Externalise dependencies?

[willfurnass]

Would the following be best managed as external dependencies rather than included in this repository? It would keep the repository much smaller and may make it easier to switch to newer versions of dependencies

• keras-contrib
• deep-belief-network
• libconfig
• ShapeBasedInterpolation
• possibly others...

[ece7048]

How can I link them with the external dependences? ( I mean with the GitHub repositories?)

[willfurnass]

For the Python packages can you use versions available from https://pypi.org/ or do you need to use very recent development versions only available from GitHub or similar? If you can find what you need on pypi.org you can add those dependencies to a requirements.txt at the top level in your project. You could then add pip install -r requirements.txt to your setup instructions.

For C++/C dependencies you may be able to acquire some/all dependencies from Conda channels. If that's not the case then you might be able to use CMake to simplify the process of finding and using dependencies when building the C++ parts of your project. For more advice on using CMake I recommend booking a Code Clinic appointment as others in the RSE are much more experienced at using CMake.

[willfurnass]

Actually, after discussing this with colleagues I think it may be simplest to leave any C/C++ dependencies as part of this repo. I still think it's a good idea to externalise Python dependencies though.