crunch-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Josh Wills (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CRUNCH-495) Fix case class/SpecificRecord interactions in Scrunch
Date Fri, 30 Jan 2015 00:40:34 GMT

     [ https://issues.apache.org/jira/browse/CRUNCH-495?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Josh Wills updated CRUNCH-495:
------------------------------
    Attachment: CRUNCH-495.patch

Patch w/the fix (just appends an "_" to the name of the generated record) and a test case
that will fail w/o the patch.

> Fix case class/SpecificRecord interactions in Scrunch
> -----------------------------------------------------
>
>                 Key: CRUNCH-495
>                 URL: https://issues.apache.org/jira/browse/CRUNCH-495
>             Project: Crunch
>          Issue Type: Bug
>          Components: Scrunch
>    Affects Versions: 0.11.0
>            Reporter: Josh Wills
>            Assignee: Josh Wills
>             Fix For: 0.12.0
>
>         Attachments: CRUNCH-495.patch
>
>
> So this is a fun one: I wrote a way to serialize case classes in Scala as Avro generic
records as part of the work for 0.11. However, if AvroMode.SPECIFIC is enabled on a MR job
(e.g., if you were doing a join between one PTable that contained specific record instances
and a different PTable that contained instances of a case class), the SpecificData object
in Avro will get confused when it sees the Avro schema I generate for the case class, b/c
the name of the Avro schema is identical to the name of the case class on the JVM, so Avro
will think that the record is an actual instance of a SpecificRecord.
> The solution I came up with is to slightly modify the name of the generated Avro generic
schema that corresponds to the case class so that it doesn't match the name of the case class
exactly so that Avro doesn't get confused.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message