`mean` returns empty DataArray for `groupby_bins` containing `datetime64`

[observingClouds]

What happened?

Applying the mean operator to an xarray dataset that has been grouped by bins, removes variables containing values of type datetime64.

What did you expect to happen?

I expect that all variables are preserved after applying the mean-operator. For min and max this works as expected. sum is not meaningful, so it makes sense that the variable is dropped in that case as well.

Minimal Complete Verifiable Example

import numpy as np
import pandas as pd
import xarray as xr

times = pd.date_range('2020-01-01', '2020-02-01', freq='1H')
index = np.arange(len(times))
bins = np.arange(0,len(index),5)
ds=xr.Dataset({'time':('index', times),'float':('index', np.linspace(0,1,len(index)))},coords={'index':index})

ds.groupby_bins('index', bins).mean()
<xarray.Dataset>
Dimensions:     (index_bins: 148)
Coordinates:
  * index_bins  (index_bins) object (0, 5] (5, 10] ... (730, 735] (735, 740]
Data variables:
    float       (index_bins) float64 0.004032 0.01075 0.01747 ... 0.9852 0.9919

ds.groupby_bins('index', bins).min()
<xarray.Dataset>
Dimensions:     (index_bins: 148)
Coordinates:
  * index_bins  (index_bins) object (0, 5] (5, 10] ... (730, 735] (735, 740]
Data variables:
    time        (index_bins) datetime64[ns] 2020-01-01T01:00:00 ... 2020-01-3...
    float       (index_bins) float64 0.001344 0.008065 0.01478 ... 0.9825 0.9892

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.

Relevant log output

No response

Anything else we need to know?

No response

Environment

INSTALLED VERSIONS ------------------ commit: None python: 3.10.6 | packaged by conda-forge | (main, Aug 22 2022, 20:43:44) [Clang 13.0.1 ] python-bits: 64 OS: Darwin OS-release: 21.6.0 machine: x86_64 processor: i386 byteorder: little LC_ALL: None LANG: None LOCALE: ('en_US', 'UTF-8') libhdf5: None libnetcdf: None

xarray: 2022.6.0
pandas: 1.4.4
numpy: 1.23.2
scipy: None
netCDF4: None
pydap: None
h5netcdf: None
h5py: None
Nio: None
zarr: None
cftime: None
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: None
distributed: None
matplotlib: 3.5.3
cartopy: None
seaborn: None
numbagg: None
fsspec: 2022.8.2
cupy: None
pint: None
sparse: None
flox: None
numpy_groupies: None
setuptools: 65.3.0
pip: 22.2.2
conda: None
pytest: None
IPython: 8.4.0
sphinx: None

[observingClouds]

This issue might be a duplicate of #5897 and it continues to exist in version 2022.09.0.

[kmuehlbauer]

This is still an issue, as datetimes are handled as "non_numeric" and are just skipped for mean/median etc.

A workaround would be to call reduce directly:

import numpy as np
import pandas as pd
import xarray as xr

times = pd.date_range('2020-01-01', '2020-02-01', freq='1H')
index = np.arange(len(times))
bins = np.arange(0,len(index),5)
ds=xr.Dataset({'time':('index', times),'float':('index', np.linspace(0,1,len(index)))},coords={'index':index})

ds.groupby_bins('index', bins).reduce(xr.core.computation.duck_array_ops.mean, dim="index", numeric_only=False)

@dcherian Is working on a solution, if I interpret the linked commits correctly.

[dcherian]

I did look in to it. I think we have to choose between

1. Have .reduce(..., numeric_only=True) pass datetime through.
2. Expose numeric_only in top-level aggregations with default True.
3. Add a new datetime_is_numeric: bool kwarg (or similar) to reduce and set it appropriately.