[ 
https://issues.apache.org/jira/browse/DATAFU-17?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13877843#comment-13877843
 ] 

Will Vaughan commented on DATAFU-17:
------------------------------------

For many of these UDFs we are using probabilistic tests that can be slow.  
Deterministic tests could be faster in some situations, but writing 
deterministic tests for these functions requires setting a seed.  While some 
UDFs take a constructor parameter for the seed, we do not have a standardized 
way setting a seed for those that do not take a constructor parameter.

I recommend setting it via a job configuration parameter.

> Improve testing of randomized functions
> ---------------------------------------
>
>                 Key: DATAFU-17
>                 URL: https://issues.apache.org/jira/browse/DATAFU-17
>             Project: DataFu
>          Issue Type: Bug
>            Reporter: Will Vaughan
>
> We have a large number of UDFs with a random component that are difficult and 
> often slow to test.  We should improve our testing standards and capabilities 
> for this class of functions.



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

Reply via email to