Struct GpuResultHandle 

pub struct GpuResultHandle { /* private fields */ }

GPU-resident query result bound to an Arrow schema and runtime.

Cloning shares the underlying cuDF batch, runtime, and frozen resident leases. Same-device consumers may borrow the result as resident input; host materialization is explicit through to_host_record_batch.

Implementations

impl GpuResultHandle

pub fn new( context: &ExecutionContext, frame: GpuDataFrame, schema: SchemaRef, ) -> Result<Self>

Bind a GPU frame to a schema and the runtime from context.

The frame must already reside on the context’s device. Hidden DataFusion order columns may be stripped before schema binding.

pub fn frame(&self) -> &DataFrame

Borrow the underlying GPU cuDF frame.

pub fn read_on(&self, context: &ExecutionContext) -> Result<GpuDataFrame>

Borrows this result as a frame read on context’s stream. Once that frame and every frame derived from it have dropped, the resident owner’s final release also waits for the work the stream had enqueued by then.

pub fn into_gpu_frame(self) -> GpuDataFrame

Consume the handle and return its GPU frame wrapper.

pub fn schema(&self) -> &SchemaRef

Arrow schema bound to this result.

pub fn runtime(&self) -> &Arc<Runtime>

Shared cuDF runtime that created the handle.

pub fn graph_residency(&self) -> Option<&GraphResidencyDescriptor>

Graph source provenance carried by this handle, when known.

pub fn placement(&self) -> ExecutionPlacement

Current cuDF execution placement for the underlying batch.

pub fn num_rows(&self) -> usize

Row count in the GPU result.

pub fn num_columns(&self) -> usize

Column count in the GPU result.

pub fn to_host_record_batch(&self) -> Result<RecordBatch>

Materialize this GPU result to host Arrow.

pub fn into_host_batch_cursor( self, cancellation: ExecutionCancellation, config: HostBatchCursorConfig, ) -> Result<GpuHostBatchCursor>

Consume this result into a bounded incremental host materializer.

pub fn into_host_batch_cursor_with_host_schema( self, output_schema: SchemaRef, cancellation: ExecutionCancellation, config: HostBatchCursorConfig, ) -> Result<GpuHostBatchCursor>

Consume this result into a bounded host materializer using an external schema whose field names, nullability, and metadata are bound by ordinal.

Trait Implementations

impl Clone for GpuResultHandle

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.