Trait TakeExecute

Source

pub trait TakeExecute: VTable {
    // Required method
    fn take(
        array: &Self::Array,
        indices: &dyn Array,
        ctx: &mut ExecutionCtx,
    ) -> VortexResult<Option<ArrayRef>>;
}

Required Methods§

Source

fn take( array: &Self::Array, indices: &dyn Array, ctx: &mut ExecutionCtx, ) -> VortexResult<Option<ArrayRef>>

Take elements from an array at the given indices, potentially reading buffers.

Unlike TakeReduce, this trait is for take implementations that may need to read and execute on the underlying buffers to produce the result.

§Preconditions

The indices are guaranteed to be non-empty.

Dyn Compatibility§

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety", so this trait is not object safe.

Implementors§

Source §

impl TakeExecute for FixedSizeListVTable

Take implementation for FixedSizeListArray.

Unlike ListView, FixedSizeListArray must rebuild the elements array because it requires that elements start at offset 0 and be perfectly packed without gaps. We expand list indices into element indices and push them down to the child elements array.

Source §

impl TakeExecute for ListVTable

Source §

impl TakeExecute for ListViewVTable

ListViewArray take implementation.

This implementation is deliberately simple and read-optimized. We just take the offsets and sizes at the requested indices and reuse the original elements array. This works because ListView (unlike List) allows non-contiguous and out-of-order lists.

We don’t slice the elements array because it would require computing min/max offsets and adjusting all offsets accordingly, which is not really worth the small potential memory we would be able to get back.

The trade-off is that we may keep unreferenced elements in memory, but this is acceptable since we’re optimizing for read performance and the data isn’t being copied.

Source §