Similarity to DataTree

[TomNicholas]

Hi @andersy005 - @jhamman showed me this repo earlier, and I've just had a look through it.

Obviously xcollection is similar to datatree, but the one-level dictionary-like design is also close to Arviz's InferenceData class (which datatree aims to replace, see arviz-devs/arviz#2015).

There are some things in xcollection that would be good inspiration for datatree, in particular:

• filter and choose methods:
  Implementing something similar to xcollections' existing filter and choose methods could solve API for filtering / subsetting xarray-contrib/datatree#79 (also filter_by_attrs Add an equivalent method to Dataset.filter_by_attrs #4 )

• plotting
  I also would like to adding some plotting functionality, but I'm also unsure what that should return (ref Add plot() function #13 and Plotting methods xarray-contrib/datatree#47)

• weighted
  DataTree should have all the same xarray aggregations (e.g. weighted) but I haven't got around to implementing that yet. I don't think I've made an issue for it either yet.

Beyond those methods I think datatree supplants xcollection.

If anyone here is interested in contributing to datatree at all then that would be great! The idea is to build it out in xarray-contrib before merging it upstream into xarray.

[andersy005]

Beyond those methods I think datatree supplants xcollection.

@TomNicholas, this is 💯 right. We started working xcollection when datatree was still an idea. Once you began working on datatree, we slowed down on xcollection development since most of the features we wanted were going to be in datatree. I think the right thing to do is (1) join effort by helping you with the development of datatree, (2) deprecate/archive xcollection.

Ccing @ncar-xdev/xcollection for thoughts/feedback...

[mgrover1]

@TomNicholas I would be interested in chatting soon. I am looking into using these hierarchal datasets with weather radar data in Py-ART, and have setup a prototype repo for exploration pyart-xarray-sandbox.

I tried both this xcollection and datatree approach which are almost identical, but would like to migrate towards datatree to collaborate with the rest of the Pangeo community. Do you have a place for "user stories" or "use cases"?

[TomNicholas]

@mgrover1 that sounds great - please feel free to add anything you think might be useful as documentation to the DataTree repo. I think we should just collect useful resources for now and worry about the "right" place to put it later.

[mgrover1]

Great! Will do!