Contiguous store with unlim dim bug fix

[jmccreight]

• Closes passing unlimited_dims to to_netcdf triggers RuntimeError: NetCDF: Invalid argument #1849
• Tests added
• Fully documented, including whats-new.rst for all changes and api.rst for new API

Not sure this needs documentation else where...

[pep8speaks]

Hello @jmccreight! 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 2019-06-04 16:16:07 UTC

[jmccreight]

I guess the additional documentation would go here. http://xarray.pydata.org/en/stable/io.html#writing-encoded-data
If we decide to keep the "gnarly" warning message, we should probably do a bit of explaination here.

[jmccreight]

I should add that changing the encoding on the variable itself was not "an easy detail". The basic thing I tried was unsuccessful, so if this is desired some more work is needed.

[dcherian]

Thanks @jmccreight. That was quick.

I think we can do without the warning but we should wait till someone else chimes in. Otherwise, this looks right to me but I don't know this bit of the code too well.

[jmccreight]

There were easy patterns to follow for this, so I just went for it.
@dcherian Are the failures fixable? I'm not sure what to make of them.

[dcherian]

The failures seem unrelated. can you merge master and remove the warning?

[jmccreight]

🎊

[dcherian]

LGTM. Just some minor changes.

I'll merge in a couple of days if no one else has comments.

[shoyer]

This should also check for not raise_on_invalid, like what we do above for chunksizes. Otherwise, if a user specified an invalid encoding by hand we would silently ignore their error.

[jmccreight]

@shoyer I think this is what you mean.

    if not raise_on_invalid and var_has_unlim_dim and 'contiguous' in encoding.keys():
        del encoding['contiguous']

[dcherian]

Thanks @jmccreight . We prefer using () to wrap long lines instead of using a \. I've made the fix.

[jmccreight]

Thanks. I was, in fact, very uncertain about the preferred approach.

[dcherian]

=). Thanks @jmccreight. Merging.