shangxinli commented on code in PR #14435: URL: https://github.com/apache/iceberg/pull/14435#discussion_r2593215231
########## parquet/src/main/java/org/apache/iceberg/parquet/ParquetFileMerger.java: ########## @@ -0,0 +1,712 @@ +/* + * 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.nio.ByteBuffer; +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.DataFile; +import org.apache.iceberg.DataFiles; +import org.apache.iceberg.FileFormat; +import org.apache.iceberg.MetadataColumns; +import org.apache.iceberg.Metrics; +import org.apache.iceberg.MetricsConfig; +import org.apache.iceberg.PartitionSpec; +import org.apache.iceberg.StructLike; +import org.apache.iceberg.hadoop.HadoopOutputFile; +import org.apache.iceberg.io.FileIO; +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.iceberg.types.Conversions; +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.ParquetOutputFormat; +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; + } + } + + /** + * Validates that DataFiles can be merged, checking both Parquet-specific and data-level + * requirements. + * + * <p>This method validates: + * + * <ul> + * <li>All Parquet-specific requirements (via {@link #canMerge(List)}) + * <li>No files have associated delete files + * <li>All files have the same partition spec + * <li>No files exceed the target output size (not splitting large files) + * </ul> + * + * <p>This validation is useful for compaction operations in Spark, Flink, or other engines that + * need to ensure files can be safely merged. + * + * @param dataFiles List of DataFiles to validate + * @param fileIO FileIO to use for reading files + * @param targetOutputSize Maximum size for output file (files larger than this cannot be merged) + * @return true if files can be merged, false otherwise + */ + public static boolean canMerge(List<DataFile> dataFiles, FileIO fileIO, long targetOutputSize) { + if (dataFiles == null || dataFiles.isEmpty()) { + return false; + } + + // Check for delete files - row-group merge cannot apply deletes + for (DataFile dataFile : dataFiles) { + List<Integer> equalityIds = dataFile.equalityFieldIds(); + if (equalityIds != null && !equalityIds.isEmpty()) { + return false; + } + } Review Comment: You're right. I just looked at the spec in format/spec.md - the manifest entry field 135 equality_ids is documented as "Required when content=2 and should be null otherwise." Since data files have content=0, equalityFieldIds should always be null for data files according to the spec. However, I think checking if it conforms to the spec won't hurt and can even add more safety - it's a defensive validation that ensures we're working with properly formatted data files. If a file somehow has equalityFieldIds set (perhaps due to a bug in writing or corruption), attempting to merge it could produce incorrect results since we can't apply equality deletes during binary row-group merging. That said, if you have concerns that this check is completely redundant and could cause issues (e.g., false rejections), I'm happy to remove it. What do you think? -- 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]
