We originally did the whole `load_database_yaml` thing because this test
wasn't cooperating and we needed to finish the namespaced rake tasks for
multiple databases.

However, it turns out that YAML can't eval ERB if you don't tell it it's
ERB so you get Pysch parse errors if you're using multi-line ERB or
ERB with conditionals. It's a hot mess.

After trying a few things and thinking it over we decided that it wasn't
worth bandaiding over, the test needed to be improved. The test was
added in #31135 to test that the env is loaded in these tasks. But it
was blowing up because we were trying to read a database name out of the
configuration - however that's not the purpose of this change. We want
to read environment files in the rake tasks, but not in the config
file.

In this PR we changed the test to test what the PR was actually fixing.
We've also deleted the `load_database_yaml` because it caused more
problems than it was worth. This should fix the issues described in
https://github.com/rails/rails/pull/32274#issuecomment-384161057. We
also had these problems at GitHub.
Co-authored-by: Nalimi <aibrahim2k2@gmail.com>

`changes_applied` should be called before continuing around callback
chain. Otherwise the mutation tracker returns old value for methods like
`changed`? or `id_in_database` in around callbacks. Also methods depend
on `id_in_database`, like `update_column`, are not working in
`around_create` callbacks.

```
class Foo < ActiveRecord::Base
  around_create :around_create_callback

  def around_create_callback
    ...
    yield
    p id_in_database # => nil
    update_column(:generated_column, generate_value) # silently fails
  end

  ...
end
```

* Add documentation for `:collation` option

The table definition supports a `:collation` option for string and text columns, but this is not documented anywhere that I could find.

I'm not sure if the "If not specified" part is accurate. From [this PR](https://github.com/rails/rails/commit/1515c4d98da3f730ef971fa5a13cad828bd9bef4), it looks like it passes `nil` and lets the database handle the collation, but I'm happy to change it if I misread the code.

[ci skip]

* FIX remove whitespace

[Nate Pinsky + Rafael Mendonça França]

There is test in has_one to test enum, but there is no for has_many.

[Rich Chen]

`Topic` isn't loaded in this test file.

This was introduced at 24f6bf0d.

https://github.com/rails/rails/issues/31190

The `Relation::Merger` has a problem that order values would be merged
as nested array.

That was caused an issue #33664 since if array value is passed to
`order` and first element in the array includes `?`, the array is
regarded as a prepared statement and bind variables.

https://api.rubyonrails.org/classes/ActiveRecord/Sanitization/ClassMethods.html#method-i-sanitize_sql_for_order

Just merging that as splat args like other values would fix the issue.

Fixes #33664.

Fixes #33520.

without define the enum in class “SpecialBook”, any string status will be casted to integer 0. Then the test have no meaning.

[Rich Chen]

https://travis-ci.org/rails/rails/jobs/417783260#L1173-L1179

If `foreign_keys` is fetched before alter table, it couldn't notice
whether foreign keys are broken or not after alter table.

To prevent style check in review like https://github.com/rails/rails/pull/33608#discussion_r211087605.

- Move changelog entry of #33530 up in order to preserve the chronology
  since we always add new entries on the top of a changelog file.
- Clarify the changelog entry
- Clarify the docs of remove_foreign_key
- Ensure reversible of `remove_foreign_key` with `:primary_key` and `:to_table`
  options.

remove_foreign_key supports
- remove_foreign_key :accounts, :branches
- remove_foreign_key :accounts, to_table: :branches

but the second one is not reversible.

This branch is to fix and allow second one to be reversible.

[Nikolay Epifanov, Rich Chen]

If a transaction is opened and closed without any queries being run, we
can safely omit the `BEGIN` and `COMMIT` statements, as they only exist
to modify the connection's behaviour inside the transaction. This
removes the overhead of those statements when saving a record with no
changes, which makes workarounds like `save if changed?` unnecessary.

This implementation buffers transactions inside the transaction manager
and materializes them the next time the connection is used. For this to
work, the adapter needs to guard all connection use with a call to
`materialize_transactions`. Because of this, adapters must opt in to get
this new behaviour by implementing `supports_lazy_transactions?`.

If `raw_connection` is used to get a reference to the underlying
database connection, the behaviour is disabled and transactions are
opened eagerly, as we can't know how the connection will be used.
However when the connection is checked back into the pool, we can assume
that the application won't use the reference again and reenable lazy
transactions. This prevents a single `raw_connection` call from
disabling lazy transactions for the lifetime of the connection.

Since fe9547b6, numericality validator would parse raw value only when a
value came from user to work type casting to a value from database.

But that was caused a regression that the validator would work against
getter value instead of parsed raw value, a getter is sometimes
customized by people. #33550

There we never guarantees that the value before type cast was going to
the used in this validation (actually here is only place that getter
value might not be used), but we should not change the behavior unless
there is some particular reason.

The purpose of fe9547b6 is to work type casting to a value from
database. We could achieve the purpose by using `read_attribute`,
without using getter value.

Fixes #33550.

Six Mocha calls prove quite resistant to Minitestification. For example,
if we replace

```
  ActiveRecord::Associations::HasManyAssociation
    .any_instance
    .expects(:reader)
    .never
```

with `assert_not_called`, Minitest wisely raises

```
NameError: undefined method `reader' for class `ActiveRecord::Associations::HasManyAssociation'
```

as `:reader` comes from a deeply embedded abstract class,
`ActiveRecord::Associations::CollectionAssociation`.

This patch tackles this difficulty by adding
`ActiveSupport::Testing::MethodCallAsserts#assert_called_on_instance_of`
which injects a stubbed method into `klass`, and verifies the number of
times it is called, similar to `assert_called`. It also adds  a convenience
method, `assert_not_called_on_instance_of`, mirroring
`assert_not_called`.

It uses the new method_call_assertions to replace the remaining Mocha
calls in `ActiveRecord` tests.

[utilum + bogdanvlviv + kspath]

Related to #33520

The first thing this method does is run  on the argument. This change passes
in a string so we don't allocate a bunch of unnecessary extra strings by
calling to_s on a symbol over and over.

What?
From now on when `accepts_nested_attributes_for` defined and `extend` option
added the overwritten `build` method being called.

[Alireza Bashiri, Martins Polakovs]

This commit follows the path we started at commit #ea4f0e2b
and continued at PR #33229.

This reverts commit d162188d, reversing
changes made to 35767828.

Reason: #24131 conflicts the #5153's default order contract, it means
that existing apps would be broken by that change.

We don't want to break existing apps without a deprecation cycle.

Since 9ac7dd47, class level `update`, `destroy`, and `delete` were placed
in the `Persistence` module as class methods.

But `Relation#update` without passing ids which was introduced at #11898
is not a class method, and it was caused the extra scoping regression
#33470.

I moved the relation method back into the `Relation` to fix the
regression.

Fixes #33470.