object-store.hon commit Merge branch 'ds/push-sparse-tree-walk' (5fda343)
   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
  10struct object_directory {
  11        struct object_directory *next;
  12
  13        /*
  14         * Used to store the results of readdir(3) calls when we are OK
  15         * sacrificing accuracy due to races for speed. That includes
  16         * object existence with OBJECT_INFO_QUICK, as well as
  17         * our search for unique abbreviated hashes. Don't use it for tasks
  18         * requiring greater accuracy!
  19         *
  20         * Be sure to call odb_load_loose_cache() before using.
  21         */
  22        char loose_objects_subdir_seen[256];
  23        struct oid_array loose_objects_cache[256];
  24
  25        /*
  26         * Path to the alternative object store. If this is a relative path,
  27         * it is relative to the current working directory.
  28         */
  29        char *path;
  30};
  31
  32void prepare_alt_odb(struct repository *r);
  33char *compute_alternate_path(const char *path, struct strbuf *err);
  34typedef int alt_odb_fn(struct object_directory *, void *);
  35int foreach_alt_odb(alt_odb_fn, void*);
  36
  37/*
  38 * Add the directory to the on-disk alternates file; the new entry will also
  39 * take effect in the current process.
  40 */
  41void add_to_alternates_file(const char *dir);
  42
  43/*
  44 * Add the directory to the in-memory list of alternates (along with any
  45 * recursive alternates it points to), but do not modify the on-disk alternates
  46 * file.
  47 */
  48void add_to_alternates_memory(const char *dir);
  49
  50/*
  51 * Populate and return the loose object cache array corresponding to the
  52 * given object ID.
  53 */
  54struct oid_array *odb_loose_cache(struct object_directory *odb,
  55                                  const struct object_id *oid);
  56
  57/* Empty the loose object cache for the specified object directory. */
  58void odb_clear_loose_cache(struct object_directory *odb);
  59
  60struct packed_git {
  61        struct packed_git *next;
  62        struct list_head mru;
  63        struct pack_window *windows;
  64        off_t pack_size;
  65        const void *index_data;
  66        size_t index_size;
  67        uint32_t num_objects;
  68        uint32_t num_bad_objects;
  69        unsigned char *bad_object_sha1;
  70        int index_version;
  71        time_t mtime;
  72        int pack_fd;
  73        int index;              /* for builtin/pack-objects.c */
  74        unsigned pack_local:1,
  75                 pack_keep:1,
  76                 pack_keep_in_core:1,
  77                 freshened:1,
  78                 do_not_close:1,
  79                 pack_promisor:1;
  80        unsigned char sha1[20];
  81        struct revindex_entry *revindex;
  82        /* something like ".git/objects/pack/xxxxx.pack" */
  83        char pack_name[FLEX_ARRAY]; /* more */
  84};
  85
  86struct multi_pack_index;
  87
  88struct raw_object_store {
  89        /*
  90         * Set of all object directories; the main directory is first (and
  91         * cannot be NULL after initialization). Subsequent directories are
  92         * alternates.
  93         */
  94        struct object_directory *odb;
  95        struct object_directory **odb_tail;
  96        int loaded_alternates;
  97
  98        /*
  99         * A list of alternate object directories loaded from the environment;
 100         * this should not generally need to be accessed directly, but will
 101         * populate the "odb" list when prepare_alt_odb() is run.
 102         */
 103        char *alternate_db;
 104
 105        /*
 106         * Objects that should be substituted by other objects
 107         * (see git-replace(1)).
 108         */
 109        struct oidmap *replace_map;
 110
 111        struct commit_graph *commit_graph;
 112        unsigned commit_graph_attempted : 1; /* if loading has been attempted */
 113
 114        /*
 115         * private data
 116         *
 117         * should only be accessed directly by packfile.c and midx.c
 118         */
 119        struct multi_pack_index *multi_pack_index;
 120
 121        /*
 122         * private data
 123         *
 124         * should only be accessed directly by packfile.c
 125         */
 126
 127        struct packed_git *packed_git;
 128        /* A most-recently-used ordered version of the packed_git list. */
 129        struct list_head packed_git_mru;
 130
 131        /*
 132         * A linked list containing all packfiles, starting with those
 133         * contained in the multi_pack_index.
 134         */
 135        struct packed_git *all_packs;
 136
 137        /*
 138         * A fast, rough count of the number of objects in the repository.
 139         * These two fields are not meant for direct access. Use
 140         * approximate_object_count() instead.
 141         */
 142        unsigned long approximate_object_count;
 143        unsigned approximate_object_count_valid : 1;
 144
 145        /*
 146         * Whether packed_git has already been populated with this repository's
 147         * packs.
 148         */
 149        unsigned packed_git_initialized : 1;
 150};
 151
 152struct raw_object_store *raw_object_store_new(void);
 153void raw_object_store_clear(struct raw_object_store *o);
 154
 155/*
 156 * Put in `buf` the name of the file in the local object database that
 157 * would be used to store a loose object with the specified sha1.
 158 */
 159const char *loose_object_path(struct repository *r, struct strbuf *buf, const unsigned char *sha1);
 160
 161void *map_sha1_file(struct repository *r, const unsigned char *sha1, unsigned long *size);
 162
 163extern void *read_object_file_extended(struct repository *r,
 164                                       const struct object_id *oid,
 165                                       enum object_type *type,
 166                                       unsigned long *size, int lookup_replace);
 167static inline void *repo_read_object_file(struct repository *r,
 168                                          const struct object_id *oid,
 169                                          enum object_type *type,
 170                                          unsigned long *size)
 171{
 172        return read_object_file_extended(r, oid, type, size, 1);
 173}
 174#ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
 175#define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
 176#endif
 177
 178/* Read and unpack an object file into memory, write memory to an object file */
 179int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
 180
 181extern int hash_object_file(const void *buf, unsigned long len,
 182                            const char *type, struct object_id *oid);
 183
 184extern int write_object_file(const void *buf, unsigned long len,
 185                             const char *type, struct object_id *oid);
 186
 187extern int hash_object_file_literally(const void *buf, unsigned long len,
 188                                      const char *type, struct object_id *oid,
 189                                      unsigned flags);
 190
 191extern int pretend_object_file(void *, unsigned long, enum object_type,
 192                               struct object_id *oid);
 193
 194extern int force_object_loose(const struct object_id *oid, time_t mtime);
 195
 196/*
 197 * Open the loose object at path, check its hash, and return the contents,
 198 * type, and size. If the object is a blob, then "contents" may return NULL,
 199 * to allow streaming of large blobs.
 200 *
 201 * Returns 0 on success, negative on error (details may be written to stderr).
 202 */
 203int read_loose_object(const char *path,
 204                      const struct object_id *expected_oid,
 205                      enum object_type *type,
 206                      unsigned long *size,
 207                      void **contents);
 208
 209/*
 210 * Convenience for sha1_object_info_extended() with a NULL struct
 211 * object_info. OBJECT_INFO_SKIP_CACHED is automatically set; pass
 212 * nonzero flags to also set other flags.
 213 */
 214int repo_has_sha1_file_with_flags(struct repository *r,
 215                                  const unsigned char *sha1, int flags);
 216static inline int repo_has_sha1_file(struct repository *r,
 217                                     const unsigned char *sha1)
 218{
 219        return repo_has_sha1_file_with_flags(r, sha1, 0);
 220}
 221
 222#ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
 223#define has_sha1_file_with_flags(sha1, flags) repo_has_sha1_file_with_flags(the_repository, sha1, flags)
 224#define has_sha1_file(sha1) repo_has_sha1_file(the_repository, sha1)
 225#endif
 226
 227/* Same as the above, except for struct object_id. */
 228int repo_has_object_file(struct repository *r, const struct object_id *oid);
 229int repo_has_object_file_with_flags(struct repository *r,
 230                                    const struct object_id *oid, int flags);
 231#ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
 232#define has_object_file(oid) repo_has_object_file(the_repository, oid)
 233#define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
 234#endif
 235
 236/*
 237 * Return true iff an alternate object database has a loose object
 238 * with the specified name.  This function does not respect replace
 239 * references.
 240 */
 241extern int has_loose_object_nonlocal(const struct object_id *);
 242
 243extern void assert_oid_type(const struct object_id *oid, enum object_type expect);
 244
 245struct object_info {
 246        /* Request */
 247        enum object_type *typep;
 248        unsigned long *sizep;
 249        off_t *disk_sizep;
 250        unsigned char *delta_base_sha1;
 251        struct strbuf *type_name;
 252        void **contentp;
 253
 254        /* Response */
 255        enum {
 256                OI_CACHED,
 257                OI_LOOSE,
 258                OI_PACKED,
 259                OI_DBCACHED
 260        } whence;
 261        union {
 262                /*
 263                 * struct {
 264                 *      ... Nothing to expose in this case
 265                 * } cached;
 266                 * struct {
 267                 *      ... Nothing to expose in this case
 268                 * } loose;
 269                 */
 270                struct {
 271                        struct packed_git *pack;
 272                        off_t offset;
 273                        unsigned int is_delta;
 274                } packed;
 275        } u;
 276};
 277
 278/*
 279 * Initializer for a "struct object_info" that wants no items. You may
 280 * also memset() the memory to all-zeroes.
 281 */
 282#define OBJECT_INFO_INIT {NULL}
 283
 284/* Invoke lookup_replace_object() on the given hash */
 285#define OBJECT_INFO_LOOKUP_REPLACE 1
 286/* Allow reading from a loose object file of unknown/bogus type */
 287#define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
 288/* Do not check cached storage */
 289#define OBJECT_INFO_SKIP_CACHED 4
 290/* Do not retry packed storage after checking packed and loose storage */
 291#define OBJECT_INFO_QUICK 8
 292/* Do not check loose object */
 293#define OBJECT_INFO_IGNORE_LOOSE 16
 294
 295int oid_object_info_extended(struct repository *r,
 296                             const struct object_id *,
 297                             struct object_info *, unsigned flags);
 298
 299/*
 300 * Iterate over the files in the loose-object parts of the object
 301 * directory "path", triggering the following callbacks:
 302 *
 303 *  - loose_object is called for each loose object we find.
 304 *
 305 *  - loose_cruft is called for any files that do not appear to be
 306 *    loose objects. Note that we only look in the loose object
 307 *    directories "objects/[0-9a-f]{2}/", so we will not report
 308 *    "objects/foobar" as cruft.
 309 *
 310 *  - loose_subdir is called for each top-level hashed subdirectory
 311 *    of the object directory (e.g., "$OBJDIR/f0"). It is called
 312 *    after the objects in the directory are processed.
 313 *
 314 * Any callback that is NULL will be ignored. Callbacks returning non-zero
 315 * will end the iteration.
 316 *
 317 * In the "buf" variant, "path" is a strbuf which will also be used as a
 318 * scratch buffer, but restored to its original contents before
 319 * the function returns.
 320 */
 321typedef int each_loose_object_fn(const struct object_id *oid,
 322                                 const char *path,
 323                                 void *data);
 324typedef int each_loose_cruft_fn(const char *basename,
 325                                const char *path,
 326                                void *data);
 327typedef int each_loose_subdir_fn(unsigned int nr,
 328                                 const char *path,
 329                                 void *data);
 330int for_each_file_in_obj_subdir(unsigned int subdir_nr,
 331                                struct strbuf *path,
 332                                each_loose_object_fn obj_cb,
 333                                each_loose_cruft_fn cruft_cb,
 334                                each_loose_subdir_fn subdir_cb,
 335                                void *data);
 336int for_each_loose_file_in_objdir(const char *path,
 337                                  each_loose_object_fn obj_cb,
 338                                  each_loose_cruft_fn cruft_cb,
 339                                  each_loose_subdir_fn subdir_cb,
 340                                  void *data);
 341int for_each_loose_file_in_objdir_buf(struct strbuf *path,
 342                                      each_loose_object_fn obj_cb,
 343                                      each_loose_cruft_fn cruft_cb,
 344                                      each_loose_subdir_fn subdir_cb,
 345                                      void *data);
 346
 347/* Flags for for_each_*_object() below. */
 348enum for_each_object_flags {
 349        /* Iterate only over local objects, not alternates. */
 350        FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
 351
 352        /* Only iterate over packs obtained from the promisor remote. */
 353        FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
 354
 355        /*
 356         * Visit objects within a pack in packfile order rather than .idx order
 357         */
 358        FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
 359};
 360
 361/*
 362 * Iterate over all accessible loose objects without respect to
 363 * reachability. By default, this includes both local and alternate objects.
 364 * The order in which objects are visited is unspecified.
 365 *
 366 * Any flags specific to packs are ignored.
 367 */
 368int for_each_loose_object(each_loose_object_fn, void *,
 369                          enum for_each_object_flags flags);
 370
 371/*
 372 * Iterate over all accessible packed objects without respect to reachability.
 373 * By default, this includes both local and alternate packs.
 374 *
 375 * Note that some objects may appear twice if they are found in multiple packs.
 376 * Each pack is visited in an unspecified order. By default, objects within a
 377 * pack are visited in pack-idx order (i.e., sorted by oid).
 378 */
 379typedef int each_packed_object_fn(const struct object_id *oid,
 380                                  struct packed_git *pack,
 381                                  uint32_t pos,
 382                                  void *data);
 383int for_each_object_in_pack(struct packed_git *p,
 384                            each_packed_object_fn, void *data,
 385                            enum for_each_object_flags flags);
 386int for_each_packed_object(each_packed_object_fn, void *,
 387                           enum for_each_object_flags flags);
 388
 389#endif /* OBJECT_STORE_H */