[jvm-packages] delete all constraints from spark layer about obj and eval metrics and handle error in jvm layer (#4560)

* temp

* prediction part

* remove supported*

* add for test

* fix param name

* add rabit

* update rabit

* return value of rabit init

* eliminate compilation warnings

* update rabit

* shutdown

* update rabit again

* check sparkcontext shutdown

* fix logic

* sleep

* fix tests

* test with relaxed threshold

* create new thread each time

* stop for job quitting

* udpate rabit

* update rabit

* update rabit

* update git modules

jvm-packages/xgboost4j-spark/src/test/scala/ml/dmlc/xgboost4j/scala/spark/ParameterSuite.scala

@@ -0,0 +1,81 @@
+/*
+ Copyright (c) 2014 by Contributors
+
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
+ */
+
+package ml.dmlc.xgboost4j.scala.spark
+
+import ml.dmlc.xgboost4j.java.XGBoostError
+import org.scalatest.{BeforeAndAfterAll, FunSuite}
+
+import org.apache.spark.ml.param.ParamMap
+
+class ParameterSuite extends FunSuite with PerTest with BeforeAndAfterAll {
+
+  test("XGBoost and Spark parameters synchronize correctly") {
+    val xgbParamMap = Map("eta" -> "1", "objective" -> "binary:logistic",
+      "objective_type" -> "classification")
+    // from xgboost params to spark params
+    val xgb = new XGBoostClassifier(xgbParamMap)
+    assert(xgb.getEta === 1.0)
+    assert(xgb.getObjective === "binary:logistic")
+    assert(xgb.getObjectiveType === "classification")
+    // from spark to xgboost params
+    val xgbCopy = xgb.copy(ParamMap.empty)
+    assert(xgbCopy.MLlib2XGBoostParams("eta").toString.toDouble === 1.0)
+    assert(xgbCopy.MLlib2XGBoostParams("objective").toString === "binary:logistic")
+    assert(xgbCopy.MLlib2XGBoostParams("objective_type").toString === "classification")
+    val xgbCopy2 = xgb.copy(ParamMap.empty.put(xgb.evalMetric, "logloss"))
+    assert(xgbCopy2.MLlib2XGBoostParams("eval_metric").toString === "logloss")
+  }
+
+  private def waitForSparkContextShutdown(): Unit = {
+    var totalWaitedTime = 0L
+    while (!ss.sparkContext.isStopped && totalWaitedTime <= 120000) {
+      Thread.sleep(10000)
+      totalWaitedTime += 10000
+    }
+    assert(ss.sparkContext.isStopped === true)
+  }
+
+  test("fail training elegantly with unsupported objective function") {
+    val paramMap = Map("eta" -> "0.1", "max_depth" -> "6", "silent" -> "1",
+      "objective" -> "wrong_objective_function", "num_class" -> "6", "num_round" -> 5,
+      "num_workers" -> numWorkers)
+    val trainingDF = buildDataFrame(MultiClassification.train)
+    val xgb = new XGBoostClassifier(paramMap)
+    try {
+      val model = xgb.fit(trainingDF)
+    } catch {
+      case e: Throwable => // swallow anything
+    } finally {
+      waitForSparkContextShutdown()
+    }
+  }
+
+  test("fail training elegantly with unsupported eval metrics") {
+    val paramMap = Map("eta" -> "0.1", "max_depth" -> "6", "silent" -> "1",
+      "objective" -> "multi:softmax", "num_class" -> "6", "num_round" -> 5,
+      "num_workers" -> numWorkers, "eval_metric" -> "wrong_eval_metrics")
+    val trainingDF = buildDataFrame(MultiClassification.train)
+    val xgb = new XGBoostClassifier(paramMap)
+    try {
+      val model = xgb.fit(trainingDF)
+    } catch {
+      case e: Throwable => // swallow anything
+    } finally {
+      waitForSparkContextShutdown()
+    }
+  }
+}

jvm-packages/xgboost4j-spark/src/test/scala/ml/dmlc/xgboost4j/scala/spark/PerTest.scala

@@ -19,10 +19,10 @@ package ml.dmlc.xgboost4j.scala.spark
 import java.io.File
 
 import ml.dmlc.xgboost4j.{LabeledPoint => XGBLabeledPoint}
-import org.apache.spark.{SparkConf, SparkContext}
+
+import org.apache.spark.{SparkConf, SparkContext, TaskFailedListener}
 import org.apache.spark.sql._
 import org.scalatest.{BeforeAndAfterEach, FunSuite}
-
 import scala.util.Random
 
 trait PerTest extends BeforeAndAfterEach { self: FunSuite =>
@@ -50,6 +50,7 @@ trait PerTest extends BeforeAndAfterEach { self: FunSuite =>
         cleanExternalCache(currentSession.sparkContext.appName)
         currentSession = null
       }
+      TaskFailedListener.killerStarted = false
     }
   }
 

jvm-packages/xgboost4j-spark/src/test/scala/ml/dmlc/xgboost4j/scala/spark/RabitRobustnessSuite.scala

@@ -29,7 +29,7 @@ import org.apache.spark.{SparkConf, SparkContext}
 import org.scalatest.FunSuite
 
 
-class RabitSuite extends FunSuite with PerTest {
+class RabitRobustnessSuite extends FunSuite with PerTest {
 
   test("training with Scala-implemented Rabit tracker") {
     val eval = new EvalError()

jvm-packages/xgboost4j-spark/src/test/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostClassifierSuite.scala

@@ -160,23 +160,6 @@ class XGBoostClassifierSuite extends FunSuite with PerTest {
     assert(model.summary.validationObjectiveHistory.isEmpty)
   }
 
-  test("XGBoost and Spark parameters synchronize correctly") {
-    val xgbParamMap = Map("eta" -> "1", "objective" -> "binary:logistic",
-      "objective_type" -> "classification")
-    // from xgboost params to spark params
-    val xgb = new XGBoostClassifier(xgbParamMap)
-    assert(xgb.getEta === 1.0)
-    assert(xgb.getObjective === "binary:logistic")
-    assert(xgb.getObjectiveType === "classification")
-    // from spark to xgboost params
-    val xgbCopy = xgb.copy(ParamMap.empty)
-    assert(xgbCopy.MLlib2XGBoostParams("eta").toString.toDouble === 1.0)
-    assert(xgbCopy.MLlib2XGBoostParams("objective").toString === "binary:logistic")
-    assert(xgbCopy.MLlib2XGBoostParams("objective_type").toString === "classification")
-    val xgbCopy2 = xgb.copy(ParamMap.empty.put(xgb.evalMetric, "logloss"))
-    assert(xgbCopy2.MLlib2XGBoostParams("eval_metric").toString === "logloss")
-  }
-
   test("multi class classification") {
     val paramMap = Map("eta" -> "0.1", "max_depth" -> "6", "silent" -> "1",
       "objective" -> "multi:softmax", "num_class" -> "6", "num_round" -> 5,