srowen commented on a change in pull request #25802: [SPARK-29095][ML] add
extractInstances
URL: https://github.com/apache/spark/pull/25802#discussion_r326706844
##########
File path: mllib/src/main/scala/org/apache/spark/ml/Predictor.scala
##########
@@ -62,6 +62,40 @@ private[ml] trait PredictorParams extends Params
}
SchemaUtils.appendColumn(schema, $(predictionCol), DoubleType)
}
+
+ /**
+ * Extract [[labelCol]], weightCol(if any) and [[featuresCol]] from the
given dataset,
+ * and put it in an RDD with strong types.
+ */
+ protected def extractInstances(dataset: Dataset[_]): RDD[Instance] = {
+ val w = this match {
+ case p: HasWeightCol =>
+ if (isDefined(p.weightCol) && $(p.weightCol).nonEmpty) {
+ col($(p.weightCol)).cast(DoubleType)
+ } else {
+ lit(1.0)
+ }
+ case _ => lit(1.0)
Review comment:
OK, I suppose I'm just concerned about the small overhead of adding an
unused column.
You're saying that it's up to implementations to call the method they need,
one with weights or not? yeah I agree, and they will call the right method in
this change? If true, then do you even need this check? it will already fail
(correctly) if the code is calling the wrong method.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
