activerecord 4.0.4 → 4.1.16

data/CHANGELOG.md

@@ -1,2847 +1,2682 @@
-## Rails 4.0.4 (March 14, 2014) ##
+## Rails 4.1.16 (July 12, 2016) ##
 
-*   Only save has_one associations if record has changes.
-    Previously after save related callbacks, such as `#after_commit`, were triggered when the has_one
-    object did not get saved to the db.
+*   Correctly pass MySQL options when using structure_dump or structure_load
 
-    *Alan Kennedy*
+    Specifically, it fixes an issue when using SSL authentication.
 
-*   Fixed STI classes not defining an attribute method if there is a
-    conflicting private method defined on its ancestors.
+    *Alex Coomans*
 
-    Fixes #11569.
 
-    *Godfrey Chan*
+## Rails 4.1.15 (March 07, 2016) ##
 
-*   Fix validation on uniqueness of empty association.
+*   Fixes custom primary keys for associations when calling `Relation#where`
 
-    *Evgeny Li*
+    Fixes #23327.
 
-*   Perform necessary deeper encoding when hstore is inside an array.
+    *Rick Song*
 
-    Fixes #11135.
+*   No longer pass deprecated option `-i` to `pg_dump`.
 
-    *Josh Goodall*, *Genadi Samokovarov*
+    *Paul Sadauskas*
 
-*   Properly detect if a connection is still active before using it
-    in multi-threaded environments.
 
-    Fixes #12867.
+## Rails 4.1.14.2 (February 26, 2016) ##
 
-    *Kevin Casey*, *Matthew Draper*, *William (B.J.) Snow Orvis*
+*   No changes.
 
-*   When inverting add_index use the index name if present instead of
-    the columns.
 
-    If there are two indices with matching columns and one of them is
-    explicitly named then reverting the migration adding the named one
-    would instead drop the unnamed one.
+## Rails 4.1.14.1 (January 25, 2015) ##
 
-    The inversion of add_index will now drop the index by its name if
-    it is present.
+*   No changes.
 
-    *Hubert Dąbrowski*
 
-*   Make sure transaction state gets reset after a commit operation on the record.
+## Rails 4.1.14 (November 12, 2015) ##
 
-    If a new transaction was open inside a callback, the record was loosing track
-    of the transaction level state, and it was leaking that state.
+*   No longer pass deprecated option `-i` to `pg_dump`.
 
-    Fixes #12566.
+    *Paul Sadauskas*
 
-    *arthurnn*
+*   Includes HABTM returns correct size now. It's caused by the join dependency
+    only instantiates one HABTM object because the join table hasn't a primary key.
 
-*   PostgreSQL implementation of `SchemaStatements#index_name_exists?`.
+    Fixes #16032.
 
-    The database agnostic implementation does not detect with indexes that are
-    not supported by the ActiveRecord schema dumper. For example, expressions
-    indexes would not be detected.
+    Examples:
 
-    Fixes #11018.
+        before:
 
-    *Jonathan Baudanza*
+        Project.first.salaried_developers.size # => 3
+        Project.includes(:salaried_developers).first.salaried_developers.size # => 1
 
-*   Parsing PostgreSQL arrays with empty strings now works correctly.
+        after:
 
-    Previously, if you tried to parse `{"1","","2","","3"}` the result
-    would be `["1","2","3"]`, removing the empty strings from the array,
-    which would be incorrect. Now it will correctly produce `["1","","2","","3"]`
-    as the result of parsing the above PostgreSQL array.
+        Project.first.salaried_developers.size # => 3
+        Project.includes(:salaried_developers).first.salaried_developers.size # => 3
 
-    Fixes #13907.
+    *Bigxiang*
 
-    *Maurício Linhares*
 
-*   Fix regressions on `select_*` methods.
-    When `select_*` methods receive a `Relation` object, they should be able to get the arel/binds from it.
-    Also fix regressions on select_rows that was ignoring the binds.
+## Rails 4.1.13 (August 24, 2015) ##
 
-    Fixes #7538, #12017, #13731, #12056.
+*   Don't define autosave association callbacks twice from
+    `accepts_nested_attributes_for`.
 
-    *arthurnn*
+    Fixes #18704.
 
-*   Correctly send an user provided statement to a `lock!()` call.
+    *Sean Griffin*
 
-        person.lock! 'FOR SHARE NOWAIT'
-        # Before: SELECT * ... LIMIT 1 FOR UPDATE
-        # After: SELECT * ... LIMIT 1 FOR SHARE NOWAIT
 
-    Fixes #13788.
+## Rails 4.1.12 (June 25, 2015) ##
 
-    *Maurício Linhares*
+*   Raises the right exception when declares a has many through
+    association with missing source.
 
-*   Active Record objects can now be correctly dumped, loaded and dumped again without issues.
+    *Mauro George*
 
-    Previously, if you did `YAML.dump`, `YAML.load` and then `YAML.dump` again
-    in an ActiveRecord model that used serialization it would fail at the last
-    dump due to the fields not being correctly serialized before being dumped
-    to YAML. Now it is possible to dump and load the same object as many times
-    as needed without any issues.
+*   Revert behavior of `db:schema:load` back to loading the full
+    environment. This ensures that initializers are run.
 
-    Fixes #13861.
+    Fixes #19545.
 
-    *Maurício Linhares*
+    *Yves Senn*
 
-*   `has_one` and `belongs_to` accessors don't add ORDER BY to the queries anymore.
+*   Rename `:class` to `:anonymous_class` in association options.
 
-    Since Rails 4.0, we add an ORDER BY in the `first` method to ensure consistent results
-    among different database engines. But for singular associations this behavior is not needed
-    since we will have one record to return. As this ORDER BY option can lead some performance
-    issues we are removing it for singular associations accessors.
+    Fixes #19659.
 
-    Fixes #12623.
+    *Andrew White*
 
-    *Rafael Mendonça França*
+*   Fix referencing wrong table aliases while joining tables of has many through
+    association (only when calling calculation methods).
 
-*   Fail early with "Primary key not included in the custom select clause"
-    in `find_in_batches`.
+    Fixes #19276.
 
-    Before this patch, the exception was raised after the first batch was
-    yielded to the block. This means that you only get it, when you hit the
-    `batch_size` treshold. This could shadow the issue in development.
+    *pinglamb*
 
-    *Alexander Balashov*
 
-*   ActiveRecord states are now correctly restored after a rollback for
-    models that did not define any transactional callbacks (i.e.
-    `after_commit`, `after_rollback` or `after_create`).
+## Rails 4.1.11 (June 16, 2015) ##
 
-    Fixes #13744.
+*   No changes.
 
-    *Godfrey Chan*
 
-*   Add back `#yaml_initialize` method for Syck deserialization
+## Rails 4.1.10 (March 19, 2015) ##
 
-    `ActiveRecord::Core#yaml_initialize` has been added back to support deserialization using
-    the Syck engine. This is to facilitate migration from Syck to Psych. Serialization support
-    will not been restored for Syck, Psych should be used to serialize YAML once its been
-    parsed with Syck.
+*   Fixed ActiveRecord::Relation#becomes! and changed_attributes issues for type column
 
-    *Blake Mesdag*, *arthurnn*
+    Fixes #17139.
 
-*   Make `touch` fire the `after_commit` and `after_rollback` callbacks.
+    *Miklos Fazekas*
 
-    *Harry Brundage*
+*   A `NullRelation` should represent nothing. This fixes a bug where
+    `Comment.where(post_id: Post.none)` returned a non-empty result.
 
-*   Don't try to get the subclass if the inheritance column doesn't exist
+    Closes #15176.
 
-    The `subclass_from_attrs` method is called even if the column specified by
-    the `inheritance_column` setting doesn't exist. This prevents setting associations
-    via the attributes hash if the association name clashes with the value of the setting,
-    typically `:type`. This worked previously in Rails 3.2.
+    *Matthew Draper*, *Yves Senn*
 
-    *Ujjwal Thaakar*
+*   Respect custom primary keys for associations when calling `Relation#where`
 
-*   Fix bug in `becomes!` when changing from the base model to a STI sub-class.
+    Fixes #18813.
 
-    Fixes #13272.
+    *Sean Griffin*
 
-    *the-web-dev*, *Yves Senn*
+*   Fixed ActiveRecord::Relation#group method when argument is SQL reserved key word:
 
-*   Initialize version on Migration objects so that it can be used in a migration,
-    and it will be included in the announce message.
+      SplitTest.group(:key).count
+      Property.group(:value).count
 
-    *Dylan Thacker-Smith*
+    *Bogdan Gusiev*
 
-*   `change_table` now uses the current adapter's `update_table_definition`
-    method to retrieve a specific table definition.
-    This ensures that `change_table` and `create_table` will use
-    similar objects.
+*   Fixed setting of foreign_key for through associations while building of new record.
 
-    Fixes #13577 and #13503.
+    Fixes #12698.
 
-    *Nishant Modak*, *Prathamesh Sonpatki*, *Rafael Mendonça França*
+    *Ivan Antropov*
 
-*   Fixed ActiveRecord::Store nil conversion TypeError when using YAML coder.
-    In case the YAML passed as paramter is nil, uses an empty string.
+*   Fixed automatic inverse_of for models nested in module.
 
-    Fixes #13570.
+    *Andrew McCloud*
 
-    *Thales Oliveira*
+*   Fix `reaping_frequency` option when the value is a string.
 
-*   Fix presence validator for association when the associated record responds to `to_a`.
+    This usually happens when it is configured using `DATABASE_URL`.
 
-    *gmarik*
+    *korbin*
 
-*   Fix bug in `belongs_to` association with `touch: true`.
-    Associated records are no longer touched, when they are already destroyed.
 
-    Fixes #13445.
+## Rails 4.1.9 (January 6, 2015) ##
 
-    Example:
+*   `db:schema:load` and `db:structure:load` no longer purge the database
+    before loading the schema. This is left for the user to do.
+    `db:test:prepare` will still purge the database.
 
-        # Given Comment has belongs_to :post, touch: true
-        comment.post.destroy
-        comment.destroy # no longer touches post
+    Closes #17945.
 
-    *Paul Nikitochkin*
+    *Yves Senn*
 
-*   Fixed `update_column`, `update_columns`, and `update_all` to correctly serialize
-    values for `array`, `hstore` and `json` column types in PostgreSQL.
+*   Bring back `db:test:prepare` to synchronize the test database schema.
 
-    Fixes #12261.
+    Manual synchronization using `bin/rake db:test:prepare` is required
+    when a migration is rolled-back, edited and reapplied.
 
-    *Tadas Tamosauskas*, *Carlos Antonio da Silva*
+    `ActiveRecord::Base.maintain_test_schema` now uses `db:test:prepare`
+    to synchronize the schema. Plugins can use this task as a hook to
+    provide custom behavior after the schema has been loaded.
 
-*   Do not consider PostgreSQL array columns as number or text columns.
+    NOTE: `test:prepare` runs before the schema was synchronized.
 
-    The code uses these checks in several places to know what to do with a
-    particular column, for instance AR attribute query methods has a branch
-    like this:
+    Fixes #17171, #15787.
 
-        if column.number?
-          !value.zero?
-        end
+    *Yves Senn*
 
-    This should never be true for array columns, since it would be the same
-    as running [].zero?, which results in a NoMethodError exception.
+*   Renaming a table in pg also renames the primary key index.
 
-    Fixing this by ensuring that array columns in PostgreSQL never return
-    true for number?/text? checks.
+    Fixes #12856
 
-    *Carlos Antonio da Silva*
+    *Sean Griffin*
 
-*   Fix a bug when assigning an array containing string numbers to a
-    PostgreSQL integer array column.
+*   Make it possible to access fixtures excluded by a `default_scope`.
 
-    Fixes #13444.
+    *Yves Senn*
 
-    Example:
+*   `timestamps` and `add_timestamps` passes additional options along.
+    (like `null: false`)
 
-        # Given Book#ratings is of type :integer, array: true
-        Book.new(ratings: [1, 2]) # worked before
-        Book.new(ratings: ['1', '2']) # now works as well
+    Closes #17624.
 
-    *Damien Mathieu*
+    *Yves Senn*
 
-*   `db:test:clone` and `db:test:prepare` must load Rails environment.
 
-    `db:test:clone` and `db:test:prepare` use `ActiveRecord::Base`. configurations,
-    so we need to load the Rails environment, otherwise the config wont be in place.
+## Rails 4.1.8 (November 16, 2014) ##
 
-    *arthurnn*
+*   Cache `CollectionAssociation#reader` proxies separately before and after
+    the owner has been saved so that the proxy is not cached without the
+    owner's id.
 
-*   Use the right column to type cast grouped calculations with custom expressions.
+    *Ben Woosley*
 
-    Fixes #13230.
+*   Fix preloading of associations which unscope a default scope.
 
-    Example:
+    Fixes #11036.
 
-            # Before
-            Account.group(:firm_name).sum('0.01 * credit_limit')
-            # => { '37signals' => '0.5' }
+    *Byron Bischoff*
 
-            # After
-            Account.group(:firm_name).sum('0.01 * credit_limit')
-            # => { '37signals' => 0.5 }
+*   Do not use `RENAME INDEX` syntax for MariaDB 10.0.
 
-    *Paul Nikitochkin*
+    Fixes #15931.
 
-*   Support deprecated finder options when performing calculations
-    on an association.
+    *Jeff Browning*
 
-    Fixes #13165.
+*   Allow included modules to override association methods.
 
-    *Garth Smedley*, *Yves Senn*
+    Fixes #16684.
 
-*   `ActiveRecord::ConnectionAdapters.string_to_time` respects
-    string with timezone (e.g. Wed, 04 Sep 2013 20:30:00 JST).
+    *Yves Senn*
 
-    Fixes #12278, #12459.
+*   Schema loading rake tasks (like `db:schema:load` and `db:setup`) maintain
+    the database connection to the current environment.
 
-    *kennyj*, *George Guimarães*
+    Fixes #16757.
 
-*   Polymorphic `belongs_to` associations with the `touch: true` option set update the timestamps of
-    the old and new owner correctly when moved between owners of different types.
+    *Joshua Cody*, *Yves Senn*
 
-    Example:
+*   `db:purge` with MySQL respects `Rails.env`.
 
-        class Rating < ActiveRecord::Base
-          belongs_to :rateable, polymorphic: true, touch: true
-        end
+    *Yves Senn*
 
-        rating = Rating.create rateable: Song.find(1)
-        rating.update_attributes rateable: Book.find(2) # => timestamps of Song(1) and Book(2) are updated
+*   Fixed automatic maintaining test schema to properly handle sql structure
+    schema format.
 
-    *Severin Schoepke*
+    Fixes #15394.
 
-*   `type_to_sql` returns a `String` for unmapped columns. This fixes an error
-    when using unmapped array types in combination with PostgreSQL's `array: true`.
+    *Wojciech Wnętrzak*
 
-    Fixes #13146.
+*   Fix has_many :through relation merging failing when dynamic conditions are
+    passed as a lambda with an arity of one.
 
-    Example:
+    Fixes #16128.
 
-        change_colum :table, :column, :bigint, array: true
+    *Agis Anastasopoulos*
 
-    *Jens Fahnenbruck*, *Yves Senn*
 
-*   Previously, the `has_one` macro incorrectly accepts the `counter_cache`
-    option due to a bug, although that options was never supported nor
-    functional on `has_one` and `has_one ... through` relationships. It now
-    correctly raises an `ArgumentError` when passed that option.
+## Rails 4.1.7.1 (November 19, 2014) ##
 
-    *Godfrey Chan*
+*   No changes.
 
-*   Update counter cache on a has_many relationship regardless of default scope
 
-    Fix #12952.
+## Rails 4.1.7 (October 29, 2014) ##
 
-    *Uku Taht*
+*   No changes.
 
-*   Fix bug when validating the uniqueness of an aliased attribute.
 
-    Fixes #12402.
+## Rails 4.1.6 (September 11, 2014) ##
 
-    *Lauro Caetano*
+*   Fixed a regression where whitespaces were stripped from DISTINCT queries in
+    PostgreSQL.
 
-*   Checks to see if the record contains the foreign key to set the inverse automatically.
+    *Agis Anastasopoulos*
 
-    *Edo Balvers*
+    Fixes #16623.
 
-*   Prevent the counter cache from being decremented twice when destroying
-    a record on a has_many :through association.
+*   Fixed an issue where custom accessor methods (such as those generated by
+    `enum`) with the same name as a global method are incorrectly overridden
+    when subclassing.
 
-    Fixes #11079.
+    Fixes #16288.
 
-    *Dmitry Dedov*
+*   Baseclass becomes! subclass.
 
-*   Fix bug where has_one associaton record update result in crash, when replaced with itself.
+    Before this change, a record which changed its STI type, could not be found when updated.
+    Setting update_record to the base class, ensures the record can be found.
 
-    Fixes #12834.
+    Fixes #14785.
 
-    *Denis Redozubov*, *Sergio Cambra*
+    *Matthew Draper*, *Earl St Sauver*, *Edo Balvers*
 
-*   Fix uninitialized constant TransactionState error when Marshall.load is used on an Active Record result.
+*   Fix regression on after_commit that didnt fire when having nested transactions.
 
-    Fixes #12790.
+    Fixes #16425.
 
-    *Jason Ayre*
+    *arthurnn*
 
-*   Do not add to scope includes values from through associations.
-    Fixed bug when providing `includes` in through association scope, and fetching targets.
+*   Define `id_was` to get the previous value of the primary key.
 
-    Example:
-        class Vendor < ActiveRecord::Base
-          has_many :relationships, -> { includes(:user) }
-          has_many :users, through: :relationships
-        end
+    Currently when we call id_was and we have a custom primary key name
+    Active Record will return the current value of the primary key. This
+    make impossible to correctly do an update operation if you change the
+    id.
 
-        vendor = Vendor.first
+    Fixes #16413.
 
-        # Before
+    *Rafael Mendonça França*
 
-        vendor.users.to_a # => Raises exception: not found `:user` for `User`
+*   Fix the schema dump generated for tables without constraints and with
+    primary key with default value of custom PostgreSQL function result.
 
-        # After
+    Fixes #16111.
 
-        vendor.users.to_a # => No exception is raised
+    *Andrey Novikov*
 
+*   Restore 4.0 behavior for using serialize attributes with `JSON` as coder.
 
-    Fixes #12242, #9517, #10240.
+    With 4.1.x, `serialize` started returning a string when `JSON` was passed as
+    the second attribute. It will now return a hash as per previous versions.
 
-    *Paul Nikitochkin*
+    Example:
 
-*   Save `has_one` association when primary key is manually set.
+        class Post < ActiveRecord::Base
+          serialize :comment, JSON
+        end
 
-    Fixes #12302.
+        class Comment
+          include ActiveModel::Model
+          attr_accessor :category, :text
+        end
 
-    *Lauro Caetano*
+        post = Post.create!
+        post.comment = Comment.new(category: "Animals", text: "This is a comment about squirrels.")
+        post.save!
 
-*    Allow any version of BCrypt when using `has_secury_password`.
+        # 4.0
+        post.comment # => {"category"=>"Animals", "text"=>"This is a comment about squirrels."}
 
-     *Mike Perham*
+        # 4.1 before
+        post.comment # => "#<Comment:0x007f80ab48ff98>"
 
-*   Assign inet/cidr attribute with `nil` value for invalid address.
+        # 4.1 after
+        post.comment # => {"category"=>"Animals", "text"=>"This is a comment about squirrels."}
 
-    Example:
+    When using `JSON` as the coder in `serialize`, Active Record will use the
+    new `ActiveRecord::Coders::JSON` coder which delegates its `dump/load` to
+    `ActiveSupport::JSON.encode/decode`. This ensures special objects are dumped
+    correctly using the `#as_json` hook.
 
-        record = User.new
+    To keep the previous behaviour, supply a custom coder instead
+    ([example](https://gist.github.com/jenncoop/8c4142bbe59da77daa63)).
 
-        record.logged_in_from_ip # is type of an inet or a cidr
+    Fixes #15594.
 
-        # Before:
-        record.logged_in_from_ip = 'bad ip address' # raise exception
+    *Jenn Cooper*
 
-        # After:
-        record.logged_in_from_ip = 'bad ip address' # do not raise exception
-        record.logged_in_from_ip # => nil
-        record.logged_in_from_ip_before_type_cast # => 'bad ip address'
+*   Fixed error in `reset_counters` when associations have `select` scope.
+    (Call to `count` generates invalid SQL.)
 
-    *Paul Nikitochkin*
+    *Cade Truitt*
 
-*   `rake db:structure:dump` no longer crashes when the port was specified as `Fixnum`.
+*   After a successful `reload`, `new_record?` is always false.
 
-    *Kenta Okamoto*
+    Fixes #12101.
 
-*   Fix a bug where rake db:structure:load crashed when the path contained
-    spaces.
+    *Matthew Draper*
 
-    *Kevin Mook*
+*   Correctly extract IPv6 addresses from `DATABASE_URI`: the square brackets
+    are part of the URI structure, not the actual host.
 
-*   `ActiveRecord::QueryMethods#unscope` unscopes negative equality
+    Fixes #15705.
 
-    Allows you to call `#unscope` on a relation with negative equality
-    operators, i.e. `Arel::Nodes::NotIn` and `Arel::Nodes::NotEqual` that have
-    been generated through the use of `where.not`.
+    *Andy Bakun*, *Aaron Stone*
 
-    *Eric Hankins*
+*   Don't error when quoting user defined types in PostgreSQL.
 
+    Fixes #15697.
 
-## Rails 4.0.3 (February 18, 2014) ##
+    *Sean Griffin*
 
-*   Correctly escape PostgreSQL arrays.
+*   Ensure both parent IDs are set on join records when both sides of a
+    through association are new.
 
-    Fixes: CVE-2014-0080
+    *Sean Griffin*
 
+*   Pluck now works when selecting columns from different tables with the same
+    name.
 
-## Rails 4.0.2 (December 02, 2013) ##
+    Fixes #15649.
 
-*No changes*
+    *Sean Griffin*
 
+*   `ActiveRecord::FinderMethods.find` with block can handle proc parameter as
+    `Enumerable#find` does.
 
-## Rails 4.0.1 (November 01, 2013) ##
+    Fixes #15382.
 
-*   `NullRelation#pluck` takes a list of columns
+    *James Yang*
 
-    The method signature in `NullRelation` was updated to mimic that in
-    `Calculations`.
+*   `ActiveRecord::SchemaMigration` has no primary key regardless of the
+    `primary_key_prefix_type` configuration.
 
-    *Derek Prior*
+    Fixes #15051.
 
-*   `scope_chain` should not be mutated for other reflections.
+    *JoseLuis Torres*, *Yves Senn*
 
-    Currently `scope_chain` uses same array for building different
-    `scope_chain` for different associations. During processing
-    these arrays are sometimes mutated and because of in-place
-    mutation the changed `scope_chain` impacts other reflections.
+*   `rake db:migrate:status` works with legacy migration numbers like `00018_xyz.rb`.
 
-    Fix is to dup the value before adding to the `scope_chain`.
+    Fixes #15538.
 
-    Fixes #3882.
+    *Yves Senn*
 
-    *Neeraj Singh*
+*   Fixed `columns_for_distinct` of postgresql adapter to work correctly
+    with orders without sort direction modifiers.
 
-*   Prevent the inversed association from being reloaded on save.
+    *Nikolay Kondratyev*
 
-    Fixes #9499.
+*   Keep PostgreSQL `hstore` and `json` attributes as `Hash` in `@attributes`.
+    Fixes duplication in combination with `store_accessor`.
 
-    *Dmitry Polushkin*
+    Fixes #15369.
 
-*   `Relation#order` quotes the column name if you pass a `Symbol`.
-    Fixes #11870.
+    *Yves Senn*
 
-    Example:
+*   `rake railties:install:migrations` respects the order of railties.
 
-        # Before
-        Post.order(:id).to_sql == '... ORDER BY "posts".id ASC'
+    *Arun Agrawal*
 
-        # After
-        Post.order(:id).to_sql == '... ORDER BY "posts"."id" ASC'
 
-    *Yves Senn*
+## Rails 4.1.5 (August 18, 2014) ##
 
-*   Generate subquery for `Relation` if it passed as array condition for `where`
-    method.
+*   No changes.
 
-    Example:
 
-        # Before
-        Blog.where('id in (?)', Blog.where(id: 1))
-        # =>  SELECT "blogs".* FROM "blogs"  WHERE "blogs"."id" = 1
-        # =>  SELECT "blogs".* FROM "blogs"  WHERE (id IN (1))
+## Rails 4.1.4 (July 2, 2014) ##
 
-        # After
-        Blog.where('id in (?)', Blog.where(id: 1).select(:id))
-        # =>  SELECT "blogs".* FROM "blogs"
-        #     WHERE "blogs"."id" IN (SELECT "blogs"."id" FROM "blogs"  WHERE "blogs"."id" = 1)
+*   Fix regression added from the latest security fix.
 
-    Fixes #12415.
+    *Sean Griffin*, *Matthew Draper*
 
-    *Paul Nikitochkin*
 
-*   For missed association exception message
-    which is raised in `ActiveRecord::Associations::Preloader` class
-    added owner record class name in order to simplify to find problem code.
+## Rails 4.1.3 (July 2, 2014) ##
 
-    *Paul Nikitochkin*
+*   Fix SQL Injection Vulnerability in 'range' quoting.
 
-*   Fixes bug when using includes combined with select, the select statement was overwritten.
+    Fixes CVE-2014-3483
 
-    Fixes #11773.
+    *Rafael Mendonça França*
 
-    *Edo Balvers*
 
-*   Objects instantiated using a null relationship will now retain the
-    attributes of the where clause.
+## Rails 4.1.2 (June 26, 2014) ##
 
-    Fixes #11676, #11675, #11376.
+*   Fix regression on eager loading association based on SQL query rather than
+    existing column.
 
-    *Paul Nikitochkin*, *Peter Brown*, *Nthalk*
+    Fixes #15480.
 
-*   Fixed `ActiveRecord::Associations::CollectionAssociation#find`
-    when using `has_many` association with `:inverse_of` and finding an array of one element,
-    it should return an array of one element too.
+    *Lauro Caetano*, *Carlos Antonio da Silva*
+
+*   Fix redefine a has_and_belongs_to_many inside inherited class
+    Fixing regression case, where redefining the same has_an_belongs_to_many
+    definition into a subclass would raise.
+
+    Fixes #14983.
 
     *arthurnn*
 
-*   Callbacks on has_many should access the in memory parent if a inverse_of is set.
+*   Fix has_and_belongs_to_many public reflection.
+    When defining a has_and_belongs_to_many, internally we convert that to two has_many.
+    But as `reflections` is a public API, people expect to see the right macro.
+
+    Fixes #14682.
 
     *arthurnn*
 
-*   Migration dump UUID default functions to schema.rb.
+*   Fixed serialization for records with an attribute named `format`.
 
-    Fixes #10751.
+    Fixes #15188.
 
-    *kennyj*
+    *Godfrey Chan*
 
-*   Fixed a bug in `ActiveRecord::Associations::CollectionAssociation#find_by_scan`
-    when using `has_many` association with `:inverse_of` option and UUID primary key.
+*   Fixed serialized fields returning serialized data after being updated with
+    `update_column`.
 
-    Fixes #10450.
+    *Simon Hørup Eskildsen*
 
-    *kennyj*
+*   When a `group` is set, `sum`, `size`, `average`, `minimum` and `maximum`
+    on a NullRelation should return a Hash.
 
-*   Fix: joins association, with defined in the scope block constraints by using several
-    where constraints and at least of them is not `Arel::Nodes::Equality`,
-    generates invalid SQL expression.
+    *Kuldeep Aggarwal*
 
-    Fixes: #11963
+*   Fixed polymorphic eager loading when using a String as foreign key.
 
-    *Paul Nikitochkin*
+    Fixes #14734.
 
-*   Make possible to run SQLite rake tasks without the `Rails` constant defined.
+    *Lauro Caetano*
 
-    *Damien Mathieu*
+*   Fixed the inferred table name of a has_and_belongs_to_many auxiliar
+    table inside a schema.
 
-*   Allow Relation#from to accept other relations with bind values.
+    Fixes #14824
 
-    *Ryan Wallace*
+    *Eric Chahin*
 
-*   Make `find_in_batches` and `find_each` work without a logger.
+*   Fix bug that added `table_name_prefix` and `table_name_suffix` to
+    extension names in PostgreSQL when migrating.
 
-    *Dmitry Polushkin*
+    *Joao Carlos*
 
-*   Fix inserts with prepared statements disabled.
+*   Floats with limit >= 25 that get turned into doubles in MySQL no longer have
+    their limit dropped from the schema.
 
-    Fixes #12023.
+    Fixes #14135.
 
-    *Rafael Mendonça França*
+    *Aaron Nelson*
 
-*   Setting a has_one association on a new record no longer causes an empty
-    transaction.
+*   Fix how to calculate associated class name when using namespaced has_and_belongs_to_many
+    association.
 
-    *Dylan Thacker-Smith*
+    Fixes #14709.
 
-*   Fix `AR::Relation#merge` sometimes failing to preserve `readonly(false)` flag.
+    *Kassio Borges*
 
-    *thedarkone*
+*   `ActiveRecord::Relation::Merger#filter_binds` now compares equivalent symbols and
+    strings in column names as equal.
 
-*   PostgreSQL adapter recognizes negative money values formatted with
-    parentheses (eg. `($1.25) # => -1.25`)).
-    Fixes #11899.
+    This fixes a rare case in which more bind values are passed than there are
+    placeholders for them in the generated SQL statement, which can make PostgreSQL
+    throw a `StatementInvalid` exception.
 
-    * Yves Senn*
+    *Nat Budin*
 
-*   Do not load all child records for inverse case.
+*   Fix `stored_attributes` to correctly merge the details of stored
+    attributes defined in parent classes.
 
-    currently `post.comments.find(Comment.first.id)` would load all
-    comments for the given post to set the inverse association.
+    Fixes #14672.
 
-    This has a huge performance penalty. Because if post has 100k
-    records and all these 100k records would be loaded in memory
-    even though the comment id was supplied.
+    *Brad Bennett*, *Jessica Yao*, *Lakshmi Parthasarathy*
 
-    Fix is to use in-memory records only if loaded? is true. Otherwise
-    load the records using full sql.
+*   `change_column_default` allows `[]` as argument to `change_column_default`.
 
-    Fixes #10509.
+    Fixes #11586.
 
-    *Neeraj Singh*
+    *Yves Senn*
 
-*   `ActiveRecord::FinderMethods#exists?` returns `true`/`false` in all cases.
+*   Fix `PostgreSQLAdapter::OID::Float#type_cast` to convert Infinity and
+    NaN PostgreSQL values into a native Ruby `Float::INFINITY` and `Float::NAN`
 
-    *Xavier Noria*
+    Example:
 
-*   Load fixtures from linked folders.
+        # Before
+        Point.create(value: 1.0/0)
+        Point.last.value # => 0.0
 
-    *Kassio Borges*
+        # After
+        Point.create(value: 1.0/0)
+        Point.last.value # => Infinity
 
-*   Create a directory for sqlite3 file if not present on the system.
+    *Innokenty Mikhailov*
 
-    *Richard Schneeman*
+*   Allow the PostgreSQL adapter to handle bigserial primary key types again.
 
-*   Removed redundant override of `xml` column definition for PG,
-    in order to use `xml` column type instead of `text`.
+    Fixes #10410.
 
-    *Paul Nikitochkin*, *Michael Nikitochkin*
+    *Patrick Robertson*
 
-*   Revert `ActiveRecord::Relation#order` change that make new order
-    prepend the old one.
+*   Fixed has_and_belongs_to_many's CollectionAssociation size calculation.
 
-    Before:
+    has_and_belongs_to_many should fall back to using the normal CollectionAssociation's
+    size calculation if the collection is not cached or loaded.
 
-        User.order("name asc").order("created_at desc")
-        # SELECT * FROM users ORDER BY created_at desc, name asc
+    Fixes #14913 and #14914.
 
-    After:
+    *Fred Wu*
 
-        User.order("name asc").order("created_at desc")
-        # SELECT * FROM users ORDER BY name asc, created_at desc
+*   Revert the behaviour of `ActiveRecord::Relation#join` changed through 4.0 => 4.1 to 4.0.
 
-    This also affects order defined in `default_scope` or any kind of associations.
+    In 4.1.0 `Relation#join` is delegated to `Arel#SelectManager`.
+    In 4.0 series it is delegated to `Array#join`.
 
-    *Rafael Mendonça França*
+    *Bogdan Gusiev*
 
-*   When using optimistic locking, `update` was not passing the column to `quote_value`
-    to allow the connection adapter to properly determine how to quote the value. This was
-    affecting certain databases that use specific column types.
+*   Log nil binary column values correctly.
 
-    Fixes #6763.
+    When an object with a binary column is updated with a nil value
+    in that column, the SQL logger would throw an exception when trying
+    to log that nil value. This only occurs when updating a record
+    that already has a non-nil value in that column since an initial nil
+    value isn't included in the SQL anyway (at least, when dirty checking
+    is enabled.) The column's new value will now be logged as `<NULL binary data>`
+    to parallel the existing `<N bytes of binary data>` for non-nil values.
 
-    *Alfred Wong*
+    *James Coleman*
 
-*   `change_column` for PostgreSQL adapter respects the `:array` option.
+*   Stringify all variables keys of MySQL connection configuration.
 
-    *Yves Senn*
+    When `sql_mode` variable for MySQL adapters set in configuration as `String`
+    was ignored and overwritten by strict mode option.
 
-*   Fixes bug introduced by #3329. Now, when autosaving associations,
-    deletions happen before inserts and saves. This prevents a
-    'duplicate unique value' database error that would occur if a record being created had
-    the same value on a unique indexed field as that of a record being destroyed.
+    Fixes #14895.
 
-    *Johnny Holton*
+    *Paul Nikitochkin*
 
-*   Flatten merged join values before building the joins.
+*   Ensure SQLite3 statements are closed on errors.
 
-    While joining values special treatment is given to string values.
-    By flattening the array it ensures that string values are detected
-    as strings and not arrays.
+    Fixes #13631.
 
-    Fixes #10669.
+    *Timur Alperovich*
 
-    *Neeraj Singh and iwiznia*
+*   Fix excluding lower bounds of PostgreSQL date and int ranges.
 
-*   Remove extra select and update queries on `save`/`touch`/`destroy` Active Record model
-    with belongs to reflection with option `touch: true`.
+    *River MacLeod*, *Yves Senn*
 
-    Fixes #11288.
+*   When using 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.
 
-    *Paul Nikitochkin*
+    Fixes #14845.
 
-*   Support array as root element in JSON fields.
+    *Kassio Borges*
 
-    *Alexey Noskov & Francesco Rodriguez*
+*   `ActiveRecord::Base.no_touching` no longer triggers callbacks or start empty transactions.
 
-*   Apply default scope when joining associations. For example:
+    Fixes #14841.
 
-        class Post < ActiveRecord::Base
-          default_scope -> { where published: true }
-        end
+    *Lucas Mazza*
 
-        class Comment
-          belongs_to :post
-        end
+*   Fix name collision with `Array#select!` with `Relation#select!`.
 
-    When calling `Comment.joins(:post)`, we expect to receive only
-    comments on published posts, since that is the default scope for
-    posts.
+    Fixes #14752.
 
-    Before this change, the default scope from `Post` was not applied,
-    so we'd get comments on unpublished posts.
+    *Earl St Sauver*
 
-    *Jon Leighton*
+*   Fixed unexpected behavior for `has_many :through` associations going through a scoped `has_many`.
 
-*   `inspect` on Active Record model classes does not initiate a
-    new connection. This means that calling `inspect`, when the
-    database is missing, will no longer raise an exception.
+    If a `has_many` association is adjusted using a scope, and another `has_many :through`
+    uses this association, then the scope adjustment is unexpectedly neglected.
 
-    Fixes #10936.
+    Fixes #14537.
+
+    *Jan Habermann*
+
+*   When a destroyed record is duped, the dup is not `destroyed?`.
+
+    *Kuldeep Aggarwal*
+
+*   Fixed has_many association to make it support irregular inflections.
+
+    Fixes #8928.
+
+    *arthurnn*, *Javier Goizueta*
+
+*   Calling `delete_all` on an unloaded `CollectionProxy` no longer
+    generates a SQL statement containing each id of the collection:
+
+    Before:
+
+        DELETE FROM `model` WHERE `model`.`parent_id` = 1
+        AND `model`.`id` IN (1, 2, 3...)
+
+    After:
+
+        DELETE FROM `model` WHERE `model`.`parent_id` = 1
+
+    *Eileen M. Uchitelle*, *Aaron Patterson*
+
+*   Fixed a problem where count used with a grouping was not returning a Hash.
+
+    Fixes #14721.
+
+    *Eric Chahin*
+
+*   Do not quote uuid default value on `change_column`.
+
+    Fixes #14604.
+
+    *Eric Chahin*
+
+*   The comparison between `Relation` and `CollectionProxy` should be consistent.
 
     Example:
 
-        Author.inspect # => "Author(no database connection)"
+        author.posts == Post.where(author_id: author.id)
+        # => true
+        Post.where(author_id: author.id) == author.posts
+        # => true
 
-    *Yves Senn*
+    Fixes #13506.
 
-*   Fix mysql2 adapter raises the correct exception when executing a query on a
-    closed connection.
+    *Lauro Caetano*
 
-    *Yves Senn*
+*   PostgreSQL adapter only warns once for every missing OID per connection.
 
-*   Fix the `:primary_key` option for `has_many` associations.
+    Fixes #14275.
 
-    Fixes #10693.
+    *Matthew Draper*, *Yves Senn*
 
-    *Yves Senn*
+*   Fixed error for aggregate methods (`empty?`, `any?`, `count`) with `select`
+    which created invalid SQL.
 
-*   Fix bug where tiny types are incorectly coerced as booleand when the length is more than 1.
+    Fixes #13648.
 
-    Fixes #10620.
+    *Simon Woker*
 
-    *Aaron Patterson*
+*   Fix insertion of records via `has_many :through` association with scope.
 
-*   Also support extensions in PostgreSQL 9.1. This feature has been supported since 9.1.
+    Fixes #3548.
 
-    *kennyj*
+    *Ivan Antropov*
 
-*   Deprecate `ConnectionAdapters::SchemaStatements#distinct`,
-    as it is no longer used by internals.
+*   Make possible to have an association called `records`.
+
+    Fixes #11645.
+
+    *prathamesh-sonpatki*
+
+*   `to_sql` on an association now matches the query that is actually executed, where it
+    could previously have incorrectly accrued additional conditions (e.g. as a result of
+    a previous query). CollectionProxy now always defers to the association scope's
+    `arel` method so the (incorrect) inherited one should be entirely concealed.
+
+    Fixes #14003.
+
+    *Jefferson Lai*
+
+*   Fixed error when using `with_options` with lambda.
+
+    Fixes #9805.
+
+    *Lauro Caetano*
 
-    *Ben Woosley#
+*   Fixed error when specifying a non-empty default value on a PostgreSQL array column.
 
-*   Remove not needed bind variables. Port of commit #5082345.
+    Fixes #10613.
 
-    Fixes #10958.
+    *Luke Steensen*
 
-    *Neeraj Singh*
+*   Fixed error where .persisted? throws SystemStackError for an unsaved model with a
+    custom primary key that didn't save due to validation error.
 
-*   Confirm a record has not already been destroyed before decrementing counter cache.
+    Fixes #14393.
 
-    *Ben Tucker*
+    *Chris Finne*
 
-*   Fixed a bug in `ActiveRecord#sanitize_sql_hash_for_conditions` in which
-    `self.class` is an argument to `PredicateBuilder#build_from_hash`
-    causing `PredicateBuilder` to call non-existent method
-    `Class#reflect_on_association`.
+*   `rake db:structure:dump` only dumps schema information if the schema
+    migration table exists.
 
-    *Zach Ohlgren*
+    Fixes #14217.
 
-*   While removing index if column option is missing then raise IrreversibleMigration exception.
+    *Yves Senn*
 
-    Following code should raise `IrreversibleMigration`. But the code was
-    failing since options is an array and not a hash.
+*   Add support for `Relation` be passed as parameter on `QueryCache#select_all`.
 
-        def change
-          change_table :users do |t|
-            t.remove_index [:name, :email]
-          end
-        end
+    Fixes #14361.
 
-    Fix was to check if the options is a Hash before operating on it.
+    *arthurnn*
 
-    Fixes #10419.
+*   Only save `has_one` associations if record has changes. Previously after save
+    related callbacks, such as `#after_commit`, were triggered when the
+    `has_one` object did not get saved to the db.
 
-    *Neeraj Singh*
+    *Alan Kennedy*
 
 
-## Rails 4.0.0 (June 25, 2013) ##
+## Rails 4.1.1 (May 6, 2014) ##
 
-*   Fix `add_column` with `array` option when using PostgreSQL. Fixes #10432
+*   No changes.
 
-*   Do not overwrite manually built records during one-to-one nested attribute assignment
 
-    For one-to-one nested associations, if you build the new (in-memory)
-    child object yourself before assignment, then the NestedAttributes
-    module will not overwrite it, e.g.:
+## Rails 4.1.0 (April 8, 2014) ##
 
-        class Member < ActiveRecord::Base
-          has_one :avatar
-          accepts_nested_attributes_for :avatar
+*   Fixed a problem where an enum would overwrite values of another enum
+    with the same name in an unrelated class.
 
-          def avatar
-            super || build_avatar(width: 200)
-          end
-        end
+    Fixes #14607.
 
-        member = Member.new
-        member.avatar_attributes = {icon: 'sad'}
-        member.avatar.width # => 200
+    *Evan Whalen*
 
-    *Olek Janiszewski*
 
-*   fixes bug introduced by #3329.  Now, when autosaving associations,
-    deletions happen before inserts and saves.  This prevents a 'duplicate
-    unique value' database error that would occur if a record being created had
-    the same value on a unique indexed field as that of a record being destroyed.
+*   Block a few default Class methods as scope name.
 
-    *Adam Anderson*
+    For instance, this will raise:
 
-*   Fix pending migrations error when loading schema and `ActiveRecord::Base.table_name_prefix`
-    is not blank.
+        scope :public, -> { where(status: 1) }
 
-    Call `assume_migrated_upto_version` on connection to prevent it from first
-    being picked up in `method_missing`.
+    *arthurnn*
 
-    In the base class, `Migration`, `method_missing` expects the argument to be a
-    table name, and calls `proper_table_name` on the arguments before sending to
-    `connection`. If `table_name_prefix` or `table_name_suffix` is used, the schema
-    version changes to `prefix_version_suffix`, breaking `rake test:prepare`.
+*   Deprecate SQLite database URLs containing an
+    authority.
 
-    Fixes #10411.
+    The current "correct" spellings for in-memory, relative, and
+    absolute URLs, respectively, are:
 
-    *Kyle Stevens*
+        sqlite3::memory:
+        sqlite3:relative/path
+        sqlite3:/full/path
 
-*   Mute `psql` output when running rake db:schema:load.
+    The previous spelling (`sqlite3:///relative/path`) continues to work
+    as it did in Rails 4.0, but with a deprecation warning: in the next
+    release, that spelling will instead be interpreted as an absolute
+    path.
 
-    *Godfrey Chan*
+    *Matthew Draper*
 
-*   Trigger a save on `has_one association=(associate)` when the associate contents have changed.
+*   `where.not` adds `references` for `includes` like normal `where` calls do.
 
-    Fix #8856.
+    Fixes #14406.
 
-    *Chris Thompson*
+    *Yves Senn*
 
-*   Allow to use databases.rake tasks without having `Rails.application`.
+*   `includes` is able to detect the right preloading strategy when string
+    joins are involved.
 
-    *Piotr Sarnacki*
+    Fixes #14109.
 
-*   Fix a `SystemStackError` problem when using time zone aware or serialized attributes.
-    In current implementation, we reuse `column_types` argument when initiating an instance.
-    If an instance has serialized or time zone aware attributes, `column_types` is
-    wrapped multiple times in `decorate_columns` method. Thus the above error occurs.
+    *Aaron Patterson*, *Yves Senn*
 
-    *Dan Erikson & kennyj*
+*   Fixed error with validation with enum fields for records where the
+    value for any enum attribute is always evaluated as 0 during
+    uniqueness validation.
 
-*   Fix for a regression bug in which counter cache columns were not being updated
-    when record was pushed into a has_many association. For example:
+    Fixes #14172.
 
-        Post.first.comments << Comment.create
+    *Vilius Luneckas* *Ahmed AbouElhamayed*
 
-    Fixes #3891.
+*   `before_add` callbacks are fired before the record is saved on
+    `has_and_belongs_to_many` assocations *and* on `has_many :through`
+    associations.  Before this change, `before_add` callbacks would be fired
+    before the record was saved on `has_and_belongs_to_many` associations, but
+    *not* on `has_many :through` associations.
 
-    *Matthew Robertson*
+    Fixes #14144.
 
-*   If a model was instantiated from the database using `select`, `respond_to?`
-    returns false for non-selected attributes. For example:
+*   Fixed STI classes not defining an attribute method if there is a
+    conflicting private method defined on its ancestors.
 
-        post = Post.select(:title).first
-        post.respond_to?(:body) # => false
+    Fixes #11569.
 
-        post = Post.select('title as post_title').first
-        post.respond_to?(:title) # => false
+    *Godfrey Chan*
 
-    Fixes #4208.
+*   Default scopes are no longer overriden by chained conditions.
 
-    *Neeraj Singh*
+    Before this change when you defined a `default_scope` in a model
+    it was overriden by chained conditions in the same field. Now it
+    is merged like any other scope.
 
-*   Run `rake migrate:down` & `rake migrate:up` in transaction if database supports.
+    Before:
 
-    *Alexander Bondarev*
+        class User < ActiveRecord::Base
+          default_scope { where state: 'pending' }
+          scope :active, -> { where state: 'active' }
+          scope :inactive, -> { where state: 'inactive' }
+        end
 
-*   `0x` prefix must be added when assigning hexadecimal string into `bit` column in PostgreSQL.
+        User.all
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'pending'
 
-    *kennyj*
+        User.active
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'active'
 
-*   Added Statement Cache to allow the caching of a single statement. The cache works by
-    duping the relation returned from yielding a statement, which allows skipping the AST
-    building phase for following executes. The cache returns results in array format.
+        User.where(state: 'inactive')
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'inactive'
 
-    Example:
+    After:
 
-        cache = ActiveRecord::StatementCache.new do
-          Book.where(name: "my book").limit(100)
+        class User < ActiveRecord::Base
+          default_scope { where state: 'pending' }
+          scope :active, -> { where state: 'active' }
+          scope :inactive, -> { where state: 'inactive' }
         end
 
-        books = cache.execute
+        User.all
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'pending'
 
-    The solution attempts to get closer to the speed of `find_by_sql` but still maintaining
-    the expressiveness of the Active Record queries.
+        User.active
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'pending' AND "users"."state" = 'active'
 
-    *Olli Rissanen*
+        User.where(state: 'inactive')
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'pending' AND "users"."state" = 'inactive'
 
-*   Preserve context while merging relations with join information.
+    To get the previous behavior it is needed to explicitly remove the
+    `default_scope` condition using `unscoped`, `unscope`, `rewhere` or
+    `except`.
 
-        class Comment < ActiveRecord::Base
-          belongs_to :post
-        end
+    Example:
 
-        class Author < ActiveRecord::Base
-          has_many :posts
+        class User < ActiveRecord::Base
+          default_scope { where state: 'pending' }
+          scope :active, -> { unscope(where: :state).where(state: 'active') }
+          scope :inactive, -> { rewhere state: 'inactive' }
         end
 
-        class Post < ActiveRecord::Base
-          belongs_to :author
-          has_many :comments
-        end
+        User.all
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'pending'
 
-    `Comment.joins(:post).merge(Post.joins(:author).merge(Author.where(:name => "Joe Blogs"))).all`
-    would fail with
-    `ActiveRecord::ConfigurationError: Association named 'author' was not found on Comment`.
+        User.active
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'active'
 
-    It is failing because `all` is being called on relation which looks like this after all
-    the merging: `{:joins=>[:post, :author], :where=>[#<Arel::Nodes::Equality: ....}`. In this
-    relation all the context that `Post` was joined with `Author` is lost and hence the error
-    that `author` was not found on `Comment`.
+        User.inactive
+        # SELECT "users".* FROM "users" WHERE "users"."state" = 'inactive'
 
-    The solution is to build `JoinAssociation` when two relations with join information are being
-    merged. And later while building the Arel use the previously built `JoinAssociation` record
-    in `JoinDependency#graft` to build the right from clause.
-    Fixes #3002.
+*   Perform necessary deeper encoding when hstore is inside an array.
 
-    *Jared Armstrong and Neeraj Singh*
+    Fixes #11135.
 
-*   `default_scopes?` is deprecated. Check for `default_scopes.empty?` instead.
+    *Josh Goodall*, *Genadi Samokovarov*
 
-    *Agis Anastasopoulos*
+*   Properly detect if a connection is still active before using it
+    in multi-threaded environments.
 
-*   Default values for PostgreSQL bigint types now get parsed and dumped to the
-    schema correctly.
+    Fixes #12867.
 
-    *Erik Peterson*
+    *Kevin Casey*, *Matthew Draper*, *William (B.J.) Snow Orvis*
 
-*   Fix associations with `:inverse_of` option when building association
-    with a block. Inside the block the parent object was different then
-    after the block.
+*   When inverting add_index use the index name if present instead of
+    the columns.
 
-    Example:
+    If there are two indices with matching columns and one of them is
+    explicitly named then reverting the migration adding the named one
+    would instead drop the unnamed one.
 
-        parent.association.build do |child|
-          child.parent.equal?(parent) # false
-        end
+    The inversion of add_index will now drop the index by its name if
+    it is present.
 
-        # vs
+    *Hubert Dąbrowski*
 
-        child = parent.association.build
-        child.parent.equal?(parent) # true
+*   Add flag to disable schema dump after migration.
 
-    *Michal Cichra*
+    Add a config parameter on Active Record named `dump_schema_after_migration`
+    which is true by default. Now schema dump does not happen at the
+    end of migration rake task if `dump_schema_after_migration` is false.
 
-*   `has_many` using `:through` now obeys the order clause mentioned in
-    through association.
-    Fixes #10016.
+    *Emil Soman*
 
-    *Neeraj Singh*
+*   `find_in_batches`, `find_each`, `Result#each` and `Enumerable#index_by` now
+    return an `Enumerator` that can calculate its size.
 
-*   `belongs_to :touch` behavior now touches old association when
-    transitioning to new association.
+    See also #13938.
 
-        class Passenger < ActiveRecord::Base
-          belongs_to :car, touch: true
-        end
+    *Marc-André Lafortune*
 
-        car_1 = Car.create
-        car_2 = Car.create
+*   Make sure transaction state gets reset after a commit operation on the record.
 
-        passenger = Passenger.create car: car_1
+    If a new transaction was open inside a callback, the record was loosing track
+    of the transaction level state, and it was leaking that state.
 
-        passenger.car = car_2
-        passenger.save
+    Fixes #12566.
 
-    Previously only car_2 would be touched. Now both car_1 and car_2
-    will be touched.
+    *arthurnn*
 
-    *Adam Gamble*
+*   Pass `has_and_belongs_to_many` `:autosave` option to
+    the underlying `has_many :through` association.
 
-*   Extract and deprecate Firebird / Sqlserver / Oracle database tasks, because
-    These tasks should be supported by 3rd-party adapter.
+    Fixes #13923.
 
-    *kennyj*
+    *Yves Senn*
 
-*   Allow `ActiveRecord::Base.connection_handler` to have thread affinity and be
-    settable, this effectively allows Active Record to be used in a multithreaded
-    setup with multiple connections to multiple databases.
+*   PostgreSQL implementation of `SchemaStatements#index_name_exists?`.
 
-    *Sam Saffron*
+    The database agnostic implementation does not detect with indexes that are
+    not supported by the ActiveRecord schema dumper. For example, expressions
+    indexes would not be detected.
 
-*   `rename_column` preserves `auto_increment` in MySQL migrations.
-    Fixes #3493.
+    Fixes #11018.
 
-    *Vipul A M*
+    *Jonathan Baudanza*
 
-*   PostgreSQL geometric type point is now supported by Active Record. Fixes #7324.
+*   Parsing PostgreSQL arrays with empty strings now works correctly.
 
-    *Martin Schuerrer*
+    Previously, if you tried to parse `{"1","","2","","3"}` the result
+    would be `["1","2","3"]`, removing the empty strings from the array,
+    which would be incorrect. Now it will correctly produce `["1","","2","","3"]`
+    as the result of parsing the above PostgreSQL array.
 
-*   Add support for concurrent indexing in PostgreSQL adapter via the
-    `algorithm: :concurrently` option.
+    Fixes #13907.
 
-        add_index(:people, :last_name, algorithm: :concurrently)
+    *Maurício Linhares*
 
-    Also add support for MySQL index algorithms (`COPY`, `INPLACE`,
-    `DEFAULT`) via the `:algorithm` option.
+*   Associations now raise `ArgumentError` on name conflicts.
 
-        add_index(:people, :last_name, algorithm: :copy) # or :inplace/:default
+    Dangerous association names conflicts include instance or class methods already
+    defined by `ActiveRecord::Base`.
 
-    *Dan McClain*
+    Example:
 
-*   Add support for fulltext and spatial indexes on MySQL tables with MyISAM database
-    engine via the `type: 'FULLTEXT'` / `type: 'SPATIAL'` option.
+        class Car < ActiveRecord::Base
+          has_many :errors
+        end
+        # Will raise ArgumentError.
 
-        add_index(:people, :last_name, type: 'FULLTEXT')
-        add_index(:people, :last_name, type: 'SPATIAL')
+    Fixes #13217.
 
-    *Ken Mazaika*
+    *Lauro Caetano*
 
-*   Add an `add_index` override in PostgreSQL adapter and MySQL adapter
-    to allow custom index type support.
-    Fixes #6101.
+*   Fix regressions on `select_*` methods.
+    When `select_*` methods receive a `Relation` object, they should be able to
+    get the arel/binds from it.
+    Also fix regressions on `select_rows` that was ignoring the binds.
 
-        add_index(:wikis, :body, :using => 'gin')
+    Fixes #7538, #12017, #13731, #12056.
 
-    *Stefan Huber* and *Doabit*
+    *arthurnn*
 
-*   After extraction of mass-assignment attributes (which protects [id, type]
-    by default) we can pass id to `update_attributes` and it will update
-    another record because id will be used in where statement. We never have
-    to change id in where statement because we try to set/replace fields for
-    already loaded record but we have to try to set new id for that record.
+*   Active Record objects can now be correctly dumped, loaded and dumped again
+    without issues.
 
-    *Dmitry Vorotilin*
+    Previously, if you did `YAML.dump`, `YAML.load` and then `YAML.dump` again
+    in an Active Record model that used serialization it would fail at the last
+    dump due to the fields not being correctly serialized before being dumped
+    to YAML. Now it is possible to dump and load the same object as many times
+    as needed without any issues.
 
-*   Models with multiple counter cache associations now update correctly on destroy.
-    See #7706.
+    Fixes #13861.
 
-    *Ian Young*
+    *Maurício Linhares*
 
-*   If `:inverse_of` is true on an association, then when one calls `find()` on
-    the association, Active Record will first look through the in-memory objects
-    in the association for a particular id. Then, it will go to the DB if it
-    is not found. This is accomplished by calling `find_by_scan` in
-    collection associations whenever `options[:inverse_of]` is not nil.
-    Fixes #9470.
+*   `find_in_batches` now returns an `Enumerator` when called without a block, so that it
+    can be chained with other `Enumerable` methods.
 
-    *John Wang*
+    *Marc-André Lafortune*
 
-*   `rake db:create` does not change permissions of the MySQL root user.
-    Fixes #8079.
+*   `enum` now raises on "dangerous" name conflicts.
 
-    *Yves Senn*
+    Dangerous name conflicts includes instance or class method conflicts
+    with methods defined within `ActiveRecord::Base` but not its ancestors,
+    as well as conflicts with methods generated by other enums on the same
+    class.
 
-*   The length of the `version` column in the `schema_migrations` table
-    created by the `mysql2` adapter is 191 if the encoding is "utf8mb4".
+    Fixes #13389.
 
-    The "utf8" encoding in MySQL has support for a maximum of 3 bytes per character,
-    and only contains characters from the BMP. The recently added
-    [utf8mb4](http://dev.mysql.com/doc/refman/5.5/en/charset-unicode-utf8mb4.html)
-    encoding extends the support to four bytes. As of this writing, said encoding
-    is supported in the betas of the `mysql2` gem.
+    *Godfrey Chan*
 
-    Setting the encoding to "utf8mb4" has
-    [a few implications](http://dev.mysql.com/doc/refman/5.5/en/charset-unicode-upgrading.html).
-    This change addresses the max length for indexes, which is 191 instead of 255.
+*   `scope` now raises on "dangerous" name conflicts.
 
-    *Xavier Noria*
+    Similar to dangerous attribute methods, a scope name conflict is
+    dangerous if it conflicts with an existing class method defined within
+    `ActiveRecord::Base` but not its ancestors.
 
-*   Counter caches on associations will now stay valid when attributes are
-    updated (not just when records are created or destroyed), for example,
-    when calling `update_attributes`. The following code now works:
+    See also #13389.
 
-        class Comment < ActiveRecord::Base
-          belongs_to :post, counter_cache: true
-        end
+    *Godfrey Chan*, *Philippe Creux*
 
-        class Post < ActiveRecord::Base
-          has_many :comments
-        end
+*   Correctly send an user provided statement to a `lock!()` call.
 
-        post = Post.create
-        comment = Comment.create
+        person.lock! 'FOR SHARE NOWAIT'
+        # Before: SELECT * ... LIMIT 1 FOR UPDATE
+        # After: SELECT * ... LIMIT 1 FOR SHARE NOWAIT
 
-        post.comments << comment
-        post.save.reload.comments_count # => 1
-        comment.update_attributes(post_id: nil)
+    Fixes #13788.
 
-        post.save.reload.comments_count # => 0
+    *Maurício Linhares*
 
-    Updating the id of a `belongs_to` object with the id of a new object will
-    also keep the count accurate.
+*   Handle aliased attributes `select()`, `order()` and `reorder()`.
 
-    *John Wang*
+    *Tsutomu Kuroda*
 
-*   Referencing join tables implicitly was deprecated. There is a
-    possibility that these deprecation warnings are shown even if you
-    don't make use of that feature. You can now disable the feature entirely.
-    Fixes #9712.
+*   Reset the collection association when calling `reset` on it.
 
-    Example:
+    Before:
 
-        # in your configuration
-        config.active_record.disable_implicit_join_references = true
+        post.comments.loaded? # => true
+        post.comments.reset
+        post.comments.loaded? # => true
 
-        # or directly
-        ActiveRecord::Base.disable_implicit_join_references = true
+    After:
 
-    *Yves Senn*
+        post.comments.loaded? # => true
+        post.comments.reset
+        post.comments.loaded? # => false
 
-*   The `:distinct` option for `Relation#count` is deprecated. You
-    should use `Relation#distinct` instead.
+    Fixes #13777.
 
-    Example:
+    *Kelsey Schlarman*
 
-        # Before
-        Post.select(:author_name).count(distinct: true)
+*   Make enum fields work as expected with the `ActiveModel::Dirty` API.
 
-        # After
-        Post.select(:author_name).distinct.count
+    Before this change, using the dirty API would have surprising results:
 
-    *Yves Senn*
+        conversation = Conversation.new
+        conversation.status = :active
+        conversation.status = :archived
+        conversation.status_was # => 0
 
-*   Rename `Relation#uniq` to `Relation#distinct`. `#uniq` is still
-    available as an alias but we encourage to use `#distinct` instead.
-    Also `Relation#uniq_value` is aliased to `Relation#distinct_value`,
-    this is a temporary solution and you should migrate to `distinct_value`.
+    After this change, the same code would result in:
 
-    *Yves Senn*
+        conversation = Conversation.new
+        conversation.status = :active
+        conversation.status = :archived
+        conversation.status_was # => "active"
 
-*   Fix quoting for sqlite migrations using `copy_table_contents` with binary
-    columns.
+    *Rafael Mendonça França*
 
-    These would fail with "SQLite3::SQLException: unrecognized token" because
-    the column was not being passed to `quote` so the data was not quoted
-    correctly.
+*   `has_one` and `belongs_to` accessors don't add ORDER BY to the queries
+    anymore.
 
-    *Matthew M. Boedicker*
+    Since Rails 4.0, we add an ORDER BY in the `first` method to ensure
+    consistent results among different database engines. But for singular
+    associations this behavior is not needed since we will have one record to
+    return. As this ORDER BY option can lead some performance issues we are
+    removing it for singular associations accessors.
 
-*   Promotes `change_column_null` to the migrations API. This macro sets/removes
-    `NOT NULL` constraints, and accepts an optional argument to replace existing
-    `NULL`s if needed. The adapters for SQLite, MySQL, PostgreSQL, and (at least)
-    Oracle, already implement this method.
+    Fixes #12623.
 
-    *Xavier Noria*
+    *Rafael Mendonça França*
 
-*   Uniqueness validation allows you to pass `:conditions` to limit
-    the constraint lookup.
+*   Prepend table name for column names passed to `Relation#select`.
 
     Example:
 
-        validates_uniqueness_of :title, conditions: -> { where('approved = ?', true) }
+        Post.select(:id)
+        # Before: => SELECT id FROM "posts"
+        # After: => SELECT "posts"."id" FROM "posts"
 
-    *Mattias Pfeiffer + Yves Senn*
+    *Yves Senn*
 
-*   `connection` is deprecated as an instance method.
-    This allows end-users to have a `connection` method on their models
-    without clashing with Active Record internals.
+*   Fail early with "Primary key not included in the custom select clause"
+    in `find_in_batches`.
 
-    *Ben Moss*
+    Before this patch, the exception was raised after the first batch was
+    yielded to the block. This means that you only get it, when you hit the
+    `batch_size` treshold. This could shadow the issue in development.
 
-*   When copying migrations, preserve their magic comments and content encoding.
+    *Alexander Balashov*
 
-    *OZAWA Sakuro*
+*   Ensure `second` through `fifth` methods act like the `first` finder.
 
-*   Fix `subclass_from_attrs` when `eager_load` is false. It cannot find
-    subclass because all classes are loaded automatically when it needs.
+    The famous ordinal Array instance methods defined in ActiveSupport
+    (`first`, `second`, `third`, `fourth`, and `fifth`) are now available as
+    full-fledged finders in ActiveRecord. The biggest benefit of this is ordering
+    of the records returned now defaults to the table's primary key in ascending order.
 
-    *Dmitry Vorotilin*
+    Fixes #13743.
 
-*   When `:name` option is provided to `remove_index`, use it if there is no
-    index by the conventional name.
+    Example:
 
-    For example, previously if an index was removed like so
-    `remove_index :values, column: :value, name: 'a_different_name'`
-    the generated SQL would not contain the specified index name,
-    and hence the migration would fail.
-    Fixes #8858.
+        User.all.second
 
-    *Ezekiel Smithburg*
+        # Before
+        # => 'SELECT  "users".* FROM "users"'
 
-*   Created block to by-pass the prepared statement bindings.
-    This will allow to compose fragments of large SQL statements to
-    avoid multiple round-trips between Ruby and the DB.
+        # After
+        # => SELECT  "users".* FROM "users"   ORDER BY "users"."id" ASC LIMIT 1 OFFSET 1'
 
-    Example:
+        User.offset(3).second
 
-        sql = Post.connection.unprepared_statement do
-          Post.first.comments.to_sql
-        end
+        # Before
+        # => 'SELECT "users".* FROM "users"  LIMIT -1 OFFSET 3' # sqlite3 gem
+        # => 'SELECT "users".* FROM "users"  OFFSET 3' # pg gem
+        # => 'SELECT `users`.* FROM `users`  LIMIT 18446744073709551615 OFFSET 3' # mysql2 gem
 
-    *Cédric Fabianski*
+        # After
+        # => SELECT  "users".* FROM "users"   ORDER BY "users"."id" ASC LIMIT 1 OFFSET 4'
 
-*   Change the semantics of combining scopes to be the same as combining
-    class methods which return scopes. For example:
+    *Jason Meller*
 
-        class User < ActiveRecord::Base
-          scope :active,   -> { where state: 'active' }
-          scope :inactive, -> { where state: 'inactive' }
-        end
+*   ActiveRecord states are now correctly restored after a rollback for
+    models that did not define any transactional callbacks (i.e.
+    `after_commit`, `after_rollback` or `after_create`).
 
-        class Post < ActiveRecord::Base
-          def self.active
-            where state: 'active'
-          end
+    Fixes #13744.
 
-          def self.inactive
-            where state: 'inactive'
-          end
-        end
+    *Godfrey Chan*
+
+*   Make `touch` fire the `after_commit` and `after_rollback` callbacks.
 
-        ### BEFORE ###
+    *Harry Brundage*
 
-        User.where(state: 'active').where(state: 'inactive')
-        # => SELECT * FROM users WHERE state = 'active' AND state = 'inactive'
+*   Enable partial indexes for `sqlite >= 3.8.0`.
 
-        User.active.inactive
-        # => SELECT * FROM users WHERE state = 'inactive'
+    See http://www.sqlite.org/partialindex.html
 
-        Post.active.inactive
-        # => SELECT * FROM posts WHERE state = 'active' AND state = 'inactive'
+    *Cody Cutrer*
 
-        ### AFTER ###
+*   Don't try to get the subclass if the inheritance column doesn't exist
 
-        User.active.inactive
-        # => SELECT * FROM posts WHERE state = 'active' AND state = 'inactive'
+    The `subclass_from_attrs` method is called even if the column specified by
+    the `inheritance_column` setting doesn't exist. This prevents setting associations
+    via the attributes hash if the association name clashes with the value of the setting,
+    typically `:type`. This worked previously in Rails 3.2.
 
-    Before this change, invoking a scope would merge it into the current
-    scope and return the result. `Relation#merge` applies "last where
-    wins" logic to de-duplicate the conditions, but this lead to
-    confusing and inconsistent behaviour. This fixes that.
+    *Ujjwal Thaakar*
 
-    If you really do want the "last where wins" logic, you can opt-in to
-    it like so:
+*   Enum mappings are now exposed via class methods instead of constants.
 
-        User.active.merge(User.inactive)
+    Example:
 
-    Fixes #7365.
+        class Conversation < ActiveRecord::Base
+          enum status: [ :active, :archived ]
+        end
 
-    *Neeraj Singh* and *Jon Leighton*
+    Before:
 
-*   Expand `#cache_key` to consult all relevant updated timestamps.
+        Conversation::STATUS # => { "active" => 0, "archived" => 1 }
 
-    Previously only `updated_at` column was checked, now it will
-    consult other columns that received updated timestamps on save,
-    such as `updated_on`.  When multiple columns are present it will
-    use the most recent timestamp.
-    Fixes #9033.
+    After:
 
-    *Brendon Murphy*
+        Conversation.statuses # => { "active" => 0, "archived" => 1 }
 
-*   Throw `NotImplementedError` when trying to instantiate `ActiveRecord::Base` or an abstract class.
+    *Godfrey Chan*
 
-    *Aaron Weiner*
+*   Set `NameError#name` when STI-class-lookup fails.
 
-*   Warn when `rake db:structure:dump` with a MySQL database and
-    `mysqldump` is not in the PATH or fails.
-    Fixes #9518.
+    *Chulki Lee*
 
-    *Yves Senn*
+*   Fix bug in `becomes!` when changing from the base model to a STI sub-class.
 
-*   Remove `connection#structure_dump`, which is no longer used. *Yves Senn*
+    Fixes #13272.
 
-*   Make it possible to execute migrations without a transaction even
-    if the database adapter supports DDL transactions.
-    Fixes #9483.
+    *the-web-dev*, *Yves Senn*
 
-    Example:
+*   Currently Active Record can be configured via the environment variable
+    `DATABASE_URL` or by manually injecting a hash of values which is what Rails does,
+    reading in `database.yml` and setting Active Record appropriately. Active Record
+    expects to be able to use `DATABASE_URL` without the use of Rails, and we cannot
+    rip out this functionality without deprecating. This presents a problem though
+    when both config is set, and a `DATABASE_URL` is present. Currently the
+    `DATABASE_URL` should "win" and none of the values in `database.yml` are
+    used. This is somewhat unexpected, if one were to set values such as
+    `pool` in the `production:` group of `database.yml` they are ignored.
+
+    There are many ways that Active Record initiates a connection today:
+
+    - Stand Alone (without rails)
+      - `rake db:<tasks>`
+      - `ActiveRecord.establish_connection`
+
+    - With Rails
+      - `rake db:<tasks>`
+      - `rails <server> | <console>`
+      - `rails dbconsole`
+
+    Now all of these behave exactly the same way. The best way to do
+    this is to put all of this logic in one place so it is guaranteed to be used.
+
+    Here is the matrix of how this behavior works:
+
+    ```
+    No database.yml
+    No DATABASE_URL
+    => Error
+    ```
+
+    ```
+    database.yml present
+    No DATABASE_URL
+    => Use database.yml configuration
+    ```
+
+    ```
+    No database.yml
+    DATABASE_URL present
+    => use DATABASE_URL configuration
+    ```
+
+    ```
+    database.yml present
+    DATABASE_URL present
+    => Merged into `url` sub key. If both specify `url` sub key, the `database.yml` `url`
+       sub key "wins". If other paramaters `adapter` or `database` are specified in YAML,
+       they are discarded as the `url` sub key "wins".
+    ```
+
+    Current implementation uses `ActiveRecord::Base.configurations` to resolve and merge
+    all connection information before returning. This is achieved through a utility
+    class: `ActiveRecord::ConnectionHandling::MergeAndResolveDefaultUrlConfig`.
+
+    To understand the exact behavior of this class, it is best to review the
+    behavior in `activerecord/test/cases/connection_adapters/connection_handler_test.rb`.
 
-        class ChangeEnum < ActiveRecord::Migration
-          disable_ddl_transaction!
+    *Richard Schneeman*
 
-          def up
-            execute "ALTER TYPE model_size ADD VALUE 'new_value'"
-          end
-        end
+*   Make `change_column_null` revertable. Fixes #13576.
 
-    *Yves Senn*
+    *Yves Senn*, *Nishant Modak*, *Prathamesh Sonpatki*
 
-*   Assigning "0.0" to a nullable numeric column does not make it dirty.
-    Fixes #9034.
+*   Don't create/drop the test database if RAILS_ENV is specified explicitly.
 
-    Example:
+    Previously, when the environment was development, we would always
+    create or drop both the test and development databases.
 
-        product = Product.create price: 0.0
-        product.price = '0.0'
-        product.changed? # => false (this used to return true)
-        product.changes # => {} (this used to return { price: [0.0, 0.0] })
+    Now, if RAILS_ENV is explicitly defined as development, we don't create
+    the test database.
 
-    *Yves Senn*
+    *Damien Mathieu*
 
-*   Added functionality to unscope relations in a relations chain. For
-    instance, if you are passed in a chain of relations as follows:
+*   Initialize version on Migration objects so that it can be used in a migration,
+    and it will be included in the announce message.
 
-        User.where(name: "John").order('id DESC')
+    *Dylan Thacker-Smith*
 
-    but you want to get rid of order, then this feature allows you to do:
+*   `change_table` now uses the current adapter's `update_table_definition`
+    method to retrieve a specific table definition.
+    This ensures that `change_table` and `create_table` will use
+    similar objects.
 
-        User.where(name: "John").order('id DESC').unscope(:order)
-            == User.where(name: "John")
+    Fixes #13577, #13503.
 
-    The .unscope() function is more general than the .except() method because
-    .except() only works on the relation it is acting on. However, .unscope()
-    works for any relation in the entire relation chain.
+    *Nishant Modak*, *Prathamesh Sonpatki*, *Rafael Mendonça França*
 
-    *John Wang*
+*   Fixed ActiveRecord::Store nil conversion TypeError when using YAML coder.
+    In case the YAML passed as paramter is nil, uses an empty string.
 
-*   PostgreSQL timestamp with time zone (timestamptz) datatype now returns a
-    ActiveSupport::TimeWithZone instance instead of a string
+    Fixes #13570.
 
-    *Troy Kruthoff*
+    *Thales Oliveira*
 
-*   The `#append` method for collection associations behaves like`<<`.
-    `#prepend` is not defined and `<<` or `#append` should be used.
-    Fixes #7364.
+*   Deprecate unused `ActiveRecord::Base.symbolized_base_class`
+    and `ActiveRecord::Base.symbolized_sti_name` without replacement.
 
     *Yves Senn*
 
-*   Added support for creating a table via Rails migration generator.
-    For example,
+*   Since the `test_help.rb` file in Railties now automatically maintains
+    your test schema, the `rake db:test:*` tasks are deprecated. This
+    doesn't stop you manually running other tasks on your test database
+    if needed:
 
-        rails g migration create_books title:string content:text
+        rake db:schema:load RAILS_ENV=test
 
-    will generate a migration that creates a table called books with
-    the listed attributes, without creating a model.
-
-    *Sammy Larbi*
+    *Jon Leighton*
 
-*   Fix bug that raises the wrong exception when the exception handled by PostgreSQL adapter
-    doesn't respond to `#result`.
-    Fixes #8617.
+*   Fix presence validator for association when the associated record responds to `to_a`.
 
-    *kennyj*
+    *gmarik*
 
-*   Support PostgreSQL specific column types when using `change_table`.
-    Fixes #9480.
+*   Fixed regression on preload/includes with multiple arguments failing in certain conditions,
+    raising a NoMethodError internally by calling `reflect_on_association` for `NilClass:Class`.
 
-    Example:
+    Fixes #13437.
 
-        change_table :authors do |t|
-          t.hstore :books
-          t.json :metadata
-        end
+    *Vipul A M*, *khustochka*
 
-    *Yves Senn*
+*   Add the ability to nullify the `enum` column.
 
-*   Revert 408227d9c5ed7d, 'quote numeric'. This introduced some regressions.
+     Example:
 
-    *Steve Klabnik*
+         class Conversation < ActiveRecord::Base
+           enum gender: [:female, :male]
+         end
 
-*   Fix calculation of `db_runtime` property in
-   `ActiveRecord::Railties::ControllerRuntime#cleanup_view_runtime`.
-    Previously, after raising `ActionView::MissingTemplate`, `db_runtime` was
-    not populated.
-    Fixes #9215.
+         Conversation::GENDER # => { female: 0, male: 1 }
 
-    *Igor Fedoronchuk*
+         # conversation.update! gender: 0
+         conversation.female!
+         conversation.female? # => true
+         conversation.gender  # => "female"
 
-*   Do not try to touch invalid (and thus not persisted) parent record
-    for a `belongs_to :parent, touch: true` association
+         # conversation.update! gender: nil
+         conversation.gender = nil
+         conversation.gender.nil? # => true
+         conversation.gender      # => nil
 
-    *Olek Janiszewski*
+     *Amr Tamimi*
 
-*   Fix when performing an ordered join query. The bug only
-    affected queries where the order was given with a symbol.
-    Fixes #9275.
+*   Connection specification now accepts a "url" key. The value of this
+    key is expected to contain a database URL. The database URL will be
+    expanded into a hash and merged.
 
-    Example:
+    *Richard Schneeman*
 
-        # This will expand the order :name to "authors".name.
-        Author.joins(:books).where('books.published = 1').order(:name)
+*   An `ArgumentError` is now raised on a call to `Relation#where.not(nil)`.
 
-*   Fix overriding of attributes by `default_scope` on `ActiveRecord::Base#dup`.
+    Example:
 
-    *Hiroshige UMINO*
+        User.where.not(nil)
 
-*   Update queries now use prepared statements.
+        # Before
+        # => 'SELECT `users`.* FROM `users`  WHERE (NOT (NULL))'
 
-    *Olli Rissanen*
+        # After
+        # => ArgumentError, 'Invalid argument for .where.not(), got nil.'
 
-*   Fixing issue #8345. Now throwing an error when one attempts to touch a
-    new object that has not yet been persisted. For instance:
+    *Kuldeep Aggarwal*
 
-    Example:
+*   Deprecated use of string argument as a configuration lookup in
+    `ActiveRecord::Base.establish_connection`. Instead, a symbol must be given.
 
-        ball = Ball.new
-        ball.touch :updated_at   # => raises error
+    *José Valim*
 
-    It is not until the ball object has been persisted that it can be touched.
-    This follows the behavior of update_column.
+*   Fixed `update_column`, `update_columns`, and `update_all` to correctly serialize
+    values for `array`, `hstore` and `json` column types in PostgreSQL.
 
-    *John Wang*
+    Fixes #12261.
 
-*   Preloading ordered `has_many :through` associations no longer applies
-    invalid ordering to the `:through` association.
-    Fixes #8663.
+    *Tadas Tamosauskas*, *Carlos Antonio da Silva*
 
-    *Yves Senn*
+*   Do not consider PostgreSQL array columns as number or text columns.
 
-*   The auto explain feature has been removed. This feature was
-    activated by configuring `config.active_record.auto_explain_threshold_in_seconds`.
-    The configuration option was deprecated and has no more effect.
+    The code uses these checks in several places to know what to do with a
+    particular column, for instance AR attribute query methods has a branch
+    like this:
 
-    You can still use `ActiveRecord::Relation#explain` to see the EXPLAIN output for
-    any given relation.
+        if column.number?
+          !value.zero?
+        end
 
-    *Yves Senn*
+    This should never be true for array columns, since it would be the same
+    as running [].zero?, which results in a NoMethodError exception.
 
-*   The `:on` option for `after_commit` and `after_rollback` now
-    accepts an Array of actions.
-    Fixes #988.
+    Fixing this by ensuring that array columns in PostgreSQL never return
+    true for number?/text? checks.
 
-    Example:
+    *Carlos Antonio da Silva*
 
-        after_commit :update_cache on: [:create, :update]
+*   When connecting to a non-existant database, the error:
+    `ActiveRecord::NoDatabaseError` will now be raised. When being used with Rails
+    the error message will include information on how to create a database:
+    `rake db:create`. Supported adapters: postgresql, mysql, mysql2, sqlite3
 
-    *Yves Senn*
+    *Richard Schneeman*
 
-*   Rename related indexes on `rename_table` and `rename_column`. This
-    does not affect indexes with custom names.
+*   Do not raise `'cannot touch on a new record object'` exception on destroying
+    already destroyed `belongs_to` association with `touch: true` option.
 
-    *Yves Senn*
+    Fixes #13445.
 
-*   Prevent the creation of indices with too long names, which cause
-    internal operations to fail (sqlite3 adapter only). The method
-    `allowed_index_name_length` defines the length limit enforced by
-    rails. It's value defaults to `index_name_length` but can vary per adapter.
-    Fixes #8264.
+    Example:
 
-    *Yves Senn*
+        # Given Comment has belongs_to :post, touch: true
+        comment.post.destroy
+        comment.destroy # no longer raises an error
 
-*   Fixing issue #776.
+    *Paul Nikitochkin*
 
-    Memory bloat in transactions is handled by having the transaction hold only
-    the AR objects which it absolutely needs to know about. These are the AR
-    objects with callbacks (they need to be updated as soon as something in the
-    transaction occurs).
+*   Fix a bug when assigning an array containing string numbers to a
+    PostgreSQL integer array column.
 
-    All other AR objects can be updated lazily by keeping a reference to a
-    TransactionState object. If an AR object gets inside a transaction, then
-    the transaction will add its TransactionState to the AR object. When the
-    user makes a call to some attribute on an AR object (which has no
-    callbacks) associated with a transaction, the AR object will call the
-    sync_with_transaction_state method and make sure it is up to date with the
-    transaction. After it has synced with the transaction state, the AR object
-    will return the attribute that was requested.
+    Fixes #13444.
 
-    Most of the logic in the changes are used to handle multiple transactions,
-    in which case the AR object has to recursively follow parent pointers of
-    TransactionState objects.
+    Example:
 
-    *John Wang*
+        # Given Book#ratings is of type :integer, array: true
+        Book.new(ratings: [1, 2]) # worked before
+        Book.new(ratings: ['1', '2']) # now works as well
 
-*   Descriptive error message when the necessary AR adapter gem was not found.
-    Fixes #7313.
+    *Damien Mathieu*
 
-    *Yves Senn*
+*   Fix `PostgreSQL` insert to properly extract table name from multiline string SQL.
 
-*   Active Record now raises an error when blank arguments are passed to query
-    methods for which blank arguments do not make sense.
+    Previously, executing an insert SQL in `PostgreSQL` with a command like this:
 
-    Example:
+        insert into articles(
+          number)
+        values(
+          5152
+        )
 
-        Post.includes()     # => raises error
+    would not work because the adapter was unable to extract the correct `articles`
+    table name.
 
-    *John Wang*
+    *Kuldeep Aggarwal*
 
-*   Simplified type casting code for timezone aware attributes to use the
-    `in_time_zone` method if it is available. This introduces a subtle change
-    of behavior when using `Date` instances as they are directly converted to
-    `ActiveSupport::TimeWithZone` instances without first being converted to
-    `Time` instances. For example:
+*   Correctly escape PostgreSQL arrays.
 
-        # Rails 3.2 behavior
-        >> Date.today.to_time.in_time_zone
-        => Wed, 13 Feb 2013 07:00:00 UTC +00:00
+    Fixes: CVE-2014-0080
 
-        # Rails 4.0 behavior
-        >> Date.today.in_time_zone
-        => Wed, 13 Feb 2013 00:00:00 UTC +00:00
+*   `Relation` no longer has mutator methods like `#map!` and `#delete_if`. Convert
+    to an `Array` by calling `#to_a` before using these methods.
 
-    On the plus side it now behaves the same whether you pass a `String` date
-    or an actual `Date` instance. For example:
+    It intends to prevent odd bugs and confusion in code that call mutator
+    methods directly on the `Relation`.
 
-        # Rails 3.2 behavior
-        >> Date.civil(2013, 2, 13).to_time.in_time_zone
-        => Wed, 13 Feb 2013 07:00:00 UTC +00:00
-        >> Time.zone.parse("2013-02-13")
-        => Wed, 13 Feb 2013 00:00:00 UTC +00:00
+    Example:
 
-        # Rails 4.0 behavior
-        >> Date.civil(2013, 2, 13).in_time_zone
-        => Wed, 13 Feb 2013 00:00:00 UTC +00:00
-        >> "2013-02-13".in_time_zone
-        => Wed, 13 Feb 2013 00:00:00 UTC +00:00
+        # Instead of this
+        Author.where(name: 'Hank Moody').compact!
 
-    If you need the old behavior you can convert the dates to times manually.
-    For example:
+        # Now you have to do this
+        authors = Author.where(name: 'Hank Moody').to_a
+        authors.compact!
 
-        >> Post.new(created_at: Date.today).created_at
-        => Wed, 13 Feb 2013 00:00:00 UTC +00:00
+    *Lauro Caetano*
 
-        >> Post.new(created_at: Date.today.to_time).created_at
-        => Wed, 13 Feb 2013 07:00:00 UTC +00:00
+*   Better support for `where()` conditions that use a `belongs_to`
+    association name.
 
-    *Andrew White*
+    Using the name of an association in `where` previously worked only
+    if the value was a single `ActiveRecord::Base` object. e.g.
 
-*   Preloading `has_many :through` associations with conditions won't
-    cache the `:through` association. This will prevent invalid
-    subsets to be cached.
-    Fixes #8423.
+        Post.where(author: Author.first)
 
-    Example:
+    Any other values, including `nil`, would cause invalid SQL to be
+    generated. This change supports arguments in the `where` query
+    conditions where the key is a `belongs_to` association name and the
+    value is `nil`, an `Array` of `ActiveRecord::Base` objects, or an
+    `ActiveRecord::Relation` object.
 
-        class User
-          has_many :posts
-          has_many :recent_comments, -> { where('created_at > ?', 1.week.ago) }, :through => :posts
+        class Post < ActiveRecord::Base
+          belongs_to :author
         end
 
-        a_user = User.includes(:recent_comments).first
-
-        # This is preloaded.
-        a_user.recent_comments
+    `nil` value finds records where the association is not set:
 
-        # This is not preloaded, fetched now.
-        a_user.posts
+        Post.where(author: nil)
+        # SELECT "posts".* FROM "posts" WHERE "posts"."author_id" IS NULL
 
-    *Yves Senn*
+    `Array` values find records where the association foreign key
+    matches the ids of the passed ActiveRecord models, resulting
+    in the same query as `Post.where(author_id: [1,2])`:
 
-*   Don't run `after_commit` callbacks when creating through an association
-    if saving the record fails.
+        authors_array = [Author.find(1), Author.find(2)]
+        Post.where(author: authors_array)
+        # SELECT "posts".* FROM "posts" WHERE "posts"."author_id" IN (1, 2)
 
-    *James Miller*
+    `ActiveRecord::Relation` values find records using the same
+    query as `Post.where(author_id: Author.where(last_name: "Emde"))`
 
-*   Allow store accessors to be overridden like other attribute methods, e.g.:
+        Post.where(author: Author.where(last_name: "Emde"))
+        # SELECT "posts".* FROM "posts"
+        # WHERE "posts"."author_id" IN (
+        #   SELECT "authors"."id" FROM "authors"
+        #   WHERE "authors"."last_name" = 'Emde')
 
-        class User < ActiveRecord::Base
-          store :settings, accessors: [ :color, :homepage ], coder: JSON
+    Polymorphic `belongs_to` associations will continue to be handled
+    appropriately, with the polymorphic `association_type` field added
+    to the query to match the base class of the value. This feature
+    previously only worked when the value was a single `ActveRecord::Base`.
 
-          def color
-            super || 'red'
-          end
+        class Post < ActiveRecord::Base
+          belongs_to :author, polymorphic: true
         end
 
-    *Sergey Nartimov*
+        Post.where(author: Author.where(last_name: "Emde"))
+        # Generates a query similar to:
+        Post.where(author_id: Author.where(last_name: "Emde"), author_type: "Author")
 
-*   Quote numeric values being compared to non-numeric columns. Otherwise,
-    in some database, the string column values will be coerced to a numeric
-    allowing 0, 0.0 or false to match any string starting with a non-digit.
+    *Martin Emde*
 
-    Example:
-
-        App.where(apikey: 0) # => SELECT * FROM users WHERE apikey = '0'
-
-    *Dylan Smith*
-
-*   Schema dumper supports dumping the enabled database extensions to `schema.rb`
-    (currently only supported by PostgreSQL).
+*   Respect temporary option when dropping tables with MySQL.
 
-    *Justin George*
+    Normal DROP TABLE also works, but commits the transaction.
 
-*   The database adapters now converts the options passed thought `DATABASE_URL`
-    environment variable to the proper Ruby types before using. For example, SQLite requires
-    that the timeout value is an integer, and PostgreSQL requires that the
-    prepared_statements option is a boolean. These now work as expected:
+        drop_table :temporary_table, temporary: true
 
-    Example:
+    *Cody Cutrer*
 
-        DATABASE_URL=sqlite3://localhost/test_db?timeout=500
-        DATABASE_URL=postgresql://localhost/test_db?prepared_statements=false
+*   Add option to create tables from a query.
 
-    *Aaron Stone + Rafael Mendonça França*
+        create_table(:long_query, temporary: true,
+          as: "SELECT * FROM orders INNER JOIN line_items ON order_id=orders.id")
 
-*   `Relation#merge` now only overwrites where values on the LHS of the
-    merge. Consider:
+    Generates:
 
-        left  = Person.where(age: [13, 14, 15])
-        right = Person.where(age: [13, 14]).where(age: [14, 15])
+        CREATE TEMPORARY TABLE long_query AS
+          SELECT * FROM orders INNER JOIN line_items ON order_id=orders.id
 
-    `left` results in the following SQL:
+    *Cody Cutrer*
 
-        WHERE age IN (13, 14, 15)
+*   `db:test:clone` and `db:test:prepare` must load Rails environment.
 
-    `right` results in the following SQL:
+    `db:test:clone` and `db:test:prepare` use `ActiveRecord::Base`. configurations,
+    so we need to load the Rails environment, otherwise the config wont be in place.
 
-        WHERE age IN (13, 14) AND age IN (14, 15)
+    *arthurnn*
 
-    Previously, `left.merge(right)` would result in all but the last
-    condition being removed:
+*   Use the right column to type cast grouped calculations with custom expressions.
 
-        WHERE age IN (14, 15)
+    Fixes #13230.
 
-    Now it results in the LHS condition(s) for `age` being removed, but
-    the RHS remains as it is:
+    Example:
 
-        WHERE age IN (13, 14) AND age IN (14, 15)
+        # Before
+        Account.group(:firm_name).sum('0.01 * credit_limit')
+        # => { '37signals' => '0.5' }
 
-    *Jon Leighton*
+        # After
+        Account.group(:firm_name).sum('0.01 * credit_limit')
+        # => { '37signals' => 0.5 }
 
-*   Fix handling of dirty time zone aware attributes
+    *Paul Nikitochkin*
 
-    Previously, when `time_zone_aware_attributes` were enabled, after
-    changing a datetime or timestamp attribute and then changing it back
-    to the original value, `changed_attributes` still tracked the
-    attribute as changed. This caused `[attribute]_changed?` and
-    `changed?` methods to return true incorrectly.
+*   Polymorphic `belongs_to` associations with the `touch: true` option set update the timestamps of
+    the old and new owner correctly when moved between owners of different types.
 
     Example:
 
-        in_time_zone 'Paris' do
-          order = Order.new
-          original_time = Time.local(2012, 10, 10)
-          order.shipped_at = original_time
-          order.save
-          order.changed? # => false
+        class Rating < ActiveRecord::Base
+          belongs_to :rateable, polymorphic: true, touch: true
+        end
 
-          # changing value
-          order.shipped_at = Time.local(2013, 1, 1)
-          order.changed? # => true
+        rating = Rating.create rateable: Song.find(1)
+        rating.update_attributes rateable: Book.find(2) # => timestamps of Song(1) and Book(2) are updated
 
-          # reverting to original value
-          order.shipped_at = original_time
-          order.changed? # => false, used to return true
-        end
+    *Severin Schoepke*
+
+*   Improve formatting of migration exception messages: make them easier to read
+    with line breaks before/after, and improve the error for pending migrations.
 
-    *Lilibeth De La Cruz*
+    *John Bachir*
 
-*   When `#count` is used in conjunction with `#uniq` we perform `count(:distinct => true)`.
-    Fixes #6865.
+*   Fix `last` with `offset` to return the proper record instead of always the last one.
 
     Example:
 
-        relation.uniq.count # => SELECT COUNT(DISTINCT *)
+        Model.offset(4).last
+        # => returns the 4th record from the end.
 
-    *Yves Senn + Kaspar Schiess*
+    Fixes #7441.
 
-*   PostgreSQL ranges type support. Includes: int4range, int8range,
-    numrange, tsrange, tstzrange, daterange
+    *kostya*, *Lauro Caetano*
 
-    Ranges can be created with inclusive and exclusive bounds.
+*   `type_to_sql` returns a `String` for unmapped columns. This fixes an error
+    when using unmapped PostgreSQL array types.
 
     Example:
 
-        create_table :Room do |t|
-          t.daterange :availability
-        end
+        change_colum :table, :column, :bigint, array: true
+
+    Fixes #13146.
 
-        Room.create(availability: (Date.today..Float::INFINITY))
-        Room.first.availability # => Wed, 19 Sep 2012..Infinity
+    *Jens Fahnenbruck*, *Yves Senn*
 
-    One thing to note: Range class does not support exclusive lower
-    bound.
+*   Fix `QueryCache` to work with nested blocks, so that it will only clear the existing cache
+    after leaving the outer block instead of clearing it right after the inner block is finished.
 
-    *Alexander Grebennik*
+    *Vipul A M*
 
-*   Added a state instance variable to each transaction. Will allow other objects
-    to know whether a transaction has been committed or rolled back.
+*   The ERB in fixture files is no longer evaluated in the context of the main
+    object. Helper methods used by multiple fixtures should be defined on the
+    class object returned by `ActiveRecord::FixtureSet.context_class`.
 
-    *John Wang*
+    *Victor Costan*
 
-*   Collection associations `#empty?` always respects built records.
-    Fixes #8879.
+*   Previously, the `has_one` macro incorrectly accepted the `counter_cache`
+    option, but never actually supported it. Now it will raise an `ArgumentError`
+    when using `has_one` with `counter_cache`.
 
-    Example:
+    *Godfrey Chan*
 
-        widget = Widget.new
-        widget.things.build
-        widget.things.empty? # => false
+*   Implement `rename_index` natively for MySQL >= 5.7.
 
-    *Yves Senn*
+    *Cody Cutrer*
 
-*   Support for PostgreSQL's `ltree` data type.
+*   Fix bug when validating the uniqueness of an aliased attribute.
 
-    *Rob Worley*
+    Fixes #12402.
 
-*   Fix undefined method `to_i` when calling `new` on a scope that uses an
-    Array; Fix FloatDomainError when setting integer column to NaN.
-    Fixes #8718, #8734, #8757.
+    *Lauro Caetano*
 
-    *Jason Stirk + Tristan Harward*
+*   Update counter cache on a `has_many` relationship regardless of default scope.
 
-*   Rename `update_attributes` to `update`, keep `update_attributes` as an alias for `update` method.
-    This is a soft-deprecation for `update_attributes`, although it will still work without any
-    deprecation message in 4.0 is recommended to start using `update` since `update_attributes` will be
-    deprecated and removed in future versions of Rails.
+    Fixes #12952.
 
-    *Amparo Luna + Guillermo Iguaran*
+    *Uku Taht*
 
-*   `after_commit` and `after_rollback` now validate the `:on` option and raise an `ArgumentError`
-    if it is not one of `:create`, `:destroy` or `:update`
+*   `rename_index` adds the new index before removing the old one. This allows to
+    rename indexes on columns with a foreign key and prevents the following error:
 
-    *Pascal Friederich*
+        Cannot drop index 'index_engines_on_car_id': needed in a foreign key constraint
 
-*   Improve ways to write `change` migrations, making the old `up` & `down` methods no longer necessary.
+    *Cody Cutrer*, *Yves Senn*
 
-    * The methods `drop_table` and `remove_column` are now reversible, as long as the necessary information is given.
-      The method `remove_column` used to accept multiple column names; instead use `remove_columns` (which is not reversible).
-      The method `change_table` is also reversible, as long as its block doesn't call `remove`, `change` or `change_default`
+*   Raise `ActiveRecord::RecordNotDestroyed` when a replaced child
+    marked with `dependent: destroy` fails to be destroyed.
 
-    * New method `reversible` makes it possible to specify code to be run when migrating up or down.
-      See the [Guide on Migration](https://github.com/rails/rails/blob/master/guides/source/migrations.md#using-the-reversible-method)
+    Fixes #12812.
 
-    * New method `revert` will revert a whole migration or the given block.
-      If migrating down, the given migration / block is run normally.
-      See the [Guide on Migration](https://github.com/rails/rails/blob/master/guides/source/migrations.md#reverting-previous-migrations)
+    *Brian Thomas Storti*
 
-    Attempting to revert the methods `execute`, `remove_columns` and `change_column` will now
-    raise an `IrreversibleMigration` instead of actually executing them without any output.
+*   Fix validation on uniqueness of empty association.
 
-    *Marc-André Lafortune*
+    *Evgeny Li*
 
-*   Serialized attributes can be serialized in integer columns.
-    Fixes #8575.
+*   Make `ActiveRecord::Relation#unscope` affect relations it is merged in to.
 
-    *Rafael Mendonça França*
+    *Jon Leighton*
 
-*   Keep index names when using `alter_table` with sqlite3.
-    Fixes #3489.
+*   Use strings to represent non-string `order_values`.
 
     *Yves Senn*
 
-*   Add ability for PostgreSQL adapter to disable user triggers in `disable_referential_integrity`.
-    Fixes #5523.
-
-    *Gary S. Weaver*
+*   Checks to see if the record contains the foreign key to set the inverse automatically.
 
-*   Added support for `validates_uniqueness_of` in PostgreSQL array columns.
-    Fixes #8075.
+    *Edo Balvers*
 
-    *Pedro Padron*
+*   Added `ActiveRecord::Base.to_param` for convenient "pretty" URLs derived from a model's attribute or method.
 
-*   Allow int4range and int8range columns to be created in PostgreSQL and properly convert to/from database.
+    Example:
 
-    *Alexey Vasiliev aka leopard*
+        class User < ActiveRecord::Base
+          to_param :name
+        end
 
-*   Do not log the binding values for binary columns.
+        user = User.find_by(name: 'Fancy Pants')
+        user.id       # => 123
+        user.to_param # => "123-fancy-pants"
 
-    *Matthew M. Boedicker*
+    *Javan Makhmali*
 
-*   Fix counter cache columns not updated when replacing `has_many :through`
-    associations.
+*   Added `ActiveRecord::Base.no_touching`, which allows ignoring touch on models.
 
-    *Matthew Robertson*
+    Example:
 
-*   Recognize migrations placed in directories containing numbers and 'rb'.
-    Fixes #8492.
+        Post.no_touching do
+          Post.first.touch
+        end
 
-    *Yves Senn*
+    *Sam Stephenson*, *Damien Mathieu*
 
-*   Add `ActiveRecord::Base.cache_timestamp_format` class attribute to control
-    the format of the timestamp value in the cache key. Defaults to `:nsec`.
-    Fixes #8195.
+*   Prevent the counter cache from being decremented twice when destroying
+    a record on a `has_many :through` association.
 
-    *Rafael Mendonça França*
+    Fixes #11079.
 
-*   Session variables can be set for the `mysql`, `mysql2`, and `postgresql` adapters
-    in the `variables: <hash>` parameter in `config/database.yml`. The key-value pairs of this
-    hash will be sent in a `SET key = value` query on new database connections. See also:
-    http://dev.mysql.com/doc/refman/5.0/en/set-statement.html
-    http://www.postgresql.org/docs/8.3/static/sql-set.html
+    *Dmitry Dedov*
 
-    *Aaron Stone*
+*   Unify boolean type casting for `MysqlAdapter` and `Mysql2Adapter`.
+    `type_cast` will return `1` for `true` and `0` for `false`.
 
-*   Allow setting of all libpq connection parameters through the PostgreSQL adapter. See also:
-    http://www.postgresql.org/docs/9.2/static/libpq-connect.html#LIBPQ-PARAMKEYWORDS
+    Fixes #11119.
 
-    *Lars Kanis*
+    *Adam Williams*, *Yves Senn*
 
-*   Allow `Relation#where` with no arguments to be chained with new `not` query method.
+*   Fix bug where `has_one` association record update result in crash, when replaced with itself.
 
-    Example:
+    Fixes #12834.
 
-        Developer.where.not(name: 'Aaron')
+    *Denis Redozubov*, *Sergio Cambra*
 
-    *Akira Matsuda*
+*   Log bind variables after they are type casted. This makes it more
+    transparent what values are actually sent to the database.
 
-*   Unscope `update_column(s)` query to ignore default scope.
+        irb(main):002:0> Event.find("im-no-integer")
+        # Before: ... WHERE "events"."id" = $1 LIMIT 1  [["id", "im-no-integer"]]
+        # After: ... WHERE "events"."id" = $1 LIMIT 1  [["id", 0]]
 
-    When applying `default_scope` to a class with a where clause, using
-    `update_column(s)` could generate a query that would not properly update
-    the record due to the where clause from the `default_scope` being applied
-    to the update query.
+    *Yves Senn*
 
-        class User < ActiveRecord::Base
-          default_scope -> { where(active: true) }
-        end
+*   Fix uninitialized constant `TransactionState` error when `Marshall.load` is used on an Active Record result.
 
-        user = User.first
-        user.active = false
-        user.save!
+    Fixes #12790.
 
-        user.update_column(:active, true) # => false
+    *Jason Ayre*
 
-    In this situation we want to skip the default_scope clause and just
-    update the record based on the primary key. With this change:
+*   `.unscope` now removes conditions specified in `default_scope`.
 
-        user.update_column(:active, true) # => true
+    *Jon Leighton*
 
-    Fixes #8436.
+*   Added `ActiveRecord::QueryMethods#rewhere` which will overwrite an existing, named where condition.
 
-    *Carlos Antonio da Silva*
+    Examples:
 
-*   SQLite adapter no longer corrupts binary data if the data contains `%00`.
+        Post.where(trashed: true).where(trashed: false)                       #=> WHERE `trashed` = 1 AND `trashed` = 0
+        Post.where(trashed: true).rewhere(trashed: false)                     #=> WHERE `trashed` = 0
+        Post.where(active: true).where(trashed: true).rewhere(trashed: false) #=> WHERE `active` = 1 AND `trashed` = 0
 
-    *Chris Feist*
+    *DHH*
 
-*   Fix performance problem with `primary_key` method in PostgreSQL adapter when having many schemas.
-    Uses `pg_constraint` table instead of `pg_depend` table which has many records in general.
-    Fixes #8414.
+*   Extend `ActiveRecord::Base#cache_key` to take an optional list of timestamp attributes of which the highest will be used.
 
-    *kennyj*
+    Example:
 
-*   Do not instantiate intermediate Active Record objects when eager loading.
-    These records caused `after_find` to run more than expected.
-    Fixes #3313.
+        # last_reviewed_at will be used, if that's more recent than updated_at, or vice versa
+        Person.find(5).cache_key(:updated_at, :last_reviewed_at)
 
-    *Yves Senn*
+    *DHH*
 
-*   Add STI support to init and building associations.
-    Allows you to do `BaseClass.new(type: "SubClass")` as well as
-    `parent.children.build(type: "SubClass")` or `parent.build_child`
-    to initialize an STI subclass. Ensures that the class name is a
-    valid class and that it is in the ancestors of the super class
-    that the association is expecting.
+*   Added `ActiveRecord::Base#enum` for declaring enum attributes where the values map to integers in the database, but can be queried by name.
 
-    *Jason Rush*
+    Example:
 
-*   Observers was extracted from Active Record as `rails-observers` gem.
+        class Conversation < ActiveRecord::Base
+          enum status: [:active, :archived]
+        end
 
-    *Rafael Mendonça França*
+        Conversation::STATUS # => { active: 0, archived: 1 }
 
-*   Ensure that associations take a symbol argument. *Steve Klabnik*
+        # conversation.update! status: 0
+        conversation.active!
+        conversation.active? # => true
+        conversation.status  # => "active"
 
-*   Fix dirty attribute checks for `TimeZoneConversion` with nil and blank
-    datetime attributes. Setting a nil datetime to a blank string should not
-    result in a change being flagged.
-    Fixes #8310.
+        # conversation.update! status: 1
+        conversation.archived!
+        conversation.archived? # => true
+        conversation.status    # => "archived"
 
-    *Alisdair McDiarmid*
+        # conversation.update! status: 1
+        conversation.status = :archived
 
-*   Prevent mass assignment to the type column of polymorphic associations when using `build`
-    Fixes #8265.
+    *DHH*
 
-    *Yves Senn*
+*   `ActiveRecord::Base#attribute_for_inspect` now truncates long arrays (more than 10 elements).
 
-*   Deprecate calling `Relation#sum` with a block. To perform a calculation over
-    the array result of the relation, use `to_a.sum(&block)`.
+    *Jan Bernacki*
 
-    *Carlos Antonio da Silva*
+*   Allow for the name of the `schema_migrations` table to be configured.
 
-*   Fix PostgreSQL adapter to handle BC timestamps correctly
+    *Jerad Phelps*
 
-        HistoryEvent.create!(name: "something", occured_at: Date.new(0) - 5.years)
+*   Do not add to scope includes values from through associations.
+    Fixed bug when providing `includes` in through association scope, and fetching targets.
 
-    *Bogdan Gusiev*
+    Example:
 
-*   When running migrations on PostgreSQL, the `:limit` option for `binary` and `text` columns is silently dropped.
-    Previously, these migrations caused sql exceptions, because PostgreSQL doesn't support limits on these types.
+        class Vendor < ActiveRecord::Base
+          has_many :relationships, -> { includes(:user) }
+          has_many :users, through: :relationships
+        end
 
-    *Victor Costan*
+        vendor = Vendor.first
 
-*   Don't change STI type when calling `ActiveRecord::Base#becomes`.
-    Add `ActiveRecord::Base#becomes!` with the previous behavior.
+        # Before
 
-    See #3023 for more information.
+        vendor.users.to_a # => Raises exception: not found `:user` for `User`
 
-    *Thomas Hollstegge*
+        # After
 
-*   `rename_index` can be used inside a `change_table` block.
+        vendor.users.to_a # => No exception is raised
 
-        change_table :accounts do |t|
-          t.rename_index :user_id, :account_id
-        end
+    Fixes #12242, #9517, #10240.
 
-    *Jarek Radosz*
+    *Paul Nikitochkin*
 
-*   `#pluck` can be used on a relation with `select` clause. Fix #7551
+*   Type cast json values on write, so that the value is consistent
+    with reading from the database.
 
     Example:
 
-        Topic.select([:approved, :id]).order(:id).pluck(:id)
-
-    *Yves Senn*
+        x = JsonDataType.new tags: {"string" => "foo", :symbol => :bar}
 
-*   Do not create useless database transaction when building `has_one` association.
+        # Before:
+        x.tags # => {"string" => "foo", :symbol => :bar}
 
-    Example:
+        # After:
+        x.tags # => {"string" => "foo", "symbol" => "bar"}
 
-        User.has_one :profile
-        User.new.build_profile
+    *Severin Schoepke*
 
-    *Bogdan Gusiev*
+*   `ActiveRecord::Store` works together with PostgreSQL `hstore` columns.
 
-*   `:counter_cache` option for `has_many` associations to support custom named counter caches.
-    Fixes #7993.
+    Fixes #12452.
 
     *Yves Senn*
 
-*   Deprecate the possibility to pass a string as third argument of `add_index`.
-    Pass `unique: true` instead.
+*   Fix bug where `ActiveRecord::Store` used a global `Hash` to keep track of
+    all registered `stored_attributes`. Now every subclass of
+    `ActiveRecord::Base` has it's own `Hash`.
 
-        add_index(:users, :organization_id, unique: true)
-
-    *Rafael Mendonça França*
+    *Yves Senn*
 
-*   Raise an `ArgumentError` when passing an invalid option to `add_index`.
+*   Save `has_one` association when primary key is manually set.
 
-    *Rafael Mendonça França*
+    Fixes #12302.
 
-*   Fix `find_in_batches` crashing when IDs are strings and start option is not specified.
+    *Lauro Caetano*
 
-    *Alexis Bernard*
+*    Allow any version of BCrypt when using `has_secure_password`.
 
-*   `AR::Base#attributes_before_type_cast` now returns unserialized values for serialized attributes.
+     *Mike Perham*
 
-    *Nikita Afanasenko*
+*    Sub-query generated for `Relation` passed as array condition did not take in account
+     bind values and have invalid syntax.
 
-*   Use query cache/uncache when using `DATABASE_URL`.
-    Fixes #6951.
+     Generate sub-query with inline bind values.
 
-    *kennyj*
+     Fixes #12586.
 
-*   Fix bug where `update_columns` and `update_column` would not let you update the primary key column.
+     *Paul Nikitochkin*
 
-    *Henrik Nyh*
+*   Fix a bug where rake db:structure:load crashed when the path contained
+    spaces.
 
-*   The `create_table` method raises an `ArgumentError` when the primary key column is redefined.
-    Fixes #6378.
+    *Kevin Mook*
 
-    *Yves Senn*
+*   `ActiveRecord::QueryMethods#unscope` unscopes negative equality
 
-*   `ActiveRecord::AttributeMethods#[]` raises `ActiveModel::MissingAttributeError`
-    error if the given attribute is missing. Fixes #5433.
+    Allows you to call `#unscope` on a relation with negative equality
+    operators, i.e. `Arel::Nodes::NotIn` and `Arel::Nodes::NotEqual` that have
+    been generated through the use of `where.not`.
 
-        class Person < ActiveRecord::Base
-          belongs_to :company
-        end
+    *Eric Hankins*
 
-        # Before:
-        person = Person.select('id').first
-        person[:name]       # => nil
-        person.name         # => ActiveModel::MissingAttributeError: missing_attribute: name
-        person[:company_id] # => nil
-        person.company      # => nil
+*   Raise an exception when model without primary key calls `.find_with_ids`.
 
-        # After:
-        person = Person.select('id').first
-        person[:name]       # => ActiveModel::MissingAttributeError: missing_attribute: name
-        person.name         # => ActiveModel::MissingAttributeError: missing_attribute: name
-        person[:company_id] # => ActiveModel::MissingAttributeError: missing_attribute: company_id
-        person.company      # => ActiveModel::MissingAttributeError: missing_attribute: company_id
+    *Shimpei Makimoto*
 
-    *Francesco Rodriguez*
+*   Make `Relation#empty?` use `exists?` instead of `count`.
 
-*   Small binary fields use the `VARBINARY` MySQL type, instead of `TINYBLOB`.
+    *Szymon Nowak*
 
-    *Victor Costan*
+*   `rake db:structure:dump` no longer crashes when the port was specified as `Fixnum`.
 
-*   Decode URI encoded attributes on database connection URLs.
+    *Kenta Okamoto*
 
-    *Shawn Veader*
+*   `NullRelation#pluck` takes a list of columns
 
-*   Add `find_or_create_by`, `find_or_create_by!` and
-    `find_or_initialize_by` methods to `Relation`.
+    The method signature in `NullRelation` was updated to mimic that in
+    `Calculations`.
 
-    These are similar to the `first_or_create` family of methods, but
-    the behaviour when a record is created is slightly different:
+    *Derek Prior*
 
-        User.where(first_name: 'Penélope').first_or_create
+*   `scope_chain` should not be mutated for other reflections.
 
-    will execute:
+    Currently `scope_chain` uses same array for building different
+    `scope_chain` for different associations. During processing
+    these arrays are sometimes mutated and because of in-place
+    mutation the changed `scope_chain` impacts other reflections.
 
-        User.where(first_name: 'Penélope').create
+    Fix is to dup the value before adding to the `scope_chain`.
 
-    Causing all the `create` callbacks to execute within the context of
-    the scope. This could affect queries that occur within callbacks.
+    Fixes #3882.
 
-        User.find_or_create_by(first_name: 'Penélope')
+    *Neeraj Singh*
 
-    will execute:
+*   Prevent the inversed association from being reloaded on save.
 
-        User.create(first_name: 'Penélope')
+    Fixes #9499.
 
-    Which obviously does not affect the scoping of queries within
-    callbacks.
+    *Dmitry Polushkin*
 
-    The `find_or_create_by` version also reads better, frankly.
+*   Generate subquery for `Relation` if it passed as array condition for `where`
+    method.
 
-    If you need to add extra attributes during create, you can do one of:
+    Example:
 
-        User.create_with(active: true).find_or_create_by(first_name: 'Jon')
-        User.find_or_create_by(first_name: 'Jon') { |u| u.active = true }
+        # Before
+        Blog.where('id in (?)', Blog.where(id: 1))
+        # =>  SELECT "blogs".* FROM "blogs"  WHERE "blogs"."id" = 1
+        # =>  SELECT "blogs".* FROM "blogs"  WHERE (id IN (1))
 
-    The `first_or_create` family of methods have been nodoc'ed in favour
-    of this API. They may be deprecated in the future but their
-    implementation is very small and it's probably not worth putting users
-    through lots of annoying deprecation warnings.
+        # After
+        Blog.where('id in (?)', Blog.where(id: 1).select(:id))
+        # =>  SELECT "blogs".* FROM "blogs"
+        #     WHERE "blogs"."id" IN (SELECT "blogs"."id" FROM "blogs"  WHERE "blogs"."id" = 1)
 
-    *Jon Leighton*
+    Fixes #12415.
 
-*   Fix bug with presence validation of associations. Would incorrectly add duplicated errors
-    when the association was blank. Bug introduced in 1fab518c6a75dac5773654646eb724a59741bc13.
+    *Paul Nikitochkin*
 
-    *Scott Willson*
+*   For missed association exception message
+    which is raised in `ActiveRecord::Associations::Preloader` class
+    added owner record class name in order to simplify to find problem code.
 
-*   Fix bug where sum(expression) returns string '0' for no matching records.
-    Fixes #7439
+    *Paul Nikitochkin*
 
-    *Tim Macfarlane*
+*   `has_and_belongs_to_many` is now transparently implemented in terms of
+    `has_many :through`.  Behavior should remain the same, if not, it is a bug.
 
-*   PostgreSQL adapter correctly fetches default values when using multiple schemas and domains in a db. Fixes #7914
+*   `create_savepoint`, `rollback_to_savepoint` and `release_savepoint` accept
+    a savepoint name.
 
-    *Arturo Pie*
+    *Yves Senn*
 
-*   Learn ActiveRecord::QueryMethods#order work with hash arguments
+*   Make `next_migration_number` accessible for third party generators.
 
-    When symbol or hash passed we convert it to Arel::Nodes::Ordering.
-    If we pass invalid direction(like name: :DeSc) ActiveRecord::QueryMethods#order will raise an exception
+    *Yves Senn*
 
-        User.order(:name, email: :desc)
-        # SELECT "users".* FROM "users" ORDER BY "users"."name" ASC, "users"."email" DESC
+*   Objects instantiated using a null relationship will now retain the
+    attributes of the where clause.
 
-    *Tima Maslyuchenko*
+    Fixes #11676, #11675, #11376.
 
-*   Rename `ActiveRecord::Fixtures` class to `ActiveRecord::FixtureSet`.
-    Instances of this class normally hold a collection of fixtures (records)
-    loaded either from a single YAML file, or from a file and a folder
-    with the same name.  This change make the class name singular and makes
-    the class easier to distinguish from the modules like
-    `ActiveRecord::TestFixtures`, which operates on multiple fixture sets,
-    or `DelegatingFixtures`, `::Fixtures`, etc.,
-    and from the class `ActiveRecord::Fixture`, which corresponds to a single
-    fixture.
+    *Paul Nikitochkin*, *Peter Brown*, *Nthalk*
 
-    *Alexey Muranov*
+*   Fixed `ActiveRecord::Associations::CollectionAssociation#find`
+    when using `has_many` association with `:inverse_of` and finding an array of one element,
+    it should return an array of one element too.
 
-*   The postgres adapter now supports tables with capital letters.
-    Fixes #5920.
+    *arthurnn*
 
-    *Yves Senn*
+*   Callbacks on has_many should access the in memory parent if a inverse_of is set.
 
-*   `CollectionAssociation#count` returns `0` without querying if the
-    parent record is not persisted.
+    *arthurnn*
 
-    Before:
+*   `ActiveRecord::ConnectionAdapters.string_to_time` respects
+    string with timezone (e.g. Wed, 04 Sep 2013 20:30:00 JST).
 
-        person.pets.count
-        # SELECT COUNT(*) FROM "pets" WHERE "pets"."person_id" IS NULL
-        # => 0
+    Fixes #12278.
 
-    After:
+    *kennyj*
 
-        person.pets.count
-        # fires without sql query
-        # => 0
+*   Calling `update_attributes` will now throw an `ArgumentError` whenever it
+    gets a `nil` argument. More specifically, it will throw an error if the
+    argument that it gets passed does not respond to to `stringify_keys`.
 
-    *Francesco Rodriguez*
+    Example:
 
-*   Fix `reset_counters` crashing on `has_many :through` associations.
-    Fixes #7822.
+        @my_comment.update_attributes(nil)  # => raises ArgumentError
 
-    *lulalala*
+    *John Wang*
 
-*   Support for partial inserts.
+*   Deprecate `quoted_locking_column` method, which isn't used anywhere.
 
-    When inserting new records, only the fields which have been changed
-    from the defaults will actually be included in the INSERT statement.
-    The other fields will be populated by the database.
+    *kennyj*
 
-    This is more efficient, and also means that it will be safe to
-    remove database columns without getting subsequent errors in running
-    app processes (so long as the code in those processes doesn't
-    contain any references to the removed column).
+*   Migration dump UUID default functions to schema.rb.
 
-    The `partial_updates` configuration option is now renamed to
-    `partial_writes` to reflect the fact that it now impacts both inserts
-    and updates.
+    Fixes #10751.
 
-    *Jon Leighton*
+    *kennyj*
 
-*   Allow before and after validations to take an array of lifecycle events
+*   Fixed a bug in `ActiveRecord::Associations::CollectionAssociation#find_by_scan`
+    when using `has_many` association with `:inverse_of` option and UUID primary key.
 
-    *John Foley*
+    Fixes #10450.
 
-*   Support for specifying transaction isolation level
+    *kennyj*
 
-    If your database supports setting the isolation level for a transaction, you can set
-    it like so:
+*   Fix: joins association, with defined in the scope block constraints by using several
+    where constraints and at least of them is not `Arel::Nodes::Equality`,
+    generates invalid SQL expression.
 
-        Post.transaction(isolation: :serializable) do
-          # ...
-        end
+    Fixes #11963.
 
-    Valid isolation levels are:
+    *Paul Nikitochkin*
 
-    * `:read_uncommitted`
-    * `:read_committed`
-    * `:repeatable_read`
-    * `:serializable`
+*   `CollectionAssociation#first`/`#last` (e.g. `has_many`) use a `LIMIT`ed
+    query to fetch results rather than loading the entire collection.
 
-    You should consult the documentation for your database to understand the
-    semantics of these different levels:
+    *Lann Martin*
 
-    * http://www.postgresql.org/docs/9.1/static/transaction-iso.html
-    * https://dev.mysql.com/doc/refman/5.0/en/set-transaction.html
+*   Make possible to run SQLite rake tasks without the `Rails` constant defined.
 
-    An `ActiveRecord::TransactionIsolationError` will be raised if:
+    *Damien Mathieu*
 
-    * The adapter does not support setting the isolation level
-    * You are joining an existing open transaction
-    * You are creating a nested (savepoint) transaction
+*   Allow Relation#from to accept other relations with bind values.
 
-    The mysql, mysql2 and postgresql adapters support setting the transaction
-    isolation level. However, support is disabled for mysql versions below 5,
-    because they are affected by a bug (http://bugs.mysql.com/bug.php?id=39170)
-    which means the isolation level gets persisted outside the transaction.
+    *Ryan Wallace*
 
-    *Jon Leighton*
+*   Fix inserts with prepared statements disabled.
 
-*   `ActiveModel::ForbiddenAttributesProtection` is included by default
-    in Active Record models. Check the docs of `ActiveModel::ForbiddenAttributesProtection`
-    for more details.
+    Fixes #12023.
 
-    *Guillermo Iguaran*
+    *Rafael Mendonça França*
 
-*   Remove integration between Active Record and
-    `ActiveModel::MassAssignmentSecurity`, `protected_attributes` gem
-    should be added to use `attr_accessible`/`attr_protected`. Mass
-    assignment options has been removed from all the AR methods that
-    used it (ex. `AR::Base.new`, `AR::Base.create`, `AR::Base#update_attributes`, etc).
+*   Setting a has_one association on a new record no longer causes an empty
+    transaction.
 
-    *Guillermo Iguaran*
+    *Dylan Thacker-Smith*
 
-*   Fix the return of querying with an empty hash.
-    Fixes #6971.
+*   Fix `AR::Relation#merge` sometimes failing to preserve `readonly(false)` flag.
 
-        User.where(token: {})
+    *thedarkone*
 
-    Before:
+*   Re-use `order` argument pre-processing for `reorder`.
 
-        #=> SELECT * FROM users;
+    *Paul Nikitochkin*
 
-    After:
+*   Fix PredicateBuilder so polymorphic association keys in `where` clause can
+    accept objects other than direct descendants of `ActiveRecord::Base` (decorated
+    models, for example).
 
-        #=> SELECT * FROM users WHERE 1=0;
+    *Mikhail Dieterle*
 
-    *Damien Mathieu*
+*   PostgreSQL adapter recognizes negative money values formatted with
+    parentheses (eg. `($1.25) # => -1.25`)).
+    Fixes #11899.
 
-*   Fix creation of through association models when using `collection=[]`
-    on a `has_many :through` association from an unsaved model.
-    Fixes #7661.
+    *Yves Senn*
 
-    *Ernie Miller*
+*   Stop interpreting SQL 'string' columns as :string type because there is no
+    common STRING datatype in SQL.
 
-*   Explain only normal CRUD sql (select / update / insert / delete).
-    Fix problem that explains unexplainable sql.
-    Fixes #7544 #6458.
+    *Ben Woosley*
 
-    *kennyj*
+*   `ActiveRecord::FinderMethods#exists?` returns `true`/`false` in all cases.
 
-*   You can now override the generated accessor methods for stored attributes
-    and reuse the original behavior with `read_store_attribute` and `write_store_attribute`,
-    which are counterparts to `read_attribute` and `write_attribute`.
+    *Xavier Noria*
 
-    *Matt Jones*
+*   Assign inet/cidr attribute with `nil` value for invalid address.
 
-*   Accept `belongs_to` (including polymorphic) association keys in queries.
+    Example:
 
-    The following queries are now equivalent:
+        record = User.new
+        record.logged_in_from_ip # is type of an inet or a cidr
 
-        Post.where(author: author)
-        Post.where(author_id: author)
+        # Before:
+        record.logged_in_from_ip = 'bad ip address' # raise exception
 
-        PriceEstimate.where(estimate_of: treasure)
-        PriceEstimate.where(estimate_of_type: 'Treasure', estimate_of_id: treasure)
+        # After:
+        record.logged_in_from_ip = 'bad ip address' # do not raise exception
+        record.logged_in_from_ip # => nil
+        record.logged_in_from_ip_before_type_cast # => 'bad ip address'
 
-    *Peter Brown*
+    *Paul Nikitochkin*
 
-*   Use native `mysqldump` command instead of `structure_dump` method
-    when dumping the database structure to a sql file. Fixes #5547.
+*   `add_to_target` now accepts a second optional `skip_callbacks` argument
 
-    *kennyj*
+    If truthy, it will skip the :before_add and :after_add callbacks.
 
-*   PostgreSQL inet and cidr types are converted to `IPAddr` objects.
+    *Ben Woosley*
 
-    *Dan McClain*
+*   Fix interactions between `:before_add` callbacks and nested attributes
+    assignment of `has_many` associations, when the association was not
+    yet loaded:
 
-*   PostgreSQL array type support. Any datatype can be used to create an
-    array column, with full migration and schema dumper support.
+    - A `:before_add` callback was being called when a nested attributes
+      assignment assigned to an existing record.
 
-    To declare an array column, use the following syntax:
+    - Nested Attributes assignment did not affect the record in the
+      association target when a `:before_add` callback triggered the
+      loading of the association
 
-        create_table :table_with_arrays do |t|
-          t.integer :int_array, array: true
-          # integer[]
-          t.integer :int_array, array: true, length: 2
-          # smallint[]
-          t.string :string_array, array: true, length: 30
-          # char varying(30)[]
-        end
+    *Jörg Schray*
 
-    This respects any other migration detail (limits, defaults, etc).
-    Active Record will serialize and deserialize the array columns on
-    their way to and from the database.
+*   Allow enable_extension migration method to be revertible.
 
-    One thing to note: PostgreSQL does not enforce any limits on the
-    number of elements, and any array can be multi-dimensional. Any
-    array that is multi-dimensional must be rectangular (each sub array
-    must have the same number of elements as its siblings).
+    *Eric Tipton*
 
-    If the `pg_array_parser` gem is available, it will be used when
-    parsing PostgreSQL's array representation.
+*   Type cast hstore values on write, so that the value is consistent
+    with reading from the database.
 
-    *Dan McClain*
+    Example:
 
-*   Attribute predicate methods, such as `article.title?`, will now raise
-    `ActiveModel::MissingAttributeError` if the attribute being queried for
-    truthiness was not read from the database, instead of just returning `false`.
+        x = Hstore.new tags: {"bool" => true, "number" => 5}
 
-    *Ernie Miller*
+        # Before:
+        x.tags # => {"bool" => true, "number" => 5}
 
-*   `ActiveRecord::SchemaDumper` uses Ruby 1.9 style hash, which means that the
-    schema.rb file will be generated using this new syntax from now on.
+        # After:
+        x.tags # => {"bool" => "true", "number" => "5"}
 
-    *Konstantin Shabanov*
+    *Yves Senn* , *Severin Schoepke*
 
-*   Map interval with precision to string datatype in PostgreSQL. Fixes #7518.
+*   Fix multidimensional PostgreSQL arrays containing non-string items.
 
     *Yves Senn*
 
-*   Fix eagerly loading associations without primary keys. Fixes #4976.
-
-    *Kelley Reynolds*
+*   Fixes bug when using includes combined with select, the select statement was overwritten.
 
-*   Rails now raise an exception when you're trying to run a migration that has an invalid
-    file name. Only lower case letters, numbers, and '_' are allowed in migration's file name.
-    Please see #7419 for more details.
+    Fixes #11773.
 
-    *Jan Bernacki*
+    *Edo Balvers*
 
-*   Fix bug when calling `store_accessor` multiple times.
-    Fixes #7532.
+*   Load fixtures from linked folders.
 
-    *Matt Jones*
+    *Kassio Borges*
 
-*   Fix store attributes that show the changes incorrectly.
-    Fixes #7532.
+*   Create a directory for sqlite3 file if not present on the system.
 
-    *Matt Jones*
+    *Richard Schneeman*
 
-*   Fix `ActiveRecord::Relation#pluck` when columns or tables are reserved words.
+*   Removed redundant override of `xml` column definition for PostgreSQL,
+    in order to use `xml` column type instead of `text`.
 
-    *Ian Lesperance*
+    *Paul Nikitochkin*, *Michael Nikitochkin*
 
-*   Allow JSON columns to be created in PostgreSQL and properly encoded/decoded.
-    to/from database.
+*   Revert `ActiveRecord::Relation#order` change that make new order
+    prepend the old one.
 
-    *Dickson S. Guedes*
+    Before:
 
-*   Fix time column type casting for invalid time string values to correctly return `nil`.
+        User.order("name asc").order("created_at desc")
+        # SELECT * FROM users ORDER BY created_at desc, name asc
 
-    *Adam Meehan*
+    After:
 
-*   Allow to pass Symbol or Proc into `:limit` option of #accepts_nested_attributes_for.
+        User.order("name asc").order("created_at desc")
+        # SELECT * FROM users ORDER BY name asc, created_at desc
 
-    *Mikhail Dieterle*
+    This also affects order defined in `default_scope` or any kind of associations.
 
-*   ActiveRecord::SessionStore has been extracted from Active Record as `activerecord-session_store`
-    gem. Please read the `README.md` file on the gem for the usage.
+*   Add ability to define how a class is converted to Arel predicates.
+    For example, adding a very vendor specific regex implementation:
 
-    *Prem Sichanugrist*
+        regex_handler = proc do |column, value|
+          Arel::Nodes::InfixOperation.new('~', column, value.source)
+        end
+        ActiveRecord::PredicateBuilder.register_handler(Regexp, regex_handler)
 
-*   Fix `reset_counters` when there are multiple `belongs_to` association with the
-    same foreign key and one of them have a counter cache.
-    Fixes #5200.
+    *Sean Griffin & @joannecheng*
 
-    *Dave Desrochers*
+*   Don't allow `quote_value` to be called without a column.
 
-*   `serialized_attributes` and `_attr_readonly` become class method only. Instance reader methods are deprecated.
+    Some adapters require column information to do their job properly.
+    By enforcing the provision of the column for this internal method
+    we ensure that those using adapters that require column information
+    will always get the proper behavior.
 
-    *kennyj*
+    *Ben Woosley*
 
-*   Round usec when comparing timestamp attributes in the dirty tracking.
-    Fixes #6975.
+*   When using optimistic locking, `update` was not passing the column to `quote_value`
+    to allow the connection adapter to properly determine how to quote the value. This was
+    affecting certain databases that use specific column types.
 
-    *kennyj*
+    Fixes #6763.
 
-*   Use inversed parent for first and last child of `has_many` association.
+    *Alfred Wong*
 
-    *Ravil Bayramgalin*
+*   rescue from all exceptions in `ConnectionManagement#call`
 
-*   Fix `Column.microseconds` and `Column.fast_string_to_time` to avoid converting
-    timestamp seconds to a float, since it occasionally results in inaccuracies
-    with microsecond-precision times. Fixes #7352.
+    Fixes #11497.
 
-    *Ari Pollak*
+    As `ActiveRecord::ConnectionAdapters::ConnectionManagement` middleware does
+    not rescue from Exception (but only from StandardError), the Connection
+    Pool quickly runs out of connections when multiple erroneous Requests come
+    in right after each other.
 
-*   Fix AR#dup to nullify the validation errors in the dup'ed object. Previously the original
-    and the dup'ed object shared the same errors.
+    Rescuing from all exceptions and not just StandardError, fixes this
+    behaviour.
 
-    *Christian Seiler*
+    *Vipul A M*
 
-*   Raise `ArgumentError` if list of attributes to change is empty in `update_all`.
+*   `change_column` for PostgreSQL adapter respects the `:array` option.
 
-    *Roman Shatsov*
+    *Yves Senn*
 
-*   Fix AR#create to return an unsaved record when AR::RecordInvalid is
-    raised. Fixes #3217.
+*   Remove deprecation warning from `attribute_missing` for attributes that are columns.
 
-    *Dave Yeu*
+    *Arun Agrawal*
 
-*   Fixed table name prefix that is generated in engines for namespaced models.
+*   Remove extra decrement of transaction deep level.
 
-    *Wojciech Wnętrzak*
+    Fixes #4566.
 
-*   Make sure `:environment` task is executed before `db:schema:load` or `db:structure:load`.
-    Fixes #4772.
+    *Paul Nikitochkin*
 
-    *Seamus Abshere*
+*   Reset @column_defaults when assigning `locking_column`.
+    We had a potential problem. For example:
 
-*   Allow Relation#merge to take a proc.
+      class Post < ActiveRecord::Base
+        self.column_defaults  # if we call this unintentionally before setting locking_column ...
+        self.locking_column = 'my_locking_column'
+      end
 
-    This was requested by DHH to allow creating of one's own custom
-    association macros.
+      Post.column_defaults["my_locking_column"]
+      => nil # expected value is 0 !
 
-    For example:
+    *kennyj*
 
-        module Commentable
-          def has_many_comments(extra)
-            has_many :comments, -> { where(:foo).merge(extra) }
-          end
-        end
+*   Remove extra select and update queries on save/touch/destroy ActiveRecord model
+    with belongs to reflection with option `touch: true`.
 
-        class Post < ActiveRecord::Base
-          extend Commentable
-          has_many_comments -> { where(:bar) }
-        end
+    Fixes #11288.
 
-    *Jon Leighton*
+    *Paul Nikitochkin*
 
-*   Add CollectionProxy#scope.
+*   Remove deprecated nil-passing to the following `SchemaCache` methods:
+    `primary_keys`, `tables`, `columns` and `columns_hash`.
 
-    This can be used to get a Relation from an association.
+    *Yves Senn*
 
-    Previously we had a #scoped method, but we're deprecating that for
-    AR::Base, so it doesn't make sense to have it here.
+*   Remove deprecated block filter from `ActiveRecord::Migrator#migrate`.
 
-    This was requested by DHH, to facilitate code like this:
+    *Yves Senn*
 
-        Project.scope.order('created_at DESC').page(current_page).tagged_with(@tag).limit(5).scoping do
-          @topics      = @project.topics.scope
-          @todolists   = @project.todolists.scope
-          @attachments = @project.attachments.scope
-          @documents   = @project.documents.scope
-        end
+*   Remove deprecated String constructor from `ActiveRecord::Migrator`.
 
-    *Jon Leighton*
+    *Yves Senn*
 
-*   Add `Relation#load`.
+*   Remove deprecated `scope` use without passing a callable object.
 
-    This method explicitly loads the records and then returns `self`.
+    *Arun Agrawal*
 
-    Rather than deciding between "do I want an array or a relation?",
-    most people are actually asking themselves "do I want to eager load
-    or lazy load?" Therefore, this method provides a way to explicitly
-    eager-load without having to switch from a `Relation` to an array.
+*   Remove deprecated `transaction_joinable=` in favor of `begin_transaction`
+    with `:joinable` option.
 
-    Example:
+    *Arun Agrawal*
 
-        @posts = Post.where(published: true).load
+*   Remove deprecated `decrement_open_transactions`.
 
-    *Jon Leighton*
+    *Arun Agrawal*
 
-*   `Relation#order`: make new order prepend old one.
+*   Remove deprecated `increment_open_transactions`.
 
-        User.order("name asc").order("created_at desc")
-        # SELECT * FROM users ORDER BY created_at desc, name asc
+    *Arun Agrawal*
 
-    This also affects order defined in `default_scope` or any kind of associations.
+*   Remove deprecated `PostgreSQLAdapter#outside_transaction?`
+    method. You can use `#transaction_open?` instead.
 
-    *Bogdan Gusiev*
+    *Yves Senn*
 
-*   `Model.all` now returns an `ActiveRecord::Relation`, rather than an
-    array of records. Use `Relation#to_a` if you really want an array.
+*   Remove deprecated `ActiveRecord::Fixtures.find_table_name` in favor of
+    `ActiveRecord::Fixtures.default_fixture_model_name`.
 
-    In some specific cases, this may cause breakage when upgrading.
-    However in most cases the `ActiveRecord::Relation` will just act as a
-    lazy-loaded array and there will be no problems.
+    *Vipul A M*
 
-    Note that calling `Model.all` with options (e.g.
-    `Model.all(conditions: '...')` was already deprecated, but it will
-    still return an array in order to make the transition easier.
+*   Removed deprecated `columns_for_remove` from `SchemaStatements`.
 
-    `Model.scoped` is deprecated in favour of `Model.all`.
+    *Neeraj Singh*
 
-    `Relation#all` still returns an array, but is deprecated (since it
-    would serve no purpose if we made it return a `Relation`).
+*   Remove deprecated `SchemaStatements#distinct`.
 
-    *Jon Leighton*
+    *Francesco Rodriguez*
 
-*   `:finder_sql` and `:counter_sql` options on collection associations
-    are deprecated. Please transition to using scopes.
+*   Move deprecated `ActiveRecord::TestCase` into the rails test
+    suite. The class is no longer public and is only used for internal
+    Rails tests.
 
-    *Jon Leighton*
+    *Yves Senn*
 
-*   `:insert_sql` and `:delete_sql` options on `has_and_belongs_to_many`
-    associations are deprecated. Please transition to using `has_many
-    :through`.
+*   Removed support for deprecated option `:restrict` for `:dependent`
+    in associations.
 
-    *Jon Leighton*
+    *Neeraj Singh*
 
-*   Added `#update_columns` method which updates the attributes from
-    the passed-in hash without calling save, hence skipping validations and
-    callbacks. `ActiveRecordError` will be raised when called on new objects
-    or when at least one of the attributes is marked as read only.
+*   Removed support for deprecated `delete_sql` in associations.
 
-        post.attributes # => {"id"=>2, "title"=>"My title", "body"=>"My content", "author"=>"Peter"}
-        post.update_columns(title: 'New title', author: 'Sebastian') # => true
-        post.attributes # => {"id"=>2, "title"=>"New title", "body"=>"My content", "author"=>"Sebastian"}
+    *Neeraj Singh*
 
-    *Sebastian Martinez + Rafael Mendonça França*
+*   Removed support for deprecated `insert_sql` in associations.
 
-*   The migration generator now creates a join table with (commented) indexes every time
-    the migration name contains the word `join_table`:
+    *Neeraj Singh*
 
-        rails g migration create_join_table_for_artists_and_musics artist_id:index music_id
+*   Removed support for deprecated `finder_sql` in associations.
 
-    *Aleksey Magusev*
+    *Neeraj Singh*
 
-*   Add `add_reference` and `remove_reference` schema statements. Aliases, `add_belongs_to`
-    and `remove_belongs_to` are acceptable. References are reversible.
+*   Support array as root element in JSON fields.
 
-    Examples:
+    *Alexey Noskov & Francesco Rodriguez*
 
-        # Create a user_id column
-        add_reference(:products, :user)
-        # Create a supplier_id, supplier_type columns and appropriate index
-        add_reference(:products, :supplier, polymorphic: true, index: true)
-        # Remove polymorphic reference
-        remove_reference(:products, :supplier, polymorphic: true)
+*   Removed support for deprecated `counter_sql` in associations.
 
-    *Aleksey Magusev*
+    *Neeraj Singh*
 
-*   Add `:default` and `:null` options to `column_exists?`.
+*   Do not invoke callbacks when `delete_all` is called on collection.
 
-        column_exists?(:testings, :taggable_id, :integer, null: false)
-        column_exists?(:testings, :taggable_type, :string, default: 'Photo')
+    Method `delete_all` should not be invoking callbacks and this
+    feature was deprecated in Rails 4.0. This is being removed.
+    `delete_all` will continue to honor the `:dependent` option. However
+    if `:dependent` value is `:destroy` then the `:delete_all` deletion
+    strategy for that collection will be applied.
 
-    *Aleksey Magusev*
+    User can also force a deletion strategy by passing parameter to
+    `delete_all`. For example you can do `@post.comments.delete_all(:nullify)`.
 
-*   `ActiveRecord::Relation#inspect` now makes it clear that you are
-    dealing with a `Relation` object rather than an array:.
+    *Neeraj Singh*
 
-        User.where(age: 30).inspect
-        # => <ActiveRecord::Relation [#<User ...>, #<User ...>, ...]>
+*   Calling default_scope without a proc will now raise `ArgumentError`.
 
-        User.where(age: 30).to_a.inspect
-        # => [#<User ...>, #<User ...>]
+    *Neeraj Singh*
 
-    The number of records displayed will be limited to 10.
+*   Removed deprecated method `type_cast_code` from Column.
 
-    *Brian Cardarella, Jon Leighton & Damien Mathieu*
+    *Neeraj Singh*
 
-*   Add `collation` and `ctype` support to PostgreSQL. These are available for PostgreSQL 8.4 or later.
-    Example:
+*   Removed deprecated options `delete_sql` and `insert_sql` from HABTM
+    association.
 
-        development:
-          adapter: postgresql
-          host: localhost
-          database: rails_development
-          username: foo
-          password: bar
-          encoding: UTF8
-          collation: ja_JP.UTF8
-          ctype: ja_JP.UTF8
+    Removed deprecated options `finder_sql` and `counter_sql` from
+    collection association.
 
-    *kennyj*
+    *Neeraj Singh*
 
-*   Changed `validates_presence_of` on an association so that children objects
-    do not validate as being present if they are marked for destruction. This
-    prevents you from saving the parent successfully and thus putting the parent
-    in an invalid state.
+*   Remove deprecated `ActiveRecord::Base#connection` method.
+    Make sure to access it via the class.
 
-    *Nick Monje & Brent Wheeldon*
+    *Yves Senn*
 
-*   `FinderMethods#exists?` now returns `false` with the `false` argument.
+*   Remove deprecation warning for `auto_explain_threshold_in_seconds`.
 
-    *Egor Lynko*
+    *Yves Senn*
 
-*   Added support for specifying the precision of a timestamp in the PostgreSQL
-    adapter. So, instead of having to incorrectly specify the precision using the
-    `:limit` option, you may use `:precision`, as intended. For example, in a migration:
+*   Remove deprecated `:distinct` option from `Relation#count`.
 
-        def change
-          create_table :foobars do |t|
-            t.timestamps precision: 0
-          end
-        end
+    *Yves Senn*
 
-    *Tony Schneider*
+*   Removed deprecated methods `partial_updates`, `partial_updates?` and
+    `partial_updates=`.
 
-*   Allow `ActiveRecord::Relation#pluck` to accept multiple columns. Returns an
-    array of arrays containing the typecasted values:
+    *Neeraj Singh*
 
-        Person.pluck(:id, :name)
-        # SELECT people.id, people.name FROM people
-        # [[1, 'David'], [2, 'Jeremy'], [3, 'Jose']]
+*   Removed deprecated method `scoped`.
 
-    *Jeroen van Ingen & Carlos Antonio da Silva*
+    *Neeraj Singh*
 
-*   Improve the derivation of HABTM join table name to take account of nesting.
-    It now takes the table names of the two models, sorts them lexically and
-    then joins them, stripping any common prefix from the second table name.
+*   Removed deprecated method `default_scopes?`.
 
-    Some examples:
+    *Neeraj Singh*
 
-        Top level models (Category <=> Product)
-        Old: categories_products
-        New: categories_products
+*   Remove implicit join references that were deprecated in 4.0.
 
-        Top level models with a global table_name_prefix (Category <=> Product)
-        Old: site_categories_products
-        New: site_categories_products
+    Example:
 
-        Nested models in a module without a table_name_prefix method (Admin::Category <=> Admin::Product)
-        Old: categories_products
-        New: categories_products
+        # before with implicit joins
+        Comment.where('posts.author_id' => 7)
 
-        Nested models in a module with a table_name_prefix method (Admin::Category <=> Admin::Product)
-        Old: categories_products
-        New: admin_categories_products
+        # after
+        Comment.references(:posts).where('posts.author_id' => 7)
 
-        Nested models in a parent model (Catalog::Category <=> Catalog::Product)
-        Old: categories_products
-        New: catalog_categories_products
+    *Yves Senn*
 
-        Nested models in different parent models (Catalog::Category <=> Content::Page)
-        Old: categories_pages
-        New: catalog_categories_content_pages
+*   Apply default scope when joining associations. For example:
 
-    *Andrew White*
+        class Post < ActiveRecord::Base
+          default_scope -> { where published: true }
+        end
 
-*   Move HABTM validity checks to `ActiveRecord::Reflection`. One side effect of
-    this is to move when the exceptions are raised from the point of declaration
-    to when the association is built. This is consistent with other association
-    validity checks.
+        class Comment
+          belongs_to :post
+        end
 
-    *Andrew White*
+    When calling `Comment.joins(:post)`, we expect to receive only
+    comments on published posts, since that is the default scope for
+    posts.
 
-*   Added `stored_attributes` hash which contains the attributes stored using
-    `ActiveRecord::Store`. This allows you to retrieve the list of attributes
-    you've defined.
+    Before this change, the default scope from `Post` was not applied,
+    so we'd get comments on unpublished posts.
 
-       class User < ActiveRecord::Base
-         store :settings, accessors: [:color, :homepage]
-       end
+    *Jon Leighton*
 
-       User.stored_attributes[:settings] # [:color, :homepage]
+*   Remove `activerecord-deprecated_finders` as a dependency.
 
-    *Joost Baaij & Carlos Antonio da Silva*
+    *Łukasz Strzałkowski*
 
-*   PostgreSQL default log level is now 'warning', to bypass the noisy notice
-    messages. You can change the log level using the `min_messages` option
-    available in your config/database.yml.
+*   Remove Oracle / Sqlserver / Firebird database tasks that were deprecated in 4.0.
 
     *kennyj*
 
-*   Add uuid datatype support to PostgreSQL adapter.
-
-    *Konstantin Shabanov*
-
-*   Added `ActiveRecord::Migration.check_pending!` that raises an error if
-    migrations are pending.
-
-    *Richard Schneeman*
+*   `find_each` now returns an `Enumerator` when called without a block, so that it
+    can be chained with other `Enumerable` methods.
 
-*   Added `#destroy!` which acts like `#destroy` but will raise an
-    `ActiveRecord::RecordNotDestroyed` exception instead of returning `false`.
+    *Ben Woosley*
 
-    *Marc-André Lafortune*
+*   `ActiveRecord::Result.each` now returns an `Enumerator` when called without
+     a block, so that it can be chained with other `Enumerable` methods.
 
-*   Added support to `CollectionAssociation#delete` for passing `fixnum`
-    or `string` values as record ids. This finds the records responding
-    to the `id` and executes delete on them.
+    *Ben Woosley*
 
-        class Person < ActiveRecord::Base
-          has_many :pets
-        end
+*   Flatten merged join_values before building the joins.
 
-        person.pets.delete("1") # => [#<Pet id: 1>]
-        person.pets.delete(2, 3) # => [#<Pet id: 2>, #<Pet id: 3>]
+    While joining_values special treatment is given to string values.
+    By flattening the array it ensures that string values are detected
+    as strings and not arrays.
 
-    *Francesco Rodriguez*
+    Fixes #10669.
 
-*   Deprecated most of the 'dynamic finder' methods. All dynamic methods
-    except for `find_by_...` and `find_by_...!` are deprecated. Here's
-    how you can rewrite the code:
+    *Neeraj Singh and iwiznia*
 
-      * `find_all_by_...` can be rewritten using `where(...)`
-      * `find_last_by_...` can be rewritten using `where(...).last`
-      * `scoped_by_...` can be rewritten using `where(...)`
-      * `find_or_initialize_by_...` can be rewritten using
-        `where(...).first_or_initialize`
-      * `find_or_create_by_...` can be rewritten using
-        `find_or_create_by(...)` or where(...).first_or_create`
-      * `find_or_create_by_...!` can be rewritten using
-        `find_or_create_by!(...) or `where(...).first_or_create!`
+*   Do not load all child records for inverse case.
 
-    The implementation of the deprecated dynamic finders has been moved
-    to the `activerecord-deprecated_finders` gem. See below for details.
+    currently `post.comments.find(Comment.first.id)` would load all
+    comments for the given post to set the inverse association.
 
-    *Jon Leighton*
+    This has a huge performance penalty. Because if post has 100k
+    records and all these 100k records would be loaded in memory
+    even though the comment id was supplied.
 
-*   Deprecated the old-style hash based finder API. This means that
-    methods which previously accepted "finder options" no longer do. For
-    example this:
+    Fix is to use in-memory records only if loaded? is true. Otherwise
+    load the records using full sql.
 
-        Post.find(:all, conditions: { comments_count: 10 }, limit: 5)
+    Fixes #10509.
 
-    Should be rewritten in the new style which has existed since Rails 3:
+    *Neeraj Singh*
 
-        Post.where(comments_count: 10).limit(5)
+*   `inspect` on Active Record model classes does not initiate a
+    new connection. This means that calling `inspect`, when the
+    database is missing, will no longer raise an exception.
+    Fixes #10936.
 
-    Note that as an interim step, it is possible to rewrite the above as:
+    Example:
 
-        Post.all.merge(where: { comments_count: 10 }, limit: 5)
+        Author.inspect # => "Author(no database connection)"
 
-    This could save you a lot of work if there is a lot of old-style
-    finder usage in your application.
+    *Yves Senn*
 
-    `Relation#merge` now accepts a hash of
-    options, but they must be identical to the names of the equivalent
-    finder method. These are mostly identical to the old-style finder
-    option names, except in the following cases:
+*   Handle single quotes in PostgreSQL default column values.
+    Fixes #10881.
 
-      * `:conditions` becomes `:where`.
-      * `:include` becomes `:includes`.
+    *Dylan Markow*
 
-    The code to implement the deprecated features has been moved out to the
-    `activerecord-deprecated_finders` gem. This gem is a dependency of Active
-    Record in Rails 4.0, so the interface works out of the box. It will no
-    longer be a dependency from Rails 4.1 (you'll need to add it to the
-    `Gemfile` in 4.1), and will be maintained until Rails 5.0.
+*   Log the sql that is actually sent to the database.
 
-    *Jon Leighton*
+    If I have a query that produces sql
+    `WHERE "users"."name" = 'a         b'` then in the log all the
+    whitespace is being squeezed. So the sql that is printed in the
+    log is `WHERE "users"."name" = 'a b'`.
 
-*   It's not possible anymore to destroy a model marked as read only.
+    Do not squeeze whitespace out of sql queries. Fixes #10982.
 
-    *Johannes Barre*
+    *Neeraj Singh*
 
-*   Added ability to ActiveRecord::Relation#from to accept other ActiveRecord::Relation objects.
+*   Fixture setup no longer depends on `ActiveRecord::Base.configurations`.
+    This is relevant when `ENV["DATABASE_URL"]` is used in place of a `database.yml`.
 
-      Record.from(subquery)
-      Record.from(subquery, :a)
+    *Yves Senn*
 
-    *Radoslav Stankov*
+*   Fix mysql2 adapter raises the correct exception when executing a query on a
+    closed connection.
 
-*   Added custom coders support for ActiveRecord::Store. Now you can set
-    your custom coder like this:
+    *Yves Senn*
 
-        store :settings, accessors: [ :color, :homepage ], coder: JSON
+*   Ambiguous reflections are on :through relationships are no longer supported.
+    For example, you need to change this:
 
-    *Andrey Voronkov*
+        class Author < ActiveRecord::Base
+          has_many :posts
+          has_many :taggings, through: :posts
+        end
 
-*   `mysql` and `mysql2` connections will set `SQL_MODE=STRICT_ALL_TABLES` by
-    default to avoid silent data loss. This can be disabled by specifying
-    `strict: false` in your `database.yml`.
+        class Post < ActiveRecord::Base
+          has_one :tagging
+          has_many :taggings
+        end
 
-    *Michael Pearson*
+        class Tagging < ActiveRecord::Base
+        end
 
-*   Added default order to `first` to assure consistent results among
-    different database engines. Introduced `take` as a replacement to
-    the old behavior of `first`.
+    To this:
 
-    *Marcelo Silveira*
+        class Author < ActiveRecord::Base
+          has_many :posts
+          has_many :taggings, through: :posts, source: :tagging
+        end
 
-*   Added an `:index` option to automatically create indexes for references
-    and belongs_to statements in migrations.
+        class Post < ActiveRecord::Base
+          has_one :tagging
+          has_many :taggings
+        end
 
-    The `references` and `belongs_to` methods now support an `index`
-    option that receives either a boolean value or an options hash
-    that is identical to options available to the add_index method:
+        class Tagging < ActiveRecord::Base
+        end
 
-      create_table :messages do |t|
-        t.references :person, index: true
-      end
+    *Aaron Patterson*
 
-      Is the same as:
+*   Remove column restrictions for `count`, let the database raise if the SQL is
+    invalid. The previous behavior was untested and surprising for the user.
+    Fixes #5554.
 
-      create_table :messages do |t|
-        t.references :person
-      end
-      add_index :messages, :person_id
+    Example:
 
-    Generators have also been updated to use the new syntax.
+        User.select("name, username").count
+        # Before => SELECT count(*) FROM users
+        # After => ActiveRecord::StatementInvalid
 
-    *Joshua Wood*
+        # you can still use `count(:all)` to perform a query unrelated to the
+        # selected columns
+        User.select("name, username").count(:all) # => SELECT count(*) FROM users
 
-*   Added `#find_by` and `#find_by!` to mirror the functionality
-    provided by dynamic finders in a way that allows dynamic input more
-    easily:
+    *Yves Senn*
 
-        Post.find_by name: 'Spartacus', rating: 4
-        Post.find_by "published_at < ?", 2.weeks.ago
-        Post.find_by! name: 'Spartacus'
+*   Rails now automatically detects inverse associations. If you do not set the
+    `:inverse_of` option on the association, then Active Record will guess the
+    inverse association based on heuristics.
 
-    *Jon Leighton*
+    Note that automatic inverse detection only works on `has_many`, `has_one`,
+    and `belongs_to` associations. Extra options on the associations will
+    also prevent the association's inverse from being found automatically.
 
-*   Added ActiveRecord::Base#slice to return a hash of the given methods with
-    their names as keys and returned values as values.
+    The automatic guessing of the inverse association uses a heuristic based
+    on the name of the class, so it may not work for all associations,
+    especially the ones with non-standard names.
 
-    *Guillermo Iguaran*
+    You can turn off the automatic detection of inverse associations by setting
+    the `:inverse_of` option to `false` like so:
 
-*   Deprecate eager-evaluated scopes.
+        class Taggable < ActiveRecord::Base
+          belongs_to :tag, inverse_of: false
+        end
 
-    Don't use this:
+    *John Wang*
 
-        scope :red, where(color: 'red')
-        default_scope where(color: 'red')
+*   Fix `add_column` with `array` option when using PostgreSQL. Fixes #10432.
 
-    Use this:
+    *Adam Anderson*
 
-        scope :red, -> { where(color: 'red') }
-        default_scope { where(color: 'red') }
+*   Usage of `implicit_readonly` is being removed`. Please use `readonly` method
+    explicitly to mark records as `readonly.
+    Fixes #10615.
 
-    The former has numerous issues. It is a common newbie gotcha to do
-    the following:
+    Example:
 
-        scope :recent, where(published_at: Time.now - 2.weeks)
+        user = User.joins(:todos).select("users.*, todos.title as todos_title").readonly(true).first
+        user.todos_title = 'clean pet'
+        user.save! # will raise error
 
-    Or a more subtle variant:
+    *Yves Senn*
 
-        scope :recent, -> { where(published_at: Time.now - 2.weeks) }
-        scope :recent_red, recent.where(color: 'red')
+*   Fix the `:primary_key` option for `has_many` associations.
 
-    Eager scopes are also very complex to implement within Active
-    Record, and there are still bugs. For example, the following does
-    not do what you expect:
+    Fixes #10693.
 
-        scope :remove_conditions, except(:where)
-        where(...).remove_conditions # => still has conditions
+    *Yves Senn*
 
-    *Jon Leighton*
+*   Fix bug where tiny types are incorrectly coerced as boolean when the length is more than 1.
 
-*   Remove IdentityMap
+    Fixes #10620.
 
-    IdentityMap has never graduated to be an "enabled-by-default" feature, due
-    to some inconsistencies with associations, as described in this commit:
+    *Aaron Patterson*
 
-       https://github.com/rails/rails/commit/302c912bf6bcd0fa200d964ec2dc4a44abe328a6
+*   Also support extensions in PostgreSQL 9.1. This feature has been supported since 9.1.
 
-    Hence the removal from the codebase, until such issues are fixed.
+    *kennyj*
 
-    *Carlos Antonio da Silva*
+*   Deprecate `ConnectionAdapters::SchemaStatements#distinct`,
+    as it is no longer used by internals.
 
-*   Added the schema cache dump feature.
+    *Ben Woosley*
 
-    `Schema cache dump` feature was implemented. This feature can dump/load internal state of `SchemaCache` instance
-    because we want to boot rails more quickly when we have many models.
+*   Fix pending migrations error when loading schema and `ActiveRecord::Base.table_name_prefix`
+    is not blank.
 
-    Usage notes:
+    Call `assume_migrated_upto_version` on connection to prevent it from first
+    being picked up in `method_missing`.
 
-      1) execute rake task.
-      RAILS_ENV=production bundle exec rake db:schema:cache:dump
-      => generate db/schema_cache.dump
+    In the base class, `Migration`, `method_missing` expects the argument to be a
+    table name, and calls `proper_table_name` on the arguments before sending to
+    `connection`. If `table_name_prefix` or `table_name_suffix` is used, the schema
+    version changes to `prefix_version_suffix`, breaking `rake test:prepare`.
 
-      2) add config.active_record.use_schema_cache_dump = true in config/production.rb. BTW, true is default.
+    Fixes #10411.
 
-      3) boot rails.
-      RAILS_ENV=production bundle exec rails server
-      => use db/schema_cache.dump
+    *Kyle Stevens*
 
-      4) If you remove clear dumped cache, execute rake task.
-      RAILS_ENV=production bundle exec rake db:schema:cache:clear
-      => remove db/schema_cache.dump
+*   Method `read_attribute_before_type_cast` should accept input as symbol.
 
-    *kennyj*
+    *Neeraj Singh*
 
-*   Added support for partial indices to PostgreSQL adapter.
+*   Confirm a record has not already been destroyed before decrementing counter cache.
 
-    The `add_index` method now supports a `where` option that receives a
-    string with the partial index criteria.
+    *Ben Tucker*
 
-        add_index(:accounts, :code, where: 'active')
+*   Fixed a bug in `ActiveRecord#sanitize_sql_hash_for_conditions` in which
+    `self.class` is an argument to `PredicateBuilder#build_from_hash`
+    causing `PredicateBuilder` to call non-existent method
+    `Class#reflect_on_association`.
 
-    generates
+    *Zach Ohlgren*
 
-        CREATE INDEX index_accounts_on_code ON accounts(code) WHERE active
+*   While removing index if column option is missing then raise IrreversibleMigration exception.
 
-    *Marcelo Silveira*
+    Following code should raise `IrreversibleMigration`. But the code was
+    failing since options is an array and not a hash.
 
-*   Implemented `ActiveRecord::Relation#none` method.
+        def change
+          change_table :users do |t|
+            t.remove_index [:name, :email]
+          end
+        end
 
-    The `none` method returns a chainable relation with zero records
-    (an instance of the NullRelation class).
+    Fix was to check if the options is a Hash before operating on it.
 
-    Any subsequent condition chained to the returned relation will continue
-    generating an empty relation and will not fire any query to the database.
+    Fixes #10419.
 
-    *Juanjo Bazán*
+    *Neeraj Singh*
 
-*   Added the `ActiveRecord::NullRelation` class implementing the null
-    object pattern for the Relation class.
+*   Do not overwrite manually built records during one-to-one nested attribute assignment
 
-    *Juanjo Bazán*
+    For one-to-one nested associations, if you build the new (in-memory)
+    child object yourself before assignment, then the NestedAttributes
+    module will not overwrite it, e.g.:
 
-*   Added new `dependent: :restrict_with_error` option. This will add
-    an error to the model, rather than raising an exception.
+        class Member < ActiveRecord::Base
+          has_one :avatar
+          accepts_nested_attributes_for :avatar
 
-    The `:restrict` option is renamed to `:restrict_with_exception` to
-    make this distinction explicit.
+          def avatar
+            super || build_avatar(width: 200)
+          end
+        end
 
-    *Manoj Kumar & Jon Leighton*
+        member = Member.new
+        member.avatar_attributes = {icon: 'sad'}
+        member.avatar.width # => 200
 
-*   Added `create_join_table` migration helper to create HABTM join tables.
+    *Olek Janiszewski*
 
-        create_join_table :products, :categories
-        # =>
-        # create_table :categories_products, id: false do |td|
-        #   td.integer :product_id,  null: false
-        #   td.integer :category_id, null: false
-        # end
+*   fixes bug introduced by #3329. Now, when autosaving associations,
+    deletions happen before inserts and saves. This prevents a 'duplicate
+    unique value' database error that would occur if a record being created had
+    the same value on a unique indexed field as that of a record being destroyed.
 
-    *Rafael Mendonça França*
+    *Johnny Holton*
 
-*   The primary key is always initialized in the @attributes hash to `nil` (unless
-    another value has been specified).
+*   Handle aliased attributes in ActiveRecord::Relation.
 
-    *Aaron Patterson*
+    When using symbol keys, ActiveRecord will now translate aliased attribute names to the actual column name used in the database:
 
-*   In previous releases, the following would generate a single query with
-    an `OUTER JOIN comments`, rather than two separate queries:
+    With the model
 
-        Post.includes(:comments)
-            .where("comments.name = 'foo'")
+        class Topic
+          alias_attribute :heading, :title
+        end
 
-    This behaviour relies on matching SQL string, which is an inherently
-    flawed idea unless we write an SQL parser, which we do not wish to
-    do.
+    The call
 
-    Therefore, it is now deprecated.
+        Topic.where(heading: 'The First Topic')
 
-    To avoid deprecation warnings and for future compatibility, you must
-    explicitly state which tables you reference, when using SQL snippets:
+    should yield the same result as
 
-        Post.includes(:comments)
-            .where("comments.name = 'foo'")
-            .references(:comments)
+        Topic.where(title: 'The First Topic')
 
-    Note that you do not need to explicitly specify references in the
-    following cases, as they can be automatically inferred:
+    This also applies to ActiveRecord::Relation::Calculations calls such as `Model.sum(:aliased)` and `Model.pluck(:aliased)`.
 
-        Post.includes(:comments).where(comments: { name: 'foo' })
-        Post.includes(:comments).where('comments.name' => 'foo')
-        Post.includes(:comments).order('comments.name')
+    This will not work with SQL fragment strings like `Model.sum('DISTINCT aliased')`.
 
-    You do not need to worry about this unless you are doing eager
-    loading. Basically, don't worry unless you see a deprecation warning
-    or (in future releases) an SQL error due to a missing JOIN.
+    *Godfrey Chan*
 
-    *Jon Leighton*
+*   Mute `psql` output when running rake db:schema:load.
 
-*   Support for the `schema_info` table has been dropped. Please
-    switch to `schema_migrations`.
+    *Godfrey Chan*
 
-    *Aaron Patterson*
+*   Trigger a save on `has_one association=(associate)` when the associate contents have changed.
 
-*   Connections *must* be closed at the end of a thread. If not, your
-    connection pool can fill and an exception will be raised.
+    Fixes #8856.
 
-    *Aaron Patterson*
+    *Chris Thompson*
 
-*   PostgreSQL hstore records can be created.
+*   Abort a rake task when missing db/structure.sql like `db:schema:load` task.
 
-    *Aaron Patterson*
+    *kennyj*
 
-*   PostgreSQL hstore types are automatically deserialized from the database.
+*   rake:db:test:prepare falls back to original environment after execution.
 
-    *Aaron Patterson*
+    *Slava Markevich*
 
-Please check [3-2-stable](https://github.com/rails/rails/blob/3-2-stable/activerecord/CHANGELOG.md) for previous changes.
+Please check [4-0-stable](https://github.com/rails/rails/blob/4-0-stable/activerecord/CHANGELOG.md) for previous changes.