1/*
2 Copyright (c) 2014 by Contributors
3
4 Licensed under the Apache License, Version 2.0 (the "License");
5 you may not use this file except in compliance with the License.
6 You may obtain a copy of the License at
7
8 http://www.apache.org/licenses/LICENSE-2.0
9
10 Unless required by applicable law or agreed to in writing, software
11 distributed under the License is distributed on an "AS IS" BASIS,
12 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 See the License for the specific language governing permissions and
14 limitations under the License.
15 */
16
17package ml.dmlc.xgboost4j.scala.spark
18
19import ml.dmlc.xgboost4j.java.XGBoostError
20import org.apache.spark.Partitioner
21import org.apache.spark.ml.feature.VectorAssembler
22import org.apache.spark.sql.SparkSession
23import org.scalatest.FunSuite
24import org.apache.spark.sql.functions._
25
26import scala.util.Random
27
28class FeatureSizeValidatingSuite extends FunSuite with PerTest {
29
30  test("transform throwing exception if feature size of dataset is greater than model's") {
31    val modelPath = getClass.getResource("/model/0.82/model").getPath
32    val model = XGBoostClassificationModel.read.load(modelPath)
33    val r = new Random(0)
34    // 0.82/model was trained with 251 features. and transform will throw exception
35    // if feature size of data is not equal to 251
36    var df = ss.createDataFrame(Seq.fill(100)(r.nextInt(2)).map(i => (i, i))).
37      toDF("feature", "label")
38    for (x <- 1 to 252) {
39      df = df.withColumn(s"feature_${x}", lit(1))
40    }
41    val assembler = new VectorAssembler()
42      .setInputCols(df.columns.filter(!_.contains("label")))
43      .setOutputCol("features")
44    val thrown = intercept[Exception] {
45      model.transform(assembler.transform(df)).show()
46    }
47    assert(thrown.getMessage.contains(
48      "Number of columns does not match number of features in booster"))
49  }
50
51  test("train throwing exception if feature size of dataset is different on distributed train") {
52    val paramMap = Map("eta" -> "1", "max_depth" -> "6", "silent" -> "1",
53      "objective" -> "binary:logistic",
54      "num_round" -> 5, "num_workers" -> 2, "use_external_memory" -> true, "missing" -> 0)
55    import DataUtils._
56    val sparkSession = SparkSession.builder().getOrCreate()
57    import sparkSession.implicits._
58    val repartitioned = sc.parallelize(Synthetic.trainWithDiffFeatureSize, 2)
59      .map(lp => (lp.label, lp)).partitionBy(
60      new Partitioner {
61        override def numPartitions: Int = 2
62
63        override def getPartition(key: Any): Int = key.asInstanceOf[Float].toInt
64      }
65    ).map(_._2).zipWithIndex().map {
66      case (lp, id) =>
67        (id, lp.label, lp.features)
68    }.toDF("id", "label", "features")
69    val xgb = new XGBoostClassifier(paramMap)
70    intercept[Exception] {
71      xgb.fit(repartitioned)
72    }
73  }
74}
75