[CRUNCH-139] PCollection#length doesn't always reduce the count to a single value

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.5.0
• Assignee: Gabriel Reid
• Reporter: Gabriel Reid
• Created: 07/Jan/13 16:21
• Updated: 22/Feb/13 06:57
• Resolved: 07/Jan/13 16:24

Description

PCollection#length doesn't explicitly set the number of reducers to 1, which means that the output of the counting mappers can be partitioned. This results only a partial count (i.e. an incorrect value) being returned in the length PObject if the input PCollection spans multiple default reduce partitions.