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/2236#issuecomment-398931150,https://api.github.com/repos/pydata/xarray/issues/2236,398931150,MDEyOklzc3VlQ29tbWVudDM5ODkzMTE1MA==,1217238,2018-06-20T23:42:04Z,2018-06-20T23:42:04Z,MEMBER,">> A module of bottleneck/numpy functions that act on numpy arrays only.
>>A module of functions that act on numpy or dask arrays (or these could be moved into duck_array_ops).

> Could you explain more detail about this idea?

OK, let me try:

1. On numpy arrays, we use bottleneck eqiuvalents of numpy functions when possible because bottleneck is faster than numpy
2. On dask arrays, we use dask equivalents of numpy functions.
3. We also want to add some extra features on top of what numpy/dask/bottleneck provide, e.g., handling of `min_count`

We could implement this with:
- `nputils.nansum()` is equivalent to `numpy.nansum()` but uses `bottleneck.nansum()` internally instead when possible.
- `duck_array_ops.nansum()` uses `numpy_nansum()` or `dask.array.nansum()`, based upon the type of the inputs.
- `duck_array_ops.sum()` uses `numpy.sum()` or `dask.array.sum()`, based upon the type of the inputs.
- `duck_array_ops.sum_with_mincount()` adds `mincount` and `skipna` support and is used in the `Dataset.sum()` implementation. Its is written using `duck_array_ops.nansum()`, `duck_array_ops.sum()`, `duck_array_ops.where()` and `duck_array_ops.isnull()`.
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,333248242
https://github.com/pydata/xarray/pull/2236#issuecomment-398150942,https://api.github.com/repos/pydata/xarray/issues/2236,398150942,MDEyOklzc3VlQ29tbWVudDM5ODE1MDk0Mg==,1217238,2018-06-18T18:28:58Z,2018-06-18T18:28:58Z,MEMBER,"Very nice!

> In my implementation, bottleneck is not used when skipna=False.
bottleneck would be advantageous when skipna=True as numpy needs to copy the entire array once, but I think numpy's method is still OK if skipna=False.

I think this is correct -- bottleneck does not speed up non-NaN skipping functions.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,333248242