_signature(struct format_commit_context *ctx)
975 struct strbuf payload = STRBUF_INIT;
976 struct strbuf signature = STRBUF_INIT;
977 struct strbuf gpg_output = STRBUF_INIT;
978 int status;
980 ctx->commit_signature_parsed = 1;
982 if (parse_signed_commit(ctx->commit->object.sha1,
983 &payload, &signature) <= 0)
984 goto out;
985 status = verify_signed_buffer(payload.buf, payload.len,
986 signature.buf, signature.len,
987 &gpg_output);
988 if (status && !gpg_output.len)
989 goto out;
990 ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
991 parse_signature_lines(ctx);
993 out:
994 strbuf_release(&gpg_output);
995 strbuf_release(&payload);
996 strbuf_release(&signature);
1000 static int format_reflog_person(struct strbuf *sb,
1001 char part,
1002 struct reflog_walk_info *log,
1003 enum date_mode dmode)
1005 const char *ident;
1007 if (!log)
1008 return 2;
1010 ident = get_reflog_ident(log);
1011 if (!ident)
1012 return 2;
1014 return format_person_part(sb, part, ident, strlen(ident), dmode);
1017 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
1018 void *context)
1020 struct format_commit_context *c = context;
1021 const struct commit *commit = c->commit;
1022 const char *msg = c->message;
1023 struct commit_list *p;
1024 int h1, h2;
1026 /* these are independent of the commit */
1027 switch (placeholder[0]) {
1028 case 'C':
1029 if (placeholder[1] == '(') {
1030 const char *begin = placeholder + 2;
1031 const char *end = strchr(begin, ')');
1032 char color[COLOR_MAXLEN];
1034 if (!end)
1035 return 0;
1036 if (!prefixcmp(begin, "auto,")) {
1037 if (!want_color(c->pretty_ctx->color))
1038 return end - placeholder + 1;
1039 begin += 5;
1041 color_parse_mem(begin,
1042 end - begin,
1043 "--pretty format", color);
1044 strbuf_addstr(sb, color);
1045 return end - placeholder + 1;
1047 if (!prefixcmp(placeholder + 1, "red")) {
1048 strbuf_addstr(sb, GIT_COLOR_RED);
1049 return 4;
1050 } else if (!prefixcmp(placeholder + 1, "green")) {
1051 strbuf_addstr(sb, GIT_COLOR_GREEN);
1052 return 6;
1053 } else if (!prefixcmp(placeholder + 1, "blue")) {
1054 strbuf_addstr(sb, GIT_COLOR_BLUE);
1055 return 5;
1056 } else if (!prefixcmp(placeholder + 1, "reset")) {
1057 strbuf_addstr(sb, GIT_COLOR_RESET);
1058 return 6;
1059 } else
1060 return 0;
1061 case 'n': /* newline */
1062 strbuf_addch(sb, '\n');
1063 return 1;
1064 case 'x':
1065 /* %x00 == NUL, %x0a == LF, etc. */
1066 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1067 h1 <= 16 &&
1068 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1069 h2 <= 16) {
1070 strbuf_addch(sb, (h1<<4)|h2);
1071 return 3;
1072 } else
1073 return 0;
1074 case 'w':
1075 if (placeholder[1] == '(') {
1076 unsigned long width = 0, indent1 = 0, indent2 = 0;
1077 char *next;
1078 const char *start = placeholder + 2;
1079 const char *end = strchr(start, ')');
1080 if (!end)
1081 return 0;
1082 if (end > start) {
1083 width = strtoul(start, &next, 10);
1084 if (*next == ',') {
1085 indent1 = strtoul(next + 1, &next, 10);
1086 if (*next == ',') {
1087 indent2 = strtoul(next + 1,
1088 &next, 10);
1091 if (*next != ')')
1092 return 0;
1094 rewrap_message_tail(sb, c, width, indent1, indent2);
1095 return end - placeholder + 1;
1096 } else
1097 return 0;
1100 /* these depend on the commit */
1101 if (!commit->object.parsed)
1102 parse_object(commit->object.sha1);
1104 switch (placeholder[0]) {
1105 case 'H': /* commit hash */
1106 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1107 return 1;
1108 case 'h': /* abbreviated commit hash */
1109 if (add_again(sb, &c->abbrev_commit_hash))
1110 return 1;
1111 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1112 c->pretty_ctx->abbrev));
1113 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1114 return 1;
1115 case 'T': /* tree hash */
1116 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1117 return 1;
1118 case 't': /* abbreviated tree hash */
1119 if (add_again(sb, &c->abbrev_tree_hash))
1120 return 1;
1121 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1122 c->pretty_ctx->abbrev));
1123 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1124 return 1;
1125 case 'P': /* parent hashes */
1126 for (p = commit->parents; p; p = p->next) {
1127 if (p != commit->parents)
1128 strbuf_addch(sb, ' ');
1129 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1131 return 1;
1132 case 'p': /* abbreviated parent hashes */
1133 if (add_again(sb, &c->abbrev_parent_hashes))
1134 return 1;
1135 for (p = commit->parents; p; p = p->next) {
1136 if (p != commit->parents)
1137 strbuf_addch(sb, ' ');
1138 strbuf_addstr(sb, find_unique_abbrev(
1139 p->item->object.sha1,
1140 c->pretty_ctx->abbrev));
1142 c->abbrev_parent_hashes.len = sb->len -
1143 c->abbrev_parent_hashes.off;
1144 return 1;
1145 case 'm': /* left/right/bottom */
1146 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1147 return 1;
1148 case 'd':
1149 format_decoration(sb, commit);
1150 return 1;
1151 case 'g': /* reflog info */
1152 switch(placeholder[1]) {
1153 case 'd': /* reflog selector */
1154 case 'D':
1155 if (c->pretty_ctx->reflog_info)
1156 get_reflog_selector(sb,
1157 c->pretty_ctx->reflog_info,
1158 c->pretty_ctx->date_mode,
1159 c->pretty_ctx->date_mode_explicit,
1160 (placeholder[1] == 'd'));
1161 return 2;
1162 case 's': /* reflog message */
1163 if (c->pretty_ctx->reflog_info)
1164 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1165 return 2;
1166 case 'n':
1167 case 'N':
1168 case 'e':
1169 case 'E':
1170 return format_reflog_person(sb,
1171 placeholder[1],
1172 c->pretty_ctx->reflog_info,
1173 c->pretty_ctx->date_mode);
1175 return 0; /* unknown %g placeholder */
1176 case 'N':
1177 if (c->pretty_ctx->notes_message) {
1178 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1179 return 1;
1181 return 0;
1184 if (placeholder[0] == 'G') {
1185 if (!c->commit_signature_parsed)
1186 parse_commit_signature(c);
1187 switch (placeholder[1]) {
1188 case 'G':
1189 if (c->signature.gpg_output)
1190 strbuf_addstr(sb, c->signature.gpg_output);
1191 break;
1192 case '?':
1193 switch (c->signature.good_bad) {
1194 case 'G':
1195 case 'B':
1196 strbuf_addch(sb, c->signature.good_bad);
1198 break;
1199 case 'S':
1200 if (c->signature.signer)
1201 strbuf_addstr(sb, c->signature.signer);
1202 break;
1204 return 2;
1208 /* For the rest we have to parse the commit header. */
1209 if (!c->commit_header_parsed)
1210 parse_commit_header(c);
1212 switch (placeholder[0]) {
1213 case 'a': /* author ... */
1214 return format_person_part(sb, placeholder[1],
1215 msg + c->author.off, c->author.len,
1216 c->pretty_ctx->date_mode);
1217 case 'c': /* committer ... */
1218 return format_person_part(sb, placeholder[1],
1219 msg + c->committer.off, c->committer.len,
1220 c->pretty_ctx->date_mode);
1221 case 'e': /* encoding */
1222 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1223 return 1;
1224 case 'B': /* raw body */
1225 /* message_off is always left at the initial newline */
1226 strbuf_addstr(sb, msg + c->message_off + 1);
1227 return 1;
1230 /* Now we need to parse the commit message. */
1231 if (!c->commit_message_parsed)
1232 parse_commit_message(c);
1234 switch (placeholder[0]) {
1235 case 's': /* subject */
1236 format_subject(sb, msg + c->subject_off, " ");
1237 return 1;
1238 case 'f': /* sanitized subject */
1239 format_sanitized_subject(sb, msg + c->subject_off);
1240 return 1;
1241 case 'b': /* body */
1242 strbuf_addstr(sb, msg + c->body_off);
1243 return 1;
1245 return 0; /* unknown placeholder */
1248 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1249 void *context)
1251 int consumed;
1252 size_t orig_len;
1253 enum {
1254 NO_MAGIC,
1255 ADD_LF_BEFORE_NON_EMPTY,
1256 DEL_LF_BEFORE_EMPTY,
1257 ADD_SP_BEFORE_NON_EMPTY
1258 } magic = NO_MAGIC;
1260 switch (placeholder[0]) {
1261 case '-':
1262 magic = DEL_LF_BEFORE_EMPTY;
1263 break;
1264 case '+':
1265 magic = ADD_LF_BEFORE_NON_EMPTY;
1266 break;
1267 case ' ':
1268 magic = ADD_SP_BEFORE_NON_EMPTY;
1269 break;
1270 default:
1271 break;
1273 if (magic != NO_MAGIC)
1274 placeholder++;
1276 orig_len = sb->len;
1277 consumed = format_commit_one(sb, placeholder, context);
1278 if (magic == NO_MAGIC)
1279 return consumed;
1281 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1282 while (sb->len && sb->buf[sb->len - 1] == '\n')
1283 strbuf_setlen(sb, sb->len - 1);
1284 } else if (orig_len != sb->len) {
1285 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1286 strbuf_insert(sb, orig_len, "\n", 1);
1287 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1288 strbuf_insert(sb, orig_len, " ", 1);
1290 return consumed + 1;
1293 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1294 void *context)
1296 struct userformat_want *w = context;
1298 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1299 placeholder++;
1301 switch (*placeholder) {
1302 case 'N':
1303 w->notes = 1;
1304 break;
1306 return 0;
1309 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1311 struct strbuf dummy = STRBUF_INIT;
1313 if (!fmt) {
1314 if (!user_format)
1315 return;
1316 fmt = user_format;
1318 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1319 strbuf_release(&dummy);
1322 void format_commit_message(const struct commit *commit,
1323 const char *format, struct strbuf *sb,
1324 const struct pretty_print_context *pretty_ctx)
1326 struct format_commit_context context;
1327 const char *output_enc = pretty_ctx->output_encoding;
1329 memset(&context, 0, sizeof(context));
1330 context.commit = commit;
1331 context.pretty_ctx = pretty_ctx;
1332 context.wrap_start = sb->len;
1333 context.message = logmsg_reencode(commit, output_enc);
1335 strbuf_expand(sb, format, format_commit_item, &context);
1336 rewrap_message_tail(sb, &context, 0, 0, 0);
1338 logmsg_free(context.message, commit);
1339 free(context.signature.gpg_output);
1340 free(context.signature.signer);
1343 static void pp_header(const struct pretty_print_context *pp,
1344 const char *encoding,
1345 const struct commit *commit,
1346 const char **msg_p,
1347 struct strbuf *sb)
1349 int parents_shown = 0;
1351 for (;;) {
1352 const char *line = *msg_p;
1353 int linelen = get_one_line(*msg_p);
1355 if (!linelen)
1356 return;
1357 *msg_p += linelen;
1359 if (linelen == 1)
1360 /* End of header */
1361 return;
1363 if (pp->fmt == CMIT_FMT_RAW) {
1364 strbuf_add(sb, line, linelen);
1365 continue;
1368 if (!prefixcmp(line, "parent ")) {
1369 if (linelen != 48)
1370 die("bad parent line in commit");
1371 continue;
1374 if (!parents_shown) {
1375 struct commit_list *parent;
1376 int num;
1377 for (parent = commit->parents, num = 0;
1378 parent;
1379 parent = parent->next, num++)
1381 /* with enough slop */
1382 strbuf_grow(sb, num * 50 + 20);
1383 add_merge_info(pp, sb, commit);
1384 parents_shown = 1;
1388 * MEDIUM == DEFAULT shows only author with dates.
1389 * FULL shows both authors but not dates.
1390 * FULLER shows both authors and dates.
1392 if (!prefixcmp(line, "author ")) {
1393 strbuf_grow(sb, linelen + 80);
1394 pp_user_info(pp, "Author", sb, line + 7, encoding);
1396 if (!prefixcmp(line, "committer ") &&
1397 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1398 strbuf_grow(sb, linelen + 80);
1399 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1404 void pp_title_line(const struct pretty_print_context *pp,
1405 const char **msg_p,
1406 struct strbuf *sb,
1407 const char *encoding,
1408 int need_8bit_cte)
1410 static const int max_length = 78; /* per rfc2047 */
1411 struct strbuf title;
1413 strbuf_init(&title, 80);
1414 *msg_p = format_subject(&title, *msg_p,
1415 pp->preserve_subject ? "\n" : " ");
1417 strbuf_grow(sb, title.len + 1024);
1418 if (pp->subject) {
1419 strbuf_addstr(sb, pp->subject);
1420 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1421 add_rfc2047(sb, title.buf, title.len,
1422 encoding, RFC2047_SUBJECT);
1423 else
1424 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1425 -last_line_length(sb), 1, max_length);
1426 } else {
1427 strbuf_addbuf(sb, &title);
1429 strbuf_addch(sb, '\n');
1431 if (need_8bit_cte > 0) {
1432 const char *header_fmt =
1433 "MIME-Version: 1.0\n"
1434 "Content-Type: text/plain; charset=%s\n"
1435 "Content-Transfer-Encoding: 8bit\n";
1436 strbuf_addf(sb, header_fmt, encoding);
1438 if (pp->after_subject) {
1439 strbuf_addstr(sb, pp->after_subject);
1441 if (pp->fmt == CMIT_FMT_EMAIL) {
1442 strbuf_addch(sb, '\n');
1444 strbuf_release(&title);
1447 void pp_remainder(const struct pretty_print_context *pp,
1448 const char **msg_p,
1449 struct strbuf *sb,
1450 int indent)
1452 int first = 1;
1453 for (;;) {
1454 const char *line = *msg_p;
1455 int linelen = get_one_line(line);
1456 *msg_p += linelen;
1458 if (!linelen)
1459 break;
1461 if (is_empty_line(line, &linelen)) {
1462 if (first)
1463 continue;
1464 if (pp->fmt == CMIT_FMT_SHORT)
1465 break;
1467 first = 0;
1469 strbuf_grow(sb, linelen + indent + 20);
1470 if (indent) {
1471 memset(sb->buf + sb->len, ' ', indent);
1472 strbuf_setlen(sb, sb->len + indent);
1474 strbuf_add(sb, line, linelen);
1475 strbuf_addch(sb, '\n');
1479 void pretty_print_commit(const struct pretty_print_context *pp,
1480 const struct commit *commit,
1481 struct strbuf *sb)
1483 unsigned long beginning_of_body;
1484 int indent = 4;
1485 const char *msg;
1486 char *reencoded;
1487 const char *encoding;
1488 int need_8bit_cte = pp->need_8bit_cte;
1490 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1491 format_commit_message(commit, user_format, sb, pp);
1492 return;
1495 encoding = get_log_output_encoding();
1496 msg = reencoded = logmsg_reencode(commit, encoding);
1498 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1499 indent = 0;
1502 * We need to check and emit Content-type: to mark it
1503 * as 8-bit if we haven't done so.
1505 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1506 int i, ch, in_body;
1508 for (in_body = i = 0; (ch = msg[i]); i++) {
1509 if (!in_body) {
1510 /* author could be non 7-bit ASCII but
1511 * the log may be so; skip over the
1512 * header part first.
1514 if (ch == '\n' && msg[i+1] == '\n')
1515 in_body = 1;
1517 else if (non_ascii(ch)) {
1518 need_8bit_cte = 1;
1519 break;
1524 pp_header(pp, encoding, commit, &msg, sb);
1525 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1526 strbuf_addch(sb, '\n');
1529 /* Skip excess blank lines at the beginning of body, if any... */
1530 msg = skip_empty_lines(msg);
1532 /* These formats treat the title line specially. */
1533 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1534 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1536 beginning_of_body = sb->len;
1537 if (pp->fmt != CMIT_FMT_ONELINE)
1538 pp_remainder(pp, &msg, sb, indent);
1539 strbuf_rtrim(sb);
1541 /* Make sure there is an EOLN for the non-oneline case */
1542 if (pp->fmt != CMIT_FMT_ONELINE)
1543 strbuf_addch(sb, '\n');
1546 * The caller may append additional body text in e-mail
1547 * format. Make sure we did not strip the blank line
1548 * between the header and the body.
1550 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1551 strbuf_addch(sb, '\n');
1553 logmsg_free(reencoded, commit);
1556 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1557 struct strbuf *sb)
1559 struct pretty_print_context pp = {0};
1560 pp.fmt = fmt;
1561 pretty_print_commit(&pp, commit, sb);