[
https://issues.apache.org/jira/browse/HIVE-24348?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17266677#comment-17266677
]
David Mollitor commented on HIVE-24348:
---------------------------------------
So, one of the things that you will have to consider with this is the utility
of the {{dfs}} command that is built into beeline.
https://cwiki.apache.org/confluence/display/Hive/HiveServer2+Clients#HiveServer2Clients-BeelineHiveCommands
If this feature is going to continue then beeline will need to be able to
include the Hadoop JARs. Beeline could explicitly include the Hadoop libraries
as a runtime dependency instead of dynamically loading it from {{HADOOP_HOME}},
and that would make things better and more standalone. Maybe that is a good
improvement, but it would add a LOT of transitive dependencies to beeline to
support HDFS, Azure, S3, etc.
I always lean in the favor of KISS. My gut says that this {{dfs}} command can
be removed and make beeline just a simple JDBC tool, but I'm not sure.
> Beeline: Isolating dependencies and execution with java
> -------------------------------------------------------
>
> Key: HIVE-24348
> URL: https://issues.apache.org/jira/browse/HIVE-24348
> Project: Hive
> Issue Type: Improvement
> Components: Beeline
> Affects Versions: 3.1.0
> Reporter: Naveen Gangam
> Assignee: Naveen Gangam
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h
> Remaining Estimate: 0h
>
> Currently, beeline code, binaries and executables are somewhat tightly
> coupled with the hive product. To be able to execute beeline from a node with
> just JRE installed and some jars in classpath is impossible.
> * beeline.sh/hive scripts rely on HADOOP_HOME to be set which are designed to
> use "hadoop" executable to run beeline.
> * Ideally, just the hive-beeline.jar and hive-jdbc-standalone jars should be
> enough but sadly they arent. The latter jar adds more problems than it solves
> because all the classfiles are shaded some dependencies cannot be resolved.
> * Beeline has many other dependencies like hive-exec, hive-common.
> hadoop-common, supercsv, jline, commons-cli, commons-io, commons-logging etc.
> While it may not be possible to eliminate some of these, we should atleast
> have a self-contains jar that contains all these to be able to make it work.
> * the underlying script used to run beeline should use JAVA as an alternate
> means to execute if HADOOP_HOME is not set
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
