html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue
https://github.com/pydata/xarray/pull/1847#issuecomment-364293763,https://api.github.com/repos/pydata/xarray/issues/1847,364293763,MDEyOklzc3VlQ29tbWVudDM2NDI5Mzc2Mw==,6815844,2018-02-09T00:18:52Z,2018-02-09T00:19:33Z,MEMBER,"It looks that our `where` does not do type-casting (inheriting from numpy's where),
```python
In [1]: import numpy as np
   ...: import xarray as xr
   ...: a = xr.Variable('x', np.array(['a', 'b', 'c']))

In [2]: a._getitem_with_mask([1, 2, -1])
Out[2]:
<xarray.Variable (x: 3)>
array(['b', 'c', 'nan'],
      dtype='<U32')
```
nan is converted to string.

I guess we need to cast arrays manually in `_getitem_with_mask` before `where` is applied.
```python
In [9]: dtype, fill_value = xr.core.dtypes.maybe_promote(a.dtype)
   ...: a.astype(dtype)._getitem_with_mask([1, 2, -1])
Out[9]: 
<xarray.Variable (x: 3)>
array(['b', 'c', nan], dtype=object)
```

","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,290320242