You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Apache Spark (JIRA)" <ji...@apache.org> on 2019/02/19 22:38:00 UTC
[jira] [Commented] (SPARK-26935) Skip DataFrameReader's CSV first
line scan when not used
[ https://issues.apache.org/jira/browse/SPARK-26935?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16772394#comment-16772394 ]
Apache Spark commented on SPARK-26935:
--------------------------------------
User 'Mister-Meeseeks' has created a pull request for this issue:
https://github.com/apache/spark/pull/23830
> Skip DataFrameReader's CSV first line scan when not used
> --------------------------------------------------------
>
> Key: SPARK-26935
> URL: https://issues.apache.org/jira/browse/SPARK-26935
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 2.4.0
> Environment: Spark Version 2.4.0
> Linux kernel 4.9.0-8-amd64 on Debian 9
> Docker container on Google Kubernetes Engine VM (n1-standard-8)
>
> Reporter: Douglas Colkitt
> Priority: Minor
> Labels: easyfix, performance, pull-request-available
> Original Estimate: 1h
> Remaining Estimate: 1h
>
> DataFrameReader always collects the first line from the first partition on every csv() call. This is used for schema inference and header filtering. But if schema is pre-specified by the user and header is false, there's no point.
> For certain cases CSV dataset generation is expensive and slow, and this clogs up the entire executor with 1 job to grab 1 line. If it's not needed, skipping the first line step is a free lunch for runtime and efficiency.
--
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