Jeremy Beard created CRUNCH-464: ----------------------------------- Summary: Empty PCollection returns null Long for length().getValue() Key: CRUNCH-464 URL: https://issues.apache.org/jira/browse/CRUNCH-464 Project: Crunch Issue Type: Bug Components: Core Affects Versions: 0.10.0 Reporter: Jeremy Beard Assignee: Josh Wills
PCollection lengths are determined by summing 1s for each record, but when the PCollection is empty there is no sum and so the PObject returns a null length. Fix is to emit a 0 at the cleanup() of each DoFn instantiation, forcing a sum when there are no input records. -- This message was sent by Atlassian JIRA (v6.2#6252)