cat-file: use oidset check-and-insert
[git/debian.git] / builtin / cat-file.c
blob04b5cda191804ce9727a1a01901c88222fb69535
1 /*
2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
5 */
6 #include "cache.h"
7 #include "config.h"
8 #include "builtin.h"
9 #include "diff.h"
10 #include "parse-options.h"
11 #include "userdiff.h"
12 #include "streaming.h"
13 #include "tree-walk.h"
14 #include "sha1-array.h"
15 #include "packfile.h"
16 #include "object-store.h"
18 struct batch_options {
19 int enabled;
20 int follow_symlinks;
21 int print_contents;
22 int buffer_output;
23 int all_objects;
24 int unordered;
25 int cmdmode; /* may be 'w' or 'c' for --filters or --textconv */
26 const char *format;
29 static const char *force_path;
31 static int filter_object(const char *path, unsigned mode,
32 const struct object_id *oid,
33 char **buf, unsigned long *size)
35 enum object_type type;
37 *buf = read_object_file(oid, &type, size);
38 if (!*buf)
39 return error(_("cannot read object %s '%s'"),
40 oid_to_hex(oid), path);
41 if ((type == OBJ_BLOB) && S_ISREG(mode)) {
42 struct strbuf strbuf = STRBUF_INIT;
43 if (convert_to_working_tree(path, *buf, *size, &strbuf)) {
44 free(*buf);
45 *size = strbuf.len;
46 *buf = strbuf_detach(&strbuf, NULL);
50 return 0;
53 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
54 int unknown_type)
56 struct object_id oid;
57 enum object_type type;
58 char *buf;
59 unsigned long size;
60 struct object_context obj_context;
61 struct object_info oi = OBJECT_INFO_INIT;
62 struct strbuf sb = STRBUF_INIT;
63 unsigned flags = OBJECT_INFO_LOOKUP_REPLACE;
64 const char *path = force_path;
66 if (unknown_type)
67 flags |= OBJECT_INFO_ALLOW_UNKNOWN_TYPE;
69 if (get_oid_with_context(obj_name, GET_OID_RECORD_PATH,
70 &oid, &obj_context))
71 die("Not a valid object name %s", obj_name);
73 if (!path)
74 path = obj_context.path;
75 if (obj_context.mode == S_IFINVALID)
76 obj_context.mode = 0100644;
78 buf = NULL;
79 switch (opt) {
80 case 't':
81 oi.type_name = &sb;
82 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
83 die("git cat-file: could not get object info");
84 if (sb.len) {
85 printf("%s\n", sb.buf);
86 strbuf_release(&sb);
87 return 0;
89 break;
91 case 's':
92 oi.sizep = &size;
93 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
94 die("git cat-file: could not get object info");
95 printf("%lu\n", size);
96 return 0;
98 case 'e':
99 return !has_object_file(&oid);
101 case 'w':
102 if (!path)
103 die("git cat-file --filters %s: <object> must be "
104 "<sha1:path>", obj_name);
106 if (filter_object(path, obj_context.mode,
107 &oid, &buf, &size))
108 return -1;
109 break;
111 case 'c':
112 if (!path)
113 die("git cat-file --textconv %s: <object> must be <sha1:path>",
114 obj_name);
116 if (textconv_object(path, obj_context.mode, &oid, 1, &buf, &size))
117 break;
118 /* else fallthrough */
120 case 'p':
121 type = oid_object_info(the_repository, &oid, NULL);
122 if (type < 0)
123 die("Not a valid object name %s", obj_name);
125 /* custom pretty-print here */
126 if (type == OBJ_TREE) {
127 const char *ls_args[3] = { NULL };
128 ls_args[0] = "ls-tree";
129 ls_args[1] = obj_name;
130 return cmd_ls_tree(2, ls_args, NULL);
133 if (type == OBJ_BLOB)
134 return stream_blob_to_fd(1, &oid, NULL, 0);
135 buf = read_object_file(&oid, &type, &size);
136 if (!buf)
137 die("Cannot read object %s", obj_name);
139 /* otherwise just spit out the data */
140 break;
142 case 0:
143 if (type_from_string(exp_type) == OBJ_BLOB) {
144 struct object_id blob_oid;
145 if (oid_object_info(the_repository, &oid, NULL) == OBJ_TAG) {
146 char *buffer = read_object_file(&oid, &type,
147 &size);
148 const char *target;
149 if (!skip_prefix(buffer, "object ", &target) ||
150 get_oid_hex(target, &blob_oid))
151 die("%s not a valid tag", oid_to_hex(&oid));
152 free(buffer);
153 } else
154 oidcpy(&blob_oid, &oid);
156 if (oid_object_info(the_repository, &blob_oid, NULL) == OBJ_BLOB)
157 return stream_blob_to_fd(1, &blob_oid, NULL, 0);
159 * we attempted to dereference a tag to a blob
160 * and failed; there may be new dereference
161 * mechanisms this code is not aware of.
162 * fall-back to the usual case.
165 buf = read_object_with_reference(&oid, exp_type, &size, NULL);
166 break;
168 default:
169 die("git cat-file: unknown option: %s", exp_type);
172 if (!buf)
173 die("git cat-file %s: bad file", obj_name);
175 write_or_die(1, buf, size);
176 free(buf);
177 free(obj_context.path);
178 return 0;
181 struct expand_data {
182 struct object_id oid;
183 enum object_type type;
184 unsigned long size;
185 off_t disk_size;
186 const char *rest;
187 struct object_id delta_base_oid;
190 * If mark_query is true, we do not expand anything, but rather
191 * just mark the object_info with items we wish to query.
193 int mark_query;
196 * Whether to split the input on whitespace before feeding it to
197 * get_sha1; this is decided during the mark_query phase based on
198 * whether we have a %(rest) token in our format.
200 int split_on_whitespace;
203 * After a mark_query run, this object_info is set up to be
204 * passed to sha1_object_info_extended. It will point to the data
205 * elements above, so you can retrieve the response from there.
207 struct object_info info;
210 * This flag will be true if the requested batch format and options
211 * don't require us to call sha1_object_info, which can then be
212 * optimized out.
214 unsigned skip_object_info : 1;
217 static int is_atom(const char *atom, const char *s, int slen)
219 int alen = strlen(atom);
220 return alen == slen && !memcmp(atom, s, alen);
223 static void expand_atom(struct strbuf *sb, const char *atom, int len,
224 void *vdata)
226 struct expand_data *data = vdata;
228 if (is_atom("objectname", atom, len)) {
229 if (!data->mark_query)
230 strbuf_addstr(sb, oid_to_hex(&data->oid));
231 } else if (is_atom("objecttype", atom, len)) {
232 if (data->mark_query)
233 data->info.typep = &data->type;
234 else
235 strbuf_addstr(sb, type_name(data->type));
236 } else if (is_atom("objectsize", atom, len)) {
237 if (data->mark_query)
238 data->info.sizep = &data->size;
239 else
240 strbuf_addf(sb, "%lu", data->size);
241 } else if (is_atom("objectsize:disk", atom, len)) {
242 if (data->mark_query)
243 data->info.disk_sizep = &data->disk_size;
244 else
245 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
246 } else if (is_atom("rest", atom, len)) {
247 if (data->mark_query)
248 data->split_on_whitespace = 1;
249 else if (data->rest)
250 strbuf_addstr(sb, data->rest);
251 } else if (is_atom("deltabase", atom, len)) {
252 if (data->mark_query)
253 data->info.delta_base_sha1 = data->delta_base_oid.hash;
254 else
255 strbuf_addstr(sb,
256 oid_to_hex(&data->delta_base_oid));
257 } else
258 die("unknown format element: %.*s", len, atom);
261 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
263 const char *end;
265 if (*start != '(')
266 return 0;
267 end = strchr(start + 1, ')');
268 if (!end)
269 die("format element '%s' does not end in ')'", start);
271 expand_atom(sb, start + 1, end - start - 1, data);
273 return end - start + 1;
276 static void batch_write(struct batch_options *opt, const void *data, int len)
278 if (opt->buffer_output) {
279 if (fwrite(data, 1, len, stdout) != len)
280 die_errno("unable to write to stdout");
281 } else
282 write_or_die(1, data, len);
285 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
287 const struct object_id *oid = &data->oid;
289 assert(data->info.typep);
291 if (data->type == OBJ_BLOB) {
292 if (opt->buffer_output)
293 fflush(stdout);
294 if (opt->cmdmode) {
295 char *contents;
296 unsigned long size;
298 if (!data->rest)
299 die("missing path for '%s'", oid_to_hex(oid));
301 if (opt->cmdmode == 'w') {
302 if (filter_object(data->rest, 0100644, oid,
303 &contents, &size))
304 die("could not convert '%s' %s",
305 oid_to_hex(oid), data->rest);
306 } else if (opt->cmdmode == 'c') {
307 enum object_type type;
308 if (!textconv_object(data->rest, 0100644, oid,
309 1, &contents, &size))
310 contents = read_object_file(oid,
311 &type,
312 &size);
313 if (!contents)
314 die("could not convert '%s' %s",
315 oid_to_hex(oid), data->rest);
316 } else
317 BUG("invalid cmdmode: %c", opt->cmdmode);
318 batch_write(opt, contents, size);
319 free(contents);
320 } else if (stream_blob_to_fd(1, oid, NULL, 0) < 0)
321 die("unable to stream %s to stdout", oid_to_hex(oid));
323 else {
324 enum object_type type;
325 unsigned long size;
326 void *contents;
328 contents = read_object_file(oid, &type, &size);
329 if (!contents)
330 die("object %s disappeared", oid_to_hex(oid));
331 if (type != data->type)
332 die("object %s changed type!?", oid_to_hex(oid));
333 if (data->info.sizep && size != data->size)
334 die("object %s changed size!?", oid_to_hex(oid));
336 batch_write(opt, contents, size);
337 free(contents);
341 static void batch_object_write(const char *obj_name, struct batch_options *opt,
342 struct expand_data *data)
344 struct strbuf buf = STRBUF_INIT;
346 if (!data->skip_object_info &&
347 oid_object_info_extended(the_repository, &data->oid, &data->info,
348 OBJECT_INFO_LOOKUP_REPLACE) < 0) {
349 printf("%s missing\n",
350 obj_name ? obj_name : oid_to_hex(&data->oid));
351 fflush(stdout);
352 return;
355 strbuf_expand(&buf, opt->format, expand_format, data);
356 strbuf_addch(&buf, '\n');
357 batch_write(opt, buf.buf, buf.len);
358 strbuf_release(&buf);
360 if (opt->print_contents) {
361 print_object_or_die(opt, data);
362 batch_write(opt, "\n", 1);
366 static void batch_one_object(const char *obj_name, struct batch_options *opt,
367 struct expand_data *data)
369 struct object_context ctx;
370 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
371 enum follow_symlinks_result result;
373 result = get_oid_with_context(obj_name, flags, &data->oid, &ctx);
374 if (result != FOUND) {
375 switch (result) {
376 case MISSING_OBJECT:
377 printf("%s missing\n", obj_name);
378 break;
379 case DANGLING_SYMLINK:
380 printf("dangling %"PRIuMAX"\n%s\n",
381 (uintmax_t)strlen(obj_name), obj_name);
382 break;
383 case SYMLINK_LOOP:
384 printf("loop %"PRIuMAX"\n%s\n",
385 (uintmax_t)strlen(obj_name), obj_name);
386 break;
387 case NOT_DIR:
388 printf("notdir %"PRIuMAX"\n%s\n",
389 (uintmax_t)strlen(obj_name), obj_name);
390 break;
391 default:
392 BUG("unknown get_sha1_with_context result %d\n",
393 result);
394 break;
396 fflush(stdout);
397 return;
400 if (ctx.mode == 0) {
401 printf("symlink %"PRIuMAX"\n%s\n",
402 (uintmax_t)ctx.symlink_path.len,
403 ctx.symlink_path.buf);
404 fflush(stdout);
405 return;
408 batch_object_write(obj_name, opt, data);
411 struct object_cb_data {
412 struct batch_options *opt;
413 struct expand_data *expand;
414 struct oidset *seen;
417 static int batch_object_cb(const struct object_id *oid, void *vdata)
419 struct object_cb_data *data = vdata;
420 oidcpy(&data->expand->oid, oid);
421 batch_object_write(NULL, data->opt, data->expand);
422 return 0;
425 static int collect_loose_object(const struct object_id *oid,
426 const char *path,
427 void *data)
429 oid_array_append(data, oid);
430 return 0;
433 static int collect_packed_object(const struct object_id *oid,
434 struct packed_git *pack,
435 uint32_t pos,
436 void *data)
438 oid_array_append(data, oid);
439 return 0;
442 static int batch_unordered_object(const struct object_id *oid, void *vdata)
444 struct object_cb_data *data = vdata;
446 if (oidset_insert(data->seen, oid))
447 return 0;
449 return batch_object_cb(oid, data);
452 static int batch_unordered_loose(const struct object_id *oid,
453 const char *path,
454 void *data)
456 return batch_unordered_object(oid, data);
459 static int batch_unordered_packed(const struct object_id *oid,
460 struct packed_git *pack,
461 uint32_t pos,
462 void *data)
464 return batch_unordered_object(oid, data);
467 static int batch_objects(struct batch_options *opt)
469 struct strbuf buf = STRBUF_INIT;
470 struct expand_data data;
471 int save_warning;
472 int retval = 0;
474 if (!opt->format)
475 opt->format = "%(objectname) %(objecttype) %(objectsize)";
478 * Expand once with our special mark_query flag, which will prime the
479 * object_info to be handed to sha1_object_info_extended for each
480 * object.
482 memset(&data, 0, sizeof(data));
483 data.mark_query = 1;
484 strbuf_expand(&buf, opt->format, expand_format, &data);
485 data.mark_query = 0;
486 if (opt->cmdmode)
487 data.split_on_whitespace = 1;
489 if (opt->all_objects) {
490 struct object_info empty = OBJECT_INFO_INIT;
491 if (!memcmp(&data.info, &empty, sizeof(empty)))
492 data.skip_object_info = 1;
496 * If we are printing out the object, then always fill in the type,
497 * since we will want to decide whether or not to stream.
499 if (opt->print_contents)
500 data.info.typep = &data.type;
502 if (opt->all_objects) {
503 struct object_cb_data cb;
505 if (repository_format_partial_clone)
506 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
508 cb.opt = opt;
509 cb.expand = &data;
511 if (opt->unordered) {
512 struct oidset seen = OIDSET_INIT;
514 cb.seen = &seen;
516 for_each_loose_object(batch_unordered_loose, &cb, 0);
517 for_each_packed_object(batch_unordered_packed, &cb,
518 FOR_EACH_OBJECT_PACK_ORDER);
520 oidset_clear(&seen);
521 } else {
522 struct oid_array sa = OID_ARRAY_INIT;
524 for_each_loose_object(collect_loose_object, &sa, 0);
525 for_each_packed_object(collect_packed_object, &sa, 0);
527 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
529 oid_array_clear(&sa);
532 return 0;
536 * We are going to call get_sha1 on a potentially very large number of
537 * objects. In most large cases, these will be actual object sha1s. The
538 * cost to double-check that each one is not also a ref (just so we can
539 * warn) ends up dwarfing the actual cost of the object lookups
540 * themselves. We can work around it by just turning off the warning.
542 save_warning = warn_on_object_refname_ambiguity;
543 warn_on_object_refname_ambiguity = 0;
545 while (strbuf_getline(&buf, stdin) != EOF) {
546 if (data.split_on_whitespace) {
548 * Split at first whitespace, tying off the beginning
549 * of the string and saving the remainder (or NULL) in
550 * data.rest.
552 char *p = strpbrk(buf.buf, " \t");
553 if (p) {
554 while (*p && strchr(" \t", *p))
555 *p++ = '\0';
557 data.rest = p;
560 batch_one_object(buf.buf, opt, &data);
563 strbuf_release(&buf);
564 warn_on_object_refname_ambiguity = save_warning;
565 return retval;
568 static const char * const cat_file_usage[] = {
569 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
570 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
571 NULL
574 static int git_cat_file_config(const char *var, const char *value, void *cb)
576 if (userdiff_config(var, value) < 0)
577 return -1;
579 return git_default_config(var, value, cb);
582 static int batch_option_callback(const struct option *opt,
583 const char *arg,
584 int unset)
586 struct batch_options *bo = opt->value;
588 if (bo->enabled) {
589 return 1;
592 bo->enabled = 1;
593 bo->print_contents = !strcmp(opt->long_name, "batch");
594 bo->format = arg;
596 return 0;
599 int cmd_cat_file(int argc, const char **argv, const char *prefix)
601 int opt = 0;
602 const char *exp_type = NULL, *obj_name = NULL;
603 struct batch_options batch = {0};
604 int unknown_type = 0;
606 const struct option options[] = {
607 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
608 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
609 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
610 OPT_CMDMODE('e', NULL, &opt,
611 N_("exit with zero when there's no error"), 'e'),
612 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
613 OPT_CMDMODE(0, "textconv", &opt,
614 N_("for blob objects, run textconv on object's content"), 'c'),
615 OPT_CMDMODE(0, "filters", &opt,
616 N_("for blob objects, run filters on object's content"), 'w'),
617 OPT_STRING(0, "path", &force_path, N_("blob"),
618 N_("use a specific path for --textconv/--filters")),
619 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
620 N_("allow -s and -t to work with broken/corrupt objects")),
621 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
622 { OPTION_CALLBACK, 0, "batch", &batch, "format",
623 N_("show info and content of objects fed from the standard input"),
624 PARSE_OPT_OPTARG, batch_option_callback },
625 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
626 N_("show info about objects fed from the standard input"),
627 PARSE_OPT_OPTARG, batch_option_callback },
628 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
629 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
630 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
631 N_("show all objects with --batch or --batch-check")),
632 OPT_BOOL(0, "unordered", &batch.unordered,
633 N_("do not order --batch-all-objects output")),
634 OPT_END()
637 git_config(git_cat_file_config, NULL);
639 batch.buffer_output = -1;
640 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
642 if (opt) {
643 if (batch.enabled && (opt == 'c' || opt == 'w'))
644 batch.cmdmode = opt;
645 else if (argc == 1)
646 obj_name = argv[0];
647 else
648 usage_with_options(cat_file_usage, options);
650 if (!opt && !batch.enabled) {
651 if (argc == 2) {
652 exp_type = argv[0];
653 obj_name = argv[1];
654 } else
655 usage_with_options(cat_file_usage, options);
657 if (batch.enabled) {
658 if (batch.cmdmode != opt || argc)
659 usage_with_options(cat_file_usage, options);
660 if (batch.cmdmode && batch.all_objects)
661 die("--batch-all-objects cannot be combined with "
662 "--textconv nor with --filters");
665 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
666 usage_with_options(cat_file_usage, options);
669 if (force_path && opt != 'c' && opt != 'w') {
670 error("--path=<path> needs --textconv or --filters");
671 usage_with_options(cat_file_usage, options);
674 if (force_path && batch.enabled) {
675 error("--path=<path> incompatible with --batch");
676 usage_with_options(cat_file_usage, options);
679 if (batch.buffer_output < 0)
680 batch.buffer_output = batch.all_objects;
682 if (batch.enabled)
683 return batch_objects(&batch);
685 if (unknown_type && opt != 't' && opt != 's')
686 die("git cat-file --allow-unknown-type: use with -s or -t");
687 return cat_one_file(opt, exp_type, obj_name, unknown_type);