Baunsgaard commented on a change in pull request #923:
URL: https://github.com/apache/systemml/pull/923#discussion_r441385213
##########
File path:
src/main/java/org/apache/sysds/runtime/instructions/spark/BinarySPInstruction.java
##########
@@ -106,7 +108,9 @@ else if (dt1 == DataType.TENSOR || dt2 == DataType.TENSOR) {
throw new DMLRuntimeException("Tensor binary
operation not yet implemented for tensor-scalar, or tensor-matrix");
}
else if( dt1 == DataType.FRAME || dt2 == DataType.FRAME ) {
- return
BinaryFrameFrameSPInstruction.parseInstruction(str);
+ if(dt1 == DataType.FRAME && dt2 == DataType.FRAME)
+ return
BinaryFrameFrameSPInstruction.parseInstruction(str);
+
Review comment:
Double check. first if (X || Y) then if (X && Y) could be reduced to (X
&& Y)
##########
File path: src/main/java/org/apache/sysds/parser/BuiltinFunctionExpression.java
##########
@@ -1529,7 +1529,16 @@ else if(this.getOpCode() == Builtins.MAX_POOL ||
this.getOpCode() == Builtins.AV
output.setValueType(ValueType.STRING);
break;
- default:
+ case DROP_INVALID_LENGTH:
+ checkNumParameters(2);
+ checkMatrixFrameParam(getFirstExpr());
+ checkMatrixFrameParam(getSecondExpr());
+ output.setDataType(DataType.FRAME);
+ output.setDimensions(id.getDim1(), id.getDim2());
+ output.setBlocksize (id.getBlocksize());
+ output.setValueType(id.getValueType());
+ break;
+ default:
Review comment:
indentation
##########
File path: src/main/java/org/apache/sysds/common/Types.java
##########
@@ -264,7 +264,7 @@ public static OpOp1 valueOfByOpcode(String opcode) {
public enum OpOp2 {
AND(true), BITWAND(true), BITWOR(true), BITWSHIFTL(true),
BITWSHIFTR(true),
BITWXOR(true), CBIND(false), CONCAT(false), COV(false),
DIV(true),
- DROP_INVALID(false), EQUAL(true), GREATER(true),
GREATEREQUAL(true),
+ DROP_INVALID_TYPE(false), DROP_INVALID_LENGTH(false),
EQUAL(true), GREATER(true), GREATEREQUAL(true),
Review comment:
oddly long line
##########
File path: src/main/java/org/apache/sysds/runtime/matrix/data/FrameBlock.java
##########
@@ -1865,6 +1870,37 @@ else if
(!dataType.toString().equals(schemaString[i].trim()))
return this;
}
+ /*
+ This method validates the frame data against an attribute
length constrain
+ if data value in any cell is greater than the specified
threshold of that attribute
+ the output frame will store a null on that cell position, thus
removing the length-violating values.
+ @param input 1) row vector of valid lengths
+ @param output FrameBlock with invalid values converted into
missing values (null)
+ */
Review comment:
Also not param for return object
##########
File path:
src/test/java/org/apache/sysds/test/functions/frame/FrameDropInvalidLengthTest.java
##########
@@ -0,0 +1,227 @@
+/*
+ * 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.functions.frame;
+
+import org.apache.sysds.api.DMLScript;
+import org.apache.sysds.common.Types;
+import org.apache.sysds.hops.OptimizerUtils;
+import org.apache.sysds.lops.LopProperties;
+import org.apache.sysds.runtime.io.FrameWriter;
+import org.apache.sysds.runtime.io.FrameWriterFactory;
+import org.apache.sysds.runtime.matrix.data.FrameBlock;
+import org.apache.sysds.runtime.util.UtilFunctions;
+import org.apache.sysds.test.AutomatedTestBase;
+import org.apache.sysds.test.TestConfiguration;
+import org.apache.sysds.test.TestUtils;
+import org.junit.Assert;
+import org.junit.Test;
+
+import java.util.ArrayList;
+import java.util.concurrent.ThreadLocalRandom;
+import java.util.stream.Collectors;
+
+public class FrameDropInvalidLengthTest extends AutomatedTestBase {
+ private final static String TEST_NAME = "DropInvalidLength";
+ private final static String TEST_DIR = "functions/frame/";
+ private static final String TEST_CLASS_DIR = TEST_DIR +
FrameDropInvalidLengthTest.class.getSimpleName() + "/";
+
+ private final static int rows = 800;
+ private final static int cols = 4;
+ private final static Types.ValueType[] schemaStrings =
{Types.ValueType.FP64, Types.ValueType.STRING, Types.ValueType.STRING,
Types.ValueType.INT64};
+
+ public static void init() {
+ TestUtils.clearDirectory(TEST_DATA_DIR + TEST_CLASS_DIR);
+ }
+
+ public static void cleanUp() {
+ if (TEST_CACHE_ENABLED) {
+ TestUtils.clearDirectory(TEST_DATA_DIR +
TEST_CLASS_DIR);
+ }
+ }
+
+ @Override
+ public void setUp() {
+ TestUtils.clearAssertionInformation();
+ addTestConfiguration(TEST_NAME, new
TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[] {"B"}));
+ if (TEST_CACHE_ENABLED) {
+ setOutAndExpectedDeletionDisabled(true);
+ }
+ }
+
+ @Test
+ public void testTwoBadColCP() {
+ double[][] invalidLength = {{-1,30,20,-1}};
+ runDropInvalidLenTest( invalidLength,1,
LopProperties.ExecType.CP);
+ }
+
+ @Test
+ public void testTwoBadColSP() {
+ double[][] invalidLength = {{-1,30,20,-1}};
+ runDropInvalidLenTest( invalidLength,1,
LopProperties.ExecType.SPARK);
+ }
+
+ @Test
+ public void testOneBadColCP() {
+ double[][] invalidLength = {{-1,-1,20,-1}};
+ runDropInvalidLenTest( invalidLength,2,
LopProperties.ExecType.CP);
+ }
+
+ @Test
+ public void testOneBadColSP() {
+ double[][] invalidLength = {{-1,-1,20,-1}};
+ runDropInvalidLenTest( invalidLength,2,
LopProperties.ExecType.SPARK);
+ }
+
+ @Test
+ public void testAllBadColCP() {
+ double[][] invalidLength = {{2,2,2,1}};
+ runDropInvalidLenTest( invalidLength,3,
LopProperties.ExecType.CP);
+ }
+
+ @Test
+ public void testAllBadColSP() {
+ double[][] invalidLength = {{2,2,2,1}};
+ runDropInvalidLenTest( invalidLength,3,
LopProperties.ExecType.SPARK);
+ }
+
+ @Test
+ public void testNoneBadColCP() {
+ double[][] invalidLength = {{-1,20,20,-1}};
+ runDropInvalidLenTest( invalidLength,4,
LopProperties.ExecType.CP);
+ }
+
+ @Test
+ public void testNoneBadColSP() {
+ double[][] invalidLength = {{-1,20,20,-1}};
+ runDropInvalidLenTest( invalidLength,4,
LopProperties.ExecType.SPARK);
+ }
+
+ private void runDropInvalidLenTest(double[][] colInvalidLength, int
test, LopProperties.ExecType et)
+ {
+ Types.ExecMode platformOld = setExecMode(et);
+ boolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;
+ boolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;
+ try {
+ getAndLoadTestConfiguration(TEST_NAME);
+ String HOME = SCRIPT_DIR + TEST_DIR;
+ fullDMLScriptName = HOME + TEST_NAME + ".dml";
+ programArgs = new String[] {"-args", input("A"),
input("M"),
+ String.valueOf(rows),
Integer.toString(cols), output("B")};
+ FrameBlock frame1 = new FrameBlock(schemaStrings);
+ double[][] A = getRandomMatrix(rows, cols, 10, 100, 1,
2373);
+ initFrameDataString(frame1,A, schemaStrings); //
initialize a frame with one column
+ FrameWriter writer =
FrameWriterFactory.createFrameWriter(Types.FileFormat.CSV);
+
+ ArrayList<Integer> badIndex = getBadIndexes(rows/4);
+ int expected = 0;
+
+ switch (test) { //Double in String
+ case 1:
+ for (int i = 0; i < badIndex.size();
i++) {
+
frame1.set(badIndex.get(i),1,"This is a very long sentence that could" +
+ " count up to
multiple characters");
+ }
+ expected += badIndex.size();
+ case 2:
+ for (int i = 0; i < badIndex.size();
i++) {
+ frame1.set(badIndex.get(i), 2,
"This is out of length");
+ }
+ expected += badIndex.size();
+ break;
+ case 3:
+ expected += rows*cols;
+ break;
+ case 4:
+ expected += 0.0;
+ break;
+ }
+ // write data frame
+ writer.writeFrameToHDFS(
+ frame1.slice(0, rows - 1, 0, cols-1,
new FrameBlock()),
+ input("A"), rows, schemaStrings.length);
+
+ // write expected feature length vector
+ writeInputMatrixWithMTD("M", colInvalidLength, true);
+
+ runTest(true, false, null, -1);
+
+ // compare output
+ FrameBlock frameOut = readDMLFrameFromHDFS("B",
Types.FileFormat.BINARY);
+
+ //read output data and compare results
+ ArrayList<Object> data = new ArrayList<>();
+ for (int i = 0; i < frameOut.getNumRows(); i++)
+ for(int j=0; j < frameOut.getNumColumns(); j++ )
+ data.add(frameOut.get(i, j));
+
+ int nullNum = Math.toIntExact(data.stream().filter(s ->
s == null).count());
+ Assert.assertEquals(expected, nullNum, 1e-5);
+ }
+ catch (Exception ex) {
+ throw new RuntimeException(ex);
+ }
+ finally {
+ rtplatform = platformOld;
+ DMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;
+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldFlag;
+ OptimizerUtils.ALLOW_AUTO_VECTORIZATION = true;
+ OptimizerUtils.ALLOW_OPERATOR_FUSION = true;
+ }
+ }
+
+ private ArrayList<Integer> getBadIndexes(int length) {
+ ArrayList<Integer> list = new ArrayList();
+ for(int i =0; i<length; i++)
+ {
+ int r = ThreadLocalRandom.current().nextInt(0, rows);
+ list.add(r);
+ }
+ return (ArrayList)
list.stream().distinct().collect(Collectors.toList());
+ }
+
+ public static void initFrameDataString(FrameBlock frame1, double[][]
data, Types.ValueType[] lschema) {
+ for (int j = 0; j < lschema.length; j++) {
+ Types.ValueType vt = lschema[j];
+ switch (vt) {
+ case STRING:
+ String[] tmp1 = new String[rows];
+ for (int i = 0; i < rows; i++)
+ tmp1[i] = (String)
UtilFunctions.doubleToObject(vt, data[i][j]);
+ frame1.appendColumn(tmp1);
+ break;
+ case INT64:
+ long[] tmp4 = new long[rows];
+ for (int i = 0; i < rows; i++)
+ data[i][j] = tmp4[i] = (Long)
UtilFunctions.doubleToObject(Types.ValueType.INT64,
+ data[i][j],
false);
+ frame1.appendColumn(tmp4);
+ break;
+ case FP64:
+ double[] tmp6 = new double[rows];
+ for (int i = 0; i < rows; i++)
+ tmp6[i] = (Double)
UtilFunctions.doubleToObject(vt, data[i][j], false);
+ frame1.appendColumn(tmp6);
+ break;
+ default:
+ throw new RuntimeException("Unsupported
value type: " + vt);
+ }
+ }
+ }
+}
Review comment:
Just because of git, i like to add a newline in the end.
##########
File path: src/main/java/org/apache/sysds/runtime/matrix/data/FrameBlock.java
##########
@@ -1830,6 +1830,11 @@ else if (isType(dataValue) == ValueType.STRING)
return fb;
}
+ /**
+ * Drop the cell value which does not confirms to the data type of its
column
+ * @input schema of the frame
Review comment:
`@input` not out standard, use param.
##########
File path: src/test/scripts/functions/frame/DropInvalidLength.dml
##########
@@ -0,0 +1,27 @@
+#-------------------------------------------------------------
+#
+# 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.
+#
+#-------------------------------------------------------------
+
+X = read($1, rows=$3, cols=$4, data_type="frame", format="csv");
+colLength = read($2); # row vector with -1 for exempted features and
+ # a valid character length for features to be processed
+R = dropInvalidLength(X, colLength);
+# print(toString(R))
+write(R, $5, format="binary");
Review comment:
again just because of git, add a newline
##########
File path: src/main/java/org/apache/sysds/runtime/matrix/data/FrameBlock.java
##########
@@ -1865,6 +1870,37 @@ else if
(!dataType.toString().equals(schemaString[i].trim()))
return this;
}
+ /*
+ This method validates the frame data against an attribute
length constrain
+ if data value in any cell is greater than the specified
threshold of that attribute
+ the output frame will store a null on that cell position, thus
removing the length-violating values.
+ @param input 1) row vector of valid lengths
+ @param output FrameBlock with invalid values converted into
missing values (null)
+ */
Review comment:
Unorthodox jdoc syntax should be:
```java
/**
* text goes here
* @param
* @return
*/
```
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
