翻译 Rails (3&4) 预加载 (preload) 的 3 种方式

qinfanpeng · 2015年12月02日 · 最后由 qinfanpeng 回复于 2015年12月05日 · 6548 次阅读

Rails（3&4）预加载（preload）的 3 种方式

作者：Robert Pankowecki 原文地址：http://blog.arkency.com/2013/12/rails4-preloading/

本文是 2013 年的一篇老文章，资深人士请绕道。前些天自己遇到了些麻烦，详见：https://ruby-china.org/topics/28225 ，本来当初也看到过这篇文章，只因自己粗心，而没及时深读本文（毕竟对我而言英文没有中文那么直观），否则将节约我不少时间。翻译此文，希望能节约他人的几秒钟。

#includes_#preload_#eager_load

和 Rails 的 ActiveRecord 打交道，你大概已经习惯用#includes来预加载数据。但你知道为啥它生成的 SQL 有时小而美，有时却是个巨型查询，且每个表、列都被重命名过（"users"."id" AS t0_r0，"addresses"."id" AS t1_r0）？你知道##preload和#eager_load 也能帮你做这事儿吗？你知道 Rails4 中预加载方面有哪些变化吗？如果都不知道的话，请耐心读下去。本文不会太长，将助你了解预加载中哪些你不熟悉的方面。

首先让我们从 Active Record 关联关系（associations）定义开始，本文通篇将以此例为基础。

class User < ActiveRecord::Base
  has_many :addresses
end

class Address < ActiveRecord::Base
  belongs_to :user
end

下面是些实验数据，以便我们验证查询结果：

rob = User.create!(name: "Robert Pankowecki", email: "[email protected]")
bob = User.create!(name: "Bob Doe", email: "[email protected]")

rob.addresses.create!(country: "Poland", city: "Wrocław", postal_code: "55-555", street: "Rynek")
rob.addresses.create!(country: "France", city: "Paris", postal_code: "75008", street: "8 rue Chambiges")
bob.addresses.create!(country: "Germany", city: "Berlin", postal_code: "10551", street: "Tiergarten")

Rails 3

预加载数据时，通常你都会选择#includes，大概由于从 Rails 2 甚至 Rails 1 开始，就推荐使用该方法了。它会魔法般地生成两个查询：

User.includes(:addresses)
#  SELECT "users".* FROM "users" 
#  SELECT "addresses".* FROM "addresses" WHERE "addresses"."user_id" IN (1, 2)

另外两个方法呢？我们先来看它们的实际表现。

User.preload(:addresses)
#  SELECT "users".* FROM "users" 
#  SELECT "addresses".* FROM "addresses" WHERE "addresses"."user_id" IN (1, 2)

到底是#preload类似于#includes呢？还是#includes类似于#preload？请继续阅读。

#eager_load行为如下：

User.eager_load(:addresses)

#  SELECT
#  "users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4, 
#  "addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7 
#  FROM "users" 
#  LEFT OUTER JOIN "addresses" ON "addresses"."user_id" = "users"."id"

结果完全不一样，对吧。问题的关键在于 Rails 有两种方式进行数据预加载：一种通过单独的（separate）查询去加载其他表的数据；另一种只用单个查询 (带left join) 一次性加载所有数据。

用#preload，就意味着你希望总是用单独的（separate）查询；用#eager_load，表明你用的是单个查询（一次性加载所有数据）。那么#includes呢？它会帮你选择从上面选择一种方式，即你把决定权交给了 Rails。那么它是根据什么决定的呢？你可能会问，是基于查询条件 (query conditions)。我们先来看这样一个例子：#includes把查询代理给了#eager_load，所以只会有一个巨型查询产生。

# 这两个用法是等效的
 User.includes(:addresses).where("addresses.country = ?", "Poland")
 User.eager_load(:addresses).where("addresses.country = ?", "Poland")

# SELECT 
# "users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4,
# "addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7 
# FROM "users"
# LEFT OUTER JOIN "addresses" 
# ON "addresses"."user_id" = "users"."id" 
# WHERE (addresses.country = 'Poland')

本例中 Rails 发现where语句用到了预加载表（addresses）中的列（country），故而将#includes代理给了#eager_load。你也可直接调用#eager_load,来到稳定实现这个效果。

如果显示调用#preload会怎样呢？

User.preload(:addresses).where("addresses.country = ?", "Poland")
#  SELECT "users".* FROM "users" WHERE (addresses.country = 'Poland')
#
#  SQLite3::SQLException: no such column: addresses.country

由于没将users表和addresses表关联（join）起来，故而抛出了异常。

目标实现了吗

如果你再看下这个例子

User.includes(:addresses).where("addresses.country = ?", "Poland")

你可能会问：这些代码初衷是啥？作者这样写想表达啥？我们将用简洁的 Rails 代码来达成以下目标：

1. 找出带波兰地址的用户，且只预加载其波兰地址； 2. 找出带波兰地址的用户，但预加载其所有地址； 3. 找出所有的用户，但只预加载其波兰地址。

你知道我们已实现了其中哪一个目标吗？第一个。下面来看看能否实现后面两个。

`#preload`有啥优势

当前任务：找出带波兰地址的用户，但预加载其所有地址。即是需要找出那些至少有一个波兰地址的用户，及其所有地址。

我们只需要那些有波兰地址的用户，这很简单：User.joins(:addresses).where("addresses.country = ?", "Poland")；要预加载地址，再加上includes(:addresses)就够了嘛，对吧？

r = User.joins(:addresses).where("addresses.country = ?", "Poland").includes(:addresses)

r[0]
#=> #<User id: 1, name: "Robert Pankowecki", email: "[email protected]", created_at: "2013-12-08 11:26:24", updated_at: "2013-12-08 11:26:24"> 

r[0].addresses
# [
#   #<Address id: 1, user_id: 1, country: "Poland", street: "Rynek", postal_code: "55-555", city: "Wrocław", created_at: "2013-12-08 11:26:50", updated_at: "2013-12-08 11:26:50">
# ]

然而，事实并非我们所想的那样：结果中少了了第二个地址，此处它本该出现的。Rails 再次探测到 Where语句使用了预加载了的表 (addresses)，因而底层还是用#eager_load来实现#includes的。唯一的区别是这里使用的是INNER JOIN而非前面的LEFT JOIN，但查询结果并非没啥区别。

SELECT 
"users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4,
"addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7 
FROM "users"
INNER JOIN "addresses" 
ON "addresses"."user_id" = "users"."id" 
WHERE (addresses.country = 'Poland')

难得有机会比 Rails 还“聪明”，我们可以显示地调用#preload而非#includes来表明我们的意图：

r = User.joins(:addresses).where("addresses.country = ?", "Poland").preload(:addresses)

# SELECT "users".* FROM "users"
# INNER JOIN "addresses" ON "addresses"."user_id" = "users"."id" 
# WHERE (addresses.country = 'Poland')

# SELECT "addresses".* FROM "addresses" WHERE "addresses"."user_id" IN (1)

r[0] 
# [#<User id: 1, name: "Robert Pankowecki", email: "[email protected]", created_at: "2013-12-08 11:26:24", updated_at: "2013-12-08 11:26:24">] 

r[0].addresses
# [
#  <Address id: 1, user_id: 1, country: "Poland", street: "Rynek", postal_code: "55-555", city: "Wrocław", created_at: "2013-12-08 11:26:50", updated_at: "2013-12-08 11:26:50">,
#  <Address id: 3, user_id: 1, country: "France", street: "8 rue Chambiges", postal_code: "75008", city: "Paris", created_at: "2013-12-08 11:36:30", updated_at: "2013-12-08 11:36:30">] 
# ]

这正是我们想要的。借助#preload，我们不必再将 user 查询条件和数据预加载混淆在一起了，查询再次变得自然而简洁。

预加载关联数据子集 (Preloading subset of association)

当前任务：查找所有所有用户，但只预加载其波兰地址。

说实话，我不喜欢仅预加载关联数据子集，因为程序其他部分还是可能会误认为关联数据已经全加载了（基于该假设做出的决策，自然也是错的）。若获取关联数据子集用于展示，倒还说得过去。

我更倾向于直接在关联上加条件：

class User < ActiveRecord::Base
  has_many :addresses
  has_many :polish_addresses, conditions: {country: "Poland"}, class_name: "Address"
end

接着，我们便可显示预加载关联数据了：

r = User.preload(:polish_addresses)

# SELECT "users".* FROM "users" 
# SELECT "addresses".* FROM "addresses" WHERE "addresses"."country" = 'Poland' AND "addresses"."user_id" IN (1, 2)

r

# [
#   <User id: 1, name: "Robert Pankowecki", email: "[email protected]", created_at: "2013-12-08 11:26:24", updated_at: "2013-12-08 11:26:24">
#   <User id: 2, name: "Bob Doe", email: "[email protected]", created_at: "2013-12-08 11:26:25", updated_at: "2013-12-08 11:26:25">
# ] 

r[0].polish_addresses

# [
#   #<Address id: 1, user_id: 1, country: "Poland", street: "Rynek", postal_code: "55-555", city: "Wrocław", created_at: "2013-12-08 11:26:50", updated_at: "2013-12-08 11:26:50">
# ] 

r[1].polish_addresses
# []

或者：

r = User.eager_load(:polish_addresses)

# SELECT "users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4, 
#        "addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7
# FROM "users" 
# LEFT OUTER JOIN "addresses" 
# ON "addresses"."user_id" = "users"."id" AND "addresses"."country" = 'Poland'

r
# [
#   #<User id: 1, name: "Robert Pankowecki", email: "[email protected]", created_at: "2013-12-08 11:26:24", updated_at: "2013-12-08 11:26:24">,
#   #<User id: 2, name: "Bob Doe", email: "[email protected]", created_at: "2013-12-08 11:26:25", updated_at: "2013-12-08 11:26:25">
# ]

r[0].polish_addresses
# [
#   #<Address id: 1, user_id: 1, country: "Poland", street: "Rynek", postal_code: "55-555", city: "Wrocław", created_at: "2013-12-08 11:26:50", updated_at: "2013-12-08 11:26:50">
# ]

r[1].polish_addresses
# []

换做是我们，只在运行时才知道要应用的关联条件，我们该怎么做？实话说我不知道，如果你知道，请在评论中告知。

最后一个目标

你可能会问：区区一个预加载为啥就这么麻烦？我也不清楚，不过我想大部分ORM初衷都是生成单条查询、从单张表中加载数据。有了预加载，情况就复杂多了：它致使我们用多条件的查询从多张表中去加载多类数据。Rails 中，我们采用链式 API 来构建多个查询 (比如#preload)。

若问我想要什么样的 API？类似下面这样的：

User.joins(:addresses).where("addresses.country = ?", "Poland").preload do |users|
  users.preload(:addresses).where("addresses.country = ?", "Germany")
  users.preload(:lists) do |lists|
    lists.preload(:tasks).where("tasks.state = ?", "unfinished")
  end
end

我希望你能明白我的心思，但这仅是理想，还是让我们回到现实中来吧。

Rails 4 中的变化

现在我们来说说 Rails 4 中有关预加载的变化。

class User < ActiveRecord::Base
  has_many :addresses
  has_many :polish_addresses, -> {where(country: "Poland")}, class_name: "Address"
end

现在 Rails 推荐你使用新的lambda语法来定义关联条件。这个改进相当棒，因为我不止一次看到这样的错误：关联条件相关代码只在类加载时被解释执行了一次 (关联中含有 Date.today 的代码就会出错)。

Rails 同样推荐你使用lambda语法或method语法去定义 scope。

# 差，Time.now 会一直停留在class加载的时刻（相对于成了一个“死”值了，而非动态变化的）
# 并且在开发环境中，这个bug还不容易发现，因为一旦类有修改，开发环境便会自动重新加载代码（掩盖了“死”值的真相）
scope :from_the_past, where("happens_at <= ?", Time.now)

# 好
scope :from_the_past, -> { where("happens_at <= ?", Time.now) }

# 好
def self.from_the_past
  where("happens_at <= ?", Time.now)
end

无论是动态解释，还是只在一开始解释一次，我们的条件where(country: "Poland")都是一样的。但 Rails 尝试统一两种场景（关联、scope）的语法总是好的，从而避免我们再犯类似的错误。

说完了语法上的变化，下面来看看行为上的改变。

User.includes(:addresses)
#  SELECT"users".* FROM "users"
#  SELECT "addresses".* FROM "addresses" WHERE "addresses"."user_id" IN (1, 2)

User.preload(:addresses)
#  SELECT "users".* FROM "users"
#  SELECT "addresses".* FROM "addresses" WHERE "addresses"."user_id" IN (1, 2)

User.eager_load(:addresses)

#  SELECT "users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4,
#         "addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7
#  FROM "users"
#  LEFT OUTER JOIN "addresses"
#  ON "addresses"."user_id" = "users"."id"

好吧，看起来还是一样，这倒不奇怪。试着加上先前让我们麻烦不断的查询条件看看：

  User.includes(:addresses).where("addresses.country = ?", "Poland")
 # 
  #弃用警告: 貌似你预加载了SQL字符串代码片段中引用的（referenced）表（users或addresses），例如：
  #
  #    Post.includes(:comments).where("comments.title = 'foo'")
  #
  # 目前，Active Record识别出来了字符串中的表（comments），并且知道将其JOIN到整个大查询中，而非单独弄一条查询出来。然而，没有成熟的SQL解析器支持的情况下，这样做是有天然缺陷的。显然我们不会实现SQL解析器，所以将移除此特性。从今以后，引用（reference）字符串中的表时，你得显示告诉Active Record：
  #
  #   Post.includes(:comments).where("comments.title = 'foo'").references(:comments)
  # 
  # 如果你用不着这个隐式join引用表的特性，你可以通过如下设置来完全禁用它：`config.active_record.disable_implicit_join_references = true`.
#
  # SELECT "users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4,
  #        "addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7
  # FROM "users" 
  # LEFT OUTER JOIN "addresses" ON "addresses"."user_id" = "users"."id" 
  # WHERE (addresses.country = 'Poland')

咦，看起来真啰嗦！不过建议你还是仔细读一下，因为它把其中情理解释得很清楚。

换句话说，Rails 不愿像以前那样聪明了——即不再会通过探测where条件来决定用哪一种查询策略了，它需要我们的帮助。我们得告诉它某张表有条件查询，就像这样：

User.includes(:addresses).where("addresses.country = ?", "Poland").references(:addresses)

我很好奇如果我们尝试预加载多张表，但只显示引用（reference）了其中一张表会怎样？

User.includes(:addresses, :places).where("addresses.country = ?", "Poland").references(:addresses)

#  SELECT "users"."id" AS t0_r0, "users"."name" AS t0_r1, "users"."email" AS t0_r2, "users"."created_at" AS t0_r3, "users"."updated_at" AS t0_r4, 
#         "addresses"."id" AS t1_r0, "addresses"."user_id" AS t1_r1, "addresses"."country" AS t1_r2, "addresses"."street" AS t1_r3, "addresses"."postal_code" AS t1_r4, "addresses"."city" AS t1_r5, "addresses"."created_at" AS t1_r6, "addresses"."updated_at" AS t1_r7,
#         "places"."id" AS t2_r0, "places"."user_id" AS t2_r1, "places"."name" AS t2_r2, "places"."created_at" AS t2_r3, "places"."updated_at" AS t2_r4 
#  FROM "users" 
#  LEFT OUTER JOIN "addresses" ON "addresses"."user_id" = "users"."id" 
#  LEFT OUTER JOIN "places" ON "places"."user_id" = "users"."id" 
#  WHERE (addresses.country = 'Poland')

我们本以为它会用#eager_load策略（通过LEFT JOIN）去加载addresses，用#preload策略（单独的查询）去加载places。但是你也看到了，事实并非如此。可能以后这个行为会改变。

如果显示使用#eager_load去预加载数据，Rails4 是不会警告你加#references的，生成的查询也一样。

User.eager_load(:addresses).where("addresses.country = ?", "Poland")

换言之，下面两种写法是等效的：

User.includes(:addresses).where("addresses.country = ?", "Poland").references(:addresses)
User.eager_load(:addresses).where("addresses.country = ?", "Poland")

对于#preload，行为还是一样（和 Rails4 以前一样）。

User.preload(:addresses).where("addresses.country = ?", "Poland")
#  SELECT "users".* FROM "users" WHERE (addresses.country = 'Poland')
#
#  SQLite3::SQLException: no such column: addresses.country: SELECT "users".* FROM "users"  WHERE (addresses.country = 'Poland')

前面提到的其他查询，在 Rails4 中表现也是一样的：

# 找出带波兰地址的用户，但预加载其所有地址；
User.joins(:addresses).where("addresses.country = ?", "Poland").preload(:addresses)

#找出所有的用户，但只预加载其波兰地址；
User.preload(:polish_addresses)