pretty.c

   1 #include "cache.h"
   2 #include "commit.h"
   3 #include "utf8.h"
   4 #include "diff.h"
   5 #include "revision.h"
   6 #include "string-list.h"
   7 #include "mailmap.h"
   8 #include "log-tree.h"
   9 #include "notes.h"
  10 #include "color.h"
  11 #include "reflog-walk.h"
  12
  13 static char *user_format;
  14 static struct cmt_fmt_map {
  15         const char *name;
  16         enum cmit_fmt format;
  17         int is_tformat;
  18         int is_alias;
  19         const char *user_format;
  20 } *commit_formats;
  21 static size_t builtin_formats_len;
  22 static size_t commit_formats_len;
  23 static size_t commit_formats_alloc;
  24 static struct cmt_fmt_map *find_commit_format(const char *sought);
  25
  26 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  27 {
  28         free(user_format);
  29         user_format = xstrdup(cp);
  30         if (is_tformat)
  31                 rev->use_terminator = 1;
  32         rev->commit_format = CMIT_FMT_USERFORMAT;
  33 }
  34
  35 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  36 {
  37         struct cmt_fmt_map *commit_format = NULL;
  38         const char *name;
  39         const char *fmt;
  40         int i;
  41
  42         if (prefixcmp(var, "pretty."))
  43                 return 0;
  44
  45         name = var + strlen("pretty.");
  46         for (i = 0; i < builtin_formats_len; i++) {
  47                 if (!strcmp(commit_formats[i].name, name))
  48                         return 0;
  49         }
  50
  51         for (i = builtin_formats_len; i < commit_formats_len; i++) {
  52                 if (!strcmp(commit_formats[i].name, name)) {
  53                         commit_format = &commit_formats[i];
  54                         break;
  55                 }
  56         }
  57
  58         if (!commit_format) {
  59                 ALLOC_GROW(commit_formats, commit_formats_len+1,
  60                            commit_formats_alloc);
  61                 commit_format = &commit_formats[commit_formats_len];
  62                 memset(commit_format, 0, sizeof(*commit_format));
  63                 commit_formats_len++;
  64         }
  65
  66         commit_format->name = xstrdup(name);
  67         commit_format->format = CMIT_FMT_USERFORMAT;
  68         git_config_string(&fmt, var, value);
  69         if (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
  70                 commit_format->is_tformat = fmt[0] == 't';
  71                 fmt = strchr(fmt, ':') + 1;
  72         } else if (strchr(fmt, '%'))
  73                 commit_format->is_tformat = 1;
  74         else
  75                 commit_format->is_alias = 1;
  76         commit_format->user_format = fmt;
  77
  78         return 0;
  79 }
  80
  81 static void setup_commit_formats(void)
  82 {
  83         struct cmt_fmt_map builtin_formats[] = {
  84                 { "raw",        CMIT_FMT_RAW,           0 },
  85                 { "medium",     CMIT_FMT_MEDIUM,        0 },
  86                 { "short",      CMIT_FMT_SHORT,         0 },
  87                 { "email",      CMIT_FMT_EMAIL,         0 },
  88                 { "fuller",     CMIT_FMT_FULLER,        0 },
  89                 { "full",       CMIT_FMT_FULL,          0 },
  90                 { "oneline",    CMIT_FMT_ONELINE,       1 }
  91         };
  92         commit_formats_len = ARRAY_SIZE(builtin_formats);
  93         builtin_formats_len = commit_formats_len;
  94         ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
  95         memcpy(commit_formats, builtin_formats,
  96                sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  97
  98         git_config(git_pretty_formats_config, NULL);
  99 }
 100
 101 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 102                                                         const char *original,
 103                                                         int num_redirections)
 104 {
 105         struct cmt_fmt_map *found = NULL;
 106         size_t found_match_len = 0;
 107         int i;
 108
 109         if (num_redirections >= commit_formats_len)
 110                 die("invalid --pretty format: "
 111                     "'%s' references an alias which points to itself",
 112                     original);
 113
 114         for (i = 0; i < commit_formats_len; i++) {
 115                 size_t match_len;
 116
 117                 if (prefixcmp(commit_formats[i].name, sought))
 118                         continue;
 119
 120                 match_len = strlen(commit_formats[i].name);
 121                 if (found == NULL || found_match_len > match_len) {
 122                         found = &commit_formats[i];
 123                         found_match_len = match_len;
 124                 }
 125         }
 126
 127         if (found && found->is_alias) {
 128                 found = find_commit_format_recursive(found->user_format,
 129                                                      original,
 130                                                      num_redirections+1);
 131         }
 132
 133         return found;
 134 }
 135
 136 static struct cmt_fmt_map *find_commit_format(const char *sought)
 137 {
 138         if (!commit_formats)
 139                 setup_commit_formats();
 140
 141         return find_commit_format_recursive(sought, sought, 0);
 142 }
 143
 144 void get_commit_format(const char *arg, struct rev_info *rev)
 145 {
 146         struct cmt_fmt_map *commit_format;
 147
 148         rev->use_terminator = 0;
 149         if (!arg || !*arg) {
 150                 rev->commit_format = CMIT_FMT_DEFAULT;
 151                 return;
 152         }
 153         if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
 154                 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 155                 return;
 156         }
 157
 158         if (strchr(arg, '%')) {
 159                 save_user_format(rev, arg, 1);
 160                 return;
 161         }
 162
 163         commit_format = find_commit_format(arg);
 164         if (!commit_format)
 165                 die("invalid --pretty format: %s", arg);
 166
 167         rev->commit_format = commit_format->format;
 168         rev->use_terminator = commit_format->is_tformat;
 169         if (commit_format->format == CMIT_FMT_USERFORMAT) {
 170                 save_user_format(rev, commit_format->user_format,
 171                                  commit_format->is_tformat);
 172         }
 173 }
 174
 175 /*
 176  * Generic support for pretty-printing the header
 177  */
 178 static int get_one_line(const char *msg)
 179 {
 180         int ret = 0;
 181
 182         for (;;) {
 183                 char c = *msg++;
 184                 if (!c)
 185                         break;
 186                 ret++;
 187                 if (c == '\n')
 188                         break;
 189         }
 190         return ret;
 191 }
 192
 193 /* High bit set, or ISO-2022-INT */
 194 static int non_ascii(int ch)
 195 {
 196         return !isascii(ch) || ch == '\033';
 197 }
 198
 199 int has_non_ascii(const char *s)
 200 {
 201         int ch;
 202         if (!s)
 203                 return 0;
 204         while ((ch = *s++) != '\0') {
 205                 if (non_ascii(ch))
 206                         return 1;
 207         }
 208         return 0;
 209 }
 210
 211 static int is_rfc822_special(char ch)
 212 {
 213         switch (ch) {
 214         case '(':
 215         case ')':
 216         case '<':
 217         case '>':
 218         case '[':
 219         case ']':
 220         case ':':
 221         case ';':
 222         case '@':
 223         case ',':
 224         case '.':
 225         case '"':
 226         case '\\':
 227                 return 1;
 228         default:
 229                 return 0;
 230         }
 231 }
 232
 233 static int has_rfc822_specials(const char *s, int len)
 234 {
 235         int i;
 236         for (i = 0; i < len; i++)
 237                 if (is_rfc822_special(s[i]))
 238                         return 1;
 239         return 0;
 240 }
 241
 242 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 243 {
 244         int i;
 245
 246         /* just a guess, we may have to also backslash-quote */
 247         strbuf_grow(out, len + 2);
 248
 249         strbuf_addch(out, '"');
 250         for (i = 0; i < len; i++) {
 251                 switch (s[i]) {
 252                 case '"':
 253                 case '\\':
 254                         strbuf_addch(out, '\\');
 255                         /* fall through */
 256                 default:
 257                         strbuf_addch(out, s[i]);
 258                 }
 259         }
 260         strbuf_addch(out, '"');
 261 }
 262
 263 static int is_rfc2047_special(char ch)
 264 {
 265         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 266 }
 267
 268 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 269                        const char *encoding)
 270 {
 271         static const int max_length = 78; /* per rfc2822 */
 272         int i;
 273         int line_len;
 274
 275         /* How many bytes are already used on the current line? */
 276         for (i = sb->len - 1; i >= 0; i--)
 277                 if (sb->buf[i] == '\n')
 278                         break;
 279         line_len = sb->len - (i+1);
 280
 281         for (i = 0; i < len; i++) {
 282                 int ch = line[i];
 283                 if (non_ascii(ch) || ch == '\n')
 284                         goto needquote;
 285                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 286                         goto needquote;
 287         }
 288         strbuf_add_wrapped_bytes(sb, line, len, 0, 1, max_length - line_len);
 289         return;
 290
 291 needquote:
 292         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 293         strbuf_addf(sb, "=?%s?q?", encoding);
 294         line_len += strlen(encoding) + 5; /* 5 for =??q? */
 295         for (i = 0; i < len; i++) {
 296                 unsigned ch = line[i] & 0xFF;
 297
 298                 if (line_len >= max_length - 2) {
 299                         strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 300                         line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 301                 }
 302
 303                 /*
 304                  * We encode ' ' using '=20' even though rfc2047
 305                  * allows using '_' for readability.  Unfortunately,
 306                  * many programs do not understand this and just
 307                  * leave the underscore in place.
 308                  */
 309                 if (is_rfc2047_special(ch) || ch == ' ' || ch == '\n') {
 310                         strbuf_addf(sb, "=%02X", ch);
 311                         line_len += 3;
 312                 }
 313                 else {
 314                         strbuf_addch(sb, ch);
 315                         line_len++;
 316                 }
 317         }
 318         strbuf_addstr(sb, "?=");
 319 }
 320
 321 void pp_user_info(const struct pretty_print_context *pp,
 322                   const char *what, struct strbuf *sb,
 323                   const char *line, const char *encoding)
 324 {
 325         char *date;
 326         int namelen;
 327         unsigned long time;
 328         int tz;
 329
 330         if (pp->fmt == CMIT_FMT_ONELINE)
 331                 return;
 332         date = strchr(line, '>');
 333         if (!date)
 334                 return;
 335         namelen = ++date - line;
 336         time = strtoul(date, &date, 10);
 337         tz = strtol(date, NULL, 10);
 338
 339         if (pp->fmt == CMIT_FMT_EMAIL) {
 340                 char *name_tail = strchr(line, '<');
 341                 int display_name_length;
 342                 int final_line;
 343                 if (!name_tail)
 344                         return;
 345                 while (line < name_tail && isspace(name_tail[-1]))
 346                         name_tail--;
 347                 display_name_length = name_tail - line;
 348                 strbuf_addstr(sb, "From: ");
 349                 if (!has_rfc822_specials(line, display_name_length)) {
 350                         add_rfc2047(sb, line, display_name_length, encoding);
 351                 } else {
 352                         struct strbuf quoted = STRBUF_INIT;
 353                         add_rfc822_quoted(&quoted, line, display_name_length);
 354                         add_rfc2047(sb, quoted.buf, quoted.len, encoding);
 355                         strbuf_release(&quoted);
 356                 }
 357                 for (final_line = 0; final_line < sb->len; final_line++)
 358                         if (sb->buf[sb->len - final_line - 1] == '\n')
 359                                 break;
 360                 if (namelen - display_name_length + final_line > 78) {
 361                         strbuf_addch(sb, '\n');
 362                         if (!isspace(name_tail[0]))
 363                                 strbuf_addch(sb, ' ');
 364                 }
 365                 strbuf_add(sb, name_tail, namelen - display_name_length);
 366                 strbuf_addch(sb, '\n');
 367         } else {
 368                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 369                               (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
 370                               "    ", namelen, line);
 371         }
 372         switch (pp->fmt) {
 373         case CMIT_FMT_MEDIUM:
 374                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, pp->date_mode));
 375                 break;
 376         case CMIT_FMT_EMAIL:
 377                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 378                 break;
 379         case CMIT_FMT_FULLER:
 380                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
 381                 break;
 382         default:
 383                 /* notin' */
 384                 break;
 385         }
 386 }
 387
 388 static int is_empty_line(const char *line, int *len_p)
 389 {
 390         int len = *len_p;
 391         while (len && isspace(line[len-1]))
 392                 len--;
 393         *len_p = len;
 394         return !len;
 395 }
 396
 397 static const char *skip_empty_lines(const char *msg)
 398 {
 399         for (;;) {
 400                 int linelen = get_one_line(msg);
 401                 int ll = linelen;
 402                 if (!linelen)
 403                         break;
 404                 if (!is_empty_line(msg, &ll))
 405                         break;
 406                 msg += linelen;
 407         }
 408         return msg;
 409 }
 410
 411 static void add_merge_info(const struct pretty_print_context *pp,
 412                            struct strbuf *sb, const struct commit *commit)
 413 {
 414         struct commit_list *parent = commit->parents;
 415
 416         if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 417             !parent || !parent->next)
 418                 return;
 419
 420         strbuf_addstr(sb, "Merge:");
 421
 422         while (parent) {
 423                 struct commit *p = parent->item;
 424                 const char *hex = NULL;
 425                 if (pp->abbrev)
 426                         hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 427                 if (!hex)
 428                         hex = sha1_to_hex(p->object.sha1);
 429                 parent = parent->next;
 430
 431                 strbuf_addf(sb, " %s", hex);
 432         }
 433         strbuf_addch(sb, '\n');
 434 }
 435
 436 static char *get_header(const struct commit *commit, const char *key)
 437 {
 438         int key_len = strlen(key);
 439         const char *line = commit->buffer;
 440
 441         for (;;) {
 442                 const char *eol = strchr(line, '\n'), *next;
 443
 444                 if (line == eol)
 445                         return NULL;
 446                 if (!eol) {
 447                         eol = line + strlen(line);
 448                         next = NULL;
 449                 } else
 450                         next = eol + 1;
 451                 if (eol - line > key_len &&
 452                     !strncmp(line, key, key_len) &&
 453                     line[key_len] == ' ') {
 454                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 455                 }
 456                 line = next;
 457         }
 458 }
 459
 460 static char *replace_encoding_header(char *buf, const char *encoding)
 461 {
 462         struct strbuf tmp = STRBUF_INIT;
 463         size_t start, len;
 464         char *cp = buf;
 465
 466         /* guess if there is an encoding header before a \n\n */
 467         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 468                 cp = strchr(cp, '\n');
 469                 if (!cp || *++cp == '\n')
 470                         return buf;
 471         }
 472         start = cp - buf;
 473         cp = strchr(cp, '\n');
 474         if (!cp)
 475                 return buf; /* should not happen but be defensive */
 476         len = cp + 1 - (buf + start);
 477
 478         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 479         if (is_encoding_utf8(encoding)) {
 480                 /* we have re-coded to UTF-8; drop the header */
 481                 strbuf_remove(&tmp, start, len);
 482         } else {
 483                 /* just replaces XXXX in 'encoding XXXX\n' */
 484                 strbuf_splice(&tmp, start + strlen("encoding "),
 485                                           len - strlen("encoding \n"),
 486                                           encoding, strlen(encoding));
 487         }
 488         return strbuf_detach(&tmp, NULL);
 489 }
 490
 491 char *logmsg_reencode(const struct commit *commit,
 492                       const char *output_encoding)
 493 {
 494         static const char *utf8 = "UTF-8";
 495         const char *use_encoding;
 496         char *encoding;
 497         char *out;
 498
 499         if (!*output_encoding)
 500                 return NULL;
 501         encoding = get_header(commit, "encoding");
 502         use_encoding = encoding ? encoding : utf8;
 503         if (!strcmp(use_encoding, output_encoding))
 504                 if (encoding) /* we'll strip encoding header later */
 505                         out = xstrdup(commit->buffer);
 506                 else
 507                         return NULL; /* nothing to do */
 508         else
 509                 out = reencode_string(commit->buffer,
 510                                       output_encoding, use_encoding);
 511         if (out)
 512                 out = replace_encoding_header(out, output_encoding);
 513
 514         free(encoding);
 515         return out;
 516 }
 517
 518 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 519 {
 520         static struct string_list *mail_map;
 521         if (!mail_map) {
 522                 mail_map = xcalloc(1, sizeof(*mail_map));
 523                 read_mailmap(mail_map, NULL);
 524         }
 525         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 526 }
 527
 528 static size_t format_person_part(struct strbuf *sb, char part,
 529                                  const char *msg, int len, enum date_mode dmode)
 530 {
 531         /* currently all placeholders have same length */
 532         const int placeholder_len = 2;
 533         int start, end, tz = 0;
 534         unsigned long date = 0;
 535         char *ep;
 536         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 537         char person_name[1024];
 538         char person_mail[1024];
 539
 540         /* advance 'end' to point to email start delimiter */
 541         for (end = 0; end < len && msg[end] != '<'; end++)
 542                 ; /* do nothing */
 543
 544         /*
 545          * When end points at the '<' that we found, it should have
 546          * matching '>' later, which means 'end' must be strictly
 547          * below len - 1.
 548          */
 549         if (end >= len - 2)
 550                 goto skip;
 551
 552         /* Seek for both name and email part */
 553         name_start = msg;
 554         name_end = msg+end;
 555         while (name_end > name_start && isspace(*(name_end-1)))
 556                 name_end--;
 557         mail_start = msg+end+1;
 558         mail_end = mail_start;
 559         while (mail_end < msg_end && *mail_end != '>')
 560                 mail_end++;
 561         if (mail_end == msg_end)
 562                 goto skip;
 563         end = mail_end-msg;
 564
 565         if (part == 'N' || part == 'E') { /* mailmap lookup */
 566                 strlcpy(person_name, name_start, name_end-name_start+1);
 567                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 568                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 569                 name_start = person_name;
 570                 name_end = name_start + strlen(person_name);
 571                 mail_start = person_mail;
 572                 mail_end = mail_start +  strlen(person_mail);
 573         }
 574         if (part == 'n' || part == 'N') {       /* name */
 575                 strbuf_add(sb, name_start, name_end-name_start);
 576                 return placeholder_len;
 577         }
 578         if (part == 'e' || part == 'E') {       /* email */
 579                 strbuf_add(sb, mail_start, mail_end-mail_start);
 580                 return placeholder_len;
 581         }
 582
 583         /* advance 'start' to point to date start delimiter */
 584         for (start = end + 1; start < len && isspace(msg[start]); start++)
 585                 ; /* do nothing */
 586         if (start >= len)
 587                 goto skip;
 588         date = strtoul(msg + start, &ep, 10);
 589         if (msg + start == ep)
 590                 goto skip;
 591
 592         if (part == 't') {      /* date, UNIX timestamp */
 593                 strbuf_add(sb, msg + start, ep - (msg + start));
 594                 return placeholder_len;
 595         }
 596
 597         /* parse tz */
 598         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 599                 ; /* do nothing */
 600         if (start + 1 < len) {
 601                 tz = strtoul(msg + start + 1, NULL, 10);
 602                 if (msg[start] == '-')
 603                         tz = -tz;
 604         }
 605
 606         switch (part) {
 607         case 'd':       /* date */
 608                 strbuf_addstr(sb, show_date(date, tz, dmode));
 609                 return placeholder_len;
 610         case 'D':       /* date, RFC2822 style */
 611                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 612                 return placeholder_len;
 613         case 'r':       /* date, relative */
 614                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 615                 return placeholder_len;
 616         case 'i':       /* date, ISO 8601 */
 617                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 618                 return placeholder_len;
 619         }
 620
 621 skip:
 622         /*
 623          * bogus commit, 'sb' cannot be updated, but we still need to
 624          * compute a valid return value.
 625          */
 626         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 627             || part == 'D' || part == 'r' || part == 'i')
 628                 return placeholder_len;
 629
 630         return 0; /* unknown placeholder */
 631 }
 632
 633 struct chunk {
 634         size_t off;
 635         size_t len;
 636 };
 637
 638 struct format_commit_context {
 639         const struct commit *commit;
 640         const struct pretty_print_context *pretty_ctx;
 641         unsigned commit_header_parsed:1;
 642         unsigned commit_message_parsed:1;
 643         char *message;
 644         size_t width, indent1, indent2;
 645
 646         /* These offsets are relative to the start of the commit message. */
 647         struct chunk author;
 648         struct chunk committer;
 649         struct chunk encoding;
 650         size_t message_off;
 651         size_t subject_off;
 652         size_t body_off;
 653
 654         /* The following ones are relative to the result struct strbuf. */
 655         struct chunk abbrev_commit_hash;
 656         struct chunk abbrev_tree_hash;
 657         struct chunk abbrev_parent_hashes;
 658         size_t wrap_start;
 659 };
 660
 661 static int add_again(struct strbuf *sb, struct chunk *chunk)
 662 {
 663         if (chunk->len) {
 664                 strbuf_adddup(sb, chunk->off, chunk->len);
 665                 return 1;
 666         }
 667
 668         /*
 669          * We haven't seen this chunk before.  Our caller is surely
 670          * going to add it the hard way now.  Remember the most likely
 671          * start of the to-be-added chunk: the current end of the
 672          * struct strbuf.
 673          */
 674         chunk->off = sb->len;
 675         return 0;
 676 }
 677
 678 static void parse_commit_header(struct format_commit_context *context)
 679 {
 680         const char *msg = context->message;
 681         int i;
 682
 683         for (i = 0; msg[i]; i++) {
 684                 int eol;
 685                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 686                         ; /* do nothing */
 687
 688                 if (i == eol) {
 689                         break;
 690                 } else if (!prefixcmp(msg + i, "author ")) {
 691                         context->author.off = i + 7;
 692                         context->author.len = eol - i - 7;
 693                 } else if (!prefixcmp(msg + i, "committer ")) {
 694                         context->committer.off = i + 10;
 695                         context->committer.len = eol - i - 10;
 696                 } else if (!prefixcmp(msg + i, "encoding ")) {
 697                         context->encoding.off = i + 9;
 698                         context->encoding.len = eol - i - 9;
 699                 }
 700                 i = eol;
 701         }
 702         context->message_off = i;
 703         context->commit_header_parsed = 1;
 704 }
 705
 706 static int istitlechar(char c)
 707 {
 708         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 709                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 710 }
 711
 712 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 713 {
 714         size_t trimlen;
 715         size_t start_len = sb->len;
 716         int space = 2;
 717
 718         for (; *msg && *msg != '\n'; msg++) {
 719                 if (istitlechar(*msg)) {
 720                         if (space == 1)
 721                                 strbuf_addch(sb, '-');
 722                         space = 0;
 723                         strbuf_addch(sb, *msg);
 724                         if (*msg == '.')
 725                                 while (*(msg+1) == '.')
 726                                         msg++;
 727                 } else
 728                         space |= 1;
 729         }
 730
 731         /* trim any trailing '.' or '-' characters */
 732         trimlen = 0;
 733         while (sb->len - trimlen > start_len &&
 734                 (sb->buf[sb->len - 1 - trimlen] == '.'
 735                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 736                 trimlen++;
 737         strbuf_remove(sb, sb->len - trimlen, trimlen);
 738 }
 739
 740 const char *format_subject(struct strbuf *sb, const char *msg,
 741                            const char *line_separator)
 742 {
 743         int first = 1;
 744
 745         for (;;) {
 746                 const char *line = msg;
 747                 int linelen = get_one_line(line);
 748
 749                 msg += linelen;
 750                 if (!linelen || is_empty_line(line, &linelen))
 751                         break;
 752
 753                 if (!sb)
 754                         continue;
 755                 strbuf_grow(sb, linelen + 2);
 756                 if (!first)
 757                         strbuf_addstr(sb, line_separator);
 758                 strbuf_add(sb, line, linelen);
 759                 first = 0;
 760         }
 761         return msg;
 762 }
 763
 764 static void parse_commit_message(struct format_commit_context *c)
 765 {
 766         const char *msg = c->message + c->message_off;
 767         const char *start = c->message;
 768
 769         msg = skip_empty_lines(msg);
 770         c->subject_off = msg - start;
 771
 772         msg = format_subject(NULL, msg, NULL);
 773         msg = skip_empty_lines(msg);
 774         c->body_off = msg - start;
 775
 776         c->commit_message_parsed = 1;
 777 }
 778
 779 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 780 {
 781         struct name_decoration *d;
 782         const char *prefix = " (";
 783
 784         load_ref_decorations(DECORATE_SHORT_REFS);
 785         d = lookup_decoration(&name_decoration, &commit->object);
 786         while (d) {
 787                 strbuf_addstr(sb, prefix);
 788                 prefix = ", ";
 789                 strbuf_addstr(sb, d->name);
 790                 d = d->next;
 791         }
 792         if (prefix[0] == ',')
 793                 strbuf_addch(sb, ')');
 794 }
 795
 796 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 797                         size_t width, size_t indent1, size_t indent2)
 798 {
 799         struct strbuf tmp = STRBUF_INIT;
 800
 801         if (pos)
 802                 strbuf_add(&tmp, sb->buf, pos);
 803         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 804                                 (int) indent1, (int) indent2, (int) width);
 805         strbuf_swap(&tmp, sb);
 806         strbuf_release(&tmp);
 807 }
 808
 809 static void rewrap_message_tail(struct strbuf *sb,
 810                                 struct format_commit_context *c,
 811                                 size_t new_width, size_t new_indent1,
 812                                 size_t new_indent2)
 813 {
 814         if (c->width == new_width && c->indent1 == new_indent1 &&
 815             c->indent2 == new_indent2)
 816                 return;
 817         if (c->wrap_start < sb->len)
 818                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 819         c->wrap_start = sb->len;
 820         c->width = new_width;
 821         c->indent1 = new_indent1;
 822         c->indent2 = new_indent2;
 823 }
 824
 825 static int format_reflog_person(struct strbuf *sb,
 826                                 char part,
 827                                 struct reflog_walk_info *log,
 828                                 enum date_mode dmode)
 829 {
 830         const char *ident;
 831
 832         if (!log)
 833                 return 2;
 834
 835         ident = get_reflog_ident(log);
 836         if (!ident)
 837                 return 2;
 838
 839         return format_person_part(sb, part, ident, strlen(ident), dmode);
 840 }
 841
 842 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 843                                 void *context)
 844 {
 845         struct format_commit_context *c = context;
 846         const struct commit *commit = c->commit;
 847         const char *msg = c->message;
 848         struct commit_list *p;
 849         int h1, h2;
 850
 851         /* these are independent of the commit */
 852         switch (placeholder[0]) {
 853         case 'C':
 854                 if (placeholder[1] == '(') {
 855                         const char *end = strchr(placeholder + 2, ')');
 856                         char color[COLOR_MAXLEN];
 857                         if (!end)
 858                                 return 0;
 859                         color_parse_mem(placeholder + 2,
 860                                         end - (placeholder + 2),
 861                                         "--pretty format", color);
 862                         strbuf_addstr(sb, color);
 863                         return end - placeholder + 1;
 864                 }
 865                 if (!prefixcmp(placeholder + 1, "red")) {
 866                         strbuf_addstr(sb, GIT_COLOR_RED);
 867                         return 4;
 868                 } else if (!prefixcmp(placeholder + 1, "green")) {
 869                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 870                         return 6;
 871                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 872                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 873                         return 5;
 874                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 875                         strbuf_addstr(sb, GIT_COLOR_RESET);
 876                         return 6;
 877                 } else
 878                         return 0;
 879         case 'n':               /* newline */
 880                 strbuf_addch(sb, '\n');
 881                 return 1;
 882         case 'x':
 883                 /* %x00 == NUL, %x0a == LF, etc. */
 884                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 885                     h1 <= 16 &&
 886                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 887                     h2 <= 16) {
 888                         strbuf_addch(sb, (h1<<4)|h2);
 889                         return 3;
 890                 } else
 891                         return 0;
 892         case 'w':
 893                 if (placeholder[1] == '(') {
 894                         unsigned long width = 0, indent1 = 0, indent2 = 0;
 895                         char *next;
 896                         const char *start = placeholder + 2;
 897                         const char *end = strchr(start, ')');
 898                         if (!end)
 899                                 return 0;
 900                         if (end > start) {
 901                                 width = strtoul(start, &next, 10);
 902                                 if (*next == ',') {
 903                                         indent1 = strtoul(next + 1, &next, 10);
 904                                         if (*next == ',') {
 905                                                 indent2 = strtoul(next + 1,
 906                                                                  &next, 10);
 907                                         }
 908                                 }
 909                                 if (*next != ')')
 910                                         return 0;
 911                         }
 912                         rewrap_message_tail(sb, c, width, indent1, indent2);
 913                         return end - placeholder + 1;
 914                 } else
 915                         return 0;
 916         }
 917
 918         /* these depend on the commit */
 919         if (!commit->object.parsed)
 920                 parse_object(commit->object.sha1);
 921
 922         switch (placeholder[0]) {
 923         case 'H':               /* commit hash */
 924                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 925                 return 1;
 926         case 'h':               /* abbreviated commit hash */
 927                 if (add_again(sb, &c->abbrev_commit_hash))
 928                         return 1;
 929                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 930                                                      c->pretty_ctx->abbrev));
 931                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 932                 return 1;
 933         case 'T':               /* tree hash */
 934                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 935                 return 1;
 936         case 't':               /* abbreviated tree hash */
 937                 if (add_again(sb, &c->abbrev_tree_hash))
 938                         return 1;
 939                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 940                                                      c->pretty_ctx->abbrev));
 941                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 942                 return 1;
 943         case 'P':               /* parent hashes */
 944                 for (p = commit->parents; p; p = p->next) {
 945                         if (p != commit->parents)
 946                                 strbuf_addch(sb, ' ');
 947                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 948                 }
 949                 return 1;
 950         case 'p':               /* abbreviated parent hashes */
 951                 if (add_again(sb, &c->abbrev_parent_hashes))
 952                         return 1;
 953                 for (p = commit->parents; p; p = p->next) {
 954                         if (p != commit->parents)
 955                                 strbuf_addch(sb, ' ');
 956                         strbuf_addstr(sb, find_unique_abbrev(
 957                                         p->item->object.sha1,
 958                                         c->pretty_ctx->abbrev));
 959                 }
 960                 c->abbrev_parent_hashes.len = sb->len -
 961                                               c->abbrev_parent_hashes.off;
 962                 return 1;
 963         case 'm':               /* left/right/bottom */
 964                 strbuf_addstr(sb, get_revision_mark(NULL, commit));
 965                 return 1;
 966         case 'd':
 967                 format_decoration(sb, commit);
 968                 return 1;
 969         case 'g':               /* reflog info */
 970                 switch(placeholder[1]) {
 971                 case 'd':       /* reflog selector */
 972                 case 'D':
 973                         if (c->pretty_ctx->reflog_info)
 974                                 get_reflog_selector(sb,
 975                                                     c->pretty_ctx->reflog_info,
 976                                                     c->pretty_ctx->date_mode,
 977                                                     (placeholder[1] == 'd'));
 978                         return 2;
 979                 case 's':       /* reflog message */
 980                         if (c->pretty_ctx->reflog_info)
 981                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
 982                         return 2;
 983                 case 'n':
 984                 case 'N':
 985                 case 'e':
 986                 case 'E':
 987                         return format_reflog_person(sb,
 988                                                     placeholder[1],
 989                                                     c->pretty_ctx->reflog_info,
 990                                                     c->pretty_ctx->date_mode);
 991                 }
 992                 return 0;       /* unknown %g placeholder */
 993         case 'N':
 994                 if (c->pretty_ctx->show_notes) {
 995                         format_display_notes(commit->object.sha1, sb,
 996                                     get_log_output_encoding(), 0);
 997                         return 1;
 998                 }
 999                 return 0;
1000         }
1001
1002         /* For the rest we have to parse the commit header. */
1003         if (!c->commit_header_parsed)
1004                 parse_commit_header(c);
1005
1006         switch (placeholder[0]) {
1007         case 'a':       /* author ... */
1008                 return format_person_part(sb, placeholder[1],
1009                                    msg + c->author.off, c->author.len,
1010                                    c->pretty_ctx->date_mode);
1011         case 'c':       /* committer ... */
1012                 return format_person_part(sb, placeholder[1],
1013                                    msg + c->committer.off, c->committer.len,
1014                                    c->pretty_ctx->date_mode);
1015         case 'e':       /* encoding */
1016                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1017                 return 1;
1018         case 'B':       /* raw body */
1019                 /* message_off is always left at the initial newline */
1020                 strbuf_addstr(sb, msg + c->message_off + 1);
1021                 return 1;
1022         }
1023
1024         /* Now we need to parse the commit message. */
1025         if (!c->commit_message_parsed)
1026                 parse_commit_message(c);
1027
1028         switch (placeholder[0]) {
1029         case 's':       /* subject */
1030                 format_subject(sb, msg + c->subject_off, " ");
1031                 return 1;
1032         case 'f':       /* sanitized subject */
1033                 format_sanitized_subject(sb, msg + c->subject_off);
1034                 return 1;
1035         case 'b':       /* body */
1036                 strbuf_addstr(sb, msg + c->body_off);
1037                 return 1;
1038         }
1039         return 0;       /* unknown placeholder */
1040 }
1041
1042 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1043                                  void *context)
1044 {
1045         int consumed;
1046         size_t orig_len;
1047         enum {
1048                 NO_MAGIC,
1049                 ADD_LF_BEFORE_NON_EMPTY,
1050                 DEL_LF_BEFORE_EMPTY,
1051                 ADD_SP_BEFORE_NON_EMPTY
1052         } magic = NO_MAGIC;
1053
1054         switch (placeholder[0]) {
1055         case '-':
1056                 magic = DEL_LF_BEFORE_EMPTY;
1057                 break;
1058         case '+':
1059                 magic = ADD_LF_BEFORE_NON_EMPTY;
1060                 break;
1061         case ' ':
1062                 magic = ADD_SP_BEFORE_NON_EMPTY;
1063                 break;
1064         default:
1065                 break;
1066         }
1067         if (magic != NO_MAGIC)
1068                 placeholder++;
1069
1070         orig_len = sb->len;
1071         consumed = format_commit_one(sb, placeholder, context);
1072         if (magic == NO_MAGIC)
1073                 return consumed;
1074
1075         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1076                 while (sb->len && sb->buf[sb->len - 1] == '\n')
1077                         strbuf_setlen(sb, sb->len - 1);
1078         } else if (orig_len != sb->len) {
1079                 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1080                         strbuf_insert(sb, orig_len, "\n", 1);
1081                 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1082                         strbuf_insert(sb, orig_len, " ", 1);
1083         }
1084         return consumed + 1;
1085 }
1086
1087 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1088                                    void *context)
1089 {
1090         struct userformat_want *w = context;
1091
1092         if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1093                 placeholder++;
1094
1095         switch (*placeholder) {
1096         case 'N':
1097                 w->notes = 1;
1098                 break;
1099         }
1100         return 0;
1101 }
1102
1103 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1104 {
1105         struct strbuf dummy = STRBUF_INIT;
1106
1107         if (!fmt) {
1108                 if (!user_format)
1109                         return;
1110                 fmt = user_format;
1111         }
1112         strbuf_expand(&dummy, fmt, userformat_want_item, w);
1113         strbuf_release(&dummy);
1114 }
1115
1116 void format_commit_message(const struct commit *commit,
1117                            const char *format, struct strbuf *sb,
1118                            const struct pretty_print_context *pretty_ctx)
1119 {
1120         struct format_commit_context context;
1121         static const char utf8[] = "UTF-8";
1122         const char *output_enc = pretty_ctx->output_encoding;
1123
1124         memset(&context, 0, sizeof(context));
1125         context.commit = commit;
1126         context.pretty_ctx = pretty_ctx;
1127         context.wrap_start = sb->len;
1128         context.message = commit->buffer;
1129         if (output_enc) {
1130                 char *enc = get_header(commit, "encoding");
1131                 if (strcmp(enc ? enc : utf8, output_enc)) {
1132                         context.message = logmsg_reencode(commit, output_enc);
1133                         if (!context.message)
1134                                 context.message = commit->buffer;
1135                 }
1136                 free(enc);
1137         }
1138
1139         strbuf_expand(sb, format, format_commit_item, &context);
1140         rewrap_message_tail(sb, &context, 0, 0, 0);
1141
1142         if (context.message != commit->buffer)
1143                 free(context.message);
1144 }
1145
1146 static void pp_header(const struct pretty_print_context *pp,
1147                       const char *encoding,
1148                       const struct commit *commit,
1149                       const char **msg_p,
1150                       struct strbuf *sb)
1151 {
1152         int parents_shown = 0;
1153
1154         for (;;) {
1155                 const char *line = *msg_p;
1156                 int linelen = get_one_line(*msg_p);
1157
1158                 if (!linelen)
1159                         return;
1160                 *msg_p += linelen;
1161
1162                 if (linelen == 1)
1163                         /* End of header */
1164                         return;
1165
1166                 if (pp->fmt == CMIT_FMT_RAW) {
1167                         strbuf_add(sb, line, linelen);
1168                         continue;
1169                 }
1170
1171                 if (!memcmp(line, "parent ", 7)) {
1172                         if (linelen != 48)
1173                                 die("bad parent line in commit");
1174                         continue;
1175                 }
1176
1177                 if (!parents_shown) {
1178                         struct commit_list *parent;
1179                         int num;
1180                         for (parent = commit->parents, num = 0;
1181                              parent;
1182                              parent = parent->next, num++)
1183                                 ;
1184                         /* with enough slop */
1185                         strbuf_grow(sb, num * 50 + 20);
1186                         add_merge_info(pp, sb, commit);
1187                         parents_shown = 1;
1188                 }
1189
1190                 /*
1191                  * MEDIUM == DEFAULT shows only author with dates.
1192                  * FULL shows both authors but not dates.
1193                  * FULLER shows both authors and dates.
1194                  */
1195                 if (!memcmp(line, "author ", 7)) {
1196                         strbuf_grow(sb, linelen + 80);
1197                         pp_user_info(pp, "Author", sb, line + 7, encoding);
1198                 }
1199                 if (!memcmp(line, "committer ", 10) &&
1200                     (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1201                         strbuf_grow(sb, linelen + 80);
1202                         pp_user_info(pp, "Commit", sb, line + 10, encoding);
1203                 }
1204         }
1205 }
1206
1207 void pp_title_line(const struct pretty_print_context *pp,
1208                    const char **msg_p,
1209                    struct strbuf *sb,
1210                    const char *encoding,
1211                    int need_8bit_cte)
1212 {
1213         struct strbuf title;
1214
1215         strbuf_init(&title, 80);
1216         *msg_p = format_subject(&title, *msg_p,
1217                                 pp->preserve_subject ? "\n" : " ");
1218
1219         strbuf_grow(sb, title.len + 1024);
1220         if (pp->subject) {
1221                 strbuf_addstr(sb, pp->subject);
1222                 add_rfc2047(sb, title.buf, title.len, encoding);
1223         } else {
1224                 strbuf_addbuf(sb, &title);
1225         }
1226         strbuf_addch(sb, '\n');
1227
1228         if (need_8bit_cte > 0) {
1229                 const char *header_fmt =
1230                         "MIME-Version: 1.0\n"
1231                         "Content-Type: text/plain; charset=%s\n"
1232                         "Content-Transfer-Encoding: 8bit\n";
1233                 strbuf_addf(sb, header_fmt, encoding);
1234         }
1235         if (pp->after_subject) {
1236                 strbuf_addstr(sb, pp->after_subject);
1237         }
1238         if (pp->fmt == CMIT_FMT_EMAIL) {
1239                 strbuf_addch(sb, '\n');
1240         }
1241         strbuf_release(&title);
1242 }
1243
1244 void pp_remainder(const struct pretty_print_context *pp,
1245                   const char **msg_p,
1246                   struct strbuf *sb,
1247                   int indent)
1248 {
1249         int first = 1;
1250         for (;;) {
1251                 const char *line = *msg_p;
1252                 int linelen = get_one_line(line);
1253                 *msg_p += linelen;
1254
1255                 if (!linelen)
1256                         break;
1257
1258                 if (is_empty_line(line, &linelen)) {
1259                         if (first)
1260                                 continue;
1261                         if (pp->fmt == CMIT_FMT_SHORT)
1262                                 break;
1263                 }
1264                 first = 0;
1265
1266                 strbuf_grow(sb, linelen + indent + 20);
1267                 if (indent) {
1268                         memset(sb->buf + sb->len, ' ', indent);
1269                         strbuf_setlen(sb, sb->len + indent);
1270                 }
1271                 strbuf_add(sb, line, linelen);
1272                 strbuf_addch(sb, '\n');
1273         }
1274 }
1275
1276 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1277 {
1278         const char *encoding;
1279
1280         encoding = get_log_output_encoding();
1281         if (encoding_p)
1282                 *encoding_p = encoding;
1283         return logmsg_reencode(commit, encoding);
1284 }
1285
1286 void pretty_print_commit(const struct pretty_print_context *pp,
1287                          const struct commit *commit,
1288                          struct strbuf *sb)
1289 {
1290         unsigned long beginning_of_body;
1291         int indent = 4;
1292         const char *msg = commit->buffer;
1293         char *reencoded;
1294         const char *encoding;
1295         int need_8bit_cte = pp->need_8bit_cte;
1296
1297         if (pp->fmt == CMIT_FMT_USERFORMAT) {
1298                 format_commit_message(commit, user_format, sb, pp);
1299                 return;
1300         }
1301
1302         reencoded = reencode_commit_message(commit, &encoding);
1303         if (reencoded) {
1304                 msg = reencoded;
1305         }
1306
1307         if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1308                 indent = 0;
1309
1310         /*
1311          * We need to check and emit Content-type: to mark it
1312          * as 8-bit if we haven't done so.
1313          */
1314         if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1315                 int i, ch, in_body;
1316
1317                 for (in_body = i = 0; (ch = msg[i]); i++) {
1318                         if (!in_body) {
1319                                 /* author could be non 7-bit ASCII but
1320                                  * the log may be so; skip over the
1321                                  * header part first.
1322                                  */
1323                                 if (ch == '\n' && msg[i+1] == '\n')
1324                                         in_body = 1;
1325                         }
1326                         else if (non_ascii(ch)) {
1327                                 need_8bit_cte = 1;
1328                                 break;
1329                         }
1330                 }
1331         }
1332
1333         pp_header(pp, encoding, commit, &msg, sb);
1334         if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1335                 strbuf_addch(sb, '\n');
1336         }
1337
1338         /* Skip excess blank lines at the beginning of body, if any... */
1339         msg = skip_empty_lines(msg);
1340
1341         /* These formats treat the title line specially. */
1342         if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1343                 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1344
1345         beginning_of_body = sb->len;
1346         if (pp->fmt != CMIT_FMT_ONELINE)
1347                 pp_remainder(pp, &msg, sb, indent);
1348         strbuf_rtrim(sb);
1349
1350         /* Make sure there is an EOLN for the non-oneline case */
1351         if (pp->fmt != CMIT_FMT_ONELINE)
1352                 strbuf_addch(sb, '\n');
1353
1354         /*
1355          * The caller may append additional body text in e-mail
1356          * format.  Make sure we did not strip the blank line
1357          * between the header and the body.
1358          */
1359         if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1360                 strbuf_addch(sb, '\n');
1361
1362         if (pp->show_notes)
1363                 format_display_notes(commit->object.sha1, sb, encoding,
1364                                      NOTES_SHOW_HEADER | NOTES_INDENT);
1365
1366         free(reencoded);
1367 }
1368
1369 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1370                     struct strbuf *sb)
1371 {
1372         struct pretty_print_context pp = {0};
1373         pp.fmt = fmt;
1374         pretty_print_commit(&pp, commit, sb);
1375 }