Struct FileObjectDatabase 

pub struct FileObjectDatabase { /* private fields */ }

Implementations

impl FileObjectDatabase

pub fn object_format(&self) -> ObjectFormat

The object-id format (hash algorithm) this database was opened with.

pub fn objects_dir(&self) -> &Path

The repository object directory this database reads from.

pub fn new(objects_dir: impl Into<PathBuf>, format: ObjectFormat) -> Self

pub fn from_git_dir(git_dir: impl AsRef<Path>, format: ObjectFormat) -> Self

pub fn with_promisor_remote_present(self, present: bool) -> Self

Declare whether the owning repository has a promisor remote configured. Only when this holds does ObjectReader::is_promised_object treat objects in .promisor packs (and their transitive references) as promised — matching git’s is_promisor_object, which is gated on repo_has_promisor_remote(). Callers that know the repo config (e.g. the fsck driver) opt in; readers built without config keep the safe default of false, so a stray .promisor sidecar never silently excuses a genuinely missing object.

pub fn refresh_read_cache(&self)

Drop cached pack registries, indexes, and decoded objects so the next read sees packs/objects installed after this handle was created (e.g. after fetch or install_pack). Long-lived [Repository] sessions call this via the owning repository’s refresh_objects hook.

pub fn loose(&self) -> &LooseObjectStore

pub fn presence_checker(&self) -> ObjectPresenceChecker

pub fn install_pack(&self, pack: &PackWrite) -> Result<PackInstallResult>

pub fn write_blob_as_pack( &self, oid: ObjectId, object: &EncodedObject, compression_level: u32, ) -> Result<ObjectId>

pub fn write_blobs_as_pack( &self, objects: &[(ObjectId, EncodedObject)], compression_level: u32, ) -> Result<()>

pub fn install_pack_with_options( &self, pack: &PackWrite, options: RawPackInstallOptions, ) -> Result<PackInstallResult>

pub fn install_written_pack( &self, pack: &PackWrite, ) -> Result<PackInstallResult>

Install a pack that was produced in this process by PackFile::write_packed.

Unlike [Self::install_raw_pack_with_options], this does not re-inflate every pack entry to rebuild the index. It validates the generated pack trailer and generated index against the writer’s object ids, CRCs, and offsets, then writes those bytes directly. Use the raw installer for arbitrary pack bytes received from an untrusted transport.

pub fn install_written_pack_with_options( &self, pack: &PackWrite, options: RawPackInstallOptions, ) -> Result<PackInstallResult>

pub fn install_raw_pack_from_reader<R>( &self, reader: &mut R, ) -> Result<PackInstallResult>

where R: Read,

pub fn begin_raw_pack_install( &self, expected_pack_id: ObjectId, expected_pack_size: u64, ) -> Result<RawPackStreamingInstall>

pub fn begin_raw_pack_install_with_options( &self, expected_pack_id: ObjectId, expected_pack_size: u64, options: RawPackInstallOptions, ) -> Result<RawPackStreamingInstall>

pub fn install_raw_pack_from_reader_with_options<R>( &self, reader: &mut R, options: RawPackInstallOptions, ) -> Result<PackInstallResult>

where R: Read,

pub fn contains(&self, oid: &ObjectId) -> Result<bool>

pub fn object_ids(&self) -> Result<Vec<ObjectId>>

pub fn object_storage_info( &self, oid: &ObjectId, ) -> Result<Option<ObjectStorageInfo>>

pub fn resolve_prefix(&self, prefix: &str) -> Result<ObjectPrefixResolution>

pub fn object_ids_with_prefix(&self, prefix: &str) -> Result<Vec<ObjectId>>

pub fn read_object_header( &self, oid: &ObjectId, ) -> Result<Option<(ObjectType, u64)>>

The object type and content size of oid without decoding its full body — git’s cat-file --batch-check fast path. Tries the decoded-object cache, then loose storage (inflating only the framing header), then packs (reading the entry header and, for deltas, only the delta’s leading varints), then alternates. Returns Ok(None) if the object is not present.

Unlike ObjectReader::read_object, this never materializes the body, so it stays cheap on huge blobs and deep delta chains. It does not populate the decoded-object cache (nothing is decoded).

Trait Implementations

impl Clone for FileObjectDatabase

fn clone(&self) -> FileObjectDatabase

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for FileObjectDatabase

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl ObjectReader for FileObjectDatabase

fn is_promised_object(&self, oid: &ObjectId) -> bool

True when oid is covered by a promisor pack. Partial clones are allowed to omit promised objects until a later on-demand fetch hydrates them; ordinary readers keep the default “no promised objects”.

fn has_shallow_grafts(&self) -> bool

Whether this reader has any shallow/graft boundaries at all. Walkers can use this to choose dense graph-only traversal when no boundary can cut parent edges.

fn is_shallow_graft(&self, oid: &ObjectId) -> bool

Graft-points seam (shallow clones today, replace refs/grafts later): true when history is cut at oid, so every walk must treat the commit as parentless even though its raw body still names parents.

fn read_object(&self, oid: &ObjectId) -> Result<Arc<EncodedObject>>

impl ObjectWriter for FileObjectDatabase

fn write_object(&self, object: EncodedObject) -> Result<ObjectId>

Write object, returning its id. Takes &self: every implementation’s write state (in-memory map, loose-object cache) is behind interior mutability, so a single handle can interleave reads and writes without a &mut borrow. This lets the merge engine read and write through one db instead of opening a second read-only handle that re-warms the caches.

impl RawPackInstaller for FileObjectDatabase

fn install_raw_pack_from_reader<R>( &self, reader: &mut R, ) -> Result<RawPackInstallResult>

where R: Read,