[
https://issues.apache.org/jira/browse/HDFS-1475?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13048660#comment-13048660
]
Allen Wittenauer commented on HDFS-1475:
----------------------------------------
My read of this JIRA in combination with HADOOP-7378 seemed to hint at a change
in -ls's default behavior. Thus why I said "suspected" above. If that isn't
the case, then we're good to go. If it does change it, we're still good to
go*... we just need to flag one of these as an incompatible change and throw a
release note in there so folks aren't surprised.
* - despite what some might think, I've actually wanted more things to break
inbetween releases than what most of the other devs have wanted. I'd rather
break things now, even in really horrible ways, before we hit 1.0 than after.
This includes -ls's default behavior, which I think is... less than useful.
> Want a -d flag in hadoop dfs -ls : Do not expand directories
> ------------------------------------------------------------
>
> Key: HDFS-1475
> URL: https://issues.apache.org/jira/browse/HDFS-1475
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs client
> Affects Versions: 0.23.0
> Environment: any
> Reporter: Greg Connor
> Assignee: Daryn Sharp
> Priority: Minor
> Attachments: HDFS-1475.patch
>
>
> I would really love it if dfs -ls had a -d flag, like unix ls -d, which would
> list the directories matching the name or pattern but *not* their contents.
> Current behavior is to expand every matching dir and list its contents, which
> is awkward if I just want to see the matching dirs themselves (and their
> permissions). Worse, if a directory exists but is empty, -ls simply returns
> no output at all, which is unhelpful.
> So far we have used some ugly workarounds to this in various scripts, such as
> -ls /path/to |grep dir # wasteful, and problematic if "dir" is a
> substring of the path
> -stat /path/to/dir "Exists" # stat has no way to get back the full path,
> sadly
> -count /path/to/dir # works but is probably overkill.
> Really there is no reliable replacement for ls -d -- the above hacks will
> work but only for certain isolated contexts. (I'm not a java programmer, or
> else I would probably submit a patch for this, or make my own jar file to do
> this since I need it a lot.)
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
