Edit on GitHub

OpenMDAO.jl Documentation

What?

Use Julia with OpenMDAO! OpenMDAO.jl is a Julia package that allows a user to:

  • Write OpenMDAO Components in Julia, and incorporate these components into a OpenMDAO model.
  • Create and run optimizations in Julia, using OpenMDAO as a library.

OpenMDAO.jl consists of three pieces of software:

  • OpenMDAOCore.jl: A small, pure-Julia package that allows users to define Julia code that will eventually be used in an OpenMDAO Problem. OpenMDAOCore.jl defines two Julia abstract types (AbstractExplicitComponent and AbstractImplicitComponent) and methods that mimic OpenMDAO's ExplicitComponent and ImplicitComponent classes.
  • omjlcomps: A Python package (actually, a OpenMDAO Plugin) that defines two classes, JuliaExplicitComp and JuliaImplicitComp, which inherit from OpenMDAO's ExplicitComponent and ImplicitComponent, respectively. These components takes instances of concrete subtypes of OpenMDAOCore.ExplicitComponent and OpenMDAOCore.ImplicitComponent and turn them into instances of JuliaExplicitComp and JuliaImplicitComp. Like any other OpenMDAO ExplicitComponent or ImplicitComponent objects, JuliaExplicitComp and JuliaImplicitComp instances can be used in an OpenMDAO model, but call Julia code in their methods (compute, apply_nonlinear, etc.).
  • OpenMDAO.jl: A Julia package that has the openmdao and omjlcomps Python packages as dependencies. Users can install OpenMDAO.jl and have the full power of the OpenMDAO framework at their disposal in Julia.

How (Installation Instructions)?

There are two approaches to getting OpenMDAO working with Julia: the Python-Centric Approach and the Julia-Centric Approach. If you like Python and just want to have a little (or a lot) of Julia buried in your OpenMDAO System, then you'll probably prefer the Python-centric approach. If you're a huge fan of Julia and would like to pretend that OpenMDAO is a Julia library, you'll want the Julia-centric approach. Either way, pick one or the other: you don't need to follow both installation instructions.

Note At the moment OpenMDAO.jl is not registered in the Julia General registry, so installing it from the Pkg prompt with 

] add OpenMDAO

will fail on a vanilla Julia installation. It is registered in the official™ DanielIngrahamRegistry, however, so if you'd like an automatic installation process, just do this from the Julia Pkg prompt:

pkg> registry add git@github.com:dingraha/DanielIngrahamRegistry.git

Python-Centric Installation

The first (and only!) step is to install omjlcomps, which is in the Python Package Index, so a simple

pip install omjlcomps

should be all you need. omjlcomps uses JuliaPkg to manage Julia dependencies, so all the Julia packages needed by omjlcomps (and even Julia itself, if necessary) will be installed automatically.

Julia-Centric Installation

The OpenMDAOCore.jl and OpenMDAO.jl Julia package are the official™ DanielIngrahamRegistry, so if you have access to that, installation should be as simple as

] add OpenMDAOCore OpenMDAO

in the Julia REPL. OpenMDAOCore.jl is a fairly small package without any Python dependencies, but OpenMDAO.jl depends on omjlcomps and openmdao itself. OpenMDAO.jl's Python dependencies are managed by CondaPkg, and should be automatically installed into a separate Conda environment specific to your current Julia environment.

Acknowledgements

  • An early version of OpenMDAO.jl was written by Daniel Ingraham, Justin Gray, and Andrew Ning while visiting Prof. Ning at Brigham Young University.
  • OpenMDAO.jl depends heavily on PythonCall and related packages, developed by Christopher Rowley.