🚸 Unique contraint on name field of most registries

[falexwolf]

The Feature and ULabel registries have a unique constraint on the name field:

• Feature
• ULabel

Let's review all other label registries towards whether they also have it.

Note: It's important to note that versioned entities shouldn't have that unique constraint because multiple versions of the same entity should be able to have the same name; e.g. Transform and Collection.

Internal Slack ref.

[falexwolf]

Indicating this in the docs:

• laminlabs/lnschema-core@589db75

[Zethson]

I looked at wetlab. It has:

1. Compound: We're using the Drug ontology for it. Tricky with different Source.
2. Experiment: I think you can have the same experiment with different dates. Not sure whether we should make name unique here.
3. Well: IDEK whether they need a name in the first place but it should be okay here. If anything, row and column should be (jointly)? unique here.
4. PerturbationTarget: Guess it's okay to have unique name.
5. GeneticPerturbation: You can target the same genes with different CRISPR sequences. Users would therefore potentially be forced to encode the sequences in the name which is bad. Not sure.
6. CompoundPerturbation: Same compound name with different concentrations surely happens in experimental setups. Learning more towards a "no".
7. EnvironmentalPerturbation: Same argument. You can use different temperatures etc. "no".
8. CombinationPerturbation: Technically hidden concentrations but w/e.
9. BioSample: Can have different batches and we might need to encode the batches in the name then. But okay I guess.
10. TechSample: Different batches but okay.
11. Donor: No. Names cannot be unique.

My conclusion is that while we could do it, it's always going to be a case by case decision and there's quite some grey area. I fear that this would lead to inconsistencies and unexpected behavior.