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 2022/04/08 02:25:17 UTC
[GitHub] [flink-ml] yunfengzhou-hub commented on a diff in pull request #82: [FLINK-27072] Add Transformer of Bucketizer
yunfengzhou-hub commented on code in PR #82:
URL: https://github.com/apache/flink-ml/pull/82#discussion_r845668376
##########
flink-ml-core/src/main/java/org/apache/flink/ml/param/ArrayArrayParam.java:
##########
@@ -0,0 +1,42 @@
+/*
+ * 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.param;
+
+import org.apache.flink.ml.util.ReadWriteUtils;
+
+import java.io.IOException;
+
+/** Class for the arrayOfArray-type parameters. */
+public class ArrayArrayParam<T> extends Param<T[][]> {
Review Comment:
Shall we just use a 1D array to represent arrays of 2 or more dimensions, or figure out another way to represent arrays of arbitrary number of dimensions? I'm not sure we would want to create a `Param` subclass for 3D, 4D parameters.
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/onehotencoder/OneHotEncoderParams.java:
##########
@@ -27,6 +27,8 @@
/**
* Params of OneHotEncoderModel.
*
+ * <p>The `keep` option of {@link HasHandleInvalid} is not supported in {@link OneHotEncoderParams}.
Review Comment:
It seems that `OneHotEncoder` only supports the `ERROR` option.
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/bucketizer/BucketizerParams.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.bucketizer;
+
+import org.apache.flink.ml.common.param.HasHandleInvalid;
+import org.apache.flink.ml.common.param.HasInputCols;
+import org.apache.flink.ml.common.param.HasOutputCols;
+import org.apache.flink.ml.param.DoubleArrayArrayParam;
+import org.apache.flink.ml.param.ParamValidators;
+
+/**
+ * Params for {@link Bucketizer}.
+ *
+ * <p>The `keep` option of {@link HasHandleInvalid} means that we put the invalid data in the last
+ * bucket of the splits, whose index is the number of the buckets.
+ *
+ * @param <T> The class type of this instance.
+ */
+public interface BucketizerParams<T>
+ extends HasInputCols<T>, HasOutputCols<T>, HasHandleInvalid<T> {
+ /**
+ * The array of split points for mapping continuous features into buckets for multiple columns.
+ *
+ * <p>Each input column is supposed to be mapped into {numberOfSplitPoints - 1} buckets. A
+ * bucket is defined by two split points. For example, bucket(x,y) contains values in the range
+ * [x,y). An exception is that the last bucket also contains y. The array should contain at
+ * least three split points and be strictly increasing.
+ */
+ DoubleArrayArrayParam SPLITS_ARRAY =
+ new DoubleArrayArrayParam(
+ "splitsArray", "Array of split points.", null, ParamValidators.nonEmptyArray());
Review Comment:
Shall we have a more detailed description for this parameter, like the first line in Javadoc?
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/onehotencoder/OneHotEncoder.java:
##########
@@ -59,7 +58,7 @@ public OneHotEncoder() {
@Override
public OneHotEncoderModel fit(Table... inputs) {
Preconditions.checkArgument(inputs.length == 1);
- Preconditions.checkArgument(getHandleInvalid().equals(HasHandleInvalid.ERROR_INVALID));
+ Preconditions.checkArgument(getHandleInvalid().equals(OneHotEncoderParams.ERROR_INVALID));
Review Comment:
Shall we just use `ERROR_INVALID`? `OneHotEncoder` is a subclass of `OneHotEncoderParams`. Same for other classes like VectorAssembler.
##########
flink-ml-lib/src/test/java/org/apache/flink/ml/feature/OneHotEncoderTest.java:
##########
@@ -205,7 +205,7 @@ public void testInputDataType() {
@Test
public void testNotSupportedHandleInvalidOptions() {
- estimator.setHandleInvalid(HasHandleInvalid.SKIP_INVALID);
+ estimator.setHandleInvalid(OneHotEncoderParams.SKIP_INVALID);
Review Comment:
I'm not sure `OneHotEncoderParams.SKIP_INVALID` would be a better choice than `HasHandleInvalid.SKIP_INVALID`. Could you please illustrate why would you like to make this change?
I think `HasHandleInvalid` might be easier for users. If they want to `setHandleInvalid`, they know that the available options are at `HasHandleInvalid`. When they use `setBatchStrategy`, the options are at `HasBatchStrategy`. This is more straightforward. `xxxParams`, on the other hand, would provide all public fields it has defined or inherited, and users have to depend on more information to judge which ones are valid options.
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/bucketizer/Bucketizer.java:
##########
@@ -0,0 +1,179 @@
+/*
+ * 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.bucketizer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+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.Transformer;
+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.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.HashMap;
+import java.util.Map;
+
+/**
+ * Bucketizer is a transformer that maps multiple columns of continuous features to multiple columns
+ * of discrete features, i.e., buckets IDs.
+ */
+public class Bucketizer implements Transformer<Bucketizer>, BucketizerParams<Bucketizer> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+
+ public Bucketizer() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Double[][] splitsArray = getSplitsArray();
+
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ Preconditions.checkArgument(inputCols.length == splitsArray.length);
+ for (Double[] splits : splitsArray) {
+ Preconditions.checkArgument(
+ splits.length >= 3,
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ for (int j = 1; j < splits.length; j++) {
+ Preconditions.checkArgument(
+ splits[j] > splits[j - 1],
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ }
+ }
Review Comment:
Shall we move this part of logic to a `ParamValidator` class? This class can be placed inside `BucketizerParams`.
Moreover, We would always need to do some checks in ParamValidator, and to do the other checks in `fit()`/`transform()`. I think we should place a validation in ParamValidator if it is generally applicable to all subclasses that inherit this parameter. If the validation needs to be computed across parameters, or depends on the input Tables, then it should be placed in `fit()`/`transform()`.
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/bucketizer/Bucketizer.java:
##########
@@ -0,0 +1,179 @@
+/*
+ * 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.bucketizer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+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.Transformer;
+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.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.HashMap;
+import java.util.Map;
+
+/**
+ * Bucketizer is a transformer that maps multiple columns of continuous features to multiple columns
+ * of discrete features, i.e., buckets IDs.
+ */
+public class Bucketizer implements Transformer<Bucketizer>, BucketizerParams<Bucketizer> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+
+ public Bucketizer() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Double[][] splitsArray = getSplitsArray();
+
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ Preconditions.checkArgument(inputCols.length == splitsArray.length);
+ for (Double[] splits : splitsArray) {
+ Preconditions.checkArgument(
+ splits.length >= 3,
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ for (int j = 1; j < splits.length; j++) {
+ Preconditions.checkArgument(
+ splits[j] > splits[j - 1],
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ }
+ }
+
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment();
+
+ 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 =
+ tEnv.toDataStream(inputs[0])
+ .flatMap(
+ new FindBucketFunction(inputCols, splitsArray, getHandleInvalid()),
+ outputTypeInfo);
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Finds the bucket index for each continuous feature of an input data point. */
+ private static class FindBucketFunction implements FlatMapFunction<Row, Row> {
+ private final String[] inputCols;
+ private final String handleInvalid;
+ private final Double[][] splitsArray;
+
+ public FindBucketFunction(
+ String[] inputCols, Double[][] splitsArray, String handleInvalid) {
+ this.inputCols = inputCols;
+ this.splitsArray = splitsArray;
+ this.handleInvalid = handleInvalid;
+ }
+
+ @Override
+ public void flatMap(Row value, Collector<Row> out) {
+ Row outputRow = new Row(inputCols.length);
+
+ for (int i = 0; i < inputCols.length; i++) {
+ double feature = ((Number) value.getField(inputCols[i])).doubleValue();
+ Double[] splits = splitsArray[i];
+ boolean isInvalid = false;
+
+ if (!Double.isNaN(feature)) {
+ int index = Arrays.binarySearch(splits, feature);
+ if (index >= 0) {
+ if (index == inputCols.length - 1) {
+ index--;
+ }
+ outputRow.setField(i, index);
+ } else {
+ index = -index - 1;
+ if (index == 0 || index == inputCols.length) {
+ isInvalid = true;
+ } else {
+ outputRow.setField(i, index - 1);
+ }
+ }
+ } else {
+ isInvalid = true;
+ }
+
+ if (isInvalid) {
+ switch (handleInvalid) {
+ case BucketizerParams.ERROR_INVALID:
+ throw new RuntimeException(
+ "The input contains invalid value. See "
+ + BucketizerParams.HANDLE_INVALID
+ + " parameter for more options.");
+ case BucketizerParams.SKIP_INVALID:
+ return;
+ case BucketizerParams.KEEP_INVALID:
+ outputRow.setField(i, splits.length - 1);
+ break;
Review Comment:
This should be `continue`.
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/bucketizer/Bucketizer.java:
##########
@@ -0,0 +1,179 @@
+/*
+ * 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.bucketizer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+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.Transformer;
+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.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.HashMap;
+import java.util.Map;
+
+/**
+ * Bucketizer is a transformer that maps multiple columns of continuous features to multiple columns
+ * of discrete features, i.e., buckets IDs.
+ */
+public class Bucketizer implements Transformer<Bucketizer>, BucketizerParams<Bucketizer> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+
+ public Bucketizer() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Double[][] splitsArray = getSplitsArray();
+
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ Preconditions.checkArgument(inputCols.length == splitsArray.length);
+ for (Double[] splits : splitsArray) {
+ Preconditions.checkArgument(
+ splits.length >= 3,
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ for (int j = 1; j < splits.length; j++) {
+ Preconditions.checkArgument(
+ splits[j] > splits[j - 1],
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ }
+ }
+
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment();
+
+ 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 =
+ tEnv.toDataStream(inputs[0])
+ .flatMap(
+ new FindBucketFunction(inputCols, splitsArray, getHandleInvalid()),
+ outputTypeInfo);
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Finds the bucket index for each continuous feature of an input data point. */
+ private static class FindBucketFunction implements FlatMapFunction<Row, Row> {
+ private final String[] inputCols;
+ private final String handleInvalid;
+ private final Double[][] splitsArray;
+
+ public FindBucketFunction(
+ String[] inputCols, Double[][] splitsArray, String handleInvalid) {
+ this.inputCols = inputCols;
+ this.splitsArray = splitsArray;
+ this.handleInvalid = handleInvalid;
+ }
+
+ @Override
+ public void flatMap(Row value, Collector<Row> out) {
+ Row outputRow = new Row(inputCols.length);
+
+ for (int i = 0; i < inputCols.length; i++) {
+ double feature = ((Number) value.getField(inputCols[i])).doubleValue();
Review Comment:
If the field value is not a Number, the operator would throw exception, which means that `HasHandleInvalid` is always `ERROR` in this case.
##########
flink-ml-lib/src/main/java/org/apache/flink/ml/feature/bucketizer/Bucketizer.java:
##########
@@ -0,0 +1,179 @@
+/*
+ * 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.bucketizer;
+
+import org.apache.flink.api.common.functions.FlatMapFunction;
+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.Transformer;
+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.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.HashMap;
+import java.util.Map;
+
+/**
+ * Bucketizer is a transformer that maps multiple columns of continuous features to multiple columns
+ * of discrete features, i.e., buckets IDs.
+ */
+public class Bucketizer implements Transformer<Bucketizer>, BucketizerParams<Bucketizer> {
+ private final Map<Param<?>, Object> paramMap = new HashMap<>();
+
+ public Bucketizer() {
+ ParamUtils.initializeMapWithDefaultValues(paramMap, this);
+ }
+
+ @Override
+ public Table[] transform(Table... inputs) {
+ Preconditions.checkArgument(inputs.length == 1);
+ String[] inputCols = getInputCols();
+ String[] outputCols = getOutputCols();
+ Double[][] splitsArray = getSplitsArray();
+
+ Preconditions.checkArgument(inputCols.length == outputCols.length);
+ Preconditions.checkArgument(inputCols.length == splitsArray.length);
+ for (Double[] splits : splitsArray) {
+ Preconditions.checkArgument(
+ splits.length >= 3,
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ for (int j = 1; j < splits.length; j++) {
+ Preconditions.checkArgument(
+ splits[j] > splits[j - 1],
+ "Illegal value for "
+ + BucketizerParams.SPLITS_ARRAY
+ + ". See param "
+ + BucketizerParams.SPLITS_ARRAY
+ + " for details.");
+ }
+ }
+
+ StreamTableEnvironment tEnv =
+ (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment();
+
+ 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 =
+ tEnv.toDataStream(inputs[0])
+ .flatMap(
+ new FindBucketFunction(inputCols, splitsArray, getHandleInvalid()),
+ outputTypeInfo);
+ return new Table[] {tEnv.fromDataStream(result)};
+ }
+
+ /** Finds the bucket index for each continuous feature of an input data point. */
+ private static class FindBucketFunction implements FlatMapFunction<Row, Row> {
+ private final String[] inputCols;
+ private final String handleInvalid;
+ private final Double[][] splitsArray;
+
+ public FindBucketFunction(
+ String[] inputCols, Double[][] splitsArray, String handleInvalid) {
+ this.inputCols = inputCols;
+ this.splitsArray = splitsArray;
+ this.handleInvalid = handleInvalid;
+ }
+
+ @Override
+ public void flatMap(Row value, Collector<Row> out) {
+ Row outputRow = new Row(inputCols.length);
+
+ for (int i = 0; i < inputCols.length; i++) {
+ double feature = ((Number) value.getField(inputCols[i])).doubleValue();
+ Double[] splits = splitsArray[i];
+ boolean isInvalid = false;
+
+ if (!Double.isNaN(feature)) {
+ int index = Arrays.binarySearch(splits, feature);
+ if (index >= 0) {
+ if (index == inputCols.length - 1) {
+ index--;
+ }
+ outputRow.setField(i, index);
+ } else {
+ index = -index - 1;
+ if (index == 0 || index == inputCols.length) {
+ isInvalid = true;
+ } else {
+ outputRow.setField(i, index - 1);
+ }
+ }
+ } else {
+ isInvalid = true;
+ }
+
+ if (isInvalid) {
+ switch (handleInvalid) {
+ case BucketizerParams.ERROR_INVALID:
+ throw new RuntimeException(
+ "The input contains invalid value. See "
+ + BucketizerParams.HANDLE_INVALID
+ + " parameter for more options.");
+ case BucketizerParams.SKIP_INVALID:
+ return;
+ case BucketizerParams.KEEP_INVALID:
+ outputRow.setField(i, splits.length - 1);
+ break;
+ default:
+ throw new IllegalStateException(
+ "Unsupported handleInvalid type: " + handleInvalid);
Review Comment:
Given that we used `BucketizerParams.SPLITS_ARRAY` in other error messages, shall we also use `BucketizerParams.HANDLE_INVALID` here?
--
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