Question regarding sharing a resource between Goroutine executions

agolangf · · 64 次点击    
这是一个分享于 的资源,其中的信息可能已经有所发展或是发生改变。
<p>If I&#39;m analysing a dataset asynchronously using several goroutines but I don&#39;t want to accidently hit the same thing more than once I want to have a blacklist that&#39;s updated as they go.</p> <p>Problem is I&#39;m wondering the best way to do this, my synchronous idea would be use a blacklist array to store ones I&#39;ve already done and check the current one I&#39;m analysing is in the blacklist.</p> <p>How do I do this with goroutines? If I don&#39;t cater for it they may find the same one before a value is added to the blacklist. </p> <p>Do I use a mutex lock? Or is there a better pattern?</p> <hr/>**评论:**<br/><br/>seankhliao: <pre><p>have a single goroutine read the data and feed it into a channel then have multiple goroutines receive from the channel to process</p> <p>if you need to save the ouputs, open another channel and have all the processing goroutines feed into that one and a have a single goroutine receive from it to write to disk/wherever </p></pre>bustyLaserCannon: <pre><p>Thanks that makes sense but what if the values are duplicated and nested inside the values I&#39;d be passing to the channel to process. The processing channel may need to do the process again but may be doing it on one that the other goroutines have already done. </p></pre>seankhliao: <pre><p>there are several ways depending on your data/processing</p> <p>if your processing doesn&#39;t require the values to be passed in together you could unroll/flatten your data and deduplicate with a simple map before you feed it in</p> <p>if it does need the entire bundle, then it might easier to use the new <code>sync.Map</code> and check in the processing goroutines (update the map before beginning to process a value)</p></pre>bustyLaserCannon: <pre><p>Which would be the faster approach assuming both were possible?</p></pre>seankhliao: <pre><p>unrolling before feeding, simply because you don&#39;t need to deal with synchronization</p> <p>you can check <a href="" rel="nofollow">this article</a> on the performance between the two</p> <p>edit: unless your unrolling takes a very long time, in which case maybe try the <code>sync.Map</code></p></pre>titpetric: <pre><p>From what I understand, you&#39;re trying to deduplicate messages that are coming into several goroutines. There are several ways to go about it, but yes, generally you would keep a &#34;seen&#34; map that will tell you if a message has already been processed. Using a mutex is a valid way to ensure that only one goroutine can read or write from it. You could also resort to the <code>sync.Map</code> construct, added in 1.9, where locking is handled implicitly.</p></pre>
64 次点击  
加入收藏 微博
添加一条新回复 (您需要 登录 后才能回复 没有账号 ?)
  • 请尽量让自己的回复能够对别人有帮助
  • 支持 Markdown 格式, **粗体**、~~删除线~~、`单行代码`
  • 支持 @ 本站用户;支持表情(输入 : 提示),见 Emoji cheat sheet
  • 图片支持拖拽、截图粘贴等方式上传