feature: support for ActiveSupport::MemCacheStore

[elhu]

Hello,

This PR adds support for ActiveSupport::Cache::MemCacheStore, which is Rails' wrapper around Dalli.
Because of Dalli lack of support for race_condition_ttl, we're stuck with MemCacheStore... hence this PR!

[elhu]

Hmm, there's a compatibility issue with Rails 3.x's ActiveSupport::Cache::MemCacheStore...
Working a fix, I hope to ship it tomorrow.

[ktheory]

Hmm, there's a compatibility issue with Rails 3.x's ActiveSupport::Cache::MemCacheStore...
Working a fix, I hope to ship it tomorrow.

Thanks for working on that.

[elhu]

I've investigated this a bit, and there's an inconsistency in Rails 3.x, with ActiveSupport::Cache::MemCacheStore.

When using the Rails 3.2 (ActiveSupport::Cache::MemCacheStore + MemCache backend), the following returns 0:

Rails.cache.delete("foo")
Rails.cache.increment("foo") # => 0

Whereas with every other cache stores, such as Dalli, it returns nil.

I'm not sure it's a good idea to deal with that specificity in Rack::Attack, but since Rails 3.2 is end-of-life, I don't think I can fix it there either.

I could create a new MemCacheProxy class to that effect, but it's only going to be useful for Rails 3.x projects with the MemCache backend... any opinion or alternative idea :) ?

[elhu]

I've addressed the issue by creating a dedicate MemCacheProxy class for the ActiveSupport::Cache::MemCacheStore + MemCache backend case.

The original reason for this PR is still solved, and now everything should work well even for Rails 3.2!

[jarthod]

👍

[ktheory]

@elhu Very cool. Didn't realize dalli didn't support race_condition_ttl! So I agree it's valuable to support MemCacheStore.

I'll review this as soon as I can (probably in the next couple weeks.)

If you can report success running this patch in a production application, that'd be helpful testing.

[elhu]

I'm on vacation this week, but I can start testing things out on a production app starting the week after that.

I'll adjust the PR if I find any issues, and report here if I don't find any.

[elhu]

I've brought my branch up to date with master, and fixed a couple more things.

I am also running the patch on a live web-app, and have no problems to report so far.

[elhu]

@ktheory did you get a chance to review this?
Once the PR is OK for you, let me know, I'll squash my commits together to avoid clutter in the master tree 😃.

[ktheory]

Can you add memcache-client as a development dependency instead?

This file is auto-generated from Appraisals file.

[elhu]

I've addressed that. Didn't know about appraisal, very interesting stuff!
I've also refactored the Rack::Attack::StoreProxy.build method, which was getting too cluttered in my opinion (but I could move that to a separate PR if you think it's necessary).

[elhu]

I had to revert my refactoring of Rack::Attack::StoreProxy.build, MRI 2.0.0 and JRuby do not behave like MRI 2.1.x and 2.2.x when calling Kernel.const_defined? with a nested constant name.

[elhu]

@ktheory Is there anything more you would like to discuss regarding this PR?

[ktheory]

@elhu Looks good! 🍰

I'll aim to publish a new release by the end of the week.

[elhu]

Great, thanks!

[ktheory]

Just shipped v4.4.0: https://rubygems.org/gems/rack-attack/versions/4.4.0

[elhu]

Brilliant, thank you very much!