pretty.c

   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
  13 static char *user_format;
  14 static struct cmt_fmt_map {
  15         const char *name;
  16         enum cmit_fmt format;
  17         int is_tformat;
  18         int is_alias;
  19         const char *user_format;
  20 } *commit_formats;
  21 static size_t builtin_formats_len;
  22 static size_t commit_formats_len;
  23 static size_t commit_formats_alloc;
  24 static struct cmt_fmt_map *find_commit_format(const char *sought);
  25
  26 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  27 {
  28         free(user_format);
  29         user_format = xstrdup(cp);
  30         if (is_tformat)
  31                 rev->use_terminator = 1;
  32         rev->commit_format = CMIT_FMT_USERFORMAT;
  33 }
  34
  35 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  36 {
  37         struct cmt_fmt_map *commit_format = NULL;
  38         const char *name;
  39         const char *fmt;
  40         int i;
  41
  42         if (prefixcmp(var, "pretty."))
  43                 return 0;
  44
  45         name = var + strlen("pretty.");
  46         for (i = 0; i < builtin_formats_len; i++) {
  47                 if (!strcmp(commit_formats[i].name, name))
  48                         return 0;
  49         }
  50
  51         for (i = builtin_formats_len; i < commit_formats_len; i++) {
  52                 if (!strcmp(commit_formats[i].name, name)) {
  53                         commit_format = &commit_formats[i];
  54                         break;
  55                 }
  56         }
  57
  58         if (!commit_format) {
  59                 ALLOC_GROW(commit_formats, commit_formats_len+1,
  60                            commit_formats_alloc);
  61                 commit_format = &commit_formats[commit_formats_len];
  62                 memset(commit_format, 0, sizeof(*commit_format));
  63                 commit_formats_len++;
  64         }
  65
  66         commit_format->name = xstrdup(name);
  67         commit_format->format = CMIT_FMT_USERFORMAT;
  68         git_config_string(&fmt, var, value);
  69         if (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
  70                 commit_format->is_tformat = fmt[0] == 't';
  71                 fmt = strchr(fmt, ':') + 1;
  72         } else if (strchr(fmt, '%'))
  73                 commit_format->is_tformat = 1;
  74         else
  75                 commit_format->is_alias = 1;
  76         commit_format->user_format = fmt;
  77
  78         return 0;
  79 }
  80
  81 static void setup_commit_formats(void)
  82 {
  83         struct cmt_fmt_map builtin_formats[] = {
  84                 { "raw",        CMIT_FMT_RAW,           0 },
  85                 { "medium",     CMIT_FMT_MEDIUM,        0 },
  86                 { "short",      CMIT_FMT_SHORT,         0 },
  87                 { "email",      CMIT_FMT_EMAIL,         0 },
  88                 { "fuller",     CMIT_FMT_FULLER,        0 },
  89                 { "full",       CMIT_FMT_FULL,          0 },
  90                 { "oneline",    CMIT_FMT_ONELINE,       1 }
  91         };
  92         commit_formats_len = ARRAY_SIZE(builtin_formats);
  93         builtin_formats_len = commit_formats_len;
  94         ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
  95         memcpy(commit_formats, builtin_formats,
  96                sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  97
  98         git_config(git_pretty_formats_config, NULL);
  99 }
 100
 101 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 102                                                         const char *original,
 103                                                         int num_redirections)
 104 {
 105         struct cmt_fmt_map *found = NULL;
 106         size_t found_match_len = 0;
 107         int i;
 108
 109         if (num_redirections >= commit_formats_len)
 110                 die("invalid --pretty format: "
 111                     "'%s' references an alias which points to itself",
 112                     original);
 113
 114         for (i = 0; i < commit_formats_len; i++) {
 115                 size_t match_len;
 116
 117                 if (prefixcmp(commit_formats[i].name, sought))
 118                         continue;
 119
 120                 match_len = strlen(commit_formats[i].name);
 121                 if (found == NULL || found_match_len > match_len) {
 122                         found = &commit_formats[i];
 123                         found_match_len = match_len;
 124                 }
 125         }
 126
 127         if (found && found->is_alias) {
 128                 found = find_commit_format_recursive(found->user_format,
 129                                                      original,
 130                                                      num_redirections+1);
 131         }
 132
 133         return found;
 134 }
 135
 136 static struct cmt_fmt_map *find_commit_format(const char *sought)
 137 {
 138         if (!commit_formats)
 139                 setup_commit_formats();
 140
 141         return find_commit_format_recursive(sought, sought, 0);
 142 }
 143
 144 void get_commit_format(const char *arg, struct rev_info *rev)
 145 {
 146         struct cmt_fmt_map *commit_format;
 147
 148         rev->use_terminator = 0;
 149         if (!arg || !*arg) {
 150                 rev->commit_format = CMIT_FMT_DEFAULT;
 151                 return;
 152         }
 153         if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
 154                 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 155                 return;
 156         }
 157
 158         if (strchr(arg, '%')) {
 159                 save_user_format(rev, arg, 1);
 160                 return;
 161         }
 162
 163         commit_format = find_commit_format(arg);
 164         if (!commit_format)
 165                 die("invalid --pretty format: %s", arg);
 166
 167         rev->commit_format = commit_format->format;
 168         rev->use_terminator = commit_format->is_tformat;
 169         if (commit_format->format == CMIT_FMT_USERFORMAT) {
 170                 save_user_format(rev, commit_format->user_format,
 171                                  commit_format->is_tformat);
 172         }
 173 }
 174
 175 /*
 176  * Generic support for pretty-printing the header
 177  */
 178 static int get_one_line(const char *msg)
 179 {
 180         int ret = 0;
 181
 182         for (;;) {
 183                 char c = *msg++;
 184                 if (!c)
 185                         break;
 186                 ret++;
 187                 if (c == '\n')
 188                         break;
 189         }
 190         return ret;
 191 }
 192
 193 /* High bit set, or ISO-2022-INT */
 194 static int non_ascii(int ch)
 195 {
 196         return !isascii(ch) || ch == '\033';
 197 }
 198
 199 int has_non_ascii(const char *s)
 200 {
 201         int ch;
 202         if (!s)
 203                 return 0;
 204         while ((ch = *s++) != '\0') {
 205                 if (non_ascii(ch))
 206                         return 1;
 207         }
 208         return 0;
 209 }
 210
 211 static int is_rfc822_special(char ch)
 212 {
 213         switch (ch) {
 214         case '(':
 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                 return 1;
 228         default:
 229                 return 0;
 230         }
 231 }
 232
 233 static int has_rfc822_specials(const char *s, int len)
 234 {
 235         int i;
 236         for (i = 0; i < len; i++)
 237                 if (is_rfc822_special(s[i]))
 238                         return 1;
 239         return 0;
 240 }
 241
 242 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 243 {
 244         int i;
 245
 246         /* just a guess, we may have to also backslash-quote */
 247         strbuf_grow(out, len + 2);
 248
 249         strbuf_addch(out, '"');
 250         for (i = 0; i < len; i++) {
 251                 switch (s[i]) {
 252                 case '"':
 253                 case '\\':
 254                         strbuf_addch(out, '\\');
 255                         /* fall through */
 256                 default:
 257                         strbuf_addch(out, s[i]);
 258                 }
 259         }
 260         strbuf_addch(out, '"');
 261 }
 262
 263 static int is_rfc2047_special(char ch)
 264 {
 265         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 266 }
 267
 268 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 269                        const char *encoding)
 270 {
 271         static const int max_length = 78; /* per rfc2822 */
 272         int i;
 273         int line_len;
 274
 275         /* How many bytes are already used on the current line? */
 276         for (i = sb->len - 1; i >= 0; i--)
 277                 if (sb->buf[i] == '\n')
 278                         break;
 279         line_len = sb->len - (i+1);
 280
 281         for (i = 0; i < len; i++) {
 282                 int ch = line[i];
 283                 if (non_ascii(ch) || ch == '\n')
 284                         goto needquote;
 285                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 286                         goto needquote;
 287         }
 288         strbuf_add_wrapped_bytes(sb, line, len, 0, 1, max_length - line_len);
 289         return;
 290
 291 needquote:
 292         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 293         strbuf_addf(sb, "=?%s?q?", encoding);
 294         line_len += strlen(encoding) + 5; /* 5 for =??q? */
 295         for (i = 0; i < len; i++) {
 296                 unsigned ch = line[i] & 0xFF;
 297
 298                 if (line_len >= max_length - 2) {
 299                         strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 300                         line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 301                 }
 302
 303                 /*
 304                  * We encode ' ' using '=20' even though rfc2047
 305                  * allows using '_' for readability.  Unfortunately,
 306                  * many programs do not understand this and just
 307                  * leave the underscore in place.
 308                  */
 309                 if (is_rfc2047_special(ch) || ch == ' ' || ch == '\n') {
 310                         strbuf_addf(sb, "=%02X", ch);
 311                         line_len += 3;
 312                 }
 313                 else {
 314                         strbuf_addch(sb, ch);
 315                         line_len++;
 316                 }
 317         }
 318         strbuf_addstr(sb, "?=");
 319 }
 320
 321 void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 322                   const char *line, enum date_mode dmode,
 323                   const char *encoding)
 324 {
 325         char *date;
 326         int namelen;
 327         unsigned long time;
 328         int tz;
 329
 330         if (fmt == CMIT_FMT_ONELINE)
 331                 return;
 332         date = strchr(line, '>');
 333         if (!date)
 334                 return;
 335         namelen = ++date - line;
 336         time = strtoul(date, &date, 10);
 337         tz = strtol(date, NULL, 10);
 338
 339         if (fmt == CMIT_FMT_EMAIL) {
 340                 char *name_tail = strchr(line, '<');
 341                 int display_name_length;
 342                 if (!name_tail)
 343                         return;
 344                 while (line < name_tail && isspace(name_tail[-1]))
 345                         name_tail--;
 346                 display_name_length = name_tail - line;
 347                 strbuf_addstr(sb, "From: ");
 348                 if (!has_rfc822_specials(line, display_name_length)) {
 349                         add_rfc2047(sb, line, display_name_length, encoding);
 350                 } else {
 351                         struct strbuf quoted = STRBUF_INIT;
 352                         add_rfc822_quoted(&quoted, line, display_name_length);
 353                         add_rfc2047(sb, quoted.buf, quoted.len, encoding);
 354                         strbuf_release(&quoted);
 355                 }
 356                 strbuf_add(sb, name_tail, namelen - display_name_length);
 357                 strbuf_addch(sb, '\n');
 358         } else {
 359                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 360                               (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 361                               "    ", namelen, line);
 362         }
 363         switch (fmt) {
 364         case CMIT_FMT_MEDIUM:
 365                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 366                 break;
 367         case CMIT_FMT_EMAIL:
 368                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 369                 break;
 370         case CMIT_FMT_FULLER:
 371                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 372                 break;
 373         default:
 374                 /* notin' */
 375                 break;
 376         }
 377 }
 378
 379 static int is_empty_line(const char *line, int *len_p)
 380 {
 381         int len = *len_p;
 382         while (len && isspace(line[len-1]))
 383                 len--;
 384         *len_p = len;
 385         return !len;
 386 }
 387
 388 static const char *skip_empty_lines(const char *msg)
 389 {
 390         for (;;) {
 391                 int linelen = get_one_line(msg);
 392                 int ll = linelen;
 393                 if (!linelen)
 394                         break;
 395                 if (!is_empty_line(msg, &ll))
 396                         break;
 397                 msg += linelen;
 398         }
 399         return msg;
 400 }
 401
 402 static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 403                         const struct commit *commit, int abbrev)
 404 {
 405         struct commit_list *parent = commit->parents;
 406
 407         if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 408             !parent || !parent->next)
 409                 return;
 410
 411         strbuf_addstr(sb, "Merge:");
 412
 413         while (parent) {
 414                 struct commit *p = parent->item;
 415                 const char *hex = NULL;
 416                 if (abbrev)
 417                         hex = find_unique_abbrev(p->object.sha1, abbrev);
 418                 if (!hex)
 419                         hex = sha1_to_hex(p->object.sha1);
 420                 parent = parent->next;
 421
 422                 strbuf_addf(sb, " %s", hex);
 423         }
 424         strbuf_addch(sb, '\n');
 425 }
 426
 427 static char *get_header(const struct commit *commit, const char *key)
 428 {
 429         int key_len = strlen(key);
 430         const char *line = commit->buffer;
 431
 432         for (;;) {
 433                 const char *eol = strchr(line, '\n'), *next;
 434
 435                 if (line == eol)
 436                         return NULL;
 437                 if (!eol) {
 438                         eol = line + strlen(line);
 439                         next = NULL;
 440                 } else
 441                         next = eol + 1;
 442                 if (eol - line > key_len &&
 443                     !strncmp(line, key, key_len) &&
 444                     line[key_len] == ' ') {
 445                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 446                 }
 447                 line = next;
 448         }
 449 }
 450
 451 static char *replace_encoding_header(char *buf, const char *encoding)
 452 {
 453         struct strbuf tmp = STRBUF_INIT;
 454         size_t start, len;
 455         char *cp = buf;
 456
 457         /* guess if there is an encoding header before a \n\n */
 458         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 459                 cp = strchr(cp, '\n');
 460                 if (!cp || *++cp == '\n')
 461                         return buf;
 462         }
 463         start = cp - buf;
 464         cp = strchr(cp, '\n');
 465         if (!cp)
 466                 return buf; /* should not happen but be defensive */
 467         len = cp + 1 - (buf + start);
 468
 469         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 470         if (is_encoding_utf8(encoding)) {
 471                 /* we have re-coded to UTF-8; drop the header */
 472                 strbuf_remove(&tmp, start, len);
 473         } else {
 474                 /* just replaces XXXX in 'encoding XXXX\n' */
 475                 strbuf_splice(&tmp, start + strlen("encoding "),
 476                                           len - strlen("encoding \n"),
 477                                           encoding, strlen(encoding));
 478         }
 479         return strbuf_detach(&tmp, NULL);
 480 }
 481
 482 char *logmsg_reencode(const struct commit *commit,
 483                       const char *output_encoding)
 484 {
 485         static const char *utf8 = "UTF-8";
 486         const char *use_encoding;
 487         char *encoding;
 488         char *out;
 489
 490         if (!*output_encoding)
 491                 return NULL;
 492         encoding = get_header(commit, "encoding");
 493         use_encoding = encoding ? encoding : utf8;
 494         if (!strcmp(use_encoding, output_encoding))
 495                 if (encoding) /* we'll strip encoding header later */
 496                         out = xstrdup(commit->buffer);
 497                 else
 498                         return NULL; /* nothing to do */
 499         else
 500                 out = reencode_string(commit->buffer,
 501                                       output_encoding, use_encoding);
 502         if (out)
 503                 out = replace_encoding_header(out, output_encoding);
 504
 505         free(encoding);
 506         return out;
 507 }
 508
 509 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 510 {
 511         static struct string_list *mail_map;
 512         if (!mail_map) {
 513                 mail_map = xcalloc(1, sizeof(*mail_map));
 514                 read_mailmap(mail_map, NULL);
 515         }
 516         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 517 }
 518
 519 static size_t format_person_part(struct strbuf *sb, char part,
 520                                  const char *msg, int len, enum date_mode dmode)
 521 {
 522         /* currently all placeholders have same length */
 523         const int placeholder_len = 2;
 524         int start, end, tz = 0;
 525         unsigned long date = 0;
 526         char *ep;
 527         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 528         char person_name[1024];
 529         char person_mail[1024];
 530
 531         /* advance 'end' to point to email start delimiter */
 532         for (end = 0; end < len && msg[end] != '<'; end++)
 533                 ; /* do nothing */
 534
 535         /*
 536          * When end points at the '<' that we found, it should have
 537          * matching '>' later, which means 'end' must be strictly
 538          * below len - 1.
 539          */
 540         if (end >= len - 2)
 541                 goto skip;
 542
 543         /* Seek for both name and email part */
 544         name_start = msg;
 545         name_end = msg+end;
 546         while (name_end > name_start && isspace(*(name_end-1)))
 547                 name_end--;
 548         mail_start = msg+end+1;
 549         mail_end = mail_start;
 550         while (mail_end < msg_end && *mail_end != '>')
 551                 mail_end++;
 552         if (mail_end == msg_end)
 553                 goto skip;
 554         end = mail_end-msg;
 555
 556         if (part == 'N' || part == 'E') { /* mailmap lookup */
 557                 strlcpy(person_name, name_start, name_end-name_start+1);
 558                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 559                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 560                 name_start = person_name;
 561                 name_end = name_start + strlen(person_name);
 562                 mail_start = person_mail;
 563                 mail_end = mail_start +  strlen(person_mail);
 564         }
 565         if (part == 'n' || part == 'N') {       /* name */
 566                 strbuf_add(sb, name_start, name_end-name_start);
 567                 return placeholder_len;
 568         }
 569         if (part == 'e' || part == 'E') {       /* email */
 570                 strbuf_add(sb, mail_start, mail_end-mail_start);
 571                 return placeholder_len;
 572         }
 573
 574         /* advance 'start' to point to date start delimiter */
 575         for (start = end + 1; start < len && isspace(msg[start]); start++)
 576                 ; /* do nothing */
 577         if (start >= len)
 578                 goto skip;
 579         date = strtoul(msg + start, &ep, 10);
 580         if (msg + start == ep)
 581                 goto skip;
 582
 583         if (part == 't') {      /* date, UNIX timestamp */
 584                 strbuf_add(sb, msg + start, ep - (msg + start));
 585                 return placeholder_len;
 586         }
 587
 588         /* parse tz */
 589         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 590                 ; /* do nothing */
 591         if (start + 1 < len) {
 592                 tz = strtoul(msg + start + 1, NULL, 10);
 593                 if (msg[start] == '-')
 594                         tz = -tz;
 595         }
 596
 597         switch (part) {
 598         case 'd':       /* date */
 599                 strbuf_addstr(sb, show_date(date, tz, dmode));
 600                 return placeholder_len;
 601         case 'D':       /* date, RFC2822 style */
 602                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 603                 return placeholder_len;
 604         case 'r':       /* date, relative */
 605                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 606                 return placeholder_len;
 607         case 'i':       /* date, ISO 8601 */
 608                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 609                 return placeholder_len;
 610         }
 611
 612 skip:
 613         /*
 614          * bogus commit, 'sb' cannot be updated, but we still need to
 615          * compute a valid return value.
 616          */
 617         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 618             || part == 'D' || part == 'r' || part == 'i')
 619                 return placeholder_len;
 620
 621         return 0; /* unknown placeholder */
 622 }
 623
 624 struct chunk {
 625         size_t off;
 626         size_t len;
 627 };
 628
 629 struct format_commit_context {
 630         const struct commit *commit;
 631         const struct pretty_print_context *pretty_ctx;
 632         unsigned commit_header_parsed:1;
 633         unsigned commit_message_parsed:1;
 634         char *message;
 635         size_t width, indent1, indent2;
 636
 637         /* These offsets are relative to the start of the commit message. */
 638         struct chunk author;
 639         struct chunk committer;
 640         struct chunk encoding;
 641         size_t message_off;
 642         size_t subject_off;
 643         size_t body_off;
 644
 645         /* The following ones are relative to the result struct strbuf. */
 646         struct chunk abbrev_commit_hash;
 647         struct chunk abbrev_tree_hash;
 648         struct chunk abbrev_parent_hashes;
 649         size_t wrap_start;
 650 };
 651
 652 static int add_again(struct strbuf *sb, struct chunk *chunk)
 653 {
 654         if (chunk->len) {
 655                 strbuf_adddup(sb, chunk->off, chunk->len);
 656                 return 1;
 657         }
 658
 659         /*
 660          * We haven't seen this chunk before.  Our caller is surely
 661          * going to add it the hard way now.  Remember the most likely
 662          * start of the to-be-added chunk: the current end of the
 663          * struct strbuf.
 664          */
 665         chunk->off = sb->len;
 666         return 0;
 667 }
 668
 669 static void parse_commit_header(struct format_commit_context *context)
 670 {
 671         const char *msg = context->message;
 672         int i;
 673
 674         for (i = 0; msg[i]; i++) {
 675                 int eol;
 676                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 677                         ; /* do nothing */
 678
 679                 if (i == eol) {
 680                         break;
 681                 } else if (!prefixcmp(msg + i, "author ")) {
 682                         context->author.off = i + 7;
 683                         context->author.len = eol - i - 7;
 684                 } else if (!prefixcmp(msg + i, "committer ")) {
 685                         context->committer.off = i + 10;
 686                         context->committer.len = eol - i - 10;
 687                 } else if (!prefixcmp(msg + i, "encoding ")) {
 688                         context->encoding.off = i + 9;
 689                         context->encoding.len = eol - i - 9;
 690                 }
 691                 i = eol;
 692         }
 693         context->message_off = i;
 694         context->commit_header_parsed = 1;
 695 }
 696
 697 static int istitlechar(char c)
 698 {
 699         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 700                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 701 }
 702
 703 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 704 {
 705         size_t trimlen;
 706         size_t start_len = sb->len;
 707         int space = 2;
 708
 709         for (; *msg && *msg != '\n'; msg++) {
 710                 if (istitlechar(*msg)) {
 711                         if (space == 1)
 712                                 strbuf_addch(sb, '-');
 713                         space = 0;
 714                         strbuf_addch(sb, *msg);
 715                         if (*msg == '.')
 716                                 while (*(msg+1) == '.')
 717                                         msg++;
 718                 } else
 719                         space |= 1;
 720         }
 721
 722         /* trim any trailing '.' or '-' characters */
 723         trimlen = 0;
 724         while (sb->len - trimlen > start_len &&
 725                 (sb->buf[sb->len - 1 - trimlen] == '.'
 726                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 727                 trimlen++;
 728         strbuf_remove(sb, sb->len - trimlen, trimlen);
 729 }
 730
 731 const char *format_subject(struct strbuf *sb, const char *msg,
 732                            const char *line_separator)
 733 {
 734         int first = 1;
 735
 736         for (;;) {
 737                 const char *line = msg;
 738                 int linelen = get_one_line(line);
 739
 740                 msg += linelen;
 741                 if (!linelen || is_empty_line(line, &linelen))
 742                         break;
 743
 744                 if (!sb)
 745                         continue;
 746                 strbuf_grow(sb, linelen + 2);
 747                 if (!first)
 748                         strbuf_addstr(sb, line_separator);
 749                 strbuf_add(sb, line, linelen);
 750                 first = 0;
 751         }
 752         return msg;
 753 }
 754
 755 static void parse_commit_message(struct format_commit_context *c)
 756 {
 757         const char *msg = c->message + c->message_off;
 758         const char *start = c->message;
 759
 760         msg = skip_empty_lines(msg);
 761         c->subject_off = msg - start;
 762
 763         msg = format_subject(NULL, msg, NULL);
 764         msg = skip_empty_lines(msg);
 765         c->body_off = msg - start;
 766
 767         c->commit_message_parsed = 1;
 768 }
 769
 770 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 771 {
 772         struct name_decoration *d;
 773         const char *prefix = " (";
 774
 775         load_ref_decorations(DECORATE_SHORT_REFS);
 776         d = lookup_decoration(&name_decoration, &commit->object);
 777         while (d) {
 778                 strbuf_addstr(sb, prefix);
 779                 prefix = ", ";
 780                 strbuf_addstr(sb, d->name);
 781                 d = d->next;
 782         }
 783         if (prefix[0] == ',')
 784                 strbuf_addch(sb, ')');
 785 }
 786
 787 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 788                         size_t width, size_t indent1, size_t indent2)
 789 {
 790         struct strbuf tmp = STRBUF_INIT;
 791
 792         if (pos)
 793                 strbuf_add(&tmp, sb->buf, pos);
 794         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 795                                 (int) indent1, (int) indent2, (int) width);
 796         strbuf_swap(&tmp, sb);
 797         strbuf_release(&tmp);
 798 }
 799
 800 static void rewrap_message_tail(struct strbuf *sb,
 801                                 struct format_commit_context *c,
 802                                 size_t new_width, size_t new_indent1,
 803                                 size_t new_indent2)
 804 {
 805         if (c->width == new_width && c->indent1 == new_indent1 &&
 806             c->indent2 == new_indent2)
 807                 return;
 808         if (c->wrap_start < sb->len)
 809                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 810         c->wrap_start = sb->len;
 811         c->width = new_width;
 812         c->indent1 = new_indent1;
 813         c->indent2 = new_indent2;
 814 }
 815
 816 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 817                                 void *context)
 818 {
 819         struct format_commit_context *c = context;
 820         const struct commit *commit = c->commit;
 821         const char *msg = c->message;
 822         struct commit_list *p;
 823         int h1, h2;
 824
 825         /* these are independent of the commit */
 826         switch (placeholder[0]) {
 827         case 'C':
 828                 if (placeholder[1] == '(') {
 829                         const char *end = strchr(placeholder + 2, ')');
 830                         char color[COLOR_MAXLEN];
 831                         if (!end)
 832                                 return 0;
 833                         color_parse_mem(placeholder + 2,
 834                                         end - (placeholder + 2),
 835                                         "--pretty format", color);
 836                         strbuf_addstr(sb, color);
 837                         return end - placeholder + 1;
 838                 }
 839                 if (!prefixcmp(placeholder + 1, "red")) {
 840                         strbuf_addstr(sb, GIT_COLOR_RED);
 841                         return 4;
 842                 } else if (!prefixcmp(placeholder + 1, "green")) {
 843                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 844                         return 6;
 845                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 846                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 847                         return 5;
 848                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 849                         strbuf_addstr(sb, GIT_COLOR_RESET);
 850                         return 6;
 851                 } else
 852                         return 0;
 853         case 'n':               /* newline */
 854                 strbuf_addch(sb, '\n');
 855                 return 1;
 856         case 'x':
 857                 /* %x00 == NUL, %x0a == LF, etc. */
 858                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 859                     h1 <= 16 &&
 860                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 861                     h2 <= 16) {
 862                         strbuf_addch(sb, (h1<<4)|h2);
 863                         return 3;
 864                 } else
 865                         return 0;
 866         case 'w':
 867                 if (placeholder[1] == '(') {
 868                         unsigned long width = 0, indent1 = 0, indent2 = 0;
 869                         char *next;
 870                         const char *start = placeholder + 2;
 871                         const char *end = strchr(start, ')');
 872                         if (!end)
 873                                 return 0;
 874                         if (end > start) {
 875                                 width = strtoul(start, &next, 10);
 876                                 if (*next == ',') {
 877                                         indent1 = strtoul(next + 1, &next, 10);
 878                                         if (*next == ',') {
 879                                                 indent2 = strtoul(next + 1,
 880                                                                  &next, 10);
 881                                         }
 882                                 }
 883                                 if (*next != ')')
 884                                         return 0;
 885                         }
 886                         rewrap_message_tail(sb, c, width, indent1, indent2);
 887                         return end - placeholder + 1;
 888                 } else
 889                         return 0;
 890         }
 891
 892         /* these depend on the commit */
 893         if (!commit->object.parsed)
 894                 parse_object(commit->object.sha1);
 895
 896         switch (placeholder[0]) {
 897         case 'H':               /* commit hash */
 898                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 899                 return 1;
 900         case 'h':               /* abbreviated commit hash */
 901                 if (add_again(sb, &c->abbrev_commit_hash))
 902                         return 1;
 903                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 904                                                      c->pretty_ctx->abbrev));
 905                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 906                 return 1;
 907         case 'T':               /* tree hash */
 908                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 909                 return 1;
 910         case 't':               /* abbreviated tree hash */
 911                 if (add_again(sb, &c->abbrev_tree_hash))
 912                         return 1;
 913                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 914                                                      c->pretty_ctx->abbrev));
 915                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 916                 return 1;
 917         case 'P':               /* parent hashes */
 918                 for (p = commit->parents; p; p = p->next) {
 919                         if (p != commit->parents)
 920                                 strbuf_addch(sb, ' ');
 921                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 922                 }
 923                 return 1;
 924         case 'p':               /* abbreviated parent hashes */
 925                 if (add_again(sb, &c->abbrev_parent_hashes))
 926                         return 1;
 927                 for (p = commit->parents; p; p = p->next) {
 928                         if (p != commit->parents)
 929                                 strbuf_addch(sb, ' ');
 930                         strbuf_addstr(sb, find_unique_abbrev(
 931                                         p->item->object.sha1,
 932                                         c->pretty_ctx->abbrev));
 933                 }
 934                 c->abbrev_parent_hashes.len = sb->len -
 935                                               c->abbrev_parent_hashes.off;
 936                 return 1;
 937         case 'm':               /* left/right/bottom */
 938                 strbuf_addstr(sb, get_revision_mark(NULL, commit));
 939                 return 1;
 940         case 'd':
 941                 format_decoration(sb, commit);
 942                 return 1;
 943         case 'g':               /* reflog info */
 944                 switch(placeholder[1]) {
 945                 case 'd':       /* reflog selector */
 946                 case 'D':
 947                         if (c->pretty_ctx->reflog_info)
 948                                 get_reflog_selector(sb,
 949                                                     c->pretty_ctx->reflog_info,
 950                                                     c->pretty_ctx->date_mode,
 951                                                     (placeholder[1] == 'd'));
 952                         return 2;
 953                 case 's':       /* reflog message */
 954                         if (c->pretty_ctx->reflog_info)
 955                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
 956                         return 2;
 957                 }
 958                 return 0;       /* unknown %g placeholder */
 959         case 'N':
 960                 if (c->pretty_ctx->show_notes) {
 961                         format_display_notes(commit->object.sha1, sb,
 962                                     get_log_output_encoding(), 0);
 963                         return 1;
 964                 }
 965                 return 0;
 966         }
 967
 968         /* For the rest we have to parse the commit header. */
 969         if (!c->commit_header_parsed)
 970                 parse_commit_header(c);
 971
 972         switch (placeholder[0]) {
 973         case 'a':       /* author ... */
 974                 return format_person_part(sb, placeholder[1],
 975                                    msg + c->author.off, c->author.len,
 976                                    c->pretty_ctx->date_mode);
 977         case 'c':       /* committer ... */
 978                 return format_person_part(sb, placeholder[1],
 979                                    msg + c->committer.off, c->committer.len,
 980                                    c->pretty_ctx->date_mode);
 981         case 'e':       /* encoding */
 982                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 983                 return 1;
 984         case 'B':       /* raw body */
 985                 /* message_off is always left at the initial newline */
 986                 strbuf_addstr(sb, msg + c->message_off + 1);
 987                 return 1;
 988         }
 989
 990         /* Now we need to parse the commit message. */
 991         if (!c->commit_message_parsed)
 992                 parse_commit_message(c);
 993
 994         switch (placeholder[0]) {
 995         case 's':       /* subject */
 996                 format_subject(sb, msg + c->subject_off, " ");
 997                 return 1;
 998         case 'f':       /* sanitized subject */
 999                 format_sanitized_subject(sb, msg + c->subject_off);
1000                 return 1;
1001         case 'b':       /* body */
1002                 strbuf_addstr(sb, msg + c->body_off);
1003                 return 1;
1004         }
1005         return 0;       /* unknown placeholder */
1006 }
1007
1008 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1009                                  void *context)
1010 {
1011         int consumed;
1012         size_t orig_len;
1013         enum {
1014                 NO_MAGIC,
1015                 ADD_LF_BEFORE_NON_EMPTY,
1016                 DEL_LF_BEFORE_EMPTY,
1017                 ADD_SP_BEFORE_NON_EMPTY
1018         } magic = NO_MAGIC;
1019
1020         switch (placeholder[0]) {
1021         case '-':
1022                 magic = DEL_LF_BEFORE_EMPTY;
1023                 break;
1024         case '+':
1025                 magic = ADD_LF_BEFORE_NON_EMPTY;
1026                 break;
1027         case ' ':
1028                 magic = ADD_SP_BEFORE_NON_EMPTY;
1029                 break;
1030         default:
1031                 break;
1032         }
1033         if (magic != NO_MAGIC)
1034                 placeholder++;
1035
1036         orig_len = sb->len;
1037         consumed = format_commit_one(sb, placeholder, context);
1038         if (magic == NO_MAGIC)
1039                 return consumed;
1040
1041         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1042                 while (sb->len && sb->buf[sb->len - 1] == '\n')
1043                         strbuf_setlen(sb, sb->len - 1);
1044         } else if (orig_len != sb->len) {
1045                 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1046                         strbuf_insert(sb, orig_len, "\n", 1);
1047                 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1048                         strbuf_insert(sb, orig_len, " ", 1);
1049         }
1050         return consumed + 1;
1051 }
1052
1053 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1054                                    void *context)
1055 {
1056         struct userformat_want *w = context;
1057
1058         if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1059                 placeholder++;
1060
1061         switch (*placeholder) {
1062         case 'N':
1063                 w->notes = 1;
1064                 break;
1065         }
1066         return 0;
1067 }
1068
1069 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1070 {
1071         struct strbuf dummy = STRBUF_INIT;
1072
1073         if (!fmt) {
1074                 if (!user_format)
1075                         return;
1076                 fmt = user_format;
1077         }
1078         strbuf_expand(&dummy, fmt, userformat_want_item, w);
1079         strbuf_release(&dummy);
1080 }
1081
1082 void format_commit_message(const struct commit *commit,
1083                            const char *format, struct strbuf *sb,
1084                            const struct pretty_print_context *pretty_ctx)
1085 {
1086         struct format_commit_context context;
1087         static const char utf8[] = "UTF-8";
1088         const char *enc;
1089         const char *output_enc = pretty_ctx->output_encoding;
1090
1091         memset(&context, 0, sizeof(context));
1092         context.commit = commit;
1093         context.pretty_ctx = pretty_ctx;
1094         context.wrap_start = sb->len;
1095         context.message = commit->buffer;
1096         if (output_enc) {
1097                 enc = get_header(commit, "encoding");
1098                 enc = enc ? enc : utf8;
1099                 if (strcmp(enc, output_enc))
1100                         context.message = logmsg_reencode(commit, output_enc);
1101         }
1102
1103         strbuf_expand(sb, format, format_commit_item, &context);
1104         rewrap_message_tail(sb, &context, 0, 0, 0);
1105
1106         if (context.message != commit->buffer)
1107                 free(context.message);
1108 }
1109
1110 static void pp_header(enum cmit_fmt fmt,
1111                       int abbrev,
1112                       enum date_mode dmode,
1113                       const char *encoding,
1114                       const struct commit *commit,
1115                       const char **msg_p,
1116                       struct strbuf *sb)
1117 {
1118         int parents_shown = 0;
1119
1120         for (;;) {
1121                 const char *line = *msg_p;
1122                 int linelen = get_one_line(*msg_p);
1123
1124                 if (!linelen)
1125                         return;
1126                 *msg_p += linelen;
1127
1128                 if (linelen == 1)
1129                         /* End of header */
1130                         return;
1131
1132                 if (fmt == CMIT_FMT_RAW) {
1133                         strbuf_add(sb, line, linelen);
1134                         continue;
1135                 }
1136
1137                 if (!memcmp(line, "parent ", 7)) {
1138                         if (linelen != 48)
1139                                 die("bad parent line in commit");
1140                         continue;
1141                 }
1142
1143                 if (!parents_shown) {
1144                         struct commit_list *parent;
1145                         int num;
1146                         for (parent = commit->parents, num = 0;
1147                              parent;
1148                              parent = parent->next, num++)
1149                                 ;
1150                         /* with enough slop */
1151                         strbuf_grow(sb, num * 50 + 20);
1152                         add_merge_info(fmt, sb, commit, abbrev);
1153                         parents_shown = 1;
1154                 }
1155
1156                 /*
1157                  * MEDIUM == DEFAULT shows only author with dates.
1158                  * FULL shows both authors but not dates.
1159                  * FULLER shows both authors and dates.
1160                  */
1161                 if (!memcmp(line, "author ", 7)) {
1162                         strbuf_grow(sb, linelen + 80);
1163                         pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
1164                 }
1165                 if (!memcmp(line, "committer ", 10) &&
1166                     (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
1167                         strbuf_grow(sb, linelen + 80);
1168                         pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
1169                 }
1170         }
1171 }
1172
1173 void pp_title_line(enum cmit_fmt fmt,
1174                    const char **msg_p,
1175                    struct strbuf *sb,
1176                    const char *subject,
1177                    const char *after_subject,
1178                    const char *encoding,
1179                    int need_8bit_cte)
1180 {
1181         struct strbuf title;
1182
1183         strbuf_init(&title, 80);
1184         *msg_p = format_subject(&title, *msg_p, " ");
1185
1186         strbuf_grow(sb, title.len + 1024);
1187         if (subject) {
1188                 strbuf_addstr(sb, subject);
1189                 add_rfc2047(sb, title.buf, title.len, encoding);
1190         } else {
1191                 strbuf_addbuf(sb, &title);
1192         }
1193         strbuf_addch(sb, '\n');
1194
1195         if (need_8bit_cte > 0) {
1196                 const char *header_fmt =
1197                         "MIME-Version: 1.0\n"
1198                         "Content-Type: text/plain; charset=%s\n"
1199                         "Content-Transfer-Encoding: 8bit\n";
1200                 strbuf_addf(sb, header_fmt, encoding);
1201         }
1202         if (after_subject) {
1203                 strbuf_addstr(sb, after_subject);
1204         }
1205         if (fmt == CMIT_FMT_EMAIL) {
1206                 strbuf_addch(sb, '\n');
1207         }
1208         strbuf_release(&title);
1209 }
1210
1211 void pp_remainder(enum cmit_fmt fmt,
1212                   const char **msg_p,
1213                   struct strbuf *sb,
1214                   int indent)
1215 {
1216         int first = 1;
1217         for (;;) {
1218                 const char *line = *msg_p;
1219                 int linelen = get_one_line(line);
1220                 *msg_p += linelen;
1221
1222                 if (!linelen)
1223                         break;
1224
1225                 if (is_empty_line(line, &linelen)) {
1226                         if (first)
1227                                 continue;
1228                         if (fmt == CMIT_FMT_SHORT)
1229                                 break;
1230                 }
1231                 first = 0;
1232
1233                 strbuf_grow(sb, linelen + indent + 20);
1234                 if (indent) {
1235                         memset(sb->buf + sb->len, ' ', indent);
1236                         strbuf_setlen(sb, sb->len + indent);
1237                 }
1238                 strbuf_add(sb, line, linelen);
1239                 strbuf_addch(sb, '\n');
1240         }
1241 }
1242
1243 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1244 {
1245         const char *encoding;
1246
1247         encoding = get_log_output_encoding();
1248         if (encoding_p)
1249                 *encoding_p = encoding;
1250         return logmsg_reencode(commit, encoding);
1251 }
1252
1253 void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
1254                          struct strbuf *sb,
1255                          const struct pretty_print_context *context)
1256 {
1257         unsigned long beginning_of_body;
1258         int indent = 4;
1259         const char *msg = commit->buffer;
1260         char *reencoded;
1261         const char *encoding;
1262         int need_8bit_cte = context->need_8bit_cte;
1263
1264         if (fmt == CMIT_FMT_USERFORMAT) {
1265                 format_commit_message(commit, user_format, sb, context);
1266                 return;
1267         }
1268
1269         reencoded = reencode_commit_message(commit, &encoding);
1270         if (reencoded) {
1271                 msg = reencoded;
1272         }
1273
1274         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1275                 indent = 0;
1276
1277         /*
1278          * We need to check and emit Content-type: to mark it
1279          * as 8-bit if we haven't done so.
1280          */
1281         if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1282                 int i, ch, in_body;
1283
1284                 for (in_body = i = 0; (ch = msg[i]); i++) {
1285                         if (!in_body) {
1286                                 /* author could be non 7-bit ASCII but
1287                                  * the log may be so; skip over the
1288                                  * header part first.
1289                                  */
1290                                 if (ch == '\n' && msg[i+1] == '\n')
1291                                         in_body = 1;
1292                         }
1293                         else if (non_ascii(ch)) {
1294                                 need_8bit_cte = 1;
1295                                 break;
1296                         }
1297                 }
1298         }
1299
1300         pp_header(fmt, context->abbrev, context->date_mode, encoding,
1301                   commit, &msg, sb);
1302         if (fmt != CMIT_FMT_ONELINE && !context->subject) {
1303                 strbuf_addch(sb, '\n');
1304         }
1305
1306         /* Skip excess blank lines at the beginning of body, if any... */
1307         msg = skip_empty_lines(msg);
1308
1309         /* These formats treat the title line specially. */
1310         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1311                 pp_title_line(fmt, &msg, sb, context->subject,
1312                               context->after_subject, encoding, need_8bit_cte);
1313
1314         beginning_of_body = sb->len;
1315         if (fmt != CMIT_FMT_ONELINE)
1316                 pp_remainder(fmt, &msg, sb, indent);
1317         strbuf_rtrim(sb);
1318
1319         /* Make sure there is an EOLN for the non-oneline case */
1320         if (fmt != CMIT_FMT_ONELINE)
1321                 strbuf_addch(sb, '\n');
1322
1323         /*
1324          * The caller may append additional body text in e-mail
1325          * format.  Make sure we did not strip the blank line
1326          * between the header and the body.
1327          */
1328         if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1329                 strbuf_addch(sb, '\n');
1330
1331         if (context->show_notes)
1332                 format_display_notes(commit->object.sha1, sb, encoding,
1333                                      NOTES_SHOW_HEADER | NOTES_INDENT);
1334
1335         free(reencoded);
1336 }