Struct BlockWriter 

pub struct BlockWriter { /* private fields */ }

The store’s block-write capability.

Only handle able to apply blocks; obtained exactly once from LoadedState::start and deliberately not cloneable, so granting it to a single task (the block builder in sequencer mode, the block sync loop in full-node mode) statically prevents every other component from writing blocks.

Exposes no read access: holders that also need to query the store receive the Arc<State> returned alongside this capability by LoadedState::start.

Implementations

impl BlockWriter

pub async fn apply_block_with_proving_inputs( &mut self, ordered_batches: OrderedBatches, block_inputs: BlockInputs, signed_block: SignedBlock, ) -> Result<(), ApplyBlockWithProvingInputsError>

Saves proving inputs for a signed block and applies it to the state.

Used by the in-process block producer after it has built and signed a block.

pub async fn save_proving_inputs( &self, block_num: BlockNumber, proving_inputs: &BlockProofRequest, ) -> Result<()>

Saves the proving inputs for the given block to the block store.

impl BlockWriter

pub async fn stop(self, writer_task: WriterTask)

Stops the store, waiting until the write worker has released the tree storage it owns.

Consumes the capability — closing the write channel the write worker listens on — and then joins the writer task returned by LoadedState::start. The drop must precede the join or the write worker never observes the closed channel; doing both here keeps that ordering out of caller hands. Read-only State references may outlive the stop.

Callers that need the storage released deterministically must use this method instead of dropping: the node’s recover command stops the store before the process exits, and the stress-test’s store seeding stops it so the same data directory can be re-loaded (or its temporary directory deleted) immediately afterwards. The running node does not use this method — its writer exits via the shutdown token passed to State::load and is joined through the node’s task set.

Panics

Panics if the writer task panicked.

pub async fn apply_block( &mut self, signed_block: SignedBlock, protocol_config: Option<ProtocolConfig>, ) -> Result<(), ApplyBlockError>

Apply changes of a new block to the DB and in-memory data structures.

Supply the active configuration if its commitment is not yet stored. The configuration must match the block header. New configurations are committed with the block. If the configuration is omitted and its commitment is not stored, this method returns DatabaseError::ProtocolConfigNotFound through ApplyBlockError.

Blocks are forwarded to the store’s write worker task, which processes them one at a time. Readers are unaffected while a block is being applied: they keep reading from the previous in-memory snapshot until the writer atomically publishes the new one.