You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@spark.apache.org by "Sean R. Owen (Jira)" <ji...@apache.org> on 2020/08/18 15:49:00 UTC

[jira] [Commented] (SPARK-26575) revisit the equality of NaN

    [ https://issues.apache.org/jira/browse/SPARK-26575?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17179713#comment-17179713 ] 

Sean R. Owen commented on SPARK-26575:
--------------------------------------

BTW as an aside to this old comment - it depends. If one needs IEEE semantics, yes. But if one needs a consistent ordering on floats, no. (Indeed Scala itself now offers two different modes for this reason). Following NaN != NaN can change lots of behaviors in sorting, grouping. NaNs could start to appear in different places in sorts depending on the algorithm.

> revisit the equality of NaN
> ---------------------------
>
>                 Key: SPARK-26575
>                 URL: https://issues.apache.org/jira/browse/SPARK-26575
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 3.0.0
>            Reporter: Wenchen Fan
>            Priority: Major
>
> In Spark SQL, we treat all the NaN values as same. This is against IEEE floating point
> standard. We should revisit it.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

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