[jira] [Commented] (PIG-4648) JRuby UDFs should have a more efficient implementation

["Brian Johnson (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/PIG-4648?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14679478#comment-14679478 ] 

Brian Johnson commented on PIG-4648:
------------------------------------

actually I think the two can be merged to work together so you can have "native" udfs alongside ruby udfs 

> JRuby UDFs should have a more efficient implementation
> ------------------------------------------------------
>
>                 Key: PIG-4648
>                 URL: https://issues.apache.org/jira/browse/PIG-4648
>             Project: Pig
>          Issue Type: New Feature
>            Reporter: Brian Johnson
>            Assignee: Brian Johnson
>            Priority: Minor
>
> The current jruby UDF implementation translates all the Java objects to ruby objects and back. This makes it very simple for ruby programmers, but it is very inefficient. I propose a second jruby implementation that passes through the Java objects to avoid this extra overhead. It should be able to re-use a lot of the existing code. The existing implementation is under org.apache.pig.scripting.jruby. I propose that the new implementation go under org.apache.pig.scripting.jruby.native and will be referenced with "using jruby-native". This is meant to reference "native" pig classes. I realize it could be confused with the Java native interface, but I couldn't think of a better name.



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