(I personally prefer writing one string in one line no matter how long it is, though)

Passing a class to `has_and_belongs_to_many` show deprecation message
three times. It is enough only once.

The idea of `class_name` as an option of reflection is that passing a
string would allow us to lazy autoload the class.

Using `belongs_to :client, class_name: Customer` is eagerloading models more than necessary
and creating possible circular dependencies.

Avoids a NoMethodError when table_name is a symbol instead of a string.

In Rails 4.2 calling `ActiveRecord::Migrator.migrate` would return an array of results. Without realizing that this return type was expected I accidentally introduced a change in https://github.com/rails/rails/commit/4d60e93174a3d6d90b1a06fc7515cb5cd749a6f3

This PR preserves the previous behavior and adds a test on the return type. This will need a backport to 5.0 branch.

Actually, private methods cannot be called with `self.`, so it's not just redundant, it's a bad habit in Ruby

Need `table_name` to avoid all databases scan.
See https://dev.mysql.com/doc/refman/5.7/en/information-schema-optimization.html.

```
> EXPLAIN SELECT fk.referenced_table_name AS 'to_table', fk.referenced_column_name AS 'primary_key', fk.column_name AS 'column', fk.constraint_name AS 'name', rc.update_rule AS 'on_update', rc.delete_rule AS 'on_delete' FROM information_schema.key_column_usage fk JOIN information_schema.referential_constraints rc USING (constraint_schema, constraint_name) WHERE fk.referenced_column_name IS NOT NULL AND fk.table_schema = 'activerecord_unittest' AND fk.table_name = 'fk_test_has_pk' AND rc.table_name = 'fk_test_has_pk'\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: fk
   partitions: NULL
         type: ALL
possible_keys: NULL
          key: TABLE_SCHEMA,TABLE_NAME
      key_len: NULL
          ref: NULL
         rows: NULL
     filtered: NULL
        Extra: Using where; Open_full_table; Scanned 0 databases
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: rc
   partitions: NULL
         type: ALL
possible_keys: NULL
          key: TABLE_NAME
      key_len: NULL
          ref: NULL
         rows: NULL
     filtered: NULL
        Extra: Using where; Open_full_table; Scanned 1 database; Using join buffer (Block Nested Loop)
2 rows in set, 1 warning (0.00 sec)
```

Fixes #27579.

https://github.com/rails/rails/issues/26834

This change raises an error if a has_many through association
is defined before the through association.

The alternative of #26213.

Currently `find_by` and `where` with AR object return inconsistent
result. This is caused by statement cache does not support AR object.
Passing to finder method to fix the issue.

Fixes #26210.

Revert "Merge pull request #21233 from mtsmfm/disable-referential-integrity-without-superuser-privileges"

This reverts commit eeac6151, reversing
changes made to 5c40239d.

Reason: Broke the isolated tests.
https://travis-ci.org/rails/rails/builds/188721346

Passing `name` to `tables` is already deprecated at #21601.
Passing `name` to `indexes` is also unused.

Related #22973, #24708.

`select_all`, `select_one`, `select_value`, and `select_values` method
signature is `(arel, name = nil, binds = [])`.
But `select_rows` is `(sql, name = nil, binds = [])`.

Per the regression commit below, the commit changes the behavior of
`#changed?`to consult the `#changed_in_place?` method on `Type::Value` classes.
Per this change, `PostgreSQL::OID::Hstore` needs to override this method
in order to compare the deserialized forms of the two arguments. In
Ruby, two hashes are considered equal even if their key order is
different. This commit helps to bring that behavior to `Hstore` values.

Fixes regression introduced by 8e633e50

Fixes #27502

Closes #27555.

[ Ben A. Morgan & Kasper Timm Hansen ]

find and exists?

In f1a0fa9e we moved backend specific timestamp behavior out of the type
and into the adapter. This was in line with our general attempt to
reduce the number of adapter specific type subclasses. However, on PG,
the array type performs all serialization, including database encoding
in its serialize method.

This means that we have converted the value into a string before
reaching the database, so no adapter specific logic can be applied (and
this also means that timestamp arrays were using the default `.to_s`
method on the given object, which likely meant timestamps were being
ignored in certain cases as well)

Ultimately I want to do a more in depth refactoring which separates
database serializer objects from the active model type objects, to give
us a less awkward API for introducing the attributes API onto Active
Model.

However, in the short term, we follow the solution we've applied
elsewhere for this. Move behavior off of the type and into the adapter,
and use a data object to allow the type to communicate information up
the stack.

Fixes #27514.

We faced a significant performance decrease when we started using STI
without storing full namespaced class name in type column (because of PostgreSQL
length limit for ENUM types).
We realized that the cause of it is the slow STI model instantiation. Problematic
method appears to be `ActiveRecord::Base.compute_type`, which is used to find
the right class for STI model on every instantiation.
It builds an array of candidate types and then iterates through it calling
`safe_constantize` on every type until it finds appropriate constant. So if
desired type isn't the first element in this array there will be at least one
unsuccessful call to `safe_constantize`, which is very expensive, since it's
defined in terms of `begin; rescue; end`.

This commit is an attempt to speed up `compute_type` method simply by caching
results of previous calls.

```ruby
class MyCompany::MyApp::Business::Accounts::Base < ApplicationRecord
  self.table_name = 'accounts'
  self.store_full_sti_class = false
end

class MyCompany::MyApp::Business::Accounts::Free < Base
end

class MyCompany::MyApp::Business::Accounts::Standard < Base
  # patch .compute_type there
end

puts '======================= .compute_type ======================='
Benchmark.ips do |x|
  x.report("original method") do
    MyCompany::MyApp::Business::Accounts::Free.send :compute_type, 'Free'
  end
  x.report("with types cached") do
    MyCompany::MyApp::Business::Accounts::Standard.send :compute_type, 'Standard'
  end
  x.compare!
end
```

```
======================= .compute_type =======================
  with types cached:  1529019.4 i/s
    original method:     2850.2 i/s - 536.46x  slower
```

```ruby
5_000.times do |i|
  MyCompany::MyApp::Business::Accounts::Standard.create!(name: "standard_#{i}")
end

5_000.times do |i|
  MyCompany::MyApp::Business::Accounts::Free.create!(name: "free_#{i}")
end

puts '====================== .limit(100).to_a ======================='
Benchmark.ips do |x|
  x.report("without .compute_type patch") do
    MyCompany::MyApp::Business::Accounts::Free.limit(100).to_a
  end
  x.report("with .compute_type patch") do
    MyCompany::MyApp::Business::Accounts::Standard.limit(100).to_a
  end
  x.compare!
end
```

```
====================== .limit(100).to_a =======================
     with .compute_type patch:      360.5 i/s
  without .compute_type patch:       24.7 i/s - 14.59x  slower
```

`current_time_from_proper_timezone` and timestamp attributes methods was
pushed up to class method.

Actually these methods don't need instantiation.

Currently does not work the following example in the doc:

```ruby
  # For the Posts with id of 10 and 15, increment the comment_count by 1
  # and update the updated_at value for each counter.
  Post.update_counters [10, 15], comment_count: 1, touch: true
  # Executes the following SQL:
  # UPDATE posts
  #    SET comment_count = COALESCE(comment_count, 0) + 1,
  #    `updated_at` = '2016-10-13T09:59:23-05:00'
  #  WHERE id IN (10, 15)
```

* Rename update -> touch to remain consistent with the other docs
  language of "touch"'ing.
* Remove the sentence that's repeated from just above and rephrase.

Today `rake db:schema:cache:dump` only supports dumping cache for a
single connection (`ActiveRecord::Base.connection`). This doesn't work
for apps with multiple databases.

This PR makes `DatabaseTasks` to provide an API for dumping schema cache
for any connection.

`touch_updates` calls `Time.now` via `current_time_from_proper_timezone`
so it is better to not invoke `touch_updates` if it is unnecessary.

* Refactor to use `touch_updates`

  Ensures we only call `current_time_from_proper_timezone` from one place.

* Clarify touch default in tests.

  Not interested in what happens when passed false but that
  nothing passed means no touching.

* Backdate touched columns in tests.

  We can't be sure a test progresses through time, so our
  touching code may be working correctly but the test
  itself is brittle.

  Fix by backdating that's further in the past akin to
  what the timestamps tests do:
  https://github.com/rails/rails/blob/d753645d40e925973724e4c3a8617b654da90e41/activerecord/test/cases/timestamp_test.rb#L17

* Expand changelog entry.

  Elaborate and show examples.

Closes #26995.

[ Jarred Trost & Kasper Timm Hansen ]

As pointed out by @matthewd this change makes ImmutableString aware
of MysqlString's existence whereas previously MysqlString was only
overriding public API.

cc @kamipo

This reverts commit e632c2fa, reversing
changes made to 334a7dcf.

Removes space from the word; is now spelled in the standard way.

[ci skip]

`timestamp_attributes_for_{create,update}` is defined as symbol but
always used as string with `to_s`. This allocates extra strings. To
avoid extra allocation, change the definitions from symbol to string.

```ruby
pp ObjectSpace::AllocationTracer.trace {
  1_000.times { |i|
    Post.create!
  }
}
```

Before:

```
["~/rails/activerecord/lib/active_record/timestamp.rb", 121]=>[1002, 0, 750, 0, 1, 18528],
["~/rails/activerecord/lib/active_record/timestamp.rb", 105]=>[1002, 0, 750, 0, 1, 7720],
["~/rails/activerecord/lib/active_record/timestamp.rb", 101]=>[1002, 0, 750, 0, 1, 7720],
["~/rails/activerecord/lib/active_record/timestamp.rb", 109]=>[1002, 0, 750, 0, 1, 13896],
["~/rails/activerecord/lib/active_record/timestamp.rb", 61]=>[4008, 0, 3000, 0, 1, 30880],
```

After:

```
["~/rails/activerecord/lib/active_record/timestamp.rb", 120]=>[1000, 0, 756, 0, 1, 17184],
["~/rails/activerecord/lib/active_record/timestamp.rb", 104]=>[1000, 0, 756, 0, 1, 7160],
["~/rails/activerecord/lib/active_record/timestamp.rb", 100]=>[1000, 0, 756, 0, 1, 7160],
["~/rails/activerecord/lib/active_record/timestamp.rb", 108]=>[1000, 0, 756, 0, 1, 12888],
```

Simply we can do `target.pluck(reflection.association_primary_key)` if
`target` is loaded.