fhueske commented on a change in pull request #8064: [FLINK-7244] Add parquet table source URL: https://github.com/apache/flink/pull/8064#discussion_r276003882
########## File path: flink-formats/flink-parquet/src/main/java/org/apache/flink/formats/parquet/ParquetTableSource.java ########## @@ -0,0 +1,482 @@ +/* + * 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.formats.parquet; + +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.api.common.typeinfo.BasicTypeInfo; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.api.java.DataSet; +import org.apache.flink.api.java.ExecutionEnvironment; +import org.apache.flink.api.java.tuple.Tuple2; +import org.apache.flink.api.java.typeutils.RowTypeInfo; +import org.apache.flink.core.fs.Path; +import org.apache.flink.formats.parquet.utils.ParquetSchemaConverter; +import org.apache.flink.table.api.TableSchema; +import org.apache.flink.table.expressions.And; +import org.apache.flink.table.expressions.Attribute; +import org.apache.flink.table.expressions.BinaryComparison; +import org.apache.flink.table.expressions.BinaryExpression; +import org.apache.flink.table.expressions.EqualTo; +import org.apache.flink.table.expressions.Expression; +import org.apache.flink.table.expressions.GreaterThan; +import org.apache.flink.table.expressions.GreaterThanOrEqual; +import org.apache.flink.table.expressions.LessThan; +import org.apache.flink.table.expressions.LessThanOrEqual; +import org.apache.flink.table.expressions.Literal; +import org.apache.flink.table.expressions.Not; +import org.apache.flink.table.expressions.NotEqualTo; +import org.apache.flink.table.expressions.Or; +import org.apache.flink.table.sources.BatchTableSource; +import org.apache.flink.table.sources.FilterableTableSource; +import org.apache.flink.table.sources.ProjectableTableSource; +import org.apache.flink.table.sources.TableSource; +import org.apache.flink.types.Row; +import org.apache.flink.util.Preconditions; + +import org.apache.hadoop.conf.Configuration; +import org.apache.parquet.filter2.predicate.FilterApi; +import org.apache.parquet.filter2.predicate.FilterPredicate; +import org.apache.parquet.filter2.predicate.Operators; +import org.apache.parquet.filter2.predicate.Operators.BinaryColumn; +import org.apache.parquet.filter2.predicate.Operators.BooleanColumn; +import org.apache.parquet.filter2.predicate.Operators.Column; +import org.apache.parquet.filter2.predicate.Operators.DoubleColumn; +import org.apache.parquet.filter2.predicate.Operators.FloatColumn; +import org.apache.parquet.filter2.predicate.Operators.IntColumn; +import org.apache.parquet.filter2.predicate.Operators.LongColumn; +import org.apache.parquet.io.api.Binary; +import org.apache.parquet.schema.MessageType; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.util.List; + +/** + * A TableSource to read Parquet files. + * + * <p>The {@link ParquetTableSource} supports projection and filter push-down.</p> + * + * <p>An {@link ParquetTableSource} is used as shown in the example below. + */ +public class ParquetTableSource + implements BatchTableSource<Row>, FilterableTableSource<Row>, ProjectableTableSource<Row> { + + private static final Logger LOG = LoggerFactory.getLogger(ParquetTableSource.class); + + // path to read Parquet files from + private final String path; + // schema of the Parquet file + private final MessageType parquetSchema; + // the schema of table + private final TableSchema tableSchema; + // the configuration to read the file + private final Configuration parquetConfig; + + // type information of the data returned by the InputFormat + private final RowTypeInfo typeInfo; + // list of selected Parquet fields to return + private final int[] selectedFields; + // list of predicates to apply + private final FilterPredicate predicate; + // flag whether a path is recursively enumerated + private final boolean recursiveEnumeration; + + private ParquetTableSource(String path, MessageType parquetSchema, Configuration configuration, + boolean recursiveEnumeration) { + this(path, parquetSchema, configuration, recursiveEnumeration, null, null); + } + + private ParquetTableSource(String path, MessageType parquetSchema, Configuration configuration, + boolean recursiveEnumeration, int[] selectedFields, FilterPredicate predicate) { + Preconditions.checkNotNull(path, "Path must not be null."); + Preconditions.checkNotNull(parquetSchema, "ParquetSchema must not be null."); + Preconditions.checkNotNull(configuration, "Configuration must not be null"); + this.path = path; + this.parquetSchema = parquetSchema; + this.parquetConfig = configuration; + this.selectedFields = selectedFields; + this.predicate = predicate; + this.recursiveEnumeration = recursiveEnumeration; + + // determine the type information from the Parquet schema + RowTypeInfo typeInfoFromSchema = (RowTypeInfo) ParquetSchemaConverter.fromParquetType(parquetSchema); + + // set return type info + if (selectedFields == null) { + this.typeInfo = typeInfoFromSchema; + } else { + this.typeInfo = RowTypeInfo.projectFields(typeInfoFromSchema, selectedFields); + } + + // create a TableSchema that corresponds to the Parquet schema + this.tableSchema = new TableSchema( + typeInfoFromSchema.getFieldNames(), + typeInfoFromSchema.getFieldTypes() + ); + } + + @Override + public TableSource<Row> projectFields(int[] fields) { + return new ParquetTableSource(path, parquetSchema, parquetConfig, recursiveEnumeration, fields, null); + } + + @Override + public DataSet<Row> getDataSet(ExecutionEnvironment executionEnvironment) { + ParquetRowInputFormat parquetRowInputFormat = buildParquetInputFormat(); + parquetRowInputFormat.setNestedFileEnumeration(recursiveEnumeration); + if (selectedFields != null) { + parquetRowInputFormat.selectFields(typeInfo.getFieldNames()); + } + + if (predicate != null) { + parquetRowInputFormat.setFilterPredicate(predicate); + } + return executionEnvironment.createInput(parquetRowInputFormat).name(explainSource()); + } + + @Override + public TableSource<Row> applyPredicate(List<Expression> predicates) { + FilterPredicate predicate = null; + if (predicates.size() > 0) { + predicate = toParquetPredicate(predicates.get(0)); + + for (Expression expression : predicates.subList(1, predicates.size())) { + FilterPredicate right = toParquetPredicate(expression); + if (predicate != null && right != null) { + predicate = FilterApi.and(predicate, right); + } else if (right != null) { + predicate = right; + } + } + } + + return new ParquetTableSource(path, parquetSchema, this.parquetConfig, recursiveEnumeration, selectedFields, predicate); + } + + @Override + public boolean isFilterPushedDown() { + return predicate != null; + } + + @Override + public TypeInformation<Row> getReturnType() { + return typeInfo; + } + + @Override + public TableSchema getTableSchema() { + return tableSchema; + } + + @Override + public String explainSource() { + return "ParquetFile[path=" + path + ", schema=" + parquetSchema + ", filter=" + predicateString() + "]"; + } + + private String predicateString() { + if (predicate != null) { + return predicate.toString(); + } else { + return "TRUE"; + } + } + + private FilterPredicate toParquetPredicate(Expression exp) { + if (exp instanceof Not) { + FilterPredicate c = toParquetPredicate(((Not) exp).child()); + if (c == null) { + return null; + } else { + return FilterApi.not(c); + } + } else if (exp instanceof BinaryComparison) { + BinaryComparison binComp = (BinaryComparison) exp; + + if (!isValid(binComp)) { + // unsupported literal Type + LOG.debug("Unsupported predict [{}] cannot be pushed to ParquetTableSource.", exp); + return null; + } + + Tuple2<Column, Comparable> columnPair = getParquetColumn(binComp); + + if (columnPair != null) { + if (exp instanceof EqualTo) { + if (columnPair.f0 instanceof IntColumn) { + return FilterApi.eq((IntColumn) columnPair.f0, (Integer) columnPair.f1); + } else if (columnPair.f0 instanceof Operators.LongColumn) { + return FilterApi.eq((Operators.LongColumn) columnPair.f0, (Long) columnPair.f1); + } else if (columnPair.f0 instanceof DoubleColumn) { + return FilterApi.eq((DoubleColumn) columnPair.f0, (Double) columnPair.f1); + } else if (columnPair.f0 instanceof FloatColumn) { + return FilterApi.eq((FloatColumn) columnPair.f0, (Float) columnPair.f1); + } else if (columnPair.f0 instanceof BooleanColumn) { + return FilterApi.eq((BooleanColumn) columnPair.f0, (Boolean) columnPair.f1); + } else if (columnPair.f0 instanceof BinaryColumn) { + return FilterApi.eq((BinaryColumn) columnPair.f0, (Binary) columnPair.f1); + } + } else if (exp instanceof NotEqualTo) { + if (columnPair.f0 instanceof IntColumn) { + return FilterApi.notEq((IntColumn) columnPair.f0, (Integer) columnPair.f1); + } else if (columnPair.f0 instanceof LongColumn) { + return FilterApi.notEq((Operators.LongColumn) columnPair.f0, (Long) columnPair.f1); + } else if (columnPair.f0 instanceof DoubleColumn) { + return FilterApi.notEq((DoubleColumn) columnPair.f0, (Double) columnPair.f1); + } else if (columnPair.f0 instanceof FloatColumn) { + return FilterApi.notEq((FloatColumn) columnPair.f0, (Float) columnPair.f1); + } else if (columnPair.f0 instanceof BooleanColumn) { + return FilterApi.notEq((BooleanColumn) columnPair.f0, (Boolean) columnPair.f1); + } else if (columnPair.f0 instanceof BinaryColumn) { + return FilterApi.notEq((BinaryColumn) columnPair.f0, (Binary) columnPair.f1); + } + } else if (exp instanceof GreaterThan) { + if (columnPair.f0 instanceof IntColumn) { + return FilterApi.gt((IntColumn) columnPair.f0, (Integer) columnPair.f1); + } else if (columnPair.f0 instanceof LongColumn) { + return FilterApi.gt((Operators.LongColumn) columnPair.f0, (Long) columnPair.f1); + } else if (columnPair.f0 instanceof DoubleColumn) { + return FilterApi.gt((DoubleColumn) columnPair.f0, (Double) columnPair.f1); + } else if (columnPair.f0 instanceof FloatColumn) { + return FilterApi.gt((FloatColumn) columnPair.f0, (Float) columnPair.f1); + } + } else if (exp instanceof GreaterThanOrEqual) { + if (columnPair.f0 instanceof IntColumn) { + return FilterApi.gtEq((IntColumn) columnPair.f0, (Integer) columnPair.f1); + } else if (columnPair.f0 instanceof LongColumn) { + return FilterApi.gtEq((Operators.LongColumn) columnPair.f0, (Long) columnPair.f1); + } else if (columnPair.f0 instanceof DoubleColumn) { + return FilterApi.gtEq((DoubleColumn) columnPair.f0, (Double) columnPair.f1); + } else if (columnPair.f0 instanceof FloatColumn) { + return FilterApi.gtEq((FloatColumn) columnPair.f0, (Float) columnPair.f1); + } + } else if (exp instanceof LessThan) { + if (columnPair.f0 instanceof IntColumn) { + return FilterApi.lt((IntColumn) columnPair.f0, (Integer) columnPair.f1); + } else if (columnPair.f0 instanceof LongColumn) { + return FilterApi.lt((Operators.LongColumn) columnPair.f0, (Long) columnPair.f1); + } else if (columnPair.f0 instanceof DoubleColumn) { + return FilterApi.lt((DoubleColumn) columnPair.f0, (Double) columnPair.f1); + } else if (columnPair.f0 instanceof FloatColumn) { + return FilterApi.lt((FloatColumn) columnPair.f0, (Float) columnPair.f1); + } + } else if (exp instanceof LessThanOrEqual) { + if (columnPair.f0 instanceof IntColumn) { + return FilterApi.ltEq((IntColumn) columnPair.f0, (Integer) columnPair.f1); + } else if (columnPair.f0 instanceof LongColumn) { + return FilterApi.ltEq((Operators.LongColumn) columnPair.f0, (Long) columnPair.f1); + } else if (columnPair.f0 instanceof DoubleColumn) { + return FilterApi.ltEq((DoubleColumn) columnPair.f0, (Double) columnPair.f1); + } else if (columnPair.f0 instanceof FloatColumn) { + return FilterApi.ltEq((FloatColumn) columnPair.f0, (Float) columnPair.f1); + } + } else { + // Unsupported Predicate + LOG.debug("Unsupported predicate [{}] cannot be pushed into ParquetTableSource.", exp); + return null; + } + } + } else if (exp instanceof BinaryExpression) { + FilterPredicate c1 = toParquetPredicate(((Or) exp).left()); + FilterPredicate c2 = toParquetPredicate(((Or) exp).right()); + + if (c1 == null || c2 == null) { + return null; + } else { + if (exp instanceof Or) { + return FilterApi.or(c1, c2); + } else if (exp instanceof And) { + return FilterApi.and(c1, c2); + } else { + // Unsupported Predicate + LOG.debug("Unsupported predicate [{}] cannot be pushed into ParquetTableSource.", exp); + return null; + } + } + } + + return null; + } + + private boolean isValid(BinaryComparison comp) { + return (comp.left() instanceof Literal && comp.right() instanceof Attribute) || + (comp.left() instanceof Attribute && comp.right() instanceof Literal); + } + + private boolean literalOnRight(BinaryComparison comp) { + if (comp.left() instanceof Literal && comp.right() instanceof Attribute) { Review comment: remove double space (twice) ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
