Github user mengxr commented on a diff in the pull request:
https://github.com/apache/spark/pull/10152#discussion_r52835896
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala ---
@@ -289,24 +301,20 @@ class Word2Vec extends Serializable with Logging {
val expTable = sc.broadcast(createExpTable())
val bcVocab = sc.broadcast(vocab)
val bcVocabHash = sc.broadcast(vocabHash)
-
- val sentences: RDD[Array[Int]] = words.mapPartitions { iter =>
- new Iterator[Array[Int]] {
- def hasNext: Boolean = iter.hasNext
-
- def next(): Array[Int] = {
- val sentence = ArrayBuilder.make[Int]
- var sentenceLength = 0
- while (iter.hasNext && sentenceLength < MAX_SENTENCE_LENGTH) {
- val word = bcVocabHash.value.get(iter.next())
- word match {
- case Some(w) =>
- sentence += w
- sentenceLength += 1
- case None =>
- }
+ // each partition is a collection of sentences,
+ // will be translated into arrays of Index integer
+ val sentences: RDD[Array[Int]] = dataset.mapPartitions { sentenceIter
=>
+ // Each sentence will map to 0 or more Array[Int]
+ sentenceIter.flatMap { sentence => {
+ // Sentence of words, some of which map to a word index
+ val wordIndexes = sentence.flatMap(bcVocabHash.value.get)
+ if (wordIndexes.nonEmpty) {
--- End diff --
They should be equivalent Scala code. We don't need to mix in their
meanings. Let us try providing an example of `sentenceIter:
Iterator[Array[String]]` and `vocabHash: Map[String, Int]` such that
~~~scala
sentenceIter.flatMap { sentence =>
val wordIndexes = sentence.flatMap(vocabHash.get)
wordIndexes.grouped(maxSentenceLength).map(_.toArray)
}
~~~~
returns a different result from your code:
~~~scala
sentenceIter.flatMap { sentence =>
val wordIndexes = sentence.flatMap(vocabHash.get)
if (wordIndexes.nonEmpty) {
val sentenceSplit = wordIndexes.grouped(maxSentenceLength)
sentenceSplit.map(_.toArray)
} else {
None
}
}
~~~
Essentially we are comparing the behavior when `wordIndexes` is an empty
`Iterator[Int]`. In this case, no matter what value `maxSentenceLength` takes,
` wordIndexes.grouped(maxSentenceLength).map(_.toArray)` returns an empty
iterator, which will be skipped by `sentenceIter.flatMap`. So it is the same as
`None` being skipped by `sentenceIter.flatMap` in the current implementation.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
