[
https://issues.apache.org/jira/browse/HADOOP-4565?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12662999#action_12662999
]
Joydeep Sen Sarma commented on HADOOP-4565:
-------------------------------------------
Comments:
- min/maxSplitSize - i found the initialization confusing. is it the intent
that values supplied by setXXXSplitSize take precedence over config values? (I
would find that logical - but the implementation seems different)
- OneFileInfo - it seems to me that this class has no use except that it's
constructor has a side-effect of populating some external structures. As a
style issue - this seems unnecessary - we could just have a simple helper
function instead.
- for(boolean doIterate ..): i found this variable pretty confusing. could we
just replace doIterator with blockToNodes.size()>0 ?
- one of the comments from the last review was that CombineFileSplit would be
nice to treat as an extension of MultiFileSplit. I would still make that
argument - there should be a single base class/interface for these multisplit
classes - because then the CombineFileRecordReader can be made to work with all
of them.
Looking at the apis - it actually seems that MultiFileSplit can perhaps be
treated as a special implementation (extension) of CombineFileSplit (with
get*offset functions always returning 0).
Any thoughts on this?
Not that familiar with miniMR stuff - would be good if someone else can review
the test code as well ..
> MultiFileInputSplit can use data locality information to create splits
> ----------------------------------------------------------------------
>
> Key: HADOOP-4565
> URL: https://issues.apache.org/jira/browse/HADOOP-4565
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: CombineMultiFile.patch, CombineMultiFile2.patch,
> CombineMultiFile3.patch, CombineMultiFile4.patch, CombineMultiFile5.patch
>
>
> The MultiFileInputFormat takes a set of paths and creates splits based on
> file sizes. Each splits contains a few files an each split are roughly equal
> in size. It would be efficient if we can extend this InputFormat to create
> splits such each all the blocks in one split and either node-local or
> rack-local.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
