Add rotated_ll to supported CF CRS grid types

[nshankar]

Should solve #458 and gives a matching result to pygrib in the notebook.

⚠️: Not 100% sure about the following mapping, it works for HRDPS but it's unclear to me whether it's correct in general

projparams["lon_0"] = da.GRIB_longitudeOfSouthernPoleInDegrees
projparams["o_lon_p"] = 0.0
projparams["o_lat_p"] = -da.GRIB_latitudeOfSouthernPoleInDegrees

⚠️: I noticed two of the models in the notebook are failing, should that be a separate issue?

📝: I also added jupyter and pygrib as dev dependencies but that should be easy to remove if unwanted.

[blaylockbk]

Thanks for the PR! I really appreciate it. Sorry it's taken some time to get to. I'd like to try this out myself before merging. Or, would you mind adding a test for this case?

I also added jupyter and pygrib as dev dependencies but that should be easy to remove if unwanted.
Hmm. I'm still thinking about this. I'd prefer pygrib to be an extra dependency not in dev (pygrib has had some issues in the past, and I'd prefer not requireing it in my dev environment). But I could be convinced otherwise. For Jupyter, I'm a VS Code user, and only need ipykernel. I'm assuming you're using JupyterLab. I'm probably fine with putting this in dev dependencies for other JupyterLab users.

[blaylockbk]

Not 100% sure about the following mapping, it works for HRDPS but it's unclear to me whether it's correct in general

I'm not aware of any other rotated_ll models asside from HRDPS, or is there?

[blaylockbk]

I noticed two of the models in the notebook are failing, should that be a separate issue?

This is probably due to not updating all my notebooks at each release. It could be a separate issue if you don't want to fix them in this PR.

[nshankar]

would you mind adding a test for this case?

Added a couple tests based on the structure you have set up. Please take a careful look.

Hmm. I'm still thinking about this. I'd prefer pygrib to be an extra dependency not in dev

Agree pygrib is sometimes problematic. I can move everything to extras, is that where you normally put testing dependencies? I had to add netcdf into dev as well in order get my tests passing locally (running uv run pytest).

It could be a separate issue if you don't want to fix them in this PR.

Yes can we track that separately please?

[nshankar]

DRAGONS:

At approx midnight UTC the HRDPS servers get cleared of the previous day's forecasts. However there is no new forecast available yet until 06Z. During that time these tests will fail.

What do you think is the right way to handle this in the Herbie package?

[nshankar]

Added this info to the module docstring

[mattw-nws]

Not 100% sure about the following mapping, it works for HRDPS but it's unclear to me whether it's correct in general

I'm not aware of any other rotated_ll models asside from HRDPS, or is there?

I'm hitting this when trying to monkey-patch a fix for the RRFS plugin... so seems to be in use there.

[mattw-nws]

I'm hitting this when trying to monkey-patch a fix for the RRFS plugin... so seems to be in use there.

I should specify, I think this is specific to the full-domain RRFS's projection (nat product). Adding this change fixed it--at least when pulling a COG into QGIS (having trouble with OpenLayers/proj4js with that projection so I dumped it for now and went back to CONUS domain).

[nshankar]

I see some tests are failing because of pygrib, however that dependency is needed to run the notebook. Is there another dependency group I can put it in @blaylockbk?

[blaylockbk]

What if we just kept it out of a dependency group and left it only in an extra dependency by itself.

[blaylockbk]

Pygrib is already an optional dependency. Maybe just keep it there and not in a group for now.

[nshankar]

Done!

[nshankar]

I followed the steps here https://github.com/blaylockbk/Herbie/blob/main/.github/workflows/tests-python.yml

uv sync --extra extras
uv run pytest

and the tests I've added are passing!

[blaylockbk]

Great! I think the error on Windows is related to cfgrib and eccodes. This can be addressed in a separate issue. The rest is good to merge. Thanks for this contribution. I really appreciate it. 😃