doc: do not use `rm .git/index` when normalizing line endings
[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 "builtin.h"
8 #include "parse-options.h"
9 #include "userdiff.h"
10 #include "streaming.h"
11 #include "tree-walk.h"
12 #include "sha1-array.h"
13
14 struct batch_options {
15 int enabled;
16 int follow_symlinks;
17 int print_contents;
18 int buffer_output;
19 int all_objects;
20 int cmdmode; /* may be 'w' or 'c' for --filters or --textconv */
21 const char *format;
22 };
23
24 static const char *force_path;
25
26 static int filter_object(const char *path, unsigned mode,
27 const struct object_id *oid,
28 char **buf, unsigned long *size)
29 {
30 enum object_type type;
31
32 *buf = read_sha1_file(oid->hash, &type, size);
33 if (!*buf)
34 return error(_("cannot read object %s '%s'"),
35 oid_to_hex(oid), path);
36 if ((type == OBJ_BLOB) && S_ISREG(mode)) {
37 struct strbuf strbuf = STRBUF_INIT;
38 if (convert_to_working_tree(path, *buf, *size, &strbuf)) {
39 free(*buf);
40 *size = strbuf.len;
41 *buf = strbuf_detach(&strbuf, NULL);
42 }
43 }
44
45 return 0;
46 }
47
48 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
49 int unknown_type)
50 {
51 struct object_id oid;
52 enum object_type type;
53 char *buf;
54 unsigned long size;
55 struct object_context obj_context;
56 struct object_info oi = OBJECT_INFO_INIT;
57 struct strbuf sb = STRBUF_INIT;
58 unsigned flags = LOOKUP_REPLACE_OBJECT;
59 const char *path = force_path;
60
61 if (unknown_type)
62 flags |= LOOKUP_UNKNOWN_OBJECT;
63
64 if (get_sha1_with_context(obj_name, 0, oid.hash, &obj_context))
65 die("Not a valid object name %s", obj_name);
66
67 if (!path)
68 path = obj_context.path;
69 if (obj_context.mode == S_IFINVALID)
70 obj_context.mode = 0100644;
71
72 buf = NULL;
73 switch (opt) {
74 case 't':
75 oi.typename = &sb;
76 if (sha1_object_info_extended(oid.hash, &oi, flags) < 0)
77 die("git cat-file: could not get object info");
78 if (sb.len) {
79 printf("%s\n", sb.buf);
80 strbuf_release(&sb);
81 return 0;
82 }
83 break;
84
85 case 's':
86 oi.sizep = &size;
87 if (sha1_object_info_extended(oid.hash, &oi, flags) < 0)
88 die("git cat-file: could not get object info");
89 printf("%lu\n", size);
90 return 0;
91
92 case 'e':
93 return !has_object_file(&oid);
94
95 case 'w':
96 if (!path[0])
97 die("git cat-file --filters %s: <object> must be "
98 "<sha1:path>", obj_name);
99
100 if (filter_object(path, obj_context.mode,
101 &oid, &buf, &size))
102 return -1;
103 break;
104
105 case 'c':
106 if (!path[0])
107 die("git cat-file --textconv %s: <object> must be <sha1:path>",
108 obj_name);
109
110 if (textconv_object(path, obj_context.mode, &oid, 1, &buf, &size))
111 break;
112
113 case 'p':
114 type = sha1_object_info(oid.hash, NULL);
115 if (type < 0)
116 die("Not a valid object name %s", obj_name);
117
118 /* custom pretty-print here */
119 if (type == OBJ_TREE) {
120 const char *ls_args[3] = { NULL };
121 ls_args[0] = "ls-tree";
122 ls_args[1] = obj_name;
123 return cmd_ls_tree(2, ls_args, NULL);
124 }
125
126 if (type == OBJ_BLOB)
127 return stream_blob_to_fd(1, &oid, NULL, 0);
128 buf = read_sha1_file(oid.hash, &type, &size);
129 if (!buf)
130 die("Cannot read object %s", obj_name);
131
132 /* otherwise just spit out the data */
133 break;
134
135 case 0:
136 if (type_from_string(exp_type) == OBJ_BLOB) {
137 struct object_id blob_oid;
138 if (sha1_object_info(oid.hash, NULL) == OBJ_TAG) {
139 char *buffer = read_sha1_file(oid.hash, &type, &size);
140 const char *target;
141 if (!skip_prefix(buffer, "object ", &target) ||
142 get_oid_hex(target, &blob_oid))
143 die("%s not a valid tag", oid_to_hex(&oid));
144 free(buffer);
145 } else
146 oidcpy(&blob_oid, &oid);
147
148 if (sha1_object_info(blob_oid.hash, NULL) == OBJ_BLOB)
149 return stream_blob_to_fd(1, &blob_oid, NULL, 0);
150 /*
151 * we attempted to dereference a tag to a blob
152 * and failed; there may be new dereference
153 * mechanisms this code is not aware of.
154 * fall-back to the usual case.
155 */
156 }
157 buf = read_object_with_reference(oid.hash, exp_type, &size, NULL);
158 break;
159
160 default:
161 die("git cat-file: unknown option: %s", exp_type);
162 }
163
164 if (!buf)
165 die("git cat-file %s: bad file", obj_name);
166
167 write_or_die(1, buf, size);
168 free(buf);
169 return 0;
170 }
171
172 struct expand_data {
173 struct object_id oid;
174 enum object_type type;
175 unsigned long size;
176 off_t disk_size;
177 const char *rest;
178 struct object_id delta_base_oid;
179
180 /*
181 * If mark_query is true, we do not expand anything, but rather
182 * just mark the object_info with items we wish to query.
183 */
184 int mark_query;
185
186 /*
187 * Whether to split the input on whitespace before feeding it to
188 * get_sha1; this is decided during the mark_query phase based on
189 * whether we have a %(rest) token in our format.
190 */
191 int split_on_whitespace;
192
193 /*
194 * After a mark_query run, this object_info is set up to be
195 * passed to sha1_object_info_extended. It will point to the data
196 * elements above, so you can retrieve the response from there.
197 */
198 struct object_info info;
199
200 /*
201 * This flag will be true if the requested batch format and options
202 * don't require us to call sha1_object_info, which can then be
203 * optimized out.
204 */
205 unsigned skip_object_info : 1;
206 };
207
208 static int is_atom(const char *atom, const char *s, int slen)
209 {
210 int alen = strlen(atom);
211 return alen == slen && !memcmp(atom, s, alen);
212 }
213
214 static void expand_atom(struct strbuf *sb, const char *atom, int len,
215 void *vdata)
216 {
217 struct expand_data *data = vdata;
218
219 if (is_atom("objectname", atom, len)) {
220 if (!data->mark_query)
221 strbuf_addstr(sb, oid_to_hex(&data->oid));
222 } else if (is_atom("objecttype", atom, len)) {
223 if (data->mark_query)
224 data->info.typep = &data->type;
225 else
226 strbuf_addstr(sb, typename(data->type));
227 } else if (is_atom("objectsize", atom, len)) {
228 if (data->mark_query)
229 data->info.sizep = &data->size;
230 else
231 strbuf_addf(sb, "%lu", data->size);
232 } else if (is_atom("objectsize:disk", atom, len)) {
233 if (data->mark_query)
234 data->info.disk_sizep = &data->disk_size;
235 else
236 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
237 } else if (is_atom("rest", atom, len)) {
238 if (data->mark_query)
239 data->split_on_whitespace = 1;
240 else if (data->rest)
241 strbuf_addstr(sb, data->rest);
242 } else if (is_atom("deltabase", atom, len)) {
243 if (data->mark_query)
244 data->info.delta_base_sha1 = data->delta_base_oid.hash;
245 else
246 strbuf_addstr(sb,
247 oid_to_hex(&data->delta_base_oid));
248 } else
249 die("unknown format element: %.*s", len, atom);
250 }
251
252 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
253 {
254 const char *end;
255
256 if (*start != '(')
257 return 0;
258 end = strchr(start + 1, ')');
259 if (!end)
260 die("format element '%s' does not end in ')'", start);
261
262 expand_atom(sb, start + 1, end - start - 1, data);
263
264 return end - start + 1;
265 }
266
267 static void batch_write(struct batch_options *opt, const void *data, int len)
268 {
269 if (opt->buffer_output) {
270 if (fwrite(data, 1, len, stdout) != len)
271 die_errno("unable to write to stdout");
272 } else
273 write_or_die(1, data, len);
274 }
275
276 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
277 {
278 const struct object_id *oid = &data->oid;
279
280 assert(data->info.typep);
281
282 if (data->type == OBJ_BLOB) {
283 if (opt->buffer_output)
284 fflush(stdout);
285 if (opt->cmdmode) {
286 char *contents;
287 unsigned long size;
288
289 if (!data->rest)
290 die("missing path for '%s'", oid_to_hex(oid));
291
292 if (opt->cmdmode == 'w') {
293 if (filter_object(data->rest, 0100644, oid,
294 &contents, &size))
295 die("could not convert '%s' %s",
296 oid_to_hex(oid), data->rest);
297 } else if (opt->cmdmode == 'c') {
298 enum object_type type;
299 if (!textconv_object(data->rest, 0100644, oid,
300 1, &contents, &size))
301 contents = read_sha1_file(oid->hash, &type,
302 &size);
303 if (!contents)
304 die("could not convert '%s' %s",
305 oid_to_hex(oid), data->rest);
306 } else
307 die("BUG: invalid cmdmode: %c", opt->cmdmode);
308 batch_write(opt, contents, size);
309 free(contents);
310 } else if (stream_blob_to_fd(1, oid, NULL, 0) < 0)
311 die("unable to stream %s to stdout", oid_to_hex(oid));
312 }
313 else {
314 enum object_type type;
315 unsigned long size;
316 void *contents;
317
318 contents = read_sha1_file(oid->hash, &type, &size);
319 if (!contents)
320 die("object %s disappeared", oid_to_hex(oid));
321 if (type != data->type)
322 die("object %s changed type!?", oid_to_hex(oid));
323 if (data->info.sizep && size != data->size)
324 die("object %s changed size!?", oid_to_hex(oid));
325
326 batch_write(opt, contents, size);
327 free(contents);
328 }
329 }
330
331 static void batch_object_write(const char *obj_name, struct batch_options *opt,
332 struct expand_data *data)
333 {
334 struct strbuf buf = STRBUF_INIT;
335
336 if (!data->skip_object_info &&
337 sha1_object_info_extended(data->oid.hash, &data->info, LOOKUP_REPLACE_OBJECT) < 0) {
338 printf("%s missing\n",
339 obj_name ? obj_name : oid_to_hex(&data->oid));
340 fflush(stdout);
341 return;
342 }
343
344 strbuf_expand(&buf, opt->format, expand_format, data);
345 strbuf_addch(&buf, '\n');
346 batch_write(opt, buf.buf, buf.len);
347 strbuf_release(&buf);
348
349 if (opt->print_contents) {
350 print_object_or_die(opt, data);
351 batch_write(opt, "\n", 1);
352 }
353 }
354
355 static void batch_one_object(const char *obj_name, struct batch_options *opt,
356 struct expand_data *data)
357 {
358 struct object_context ctx;
359 int flags = opt->follow_symlinks ? GET_SHA1_FOLLOW_SYMLINKS : 0;
360 enum follow_symlinks_result result;
361
362 result = get_sha1_with_context(obj_name, flags, data->oid.hash, &ctx);
363 if (result != FOUND) {
364 switch (result) {
365 case MISSING_OBJECT:
366 printf("%s missing\n", obj_name);
367 break;
368 case DANGLING_SYMLINK:
369 printf("dangling %"PRIuMAX"\n%s\n",
370 (uintmax_t)strlen(obj_name), obj_name);
371 break;
372 case SYMLINK_LOOP:
373 printf("loop %"PRIuMAX"\n%s\n",
374 (uintmax_t)strlen(obj_name), obj_name);
375 break;
376 case NOT_DIR:
377 printf("notdir %"PRIuMAX"\n%s\n",
378 (uintmax_t)strlen(obj_name), obj_name);
379 break;
380 default:
381 die("BUG: unknown get_sha1_with_context result %d\n",
382 result);
383 break;
384 }
385 fflush(stdout);
386 return;
387 }
388
389 if (ctx.mode == 0) {
390 printf("symlink %"PRIuMAX"\n%s\n",
391 (uintmax_t)ctx.symlink_path.len,
392 ctx.symlink_path.buf);
393 fflush(stdout);
394 return;
395 }
396
397 batch_object_write(obj_name, opt, data);
398 }
399
400 struct object_cb_data {
401 struct batch_options *opt;
402 struct expand_data *expand;
403 };
404
405 static int batch_object_cb(const struct object_id *oid, void *vdata)
406 {
407 struct object_cb_data *data = vdata;
408 oidcpy(&data->expand->oid, oid);
409 batch_object_write(NULL, data->opt, data->expand);
410 return 0;
411 }
412
413 static int batch_loose_object(const struct object_id *oid,
414 const char *path,
415 void *data)
416 {
417 oid_array_append(data, oid);
418 return 0;
419 }
420
421 static int batch_packed_object(const struct object_id *oid,
422 struct packed_git *pack,
423 uint32_t pos,
424 void *data)
425 {
426 oid_array_append(data, oid);
427 return 0;
428 }
429
430 static int batch_objects(struct batch_options *opt)
431 {
432 struct strbuf buf = STRBUF_INIT;
433 struct expand_data data;
434 int save_warning;
435 int retval = 0;
436
437 if (!opt->format)
438 opt->format = "%(objectname) %(objecttype) %(objectsize)";
439
440 /*
441 * Expand once with our special mark_query flag, which will prime the
442 * object_info to be handed to sha1_object_info_extended for each
443 * object.
444 */
445 memset(&data, 0, sizeof(data));
446 data.mark_query = 1;
447 strbuf_expand(&buf, opt->format, expand_format, &data);
448 data.mark_query = 0;
449 if (opt->cmdmode)
450 data.split_on_whitespace = 1;
451
452 if (opt->all_objects) {
453 struct object_info empty = OBJECT_INFO_INIT;
454 if (!memcmp(&data.info, &empty, sizeof(empty)))
455 data.skip_object_info = 1;
456 }
457
458 /*
459 * If we are printing out the object, then always fill in the type,
460 * since we will want to decide whether or not to stream.
461 */
462 if (opt->print_contents)
463 data.info.typep = &data.type;
464
465 if (opt->all_objects) {
466 struct oid_array sa = OID_ARRAY_INIT;
467 struct object_cb_data cb;
468
469 for_each_loose_object(batch_loose_object, &sa, 0);
470 for_each_packed_object(batch_packed_object, &sa, 0);
471
472 cb.opt = opt;
473 cb.expand = &data;
474 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
475
476 oid_array_clear(&sa);
477 return 0;
478 }
479
480 /*
481 * We are going to call get_sha1 on a potentially very large number of
482 * objects. In most large cases, these will be actual object sha1s. The
483 * cost to double-check that each one is not also a ref (just so we can
484 * warn) ends up dwarfing the actual cost of the object lookups
485 * themselves. We can work around it by just turning off the warning.
486 */
487 save_warning = warn_on_object_refname_ambiguity;
488 warn_on_object_refname_ambiguity = 0;
489
490 while (strbuf_getline(&buf, stdin) != EOF) {
491 if (data.split_on_whitespace) {
492 /*
493 * Split at first whitespace, tying off the beginning
494 * of the string and saving the remainder (or NULL) in
495 * data.rest.
496 */
497 char *p = strpbrk(buf.buf, " \t");
498 if (p) {
499 while (*p && strchr(" \t", *p))
500 *p++ = '\0';
501 }
502 data.rest = p;
503 }
504
505 batch_one_object(buf.buf, opt, &data);
506 }
507
508 strbuf_release(&buf);
509 warn_on_object_refname_ambiguity = save_warning;
510 return retval;
511 }
512
513 static const char * const cat_file_usage[] = {
514 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
515 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
516 NULL
517 };
518
519 static int git_cat_file_config(const char *var, const char *value, void *cb)
520 {
521 if (userdiff_config(var, value) < 0)
522 return -1;
523
524 return git_default_config(var, value, cb);
525 }
526
527 static int batch_option_callback(const struct option *opt,
528 const char *arg,
529 int unset)
530 {
531 struct batch_options *bo = opt->value;
532
533 if (bo->enabled) {
534 return 1;
535 }
536
537 bo->enabled = 1;
538 bo->print_contents = !strcmp(opt->long_name, "batch");
539 bo->format = arg;
540
541 return 0;
542 }
543
544 int cmd_cat_file(int argc, const char **argv, const char *prefix)
545 {
546 int opt = 0;
547 const char *exp_type = NULL, *obj_name = NULL;
548 struct batch_options batch = {0};
549 int unknown_type = 0;
550
551 const struct option options[] = {
552 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
553 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
554 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
555 OPT_CMDMODE('e', NULL, &opt,
556 N_("exit with zero when there's no error"), 'e'),
557 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
558 OPT_CMDMODE(0, "textconv", &opt,
559 N_("for blob objects, run textconv on object's content"), 'c'),
560 OPT_CMDMODE(0, "filters", &opt,
561 N_("for blob objects, run filters on object's content"), 'w'),
562 OPT_STRING(0, "path", &force_path, N_("blob"),
563 N_("use a specific path for --textconv/--filters")),
564 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
565 N_("allow -s and -t to work with broken/corrupt objects")),
566 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
567 { OPTION_CALLBACK, 0, "batch", &batch, "format",
568 N_("show info and content of objects fed from the standard input"),
569 PARSE_OPT_OPTARG, batch_option_callback },
570 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
571 N_("show info about objects fed from the standard input"),
572 PARSE_OPT_OPTARG, batch_option_callback },
573 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
574 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
575 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
576 N_("show all objects with --batch or --batch-check")),
577 OPT_END()
578 };
579
580 git_config(git_cat_file_config, NULL);
581
582 batch.buffer_output = -1;
583 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
584
585 if (opt) {
586 if (batch.enabled && (opt == 'c' || opt == 'w'))
587 batch.cmdmode = opt;
588 else if (argc == 1)
589 obj_name = argv[0];
590 else
591 usage_with_options(cat_file_usage, options);
592 }
593 if (!opt && !batch.enabled) {
594 if (argc == 2) {
595 exp_type = argv[0];
596 obj_name = argv[1];
597 } else
598 usage_with_options(cat_file_usage, options);
599 }
600 if (batch.enabled) {
601 if (batch.cmdmode != opt || argc)
602 usage_with_options(cat_file_usage, options);
603 if (batch.cmdmode && batch.all_objects)
604 die("--batch-all-objects cannot be combined with "
605 "--textconv nor with --filters");
606 }
607
608 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
609 usage_with_options(cat_file_usage, options);
610 }
611
612 if (force_path && opt != 'c' && opt != 'w') {
613 error("--path=<path> needs --textconv or --filters");
614 usage_with_options(cat_file_usage, options);
615 }
616
617 if (force_path && batch.enabled) {
618 error("--path=<path> incompatible with --batch");
619 usage_with_options(cat_file_usage, options);
620 }
621
622 if (batch.buffer_output < 0)
623 batch.buffer_output = batch.all_objects;
624
625 if (batch.enabled)
626 return batch_objects(&batch);
627
628 if (unknown_type && opt != 't' && opt != 's')
629 die("git cat-file --allow-unknown-type: use with -s or -t");
630 return cat_one_file(opt, exp_type, obj_name, unknown_type);
631 }