DiffEq Developer Documentation
This is the developer documentation and Contributor's Guide for the DiffEq ecosystem. It explains the common interface and some the package internals to help developers contribute.
If you have any questions, or just want to chat about solvers/using the package, please feel free to use the Gitter channel. For bug reports, feature requests, etc., please submit an issue.
The DiffEq ecosystem is built around the common interface. The common interface is a type-based interface where users define problems as a type, and solvers plug into the ecosystem by defining an algorithm to give a new dispatch to
There is a top level
init function which is in DiffEqBase.jl that handles distribution and function input (along with extra warnings) before sending the problems to the packages.
There is then an ecosystem of add-on components which use the common solver interface to add analysis tools for differential equations.
Contributing to the Ecosystem
There are many ways to help the ecosystem. One way you can contribute is to give pull requests (PRs) to existing packages. Another way to contribute is to add your own package to the ecosystem. Adding your own package to the ecosystem allows you to keep executive control and licensing over your methods, but allows users of DifferentialEquations.jl to use your methods via the common interface, and makes your package compatible with the add-on tools (sensitivity analysis, parameter estimation, etc). Note that, in order for the method to be used as a default, one is required to move their package to the JuliaDiffEq organization so that way common maintenance (such as fixing deprication warnings, updating tests to newer versions, and emergency fixes / disabling) can be allowed by JuliaDiffEq members. However, the lead developer of the package maintains administrative control, and thus any change to the core algorithms by other JuliaDiffEq members will only be given through PRs.
Even if you don't have the time to contribute new solver algorithms or add-on tools, there's always ways to help! Improved plot recipes and new series recipes are always nice to add more default plots. It is always helpful to have benchmarks between different algorithms to see "which is best". Adding examples IJulia notebooks to DiffEqTutorials.jl is a good way to share knowledge about DifferentialEquations.jl. Also, please feel free to comb through the solvers and look for ways to make them more efficient. Lastly, the documentation could always use improvements. If you have any questions on how to help, just ask them in the Gitter!
- Ecosystem Overview
- Adding a new package to the common interface
- Adding Algorithms
- Developing A New Problem
- The DiffEq Internals
- Type Traits
Algorithm Development Tools
The following algorithm development tools are provided by DiffEqDevTools.jl
- Test Problems
- Convergence Simulations
- Benchmark Suite