midx: fix bug that skips midx with alternates
[git/git.git] / object-store.h
1 #ifndef OBJECT_STORE_H
2 #define OBJECT_STORE_H
3
4 #include "cache.h"
5 #include "oidmap.h"
6 #include "list.h"
7 #include "sha1-array.h"
8 #include "strbuf.h"
9
10 struct alternate_object_database {
11 struct alternate_object_database *next;
12
13 /* see alt_scratch_buf() */
14 struct strbuf scratch;
15 size_t base_len;
16
17 /*
18 * Used to store the results of readdir(3) calls when searching
19 * for unique abbreviated hashes. This cache is never
20 * invalidated, thus it's racy and not necessarily accurate.
21 * That's fine for its purpose; don't use it for tasks requiring
22 * greater accuracy!
23 */
24 char loose_objects_subdir_seen[256];
25 struct oid_array loose_objects_cache;
26
27 /*
28 * Path to the alternative object store. If this is a relative path,
29 * it is relative to the current working directory.
30 */
31 char path[FLEX_ARRAY];
32 };
33 void prepare_alt_odb(struct repository *r);
34 char *compute_alternate_path(const char *path, struct strbuf *err);
35 typedef int alt_odb_fn(struct alternate_object_database *, void *);
36 int foreach_alt_odb(alt_odb_fn, void*);
37
38 /*
39 * Allocate a "struct alternate_object_database" but do _not_ actually
40 * add it to the list of alternates.
41 */
42 struct alternate_object_database *alloc_alt_odb(const char *dir);
43
44 /*
45 * Add the directory to the on-disk alternates file; the new entry will also
46 * take effect in the current process.
47 */
48 void add_to_alternates_file(const char *dir);
49
50 /*
51 * Add the directory to the in-memory list of alternates (along with any
52 * recursive alternates it points to), but do not modify the on-disk alternates
53 * file.
54 */
55 void add_to_alternates_memory(const char *dir);
56
57 /*
58 * Returns a scratch strbuf pre-filled with the alternate object directory,
59 * including a trailing slash, which can be used to access paths in the
60 * alternate. Always use this over direct access to alt->scratch, as it
61 * cleans up any previous use of the scratch buffer.
62 */
63 struct strbuf *alt_scratch_buf(struct alternate_object_database *alt);
64
65 struct packed_git {
66 struct packed_git *next;
67 struct list_head mru;
68 struct pack_window *windows;
69 off_t pack_size;
70 const void *index_data;
71 size_t index_size;
72 uint32_t num_objects;
73 uint32_t num_bad_objects;
74 unsigned char *bad_object_sha1;
75 int index_version;
76 time_t mtime;
77 int pack_fd;
78 int index; /* for builtin/pack-objects.c */
79 unsigned pack_local:1,
80 pack_keep:1,
81 pack_keep_in_core:1,
82 freshened:1,
83 do_not_close:1,
84 pack_promisor:1;
85 unsigned char sha1[20];
86 struct revindex_entry *revindex;
87 /* something like ".git/objects/pack/xxxxx.pack" */
88 char pack_name[FLEX_ARRAY]; /* more */
89 };
90
91 struct multi_pack_index;
92
93 struct raw_object_store {
94 /*
95 * Path to the repository's object store.
96 * Cannot be NULL after initialization.
97 */
98 char *objectdir;
99
100 /* Path to extra alternate object database if not NULL */
101 char *alternate_db;
102
103 struct alternate_object_database *alt_odb_list;
104 struct alternate_object_database **alt_odb_tail;
105
106 /*
107 * Objects that should be substituted by other objects
108 * (see git-replace(1)).
109 */
110 struct oidmap *replace_map;
111
112 struct commit_graph *commit_graph;
113 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
114
115 /*
116 * private data
117 *
118 * should only be accessed directly by packfile.c and midx.c
119 */
120 struct multi_pack_index *multi_pack_index;
121
122 /*
123 * private data
124 *
125 * should only be accessed directly by packfile.c
126 */
127
128 struct packed_git *packed_git;
129 /* A most-recently-used ordered version of the packed_git list. */
130 struct list_head packed_git_mru;
131
132 /*
133 * A fast, rough count of the number of objects in the repository.
134 * These two fields are not meant for direct access. Use
135 * approximate_object_count() instead.
136 */
137 unsigned long approximate_object_count;
138 unsigned approximate_object_count_valid : 1;
139
140 /*
141 * Whether packed_git has already been populated with this repository's
142 * packs.
143 */
144 unsigned packed_git_initialized : 1;
145 };
146
147 struct raw_object_store *raw_object_store_new(void);
148 void raw_object_store_clear(struct raw_object_store *o);
149
150 /*
151 * Put in `buf` the name of the file in the local object database that
152 * would be used to store a loose object with the specified sha1.
153 */
154 void sha1_file_name(struct repository *r, struct strbuf *buf, const unsigned char *sha1);
155
156 void *map_sha1_file(struct repository *r, const unsigned char *sha1, unsigned long *size);
157
158 extern void *read_object_file_extended(const struct object_id *oid,
159 enum object_type *type,
160 unsigned long *size, int lookup_replace);
161 static inline void *read_object_file(const struct object_id *oid, enum object_type *type, unsigned long *size)
162 {
163 return read_object_file_extended(oid, type, size, 1);
164 }
165
166 /* Read and unpack an object file into memory, write memory to an object file */
167 int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
168
169 extern int hash_object_file(const void *buf, unsigned long len,
170 const char *type, struct object_id *oid);
171
172 extern int write_object_file(const void *buf, unsigned long len,
173 const char *type, struct object_id *oid);
174
175 extern int hash_object_file_literally(const void *buf, unsigned long len,
176 const char *type, struct object_id *oid,
177 unsigned flags);
178
179 extern int pretend_object_file(void *, unsigned long, enum object_type,
180 struct object_id *oid);
181
182 extern int force_object_loose(const struct object_id *oid, time_t mtime);
183
184 /*
185 * Open the loose object at path, check its hash, and return the contents,
186 * type, and size. If the object is a blob, then "contents" may return NULL,
187 * to allow streaming of large blobs.
188 *
189 * Returns 0 on success, negative on error (details may be written to stderr).
190 */
191 int read_loose_object(const char *path,
192 const struct object_id *expected_oid,
193 enum object_type *type,
194 unsigned long *size,
195 void **contents);
196
197 /*
198 * Convenience for sha1_object_info_extended() with a NULL struct
199 * object_info. OBJECT_INFO_SKIP_CACHED is automatically set; pass
200 * nonzero flags to also set other flags.
201 */
202 extern int has_sha1_file_with_flags(const unsigned char *sha1, int flags);
203 static inline int has_sha1_file(const unsigned char *sha1)
204 {
205 return has_sha1_file_with_flags(sha1, 0);
206 }
207
208 /* Same as the above, except for struct object_id. */
209 extern int has_object_file(const struct object_id *oid);
210 extern int has_object_file_with_flags(const struct object_id *oid, int flags);
211
212 /*
213 * Return true iff an alternate object database has a loose object
214 * with the specified name. This function does not respect replace
215 * references.
216 */
217 extern int has_loose_object_nonlocal(const struct object_id *);
218
219 extern void assert_oid_type(const struct object_id *oid, enum object_type expect);
220
221 struct object_info {
222 /* Request */
223 enum object_type *typep;
224 unsigned long *sizep;
225 off_t *disk_sizep;
226 unsigned char *delta_base_sha1;
227 struct strbuf *type_name;
228 void **contentp;
229
230 /* Response */
231 enum {
232 OI_CACHED,
233 OI_LOOSE,
234 OI_PACKED,
235 OI_DBCACHED
236 } whence;
237 union {
238 /*
239 * struct {
240 * ... Nothing to expose in this case
241 * } cached;
242 * struct {
243 * ... Nothing to expose in this case
244 * } loose;
245 */
246 struct {
247 struct packed_git *pack;
248 off_t offset;
249 unsigned int is_delta;
250 } packed;
251 } u;
252 };
253
254 /*
255 * Initializer for a "struct object_info" that wants no items. You may
256 * also memset() the memory to all-zeroes.
257 */
258 #define OBJECT_INFO_INIT {NULL}
259
260 /* Invoke lookup_replace_object() on the given hash */
261 #define OBJECT_INFO_LOOKUP_REPLACE 1
262 /* Allow reading from a loose object file of unknown/bogus type */
263 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
264 /* Do not check cached storage */
265 #define OBJECT_INFO_SKIP_CACHED 4
266 /* Do not retry packed storage after checking packed and loose storage */
267 #define OBJECT_INFO_QUICK 8
268 /* Do not check loose object */
269 #define OBJECT_INFO_IGNORE_LOOSE 16
270
271 int oid_object_info_extended(struct repository *r,
272 const struct object_id *,
273 struct object_info *, unsigned flags);
274
275 /*
276 * Iterate over the files in the loose-object parts of the object
277 * directory "path", triggering the following callbacks:
278 *
279 * - loose_object is called for each loose object we find.
280 *
281 * - loose_cruft is called for any files that do not appear to be
282 * loose objects. Note that we only look in the loose object
283 * directories "objects/[0-9a-f]{2}/", so we will not report
284 * "objects/foobar" as cruft.
285 *
286 * - loose_subdir is called for each top-level hashed subdirectory
287 * of the object directory (e.g., "$OBJDIR/f0"). It is called
288 * after the objects in the directory are processed.
289 *
290 * Any callback that is NULL will be ignored. Callbacks returning non-zero
291 * will end the iteration.
292 *
293 * In the "buf" variant, "path" is a strbuf which will also be used as a
294 * scratch buffer, but restored to its original contents before
295 * the function returns.
296 */
297 typedef int each_loose_object_fn(const struct object_id *oid,
298 const char *path,
299 void *data);
300 typedef int each_loose_cruft_fn(const char *basename,
301 const char *path,
302 void *data);
303 typedef int each_loose_subdir_fn(unsigned int nr,
304 const char *path,
305 void *data);
306 int for_each_file_in_obj_subdir(unsigned int subdir_nr,
307 struct strbuf *path,
308 each_loose_object_fn obj_cb,
309 each_loose_cruft_fn cruft_cb,
310 each_loose_subdir_fn subdir_cb,
311 void *data);
312 int for_each_loose_file_in_objdir(const char *path,
313 each_loose_object_fn obj_cb,
314 each_loose_cruft_fn cruft_cb,
315 each_loose_subdir_fn subdir_cb,
316 void *data);
317 int for_each_loose_file_in_objdir_buf(struct strbuf *path,
318 each_loose_object_fn obj_cb,
319 each_loose_cruft_fn cruft_cb,
320 each_loose_subdir_fn subdir_cb,
321 void *data);
322
323 /* Flags for for_each_*_object() below. */
324 enum for_each_object_flags {
325 /* Iterate only over local objects, not alternates. */
326 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
327
328 /* Only iterate over packs obtained from the promisor remote. */
329 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
330
331 /*
332 * Visit objects within a pack in packfile order rather than .idx order
333 */
334 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
335 };
336
337 /*
338 * Iterate over all accessible loose objects without respect to
339 * reachability. By default, this includes both local and alternate objects.
340 * The order in which objects are visited is unspecified.
341 *
342 * Any flags specific to packs are ignored.
343 */
344 int for_each_loose_object(each_loose_object_fn, void *,
345 enum for_each_object_flags flags);
346
347 /*
348 * Iterate over all accessible packed objects without respect to reachability.
349 * By default, this includes both local and alternate packs.
350 *
351 * Note that some objects may appear twice if they are found in multiple packs.
352 * Each pack is visited in an unspecified order. By default, objects within a
353 * pack are visited in pack-idx order (i.e., sorted by oid).
354 */
355 typedef int each_packed_object_fn(const struct object_id *oid,
356 struct packed_git *pack,
357 uint32_t pos,
358 void *data);
359 int for_each_object_in_pack(struct packed_git *p,
360 each_packed_object_fn, void *data,
361 enum for_each_object_flags flags);
362 int for_each_packed_object(each_packed_object_fn, void *,
363 enum for_each_object_flags flags);
364
365 #endif /* OBJECT_STORE_H */