Andrew's git / gitweb.git / pretty.c
?
pretty.con commit merge script: handle many-way octopus (d9a680a)
   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
  13static char *user_format;
  14static 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;
  21static size_t builtin_formats_len;
  22static size_t commit_formats_len;
  23static size_t commit_formats_alloc;
  24static struct cmt_fmt_map *find_commit_format(const char *sought);
  25
  26static 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
  35static 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
  81static 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
 101static 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
 136static 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
 144void 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 */
 178static 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 */
 194static int non_ascii(int ch)
 195{
 196        return !isascii(ch) || ch == '\033';
 197}
 198
 199int 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
 211static int is_rfc2047_special(char ch)
 212{
 213        return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 214}
 215
 216static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 217                       const char *encoding)
 218{
 219        int i, last;
 220
 221        for (i = 0; i < len; i++) {
 222                int ch = line[i];
 223                if (non_ascii(ch))
 224                        goto needquote;
 225                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 226                        goto needquote;
 227        }
 228        strbuf_add(sb, line, len);
 229        return;
 230
 231needquote:
 232        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 233        strbuf_addf(sb, "=?%s?q?", encoding);
 234        for (i = last = 0; i < len; i++) {
 235                unsigned ch = line[i] & 0xFF;
 236                /*
 237                 * We encode ' ' using '=20' even though rfc2047
 238                 * allows using '_' for readability.  Unfortunately,
 239                 * many programs do not understand this and just
 240                 * leave the underscore in place.
 241                 */
 242                if (is_rfc2047_special(ch) || ch == ' ') {
 243                        strbuf_add(sb, line + last, i - last);
 244                        strbuf_addf(sb, "=%02X", ch);
 245                        last = i + 1;
 246                }
 247        }
 248        strbuf_add(sb, line + last, len - last);
 249        strbuf_addstr(sb, "?=");
 250}
 251
 252void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 253                  const char *line, enum date_mode dmode,
 254                  const char *encoding)
 255{
 256        char *date;
 257        int namelen;
 258        unsigned long time;
 259        int tz;
 260
 261        if (fmt == CMIT_FMT_ONELINE)
 262                return;
 263        date = strchr(line, '>');
 264        if (!date)
 265                return;
 266        namelen = ++date - line;
 267        time = strtoul(date, &date, 10);
 268        tz = strtol(date, NULL, 10);
 269
 270        if (fmt == CMIT_FMT_EMAIL) {
 271                char *name_tail = strchr(line, '<');
 272                int display_name_length;
 273                if (!name_tail)
 274                        return;
 275                while (line < name_tail && isspace(name_tail[-1]))
 276                        name_tail--;
 277                display_name_length = name_tail - line;
 278                strbuf_addstr(sb, "From: ");
 279                add_rfc2047(sb, line, display_name_length, encoding);
 280                strbuf_add(sb, name_tail, namelen - display_name_length);
 281                strbuf_addch(sb, '\n');
 282        } else {
 283                strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 284                              (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 285                              "    ", namelen, line);
 286        }
 287        switch (fmt) {
 288        case CMIT_FMT_MEDIUM:
 289                strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 290                break;
 291        case CMIT_FMT_EMAIL:
 292                strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 293                break;
 294        case CMIT_FMT_FULLER:
 295                strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 296                break;
 297        default:
 298                /* notin' */
 299                break;
 300        }
 301}
 302
 303static int is_empty_line(const char *line, int *len_p)
 304{
 305        int len = *len_p;
 306        while (len && isspace(line[len-1]))
 307                len--;
 308        *len_p = len;
 309        return !len;
 310}
 311
 312static const char *skip_empty_lines(const char *msg)
 313{
 314        for (;;) {
 315                int linelen = get_one_line(msg);
 316                int ll = linelen;
 317                if (!linelen)
 318                        break;
 319                if (!is_empty_line(msg, &ll))
 320                        break;
 321                msg += linelen;
 322        }
 323        return msg;
 324}
 325
 326static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 327                        const struct commit *commit, int abbrev)
 328{
 329        struct commit_list *parent = commit->parents;
 330
 331        if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 332            !parent || !parent->next)
 333                return;
 334
 335        strbuf_addstr(sb, "Merge:");
 336
 337        while (parent) {
 338                struct commit *p = parent->item;
 339                const char *hex = NULL;
 340                if (abbrev)
 341                        hex = find_unique_abbrev(p->object.sha1, abbrev);
 342                if (!hex)
 343                        hex = sha1_to_hex(p->object.sha1);
 344                parent = parent->next;
 345
 346                strbuf_addf(sb, " %s", hex);
 347        }
 348        strbuf_addch(sb, '\n');
 349}
 350
 351static char *get_header(const struct commit *commit, const char *key)
 352{
 353        int key_len = strlen(key);
 354        const char *line = commit->buffer;
 355
 356        for (;;) {
 357                const char *eol = strchr(line, '\n'), *next;
 358
 359                if (line == eol)
 360                        return NULL;
 361                if (!eol) {
 362                        eol = line + strlen(line);
 363                        next = NULL;
 364                } else
 365                        next = eol + 1;
 366                if (eol - line > key_len &&
 367                    !strncmp(line, key, key_len) &&
 368                    line[key_len] == ' ') {
 369                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 370                }
 371                line = next;
 372        }
 373}
 374
 375static char *replace_encoding_header(char *buf, const char *encoding)
 376{
 377        struct strbuf tmp = STRBUF_INIT;
 378        size_t start, len;
 379        char *cp = buf;
 380
 381        /* guess if there is an encoding header before a \n\n */
 382        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 383                cp = strchr(cp, '\n');
 384                if (!cp || *++cp == '\n')
 385                        return buf;
 386        }
 387        start = cp - buf;
 388        cp = strchr(cp, '\n');
 389        if (!cp)
 390                return buf; /* should not happen but be defensive */
 391        len = cp + 1 - (buf + start);
 392
 393        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 394        if (is_encoding_utf8(encoding)) {
 395                /* we have re-coded to UTF-8; drop the header */
 396                strbuf_remove(&tmp, start, len);
 397        } else {
 398                /* just replaces XXXX in 'encoding XXXX\n' */
 399                strbuf_splice(&tmp, start + strlen("encoding "),
 400                                          len - strlen("encoding \n"),
 401                                          encoding, strlen(encoding));
 402        }
 403        return strbuf_detach(&tmp, NULL);
 404}
 405
 406static char *logmsg_reencode(const struct commit *commit,
 407                             const char *output_encoding)
 408{
 409        static const char *utf8 = "UTF-8";
 410        const char *use_encoding;
 411        char *encoding;
 412        char *out;
 413
 414        if (!*output_encoding)
 415                return NULL;
 416        encoding = get_header(commit, "encoding");
 417        use_encoding = encoding ? encoding : utf8;
 418        if (!strcmp(use_encoding, output_encoding))
 419                if (encoding) /* we'll strip encoding header later */
 420                        out = xstrdup(commit->buffer);
 421                else
 422                        return NULL; /* nothing to do */
 423        else
 424                out = reencode_string(commit->buffer,
 425                                      output_encoding, use_encoding);
 426        if (out)
 427                out = replace_encoding_header(out, output_encoding);
 428
 429        free(encoding);
 430        return out;
 431}
 432
 433static int mailmap_name(char *email, int email_len, char *name, int name_len)
 434{
 435        static struct string_list *mail_map;
 436        if (!mail_map) {
 437                mail_map = xcalloc(1, sizeof(*mail_map));
 438                read_mailmap(mail_map, NULL);
 439        }
 440        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 441}
 442
 443static size_t format_person_part(struct strbuf *sb, char part,
 444                                 const char *msg, int len, enum date_mode dmode)
 445{
 446        /* currently all placeholders have same length */
 447        const int placeholder_len = 2;
 448        int start, end, tz = 0;
 449        unsigned long date = 0;
 450        char *ep;
 451        const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 452        char person_name[1024];
 453        char person_mail[1024];
 454
 455        /* advance 'end' to point to email start delimiter */
 456        for (end = 0; end < len && msg[end] != '<'; end++)
 457                ; /* do nothing */
 458
 459        /*
 460         * When end points at the '<' that we found, it should have
 461         * matching '>' later, which means 'end' must be strictly
 462         * below len - 1.
 463         */
 464        if (end >= len - 2)
 465                goto skip;
 466
 467        /* Seek for both name and email part */
 468        name_start = msg;
 469        name_end = msg+end;
 470        while (name_end > name_start && isspace(*(name_end-1)))
 471                name_end--;
 472        mail_start = msg+end+1;
 473        mail_end = mail_start;
 474        while (mail_end < msg_end && *mail_end != '>')
 475                mail_end++;
 476        if (mail_end == msg_end)
 477                goto skip;
 478        end = mail_end-msg;
 479
 480        if (part == 'N' || part == 'E') { /* mailmap lookup */
 481                strlcpy(person_name, name_start, name_end-name_start+1);
 482                strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 483                mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 484                name_start = person_name;
 485                name_end = name_start + strlen(person_name);
 486                mail_start = person_mail;
 487                mail_end = mail_start +  strlen(person_mail);
 488        }
 489        if (part == 'n' || part == 'N') {       /* name */
 490                strbuf_add(sb, name_start, name_end-name_start);
 491                return placeholder_len;
 492        }
 493        if (part == 'e' || part == 'E') {       /* email */
 494                strbuf_add(sb, mail_start, mail_end-mail_start);
 495                return placeholder_len;
 496        }
 497
 498        /* advance 'start' to point to date start delimiter */
 499        for (start = end + 1; start < len && isspace(msg[start]); start++)
 500                ; /* do nothing */
 501        if (start >= len)
 502                goto skip;
 503        date = strtoul(msg + start, &ep, 10);
 504        if (msg + start == ep)
 505                goto skip;
 506
 507        if (part == 't') {      /* date, UNIX timestamp */
 508                strbuf_add(sb, msg + start, ep - (msg + start));
 509                return placeholder_len;
 510        }
 511
 512        /* parse tz */
 513        for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 514                ; /* do nothing */
 515        if (start + 1 < len) {
 516                tz = strtoul(msg + start + 1, NULL, 10);
 517                if (msg[start] == '-')
 518                        tz = -tz;
 519        }
 520
 521        switch (part) {
 522        case 'd':       /* date */
 523                strbuf_addstr(sb, show_date(date, tz, dmode));
 524                return placeholder_len;
 525        case 'D':       /* date, RFC2822 style */
 526                strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 527                return placeholder_len;
 528        case 'r':       /* date, relative */
 529                strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 530                return placeholder_len;
 531        case 'i':       /* date, ISO 8601 */
 532                strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 533                return placeholder_len;
 534        }
 535
 536skip:
 537        /*
 538         * bogus commit, 'sb' cannot be updated, but we still need to
 539         * compute a valid return value.
 540         */
 541        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 542            || part == 'D' || part == 'r' || part == 'i')
 543                return placeholder_len;
 544
 545        return 0; /* unknown placeholder */
 546}
 547
 548struct chunk {
 549        size_t off;
 550        size_t len;
 551};
 552
 553struct format_commit_context {
 554        const struct commit *commit;
 555        const struct pretty_print_context *pretty_ctx;
 556        unsigned commit_header_parsed:1;
 557        unsigned commit_message_parsed:1;
 558        size_t width, indent1, indent2;
 559
 560        /* These offsets are relative to the start of the commit message. */
 561        struct chunk author;
 562        struct chunk committer;
 563        struct chunk encoding;
 564        size_t message_off;
 565        size_t subject_off;
 566        size_t body_off;
 567
 568        /* The following ones are relative to the result struct strbuf. */
 569        struct chunk abbrev_commit_hash;
 570        struct chunk abbrev_tree_hash;
 571        struct chunk abbrev_parent_hashes;
 572        size_t wrap_start;
 573};
 574
 575static int add_again(struct strbuf *sb, struct chunk *chunk)
 576{
 577        if (chunk->len) {
 578                strbuf_adddup(sb, chunk->off, chunk->len);
 579                return 1;
 580        }
 581
 582        /*
 583         * We haven't seen this chunk before.  Our caller is surely
 584         * going to add it the hard way now.  Remember the most likely
 585         * start of the to-be-added chunk: the current end of the
 586         * struct strbuf.
 587         */
 588        chunk->off = sb->len;
 589        return 0;
 590}
 591
 592static void parse_commit_header(struct format_commit_context *context)
 593{
 594        const char *msg = context->commit->buffer;
 595        int i;
 596
 597        for (i = 0; msg[i]; i++) {
 598                int eol;
 599                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 600                        ; /* do nothing */
 601
 602                if (i == eol) {
 603                        break;
 604                } else if (!prefixcmp(msg + i, "author ")) {
 605                        context->author.off = i + 7;
 606                        context->author.len = eol - i - 7;
 607                } else if (!prefixcmp(msg + i, "committer ")) {
 608                        context->committer.off = i + 10;
 609                        context->committer.len = eol - i - 10;
 610                } else if (!prefixcmp(msg + i, "encoding ")) {
 611                        context->encoding.off = i + 9;
 612                        context->encoding.len = eol - i - 9;
 613                }
 614                i = eol;
 615        }
 616        context->message_off = i;
 617        context->commit_header_parsed = 1;
 618}
 619
 620static int istitlechar(char c)
 621{
 622        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 623                (c >= '0' && c <= '9') || c == '.' || c == '_';
 624}
 625
 626static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 627{
 628        size_t trimlen;
 629        size_t start_len = sb->len;
 630        int space = 2;
 631
 632        for (; *msg && *msg != '\n'; msg++) {
 633                if (istitlechar(*msg)) {
 634                        if (space == 1)
 635                                strbuf_addch(sb, '-');
 636                        space = 0;
 637                        strbuf_addch(sb, *msg);
 638                        if (*msg == '.')
 639                                while (*(msg+1) == '.')
 640                                        msg++;
 641                } else
 642                        space |= 1;
 643        }
 644
 645        /* trim any trailing '.' or '-' characters */
 646        trimlen = 0;
 647        while (sb->len - trimlen > start_len &&
 648                (sb->buf[sb->len - 1 - trimlen] == '.'
 649                || sb->buf[sb->len - 1 - trimlen] == '-'))
 650                trimlen++;
 651        strbuf_remove(sb, sb->len - trimlen, trimlen);
 652}
 653
 654const char *format_subject(struct strbuf *sb, const char *msg,
 655                           const char *line_separator)
 656{
 657        int first = 1;
 658
 659        for (;;) {
 660                const char *line = msg;
 661                int linelen = get_one_line(line);
 662
 663                msg += linelen;
 664                if (!linelen || is_empty_line(line, &linelen))
 665                        break;
 666
 667                if (!sb)
 668                        continue;
 669                strbuf_grow(sb, linelen + 2);
 670                if (!first)
 671                        strbuf_addstr(sb, line_separator);
 672                strbuf_add(sb, line, linelen);
 673                first = 0;
 674        }
 675        return msg;
 676}
 677
 678static void parse_commit_message(struct format_commit_context *c)
 679{
 680        const char *msg = c->commit->buffer + c->message_off;
 681        const char *start = c->commit->buffer;
 682
 683        msg = skip_empty_lines(msg);
 684        c->subject_off = msg - start;
 685
 686        msg = format_subject(NULL, msg, NULL);
 687        msg = skip_empty_lines(msg);
 688        c->body_off = msg - start;
 689
 690        c->commit_message_parsed = 1;
 691}
 692
 693static void format_decoration(struct strbuf *sb, const struct commit *commit)
 694{
 695        struct name_decoration *d;
 696        const char *prefix = " (";
 697
 698        load_ref_decorations(DECORATE_SHORT_REFS);
 699        d = lookup_decoration(&name_decoration, &commit->object);
 700        while (d) {
 701                strbuf_addstr(sb, prefix);
 702                prefix = ", ";
 703                strbuf_addstr(sb, d->name);
 704                d = d->next;
 705        }
 706        if (prefix[0] == ',')
 707                strbuf_addch(sb, ')');
 708}
 709
 710static void strbuf_wrap(struct strbuf *sb, size_t pos,
 711                        size_t width, size_t indent1, size_t indent2)
 712{
 713        struct strbuf tmp = STRBUF_INIT;
 714
 715        if (pos)
 716                strbuf_add(&tmp, sb->buf, pos);
 717        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 718                                (int) indent1, (int) indent2, (int) width);
 719        strbuf_swap(&tmp, sb);
 720        strbuf_release(&tmp);
 721}
 722
 723static void rewrap_message_tail(struct strbuf *sb,
 724                                struct format_commit_context *c,
 725                                size_t new_width, size_t new_indent1,
 726                                size_t new_indent2)
 727{
 728        if (c->width == new_width && c->indent1 == new_indent1 &&
 729            c->indent2 == new_indent2)
 730                return;
 731        if (c->wrap_start < sb->len)
 732                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 733        c->wrap_start = sb->len;
 734        c->width = new_width;
 735        c->indent1 = new_indent1;
 736        c->indent2 = new_indent2;
 737}
 738
 739static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 740                                void *context)
 741{
 742        struct format_commit_context *c = context;
 743        const struct commit *commit = c->commit;
 744        const char *msg = commit->buffer;
 745        struct commit_list *p;
 746        int h1, h2;
 747
 748        /* these are independent of the commit */
 749        switch (placeholder[0]) {
 750        case 'C':
 751                if (placeholder[1] == '(') {
 752                        const char *end = strchr(placeholder + 2, ')');
 753                        char color[COLOR_MAXLEN];
 754                        if (!end)
 755                                return 0;
 756                        color_parse_mem(placeholder + 2,
 757                                        end - (placeholder + 2),
 758                                        "--pretty format", color);
 759                        strbuf_addstr(sb, color);
 760                        return end - placeholder + 1;
 761                }
 762                if (!prefixcmp(placeholder + 1, "red")) {
 763                        strbuf_addstr(sb, GIT_COLOR_RED);
 764                        return 4;
 765                } else if (!prefixcmp(placeholder + 1, "green")) {
 766                        strbuf_addstr(sb, GIT_COLOR_GREEN);
 767                        return 6;
 768                } else if (!prefixcmp(placeholder + 1, "blue")) {
 769                        strbuf_addstr(sb, GIT_COLOR_BLUE);
 770                        return 5;
 771                } else if (!prefixcmp(placeholder + 1, "reset")) {
 772                        strbuf_addstr(sb, GIT_COLOR_RESET);
 773                        return 6;
 774                } else
 775                        return 0;
 776        case 'n':               /* newline */
 777                strbuf_addch(sb, '\n');
 778                return 1;
 779        case 'x':
 780                /* %x00 == NUL, %x0a == LF, etc. */
 781                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 782                    h1 <= 16 &&
 783                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 784                    h2 <= 16) {
 785                        strbuf_addch(sb, (h1<<4)|h2);
 786                        return 3;
 787                } else
 788                        return 0;
 789        case 'w':
 790                if (placeholder[1] == '(') {
 791                        unsigned long width = 0, indent1 = 0, indent2 = 0;
 792                        char *next;
 793                        const char *start = placeholder + 2;
 794                        const char *end = strchr(start, ')');
 795                        if (!end)
 796                                return 0;
 797                        if (end > start) {
 798                                width = strtoul(start, &next, 10);
 799                                if (*next == ',') {
 800                                        indent1 = strtoul(next + 1, &next, 10);
 801                                        if (*next == ',') {
 802                                                indent2 = strtoul(next + 1,
 803                                                                 &next, 10);
 804                                        }
 805                                }
 806                                if (*next != ')')
 807                                        return 0;
 808                        }
 809                        rewrap_message_tail(sb, c, width, indent1, indent2);
 810                        return end - placeholder + 1;
 811                } else
 812                        return 0;
 813        }
 814
 815        /* these depend on the commit */
 816        if (!commit->object.parsed)
 817                parse_object(commit->object.sha1);
 818
 819        switch (placeholder[0]) {
 820        case 'H':               /* commit hash */
 821                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 822                return 1;
 823        case 'h':               /* abbreviated commit hash */
 824                if (add_again(sb, &c->abbrev_commit_hash))
 825                        return 1;
 826                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 827                                                     c->pretty_ctx->abbrev));
 828                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 829                return 1;
 830        case 'T':               /* tree hash */
 831                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 832                return 1;
 833        case 't':               /* abbreviated tree hash */
 834                if (add_again(sb, &c->abbrev_tree_hash))
 835                        return 1;
 836                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 837                                                     c->pretty_ctx->abbrev));
 838                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 839                return 1;
 840        case 'P':               /* parent hashes */
 841                for (p = commit->parents; p; p = p->next) {
 842                        if (p != commit->parents)
 843                                strbuf_addch(sb, ' ');
 844                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 845                }
 846                return 1;
 847        case 'p':               /* abbreviated parent hashes */
 848                if (add_again(sb, &c->abbrev_parent_hashes))
 849                        return 1;
 850                for (p = commit->parents; p; p = p->next) {
 851                        if (p != commit->parents)
 852                                strbuf_addch(sb, ' ');
 853                        strbuf_addstr(sb, find_unique_abbrev(
 854                                        p->item->object.sha1,
 855                                        c->pretty_ctx->abbrev));
 856                }
 857                c->abbrev_parent_hashes.len = sb->len -
 858                                              c->abbrev_parent_hashes.off;
 859                return 1;
 860        case 'm':               /* left/right/bottom */
 861                strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 862                                 ? '-'
 863                                 : (commit->object.flags & SYMMETRIC_LEFT)
 864                                 ? '<'
 865                                 : '>');
 866                return 1;
 867        case 'd':
 868                format_decoration(sb, commit);
 869                return 1;
 870        case 'g':               /* reflog info */
 871                switch(placeholder[1]) {
 872                case 'd':       /* reflog selector */
 873                case 'D':
 874                        if (c->pretty_ctx->reflog_info)
 875                                get_reflog_selector(sb,
 876                                                    c->pretty_ctx->reflog_info,
 877                                                    c->pretty_ctx->date_mode,
 878                                                    (placeholder[1] == 'd'));
 879                        return 2;
 880                case 's':       /* reflog message */
 881                        if (c->pretty_ctx->reflog_info)
 882                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
 883                        return 2;
 884                }
 885                return 0;       /* unknown %g placeholder */
 886        case 'N':
 887                if (c->pretty_ctx->show_notes) {
 888                        format_display_notes(commit->object.sha1, sb,
 889                                    git_log_output_encoding ? git_log_output_encoding
 890                                                            : git_commit_encoding, 0);
 891                        return 1;
 892                }
 893                return 0;
 894        }
 895
 896        /* For the rest we have to parse the commit header. */
 897        if (!c->commit_header_parsed)
 898                parse_commit_header(c);
 899
 900        switch (placeholder[0]) {
 901        case 'a':       /* author ... */
 902                return format_person_part(sb, placeholder[1],
 903                                   msg + c->author.off, c->author.len,
 904                                   c->pretty_ctx->date_mode);
 905        case 'c':       /* committer ... */
 906                return format_person_part(sb, placeholder[1],
 907                                   msg + c->committer.off, c->committer.len,
 908                                   c->pretty_ctx->date_mode);
 909        case 'e':       /* encoding */
 910                strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 911                return 1;
 912        case 'B':       /* raw body */
 913                /* message_off is always left at the initial newline */
 914                strbuf_addstr(sb, msg + c->message_off + 1);
 915                return 1;
 916        }
 917
 918        /* Now we need to parse the commit message. */
 919        if (!c->commit_message_parsed)
 920                parse_commit_message(c);
 921
 922        switch (placeholder[0]) {
 923        case 's':       /* subject */
 924                format_subject(sb, msg + c->subject_off, " ");
 925                return 1;
 926        case 'f':       /* sanitized subject */
 927                format_sanitized_subject(sb, msg + c->subject_off);
 928                return 1;
 929        case 'b':       /* body */
 930                strbuf_addstr(sb, msg + c->body_off);
 931                return 1;
 932        }
 933        return 0;       /* unknown placeholder */
 934}
 935
 936static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 937                                 void *context)
 938{
 939        int consumed;
 940        size_t orig_len;
 941        enum {
 942                NO_MAGIC,
 943                ADD_LF_BEFORE_NON_EMPTY,
 944                DEL_LF_BEFORE_EMPTY,
 945                ADD_SP_BEFORE_NON_EMPTY
 946        } magic = NO_MAGIC;
 947
 948        switch (placeholder[0]) {
 949        case '-':
 950                magic = DEL_LF_BEFORE_EMPTY;
 951                break;
 952        case '+':
 953                magic = ADD_LF_BEFORE_NON_EMPTY;
 954                break;
 955        case ' ':
 956                magic = ADD_SP_BEFORE_NON_EMPTY;
 957                break;
 958        default:
 959                break;
 960        }
 961        if (magic != NO_MAGIC)
 962                placeholder++;
 963
 964        orig_len = sb->len;
 965        consumed = format_commit_one(sb, placeholder, context);
 966        if (magic == NO_MAGIC)
 967                return consumed;
 968
 969        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 970                while (sb->len && sb->buf[sb->len - 1] == '\n')
 971                        strbuf_setlen(sb, sb->len - 1);
 972        } else if (orig_len != sb->len) {
 973                if (magic == ADD_LF_BEFORE_NON_EMPTY)
 974                        strbuf_insert(sb, orig_len, "\n", 1);
 975                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
 976                        strbuf_insert(sb, orig_len, " ", 1);
 977        }
 978        return consumed + 1;
 979}
 980
 981static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
 982                                   void *context)
 983{
 984        struct userformat_want *w = context;
 985
 986        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
 987                placeholder++;
 988
 989        switch (*placeholder) {
 990        case 'N':
 991                w->notes = 1;
 992                break;
 993        }
 994        return 0;
 995}
 996
 997void userformat_find_requirements(const char *fmt, struct userformat_want *w)
 998{
 999        struct strbuf dummy = STRBUF_INIT;
1000
1001        if (!fmt) {
1002                if (!user_format)
1003                        return;
1004                fmt = user_format;
1005        }
1006        strbuf_expand(&dummy, user_format, userformat_want_item, w);
1007        strbuf_release(&dummy);
1008}
1009
1010void format_commit_message(const struct commit *commit,
1011                           const char *format, struct strbuf *sb,
1012                           const struct pretty_print_context *pretty_ctx)
1013{
1014        struct format_commit_context context;
1015
1016        memset(&context, 0, sizeof(context));
1017        context.commit = commit;
1018        context.pretty_ctx = pretty_ctx;
1019        context.wrap_start = sb->len;
1020        strbuf_expand(sb, format, format_commit_item, &context);
1021        rewrap_message_tail(sb, &context, 0, 0, 0);
1022}
1023
1024static void pp_header(enum cmit_fmt fmt,
1025                      int abbrev,
1026                      enum date_mode dmode,
1027                      const char *encoding,
1028                      const struct commit *commit,
1029                      const char **msg_p,
1030                      struct strbuf *sb)
1031{
1032        int parents_shown = 0;
1033
1034        for (;;) {
1035                const char *line = *msg_p;
1036                int linelen = get_one_line(*msg_p);
1037
1038                if (!linelen)
1039                        return;
1040                *msg_p += linelen;
1041
1042                if (linelen == 1)
1043                        /* End of header */
1044                        return;
1045
1046                if (fmt == CMIT_FMT_RAW) {
1047                        strbuf_add(sb, line, linelen);
1048                        continue;
1049                }
1050
1051                if (!memcmp(line, "parent ", 7)) {
1052                        if (linelen != 48)
1053                                die("bad parent line in commit");
1054                        continue;
1055                }
1056
1057                if (!parents_shown) {
1058                        struct commit_list *parent;
1059                        int num;
1060                        for (parent = commit->parents, num = 0;
1061                             parent;
1062                             parent = parent->next, num++)
1063                                ;
1064                        /* with enough slop */
1065                        strbuf_grow(sb, num * 50 + 20);
1066                        add_merge_info(fmt, sb, commit, abbrev);
1067                        parents_shown = 1;
1068                }
1069
1070                /*
1071                 * MEDIUM == DEFAULT shows only author with dates.
1072                 * FULL shows both authors but not dates.
1073                 * FULLER shows both authors and dates.
1074                 */
1075                if (!memcmp(line, "author ", 7)) {
1076                        strbuf_grow(sb, linelen + 80);
1077                        pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
1078                }
1079                if (!memcmp(line, "committer ", 10) &&
1080                    (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
1081                        strbuf_grow(sb, linelen + 80);
1082                        pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
1083                }
1084        }
1085}
1086
1087void pp_title_line(enum cmit_fmt fmt,
1088                   const char **msg_p,
1089                   struct strbuf *sb,
1090                   const char *subject,
1091                   const char *after_subject,
1092                   const char *encoding,
1093                   int need_8bit_cte)
1094{
1095        const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
1096        struct strbuf title;
1097
1098        strbuf_init(&title, 80);
1099        *msg_p = format_subject(&title, *msg_p, line_separator);
1100
1101        strbuf_grow(sb, title.len + 1024);
1102        if (subject) {
1103                strbuf_addstr(sb, subject);
1104                add_rfc2047(sb, title.buf, title.len, encoding);
1105        } else {
1106                strbuf_addbuf(sb, &title);
1107        }
1108        strbuf_addch(sb, '\n');
1109
1110        if (need_8bit_cte > 0) {
1111                const char *header_fmt =
1112                        "MIME-Version: 1.0\n"
1113                        "Content-Type: text/plain; charset=%s\n"
1114                        "Content-Transfer-Encoding: 8bit\n";
1115                strbuf_addf(sb, header_fmt, encoding);
1116        }
1117        if (after_subject) {
1118                strbuf_addstr(sb, after_subject);
1119        }
1120        if (fmt == CMIT_FMT_EMAIL) {
1121                strbuf_addch(sb, '\n');
1122        }
1123        strbuf_release(&title);
1124}
1125
1126void pp_remainder(enum cmit_fmt fmt,
1127                  const char **msg_p,
1128                  struct strbuf *sb,
1129                  int indent)
1130{
1131        int first = 1;
1132        for (;;) {
1133                const char *line = *msg_p;
1134                int linelen = get_one_line(line);
1135                *msg_p += linelen;
1136
1137                if (!linelen)
1138                        break;
1139
1140                if (is_empty_line(line, &linelen)) {
1141                        if (first)
1142                                continue;
1143                        if (fmt == CMIT_FMT_SHORT)
1144                                break;
1145                }
1146                first = 0;
1147
1148                strbuf_grow(sb, linelen + indent + 20);
1149                if (indent) {
1150                        memset(sb->buf + sb->len, ' ', indent);
1151                        strbuf_setlen(sb, sb->len + indent);
1152                }
1153                strbuf_add(sb, line, linelen);
1154                strbuf_addch(sb, '\n');
1155        }
1156}
1157
1158char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1159{
1160        const char *encoding;
1161
1162        encoding = (git_log_output_encoding
1163                    ? git_log_output_encoding
1164                    : git_commit_encoding);
1165        if (!encoding)
1166                encoding = "UTF-8";
1167        if (encoding_p)
1168                *encoding_p = encoding;
1169        return logmsg_reencode(commit, encoding);
1170}
1171
1172void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
1173                         struct strbuf *sb,
1174                         const struct pretty_print_context *context)
1175{
1176        unsigned long beginning_of_body;
1177        int indent = 4;
1178        const char *msg = commit->buffer;
1179        char *reencoded;
1180        const char *encoding;
1181        int need_8bit_cte = context->need_8bit_cte;
1182
1183        if (fmt == CMIT_FMT_USERFORMAT) {
1184                format_commit_message(commit, user_format, sb, context);
1185                return;
1186        }
1187
1188        reencoded = reencode_commit_message(commit, &encoding);
1189        if (reencoded) {
1190                msg = reencoded;
1191        }
1192
1193        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1194                indent = 0;
1195
1196        /*
1197         * We need to check and emit Content-type: to mark it
1198         * as 8-bit if we haven't done so.
1199         */
1200        if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1201                int i, ch, in_body;
1202
1203                for (in_body = i = 0; (ch = msg[i]); i++) {
1204                        if (!in_body) {
1205                                /* author could be non 7-bit ASCII but
1206                                 * the log may be so; skip over the
1207                                 * header part first.
1208                                 */
1209                                if (ch == '\n' && msg[i+1] == '\n')
1210                                        in_body = 1;
1211                        }
1212                        else if (non_ascii(ch)) {
1213                                need_8bit_cte = 1;
1214                                break;
1215                        }
1216                }
1217        }
1218
1219        pp_header(fmt, context->abbrev, context->date_mode, encoding,
1220                  commit, &msg, sb);
1221        if (fmt != CMIT_FMT_ONELINE && !context->subject) {
1222                strbuf_addch(sb, '\n');
1223        }
1224
1225        /* Skip excess blank lines at the beginning of body, if any... */
1226        msg = skip_empty_lines(msg);
1227
1228        /* These formats treat the title line specially. */
1229        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1230                pp_title_line(fmt, &msg, sb, context->subject,
1231                              context->after_subject, encoding, need_8bit_cte);
1232
1233        beginning_of_body = sb->len;
1234        if (fmt != CMIT_FMT_ONELINE)
1235                pp_remainder(fmt, &msg, sb, indent);
1236        strbuf_rtrim(sb);
1237
1238        /* Make sure there is an EOLN for the non-oneline case */
1239        if (fmt != CMIT_FMT_ONELINE)
1240                strbuf_addch(sb, '\n');
1241
1242        /*
1243         * The caller may append additional body text in e-mail
1244         * format.  Make sure we did not strip the blank line
1245         * between the header and the body.
1246         */
1247        if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1248                strbuf_addch(sb, '\n');
1249
1250        if (context->show_notes)
1251                format_display_notes(commit->object.sha1, sb, encoding,
1252                                     NOTES_SHOW_HEADER | NOTES_INDENT);
1253
1254        free(reencoded);
1255}