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 } *commit_formats;
  19 static size_t commit_formats_len;
  20 static struct cmt_fmt_map *find_commit_format(const char *sought);
  21
  22 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  23 {
  24         free(user_format);
  25         user_format = xstrdup(cp);
  26         if (is_tformat)
  27                 rev->use_terminator = 1;
  28         rev->commit_format = CMIT_FMT_USERFORMAT;
  29 }
  30
  31 static void setup_commit_formats(void)
  32 {
  33         struct cmt_fmt_map builtin_formats[] = {
  34                 { "raw",        CMIT_FMT_RAW,           0 },
  35                 { "medium",     CMIT_FMT_MEDIUM,        0 },
  36                 { "short",      CMIT_FMT_SHORT,         0 },
  37                 { "email",      CMIT_FMT_EMAIL,         0 },
  38                 { "fuller",     CMIT_FMT_FULLER,        0 },
  39                 { "full",       CMIT_FMT_FULL,          0 },
  40                 { "oneline",    CMIT_FMT_ONELINE,       1 }
  41         };
  42         commit_formats_len = ARRAY_SIZE(builtin_formats);
  43         commit_formats = xmalloc(commit_formats_len *
  44                                  sizeof(*builtin_formats));
  45         memcpy(commit_formats, builtin_formats,
  46                sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  47 }
  48
  49 static struct cmt_fmt_map *find_commit_format(const char *sought)
  50 {
  51         struct cmt_fmt_map *found = NULL;
  52         size_t found_match_len = 0;
  53         int i;
  54
  55         if (!commit_formats)
  56                 setup_commit_formats();
  57
  58         for (i = 0; i < commit_formats_len; i++) {
  59                 size_t match_len;
  60
  61                 if (prefixcmp(commit_formats[i].name, sought))
  62                         continue;
  63
  64                 match_len = strlen(commit_formats[i].name);
  65                 if (found == NULL || found_match_len > match_len) {
  66                         found = &commit_formats[i];
  67                         found_match_len = match_len;
  68                 }
  69         }
  70         return found;
  71 }
  72
  73 void get_commit_format(const char *arg, struct rev_info *rev)
  74 {
  75         struct cmt_fmt_map *commit_format;
  76
  77         rev->use_terminator = 0;
  78         if (!arg || !*arg) {
  79                 rev->commit_format = CMIT_FMT_DEFAULT;
  80                 return;
  81         }
  82         if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
  83                 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
  84                 return;
  85         }
  86
  87         if (strchr(arg, '%')) {
  88                 save_user_format(rev, arg, 1);
  89                 return;
  90         }
  91
  92         commit_format = find_commit_format(arg);
  93         if (!commit_format)
  94                 die("invalid --pretty format: %s", arg);
  95
  96         rev->commit_format = commit_format->format;
  97         rev->use_terminator = commit_format->is_tformat;
  98 }
  99
 100 /*
 101  * Generic support for pretty-printing the header
 102  */
 103 static int get_one_line(const char *msg)
 104 {
 105         int ret = 0;
 106
 107         for (;;) {
 108                 char c = *msg++;
 109                 if (!c)
 110                         break;
 111                 ret++;
 112                 if (c == '\n')
 113                         break;
 114         }
 115         return ret;
 116 }
 117
 118 /* High bit set, or ISO-2022-INT */
 119 static int non_ascii(int ch)
 120 {
 121         return !isascii(ch) || ch == '\033';
 122 }
 123
 124 int has_non_ascii(const char *s)
 125 {
 126         int ch;
 127         if (!s)
 128                 return 0;
 129         while ((ch = *s++) != '\0') {
 130                 if (non_ascii(ch))
 131                         return 1;
 132         }
 133         return 0;
 134 }
 135
 136 static int is_rfc2047_special(char ch)
 137 {
 138         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 139 }
 140
 141 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 142                        const char *encoding)
 143 {
 144         int i, last;
 145
 146         for (i = 0; i < len; i++) {
 147                 int ch = line[i];
 148                 if (non_ascii(ch))
 149                         goto needquote;
 150                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 151                         goto needquote;
 152         }
 153         strbuf_add(sb, line, len);
 154         return;
 155
 156 needquote:
 157         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 158         strbuf_addf(sb, "=?%s?q?", encoding);
 159         for (i = last = 0; i < len; i++) {
 160                 unsigned ch = line[i] & 0xFF;
 161                 /*
 162                  * We encode ' ' using '=20' even though rfc2047
 163                  * allows using '_' for readability.  Unfortunately,
 164                  * many programs do not understand this and just
 165                  * leave the underscore in place.
 166                  */
 167                 if (is_rfc2047_special(ch) || ch == ' ') {
 168                         strbuf_add(sb, line + last, i - last);
 169                         strbuf_addf(sb, "=%02X", ch);
 170                         last = i + 1;
 171                 }
 172         }
 173         strbuf_add(sb, line + last, len - last);
 174         strbuf_addstr(sb, "?=");
 175 }
 176
 177 void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 178                   const char *line, enum date_mode dmode,
 179                   const char *encoding)
 180 {
 181         char *date;
 182         int namelen;
 183         unsigned long time;
 184         int tz;
 185
 186         if (fmt == CMIT_FMT_ONELINE)
 187                 return;
 188         date = strchr(line, '>');
 189         if (!date)
 190                 return;
 191         namelen = ++date - line;
 192         time = strtoul(date, &date, 10);
 193         tz = strtol(date, NULL, 10);
 194
 195         if (fmt == CMIT_FMT_EMAIL) {
 196                 char *name_tail = strchr(line, '<');
 197                 int display_name_length;
 198                 if (!name_tail)
 199                         return;
 200                 while (line < name_tail && isspace(name_tail[-1]))
 201                         name_tail--;
 202                 display_name_length = name_tail - line;
 203                 strbuf_addstr(sb, "From: ");
 204                 add_rfc2047(sb, line, display_name_length, encoding);
 205                 strbuf_add(sb, name_tail, namelen - display_name_length);
 206                 strbuf_addch(sb, '\n');
 207         } else {
 208                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 209                               (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 210                               "    ", namelen, line);
 211         }
 212         switch (fmt) {
 213         case CMIT_FMT_MEDIUM:
 214                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 215                 break;
 216         case CMIT_FMT_EMAIL:
 217                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 218                 break;
 219         case CMIT_FMT_FULLER:
 220                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 221                 break;
 222         default:
 223                 /* notin' */
 224                 break;
 225         }
 226 }
 227
 228 static int is_empty_line(const char *line, int *len_p)
 229 {
 230         int len = *len_p;
 231         while (len && isspace(line[len-1]))
 232                 len--;
 233         *len_p = len;
 234         return !len;
 235 }
 236
 237 static const char *skip_empty_lines(const char *msg)
 238 {
 239         for (;;) {
 240                 int linelen = get_one_line(msg);
 241                 int ll = linelen;
 242                 if (!linelen)
 243                         break;
 244                 if (!is_empty_line(msg, &ll))
 245                         break;
 246                 msg += linelen;
 247         }
 248         return msg;
 249 }
 250
 251 static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 252                         const struct commit *commit, int abbrev)
 253 {
 254         struct commit_list *parent = commit->parents;
 255
 256         if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 257             !parent || !parent->next)
 258                 return;
 259
 260         strbuf_addstr(sb, "Merge:");
 261
 262         while (parent) {
 263                 struct commit *p = parent->item;
 264                 const char *hex = NULL;
 265                 if (abbrev)
 266                         hex = find_unique_abbrev(p->object.sha1, abbrev);
 267                 if (!hex)
 268                         hex = sha1_to_hex(p->object.sha1);
 269                 parent = parent->next;
 270
 271                 strbuf_addf(sb, " %s", hex);
 272         }
 273         strbuf_addch(sb, '\n');
 274 }
 275
 276 static char *get_header(const struct commit *commit, const char *key)
 277 {
 278         int key_len = strlen(key);
 279         const char *line = commit->buffer;
 280
 281         for (;;) {
 282                 const char *eol = strchr(line, '\n'), *next;
 283
 284                 if (line == eol)
 285                         return NULL;
 286                 if (!eol) {
 287                         eol = line + strlen(line);
 288                         next = NULL;
 289                 } else
 290                         next = eol + 1;
 291                 if (eol - line > key_len &&
 292                     !strncmp(line, key, key_len) &&
 293                     line[key_len] == ' ') {
 294                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 295                 }
 296                 line = next;
 297         }
 298 }
 299
 300 static char *replace_encoding_header(char *buf, const char *encoding)
 301 {
 302         struct strbuf tmp = STRBUF_INIT;
 303         size_t start, len;
 304         char *cp = buf;
 305
 306         /* guess if there is an encoding header before a \n\n */
 307         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 308                 cp = strchr(cp, '\n');
 309                 if (!cp || *++cp == '\n')
 310                         return buf;
 311         }
 312         start = cp - buf;
 313         cp = strchr(cp, '\n');
 314         if (!cp)
 315                 return buf; /* should not happen but be defensive */
 316         len = cp + 1 - (buf + start);
 317
 318         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 319         if (is_encoding_utf8(encoding)) {
 320                 /* we have re-coded to UTF-8; drop the header */
 321                 strbuf_remove(&tmp, start, len);
 322         } else {
 323                 /* just replaces XXXX in 'encoding XXXX\n' */
 324                 strbuf_splice(&tmp, start + strlen("encoding "),
 325                                           len - strlen("encoding \n"),
 326                                           encoding, strlen(encoding));
 327         }
 328         return strbuf_detach(&tmp, NULL);
 329 }
 330
 331 static char *logmsg_reencode(const struct commit *commit,
 332                              const char *output_encoding)
 333 {
 334         static const char *utf8 = "UTF-8";
 335         const char *use_encoding;
 336         char *encoding;
 337         char *out;
 338
 339         if (!*output_encoding)
 340                 return NULL;
 341         encoding = get_header(commit, "encoding");
 342         use_encoding = encoding ? encoding : utf8;
 343         if (!strcmp(use_encoding, output_encoding))
 344                 if (encoding) /* we'll strip encoding header later */
 345                         out = xstrdup(commit->buffer);
 346                 else
 347                         return NULL; /* nothing to do */
 348         else
 349                 out = reencode_string(commit->buffer,
 350                                       output_encoding, use_encoding);
 351         if (out)
 352                 out = replace_encoding_header(out, output_encoding);
 353
 354         free(encoding);
 355         return out;
 356 }
 357
 358 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 359 {
 360         static struct string_list *mail_map;
 361         if (!mail_map) {
 362                 mail_map = xcalloc(1, sizeof(*mail_map));
 363                 read_mailmap(mail_map, NULL);
 364         }
 365         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 366 }
 367
 368 static size_t format_person_part(struct strbuf *sb, char part,
 369                                  const char *msg, int len, enum date_mode dmode)
 370 {
 371         /* currently all placeholders have same length */
 372         const int placeholder_len = 2;
 373         int start, end, tz = 0;
 374         unsigned long date = 0;
 375         char *ep;
 376         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 377         char person_name[1024];
 378         char person_mail[1024];
 379
 380         /* advance 'end' to point to email start delimiter */
 381         for (end = 0; end < len && msg[end] != '<'; end++)
 382                 ; /* do nothing */
 383
 384         /*
 385          * When end points at the '<' that we found, it should have
 386          * matching '>' later, which means 'end' must be strictly
 387          * below len - 1.
 388          */
 389         if (end >= len - 2)
 390                 goto skip;
 391
 392         /* Seek for both name and email part */
 393         name_start = msg;
 394         name_end = msg+end;
 395         while (name_end > name_start && isspace(*(name_end-1)))
 396                 name_end--;
 397         mail_start = msg+end+1;
 398         mail_end = mail_start;
 399         while (mail_end < msg_end && *mail_end != '>')
 400                 mail_end++;
 401         if (mail_end == msg_end)
 402                 goto skip;
 403         end = mail_end-msg;
 404
 405         if (part == 'N' || part == 'E') { /* mailmap lookup */
 406                 strlcpy(person_name, name_start, name_end-name_start+1);
 407                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 408                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 409                 name_start = person_name;
 410                 name_end = name_start + strlen(person_name);
 411                 mail_start = person_mail;
 412                 mail_end = mail_start +  strlen(person_mail);
 413         }
 414         if (part == 'n' || part == 'N') {       /* name */
 415                 strbuf_add(sb, name_start, name_end-name_start);
 416                 return placeholder_len;
 417         }
 418         if (part == 'e' || part == 'E') {       /* email */
 419                 strbuf_add(sb, mail_start, mail_end-mail_start);
 420                 return placeholder_len;
 421         }
 422
 423         /* advance 'start' to point to date start delimiter */
 424         for (start = end + 1; start < len && isspace(msg[start]); start++)
 425                 ; /* do nothing */
 426         if (start >= len)
 427                 goto skip;
 428         date = strtoul(msg + start, &ep, 10);
 429         if (msg + start == ep)
 430                 goto skip;
 431
 432         if (part == 't') {      /* date, UNIX timestamp */
 433                 strbuf_add(sb, msg + start, ep - (msg + start));
 434                 return placeholder_len;
 435         }
 436
 437         /* parse tz */
 438         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 439                 ; /* do nothing */
 440         if (start + 1 < len) {
 441                 tz = strtoul(msg + start + 1, NULL, 10);
 442                 if (msg[start] == '-')
 443                         tz = -tz;
 444         }
 445
 446         switch (part) {
 447         case 'd':       /* date */
 448                 strbuf_addstr(sb, show_date(date, tz, dmode));
 449                 return placeholder_len;
 450         case 'D':       /* date, RFC2822 style */
 451                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 452                 return placeholder_len;
 453         case 'r':       /* date, relative */
 454                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 455                 return placeholder_len;
 456         case 'i':       /* date, ISO 8601 */
 457                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 458                 return placeholder_len;
 459         }
 460
 461 skip:
 462         /*
 463          * bogus commit, 'sb' cannot be updated, but we still need to
 464          * compute a valid return value.
 465          */
 466         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 467             || part == 'D' || part == 'r' || part == 'i')
 468                 return placeholder_len;
 469
 470         return 0; /* unknown placeholder */
 471 }
 472
 473 struct chunk {
 474         size_t off;
 475         size_t len;
 476 };
 477
 478 struct format_commit_context {
 479         const struct commit *commit;
 480         const struct pretty_print_context *pretty_ctx;
 481         unsigned commit_header_parsed:1;
 482         unsigned commit_message_parsed:1;
 483         size_t width, indent1, indent2;
 484
 485         /* These offsets are relative to the start of the commit message. */
 486         struct chunk author;
 487         struct chunk committer;
 488         struct chunk encoding;
 489         size_t message_off;
 490         size_t subject_off;
 491         size_t body_off;
 492
 493         /* The following ones are relative to the result struct strbuf. */
 494         struct chunk abbrev_commit_hash;
 495         struct chunk abbrev_tree_hash;
 496         struct chunk abbrev_parent_hashes;
 497         size_t wrap_start;
 498 };
 499
 500 static int add_again(struct strbuf *sb, struct chunk *chunk)
 501 {
 502         if (chunk->len) {
 503                 strbuf_adddup(sb, chunk->off, chunk->len);
 504                 return 1;
 505         }
 506
 507         /*
 508          * We haven't seen this chunk before.  Our caller is surely
 509          * going to add it the hard way now.  Remember the most likely
 510          * start of the to-be-added chunk: the current end of the
 511          * struct strbuf.
 512          */
 513         chunk->off = sb->len;
 514         return 0;
 515 }
 516
 517 static void parse_commit_header(struct format_commit_context *context)
 518 {
 519         const char *msg = context->commit->buffer;
 520         int i;
 521
 522         for (i = 0; msg[i]; i++) {
 523                 int eol;
 524                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 525                         ; /* do nothing */
 526
 527                 if (i == eol) {
 528                         break;
 529                 } else if (!prefixcmp(msg + i, "author ")) {
 530                         context->author.off = i + 7;
 531                         context->author.len = eol - i - 7;
 532                 } else if (!prefixcmp(msg + i, "committer ")) {
 533                         context->committer.off = i + 10;
 534                         context->committer.len = eol - i - 10;
 535                 } else if (!prefixcmp(msg + i, "encoding ")) {
 536                         context->encoding.off = i + 9;
 537                         context->encoding.len = eol - i - 9;
 538                 }
 539                 i = eol;
 540         }
 541         context->message_off = i;
 542         context->commit_header_parsed = 1;
 543 }
 544
 545 static int istitlechar(char c)
 546 {
 547         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 548                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 549 }
 550
 551 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 552 {
 553         size_t trimlen;
 554         size_t start_len = sb->len;
 555         int space = 2;
 556
 557         for (; *msg && *msg != '\n'; msg++) {
 558                 if (istitlechar(*msg)) {
 559                         if (space == 1)
 560                                 strbuf_addch(sb, '-');
 561                         space = 0;
 562                         strbuf_addch(sb, *msg);
 563                         if (*msg == '.')
 564                                 while (*(msg+1) == '.')
 565                                         msg++;
 566                 } else
 567                         space |= 1;
 568         }
 569
 570         /* trim any trailing '.' or '-' characters */
 571         trimlen = 0;
 572         while (sb->len - trimlen > start_len &&
 573                 (sb->buf[sb->len - 1 - trimlen] == '.'
 574                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 575                 trimlen++;
 576         strbuf_remove(sb, sb->len - trimlen, trimlen);
 577 }
 578
 579 const char *format_subject(struct strbuf *sb, const char *msg,
 580                            const char *line_separator)
 581 {
 582         int first = 1;
 583
 584         for (;;) {
 585                 const char *line = msg;
 586                 int linelen = get_one_line(line);
 587
 588                 msg += linelen;
 589                 if (!linelen || is_empty_line(line, &linelen))
 590                         break;
 591
 592                 if (!sb)
 593                         continue;
 594                 strbuf_grow(sb, linelen + 2);
 595                 if (!first)
 596                         strbuf_addstr(sb, line_separator);
 597                 strbuf_add(sb, line, linelen);
 598                 first = 0;
 599         }
 600         return msg;
 601 }
 602
 603 static void parse_commit_message(struct format_commit_context *c)
 604 {
 605         const char *msg = c->commit->buffer + c->message_off;
 606         const char *start = c->commit->buffer;
 607
 608         msg = skip_empty_lines(msg);
 609         c->subject_off = msg - start;
 610
 611         msg = format_subject(NULL, msg, NULL);
 612         msg = skip_empty_lines(msg);
 613         c->body_off = msg - start;
 614
 615         c->commit_message_parsed = 1;
 616 }
 617
 618 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 619 {
 620         struct name_decoration *d;
 621         const char *prefix = " (";
 622
 623         load_ref_decorations(DECORATE_SHORT_REFS);
 624         d = lookup_decoration(&name_decoration, &commit->object);
 625         while (d) {
 626                 strbuf_addstr(sb, prefix);
 627                 prefix = ", ";
 628                 strbuf_addstr(sb, d->name);
 629                 d = d->next;
 630         }
 631         if (prefix[0] == ',')
 632                 strbuf_addch(sb, ')');
 633 }
 634
 635 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 636                         size_t width, size_t indent1, size_t indent2)
 637 {
 638         struct strbuf tmp = STRBUF_INIT;
 639
 640         if (pos)
 641                 strbuf_add(&tmp, sb->buf, pos);
 642         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 643                                 (int) indent1, (int) indent2, (int) width);
 644         strbuf_swap(&tmp, sb);
 645         strbuf_release(&tmp);
 646 }
 647
 648 static void rewrap_message_tail(struct strbuf *sb,
 649                                 struct format_commit_context *c,
 650                                 size_t new_width, size_t new_indent1,
 651                                 size_t new_indent2)
 652 {
 653         if (c->width == new_width && c->indent1 == new_indent1 &&
 654             c->indent2 == new_indent2)
 655                 return;
 656         if (c->wrap_start < sb->len)
 657                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 658         c->wrap_start = sb->len;
 659         c->width = new_width;
 660         c->indent1 = new_indent1;
 661         c->indent2 = new_indent2;
 662 }
 663
 664 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 665                                 void *context)
 666 {
 667         struct format_commit_context *c = context;
 668         const struct commit *commit = c->commit;
 669         const char *msg = commit->buffer;
 670         struct commit_list *p;
 671         int h1, h2;
 672
 673         /* these are independent of the commit */
 674         switch (placeholder[0]) {
 675         case 'C':
 676                 if (placeholder[1] == '(') {
 677                         const char *end = strchr(placeholder + 2, ')');
 678                         char color[COLOR_MAXLEN];
 679                         if (!end)
 680                                 return 0;
 681                         color_parse_mem(placeholder + 2,
 682                                         end - (placeholder + 2),
 683                                         "--pretty format", color);
 684                         strbuf_addstr(sb, color);
 685                         return end - placeholder + 1;
 686                 }
 687                 if (!prefixcmp(placeholder + 1, "red")) {
 688                         strbuf_addstr(sb, GIT_COLOR_RED);
 689                         return 4;
 690                 } else if (!prefixcmp(placeholder + 1, "green")) {
 691                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 692                         return 6;
 693                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 694                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 695                         return 5;
 696                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 697                         strbuf_addstr(sb, GIT_COLOR_RESET);
 698                         return 6;
 699                 } else
 700                         return 0;
 701         case 'n':               /* newline */
 702                 strbuf_addch(sb, '\n');
 703                 return 1;
 704         case 'x':
 705                 /* %x00 == NUL, %x0a == LF, etc. */
 706                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 707                     h1 <= 16 &&
 708                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 709                     h2 <= 16) {
 710                         strbuf_addch(sb, (h1<<4)|h2);
 711                         return 3;
 712                 } else
 713                         return 0;
 714         case 'w':
 715                 if (placeholder[1] == '(') {
 716                         unsigned long width = 0, indent1 = 0, indent2 = 0;
 717                         char *next;
 718                         const char *start = placeholder + 2;
 719                         const char *end = strchr(start, ')');
 720                         if (!end)
 721                                 return 0;
 722                         if (end > start) {
 723                                 width = strtoul(start, &next, 10);
 724                                 if (*next == ',') {
 725                                         indent1 = strtoul(next + 1, &next, 10);
 726                                         if (*next == ',') {
 727                                                 indent2 = strtoul(next + 1,
 728                                                                  &next, 10);
 729                                         }
 730                                 }
 731                                 if (*next != ')')
 732                                         return 0;
 733                         }
 734                         rewrap_message_tail(sb, c, width, indent1, indent2);
 735                         return end - placeholder + 1;
 736                 } else
 737                         return 0;
 738         }
 739
 740         /* these depend on the commit */
 741         if (!commit->object.parsed)
 742                 parse_object(commit->object.sha1);
 743
 744         switch (placeholder[0]) {
 745         case 'H':               /* commit hash */
 746                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 747                 return 1;
 748         case 'h':               /* abbreviated commit hash */
 749                 if (add_again(sb, &c->abbrev_commit_hash))
 750                         return 1;
 751                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 752                                                      DEFAULT_ABBREV));
 753                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 754                 return 1;
 755         case 'T':               /* tree hash */
 756                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 757                 return 1;
 758         case 't':               /* abbreviated tree hash */
 759                 if (add_again(sb, &c->abbrev_tree_hash))
 760                         return 1;
 761                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 762                                                      DEFAULT_ABBREV));
 763                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 764                 return 1;
 765         case 'P':               /* parent hashes */
 766                 for (p = commit->parents; p; p = p->next) {
 767                         if (p != commit->parents)
 768                                 strbuf_addch(sb, ' ');
 769                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 770                 }
 771                 return 1;
 772         case 'p':               /* abbreviated parent hashes */
 773                 if (add_again(sb, &c->abbrev_parent_hashes))
 774                         return 1;
 775                 for (p = commit->parents; p; p = p->next) {
 776                         if (p != commit->parents)
 777                                 strbuf_addch(sb, ' ');
 778                         strbuf_addstr(sb, find_unique_abbrev(
 779                                         p->item->object.sha1, DEFAULT_ABBREV));
 780                 }
 781                 c->abbrev_parent_hashes.len = sb->len -
 782                                               c->abbrev_parent_hashes.off;
 783                 return 1;
 784         case 'm':               /* left/right/bottom */
 785                 strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 786                                  ? '-'
 787                                  : (commit->object.flags & SYMMETRIC_LEFT)
 788                                  ? '<'
 789                                  : '>');
 790                 return 1;
 791         case 'd':
 792                 format_decoration(sb, commit);
 793                 return 1;
 794         case 'g':               /* reflog info */
 795                 switch(placeholder[1]) {
 796                 case 'd':       /* reflog selector */
 797                 case 'D':
 798                         if (c->pretty_ctx->reflog_info)
 799                                 get_reflog_selector(sb,
 800                                                     c->pretty_ctx->reflog_info,
 801                                                     c->pretty_ctx->date_mode,
 802                                                     (placeholder[1] == 'd'));
 803                         return 2;
 804                 case 's':       /* reflog message */
 805                         if (c->pretty_ctx->reflog_info)
 806                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
 807                         return 2;
 808                 }
 809                 return 0;       /* unknown %g placeholder */
 810         case 'N':
 811                 if (c->pretty_ctx->show_notes) {
 812                         format_display_notes(commit->object.sha1, sb,
 813                                     git_log_output_encoding ? git_log_output_encoding
 814                                                             : git_commit_encoding, 0);
 815                         return 1;
 816                 }
 817                 return 0;
 818         }
 819
 820         /* For the rest we have to parse the commit header. */
 821         if (!c->commit_header_parsed)
 822                 parse_commit_header(c);
 823
 824         switch (placeholder[0]) {
 825         case 'a':       /* author ... */
 826                 return format_person_part(sb, placeholder[1],
 827                                    msg + c->author.off, c->author.len,
 828                                    c->pretty_ctx->date_mode);
 829         case 'c':       /* committer ... */
 830                 return format_person_part(sb, placeholder[1],
 831                                    msg + c->committer.off, c->committer.len,
 832                                    c->pretty_ctx->date_mode);
 833         case 'e':       /* encoding */
 834                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 835                 return 1;
 836         }
 837
 838         /* Now we need to parse the commit message. */
 839         if (!c->commit_message_parsed)
 840                 parse_commit_message(c);
 841
 842         switch (placeholder[0]) {
 843         case 's':       /* subject */
 844                 format_subject(sb, msg + c->subject_off, " ");
 845                 return 1;
 846         case 'f':       /* sanitized subject */
 847                 format_sanitized_subject(sb, msg + c->subject_off);
 848                 return 1;
 849         case 'b':       /* body */
 850                 strbuf_addstr(sb, msg + c->body_off);
 851                 return 1;
 852         }
 853         return 0;       /* unknown placeholder */
 854 }
 855
 856 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 857                                  void *context)
 858 {
 859         int consumed;
 860         size_t orig_len;
 861         enum {
 862                 NO_MAGIC,
 863                 ADD_LF_BEFORE_NON_EMPTY,
 864                 DEL_LF_BEFORE_EMPTY,
 865         } magic = NO_MAGIC;
 866
 867         switch (placeholder[0]) {
 868         case '-':
 869                 magic = DEL_LF_BEFORE_EMPTY;
 870                 break;
 871         case '+':
 872                 magic = ADD_LF_BEFORE_NON_EMPTY;
 873                 break;
 874         default:
 875                 break;
 876         }
 877         if (magic != NO_MAGIC)
 878                 placeholder++;
 879
 880         orig_len = sb->len;
 881         consumed = format_commit_one(sb, placeholder, context);
 882         if (magic == NO_MAGIC)
 883                 return consumed;
 884
 885         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 886                 while (sb->len && sb->buf[sb->len - 1] == '\n')
 887                         strbuf_setlen(sb, sb->len - 1);
 888         } else if ((orig_len != sb->len) && magic == ADD_LF_BEFORE_NON_EMPTY) {
 889                 strbuf_insert(sb, orig_len, "\n", 1);
 890         }
 891         return consumed + 1;
 892 }
 893
 894 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
 895                                    void *context)
 896 {
 897         struct userformat_want *w = context;
 898
 899         if (*placeholder == '+' || *placeholder == '-')
 900                 placeholder++;
 901
 902         switch (*placeholder) {
 903         case 'N':
 904                 w->notes = 1;
 905                 break;
 906         }
 907         return 0;
 908 }
 909
 910 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
 911 {
 912         struct strbuf dummy = STRBUF_INIT;
 913
 914         if (!fmt) {
 915                 if (!user_format)
 916                         return;
 917                 fmt = user_format;
 918         }
 919         strbuf_expand(&dummy, user_format, userformat_want_item, w);
 920         strbuf_release(&dummy);
 921 }
 922
 923 void format_commit_message(const struct commit *commit,
 924                            const char *format, struct strbuf *sb,
 925                            const struct pretty_print_context *pretty_ctx)
 926 {
 927         struct format_commit_context context;
 928
 929         memset(&context, 0, sizeof(context));
 930         context.commit = commit;
 931         context.pretty_ctx = pretty_ctx;
 932         context.wrap_start = sb->len;
 933         strbuf_expand(sb, format, format_commit_item, &context);
 934         rewrap_message_tail(sb, &context, 0, 0, 0);
 935 }
 936
 937 static void pp_header(enum cmit_fmt fmt,
 938                       int abbrev,
 939                       enum date_mode dmode,
 940                       const char *encoding,
 941                       const struct commit *commit,
 942                       const char **msg_p,
 943                       struct strbuf *sb)
 944 {
 945         int parents_shown = 0;
 946
 947         for (;;) {
 948                 const char *line = *msg_p;
 949                 int linelen = get_one_line(*msg_p);
 950
 951                 if (!linelen)
 952                         return;
 953                 *msg_p += linelen;
 954
 955                 if (linelen == 1)
 956                         /* End of header */
 957                         return;
 958
 959                 if (fmt == CMIT_FMT_RAW) {
 960                         strbuf_add(sb, line, linelen);
 961                         continue;
 962                 }
 963
 964                 if (!memcmp(line, "parent ", 7)) {
 965                         if (linelen != 48)
 966                                 die("bad parent line in commit");
 967                         continue;
 968                 }
 969
 970                 if (!parents_shown) {
 971                         struct commit_list *parent;
 972                         int num;
 973                         for (parent = commit->parents, num = 0;
 974                              parent;
 975                              parent = parent->next, num++)
 976                                 ;
 977                         /* with enough slop */
 978                         strbuf_grow(sb, num * 50 + 20);
 979                         add_merge_info(fmt, sb, commit, abbrev);
 980                         parents_shown = 1;
 981                 }
 982
 983                 /*
 984                  * MEDIUM == DEFAULT shows only author with dates.
 985                  * FULL shows both authors but not dates.
 986                  * FULLER shows both authors and dates.
 987                  */
 988                 if (!memcmp(line, "author ", 7)) {
 989                         strbuf_grow(sb, linelen + 80);
 990                         pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
 991                 }
 992                 if (!memcmp(line, "committer ", 10) &&
 993                     (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
 994                         strbuf_grow(sb, linelen + 80);
 995                         pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
 996                 }
 997         }
 998 }
 999
1000 void pp_title_line(enum cmit_fmt fmt,
1001                    const char **msg_p,
1002                    struct strbuf *sb,
1003                    const char *subject,
1004                    const char *after_subject,
1005                    const char *encoding,
1006                    int need_8bit_cte)
1007 {
1008         const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
1009         struct strbuf title;
1010
1011         strbuf_init(&title, 80);
1012         *msg_p = format_subject(&title, *msg_p, line_separator);
1013
1014         strbuf_grow(sb, title.len + 1024);
1015         if (subject) {
1016                 strbuf_addstr(sb, subject);
1017                 add_rfc2047(sb, title.buf, title.len, encoding);
1018         } else {
1019                 strbuf_addbuf(sb, &title);
1020         }
1021         strbuf_addch(sb, '\n');
1022
1023         if (need_8bit_cte > 0) {
1024                 const char *header_fmt =
1025                         "MIME-Version: 1.0\n"
1026                         "Content-Type: text/plain; charset=%s\n"
1027                         "Content-Transfer-Encoding: 8bit\n";
1028                 strbuf_addf(sb, header_fmt, encoding);
1029         }
1030         if (after_subject) {
1031                 strbuf_addstr(sb, after_subject);
1032         }
1033         if (fmt == CMIT_FMT_EMAIL) {
1034                 strbuf_addch(sb, '\n');
1035         }
1036         strbuf_release(&title);
1037 }
1038
1039 void pp_remainder(enum cmit_fmt fmt,
1040                   const char **msg_p,
1041                   struct strbuf *sb,
1042                   int indent)
1043 {
1044         int first = 1;
1045         for (;;) {
1046                 const char *line = *msg_p;
1047                 int linelen = get_one_line(line);
1048                 *msg_p += linelen;
1049
1050                 if (!linelen)
1051                         break;
1052
1053                 if (is_empty_line(line, &linelen)) {
1054                         if (first)
1055                                 continue;
1056                         if (fmt == CMIT_FMT_SHORT)
1057                                 break;
1058                 }
1059                 first = 0;
1060
1061                 strbuf_grow(sb, linelen + indent + 20);
1062                 if (indent) {
1063                         memset(sb->buf + sb->len, ' ', indent);
1064                         strbuf_setlen(sb, sb->len + indent);
1065                 }
1066                 strbuf_add(sb, line, linelen);
1067                 strbuf_addch(sb, '\n');
1068         }
1069 }
1070
1071 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1072 {
1073         const char *encoding;
1074
1075         encoding = (git_log_output_encoding
1076                     ? git_log_output_encoding
1077                     : git_commit_encoding);
1078         if (!encoding)
1079                 encoding = "UTF-8";
1080         if (encoding_p)
1081                 *encoding_p = encoding;
1082         return logmsg_reencode(commit, encoding);
1083 }
1084
1085 void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
1086                          struct strbuf *sb,
1087                          const struct pretty_print_context *context)
1088 {
1089         unsigned long beginning_of_body;
1090         int indent = 4;
1091         const char *msg = commit->buffer;
1092         char *reencoded;
1093         const char *encoding;
1094         int need_8bit_cte = context->need_8bit_cte;
1095
1096         if (fmt == CMIT_FMT_USERFORMAT) {
1097                 format_commit_message(commit, user_format, sb, context);
1098                 return;
1099         }
1100
1101         reencoded = reencode_commit_message(commit, &encoding);
1102         if (reencoded) {
1103                 msg = reencoded;
1104         }
1105
1106         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1107                 indent = 0;
1108
1109         /*
1110          * We need to check and emit Content-type: to mark it
1111          * as 8-bit if we haven't done so.
1112          */
1113         if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1114                 int i, ch, in_body;
1115
1116                 for (in_body = i = 0; (ch = msg[i]); i++) {
1117                         if (!in_body) {
1118                                 /* author could be non 7-bit ASCII but
1119                                  * the log may be so; skip over the
1120                                  * header part first.
1121                                  */
1122                                 if (ch == '\n' && msg[i+1] == '\n')
1123                                         in_body = 1;
1124                         }
1125                         else if (non_ascii(ch)) {
1126                                 need_8bit_cte = 1;
1127                                 break;
1128                         }
1129                 }
1130         }
1131
1132         pp_header(fmt, context->abbrev, context->date_mode, encoding,
1133                   commit, &msg, sb);
1134         if (fmt != CMIT_FMT_ONELINE && !context->subject) {
1135                 strbuf_addch(sb, '\n');
1136         }
1137
1138         /* Skip excess blank lines at the beginning of body, if any... */
1139         msg = skip_empty_lines(msg);
1140
1141         /* These formats treat the title line specially. */
1142         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1143                 pp_title_line(fmt, &msg, sb, context->subject,
1144                               context->after_subject, encoding, need_8bit_cte);
1145
1146         beginning_of_body = sb->len;
1147         if (fmt != CMIT_FMT_ONELINE)
1148                 pp_remainder(fmt, &msg, sb, indent);
1149         strbuf_rtrim(sb);
1150
1151         /* Make sure there is an EOLN for the non-oneline case */
1152         if (fmt != CMIT_FMT_ONELINE)
1153                 strbuf_addch(sb, '\n');
1154
1155         /*
1156          * The caller may append additional body text in e-mail
1157          * format.  Make sure we did not strip the blank line
1158          * between the header and the body.
1159          */
1160         if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1161                 strbuf_addch(sb, '\n');
1162
1163         if (context->show_notes)
1164                 format_display_notes(commit->object.sha1, sb, encoding,
1165                                      NOTES_SHOW_HEADER | NOTES_INDENT);
1166
1167         free(reencoded);
1168 }