pvary commented on code in PR #14435: URL: https://github.com/apache/iceberg/pull/14435#discussion_r2576716711
########## parquet/src/main/java/org/apache/iceberg/parquet/ParquetFileMerger.java: ########## @@ -0,0 +1,643 @@ +/* + * 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.parquet; + +import static java.util.Collections.emptyMap; + +import java.io.IOException; +import java.util.List; +import java.util.Locale; +import java.util.Map; +import java.util.function.LongUnaryOperator; +import org.apache.hadoop.conf.Configuration; +import org.apache.iceberg.MetadataColumns; +import org.apache.iceberg.encryption.EncryptedOutputFile; +import org.apache.iceberg.io.InputFile; +import org.apache.iceberg.io.OutputFile; +import org.apache.iceberg.io.SeekableInputStream; +import org.apache.iceberg.relocated.com.google.common.collect.Lists; +import org.apache.parquet.bytes.BytesInput; +import org.apache.parquet.bytes.DirectByteBufferAllocator; +import org.apache.parquet.bytes.HeapByteBufferAllocator; +import org.apache.parquet.column.ColumnDescriptor; +import org.apache.parquet.column.Encoding; +import org.apache.parquet.column.ParquetProperties; +import org.apache.parquet.column.statistics.LongStatistics; +import org.apache.parquet.column.statistics.Statistics; +import org.apache.parquet.column.values.ValuesWriter; +import org.apache.parquet.column.values.delta.DeltaBinaryPackingValuesWriterForLong; +import org.apache.parquet.crypto.InternalFileEncryptor; +import org.apache.parquet.hadoop.CodecFactory; +import org.apache.parquet.hadoop.ParquetFileReader; +import org.apache.parquet.hadoop.ParquetFileWriter; +import org.apache.parquet.hadoop.metadata.BlockMetaData; +import org.apache.parquet.hadoop.metadata.ColumnChunkMetaData; +import org.apache.parquet.hadoop.metadata.CompressionCodecName; +import org.apache.parquet.io.DelegatingSeekableInputStream; +import org.apache.parquet.schema.MessageType; +import org.apache.parquet.schema.PrimitiveType; +import org.apache.parquet.schema.Type; +import org.apache.parquet.schema.Types; + +/** + * Utility class for performing strict schema validation and merging of Parquet files at the + * row-group level. + * + * <p>This class ensures that all input files have identical Parquet schemas before merging. The + * merge operation is performed by copying row groups directly without + * serialization/deserialization, providing significant performance benefits over traditional + * read-rewrite approaches. + * + * <p>This class works with any Iceberg FileIO implementation (HadoopFileIO, S3FileIO, GCSFileIO, + * etc.), making it cloud-agnostic. + * + * <p>TODO: Encrypted tables are not supported + * + * <p>Key features: + * + * <ul> + * <li>Row group merging without deserialization using {@link ParquetFileWriter#appendFile} + * <li>Strict schema validation - all files must have identical {@link MessageType} + * <li>Metadata merging for Iceberg-specific footer data + * <li>Works with any FileIO implementation (local, S3, GCS, Azure, etc.) + * </ul> + * + * <p>Restrictions: + * + * <ul> + * <li>All files must have compatible schemas (identical {@link MessageType}) + * <li>Files must not be encrypted + * <li>Files must not have associated delete files or delete vectors + * <li>Table must not have a sort order (including z-ordered tables) + * </ul> + * + * <p>Typical usage: + * + * <pre> + * ValidationResult result = ParquetFileMerger.readAndValidateSchema(inputFiles); + * if (result != null) { + * ParquetFileMerger.mergeFiles( + * inputFiles, encryptedOutputFile, result.schema(), firstRowIds, + * rowGroupSize, columnIndexTruncateLength, result.metadata()); + * } + * </pre> + */ +public class ParquetFileMerger { + // Default buffer sizes for DeltaBinaryPackingValuesWriter + private static final int DEFAULT_INITIAL_BUFFER_SIZE = 64 * 1024; // 64KB + private static final int DEFAULT_PAGE_SIZE_FOR_ENCODING = 64 * 1024; // 64KB + + private ParquetFileMerger() { + // Utility class - prevent instantiation + } + + /** + * Validates that all input files can be merged. + * + * <p>This method validates that: + * + * <ul> + * <li>All files are valid Parquet format (detected by reading Parquet footer) + * <li>All files have identical schemas + * <li>Files are not encrypted (detected by ParquetCryptoRuntimeException) + * <li>If physical row lineage columns exist, all values are non-null + * </ul> + * + * <p>This method works with any Iceberg FileIO implementation (S3FileIO, GCSFileIO, etc.). + * + * @param inputFiles List of Iceberg input files to validate + * @return true if files can be merged, false otherwise + */ + public static boolean canMerge(List<InputFile> inputFiles) { + try { + if (inputFiles == null || inputFiles.isEmpty()) { + return false; + } + + // Read schema from the first file + MessageType firstSchema = readSchema(inputFiles.get(0)); + + // Check if schema has physical row lineage columns + boolean hasRowIdColumn = firstSchema.containsField(MetadataColumns.ROW_ID.name()); + boolean hasSeqNumColumn = + firstSchema.containsField(MetadataColumns.LAST_UPDATED_SEQUENCE_NUMBER.name()); + + // Validate all files have the same schema + for (int i = 1; i < inputFiles.size(); i++) { + MessageType currentSchema = readSchema(inputFiles.get(i)); + + if (!firstSchema.equals(currentSchema)) { + return false; + } + } + + // If there are physical row lineage columns, validate no nulls + if (hasRowIdColumn || hasSeqNumColumn) { + validateRowLineageColumnsHaveNoNulls(inputFiles); + } + + return true; + } catch (RuntimeException | IOException e) { + // Returns false for: + // - Non-Parquet files (IOException when reading Parquet footer) + // - Encrypted files (ParquetCryptoRuntimeException extends RuntimeException) + // - Files with null row lineage values (IllegalArgumentException from + // validateRowLineageColumnsHaveNoNulls) + // - Any other validation failures + return false; + } + } + + /** + * Reads the Parquet schema from an Iceberg InputFile. + * + * @param inputFile Iceberg input file to read schema from + * @return MessageType schema of the Parquet file + * @throws IOException if reading fails + */ + public static MessageType readSchema(InputFile inputFile) throws IOException { Review Comment: If you follow my suggestion here: https://github.com/apache/iceberg/pull/14435#discussion_r2576659817, you can eliminate the external calls for readMetadata. For `readSchema`, consider invoking it internally in `ParquetFileMerger.mergeFiles`, which removes the need for an extra parameter. I’d like to see these changes incorporated in the PR. Another idea (more debatable): With some additional effort, we might avoid the extra file open: - `canMerge` could return an enum (e.g., FORBIDDEN, SIMPLE_MERGE, SYNTHETIC_LINEAGE). - Based on the enum, decide in `ParquetFileMerger.mergeFiles` which merge operation to use. - In `mergeFilesWithRowLineageAndSchema` and `mergeFilesWithSchema`, open the first file, read schema/metadata, and reuse the opened file during the loop. I’m not fully convinced this is worth it—the code becomes complex, and we only save one footer read. If you have a better approach, I’m open to it. Otherwise, I’m fine accepting the duplicated footer read. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
