[PARQUET-326] Binary statistics are invalid if buffers are reused

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Duplicate
• Affects Version/s: 1.7.0
• Component/s: parquet-mr
• Assignee: Unassigned
• Reporter: Thomas White
• Created: 07/Jul/15 15:24
• Updated: 07/Jul/15 16:02
• Resolved: 07/Jul/15 16:02

Description

BinaryStatistics does not copy the min or max binary objects, so if the buffers are shared (as they usually are in MapReduce), then the statistics are incorrect.

Issue Links

• duplicates: PARQUET-251 Binary column statistics error when reuse byte[] among rows
• is related to: PARQUET-62 DictionaryValuesWriter dictionaries are corrupted by user changes.