Fix Database.get() for non-existent schemes

[hagenw]

Closes #426

Add two tests that cover #246. The code itself is fixed by removing an if statement, that explicitly was setting an requested additional scheme to an empty frame, if the additional scheme requested before was also empty. I guess this was some left over code, that was no longer needed, but did also not fail the tests we had so far.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.0%. Comparing base (6b16209) to head (8ab1038).

Additional details and impacted files

Files	Coverage Δ
audformat/core/database.py	100.0% <100.0%> (ø)

[audeerington]

Thanks for fixing this!

I would guess that the original intention of the line is that you don't need to load the data at all to check for additional schemes if you already know that the length of the dataframe containing the main scheme is 0. So maybe this would save some time. But since obj is overwritten in the loop, this leads to the bug.

It seems to me that this case wouldn't come up very often, so I think it is fine to just remove the check like this.

[hagenw]

Good point.
I think, this is still of interest as it can save execution time. The problem before was, that I used the same variable name for the additional schemes. I updated the code now to use the structure from main, but use additional_obj as variable name for the additional schemes, instead of obj.

[audeerington]

Great, then everything looks good to me.