Add audformat.utils.join_labels()

[hagenw]

As discussed in #66 (review) this adds audformat.utils.join_labels() as a helper function to join scheme labels before merging two databases.

[frankenjoe]

This is a proposal of how to add the join_scheme_labels argument suggested in #62 (comment).

Puhh, good you pushed this early, a minute ago I started to work on the same issue :)

Glad to only have to review this one :)

[codecov]

Codecov Report

Merging #66 (a5eaef4) into master (022865e) will not change coverage.
The diff coverage is 100.0%.

Impacted Files	Coverage Δ
audformat/utils/__init__.py	100.0% <ø> (ø)
audformat/core/utils.py	100.0% <100.0%> (ø)

[frankenjoe]

Currently the following is possible:

labels_1 = {
    0: {'age': 20},
}

labels_2 = {
    '0': {'age': 30},
}

labels = audformat.utils.join_labels([labels_1, labels_2])
print(labels)

{0: {'age': 20}, '0': {'age': 30}}

So I wonder if we should check for same key type when we merge dictionaries? An easy way to do so is trying to create a Scheme from the combined labels, which will raise an error if the new labels are malformed:

audformat.Scheme(labels=labels)

ValueError: All labels must be of the same data type.

[frankenjoe]

Currently, it's not possible to update dictionaries with new values:

labels_1 = {
    0: {'age': 20},
}

labels_2 = {
    0: {'gender': 'male'},
}

labels = audformat.utils.join_labels([labels_1, labels_2])
print(labels)

ValueError: Values for key '0' are different:
{'age': 20}
!=
{'gender': 'male'}

It would be nice if we get:

{0: {'age': 20, 'gender': 'male'}}

[hagenw]

Currently, it's not possible to update dictionaries with new values:

labels_1 = {
    0: {'age': 20},
}

labels_2 = {
    0: {'gender': 'male'},
}

labels = audformat.utils.join_labels([labels_1, labels_2])
print(labels)

ValueError: Values for key '0' are different:
{'age': 20}
!=
{'gender': 'male'}

It would be nice if we get:

{0: {'age': 20, 'gender': 'male'}}

Good point, I now allow for overwriting and updated the documentation screenshot in the description.

[hagenw]

Currently the following is possible:

labels_1 = {
    0: {'age': 20},
}

labels_2 = {
    '0': {'age': 30},
}

labels = audformat.utils.join_labels([labels_1, labels_2])
print(labels)

{0: {'age': 20}, '0': {'age': 30}}

So I wonder if we should check for same key type when we merge dictionaries? An easy way to do so is trying to create a Scheme from the combined labels, which will raise an error if the new labels are malformed:

audformat.Scheme(labels=labels)

ValueError: All labels must be of the same data type.

I agree, will update the code accordingly.

[hagenw]

'0': {'age': 30}

Implemented it:

>>> audformat.utils.join_labels([{0: {'age': 20}}, {'0': {'age': 30}}])
...
ValueError: All labels must be of the same data type.

also added a test for it.