2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
8 #include "parse-options.h"
10 #include "streaming.h"
12 static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
14 unsigned char sha1[20];
15 enum object_type type;
18 struct object_context obj_context;
20 if (get_sha1_with_context(obj_name, 0, sha1, &obj_context))
21 die("Not a valid object name %s", obj_name);
26 type = sha1_object_info(sha1, NULL);
28 printf("%s\n", typename(type));
34 type = sha1_object_info(sha1, &size);
36 printf("%lu\n", size);
42 return !has_sha1_file(sha1);
45 if (!obj_context.path[0])
46 die("git cat-file --textconv %s: <object> must be <sha1:path>",
49 if (textconv_object(obj_context.path, obj_context.mode, sha1, 1, &buf, &size))
53 type = sha1_object_info(sha1, NULL);
55 die("Not a valid object name %s", obj_name);
57 /* custom pretty-print here */
58 if (type == OBJ_TREE) {
59 const char *ls_args[3] = { NULL };
60 ls_args[0] = "ls-tree";
61 ls_args[1] = obj_name;
62 return cmd_ls_tree(2, ls_args, NULL);
66 return stream_blob_to_fd(1, sha1, NULL, 0);
67 buf = read_sha1_file(sha1, &type, &size);
69 die("Cannot read object %s", obj_name);
71 /* otherwise just spit out the data */
75 if (type_from_string(exp_type) == OBJ_BLOB) {
76 unsigned char blob_sha1[20];
77 if (sha1_object_info(sha1, NULL) == OBJ_TAG) {
78 enum object_type type;
80 char *buffer = read_sha1_file(sha1, &type, &size);
82 if (!skip_prefix(buffer, "object ", &target) ||
83 get_sha1_hex(target, blob_sha1))
84 die("%s not a valid tag", sha1_to_hex(sha1));
87 hashcpy(blob_sha1, sha1);
89 if (sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
90 return stream_blob_to_fd(1, blob_sha1, NULL, 0);
92 * we attempted to dereference a tag to a blob
93 * and failed; there may be new dereference
94 * mechanisms this code is not aware of.
95 * fall-back to the usual case.
98 buf = read_object_with_reference(sha1, exp_type, &size, NULL);
102 die("git cat-file: unknown option: %s", exp_type);
106 die("git cat-file %s: bad file", obj_name);
108 write_or_die(1, buf, size);
113 unsigned char sha1[20];
114 enum object_type type;
116 unsigned long disk_size;
118 unsigned char delta_base_sha1[20];
121 * If mark_query is true, we do not expand anything, but rather
122 * just mark the object_info with items we wish to query.
127 * Whether to split the input on whitespace before feeding it to
128 * get_sha1; this is decided during the mark_query phase based on
129 * whether we have a %(rest) token in our format.
131 int split_on_whitespace;
134 * After a mark_query run, this object_info is set up to be
135 * passed to sha1_object_info_extended. It will point to the data
136 * elements above, so you can retrieve the response from there.
138 struct object_info info;
141 static int is_atom(const char *atom, const char *s, int slen)
143 int alen = strlen(atom);
144 return alen == slen && !memcmp(atom, s, alen);
147 static void expand_atom(struct strbuf *sb, const char *atom, int len,
150 struct expand_data *data = vdata;
152 if (is_atom("objectname", atom, len)) {
153 if (!data->mark_query)
154 strbuf_addstr(sb, sha1_to_hex(data->sha1));
155 } else if (is_atom("objecttype", atom, len)) {
156 if (data->mark_query)
157 data->info.typep = &data->type;
159 strbuf_addstr(sb, typename(data->type));
160 } else if (is_atom("objectsize", atom, len)) {
161 if (data->mark_query)
162 data->info.sizep = &data->size;
164 strbuf_addf(sb, "%lu", data->size);
165 } else if (is_atom("objectsize:disk", atom, len)) {
166 if (data->mark_query)
167 data->info.disk_sizep = &data->disk_size;
169 strbuf_addf(sb, "%lu", data->disk_size);
170 } else if (is_atom("rest", atom, len)) {
171 if (data->mark_query)
172 data->split_on_whitespace = 1;
174 strbuf_addstr(sb, data->rest);
175 } else if (is_atom("deltabase", atom, len)) {
176 if (data->mark_query)
177 data->info.delta_base_sha1 = data->delta_base_sha1;
179 strbuf_addstr(sb, sha1_to_hex(data->delta_base_sha1));
181 die("unknown format element: %.*s", len, atom);
184 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
190 end = strchr(start + 1, ')');
192 die("format element '%s' does not end in ')'", start);
194 expand_atom(sb, start + 1, end - start - 1, data);
196 return end - start + 1;
199 static void print_object_or_die(int fd, struct expand_data *data)
201 const unsigned char *sha1 = data->sha1;
203 assert(data->info.typep);
205 if (data->type == OBJ_BLOB) {
206 if (stream_blob_to_fd(fd, sha1, NULL, 0) < 0)
207 die("unable to stream %s to stdout", sha1_to_hex(sha1));
210 enum object_type type;
214 contents = read_sha1_file(sha1, &type, &size);
216 die("object %s disappeared", sha1_to_hex(sha1));
217 if (type != data->type)
218 die("object %s changed type!?", sha1_to_hex(sha1));
219 if (data->info.sizep && size != data->size)
220 die("object %s changed size!?", sha1_to_hex(sha1));
222 write_or_die(fd, contents, size);
227 struct batch_options {
233 static int batch_one_object(const char *obj_name, struct batch_options *opt,
234 struct expand_data *data)
236 struct strbuf buf = STRBUF_INIT;
241 if (get_sha1(obj_name, data->sha1)) {
242 printf("%s missing\n", obj_name);
247 if (sha1_object_info_extended(data->sha1, &data->info, LOOKUP_REPLACE_OBJECT) < 0) {
248 printf("%s missing\n", obj_name);
253 strbuf_expand(&buf, opt->format, expand_format, data);
254 strbuf_addch(&buf, '\n');
255 write_or_die(1, buf.buf, buf.len);
256 strbuf_release(&buf);
258 if (opt->print_contents) {
259 print_object_or_die(1, data);
260 write_or_die(1, "\n", 1);
265 static int batch_objects(struct batch_options *opt)
267 struct strbuf buf = STRBUF_INIT;
268 struct expand_data data;
273 opt->format = "%(objectname) %(objecttype) %(objectsize)";
276 * Expand once with our special mark_query flag, which will prime the
277 * object_info to be handed to sha1_object_info_extended for each
280 memset(&data, 0, sizeof(data));
282 strbuf_expand(&buf, opt->format, expand_format, &data);
286 * If we are printing out the object, then always fill in the type,
287 * since we will want to decide whether or not to stream.
289 if (opt->print_contents)
290 data.info.typep = &data.type;
293 * We are going to call get_sha1 on a potentially very large number of
294 * objects. In most large cases, these will be actual object sha1s. The
295 * cost to double-check that each one is not also a ref (just so we can
296 * warn) ends up dwarfing the actual cost of the object lookups
297 * themselves. We can work around it by just turning off the warning.
299 save_warning = warn_on_object_refname_ambiguity;
300 warn_on_object_refname_ambiguity = 0;
302 while (strbuf_getline(&buf, stdin, '\n') != EOF) {
303 if (data.split_on_whitespace) {
305 * Split at first whitespace, tying off the beginning
306 * of the string and saving the remainder (or NULL) in
309 char *p = strpbrk(buf.buf, " \t");
311 while (*p && strchr(" \t", *p))
317 retval = batch_one_object(buf.buf, opt, &data);
322 strbuf_release(&buf);
323 warn_on_object_refname_ambiguity = save_warning;
327 static const char * const cat_file_usage[] = {
328 N_("git cat-file (-t|-s|-e|-p|<type>|--textconv) <object>"),
329 N_("git cat-file (--batch|--batch-check) < <list_of_objects>"),
333 static int git_cat_file_config(const char *var, const char *value, void *cb)
335 if (userdiff_config(var, value) < 0)
338 return git_default_config(var, value, cb);
341 static int batch_option_callback(const struct option *opt,
345 struct batch_options *bo = opt->value;
348 memset(bo, 0, sizeof(*bo));
353 bo->print_contents = !strcmp(opt->long_name, "batch");
359 int cmd_cat_file(int argc, const char **argv, const char *prefix)
362 const char *exp_type = NULL, *obj_name = NULL;
363 struct batch_options batch = {0};
365 const struct option options[] = {
366 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
367 OPT_SET_INT('t', NULL, &opt, N_("show object type"), 't'),
368 OPT_SET_INT('s', NULL, &opt, N_("show object size"), 's'),
369 OPT_SET_INT('e', NULL, &opt,
370 N_("exit with zero when there's no error"), 'e'),
371 OPT_SET_INT('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
372 OPT_SET_INT(0, "textconv", &opt,
373 N_("for blob objects, run textconv on object's content"), 'c'),
374 { OPTION_CALLBACK, 0, "batch", &batch, "format",
375 N_("show info and content of objects fed from the standard input"),
376 PARSE_OPT_OPTARG, batch_option_callback },
377 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
378 N_("show info about objects fed from the standard input"),
379 PARSE_OPT_OPTARG, batch_option_callback },
383 git_config(git_cat_file_config, NULL);
385 if (argc != 3 && argc != 2)
386 usage_with_options(cat_file_usage, options);
388 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
394 usage_with_options(cat_file_usage, options);
396 if (!opt && !batch.enabled) {
401 usage_with_options(cat_file_usage, options);
403 if (batch.enabled && (opt || argc)) {
404 usage_with_options(cat_file_usage, options);
408 return batch_objects(&batch);
410 return cat_one_file(opt, exp_type, obj_name);