TypedMatrices.jl: An Extensible and Type-Based Matrix Collection for Julia

TL;DR

TypedMatrices.jl provides an extensible, type-based collection of test matrices for Julia, enabling dynamic on-demand generation and property-driven organization. By encoding matrices as dedicated Julia types and associating properties as distinct types, it achieves memory efficiency and seamless integration with Julia's multiple dispatch and LinearAlgebra ecosystem. The framework supports grouping, property-based searches, and a testing interface for algorithm validation, delivering reproducible workflows and scalable benchmarking. While it surpasses prior collections in dynamism and type-awareness, it does not yet support static collections like Matrix Market, suggesting directions for future expansion and cross-language applicability.

Abstract

TypedMatrices.jl is a Julia package to organize test matrices. By default, the package comes with a number of built-in matrices and interfaces to help users select test cases based on their properties. The package is designed to be extensible, allowing users to define their own matrix types. We discuss the design and implementation of the package and demonstrate its usage with a number of examples.

Figures (1)

• Figure 1: The Minij matrix is a symmetric matrix with elements defined as $i+j$.