Add a tutorial dataset loader

[alioacar]

Proposed new feature or change:

Apologies if this is a duplicate or if I missed an existing way to do this.
It would be nice to have an xarray-like tutorial dataset API, maybe using pooch to download and cache small sample grid/data files. This would make the example notebooks easier to reproduce from outside the repository.

For example

import uxarray as ux

uxds = ux.tutorial.open_dataset("quad-hexagon")
uxgrid = ux.tutorial.open_grid("outCSne30")

Something like ux.tutorial.available_datasets() could also be useful, so users can quickly see which sample datasets are available.

[erogluorhan]

Hi @alioacar , this is great suggestion! We actually discussed this a few times, and then it fell through the cracks. We'll look into it again, or please let us know if you'd be interested in implementing it (we can help with that, too)

[alioacar]

Hi @erogluorhan, thanks! I think the main design decision here would be where to host the sample data. Maybe uxarray could follow a similar pattern to xarray (xarray-data) and keep the sample files in a small separate data repo (maybe uxarray-data), so the main package does not need to include the example datasets directly. I'm not sure if this fits with your plans but I'd be happy to start with a small prototype and adjust the approach as needed

[erogluorhan]

I'd imagine it to be similar to xarray's tutorial module with a slight difference (for now) on where the data resides:

• Rather than a separate data repo, we might stick to our test/meshfiles dir for now, which is working well for our small data needs so far
• With tutorial.open_dataset() in place, it'd be great to eventually go over all our docs to use it instead of relative path access, but having all our test data sets stay where they are would help us table that for however long we want.

Does this make sense?

@rajeeja thoughts?