* quoted_date converts time-like objects to ActiveRecord::Base.default_timezone before serialization. This allows you to use Time.now in find conditions and have it correctly be serialized as the current time in UTC when default_timezone == :utc. #2946 [Geoff Buesing]
* Base.skip_time_zone_conversion_for_attributes uses class_inheritable_accessor, so that subclasses don't overwrite Base [#346 state:resolved] [Emilio Tagua]
* Internal API: configurable association options and build_association method for reflections so plugins may extend and override. #985 [Hongli Lai (Phusion)]
* before_save, before_validation and before_destroy callbacks that return false will now ROLLBACK the transaction. Previously this would have been committed before the processing was aborted. #891 [Xavier Noria]
* Set config.active_record.timestamped_migrations = false to have migrations with numeric prefix instead of UTC timestamp. #446. [Andrew Stone, Nik Wakelin]
* Base#instantiate_time_object: eliminate check for Time.zone, since we can assume this is set if time_zone_aware_attributes is set to true [Geoff Buesing]
* Time zone aware attribute methods use Time.zone.parse instead of #to_time for String arguments, so that offset information in String is respected. Resolves #105. [Scott Fleckenstein, Geoff Buesing]
* Added that the MySQL adapter should map integer to either smallint, int, or bigint depending on the :limit just like PostgreSQL [David Heinemeier Hansson]
* Refactor HasManyThroughAssociation to inherit from HasManyAssociation. Association callbacks and <association>_ids= now work with hm:t. #11516 [Ruy Asan]
* Switched to UTC-timebased version numbers for migrations and the schema. This will as good as eliminate the problem of multiple migrations getting the same version assigned in different branches. Also added rake db:migrate:up/down to apply individual migrations that may need to be run when you merge branches #11458 [John Barnette]
* Base#instantiate_time_object only uses Time.zone when Base.time_zone_aware_attributes is true; leverages Time#time_with_datetime_fallback for readability [Geoff Buesing]
* update_all ignores scoped :order and :limit, so post.comments.update_all doesn't try to include the comment order in the update statement. #10686 [Brendan Ribera]
* Added ActiveRecord::Base.cache_key to make it easier to cache Active Records in combination with the new ActiveSupport::Cache::* libraries [David Heinemeier Hansson]
* Anchor DateTimeTest to fixed DateTime instead of a variable value based on Time.now#advance#to_datetime, so that this test passes on 64-bit platforms running Ruby 1.8.6+ [Geoff Buesing]
* Fixed that the Query Cache should just be ignored if the database is misconfigured (so that the "About your applications environment" works even before the database has been created) [David Heinemeier Hansson]
* Reveal that the type option in migrations can be any supported column type for your database but also include caveat about agnosticism. Closes #7531 [adamwiggins, mikong]
* Assigning an instance of a foreign class to a composed_of aggregate calls an optional conversion block. Refactor and simplify composed_of implementation. #6322 [brandon, Chris Cruft]
* Fix regression where the association would not construct new finder SQL on save causing bogus queries for "WHERE owner_id = NULL" even after owner was saved. #8713 [Bryan Helmkamp]
* Fix that ActiveRecord would create attribute methods and override custom attribute getters if the method is also defined in Kernel.methods. [Rick Olson]
* Don't call attr_readonly on polymorphic belongs_to associations, in case it matches the name of some other non-ActiveRecord class/module. [Rick Olson]
* Try loading activerecord-<adaptername>-adapter gem before trying a plain require so you can use custom gems for the bundled adapters. Also stops gems from requiring an adapter from an old Active Record gem. [Jeremy Kemper, Derrick Spell]
* RubyGem database adapters: expects a gem named activerecord-<database>-adapter with active_record/connection_adapters/<database>_adapter.rb in its load path. [Jeremy Kemper]
- Generate Reader and Writer methods which cache attribute values in hashes. This is to avoid repeatedly parsing the same date or integer columns.
- Change exception raised when users use find with :select then try to access a skipped column. Plugins could override missing_attribute() to lazily load the columns.
- Move method definition to the class, instead of the instance
- Always generate the readers, writers and predicate methods.
* Remove spurious tests from deprecated_associations_test, most of these aren't deprecated, and are duplicated in associations_test. Closes #8987 [Pratik Naik]
* Load database adapters on demand. Eliminates config.connection_adapters and RAILS_CONNECTION_ADAPTERS. Add your lib directory to the $LOAD_PATH and put your custom adapter in lib/active_record/connection_adapters/adaptername_adapter.rb. This way you can provide custom adapters as plugins or gems without modifying Rails. [Jeremy Kemper]
* Use association name for the wrapper element when using .to_xml. Previous behavior lead to non-deterministic situations with STI and polymorphic associations. [Michael Koziarski, jstrachan]
* Add helpful debugging info to the ActiveRecord::StatementInvalid exception in ActiveRecord::ConnectionAdapters::SqliteAdapter#table_structure. Closes #7925. [court3nay]
* Remove deprecated object transactions. People relying on this functionality should install the object_transactions plugin at http://code.bitsweat.net/svn/object_transactions. Closes #5637 [Michael Koziarski, Jeremy Kemper]
* PostgreSQL: remove DateTime -> Time downcast. Warning: do not enable translate_results for the C bindings if you have timestamps outside Time's domain. [Jeremy Kemper]
* find_or_create_by_* takes a hash so you can create with more attributes than are in the method name. For example, Person.find_or_create_by_name(:name => 'Henry', :comments => 'Hi new user!') is equivalent to Person.find_by_name('Henry') || Person.create(:name => 'Henry', :comments => 'Hi new user!'). #7368 [Josh Susser]
* Make sure with_scope takes both :select and :joins into account when setting :readonly. Allows you to save records you retrieve using method_missing on a has_many :through associations. [Michael Koziarski]
* Reworked David's query cache to be available as Model.cache {...}. For the duration of the block no select query should be run more then once. Any inserts/deletes/executes will flush the whole cache however [Tobias Lütke]
* When dealing with SQLite3, use the table_info pragma helper, so that the bindings can do some translation for when sqlite3 breaks incompatibly between point releases. [Jamis Buck]
* update_all can take a Hash argument. sanitize_sql splits into two methods for conditions and assignment since NULL values and delimiters are handled differently. #6583, #7365 [sandofsky, Assaf]
* fix faulty inheritance tests and that eager loading grabs the wrong inheritance column when the class of your association is an STI subclass. Closes #6859 [protocool]
* Consolidated different create and create! versions to call through to the base class with scope. This fixes inconsistencies, especially related to protected attribtues. Closes #5847 [Alexander Dymo, Tobias Lütke]
* find supports :lock with :include. Check whether your database allows SELECT ... FOR UPDATE with outer joins before using. #6764 [vitaly, Jeremy Kemper]
* Add AssociationCollection#create! to be consistent with AssociationCollection#create when dealing with a foreign key that is a protected attribute [Cody Fauser]
* Added counter optimization for AssociationCollection#any? so person.friends.any? won't actually load the full association if we have the count in a cheaper form [David Heinemeier Hansson]
* validates_numericality_of uses \A \Z to ensure the entire string matches rather than ^ $ which may match one valid line of a multiline string. #5716 [Andreas Schwarz]
* MySQL: detect when a NOT NULL column without a default value is misreported as default ''. Can't detect for string, text, and binary columns since '' is a legitimate default. #6156 [simon@redhillconsulting.com.au, obrie, Jonathan Viney, Jeremy Kemper]
* Oracle: to increase performance, prefetch 100 rows and enable similar cursor sharing. Both are configurable in database.yml. #6607 [philbogle@gmail.com, ray.fortna@jobster.com, Michael Schoen]
* Consistently use LOWER() for uniqueness validations (rather than mixing with UPPER()) so the database can always use a functional index on the lowercased column. #6495 [Si]
* Oracle: resolve test failures, use prefetched primary key for inserts, check for null defaults, fix limited id selection for eager loading. Factor out some common methods from all adapters. #6515 [Michael Schoen]
* Don't rollback in teardown unless a transaction was started. Don't start a transaction in create_fixtures if a transaction is started. #6282 [Jacob Fugal, Jeremy Kemper]
* Removes the ability for eager loaded conditions to be interpolated, since there is no model instance to use as a context for interpolation. #5553 [turnip@turnipspatch.com]
* Added timeout option to SQLite3 configurations to deal more gracefully with SQLite3::BusyException, now the connection can instead retry for x seconds to see if the db clears up before throwing that exception #6126 [wreese@gmail.com]
* Association collections have an _ids reader method to match the existing writer for collection_select convenience (e.g. employee.task_ids). The writer method skips blank ids so you can safely do @employee.task_ids = params[:tasks] without checking every time for an empty list or blank values. #1887, #5780 [Michael Schuerig]
* Fixed the Ruby/MySQL adapter we ship with Active Record to work with the new authentication handshake that was introduced in MySQL 4.1, along with the other protocol changes made at that time #5723 [jimw@mysql.com]
* has_one supports the :dependent => :delete option which skips the typical callback chain and deletes the associated object directly from the database. #5927 [Chris Mear, Jonathan Viney]
* SQLServer: work around bug where some unambiguous date formats are not correctly identified if the session language is set to german. #5894 [Tom Ward, kruth@bfpi]
* Pushing a record onto a has_many :through sets the association's foreign key to the associate's primary key and adds it to the correct association. #5815, #5829 [Josh Susser]
* Add records to has_many :through using <<, push, and concat by creating the association record. Raise if base or associate are new records since both ids are required to create the association. #build raises since you can't associate an unsaved record. #create! takes an attributes hash and creates the associated record and its association in a transaction. [Jeremy Kemper]
* PostgreSQL: autodetected sequences work correctly with multiple schemas. Rely on the schema search_path instead of explicitly qualifying the sequence name with its schema. #5280 [guy.naor@famundo.com]
* SQLServer: added tests to ensure all database statements are closed, refactored identity_insert management code to use blocks, removed update/delete rowcount code out of execute and into update/delete, changed insert to go through execute method, removed unused quoting methods, disabled pessimistic locking tests as feature is currently unsupported, fixed RakeFile to load sqlserver specific tests whether running in ado or odbc mode, fixed support for recently added decimal types, added support for limits on integer types. #5670 [Tom Ward]
* Numeric and decimal columns map to BigDecimal instead of Float. Those with scale 0 map to Integer. #5454 [robbat2@gentoo.org, work@ashleymoran.me.uk]
* Oracle: use nonblocking queries if allow_concurrency is set, fix pessimistic locking, don't guess date vs. time by default (set OracleAdapter.emulate_dates = true for the old behavior), adapter cleanup. #5635 [Michael Schoen]
* Fixed a few Oracle issues: Allows Oracle's odd date handling to still work consistently within #to_xml, Passes test that hardcode insert statement by dropping the :id column, Updated RUNNING_UNIT_TESTS with Oracle instructions, Corrects method signature for #exec #5294 [Michael Schoen]
* Row locking. Provide a locking clause with the :lock finder option or true for the default "FOR UPDATE". Use the #lock! method to obtain a row lock on a single record (reloads the record with :lock => true). [Shugo Maeda]
# Obtain an exclusive lock on person 1 so we can safely increment visits.
Person.transaction do
# select * from people where id=1 for update
person = Person.find(1, :lock => true)
person.visits += 1
person.save!
end
* PostgreSQL: introduce allow_concurrency option which determines whether to use blocking or asynchronous #execute. Adapters with blocking #execute will deadlock Ruby threads. The default value is ActiveRecord::Base.allow_concurrency. [Jeremy Kemper]
* Use a per-thread (rather than global) transaction mutex so you may execute concurrent transactions on separate connections. [Jeremy Kemper]
* Refactored ActiveRecord::Base.to_xml to become a delegate for XmlSerializer, which restores sanity to the mega method. This refactoring also reinstates the opinions that type="string" is redundant and ugly and nil-differentiation is not a concern of serialization [David Heinemeier Hansson]
* Fixed that :includes merged from with_scope won't cause the same association to be loaded more than once if repetition occurs in the clauses #5253 [alex@purefiction.net]
* create! no longer blows up when no attributes are passed and a :create scope is in effect (e.g. foo.bars.create! failed whereas foo.bars.create!({}) didn't.) [Jeremy Kemper]
* PostgreSQL: migrations support :limit with :integer columns by mapping limit < 4 to smallint, > 4 to bigint, and anything else to integer. #2900 [keegan@thebasement.org]
* Add support for FrontBase (http://www.frontbase.com/) with a new adapter thanks to the hard work of one Mike Laster. Closes #4093. [mlaster@metavillage.com]
* When dealing with SQLite3, use the table_info pragma helper, so that the bindings can do some translation for when sqlite3 breaks incompatibly between point releases. [Jamis Buck]
* Consistently use LOWER() for uniqueness validations (rather than mixing with UPPER()) so the database can always use a functional index on the lowercased column. #6495 [Si]
* MySQL: SET SQL_AUTO_IS_NULL=0 so 'where id is null' doesn't select the last inserted id. #6778 [Jonathan Viney, timc]
* Fixtures use the table name and connection from set_fixture_class. #7330 [Anthony Eden]
* SQLServer: quote table name in indexes query. #2928 [keithm@infused.org]
* Subclass instantiation doesn't try to explicitly require the corresponding subclass. #6840 [leei, Jeremy Kemper]
* fix faulty inheritance tests and that eager loading grabs the wrong inheritance column when the class of your association is an STI subclass. Closes #6859 [protocool]
* find supports :lock with :include. Check whether your database allows SELECT ... FOR UPDATE with outer joins before using. #6764 [vitaly, Jeremy Kemper]
* Support nil and Array in :conditions => { attr => value } hashes. #6548 [Assaf, Jeremy Kemper]
* MySQL: detect when a NOT NULL column without a default value is misreported as default ''. Can't detect for string, text, and binary columns since '' is a legitimate default. #6156 [simon@redhillconsulting.com.au, obrie, Jonathan Viney, Jeremy Kemper]
* validates_numericality_of uses \A \Z to ensure the entire string matches rather than ^ $ which may match one valid line of a multiline string. #5716 [Andreas Schwarz]
* Oracle: automatically detect the primary key. #6594 [vesaria, Michael Schoen]
* Oracle: to increase performance, prefetch 100 rows and enable similar cursor sharing. Both are configurable in database.yml. #6607 [philbogle@gmail.com, ray.fortna@jobster.com, Michael Schoen]
* Oracle: resolve test failures, use prefetched primary key for inserts, check for null defaults, fix limited id selection for eager loading. Factor out some common methods from all adapters. #6515 [Michael Schoen]
* Make add_column use the options hash with the Sqlite Adapter. Closes #6464 [obrie]
* Document other options available to migration's add_column. #6419 [grg]
* MySQL: all_hashes compatibility with old MysqlRes class. #6429, #6601 [Jeremy Kemper]
* Fix has_many :through to add the appropriate conditions when going through an association using STI. Closes #5783. [Jonathan Viney]
* Don't rollback in teardown unless a transaction was started. Don't start a transaction in create_fixtures if a transaction is started. #6282 [Jacob Fugal, Jeremy Kemper]
* Add #delete support to has_many :through associations. Closes #6049 [Martin Landers]
* Removes the ability for eager loaded conditions to be interpolated, since there is no model instance to use as a context for interpolation. #5553 [turnip@turnipspatch.com]
* Added timeout option to SQLite3 configurations to deal more gracefully with SQLite3::BusyException, now the connection can instead retry for x seconds to see if the db clears up before throwing that exception #6126 [wreese@gmail.com]
* Added update_attributes! which uses save! to raise an exception if a validation error prevents saving #6192 [jonathan]
* The has_many create method works with polymorphic associations. #6361 [Dan Peterson]
* MySQL: introduce Mysql::Result#all_hashes to support further optimization. #5581 [Stefan Kaes]
* save! shouldn't validate twice. #6324 [maiha, Bob Silva]
* Association collections have an _ids reader method to match the existing writer for collection_select convenience (e.g. employee.task_ids). The writer method skips blank ids so you can safely do @employee.task_ids = params[:tasks] without checking every time for an empty list or blank values. #1887, #5780 [Michael Schuerig]
* Add an attribute reader method for ActiveRecord::Base.observers [Rick Olson]
* Deprecation: count class method should be called with an options hash rather than two args for conditions and joins. #6287 [Bob Silva]
* has_one associations with a nil target may be safely marshaled. #6279 [norbauer, Jeremy Kemper]
* Fixed the Ruby/MySQL adapter we ship with Active Record to work with the new authentication handshake that was introduced in MySQL 4.1, along with the other protocol changes made at that time #5723 [jimw@mysql.com]
* Deprecation: use :dependent => :delete_all rather than :exclusively_dependent => true. #6024 [Josh Susser]
* to_xml: the :methods option works on arrays of records. #5845 [Josh Starcher]
* has_many :through conditions are sanitized by the associating class. #5971 [martin.emde@gmail.com]
* Fix spurious newlines and spaces in AR::Base#to_xml output [Jamis Buck]
* has_one supports the :dependent => :delete option which skips the typical callback chain and deletes the associated object directly from the database. #5927 [Chris Mear, Jonathan Viney]
* Nested subclasses are not prefixed with the parent class' table_name since they should always use the base class' table_name. #5911 [Jonathan Viney]
* SQLServer: work around bug where some unambiguous date formats are not correctly identified if the session language is set to german. #5894 [Tom Ward, kruth@bfpi]
* Clashing type columns due to a sloppy join shouldn't wreck single-table inheritance. #5838 [Kevin Clark]
* Fixtures: correct escaping of \n and \r. #5859 [evgeny.zislis@gmail.com]
* Pushing a record onto a has_many :through sets the association's foreign key to the associate's primary key and adds it to the correct association. #5815, #5829 [Josh Susser]
* Add records to has_many :through using <<, push, and concat by creating the association record. Raise if base or associate are new records since both ids are required to create the association. #build raises since you can't associate an unsaved record. #create! takes an attributes hash and creates the associated record and its association in a transaction. [Jeremy Kemper]
WARNING: backward-incompatibility. Multicolumn indexes created before this
revision were named using the first column name only. Now they're uniquely
named using all indexed columns.
To remove an old multicolumn index, remove_index :table_name, :first_column
* Fix for deep includes on the same association. [richcollins@gmail.com]
* Tweak fixtures so they don't try to use a non-ActiveRecord class. [Kevin Clark]
* Remove ActiveRecord::Base.reset since Dispatcher doesn't use it anymore. [Rick Olson]
* PostgreSQL: autodetected sequences work correctly with multiple schemas. Rely on the schema search_path instead of explicitly qualifying the sequence name with its schema. #5280 [guy.naor@famundo.com]
* Replace Reloadable with Reloadable::Deprecated. [Nicholas Seckar]
* PostgreSQL: return array fields as strings. #4664 [Robby Russell]
* SQLServer: added tests to ensure all database statements are closed, refactored identity_insert management code to use blocks, removed update/delete rowcount code out of execute and into update/delete, changed insert to go through execute method, removed unused quoting methods, disabled pessimistic locking tests as feature is currently unsupported, fixed RakeFile to load sqlserver specific tests whether running in ado or odbc mode, fixed support for recently added decimal types, added support for limits on integer types. #5670 [Tom Ward]
* Numeric and decimal columns map to BigDecimal instead of Float. Those with scale 0 map to Integer. #5454 [robbat2@gentoo.org, work@ashleymoran.me.uk]
* Oracle: use nonblocking queries if allow_concurrency is set, fix pessimistic locking, don't guess date vs. time by default (set OracleAdapter.emulate_dates = true for the old behavior), adapter cleanup. #5635 [Michael Schoen]
* Fixed a few Oracle issues: Allows Oracle's odd date handling to still work consistently within #to_xml, Passes test that hardcode insert statement by dropping the :id column, Updated RUNNING_UNIT_TESTS with Oracle instructions, Corrects method signature for #exec #5294 [Michael Schoen]
* Added :group to available options for finds done on associations #5516 [mike@michaeldewey.org]
* Observers also watch subclasses created after they are declared. #5535 [daniels@pronto.com.au]
* Removed deprecated timestamps_gmt class methods. [Jeremy Kemper]
* rake build_mysql_database grants permissions to rails@localhost. #5501 [brianegge@yahoo.com]
* PostgreSQL: support microsecond time resolution. #5492 [alex@msgpad.com]
* Add AssociationCollection#sum since the method_missing invokation has been shadowed by Enumerable#sum.
* Added find_or_initialize_by_X which works like find_or_create_by_X but doesn't save the newly instantiated record. [Sam Stephenson]
* Row locking. Provide a locking clause with the :lock finder option or true for the default "FOR UPDATE". Use the #lock! method to obtain a row lock on a single record (reloads the record with :lock => true). [Shugo Maeda]
# Obtain an exclusive lock on person 1 so we can safely increment visits.
Person.transaction do
# select * from people where id=1 for update
person = Person.find(1, :lock => true)
person.visits += 1
person.save!
end
* PostgreSQL: introduce allow_concurrency option which determines whether to use blocking or asynchronous #execute. Adapters with blocking #execute will deadlock Ruby threads. The default value is ActiveRecord::Base.allow_concurrency. [Jeremy Kemper]
* Use a per-thread (rather than global) transaction mutex so you may execute concurrent transactions on separate connections. [Jeremy Kemper]
* Change AR::Base#to_param to return a String instead of a Fixnum. Closes #5320. [Nicholas Seckar]
* Use explicit delegation instead of method aliasing for AR::Base.to_param -> AR::Base.id. #5299 (skaes@web.de)
* Refactored ActiveRecord::Base.to_xml to become a delegate for XmlSerializer, which restores sanity to the mega method. This refactoring also reinstates the opinions that type="string" is redundant and ugly and nil-differentiation is not a concern of serialization [David Heinemeier Hansson]
Person.find(:all, :conditions => [ "last_name = ? and status = ?", "Catlin", 1 ], :limit => 2)
This makes it easier to pass in the options from a form or otherwise outside.
* Fixed issues with BLOB limits, charsets, and booleans for Firebird #5194, #5191, #5189 [kennethkunz@gmail.com]
* Fixed usage of :limit and with_scope when the association in scope is a 1:m #5208 [alex@purefiction.net]
* Fixed migration trouble with SQLite when NOT NULL is used in the new definition #5215 [greg@lapcominc.com]
* Fixed problems with eager loading and counting on SQL Server #5212 [kajism@yahoo.com]
* Fixed that count distinct should use the selected column even when using :include #5251 [anna@wota.jp]
* Fixed that :includes merged from with_scope won't cause the same association to be loaded more than once if repetition occurs in the clauses #5253 [alex@purefiction.net]
* Allow models to override to_xml. #4989 [Blair Zajac <blair@orcaware.com>]
* PostgreSQL: don't ignore port when host is nil since it's often used to label the domain socket. #5247 [shimbo@is.naist.jp]
* Records and arrays of records are bound as quoted ids. [Jeremy Kemper]
Foo.find(:all, :conditions => ['bar_id IN (?)', bars])
* create! no longer blows up when no attributes are passed and a :create scope is in effect (e.g. foo.bars.create! failed whereas foo.bars.create!({}) didn't.) [Jeremy Kemper]
* Call Inflector#demodulize on the class name when eagerly including an STI model. Closes #5077 [info@loobmedia.com]
* Preserve MySQL boolean column defaults when changing a column in a migration. Closes #5015. [pdcawley@bofh.org.uk]
* PostgreSQL: migrations support :limit with :integer columns by mapping limit < 4 to smallint, > 4 to bigint, and anything else to integer. #2900 [keegan@thebasement.org]
* Dates and times interpret empty strings as nil rather than 2000-01-01. #4830 [kajism@yahoo.com]
* When grouping, use the appropriate option key. [Marcel Molina Jr.]
* Add support for FrontBase (http://www.frontbase.com/) with a new adapter thanks to the hard work of one Mike Laster. Closes #4093. [mlaster@metavillage.com]
* Add warning about the proper way to validate the presence of a foreign key. Closes #4147. [Francois Beausoleil <francois.beausoleil@gmail.com>]
* Multiple fixes and optimizations in PostgreSQL adapter, allowing ruby-postgres gem to work properly. [ruben.nine@gmail.com]
* Fixed that AssociationCollection#delete_all should work even if the records of the association are not loaded yet. [Florian Weber]
* Changed those private ActiveRecord methods to take optional third argument :auto instead of nil for performance optimizations. (closes #4456) [Stefan]
* DEPRECATED: Using additional attributes on has_and_belongs_to_many associations. Instead upgrade your association to be a real join model [David Heinemeier Hansson]
* Fixed that records returned from has_and_belongs_to_many associations with additional attributes should be marked as read only (fixes #4512) [David Heinemeier Hansson]
* Do not implicitly mark recordss of has_many :through as readonly but do mark habtm records as readonly (eventually only on join tables without rich attributes). [Marcel Mollina Jr.]
* Change has_many :through to use the :source option to specify the source association. :class_name is now ignored. [Rick Olson]
class Connection < ActiveRecord::Base
belongs_to :user
belongs_to :channel
end
class Channel < ActiveRecord::Base
has_many :connections
has_many :contacts, :through => :connections, :class_name => 'User' # OLD
has_many :contacts, :through => :connections, :source => :user # NEW
end
* Fixed DB2 adapter so nullable columns will be determines correctly now and quotes from column default values will be removed #4350 [contact@maik-schmidt.de]
* Allow overriding of find parameters in scoped has_many :through calls [Rick Olson]
In this example, :include => false disables the default eager association from loading. :select changes the standard
select clause. :joins specifies a join that is added to the end of the has_many :through query.
class Post < ActiveRecord::Base
has_many :tags, :through => :taggings, :include => :tagging do
* Fixed that schema changes while the database was open would break any connections to a SQLite database (now we reconnect if that error is throw) [David Heinemeier Hansson]
* Fix that has_many :through honors the foreign key set by the belongs_to association in the join model (closes #4259) [andylien@gmail.com / Rick Olson]
* Added OpenBase database adapter that builds on top of the http://www.spice-of-life.net/ruby-openbase/ driver. All functionality except LIMIT/OFFSET is supported #3528 [derrickspell@cdmplus.com]
* Rework table aliasing to account for truncated table aliases. Add smarter table aliasing when doing eager loading of STI associations. This allows you to use the association name in the order/where clause. [Jonathan Viney / Rick Olson] #4108 Example (SpecialComment is using STI):
* Provide access to the underlying database connection through Adapter#raw_connection. Enables the use of db-specific methods without complicating the adapters. #2090 [Michael Koziarski]
* Added connection#current_database that'll return of the current database (only works in MySQL, SQL Server, and Oracle so far -- please help implement for the rest of the adapters) #3663 [Tom Ward]
You can even do load first-level associations as part of the document:
firm.to_xml :include => [ :account, :clients ]
...that'll return something like:
<?xml version="1.0" encoding="UTF-8"?>
<firm>
<id type="integer">1</id>
<rating type="integer">1</rating>
<name>37signals</name>
<clients>
<client>
<rating type="integer">1</rating>
<name>Summit</name>
</client>
<client>
<rating type="integer">1</rating>
<name>Microsoft</name>
</client>
</clients>
<account>
<id type="integer">1</id>
<credit-limit type="integer">50</credit-limit>
</account>
</firm>
* Allow :counter_cache to take a column name for custom counter cache columns [Jamis Buck]
* Documentation fixes for :dependent [robby@planetargon.com]
* Stop the MySQL adapter crashing when views are present. #3782 [Jonathan Viney]
* Don't classify the belongs_to class, it is already singular #4117 [keithm@infused.org]
* Allow set_fixture_class to take Classes instead of strings for a class in a module. Raise FixtureClassNotFound if a fixture can't load. [Rick Olson]
* Fix quoting of inheritance column for STI eager loading #4098 [Jonathan Viney <jonathan@bluewire.net.nz>]
* Added smarter table aliasing for eager associations for multiple self joins #3580 [Rick Olson]
* The first time a table is referenced in a join, no alias is used.
* After that, the parent class name and the reflection name are used.
Tree.find(:all, :include => :children) # LEFT OUTER JOIN trees AS tree_children ...
* Any additional join references get a numerical suffix like '_2', '_3', etc.
* Fixed eager loading problems with single-table inheritance #3580 [Rick Olson]. Post.find(:all, :include => :special_comments) now returns all posts, and any special comments that the posts may have. And made STI work with has_many :through and polymorphic belongs_to.
* Added cascading eager loading that allows for queries like Author.find(:all, :include=> { :posts=> :comments }), which will fetch all authors, their posts, and the comments belonging to those posts in a single query (using LEFT OUTER JOIN) #3913 [anna@wota.jp]. Examples:
* Make counter cache work when replacing an association #3245 [eugenol@gmail.com]
* Make migrations verbose [Jamis Buck]
* Make counter_cache work with polymorphic belongs_to [Jamis Buck]
* Fixed that calling HasOneProxy#build_model repeatedly would cause saving to happen #4058 [anna@wota.jp]
* Added Sybase database adapter that relies on the Sybase Open Client bindings (see http://raa.ruby-lang.org/project/sybase-ctlib) #3765 [John Sheets]. It's almost completely Active Record compliant (including migrations), but has the following caveats:
* Does not support DATE SQL column types; use DATETIME instead.
* Date columns on HABTM join tables are returned as String, not Time.
* Insertions are potentially broken for :polymorphic join tables
* BLOB column access not yet fully supported
* Clear stale, cached connections left behind by defunct threads. [Jeremy Kemper]
* CHANGED DEFAULT: set ActiveRecord::Base.allow_concurrency to false. Most AR usage is in single-threaded applications. [Jeremy Kemper]
* Fixed that Base.save should always return false if the save didn't succeed, including if it has halted by before_save's #1861, #2477 [David Heinemeier Hansson]
* ActiveRecord::Base.remove_connection explicitly closes database connections and doesn't corrupt the connection cache. Introducing the disconnect! instance method for the PostgreSQL, MySQL, and SQL Server adapters; implementations for the others are welcome. #3591 [Simon Stapleton, Tom Ward]
* Added support for nested scopes #3407 [anna@wota.jp]. Examples:
* Added calculations: Base.count, Base.average, Base.sum, Base.minimum, Base.maxmium, and the generic Base.calculate. All can be used with :group and :having. Calculations and statitics need no longer require custom SQL. #3958 [Rick Olson]. Examples:
* Reflections don't attempt to resolve module nesting of association classes. Simplify type computation. [Jeremy Kemper]
* Improved the Oracle OCI Adapter with better performance for column reflection (from #3210), fixes to migrations (from #3476 and #3742), tweaks to unit tests (from #3610), and improved documentation (from #2446) #3879 [Aggregated by schoenm@earthlink.net]
* Fixed that the schema_info table used by ActiveRecord::Schema.define should respect table pre- and suffixes #3834 [rubyonrails@atyp.de]
* Added :select option to Base.count that'll allow you to select something else than * to be counted on. Especially important for count queries using DISTINCT #3839 [Stefan Kaes]
* Remove definition of reloadable? from ActiveRecord::Base to make way for new Reloadable code. [Nicholas Seckar]
* Fixed schema handling for DB2 adapter that didn't work: an initial schema could be set, but it wasn't used when getting tables and indexes #3678 [Maik Schmidt]
* Added option to Base.reflection_of_all_associations to specify a specific association to scope the call. For example Base.reflection_of_all_associations(:has_many) [David Heinemeier Hansson]
* Added ActiveRecord::SchemaDumper.ignore_tables which tells SchemaDumper which tables to ignore. Useful for tables with funky column like the ones required for tsearch2. [Tobias Lütke]
* SchemaDumper now doesn't fail anymore when there are unknown column types in the schema. Instead the table is ignored and a Comment is left in the schema.rb. [Tobias Lütke]
* Fixed that saving a model with multiple habtm associations would only save the first one. #3244 [yanowitz-rubyonrails@quantumfoam.org, Florian Weber]
* Fix change_column to work with PostgreSQL 7.x and 8.x. #3141 [wejn@box.cz, Rick Olson, Scott Barron]
* Added option to specify :group, :limit, :offset, and :select options from find on has_and_belongs_to_many and has_many assosociations [David Heinemeier Hansson]
* Fixed that using :include together with :conditions array in Base.find would cause NoMethodError #2887 [Paul Hammmond]
* PostgreSQL: more robust sequence name discovery. #3087 [Rick Olson]
* Oracle: use syntax compatible with Oracle 8. #3131 [Michael Schoen]
* MySQL: work around ruby-mysql/mysql-ruby inconsistency with mysql.stat. Eliminate usage of mysql.ping because it doesn't guarantee reconnect. Explicitly close and reopen the connection instead. [Jeremy Kemper]
* Allow validate_uniqueness_of to be scoped by more than just one column. #1559. [jeremy@jthopple.com, Marcel Molina Jr.]
* Firebird: active? and reconnect! methods for handling stale connections. #428 [Ken Kunz <kennethkunz@gmail.com>]
* Firebird: updated for FireRuby 0.4.0. #3009 [Ken Kunz <kennethkunz@gmail.com>]
* MySQL and PostgreSQL: active? compatibility with the pure-Ruby driver. #428 [Jeremy Kemper]
* Oracle: active? check pings the database rather than testing the last command status. #428 [Michael Schoen]
* SQLServer: resolve column aliasing/quoting collision when using limit or offset in an eager find. #2974 [kajism@yahoo.com]
* Reloading a model doesn't lose track of its connection. #2996 [junk@miriamtech.com, Jeremy Kemper]
* Fixed bug where using update_attribute after pushing a record to a habtm association of the object caused duplicate rows in the join table. #2888 [colman@rominato.com, Florian Weber, Michael Schoen]
* MySQL, PostgreSQL: reconnect! also reconfigures the connection. Otherwise, the connection 'loses' its settings if it times out and is reconnected. #2978 [Shugo Maeda]
* has_and_belongs_to_many: use JOIN instead of LEFT JOIN. [Jeremy Kemper]
* MySQL: introduce :encoding option to specify the character set for client, connection, and results. Only available for MySQL 4.1 and later with the mysql-ruby driver. Do SHOW CHARACTER SET in mysql client to see available encodings. #2975 [Shugo Maeda]
* Add tasks to create, drop and rebuild the MySQL and PostgreSQL test databases. [Marcel Molina Jr.]
* Don't generate read methods for columns whose names are not valid ruby method names. #2946 [Stefan Kaes]
* Document :force option to create_table. #2921 [Blair Zajac <blair@orcaware.com>]
* Don't add the same conditions twice in has_one finder sql. #2916 [Jeremy Evans]
* Rename Version constant to VERSION. #2802 [Marcel Molina Jr.]
* Introducing the Firebird adapter. Quote columns and use attribute_condition more consistently. Setup guide: http://wiki.rubyonrails.com/rails/pages/Firebird+Adapter #1874 [Ken Kunz <kennethkunz@gmail.com>]
* SQLServer: active? and reconnect! methods for handling stale connections. #428 [kajism@yahoo.com, Tom Ward <tom@popdog.net>]
* Associations handle case-equality more consistently: item.parts.is_a?(Array) and item.parts === Array. #1345 [MarkusQ@reality.com]
* SQLServer: insert uses given primary key value if not nil rather than SELECT @@IDENTITY. #2866 [kajism@yahoo.com, Tom Ward <tom@popdog.net>]
* Oracle: active? and reconnect! methods for handling stale connections. Optionally retry queries after reconnect. #428 [Michael Schoen <schoenm@earthlink.net>]
* Correct documentation for Base.delete_all. #1568 [Newhydra]
* Oracle: test case for column default parsing. #2788 [Michael Schoen <schoenm@earthlink.net>]
* Update documentation for Migrations. #2861 [Tom Werner <tom@cube6media.com>]
* When AbstractAdapter#log rescues an exception, attempt to detect and reconnect to an inactive database connection. Connection adapter must respond to the active? and reconnect! instance methods. Initial support for PostgreSQL, MySQL, and SQLite. Make certain that all statements which may need reconnection are performed within a logged block: for example, this means no avoiding log(sql, name) { } if @logger.nil? #428 [Jeremy Kemper]
* Oracle: Much faster column reflection. #2848 [Michael Schoen <schoenm@earthlink.net>]
* Base.reset_sequence_name analogous to reset_table_name (mostly useful for testing). Base.define_attr_method allows nil values. [Jeremy Kemper]
* PostgreSQL: smarter sequence name defaults, stricter last_insert_id, warn on pk without sequence. [Jeremy Kemper]
* Upgrade bundled ruby-mysql 0.2.4 with mysql411 shim (see #440) to ruby-mysql 0.2.6 with a patchset for 4.1 protocol support. Local change [301] is now a part of the main driver; reapplied local change [2182]. Removed GC.start from Result.free. [tommy@tmtm.org, akuroda@gmail.com, Doug Fales <doug.fales@gmail.com>, Jeremy Kemper]
* Correct handling of complex order clauses with SQL Server limit emulation. #2770 [Tom Ward <tom@popdog.net>, Matt B.]
* Correct whitespace problem in Oracle default column value parsing. #2788 [rick@rickbradley.com]
* Destroy associated has_and_belongs_to_many records after all before_destroy callbacks but before destroy. This allows you to act on the habtm association as you please while preserving referential integrity. #2065 [larrywilliams1@gmail.com, sam.kirchmeier@gmail.com, elliot@townx.org, Jeremy Kemper]
* Deprecate the old, confusing :exclusively_dependent option in favor of :dependent => :delete_all. [Jeremy Kemper]
* More compatible Oracle column reflection. #2771 [Ryan Davis <ryand-ruby@zenspider.com>, Michael Schoen <schoenm@earthlink.net>]
* Rename Base.constrain to Base.with_scope so it doesn't conflict with existing concept of database constraints. Make scoping more robust: uniform method => parameters, validated method names and supported finder parameters, raise exception on nested scopes. [Jeremy Kemper] Example:
* Added find_or_create_by_X as a second type of dynamic finder that'll create the record if it doesn't already exist [David Heinemeier Hansson]. Example:
* Correct the pure-Ruby MySQL 4.1.1 shim's version test. #2718 [Jeremy Kemper]
* Add Model.create! to match existing model.save! method. When save! raises RecordInvalid, you can catch the exception, retrieve the invalid record (invalid_exception.record), and see its errors (invalid_exception.record.errors). [Jeremy Kemper]
* Correct fixture behavior when table name pluralization is off. #2719 [Rick Bradley <rick@rickbradley.com>]
* Changed :dbfile to :database for SQLite adapter for consistency (old key still works as an alias) #2644 [Dan Peterson]
* Added migration support for Oracle #2647 [Michael Schoen]
* Worked around that connection can't be reset if allow_concurrency is off. #2648 [Michael Schoen <schoenm@earthlink.net>]
* Fixed SQL Server adapter so it honors options[:conditions] when applying :limits #1978 [Tom Ward]
* Added migration support to SQL Server adapter (please someone do the same for Oracle and DB2) #2625 [Tom Ward]
* Use AR::Base.silence rather than AR::Base.logger.silence in fixtures to preserve Log4r compatibility. #2618 [dansketcher@gmail.com]
* Constraints are cloned so they can't be inadvertently modified while they're
in effect. Added :readonly finder constraint. Calling an association collection's class method (Part.foobar via item.parts.foobar) constrains :readonly => false since the collection's :joins constraint would otherwise force it to true. [Jeremy Kemper <rails@bitsweat.net>]
* Added :offset and :limit to the kinds of options that Base.constrain can use #2466 [duane.johnson@gmail.com]
* Added support for using limits in eager loads that involve has_many and has_and_belongs_to_many associations
*1.12.0* (October 16th, 2005)
* Update/clean up documentation (rdoc)
* PostgreSQL sequence support. Use set_sequence_name in your model class to specify its primary key sequence. #2292 [Rick Olson <technoweenie@gmail.com>, Robby Russell <robby@planetargon.com>]
* Change default logging colors to work on both white and black backgrounds. [Sam Stephenson]
* YAML fixtures support ordered hashes for fixtures with foreign key dependencies in the same table. #1896 [purestorm@ggnore.net]
* :dependent now accepts :nullify option. Sets the foreign key of the related objects to NULL instead of deleting them. #2015 [Robby Russell <robby@planetargon.com>]
* Introduce read-only records. If you call object.readonly! then it will mark the object as read-only and raise ReadOnlyRecord if you call object.save. object.readonly? reports whether the object is read-only. Passing :readonly => true to any finder method will mark returned records as read-only. The :joins option now implies :readonly, so if you use this option, saving the same record will now fail. Use find_by_sql to work around.
* Avoid memleak in dev mode when using fcgi
* Simplified .clear on active record associations by using the existing delete_records method. #1906 [Caleb <me@cpb.ca>]
* Delegate access to a customized primary key to the conventional id method. #2444. [Blair Zajac <blair@orcaware.com>]
* Fix errors caused by assigning a has-one or belongs-to property to itself
* Add ActiveRecord::Base.schema_format setting which specifies how databases should be dumped [Sam Stephenson]
* Add option (true by default) to generate reader methods for each attribute of a record to avoid the overhead of calling method missing. In partial fullfilment of #1236. [Stefan Kaes]
* Raise errors when invalid hash keys are passed to ActiveRecord::Base.find. #2363 [Chad Fowler <chad@chadfowler.com>, Nicholas Seckar]
* Added :force option to create_table that'll try to drop the table if it already exists before creating
* Fix transactions so that calling return while inside a transaction will not leave an open transaction on the connection. [Nicholas Seckar]
* Use foreign_key inflection uniformly. #2156 [Blair Zajac <blair@orcaware.com>]
* model.association.clear should destroy associated objects if :dependent => true instead of nullifying their foreign keys. #2221 [joergd@pobox.com, ObieFernandez <obiefernandez@gmail.com>]
* Returning false from before_destroy should cancel the action. #1829 [Jeremy Huffman]
* Recognize PostgreSQL NOW() default as equivalent to CURRENT_TIMESTAMP or CURRENT_DATE, depending on the column's type. #2256 [mat <mat@absolight.fr>]
* Extensive documentation for the abstract database adapter. #2250 [François Beausoleil <fbeausoleil@ftml.net>]
* Clean up Fixtures.reset_sequences for PostgreSQL. Handle tables with no rows and models with custom primary keys. #2174, #2183 [jay@jay.fm, Blair Zajac <blair@orcaware.com>]
* Improve error message when nil is assigned to an attr which validates_size_of within a range. #2022 [Manuel Holtgrewe <purestorm@ggnore.net>]
* Make update_attribute use the same writer method that update_attributes uses.
* Fixtures should only reset a PostgreSQL sequence if it corresponds to an integer primary key named id. #1749 [chris@chrisbrinker.com]
* Standardize the interpretation of boolean columns in the Mysql and Sqlite adapters. (Use MysqlAdapter.emulate_booleans = false to disable this behavior)
* Added AbstractAdapter#select_value and AbstractAdapter#select_values as convenience methods for selecting single values, instead of hashes, of the first column in a SELECT #2283 [solo@gatelys.com]
* Wrap :conditions in parentheses to prevent problems with OR's #1871 [Jamis Buck]
* Allow the postgresql adapter to work with the SchemaDumper. [Jamis Buck]
* Add ActiveRecord::SchemaDumper for dumping a DB schema to a pure-ruby file, making it easier to consolidate large migration lists and port database schemas between databases. [Jamis Buck]
* Fixed migrations for Windows when using more than 10 [David Naseby]
* Fixed that the create_x method from belongs_to wouldn't save the association properly #2042 [Florian Weber]
* Added check for RAILS_CONNECTION_ADAPTERS on startup and only load the connection adapters specified within if its present (available in Rails through config.connection_adapters using the new config) #1958 [skae]
* Fixed various problems with has_and_belongs_to_many when using customer finder_sql #2094 [Florian Weber]
* Fixed "connection lost" issue with the bundled Ruby/MySQL driver (would kill the app after 8 hours of inactivity) #2163, #428 [kajism@yahoo.com]
* Fixed comparison of Active Record objects so two new objects are not equal #2099 [deberg]
* Fixed that the SQL Server adapter would sometimes return DBI::Timestamp objects instead of Time #2127 [Tom Ward]
* Added the instance methods #root and #ancestors on acts_as_tree and fixed siblings to not include the current node #2142, #2140 [coffee2code]
* Fixed that Active Record would call SHOW FIELDS twice (or more) for the same model when the cached results were available #1947 [sd@notso.net]
* Added log_level and use_silence parameter to ActiveRecord::Base.benchmark. The first controls at what level the benchmark statement will be logged (now as debug, instead of info) and the second that can be passed false to include all logging statements during the benchmark block/
* Make sure the schema_info table is created before querying the current version #1903
* Fix typo in validations documentation #1938 [court3nay]
* Make acts_as_list work for insert_at(1) #1966 [hensleyl@papermountain.org]
* Fix typo in count_by_sql documentation #1969 [Alexey Verkhovsky]
* Allow add_column and create_table to specify NOT NULL #1712 [emptysands@gmail.com]
* Fix create_table so that id column is implicitly added [Rick Olson]
* Default sequence names for Oracle changed to #{table_name}_seq, which is the most commonly used standard. In addition, a new method ActiveRecord::Base#set_sequence_name allows the developer to set the sequence name per model. This is a non-backwards-compatible change -- anyone using the old-style "rails_sequence" will need to either create new sequences, or set: ActiveRecord::Base.set_sequence_name = "rails_sequence" #1798
* OCIAdapter now properly handles synonyms, which are commonly used to separate out the schema owner from the application user #1798
* Fixed the handling of camelCase columns names in Oracle #1798
* Implemented for OCI the Rakefile tasks of :clone_structure_to_test, :db_structure_dump, and :purge_test_database, which enable Oracle folks to enjoy all the agile goodness of Rails for testing. Note that the current implementation is fairly limited -- only tables and sequences are cloned, not constraints or indexes. A full clone in Oracle generally requires some manual effort, and is version-specific. Post 9i, Oracle recommends the use of the DBMS_METADATA package, though that approach requires editing of the physical characteristics generated #1798
* Fixed the handling of multiple blob columns in Oracle if one or more of them are null #1798
Post.find(1).comments.search('hi') # => SELECT * from comments WHERE post_id = 1 AND body = 'hi'
NOTICE: This patch changes the underlying SQL generated by has_and_belongs_to_many queries. If your relying on that, such as
by explicitly referencing the old t and j aliases, you'll need to update your code. Of course, you _shouldn't_ be relying on
details like that no less than you should be diving in to touch private variables. But just in case you do, consider yourself
noticed :)
* Added migration support for SQLite (using temporary tables to simulate ALTER TABLE) #1771 [Sam Stephenson]
* Remove extra definition of supports_migrations? from abstract_adaptor.rb [Nicholas Seckar]
* Fix acts_as_list so that moving next-to-last item to the bottom does not result in duplicate item positions
* Fixed incompatibility in DB2 adapter with the new limit/offset approach #1718 [Maik Schmidt]
* Added :select option to find which can specify a different value than the default *, like find(:all, :select => "first_name, last_name"), if you either only want to select part of the columns or exclude columns otherwise included from a join #1338 [Stefan Kaes]
* Added support for limit and offset with eager loading of has_one and belongs_to associations. Using the options with has_many and has_and_belongs_to_many associations will now raise an ActiveRecord::ConfigurationError #1692 [Rick Olson]
* Fixed that assume_bottom_position (in acts_as_list) could be called on items already last in the list and they would move one position away from the list #1648 [tyler@kianta.com]
* Added ActiveRecord::Base.threaded_connections flag to turn off 1-connection per thread (required for thread safety). By default it's on, but WEBrick in Rails need it off #1685 [Sam Stephenson]
* Added new Migrations framework for describing schema transformations in a way that can be easily applied across multiple databases #1604 [Tobias Lütke] See documentation under ActiveRecord::Migration and the additional support in the Rails rakefile/generator.
* Added support for limit/offset with the MS SQL Server driver so that pagination will now work #1569 [DeLynn Berry]
* Added support for ODBC connections to MS SQL Server so you can connect from a non-Windows machine #1569 [Mark Imbriaco/DeLynn Berry]
* Fixed that multiparameter posts ignored attr_protected #1532 [alec+rails@veryclever.net]
* Fixed problem with eager loading when using a has_and_belongs_to_many association using :association_foreign_key #1504 [flash@vanklinkenbergsoftware.nl]
* Fixed Base#find to honor the documentation on how :joins work and make them consistent with Base#count #1405 [pritchie@gmail.com]. What used to be:
Developer.find :all, :joins => 'developers_projects', :conditions => 'id=developer_id AND project_id=1'
...should instead be:
Developer.find(
:all,
:joins => 'LEFT JOIN developers_projects ON developers.id = developers_projects.developer_id',
* Fixed that clear_association_cache doesn't delete new associations on new records (so you can safely place new records in the session with Action Pack without having new associations wiped) #1494 [cluon]
* Fixed that calling Model.find([]) returns [] and doesn't throw an exception #1379
* Fixed that adding a record to a has_and_belongs_to collection would always save it -- now it only saves if its a new record #1203 [Alisdair McDiarmid]
* Fixed saving of in-memory association structures to happen as a after_create/after_update callback instead of after_save -- that way you can add new associations in after_create/after_update callbacks without getting them saved twice
* Allow any Enumerable, not just Array, to work as bind variables #1344 [Jeremy Kemper]
* Added actual database-changing behavior to collection assigment for has_many and has_and_belongs_to_many #1425 [Sebastian Kanthak].
* Comprehensive PostgreSQL schema support. Use the optional schema_search_path directive in database.yml to give a comma-separated list of schemas to search for your tables. This allows you, for example, to have tables in a shared schema without having to use a custom table name. See http://www.postgresql.org/docs/8.0/interactive/ddl-schemas.html to learn more. #827 [dave@cherryville.org]
* Return PostgreSQL columns in the order they were declared #1374 [perlguy@gmail.com]
* Allow before/after update hooks to work on models using optimistic locking
* Eager loading of dependent has_one associations won't delete the association #1212
* Added a second parameter to the build and create method for has_one that controls whether the existing association should be replaced (which means nullifying its foreign key as well). By default this is true, but false can be passed to prevent it.
* Using transactional fixtures now causes the data to be loaded only once.
* Added fixture accessor methods that can be used when instantiated fixtures are disabled.
fixtures :web_sites
def test_something
assert_equal "Ruby on Rails", web_sites(:rubyonrails).name
end
* Added DoubleRenderError exception that'll be raised if render* is called twice #518 [Nicholas Seckar]
* Fixed exceptions occuring after render has been called #1096 [Nicholas Seckar]
* CHANGED: validates_presence_of now uses Errors#add_on_blank, which will make " " fail the validation where it didn't before #1309
* Added Errors#add_on_blank which works like Errors#add_on_empty, but uses Object#blank? instead
* Added the :if option to all validations that can either use a block or a method pointer to determine whether the validation should be run or not. #1324 [Duane Johnson/jhosteny]. Examples:
Conditional validations such as the following are made possible:
* Fixed use of construct_finder_sql when using :join #1288 [dwlt@dwlt.net]
* Fixed that :delete_sql in has_and_belongs_to_many associations couldn't access record properties #1299 [Rick Olson]
* Fixed that clone would break when an aggregate had the same name as one of its attributes #1307 [Jeremy Kemper]
* Changed that destroying an object will only freeze the attributes hash, which keeps the object from having attributes changed (as that wouldn't make sense), but allows for the querying of associations after it has been destroyed.
* Changed the callbacks such that observers are notified before the in-object callbacks are triggered. Without this change, it wasn't possible to act on the whole object in something like a before_destroy observer without having the objects own callbacks (like deleting associations) called first.
* Added option for passing an array to the find_all version of the dynamic finders and have it evaluated as an IN fragment. Example:
# SELECT * FROM topics WHERE title IN ('First', 'Second')
* Added Base.save! that attempts to save the record just like Base.save but will raise a RecordInvalid exception instead of returning false if the record is not valid [Dave Thomas]
* Fixed PostgreSQL usage of fixtures with regards to public schemas and table names with dots #962 [gnuman1@gmail.com]
* Fixed that fixtures were being deleted in the same order as inserts causing FK errors #890 [andrew.john.peters@gmail.com]
* Fixed loading of fixtures in to be in the right order (or PostgreSQL would bark) #1047 [stephenh@chase3000.com]
* Fixed page caching for non-vhost applications living underneath the root #1004 [Ben Schumacher]
* Fixes a problem with the SQL Adapter which was resulting in IDENTITY_INSERT not being set to ON when it should be #1104 [adelle]
* Added the option to specify the acceptance string in validates_acceptance_of #1106 [caleb@aei-tech.com]
* Added insert_at(position) to acts_as_list #1083 [DeLynnB]
* Removed the default order by id on has_and_belongs_to_many queries as it could kill performance on large sets (you can still specify by hand with :order)
* Fixed that Base.silence should restore the old logger level when done, not just set it to DEBUG #1084 [yon@milliped.com]
* Fixed boolean saving on Oracle #1093 [mparrish@pearware.org]
* Moved build_association and create_association for has_one and belongs_to out of deprecation as they work when the association is nil unlike association.build and association.create, which require the association to be already in place #864
* Added rollbacks of transactions if they're active as the dispatcher is killed gracefully (TERM signal) #1054 [Leon Bredt]
* Added quoting of column names for fixtures #997 [jcfischer@gmail.com]
* Fixed counter_sql when no records exist in database for PostgreSQL (would give error, not 0) #1039 [Caleb Tennis]
* Fixed boolean queries for t/f fields in PostgreSQL #995 [dave@cherryville.org]
* Added that model.items.delete(child) will delete the child, not just set the foreign key to nil, if the child is dependent on the model #978 [Jeremy Kemper]
* Fixed auto-stamping of dates (created_on/updated_on) for PostgreSQL #985 [dave@cherryville.org]
* Fixed bug in Base#hash method that would treat records with the same string-based id as different [Dave Thomas]
* Renamed DateHelper#distance_of_time_in_words_to_now to DateHelper#time_ago_in_words (old method name is still available as a deprecated alias)
*1.9.1* (27th March, 2005)
* Fixed that Active Record objects with float attribute could not be cloned #808
* Fixed that MissingSourceFile's wasn't properly detected in production mode #925 [Nicholas Seckar]
* Fixed that :counter_cache option would look for a line_items_count column for a LineItem object instead of lineitems_count
* Fixed that AR exists?() would explode on postgresql if the passed id did not match the PK type #900 [Scott Barron]
* Fixed the MS SQL adapter to work with the new limit/offset approach and with binary data (still suffering from 7KB limit, though) #901 [delynnb]
*1.9.0* (22th March, 2005)
* Added adapter independent limit clause as a two-element array with the first being the limit, the second being the offset #795 [Sam Stephenson]. Example:
Developer.find_all nil, 'id ASC', 5 # return the first five developers
Developer.find_all nil, 'id ASC', [3, 8] # return three developers, starting from #8 and forward
This doesn't yet work with the DB2 or MS SQL adapters. Patches to make that happen are encouraged.
* Added alias_method :to_param, :id to Base, such that Active Record objects to be used as URL parameters in Action Pack automatically #812 [Nicholas Seckar/Sam Stephenson]
* Improved the performance of the OCI8 adapter for Oracle #723 [pilx/gjenkins]
* Added type conversion before saving a record, so string-based values like "10.0" aren't left for the database to convert #820 [dave@cherryville.org]
* Added with additional settings for working with transactional fixtures and pre-loaded test databases #865 [mindel]
* Fixed acts_as_list to trigger remove_from_list on destroy after the fact, not before, so a unique position can be maintained #871 [Alisdair McDiarmid]
* Added the possibility of specifying fixtures in multiple calls #816 [kim@tinker.com]
* Added Base.exists?(id) that'll return true if an object of the class with the given id exists #854 [stian@grytoyr.net]
* Added optionally allow for nil or empty strings with validates_numericality_of #801 [Sebastian Kanthak]
* Fixed problem with using slashes in validates_format_of regular expressions #801 [Sebastian Kanthak]
* Fixed that SQLite3 exceptions are caught and reported properly #823 [yerejm]
* Added that all types of after_find/after_initialized callbacks are triggered if the explicit implementation is present, not only the explicit implementation itself
* Fixed that symbols can be used on attribute assignment, like page.emails.create(:subject => data.subject, :body => data.body)
*1.8.0* (7th March, 2005)
* Added ActiveRecord::Base.colorize_logging to control whether to use colors in logs or not (on by default)
* Added support for timestamp with time zone in PostgreSQL #560 [Scott Barron]
* Added MultiparameterAssignmentErrors and AttributeAssignmentError exceptions #777 [demetrius]. Documentation:
* +MultiparameterAssignmentErrors+ -- collection of errors that occurred during a mass assignment using the
+attributes=+ method. The +errors+ property of this exception contains an array of +AttributeAssignmentError+
objects that should be inspected to determine which attributes triggered the errors.
* +AttributeAssignmentError+ -- an error occurred while doing a mass assignment through the +attributes=+ method.
You can inspect the +attribute+ property of the exception object to determine which attribute triggered the error.
* Fixed that postgresql adapter would fails when reading bytea fields with null value #771 [rodrigo k]
* Added transactional fixtures that uses rollback to undo changes to fixtures instead of DELETE/INSERT -- it's much faster. See documentation under Fixtures #760 [Jeremy Kemper]
* Added destruction of dependent objects in has_one associations when a new assignment happens #742 [mindel]. Example:
class Account < ActiveRecord::Base
has_one :credit_card, :dependent => true
end
class CreditCard < ActiveRecord::Base
belongs_to :account
end
account.credit_card # => returns existing credit card, lets say id = 12
Validates whether the value of the specified attribute is numeric by trying to convert it to
a float with Kernel.Float (if <tt>integer</tt> is false) or applying it to the regular expression
<tt>/^[\+\-]?\d+$/</tt> (if <tt>integer</tt> is set to true).
class Person < ActiveRecord::Base
validates_numericality_of :value, :on => :create
end
Configuration options:
* <tt>message</tt> - A custom error message (default is: "is not a number")
* <tt>on</tt> Specifies when this validation is active (default is :save, other options :create, :update)
* <tt>only_integer</tt> Specifies whether the value has to be an integer, e.g. an integral value (default is false)
* Fixed that HasManyAssociation#count was using :finder_sql rather than :counter_sql if it was available #445 [Scott Barron]
* Added better defaults for composed_of, so statements like composed_of :time_zone, :mapping => %w( time_zone time_zone ) can be written without the mapping part (it's now assumed)
* Added MacroReflection#macro which will return a symbol describing the macro used (like :composed_of or :has_many) #718, #248 [james@slashetc.com]
*1.7.0* (24th February, 2005)
* Changed the auto-timestamping feature to use ActiveRecord::Base.default_timezone instead of entertaining the parallel ActiveRecord::Base.timestamps_gmt method. The latter is now deprecated and will throw a warning on use (but still work) #710 [Jamis Buck]
* Added a OCI8-based Oracle adapter that has been verified to work with Oracle 8 and 9 #629 [Graham Jenkins]. Usage notes:
1. Key generation uses a sequence "rails_sequence" for all tables. (I couldn't find a simple
and safe way of passing table-specific sequence information to the adapter.)
2. Oracle uses DATE or TIMESTAMP datatypes for both dates and times. Consequently I have had to
resort to some hacks to get data converted to Date or Time in Ruby.
If the column_name ends in _at (like created_at, updated_at) it's created as a Ruby Time. Else if the
hours/minutes/seconds are 0, I make it a Ruby Date. Else it's a Ruby Time.
This is nasty - but if you use Duck Typing you'll probably not care very much.
In 9i it's tempting to map DATE to Date and TIMESTAMP to Time but I don't think that is
valid - too many databases use DATE for both.
Timezones and sub-second precision on timestamps are not supported.
3. Default values that are functions (such as "SYSDATE") are not supported. This is a
restriction of the way active record supports default values.
4. Referential integrity constraints are not fully supported. Under at least
some circumstances, active record appears to delete parent and child records out of
sequence and out of transaction scope. (Or this may just be a problem of test setup.)
The OCI8 driver can be retrieved from http://rubyforge.org/projects/ruby-oci8/
* Added option :schema_order to the PostgreSQL adapter to support the use of multiple schemas per database #697 [YuriSchimke]
* Optimized the SQL used to generate has_and_belongs_to_many queries by listing the join table first #693 [yerejm]
* Fixed that when using validation macros with a custom message, if you happened to use single quotes in the message string you would get a parsing error #657 [tonka]
* Fixed that Active Record would throw Broken Pipe errors with FCGI when the MySQL connection timed out instead of reconnecting #428 [Nicholas Seckar]
* Added options to specify an SSL connection for MySQL. Define the following attributes in the connection config (config/database.yml in Rails) to use it: sslkey, sslcert, sslca, sslcapath, sslcipher. To use SSL with no client certs, just set :sslca = '/dev/null'. http://dev.mysql.com/doc/mysql/en/secure-connections.html #604 [daniel@nightrunner.com]
* Added automatic dropping/creating of test tables for running the unit tests on all databases #587 [adelle@bullet.net.au]
* Fixed that find_by_* would fail when column names had numbers #670 [demetrius]
* Fixed the SQL Server adapter on a bunch of issues #667 [DeLynn]
1. Created a new columns method that is much cleaner.
2. Corrected a problem with the select and select_all methods
that didn't account for the LIMIT clause being passed into raw SQL statements.
3. Implemented the string_to_time method in order to create proper instances of the time class.
4. Added logic to the simplified_type method that allows the database to specify the scale of float data.
5. Adjusted the quote_column_name to account for the fact that MS SQL is bothered by a forward slash in the data string.
* Added validates_each that validates each specified attribute against a block #610 [Jeremy Kemper]. Example:
class Person < ActiveRecord::Base
validates_each :first_name, :last_name do |record, attr|
record.errors.add attr, 'starts with z.' if attr[0] == ?z
end
end
* Added :allow_nil as an explicit option for validates_length_of, so unless that's set to true having the attribute as nil will also return an error if a range is specified as :within #610 [Jeremy Kemper]
* Added that validates_* now accept blocks to perform validations #618 [Tim Bates]. Example:
class Person < ActiveRecord::Base
validate { |person| person.errors.add("title", "will never be valid") if SHOULD_NEVER_BE_VALID }
end
* Addded validation for validate all the associated objects before declaring failure with validates_associated #618 [Tim Bates]
* Added keyword-style approach to defining the custom relational bindings #545 [Jamis Buck]. Example:
* Fixed Base#clone for use with PostgreSQL #565 [hanson@surgery.wisc.edu]
*1.6.0* (January 25th, 2005)
* Added that has_many association build and create methods can take arrays of record data like Base#create and Base#build to build/create multiple records at once.
* Added that Base#delete and Base#destroy both can take an array of ids to delete/destroy #336
* Added the option of supplying an array of attributes to Base#create, so that multiple records can be created at once.
* Added the option of supplying an array of ids and attributes to Base#update, so that multiple records can be updated at once (inspired by #526/Duane Johnson). Example
* Added ActiveRecord::Base.timestamps_gmt that can be set to true to make the automated timestamping use GMT instead of local time #520 [Scott Baron]
* Added that update_all calls sanitize_sql on its updates argument, so stuff like MyRecord.update_all(['time = ?', Time.now]) works #519 [notahat]
* Fixed that the dynamic finders didn't treat nil as a "IS NULL" but rather "= NULL" case #515 [Demetrius]
* Added bind-named arrays for interpolating a group of ids or strings in conditions #528 [Jeremy Kemper]
* Added that has_and_belongs_to_many associations with additional attributes also can be created between unsaved objects and only committed to the database when Base#save is called on the associator #524 [Eric Anderson]
* Fixed that records fetched with piggy-back attributes or through rich has_and_belongs_to_many associations couldn't be saved due to the extra attributes not part of the table #522 [Eric Anderson]
* Added mass-assignment protection for the inheritance column -- regardless of a custom column is used or not
* Fixed that association proxies would fail === tests like PremiumSubscription === @account.subscription
* Fixed that column aliases didn't work as expected with the new MySql411 driver #507 [Demetrius]
* Fixed that find_all would produce invalid sql when called sequentialy #490 [Scott Baron]
*1.5.1* (January 18th, 2005)
* Fixed that the belongs_to and has_one proxy would fail a test like 'if project.manager' -- this unfortunately also means that you can't call methods like project.manager.build unless there already is a manager on the project #492 [Tim Bates]
* Fixed that the Ruby/MySQL adapter wouldn't connect if the password was empty #503 [Pelle]
*1.5.0* (January 17th, 2005)
* Fixed that unit tests for MySQL are now run as the "rails" user instead of root #455 [Eric Hodel]
* Added validates_associated that enables validation of objects in an unsaved association #398 [Tim Bates]. Example:
class Book < ActiveRecord::Base
has_many :pages
belongs_to :library
validates_associated :pages, :library
end
* Added support for associating unsaved objects #402 [Tim Bates]. Rules that govern this addition:
== Unsaved objects and associations
You can manipulate objects and associations before they are saved to the database, but there is some special behaviour you should be
aware of, mostly involving the saving of associated objects.
=== One-to-one associations
* Assigning an object to a has_one association automatically saves that object, and the object being replaced (if there is one), in
order to update their primary keys - except if the parent object is unsaved (new_record? == true).
* If either of these saves fail (due to one of the objects being invalid) the assignment statement returns false and the assignment
is cancelled.
* If you wish to assign an object to a has_one association without saving it, use the #association.build method (documented below).
* Assigning an object to a belongs_to association does not save the object, since the foreign key field belongs on the parent. It does
not save the parent either.
=== Collections
* Adding an object to a collection (has_many or has_and_belongs_to_many) automatically saves that object, except if the parent object
(the owner of the collection) is not yet stored in the database.
* If saving any of the objects being added to a collection (via #push or similar) fails, then #push returns false.
* You can add an object to a collection without automatically saving it by using the #collection.build method (documented below).
* All unsaved (new_record? == true) members of the collection are automatically saved when the parent is saved.
* Added replace to associations, so you can do project.manager.replace(new_manager) or project.milestones.replace(new_milestones) #402 [Tim Bates]
* Added build and create methods to has_one and belongs_to associations, so you can now do project.manager.build(attributes) #402 [Tim Bates]
* Added that if a before_* callback returns false, all the later callbacks and the associated action are cancelled. If an after_* callback returns false, all the later callbacks are cancelled. Callbacks are generally run in the order they are defined, with the exception of callbacks defined as methods on the model, which are called last. #402 [Tim Bates]
* Fixed that Base#== wouldn't work for multiple references to the same unsaved object #402 [Tim Bates]
* Fixed binary support for PostgreSQL #444 [alex@byzantine.no]
* Added a differenciation between AssociationCollection#size and -length. Now AssociationCollection#size returns the size of the
collection by executing a SELECT COUNT(*) query if the collection hasn't been loaded and calling collection.size if it has. If
it's more likely than not that the collection does have a size larger than zero and you need to fetch that collection afterwards,
it'll take one less SELECT query if you use length.
* Added Base#attributes that returns a hash of all the attributes with their names as keys and clones of their objects as values #433 [atyp.de]
* Fixed that foreign keys named the same as the association would cause stack overflow #437 [Eric Anderson]
* Fixed default scope of acts_as_list from "1" to "1 = 1", so it'll work in PostgreSQL (among other places) #427 [Alexey]
* Added Base#reload that reloads the attributes of an object from the database #422 [Andreas Schwarz]
* Added SQLite3 compatibility through the sqlite3-ruby adapter by Jamis Buck #381 [Jeremy Kemper]
* Added support for the new protocol spoken by MySQL 4.1.1+ servers for the Ruby/MySQL adapter that ships with Rails #440 [Matt Mower]
* Added that Observers can use the observes class method instead of overwriting self.observed_class().
Before:
class ListSweeper < ActiveRecord::Base
def self.observed_class() [ List, Item ]
end
After:
class ListSweeper < ActiveRecord::Base
observes List, Item
end
* Fixed that conditions in has_many and has_and_belongs_to_many should be interpolated just like the finder_sql is
* Fixed Base#update_attribute to be indifferent to whether a string or symbol is used to describe the name
* Added Base#toggle(attribute) and Base#toggle!(attribute) that makes it easier to flip a switch or flag.
* Changed those private ActiveRecord methods to take optional third argument :auto instead of nil for performance optimizations. (closes #4456) [Stefan]
* DEPRECATED: Using additional attributes on has_and_belongs_to_many associations. Instead upgrade your association to be a real join model [David Heinemeier Hansson]
* Fixed that records returned from has_and_belongs_to_many associations with additional attributes should be marked as read only (fixes #4512) [David Heinemeier Hansson]
* Do not implicitly mark recordss of has_many :through as readonly but do mark habtm records as readonly (eventually only on join tables without rich attributes). [Marcel Mollina Jr.]
* Fixed DB2 adapter so nullable columns will be determines correctly now and quotes from column default values will be removed #4350 [contact@maik-schmidt.de]
* Fixed that schema changes while the database was open would break any connections to a SQLite database (now we reconnect if that error is throw) [David Heinemeier Hansson]
* Fix that has_many :through honors the foreign key set by the belongs_to association in the join model (closes #4259) [andylien@gmail.com / Rick Olson]
* Added OpenBase database adapter that builds on top of the http://www.spice-of-life.net/ruby-openbase/ driver. All functionality except LIMIT/OFFSET is supported #3528 [derrickspell@cdmplus.com]
* Rework table aliasing to account for truncated table aliases. Add smarter table aliasing when doing eager loading of STI associations. This allows you to use the association name in the order/where clause. [Jonathan Viney / Rick Olson] #4108 Example (SpecialComment is using STI):
* Provide access to the underlying database connection through Adapter#raw_connection. Enables the use of db-specific methods without complicating the adapters. #2090 [Michael Koziarski]
* Added connection#current_database that'll return of the current database (only works in MySQL, SQL Server, and Oracle so far -- please help implement for the rest of the adapters) #3663 [Tom Ward]
* Any additional join references get a numerical suffix like '_2', '_3', etc.
* Fixed eager loading problems with single-table inheritance #3580 [Rick Olson]. Post.find(:all, :include => :special_comments) now returns all posts, and any special comments that the posts may have. And made STI work with has_many :through and polymorphic belongs_to.
* Added cascading eager loading that allows for queries like Author.find(:all, :include=> { :posts=> :comments }), which will fetch all authors, their posts, and the comments belonging to those posts in a single query (using LEFT OUTER JOIN) #3913 [anna@wota.jp]. Examples:
* Added Sybase database adapter that relies on the Sybase Open Client bindings (see http://raa.ruby-lang.org/project/sybase-ctlib) #3765 [John Sheets]. It's almost completely Active Record compliant (including migrations), but has the following caveats:
* Fixed that Base.save should always return false if the save didn't succeed, including if it has halted by before_save's #1861, #2477 [David Heinemeier Hansson]
* ActiveRecord::Base.remove_connection explicitly closes database connections and doesn't corrupt the connection cache. Introducing the disconnect! instance method for the PostgreSQL, MySQL, and SQL Server adapters; implementations for the others are welcome. #3591 [Simon Stapleton, Tom Ward]
* Added calculations: Base.count, Base.average, Base.sum, Base.minimum, Base.maxmium, and the generic Base.calculate. All can be used with :group and :having. Calculations and statitics need no longer require custom SQL. #3958 [Rick Olson]. Examples:
* Improved the Oracle OCI Adapter with better performance for column reflection (from #3210), fixes to migrations (from #3476 and #3742), tweaks to unit tests (from #3610), and improved documentation (from #2446) #3879 [Aggregated by schoenm@earthlink.net]
* Fixed that the schema_info table used by ActiveRecord::Schema.define should respect table pre- and suffixes #3834 [rubyonrails@atyp.de]
* Added :select option to Base.count that'll allow you to select something else than * to be counted on. Especially important for count queries using DISTINCT #3839 [Stefan Kaes]
* Fixed schema handling for DB2 adapter that didn't work: an initial schema could be set, but it wasn't used when getting tables and indexes #3678 [Maik Schmidt]
* Added option to Base.reflection_of_all_associations to specify a specific association to scope the call. For example Base.reflection_of_all_associations(:has_many) [David Heinemeier Hansson]
* Added ActiveRecord::SchemaDumper.ignore_tables which tells SchemaDumper which tables to ignore. Useful for tables with funky column like the ones required for tsearch2. [Tobias Lütke]
* SchemaDumper now doesn't fail anymore when there are unknown column types in the schema. Instead the table is ignored and a Comment is left in the schema.rb. [Tobias Lütke]
* Fixed that saving a model with multiple habtm associations would only save the first one. #3244 [yanowitz-rubyonrails@quantumfoam.org, Florian Weber]
* Added option to specify :group, :limit, :offset, and :select options from find on has_and_belongs_to_many and has_many assosociations [David Heinemeier Hansson]
* MySQL: work around ruby-mysql/mysql-ruby inconsistency with mysql.stat. Eliminate usage of mysql.ping because it doesn't guarantee reconnect. Explicitly close and reopen the connection instead. [Jeremy Kemper]
* Fixed bug where using update_attribute after pushing a record to a habtm association of the object caused duplicate rows in the join table. #2888 [colman@rominato.com, Florian Weber, Michael Schoen]
* MySQL, PostgreSQL: reconnect! also reconfigures the connection. Otherwise, the connection 'loses' its settings if it times out and is reconnected. #2978 [Shugo Maeda]
* MySQL: introduce :encoding option to specify the character set for client, connection, and results. Only available for MySQL 4.1 and later with the mysql-ruby driver. Do SHOW CHARACTER SET in mysql client to see available encodings. #2975 [Shugo Maeda]
* Introducing the Firebird adapter. Quote columns and use attribute_condition more consistently. Setup guide: http://wiki.rubyonrails.com/rails/pages/Firebird+Adapter #1874 [Ken Kunz <kennethkunz@gmail.com>]
* Oracle: active? and reconnect! methods for handling stale connections. Optionally retry queries after reconnect. #428 [Michael Schoen <schoenm@earthlink.net>]
* When AbstractAdapter#log rescues an exception, attempt to detect and reconnect to an inactive database connection. Connection adapter must respond to the active? and reconnect! instance methods. Initial support for PostgreSQL, MySQL, and SQLite. Make certain that all statements which may need reconnection are performed within a logged block: for example, this means no avoiding log(sql, name) { } if @logger.nil? #428 [Jeremy Kemper]
* Upgrade bundled ruby-mysql 0.2.4 with mysql411 shim (see #440) to ruby-mysql 0.2.6 with a patchset for 4.1 protocol support. Local change [301] is now a part of the main driver; reapplied local change [2182]. Removed GC.start from Result.free. [tommy@tmtm.org, akuroda@gmail.com, Doug Fales <doug.fales@gmail.com>, Jeremy Kemper]
* Destroy associated has_and_belongs_to_many records after all before_destroy callbacks but before destroy. This allows you to act on the habtm association as you please while preserving referential integrity. #2065 [larrywilliams1@gmail.com, sam.kirchmeier@gmail.com, elliot@townx.org, Jeremy Kemper]
* Rename Base.constrain to Base.with_scope so it doesn't conflict with existing concept of database constraints. Make scoping more robust: uniform method => parameters, validated method names and supported finder parameters, raise exception on nested scopes. [Jeremy Kemper] Example:
* Added find_or_create_by_X as a second type of dynamic finder that'll create the record if it doesn't already exist [David Heinemeier Hansson]. Example:
* Correct the pure-Ruby MySQL 4.1.1 shim's version test. #2718 [Jeremy Kemper]
* Add Model.create! to match existing model.save! method. When save! raises RecordInvalid, you can catch the exception, retrieve the invalid record (invalid_exception.record), and see its errors (invalid_exception.record.errors). [Jeremy Kemper]
* Correct fixture behavior when table name pluralization is off. #2719 [Rick Bradley <rick@rickbradley.com>]
* Constraints are cloned so they can't be inadvertently modified while they're
in effect. Added :readonly finder constraint. Calling an association collection's class method (Part.foobar via item.parts.foobar) constrains :readonly => false since the collection's :joins constraint would otherwise force it to true. [Jeremy Kemper <rails@bitsweat.net>]
* PostgreSQL sequence support. Use set_sequence_name in your model class to specify its primary key sequence. #2292 [Rick Olson <technoweenie@gmail.com>, Robby Russell <robby@planetargon.com>]
* :dependent now accepts :nullify option. Sets the foreign key of the related objects to NULL instead of deleting them. #2015 [Robby Russell <robby@planetargon.com>]
* Introduce read-only records. If you call object.readonly! then it will mark the object as read-only and raise ReadOnlyRecord if you call object.save. object.readonly? reports whether the object is read-only. Passing :readonly => true to any finder method will mark returned records as read-only. The :joins option now implies :readonly, so if you use this option, saving the same record will now fail. Use find_by_sql to work around.
* Add option (true by default) to generate reader methods for each attribute of a record to avoid the overhead of calling method missing. In partial fullfilment of #1236. [Stefan Kaes]
* model.association.clear should destroy associated objects if :dependent => true instead of nullifying their foreign keys. #2221 [joergd@pobox.com, ObieFernandez <obiefernandez@gmail.com>]
* Recognize PostgreSQL NOW() default as equivalent to CURRENT_TIMESTAMP or CURRENT_DATE, depending on the column's type. #2256 [mat <mat@absolight.fr>]
* Clean up Fixtures.reset_sequences for PostgreSQL. Handle tables with no rows and models with custom primary keys. #2174, #2183 [jay@jay.fm, Blair Zajac <blair@orcaware.com>]
* Standardize the interpretation of boolean columns in the Mysql and Sqlite adapters. (Use MysqlAdapter.emulate_booleans = false to disable this behavior)
* Added AbstractAdapter#select_value and AbstractAdapter#select_values as convenience methods for selecting single values, instead of hashes, of the first column in a SELECT #2283 [solo@gatelys.com]
* Wrap :conditions in parentheses to prevent problems with OR's #1871 [Jamis Buck]
* Allow the postgresql adapter to work with the SchemaDumper. [Jamis Buck]
* Add ActiveRecord::SchemaDumper for dumping a DB schema to a pure-ruby file, making it easier to consolidate large migration lists and port database schemas between databases. [Jamis Buck]
* Added check for RAILS_CONNECTION_ADAPTERS on startup and only load the connection adapters specified within if its present (available in Rails through config.connection_adapters using the new config) #1958 [skae]
* Fixed that Active Record would call SHOW FIELDS twice (or more) for the same model when the cached results were available #1947 [sd@notso.net]
* Added log_level and use_silence parameter to ActiveRecord::Base.benchmark. The first controls at what level the benchmark statement will be logged (now as debug, instead of info) and the second that can be passed false to include all logging statements during the benchmark block/
* Default sequence names for Oracle changed to #{table_name}_seq, which is the most commonly used standard. In addition, a new method ActiveRecord::Base#set_sequence_name allows the developer to set the sequence name per model. This is a non-backwards-compatible change -- anyone using the old-style "rails_sequence" will need to either create new sequences, or set: ActiveRecord::Base.set_sequence_name = "rails_sequence" #1798
* OCIAdapter now properly handles synonyms, which are commonly used to separate out the schema owner from the application user #1798
* Fixed the handling of camelCase columns names in Oracle #1798
* Implemented for OCI the Rakefile tasks of :clone_structure_to_test, :db_structure_dump, and :purge_test_database, which enable Oracle folks to enjoy all the agile goodness of Rails for testing. Note that the current implementation is fairly limited -- only tables and sequences are cloned, not constraints or indexes. A full clone in Oracle generally requires some manual effort, and is version-specific. Post 9i, Oracle recommends the use of the DBMS_METADATA package, though that approach requires editing of the physical characteristics generated #1798
* Fixed the handling of multiple blob columns in Oracle if one or more of them are null #1798
* Added :select option to find which can specify a different value than the default *, like find(:all, :select => "first_name, last_name"), if you either only want to select part of the columns or exclude columns otherwise included from a join #1338 [Stefan Kaes]
* Added support for limit and offset with eager loading of has_one and belongs_to associations. Using the options with has_many and has_and_belongs_to_many associations will now raise an ActiveRecord::ConfigurationError #1692 [Rick Olson]
* Fixed that assume_bottom_position (in acts_as_list) could be called on items already last in the list and they would move one position away from the list #1648 [tyler@kianta.com]
* Added ActiveRecord::Base.threaded_connections flag to turn off 1-connection per thread (required for thread safety). By default it's on, but WEBrick in Rails need it off #1685 [Sam Stephenson]
* Added new Migrations framework for describing schema transformations in a way that can be easily applied across multiple databases #1604 [Tobias Lütke] See documentation under ActiveRecord::Migration and the additional support in the Rails rakefile/generator.
* Fixed problem with eager loading when using a has_and_belongs_to_many association using :association_foreign_key #1504 [flash@vanklinkenbergsoftware.nl]
* Fixed that clear_association_cache doesn't delete new associations on new records (so you can safely place new records in the session with Action Pack without having new associations wiped) #1494 [cluon]
* Fixed that adding a record to a has_and_belongs_to collection would always save it -- now it only saves if its a new record #1203 [Alisdair McDiarmid]
* Fixed saving of in-memory association structures to happen as a after_create/after_update callback instead of after_save -- that way you can add new associations in after_create/after_update callbacks without getting them saved twice
* Comprehensive PostgreSQL schema support. Use the optional schema_search_path directive in database.yml to give a comma-separated list of schemas to search for your tables. This allows you, for example, to have tables in a shared schema without having to use a custom table name. See http://www.postgresql.org/docs/8.0/interactive/ddl-schemas.html to learn more. #827 [dave@cherryville.org]
* Added a second parameter to the build and create method for has_one that controls whether the existing association should be replaced (which means nullifying its foreign key as well). By default this is true, but false can be passed to prevent it.
* Added the :if option to all validations that can either use a block or a method pointer to determine whether the validation should be run or not. #1324 [Duane Johnson/jhosteny]. Examples:
Conditional validations such as the following are made possible:
* Changed that destroying an object will only freeze the attributes hash, which keeps the object from having attributes changed (as that wouldn't make sense), but allows for the querying of associations after it has been destroyed.
* Changed the callbacks such that observers are notified before the in-object callbacks are triggered. Without this change, it wasn't possible to act on the whole object in something like a before_destroy observer without having the objects own callbacks (like deleting associations) called first.
* Added Base.save! that attempts to save the record just like Base.save but will raise a RecordInvalid exception instead of returning false if the record is not valid [Dave Thomas]
* Removed the default order by id on has_and_belongs_to_many queries as it could kill performance on large sets (you can still specify by hand with :order)
* Moved build_association and create_association for has_one and belongs_to out of deprecation as they work when the association is nil unlike association.build and association.create, which require the association to be already in place #864
* Added that model.items.delete(child) will delete the child, not just set the foreign key to nil, if the child is dependent on the model #978 [Jeremy Kemper]
* Added adapter independent limit clause as a two-element array with the first being the limit, the second being the offset #795 [Sam Stephenson]. Example:
Developer.find_all nil, 'id ASC', 5 # return the first five developers
Developer.find_all nil, 'id ASC', [3, 8] # return three developers, starting from #8 and forward
This doesn't yet work with the DB2 or MS SQL adapters. Patches to make that happen are encouraged.
* Added alias_method :to_param, :id to Base, such that Active Record objects to be used as URL parameters in Action Pack automatically #812 [Nicholas Seckar/Sam Stephenson]
* Fixed acts_as_list to trigger remove_from_list on destroy after the fact, not before, so a unique position can be maintained #871 [Alisdair McDiarmid]
* Added that all types of after_find/after_initialized callbacks are triggered if the explicit implementation is present, not only the explicit implementation itself
* Added transactional fixtures that uses rollback to undo changes to fixtures instead of DELETE/INSERT -- it's much faster. See documentation under Fixtures #760 [Jeremy Kemper]
* Added better defaults for composed_of, so statements like composed_of :time_zone, :mapping => %w( time_zone time_zone ) can be written without the mapping part (it's now assumed)
* Changed the auto-timestamping feature to use ActiveRecord::Base.default_timezone instead of entertaining the parallel ActiveRecord::Base.timestamps_gmt method. The latter is now deprecated and will throw a warning on use (but still work) #710 [Jamis Buck]
* Fixed that when using validation macros with a custom message, if you happened to use single quotes in the message string you would get a parsing error #657 [tonka]
* Added options to specify an SSL connection for MySQL. Define the following attributes in the connection config (config/database.yml in Rails) to use it: sslkey, sslcert, sslca, sslcapath, sslcipher. To use SSL with no client certs, just set :sslca = '/dev/null'. http://dev.mysql.com/doc/mysql/en/secure-connections.html #604 [daniel@nightrunner.com]
* Added :allow_nil as an explicit option for validates_length_of, so unless that's set to true having the attribute as nil will also return an error if a range is specified as :within #610 [Jeremy Kemper]
* Added that has_many association build and create methods can take arrays of record data like Base#create and Base#build to build/create multiple records at once.
* Added that Base#delete and Base#destroy both can take an array of ids to delete/destroy #336
* Added the option of supplying an array of attributes to Base#create, so that multiple records can be created at once.
* Added the option of supplying an array of ids and attributes to Base#update, so that multiple records can be updated at once (inspired by #526/Duane Johnson). Example
* Added that has_and_belongs_to_many associations with additional attributes also can be created between unsaved objects and only committed to the database when Base#save is called on the associator #524 [Eric Anderson]
* Fixed that records fetched with piggy-back attributes or through rich has_and_belongs_to_many associations couldn't be saved due to the extra attributes not part of the table #522 [Eric Anderson]
* Fixed that the belongs_to and has_one proxy would fail a test like 'if project.manager' -- this unfortunately also means that you can't call methods like project.manager.build unless there already is a manager on the project #492 [Tim Bates]
* Added support for associating unsaved objects #402 [Tim Bates]. Rules that govern this addition:
== Unsaved objects and associations
You can manipulate objects and associations before they are saved to the database, but there is some special behaviour you should be
aware of, mostly involving the saving of associated objects.
=== One-to-one associations
* Assigning an object to a has_one association automatically saves that object, and the object being replaced (if there is one), in
order to update their primary keys - except if the parent object is unsaved (new_record? == true).
* If either of these saves fail (due to one of the objects being invalid) the assignment statement returns false and the assignment
is cancelled.
* If you wish to assign an object to a has_one association without saving it, use the #association.build method (documented below).
* Assigning an object to a belongs_to association does not save the object, since the foreign key field belongs on the parent. It does
not save the parent either.
=== Collections
* Adding an object to a collection (has_many or has_and_belongs_to_many) automatically saves that object, except if the parent object
(the owner of the collection) is not yet stored in the database.
* If saving any of the objects being added to a collection (via #push or similar) fails, then #push returns false.
* You can add an object to a collection without automatically saving it by using the #collection.build method (documented below).
* All unsaved (new_record? == true) members of the collection are automatically saved when the parent is saved.
* Added replace to associations, so you can do project.manager.replace(new_manager) or project.milestones.replace(new_milestones) #402 [Tim Bates]
* Added build and create methods to has_one and belongs_to associations, so you can now do project.manager.build(attributes) #402 [Tim Bates]
* Added that if a before_* callback returns false, all the later callbacks and the associated action are cancelled. If an after_* callback returns false, all the later callbacks are cancelled. Callbacks are generally run in the order they are defined, with the exception of callbacks defined as methods on the model, which are called last. #402 [Tim Bates]
* Fixed that Base#== wouldn't work for multiple references to the same unsaved object #402 [Tim Bates]
* Added the possibility for adapters to overwrite add_limit! to implement a different limiting scheme than "LIMIT X" used by MySQL, PostgreSQL, and SQLite.
* Added that Base#find takes an optional options hash, including :conditions. Base#find_on_conditions deprecated in favor of #find with :conditions #407 [Jeremy Kemper]
* Added a require_association hook on const_missing that makes it possible to use any model class without requiring it first. This makes STI look like:
before:
require_association 'person'
class Employee < Person
end
after:
class Employee < Person
end
This also reduces the usefulness of Controller.model in Action Pack to currently only being for documentation purposes.
* Added the possibility of marking fields as being in error without adding a message (using nil) to it that'll get displayed wth full_messages #208 [mjobin]
* Added Base.validates_confirmation_of that encapsulates the pattern of wanting to validate a password or email address field with a confirmation. Example:
* Added Base.validates_acceptance_of that encapsulates the pattern of wanting to validate the acceptance of a terms of service check box (or similar agreement). Example:
* Fixed problems with primary keys and postgresql sequences (#230) [Tim Bates]
* Added reloading for associations under cached environments like FastCGI and mod_ruby. This makes it possible to use those environments for development.
This is turned on by default, but can be turned off with ActiveRecord::Base.reload_dependencies = false in production environments.
NOTE: This will only have an effect if you let the associations manage the requiring of model classes. All libraries loaded through
require will be "forever" cached. You can, however, use ActiveRecord::Base.load_or_require("library") to get this behavior outside of the
auto-loading associations.
* Added ERB capabilities to the fixture files for dynamic fixture generation. You don't need to do anything, just include ERB blocks like:
david:
id: 1
name: David
jamis:
id: 2
name: Jamis
<% for digit in 3..10 %>
dev_<%= digit %>:
id: <%= digit %>
name: fixture_<%= digit %>
<% end %>
* Changed the yaml fixture searcher to look in the root of the fixtures directory, so when you before could have something like:
fixtures/developers/fixtures.yaml
fixtures/accounts/fixtures.yaml
...you now need to do:
fixtures/developers.yaml
fixtures/accounts.yaml
* Changed the fixture format from:
name: david
data:
id: 1
name: David Heinemeier Hansson
birthday: 1979-10-15
profession: Systems development
---
name: steve
data:
id: 2
name: Steve Ross Kellock
birthday: 1974-09-27
profession: guy with keyboard
...to:
david:
id: 1
name: David Heinemeier Hansson
birthday: 1979-10-15
profession: Systems development
steve:
id: 2
name: Steve Ross Kellock
birthday: 1974-09-27
profession: guy with keyboard
The change is NOT backwards compatible. Fixtures written in the old YAML style needs to be rewritten!
* All associations will now attempt to require the classes that they associate to. Relieving the need for most explicit 'require' statements.
* Added HasAndBelongsToManyAssociation#push_with_attributes(object, join_attributes) that can create associations in the join table with additional
attributes. This is really useful when you have information that's only relevant to the join itself, such as a "added_on" column for an association
between post and category. The added attributes will automatically be injected into objects retrieved through the association similar to the piggy-back
* Added HasAndBelongsToManyAssociation#find(id) that'll search inside the collection and find the object or record with that id
* Added :conditions option to has_and_belongs_to_many that works just like the one on all the other associations
* Added AssociationCollection#clear to remove all associations from has_many and has_and_belongs_to_many associations without destroying the records [geech]
* Added type-checking and remove in 1-instead-of-N sql statements to AssociationCollection#delete [geech]
* Added a return of self to AssociationCollection#<< so appending can be chained, like project << Milestone.create << Milestone.create [geech]
* Added Base#hash and Base#eql? which means that all of the equality using features of array and other containers now works:
