[PIG-3133] Revamp algebraic interface to actually return classes

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Fix Version/s: 0.18.0
• Assignee: Unassigned
• Reporter: Jonathan Coveney
• Created: 23/Jan/13 00:48
• Updated: 26/May/17 22:35

Description

The current algebraic interface is a bit weird to work with. It would make a lot more sense to let people return Class<? extends EvalFunc<Tuple>> or what have you, or even a FuncSpec, but the current string based approach circumvents the whole point of using Java and is annoying. I think we should have an abstract EFInitial, EFIntermediate, EFFinal which implemented the exec function for the user, but in terms of a simpler, clearer interface. This way if people really want the old way they can, but we can present them something less ugly.

This would also be a good time to clarify the contracts of Algebraics and simplify them (the initial function's "a tuple which contains a bag which contains 1 tuple" is super whack).

If anyone wants to work on this let me know because this is the sort of thing I will probably bang out when procrastinating something else.