spark WriteStatsTracker 源码

spark WriteStatsTracker 代码

文件路径：/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/WriteStatsTracker.scala

/*
 * 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.spark.sql.execution.datasources

import org.apache.spark.sql.catalyst.InternalRow


/**
 * To be implemented by classes that represent data statistics collected during a Write Task.
 * It is important that instances of this type are [[Serializable]], as they will be gathered
 * on the driver from all executors.
 */
trait WriteTaskStats extends Serializable


/**
 * A trait for classes that are capable of collecting statistics on data that's being processed by
 * a single write task in [[FileFormatWriter]] - i.e. there should be one instance per executor.
 *
 * newPartition event is only triggered if the relation to be written out is partitioned.
 */
trait WriteTaskStatsTracker {

  /**
   * Process the fact that a new partition is about to be written.
   * Only triggered when the relation is partitioned by a (non-empty) sequence of columns.
   * @param partitionValues The values that define this new partition.
   */
  def newPartition(partitionValues: InternalRow): Unit

  /**
   * Process the fact that a new file is about to be written.
   * @param filePath Path of the file into which future rows will be written.
   */
  def newFile(filePath: String): Unit

  /**
   * Process the fact that a file is finished to be written and closed.
   * @param filePath Path of the file.
   */
  def closeFile(filePath: String): Unit

  /**
   * Process the fact that a new row to update the tracked statistics accordingly.
   * @note Keep in mind that any overhead here is per-row, obviously,
   *       so implementations should be as lightweight as possible.
   * @param filePath Path of the file which the row is written to.
   * @param row Current data row to be processed.
   */
  def newRow(filePath: String, row: InternalRow): Unit

  /**
   * Returns the final statistics computed so far.
   * @param taskCommitTime Time of committing the task.
   * @note This may only be called once. Further use of the object may lead to undefined behavior.
   * @return An object of subtype of [[WriteTaskStats]], to be sent to the driver.
   */
  def getFinalStats(taskCommitTime: Long): WriteTaskStats
}


/**
 * A class implementing this trait is basically a collection of parameters that are necessary
 * for instantiating a (derived type of) [[WriteTaskStatsTracker]] on all executors and then
 * process the statistics produced by them (e.g. save them to memory/disk, issue warnings, etc).
 * It is therefore important that such an objects is [[Serializable]], as it will be sent
 * from the driver to all executors.
 */
trait WriteJobStatsTracker extends Serializable {

  /**
   * Instantiates a [[WriteTaskStatsTracker]], based on (non-transient) members of this class.
   * To be called by executors.
   * @return A [[WriteTaskStatsTracker]] instance to be used for computing stats during a write task
   */
  def newTaskInstance(): WriteTaskStatsTracker

  /**
   * Process the given collection of stats computed during this job.
   * E.g. aggregate them, write them to memory / disk, issue warnings, whatever.
   * @param stats One [[WriteTaskStats]] object from each successful write task.
   * @param jobCommitTime Time of committing the job.
   * @note The type of @param `stats` is too generic. These classes should probably be parametrized:
   *   WriteTaskStatsTracker[S <: WriteTaskStats]
   *   WriteJobStatsTracker[S <: WriteTaskStats, T <: WriteTaskStatsTracker[S]]
   * and this would then be:
   *   def processStats(stats: Seq[S]): Unit
   * but then we wouldn't be able to have a Seq[WriteJobStatsTracker] due to type
   * co-/contra-variance considerations. Instead, you may feel free to just cast `stats`
   * to the expected derived type when implementing this method in a derived class.
   * The framework will make sure to call this with the right arguments.
   */
  def processStats(stats: Seq[WriteTaskStats], jobCommitTime: Long): Unit
}

相关信息

spark 源码目录

相关文章

spark AggregatePushDownUtils 源码

spark ApplyCharTypePadding 源码

spark BasicWriteStatsTracker 源码

spark BucketingUtils 源码

spark CatalogFileIndex 源码

spark CodecStreams 源码

spark DataSource 源码

spark DataSourceStrategy 源码

spark DataSourceUtils 源码

spark DaysWritable 源码