lbooker42
commented
6 months ago From #5277, implementation notes:
Do we want to move from exposing Namespace and TableIdentifier and just use Strings?
I think Java uses the POJOs, since they can be made from Strings by the user.
Python, I THINK, should accept a single string (and delegate to TableIdenfier.parse) unioned with a sequence of strings (TableIdentifier.of).
We may have to do our own splitting for methods that take just a Namespace, since there's no Namespace.parse, but it's literally just dot-split.
Provide access to the Iceberg catalog adapter and tools for python users.
Should support S3 + REST and AWSGlue as part of the initial implementation.