vinishjail97 commented on code in PR #935: URL: https://github.com/apache/incubator-xtable/pull/935#discussion_r4226928628
########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = + TableFormatUtils.getTableDataLocation( + TableFormat.ICEBERG, tableLocation, icebergTable.properties()); + // the Hudi table lives under the data location, which Iceberg resolves from the table + // properties + this.sourceTable = sourceTable.toBuilder().dataPath(dataPath).build(); + this.targetTableProperties = new Properties(); + this.targetTableProperties.putAll(targetTableProperties); + this.targetTableProperties.setProperty( + HudiTargetConfig.EXECUTION_ENGINE, HudiTargetConfig.EXECUTION_ENGINE_SPARK); + // the secondary index is only available from Hudi table version 8 onwards + this.targetTableProperties.setProperty( + HudiTargetConfig.HUDI_TABLE_VERSION, String.valueOf(HoodieTableVersion.NINE.versionCode())); + this.indexedColumns = + HudiTargetConfig.fromProperties(this.targetTableProperties).getSecondaryIndexColumns(); + Preconditions.checkArgument( + !indexedColumns.isEmpty(), + "Set %s to the columns to index", + HudiTargetConfig.SECONDARY_INDEX_COLUMNS); + this.sparkSession = sparkSession; + this.javaSparkContext = JavaSparkContext.fromSparkContext(sparkSession.sparkContext()); + this.engineContext = new HoodieSparkEngineContext(javaSparkContext); + } + + @Override + public boolean doesIndexExist(String columnName) { + return getCompletedIndexPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + } + + @Override + public void syncIndex(Table icebergTable) { + Configuration configuration = javaSparkContext.hadoopConfiguration(); + IcebergConversionSourceProvider sourceProvider = new IcebergConversionSourceProvider(); + sourceProvider.init(configuration); + // The sync reports a failed target in its result instead of throwing. It returns no result + // for the target when the table has no new commits. + SyncResult syncResult = + new ConversionController(configuration) + .sync(getConversionConfig(icebergTable), sourceProvider) + .get(TableFormat.HUDI); + if (syncResult != null + && (syncResult.getTableFormatSyncStatus() == null + || syncResult.getTableFormatSyncStatus().getStatusCode() != SyncStatusCode.SUCCESS)) { + throw new IllegalStateException( + "Failed to sync the secondary indexes of columns " + + indexedColumns + + " of table " + + tableLocation + + ": " + + syncResult.getTableFormatSyncStatus()); + } + Set<String> completedPartitions = getCompletedIndexPartitions(); + List<String> notIndexed = + indexedColumns.stream() + .filter( + column -> + !completedPartitions.contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + column)) + .collect(Collectors.toList()); + if (!notIndexed.isEmpty()) { + throw new IllegalStateException( + "The secondary indexes of columns " + + notIndexed + + " of table " + + tableLocation + + " do not exist. The indexes are built when the index is created, and adding a" + + " column to an existing index is not supported yet."); + } + log.info( + "Synced the secondary indexes of columns {} of table {}", indexedColumns, tableLocation); + } + + /** + * Returns the metadata table partitions whose build has completed. A partition directory can + * exist while the build is in flight or after it failed, so the directory alone does not show + * that the index is usable. + */ + private Set<String> getCompletedIndexPartitions() { + try { + return getMetaClient().getTableConfig().getMetadataPartitions(); + } catch (TableNotFoundException e) { + return Collections.emptySet(); + } + } + + private HoodieTableMetaClient getMetaClient() { + HoodieStorage storage = + new HoodieHadoopStorage(dataPath, javaSparkContext.hadoopConfiguration()); + return HoodieTableMetaClient.builder().setStorage(storage).setBasePath(dataPath).build(); + } + + private ConversionConfig getConversionConfig(Table icebergTable) { + TargetTable targetTable = + TargetTable.builder() + .name(icebergTable.name()) + .basePath(dataPath) + .formatName(TableFormat.HUDI) + .additionalProperties(targetTableProperties) + .build(); + return ConversionConfig.builder() + .sourceTable(sourceTable) + .targetTables(Collections.singletonList(targetTable)) + .build(); + } + + @Override + public Dataset<Row> lookup(Table icebergTable, Dataset<Row> keys, String columnName) { + Preconditions.checkArgument( + indexedColumns.contains(columnName), + "Column %s is not one of the indexed columns %s", + columnName, + indexedColumns); + StructType resultSchema = + new StructType() + .add(columnName, DataTypes.StringType, false) + .add(FILE_COLUMN, DataTypes.StringType, false) + .add(POSITION_COLUMN, DataTypes.LongType, false); + // the secondary index stores the values of the indexed column as strings + JavaRDD<String> secondaryKeys = + keys.where(col(columnName).isNotNull()) + .select(col(columnName).cast(DataTypes.StringType)) + .as(Encoders.STRING()) + .toJavaRDD(); + HoodieTableMetaClient metaClient = getMetaClient(); + HoodieStorage storage = metaClient.getStorage(); + if (!metaClient + .getTableConfig() + .getMetadataPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName)) { + throw new IllegalStateException( + "The secondary index of column " + + columnName + + " of table " + + tableLocation + + " is not built"); + } + HoodieMetadataConfig metadataConfig = Review Comment: The table config does not store the writer's metadata config, but the reader does not read `withSecondaryIndexEnabled` either, so ee95125b removes it. `enable(true)` stays, because without it the factory returns `FileSystemBackedTableMetadata` and the cast fails. ########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = + TableFormatUtils.getTableDataLocation( + TableFormat.ICEBERG, tableLocation, icebergTable.properties()); + // the Hudi table lives under the data location, which Iceberg resolves from the table + // properties + this.sourceTable = sourceTable.toBuilder().dataPath(dataPath).build(); + this.targetTableProperties = new Properties(); + this.targetTableProperties.putAll(targetTableProperties); + this.targetTableProperties.setProperty( + HudiTargetConfig.EXECUTION_ENGINE, HudiTargetConfig.EXECUTION_ENGINE_SPARK); + // the secondary index is only available from Hudi table version 8 onwards + this.targetTableProperties.setProperty( + HudiTargetConfig.HUDI_TABLE_VERSION, String.valueOf(HoodieTableVersion.NINE.versionCode())); + this.indexedColumns = + HudiTargetConfig.fromProperties(this.targetTableProperties).getSecondaryIndexColumns(); + Preconditions.checkArgument( + !indexedColumns.isEmpty(), + "Set %s to the columns to index", + HudiTargetConfig.SECONDARY_INDEX_COLUMNS); + this.sparkSession = sparkSession; + this.javaSparkContext = JavaSparkContext.fromSparkContext(sparkSession.sparkContext()); + this.engineContext = new HoodieSparkEngineContext(javaSparkContext); + } + + @Override + public boolean doesIndexExist(String columnName) { + return getCompletedIndexPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + } + + @Override + public void syncIndex(Table icebergTable) { + Configuration configuration = javaSparkContext.hadoopConfiguration(); + IcebergConversionSourceProvider sourceProvider = new IcebergConversionSourceProvider(); + sourceProvider.init(configuration); + // The sync reports a failed target in its result instead of throwing. It returns no result + // for the target when the table has no new commits. + SyncResult syncResult = + new ConversionController(configuration) + .sync(getConversionConfig(icebergTable), sourceProvider) + .get(TableFormat.HUDI); + if (syncResult != null + && (syncResult.getTableFormatSyncStatus() == null + || syncResult.getTableFormatSyncStatus().getStatusCode() != SyncStatusCode.SUCCESS)) { + throw new IllegalStateException( + "Failed to sync the secondary indexes of columns " + + indexedColumns + + " of table " + + tableLocation + + ": " + + syncResult.getTableFormatSyncStatus()); + } + Set<String> completedPartitions = getCompletedIndexPartitions(); + List<String> notIndexed = + indexedColumns.stream() + .filter( + column -> + !completedPartitions.contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + column)) + .collect(Collectors.toList()); + if (!notIndexed.isEmpty()) { + throw new IllegalStateException( + "The secondary indexes of columns " + + notIndexed + + " of table " + + tableLocation + + " do not exist. The indexes are built when the index is created, and adding a" + + " column to an existing index is not supported yet."); + } + log.info( + "Synced the secondary indexes of columns {} of table {}", indexedColumns, tableLocation); + } + + /** + * Returns the metadata table partitions whose build has completed. A partition directory can + * exist while the build is in flight or after it failed, so the directory alone does not show + * that the index is usable. + */ + private Set<String> getCompletedIndexPartitions() { + try { + return getMetaClient().getTableConfig().getMetadataPartitions(); + } catch (TableNotFoundException e) { + return Collections.emptySet(); + } + } + + private HoodieTableMetaClient getMetaClient() { + HoodieStorage storage = + new HoodieHadoopStorage(dataPath, javaSparkContext.hadoopConfiguration()); + return HoodieTableMetaClient.builder().setStorage(storage).setBasePath(dataPath).build(); + } + + private ConversionConfig getConversionConfig(Table icebergTable) { + TargetTable targetTable = + TargetTable.builder() + .name(icebergTable.name()) + .basePath(dataPath) + .formatName(TableFormat.HUDI) + .additionalProperties(targetTableProperties) + .build(); + return ConversionConfig.builder() + .sourceTable(sourceTable) + .targetTables(Collections.singletonList(targetTable)) + .build(); + } + + @Override + public Dataset<Row> lookup(Table icebergTable, Dataset<Row> keys, String columnName) { + Preconditions.checkArgument( + indexedColumns.contains(columnName), + "Column %s is not one of the indexed columns %s", + columnName, + indexedColumns); + StructType resultSchema = Review Comment: Done in ee95125b. ########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = + TableFormatUtils.getTableDataLocation( + TableFormat.ICEBERG, tableLocation, icebergTable.properties()); + // the Hudi table lives under the data location, which Iceberg resolves from the table + // properties + this.sourceTable = sourceTable.toBuilder().dataPath(dataPath).build(); + this.targetTableProperties = new Properties(); + this.targetTableProperties.putAll(targetTableProperties); + this.targetTableProperties.setProperty( + HudiTargetConfig.EXECUTION_ENGINE, HudiTargetConfig.EXECUTION_ENGINE_SPARK); + // the secondary index is only available from Hudi table version 8 onwards + this.targetTableProperties.setProperty( + HudiTargetConfig.HUDI_TABLE_VERSION, String.valueOf(HoodieTableVersion.NINE.versionCode())); + this.indexedColumns = + HudiTargetConfig.fromProperties(this.targetTableProperties).getSecondaryIndexColumns(); + Preconditions.checkArgument( + !indexedColumns.isEmpty(), + "Set %s to the columns to index", + HudiTargetConfig.SECONDARY_INDEX_COLUMNS); + this.sparkSession = sparkSession; + this.javaSparkContext = JavaSparkContext.fromSparkContext(sparkSession.sparkContext()); + this.engineContext = new HoodieSparkEngineContext(javaSparkContext); + } + + @Override + public boolean doesIndexExist(String columnName) { + return getCompletedIndexPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + } + + @Override + public void syncIndex(Table icebergTable) { + Configuration configuration = javaSparkContext.hadoopConfiguration(); + IcebergConversionSourceProvider sourceProvider = new IcebergConversionSourceProvider(); + sourceProvider.init(configuration); + // The sync reports a failed target in its result instead of throwing. It returns no result + // for the target when the table has no new commits. + SyncResult syncResult = + new ConversionController(configuration) + .sync(getConversionConfig(icebergTable), sourceProvider) + .get(TableFormat.HUDI); + if (syncResult != null + && (syncResult.getTableFormatSyncStatus() == null + || syncResult.getTableFormatSyncStatus().getStatusCode() != SyncStatusCode.SUCCESS)) { + throw new IllegalStateException( + "Failed to sync the secondary indexes of columns " + + indexedColumns + + " of table " + + tableLocation + + ": " + + syncResult.getTableFormatSyncStatus()); + } + Set<String> completedPartitions = getCompletedIndexPartitions(); + List<String> notIndexed = + indexedColumns.stream() + .filter( + column -> + !completedPartitions.contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + column)) + .collect(Collectors.toList()); + if (!notIndexed.isEmpty()) { + throw new IllegalStateException( + "The secondary indexes of columns " + + notIndexed + + " of table " + + tableLocation + + " do not exist. The indexes are built when the index is created, and adding a" + + " column to an existing index is not supported yet."); + } + log.info( + "Synced the secondary indexes of columns {} of table {}", indexedColumns, tableLocation); + } + + /** + * Returns the metadata table partitions whose build has completed. A partition directory can + * exist while the build is in flight or after it failed, so the directory alone does not show + * that the index is usable. + */ + private Set<String> getCompletedIndexPartitions() { + try { + return getMetaClient().getTableConfig().getMetadataPartitions(); + } catch (TableNotFoundException e) { + return Collections.emptySet(); + } + } + + private HoodieTableMetaClient getMetaClient() { + HoodieStorage storage = + new HoodieHadoopStorage(dataPath, javaSparkContext.hadoopConfiguration()); + return HoodieTableMetaClient.builder().setStorage(storage).setBasePath(dataPath).build(); + } + + private ConversionConfig getConversionConfig(Table icebergTable) { + TargetTable targetTable = + TargetTable.builder() + .name(icebergTable.name()) + .basePath(dataPath) + .formatName(TableFormat.HUDI) + .additionalProperties(targetTableProperties) + .build(); + return ConversionConfig.builder() + .sourceTable(sourceTable) + .targetTables(Collections.singletonList(targetTable)) + .build(); + } + + @Override + public Dataset<Row> lookup(Table icebergTable, Dataset<Row> keys, String columnName) { + Preconditions.checkArgument( + indexedColumns.contains(columnName), + "Column %s is not one of the indexed columns %s", + columnName, + indexedColumns); + StructType resultSchema = + new StructType() + .add(columnName, DataTypes.StringType, false) + .add(FILE_COLUMN, DataTypes.StringType, false) + .add(POSITION_COLUMN, DataTypes.LongType, false); + // the secondary index stores the values of the indexed column as strings + JavaRDD<String> secondaryKeys = + keys.where(col(columnName).isNotNull()) + .select(col(columnName).cast(DataTypes.StringType)) + .as(Encoders.STRING()) + .toJavaRDD(); + HoodieTableMetaClient metaClient = getMetaClient(); + HoodieStorage storage = metaClient.getStorage(); + if (!metaClient + .getTableConfig() + .getMetadataPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName)) { + throw new IllegalStateException( + "The secondary index of column " + + columnName + + " of table " + + tableLocation + + " is not built"); + } + HoodieMetadataConfig metadataConfig = + HoodieMetadataConfig.newBuilder().enable(true).withSecondaryIndexEnabled(true).build(); + try (HoodieBackedTableMetadata tableMetadata = + (HoodieBackedTableMetadata) + metaClient + .getTableFormat() + .getMetadataFactory() + .create(engineContext, storage, metadataConfig, dataPath)) { + HoodiePairData<String, String> recordKeysBySecondaryKey = + tableMetadata.readSecondaryIndexDataTableRecordKeysWithKeys( + HoodieJavaRDD.of(secondaryKeys), PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + // capture the field so the closure does not serialize the index instance + String dataPath = this.dataPath; + JavaRDD<Row> lookupResults = + toJavaPairRDD(recordKeysBySecondaryKey) + .map( + recordKeyBySecondaryKey -> + toLookupResult( + dataPath, recordKeyBySecondaryKey._1(), recordKeyBySecondaryKey._2())); + return sparkSession.createDataFrame(lookupResults, resultSchema); + } + } + + /** + * Hudi returns an RDD backed result for keys in an RDD, but a list backed result when there are + * no keys to look up, for example when every key is null. + */ + private JavaPairRDD<String, String> toJavaPairRDD(HoodiePairData<String, String> pairData) { + if (pairData instanceof HoodieJavaPairRDD) { + return HoodieJavaPairRDD.getJavaPairRDD(pairData); + } + List<Tuple2<String, String>> pairs = + pairData.collectAsList().stream() + .map(pair -> new Tuple2<>(pair.getKey(), pair.getValue())) + .collect(Collectors.toList()); Review Comment: Hudi returns a list only for no keys and for an index with one file group. In the one-group case Hudi already collects the keys and reads that file group on the driver, and one group needs a minimum below the default of 10. ee95125b corrects the comment to name both cases. ########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = + TableFormatUtils.getTableDataLocation( + TableFormat.ICEBERG, tableLocation, icebergTable.properties()); + // the Hudi table lives under the data location, which Iceberg resolves from the table + // properties + this.sourceTable = sourceTable.toBuilder().dataPath(dataPath).build(); + this.targetTableProperties = new Properties(); + this.targetTableProperties.putAll(targetTableProperties); + this.targetTableProperties.setProperty( + HudiTargetConfig.EXECUTION_ENGINE, HudiTargetConfig.EXECUTION_ENGINE_SPARK); + // the secondary index is only available from Hudi table version 8 onwards + this.targetTableProperties.setProperty( + HudiTargetConfig.HUDI_TABLE_VERSION, String.valueOf(HoodieTableVersion.NINE.versionCode())); + this.indexedColumns = + HudiTargetConfig.fromProperties(this.targetTableProperties).getSecondaryIndexColumns(); + Preconditions.checkArgument( + !indexedColumns.isEmpty(), + "Set %s to the columns to index", + HudiTargetConfig.SECONDARY_INDEX_COLUMNS); + this.sparkSession = sparkSession; + this.javaSparkContext = JavaSparkContext.fromSparkContext(sparkSession.sparkContext()); + this.engineContext = new HoodieSparkEngineContext(javaSparkContext); + } + + @Override + public boolean doesIndexExist(String columnName) { + return getCompletedIndexPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + } + + @Override + public void syncIndex(Table icebergTable) { + Configuration configuration = javaSparkContext.hadoopConfiguration(); + IcebergConversionSourceProvider sourceProvider = new IcebergConversionSourceProvider(); + sourceProvider.init(configuration); + // The sync reports a failed target in its result instead of throwing. It returns no result + // for the target when the table has no new commits. + SyncResult syncResult = + new ConversionController(configuration) + .sync(getConversionConfig(icebergTable), sourceProvider) + .get(TableFormat.HUDI); + if (syncResult != null + && (syncResult.getTableFormatSyncStatus() == null + || syncResult.getTableFormatSyncStatus().getStatusCode() != SyncStatusCode.SUCCESS)) { + throw new IllegalStateException( + "Failed to sync the secondary indexes of columns " + + indexedColumns + + " of table " + + tableLocation + + ": " + + syncResult.getTableFormatSyncStatus()); + } + Set<String> completedPartitions = getCompletedIndexPartitions(); + List<String> notIndexed = + indexedColumns.stream() + .filter( + column -> + !completedPartitions.contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + column)) + .collect(Collectors.toList()); + if (!notIndexed.isEmpty()) { + throw new IllegalStateException( + "The secondary indexes of columns " + + notIndexed + + " of table " + + tableLocation + + " do not exist. The indexes are built when the index is created, and adding a" + + " column to an existing index is not supported yet."); + } + log.info( + "Synced the secondary indexes of columns {} of table {}", indexedColumns, tableLocation); + } + + /** + * Returns the metadata table partitions whose build has completed. A partition directory can + * exist while the build is in flight or after it failed, so the directory alone does not show + * that the index is usable. + */ + private Set<String> getCompletedIndexPartitions() { Review Comment: Done in ee95125b. `getIndexedColumns()` returns the columns with a completed index, which also removes the prefix checks in `doesIndexExist`, `syncIndex` and `lookup`. ########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = + TableFormatUtils.getTableDataLocation( + TableFormat.ICEBERG, tableLocation, icebergTable.properties()); + // the Hudi table lives under the data location, which Iceberg resolves from the table + // properties + this.sourceTable = sourceTable.toBuilder().dataPath(dataPath).build(); + this.targetTableProperties = new Properties(); + this.targetTableProperties.putAll(targetTableProperties); + this.targetTableProperties.setProperty( + HudiTargetConfig.EXECUTION_ENGINE, HudiTargetConfig.EXECUTION_ENGINE_SPARK); + // the secondary index is only available from Hudi table version 8 onwards + this.targetTableProperties.setProperty( + HudiTargetConfig.HUDI_TABLE_VERSION, String.valueOf(HoodieTableVersion.NINE.versionCode())); + this.indexedColumns = + HudiTargetConfig.fromProperties(this.targetTableProperties).getSecondaryIndexColumns(); + Preconditions.checkArgument( + !indexedColumns.isEmpty(), + "Set %s to the columns to index", + HudiTargetConfig.SECONDARY_INDEX_COLUMNS); + this.sparkSession = sparkSession; + this.javaSparkContext = JavaSparkContext.fromSparkContext(sparkSession.sparkContext()); + this.engineContext = new HoodieSparkEngineContext(javaSparkContext); + } + + @Override + public boolean doesIndexExist(String columnName) { + return getCompletedIndexPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + } + + @Override + public void syncIndex(Table icebergTable) { + Configuration configuration = javaSparkContext.hadoopConfiguration(); + IcebergConversionSourceProvider sourceProvider = new IcebergConversionSourceProvider(); + sourceProvider.init(configuration); + // The sync reports a failed target in its result instead of throwing. It returns no result + // for the target when the table has no new commits. + SyncResult syncResult = + new ConversionController(configuration) + .sync(getConversionConfig(icebergTable), sourceProvider) + .get(TableFormat.HUDI); + if (syncResult != null + && (syncResult.getTableFormatSyncStatus() == null + || syncResult.getTableFormatSyncStatus().getStatusCode() != SyncStatusCode.SUCCESS)) { + throw new IllegalStateException( + "Failed to sync the secondary indexes of columns " + + indexedColumns + + " of table " + + tableLocation + + ": " + + syncResult.getTableFormatSyncStatus()); + } + Set<String> completedPartitions = getCompletedIndexPartitions(); + List<String> notIndexed = + indexedColumns.stream() + .filter( + column -> + !completedPartitions.contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + column)) + .collect(Collectors.toList()); + if (!notIndexed.isEmpty()) { + throw new IllegalStateException( + "The secondary indexes of columns " + + notIndexed + + " of table " + + tableLocation + + " do not exist. The indexes are built when the index is created, and adding a" + + " column to an existing index is not supported yet."); + } Review Comment: The check is not about atomicity. A sync without a new Iceberg snapshot returns no result, so this is the only place that notices an index that was never built. A changed column now fails inside the commit, see `checkSecondaryIndexColumnUnchanged` in ee95125b. ########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = + TableFormatUtils.getTableDataLocation( + TableFormat.ICEBERG, tableLocation, icebergTable.properties()); + // the Hudi table lives under the data location, which Iceberg resolves from the table + // properties + this.sourceTable = sourceTable.toBuilder().dataPath(dataPath).build(); + this.targetTableProperties = new Properties(); + this.targetTableProperties.putAll(targetTableProperties); + this.targetTableProperties.setProperty( + HudiTargetConfig.EXECUTION_ENGINE, HudiTargetConfig.EXECUTION_ENGINE_SPARK); + // the secondary index is only available from Hudi table version 8 onwards + this.targetTableProperties.setProperty( + HudiTargetConfig.HUDI_TABLE_VERSION, String.valueOf(HoodieTableVersion.NINE.versionCode())); + this.indexedColumns = + HudiTargetConfig.fromProperties(this.targetTableProperties).getSecondaryIndexColumns(); + Preconditions.checkArgument( + !indexedColumns.isEmpty(), + "Set %s to the columns to index", + HudiTargetConfig.SECONDARY_INDEX_COLUMNS); + this.sparkSession = sparkSession; + this.javaSparkContext = JavaSparkContext.fromSparkContext(sparkSession.sparkContext()); + this.engineContext = new HoodieSparkEngineContext(javaSparkContext); + } + + @Override + public boolean doesIndexExist(String columnName) { + return getCompletedIndexPartitions() + .contains(PARTITION_NAME_SECONDARY_INDEX_PREFIX + columnName); + } + + @Override + public void syncIndex(Table icebergTable) { + Configuration configuration = javaSparkContext.hadoopConfiguration(); + IcebergConversionSourceProvider sourceProvider = new IcebergConversionSourceProvider(); + sourceProvider.init(configuration); + // The sync reports a failed target in its result instead of throwing. It returns no result + // for the target when the table has no new commits. + SyncResult syncResult = + new ConversionController(configuration) + .sync(getConversionConfig(icebergTable), sourceProvider) + .get(TableFormat.HUDI); + if (syncResult != null + && (syncResult.getTableFormatSyncStatus() == null + || syncResult.getTableFormatSyncStatus().getStatusCode() != SyncStatusCode.SUCCESS)) { + throw new IllegalStateException( Review Comment: Done in ee95125b. There was no such helper, so `SyncResult.getErrorMessage` now returns the error message of the table format sync, with a test in `xtable-api`. ########## xtable-core/src/main/java/org/apache/xtable/index/HudiBackedIcebergSecondaryIndex.java: ########## @@ -0,0 +1,299 @@ +/* + * 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.xtable.index; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.PARTITION_NAME_SECONDARY_INDEX_PREFIX; +import static org.apache.spark.sql.functions.col; + +import java.util.Collections; +import java.util.List; +import java.util.Properties; +import java.util.Set; +import java.util.stream.Collectors; + +import lombok.extern.log4j.Log4j2; + +import org.apache.hadoop.conf.Configuration; +import org.apache.spark.api.java.JavaPairRDD; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Encoders; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.RowFactory; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.StructType; + +import org.apache.hudi.client.common.HoodieSparkEngineContext; +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodiePairData; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.table.HoodieTableVersion; +import org.apache.hudi.data.HoodieJavaPairRDD; +import org.apache.hudi.data.HoodieJavaRDD; +import org.apache.hudi.exception.TableNotFoundException; +import org.apache.hudi.metadata.HoodieBackedTableMetadata; +import org.apache.hudi.storage.HoodieStorage; +import org.apache.hudi.storage.hadoop.HoodieHadoopStorage; + +import org.apache.iceberg.Table; + +import scala.Tuple2; + +import com.google.common.base.Preconditions; + +import org.apache.xtable.catalog.TableFormatUtils; +import org.apache.xtable.conversion.ConversionConfig; +import org.apache.xtable.conversion.ConversionController; +import org.apache.xtable.conversion.SourceTable; +import org.apache.xtable.conversion.TargetTable; +import org.apache.xtable.hudi.HudiTargetConfig; +import org.apache.xtable.iceberg.IcebergConversionSourceProvider; +import org.apache.xtable.model.storage.TableFormat; +import org.apache.xtable.model.sync.SyncResult; +import org.apache.xtable.model.sync.SyncStatusCode; + +/** + * Secondary index for an Iceberg table backed by the Hudi metadata table. The index is built by + * syncing the Iceberg table to Hudi with the record index and a secondary index enabled; the Hudi + * metadata lives under {@code <data location>/.hoodie/metadata}. Because the Iceberg data files + * carry no Hudi record keys, the record key of a row is {@code <file path relative to the data + * location>_<row position>}, which a lookup resolves back to a file path and row position. + */ +@Log4j2 +public class HudiBackedIcebergSecondaryIndex implements Index<Table> { + private final String tableLocation; + private final String dataPath; + private final SourceTable sourceTable; + private final Properties targetTableProperties; + private final List<String> indexedColumns; + private final SparkSession sparkSession; + private final JavaSparkContext javaSparkContext; + private final HoodieSparkEngineContext engineContext; + + /** + * @param icebergTable the table to index + * @param sourceTable how XTable loads the table for a sync, because a loaded table does not + * expose the catalog it came from. A table of a catalog sets its catalog config, name and + * namespace, and a table without a catalog config is loaded from its location + * @param sparkSession the Spark session used to build and query the index + * @param targetTableProperties {@link HudiTargetConfig} properties for the sync. {@link + * HudiTargetConfig#SECONDARY_INDEX_COLUMNS} lists the columns to index, and other properties + * such as the record index file group counts are optional + */ + public HudiBackedIcebergSecondaryIndex( + Table icebergTable, + SourceTable sourceTable, + SparkSession sparkSession, + Properties targetTableProperties) { + Preconditions.checkArgument( + TableFormat.ICEBERG.equals(sourceTable.getFormatName()), + "Expected an Iceberg source table but got format %s", + sourceTable.getFormatName()); + this.tableLocation = icebergTable.location(); + this.dataPath = Review Comment: It does not append blindly: `TableFormatUtils.getTableDataLocation` reads `write.data.path`, `write.folder-storage.path` and `write.object-storage.path` before it falls back to `<location>/data`. Files outside that location are not supported, because Hudi needs every file under its base path, and ee95125b documents this on the class. There is no IT for a custom `write.data.path` yet. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
