Struct TensorSliceView 

pub struct TensorSliceView {
    pub data: Vec<f64>,
    pub descriptor: ViewDescriptor,
    pub stats: SliceViewStats,
}

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

All slice/broadcast operations produce new ViewDescriptors that share the same backing data — no elements are ever duplicated.

Examples

use ipfrs_tensorlogic::slice_view::{TensorSliceView, SliceRange};

// 2×3 tensor filled with 0..6
let data: Vec<f64> = (0..6).map(|x| x as f64).collect();
let mut view = TensorSliceView::new(data, vec![2, 3]);

assert_eq!(view.get(&[0, 0]), Some(0.0));
assert_eq!(view.get(&[1, 2]), Some(5.0));

// Slice row 1 only
let slice_desc = view.slice(0, SliceRange::contiguous(1, 2)).expect("example: should succeed in docs");
assert_eq!(slice_desc.base_offset, 3);
assert_eq!(slice_desc.shape, vec![1, 3]);

Fields

data: Vec<f64>

Flat backing buffer.

descriptor: ViewDescriptor

Current logical view descriptor.

stats: SliceViewStats

Cumulative statistics (slices, broadcasts, views created).

Implementations

impl TensorSliceView

pub fn new(data: Vec<f64>, shape: Vec<usize>) -> Self

Create a new TensorSliceView over data with the given logical shape.

Row-major strides are computed automatically and base_offset is set to 0. stats.total_views_created is incremented.

pub fn get(&self, indices: &[usize]) -> Option<f64>

Retrieve a single element by its multi-dimensional index.

Returns None if the index is out of bounds for the logical shape or if the computed flat index exceeds the backing buffer length.

pub fn slice(&mut self, dim: usize, range: SliceRange) -> Option<ViewDescriptor>

Apply a slice along dim and return the resulting ViewDescriptor.

The data buffer is not modified; the new descriptor adjusts base_offset and strides[dim] to reflect the slice.

Returns None if:

• dim >= self.descriptor.ndim(), or
• range.start >= self.descriptor.shape[dim].

stats.total_slices is incremented on success.

pub fn broadcast_strides( &mut self, target_shape: &[usize], ) -> Option<Vec<usize>>

Compute broadcast strides to view the current logical shape as target_shape.

Uses BroadcastShape internally. Returns None if the shapes are incompatible. stats.total_broadcasts is incremented on success.

pub fn stats(&self) -> &SliceViewStats

Return a reference to the cumulative statistics.

Trait Implementations

impl Debug for TensorSliceView

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

Formats the value using the given formatter.