[SPARK-11766][MLLIB] add toJson/fromJson to Vector/Vectors

This is to support JSON serialization of Param[Vector] in the pipeline API. It could be used for other purposes too. The schema is the same as `VectorUDT`. jkbradley

Author: Xiangrui Meng <meng@databricks.com>

Closes #9751 from mengxr/SPARK-11766.

1 files changed, 17 insertions, 0 deletions

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 6508ddeba4..f895e2a8e4 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
@@ -20,6 +20,7 @@ package org.apache.spark.mllib.linalg
 import scala.util.Random
 
 import breeze.linalg.{DenseMatrix => BDM, squaredDistance => breezeSquaredDistance}
+import org.json4s.jackson.JsonMethods.{parse => parseJson}
 
 import org.apache.spark.{Logging, SparkException, SparkFunSuite}
 import org.apache.spark.mllib.util.TestingUtils._
@@ -374,4 +375,20 @@ class VectorsSuite extends SparkFunSuite with Logging {
     assert(v.slice(Array(2, 0)) === new SparseVector(2, Array(0), Array(2.2)))
     assert(v.slice(Array(2, 0, 3, 4)) === new SparseVector(4, Array(0, 3), Array(2.2, 4.4)))
   }
+
+  test("toJson/fromJson") {
+    val sv0 = Vectors.sparse(0, Array.empty, Array.empty)
+    val sv1 = Vectors.sparse(1, Array.empty, Array.empty)
+    val sv2 = Vectors.sparse(2, Array(1), Array(2.0))
+    val dv0 = Vectors.dense(Array.empty[Double])
+    val dv1 = Vectors.dense(1.0)
+    val dv2 = Vectors.dense(0.0, 2.0)
+    for (v <- Seq(sv0, sv1, sv2, dv0, dv1, dv2)) {
+      val json = v.toJson
+      parseJson(json) // `json` should be a valid JSON string
+      val u = Vectors.fromJson(json)
+      assert(u.getClass === v.getClass, "toJson/fromJson should preserve vector types.")
+      assert(u === v, "toJson/fromJson should preserve vector values.")
+    }
+  }
 }