[SPARK-4328] Python serialization updates make Python ML API more brittle to types

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 1.2.0
• Component/s: MLlib, PySpark
• Assignee: Unassigned
• Reporter: Joseph K. Bradley
• Created: 10/Nov/14 22:43
• Updated: 10/Nov/14 22:56
• Resolved: 10/Nov/14 22:56

Description

In Spark 1.1, you could create a LabeledPoint with labels specified as integers, and then use it with LinearRegression. This was broken by the Python API updates since then. E.g., this code runs in the 1.1 branch but not in the current master:

from pyspark.mllib.regression import *
import numpy
features = numpy.ndarray((3))
data = sc.parallelize([LabeledPoint(1, features)])
LinearRegressionWithSGD.train(data)

Recommendation: Allow users to use integers from Python.

The error message you get is:

py4j.protocol.Py4JJavaError: An error occurred while calling o55.trainLinearRegressionModelWithSGD.
: org.apache.spark.SparkException: Job aborted due to stage failure: Task 7 in stage 3.0 failed 1 times, most recent failure: Lost task 7.0 in stage 3.0 (TID 15, localhost): java.lang.ClassCastException: java.lang.Integer cannot be cast to java.lang.Double
	at scala.runtime.BoxesRunTime.unboxToDouble(BoxesRunTime.java:119)
	at org.apache.spark.mllib.api.python.SerDe$LabeledPointPickler.construct(PythonMLLibAPI.scala:727)
	at net.razorvine.pickle.Unpickler.load_reduce(Unpickler.java:617)
	at net.razorvine.pickle.Unpickler.dispatch(Unpickler.java:170)
	at net.razorvine.pickle.Unpickler.load(Unpickler.java:84)
	at net.razorvine.pickle.Unpickler.loads(Unpickler.java:97)
	at org.apache.spark.mllib.api.python.SerDe$$anonfun$pythonToJava$1$$anonfun$apply$1.apply(PythonMLLibAPI.scala:804)
	at org.apache.spark.mllib.api.python.SerDe$$anonfun$pythonToJava$1$$anonfun$apply$1.apply(PythonMLLibAPI.scala:803)
	at scala.collection.Iterator$$anon$13.hasNext(Iterator.scala:371)
	at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:327)
	at org.apache.spark.util.Utils$.getIteratorSize(Utils.scala:1309)
	at org.apache.spark.rdd.RDD$$anonfun$count$1.apply(RDD.scala:910)
	at org.apache.spark.rdd.RDD$$anonfun$count$1.apply(RDD.scala:910)
	at org.apache.spark.SparkContext$$anonfun$runJob$3.apply(SparkContext.scala:1223)
	at org.apache.spark.SparkContext$$anonfun$runJob$3.apply(SparkContext.scala:1223)
	at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:61)
	at org.apache.spark.scheduler.Task.run(Task.scala:56)
	at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:195)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
	at java.lang.Thread.run(Thread.java:745)

Issue Links

• duplicates: SPARK-4324 Support numpy/scipy in all Python API of MLlib
• is related to: SPARK-4324 Support numpy/scipy in all Python API of MLlib