You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Harsh J (JIRA)" <ji...@apache.org> on 2012/09/22 10:11:07 UTC

[jira] [Commented] (HADOOP-7779) bin/hadoop classpath CLI does not display the complete classpath

    [ https://issues.apache.org/jira/browse/HADOOP-7779?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13461059#comment-13461059 ] 

Harsh J commented on HADOOP-7779:
---------------------------------

I tend to agree with Ravi. I don't find it useful to see a very huge list of jars thrown on my screen when I demand a {{classpath}}. I just want a classpath I can use via {{-cp/-classpath}}.

bq. For cases where users want to include a jar in classpath by checking if one already exists or not, one has to manually traverse each and every directory just to locate the jar.

But there's maven for that purpose today. We should really be moving people down the proper project-based paths?

bq. IMO, the idea of exposing "hadoop classpath" was to let users learn about all the jars included in the classpath. Masking that by just printing wildcards would not serve the purpose.

That may have been so, but all I've seen as the use of this command from both dev and op side (specifically of those who've been running 2.x based version distributions) has been to rather check for the presence of certain projects or for use with a quick {{java/javac}} command. For these purposes, knowing the jar is not very important.

bq. Also, this was the behavior in hadoop-0.20.x. I would not want to change this behavior unless we agree to introduce a regression.

Functionally we're still printing a usable classpath (in whatever form - wildcards or no wildcards) for use with other programs, that much we haven't broken I think?

Also to note is that now we have tons of jars on the classpath, its not sane to print or read through them all anymore.

Let me know how you still feel about this change - I feel that we may instead devote more time to documenting how to use the apache hadoop maven-based projects, than anything else such as teaching people to use jars (which is so old style today!).
                
> bin/hadoop classpath CLI does not display the complete classpath
> ----------------------------------------------------------------
>
>                 Key: HADOOP-7779
>                 URL: https://issues.apache.org/jira/browse/HADOOP-7779
>             Project: Hadoop Common
>          Issue Type: Bug
>    Affects Versions: 0.23.0
>            Reporter: Ramya Sunil
>
> bin/hadoop classpath does not display the complete classpath. The classpath looks like:
> {noformat}
> $HADOOP_CONF_DIR:$HADOOP_COMMON_HOME/libexec/../share/hadoop/common/lib/*:$HADOOP_COMMON_HOME/libexec/../share/hadoop/common/*:/$HADOOP_HDFS_HOME/share/hadoop/hdfs:$HADOOP_HDFS_HOME/share/hadoop/hdfs/lib/*:$HADOOP_HDFS_HOME/share/hadoop/hdfs/*
> {noformat}
> The jar names have to be expanded instead of "*" to display the complete classpath.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira