[GitHub] [arrow-datafusion] alamb commented on issue #120: Decide on CAST behaviour for invalid inputs

[GitBox <gi...@apache.org>]

alamb commented on issue #120:
URL: https://github.com/apache/arrow-datafusion/issues/120#issuecomment-826831994


   Comment from Mike Seddon(MikeSeddonAU) @ 2020-12-02T22:44:31.207+0000:
   <pre>I strongly feel that DataFusion should adopt the ANSI-style strict typing rather than silent error suppression and return of NULL values as intuitively (due to years of using DBMS) users expect that if an error was not thrown then all operations were completed successfully.
   
   The default Spark behavior was inherited from Hive SQL which I assume was originally built to support a business where absolute precision was not necessarily important. As part of the Spark 3.0 release a huge amount of effort was put in to comply with ANSI standard SQL ([https://spark.apache.org/docs/3.0.0/sql-ref-ansi-compliance.html|https://spark.apache.org/docs/3.0.0/sql-ref-ansi-compliance.html)]) which is obviously a lot harder to retrofit than start with.
   
   This also goes wider than type conversions as adopting ANSI SQL standards (including functionality like [https://github.com/apache/arrow/pull/8688] which I think requires a CASE statement in ANSI SQL) should maybe be agreed by the PMC to give a framework for assessing PRs against. Perhaps this ticket should be changed to a discussion of which dialect of SQL DataFusion aims to support.</pre>


-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org