[SPARK-28724] Throw error message when cast out range decimal to long

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Incomplete
• Affects Version/s: 2.3.0
• Component/s: SQL
• Labels: bulk-closed
• Assignee: Unassigned
• Reporter: ShuMing Li
• Created: 14/Aug/19 03:36
• Updated: 25/May/21 01:55
• Resolved: 25/May/21 01:41

Description

Maybe this is a bug in `Scala` when convert `BigDecimal` to `Long`, however Spark should keep the result correct when query the sqls below:

spark-sql> select cast(20190801002382000052000000017638 as int);
-1493203738

spark-sql> select cast(20190801002382000052000000017638 as bigint);
4671677505944388838

After this patch, the result will throw AnalysisException :

spark-sql> select cast(20190801002382000052000000017638 as bigint);
Error in query: Decimal 20190801002382000052000000017638 does not fit in range [-9223372036854775808, 9223372036854775807] for type Long;

 

For `toFloat/toDouble`, the result is reasonable:

spark-sql> select cast(201908010023820000520000000176380000000000000000.0 as double);
Error in query: DecimalType can only support precision up to 38 == SQL == select cast(201908010023820000520000000176380000000000000000.0 as double)

spark-sql> select cast(201908010023820000520000000176380000000000000000.0 as float); Error in query: DecimalType can only support precision up to 38 == SQL == select cast(201908010023820000520000000176380000000000000000.0 as float)

 

 

Issue Links

• links to: GitHub Pull Request #25446