[
https://issues.apache.org/jira/browse/SPARK-16521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15374612#comment-15374612
]
Saisai Shao commented on SPARK-16521:
-------------------------------------
I see, sorry about the duplication.
> Add support of parameterized configuration for SparkConf
> --------------------------------------------------------
>
> Key: SPARK-16521
> URL: https://issues.apache.org/jira/browse/SPARK-16521
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Reporter: Saisai Shao
> Priority: Minor
>
> Current SparkConf is a key value pair mechanism, in which value is a literal
> string cannot be changed. In most of the use cases this key value pair system
> is enough to express the meanings, while in some cases it would be more
> convenient to make value as a parameterized variable which can be replaceable
> by other configurations.
> One case is {{spark.sql.warehouse.dir}}, here the default value is is
> "file:${system:user.dir}/spark-warehouse" in which {{user.dir}} is replaced
> with system property in the runtime.
> Also several configuration like:
> {code}
> spark.dynamicAllocation.minExecutors 1
> spark.dynamicAllocation.initialExecutors 1
> {code}
> can also be configured as:
> {code}
> spark.dynamicAllocation.minExecutors 1
> spark.dynamicAllocation.initialExecutors
> ${spark.dynamicAllocation.minExecutors}
> {code}
> So here propose to add parameterized configuration support in SparkConf, this
> will not change the original semantics of SparkConf, just add a more option
> to do the configuration.
> This feature is quite useful in our environment, since we have some
> configurations that are version dependent, it is error prone and tedious to
> change it when environment is changed.
> Please suggest and comment, thanks a lot.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
