aokolnychyi commented on a change in pull request #2210: URL: https://github.com/apache/iceberg/pull/2210#discussion_r588694813
########## File path: spark3/src/main/java/org/apache/iceberg/spark/procedures/AddFilesProcedure.java ########## @@ -0,0 +1,319 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + + +package org.apache.iceberg.spark.procedures; + +import java.io.IOException; +import java.util.Collections; +import java.util.List; +import java.util.Locale; +import java.util.Map; +import java.util.stream.Collectors; +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.fs.FileSystem; +import org.apache.hadoop.fs.Path; +import org.apache.iceberg.AppendFiles; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.MetricsConfig; +import org.apache.iceberg.PartitionField; +import org.apache.iceberg.PartitionSpec; +import org.apache.iceberg.Snapshot; +import org.apache.iceberg.SnapshotSummary; +import org.apache.iceberg.Table; +import org.apache.iceberg.TableProperties; +import org.apache.iceberg.mapping.MappingUtil; +import org.apache.iceberg.mapping.NameMapping; +import org.apache.iceberg.mapping.NameMappingParser; +import org.apache.iceberg.relocated.com.google.common.collect.ImmutableList; +import org.apache.iceberg.relocated.com.google.common.collect.Maps; +import org.apache.iceberg.spark.Spark3Util; +import org.apache.iceberg.spark.SparkTableUtil; +import org.apache.spark.sql.catalyst.InternalRow; +import org.apache.spark.sql.catalyst.catalog.CatalogTable; +import org.apache.spark.sql.connector.catalog.Identifier; +import org.apache.spark.sql.connector.catalog.TableCatalog; +import org.apache.spark.sql.connector.iceberg.catalog.ProcedureParameter; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.Metadata; +import org.apache.spark.sql.types.StructField; +import org.apache.spark.sql.types.StructType; +import scala.Option; +import scala.runtime.BoxedUnit; + +public class AddFilesProcedure extends BaseProcedure { + + enum Format { + catalog, + orc, + parquet + } + + private static final ProcedureParameter[] PARAMETERS = new ProcedureParameter[]{ + ProcedureParameter.required("table", DataTypes.StringType), + ProcedureParameter.required("source_table", DataTypes.StringType), + ProcedureParameter.optional("partition", STRING_MAP) + }; + + private static final StructType OUTPUT_TYPE = new StructType(new StructField[]{ + new StructField("added_files_count", DataTypes.LongType, false, Metadata.empty()) + }); + + private AddFilesProcedure(TableCatalog tableCatalog) { + super(tableCatalog); + } + + public static SparkProcedures.ProcedureBuilder builder() { + return new BaseProcedure.Builder<AddFilesProcedure>() { + @Override + protected AddFilesProcedure doBuild() { + return new AddFilesProcedure(tableCatalog()); + } + }; + } + + @Override + public ProcedureParameter[] parameters() { + return PARAMETERS; + } + + @Override + public StructType outputType() { + return OUTPUT_TYPE; + } + + @Override + public InternalRow[] call(InternalRow args) { + Identifier tableIdent = toIdentifier(args.getString(0), PARAMETERS[0].name()); + + String sourceTable = args.getString(1); + Format format = parseFormat(sourceTable); + + Map<String, String> partition = Maps.newHashMap(); + if (!args.isNullAt(2)) { + args.getMap(2).foreach(DataTypes.StringType, DataTypes.StringType, + (k, v) -> { + partition.put(k.toString(), v.toString()); + return BoxedUnit.UNIT; + }); + } + + Long filesAdded = importToIceberg(tableIdent, sourceTable, format, partition); + return new InternalRow[]{newInternalRow(filesAdded)}; + } + + private Format parseFormat(String source) { + String[] parts = source.split("\\.", 2); + if (parts.length == 2) { + if (parts[0].toLowerCase(Locale.ROOT).equals("orc")) { + return Format.orc; + } + if (parts[0].toLowerCase(Locale.ROOT).equals("parquet")) { + return Format.parquet; + } + return Format.catalog; + } + return Format.catalog; + } + + private long importToIceberg(Identifier destIdent, String sourceIdent, Format format, Map<String, String> partition) { + return modifyIcebergTable(destIdent, table -> { + + validatePartitionSpec(table, partition); + applyNameMappingIfMissing(table); + + if (format != Format.catalog) { + String[] parts = sourceIdent.split("\\.", 2); + Path sourcePath = new Path(parts[1]); + + Configuration conf = spark().sessionState().newHadoopConf(); + FileSystem fs; + Boolean isFile; + try { + fs = sourcePath.getFileSystem(conf); + isFile = fs.getFileStatus(sourcePath).isFile(); + } catch (IOException e) { + throw new RuntimeException("Unable to access add_file path", e); + } + + if (isFile) { + importFile(table, sourcePath, format, partition); + } else { + importFileTable(table, sourcePath, format, partition); + } + } else { + importCatalogTable(table, sourceIdent, partition); + } + + Snapshot snapshot = table.currentSnapshot(); + long numAddedFiles = Long.parseLong(snapshot.summary().get(SnapshotSummary.ADDED_FILES_PROP)); + return numAddedFiles; + }); + } + + private static void applyNameMappingIfMissing(Table table) { + if (table.properties().get(TableProperties.DEFAULT_NAME_MAPPING) == null) { + // Forces Name based resolution instead of position based resolution + NameMapping mapping = MappingUtil.create(table.schema()); + String mappingJson = NameMappingParser.toJson(mapping); + table.updateProperties() + .set(TableProperties.DEFAULT_NAME_MAPPING, mappingJson) + .commit(); + } + } + + private int importFile(Table table, Path file, Format format, Map<String, String> partition) { + if (partition.isEmpty() && !table.spec().isUnpartitioned()) { + throw new IllegalArgumentException("Cannot add a file to a partitioned table without specifying the partition " + + "which it should be placed in"); + } + if (!partition.isEmpty() && table.spec().isUnpartitioned()) { + throw new IllegalArgumentException("Cannot add a file to an unpartitioned table while specifying the partition " + + "which it should be placed in"); + } + + PartitionSpec spec = table.spec(); + MetricsConfig metricsConfig = MetricsConfig.fromProperties(table.properties()); + String partitionURI = file.toString(); + Configuration conf = spark().sessionState().newHadoopConf(); + + List<DataFile> files = + SparkTableUtil.listPartition(partition, partitionURI, format.name(), spec, conf, metricsConfig); + + if (files.isEmpty()) { + throw new IllegalArgumentException( + String.format("No file found for add_file command. Looking for a file at URI %s", partitionURI)); + } + + // Add Snapshot Summary Info? + AppendFiles append = table.newAppend(); + files.forEach(append::appendFile); + append.commit(); + return 1; + } + + private void importFileTable(Table table, Path tableLocation, Format format, Map<String, String> partition) { + // List Partitions via Spark InMemory file search interface + List<SparkTableUtil.SparkPartition> partitions = + filterPartitions(Spark3Util.getPartitions(spark(), tableLocation, format.name()), partition); + + importPartitions(table, partitions); + } + + private void importCatalogTable(Table table, String sourceTable, Map<String, String> partition) { + CatalogTable catalogTable = SparkTableUtil.getCatalogTable(spark(), sourceTable); + List<SparkTableUtil.SparkPartition> partitions; + + if (catalogTable.partitionColumnNames().isEmpty()) { + // Unpartitioned table + if (!partition.isEmpty()) { + throw new IllegalArgumentException("Cannot import files from a unpartitioned Hive table when a partition " + + "is specified"); + } + Option<String> format = + catalogTable.storage().serde().nonEmpty() ? catalogTable.storage().serde() : catalogTable.provider(); + partitions = ImmutableList.of(new SparkTableUtil.SparkPartition( + Collections.emptyMap(), + catalogTable.location().toString(), + catalogTable.storage().serde().get())); + + } else { + // Read partitions Spark via Catalog Interface + partitions = filterPartitions(SparkTableUtil.getPartitions(spark(), sourceTable), partition); + } + + importPartitions(table, partitions); + } + + private void importPartitions(Table table, List<SparkTableUtil.SparkPartition> partitions) { + String stagingLocation = table.properties() + .getOrDefault(TableProperties.WRITE_METADATA_LOCATION, table.location() + "/metadata"); + SparkTableUtil.importSparkPartitions(spark(), partitions, table, table.spec(), stagingLocation); + } + + private List<SparkTableUtil.SparkPartition> filterPartitions(List<SparkTableUtil.SparkPartition> partitions, + Map<String, String> partition) { + if (partition.isEmpty()) { + // No partition filter arg + + if (partitions.isEmpty()) { + throw new IllegalArgumentException("Cannot add files, no files found in the table."); + } + return partitions; + } else { + // Partition filter arg passed + + List<SparkTableUtil.SparkPartition> filteredPartitions = partitions + .stream().filter(p -> p.getValues().equals(partition)) // Todo Support Wildcards + .collect(Collectors.toList()); + + if (filteredPartitions.isEmpty()) { + throw new IllegalArgumentException( + String.format("No partitions found in table for add_file command. " + + "Looking for a partition with the value %s", + partition.entrySet().stream().map(e -> e.getKey() + "=" + e.getValue()) + .collect(Collectors.joining(","))) + ); + } + + return filteredPartitions; + } + } + + @Override + public String description() { + return null; + } + + private void validatePartitionSpec(Table table, Map<String, String> partition) { + List<PartitionField> partitionFields = table.spec().fields(); + + boolean tablePartitioned = !partitionFields.isEmpty(); + boolean partitionSpecPassed = !partition.isEmpty(); + + if (tablePartitioned && partitionSpecPassed) { Review comment: Since we use the map as a filter, we no longer should require the map to contain all arguments. This should be valid, I guess: ``` // table partitioned by p1, p2 add_files( table => 'db.iceberg_table', -- partitioned Iceberg table source_table => '`parquet`.`path/to/table/`' partition_filter => map('p1', ‘x’) ) ``` That is instead of using *, which is more flexible. ---------------------------------------------------------------- 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: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
