Enum StatisticalAggregateExecutionMode 

pub enum StatisticalAggregateExecutionMode {
    DataFusionExactHostWelford,
    GpuNativeTolerant,
}

Execution contract for floating statistical aggregates such as STDDEV_SAMP, STDDEV_POP, VAR_SAMP, and VAR_POP.

Self::DataFusionExactHostWelford requires the supported DataFusion aggregate semantics and a validated input-order and evaluation contract. Bit-for-bit equivalence is claimed only for that contract. If the required contract is unavailable, planning rejects the native candidate, and a surrounding join or final sort does not supply exactness. Self::GpuNativeTolerant remains the default and does not promise bit-for-bit equivalence with DataFusion.

Variants

DataFusionExactHostWelford

Evaluate DataFusion’s Welford accumulator on the host over the aggregate input, in DataFusion input order where the plan carries it.

Aggregates in the same spec keep DataFusion’s integer and NaN semantics. Planning rejects group-key and aggregate types that this host kernel does not cover, as well as ungrouped floating SUM and AVG, which DataFusion accumulates batch by batch. Those plans remain on DataFusion. A NaN result matches as a NaN, with no bit-for-bit equality. An add of two NaN operands keeps the sign of the operand the compiler places first, in DataFusion and in this kernel alike.

GpuNativeTolerant

Uses libcudf native grouped and scalar statistical kernels with each aggregate’s declared degrees of freedom. This mode is the default.

Variance is derived from sum and sum-of-squares moments, while DataFusion uses Welford updates. Floating-point results can differ from DataFusion, and precision degrades when values are large relative to their spread.

Implementations

impl StatisticalAggregateExecutionMode

pub const fn as_str(self) -> &'static str

Stable snake-case mode label used in env vars and reports.

pub const fn tolerance_contract(self) -> StatisticalAggregateToleranceContract

Floating parity contract required by this execution mode.

pub const fn is_datafusion_exact(self) -> bool

Returns true when execution must preserve DataFusion’s exact Welford bits.

Admission reads this to decide whether a plan with floating statistical aggregates is rejected for group-key or aggregate types the host kernel does not cover; under the tolerant mode those plans run on libcudf.

Trait Implementations

impl Clone for StatisticalAggregateExecutionMode

fn clone(&self) -> StatisticalAggregateExecutionMode

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for StatisticalAggregateExecutionMode

impl Debug for StatisticalAggregateExecutionMode

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

Formats the value using the given formatter.

impl Default for StatisticalAggregateExecutionMode

fn default() -> StatisticalAggregateExecutionMode

Returns the “default value” for a type.

impl Display for StatisticalAggregateExecutionMode

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

Formats the value using the given formatter.

impl Eq for StatisticalAggregateExecutionMode

impl FromStr for StatisticalAggregateExecutionMode

type Err = StatisticalAggregateExecutionModeParseError

The associated error which can be returned from parsing.

fn from_str( value: &str, ) -> Result<StatisticalAggregateExecutionMode, <StatisticalAggregateExecutionMode as FromStr>::Err>

Parses a string s to return a value of this type.

impl Hash for StatisticalAggregateExecutionMode

fn hash<__H>(&self, state: &mut __H)

where __H: Hasher,

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for StatisticalAggregateExecutionMode

fn eq(&self, other: &StatisticalAggregateExecutionMode) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for StatisticalAggregateExecutionMode