Reshape

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

Purpose

Provides shape-changing helpers that preserve flat storage order.

Main API

Reshape

• reshape(array, shape)

Flatten

• flatten(array)

Invariants

• reshape preserves flat element order.
• reshape does not change element count.
• reshape always returns an owning Tensor.
• flatten always returns an owning rank-1 Vector.
• flatten(array).size() == array.size().
• No reshape or flatten helper returns a view.

Validation Notes

• reshape requires array.size() == shape.elements() through core::validation::require_equal_size().
• Shape mismatch throws Exceptions::ShapeError.
• Overflow from Shape::elements() propagates as Exceptions::DimensionError.
• Empty and zero-dimension shapes are allowed when the element count matches.

Implementation Notes

• reshape always returns a Tensor<value_type>.
• flatten always returns a rank-1 Vector<value_type>.
• Values are copied in flat storage order using operator[].
• No view semantics are implemented; returned containers own their buffers.
• Element-count validation should stay routed through Validation.hpp.

Time Complexity

• reshape(array, shape) is O(n + r).
• flatten(array) is O(n).
• Shape element-count validation is O(r).

Future Work

• Add view-based reshape if non-owning containers are introduced.
• Add inferred dimension support, such as {2, -1}, only if signed shape syntax is added.
• Consider preserving container type where possible.