[
https://issues.apache.org/jira/browse/SPARK-5916?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14332001#comment-14332001
]
Patrick Wendell commented on SPARK-5916:
----------------------------------------
The naming conflict is unfortunate. However, scripts like this are a public API
in Spark, so I don't think we can remove this API randomly in a minor release
given our versioning policies. For instance, people write scripts that invoke
"beeline" to submit queries, etc, and it would break those.
One thing we could potentially do is rename this to spark-beeline (or even to
something else) and update references accordingly. And then leave ./bin/beeline
as a backwards-compatible alias. And if individual distributors want to order
their path so that the "beeline" hits the Hive version first, they can do that.
However, that approach is also confusing because we'd have "beeline" and
"spark-beeline" both there.
Ping also [~marmbrus] and [~matei] regarding how we approach changes to scripts
like this. In the past I don't think we've ever changed one of these in a minor
release.
> $SPARK_HOME/bin/beeline conflicts with $HIVE_HOME/bin/beeline
> -------------------------------------------------------------
>
> Key: SPARK-5916
> URL: https://issues.apache.org/jira/browse/SPARK-5916
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Reporter: Carl Steinbach
> Priority: Minor
>
> Hive provides a JDBC CLI named "beeline". Spark currently depends on beeline,
> but provides its own "beeline" wrapper script for launching it. This results
> in a conflict when both $HIVE_HOME/bin and $SPARK_HOME/bin appear on a user's
> PATH.
> In order to eliminate the potential for conflict I propose changing the name
> of Spark's beeline wrapper script to sparkline.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
