SmashCache

A simple caching and sweeping utility

There are only two hard things in Computer Science: cache invalidation and naming things. -Phil Karlton

After much research at the time I could not find a simple caching utility that handled sweeping in a good way (if at all).

For better or worse, I opted to write my own and started with a sample file caching utility and updated it to use rails_cache and mem_cache.

Smash Cache allows individual and tag based sweeping, so you can basically cache an item with multiple keys.

I use the tagging to save data under data id specific keys and under general keys to sweep/clear entire groups of cached items.

When data is updated or created, I can use the generic restful endpoints to sweep all the necessary cached data - I generally put these sweep helper in a separate class.

The file caching is not as fully implemented as the other two at this time and will probably be removed as it was more of an experiment so treat it as Deprecated

See list at the bottom for other TODOs and deficiencies

##Example use:

Include this file somewhere, I put mine in lib/utility/smash_cache.rb -- this example is structured around that

Place in the config, and update the appropriate parent class to create these as config vars

environment config

config.smash_cache.enabled = true
config.smash_cache.default_cache_type = :rails_cache  (or :file or :mem_cache)

configuration.rb

@smash_cache = ActiveSupport::OrderedOptions.new
@smash_cache.enabled              = false
@smash_cache.default_cache_type   = :rails_cache

Prep in an initializer

 API_CACHE = Utility::SmashCache.new 'api_cache', 30.minutes, true, 50, 'app_monitors/api_cache/hit_count', 'app_monitors/api_cache/miss_count', 'app_monitors/api_cache/cached_object_count'
 Log a cache creation time if you want...
 API_CACHE.replace '/started_cache', '{ "api_cache_started":' + DateTime.now.strftime('%Y_%m_%d %H:%M %Ss').to_s.to_s + '}', nil

Use it...

Place (something like) these in the class or base class where @data is the data to cache or retrieve from the cache

 def cache_api_call expires = 1.hour
   tag_array = Array.new  add a tag if you need, this one removes the query string to group all similar restful routes in one group to make sweeping endpoints with paging easier
   tag_array.push request.original_fullpath.split('?').first.to_s  split off the query string
   API_CACHE.add(request.original_fullpath, @data.to_json, expires, tag_array)
 end

 def check_api_cache
  Add CORS checking if needed for cross domin calls
  cached_results = API_CACHE.find request.original_fullpath
  render :json => cached_results, :status => :ok and return unless cached_results.blank?
 end

In your controllers you can add this, which will only be called if check_api_cache fails

If your cache does fail, cache_api_call will be called at the end of this controller call, caching the call

 def show
   @data = Dog.find params[:id]
   if @data.nil?
     render :json => Errors.unprocessable_entity('Dog', params[:id]), :status => :unprocessable_entity
   else
     render :json => @data, :status => :ok
   end
 end

call (something like) these methods using the appropriate before_filter or after_filter

 before_filter :check_api_cache, :except => [:create_dog,dog :properties] if Utility::SmashCache.enabled?
 after_filter :cache_api_call, :except => [:create_dog, :properties] if Utility::SmashCache.enabled? && [email protected]?

Sweeping

by key, key pattern and tag...

key = exact match

pattern, :rails_cache and :file only - not recommended for :rails_cache (slow - use tagging functionality instead), recommended for :file (directory deletion)

tag, :rails_cache and :mem_cache only - use in lieu of pattern for mem_cache (mem_cache doesn't support it) - can push in multiple tags for one set of @data results

Example Sweeping

  def self.sweep_dog_data dog_id
    return unless Utility::SmashCache.enabled?
    API_CACHE.smash_by_pattern('/v1/dogs/' + dog_id.to_s) unless dog_id.blank?
  end
 
  def self.sweep_all_dog_endpoints dog_id=nil
    return unless Utility::SmashCache.enabled?
 
    API_CACHE.smash_by_pattern('/v1/dogs')
    API_CACHE.smash_by_pattern('/v1/dogs/' + dog_id.to_s) unless dog_id.blank?
 
    dogshow = DogShow.find_by_dog_id dog_id
    API_CACHE.smash_by_pattern('/v1/dog_shows')
    API_CACHE.smash_by_pattern('/v1/dog_shows/' + dogshow.to_s) unless dogshow.blank?
    end
  end
 

TODO

Refactor, restructure, consolidate multiple/confusing concepts (patterns vs tags)

Better instructions and samples

Tests

Cleanup sweeping and functionality deficiencies in different types

File caching issues, race conditions, tagging, etc - or remove file caching?