[
https://issues.apache.org/jira/browse/PIG-3679?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Cheolsoo Park updated PIG-3679:
-------------------------------
Attachment: PIG-3679-4.patch
I am uploading yet another patch that fixes the e2e test. This time, I took
Daniel's patch and added a null check to the exec() of StreamingUDF.
On a second thought about backward incompatibility of PIG-3568, looks like
we're safe with changing POUserFunc to filter out nulls for the following
reasons-
# The changes in all the other PO operators except POUserFunc should be
transparent to users because in the end, the result remains unchanged. The only
difference is whether nulls travel the full or part of pipeline.
# Since POUserFunc now filters out nulls, some UDFs that used to throw NPE will
no longer throw NPE (e.g. DoubleRound). But this is more of an improvement than
an backward incompatible change.
So I will commit PIG-3679-4.patch if there is no objection. Thanks!
> e2e StreamingPythonUDFs_10 fails in trunk
> -----------------------------------------
>
> Key: PIG-3679
> URL: https://issues.apache.org/jira/browse/PIG-3679
> Project: Pig
> Issue Type: Bug
> Reporter: Cheolsoo Park
> Assignee: Cheolsoo Park
> Fix For: 0.13.0
>
> Attachments: PIG-3679-1.patch, PIG-3679-2.patch, PIG-3679-3.patch,
> PIG-3679-4.patch
>
>
> The e2e test StreamingPythonUDFs_10 fails in trunk with NPE-
> {code}
> Caused by: java.lang.NullPointerException
> at org.apache.pig.builtin.DoubleRound.exec(DoubleRound.java:45)
> {code}
> The test query is as follows-
> {code}
> a = load '/user/pig/tests/data/singlefile/allscalar10k' using PigStorage() as
> (name:chararray, age:int, gpa:double, instate:chararray);
> b = foreach a generate name,
> ((double)ROUND((instate=='true'?gpa:gpa+1)*10000)) / 10000.0;
> store b into
> '/user/pig/out/cheolsoop-1390330024-nightly.conf-StreamingPythonUDFs/StreamingPythonUDFs_10_benchmark.out';
> {code}
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
