[PIG-4001] POPartialAgg aggregates too aggressively when multiple values aggregated

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.13.0
• Assignee: Travis Woodruff
• Reporter: Travis Woodruff
• Created: 09/Jun/14 18:58
• Updated: 19/Jun/14 23:59
• Resolved: 19/Jun/14 23:59

Description

POPartialAgg.aggregateRawRow() is counting values in numEntriesInTarget, but all the values that numEntriesInTarget is compared to count tuples.

This is similar to PIG-3649 but has less of a performance impact, as it generally does not cause map-side aggregation to be disabled entirely (since aggregateRawRow() is unlikely to be called before checkSizeReduction(); however, it can cause second-level aggregation to be run too often.