[jira] [Resolved] (SPARK-30413) Avoid unnecessary WrappedArray roundtrip in GenericArrayData constructor

["Dongjoon Hyun (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SPARK-30413?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dongjoon Hyun resolved SPARK-30413.
-----------------------------------
    Fix Version/s: 3.0.0
       Resolution: Fixed

Issue resolved by pull request 27088
[https://github.com/apache/spark/pull/27088]

> Avoid unnecessary WrappedArray roundtrip in GenericArrayData constructor
> ------------------------------------------------------------------------
>
>                 Key: SPARK-30413
>                 URL: https://issues.apache.org/jira/browse/SPARK-30413
>             Project: Spark
>          Issue Type: Improvement
>          Components: SQL
>    Affects Versions: 3.0.0
>            Reporter: Josh Rosen
>            Assignee: Josh Rosen
>            Priority: Major
>             Fix For: 3.0.0
>
>
> GenericArrayData has a constructor which accepts a {{seqOrArray: Any}} parameter. This constructor was originally added for use in situations where we don't know the actual type at compile-time (e.g. when converting UDF outputs). It's also called (perhaps unintentionally) in code paths where we could plausibly and statically know that the type is {{Array[Any]}} (in which case we could simply call the primary constructor).
> In the current version of this code there's an unnecessary performance penalty for going through this path when {{seqOrArray}} is an {{Array[Any]}}: we end up converting the array into a WrappedArray, then call a method to unwrap it back into an array: this results in a bunch of unnecessary method calls. See https://scastie.scala-lang.org/7jOHydbNTaGSU677FWA8nA for an example of situations where this can crop up.
> Via a small modification to this constructor's implementation, I think we can effectively remove this penalty.



--
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