Inability to slice with a step messes with the download model about dkist HOT 3 CLOSED

In the process of working on the acceptance tests I realised the following workflow also doesn't work:

ds = Dataset...

ds[0].download()

ds.plot()

i.e. I want to keep working with this large dataset, but only download some of the files (for now?)

My initial thought on this is to have a .files or similar attribute which is where the .download() method lives. This .files attribute could be sliceable, but also could allow swapping the root path (this would involve linking to the .data) and showing the files which are present and which are not downloaded in some nice summary.

from dkist.

What we want is for all subsets of a dataset (i.e. all slicing operations) to keep the same files object so that they all share the same root path and downloads are cumulative.

from dkist.

This is now fixed by enabling slicing of the .files property and not the NDCube

from dkist.