Sam Meder created MRUNIT-216:
--------------------------------
Summary: MR Unit's call to key grouping comparator passes
different data than Hadoop
Key: MRUNIT-216
URL: https://issues.apache.org/jira/browse/MRUNIT-216
Project: MRUnit
Issue Type: Bug
Affects Versions: 1.1.0
Reporter: Sam Meder
I'm packing a composite key into a BytesWritable and am combining this with a
grouping comparator and am finding that the bytes passed to the comparator are
different in MRUnit vs on a real Hadoop cluster. On the hadoop cluster I am
having to do things similar to the following comparator code from BytesWritable:
return compareBytes(b1, s1+LENGTH_BYTES, l1-LENGTH_BYTES,
b2, s2+LENGTH_BYTES, l2-LENGTH_BYTES);
i.e. need to strip some length bytes that are being passed along. In the mr
unit test on the other hand those bytes don't seemed to be passed to the
comparator...
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
