Weird behaviour of keep_attrs in xr.where

[yt87]

I want to categorize DataArray values with xr.where. Here is a simple case:

import numpy as np
import xarray as xr
ds = xr.Dataset(
    {'sst': ('lat', 0.1 * np.arange(10), {'standard_name': 'sea_surface_temperature'})},
    coords={"lat": ('lat', np.arange(10), {'axis': 'Y'})},
)
da = xr.where(ds['sst'] < 0.5, 0, 1, keep_attrs=True)
da.attrs
da.lat.attrs
{}
{}

I understand that da.attrs should be empty, number 0 has no attributes. However, coordinate attributes are also lost. What surprised me is the output when I set keep_attrs=False:

da2 = xr.where(ds['sst'] < 0.5, 0, 1, keep_attrs=False)
da2.attrs
da2.lat.attrs
{}
{'axis': 'Y'}

Here, coordinate attributes are preserved. This is what I want. Can I rely on this behaviour?

[yt87]

I think this can be justified by interpreting keep_attrs=True as: copy attributes of x, otherwise use those of cond. Maybe use_attrs with options 'x', 'y' and 'cond' (or None) would be a better name.