test-input and example data and it's path and availability

[mathause]

We currently have coarse-grid cmip6 data at tests/test-data/calibrate-coarse-grid/cmip6-ng/. This is fine for the tests but we also use it in the examples and there are several issues with this.

1. the path is awkward and too long (even when relative)
2. constructing the path is cumbersome (see below)
3. the data is not available without cloning the repo, i.e. cannot run examples when the package is installed from pypi or conda. (However, the examples are also not available...)
4. We need to check the license of the data and if we are allowed to distribute it 🙁

Suggestion

• move cmip6-ng folder into the ./data folder
• we make the example data path available from the top level

Questions

• What is a good name?
  • mesmer.paths.cmip6_ng
  • mesmer.example_data.cmip6_ng
  • mesmer.tutorial.cmip6_ng_path
  • mesmer.cmip6_ng_coarse_grid
  • ???
• Should we rename the resolution folder (currently g025) because we regridded it again? (Unfortunately we have a different lon & lat resolution 🤷‍♂)
• Should we make the data available via automatic download if it's not?
  • I think we would do this via pooch and have every file individually. So to download we have to loop over them. A tar-archive would be much simpler but would double the required disk space and would probably double it every time we have to add a new file. pooch should be able to handle subfolders.

---

• We construct the path using importlib.resources - need to add some sanity checks but can use this as starting point (from Add Example for several Scenarios #521):

  # mesmer provides example data under "tests/test-data"
  data_path = importlib.resources.files("mesmer").parent / "tests" / "test-data"
  
  # make `data_path` relative to the current working directory
  # this leads to slightly nicer paths but is not really necessary
  data_path = data_path.relative_to(pathlib.Path.cwd(), walk_up=True)
  
  cmip_path = data_path / "calibrate-coarse-grid" / "cmip6-ng"

[mathause]

@veni-vidi-vici-dormivi would be great to get your opinion and naming-prowess here 🙂

[veni-vidi-vici-dormivi]

Ugh that is cumbersome.
I like mesmer.example_data.cmip6_ng best. We should check their license but I would think we are allowed to distribute it.
We could also just leave out the resolution folder, since we don't have any other, only keep it in the file name
I think looping through them is not so bad considering the excess space on the tar option. I had a quick look at pymagicc because they somehow provide data but it seems it is shipped automatically with the package.

[mathause]

I like mesmer.example_data.cmip6_ng

Me too. Now I saw that we already have mesmer.data so I am considering if we can use mesmer.data.something but did not yet come up with something convincing.

We should check their license but I would think we are allowed to distribute it.

Yes we are (luckily...)

We could also just leave out the resolution folder, since we don't have any other, only keep it in the file name

True, but I like to mirror the real cmip6_ng. Could maybe name it coarse_resolution or
r20 (from cdo remapbil,r20x20 ...), or just keep it

I had a quick look at pymagicc because they somehow provide data but it seems it is shipped automatically with the package.

Yes the define it as package_data

https://github.com/openscm/pymagicc/blob/0fa29c960769563a596356bf4b6ea9463f5a717e/setup.py#L161-L162