[
https://issues.apache.org/jira/browse/PARQUET-1381?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17744984#comment-17744984
]
ASF GitHub Bot commented on PARQUET-1381:
-----------------------------------------
advancedxy commented on code in PR #1121:
URL: https://github.com/apache/parquet-mr/pull/1121#discussion_r1269155063
##########
parquet-hadoop/src/main/java/org/apache/parquet/hadoop/rewrite/ParquetRewriter.java:
##########
@@ -751,6 +764,27 @@ public GroupConverter asGroupConverter() {
}
}
+ private void mergeRowGroups() throws IOException {
+ if (null == reader) {
+ return;
+ }
+
+ boolean v2EncodingHint = meta.getBlocks().stream()
+ .flatMap(b -> b.getColumns().stream())
+ .anyMatch(chunk -> {
+ EncodingStats stats = chunk.getEncodingStats();
+ return stats != null && stats.usesV2Pages();
+ });
+
+ List<ParquetFileReader> readers = new ArrayList<>();
+ do {
+ readers.add(reader);
+ initNextReader();
Review Comment:
Looks like `v2EncodingHint` only checks the first parquet file..
Should all the files to be checked?
##########
parquet-hadoop/src/main/java/org/apache/parquet/hadoop/rewrite/RowGroupMerger.java:
##########
@@ -0,0 +1,657 @@
+/*
+ * 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.parquet.hadoop.rewrite;
+
+import static java.lang.String.format;
+import static org.apache.parquet.column.ValuesType.DEFINITION_LEVEL;
+import static org.apache.parquet.column.ValuesType.REPETITION_LEVEL;
+import static org.apache.parquet.column.ValuesType.VALUES;
+
+import java.io.IOException;
+import java.io.UncheckedIOException;
+import java.nio.ByteBuffer;
+import java.util.List;
+import java.util.Collections;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.Map;
+import java.util.AbstractMap.SimpleEntry;
+import java.util.Map.Entry;
+import java.util.function.BiConsumer;
+import java.util.function.Consumer;
+import java.util.stream.Collectors;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.parquet.Preconditions;
+import org.apache.parquet.bytes.ByteBufferInputStream;
+import org.apache.parquet.bytes.BytesInput;
+import org.apache.parquet.column.ColumnDescriptor;
+import org.apache.parquet.column.Dictionary;
+import org.apache.parquet.column.Encoding;
+import org.apache.parquet.column.ParquetProperties;
+import org.apache.parquet.column.page.DataPage;
+import org.apache.parquet.column.page.DataPageV1;
+import org.apache.parquet.column.page.DataPageV2;
+import org.apache.parquet.column.page.DictionaryPage;
+import org.apache.parquet.column.page.PageReadStore;
+import org.apache.parquet.column.page.PageReader;
+import org.apache.parquet.column.values.ValuesReader;
+import org.apache.parquet.column.values.ValuesWriter;
+import org.apache.parquet.compression.CompressionCodecFactory;
+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.ParquetDecodingException;
+import org.apache.parquet.schema.MessageType;
+import org.apache.parquet.schema.PrimitiveType;
+
+class RowGroupMerger {
+
+ private final MessageType schema;
+ private final CompressionCodecFactory.BytesInputCompressor compressor;
+ private final ParquetProperties parquetProperties;
+
+ public RowGroupMerger(MessageType schema, CompressionCodecName compression,
boolean useV2ValueWriter) {
+ this(schema, new Configuration(), compression, useV2ValueWriter);
+ }
+
+ RowGroupMerger(MessageType schema, Configuration conf, CompressionCodecName
compression, boolean useV2ValueWriter) {
+ this(schema, conf, compression, createParquetProperties(useV2ValueWriter));
+ }
+
+ RowGroupMerger(MessageType schema, Configuration conf, CompressionCodecName
compression, ParquetProperties parquetProperties) {
+ this.schema = schema;
+ this.parquetProperties = parquetProperties;
+ this.compressor = new CodecFactory(conf,
this.parquetProperties.getPageSizeThreshold()).getCompressor(compression);
+ }
+
+ /**
+ * Merges the row groups making sure that new row groups do not exceed the
supplied maxRowGroupSize
+ *
+ * @param inputFiles input files to merge
+ * @param maxRowGroupSize the max limit for new blocks
+ * @param writer writer to write the new blocks to
+ * @throws IOException if an IO error occurs
+ */
+ public void merge(List<ParquetFileReader> inputFiles, final long
maxRowGroupSize,
+ ParquetFileWriter writer) throws IOException {
+
+ SizeEstimator estimator = new SizeEstimator(compressor.getCodecName() !=
CompressionCodecName.UNCOMPRESSED);
+ MutableMergedBlock mergedBlock = null;
+ for (ParquetFileReader reader : inputFiles) {
+ for (BlockMetaData blockMeta : reader.getRowGroups()) {
+ PageReadStore group = reader.readNextRowGroup();
+ Preconditions.checkState(group != null,
+ "number of groups returned by FileReader does not match metadata");
+
+ if (mergedBlock != null && mergedBlock.getCompressedSize() +
estimator.estimate(blockMeta) > maxRowGroupSize) {
+ saveBlockTo(mergedBlock, writer);
+ mergedBlock = null;
+ }
+
+ if (mergedBlock == null && estimator.estimate(blockMeta) >
maxRowGroupSize) {
+ //save it directly without re encoding it
+ saveBlockTo(ReadOnlyMergedBlock.of(blockMeta, group, schema,
compressor), writer);
+ continue;
+ }
+
+ if (mergedBlock == null) {
+ mergedBlock = new MutableMergedBlock(schema);
+ }
+
+ long sizeBeforeMerge = mergedBlock.getCompressedSize();
+ mergedBlock.merge(blockMeta, group);
+ //update our estimator
+ long currentBlockEffect = mergedBlock.getCompressedSize() -
sizeBeforeMerge;
+ estimator.update(currentBlockEffect, blockMeta);
+ }
+ }
+ if (mergedBlock != null) {
+ saveBlockTo(mergedBlock, writer);
+ }
+ mergedBlock = null;
+ }
+
+
+ private void saveBlockTo(MergedBlock block, ParquetFileWriter writer) {
+ try {
+ writer.startBlock(block.rowCount());
+
+ for (MergedColumn col : block.columnsInOrder()) {
+ writer.startColumn(col.getColumnDesc(), col.getValueCount(),
col.getCompression());
+
+ col.writeDictionaryPageTo(writer);
+ col.writeDataPagesTo(writer);
+
+ writer.endColumn();
+ }
+
+ writer.endBlock();
+ } catch (IOException e) {
+ throw new UncheckedIOException(e);
+ }
+ }
+
+ private static ParquetProperties createParquetProperties(boolean
useV2Writer) {
+ ParquetProperties.Builder builder = ParquetProperties.builder();
+ if (useV2Writer) {
+ builder.withWriterVersion(ParquetProperties.WriterVersion.PARQUET_2_0);
+ }
+ return builder.build();
+ }
+
+ private BytesInput compress(BytesInput bytes) {
+ return compress(bytes, compressor);
+ }
+
+ private static BytesInput compress(BytesInput bytes,
CompressionCodecFactory.BytesInputCompressor compressor) {
+ try {
+ //we copy as some compressors use shared memory
+ return BytesInput.copy(compressor.compress(bytes));
+ } catch (IOException e) {
+ throw new UncheckedIOException(e);
+ }
+ }
+
+ private BiConsumer<ValuesReader, ValuesWriter>
createWritingBridge(PrimitiveType.PrimitiveTypeName typeName) {
+ switch (typeName) {
+ case FIXED_LEN_BYTE_ARRAY:
+ case INT96:
+ case BINARY:
+ return (src, dest) -> dest.writeBytes(src.readBytes());
+ case BOOLEAN:
+ return (src, dest) -> dest.writeBoolean(src.readBoolean());
+ case DOUBLE:
+ return (src, dest) -> dest.writeDouble(src.readDouble());
+ case FLOAT:
+ return (src, dest) -> dest.writeFloat(src.readFloat());
+ case INT32:
+ return (src, dest) -> dest.writeInteger(src.readInteger());
+ case INT64:
+ return (src, dest) -> dest.writeLong(src.readLong());
+ default:
+ throw new RuntimeException("Unsupported column primitive type: " +
typeName.name());
+ }
+ }
+
+ private static void writePageTo(DataPage dataPage, ParquetFileWriter writer)
{
+ dataPage.accept(new DataPage.Visitor<Void>() {
+ @Override
+ public Void visit(DataPageV1 page) {
+ try {
+ if (page.getIndexRowCount().isPresent()) {
+ writer.writeDataPage(page.getValueCount(),
page.getUncompressedSize(),
+ page.getBytes(), page.getStatistics(),
page.getIndexRowCount().get(), page.getRlEncoding(),
+ page.getDlEncoding(), page.getValueEncoding());
+
+ } else {
+ writer.writeDataPage(page.getValueCount(),
page.getUncompressedSize(),
+ page.getBytes(), page.getStatistics(), page.getRlEncoding(),
+ page.getDlEncoding(), page.getValueEncoding());
+ }
+ } catch (IOException e) {
+ throw new UncheckedIOException(e);
+ }
+ return null;
+ }
+
+ @Override
+ public Void visit(DataPageV2 page) {
+ try {
+ writer.writeDataPageV2(page.getRowCount(), page.getNullCount(),
page.getValueCount(),
+ page.getRepetitionLevels(), page.getDefinitionLevels(),
page.getDataEncoding(),
+ page.getData(), page.getUncompressedSize(), page.getStatistics());
+ } catch (IOException e) {
+ throw new UncheckedIOException(e);
+ }
+ return null;
+ }
+ });
+ }
+
+ private static DictionaryPage getCompressedDictionary(DictionaryPage
dictionary, CompressionCodecFactory.BytesInputCompressor compressor) {
+ return new DictionaryPage(
+ compress(dictionary.getBytes(), compressor),
+ dictionary.getUncompressedSize(),
+ dictionary.getDictionarySize(),
+ dictionary.getEncoding());
+ }
+
+ private interface MergedBlock {
+ long rowCount();
+
+ List<MergedColumn> columnsInOrder();
+ }
+
+ private interface MergedColumn {
+ long getValueCount();
+
+ CompressionCodecName getCompression();
+
+ ColumnDescriptor getColumnDesc();
+
+ void writeDataPagesTo(ParquetFileWriter writer);
+
+ void writeDictionaryPageTo(ParquetFileWriter writer) throws IOException;
+ }
+
+ private class MutableMergedBlock implements MergedBlock {
+
+ private final Map<ColumnDescriptor, MutableMergedColumn> columns = new
HashMap<>();
+ private final MessageType schema;
+ private long recordCount;
+ private long compressedSize;
+
+ private MutableMergedBlock(MessageType schema) {
+ this.schema = schema;
+ }
+
+ @Override
+ public long rowCount() {
+ return recordCount;
+ }
+
+ private long getCompressedSize() {
+ return compressedSize;
+ }
+
+ @Override
+ public List<MergedColumn> columnsInOrder() {
+ return schema.getColumns()
+ .stream()
+ .map(columns::get)
+ .collect(Collectors.toList());
+ }
+
+ MutableMergedColumn getOrCreateColumn(ColumnDescriptor column) {
+ return columns.computeIfAbsent(column, desc -> new
MutableMergedColumn(desc, this::addCompressedBytes));
+ }
+
+ void addRowCount(long rowCount) {
+ recordCount += rowCount;
+ }
+
+ void addCompressedBytes(long size) {
+ compressedSize += size;
+ }
+
+ void merge(BlockMetaData blockMeta, PageReadStore group) throws
IOException {
+ for (Entry<ColumnDescriptor, ColumnChunkMetaData> col :
getColumnsInOrder(blockMeta, schema)) {
+
+ MutableMergedColumn column = getOrCreateColumn(col.getKey());
+ PageReader columnReader = group.getPageReader(col.getKey());
+
+ DictionaryPage dictPage = columnReader.readDictionaryPage();
+ Dictionary decodedDictionary = null;
+ if (dictPage != null) {
+ decodedDictionary =
dictPage.getEncoding().initDictionary(column.getColumnDesc(), dictPage);
+ }
Review Comment:
If I understand the process of page encoding correctly: parquet tries to use
dictionary encoding by default, If the dictionary grows too big, whether in
size or number of distinct values, the encoding will fall back to the plain
encoding. The check and fallback logic happens when emit the first page.
So when we are merging multiple column chunks from different row groups, if
the first column chunks is dictionary encoded and others are not because it
fallbacks to plain encoding, we should disable the dictionary encoding for that
column on purpose to avoid introducing overhead.
Current logic doesn't handle that, it will use dictionary encoding if the
column chunk in the first row group to be merged use dictionary encoding.
##########
parquet-hadoop/src/main/java/org/apache/parquet/hadoop/rewrite/RowGroupMerger.java:
##########
@@ -0,0 +1,657 @@
+/*
+ * 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.parquet.hadoop.rewrite;
+
+import static java.lang.String.format;
+import static org.apache.parquet.column.ValuesType.DEFINITION_LEVEL;
+import static org.apache.parquet.column.ValuesType.REPETITION_LEVEL;
+import static org.apache.parquet.column.ValuesType.VALUES;
+
+import java.io.IOException;
+import java.io.UncheckedIOException;
+import java.nio.ByteBuffer;
+import java.util.List;
+import java.util.Collections;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.Map;
+import java.util.AbstractMap.SimpleEntry;
+import java.util.Map.Entry;
+import java.util.function.BiConsumer;
+import java.util.function.Consumer;
+import java.util.stream.Collectors;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.parquet.Preconditions;
+import org.apache.parquet.bytes.ByteBufferInputStream;
+import org.apache.parquet.bytes.BytesInput;
+import org.apache.parquet.column.ColumnDescriptor;
+import org.apache.parquet.column.Dictionary;
+import org.apache.parquet.column.Encoding;
+import org.apache.parquet.column.ParquetProperties;
+import org.apache.parquet.column.page.DataPage;
+import org.apache.parquet.column.page.DataPageV1;
+import org.apache.parquet.column.page.DataPageV2;
+import org.apache.parquet.column.page.DictionaryPage;
+import org.apache.parquet.column.page.PageReadStore;
+import org.apache.parquet.column.page.PageReader;
+import org.apache.parquet.column.values.ValuesReader;
+import org.apache.parquet.column.values.ValuesWriter;
+import org.apache.parquet.compression.CompressionCodecFactory;
+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.ParquetDecodingException;
+import org.apache.parquet.schema.MessageType;
+import org.apache.parquet.schema.PrimitiveType;
+
+class RowGroupMerger {
+
+ private final MessageType schema;
+ private final CompressionCodecFactory.BytesInputCompressor compressor;
+ private final ParquetProperties parquetProperties;
+
+ public RowGroupMerger(MessageType schema, CompressionCodecName compression,
boolean useV2ValueWriter) {
+ this(schema, new Configuration(), compression, useV2ValueWriter);
+ }
+
+ RowGroupMerger(MessageType schema, Configuration conf, CompressionCodecName
compression, boolean useV2ValueWriter) {
+ this(schema, conf, compression, createParquetProperties(useV2ValueWriter));
+ }
+
+ RowGroupMerger(MessageType schema, Configuration conf, CompressionCodecName
compression, ParquetProperties parquetProperties) {
+ this.schema = schema;
+ this.parquetProperties = parquetProperties;
+ this.compressor = new CodecFactory(conf,
this.parquetProperties.getPageSizeThreshold()).getCompressor(compression);
+ }
+
+ /**
+ * Merges the row groups making sure that new row groups do not exceed the
supplied maxRowGroupSize
+ *
+ * @param inputFiles input files to merge
+ * @param maxRowGroupSize the max limit for new blocks
+ * @param writer writer to write the new blocks to
+ * @throws IOException if an IO error occurs
+ */
+ public void merge(List<ParquetFileReader> inputFiles, final long
maxRowGroupSize,
+ ParquetFileWriter writer) throws IOException {
+
+ SizeEstimator estimator = new SizeEstimator(compressor.getCodecName() !=
CompressionCodecName.UNCOMPRESSED);
+ MutableMergedBlock mergedBlock = null;
+ for (ParquetFileReader reader : inputFiles) {
+ for (BlockMetaData blockMeta : reader.getRowGroups()) {
+ PageReadStore group = reader.readNextRowGroup();
+ Preconditions.checkState(group != null,
+ "number of groups returned by FileReader does not match metadata");
+
+ if (mergedBlock != null && mergedBlock.getCompressedSize() +
estimator.estimate(blockMeta) > maxRowGroupSize) {
+ saveBlockTo(mergedBlock, writer);
+ mergedBlock = null;
+ }
+
+ if (mergedBlock == null && estimator.estimate(blockMeta) >
maxRowGroupSize) {
+ //save it directly without re encoding it
+ saveBlockTo(ReadOnlyMergedBlock.of(blockMeta, group, schema,
compressor), writer);
Review Comment:
I checked related code, seems that `startColumn` and `endColumn` doesn't
maintain bloom filter....
It might be hard to maintain bloom filters when merging multiple row groups,
but it should be possible and easy to maintain bloom filter for only one row
group. See ParquetWriter#L337 for related code.
> Add merge blocks command to parquet-tools
> -----------------------------------------
>
> Key: PARQUET-1381
> URL: https://issues.apache.org/jira/browse/PARQUET-1381
> Project: Parquet
> Issue Type: New Feature
> Components: parquet-mr
> Affects Versions: 1.10.0
> Reporter: Ekaterina Galieva
> Assignee: Ekaterina Galieva
> Priority: Major
> Labels: pull-request-available
>
> Current implementation of merge command in parquet-tools doesn't merge row
> groups, just places one after the other. Add API and command option to be
> able to merge small blocks into larger ones up to specified size limit.
> h6. Implementation details:
> Blocks are not reordered not to break possible initial predicate pushdown
> optimizations.
> Blocks are not divided to fit upper bound perfectly.
> This is an intentional performance optimization.
> This gives an opportunity to form new blocks by coping full content of
> smaller blocks by column, not by row.
> h6. Examples:
> # Input files with blocks sizes:
> {code:java}
> [128 | 35], [128 | 40], [120]{code}
> Expected output file blocks sizes:
> {{merge }}
> {code:java}
> [128 | 35 | 128 | 40 | 120]
> {code}
> {{merge -b}}
> {code:java}
> [128 | 35 | 128 | 40 | 120]
> {code}
> {{merge -b -l 256 }}
> {code:java}
> [163 | 168 | 120]
> {code}
> # Input files with blocks sizes:
> {code:java}
> [128 | 35], [40], [120], [6] {code}
> Expected output file blocks sizes:
> {{merge}}
> {code:java}
> [128 | 35 | 40 | 120 | 6]
> {code}
> {{merge -b}}
> {code:java}
> [128 | 75 | 126]
> {code}
> {{merge -b -l 256}}
> {code:java}
> [203 | 126]{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
