diff options
Diffstat (limited to 'mllib')
-rw-r--r-- | mllib/pom.xml | 5 | ||||
-rw-r--r-- | mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala | 69 | ||||
-rw-r--r-- | mllib/src/test/scala/org/apache/spark/mllib/linalg/VectorsSuite.scala | 11 |
3 files changed, 83 insertions, 2 deletions
diff --git a/mllib/pom.xml b/mllib/pom.xml index fb7239e779..87a7ddaba9 100644 --- a/mllib/pom.xml +++ b/mllib/pom.xml @@ -46,6 +46,11 @@ <version>${project.version}</version> </dependency> <dependency> + <groupId>org.apache.spark</groupId> + <artifactId>spark-sql_${scala.binary.version}</artifactId> + <version>${project.version}</version> + </dependency> + <dependency> <groupId>org.eclipse.jetty</groupId> <artifactId>jetty-server</artifactId> </dependency> diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala index 6af225b7f4..ac217edc61 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala @@ -17,22 +17,26 @@ package org.apache.spark.mllib.linalg -import java.lang.{Double => JavaDouble, Integer => JavaInteger, Iterable => JavaIterable} import java.util +import java.lang.{Double => JavaDouble, Integer => JavaInteger, Iterable => JavaIterable} import scala.annotation.varargs import scala.collection.JavaConverters._ import breeze.linalg.{DenseVector => BDV, SparseVector => BSV, Vector => BV} -import org.apache.spark.mllib.util.NumericParser import org.apache.spark.SparkException +import org.apache.spark.mllib.util.NumericParser +import org.apache.spark.sql.catalyst.annotation.SQLUserDefinedType +import org.apache.spark.sql.catalyst.expressions.{GenericMutableRow, Row} +import org.apache.spark.sql.catalyst.types._ /** * Represents a numeric vector, whose index type is Int and value type is Double. * * Note: Users should not implement this interface. */ +@SQLUserDefinedType(udt = classOf[VectorUDT]) sealed trait Vector extends Serializable { /** @@ -75,6 +79,65 @@ sealed trait Vector extends Serializable { } /** + * User-defined type for [[Vector]] which allows easy interaction with SQL + * via [[org.apache.spark.sql.SchemaRDD]]. + */ +private[spark] class VectorUDT extends UserDefinedType[Vector] { + + override def sqlType: StructType = { + // type: 0 = sparse, 1 = dense + // We only use "values" for dense vectors, and "size", "indices", and "values" for sparse + // vectors. The "values" field is nullable because we might want to add binary vectors later, + // which uses "size" and "indices", but not "values". + StructType(Seq( + StructField("type", ByteType, nullable = false), + StructField("size", IntegerType, nullable = true), + StructField("indices", ArrayType(IntegerType, containsNull = false), nullable = true), + StructField("values", ArrayType(DoubleType, containsNull = false), nullable = true))) + } + + override def serialize(obj: Any): Row = { + val row = new GenericMutableRow(4) + obj match { + case sv: SparseVector => + row.setByte(0, 0) + row.setInt(1, sv.size) + row.update(2, sv.indices.toSeq) + row.update(3, sv.values.toSeq) + case dv: DenseVector => + row.setByte(0, 1) + row.setNullAt(1) + row.setNullAt(2) + row.update(3, dv.values.toSeq) + } + row + } + + override def deserialize(datum: Any): Vector = { + datum match { + case row: Row => + require(row.length == 4, + s"VectorUDT.deserialize given row with length ${row.length} but requires length == 4") + val tpe = row.getByte(0) + tpe match { + case 0 => + val size = row.getInt(1) + val indices = row.getAs[Iterable[Int]](2).toArray + val values = row.getAs[Iterable[Double]](3).toArray + new SparseVector(size, indices, values) + case 1 => + val values = row.getAs[Iterable[Double]](3).toArray + new DenseVector(values) + } + } + } + + override def pyUDT: String = "pyspark.mllib.linalg.VectorUDT" + + override def userClass: Class[Vector] = classOf[Vector] +} + +/** * Factory methods for [[org.apache.spark.mllib.linalg.Vector]]. * We don't use the name `Vector` because Scala imports * [[scala.collection.immutable.Vector]] by default. @@ -191,6 +254,7 @@ object Vectors { /** * A dense vector represented by a value array. */ +@SQLUserDefinedType(udt = classOf[VectorUDT]) class DenseVector(val values: Array[Double]) extends Vector { override def size: Int = values.length @@ -215,6 +279,7 @@ class DenseVector(val values: Array[Double]) extends Vector { * @param indices index array, assume to be strictly increasing. * @param values value array, must have the same length as the index array. */ +@SQLUserDefinedType(udt = classOf[VectorUDT]) class SparseVector( override val size: Int, val indices: Array[Int], diff --git a/mllib/src/test/scala/org/apache/spark/mllib/linalg/VectorsSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/linalg/VectorsSuite.scala index cd651fe2d2..93a84fe07b 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/linalg/VectorsSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/linalg/VectorsSuite.scala @@ -155,4 +155,15 @@ class VectorsSuite extends FunSuite { throw new RuntimeException(s"copy returned ${dvCopy.getClass} on ${dv.getClass}.") } } + + test("VectorUDT") { + val dv0 = Vectors.dense(Array.empty[Double]) + val dv1 = Vectors.dense(1.0, 2.0) + val sv0 = Vectors.sparse(2, Array.empty, Array.empty) + val sv1 = Vectors.sparse(2, Array(1), Array(2.0)) + val udt = new VectorUDT() + for (v <- Seq(dv0, dv1, sv0, sv1)) { + assert(v === udt.deserialize(udt.serialize(v))) + } + } } |