Provide some linguistic guidance for the documentation.
[git/mjg.git] / builtin / cat-file.c
blob163ce6c77cea19b8c209e52ea2694f98d6cb0d05
1 /*
2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
5 */
6 #include "cache.h"
7 #include "exec_cmd.h"
8 #include "tag.h"
9 #include "tree.h"
10 #include "builtin.h"
11 #include "parse-options.h"
12 #include "diff.h"
13 #include "userdiff.h"
14 #include "streaming.h"
16 static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
18 unsigned char sha1[20];
19 enum object_type type;
20 char *buf;
21 unsigned long size;
22 struct object_context obj_context;
24 if (get_sha1_with_context(obj_name, 0, sha1, &obj_context))
25 die("Not a valid object name %s", obj_name);
27 buf = NULL;
28 switch (opt) {
29 case 't':
30 type = sha1_object_info(sha1, NULL);
31 if (type > 0) {
32 printf("%s\n", typename(type));
33 return 0;
35 break;
37 case 's':
38 type = sha1_object_info(sha1, &size);
39 if (type > 0) {
40 printf("%lu\n", size);
41 return 0;
43 break;
45 case 'e':
46 return !has_sha1_file(sha1);
48 case 'p':
49 type = sha1_object_info(sha1, NULL);
50 if (type < 0)
51 die("Not a valid object name %s", obj_name);
53 /* custom pretty-print here */
54 if (type == OBJ_TREE) {
55 const char *ls_args[3] = { NULL };
56 ls_args[0] = "ls-tree";
57 ls_args[1] = obj_name;
58 return cmd_ls_tree(2, ls_args, NULL);
61 if (type == OBJ_BLOB)
62 return stream_blob_to_fd(1, sha1, NULL, 0);
63 buf = read_sha1_file(sha1, &type, &size);
64 if (!buf)
65 die("Cannot read object %s", obj_name);
67 /* otherwise just spit out the data */
68 break;
70 case 'c':
71 if (!obj_context.path[0])
72 die("git cat-file --textconv %s: <object> must be <sha1:path>",
73 obj_name);
75 if (!textconv_object(obj_context.path, obj_context.mode, sha1, 1, &buf, &size))
76 die("git cat-file --textconv: unable to run textconv on %s",
77 obj_name);
78 break;
80 case 0:
81 if (type_from_string(exp_type) == OBJ_BLOB) {
82 unsigned char blob_sha1[20];
83 if (sha1_object_info(sha1, NULL) == OBJ_TAG) {
84 enum object_type type;
85 unsigned long size;
86 char *buffer = read_sha1_file(sha1, &type, &size);
87 if (memcmp(buffer, "object ", 7) ||
88 get_sha1_hex(buffer + 7, blob_sha1))
89 die("%s not a valid tag", sha1_to_hex(sha1));
90 free(buffer);
91 } else
92 hashcpy(blob_sha1, sha1);
94 if (sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
95 return stream_blob_to_fd(1, blob_sha1, NULL, 0);
97 * we attempted to dereference a tag to a blob
98 * and failed; there may be new dereference
99 * mechanisms this code is not aware of.
100 * fall-back to the usual case.
103 buf = read_object_with_reference(sha1, exp_type, &size, NULL);
104 break;
106 default:
107 die("git cat-file: unknown option: %s", exp_type);
110 if (!buf)
111 die("git cat-file %s: bad file", obj_name);
113 write_or_die(1, buf, size);
114 return 0;
117 struct expand_data {
118 unsigned char sha1[20];
119 enum object_type type;
120 unsigned long size;
121 unsigned long disk_size;
122 const char *rest;
125 * If mark_query is true, we do not expand anything, but rather
126 * just mark the object_info with items we wish to query.
128 int mark_query;
131 * After a mark_query run, this object_info is set up to be
132 * passed to sha1_object_info_extended. It will point to the data
133 * elements above, so you can retrieve the response from there.
135 struct object_info info;
138 static int is_atom(const char *atom, const char *s, int slen)
140 int alen = strlen(atom);
141 return alen == slen && !memcmp(atom, s, alen);
144 static void expand_atom(struct strbuf *sb, const char *atom, int len,
145 void *vdata)
147 struct expand_data *data = vdata;
149 if (is_atom("objectname", atom, len)) {
150 if (!data->mark_query)
151 strbuf_addstr(sb, sha1_to_hex(data->sha1));
152 } else if (is_atom("objecttype", atom, len)) {
153 if (data->mark_query)
154 data->info.typep = &data->type;
155 else
156 strbuf_addstr(sb, typename(data->type));
157 } else if (is_atom("objectsize", atom, len)) {
158 if (data->mark_query)
159 data->info.sizep = &data->size;
160 else
161 strbuf_addf(sb, "%lu", data->size);
162 } else if (is_atom("objectsize:disk", atom, len)) {
163 if (data->mark_query)
164 data->info.disk_sizep = &data->disk_size;
165 else
166 strbuf_addf(sb, "%lu", data->disk_size);
167 } else if (is_atom("rest", atom, len)) {
168 if (!data->mark_query && data->rest)
169 strbuf_addstr(sb, data->rest);
170 } else
171 die("unknown format element: %.*s", len, atom);
174 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
176 const char *end;
178 if (*start != '(')
179 return 0;
180 end = strchr(start + 1, ')');
181 if (!end)
182 die("format element '%s' does not end in ')'", start);
184 expand_atom(sb, start + 1, end - start - 1, data);
186 return end - start + 1;
189 static void print_object_or_die(int fd, const unsigned char *sha1,
190 enum object_type type, unsigned long size)
192 if (type == OBJ_BLOB) {
193 if (stream_blob_to_fd(fd, sha1, NULL, 0) < 0)
194 die("unable to stream %s to stdout", sha1_to_hex(sha1));
196 else {
197 enum object_type rtype;
198 unsigned long rsize;
199 void *contents;
201 contents = read_sha1_file(sha1, &rtype, &rsize);
202 if (!contents)
203 die("object %s disappeared", sha1_to_hex(sha1));
204 if (rtype != type)
205 die("object %s changed type!?", sha1_to_hex(sha1));
206 if (rsize != size)
207 die("object %s change size!?", sha1_to_hex(sha1));
209 write_or_die(fd, contents, size);
210 free(contents);
214 struct batch_options {
215 int enabled;
216 int print_contents;
217 const char *format;
220 static int batch_one_object(const char *obj_name, struct batch_options *opt,
221 struct expand_data *data)
223 struct strbuf buf = STRBUF_INIT;
225 if (!obj_name)
226 return 1;
228 if (get_sha1(obj_name, data->sha1)) {
229 printf("%s missing\n", obj_name);
230 fflush(stdout);
231 return 0;
234 if (sha1_object_info_extended(data->sha1, &data->info) < 0) {
235 printf("%s missing\n", obj_name);
236 fflush(stdout);
237 return 0;
240 strbuf_expand(&buf, opt->format, expand_format, data);
241 strbuf_addch(&buf, '\n');
242 write_or_die(1, buf.buf, buf.len);
243 strbuf_release(&buf);
245 if (opt->print_contents) {
246 print_object_or_die(1, data->sha1, data->type, data->size);
247 write_or_die(1, "\n", 1);
249 return 0;
252 static int batch_objects(struct batch_options *opt)
254 struct strbuf buf = STRBUF_INIT;
255 struct expand_data data;
257 if (!opt->format)
258 opt->format = "%(objectname) %(objecttype) %(objectsize)";
261 * Expand once with our special mark_query flag, which will prime the
262 * object_info to be handed to sha1_object_info_extended for each
263 * object.
265 memset(&data, 0, sizeof(data));
266 data.mark_query = 1;
267 strbuf_expand(&buf, opt->format, expand_format, &data);
268 data.mark_query = 0;
271 * We are going to call get_sha1 on a potentially very large number of
272 * objects. In most large cases, these will be actual object sha1s. The
273 * cost to double-check that each one is not also a ref (just so we can
274 * warn) ends up dwarfing the actual cost of the object lookups
275 * themselves. We can work around it by just turning off the warning.
277 warn_on_object_refname_ambiguity = 0;
279 while (strbuf_getline(&buf, stdin, '\n') != EOF) {
280 char *p;
281 int error;
284 * Split at first whitespace, tying off the beginning of the
285 * string and saving the remainder (or NULL) in data.rest.
287 p = strpbrk(buf.buf, " \t");
288 if (p) {
289 while (*p && strchr(" \t", *p))
290 *p++ = '\0';
292 data.rest = p;
294 error = batch_one_object(buf.buf, opt, &data);
295 if (error)
296 return error;
299 return 0;
302 static const char * const cat_file_usage[] = {
303 N_("git cat-file (-t|-s|-e|-p|<type>|--textconv) <object>"),
304 N_("git cat-file (--batch|--batch-check) < <list_of_objects>"),
305 NULL
308 static int git_cat_file_config(const char *var, const char *value, void *cb)
310 if (userdiff_config(var, value) < 0)
311 return -1;
313 return git_default_config(var, value, cb);
316 static int batch_option_callback(const struct option *opt,
317 const char *arg,
318 int unset)
320 struct batch_options *bo = opt->value;
322 if (unset) {
323 memset(bo, 0, sizeof(*bo));
324 return 0;
327 bo->enabled = 1;
328 bo->print_contents = !strcmp(opt->long_name, "batch");
329 bo->format = arg;
331 return 0;
334 int cmd_cat_file(int argc, const char **argv, const char *prefix)
336 int opt = 0;
337 const char *exp_type = NULL, *obj_name = NULL;
338 struct batch_options batch = {0};
340 const struct option options[] = {
341 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
342 OPT_SET_INT('t', NULL, &opt, N_("show object type"), 't'),
343 OPT_SET_INT('s', NULL, &opt, N_("show object size"), 's'),
344 OPT_SET_INT('e', NULL, &opt,
345 N_("exit with zero when there's no error"), 'e'),
346 OPT_SET_INT('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
347 OPT_SET_INT(0, "textconv", &opt,
348 N_("for blob objects, run textconv on object's content"), 'c'),
349 { OPTION_CALLBACK, 0, "batch", &batch, "format",
350 N_("show info and content of objects fed from the standard input"),
351 PARSE_OPT_OPTARG, batch_option_callback },
352 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
353 N_("show info about objects fed from the standard input"),
354 PARSE_OPT_OPTARG, batch_option_callback },
355 OPT_END()
358 git_config(git_cat_file_config, NULL);
360 if (argc != 3 && argc != 2)
361 usage_with_options(cat_file_usage, options);
363 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
365 if (opt) {
366 if (argc == 1)
367 obj_name = argv[0];
368 else
369 usage_with_options(cat_file_usage, options);
371 if (!opt && !batch.enabled) {
372 if (argc == 2) {
373 exp_type = argv[0];
374 obj_name = argv[1];
375 } else
376 usage_with_options(cat_file_usage, options);
378 if (batch.enabled && (opt || argc)) {
379 usage_with_options(cat_file_usage, options);
382 if (batch.enabled)
383 return batch_objects(&batch);
385 return cat_one_file(opt, exp_type, obj_name);