Struct Distinct 

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

Stable SQL duplicate elimination.

With no key expressions, the complete positional output row is the key. With expressions, the operator implements DISTINCT ON: it preserves the first row for each evaluated key in child order.

Implementations

impl<'a> Distinct<'a>

pub fn all(child: Box<dyn PhysicalOperator + 'a>) -> Self

Construct a bounded full-row DISTINCT with the compatibility default work-memory budget.

pub fn all_with_work_mem( child: Box<dyn PhysicalOperator + 'a>, work_mem_bytes: usize, ) -> Self

Construct a bounded full-row DISTINCT with an explicit byte budget.

pub fn on( child: Box<dyn PhysicalOperator + 'a>, keys: Vec<ScalarExpr>, evaluator: SharedExpressionEvaluator<'a>, ) -> Self

Construct a bounded DISTINCT ON with the compatibility default work-memory budget.

pub fn on_with_work_mem( child: Box<dyn PhysicalOperator + 'a>, keys: Vec<ScalarExpr>, evaluator: SharedExpressionEvaluator<'a>, work_mem_bytes: usize, ) -> Self

Construct a bounded DISTINCT ON with an explicit byte budget.

pub fn with_spill_directory(self, directory: impl Into<PathBuf>) -> Self

Place the exact-set files in a caller-selected temporary-data directory. The directory must already exist; a private child directory is created lazily on the first spill and removed through RAII.

pub fn has_spilled(&self) -> bool

Whether this invocation has migrated its key set to disk.

pub fn in_memory_key_bytes(&self) -> usize

Exact encoded key bytes retained by the in-memory set.

pub fn spill_path(&self) -> Option<&Path>

Live private spill directory, for diagnostics and cleanup tests.

Trait Implementations

impl PhysicalOperator for Distinct<'_>

fn row_schema(&self) -> &RowSchema

Complete logical-to-physical row layout emitted by this operator. Every Batch returned by Self::next must carry this exact schema; operators must reject a child that violates that invariant.

fn open(&mut self) -> ExecResult<()>

fn next(&mut self) -> ExecResult<Option<Batch>>

fn close(&mut self) -> ExecResult<()>

fn schema(&self) -> &[String]

Schema column names in logical output order.

fn estimated_cardinality(&self) -> Option<u64>

Planner/runtime cardinality estimate for choosing physical strategies. None means the operator cannot provide a useful estimate. The value is advisory rather than a correctness bound.

fn output_ordering(&self) -> &[PhysicalOrder]

Leading output ordering known to be preserved by this operator.

fn consume_into_aggregate( &mut self, _executor: &mut dyn AggregateExecutor, ) -> ExecResult<bool>

Let a leaf consume its native projected rows directly into an aggregate executor. Returning false promises that no input was consumed, so the caller can fall back to ordinary Batch pulls.