石锅拌饭

rails缓存机制的几个问题

by Robin Lu on Apr.28, 2007, about , ,

ruby on rails提供了一些内建的cache机制,我们比较多的用到了其中的fragment cache。在实际使用过程中,发现了一些问题,如果你不注意,performance和正确性可能都会受到影响。

cache的重复读取问题

这是从Agile Web Development with Rails中抄下来的一段代码:

1
2
3
4
5
6
7
def list 
  @dynamic_content = Time.now.to_s 
  unless read_fragment(:action => 'list') 
    logger.info("Creating fragment") 
    @articles = Article.find_recent 
  end 
end

相信大家也都是用判断read_fragment返回值的方法来看cache是否存在。但你有没有想过,read_fragment这个函数的真正含义是读取cache。在view里,helper cache又会把cache的内容读一遍。假设你的cache用的是FileStore,下面是cache.rb中read的实现:

1
2
3
def read(name, options = nil) #:nodoc:
    File.open(real_file_path(name), 'rb') { |f| f.read } rescue nil
end

两次调用read_fragment就是两次独立的文件IO,而且每次都全文读入。第一次读完全是浪费。
解决方案:
一个是在controller里将read_fragment返回的内容保存在一个instance variable里,再到view里显示。这个做法的缺点是无法在view里使用现成的helper cache了,需要重写一个类似的helper,判断相关的instance variable而不是再去read。
另一个方法是自己写一个check_fragment,用File.exist?去判断,效率要高得多。

cache判断的原子性问题

还是上面那段代码:

1
2
3
4
5
6
7
def list 
  @dynamic_content = Time.now.to_s 
  unless read_fragment(:action => 'list') 
    logger.info("Creating fragment") 
    @articles = Article.find_recent 
  end 
end

假设controller在read_fragment的时候,cache存在,但就在运行出了这个unless之后,render view之前,cache被清除了,这完全有可能发生。就在render view的时候,helper cache发现cache不存在,看看下面的代码(仍然来自Agile Web Development with Rails),会发生什么?

1
2
3
4
5
6
7
< % cache do %> < !- Here's the content we cache -> 
  <ul> 
    < % for article in @articles -%> 
      <li><p>< %= h(article.body) %></p></li> 
    < % end -%> 
  </ul> 
< % end %> < !- End of the cached content ->

@articles是nil,你的程序抛出异常了!
解决方法:
没有什么万灵药,要根据具体的情况来分析。最简单的就是在view里判断@articles是否存在,总比抛异常好。还有一个方法就是把@articles的初始化放到view的cache do里面,如果嫌view里代码太多,就放一个helper里。

可怕的expire_fragment

其实只有一半是可怕的。
如果用FileStore,expire_fragment做的其实就是删除文件。假如你指定了是哪个cache,没问题。但expire_fragment还支持正则表达式,假如你指定的是一个正则表达式,它会干什么?看action_controller/caching.rb里的实现:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
def delete_matched(matcher, options) #:nodoc:
  search_dir(@cache_path) do |f|
    if f =~ matcher
      begin
        File.delete(f)
      rescue Object => e
        # If there's no cache, then there's nothing to complain about
      end
    end
  end
end
...
def search_dir(dir, &amp;callback)
  Dir.foreach(dir) do |d|
    next if d == "." || d == ".."
    name = File.join(dir, d)
    if File.directory?(name)
      search_dir(name, &amp;callback)
    else
      callback.call name
    end
  end
end

遍历所有cache目录下的文件,所有。然后一个一个去和正则表达式匹配。如果你不幸有上千个cache fragment,知道有多慢么?自己试试吧。
解决方案:
首先不是一个解决方案,是一个忠告:不要传给expire_fragment一个正则表达式。
其次,你可能可以写一个expire_fragment_from(root_path, reg),让它从一个指定的路径下做这样的遍历。然后优化cache fragment的目录结构,尽量缩小作用域。

Ruby On Rails是一个很好的框架,但仍然有要改进的地方。如果你在使用rails cache上有什么心得,欢迎交流。

:, ,

6 Comments for this entry

  • Caiwangqin

    谢谢分享,robin 是一个非常细致入微的程序员,这是优秀程序员的标致吧 :)

  • Kevin

    我的做法是, 在 controller 裡建一個 late finder (也有人叫 lazy data fetcher), 把實際從Database讀資料的工作延後到 rendering view 時才做, 這樣就不需要在 controller 裡判斷 cache 是否存在, 也不需要擔心上文提到的問題.

  • confach

    研究果真很透彻,有些问题我也遇见过!谢谢分享啦

  • bujiande

    关于那个 read 的 问题 , 好像社区里面 不推荐在 controller 里面 判断 , 直接在 helper 里面写个方法 ,然后:

    ‘fragment_url’ , :object => helper_method %>

  • Robin Lu

    楼上的方法不错

  • cricy

    ‘fragment_url’ , :object => helper_method %>

    ‘fragment_url’是指什么啊,能说清楚点吗?意思是也在view中判断,如果没有再查找

1 Trackback or Pingback for this entry

Archives

Browse by tags