Merge branch 'ah/doc-gitattributes-empty-index' into maint
[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, GET_SHA1_RECORD_PATH,
65 oid.hash, &obj_context))
66 die("Not a valid object name %s", obj_name);
67
68 if (!path)
69 path = obj_context.path;
70 if (obj_context.mode == S_IFINVALID)
71 obj_context.mode = 0100644;
72
73 buf = NULL;
74 switch (opt) {
75 case 't':
76 oi.typename = &sb;
77 if (sha1_object_info_extended(oid.hash, &oi, flags) < 0)
78 die("git cat-file: could not get object info");
79 if (sb.len) {
80 printf("%s\n", sb.buf);
81 strbuf_release(&sb);
82 return 0;
83 }
84 break;
85
86 case 's':
87 oi.sizep = &size;
88 if (sha1_object_info_extended(oid.hash, &oi, flags) < 0)
89 die("git cat-file: could not get object info");
90 printf("%lu\n", size);
91 return 0;
92
93 case 'e':
94 return !has_object_file(&oid);
95
96 case 'w':
97 if (!path[0])
98 die("git cat-file --filters %s: <object> must be "
99 "<sha1:path>", obj_name);
100
101 if (filter_object(path, obj_context.mode,
102 &oid, &buf, &size))
103 return -1;
104 break;
105
106 case 'c':
107 if (!path[0])
108 die("git cat-file --textconv %s: <object> must be <sha1:path>",
109 obj_name);
110
111 if (textconv_object(path, obj_context.mode, &oid, 1, &buf, &size))
112 break;
113
114 case 'p':
115 type = sha1_object_info(oid.hash, NULL);
116 if (type < 0)
117 die("Not a valid object name %s", obj_name);
118
119 /* custom pretty-print here */
120 if (type == OBJ_TREE) {
121 const char *ls_args[3] = { NULL };
122 ls_args[0] = "ls-tree";
123 ls_args[1] = obj_name;
124 return cmd_ls_tree(2, ls_args, NULL);
125 }
126
127 if (type == OBJ_BLOB)
128 return stream_blob_to_fd(1, &oid, NULL, 0);
129 buf = read_sha1_file(oid.hash, &type, &size);
130 if (!buf)
131 die("Cannot read object %s", obj_name);
132
133 /* otherwise just spit out the data */
134 break;
135
136 case 0:
137 if (type_from_string(exp_type) == OBJ_BLOB) {
138 struct object_id blob_oid;
139 if (sha1_object_info(oid.hash, NULL) == OBJ_TAG) {
140 char *buffer = read_sha1_file(oid.hash, &type, &size);
141 const char *target;
142 if (!skip_prefix(buffer, "object ", &target) ||
143 get_oid_hex(target, &blob_oid))
144 die("%s not a valid tag", oid_to_hex(&oid));
145 free(buffer);
146 } else
147 oidcpy(&blob_oid, &oid);
148
149 if (sha1_object_info(blob_oid.hash, NULL) == OBJ_BLOB)
150 return stream_blob_to_fd(1, &blob_oid, NULL, 0);
151 /*
152 * we attempted to dereference a tag to a blob
153 * and failed; there may be new dereference
154 * mechanisms this code is not aware of.
155 * fall-back to the usual case.
156 */
157 }
158 buf = read_object_with_reference(oid.hash, exp_type, &size, NULL);
159 break;
160
161 default:
162 die("git cat-file: unknown option: %s", exp_type);
163 }
164
165 if (!buf)
166 die("git cat-file %s: bad file", obj_name);
167
168 write_or_die(1, buf, size);
169 free(buf);
170 free(obj_context.path);
171 return 0;
172 }
173
174 struct expand_data {
175 struct object_id oid;
176 enum object_type type;
177 unsigned long size;
178 off_t disk_size;
179 const char *rest;
180 struct object_id delta_base_oid;
181
182 /*
183 * If mark_query is true, we do not expand anything, but rather
184 * just mark the object_info with items we wish to query.
185 */
186 int mark_query;
187
188 /*
189 * Whether to split the input on whitespace before feeding it to
190 * get_sha1; this is decided during the mark_query phase based on
191 * whether we have a %(rest) token in our format.
192 */
193 int split_on_whitespace;
194
195 /*
196 * After a mark_query run, this object_info is set up to be
197 * passed to sha1_object_info_extended. It will point to the data
198 * elements above, so you can retrieve the response from there.
199 */
200 struct object_info info;
201
202 /*
203 * This flag will be true if the requested batch format and options
204 * don't require us to call sha1_object_info, which can then be
205 * optimized out.
206 */
207 unsigned skip_object_info : 1;
208 };
209
210 static int is_atom(const char *atom, const char *s, int slen)
211 {
212 int alen = strlen(atom);
213 return alen == slen && !memcmp(atom, s, alen);
214 }
215
216 static void expand_atom(struct strbuf *sb, const char *atom, int len,
217 void *vdata)
218 {
219 struct expand_data *data = vdata;
220
221 if (is_atom("objectname", atom, len)) {
222 if (!data->mark_query)
223 strbuf_addstr(sb, oid_to_hex(&data->oid));
224 } else if (is_atom("objecttype", atom, len)) {
225 if (data->mark_query)
226 data->info.typep = &data->type;
227 else
228 strbuf_addstr(sb, typename(data->type));
229 } else if (is_atom("objectsize", atom, len)) {
230 if (data->mark_query)
231 data->info.sizep = &data->size;
232 else
233 strbuf_addf(sb, "%lu", data->size);
234 } else if (is_atom("objectsize:disk", atom, len)) {
235 if (data->mark_query)
236 data->info.disk_sizep = &data->disk_size;
237 else
238 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
239 } else if (is_atom("rest", atom, len)) {
240 if (data->mark_query)
241 data->split_on_whitespace = 1;
242 else if (data->rest)
243 strbuf_addstr(sb, data->rest);
244 } else if (is_atom("deltabase", atom, len)) {
245 if (data->mark_query)
246 data->info.delta_base_sha1 = data->delta_base_oid.hash;
247 else
248 strbuf_addstr(sb,
249 oid_to_hex(&data->delta_base_oid));
250 } else
251 die("unknown format element: %.*s", len, atom);
252 }
253
254 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
255 {
256 const char *end;
257
258 if (*start != '(')
259 return 0;
260 end = strchr(start + 1, ')');
261 if (!end)
262 die("format element '%s' does not end in ')'", start);
263
264 expand_atom(sb, start + 1, end - start - 1, data);
265
266 return end - start + 1;
267 }
268
269 static void batch_write(struct batch_options *opt, const void *data, int len)
270 {
271 if (opt->buffer_output) {
272 if (fwrite(data, 1, len, stdout) != len)
273 die_errno("unable to write to stdout");
274 } else
275 write_or_die(1, data, len);
276 }
277
278 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
279 {
280 const struct object_id *oid = &data->oid;
281
282 assert(data->info.typep);
283
284 if (data->type == OBJ_BLOB) {
285 if (opt->buffer_output)
286 fflush(stdout);
287 if (opt->cmdmode) {
288 char *contents;
289 unsigned long size;
290
291 if (!data->rest)
292 die("missing path for '%s'", oid_to_hex(oid));
293
294 if (opt->cmdmode == 'w') {
295 if (filter_object(data->rest, 0100644, oid,
296 &contents, &size))
297 die("could not convert '%s' %s",
298 oid_to_hex(oid), data->rest);
299 } else if (opt->cmdmode == 'c') {
300 enum object_type type;
301 if (!textconv_object(data->rest, 0100644, oid,
302 1, &contents, &size))
303 contents = read_sha1_file(oid->hash, &type,
304 &size);
305 if (!contents)
306 die("could not convert '%s' %s",
307 oid_to_hex(oid), data->rest);
308 } else
309 die("BUG: invalid cmdmode: %c", opt->cmdmode);
310 batch_write(opt, contents, size);
311 free(contents);
312 } else if (stream_blob_to_fd(1, oid, NULL, 0) < 0)
313 die("unable to stream %s to stdout", oid_to_hex(oid));
314 }
315 else {
316 enum object_type type;
317 unsigned long size;
318 void *contents;
319
320 contents = read_sha1_file(oid->hash, &type, &size);
321 if (!contents)
322 die("object %s disappeared", oid_to_hex(oid));
323 if (type != data->type)
324 die("object %s changed type!?", oid_to_hex(oid));
325 if (data->info.sizep && size != data->size)
326 die("object %s changed size!?", oid_to_hex(oid));
327
328 batch_write(opt, contents, size);
329 free(contents);
330 }
331 }
332
333 static void batch_object_write(const char *obj_name, struct batch_options *opt,
334 struct expand_data *data)
335 {
336 struct strbuf buf = STRBUF_INIT;
337
338 if (!data->skip_object_info &&
339 sha1_object_info_extended(data->oid.hash, &data->info, LOOKUP_REPLACE_OBJECT) < 0) {
340 printf("%s missing\n",
341 obj_name ? obj_name : oid_to_hex(&data->oid));
342 fflush(stdout);
343 return;
344 }
345
346 strbuf_expand(&buf, opt->format, expand_format, data);
347 strbuf_addch(&buf, '\n');
348 batch_write(opt, buf.buf, buf.len);
349 strbuf_release(&buf);
350
351 if (opt->print_contents) {
352 print_object_or_die(opt, data);
353 batch_write(opt, "\n", 1);
354 }
355 }
356
357 static void batch_one_object(const char *obj_name, struct batch_options *opt,
358 struct expand_data *data)
359 {
360 struct object_context ctx;
361 int flags = opt->follow_symlinks ? GET_SHA1_FOLLOW_SYMLINKS : 0;
362 enum follow_symlinks_result result;
363
364 result = get_sha1_with_context(obj_name, flags, data->oid.hash, &ctx);
365 if (result != FOUND) {
366 switch (result) {
367 case MISSING_OBJECT:
368 printf("%s missing\n", obj_name);
369 break;
370 case DANGLING_SYMLINK:
371 printf("dangling %"PRIuMAX"\n%s\n",
372 (uintmax_t)strlen(obj_name), obj_name);
373 break;
374 case SYMLINK_LOOP:
375 printf("loop %"PRIuMAX"\n%s\n",
376 (uintmax_t)strlen(obj_name), obj_name);
377 break;
378 case NOT_DIR:
379 printf("notdir %"PRIuMAX"\n%s\n",
380 (uintmax_t)strlen(obj_name), obj_name);
381 break;
382 default:
383 die("BUG: unknown get_sha1_with_context result %d\n",
384 result);
385 break;
386 }
387 fflush(stdout);
388 return;
389 }
390
391 if (ctx.mode == 0) {
392 printf("symlink %"PRIuMAX"\n%s\n",
393 (uintmax_t)ctx.symlink_path.len,
394 ctx.symlink_path.buf);
395 fflush(stdout);
396 return;
397 }
398
399 batch_object_write(obj_name, opt, data);
400 }
401
402 struct object_cb_data {
403 struct batch_options *opt;
404 struct expand_data *expand;
405 };
406
407 static int batch_object_cb(const struct object_id *oid, void *vdata)
408 {
409 struct object_cb_data *data = vdata;
410 oidcpy(&data->expand->oid, oid);
411 batch_object_write(NULL, data->opt, data->expand);
412 return 0;
413 }
414
415 static int batch_loose_object(const struct object_id *oid,
416 const char *path,
417 void *data)
418 {
419 oid_array_append(data, oid);
420 return 0;
421 }
422
423 static int batch_packed_object(const struct object_id *oid,
424 struct packed_git *pack,
425 uint32_t pos,
426 void *data)
427 {
428 oid_array_append(data, oid);
429 return 0;
430 }
431
432 static int batch_objects(struct batch_options *opt)
433 {
434 struct strbuf buf = STRBUF_INIT;
435 struct expand_data data;
436 int save_warning;
437 int retval = 0;
438
439 if (!opt->format)
440 opt->format = "%(objectname) %(objecttype) %(objectsize)";
441
442 /*
443 * Expand once with our special mark_query flag, which will prime the
444 * object_info to be handed to sha1_object_info_extended for each
445 * object.
446 */
447 memset(&data, 0, sizeof(data));
448 data.mark_query = 1;
449 strbuf_expand(&buf, opt->format, expand_format, &data);
450 data.mark_query = 0;
451 if (opt->cmdmode)
452 data.split_on_whitespace = 1;
453
454 if (opt->all_objects) {
455 struct object_info empty = OBJECT_INFO_INIT;
456 if (!memcmp(&data.info, &empty, sizeof(empty)))
457 data.skip_object_info = 1;
458 }
459
460 /*
461 * If we are printing out the object, then always fill in the type,
462 * since we will want to decide whether or not to stream.
463 */
464 if (opt->print_contents)
465 data.info.typep = &data.type;
466
467 if (opt->all_objects) {
468 struct oid_array sa = OID_ARRAY_INIT;
469 struct object_cb_data cb;
470
471 for_each_loose_object(batch_loose_object, &sa, 0);
472 for_each_packed_object(batch_packed_object, &sa, 0);
473
474 cb.opt = opt;
475 cb.expand = &data;
476 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
477
478 oid_array_clear(&sa);
479 return 0;
480 }
481
482 /*
483 * We are going to call get_sha1 on a potentially very large number of
484 * objects. In most large cases, these will be actual object sha1s. The
485 * cost to double-check that each one is not also a ref (just so we can
486 * warn) ends up dwarfing the actual cost of the object lookups
487 * themselves. We can work around it by just turning off the warning.
488 */
489 save_warning = warn_on_object_refname_ambiguity;
490 warn_on_object_refname_ambiguity = 0;
491
492 while (strbuf_getline(&buf, stdin) != EOF) {
493 if (data.split_on_whitespace) {
494 /*
495 * Split at first whitespace, tying off the beginning
496 * of the string and saving the remainder (or NULL) in
497 * data.rest.
498 */
499 char *p = strpbrk(buf.buf, " \t");
500 if (p) {
501 while (*p && strchr(" \t", *p))
502 *p++ = '\0';
503 }
504 data.rest = p;
505 }
506
507 batch_one_object(buf.buf, opt, &data);
508 }
509
510 strbuf_release(&buf);
511 warn_on_object_refname_ambiguity = save_warning;
512 return retval;
513 }
514
515 static const char * const cat_file_usage[] = {
516 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
517 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
518 NULL
519 };
520
521 static int git_cat_file_config(const char *var, const char *value, void *cb)
522 {
523 if (userdiff_config(var, value) < 0)
524 return -1;
525
526 return git_default_config(var, value, cb);
527 }
528
529 static int batch_option_callback(const struct option *opt,
530 const char *arg,
531 int unset)
532 {
533 struct batch_options *bo = opt->value;
534
535 if (bo->enabled) {
536 return 1;
537 }
538
539 bo->enabled = 1;
540 bo->print_contents = !strcmp(opt->long_name, "batch");
541 bo->format = arg;
542
543 return 0;
544 }
545
546 int cmd_cat_file(int argc, const char **argv, const char *prefix)
547 {
548 int opt = 0;
549 const char *exp_type = NULL, *obj_name = NULL;
550 struct batch_options batch = {0};
551 int unknown_type = 0;
552
553 const struct option options[] = {
554 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
555 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
556 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
557 OPT_CMDMODE('e', NULL, &opt,
558 N_("exit with zero when there's no error"), 'e'),
559 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
560 OPT_CMDMODE(0, "textconv", &opt,
561 N_("for blob objects, run textconv on object's content"), 'c'),
562 OPT_CMDMODE(0, "filters", &opt,
563 N_("for blob objects, run filters on object's content"), 'w'),
564 OPT_STRING(0, "path", &force_path, N_("blob"),
565 N_("use a specific path for --textconv/--filters")),
566 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
567 N_("allow -s and -t to work with broken/corrupt objects")),
568 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
569 { OPTION_CALLBACK, 0, "batch", &batch, "format",
570 N_("show info and content of objects fed from the standard input"),
571 PARSE_OPT_OPTARG, batch_option_callback },
572 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
573 N_("show info about objects fed from the standard input"),
574 PARSE_OPT_OPTARG, batch_option_callback },
575 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
576 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
577 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
578 N_("show all objects with --batch or --batch-check")),
579 OPT_END()
580 };
581
582 git_config(git_cat_file_config, NULL);
583
584 batch.buffer_output = -1;
585 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
586
587 if (opt) {
588 if (batch.enabled && (opt == 'c' || opt == 'w'))
589 batch.cmdmode = opt;
590 else if (argc == 1)
591 obj_name = argv[0];
592 else
593 usage_with_options(cat_file_usage, options);
594 }
595 if (!opt && !batch.enabled) {
596 if (argc == 2) {
597 exp_type = argv[0];
598 obj_name = argv[1];
599 } else
600 usage_with_options(cat_file_usage, options);
601 }
602 if (batch.enabled) {
603 if (batch.cmdmode != opt || argc)
604 usage_with_options(cat_file_usage, options);
605 if (batch.cmdmode && batch.all_objects)
606 die("--batch-all-objects cannot be combined with "
607 "--textconv nor with --filters");
608 }
609
610 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
611 usage_with_options(cat_file_usage, options);
612 }
613
614 if (force_path && opt != 'c' && opt != 'w') {
615 error("--path=<path> needs --textconv or --filters");
616 usage_with_options(cat_file_usage, options);
617 }
618
619 if (force_path && batch.enabled) {
620 error("--path=<path> incompatible with --batch");
621 usage_with_options(cat_file_usage, options);
622 }
623
624 if (batch.buffer_output < 0)
625 batch.buffer_output = batch.all_objects;
626
627 if (batch.enabled)
628 return batch_objects(&batch);
629
630 if (unknown_type && opt != 't' && opt != 's')
631 die("git cat-file --allow-unknown-type: use with -s or -t");
632 return cat_one_file(opt, exp_type, obj_name, unknown_type);
633 }