completion: improve ls-files filter performance
[alt-git.git] / builtin / cat-file.c
blobd90170f070f4f6ab7750c31fdcfac462c934b26d
1 /*
2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
5 */
6 #include "cache.h"
7 #include "config.h"
8 #include "builtin.h"
9 #include "diff.h"
10 #include "parse-options.h"
11 #include "userdiff.h"
12 #include "streaming.h"
13 #include "tree-walk.h"
14 #include "sha1-array.h"
15 #include "packfile.h"
17 struct batch_options {
18 int enabled;
19 int follow_symlinks;
20 int print_contents;
21 int buffer_output;
22 int all_objects;
23 int cmdmode; /* may be 'w' or 'c' for --filters or --textconv */
24 const char *format;
27 static const char *force_path;
29 static int filter_object(const char *path, unsigned mode,
30 const struct object_id *oid,
31 char **buf, unsigned long *size)
33 enum object_type type;
35 *buf = read_sha1_file(oid->hash, &type, size);
36 if (!*buf)
37 return error(_("cannot read object %s '%s'"),
38 oid_to_hex(oid), path);
39 if ((type == OBJ_BLOB) && S_ISREG(mode)) {
40 struct strbuf strbuf = STRBUF_INIT;
41 if (convert_to_working_tree(path, *buf, *size, &strbuf)) {
42 free(*buf);
43 *size = strbuf.len;
44 *buf = strbuf_detach(&strbuf, NULL);
48 return 0;
51 static int cat_one_file(int opt, const char *exp_type, const char *obj_name,
52 int unknown_type)
54 struct object_id oid;
55 enum object_type type;
56 char *buf;
57 unsigned long size;
58 struct object_context obj_context;
59 struct object_info oi = OBJECT_INFO_INIT;
60 struct strbuf sb = STRBUF_INIT;
61 unsigned flags = OBJECT_INFO_LOOKUP_REPLACE;
62 const char *path = force_path;
64 if (unknown_type)
65 flags |= OBJECT_INFO_ALLOW_UNKNOWN_TYPE;
67 if (get_oid_with_context(obj_name, GET_OID_RECORD_PATH,
68 &oid, &obj_context))
69 die("Not a valid object name %s", obj_name);
71 if (!path)
72 path = obj_context.path;
73 if (obj_context.mode == S_IFINVALID)
74 obj_context.mode = 0100644;
76 buf = NULL;
77 switch (opt) {
78 case 't':
79 oi.type_name = &sb;
80 if (sha1_object_info_extended(oid.hash, &oi, flags) < 0)
81 die("git cat-file: could not get object info");
82 if (sb.len) {
83 printf("%s\n", sb.buf);
84 strbuf_release(&sb);
85 return 0;
87 break;
89 case 's':
90 oi.sizep = &size;
91 if (sha1_object_info_extended(oid.hash, &oi, flags) < 0)
92 die("git cat-file: could not get object info");
93 printf("%lu\n", size);
94 return 0;
96 case 'e':
97 return !has_object_file(&oid);
99 case 'w':
100 if (!path)
101 die("git cat-file --filters %s: <object> must be "
102 "<sha1:path>", obj_name);
104 if (filter_object(path, obj_context.mode,
105 &oid, &buf, &size))
106 return -1;
107 break;
109 case 'c':
110 if (!path)
111 die("git cat-file --textconv %s: <object> must be <sha1:path>",
112 obj_name);
114 if (textconv_object(path, obj_context.mode, &oid, 1, &buf, &size))
115 break;
116 /* else fallthrough */
118 case 'p':
119 type = sha1_object_info(oid.hash, NULL);
120 if (type < 0)
121 die("Not a valid object name %s", obj_name);
123 /* custom pretty-print here */
124 if (type == OBJ_TREE) {
125 const char *ls_args[3] = { NULL };
126 ls_args[0] = "ls-tree";
127 ls_args[1] = obj_name;
128 return cmd_ls_tree(2, ls_args, NULL);
131 if (type == OBJ_BLOB)
132 return stream_blob_to_fd(1, &oid, NULL, 0);
133 buf = read_sha1_file(oid.hash, &type, &size);
134 if (!buf)
135 die("Cannot read object %s", obj_name);
137 /* otherwise just spit out the data */
138 break;
140 case 0:
141 if (type_from_string(exp_type) == OBJ_BLOB) {
142 struct object_id blob_oid;
143 if (sha1_object_info(oid.hash, NULL) == OBJ_TAG) {
144 char *buffer = read_sha1_file(oid.hash, &type, &size);
145 const char *target;
146 if (!skip_prefix(buffer, "object ", &target) ||
147 get_oid_hex(target, &blob_oid))
148 die("%s not a valid tag", oid_to_hex(&oid));
149 free(buffer);
150 } else
151 oidcpy(&blob_oid, &oid);
153 if (sha1_object_info(blob_oid.hash, NULL) == OBJ_BLOB)
154 return stream_blob_to_fd(1, &blob_oid, NULL, 0);
156 * we attempted to dereference a tag to a blob
157 * and failed; there may be new dereference
158 * mechanisms this code is not aware of.
159 * fall-back to the usual case.
162 buf = read_object_with_reference(oid.hash, exp_type, &size, NULL);
163 break;
165 default:
166 die("git cat-file: unknown option: %s", exp_type);
169 if (!buf)
170 die("git cat-file %s: bad file", obj_name);
172 write_or_die(1, buf, size);
173 free(buf);
174 free(obj_context.path);
175 return 0;
178 struct expand_data {
179 struct object_id oid;
180 enum object_type type;
181 unsigned long size;
182 off_t disk_size;
183 const char *rest;
184 struct object_id delta_base_oid;
187 * If mark_query is true, we do not expand anything, but rather
188 * just mark the object_info with items we wish to query.
190 int mark_query;
193 * Whether to split the input on whitespace before feeding it to
194 * get_sha1; this is decided during the mark_query phase based on
195 * whether we have a %(rest) token in our format.
197 int split_on_whitespace;
200 * After a mark_query run, this object_info is set up to be
201 * passed to sha1_object_info_extended. It will point to the data
202 * elements above, so you can retrieve the response from there.
204 struct object_info info;
207 * This flag will be true if the requested batch format and options
208 * don't require us to call sha1_object_info, which can then be
209 * optimized out.
211 unsigned skip_object_info : 1;
214 static int is_atom(const char *atom, const char *s, int slen)
216 int alen = strlen(atom);
217 return alen == slen && !memcmp(atom, s, alen);
220 static void expand_atom(struct strbuf *sb, const char *atom, int len,
221 void *vdata)
223 struct expand_data *data = vdata;
225 if (is_atom("objectname", atom, len)) {
226 if (!data->mark_query)
227 strbuf_addstr(sb, oid_to_hex(&data->oid));
228 } else if (is_atom("objecttype", atom, len)) {
229 if (data->mark_query)
230 data->info.typep = &data->type;
231 else
232 strbuf_addstr(sb, type_name(data->type));
233 } else if (is_atom("objectsize", atom, len)) {
234 if (data->mark_query)
235 data->info.sizep = &data->size;
236 else
237 strbuf_addf(sb, "%lu", data->size);
238 } else if (is_atom("objectsize:disk", atom, len)) {
239 if (data->mark_query)
240 data->info.disk_sizep = &data->disk_size;
241 else
242 strbuf_addf(sb, "%"PRIuMAX, (uintmax_t)data->disk_size);
243 } else if (is_atom("rest", atom, len)) {
244 if (data->mark_query)
245 data->split_on_whitespace = 1;
246 else if (data->rest)
247 strbuf_addstr(sb, data->rest);
248 } else if (is_atom("deltabase", atom, len)) {
249 if (data->mark_query)
250 data->info.delta_base_sha1 = data->delta_base_oid.hash;
251 else
252 strbuf_addstr(sb,
253 oid_to_hex(&data->delta_base_oid));
254 } else
255 die("unknown format element: %.*s", len, atom);
258 static size_t expand_format(struct strbuf *sb, const char *start, void *data)
260 const char *end;
262 if (*start != '(')
263 return 0;
264 end = strchr(start + 1, ')');
265 if (!end)
266 die("format element '%s' does not end in ')'", start);
268 expand_atom(sb, start + 1, end - start - 1, data);
270 return end - start + 1;
273 static void batch_write(struct batch_options *opt, const void *data, int len)
275 if (opt->buffer_output) {
276 if (fwrite(data, 1, len, stdout) != len)
277 die_errno("unable to write to stdout");
278 } else
279 write_or_die(1, data, len);
282 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
284 const struct object_id *oid = &data->oid;
286 assert(data->info.typep);
288 if (data->type == OBJ_BLOB) {
289 if (opt->buffer_output)
290 fflush(stdout);
291 if (opt->cmdmode) {
292 char *contents;
293 unsigned long size;
295 if (!data->rest)
296 die("missing path for '%s'", oid_to_hex(oid));
298 if (opt->cmdmode == 'w') {
299 if (filter_object(data->rest, 0100644, oid,
300 &contents, &size))
301 die("could not convert '%s' %s",
302 oid_to_hex(oid), data->rest);
303 } else if (opt->cmdmode == 'c') {
304 enum object_type type;
305 if (!textconv_object(data->rest, 0100644, oid,
306 1, &contents, &size))
307 contents = read_sha1_file(oid->hash, &type,
308 &size);
309 if (!contents)
310 die("could not convert '%s' %s",
311 oid_to_hex(oid), data->rest);
312 } else
313 die("BUG: invalid cmdmode: %c", opt->cmdmode);
314 batch_write(opt, contents, size);
315 free(contents);
316 } else if (stream_blob_to_fd(1, oid, NULL, 0) < 0)
317 die("unable to stream %s to stdout", oid_to_hex(oid));
319 else {
320 enum object_type type;
321 unsigned long size;
322 void *contents;
324 contents = read_sha1_file(oid->hash, &type, &size);
325 if (!contents)
326 die("object %s disappeared", oid_to_hex(oid));
327 if (type != data->type)
328 die("object %s changed type!?", oid_to_hex(oid));
329 if (data->info.sizep && size != data->size)
330 die("object %s changed size!?", oid_to_hex(oid));
332 batch_write(opt, contents, size);
333 free(contents);
337 static void batch_object_write(const char *obj_name, struct batch_options *opt,
338 struct expand_data *data)
340 struct strbuf buf = STRBUF_INIT;
342 if (!data->skip_object_info &&
343 sha1_object_info_extended(data->oid.hash, &data->info,
344 OBJECT_INFO_LOOKUP_REPLACE) < 0) {
345 printf("%s missing\n",
346 obj_name ? obj_name : oid_to_hex(&data->oid));
347 fflush(stdout);
348 return;
351 strbuf_expand(&buf, opt->format, expand_format, data);
352 strbuf_addch(&buf, '\n');
353 batch_write(opt, buf.buf, buf.len);
354 strbuf_release(&buf);
356 if (opt->print_contents) {
357 print_object_or_die(opt, data);
358 batch_write(opt, "\n", 1);
362 static void batch_one_object(const char *obj_name, struct batch_options *opt,
363 struct expand_data *data)
365 struct object_context ctx;
366 int flags = opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0;
367 enum follow_symlinks_result result;
369 result = get_oid_with_context(obj_name, flags, &data->oid, &ctx);
370 if (result != FOUND) {
371 switch (result) {
372 case MISSING_OBJECT:
373 printf("%s missing\n", obj_name);
374 break;
375 case DANGLING_SYMLINK:
376 printf("dangling %"PRIuMAX"\n%s\n",
377 (uintmax_t)strlen(obj_name), obj_name);
378 break;
379 case SYMLINK_LOOP:
380 printf("loop %"PRIuMAX"\n%s\n",
381 (uintmax_t)strlen(obj_name), obj_name);
382 break;
383 case NOT_DIR:
384 printf("notdir %"PRIuMAX"\n%s\n",
385 (uintmax_t)strlen(obj_name), obj_name);
386 break;
387 default:
388 die("BUG: unknown get_sha1_with_context result %d\n",
389 result);
390 break;
392 fflush(stdout);
393 return;
396 if (ctx.mode == 0) {
397 printf("symlink %"PRIuMAX"\n%s\n",
398 (uintmax_t)ctx.symlink_path.len,
399 ctx.symlink_path.buf);
400 fflush(stdout);
401 return;
404 batch_object_write(obj_name, opt, data);
407 struct object_cb_data {
408 struct batch_options *opt;
409 struct expand_data *expand;
412 static int batch_object_cb(const struct object_id *oid, void *vdata)
414 struct object_cb_data *data = vdata;
415 oidcpy(&data->expand->oid, oid);
416 batch_object_write(NULL, data->opt, data->expand);
417 return 0;
420 static int batch_loose_object(const struct object_id *oid,
421 const char *path,
422 void *data)
424 oid_array_append(data, oid);
425 return 0;
428 static int batch_packed_object(const struct object_id *oid,
429 struct packed_git *pack,
430 uint32_t pos,
431 void *data)
433 oid_array_append(data, oid);
434 return 0;
437 static int batch_objects(struct batch_options *opt)
439 struct strbuf buf = STRBUF_INIT;
440 struct expand_data data;
441 int save_warning;
442 int retval = 0;
444 if (!opt->format)
445 opt->format = "%(objectname) %(objecttype) %(objectsize)";
448 * Expand once with our special mark_query flag, which will prime the
449 * object_info to be handed to sha1_object_info_extended for each
450 * object.
452 memset(&data, 0, sizeof(data));
453 data.mark_query = 1;
454 strbuf_expand(&buf, opt->format, expand_format, &data);
455 data.mark_query = 0;
456 if (opt->cmdmode)
457 data.split_on_whitespace = 1;
459 if (opt->all_objects) {
460 struct object_info empty = OBJECT_INFO_INIT;
461 if (!memcmp(&data.info, &empty, sizeof(empty)))
462 data.skip_object_info = 1;
466 * If we are printing out the object, then always fill in the type,
467 * since we will want to decide whether or not to stream.
469 if (opt->print_contents)
470 data.info.typep = &data.type;
472 if (opt->all_objects) {
473 struct oid_array sa = OID_ARRAY_INIT;
474 struct object_cb_data cb;
476 for_each_loose_object(batch_loose_object, &sa, 0);
477 for_each_packed_object(batch_packed_object, &sa, 0);
478 if (repository_format_partial_clone)
479 warning("This repository has extensions.partialClone set. Some objects may not be loaded.");
481 cb.opt = opt;
482 cb.expand = &data;
483 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
485 oid_array_clear(&sa);
486 return 0;
490 * We are going to call get_sha1 on a potentially very large number of
491 * objects. In most large cases, these will be actual object sha1s. The
492 * cost to double-check that each one is not also a ref (just so we can
493 * warn) ends up dwarfing the actual cost of the object lookups
494 * themselves. We can work around it by just turning off the warning.
496 save_warning = warn_on_object_refname_ambiguity;
497 warn_on_object_refname_ambiguity = 0;
499 while (strbuf_getline(&buf, stdin) != EOF) {
500 if (data.split_on_whitespace) {
502 * Split at first whitespace, tying off the beginning
503 * of the string and saving the remainder (or NULL) in
504 * data.rest.
506 char *p = strpbrk(buf.buf, " \t");
507 if (p) {
508 while (*p && strchr(" \t", *p))
509 *p++ = '\0';
511 data.rest = p;
514 batch_one_object(buf.buf, opt, &data);
517 strbuf_release(&buf);
518 warn_on_object_refname_ambiguity = save_warning;
519 return retval;
522 static const char * const cat_file_usage[] = {
523 N_("git cat-file (-t [--allow-unknown-type] | -s [--allow-unknown-type] | -e | -p | <type> | --textconv | --filters) [--path=<path>] <object>"),
524 N_("git cat-file (--batch | --batch-check) [--follow-symlinks] [--textconv | --filters]"),
525 NULL
528 static int git_cat_file_config(const char *var, const char *value, void *cb)
530 if (userdiff_config(var, value) < 0)
531 return -1;
533 return git_default_config(var, value, cb);
536 static int batch_option_callback(const struct option *opt,
537 const char *arg,
538 int unset)
540 struct batch_options *bo = opt->value;
542 if (bo->enabled) {
543 return 1;
546 bo->enabled = 1;
547 bo->print_contents = !strcmp(opt->long_name, "batch");
548 bo->format = arg;
550 return 0;
553 int cmd_cat_file(int argc, const char **argv, const char *prefix)
555 int opt = 0;
556 const char *exp_type = NULL, *obj_name = NULL;
557 struct batch_options batch = {0};
558 int unknown_type = 0;
560 const struct option options[] = {
561 OPT_GROUP(N_("<type> can be one of: blob, tree, commit, tag")),
562 OPT_CMDMODE('t', NULL, &opt, N_("show object type"), 't'),
563 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
564 OPT_CMDMODE('e', NULL, &opt,
565 N_("exit with zero when there's no error"), 'e'),
566 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print object's content"), 'p'),
567 OPT_CMDMODE(0, "textconv", &opt,
568 N_("for blob objects, run textconv on object's content"), 'c'),
569 OPT_CMDMODE(0, "filters", &opt,
570 N_("for blob objects, run filters on object's content"), 'w'),
571 OPT_STRING(0, "path", &force_path, N_("blob"),
572 N_("use a specific path for --textconv/--filters")),
573 OPT_BOOL(0, "allow-unknown-type", &unknown_type,
574 N_("allow -s and -t to work with broken/corrupt objects")),
575 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
576 { OPTION_CALLBACK, 0, "batch", &batch, "format",
577 N_("show info and content of objects fed from the standard input"),
578 PARSE_OPT_OPTARG, batch_option_callback },
579 { OPTION_CALLBACK, 0, "batch-check", &batch, "format",
580 N_("show info about objects fed from the standard input"),
581 PARSE_OPT_OPTARG, batch_option_callback },
582 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
583 N_("follow in-tree symlinks (used with --batch or --batch-check)")),
584 OPT_BOOL(0, "batch-all-objects", &batch.all_objects,
585 N_("show all objects with --batch or --batch-check")),
586 OPT_END()
589 git_config(git_cat_file_config, NULL);
591 batch.buffer_output = -1;
592 argc = parse_options(argc, argv, prefix, options, cat_file_usage, 0);
594 if (opt) {
595 if (batch.enabled && (opt == 'c' || opt == 'w'))
596 batch.cmdmode = opt;
597 else if (argc == 1)
598 obj_name = argv[0];
599 else
600 usage_with_options(cat_file_usage, options);
602 if (!opt && !batch.enabled) {
603 if (argc == 2) {
604 exp_type = argv[0];
605 obj_name = argv[1];
606 } else
607 usage_with_options(cat_file_usage, options);
609 if (batch.enabled) {
610 if (batch.cmdmode != opt || argc)
611 usage_with_options(cat_file_usage, options);
612 if (batch.cmdmode && batch.all_objects)
613 die("--batch-all-objects cannot be combined with "
614 "--textconv nor with --filters");
617 if ((batch.follow_symlinks || batch.all_objects) && !batch.enabled) {
618 usage_with_options(cat_file_usage, options);
621 if (force_path && opt != 'c' && opt != 'w') {
622 error("--path=<path> needs --textconv or --filters");
623 usage_with_options(cat_file_usage, options);
626 if (force_path && batch.enabled) {
627 error("--path=<path> incompatible with --batch");
628 usage_with_options(cat_file_usage, options);
631 if (batch.buffer_output < 0)
632 batch.buffer_output = batch.all_objects;
634 if (batch.enabled)
635 return batch_objects(&batch);
637 if (unknown_type && opt != 't' && opt != 's')
638 die("git cat-file --allow-unknown-type: use with -s or -t");
639 return cat_one_file(opt, exp_type, obj_name, unknown_type);