Re-add support for memory_store

[fbogsany]

PR #154 removed support for memory_store, due to the requirement for cas/cas_multi.

This adds a MemoryFetcher ands uses it instead of CacheFetcher if the cache_backend doesn't respond to cas and cas_multi.

Fixes #160

@arthurnn @dylanahsmith for review /cc @camilo

[dylanahsmith]

Could you also add a test for fetch_multi when using the MemoryStore?

[dylanahsmith]

I don't think this is necessary. I think it omits keys from the hash if they aren't found in the cache.

[dylanahsmith]

LGTM other than my 2 comments.

[RobWu]

Get the same issue with dalli gem

/gems/identity_cache-0.2.0/lib/identity_cache/cache_fetcher.rb:49:in `cas_multi': undefined method `cas_multi' for #<ActiveSupport::Cache::DalliStore:0x007fb15c667000> (NoMethodError)

Would be great if it still would be possible to use alternativ memcached clients.

[camilo]

@RobWu we are moving towards CAS over SET to avoid race conditions that can potentially return wrong data, is there a version of Dalli that supports CAS?

[RobWu]

There is cas but can't find multi_cas

http://www.rubydoc.info/github/mperham/dalli/Dalli/Client#cas-instance_method

[camilo]

@RobWu yeah. @fbogsany had to add multi cas support to memcached for this to work, so dalli would need the same, we probably don't have the time now to add it to dalli, but that is about all that would need to happen, for it to work with master.

[RobWu]

Found something:
https://github.com/mperham/dalli/blob/master/lib/dalli/cas/client.rb

[RobWu]

Will look into it tomorrow..
Use version 0.1.0 for now.

[arthurnn]

I am not sure if I like this approach.
What we are trying to solve here should be, make IDC work with other adapters(which include the memory one). So IMO, we should have a fallback backend which still uses sets, for people that dont use memcached gem, and not only create a fallback for memory stores.

If thats too hard, we maybe should drop support to the other adapters, and raise an error if the adapter is not memcached_store.

[fbogsany]

@arthurnn is that just a critique of the MemoryFetcher name? Other than the name, I think this PR does what you suggest.

[arthurnn]

@fbogsany sorry, I should've been more clear about it.. Yep.. the name MemoryFetcher is not ideal. But thats my point only.. the rest looks good.

[fbogsany]

@arthurnn groovy - does FallbackFetcher work? Or LegacyCacheFetcher? Naming is hard...

[arthurnn]

I like FallbackFetcher better. but both would work for me.. ❤️

[fbogsany]

FallbackFetcher it is.

[camilo]

🐑

[arthurnn]

❤️ ❤️ ❤️