I'm not sure why `Mysql2VirtualColumnTest#test_schema_dumping` passed
previously. But now the test not pass at least in MariaDB 10.1.9.
I fixed the regexp to respect `COLLATE`.

```
% ARCONN=mysql2 be ruby -w -Itest test/cases/adapters/mysql2/virtual_column_test.rb -n test_schema_dumping
Using mysql2
Run options: -n test_schema_dumping --seed 7131

F

Finished in 0.466304s, 2.1445 runs/s, 4.2890 assertions/s.

  1) Failure:
Mysql2VirtualColumnTest#test_schema_dumping [test/cases/adapters/mysql2/virtual_column_test.rb:55]:
Expected /t\.virtual\s+"upper_name",\s+type: :string,\s+as: "UPPER\(`name`\)"$/i to match "# This file is auto-generated from the current state of the database. Instead\n# of editing this file, please use the migrations feature of Active Record to\n# incrementally modify your database, and then regenerate this schema definition.\n#\n# Note that this schema.rb definition is the authoritative source for your\n# database schema. If you need to create the application database on another\n# system, you should be using db:schema:load, not running all the migrations\n# from scratch. The latter is a flawed and unsustainable approach (the more migrations\n# you'll amass, the slower it'll run and the greater likelihood for issues).\n#\n# It's strongly recommended that you check this file into your version control system.\n\nActiveRecord::Schema.define(version: 0) do\n\n  create_table \"virtual_columns\", force: :cascade, options: \"ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci\" do |t|\n    t.string \"name\"\n    t.virtual \"upper_name\", type: :string, as: \n    t.virtual \"name_length\", type: :integer, as: \"LENGTH(`name`)\", stored: true\n  end\n\nend\n".

1 runs, 2 assertions, 1 failures, 0 errors, 0 skips
```

```
> select @@version;
+--------------------+
| @@version          |
+--------------------+
| 10.1.9-MariaDB-log |
+--------------------+
1 row in set (0.00 sec)
```

Related #28399.

Remove datetime fields from helper list [ci skip]

The `datetime_field` is an alias of the `datetime_local_field`, there is
no helper to generate datetime fields.
Ref: #25469

Add test for  method `#attributes`

Explicitly require AS::Time in AS::Testing::TimeHelpers

Fix Enumerable#sum redefined warning

If you just try to use `ActiveSupport::Testing::TimeHelpers` standalone by requiring `active_support/testing/time_helpers`, you currently get an error: `NoMethodError: undefined method `change' for 2017-12-14 01:04:44 -0500:Time`

9f6e82ee added a dependency on `AS::Time` by using `AS::Time#change`.

Here's a script to reproduce the error:

```ruby
require "bundler/inline"

gemfile(true) do
  source "https://rubygems.org"
  gem "activesupport", github: "rails/rails"
end

require "active_support/testing/time_helpers"

require "minitest/autorun"

class BugTest < Minitest::Test
  include ActiveSupport::Testing::TimeHelpers

  def test_stuff
    travel_to Time.new(2017, 12, 14, 01, 04, 44) do
      assert true
    end
  end
end
```

It currently fails for all 5.x.x versions and master.  Ideally, this would be backported to `5-0-stable` and `5-1-stable` as well.

This reverts commit b86631d9, reversing
changes made to 8776a713.

ActiveSupport::Notifications is loaded using autoload that is defined by
the top level file of `active_support`. All the frameworks of Rails
requires the top level files before using any of the others files inside
the framework because the top level file is what setup the autoload
hooks and require the common dependencies.

Explicitly require AS::Notifications in AS::Cache

If we require 'active_support/core_ext/enumerable' on Ruby 2.4,
we'll see following warning because `Enumerable#sum` and `Array#sum`
are added in Ruby 2.4.

```
rails/rails/activesupport/lib/active_support/core_ext/enumerable.rb:20: warning: method redefined; discarding old sum
```

The minimal way to fix the warning is `alias sum sum`.

```
$ ruby -v
ruby 2.4.1p111 (2017-03-22 revision 58053) [x86_64-linux]

$ ruby -w -e "def a; end; def a; end"
-e:1: warning: method redefined; discarding old a
-e:1: warning: previous definition of a was here

$ ruby -w -e "def a; end; alias a a; def a; end"
```

But this behavior is not intended. (@amatsuda was told by @ko1)
So we should use `alias` as a meaningful way.

Ruby 2.4's `sum`  (`orig_sum`) assumes an `identity` is `0` when we omit `identity`
so we can delegate to `orig_sum` with explicit `identity` only.
In a strict sense, we can detect `identity` by check instance's class
but we don't care at this time about that because calling `Enumerable#sum` is rare.
In many cases, we will call `Array#sum`.

The current check for whether to add an optional format to the path
is very lax and will match things like `:format_id` where there are
nested resources, e.g:

    resources :formats do
      resources :items
    end

Fix this by using a more restrictive regex pattern that looks for
the patterns `(.:format)`, `.:format` or `/` at the end of the path.
Note that we need to allow for multiple closing parenthesis since
the route may be of this form:

    get "/books(/:action(.:format))", controller: "books"

This probably isn't what's intended since it means that the default
index action route doesn't support a format but we have a test for
it so we need to allow it.

Fixes #28517.

[ci skip]

Add more explanation to to_time about timezones

Add missing word "programming", to clarify what type of language Ruby is
;)

[ci skip]

There's a difference in the results between `Date#to_time(:local)` and
`Date#in_time_zone` but it is subtle and can confuse users (like me :-).

Right now, if you install the current release version of Active Support (5.0.2) and try to use its cache implementation standalone by requiring `active_support/cache`, it crashes with `NameError: uninitialized constant ActiveSupport::Notifications`.  `AS::Notifications` is used in `cache.rb` down around [line 555](https://github.com/rails/rails/blob/8776a7139757d0b264785c774d4e7f37d4bc1ac7/activesupport/lib/active_support/cache.rb#L555).

Here's a quick repro script:

```ruby
require "bundler/inline"

gemfile(true) do
  source "https://rubygems.org"
  gem "activesupport", "5.0.2"
end

require "active_support/cache"

cache_store = ActiveSupport::Cache::MemoryStore.new

cache_store.write('test', 'okay')
puts cache_store.read('test')
```

However, any version _newer_ than 5.0.2 passes.

This is because [this commit](https://github.com/rails/rails/commit/75924c4517c8f87712d3f59c11f10152ed57b9d8) inadvertently included `AS::Notifications` into `AS::Cache` (thus fixing the issue) by mixing [`AS::Deprecation` into `AS::Duration`](https://github.com/rails/rails/blob/75924c4517c8f87712d3f59c11f10152ed57b9d8/activesupport/lib/active_support/duration.rb#L4), giving you a nice require chain of [`Cache` including `Time`][1] [including `Duration`][2] [including `Deprecation`][3] [including `Behaviors`][4] [including `Notifications`][5].  Phew.

Aside from being not very explicit, the fact that the fixing is specifically done by `AS::Deprecation` means that this fix is probably only temporary (until the deprecation is removed).

This PR just makes the inclusion explicit to future-proof against this breakage.

(Ideally, this would also be backported to `5-0-stable` to get picked up in any subsequent point release.)

See also: https://github.com/rails/rails/pull/14667

[1]: https://github.com/rails/rails/blob/75924c4517c8f87712d3f59c11f10152ed57b9d8/activesupport/lib/active_support/cache.rb#L6
[2]: https://github.com/rails/rails/blob/75924c4517c8f87712d3f59c11f10152ed57b9d8/activesupport/lib/active_support/time.rb#L2
[3]: https://github.com/rails/rails/blob/75924c4517c8f87712d3f59c11f10152ed57b9d8/activesupport/lib/active_support/duration.rb#L4
[4]: https://github.com/rails/rails/blob/75924c4517c8f87712d3f59c11f10152ed57b9d8/activesupport/lib/active_support/deprecation.rb#L16
[5]: https://github.com/rails/rails/blob/75924c4517c8f87712d3f59c11f10152ed57b9d8/activesupport/lib/active_support/deprecation/behaviors.rb#L1

Add Puerto Rico support to ActiveSupport::TimeZone

Add missing word 'to' [ci skip]

Smooth form with upgrading

While the original sentence is easy to understand, adding the 'to' is
imho a bit more correct.

Sources:
* https://ell.stackexchange.com/q/16964
* https://www.usingenglish.com/forum/threads/127099-It-is-referred-as-or-It-is-referred-to-as

Since 1a92ae83 all `indexes` methods are under the `SchemaStatements`.

[ci skip]

[ci skip]

Support Descending Indexes for MySQL

Allows users to not have remote forms by default, since there's
more JS harness, e.g. bundling rails-ujs, otherwise.

Also don't skip creating defaults file anymore. Sprockets isn't the only new config.

Effectively treat nil values as "auto", e.g. whatever a form helper
chooses to interpret it as.

But treat an explicitly assigned false value as disabling.

Use the config value directly when call `secrets`

Currently, `read_encrypted_secrets` is set with initializer.
Therefore if refer to `secrets` in config, `read_encrypted_secrets` is false,
so can not get the value of `secrets.yml.enc`.

In order to be able to refer to secrets in config, modified to refer to
`config.read_encrypted_secrets` when calling `secrets`.

Fixes #28618.

MySQL 8.0.1 and higher supports descending indexes: `DESC` in an index
definition is no longer ignored.

See https://dev.mysql.com/doc/refman/8.0/en/descending-indexes.html.

* Use keyword arguments in `IndexDefinition` to ease to ignore unused
  options and to avoid to initialize incorrect empty value.
* Place it in `SchemaStatements` for consistency.
* And tiny tweaks.

Changelog links at the top of the components, and lowercased the 'c' in
notable changes. Personally, the styling isn't super important to me, but
it's what we always done in previous guides, so I figured it's better to
keep with what we've always done.

[ci skip]

Prevent aliases Array#append and Array#prepend

Started adding release notes for Rails 5.1 [ci skip]

Rename `association_query_handler.rb` to `association_query_value.rb`

Early return in `PredicateBuilder::ArrayHandler`