[ 
https://issues.apache.org/jira/browse/CRUNCH-358?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Tom White updated CRUNCH-358:
-----------------------------

    Attachment: CRUNCH-358.patch

I discovered this problem when  trying a later build of Parquet that has 
https://github.com/Parquet/parquet-mr/pull/299 in it. The problem I was hitting 
was that since the first type in the union was declared as an array of string 
the default value was being interpreted as an empty list, rather than null, 
which caused 
testProjectionSpecific(org.apache.crunch.io.parquet.AvroParquetFileSourceTargetIT)
 to fail.

I've attached a patch with a trivial fix. Crunch tests still pass with it, and 
testProjectionSpecific does too when using the later build of Parquet.

> Invalid optional field declaration in person.avsc
> -------------------------------------------------
>
>                 Key: CRUNCH-358
>                 URL: https://issues.apache.org/jira/browse/CRUNCH-358
>             Project: Crunch
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.9.0
>            Reporter: Tom White
>            Assignee: Tom White
>         Attachments: CRUNCH-358.patch
>
>
> The siblingnames field in person.avsc has a default value of null, even 
> though the first branch of the union is an array of string (not null). This 
> is not a valid Avro schema (see AVRO-1449) and should be fixed by reversing 
> the types in the union.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Reply via email to