Merge branch 'master' of git://git.kernel.org/pub/scm/git/git
[git/jnareb-git.git] / pretty.c
blobd3a82d22d398ae7234f8917f4d8b0770ba4d0c47
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 void pp_user_info(const struct pretty_print_context *pp,
397 const char *what, struct strbuf *sb,
398 const char *line, const char *encoding)
400 struct strbuf name;
401 struct strbuf mail;
402 struct ident_split ident;
403 int linelen;
404 char *line_end, *date;
405 const char *mailbuf, *namebuf;
406 size_t namelen, maillen;
407 int max_length = 78; /* per rfc2822 */
408 unsigned long time;
409 int tz;
411 if (pp->fmt == CMIT_FMT_ONELINE)
412 return;
414 line_end = strchr(line, '\n');
415 if (!line_end) {
416 line_end = strchr(line, '\0');
417 if (!line_end)
418 return;
421 linelen = ++line_end - line;
422 if (split_ident_line(&ident, line, linelen))
423 return;
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;
431 if (pp->mailmap)
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(&quoted, name.buf, name.len);
453 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
454 -6, 1, max_length);
455 strbuf_release(&quoted);
456 } else {
457 strbuf_add_wrapped_bytes(sb, name.buf, name.len,
458 -6, 1, max_length);
460 if (namelen - name.len + last_line_length(sb) > max_length)
461 strbuf_addch(sb, '\n');
463 strbuf_addf(sb, " <%s>\n", mail.buf);
464 } else {
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);
473 switch (pp->fmt) {
474 case CMIT_FMT_MEDIUM:
475 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, pp->date_mode));
476 break;
477 case CMIT_FMT_EMAIL:
478 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
479 break;
480 case CMIT_FMT_FULLER:
481 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
482 break;
483 default:
484 /* notin' */
485 break;
489 static int is_empty_line(const char *line, int *len_p)
491 int len = *len_p;
492 while (len && isspace(line[len-1]))
493 len--;
494 *len_p = len;
495 return !len;
498 static const char *skip_empty_lines(const char *msg)
500 for (;;) {
501 int linelen = get_one_line(msg);
502 int ll = linelen;
503 if (!linelen)
504 break;
505 if (!is_empty_line(msg, &ll))
506 break;
507 msg += linelen;
509 return msg;
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)
519 return;
521 strbuf_addstr(sb, "Merge:");
523 while (parent) {
524 struct commit *p = parent->item;
525 const char *hex = NULL;
526 if (pp->abbrev)
527 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
528 if (!hex)
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,
538 const char *key)
540 int key_len = strlen(key);
541 const char *line = msg;
543 while (line) {
544 const char *eol = strchr(line, '\n'), *next;
546 if (line == eol)
547 return NULL;
548 if (!eol) {
549 warning("malformed commit (header is missing newline): %s",
550 sha1_to_hex(commit->object.sha1));
551 eol = line + strlen(line);
552 next = NULL;
553 } else
554 next = eol + 1;
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);
560 line = next;
562 return NULL;
565 static char *replace_encoding_header(char *buf, const char *encoding)
567 struct strbuf tmp = STRBUF_INIT;
568 size_t start, len;
569 char *cp = buf;
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')
575 return buf;
577 start = cp - buf;
578 cp = strchr(cp, '\n');
579 if (!cp)
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);
587 } else {
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 const char *output_encoding)
599 static const char *utf8 = "UTF-8";
600 const char *use_encoding;
601 char *encoding;
602 char *msg = commit->buffer;
603 char *out;
605 if (!msg) {
606 enum object_type type;
607 unsigned long size;
609 msg = read_sha1_file(commit->object.sha1, &type, &size);
610 if (!msg)
611 die("Cannot read commit object %s",
612 sha1_to_hex(commit->object.sha1));
613 if (type != OBJ_COMMIT)
614 die("Expected commit for '%s', got %s",
615 sha1_to_hex(commit->object.sha1), typename(type));
618 if (!output_encoding || !*output_encoding)
619 return msg;
620 encoding = get_header(commit, msg, "encoding");
621 use_encoding = encoding ? encoding : utf8;
622 if (same_encoding(use_encoding, output_encoding)) {
624 * No encoding work to be done. If we have no encoding header
625 * at all, then there's nothing to do, and we can return the
626 * message verbatim (whether newly allocated or not).
628 if (!encoding)
629 return msg;
632 * Otherwise, we still want to munge the encoding header in the
633 * result, which will be done by modifying the buffer. If we
634 * are using a fresh copy, we can reuse it. But if we are using
635 * the cached copy from commit->buffer, we need to duplicate it
636 * to avoid munging commit->buffer.
638 out = msg;
639 if (out == commit->buffer)
640 out = xstrdup(out);
642 else {
644 * There's actual encoding work to do. Do the reencoding, which
645 * still leaves the header to be replaced in the next step. At
646 * this point, we are done with msg. If we allocated a fresh
647 * copy, we can free it.
649 out = reencode_string(msg, output_encoding, use_encoding);
650 if (out && msg != commit->buffer)
651 free(msg);
655 * This replacement actually consumes the buffer we hand it, so we do
656 * not have to worry about freeing the old "out" here.
658 if (out)
659 out = replace_encoding_header(out, output_encoding);
661 free(encoding);
663 * If the re-encoding failed, out might be NULL here; in that
664 * case we just return the commit message verbatim.
666 return out ? out : msg;
669 void logmsg_free(char *msg, const struct commit *commit)
671 if (msg != commit->buffer)
672 free(msg);
675 static int mailmap_name(const char **email, size_t *email_len,
676 const char **name, size_t *name_len)
678 static struct string_list *mail_map;
679 if (!mail_map) {
680 mail_map = xcalloc(1, sizeof(*mail_map));
681 read_mailmap(mail_map, NULL);
683 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
686 static size_t format_person_part(struct strbuf *sb, char part,
687 const char *msg, int len, enum date_mode dmode)
689 /* currently all placeholders have same length */
690 const int placeholder_len = 2;
691 int tz;
692 unsigned long date = 0;
693 struct ident_split s;
694 const char *name, *mail;
695 size_t maillen, namelen;
697 if (split_ident_line(&s, msg, len) < 0)
698 goto skip;
700 name = s.name_begin;
701 namelen = s.name_end - s.name_begin;
702 mail = s.mail_begin;
703 maillen = s.mail_end - s.mail_begin;
705 if (part == 'N' || part == 'E') /* mailmap lookup */
706 mailmap_name(&mail, &maillen, &name, &namelen);
707 if (part == 'n' || part == 'N') { /* name */
708 strbuf_add(sb, name, namelen);
709 return placeholder_len;
711 if (part == 'e' || part == 'E') { /* email */
712 strbuf_add(sb, mail, maillen);
713 return placeholder_len;
716 if (!s.date_begin)
717 goto skip;
719 date = strtoul(s.date_begin, NULL, 10);
721 if (part == 't') { /* date, UNIX timestamp */
722 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
723 return placeholder_len;
726 /* parse tz */
727 tz = strtoul(s.tz_begin + 1, NULL, 10);
728 if (*s.tz_begin == '-')
729 tz = -tz;
731 switch (part) {
732 case 'd': /* date */
733 strbuf_addstr(sb, show_date(date, tz, dmode));
734 return placeholder_len;
735 case 'D': /* date, RFC2822 style */
736 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
737 return placeholder_len;
738 case 'r': /* date, relative */
739 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
740 return placeholder_len;
741 case 'i': /* date, ISO 8601 */
742 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
743 return placeholder_len;
746 skip:
748 * reading from either a bogus commit, or a reflog entry with
749 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
750 * to compute a valid return value.
752 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
753 || part == 'D' || part == 'r' || part == 'i')
754 return placeholder_len;
756 return 0; /* unknown placeholder */
759 struct chunk {
760 size_t off;
761 size_t len;
764 struct format_commit_context {
765 const struct commit *commit;
766 const struct pretty_print_context *pretty_ctx;
767 unsigned commit_header_parsed:1;
768 unsigned commit_message_parsed:1;
769 struct signature_check signature_check;
770 char *message;
771 size_t width, indent1, indent2;
773 /* These offsets are relative to the start of the commit message. */
774 struct chunk author;
775 struct chunk committer;
776 struct chunk encoding;
777 size_t message_off;
778 size_t subject_off;
779 size_t body_off;
781 /* The following ones are relative to the result struct strbuf. */
782 struct chunk abbrev_commit_hash;
783 struct chunk abbrev_tree_hash;
784 struct chunk abbrev_parent_hashes;
785 size_t wrap_start;
788 static int add_again(struct strbuf *sb, struct chunk *chunk)
790 if (chunk->len) {
791 strbuf_adddup(sb, chunk->off, chunk->len);
792 return 1;
796 * We haven't seen this chunk before. Our caller is surely
797 * going to add it the hard way now. Remember the most likely
798 * start of the to-be-added chunk: the current end of the
799 * struct strbuf.
801 chunk->off = sb->len;
802 return 0;
805 static void parse_commit_header(struct format_commit_context *context)
807 const char *msg = context->message;
808 int i;
810 for (i = 0; msg[i]; i++) {
811 int eol;
812 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
813 ; /* do nothing */
815 if (i == eol) {
816 break;
817 } else if (!prefixcmp(msg + i, "author ")) {
818 context->author.off = i + 7;
819 context->author.len = eol - i - 7;
820 } else if (!prefixcmp(msg + i, "committer ")) {
821 context->committer.off = i + 10;
822 context->committer.len = eol - i - 10;
823 } else if (!prefixcmp(msg + i, "encoding ")) {
824 context->encoding.off = i + 9;
825 context->encoding.len = eol - i - 9;
827 i = eol;
829 context->message_off = i;
830 context->commit_header_parsed = 1;
833 static int istitlechar(char c)
835 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
836 (c >= '0' && c <= '9') || c == '.' || c == '_';
839 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
841 size_t trimlen;
842 size_t start_len = sb->len;
843 int space = 2;
845 for (; *msg && *msg != '\n'; msg++) {
846 if (istitlechar(*msg)) {
847 if (space == 1)
848 strbuf_addch(sb, '-');
849 space = 0;
850 strbuf_addch(sb, *msg);
851 if (*msg == '.')
852 while (*(msg+1) == '.')
853 msg++;
854 } else
855 space |= 1;
858 /* trim any trailing '.' or '-' characters */
859 trimlen = 0;
860 while (sb->len - trimlen > start_len &&
861 (sb->buf[sb->len - 1 - trimlen] == '.'
862 || sb->buf[sb->len - 1 - trimlen] == '-'))
863 trimlen++;
864 strbuf_remove(sb, sb->len - trimlen, trimlen);
867 const char *format_subject(struct strbuf *sb, const char *msg,
868 const char *line_separator)
870 int first = 1;
872 for (;;) {
873 const char *line = msg;
874 int linelen = get_one_line(line);
876 msg += linelen;
877 if (!linelen || is_empty_line(line, &linelen))
878 break;
880 if (!sb)
881 continue;
882 strbuf_grow(sb, linelen + 2);
883 if (!first)
884 strbuf_addstr(sb, line_separator);
885 strbuf_add(sb, line, linelen);
886 first = 0;
888 return msg;
891 static void parse_commit_message(struct format_commit_context *c)
893 const char *msg = c->message + c->message_off;
894 const char *start = c->message;
896 msg = skip_empty_lines(msg);
897 c->subject_off = msg - start;
899 msg = format_subject(NULL, msg, NULL);
900 msg = skip_empty_lines(msg);
901 c->body_off = msg - start;
903 c->commit_message_parsed = 1;
906 static void format_decoration(struct strbuf *sb, const struct commit *commit)
908 struct name_decoration *d;
909 const char *prefix = " (";
911 load_ref_decorations(DECORATE_SHORT_REFS);
912 d = lookup_decoration(&name_decoration, &commit->object);
913 while (d) {
914 strbuf_addstr(sb, prefix);
915 prefix = ", ";
916 strbuf_addstr(sb, d->name);
917 d = d->next;
919 if (prefix[0] == ',')
920 strbuf_addch(sb, ')');
923 static void strbuf_wrap(struct strbuf *sb, size_t pos,
924 size_t width, size_t indent1, size_t indent2)
926 struct strbuf tmp = STRBUF_INIT;
928 if (pos)
929 strbuf_add(&tmp, sb->buf, pos);
930 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
931 (int) indent1, (int) indent2, (int) width);
932 strbuf_swap(&tmp, sb);
933 strbuf_release(&tmp);
936 static void rewrap_message_tail(struct strbuf *sb,
937 struct format_commit_context *c,
938 size_t new_width, size_t new_indent1,
939 size_t new_indent2)
941 if (c->width == new_width && c->indent1 == new_indent1 &&
942 c->indent2 == new_indent2)
943 return;
944 if (c->wrap_start < sb->len)
945 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
946 c->wrap_start = sb->len;
947 c->width = new_width;
948 c->indent1 = new_indent1;
949 c->indent2 = new_indent2;
952 static int format_reflog_person(struct strbuf *sb,
953 char part,
954 struct reflog_walk_info *log,
955 enum date_mode dmode)
957 const char *ident;
959 if (!log)
960 return 2;
962 ident = get_reflog_ident(log);
963 if (!ident)
964 return 2;
966 return format_person_part(sb, part, ident, strlen(ident), dmode);
969 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
970 void *context)
972 struct format_commit_context *c = context;
973 const struct commit *commit = c->commit;
974 const char *msg = c->message;
975 struct commit_list *p;
976 int h1, h2;
978 /* these are independent of the commit */
979 switch (placeholder[0]) {
980 case 'C':
981 if (placeholder[1] == '(') {
982 const char *begin = placeholder + 2;
983 const char *end = strchr(begin, ')');
984 char color[COLOR_MAXLEN];
986 if (!end)
987 return 0;
988 if (!prefixcmp(begin, "auto,")) {
989 if (!want_color(c->pretty_ctx->color))
990 return end - placeholder + 1;
991 begin += 5;
993 color_parse_mem(begin,
994 end - begin,
995 "--pretty format", color);
996 strbuf_addstr(sb, color);
997 return end - placeholder + 1;
999 if (!prefixcmp(placeholder + 1, "red")) {
1000 strbuf_addstr(sb, GIT_COLOR_RED);
1001 return 4;
1002 } else if (!prefixcmp(placeholder + 1, "green")) {
1003 strbuf_addstr(sb, GIT_COLOR_GREEN);
1004 return 6;
1005 } else if (!prefixcmp(placeholder + 1, "blue")) {
1006 strbuf_addstr(sb, GIT_COLOR_BLUE);
1007 return 5;
1008 } else if (!prefixcmp(placeholder + 1, "reset")) {
1009 strbuf_addstr(sb, GIT_COLOR_RESET);
1010 return 6;
1011 } else
1012 return 0;
1013 case 'n': /* newline */
1014 strbuf_addch(sb, '\n');
1015 return 1;
1016 case 'x':
1017 /* %x00 == NUL, %x0a == LF, etc. */
1018 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1019 h1 <= 16 &&
1020 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1021 h2 <= 16) {
1022 strbuf_addch(sb, (h1<<4)|h2);
1023 return 3;
1024 } else
1025 return 0;
1026 case 'w':
1027 if (placeholder[1] == '(') {
1028 unsigned long width = 0, indent1 = 0, indent2 = 0;
1029 char *next;
1030 const char *start = placeholder + 2;
1031 const char *end = strchr(start, ')');
1032 if (!end)
1033 return 0;
1034 if (end > start) {
1035 width = strtoul(start, &next, 10);
1036 if (*next == ',') {
1037 indent1 = strtoul(next + 1, &next, 10);
1038 if (*next == ',') {
1039 indent2 = strtoul(next + 1,
1040 &next, 10);
1043 if (*next != ')')
1044 return 0;
1046 rewrap_message_tail(sb, c, width, indent1, indent2);
1047 return end - placeholder + 1;
1048 } else
1049 return 0;
1052 /* these depend on the commit */
1053 if (!commit->object.parsed)
1054 parse_object(commit->object.sha1);
1056 switch (placeholder[0]) {
1057 case 'H': /* commit hash */
1058 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1059 return 1;
1060 case 'h': /* abbreviated commit hash */
1061 if (add_again(sb, &c->abbrev_commit_hash))
1062 return 1;
1063 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1064 c->pretty_ctx->abbrev));
1065 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1066 return 1;
1067 case 'T': /* tree hash */
1068 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1069 return 1;
1070 case 't': /* abbreviated tree hash */
1071 if (add_again(sb, &c->abbrev_tree_hash))
1072 return 1;
1073 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1074 c->pretty_ctx->abbrev));
1075 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1076 return 1;
1077 case 'P': /* parent hashes */
1078 for (p = commit->parents; p; p = p->next) {
1079 if (p != commit->parents)
1080 strbuf_addch(sb, ' ');
1081 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1083 return 1;
1084 case 'p': /* abbreviated parent hashes */
1085 if (add_again(sb, &c->abbrev_parent_hashes))
1086 return 1;
1087 for (p = commit->parents; p; p = p->next) {
1088 if (p != commit->parents)
1089 strbuf_addch(sb, ' ');
1090 strbuf_addstr(sb, find_unique_abbrev(
1091 p->item->object.sha1,
1092 c->pretty_ctx->abbrev));
1094 c->abbrev_parent_hashes.len = sb->len -
1095 c->abbrev_parent_hashes.off;
1096 return 1;
1097 case 'm': /* left/right/bottom */
1098 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1099 return 1;
1100 case 'd':
1101 format_decoration(sb, commit);
1102 return 1;
1103 case 'g': /* reflog info */
1104 switch(placeholder[1]) {
1105 case 'd': /* reflog selector */
1106 case 'D':
1107 if (c->pretty_ctx->reflog_info)
1108 get_reflog_selector(sb,
1109 c->pretty_ctx->reflog_info,
1110 c->pretty_ctx->date_mode,
1111 c->pretty_ctx->date_mode_explicit,
1112 (placeholder[1] == 'd'));
1113 return 2;
1114 case 's': /* reflog message */
1115 if (c->pretty_ctx->reflog_info)
1116 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1117 return 2;
1118 case 'n':
1119 case 'N':
1120 case 'e':
1121 case 'E':
1122 return format_reflog_person(sb,
1123 placeholder[1],
1124 c->pretty_ctx->reflog_info,
1125 c->pretty_ctx->date_mode);
1127 return 0; /* unknown %g placeholder */
1128 case 'N':
1129 if (c->pretty_ctx->notes_message) {
1130 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1131 return 1;
1133 return 0;
1136 if (placeholder[0] == 'G') {
1137 if (!c->signature_check.result)
1138 check_commit_signature(c->commit, &(c->signature_check));
1139 switch (placeholder[1]) {
1140 case 'G':
1141 if (c->signature_check.gpg_output)
1142 strbuf_addstr(sb, c->signature_check.gpg_output);
1143 break;
1144 case '?':
1145 switch (c->signature_check.result) {
1146 case 'G':
1147 case 'B':
1148 case 'U':
1149 case 'N':
1150 strbuf_addch(sb, c->signature_check.result);
1152 break;
1153 case 'S':
1154 if (c->signature_check.signer)
1155 strbuf_addstr(sb, c->signature_check.signer);
1156 break;
1157 case 'K':
1158 if (c->signature_check.key)
1159 strbuf_addstr(sb, c->signature_check.key);
1160 break;
1162 return 2;
1166 /* For the rest we have to parse the commit header. */
1167 if (!c->commit_header_parsed)
1168 parse_commit_header(c);
1170 switch (placeholder[0]) {
1171 case 'a': /* author ... */
1172 return format_person_part(sb, placeholder[1],
1173 msg + c->author.off, c->author.len,
1174 c->pretty_ctx->date_mode);
1175 case 'c': /* committer ... */
1176 return format_person_part(sb, placeholder[1],
1177 msg + c->committer.off, c->committer.len,
1178 c->pretty_ctx->date_mode);
1179 case 'e': /* encoding */
1180 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1181 return 1;
1182 case 'B': /* raw body */
1183 /* message_off is always left at the initial newline */
1184 strbuf_addstr(sb, msg + c->message_off + 1);
1185 return 1;
1188 /* Now we need to parse the commit message. */
1189 if (!c->commit_message_parsed)
1190 parse_commit_message(c);
1192 switch (placeholder[0]) {
1193 case 's': /* subject */
1194 format_subject(sb, msg + c->subject_off, " ");
1195 return 1;
1196 case 'f': /* sanitized subject */
1197 format_sanitized_subject(sb, msg + c->subject_off);
1198 return 1;
1199 case 'b': /* body */
1200 strbuf_addstr(sb, msg + c->body_off);
1201 return 1;
1203 return 0; /* unknown placeholder */
1206 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1207 void *context)
1209 int consumed;
1210 size_t orig_len;
1211 enum {
1212 NO_MAGIC,
1213 ADD_LF_BEFORE_NON_EMPTY,
1214 DEL_LF_BEFORE_EMPTY,
1215 ADD_SP_BEFORE_NON_EMPTY
1216 } magic = NO_MAGIC;
1218 switch (placeholder[0]) {
1219 case '-':
1220 magic = DEL_LF_BEFORE_EMPTY;
1221 break;
1222 case '+':
1223 magic = ADD_LF_BEFORE_NON_EMPTY;
1224 break;
1225 case ' ':
1226 magic = ADD_SP_BEFORE_NON_EMPTY;
1227 break;
1228 default:
1229 break;
1231 if (magic != NO_MAGIC)
1232 placeholder++;
1234 orig_len = sb->len;
1235 consumed = format_commit_one(sb, placeholder, context);
1236 if (magic == NO_MAGIC)
1237 return consumed;
1239 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1240 while (sb->len && sb->buf[sb->len - 1] == '\n')
1241 strbuf_setlen(sb, sb->len - 1);
1242 } else if (orig_len != sb->len) {
1243 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1244 strbuf_insert(sb, orig_len, "\n", 1);
1245 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1246 strbuf_insert(sb, orig_len, " ", 1);
1248 return consumed + 1;
1251 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1252 void *context)
1254 struct userformat_want *w = context;
1256 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1257 placeholder++;
1259 switch (*placeholder) {
1260 case 'N':
1261 w->notes = 1;
1262 break;
1264 return 0;
1267 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1269 struct strbuf dummy = STRBUF_INIT;
1271 if (!fmt) {
1272 if (!user_format)
1273 return;
1274 fmt = user_format;
1276 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1277 strbuf_release(&dummy);
1280 void format_commit_message(const struct commit *commit,
1281 const char *format, struct strbuf *sb,
1282 const struct pretty_print_context *pretty_ctx)
1284 struct format_commit_context context;
1285 const char *output_enc = pretty_ctx->output_encoding;
1287 memset(&context, 0, sizeof(context));
1288 context.commit = commit;
1289 context.pretty_ctx = pretty_ctx;
1290 context.wrap_start = sb->len;
1291 context.message = logmsg_reencode(commit, output_enc);
1293 strbuf_expand(sb, format, format_commit_item, &context);
1294 rewrap_message_tail(sb, &context, 0, 0, 0);
1296 logmsg_free(context.message, commit);
1297 free(context.signature_check.gpg_output);
1298 free(context.signature_check.signer);
1301 static void pp_header(const struct pretty_print_context *pp,
1302 const char *encoding,
1303 const struct commit *commit,
1304 const char **msg_p,
1305 struct strbuf *sb)
1307 int parents_shown = 0;
1309 for (;;) {
1310 const char *line = *msg_p;
1311 int linelen = get_one_line(*msg_p);
1313 if (!linelen)
1314 return;
1315 *msg_p += linelen;
1317 if (linelen == 1)
1318 /* End of header */
1319 return;
1321 if (pp->fmt == CMIT_FMT_RAW) {
1322 strbuf_add(sb, line, linelen);
1323 continue;
1326 if (!prefixcmp(line, "parent ")) {
1327 if (linelen != 48)
1328 die("bad parent line in commit");
1329 continue;
1332 if (!parents_shown) {
1333 struct commit_list *parent;
1334 int num;
1335 for (parent = commit->parents, num = 0;
1336 parent;
1337 parent = parent->next, num++)
1339 /* with enough slop */
1340 strbuf_grow(sb, num * 50 + 20);
1341 add_merge_info(pp, sb, commit);
1342 parents_shown = 1;
1346 * MEDIUM == DEFAULT shows only author with dates.
1347 * FULL shows both authors but not dates.
1348 * FULLER shows both authors and dates.
1350 if (!prefixcmp(line, "author ")) {
1351 strbuf_grow(sb, linelen + 80);
1352 pp_user_info(pp, "Author", sb, line + 7, encoding);
1354 if (!prefixcmp(line, "committer ") &&
1355 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1356 strbuf_grow(sb, linelen + 80);
1357 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1362 void pp_title_line(const struct pretty_print_context *pp,
1363 const char **msg_p,
1364 struct strbuf *sb,
1365 const char *encoding,
1366 int need_8bit_cte)
1368 static const int max_length = 78; /* per rfc2047 */
1369 struct strbuf title;
1371 strbuf_init(&title, 80);
1372 *msg_p = format_subject(&title, *msg_p,
1373 pp->preserve_subject ? "\n" : " ");
1375 strbuf_grow(sb, title.len + 1024);
1376 if (pp->subject) {
1377 strbuf_addstr(sb, pp->subject);
1378 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1379 add_rfc2047(sb, title.buf, title.len,
1380 encoding, RFC2047_SUBJECT);
1381 else
1382 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1383 -last_line_length(sb), 1, max_length);
1384 } else {
1385 strbuf_addbuf(sb, &title);
1387 strbuf_addch(sb, '\n');
1389 if (need_8bit_cte > 0) {
1390 const char *header_fmt =
1391 "MIME-Version: 1.0\n"
1392 "Content-Type: text/plain; charset=%s\n"
1393 "Content-Transfer-Encoding: 8bit\n";
1394 strbuf_addf(sb, header_fmt, encoding);
1396 if (pp->after_subject) {
1397 strbuf_addstr(sb, pp->after_subject);
1399 if (pp->fmt == CMIT_FMT_EMAIL) {
1400 strbuf_addch(sb, '\n');
1402 strbuf_release(&title);
1405 void pp_remainder(const struct pretty_print_context *pp,
1406 const char **msg_p,
1407 struct strbuf *sb,
1408 int indent)
1410 int first = 1;
1411 for (;;) {
1412 const char *line = *msg_p;
1413 int linelen = get_one_line(line);
1414 *msg_p += linelen;
1416 if (!linelen)
1417 break;
1419 if (is_empty_line(line, &linelen)) {
1420 if (first)
1421 continue;
1422 if (pp->fmt == CMIT_FMT_SHORT)
1423 break;
1425 first = 0;
1427 strbuf_grow(sb, linelen + indent + 20);
1428 if (indent) {
1429 memset(sb->buf + sb->len, ' ', indent);
1430 strbuf_setlen(sb, sb->len + indent);
1432 strbuf_add(sb, line, linelen);
1433 strbuf_addch(sb, '\n');
1437 void pretty_print_commit(const struct pretty_print_context *pp,
1438 const struct commit *commit,
1439 struct strbuf *sb)
1441 unsigned long beginning_of_body;
1442 int indent = 4;
1443 const char *msg;
1444 char *reencoded;
1445 const char *encoding;
1446 int need_8bit_cte = pp->need_8bit_cte;
1448 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1449 format_commit_message(commit, user_format, sb, pp);
1450 return;
1453 encoding = get_log_output_encoding();
1454 msg = reencoded = logmsg_reencode(commit, encoding);
1456 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1457 indent = 0;
1460 * We need to check and emit Content-type: to mark it
1461 * as 8-bit if we haven't done so.
1463 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1464 int i, ch, in_body;
1466 for (in_body = i = 0; (ch = msg[i]); i++) {
1467 if (!in_body) {
1468 /* author could be non 7-bit ASCII but
1469 * the log may be so; skip over the
1470 * header part first.
1472 if (ch == '\n' && msg[i+1] == '\n')
1473 in_body = 1;
1475 else if (non_ascii(ch)) {
1476 need_8bit_cte = 1;
1477 break;
1482 pp_header(pp, encoding, commit, &msg, sb);
1483 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1484 strbuf_addch(sb, '\n');
1487 /* Skip excess blank lines at the beginning of body, if any... */
1488 msg = skip_empty_lines(msg);
1490 /* These formats treat the title line specially. */
1491 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1492 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1494 beginning_of_body = sb->len;
1495 if (pp->fmt != CMIT_FMT_ONELINE)
1496 pp_remainder(pp, &msg, sb, indent);
1497 strbuf_rtrim(sb);
1499 /* Make sure there is an EOLN for the non-oneline case */
1500 if (pp->fmt != CMIT_FMT_ONELINE)
1501 strbuf_addch(sb, '\n');
1504 * The caller may append additional body text in e-mail
1505 * format. Make sure we did not strip the blank line
1506 * between the header and the body.
1508 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1509 strbuf_addch(sb, '\n');
1511 logmsg_free(reencoded, commit);
1514 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1515 struct strbuf *sb)
1517 struct pretty_print_context pp = {0};
1518 pp.fmt = fmt;
1519 pretty_print_commit(&pp, commit, sb);