[jira] [Commented] (SPARK-27336) Incorrect DataSet.summary() result

["Chakravarthi (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-27336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16806947#comment-16806947 ] 

Chakravarthi commented on SPARK-27336:
--------------------------------------

I'm checking this issue.

> Incorrect DataSet.summary() result
> ----------------------------------
>
>                 Key: SPARK-27336
>                 URL: https://issues.apache.org/jira/browse/SPARK-27336
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 3.0.0
>            Reporter: Gengliang Wang
>            Priority: Major
>         Attachments: test.csv
>
>
> There is a single data point in the minimum_nights column that is 1.0E8 out of 8k records, but .summary() says it is the 75% and the max.
> I compared this with approxQuantile, and approxQuantile for 75% gave the correct value of 30.0.
> To reproduce:
> {code:java}
> scala> val df = spark.read.format("csv").load("test.csv").withColumn("minimum_nights", '_c0.cast("Int"))
> df: org.apache.spark.sql.DataFrame = [_c0: string, minimum_nights: int]
> scala> df.select("minimum_nights").summary().show()
> +-------+------------------+
> |summary|    minimum_nights|
> +-------+------------------+
> |  count|              7072|
> |   mean| 14156.35407239819|
> | stddev|1189128.5444975856|
> |    min|                 1|
> |    25%|                 2|
> |    50%|                 4|
> |    75%|         100000000|
> |    max|         100000000|
> +-------+------------------+
> scala> df.stat.approxQuantile("minimum_nights", Array(0.75), 0.1)
> res1: Array[Double] = Array(30.0)
> scala> df.stat.approxQuantile("minimum_nights", Array(0.75), 0.001)
> res2: Array[Double] = Array(30.0)
> scala> df.stat.approxQuantile("minimum_nights", Array(0.75), 0.0001)
> res3: Array[Double] = Array(1.0E8)
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org