Matthew Tovbin created SPARK-23152:
--------------------------------------
Summary: Invalid guard condition in
org.apache.spark.ml.classification.Classifier
Key: SPARK-23152
URL: https://issues.apache.org/jira/browse/SPARK-23152
Project: Spark
Issue Type: Bug
Components: ML, MLlib
Affects Versions: 2.1.2, 2.1.1, 2.1.0, 2.0.2, 2.0.1, 2.0.0, 2.1.3, 2.3.0,
2.3.1
Reporter: Matthew Tovbin
When fitting a classifier that extends
"org.apache.spark.ml.classification.Classifier" (NaiveBayes,
DecisionTreeClassifier, RandomForestClassifier) a NullPointerException is
thrown.
Steps to reproduce:
{code:java}
val data = spark.createDataset(Seq.empty[(Double,
org.apache.spark.ml.linalg.Vector)])
new DecisionTreeClassifier()
.setLabelCol("_1").setFeaturesCol("_2").fit(data)
{code}
The error:
{code:java}
java.lang.NullPointerException: Value at index 0 is null
at org.apache.spark.sql.Row$class.getAnyValAs(Row.scala:472)
at org.apache.spark.sql.Row$class.getDouble(Row.scala:248)
at
org.apache.spark.sql.catalyst.expressions.GenericRow.getDouble(rows.scala:165)
at
org.apache.spark.ml.classification.Classifier.getNumClasses(Classifier.scala:115)
at
org.apache.spark.ml.classification.DecisionTreeClassifier.train(DecisionTreeClassifier.scala:102)
at
org.apache.spark.ml.classification.DecisionTreeClassifier.train(DecisionTreeClassifier.scala:45)
at org.apache.spark.ml.Predictor.fit(Predictor.scala:118){code}
The problem happens due to an incorrect guard condition in
org.apache.spark.ml.classification.Classifier:getNumClasses
{code:java}
val maxLabelRow: Array[Row] = dataset.select(max($(labelCol))).take(1)
if (maxLabelRow.isEmpty) {
throw new SparkException("ML algorithm was given empty dataset.")
}
{code}
When the input data is empty the "maxLabelRow" array is not. It contains a
single element with no columns in it. Therefore the condition has to be
modified to verify that.
{code:java}
maxLabelRow.size <= 1
{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
