DataQuery matches against all unmodified DataIDs

[djhoese]

Describe the bug
I'm working on the nucaps reader and am getting exceptions of "TooManyResults". This happens when the DependencyTree is trying to load composite dependencies so it searches for the wavelength for one composite in the reader datasets. The DataID matching is matching against all DataIDs even though none of them actually match because it is only using the modifiers key set to an empty tuple.

To Reproduce

In [9]: from satpy.dataset import ModifierTuple

In [10]: from satpy import DataID, DataQuery

In [11]: query1 = DataQuery(wavelength=1.6, modifiers=())

In [12]: query2 = DataQuery(wavelength=1.6)

In [13]: data_id = DataID({"name": {"required": True}, "level": {}, "modifiers": {"default": [], "type": ModifierTuple}}, name="test", modifiers=tuple())

In [14]: query1.filter_dataids([data_id])
Out[14]: [DataID(name='test', modifiers=())]

In [16]: query2.filter_dataids([data_id])
Out[16]: []

Expected behavior
No match and just ignore the composite when checking if it can be loaded.

Actual results
See above. When loading the actual Scene I see:

WARNING  : Too many possible datasets to load for DataQuery(wavelength=0.85, modifiers=())

Additional context
The nucaps reader does not define wavelength in its DataID. I'm going to try and define it and see if that fixes this.

[djhoese]

Adding this to the reader yaml fixes it:

    wavelength:
      type: !!python/name:satpy.dataset.dataid.WavelengthRange

What are your thoughts @mraspaud? It would seem every reader with a custom DataID needs this defined.

[mraspaud]

That's not good, and not the way I envisioned this being used. The wavelength should not be needed for custom DataIDs to work.

[mraspaud]

I'm pretty sure that's because the wavelength is being hardcoded somewhere in the DataID comparisons, but unfortunately I don't have time to look a this before the week end

[djhoese]

I'll see if I can figure it out. Now looking at it it isn't quite clear to me why adding wavelength to the reader's identifiers fixes this. 🤔 Actually maybe it makes sense...I'll see what I can do.