Slice

Developer notes for include/stratax/ops/Slice.hpp.

Purpose

Implements range-based slicing for vectors, matrices, and tensors using core::Slice.

Main API

Vector Slicing

• slice(vector, slice)

Matrix Slicing

• slice(matrix, row_slice, col_slice)

Tensor Slicing

• slice(tensor, slice_0, slice_1, ...)

Invariants

• Slicing uses half-open ranges [start, stop).
• Slice results are owning containers, not views.
• Vector slice rank is always 1.
• Matrix slice rank is always 2.
• Tensor slice rank matches the input tensor rank.
• Result shape dimensions are the sizes of the supplied slices.
• Slicing preserves source element order within the selected region.

Validation Notes

• core::Slice rejects reversed ranges at construction.
• Vector slice stop must be within vector.size() through core::validation::require_at_most().
• Matrix row and column slice stops must be within rows() and cols() through core::validation::require_at_most().
• Tensor slice count must match tensor rank through core::validation::require_rank().
• Tensor rank mismatch throws Exceptions::DimensionError.
• Out-of-bounds slices throw Exceptions::IndexError.

Implementation Notes

• Slices are half-open: [start, stop).
• Vector and matrix slicing copy values into new owning containers.
• Tensor slicing computes result shape from slice sizes.
• Empty result slices return immediately to avoid indexing through zero strides.
• Tensor source offsets are computed from source strides.
• Rank and slice-stop checks should stay routed through Validation.hpp.

Time Complexity

• Vector slicing is O(k), where k is the slice length.
• Matrix slicing is O(rows_selected * cols_selected).
• Tensor slicing is O(result_size * r), where r is tensor rank.
• Empty result slices are O(r) for validation and shape construction.

Future Work

• Add step support to core::Slice.
• Add negative indexing only if signed indexing becomes part of the API.
• Consider slice views to avoid copies.