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/issues/1234#issuecomment-457209817,https://api.github.com/repos/pydata/xarray/issues/1234,457209817,MDEyOklzc3VlQ29tbWVudDQ1NzIwOTgxNw==,5635139,2019-01-24T14:11:12Z,2019-01-24T15:38:36Z,MEMBER,"> Alternatively, we might simply skip variables in Dataset.where if they don't have overlapping dimensions. The logic gets a good deal more complex, though. Currently where is a thin wrapper around np.where.

Would skipping variables without the dimension be preferable? Or do people expect the broadcast?

We recently had a discussion re skipping variables that don't have the dimension supplied using `.reduce`, although that was for a clearer example, where broadcasting would be less relevant - e.g. rolling along a dimension","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,203630267
https://github.com/pydata/xarray/issues/1234#issuecomment-275721597,https://api.github.com/repos/pydata/xarray/issues/1234,275721597,MDEyOklzc3VlQ29tbWVudDI3NTcyMTU5Nw==,1217238,2017-01-27T17:23:03Z,2017-01-27T17:23:03Z,MEMBER,"You're not the first person to be confused about this (see https://github.com/pydata/xarray/issues/1217). This is indeed working as expected, though maybe not in the most useful way.

`where` works by broadcasting its arguments, and when `ds.c` (along the `a` dimension) gets broadcasted with `ds.y` (along the `b` dimension), the result gets the union of both dimensions.

Alternatively, we might simply skip variables in Dataset.where if they don't have overlapping dimensions. The logic gets a good deal more complex, though. Currently `where` is a thin wrapper around `np.where`.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,203630267