j1wonpark commented on PR #56447:
URL: https://github.com/apache/spark/pull/56447#issuecomment-4686873917

   Thanks @pan3793! Surveyed Trino, pgjdbc, and Databricks (databricks-jdbc 
OSS) in addition to Hive. All current drivers map ARRAY to `Types.ARRAY`, and 
MAP to `Types.JAVA_OBJECT` since JDBC has no `Types.MAP` (Hive, Trino), so this 
PR follows that.
   
   The only divergence is STRUCT: Trino maps ROW to `JAVA_OBJECT`, arguably 
because `Types.STRUCT` was designed for named UDTs - which Spark's anonymous 
STRUCT isn't either. I still chose `Types.STRUCT` because with `JAVA_OBJECT`, 
MAP and STRUCT become indistinguishable by type code (in Trino, both report 
2000 and clients must parse the type name string to tell them apart), and 2002 
matches what Hive JDBC reports for struct columns, so type-code-based dispatch 
in existing clients keeps working.
   
   Also added nested type cases (`ARRAY<STRUCT>`, `MAP<STRING, ARRAY<INT>>`, 
`STRUCT` containing `MAP`) to the UT.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to