Uniqueness of names in schemas

[KevinDCarlson]

Daniel Filonik suggests making it so that the same name can be used for more than one attribute in a schema. This is common in databases, for instance multiple tables might have an "address" column.

[epatters]

Having names be unique only within each hom-set is definitely a feature I've often found myself wanting. We might even have an old issue about this; I can't remember.

Implementation might be tricky because the assumption of global uniqueness of names in the schema is probably implicitly made all over the codebase.

[filonik-cmu]

Perhaps there is a way this could be implemented such that global uniqueness remains intact? I am thinking of some kind of "name mangling" that would automatically prepend the domain (which I am doing these days manually), i.e. turn :name into :Person_name. It's not pretty, but it might get the job done while minimizing changes to the codebase.

[TheCedarPrince]

Just bumping this discussion a bit after a discussion about a use case I am exploring. It would be nice for the API to differentiate attribute or hom names based on type specification. For example:

person_id::Hom(DeviceExposure, Person)
person_id::Hom(VisitOccurrence, Person)

Are two different hom's but the current API does not disambiguate these two leading to an error. I think having this would be great as it is less logistical overhead of ingesting database schemas that may share the same names across tables and let users have their schemas match exactly to their code.

This would especially be nice with the CatColab JSON exporting (and importing? 👀 ) functionality being worked on ToposInstitute/CatColab#303 .

Just some thoughts from my side. 😃