You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by GitBox <gi...@apache.org> on 2022/01/05 09:50:33 UTC
[GitHub] [flink-ml] zhipeng93 opened a new pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer in FlinkML
zhipeng93 opened a new pull request #52:
URL: https://github.com/apache/flink-ml/pull/52
## What is the purpose of the change
- Add Transformer and Estimator of StringIndexer in FlinkML.
## Brief change log
- Add StringIndexer.
- Add StringIndexerModel.
- Add IndexToStringModel.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (no)
- The public API, i.e., is any changed class annotated with @Public(Evolving): (no)
- Does this pull request introduce a new feature? (no)
- If yes, how is the feature documented? (Java doc)
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831845494
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
Review comment:
If there is only one partition, the result is not `random`.
After reading [1], we agree that `arbitrary` is better here.
[1] http://johnthedebs.blogspot.com/2008/01/arbitrary-vs-random.html
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831835128
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,204 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerParams<StringIndexerModel> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+ private Table modelDataTable;
+
+ public StringIndexerModel() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ ReadWriteUtils.saveModelData(
+ StringIndexerModelData.getModelDataStream(modelDataTable),
+ path,
+ new StringIndexerModelData.ModelDataEncoder());
+ }
+
+ public static StringIndexerModel load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ StreamTableEnvironment tEnv = StreamTableEnvironment.create(env);
+ StringIndexerModel model = ReadWriteUtils.loadStageParam(path);
+ DataStream<StringIndexerModelData> modelData =
+ ReadWriteUtils.loadModelData(
+ env, path, new StringIndexerModelData.ModelDataDecoder());
+ return model.setModelData(tEnv.fromDataStream(modelData));
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel setModelData(Table... inputs) {
+ modelDataTable = inputs[0];
+ return this;
+ }
+
+ @Override
+ public Table[] getModelData() {
+ return new Table[] {modelDataTable};
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) modelDataTable).getTableEnvironment();
+
+ final String broadcastModelKey = "broadcastModelKey";
+ DataStream<StringIndexerModelData> modelDataStream =
+ StringIndexerModelData.getModelDataStream(modelDataTable);
+ RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema());
+ TypeInformation<?>[] outputTypes = new TypeInformation[outputCols.length];
+ Arrays.fill(outputTypes, BasicTypeInfo.INT_TYPE_INFO);
+ RowTypeInfo outputTypeInfo =
+ new RowTypeInfo(
+ ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), outputTypes),
+ ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCols()));
+ DataStream<Row> result =
+ BroadcastUtils.withBroadcastStream(
+ Collections.singletonList(tEnv.toDataStream(inputs[0])),
+ Collections.singletonMap(broadcastModelKey, modelDataStream),
+ inputList -> {
+ DataStream inputData = inputList.get(0);
+ return inputData.flatMap(
+ new String2Index(
+ broadcastModelKey, inputCols, getHandleInvalid()),
+ outputTypeInfo);
+ });
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Maps the input columns to integer values according to the model data. */
+ private static class String2Index extends RichFlatMapFunction<Row, Row> {
+ private HashMap<String, Integer>[] modelDataMap;
+ private final String broadcastModelKey;
+ private final String[] inputCols;
+ private final String handleInValid;
+
+ public String2Index(String broadcastModelKey, String[] inputCols, String handleInValid) {
+ this.broadcastModelKey = broadcastModelKey;
+ this.inputCols = inputCols;
+ this.handleInValid = handleInValid;
+ }
+
+ @Override
+ public void flatMap(Row input, Collector<Row> out) throws Exception {
+ if (modelDataMap == null) {
+ modelDataMap = new HashMap[inputCols.length];
+ StringIndexerModelData modelData =
+ (StringIndexerModelData)
+ getRuntimeContext().getBroadcastVariable(broadcastModelKey).get(0);
+ String[][] stringsArray = modelData.stringsArray;
+ for (int i = 0; i < stringsArray.length; i++) {
+ int idx = 0;
+ modelDataMap[i] = new HashMap<>(stringsArray[i].length);
+ for (String string : stringsArray[i]) {
+ modelDataMap[i].put(string, idx++);
+ }
+ }
+ }
+ Row outputIndices = new Row(inputCols.length);
+ for (int i = 0; i < inputCols.length; i++) {
+ Object objVal = input.getField(inputCols[i]);
+ String stringVal;
+ if (objVal instanceof String) {
+ stringVal = (String) objVal;
+ } else if (objVal instanceof Number) {
+ stringVal = String.valueOf(objVal);
+ } else {
+ throw new RuntimeException(
+ "The input column only supports string type and numeric type.");
+ }
+ if (modelDataMap[i].containsKey(stringVal)) {
+ outputIndices.setField(i, modelDataMap[i].get(stringVal));
+ } else {
+ switch (handleInValid) {
+ case StringIndexerModelParams.SKIP_INVALID:
+ return;
+ case StringIndexerModelParams.ERROR_INVALID:
+ throw new RuntimeException(
Review comment:
I mean the default branch. It is fixed now.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 merged pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 merged pull request #52:
URL: https://github.com/apache/flink-ml/pull/52
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831810341
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,204 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerParams<StringIndexerModel> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+ private Table modelDataTable;
+
+ public StringIndexerModel() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ ReadWriteUtils.saveModelData(
+ StringIndexerModelData.getModelDataStream(modelDataTable),
+ path,
+ new StringIndexerModelData.ModelDataEncoder());
+ }
+
+ public static StringIndexerModel load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ StreamTableEnvironment tEnv = StreamTableEnvironment.create(env);
+ StringIndexerModel model = ReadWriteUtils.loadStageParam(path);
+ DataStream<StringIndexerModelData> modelData =
+ ReadWriteUtils.loadModelData(
+ env, path, new StringIndexerModelData.ModelDataDecoder());
+ return model.setModelData(tEnv.fromDataStream(modelData));
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel setModelData(Table... inputs) {
+ modelDataTable = inputs[0];
+ return this;
+ }
+
+ @Override
+ public Table[] getModelData() {
+ return new Table[] {modelDataTable};
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) modelDataTable).getTableEnvironment();
+
+ final String broadcastModelKey = "broadcastModelKey";
+ DataStream<StringIndexerModelData> modelDataStream =
+ StringIndexerModelData.getModelDataStream(modelDataTable);
+ RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema());
+ TypeInformation<?>[] outputTypes = new TypeInformation[outputCols.length];
+ Arrays.fill(outputTypes, BasicTypeInfo.INT_TYPE_INFO);
+ RowTypeInfo outputTypeInfo =
+ new RowTypeInfo(
+ ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), outputTypes),
+ ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCols()));
+ DataStream<Row> result =
+ BroadcastUtils.withBroadcastStream(
+ Collections.singletonList(tEnv.toDataStream(inputs[0])),
+ Collections.singletonMap(broadcastModelKey, modelDataStream),
+ inputList -> {
+ DataStream inputData = inputList.get(0);
+ return inputData.flatMap(
+ new String2Index(
+ broadcastModelKey, inputCols, getHandleInvalid()),
+ outputTypeInfo);
+ });
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Maps the input columns to integer values according to the model data. */
+ private static class String2Index extends RichFlatMapFunction<Row, Row> {
+ private HashMap<String, Integer>[] modelDataMap;
+ private final String broadcastModelKey;
+ private final String[] inputCols;
+ private final String handleInValid;
+
+ public String2Index(String broadcastModelKey, String[] inputCols, String handleInValid) {
+ this.broadcastModelKey = broadcastModelKey;
+ this.inputCols = inputCols;
+ this.handleInValid = handleInValid;
+ }
+
+ @Override
+ public void flatMap(Row input, Collector<Row> out) throws Exception {
+ if (modelDataMap == null) {
+ modelDataMap = new HashMap[inputCols.length];
+ StringIndexerModelData modelData =
+ (StringIndexerModelData)
+ getRuntimeContext().getBroadcastVariable(broadcastModelKey).get(0);
+ String[][] stringsArray = modelData.stringsArray;
+ for (int i = 0; i < stringsArray.length; i++) {
+ int idx = 0;
+ modelDataMap[i] = new HashMap<>(stringsArray[i].length);
+ for (String string : stringsArray[i]) {
+ modelDataMap[i].put(string, idx++);
+ }
+ }
+ }
+ Row outputIndices = new Row(inputCols.length);
+ for (int i = 0; i < inputCols.length; i++) {
+ Object objVal = input.getField(inputCols[i]);
+ String stringVal;
+ if (objVal instanceof String) {
+ stringVal = (String) objVal;
+ } else if (objVal instanceof Number) {
+ stringVal = String.valueOf(objVal);
+ } else {
+ throw new RuntimeException(
+ "The input column only supports string type and numeric type.");
+ }
+ if (modelDataMap[i].containsKey(stringVal)) {
+ outputIndices.setField(i, modelDataMap[i].get(stringVal));
+ } else {
+ switch (handleInValid) {
+ case StringIndexerModelParams.SKIP_INVALID:
+ return;
+ case StringIndexerModelParams.ERROR_INVALID:
+ throw new RuntimeException(
Review comment:
Do you mean the default branch or the `ERROR_INVALID` branch?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831815876
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/StringIndexerTest.java
##########
@@ -0,0 +1,319 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.TestBaseUtils;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link StringIndexer} and {@link StringIndexerModel}. */
+public class StringIndexerTest {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table trainTable;
+ private Table predictTable;
+
+ private String[][] expectedAlphabeticAscModelData;
+ private List<Row> expectedAlphabeticAscPredictData;
+ private List<Row> expectedAlphabeticDescPredictData;
+ private List<Row> expectedFreqAscPredictData;
+ private List<Row> expectedFreqDescPredictData;
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ List<Row> trainData =
+ Arrays.asList(
+ Row.of("a", 1.),
+ Row.of("b", 1.),
+ Row.of("b", 2.0),
+ Row.of("c", 0.0),
+ Row.of("d", 2.),
+ Row.of("a", 2.),
+ Row.of("b", 2.),
+ Row.of("b", -1.),
+ Row.of("a", -1.),
+ Row.of("c", -1.));
+ trainTable =
+ tEnv.fromDataStream(env.fromCollection(trainData)).as("inputCol1", "inputCol2");
+
+ List<Row> predictData = Arrays.asList(Row.of("a", 2.), Row.of("b", 1.), Row.of("e", 2.));
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(predictData)).as("inputCol1", "inputCol2");
+
+ expectedAlphabeticAscModelData =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+ expectedAlphabeticAscPredictData =
+ Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2), Row.of("e", 2.0, 4, 3));
+ expectedAlphabeticDescPredictData =
+ Arrays.asList(Row.of("a", 2., 3, 0), Row.of("b", 1., 2, 1), Row.of("e", 2.0, 4, 0));
+ expectedFreqAscPredictData =
+ Arrays.asList(Row.of("a", 2., 2, 3), Row.of("b", 1., 3, 1), Row.of("e", 2.0, 4, 3));
+ expectedFreqDescPredictData =
+ Arrays.asList(Row.of("a", 2., 1, 0), Row.of("b", 1., 0, 2), Row.of("e", 2.0, 4, 0));
+ }
+
+ @Test
+ public void testFitParam() {
+ StringIndexer stringIndexer = new StringIndexer();
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.RANDOM_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.ERROR_INVALID);
+
+ stringIndexer
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol1")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+
+ assertArrayEquals(new String[] {"inputCol1", "inputCol2"}, stringIndexer.getInputCols());
+ assertArrayEquals(new String[] {"outputCol1", "outputCol1"}, stringIndexer.getOutputCols());
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.ALPHABET_ASC_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.SKIP_INVALID);
+ }
+
+ @Test
+ public void testPredictParam() {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testStringOrderType() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+
+ Table output;
+ List<Row> predictedResult;
+
+ // alphabetAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // alphabetDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticDescPredictData, predictedResult);
+
+ // frequencyAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqAscPredictData, predictedResult);
+
+ // frequencyDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqDescPredictData, predictedResult);
+
+ // random
+ stringIndexer.setStringOrderType(StringIndexerParams.RANDOM_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ Set<Integer> distinctStringsCol1 = new HashSet<>();
+ Set<Integer> distinctStringsCol2 = new HashSet<>();
+ int index;
+ for (Row r : predictedResult) {
+ index = (Integer) r.getField(2);
+ distinctStringsCol1.add(index);
+ assertTrue(index >= 0 && index <= 4);
+ index = (Integer) r.getField(3);
+ assertTrue(index >= 0 && index <= 3);
+ distinctStringsCol2.add(index);
+ }
+ assertEquals(3, distinctStringsCol1.size());
+ assertEquals(2, distinctStringsCol2.size());
+ }
+
+ @Test
+ public void testHandleInvalid() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+
+ Table output;
+ List<Row> expectedResult;
+
+ // keep invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // skip invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ expectedResult = Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2));
+ verifyPredictionResult(expectedResult, predictedResult);
+
+ // throw exception on invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.ERROR_INVALID);
+ try {
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen string: e. "
+ + "To handle unseen strings, set Param handleInvalid to keep.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ public void testFitAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testSaveLoadAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ stringIndexer =
+ StageTestUtils.saveAndReload(
+ env, stringIndexer, tempFolder.newFolder().getAbsolutePath());
+ StringIndexerModel model = stringIndexer.fit(trainTable);
+ model = StageTestUtils.saveAndReload(env, model, tempFolder.newFolder().getAbsolutePath());
+ assertEquals(
+ Collections.singletonList("stringsArray"),
+ model.getModelData()[0].getResolvedSchema().getColumnNames());
+ Table output = model.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testGetModelData() throws Exception {
+ StringIndexerModel model =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .fit(trainTable);
+ StringIndexerModelData modelData =
+ StringIndexerModelData.getModelDataStream(model.getModelData()[0])
Review comment:
Thanks for the comment. The check for schema is added in StringIndexerTest#testGetModelData.
It is not added in `IndexToStringModelTest#testGetModelData()` because the model data is generated by `StringIndexer`. So I believe that the check for schema is already covered by StringIndexerTest#testGetModelData.
What do you think?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#issuecomment-1075102827
Thanks for the reviews @lindong28 . I have addressed your comments in the latest reviews :)
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831810341
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,204 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerParams<StringIndexerModel> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+ private Table modelDataTable;
+
+ public StringIndexerModel() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ ReadWriteUtils.saveModelData(
+ StringIndexerModelData.getModelDataStream(modelDataTable),
+ path,
+ new StringIndexerModelData.ModelDataEncoder());
+ }
+
+ public static StringIndexerModel load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ StreamTableEnvironment tEnv = StreamTableEnvironment.create(env);
+ StringIndexerModel model = ReadWriteUtils.loadStageParam(path);
+ DataStream<StringIndexerModelData> modelData =
+ ReadWriteUtils.loadModelData(
+ env, path, new StringIndexerModelData.ModelDataDecoder());
+ return model.setModelData(tEnv.fromDataStream(modelData));
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel setModelData(Table... inputs) {
+ modelDataTable = inputs[0];
+ return this;
+ }
+
+ @Override
+ public Table[] getModelData() {
+ return new Table[] {modelDataTable};
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) modelDataTable).getTableEnvironment();
+
+ final String broadcastModelKey = "broadcastModelKey";
+ DataStream<StringIndexerModelData> modelDataStream =
+ StringIndexerModelData.getModelDataStream(modelDataTable);
+ RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema());
+ TypeInformation<?>[] outputTypes = new TypeInformation[outputCols.length];
+ Arrays.fill(outputTypes, BasicTypeInfo.INT_TYPE_INFO);
+ RowTypeInfo outputTypeInfo =
+ new RowTypeInfo(
+ ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), outputTypes),
+ ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCols()));
+ DataStream<Row> result =
+ BroadcastUtils.withBroadcastStream(
+ Collections.singletonList(tEnv.toDataStream(inputs[0])),
+ Collections.singletonMap(broadcastModelKey, modelDataStream),
+ inputList -> {
+ DataStream inputData = inputList.get(0);
+ return inputData.flatMap(
+ new String2Index(
+ broadcastModelKey, inputCols, getHandleInvalid()),
+ outputTypeInfo);
+ });
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Maps the input columns to integer values according to the model data. */
+ private static class String2Index extends RichFlatMapFunction<Row, Row> {
+ private HashMap<String, Integer>[] modelDataMap;
+ private final String broadcastModelKey;
+ private final String[] inputCols;
+ private final String handleInValid;
+
+ public String2Index(String broadcastModelKey, String[] inputCols, String handleInValid) {
+ this.broadcastModelKey = broadcastModelKey;
+ this.inputCols = inputCols;
+ this.handleInValid = handleInValid;
+ }
+
+ @Override
+ public void flatMap(Row input, Collector<Row> out) throws Exception {
+ if (modelDataMap == null) {
+ modelDataMap = new HashMap[inputCols.length];
+ StringIndexerModelData modelData =
+ (StringIndexerModelData)
+ getRuntimeContext().getBroadcastVariable(broadcastModelKey).get(0);
+ String[][] stringsArray = modelData.stringsArray;
+ for (int i = 0; i < stringsArray.length; i++) {
+ int idx = 0;
+ modelDataMap[i] = new HashMap<>(stringsArray[i].length);
+ for (String string : stringsArray[i]) {
+ modelDataMap[i].put(string, idx++);
+ }
+ }
+ }
+ Row outputIndices = new Row(inputCols.length);
+ for (int i = 0; i < inputCols.length; i++) {
+ Object objVal = input.getField(inputCols[i]);
+ String stringVal;
+ if (objVal instanceof String) {
+ stringVal = (String) objVal;
+ } else if (objVal instanceof Number) {
+ stringVal = String.valueOf(objVal);
+ } else {
+ throw new RuntimeException(
+ "The input column only supports string type and numeric type.");
+ }
+ if (modelDataMap[i].containsKey(stringVal)) {
+ outputIndices.setField(i, modelDataMap[i].get(stringVal));
+ } else {
+ switch (handleInValid) {
+ case StringIndexerModelParams.SKIP_INVALID:
+ return;
+ case StringIndexerModelParams.ERROR_INVALID:
+ throw new RuntimeException(
Review comment:
Do you mean the default branch or the `ERROR_INVALID` branch? The default branch is updated.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831799753
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
+ * the same. The indices are in [0, numDistinctValuesThisColumn).
Review comment:
Thanks for pointing this out. The java doc is updated.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 edited a comment on pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 edited a comment on pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#issuecomment-1075102827
Thanks for the reviews @lindong28 . I have addressed your comments in the latest commit :)
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r832739713
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/IndexToStringModelTest.java
##########
@@ -0,0 +1,165 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.AbstractTestBase;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.List;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link IndexToStringModel}. */
+public class IndexToStringModelTest extends AbstractTestBase {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table predictTable;
+ private Table modelTable;
+ private Table predictTableWithUnseenValues;
+
+ private final List<Row> expectedPrediction =
+ Arrays.asList(Row.of(0, 3, "a", "2.0"), Row.of(1, 2, "b", "1.0"));
+ private final String[][] stringArrays =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ modelTable =
+ tEnv.fromDataStream(env.fromElements(new StringIndexerModelData(stringArrays)))
+ .as("stringArrays");
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(Arrays.asList(Row.of(0, 3), Row.of(1, 2))))
+ .as("inputCol1", "inputCol2");
+ predictTableWithUnseenValues =
+ tEnv.fromDataStream(
+ env.fromCollection(
+ Arrays.asList(Row.of(0, 3), Row.of(1, 2), Row.of(4, 1))))
+ .as("inputCol1", "inputCol2");
+ }
+
+ @Test
+ public void testPredictParam() {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testInputWithUnseenValues() {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTableWithUnseenValues)[0];
+ try {
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen index: 4.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testPredict() throws Exception {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ StringIndexerTest.verifyPredictionResult(expectedPrediction, predictedResult);
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testSaveLoadAndPredict() throws Exception {
Review comment:
nits: The code seems a bit too crowded as there is no line break in this method body.
Any chance we could make it less crowded? Maybe see `KMeansTest` or Spark's `KMeansSuite` for example?
Same for other tests.
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModelParams.java
##########
@@ -0,0 +1,60 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.common.param.HasInputCols;
+import org.apache.flink.ml.common.param.HasOutputCols;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexerModel}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerModelParams<T> extends HasInputCols<T>, HasOutputCols<T> {
+ String ERROR_INVALID = "error";
+ String SKIP_INVALID = "skip";
+ String KEEP_INVALID = "keep";
+ /**
+ * Supported options and the corresponding behavior to handle invalid entries in
+ * StringIndexerModel are listed as follows.
+ *
+ * <ul>
+ * <li>error: raise an exception.
+ * <li>skip: filter out rows with bad values.
+ * <li>keep: put the invalid entries in a special bucket.
Review comment:
Would it be useful to mention what this `special bucket` is? For example, we can say the index of these invalid entries is the expected number of distinct values in this column.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831815876
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/StringIndexerTest.java
##########
@@ -0,0 +1,319 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.TestBaseUtils;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link StringIndexer} and {@link StringIndexerModel}. */
+public class StringIndexerTest {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table trainTable;
+ private Table predictTable;
+
+ private String[][] expectedAlphabeticAscModelData;
+ private List<Row> expectedAlphabeticAscPredictData;
+ private List<Row> expectedAlphabeticDescPredictData;
+ private List<Row> expectedFreqAscPredictData;
+ private List<Row> expectedFreqDescPredictData;
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ List<Row> trainData =
+ Arrays.asList(
+ Row.of("a", 1.),
+ Row.of("b", 1.),
+ Row.of("b", 2.0),
+ Row.of("c", 0.0),
+ Row.of("d", 2.),
+ Row.of("a", 2.),
+ Row.of("b", 2.),
+ Row.of("b", -1.),
+ Row.of("a", -1.),
+ Row.of("c", -1.));
+ trainTable =
+ tEnv.fromDataStream(env.fromCollection(trainData)).as("inputCol1", "inputCol2");
+
+ List<Row> predictData = Arrays.asList(Row.of("a", 2.), Row.of("b", 1.), Row.of("e", 2.));
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(predictData)).as("inputCol1", "inputCol2");
+
+ expectedAlphabeticAscModelData =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+ expectedAlphabeticAscPredictData =
+ Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2), Row.of("e", 2.0, 4, 3));
+ expectedAlphabeticDescPredictData =
+ Arrays.asList(Row.of("a", 2., 3, 0), Row.of("b", 1., 2, 1), Row.of("e", 2.0, 4, 0));
+ expectedFreqAscPredictData =
+ Arrays.asList(Row.of("a", 2., 2, 3), Row.of("b", 1., 3, 1), Row.of("e", 2.0, 4, 3));
+ expectedFreqDescPredictData =
+ Arrays.asList(Row.of("a", 2., 1, 0), Row.of("b", 1., 0, 2), Row.of("e", 2.0, 4, 0));
+ }
+
+ @Test
+ public void testFitParam() {
+ StringIndexer stringIndexer = new StringIndexer();
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.RANDOM_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.ERROR_INVALID);
+
+ stringIndexer
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol1")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+
+ assertArrayEquals(new String[] {"inputCol1", "inputCol2"}, stringIndexer.getInputCols());
+ assertArrayEquals(new String[] {"outputCol1", "outputCol1"}, stringIndexer.getOutputCols());
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.ALPHABET_ASC_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.SKIP_INVALID);
+ }
+
+ @Test
+ public void testPredictParam() {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testStringOrderType() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+
+ Table output;
+ List<Row> predictedResult;
+
+ // alphabetAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // alphabetDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticDescPredictData, predictedResult);
+
+ // frequencyAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqAscPredictData, predictedResult);
+
+ // frequencyDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqDescPredictData, predictedResult);
+
+ // random
+ stringIndexer.setStringOrderType(StringIndexerParams.RANDOM_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ Set<Integer> distinctStringsCol1 = new HashSet<>();
+ Set<Integer> distinctStringsCol2 = new HashSet<>();
+ int index;
+ for (Row r : predictedResult) {
+ index = (Integer) r.getField(2);
+ distinctStringsCol1.add(index);
+ assertTrue(index >= 0 && index <= 4);
+ index = (Integer) r.getField(3);
+ assertTrue(index >= 0 && index <= 3);
+ distinctStringsCol2.add(index);
+ }
+ assertEquals(3, distinctStringsCol1.size());
+ assertEquals(2, distinctStringsCol2.size());
+ }
+
+ @Test
+ public void testHandleInvalid() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+
+ Table output;
+ List<Row> expectedResult;
+
+ // keep invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // skip invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ expectedResult = Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2));
+ verifyPredictionResult(expectedResult, predictedResult);
+
+ // throw exception on invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.ERROR_INVALID);
+ try {
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen string: e. "
+ + "To handle unseen strings, set Param handleInvalid to keep.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ public void testFitAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testSaveLoadAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ stringIndexer =
+ StageTestUtils.saveAndReload(
+ env, stringIndexer, tempFolder.newFolder().getAbsolutePath());
+ StringIndexerModel model = stringIndexer.fit(trainTable);
+ model = StageTestUtils.saveAndReload(env, model, tempFolder.newFolder().getAbsolutePath());
+ assertEquals(
+ Collections.singletonList("stringsArray"),
+ model.getModelData()[0].getResolvedSchema().getColumnNames());
+ Table output = model.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testGetModelData() throws Exception {
+ StringIndexerModel model =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .fit(trainTable);
+ StringIndexerModelData modelData =
+ StringIndexerModelData.getModelDataStream(model.getModelData()[0])
Review comment:
Thanks for the comment. The check for schema is added in StringIndexerTest#testGetModelData.
It is not added in `IndexToStringModelTest#testGetModelData()` because the model data is generated by `StringIndexer`. I believe that the check for schema is already covered by `StringIndexerTest#testGetModelData`.
What do you think?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831813941
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
Review comment:
Thanks for the comment. It is updated as follows:
`A string indexer maps one/multiple columns (string/numerical value) of the input to
one/multiple indexed output columns (integer value). The output indices of two data points are
same iff their corresponding input columns are the same. The indices are in [0,
numDistinctValuesInThisColumn].
`
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831836535
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
+ String FREQUENCY_DESC_ORDER = "frequencyDesc";
+ String FREQUENCY_ASC_ORDER = "frequencyAsc";
+ String ALPHABET_DESC_ORDER = "alphabetDesc";
+ String ALPHABET_ASC_ORDER = "alphabetAsc";
+
+ /** Param for {@link StringIndexer} to decide the order of the strings of each column. */
+ Param<String> STRING_ORDER_TYPE =
Review comment:
Thanks for the update.
Would it be useful to also mention "The first label after ordering is assigned an index of 0"?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#issuecomment-1075850626
Thanks for the update! The PR has been merged to the master branch.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer in FlinkML
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#issuecomment-1005741159
@gaoyunhaii @lindong28 Can you help review this PR?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831837017
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/StringIndexerTest.java
##########
@@ -0,0 +1,319 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.TestBaseUtils;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link StringIndexer} and {@link StringIndexerModel}. */
+public class StringIndexerTest {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table trainTable;
+ private Table predictTable;
+
+ private String[][] expectedAlphabeticAscModelData;
+ private List<Row> expectedAlphabeticAscPredictData;
+ private List<Row> expectedAlphabeticDescPredictData;
+ private List<Row> expectedFreqAscPredictData;
+ private List<Row> expectedFreqDescPredictData;
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ List<Row> trainData =
+ Arrays.asList(
+ Row.of("a", 1.),
+ Row.of("b", 1.),
+ Row.of("b", 2.0),
+ Row.of("c", 0.0),
+ Row.of("d", 2.),
+ Row.of("a", 2.),
+ Row.of("b", 2.),
+ Row.of("b", -1.),
+ Row.of("a", -1.),
+ Row.of("c", -1.));
+ trainTable =
+ tEnv.fromDataStream(env.fromCollection(trainData)).as("inputCol1", "inputCol2");
+
+ List<Row> predictData = Arrays.asList(Row.of("a", 2.), Row.of("b", 1.), Row.of("e", 2.));
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(predictData)).as("inputCol1", "inputCol2");
+
+ expectedAlphabeticAscModelData =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+ expectedAlphabeticAscPredictData =
+ Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2), Row.of("e", 2.0, 4, 3));
+ expectedAlphabeticDescPredictData =
+ Arrays.asList(Row.of("a", 2., 3, 0), Row.of("b", 1., 2, 1), Row.of("e", 2.0, 4, 0));
+ expectedFreqAscPredictData =
+ Arrays.asList(Row.of("a", 2., 2, 3), Row.of("b", 1., 3, 1), Row.of("e", 2.0, 4, 3));
+ expectedFreqDescPredictData =
+ Arrays.asList(Row.of("a", 2., 1, 0), Row.of("b", 1., 0, 2), Row.of("e", 2.0, 4, 0));
+ }
+
+ @Test
+ public void testFitParam() {
+ StringIndexer stringIndexer = new StringIndexer();
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.RANDOM_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.ERROR_INVALID);
+
+ stringIndexer
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol1")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+
+ assertArrayEquals(new String[] {"inputCol1", "inputCol2"}, stringIndexer.getInputCols());
+ assertArrayEquals(new String[] {"outputCol1", "outputCol1"}, stringIndexer.getOutputCols());
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.ALPHABET_ASC_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.SKIP_INVALID);
+ }
+
+ @Test
+ public void testPredictParam() {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testStringOrderType() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+
+ Table output;
+ List<Row> predictedResult;
+
+ // alphabetAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // alphabetDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticDescPredictData, predictedResult);
+
+ // frequencyAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqAscPredictData, predictedResult);
+
+ // frequencyDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqDescPredictData, predictedResult);
+
+ // random
+ stringIndexer.setStringOrderType(StringIndexerParams.RANDOM_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ Set<Integer> distinctStringsCol1 = new HashSet<>();
+ Set<Integer> distinctStringsCol2 = new HashSet<>();
+ int index;
+ for (Row r : predictedResult) {
+ index = (Integer) r.getField(2);
+ distinctStringsCol1.add(index);
+ assertTrue(index >= 0 && index <= 4);
+ index = (Integer) r.getField(3);
+ assertTrue(index >= 0 && index <= 3);
+ distinctStringsCol2.add(index);
+ }
+ assertEquals(3, distinctStringsCol1.size());
+ assertEquals(2, distinctStringsCol2.size());
+ }
+
+ @Test
+ public void testHandleInvalid() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+
+ Table output;
+ List<Row> expectedResult;
+
+ // keep invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // skip invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ expectedResult = Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2));
+ verifyPredictionResult(expectedResult, predictedResult);
+
+ // throw exception on invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.ERROR_INVALID);
+ try {
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen string: e. "
+ + "To handle unseen strings, set Param handleInvalid to keep.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ public void testFitAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testSaveLoadAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ stringIndexer =
+ StageTestUtils.saveAndReload(
+ env, stringIndexer, tempFolder.newFolder().getAbsolutePath());
+ StringIndexerModel model = stringIndexer.fit(trainTable);
+ model = StageTestUtils.saveAndReload(env, model, tempFolder.newFolder().getAbsolutePath());
+ assertEquals(
+ Collections.singletonList("stringsArray"),
+ model.getModelData()[0].getResolvedSchema().getColumnNames());
+ Table output = model.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testGetModelData() throws Exception {
+ StringIndexerModel model =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .fit(trainTable);
+ StringIndexerModelData modelData =
+ StringIndexerModelData.getModelDataStream(model.getModelData()[0])
Review comment:
Sounds good. Thanks.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r832086803
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,198 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerModelParams<StringIndexerModel> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+ private Table modelDataTable;
+
+ public StringIndexerModel() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ ReadWriteUtils.saveModelData(
+ StringIndexerModelData.getModelDataStream(modelDataTable),
+ path,
+ new StringIndexerModelData.ModelDataEncoder());
+ }
+
+ public static StringIndexerModel load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ StreamTableEnvironment tEnv = StreamTableEnvironment.create(env);
+ StringIndexerModel model = ReadWriteUtils.loadStageParam(path);
+ DataStream<StringIndexerModelData> modelData =
+ ReadWriteUtils.loadModelData(
+ env, path, new StringIndexerModelData.ModelDataDecoder());
+ return model.setModelData(tEnv.fromDataStream(modelData));
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel setModelData(Table... inputs) {
+ modelDataTable = inputs[0];
+ return this;
+ }
+
+ @Override
+ public Table[] getModelData() {
+ return new Table[] {modelDataTable};
+ }
+
+ @Override
+ @SuppressWarnings("unchecked, rawtypes")
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) modelDataTable).getTableEnvironment();
+
+ final String broadcastModelKey = "broadcastModelKey";
+ DataStream<StringIndexerModelData> modelDataStream =
+ StringIndexerModelData.getModelDataStream(modelDataTable);
+ RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema());
+ TypeInformation<?>[] outputTypes = new TypeInformation[outputCols.length];
+ Arrays.fill(outputTypes, BasicTypeInfo.INT_TYPE_INFO);
+ RowTypeInfo outputTypeInfo =
+ new RowTypeInfo(
+ ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), outputTypes),
+ ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCols()));
+ DataStream<Row> result =
+ BroadcastUtils.withBroadcastStream(
+ Collections.singletonList(tEnv.toDataStream(inputs[0])),
+ Collections.singletonMap(broadcastModelKey, modelDataStream),
+ inputList -> {
+ DataStream inputData = inputList.get(0);
+ return inputData.flatMap(
+ new String2Index(
+ broadcastModelKey, inputCols, getHandleInvalid()),
+ outputTypeInfo);
+ });
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Maps the input columns to integer values according to the model data. */
+ private static class String2Index extends RichFlatMapFunction<Row, Row> {
+ private HashMap<String, Integer>[] modelDataMap;
+ private final String broadcastModelKey;
+ private final String[] inputCols;
+ private final String handleInValid;
+
+ public String2Index(String broadcastModelKey, String[] inputCols, String handleInValid) {
+ this.broadcastModelKey = broadcastModelKey;
+ this.inputCols = inputCols;
+ this.handleInValid = handleInValid;
+ }
+
+ @Override
+ @SuppressWarnings("unchecked")
+ public void flatMap(Row input, Collector<Row> out) {
+ if (modelDataMap == null) {
+ modelDataMap = new HashMap[inputCols.length];
+ StringIndexerModelData modelData =
+ (StringIndexerModelData)
+ getRuntimeContext().getBroadcastVariable(broadcastModelKey).get(0);
+ String[][] stringsArray = modelData.stringArrays;
+ for (int i = 0; i < stringsArray.length; i++) {
+ int idx = 0;
+ modelDataMap[i] = new HashMap<>(stringsArray[i].length);
+ for (String string : stringsArray[i]) {
+ modelDataMap[i].put(string, idx++);
+ }
+ }
+ }
+ Row outputIndices = new Row(inputCols.length);
+ for (int i = 0; i < inputCols.length; i++) {
+ Object objVal = input.getField(inputCols[i]);
+ String stringVal;
+ if (objVal instanceof String) {
+ stringVal = (String) objVal;
+ } else if (objVal instanceof Number) {
+ stringVal = String.valueOf(objVal);
+ } else {
+ throw new RuntimeException(
+ "The input column only supports string and numeric type.");
+ }
+ if (modelDataMap[i].containsKey(stringVal)) {
+ outputIndices.setField(i, modelDataMap[i].get(stringVal));
+ } else {
+ switch (handleInValid) {
+ case StringIndexerModelParams.SKIP_INVALID:
+ return;
+ case StringIndexerModelParams.ERROR_INVALID:
+ throw new RuntimeException(
+ "The input contains unseen string: "
+ + stringVal
+ + ". To handle unseen strings, set Param "
Review comment:
Sounds good :)
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831980106
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/StringIndexerTest.java
##########
@@ -0,0 +1,317 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.AbstractTestBase;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link StringIndexer} and {@link StringIndexerModel}. */
+public class StringIndexerTest extends AbstractTestBase {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table trainTable;
+ private Table predictTable;
+
+ private final String[][] expectedAlphabeticAscModelData =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+ private final List<Row> expectedAlphabeticAscPredictData =
+ Arrays.asList(Row.of("a", 2.0, 0, 3), Row.of("b", 1.0, 1, 2), Row.of("e", 2.0, 4, 3));
+ private final List<Row> expectedAlphabeticDescPredictData =
+ Arrays.asList(Row.of("a", 2.0, 3, 0), Row.of("b", 1.0, 2, 1), Row.of("e", 2.0, 4, 0));
+ private final List<Row> expectedFreqAscPredictData =
+ Arrays.asList(Row.of("a", 2.0, 2, 3), Row.of("b", 1.0, 3, 1), Row.of("e", 2.0, 4, 3));
+ private final List<Row> expectedFreqDescPredictData =
+ Arrays.asList(Row.of("a", 2.0, 1, 0), Row.of("b", 1.0, 0, 2), Row.of("e", 2.0, 4, 0));
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+ List<Row> trainData =
+ Arrays.asList(
+ Row.of("a", 1.0),
+ Row.of("b", 1.0),
+ Row.of("b", 2.0),
+ Row.of("c", 0.0),
+ Row.of("d", 2.0),
+ Row.of("a", 2.0),
+ Row.of("b", 2.0),
+ Row.of("b", -1.0),
+ Row.of("a", -1.0),
+ Row.of("c", -1.0));
+ trainTable =
+ tEnv.fromDataStream(env.fromCollection(trainData)).as("inputCol1", "inputCol2");
+ List<Row> predictData = Arrays.asList(Row.of("a", 2.0), Row.of("b", 1.0), Row.of("e", 2.0));
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(predictData)).as("inputCol1", "inputCol2");
+ }
+
+ @Test
+ public void testFitParam() {
+ StringIndexer stringIndexer = new StringIndexer();
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.ARBITRARY_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.ERROR_INVALID);
+
+ stringIndexer
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ assertArrayEquals(new String[] {"inputCol1", "inputCol2"}, stringIndexer.getInputCols());
+ assertArrayEquals(new String[] {"outputCol1", "outputCol2"}, stringIndexer.getOutputCols());
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.ALPHABET_ASC_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.SKIP_INVALID);
+ }
+
+ @Test
+ public void testPredictParam() {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ @SuppressWarnings("all")
+ public void testStringOrderType() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ Table output;
+ List<Row> predictedResult;
+
+ // alphabetAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // alphabetDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticDescPredictData, predictedResult);
+
+ // frequencyAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqAscPredictData, predictedResult);
+
+ // frequencyDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqDescPredictData, predictedResult);
+
+ // arbitrary
+ stringIndexer.setStringOrderType(StringIndexerParams.ARBITRARY_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ Set<Integer> distinctStringsCol1 = new HashSet<>();
+ Set<Integer> distinctStringsCol2 = new HashSet<>();
+ int index;
+ for (Row r : predictedResult) {
+ index = (Integer) r.getField(2);
+ distinctStringsCol1.add(index);
+ assertTrue(index >= 0 && index <= 4);
+ index = (Integer) r.getField(3);
+ assertTrue(index >= 0 && index <= 3);
+ distinctStringsCol2.add(index);
+ }
+ assertEquals(3, distinctStringsCol1.size());
+ assertEquals(2, distinctStringsCol2.size());
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testHandleInvalid() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+
+ Table output;
+ List<Row> expectedResult;
+
+ // keep invalid data
Review comment:
nits: How about we always start a comment with upper-case letter and ends with a dot?
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/IndexToStringModelTest.java
##########
@@ -0,0 +1,162 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.AbstractTestBase;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.List;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link IndexToStringModel}. */
+public class IndexToStringModelTest extends AbstractTestBase {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table illegalPredictTable;
+ private Table predictTable;
+ private Table modelTable;
+
+ private final List<Row> expectedPrediction =
+ Arrays.asList(Row.of(0, 3, "a", "2.0"), Row.of(1, 2, "b", "1.0"));
+ private final String[][] stringArrays =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ modelTable =
+ tEnv.fromDataStream(env.fromElements(new StringIndexerModelData(stringArrays)))
+ .as("stringArrays");
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(Arrays.asList(Row.of(0, 3), Row.of(1, 2))))
+ .as("inputCol1", "inputCol2");
+ illegalPredictTable =
+ tEnv.fromDataStream(
+ env.fromCollection(
+ Arrays.asList(Row.of(0, 3), Row.of(1, 2), Row.of(4, 1))))
+ .as("inputCol1", "inputCol2");
+ }
+
+ @Test
+ public void testPredictParam() {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testIllegalInput() {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(illegalPredictTable)[0];
+ try {
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen index: 4.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testPredict() throws Exception {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ StringIndexerTest.verifyPredictionResult(expectedPrediction, predictedResult);
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testSaveLoadAndPredict() throws Exception {
+ IndexToStringModel model =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ model = StageTestUtils.saveAndReload(env, model, tempFolder.newFolder().getAbsolutePath());
+ assertEquals(
+ Collections.singletonList("stringArrays"),
+ model.getModelData()[0].getResolvedSchema().getColumnNames());
+ Table output = model.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ StringIndexerTest.verifyPredictionResult(expectedPrediction, predictedResult);
+ }
+
+ @Test
+ public void testGetModelData() throws Exception {
+ IndexToStringModel model =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+
+ StringIndexerModelData modelData =
+ StringIndexerModelData.getModelDataStream(model.getModelData()[0])
+ .executeAndCollect()
Review comment:
Should we replace `next()` with `IteratorUtils.toList(...)` for consistency with other existing tests?
This approach would also verify that there is exactly one record in the model data stream in this test.
Same for `StringIndexerTest`.
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,198 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerModelParams<StringIndexerModel> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+ private Table modelDataTable;
+
+ public StringIndexerModel() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ ReadWriteUtils.saveModelData(
+ StringIndexerModelData.getModelDataStream(modelDataTable),
+ path,
+ new StringIndexerModelData.ModelDataEncoder());
+ }
+
+ public static StringIndexerModel load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ StreamTableEnvironment tEnv = StreamTableEnvironment.create(env);
+ StringIndexerModel model = ReadWriteUtils.loadStageParam(path);
+ DataStream<StringIndexerModelData> modelData =
+ ReadWriteUtils.loadModelData(
+ env, path, new StringIndexerModelData.ModelDataDecoder());
+ return model.setModelData(tEnv.fromDataStream(modelData));
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel setModelData(Table... inputs) {
+ modelDataTable = inputs[0];
+ return this;
+ }
+
+ @Override
+ public Table[] getModelData() {
+ return new Table[] {modelDataTable};
+ }
+
+ @Override
+ @SuppressWarnings("unchecked, rawtypes")
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) modelDataTable).getTableEnvironment();
+
+ final String broadcastModelKey = "broadcastModelKey";
+ DataStream<StringIndexerModelData> modelDataStream =
+ StringIndexerModelData.getModelDataStream(modelDataTable);
+ RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema());
+ TypeInformation<?>[] outputTypes = new TypeInformation[outputCols.length];
+ Arrays.fill(outputTypes, BasicTypeInfo.INT_TYPE_INFO);
+ RowTypeInfo outputTypeInfo =
+ new RowTypeInfo(
+ ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), outputTypes),
+ ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCols()));
+ DataStream<Row> result =
+ BroadcastUtils.withBroadcastStream(
+ Collections.singletonList(tEnv.toDataStream(inputs[0])),
+ Collections.singletonMap(broadcastModelKey, modelDataStream),
+ inputList -> {
+ DataStream inputData = inputList.get(0);
+ return inputData.flatMap(
+ new String2Index(
+ broadcastModelKey, inputCols, getHandleInvalid()),
+ outputTypeInfo);
+ });
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Maps the input columns to integer values according to the model data. */
+ private static class String2Index extends RichFlatMapFunction<Row, Row> {
+ private HashMap<String, Integer>[] modelDataMap;
+ private final String broadcastModelKey;
+ private final String[] inputCols;
+ private final String handleInValid;
+
+ public String2Index(String broadcastModelKey, String[] inputCols, String handleInValid) {
+ this.broadcastModelKey = broadcastModelKey;
+ this.inputCols = inputCols;
+ this.handleInValid = handleInValid;
+ }
+
+ @Override
+ @SuppressWarnings("unchecked")
+ public void flatMap(Row input, Collector<Row> out) {
+ if (modelDataMap == null) {
+ modelDataMap = new HashMap[inputCols.length];
+ StringIndexerModelData modelData =
+ (StringIndexerModelData)
+ getRuntimeContext().getBroadcastVariable(broadcastModelKey).get(0);
+ String[][] stringsArray = modelData.stringArrays;
+ for (int i = 0; i < stringsArray.length; i++) {
+ int idx = 0;
+ modelDataMap[i] = new HashMap<>(stringsArray[i].length);
+ for (String string : stringsArray[i]) {
+ modelDataMap[i].put(string, idx++);
+ }
+ }
+ }
+ Row outputIndices = new Row(inputCols.length);
+ for (int i = 0; i < inputCols.length; i++) {
+ Object objVal = input.getField(inputCols[i]);
+ String stringVal;
+ if (objVal instanceof String) {
+ stringVal = (String) objVal;
+ } else if (objVal instanceof Number) {
+ stringVal = String.valueOf(objVal);
+ } else {
+ throw new RuntimeException(
+ "The input column only supports string and numeric type.");
+ }
+ if (modelDataMap[i].containsKey(stringVal)) {
+ outputIndices.setField(i, modelDataMap[i].get(stringVal));
+ } else {
+ switch (handleInValid) {
+ case StringIndexerModelParams.SKIP_INVALID:
+ return;
+ case StringIndexerModelParams.ERROR_INVALID:
+ throw new RuntimeException(
+ "The input contains unseen string: "
+ + stringVal
+ + ". To handle unseen strings, set Param "
Review comment:
The error message seems incorrect. In order to `handle` the unseen strings, any one of three options work, right?
I guess you mean that `To keep unseen strings in the output, set ...`. But if we say so, do we also need to say `To drop unseen strings from the output, set ...`? Would that be too verbose?
It is probably simpler to just mention "see handleInvalid parameter for more options".
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/IndexToStringModelTest.java
##########
@@ -0,0 +1,162 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.AbstractTestBase;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.List;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link IndexToStringModel}. */
+public class IndexToStringModelTest extends AbstractTestBase {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table illegalPredictTable;
Review comment:
It is not clear what `illegal` means here. Would it be more intuitive to name it `predictTableWithUnseenValues`?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831693277
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
Review comment:
There appears to be typo in this sentence.
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,204 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerParams<StringIndexerModel> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+ private Table modelDataTable;
+
+ public StringIndexerModel() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ ReadWriteUtils.saveModelData(
+ StringIndexerModelData.getModelDataStream(modelDataTable),
+ path,
+ new StringIndexerModelData.ModelDataEncoder());
+ }
+
+ public static StringIndexerModel load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ StreamTableEnvironment tEnv = StreamTableEnvironment.create(env);
+ StringIndexerModel model = ReadWriteUtils.loadStageParam(path);
+ DataStream<StringIndexerModelData> modelData =
+ ReadWriteUtils.loadModelData(
+ env, path, new StringIndexerModelData.ModelDataDecoder());
+ return model.setModelData(tEnv.fromDataStream(modelData));
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel setModelData(Table... inputs) {
+ modelDataTable = inputs[0];
+ return this;
+ }
+
+ @Override
+ public Table[] getModelData() {
+ return new Table[] {modelDataTable};
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) modelDataTable).getTableEnvironment();
+
+ final String broadcastModelKey = "broadcastModelKey";
+ DataStream<StringIndexerModelData> modelDataStream =
+ StringIndexerModelData.getModelDataStream(modelDataTable);
+ RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema());
+ TypeInformation<?>[] outputTypes = new TypeInformation[outputCols.length];
+ Arrays.fill(outputTypes, BasicTypeInfo.INT_TYPE_INFO);
+ RowTypeInfo outputTypeInfo =
+ new RowTypeInfo(
+ ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), outputTypes),
+ ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCols()));
+ DataStream<Row> result =
+ BroadcastUtils.withBroadcastStream(
+ Collections.singletonList(tEnv.toDataStream(inputs[0])),
+ Collections.singletonMap(broadcastModelKey, modelDataStream),
+ inputList -> {
+ DataStream inputData = inputList.get(0);
+ return inputData.flatMap(
+ new String2Index(
+ broadcastModelKey, inputCols, getHandleInvalid()),
+ outputTypeInfo);
+ });
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Maps the input columns to integer values according to the model data. */
+ private static class String2Index extends RichFlatMapFunction<Row, Row> {
+ private HashMap<String, Integer>[] modelDataMap;
+ private final String broadcastModelKey;
+ private final String[] inputCols;
+ private final String handleInValid;
+
+ public String2Index(String broadcastModelKey, String[] inputCols, String handleInValid) {
+ this.broadcastModelKey = broadcastModelKey;
+ this.inputCols = inputCols;
+ this.handleInValid = handleInValid;
+ }
+
+ @Override
+ public void flatMap(Row input, Collector<Row> out) throws Exception {
+ if (modelDataMap == null) {
+ modelDataMap = new HashMap[inputCols.length];
+ StringIndexerModelData modelData =
+ (StringIndexerModelData)
+ getRuntimeContext().getBroadcastVariable(broadcastModelKey).get(0);
+ String[][] stringsArray = modelData.stringsArray;
+ for (int i = 0; i < stringsArray.length; i++) {
+ int idx = 0;
+ modelDataMap[i] = new HashMap<>(stringsArray[i].length);
+ for (String string : stringsArray[i]) {
+ modelDataMap[i].put(string, idx++);
+ }
+ }
+ }
+ Row outputIndices = new Row(inputCols.length);
+ for (int i = 0; i < inputCols.length; i++) {
+ Object objVal = input.getField(inputCols[i]);
+ String stringVal;
+ if (objVal instanceof String) {
+ stringVal = (String) objVal;
+ } else if (objVal instanceof Number) {
+ stringVal = String.valueOf(objVal);
+ } else {
+ throw new RuntimeException(
+ "The input column only supports string type and numeric type.");
+ }
+ if (modelDataMap[i].containsKey(stringVal)) {
+ outputIndices.setField(i, modelDataMap[i].get(stringVal));
+ } else {
+ switch (handleInValid) {
+ case StringIndexerModelParams.SKIP_INVALID:
+ return;
+ case StringIndexerModelParams.ERROR_INVALID:
+ throw new RuntimeException(
Review comment:
The validator of the stringOrderType parameter should already guarantee that this order that is in the given list of values. Would it be simpler to throw `IllegalStateException` here without listing the expected values?
Same for `handleInvalid` used in `StringIndexerModel::String2Index`.
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
+ * the same. The indices are in [0, numDistinctValuesThisColumn).
+ *
+ * <p>The input columns are cast to string if they are numeric values. By default, the output model
+ * is random ordered. Users can control this by setting {@link
+ * StringIndexerParams#STRING_ORDER_TYPE}.
+ */
+public class StringIndexer
+ implements Estimator<StringIndexer, StringIndexerModel>,
+ StringIndexerParams<StringIndexer> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+
+ public StringIndexer() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ }
+
+ public static StringIndexer load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ return ReadWriteUtils.loadStageParam(path);
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel fit(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment();
+
+ DataStream<Tuple2<Integer, String>> columnIdAndString =
+ tEnv.toDataStream(inputs[0]).flatMap(new ExtractColumnIdAndString(inputCols));
+ DataStream<Tuple3<Integer, String, Long>> columnIdAndStringAndCnt =
+ DataStreamUtils.mapPartition(
+ columnIdAndString.keyBy(
+ (KeySelector<Tuple2<Integer, String>, Integer>) Tuple2::hashCode),
+ new CountStringsByColumn(inputCols.length));
+ DataStream<StringIndexerModelData> modelData =
+ DataStreamUtils.mapPartition(
+ columnIdAndStringAndCnt,
+ new GenerateModel(inputCols.length, getStringOrderType()));
+ modelData.getTransformation().setParallelism(1);
+ StringIndexerModel model =
+ new StringIndexerModel().setModelData(tEnv.fromDataStream(modelData));
+ ReadWriteUtils.updateExistingParams(model, paramMap);
+ return model;
+ }
+
+ /**
+ * Merges all the extracted strings and generates the {@link StringIndexerModelData} according
+ * to the specified string order type.
+ */
+ private static class GenerateModel
+ implements MapPartitionFunction<Tuple3<Integer, String, Long>, StringIndexerModelData> {
+ private final int numCols;
+ private final String stringOrderType;
+
+ public GenerateModel(int numCols, String stringOrderType) {
+ this.numCols = numCols;
+ this.stringOrderType = stringOrderType;
+ }
+
+ @Override
+ public void mapPartition(
+ Iterable<Tuple3<Integer, String, Long>> values,
+ Collector<StringIndexerModelData> out) {
+ String[][] stringsArray = new String[numCols][];
+ ArrayList<Tuple2<String, Long>>[] stringsAndCntsByColumn = new ArrayList[numCols];
+ for (int i = 0; i < numCols; i++) {
+ stringsAndCntsByColumn[i] = new ArrayList<>();
+ }
+ for (Tuple3<Integer, String, Long> colIdAndStringAndCnt : values) {
+ stringsAndCntsByColumn[colIdAndStringAndCnt.f0].add(
+ Tuple2.of(colIdAndStringAndCnt.f1, colIdAndStringAndCnt.f2));
+ }
+ for (int i = 0; i < stringsAndCntsByColumn.length; i++) {
+ List<Tuple2<String, Long>> stringsAndCnts = stringsAndCntsByColumn[i];
+ switch (stringOrderType) {
+ case StringIndexerParams.ALPHABET_ASC_ORDER:
+ stringsAndCnts.sort(Comparator.comparing(valAndCnt -> valAndCnt.f0));
+ break;
+ case StringIndexerParams.ALPHABET_DESC_ORDER:
+ stringsAndCnts.sort(
+ (valAndCnt1, valAndCnt2) ->
+ -valAndCnt1.f0.compareTo(valAndCnt2.f0));
+ break;
+ case StringIndexerParams.FREQUENCY_ASC_ORDER:
+ stringsAndCnts.sort(Comparator.comparing(valAndCnt -> valAndCnt.f1));
+ break;
+ case StringIndexerParams.FREQUENCY_DESC_ORDER:
+ stringsAndCnts.sort(
+ (valAndCnt1, valAndCnt2) ->
+ -valAndCnt1.f1.compareTo(valAndCnt2.f1));
+ break;
+ case StringIndexerParams.RANDOM_ORDER:
+ break;
+ default:
+ throw new RuntimeException(
+ "Unsupported string order type: "
+ + stringOrderType
+ + ". Supported options are: ["
+ + StringIndexerParams.RANDOM_ORDER
+ + ", "
+ + StringIndexerParams.ALPHABET_ASC_ORDER
+ + ", "
+ + StringIndexerParams.ALPHABET_DESC_ORDER
+ + ", "
+ + StringIndexerParams.FREQUENCY_ASC_ORDER
+ + ", "
+ + StringIndexerParams.ALPHABET_DESC_ORDER
+ + "].");
+ }
+ stringsArray[i] = new String[stringsAndCnts.size()];
+ for (int stringId = 0; stringId < stringsArray[i].length; stringId++) {
+ stringsArray[i][stringId] = stringsAndCnts.get(stringId).f0;
+ }
+ }
+
+ out.collect(new StringIndexerModelData(stringsArray));
+ }
+ }
+
+ /** Computes the frequency of strings in each column. */
+ private static class CountStringsByColumn
+ implements MapPartitionFunction<
+ Tuple2<Integer, String>, Tuple3<Integer, String, Long>> {
+ private final int numCols;
+
+ public CountStringsByColumn(int numCols) {
+ this.numCols = numCols;
+ }
+
+ @Override
+ public void mapPartition(
+ Iterable<Tuple2<Integer, String>> values,
+ Collector<Tuple3<Integer, String, Long>> out) {
+ HashMap<String, Long>[] string2CntByColumn = new HashMap[numCols];
+ for (int i = 0; i < numCols; i++) {
+ string2CntByColumn[i] = new HashMap<>();
+ }
+ for (Tuple2<Integer, String> columnIdAndString : values) {
+ int colId = columnIdAndString.f0;
+ String stringVal = columnIdAndString.f1;
+ if (string2CntByColumn[colId].containsKey(stringVal)) {
Review comment:
It might be more performant and readable to do the following:
```
Long cnt = string2CntByColumn[colId].getOrDefault(stringVal, 0L) + 1;
string2CntByColumn[colId].put(stringVal, cnt);
```
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModel.java
##########
@@ -0,0 +1,204 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.RichFlatMapFunction;
+import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.java.typeutils.RowTypeInfo;
+import org.apache.flink.ml.api.Model;
+import org.apache.flink.ml.common.broadcast.BroadcastUtils;
+import org.apache.flink.ml.common.datastream.TableUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import org.apache.commons.lang3.ArrayUtils;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * A Model which transforms input string/numeric column(s) to integer column(s) using the model data
+ * computed by {@link StringIndexer}.
+ */
+public class StringIndexerModel
+ implements Model<StringIndexerModel>, StringIndexerParams<StringIndexerModel> {
Review comment:
Should this class implement `StringIndexerModelParams` instead of `StringIndexerParams`?
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
+ * the same. The indices are in [0, numDistinctValuesThisColumn).
Review comment:
Is this statement still correct if `handleInvalid = keep`?
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/StringIndexerTest.java
##########
@@ -0,0 +1,319 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.TestBaseUtils;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link StringIndexer} and {@link StringIndexerModel}. */
+public class StringIndexerTest {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table trainTable;
+ private Table predictTable;
+
+ private String[][] expectedAlphabeticAscModelData;
+ private List<Row> expectedAlphabeticAscPredictData;
+ private List<Row> expectedAlphabeticDescPredictData;
+ private List<Row> expectedFreqAscPredictData;
+ private List<Row> expectedFreqDescPredictData;
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ List<Row> trainData =
+ Arrays.asList(
+ Row.of("a", 1.),
+ Row.of("b", 1.),
+ Row.of("b", 2.0),
Review comment:
It appears that we use both `2.` and `2.0` in the code. Could we use the same style for consistency?
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerModelData.java
##########
@@ -0,0 +1,122 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.serialization.Encoder;
+import org.apache.flink.api.common.typeinfo.TypeInformation;
+import org.apache.flink.api.common.typeutils.base.IntSerializer;
+import org.apache.flink.api.common.typeutils.base.array.StringArraySerializer;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.connector.file.src.reader.SimpleStreamFormat;
+import org.apache.flink.core.fs.FSDataInputStream;
+import org.apache.flink.core.memory.DataInputViewStreamWrapper;
+import org.apache.flink.core.memory.DataOutputViewStreamWrapper;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+
+import java.io.EOFException;
+import java.io.IOException;
+import java.io.OutputStream;
+
+/**
+ * Model data of {@link StringIndexerModel} and {@link IndexToStringModel}.
+ *
+ * <p>This class also provides methods to convert model data from Table to DataStream, and classes
+ * to save/load model data.
+ */
+public class StringIndexerModelData {
+ /** Ordered strings of each input column. */
+ public String[][] stringsArray;
Review comment:
Given that it is an array of array of strings, would it be more self-explanatory to rename this variable as `stringArrays`?
Note that it seems redundant to name it stringsArrays given that an array must be `plural`.
If we agree to change this variable name, could you also update variables names used in the rest of this PR for consistency?
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/StringIndexerTest.java
##########
@@ -0,0 +1,319 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.TestBaseUtils;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertTrue;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link StringIndexer} and {@link StringIndexerModel}. */
+public class StringIndexerTest {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table trainTable;
+ private Table predictTable;
+
+ private String[][] expectedAlphabeticAscModelData;
+ private List<Row> expectedAlphabeticAscPredictData;
+ private List<Row> expectedAlphabeticDescPredictData;
+ private List<Row> expectedFreqAscPredictData;
+ private List<Row> expectedFreqDescPredictData;
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ List<Row> trainData =
+ Arrays.asList(
+ Row.of("a", 1.),
+ Row.of("b", 1.),
+ Row.of("b", 2.0),
+ Row.of("c", 0.0),
+ Row.of("d", 2.),
+ Row.of("a", 2.),
+ Row.of("b", 2.),
+ Row.of("b", -1.),
+ Row.of("a", -1.),
+ Row.of("c", -1.));
+ trainTable =
+ tEnv.fromDataStream(env.fromCollection(trainData)).as("inputCol1", "inputCol2");
+
+ List<Row> predictData = Arrays.asList(Row.of("a", 2.), Row.of("b", 1.), Row.of("e", 2.));
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(predictData)).as("inputCol1", "inputCol2");
+
+ expectedAlphabeticAscModelData =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+ expectedAlphabeticAscPredictData =
+ Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2), Row.of("e", 2.0, 4, 3));
+ expectedAlphabeticDescPredictData =
+ Arrays.asList(Row.of("a", 2., 3, 0), Row.of("b", 1., 2, 1), Row.of("e", 2.0, 4, 0));
+ expectedFreqAscPredictData =
+ Arrays.asList(Row.of("a", 2., 2, 3), Row.of("b", 1., 3, 1), Row.of("e", 2.0, 4, 3));
+ expectedFreqDescPredictData =
+ Arrays.asList(Row.of("a", 2., 1, 0), Row.of("b", 1., 0, 2), Row.of("e", 2.0, 4, 0));
+ }
+
+ @Test
+ public void testFitParam() {
+ StringIndexer stringIndexer = new StringIndexer();
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.RANDOM_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.ERROR_INVALID);
+
+ stringIndexer
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol1")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+
+ assertArrayEquals(new String[] {"inputCol1", "inputCol2"}, stringIndexer.getInputCols());
+ assertArrayEquals(new String[] {"outputCol1", "outputCol1"}, stringIndexer.getOutputCols());
+ assertEquals(stringIndexer.getStringOrderType(), StringIndexerParams.ALPHABET_ASC_ORDER);
+ assertEquals(stringIndexer.getHandleInvalid(), StringIndexerParams.SKIP_INVALID);
+ }
+
+ @Test
+ public void testPredictParam() {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testStringOrderType() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+
+ Table output;
+ List<Row> predictedResult;
+
+ // alphabetAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // alphabetDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.ALPHABET_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticDescPredictData, predictedResult);
+
+ // frequencyAsc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_ASC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqAscPredictData, predictedResult);
+
+ // frequencyDesc
+ stringIndexer.setStringOrderType(StringIndexerParams.FREQUENCY_DESC_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedFreqDescPredictData, predictedResult);
+
+ // random
+ stringIndexer.setStringOrderType(StringIndexerParams.RANDOM_ORDER);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ Set<Integer> distinctStringsCol1 = new HashSet<>();
+ Set<Integer> distinctStringsCol2 = new HashSet<>();
+ int index;
+ for (Row r : predictedResult) {
+ index = (Integer) r.getField(2);
+ distinctStringsCol1.add(index);
+ assertTrue(index >= 0 && index <= 4);
+ index = (Integer) r.getField(3);
+ assertTrue(index >= 0 && index <= 3);
+ distinctStringsCol2.add(index);
+ }
+ assertEquals(3, distinctStringsCol1.size());
+ assertEquals(2, distinctStringsCol2.size());
+ }
+
+ @Test
+ public void testHandleInvalid() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER);
+
+ Table output;
+ List<Row> expectedResult;
+
+ // keep invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+
+ // skip invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.SKIP_INVALID);
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ predictedResult = IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ expectedResult = Arrays.asList(Row.of("a", 2., 0, 3), Row.of("b", 1., 1, 2));
+ verifyPredictionResult(expectedResult, predictedResult);
+
+ // throw exception on invalid data
+ stringIndexer.setHandleInvalid(StringIndexerParams.ERROR_INVALID);
+ try {
+ output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen string: e. "
+ + "To handle unseen strings, set Param handleInvalid to keep.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ public void testFitAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ Table output = stringIndexer.fit(trainTable).transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testSaveLoadAndPredict() throws Exception {
+ StringIndexer stringIndexer =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .setHandleInvalid(StringIndexerParams.KEEP_INVALID);
+ stringIndexer =
+ StageTestUtils.saveAndReload(
+ env, stringIndexer, tempFolder.newFolder().getAbsolutePath());
+ StringIndexerModel model = stringIndexer.fit(trainTable);
+ model = StageTestUtils.saveAndReload(env, model, tempFolder.newFolder().getAbsolutePath());
+ assertEquals(
+ Collections.singletonList("stringsArray"),
+ model.getModelData()[0].getResolvedSchema().getColumnNames());
+ Table output = model.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ verifyPredictionResult(expectedAlphabeticAscPredictData, predictedResult);
+ }
+
+ @Test
+ public void testGetModelData() throws Exception {
+ StringIndexerModel model =
+ new StringIndexer()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setStringOrderType(StringIndexerParams.ALPHABET_ASC_ORDER)
+ .fit(trainTable);
+ StringIndexerModelData modelData =
+ StringIndexerModelData.getModelDataStream(model.getModelData()[0])
Review comment:
Given that the focus of this test is on the `getModelData(..)` API, could we also verify that the model data table has the expected schema?
Same for `IndexToStringModelTest::testGetModelData()`.
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
+ String FREQUENCY_DESC_ORDER = "frequencyDesc";
+ String FREQUENCY_ASC_ORDER = "frequencyAsc";
+ String ALPHABET_DESC_ORDER = "alphabetDesc";
+ String ALPHABET_ASC_ORDER = "alphabetAsc";
+
+ /** Param for {@link StringIndexer} to decide the order of the strings of each column. */
+ Param<String> STRING_ORDER_TYPE =
Review comment:
Should we add Javadoc to explain the semantics for each of these options? Maybe follow Spark's Javadoc for example.
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
+ * the same. The indices are in [0, numDistinctValuesThisColumn).
+ *
+ * <p>The input columns are cast to string if they are numeric values. By default, the output model
+ * is random ordered. Users can control this by setting {@link
+ * StringIndexerParams#STRING_ORDER_TYPE}.
+ */
+public class StringIndexer
+ implements Estimator<StringIndexer, StringIndexerModel>,
+ StringIndexerParams<StringIndexer> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+
+ public StringIndexer() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public void save(String path) throws IOException {
+ ReadWriteUtils.saveMetadata(this, path);
+ }
+
+ public static StringIndexer load(StreamExecutionEnvironment env, String path)
+ throws IOException {
+ return ReadWriteUtils.loadStageParam(path);
+ }
+
+ @Override
+ public Map<Param<?>, Object> getParamMap() {
+ return paramMap;
+ }
+
+ @Override
+ public StringIndexerModel fit(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment();
+
+ DataStream<Tuple2<Integer, String>> columnIdAndString =
+ tEnv.toDataStream(inputs[0]).flatMap(new ExtractColumnIdAndString(inputCols));
+ DataStream<Tuple3<Integer, String, Long>> columnIdAndStringAndCnt =
+ DataStreamUtils.mapPartition(
+ columnIdAndString.keyBy(
+ (KeySelector<Tuple2<Integer, String>, Integer>) Tuple2::hashCode),
+ new CountStringsByColumn(inputCols.length));
+ DataStream<StringIndexerModelData> modelData =
+ DataStreamUtils.mapPartition(
+ columnIdAndStringAndCnt,
+ new GenerateModel(inputCols.length, getStringOrderType()));
+ modelData.getTransformation().setParallelism(1);
+ StringIndexerModel model =
+ new StringIndexerModel().setModelData(tEnv.fromDataStream(modelData));
+ ReadWriteUtils.updateExistingParams(model, paramMap);
+ return model;
+ }
+
+ /**
+ * Merges all the extracted strings and generates the {@link StringIndexerModelData} according
+ * to the specified string order type.
+ */
+ private static class GenerateModel
+ implements MapPartitionFunction<Tuple3<Integer, String, Long>, StringIndexerModelData> {
+ private final int numCols;
+ private final String stringOrderType;
+
+ public GenerateModel(int numCols, String stringOrderType) {
+ this.numCols = numCols;
+ this.stringOrderType = stringOrderType;
+ }
+
+ @Override
+ public void mapPartition(
+ Iterable<Tuple3<Integer, String, Long>> values,
+ Collector<StringIndexerModelData> out) {
+ String[][] stringsArray = new String[numCols][];
+ ArrayList<Tuple2<String, Long>>[] stringsAndCntsByColumn = new ArrayList[numCols];
+ for (int i = 0; i < numCols; i++) {
+ stringsAndCntsByColumn[i] = new ArrayList<>();
+ }
+ for (Tuple3<Integer, String, Long> colIdAndStringAndCnt : values) {
+ stringsAndCntsByColumn[colIdAndStringAndCnt.f0].add(
+ Tuple2.of(colIdAndStringAndCnt.f1, colIdAndStringAndCnt.f2));
+ }
+ for (int i = 0; i < stringsAndCntsByColumn.length; i++) {
+ List<Tuple2<String, Long>> stringsAndCnts = stringsAndCntsByColumn[i];
+ switch (stringOrderType) {
+ case StringIndexerParams.ALPHABET_ASC_ORDER:
+ stringsAndCnts.sort(Comparator.comparing(valAndCnt -> valAndCnt.f0));
+ break;
+ case StringIndexerParams.ALPHABET_DESC_ORDER:
+ stringsAndCnts.sort(
+ (valAndCnt1, valAndCnt2) ->
+ -valAndCnt1.f0.compareTo(valAndCnt2.f0));
+ break;
+ case StringIndexerParams.FREQUENCY_ASC_ORDER:
+ stringsAndCnts.sort(Comparator.comparing(valAndCnt -> valAndCnt.f1));
+ break;
+ case StringIndexerParams.FREQUENCY_DESC_ORDER:
+ stringsAndCnts.sort(
+ (valAndCnt1, valAndCnt2) ->
+ -valAndCnt1.f1.compareTo(valAndCnt2.f1));
+ break;
+ case StringIndexerParams.RANDOM_ORDER:
+ break;
+ default:
+ throw new RuntimeException(
+ "Unsupported string order type: "
+ + stringOrderType
+ + ". Supported options are: ["
+ + StringIndexerParams.RANDOM_ORDER
+ + ", "
+ + StringIndexerParams.ALPHABET_ASC_ORDER
+ + ", "
+ + StringIndexerParams.ALPHABET_DESC_ORDER
+ + ", "
+ + StringIndexerParams.FREQUENCY_ASC_ORDER
+ + ", "
+ + StringIndexerParams.ALPHABET_DESC_ORDER
+ + "].");
+ }
+ stringsArray[i] = new String[stringsAndCnts.size()];
+ for (int stringId = 0; stringId < stringsArray[i].length; stringId++) {
+ stringsArray[i][stringId] = stringsAndCnts.get(stringId).f0;
+ }
+ }
+
+ out.collect(new StringIndexerModelData(stringsArray));
+ }
+ }
+
+ /** Computes the frequency of strings in each column. */
+ private static class CountStringsByColumn
+ implements MapPartitionFunction<
+ Tuple2<Integer, String>, Tuple3<Integer, String, Long>> {
+ private final int numCols;
+
+ public CountStringsByColumn(int numCols) {
+ this.numCols = numCols;
+ }
+
+ @Override
+ public void mapPartition(
+ Iterable<Tuple2<Integer, String>> values,
+ Collector<Tuple3<Integer, String, Long>> out) {
+ HashMap<String, Long>[] string2CntByColumn = new HashMap[numCols];
+ for (int i = 0; i < numCols; i++) {
+ string2CntByColumn[i] = new HashMap<>();
+ }
+ for (Tuple2<Integer, String> columnIdAndString : values) {
+ int colId = columnIdAndString.f0;
+ String stringVal = columnIdAndString.f1;
+ if (string2CntByColumn[colId].containsKey(stringVal)) {
+ string2CntByColumn[colId].put(
+ stringVal, string2CntByColumn[colId].get(stringVal) + 1);
+ } else {
+ string2CntByColumn[colId].put(stringVal, 1L);
+ }
+ }
+ for (int i = 0; i < numCols; i++) {
+ if (string2CntByColumn[i].size() != 0) {
Review comment:
Would it be simpler to remove the `if (...)` here?
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
Review comment:
Strictly speaking it is not random. Random means when the program is executed multiple time with exactly the same input, the output will likely still be different. This usually happen when the code explicitly permutes the output before returning output to the caller.
It seems better to name it `arbitrary`. What do you think?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831808440
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
Review comment:
Hmm, I would say it is indeed `random`. If you run the program with `RANDOM_ORDER` with the same input multiple times, you are supposed to get different model data.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831812871
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
+ String FREQUENCY_DESC_ORDER = "frequencyDesc";
+ String FREQUENCY_ASC_ORDER = "frequencyAsc";
+ String ALPHABET_DESC_ORDER = "alphabetDesc";
+ String ALPHABET_ASC_ORDER = "alphabetAsc";
+
+ /** Param for {@link StringIndexer} to decide the order of the strings of each column. */
+ Param<String> STRING_ORDER_TYPE =
Review comment:
Thanks for pointing this out. The java doc is added.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r832775562
##########
File path: flink-ml-lib/src/test/java/org/apache/flink/ml/feature/stringindexer/IndexToStringModelTest.java
##########
@@ -0,0 +1,165 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.restartstrategy.RestartStrategies;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.ml.util.StageTestUtils;
+import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.test.util.AbstractTestBase;
+import org.apache.flink.types.Row;
+
+import org.apache.commons.collections.IteratorUtils;
+import org.junit.Before;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.List;
+
+import static org.junit.Assert.assertArrayEquals;
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.fail;
+
+/** Tests the {@link IndexToStringModel}. */
+public class IndexToStringModelTest extends AbstractTestBase {
+ @Rule public final TemporaryFolder tempFolder = new TemporaryFolder();
+
+ private StreamExecutionEnvironment env;
+ private StreamTableEnvironment tEnv;
+ private Table predictTable;
+ private Table modelTable;
+ private Table predictTableWithUnseenValues;
+
+ private final List<Row> expectedPrediction =
+ Arrays.asList(Row.of(0, 3, "a", "2.0"), Row.of(1, 2, "b", "1.0"));
+ private final String[][] stringArrays =
+ new String[][] {{"a", "b", "c", "d"}, {"-1.0", "0.0", "1.0", "2.0"}};
+
+ @Before
+ public void before() {
+ Configuration config = new Configuration();
+ config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true);
+ env = StreamExecutionEnvironment.getExecutionEnvironment(config);
+ env.setParallelism(4);
+ env.enableCheckpointing(100);
+ env.setRestartStrategy(RestartStrategies.noRestart());
+ tEnv = StreamTableEnvironment.create(env);
+
+ modelTable =
+ tEnv.fromDataStream(env.fromElements(new StringIndexerModelData(stringArrays)))
+ .as("stringArrays");
+ predictTable =
+ tEnv.fromDataStream(env.fromCollection(Arrays.asList(Row.of(0, 3), Row.of(1, 2))))
+ .as("inputCol1", "inputCol2");
+ predictTableWithUnseenValues =
+ tEnv.fromDataStream(
+ env.fromCollection(
+ Arrays.asList(Row.of(0, 3), Row.of(1, 2), Row.of(4, 1))))
+ .as("inputCol1", "inputCol2");
+ }
+
+ @Test
+ public void testPredictParam() {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTable)[0];
+ assertEquals(
+ Arrays.asList("inputCol1", "inputCol2", "outputCol1", "outputCol2"),
+ output.getResolvedSchema().getColumnNames());
+ }
+
+ @Test
+ public void testInputWithUnseenValues() {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTableWithUnseenValues)[0];
+ try {
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ fail();
+ } catch (Exception e) {
+ assertEquals(
+ "The input contains unseen index: 4.",
+ e.getCause().getCause().getCause().getCause().getCause().getMessage());
+ }
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testPredict() throws Exception {
+ IndexToStringModel indexToStringModel =
+ new IndexToStringModel()
+ .setInputCols("inputCol1", "inputCol2")
+ .setOutputCols("outputCol1", "outputCol2")
+ .setModelData(modelTable);
+ Table output = indexToStringModel.transform(predictTable)[0];
+ List<Row> predictedResult =
+ IteratorUtils.toList(tEnv.toDataStream(output).executeAndCollect());
+ StringIndexerTest.verifyPredictionResult(expectedPrediction, predictedResult);
+ }
+
+ @Test
+ @SuppressWarnings("unchecked")
+ public void testSaveLoadAndPredict() throws Exception {
Review comment:
Thanks for pointing this out. I have done a pass for all code snippets in this PR.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831838053
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
Review comment:
Hmm... do you mean that even if there is only one partition and all operator has parallelism, the program could still produce different inputs when it is invoked on the same input multiple times?
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] lindong28 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
lindong28 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831834508
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexer.java
##########
@@ -0,0 +1,247 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+import org.apache.flink.api.common.functions.MapPartitionFunction;
+import org.apache.flink.api.java.functions.KeySelector;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.api.java.tuple.Tuple3;
+import org.apache.flink.ml.api.Estimator;
+import org.apache.flink.ml.common.datastream.DataStreamUtils;
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.util.ParamUtils;
+import org.apache.flink.ml.util.ReadWriteUtils;
+import org.apache.flink.streaming.api.datastream.DataStream;
+import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
+import org.apache.flink.table.api.Table;
+import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;
+import org.apache.flink.table.api.internal.TableImpl;
+import org.apache.flink.types.Row;
+import org.apache.flink.util.Collector;
+import org.apache.flink.util.Preconditions;
+
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * An Estimator which implements the string indexing algorithm.
+ *
+ * <p>A string indexer maps each input column (string/numerical value) to a index column (integer
+ * value) such that if the indices of two input are same iff their corresponding input columns are
Review comment:
It looks much better now.
Some followup minor comments:
one/multiple -> one or more
two data points are same -> two data points are the same
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [flink-ml] zhipeng93 commented on a change in pull request #52: [FLINK-25527] Add Transformer and Estimator for StringIndexer
Posted by GitBox <gi...@apache.org>.
zhipeng93 commented on a change in pull request #52:
URL: https://github.com/apache/flink-ml/pull/52#discussion_r831845494
##########
File path: flink-ml-lib/src/main/java/org/apache/flink/ml/feature/stringindexer/StringIndexerParams.java
##########
@@ -0,0 +1,57 @@
+/*
+ * 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.flink.ml.feature.stringindexer;
+
+import org.apache.flink.ml.param.Param;
+import org.apache.flink.ml.param.ParamValidators;
+import org.apache.flink.ml.param.StringParam;
+
+/**
+ * Params of {@link StringIndexer}.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface StringIndexerParams<T> extends StringIndexerModelParams<T> {
+ String RANDOM_ORDER = "random";
Review comment:
If there is only one partition, the result is not `random`.
If the parallelism of the job is not one, then the program produces different outputs when it is invoked on the same input multiple times.
--
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.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org