2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
10 #include "parse-options.h"
12 #include "streaming.h"
13 #include "tree-walk.h"
14 #include "sha1-array.h"
16 #include "object-store.h"
18 struct batch_options
{
24 int cmdmode
; /* may be 'w' or 'c' for --filters or --textconv */
28 static const char *force_path
;
30 static int filter_object(const char *path
, unsigned mode
,
31 const struct object_id
*oid
,
32 char **buf
, unsigned long *size
)
34 enum object_type type
;
36 *buf
= read_object_file(oid
, &type
, size
);
38 return error(_("cannot read object %s '%s'"),
39 oid_to_hex(oid
), path
);
40 if ((type
== OBJ_BLOB
) && S_ISREG(mode
)) {
41 struct strbuf strbuf
= STRBUF_INIT
;
42 if (convert_to_working_tree(path
, *buf
, *size
, &strbuf
)) {
45 *buf
= strbuf_detach(&strbuf
, NULL
);
52 static int cat_one_file(int opt
, const char *exp_type
, const char *obj_name
,
56 enum object_type type
;
59 struct object_context obj_context
;
60 struct object_info oi
= OBJECT_INFO_INIT
;
61 struct strbuf sb
= STRBUF_INIT
;
62 unsigned flags
= OBJECT_INFO_LOOKUP_REPLACE
;
63 const char *path
= force_path
;
66 flags
|= OBJECT_INFO_ALLOW_UNKNOWN_TYPE
;
68 if (get_oid_with_context(obj_name
, GET_OID_RECORD_PATH
,
70 die("Not a valid object name %s", obj_name
);
73 path
= obj_context
.path
;
74 if (obj_context
.mode
== S_IFINVALID
)
75 obj_context
.mode
= 0100644;
81 if (oid_object_info_extended(the_repository
, &oid
, &oi
, flags
) < 0)
82 die("git cat-file: could not get object info");
84 printf("%s\n", sb
.buf
);
92 if (oid_object_info_extended(the_repository
, &oid
, &oi
, flags
) < 0)
93 die("git cat-file: could not get object info");
94 printf("%lu\n", size
);
98 return !has_object_file(&oid
);
102 die("git cat-file --filters %s: <object> must be "
103 "<sha1:path>", obj_name
);
105 if (filter_object(path
, obj_context
.mode
,
112 die("git cat-file --textconv %s: <object> must be <sha1:path>",
115 if (textconv_object(path
, obj_context
.mode
, &oid
, 1, &buf
, &size
))
117 /* else fallthrough */
120 type
= oid_object_info(the_repository
, &oid
, NULL
);
122 die("Not a valid object name %s", obj_name
);
124 /* custom pretty-print here */
125 if (type
== OBJ_TREE
) {
126 const char *ls_args
[3] = { NULL
};
127 ls_args
[0] = "ls-tree";
128 ls_args
[1] = obj_name
;
129 return cmd_ls_tree(2, ls_args
, NULL
);
132 if (type
== OBJ_BLOB
)
133 return stream_blob_to_fd(1, &oid
, NULL
, 0);
134 buf
= read_object_file(&oid
, &type
, &size
);
136 die("Cannot read object %s", obj_name
);
138 /* otherwise just spit out the data */
142 if (type_from_string(exp_type
) == OBJ_BLOB
) {
143 struct object_id blob_oid
;
144 if (oid_object_info(the_repository
, &oid
, NULL
) == OBJ_TAG
) {
145 char *buffer
= read_object_file(&oid
, &type
,
148 if (!skip_prefix(buffer
, "object ", &target
) ||
149 get_oid_hex(target
, &blob_oid
))
150 die("%s not a valid tag", oid_to_hex(&oid
));
153 oidcpy(&blob_oid
, &oid
);
155 if (oid_object_info(the_repository
, &blob_oid
, NULL
) == OBJ_BLOB
)
156 return stream_blob_to_fd(1, &blob_oid
, NULL
, 0);
158 * we attempted to dereference a tag to a blob
159 * and failed; there may be new dereference
160 * mechanisms this code is not aware of.
161 * fall-back to the usual case.
164 buf
= read_object_with_reference(&oid
, exp_type
, &size
, NULL
);
168 die("git cat-file: unknown option: %s", exp_type
);
172 die("git cat-file %s: bad file", obj_name
);
174 write_or_die(1, buf
, size
);
176 free(obj_context
.path
);
181 struct object_id oid
;
182 enum object_type type
;
186 struct object_id delta_base_oid
;
189 * If mark_query is true, we do not expand anything, but rather
190 * just mark the object_info with items we wish to query.
195 * Whether to split the input on whitespace before feeding it to
196 * get_sha1; this is decided during the mark_query phase based on
197 * whether we have a %(rest) token in our format.
199 int split_on_whitespace
;
202 * After a mark_query run, this object_info is set up to be
203 * passed to sha1_object_info_extended. It will point to the data
204 * elements above, so you can retrieve the response from there.
206 struct object_info info
;
209 * This flag will be true if the requested batch format and options
210 * don't require us to call sha1_object_info, which can then be
213 unsigned skip_object_info
: 1;
216 static int is_atom(const char *atom
, const char *s
, int slen
)
218 int alen
= strlen(atom
);
219 return alen
== slen
&& !memcmp(atom
, s
, alen
);
222 static void expand_atom(struct strbuf
*sb
, const char *atom
, int len
,
225 struct expand_data
*data
= vdata
;
227 if (is_atom("objectname", atom
, len
)) {
228 if (!data
->mark_query
)
229 strbuf_addstr(sb
, oid_to_hex(&data
->oid
));
230 } else if (is_atom("objecttype", atom
, len
)) {
231 if (data
->mark_query
)
232 data
->info
.typep
= &data
->type
;
234 strbuf_addstr(sb
, type_name(data
->type
));
235 } else if (is_atom("objectsize", atom
, len
)) {
236 if (data
->mark_query
)
237 data
->info
.sizep
= &data
->size
;
239 strbuf_addf(sb
, "%lu", data
->size
);
240 } else if (is_atom("objectsize:disk", atom
, len
)) {
241 if (data
->mark_query
)
242 data
->info
.disk_sizep
= &data
->disk_size
;
244 strbuf_addf(sb
, "%"PRIuMAX
, (uintmax_t)data
->disk_size
);
245 } else if (is_atom("rest", atom
, len
)) {
246 if (data
->mark_query
)
247 data
->split_on_whitespace
= 1;
249 strbuf_addstr(sb
, data
->rest
);
250 } else if (is_atom("deltabase", atom
, len
)) {
251 if (data
->mark_query
)
252 data
->info
.delta_base_sha1
= data
->delta_base_oid
.hash
;
255 oid_to_hex(&data
->delta_base_oid
));
257 die("unknown format element: %.*s", len
, atom
);
260 static size_t expand_format(struct strbuf
*sb
, const char *start
, void *data
)
266 end
= strchr(start
+ 1, ')');
268 die("format element '%s' does not end in ')'", start
);
270 expand_atom(sb
, start
+ 1, end
- start
- 1, data
);
272 return end
- start
+ 1;
275 static void batch_write(struct batch_options
*opt
, const void *data
, int len
)
277 if (opt
->buffer_output
) {
278 if (fwrite(data
, 1, len
, stdout
) != len
)
279 die_errno("unable to write to stdout");
281 write_or_die(1, data
, len
);
284 static void print_object_or_die(struct batch_options
*opt
, struct expand_data
*data
)
286 const struct object_id
*oid
= &data
->oid
;
288 assert(data
->info
.typep
);
290 if (data
->type
== OBJ_BLOB
) {
291 if (opt
->buffer_output
)
298 die("missing path for '%s'", oid_to_hex(oid
));
300 if (opt
->cmdmode
== 'w') {
301 if (filter_object(data
->rest
, 0100644, oid
,
303 die("could not convert '%s' %s",
304 oid_to_hex(oid
), data
->rest
);
305 } else if (opt
->cmdmode
== 'c') {
306 enum object_type type
;
307 if (!textconv_object(data
->rest
, 0100644, oid
,
308 1, &contents
, &size
))
309 contents
= read_object_file(oid
,
313 die("could not convert '%s' %s",
314 oid_to_hex(oid
), data
->rest
);
316 BUG("invalid cmdmode: %c", opt
->cmdmode
);
317 batch_write(opt
, contents
, size
);
319 } else if (stream_blob_to_fd(1, oid
, NULL
, 0) < 0)
320 die("unable to stream %s to stdout", oid_to_hex(oid
));
323 enum object_type type
;
327 contents
= read_object_file(oid
, &type
, &size
);
329 die("object %s disappeared", oid_to_hex(oid
));
330 if (type
!= data
->type
)
331 die("object %s changed type!?", oid_to_hex(oid
));
332 if (data
->info
.sizep
&& size
!= data
->size
)
333 die("object %s changed size!?", oid_to_hex(oid
));
335 batch_write(opt
, contents
, size
);
340 static void batch_object_write(const char *obj_name
, struct batch_options
*opt
,
341 struct expand_data
*data
)
343 struct strbuf buf
= STRBUF_INIT
;
345 if (!data
->skip_object_info
&&
346 oid_object_info_extended(the_repository
, &data
->oid
, &data
->info
,
347 OBJECT_INFO_LOOKUP_REPLACE
) < 0) {
348 printf("%s missing\n",
349 obj_name
? obj_name
: oid_to_hex(&data
->oid
));
354 strbuf_expand(&buf
, opt
->format
, expand_format
, data
);
355 strbuf_addch(&buf
, '\n');
356 batch_write(opt
, buf
.buf
, buf
.len
);
357 strbuf_release(&buf
);
359 if (opt
->print_contents
) {
360 print_object_or_die(opt
, data
);
361 batch_write(opt
, "\n", 1);
365 static void batch_one_object(const char *obj_name
, struct batch_options
*opt
,
366 struct expand_data
*data
)
368 struct object_context ctx
;
369 int flags
= opt
->follow_symlinks
? GET_OID_FOLLOW_SYMLINKS
: 0;
370 enum follow_symlinks_result result
;
372 result
= get_oid_with_context(obj_name
, flags
, &data
->oid
, &ctx
);
373 if (result
!= FOUND
) {
376 printf("%s missing\n", obj_name
);
378 case DANGLING_SYMLINK
:
379 printf("dangling %"PRIuMAX
"\n%s\n",
380 (uintmax_t)strlen(obj_name
), obj_name
);
383 printf("loop %"PRIuMAX
"\n%s\n",
384 (uintmax_t)strlen(obj_name
), obj_name
);
387 printf("notdir %"PRIuMAX
"\n%s\n",
388 (uintmax_t)strlen(obj_name
), obj_name
);
391 BUG("unknown get_sha1_with_context result %d\n",
400 printf("symlink %"PRIuMAX
"\n%s\n",
401 (uintmax_t)ctx
.symlink_path
.len
,
402 ctx
.symlink_path
.buf
);
407 batch_object_write(obj_name
, opt
, data
);
410 struct object_cb_data
{
411 struct batch_options
*opt
;
412 struct expand_data
*expand
;
415 static int batch_object_cb(const struct object_id
*oid
, void *vdata
)
417 struct object_cb_data
*data
= vdata
;
418 oidcpy(&data
->expand
->oid
, oid
);
419 batch_object_write(NULL
, data
->opt
, data
->expand
);
423 static int batch_loose_object(const struct object_id
*oid
,
427 oid_array_append(data
, oid
);
431 static int batch_packed_object(const struct object_id
*oid
,
432 struct packed_git
*pack
,
436 oid_array_append(data
, oid
);
440 static int batch_objects(struct batch_options
*opt
)
442 struct strbuf buf
= STRBUF_INIT
;
443 struct expand_data data
;
448 opt
->format
= "%(objectname) %(objecttype) %(objectsize)";
451 * Expand once with our special mark_query flag, which will prime the
452 * object_info to be handed to sha1_object_info_extended for each
455 memset(&data
, 0, sizeof(data
));
457 strbuf_expand(&buf
, opt
->format
, expand_format
, &data
);
460 data
.split_on_whitespace
= 1;
462 if (opt
->all_objects
) {
463 struct object_info empty
= OBJECT_INFO_INIT
;
464 if (!memcmp(&data
.info
, &empty
, sizeof(empty
)))
465 data
.skip_object_info
= 1;
469 * If we are printing out the object, then always fill in the type,
470 * since we will want to decide whether or not to stream.
472 if (opt
->print_contents
)
473 data
.info
.typep
= &data
.type
;
475 if (opt
->all_objects
) {
476 struct oid_array sa
= OID_ARRAY_INIT
;
477 struct object_cb_data cb
;
479 for_each_loose_object(batch_loose_object
, &sa
, 0);
480 for_each_packed_object(batch_packed_object
, &sa
, 0);
481 if (repository_format_partial_clone
)
482 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
486 oid_array_for_each_unique(&sa
, batch_object_cb
, &cb
);
488 oid_array_clear(&sa
);
493 * We are going to call get_sha1 on a potentially very large number of
494 * objects. In most large cases, these will be actual object sha1s. The
495 * cost to double-check that each one is not also a ref (just so we can
496 * warn) ends up dwarfing the actual cost of the object lookups
497 * themselves. We can work around it by just turning off the warning.
499 save_warning
= warn_on_object_refname_ambiguity
;
500 warn_on_object_refname_ambiguity
= 0;
502 while (strbuf_getline(&buf
, stdin
) != EOF
) {
503 if (data
.split_on_whitespace
) {
505 * Split at first whitespace, tying off the beginning
506 * of the string and saving the remainder (or NULL) in
509 char *p
= strpbrk(buf
.buf
, " \t");
511 while (*p
&& strchr(" \t", *p
))
517 batch_one_object(buf
.buf
, opt
, &data
);
520 strbuf_release(&buf
);
521 warn_on_object_refname_ambiguity
= save_warning
;
525 static const char * const cat_file_usage
[] = {
526 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
527 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
531 static int git_cat_file_config(const char *var
, const char *value
, void *cb
)
533 if (userdiff_config(var
, value
) < 0)
536 return git_default_config(var
, value
, cb
);
539 static int batch_option_callback(const struct option
*opt
,
543 struct batch_options
*bo
= opt
->value
;
550 bo
->print_contents
= !strcmp(opt
->long_name
, "batch");
556 int cmd_cat_file(int argc
, const char **argv
, const char *prefix
)
559 const char *exp_type
= NULL
, *obj_name
= NULL
;
560 struct batch_options batch
= {0};
561 int unknown_type
= 0;
563 const struct option options
[] = {
564 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
565 OPT_CMDMODE('t', NULL
, &opt
, N_("show object type"), 't'),
566 OPT_CMDMODE('s', NULL
, &opt
, N_("show object size"), 's'),
567 OPT_CMDMODE('e', NULL
, &opt
,
568 N_("exit with zero when there's no error"), 'e'),
569 OPT_CMDMODE('p', NULL
, &opt
, N_("pretty-print object's content"), 'p'),
570 OPT_CMDMODE(0, "textconv", &opt
,
571 N_("for blob objects, run textconv on object's content"), 'c'),
572 OPT_CMDMODE(0, "filters", &opt
,
573 N_("for blob objects, run filters on object's content"), 'w'),
574 OPT_STRING(0, "path", &force_path
, N_("blob"),
575 N_("use a specific path for --textconv/--filters")),
576 OPT_BOOL(0, "allow-unknown-type", &unknown_type
,
577 N_("allow -s and -t to work with broken/corrupt objects")),
578 OPT_BOOL(0, "buffer", &batch
.buffer_output
, N_("buffer --batch output")),
579 { OPTION_CALLBACK
, 0, "batch", &batch
, "format",
580 N_("show info and content of objects fed from the standard input"),
581 PARSE_OPT_OPTARG
, batch_option_callback
},
582 { OPTION_CALLBACK
, 0, "batch-check", &batch
, "format",
583 N_("show info about objects fed from the standard input"),
584 PARSE_OPT_OPTARG
, batch_option_callback
},
585 OPT_BOOL(0, "follow-symlinks", &batch
.follow_symlinks
,
586 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
587 OPT_BOOL(0, "batch-all-objects", &batch
.all_objects
,
588 N_("show all objects with --batch or --batch-check")),
592 git_config(git_cat_file_config
, NULL
);
594 batch
.buffer_output
= -1;
595 argc
= parse_options(argc
, argv
, prefix
, options
, cat_file_usage
, 0);
598 if (batch
.enabled
&& (opt
== 'c' || opt
== 'w'))
603 usage_with_options(cat_file_usage
, options
);
605 if (!opt
&& !batch
.enabled
) {
610 usage_with_options(cat_file_usage
, options
);
613 if (batch
.cmdmode
!= opt
|| argc
)
614 usage_with_options(cat_file_usage
, options
);
615 if (batch
.cmdmode
&& batch
.all_objects
)
616 die("--batch-all-objects cannot be combined with "
617 "--textconv nor with --filters");
620 if ((batch
.follow_symlinks
|| batch
.all_objects
) && !batch
.enabled
) {
621 usage_with_options(cat_file_usage
, options
);
624 if (force_path
&& opt
!= 'c' && opt
!= 'w') {
625 error("--path=<path> needs --textconv or --filters");
626 usage_with_options(cat_file_usage
, options
);
629 if (force_path
&& batch
.enabled
) {
630 error("--path=<path> incompatible with --batch");
631 usage_with_options(cat_file_usage
, options
);
634 if (batch
.buffer_output
< 0)
635 batch
.buffer_output
= batch
.all_objects
;
638 return batch_objects(&batch
);
640 if (unknown_type
&& opt
!= 't' && opt
!= 's')
641 die("git cat-file --allow-unknown-type: use with -s or -t");
642 return cat_one_file(opt
, exp_type
, obj_name
, unknown_type
);