[
https://issues.apache.org/jira/browse/SPARK-22036?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16168989#comment-16168989
]
Olivier Blanvillain edited comment on SPARK-22036 at 9/16/17 5:22 PM:
----------------------------------------------------------------------
It's surprising because in this case the resulting value seems fits within the
range of representable values:
{code:java}
scala> val result = BigDecimal(-0.1267333984375) * BigDecimal(-1000.1)
result: scala.math.BigDecimal = 126.74607177734375
scala> sqlContenxt.createDataset(List(result)).head == result
res10: Boolean = true
{code}
Also Spark will silently loses BigDecimal precision in other circumstances:
{code:java}
scala> val tooPrecise = BigDecimal("126.74607177734375111111111")
tooPrecise: scala.math.BigDecimal = 126.74607177734375111111111
scala> val ds = sqlContenxt.createDataset(List(tooPrecise))
ds: org.apache.spark.sql.Dataset[scala.math.BigDecimal] = [value:
decimal(38,18)]
scala> ds.head
res14: scala.math.BigDecimal = 126.746071777343751111
scala> ds.select(ds("value") * BigDecimal(1)).head
res15: org.apache.spark.sql.Row = [126.746071777343751111]
{code}
> I am not sure of what should be done in this case
Given that Sparks' BigDecimal have bounded precision I would consider following
that is done for other numeric representations and return the closest
representable value in case of overflow.
was (Author: olivierblanvillain):
It's surprising because in this case the resulting value fits within the range
of representable values:
{code:java}
scala> val result = BigDecimal(-0.1267333984375) * BigDecimal(-1000.1)
result: scala.math.BigDecimal = 126.74607177734375
scala> sqlContenxt.createDataset(List(result)).head == result
res10: Boolean = true
{code}
Also Spark will silently loses BigDecimal precision in other circumstances:
{code:java}
scala> val tooPrecise = BigDecimal("126.74607177734375111111111")
tooPrecise: scala.math.BigDecimal = 126.74607177734375111111111
scala> val ds = sqlContenxt.createDataset(List(tooPrecise))
ds: org.apache.spark.sql.Dataset[scala.math.BigDecimal] = [value:
decimal(38,18)]
scala> ds.head
res14: scala.math.BigDecimal = 126.746071777343751111
scala> ds.select(ds("value") * BigDecimal(1)).head
res15: org.apache.spark.sql.Row = [126.746071777343751111]
{code}
> I am not sure of what should be done in this case
Given that Sparks' BigDecimal have bounded precision I would consider following
that is done for other numeric representations and return the closest
representable value in case of overflow.
> BigDecimal multiplication sometimes returns null
> ------------------------------------------------
>
> Key: SPARK-22036
> URL: https://issues.apache.org/jira/browse/SPARK-22036
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 2.2.0
> Reporter: Olivier Blanvillain
>
> The multiplication of two BigDecimal numbers sometimes returns null. This
> issue we discovered while doing property based testing for the frameless
> project. Here is a minimal reproduction:
> {code:java}
> object Main extends App {
> import org.apache.spark.{SparkConf, SparkContext}
> import org.apache.spark.sql.SparkSession
> import spark.implicits._
> val conf = new
> SparkConf().setMaster("local[*]").setAppName("REPL").set("spark.ui.enabled",
> "false")
> val spark =
> SparkSession.builder().config(conf).appName("REPL").getOrCreate()
> implicit val sqlContext = spark.sqlContext
> case class X2(a: BigDecimal, b: BigDecimal)
> val ds = sqlContext.createDataset(List(X2(BigDecimal(-0.1267333984375),
> BigDecimal(-1000.1))))
> val result = ds.select(ds("a") * ds("b")).collect.head
> println(result) // [null]
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
