issue_comments

Thanks @crusaderky ! +1 to @dcherian 's comment!

Thanks @crusaderky. That was a major lift.

Thanks all!

Thanks @crusaderky. Good to have this policy solidified.

@shoyer any comments from you?

Okay, planning to merge c.COB tomorrow unless there are additional comments.

@jhamman it's ready for merge if everybody's happy with it

@crusaderky - this is getting hard to review at this point. Is there more to do or can we merge this in its current state?

I'm +0.5 on moving to 0.14 given the benefits of this and the breaking changes. Any other thoughts?

See #3380

The only thing I see is the big bump in matplotlib. Should we make that a 12 month policy like scipy and pandas? ( I don't have strong opinions on this)

This is super! I think it'd be fine to merge before all the dependency changes are in; I see this is getting to be a mammoth PR.

I'm +0.5 on moving to 0.14 given the benefits of this and the breaking changes. Any other thoughts?

RTD build successful! https://crusaderky-xarray.readthedocs.io/en/mindeps

I think I worked around it - logged in #3369

Not sure if the above is caused by the fact that I'm not running RTD in the main xarray repo?

I'm getting a crash in readthedocs because of versioneer: https://readthedocs.org/projects/crusaderky-xarray/builds/9748865/

In the RTD CI, versioneer incorrectly sets the version as 0+untagged.111.g6d60700, which while it is compliant to PEP-440, it obviously fails a minimum version check in pandas.DataFrame.to_xarray. On top of this, there's a bug in distutils (!!!) that causes an ugly crash instead of a clean error message:

``` pandas/compat/_optional.py(100)import_optional_dependency()

 98     if minimum_version:
 99         version = _get_version(module)

--> 100 if distutils.version.LooseVersion(version) < minimum_version: 101 assert on_version in {"warn", "raise", "ignore"}

``` if distutils.version.LooseVersion(version) < minimum_version:

To replicate locally: manually

``` xarray: 0+untagged.111.g6d60700 [...]

---

Exception in /home/docs/checkouts/readthedocs.org/user_builds/crusaderky-xarray/checkouts/mindeps/doc/pandas.rst at block ending on line 105 Specify :okexcept: as an option in the ipython:: block to suppress this message

TypeError Traceback (most recent call last) <ipython-input-10-7f86783d8836> in <module> ----> 1 s[::2].to_xarray()

~/checkouts/readthedocs.org/user_builds/crusaderky-xarray/conda/mindeps/lib/python3.7/site-packages/pandas/core/generic.py in to_xarray(self) 2904 speed (date, animal) int64 350 18 361 15 2905 """ -> 2906 xarray = import_optional_dependency("xarray") 2907 2908 if self.ndim == 1:

~/checkouts/readthedocs.org/user_builds/crusaderky-xarray/conda/mindeps/lib/python3.7/site-packages/pandas/compat/_optional.py in import_optional_dependency(name, extra, raise_on_missing, on_version) 98 if minimum_version: 99 version = _get_version(module) --> 100 if distutils.version.LooseVersion(version) < minimum_version: 101 assert on_version in {"warn", "raise", "ignore"} 102 msg = version_message.format(

~/checkouts/readthedocs.org/user_builds/crusaderky-xarray/conda/mindeps/lib/python3.7/distutils/version.py in lt(self, other) 50 51 def lt(self, other): ---> 52 c = self._cmp(other) 53 if c is NotImplemented: 54 return c

~/checkouts/readthedocs.org/user_builds/crusaderky-xarray/conda/mindeps/lib/python3.7/distutils/version.py in _cmp(self, other) 335 if self.version == other.version: 336 return 0 --> 337 if self.version < other.version: 338 return -1 339 if self.version > other.version:

TypeError: '<' not supported between instances of 'str' and 'int' ```

Ready for final review and merge.

I'm trying to test the readthedocs CI but it's getting stuck: https://readthedocs.org/projects/crusaderky-xarray/builds/ Maybe it's because I'm not a paying user?

Thanks @crusaderky . Can you replace the last line with just loffset=loffset https://github.com/pydata/xarray/blob/dd2b803a28cdf2f36210f5a7a46897cc47584ea6/xarray/core/common.py#L1043-L1049

@shoyer how do we test the readthedocs CI build before merging into main?

I have one outstanding error ```

  abs_diff = np.atleast_1d(abs(actual - expected)).astype(np.timedelta64)

E TypeError: Cannot cast datetime.timedelta object from metadata [Y] to [us] according to the rule 'same_kind'

xarray/tests/test_coding_times.py:122: TypeError ` The error disappears if I change the repositories of py36-min-all-deps from[conda-forge]to[anaconda, conda-forge]. I tried comparing the twoconda list`` but I could not spot any glaring differences. Help appreciated!

Evidence of the new CI script at work: https://dev.azure.com/xarray/xarray/_build/results?buildId=975&view=logs

Missing only CI integration of the new tool; waiting for #3366. All tests pass locally. You can start reviewing...

That's awesome, I've never seen this done so systematically—great idea!

Let's put dropping 3.5 fairly prominently in the release notes

Thanks @crusaderky !

Hello @crusaderky! 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! :beers:

Comment last updated at 2019-10-01 18:07:39 UTC

I was bored, so I wrote a CI tool to keep track of the situation :)

``` Package Required Policy Status

python 3.6 (2016-12-23) 3.6 (2016-12-23) = boto3 1.9 (2018-09-10) 1.9 (2018-09-10) = bottleneck 1.2 (2017-09-14) 1.2 (2017-09-14) = cartopy 0.17 (2018-11-17) 0.17 (2018-11-17) = cdms2 3.1 (2018-11-02) 3.1 (2018-11-02) = cfgrib 0.9 (2019-02-25) 0.9 (2019-02-25) = cftime 1.0 (2018-04-27) 1.0 (2018-04-27) = dask 1.2 (2019-04-13) 1.2 (2019-04-13) = distributed 1.27 (2019-04-12) 1.27 (2019-04-12) = h5netcdf 0.7 (2019-02-27) 0.7 (2019-02-27) = h5py 2.9 (2018-12-20) 2.10 (2019-09-10) < hdf5 1.10 (2017-09-22) 1.10 (2017-09-22) = iris 2.2 (2018-10-12) 2.2 (2018-10-12) = lxml 4.4 (2019-07-27) 4.4 (2019-07-27) = matplotlib 3.1 (2019-05-18) 3.1 (2019-05-18) = nc-time-axis 1.2 (2019-01-25) 1.2 (2019-01-25) = netcdf4 1.4 (2018-05-08) 1.4 (2018-05-08) = numba 0.44 (2019-06-18) 0.44 (2019-06-18) = numpy 1.14 (2018-01-09) 1.14 (2018-01-09) = pandas 0.21 (2017-11-04) 0.21 (2017-11-04) = pseudonetcdf 3.0 (2019-07-18) 3.0 (2019-07-18) = pydap 3.2 (2018-10-16) 3.2 (2018-10-16) = pynio 1.5 (2017-12-01) 1.5 (2017-12-01) = rasterio 1.0 (2018-12-21) 1.0 (2018-12-21) = scipy 1.0 (2017-11-07) 1.0 (2017-11-07) = seaborn 0.9 (2018-07-17) 0.9 (2018-07-17) = toolz 0.10 (2019-07-12) 0.10 (2019-07-12) = zarr 2.3 (2019-03-25) 2.3 (2019-03-25) = ```

A note: it seems impossible to install dask >=2.0 together with these old versions of numpy/pandas - at least with conda. I did not investigate why.