Merge branch 'ab/checkout-branch-info-leakfix'
[git/debian.git] / builtin / cat-file.c
blobd94050e6c188ff4594a065da87695c67c560ac94
1 /*
2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
5 */
6 #define USE_THE_INDEX_COMPATIBILITY_MACROS
7 #include "cache.h"
8 #include "config.h"
9 #include "builtin.h"
10 #include "diff.h"
11 #include "parse-options.h"
12 #include "userdiff.h"
13 #include "streaming.h"
14 #include "tree-walk.h"
15 #include "oid-array.h"
16 #include "packfile.h"
17 #include "object-store.h"
18 #include "promisor-remote.h"
20 struct batch_options {
21 int enabled;
22 int follow_symlinks;
23 int print_contents;
24 int buffer_output;
25 int all_objects;
26 int unordered;
27 int cmdmode; /* may be 'w' or 'c' for --filters or --textconv */
28 const char *format;
31 static const char *force_path;
33 static int filter_object(const char *path, unsigned mode,
34 const struct object_id *oid,
35 char **buf, unsigned long *size)
37 enum object_type type;
39 *buf = read_object_file(oid, &type, size);
40 if (!*buf)
41 return error(_("cannot read object %s '%s'"),
42 oid_to_hex(oid), path);
43 if ((type == OBJ_BLOB) && S_ISREG(mode)) {
44 struct strbuf strbuf = STRBUF_INIT;
45 struct checkout_metadata meta;
47 init_checkout_metadata(&meta, NULL, NULL, oid);
48 if (convert_to_working_tree(&the_index, path, *buf, *size, &strbuf, &meta)) {
49 free(*buf);
50 *size = strbuf.len;
51 *buf = strbuf_detach(&strbuf, NULL);
55 return 0;
58 static int stream_blob(const struct object_id *oid)
60 if (stream_blob_to_fd(1, oid, NULL, 0))
61 die("unable to stream %s to stdout", oid_to_hex(oid));
62 return 0;
65 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
66 int unknown_type)
68 struct object_id oid;
69 enum object_type type;
70 char *buf;
71 unsigned long size;
72 struct object_context obj_context;
73 struct object_info oi = OBJECT_INFO_INIT;
74 struct strbuf sb = STRBUF_INIT;
75 unsigned flags = OBJECT_INFO_LOOKUP_REPLACE;
76 const char *path = force_path;
78 if (unknown_type)
79 flags |= OBJECT_INFO_ALLOW_UNKNOWN_TYPE;
81 if (get_oid_with_context(the_repository, obj_name,
82 GET_OID_RECORD_PATH,
83 &oid, &obj_context))
84 die("Not a valid object name %s", obj_name);
86 if (!path)
87 path = obj_context.path;
88 if (obj_context.mode == S_IFINVALID)
89 obj_context.mode = 0100644;
91 buf = NULL;
92 switch (opt) {
93 case 't':
94 oi.type_name = &sb;
95 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
96 die("git cat-file: could not get object info");
97 if (sb.len) {
98 printf("%s\n", sb.buf);
99 strbuf_release(&sb);
100 return 0;
102 break;
104 case 's':
105 oi.sizep = &size;
106 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
107 die("git cat-file: could not get object info");
108 printf("%"PRIuMAX"\n", (uintmax_t)size);
109 return 0;
111 case 'e':
112 return !has_object_file(&oid);
114 case 'w':
115 if (!path)
116 die("git cat-file --filters %s: <object> must be "
117 "<sha1:path>", obj_name);
119 if (filter_object(path, obj_context.mode,
120 &oid, &buf, &size))
121 return -1;
122 break;
124 case 'c':
125 if (!path)
126 die("git cat-file --textconv %s: <object> must be <sha1:path>",
127 obj_name);
129 if (textconv_object(the_repository, path, obj_context.mode,
130 &oid, 1, &buf, &size))
131 break;
132 /* else fallthrough */
134 case 'p':
135 type = oid_object_info(the_repository, &oid, NULL);
136 if (type < 0)
137 die("Not a valid object name %s", obj_name);
139 /* custom pretty-print here */
140 if (type == OBJ_TREE) {
141 const char *ls_args[3] = { NULL };
142 ls_args[0] = "ls-tree";
143 ls_args[1] = obj_name;
144 return cmd_ls_tree(2, ls_args, NULL);
147 if (type == OBJ_BLOB)
148 return stream_blob(&oid);
149 buf = read_object_file(&oid, &type, &size);
150 if (!buf)
151 die("Cannot read object %s", obj_name);
153 /* otherwise just spit out the data */
154 break;
156 case 0:
157 if (type_from_string(exp_type) == OBJ_BLOB) {
158 struct object_id blob_oid;
159 if (oid_object_info(the_repository, &oid, NULL) == OBJ_TAG) {
160 char *buffer = read_object_file(&oid, &type,
161 &size);
162 const char *target;
163 if (!skip_prefix(buffer, "object ", &target) ||
164 get_oid_hex(target, &blob_oid))
165 die("%s not a valid tag", oid_to_hex(&oid));
166 free(buffer);
167 } else
168 oidcpy(&blob_oid, &oid);
170 if (oid_object_info(the_repository, &blob_oid, NULL) == OBJ_BLOB)
171 return stream_blob(&blob_oid);
173 * we attempted to dereference a tag to a blob
174 * and failed; there may be new dereference
175 * mechanisms this code is not aware of.
176 * fall-back to the usual case.
179 buf = read_object_with_reference(the_repository,
180 &oid, exp_type, &size, NULL);
181 break;
183 default:
184 die("git cat-file: unknown option: %s", exp_type);
187 if (!buf)
188 die("git cat-file %s: bad file", obj_name);
190 write_or_die(1, buf, size);
191 free(buf);
192 free(obj_context.path);
193 return 0;
196 struct expand_data {
197 struct object_id oid;
198 enum object_type type;
199 unsigned long size;
200 off_t disk_size;
201 const char *rest;
202 struct object_id delta_base_oid;
205 * If mark_query is true, we do not expand anything, but rather
206 * just mark the object_info with items we wish to query.
208 int mark_query;
211 * Whether to split the input on whitespace before feeding it to
212 * get_sha1; this is decided during the mark_query phase based on
213 * whether we have a %(rest) token in our format.
215 int split_on_whitespace;
218 * After a mark_query run, this object_info is set up to be
219 * passed to oid_object_info_extended. It will point to the data
220 * elements above, so you can retrieve the response from there.
222 struct object_info info;
225 * This flag will be true if the requested batch format and options
226 * don't require us to call oid_object_info, which can then be
227 * optimized out.
229 unsigned skip_object_info : 1;
232 static int is_atom(const char *atom, const char *s, int slen)
234 int alen = strlen(atom);
235 return alen == slen && !memcmp(atom, s, alen);
238 static void expand_atom(struct strbuf *sb, const char *atom, int len,
239 void *vdata)
241 struct expand_data *data = vdata;
243 if (is_atom("objectname", atom, len)) {
244 if (!data->mark_query)
245 strbuf_addstr(sb, oid_to_hex(&data->oid));
246 } else if (is_atom("objecttype", atom, len)) {
247 if (data->mark_query)
248 data->info.typep = &data->type;
249 else
250 strbuf_addstr(sb, type_name(data->type));
251 } else if (is_atom("objectsize", atom, len)) {
252 if (data->mark_query)
253 data->info.sizep = &data->size;
254 else
255 strbuf_addf(sb, "%"PRIuMAX , (uintmax_t)data->size);
256 } else if (is_atom("objectsize:disk", atom, len)) {
257 if (data->mark_query)
258 data->info.disk_sizep = &data->disk_size;
259 else
260 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
261 } else if (is_atom("rest", atom, len)) {
262 if (data->mark_query)
263 data->split_on_whitespace = 1;
264 else if (data->rest)
265 strbuf_addstr(sb, data->rest);
266 } else if (is_atom("deltabase", atom, len)) {
267 if (data->mark_query)
268 data->info.delta_base_oid = &data->delta_base_oid;
269 else
270 strbuf_addstr(sb,
271 oid_to_hex(&data->delta_base_oid));
272 } else
273 die("unknown format element: %.*s", len, atom);
276 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
278 const char *end;
280 if (*start != '(')
281 return 0;
282 end = strchr(start + 1, ')');
283 if (!end)
284 die("format element '%s' does not end in ')'", start);
286 expand_atom(sb, start + 1, end - start - 1, data);
288 return end - start + 1;
291 static void batch_write(struct batch_options *opt, const void *data, int len)
293 if (opt->buffer_output) {
294 if (fwrite(data, 1, len, stdout) != len)
295 die_errno("unable to write to stdout");
296 } else
297 write_or_die(1, data, len);
300 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
302 const struct object_id *oid = &data->oid;
304 assert(data->info.typep);
306 if (data->type == OBJ_BLOB) {
307 if (opt->buffer_output)
308 fflush(stdout);
309 if (opt->cmdmode) {
310 char *contents;
311 unsigned long size;
313 if (!data->rest)
314 die("missing path for '%s'", oid_to_hex(oid));
316 if (opt->cmdmode == 'w') {
317 if (filter_object(data->rest, 0100644, oid,
318 &contents, &size))
319 die("could not convert '%s' %s",
320 oid_to_hex(oid), data->rest);
321 } else if (opt->cmdmode == 'c') {
322 enum object_type type;
323 if (!textconv_object(the_repository,
324 data->rest, 0100644, oid,
325 1, &contents, &size))
326 contents = read_object_file(oid,
327 &type,
328 &size);
329 if (!contents)
330 die("could not convert '%s' %s",
331 oid_to_hex(oid), data->rest);
332 } else
333 BUG("invalid cmdmode: %c", opt->cmdmode);
334 batch_write(opt, contents, size);
335 free(contents);
336 } else {
337 stream_blob(oid);
340 else {
341 enum object_type type;
342 unsigned long size;
343 void *contents;
345 contents = read_object_file(oid, &type, &size);
346 if (!contents)
347 die("object %s disappeared", oid_to_hex(oid));
348 if (type != data->type)
349 die("object %s changed type!?", oid_to_hex(oid));
350 if (data->info.sizep && size != data->size)
351 die("object %s changed size!?", oid_to_hex(oid));
353 batch_write(opt, contents, size);
354 free(contents);
359 * If "pack" is non-NULL, then "offset" is the byte offset within the pack from
360 * which the object may be accessed (though note that we may also rely on
361 * data->oid, too). If "pack" is NULL, then offset is ignored.
363 static void batch_object_write(const char *obj_name,
364 struct strbuf *scratch,
365 struct batch_options *opt,
366 struct expand_data *data,
367 struct packed_git *pack,
368 off_t offset)
370 if (!data->skip_object_info) {
371 int ret;
373 if (pack)
374 ret = packed_object_info(the_repository, pack, offset,
375 &data->info);
376 else
377 ret = oid_object_info_extended(the_repository,
378 &data->oid, &data->info,
379 OBJECT_INFO_LOOKUP_REPLACE);
380 if (ret < 0) {
381 printf("%s missing\n",
382 obj_name ? obj_name : oid_to_hex(&data->oid));
383 fflush(stdout);
384 return;
388 strbuf_reset(scratch);
389 strbuf_expand(scratch, opt->format, expand_format, data);
390 strbuf_addch(scratch, '\n');
391 batch_write(opt, scratch->buf, scratch->len);
393 if (opt->print_contents) {
394 print_object_or_die(opt, data);
395 batch_write(opt, "\n", 1);
399 static void batch_one_object(const char *obj_name,
400 struct strbuf *scratch,
401 struct batch_options *opt,
402 struct expand_data *data)
404 struct object_context ctx;
405 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
406 enum get_oid_result result;
408 result = get_oid_with_context(the_repository, obj_name,
409 flags, &data->oid, &ctx);
410 if (result != FOUND) {
411 switch (result) {
412 case MISSING_OBJECT:
413 printf("%s missing\n", obj_name);
414 break;
415 case SHORT_NAME_AMBIGUOUS:
416 printf("%s ambiguous\n", obj_name);
417 break;
418 case DANGLING_SYMLINK:
419 printf("dangling %"PRIuMAX"\n%s\n",
420 (uintmax_t)strlen(obj_name), obj_name);
421 break;
422 case SYMLINK_LOOP:
423 printf("loop %"PRIuMAX"\n%s\n",
424 (uintmax_t)strlen(obj_name), obj_name);
425 break;
426 case NOT_DIR:
427 printf("notdir %"PRIuMAX"\n%s\n",
428 (uintmax_t)strlen(obj_name), obj_name);
429 break;
430 default:
431 BUG("unknown get_sha1_with_context result %d\n",
432 result);
433 break;
435 fflush(stdout);
436 return;
439 if (ctx.mode == 0) {
440 printf("symlink %"PRIuMAX"\n%s\n",
441 (uintmax_t)ctx.symlink_path.len,
442 ctx.symlink_path.buf);
443 fflush(stdout);
444 return;
447 batch_object_write(obj_name, scratch, opt, data, NULL, 0);
450 struct object_cb_data {
451 struct batch_options *opt;
452 struct expand_data *expand;
453 struct oidset *seen;
454 struct strbuf *scratch;
457 static int batch_object_cb(const struct object_id *oid, void *vdata)
459 struct object_cb_data *data = vdata;
460 oidcpy(&data->expand->oid, oid);
461 batch_object_write(NULL, data->scratch, data->opt, data->expand,
462 NULL, 0);
463 return 0;
466 static int collect_loose_object(const struct object_id *oid,
467 const char *path,
468 void *data)
470 oid_array_append(data, oid);
471 return 0;
474 static int collect_packed_object(const struct object_id *oid,
475 struct packed_git *pack,
476 uint32_t pos,
477 void *data)
479 oid_array_append(data, oid);
480 return 0;
483 static int batch_unordered_object(const struct object_id *oid,
484 struct packed_git *pack, off_t offset,
485 void *vdata)
487 struct object_cb_data *data = vdata;
489 if (oidset_insert(data->seen, oid))
490 return 0;
492 oidcpy(&data->expand->oid, oid);
493 batch_object_write(NULL, data->scratch, data->opt, data->expand,
494 pack, offset);
495 return 0;
498 static int batch_unordered_loose(const struct object_id *oid,
499 const char *path,
500 void *data)
502 return batch_unordered_object(oid, NULL, 0, data);
505 static int batch_unordered_packed(const struct object_id *oid,
506 struct packed_git *pack,
507 uint32_t pos,
508 void *data)
510 return batch_unordered_object(oid, pack,
511 nth_packed_object_offset(pack, pos),
512 data);
515 static int batch_objects(struct batch_options *opt)
517 struct strbuf input = STRBUF_INIT;
518 struct strbuf output = STRBUF_INIT;
519 struct expand_data data;
520 int save_warning;
521 int retval = 0;
523 if (!opt->format)
524 opt->format = "%(objectname) %(objecttype) %(objectsize)";
527 * Expand once with our special mark_query flag, which will prime the
528 * object_info to be handed to oid_object_info_extended for each
529 * object.
531 memset(&data, 0, sizeof(data));
532 data.mark_query = 1;
533 strbuf_expand(&output, opt->format, expand_format, &data);
534 data.mark_query = 0;
535 strbuf_release(&output);
536 if (opt->cmdmode)
537 data.split_on_whitespace = 1;
540 * If we are printing out the object, then always fill in the type,
541 * since we will want to decide whether or not to stream.
543 if (opt->print_contents)
544 data.info.typep = &data.type;
546 if (opt->all_objects) {
547 struct object_cb_data cb;
548 struct object_info empty = OBJECT_INFO_INIT;
550 if (!memcmp(&data.info, &empty, sizeof(empty)))
551 data.skip_object_info = 1;
553 if (has_promisor_remote())
554 warning("This repository uses promisor remotes. Some objects may not be loaded.");
556 read_replace_refs = 0;
558 cb.opt = opt;
559 cb.expand = &data;
560 cb.scratch = &output;
562 if (opt->unordered) {
563 struct oidset seen = OIDSET_INIT;
565 cb.seen = &seen;
567 for_each_loose_object(batch_unordered_loose, &cb, 0);
568 for_each_packed_object(batch_unordered_packed, &cb,
569 FOR_EACH_OBJECT_PACK_ORDER);
571 oidset_clear(&seen);
572 } else {
573 struct oid_array sa = OID_ARRAY_INIT;
575 for_each_loose_object(collect_loose_object, &sa, 0);
576 for_each_packed_object(collect_packed_object, &sa, 0);
578 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
580 oid_array_clear(&sa);
583 strbuf_release(&output);
584 return 0;
588 * We are going to call get_sha1 on a potentially very large number of
589 * objects. In most large cases, these will be actual object sha1s. The
590 * cost to double-check that each one is not also a ref (just so we can
591 * warn) ends up dwarfing the actual cost of the object lookups
592 * themselves. We can work around it by just turning off the warning.
594 save_warning = warn_on_object_refname_ambiguity;
595 warn_on_object_refname_ambiguity = 0;
597 while (strbuf_getline(&input, stdin) != EOF) {
598 if (data.split_on_whitespace) {
600 * Split at first whitespace, tying off the beginning
601 * of the string and saving the remainder (or NULL) in
602 * data.rest.
604 char *p = strpbrk(input.buf, " \t");
605 if (p) {
606 while (*p && strchr(" \t", *p))
607 *p++ = '\0';
609 data.rest = p;
612 batch_one_object(input.buf, &output, opt, &data);
615 strbuf_release(&input);
616 strbuf_release(&output);
617 warn_on_object_refname_ambiguity = save_warning;
618 return retval;
621 static const char * const cat_file_usage[] = {
622 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
623 N_("git cat-file (--batch[=<format>] | --batch-check[=<format>]) [--follow-symlinks] [--textconv | --filters]"),
624 NULL
627 static int git_cat_file_config(const char *var, const char *value, void *cb)
629 if (userdiff_config(var, value) < 0)
630 return -1;
632 return git_default_config(var, value, cb);
635 static int batch_option_callback(const struct option *opt,
636 const char *arg,
637 int unset)
639 struct batch_options *bo = opt->value;
641 BUG_ON_OPT_NEG(unset);
643 if (bo->enabled) {
644 return error(_("only one batch option may be specified"));
647 bo->enabled = 1;
648 bo->print_contents = !strcmp(opt->long_name, "batch");
649 bo->format = arg;
651 return 0;
654 int cmd_cat_file(int argc, const char **argv, const char *prefix)
656 int opt = 0;
657 const char *exp_type = NULL, *obj_name = NULL;
658 struct batch_options batch = {0};
659 int unknown_type = 0;
661 const struct option options[] = {
662 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
663 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
664 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
665 OPT_CMDMODE('e', NULL, &opt,
666 N_("exit with zero when there's no error"), 'e'),
667 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
668 OPT_CMDMODE(0, "textconv", &opt,
669 N_("for blob objects, run textconv on object's content"), 'c'),
670 OPT_CMDMODE(0, "filters", &opt,
671 N_("for blob objects, run filters on object's content"), 'w'),
672 OPT_STRING(0, "path", &force_path, N_("blob"),
673 N_("use a specific path for --textconv/--filters")),
674 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
675 N_("allow -s and -t to work with broken/corrupt objects")),
676 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
677 OPT_CALLBACK_F(0, "batch", &batch, "format",
678 N_("show info and content of objects fed from the standard input"),
679 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
680 batch_option_callback),
681 OPT_CALLBACK_F(0, "batch-check", &batch, "format",
682 N_("show info about objects fed from the standard input"),
683 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
684 batch_option_callback),
685 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
686 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
687 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
688 N_("show all objects with --batch or --batch-check")),
689 OPT_BOOL(0, "unordered", &batch.unordered,
690 N_("do not order --batch-all-objects output")),
691 OPT_END()
694 git_config(git_cat_file_config, NULL);
696 batch.buffer_output = -1;
697 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
699 if (opt) {
700 if (batch.enabled && (opt == 'c' || opt == 'w'))
701 batch.cmdmode = opt;
702 else if (argc == 1)
703 obj_name = argv[0];
704 else
705 usage_with_options(cat_file_usage, options);
707 if (!opt && !batch.enabled) {
708 if (argc == 2) {
709 exp_type = argv[0];
710 obj_name = argv[1];
711 } else
712 usage_with_options(cat_file_usage, options);
714 if (batch.enabled) {
715 if (batch.cmdmode != opt || argc)
716 usage_with_options(cat_file_usage, options);
717 if (batch.cmdmode && batch.all_objects)
718 die("--batch-all-objects cannot be combined with "
719 "--textconv nor with --filters");
722 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
723 usage_with_options(cat_file_usage, options);
726 if (force_path && opt != 'c' && opt != 'w') {
727 error("--path=<path> needs --textconv or --filters");
728 usage_with_options(cat_file_usage, options);
731 if (force_path && batch.enabled) {
732 error("options '--path=<path>' and '--batch' cannot be used together");
733 usage_with_options(cat_file_usage, options);
736 if (batch.buffer_output < 0)
737 batch.buffer_output = batch.all_objects;
739 if (batch.enabled)
740 return batch_objects(&batch);
742 if (unknown_type && opt != 't' && opt != 's')
743 die("git cat-file --allow-unknown-type: use with -s or -t");
744 return cat_one_file(opt, exp_type, obj_name, unknown_type);