[PARQUET-2381] Deprecate methods relying on Hadoop classes when alternatives using more generic Parquet interfaces are available

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: parquet-mr
• Assignee: Unassigned
• Reporter: Atour Mousavi Gourabi
• Created: 15/Nov/23 13:55
• Updated: 23/Jun/24 03:33

Description

Deprecate methods relying on Hadoop classes when alternatives using Parquet interfaces are available. This would allow for fully decoupling from Hadoop for simple read/write (and potentially rewrite) operations in the future, where we are now only on track to decouple from hadoop-client-runtime in the near future.

We could decide to still keep Hadoop methods available through convenience classes in their own Hadooped module, but for core read/write logic to be available without forcing users into the Hadoop dependency, we will need to fully strip the core readers and writers of these references.

Issue Links

• is blocked by: PARQUET-2402 Create specialized HadoopParquetWriter
• is blocked by: PARQUET-2403 Create specialized HadoopParquetReader