Collections

[stuartmcalpine]

Similar to Rucio, we could add "collections". An ability to tag datasets and executions jointly under a single collection.

For example datareg.Registrar.collection.register(dataset_ids=[1,2,3], execution_ids=[4,5,6]).

This would be a way of tagging all the components of a single pipeline for example, making it easier to query.

[JoanneBogart]

Some form of collections could address a number of potential problem use cases. I can see uses for collections of datasets and for collections of executions. I'm not convinced datasets and executions should be lumped together in the same collection, however. Given an execution collection, one can already find all the datasets associated with them. Also, at some level the implementation will involve foreign keys, which will have to be defined as referencing a specific table.