andygrove opened a new issue, #464:
URL: https://github.com/apache/datafusion-comet/issues/464

   ### Describe the bug
   
   Comet currently fails with an overflow when ANSI mode is disabled, but we 
should return the original value instead.
   
   ## Create test data
   
   ```
   val df = Seq(Int.MaxValue, Int.MinValue).toDF("a")
   df.write.parquet("/tmp/int.parquet")
   spark.read.parquet("/tmp/int.parquet").createTempView("t")
   ```
   
   ## Test with ansi mode disabled
   
   With ansi mode disabled we expect abs to return the original value if there 
would be an overflow calculating the abs value, but Comet fails with an 
overflow.
   
   ```
   scala> spark.conf.set("spark.comet.enabled", false)
   
   scala> spark.sql("select a, abs(a) from t").show
   +-----------+-----------+
   |          a|     abs(a)|
   +-----------+-----------+
   | 2147483647| 2147483647|
   |-2147483648|-2147483648|
   +-----------+-----------+
   
   scala> spark.conf.set("spark.comet.enabled", true)
   
   scala> spark.sql("select a, abs(a) from t").show
   24/05/23 13:26:17 WARN CometSparkSessionExtensions$CometExecRule: Comet 
cannot execute some parts of this plan natively because CollectLimit is not 
supported
   24/05/23 13:26:17 ERROR Executor: Exception in task 0.0 in stage 7.0 (TID 8)
   org.apache.comet.CometNativeException: Arrow error: Compute error: 
Int32Array overflow on abs(-2147483648)
   ```
   
   With ansi mode enabled, we do expect an exception, but we should try and 
match Spark's exception.
   
   ```
   org.apache.spark.SparkArithmeticException: [ARITHMETIC_OVERFLOW] integer 
overflow. If necessary set "spark.sql.ansi.enabled" to "false" to bypass this 
error.
   ```
   
   Comet currently has this error:
   
   ```
   org.apache.comet.CometNativeException: Arrow error: Compute error: 
Int32Array overflow on abs(-2147483648)
   ```
   
   ### Steps to reproduce
   
   _No response_
   
   ### Expected behavior
   
   _No response_
   
   ### Additional context
   
   _No response_


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to