Merge branch 'nd/no-the-index'
[git/git.git] / builtin / cat-file.c
1 /*
2 * GIT - The information manager from hell
3 *
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"
17
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;
27 };
28
29 static const char *force_path;
30
31 static int filter_object(const char *path, unsigned mode,
32 const struct object_id *oid,
33 char **buf, unsigned long *size)
34 {
35 enum object_type type;
36
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(&the_index, path, *buf, *size, &strbuf)) {
44 free(*buf);
45 *size = strbuf.len;
46 *buf = strbuf_detach(&strbuf, NULL);
47 }
48 }
49
50 return 0;
51 }
52
53 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
54 int unknown_type)
55 {
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;
65
66 if (unknown_type)
67 flags |= OBJECT_INFO_ALLOW_UNKNOWN_TYPE;
68
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);
72
73 if (!path)
74 path = obj_context.path;
75 if (obj_context.mode == S_IFINVALID)
76 obj_context.mode = 0100644;
77
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;
88 }
89 break;
90
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;
97
98 case 'e':
99 return !has_object_file(&oid);
100
101 case 'w':
102 if (!path)
103 die("git cat-file --filters %s: <object> must be "
104 "<sha1:path>", obj_name);
105
106 if (filter_object(path, obj_context.mode,
107 &oid, &buf, &size))
108 return -1;
109 break;
110
111 case 'c':
112 if (!path)
113 die("git cat-file --textconv %s: <object> must be <sha1:path>",
114 obj_name);
115
116 if (textconv_object(path, obj_context.mode, &oid, 1, &buf, &size))
117 break;
118 /* else fallthrough */
119
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);
124
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);
131 }
132
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);
138
139 /* otherwise just spit out the data */
140 break;
141
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);
155
156 if (oid_object_info(the_repository, &blob_oid, NULL) == OBJ_BLOB)
157 return stream_blob_to_fd(1, &blob_oid, NULL, 0);
158 /*
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.
163 */
164 }
165 buf = read_object_with_reference(&oid, exp_type, &size, NULL);
166 break;
167
168 default:
169 die("git cat-file: unknown option: %s", exp_type);
170 }
171
172 if (!buf)
173 die("git cat-file %s: bad file", obj_name);
174
175 write_or_die(1, buf, size);
176 free(buf);
177 free(obj_context.path);
178 return 0;
179 }
180
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;
188
189 /*
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.
192 */
193 int mark_query;
194
195 /*
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.
199 */
200 int split_on_whitespace;
201
202 /*
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.
206 */
207 struct object_info info;
208
209 /*
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.
213 */
214 unsigned skip_object_info : 1;
215 };
216
217 static int is_atom(const char *atom, const char *s, int slen)
218 {
219 int alen = strlen(atom);
220 return alen == slen && !memcmp(atom, s, alen);
221 }
222
223 static void expand_atom(struct strbuf *sb, const char *atom, int len,
224 void *vdata)
225 {
226 struct expand_data *data = vdata;
227
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);
259 }
260
261 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
262 {
263 const char *end;
264
265 if (*start != '(')
266 return 0;
267 end = strchr(start + 1, ')');
268 if (!end)
269 die("format element '%s' does not end in ')'", start);
270
271 expand_atom(sb, start + 1, end - start - 1, data);
272
273 return end - start + 1;
274 }
275
276 static void batch_write(struct batch_options *opt, const void *data, int len)
277 {
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);
283 }
284
285 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
286 {
287 const struct object_id *oid = &data->oid;
288
289 assert(data->info.typep);
290
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;
297
298 if (!data->rest)
299 die("missing path for '%s'", oid_to_hex(oid));
300
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));
322 }
323 else {
324 enum object_type type;
325 unsigned long size;
326 void *contents;
327
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));
335
336 batch_write(opt, contents, size);
337 free(contents);
338 }
339 }
340
341 static void batch_object_write(const char *obj_name,
342 struct strbuf *scratch,
343 struct batch_options *opt,
344 struct expand_data *data)
345 {
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;
353 }
354
355 strbuf_reset(scratch);
356 strbuf_expand(scratch, opt->format, expand_format, data);
357 strbuf_addch(scratch, '\n');
358 batch_write(opt, scratch->buf, scratch->len);
359
360 if (opt->print_contents) {
361 print_object_or_die(opt, data);
362 batch_write(opt, "\n", 1);
363 }
364 }
365
366 static void batch_one_object(const char *obj_name,
367 struct strbuf *scratch,
368 struct batch_options *opt,
369 struct expand_data *data)
370 {
371 struct object_context ctx;
372 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
373 enum follow_symlinks_result result;
374
375 result = get_oid_with_context(obj_name, flags, &data->oid, &ctx);
376 if (result != FOUND) {
377 switch (result) {
378 case MISSING_OBJECT:
379 printf("%s missing\n", obj_name);
380 break;
381 case DANGLING_SYMLINK:
382 printf("dangling %"PRIuMAX"\n%s\n",
383 (uintmax_t)strlen(obj_name), obj_name);
384 break;
385 case SYMLINK_LOOP:
386 printf("loop %"PRIuMAX"\n%s\n",
387 (uintmax_t)strlen(obj_name), obj_name);
388 break;
389 case NOT_DIR:
390 printf("notdir %"PRIuMAX"\n%s\n",
391 (uintmax_t)strlen(obj_name), obj_name);
392 break;
393 default:
394 BUG("unknown get_sha1_with_context result %d\n",
395 result);
396 break;
397 }
398 fflush(stdout);
399 return;
400 }
401
402 if (ctx.mode == 0) {
403 printf("symlink %"PRIuMAX"\n%s\n",
404 (uintmax_t)ctx.symlink_path.len,
405 ctx.symlink_path.buf);
406 fflush(stdout);
407 return;
408 }
409
410 batch_object_write(obj_name, scratch, opt, data);
411 }
412
413 struct object_cb_data {
414 struct batch_options *opt;
415 struct expand_data *expand;
416 struct oidset *seen;
417 struct strbuf *scratch;
418 };
419
420 static int batch_object_cb(const struct object_id *oid, void *vdata)
421 {
422 struct object_cb_data *data = vdata;
423 oidcpy(&data->expand->oid, oid);
424 batch_object_write(NULL, data->scratch, data->opt, data->expand);
425 return 0;
426 }
427
428 static int collect_loose_object(const struct object_id *oid,
429 const char *path,
430 void *data)
431 {
432 oid_array_append(data, oid);
433 return 0;
434 }
435
436 static int collect_packed_object(const struct object_id *oid,
437 struct packed_git *pack,
438 uint32_t pos,
439 void *data)
440 {
441 oid_array_append(data, oid);
442 return 0;
443 }
444
445 static int batch_unordered_object(const struct object_id *oid, void *vdata)
446 {
447 struct object_cb_data *data = vdata;
448
449 if (oidset_insert(data->seen, oid))
450 return 0;
451
452 return batch_object_cb(oid, data);
453 }
454
455 static int batch_unordered_loose(const struct object_id *oid,
456 const char *path,
457 void *data)
458 {
459 return batch_unordered_object(oid, data);
460 }
461
462 static int batch_unordered_packed(const struct object_id *oid,
463 struct packed_git *pack,
464 uint32_t pos,
465 void *data)
466 {
467 return batch_unordered_object(oid, data);
468 }
469
470 static int batch_objects(struct batch_options *opt)
471 {
472 struct strbuf input = STRBUF_INIT;
473 struct strbuf output = STRBUF_INIT;
474 struct expand_data data;
475 int save_warning;
476 int retval = 0;
477
478 if (!opt->format)
479 opt->format = "%(objectname) %(objecttype) %(objectsize)";
480
481 /*
482 * Expand once with our special mark_query flag, which will prime the
483 * object_info to be handed to sha1_object_info_extended for each
484 * object.
485 */
486 memset(&data, 0, sizeof(data));
487 data.mark_query = 1;
488 strbuf_expand(&output, opt->format, expand_format, &data);
489 data.mark_query = 0;
490 strbuf_release(&output);
491 if (opt->cmdmode)
492 data.split_on_whitespace = 1;
493
494 if (opt->all_objects) {
495 struct object_info empty = OBJECT_INFO_INIT;
496 if (!memcmp(&data.info, &empty, sizeof(empty)))
497 data.skip_object_info = 1;
498 }
499
500 /*
501 * If we are printing out the object, then always fill in the type,
502 * since we will want to decide whether or not to stream.
503 */
504 if (opt->print_contents)
505 data.info.typep = &data.type;
506
507 if (opt->all_objects) {
508 struct object_cb_data cb;
509
510 if (repository_format_partial_clone)
511 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
512
513 cb.opt = opt;
514 cb.expand = &data;
515 cb.scratch = &output;
516
517 if (opt->unordered) {
518 struct oidset seen = OIDSET_INIT;
519
520 cb.seen = &seen;
521
522 for_each_loose_object(batch_unordered_loose, &cb, 0);
523 for_each_packed_object(batch_unordered_packed, &cb,
524 FOR_EACH_OBJECT_PACK_ORDER);
525
526 oidset_clear(&seen);
527 } else {
528 struct oid_array sa = OID_ARRAY_INIT;
529
530 for_each_loose_object(collect_loose_object, &sa, 0);
531 for_each_packed_object(collect_packed_object, &sa, 0);
532
533 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
534
535 oid_array_clear(&sa);
536 }
537
538 strbuf_release(&output);
539 return 0;
540 }
541
542 /*
543 * We are going to call get_sha1 on a potentially very large number of
544 * objects. In most large cases, these will be actual object sha1s. The
545 * cost to double-check that each one is not also a ref (just so we can
546 * warn) ends up dwarfing the actual cost of the object lookups
547 * themselves. We can work around it by just turning off the warning.
548 */
549 save_warning = warn_on_object_refname_ambiguity;
550 warn_on_object_refname_ambiguity = 0;
551
552 while (strbuf_getline(&input, stdin) != EOF) {
553 if (data.split_on_whitespace) {
554 /*
555 * Split at first whitespace, tying off the beginning
556 * of the string and saving the remainder (or NULL) in
557 * data.rest.
558 */
559 char *p = strpbrk(input.buf, " \t");
560 if (p) {
561 while (*p && strchr(" \t", *p))
562 *p++ = '\0';
563 }
564 data.rest = p;
565 }
566
567 batch_one_object(input.buf, &output, opt, &data);
568 }
569
570 strbuf_release(&input);
571 strbuf_release(&output);
572 warn_on_object_refname_ambiguity = save_warning;
573 return retval;
574 }
575
576 static const char * const cat_file_usage[] = {
577 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
578 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
579 NULL
580 };
581
582 static int git_cat_file_config(const char *var, const char *value, void *cb)
583 {
584 if (userdiff_config(var, value) < 0)
585 return -1;
586
587 return git_default_config(var, value, cb);
588 }
589
590 static int batch_option_callback(const struct option *opt,
591 const char *arg,
592 int unset)
593 {
594 struct batch_options *bo = opt->value;
595
596 if (bo->enabled) {
597 return 1;
598 }
599
600 bo->enabled = 1;
601 bo->print_contents = !strcmp(opt->long_name, "batch");
602 bo->format = arg;
603
604 return 0;
605 }
606
607 int cmd_cat_file(int argc, const char **argv, const char *prefix)
608 {
609 int opt = 0;
610 const char *exp_type = NULL, *obj_name = NULL;
611 struct batch_options batch = {0};
612 int unknown_type = 0;
613
614 const struct option options[] = {
615 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
616 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
617 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
618 OPT_CMDMODE('e', NULL, &opt,
619 N_("exit with zero when there's no error"), 'e'),
620 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
621 OPT_CMDMODE(0, "textconv", &opt,
622 N_("for blob objects, run textconv on object's content"), 'c'),
623 OPT_CMDMODE(0, "filters", &opt,
624 N_("for blob objects, run filters on object's content"), 'w'),
625 OPT_STRING(0, "path", &force_path, N_("blob"),
626 N_("use a specific path for --textconv/--filters")),
627 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
628 N_("allow -s and -t to work with broken/corrupt objects")),
629 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
630 { OPTION_CALLBACK, 0, "batch", &batch, "format",
631 N_("show info and content of objects fed from the standard input"),
632 PARSE_OPT_OPTARG, batch_option_callback },
633 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
634 N_("show info about objects fed from the standard input"),
635 PARSE_OPT_OPTARG, batch_option_callback },
636 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
637 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
638 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
639 N_("show all objects with --batch or --batch-check")),
640 OPT_BOOL(0, "unordered", &batch.unordered,
641 N_("do not order --batch-all-objects output")),
642 OPT_END()
643 };
644
645 git_config(git_cat_file_config, NULL);
646
647 batch.buffer_output = -1;
648 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
649
650 if (opt) {
651 if (batch.enabled && (opt == 'c' || opt == 'w'))
652 batch.cmdmode = opt;
653 else if (argc == 1)
654 obj_name = argv[0];
655 else
656 usage_with_options(cat_file_usage, options);
657 }
658 if (!opt && !batch.enabled) {
659 if (argc == 2) {
660 exp_type = argv[0];
661 obj_name = argv[1];
662 } else
663 usage_with_options(cat_file_usage, options);
664 }
665 if (batch.enabled) {
666 if (batch.cmdmode != opt || argc)
667 usage_with_options(cat_file_usage, options);
668 if (batch.cmdmode && batch.all_objects)
669 die("--batch-all-objects cannot be combined with "
670 "--textconv nor with --filters");
671 }
672
673 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
674 usage_with_options(cat_file_usage, options);
675 }
676
677 if (force_path && opt != 'c' && opt != 'w') {
678 error("--path=<path> needs --textconv or --filters");
679 usage_with_options(cat_file_usage, options);
680 }
681
682 if (force_path && batch.enabled) {
683 error("--path=<path> incompatible with --batch");
684 usage_with_options(cat_file_usage, options);
685 }
686
687 if (batch.buffer_output < 0)
688 batch.buffer_output = batch.all_objects;
689
690 if (batch.enabled)
691 return batch_objects(&batch);
692
693 if (unknown_type && opt != 't' && opt != 's')
694 die("git cat-file --allow-unknown-type: use with -s or -t");
695 return cat_one_file(opt, exp_type, obj_name, unknown_type);
696 }