rdblue commented on a change in pull request #1145: URL: https://github.com/apache/iceberg/pull/1145#discussion_r449712596
########## File path: flink/src/main/java/org/apache/iceberg/flink/IcebergStreamWriter.java ########## @@ -0,0 +1,225 @@ +/* + * 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.flink; + +import java.io.IOException; +import java.util.List; +import java.util.Locale; +import java.util.Map; +import org.apache.flink.api.common.ExecutionConfig; +import org.apache.flink.api.java.ClosureCleaner; +import org.apache.flink.streaming.api.datastream.DataStream; +import org.apache.flink.streaming.api.functions.sink.SinkFunction; +import org.apache.flink.streaming.api.operators.AbstractStreamOperator; +import org.apache.flink.streaming.api.operators.OneInputStreamOperator; +import org.apache.flink.streaming.runtime.streamrecord.StreamRecord; +import org.apache.flink.table.api.TableSchema; +import org.apache.flink.types.Row; +import org.apache.hadoop.conf.Configuration; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.FileFormat; +import org.apache.iceberg.MetricsConfig; +import org.apache.iceberg.Schema; +import org.apache.iceberg.Table; +import org.apache.iceberg.exceptions.RuntimeIOException; +import org.apache.iceberg.flink.data.FlinkParquetWriters; +import org.apache.iceberg.flink.writer.FileAppenderFactory; +import org.apache.iceberg.flink.writer.OutputFileFactory; +import org.apache.iceberg.flink.writer.TaskWriter; +import org.apache.iceberg.flink.writer.TaskWriterFactory; +import org.apache.iceberg.hadoop.SerializableConfiguration; +import org.apache.iceberg.io.FileAppender; +import org.apache.iceberg.io.OutputFile; +import org.apache.iceberg.parquet.Parquet; +import org.apache.iceberg.relocated.com.google.common.base.Preconditions; +import org.apache.iceberg.types.TypeUtil; +import org.apache.iceberg.util.PropertyUtil; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import static org.apache.iceberg.TableProperties.DEFAULT_FILE_FORMAT; +import static org.apache.iceberg.TableProperties.DEFAULT_FILE_FORMAT_DEFAULT; +import static org.apache.iceberg.TableProperties.WRITE_TARGET_FILE_SIZE_BYTES; +import static org.apache.iceberg.TableProperties.WRITE_TARGET_FILE_SIZE_BYTES_DEFAULT; + +class IcebergStreamWriter extends AbstractStreamOperator<SerializableDataFile> + implements OneInputStreamOperator<Row, SerializableDataFile> { + + private static final long serialVersionUID = 1L; + private static final Logger LOG = LoggerFactory.getLogger(IcebergStreamWriter.class); + + private final String tablePath; + private final SerializableConfiguration conf; + private Schema readSchema; + + private transient Table table; + private transient TaskWriter<Row> writer; + private transient int subTaskId; + + /** + * Be careful to do the initialization in this constructor, because in {@link DataStream#addSink(SinkFunction)} + * it will call {@link ClosureCleaner#clean(Object, ExecutionConfig.ClosureCleanerLevel, boolean)} to set all the + * non-serializable members to be null. + * + * @param tablePath The base path of the iceberg table. + * @param readSchema The schema of source data. + * @param conf The hadoop's configuration. + */ + private IcebergStreamWriter(String tablePath, Schema readSchema, Configuration conf) { + this.tablePath = tablePath; + this.conf = new SerializableConfiguration(conf); + this.readSchema = readSchema; + } + + @Override + public void open() { + this.table = TableUtil.findTable(tablePath, conf.get()); + if (this.readSchema != null) { + // reassign ids to match the existing table schema + readSchema = TypeUtil.reassignIds(readSchema, table.schema()); + TypeUtil.validateWriteSchema(readSchema, table.schema(), true, true); + } + + this.subTaskId = getRuntimeContext().getIndexOfThisSubtask(); + + // Initialize the task writer. + FileFormat fileFormat = getFileFormat(); + FileAppenderFactory<Row> appenderFactory = new FlinkFileAppenderFactory(table); + OutputFileFactory outputFileFactory = new OutputFileFactory(table, fileFormat, subTaskId); + this.writer = TaskWriterFactory.createTaskWriter(table.spec(), + appenderFactory, + outputFileFactory, + getTargetFileSizeBytes(), + fileFormat); + } + + @Override + public void prepareSnapshotPreBarrier(long checkpointId) throws Exception { + LOG.info("Iceberg writer {} subtask {} begin preparing for checkpoint {}", tablePath, subTaskId, checkpointId); + // close all open files and emit files to downstream committer operator + writer.close(); + for (DataFile dataFile : writer.getCompleteFiles()) { + emit(dataFile); + } + // Remember to clear the writer's cached complete files. + writer.reset(); + LOG.info("Iceberg writer {} subtask {} completed preparing for checkpoint {}", tablePath, subTaskId, checkpointId); + } + + @Override + public void processElement(StreamRecord<Row> element) throws Exception { + Row value = element.getValue(); + writer.append(value); + + // Emit the data file entries to downstream committer operator if there exist any complete files. + List<DataFile> completeFiles = writer.getCompleteFiles(); + if (!completeFiles.isEmpty()) { + completeFiles.forEach(this::emit); + // Remember to clear the writer's cached complete files. + writer.reset(); Review comment: I would normally expect `reset` to restore the writer to a new state for reuse. Looks like in this case it is clearing the files that have already been consumed. I think that this could have a better name for the operation. Also, I would rather improve this API. It seems brittle to get the list and then reset it through the writer. I think it would be cleaner to have a `poll` call that retrieves one data file and removes it from the internal tracking, if one is ready. Then this could call `poll` until there are no more files. ---------------------------------------------------------------- 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]
