[SPARK-38929] Improve error messages for cast failures in ANSI

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.3.0
• Fix Version/s: 3.3.0, 3.4.0
• Component/s: SQL
• Epic Link: Error Message Improvements in Spark 3.3
• Assignee: Xinyi Yu
• Reporter: Xinyi Yu
• Created: 18/Apr/22 04:35
• Updated: 20/Apr/22 10:09
• Resolved: 19/Apr/22 19:25

Description

Improve several error messages for cast failures in ANSI.

Cast to numeric types

java.lang.NumberFormatException: invalid input syntax for type numeric: 1.0. To return NULL instead, use 'try_cast'. ...

This is confusing as 1.0 is numeric to an average user. Need to mention the specific target type (integer in this case) and put 1.0 in single quotes. If we can mention this is a cast from string to an integer that’s even better.

Proposed change

Invalid `int` literal: '1.0'. To return NULL instead, use 'try_cast'.

Cast to date types

java.time.DateTimeException: Cannot cast 2021-09- 2 to DateType.

Can align with the above change.

Proposed change

Invalid `date` literal: '2021-09- 2'. To return NULL instead, use 'try_cast'.

Issue Links

• links to: [Github] Pull Request #36241 (anchovYu)
• links to: [Github] Pull Request #36275 (anchovYu)

Sub-Tasks

• Use SparkDateTimeException when casting to datetime types failed: Resolved