51f4c546ee97847022d11997c07f5e150481e707
   1/*
   2 * GIT - The information manager from hell
   3 *
   4 * Copyright (C) Linus Torvalds, 2005
   5 */
   6#include "cache.h"
   7#include "builtin.h"
   8#include "parse-options.h"
   9#include "userdiff.h"
  10#include "streaming.h"
  11#include "tree-walk.h"
  12#include "sha1-array.h"
  13
  14struct batch_options {
  15        int enabled;
  16        int follow_symlinks;
  17        int print_contents;
  18        int buffer_output;
  19        int all_objects;
  20        const char *format;
  21};
  22
  23static const char *force_path;
  24
  25static int filter_object(const char *path, unsigned mode,
  26                         const unsigned char *sha1,
  27                         char **buf, unsigned long *size)
  28{
  29        enum object_type type;
  30
  31        *buf = read_sha1_file(sha1, &type, size);
  32        if (!*buf)
  33                return error(_("cannot read object %s '%s'"),
  34                        sha1_to_hex(sha1), path);
  35        if ((type == OBJ_BLOB) && S_ISREG(mode)) {
  36                struct strbuf strbuf = STRBUF_INIT;
  37                if (convert_to_working_tree(path, *buf, *size, &strbuf)) {
  38                        free(*buf);
  39                        *size = strbuf.len;
  40                        *buf = strbuf_detach(&strbuf, NULL);
  41                }
  42        }
  43
  44        return 0;
  45}
  46
  47static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
  48                        int unknown_type)
  49{
  50        unsigned char sha1[20];
  51        enum object_type type;
  52        char *buf;
  53        unsigned long size;
  54        struct object_context obj_context;
  55        struct object_info oi = {NULL};
  56        struct strbuf sb = STRBUF_INIT;
  57        unsigned flags = LOOKUP_REPLACE_OBJECT;
  58        const char *path = force_path;
  59
  60        if (unknown_type)
  61                flags |= LOOKUP_UNKNOWN_OBJECT;
  62
  63        if (get_sha1_with_context(obj_name, 0, sha1, &obj_context))
  64                die("Not a valid object name %s", obj_name);
  65
  66        if (!path)
  67                path = obj_context.path;
  68        if (obj_context.mode == S_IFINVALID)
  69                obj_context.mode = 0100644;
  70
  71        buf = NULL;
  72        switch (opt) {
  73        case 't':
  74                oi.typename = &sb;
  75                if (sha1_object_info_extended(sha1, &oi, flags) < 0)
  76                        die("git cat-file: could not get object info");
  77                if (sb.len) {
  78                        printf("%s\n", sb.buf);
  79                        strbuf_release(&sb);
  80                        return 0;
  81                }
  82                break;
  83
  84        case 's':
  85                oi.sizep = &size;
  86                if (sha1_object_info_extended(sha1, &oi, flags) < 0)
  87                        die("git cat-file: could not get object info");
  88                printf("%lu\n", size);
  89                return 0;
  90
  91        case 'e':
  92                return !has_sha1_file(sha1);
  93
  94        case 'w':
  95                if (!path[0])
  96                        die("git cat-file --filters %s: <object> must be "
  97                            "<sha1:path>", obj_name);
  98
  99                if (filter_object(path, obj_context.mode,
 100                                  sha1, &buf, &size))
 101                        return -1;
 102                break;
 103
 104        case 'c':
 105                if (!path[0])
 106                        die("git cat-file --textconv %s: <object> must be <sha1:path>",
 107                            obj_name);
 108
 109                if (textconv_object(path, obj_context.mode,
 110                                    sha1, 1, &buf, &size))
 111                        break;
 112
 113        case 'p':
 114                type = sha1_object_info(sha1, NULL);
 115                if (type < 0)
 116                        die("Not a valid object name %s", obj_name);
 117
 118                /* custom pretty-print here */
 119                if (type == OBJ_TREE) {
 120                        const char *ls_args[3] = { NULL };
 121                        ls_args[0] =  "ls-tree";
 122                        ls_args[1] =  obj_name;
 123                        return cmd_ls_tree(2, ls_args, NULL);
 124                }
 125
 126                if (type == OBJ_BLOB)
 127                        return stream_blob_to_fd(1, sha1, NULL, 0);
 128                buf = read_sha1_file(sha1, &type, &size);
 129                if (!buf)
 130                        die("Cannot read object %s", obj_name);
 131
 132                /* otherwise just spit out the data */
 133                break;
 134
 135        case 0:
 136                if (type_from_string(exp_type) == OBJ_BLOB) {
 137                        unsigned char blob_sha1[20];
 138                        if (sha1_object_info(sha1, NULL) == OBJ_TAG) {
 139                                char *buffer = read_sha1_file(sha1, &type, &size);
 140                                const char *target;
 141                                if (!skip_prefix(buffer, "object ", &target) ||
 142                                    get_sha1_hex(target, blob_sha1))
 143                                        die("%s not a valid tag", sha1_to_hex(sha1));
 144                                free(buffer);
 145                        } else
 146                                hashcpy(blob_sha1, sha1);
 147
 148                        if (sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
 149                                return stream_blob_to_fd(1, blob_sha1, NULL, 0);
 150                        /*
 151                         * we attempted to dereference a tag to a blob
 152                         * and failed; there may be new dereference
 153                         * mechanisms this code is not aware of.
 154                         * fall-back to the usual case.
 155                         */
 156                }
 157                buf = read_object_with_reference(sha1, exp_type, &size, NULL);
 158                break;
 159
 160        default:
 161                die("git cat-file: unknown option: %s", exp_type);
 162        }
 163
 164        if (!buf)
 165                die("git cat-file %s: bad file", obj_name);
 166
 167        write_or_die(1, buf, size);
 168        return 0;
 169}
 170
 171struct expand_data {
 172        unsigned char sha1[20];
 173        enum object_type type;
 174        unsigned long size;
 175        off_t disk_size;
 176        const char *rest;
 177        unsigned char delta_base_sha1[20];
 178
 179        /*
 180         * If mark_query is true, we do not expand anything, but rather
 181         * just mark the object_info with items we wish to query.
 182         */
 183        int mark_query;
 184
 185        /*
 186         * Whether to split the input on whitespace before feeding it to
 187         * get_sha1; this is decided during the mark_query phase based on
 188         * whether we have a %(rest) token in our format.
 189         */
 190        int split_on_whitespace;
 191
 192        /*
 193         * After a mark_query run, this object_info is set up to be
 194         * passed to sha1_object_info_extended. It will point to the data
 195         * elements above, so you can retrieve the response from there.
 196         */
 197        struct object_info info;
 198
 199        /*
 200         * This flag will be true if the requested batch format and options
 201         * don't require us to call sha1_object_info, which can then be
 202         * optimized out.
 203         */
 204        unsigned skip_object_info : 1;
 205};
 206
 207static int is_atom(const char *atom, const char *s, int slen)
 208{
 209        int alen = strlen(atom);
 210        return alen == slen && !memcmp(atom, s, alen);
 211}
 212
 213static void expand_atom(struct strbuf *sb, const char *atom, int len,
 214                        void *vdata)
 215{
 216        struct expand_data *data = vdata;
 217
 218        if (is_atom("objectname", atom, len)) {
 219                if (!data->mark_query)
 220                        strbuf_addstr(sb, sha1_to_hex(data->sha1));
 221        } else if (is_atom("objecttype", atom, len)) {
 222                if (data->mark_query)
 223                        data->info.typep = &data->type;
 224                else
 225                        strbuf_addstr(sb, typename(data->type));
 226        } else if (is_atom("objectsize", atom, len)) {
 227                if (data->mark_query)
 228                        data->info.sizep = &data->size;
 229                else
 230                        strbuf_addf(sb, "%lu", data->size);
 231        } else if (is_atom("objectsize:disk", atom, len)) {
 232                if (data->mark_query)
 233                        data->info.disk_sizep = &data->disk_size;
 234                else
 235                        strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
 236        } else if (is_atom("rest", atom, len)) {
 237                if (data->mark_query)
 238                        data->split_on_whitespace = 1;
 239                else if (data->rest)
 240                        strbuf_addstr(sb, data->rest);
 241        } else if (is_atom("deltabase", atom, len)) {
 242                if (data->mark_query)
 243                        data->info.delta_base_sha1 = data->delta_base_sha1;
 244                else
 245                        strbuf_addstr(sb, sha1_to_hex(data->delta_base_sha1));
 246        } else
 247                die("unknown format element: %.*s", len, atom);
 248}
 249
 250static size_t expand_format(struct strbuf *sb, const char *start, void *data)
 251{
 252        const char *end;
 253
 254        if (*start != '(')
 255                return 0;
 256        end = strchr(start + 1, ')');
 257        if (!end)
 258                die("format element '%s' does not end in ')'", start);
 259
 260        expand_atom(sb, start + 1, end - start - 1, data);
 261
 262        return end - start + 1;
 263}
 264
 265static void batch_write(struct batch_options *opt, const void *data, int len)
 266{
 267        if (opt->buffer_output) {
 268                if (fwrite(data, 1, len, stdout) != len)
 269                        die_errno("unable to write to stdout");
 270        } else
 271                write_or_die(1, data, len);
 272}
 273
 274static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
 275{
 276        const unsigned char *sha1 = data->sha1;
 277
 278        assert(data->info.typep);
 279
 280        if (data->type == OBJ_BLOB) {
 281                if (opt->buffer_output)
 282                        fflush(stdout);
 283                if (stream_blob_to_fd(1, sha1, NULL, 0) < 0)
 284                        die("unable to stream %s to stdout", sha1_to_hex(sha1));
 285        }
 286        else {
 287                enum object_type type;
 288                unsigned long size;
 289                void *contents;
 290
 291                contents = read_sha1_file(sha1, &type, &size);
 292                if (!contents)
 293                        die("object %s disappeared", sha1_to_hex(sha1));
 294                if (type != data->type)
 295                        die("object %s changed type!?", sha1_to_hex(sha1));
 296                if (data->info.sizep && size != data->size)
 297                        die("object %s changed size!?", sha1_to_hex(sha1));
 298
 299                batch_write(opt, contents, size);
 300                free(contents);
 301        }
 302}
 303
 304static void batch_object_write(const char *obj_name, struct batch_options *opt,
 305                               struct expand_data *data)
 306{
 307        struct strbuf buf = STRBUF_INIT;
 308
 309        if (!data->skip_object_info &&
 310            sha1_object_info_extended(data->sha1, &data->info, LOOKUP_REPLACE_OBJECT) < 0) {
 311                printf("%s missing\n", obj_name ? obj_name : sha1_to_hex(data->sha1));
 312                fflush(stdout);
 313                return;
 314        }
 315
 316        strbuf_expand(&buf, opt->format, expand_format, data);
 317        strbuf_addch(&buf, '\n');
 318        batch_write(opt, buf.buf, buf.len);
 319        strbuf_release(&buf);
 320
 321        if (opt->print_contents) {
 322                print_object_or_die(opt, data);
 323                batch_write(opt, "\n", 1);
 324        }
 325}
 326
 327static void batch_one_object(const char *obj_name, struct batch_options *opt,
 328                             struct expand_data *data)
 329{
 330        struct object_context ctx;
 331        int flags = opt->follow_symlinks ? GET_SHA1_FOLLOW_SYMLINKS : 0;
 332        enum follow_symlinks_result result;
 333
 334        result = get_sha1_with_context(obj_name, flags, data->sha1, &ctx);
 335        if (result != FOUND) {
 336                switch (result) {
 337                case MISSING_OBJECT:
 338                        printf("%s missing\n", obj_name);
 339                        break;
 340                case DANGLING_SYMLINK:
 341                        printf("dangling %"PRIuMAX"\n%s\n",
 342                               (uintmax_t)strlen(obj_name), obj_name);
 343                        break;
 344                case SYMLINK_LOOP:
 345                        printf("loop %"PRIuMAX"\n%s\n",
 346                               (uintmax_t)strlen(obj_name), obj_name);
 347                        break;
 348                case NOT_DIR:
 349                        printf("notdir %"PRIuMAX"\n%s\n",
 350                               (uintmax_t)strlen(obj_name), obj_name);
 351                        break;
 352                default:
 353                        die("BUG: unknown get_sha1_with_context result %d\n",
 354                               result);
 355                        break;
 356                }
 357                fflush(stdout);
 358                return;
 359        }
 360
 361        if (ctx.mode == 0) {
 362                printf("symlink %"PRIuMAX"\n%s\n",
 363                       (uintmax_t)ctx.symlink_path.len,
 364                       ctx.symlink_path.buf);
 365                fflush(stdout);
 366                return;
 367        }
 368
 369        batch_object_write(obj_name, opt, data);
 370}
 371
 372struct object_cb_data {
 373        struct batch_options *opt;
 374        struct expand_data *expand;
 375};
 376
 377static void batch_object_cb(const unsigned char sha1[20], void *vdata)
 378{
 379        struct object_cb_data *data = vdata;
 380        hashcpy(data->expand->sha1, sha1);
 381        batch_object_write(NULL, data->opt, data->expand);
 382}
 383
 384static int batch_loose_object(const unsigned char *sha1,
 385                              const char *path,
 386                              void *data)
 387{
 388        sha1_array_append(data, sha1);
 389        return 0;
 390}
 391
 392static int batch_packed_object(const unsigned char *sha1,
 393                               struct packed_git *pack,
 394                               uint32_t pos,
 395                               void *data)
 396{
 397        sha1_array_append(data, sha1);
 398        return 0;
 399}
 400
 401static int batch_objects(struct batch_options *opt)
 402{
 403        struct strbuf buf = STRBUF_INIT;
 404        struct expand_data data;
 405        int save_warning;
 406        int retval = 0;
 407
 408        if (!opt->format)
 409                opt->format = "%(objectname) %(objecttype) %(objectsize)";
 410
 411        /*
 412         * Expand once with our special mark_query flag, which will prime the
 413         * object_info to be handed to sha1_object_info_extended for each
 414         * object.
 415         */
 416        memset(&data, 0, sizeof(data));
 417        data.mark_query = 1;
 418        strbuf_expand(&buf, opt->format, expand_format, &data);
 419        data.mark_query = 0;
 420
 421        if (opt->all_objects) {
 422                struct object_info empty;
 423                memset(&empty, 0, sizeof(empty));
 424                if (!memcmp(&data.info, &empty, sizeof(empty)))
 425                        data.skip_object_info = 1;
 426        }
 427
 428        /*
 429         * If we are printing out the object, then always fill in the type,
 430         * since we will want to decide whether or not to stream.
 431         */
 432        if (opt->print_contents)
 433                data.info.typep = &data.type;
 434
 435        if (opt->all_objects) {
 436                struct sha1_array sa = SHA1_ARRAY_INIT;
 437                struct object_cb_data cb;
 438
 439                for_each_loose_object(batch_loose_object, &sa, 0);
 440                for_each_packed_object(batch_packed_object, &sa, 0);
 441
 442                cb.opt = opt;
 443                cb.expand = &data;
 444                sha1_array_for_each_unique(&sa, batch_object_cb, &cb);
 445
 446                sha1_array_clear(&sa);
 447                return 0;
 448        }
 449
 450        /*
 451         * We are going to call get_sha1 on a potentially very large number of
 452         * objects. In most large cases, these will be actual object sha1s. The
 453         * cost to double-check that each one is not also a ref (just so we can
 454         * warn) ends up dwarfing the actual cost of the object lookups
 455         * themselves. We can work around it by just turning off the warning.
 456         */
 457        save_warning = warn_on_object_refname_ambiguity;
 458        warn_on_object_refname_ambiguity = 0;
 459
 460        while (strbuf_getline(&buf, stdin) != EOF) {
 461                if (data.split_on_whitespace) {
 462                        /*
 463                         * Split at first whitespace, tying off the beginning
 464                         * of the string and saving the remainder (or NULL) in
 465                         * data.rest.
 466                         */
 467                        char *p = strpbrk(buf.buf, " \t");
 468                        if (p) {
 469                                while (*p && strchr(" \t", *p))
 470                                        *p++ = '\0';
 471                        }
 472                        data.rest = p;
 473                }
 474
 475                batch_one_object(buf.buf, opt, &data);
 476        }
 477
 478        strbuf_release(&buf);
 479        warn_on_object_refname_ambiguity = save_warning;
 480        return retval;
 481}
 482
 483static const char * const cat_file_usage[] = {
 484        N_("git cat-file (-t [--allow-unknown-type]|-s [--allow-unknown-type]|-e|-p|<type>|--textconv|--filters) [--path=<path>] <object>"),
 485        N_("git cat-file (--batch | --batch-check) [--follow-symlinks]"),
 486        NULL
 487};
 488
 489static int git_cat_file_config(const char *var, const char *value, void *cb)
 490{
 491        if (userdiff_config(var, value) < 0)
 492                return -1;
 493
 494        return git_default_config(var, value, cb);
 495}
 496
 497static int batch_option_callback(const struct option *opt,
 498                                 const char *arg,
 499                                 int unset)
 500{
 501        struct batch_options *bo = opt->value;
 502
 503        if (bo->enabled) {
 504                return 1;
 505        }
 506
 507        bo->enabled = 1;
 508        bo->print_contents = !strcmp(opt->long_name, "batch");
 509        bo->format = arg;
 510
 511        return 0;
 512}
 513
 514int cmd_cat_file(int argc, const char **argv, const char *prefix)
 515{
 516        int opt = 0;
 517        const char *exp_type = NULL, *obj_name = NULL;
 518        struct batch_options batch = {0};
 519        int unknown_type = 0;
 520
 521        const struct option options[] = {
 522                OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
 523                OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
 524                OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
 525                OPT_CMDMODE('e', NULL, &opt,
 526                            N_("exit with zero when there's no error"), 'e'),
 527                OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
 528                OPT_CMDMODE(0, "textconv", &opt,
 529                            N_("for blob objects, run textconv on object's content"), 'c'),
 530                OPT_CMDMODE(0, "filters", &opt,
 531                            N_("for blob objects, run filters on object's content"), 'w'),
 532                OPT_STRING(0, "path", &force_path, N_("blob"),
 533                           N_("use a specific path for --textconv/--filters")),
 534                OPT_BOOL(0, "allow-unknown-type", &unknown_type,
 535                          N_("allow -s and -t to work with broken/corrupt objects")),
 536                OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
 537                { OPTION_CALLBACK, 0, "batch", &batch, "format",
 538                        N_("show info and content of objects fed from the standard input"),
 539                        PARSE_OPT_OPTARG, batch_option_callback },
 540                { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
 541                        N_("show info about objects fed from the standard input"),
 542                        PARSE_OPT_OPTARG, batch_option_callback },
 543                OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
 544                         N_("follow in-tree symlinks (used with --batch or --batch-check)")),
 545                OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
 546                         N_("show all objects with --batch or --batch-check")),
 547                OPT_END()
 548        };
 549
 550        git_config(git_cat_file_config, NULL);
 551
 552        batch.buffer_output = -1;
 553        argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
 554
 555        if (opt) {
 556                if (argc == 1)
 557                        obj_name = argv[0];
 558                else
 559                        usage_with_options(cat_file_usage, options);
 560        }
 561        if (!opt && !batch.enabled) {
 562                if (argc == 2) {
 563                        exp_type = argv[0];
 564                        obj_name = argv[1];
 565                } else
 566                        usage_with_options(cat_file_usage, options);
 567        }
 568        if (batch.enabled && (opt || argc)) {
 569                usage_with_options(cat_file_usage, options);
 570        }
 571
 572        if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
 573                usage_with_options(cat_file_usage, options);
 574        }
 575
 576        if (force_path && opt != 'c' && opt != 'w') {
 577                error("--path=<path> needs --textconv or --filters");
 578                usage_with_options(cat_file_usage, options);
 579        }
 580
 581        if (batch.buffer_output < 0)
 582                batch.buffer_output = batch.all_objects;
 583
 584        if (batch.enabled)
 585                return batch_objects(&batch);
 586
 587        if (unknown_type && opt != 't' && opt != 's')
 588                die("git cat-file --allow-unknown-type: use with -s or -t");
 589        return cat_one_file(opt, exp_type, obj_name, unknown_type);
 590}