diff options
author | DB Tsai <dbt@netflix.com> | 2016-04-15 01:17:03 -0700 |
---|---|---|
committer | Xiangrui Meng <meng@databricks.com> | 2016-04-15 01:17:03 -0700 |
commit | 96534aa47c39e0ec40bc38be566455d11e21adb2 (patch) | |
tree | 3702b1573b3f6fb5b0bd682e7c70e55ab1868b51 /mllib-local/src/test/scala/org | |
parent | a9324a06ef0e3646410dc9b3d4f21d66b9064303 (diff) | |
download | spark-96534aa47c39e0ec40bc38be566455d11e21adb2.tar.gz spark-96534aa47c39e0ec40bc38be566455d11e21adb2.tar.bz2 spark-96534aa47c39e0ec40bc38be566455d11e21adb2.zip |
[SPARK-14549][ML] Copy the Vector and Matrix classes from mllib to ml in mllib-local
## What changes were proposed in this pull request?
This task will copy the Vector and Matrix classes from mllib to ml package in mllib-local jar. The UDTs and `since` annotation in ml vector and matrix will be removed from now. UDTs will be achieved by #SPARK-14487, and `since` will be replaced by /* since 1.2.0 */
The BLAS implementation will be copied, and some of the test utilities will be copies as well.
Summary of changes:
1. In mllib-local/src/main/scala/org/apache/spark/**ml**/linalg/BLAS.scala
- Copied from mllib/src/main/scala/org/apache/spark/**mllib**/linalg/BLAS.scala
- logDebug("gemm: alpha is equal to 0 and beta is equal to 1. Returning C.") is removed in ml version.
2. In mllib-local/src/main/scala/org/apache/spark/**ml**/linalg/Matrices.scala
- Copied from mllib/src/main/scala/org/apache/spark/**mllib**/linalg/Matrices.scala
- `Since` was removed, and we'll use standard `/* Since /*` Java doc. Will be in another PR.
- `UDT` related code was removed, and will use `SPARK-13944` https://github.com/apache/spark/pull/12259 to replace the annotation.
3. In mllib-local/src/main/scala/org/apache/spark/**ml**/linalg/Vectors.scala
- Copied from mllib/src/main/scala/org/apache/spark/**mllib**/linalg/Vectors.scala
- `Since` was removed.
- `UDT` related code was removed.
- In `def parseNumeric`, it was throwing `throw new SparkException(s"Cannot parse $other.")`, and now it's throwing `throw new IllegalArgumentException(s"Cannot parse $other.")`
4. In mllib/src/main/scala/org/apache/spark/**mllib**/linalg/Vectors.scala
- For consistency with ML version of vector, `def parseNumeric` is now throwing `throw new IllegalArgumentException(s"Cannot parse $other.")`
5. mllib/src/main/scala/org/apache/spark/**mllib**/util/NumericParser.scala is moved to mllib-local/src/main/scala/org/apache/spark/**ml**/util/NumericParser.scala
- All the `throw new SparkException` were replaced by `throw new IllegalArgumentException`
## How was this patch tested?
unit tests
Author: DB Tsai <dbt@netflix.com>
Closes #12317 from dbtsai/dbtsai-ml-vector.
Diffstat (limited to 'mllib-local/src/test/scala/org')
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/SparkMLFunSuite.scala (renamed from mllib-local/src/test/scala/org/apache/spark/ml/DummyTestingSuite.scala) | 16 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/linalg/BLASSuite.scala | 408 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeMatrixConversionSuite.scala | 71 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeVectorConversionSuite.scala | 67 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/linalg/MatricesSuite.scala | 511 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/linalg/VectorsSuite.scala | 358 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtils.scala | 236 | ||||
-rw-r--r-- | mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtilsSuite.scala | 187 |
8 files changed, 1847 insertions, 7 deletions
diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/DummyTestingSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/SparkMLFunSuite.scala index 51b7c2409f..cb3b56bba8 100644 --- a/mllib-local/src/test/scala/org/apache/spark/ml/DummyTestingSuite.scala +++ b/mllib-local/src/test/scala/org/apache/spark/ml/SparkMLFunSuite.scala @@ -17,12 +17,14 @@ package org.apache.spark.ml -import org.scalatest.FunSuite // scalastyle:ignore funsuite +// scalastyle:off +import org.scalatest.{BeforeAndAfterAll, FunSuite} -// This is testing if the new build works. To be removed soon. -class DummyTestingSuite extends FunSuite { // scalastyle:ignore funsuite - - test("This is testing if the new build works.") { - assert(DummyTesting.add10(15) === 25) - } +/** + * Base abstract class for all unit tests in Spark for handling common functionality. + */ +private[spark] abstract class SparkMLFunSuite + extends FunSuite + with BeforeAndAfterAll { + // scalastyle:on } diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BLASSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BLASSuite.scala new file mode 100644 index 0000000000..8a9f49792c --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BLASSuite.scala @@ -0,0 +1,408 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.linalg + +import org.apache.spark.ml.SparkMLFunSuite +import org.apache.spark.ml.linalg.BLAS._ +import org.apache.spark.ml.util.TestingUtils._ + +class BLASSuite extends SparkMLFunSuite { + + test("copy") { + val sx = Vectors.sparse(4, Array(0, 2), Array(1.0, -2.0)) + val dx = Vectors.dense(1.0, 0.0, -2.0, 0.0) + val sy = Vectors.sparse(4, Array(0, 1, 3), Array(2.0, 1.0, 1.0)) + val dy = Array(2.0, 1.0, 0.0, 1.0) + + val dy1 = Vectors.dense(dy.clone()) + copy(sx, dy1) + assert(dy1 ~== dx absTol 1e-15) + + val dy2 = Vectors.dense(dy.clone()) + copy(dx, dy2) + assert(dy2 ~== dx absTol 1e-15) + + intercept[IllegalArgumentException] { + copy(sx, sy) + } + + intercept[IllegalArgumentException] { + copy(dx, sy) + } + + withClue("vector sizes must match") { + intercept[Exception] { + copy(sx, Vectors.dense(0.0, 1.0, 2.0)) + } + } + } + + test("scal") { + val a = 0.1 + val sx = Vectors.sparse(3, Array(0, 2), Array(1.0, -2.0)) + val dx = Vectors.dense(1.0, 0.0, -2.0) + + scal(a, sx) + assert(sx ~== Vectors.sparse(3, Array(0, 2), Array(0.1, -0.2)) absTol 1e-15) + + scal(a, dx) + assert(dx ~== Vectors.dense(0.1, 0.0, -0.2) absTol 1e-15) + } + + test("axpy") { + val alpha = 0.1 + val sx = Vectors.sparse(3, Array(0, 2), Array(1.0, -2.0)) + val dx = Vectors.dense(1.0, 0.0, -2.0) + val dy = Array(2.0, 1.0, 0.0) + val expected = Vectors.dense(2.1, 1.0, -0.2) + + val dy1 = Vectors.dense(dy.clone()) + axpy(alpha, sx, dy1) + assert(dy1 ~== expected absTol 1e-15) + + val dy2 = Vectors.dense(dy.clone()) + axpy(alpha, dx, dy2) + assert(dy2 ~== expected absTol 1e-15) + + val sy = Vectors.sparse(4, Array(0, 1), Array(2.0, 1.0)) + + intercept[IllegalArgumentException] { + axpy(alpha, sx, sy) + } + + intercept[IllegalArgumentException] { + axpy(alpha, dx, sy) + } + + withClue("vector sizes must match") { + intercept[Exception] { + axpy(alpha, sx, Vectors.dense(1.0, 2.0)) + } + } + } + + test("dot") { + val sx = Vectors.sparse(3, Array(0, 2), Array(1.0, -2.0)) + val dx = Vectors.dense(1.0, 0.0, -2.0) + val sy = Vectors.sparse(3, Array(0, 1), Array(2.0, 1.0)) + val dy = Vectors.dense(2.0, 1.0, 0.0) + + assert(dot(sx, sy) ~== 2.0 absTol 1e-15) + assert(dot(sy, sx) ~== 2.0 absTol 1e-15) + assert(dot(sx, dy) ~== 2.0 absTol 1e-15) + assert(dot(dy, sx) ~== 2.0 absTol 1e-15) + assert(dot(dx, dy) ~== 2.0 absTol 1e-15) + assert(dot(dy, dx) ~== 2.0 absTol 1e-15) + + assert(dot(sx, sx) ~== 5.0 absTol 1e-15) + assert(dot(dx, dx) ~== 5.0 absTol 1e-15) + assert(dot(sx, dx) ~== 5.0 absTol 1e-15) + assert(dot(dx, sx) ~== 5.0 absTol 1e-15) + + val sx1 = Vectors.sparse(10, Array(0, 3, 5, 7, 8), Array(1.0, 2.0, 3.0, 4.0, 5.0)) + val sx2 = Vectors.sparse(10, Array(1, 3, 6, 7, 9), Array(1.0, 2.0, 3.0, 4.0, 5.0)) + assert(dot(sx1, sx2) ~== 20.0 absTol 1e-15) + assert(dot(sx2, sx1) ~== 20.0 absTol 1e-15) + + withClue("vector sizes must match") { + intercept[Exception] { + dot(sx, Vectors.dense(2.0, 1.0)) + } + } + } + + test("spr") { + // test dense vector + val alpha = 0.1 + val x = new DenseVector(Array(1.0, 2, 2.1, 4)) + val U = new DenseVector(Array(1.0, 2, 2, 3, 3, 3, 4, 4, 4, 4)) + val expected = new DenseVector(Array(1.1, 2.2, 2.4, 3.21, 3.42, 3.441, 4.4, 4.8, 4.84, 5.6)) + + spr(alpha, x, U) + assert(U ~== expected absTol 1e-9) + + val matrix33 = new DenseVector(Array(1.0, 2, 3, 4, 5)) + withClue("Size of vector must match the rank of matrix") { + intercept[Exception] { + spr(alpha, x, matrix33) + } + } + + // test sparse vector + val sv = new SparseVector(4, Array(0, 3), Array(1.0, 2)) + val U2 = new DenseVector(Array(1.0, 2, 2, 3, 3, 3, 4, 4, 4, 4)) + spr(0.1, sv, U2) + val expectedSparse = new DenseVector(Array(1.1, 2.0, 2.0, 3.0, 3.0, 3.0, 4.2, 4.0, 4.0, 4.4)) + assert(U2 ~== expectedSparse absTol 1e-15) + } + + test("syr") { + val dA = new DenseMatrix(4, 4, + Array(0.0, 1.2, 2.2, 3.1, 1.2, 3.2, 5.3, 4.6, 2.2, 5.3, 1.8, 3.0, 3.1, 4.6, 3.0, 0.8)) + val x = new DenseVector(Array(0.0, 2.7, 3.5, 2.1)) + val alpha = 0.15 + + val expected = new DenseMatrix(4, 4, + Array(0.0, 1.2, 2.2, 3.1, 1.2, 4.2935, 6.7175, 5.4505, 2.2, 6.7175, 3.6375, 4.1025, 3.1, + 5.4505, 4.1025, 1.4615)) + + syr(alpha, x, dA) + + assert(dA ~== expected absTol 1e-15) + + val dB = + new DenseMatrix(3, 4, Array(0.0, 1.2, 2.2, 3.1, 1.2, 3.2, 5.3, 4.6, 2.2, 5.3, 1.8, 3.0)) + + withClue("Matrix A must be a symmetric Matrix") { + intercept[Exception] { + syr(alpha, x, dB) + } + } + + val dC = + new DenseMatrix(3, 3, Array(0.0, 1.2, 2.2, 1.2, 3.2, 5.3, 2.2, 5.3, 1.8)) + + withClue("Size of vector must match the rank of matrix") { + intercept[Exception] { + syr(alpha, x, dC) + } + } + + val y = new DenseVector(Array(0.0, 2.7, 3.5, 2.1, 1.5)) + + withClue("Size of vector must match the rank of matrix") { + intercept[Exception] { + syr(alpha, y, dA) + } + } + + val xSparse = new SparseVector(4, Array(0, 2, 3), Array(1.0, 3.0, 4.0)) + val dD = new DenseMatrix(4, 4, + Array(0.0, 1.2, 2.2, 3.1, 1.2, 3.2, 5.3, 4.6, 2.2, 5.3, 1.8, 3.0, 3.1, 4.6, 3.0, 0.8)) + syr(0.1, xSparse, dD) + val expectedSparse = new DenseMatrix(4, 4, + Array(0.1, 1.2, 2.5, 3.5, 1.2, 3.2, 5.3, 4.6, 2.5, 5.3, 2.7, 4.2, 3.5, 4.6, 4.2, 2.4)) + assert(dD ~== expectedSparse absTol 1e-15) + } + + test("gemm") { + val dA = + new DenseMatrix(4, 3, Array(0.0, 1.0, 0.0, 0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 3.0)) + val sA = new SparseMatrix(4, 3, Array(0, 1, 3, 4), Array(1, 0, 2, 3), Array(1.0, 2.0, 1.0, 3.0)) + + val B = new DenseMatrix(3, 2, Array(1.0, 0.0, 0.0, 0.0, 2.0, 1.0)) + val expected = new DenseMatrix(4, 2, Array(0.0, 1.0, 0.0, 0.0, 4.0, 0.0, 2.0, 3.0)) + val BTman = new DenseMatrix(2, 3, Array(1.0, 0.0, 0.0, 2.0, 0.0, 1.0)) + val BT = B.transpose + + assert(dA.multiply(B) ~== expected absTol 1e-15) + assert(sA.multiply(B) ~== expected absTol 1e-15) + + val C1 = new DenseMatrix(4, 2, Array(1.0, 0.0, 2.0, 1.0, 0.0, 0.0, 1.0, 0.0)) + val C2 = C1.copy + val C3 = C1.copy + val C4 = C1.copy + val C5 = C1.copy + val C6 = C1.copy + val C7 = C1.copy + val C8 = C1.copy + val C9 = C1.copy + val C10 = C1.copy + val C11 = C1.copy + val C12 = C1.copy + val C13 = C1.copy + val C14 = C1.copy + val C15 = C1.copy + val C16 = C1.copy + val C17 = C1.copy + val expected2 = new DenseMatrix(4, 2, Array(2.0, 1.0, 4.0, 2.0, 4.0, 0.0, 4.0, 3.0)) + val expected3 = new DenseMatrix(4, 2, Array(2.0, 2.0, 4.0, 2.0, 8.0, 0.0, 6.0, 6.0)) + val expected4 = new DenseMatrix(4, 2, Array(5.0, 0.0, 10.0, 5.0, 0.0, 0.0, 5.0, 0.0)) + val expected5 = C1.copy + + gemm(1.0, dA, B, 2.0, C1) + gemm(1.0, sA, B, 2.0, C2) + gemm(2.0, dA, B, 2.0, C3) + gemm(2.0, sA, B, 2.0, C4) + assert(C1 ~== expected2 absTol 1e-15) + assert(C2 ~== expected2 absTol 1e-15) + assert(C3 ~== expected3 absTol 1e-15) + assert(C4 ~== expected3 absTol 1e-15) + gemm(1.0, dA, B, 0.0, C17) + assert(C17 ~== expected absTol 1e-15) + gemm(1.0, sA, B, 0.0, C17) + assert(C17 ~== expected absTol 1e-15) + + withClue("columns of A don't match the rows of B") { + intercept[Exception] { + gemm(1.0, dA.transpose, B, 2.0, C1) + } + } + + val dATman = + new DenseMatrix(3, 4, Array(0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 3.0)) + val sATman = + new SparseMatrix(3, 4, Array(0, 1, 2, 3, 4), Array(1, 0, 1, 2), Array(2.0, 1.0, 1.0, 3.0)) + + val dATT = dATman.transpose + val sATT = sATman.transpose + val BTT = BTman.transpose.asInstanceOf[DenseMatrix] + + assert(dATT.multiply(B) ~== expected absTol 1e-15) + assert(sATT.multiply(B) ~== expected absTol 1e-15) + assert(dATT.multiply(BTT) ~== expected absTol 1e-15) + assert(sATT.multiply(BTT) ~== expected absTol 1e-15) + + gemm(1.0, dATT, BTT, 2.0, C5) + gemm(1.0, sATT, BTT, 2.0, C6) + gemm(2.0, dATT, BTT, 2.0, C7) + gemm(2.0, sATT, BTT, 2.0, C8) + gemm(1.0, dA, BTT, 2.0, C9) + gemm(1.0, sA, BTT, 2.0, C10) + gemm(2.0, dA, BTT, 2.0, C11) + gemm(2.0, sA, BTT, 2.0, C12) + assert(C5 ~== expected2 absTol 1e-15) + assert(C6 ~== expected2 absTol 1e-15) + assert(C7 ~== expected3 absTol 1e-15) + assert(C8 ~== expected3 absTol 1e-15) + assert(C9 ~== expected2 absTol 1e-15) + assert(C10 ~== expected2 absTol 1e-15) + assert(C11 ~== expected3 absTol 1e-15) + assert(C12 ~== expected3 absTol 1e-15) + + gemm(0, dA, B, 5, C13) + gemm(0, sA, B, 5, C14) + gemm(0, dA, B, 1, C15) + gemm(0, sA, B, 1, C16) + assert(C13 ~== expected4 absTol 1e-15) + assert(C14 ~== expected4 absTol 1e-15) + assert(C15 ~== expected5 absTol 1e-15) + assert(C16 ~== expected5 absTol 1e-15) + + } + + test("gemv") { + + val dA = + new DenseMatrix(4, 3, Array(0.0, 1.0, 0.0, 0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 3.0)) + val sA = new SparseMatrix(4, 3, Array(0, 1, 3, 4), Array(1, 0, 2, 3), Array(1.0, 2.0, 1.0, 3.0)) + + val dA2 = + new DenseMatrix(4, 3, Array(0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 3.0), true) + val sA2 = + new SparseMatrix(4, 3, Array(0, 1, 2, 3, 4), Array(1, 0, 1, 2), Array(2.0, 1.0, 1.0, 3.0), + true) + + val dx = new DenseVector(Array(1.0, 2.0, 3.0)) + val sx = dx.toSparse + val expected = new DenseVector(Array(4.0, 1.0, 2.0, 9.0)) + + assert(dA.multiply(dx) ~== expected absTol 1e-15) + assert(sA.multiply(dx) ~== expected absTol 1e-15) + assert(dA.multiply(sx) ~== expected absTol 1e-15) + assert(sA.multiply(sx) ~== expected absTol 1e-15) + + val y1 = new DenseVector(Array(1.0, 3.0, 1.0, 0.0)) + val y2 = y1.copy + val y3 = y1.copy + val y4 = y1.copy + val y5 = y1.copy + val y6 = y1.copy + val y7 = y1.copy + val y8 = y1.copy + val y9 = y1.copy + val y10 = y1.copy + val y11 = y1.copy + val y12 = y1.copy + val y13 = y1.copy + val y14 = y1.copy + val y15 = y1.copy + val y16 = y1.copy + + val expected2 = new DenseVector(Array(6.0, 7.0, 4.0, 9.0)) + val expected3 = new DenseVector(Array(10.0, 8.0, 6.0, 18.0)) + + gemv(1.0, dA, dx, 2.0, y1) + gemv(1.0, sA, dx, 2.0, y2) + gemv(1.0, dA, sx, 2.0, y3) + gemv(1.0, sA, sx, 2.0, y4) + + gemv(1.0, dA2, dx, 2.0, y5) + gemv(1.0, sA2, dx, 2.0, y6) + gemv(1.0, dA2, sx, 2.0, y7) + gemv(1.0, sA2, sx, 2.0, y8) + + gemv(2.0, dA, dx, 2.0, y9) + gemv(2.0, sA, dx, 2.0, y10) + gemv(2.0, dA, sx, 2.0, y11) + gemv(2.0, sA, sx, 2.0, y12) + + gemv(2.0, dA2, dx, 2.0, y13) + gemv(2.0, sA2, dx, 2.0, y14) + gemv(2.0, dA2, sx, 2.0, y15) + gemv(2.0, sA2, sx, 2.0, y16) + + assert(y1 ~== expected2 absTol 1e-15) + assert(y2 ~== expected2 absTol 1e-15) + assert(y3 ~== expected2 absTol 1e-15) + assert(y4 ~== expected2 absTol 1e-15) + + assert(y5 ~== expected2 absTol 1e-15) + assert(y6 ~== expected2 absTol 1e-15) + assert(y7 ~== expected2 absTol 1e-15) + assert(y8 ~== expected2 absTol 1e-15) + + assert(y9 ~== expected3 absTol 1e-15) + assert(y10 ~== expected3 absTol 1e-15) + assert(y11 ~== expected3 absTol 1e-15) + assert(y12 ~== expected3 absTol 1e-15) + + assert(y13 ~== expected3 absTol 1e-15) + assert(y14 ~== expected3 absTol 1e-15) + assert(y15 ~== expected3 absTol 1e-15) + assert(y16 ~== expected3 absTol 1e-15) + + withClue("columns of A don't match the rows of B") { + intercept[Exception] { + gemv(1.0, dA.transpose, dx, 2.0, y1) + } + intercept[Exception] { + gemv(1.0, sA.transpose, dx, 2.0, y1) + } + intercept[Exception] { + gemv(1.0, dA.transpose, sx, 2.0, y1) + } + intercept[Exception] { + gemv(1.0, sA.transpose, sx, 2.0, y1) + } + } + + val dAT = + new DenseMatrix(3, 4, Array(0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 3.0)) + val sAT = + new SparseMatrix(3, 4, Array(0, 1, 2, 3, 4), Array(1, 0, 1, 2), Array(2.0, 1.0, 1.0, 3.0)) + + val dATT = dAT.transpose + val sATT = sAT.transpose + + assert(dATT.multiply(dx) ~== expected absTol 1e-15) + assert(sATT.multiply(dx) ~== expected absTol 1e-15) + assert(dATT.multiply(sx) ~== expected absTol 1e-15) + assert(sATT.multiply(sx) ~== expected absTol 1e-15) + } +} diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeMatrixConversionSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeMatrixConversionSuite.scala new file mode 100644 index 0000000000..70a21e41bf --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeMatrixConversionSuite.scala @@ -0,0 +1,71 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.linalg + +import breeze.linalg.{CSCMatrix => BSM, DenseMatrix => BDM} + +import org.apache.spark.ml.SparkMLFunSuite + +class BreezeMatrixConversionSuite extends SparkMLFunSuite { + test("dense matrix to breeze") { + val mat = Matrices.dense(3, 2, Array(0.0, 1.0, 2.0, 3.0, 4.0, 5.0)) + val breeze = mat.toBreeze.asInstanceOf[BDM[Double]] + assert(breeze.rows === mat.numRows) + assert(breeze.cols === mat.numCols) + assert(breeze.data.eq(mat.asInstanceOf[DenseMatrix].values), "should not copy data") + } + + test("dense breeze matrix to matrix") { + val breeze = new BDM[Double](3, 2, Array(0.0, 1.0, 2.0, 3.0, 4.0, 5.0)) + val mat = Matrices.fromBreeze(breeze).asInstanceOf[DenseMatrix] + assert(mat.numRows === breeze.rows) + assert(mat.numCols === breeze.cols) + assert(mat.values.eq(breeze.data), "should not copy data") + // transposed matrix + val matTransposed = Matrices.fromBreeze(breeze.t).asInstanceOf[DenseMatrix] + assert(matTransposed.numRows === breeze.cols) + assert(matTransposed.numCols === breeze.rows) + assert(matTransposed.values.eq(breeze.data), "should not copy data") + } + + test("sparse matrix to breeze") { + val values = Array(1.0, 2.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(1, 2, 1, 2) + val mat = Matrices.sparse(3, 2, colPtrs, rowIndices, values) + val breeze = mat.toBreeze.asInstanceOf[BSM[Double]] + assert(breeze.rows === mat.numRows) + assert(breeze.cols === mat.numCols) + assert(breeze.data.eq(mat.asInstanceOf[SparseMatrix].values), "should not copy data") + } + + test("sparse breeze matrix to sparse matrix") { + val values = Array(1.0, 2.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(1, 2, 1, 2) + val breeze = new BSM[Double](values, 3, 2, colPtrs, rowIndices) + val mat = Matrices.fromBreeze(breeze).asInstanceOf[SparseMatrix] + assert(mat.numRows === breeze.rows) + assert(mat.numCols === breeze.cols) + assert(mat.values.eq(breeze.data), "should not copy data") + val matTransposed = Matrices.fromBreeze(breeze.t).asInstanceOf[SparseMatrix] + assert(matTransposed.numRows === breeze.cols) + assert(matTransposed.numCols === breeze.rows) + assert(!matTransposed.values.eq(breeze.data), "has to copy data") + } +} diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeVectorConversionSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeVectorConversionSuite.scala new file mode 100644 index 0000000000..00c9ee79eb --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BreezeVectorConversionSuite.scala @@ -0,0 +1,67 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.linalg + +import breeze.linalg.{DenseVector => BDV, SparseVector => BSV} + +import org.apache.spark.ml.SparkMLFunSuite + +/** + * Test Breeze vector conversions. + */ +class BreezeVectorConversionSuite extends SparkMLFunSuite { + + val arr = Array(0.1, 0.2, 0.3, 0.4) + val n = 20 + val indices = Array(0, 3, 5, 10, 13) + val values = Array(0.1, 0.5, 0.3, -0.8, -1.0) + + test("dense to breeze") { + val vec = Vectors.dense(arr) + assert(vec.toBreeze === new BDV[Double](arr)) + } + + test("sparse to breeze") { + val vec = Vectors.sparse(n, indices, values) + assert(vec.toBreeze === new BSV[Double](indices, values, n)) + } + + test("dense breeze to vector") { + val breeze = new BDV[Double](arr) + val vec = Vectors.fromBreeze(breeze).asInstanceOf[DenseVector] + assert(vec.size === arr.length) + assert(vec.values.eq(arr), "should not copy data") + } + + test("sparse breeze to vector") { + val breeze = new BSV[Double](indices, values, n) + val vec = Vectors.fromBreeze(breeze).asInstanceOf[SparseVector] + assert(vec.size === n) + assert(vec.indices.eq(indices), "should not copy data") + assert(vec.values.eq(values), "should not copy data") + } + + test("sparse breeze with partially-used arrays to vector") { + val activeSize = 3 + val breeze = new BSV[Double](indices, values, activeSize, n) + val vec = Vectors.fromBreeze(breeze).asInstanceOf[SparseVector] + assert(vec.size === n) + assert(vec.indices === indices.slice(0, activeSize)) + assert(vec.values === values.slice(0, activeSize)) + } +} diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/linalg/MatricesSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/MatricesSuite.scala new file mode 100644 index 0000000000..5c69c5ed7b --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/MatricesSuite.scala @@ -0,0 +1,511 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.linalg + +import java.util.Random + +import breeze.linalg.{CSCMatrix, Matrix => BM} +import org.mockito.Mockito.when +import org.scalatest.mock.MockitoSugar._ +import scala.collection.mutable.{Map => MutableMap} + +import org.apache.spark.ml.SparkMLFunSuite +import org.apache.spark.ml.util.TestingUtils._ + +class MatricesSuite extends SparkMLFunSuite { + test("dense matrix construction") { + val m = 3 + val n = 2 + val values = Array(0.0, 1.0, 2.0, 3.0, 4.0, 5.0) + val mat = Matrices.dense(m, n, values).asInstanceOf[DenseMatrix] + assert(mat.numRows === m) + assert(mat.numCols === n) + assert(mat.values.eq(values), "should not copy data") + } + + test("dense matrix construction with wrong dimension") { + intercept[RuntimeException] { + Matrices.dense(3, 2, Array(0.0, 1.0, 2.0)) + } + } + + test("sparse matrix construction") { + val m = 3 + val n = 4 + val values = Array(1.0, 2.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 2, 4, 4) + val rowIndices = Array(1, 2, 1, 2) + val mat = Matrices.sparse(m, n, colPtrs, rowIndices, values).asInstanceOf[SparseMatrix] + assert(mat.numRows === m) + assert(mat.numCols === n) + assert(mat.values.eq(values), "should not copy data") + assert(mat.colPtrs.eq(colPtrs), "should not copy data") + assert(mat.rowIndices.eq(rowIndices), "should not copy data") + + val entries: Array[(Int, Int, Double)] = Array((2, 2, 3.0), (1, 0, 1.0), (2, 0, 2.0), + (1, 2, 2.0), (2, 2, 2.0), (1, 2, 2.0), (0, 0, 0.0)) + + val mat2 = SparseMatrix.fromCOO(m, n, entries) + assert(mat.toBreeze === mat2.toBreeze) + assert(mat2.values.length == 4) + } + + test("sparse matrix construction with wrong number of elements") { + intercept[IllegalArgumentException] { + Matrices.sparse(3, 2, Array(0, 1), Array(1, 2, 1), Array(0.0, 1.0, 2.0)) + } + + intercept[IllegalArgumentException] { + Matrices.sparse(3, 2, Array(0, 1, 2), Array(1, 2), Array(0.0, 1.0, 2.0)) + } + } + + test("index in matrices incorrect input") { + val sm = Matrices.sparse(3, 2, Array(0, 2, 3), Array(1, 2, 1), Array(0.0, 1.0, 2.0)) + val dm = Matrices.dense(3, 2, Array(0.0, 2.3, 1.4, 3.2, 1.0, 9.1)) + Array(sm, dm).foreach { mat => + intercept[IllegalArgumentException] { mat.index(4, 1) } + intercept[IllegalArgumentException] { mat.index(1, 4) } + intercept[IllegalArgumentException] { mat.index(-1, 2) } + intercept[IllegalArgumentException] { mat.index(1, -2) } + } + } + + test("equals") { + val dm1 = Matrices.dense(2, 2, Array(0.0, 1.0, 2.0, 3.0)) + assert(dm1 === dm1) + assert(dm1 !== dm1.transpose) + + val dm2 = Matrices.dense(2, 2, Array(0.0, 2.0, 1.0, 3.0)) + assert(dm1 === dm2.transpose) + + val sm1 = dm1.asInstanceOf[DenseMatrix].toSparse + assert(sm1 === sm1) + assert(sm1 === dm1) + assert(sm1 !== sm1.transpose) + + val sm2 = dm2.asInstanceOf[DenseMatrix].toSparse + assert(sm1 === sm2.transpose) + assert(sm1 === dm2.transpose) + } + + test("matrix copies are deep copies") { + val m = 3 + val n = 2 + + val denseMat = Matrices.dense(m, n, Array(0.0, 1.0, 2.0, 3.0, 4.0, 5.0)) + val denseCopy = denseMat.copy + + assert(!denseMat.toArray.eq(denseCopy.toArray)) + + val values = Array(1.0, 2.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(1, 2, 1, 2) + val sparseMat = Matrices.sparse(m, n, colPtrs, rowIndices, values) + val sparseCopy = sparseMat.copy + + assert(!sparseMat.toArray.eq(sparseCopy.toArray)) + } + + test("matrix indexing and updating") { + val m = 3 + val n = 2 + val allValues = Array(0.0, 1.0, 2.0, 3.0, 4.0, 0.0) + + val denseMat = new DenseMatrix(m, n, allValues) + + assert(denseMat(0, 1) === 3.0) + assert(denseMat(0, 1) === denseMat.values(3)) + assert(denseMat(0, 1) === denseMat(3)) + assert(denseMat(0, 0) === 0.0) + + denseMat.update(0, 0, 10.0) + assert(denseMat(0, 0) === 10.0) + assert(denseMat.values(0) === 10.0) + + val sparseValues = Array(1.0, 2.0, 3.0, 4.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(1, 2, 0, 1) + val sparseMat = new SparseMatrix(m, n, colPtrs, rowIndices, sparseValues) + + assert(sparseMat(0, 1) === 3.0) + assert(sparseMat(0, 1) === sparseMat.values(2)) + assert(sparseMat(0, 0) === 0.0) + + intercept[NoSuchElementException] { + sparseMat.update(0, 0, 10.0) + } + + intercept[NoSuchElementException] { + sparseMat.update(2, 1, 10.0) + } + + sparseMat.update(0, 1, 10.0) + assert(sparseMat(0, 1) === 10.0) + assert(sparseMat.values(2) === 10.0) + } + + test("toSparse, toDense") { + val m = 3 + val n = 2 + val values = Array(1.0, 2.0, 4.0, 5.0) + val allValues = Array(1.0, 2.0, 0.0, 0.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(0, 1, 1, 2) + + val spMat1 = new SparseMatrix(m, n, colPtrs, rowIndices, values) + val deMat1 = new DenseMatrix(m, n, allValues) + + val spMat2 = deMat1.toSparse + val deMat2 = spMat1.toDense + + assert(spMat1.toBreeze === spMat2.toBreeze) + assert(deMat1.toBreeze === deMat2.toBreeze) + } + + test("map, update") { + val m = 3 + val n = 2 + val values = Array(1.0, 2.0, 4.0, 5.0) + val allValues = Array(1.0, 2.0, 0.0, 0.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(0, 1, 1, 2) + + val spMat1 = new SparseMatrix(m, n, colPtrs, rowIndices, values) + val deMat1 = new DenseMatrix(m, n, allValues) + val deMat2 = deMat1.map(_ * 2) + val spMat2 = spMat1.map(_ * 2) + deMat1.update(_ * 2) + spMat1.update(_ * 2) + + assert(spMat1.toArray === spMat2.toArray) + assert(deMat1.toArray === deMat2.toArray) + } + + test("transpose") { + val dA = + new DenseMatrix(4, 3, Array(0.0, 1.0, 0.0, 0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 3.0)) + val sA = new SparseMatrix(4, 3, Array(0, 1, 3, 4), Array(1, 0, 2, 3), Array(1.0, 2.0, 1.0, 3.0)) + + val dAT = dA.transpose.asInstanceOf[DenseMatrix] + val sAT = sA.transpose.asInstanceOf[SparseMatrix] + val dATexpected = + new DenseMatrix(3, 4, Array(0.0, 2.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 3.0)) + val sATexpected = + new SparseMatrix(3, 4, Array(0, 1, 2, 3, 4), Array(1, 0, 1, 2), Array(2.0, 1.0, 1.0, 3.0)) + + assert(dAT.toBreeze === dATexpected.toBreeze) + assert(sAT.toBreeze === sATexpected.toBreeze) + assert(dA(1, 0) === dAT(0, 1)) + assert(dA(2, 1) === dAT(1, 2)) + assert(sA(1, 0) === sAT(0, 1)) + assert(sA(2, 1) === sAT(1, 2)) + + assert(!dA.toArray.eq(dAT.toArray), "has to have a new array") + assert(dA.values.eq(dAT.transpose.asInstanceOf[DenseMatrix].values), "should not copy array") + + assert(dAT.toSparse.toBreeze === sATexpected.toBreeze) + assert(sAT.toDense.toBreeze === dATexpected.toBreeze) + } + + test("foreachActive") { + val m = 3 + val n = 2 + val values = Array(1.0, 2.0, 4.0, 5.0) + val allValues = Array(1.0, 2.0, 0.0, 0.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(0, 1, 1, 2) + + val sp = new SparseMatrix(m, n, colPtrs, rowIndices, values) + val dn = new DenseMatrix(m, n, allValues) + + val dnMap = MutableMap[(Int, Int), Double]() + dn.foreachActive { (i, j, value) => + dnMap.put((i, j), value) + } + assert(dnMap.size === 6) + assert(dnMap(0, 0) === 1.0) + assert(dnMap(1, 0) === 2.0) + assert(dnMap(2, 0) === 0.0) + assert(dnMap(0, 1) === 0.0) + assert(dnMap(1, 1) === 4.0) + assert(dnMap(2, 1) === 5.0) + + val spMap = MutableMap[(Int, Int), Double]() + sp.foreachActive { (i, j, value) => + spMap.put((i, j), value) + } + assert(spMap.size === 4) + assert(spMap(0, 0) === 1.0) + assert(spMap(1, 0) === 2.0) + assert(spMap(1, 1) === 4.0) + assert(spMap(2, 1) === 5.0) + } + + test("horzcat, vertcat, eye, speye") { + val m = 3 + val n = 2 + val values = Array(1.0, 2.0, 4.0, 5.0) + val allValues = Array(1.0, 2.0, 0.0, 0.0, 4.0, 5.0) + val colPtrs = Array(0, 2, 4) + val rowIndices = Array(0, 1, 1, 2) + // transposed versions + val allValuesT = Array(1.0, 0.0, 2.0, 4.0, 0.0, 5.0) + val colPtrsT = Array(0, 1, 3, 4) + val rowIndicesT = Array(0, 0, 1, 1) + + val spMat1 = new SparseMatrix(m, n, colPtrs, rowIndices, values) + val deMat1 = new DenseMatrix(m, n, allValues) + val spMat1T = new SparseMatrix(n, m, colPtrsT, rowIndicesT, values) + val deMat1T = new DenseMatrix(n, m, allValuesT) + + // should equal spMat1 & deMat1 respectively + val spMat1TT = spMat1T.transpose + val deMat1TT = deMat1T.transpose + + val deMat2 = Matrices.eye(3) + val spMat2 = Matrices.speye(3) + val deMat3 = Matrices.eye(2) + val spMat3 = Matrices.speye(2) + + val spHorz = Matrices.horzcat(Array(spMat1, spMat2)) + val spHorz2 = Matrices.horzcat(Array(spMat1, deMat2)) + val spHorz3 = Matrices.horzcat(Array(deMat1, spMat2)) + val deHorz1 = Matrices.horzcat(Array(deMat1, deMat2)) + val deHorz2 = Matrices.horzcat(Array[Matrix]()) + + assert(deHorz1.numRows === 3) + assert(spHorz2.numRows === 3) + assert(spHorz3.numRows === 3) + assert(spHorz.numRows === 3) + assert(deHorz1.numCols === 5) + assert(spHorz2.numCols === 5) + assert(spHorz3.numCols === 5) + assert(spHorz.numCols === 5) + assert(deHorz2.numRows === 0) + assert(deHorz2.numCols === 0) + assert(deHorz2.toArray.length === 0) + + assert(deHorz1 ~== spHorz2.asInstanceOf[SparseMatrix].toDense absTol 1e-15) + assert(spHorz2 ~== spHorz3 absTol 1e-15) + assert(spHorz(0, 0) === 1.0) + assert(spHorz(2, 1) === 5.0) + assert(spHorz(0, 2) === 1.0) + assert(spHorz(1, 2) === 0.0) + assert(spHorz(1, 3) === 1.0) + assert(spHorz(2, 4) === 1.0) + assert(spHorz(1, 4) === 0.0) + assert(deHorz1(0, 0) === 1.0) + assert(deHorz1(2, 1) === 5.0) + assert(deHorz1(0, 2) === 1.0) + assert(deHorz1(1, 2) == 0.0) + assert(deHorz1(1, 3) === 1.0) + assert(deHorz1(2, 4) === 1.0) + assert(deHorz1(1, 4) === 0.0) + + // containing transposed matrices + val spHorzT = Matrices.horzcat(Array(spMat1TT, spMat2)) + val spHorz2T = Matrices.horzcat(Array(spMat1TT, deMat2)) + val spHorz3T = Matrices.horzcat(Array(deMat1TT, spMat2)) + val deHorz1T = Matrices.horzcat(Array(deMat1TT, deMat2)) + + assert(deHorz1T ~== deHorz1 absTol 1e-15) + assert(spHorzT ~== spHorz absTol 1e-15) + assert(spHorz2T ~== spHorz2 absTol 1e-15) + assert(spHorz3T ~== spHorz3 absTol 1e-15) + + intercept[IllegalArgumentException] { + Matrices.horzcat(Array(spMat1, spMat3)) + } + + intercept[IllegalArgumentException] { + Matrices.horzcat(Array(deMat1, spMat3)) + } + + val spVert = Matrices.vertcat(Array(spMat1, spMat3)) + val deVert1 = Matrices.vertcat(Array(deMat1, deMat3)) + val spVert2 = Matrices.vertcat(Array(spMat1, deMat3)) + val spVert3 = Matrices.vertcat(Array(deMat1, spMat3)) + val deVert2 = Matrices.vertcat(Array[Matrix]()) + + assert(deVert1.numRows === 5) + assert(spVert2.numRows === 5) + assert(spVert3.numRows === 5) + assert(spVert.numRows === 5) + assert(deVert1.numCols === 2) + assert(spVert2.numCols === 2) + assert(spVert3.numCols === 2) + assert(spVert.numCols === 2) + assert(deVert2.numRows === 0) + assert(deVert2.numCols === 0) + assert(deVert2.toArray.length === 0) + + assert(deVert1 ~== spVert2.asInstanceOf[SparseMatrix].toDense absTol 1e-15) + assert(spVert2 ~== spVert3 absTol 1e-15) + assert(spVert(0, 0) === 1.0) + assert(spVert(2, 1) === 5.0) + assert(spVert(3, 0) === 1.0) + assert(spVert(3, 1) === 0.0) + assert(spVert(4, 1) === 1.0) + assert(deVert1(0, 0) === 1.0) + assert(deVert1(2, 1) === 5.0) + assert(deVert1(3, 0) === 1.0) + assert(deVert1(3, 1) === 0.0) + assert(deVert1(4, 1) === 1.0) + + // containing transposed matrices + val spVertT = Matrices.vertcat(Array(spMat1TT, spMat3)) + val deVert1T = Matrices.vertcat(Array(deMat1TT, deMat3)) + val spVert2T = Matrices.vertcat(Array(spMat1TT, deMat3)) + val spVert3T = Matrices.vertcat(Array(deMat1TT, spMat3)) + + assert(deVert1T ~== deVert1 absTol 1e-15) + assert(spVertT ~== spVert absTol 1e-15) + assert(spVert2T ~== spVert2 absTol 1e-15) + assert(spVert3T ~== spVert3 absTol 1e-15) + + intercept[IllegalArgumentException] { + Matrices.vertcat(Array(spMat1, spMat2)) + } + + intercept[IllegalArgumentException] { + Matrices.vertcat(Array(deMat1, spMat2)) + } + } + + test("zeros") { + val mat = Matrices.zeros(2, 3).asInstanceOf[DenseMatrix] + assert(mat.numRows === 2) + assert(mat.numCols === 3) + assert(mat.values.forall(_ == 0.0)) + } + + test("ones") { + val mat = Matrices.ones(2, 3).asInstanceOf[DenseMatrix] + assert(mat.numRows === 2) + assert(mat.numCols === 3) + assert(mat.values.forall(_ == 1.0)) + } + + test("eye") { + val mat = Matrices.eye(2).asInstanceOf[DenseMatrix] + assert(mat.numCols === 2) + assert(mat.numCols === 2) + assert(mat.values.toSeq === Seq(1.0, 0.0, 0.0, 1.0)) + } + + test("rand") { + val rng = mock[Random] + when(rng.nextDouble()).thenReturn(1.0, 2.0, 3.0, 4.0) + val mat = Matrices.rand(2, 2, rng).asInstanceOf[DenseMatrix] + assert(mat.numRows === 2) + assert(mat.numCols === 2) + assert(mat.values.toSeq === Seq(1.0, 2.0, 3.0, 4.0)) + } + + test("randn") { + val rng = mock[Random] + when(rng.nextGaussian()).thenReturn(1.0, 2.0, 3.0, 4.0) + val mat = Matrices.randn(2, 2, rng).asInstanceOf[DenseMatrix] + assert(mat.numRows === 2) + assert(mat.numCols === 2) + assert(mat.values.toSeq === Seq(1.0, 2.0, 3.0, 4.0)) + } + + test("diag") { + val mat = Matrices.diag(Vectors.dense(1.0, 2.0)).asInstanceOf[DenseMatrix] + assert(mat.numRows === 2) + assert(mat.numCols === 2) + assert(mat.values.toSeq === Seq(1.0, 0.0, 0.0, 2.0)) + } + + test("sprand") { + val rng = mock[Random] + when(rng.nextInt(4)).thenReturn(0, 1, 1, 3, 2, 2, 0, 1, 3, 0) + when(rng.nextDouble()).thenReturn(1.0, 2.0, 3.0, 4.0, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0) + val mat = SparseMatrix.sprand(4, 4, 0.25, rng) + assert(mat.numRows === 4) + assert(mat.numCols === 4) + assert(mat.rowIndices.toSeq === Seq(3, 0, 2, 1)) + assert(mat.values.toSeq === Seq(1.0, 2.0, 3.0, 4.0)) + val mat2 = SparseMatrix.sprand(2, 3, 1.0, rng) + assert(mat2.rowIndices.toSeq === Seq(0, 1, 0, 1, 0, 1)) + assert(mat2.colPtrs.toSeq === Seq(0, 2, 4, 6)) + } + + test("sprandn") { + val rng = mock[Random] + when(rng.nextInt(4)).thenReturn(0, 1, 1, 3, 2, 2, 0, 1, 3, 0) + when(rng.nextGaussian()).thenReturn(1.0, 2.0, 3.0, 4.0) + val mat = SparseMatrix.sprandn(4, 4, 0.25, rng) + assert(mat.numRows === 4) + assert(mat.numCols === 4) + assert(mat.rowIndices.toSeq === Seq(3, 0, 2, 1)) + assert(mat.values.toSeq === Seq(1.0, 2.0, 3.0, 4.0)) + } + + test("toString") { + val empty = Matrices.ones(0, 0) + empty.toString(0, 0) + + val mat = Matrices.rand(5, 10, new Random()) + mat.toString(-1, -5) + mat.toString(0, 0) + mat.toString(Int.MinValue, Int.MinValue) + mat.toString(Int.MaxValue, Int.MaxValue) + var lines = mat.toString(6, 50).lines.toArray + assert(lines.size == 5 && lines.forall(_.size <= 50)) + + lines = mat.toString(5, 100).lines.toArray + assert(lines.size == 5 && lines.forall(_.size <= 100)) + } + + test("numNonzeros and numActives") { + val dm1 = Matrices.dense(3, 2, Array(0, 0, -1, 1, 0, 1)) + assert(dm1.numNonzeros === 3) + assert(dm1.numActives === 6) + + val sm1 = Matrices.sparse(3, 2, Array(0, 2, 3), Array(0, 2, 1), Array(0.0, -1.2, 0.0)) + assert(sm1.numNonzeros === 1) + assert(sm1.numActives === 3) + } + + test("fromBreeze with sparse matrix") { + // colPtr.last does NOT always equal to values.length in breeze SCSMatrix and + // invocation of compact() may be necessary. Refer to SPARK-11507 + val bm1: BM[Double] = new CSCMatrix[Double]( + Array(1.0, 1, 1), 3, 3, Array(0, 1, 2, 3), Array(0, 1, 2)) + val bm2: BM[Double] = new CSCMatrix[Double]( + Array(1.0, 2, 2, 4), 3, 3, Array(0, 0, 2, 4), Array(1, 2, 1, 2)) + val sum = bm1 + bm2 + Matrices.fromBreeze(sum) + } + + test("row/col iterator") { + val dm = new DenseMatrix(3, 2, Array(0, 1, 2, 3, 4, 0)) + val sm = dm.toSparse + val rows = Seq(Vectors.dense(0, 3), Vectors.dense(1, 4), Vectors.dense(2, 0)) + val cols = Seq(Vectors.dense(0, 1, 2), Vectors.dense(3, 4, 0)) + for (m <- Seq(dm, sm)) { + assert(m.rowIter.toSeq === rows) + assert(m.colIter.toSeq === cols) + assert(m.transpose.rowIter.toSeq === cols) + assert(m.transpose.colIter.toSeq === rows) + } + } +} 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 new file mode 100644 index 0000000000..504be36413 --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/linalg/VectorsSuite.scala @@ -0,0 +1,358 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.linalg + +import scala.util.Random + +import breeze.linalg.{squaredDistance => breezeSquaredDistance, DenseMatrix => BDM} +import org.json4s.jackson.JsonMethods.{parse => parseJson} + +import org.apache.spark.ml.SparkMLFunSuite +import org.apache.spark.ml.util.TestingUtils._ + +class VectorsSuite extends SparkMLFunSuite { + + val arr = Array(0.1, 0.0, 0.3, 0.4) + val n = 4 + val indices = Array(0, 2, 3) + val values = Array(0.1, 0.3, 0.4) + + test("dense vector construction with varargs") { + val vec = Vectors.dense(arr).asInstanceOf[DenseVector] + assert(vec.size === arr.length) + assert(vec.values.eq(arr)) + } + + test("dense vector construction from a double array") { + val vec = Vectors.dense(arr).asInstanceOf[DenseVector] + assert(vec.size === arr.length) + assert(vec.values.eq(arr)) + } + + test("sparse vector construction") { + val vec = Vectors.sparse(n, indices, values).asInstanceOf[SparseVector] + assert(vec.size === n) + assert(vec.indices.eq(indices)) + assert(vec.values.eq(values)) + } + + test("sparse vector construction with unordered elements") { + val vec = Vectors.sparse(n, indices.zip(values).reverse).asInstanceOf[SparseVector] + assert(vec.size === n) + assert(vec.indices === indices) + assert(vec.values === values) + } + + test("sparse vector construction with mismatched indices/values array") { + intercept[IllegalArgumentException] { + Vectors.sparse(4, Array(1, 2, 3), Array(3.0, 5.0, 7.0, 9.0)) + } + intercept[IllegalArgumentException] { + Vectors.sparse(4, Array(1, 2, 3), Array(3.0, 5.0)) + } + } + + test("sparse vector construction with too many indices vs size") { + intercept[IllegalArgumentException] { + Vectors.sparse(3, Array(1, 2, 3, 4), Array(3.0, 5.0, 7.0, 9.0)) + } + } + + test("dense to array") { + val vec = Vectors.dense(arr).asInstanceOf[DenseVector] + assert(vec.toArray.eq(arr)) + } + + test("dense argmax") { + val vec = Vectors.dense(Array.empty[Double]).asInstanceOf[DenseVector] + assert(vec.argmax === -1) + + val vec2 = Vectors.dense(arr).asInstanceOf[DenseVector] + assert(vec2.argmax === 3) + + val vec3 = Vectors.dense(Array(-1.0, 0.0, -2.0, 1.0)).asInstanceOf[DenseVector] + assert(vec3.argmax === 3) + } + + test("sparse to array") { + val vec = Vectors.sparse(n, indices, values).asInstanceOf[SparseVector] + assert(vec.toArray === arr) + } + + test("sparse argmax") { + val vec = Vectors.sparse(0, Array.empty[Int], Array.empty[Double]).asInstanceOf[SparseVector] + assert(vec.argmax === -1) + + val vec2 = Vectors.sparse(n, indices, values).asInstanceOf[SparseVector] + assert(vec2.argmax === 3) + + val vec3 = Vectors.sparse(5, Array(2, 3, 4), Array(1.0, 0.0, -.7)) + assert(vec3.argmax === 2) + + // check for case that sparse vector is created with + // only negative values {0.0, 0.0,-1.0, -0.7, 0.0} + val vec4 = Vectors.sparse(5, Array(2, 3), Array(-1.0, -.7)) + assert(vec4.argmax === 0) + + val vec5 = Vectors.sparse(11, Array(0, 3, 10), Array(-1.0, -.7, 0.0)) + assert(vec5.argmax === 1) + + val vec6 = Vectors.sparse(11, Array(0, 1, 2), Array(-1.0, -.7, 0.0)) + assert(vec6.argmax === 2) + + val vec7 = Vectors.sparse(5, Array(0, 1, 3), Array(-1.0, 0.0, -.7)) + assert(vec7.argmax === 1) + + val vec8 = Vectors.sparse(5, Array(1, 2), Array(0.0, -1.0)) + assert(vec8.argmax === 0) + } + + test("vector equals") { + val dv1 = Vectors.dense(arr.clone()) + val dv2 = Vectors.dense(arr.clone()) + val sv1 = Vectors.sparse(n, indices.clone(), values.clone()) + val sv2 = Vectors.sparse(n, indices.clone(), values.clone()) + + val vectors = Seq(dv1, dv2, sv1, sv2) + + for (v <- vectors; u <- vectors) { + assert(v === u) + assert(v.## === u.##) + } + + val another = Vectors.dense(0.1, 0.2, 0.3, 0.4) + + for (v <- vectors) { + assert(v != another) + assert(v.## != another.##) + } + } + + test("vectors equals with explicit 0") { + val dv1 = Vectors.dense(Array(0, 0.9, 0, 0.8, 0)) + val sv1 = Vectors.sparse(5, Array(1, 3), Array(0.9, 0.8)) + val sv2 = Vectors.sparse(5, Array(0, 1, 2, 3, 4), Array(0, 0.9, 0, 0.8, 0)) + + val vectors = Seq(dv1, sv1, sv2) + for (v <- vectors; u <- vectors) { + assert(v === u) + assert(v.## === u.##) + } + + val another = Vectors.sparse(5, Array(0, 1, 3), Array(0, 0.9, 0.2)) + for (v <- vectors) { + assert(v != another) + assert(v.## != another.##) + } + } + + test("indexing dense vectors") { + val vec = Vectors.dense(1.0, 2.0, 3.0, 4.0) + assert(vec(0) === 1.0) + assert(vec(3) === 4.0) + } + + test("indexing sparse vectors") { + val vec = Vectors.sparse(7, Array(0, 2, 4, 6), Array(1.0, 2.0, 3.0, 4.0)) + assert(vec(0) === 1.0) + assert(vec(1) === 0.0) + assert(vec(2) === 2.0) + assert(vec(3) === 0.0) + assert(vec(6) === 4.0) + val vec2 = Vectors.sparse(8, Array(0, 2, 4, 6), Array(1.0, 2.0, 3.0, 4.0)) + assert(vec2(6) === 4.0) + assert(vec2(7) === 0.0) + } + + test("zeros") { + assert(Vectors.zeros(3) === Vectors.dense(0.0, 0.0, 0.0)) + } + + test("Vector.copy") { + val sv = Vectors.sparse(4, Array(0, 2), Array(1.0, 2.0)) + val svCopy = sv.copy + (sv, svCopy) match { + case (sv: SparseVector, svCopy: SparseVector) => + assert(sv.size === svCopy.size) + assert(sv.indices === svCopy.indices) + assert(sv.values === svCopy.values) + assert(!sv.indices.eq(svCopy.indices)) + assert(!sv.values.eq(svCopy.values)) + case _ => + throw new RuntimeException(s"copy returned ${svCopy.getClass} on ${sv.getClass}.") + } + + val dv = Vectors.dense(1.0, 0.0, 2.0) + val dvCopy = dv.copy + (dv, dvCopy) match { + case (dv: DenseVector, dvCopy: DenseVector) => + assert(dv.size === dvCopy.size) + assert(dv.values === dvCopy.values) + assert(!dv.values.eq(dvCopy.values)) + case _ => + throw new RuntimeException(s"copy returned ${dvCopy.getClass} on ${dv.getClass}.") + } + } + + test("fromBreeze") { + val x = BDM.zeros[Double](10, 10) + val v = Vectors.fromBreeze(x(::, 0)) + assert(v.size === x.rows) + } + + test("sqdist") { + val random = new Random() + for (m <- 1 until 1000 by 100) { + val nnz = random.nextInt(m) + + val indices1 = random.shuffle(0 to m - 1).slice(0, nnz).sorted.toArray + val values1 = Array.fill(nnz)(random.nextDouble) + val sparseVector1 = Vectors.sparse(m, indices1, values1) + + val indices2 = random.shuffle(0 to m - 1).slice(0, nnz).sorted.toArray + val values2 = Array.fill(nnz)(random.nextDouble) + val sparseVector2 = Vectors.sparse(m, indices2, values2) + + val denseVector1 = Vectors.dense(sparseVector1.toArray) + val denseVector2 = Vectors.dense(sparseVector2.toArray) + + val squaredDist = breezeSquaredDistance(sparseVector1.toBreeze, sparseVector2.toBreeze) + + // SparseVector vs. SparseVector + assert(Vectors.sqdist(sparseVector1, sparseVector2) ~== squaredDist relTol 1E-8) + // DenseVector vs. SparseVector + assert(Vectors.sqdist(denseVector1, sparseVector2) ~== squaredDist relTol 1E-8) + // DenseVector vs. DenseVector + assert(Vectors.sqdist(denseVector1, denseVector2) ~== squaredDist relTol 1E-8) + } + } + + test("foreachActive") { + val dv = Vectors.dense(0.0, 1.2, 3.1, 0.0) + val sv = Vectors.sparse(4, Seq((1, 1.2), (2, 3.1), (3, 0.0))) + + val dvMap = scala.collection.mutable.Map[Int, Double]() + dv.foreachActive { (index, value) => + dvMap.put(index, value) + } + assert(dvMap.size === 4) + assert(dvMap.get(0) === Some(0.0)) + assert(dvMap.get(1) === Some(1.2)) + assert(dvMap.get(2) === Some(3.1)) + assert(dvMap.get(3) === Some(0.0)) + + val svMap = scala.collection.mutable.Map[Int, Double]() + sv.foreachActive { (index, value) => + svMap.put(index, value) + } + assert(svMap.size === 3) + assert(svMap.get(1) === Some(1.2)) + assert(svMap.get(2) === Some(3.1)) + assert(svMap.get(3) === Some(0.0)) + } + + test("vector p-norm") { + val dv = Vectors.dense(0.0, -1.2, 3.1, 0.0, -4.5, 1.9) + val sv = Vectors.sparse(6, Seq((1, -1.2), (2, 3.1), (3, 0.0), (4, -4.5), (5, 1.9))) + + assert(Vectors.norm(dv, 1.0) ~== dv.toArray.foldLeft(0.0)((a, v) => + a + math.abs(v)) relTol 1E-8) + assert(Vectors.norm(sv, 1.0) ~== sv.toArray.foldLeft(0.0)((a, v) => + a + math.abs(v)) relTol 1E-8) + + assert(Vectors.norm(dv, 2.0) ~== math.sqrt(dv.toArray.foldLeft(0.0)((a, v) => + a + v * v)) relTol 1E-8) + assert(Vectors.norm(sv, 2.0) ~== math.sqrt(sv.toArray.foldLeft(0.0)((a, v) => + a + v * v)) relTol 1E-8) + + assert(Vectors.norm(dv, Double.PositiveInfinity) ~== dv.toArray.map(math.abs).max relTol 1E-8) + assert(Vectors.norm(sv, Double.PositiveInfinity) ~== sv.toArray.map(math.abs).max relTol 1E-8) + + assert(Vectors.norm(dv, 3.7) ~== math.pow(dv.toArray.foldLeft(0.0)((a, v) => + a + math.pow(math.abs(v), 3.7)), 1.0 / 3.7) relTol 1E-8) + assert(Vectors.norm(sv, 3.7) ~== math.pow(sv.toArray.foldLeft(0.0)((a, v) => + a + math.pow(math.abs(v), 3.7)), 1.0 / 3.7) relTol 1E-8) + } + + test("Vector numActive and numNonzeros") { + val dv = Vectors.dense(0.0, 2.0, 3.0, 0.0) + assert(dv.numActives === 4) + assert(dv.numNonzeros === 2) + + val sv = Vectors.sparse(4, Array(0, 1, 2), Array(0.0, 2.0, 3.0)) + assert(sv.numActives === 3) + assert(sv.numNonzeros === 2) + } + + test("Vector toSparse and toDense") { + val dv0 = Vectors.dense(0.0, 2.0, 3.0, 0.0) + assert(dv0.toDense === dv0) + val dv0s = dv0.toSparse + assert(dv0s.numActives === 2) + assert(dv0s === dv0) + + val sv0 = Vectors.sparse(4, Array(0, 1, 2), Array(0.0, 2.0, 3.0)) + assert(sv0.toDense === sv0) + val sv0s = sv0.toSparse + assert(sv0s.numActives === 2) + assert(sv0s === sv0) + } + + test("Vector.compressed") { + val dv0 = Vectors.dense(1.0, 2.0, 3.0, 0.0) + val dv0c = dv0.compressed.asInstanceOf[DenseVector] + assert(dv0c === dv0) + + val dv1 = Vectors.dense(0.0, 2.0, 0.0, 0.0) + val dv1c = dv1.compressed.asInstanceOf[SparseVector] + assert(dv1 === dv1c) + assert(dv1c.numActives === 1) + + val sv0 = Vectors.sparse(4, Array(1, 2), Array(2.0, 0.0)) + val sv0c = sv0.compressed.asInstanceOf[SparseVector] + assert(sv0 === sv0c) + assert(sv0c.numActives === 1) + + val sv1 = Vectors.sparse(4, Array(0, 1, 2), Array(1.0, 2.0, 3.0)) + val sv1c = sv1.compressed.asInstanceOf[DenseVector] + assert(sv1 === sv1c) + } + + test("SparseVector.slice") { + val v = new SparseVector(5, Array(1, 2, 4), Array(1.1, 2.2, 4.4)) + assert(v.slice(Array(0, 2)) === new SparseVector(2, Array(1), Array(2.2))) + 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.") + } + } +} diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtils.scala b/mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtils.scala new file mode 100644 index 0000000000..2bebaa35ba --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtils.scala @@ -0,0 +1,236 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.util + +import org.scalatest.exceptions.TestFailedException + +import org.apache.spark.ml.linalg.{Matrix, Vector} + +object TestingUtils { + + val ABS_TOL_MSG = " using absolute tolerance" + val REL_TOL_MSG = " using relative tolerance" + + /** + * Private helper function for comparing two values using relative tolerance. + * Note that if x or y is extremely close to zero, i.e., smaller than Double.MinPositiveValue, + * the relative tolerance is meaningless, so the exception will be raised to warn users. + */ + private def RelativeErrorComparison(x: Double, y: Double, eps: Double): Boolean = { + val absX = math.abs(x) + val absY = math.abs(y) + val diff = math.abs(x - y) + if (x == y) { + true + } else if (absX < Double.MinPositiveValue || absY < Double.MinPositiveValue) { + throw new TestFailedException( + s"$x or $y is extremely close to zero, so the relative tolerance is meaningless.", 0) + } else { + diff < eps * math.min(absX, absY) + } + } + + /** + * Private helper function for comparing two values using absolute tolerance. + */ + private def AbsoluteErrorComparison(x: Double, y: Double, eps: Double): Boolean = { + math.abs(x - y) < eps + } + + case class CompareDoubleRightSide( + fun: (Double, Double, Double) => Boolean, y: Double, eps: Double, method: String) + + /** + * Implicit class for comparing two double values using relative tolerance or absolute tolerance. + */ + implicit class DoubleWithAlmostEquals(val x: Double) { + + /** + * When the difference of two values are within eps, returns true; otherwise, returns false. + */ + def ~=(r: CompareDoubleRightSide): Boolean = r.fun(x, r.y, r.eps) + + /** + * When the difference of two values are within eps, returns false; otherwise, returns true. + */ + def !~=(r: CompareDoubleRightSide): Boolean = !r.fun(x, r.y, r.eps) + + /** + * Throws exception when the difference of two values are NOT within eps; + * otherwise, returns true. + */ + def ~==(r: CompareDoubleRightSide): Boolean = { + if (!r.fun(x, r.y, r.eps)) { + throw new TestFailedException( + s"Expected $x and ${r.y} to be within ${r.eps}${r.method}.", 0) + } + true + } + + /** + * Throws exception when the difference of two values are within eps; otherwise, returns true. + */ + def !~==(r: CompareDoubleRightSide): Boolean = { + if (r.fun(x, r.y, r.eps)) { + throw new TestFailedException( + s"Did not expect $x and ${r.y} to be within ${r.eps}${r.method}.", 0) + } + true + } + + /** + * Comparison using absolute tolerance. + */ + def absTol(eps: Double): CompareDoubleRightSide = + CompareDoubleRightSide(AbsoluteErrorComparison, x, eps, ABS_TOL_MSG) + + /** + * Comparison using relative tolerance. + */ + def relTol(eps: Double): CompareDoubleRightSide = + CompareDoubleRightSide(RelativeErrorComparison, x, eps, REL_TOL_MSG) + + override def toString: String = x.toString + } + + case class CompareVectorRightSide( + fun: (Vector, Vector, Double) => Boolean, y: Vector, eps: Double, method: String) + + /** + * Implicit class for comparing two vectors using relative tolerance or absolute tolerance. + */ + implicit class VectorWithAlmostEquals(val x: Vector) { + + /** + * When the difference of two vectors are within eps, returns true; otherwise, returns false. + */ + def ~=(r: CompareVectorRightSide): Boolean = r.fun(x, r.y, r.eps) + + /** + * When the difference of two vectors are within eps, returns false; otherwise, returns true. + */ + def !~=(r: CompareVectorRightSide): Boolean = !r.fun(x, r.y, r.eps) + + /** + * Throws exception when the difference of two vectors are NOT within eps; + * otherwise, returns true. + */ + def ~==(r: CompareVectorRightSide): Boolean = { + if (!r.fun(x, r.y, r.eps)) { + throw new TestFailedException( + s"Expected $x and ${r.y} to be within ${r.eps}${r.method} for all elements.", 0) + } + true + } + + /** + * Throws exception when the difference of two vectors are within eps; otherwise, returns true. + */ + def !~==(r: CompareVectorRightSide): Boolean = { + if (r.fun(x, r.y, r.eps)) { + throw new TestFailedException( + s"Did not expect $x and ${r.y} to be within ${r.eps}${r.method} for all elements.", 0) + } + true + } + + /** + * Comparison using absolute tolerance. + */ + def absTol(eps: Double): CompareVectorRightSide = CompareVectorRightSide( + (x: Vector, y: Vector, eps: Double) => { + x.toArray.zip(y.toArray).forall(x => x._1 ~= x._2 absTol eps) + }, x, eps, ABS_TOL_MSG) + + /** + * Comparison using relative tolerance. Note that comparing against sparse vector + * with elements having value of zero will raise exception because it involves with + * comparing against zero. + */ + def relTol(eps: Double): CompareVectorRightSide = CompareVectorRightSide( + (x: Vector, y: Vector, eps: Double) => { + x.toArray.zip(y.toArray).forall(x => x._1 ~= x._2 relTol eps) + }, x, eps, REL_TOL_MSG) + + override def toString: String = x.toString + } + + case class CompareMatrixRightSide( + fun: (Matrix, Matrix, Double) => Boolean, y: Matrix, eps: Double, method: String) + + /** + * Implicit class for comparing two matrices using relative tolerance or absolute tolerance. + */ + implicit class MatrixWithAlmostEquals(val x: Matrix) { + + /** + * When the difference of two matrices are within eps, returns true; otherwise, returns false. + */ + def ~=(r: CompareMatrixRightSide): Boolean = r.fun(x, r.y, r.eps) + + /** + * When the difference of two matrices are within eps, returns false; otherwise, returns true. + */ + def !~=(r: CompareMatrixRightSide): Boolean = !r.fun(x, r.y, r.eps) + + /** + * Throws exception when the difference of two matrices are NOT within eps; + * otherwise, returns true. + */ + def ~==(r: CompareMatrixRightSide): Boolean = { + if (!r.fun(x, r.y, r.eps)) { + throw new TestFailedException( + s"Expected \n$x\n and \n${r.y}\n to be within ${r.eps}${r.method} for all elements.", 0) + } + true + } + + /** + * Throws exception when the difference of two matrices are within eps; otherwise, returns true. + */ + def !~==(r: CompareMatrixRightSide): Boolean = { + if (r.fun(x, r.y, r.eps)) { + throw new TestFailedException( + s"Did not expect \n$x\n and \n${r.y}\n to be within " + + "${r.eps}${r.method} for all elements.", 0) + } + true + } + + /** + * Comparison using absolute tolerance. + */ + def absTol(eps: Double): CompareMatrixRightSide = CompareMatrixRightSide( + (x: Matrix, y: Matrix, eps: Double) => { + x.toArray.zip(y.toArray).forall(x => x._1 ~= x._2 absTol eps) + }, x, eps, ABS_TOL_MSG) + + /** + * Comparison using relative tolerance. Note that comparing against sparse vector + * with elements having value of zero will raise exception because it involves with + * comparing against zero. + */ + def relTol(eps: Double): CompareMatrixRightSide = CompareMatrixRightSide( + (x: Matrix, y: Matrix, eps: Double) => { + x.toArray.zip(y.toArray).forall(x => x._1 ~= x._2 relTol eps) + }, x, eps, REL_TOL_MSG) + + override def toString: String = x.toString + } + +} diff --git a/mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtilsSuite.scala b/mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtilsSuite.scala new file mode 100644 index 0000000000..e374165f75 --- /dev/null +++ b/mllib-local/src/test/scala/org/apache/spark/ml/util/TestingUtilsSuite.scala @@ -0,0 +1,187 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.ml.util + +import org.scalatest.exceptions.TestFailedException + +import org.apache.spark.ml.SparkMLFunSuite +import org.apache.spark.ml.linalg.Vectors +import org.apache.spark.ml.util.TestingUtils._ + +class TestingUtilsSuite extends SparkMLFunSuite { + + test("Comparing doubles using relative error.") { + + assert(23.1 ~== 23.52 relTol 0.02) + assert(23.1 ~== 22.74 relTol 0.02) + assert(23.1 ~= 23.52 relTol 0.02) + assert(23.1 ~= 22.74 relTol 0.02) + assert(!(23.1 !~= 23.52 relTol 0.02)) + assert(!(23.1 !~= 22.74 relTol 0.02)) + + // Should throw exception with message when test fails. + intercept[TestFailedException](23.1 !~== 23.52 relTol 0.02) + intercept[TestFailedException](23.1 !~== 22.74 relTol 0.02) + intercept[TestFailedException](23.1 ~== 23.63 relTol 0.02) + intercept[TestFailedException](23.1 ~== 22.34 relTol 0.02) + + assert(23.1 !~== 23.63 relTol 0.02) + assert(23.1 !~== 22.34 relTol 0.02) + assert(23.1 !~= 23.63 relTol 0.02) + assert(23.1 !~= 22.34 relTol 0.02) + assert(!(23.1 ~= 23.63 relTol 0.02)) + assert(!(23.1 ~= 22.34 relTol 0.02)) + + // Comparing against zero should fail the test and throw exception with message + // saying that the relative error is meaningless in this situation. + intercept[TestFailedException](0.1 ~== 0.0 relTol 0.032) + intercept[TestFailedException](0.1 ~= 0.0 relTol 0.032) + intercept[TestFailedException](0.1 !~== 0.0 relTol 0.032) + intercept[TestFailedException](0.1 !~= 0.0 relTol 0.032) + intercept[TestFailedException](0.0 ~== 0.1 relTol 0.032) + intercept[TestFailedException](0.0 ~= 0.1 relTol 0.032) + intercept[TestFailedException](0.0 !~== 0.1 relTol 0.032) + intercept[TestFailedException](0.0 !~= 0.1 relTol 0.032) + + // Comparisons of numbers very close to zero. + assert(10 * Double.MinPositiveValue ~== 9.5 * Double.MinPositiveValue relTol 0.01) + assert(10 * Double.MinPositiveValue !~== 11 * Double.MinPositiveValue relTol 0.01) + + assert(-Double.MinPositiveValue ~== 1.18 * -Double.MinPositiveValue relTol 0.012) + assert(-Double.MinPositiveValue ~== 1.38 * -Double.MinPositiveValue relTol 0.012) + } + + test("Comparing doubles using absolute error.") { + + assert(17.8 ~== 17.99 absTol 0.2) + assert(17.8 ~== 17.61 absTol 0.2) + assert(17.8 ~= 17.99 absTol 0.2) + assert(17.8 ~= 17.61 absTol 0.2) + assert(!(17.8 !~= 17.99 absTol 0.2)) + assert(!(17.8 !~= 17.61 absTol 0.2)) + + // Should throw exception with message when test fails. + intercept[TestFailedException](17.8 !~== 17.99 absTol 0.2) + intercept[TestFailedException](17.8 !~== 17.61 absTol 0.2) + intercept[TestFailedException](17.8 ~== 18.01 absTol 0.2) + intercept[TestFailedException](17.8 ~== 17.59 absTol 0.2) + + assert(17.8 !~== 18.01 absTol 0.2) + assert(17.8 !~== 17.59 absTol 0.2) + assert(17.8 !~= 18.01 absTol 0.2) + assert(17.8 !~= 17.59 absTol 0.2) + assert(!(17.8 ~= 18.01 absTol 0.2)) + assert(!(17.8 ~= 17.59 absTol 0.2)) + + // Comparisons of numbers very close to zero, and both side of zeros + assert( + Double.MinPositiveValue ~== 4 * Double.MinPositiveValue absTol 5 * Double.MinPositiveValue) + assert( + Double.MinPositiveValue !~== 6 * Double.MinPositiveValue absTol 5 * Double.MinPositiveValue) + + assert( + -Double.MinPositiveValue ~== 3 * Double.MinPositiveValue absTol 5 * Double.MinPositiveValue) + assert( + Double.MinPositiveValue !~== -4 * Double.MinPositiveValue absTol 5 * Double.MinPositiveValue) + } + + test("Comparing vectors using relative error.") { + + // Comparisons of two dense vectors + assert(Vectors.dense(Array(3.1, 3.5)) ~== Vectors.dense(Array(3.130, 3.534)) relTol 0.01) + assert(Vectors.dense(Array(3.1, 3.5)) !~== Vectors.dense(Array(3.135, 3.534)) relTol 0.01) + assert(Vectors.dense(Array(3.1, 3.5)) ~= Vectors.dense(Array(3.130, 3.534)) relTol 0.01) + assert(Vectors.dense(Array(3.1, 3.5)) !~= Vectors.dense(Array(3.135, 3.534)) relTol 0.01) + assert(!(Vectors.dense(Array(3.1, 3.5)) !~= Vectors.dense(Array(3.130, 3.534)) relTol 0.01)) + assert(!(Vectors.dense(Array(3.1, 3.5)) ~= Vectors.dense(Array(3.135, 3.534)) relTol 0.01)) + + // Should throw exception with message when test fails. + intercept[TestFailedException]( + Vectors.dense(Array(3.1, 3.5)) !~== Vectors.dense(Array(3.130, 3.534)) relTol 0.01) + + intercept[TestFailedException]( + Vectors.dense(Array(3.1, 3.5)) ~== Vectors.dense(Array(3.135, 3.534)) relTol 0.01) + + // Comparing against zero should fail the test and throw exception with message + // saying that the relative error is meaningless in this situation. + intercept[TestFailedException]( + Vectors.dense(Array(3.1, 0.01)) ~== Vectors.dense(Array(3.13, 0.0)) relTol 0.01) + + intercept[TestFailedException]( + Vectors.dense(Array(3.1, 0.01)) ~== Vectors.sparse(2, Array(0), Array(3.13)) relTol 0.01) + + // Comparisons of two sparse vectors + assert(Vectors.dense(Array(3.1, 3.5)) ~== + Vectors.sparse(2, Array(0, 1), Array(3.130, 3.534)) relTol 0.01) + + assert(Vectors.dense(Array(3.1, 3.5)) !~== + Vectors.sparse(2, Array(0, 1), Array(3.135, 3.534)) relTol 0.01) + } + + test("Comparing vectors using absolute error.") { + + // Comparisons of two dense vectors + assert(Vectors.dense(Array(3.1, 3.5, 0.0)) ~== + Vectors.dense(Array(3.1 + 1E-8, 3.5 + 2E-7, 1E-8)) absTol 1E-6) + + assert(Vectors.dense(Array(3.1, 3.5, 0.0)) !~== + Vectors.dense(Array(3.1 + 1E-5, 3.5 + 2E-7, 1 + 1E-3)) absTol 1E-6) + + assert(Vectors.dense(Array(3.1, 3.5, 0.0)) ~= + Vectors.dense(Array(3.1 + 1E-8, 3.5 + 2E-7, 1E-8)) absTol 1E-6) + + assert(Vectors.dense(Array(3.1, 3.5, 0.0)) !~= + Vectors.dense(Array(3.1 + 1E-5, 3.5 + 2E-7, 1 + 1E-3)) absTol 1E-6) + + assert(!(Vectors.dense(Array(3.1, 3.5, 0.0)) !~= + Vectors.dense(Array(3.1 + 1E-8, 3.5 + 2E-7, 1E-8)) absTol 1E-6)) + + assert(!(Vectors.dense(Array(3.1, 3.5, 0.0)) ~= + Vectors.dense(Array(3.1 + 1E-5, 3.5 + 2E-7, 1 + 1E-3)) absTol 1E-6)) + + // Should throw exception with message when test fails. + intercept[TestFailedException](Vectors.dense(Array(3.1, 3.5, 0.0)) !~== + Vectors.dense(Array(3.1 + 1E-8, 3.5 + 2E-7, 1E-8)) absTol 1E-6) + + intercept[TestFailedException](Vectors.dense(Array(3.1, 3.5, 0.0)) ~== + Vectors.dense(Array(3.1 + 1E-5, 3.5 + 2E-7, 1 + 1E-3)) absTol 1E-6) + + // Comparisons of two sparse vectors + assert(Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) ~== + Vectors.sparse(3, Array(0, 2), Array(3.1 + 1E-8, 2.4 + 1E-7)) absTol 1E-6) + + assert(Vectors.sparse(3, Array(0, 2), Array(3.1 + 1E-8, 2.4 + 1E-7)) ~== + Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) absTol 1E-6) + + assert(Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) !~== + Vectors.sparse(3, Array(0, 2), Array(3.1 + 1E-3, 2.4)) absTol 1E-6) + + assert(Vectors.sparse(3, Array(0, 2), Array(3.1 + 1E-3, 2.4)) !~== + Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) absTol 1E-6) + + // Comparisons of a dense vector and a sparse vector + assert(Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) ~== + Vectors.dense(Array(3.1 + 1E-8, 0, 2.4 + 1E-7)) absTol 1E-6) + + assert(Vectors.dense(Array(3.1 + 1E-8, 0, 2.4 + 1E-7)) ~== + Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) absTol 1E-6) + + assert(Vectors.sparse(3, Array(0, 2), Array(3.1, 2.4)) !~== + Vectors.dense(Array(3.1, 1E-3, 2.4)) absTol 1E-6) + } +} |