From be53a78352ae7c70d8a07d0df24574b3e3129b4a Mon Sep 17 00:00:00 2001 From: Jon McLean Date: Tue, 9 May 2017 09:47:50 +0100 Subject: [PATCH] [SPARK-20615][ML][TEST] SparseVector.argmax throws IndexOutOfBoundsException ## What changes were proposed in this pull request? Added a check for for the number of defined values. Previously the argmax function assumed that at least one value was defined if the vector size was greater than zero. ## How was this patch tested? Tests were added to the existing VectorsSuite to cover this case. Author: Jon McLean Closes #17877 from jonmclean/vectorArgmaxIndexBug. --- .../main/scala/org/apache/spark/ml/linalg/Vectors.scala | 2 ++ .../scala/org/apache/spark/ml/linalg/VectorsSuite.scala | 7 +++++++ .../main/scala/org/apache/spark/mllib/linalg/Vectors.scala | 2 ++ .../scala/org/apache/spark/mllib/linalg/VectorsSuite.scala | 7 +++++++ 4 files changed, 18 insertions(+) diff --git a/mllib-local/src/main/scala/org/apache/spark/ml/linalg/Vectors.scala b/mllib-local/src/main/scala/org/apache/spark/ml/linalg/Vectors.scala index 8e166ba0ff51a..3fbc0958a0f11 100644 --- a/mllib-local/src/main/scala/org/apache/spark/ml/linalg/Vectors.scala +++ b/mllib-local/src/main/scala/org/apache/spark/ml/linalg/Vectors.scala @@ -657,6 +657,8 @@ class SparseVector @Since("2.0.0") ( override def argmax: Int = { if (size == 0) { -1 + } else if (numActives == 0) { + 0 } else { // Find the max active entry. var maxIdx = indices(0) diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/linalg/VectorsSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/VectorsSuite.scala index dfbdaf19d374b..4cd91afd6d7fc 100644 --- a/mllib-local/src/test/scala/org/apache/spark/ml/linalg/VectorsSuite.scala +++ b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/VectorsSuite.scala @@ -125,6 +125,13 @@ class VectorsSuite extends SparkMLFunSuite { val vec8 = Vectors.sparse(5, Array(1, 2), Array(0.0, -1.0)) assert(vec8.argmax === 0) + + // Check for case when sparse vector is non-empty but the values are empty + val vec9 = Vectors.sparse(100, Array.empty[Int], Array.empty[Double]).asInstanceOf[SparseVector] + assert(vec9.argmax === 0) + + val vec10 = Vectors.sparse(1, Array.empty[Int], Array.empty[Double]).asInstanceOf[SparseVector] + assert(vec10.argmax === 0) } test("vector equals") { 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 723addc7150dd..f063420bec143 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 @@ -846,6 +846,8 @@ class SparseVector @Since("1.0.0") ( override def argmax: Int = { if (size == 0) { -1 + } else if (numActives == 0) { + 0 } else { // Find the max active entry. var maxIdx = indices(0) 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 71a3ceac1b947..6172cffee861c 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 @@ -122,6 +122,13 @@ class VectorsSuite extends SparkFunSuite with Logging { val vec8 = Vectors.sparse(5, Array(1, 2), Array(0.0, -1.0)) assert(vec8.argmax === 0) + + // Check for case when sparse vector is non-empty but the values are empty + val vec9 = Vectors.sparse(100, Array.empty[Int], Array.empty[Double]).asInstanceOf[SparseVector] + assert(vec9.argmax === 0) + + val vec10 = Vectors.sparse(1, Array.empty[Int], Array.empty[Double]).asInstanceOf[SparseVector] + assert(vec10.argmax === 0) } test("vector equals") {