[HIVE-8120] Umbrella JIRA tracking Parquet improvements

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Assignee: Brock Noland
• Reporter: Brock Noland
• Created: 16/Sep/14 04:53
• Updated: 18/Aug/15 23:28

Issue Links

• blocks: HIVE-8128 Improve Parquet Vectorization
• incorporates: HIVE-10666 Improvement for Parquet Predicate Push down on Hive
• is blocked by: HIVE-7685 Parquet memory manager
• is blocked by: HIVE-7858 Parquet compression should be configurable via table property
• is blocked by: HIVE-9442 Make sure all data types work for PARQUET
• is blocked by: HIVE-6384 Implement all Hive data types in Parquet
• is related to: HIVE-6914 parquet-hive cannot write nested map (map value is map)
• is related to: HIVE-7800 Parquet Column Index Access Schema Size Checking
• is related to: HIVE-8838 Support Parquet through HCatalog
• relates to: HIVE-5998 Add vectorized reader for Parquet files
• relates to: HIVE-4329 HCatalog should use getHiveRecordWriter rather than getRecordWriter
• relates to: HIVE-11598 Document Configuration for Parquet Files

Sub-Tasks

• Improve Parquet Vectorization: Patch Available
• Implement the bloom filter for the ParquetSerde: Open
• Warn user when parquet mm kicks in: Open
• Improve the speed of select count(*) statement for a parquet table with big input(~1GB): Open
• Reduce parquet memory usage by bypassing java primitive objects on ETypeConverter: Patch Available
• Avoid reading file footers in ParquetRecordReaderWrapper: Open
• Allocate only parquet selected columns in HiveStructConverter class: Patch Available
• Turn on Parquet vectorization in parquet branch: Patch Available
• Use new ParquetInputSplit constructor API: Patch Available
• A bad performance regression issue with Parquet happens if Hive does not select any columns: Reopened
• Map instances with null keys are not properly handled for Parquet tables: Open