Struct HashPartitionSpiller 

pub struct HashPartitionSpiller { /* private fields */ }

A partition spill writer that routes rows by hash into separate files.

Each partition gets its own RunWriter for sorted/typed row data. The spiller handles skew detection and recursive repartitioning.

Implementations

impl HashPartitionSpiller

pub fn new( config: HashPartitionConfig, manager: &SpillManager, schema_fingerprint: u64, ) -> Result<Self, QueryError>

Create a new hash partition spiller.

pub fn insert_row( &mut self, row: &[Value], manager: &SpillManager, ) -> Result<(), QueryError>

Insert a row into the appropriate partition.

pub fn insert_row_to_partition( &mut self, row: &[Value], partition: usize, manager: &SpillManager, ) -> Result<(), QueryError>

Insert a row into a specific partition (caller computes the hash).

Useful when the caller needs to partition by a derived key (e.g., group key for aggregate) rather than the row itself.

pub fn finalize_with_manager( self, manager: &SpillManager, ) -> Result<Vec<Option<SpilledRun>>, QueryError>

Finalize all partitions through the manager so disk quota is enforced.

pub fn partition_counts(&self) -> &[u64]

Current partition row counts.

pub fn num_partitions(&self) -> u64

Number of partitions.

Trait Implementations

impl Debug for HashPartitionSpiller

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

Formats the value using the given formatter.