sha1-name.c: remove implicit dependency on 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 stream_blob(const struct object_id *oid)
54 {
55 if (stream_blob_to_fd(1, oid, NULL, 0))
56 die("unable to stream %s to stdout", oid_to_hex(oid));
57 return 0;
58 }
59
60 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
61 int unknown_type)
62 {
63 struct object_id oid;
64 enum object_type type;
65 char *buf;
66 unsigned long size;
67 struct object_context obj_context;
68 struct object_info oi = OBJECT_INFO_INIT;
69 struct strbuf sb = STRBUF_INIT;
70 unsigned flags = OBJECT_INFO_LOOKUP_REPLACE;
71 const char *path = force_path;
72
73 if (unknown_type)
74 flags |= OBJECT_INFO_ALLOW_UNKNOWN_TYPE;
75
76 if (get_oid_with_context(the_repository, obj_name,
77 GET_OID_RECORD_PATH,
78 &oid, &obj_context))
79 die("Not a valid object name %s", obj_name);
80
81 if (!path)
82 path = obj_context.path;
83 if (obj_context.mode == S_IFINVALID)
84 obj_context.mode = 0100644;
85
86 buf = NULL;
87 switch (opt) {
88 case 't':
89 oi.type_name = &sb;
90 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
91 die("git cat-file: could not get object info");
92 if (sb.len) {
93 printf("%s\n", sb.buf);
94 strbuf_release(&sb);
95 return 0;
96 }
97 break;
98
99 case 's':
100 oi.sizep = &size;
101 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
102 die("git cat-file: could not get object info");
103 printf("%"PRIuMAX"\n", (uintmax_t)size);
104 return 0;
105
106 case 'e':
107 return !has_object_file(&oid);
108
109 case 'w':
110 if (!path)
111 die("git cat-file --filters %s: <object> must be "
112 "<sha1:path>", obj_name);
113
114 if (filter_object(path, obj_context.mode,
115 &oid, &buf, &size))
116 return -1;
117 break;
118
119 case 'c':
120 if (!path)
121 die("git cat-file --textconv %s: <object> must be <sha1:path>",
122 obj_name);
123
124 if (textconv_object(the_repository, path, obj_context.mode,
125 &oid, 1, &buf, &size))
126 break;
127 /* else fallthrough */
128
129 case 'p':
130 type = oid_object_info(the_repository, &oid, NULL);
131 if (type < 0)
132 die("Not a valid object name %s", obj_name);
133
134 /* custom pretty-print here */
135 if (type == OBJ_TREE) {
136 const char *ls_args[3] = { NULL };
137 ls_args[0] = "ls-tree";
138 ls_args[1] = obj_name;
139 return cmd_ls_tree(2, ls_args, NULL);
140 }
141
142 if (type == OBJ_BLOB)
143 return stream_blob(&oid);
144 buf = read_object_file(&oid, &type, &size);
145 if (!buf)
146 die("Cannot read object %s", obj_name);
147
148 /* otherwise just spit out the data */
149 break;
150
151 case 0:
152 if (type_from_string(exp_type) == OBJ_BLOB) {
153 struct object_id blob_oid;
154 if (oid_object_info(the_repository, &oid, NULL) == OBJ_TAG) {
155 char *buffer = read_object_file(&oid, &type,
156 &size);
157 const char *target;
158 if (!skip_prefix(buffer, "object ", &target) ||
159 get_oid_hex(target, &blob_oid))
160 die("%s not a valid tag", oid_to_hex(&oid));
161 free(buffer);
162 } else
163 oidcpy(&blob_oid, &oid);
164
165 if (oid_object_info(the_repository, &blob_oid, NULL) == OBJ_BLOB)
166 return stream_blob(&blob_oid);
167 /*
168 * we attempted to dereference a tag to a blob
169 * and failed; there may be new dereference
170 * mechanisms this code is not aware of.
171 * fall-back to the usual case.
172 */
173 }
174 buf = read_object_with_reference(&oid, exp_type, &size, NULL);
175 break;
176
177 default:
178 die("git cat-file: unknown option: %s", exp_type);
179 }
180
181 if (!buf)
182 die("git cat-file %s: bad file", obj_name);
183
184 write_or_die(1, buf, size);
185 free(buf);
186 free(obj_context.path);
187 return 0;
188 }
189
190 struct expand_data {
191 struct object_id oid;
192 enum object_type type;
193 unsigned long size;
194 off_t disk_size;
195 const char *rest;
196 struct object_id delta_base_oid;
197
198 /*
199 * If mark_query is true, we do not expand anything, but rather
200 * just mark the object_info with items we wish to query.
201 */
202 int mark_query;
203
204 /*
205 * Whether to split the input on whitespace before feeding it to
206 * get_sha1; this is decided during the mark_query phase based on
207 * whether we have a %(rest) token in our format.
208 */
209 int split_on_whitespace;
210
211 /*
212 * After a mark_query run, this object_info is set up to be
213 * passed to sha1_object_info_extended. It will point to the data
214 * elements above, so you can retrieve the response from there.
215 */
216 struct object_info info;
217
218 /*
219 * This flag will be true if the requested batch format and options
220 * don't require us to call sha1_object_info, which can then be
221 * optimized out.
222 */
223 unsigned skip_object_info : 1;
224 };
225
226 static int is_atom(const char *atom, const char *s, int slen)
227 {
228 int alen = strlen(atom);
229 return alen == slen && !memcmp(atom, s, alen);
230 }
231
232 static void expand_atom(struct strbuf *sb, const char *atom, int len,
233 void *vdata)
234 {
235 struct expand_data *data = vdata;
236
237 if (is_atom("objectname", atom, len)) {
238 if (!data->mark_query)
239 strbuf_addstr(sb, oid_to_hex(&data->oid));
240 } else if (is_atom("objecttype", atom, len)) {
241 if (data->mark_query)
242 data->info.typep = &data->type;
243 else
244 strbuf_addstr(sb, type_name(data->type));
245 } else if (is_atom("objectsize", atom, len)) {
246 if (data->mark_query)
247 data->info.sizep = &data->size;
248 else
249 strbuf_addf(sb, "%"PRIuMAX , (uintmax_t)data->size);
250 } else if (is_atom("objectsize:disk", atom, len)) {
251 if (data->mark_query)
252 data->info.disk_sizep = &data->disk_size;
253 else
254 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
255 } else if (is_atom("rest", atom, len)) {
256 if (data->mark_query)
257 data->split_on_whitespace = 1;
258 else if (data->rest)
259 strbuf_addstr(sb, data->rest);
260 } else if (is_atom("deltabase", atom, len)) {
261 if (data->mark_query)
262 data->info.delta_base_sha1 = data->delta_base_oid.hash;
263 else
264 strbuf_addstr(sb,
265 oid_to_hex(&data->delta_base_oid));
266 } else
267 die("unknown format element: %.*s", len, atom);
268 }
269
270 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
271 {
272 const char *end;
273
274 if (*start != '(')
275 return 0;
276 end = strchr(start + 1, ')');
277 if (!end)
278 die("format element '%s' does not end in ')'", start);
279
280 expand_atom(sb, start + 1, end - start - 1, data);
281
282 return end - start + 1;
283 }
284
285 static void batch_write(struct batch_options *opt, const void *data, int len)
286 {
287 if (opt->buffer_output) {
288 if (fwrite(data, 1, len, stdout) != len)
289 die_errno("unable to write to stdout");
290 } else
291 write_or_die(1, data, len);
292 }
293
294 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
295 {
296 const struct object_id *oid = &data->oid;
297
298 assert(data->info.typep);
299
300 if (data->type == OBJ_BLOB) {
301 if (opt->buffer_output)
302 fflush(stdout);
303 if (opt->cmdmode) {
304 char *contents;
305 unsigned long size;
306
307 if (!data->rest)
308 die("missing path for '%s'", oid_to_hex(oid));
309
310 if (opt->cmdmode == 'w') {
311 if (filter_object(data->rest, 0100644, oid,
312 &contents, &size))
313 die("could not convert '%s' %s",
314 oid_to_hex(oid), data->rest);
315 } else if (opt->cmdmode == 'c') {
316 enum object_type type;
317 if (!textconv_object(the_repository,
318 data->rest, 0100644, oid,
319 1, &contents, &size))
320 contents = read_object_file(oid,
321 &type,
322 &size);
323 if (!contents)
324 die("could not convert '%s' %s",
325 oid_to_hex(oid), data->rest);
326 } else
327 BUG("invalid cmdmode: %c", opt->cmdmode);
328 batch_write(opt, contents, size);
329 free(contents);
330 } else {
331 stream_blob(oid);
332 }
333 }
334 else {
335 enum object_type type;
336 unsigned long size;
337 void *contents;
338
339 contents = read_object_file(oid, &type, &size);
340 if (!contents)
341 die("object %s disappeared", oid_to_hex(oid));
342 if (type != data->type)
343 die("object %s changed type!?", oid_to_hex(oid));
344 if (data->info.sizep && size != data->size)
345 die("object %s changed size!?", oid_to_hex(oid));
346
347 batch_write(opt, contents, size);
348 free(contents);
349 }
350 }
351
352 static void batch_object_write(const char *obj_name,
353 struct strbuf *scratch,
354 struct batch_options *opt,
355 struct expand_data *data)
356 {
357 if (!data->skip_object_info &&
358 oid_object_info_extended(the_repository, &data->oid, &data->info,
359 OBJECT_INFO_LOOKUP_REPLACE) < 0) {
360 printf("%s missing\n",
361 obj_name ? obj_name : oid_to_hex(&data->oid));
362 fflush(stdout);
363 return;
364 }
365
366 strbuf_reset(scratch);
367 strbuf_expand(scratch, opt->format, expand_format, data);
368 strbuf_addch(scratch, '\n');
369 batch_write(opt, scratch->buf, scratch->len);
370
371 if (opt->print_contents) {
372 print_object_or_die(opt, data);
373 batch_write(opt, "\n", 1);
374 }
375 }
376
377 static void batch_one_object(const char *obj_name,
378 struct strbuf *scratch,
379 struct batch_options *opt,
380 struct expand_data *data)
381 {
382 struct object_context ctx;
383 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
384 enum follow_symlinks_result result;
385
386 result = get_oid_with_context(the_repository, obj_name,
387 flags, &data->oid, &ctx);
388 if (result != FOUND) {
389 switch (result) {
390 case MISSING_OBJECT:
391 printf("%s missing\n", obj_name);
392 break;
393 case DANGLING_SYMLINK:
394 printf("dangling %"PRIuMAX"\n%s\n",
395 (uintmax_t)strlen(obj_name), obj_name);
396 break;
397 case SYMLINK_LOOP:
398 printf("loop %"PRIuMAX"\n%s\n",
399 (uintmax_t)strlen(obj_name), obj_name);
400 break;
401 case NOT_DIR:
402 printf("notdir %"PRIuMAX"\n%s\n",
403 (uintmax_t)strlen(obj_name), obj_name);
404 break;
405 default:
406 BUG("unknown get_sha1_with_context result %d\n",
407 result);
408 break;
409 }
410 fflush(stdout);
411 return;
412 }
413
414 if (ctx.mode == 0) {
415 printf("symlink %"PRIuMAX"\n%s\n",
416 (uintmax_t)ctx.symlink_path.len,
417 ctx.symlink_path.buf);
418 fflush(stdout);
419 return;
420 }
421
422 batch_object_write(obj_name, scratch, opt, data);
423 }
424
425 struct object_cb_data {
426 struct batch_options *opt;
427 struct expand_data *expand;
428 struct oidset *seen;
429 struct strbuf *scratch;
430 };
431
432 static int batch_object_cb(const struct object_id *oid, void *vdata)
433 {
434 struct object_cb_data *data = vdata;
435 oidcpy(&data->expand->oid, oid);
436 batch_object_write(NULL, data->scratch, data->opt, data->expand);
437 return 0;
438 }
439
440 static int collect_loose_object(const struct object_id *oid,
441 const char *path,
442 void *data)
443 {
444 oid_array_append(data, oid);
445 return 0;
446 }
447
448 static int collect_packed_object(const struct object_id *oid,
449 struct packed_git *pack,
450 uint32_t pos,
451 void *data)
452 {
453 oid_array_append(data, oid);
454 return 0;
455 }
456
457 static int batch_unordered_object(const struct object_id *oid, void *vdata)
458 {
459 struct object_cb_data *data = vdata;
460
461 if (oidset_insert(data->seen, oid))
462 return 0;
463
464 return batch_object_cb(oid, data);
465 }
466
467 static int batch_unordered_loose(const struct object_id *oid,
468 const char *path,
469 void *data)
470 {
471 return batch_unordered_object(oid, data);
472 }
473
474 static int batch_unordered_packed(const struct object_id *oid,
475 struct packed_git *pack,
476 uint32_t pos,
477 void *data)
478 {
479 return batch_unordered_object(oid, data);
480 }
481
482 static int batch_objects(struct batch_options *opt)
483 {
484 struct strbuf input = STRBUF_INIT;
485 struct strbuf output = STRBUF_INIT;
486 struct expand_data data;
487 int save_warning;
488 int retval = 0;
489
490 if (!opt->format)
491 opt->format = "%(objectname) %(objecttype) %(objectsize)";
492
493 /*
494 * Expand once with our special mark_query flag, which will prime the
495 * object_info to be handed to sha1_object_info_extended for each
496 * object.
497 */
498 memset(&data, 0, sizeof(data));
499 data.mark_query = 1;
500 strbuf_expand(&output, opt->format, expand_format, &data);
501 data.mark_query = 0;
502 strbuf_release(&output);
503 if (opt->cmdmode)
504 data.split_on_whitespace = 1;
505
506 if (opt->all_objects) {
507 struct object_info empty = OBJECT_INFO_INIT;
508 if (!memcmp(&data.info, &empty, sizeof(empty)))
509 data.skip_object_info = 1;
510 }
511
512 /*
513 * If we are printing out the object, then always fill in the type,
514 * since we will want to decide whether or not to stream.
515 */
516 if (opt->print_contents)
517 data.info.typep = &data.type;
518
519 if (opt->all_objects) {
520 struct object_cb_data cb;
521
522 if (repository_format_partial_clone)
523 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
524
525 cb.opt = opt;
526 cb.expand = &data;
527 cb.scratch = &output;
528
529 if (opt->unordered) {
530 struct oidset seen = OIDSET_INIT;
531
532 cb.seen = &seen;
533
534 for_each_loose_object(batch_unordered_loose, &cb, 0);
535 for_each_packed_object(batch_unordered_packed, &cb,
536 FOR_EACH_OBJECT_PACK_ORDER);
537
538 oidset_clear(&seen);
539 } else {
540 struct oid_array sa = OID_ARRAY_INIT;
541
542 for_each_loose_object(collect_loose_object, &sa, 0);
543 for_each_packed_object(collect_packed_object, &sa, 0);
544
545 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
546
547 oid_array_clear(&sa);
548 }
549
550 strbuf_release(&output);
551 return 0;
552 }
553
554 /*
555 * We are going to call get_sha1 on a potentially very large number of
556 * objects. In most large cases, these will be actual object sha1s. The
557 * cost to double-check that each one is not also a ref (just so we can
558 * warn) ends up dwarfing the actual cost of the object lookups
559 * themselves. We can work around it by just turning off the warning.
560 */
561 save_warning = warn_on_object_refname_ambiguity;
562 warn_on_object_refname_ambiguity = 0;
563
564 while (strbuf_getline(&input, stdin) != EOF) {
565 if (data.split_on_whitespace) {
566 /*
567 * Split at first whitespace, tying off the beginning
568 * of the string and saving the remainder (or NULL) in
569 * data.rest.
570 */
571 char *p = strpbrk(input.buf, " \t");
572 if (p) {
573 while (*p && strchr(" \t", *p))
574 *p++ = '\0';
575 }
576 data.rest = p;
577 }
578
579 batch_one_object(input.buf, &output, opt, &data);
580 }
581
582 strbuf_release(&input);
583 strbuf_release(&output);
584 warn_on_object_refname_ambiguity = save_warning;
585 return retval;
586 }
587
588 static const char * const cat_file_usage[] = {
589 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
590 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
591 NULL
592 };
593
594 static int git_cat_file_config(const char *var, const char *value, void *cb)
595 {
596 if (userdiff_config(var, value) < 0)
597 return -1;
598
599 return git_default_config(var, value, cb);
600 }
601
602 static int batch_option_callback(const struct option *opt,
603 const char *arg,
604 int unset)
605 {
606 struct batch_options *bo = opt->value;
607
608 BUG_ON_OPT_NEG(unset);
609
610 if (bo->enabled) {
611 return error(_("only one batch option may be specified"));
612 }
613
614 bo->enabled = 1;
615 bo->print_contents = !strcmp(opt->long_name, "batch");
616 bo->format = arg;
617
618 return 0;
619 }
620
621 int cmd_cat_file(int argc, const char **argv, const char *prefix)
622 {
623 int opt = 0;
624 const char *exp_type = NULL, *obj_name = NULL;
625 struct batch_options batch = {0};
626 int unknown_type = 0;
627
628 const struct option options[] = {
629 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
630 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
631 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
632 OPT_CMDMODE('e', NULL, &opt,
633 N_("exit with zero when there's no error"), 'e'),
634 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
635 OPT_CMDMODE(0, "textconv", &opt,
636 N_("for blob objects, run textconv on object's content"), 'c'),
637 OPT_CMDMODE(0, "filters", &opt,
638 N_("for blob objects, run filters on object's content"), 'w'),
639 OPT_STRING(0, "path", &force_path, N_("blob"),
640 N_("use a specific path for --textconv/--filters")),
641 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
642 N_("allow -s and -t to work with broken/corrupt objects")),
643 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
644 { OPTION_CALLBACK, 0, "batch", &batch, "format",
645 N_("show info and content of objects fed from the standard input"),
646 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
647 batch_option_callback },
648 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
649 N_("show info about objects fed from the standard input"),
650 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
651 batch_option_callback },
652 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
653 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
654 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
655 N_("show all objects with --batch or --batch-check")),
656 OPT_BOOL(0, "unordered", &batch.unordered,
657 N_("do not order --batch-all-objects output")),
658 OPT_END()
659 };
660
661 git_config(git_cat_file_config, NULL);
662
663 batch.buffer_output = -1;
664 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
665
666 if (opt) {
667 if (batch.enabled && (opt == 'c' || opt == 'w'))
668 batch.cmdmode = opt;
669 else if (argc == 1)
670 obj_name = argv[0];
671 else
672 usage_with_options(cat_file_usage, options);
673 }
674 if (!opt && !batch.enabled) {
675 if (argc == 2) {
676 exp_type = argv[0];
677 obj_name = argv[1];
678 } else
679 usage_with_options(cat_file_usage, options);
680 }
681 if (batch.enabled) {
682 if (batch.cmdmode != opt || argc)
683 usage_with_options(cat_file_usage, options);
684 if (batch.cmdmode && batch.all_objects)
685 die("--batch-all-objects cannot be combined with "
686 "--textconv nor with --filters");
687 }
688
689 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
690 usage_with_options(cat_file_usage, options);
691 }
692
693 if (force_path && opt != 'c' && opt != 'w') {
694 error("--path=<path> needs --textconv or --filters");
695 usage_with_options(cat_file_usage, options);
696 }
697
698 if (force_path && batch.enabled) {
699 error("--path=<path> incompatible with --batch");
700 usage_with_options(cat_file_usage, options);
701 }
702
703 if (batch.buffer_output < 0)
704 batch.buffer_output = batch.all_objects;
705
706 if (batch.enabled)
707 return batch_objects(&batch);
708
709 if (unknown_type && opt != 't' && opt != 's')
710 die("git cat-file --allow-unknown-type: use with -s or -t");
711 return cat_one_file(opt, exp_type, obj_name, unknown_type);
712 }