Identity Map in Rails3.1

Identity Map是Rails3.1的又一个新特性。
一、什么是Identity Map

下面的代码是ActiveRecord在未开启Identify Map时的表现：

>> user = User.first
  User Load (0.3ms)  SELECT "users".* FROM "users" LIMIT 1
>> same_user = User.find 1
  User Load (11.3ms)  SELECT "users".* FROM "users" WHERE "users"."id" = ? LIMIT 1  [["id", 1]]
>> user == same_user
=> true
>> user.object_id == same_user.object_id
=> false

上面的user和same_user是都是id为1的User对象，但是却是两个对象。

下面开启Identity Map：

>> ActiveRecord::IdentityMap.enabled = true
>> user = User.first
  User Load (0.4ms)  SELECT "users".* FROM "users" LIMIT 1
>> same_user = User.find 1
  User Loaded  From Identity Map (id: 1)
>> user == same_user
=> true
>> user.object_id == same_user.object_id
=> true

当ActiveRecord开启了Identity Map后，每一个model会维持一个Map，每次实例化对象之前会通过主键去这个model对应的Map里找到对应的对象。如果存在，直接从Map里返回这个对象，否则才进行查询，实例化对象后放到Map里。
二、和Query Cache/Cache Money的差别
听起来和ActiveRecord的query cache或是cache money类似。实际上差别蛮大的：
从减少db查询方面和query cache一样，缓存池是在一次请求结束后销毁。各个请求之间不共享，而cache money是全局的，缓存结果可以被持续利用。
query cache是根据sql来做主键，而identity map和cache money是根据model的主键。
Identity Map还具有减少内存使用和保持一致性的作用。

三、Identity Map的优点
1.一致性：一个model对象保持完全唯一的引用，避免了更新同一个对象时候产生的不一致现象。
下面的例子是未开启Identity Map时，同一条记录的num被两次增加1，但是结果只增加了1.造成最终结果和实际不一致。而开启了Identity Map的结果最终是正确的。

>> user = User.first
  User Load (0.3ms)  SELECT "users".* FROM "users" LIMIT 1
>> same_user = User.find 1
  User Load (12.2ms)  SELECT "users".* FROM "users" WHERE "users"."id" = ? LIMIT 1  [["id", 1]]
>> user.num = user.num + 1
=> 1
>> user.save
   (0.9ms)  UPDATE "users" SET "updated_at" = '2011-06-12 09:46:32.722893', "num" = 1 WHERE "users"."id" = 1
=> true
>> same_user.num = same_user.num + 1
=> 1
>> same_user.save
   (0.4ms)  UPDATE "users" SET "updated_at" = '2011-06-12 09:46:53.093752', "num" = 1 WHERE "users"."id" = 1
=> true
>> user.reload
=> #<User id: 1, name: nil, num: 1, created_at: "2011-06-12 03:00:30", updated_at: "2011-06-12 09:46:53">#这里结果不一致了
>> ActiveRecord::IdentityMap.enabled = true
=> true
>> same_user = User.find 1
  User Load (0.2ms)  SELECT "users".* FROM "users" WHERE "users"."id" = ? LIMIT 1  [["id", 1]]
=> #<User id: 1, name: nil, num: 1, created_at: "2011-06-12 03:00:30", updated_at: "2011-06-12 09:46:53">
>> user = User.first
  User Load (0.3ms)  SELECT "users".* FROM "users" LIMIT 1
  User Loaded  From Identity Map (id: 1)
=> #<User id: 1, name: nil, num: 1, created_at: "2011-06-12 03:00:30", updated_at: "2011-06-12 09:46:53">
>> same_user.num = same_user.num + 1
=> 2
>> same_user.save
   (0.4ms)  UPDATE "users" SET "updated_at" = '2011-06-12 09:51:59.203676', "num" = 2 WHERE "users"."id" = 1
>> user.num = user.num + 1
=> 3
>> user.save
   (0.5ms)  UPDATE "users" SET "updated_at" = '2011-06-12 09:52:21.643577', "num" = 3 WHERE "users"."id" = 1 #加了两次1结果为3，正确！

2.节省内存：不会重复创建相同的对象。
3.提高速度：和ActiveRecord的query cache一样，减少db hits

四、存在的问题
目前默认还是未开启的。因为最近发现的在处理关联时的问题：

  # Active Record Identity Map does not track associations yet. For example:
  #   comment = @post.comments.first
  #   comment.post = nil
  #   @post.comments.include?(comment) #=> true
  #
  # Ideally, the example above would return false, removing the comment object from the
  # post association when the association is nullified. This may cause side effects, as
  # in the situation below, if Identity Map is enabled:
  #
  #   Post.has_many :comments, :dependent => :destroy
  #
  #   comment = @post.comments.first
  #   comment.post = nil
  #   comment.save
  #   Post.destroy(@post.id)
  #
  # Without using Identity Map, the code above will destroy the @post object leaving
  # the comment object intact. However, once we enable Identity Map, the post loaded
  # by Post.destroy is exactly the same object as the object @post. As the object @post
  # still has the comment object in @post.comments, once Identity Map is enabled, the
  # comment object will be accidently removed.
  #
  # This inconsistency is meant to be fixed in future Rails releases.

ps：rails 3.1的rails console默认可以打印sql日志，不需要hack irbrc了，cool～

期待３.１！
希望rails早日摘掉“低性能”的帽子