From 3893e8c576cf1a6decc18701267ce7cd8caaf521 Mon Sep 17 00:00:00 2001
From: Sean Owen <sowen@cloudera.com>
Date: Thu, 1 Sep 2016 12:13:07 -0700
Subject: [SPARK-17331][CORE][MLLIB] Avoid allocating 0-length arrays

## What changes were proposed in this pull request?

Avoid allocating some 0-length arrays, esp. in UTF8String, and by using Array.empty in Scala over Array[T]()

## How was this patch tested?

Jenkins

Author: Sean Owen <sowen@cloudera.com>

Closes #14895 from srowen/SPARK-17331.
---
 .../spark/ml/classification/MultinomialLogisticRegression.scala     | 2 +-
 .../src/main/scala/org/apache/spark/ml/feature/OneHotEncoder.scala  | 4 ++--
 mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala   | 6 +++---
 .../src/main/scala/org/apache/spark/mllib/stat/test/ChiSqTest.scala | 2 +-
 4 files changed, 7 insertions(+), 7 deletions(-)

(limited to 'mllib')

diff --git a/mllib/src/main/scala/org/apache/spark/ml/classification/MultinomialLogisticRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/classification/MultinomialLogisticRegression.scala
index f85ac76a8d..006f57c0ce 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/classification/MultinomialLogisticRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/classification/MultinomialLogisticRegression.scala
@@ -363,7 +363,7 @@ class MultinomialLogisticRegression @Since("2.1.0") (
             rawCoefficients(coefIndex)
           }
         } else {
-          Array[Double]()
+          Array.empty
         }
 
         val coefficientArray: Array[Double] = Array.tabulate(numClasses * numFeatures) { i =>
diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/OneHotEncoder.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/OneHotEncoder.scala
index 8b04b5de6f..e8e28ba29c 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/OneHotEncoder.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/OneHotEncoder.scala
@@ -164,8 +164,8 @@ class OneHotEncoder @Since("1.4.0") (@Since("1.4.0") override val uid: String) e
     // data transformation
     val size = outputAttrGroup.size
     val oneValue = Array(1.0)
-    val emptyValues = Array[Double]()
-    val emptyIndices = Array[Int]()
+    val emptyValues = Array.empty[Double]
+    val emptyIndices = Array.empty[Int]
     val encode = udf { label: Double =>
       if (label < size) {
         Vectors.sparse(size, Array(label.toInt), oneValue)
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala
index 4c39cf17f4..ad882c969a 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala
@@ -842,7 +842,7 @@ object SparseMatrix {
       "The expected number of nonzeros cannot be greater than Int.MaxValue.")
     val nnz = math.ceil(expected).toInt
     if (density == 0.0) {
-      new SparseMatrix(numRows, numCols, new Array[Int](numCols + 1), Array[Int](), Array[Double]())
+      new SparseMatrix(numRows, numCols, new Array[Int](numCols + 1), Array.empty, Array.empty)
     } else if (density == 1.0) {
       val colPtrs = Array.tabulate(numCols + 1)(j => j * numRows)
       val rowIndices = Array.tabulate(size.toInt)(idx => idx % numRows)
@@ -1098,7 +1098,7 @@ object Matrices {
   @Since("1.3.0")
   def horzcat(matrices: Array[Matrix]): Matrix = {
     if (matrices.isEmpty) {
-      return new DenseMatrix(0, 0, Array[Double]())
+      return new DenseMatrix(0, 0, Array.empty)
     } else if (matrices.length == 1) {
       return matrices(0)
     }
@@ -1157,7 +1157,7 @@ object Matrices {
   @Since("1.3.0")
   def vertcat(matrices: Array[Matrix]): Matrix = {
     if (matrices.isEmpty) {
-      return new DenseMatrix(0, 0, Array[Double]())
+      return new DenseMatrix(0, 0, Array.empty[Double])
     } else if (matrices.length == 1) {
       return matrices(0)
     }
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/stat/test/ChiSqTest.scala b/mllib/src/main/scala/org/apache/spark/mllib/stat/test/ChiSqTest.scala
index da5df9bf45..9a63b8a5d6 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/stat/test/ChiSqTest.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/stat/test/ChiSqTest.scala
@@ -146,7 +146,7 @@ private[stat] object ChiSqTest extends Logging {
    * Uniform distribution is assumed when `expected` is not passed in.
    */
   def chiSquared(observed: Vector,
-      expected: Vector = Vectors.dense(Array[Double]()),
+      expected: Vector = Vectors.dense(Array.empty[Double]),
       methodName: String = PEARSON.name): ChiSqTestResult = {
 
     // Validate input arguments
-- 
cgit v1.2.3