Meaningfully compare Scene and MultiScene objects for equality

[gerritholl]

Feature Request

Is your feature request related to a problem? Please describe.

For a unit test, I wanted to ensure that my code was producing the right MultiScene object, so I tested for equality. However, neither Scene nor MultiScene objects have useful equality checks implemented. The following both evaluate to False:

• Scene() == Scene()
• MultiScene() == MultiScene()

Describe the solution you'd like

Scene and MultiScene objects are containers. Either containers should compare equal if all their contents and, if applicable, metadata compare equal; or the equality operator should itself be a container, like for numpy arrays or xarray datasets.

sc1 = satpy.tests.utils.make_fake_scene({"a": arange(25).reshape(5, 5)})
sc2 = satpy.tests.utils.make_fake_scene({"a": arange(25).reshape(5, 5)})
print(sc1 == sc2)   # False
(sc1.to_xarray_dataset() == sc2.to_xarray_dataset()).all()  # True (as xarray Dataset)

Describe any changes to existing user workflow

• Currently comparing Scene or MultiScene objects reduces to object identity. The meaning would change.
• Currently such comparisons are very cheap. If the equality operator compared the content, the comparison would become quite expensive. If triggering a dask computation, it would become very expensive.

Additional context

This would have to trigger a dask computation, which might be problematic.

[djhoese]

My preference would be to leverage xarray here as you have in the various comparison methods. However, not all Scene's can be turned into Dataset objects (different areas/resolutions) iirc. And a MultiScene will likely always be generator or a series of dask Delayed objects (my hope in the future) so I'm not sure how accurate it would be to compare them.

[gerritholl]

It cannot always be a Dataset, but can it always be a dictionary? The keys of a Scene are always DataID, and DataID is always hashable?

[djhoese]

True.