2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
8 #include "parse-options.h"
10 #include "streaming.h"
11 #include "tree-walk.h"
12 #include "sha1-array.h"
14 struct batch_options
{
23 static int cat_one_file(int opt
, const char *exp_type
, const char *obj_name
,
27 enum object_type type
;
30 struct object_context obj_context
;
31 struct object_info oi
= {NULL
};
32 struct strbuf sb
= STRBUF_INIT
;
33 unsigned flags
= LOOKUP_REPLACE_OBJECT
;
36 flags
|= LOOKUP_UNKNOWN_OBJECT
;
38 if (get_sha1_with_context(obj_name
, 0, oid
.hash
, &obj_context
))
39 die("Not a valid object name %s", obj_name
);
45 if (sha1_object_info_extended(oid
.hash
, &oi
, flags
) < 0)
46 die("git cat-file: could not get object info");
48 printf("%s\n", sb
.buf
);
56 if (sha1_object_info_extended(oid
.hash
, &oi
, flags
) < 0)
57 die("git cat-file: could not get object info");
58 printf("%lu\n", size
);
62 return !has_object_file(&oid
);
65 if (!obj_context
.path
[0])
66 die("git cat-file --textconv %s: <object> must be <sha1:path>",
69 if (textconv_object(obj_context
.path
, obj_context
.mode
, &oid
, 1, &buf
, &size
))
73 type
= sha1_object_info(oid
.hash
, NULL
);
75 die("Not a valid object name %s", obj_name
);
77 /* custom pretty-print here */
78 if (type
== OBJ_TREE
) {
79 const char *ls_args
[3] = { NULL
};
80 ls_args
[0] = "ls-tree";
81 ls_args
[1] = obj_name
;
82 return cmd_ls_tree(2, ls_args
, NULL
);
86 return stream_blob_to_fd(1, &oid
, NULL
, 0);
87 buf
= read_sha1_file(oid
.hash
, &type
, &size
);
89 die("Cannot read object %s", obj_name
);
91 /* otherwise just spit out the data */
95 if (type_from_string(exp_type
) == OBJ_BLOB
) {
96 struct object_id blob_oid
;
97 if (sha1_object_info(oid
.hash
, NULL
) == OBJ_TAG
) {
98 char *buffer
= read_sha1_file(oid
.hash
, &type
, &size
);
100 if (!skip_prefix(buffer
, "object ", &target
) ||
101 get_oid_hex(target
, &blob_oid
))
102 die("%s not a valid tag", oid_to_hex(&oid
));
105 oidcpy(&blob_oid
, &oid
);
107 if (sha1_object_info(blob_oid
.hash
, NULL
) == OBJ_BLOB
)
108 return stream_blob_to_fd(1, &blob_oid
, NULL
, 0);
110 * we attempted to dereference a tag to a blob
111 * and failed; there may be new dereference
112 * mechanisms this code is not aware of.
113 * fall-back to the usual case.
116 buf
= read_object_with_reference(oid
.hash
, exp_type
, &size
, NULL
);
120 die("git cat-file: unknown option: %s", exp_type
);
124 die("git cat-file %s: bad file", obj_name
);
126 write_or_die(1, buf
, size
);
131 struct object_id oid
;
132 enum object_type type
;
136 struct object_id delta_base_oid
;
139 * If mark_query is true, we do not expand anything, but rather
140 * just mark the object_info with items we wish to query.
145 * Whether to split the input on whitespace before feeding it to
146 * get_sha1; this is decided during the mark_query phase based on
147 * whether we have a %(rest) token in our format.
149 int split_on_whitespace
;
152 * After a mark_query run, this object_info is set up to be
153 * passed to sha1_object_info_extended. It will point to the data
154 * elements above, so you can retrieve the response from there.
156 struct object_info info
;
159 * This flag will be true if the requested batch format and options
160 * don't require us to call sha1_object_info, which can then be
163 unsigned skip_object_info
: 1;
166 static int is_atom(const char *atom
, const char *s
, int slen
)
168 int alen
= strlen(atom
);
169 return alen
== slen
&& !memcmp(atom
, s
, alen
);
172 static void expand_atom(struct strbuf
*sb
, const char *atom
, int len
,
175 struct expand_data
*data
= vdata
;
177 if (is_atom("objectname", atom
, len
)) {
178 if (!data
->mark_query
)
179 strbuf_addstr(sb
, oid_to_hex(&data
->oid
));
180 } else if (is_atom("objecttype", atom
, len
)) {
181 if (data
->mark_query
)
182 data
->info
.typep
= &data
->type
;
184 strbuf_addstr(sb
, typename(data
->type
));
185 } else if (is_atom("objectsize", atom
, len
)) {
186 if (data
->mark_query
)
187 data
->info
.sizep
= &data
->size
;
189 strbuf_addf(sb
, "%lu", data
->size
);
190 } else if (is_atom("objectsize:disk", atom
, len
)) {
191 if (data
->mark_query
)
192 data
->info
.disk_sizep
= &data
->disk_size
;
194 strbuf_addf(sb
, "%"PRIuMAX
, (uintmax_t)data
->disk_size
);
195 } else if (is_atom("rest", atom
, len
)) {
196 if (data
->mark_query
)
197 data
->split_on_whitespace
= 1;
199 strbuf_addstr(sb
, data
->rest
);
200 } else if (is_atom("deltabase", atom
, len
)) {
201 if (data
->mark_query
)
202 data
->info
.delta_base_sha1
= data
->delta_base_oid
.hash
;
205 oid_to_hex(&data
->delta_base_oid
));
207 die("unknown format element: %.*s", len
, atom
);
210 static size_t expand_format(struct strbuf
*sb
, const char *start
, void *data
)
216 end
= strchr(start
+ 1, ')');
218 die("format element '%s' does not end in ')'", start
);
220 expand_atom(sb
, start
+ 1, end
- start
- 1, data
);
222 return end
- start
+ 1;
225 static void batch_write(struct batch_options
*opt
, const void *data
, int len
)
227 if (opt
->buffer_output
) {
228 if (fwrite(data
, 1, len
, stdout
) != len
)
229 die_errno("unable to write to stdout");
231 write_or_die(1, data
, len
);
234 static void print_object_or_die(struct batch_options
*opt
, struct expand_data
*data
)
236 const struct object_id
*oid
= &data
->oid
;
238 assert(data
->info
.typep
);
240 if (data
->type
== OBJ_BLOB
) {
241 if (opt
->buffer_output
)
243 if (stream_blob_to_fd(1, oid
, NULL
, 0) < 0)
244 die("unable to stream %s to stdout", oid_to_hex(oid
));
247 enum object_type type
;
251 contents
= read_sha1_file(oid
->hash
, &type
, &size
);
253 die("object %s disappeared", oid_to_hex(oid
));
254 if (type
!= data
->type
)
255 die("object %s changed type!?", oid_to_hex(oid
));
256 if (data
->info
.sizep
&& size
!= data
->size
)
257 die("object %s changed size!?", oid_to_hex(oid
));
259 batch_write(opt
, contents
, size
);
264 static void batch_object_write(const char *obj_name
, struct batch_options
*opt
,
265 struct expand_data
*data
)
267 struct strbuf buf
= STRBUF_INIT
;
269 if (!data
->skip_object_info
&&
270 sha1_object_info_extended(data
->oid
.hash
, &data
->info
, LOOKUP_REPLACE_OBJECT
) < 0) {
271 printf("%s missing\n",
272 obj_name
? obj_name
: oid_to_hex(&data
->oid
));
277 strbuf_expand(&buf
, opt
->format
, expand_format
, data
);
278 strbuf_addch(&buf
, '\n');
279 batch_write(opt
, buf
.buf
, buf
.len
);
280 strbuf_release(&buf
);
282 if (opt
->print_contents
) {
283 print_object_or_die(opt
, data
);
284 batch_write(opt
, "\n", 1);
288 static void batch_one_object(const char *obj_name
, struct batch_options
*opt
,
289 struct expand_data
*data
)
291 struct object_context ctx
;
292 int flags
= opt
->follow_symlinks
? GET_SHA1_FOLLOW_SYMLINKS
: 0;
293 enum follow_symlinks_result result
;
295 result
= get_sha1_with_context(obj_name
, flags
, data
->oid
.hash
, &ctx
);
296 if (result
!= FOUND
) {
299 printf("%s missing\n", obj_name
);
301 case DANGLING_SYMLINK
:
302 printf("dangling %"PRIuMAX
"\n%s\n",
303 (uintmax_t)strlen(obj_name
), obj_name
);
306 printf("loop %"PRIuMAX
"\n%s\n",
307 (uintmax_t)strlen(obj_name
), obj_name
);
310 printf("notdir %"PRIuMAX
"\n%s\n",
311 (uintmax_t)strlen(obj_name
), obj_name
);
314 die("BUG: unknown get_sha1_with_context result %d\n",
323 printf("symlink %"PRIuMAX
"\n%s\n",
324 (uintmax_t)ctx
.symlink_path
.len
,
325 ctx
.symlink_path
.buf
);
330 batch_object_write(obj_name
, opt
, data
);
333 struct object_cb_data
{
334 struct batch_options
*opt
;
335 struct expand_data
*expand
;
338 static void batch_object_cb(const unsigned char sha1
[20], void *vdata
)
340 struct object_cb_data
*data
= vdata
;
341 hashcpy(data
->expand
->oid
.hash
, sha1
);
342 batch_object_write(NULL
, data
->opt
, data
->expand
);
345 static int batch_loose_object(const unsigned char *sha1
,
349 sha1_array_append(data
, sha1
);
353 static int batch_packed_object(const unsigned char *sha1
,
354 struct packed_git
*pack
,
358 sha1_array_append(data
, sha1
);
362 static int batch_objects(struct batch_options
*opt
)
364 struct strbuf buf
= STRBUF_INIT
;
365 struct expand_data data
;
370 opt
->format
= "%(objectname) %(objecttype) %(objectsize)";
373 * Expand once with our special mark_query flag, which will prime the
374 * object_info to be handed to sha1_object_info_extended for each
377 memset(&data
, 0, sizeof(data
));
379 strbuf_expand(&buf
, opt
->format
, expand_format
, &data
);
382 if (opt
->all_objects
) {
383 struct object_info empty
;
384 memset(&empty
, 0, sizeof(empty
));
385 if (!memcmp(&data
.info
, &empty
, sizeof(empty
)))
386 data
.skip_object_info
= 1;
390 * If we are printing out the object, then always fill in the type,
391 * since we will want to decide whether or not to stream.
393 if (opt
->print_contents
)
394 data
.info
.typep
= &data
.type
;
396 if (opt
->all_objects
) {
397 struct sha1_array sa
= SHA1_ARRAY_INIT
;
398 struct object_cb_data cb
;
400 for_each_loose_object(batch_loose_object
, &sa
, 0);
401 for_each_packed_object(batch_packed_object
, &sa
, 0);
405 sha1_array_for_each_unique(&sa
, batch_object_cb
, &cb
);
407 sha1_array_clear(&sa
);
412 * We are going to call get_sha1 on a potentially very large number of
413 * objects. In most large cases, these will be actual object sha1s. The
414 * cost to double-check that each one is not also a ref (just so we can
415 * warn) ends up dwarfing the actual cost of the object lookups
416 * themselves. We can work around it by just turning off the warning.
418 save_warning
= warn_on_object_refname_ambiguity
;
419 warn_on_object_refname_ambiguity
= 0;
421 while (strbuf_getline(&buf
, stdin
) != EOF
) {
422 if (data
.split_on_whitespace
) {
424 * Split at first whitespace, tying off the beginning
425 * of the string and saving the remainder (or NULL) in
428 char *p
= strpbrk(buf
.buf
, " \t");
430 while (*p
&& strchr(" \t", *p
))
436 batch_one_object(buf
.buf
, opt
, &data
);
439 strbuf_release(&buf
);
440 warn_on_object_refname_ambiguity
= save_warning
;
444 static const char * const cat_file_usage
[] = {
445 N_("git cat-file (-t [--allow-unknown-type]|-s [--allow-unknown-type]|-e|-p|<type>|--textconv) <object>"),
446 N_("git cat-file (--batch | --batch-check) [--follow-symlinks]"),
450 static int git_cat_file_config(const char *var
, const char *value
, void *cb
)
452 if (userdiff_config(var
, value
) < 0)
455 return git_default_config(var
, value
, cb
);
458 static int batch_option_callback(const struct option
*opt
,
462 struct batch_options
*bo
= opt
->value
;
469 bo
->print_contents
= !strcmp(opt
->long_name
, "batch");
475 int cmd_cat_file(int argc
, const char **argv
, const char *prefix
)
478 const char *exp_type
= NULL
, *obj_name
= NULL
;
479 struct batch_options batch
= {0};
480 int unknown_type
= 0;
482 const struct option options
[] = {
483 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
484 OPT_CMDMODE('t', NULL
, &opt
, N_("show object type"), 't'),
485 OPT_CMDMODE('s', NULL
, &opt
, N_("show object size"), 's'),
486 OPT_CMDMODE('e', NULL
, &opt
,
487 N_("exit with zero when there's no error"), 'e'),
488 OPT_CMDMODE('p', NULL
, &opt
, N_("pretty-print object's content"), 'p'),
489 OPT_CMDMODE(0, "textconv", &opt
,
490 N_("for blob objects, run textconv on object's content"), 'c'),
491 OPT_BOOL(0, "allow-unknown-type", &unknown_type
,
492 N_("allow -s and -t to work with broken/corrupt objects")),
493 OPT_BOOL(0, "buffer", &batch
.buffer_output
, N_("buffer --batch output")),
494 { OPTION_CALLBACK
, 0, "batch", &batch
, "format",
495 N_("show info and content of objects fed from the standard input"),
496 PARSE_OPT_OPTARG
, batch_option_callback
},
497 { OPTION_CALLBACK
, 0, "batch-check", &batch
, "format",
498 N_("show info about objects fed from the standard input"),
499 PARSE_OPT_OPTARG
, batch_option_callback
},
500 OPT_BOOL(0, "follow-symlinks", &batch
.follow_symlinks
,
501 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
502 OPT_BOOL(0, "batch-all-objects", &batch
.all_objects
,
503 N_("show all objects with --batch or --batch-check")),
507 git_config(git_cat_file_config
, NULL
);
509 batch
.buffer_output
= -1;
510 argc
= parse_options(argc
, argv
, prefix
, options
, cat_file_usage
, 0);
516 usage_with_options(cat_file_usage
, options
);
518 if (!opt
&& !batch
.enabled
) {
523 usage_with_options(cat_file_usage
, options
);
525 if (batch
.enabled
&& (opt
|| argc
)) {
526 usage_with_options(cat_file_usage
, options
);
529 if ((batch
.follow_symlinks
|| batch
.all_objects
) && !batch
.enabled
) {
530 usage_with_options(cat_file_usage
, options
);
533 if (batch
.buffer_output
< 0)
534 batch
.buffer_output
= batch
.all_objects
;
537 return batch_objects(&batch
);
539 if (unknown_type
&& opt
!= 't' && opt
!= 's')
540 die("git cat-file --allow-unknown-type: use with -s or -t");
541 return cat_one_file(opt
, exp_type
, obj_name
, unknown_type
);