PERFORMANCE: streaming data to the UDFs in foreach
--------------------------------------------------
Key: PIG-844
URL: https://issues.apache.org/jira/browse/PIG-844
Project: Pig
Issue Type: Improvement
Reporter: Olga Natkovich
Currently, Pig places the data passed to UDFs into a bag. This can cause the
process to use more memory than actually needed as in many cases it would be
better to push the data one tuple at a time to the UDFs.
For the case where combiner is invoked, this might not be that important;
however, for non-algebraic UDFs as well as other cases where combiner can't be
used, this can provide significant memory improvement.
Another possible use case is where the data is already grouped going into pig
and we don't need to group it again.
How this will effect UDF interface needs to be further discussed.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
