databricks fat jar scala 2.13 compatible

.travis.yml

@@ -22,7 +22,7 @@ services:
 
 language: scala
 scala:
-  - 2.12.13
+  - 2.13.11
 jdk:
   - openjdk11
 

bundle-hdfs/src/main/scala/ml/bundle/hdfs/HadoopBundleFileSystem.scala

@@ -30,7 +30,7 @@ object HadoopBundleFileSystem {
   }
 
   def createSchemes(config: Config): Seq[String] = if (config.hasPath("schemes")) {
-    config.getStringList("schemes").asScala
+    config.getStringList("schemes").asScala.toSeq
   } else { Seq("hdfs") }
 }
 

bundle-ml/src/main/scala/ml/combust/bundle/json/JsonSupport.scala

@@ -133,7 +133,7 @@ trait JsonSupport {
       obj.ds.foreach(ds => fb += ("data_shape" -> ds.toJson))
       obj.m.foreach(m => fb += ("model" -> m.toJson))
 
-      JsObject(fb.result(): _*)
+      JsObject(fb.result().toSeq: _*)
     }
 
     override def read(json: JsValue): Scalar = json match {
@@ -220,7 +220,7 @@ trait JsonSupport {
       if(obj.ds.nonEmpty) { fb += ("data_shape" -> obj.ds.toJson) }
       if(obj.m.nonEmpty) { fb += ("model" -> obj.m.toJson) }
 
-      JsObject(fb.result(): _*)
+      JsObject(fb.result().toSeq: _*)
     }
 
     override def read(json: JsValue): List = json match {

bundle-ml/src/main/scala/ml/combust/bundle/serializer/ModelSerializer.scala

@@ -43,7 +43,7 @@ object FormatModelSerializer {
 
 /** Object for serializing/deserializing model definitions with JSON.
   */
-case class JsonFormatModelSerializer(implicit hr: HasBundleRegistry) extends FormatModelSerializer {
+case class JsonFormatModelSerializer()(implicit hr: HasBundleRegistry) extends FormatModelSerializer {
   override def write(path: Path, model: Model): Unit = {
     Files.write(path, model.asBundle.toJson.prettyPrint.getBytes("UTF-8"))
   }
@@ -55,7 +55,7 @@ case class JsonFormatModelSerializer(implicit hr: HasBundleRegistry) extends For
 
 /** Object for serializing/deserializing model definitions with Protobuf.
   */
-case class ProtoFormatModelSerializer(implicit hr: HasBundleRegistry) extends FormatModelSerializer {
+case class ProtoFormatModelSerializer()(implicit hr: HasBundleRegistry) extends FormatModelSerializer {
   override def write(path: Path, model: Model): Unit = {
     Files.write(path, model.asBundle.toByteArray)
   }

mleap-avro/src/main/scala/ml/combust/mleap/avro/DefaultFrameReader.scala

@@ -29,11 +29,11 @@ class DefaultFrameReader extends FrameReader {
     var rows = mutable.Seq[Row]()
     while(Try(reader.hasNext).getOrElse(false)) {
       record = reader.next(record)
-      val row = ArrayRow(new Array[Any](schema.fields.length))
+      val row = ArrayRow((new Array[Any](schema.fields.length)).toSeq)
       for(i <- schema.fields.indices) { row.set(i, readers(i)(record.get(i))) }
       rows :+= row
     }
 
-    DefaultLeapFrame(schema, rows)
+    DefaultLeapFrame(schema, rows.toSeq)
   }
 }

mleap-avro/src/main/scala/ml/combust/mleap/avro/DefaultRowReader.scala

@@ -26,7 +26,7 @@ class DefaultRowReader(override val schema: StructType) extends RowReader {
   override def fromBytes(bytes: Array[Byte], charset: Charset = BuiltinFormats.charset): Try[Row] = Try {
     decoder = DecoderFactory.get().binaryDecoder(bytes, decoder)
     record = datumReader.read(record, decoder)
-    val row = ArrayRow(new Array[Any](schema.fields.length))
+    val row = ArrayRow((new Array[Any](schema.fields.length)).toSeq)
     for(i <- schema.fields.indices) { row.set(i, readers(i)(record.get(i))) }
     row
   }

mleap-base/src/main/scala/ml/combust/mleap/ClassLoaderUtil.scala

@@ -8,15 +8,15 @@ import scala.util.control.NonFatal
 object ClassLoaderUtil {
   def findClassLoader(baseName: String): ClassLoader = {
     def findCaller(get: Int ⇒ Class[_]): ClassLoader =
-      Iterator.from(2 /*is the magic number, promise*/ ).map(get) dropWhile { c ⇒
+      Iterator.from(2 /*is the magic number, promise*/ ).map(get).dropWhile { c ⇒
         c != null &&
           (c.getName.startsWith(baseName) ||
             c.getName.startsWith("scala.Option") ||
             c.getName.startsWith("scala.collection.Iterator") ||
             c.getName.startsWith("ml.combust.bundle.util.ClassLoaderUtil"))
-      } next () match {
-        case null => getClass.getClassLoader
-        case c => c.getClassLoader
+      }.nextOption() match {
+        case Some(c) ⇒ c.getClassLoader
+        case None ⇒ getClass.getClassLoader
       }
 
     Option(Thread.currentThread.getContextClassLoader) orElse

mleap-core/src/main/scala/ml/combust/mleap/core/feature/VectorSlicerModel.scala

@@ -13,7 +13,7 @@ import org.apache.spark.ml.linalg.mleap.VectorUtil._
 case class VectorSlicerModel(indices: Array[Int],
                              namedIndices: Array[(String, Int)] = Array(),
                             inputSize: Int) extends Model {
-  val allIndices: Array[Int] = indices.union(namedIndices.map(_._2))
+  val allIndices: Array[Int] = indices.union(namedIndices.map(_._2)).toArray
 
   def apply(features: Vector): Vector = features match {
     case features: DenseVector => Vectors.dense(allIndices.map(features.apply))

mleap-core/src/main/scala/ml/combust/mleap/core/feature/WordToVectorModel.scala

@@ -68,7 +68,7 @@ case class WordToVectorModel(wordIndex: Map[String, Int],
     wordIndex.map { case (word, ind) =>
       (word, wordVectors.slice(vectorSize * ind, vectorSize * ind + vectorSize))
     }
-  }.mapValues(Vectors.dense).map(identity)
+  }.mapValues(Vectors.dense).map(identity).toMap
 
   def apply(sentence: Seq[String]): Vector = {
     if (sentence.isEmpty) {

mleap-runtime/src/main/java/ml/combust/mleap/runtime/javadsl/LeapFrameBuilderSupport.scala

@@ -35,7 +35,7 @@ class LeapFrameBuilderSupport {
   def createByteString(): BasicType = BasicType.ByteString
 
   def createTensorDimensions(dims : java.util.List[Integer]): Option[Seq[Int]] = {
-    Some(dims.asScala.map(_.intValue()))
+    Some(dims.asScala.toSeq.map(_.intValue()))
   }
 
   def createSchema(json: String): StructType = json.parseJson.convertTo[StructType]

mleap-runtime/src/main/java/ml/combust/mleap/runtime/javadsl/LeapFrameSupport.scala

@@ -12,11 +12,11 @@ class LeapFrameSupport {
   }
 
   def select(frame: DefaultLeapFrame, fieldNames: java.util.List[String]): DefaultLeapFrame = {
-    frame.select(fieldNames.asScala: _*).get
+    frame.select(fieldNames.asScala.toSeq: _*).get
   }
 
   def drop(frame: DefaultLeapFrame, names: java.util.List[String]): DefaultLeapFrame = {
-    frame.drop(names.asScala: _*).get
+    frame.drop(names.asScala.toSeq: _*).get
   }
 
   def getFields(schema: StructType): java.util.List[StructField]  = {

mleap-runtime/src/main/scala/ml/combust/mleap/binary/DefaultFrameReader.scala

@@ -41,7 +41,7 @@ class DefaultFrameReader extends FrameReader {
         rows(i) = row
       }
 
-      DefaultLeapFrame(schema, rows)
+      DefaultLeapFrame(schema, rows.toSeq)
     }).tried
   }
 }

mleap-runtime/src/main/scala/ml/combust/mleap/binary/DefaultRowReader.scala

@@ -19,7 +19,7 @@ class DefaultRowReader(override val schema: StructType) extends RowReader {
   override def fromBytes(bytes: Array[Byte], charset: Charset = BuiltinFormats.charset): Try[Row] = {
     (for(in <- managed(new ByteArrayInputStream(bytes))) yield {
       val din = new DataInputStream(in)
-      val row = ArrayRow(new Array[Any](schema.fields.length))
+      val row = ArrayRow((new Array[Any](schema.fields.length)).toSeq)
       var i = 0
       for(s <- serializers) {
         row.set(i, s.read(din))

mleap-runtime/src/main/scala/ml/combust/mleap/runtime/frame/ArrayRow.scala

@@ -24,7 +24,7 @@ case class ArrayRow(values: mutable.WrappedArray[Any]) extends Row {
   override def withValue(value: Any): ArrayRow = ArrayRow(values :+ value)
   override def withValues(values: Seq[Any]): ArrayRow = ArrayRow(this.values ++ values)
 
-  override def selectIndices(indices: Int *): ArrayRow = ArrayRow(indices.toArray.map(values))
+  override def selectIndices(indices: Int *): ArrayRow = ArrayRow(indices.toArray.map(values).toSeq)
 
   override def dropIndices(indices: Int *): ArrayRow = {
     val drops = Set(indices: _*)

mleap-runtime/src/main/scala/ml/combust/mleap/runtime/frame/Row.scala

@@ -16,7 +16,7 @@ object Row {
     * @param values values in the row
     * @return default row implementation with values
     */
-  def apply(values: Any *): Row = ArrayRow(values.toArray)
+  def apply(values: Any *): Row = ArrayRow(values.toArray.toSeq)
 }
 
 /** Base trait for row data.

mleap-runtime/src/main/scala/ml/combust/mleap/runtime/frame/RowTransformer.scala

@@ -169,7 +169,7 @@ case class RowTransformer private (inputSchema: StructType,
   def transformOption(row: Row): Option[ArrayRow] = {
     val arr = new Array[Any](maxSize)
     row.toArray.copyToArray(arr)
-    val arrRow = ArrayRow(arr)
+    val arrRow = ArrayRow(arr.toSeq)
 
     val r = transforms.foldLeft(Option(arrRow)) {
       (r, transform) => r.flatMap(transform)

mleap-spark-extension/src/main/scala/org/apache/spark/ml/mleap/classification/OneVsRest.scala

@@ -23,6 +23,7 @@ import java.util.UUID
 import ml.combust.mleap.core.annotation.SparkCode
 
 import scala.collection.JavaConverters._
+import scala.collection.parallel.CollectionConverters._
 import scala.language.existentials
 import org.apache.hadoop.fs.Path
 import org.json4s.{DefaultFormats, JObject, _}
@@ -355,7 +356,7 @@ final class OneVsRest @Since("1.4.0") (
     }
 
     // create k columns, one for each binary classifier.
-    val models = Range(0, numClasses).par.map { index =>
+    val models = (0 until numClasses).par.map { index =>
       // generate new label metadata for the binary problem.
       val newLabelMeta = BinaryAttribute.defaultAttr.withName("label").toMetadata()
       val labelColName = "mc2b$" + index

mleap-tensorflow/src/main/scala/ml/combust/mleap/tensorflow/TensorflowModel.scala

@@ -13,6 +13,7 @@ import java.nio.file.Files
 import java.util.zip.ZipInputStream
 import scala.collection.JavaConverters._
 import scala.collection.mutable
+import scala.reflect.ClassTag
 import scala.util.Try
 
 /**
@@ -28,7 +29,7 @@ case class TensorflowModel( @transient var graph: Option[tensorflow.Graph] = Non
                           ) extends Model with AutoCloseable {
 
   def apply(values: Tensor[_] *): Seq[Any] = {
-    val garbage: mutable.ArrayBuilder[tensorflow.Tensor] = mutable.ArrayBuilder.make[tensorflow.Tensor]()
+    val garbage: mutable.ArrayBuilder[tensorflow.Tensor] = mutable.ArrayBuilder.make[tensorflow.Tensor](ClassTag(classOf[Tensor[_]]))
 
     val result = Try {
       val tensors = values.zip(inputs).map {

project/Common.scala

@@ -16,7 +16,8 @@ object Common {
   lazy val defaultSettings = buildSettings ++ sonatypeSettings
 
   lazy val buildSettings: Seq[Def.Setting[_]] = Seq(
-    scalaVersion := "2.12.13",
+    scalaVersion := "2.13.11",
+    crossScalaVersions := Seq("2.12.13", "2.13.11"),
     scalacOptions ++= Seq("-unchecked", "-deprecation", "-feature"),
     fork in Test := true,
     javaOptions in test += sys.env.getOrElse("JVM_OPTS", ""),
@@ -32,7 +33,8 @@ object Common {
       } else {
         Seq()
       }
-    }
+    },
+    resolvers += "XGBoost4J Snapshot Repo" at "https://s3-us-west-2.amazonaws.com/xgboost-maven-repo/snapshot/"
   )
 
   lazy val mleapSettings: Seq[Def.Setting[_]] = Seq(organization := "ml.combust.mleap")

project/Dependencies.scala

@@ -17,7 +17,7 @@ object Dependencies {
   lazy val slf4jVersion = "2.0.6"
   lazy val awsSdkVersion = "1.11.1033"
   val tensorflowJavaVersion = "0.5.0" // Match Tensorflow 2.10.1 https://github.com/tensorflow/java/#tensorflow-version-support
-  val xgboostVersion = "1.7.3"
+  val xgboostVersion = "2.0.0-SNAPSHOT"
   val breezeVersion = "2.1.0"
   val hadoopVersion = "3.3.4" // matches spark version
   val platforms = "windows-x86_64,linux-x86_64,macosx-x86_64"
@@ -33,8 +33,9 @@ object Dependencies {
       "org.apache.spark" %% "spark-avro" % sparkVersion
     )
     val avroDep = "org.apache.avro" % "avro" % "1.11.1"
-    val sprayJson = "io.spray" %% "spray-json" % "1.3.2"
-    val arm = "com.jsuereth" %% "scala-arm" % "2.0"
+    val sprayJson = "io.spray" %% "spray-json" % "1.3.6"
+    // https://github.com/jsuereth/scala-arm/issues/79
+    val arm = "com.michaelpollmeier" %% "scala-arm" % "2.1"
     val config = "com.typesafe" % "config" % "1.3.0"
     val scalaReflect = ScalaVersionDependentModuleID.versioned("org.scala-lang" % "scala-reflect" % _)
     val scalaTest = "org.scalatest" %% "scalatest" % scalaTestVersion
@@ -72,8 +73,8 @@ object Dependencies {
 
     val breeze = "org.scalanlp" %% "breeze" % breezeVersion
 
-    val xgboostDep = "ml.dmlc" %% "xgboost4j" % xgboostVersion
-    val xgboostSparkDep = "ml.dmlc" %% "xgboost4j-spark" % xgboostVersion
+    val xgboostDep = "ml.dmlc" %% "xgboost4j" % xgboostVersion exclude("org.scala-lang.modules", "scala-collection-compat_2.12")
+    val xgboostSparkDep = "ml.dmlc" %% "xgboost4j-spark" % xgboostVersion exclude("org.scala-lang.modules", "scala-collection-compat_2.12") exclude("ml.dmlc", "xgboost4j_2.12")
     val xgboostPredictorDep = "ai.h2o" % "xgboost-predictor" % "0.3.18" exclude("com.esotericsoftware.kryo", "kryo")
 
     val hadoop = "org.apache.hadoop" % "hadoop-client" % hadoopVersion
@@ -127,7 +128,7 @@ object Dependencies {
 
   val xgboostRuntime = l ++= Seq(xgboostDep) ++ Seq(xgboostPredictorDep) ++ Test.spark ++ Test.sparkTest ++ Seq(Test.scalaTest)
 
-  val xgboostSpark = l ++= Seq(xgboostSparkDep) ++ Provided.spark ++ Test.spark ++ Test.sparkTest
+  val xgboostSpark = l ++= Seq(xgboostDep) ++ Seq(xgboostSparkDep) ++ Provided.spark ++ Test.spark ++ Test.sparkTest
 
   val serving = l ++= Seq(akkaHttp, akkaHttpSprayJson, config, Test.scalaTest, Test.akkaHttpTestkit)