mguymon-cache-money 0.2.12

data/README

@@ -0,0 +1,210 @@
+## What is Cache Money ##
+
+Cache Money is a write-through and read-through caching library for ActiveRecord.
+
+Read-Through: Queries like `User.find(:all, :conditions => ...)` will first look in Memcached and then look in the database for the results of that query. If there is a cache miss, it will populate the cache.
+
+Write-Through: As objects are created, updated, and deleted, all of the caches are *automatically* kept up-to-date and coherent.
+
+## Howto ##
+### What kinds of queries are supported? ###
+
+Many styles of ActiveRecord usage are supported:
+
+ * `User.find`
+ * `User.find_by_id`
+ * `User.find(:conditions => {:id => ...})`
+ * `User.find(:conditions => ['id = ?', ...])`
+ * `User.find(:conditions => 'id = ...')`
+ * `User.find(:conditions => 'users.id = ...')`
+
+As you can see, the `find_by_`, `find_all_by`, hash, array, and string forms are all supported.
+
+Queries with joins/includes are unsupported at this time. In general, any query involving just equality (=) and conjunction (AND) is supported by `Cache Money`. Disjunction (OR) and inequality (!=, <=, etc.) are not typically materialized in a hash table style index and are unsupported at this time.
+
+Queries with limits and offsets are supported. In general, however, if you are running queries with limits and offsets you are dealing with large datasets. It's more performant to place a limit on the size of the `Cache Money` index like so:
+
+    DirectMessage.index :user_id, :limit => 1000
+    
+In this example, only queries whose limit and offset are less than 1000 will use the cache.
+
+### Multiple indices are supported ###
+
+    class User < ActiveRecord::Base
+      index :screen_name
+      index :email
+    end
+
+#### `with_scope` support ####
+
+`with_scope` and the like (`named_scope`, `has_many`, `belongs_to`, etc.) are fully supported. For example, `user.devices.find(1)` will first look in the cache if there is an index like this:
+
+    class Device < ActiveRecord::Base
+     index [:user_id, :id]
+    end
+
+### Ordered indices ###
+
+    class Message < ActiveRecord::Base
+      index :sender_id, :order => :desc
+    end
+
+The order declaration will ensure that the index is kept in the correctly sorted order. Only queries with order clauses compatible with the ordering in the index will use the cache:
+
+ * `Message.find(:all, :conditions => {:sender_id => ...}, :order => 'id DESC')`.
+
+Order clauses can be specified in many formats ("`messages`.id DESC", "`messages`.`id` DESC", and so forth), but ordering MUST be on the primary key column. 
+
+    class Message < ActiveRecord::Base
+      index :sender_id, :order => :asc
+    end
+    
+will support queries like:
+
+ * `Message.find(:all, :conditions => {:sender_id => ...}, :order => 'id ASC')`
+ * `Message.find(:all, :conditions => {:sender_id => ...})`
+ 
+Note that ascending order is implicit in index declarations (i.e., not specifying an order is the same as ascending). This is also true of queries (order is not nondeterministic as in MySQL).
+
+### Window indices ###
+
+    class Message < ActiveRecord::Base
+      index :sender_id, :limit => 500, :buffer => 100
+    end
+
+With a limit attribute, indices will only store limit + buffer in the cache. As new objects are created the index will be truncated, and as objects are destroyed, the cache will be refreshed if it has fewer than the limit of items. The buffer is how many "extra" items to keep around in case of deletes.
+
+It is particularly in conjunction with window indices that the `:order` attribute is useful.
+
+### Calculations ###
+
+`Message.count(:all, :conditions => {:sender_id => ...})` will use the cache rather than the database. This happens for "free" -- no additional declarations are necessary.
+
+### Version Numbers ###
+
+    class User < ActiveRecord::Base
+      version 7
+      index ...
+    end
+    
+You can increment the version number as you migrate your schema. Be careful how you deploy changes like this as during deployment independent mongrels may be using different versions of your code. Indices can be corrupted if you do not plan accordingly.
+
+### Transactions ###
+
+Because of the parallel requests writing to the same indices, race conditions are possible. We have created a pessimistic "transactional" memcache client to handle the locking issues.
+
+The memcache client library has been enhanced to simulate transactions.
+
+    $cache.transaction do
+      $cache.set(key1, value1)
+      $cache.set(key2, value2)
+    end
+
+The writes to the cache are buffered until the transaction is committed. Reads within the transaction read from the buffer. The writes are performed as if atomically, by acquiring locks, performing writes, and finally releasing locks. Special attention has been paid to ensure that deadlocks cannot occur and that the critical region (the duration of lock ownership) is as small as possible.
+
+Writes are not truly atomic as reads do not pay attention to locks. Therefore, it is possible to peak inside a partially committed transaction. This is a performance compromise, since acquiring a lock for a read was deemed too expensive. Again, the critical region is as small as possible, reducing the frequency of such "peeks".
+
+#### Rollbacks ####
+
+    $cache.transaction do
+      $cache.set(k, v)
+      raise
+    end
+
+Because transactions buffer writes, an exception in a transaction ensures that the writes are cleanly rolled-back (i.e., never committed to memcache). Database transactions are wrapped in memcache transactions, ensuring a database rollback also rolls back cache transactions.
+
+Nested transactions are fully supported, with partial rollback and (apparent) partial commitment (this is simulated with nested buffers).
+
+### Mocks ###
+
+For your unit tests, it is faster to use a Memcached mock than the real deal. Just place this in your initializer for your test environment:
+
+    $memcache = Cash::Mock.new
+    
+### Locks ###
+
+In most cases locks are unnecessary; the transactional Memcached client will take care locks for you automatically and guarantees that no deadlocks can occur. But for very complex distributed transactions, shared locks are necessary.
+
+    $lock.synchronize('lock_name') do
+      $memcache.set("key", "value")
+    end
+    
+### Local Cache ###
+
+Sometimes your code will request the same cache key twice in one request. You can avoid a round trip to the Memcached server by using a local, per-request cache. Add this to your initializer:
+
+  $local = Cash::Local.new($memcache)
+  $cache = Cash::Transactional.new($local, $lock)
+
+## Installation ##
+
+#### Step 0: Install MemCached
+
+#### Step 1: Get the GEM ####
+
+    % gem sources -a http://gems.github.com
+    % sudo gem install ngmoco-cache-money
+    
+#### Step 2: Configure MemCached.
+
+Place a YAML file in `config/memcached.yml` with contents like:
+
+    test:
+      ttl: 604800
+      namespace: ...
+      sessions: false
+      debug: false
+      servers: localhost:11211
+      cache_money: true
+
+    development: 
+       ....
+
+#### Step 3: `config/environment.rb` ####
+ config.gem "ngmoco-cache-money",
+   :lib => "cache_money",
+   :source => 'http://gems.github.com',
+   :version => '0.2.9'
+
+#### Step 4: Add indices to your ActiveRecord models ####
+
+Queries like `User.find(1)` will use the cache automatically. For more complex queries you must add indices on the attributes that you will query on. For example, a query like `User.find(:all, :conditions => {:name => 'bob'})` will require an index like:
+
+    class User < ActiveRecord::Base
+      index :name
+    end
+    
+For queries on multiple attributes, combination indexes are necessary. For example, `User.find(:all, :conditions => {:name => 'bob', :age => 26})`
+
+    class User < ActiveRecord::Base
+      index [:name, :age]
+    end
+
+#### Optional: Selectively cache specific models
+
+There may be times where you only want to cache some of your models instead of everything.
+
+In that case, you can omit the following from your `config/initializers/cache_money.rb`
+
+	class ActiveRecord::Base
+	  is_cached :repository => $cache
+	end
+		
+After that is removed, you can simple put this at the top of your models you wish to cache:
+
+	is_cached :repository => $cache
+
+Just make sure that you put that line before any of your index directives.
+
+## Version ##
+
+WARNING: This is currently a RELEASE CANDIDATE. A version of this code is in production use at Twitter but the extraction and refactoring process may have introduced bugs and/or performance problems. There are no known major defects at this point, but still.
+
+## Acknowledgments ##
+
+Thanks to
+
+ * Twitter for commissioning the development of this library and supporting the effort to open-source it.
+ * Sam Luckenbill for pairing with me on most of the hard stuff.
+ * Matthew and Chris for pairing a few days, offering useful feedback on the readability of the code, and the initial implementation of the Memcached mock.
+ * Evan Weaver for helping to reason-through software and testing strategies to deal with replication lag, and the initial implementation of the Memcached lock.

data/TODO

@@ -0,0 +1,17 @@
+TOP PRIORITY
+
+REFACTOR
+* Clarify terminology around cache/key/index, etc.
+
+INFRASTRUCTURE
+
+NEW FEATURES
+* transactional get multi isn't really multi
+
+BUGS
+* Handle append strategy (using add rather than set?) to avoid race condition
+
+MISSING TESTS:
+* missing tests for Klass.transaction do ... end
+* non "id" pks work but lack test coverage
+* expire_cache

data/UNSUPPORTED_FEATURES

@@ -0,0 +1,13 @@
+* does not work with :dependent => nullify because
+  def nullify_has_many_dependencies(record, reflection_name, association_class, primary_key_name, dependent_conditions)
+    association_class.update_all("#{primary_key_name} = NULL", dependent_conditions)
+  end
+  This does not trigger callbacks
+* update_all, delete, update_counter, increment_counter, decrement_counter, counter_caches in general - counter caches are replaced by this gem, bear that in mind.
+* attr_readonly - no technical obstacle, just not yet supported
+* attributes before typecast behave unpredictably - hard to support
+* ActiveRecord::Rollback is unsupported - the exception gets swallowed so there isn't an opportunity to rollback the cache transaction - not hard to support
+* printf style binds: :conditions => ["name = '%s'", "37signals!"] - not too hard to support
+* objects as attributes that are serialized. story.title = {:foo => :bar}; customer.balance = Money.new(...) - these could be coerced using Column#type_cast?
+
+With a lot of these features the issue is not technical but performance. Every special case costs some overhead.

data/config/environment.rb

@@ -0,0 +1,14 @@
+require 'rubygems'
+gem 'activesupport', '~> 2.3.0'
+gem 'activerecord', '~> 2.3.0'
+gem 'actionpack', '~> 2.3.0'
+gem 'rspec', '>= 1.3.0'
+
+require 'action_controller'
+require 'active_record'
+require 'active_record/session_store'
+
+ActiveRecord::Base.establish_connection(
+  :adapter => 'sqlite3',
+  :database => ':memory:'
+)

data/config/memcached.yml

@@ -0,0 +1,6 @@
+test:
+  ttl: 604800
+  namespace: cache
+  sessions: false
+  debug: false
+  servers: localhost:11211

data/db/schema.rb

@@ -0,0 +1,18 @@
+ActiveRecord::Schema.define(:version => 2) do
+  create_table "stories", :force => true do |t|
+    t.string "title", "subtitle"
+    t.string  "type"
+    t.boolean "published"
+  end
+
+  create_table "characters", :force => true do |t|
+    t.integer "story_id"
+    t.string "name"
+  end
+  
+  create_table :sessions, :force => true do |t|
+    t.string :session_id
+    t.text   :data
+    t.timestamps
+  end
+end

data/init.rb

@@ -0,0 +1 @@
+require File.join(File.basedir(__FILE__),'rails','init')

data/lib/cache_money.rb

@@ -0,0 +1,84 @@
+require 'active_support'
+require 'active_record'
+
+require 'cash/lock'
+require 'cash/transactional'
+require 'cash/write_through'
+require 'cash/finders'
+require 'cash/buffered'
+require 'cash/index'
+require 'cash/config'
+require 'cash/accessor'
+
+require 'cash/request'
+require 'cash/fake'
+require 'cash/local'
+
+require 'cash/query/abstract'
+require 'cash/query/select'
+require 'cash/query/primary_key'
+require 'cash/query/calculation'
+
+require 'cash/util/array'
+require 'cash/util/marshal'
+
+class ActiveRecord::Base
+  def self.is_cached(options = {})
+    if options == false
+      include NoCash
+    else
+      options.assert_valid_keys(:ttl, :repository, :version)
+      include Cash unless ancestors.include?(Cash)
+      Cash::Config.create(self, options)
+    end
+  end
+
+  def <=>(other)
+    if self.id == other.id then 
+      0
+    else
+      self.id < other.id ? -1 : 1
+    end
+  end
+end
+
+module Cash
+  def self.included(active_record_class)
+    active_record_class.class_eval do
+      include Config, Accessor, WriteThrough, Finders
+      extend ClassMethods
+    end
+  end
+
+  module ClassMethods
+    def self.extended(active_record_class)
+      class << active_record_class
+        alias_method_chain :transaction, :cache_transaction
+      end
+    end
+
+    def transaction_with_cache_transaction(&block)
+      if cache_config
+        repository.transaction { transaction_without_cache_transaction(&block) }
+      else
+        transaction_without_cache_transaction(&block)
+      end
+    end
+
+    def cacheable?(*args)
+      true
+    end
+  end
+end
+module NoCash
+  def self.included(active_record_class)
+    active_record_class.class_eval do
+      extend ClassMethods
+    end
+  end
+  module ClassMethods
+    def cachable?(*args)
+      false
+    end
+  end
+end

data/lib/cash/accessor.rb

@@ -0,0 +1,83 @@
+module Cash
+  module Accessor
+    def self.included(a_module)
+      a_module.module_eval do
+        extend ClassMethods
+        include InstanceMethods
+      end
+    end
+
+    module ClassMethods
+      def fetch(keys, options = {}, &block)
+        case keys
+        when Array
+          return {} if keys.empty?
+          
+          keys = keys.collect { |key| cache_key(key) }
+          hits = repository.get_multi(*keys)
+          if (missed_keys = keys - hits.keys).any?
+            missed_values = block.call(missed_keys)
+            hits.merge!(missed_keys.zip(Array(missed_values)).to_hash_without_nils)
+          end
+          hits
+        else
+          repository.get(cache_key(keys), options[:raw]) || (block ? block.call : nil)
+        end
+      end
+
+      def get(keys, options = {}, &block)
+        case keys
+        when Array
+          fetch(keys, options, &block)
+        else
+          fetch(keys, options) do
+            if block_given?
+              add(keys, result = yield(keys), options)
+              result
+            end
+          end
+        end
+      end
+
+      def add(key, value, options = {})
+        if repository.add(cache_key(key), value, options[:ttl] || cache_config.ttl, options[:raw]) == "NOT_STORED\r\n"
+          yield if block_given?
+        end
+      end
+
+      def set(key, value, options = {})
+        repository.set(cache_key(key), value, options[:ttl] || cache_config.ttl, options[:raw])
+      end
+
+      def incr(key, delta = 1, ttl = nil)
+        ttl ||= cache_config.ttl
+        repository.incr(cache_key = cache_key(key), delta) || begin
+          repository.add(cache_key, (result = yield).to_s, ttl, true) { repository.incr(cache_key) }
+          result
+        end
+      end
+
+      def decr(key, delta = 1, ttl = nil)
+        ttl ||= cache_config.ttl
+        repository.decr(cache_key = cache_key(key), delta) || begin
+          repository.add(cache_key, (result = yield).to_s, ttl, true) { repository.decr(cache_key) }
+          result
+        end
+      end
+
+      def expire(key)
+        repository.delete(cache_key(key))
+      end
+
+      def cache_key(key)
+        "#{name}:#{cache_config.version}/#{key.to_s.gsub(' ', '+')}"
+      end
+    end
+
+    module InstanceMethods
+      def expire
+        self.class.expire(id)
+      end
+    end
+  end
+end