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
18struct 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
29static const char *force_path;
30
31static 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
53static 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
60static 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(obj_name, GET_OID_RECORD_PATH,
77 &oid, &obj_context))
78 die("Not a valid object name %s", obj_name);
79
80 if (!path)
81 path = obj_context.path;
82 if (obj_context.mode == S_IFINVALID)
83 obj_context.mode = 0100644;
84
85 buf = NULL;
86 switch (opt) {
87 case 't':
88 oi.type_name = &sb;
89 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
90 die("git cat-file: could not get object info");
91 if (sb.len) {
92 printf("%s\n", sb.buf);
93 strbuf_release(&sb);
94 return 0;
95 }
96 break;
97
98 case 's':
99 oi.sizep = &size;
100 if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
101 die("git cat-file: could not get object info");
102 printf("%lu\n", size);
103 return 0;
104
105 case 'e':
106 return !has_object_file(&oid);
107
108 case 'w':
109 if (!path)
110 die("git cat-file --filters %s: <object> must be "
111 "<sha1:path>", obj_name);
112
113 if (filter_object(path, obj_context.mode,
114 &oid, &buf, &size))
115 return -1;
116 break;
117
118 case 'c':
119 if (!path)
120 die("git cat-file --textconv %s: <object> must be <sha1:path>",
121 obj_name);
122
123 if (textconv_object(the_repository, path, obj_context.mode,
124 &oid, 1, &buf, &size))
125 break;
126 /* else fallthrough */
127
128 case 'p':
129 type = oid_object_info(the_repository, &oid, NULL);
130 if (type < 0)
131 die("Not a valid object name %s", obj_name);
132
133 /* custom pretty-print here */
134 if (type == OBJ_TREE) {
135 const char *ls_args[3] = { NULL };
136 ls_args[0] = "ls-tree";
137 ls_args[1] = obj_name;
138 return cmd_ls_tree(2, ls_args, NULL);
139 }
140
141 if (type == OBJ_BLOB)
142 return stream_blob(&oid);
143 buf = read_object_file(&oid, &type, &size);
144 if (!buf)
145 die("Cannot read object %s", obj_name);
146
147 /* otherwise just spit out the data */
148 break;
149
150 case 0:
151 if (type_from_string(exp_type) == OBJ_BLOB) {
152 struct object_id blob_oid;
153 if (oid_object_info(the_repository, &oid, NULL) == OBJ_TAG) {
154 char *buffer = read_object_file(&oid, &type,
155 &size);
156 const char *target;
157 if (!skip_prefix(buffer, "object ", &target) ||
158 get_oid_hex(target, &blob_oid))
159 die("%s not a valid tag", oid_to_hex(&oid));
160 free(buffer);
161 } else
162 oidcpy(&blob_oid, &oid);
163
164 if (oid_object_info(the_repository, &blob_oid, NULL) == OBJ_BLOB)
165 return stream_blob(&blob_oid);
166 /*
167 * we attempted to dereference a tag to a blob
168 * and failed; there may be new dereference
169 * mechanisms this code is not aware of.
170 * fall-back to the usual case.
171 */
172 }
173 buf = read_object_with_reference(&oid, exp_type, &size, NULL);
174 break;
175
176 default:
177 die("git cat-file: unknown option: %s", exp_type);
178 }
179
180 if (!buf)
181 die("git cat-file %s: bad file", obj_name);
182
183 write_or_die(1, buf, size);
184 free(buf);
185 free(obj_context.path);
186 return 0;
187}
188
189struct expand_data {
190 struct object_id oid;
191 enum object_type type;
192 unsigned long size;
193 off_t disk_size;
194 const char *rest;
195 struct object_id delta_base_oid;
196
197 /*
198 * If mark_query is true, we do not expand anything, but rather
199 * just mark the object_info with items we wish to query.
200 */
201 int mark_query;
202
203 /*
204 * Whether to split the input on whitespace before feeding it to
205 * get_sha1; this is decided during the mark_query phase based on
206 * whether we have a %(rest) token in our format.
207 */
208 int split_on_whitespace;
209
210 /*
211 * After a mark_query run, this object_info is set up to be
212 * passed to sha1_object_info_extended. It will point to the data
213 * elements above, so you can retrieve the response from there.
214 */
215 struct object_info info;
216
217 /*
218 * This flag will be true if the requested batch format and options
219 * don't require us to call sha1_object_info, which can then be
220 * optimized out.
221 */
222 unsigned skip_object_info : 1;
223};
224
225static int is_atom(const char *atom, const char *s, int slen)
226{
227 int alen = strlen(atom);
228 return alen == slen && !memcmp(atom, s, alen);
229}
230
231static void expand_atom(struct strbuf *sb, const char *atom, int len,
232 void *vdata)
233{
234 struct expand_data *data = vdata;
235
236 if (is_atom("objectname", atom, len)) {
237 if (!data->mark_query)
238 strbuf_addstr(sb, oid_to_hex(&data->oid));
239 } else if (is_atom("objecttype", atom, len)) {
240 if (data->mark_query)
241 data->info.typep = &data->type;
242 else
243 strbuf_addstr(sb, type_name(data->type));
244 } else if (is_atom("objectsize", atom, len)) {
245 if (data->mark_query)
246 data->info.sizep = &data->size;
247 else
248 strbuf_addf(sb, "%lu", data->size);
249 } else if (is_atom("objectsize:disk", atom, len)) {
250 if (data->mark_query)
251 data->info.disk_sizep = &data->disk_size;
252 else
253 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
254 } else if (is_atom("rest", atom, len)) {
255 if (data->mark_query)
256 data->split_on_whitespace = 1;
257 else if (data->rest)
258 strbuf_addstr(sb, data->rest);
259 } else if (is_atom("deltabase", atom, len)) {
260 if (data->mark_query)
261 data->info.delta_base_sha1 = data->delta_base_oid.hash;
262 else
263 strbuf_addstr(sb,
264 oid_to_hex(&data->delta_base_oid));
265 } else
266 die("unknown format element: %.*s", len, atom);
267}
268
269static size_t expand_format(struct strbuf *sb, const char *start, void *data)
270{
271 const char *end;
272
273 if (*start != '(')
274 return 0;
275 end = strchr(start + 1, ')');
276 if (!end)
277 die("format element '%s' does not end in ')'", start);
278
279 expand_atom(sb, start + 1, end - start - 1, data);
280
281 return end - start + 1;
282}
283
284static void batch_write(struct batch_options *opt, const void *data, int len)
285{
286 if (opt->buffer_output) {
287 if (fwrite(data, 1, len, stdout) != len)
288 die_errno("unable to write to stdout");
289 } else
290 write_or_die(1, data, len);
291}
292
293static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
294{
295 const struct object_id *oid = &data->oid;
296
297 assert(data->info.typep);
298
299 if (data->type == OBJ_BLOB) {
300 if (opt->buffer_output)
301 fflush(stdout);
302 if (opt->cmdmode) {
303 char *contents;
304 unsigned long size;
305
306 if (!data->rest)
307 die("missing path for '%s'", oid_to_hex(oid));
308
309 if (opt->cmdmode == 'w') {
310 if (filter_object(data->rest, 0100644, oid,
311 &contents, &size))
312 die("could not convert '%s' %s",
313 oid_to_hex(oid), data->rest);
314 } else if (opt->cmdmode == 'c') {
315 enum object_type type;
316 if (!textconv_object(the_repository,
317 data->rest, 0100644, oid,
318 1, &contents, &size))
319 contents = read_object_file(oid,
320 &type,
321 &size);
322 if (!contents)
323 die("could not convert '%s' %s",
324 oid_to_hex(oid), data->rest);
325 } else
326 BUG("invalid cmdmode: %c", opt->cmdmode);
327 batch_write(opt, contents, size);
328 free(contents);
329 } else {
330 stream_blob(oid);
331 }
332 }
333 else {
334 enum object_type type;
335 unsigned long size;
336 void *contents;
337
338 contents = read_object_file(oid, &type, &size);
339 if (!contents)
340 die("object %s disappeared", oid_to_hex(oid));
341 if (type != data->type)
342 die("object %s changed type!?", oid_to_hex(oid));
343 if (data->info.sizep && size != data->size)
344 die("object %s changed size!?", oid_to_hex(oid));
345
346 batch_write(opt, contents, size);
347 free(contents);
348 }
349}
350
351static void batch_object_write(const char *obj_name,
352 struct strbuf *scratch,
353 struct batch_options *opt,
354 struct expand_data *data)
355{
356 if (!data->skip_object_info &&
357 oid_object_info_extended(the_repository, &data->oid, &data->info,
358 OBJECT_INFO_LOOKUP_REPLACE) < 0) {
359 printf("%s missing\n",
360 obj_name ? obj_name : oid_to_hex(&data->oid));
361 fflush(stdout);
362 return;
363 }
364
365 strbuf_reset(scratch);
366 strbuf_expand(scratch, opt->format, expand_format, data);
367 strbuf_addch(scratch, '\n');
368 batch_write(opt, scratch->buf, scratch->len);
369
370 if (opt->print_contents) {
371 print_object_or_die(opt, data);
372 batch_write(opt, "\n", 1);
373 }
374}
375
376static void batch_one_object(const char *obj_name,
377 struct strbuf *scratch,
378 struct batch_options *opt,
379 struct expand_data *data)
380{
381 struct object_context ctx;
382 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
383 enum follow_symlinks_result result;
384
385 result = get_oid_with_context(obj_name, flags, &data->oid, &ctx);
386 if (result != FOUND) {
387 switch (result) {
388 case MISSING_OBJECT:
389 printf("%s missing\n", obj_name);
390 break;
391 case DANGLING_SYMLINK:
392 printf("dangling %"PRIuMAX"\n%s\n",
393 (uintmax_t)strlen(obj_name), obj_name);
394 break;
395 case SYMLINK_LOOP:
396 printf("loop %"PRIuMAX"\n%s\n",
397 (uintmax_t)strlen(obj_name), obj_name);
398 break;
399 case NOT_DIR:
400 printf("notdir %"PRIuMAX"\n%s\n",
401 (uintmax_t)strlen(obj_name), obj_name);
402 break;
403 default:
404 BUG("unknown get_sha1_with_context result %d\n",
405 result);
406 break;
407 }
408 fflush(stdout);
409 return;
410 }
411
412 if (ctx.mode == 0) {
413 printf("symlink %"PRIuMAX"\n%s\n",
414 (uintmax_t)ctx.symlink_path.len,
415 ctx.symlink_path.buf);
416 fflush(stdout);
417 return;
418 }
419
420 batch_object_write(obj_name, scratch, opt, data);
421}
422
423struct object_cb_data {
424 struct batch_options *opt;
425 struct expand_data *expand;
426 struct oidset *seen;
427 struct strbuf *scratch;
428};
429
430static int batch_object_cb(const struct object_id *oid, void *vdata)
431{
432 struct object_cb_data *data = vdata;
433 oidcpy(&data->expand->oid, oid);
434 batch_object_write(NULL, data->scratch, data->opt, data->expand);
435 return 0;
436}
437
438static int collect_loose_object(const struct object_id *oid,
439 const char *path,
440 void *data)
441{
442 oid_array_append(data, oid);
443 return 0;
444}
445
446static int collect_packed_object(const struct object_id *oid,
447 struct packed_git *pack,
448 uint32_t pos,
449 void *data)
450{
451 oid_array_append(data, oid);
452 return 0;
453}
454
455static int batch_unordered_object(const struct object_id *oid, void *vdata)
456{
457 struct object_cb_data *data = vdata;
458
459 if (oidset_insert(data->seen, oid))
460 return 0;
461
462 return batch_object_cb(oid, data);
463}
464
465static int batch_unordered_loose(const struct object_id *oid,
466 const char *path,
467 void *data)
468{
469 return batch_unordered_object(oid, data);
470}
471
472static int batch_unordered_packed(const struct object_id *oid,
473 struct packed_git *pack,
474 uint32_t pos,
475 void *data)
476{
477 return batch_unordered_object(oid, data);
478}
479
480static int batch_objects(struct batch_options *opt)
481{
482 struct strbuf input = STRBUF_INIT;
483 struct strbuf output = STRBUF_INIT;
484 struct expand_data data;
485 int save_warning;
486 int retval = 0;
487
488 if (!opt->format)
489 opt->format = "%(objectname) %(objecttype) %(objectsize)";
490
491 /*
492 * Expand once with our special mark_query flag, which will prime the
493 * object_info to be handed to sha1_object_info_extended for each
494 * object.
495 */
496 memset(&data, 0, sizeof(data));
497 data.mark_query = 1;
498 strbuf_expand(&output, opt->format, expand_format, &data);
499 data.mark_query = 0;
500 strbuf_release(&output);
501 if (opt->cmdmode)
502 data.split_on_whitespace = 1;
503
504 if (opt->all_objects) {
505 struct object_info empty = OBJECT_INFO_INIT;
506 if (!memcmp(&data.info, &empty, sizeof(empty)))
507 data.skip_object_info = 1;
508 }
509
510 /*
511 * If we are printing out the object, then always fill in the type,
512 * since we will want to decide whether or not to stream.
513 */
514 if (opt->print_contents)
515 data.info.typep = &data.type;
516
517 if (opt->all_objects) {
518 struct object_cb_data cb;
519
520 if (repository_format_partial_clone)
521 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
522
523 cb.opt = opt;
524 cb.expand = &data;
525 cb.scratch = &output;
526
527 if (opt->unordered) {
528 struct oidset seen = OIDSET_INIT;
529
530 cb.seen = &seen;
531
532 for_each_loose_object(batch_unordered_loose, &cb, 0);
533 for_each_packed_object(batch_unordered_packed, &cb,
534 FOR_EACH_OBJECT_PACK_ORDER);
535
536 oidset_clear(&seen);
537 } else {
538 struct oid_array sa = OID_ARRAY_INIT;
539
540 for_each_loose_object(collect_loose_object, &sa, 0);
541 for_each_packed_object(collect_packed_object, &sa, 0);
542
543 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
544
545 oid_array_clear(&sa);
546 }
547
548 strbuf_release(&output);
549 return 0;
550 }
551
552 /*
553 * We are going to call get_sha1 on a potentially very large number of
554 * objects. In most large cases, these will be actual object sha1s. The
555 * cost to double-check that each one is not also a ref (just so we can
556 * warn) ends up dwarfing the actual cost of the object lookups
557 * themselves. We can work around it by just turning off the warning.
558 */
559 save_warning = warn_on_object_refname_ambiguity;
560 warn_on_object_refname_ambiguity = 0;
561
562 while (strbuf_getline(&input, stdin) != EOF) {
563 if (data.split_on_whitespace) {
564 /*
565 * Split at first whitespace, tying off the beginning
566 * of the string and saving the remainder (or NULL) in
567 * data.rest.
568 */
569 char *p = strpbrk(input.buf, " \t");
570 if (p) {
571 while (*p && strchr(" \t", *p))
572 *p++ = '\0';
573 }
574 data.rest = p;
575 }
576
577 batch_one_object(input.buf, &output, opt, &data);
578 }
579
580 strbuf_release(&input);
581 strbuf_release(&output);
582 warn_on_object_refname_ambiguity = save_warning;
583 return retval;
584}
585
586static const char * const cat_file_usage[] = {
587 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
588 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
589 NULL
590};
591
592static int git_cat_file_config(const char *var, const char *value, void *cb)
593{
594 if (userdiff_config(var, value) < 0)
595 return -1;
596
597 return git_default_config(var, value, cb);
598}
599
600static int batch_option_callback(const struct option *opt,
601 const char *arg,
602 int unset)
603{
604 struct batch_options *bo = opt->value;
605
606 if (bo->enabled) {
607 return 1;
608 }
609
610 bo->enabled = 1;
611 bo->print_contents = !strcmp(opt->long_name, "batch");
612 bo->format = arg;
613
614 return 0;
615}
616
617int cmd_cat_file(int argc, const char **argv, const char *prefix)
618{
619 int opt = 0;
620 const char *exp_type = NULL, *obj_name = NULL;
621 struct batch_options batch = {0};
622 int unknown_type = 0;
623
624 const struct option options[] = {
625 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
626 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
627 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
628 OPT_CMDMODE('e', NULL, &opt,
629 N_("exit with zero when there's no error"), 'e'),
630 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
631 OPT_CMDMODE(0, "textconv", &opt,
632 N_("for blob objects, run textconv on object's content"), 'c'),
633 OPT_CMDMODE(0, "filters", &opt,
634 N_("for blob objects, run filters on object's content"), 'w'),
635 OPT_STRING(0, "path", &force_path, N_("blob"),
636 N_("use a specific path for --textconv/--filters")),
637 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
638 N_("allow -s and -t to work with broken/corrupt objects")),
639 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
640 { OPTION_CALLBACK, 0, "batch", &batch, "format",
641 N_("show info and content of objects fed from the standard input"),
642 PARSE_OPT_OPTARG, batch_option_callback },
643 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
644 N_("show info about objects fed from the standard input"),
645 PARSE_OPT_OPTARG, batch_option_callback },
646 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
647 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
648 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
649 N_("show all objects with --batch or --batch-check")),
650 OPT_BOOL(0, "unordered", &batch.unordered,
651 N_("do not order --batch-all-objects output")),
652 OPT_END()
653 };
654
655 git_config(git_cat_file_config, NULL);
656
657 batch.buffer_output = -1;
658 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
659
660 if (opt) {
661 if (batch.enabled && (opt == 'c' || opt == 'w'))
662 batch.cmdmode = opt;
663 else if (argc == 1)
664 obj_name = argv[0];
665 else
666 usage_with_options(cat_file_usage, options);
667 }
668 if (!opt && !batch.enabled) {
669 if (argc == 2) {
670 exp_type = argv[0];
671 obj_name = argv[1];
672 } else
673 usage_with_options(cat_file_usage, options);
674 }
675 if (batch.enabled) {
676 if (batch.cmdmode != opt || argc)
677 usage_with_options(cat_file_usage, options);
678 if (batch.cmdmode && batch.all_objects)
679 die("--batch-all-objects cannot be combined with "
680 "--textconv nor with --filters");
681 }
682
683 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
684 usage_with_options(cat_file_usage, options);
685 }
686
687 if (force_path && opt != 'c' && opt != 'w') {
688 error("--path=<path> needs --textconv or --filters");
689 usage_with_options(cat_file_usage, options);
690 }
691
692 if (force_path && batch.enabled) {
693 error("--path=<path> incompatible with --batch");
694 usage_with_options(cat_file_usage, options);
695 }
696
697 if (batch.buffer_output < 0)
698 batch.buffer_output = batch.all_objects;
699
700 if (batch.enabled)
701 return batch_objects(&batch);
702
703 if (unknown_type && opt != 't' && opt != 's')
704 die("git cat-file --allow-unknown-type: use with -s or -t");
705 return cat_one_file(opt, exp_type, obj_name, unknown_type);
706}