ActiveRecord::Relation::Merger's filter_binds method does not filter out bind
variables when one of the attribute nodes has a string name, but the other has
a symbol name, even when those names are actually equal.

This can result in there being more bind variables than placeholders in the
generated SQL.  This is particularly an issue for PostgreSQL, where this is
treated as an error.

This patch changes the filter_binds method to make it convert both attribute
names to strings before comparing.

Closes #10802.

Before this patch `Infinity`, `-Infinity` and `Nan` were read as `0`.

In #10410 it was noted that you can no longer create PK's with the
type of bigserial in PostgreSQL in 4.0.0.rc1. This is mostly
because the newer adapter is checking for column type with the
id column instead of just letting it pass through like it did
before.

Side effects:
You may just create a PK column of a type that you really don't
want to be your PK. As far as I can tell this was allowed in 3.2.X
and perhaps an exception should be raised if you try and do
something extremely dumb.

Closes #15024.

These operations happen before instances are created.
The current behavior is misleading and can result in broken behavior.

We are reverting these commits, because there are other caveats related
to dirty attributes not being restored when a transaction is rollback.
For instance, nested transactions cannot proper restore the dirty
attributes state after a rollback.

At the moment, the dirty attributes are scoped by the transaction.
When we call `.save` on a record, the dirty attributes will be reset even
if the transaction gets rollback.

[related #13166]
[related #15018]
[related #15016]
[closes #15019]

This reverts commits
* bab48f0a
  * b0fa7cf3.
* 73fb39b6
  * 37c23892.
  * 8d8d4f15

Revert "Merge pull request #13166 from bogdan/transaction-magic"

HABTM should fall back to using the normal CollectionAssociation's size calculation if the collection is not cached or loaded.

This addresses both #14913 and #14914 for master.

Closes #14989

[related #13166]

This makes table_name_suffix work the same as table_name_prefix when
using namespaced models. Pretty much the same as 67d1cec4.

This is a regression 4.0 -> 4.1 fix.
In 4.1.0 Relation#join is delegated to Arel#SelectManager.
In 4.0 series it is delegated to Array#join

This patch puts back the behaviour of 4.0

associations were not being saved.

Fixes #13854.

[Eric Chahin, Aaron Nelson, & Kevin Casey]

For mysql2/mysql adapters, `sql_mode` variable name set in `database.yml`
as string, was ignored and `sql_mode` was set to use strict mode.

Fixes #14895

When used a custom join_table name on a habtm, rails was not saving it
on Reflections. This causes a problem when rails loads fixtures, because
it uses the reflections to set database with fixtures.

/cc @tenderlove

The PG Adapter should use `:datetime` consistently instead of mapping
mispellings to `:timestamp`.

See #14513

Closes #14841.

Also add a Changelog entry

[related #9702]
[fixes #8928]

Changed the NullRelation so that when count is called with #group it will properly return an empty hash instead of zero. Fixes issue #14721

Conflicts:
	activerecord/CHANGELOG.md

Closes #14222.

This is a follow up to #6104

This does not have the backwards compatibility issues brought up in
implementation to break.

Example:

    author.posts == Post.where(author_id: author.id)
    # => true
      Post.where(author_id: author.id) == author.posts
    # => true

Fixes #13506

[Yves Senn & Matthew Draper]

[Yves Senn & Matthew Draper]

There is no reason for the PG adapter to have a default limit of 255 on :string
columns. See this snippet from the PG docs:

    Tip: There is no performance difference among these three types, apart
    from increased storage space when using the blank-padded type, and a
    few extra CPU cycles to check the length when storing into a
    length-constrained column. While character(n) has performance
    advantages in some other database systems, there is no such advantage
    in PostgreSQL; in fact character(n) is usually the slowest of the
    three because of its additional storage costs. In most situations text
    or character varying should be used instead.

Add tests to make sure scopes cannot be create with names such as:
private, protected, public.
Make sure enum values don't collide with those methods too.

It was causing error when using `with_options` passing a lambda as its
last argument.

    class User < ActiveRecord::Base
      with_options dependent: :destroy do |assoc|
        assoc.has_many :profiles, -> { where(active: true) }
      end
    end

It was happening because the `option_merger` was taking the last
argument and checking if it was a Hash. This breaks the HasMany usage,
because its last argument can be a Hash or a Proc.

As the behavior described in this test:
https://github.com/rails/rails/blob/master/activesupport/test/option_merger_test.rb#L69
the method will only accept the lambda, this way it will keep the expected behavior. See 9eaa0a34