Example recipe is too specific: will not work with DRS: BADC or DKRZ

[valeriupredoi]

ESMValTool/esmvaltool/recipes/examples/recipe_python.yml

Line 7 in f6dc985

- {dataset: ERA-Interim, project: OBS, tier: 3, type: reanaly, version: 1, start_year: 2000, end_year: 2002}

This has to be a general purpose example recipe and the fact that ERA-Interim is listed under the dataset section makes it fail when running it with drs set to e.g. BADC; suggest removing the ERA-Interim data request

[bouweandela]

If it fails because drs is set to BADC, that is a problem with _data_finder.py, not with the recipe.

I agree that having an example that uses observational data that is difficult to obtain is not ideal though, maybe we should remove the NCEP dataset as well.

[mattiarighi]

I agree. The easiest solution if you want to have observational data in the example recipes is to use the obs4mips data, which are freely available and already cmorized.

Check this recipe for example.

But we can also have examples using only models, that's fine as well.

[mattiarighi]

@valeriupredoi can you take care of this and then close this issue? Thanks! 👍

[valeriupredoi]

aye! @bouweandela @mattiarighi

[fserva]

I am also trying to run the python example. However, there is a fatal error when self._check_rank()
is called, causing File "ESMValTool/esmvaltool/cmor/check.py", line 125, in report_errors raise CMORCheckError(msg) esmvaltool.cmor.check.CMORCheckError: There were errors in variable ta: ta: does not match coordinate rank in cube:. Do you think it is safe to simply comment/remove this check?
In that case, there is a plotting error with pcolormesh
result = self._pcolormesh_patched(*args, **kwargs) File "/home/serva/miniconda3/envs/esmvaltool/lib/python3.6/site-packages/cartopy/mpl/geoaxes.py", line 1469, in _pcolormesh_patched if not self._hold: AttributeError: 'GeoAxesSubplot' object has no attribute '_hold'
A related discussion is here: SciTools/cartopy#1120
Any suggestion? Many thanks

[mattiarighi]

The issue with _check_rank has been solved by #627. You probably need to merge the latest version2_development into your branch.

@valeriupredoi can you help with the other issue?

[valeriupredoi]

I am unable to reproduce the (second) issue; I have, however, removed the OBS entries in the recipe and tested - it works fine. Can you give it a test, please? #633 it is surely not related to the OBS files, but a check run of the recipe is still needed so we can push the PR. Can you tell me what matplotlib version you working with pls? 2.2.2 (one I have) is not complaining of anything

[fserva]

The environment is using matplotlib 3.0.0, and basemap 1.2.0. I will try merging the latest version2_development and removing OBS from the recipe. Thanks

[fserva]

Ok, after pulling the latest version2_development the rank error is gone. Thanks. The only error I have appears to be related with the _hold at the plotting stage.

[valeriupredoi]

yes, I can confirm that with matplotlib=3.0.0 I get the same error as you. It looks like the patch (code patch, not the graphical one) is not applied in the latest version. basemap can be whatever version, no effect on the issue at hand. I can suggest downgrading to matplotlib=2.2.2, that will avert this issue. Note that to do that you can not just conda install matplotlib=2.2.2 but rather need to downgrade the overarching package which is ncl (which is 6.5.0 for you) so need to do `conda install -c conda-forge ncl=6.4.0'