issue_comments: 504058279

This data as json

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/2940#issuecomment-504058279 https://api.github.com/repos/pydata/xarray/issues/2940 504058279 MDEyOklzc3VlQ29tbWVudDUwNDA1ODI3OQ== 1217238 2019-06-20T14:55:04Z 2019-06-20T14:55:04Z MEMBER

Looking into this in more detail, the fix looks somewhat non-trivial. For example, we definitely are not padding by enough inside dask_rolling_wrapper if center=False: https://github.com/pydata/xarray/blob/145f25f69078f245313bb8e07b7c6af7509d0de8/xarray/core/dask_array_ops.py#L29

I'm not sure if this was ever tested properly, at least for dask arrays with multiple chunks. My guess is that this previously worked by consolidating dask arrays into a single chunk, which would simply fail for arrays that are larger than fit into memory.

For now, I think it would be safest to issue a new release that simply disables rolling() methods on dask arrays (by raising an error), and to save a full fix for later. I am concerned about letting the current behavior stick around, which silently calculates the wrong result.

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   440233667