Add hypothesis test for netCDF4 roundtrip

[takluyver]

Part of #1846: add a property-based test for reading & writing netCDF4 files.

This is the first time I've played with Hypothesis, but it seems to be working - e.g. I got an error with float16, and the netCDF docs show that 16-bit floats are not a supported data type.

However:

• This currently only tests a dataset with a single variable - it could be extended to multiple variables if that's useful.
• It looks like netCDF4 should support unicode characters, but it failed when I didn't have max_codepoint=255 in there. I don't know if that's an expected limitation I'm not aware of, or a bug somewhere. But I thought I'd make the test pass for now.

[pep8speaks]

Hello @takluyver! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2019-10-12 09:28:07 UTC

[dcherian]

Thanks @takluyver. LGTM but maybe @Zac-HD can review this and #3285

[Zac-HD]

I'd delete this part - it's at best redundant given the changes in #3285 and I'd prefer that config.

[Zac-HD]

Where is this specified? It would be useful to provide a link and prehaps explain what's going on with the categories too 😄

[jhamman]

@takluyver - just checking in on this one to see if there is interested in wrapping this up.

[takluyver]

Yes, sorry, it dropped off my radar. I've implemented all the changes suggested above.

The datetime64 & timedelta64 cases don't roundtrip completely, so there are test failures. I've added these in a separate commit at the end, so it's easy to cut that back out if you prefer.

[Zac-HD]

The datetime64 & timedelta64 cases don't roundtrip completely, so there are test failures. I've added these in a separate commit at the end, so it's easy to cut that back out if you prefer.

The basic problem seems to be that Hypothesis is generating timedelta64[Y], but Xarray is dealing with timedelta64[ns]:

____________________________ test_netcdf_roundtrip _____________________________
tmp_path = PosixPath('/tmp/pytest-of-vsts/pytest-0/test_netcdf_roundtrip0')
data = data(...), arr = array([293], dtype='timedelta64[Y]')
        ...
        with xr.open_dataset(tmp_path / "test.nc") as roundtripped:
>           xr.testing.assert_identical(original, roundtripped)
E           AssertionError: Left and right Dataset objects are not identical
E           
E           Differing data variables:
E           L   data     (0) timedelta64[ns] -106488 days +01:41:02.290448
E           R   data     (0) timedelta64[ns] -106488 days +01:41:02.290449

properties/test_netcdf_roundtrip.py:51: AssertionError
---------------------------------- Hypothesis ----------------------------------
Falsifying example: test_netcdf_roundtrip(
    tmp_path=PosixPath('/tmp/pytest-of-vsts/pytest-0/test_netcdf_roundtrip0'), 
    data=data(...), 
    arr=array([293], dtype='timedelta64[Y]')
)

So either that's a pretty serious bug, or you should specify the max_period of the timedelta dtype strategy.