Revert "[SPARK-48883][ML][R] Replace RDD read / write API invocation …

…with Dataframe read / write API" This reverts commit 0fa5787.
apache · Jul 13, 2024 · cc32137 · cc32137
1 parent b4cd2ec
commit cc32137
Show file tree

Hide file tree

Showing 24 changed files with 45 additions and 114 deletions.
diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/AFTSurvivalRegressionWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/AFTSurvivalRegressionWrapper.scala
@@ -129,9 +129,7 @@ private[r] object AFTSurvivalRegressionWrapper extends MLReadable[AFTSurvivalReg
       val rMetadata = ("class" -> instance.getClass.getName) ~
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -144,8 +142,7 @@ private[r] object AFTSurvivalRegressionWrapper extends MLReadable[AFTSurvivalReg
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
 

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/ALSWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/ALSWrapper.scala
@@ -94,9 +94,7 @@ private[r] object ALSWrapper extends MLReadable[ALSWrapper] {
       val rMetadata = ("class" -> instance.getClass.getName) ~
         ("ratingCol" -> instance.ratingCol)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.alsModel.save(modelPath)
     }
@@ -109,8 +107,7 @@ private[r] object ALSWrapper extends MLReadable[ALSWrapper] {
       val rMetadataPath = new Path(path, "rMetadata").toString
       val modelPath = new Path(path, "model").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val ratingCol = (rMetadata \ "ratingCol").extract[String]
       val alsModel = ALSModel.load(modelPath)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/BisectingKMeansWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/BisectingKMeansWrapper.scala
@@ -120,9 +120,7 @@ private[r] object BisectingKMeansWrapper extends MLReadable[BisectingKMeansWrapp
         ("size" -> instance.size.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -135,8 +133,7 @@ private[r] object BisectingKMeansWrapper extends MLReadable[BisectingKMeansWrapp
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
       val size = (rMetadata \ "size").extract[Array[Long]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/DecisionTreeClassifierWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/DecisionTreeClassifierWrapper.scala
@@ -131,9 +131,7 @@ private[r] object DecisionTreeClassifierWrapper extends MLReadable[DecisionTreeC
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -146,8 +144,7 @@ private[r] object DecisionTreeClassifierWrapper extends MLReadable[DecisionTreeC
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val formula = (rMetadata \ "formula").extract[String]
       val features = (rMetadata \ "features").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/DecisionTreeRegressorWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/DecisionTreeRegressorWrapper.scala
@@ -114,9 +114,7 @@ private[r] object DecisionTreeRegressorWrapper extends MLReadable[DecisionTreeRe
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -129,8 +127,7 @@ private[r] object DecisionTreeRegressorWrapper extends MLReadable[DecisionTreeRe
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val formula = (rMetadata \ "formula").extract[String]
       val features = (rMetadata \ "features").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/FMClassifierWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/FMClassifierWrapper.scala
@@ -151,9 +151,7 @@ private[r] object FMClassifierWrapper
         ("features" -> instance.features.toImmutableArraySeq) ~
         ("labels" -> instance.labels.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -166,8 +164,7 @@ private[r] object FMClassifierWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
       val labels = (rMetadata \ "labels").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/FMRegressorWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/FMRegressorWrapper.scala
@@ -132,9 +132,7 @@ private[r] object FMRegressorWrapper
       val rMetadata = ("class" -> instance.getClass.getName) ~
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -147,8 +145,7 @@ private[r] object FMRegressorWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
 

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/FPGrowthWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/FPGrowthWrapper.scala
@@ -78,9 +78,7 @@ private[r] object FPGrowthWrapper extends MLReadable[FPGrowthWrapper] {
         "class" -> instance.getClass.getName
       ))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.fpGrowthModel.save(modelPath)
     }

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/GBTClassifierWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/GBTClassifierWrapper.scala
@@ -138,9 +138,7 @@ private[r] object GBTClassifierWrapper extends MLReadable[GBTClassifierWrapper]
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -153,8 +151,7 @@ private[r] object GBTClassifierWrapper extends MLReadable[GBTClassifierWrapper]
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val formula = (rMetadata \ "formula").extract[String]
       val features = (rMetadata \ "features").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/GBTRegressorWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/GBTRegressorWrapper.scala
@@ -122,9 +122,7 @@ private[r] object GBTRegressorWrapper extends MLReadable[GBTRegressorWrapper] {
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -137,8 +135,7 @@ private[r] object GBTRegressorWrapper extends MLReadable[GBTRegressorWrapper] {
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val formula = (rMetadata \ "formula").extract[String]
       val features = (rMetadata \ "features").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/GaussianMixtureWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/GaussianMixtureWrapper.scala
@@ -113,9 +113,7 @@ private[r] object GaussianMixtureWrapper extends MLReadable[GaussianMixtureWrapp
         ("logLikelihood" -> instance.logLikelihood)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -128,8 +126,7 @@ private[r] object GaussianMixtureWrapper extends MLReadable[GaussianMixtureWrapp
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val dim = (rMetadata \ "dim").extract[Int]
       val logLikelihood = (rMetadata \ "logLikelihood").extract[Double]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/GeneralizedLinearRegressionWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/GeneralizedLinearRegressionWrapper.scala
@@ -170,9 +170,7 @@ private[r] object GeneralizedLinearRegressionWrapper
         ("rAic" -> instance.rAic) ~
         ("rNumIterations" -> instance.rNumIterations)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -186,8 +184,7 @@ private[r] object GeneralizedLinearRegressionWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val rFeatures = (rMetadata \ "rFeatures").extract[Array[String]]
       val rCoefficients = (rMetadata \ "rCoefficients").extract[Array[Double]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/IsotonicRegressionWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/IsotonicRegressionWrapper.scala
@@ -99,9 +99,7 @@ private[r] object IsotonicRegressionWrapper
       val rMetadata = ("class" -> instance.getClass.getName) ~
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -114,8 +112,7 @@ private[r] object IsotonicRegressionWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
 

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/KMeansWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/KMeansWrapper.scala
@@ -123,9 +123,7 @@ private[r] object KMeansWrapper extends MLReadable[KMeansWrapper] {
         ("size" -> instance.size.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
 
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
       instance.pipeline.save(pipelinePath)
     }
   }
@@ -138,8 +136,7 @@ private[r] object KMeansWrapper extends MLReadable[KMeansWrapper] {
       val pipelinePath = new Path(path, "pipeline").toString
       val pipeline = PipelineModel.load(pipelinePath)
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
       val size = (rMetadata \ "size").extract[Array[Long]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/LDAWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/LDAWrapper.scala
@@ -198,9 +198,7 @@ private[r] object LDAWrapper extends MLReadable[LDAWrapper] {
         ("logPerplexity" -> instance.logPerplexity) ~
         ("vocabulary" -> instance.vocabulary.toList)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -213,8 +211,7 @@ private[r] object LDAWrapper extends MLReadable[LDAWrapper] {
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val logLikelihood = (rMetadata \ "logLikelihood").extract[Double]
       val logPerplexity = (rMetadata \ "logPerplexity").extract[Double]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/LinearRegressionWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/LinearRegressionWrapper.scala
@@ -127,9 +127,7 @@ private[r] object LinearRegressionWrapper
       val rMetadata = ("class" -> instance.getClass.getName) ~
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -142,8 +140,7 @@ private[r] object LinearRegressionWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
 

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/LinearSVCWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/LinearSVCWrapper.scala
@@ -137,9 +137,7 @@ private[r] object LinearSVCWrapper
         ("features" -> instance.features.toImmutableArraySeq) ~
         ("labels" -> instance.labels.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -152,8 +150,7 @@ private[r] object LinearSVCWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
       val labels = (rMetadata \ "labels").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/LogisticRegressionWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/LogisticRegressionWrapper.scala
@@ -192,9 +192,7 @@ private[r] object LogisticRegressionWrapper
         ("features" -> instance.features.toImmutableArraySeq) ~
         ("labels" -> instance.labels.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -207,8 +205,7 @@ private[r] object LogisticRegressionWrapper
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val features = (rMetadata \ "features").extract[Array[String]]
       val labels = (rMetadata \ "labels").extract[Array[String]]

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/MultilayerPerceptronClassifierWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/MultilayerPerceptronClassifierWrapper.scala
@@ -142,9 +142,7 @@ private[r] object MultilayerPerceptronClassifierWrapper
 
       val rMetadata = "class" -> instance.getClass.getName
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }

diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/NaiveBayesWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/NaiveBayesWrapper.scala
@@ -102,9 +102,7 @@ private[r] object NaiveBayesWrapper extends MLReadable[NaiveBayesWrapper] {
         ("labels" -> instance.labels.toImmutableArraySeq) ~
         ("features" -> instance.features.toImmutableArraySeq)
       val rMetadataJson: String = compact(render(rMetadata))
-      sparkSession.createDataFrame(
-        Seq(Tuple1(rMetadataJson))
-      ).repartition(1).write.text(rMetadataPath)
+      sc.parallelize(Seq(rMetadataJson), 1).saveAsTextFile(rMetadataPath)
 
       instance.pipeline.save(pipelinePath)
     }
@@ -117,8 +115,7 @@ private[r] object NaiveBayesWrapper extends MLReadable[NaiveBayesWrapper] {
       val rMetadataPath = new Path(path, "rMetadata").toString
       val pipelinePath = new Path(path, "pipeline").toString
 
-      val rMetadataStr = sparkSession.read.text(rMetadataPath)
-        .first().getString(0)
+      val rMetadataStr = sc.textFile(rMetadataPath, 1).first()
       val rMetadata = parse(rMetadataStr)
       val labels = (rMetadata \ "labels").extract[Array[String]]
       val features = (rMetadata \ "features").extract[Array[String]]