mongoid query caching

问题:

Rails' ActiveRecord has a feature called Query Caching (ActiveRecord::QueryCache) which saves the result of SQL query for the life-span of a request. While I'm not very familiar with the internals of the implementation, I think that it saves the query results somewhere in the Rack env, which is discarded in the end of the request.

The Mongoid, unfortunately, doesn't currently provide such feature, and this is exacerbated by the fact, that some queries occur implicitly (references). I'm considering to implement this feature, and I'm curious, where and how Mongoid (or, perhaps, mongo driver?) should be hooked in order to implement this.

回答1:

Mongoid has caching, described under http://mongoid.org/en/mongoid/docs/extras.html

Also MongoDB itself has caching ability: http://www.mongodb.org/display/DOCS/Caching

The mongoid caching extra knows 2 different cases: Caching of all queries of a model or caching of a query.

Mongoid caching seems to work slightly different: it looks like mongoid delegates caching to mongodb. (In the sources of mongoid I only can find option settings for caching but no cache module.)

Finally would say, there is no real difference in the caching in general -- in memory is in fact in memory! No matter if it's in the app or in the database.

I don't prefer to implement an extra caching algorithm, because this seems to be redundant and a RAM killer.

BTW: If your really want to cache results in-app you could try Rails.cache or another cache gem as a workaround.

回答2:

The other answer is obviously wrong. Not only mongoid or mongo driver doesn't cache the query, even if mongo would - it still might be on other machine across the network.

My solution was to wrap the receive_message in Mongo::Connection. Pros: one definite place Cons: deserialization still takes place

require 'mongo'
module Mongo
  class Connection
    module QueryCache
      extend ActiveSupport::Concern

      module InstanceMethods

        # Enable the selector cache within the block.
        def cache
          @query_cache ||= {}
          old, @query_cache_enabled = @query_cache_enabled, true
          yield
        ensure
          clear_query_cache
          @query_cache_enabled = old
        end

        # Disable the selector cache within the block.
        def uncached
          old, @query_cache_enabled = @query_cache_enabled, false
          yield
        ensure
          @query_cache_enabled = old
        end

        def clear_query_cache
          @query_cache.clear
        end

        def cache_receive_message(operation, message)
          @query_cache[operation] ||= {}
          key = message.to_s.hash
          log = "[MONGO] CACHE %s"
          if entry = @query_cache[operation][key]
            Mongoid.logger.debug log % 'HIT'
            entry
          else
            Mongoid.logger.debug log % 'MISS'
            @query_cache[operation][key] = yield
          end
        end

        def receive_message_with_cache(operation, message, log_message=nil, socket=nil, command=false)
          if query_cache_enabled
            cache_receive_message(operation, message) do
              receive_message_without_cache(operation, message, log_message, socket, command)
            end
          else
            receive_message_without_cache(operation, message, log_message, socket, command)
          end
        end
      end # module InstanceMethods

      included do
        alias_method_chain :receive_message, :cache
        attr_reader :query_cache, :query_cache_enabled
      end
    end # module QueryCache
  end # class Connection
end

Mongo::Connection.send(:include, Mongo::Connection::QueryCache)

回答3:

OK, Mongoid 4 supports QueryCache middleware.

Just add middleware in application.rb

config.middleware.use "Mongoid::QueryCache::Middleware"

And then profit:

  MOPED: 127.0.0.1:27017 QUERY        database=XXX collection=page_variants selector={"$query"=>{"_id"=>BSON::ObjectId('5564dabb6d61631e21d70000')}, "$orderby"=>{:_id=>1}} flags=[] limit=-1 skip=0 batch_size=nil fields=nil runtime: 0.4397ms
  MOPED: 127.0.0.1:27017 QUERY        database=XXX collection=page_variants selector={"$query"=>{"_id"=>BSON::ObjectId('5564dacf6d61631e21dc0000')}, "$orderby"=>{:_id=>1}} flags=[] limit=-1 skip=0 batch_size=nil fields=nil runtime: 0.4590ms
  QUERY CACHE                         database=XXX collection=page_variants selector={"$query"=>{"_id"=>BSON::ObjectId('5564c9596d61631e21d30000')}, "$orderby"=>{:_id=>1}}
  QUERY CACHE                         database=XXX collection=page_variants selector={"$query"=>{"_id"=>BSON::ObjectId('5564dabb6d61631e21d70000')}, "$orderby"=>{:_id=>1}}

---

Source:

Mongoid changelog

https://github.com/mongoid/mongoid/blob/master/CHANGELOG.md#new-features-2

3410 Mongoid now has a query cache that can be used as a middleware in Rack applications. (Arthur Neves)

For Rails:

config.middleware.use(Mongoid::QueryCache::Middleware)

回答4:

Mongoid 4.0+ now has a QueryCaching module: http://www.rubydoc.info/github/mongoid/mongoid/Mongoid/QueryCache

You can use it on finds by wrapping your lookups like so:

QueryCache.cache { MyCollection.find("xyz") }