[
https://issues.apache.org/jira/browse/HADOOP-2151?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12540306
]
Raghu Angadi commented on HADOOP-2151:
--------------------------------------
This is because globPaths decides invoke {{listPaths}} on directories only if
there is a glob specified for that 'component' of path.
I think we should change it to use {{listPaths}} at every component if there is
a glob specified anywhere to start with. Otherwise just return the path.
Minimizing {{listPaths}} is not required when a glob is specified since the
user is already willing to pay the cost.
> FileSyste.globPaths does not validate the return list of Paths
> --------------------------------------------------------------
>
> Key: HADOOP-2151
> URL: https://issues.apache.org/jira/browse/HADOOP-2151
> Project: Hadoop
> Issue Type: Bug
> Components: dfs
> Affects Versions: 0.14.3, 0.15.0
> Reporter: lohit vijayarenu
> Fix For: 0.16.0
>
>
> FileSystem.globPaths does not validate the return list of Paths.
> Here is an example.
> Consider a directory structure like
> /user/foo/DIR1/FILE1
> /user/foo/DIR2
> now if we pass an input path like "/user/foo/*/FILE1" to
> FileSystem.globPaths()
> It returns 2 entries as shown below
> /user/foo/DIR1/FILE1
> /user/foo/DIR2/FILE1
> Should globPaths validate this and return only valid Paths? This behavior was
> caught in FileSystem.validateInput() where an IOException is thrown while
> processing such a directory structure.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
