Adjust for 2.19.x series
[git.git] / builtin / cat-file.c
blob0520cecc9a1d963841b384de6173d752e6d78fb7
1 /*
2 * GIT - The information manager from hell
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"
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;
29 static const char *force_path;
31 static int filter_object(const char *path, unsigned mode,
32 const struct object_id *oid,
33 char **buf, unsigned long *size)
35 enum object_type type;
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);
50 return 0;
53 static int stream_blob(const struct object_id *oid)
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;
60 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
61 int unknown_type)
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;
73 if (unknown_type)
74 flags |= OBJECT_INFO_ALLOW_UNKNOWN_TYPE;
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);
80 if (!path)
81 path = obj_context.path;
82 if (obj_context.mode == S_IFINVALID)
83 obj_context.mode = 0100644;
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;
96 break;
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;
105 case 'e':
106 return !has_object_file(&oid);
108 case 'w':
109 if (!path)
110 die("git cat-file --filters %s: <object> must be "
111 "<sha1:path>", obj_name);
113 if (filter_object(path, obj_context.mode,
114 &oid, &buf, &size))
115 return -1;
116 break;
118 case 'c':
119 if (!path)
120 die("git cat-file --textconv %s: <object> must be <sha1:path>",
121 obj_name);
123 if (textconv_object(path, obj_context.mode, &oid, 1, &buf, &size))
124 break;
125 /* else fallthrough */
127 case 'p':
128 type = oid_object_info(the_repository, &oid, NULL);
129 if (type < 0)
130 die("Not a valid object name %s", obj_name);
132 /* custom pretty-print here */
133 if (type == OBJ_TREE) {
134 const char *ls_args[3] = { NULL };
135 ls_args[0] = "ls-tree";
136 ls_args[1] = obj_name;
137 return cmd_ls_tree(2, ls_args, NULL);
140 if (type == OBJ_BLOB)
141 return stream_blob(&oid);
142 buf = read_object_file(&oid, &type, &size);
143 if (!buf)
144 die("Cannot read object %s", obj_name);
146 /* otherwise just spit out the data */
147 break;
149 case 0:
150 if (type_from_string(exp_type) == OBJ_BLOB) {
151 struct object_id blob_oid;
152 if (oid_object_info(the_repository, &oid, NULL) == OBJ_TAG) {
153 char *buffer = read_object_file(&oid, &type,
154 &size);
155 const char *target;
156 if (!skip_prefix(buffer, "object ", &target) ||
157 get_oid_hex(target, &blob_oid))
158 die("%s not a valid tag", oid_to_hex(&oid));
159 free(buffer);
160 } else
161 oidcpy(&blob_oid, &oid);
163 if (oid_object_info(the_repository, &blob_oid, NULL) == OBJ_BLOB)
164 return stream_blob(&blob_oid);
166 * we attempted to dereference a tag to a blob
167 * and failed; there may be new dereference
168 * mechanisms this code is not aware of.
169 * fall-back to the usual case.
172 buf = read_object_with_reference(&oid, exp_type, &size, NULL);
173 break;
175 default:
176 die("git cat-file: unknown option: %s", exp_type);
179 if (!buf)
180 die("git cat-file %s: bad file", obj_name);
182 write_or_die(1, buf, size);
183 free(buf);
184 free(obj_context.path);
185 return 0;
188 struct expand_data {
189 struct object_id oid;
190 enum object_type type;
191 unsigned long size;
192 off_t disk_size;
193 const char *rest;
194 struct object_id delta_base_oid;
197 * If mark_query is true, we do not expand anything, but rather
198 * just mark the object_info with items we wish to query.
200 int mark_query;
203 * Whether to split the input on whitespace before feeding it to
204 * get_sha1; this is decided during the mark_query phase based on
205 * whether we have a %(rest) token in our format.
207 int split_on_whitespace;
210 * After a mark_query run, this object_info is set up to be
211 * passed to sha1_object_info_extended. It will point to the data
212 * elements above, so you can retrieve the response from there.
214 struct object_info info;
217 * This flag will be true if the requested batch format and options
218 * don't require us to call sha1_object_info, which can then be
219 * optimized out.
221 unsigned skip_object_info : 1;
224 static int is_atom(const char *atom, const char *s, int slen)
226 int alen = strlen(atom);
227 return alen == slen && !memcmp(atom, s, alen);
230 static void expand_atom(struct strbuf *sb, const char *atom, int len,
231 void *vdata)
233 struct expand_data *data = vdata;
235 if (is_atom("objectname", atom, len)) {
236 if (!data->mark_query)
237 strbuf_addstr(sb, oid_to_hex(&data->oid));
238 } else if (is_atom("objecttype", atom, len)) {
239 if (data->mark_query)
240 data->info.typep = &data->type;
241 else
242 strbuf_addstr(sb, type_name(data->type));
243 } else if (is_atom("objectsize", atom, len)) {
244 if (data->mark_query)
245 data->info.sizep = &data->size;
246 else
247 strbuf_addf(sb, "%lu", data->size);
248 } else if (is_atom("objectsize:disk", atom, len)) {
249 if (data->mark_query)
250 data->info.disk_sizep = &data->disk_size;
251 else
252 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
253 } else if (is_atom("rest", atom, len)) {
254 if (data->mark_query)
255 data->split_on_whitespace = 1;
256 else if (data->rest)
257 strbuf_addstr(sb, data->rest);
258 } else if (is_atom("deltabase", atom, len)) {
259 if (data->mark_query)
260 data->info.delta_base_sha1 = data->delta_base_oid.hash;
261 else
262 strbuf_addstr(sb,
263 oid_to_hex(&data->delta_base_oid));
264 } else
265 die("unknown format element: %.*s", len, atom);
268 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
270 const char *end;
272 if (*start != '(')
273 return 0;
274 end = strchr(start + 1, ')');
275 if (!end)
276 die("format element '%s' does not end in ')'", start);
278 expand_atom(sb, start + 1, end - start - 1, data);
280 return end - start + 1;
283 static void batch_write(struct batch_options *opt, const void *data, int len)
285 if (opt->buffer_output) {
286 if (fwrite(data, 1, len, stdout) != len)
287 die_errno("unable to write to stdout");
288 } else
289 write_or_die(1, data, len);
292 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
294 const struct object_id *oid = &data->oid;
296 assert(data->info.typep);
298 if (data->type == OBJ_BLOB) {
299 if (opt->buffer_output)
300 fflush(stdout);
301 if (opt->cmdmode) {
302 char *contents;
303 unsigned long size;
305 if (!data->rest)
306 die("missing path for '%s'", oid_to_hex(oid));
308 if (opt->cmdmode == 'w') {
309 if (filter_object(data->rest, 0100644, oid,
310 &contents, &size))
311 die("could not convert '%s' %s",
312 oid_to_hex(oid), data->rest);
313 } else if (opt->cmdmode == 'c') {
314 enum object_type type;
315 if (!textconv_object(data->rest, 0100644, oid,
316 1, &contents, &size))
317 contents = read_object_file(oid,
318 &type,
319 &size);
320 if (!contents)
321 die("could not convert '%s' %s",
322 oid_to_hex(oid), data->rest);
323 } else
324 BUG("invalid cmdmode: %c", opt->cmdmode);
325 batch_write(opt, contents, size);
326 free(contents);
327 } else {
328 stream_blob(oid);
331 else {
332 enum object_type type;
333 unsigned long size;
334 void *contents;
336 contents = read_object_file(oid, &type, &size);
337 if (!contents)
338 die("object %s disappeared", oid_to_hex(oid));
339 if (type != data->type)
340 die("object %s changed type!?", oid_to_hex(oid));
341 if (data->info.sizep && size != data->size)
342 die("object %s changed size!?", oid_to_hex(oid));
344 batch_write(opt, contents, size);
345 free(contents);
349 static void batch_object_write(const char *obj_name,
350 struct strbuf *scratch,
351 struct batch_options *opt,
352 struct expand_data *data)
354 if (!data->skip_object_info &&
355 oid_object_info_extended(the_repository, &data->oid, &data->info,
356 OBJECT_INFO_LOOKUP_REPLACE) < 0) {
357 printf("%s missing\n",
358 obj_name ? obj_name : oid_to_hex(&data->oid));
359 fflush(stdout);
360 return;
363 strbuf_reset(scratch);
364 strbuf_expand(scratch, opt->format, expand_format, data);
365 strbuf_addch(scratch, '\n');
366 batch_write(opt, scratch->buf, scratch->len);
368 if (opt->print_contents) {
369 print_object_or_die(opt, data);
370 batch_write(opt, "\n", 1);
374 static void batch_one_object(const char *obj_name,
375 struct strbuf *scratch,
376 struct batch_options *opt,
377 struct expand_data *data)
379 struct object_context ctx;
380 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
381 enum follow_symlinks_result result;
383 result = get_oid_with_context(obj_name, flags, &data->oid, &ctx);
384 if (result != FOUND) {
385 switch (result) {
386 case MISSING_OBJECT:
387 printf("%s missing\n", obj_name);
388 break;
389 case DANGLING_SYMLINK:
390 printf("dangling %"PRIuMAX"\n%s\n",
391 (uintmax_t)strlen(obj_name), obj_name);
392 break;
393 case SYMLINK_LOOP:
394 printf("loop %"PRIuMAX"\n%s\n",
395 (uintmax_t)strlen(obj_name), obj_name);
396 break;
397 case NOT_DIR:
398 printf("notdir %"PRIuMAX"\n%s\n",
399 (uintmax_t)strlen(obj_name), obj_name);
400 break;
401 default:
402 BUG("unknown get_sha1_with_context result %d\n",
403 result);
404 break;
406 fflush(stdout);
407 return;
410 if (ctx.mode == 0) {
411 printf("symlink %"PRIuMAX"\n%s\n",
412 (uintmax_t)ctx.symlink_path.len,
413 ctx.symlink_path.buf);
414 fflush(stdout);
415 return;
418 batch_object_write(obj_name, scratch, opt, data);
421 struct object_cb_data {
422 struct batch_options *opt;
423 struct expand_data *expand;
424 struct oidset *seen;
425 struct strbuf *scratch;
428 static int batch_object_cb(const struct object_id *oid, void *vdata)
430 struct object_cb_data *data = vdata;
431 oidcpy(&data->expand->oid, oid);
432 batch_object_write(NULL, data->scratch, data->opt, data->expand);
433 return 0;
436 static int collect_loose_object(const struct object_id *oid,
437 const char *path,
438 void *data)
440 oid_array_append(data, oid);
441 return 0;
444 static int collect_packed_object(const struct object_id *oid,
445 struct packed_git *pack,
446 uint32_t pos,
447 void *data)
449 oid_array_append(data, oid);
450 return 0;
453 static int batch_unordered_object(const struct object_id *oid, void *vdata)
455 struct object_cb_data *data = vdata;
457 if (oidset_insert(data->seen, oid))
458 return 0;
460 return batch_object_cb(oid, data);
463 static int batch_unordered_loose(const struct object_id *oid,
464 const char *path,
465 void *data)
467 return batch_unordered_object(oid, data);
470 static int batch_unordered_packed(const struct object_id *oid,
471 struct packed_git *pack,
472 uint32_t pos,
473 void *data)
475 return batch_unordered_object(oid, data);
478 static int batch_objects(struct batch_options *opt)
480 struct strbuf input = STRBUF_INIT;
481 struct strbuf output = STRBUF_INIT;
482 struct expand_data data;
483 int save_warning;
484 int retval = 0;
486 if (!opt->format)
487 opt->format = "%(objectname) %(objecttype) %(objectsize)";
490 * Expand once with our special mark_query flag, which will prime the
491 * object_info to be handed to sha1_object_info_extended for each
492 * object.
494 memset(&data, 0, sizeof(data));
495 data.mark_query = 1;
496 strbuf_expand(&output, opt->format, expand_format, &data);
497 data.mark_query = 0;
498 strbuf_release(&output);
499 if (opt->cmdmode)
500 data.split_on_whitespace = 1;
502 if (opt->all_objects) {
503 struct object_info empty = OBJECT_INFO_INIT;
504 if (!memcmp(&data.info, &empty, sizeof(empty)))
505 data.skip_object_info = 1;
509 * If we are printing out the object, then always fill in the type,
510 * since we will want to decide whether or not to stream.
512 if (opt->print_contents)
513 data.info.typep = &data.type;
515 if (opt->all_objects) {
516 struct object_cb_data cb;
518 if (repository_format_partial_clone)
519 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
521 cb.opt = opt;
522 cb.expand = &data;
523 cb.scratch = &output;
525 if (opt->unordered) {
526 struct oidset seen = OIDSET_INIT;
528 cb.seen = &seen;
530 for_each_loose_object(batch_unordered_loose, &cb, 0);
531 for_each_packed_object(batch_unordered_packed, &cb,
532 FOR_EACH_OBJECT_PACK_ORDER);
534 oidset_clear(&seen);
535 } else {
536 struct oid_array sa = OID_ARRAY_INIT;
538 for_each_loose_object(collect_loose_object, &sa, 0);
539 for_each_packed_object(collect_packed_object, &sa, 0);
541 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
543 oid_array_clear(&sa);
546 strbuf_release(&output);
547 return 0;
551 * We are going to call get_sha1 on a potentially very large number of
552 * objects. In most large cases, these will be actual object sha1s. The
553 * cost to double-check that each one is not also a ref (just so we can
554 * warn) ends up dwarfing the actual cost of the object lookups
555 * themselves. We can work around it by just turning off the warning.
557 save_warning = warn_on_object_refname_ambiguity;
558 warn_on_object_refname_ambiguity = 0;
560 while (strbuf_getline(&input, stdin) != EOF) {
561 if (data.split_on_whitespace) {
563 * Split at first whitespace, tying off the beginning
564 * of the string and saving the remainder (or NULL) in
565 * data.rest.
567 char *p = strpbrk(input.buf, " \t");
568 if (p) {
569 while (*p && strchr(" \t", *p))
570 *p++ = '\0';
572 data.rest = p;
575 batch_one_object(input.buf, &output, opt, &data);
578 strbuf_release(&input);
579 strbuf_release(&output);
580 warn_on_object_refname_ambiguity = save_warning;
581 return retval;
584 static const char * const cat_file_usage[] = {
585 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
586 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
587 NULL
590 static int git_cat_file_config(const char *var, const char *value, void *cb)
592 if (userdiff_config(var, value) < 0)
593 return -1;
595 return git_default_config(var, value, cb);
598 static int batch_option_callback(const struct option *opt,
599 const char *arg,
600 int unset)
602 struct batch_options *bo = opt->value;
604 if (bo->enabled) {
605 return 1;
608 bo->enabled = 1;
609 bo->print_contents = !strcmp(opt->long_name, "batch");
610 bo->format = arg;
612 return 0;
615 int cmd_cat_file(int argc, const char **argv, const char *prefix)
617 int opt = 0;
618 const char *exp_type = NULL, *obj_name = NULL;
619 struct batch_options batch = {0};
620 int unknown_type = 0;
622 const struct option options[] = {
623 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
624 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
625 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
626 OPT_CMDMODE('e', NULL, &opt,
627 N_("exit with zero when there's no error"), 'e'),
628 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
629 OPT_CMDMODE(0, "textconv", &opt,
630 N_("for blob objects, run textconv on object's content"), 'c'),
631 OPT_CMDMODE(0, "filters", &opt,
632 N_("for blob objects, run filters on object's content"), 'w'),
633 OPT_STRING(0, "path", &force_path, N_("blob"),
634 N_("use a specific path for --textconv/--filters")),
635 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
636 N_("allow -s and -t to work with broken/corrupt objects")),
637 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
638 { OPTION_CALLBACK, 0, "batch", &batch, "format",
639 N_("show info and content of objects fed from the standard input"),
640 PARSE_OPT_OPTARG, batch_option_callback },
641 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
642 N_("show info about objects fed from the standard input"),
643 PARSE_OPT_OPTARG, batch_option_callback },
644 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
645 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
646 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
647 N_("show all objects with --batch or --batch-check")),
648 OPT_BOOL(0, "unordered", &batch.unordered,
649 N_("do not order --batch-all-objects output")),
650 OPT_END()
653 git_config(git_cat_file_config, NULL);
655 batch.buffer_output = -1;
656 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
658 if (opt) {
659 if (batch.enabled && (opt == 'c' || opt == 'w'))
660 batch.cmdmode = opt;
661 else if (argc == 1)
662 obj_name = argv[0];
663 else
664 usage_with_options(cat_file_usage, options);
666 if (!opt && !batch.enabled) {
667 if (argc == 2) {
668 exp_type = argv[0];
669 obj_name = argv[1];
670 } else
671 usage_with_options(cat_file_usage, options);
673 if (batch.enabled) {
674 if (batch.cmdmode != opt || argc)
675 usage_with_options(cat_file_usage, options);
676 if (batch.cmdmode && batch.all_objects)
677 die("--batch-all-objects cannot be combined with "
678 "--textconv nor with --filters");
681 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
682 usage_with_options(cat_file_usage, options);
685 if (force_path && opt != 'c' && opt != 'w') {
686 error("--path=<path> needs --textconv or --filters");
687 usage_with_options(cat_file_usage, options);
690 if (force_path && batch.enabled) {
691 error("--path=<path> incompatible with --batch");
692 usage_with_options(cat_file_usage, options);
695 if (batch.buffer_output < 0)
696 batch.buffer_output = batch.all_objects;
698 if (batch.enabled)
699 return batch_objects(&batch);
701 if (unknown_type && opt != 't' && opt != 's')
702 die("git cat-file --allow-unknown-type: use with -s or -t");
703 return cat_one_file(opt, exp_type, obj_name, unknown_type);