Skip to main content

ObjectStore

grit_lib::objects::odb

Trait ObjectStore 

Source 
pub trait ObjectStore:
    Send
    + Sync
    + Debug {
    // Required methods
    fn hash_algo(&self) -> HashAlgo;
    fn read(&self, oid: &ObjectId) -> Result<Option<Object>>;
    fn read_info(&self, oid: &ObjectId) -> Result<Option<ObjectInfo>>;
    fn for_each_object(
        &self,
        f: &mut dyn FnMut(&ObjectId) -> ControlFlow<()>,
    ) -> Result<()>;

    // Provided methods
    fn read_into(
        &self,
        oid: &ObjectId,
        buf: &mut Vec<u8>,
    ) -> Result<Option<ObjectKind>> { ... }
    fn contains(&self, oid: &ObjectId) -> Result<bool> { ... }
    fn contains_many(&self, oids: &[ObjectId]) -> Result<Vec<bool>> { ... }
    fn open_stream(&self, oid: &ObjectId) -> Result<Option<ObjectStream<'_>>> { ... }
    fn lookup_prefix(
        &self,
        prefix: &str,
        limit: usize,
        out: &mut Vec<ObjectId>,
    ) -> Result<()> { ... }
    fn refresh(&self) -> Result<bool> { ... }
}
Expand description

Read-only object storage: lookup, metadata, streaming, and enumeration.

Implementations must be safe to share across threads (Send + Sync). Missing objects are reported as Ok(None), not as errors.

Required Methods§

Source

fn hash_algo(&self) -> HashAlgo

Hash algorithm for objects written through this store.

Source

fn read(&self, oid: &ObjectId) -> Result<Option<Object>>

Load the full object, if present.

§Errors

Returns I/O or corruption errors from the backend; never Error::ObjectNotFound.

Source

fn read_info(&self, oid: &ObjectId) -> Result<Option<ObjectInfo>>

Object kind and size without loading the full payload.

§Errors

Same as Self::read.

Source

fn for_each_object( &self, f: &mut dyn FnMut(&ObjectId) -> ControlFlow<()>, ) -> Result<()>

Invoke f for every object id in this store until f returns ControlFlow::Break.

§Errors

Propagates backend enumeration failures.

Provided Methods§

Source

fn read_into( &self, oid: &ObjectId, buf: &mut Vec<u8>, ) -> Result<Option<ObjectKind>>

Load object payload into buf, reusing its capacity when possible.

Default implementation delegates to Self::read and copies into buf. Native backends should inflate or resolve deltas directly into buf.

§Errors

Same as Self::read.

Source

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

Whether oid is stored in this backend (without reading the payload).

§Errors

Propagates backend failures other than “not found”.

Source

fn contains_many(&self, oids: &[ObjectId]) -> Result<Vec<bool>>

For each id in oids, whether it is stored in this backend; the result has one entry per input id, in input order.

Callers that need to check many objects at once (for example write-tree confirming that every reusable cached subtree still exists) use this instead of one Self::contains call per id. The default implementation calls Self::contains for each id. A database- or network-backed store should override it to answer the whole batch in a single round trip.

§Errors

Propagates the first backend failure, like Self::contains.

Source

fn open_stream(&self, oid: &ObjectId) -> Result<Option<ObjectStream<'_>>>

Open a stream over the object’s uncompressed payload.

Default implementation buffers via Self::read.

§Errors

Same as Self::read.

Source

fn lookup_prefix( &self, prefix: &str, limit: usize, out: &mut Vec<ObjectId>, ) -> Result<()>

Fill out with object ids whose hex form starts with prefix (case-insensitive).

At most limit ids are appended (0 means no limit). When more than one object matches and limit is 1, callers treat the result as ambiguous if out would have held multiple matches.

The default implementation parses prefix once into an ObjectIdPrefix and compares it against the raw bytes of every id from Self::for_each_object: one allocation-free pass over the whole store per call. Abbreviating ids and resolving short ids call this once per id, so a store with many objects and a cheaper way to find ids by leading bytes (a sorted index, an ordered key-value range from ObjectIdPrefix::lowest to ObjectIdPrefix::highest) should override it.

§Errors

Returns Error::InvalidObjectId when prefix is not valid hex or is longer than a full object id for this store’s hash algorithm.

Source

fn refresh(&self) -> Result<bool>

Reload indexes or caches after on-disk changes (packs, alternates, MIDX).

Returns true when the store’s view of objects changed. Default is no-op.

§Errors

Propagates backend refresh failures.

Dyn Compatibility§

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors§