How to distribute the examples

[jklymak]

python packages can have data, but usually that is for things the routines need at runtime. Not 100% sure how we should distribute the examples here. Maybe as a tarball in the doc/_static/? Not sure what other libraries do for examples like this that require significant data.

[callumrollo]

xarray has a seperate data repo that's downloaded from when a user runs their tutorial scripts. This would be an efficient way to keep bloat out of the repo.

It uses a script tutorial.py in the library to manage these calls

They use pooch for fetching and caching the data

[jklymak]

Ok that's a little different because they just need a file at a time and they want the datasets to run in an example.

We have to serve a whole directory, and I'm not sure we need to show much after the netcdf files are created. At that point they are just netcdf files.

Keeping the test data in a separate repo isn't a bad idea and maybe is all we need. Users and the tests can just install that repo locally with git clone. My only concern with that is some folks won't be thrilled with git clone and we prob also have to serve a tarball somewhere.

[callumrollo]

I think a separate repo would be a good idea. Should we just make a new repo with the current example directories? We could easily tarball the whole thing and make it available with a link in the README of pyglider

[jklymak]

Yeah, I think that's easiest. I'm a little ignorant how to get the tests to download that directory, but I think I can figure it out.

[jklymak]

OK, repo here... https://github.com/c-proof/pyglider-example-data

[jklymak]

We need a way to serve this and have it updated, probably on release of pyglider....