[
https://issues.apache.org/jira/browse/SPARK-9550?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Reynold Xin updated SPARK-9550:
-------------------------------
Priority: Major (was: Blocker)
> Configuration renaming, defaults changes, and deprecation for 1.5.0 (master
> ticket)
> -----------------------------------------------------------------------------------
>
> Key: SPARK-9550
> URL: https://issues.apache.org/jira/browse/SPARK-9550
> Project: Spark
> Issue Type: Task
> Components: Spark Core, SQL
> Affects Versions: 1.5.0
> Reporter: Josh Rosen
>
> This ticket tracks configurations which need to be renamed, deprecated, or
> have their defaults changed for Spark 1.5.0.
> Note that subtasks / comments here do not necessarily need to reflect changes
> that must be performed. Rather, tasks should be added here to make sure that
> the relevant configurations are at least checked before we cut releases.
> This ticket will also help us to track configuration changes which must make
> it into the release notes.
> *Configuration renaming*
> - Consider renaming {{spark.shuffle.memoryFraction}} to
> {{spark.execution.memoryFraction}}
> ([discussion|https://github.com/apache/spark/pull/7770#discussion-diff-36019144]).
> - Rename all public-facing uses of {{unsafe}} to something less scary, such
> as {{tungsten}}
> *Defaults changes*
> - Codegen is now enabled by default.
> - Tungsten is now enabled by default.
> - Parquet schema merging is now disabled by default.
> - In-memory relation partition pruning should be enabled by default
> (SPARK-9554).
> *Deprecation*
> - Local execution has been removed.
> *Behavior Changes*
> - Canonical name of SQL/DataFrame functions are now lower case (e.g. sum vs
> SUM)
> - DirectOutputCommitter is not safe to use with speculation
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
