[
https://issues.apache.org/jira/browse/PIG-1868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13007663#comment-13007663
]
Daniel Dai commented on PIG-1868:
---------------------------------
The reason for the failure on trunk is because in this query,
TransformToMyDataType returns more specific error message than user declared
schema "seq: { t: ( previous, current, next ) }", and now we only take user
declared schema. This is a regression, I will attach a patch to fix it.
> New logical plan fails when I have complex data types from udf
> --------------------------------------------------------------
>
> Key: PIG-1868
> URL: https://issues.apache.org/jira/browse/PIG-1868
> Project: Pig
> Issue Type: Bug
> Components: impl
> Affects Versions: 0.8.0
> Reporter: Vivek Padmanabhan
> Assignee: Daniel Dai
> Fix For: 0.8.0
>
>
> The new logical plan fails when I have complex data types returning from my
> eval function.
> The below is my script :
> {code}
> register myudf.jar;
> B1 = load 'myinput' as (id:chararray,ts:int,url:chararray);
> B2 = group B1 by id;
> B = foreach B2 {
> Tuples = order B1 by ts;
> generate Tuples;
> };
> C1 = foreach B generate TransformToMyDataType(Tuples,-1,0,1) as seq: { t: (
> previous, current, next ) };
> C2 = foreach C1 generate FLATTEN(seq);
> C3 = foreach C2 generate current.id as id;
> dump C3;
> {code}
> On C3 it fails with below message :
> {code}
> Couldn't find matching uid -1 for project (Name: Project Type: bytearray Uid:
> 45 Input: 0 Column: 1)
> {code}
> The below is the describe on C1 ;
> {code}
> C1: {seq: {t: (previous: (id: chararray,ts: int,url: chararray),current: (id:
> chararray,ts: int,url: chararray),next: (id: chararray,ts: int,url:
> chararray))}}
> {code}
> The script works if I turn off new logical plan or use Pig 0.7.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
