[
https://issues.apache.org/jira/browse/HADOOP-3498?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tom White updated HADOOP-3498:
------------------------------
Attachment: hadoop-3498-v2.patch
Hairong, Thanks for the feedback.
I wanted to avoid expanding all alternations so that we didn't get an explosion
of paths for things like:
{a,b,c,d,e}{f,g,h,i,j}{k,l,m,n,o}
This would continue to be processed as a single path and matched using regular
expressions.
The patch does scan filePattern multiple times, so I've changed it so that it
doesn't scan previously-expanded parts of the path. This change dramatically
reduces the number of character scans.
> File globbing alternation should be able to span path components
> ----------------------------------------------------------------
>
> Key: HADOOP-3498
> URL: https://issues.apache.org/jira/browse/HADOOP-3498
> Project: Hadoop Core
> Issue Type: Improvement
> Components: fs
> Reporter: Tom White
> Assignee: Tom White
> Attachments: hadoop-3498-v2.patch, hadoop-3498.patch
>
>
> For example, {/a/b,/c/d} should expand to /a/b and /c/d. This change would
> also permit a consistent syntax for specifying multiple input paths to
> MapReduce, streaming and Pig by specification of a single glob path with
> alternation {/a/b,/c/d}, rather than a collection of comma separated glob
> paths /a/b,/c/d.
> This change would also make globbing more consistent with bash, which
> supports this feature.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
