Struct ArrowPredicateFn

pub struct ArrowPredicateFn<F> {
    f: F,
    projection: ProjectionMask,
}

An ArrowPredicate created from an FnMut and a ProjectionMask

See RowFilter for more information on applying filters during the Parquet decoding process.

The function is passed RecordBatches with only the columns specified in the ProjectionMask.

The function must return a [BooleanArray] that has the same length as the input batch where each row indicates whether the row should be returned:

• true: the row should be returned
• false or null: the row should not be returned

Example:

Given an input schema: "a:int64", "b:int64", you can create a predicate that evaluates b > 0 like this:

// Create a mask for selecting only the second column "b" (index 1)
let projection_mask = ProjectionMask::leaves(&descriptor, [1]);
// Closure that evaluates "b > 0"
let predicate = |batch: RecordBatch| {
   let scalar_0 = Int64Array::new_scalar(0);
   let column = batch.column(0).as_primitive::<Int64Type>();
   // call the gt kernel to compute `>` which returns a BooleanArray
   gt(column, &scalar_0)
 };
// Create ArrowPredicateFn that can be passed to RowFilter
let arrow_predicate = ArrowPredicateFn::new(projection_mask, predicate);

Fields

f: F

projection: ProjectionMask

Implementations

impl<F> ArrowPredicateFn<F>

where F: FnMut(RecordBatch) -> Result<BooleanArray, ArrowError> + Send + 'static,

pub fn new(projection: ProjectionMask, f: F) -> Self

Create a new ArrowPredicateFn that invokes f on the columns specified in projection.

Trait Implementations

impl<F> ArrowPredicate for ArrowPredicateFn<F>

where F: FnMut(RecordBatch) -> Result<BooleanArray, ArrowError> + Send + 'static,

fn projection(&self) -> &ProjectionMask

Returns the ProjectionMask that describes the columns required to evaluate this predicate.

fn evaluate(&mut self, batch: RecordBatch) -> Result<BooleanArray, ArrowError>

Evaluate this predicate for the given [RecordBatch] containing the columns identified by Self::projection