src/test/java/org/apache/sysds/test/applications/UnivariateStatsBase.java - systemds - Git at Google

 /*
  * 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.sysds.test.applications;

 import java.util.HashMap;

 import org.apache.sysds.common.Types.ExecMode;
 import org.apache.sysds.runtime.matrix.data.MatrixValue.CellIndex;
 import org.apache.sysds.test.AutomatedTestBase;
 import org.apache.sysds.test.TestConfiguration;
 import org.apache.sysds.test.TestUtils;

 /**
  * Shared methods and fields from the different univariate stats test harness
  * classes.
  */
 public abstract class UnivariateStatsBase extends AutomatedTestBase {

 	/**
 	 * Hard-coded path to the directory where this test keeps all of its files,
 	 * generated and static
 	 */
 	protected final static String TEST_DIR = "applications/descriptivestats/";
 	protected String TEST_CLASS_DIR = TEST_DIR + UnivariateStatsBase.class.getSimpleName() + "/";

 	/** Fudge factor for comparisons against R's output */
 	protected final static double epsilon = 0.000000001;

 	/**
 	 * Number of rows in the primary input to each aggregate ("V" in the test
 	 * script, "X" in the SystemDS documentation)
 	 */
 	protected final static int rows1 = 2000;

 	/**
 	 * Number of rows in the secondary input to aggregates that take two input
 	 * vectors ("P" in the test script and the documentation)
 	 */
 	protected final static int rows2 = 5;

 	/**
 	 * Sizes of inputs used in tests, measured as number of rows in primary
 	 * input vector. DIV4=divisible by 4; DIV4P1=divisible by 4 plus 1
 	 */
 	protected enum SIZE {
 		DIV4(2000), DIV4P1(2001), DIV4P2(2002), DIV4P3(2003);
 		int size = -1;

 		SIZE(int s) {
 			size = s;
 		}
 	}

 	/** Ranges of values passed to aggregates in different tests. */
 	protected enum RANGE {
 		NEG(-255, -2), MIXED(-200, 200), POS(2, 255);
 		double min, max;

 		RANGE(double mn, double mx) {
 			min = mn;
 			max = mx;
 		}
 	}

 	/**
 	 * Actual sparsity values used in the "dense" and "sparse" variants of the
 	 * tests.
 	 */
 	protected enum SPARSITY {
 		SPARSE(0.3), DENSE(0.8);
 		double sparsity;

 		SPARSITY(double sp) {
 			sparsity = sp;
 		}
 	}

 	/** Shared setup code for test harness configurations */
 	@Override
 	public void setUp() {
 		TestUtils.clearAssertionInformation();

 		addTestConfiguration("Scale", new TestConfiguration(TEST_CLASS_DIR, "Scale",
 				new String[] { "mean" + ".scalar", "std" + ".scalar",
 						"se" + ".scalar", "var" + ".scalar", "cv" + ".scalar",
 						/* "har", "geom", */
 						"min" + ".scalar", "max" + ".scalar",
 						"rng" + ".scalar", "g1" + ".scalar",
 						"se_g1" + ".scalar", "g2" + ".scalar",
 						"se_g2" + ".scalar", "out_minus", "out_plus",
 						"median" + ".scalar", "quantile", "iqm" + ".scalar" }));
 		addTestConfiguration("WeightedScaleTest", new TestConfiguration(
 				TEST_CLASS_DIR, "WeightedScaleTest", new String[] {
 						"mean" + ".scalar", "std" + ".scalar",
 						"se" + ".scalar", "var" + ".scalar", "cv" + ".scalar",
 						/* "har", "geom", */
 						"min" + ".scalar", "max" + ".scalar",
 						"rng" + ".scalar", "g1" + ".scalar",
 						"se_g1" + ".scalar", "g2" + ".scalar",
 						"se_g2" + ".scalar", "out_minus", "out_plus",
 						"median" + ".scalar", "quantile", "iqm" + ".scalar" }));
 		addTestConfiguration("Categorical", new TestConfiguration(TEST_CLASS_DIR,
 				"Categorical", new String[] { "Nc", "R" + ".scalar", "Pc", "C",
 						"Mode" })); // Indicate some file is scalar
 		addTestConfiguration("WeightedCategoricalTest", new TestConfiguration(
 				TEST_CLASS_DIR, "WeightedCategoricalTest", new String[] { "Nc",
 						"R" + ".scalar", "Pc", "C", "Mode" }));
 	}

 	/**
 	 * Shared test driver for tests of univariate statistics over continuous,
 	 * scaled, but not weighted, data
 	 *
 	 * @param sz
 	 *            size of primary input vector
 	 * @param rng
 	 *            range of randomly-generated values to use
 	 * @param sp
 	 *            sparsity of generated data
 	 * @param rt
 	 *            backend platform to test
 	 */
 	protected void testScaleWithR(SIZE sz, RANGE rng, SPARSITY sp, ExecMode rt) {
 		ExecMode oldrt = rtplatform;
 		rtplatform = rt;

 		try {
 			TestConfiguration config = getTestConfiguration("Scale");
 			config.addVariable("rows1", sz.size);
 			config.addVariable("rows2", rows2);
 			loadTestConfiguration(config);

 			// This is for running the junit test the new way, i.e., construct
 			// the arguments directly
 			String S_HOME = SCRIPT_DIR + TEST_DIR;
 			fullDMLScriptName = S_HOME + "Scale" + ".dml";
 			programArgs = new String[] { "-args",
 					input("vector"), Integer.toString(sz.size),
 					input("prob"), Integer.toString(rows2),
 					output("mean"), output("std"),
 					output("se"), output("var"),
 					output("cv"), output("min"),
 					output("max"), output("rng"),
 					output("g1"), output("se_g1"),
 					output("g2"), output("se_g2"),
 					output("median"),
 					output("iqm"),
 					output("out_minus"),
 					output("out_plus"),
 					output("quantile") };

 			fullRScriptName = S_HOME + "Scale" + ".R";
 			rCmd = "Rscript" + " " + fullRScriptName + " " + inputDir() + " " + expectedDir();

 			double[][] vector = getRandomMatrix(sz.size, 1, rng.min, rng.max, sp.sparsity, 20160124165501L);
 			double[][] prob = getRandomMatrix(rows2, 1, 0, 1, 1, 20160124165502L);

 			writeInputMatrix("vector", vector, true);
 			writeInputMatrix("prob", prob, true);

 			runTest(true, false, null, -1);
 			runRScript(true);

 			for (String file : config.getOutputFiles()) {
 				// NOte that some files do not contain matrix, but just a single
 				// scalar value inside
 				HashMap<CellIndex, Double> dmlfile;
 				HashMap<CellIndex, Double> rfile;
 				if (file.endsWith(".scalar")) {
 					file = file.replace(".scalar", "");
 					dmlfile = readDMLScalarFromOutputDir(file);
 					rfile = readRScalarFromExpectedDir(file);
 				} else {
 					dmlfile = readDMLMatrixFromOutputDir(file);
 					rfile = readRMatrixFromExpectedDir(file);
 				}
 				TestUtils.compareMatrices(dmlfile, rfile, epsilon, file
 						+ "-DML", file + "-R");
 			}
 		} finally {
 			rtplatform = oldrt;
 		}
 	}

 	/**
 	 * Shared test driver for tests of univariate statistics over continuous,
 	 * scaled, and weighted data
 	 *
 	 * @param sz
 	 *            size of primary input vector
 	 * @param rng
 	 *            range of randomly-generated values to use
 	 * @param sp
 	 *            sparsity of generated data
 	 * @param rt
 	 *            backend platform to test
 	 */
 	protected void testWeightedScaleWithR(SIZE sz, RANGE rng, SPARSITY sp,
 			ExecMode rt) {

 		ExecMode oldrt = rtplatform;
 		rtplatform = rt;

 		try {
 			TestConfiguration config = getTestConfiguration("WeightedScaleTest");
 			config.addVariable("rows1", sz.size);
 			config.addVariable("rows2", rows2);
 			loadTestConfiguration(config);

 			// This is for running the junit test the new way, i.e., construct
 			// the arguments directly
 			String S_HOME = SCRIPT_DIR + TEST_DIR;
 			fullDMLScriptName = S_HOME + "WeightedScaleTest" + ".dml";
 			programArgs = new String[] { "-args",
 					input("vector"), Integer.toString(sz.size),
 					input("weight"), input("prob"),
 					Integer.toString(rows2), output("mean"),
 					output("std"), output("se"),
 					output("var"), output("cv"),
 					output("min"), output("max"),
 					output("rng"), output("g1"),
 					output("se_g1"), output("g2"),
 					output("se_g2"),
 					output("median"),
 					output("iqm"),
 					output("out_minus"),
 					output("out_plus"),
 					output("quantile") };

 			fullRScriptName = S_HOME + "WeightedScaleTest" + ".R";
 			rCmd = "Rscript" + " " + fullRScriptName + " " + inputDir() + " " + expectedDir();

 			createHelperMatrix();
 			double[][] vector = getRandomMatrix(sz.size, 1, rng.min, rng.max, sp.sparsity, 20160124165503L);
 			double[][] weight = getRandomMatrix(sz.size, 1, 1, 10, 1, 20160124165504L);
 			OrderStatisticsTest.round(weight);
 			double[][] prob = getRandomMatrix(rows2, 1, 0, 1, 1, 20160124165505L);

 			writeInputMatrix("vector", vector, true);
 			writeInputMatrix("weight", weight, true);
 			writeInputMatrix("prob", prob, true);

 			runTest(true, false, null, -1);
 			runRScript(true);

 			for (String file : config.getOutputFiles()) {
 				// NOte that some files do not contain matrix, but just a single
 				// scalar value inside
 				HashMap<CellIndex, Double> dmlfile;
 				HashMap<CellIndex, Double> rfile;
 				if (file.endsWith(".scalar")) {
 					file = file.replace(".scalar", "");
 					dmlfile = readDMLScalarFromOutputDir(file);
 					rfile = readRScalarFromExpectedDir(file);
 				} else {
 					dmlfile = readDMLMatrixFromOutputDir(file);
 					rfile = readRMatrixFromExpectedDir(file);
 				}
 				TestUtils.compareMatrices(dmlfile, rfile, epsilon, file
 						+ "-DML", file + "-R");
 			}
 		} finally {
 			// reset runtime platform
 			rtplatform = oldrt;
 		}
 	}
 }
	/*
	* 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.sysds.test.applications;

	import java.util.HashMap;

	import org.apache.sysds.common.Types.ExecMode;
	import org.apache.sysds.runtime.matrix.data.MatrixValue.CellIndex;
	import org.apache.sysds.test.AutomatedTestBase;
	import org.apache.sysds.test.TestConfiguration;
	import org.apache.sysds.test.TestUtils;

	/**
	* Shared methods and fields from the different univariate stats test harness
	* classes.
	*/
	public abstract class UnivariateStatsBase extends AutomatedTestBase {

	/**
	* Hard-coded path to the directory where this test keeps all of its files,
	* generated and static
	*/
	protected final static String TEST_DIR = "applications/descriptivestats/";
	protected String TEST_CLASS_DIR = TEST_DIR + UnivariateStatsBase.class.getSimpleName() + "/";

	/** Fudge factor for comparisons against R's output */
	protected final static double epsilon = 0.000000001;

	/**
	* Number of rows in the primary input to each aggregate ("V" in the test
	* script, "X" in the SystemDS documentation)
	*/
	protected final static int rows1 = 2000;

	/**
	* Number of rows in the secondary input to aggregates that take two input
	* vectors ("P" in the test script and the documentation)
	*/
	protected final static int rows2 = 5;

	/**
	* Sizes of inputs used in tests, measured as number of rows in primary
	* input vector. DIV4=divisible by 4; DIV4P1=divisible by 4 plus 1
	*/
	protected enum SIZE {
	DIV4(2000), DIV4P1(2001), DIV4P2(2002), DIV4P3(2003);
	int size = -1;

	SIZE(int s) {
	size = s;
	}
	}

	/** Ranges of values passed to aggregates in different tests. */
	protected enum RANGE {
	NEG(-255, -2), MIXED(-200, 200), POS(2, 255);
	double min, max;

	RANGE(double mn, double mx) {
	min = mn;
	max = mx;
	}
	}

	/**
	* Actual sparsity values used in the "dense" and "sparse" variants of the
	* tests.
	*/
	protected enum SPARSITY {
	SPARSE(0.3), DENSE(0.8);
	double sparsity;

	SPARSITY(double sp) {
	sparsity = sp;
	}
	}

	/** Shared setup code for test harness configurations */
	@Override
	public void setUp() {
	TestUtils.clearAssertionInformation();

	addTestConfiguration("Scale", new TestConfiguration(TEST_CLASS_DIR, "Scale",
	new String[] { "mean" + ".scalar", "std" + ".scalar",
	"se" + ".scalar", "var" + ".scalar", "cv" + ".scalar",
	/* "har", "geom", */
	"min" + ".scalar", "max" + ".scalar",
	"rng" + ".scalar", "g1" + ".scalar",
	"se_g1" + ".scalar", "g2" + ".scalar",
	"se_g2" + ".scalar", "out_minus", "out_plus",
	"median" + ".scalar", "quantile", "iqm" + ".scalar" }));
	addTestConfiguration("WeightedScaleTest", new TestConfiguration(
	TEST_CLASS_DIR, "WeightedScaleTest", new String[] {
	"mean" + ".scalar", "std" + ".scalar",
	"se" + ".scalar", "var" + ".scalar", "cv" + ".scalar",
	/* "har", "geom", */
	"min" + ".scalar", "max" + ".scalar",
	"rng" + ".scalar", "g1" + ".scalar",
	"se_g1" + ".scalar", "g2" + ".scalar",
	"se_g2" + ".scalar", "out_minus", "out_plus",
	"median" + ".scalar", "quantile", "iqm" + ".scalar" }));
	addTestConfiguration("Categorical", new TestConfiguration(TEST_CLASS_DIR,
	"Categorical", new String[] { "Nc", "R" + ".scalar", "Pc", "C",
	"Mode" })); // Indicate some file is scalar
	addTestConfiguration("WeightedCategoricalTest", new TestConfiguration(
	TEST_CLASS_DIR, "WeightedCategoricalTest", new String[] { "Nc",
	"R" + ".scalar", "Pc", "C", "Mode" }));
	}

	/**
	* Shared test driver for tests of univariate statistics over continuous,
	* scaled, but not weighted, data
	*
	* @param sz
	* size of primary input vector
	* @param rng
	* range of randomly-generated values to use
	* @param sp
	* sparsity of generated data
	* @param rt
	* backend platform to test
	*/
	protected void testScaleWithR(SIZE sz, RANGE rng, SPARSITY sp, ExecMode rt) {
	ExecMode oldrt = rtplatform;
	rtplatform = rt;

	try {
	TestConfiguration config = getTestConfiguration("Scale");
	config.addVariable("rows1", sz.size);
	config.addVariable("rows2", rows2);
	loadTestConfiguration(config);

	// This is for running the junit test the new way, i.e., construct
	// the arguments directly
	String S_HOME = SCRIPT_DIR + TEST_DIR;
	fullDMLScriptName = S_HOME + "Scale" + ".dml";
	programArgs = new String[] { "-args",
	input("vector"), Integer.toString(sz.size),
	input("prob"), Integer.toString(rows2),
	output("mean"), output("std"),
	output("se"), output("var"),
	output("cv"), output("min"),
	output("max"), output("rng"),
	output("g1"), output("se_g1"),
	output("g2"), output("se_g2"),
	output("median"),
	output("iqm"),
	output("out_minus"),
	output("out_plus"),
	output("quantile") };

	fullRScriptName = S_HOME + "Scale" + ".R";
	rCmd = "Rscript" + " " + fullRScriptName + " " + inputDir() + " " + expectedDir();

	double[][] vector = getRandomMatrix(sz.size, 1, rng.min, rng.max, sp.sparsity, 20160124165501L);
	double[][] prob = getRandomMatrix(rows2, 1, 0, 1, 1, 20160124165502L);

	writeInputMatrix("vector", vector, true);
	writeInputMatrix("prob", prob, true);

	runTest(true, false, null, -1);
	runRScript(true);

	for (String file : config.getOutputFiles()) {
	// NOte that some files do not contain matrix, but just a single
	// scalar value inside
	HashMap<CellIndex, Double> dmlfile;
	HashMap<CellIndex, Double> rfile;
	if (file.endsWith(".scalar")) {
	file = file.replace(".scalar", "");
	dmlfile = readDMLScalarFromOutputDir(file);
	rfile = readRScalarFromExpectedDir(file);
	} else {
	dmlfile = readDMLMatrixFromOutputDir(file);
	rfile = readRMatrixFromExpectedDir(file);
	}
	TestUtils.compareMatrices(dmlfile, rfile, epsilon, file
	+ "-DML", file + "-R");
	}
	} finally {
	rtplatform = oldrt;
	}
	}

	/**
	* Shared test driver for tests of univariate statistics over continuous,
	* scaled, and weighted data
	*
	* @param sz
	* size of primary input vector
	* @param rng
	* range of randomly-generated values to use
	* @param sp
	* sparsity of generated data
	* @param rt
	* backend platform to test
	*/
	protected void testWeightedScaleWithR(SIZE sz, RANGE rng, SPARSITY sp,
	ExecMode rt) {

	ExecMode oldrt = rtplatform;
	rtplatform = rt;

	try {
	TestConfiguration config = getTestConfiguration("WeightedScaleTest");
	config.addVariable("rows1", sz.size);
	config.addVariable("rows2", rows2);
	loadTestConfiguration(config);

	// This is for running the junit test the new way, i.e., construct
	// the arguments directly
	String S_HOME = SCRIPT_DIR + TEST_DIR;
	fullDMLScriptName = S_HOME + "WeightedScaleTest" + ".dml";
	programArgs = new String[] { "-args",
	input("vector"), Integer.toString(sz.size),
	input("weight"), input("prob"),
	Integer.toString(rows2), output("mean"),
	output("std"), output("se"),
	output("var"), output("cv"),
	output("min"), output("max"),
	output("rng"), output("g1"),
	output("se_g1"), output("g2"),
	output("se_g2"),
	output("median"),
	output("iqm"),
	output("out_minus"),
	output("out_plus"),
	output("quantile") };

	fullRScriptName = S_HOME + "WeightedScaleTest" + ".R";
	rCmd = "Rscript" + " " + fullRScriptName + " " + inputDir() + " " + expectedDir();

	createHelperMatrix();
	double[][] vector = getRandomMatrix(sz.size, 1, rng.min, rng.max, sp.sparsity, 20160124165503L);
	double[][] weight = getRandomMatrix(sz.size, 1, 1, 10, 1, 20160124165504L);
	OrderStatisticsTest.round(weight);
	double[][] prob = getRandomMatrix(rows2, 1, 0, 1, 1, 20160124165505L);

	writeInputMatrix("vector", vector, true);
	writeInputMatrix("weight", weight, true);
	writeInputMatrix("prob", prob, true);

	runTest(true, false, null, -1);
	runRScript(true);

	for (String file : config.getOutputFiles()) {
	// NOte that some files do not contain matrix, but just a single
	// scalar value inside
	HashMap<CellIndex, Double> dmlfile;
	HashMap<CellIndex, Double> rfile;
	if (file.endsWith(".scalar")) {
	file = file.replace(".scalar", "");
	dmlfile = readDMLScalarFromOutputDir(file);
	rfile = readRScalarFromExpectedDir(file);
	} else {
	dmlfile = readDMLMatrixFromOutputDir(file);
	rfile = readRMatrixFromExpectedDir(file);
	}
	TestUtils.compareMatrices(dmlfile, rfile, epsilon, file
	+ "-DML", file + "-R");
	}
	} finally {
	// reset runtime platform
	rtplatform = oldrt;
	}
	}
	}