Make labels unique

[DaniBodor]

Is it necessary to make labels (and names?) unique throughout a single Sequence? If we want to make the data machine searchable, then it's hard to deal with multiple data inputs (what's the generic term for all our data classes?) with identical labels. The same goes for names and human readability.
Would this potentially lead to problems if data is transposed from one Sequence to another? Is this even an operation that would make sense to do?

@psomhorst @wbaccinelli , what do you think?

[psomhorst]

I think it makes sense to make it a standard practice to have unique labels (and names?) throughout the Sequence. However, it is very difficult to enforce this, I think.

I think it would be a nice QoL addition to allow 'searching' or selecting a dataset without knowing what type it is. E.g. Sequence.data["airway_pressure"] or even Sequence["airway_pressure"] to refer to Sequence.continuous_data["airway_pressure"]. This could also be useful when adding data: Sequence.add(SparseData(...)) instead of Sequence.sparse_data.add(SparseData(...)).

Having unique labels also simplifies reporting, removing the requirement to explicitly state the collection type.

For each type of data (EIT, Continuous, Sparse and Interval) the label already needs to be unique within the data collection. However, when adding data with a label to DataCollection[EITData], there is no easy way to check whether that label already exists as DataCollection[Continuous/Sparse/IntervalData], because there is no reference to those within the DataCollection[EITData]. We could solve this by 'binding'* the DataCollection to a Sequence at creation of the Sequence (DataCollection.bound: Sequence | None). Then we have a reference to the other data collections and can enforce a unique ID.

*I chose the word bound as it also is used to refer to methods as 'function bound to objects', but there might be a better word for it.

[DaniBodor]

I didnt realize that within the collection it's already enforced to be unique. I think for now that resolves the most dramatic "problem". I agree that having it unique within the sequence (which is what I referred to in my original post) is the ideal, but not a priority for now.
It will probably indeed be a bit tricky/wordy, but should be doable I think once higher priority issues are resolved.

[psomhorst]

The .data-object (#364) enforces uniqueness if data is added through that protocol.

My main question is still: how to generate labels. I propose you should always be able to pass label="<str>" to a function that generates any code, and the label can be updated using replace(). However, if no explicit label is passed, should an implicit label be set? E.g. "filtered impedance a2cf1" (name + short sha1 hash). Or should it be empty by default? The same goes when copying an object. A label like "copy of <some other label>" is quite verbose. Maybe "<some other label> copy <hash>"?