blocks|key|1536871|text|我不认为Flink有一种状态类型可以支持这一点。我能想到的最接近的事情就是使用ListState。使用ListState，您可以像添加常规列表一样追加元素。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|1536872|对于用例，您可以读取状态，调用.get()，这将使您可以迭代，删除要删除的项，然后将状态推回。|style|CODE|1536873|从性能的角度来看，迭代可能并不理想，但另一方面，与从磁盘访问状态(如果您使用RocksDB作为状态后端)相比，迭代并不重要，因为序列化和反序列化会带来很大的成本。|1536874|entityMap|0|LINK|mutability|MUTABLE|url|https://nightlies.apache.org/flink/flink-docs-master/api/java/org/apache/flink/api/common/state/ListState.html^0|13|9|0|0|F|6|0|0^^$0|@$1|2|3|4|5|6|7|R|8|@]|9|@$A|S|B|T|1|U]]|C|$]]|$1|D|3|E|5|6|7|V|8|@$A|W|B|X|F|G]]|9|@]|C|$]]|$1|H|3|I|5|6|7|Y|8|@]|9|@]|C|$]]|$1|J|3|-4|5|6|7|Z|8|@]|9|@]|C|$]]]|K|$L|$5|M|N|O|C|$P|Q]]]]

<p>I don't think Flink has a state type that would support this out of the box.
The closest thing I can think of is to use <a href="https://nightlies.apache.org/flink/flink-docs-master/api/java/org/apache/flink/api/common/state/ListState.html" rel="nofollow noreferrer">ListState</a>. With ListState you can append elements like you would a regular list.</p>
<p>For your use case, you would read the state, call <code>.get()</code> which would give you an iterable that you can iterate on, removing the item you'd like to drop and then pushing the state back.</p>
<p>From a performance perspective, the iteration may not be ideal but on the other hand, it would not be significant in comparison to accessing state from disk (in case you're using RocksDB as a state backend) which incurs a heavy cost due to serialization and deserialization.</p>


blocks|key|761999|text|虽然这不是直接提供的，但是您可以使用MapState<Long,+Event>加上几个附加的ValueState<Long>值来跟踪当前活动的MapState索引范围。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|762000|当事件到达时，做一些大致类似的事情(但是使用Flink状态而不是这个伪代码)：|762001|map[nextIndex%2B%2B]+=+thisEvent;
if+(nextIndex+-+oldestIndex+>+n)+{
++map[oldestIndex%2B%2B].clear();
}|code-block|syntax|javascript|762002|entityMap^0|I|L|1A|G|0|0|0^^$0|@$1|2|3|4|5|6|7|O|8|@$9|P|A|Q|B|C]|$9|R|A|S|B|C]]|D|@]|E|$]]|$1|F|3|G|5|6|7|T|8|@]|D|@]|E|$]]|$1|H|3|I|5|J|7|U|8|@]|D|@]|E|$K|L]]|$1|M|3|-4|5|6|7|V|8|@]|D|@]|E|$]]]|N|$]]

<p>While this isn't directly provided, you could achieve this with <code>MapState&lt;Long, Event&gt;</code> plus a couple of additional <code>ValueState&lt;Long&gt;</code> values to keep track of the currently active range of indexes into the MapState.</p>
<p>As events arrive, do something roughly like this (but using Flink state rather than this pseudocode):</p>
<pre class="lang-java prettyprint-override"><code>map[nextIndex++] = thisEvent;
if (nextIndex - oldestIndex &gt; n) {
  map[oldestIndex++].clear();
}
</code></pre>


blocks|key|1536920|text|我认为您可以通过实现LinkedHashMap的某种"Flink调整“版本来创建这个功能|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|1536921|您必须记住，MapState并不是一个经典的Java，而是一个支持状态管理的详细实现。|1536922|本质上--我认为您可以使用MapState和ListState的组合来实现这样的事情，遵循一些关于如何实现这一点的指导方针--下面的例子是：Java提示和技巧。：Queue混合--在Java中创建数据结构(techtipsjava.blogspot.com)。|1536923|关于TTL+-我不确定键移除(因为它是定时的)会如何影响这样的结构，这必须仔细考虑。|1536924|entityMap|0|LINK|mutability|MUTABLE|url|http://techtipsjava.blogspot.com/2013/05/queue-map-hybrid-creating-data.html^0|A|D|0|0|D|8|M|9|1Y|1N|0|0|0^^$0|@$1|2|3|4|5|6|7|T|8|@$9|U|A|V|B|C]]|D|@]|E|$]]|$1|F|3|G|5|6|7|W|8|@]|D|@]|E|$]]|$1|H|3|I|5|6|7|X|8|@$9|Y|A|Z|B|C]|$9|10|A|11|B|C]]|D|@$9|12|A|13|1|14]]|E|$]]|$1|J|3|K|5|6|7|15|8|@]|D|@]|E|$]]|$1|L|3|-4|5|6|7|16|8|@]|D|@]|E|$]]]|M|$N|$5|O|P|Q|E|$R|S]]]]

<p>I think you can create this functionality yourself by implementing sort of &quot;Flink adjusted&quot; version of Java's <code>LinkedHashMap</code></p>
<p>You have to remember that MapState is not exactly a classic Java Map, but an elaborate implementation of it that enables State management.</p>
<p>Essentially - I think you can implement such a thing using a combiniation of <code>MapState</code> and <code>ListState</code>, following some guidelines on how to implement this - something along this example: <a href="http://techtipsjava.blogspot.com/2013/05/queue-map-hybrid-creating-data.html" rel="nofollow noreferrer">Java tips and tricks.: Queue Map Hybrid -- Creating Data Structures in Java (techtipsjava.blogspot.com)</a></p>
<p>Regarding TTL - I am not sure how would key removal (because it was timed-out) would affect such structure, this has to be carefully thought of.</p>


<h2>Background</h2>
<p>We want to keep in a Flink operator's state the last <code>n</code> unique id's.
When the <code>n+1</code> unique id arrives, we want to keep it and drop the oldest unique id in the state. This is in order to avoid an ever-growing state.</p>
<p>We already have a TTL (expiration time) mechanism in place. The size limit is another restriction we're looking to put in place.</p>
<p>Not every element holds a unique id.</p>
<h2>Question</h2>
<p>Does Flink provide an API that limits the number of elements in the state?</p>
<h2>Things tried</h2>
<ol>
<li>Using <code>MapState</code> with a <code>StateTtlConfig</code> generated TTL/expiration mechanism.</li>
<li>Window limited the number of processed elements, but not the number of elements in state.</li>
</ol>


Flink capped MapState

翻译质量差，导致语言生硬或混乱。

没有提供实际的解决方法或示例。

解答不清晰，无法理解或解决问题。

页面排版不美观，阅读体验差。

文章

问答

视频

学习中心

腾讯云实验室

直播

竞赛

腾讯云代码分析专区

腾讯iOA零信任安全管理系统专区

腾讯云架构师技术同盟交流圈

腾讯云数据库专区

腾讯云顾问专区

腾讯云原生专区

腾讯混元专区

腾讯云TCE专区

腾讯云Lighthouse专区

腾讯云HAI专区

腾讯云Edgeone专区

腾讯云存储专区

腾讯云智能专区

腾讯轻联专区 

腾讯云开发专区

TAPD专区

腾讯轻量云游戏服专区

腾讯云最具价值专家

腾讯云架构师技术同盟

腾讯云创作之星

腾讯云开发者先锋

腾讯云代码助手

云原生构建

TAPD 敏捷项目管理

Cloud Studio

SDK中心

API中心

命令行工具

涵盖代码开发、场景应用、自动测试全流程，助你从零构建专属AI助手

一站式MCP教程库，解锁AI应用新玩法

背景我们希望保持Flink操作符的状态，即最后一个n唯一id。当n+1唯一id到达时，我们希望保留它并删除状态中最古老的唯一id。这是为了避免一个不断增长的状态。我们已经有了一个TTL (过期时间)机制。尺寸限制是我们想要实施的另一个限制。并不是每个元素都拥有唯一的id。问题Flink是否提供了一个API来限制状态中元...

问Flink上限MapState
EN

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问Flink上限MapStateEN

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问Flink上限MapState
EN