[ https://issues.apache.org/jira/browse/CRUNCH-315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13857929#comment-13857929
]
Micah Whitacre commented on CRUNCH-315:
---------------------------------------
I would assume it would have to be serialized to disk. Most of the time in our cases where
we've needed a collection like that since the data size is typically a lot smaller we've done
mapside joins or injected that in memory collection into the DoFn with custom serialization
logic if needed.
> Empty collection
> ----------------
>
> Key: CRUNCH-315
> URL: https://issues.apache.org/jira/browse/CRUNCH-315
> Project: Crunch
> Issue Type: New Feature
> Reporter: Chao Shi
> Attachments: CRUNCH-315.patch
>
>
> As discussed in the mailing list [1] and [2], I'd like to add an empty collection feature.
On the API side, I think we can add a new method in Pipeline to create an empty collection.
The collection should be a subclass of PCollection and behaves like other normal PCollecitons.
There are also some optimization points that Josh mentioned in [2].
> I haven't thought it clearly. Just put a ticket here and see if anyone else has a better
idea.
> [1] http://mail-archives.apache.org/mod_mbox/crunch-dev/201312.mbox/%3CBLU0-SMTP1337A04FAC6B5F497F7473EADC10%40phx.gbl%3E
> [2] http://mail-archives.apache.org/mod_mbox/crunch-dev/201312.mbox/%3CCAH29n6MbSK9gapoC2DgVnhofjAobyasCuZh_0475DuSajV%3DCPg%40mail.gmail.com%3E
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
|