prompt: fix show upstream with svn and zsh
[git.git] / pretty.c
blobacbfceb5fea49e49f5cc8eaef816c4b0a6ffe05a
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, size_t 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? */
359 while (len) {
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,
397 enum date_mode mode)
399 unsigned long date = 0;
400 int tz = 0;
402 if (ident->date_begin && ident->date_end)
403 date = strtoul(ident->date_begin, NULL, 10);
404 if (ident->tz_begin && ident->tz_end)
405 tz = strtol(ident->tz_begin, NULL, 10);
406 return show_date(date, tz, mode);
409 void pp_user_info(const struct pretty_print_context *pp,
410 const char *what, struct strbuf *sb,
411 const char *line, const char *encoding)
413 struct strbuf name;
414 struct strbuf mail;
415 struct ident_split ident;
416 int linelen;
417 char *line_end;
418 const char *mailbuf, *namebuf;
419 size_t namelen, maillen;
420 int max_length = 78; /* per rfc2822 */
422 if (pp->fmt == CMIT_FMT_ONELINE)
423 return;
425 line_end = strchr(line, '\n');
426 if (!line_end) {
427 line_end = strchr(line, '\0');
428 if (!line_end)
429 return;
432 linelen = ++line_end - line;
433 if (split_ident_line(&ident, line, linelen))
434 return;
437 mailbuf = ident.mail_begin;
438 maillen = ident.mail_end - ident.mail_begin;
439 namebuf = ident.name_begin;
440 namelen = ident.name_end - ident.name_begin;
442 if (pp->mailmap)
443 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
445 strbuf_init(&mail, 0);
446 strbuf_init(&name, 0);
448 strbuf_add(&mail, mailbuf, maillen);
449 strbuf_add(&name, namebuf, namelen);
451 namelen = name.len + mail.len + 3; /* ' ' + '<' + '>' */
453 if (pp->fmt == CMIT_FMT_EMAIL) {
454 strbuf_addstr(sb, "From: ");
455 if (needs_rfc2047_encoding(name.buf, name.len, RFC2047_ADDRESS)) {
456 add_rfc2047(sb, name.buf, name.len,
457 encoding, RFC2047_ADDRESS);
458 max_length = 76; /* per rfc2047 */
459 } else if (needs_rfc822_quoting(name.buf, name.len)) {
460 struct strbuf quoted = STRBUF_INIT;
461 add_rfc822_quoted(&quoted, name.buf, name.len);
462 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
463 -6, 1, max_length);
464 strbuf_release(&quoted);
465 } else {
466 strbuf_add_wrapped_bytes(sb, name.buf, name.len,
467 -6, 1, max_length);
469 if (namelen - name.len + last_line_length(sb) > max_length)
470 strbuf_addch(sb, '\n');
472 strbuf_addf(sb, " <%s>\n", mail.buf);
473 } else {
474 strbuf_addf(sb, "%s: %.*s%s <%s>\n", what,
475 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
476 " ", name.buf, mail.buf);
479 strbuf_release(&mail);
480 strbuf_release(&name);
482 switch (pp->fmt) {
483 case CMIT_FMT_MEDIUM:
484 strbuf_addf(sb, "Date: %s\n",
485 show_ident_date(&ident, pp->date_mode));
486 break;
487 case CMIT_FMT_EMAIL:
488 strbuf_addf(sb, "Date: %s\n",
489 show_ident_date(&ident, DATE_RFC2822));
490 break;
491 case CMIT_FMT_FULLER:
492 strbuf_addf(sb, "%sDate: %s\n", what,
493 show_ident_date(&ident, pp->date_mode));
494 break;
495 default:
496 /* notin' */
497 break;
501 static int is_empty_line(const char *line, int *len_p)
503 int len = *len_p;
504 while (len && isspace(line[len-1]))
505 len--;
506 *len_p = len;
507 return !len;
510 static const char *skip_empty_lines(const char *msg)
512 for (;;) {
513 int linelen = get_one_line(msg);
514 int ll = linelen;
515 if (!linelen)
516 break;
517 if (!is_empty_line(msg, &ll))
518 break;
519 msg += linelen;
521 return msg;
524 static void add_merge_info(const struct pretty_print_context *pp,
525 struct strbuf *sb, const struct commit *commit)
527 struct commit_list *parent = commit->parents;
529 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
530 !parent || !parent->next)
531 return;
533 strbuf_addstr(sb, "Merge:");
535 while (parent) {
536 struct commit *p = parent->item;
537 const char *hex = NULL;
538 if (pp->abbrev)
539 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
540 if (!hex)
541 hex = sha1_to_hex(p->object.sha1);
542 parent = parent->next;
544 strbuf_addf(sb, " %s", hex);
546 strbuf_addch(sb, '\n');
549 static char *get_header(const struct commit *commit, const char *msg,
550 const char *key)
552 int key_len = strlen(key);
553 const char *line = msg;
555 while (line) {
556 const char *eol = strchr(line, '\n'), *next;
558 if (line == eol)
559 return NULL;
560 if (!eol) {
561 warning("malformed commit (header is missing newline): %s",
562 sha1_to_hex(commit->object.sha1));
563 eol = line + strlen(line);
564 next = NULL;
565 } else
566 next = eol + 1;
567 if (eol - line > key_len &&
568 !strncmp(line, key, key_len) &&
569 line[key_len] == ' ') {
570 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
572 line = next;
574 return NULL;
577 static char *replace_encoding_header(char *buf, const char *encoding)
579 struct strbuf tmp = STRBUF_INIT;
580 size_t start, len;
581 char *cp = buf;
583 /* guess if there is an encoding header before a \n\n */
584 while (strncmp(cp, "encoding ", strlen("encoding "))) {
585 cp = strchr(cp, '\n');
586 if (!cp || *++cp == '\n')
587 return buf;
589 start = cp - buf;
590 cp = strchr(cp, '\n');
591 if (!cp)
592 return buf; /* should not happen but be defensive */
593 len = cp + 1 - (buf + start);
595 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
596 if (is_encoding_utf8(encoding)) {
597 /* we have re-coded to UTF-8; drop the header */
598 strbuf_remove(&tmp, start, len);
599 } else {
600 /* just replaces XXXX in 'encoding XXXX\n' */
601 strbuf_splice(&tmp, start + strlen("encoding "),
602 len - strlen("encoding \n"),
603 encoding, strlen(encoding));
605 return strbuf_detach(&tmp, NULL);
608 char *logmsg_reencode(const struct commit *commit,
609 const char *output_encoding)
611 static const char *utf8 = "UTF-8";
612 const char *use_encoding;
613 char *encoding;
614 char *msg = commit->buffer;
615 char *out;
617 if (!msg) {
618 enum object_type type;
619 unsigned long size;
621 msg = read_sha1_file(commit->object.sha1, &type, &size);
622 if (!msg)
623 die("Cannot read commit object %s",
624 sha1_to_hex(commit->object.sha1));
625 if (type != OBJ_COMMIT)
626 die("Expected commit for '%s', got %s",
627 sha1_to_hex(commit->object.sha1), typename(type));
630 if (!output_encoding || !*output_encoding)
631 return msg;
632 encoding = get_header(commit, msg, "encoding");
633 use_encoding = encoding ? encoding : utf8;
634 if (same_encoding(use_encoding, output_encoding)) {
636 * No encoding work to be done. If we have no encoding header
637 * at all, then there's nothing to do, and we can return the
638 * message verbatim (whether newly allocated or not).
640 if (!encoding)
641 return msg;
644 * Otherwise, we still want to munge the encoding header in the
645 * result, which will be done by modifying the buffer. If we
646 * are using a fresh copy, we can reuse it. But if we are using
647 * the cached copy from commit->buffer, we need to duplicate it
648 * to avoid munging commit->buffer.
650 out = msg;
651 if (out == commit->buffer)
652 out = xstrdup(out);
654 else {
656 * There's actual encoding work to do. Do the reencoding, which
657 * still leaves the header to be replaced in the next step. At
658 * this point, we are done with msg. If we allocated a fresh
659 * copy, we can free it.
661 out = reencode_string(msg, output_encoding, use_encoding);
662 if (out && msg != commit->buffer)
663 free(msg);
667 * This replacement actually consumes the buffer we hand it, so we do
668 * not have to worry about freeing the old "out" here.
670 if (out)
671 out = replace_encoding_header(out, output_encoding);
673 free(encoding);
675 * If the re-encoding failed, out might be NULL here; in that
676 * case we just return the commit message verbatim.
678 return out ? out : msg;
681 void logmsg_free(char *msg, const struct commit *commit)
683 if (msg != commit->buffer)
684 free(msg);
687 static int mailmap_name(const char **email, size_t *email_len,
688 const char **name, size_t *name_len)
690 static struct string_list *mail_map;
691 if (!mail_map) {
692 mail_map = xcalloc(1, sizeof(*mail_map));
693 read_mailmap(mail_map, NULL);
695 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
698 static size_t format_person_part(struct strbuf *sb, char part,
699 const char *msg, int len, enum date_mode dmode)
701 /* currently all placeholders have same length */
702 const int placeholder_len = 2;
703 struct ident_split s;
704 const char *name, *mail;
705 size_t maillen, namelen;
707 if (split_ident_line(&s, msg, len) < 0)
708 goto skip;
710 name = s.name_begin;
711 namelen = s.name_end - s.name_begin;
712 mail = s.mail_begin;
713 maillen = s.mail_end - s.mail_begin;
715 if (part == 'N' || part == 'E') /* mailmap lookup */
716 mailmap_name(&mail, &maillen, &name, &namelen);
717 if (part == 'n' || part == 'N') { /* name */
718 strbuf_add(sb, name, namelen);
719 return placeholder_len;
721 if (part == 'e' || part == 'E') { /* email */
722 strbuf_add(sb, mail, maillen);
723 return placeholder_len;
726 if (!s.date_begin)
727 goto skip;
729 if (part == 't') { /* date, UNIX timestamp */
730 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
731 return placeholder_len;
734 switch (part) {
735 case 'd': /* date */
736 strbuf_addstr(sb, show_ident_date(&s, dmode));
737 return placeholder_len;
738 case 'D': /* date, RFC2822 style */
739 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
740 return placeholder_len;
741 case 'r': /* date, relative */
742 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
743 return placeholder_len;
744 case 'i': /* date, ISO 8601 */
745 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
746 return placeholder_len;
749 skip:
751 * reading from either a bogus commit, or a reflog entry with
752 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
753 * to compute a valid return value.
755 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
756 || part == 'D' || part == 'r' || part == 'i')
757 return placeholder_len;
759 return 0; /* unknown placeholder */
762 struct chunk {
763 size_t off;
764 size_t len;
767 struct format_commit_context {
768 const struct commit *commit;
769 const struct pretty_print_context *pretty_ctx;
770 unsigned commit_header_parsed:1;
771 unsigned commit_message_parsed:1;
772 unsigned commit_signature_parsed:1;
773 struct {
774 char *gpg_output;
775 char *gpg_status;
776 char good_bad;
777 char *signer;
778 char *key;
779 } signature;
780 char *message;
781 size_t width, indent1, indent2;
783 /* These offsets are relative to the start of the commit message. */
784 struct chunk author;
785 struct chunk committer;
786 struct chunk encoding;
787 size_t message_off;
788 size_t subject_off;
789 size_t body_off;
791 /* The following ones are relative to the result struct strbuf. */
792 struct chunk abbrev_commit_hash;
793 struct chunk abbrev_tree_hash;
794 struct chunk abbrev_parent_hashes;
795 size_t wrap_start;
798 static int add_again(struct strbuf *sb, struct chunk *chunk)
800 if (chunk->len) {
801 strbuf_adddup(sb, chunk->off, chunk->len);
802 return 1;
806 * We haven't seen this chunk before. Our caller is surely
807 * going to add it the hard way now. Remember the most likely
808 * start of the to-be-added chunk: the current end of the
809 * struct strbuf.
811 chunk->off = sb->len;
812 return 0;
815 static void parse_commit_header(struct format_commit_context *context)
817 const char *msg = context->message;
818 int i;
820 for (i = 0; msg[i]; i++) {
821 int eol;
822 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
823 ; /* do nothing */
825 if (i == eol) {
826 break;
827 } else if (!prefixcmp(msg + i, "author ")) {
828 context->author.off = i + 7;
829 context->author.len = eol - i - 7;
830 } else if (!prefixcmp(msg + i, "committer ")) {
831 context->committer.off = i + 10;
832 context->committer.len = eol - i - 10;
833 } else if (!prefixcmp(msg + i, "encoding ")) {
834 context->encoding.off = i + 9;
835 context->encoding.len = eol - i - 9;
837 i = eol;
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)
851 size_t trimlen;
852 size_t start_len = sb->len;
853 int space = 2;
855 for (; *msg && *msg != '\n'; msg++) {
856 if (istitlechar(*msg)) {
857 if (space == 1)
858 strbuf_addch(sb, '-');
859 space = 0;
860 strbuf_addch(sb, *msg);
861 if (*msg == '.')
862 while (*(msg+1) == '.')
863 msg++;
864 } else
865 space |= 1;
868 /* trim any trailing '.' or '-' characters */
869 trimlen = 0;
870 while (sb->len - trimlen > start_len &&
871 (sb->buf[sb->len - 1 - trimlen] == '.'
872 || sb->buf[sb->len - 1 - trimlen] == '-'))
873 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)
880 int first = 1;
882 for (;;) {
883 const char *line = msg;
884 int linelen = get_one_line(line);
886 msg += linelen;
887 if (!linelen || is_empty_line(line, &linelen))
888 break;
890 if (!sb)
891 continue;
892 strbuf_grow(sb, linelen + 2);
893 if (!first)
894 strbuf_addstr(sb, line_separator);
895 strbuf_add(sb, line, linelen);
896 first = 0;
898 return msg;
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 format_decoration(struct strbuf *sb, const struct commit *commit)
918 struct name_decoration *d;
919 const char *prefix = " (";
921 load_ref_decorations(DECORATE_SHORT_REFS);
922 d = lookup_decoration(&name_decoration, &commit->object);
923 while (d) {
924 strbuf_addstr(sb, prefix);
925 prefix = ", ";
926 strbuf_addstr(sb, d->name);
927 d = d->next;
929 if (prefix[0] == ',')
930 strbuf_addch(sb, ')');
933 static void strbuf_wrap(struct strbuf *sb, size_t pos,
934 size_t width, size_t indent1, size_t indent2)
936 struct strbuf tmp = STRBUF_INIT;
938 if (pos)
939 strbuf_add(&tmp, sb->buf, pos);
940 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
941 (int) indent1, (int) indent2, (int) width);
942 strbuf_swap(&tmp, sb);
943 strbuf_release(&tmp);
946 static void rewrap_message_tail(struct strbuf *sb,
947 struct format_commit_context *c,
948 size_t new_width, size_t new_indent1,
949 size_t new_indent2)
951 if (c->width == new_width && c->indent1 == new_indent1 &&
952 c->indent2 == new_indent2)
953 return;
954 if (c->wrap_start < sb->len)
955 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
956 c->wrap_start = sb->len;
957 c->width = new_width;
958 c->indent1 = new_indent1;
959 c->indent2 = new_indent2;
962 static struct {
963 char result;
964 const char *check;
965 } signature_check[] = {
966 { 'G', "\n[GNUPG:] GOODSIG " },
967 { 'B', "\n[GNUPG:] BADSIG " },
970 static void parse_signature_lines(struct format_commit_context *ctx)
972 const char *buf = ctx->signature.gpg_status;
973 int i;
975 for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
976 const char *found = strstr(buf, signature_check[i].check);
977 const char *next;
978 if (!found)
979 continue;
980 ctx->signature.good_bad = signature_check[i].result;
981 found += strlen(signature_check[i].check);
982 ctx->signature.key = xmemdupz(found, 16);
983 found += 17;
984 next = strchrnul(found, '\n');
985 ctx->signature.signer = xmemdupz(found, next - found);
986 break;
990 static void parse_commit_signature(struct format_commit_context *ctx)
992 struct strbuf payload = STRBUF_INIT;
993 struct strbuf signature = STRBUF_INIT;
994 struct strbuf gpg_output = STRBUF_INIT;
995 struct strbuf gpg_status = STRBUF_INIT;
996 int status;
998 ctx->commit_signature_parsed = 1;
1000 if (parse_signed_commit(ctx->commit->object.sha1,
1001 &payload, &signature) <= 0)
1002 goto out;
1003 status = verify_signed_buffer(payload.buf, payload.len,
1004 signature.buf, signature.len,
1005 &gpg_output, &gpg_status);
1006 if (status && !gpg_output.len)
1007 goto out;
1008 ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
1009 ctx->signature.gpg_status = strbuf_detach(&gpg_status, NULL);
1010 parse_signature_lines(ctx);
1012 out:
1013 strbuf_release(&gpg_status);
1014 strbuf_release(&gpg_output);
1015 strbuf_release(&payload);
1016 strbuf_release(&signature);
1020 static int format_reflog_person(struct strbuf *sb,
1021 char part,
1022 struct reflog_walk_info *log,
1023 enum date_mode dmode)
1025 const char *ident;
1027 if (!log)
1028 return 2;
1030 ident = get_reflog_ident(log);
1031 if (!ident)
1032 return 2;
1034 return format_person_part(sb, part, ident, strlen(ident), dmode);
1037 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
1038 void *context)
1040 struct format_commit_context *c = context;
1041 const struct commit *commit = c->commit;
1042 const char *msg = c->message;
1043 struct commit_list *p;
1044 int h1, h2;
1046 /* these are independent of the commit */
1047 switch (placeholder[0]) {
1048 case 'C':
1049 if (placeholder[1] == '(') {
1050 const char *begin = placeholder + 2;
1051 const char *end = strchr(begin, ')');
1052 char color[COLOR_MAXLEN];
1054 if (!end)
1055 return 0;
1056 if (!prefixcmp(begin, "auto,")) {
1057 if (!want_color(c->pretty_ctx->color))
1058 return end - placeholder + 1;
1059 begin += 5;
1061 color_parse_mem(begin,
1062 end - begin,
1063 "--pretty format", color);
1064 strbuf_addstr(sb, color);
1065 return end - placeholder + 1;
1067 if (!prefixcmp(placeholder + 1, "red")) {
1068 strbuf_addstr(sb, GIT_COLOR_RED);
1069 return 4;
1070 } else if (!prefixcmp(placeholder + 1, "green")) {
1071 strbuf_addstr(sb, GIT_COLOR_GREEN);
1072 return 6;
1073 } else if (!prefixcmp(placeholder + 1, "blue")) {
1074 strbuf_addstr(sb, GIT_COLOR_BLUE);
1075 return 5;
1076 } else if (!prefixcmp(placeholder + 1, "reset")) {
1077 strbuf_addstr(sb, GIT_COLOR_RESET);
1078 return 6;
1079 } else
1080 return 0;
1081 case 'n': /* newline */
1082 strbuf_addch(sb, '\n');
1083 return 1;
1084 case 'x':
1085 /* %x00 == NUL, %x0a == LF, etc. */
1086 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1087 h1 <= 16 &&
1088 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1089 h2 <= 16) {
1090 strbuf_addch(sb, (h1<<4)|h2);
1091 return 3;
1092 } else
1093 return 0;
1094 case 'w':
1095 if (placeholder[1] == '(') {
1096 unsigned long width = 0, indent1 = 0, indent2 = 0;
1097 char *next;
1098 const char *start = placeholder + 2;
1099 const char *end = strchr(start, ')');
1100 if (!end)
1101 return 0;
1102 if (end > start) {
1103 width = strtoul(start, &next, 10);
1104 if (*next == ',') {
1105 indent1 = strtoul(next + 1, &next, 10);
1106 if (*next == ',') {
1107 indent2 = strtoul(next + 1,
1108 &next, 10);
1111 if (*next != ')')
1112 return 0;
1114 rewrap_message_tail(sb, c, width, indent1, indent2);
1115 return end - placeholder + 1;
1116 } else
1117 return 0;
1120 /* these depend on the commit */
1121 if (!commit->object.parsed)
1122 parse_object(commit->object.sha1);
1124 switch (placeholder[0]) {
1125 case 'H': /* commit hash */
1126 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1127 return 1;
1128 case 'h': /* abbreviated commit hash */
1129 if (add_again(sb, &c->abbrev_commit_hash))
1130 return 1;
1131 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1132 c->pretty_ctx->abbrev));
1133 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1134 return 1;
1135 case 'T': /* tree hash */
1136 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1137 return 1;
1138 case 't': /* abbreviated tree hash */
1139 if (add_again(sb, &c->abbrev_tree_hash))
1140 return 1;
1141 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1142 c->pretty_ctx->abbrev));
1143 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1144 return 1;
1145 case 'P': /* parent hashes */
1146 for (p = commit->parents; p; p = p->next) {
1147 if (p != commit->parents)
1148 strbuf_addch(sb, ' ');
1149 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1151 return 1;
1152 case 'p': /* abbreviated parent hashes */
1153 if (add_again(sb, &c->abbrev_parent_hashes))
1154 return 1;
1155 for (p = commit->parents; p; p = p->next) {
1156 if (p != commit->parents)
1157 strbuf_addch(sb, ' ');
1158 strbuf_addstr(sb, find_unique_abbrev(
1159 p->item->object.sha1,
1160 c->pretty_ctx->abbrev));
1162 c->abbrev_parent_hashes.len = sb->len -
1163 c->abbrev_parent_hashes.off;
1164 return 1;
1165 case 'm': /* left/right/bottom */
1166 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1167 return 1;
1168 case 'd':
1169 format_decoration(sb, commit);
1170 return 1;
1171 case 'g': /* reflog info */
1172 switch(placeholder[1]) {
1173 case 'd': /* reflog selector */
1174 case 'D':
1175 if (c->pretty_ctx->reflog_info)
1176 get_reflog_selector(sb,
1177 c->pretty_ctx->reflog_info,
1178 c->pretty_ctx->date_mode,
1179 c->pretty_ctx->date_mode_explicit,
1180 (placeholder[1] == 'd'));
1181 return 2;
1182 case 's': /* reflog message */
1183 if (c->pretty_ctx->reflog_info)
1184 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1185 return 2;
1186 case 'n':
1187 case 'N':
1188 case 'e':
1189 case 'E':
1190 return format_reflog_person(sb,
1191 placeholder[1],
1192 c->pretty_ctx->reflog_info,
1193 c->pretty_ctx->date_mode);
1195 return 0; /* unknown %g placeholder */
1196 case 'N':
1197 if (c->pretty_ctx->notes_message) {
1198 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1199 return 1;
1201 return 0;
1204 if (placeholder[0] == 'G') {
1205 if (!c->commit_signature_parsed)
1206 parse_commit_signature(c);
1207 switch (placeholder[1]) {
1208 case 'G':
1209 if (c->signature.gpg_output)
1210 strbuf_addstr(sb, c->signature.gpg_output);
1211 break;
1212 case '?':
1213 switch (c->signature.good_bad) {
1214 case 'G':
1215 case 'B':
1216 strbuf_addch(sb, c->signature.good_bad);
1218 break;
1219 case 'S':
1220 if (c->signature.signer)
1221 strbuf_addstr(sb, c->signature.signer);
1222 break;
1223 case 'K':
1224 if (c->signature.key)
1225 strbuf_addstr(sb, c->signature.key);
1226 break;
1228 return 2;
1232 /* For the rest we have to parse the commit header. */
1233 if (!c->commit_header_parsed)
1234 parse_commit_header(c);
1236 switch (placeholder[0]) {
1237 case 'a': /* author ... */
1238 return format_person_part(sb, placeholder[1],
1239 msg + c->author.off, c->author.len,
1240 c->pretty_ctx->date_mode);
1241 case 'c': /* committer ... */
1242 return format_person_part(sb, placeholder[1],
1243 msg + c->committer.off, c->committer.len,
1244 c->pretty_ctx->date_mode);
1245 case 'e': /* encoding */
1246 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1247 return 1;
1248 case 'B': /* raw body */
1249 /* message_off is always left at the initial newline */
1250 strbuf_addstr(sb, msg + c->message_off + 1);
1251 return 1;
1254 /* Now we need to parse the commit message. */
1255 if (!c->commit_message_parsed)
1256 parse_commit_message(c);
1258 switch (placeholder[0]) {
1259 case 's': /* subject */
1260 format_subject(sb, msg + c->subject_off, " ");
1261 return 1;
1262 case 'f': /* sanitized subject */
1263 format_sanitized_subject(sb, msg + c->subject_off);
1264 return 1;
1265 case 'b': /* body */
1266 strbuf_addstr(sb, msg + c->body_off);
1267 return 1;
1269 return 0; /* unknown placeholder */
1272 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1273 void *context)
1275 int consumed;
1276 size_t orig_len;
1277 enum {
1278 NO_MAGIC,
1279 ADD_LF_BEFORE_NON_EMPTY,
1280 DEL_LF_BEFORE_EMPTY,
1281 ADD_SP_BEFORE_NON_EMPTY
1282 } magic = NO_MAGIC;
1284 switch (placeholder[0]) {
1285 case '-':
1286 magic = DEL_LF_BEFORE_EMPTY;
1287 break;
1288 case '+':
1289 magic = ADD_LF_BEFORE_NON_EMPTY;
1290 break;
1291 case ' ':
1292 magic = ADD_SP_BEFORE_NON_EMPTY;
1293 break;
1294 default:
1295 break;
1297 if (magic != NO_MAGIC)
1298 placeholder++;
1300 orig_len = sb->len;
1301 consumed = format_commit_one(sb, placeholder, context);
1302 if (magic == NO_MAGIC)
1303 return consumed;
1305 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1306 while (sb->len && sb->buf[sb->len - 1] == '\n')
1307 strbuf_setlen(sb, sb->len - 1);
1308 } else if (orig_len != sb->len) {
1309 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1310 strbuf_insert(sb, orig_len, "\n", 1);
1311 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1312 strbuf_insert(sb, orig_len, " ", 1);
1314 return consumed + 1;
1317 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1318 void *context)
1320 struct userformat_want *w = context;
1322 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1323 placeholder++;
1325 switch (*placeholder) {
1326 case 'N':
1327 w->notes = 1;
1328 break;
1330 return 0;
1333 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1335 struct strbuf dummy = STRBUF_INIT;
1337 if (!fmt) {
1338 if (!user_format)
1339 return;
1340 fmt = user_format;
1342 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1343 strbuf_release(&dummy);
1346 void format_commit_message(const struct commit *commit,
1347 const char *format, struct strbuf *sb,
1348 const struct pretty_print_context *pretty_ctx)
1350 struct format_commit_context context;
1351 const char *output_enc = pretty_ctx->output_encoding;
1353 memset(&context, 0, sizeof(context));
1354 context.commit = commit;
1355 context.pretty_ctx = pretty_ctx;
1356 context.wrap_start = sb->len;
1357 context.message = logmsg_reencode(commit, output_enc);
1359 strbuf_expand(sb, format, format_commit_item, &context);
1360 rewrap_message_tail(sb, &context, 0, 0, 0);
1362 logmsg_free(context.message, commit);
1363 free(context.signature.gpg_output);
1364 free(context.signature.signer);
1367 static void pp_header(const struct pretty_print_context *pp,
1368 const char *encoding,
1369 const struct commit *commit,
1370 const char **msg_p,
1371 struct strbuf *sb)
1373 int parents_shown = 0;
1375 for (;;) {
1376 const char *line = *msg_p;
1377 int linelen = get_one_line(*msg_p);
1379 if (!linelen)
1380 return;
1381 *msg_p += linelen;
1383 if (linelen == 1)
1384 /* End of header */
1385 return;
1387 if (pp->fmt == CMIT_FMT_RAW) {
1388 strbuf_add(sb, line, linelen);
1389 continue;
1392 if (!prefixcmp(line, "parent ")) {
1393 if (linelen != 48)
1394 die("bad parent line in commit");
1395 continue;
1398 if (!parents_shown) {
1399 struct commit_list *parent;
1400 int num;
1401 for (parent = commit->parents, num = 0;
1402 parent;
1403 parent = parent->next, num++)
1405 /* with enough slop */
1406 strbuf_grow(sb, num * 50 + 20);
1407 add_merge_info(pp, sb, commit);
1408 parents_shown = 1;
1412 * MEDIUM == DEFAULT shows only author with dates.
1413 * FULL shows both authors but not dates.
1414 * FULLER shows both authors and dates.
1416 if (!prefixcmp(line, "author ")) {
1417 strbuf_grow(sb, linelen + 80);
1418 pp_user_info(pp, "Author", sb, line + 7, encoding);
1420 if (!prefixcmp(line, "committer ") &&
1421 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1422 strbuf_grow(sb, linelen + 80);
1423 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1428 void pp_title_line(const struct pretty_print_context *pp,
1429 const char **msg_p,
1430 struct strbuf *sb,
1431 const char *encoding,
1432 int need_8bit_cte)
1434 static const int max_length = 78; /* per rfc2047 */
1435 struct strbuf title;
1437 strbuf_init(&title, 80);
1438 *msg_p = format_subject(&title, *msg_p,
1439 pp->preserve_subject ? "\n" : " ");
1441 strbuf_grow(sb, title.len + 1024);
1442 if (pp->subject) {
1443 strbuf_addstr(sb, pp->subject);
1444 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1445 add_rfc2047(sb, title.buf, title.len,
1446 encoding, RFC2047_SUBJECT);
1447 else
1448 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1449 -last_line_length(sb), 1, max_length);
1450 } else {
1451 strbuf_addbuf(sb, &title);
1453 strbuf_addch(sb, '\n');
1455 if (need_8bit_cte > 0) {
1456 const char *header_fmt =
1457 "MIME-Version: 1.0\n"
1458 "Content-Type: text/plain; charset=%s\n"
1459 "Content-Transfer-Encoding: 8bit\n";
1460 strbuf_addf(sb, header_fmt, encoding);
1462 if (pp->after_subject) {
1463 strbuf_addstr(sb, pp->after_subject);
1465 if (pp->fmt == CMIT_FMT_EMAIL) {
1466 strbuf_addch(sb, '\n');
1468 strbuf_release(&title);
1471 void pp_remainder(const struct pretty_print_context *pp,
1472 const char **msg_p,
1473 struct strbuf *sb,
1474 int indent)
1476 int first = 1;
1477 for (;;) {
1478 const char *line = *msg_p;
1479 int linelen = get_one_line(line);
1480 *msg_p += linelen;
1482 if (!linelen)
1483 break;
1485 if (is_empty_line(line, &linelen)) {
1486 if (first)
1487 continue;
1488 if (pp->fmt == CMIT_FMT_SHORT)
1489 break;
1491 first = 0;
1493 strbuf_grow(sb, linelen + indent + 20);
1494 if (indent) {
1495 memset(sb->buf + sb->len, ' ', indent);
1496 strbuf_setlen(sb, sb->len + indent);
1498 strbuf_add(sb, line, linelen);
1499 strbuf_addch(sb, '\n');
1503 void pretty_print_commit(const struct pretty_print_context *pp,
1504 const struct commit *commit,
1505 struct strbuf *sb)
1507 unsigned long beginning_of_body;
1508 int indent = 4;
1509 const char *msg;
1510 char *reencoded;
1511 const char *encoding;
1512 int need_8bit_cte = pp->need_8bit_cte;
1514 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1515 format_commit_message(commit, user_format, sb, pp);
1516 return;
1519 encoding = get_log_output_encoding();
1520 msg = reencoded = logmsg_reencode(commit, encoding);
1522 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1523 indent = 0;
1526 * We need to check and emit Content-type: to mark it
1527 * as 8-bit if we haven't done so.
1529 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1530 int i, ch, in_body;
1532 for (in_body = i = 0; (ch = msg[i]); i++) {
1533 if (!in_body) {
1534 /* author could be non 7-bit ASCII but
1535 * the log may be so; skip over the
1536 * header part first.
1538 if (ch == '\n' && msg[i+1] == '\n')
1539 in_body = 1;
1541 else if (non_ascii(ch)) {
1542 need_8bit_cte = 1;
1543 break;
1548 pp_header(pp, encoding, commit, &msg, sb);
1549 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1550 strbuf_addch(sb, '\n');
1553 /* Skip excess blank lines at the beginning of body, if any... */
1554 msg = skip_empty_lines(msg);
1556 /* These formats treat the title line specially. */
1557 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1558 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1560 beginning_of_body = sb->len;
1561 if (pp->fmt != CMIT_FMT_ONELINE)
1562 pp_remainder(pp, &msg, sb, indent);
1563 strbuf_rtrim(sb);
1565 /* Make sure there is an EOLN for the non-oneline case */
1566 if (pp->fmt != CMIT_FMT_ONELINE)
1567 strbuf_addch(sb, '\n');
1570 * The caller may append additional body text in e-mail
1571 * format. Make sure we did not strip the blank line
1572 * between the header and the body.
1574 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1575 strbuf_addch(sb, '\n');
1577 logmsg_free(reencoded, commit);
1580 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1581 struct strbuf *sb)
1583 struct pretty_print_context pp = {0};
1584 pp.fmt = fmt;
1585 pretty_print_commit(&pp, commit, sb);