window函数 resetAccumulator
/*
* 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.flink.table.runtime.aggregate import java.lang.Iterable import org.apache.flink.api.common.functions.{MapPartitionFunction, RichGroupReduceFunction}
import org.apache.flink.configuration.Configuration
import org.apache.flink.table.codegen.{Compiler, GeneratedAggregationsFunction}
import org.apache.flink.table.util.Logging
import org.apache.flink.types.Row
import org.apache.flink.util.Collector /**
* [[RichGroupReduceFunction]] and [[MapPartitionFunction]] to compute aggregates that do
* not support pre-aggregation for batch(DataSet) queries.
*
* @param genAggregations Code-generated [[GeneratedAggregations]]
*/
class DataSetAggFunction(
private val genAggregations: GeneratedAggregationsFunction)
extends RichGroupReduceFunction[Row, Row]
with MapPartitionFunction[Row, Row]
with Compiler[GeneratedAggregations] with Logging { private var output: Row = _
private var accumulators: Row = _ private var function: GeneratedAggregations = _ override def open(config: Configuration) {
LOG.debug(s"Compiling AggregateHelper: $genAggregations.name \n\n " +
s"Code:\n$genAggregations.code")
val clazz = compile(
getRuntimeContext.getUserCodeClassLoader,
genAggregations.name,
genAggregations.code)
LOG.debug("Instantiating AggregateHelper.")
function = clazz.newInstance() output = function.createOutputRow()
accumulators = function.createAccumulators()
} /**
* Computes a non-pre-aggregated aggregation.
*
* @param records An iterator over all records of the group.
* @param out The collector to hand results to.
*/
override def reduce(records: Iterable[Row], out: Collector[Row]): Unit = { // reset accumulators
function.resetAccumulator(accumulators) val iterator = records.iterator() var record: Row = null
while (iterator.hasNext) {
record = iterator.next() // accumulate
function.accumulate(accumulators, record)
} // set group keys value to final output
function.setForwardedFields(record, output) // set agg results to output
function.setAggregationResults(accumulators, output) out.collect(output)
} /**
* Computes a non-pre-aggregated aggregation and returns a row even if the input is empty.
*
* @param records An iterator over all records of the partition.
* @param out The collector to hand results to.
*/
override def mapPartition(records: Iterable[Row], out: Collector[Row]): Unit = {
reduce(records, out)
} } 、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、、
/*
* 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.flink.table.runtime.aggregate import java.lang.Iterable import org.apache.flink.api.common.functions.RichGroupReduceFunction
import org.apache.flink.configuration.Configuration
import org.apache.flink.table.codegen.{Compiler, GeneratedAggregationsFunction}
import org.apache.flink.table.util.Logging
import org.apache.flink.types.Row
import org.apache.flink.util.Collector /**
* It wraps the aggregate logic inside of
* [[org.apache.flink.api.java.operators.GroupReduceOperator]]. It is used for tumbling time-window
* on batch.
*
* @param genAggregations Code-generated [[GeneratedAggregations]]
* @param windowSize Tumbling time window size
* @param windowStartPos The relative window-start field position to the last field of output row
* @param windowEndPos The relative window-end field position to the last field of output row
* @param windowRowtimePos The relative window-rowtime field position to the last field of
* output row
* @param keysAndAggregatesArity The total arity of keys and aggregates
*/
class DataSetTumbleTimeWindowAggReduceGroupFunction(
genAggregations: GeneratedAggregationsFunction,
windowSize: Long,
windowStartPos: Option[Int],
windowEndPos: Option[Int],
windowRowtimePos: Option[Int],
keysAndAggregatesArity: Int)
extends RichGroupReduceFunction[Row, Row]
with Compiler[GeneratedAggregations]
with Logging { private var collector: DataSetTimeWindowPropertyCollector = _
protected var aggregateBuffer: Row = new Row(keysAndAggregatesArity + 1) private var output: Row = _
protected var accumulators: Row = _ protected var function: GeneratedAggregations = _ override def open(config: Configuration) {
LOG.debug(s"Compiling AggregateHelper: $genAggregations.name \n\n " +
s"Code:\n$genAggregations.code")
val clazz = compile(
getRuntimeContext.getUserCodeClassLoader,
genAggregations.name,
genAggregations.code)
LOG.debug("Instantiating AggregateHelper.")
function = clazz.newInstance() output = function.createOutputRow()
accumulators = function.createAccumulators()
collector = new DataSetTimeWindowPropertyCollector(
windowStartPos,
windowEndPos,
windowRowtimePos)
} override def reduce(records: Iterable[Row], out: Collector[Row]): Unit = { var last: Row = null
val iterator = records.iterator() // reset accumulator
function.resetAccumulator(accumulators) while (iterator.hasNext) {
val record = iterator.next()
function.mergeAccumulatorsPair(accumulators, record)
last = record
} // set group keys value to final output.
function.setForwardedFields(last, output) // get final aggregate value and set to output.
function.setAggregationResults(accumulators, output) // get window start timestamp
val startTs: Long = last.getField(keysAndAggregatesArity).asInstanceOf[Long] // set collector and window
collector.wrappedCollector = out
collector.windowStart = startTs
collector.windowEnd = startTs + windowSize collector.collect(output)
} }
window函数 resetAccumulator的更多相关文章
- DStream-04 Window函数的原理和源码
DStream 中 window 函数有两种,一种是普通 WindowedDStream,另外一种是针对 window聚合 优化的 ReducedWindowedDStream. Demo objec ...
- MySQL 对window函数执行sum函数疑似Bug
MySQL 对window函数执行sum函数疑似Bug 使用MySql的窗口函数统计数据时,发现一个小的问题,与大家一起探讨下. 环境配置: mysql-installer-community-8.0 ...
- 使用streaming window函数统计用户不同时间段平均消费金额等指标
场景 现在餐厅老板已经不满足仅仅统计历史用户消费金额总数了,他想知道每个用户半年,每个月,每天,或者一小时消费的总额,来店消费的次数以及平均金额. 给出的例子计算的是每5秒,每30秒,每1分钟的用户消 ...
- javascript中的this与函数讲解
前言 javascript中没有块级作用域(es6以前),javascript中作用域分为函数作用域和全局作用域.并且,大家可以认为全局作用域其实就是Window函数的函数作用域,我们编写的js代码, ...
- avascript中的this与函数讲解
徐某某 一个半路出家的野生程序员 javascript中的this与函数讲解 前言 javascript中没有块级作用域(es6以前),javascript中作用域分为函数作用域和全局作用域.并且,大 ...
- javascript篇-----函数作用域,函数作用域链和声明提前
在一些类似C语言的编程语言中,花括号内的每一段代码都具有各自的作用域,而且变量在声明它们的代码段之外是不可见的(也就是我们不能在代码段外直接访问代码段内声明的变量),我们称之为块级作用域,然而,不同于 ...
- javascript 函数声明与函数表达式的区别
先看一段代码 var f = function g() { return 1; }; if (false) { f = function g(){ return 2; }; } alert(g()); ...
- javascript + jquery函数大全
JAVASCRIPT Array 函数 array创建数组 concat()连接两个或更多的数组,并返回结果. join()把数组中所有元素组成字符串. pop()删除并返回数组的最后一个元素 s ...
- JavaScript window
window -- window对象是BOM中所有对象的核心 window,中文"窗口" window对象除了是BOM中所有对象的父对象外,还包含一些窗口控制函数 全局的windo ...
随机推荐
- Megcup2017 Dogfood
问题描述 小强每天会在小区的某些位置摆一些狗盆,并在狗盆里倒入不同口味的狗粮.而所有的流浪狗都会跑到离自己第k近的狗盆那里吃狗粮,一定的跑动可以帮助狗保持身材. 已知小强牌狗粮目前只有10种口味,我们 ...
- Zabbix检测Mysql数据库的主从同步
在高并发网站架构中,MySQL数据库主从同步是不可或缺的,不过经常会发生由于网络原因或者操作错误,MySQL主从经常会出现不同步的情况,那么如何监控MySQL主从同步,也变成检测网站正常运行的重要环节 ...
- MachineLearning之Logistic回归
一.概述 假设现在有一些数据点,我们用一条直线对这些点进行拟合(该线称为最佳拟合直线),这个拟合过程就称为回归: 利用Logistic回归进行分类的主要思想是: 根据现有数据对分类边界线建立回归公式, ...
- MVC几种找不到资源的解决方式
在MVC中,controller中的Action和View中的.cshtml文件名称有一个对应的关系. 当不对应时,有以下几种情况发生: 一.找不到视图的错误 请求URL:http://localho ...
- Php廖雪峰教程学习与实战
https://www.liaoxuefeng.com/wiki/0014316089557264a6b348958f449949df42a6d3a2e542c000 目录 Python教程 Pyth ...
- IDEA使用笔记(一)——使用前的基本设置
前言:记忆不好,有些东西需要的时候又需要找一找,那就不如让“纸和笔”来帮忙记录一下啦!到时候查找也方便,而且是自己的东西印象更加的深刻,说不定还能帮助到他人多好玩的事情! 软件的下载.安装就不记啦!自 ...
- 如何调试bash脚本
Bash 是Linux操作系统的默认Shell脚本.Shell是用来处理操作系统和用户交互的一个程序.Shell的脚本可以帮助用户自动化地和操作系统进行交互.你也可以理解为一种脚本式的编程.即然有编程 ...
- DrawItem
原文链接: http://blog.csdn.net/jiftlixu/article/details/4893505 今天从CButton派生了一个类CUIButton,主要用于自绘,按照基本的流程 ...
- 利用 T-sql 的从句 for xml path('') 实现多行合并到一行, 并带有分隔符
T-sql 有一个for xml path('')的从句能把多行结果合并到一行,并成为xml 格式 比如有一张表tb有两列,其格式和数据为: id value—————1 aa1 bb2 aaa2 b ...
- SeqGAN 原理简述
1. 背景GAN在之前发的文章里已经说过了,虽然现在GAN的变种越来越多,用途广泛,但是它们的对抗思想都是没有变化的.简单来说,就是在生成的过程中加入一个可以鉴别真实数据和生成数据的鉴别器,使生成器G ...