[
https://issues.apache.org/jira/browse/MAHOUT-1615?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14133581#comment-14133581
]
ASF GitHub Bot commented on MAHOUT-1615:
----------------------------------------
Github user dlyubimov commented on the pull request:
https://github.com/apache/mahout/pull/52#issuecomment-55552838
yeah i dont think cloning is needed, since we unwrapping the actual object
(string, int etc). So by the time it is cached, it is a collection of ints or
stings but not writables. So what Matei is saying there, should not already be
applicable.
More likely, i was thinking, spark is creating a strict collection when it
fuses two maps, so more likely fix would be simply removing extra map,
something along the lines
val rdd = sc.sequenceFile(path, classOf[Writable],
classOf[VectorWritable], minPartitions = parMin)
...
val drmRdd = rdd.map { t => key2valFunc(t._1) -> t._2.get}
...
(so there's no forced map fusion). But it doesn't seem to work in my tests
either. Weird.
Ok i think i want to stop spending my neurons on this for now and to have
a look at it tomorrow again.
But i don't think we should clone anything. we just need to transform away
from writable before attempting any caches.
> 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)
