Struct BulkWriter 

pub struct BulkWriter<'a> { /* private fields */ }

Deferred-fsync writer returned by ObjectStore::bulk_writer. See that method for the crash-safety contract.

Implementations

impl BulkWriter<'_>

pub fn write(&mut self, bytes: &[u8]) -> StoreResult<Hash>

Write one object durable-before-visible: contents are fsynced before the rename publishes the object, and only the shard-dir fsync (rename durability) is deferred to commit. This upholds the store’s global invariant (an object is never visible before its bytes are durable), so another process’s contains() dedup can never reference a half-written object. An existing path is byte-verified: matched objects are left in place (but still fsynced at commit, see below), torn ones are rewritten.

Panics

Never in practice: object paths always have a 2-hex shard parent by construction.

pub fn commit(self) -> StoreResult<()>

Make the session durable: fsync the CONTENTS of any pre-existing objects this batch re-used (newly written objects were already content-fsynced before their rename in write), then every touched shard directory (renames become durable).

Trait Implementations

impl<'a> Debug for BulkWriter<'a>

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

Formats the value using the given formatter.