[AUDIT] [SPARK-45786] Inaccurate Decimal multiplication and division results

[tgravescs]

Describe the bug
https://issues.apache.org/jira/browse/SPARK-45786 fixes inaccurate decimal multiplication and division in Spark. This looks like a dup of https://issues.apache.org/jira/browse/SPARK-40129.

I believe we explicitly copied the behavior of Spark for the multiplication side so we should look to see if we need to change it.

[tgravescs]

Note a repro case of this is:

pyspark
 
from pyspark.sql.functions import *
df = spark.createDataFrame([("-8533444864753048107770677711.1312637916", )], ["col_1"])
df = df.withColumn("col_2", lit(-12).cast("decimal(38,10)"))
df = df.withColumn("col_3", col("col_2").cast("decimal(38,10)") * col("col_1").cast("decimal(38,10)"))
df.show(truncate=False)
 
(This was run just on my CPU with Apache Spark) before the fix:
+----------------------------------------+--------------+-------------------------------------+
|col_1                                   |col_2         |col_3                                |
+----------------------------------------+--------------+-------------------------------------+
|-8533444864753048107770677711.1312637916|-12.0000000000|102401338377036577293248132533.575166|
+----------------------------------------+--------------+-------------------------------------+
 
The value after the fix is:
 got  102401338377036577293248132533.575165

[razajafri]

There is a follow-on bug that specifically deals with Decimal division on Spark versions 330db+ and 340+