[
https://issues.apache.org/jira/browse/CRUNCH-239?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13865115#comment-13865115
]
Josh Wills commented on CRUNCH-239:
-----------------------------------
So I modeled UnionTuple after TupleN in the cogroup case; i.e., where we were
writing a TupleN that had at most one non-null value. TupleN.size() returns
values.length (where values is an Object[]), so that was how I came down there.
Re: Union implementing Tuple, that's more of an artifact of how I did the
initial refactoring than anything else; I don't know that there's any obvious
upside to having Union implement Tuple in the general case. I'm fine to drop
that if you think it's cognitive overhead we don't really need.
> Add support for unions of PTypes
> --------------------------------
>
> Key: CRUNCH-239
> URL: https://issues.apache.org/jira/browse/CRUNCH-239
> Project: Crunch
> Issue Type: Improvement
> Components: Core
> Reporter: Josh Wills
> Assignee: Josh Wills
> Attachments: CRUNCH-239.patch
>
>
> There are a few core use cases (e.g., cogroups, as in CRUNCH-174) where it
> would be useful to construct a union of multiple PTypes into a single PType,
> where any given observation could only be of one of multiple/optional types.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
