scipy 2025: xarray computations and workflows for hierarchical data

[keewis]

The main topic of the second section of the scipy 2025 tutorial is

Understand how to map typical xarray computations and workflows over hierarchical data

and the proposed subsections are

• Non-datatree computations
• Datatree computations
  • map_over_subtree
  • Escape hatches

I'll use this issue as a way to figure out more details, and to make the writing of this section as efficient as possible.

Here's a checklist (feel free to edit):

• Non-datatree computations
  • indexing (isel, sel), nan-ops, arithmetic
  • reductions (mean, median, argmin / idxmin, count)
  • groupby?
• Datatree computations
  • map_over_subtree
  • Escape hatches

It might also be interesting to have a figure that shows the (slightly inaccurate but really useful) mental model of a "tree of dataset objects", and use that to explain how operations apply to a DataTree object.

[keewis]

@TomNicholas, I assume "non-datatree computations" refers to operations on Dataset / DataArray / Variable? If so, I guess we have most if not all of the material already, and would only need to pick a few representative examples. Then we'd need to translate these using map_over_datasets to DataTree (or the DataTree methods).

[TomNicholas]

Yes we should be able to re-use plenty of material. If you see any way to just re-use entire sections then just do that - we can just link to whatever sections of content and present them in whatever order we like.