Revert back to Spark 2.3 (#399)

README.md

@@ -128,7 +128,7 @@ Start by picking TransmogrifAI version to match your project dependencies from t
 
 | TransmogrifAI Version | Spark Version | Scala Version | Java Version |
 |-------------------------------------------------|:-------------:|:-------------:|:------------:|
-| 0.6.1 (unreleased, master) | 2.4 | 2.11 | 1.8 |
+| 0.6.1 (unreleased, master) | 2.3 | 2.11 | 1.8 |
 | **0.6.0 (stable)**, 0.5.3, 0.5.2, 0.5.1, 0.5.0 | **2.3** | **2.11** | **1.8** |
 | 0.4.0, 0.3.4 | 2.2 | 2.11 | 1.8 |
 

build.gradle

@@ -1,16 +1,15 @@
 buildscript {
  repositories {
- mavenCentral()
- jcenter()
  maven { url "https://plugins.gradle.org/m2/" }
+ mavenCentral()
  }
  dependencies {
  classpath 'org.github.ngbinh.scalastyle:gradle-scalastyle-plugin_2.11:1.0.1'
- classpath 'com.commercehub.gradle.plugin:gradle-avro-plugin:0.16.0'
  }
 }
 
 plugins {
+ id 'com.commercehub.gradle.plugin.avro' version '0.8.0'
  id 'org.scoverage' version '2.5.0'
  id 'net.minecrell.licenser' version '0.4.1'
  id 'com.github.jk1.dependency-license-report' version '0.5.0'
@@ -58,13 +57,14 @@ configure(allProjs) {
  scalaVersionRevision = '12'
  scalaTestVersion = '3.0.5'
  scalaCheckVersion = '1.14.0'
- junitVersion = '4.12'
- avroVersion = '1.8.2'
- sparkVersion = '2.4.3'
+ junitVersion = '4.11'
+ avroVersion = '1.7.7'
+ sparkVersion = '2.3.2'
+ sparkAvroVersion = '4.0.0'
  scalaGraphVersion = '1.12.5'
  scalafmtVersion = '1.5.1'
  hadoopVersion = 'hadoop2'
- json4sVersion = '3.5.3' // matches Spark dependency version
+ json4sVersion = '3.2.11' // matches Spark dependency version
  jodaTimeVersion = '2.9.4'
  jodaConvertVersion = '1.8.1'
  algebirdVersion = '0.13.4'
@@ -75,20 +75,20 @@ configure(allProjs) {
  googleLibPhoneNumberVersion = '8.8.5'
  googleGeoCoderVersion = '2.82'
  googleCarrierVersion = '1.72'
- chillVersion = '0.9.3'
+ chillVersion = '0.8.4'
  reflectionsVersion = '0.9.11'
  collectionsVersion = '3.2.2'
  optimaizeLangDetectorVersion = '0.0.1'
  tikaVersion = '1.22'
- sparkTestingBaseVersion = '2.4.3_0.12.0'
+ sparkTestingBaseVersion = '2.3.1_0.10.0'
  sourceCodeVersion = '0.1.3'
  pegdownVersion = '1.4.2'
  commonsValidatorVersion = '1.6'
  commonsIOVersion = '2.6'
  scoveragePluginVersion = '1.3.1'
- xgboostVersion = '0.90'
+ xgboostVersion = '0.81'
  akkaSlf4jVersion = '2.3.11'
- mleapVersion = '0.14.0'
+ mleapVersion = '0.13.0'
  memoryFilesystemVersion = '2.1.0'
  }
 

cli/build.gradle

@@ -69,16 +69,20 @@ task copyTemplates(type: Copy) {
  fileName.replace(".gradle.template", ".gradle")
  }
  expand([
+ databaseHostname: 'db.company.com',
  version: scalaVersion,
  scalaVersion: scalaVersion,
  scalaVersionRevision: scalaVersionRevision,
  scalaTestVersion: scalaTestVersion,
  junitVersion: junitVersion,
  sparkVersion: sparkVersion,
  avroVersion: avroVersion,
+ sparkAvroVersion: sparkAvroVersion,
  hadoopVersion: hadoopVersion,
  collectionsVersion: collectionsVersion,
- transmogrifaiVersion: version
+ transmogrifaiVersion: version,
+ buildNumber: (int)(Math.random() * 1000),
+ date: new Date()
  ])
 }
 

cli/src/main/scala/com/salesforce/op/cli/SchemaSource.scala

@@ -138,7 +138,7 @@ case class AutomaticSchema(recordClassName: String)(dataFile: File) extends Sche
  case Some(actualType) =>
  val newSchema = Schema.create(actualType)
  val schemaField =
- new Schema.Field(field.name, newSchema, "auto-generated", orgSchemaField.defaultVal())
+ new Schema.Field(field.name, newSchema, "auto-generated", orgSchemaField.defaultValue)
  AvroField.from(schemaField)
  }
  } else field

cli/src/test/scala/com/salesforce/op/cli/gen/AvroFieldTest.scala

@@ -69,7 +69,7 @@ class AvroFieldTest extends FlatSpec with TestCommon with Assertions {
  val allSchemas = (enum::unions)++simpleSchemas // NULL does not work
 
  val fields = allSchemas.zipWithIndex map {
- case (s, i) => new Schema.Field("x" + i, s, "Who", null: Object)
+ case (s, i) => new Schema.Field("x" + i, s, "Who", null)
  }
 
  val expected = List(
@@ -86,7 +86,7 @@ class AvroFieldTest extends FlatSpec with TestCommon with Assertions {
 
  an[IllegalArgumentException] should be thrownBy {
  val nullSchema = Schema.create(Schema.Type.NULL)
- val nullField = new Schema.Field("xxx", null, "Nobody", null: Object)
+ val nullField = new Schema.Field("xxx", null, "Nobody", null)
  AvroField from nullField
  }
 

core/src/main/scala/com/salesforce/op/stages/impl/classification/OpLinearSVC.scala

@@ -152,14 +152,14 @@ class OpLinearSVCModel
 ) extends OpPredictorWrapperModel[LinearSVCModel](uid = uid, operationName = operationName, sparkModel = sparkModel) {
 
  @transient lazy private val predictRaw = reflectMethod(getSparkMlStage().get, "predictRaw")
- @transient lazy private val predict: Vector => Double = getSparkMlStage().get.predict(_)
+ @transient lazy private val predict = reflectMethod(getSparkMlStage().get, "predict")
 
  /**
  * Function used to convert input to output
  */
  override def transformFn: (RealNN, OPVector) => Prediction = (label, features) => {
  val raw = predictRaw(features.value).asInstanceOf[Vector]
- val pred = predict(features.value)
+ val pred = predict(features.value).asInstanceOf[Double]
 
  Prediction(rawPrediction = raw, prediction = pred)
  }

core/src/main/scala/com/salesforce/op/stages/impl/classification/OpXGBoostClassifier.scala

@@ -235,11 +235,6 @@ class OpXGBoostClassifier(uid: String = UID[OpXGBoostClassifier])
  */
  def setMaxBins(value: Int): this.type = set(maxBins, value)
 
- /**
- * Maximum number of nodes to be added. Only relevant when grow_policy=lossguide is set.
- */
- def setMaxLeaves(value: Int): this.type = set(maxLeaves, value)
-
  /**
  * This is only used for approximate greedy algorithm.
  * This roughly translated into O(1 / sketch_eps) number of bins. Compared to directly select
@@ -287,19 +282,8 @@ class OpXGBoostClassifier(uid: String = UID[OpXGBoostClassifier])
  def setLambdaBias(value: Double): this.type = set(lambdaBias, value)
 
  // setters for learning params
-
- /**
- * Specify the learning task and the corresponding learning objective.
- * options: reg:squarederror, reg:logistic, binary:logistic, binary:logitraw, count:poisson,
- * multi:softmax, multi:softprob, rank:pairwise, reg:gamma. default: reg:squarederror
- */
  def setObjective(value: String): this.type = set(objective, value)
 
- /**
- * Objective type used for training. For options see [[ml.dmlc.xgboost4j.scala.spark.params.LearningTaskParams]]
- */
- def setObjectiveType(value: String): this.type = set(objectiveType, value)
-
  /**
  * Specify the learning task and the corresponding learning objective.
  * options: reg:linear, reg:logistic, binary:logistic, binary:logitraw, count:poisson,
@@ -326,11 +310,6 @@ class OpXGBoostClassifier(uid: String = UID[OpXGBoostClassifier])
  */
  def setNumEarlyStoppingRounds(value: Int): this.type = set(numEarlyStoppingRounds, value)
 
- /**
- * Define the expected optimization to the evaluation metrics, true to maximize otherwise minimize it
- */
- def setMaximizeEvaluationMetrics(value: Boolean): this.type = set(maximizeEvaluationMetrics, value)
-
  /**
  * Customized objective function provided by user. default: null
  */
@@ -380,18 +359,17 @@ class OpXGBoostClassificationModel
 
  private lazy val model = getSparkMlStage().get
  private lazy val booster = model.nativeBooster
- private lazy val treeLimit = model.getTreeLimit
+ private lazy val treeLimit = model.getTreeLimit.toInt
  private lazy val missing = model.getMissing
 
  override def transformFn: (RealNN, OPVector) => Prediction = (label, features) => {
- val data = processMissingValues(Iterator(features.value.asXGB), missing)
+ val data = removeMissingValues(Iterator(features.value.asXGB), missing)
  val dm = new DMatrix(dataIter = data)
  val rawPred = booster.predict(dm, outPutMargin = true, treeLimit = treeLimit)(0).map(_.toDouble)
  val rawPrediction = if (model.numClasses == 2) Array(-rawPred(0), rawPred(0)) else rawPred
  val prob = booster.predict(dm, outPutMargin = false, treeLimit = treeLimit)(0).map(_.toDouble)
  val probability = if (model.numClasses == 2) Array(1.0 - prob(0), prob(0)) else prob
  val prediction = probability2predictionMirror(Vectors.dense(probability)).asInstanceOf[Double]
-
  Prediction(prediction = prediction, rawPrediction = rawPrediction, probability = probability)
  }
 }

core/src/main/scala/com/salesforce/op/stages/impl/regression/OpDecisionTreeRegressor.scala

@@ -34,6 +34,7 @@ import com.salesforce.op.UID
 import com.salesforce.op.features.types.{OPVector, Prediction, RealNN}
 import com.salesforce.op.stages.impl.CheckIsResponseValues
 import com.salesforce.op.stages.sparkwrappers.specific.{OpPredictionModel, OpPredictorWrapper}
+import com.salesforce.op.utils.reflection.ReflectionUtils.reflectMethod
 import org.apache.spark.ml.regression.{DecisionTreeRegressionModel, DecisionTreeRegressor, OpDecisionTreeRegressorParams}
 
 import scala.reflect.runtime.universe.TypeTag
@@ -112,4 +113,7 @@ class OpDecisionTreeRegressionModel
  ttov: TypeTag[Prediction#Value]
 ) extends OpPredictionModel[DecisionTreeRegressionModel](
  sparkModel = sparkModel, uid = uid, operationName = operationName
-)
+) {
+ @transient lazy val predictMirror = reflectMethod(getSparkMlStage().get, "predict")
+}
+

core/src/main/scala/com/salesforce/op/stages/impl/regression/OpGBTRegressor.scala

@@ -34,6 +34,7 @@ import com.salesforce.op.UID
 import com.salesforce.op.features.types.{OPVector, Prediction, RealNN}
 import com.salesforce.op.stages.impl.CheckIsResponseValues
 import com.salesforce.op.stages.sparkwrappers.specific.{OpPredictionModel, OpPredictorWrapper}
+import com.salesforce.op.utils.reflection.ReflectionUtils.reflectMethod
 import org.apache.spark.ml.regression.{GBTRegressionModel, GBTRegressor, OpGBTRegressorParams}
 
 import scala.reflect.runtime.universe.TypeTag
@@ -138,4 +139,7 @@ class OpGBTRegressionModel
  ttov: TypeTag[Prediction#Value]
 ) extends OpPredictionModel[GBTRegressionModel](
  sparkModel = sparkModel, uid = uid, operationName = operationName
-)
+) {
+ @transient lazy val predictMirror = reflectMethod(getSparkMlStage().get, "predict")
+}
+

core/src/main/scala/com/salesforce/op/stages/impl/regression/OpLinearRegression.scala

@@ -34,6 +34,7 @@ import com.salesforce.op._
 import com.salesforce.op.features.types.{OPVector, Prediction, RealNN}
 import com.salesforce.op.stages.impl.CheckIsResponseValues
 import com.salesforce.op.stages.sparkwrappers.specific.{OpPredictionModel, OpPredictorWrapper}
+import com.salesforce.op.utils.reflection.ReflectionUtils.reflectMethod
 import org.apache.spark.ml.regression.{LinearRegression, LinearRegressionModel, OpLinearRegressionParams}
 
 import scala.reflect.runtime.universe.TypeTag
@@ -179,4 +180,7 @@ class OpLinearRegressionModel
  ttov: TypeTag[Prediction#Value]
 ) extends OpPredictionModel[LinearRegressionModel](
  sparkModel = sparkModel, uid = uid, operationName = operationName
-)
+) {
+ @transient lazy val predictMirror = reflectMethod(getSparkMlStage().get, "predict")
+}
+

core/src/main/scala/com/salesforce/op/stages/impl/regression/OpRandomForestRegressor.scala

@@ -34,6 +34,7 @@ import com.salesforce.op.UID
 import com.salesforce.op.features.types.{OPVector, Prediction, RealNN}
 import com.salesforce.op.stages.impl.CheckIsResponseValues
 import com.salesforce.op.stages.sparkwrappers.specific.{OpPredictionModel, OpPredictorWrapper}
+import com.salesforce.op.utils.reflection.ReflectionUtils.reflectMethod
 import org.apache.spark.ml.regression.{OpRandomForestRegressorParams, RandomForestRegressionModel, RandomForestRegressor}
 
 import scala.reflect.runtime.universe.TypeTag
@@ -125,4 +126,8 @@ class OpRandomForestRegressionModel
  ttov: TypeTag[Prediction#Value]
 ) extends OpPredictionModel[RandomForestRegressionModel](
  sparkModel = sparkModel, uid = uid, operationName = operationName
-)
+) {
+ @transient lazy val predictMirror = reflectMethod(getSparkMlStage().get, "predict")
+}
+
+

core/src/main/scala/com/salesforce/op/stages/impl/regression/OpXGBoostRegressor.scala

@@ -34,6 +34,7 @@ import com.salesforce.op.UID
 import com.salesforce.op.features.types.{OPVector, Prediction, RealNN}
 import com.salesforce.op.stages.impl.CheckIsResponseValues
 import com.salesforce.op.stages.sparkwrappers.specific.{OpPredictionModel, OpPredictorWrapper}
+import com.salesforce.op.utils.reflection.ReflectionUtils.reflectMethod
 import ml.dmlc.xgboost4j.scala.{EvalTrait, ObjectiveTrait}
 import ml.dmlc.xgboost4j.scala.spark.{OpXGBoostRegressorParams, TrackerConf, XGBoostRegressionModel, XGBoostRegressor}
 
@@ -233,11 +234,6 @@ class OpXGBoostRegressor(uid: String = UID[OpXGBoostRegressor])
  */
  def setMaxBins(value: Int): this.type = set(maxBins, value)
 
- /**
- * Maximum number of nodes to be added. Only relevant when grow_policy=lossguide is set.
- */
- def setMaxLeaves(value: Int): this.type = set(maxLeaves, value)
-
  /**
  * This is only used for approximate greedy algorithm.
  * This roughly translated into O(1 / sketch_eps) number of bins. Compared to directly select
@@ -285,19 +281,8 @@ class OpXGBoostRegressor(uid: String = UID[OpXGBoostRegressor])
  def setLambdaBias(value: Double): this.type = set(lambdaBias, value)
 
  // setters for learning params
-
- /**
- * Specify the learning task and the corresponding learning objective.
- * options: reg:squarederror, reg:logistic, binary:logistic, binary:logitraw, count:poisson,
- * multi:softmax, multi:softprob, rank:pairwise, reg:gamma. default: reg:squarederror
- */
  def setObjective(value: String): this.type = set(objective, value)
 
- /**
- * Objective type used for training. For options see [[ml.dmlc.xgboost4j.scala.spark.params.LearningTaskParams]]
- */
- def setObjectiveType(value: String): this.type = set(objectiveType, value)
-
  /**
  * Specify the learning task and the corresponding learning objective.
  * options: reg:linear, reg:logistic, binary:logistic, binary:logitraw, count:poisson,
@@ -324,11 +309,6 @@ class OpXGBoostRegressor(uid: String = UID[OpXGBoostRegressor])
  */
  def setNumEarlyStoppingRounds(value: Int): this.type = set(numEarlyStoppingRounds, value)
 
- /**
- * Define the expected optimization to the evaluation metrics, true to maximize otherwise minimize it
- */
- def setMaximizeEvaluationMetrics(value: Boolean): this.type = set(maximizeEvaluationMetrics, value)
-
  /**
  * Customized objective function provided by user. default: null
  */
@@ -361,4 +341,6 @@ class OpXGBoostRegressionModel
  ttov: TypeTag[Prediction#Value]
 ) extends OpPredictionModel[XGBoostRegressionModel](
  sparkModel = sparkModel, uid = uid, operationName = operationName
-)
+) {
+ @transient lazy val predictMirror = reflectMethod(getSparkMlStage().get, "predict")
+}

core/src/main/scala/com/salesforce/op/stages/impl/selector/ModelSelector.scala

@@ -41,6 +41,7 @@ import com.salesforce.op.stages.impl.tuning.{BestEstimator, _}
 import com.salesforce.op.stages.sparkwrappers.generic.SparkWrapperParams
 import com.salesforce.op.stages.sparkwrappers.specific.{OpPredictorWrapperModel, SparkModelConverter}
 import com.salesforce.op.utils.spark.RichMetadata._
+import com.salesforce.op.utils.spark.RichDataset._
 import com.salesforce.op.utils.spark.RichParamMap._
 import com.salesforce.op.utils.stages.FitStagesUtil._
 import org.apache.spark.ml.param._

core/src/main/scala/com/salesforce/op/stages/sparkwrappers/specific/OpPredictionModel.scala

@@ -52,10 +52,9 @@ abstract class OpPredictionModel[T <: PredictionModel[Vector, T]]
  operationName: String
 ) extends OpPredictorWrapperModel[T](uid = uid, operationName = operationName, sparkModel = sparkModel) {
 
- /**
- * Predict label for the given features
- */
- @transient protected lazy val predict: Vector => Double = getSparkMlStage().get.predict(_)
+ protected def predictMirror: MethodMirror
+
+ protected def predict(features: Vector): Double = predictMirror.apply(features).asInstanceOf[Double]
 
  /**
  * Function used to convert input to output

core/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostParams.scala

@@ -104,8 +104,8 @@ case object OpXGBoost {
  }
 
  /**
- * Hack to access [[ml.dmlc.xgboost4j.scala.spark.XGBoost.processMissingValues]] private method
+ * Hack to access [[ml.dmlc.xgboost4j.scala.spark.XGBoost.removeMissingValues]] private method
  */
- def processMissingValues(xgbLabelPoints: Iterator[LabeledPoint], missing: Float): Iterator[LabeledPoint] =
- XGBoost.processMissingValues(xgbLabelPoints, missing)
+ def removeMissingValues(xgbLabelPoints: Iterator[LabeledPoint], missing: Float): Iterator[LabeledPoint] =
+ XGBoost.removeMissingValues(xgbLabelPoints, missing)
 }