[PARQUET-362] Parquet buffered writer is too sensitive regarding unions and unknown fields

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.7.0, 1.8.0, 1.8.1
• Fix Version/s: 2.0.0
• Component/s: parquet-mr
• Assignee: Unassigned
• Reporter: Laurent Goujon
• Created: 20/Aug/15 20:43
• Updated: 20/Aug/15 20:53
• Resolved: 20/Aug/15 20:53

Description

Parquet does prevent records with unknown union fields to be written as it would create a TProtocol violation. But it also prevents records with unions having one their field itself having an unknown field (which is acceptable if it is a struct).

The recursive check should be removed, and only when a union field is unknown should the record be deemed invalid.