Introduction

Libejovo is the synthesis of my Applied Mathematics and Computer Science education. Centering around the fundamental type Matrix, Libejovo contains a number of general purpose routines that span the following subjects:

• Numeric Linear Algebra
  • Matrix Decompositions
  • Iterative Methods (Ax = b, eigenvalue methods)
  • Direct Methods (Ax = b, eigenvalue methods)
  • Random Matrix generation
  • Basic framework of matrix operations, additions, norms
  • inquiry functions like is Hermitian, rank, dim, etc..
  • Polynomial interpolation
• Statistics
  • Standard routines (std, mean, var, cor)
  • Functional patterns to operate on columns
  • DataFrame structure to write data as csv
• Differential Equations
  • Basic solvers like euler's method
• Numerical Quadrature
  • Also not implemented in this C library
  • Adaptive methods
• Probability
  • Non-naive psuedo random number generators
  • Inverse transform, rejection, box-muller methods for sampling from probability distributions
  • Matrices sampled from the Gaussian Orthogonal Ensemble

Documentation

A high-level API hosted at readthedocs is available here. This documentation tells a story about what this library is and snippets of code that demonstrate how to use and think with this library. For a more technical API, consult the doxygen documentation hosted with github pages.

Build

Build out of source with CMake

git clone [email protected]:ejovo13/libejovo.git

# configure build
cmake -B ./build
cmake --build ./build

To test that the functions are working properly, run CTest in the build directory.

cd ./build
ctest

We can inspect the symbols of libejovo.a with

nm ./lib/libejovo.a

where ./ is the top-level source directory.

As a dependency

To include this project as a dependency, there are two super easy methods. The first consists of building and installing this project, followed by adding the header file paths to your $C_INCLUDE_PATH. When including this library in another CMake project, simply link the ejovo library with the target_add_libary command.

cmake --install build

Alternatively, we can use gitsubmodules to lock a reference to a commit of this repository in another project. For an example, see this project that uses libejovo as a dependency. I was blown away with how simple it was. All you have to do when cloning is make sure you download the submodule's files and then add the submodule as a subdirectory to your cmake project with add_subdirectory. CMake will take care of the rest!

In Progress

I am currently working on getting a proper High-level API established at readthedocs. The skeleton is already in place and provides a brief introduction to this library. The technical documentation is also a work in progress, with about half of all publicly exposed functions being documented.

TODO

Matrices

• Finish technical documentation with doxygen
• Add general API documentation with readthedocs
• Add complete documentation to ejovo_matrix.h
• Provide more conversion functions between a Matrix and a normal c array
• Finish Complete testing suite of Matrix.

Rand

• Please for the love of god add floating-point functions

Future

I might want to add a bitflag to each matrix to be able to keep track of whether or nor it is a transpose, if it's hermitian, symmetric, etc. etc.

Long term project goals

I want to make a really nice finished product that has the majority of fundamental linear algebra routines:

SVD, QR, LU, Cholesky, Eigenvalue decomposition, Gaussian Elimination.

I'd like to eventually (Throughout the next quarter) implement more of the iterative algorithms used to solve large systems of equations and eigenproblems

API

I'd like to be able to switch from Column-major and Row-major indexing seamlessly. The hookup is that I would have to change how rows and cols are copied, although this would require an adjustment to just a few very low-level functions.