org.apache.hudi.io.HoodieWriteHandle Maven / Gradle / Ivy
/*
* 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.hudi.io;
import org.apache.hudi.avro.HoodieAvroUtils;
import org.apache.hudi.client.WriteStatus;
import org.apache.hudi.common.engine.TaskContextSupplier;
import org.apache.hudi.common.fs.FSUtils;
import org.apache.hudi.common.model.HoodieRecord;
import org.apache.hudi.common.model.HoodieRecordPayload;
import org.apache.hudi.common.model.IOType;
import org.apache.hudi.common.util.HoodieTimer;
import org.apache.hudi.common.util.Option;
import org.apache.hudi.common.util.ReflectionUtils;
import org.apache.hudi.config.HoodieWriteConfig;
import org.apache.hudi.exception.HoodieIOException;
import org.apache.hudi.io.storage.HoodieFileWriter;
import org.apache.hudi.io.storage.HoodieFileWriterFactory;
import org.apache.hudi.table.HoodieTable;
import org.apache.hudi.table.marker.WriteMarkersFactory;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.generic.IndexedRecord;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
import java.io.IOException;
import java.util.Collections;
import java.util.List;
/**
* Base class for all write operations logically performed at the file group level.
*/
public abstract class HoodieWriteHandle extends HoodieIOHandle {
private static final Logger LOG = LogManager.getLogger(HoodieWriteHandle.class);
/**
* A special record returned by {@link HoodieRecordPayload}, which means
* {@link HoodieWriteHandle} should just skip this record.
* This record is only used for {@link HoodieRecordPayload} currently, so it should not
* shuffle though network, we can compare the record locally by the equal method.
* The HoodieRecordPayload#combineAndGetUpdateValue and HoodieRecordPayload#getInsertValue
* have 3 kind of return:
* 1、Option.empty
* This means we should delete this record.
* 2、IGNORE_RECORD
* This means we should not process this record,just skip.
* 3、Other non-empty record
* This means we should process this record.
*
* We can see the usage of IGNORE_RECORD in
* org.apache.spark.sql.hudi.command.payload.ExpressionPayload
*/
public static IgnoreRecord IGNORE_RECORD = new IgnoreRecord();
/**
* The specified schema of the table. ("specified" denotes that this is configured by the client,
* as opposed to being implicitly fetched out of the commit metadata)
*/
protected final Schema tableSchema;
protected final Schema tableSchemaWithMetaFields;
/**
* The write schema. In most case the write schema is the same to the
* input schema. But if HoodieWriteConfig#WRITE_SCHEMA is specified,
* we use the WRITE_SCHEMA as the write schema.
*
* This is useful for the case of custom HoodieRecordPayload which do some conversion
* to the incoming record in it. e.g. the ExpressionPayload do the sql expression conversion
* to the input.
*/
protected final Schema writeSchema;
protected final Schema writeSchemaWithMetaFields;
protected HoodieTimer timer;
protected WriteStatus writeStatus;
protected final String partitionPath;
protected final String fileId;
protected final String writeToken;
protected final TaskContextSupplier taskContextSupplier;
public HoodieWriteHandle(HoodieWriteConfig config, String instantTime, String partitionPath,
String fileId, HoodieTable hoodieTable, TaskContextSupplier taskContextSupplier) {
this(config, instantTime, partitionPath, fileId, hoodieTable,
Option.empty(), taskContextSupplier);
}
protected HoodieWriteHandle(HoodieWriteConfig config, String instantTime, String partitionPath, String fileId,
HoodieTable hoodieTable, Option overriddenSchema,
TaskContextSupplier taskContextSupplier) {
super(config, instantTime, hoodieTable);
this.partitionPath = partitionPath;
this.fileId = fileId;
this.tableSchema = overriddenSchema.orElseGet(() -> getSpecifiedTableSchema(config));
this.tableSchemaWithMetaFields = HoodieAvroUtils.addMetadataFields(tableSchema, config.allowOperationMetadataField());
this.writeSchema = overriddenSchema.orElseGet(() -> getWriteSchema(config));
this.writeSchemaWithMetaFields = HoodieAvroUtils.addMetadataFields(writeSchema, config.allowOperationMetadataField());
this.timer = new HoodieTimer().startTimer();
this.writeStatus = (WriteStatus) ReflectionUtils.loadClass(config.getWriteStatusClassName(),
!hoodieTable.getIndex().isImplicitWithStorage(), config.getWriteStatusFailureFraction());
this.taskContextSupplier = taskContextSupplier;
this.writeToken = makeWriteToken();
}
/**
* Get the specified table schema.
* @param config
* @return
*/
private static Schema getSpecifiedTableSchema(HoodieWriteConfig config) {
return new Schema.Parser().parse(config.getSchema());
}
/**
* Get the schema, of the actual write.
*
* @param config
* @return
*/
private static Schema getWriteSchema(HoodieWriteConfig config) {
return new Schema.Parser().parse(config.getWriteSchema());
}
/**
* Generate a write token based on the currently running spark task and its place in the spark dag.
*/
private String makeWriteToken() {
return FSUtils.makeWriteToken(getPartitionId(), getStageId(), getAttemptId());
}
public Path makeNewPath(String partitionPath) {
Path path = FSUtils.getPartitionPath(config.getBasePath(), partitionPath);
try {
if (!fs.exists(path)) {
fs.mkdirs(path); // create a new partition as needed.
}
} catch (IOException e) {
throw new HoodieIOException("Failed to make dir " + path, e);
}
return new Path(path.toString(), FSUtils.makeDataFileName(instantTime, writeToken, fileId,
hoodieTable.getMetaClient().getTableConfig().getBaseFileFormat().getFileExtension()));
}
/**
* Make new file path with given file name.
*/
protected Path makeNewFilePath(String partitionPath, String fileName) {
String relativePath = new Path((partitionPath.isEmpty() ? "" : partitionPath + "/")
+ fileName).toString();
return new Path(config.getBasePath(), relativePath);
}
/**
* Creates an empty marker file corresponding to storage writer path.
*
* @param partitionPath Partition path
*/
protected void createMarkerFile(String partitionPath, String dataFileName) {
WriteMarkersFactory.get(config.getMarkersType(), hoodieTable, instantTime)
.create(partitionPath, dataFileName, getIOType());
}
public Schema getWriterSchemaWithMetaFields() {
return writeSchemaWithMetaFields;
}
/**
* Determines whether we can accept the incoming records, into the current file. Depending on
*
* - Whether it belongs to the same partitionPath as existing records - Whether the current file written bytes lt max
* file size
*/
public boolean canWrite(HoodieRecord record) {
return false;
}
/**
* Perform the actual writing of the given record into the backing file.
*/
public void write(HoodieRecord record, Option insertValue) {
// NO_OP
}
/**
* Perform the actual writing of the given record into the backing file.
*/
public void write(HoodieRecord record, Option avroRecord, Option exception) {
Option recordMetadata = record.getData().getMetadata();
if (exception.isPresent() && exception.get() instanceof Throwable) {
// Not throwing exception from here, since we don't want to fail the entire job for a single record
writeStatus.markFailure(record, exception.get(), recordMetadata);
LOG.error("Error writing record " + record, exception.get());
} else {
write(record, avroRecord);
}
}
/**
* Rewrite the GenericRecord with the Schema containing the Hoodie Metadata fields.
*/
protected GenericRecord rewriteRecord(GenericRecord record) {
return HoodieAvroUtils.rewriteRecord(record, writeSchemaWithMetaFields);
}
public abstract List close();
public List writeStatuses() {
return Collections.singletonList(writeStatus);
}
public String getPartitionPath() {
return partitionPath;
}
public abstract IOType getIOType();
@Override
protected FileSystem getFileSystem() {
return hoodieTable.getMetaClient().getFs();
}
protected int getPartitionId() {
return taskContextSupplier.getPartitionIdSupplier().get();
}
protected int getStageId() {
return taskContextSupplier.getStageIdSupplier().get();
}
protected long getAttemptId() {
return taskContextSupplier.getAttemptIdSupplier().get();
}
protected HoodieFileWriter createNewFileWriter(String instantTime, Path path, HoodieTable hoodieTable,
HoodieWriteConfig config, Schema schema, TaskContextSupplier taskContextSupplier) throws IOException {
return HoodieFileWriterFactory.getFileWriter(instantTime, path, hoodieTable, config, schema, taskContextSupplier);
}
private static class IgnoreRecord implements GenericRecord {
@Override
public void put(int i, Object v) {
}
@Override
public Object get(int i) {
return null;
}
@Override
public Schema getSchema() {
return null;
}
@Override
public void put(String key, Object v) {
}
@Override
public Object get(String key) {
return null;
}
}
}