issue_comments: 317247671

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/pull/1473#issuecomment-317247671 https://api.github.com/repos/pydata/xarray/issues/1473 317247671 MDEyOklzc3VlQ29tbWVudDMxNzI0NzY3MQ== 6815844 2017-07-23T11:51:00Z 2017-07-23T11:51:00Z MEMBER

1 Backends support only "basic indexing "(int and slice). This is pretty common. 2 Backends support some of the "advanced indexing" use cases but not everything (e.g., restricted to most one list). This is also pretty common (e.g., dask and h5py). 3 Backends support "orthogonal indexing" instead of advanced indexing. NetCDF4 does this (but perform can be pretty terrible). 4 Backends support NumPy's fully vectorized "advanced indexing". This is quite rare -- I've only seen this for backends that actually store their data in the form of NumPy arrays (e.g., scipy.io.netcdf).

I am wondering what the cleanest design is. Because the cases 3 and 4 you suggested are pretty exculsive, I tried to distinguish cases 1, 2, and 4 in Variable._broadcast_indexes(key) method. For backends that only accept orthogonal indexing, I think case 4 indexers can be orthogonalized in each ArrayWrappers (by indexing._unbroadcast_indexes function).

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   241578773