[
https://issues.apache.org/jira/browse/MAHOUT-1615?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14143927#comment-14143927
]
ASF GitHub Bot commented on MAHOUT-1615:
----------------------------------------
GitHub user andrewpalumbo reopened a pull request:
https://github.com/apache/mahout/pull/52
MAHOUT-1615: drmFromHDFS returning the same Key for all Key,Vec Pairs for
Text-Keyed SequenceFiles
SparkContext.sequenceFile(...) will yield the same key per partition for
Text-Keyed Sequence files if the key a new copy of the key is not created when
mapping to an RDD. This patch checks for Text Keys and creates a copy of each
Key if necessary.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/andrewpalumbo/mahout MAHOUT-1615
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/mahout/pull/52.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #52
----
commit 6adb01ee53ce591962b97a4ed474c111635f7c47
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-14T20:50:50Z
Create copy of Key for Text Keys
commit a81c34000e0152a7ad7349afdba8a9e854380653
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-14T21:11:45Z
use y: Writable instead of y: VectorWritable
commit 2d431a40bd7a62a117487451b3e255c2e56c7d1a
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-14T22:55:34Z
scala not java
commit 5541d500ef8cfb53c9e18da1c760ea8c39dd5409
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-17T19:57:35Z
Read SequenceFile Header to get Key/Value Classes
commit 79468e7a83e7af1f9ab10689b46a900bd463aa38
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-17T19:59:34Z
Use our new method to get the ClassTag
commit ffb40161173df145041cf60d9b2878af3cab911b
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-18T23:07:33Z
Very Clunky use of getKeyClass. This commit solves inital problems. Needs
to be gutted. Hadoop configuration needs to be set up correctly. Spark I/O
DFS tests fail because reader doesnt know weather to read locally or from HDFS
commit 56b5db1b2ee5cdef3700fb9199c75637d5e3b570
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-19T19:10:00Z
Use case matching and val2keyFunc to map rdd
commit 0edeffe30c64e78297335f06516d8cb3ff6b36a3
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-21T21:58:29Z
Cleanup/temporarily hardcode spark kryo buffer property
commit 0d64b0be427998e74cc6d57bf653573b235e0a31
Author: Dmitriy Lyubimov <[email protected]>
Date: 2014-09-22T17:13:15Z
Adding key class tag extraction from checkpointed class;
adding assert to `DRM DFS i/o (local)` test to fail if key class tag is
incorrectly loaded.
temporary commended h20 module since it doesn't compile
commit c63468e10b9ae7e7b0b50728b0d2b71883f894b8
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-22T17:40:20Z
Merge branch 'MAHOUT-1615-drmFromHdfs' of
http://github.com/dlyubimov/mahout into MAHOUT-1615
commit 36b40615619c12dcd7820e7628b3bbf86329bd76
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-22T21:42:29Z
Read a sequence FILE not a directory- like the error says
commit da014759404db1398f8e334e2801b5159e816d56
Author: Andrew Palumbo <[email protected]>
Date: 2014-09-22T21:50:46Z
cleanup
----
> SparkEngine drmFromHDFS returning the same Key for all Key,Vec Pairs for
> Text-Keyed SequenceFiles
> -------------------------------------------------------------------------------------------------
>
> Key: MAHOUT-1615
> URL: https://issues.apache.org/jira/browse/MAHOUT-1615
> Project: Mahout
> Issue Type: Bug
> Reporter: Andrew Palumbo
> Fix For: 1.0
>
>
> When reading in seq2sparse output from HDFS in the spark-shell of form
> <Text,VectorWriteable> SparkEngine's drmFromHDFS method is creating rdds
> with the same Key for all Pairs:
> {code}
> mahout> val drmTFIDF= drmFromHDFS( path =
> "/tmp/mahout-work-andy/20news-test-vectors/part-r-00000")
> {code}
> Has keys:
> {...}
> key: /talk.religion.misc/84570
> key: /talk.religion.misc/84570
> key: /talk.religion.misc/84570
> {...}
> for the entire set. This is the last Key in the set.
> The problem can be traced to the first line of drmFromHDFS(...) in
> SparkEngine.scala:
> {code}
> val rdd = sc.sequenceFile(path, classOf[Writable], classOf[VectorWritable],
> minPartitions = parMin)
> // Get rid of VectorWritable
> .map(t => (t._1, t._2.get()))
> {code}
> which gives the same key for all t._1.
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
