Github user huaxingao commented on the issue:
https://github.com/apache/spark/pull/22790
I added a regression test in
```org.apache.spark.mllib.clustering.BisectingKMeansSuite```
I could add the following test in ml package.
```
test("SPARK-25793") {
val bisectingKMeans = new BisectingKMeans()
bisectingKMeans.setDistanceMeasure(DistanceMeasure.COSINE)
val readBisectingKMeans = testDefaultReadWrite(bisectingKMeans)
assert(bisectingKMeans.distanceMeasure ===
readBisectingKMeans.distanceMeasure)
}
```
But the bug doesn't really affect the above test. With the bug, even though
mllib ```BisectingKMeansModel.load``` will call V1_0 load and gives a model
with default value of distanceMeasure, in ml package,
```BisectingKMeansModelReader.load``` will call
```metadata.getAndSetParams(model)``` which will set the distanceMeasure to the
correct value (DistanceMeasure.COSINE).
```
override def load(path: String): BisectingKMeansModel = {
val metadata = DefaultParamsReader.loadMetadata(path, sc, className)
val dataPath = new Path(path, "data").toString
val mllibModel = MLlibBisectingKMeansModel.load(sc, dataPath)
val model = new BisectingKMeansModel(metadata.uid, mllibModel)
metadata.getAndSetParams(model)
model
}
```
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
