[ 
https://issues.apache.org/jira/browse/PIG-805?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12996051#comment-12996051
 ] 

Olga Natkovich commented on PIG-805:
------------------------------------

Here is what we have now:

(1) We publish Pig without any dependencies to maven since all the dependencies 
can be picked up from there. This is mostly intended for compiling UDFs or any 
other code that depends on Pig
(2) Through our regular build process we produce 3 jars:
- fully bundled jar - for ease of use
- bundled jar without hadoop - to be able to run with different versions of 
hadoop
- completely unbundled one for (1)
(3) In our release we only include the fully bundled one for a number of 
reasons:
- historically that's all we have done and up till now we have not hand 
requests for anything else
- Currently, any release of Pig only works and with a single major version of 
Pig. Minor versions are suppose to be backward compatible and should not 
require a recompile (though in the past we have experienced problems with 
this.) We only test with this one bundled version so we can't make guarantees 
that the release will work with anything else.

I am ok if we want, in addition to the fully bundled pig.jar to include the one 
without hadoop in Pig 0.9 release with the caveats listed above and also the 
concern that our already large release distribution will get even larger.

> removing dependency on consolidated hadoop.jar from pig.jar
> -----------------------------------------------------------
>
>                 Key: PIG-805
>                 URL: https://issues.apache.org/jira/browse/PIG-805
>             Project: Pig
>          Issue Type: Improvement
>            Reporter: Olga Natkovich
>            Assignee: Giridharan Kesavan
>
> The proposal is to
> - for compilation always use dependencies from Ivy
> - for packaging to have 2 targets:
> (1) The current jar target will not package it and rely on having it 
> available at runtime
> (2) Add a new target jar_with_hadoop that does what the current jar does
> (3) Add a property that allows to build the release tar with and without 
> hadoop.jar embeded into pig.

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to