Github user tashoyan commented on a diff in the pull request:
https://github.com/apache/spark/pull/20578#discussion_r167442376
--- Diff: mllib/src/main/scala/org/apache/spark/ml/fpm/FPGrowth.scala ---
@@ -158,18 +159,30 @@ class FPGrowth @Since("2.2.0") (
}
private def genericFit[T: ClassTag](dataset: Dataset[_]): FPGrowthModel
= {
+ val handlePersistence = dataset.storageLevel == StorageLevel.NONE
+
val data = dataset.select($(itemsCol))
- val items = data.where(col($(itemsCol)).isNotNull).rdd.map(r =>
r.getSeq[T](0).toArray)
+ val items = data.where(col($(itemsCol)).isNotNull).rdd.map(r =>
r.getSeq[Any](0).toArray)
--- End diff --
An interesting curiosity for me: why FPGrowth contract requires `Array` of
items, not `Seq`? First, it's strange for the contract to require a specific
implementation rather than an interface. Second, this leads to redundant
`toArray` and back `toSeq` transformations. `Seq` would be more convenient, as
`Row` class has `getSeq` method but does not have `getArray`.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
