[CALCITE-5862] Incorrect semantics of ARRAY function (Spark library) when elements have Numeric and Character types

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.34.0
• Fix Version/s: 1.36.0
• Component/s: core
• Labels: pull-request-available
• Assignee: Ran Tao
• Reporter: Ran Tao
• Created: 19/Jul/23 07:44
• Updated: 21/Nov/23 10:07
• Resolved: 12/Sep/23 15:52

Description

when run select array(1, 2, 'Hi')

spark-sql (default)> select array(1, 2, 'Hi');
["1","2","Hi"] 

and calcite will cause java.lang.NullPointerException: inferred array element type.

Spark supports both character and numeric types in the array, and the return type is character.
In fact, calcite also allows both character and numeric types to exist in the operand type checker, but there is no corresponding processing logic in the return type inference, so an error is reported.

We should fix this bug.

Issue Links

• links to: GitHub Pull Request #3324