Module slice_view 

TensorSliceView — zero-copy logical views into tensor data via offset+stride descriptors, supporting slicing, broadcasting, and element access without data duplication.

Overview

A TensorSliceView holds a flat Vec<f64> as the backing store and a ViewDescriptor that describes the logical shape, strides, and starting offset into that store. All slice and broadcast operations return new descriptors that share the same flat buffer – no data is ever copied.

Key types

Type	Purpose
SliceRange	A half-open [start, stop) range with a step
ViewDescriptor	Offset + shape + strides descriptor for a view
BroadcastShape	NumPy-style broadcast compatibility and stride computation
SliceViewStats	Cumulative counters for views, slices, and broadcasts
TensorSliceView	The main manager that owns the data and exposes the API

Structs

BroadcastShape

NumPy-style broadcast compatibility checker and stride generator.

SliceRange

A half-open range [start, stop) with a positive step.

SliceViewStats

Cumulative statistics for a TensorSliceView.

TensorSliceView

Zero-copy logical view manager over a flat f64 data buffer.

ViewDescriptor

Describes a logical view into a flat buffer via an offset, shape, and strides.