`migration_keys` includes `name` but `name` is not a column option.

```ruby
  # Before
  t.index ["firm_id", "type", "rating"], name: "company_index", order: {"rating"=>:desc}, using: :btree

  # After
  t.index ["firm_id", "type", "rating"], name: "company_index", order: { rating: :desc }, using: :btree
```

Currently blank comment does not dump to `db/schema.rb`. But created it
even if specified blank.

Per discussion in pull request #26622:

"Let's change it to PG::Error. The more specific classes mentioned are
subclasses, and the fact the raised exception is a PG::UndefinedColumn
doesn't change the fact that it's also a PG::Error." - matthewd

This clarifies the object that +ActiveRecord::Base.connection.execute+
will return when using Postgresql.

[fixes #26441]

All indentation was normalized by rubocop auto-correct at 80e66cc4.
But comments was still kept absolute position. This commit aligns
comments with method definitions for consistency.

By doing `@connection = nil` that means that we need nil checks before it
is used anywhere, but we weren't doing those checks.  Instead, we get a
NoMethodError after using a connection after it fails to reconnect.

Neither of the other adapters set @connection to nil, just the mysql2
adapter. By just closing it, we avoid the need to check if we have a
connection object and it will produce an appropriate exception when used.

Recently, the Rails team made an effort to keep the source code consistent, using Ruboco
(bb1ecdcc and below). Some of the case
statements were missed.

This changes the case statements' formatting and is consistent with changes
in 810dff7c and db63406c.

There are some minor changes to the point type as I had forgotten that
this will affect the behavior of `t.point` in migrations and the schema
dumper so we need to handle those as well.

I'll say this again so I can convince myself to come up with a better
structure... TYPES SHOULD NOT CARE ABOUT SCHEMA DUMPING AND WE NEED TO
BETTER SEPARATE THESE.

This reverts commit 3a1f6fe7.

This commit takes the code in a direction that I am looking to avoid.
The predicate builder should be purely concerned with AST construction
as it matters to methods like `where`. Things like case sensitivity
should continue to be handled elsewhere.

I still think that this is something that should be handled in the pg
gem, but it's not going to end up happening there so we'll do it here
instead. Once we bump to pg 0.19 we can pass the encoding to the
`encode` method instead.

This issue occurs because C has no concept of encoding (or strings,
really). The bytes that we pass here when sending the value to the
database will always be interpreted as whatever encoding the connection
is currently configured to use. That means that roundtripping to the
database will lose no information

However, after assigning we round trip through our type system without
hitting the database. The only way that we can do the "correct" thin
here would be to actually give a reference to the connection to the
array type and have it check the current value of the connection's
encoding -- which I'm strongly opposed to. We could also pass in the
encoding when it's constructed, but since that can change independently
of the type I'm not a huge fan of that either.

This feels like a reasonable middle ground, where if we have an array of
strings we simply use the encoding of the string we're given.

Fixes #26326.

For reduce instantiating `Type::Value`.

Fixes #26137.

This reverts #23067. #23067 is for propagating `pk` value from
`sql_for_insert` to `exec_insert` (avoiding extra query for pg adapter).
Now `exec_insert` includes `sql_for_insert` since #26002 therefore this
propagating is no longer needed.

Strict mode controls how MySQL handles invalid or missing values in
data-change statements such as INSERT or UPDATE. If strict mode is not
in effect, MySQL inserts adjusted values for invalid or missing values
and produces warnings.

```ruby
  def test_mysql_not_null_defaults_non_strict
    using_strict(false) do
      with_mysql_not_null_table do |klass|
        record = klass.new
        assert_nil record.non_null_integer
        assert_nil record.non_null_string
        assert_nil record.non_null_text
        assert_nil record.non_null_blob

        record.save!
        record.reload

        assert_equal 0,  record.non_null_integer
        assert_equal "", record.non_null_string
        assert_equal "", record.non_null_text
        assert_equal "", record.non_null_blob
      end
    end
  end
```

It is inconsistent with other types that only text/blob defaults treated
as an empty string. This commit fixes the inconsistency.

invalid column type
    /activerecord/lib/active_record/connection_adapters
    /abstract/schema_definitions.rb:306
    type = type.to_sym

Changed to the following to handle nil case:
    type = type.to_sym if type

Added regression test for this case:
  /activerecord/test/cases/migration_test.rb:554
  if current_adapter?(:SQLite3Adapter)
    def test_allows_sqlite3_rollback_on_invalid_column_type
      Person.connection.create_table :something, force: true do |t|
        t.column :number, :integer
        t.column :name, :string
        t.column :foo, :bar
      end
      assert Person.connection.column_exists?(:something, :foo)
      assert_nothing_raised { Person.connection.remove_column :something, :foo, :bar }
      assert !Person.connection.column_exists?(:something, :foo)
      assert Person.connection.column_exists?(:something, :name)
      assert Person.connection.column_exists?(:something, :number)
    ensure
      Person.connection.drop_table :something, if_exists: true
    end
  end

Because `sql_for_insert` is only called in `use_insert_returning?` is
true since #26002.

Currently uniqueness validator is coupled with building Arel ASTs.
This commit extracts `PredicateBuilder::CaseSensitiveHandler` for
decouple the building Arel ASTs.

Style/SpaceBeforeBlockBraces
Style/SpaceInsideBlockBraces
Style/SpaceInsideHashLiteralBraces

Fix all violations in the repository.

Unfortunately `t.timestamps` in `create_table` allows unused splat args.
But the same one in `change_table` does not allow them.
This commit fixes the inconsistent behavior.

If you call `remove_index` with wrong options, say a type, like I did,
you get:

```
== 20160810072541 RemoveUniqueIndexOnGoals: migrating =========================
-- remove_index(:goal, {:coulmn=>:kid_id, :unique=>true})
rails aborted!
StandardError: An error has occurred, this and all later migrations canceled:

undefined method `ArgumentError' for #<ActiveRecord::ConnectionAdapters::PostgreSQLAdapter:0x007fb7dec91b28>
```

What happened is that I mistyped column (coulmn) and got a
`NoMethodError`, because of a missing comma during the raise. This made
Ruby think we're calling the method `ArgumentError`.

Hash syntax auto-correcting breaks alignments. 411ccbda

Related f27a932d, #25391.

The current code base is not uniform. After some discussion,
we have chosen to go with double quotes by default.

A `value` is only used for checking `value.nil?`. It is unnecessary if
immediately return when `value.nil?`.

`exec_insert` cannot return last inserted id if `use_insert_returning?`
is true. `sql_for_insert` should be called inside `exec_insert`.