groupby_bins sorts bin labels as strings

[shoyer]

This sometimes gets them out of order:

import xarray as xr
import numpy as np
data = xr.DataArray(np.arange(100), dims='x', coords={'x': np.linspace(-100, 100, num=100)})
print(data.groupby_bins('x', bins=11).mean().to_series())

x_bins
(-100.2, -81.818]      4.5
(-27.273, -9.0909]    41.0
(-45.455, -27.273]    32.0
(-63.636, -45.455]    23.0
(-81.818, -63.636]    14.0
(-9.0909, 9.0909]     50.0
(27.273, 45.455]      68.0
(45.455, 63.636]      77.0
(63.636, 81.818]      86.0
(81.818, 100]         95.0
(9.0909, 27.273]      59.0
dtype: float64

We should pass through sort=False to pd.factorize in groupby.unique_value_groups when using bins to avoid this.

CC @chrisroat @rabernat

[rabernat]

Good catch. Should be an easy fix. I can put together a PR.