[
https://issues.apache.org/jira/browse/PIG-1776?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Thejas M Nair updated PIG-1776:
-------------------------------
Attachment: PIG-1776.1.patch
The root cause of the problem was that the UDFContext objects were not reset
between plan regenerations. The explain command in the query set the
requiredfields property of the load function to require only first two fields.
When the plan was regenerated during the dump command, the optimizer rules
figured that all columns in load statement are required, and it did not set the
requiredfields property.
To fix this bug, each time a clone of the logical plan is created for
regenerating the plan, the UDFContext is being reset.
> changing statement corresponding to alias after explain , then doing dump
> gives incorrect result
> ------------------------------------------------------------------------------------------------
>
> Key: PIG-1776
> URL: https://issues.apache.org/jira/browse/PIG-1776
> Project: Pig
> Issue Type: Bug
> Affects Versions: 0.8.0
> Reporter: Thejas M Nair
> Assignee: Thejas M Nair
> Fix For: 0.8.0
>
> Attachments: PIG-1776.1.patch
>
>
> {code}
> grunt> a = load '/tmp/t2.txt' as (str:chararray, num1:int, alph : chararray);
> grunt> dump a;
> (ABC,1,a)
> (ABC,1,b)
> (ABC,1,a)
> (ABC,2,b)
> (DEF,1,d)
> (XYZ,1,x)
> grunt> c = foreach b generate group.str, group.$1, COUNT(a.alph) ;
> grunt> dump c; -- gives correct results
> (ABC,1,3)
> (ABC,2,1)
> (DEF,1,1)
> (XYZ,1,1)
> /* but dumping c after following steps gives incorrect results */
> grunt> c = foreach b generate group.$0 , (CHARARRAY)group.$1;
>
> grunt> explain c;
> ...
> ...
> grunt> c = foreach b generate group.str, group.$1, COUNT(a.alph) ;
> grunt> dump c;
> (ABC,1,0)
> (ABC,2,0)
> (DEF,1,0)
> (XYZ,1,0)
> {code}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
