[SPARK-38710][SQL] Use SparkArithmeticException for arithmetic overflow runtime errors

[gengliangwang]

What changes were proposed in this pull request?

On arithmetic overflow runtime errors, Spark should throw SparkArithmeticException instead of java.lang.ArithmeticException

Why are the changes needed?

Use a better error exception type.

Does this PR introduce any user-facing change?

Yes, trivial change on the exception type: on arithmetic overflow runtime errors, Spark will throw SparkArithmeticException instead of java.lang.ArithmeticException

How was this patch tested?

UT

[gengliangwang]

SQLSTATE	Class	Condition	Subclass	Subcondition
22003	22	data exception	003	numeric value out of range

[gengliangwang]

Merging to master/3.3

[tgravescs]

question, I haven't kept up with all the error changes, but this only changes 1 place and there are other places in the QueryExecutionErrors that still use the ArtithmeticException, like right above the one you changed:

 def arithmeticOverflowError(e: ArithmeticException): ArithmeticException = {
    new ArithmeticException(s"${e.getMessage}. If necessary set ${SQLConf.ANSI_ENABLED.key} " +
      s"to false to bypass this error.")
  }

is there a reason we are now inconsistent?

[gengliangwang]

@tgravescs Good point.
I found this during working on the project https://issues.apache.org/jira/browse/SPARK-38615. Currently my major focus is still on SPARK-38615.
The method you mentioned seems not to be used anywhere. There are also a few places using ArithmeticException as well. I created https://issues.apache.org/jira/browse/SPARK-38842 as a follow-up.

[tgravescs]

thanks for the clarification and followup @gengliangwang