Conversions

Version: v0.2.0

Status: Complete

Header: include/stratax/algorithms/Conversion.hpp

Overview

Conversions.hpp provides shape conversion and value-type casting helpers for Stratax array containers.

Helpers preserve flat storage order and return new owning containers.

Responsibilities

The conversions module is responsible for:

Converting array-like containers to vector, matrix, or tensor forms
Supporting shape compatibility checks for vector/matrix conversion
Casting element types across vector/matrix/tensor containers

The conversions module is not responsible for:

View-based conversion
Implicit numeric safety checks beyond static_cast
Runtime dtype-policy configuration

Relationships

to_vector / to_matrix / to_tensor
└── flat index copy loop
 
astype<To>(...)
└── per-element static_cast<To>

Depends on:

Invariants

The following conditions are always true:

Conversions return new owning containers.
Element order is preserved in flat storage order.
to_vector accepts shapes that are rank-1 or have exactly one non-singleton dimension.
to_matrix accepts shapes that are rank-2 or have exactly two non-singleton dimensions.
to_tensor preserves original shape exactly.
astype preserves shape and element count.

Public Interface

Shape helpers

bool is_vector_shape(const stratax::core::Shape& shape);
bool is_matrix_shape(const stratax::core::Shape& shape);
stratax::core::Shape matrix_shape(const stratax::core::Shape& shape);

Container conversions

template<Array A>
stratax::container::Vector<typename A::value_type> to_vector(const A& arr);
 
template<Array A>
stratax::container::Matrix<typename A::value_type> to_matrix(const A& arr);
 
template<Array A>
stratax::container::Tensor<typename A::value_type> to_tensor(const A& arr);

Throws

Exceptions::ShapeError for unsupported source shape in to_vector/to_matrix

Type casting

template<typename To, typename From>
requires Numeric<To> && Numeric<From>
stratax::container::Vector<To> astype(const stratax::container::Vector<From>& vec);
 
template<typename To, typename From>
requires Numeric<To> && Numeric<From>
stratax::container::Matrix<To> astype(const stratax::container::Matrix<From>& mat);
 
template<typename To, typename From>
requires Numeric<To> && Numeric<From>
stratax::container::Tensor<To> astype(const stratax::container::Tensor<From>& tensor);

Behavior

Uses static_cast<To> per element

Complexity Summary

Operation	Complexity
`is_vector_shape` / `is_matrix_shape`	O(r)
`matrix_shape`	O(r)
`to_vector` / `to_matrix` / `to_tensor`	O(n + r)
`astype` overloads	O(n)

n is element count and r is rank.

Examples

const auto v = to_vector(tensor_like);
const auto m = to_matrix(tensor_like);
const auto t = to_tensor(matrix_like);
 
const auto as_double = astype<double>(v);

Design Notes

Vector/matrix conversion is intentionally permissive for singleton dimensions, which simplifies interoperability with tensor-shaped data that carries redundant axes.

Future Improvements

Add explicit policy helpers for strict rank-only conversion
Add optional checked-cast helpers for narrowing conversions