pretty: get the correct encoding for --pretty:format=%e
[alt-git.git] / pretty.c
blobe59688bcf5d07b562d11d25fc2d2d70a4388fda7
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 char **commit_encoding,
598 const char *output_encoding)
600 static const char *utf8 = "UTF-8";
601 const char *use_encoding;
602 char *encoding;
603 char *msg = commit->buffer;
604 char *out;
606 if (!msg) {
607 enum object_type type;
608 unsigned long size;
610 msg = read_sha1_file(commit->object.sha1, &type, &size);
611 if (!msg)
612 die("Cannot read commit object %s",
613 sha1_to_hex(commit->object.sha1));
614 if (type != OBJ_COMMIT)
615 die("Expected commit for '%s', got %s",
616 sha1_to_hex(commit->object.sha1), typename(type));
619 if (!output_encoding || !*output_encoding) {
620 if (commit_encoding)
621 *commit_encoding =
622 get_header(commit, msg, "encoding");
623 return msg;
625 encoding = get_header(commit, msg, "encoding");
626 if (commit_encoding)
627 *commit_encoding = encoding;
628 use_encoding = encoding ? encoding : utf8;
629 if (same_encoding(use_encoding, output_encoding)) {
631 * No encoding work to be done. If we have no encoding header
632 * at all, then there's nothing to do, and we can return the
633 * message verbatim (whether newly allocated or not).
635 if (!encoding)
636 return msg;
639 * Otherwise, we still want to munge the encoding header in the
640 * result, which will be done by modifying the buffer. If we
641 * are using a fresh copy, we can reuse it. But if we are using
642 * the cached copy from commit->buffer, we need to duplicate it
643 * to avoid munging commit->buffer.
645 out = msg;
646 if (out == commit->buffer)
647 out = xstrdup(out);
649 else {
651 * There's actual encoding work to do. Do the reencoding, which
652 * still leaves the header to be replaced in the next step. At
653 * this point, we are done with msg. If we allocated a fresh
654 * copy, we can free it.
656 out = reencode_string(msg, output_encoding, use_encoding);
657 if (out && msg != commit->buffer)
658 free(msg);
662 * This replacement actually consumes the buffer we hand it, so we do
663 * not have to worry about freeing the old "out" here.
665 if (out)
666 out = replace_encoding_header(out, output_encoding);
668 if (!commit_encoding)
669 free(encoding);
671 * If the re-encoding failed, out might be NULL here; in that
672 * case we just return the commit message verbatim.
674 return out ? out : msg;
677 void logmsg_free(char *msg, const struct commit *commit)
679 if (msg != commit->buffer)
680 free(msg);
683 static int mailmap_name(const char **email, size_t *email_len,
684 const char **name, size_t *name_len)
686 static struct string_list *mail_map;
687 if (!mail_map) {
688 mail_map = xcalloc(1, sizeof(*mail_map));
689 read_mailmap(mail_map, NULL);
691 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
694 static size_t format_person_part(struct strbuf *sb, char part,
695 const char *msg, int len, enum date_mode dmode)
697 /* currently all placeholders have same length */
698 const int placeholder_len = 2;
699 int tz;
700 unsigned long date = 0;
701 struct ident_split s;
702 const char *name, *mail;
703 size_t maillen, namelen;
705 if (split_ident_line(&s, msg, len) < 0)
706 goto skip;
708 name = s.name_begin;
709 namelen = s.name_end - s.name_begin;
710 mail = s.mail_begin;
711 maillen = s.mail_end - s.mail_begin;
713 if (part == 'N' || part == 'E') /* mailmap lookup */
714 mailmap_name(&mail, &maillen, &name, &namelen);
715 if (part == 'n' || part == 'N') { /* name */
716 strbuf_add(sb, name, namelen);
717 return placeholder_len;
719 if (part == 'e' || part == 'E') { /* email */
720 strbuf_add(sb, mail, maillen);
721 return placeholder_len;
724 if (!s.date_begin)
725 goto skip;
727 date = strtoul(s.date_begin, NULL, 10);
729 if (part == 't') { /* date, UNIX timestamp */
730 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
731 return placeholder_len;
734 /* parse tz */
735 tz = strtoul(s.tz_begin + 1, NULL, 10);
736 if (*s.tz_begin == '-')
737 tz = -tz;
739 switch (part) {
740 case 'd': /* date */
741 strbuf_addstr(sb, show_date(date, tz, dmode));
742 return placeholder_len;
743 case 'D': /* date, RFC2822 style */
744 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
745 return placeholder_len;
746 case 'r': /* date, relative */
747 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
748 return placeholder_len;
749 case 'i': /* date, ISO 8601 */
750 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
751 return placeholder_len;
754 skip:
756 * reading from either a bogus commit, or a reflog entry with
757 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
758 * to compute a valid return value.
760 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
761 || part == 'D' || part == 'r' || part == 'i')
762 return placeholder_len;
764 return 0; /* unknown placeholder */
767 struct chunk {
768 size_t off;
769 size_t len;
772 struct format_commit_context {
773 const struct commit *commit;
774 const struct pretty_print_context *pretty_ctx;
775 unsigned commit_header_parsed:1;
776 unsigned commit_message_parsed:1;
777 struct signature_check signature_check;
778 char *message;
779 char *commit_encoding;
780 size_t width, indent1, indent2;
782 /* These offsets are relative to the start of the commit message. */
783 struct chunk author;
784 struct chunk committer;
785 size_t message_off;
786 size_t subject_off;
787 size_t body_off;
789 /* The following ones are relative to the result struct strbuf. */
790 struct chunk abbrev_commit_hash;
791 struct chunk abbrev_tree_hash;
792 struct chunk abbrev_parent_hashes;
793 size_t wrap_start;
796 static int add_again(struct strbuf *sb, struct chunk *chunk)
798 if (chunk->len) {
799 strbuf_adddup(sb, chunk->off, chunk->len);
800 return 1;
804 * We haven't seen this chunk before. Our caller is surely
805 * going to add it the hard way now. Remember the most likely
806 * start of the to-be-added chunk: the current end of the
807 * struct strbuf.
809 chunk->off = sb->len;
810 return 0;
813 static void parse_commit_header(struct format_commit_context *context)
815 const char *msg = context->message;
816 int i;
818 for (i = 0; msg[i]; i++) {
819 int eol;
820 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
821 ; /* do nothing */
823 if (i == eol) {
824 break;
825 } else if (!prefixcmp(msg + i, "author ")) {
826 context->author.off = i + 7;
827 context->author.len = eol - i - 7;
828 } else if (!prefixcmp(msg + i, "committer ")) {
829 context->committer.off = i + 10;
830 context->committer.len = eol - i - 10;
832 i = eol;
834 context->message_off = i;
835 context->commit_header_parsed = 1;
838 static int istitlechar(char c)
840 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
841 (c >= '0' && c <= '9') || c == '.' || c == '_';
844 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
846 size_t trimlen;
847 size_t start_len = sb->len;
848 int space = 2;
850 for (; *msg && *msg != '\n'; msg++) {
851 if (istitlechar(*msg)) {
852 if (space == 1)
853 strbuf_addch(sb, '-');
854 space = 0;
855 strbuf_addch(sb, *msg);
856 if (*msg == '.')
857 while (*(msg+1) == '.')
858 msg++;
859 } else
860 space |= 1;
863 /* trim any trailing '.' or '-' characters */
864 trimlen = 0;
865 while (sb->len - trimlen > start_len &&
866 (sb->buf[sb->len - 1 - trimlen] == '.'
867 || sb->buf[sb->len - 1 - trimlen] == '-'))
868 trimlen++;
869 strbuf_remove(sb, sb->len - trimlen, trimlen);
872 const char *format_subject(struct strbuf *sb, const char *msg,
873 const char *line_separator)
875 int first = 1;
877 for (;;) {
878 const char *line = msg;
879 int linelen = get_one_line(line);
881 msg += linelen;
882 if (!linelen || is_empty_line(line, &linelen))
883 break;
885 if (!sb)
886 continue;
887 strbuf_grow(sb, linelen + 2);
888 if (!first)
889 strbuf_addstr(sb, line_separator);
890 strbuf_add(sb, line, linelen);
891 first = 0;
893 return msg;
896 static void parse_commit_message(struct format_commit_context *c)
898 const char *msg = c->message + c->message_off;
899 const char *start = c->message;
901 msg = skip_empty_lines(msg);
902 c->subject_off = msg - start;
904 msg = format_subject(NULL, msg, NULL);
905 msg = skip_empty_lines(msg);
906 c->body_off = msg - start;
908 c->commit_message_parsed = 1;
911 static void format_decoration(struct strbuf *sb, const struct commit *commit)
913 struct name_decoration *d;
914 const char *prefix = " (";
916 load_ref_decorations(DECORATE_SHORT_REFS);
917 d = lookup_decoration(&name_decoration, &commit->object);
918 while (d) {
919 strbuf_addstr(sb, prefix);
920 prefix = ", ";
921 strbuf_addstr(sb, d->name);
922 d = d->next;
924 if (prefix[0] == ',')
925 strbuf_addch(sb, ')');
928 static void strbuf_wrap(struct strbuf *sb, size_t pos,
929 size_t width, size_t indent1, size_t indent2)
931 struct strbuf tmp = STRBUF_INIT;
933 if (pos)
934 strbuf_add(&tmp, sb->buf, pos);
935 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
936 (int) indent1, (int) indent2, (int) width);
937 strbuf_swap(&tmp, sb);
938 strbuf_release(&tmp);
941 static void rewrap_message_tail(struct strbuf *sb,
942 struct format_commit_context *c,
943 size_t new_width, size_t new_indent1,
944 size_t new_indent2)
946 if (c->width == new_width && c->indent1 == new_indent1 &&
947 c->indent2 == new_indent2)
948 return;
949 if (c->wrap_start < sb->len)
950 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
951 c->wrap_start = sb->len;
952 c->width = new_width;
953 c->indent1 = new_indent1;
954 c->indent2 = new_indent2;
957 static int format_reflog_person(struct strbuf *sb,
958 char part,
959 struct reflog_walk_info *log,
960 enum date_mode dmode)
962 const char *ident;
964 if (!log)
965 return 2;
967 ident = get_reflog_ident(log);
968 if (!ident)
969 return 2;
971 return format_person_part(sb, part, ident, strlen(ident), dmode);
974 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
975 void *context)
977 struct format_commit_context *c = context;
978 const struct commit *commit = c->commit;
979 const char *msg = c->message;
980 struct commit_list *p;
981 int h1, h2;
983 /* these are independent of the commit */
984 switch (placeholder[0]) {
985 case 'C':
986 if (placeholder[1] == '(') {
987 const char *begin = placeholder + 2;
988 const char *end = strchr(begin, ')');
989 char color[COLOR_MAXLEN];
991 if (!end)
992 return 0;
993 if (!prefixcmp(begin, "auto,")) {
994 if (!want_color(c->pretty_ctx->color))
995 return end - placeholder + 1;
996 begin += 5;
998 color_parse_mem(begin,
999 end - begin,
1000 "--pretty format", color);
1001 strbuf_addstr(sb, color);
1002 return end - placeholder + 1;
1004 if (!prefixcmp(placeholder + 1, "red")) {
1005 strbuf_addstr(sb, GIT_COLOR_RED);
1006 return 4;
1007 } else if (!prefixcmp(placeholder + 1, "green")) {
1008 strbuf_addstr(sb, GIT_COLOR_GREEN);
1009 return 6;
1010 } else if (!prefixcmp(placeholder + 1, "blue")) {
1011 strbuf_addstr(sb, GIT_COLOR_BLUE);
1012 return 5;
1013 } else if (!prefixcmp(placeholder + 1, "reset")) {
1014 strbuf_addstr(sb, GIT_COLOR_RESET);
1015 return 6;
1016 } else
1017 return 0;
1018 case 'n': /* newline */
1019 strbuf_addch(sb, '\n');
1020 return 1;
1021 case 'x':
1022 /* %x00 == NUL, %x0a == LF, etc. */
1023 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1024 h1 <= 16 &&
1025 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1026 h2 <= 16) {
1027 strbuf_addch(sb, (h1<<4)|h2);
1028 return 3;
1029 } else
1030 return 0;
1031 case 'w':
1032 if (placeholder[1] == '(') {
1033 unsigned long width = 0, indent1 = 0, indent2 = 0;
1034 char *next;
1035 const char *start = placeholder + 2;
1036 const char *end = strchr(start, ')');
1037 if (!end)
1038 return 0;
1039 if (end > start) {
1040 width = strtoul(start, &next, 10);
1041 if (*next == ',') {
1042 indent1 = strtoul(next + 1, &next, 10);
1043 if (*next == ',') {
1044 indent2 = strtoul(next + 1,
1045 &next, 10);
1048 if (*next != ')')
1049 return 0;
1051 rewrap_message_tail(sb, c, width, indent1, indent2);
1052 return end - placeholder + 1;
1053 } else
1054 return 0;
1057 /* these depend on the commit */
1058 if (!commit->object.parsed)
1059 parse_object(commit->object.sha1);
1061 switch (placeholder[0]) {
1062 case 'H': /* commit hash */
1063 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1064 return 1;
1065 case 'h': /* abbreviated commit hash */
1066 if (add_again(sb, &c->abbrev_commit_hash))
1067 return 1;
1068 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1069 c->pretty_ctx->abbrev));
1070 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1071 return 1;
1072 case 'T': /* tree hash */
1073 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1074 return 1;
1075 case 't': /* abbreviated tree hash */
1076 if (add_again(sb, &c->abbrev_tree_hash))
1077 return 1;
1078 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1079 c->pretty_ctx->abbrev));
1080 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1081 return 1;
1082 case 'P': /* parent hashes */
1083 for (p = commit->parents; p; p = p->next) {
1084 if (p != commit->parents)
1085 strbuf_addch(sb, ' ');
1086 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1088 return 1;
1089 case 'p': /* abbreviated parent hashes */
1090 if (add_again(sb, &c->abbrev_parent_hashes))
1091 return 1;
1092 for (p = commit->parents; p; p = p->next) {
1093 if (p != commit->parents)
1094 strbuf_addch(sb, ' ');
1095 strbuf_addstr(sb, find_unique_abbrev(
1096 p->item->object.sha1,
1097 c->pretty_ctx->abbrev));
1099 c->abbrev_parent_hashes.len = sb->len -
1100 c->abbrev_parent_hashes.off;
1101 return 1;
1102 case 'm': /* left/right/bottom */
1103 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1104 return 1;
1105 case 'd':
1106 format_decoration(sb, commit);
1107 return 1;
1108 case 'g': /* reflog info */
1109 switch(placeholder[1]) {
1110 case 'd': /* reflog selector */
1111 case 'D':
1112 if (c->pretty_ctx->reflog_info)
1113 get_reflog_selector(sb,
1114 c->pretty_ctx->reflog_info,
1115 c->pretty_ctx->date_mode,
1116 c->pretty_ctx->date_mode_explicit,
1117 (placeholder[1] == 'd'));
1118 return 2;
1119 case 's': /* reflog message */
1120 if (c->pretty_ctx->reflog_info)
1121 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1122 return 2;
1123 case 'n':
1124 case 'N':
1125 case 'e':
1126 case 'E':
1127 return format_reflog_person(sb,
1128 placeholder[1],
1129 c->pretty_ctx->reflog_info,
1130 c->pretty_ctx->date_mode);
1132 return 0; /* unknown %g placeholder */
1133 case 'N':
1134 if (c->pretty_ctx->notes_message) {
1135 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1136 return 1;
1138 return 0;
1141 if (placeholder[0] == 'G') {
1142 if (!c->signature_check.result)
1143 check_commit_signature(c->commit, &(c->signature_check));
1144 switch (placeholder[1]) {
1145 case 'G':
1146 if (c->signature_check.gpg_output)
1147 strbuf_addstr(sb, c->signature_check.gpg_output);
1148 break;
1149 case '?':
1150 switch (c->signature_check.result) {
1151 case 'G':
1152 case 'B':
1153 case 'U':
1154 case 'N':
1155 strbuf_addch(sb, c->signature_check.result);
1157 break;
1158 case 'S':
1159 if (c->signature_check.signer)
1160 strbuf_addstr(sb, c->signature_check.signer);
1161 break;
1162 case 'K':
1163 if (c->signature_check.key)
1164 strbuf_addstr(sb, c->signature_check.key);
1165 break;
1167 return 2;
1171 /* For the rest we have to parse the commit header. */
1172 if (!c->commit_header_parsed)
1173 parse_commit_header(c);
1175 switch (placeholder[0]) {
1176 case 'a': /* author ... */
1177 return format_person_part(sb, placeholder[1],
1178 msg + c->author.off, c->author.len,
1179 c->pretty_ctx->date_mode);
1180 case 'c': /* committer ... */
1181 return format_person_part(sb, placeholder[1],
1182 msg + c->committer.off, c->committer.len,
1183 c->pretty_ctx->date_mode);
1184 case 'e': /* encoding */
1185 if (c->commit_encoding)
1186 strbuf_addstr(sb, c->commit_encoding);
1187 return 1;
1188 case 'B': /* raw body */
1189 /* message_off is always left at the initial newline */
1190 strbuf_addstr(sb, msg + c->message_off + 1);
1191 return 1;
1194 /* Now we need to parse the commit message. */
1195 if (!c->commit_message_parsed)
1196 parse_commit_message(c);
1198 switch (placeholder[0]) {
1199 case 's': /* subject */
1200 format_subject(sb, msg + c->subject_off, " ");
1201 return 1;
1202 case 'f': /* sanitized subject */
1203 format_sanitized_subject(sb, msg + c->subject_off);
1204 return 1;
1205 case 'b': /* body */
1206 strbuf_addstr(sb, msg + c->body_off);
1207 return 1;
1209 return 0; /* unknown placeholder */
1212 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1213 void *context)
1215 int consumed;
1216 size_t orig_len;
1217 enum {
1218 NO_MAGIC,
1219 ADD_LF_BEFORE_NON_EMPTY,
1220 DEL_LF_BEFORE_EMPTY,
1221 ADD_SP_BEFORE_NON_EMPTY
1222 } magic = NO_MAGIC;
1224 switch (placeholder[0]) {
1225 case '-':
1226 magic = DEL_LF_BEFORE_EMPTY;
1227 break;
1228 case '+':
1229 magic = ADD_LF_BEFORE_NON_EMPTY;
1230 break;
1231 case ' ':
1232 magic = ADD_SP_BEFORE_NON_EMPTY;
1233 break;
1234 default:
1235 break;
1237 if (magic != NO_MAGIC)
1238 placeholder++;
1240 orig_len = sb->len;
1241 consumed = format_commit_one(sb, placeholder, context);
1242 if (magic == NO_MAGIC)
1243 return consumed;
1245 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1246 while (sb->len && sb->buf[sb->len - 1] == '\n')
1247 strbuf_setlen(sb, sb->len - 1);
1248 } else if (orig_len != sb->len) {
1249 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1250 strbuf_insert(sb, orig_len, "\n", 1);
1251 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1252 strbuf_insert(sb, orig_len, " ", 1);
1254 return consumed + 1;
1257 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1258 void *context)
1260 struct userformat_want *w = context;
1262 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1263 placeholder++;
1265 switch (*placeholder) {
1266 case 'N':
1267 w->notes = 1;
1268 break;
1270 return 0;
1273 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1275 struct strbuf dummy = STRBUF_INIT;
1277 if (!fmt) {
1278 if (!user_format)
1279 return;
1280 fmt = user_format;
1282 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1283 strbuf_release(&dummy);
1286 void format_commit_message(const struct commit *commit,
1287 const char *format, struct strbuf *sb,
1288 const struct pretty_print_context *pretty_ctx)
1290 struct format_commit_context context;
1291 const char *output_enc = pretty_ctx->output_encoding;
1293 memset(&context, 0, sizeof(context));
1294 context.commit = commit;
1295 context.pretty_ctx = pretty_ctx;
1296 context.wrap_start = sb->len;
1297 context.message = logmsg_reencode(commit,
1298 &context.commit_encoding,
1299 output_enc);
1301 strbuf_expand(sb, format, format_commit_item, &context);
1302 rewrap_message_tail(sb, &context, 0, 0, 0);
1304 free(context.commit_encoding);
1305 logmsg_free(context.message, commit);
1306 free(context.signature_check.gpg_output);
1307 free(context.signature_check.signer);
1310 static void pp_header(const struct pretty_print_context *pp,
1311 const char *encoding,
1312 const struct commit *commit,
1313 const char **msg_p,
1314 struct strbuf *sb)
1316 int parents_shown = 0;
1318 for (;;) {
1319 const char *line = *msg_p;
1320 int linelen = get_one_line(*msg_p);
1322 if (!linelen)
1323 return;
1324 *msg_p += linelen;
1326 if (linelen == 1)
1327 /* End of header */
1328 return;
1330 if (pp->fmt == CMIT_FMT_RAW) {
1331 strbuf_add(sb, line, linelen);
1332 continue;
1335 if (!prefixcmp(line, "parent ")) {
1336 if (linelen != 48)
1337 die("bad parent line in commit");
1338 continue;
1341 if (!parents_shown) {
1342 struct commit_list *parent;
1343 int num;
1344 for (parent = commit->parents, num = 0;
1345 parent;
1346 parent = parent->next, num++)
1348 /* with enough slop */
1349 strbuf_grow(sb, num * 50 + 20);
1350 add_merge_info(pp, sb, commit);
1351 parents_shown = 1;
1355 * MEDIUM == DEFAULT shows only author with dates.
1356 * FULL shows both authors but not dates.
1357 * FULLER shows both authors and dates.
1359 if (!prefixcmp(line, "author ")) {
1360 strbuf_grow(sb, linelen + 80);
1361 pp_user_info(pp, "Author", sb, line + 7, encoding);
1363 if (!prefixcmp(line, "committer ") &&
1364 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1365 strbuf_grow(sb, linelen + 80);
1366 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1371 void pp_title_line(const struct pretty_print_context *pp,
1372 const char **msg_p,
1373 struct strbuf *sb,
1374 const char *encoding,
1375 int need_8bit_cte)
1377 static const int max_length = 78; /* per rfc2047 */
1378 struct strbuf title;
1380 strbuf_init(&title, 80);
1381 *msg_p = format_subject(&title, *msg_p,
1382 pp->preserve_subject ? "\n" : " ");
1384 strbuf_grow(sb, title.len + 1024);
1385 if (pp->subject) {
1386 strbuf_addstr(sb, pp->subject);
1387 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1388 add_rfc2047(sb, title.buf, title.len,
1389 encoding, RFC2047_SUBJECT);
1390 else
1391 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1392 -last_line_length(sb), 1, max_length);
1393 } else {
1394 strbuf_addbuf(sb, &title);
1396 strbuf_addch(sb, '\n');
1398 if (need_8bit_cte > 0) {
1399 const char *header_fmt =
1400 "MIME-Version: 1.0\n"
1401 "Content-Type: text/plain; charset=%s\n"
1402 "Content-Transfer-Encoding: 8bit\n";
1403 strbuf_addf(sb, header_fmt, encoding);
1405 if (pp->after_subject) {
1406 strbuf_addstr(sb, pp->after_subject);
1408 if (pp->fmt == CMIT_FMT_EMAIL) {
1409 strbuf_addch(sb, '\n');
1411 strbuf_release(&title);
1414 void pp_remainder(const struct pretty_print_context *pp,
1415 const char **msg_p,
1416 struct strbuf *sb,
1417 int indent)
1419 int first = 1;
1420 for (;;) {
1421 const char *line = *msg_p;
1422 int linelen = get_one_line(line);
1423 *msg_p += linelen;
1425 if (!linelen)
1426 break;
1428 if (is_empty_line(line, &linelen)) {
1429 if (first)
1430 continue;
1431 if (pp->fmt == CMIT_FMT_SHORT)
1432 break;
1434 first = 0;
1436 strbuf_grow(sb, linelen + indent + 20);
1437 if (indent) {
1438 memset(sb->buf + sb->len, ' ', indent);
1439 strbuf_setlen(sb, sb->len + indent);
1441 strbuf_add(sb, line, linelen);
1442 strbuf_addch(sb, '\n');
1446 void pretty_print_commit(const struct pretty_print_context *pp,
1447 const struct commit *commit,
1448 struct strbuf *sb)
1450 unsigned long beginning_of_body;
1451 int indent = 4;
1452 const char *msg;
1453 char *reencoded;
1454 const char *encoding;
1455 int need_8bit_cte = pp->need_8bit_cte;
1457 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1458 format_commit_message(commit, user_format, sb, pp);
1459 return;
1462 encoding = get_log_output_encoding();
1463 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1465 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1466 indent = 0;
1469 * We need to check and emit Content-type: to mark it
1470 * as 8-bit if we haven't done so.
1472 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1473 int i, ch, in_body;
1475 for (in_body = i = 0; (ch = msg[i]); i++) {
1476 if (!in_body) {
1477 /* author could be non 7-bit ASCII but
1478 * the log may be so; skip over the
1479 * header part first.
1481 if (ch == '\n' && msg[i+1] == '\n')
1482 in_body = 1;
1484 else if (non_ascii(ch)) {
1485 need_8bit_cte = 1;
1486 break;
1491 pp_header(pp, encoding, commit, &msg, sb);
1492 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1493 strbuf_addch(sb, '\n');
1496 /* Skip excess blank lines at the beginning of body, if any... */
1497 msg = skip_empty_lines(msg);
1499 /* These formats treat the title line specially. */
1500 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1501 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1503 beginning_of_body = sb->len;
1504 if (pp->fmt != CMIT_FMT_ONELINE)
1505 pp_remainder(pp, &msg, sb, indent);
1506 strbuf_rtrim(sb);
1508 /* Make sure there is an EOLN for the non-oneline case */
1509 if (pp->fmt != CMIT_FMT_ONELINE)
1510 strbuf_addch(sb, '\n');
1513 * The caller may append additional body text in e-mail
1514 * format. Make sure we did not strip the blank line
1515 * between the header and the body.
1517 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1518 strbuf_addch(sb, '\n');
1520 logmsg_free(reencoded, commit);
1523 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1524 struct strbuf *sb)
1526 struct pretty_print_context pp = {0};
1527 pp.fmt = fmt;
1528 pretty_print_commit(&pp, commit, sb);