TheWebFellas - Ruby on Rails Development Specialists UK - Rails 2.1: now with better integrated caching Comments

Comment on 'Rails 2.1: now with better integrated caching' by hosiawak

2008-09-03T16:26:05Z

This description is the one I've been looking for. It should be added to the offical Rails doc (or at least appended as notes via ApiDock.com)

Comment on 'Rails 2.1: now with better integrated caching' by Rob Anderton

2008-08-20T07:56:29Z

Thanks for compliments. I'd be happy for it to be used to improve the API docs: it'll need a few tweaks to reflect the most recent changes but they're nothing too major. I've not got time at the moment to do it (pesky clients keep insisting on paying us to develop their sites), but if someone feels like doing it before I get around to it then feel free!

Comment on 'Rails 2.1: now with better integrated caching' by karmi

2008-08-19T09:47:15Z

Hi Rob,

thanks!!, still the best documentation for this feature online :)

Would you mind to add it to http://github.com/lifo/docrails/tree/master/actionpack/lib/action_controller/caching.rb so it can be merged into the API docs?

Karel

Comment on 'Rails 2.1: now with better integrated caching' by Rob Anderton

2008-08-16T17:55:43Z

You might want to take a look at FiveRuns’ enhanced memcache-client gem.

Comment on 'Rails 2.1: now with better integrated caching' by Wes

2008-08-15T18:39:14Z

Is there a way to have Rails automatically attempt to reconnect to the memcache server if memcache stops and is then restarted?

I am trying to figure out how to make my application still use memcache if the server stops.

Comment on 'Rails 2.1: now with better integrated caching' by Nathan Leavitt

2008-08-14T18:31:16Z

Just want to post this for those who are implementing a different way to store the session.

For those implementing this for the first time in Rails 2.1 don’t forget to uncomment the :secret from the protect_from_forgery line in your application.rb file. I probably spent about 2 hours figuring this out :) Thanks for the post!

Comment on 'Rails 2.1: now with better integrated caching' by Rob Anderton

2008-08-12T07:54:54Z

There are now a number of tickets and patches on Lighthouse that attempt to clean up some of the inconsistencies in Rails caching:

#785 looks at the problem of reloading classes in development mode breaking the MemoryStore. There seems to be some reluctance to fix it though, which is a shame.
#799 looks at making objects in the MemoryStore immutable.
#800 fixes a major problem with the FileStore in that it currently can only store strings because it doesn't marshal objects (something I posted about on RailsForum a while back but didn't get chance to turn into a patch).

So take a look, make a comment and maybe we'll get some of these much-needed fixes implemented.

Comment on 'Rails 2.1: now with better integrated caching' by Matt Hodgson

2008-07-29T23:48:56Z

Hey there, thanks for this article! One small fix to the memcache initialization is needed. The hash parameters need to be wrapped in curly brackets or else rails complains and wont initialize. Your example should look like so:

config.cache_store = :mem_cache_store, 'localhost', '192.168.1.1:1001', { :namespace => 'test' }

Thanks for the documentation!

Comment on 'Rails 2.1: now with better integrated caching' by Russ Smith

2008-07-25T03:24:04Z

I just wrote this really quick to get around the problem. Seems there should be a setting to not perform any caching using the new cache module.


if RAILS_ENV == 'development'
  module ActiveSupport
    module Cache
      class Store
        def fetch(key, options = {})
          log("write (will save)", key, nil)
          (block_given?) ? yield : nil
        end
      end
    end
  end
end

Comment on 'Rails 2.1: now with better integrated caching' by Jochen

2008-07-11T20:33:24Z

@Joel: I had the same issue with the default memory_store in Rails 2.1. Changing to mem_cache_store fixed that issue and now it works as it should (after fixing the undefined class/module issue as Rob posted). So maybe you still use the default memory_store instead of the mem_cache_store.

Comment on 'Rails 2.1: now with better integrated caching' by Rob Anderton

2008-07-10T07:46:30Z

I've not come across that problem yet, but I have hit an undefined class/module error when storing ActiveRecord objects in the cache. This one is caused when reading from the cache: the memcache-client gem tries to unmarshal the object (using Marshal.load) but doesn't know about the class and so it fails.

It pre-dates the integration of caching into Rails and it looks like it can be fixed either by referencing the class before trying to fetch an object from the cache (as described here) or by using require_dependency.

Maybe it is something similar causing the weirdness you're seeing?

Comment on 'Rails 2.1: now with better integrated caching' by Joel Nylund

2008-07-09T21:04:42Z

Im having some trouble using memcached in the new stuff, when I put an array of active records in and then get back out of the cache, it barfs if I try to call a method on any of the objects with no method error, also if I try to get the id I get stack too deep, is there a rule against storing active records in memcached, is it only for primitive types? Is there a setting im missing? When I open in the debugger, the object appears to be there, I can see the attributes, I just cant call a method on it.

Comment on 'Rails 2.1: now with better integrated caching' by Peter Abrahamsen

2008-07-09T16:22:53Z

I'm really glad caching is making it into mainline ActiveRecord. cache_fu seems a bit stalled.

Does AR optimistic locking support still work?

There's some concern around these parts of race conditions. The suggestion was to flag any cache items that would be updated during an upcoming transaction. Other processes that try to hit any such cache item would hit the DB upon seeing the flag, and block on the transaction as appropriate, or use the known-stale data if they can tolerate it.

There's a plugin that gives you an after commit hook, which we could use to ensure that only actually committed data makes it into memcache, and to clear the flagged entries (which would have a low TTL in case the updating process dies mid-transaction).

Make sense? Is anyone doing anything like this? I think it would work best with memcache check-and-set (CAS) support, and of course flags, neither of which is exposed by the pure-ruby memcache-client library, unfortunately. They are supported by libmemcached, though.

We're planning on doing this work ourselves, and will toss a patch up somewhere if we can convince ourselves that we're not crazy (might take a while).

Comment on 'Rails 2.1: now with better integrated caching' by Rob Anderton

2008-07-08T08:35:31Z

Yes, the MemoryStore just ignores the options hash, so the :expires_in setting has no effect. The same is true of the FileStore.

It wouldn't be too hard to add support for options like :expires_in to these stores so that they really could become drop-in replacements for more advanced stores like Memcached - anybody fancy writing a patch?

Comment on 'Rails 2.1: now with better integrated caching' by Joel Nylund

2008-07-08T02:04:36Z

so what happens if I use :expires_in => 15.minutes but don't have memcached as the setting, it doesn't seem to barf when I use it, is it just ignored. I would like to figure out a way to use memory_store in dev and memcached in prod and qa, so each developer doesnt need to setup and run memcached.

Comment on 'Rails 2.1: now with better integrated caching' by Rob Anderton

2008-06-30T10:57:09Z

@Peter: you’re quite right - due to a missing call to default_cache_store in the Rails::Configuration class constructor then the tmp/cache check is never carried out - I’ve submitted a bug report.

@Ryan: glad I could be an inspiration :D Thanks for pointing out fetch - in my excitement to write this up I completely overlooked it!

For those who haven’t yet seen Ryan’s 115th Railscast you can watch it here.

Comment on 'Rails 2.1: now with better integrated caching' by Peter Wagenet

2008-06-27T20:23:42Z

You say: "By default if the tmp/cache directory exists in your application root then Rails will use a FileStore otherwise it’ll use a MemoryStore as the global cache."

In my experience this used to be the case, but it doesn't seem to be as of Rails 2.1. I did some poking around in the source and discovered that though Rails::Initializer has a default_cache_store method which has this behavior, this method is never referenced.

Instead initialize_cache is run which, in turn, runs ActiveSupport::Cache.lookup_store. If no cache_store is set lookup_store will default to ActiveSupport::Cache::MemoryStore.new. It does not check for /tmp/cache.

Hope this will save someone else some time.

Comment on 'Rails 2.1: now with better integrated caching' by jeff

2008-06-19T19:54:41Z

great tutorial, thanks.

Comment on 'Rails 2.1: now with better integrated caching' by Ryan Bates

2008-06-19T15:39:48Z

Excellent! I was waiting for someone to shed the light on this. I'll probably be doing a Railscasts episode on the subject now. :)

BTW, it appears there's also a Rails.cache.fetch method which can be used like this.

Rails.cache.fetch('some_key') do
  'initial value' # used to first set the key if it doesn't exist
end

Haven't tested that yet though.

Comment on 'Rails 2.1: now with better integrated caching' by Jim Neath

2008-06-18T08:55:28Z

Thank god someone has finally covered all this.

Great work, Rob!

Comment on 'Rails 2.1: now with better integrated caching' by Piyush

2008-06-10T03:20:06Z

wow u saved my day ..cheers !!!!!!!!!!

Comment on 'Rails 2.1: now with better integrated caching' by Mitchell Hashimoto

2008-06-10T00:17:08Z

Amazing! I can't believe I can't find anyone who has documented this feature other than this blog as of the time of this writing. I am implementing a web service at the moment that pulls a lot of data from Amazon's S3 that I would like to cache and in researching rails cacheing capabilities I stumbled across this post and it is exactly what I needed.

Thank you very much.