Regridded output now same dtype as input

[brews]

Output regridded data is now the same type as input data. If input is xarray.Dataset with multiple unique dtypes, raises ValueError.

PR includes integration tests for new behavior under different input types/containers.

One pre-existing test needed to be updated because it output ints when the test was expecting precision from float.

Close #99.

[brews]

I think this is ready for review!

I'd especially appreciate input in how this is throwing a ValueError if a multi-dtype xr.Dataset is input.

Also, the change I needed to make to the pre-existing unit test makes me a bit nervous because this PR is clearly breaking from past behavior. Just another thing to note. Anyone else nervous about this?

[brews]

I think I could use maintainer input on this. pre-commit is giving me some trouble. It looks like it might be a pre-commit/GH actions configuration issue?

Specifically, I'm looking at

blackdoc.................................................................Failed
- hook id: blackdoc
- exit code: 1

Traceback (most recent call last):
  File "/home/runner/.cache/pre-commit/repotmc3t5mu/py_env-python3/bin/blackdoc", line 5, in <module>
    from blackdoc.__main__ import main
  File "/home/runner/.cache/pre-commit/repotmc3t5mu/py_env-python3/lib/python3.9/site-packages/blackdoc/__main__.py", line 10, in <module>
    from .blackcompat import find_project_root, gen_python_files, read_pyproject_toml
  File "/home/runner/.cache/pre-commit/repotmc3t5mu/py_env-python3/lib/python3.9/site-packages/blackdoc/blackcompat.py", line 9, in <module>
    import toml
ModuleNotFoundError: No module named 'toml'

from https://github.com/pangeo-data/xESMF/pull/102/checks?check_run_id=3091031309.

This PR hasn't added or changed dependencies.

I fixed other apparent linting issues by deactivating pre-commit.

[raphaeldussin]

@brews please feel free to add toml package to the various environment yml files under ci
it looks like black does not depend on toml anymore but blackdoc does:

keewis/blackdoc#100

this should take care of the linting test

[raphaeldussin]

sorry, this is not necessary anymore, fixed in master

[brews]

@brews please feel free to add toml package to the various environment yml files under ci
it looks like black does not depend on toml anymore but blackdoc does:

keewis/blackdoc#100

this should take care of the linting test

toml is in the environments as of 87285ce. This looks good to go on my end, unless CI/CD says otherwise.

[raphaeldussin]

the handling in the dataset is finicky and decisions need to be made in terms of desired behavior.
the rest looks good.

[huard]

@raphaeldussin Could you clarify what needs to be done to merge this ? What decisions do we need to make?

[raphaeldussin]

@brews I have made a PR to your branch here: brews#2
passing a list of different dtypes in the dataset case works fine (see test), please check and update your PR if you are satisfied with the modifications then we can move forward and merge your PR into the main code.

[raphaeldussin]

I'd recommend updating PR with modifs from here brews#2

[brews]

Thanks for the bump, @huard. I merged in the suggested corrections from @raphaeldussin. I agree with @raphaeldussin that accepting multiple dtypes from Datasets is ideal behavior. It appears to work in testing so 👍 .

Pending CI/CD checks, I think this is ready for re-review, @raphaeldussin.

[raphaeldussin]

LGTM