skainet-lang-core/sk.ainet.lang.tensor/Slice

Slice

sealed class Slice<T : DType, V>(source)

Represents a slice descriptor for tensor slicing operations.

Slice is a sealed class hierarchy that describes different ways to select elements along a single tensor dimension. It provides the building blocks for complex tensor slicing operations while maintaining type safety and enabling compile-time optimization opportunities.

Slice Types

The sealed class hierarchy includes:

Range: Contiguous range of indices [start, end)
At: Single specific index selection
All: Full dimension selection (equivalent to :)
Step: Strided access with start, end, and step size

Design Principles

Type Safety: Compile-time guarantees about slice validity
Composability: Slices can be combined and normalized
Performance: Enable optimization based on slice patterns
Expressiveness: Support common slicing use cases

Parameters

T

the data type constraint extending DType

V

the actual value type

Inheritors

Types

class All<T : DType, V> : Slice<T, V>

Represents selection of the entire dimension.

data class At<T : DType, V>(val index: Int) : Slice<T, V>

Represents selection of a single specific index.

data class Range<T : DType, V>(val start: Int, val end: Int) : Slice<T, V>

Represents a contiguous range slice [start, end).

data class Step<T : DType, V>(val start: Int, val end: Int, val step: Int) : Slice<T, V>

Represents strided access with custom step size.

Functions

fun getResultSize(dimensionSize: Int): Int

Calculates the number of elements this slice will select.

hasNonTrivialStride

fun hasNonTrivialStride(): Boolean

Checks if this slice uses non-trivial stride patterns.

fun isContiguous(): Boolean

Determines if this slice represents contiguous memory access.

fun isEmpty(): Boolean

Determines if this slice selects zero elements.

fun isValid(dimensionSize: Int): Boolean

Validates that this slice is compatible with the given dimension size.

fun normalize(dimensionSize: Int): Slice<T, V>

Normalizes this slice against a specific dimension size.