Merge branch 'ah/usage-strings'
[git.git] / builtin / cat-file.c
blobdf99df4db1ddb058368fe88f07c890d87600e6d9
1 /*
2 * GIT - The information manager from hell
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"
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;
16 char *buf;
17 unsigned long size;
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);
23 buf = NULL;
24 switch (opt) {
25 case 't':
26 type = sha1_object_info(sha1, NULL);
27 if (type > 0) {
28 printf("%s\n", typename(type));
29 return 0;
31 break;
33 case 's':
34 type = sha1_object_info(sha1, &size);
35 if (type > 0) {
36 printf("%lu\n", size);
37 return 0;
39 break;
41 case 'e':
42 return !has_sha1_file(sha1);
44 case 'c':
45 if (!obj_context.path[0])
46 die("git cat-file --textconv %s: <object> must be <sha1:path>",
47 obj_name);
49 if (textconv_object(obj_context.path, obj_context.mode, sha1, 1, &buf, &size))
50 break;
52 case 'p':
53 type = sha1_object_info(sha1, NULL);
54 if (type < 0)
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);
65 if (type == OBJ_BLOB)
66 return stream_blob_to_fd(1, sha1, NULL, 0);
67 buf = read_sha1_file(sha1, &type, &size);
68 if (!buf)
69 die("Cannot read object %s", obj_name);
71 /* otherwise just spit out the data */
72 break;
74 case 0:
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 char *buffer = read_sha1_file(sha1, &type, &size);
79 const char *target;
80 if (!skip_prefix(buffer, "object ", &target) ||
81 get_sha1_hex(target, blob_sha1))
82 die("%s not a valid tag", sha1_to_hex(sha1));
83 free(buffer);
84 } else
85 hashcpy(blob_sha1, sha1);
87 if (sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
88 return stream_blob_to_fd(1, blob_sha1, NULL, 0);
90 * we attempted to dereference a tag to a blob
91 * and failed; there may be new dereference
92 * mechanisms this code is not aware of.
93 * fall-back to the usual case.
96 buf = read_object_with_reference(sha1, exp_type, &size, NULL);
97 break;
99 default:
100 die("git cat-file: unknown option: %s", exp_type);
103 if (!buf)
104 die("git cat-file %s: bad file", obj_name);
106 write_or_die(1, buf, size);
107 return 0;
110 struct expand_data {
111 unsigned char sha1[20];
112 enum object_type type;
113 unsigned long size;
114 unsigned long disk_size;
115 const char *rest;
116 unsigned char delta_base_sha1[20];
119 * If mark_query is true, we do not expand anything, but rather
120 * just mark the object_info with items we wish to query.
122 int mark_query;
125 * Whether to split the input on whitespace before feeding it to
126 * get_sha1; this is decided during the mark_query phase based on
127 * whether we have a %(rest) token in our format.
129 int split_on_whitespace;
132 * After a mark_query run, this object_info is set up to be
133 * passed to sha1_object_info_extended. It will point to the data
134 * elements above, so you can retrieve the response from there.
136 struct object_info info;
139 static int is_atom(const char *atom, const char *s, int slen)
141 int alen = strlen(atom);
142 return alen == slen && !memcmp(atom, s, alen);
145 static void expand_atom(struct strbuf *sb, const char *atom, int len,
146 void *vdata)
148 struct expand_data *data = vdata;
150 if (is_atom("objectname", atom, len)) {
151 if (!data->mark_query)
152 strbuf_addstr(sb, sha1_to_hex(data->sha1));
153 } else if (is_atom("objecttype", atom, len)) {
154 if (data->mark_query)
155 data->info.typep = &data->type;
156 else
157 strbuf_addstr(sb, typename(data->type));
158 } else if (is_atom("objectsize", atom, len)) {
159 if (data->mark_query)
160 data->info.sizep = &data->size;
161 else
162 strbuf_addf(sb, "%lu", data->size);
163 } else if (is_atom("objectsize:disk", atom, len)) {
164 if (data->mark_query)
165 data->info.disk_sizep = &data->disk_size;
166 else
167 strbuf_addf(sb, "%lu", data->disk_size);
168 } else if (is_atom("rest", atom, len)) {
169 if (data->mark_query)
170 data->split_on_whitespace = 1;
171 else if (data->rest)
172 strbuf_addstr(sb, data->rest);
173 } else if (is_atom("deltabase", atom, len)) {
174 if (data->mark_query)
175 data->info.delta_base_sha1 = data->delta_base_sha1;
176 else
177 strbuf_addstr(sb, sha1_to_hex(data->delta_base_sha1));
178 } else
179 die("unknown format element: %.*s", len, atom);
182 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
184 const char *end;
186 if (*start != '(')
187 return 0;
188 end = strchr(start + 1, ')');
189 if (!end)
190 die("format element '%s' does not end in ')'", start);
192 expand_atom(sb, start + 1, end - start - 1, data);
194 return end - start + 1;
197 static void print_object_or_die(int fd, struct expand_data *data)
199 const unsigned char *sha1 = data->sha1;
201 assert(data->info.typep);
203 if (data->type == OBJ_BLOB) {
204 if (stream_blob_to_fd(fd, sha1, NULL, 0) < 0)
205 die("unable to stream %s to stdout", sha1_to_hex(sha1));
207 else {
208 enum object_type type;
209 unsigned long size;
210 void *contents;
212 contents = read_sha1_file(sha1, &type, &size);
213 if (!contents)
214 die("object %s disappeared", sha1_to_hex(sha1));
215 if (type != data->type)
216 die("object %s changed type!?", sha1_to_hex(sha1));
217 if (data->info.sizep && size != data->size)
218 die("object %s changed size!?", sha1_to_hex(sha1));
220 write_or_die(fd, contents, size);
221 free(contents);
225 struct batch_options {
226 int enabled;
227 int print_contents;
228 const char *format;
231 static int batch_one_object(const char *obj_name, struct batch_options *opt,
232 struct expand_data *data)
234 struct strbuf buf = STRBUF_INIT;
236 if (!obj_name)
237 return 1;
239 if (get_sha1(obj_name, data->sha1)) {
240 printf("%s missing\n", obj_name);
241 fflush(stdout);
242 return 0;
245 if (sha1_object_info_extended(data->sha1, &data->info, LOOKUP_REPLACE_OBJECT) < 0) {
246 printf("%s missing\n", obj_name);
247 fflush(stdout);
248 return 0;
251 strbuf_expand(&buf, opt->format, expand_format, data);
252 strbuf_addch(&buf, '\n');
253 write_or_die(1, buf.buf, buf.len);
254 strbuf_release(&buf);
256 if (opt->print_contents) {
257 print_object_or_die(1, data);
258 write_or_die(1, "\n", 1);
260 return 0;
263 static int batch_objects(struct batch_options *opt)
265 struct strbuf buf = STRBUF_INIT;
266 struct expand_data data;
267 int save_warning;
268 int retval = 0;
270 if (!opt->format)
271 opt->format = "%(objectname) %(objecttype) %(objectsize)";
274 * Expand once with our special mark_query flag, which will prime the
275 * object_info to be handed to sha1_object_info_extended for each
276 * object.
278 memset(&data, 0, sizeof(data));
279 data.mark_query = 1;
280 strbuf_expand(&buf, opt->format, expand_format, &data);
281 data.mark_query = 0;
284 * If we are printing out the object, then always fill in the type,
285 * since we will want to decide whether or not to stream.
287 if (opt->print_contents)
288 data.info.typep = &data.type;
291 * We are going to call get_sha1 on a potentially very large number of
292 * objects. In most large cases, these will be actual object sha1s. The
293 * cost to double-check that each one is not also a ref (just so we can
294 * warn) ends up dwarfing the actual cost of the object lookups
295 * themselves. We can work around it by just turning off the warning.
297 save_warning = warn_on_object_refname_ambiguity;
298 warn_on_object_refname_ambiguity = 0;
300 while (strbuf_getline(&buf, stdin, '\n') != EOF) {
301 if (data.split_on_whitespace) {
303 * Split at first whitespace, tying off the beginning
304 * of the string and saving the remainder (or NULL) in
305 * data.rest.
307 char *p = strpbrk(buf.buf, " \t");
308 if (p) {
309 while (*p && strchr(" \t", *p))
310 *p++ = '\0';
312 data.rest = p;
315 retval = batch_one_object(buf.buf, opt, &data);
316 if (retval)
317 break;
320 strbuf_release(&buf);
321 warn_on_object_refname_ambiguity = save_warning;
322 return retval;
325 static const char * const cat_file_usage[] = {
326 N_("git cat-file (-t | -s | -e | -p | <type> | --textconv) <object>"),
327 N_("git cat-file (--batch | --batch-check) < <list-of-objects>"),
328 NULL
331 static int git_cat_file_config(const char *var, const char *value, void *cb)
333 if (userdiff_config(var, value) < 0)
334 return -1;
336 return git_default_config(var, value, cb);
339 static int batch_option_callback(const struct option *opt,
340 const char *arg,
341 int unset)
343 struct batch_options *bo = opt->value;
345 if (unset) {
346 memset(bo, 0, sizeof(*bo));
347 return 0;
350 bo->enabled = 1;
351 bo->print_contents = !strcmp(opt->long_name, "batch");
352 bo->format = arg;
354 return 0;
357 int cmd_cat_file(int argc, const char **argv, const char *prefix)
359 int opt = 0;
360 const char *exp_type = NULL, *obj_name = NULL;
361 struct batch_options batch = {0};
363 const struct option options[] = {
364 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
365 OPT_SET_INT('t', NULL, &opt, N_("show object type"), 't'),
366 OPT_SET_INT('s', NULL, &opt, N_("show object size"), 's'),
367 OPT_SET_INT('e', NULL, &opt,
368 N_("exit with zero when there's no error"), 'e'),
369 OPT_SET_INT('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
370 OPT_SET_INT(0, "textconv", &opt,
371 N_("for blob objects, run textconv on object's content"), 'c'),
372 { OPTION_CALLBACK, 0, "batch", &batch, "format",
373 N_("show info and content of objects fed from the standard input"),
374 PARSE_OPT_OPTARG, batch_option_callback },
375 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
376 N_("show info about objects fed from the standard input"),
377 PARSE_OPT_OPTARG, batch_option_callback },
378 OPT_END()
381 git_config(git_cat_file_config, NULL);
383 if (argc != 3 && argc != 2)
384 usage_with_options(cat_file_usage, options);
386 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
388 if (opt) {
389 if (argc == 1)
390 obj_name = argv[0];
391 else
392 usage_with_options(cat_file_usage, options);
394 if (!opt && !batch.enabled) {
395 if (argc == 2) {
396 exp_type = argv[0];
397 obj_name = argv[1];
398 } else
399 usage_with_options(cat_file_usage, options);
401 if (batch.enabled && (opt || argc)) {
402 usage_with_options(cat_file_usage, options);
405 if (batch.enabled)
406 return batch_objects(&batch);
408 return cat_one_file(opt, exp_type, obj_name);