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

So this bug is kinda funky. The code path is basically "if we weren't passed an
instance of the class we compose to, and we have a converter, call that".
Ignoring the hash case for a moment, everything after that was roughly intended
to be the "else" clause, meaning that we are expected to have an instance of
the class we compose to. Really, we should be blowing up in that case, as we
can give a much better error message than what they user will likely get (e.g.
`NameError: No method first for String` or something). Still, Ruby is duck
typed, so if the object you're assigning responds to the same methods as the
type you compose to, knock yourself out.

The hash case was added in 36e9be85 to remove a bunch of special cased code from
multiparameter assignment. I wrongly assumed that the only time we'd get a hash
there is in that case. Multiparameter assignment will construct a very specific
hash though, where the keys are integers, and we will have a set of keys
covering `1..part.size` exactly. I'm pretty sure this could actually be passed
around as an array, but that's a different story. Really I should convert this
to something like `class MultiParameterAssignment < Hash; end`, which I might
do soon. However for a change that I'm willing to backport to 4-2-stable, this
is what I want to go with for the time being.

Fixes #25978

Fixes #25892.

Fixes #25653.

If reuse `QUOTED_TRUE` and `QUOTED_FALSE` without frozen, causing the
following issue.

```
Loading development environment (Rails 5.1.0.alpha)
irb(main):001:0> ActiveRecord::Base.connection.quote(true) << ' foo'
=> "1 foo"
irb(main):002:0> ActiveRecord::Base.connection.quote(true) << ' foo'
=> "1 foo foo"
irb(main):003:0> type = ActiveRecord::ConnectionAdapters::AbstractMysqlAdapter::MysqlString.new
=> #<ActiveRecord::ConnectionAdapters::AbstractMysqlAdapter::MysqlString:0x007fd40c15e018 @precision=nil, @scale=nil, @limit=nil>
irb(main):004:0> type.serialize(true) << ' bar'
=> "1 foo foo bar"
irb(main):005:0> type.cast(true) << ' bar'
=> "1 foo foo bar bar"
```

These test cases tests exactly mutating loaded target.

This is not a test case.

Previously we had primarily tested the behavior of these attributes by
calling `.new`, allowing this to slip through the cracks. There were a
few ways in which they were behaving incorrectly.

The biggest issue was that attempting to read the attribute would
through a `MissingAttribute` error. We've corrected this by returning
the default value when the attribute isn't backed by a database column.
This is super special cased, but I don't see a way to avoid this
conditional. I had considered handling this higher up in
`define_default_attribute`, but we don't have the relevant information
there as users can provide new defaults for database columns as well.

Once I corrected this, I had noticed that the attributes were always
being marked as changed. This is because the behavior of
`define_default_attribute` was treating them as assigned from
`Attribute::Null`.

Finally, with our new implementation, `LazyAttributeHash` could no
longer be marshalled, as it holds onto a proc. This has been corrected
as well. I've not handled YAML in that class, as we do additional work
higher up to avoid YAML dumping it at all.

Fixes #25787
Close #25841

It all started noticing some foo.method({ ... }) method calls
in passing. This is a whole pass modernizing this file.

While some string literals are edited where I touched code,
this pass does not uniformizes quotes. A ton are left untouched
on purposes. We have no defined style.

Where appropriatei, prefer the more concise Regexp#match?,
String#include?, String#start_with?, or String#end_with?

`send` is unnecessary.

 - Refer https://github.com/rsim/oracle-enhanced/pull/845
Remove `set_date_columns` which has been deprecated in Oracle enhanced adapter
 - Refer https://github.com/rsim/oracle-enhanced/pull/869

`binds` is an array of a query attribute since Active Record 5.0.

Fixes #25732.

Address to https://github.com/rails/rails/commit/5a302bf553af0e6fedfc63299fc5cd6e79599ef3#commitcomment-18288388.

Changing the order of method chaining `merge` and other query
method such as `joins` should produce the same result.

```ruby
class Topic < ApplicationRecord
  scope :safe_chaininig,   -> { joins(:comments).merge(Comment.newest) }
  scope :unsafe_chaininig, -> { merge(Comment.newest).joins(:comments) } #=> NoMethodError
end
```

Albeit the previous existence of this method is not seen in
the patch itself, the declaration

    mattr_accessor :error_on_ignored_order_or_limit, instance_writer: false

was present before. It was removed recently in 210012f6.

Follow up to #20818.

`retrieve_connection` is passed `spec_name` instead of `klass` since #24844.

- Followup of https://github.com/rails/rails/pull/23179

loaded model classes have their connections wrapped in transactions.

See #17776

In Rails 4 config.eager_load was changed to false in the test environment. This
means that model classes that connect to alternate databases with
establish_connection are not loaded at start up. If use_transactional_fixtures
is enabled, transactions are wrapped around the connections that have been
established only at the start of the test suite. So model classes loaded later
don't have transactions causing data created in the alternate database not to
be removed.

This change resolves that by creating a new connection.active_record
notification that gets fired whenever a connection is established. I then added
a subscriber after we set up transactions in the test environment to listen for
additional connections and wrap those in transactions as well.

Caused at a45363a2.

When `group` is used in combination with any calculation method, the
resulting hash uses the grouping expression as the key. Currently we're
incorrectly always favoring the type reported by the query, instead of
the type known by the class. This causes differing behavior depending on
whether the adaptor actually gives proper types with the query or not.
After this change, the behavior will be the same on all adaptors -- we
see if we know the type from the class, fall back to the type from the
query, and finally fall back to the identity type.

Fixes #25595

`AssociationQueryHandler` requires `association` initialized
`TableMetadata` even if `table_name == arel_table.name`.

Fixes #25689.

The documentation states that parameter values longer than 20 characters
will be truncated by words, but the example shows that a parameter based
on "David Heinemeier Hansson" (with id: 125) becomes "125-david" when
"David Heinemeier".length == 16 so why so short?

The answer lies in the use of the #truncate option omission: nil which
seems to have been intended to mean "nothing", but which actually causes
the default string "..." to be used. This causes #truncate to cleave
words until the "..." can be added and still remain within the requested
size of 20 characters.

The better option is omission: '' (which is probably what was originally
intended).

Furthermore, since the use of #parameterize will remove non-alphanumeric
characters, we can maximize the useful content of the output by calling
parameterize first and then giving truncate a separator: /-/ rather than
a space.

     as this can lead to confusing time stubbing.

     Instead of:

         travel_to 2.days.from_now do
           # 2 days from today
           travel_to 3.days.from_now do
             # 5 days from today
           end
         end

     preferred way to achieve above is:

         travel_to 2.days.from_now
         # 2 days from today

         travel_back
         travel_to 5.days.from_now
         # 5 days from today

Closes #24690
Fixes #24689

Actually schema dumper/creation supports composite primary key (#21614).
Therefore it should not show the warning about composite primary key in
connection adapter.

This change moves the warning to `AttributeMethods::PrimaryKey` and
suppress the warning for habtm join table.

Fixes #25388.

This reverts commit 407e0ab5.