前端技术
HTML
CSS
Javascript
前端框架和UI库
VUE
ReactJS
AngularJS
JQuery
NodeJS
JSON
Element-UI
Bootstrap
Material UI
服务端和客户端
Java
Python
PHP
Golang
Scala
Kotlin
Groovy
Ruby
Lua
.net
c#
c++
后端WEB和工程框架
SpringBoot
SpringCloud
Struts2
MyBatis
Hibernate
Tornado
Beego
Go-Spring
Go Gin
Go Iris
Dubbo
HessianRPC
Maven
Gradle
数据库
MySQL
Oracle
Mongo
中间件与web容器
Redis
MemCache
Etcd
Cassandra
Kafka
RabbitMQ
RocketMQ
ActiveMQ
Nacos
Consul
Tomcat
Nginx
Netty
大数据技术
Hive
Impala
ClickHouse
DorisDB
Greenplum
PostgreSQL
HBase
Kylin
Hadoop
Apache Pig
ZooKeeper
SeaTunnel
Sqoop
Datax
Flink
Spark
Mahout
数据搜索与日志
ElasticSearch
Apache Lucene
Apache Solr
Kibana
Logstash
数据可视化与OLAP
Apache Atlas
Superset
Saiku
Tesseract
系统与容器
Linux
Shell
Docker
Kubernetes
[配置回滚策略]的搜索结果
这里是文章列表。热门标签的颜色随机变换,标签颜色没有特殊含义。
点击某个标签可搜索标签相关的文章。
点击某个标签可搜索标签相关的文章。
转载文章
...的模式,如状态模式、策略模式、访问者模式本质上是在更特殊的场合采用了委托模式。委托模式使得我们可以用聚合来替代继承。 Java中委托: interface Printer { Kotlin: interface Printer { by表示 p 将会在 PrintImpl 中内部存储, 并且编译器将自动生成转发给 p 的所有 Printer 的方法。 委托属性 有一些常见的属性类型,虽然我们可以在每次需要的时候手动实现它们, 但是如果能够为大家把他们只实现一次并放入一个库会更好。例如包括: 延迟属性(lazy properties): 其值只在首次访问时计算; 可观察属性(observable properties): 监听器会收到有关此属性变更的通知; 把多个属性储存在一个映射(map)中,而不是每个存在单独的字段中。 为了涵盖这些(以及其他)情况,Kotlin 支持 委托属性 。 委托属性的语法是: var : 在 by 后面的表达式是该 委托, 因为属性对应的 get()(和 set())会被委托给它的 getValue() 和 setValue() 方法。 标准委托: Kotlin 标准库为几种有用的委托提供了工厂方法。 延迟属性 Lazy lazy() 接受一个 lambda 并返回一个 Lazy 实例的函数,返回的实例可以作为实现延迟属性的委托:第一次调用 get() 会执行已传递给 lazy() 的 lambda 表达式并记录结果, 后续调用 get() 只是返回记录的结果。例如: val lazyValue: String 可观察属性 Observable Delegates.observable() 接受两个参数:初始值和修改时处理程序(handler)。每当我们给属性赋值时会调用该处理程序(在赋值后执行)。它有三个参数:被赋值的属性、旧值和新值: class User { 如果想拦截赋的新值,并根据你是不是想要这个值来决定是否给属性赋新值,可以使用 vetoable() 取代 observable(),接收的参数和 observable 一样,不过处理程序 返回值是 Boolean 来决定是否采用新值,即在属性被赋新值生效之前 会调用传递给 vetoable 的处理程序。例如: class User { 把属性存在map 中 一个常见的用例是在一个映射(map)里存储属性的值。这经常出现在像解析 JSON 或者做其他“动态”事情的应用中。在这种情况下,你可以使用映射实例自身作为委托来实现委托属性。 例如: class User(map: Map 在上例中,委托属性会从构造函数传入的map中取值(通过字符串键——属性的名称),如果遇到声明的属性名在map 中找不到对应的key 名,或者key 对应的value 值的类型与声明的属性的类型不一致,会抛出异常。 内联函数 当一个函数被声明为inline时,它的函数体是内联的,也就是说,函数体会被直接替换到函数被调用地方 inline函数(内联函数)从概念上讲是编译器使用函数实现的真实代码来替换每一次的函数调用,带来的最直接的好处就是节省了函数调用的开销,而缺点就是增加了所生成字节码的尺寸。基于此,在代码量不是很大的情况下,我们是否有必要将所有的函数定义为内联?让我们分两种情况进行说明: 将普通函数定义为内联:众所周知,JVM内部已经实现了内联优化,它会在任何可以通过内联来提升性能的地方将函数调用内联化,并且相对于手动将普通函数定义为内联,通过JVM内联优化所生成的字节码,每个函数的实现只会出现一次,这样在保证减少运行时开销的同时,也没有增加字节码的尺寸;所以我们可以得出结论,对于普通函数,我们没有必要将其声明为内联函数,而是交给JVM自行优化。 将带有lambda参数的函数定义为内联:是的,这种情况下确实可以提高性能;但在使用的过程中,我们会发现它是有诸多限制的,让我们从下面的例子开始展开说明: inline 假如我们这样调用doSomething: fun main(args: Array<String>) { 上面的调用会被编译成: fun main(args: Array<String>) { 从上面编译的结果可以看出,无论doSomething函数还是action参数都被内联了,很棒,那让我们换一种调用方式: fun main(args: Array<String>) { 上面的调用会被编译成: fun main(args: Array<String>) { doSomething函数被内联,而action参数没有被内联,这是因为以函数型变量的形式传递给doSomething的lambda在函数的调用点是不可用的,只有等到doSomething被内联后,该lambda才可以正常使用。 通过上面的例子,我们对lambda表达式何时被内联做一下简单的总结: 当lambda表达式以参数的形式直接传递给内联函数,那么lambda表达式的代码会被直接替换到最终生成的代码中。 当lambda表达式在某个地方被保存起来,然后以变量形式传递给内联函数,那么此时的lambda表达式的代码将不会被内联。 上面对lambda的内联时机进行了讨论,消化片刻后让我们再看最后一个例子: inline 上面的例子是否有问题?是的,编译器会抛出“Illegal usage of inline-parameter”的错误,这是因为Kotlin规定内联函数中的lambda参数只能被直接调用或者传递给另外一个内联函数,除此之外不能作为他用;那我们如果确实想要将某一个lambda传递给一个非内联函数怎么办?我们只需将上述代码这样改造即可: inline 很简单,在不需要内联的lambda参数前加上noinline修饰符就可以了。 以上便是我对内联函数的全部理解,通过掌握该特性的运行机制,相信大家可以做到在正确的时机使用该特性,而非滥用或因恐惧弃而不用。 Kotlin下单例模式 饿汉式实现 //Java实现 懒汉式 //Java实现 上述代码中,我们可以发现在Kotlin实现中,我们让其主构造函数私有化并自定义了其属性访问器,其余内容大同小异。 如果有小伙伴不清楚Kotlin构造函数的使用方式。请点击 - - - 构造函数 不清楚Kotlin的属性与访问器,请点击 - - -属性和字段 线程安全的懒汉式 //Java实现 大家都知道在使用懒汉式会出现线程安全的问题,需要使用使用同步锁,在Kotlin中,如果你需要将方法声明为同步,需要添加@Synchronized注解。 双重校验锁式 //Java实现 哇!小伙伴们惊喜不,感不感动啊。我们居然几行代码就实现了多行的Java代码。其中我们运用到了Kotlin的延迟属性 Lazy。 Lazy内部实现 public 观察上述代码,因为我们传入的mode = LazyThreadSafetyMode.SYNCHRONIZED, 那么会直接走 SynchronizedLazyImpl,我们继续观察SynchronizedLazyImpl。 Lazy接口 SynchronizedLazyImpl实现了Lazy接口,Lazy具体接口如下: public 继续查看SynchronizedLazyImpl,具体实现如下: SynchronizedLazyImpl内部实现 private 通过上述代码,我们发现 SynchronizedLazyImpl 覆盖了Lazy接口的value属性,并且重新了其属性访问器。其具体逻辑与Java的双重检验是类似的。 到里这里其实大家还是肯定有疑问,我这里只是实例化了SynchronizedLazyImpl对象,并没有进行值的获取,它是怎么拿到高阶函数的返回值呢?。这里又涉及到了委托属性。 委托属性语法是:val/var : by 。在 by 后面的表达式是该 委托, 因为属性对应的 get()(和 set())会被委托给它的 getValue() 和 setValue() 方法。属性的委托不必实现任何的接口,但是需要提供一个 getValue() 函数(和 setValue()——对于 var 属性)。 而Lazy.kt文件中,声明了Lazy接口的getValue扩展函数。故在最终赋值的时候会调用该方法。 internal.InlineOnly 静态内部类式 //Java实现 静态内部类的实现方式,也没有什么好说的。Kotlin与Java实现基本雷同。 补充 在该篇文章结束后,有很多小伙伴咨询,如何在Kotlin版的Double Check,给单例添加一个属性,这里我给大家提供了一个实现的方式。(不好意思,最近才抽出时间来解决这个问题) class SingletonDemo private constructor( 其中关于?:操作符,如果 ?: 左侧表达式非空,就返回其左侧表达式,否则返回右侧表达式。请注意,当且仅当左侧为空时,才会对右侧表达式求值。 Kotlin 智能类型转换 对于子父类之间的类型转换 先看这样一段 Java 代码 public 尽管在 main 函数中,对 person 这个对象进行了类型判断,但是在使用的时候还是需要强制转换成 Student 类型,这样是不是很不智能? 同样的情况在 Kotlin 中就变得简单多了 fun main(args: Array<String>) { 在 Kotlin 中,只要对类型进行了判断,就可以直接通过父类的对象去调用子类的函数了 安全的类型转换 还是上面的那个例子,如果我们没有进行类型判断,并且直接进行强转,会怎么样呢? public static void main(String[] args) { 结果就只能是 Exception in thread "main" java.lang.ClassCastException 那么在 Kotlin 中是不是会有更好的解决方法呢? val person: Person = Person() 在转换操作符后面添加一个 ?,就不会把程序 crash 掉了,当转化失败的时候,就会返回一个 null 在空类型中的智能转换 需要提前了解 Kotlin 类型安全的相关知识(Kotlin 中的类型安全(对空指针的优化处理)) String? = aString 在定义的时候定义成了有可能为 null,按照之前的写法,我们需要这样写 String? = 但是已经进行了是否为 String 类型的判断,所以就一定 不是 空类型了,也就可以直接输出它的长度了 T.()->Unit 、 ()->Unit 在做kotlin开发中,经常看到一些系统函数里,用函数作为参数 public .()-Unit与()->Unit的区别是我们调用时,在代码块里面写this,的时候,两个this代表的含义不一样,T.()->Unit里的this代表的是自身实例,而()->Unit里,this代表的是外部类的实例。 推荐阅读 对 Kotlin 与 Java 编程语言的思考 使用 Kotlin 做开发一个月后的感想 扫一扫 关注我的公众号如果你想要跟大家分享你的文章,欢迎投稿~ 本篇文章为转载内容。原文链接:https://blog.csdn.net/weixin_39611037/article/details/109984124。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-06-23 23:56:14
470
转载
转载文章
...了一种改进的内存映射策略,旨在减少在高负载环境下由于频繁的内存映射操作导致的系统开销,并已在分布式数据库和大数据分析应用中取得了显著效果。 同时,内存映射的安全性问题也引起了业界的关注。今年早些时候,一项安全研究报告揭示了利用mmap进行提权攻击的新方法,再次提醒开发者在享受内存映射带来的便利时,也需要关注其潜在的安全风险,并采取相应的防御措施。 总之,内存映射作为底层系统调用的重要组成部分,其发展与优化将持续影响着整个软件生态系统的性能表现与安全性,值得广大开发者和技术研究者深入探究和实践。
2023-09-20 22:49:12
464
转载
转载文章
...能看到每款设备的详细配置说明。如果您使用 Pixel 设备,现在就可以加入 Android Beta program,然后自动获得最新的 Android P Beta。 马上开始在您喜欢的设备上体验 Android P Beta 吧,欢迎您向我们反馈意见和建议!并请继续关注 Project Treble 的最新动态。 确保 app 兼容 随着越来越多的用户开始体验 Android P Beta,是时候开始测试您 app 的兼容性,以尽早解决在测试中发现的问题并尽快发布更新。请查看迁移手册了解操作步骤以及 Android P 的时间推进表。 请从 Google Play 下载您的应用,并在运行 Android P Beta 的设备或模拟器上测试用户流程。确保您的应用体验良好,并正确处理 Android P 的行为变更。尤其注意动态电量管理、Wi-Fi 权限变化、后台调用摄像头以及传感器的限制、针对应用数据的 SELinux 政策、默认启用 TLS 的变化,以及 Build.SERIAL 限制。 · 公开 API 的兼容性 (Compatibility through public APIs) 针对非 SDK 接口的测试十分重要。正如我们之前所强调的,在 Android P 中,我们将逐渐收紧一些非 SDK 接口的使用,这也要求广大的开发者们,包括 Google 内部的应用团队,使用公开 API。 如果您的应用正在使用私有 Android API 或者库,您需要改为使用 Android SDK 或 NDK 公开的 API。我们在 DP1 里已经对使用私有接口的开发者发出了警告信息,从 Android P Beta 开始,调用非 SDK 接口将会报错 (部分被豁免的私有 API 除外) —— 也就是说您的应用将会遭遇异常,而不再只是警告了。 为了帮助您定位非 SDK API 的使用情况,我们在 StrictMode 里加入了两个新的方法。您可以使用 detectNonSdkApiUsage() 在应用通过反射或 JNI 调用非 SDK API 的时候收到警报,您还可以使用 permitNonSdkApiUsage() 来阻止 StrictMode 针对这些调用报错。这些方法都可助您了解应用调用非 SDK API 的情况,但请注意,即便调用的 API 暂时得到了豁免,最保险的做法依然是尽快放弃对它们的使用。 如果您确实遇到了公开 API 无法满足需求的情况,请立刻告知我们。更多详细内容请查看相关文档。 · 凹口屏测试 (Test with display cutout) 针对凹口屏测试您的应用也十分重要。现在您可以在运行 Android P Beta 的合作伙伴机型上测试,确保您的应用在凹口屏上表现良好。同时,您也可以在 Android P 设备的开发者选项里打开对凹口屏的模拟,对您的应用做相应测试。 体验 Android P 在准备好开发条件后,请深入了解 Android P 并学习可以在您的应用中使用到的全新功能和 API。为了帮助您更轻松地探索和使用新 API,请查阅 API 变化报告 (API 27->DP2, DP1->DP2) 以及 Android P API 文档。访问开发者预览版网站了解详情。 下载/更新 Android P 开发者预览版 SDK 和工具包至 Android Studio 3.1,或使用最新版本的 Android Studio 3.2。如果您手边没有 Android P Beta 设备 (或查看今天推送的次条文章),请使用 Android P 模拟器来运行和测试您的应用。 您的反馈一直都至关重要,我们欢迎您畅所欲言。如果您在开发或测试过程中遇到了问题,请在文章下方留言给我们。再次感谢大家一路以来的支持。 请点击蓝色字体前往 “Android Developers 官方文档”查看详细说明 本篇文章为转载内容。原文链接:https://blog.csdn.net/weixin_34258782/article/details/87952581。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-04-10 18:19:36
338
转载
转载文章
...模型采用了多任务学习策略,旨在提高对图像中特定区域(例如手部)的分割精度和整体效果。在本文研究中,作者选取了SeedNet网络的第一阶段进行实验,并展示了其在证件照生成背景分割上的应用效果。
2023-07-11 23:36:51
131
转载
转载文章
...M实现高效的UI更新策略,大大提升了网页渲染性能。 同时,在Web组件化开发领域,Custom Elements V1规范已得到广泛支持,开发者可以通过自定义HTML元素并扩展其行为,这背后离不开DOM API的强力支撑。例如,借助MutationObserver接口可以监听DOM树的变化,实现实时响应式布局。 此外,无障碍性(Accessibility)也是当前Web开发的重要考量因素之一,正确且高效的DOM操作有助于提升网站对屏幕阅读器等辅助技术的支持,确保信息能够无障碍地传达给所有用户。 近期,W3C还在持续推动DOM标准的发展,如Shadow DOM v1规范让组件样式和DOM结构更加独立和可控,对于构建复杂Web应用具有重要意义。了解和掌握这些前沿技术和标准,将有助于开发者更好地利用DOM API创建高性能、可维护且符合现代Web标准的页面和应用。
2023-08-04 13:36:05
247
转载
转载文章
...达式优化用户输入验证策略”的文章详尽探讨了如何结合现代浏览器特性,如约束验证API,配合正则表达式进行高效的数据校验。 此外,对于正则表达式的性能优化也是值得关注的话题。有研究指出,在处理大量数据时,某些复杂的正则可能导致性能瓶颈。阅读相关的性能分析报告和技术分享,可以帮助开发者掌握编写高性能正则表达式的技巧,并避免潜在的性能陷阱。 最后,关于UTC时间戳在跨时区开发中的重要性,可参考有关国际协作项目中如何妥善处理时间问题的文章,了解如何借助JavaScript Date对象正确转换和处理不同时区的时间信息,从而确保在全球范围内应用程序的正常运行。尤其在当前全球化的互联网环境下,理解和掌握这一技能愈发关键。
2023-01-24 13:01:25
529
转载
转载文章
...外,在图片加载与缓存策略方面,除了文中提及的开源库如universal-image-loader和Volley,Google自家的Glide库凭借其高度优化的内存管理和磁盘缓存策略,已成为众多开发者首选的图片加载工具。Glide不仅实现了三级缓存,还特别针对Android设备的特性进行了深度优化,进一步提升了应用的流畅度和用户体验。 而对于面试中的XMPP协议通信技术,尽管仍被广泛采用,但近年来WebRTC和MQTT等新兴通信协议也逐渐崭露头角,特别是在实时音视频通话及物联网场景中,它们因更低延迟和更高效率受到业界青睐。因此,Android开发者应紧跟行业趋势,了解并掌握多种通信协议及其应用场景,以适应不断变化的技术需求。 总的来说,无论是面试技巧还是技术储备,持续学习和积累都是提升竞争力的关键。在实际工作中深入理解Android系统原理,关注行业最新动态和技术演进方向,将有助于求职者更好地应对各类面试挑战,并在未来的职业道路上取得成功。
2023-06-19 17:42:52
336
转载
转载文章
...互比较同一算法的不同配置。 创建比较算法的结果图。 下面的示例在皮马印第安人发病的糖尿病数据集中将Logistic回归和线性判别分析进行了比较。 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Compare Algorithms from pandas import read_csv from sklearn.model_selection import KFold from sklearn.model_selection import cross_val_score from sklearn.linear_model import LogisticRegression from sklearn.discriminant_analysis import LinearDiscriminantAnalysis load dataset url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/pima-indians-diabetes.data.csv" names = ['preg', 'plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', 'class'] dataframe = read_csv(url, names=names) array = dataframe.values X = array[:,0:8] Y = array[:,8] prepare models models = [] models.append(('LR', LogisticRegression(solver='liblinear'))) models.append(('LDA', LinearDiscriminantAnalysis())) evaluate each model in turn results = [] names = [] scoring = 'accuracy' for name, model in models: kfold = KFold(n_splits=10, random_state=7) cv_results = cross_val_score(model, X, Y, cv=kfold, scoring=scoring) results.append(cv_results) names.append(name) msg = "%s: %f (%f)" % (name, cv_results.mean(), cv_results.std()) print(msg) 哪种算法效果更好?你能做得更好吗?在评论中让我知道。 第11课:通过算法调整提高准确性 一旦找到一种或两种在数据集上表现良好的算法,您可能希望提高这些模型的性能。 提高算法性能的一种方法是将其参数调整为特定的数据集。 scikit-learn库提供了两种方法来搜索机器学习算法的参数组合。在今天的课程中,您的目标是练习每个。 使用您指定的网格搜索来调整算法的参数。 使用随机搜索调整算法的参数。 下面使用的代码段是一个示例,该示例使用网格搜索在Pima Indians糖尿病发病数据集上的Ridge回归算法。 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Grid Search for Algorithm Tuning from pandas import read_csv import numpy from sklearn.linear_model import Ridge from sklearn.model_selection import GridSearchCV url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/pima-indians-diabetes.data.csv" names = ['preg', 'plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', 'class'] dataframe = read_csv(url, names=names) array = dataframe.values X = array[:,0:8] Y = array[:,8] alphas = numpy.array([1,0.1,0.01,0.001,0.0001,0]) param_grid = dict(alpha=alphas) model = Ridge() grid = GridSearchCV(estimator=model, param_grid=param_grid, cv=3) grid.fit(X, Y) print(grid.best_score_) print(grid.best_estimator_.alpha) 哪些参数取得最佳效果?你能做得更好吗?在评论中让我知道。 第12课:利用集合预测提高准确性 您可以提高模型性能的另一种方法是组合来自多个模型的预测。 一些模型提供了内置的此功能,例如用于装袋的随机森林和用于增强的随机梯度增强。可以使用另一种称为投票的合奏将来自多个不同模型的预测组合在一起。 在今天的课程中,您将练习使用合奏方法。 使用随机森林和多余树木算法练习装袋。 使用梯度增强机和AdaBoost算法练习增强合奏。 通过将来自多个模型的预测组合在一起来练习投票合奏。 下面的代码段演示了如何在Pima Indians糖尿病发病数据集上使用随机森林算法(袋装决策树集合)。 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Random Forest Classification from pandas import read_csv from sklearn.model_selection import KFold from sklearn.model_selection import cross_val_score from sklearn.ensemble import RandomForestClassifier url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/pima-indians-diabetes.data.csv" names = ['preg', 'plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', 'class'] dataframe = read_csv(url, names=names) array = dataframe.values X = array[:,0:8] Y = array[:,8] num_trees = 100 max_features = 3 kfold = KFold(n_splits=10, random_state=7) model = RandomForestClassifier(n_estimators=num_trees, max_features=max_features) results = cross_val_score(model, X, Y, cv=kfold) print(results.mean()) 你能设计出更好的合奏吗?在评论中让我知道。 第13课:完成并保存模型 找到有关机器学习问题的良好模型后,您需要完成该模型。 在今天的课程中,您将练习与完成模型有关的任务。 练习使用模型对新数据(在训练和测试过程中看不到的数据)进行预测。 练习将经过训练的模型保存到文件中,然后再次加载。 例如,下面的代码片段显示了如何创建Logistic回归模型,将其保存到文件中,之后再加载它以及对看不见的数据进行预测。 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Save Model Using Pickle from pandas import read_csv from sklearn.model_selection import train_test_split from sklearn.linear_model import LogisticRegression import pickle url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/pima-indians-diabetes.data.csv" names = ['preg', 'plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', 'class'] dataframe = read_csv(url, names=names) array = dataframe.values X = array[:,0:8] Y = array[:,8] test_size = 0.33 seed = 7 X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=test_size, random_state=seed) Fit the model on 33% model = LogisticRegression(solver='liblinear') model.fit(X_train, Y_train) save the model to disk filename = 'finalized_model.sav' pickle.dump(model, open(filename, 'wb')) some time later... load the model from disk loaded_model = pickle.load(open(filename, 'rb')) result = loaded_model.score(X_test, Y_test) print(result) 第14课:Hello World端到端项目 您现在知道如何完成预测建模机器学习问题的每个任务。 在今天的课程中,您需要练习将各个部分组合在一起,并通过端到端的标准机器学习数据集进行操作。 端到端遍历虹膜数据集(机器学习的世界) 这包括以下步骤: 使用描述性统计数据和可视化了解您的数据。 预处理数据以最好地揭示问题的结构。 使用您自己的测试工具抽查多种算法。 使用算法参数调整来改善结果。 使用集成方法改善结果。 最终确定模型以备将来使用。 慢慢进行,并记录结果。 您使用什么型号?您得到了什么结果?在评论中让我知道。 结束! (看你走了多远) 你做到了。做得好! 花一点时间,回头看看你已经走了多远。 您最初对机器学习感兴趣,并强烈希望能够使用Python练习和应用机器学习。 您可能是第一次下载,安装并启动Python,并开始熟悉该语言的语法。 在许多课程中,您逐渐地,稳定地学习了预测建模机器学习项目的标准任务如何映射到Python平台上。 基于常见机器学习任务的配方,您使用Python端到端解决了第一个机器学习问题。 使用标准模板,您所收集的食谱和经验现在可以自行解决新的和不同的预测建模机器学习问题。 不要轻描淡写,您在短时间内就取得了长足的进步。 这只是您使用Python进行机器学习的起点。继续练习和发展自己的技能。 喜欢点下关注,你的关注是我写作的最大支持 本篇文章为转载内容。原文链接:https://blog.csdn.net/m0_37337849/article/details/104016531。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-07-11 10:04:06
92
转载
转载文章
... xfree86命令配置的X Windows效果比Xconfigurator好很多的。界面最友好的,当属XF86Setup工具,若你有安装这个工具的话,不妨就用这个工具来设置你的显卡吧。 八、如何设置网卡 九、如何播放CD音乐 声卡设置好了,可以在光驱中插入一张CD,用装载命令mount将光盘载入,然后输入cdp命令就可以播放了。在屏幕上可以看到CD的音轨清单,小键盘的9是播放,7是停止,6进到下一音轨,4退到前一音轨,0退出播放程序,2弹出CD,8是暂停。 如果是在X-Window中,可以用xplaycd播放CD,这是一个有图形界面的播放器,按钮及功能一目了然,这里就不多说了。 10、如何播放VCD影碟 11、如何拨号上网 12、Linux下能玩游戏吗 光盘中有bsd-games-2.1-3.src.rpm文件包,用RPM命令进行安装,然后到安装目录中去找游戏吧。不过这些游戏都不够精致,如果你想玩好的,就须要进入X-Window,acm是空战模拟游戏,paradise和xpilot是联网战斗游戏,xdemineur是挖地雷(没想到吧),xjewel是俄罗斯方块,xboard是国际象棋,xpat2是扑克牌游戏,xboing是弹珠台游戏,还有Doom——大名鼎鼎的第一人称射击游戏!这些游戏有的可以直接调出,有的须要用RPM命令安装。所有的RPM包都在安装光盘中的srpms目录下,自己去看看吧。 十.用xvidtune调整你的显示器 大家会发现装了linux之后在windows下用的好好的显示器有时进到linux的xwindows里后就歪掉了,调整好之后回到windows后windows的桌面也外调了,来回启动系统每次都要调整很麻烦的,这里介绍一个办法一劳永逸 j进入linux启动x在xterm里执行xvidtune,会弹出这个软件的窗口,点Auto然后点Left,Right等按钮调整你的显示器到最佳的位置,然后点界面上的Show按钮会得到类似这样的输出: "1152x864" 121.50 1152 1232 1360 1568 864 865 868 911 +hsync -vsync 然后退出这个软件,修改你的/etc/X11/XF86Config-4文件在 Section "Monitor" Identifier "AS 786T" VendorName "Unknown" ModelName "Unknown" HorizSync 30 - 87 VertRefresh 50 - 160 Option "dpms" EndSection 里加上刚才的输出,我的是: Section "Monitor" Identifier "AS 786T" VendorName "Unknown" ModelName "Unknown" HorizSync 30 - 87 VertRefresh 50 - 160 Modeline "1152x864" 121.50 1152 1232 1360 1568 864 865 868 911 +hsync -vsync Option "dpms" EndSection 保存然后重起试试看吧 十一.问:我的机器是windows和linux双系统,如何改变grub默认启动的系统? 答:这需要修改/boot/grub/grub.conf。举一个例子你就明白了。假设你的/boot/grub/grub.conf是这样子的: default=0 timeout=10 splashimage=(hd0,7)/grub/splash.xpm.gz title Red Hat Linux (2.4.18-14) root (hd0,7) kernel /vmlinuz-2.4.18-14 ro root=LABEL=/ initrd /initrd-2.4.18-14.img title DOS rootnoverify (hd0,0) chainloader +1 那么你的grub会默认启动Red Hat Linux (2.4.18-14)这个系统,把default=0改成default=1,那么grub会默认启动DOS这个系统。注意,这里的要点是:你想默认启动第n个title所指的系统,那么default应该是等于n-1 十二.问:我的文本控制台怎么总是出现乱码呢? 答:这是因为你安装了中文支持的缘故。解决的方法是安装一个zhcon(一个快速地外挂式CJK(中文/日文/韩文)的多内码平台),我把他放在附件中提供大家下载。关于zhcon的更进一步的消息,大家可以到他的官方主页zhcon.gnuchina.org查看。安装和使用请参考这个连接 http://hepg.sdu.edu.cn/Service/tips/zhcon_manual.html zhcon下载连接: http://zhcon.gnuchina.org/download/...on-0.2.1.tar.gz 十三.问:我在安装一个软件的时候,提示我缺少一个.so文件,安装无法继续,怎么办? 答:.so文件就像windows中的.dll文件一样,是库文件。一个程序的正常安装和运行需要特定的库文件的支持。所以你需要去找到包含这个.so的包装上。去 http://www.rpmfind.net用你缺的那个.....剿枰?rpm包 十四.我访问windows分区时发现所有windows分区中的文件和文件夹名中的中文全变成问号,怎么办? 答:在第三贴中我们讲解了通过编辑/etc/fstab实现在linux中访问windows的fat32分区。同样,我们可以通过进一步修改 /etc/fstab来实现中文文件名显示。只要把/dev/hda1 /mnt/c vfat default 0 0中的default全改为iocharset=cp936就行了。 十五.我的rh8.0中的XMMS不好使,不能播放MP3,怎么办? 答:这是因为rh公司怕别人告他侵权,所以在rh8.0中去掉了XMMS对MP3的支持,8.0以前的版本都是好使的。在8.0中要解决也很简单,装一个插件就行了。这个插件我放在本贴的附件里,rpm格式,经winrar压缩 附件: http://www.chinalinuxpub.com/vbbfor...s=&postid=86299 十六.问:我在linux中怎样才能使用windows分区呢? 答:先说一点背景知识。linux支持很多种文件系统,包括windows的fat32和ntfs。对fat32的支持已经很好,可以直接使用,而对ntfs 的支持还不是太好,只能读,而写是极危险的,并且对ntfs的支持不是默认的,也就是说你想要使用ntfs的话,需要重新编译内核。鉴于重编内核对于新手的复杂性,这里只讲解使用fat32分区的方法下面给出上述问题的两种解决方案:1.在安装系统(linux),进行到分区选择挂载点时,你可以建立几个挂载点,如/mnt/c,/mnt/d等,然后选择你的windows fat32分区,把它们分别挂载到前面建立的挂载点即可。(注意,正如前面所说,在这里你不能把一个ntfs分区挂载到一个挂载点,应为ntfs不是默认支持的。)这样你装好系统后就能直接使用你的windows fat32分区了。例如,你把windows的c盘(linux中的/dev/hda1)挂载到/mnt/c,那么你就能在/mnt/c目录中找到你的c 盘中的全部数据。2. 如果你在安装系统时没有像方案1所说的那样挂载上你的fat32分区,没关系,仍然能够很方便的解决这个问题。首先,用一个文本编辑器(如vi)打开 /etc/fstab,在文件的最后加入类似如下的几行 /dev/hda1 /mnt/c vfat default 0 0 你所要做的修改就是,把/dev/hda1改成你要挂载的fat32分区在linux中的设备号,把/mnt/c改成相应的挂载点即可。注意,挂载点就是一个目录,这个目录要事先建立。举一个例子,我有三个fat32分区,在windows中是c,d,e盘,在linux中的设备号分别为 /dev/hda1,/dev/hda5,/dev/hda6。那么我就要先建立3个挂载点,如/mnt/c,/mnt/d,/mnt/e,然后在 /etc/fstab中加上这么几行: /dev/hda1 /mnt/c vfat default 0 0 /dev/hda5 /mnt/d vfat default 0 0 /dev/hda6 /mnt/e vfat default 0 0 保存一下退出编辑器。这样以后你重启机器后就能直接使用c,d,e这三个fat32格式的windows分区了 十七.问:我的机器重装windows后,开机启动就直接进入了windows,原来的linux进不去了,怎么办? 答:这是由于windows的霸道。重装windows后,windows重写了你的mbr,覆盖掉了grub。解决方法很简单:用你的linux第一张安装盘引导进入linx rescue模式(如何进入?你注意一下系统的提示信息就知道了),执行下面两条命令就可以了 chroot /mnt/sysimage 改变你的根目录 grub-install /dev/hda 安装grub到mbr 十八.问:我的linux开机直接进入文本界面,怎样才能让它默认进入图形界面? 答:修改/etc/inittab文件,其中有一行id:3:initdefault,意思是说开机默认进入运行级别3(多用户的文本界面),把它改成id:5:initdefault,既开机默认进入运行级别5(多用户的图形界面)。这样就行了。 十九.如何同时启动多个x 以前的帖子,估计很多人没看过,贴出来温习一下 Linux里的X-Windows以其独特的面貌和强大的功能吸引了很多原先对linux不感兴趣的人,特别是KDE和GNOME,功能强大不说,而且自带了很多很棒的软件,界面非常友好,很适合于初学者。下面告诉大家一个同时启动6个X的小技巧: 在~/.bashrc中加入 以下几行: alias X=startx -- -bpp 32 -quiet& alias X1=startx -- :1 -bpp 32 -quiet& alias X2=startx -- :2 -bpp 32 -quiet& alias X3=startx -- :3 -bpp 32 -quiet& alias X4=startx -- :4 -bpp 32 -quiet& alias X5=startx -- :5 -bpp 32 -quiet& 其中32是显示器的色彩深度,你应该根据自己的实际情况设置。 之后运行 bash 使改变生效,以后只要依次运行X,X1,X2,X3,X4,X5就可以启动6个X-Windows了。 二十.装了rpm的postgresql之后启动 /etc/init.d/postgresql start 是不能启动postgresql的tcp/ip连接支持的,所以打开/etc/init.d/postgresql这个文件把 su -l postgres -s /bin/sh -c "/usr/bin/pg_ctl -D $PGDATA -p /usr/bin/postmaster start > /dev/null 2>&1" < /dev/null 改为: su -l postgres -s /bin/sh -c "/usr/bin/pg_ctl -o -o -F -i -w -D $PGDATA -p /usr/bin/postmaster start > /dev/null 2>&1" < /dev/null 这样就可以启动数据库的tcp/ip链接了 二十一.如何将man转存为文本文件 以ls的man为例 man ls |col -b >ls.txt 将info变成文本,以make为例 info make -o make.txt -s 二十二.如何在文本模式下发送2进制文件 首先检查系统有没有uuencode 和 uudecode如果没有从光盘上装 rpm -ivh sharutils-x.xx.x-x.rpm 假设要发送的文件是vpopmail-5.2.1.tar.gz执行 uuencode -m vpopmail-5.2.1.tar.gz vpopmail.tar.gz>encodefile 说明: uuenode是编码命令,-m是使用mime64编码,vpopmail-5.2.1.tar.gz是要编码的文件,vpopmail.tar.gz是如果解码后得到的文件名,encodefile是编码后的文件名。 执行上述命令之后就可以通过mail命令发送编码后的文件了 mail chenlf@chinalinuxpub.com<encodefile 好了,现在我来接收邮件 在控制台上输入mail命令: mail Mail version 8.1 6/6/93. Type ? for help. "/var/spool/mail/chenlf": 2 messages 2 new >N 1 chenlf@ns1.catv.net Mon Jun 10 16:44 17/363 N 2 root@ns2.catv.net Mon Jun 10 16:45 6091/371145 & 2 Message 2: From root@ns2.catv.net Mon Jun 10 16:45:28 2002 Date: Mon, 10 Jun 2002 16:44:51 +0800 From: root <root@ns2.catv.net> To: chenlf@chinalinuxpub.com begin-base64 644 vpopmai.tar.gz H4sIABr15TwAA+w9a2PbNpL7NfwVqNPbWIlFPSzbiR2n9SuxE7/OcuLNtdmU EiGLMUWqfFhWt7u//eYBgKRE2U7iTa+3VndjiQQGg5nBYDAYDC6H4XDgeH51 yW7ajdpf/h2fer1VX1lagr/1+spyq/BXff5SX2mtNBZXmovN5l/qjWZrqfEX sfRvwWbik8aJEwnxl7ifDofXlLvp/Z/0c1nk/8uN/777NuqNen251ZrB/+XF pcUG8r/ZbC0vL9ZXoPwi/O8von73qEx//sP5bwHHxanT8aUIe2IrDBIZJLFl 7QVJFFovpZOkkYxFL4yEFhVLCKhk1W2xG45E1wnEnohlIsJAiksvSlLHF24I JQORhKIjRdKXYhh5Ayca6xcAD8DQm4HT7XuB/EGcSXgbPErEyAkSrNp3LqVw grGoyaRbGzpxPHJFGssotq0Gtw6l9gTgJbixode9EOlQDMaTmEjE/AerydVc rAY4jJzIFY7vC3wL2DgJvJIxIjFwkm6fWkfw1KoAIti/EgkWc3A6YRp05ReB aeXAQH34GoXOwAvOVUnoEnwRYRqJeJAMgczRpYzEyEv6YQoUH8oACltLtjjD Rr1YOCJ2BkPgJop1IuJu5A0TYh9xIdQwfrCWTdt9pMKvaZg4j5jT3PgojC5+ sFZswM0LAJzvSyhGXQSCOmLoO9DtEOAicBCD2qUT1agAg44BSd+1niIEzVPs ................. ................. ................. & s 2 encodefile "encode" [New file] & q 然后进行解码 uudecode encodefile ls encodefile vpopmai.tar.gz tar zxvf vpopmail.tar.gz OK了 二十三.将 man page 转成 HTML 格式 使用 man2html 这个指令,就可以将 man page 转成 HTML 格式了。用法是: man2html filename > htmlfile.html 二十四.如何在gnome和kde之间切换。 如果你是以图形登录方式登录linux,那么点击登录界面上的session(任务)即可以选择gnome和kde。如果你是以文本方式登录,那执行switchdesk gnome或switchdesk kde,然后再startx就可以进入gnome或kde。 25...tar,.tar.gz,.bz2,.tar.bz2,.bz,.gz是什么文件,如何解开他们? 他们都是文件(压缩)包。 .tar:把文件打包,不压缩:tar cvf .tar dirName 解开:tar xvf .tar .tar.gz:把文件打包并压缩:tar czvf .tar.gz dirName 解开:tar xzvf .tar.gz .bz2:解开:bzip2 -d .bz2 .bz:解开:bzip -d .bz .gz:解开:gzip -d .gz 26.linux下如何解开.zip,.rar压缩文件? rh8下有一个图形界面的软件file-roller可以做这件事。令外可以用unzip .zip解开zip文件,unrar .rar解开rar文件,不过unrar一般系统不自带,要到网上下载。 27.linux下如何浏览.iso光盘镜像文件? a.建一个目录,如:mkdir a b.把iso文件挂载到该目录上:mount -o loop xxxx.iso a 现在目录a里的内容就是iso文件里的内容了。 28.linux下如何配置网络? 用netconfig。“IP address:”就是要配置的IP地址,“Netmask:”子网掩码,“Default gateway (IP):”网关,“Primary nameserver:”DNS服务器IP。 29.如何让鼠标支持滚轮? 在配置鼠标时,选择微软的鼠标,并正确选择端口如ps2,usb等 30.如何让控制台支持中文显示? 安装zhcon。zhcon需要libimm_server.so和libpth.so.13这两个库支持。一般的中文输入法应该都有libimm_server.so。libpth.so.13出自pth-1.3.x。把这两个文件放到/usr/lib下就行了。 31.如何配置grub? 修改/boot/grub/grub.conf文件。其中 “default=n”(n是个数字)是grub引导菜单默认被选中的项,n从0开始,0表示第一项,1表示第二项,依此类推。 “timeout=x”(x是一个数)是超时时间,单位是妙。也就是引导菜单显示后,如果x秒内用户不进行选择,那么grub将启动默认项。 “splashimage =xxxxxx”,这是引导菜单的背景图,先不理他。 其它常用项我用下面的例子来说明: title Red Hat 8.0 root (hd1,6) kernel /boot/vmlinuz-2.4.18-14 ro root=/dev/hdb7 initrd /boot/initrd-2.4.18-14.img 其中"Red Hat 8.0"是在启动菜单列表里显示的名字 root (hdx,y)用来指定你的boot分区位置,如果你没有分boot分区(本例就没分boot分区),那就指向根分区就行了,hdx是linux所在硬盘,hd0是第一块硬盘,hd1是第二块,依此类推。y是分区位置,从0开始,也就是等于分区号减一,比如你要指向的分区是hdx7,那么y就是6,如果是hdx1,那y就是0。注意root后面要有一个空格。 kernel /boot/vmlinuz-2.4.18-14,其中"/boot/vmlinuz-2.4.18-14"是你要用的内核路径,如果你编译了心内核,把它改成你的新内核的路径就行了。 ro就不用管,写上不会有错。 root=/dev/hdxx指定根分区,本例是hdb7,所以root=/dev/hdb7 initrd xxxxxxxxxxxxx这行不要也行,目前我还不清楚它是做什么用的。 上面是linux的,下面是windows的 title windows 98 rootnoverify (hd0,0) chainloader +1 title xxxxxxx不用解释了,上面有解释。 rootnoverify (hdx,y)用来指定windows所在分区,x,y跟上面一样,注意rootnoverify后有空格。 chainloader +1照抄就行,注意空格。 本篇文章为转载内容。原文链接:https://blog.csdn.net/gudulyn/article/details/764890。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-10-27 09:27:49
255
转载
转载文章
...所以需要设计一些召回策略,每次推荐时从海量内容中筛选出千级别的内容库。召回策略最重要的要求是性能要极致,一般超时不能超过50毫秒。 召回策略种类有很多,我们主要用的是倒排的思路。离线维护一个倒排,这个倒排的key可以是分类,topic,实体,来源等。 排序考虑热度、新鲜度、动作等。线上召回可以迅速从倒排中根据用户兴趣标签对内容做截断,高效的从很大的内容库中筛选比较靠谱的一小部分内容。 二、内容分析 内容分析包括文本分析,图片分析和视频分析。头条一开始主要做资讯,今天我们主要讲一下文本分析。文本分析在推荐系统中一个很重要的作用是用户兴趣建模。 没有内容及文本标签,无法得到用户兴趣标签。举个例子,只有知道文章标签是互联网,用户看了互联网标签的文章,才能知道用户有互联网标签,其他关键词也一样。 另一方面,文本内容的标签可以直接帮助推荐特征,比如魅族的内容可以推荐给关注魅族的用户,这是用户标签的匹配。 如果某段时间推荐主频道效果不理想,出现推荐窄化,用户会发现到具体的频道推荐(如科技、体育、娱乐、军事等)中阅读后,再回主feed,推荐效果会更好。 因为整个模型是打通的,子频道探索空间较小,更容易满足用户需求。只通过单一信道反馈提高推荐准确率难度会比较大,子频道做的好很重要。而这也需要好的内容分析。 上图是今日头条的一个实际文本case。可以看到,这篇文章有分类、关键词、topic、实体词等文本特征。 当然不是没有文本特征,推荐系统就不能工作,推荐系统最早期应用在Amazon,甚至沃尔玛时代就有,包括Netfilx做视频推荐也没有文本特征直接协同过滤推荐。 但对资讯类产品而言,大部分是消费当天内容,没有文本特征新内容冷启动非常困难,协同类特征无法解决文章冷启动问题。 今日头条推荐系统主要抽取的文本特征包括以下几类。首先是语义标签类特征,显式为文章打上语义标签。 这部分标签是由人定义的特征,每个标签有明确的意义,标签体系是预定义的。 此外还有隐式语义特征,主要是topic特征和关键词特征,其中topic特征是对于词概率分布的描述,无明确意义;而关键词特征会基于一些统一特征描述,无明确集合。 另外文本相似度特征也非常重要。在头条,曾经用户反馈最大的问题之一就是为什么总推荐重复的内容。这个问题的难点在于,每个人对重复的定义不一样。 举个例子,有人觉得这篇讲皇马和巴萨的文章,昨天已经看过类似内容,今天还说这两个队那就是重复。 但对于一个重度球迷而言,尤其是巴萨的球迷,恨不得所有报道都看一遍。解决这一问题需要根据判断相似文章的主题、行文、主体等内容,根据这些特征做线上策略。 同样,还有时空特征,分析内容的发生地点以及时效性。比如武汉限行的事情推给北京用户可能就没有意义。 最后还要考虑质量相关特征,判断内容是否低俗,色情,是否是软文,鸡汤? 上图是头条语义标签的特征和使用场景。他们之间层级不同,要求不同。 分类的目标是覆盖全面,希望每篇内容每段视频都有分类;而实体体系要求精准,相同名字或内容要能明确区分究竟指代哪一个人或物,但不用覆盖很全。 概念体系则负责解决比较精确又属于抽象概念的语义。这是我们最初的分类,实践中发现分类和概念在技术上能互用,后来统一用了一套技术架构。 目前,隐式语义特征已经可以很好的帮助推荐,而语义标签需要持续标注,新名词新概念不断出现,标注也要不断迭代。其做好的难度和资源投入要远大于隐式语义特征,那为什么还需要语义标签? 有一些产品上的需要,比如频道需要有明确定义的分类内容和容易理解的文本标签体系。语义标签的效果是检查一个公司NLP技术水平的试金石。 今日头条推荐系统的线上分类采用典型的层次化文本分类算法。 最上面Root,下面第一层的分类是像科技、体育、财经、娱乐,体育这样的大类,再下面细分足球、篮球、乒乓球、网球、田径、游泳…,足球再细分国际足球、中国足球,中国足球又细分中甲、中超、国家队…,相比单独的分类器,利用层次化文本分类算法能更好地解决数据倾斜的问题。 有一些例外是,如果要提高召回,可以看到我们连接了一些飞线。这套架构通用,但根据不同的问题难度,每个元分类器可以异构,像有些分类SVM效果很好,有些要结合CNN,有些要结合RNN再处理一下。 上图是一个实体词识别算法的case。基于分词结果和词性标注选取候选,期间可能需要根据知识库做一些拼接,有些实体是几个词的组合,要确定哪几个词结合在一起能映射实体的描述。 如果结果映射多个实体还要通过词向量、topic分布甚至词频本身等去歧,最后计算一个相关性模型。 三、用户标签 内容分析和用户标签是推荐系统的两大基石。内容分析涉及到机器学习的内容多一些,相比而言,用户标签工程挑战更大。 今日头条常用的用户标签包括用户感兴趣的类别和主题、关键词、来源、基于兴趣的用户聚类以及各种垂直兴趣特征(车型,体育球队,股票等)。还有性别、年龄、地点等信息。 性别信息通过用户第三方社交账号登录得到。年龄信息通常由模型预测,通过机型、阅读时间分布等预估。 常驻地点来自用户授权访问位置信息,在位置信息的基础上通过传统聚类的方法拿到常驻点。 常驻点结合其他信息,可以推测用户的工作地点、出差地点、旅游地点。这些用户标签非常有助于推荐。 当然最简单的用户标签是浏览过的内容标签。但这里涉及到一些数据处理策略。 主要包括: 一、过滤噪声。通过停留时间短的点击,过滤标题党。 二、热点惩罚。对用户在一些热门文章(如前段时间PG One的新闻)上的动作做降权处理。理论上,传播范围较大的内容,置信度会下降。 三、时间衰减。用户兴趣会发生偏移,因此策略更偏向新的用户行为。因此,随着用户动作的增加,老的特征权重会随时间衰减,新动作贡献的特征权重会更大。 四、惩罚展现。如果一篇推荐给用户的文章没有被点击,相关特征(类别,关键词,来源)权重会被惩罚。当 然同时,也要考虑全局背景,是不是相关内容推送比较多,以及相关的关闭和dislike信号等。 用户标签挖掘总体比较简单,主要还是刚刚提到的工程挑战。头条用户标签第一版是批量计算框架,流程比较简单,每天抽取昨天的日活用户过去两个月的动作数据,在Hadoop集群上批量计算结果。 但问题在于,随着用户高速增长,兴趣模型种类和其他批量处理任务都在增加,涉及到的计算量太大。 2014年,批量处理任务几百万用户标签更新的Hadoop任务,当天完成已经开始勉强。集群计算资源紧张很容易影响其它工作,集中写入分布式存储系统的压力也开始增大,并且用户兴趣标签更新延迟越来越高。 面对这些挑战。2014年底今日头条上线了用户标签Storm集群流式计算系统。改成流式之后,只要有用户动作更新就更新标签,CPU代价比较小,可以节省80%的CPU时间,大大降低了计算资源开销。 同时,只需几十台机器就可以支撑每天数千万用户的兴趣模型更新,并且特征更新速度非常快,基本可以做到准实时。这套系统从上线一直使用至今。 当然,我们也发现并非所有用户标签都需要流式系统。像用户的性别、年龄、常驻地点这些信息,不需要实时重复计算,就仍然保留daily更新。 四、评估分析 上面介绍了推荐系统的整体架构,那么如何评估推荐效果好不好? 有一句我认为非常有智慧的话,“一个事情没法评估就没法优化”。对推荐系统也是一样。 事实上,很多因素都会影响推荐效果。比如侯选集合变化,召回模块的改进或增加,推荐特征的增加,模型架构的改进在,算法参数的优化等等,不一一举例。 评估的意义就在于,很多优化最终可能是负向效果,并不是优化上线后效果就会改进。 全面的评估推荐系统,需要完备的评估体系、强大的实验平台以及易用的经验分析工具。 所谓完备的体系就是并非单一指标衡量,不能只看点击率或者停留时长等,需要综合评估。 很多公司算法做的不好,并非是工程师能力不够,而是需要一个强大的实验平台,还有便捷的实验分析工具,可以智能分析数据指标的置信度。 一个良好的评估体系建立需要遵循几个原则,首先是兼顾短期指标与长期指标。我在之前公司负责电商方向的时候观察到,很多策略调整短期内用户觉得新鲜,但是长期看其实没有任何助益。 其次,要兼顾用户指标和生态指标。既要为内容创作者提供价值,让他更有尊严的创作,也有义务满足用户,这两者要平衡。 还有广告主利益也要考虑,这是多方博弈和平衡的过程。 另外,要注意协同效应的影响。实验中严格的流量隔离很难做到,要注意外部效应。 强大的实验平台非常直接的优点是,当同时在线的实验比较多时,可以由平台自动分配流量,无需人工沟通,并且实验结束流量立即回收,提高管理效率。 这能帮助公司降低分析成本,加快算法迭代效应,使整个系统的算法优化工作能够快速往前推进。 这是头条A/B Test实验系统的基本原理。首先我们会做在离线状态下做好用户分桶,然后线上分配实验流量,将桶里用户打上标签,分给实验组。 举个例子,开一个10%流量的实验,两个实验组各5%,一个5%是基线,策略和线上大盘一样,另外一个是新的策略。 实验过程中用户动作会被搜集,基本上是准实时,每小时都可以看到。但因为小时数据有波动,通常是以天为时间节点来看。动作搜集后会有日志处理、分布式统计、写入数据库,非常便捷。 在这个系统下工程师只需要设置流量需求、实验时间、定义特殊过滤条件,自定义实验组ID。系统可以自动生成:实验数据对比、实验数据置信度、实验结论总结以及实验优化建议。 当然,只有实验平台是远远不够的。线上实验平台只能通过数据指标变化推测用户体验的变化,但数据指标和用户体验存在差异,很多指标不能完全量化。 很多改进仍然要通过人工分析,重大改进需要人工评估二次确认。 五、内容安全 最后要介绍今日头条在内容安全上的一些举措。头条现在已经是国内最大的内容创作与分发凭条,必须越来越重视社会责任和行业领导者的责任。如果1%的推荐内容出现问题,就会产生较大的影响。 现在,今日头条的内容主要来源于两部分,一是具有成熟内容生产能力的PGC平台 一是UGC用户内容,如问答、用户评论、微头条。这两部分内容需要通过统一的审核机制。如果是数量相对少的PGC内容,会直接进行风险审核,没有问题会大范围推荐。 UGC内容需要经过一个风险模型的过滤,有问题的会进入二次风险审核。审核通过后,内容会被真正进行推荐。这时如果收到一定量以上的评论或者举报负向反馈,还会再回到复审环节,有问题直接下架。 整个机制相对而言比较健全,作为行业领先者,在内容安全上,今日头条一直用最高的标准要求自己。 分享内容识别技术主要鉴黄模型,谩骂模型以及低俗模型。今日头条的低俗模型通过深度学习算法训练,样本库非常大,图片、文本同时分析。 这部分模型更注重召回率,准确率甚至可以牺牲一些。谩骂模型的样本库同样超过百万,召回率高达95%+,准确率80%+。如果用户经常出言不讳或者不当的评论,我们有一些惩罚机制。 泛低质识别涉及的情况非常多,像假新闻、黑稿、题文不符、标题党、内容质量低等等,这部分内容由机器理解是非常难的,需要大量反馈信息,包括其他样本信息比对。 目前低质模型的准确率和召回率都不是特别高,还需要结合人工复审,将阈值提高。目前最终的召回已达到95%,这部分其实还有非常多的工作可以做。别平台。 如果需要机器学习视频,可以在公众号后台聊天框回复【机器学习】,可以免费获取编程视频 。 你可能还喜欢 数学在机器学习中到底有多重要? AI 新手学习路线,附上最详细的资源整理! 提升机器学习数学基础,推荐7本书 酷爆了!围观2020年十大科技趋势 机器学习该如何入门,听听过来人的经验! 长按加入T圈,接触人工智能 觉得内容还不错的话,给我点个“在看”呗 本篇文章为转载内容。原文链接:https://blog.csdn.net/itcodexy/article/details/109574173。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2024-01-13 09:21:23
322
转载
转载文章
...整个应用的启动、自动配置、组件扫描等功能,使得开发者能够快速搭建稳定、高效且易于维护的后端服务,例如定义Service和Controller层接口并实现相关业务逻辑。 Timestamp , Timestamp是一种数据库中的时间戳类型,表示从1970年1月1日(UTC/GMT的午夜)开始所经过的秒数,精确到微秒级别。在文中提到的SeckillGoods实体类中,startDate和endDate字段采用了Timestamp类型,以便精确记录秒杀活动的开始和结束时间,并使用DateTimeFormat注解进行格式化处理,确保与前端展示的时间格式一致。 VO(Value Object) , VO是值对象(Value Object)的简称,在面向对象编程领域中,VO通常用来封装从数据库查询或由用户输入的数据,仅包含属性以及它们的getter和setter方法,不包含行为。在本文中,创建了SeckillGoodsVo这个实体类VO,用于连表查询时接收和展示商品名字等多张表的关联数据,便于前后端之间的数据传输和展示。 前后端分离架构 , 前后端分离架构是一种常见的Web应用程序设计模式,其中前端专注于用户界面的设计和交互逻辑,而后端则关注业务逻辑处理、数据存储和API接口提供。在本篇文章中,前端通过Ajax请求调用后端提供的RESTful API获取数据并渲染页面,实现了前后端职责清晰、开发并行且可独立部署升级的现代Web应用架构。
2023-02-25 23:20:34
121
转载
转载文章
...文分词采用了多种算法策略,如基于前缀词典构建有向无环图(DAG)进行扫描、动态规划查找最大概率路径以及HMM模型处理未登录词等技术,旨在准确高效地识别和切分出文本中的词汇单元。 TextRank算法 , TextRank是一种基于图排序理论的关键词抽取算法,其基本思想来源于PageRank算法,常用于信息检索和文本摘要等领域。在jieba库中,TextRank算法被应用于提取句子或文档中的关键词,通过统计词语间的共现关系构建网络,并计算节点的PageRank值来确定关键词的重要性。 PaddlePaddle , PaddlePaddle是由百度公司研发的开源深度学习框架,全称为“PArallel Distributed Deep LEarning”,适用于大规模数据训练和高性能推理场景。在jieba分词组件中,paddle模式利用PaddlePaddle框架训练序列标注模型(如双向GRU),实现更高级别的中文分词功能,同时支持词性标注,提升了对复杂语境下词汇切分与理解的能力。 TF-IDF算法 , TF-IDF(Term Frequency-Inverse Document Frequency)是一种常用的文本挖掘技术,用于评估一个词语对于一份文档或者一组文档集的重要性。在jieba.analyse模块中,通过TF-IDF算法可以为文本中的词语计算权重,从而有效地从大量文本中提取最具代表性和区分度的关键词,帮助用户快速了解文本主题和关键信息。
2023-12-02 10:38:37
500
转载
转载文章
...下的适应性及精度提升策略,提出了一种结合深度学习算法进行谱线解卷积和背景扣除的新方法,有望进一步提高LIBS-LIF在实际环境监测中的准确性和可靠性。 综上所述,LIBS-LIF技术作为前沿的元素分析手段,在环境监测方面的潜力正逐渐被挖掘并广泛应用,未来将在更广泛的环境污染治理、生态保护以及环境风险评估等领域发挥重要作用。
2023-08-13 12:41:47
360
转载
转载文章
...zlib选项来禁用 配置hosts 所有主机上都配置 [root@xl2 11] cat /etc/hosts127.0.0.1 localhost192.168.20.132 gtm192.168.20.133 xl1192.168.20.134 xl2 关闭防火墙、Selinux 所有主机都执行 关闭防火墙: [root@gtm ~] systemctl stop firewalld.service[root@gtm ~] systemctl disable firewalld.service selinux设置: [root@gtm ~]vim /etc/selinux/config 设置SELINUX=disabled,保存退出。 This file controls the state of SELinux on the system. SELINUX= can take one of these three values: enforcing - SELinux security policy is enforced. permissive - SELinux prints warnings instead of enforcing. disabled - No SELinux policy is loaded.SELINUX=disabled SELINUXTYPE= can take one of three two values: targeted - Targeted processes are protected, minimum - Modification of targeted policy. Only selected processes are protected. mls - Multi Level Security protection. 安装依赖包 所有主机上都执行 yum install -y flex bison readline-devel zlib-devel openjade docbook-style-dsssl gcc 创建用户 所有主机上都执行 [root@gtm ~] useradd postgres[root@gtm ~] passwd postgres[root@gtm ~] su - postgres[root@gtm ~] mkdir ~/.ssh[root@gtm ~] chmod 700 ~/.ssh 配置SSH免密登录 仅仅在gtm节点配置如下操作: [root@gtm ~] su - postgres[postgres@gtm ~] ssh-keygen -t rsa[postgres@gtm ~] cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys[postgres@gtm ~] chmod 600 ~/.ssh/authorized_keys 将刚生成的认证文件拷贝到xl1到xl2中,使得gtm节点可以免密码登录xl1~xl2的任意一个节点: [postgres@gtm ~] scp ~/.ssh/authorized_keys postgres@xl1:~/.ssh/[postgres@gtm ~] scp ~/.ssh/authorized_keys postgres@xl2:~/.ssh/ 对所有提示都不要输入,直接enter下一步。直到最后,因为第一次要求输入目标机器的用户密码,输入即可。 下载源码 下载地址:https://www.postgres-xl.org/download/ [root@slave ~] ll postgres-xl-10r1.1.tar.gz-rw-r--r-- 1 root root 28121666 May 30 05:21 postgres-xl-10r1.1.tar.gz 编译、安装Postgres-XL 所有节点都安装,编译需要一点时间,最好同时进行编译。 [root@slave ~] tar xvf postgres-xl-10r1.1.tar.gz[root@slave ~] ./configure --prefix=/home/postgres/pgxl/[root@slave ~] make[root@slave ~] make install[root@slave ~] cd contrib/ --安装必要的工具,在gtm节点上安装即可[root@slave ~] make[root@slave ~] make install 配置环境变量 所有节点都要配置 进入postgres用户,修改其环境变量,开始编辑 [root@gtm ~]su - postgres[postgres@gtm ~]vi .bashrc --不是.bash_profile 在打开的文件末尾,新增如下变量配置: export PGHOME=/home/postgres/pgxlexport LD_LIBRARY_PATH=$PGHOME/lib:$LD_LIBRARY_PATHexport PATH=$PGHOME/bin:$PATH 按住esc,然后输入:wq!保存退出。输入以下命令对更改重启生效。 [postgres@gtm ~] source .bashrc --不是.bash_profile 输入以下语句,如果输出变量结果,代表生效 [postgres@gtm ~] echo $PGHOME 应该输出/home/postgres/pgxl代表生效 配置集群 生成pgxc_ctl.conf配置文件 [postgres@gtm ~] pgxc_ctl prepare/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxl/pgxc_ctl/pgxc_ctl_bash.ERROR: File "/home/postgres/pgxl/pgxc_ctl/pgxc_ctl.conf" not found or not a regular file. No such file or directoryInstalling pgxc_ctl_bash script as /home/postgres/pgxl/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxl/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxl/pgxc_ctl --configuration /home/postgres/pgxl/pgxc_ctl/pgxc_ctl.confFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxl/pgxc_ctl 配置pgxc_ctl.conf 新建/home/postgres/pgxc_ctl/pgxc_ctl.conf文件,编辑如下: 对着模板文件一个一个修改,否则会造成初始化过程出现各种神奇问题。 pgxcInstallDir=$PGHOMEpgxlDATA=$PGHOME/data pgxcOwner=postgres---- GTM Master -----------------------------------------gtmName=gtmgtmMasterServer=gtmgtmMasterPort=6666gtmMasterDir=$pgxlDATA/nodes/gtmgtmSlave=y Specify y if you configure GTM Slave. Otherwise, GTM slave will not be configured and all the following variables will be reset.gtmSlaveName=gtmSlavegtmSlaveServer=gtm value none means GTM slave is not available. Give none if you don't configure GTM Slave.gtmSlavePort=20001 Not used if you don't configure GTM slave.gtmSlaveDir=$pgxlDATA/nodes/gtmSlave Not used if you don't configure GTM slave.---- GTM-Proxy Master -------gtmProxyDir=$pgxlDATA/nodes/gtm_proxygtmProxy=y gtmProxyNames=(gtm_pxy1 gtm_pxy2) gtmProxyServers=(xl1 xl2) gtmProxyPorts=(6666 6666) gtmProxyDirs=($gtmProxyDir $gtmProxyDir) ---- Coordinators ---------coordMasterDir=$pgxlDATA/nodes/coordcoordNames=(coord1 coord2) coordPorts=(5432 5432) poolerPorts=(6667 6667) coordPgHbaEntries=(0.0.0.0/0)coordMasterServers=(xl1 xl2) coordMasterDirs=($coordMasterDir $coordMasterDir)coordMaxWALsernder=0 没设置备份节点,设置为0coordMaxWALSenders=($coordMaxWALsernder $coordMaxWALsernder) 数量保持和coordMasterServers一致coordSlave=n---- Datanodes ----------datanodeMasterDir=$pgxlDATA/nodes/dn_masterprimaryDatanode=xl1 主数据节点datanodeNames=(node1 node2)datanodePorts=(5433 5433) datanodePoolerPorts=(6668 6668) datanodePgHbaEntries=(0.0.0.0/0)datanodeMasterServers=(xl1 xl2)datanodeMasterDirs=($datanodeMasterDir $datanodeMasterDir)datanodeMaxWalSender=4datanodeMaxWALSenders=($datanodeMaxWalSender $datanodeMaxWalSender) 集群初始化,启动,停止 初始化 pgxc_ctl -c /home/postgres/pgxc_ctl/pgxc_ctl.conf init all 输出结果: /bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.conf/home/postgres/pgxc_ctl/pgxc_ctl.conf: line 189: $coordExtraConfig: ambiguous redirectFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlStopping all the coordinator masters.Stopping coordinator master coord1.Stopping coordinator master coord2.pg_ctl: directory "/home/postgres/pgxc/nodes/coord/coord1" does not existpg_ctl: directory "/home/postgres/pgxc/nodes/coord/coord2" does not existDone.Stopping all the datanode masters.Stopping datanode master datanode1.Stopping datanode master datanode2.pg_ctl: PID file "/home/postgres/pgxc/nodes/datanode/datanode1/postmaster.pid" does not existIs server running?Done.Stop GTM masterwaiting for server to shut down.... doneserver stopped[postgres@gtm ~]$ echo $PGHOME/home/postgres/pgxl[postgres@gtm ~]$ ll /home/postgres/pgxl/pgxc/nodes/gtm/gtm.^C[postgres@gtm ~]$ pgxc_ctl -c /home/postgres/pgxc_ctl/pgxc_ctl.conf init all/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.conf/home/postgres/pgxc_ctl/pgxc_ctl.conf: line 189: $coordExtraConfig: ambiguous redirectFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlInitialize GTM masterERROR: target directory (/home/postgres/pgxc/nodes/gtm) exists and not empty. Skip GTM initilializationDone.Start GTM masterserver startingInitialize all the coordinator masters.Initialize coordinator master coord1.ERROR: target coordinator master coord1 is running now. Skip initilialization.Initialize coordinator master coord2.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/coord/coord2 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.Done.Starting coordinator master.Starting coordinator master coord1ERROR: target coordinator master coord1 is already running now. Skip initialization.Starting coordinator master coord22019-05-30 21:09:25.562 EDT [2148] LOG: listening on IPv4 address "0.0.0.0", port 54322019-05-30 21:09:25.562 EDT [2148] LOG: listening on IPv6 address "::", port 54322019-05-30 21:09:25.563 EDT [2148] LOG: listening on Unix socket "/tmp/.s.PGSQL.5432"2019-05-30 21:09:25.601 EDT [2149] LOG: database system was shut down at 2019-05-30 21:09:22 EDT2019-05-30 21:09:25.605 EDT [2148] LOG: database system is ready to accept connections2019-05-30 21:09:25.612 EDT [2156] LOG: cluster monitor startedDone.Initialize all the datanode masters.Initialize the datanode master datanode1.Initialize the datanode master datanode2.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/datanode/datanode1 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/datanode/datanode2 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.Done.Starting all the datanode masters.Starting datanode master datanode1.WARNING: datanode master datanode1 is running now. Skipping.Starting datanode master datanode2.2019-05-30 21:09:33.352 EDT [2404] LOG: listening on IPv4 address "0.0.0.0", port 154322019-05-30 21:09:33.352 EDT [2404] LOG: listening on IPv6 address "::", port 154322019-05-30 21:09:33.355 EDT [2404] LOG: listening on Unix socket "/tmp/.s.PGSQL.15432"2019-05-30 21:09:33.392 EDT [2404] LOG: redirecting log output to logging collector process2019-05-30 21:09:33.392 EDT [2404] HINT: Future log output will appear in directory "pg_log".Done.psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"Done.psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"Done.[postgres@gtm ~]$ pgxc_ctl -c /home/postgres/pgxc_ctl/pgxc_ctl.conf stop all/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.conf/home/postgres/pgxc_ctl/pgxc_ctl.conf: line 189: $coordExtraConfig: ambiguous redirectFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlStopping all the coordinator masters.Stopping coordinator master coord1.Stopping coordinator master coord2.pg_ctl: directory "/home/postgres/pgxc/nodes/coord/coord1" does not existDone.Stopping all the datanode masters.Stopping datanode master datanode1.Stopping datanode master datanode2.pg_ctl: PID file "/home/postgres/pgxc/nodes/datanode/datanode1/postmaster.pid" does not existIs server running?Done.Stop GTM masterwaiting for server to shut down.... doneserver stopped[postgres@gtm ~]$ pgxc_ctl/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.conf/home/postgres/pgxc_ctl/pgxc_ctl.conf: line 189: $coordExtraConfig: ambiguous redirectFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlPGXC monitor allNot running: gtm masterRunning: coordinator master coord1Not running: coordinator master coord2Running: datanode master datanode1Not running: datanode master datanode2PGXC stop coordinator master coord1Stopping coordinator master coord1.pg_ctl: directory "/home/postgres/pgxc/nodes/coord/coord1" does not existDone.PGXC stop datanode master datanode1Stopping datanode master datanode1.pg_ctl: PID file "/home/postgres/pgxc/nodes/datanode/datanode1/postmaster.pid" does not existIs server running?Done.PGXC monitor allNot running: gtm masterRunning: coordinator master coord1Not running: coordinator master coord2Running: datanode master datanode1Not running: datanode master datanode2PGXC monitor allNot running: gtm masterNot running: coordinator master coord1Not running: coordinator master coord2Not running: datanode master datanode1Not running: datanode master datanode2PGXC exit[postgres@gtm ~]$ pgxc_ctl -c /home/postgres/pgxc_ctl/pgxc_ctl.conf init all/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.conf/home/postgres/pgxc_ctl/pgxc_ctl.conf: line 189: $coordExtraConfig: ambiguous redirectFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlInitialize GTM masterERROR: target directory (/home/postgres/pgxc/nodes/gtm) exists and not empty. Skip GTM initilializationDone.Start GTM masterserver startingInitialize all the coordinator masters.Initialize coordinator master coord1.Initialize coordinator master coord2.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/coord/coord1 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/coord/coord2 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.Done.Starting coordinator master.Starting coordinator master coord1Starting coordinator master coord22019-05-30 21:13:03.998 EDT [25137] LOG: listening on IPv4 address "0.0.0.0", port 54322019-05-30 21:13:03.998 EDT [25137] LOG: listening on IPv6 address "::", port 54322019-05-30 21:13:04.000 EDT [25137] LOG: listening on Unix socket "/tmp/.s.PGSQL.5432"2019-05-30 21:13:04.038 EDT [25138] LOG: database system was shut down at 2019-05-30 21:13:00 EDT2019-05-30 21:13:04.042 EDT [25137] LOG: database system is ready to accept connections2019-05-30 21:13:04.049 EDT [25145] LOG: cluster monitor started2019-05-30 21:13:04.020 EDT [2730] LOG: listening on IPv4 address "0.0.0.0", port 54322019-05-30 21:13:04.020 EDT [2730] LOG: listening on IPv6 address "::", port 54322019-05-30 21:13:04.021 EDT [2730] LOG: listening on Unix socket "/tmp/.s.PGSQL.5432"2019-05-30 21:13:04.057 EDT [2731] LOG: database system was shut down at 2019-05-30 21:13:00 EDT2019-05-30 21:13:04.061 EDT [2730] LOG: database system is ready to accept connections2019-05-30 21:13:04.062 EDT [2738] LOG: cluster monitor startedDone.Initialize all the datanode masters.Initialize the datanode master datanode1.Initialize the datanode master datanode2.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/datanode/datanode1 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.The files belonging to this database system will be owned by user "postgres".This user must also own the server process.The database cluster will be initialized with locale "en_US.UTF-8".The default database encoding has accordingly been set to "UTF8".The default text search configuration will be set to "english".Data page checksums are disabled.fixing permissions on existing directory /home/postgres/pgxc/nodes/datanode/datanode2 ... okcreating subdirectories ... okselecting default max_connections ... 100selecting default shared_buffers ... 128MBselecting dynamic shared memory implementation ... posixcreating configuration files ... okrunning bootstrap script ... okperforming post-bootstrap initialization ... creating cluster information ... oksyncing data to disk ... okfreezing database template0 ... okfreezing database template1 ... okfreezing database postgres ... okWARNING: enabling "trust" authentication for local connectionsYou can change this by editing pg_hba.conf or using the option -A, or--auth-local and --auth-host, the next time you run initdb.Success.Done.Starting all the datanode masters.Starting datanode master datanode1.Starting datanode master datanode2.2019-05-30 21:13:12.077 EDT [25392] LOG: listening on IPv4 address "0.0.0.0", port 154322019-05-30 21:13:12.077 EDT [25392] LOG: listening on IPv6 address "::", port 154322019-05-30 21:13:12.079 EDT [25392] LOG: listening on Unix socket "/tmp/.s.PGSQL.15432"2019-05-30 21:13:12.114 EDT [25392] LOG: redirecting log output to logging collector process2019-05-30 21:13:12.114 EDT [25392] HINT: Future log output will appear in directory "pg_log".2019-05-30 21:13:12.079 EDT [2985] LOG: listening on IPv4 address "0.0.0.0", port 154322019-05-30 21:13:12.079 EDT [2985] LOG: listening on IPv6 address "::", port 154322019-05-30 21:13:12.081 EDT [2985] LOG: listening on Unix socket "/tmp/.s.PGSQL.15432"2019-05-30 21:13:12.117 EDT [2985] LOG: redirecting log output to logging collector process2019-05-30 21:13:12.117 EDT [2985] HINT: Future log output will appear in directory "pg_log".Done.psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"Done.psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"psql: FATAL: no pg_hba.conf entry for host "192.168.20.132", user "postgres", database "postgres"Done. 启动 pgxc_ctl -c /home/postgres/pgxc_ctl/pgxc_ctl.conf start all 关闭 pgxc_ctl -c /home/postgres/pgxc_ctl/pgxc_ctl.conf stop all 查看集群状态 [postgres@gtm ~]$ pgxc_ctl monitor all/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.conf/home/postgres/pgxc_ctl/pgxc_ctl.conf: line 189: $coordExtraConfig: ambiguous redirectFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlRunning: gtm masterRunning: coordinator master coord1Running: coordinator master coord2Running: datanode master datanode1Running: datanode master datanode2 配置集群信息 分别在数据节点、协调器节点上分别执行以下命令: 注:本节点只执行修改操作即可(alert node),其他节点执行创建命令(create node)。因为本节点已经包含本节点的信息。 create node coord1 with (type=coordinator,host=xl1, port=5432);create node coord2 with (type=coordinator,host=xl2, port=5432);alter node coord1 with (type=coordinator,host=xl1, port=5432);alter node coord2 with (type=coordinator,host=xl2, port=5432);create node datanode1 with (type=datanode, host=xl1,port=15432,primary=true,PREFERRED);create node datanode2 with (type=datanode, host=xl2,port=15432);alter node datanode1 with (type=datanode, host=xl1,port=15432,primary=true,PREFERRED);alter node datanode2 with (type=datanode, host=xl2,port=15432);select pgxc_pool_reload(); 分别登陆数据节点、协调器节点验证 postgres= select from pgxc_node;node_name | node_type | node_port | node_host | nodeis_primary | nodeis_preferred | node_id-----------+-----------+-----------+-----------+----------------+------------------+-------------coord1 | C | 5432 | xl1 | f | f | 1885696643coord2 | C | 5432 | xl2 | f | f | -1197102633datanode2 | D | 15432 | xl2 | f | f | -905831925datanode1 | D | 15432 | xl1 | t | f | 888802358(4 rows) 测试 插入数据 在数据节点1,执行相关操作。 通过协调器端口登录PG [postgres@xl1 ~]$ psql -p 5432psql (PGXL 10r1.1, based on PG 10.6 (Postgres-XL 10r1.1))Type "help" for help.postgres= create database lei;CREATE DATABASEpostgres= \c lei;You are now connected to database "lei" as user "postgres".lei= create table test1(id int,name text);CREATE TABLElei= insert into test1(id,name) select generate_series(1,8),'测试';INSERT 0 8lei= select from test1;id | name----+------1 | 测试2 | 测试5 | 测试6 | 测试8 | 测试3 | 测试4 | 测试7 | 测试(8 rows) 注:默认创建的表为分布式表,也就是每个数据节点值存储表的部分数据。关于表类型具体说明,下面有说明。 通过15432端口登录数据节点,查看数据 有5条数据 [postgres@xl1 ~]$ psql -p 15432psql (PGXL 10r1.1, based on PG 10.6 (Postgres-XL 10r1.1))Type "help" for help.postgres= \c lei;You are now connected to database "lei" as user "postgres".lei= select from test1;id | name----+------1 | 测试2 | 测试5 | 测试6 | 测试8 | 测试(5 rows) 登录到节点2,查看数据 有3条数据 [postgres@xl2 ~]$ psql -p15432psql (PGXL 10r1.1, based on PG 10.6 (Postgres-XL 10r1.1))Type "help" for help.postgres= \c lei;You are now connected to database "lei" as user "postgres".lei= select from test1;id | name----+------3 | 测试4 | 测试7 | 测试(3 rows) 两个节点的数据加起来整个8条,没有问题。 至此Postgre-XL集群搭建完成。 创建数据库、表时可能会出现以下错误: ERROR: Failed to get pooled connections 是因为pg_hba.conf配置不对,所有节点加上host all all 192.168.20.0/0 trust并重启集群即可。 ERROR: No Datanode defined in cluster 首先确认是否创建了数据节点,也就是create node相关的命令。如果创建了则执行select pgxc_pool_reload();使其生效即可。 集群管理与应用 表类型说明 REPLICATION表:各个datanode节点中,表的数据完全相同,也就是说,插入数据时,会分别在每个datanode节点插入相同数据。读数据时,只需要读任意一个datanode节点上的数据。 建表语法: CREATE TABLE repltab (col1 int, col2 int) DISTRIBUTE BY REPLICATION; DISTRIBUTE :会将插入的数据,按照拆分规则,分配到不同的datanode节点中存储,也就是sharding技术。每个datanode节点只保存了部分数据,通过coordinate节点可以查询完整的数据视图。 CREATE TABLE disttab(col1 int, col2 int, col3 text) DISTRIBUTE BY HASH(col1); 模拟数据插入 任意登录一个coordinate节点进行建表操作 [postgres@gtm ~]$ psql -h xl1 -p 5432 -U postgrespostgres= INSERT INTO disttab SELECT generate_series(1,100), generate_series(101, 200), 'foo';INSERT 0 100postgres= INSERT INTO repltab SELECT generate_series(1,100), generate_series(101, 200);INSERT 0 100 查看数据分布结果: DISTRIBUTE表分布结果 postgres= SELECT xc_node_id, count() FROM disttab GROUP BY xc_node_id;xc_node_id | count ------------+-------1148549230 | 42-927910690 | 58(2 rows) REPLICATION表分布结果 postgres= SELECT count() FROM repltab;count -------100(1 row) 查看另一个datanode2中repltab表结果 [postgres@datanode2 pgxl9.5]$ psql -p 15432psql (PGXL 10r1.1, based on PG 10.6 (Postgres-XL 10r1.1))Type "help" for help.postgres= SELECT count() FROM repltab;count -------100(1 row) 结论:REPLICATION表中,datanode1,datanode2中表是全部数据,一模一样。而DISTRIBUTE表,数据散落近乎平均分配到了datanode1,datanode2节点中。 新增数据节点与数据重分布 在线新增节点、并重新分布数据。 新增datanode节点 在gtm集群管理节点上执行pgxc_ctl命令 [postgres@gtm ~]$ pgxc_ctl/bin/bashInstalling pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Installing pgxc_ctl_bash script as /home/postgres/pgxc_ctl/pgxc_ctl_bash.Reading configuration using /home/postgres/pgxc_ctl/pgxc_ctl_bash --home /home/postgres/pgxc_ctl --configuration /home/postgres/pgxc_ctl/pgxc_ctl.confFinished reading configuration. PGXC_CTL START Current directory: /home/postgres/pgxc_ctlPGXC 在服务器xl3上,新增一个master角色的datanode节点,名称是datanode3 端口号暂定5430,pool master暂定6669 ,指定好数据目录位置,从两个节点升级到3个节点,之后要写3个none none应该是datanodeSpecificExtraConfig或者datanodeSpecificExtraPgHba配置PGXC add datanode master datanode3 xl3 15432 6671 /home/postgres/pgxc/nodes/datanode/datanode3 none none none 等待新增完成后,查询集群节点状态: postgres= select from pgxc_node;node_name | node_type | node_port | node_host | nodeis_primary | nodeis_preferred | node_id-----------+-----------+-----------+-----------+----------------+------------------+-------------datanode1 | D | 15432 | xl1 | t | f | 888802358datanode2 | D | 15432 | xl2 | f | f | -905831925datanode3 | D | 15432 | xl3 | f | f | -705831925coord1 | C | 5432 | xl1 | f | f | 1885696643coord2 | C | 5432 | xl2 | f | f | -1197102633(4 rows) 节点新增完毕 数据重新分布 由于新增节点后无法自动完成数据重新分布,需要手动操作。 DISTRIBUTE表分布在了node1,node2节点上,如下: postgres= SELECT xc_node_id, count() FROM disttab GROUP BY xc_node_id;xc_node_id | count ------------+-------1148549230 | 42-927910690 | 58(2 rows) 新增一个节点后,将sharding表数据重新分配到三个节点上,将repl表复制到新节点 重分布sharding表postgres= ALTER TABLE disttab ADD NODE (datanode3);ALTER TABLE 复制数据到新节点postgres= ALTER TABLE repltab ADD NODE (datanode3);ALTER TABLE 查看新的数据分布: postgres= SELECT xc_node_id, count() FROM disttab GROUP BY xc_node_id;xc_node_id | count ------------+--------700122826 | 36-927910690 | 321148549230 | 32(3 rows) 登录datanode3(新增的时候,放在了xl3服务器上,端口15432)节点查看数据: [postgres@gtm ~]$ psql -h xl3 -p 15432 -U postgrespsql (PGXL 10r1.1, based on PG 10.6 (Postgres-XL 10r1.1))Type "help" for help.postgres= select count() from repltab;count -------100(1 row) 很明显,通过 ALTER TABLE tt ADD NODE (dn)命令,可以将DISTRIBUTE表数据重新分布到新节点,重分布过程中会中断所有事务。可以将REPLICATION表数据复制到新节点。 从datanode节点中回收数据 postgres= ALTER TABLE disttab DELETE NODE (datanode3);ALTER TABLEpostgres= ALTER TABLE repltab DELETE NODE (datanode3);ALTER TABLE 删除数据节点 Postgresql-XL并没有检查将被删除的datanode节点是否有replicated/distributed表的数据,为了数据安全,在删除之前需要检查下被删除节点上的数据,有数据的话,要回收掉分配到其他节点,然后才能安全删除。删除数据节点分为四步骤: 1.查询要删除节点dn3的oid postgres= SELECT oid, FROM pgxc_node;oid | node_name | node_type | node_port | node_host | nodeis_primary | nodeis_preferred | node_id -------+-----------+-----------+-----------+-----------+----------------+------------------+-------------11819 | coord1 | C | 5432 | datanode1 | f | f | 188569664316384 | coord2 | C | 5432 | datanode2 | f | f | -119710263316385 | node1 | D | 5433 | datanode1 | f | t | 114854923016386 | node2 | D | 5433 | datanode2 | f | f | -92791069016397 | dn3 | D | 5430 | datanode1 | f | f | -700122826(5 rows) 2.查询dn3对应的oid中是否有数据 testdb= SELECT FROM pgxc_class WHERE nodeoids::integer[] @> ARRAY[16397];pcrelid | pclocatortype | pcattnum | pchashalgorithm | pchashbuckets | nodeoids ---------+---------------+----------+-----------------+---------------+-------------------16388 | H | 1 | 1 | 4096 | 16397 16385 1638616394 | R | 0 | 0 | 0 | 16397 16385 16386(2 rows) 3.有数据的先回收数据 postgres= ALTER TABLE disttab DELETE NODE (dn3);ALTER TABLEpostgres= ALTER TABLE repltab DELETE NODE (dn3);ALTER TABLEpostgres= SELECT FROM pgxc_class WHERE nodeoids::integer[] @> ARRAY[16397];pcrelid | pclocatortype | pcattnum | pchashalgorithm | pchashbuckets | nodeoids ---------+---------------+----------+-----------------+---------------+----------(0 rows) 4.安全删除dn3 PGXC$ remove datanode master dn3 clean 故障节点FAILOVER 1.查看当前集群状态 [postgres@gtm ~]$ psql -h xl1 -p 5432psql (PGXL 10r1.1, based on PG 10.6 (Postgres-XL 10r1.1))Type "help" for help.postgres= SELECT oid, FROM pgxc_node;oid | node_name | node_type | node_port | node_host | nodeis_primary | nodeis_preferred | node_id-------+-----------+-----------+-----------+-----------+----------------+------------------+-------------11739 | coord1 | C | 5432 | xl1 | f | f | 188569664316384 | coord2 | C | 5432 | xl2 | f | f | -119710263316387 | datanode2 | D | 15432 | xl2 | f | f | -90583192516388 | datanode1 | D | 15432 | xl1 | t | t | 888802358(4 rows) 2.模拟datanode1节点故障 直接关闭即可 PGXC stop -m immediate datanode master datanode1Stopping datanode master datanode1.Done. 3.测试查询 只要查询涉及到datanode1上的数据,那么该查询就会报错 postgres= SELECT xc_node_id, count() FROM disttab GROUP BY xc_node_id;WARNING: failed to receive file descriptors for connectionsERROR: Failed to get pooled connectionsHINT: This may happen because one or more nodes are currently unreachable, either because of node or network failure.Its also possible that the target node may have hit the connection limit or the pooler is configured with low connections.Please check if all nodes are running fine and also review max_connections and max_pool_size configuration parameterspostgres= SELECT xc_node_id, FROM disttab WHERE col1 = 3;xc_node_id | col1 | col2 | col3------------+------+------+-------905831925 | 3 | 103 | foo(1 row) 测试发现,查询范围如果涉及到故障的node1节点,会报错,而查询的数据范围不在node1上的话,仍然可以查询。 4.手动切换 要想切换,必须要提前配置slave节点。 PGXC$ failover datanode node1 切换完成后,查询集群 postgres= SELECT oid, FROM pgxc_node;oid | node_name | node_type | node_port | node_host | nodeis_primary | nodeis_preferred | node_id -------+-----------+-----------+-----------+-----------+----------------+------------------+-------------11819 | coord1 | C | 5432 | datanode1 | f | f | 188569664316384 | coord2 | C | 5432 | datanode2 | f | f | -119710263316386 | node2 | D | 15432 | datanode2 | f | f | -92791069016385 | node1 | D | 15433 | datanode2 | f | t | 1148549230(4 rows) 发现datanode1节点的ip和端口都已经替换为配置的slave了。 本篇文章为转载内容。原文链接:https://blog.csdn.net/qianglei6077/article/details/94379331。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-01-30 11:09:03
94
转载
转载文章
...QLS确定最佳的查询策略:建立查询计划和是否使用索引以及使用什么样的索引。 在创建索引时,SQLS会创建分布数据页来存放有关索引的两种统计信息:分布表和密度表。查询优化器使用这些统计信息估算使用该索引进行查询的成本(Cost),并在此基础上判断该索引对某个特定查询是否有用。 随着表中的数据发生变化,SQLS自动定期更新这些统计信息。采样是在各个数据页上随机进行。从磁盘读取一个数据页后,该数据页上的所有行都被用来更新统计信息。统计信息更新的频率取决于字段或索引中的数据量以及数据更改量。比如,对于有一万条记录的表,当1000个索引键值发生改变时,该表的统计信息便可能需要更新,因为1000 个值在该表中占了10%,这是一个很大的比例。而对于有1千万条记录的表来说,1000个索引值发生更改的意义则可以忽略不计,因此统计信息就不会自动更新。 至于它们帮助SQLS建立查询计划的具体过程,限于篇幅,这里就省略了,请有兴趣的朋友们自己研究。 顺便多说一句,SQLS除了能自动记录统计信息之外,还可以记录服务器中所发生的其它活动的详细信息,包括I/O 统计信息、CPU 统计信息、锁定请求、T-SQL 和 RPC 统计信息、索引和表扫描、警告和引发的错误、数据库对象的创建/除去、连接/断开、存储过程操作、游标操作等等。这些信息的读取、设置请朋友们在SQLS联机帮助文档(SQL Server Books Online)中搜索字符串“Profiler”查找。 五、索引的人工维护 上面讲到,某些不合适的索引将影响到SQLS的性能,随着应用系统的运行,数据不断地发生变化,当数据变化达到某一个程度时将会影响到索引的使用。这时需要用户自己来维护索引。 随着数据行的插入、删除和数据页的分裂,有些索引页可能只包含几页数据,另外应用在执行大量I/O的时候,重建非聚聚集索引可以维护I/O的效率。重建索引实质上是重新组织B树。需要重建索引的情况有: 1) 数据和使用模式大幅度变化; 2)排序的顺序发生改变; 3)要进行大量插入操作或已经完成; 4)使用I/O查询的磁盘读次数比预料的要多; 5)由于大量数据修改,使得数据页和索引页没有充分使用而导致空间的使用超出估算; 6)dbcc检查出索引有问题。 六、索引的使用原则 接近尾声的时候,让我们再从另一个角度认识索引的两个重要属性----唯一性索引和复合性索引。 在设计表的时候,可以对字段值进行某些限制,比如可以对字段进行主键约束或唯一性约束。 主键约束是指定某个或多个字段不允许重复,用于防止表中出现两条完全相同的记录,这样的字段称为主键,每张表都可以建立并且只能建立一个主键,构成主键的字段不允许空值。例如职员表中“身份证号”字段或成绩表中“学号、课程编号”字段组合。 而唯一性约束与主键约束类似,区别只在于构成唯一性约束的字段允许出现空值。 建立在主键约束和唯一性约束上的索引,由于其字段值具有唯一性,于是我们将这种索引叫做“唯一性索引”,如果这个唯一性索引是由两个以上字段的组合建立的,那么它又叫“复合性索引”。 注意,唯一索引不是聚集索引,如果对一个字段建立了唯一索引,你仅仅不能向这个字段输入重复的值。并不妨碍你可以对其它类型的字段也建立一个唯一性索引,它们可以是聚集的,也可以是非聚集的。 唯一性索引保证在索引列中的全部数据是唯一的,不会包含冗余数据。如果表中已经有一个主键约束或者唯一性约束,那么当创建表或者修改表时,SQLS自动创建一个唯一性索引。但出于必须保证唯一性,那么应该创建主键约束或者唯一性键约束,而不是创建一个唯一性索引。当创建唯一性索引时,应该认真考虑这些规则:当在表中创建主键约束或者唯一性键约束时, SQLS钭自动创建一个唯一性索引;如果表中已经包含有数据,那么当创建索引时,SQLS检查表中已有数据的冗余性,如果发现冗余值,那么SQLS就取消该语句的执行,并且返回一个错误消息,确保表中的每一行数据都有一个唯一值。 复合索引就是一个索引创建在两个列或者多个列上。在搜索时,当两个或者多个列作为一个关键值时,最好在这些列上创建复合索引。当创建复合索引时,应该考虑这些规则:最多可以把16个列合并成一个单独的复合索引,构成复合索引的列的总长度不能超过900字节,也就是说复合列的长度不能太长;在复合索引中,所有的列必须来自同一个表中,不能跨表建立复合列;在复合索引中,列的排列顺序是非常重要的,原则上,应该首先定义最唯一的列,例如在(COL1,COL2)上的索引与在(COL2,COL1)上的索引是不相同的,因为两个索引的列的顺序不同;为了使查询优化器使用复合索引,查询语句中的WHERE子句必须参考复合索引中第一个列;当表中有多个关键列时,复合索引是非常有用的;使用复合索引可以提高查询性能,减少在一个表中所创建的索引数量。 综上所述,我们总结了如下索引使用原则: 1)逻辑主键使用唯一的成组索引,对系统键(作为存储过程)采用唯一的非成组索引,对任何外键列采用非成组索引。考虑数据库的空间有多大,表如何进行访问,还有这些访问是否主要用作读写。 2)不要索引memo/note 字段,不要索引大型字段(有很多字符),这样作会让索引占用太多的存储空间。 3)不要索引常用的小型表 4)一般不要为小型数据表设置过多的索引,假如它们经常有插入和删除操作就更别这样作了,SQLS对这些插入和删除操作提供的索引维护可能比扫描表空间消耗更多的时间。 七、大结局 查询是一个物理过程,表面上是SQLS在东跑西跑,其实真正大部分压马路的工作是由磁盘输入输出系统(I/O)完成,全表扫描需要从磁盘上读表的每一个数据页,如果有索引指向数据值,则I/O读几次磁盘就可以了。但是,在随时发生的增、删、改操作中,索引的存在会大大增加工作量,因此,合理的索引设计是建立在对各种查询的分析和预测上的,只有正确地使索引与程序结合起来,才能产生最佳的优化方案。 一般来说建立索引的思路是: (1)主键时常作为where子句的条件,应在表的主键列上建立聚聚集索引,尤其当经常用它作为连接的时候。 (2)有大量重复值且经常有范围查询和排序、分组发生的列,或者非常频繁地被访问的列,可考虑建立聚聚集索引。 (3)经常同时存取多列,且每列都含有重复值可考虑建立复合索引来覆盖一个或一组查询,并把查询引用最频繁的列作为前导列,如果可能尽量使关键查询形成覆盖查询。 (4)如果知道索引键的所有值都是唯一的,那么确保把索引定义成唯一索引。 (5)在一个经常做插入操作的表上建索引时,使用fillfactor(填充因子)来减少页分裂,同时提高并发度降低死锁的发生。如果在只读表上建索引,则可以把fillfactor置为100。 (6)在选择索引字段时,尽量选择那些小数据类型的字段作为索引键,以使每个索引页能够容纳尽可能多的索引键和指针,通过这种方式,可使一个查询必须遍历的索引页面降到最小。此外,尽可能地使用整数为键值,因为它能够提供比任何数据类型都快的访问速度。 SQLS是一个很复杂的系统,让索引以及查询背后的东西真相大白,可以帮助我们更为深刻的了解我们的系统。一句话,索引就象盐,少则无味多则咸。 本篇文章为转载内容。原文链接:https://blog.csdn.net/qq_28052907/article/details/75194926。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-04-30 23:10:07
97
转载
转载文章
...用,实现企业资源优化配置,提升运营效率,增强竞争力的过程。在本文中,企业信息化是软件开发者就业的热门领域之一。 通信领域 , 涉及信息传递与交换技术的行业,包括但不限于电信服务、网络通信、无线通信、数据通信等子领域。在本文背景下,通信领域因门槛高、薪水高等特点吸引了大量软件开发者参与相关技术研发与项目实施,成为开发者主要从业方向之一。 系统集成 , 是指将不同功能、不同品牌或供应商的硬件设备、软件系统以及网络设备等按照一定的架构标准和规范,进行整合、协调和优化,形成一个统一、高效、稳定运行的信息系统解决方案的过程。在本文中,系统集成作为软件开发的重要组成部分,是部分开发者从事的工作内容之一。 高级程序员 , 在软件开发行业中,具备较深厚的专业技能、丰富的项目经验和较高技术水平的编程人员。他们不仅能够独立完成复杂模块的设计与编码工作,还能在项目中起到技术引领与指导作用,对项目的整体质量和进度有直接影响,通常其薪资待遇高于普通程序员。 技术总监(CTO) , Chief Technology Officer 的缩写,是企业中负责技术方向决策、技术团队管理、技术研发规划与实施的关键角色。技术总监需要具有深厚的技术背景、前瞻性的战略眼光以及出色的组织协调能力,确保企业的技术发展方向与业务需求保持一致,并通过技术创新推动企业发展。在本文中,技术总监的角色由于其综合能力和职责要求,在软件行业内占据重要地位,但人数相对较少。
2023-12-24 09:01:26
286
转载
转载文章
...司也需要有自己的开源策略,而拥有一个开源计划办公室则可以为其提供帮助。 “在使用开源软件方面,OPSO为公司提供了一个至关重要的能力中心”他说。 这与公司拥有安全运营中心的方式类似,他说。 “围绕一个项目而发展的开发人员社区,有助于代码库的形成,并吸引更多的开发人员参与,这可以变成一个良性循环。” ——Tom Hickman,ThreatX 首席创新官 “如果你对安全团队进行相应投资,你通常是不会期望你的软件是安全的,也无法及时应对安全事件。”他说。 “同样的逻辑也适用于 OSPO,这就是为什么你会看到许多领先的公司,例如Apple、Meta、Twitter、Goldman Sachs、Bloomberg 和 Google 都拥有 OSPO。他们走在了趋势的前面。” 而对组织内的开源活动的支持态度亦可成为软件供应商们的差异化原因与营销的机会。 根据Red Hat 2月分发布的一项调查,82%的IT领导者更倾向于选择为开源社区作出贡献的软件供应商。 受访者表示,当供应商支持开源社区时,就表示着他们更熟悉开源的流程并且在客户遇到技术难题时会更加有效。 但收益的不仅仅是软件供应商们。 根据 The New Stack、Linux Foundation Research 和 TODO Group 9 月份的调查,57% 拥有 OSPO 的组织将使用它们来进一步发展战略关系和建立合作伙伴关系。 十年前,Mark Hinkle 在 Citrix 工作时创办了一个开源计划办公室。他指出了在内部拥有一个 OSPO将如何使公司受益。 “对于我们来说,最大的工作是让不熟悉开源的员工学会并参与其中,成为优秀的社区成员”,他说,“我们还就如何确保我们的IP不会在没有正确理解的情况下进入项目的情况提供了指导,并确保我们没有与我们企业软件许可相冲突的开源项目合作。” 他说,OSPO还帮助Citrix确定了公司参与开源项目和Linux基金会等贸易组织的战略机会。 如今,他是云原生开源集成平台 TriggerMesh 的首席执行官兼联合创始人。 他说,参与开源系统对公司来说有着重大的经济效益。 “我们参与Knative是为了分享我们基础底层平台的开发,但作为业务的一部分,我们也拥有相关的增值服务。”他说,“通过共享该平台的研发,这为我们提供了更多的资源来改进我们自己的差异化技术。” Part4如何入门开源 在 The New Stack、Linux Foundation Research 和 TODO Group 的 9 月份调查中,有 63% 的公司表示,拥有OSPO 对其工程或产品团队的成功至关重要,高于上一年度该项研究数据的 54%。 其中77% 的人表示他们的开源程序对他们的软件实践产生了积极影响,例如提高了代码质量。 但公司也不可能总是为他们使用的每一个开源项目而花费精力。 “首先,节流一下”,VMware 的 Ambiel 建议道。 公司应该关注投入使用中最有意义的项目。而这也是OSPO可以帮助确定优先事项并确保技术与战略一致性的领域。 之后,开发人员应该自己去了解一下。项目通常提供相关在线文档,一般包含贡献着指南、治理文档和未解决问题列表。 “对于那些你较感兴趣的项目中,你可以介绍一下自己----打个招呼”,她说。“然后转到Slack频道或者分发列表,询问他们需要帮助的地方。也许他们不需要帮助,一切完好;又或者他们也有可能使用新人来审查核验代码。” Ambiel 说,开源计划办公室不仅可以帮助制定为开源社区做出贡献的商业案例,还可以帮助公司以安全、可靠和健全的方式来做这件事。 “如果我为一家公司工作,并想为开源做出贡献,我不想意外披露、泄露或破坏任何专利,”她说。“而OSPO可以帮助您做出明智的选择。” 她说,OSPO还可以在开源方面提供领导力和指导理念的支持。“它可以提供引领、指导、辅导和最佳实践的作用。” Aqua Security的开发人员倡导者Anaïs Urlichs则认为,支持开源的承诺必须从高层开始。 她说,“公司在多数时候往往不重视对开源的投资,所以员工自然而然不被鼓励对此作出贡献。” 在这些情况下,员工对于开源的热情也会在空闲时间里对开源的建设而消散殆尽,这对于开源的发展来说是不可持续的。 “如果公司对开源项目依赖度高,那么将开源贡献纳入工程师的日程安排是很重要的,”她说。“一些公司定义了员工可以为开源建设的时间百分比,将其作为他们正常工作日的一部分。” The New Stack 是 Insight Partners 的全资子公司,Insight Partners 是本文提到的以下公司的投资者:Sysdig、Aqua Security。 中英对照版 How an OSPO Can Help Your Engineers Give Back to Open Source OSPO (开源项目办公室)是如何使工程师回馈开源的 When it comes to open source software, there’s a big and growing problem: most organizations are takers, not givers. 谈到开源软件,有一个较大且日益严重的问题:大多数组织都是索取者,而不是给予者。 There’s a classic XKCD comic that shows a giant structure representing modern digital infrastructure, dependent on a tiny component created by “some random person in Nebraska” who has been “thanklessly maintaining since 2003.” 经典漫画XKCD展示了一个代表现代数字基础设施的巨大结构,它依赖于“内布拉斯加州的某位人士”创建的微小组件,该组件“自2003年来一直都处于吃力不讨好的状态”。 Randall Monroe’s XKCD comic illustrates the open source dilemma: overreliance on a small number of volunteer project maintainers. Randall Monroe 的XKCD漫画展示了目前开源面临的窘境:过度依赖少数项目维护志愿者的志愿服务。 This would have been funny, except that this is exactly what happened when security vulnerabilities were discovered in Log4j last December. (开源项目由志愿者自发来维护,)这听起来像是一件很滑稽的事情,但事实上去年十二月在Log4j中发现的安全漏洞也确实存在着上述情况。 The Java-based logging tool is ubiquitous in enterprise publications. In the last three months, for example, Log4j has been downloaded more than 30 million times, according to a report by the enterprise software company Sonatype. 然而这个基于Java的日志记录工具已经在企业内部刊物中无处不在。例如根据软件公司Sonatype的一份报告显示,在过去的三个月里,Log4j的下载量就已经超过3000万次。 The tool has 440,000 lines of code, according to Synopsys‘ Black Duck Open Hub research tool, with nearly 24,000 contributions by nearly 200 developers. That’s a large dev team compared to other open source projects. But looking closer at the numbers, more than 70% of commits were by just five people. 根据Synopsys(新思)公司旗下的Black Duck Open Hub 研究工具显示。Log4j有着440,000行代码,由近200名开发人员贡献了将近24,000行代码。其实与其他开源项目相比,这是一个庞大的开发团队。但是如果关注数据的话,就会发现超过70%的提交是仅仅靠五个人来完成的。 Log4j’s home page lists about a dozen members on its project team. Most projects have far fewer developers working on them — and that presents a problem for the organizations that depend on them. Log4j的主页上展示了十几位项目团队的成员。而大多项目的开发人员要比其原本需要的少得多----这是高度依赖开发人员团队所呈现出来的问题。 “There is little incentive for anyone today to contribute to an existing open source project,” said Jeremy Stretch, distinguished engineer at NS1, a DNS network company. “There’s usually no direct compensation, and few accolades are offered — most users don’t even know who maintains the software that they use.” “如今的人没有什么动力去为现有的开源项目做贡献”,来自DNS网络公司NS1的杰出工程师Jeremy Strech说,“因为通常来说,这没有直接的物质回报,也很少提供荣誉----大多数用户甚至不知道他们所用的软件是谁维护的。” The most common motivation among open source contributors is to add a feature that they themselves want to see, he said. “Once this has been achieved, the contributor rarely sticks around.” 他说,开源贡献者们最常见的动机就是添加他们自己想要的功能。“一旦实现了这一点,他们几乎都不会留下来。” Meanwhile, as a project becomes more popular, the burden on the core team of maintainers keeps increasing. 与此同时,随着项目的逐渐流行,对于维护方面的核心团队来说,他们的负担也在不断增加。 “More users means more feature requests and more bug reports — but not more maintainers,” Stretch said. “What was once an enjoyable hobby can quickly become a tedious chore, and many maintainers understandably opt to simply abandon their projects altogether.” “更多的用户意味有着更多的功能需求和错误报告----但不是更多的维护人员”,Stretch说。“曾经令人愉快的爱好很快就会变成一项乏味的项目,所以很多维护人员选择干脆完全放弃他们的项目,这也是可以理解的。” Part1The Tragedy of the Commons The open source software ecosystem is a perfect example of the “tragedy of the commons.” 开源软件的生态系统,就是“公地悲剧”的一个完美例子。 And the tragedy is — when everyone uses, but no one contributes, that resource — whether it’s an overrun park or an open source project — eventually collapses from overuse and underinvestment. Everyone loves using free stuff, but everyone expects someone else to take care of it. 这个悲剧就是---当一种资源,无论是一个超限的公园还是一个开源项目,所有人都在使用而没有人贡献之时,最终都会因为过度使用和投入不足而崩溃坍塌。 This approach can save you money in the short term, but it can become a fatal flaw over time. Especially since open source software is everywhere, running everything. 这种方式可以在短期内为你节省资金,但随着时间的推移,它可能会变成项目里致命的缺陷。 Linux, for example, the open source operating system, runs on 96% of the world’s top 1 million servers, and 90% of all cloud infrastructure is on Linux. Not to mention that 85% of all smartphones in the world run Linux, in the form of the Android OS. 拿Linux来说,这个开源操作系统在全球前100万台服务器中运行率在96%以上,且这些服务器90%的云基础设施也都在Linux上。更不用说世界上85%的智能手机都运行着Linux,即Android操作系统。 Then there’s Java, Apache, WordPress, Cassandra, Hadoop, MySQL, PHP, ElasticSearch, Kubernetes — the list of ubiquitous open source projects goes on and on. 还有Java, Apache, WordPress, Cassandra, Hadoop, MySQL, PHP, ElasticSearch, Kubernetes--这些常见开源项目的列表还在逐渐增加着。 Without open source, much of today’s technical infrastructure would immediately grind to a halt. 如果没有开源,今天的大部分技术基础设施的建设也将会戛然而止。 “It is a real problem,” said Danil Mikhailov, executive director at Data.org, a nonprofit backed by the Mastercard Center for Inclusive Growth and The Rockefeller Foundation that promotes the use of data science to tackle society’s greatest challenges. “这是一个很现实的问题”,Data.org的执行董事Danil Mikhailov说,该组织是由万事达包容性发展中心和洛克菲勒基金会支持,旨在促进使用数据科学来应对当今社会所面临的巨大挑战的非营利性组织。 While nearly all organizations use open source software, only a minority contribute to those projects. Forty-two percent of participants in a survey released in September by The New Stack, Linux Foundation Research, and the TODO Group said tthey contribute at least sometimes to open source projects. 虽然几乎所有组织都在使用着开源软件,但只有少数组织为这些项目作出了贡献。The New Stack、Linux Foundation Research 和 TODO Group 在 9 月发布的一项调查中,42% 的参与者表示,他们至少有时会为开源项目做出贡献。 The same study showed that only 36% of organizations train their engineers to contribute to open source. 而同一项研究表明,只有36%的组织会培训他们的工程师为开源作出贡献。 Individual companies should support projects that they use the most and are critical to their success, Mikhailov said: “If you use, you contribute.” 个体公司应该支持贡献这些他们使用最多且对他们成功至关重要的项目,Mikhailov认为:“如果你使用开源,你就应该为他做出属于你自己的贡献。” Part2OSPO Benefits:Less Tech Debt,Better Recruiting Participating in open source communities — especially when guided by an in-house open source program office (OSPO) — can help ensure the health of projects critical to your organization’s success, improve those projects’ security, and allow your engineers to have more impact in the projects’ development road map. 参与开源社区——特别是在内部开源项目办公室(OSPO)的指导下——不仅可以保证对组织成功至关重要项目的健康发展,还可以提高项目安全性,同时可以允许工程师在项目发展规划中起到更大的影响。 Say, for example, a company uses an open source tool and modifies it a little to make it better. If that improvement isn’t contributed back to the community, then the official version of the open source project will start to diverge from what the company is using 例如,如果一家公司使用了开源工具,并对其进行了一些调整使其变得更好。但如果这项改进没有反馈到开源社区,那么开源项目的正式版本就会一开始与该公司所使用的版本有所不同。 “You start to grow technical debt because when the original source changes and you’ve got a different version. Those differences grow rapidly, compounding daily. It doesn’t take long for you to be the proud user and maintainer of a one-of-a-kind open source project variant,” said Suzanne Ambiel, director, open source marketing and strategy at VMware. “当原始代码来源发生变化且你所使用的是不同的版本时,你的技术负债将越来越多。而这些差异是以天为单位迅速增长的。”VMware 开源营销和战略总监 Suzanne Ambiel 表示,“所以你很快就会变成一个开源项目里独一无二变体的‘自豪’用户和维护人员。” “The technical debt gets bigger and bigger and it gets very expensive for a company to manage.” “如果技术负债越来越多,那么公司的管理成本则会非常昂贵”。 Support for open source activity can also be a recruiting tool. “It’s really a talent magnet,” said Ambiel. “It’s one of the things that new hires look for.” 实际上对于开源活动的支持也变成了一种招聘途径。“这真是一块吸引人才的磁铁,”Ambiel说,“这也是新员工所寻求的“。 Some engineering managers might worry that open source contributions will detract from core product development, she said. Their rationale, she added, might run along the lines of, “I only have so much talent, and so many hours, and I need them to only work on things where I can measure and see the return on investment.” 她还提到,一些工程经理可能会对贡献开源而减损核心产品的开发的精力而感到担忧。她补充到,他们的理由有可能是这样的:“我只有有限的才华与时间,且我需要这些只做我认为可以度量且看到投资回报的事情。” But that attitude, she said, is shortsighted. Supporting employees who contribute to open source communities can build skills and develop talent, she said. 但她说,这是一种鼠目寸光的态度。支持开源社区并且作出贡献的员工,可以从中培养技能与增长才华。 Loris Degionni, chief technology officer and founder at Sysdig, a cloud security vendor, echoed this notion: “Finding employees who contribute to open source is a gold mine,” said. 云安全供应商 Sysdig 的首席技术官兼创始人 Loris Degionni 也赞同这一观点:“找出为开源做出贡献的员工无疑就找到一座金矿,”他说。 These employees are more capable of delivering features a company wants to use and merge them into community-supported standards, he said. And in a war for talent, companies that embrace open source are more attractive to developers. 他认为,这些参与开源的员工更具备公司想拥有的竞争力并将一些功能融入至社区所支持的标准中。且在人才争夺战中,拥抱开源的公司也更受到开发人员的青睐。 “Lastly, open source is driven by a community of technical experts you may not be able to hire,” he said. “When employees actively contribute and collaborate with these experts, they’ll be better informed of best practices and bring them back to your organization. “最后,开源项目是由你可能无法聘请的技术专家社区推动的”,他说,“当员工积极参与并于这些专家合作时,他们将能更好地深入这些最佳实践,并将这些收获带回到你的组织之中。” “You start to grow technical debt because when the original source changes and you’ve got a different version … It doesn’t take long for you to be the proud user and maintainer of a one-of-a-kind open source project variant.” —Suzanne Ambiel, director, open source marketing and strategy, VMware “当原始数据来源发生变化且你所使用的是不同的版本时,你的技术负债将越来越多...所以你很快就会变成一个开源项目里独一无二变体的”自豪“用户和维护人员。” — Suzanne Ambiel,VMware 开源营销和战略总监 “All of this should be rewarded — developers shouldn’t have to spend their free time honing their skills, as your company will quickly see benefits from their efforts.” “但是这一切终究不会白费--开发人员不应该把业余时间用在磨练他们的技能上,因为你的公司很快就会在他们的努力中看到好处。” An OSPO, Degionni suggested, can help achieve these goals, as well as help prioritize contributions and ensure collaboration. In addition, they can help provide governance that mirrors what companies would have for internally developed applications. Degionni认为,OSPO(开源计划办公室)可以帮助公司实现这些目标,以及帮助确定贡献的优先级并确保合作的进行。除此之外,他们也可以对公司内部开发应用程序方面的治理提供相关帮助。 “Members of the open source team are also in a position to be great internal evangelists for open source technologies, and act as bridges between the organization and the broader community,” he added. “开源团队的成员也可以成为开源技术的伟大内部布道师,并充当组织与更广泛社区之间的桥梁。”他补充道。 In the September survey from The New Stack, Linux Foundation Research and the TODO Group, nearly 53% of organizations with OSPOs said they saw more innovation as a result of having an OSPO, while almost 43% said they saw increased participation in external open source projects. 在 The New Stack、Linux Foundation Research 和 TODO Group 的 9 月调查中,近 53% 的拥有 OSPO的组织表示,由于拥有了OSPO,他们看到了更多创新,而近 43% 的组织表示,他们在外部开源项目的参与度上有所增加。 Part3More OSPO Benefits:A Business Edge Contributing to open source communities doesn’t just help the communities, but the companies that contribute to them, said Tom Hickman, chief innovation officer at ThreatX, a cybersecurity firm. 网络安全公司 ThreatX 的首席创新官 Tom Hickman 表示,为开源社区做出贡献,不仅有助于社区,还有助于为社区做出贡献的公司。 “Growing the community of developers around a project helps the code base, and attracts more developers,” he said. “It can become a virtuous circle.” “围绕一个项目而发展的开发人员社区,有助于代码库的形成,并吸引更多的开发人员参与”,他说,“这可以变成一个良性循环。” Also, companies that contribute to open source projects get twice the productive value from their use of open source than companies that don’t, according to research by Harvard Business School. 此外,根据哈佛商学院的研究,为开源项目作出贡献的公司从使用开源的项目中获得的生产价值,是不参与开源项目公司的两倍。 Many of the biggest companies in the world are contributing to open source, said Chris Aniszczyk, chief technology officer at Cloud Native Computing Foundation. He pointed to the Open Source Contributor Index as a reference for exactly just how much companies are doing. Cloud Native Computing Foundation 的首席技术官 Chris Aniszczyk 说,世界上许多巨头公司都为开源作出了贡献。他还提到,开源贡献者的指数是作为公司是否有所作为的参考。 The tech giants dominate the list: Google, Microsoft, Red Hat, Intel, IBM, Amazon, Facebook, VMware, GitHub and SAP are the top 10 contributors, in that order. But there are also a lot of end users on the top 100 list, said Aniszczyk, including Uber, the BBC, Orange, Netflix, and Square. 科技巨头占据了这份榜单的主导地位:谷歌、微软、红帽、英特尔、IBM、亚马逊、Facebook、VMware、GitHub 和 SAP 依次是排名前 10 的贡献者。但Aniszczyk 表示,但也有很多终端用户公司进入前 100 名,包括 Uber、BBC、Orange、Netflix 和 Square。 “We’ve always known working in upstream projects is not just the right thing to do —it’s the best approach to open source software development and the best way to deliver open source benefits to our customers,” he said. “It’s great to see that IT leaders recognize this as well.” “我们一直知道,在上游项目中工作不仅仅是关正确与否----它是开源软件开发的最佳方法,也是向客户提供开源福利的最佳方式“他说,“很高兴看到IT领导者们也认识到了这一点。” To contribute alongside these giants, companies need to have their own open source strategies, and having an open source program office can help. 为了和这些公司一起作出贡献,公司也需要有自己的开源策略,而拥有一个开源项目办公室则可以为其提供帮助。 “OSPOs provide a critical center of competency in a company when it comes to utilizing open source software,” he said. “在使用开源软件方面,OPSO为公司提供了一个至关重要的能力中心”他说。 It’s similar to the way that companies have security operations centers, he said. 这与公司拥有安全运营中心的方式类似,他说。 “Growing the community of developers around a project helps the code base, and attracts more developers. It can become a virtuous circle.” —Tom Hickman, chief innovation officer, ThreatX “围绕一个项目而发展的开发人员社区,有助于代码库的形成,并吸引更多的开发人员参与,这可以变成一个良性循环。” ——Tom Hickman,ThreatX 首席创新官 “If you don’t make the investment in a security team, you generally don’t expect your software to be secure or be able to respond to security incidents in a timely fashion,” he said. “如果你没有对安全团队进行相应投资,你通常是不会期望你的软件是安全的,也无法及时响应安全事件。”他说。 “The same logic applies to OSPOs and is why you see many leading companies out there such as Apple, Meta, Twitter, Goldman Sachs, Bloomberg, and Google all have OSPOs. They are ahead of the curve.” “同样的逻辑也适用于 OSPO,这就是为什么你会看到许多领先的公司,例如 Apple、Meta、Twitter、Goldman Sachs、Bloomberg 和 Google 都拥有 OSPO。他们走在了趋势的前面。” Support for open source activity within your organization can become a differentiator and marketing opportunity for software vendors. 而对组织内的开源活动的支持态度亦可成为软件供应商们的差异化原因与营销的机会。 According to a Red Hat survey released in February, 82% of IT leaders are more likely to select a vendor who contributes to the open source community. 根据Red Hat2月分发布的一项调查,82%的IT领导者更倾向于选择为开源社区作出贡献的软件供应商。 Respondents said that when vendors support open source communities they are more familiar with open source processes and are more effective if customers have technical challenges. 受访者表示,当供应商支持开源社区时,就表示着他们更熟悉开源的流程并且在客户遇到技术难题时会更加有效。 But it’s not just software vendors who benefit. 但收益的不仅仅是软件供应商们。 According to September’s survey by The New Stack, Linux Foundation Research, and the TODO Group, 57% of organizations with OSPOs use them to further strategic relationships and build partnerships. 根据 The New Stack、Linux Foundation Research 和 TODO Group 9 月份的调查,57% 拥有 OSPO 的组织将使用它们来进一步发展战略关系和建立合作伙伴关系。 Mark Hinkle started an open source program office back when he worked at Citrix a decade ago. He pointed out how having an OSPO in-house benefited the company. 十年前,Mark Hinkle 在 Citrix 工作时创办了一个开源计划办公室。他指出了在内部拥有一个 OSPO将如何使公司受益。 “For us the biggest job was to educate our employees who weren’t familiar with open source to get involved and be good community members,” he said. “We also provided guidance on how to make sure our IP didn’t enter projects without proper understanding and we made sure we didn’t incorporate open source that conflicted with our enterprise software licensing.” “对于我们来说,最大的工作是让不熟悉开源的员工学会并参与其中,成为优秀的社区成员”,他说,“我们还就如何确保我们的IP不会在没有正确理解的情况下进入项目的情况提供了指导,并确保我们没有与我们企业软件许可相冲突的开源项目合作。” The OSPO also helped Citrix identify strategic opportunities for the company to participate in open source projects and trade organizations like The Linux Foundation, he said. 他说,OSPO还帮助Citrix确定了公司参与开源项目和Linux基金会等贸易组织的战略机会。 Today, he’s the CEO and co-founder of TriggerMesh, a cloud native, open source integration platform. 如今,他是云原生开源集成平台 TriggerMesh 的首席执行官兼联合创始人。 There are some significant economic benefits to participating in the open source ecosystem, he said. 他说,参与开源系统对公司来说有着重大的经济效益。 “We participate in Knative to share the development of our underlying platform but we develop value-added services as part of our business,” he said. “By sharing the R and D for the platform, it gives us more resources to develop our own differentiated technology.” “我们参与Knative是为了分享我们基础底层平台的开发,但作为业务的一部分,我们也拥有相关的增值服务。”他说,“通过共享该平台的研发,这为我们提供了更多的资源来改进我们自己的差异化技术。” Part4How to Get Started in Open Source Sixty-three percent of companies in the September survey from The New Stack, Linux Foundation Research and the TODO Group said that having an OSPO was very or extremely critical to the success of their engineering or product teams, up from 54% in the previous annual study. 在 The New Stack、Linux Foundation Research 和 TODO Group 的 9 月份调查中,有 63% 的公司表示,拥有OSPO 对其工程或产品团队的成功至关重要,高于上一年度该项研究数据的 54%。 In particular, 77% said that their open source program had a positive impact on their software practices, such as improved code quality. 其中77% 的人表示他们的开源程序对他们的软件实践产生了积极影响,例如提高了代码质量。 But companies can’t always contribute to every single open source project that they use. 但公司也不可能总是为他们使用的每一个开源项目而花费精力。 “First, thin the herd a little bit,” advised VMware’s Ambiel. “首先,节流一下”,VMware 的 Ambiel 建议道。 Companies should look at the projects that make the most sense for their use cases. This is an area where an OSPO can help set priorities and ensure technical and strategic alignment. 公司应该关注投入使用中最有意义的项目。而这也是OSPO可以帮助确定优先事项并确保技术与战略一致性的领域。 Then, developers should go and check out the projects themselves. Projects typically offer online documentation, often with contributor guides, governance documents, and lists of open issues. 之后,开发人员应该自己去了解一下。项目通常提供相关在线文档,一般包含贡献着指南、治理文档和未解决问题列表。 “For the projects that rise to the top of your strategic list, introduce yourself — say hello,” she said. “Go to the Slack channel or the distribution list and ask where they need help. Maybe they don’t need help and everything is good. Or maybe they can use a new person to review code.” “对于那些上升到你的战略清单顶端的项目,你可以介绍一下自己----打个招呼”,她说。“然后转到Slack频道或者分发列表,询问他们需要帮助的地方。也许他们不需要帮助,一切完好;又或者他们也有可能使用新人来审查核验代码。” An open source program office can not only help make a business case for contributing to the open source community, Ambiel said, but can help companies do it in a way that’s safe, secure and sound. Ambiel 说,开源项目办公室不仅可以帮助制定为开源社区做出贡献的商业案例,还可以帮助公司以安全、可靠和健全的方式来做这件事。 “If I work for a company and want to contribute to open source, I don’t want to accidentally disclose, divulge or undermine any patents,” she said. “An OSPO helps you make smart choices.” “如果我为一家公司工作,并想为开源做出贡献,我不想意外披露、泄露或破坏任何专利,”她说。“而OSPO可以帮助您做出明智的选择。” An OSPO can also help provide leadership and the guiding philosophy about supporting open source, she said. “It can provide guidance, mentorship, coaching and best practices.” 她说,OSPO还可以在开源方面提供领导力和指导理念的支持。“它可以提供引领、指导、辅导和最佳实践的作用。” Commitment to support open source has to start at the top, said Anaïs Urlichs, developer advocate at Aqua Security. Aqua Security的开发人员倡导者Anaïs Urlichs则认为,支持开源的承诺必须从高层开始。 “Too often,” she said, “companies do not value investment into open source, so employees are not encouraged to contribute to it.” 她说,“公司在多数时候往往不重视对开源的投资,所以员工自然而然不被鼓励对此作出贡献。” In those cases, employees with a passion for open source end up contributing during their free time, which is not sustainable. 在这些情况下,员工对于开源的热情也会在空闲时间里对开源的建设而消散殆尽,这对于开源的发展来说是不可持续的。 “If companies rely on open source projects, it is important to make open source contributions part of an engineer’s work schedule,” she said. “Some companies define a time percentage that employees can contribute to open source as part of their normal workday.” “如果公司对开源项目依赖度高,那么将开源贡献纳入工程师的日程安排是很重要的,”她说。“一些公司定义了员工可以为开源建设的时间百分比,将其作为他们正常工作日的一部分。” The New Stack is a wholly owned subsidiary of Insight Partners, an investor in the following companies mentioned in this article: Sysdig, Aqua Security. The New Stack 是 Insight Partners 的全资子公司,Insight Partners 是本文提到的以下公司的投资者:Sysdig、Aqua Security。 相关阅读 | Related Reading 《开源合规指南(企业篇)》正式发布,为推动我国开源合规建设提供参考 “目标->用户->指标”——企业开源运营之道|瞰道@谭中意 开源之夏邀请函——仅限高校学子开启 开源社简介 开源社成立于 2014 年,是由志愿贡献于开源事业的个人成员,依 “贡献、共识、共治” 原则所组成,始终维持厂商中立、公益、非营利的特点,是最早以 “开源治理、国际接轨、社区发展、开源项目” 为使命的开源社区联合体。开源社积极与支持开源的社区、企业以及政府相关单位紧密合作,以 “立足中国、贡献全球” 为愿景,旨在共创健康可持续发展的开源生态,推动中国开源社区成为全球开源体系的积极参与及贡献者。 2017 年,开源社转型为完全由个人成员组成,参照 ASF 等国际顶级开源基金会的治理模式运作。近八年来,链接了数万名开源人,集聚了上千名社区成员及志愿者、海内外数百位讲师,合作了近百家赞助、媒体、社区伙伴。 本篇文章为转载内容。原文链接:https://blog.csdn.net/kaiyuanshe/article/details/124976824。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-05-03 09:19:23
273
转载
转载文章
...了线程,将会执行拒绝策略 上面的细节涉及的所有步骤内容,均由创建线程池的参数执行 下面是ThreadPoolExecutor构造方法参数的源码注释 / 用给定的初始值,创建一个新的线程池 @param corePoolSize 核心线程数量 @param maximumPoolSize 最大线程数量 @param keepAliveTime 当线程数大于核心线程数量时,空闲的线程可生存的时间 @param unit 时间单位 @param workQueue 任务队列,只能包含由execute提交的Runnable任务 @param threadFactory 工厂,用于创建线程给线程池调度的工厂,可以自定义 @param handler 拒绝策略(可以自定义,JDK默认提供4种),当线程边界和队列容量已经满了,新来线程被阻塞时使用的处理程序/public ThreadPoolExecutor(int corePoolSize,int maximumPoolSize,long keepAliveTime,TimeUnit unit,BlockingQueue<Runnable> workQueue,ThreadFactory threadFactory,RejectedExecutionHandler handler) JDK提供的4种拒绝策略,不常用,一般都是自己定义拒绝策略 Abort:抛异常 Discard:扔掉,不抛异常 DiscardOldest:扔掉排队时间最久的(将队列中排队时间最久的扔掉,然后让新来的进来) CallerRuns:调用者处理任务(谁通过execute方法提交任务,谁处理) ThreadPoolExecutor继承关系 继承关系:ThreadPoolExecutor->AbstractExectorService类->ExectorService接口->Exector接口 Executors(注意这后面有s) 它可以说是线程池工厂类,我们一般通过它创建线程池,并且它为我们封装了线程 看看下面创建线程池,哪里用到了它 使用实例 import java.io.IOException;import java.util.concurrent.;public class T05_00_HelloThreadPool {static class Task implements Runnable {private int i;public Task(int i) {this.i = i;}@Overridepublic void run() {System.out.println(Thread.currentThread().getName() + " Task " + i);try {System.in.read();} catch (IOException e) {e.printStackTrace();} }@Overridepublic String toString() {return "Task{" +"i=" + i +'}';} }public static void main(String[] args) {ThreadPoolExecutor tpe = new ThreadPoolExecutor(2, 4,60, TimeUnit.SECONDS,new ArrayBlockingQueue<Runnable>(4),Executors.defaultThreadFactory(),new ThreadPoolExecutor.CallerRunsPolicy());//创建线程池,核心2个,最大4个,空闲线程存活时间60s,任务队列容量4,使用默认线程工程,创建线程。拒绝策略是JDK提供的for (int i = 0; i < 8; i++) {tpe.execute(new Task(i));//供提交8次任务}System.out.println(tpe.getQueue());//查看任务队列tpe.execute(new Task(100));//提交新的任务System.out.println(tpe.getQueue());tpe.shutdown();//关闭线程池} } 5、TPE型线程池2:SingleThreadPool 单例线程池(只有一个线程) 为什么有单例线程池 有任务队列,有线程池管理机制 Executors(注意这后面有s) 它可以说是线程池工厂类,我们一般通过它创建线程池,并且它为我们封装了线程 看看下面哪里用到了它 /创建单例线程池,扔5个任务进去,查看输出结果,看看有几个线程执行任务/import java.util.concurrent.ExecutorService;import java.util.concurrent.Executors;public class T07_SingleThreadPool {public static void main(String[] args) {ExecutorService service = Executors.newSingleThreadExecutor();for(int i=0; i<5; i++) {final int j = i;service.execute(()->{System.out.println(j + " " + Thread.currentThread().getName());});} }} 6、TPE型线程池3:CachedPool 缓存,存储线程池 此线程池没有核心线程,来一个任务启动一个线程(最多Integer.MaxValue,不会放在任务队列,因为任务队列容量为0),每个线程空闲后,只能活60s 实例 import java.util.concurrent.ExecutorService;import java.util.concurrent.Executors;public class T07_SingleThreadPool {public static void main(String[] args) {ExecutorService service = Executors.newSingleThreadExecutor();//通过Executors获取池子for(int i=0; i<5; i++) {final int j = i;service.execute(()->{//提交任务System.out.println(j + " " + Thread.currentThread().getName());});}service.shutdown();} } 7、TPE型线程池4:FixedThreadPool 固定线程池 此线次池,用于创建一个固定线程数量的线程池,不会回收 实例 import java.util.ArrayList;import java.util.List;import java.util.concurrent.Callable;import java.util.concurrent.ExecutionException;import java.util.concurrent.ExecutorService;import java.util.concurrent.Executors;import java.util.concurrent.Future;public class T09_FixedThreadPool {public static void main(String[] args) throws InterruptedException, ExecutionException {//并发执行long start = System.currentTimeMillis();getPrime(1, 200000); long end = System.currentTimeMillis();System.out.println(end - start);//输出并发执行耗费时间final int cpuCoreNum = 4;//并行执行ExecutorService service = Executors.newFixedThreadPool(cpuCoreNum);MyTask t1 = new MyTask(1, 80000); //1-5 5-10 10-15 15-20MyTask t2 = new MyTask(80001, 130000);MyTask t3 = new MyTask(130001, 170000);MyTask t4 = new MyTask(170001, 200000);Future<List<Integer>> f1 = service.submit(t1);Future<List<Integer>> f2 = service.submit(t2);Future<List<Integer>> f3 = service.submit(t3);Future<List<Integer>> f4 = service.submit(t4);start = System.currentTimeMillis();f1.get();f2.get();f3.get();f4.get();end = System.currentTimeMillis();System.out.println(end - start);//输出并行耗费时间}static class MyTask implements Callable<List<Integer>> {int startPos, endPos;MyTask(int s, int e) {this.startPos = s;this.endPos = e;}@Overridepublic List<Integer> call() throws Exception {List<Integer> r = getPrime(startPos, endPos);return r;} }static boolean isPrime(int num) {for(int i=2; i<=num/2; i++) {if(num % i == 0) return false;}return true;}static List<Integer> getPrime(int start, int end) {List<Integer> results = new ArrayList<>();for(int i=start; i<=end; i++) {if(isPrime(i)) results.add(i);}return results;} } 8、TPE型线程池5:ScheduledPool 预定,延时线程池 根据延时时间(隔多长时间后运行),排序,哪个线程先执行,用户只需要指定核心线程数量 此线程池返回的池对象,和提交任务方法都不一样,比较涉及到时间 import java.util.Random;import java.util.concurrent.Executors;import java.util.concurrent.ScheduledExecutorService;import java.util.concurrent.TimeUnit;public class T10_ScheduledPool {public static void main(String[] args) {ScheduledExecutorService service = Executors.newScheduledThreadPool(4);service.scheduleAtFixedRate(()->{//提交延时任务try {TimeUnit.MILLISECONDS.sleep(new Random().nextInt(1000));} catch (InterruptedException e) {e.printStackTrace();}System.out.println(Thread.currentThread().getName());}, 0, 500, TimeUnit.MILLISECONDS);//指定延时时间和单位,第一个任务延时0毫秒,之后的任务,延时500毫秒} } 9、手写拒绝策略小例子 import java.util.concurrent.;public class T14_MyRejectedHandler {public static void main(String[] args) {ExecutorService service = new ThreadPoolExecutor(4, 4,0, TimeUnit.SECONDS, new ArrayBlockingQueue<>(6),Executors.defaultThreadFactory(),new MyHandler());//将手写拒绝策略传入}static class MyHandler implements RejectedExecutionHandler {//1、继承RejectedExecutionHandler@Overridepublic void rejectedExecution(Runnable r, ThreadPoolExecutor executor) {//2、重写方法//log("r rejected")//伪代码,表示通过log4j.log()报一下日志,拒绝的时间,线程名//save r kafka mysql redis//可以尝试保存队列//try 3 times //可以尝试几次,比如3次,重新去抢队列,3次还不行就丢弃if(executor.getQueue().size() < 10000) {//尝试条件,如果size>10000了,就执行拒绝策略//try put again();//如果小于10000,尝试将其放到队列中} }} } 10、ForkJoinPool线程池1:ForkJoinPool 前面我们讲过线程分为两大类,TPE和FJP ForkJoinPool(分解汇总任务(将任务细化,最后汇总结果),少量线程执行多个任务(子任务,TPE做不到先执行子任务),CPU密集型) 适合将大任务切分成多个小任务运行 两个方法,fork():分子任务,将子任务分配到线程池中 join():当前任务的计算结果,如果有子任务,等子任务结果返回后再汇总 下面实例实现,一百万个随机数求和,由两种方法实现,一种ForkJoinPool分任务并行,一种使用单线程做 import java.io.IOException;import java.util.Arrays;import java.util.Random;import java.util.concurrent.ForkJoinPool;import java.util.concurrent.RecursiveAction;import java.util.concurrent.RecursiveTask;public class T12_ForkJoinPool {//1000000个随机数求和static int[] nums = new int[1000000];//一堆数static final int MAX_NUM = 50000;//分任务时,每个任务的操作量不能多于50000个,否则就继续细分static Random r = new Random();//使用随机数将数组初始化static {for(int i=0; i<nums.length; i++) {nums[i] = r.nextInt(100);}System.out.println("---" + Arrays.stream(nums).sum()); //stream api 单线程就这么做,一个一个加}//分任务,需要继承,可以继承RecursiveAction(不需要返回值,一般用在不需要返回值的场景)或//RecursiveTask(需要返回值,我们用这个,因为我们需要最后获取求和结果)两个更好实现的类,//他俩继承与ForkJoinTaskstatic class AddTaskRet extends RecursiveTask<Long> {private static final long serialVersionUID = 1L;int start, end;AddTaskRet(int s, int e) {start = s;end = e;}@Overrideprotected Long compute() {if(end-start <= MAX_NUM) {//如果任务操作数小于规定的最大操作数,就进行运算,long sum = 0L;for(int i=start; i<end; i++) sum += nums[i];return sum;//返回结果} //如果分配的操作数大于规定,就继续细分(简单的重中点分,两半)int middle = start + (end-start)/2;//获取中间值AddTaskRet subTask1 = new AddTaskRet(start, middle);//传入起始值和中间值,表示一个子任务AddTaskRet subTask2 = new AddTaskRet(middle, end);//中间值和结尾值,表示一个子任务subTask1.fork();//分任务subTask2.fork();//分任务return subTask1.join() + subTask2.join();//最后返回结果汇总} }public static void main(String[] args) throws IOException {/ForkJoinPool fjp = new ForkJoinPool();AddTask task = new AddTask(0, nums.length);fjp.execute(task);/ForkJoinPool fjp = new ForkJoinPool();//创建线程池AddTaskRet task = new AddTaskRet(0, nums.length);//创建任务fjp.execute(task);//传入任务long result = task.join();//返回汇总结果System.out.println(result);//System.in.read();} } 11、ForkJoinPool线程池2:WorkStealingPool 任务偷取线程池 原来的线程池,都是有一个任务队列,而这个不同,它给每个线程都分配了一个任务队列 当某一个线程的任务队列没有任务,并且自己空闲,它就去其它线程的任务队列中偷任务,所以叫任务偷取线程池 细节:当线程自己从自己的任务队列拿任务时,不需要加锁,但是偷任务时,因为有两个线程,可能发生同步问题,需要加锁 此线程继承FJP 实例 import java.io.IOException;import java.util.concurrent.ExecutorService;import java.util.concurrent.Executors;import java.util.concurrent.TimeUnit;public class T11_WorkStealingPool {public static void main(String[] args) throws IOException {ExecutorService service = Executors.newWorkStealingPool();System.out.println(Runtime.getRuntime().availableProcessors());service.execute(new R(1000));service.execute(new R(2000));service.execute(new R(2000));service.execute(new R(2000)); //daemonservice.execute(new R(2000));//由于产生的是精灵线程(守护线程、后台线程),主线程不阻塞的话,看不到输出System.in.read(); }static class R implements Runnable {int time;R(int t) {this.time = t;}@Overridepublic void run() {try {TimeUnit.MILLISECONDS.sleep(time);} catch (InterruptedException e) {e.printStackTrace();}System.out.println(time + " " + Thread.currentThread().getName());} }} 12、流式API:ParallelStreamAPI 不懂的请参考:https://blog.csdn.net/grd_java/article/details/110265219 实例 import java.util.ArrayList;import java.util.List;import java.util.Random;public class T13_ParallelStreamAPI {public static void main(String[] args) {List<Integer> nums = new ArrayList<>();Random r = new Random();for(int i=0; i<10000; i++) nums.add(1000000 + r.nextInt(1000000));//System.out.println(nums);long start = System.currentTimeMillis();nums.forEach(v->isPrime(v));long end = System.currentTimeMillis();System.out.println(end - start);//使用parallel stream apistart = System.currentTimeMillis();nums.parallelStream().forEach(T13_ParallelStreamAPI::isPrime);//并行流,将任务切分成子任务执行end = System.currentTimeMillis();System.out.println(end - start);}static boolean isPrime(int num) {for(int i=2; i<=num/2; i++) {if(num % i == 0) return false;}return true;} } 13、总结 总结 Callable相当于一Runnable但是它有返回值 Future:存储执行完产生的结果 FutureTask 相当于Future+Runnable,既可以执行任务,又能获取任务执行的Future结果 CompletableFuture 可以多任务异步,并对多任务控制,整合任务结果,细化完美,比如可以一个任务完成就可以整合结果,也可以所有任务完成才整合结果 4、ThreadPoolExecutor源码解析 依然只讲重点,实际还需要大家按照上篇博客中看源码的方式来看 1、常用变量的解释 // 1. ctl,可以看做一个int类型的数字,高3位表示线程池状态,低29位表示worker数量private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0));// 2. COUNT_BITS,Integer.SIZE为32,所以COUNT_BITS为29private static final int COUNT_BITS = Integer.SIZE - 3;// 3. CAPACITY,线程池允许的最大线程数。1左移29位,然后减1,即为 2^29 - 1private static final int CAPACITY = (1 << COUNT_BITS) - 1;// runState is stored in the high-order bits// 4. 线程池有5种状态,按大小排序如下:RUNNING < SHUTDOWN < STOP < TIDYING < TERMINATEDprivate static final int RUNNING = -1 << COUNT_BITS;private static final int SHUTDOWN = 0 << COUNT_BITS;private static final int STOP = 1 << COUNT_BITS;private static final int TIDYING = 2 << COUNT_BITS;private static final int TERMINATED = 3 << COUNT_BITS;// Packing and unpacking ctl// 5. runStateOf(),获取线程池状态,通过按位与操作,低29位将全部变成0private static int runStateOf(int c) { return c & ~CAPACITY; }// 6. workerCountOf(),获取线程池worker数量,通过按位与操作,高3位将全部变成0private static int workerCountOf(int c) { return c & CAPACITY; }// 7. ctlOf(),根据线程池状态和线程池worker数量,生成ctl值private static int ctlOf(int rs, int wc) { return rs | wc; }/ Bit field accessors that don't require unpacking ctl. These depend on the bit layout and on workerCount being never negative./// 8. runStateLessThan(),线程池状态小于xxprivate static boolean runStateLessThan(int c, int s) {return c < s;}// 9. runStateAtLeast(),线程池状态大于等于xxprivate static boolean runStateAtLeast(int c, int s) {return c >= s;} 2、构造方法 public ThreadPoolExecutor(int corePoolSize,int maximumPoolSize,long keepAliveTime,TimeUnit unit,BlockingQueue<Runnable> workQueue,ThreadFactory threadFactory,RejectedExecutionHandler handler) {// 基本类型参数校验if (corePoolSize < 0 ||maximumPoolSize <= 0 ||maximumPoolSize < corePoolSize ||keepAliveTime < 0)throw new IllegalArgumentException();// 空指针校验if (workQueue == null || threadFactory == null || handler == null)throw new NullPointerException();this.corePoolSize = corePoolSize;this.maximumPoolSize = maximumPoolSize;this.workQueue = workQueue;// 根据传入参数unit和keepAliveTime,将存活时间转换为纳秒存到变量keepAliveTime 中this.keepAliveTime = unit.toNanos(keepAliveTime);this.threadFactory = threadFactory;this.handler = handler;} 3、提交执行task的过程 public void execute(Runnable command) {if (command == null)throw new NullPointerException();/ Proceed in 3 steps: 1. If fewer than corePoolSize threads are running, try to start a new thread with the given command as its first task. The call to addWorker atomically checks runState and workerCount, and so prevents false alarms that would add threads when it shouldn't, by returning false. 2. If a task can be successfully queued, then we still need to double-check whether we should have added a thread (because existing ones died since last checking) or that the pool shut down since entry into this method. So we recheck state and if necessary roll back the enqueuing if stopped, or start a new thread if there are none. 3. If we cannot queue task, then we try to add a new thread. If it fails, we know we are shut down or saturated and so reject the task./int c = ctl.get();// worker数量比核心线程数小,直接创建worker执行任务if (workerCountOf(c) < corePoolSize) {if (addWorker(command, true))return;c = ctl.get();}// worker数量超过核心线程数,任务直接进入队列if (isRunning(c) && workQueue.offer(command)) {int recheck = ctl.get();// 线程池状态不是RUNNING状态,说明执行过shutdown命令,需要对新加入的任务执行reject()操作。// 这儿为什么需要recheck,是因为任务入队列前后,线程池的状态可能会发生变化。if (! isRunning(recheck) && remove(command))reject(command);// 这儿为什么需要判断0值,主要是在线程池构造方法中,核心线程数允许为0else if (workerCountOf(recheck) == 0)addWorker(null, false);}// 如果线程池不是运行状态,或者任务进入队列失败,则尝试创建worker执行任务。// 这儿有3点需要注意:// 1. 线程池不是运行状态时,addWorker内部会判断线程池状态// 2. addWorker第2个参数表示是否创建核心线程// 3. addWorker返回false,则说明任务执行失败,需要执行reject操作else if (!addWorker(command, false))reject(command);} 4、addworker源码解析 private boolean addWorker(Runnable firstTask, boolean core) {retry:// 外层自旋for (;;) {int c = ctl.get();int rs = runStateOf(c);// 这个条件写得比较难懂,我对其进行了调整,和下面的条件等价// (rs > SHUTDOWN) || // (rs == SHUTDOWN && firstTask != null) || // (rs == SHUTDOWN && workQueue.isEmpty())// 1. 线程池状态大于SHUTDOWN时,直接返回false// 2. 线程池状态等于SHUTDOWN,且firstTask不为null,直接返回false// 3. 线程池状态等于SHUTDOWN,且队列为空,直接返回false// Check if queue empty only if necessary.if (rs >= SHUTDOWN &&! (rs == SHUTDOWN &&firstTask == null &&! workQueue.isEmpty()))return false;// 内层自旋for (;;) {int wc = workerCountOf(c);// worker数量超过容量,直接返回falseif (wc >= CAPACITY ||wc >= (core ? corePoolSize : maximumPoolSize))return false;// 使用CAS的方式增加worker数量。// 若增加成功,则直接跳出外层循环进入到第二部分if (compareAndIncrementWorkerCount(c))break retry;c = ctl.get(); // Re-read ctl// 线程池状态发生变化,对外层循环进行自旋if (runStateOf(c) != rs)continue retry;// 其他情况,直接内层循环进行自旋即可// else CAS failed due to workerCount change; retry inner loop} }boolean workerStarted = false;boolean workerAdded = false;Worker w = null;try {w = new Worker(firstTask);final Thread t = w.thread;if (t != null) {final ReentrantLock mainLock = this.mainLock;// worker的添加必须是串行的,因此需要加锁mainLock.lock();try {// Recheck while holding lock.// Back out on ThreadFactory failure or if// shut down before lock acquired.// 这儿需要重新检查线程池状态int rs = runStateOf(ctl.get());if (rs < SHUTDOWN ||(rs == SHUTDOWN && firstTask == null)) {// worker已经调用过了start()方法,则不再创建workerif (t.isAlive()) // precheck that t is startablethrow new IllegalThreadStateException();// worker创建并添加到workers成功workers.add(w);// 更新largestPoolSize变量int s = workers.size();if (s > largestPoolSize)largestPoolSize = s;workerAdded = true;} } finally {mainLock.unlock();}// 启动worker线程if (workerAdded) {t.start();workerStarted = true;} }} finally {// worker线程启动失败,说明线程池状态发生了变化(关闭操作被执行),需要进行shutdown相关操作if (! workerStarted)addWorkerFailed(w);}return workerStarted;} 5、线程池worker任务单元 private final class Workerextends AbstractQueuedSynchronizerimplements Runnable{/ This class will never be serialized, but we provide a serialVersionUID to suppress a javac warning./private static final long serialVersionUID = 6138294804551838833L;/ Thread this worker is running in. Null if factory fails. /final Thread thread;/ Initial task to run. Possibly null. /Runnable firstTask;/ Per-thread task counter /volatile long completedTasks;/ Creates with given first task and thread from ThreadFactory. @param firstTask the first task (null if none)/Worker(Runnable firstTask) {setState(-1); // inhibit interrupts until runWorkerthis.firstTask = firstTask;// 这儿是Worker的关键所在,使用了线程工厂创建了一个线程。传入的参数为当前workerthis.thread = getThreadFactory().newThread(this);}/ Delegates main run loop to outer runWorker /public void run() {runWorker(this);}// 省略代码...} 6、核心线程执行逻辑-runworker final void runWorker(Worker w) {Thread wt = Thread.currentThread();Runnable task = w.firstTask;w.firstTask = null;// 调用unlock()是为了让外部可以中断w.unlock(); // allow interrupts// 这个变量用于判断是否进入过自旋(while循环)boolean completedAbruptly = true;try {// 这儿是自旋// 1. 如果firstTask不为null,则执行firstTask;// 2. 如果firstTask为null,则调用getTask()从队列获取任务。// 3. 阻塞队列的特性就是:当队列为空时,当前线程会被阻塞等待while (task != null || (task = getTask()) != null) {// 这儿对worker进行加锁,是为了达到下面的目的// 1. 降低锁范围,提升性能// 2. 保证每个worker执行的任务是串行的w.lock();// If pool is stopping, ensure thread is interrupted;// if not, ensure thread is not interrupted. This// requires a recheck in second case to deal with// shutdownNow race while clearing interrupt// 如果线程池正在停止,则对当前线程进行中断操作if ((runStateAtLeast(ctl.get(), STOP) ||(Thread.interrupted() &&runStateAtLeast(ctl.get(), STOP))) &&!wt.isInterrupted())wt.interrupt();// 执行任务,且在执行前后通过beforeExecute()和afterExecute()来扩展其功能。// 这两个方法在当前类里面为空实现。try {beforeExecute(wt, task);Throwable thrown = null;try {task.run();} catch (RuntimeException x) {thrown = x; throw x;} catch (Error x) {thrown = x; throw x;} catch (Throwable x) {thrown = x; throw new Error(x);} finally {afterExecute(task, thrown);} } finally {// 帮助gctask = null;// 已完成任务数加一 w.completedTasks++;w.unlock();} }completedAbruptly = false;} finally {// 自旋操作被退出,说明线程池正在结束processWorkerExit(w, completedAbruptly);} } 本篇文章为转载内容。原文链接:https://blog.csdn.net/grd_java/article/details/113116244。 该文由互联网用户投稿提供,文中观点代表作者本人意见,并不代表本站的立场。 作为信息平台,本站仅提供文章转载服务,并不拥有其所有权,也不对文章内容的真实性、准确性和合法性承担责任。 如发现本文存在侵权、违法、违规或事实不符的情况,请及时联系我们,我们将第一时间进行核实并删除相应内容。
2023-07-21 16:19:45
328
转载
转载文章
...密协议强化与新型证书策略:随着网络安全技术的发展,越来越多的网站采用更高级别的SSL/TLS证书以增强安全性。例如,Let's Encrypt等项目推动了免费SSL证书的广泛部署,使得HTTP站点向HTTPS过渡成为主流趋势。这对网络爬虫而言意味着必须更新应对策略,理解和适配不同类型的SSL证书验证机制。 3. 反爬策略的技术演进与对策研究:面对日益复杂的网站反爬机制,诸如基于用户行为分析、动态验证码、IP封锁等手段层出不穷。研究人员正在探索更先进的模拟登录方法和维持session活性技术,同时利用AI图像识别技术破解复杂验证码也成为业界热门话题。 4. 网络爬虫伦理与法律边界探讨:在实际应用中,网络爬虫技术往往涉及道德和法律问题。例如,未经许可抓取受版权保护的内容或侵犯用户隐私。相关案例引发了关于合理使用网络爬虫、尊重数据来源权和用户知情权的深入讨论,这对于指导开发者正确运用cookie和session管理用户状态具有重要意义。 综上所述,无论是从技术层面还是法律伦理角度,处理不信任SSL证书、cookie和session的相关知识都是网络爬虫领域发展的重要组成部分。不断跟进相关政策变化和技术演进,将有助于我们更好地在遵守规则的前提下进行有效的数据采集和分析工作。
2023-03-01 12:40:55
563
转载
站内搜索
用于搜索本网站内部文章,支持栏目切换。
知识学习
实践的时候请根据实际情况谨慎操作。
随机学习一条linux命令:
pkill -9 process_name
- 强制终止指定进程。
推荐内容
推荐本栏目内的其它文章,看看还有哪些文章让你感兴趣。
2023-04-28
2023-08-09
2023-06-18
2023-04-14
2023-02-18
2023-04-17
2024-01-11
2023-10-03
2023-09-09
2023-06-13
2023-08-07
2023-03-11
历史内容
快速导航到对应月份的历史文章列表。
随便看看
拉到页底了吧,随便看看还有哪些文章你可能感兴趣。
时光飞逝
"流光容易把人抛,红了樱桃,绿了芭蕉。"