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_rfc2047_special(char ch)
 212 {
 213         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 214 }
 215
 216 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 217                        const char *encoding)
 218 {
 219         static const int max_length = 78; /* per rfc2822 */
 220         int i;
 221         int line_len;
 222
 223         /* How many bytes are already used on the current line? */
 224         for (i = sb->len - 1; i >= 0; i--)
 225                 if (sb->buf[i] == '\n')
 226                         break;
 227         line_len = sb->len - (i+1);
 228
 229         for (i = 0; i < len; i++) {
 230                 int ch = line[i];
 231                 if (non_ascii(ch) || ch == '\n')
 232                         goto needquote;
 233                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 234                         goto needquote;
 235         }
 236         strbuf_add_wrapped_bytes(sb, line, len, 0, 1, max_length - line_len);
 237         return;
 238
 239 needquote:
 240         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 241         strbuf_addf(sb, "=?%s?q?", encoding);
 242         line_len += strlen(encoding) + 5; /* 5 for =??q? */
 243         for (i = 0; i < len; i++) {
 244                 unsigned ch = line[i] & 0xFF;
 245
 246                 if (line_len >= max_length - 2) {
 247                         strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 248                         line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 249                 }
 250
 251                 /*
 252                  * We encode ' ' using '=20' even though rfc2047
 253                  * allows using '_' for readability.  Unfortunately,
 254                  * many programs do not understand this and just
 255                  * leave the underscore in place.
 256                  */
 257                 if (is_rfc2047_special(ch) || ch == ' ' || ch == '\n') {
 258                         strbuf_addf(sb, "=%02X", ch);
 259                         line_len += 3;
 260                 }
 261                 else {
 262                         strbuf_addch(sb, ch);
 263                         line_len++;
 264                 }
 265         }
 266         strbuf_addstr(sb, "?=");
 267 }
 268
 269 void pp_user_info(const struct pretty_print_context *pp,
 270                   const char *what, struct strbuf *sb,
 271                   const char *line, const char *encoding)
 272 {
 273         char *date;
 274         int namelen;
 275         unsigned long time;
 276         int tz;
 277
 278         if (pp->fmt == CMIT_FMT_ONELINE)
 279                 return;
 280         date = strchr(line, '>');
 281         if (!date)
 282                 return;
 283         namelen = ++date - line;
 284         time = strtoul(date, &date, 10);
 285         tz = strtol(date, NULL, 10);
 286
 287         if (pp->fmt == CMIT_FMT_EMAIL) {
 288                 char *name_tail = strchr(line, '<');
 289                 int display_name_length;
 290                 int final_line;
 291                 if (!name_tail)
 292                         return;
 293                 while (line < name_tail && isspace(name_tail[-1]))
 294                         name_tail--;
 295                 display_name_length = name_tail - line;
 296                 strbuf_addstr(sb, "From: ");
 297                 add_rfc2047(sb, line, display_name_length, encoding);
 298                 for (final_line = 0; final_line < sb->len; final_line++)
 299                         if (sb->buf[sb->len - final_line - 1] == '\n')
 300                                 break;
 301                 if (namelen - display_name_length + final_line > 78) {
 302                         strbuf_addch(sb, '\n');
 303                         if (!isspace(name_tail[0]))
 304                                 strbuf_addch(sb, ' ');
 305                 }
 306                 strbuf_add(sb, name_tail, namelen - display_name_length);
 307                 strbuf_addch(sb, '\n');
 308         } else {
 309                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 310                               (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
 311                               "    ", namelen, line);
 312         }
 313         switch (pp->fmt) {
 314         case CMIT_FMT_MEDIUM:
 315                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, pp->date_mode));
 316                 break;
 317         case CMIT_FMT_EMAIL:
 318                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 319                 break;
 320         case CMIT_FMT_FULLER:
 321                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
 322                 break;
 323         default:
 324                 /* notin' */
 325                 break;
 326         }
 327 }
 328
 329 static int is_empty_line(const char *line, int *len_p)
 330 {
 331         int len = *len_p;
 332         while (len && isspace(line[len-1]))
 333                 len--;
 334         *len_p = len;
 335         return !len;
 336 }
 337
 338 static const char *skip_empty_lines(const char *msg)
 339 {
 340         for (;;) {
 341                 int linelen = get_one_line(msg);
 342                 int ll = linelen;
 343                 if (!linelen)
 344                         break;
 345                 if (!is_empty_line(msg, &ll))
 346                         break;
 347                 msg += linelen;
 348         }
 349         return msg;
 350 }
 351
 352 static void add_merge_info(const struct pretty_print_context *pp,
 353                            struct strbuf *sb, const struct commit *commit)
 354 {
 355         struct commit_list *parent = commit->parents;
 356
 357         if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 358             !parent || !parent->next)
 359                 return;
 360
 361         strbuf_addstr(sb, "Merge:");
 362
 363         while (parent) {
 364                 struct commit *p = parent->item;
 365                 const char *hex = NULL;
 366                 if (pp->abbrev)
 367                         hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 368                 if (!hex)
 369                         hex = sha1_to_hex(p->object.sha1);
 370                 parent = parent->next;
 371
 372                 strbuf_addf(sb, " %s", hex);
 373         }
 374         strbuf_addch(sb, '\n');
 375 }
 376
 377 static char *get_header(const struct commit *commit, const char *key)
 378 {
 379         int key_len = strlen(key);
 380         const char *line = commit->buffer;
 381
 382         for (;;) {
 383                 const char *eol = strchr(line, '\n'), *next;
 384
 385                 if (line == eol)
 386                         return NULL;
 387                 if (!eol) {
 388                         eol = line + strlen(line);
 389                         next = NULL;
 390                 } else
 391                         next = eol + 1;
 392                 if (eol - line > key_len &&
 393                     !strncmp(line, key, key_len) &&
 394                     line[key_len] == ' ') {
 395                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 396                 }
 397                 line = next;
 398         }
 399 }
 400
 401 static char *replace_encoding_header(char *buf, const char *encoding)
 402 {
 403         struct strbuf tmp = STRBUF_INIT;
 404         size_t start, len;
 405         char *cp = buf;
 406
 407         /* guess if there is an encoding header before a \n\n */
 408         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 409                 cp = strchr(cp, '\n');
 410                 if (!cp || *++cp == '\n')
 411                         return buf;
 412         }
 413         start = cp - buf;
 414         cp = strchr(cp, '\n');
 415         if (!cp)
 416                 return buf; /* should not happen but be defensive */
 417         len = cp + 1 - (buf + start);
 418
 419         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 420         if (is_encoding_utf8(encoding)) {
 421                 /* we have re-coded to UTF-8; drop the header */
 422                 strbuf_remove(&tmp, start, len);
 423         } else {
 424                 /* just replaces XXXX in 'encoding XXXX\n' */
 425                 strbuf_splice(&tmp, start + strlen("encoding "),
 426                                           len - strlen("encoding \n"),
 427                                           encoding, strlen(encoding));
 428         }
 429         return strbuf_detach(&tmp, NULL);
 430 }
 431
 432 char *logmsg_reencode(const struct commit *commit,
 433                       const char *output_encoding)
 434 {
 435         static const char *utf8 = "UTF-8";
 436         const char *use_encoding;
 437         char *encoding;
 438         char *out;
 439
 440         if (!*output_encoding)
 441                 return NULL;
 442         encoding = get_header(commit, "encoding");
 443         use_encoding = encoding ? encoding : utf8;
 444         if (!strcmp(use_encoding, output_encoding))
 445                 if (encoding) /* we'll strip encoding header later */
 446                         out = xstrdup(commit->buffer);
 447                 else
 448                         return NULL; /* nothing to do */
 449         else
 450                 out = reencode_string(commit->buffer,
 451                                       output_encoding, use_encoding);
 452         if (out)
 453                 out = replace_encoding_header(out, output_encoding);
 454
 455         free(encoding);
 456         return out;
 457 }
 458
 459 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 460 {
 461         static struct string_list *mail_map;
 462         if (!mail_map) {
 463                 mail_map = xcalloc(1, sizeof(*mail_map));
 464                 read_mailmap(mail_map, NULL);
 465         }
 466         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 467 }
 468
 469 static size_t format_person_part(struct strbuf *sb, char part,
 470                                  const char *msg, int len, enum date_mode dmode)
 471 {
 472         /* currently all placeholders have same length */
 473         const int placeholder_len = 2;
 474         int start, end, tz = 0;
 475         unsigned long date = 0;
 476         char *ep;
 477         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 478         char person_name[1024];
 479         char person_mail[1024];
 480
 481         /* advance 'end' to point to email start delimiter */
 482         for (end = 0; end < len && msg[end] != '<'; end++)
 483                 ; /* do nothing */
 484
 485         /*
 486          * When end points at the '<' that we found, it should have
 487          * matching '>' later, which means 'end' must be strictly
 488          * below len - 1.
 489          */
 490         if (end >= len - 2)
 491                 goto skip;
 492
 493         /* Seek for both name and email part */
 494         name_start = msg;
 495         name_end = msg+end;
 496         while (name_end > name_start && isspace(*(name_end-1)))
 497                 name_end--;
 498         mail_start = msg+end+1;
 499         mail_end = mail_start;
 500         while (mail_end < msg_end && *mail_end != '>')
 501                 mail_end++;
 502         if (mail_end == msg_end)
 503                 goto skip;
 504         end = mail_end-msg;
 505
 506         if (part == 'N' || part == 'E') { /* mailmap lookup */
 507                 strlcpy(person_name, name_start, name_end-name_start+1);
 508                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 509                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 510                 name_start = person_name;
 511                 name_end = name_start + strlen(person_name);
 512                 mail_start = person_mail;
 513                 mail_end = mail_start +  strlen(person_mail);
 514         }
 515         if (part == 'n' || part == 'N') {       /* name */
 516                 strbuf_add(sb, name_start, name_end-name_start);
 517                 return placeholder_len;
 518         }
 519         if (part == 'e' || part == 'E') {       /* email */
 520                 strbuf_add(sb, mail_start, mail_end-mail_start);
 521                 return placeholder_len;
 522         }
 523
 524         /* advance 'start' to point to date start delimiter */
 525         for (start = end + 1; start < len && isspace(msg[start]); start++)
 526                 ; /* do nothing */
 527         if (start >= len)
 528                 goto skip;
 529         date = strtoul(msg + start, &ep, 10);
 530         if (msg + start == ep)
 531                 goto skip;
 532
 533         if (part == 't') {      /* date, UNIX timestamp */
 534                 strbuf_add(sb, msg + start, ep - (msg + start));
 535                 return placeholder_len;
 536         }
 537
 538         /* parse tz */
 539         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 540                 ; /* do nothing */
 541         if (start + 1 < len) {
 542                 tz = strtoul(msg + start + 1, NULL, 10);
 543                 if (msg[start] == '-')
 544                         tz = -tz;
 545         }
 546
 547         switch (part) {
 548         case 'd':       /* date */
 549                 strbuf_addstr(sb, show_date(date, tz, dmode));
 550                 return placeholder_len;
 551         case 'D':       /* date, RFC2822 style */
 552                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 553                 return placeholder_len;
 554         case 'r':       /* date, relative */
 555                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 556                 return placeholder_len;
 557         case 'i':       /* date, ISO 8601 */
 558                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 559                 return placeholder_len;
 560         }
 561
 562 skip:
 563         /*
 564          * bogus commit, 'sb' cannot be updated, but we still need to
 565          * compute a valid return value.
 566          */
 567         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 568             || part == 'D' || part == 'r' || part == 'i')
 569                 return placeholder_len;
 570
 571         return 0; /* unknown placeholder */
 572 }
 573
 574 struct chunk {
 575         size_t off;
 576         size_t len;
 577 };
 578
 579 struct format_commit_context {
 580         const struct commit *commit;
 581         const struct pretty_print_context *pretty_ctx;
 582         unsigned commit_header_parsed:1;
 583         unsigned commit_message_parsed:1;
 584         char *message;
 585         size_t width, indent1, indent2;
 586
 587         /* These offsets are relative to the start of the commit message. */
 588         struct chunk author;
 589         struct chunk committer;
 590         struct chunk encoding;
 591         size_t message_off;
 592         size_t subject_off;
 593         size_t body_off;
 594
 595         /* The following ones are relative to the result struct strbuf. */
 596         struct chunk abbrev_commit_hash;
 597         struct chunk abbrev_tree_hash;
 598         struct chunk abbrev_parent_hashes;
 599         size_t wrap_start;
 600 };
 601
 602 static int add_again(struct strbuf *sb, struct chunk *chunk)
 603 {
 604         if (chunk->len) {
 605                 strbuf_adddup(sb, chunk->off, chunk->len);
 606                 return 1;
 607         }
 608
 609         /*
 610          * We haven't seen this chunk before.  Our caller is surely
 611          * going to add it the hard way now.  Remember the most likely
 612          * start of the to-be-added chunk: the current end of the
 613          * struct strbuf.
 614          */
 615         chunk->off = sb->len;
 616         return 0;
 617 }
 618
 619 static void parse_commit_header(struct format_commit_context *context)
 620 {
 621         const char *msg = context->message;
 622         int i;
 623
 624         for (i = 0; msg[i]; i++) {
 625                 int eol;
 626                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 627                         ; /* do nothing */
 628
 629                 if (i == eol) {
 630                         break;
 631                 } else if (!prefixcmp(msg + i, "author ")) {
 632                         context->author.off = i + 7;
 633                         context->author.len = eol - i - 7;
 634                 } else if (!prefixcmp(msg + i, "committer ")) {
 635                         context->committer.off = i + 10;
 636                         context->committer.len = eol - i - 10;
 637                 } else if (!prefixcmp(msg + i, "encoding ")) {
 638                         context->encoding.off = i + 9;
 639                         context->encoding.len = eol - i - 9;
 640                 }
 641                 i = eol;
 642         }
 643         context->message_off = i;
 644         context->commit_header_parsed = 1;
 645 }
 646
 647 static int istitlechar(char c)
 648 {
 649         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 650                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 651 }
 652
 653 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 654 {
 655         size_t trimlen;
 656         size_t start_len = sb->len;
 657         int space = 2;
 658
 659         for (; *msg && *msg != '\n'; msg++) {
 660                 if (istitlechar(*msg)) {
 661                         if (space == 1)
 662                                 strbuf_addch(sb, '-');
 663                         space = 0;
 664                         strbuf_addch(sb, *msg);
 665                         if (*msg == '.')
 666                                 while (*(msg+1) == '.')
 667                                         msg++;
 668                 } else
 669                         space |= 1;
 670         }
 671
 672         /* trim any trailing '.' or '-' characters */
 673         trimlen = 0;
 674         while (sb->len - trimlen > start_len &&
 675                 (sb->buf[sb->len - 1 - trimlen] == '.'
 676                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 677                 trimlen++;
 678         strbuf_remove(sb, sb->len - trimlen, trimlen);
 679 }
 680
 681 const char *format_subject(struct strbuf *sb, const char *msg,
 682                            const char *line_separator)
 683 {
 684         int first = 1;
 685
 686         for (;;) {
 687                 const char *line = msg;
 688                 int linelen = get_one_line(line);
 689
 690                 msg += linelen;
 691                 if (!linelen || is_empty_line(line, &linelen))
 692                         break;
 693
 694                 if (!sb)
 695                         continue;
 696                 strbuf_grow(sb, linelen + 2);
 697                 if (!first)
 698                         strbuf_addstr(sb, line_separator);
 699                 strbuf_add(sb, line, linelen);
 700                 first = 0;
 701         }
 702         return msg;
 703 }
 704
 705 static void parse_commit_message(struct format_commit_context *c)
 706 {
 707         const char *msg = c->message + c->message_off;
 708         const char *start = c->message;
 709
 710         msg = skip_empty_lines(msg);
 711         c->subject_off = msg - start;
 712
 713         msg = format_subject(NULL, msg, NULL);
 714         msg = skip_empty_lines(msg);
 715         c->body_off = msg - start;
 716
 717         c->commit_message_parsed = 1;
 718 }
 719
 720 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 721 {
 722         struct name_decoration *d;
 723         const char *prefix = " (";
 724
 725         load_ref_decorations(DECORATE_SHORT_REFS);
 726         d = lookup_decoration(&name_decoration, &commit->object);
 727         while (d) {
 728                 strbuf_addstr(sb, prefix);
 729                 prefix = ", ";
 730                 strbuf_addstr(sb, d->name);
 731                 d = d->next;
 732         }
 733         if (prefix[0] == ',')
 734                 strbuf_addch(sb, ')');
 735 }
 736
 737 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 738                         size_t width, size_t indent1, size_t indent2)
 739 {
 740         struct strbuf tmp = STRBUF_INIT;
 741
 742         if (pos)
 743                 strbuf_add(&tmp, sb->buf, pos);
 744         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 745                                 (int) indent1, (int) indent2, (int) width);
 746         strbuf_swap(&tmp, sb);
 747         strbuf_release(&tmp);
 748 }
 749
 750 static void rewrap_message_tail(struct strbuf *sb,
 751                                 struct format_commit_context *c,
 752                                 size_t new_width, size_t new_indent1,
 753                                 size_t new_indent2)
 754 {
 755         if (c->width == new_width && c->indent1 == new_indent1 &&
 756             c->indent2 == new_indent2)
 757                 return;
 758         if (c->wrap_start < sb->len)
 759                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 760         c->wrap_start = sb->len;
 761         c->width = new_width;
 762         c->indent1 = new_indent1;
 763         c->indent2 = new_indent2;
 764 }
 765
 766 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 767                                 void *context)
 768 {
 769         struct format_commit_context *c = context;
 770         const struct commit *commit = c->commit;
 771         const char *msg = c->message;
 772         struct commit_list *p;
 773         int h1, h2;
 774
 775         /* these are independent of the commit */
 776         switch (placeholder[0]) {
 777         case 'C':
 778                 if (placeholder[1] == '(') {
 779                         const char *end = strchr(placeholder + 2, ')');
 780                         char color[COLOR_MAXLEN];
 781                         if (!end)
 782                                 return 0;
 783                         color_parse_mem(placeholder + 2,
 784                                         end - (placeholder + 2),
 785                                         "--pretty format", color);
 786                         strbuf_addstr(sb, color);
 787                         return end - placeholder + 1;
 788                 }
 789                 if (!prefixcmp(placeholder + 1, "red")) {
 790                         strbuf_addstr(sb, GIT_COLOR_RED);
 791                         return 4;
 792                 } else if (!prefixcmp(placeholder + 1, "green")) {
 793                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 794                         return 6;
 795                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 796                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 797                         return 5;
 798                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 799                         strbuf_addstr(sb, GIT_COLOR_RESET);
 800                         return 6;
 801                 } else
 802                         return 0;
 803         case 'n':               /* newline */
 804                 strbuf_addch(sb, '\n');
 805                 return 1;
 806         case 'x':
 807                 /* %x00 == NUL, %x0a == LF, etc. */
 808                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 809                     h1 <= 16 &&
 810                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 811                     h2 <= 16) {
 812                         strbuf_addch(sb, (h1<<4)|h2);
 813                         return 3;
 814                 } else
 815                         return 0;
 816         case 'w':
 817                 if (placeholder[1] == '(') {
 818                         unsigned long width = 0, indent1 = 0, indent2 = 0;
 819                         char *next;
 820                         const char *start = placeholder + 2;
 821                         const char *end = strchr(start, ')');
 822                         if (!end)
 823                                 return 0;
 824                         if (end > start) {
 825                                 width = strtoul(start, &next, 10);
 826                                 if (*next == ',') {
 827                                         indent1 = strtoul(next + 1, &next, 10);
 828                                         if (*next == ',') {
 829                                                 indent2 = strtoul(next + 1,
 830                                                                  &next, 10);
 831                                         }
 832                                 }
 833                                 if (*next != ')')
 834                                         return 0;
 835                         }
 836                         rewrap_message_tail(sb, c, width, indent1, indent2);
 837                         return end - placeholder + 1;
 838                 } else
 839                         return 0;
 840         }
 841
 842         /* these depend on the commit */
 843         if (!commit->object.parsed)
 844                 parse_object(commit->object.sha1);
 845
 846         switch (placeholder[0]) {
 847         case 'H':               /* commit hash */
 848                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 849                 return 1;
 850         case 'h':               /* abbreviated commit hash */
 851                 if (add_again(sb, &c->abbrev_commit_hash))
 852                         return 1;
 853                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 854                                                      c->pretty_ctx->abbrev));
 855                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 856                 return 1;
 857         case 'T':               /* tree hash */
 858                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 859                 return 1;
 860         case 't':               /* abbreviated tree hash */
 861                 if (add_again(sb, &c->abbrev_tree_hash))
 862                         return 1;
 863                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 864                                                      c->pretty_ctx->abbrev));
 865                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 866                 return 1;
 867         case 'P':               /* parent hashes */
 868                 for (p = commit->parents; p; p = p->next) {
 869                         if (p != commit->parents)
 870                                 strbuf_addch(sb, ' ');
 871                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 872                 }
 873                 return 1;
 874         case 'p':               /* abbreviated parent hashes */
 875                 if (add_again(sb, &c->abbrev_parent_hashes))
 876                         return 1;
 877                 for (p = commit->parents; p; p = p->next) {
 878                         if (p != commit->parents)
 879                                 strbuf_addch(sb, ' ');
 880                         strbuf_addstr(sb, find_unique_abbrev(
 881                                         p->item->object.sha1,
 882                                         c->pretty_ctx->abbrev));
 883                 }
 884                 c->abbrev_parent_hashes.len = sb->len -
 885                                               c->abbrev_parent_hashes.off;
 886                 return 1;
 887         case 'm':               /* left/right/bottom */
 888                 strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 889                                  ? '-'
 890                                  : (commit->object.flags & SYMMETRIC_LEFT)
 891                                  ? '<'
 892                                  : '>');
 893                 return 1;
 894         case 'd':
 895                 format_decoration(sb, commit);
 896                 return 1;
 897         case 'g':               /* reflog info */
 898                 switch(placeholder[1]) {
 899                 case 'd':       /* reflog selector */
 900                 case 'D':
 901                         if (c->pretty_ctx->reflog_info)
 902                                 get_reflog_selector(sb,
 903                                                     c->pretty_ctx->reflog_info,
 904                                                     c->pretty_ctx->date_mode,
 905                                                     (placeholder[1] == 'd'));
 906                         return 2;
 907                 case 's':       /* reflog message */
 908                         if (c->pretty_ctx->reflog_info)
 909                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
 910                         return 2;
 911                 }
 912                 return 0;       /* unknown %g placeholder */
 913         case 'N':
 914                 if (c->pretty_ctx->show_notes) {
 915                         format_display_notes(commit->object.sha1, sb,
 916                                     get_log_output_encoding(), 0);
 917                         return 1;
 918                 }
 919                 return 0;
 920         }
 921
 922         /* For the rest we have to parse the commit header. */
 923         if (!c->commit_header_parsed)
 924                 parse_commit_header(c);
 925
 926         switch (placeholder[0]) {
 927         case 'a':       /* author ... */
 928                 return format_person_part(sb, placeholder[1],
 929                                    msg + c->author.off, c->author.len,
 930                                    c->pretty_ctx->date_mode);
 931         case 'c':       /* committer ... */
 932                 return format_person_part(sb, placeholder[1],
 933                                    msg + c->committer.off, c->committer.len,
 934                                    c->pretty_ctx->date_mode);
 935         case 'e':       /* encoding */
 936                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 937                 return 1;
 938         case 'B':       /* raw body */
 939                 /* message_off is always left at the initial newline */
 940                 strbuf_addstr(sb, msg + c->message_off + 1);
 941                 return 1;
 942         }
 943
 944         /* Now we need to parse the commit message. */
 945         if (!c->commit_message_parsed)
 946                 parse_commit_message(c);
 947
 948         switch (placeholder[0]) {
 949         case 's':       /* subject */
 950                 format_subject(sb, msg + c->subject_off, " ");
 951                 return 1;
 952         case 'f':       /* sanitized subject */
 953                 format_sanitized_subject(sb, msg + c->subject_off);
 954                 return 1;
 955         case 'b':       /* body */
 956                 strbuf_addstr(sb, msg + c->body_off);
 957                 return 1;
 958         }
 959         return 0;       /* unknown placeholder */
 960 }
 961
 962 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 963                                  void *context)
 964 {
 965         int consumed;
 966         size_t orig_len;
 967         enum {
 968                 NO_MAGIC,
 969                 ADD_LF_BEFORE_NON_EMPTY,
 970                 DEL_LF_BEFORE_EMPTY,
 971                 ADD_SP_BEFORE_NON_EMPTY
 972         } magic = NO_MAGIC;
 973
 974         switch (placeholder[0]) {
 975         case '-':
 976                 magic = DEL_LF_BEFORE_EMPTY;
 977                 break;
 978         case '+':
 979                 magic = ADD_LF_BEFORE_NON_EMPTY;
 980                 break;
 981         case ' ':
 982                 magic = ADD_SP_BEFORE_NON_EMPTY;
 983                 break;
 984         default:
 985                 break;
 986         }
 987         if (magic != NO_MAGIC)
 988                 placeholder++;
 989
 990         orig_len = sb->len;
 991         consumed = format_commit_one(sb, placeholder, context);
 992         if (magic == NO_MAGIC)
 993                 return consumed;
 994
 995         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 996                 while (sb->len && sb->buf[sb->len - 1] == '\n')
 997                         strbuf_setlen(sb, sb->len - 1);
 998         } else if (orig_len != sb->len) {
 999                 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1000                         strbuf_insert(sb, orig_len, "\n", 1);
1001                 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1002                         strbuf_insert(sb, orig_len, " ", 1);
1003         }
1004         return consumed + 1;
1005 }
1006
1007 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1008                                    void *context)
1009 {
1010         struct userformat_want *w = context;
1011
1012         if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1013                 placeholder++;
1014
1015         switch (*placeholder) {
1016         case 'N':
1017                 w->notes = 1;
1018                 break;
1019         }
1020         return 0;
1021 }
1022
1023 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1024 {
1025         struct strbuf dummy = STRBUF_INIT;
1026
1027         if (!fmt) {
1028                 if (!user_format)
1029                         return;
1030                 fmt = user_format;
1031         }
1032         strbuf_expand(&dummy, user_format, userformat_want_item, w);
1033         strbuf_release(&dummy);
1034 }
1035
1036 void format_commit_message(const struct commit *commit,
1037                            const char *format, struct strbuf *sb,
1038                            const struct pretty_print_context *pretty_ctx)
1039 {
1040         struct format_commit_context context;
1041         static const char utf8[] = "UTF-8";
1042         const char *enc;
1043         const char *output_enc = pretty_ctx->output_encoding;
1044
1045         memset(&context, 0, sizeof(context));
1046         context.commit = commit;
1047         context.pretty_ctx = pretty_ctx;
1048         context.wrap_start = sb->len;
1049         context.message = commit->buffer;
1050         if (output_enc) {
1051                 enc = get_header(commit, "encoding");
1052                 enc = enc ? enc : utf8;
1053                 if (strcmp(enc, output_enc))
1054                         context.message = logmsg_reencode(commit, output_enc);
1055         }
1056
1057         strbuf_expand(sb, format, format_commit_item, &context);
1058         rewrap_message_tail(sb, &context, 0, 0, 0);
1059
1060         if (context.message != commit->buffer)
1061                 free(context.message);
1062 }
1063
1064 static void pp_header(const struct pretty_print_context *pp,
1065                       const char *encoding,
1066                       const struct commit *commit,
1067                       const char **msg_p,
1068                       struct strbuf *sb)
1069 {
1070         int parents_shown = 0;
1071
1072         for (;;) {
1073                 const char *line = *msg_p;
1074                 int linelen = get_one_line(*msg_p);
1075
1076                 if (!linelen)
1077                         return;
1078                 *msg_p += linelen;
1079
1080                 if (linelen == 1)
1081                         /* End of header */
1082                         return;
1083
1084                 if (pp->fmt == CMIT_FMT_RAW) {
1085                         strbuf_add(sb, line, linelen);
1086                         continue;
1087                 }
1088
1089                 if (!memcmp(line, "parent ", 7)) {
1090                         if (linelen != 48)
1091                                 die("bad parent line in commit");
1092                         continue;
1093                 }
1094
1095                 if (!parents_shown) {
1096                         struct commit_list *parent;
1097                         int num;
1098                         for (parent = commit->parents, num = 0;
1099                              parent;
1100                              parent = parent->next, num++)
1101                                 ;
1102                         /* with enough slop */
1103                         strbuf_grow(sb, num * 50 + 20);
1104                         add_merge_info(pp, sb, commit);
1105                         parents_shown = 1;
1106                 }
1107
1108                 /*
1109                  * MEDIUM == DEFAULT shows only author with dates.
1110                  * FULL shows both authors but not dates.
1111                  * FULLER shows both authors and dates.
1112                  */
1113                 if (!memcmp(line, "author ", 7)) {
1114                         strbuf_grow(sb, linelen + 80);
1115                         pp_user_info(pp, "Author", sb, line + 7, encoding);
1116                 }
1117                 if (!memcmp(line, "committer ", 10) &&
1118                     (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1119                         strbuf_grow(sb, linelen + 80);
1120                         pp_user_info(pp, "Commit", sb, line + 10, encoding);
1121                 }
1122         }
1123 }
1124
1125 void pp_title_line(const struct pretty_print_context *pp,
1126                    const char **msg_p,
1127                    struct strbuf *sb,
1128                    const char *encoding,
1129                    int need_8bit_cte)
1130 {
1131         struct strbuf title;
1132
1133         strbuf_init(&title, 80);
1134         *msg_p = format_subject(&title, *msg_p,
1135                                 pp->preserve_subject ? "\n" : " ");
1136
1137         strbuf_grow(sb, title.len + 1024);
1138         if (pp->subject) {
1139                 strbuf_addstr(sb, pp->subject);
1140                 add_rfc2047(sb, title.buf, title.len, encoding);
1141         } else {
1142                 strbuf_addbuf(sb, &title);
1143         }
1144         strbuf_addch(sb, '\n');
1145
1146         if (need_8bit_cte > 0) {
1147                 const char *header_fmt =
1148                         "MIME-Version: 1.0\n"
1149                         "Content-Type: text/plain; charset=%s\n"
1150                         "Content-Transfer-Encoding: 8bit\n";
1151                 strbuf_addf(sb, header_fmt, encoding);
1152         }
1153         if (pp->after_subject) {
1154                 strbuf_addstr(sb, pp->after_subject);
1155         }
1156         if (pp->fmt == CMIT_FMT_EMAIL) {
1157                 strbuf_addch(sb, '\n');
1158         }
1159         strbuf_release(&title);
1160 }
1161
1162 void pp_remainder(const struct pretty_print_context *pp,
1163                   const char **msg_p,
1164                   struct strbuf *sb,
1165                   int indent)
1166 {
1167         int first = 1;
1168         for (;;) {
1169                 const char *line = *msg_p;
1170                 int linelen = get_one_line(line);
1171                 *msg_p += linelen;
1172
1173                 if (!linelen)
1174                         break;
1175
1176                 if (is_empty_line(line, &linelen)) {
1177                         if (first)
1178                                 continue;
1179                         if (pp->fmt == CMIT_FMT_SHORT)
1180                                 break;
1181                 }
1182                 first = 0;
1183
1184                 strbuf_grow(sb, linelen + indent + 20);
1185                 if (indent) {
1186                         memset(sb->buf + sb->len, ' ', indent);
1187                         strbuf_setlen(sb, sb->len + indent);
1188                 }
1189                 strbuf_add(sb, line, linelen);
1190                 strbuf_addch(sb, '\n');
1191         }
1192 }
1193
1194 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1195 {
1196         const char *encoding;
1197
1198         encoding = get_log_output_encoding();
1199         if (encoding_p)
1200                 *encoding_p = encoding;
1201         return logmsg_reencode(commit, encoding);
1202 }
1203
1204 void pretty_print_commit(const struct pretty_print_context *pp,
1205                          const struct commit *commit,
1206                          struct strbuf *sb)
1207 {
1208         unsigned long beginning_of_body;
1209         int indent = 4;
1210         const char *msg = commit->buffer;
1211         char *reencoded;
1212         const char *encoding;
1213         int need_8bit_cte = pp->need_8bit_cte;
1214
1215         if (pp->fmt == CMIT_FMT_USERFORMAT) {
1216                 format_commit_message(commit, user_format, sb, pp);
1217                 return;
1218         }
1219
1220         reencoded = reencode_commit_message(commit, &encoding);
1221         if (reencoded) {
1222                 msg = reencoded;
1223         }
1224
1225         if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1226                 indent = 0;
1227
1228         /*
1229          * We need to check and emit Content-type: to mark it
1230          * as 8-bit if we haven't done so.
1231          */
1232         if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1233                 int i, ch, in_body;
1234
1235                 for (in_body = i = 0; (ch = msg[i]); i++) {
1236                         if (!in_body) {
1237                                 /* author could be non 7-bit ASCII but
1238                                  * the log may be so; skip over the
1239                                  * header part first.
1240                                  */
1241                                 if (ch == '\n' && msg[i+1] == '\n')
1242                                         in_body = 1;
1243                         }
1244                         else if (non_ascii(ch)) {
1245                                 need_8bit_cte = 1;
1246                                 break;
1247                         }
1248                 }
1249         }
1250
1251         pp_header(pp, encoding, commit, &msg, sb);
1252         if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1253                 strbuf_addch(sb, '\n');
1254         }
1255
1256         /* Skip excess blank lines at the beginning of body, if any... */
1257         msg = skip_empty_lines(msg);
1258
1259         /* These formats treat the title line specially. */
1260         if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1261                 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1262
1263         beginning_of_body = sb->len;
1264         if (pp->fmt != CMIT_FMT_ONELINE)
1265                 pp_remainder(pp, &msg, sb, indent);
1266         strbuf_rtrim(sb);
1267
1268         /* Make sure there is an EOLN for the non-oneline case */
1269         if (pp->fmt != CMIT_FMT_ONELINE)
1270                 strbuf_addch(sb, '\n');
1271
1272         /*
1273          * The caller may append additional body text in e-mail
1274          * format.  Make sure we did not strip the blank line
1275          * between the header and the body.
1276          */
1277         if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1278                 strbuf_addch(sb, '\n');
1279
1280         if (pp->show_notes)
1281                 format_display_notes(commit->object.sha1, sb, encoding,
1282                                      NOTES_SHOW_HEADER | NOTES_INDENT);
1283
1284         free(reencoded);
1285 }
1286
1287 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1288                     struct strbuf *sb)
1289 {
1290         struct pretty_print_context pp = {0};
1291         pp.fmt = fmt;
1292         pretty_print_commit(&pp, commit, sb);
1293 }