CHANGELOG

*SVN*

* 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 [bitsweat]

* 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
    account.credit_card = CreditCard.create("number" => "123")
    account.save # => CC with id = 12 is destroyed


* Added validates_numericality_of #716 [skanthak/c.r.mcgrath]. Docuemntation:

    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.

* Fixed that the dynamic finder like find_all_by_something_boolean(false) didn't work #649 [lmarlow@yahoo.com]

* Added validates_each that validates each specified attribute against a block #610 [bitsweat]. 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 [bitsweat]

* 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:

    class Project < ActiveRecord::Base
      primary_key "sysid"
      table_name "XYZ_PROJECT"
      inheritance_column { original_inheritance_column + "_id" }
    end

* 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

    people = { 1 => { "first_name" => "David" }, 2 => { "first_name" => "Jeremy"} }
    Person.update(people.keys, people.values)

* 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 [bitsweat]

* 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 [bitsweat]

* 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.

    Before: topic.update_attribute(:approved, !approved?)
    After : topic.toggle!(:approved)

* Added Base#increment!(attribute) and Base#decrement!(attribute) that also saves the records. Example:

    page.views # => 1
    page.increment!(:views) # executes an UPDATE statement
    page.views # => 2
    
    page.increment(:views).increment!(:views)
    page.views # => 4

* Added Base#increment(attribute) and Base#decrement(attribute) that encapsulates the += 1 and -= 1 patterns.


*1.4.0* (January 4th, 2005)

* Added automated optimistic locking if the field <tt>lock_version</tt> is present.  Each update to the
  record increments the lock_version column and the locking facilities ensure that records instantiated twice
  will let the last one saved raise a StaleObjectError if the first was also updated. Example:
  
    p1 = Person.find(1)
    p2 = Person.find(1)
    
    p1.first_name = "Michael"
    p1.save
    
    p2.first_name = "should fail"
    p2.save # Raises a ActiveRecord::StaleObjectError
  
  You're then responsible for dealing with the conflict by rescuing the exception and either rolling back, merging,
  or otherwise apply the business logic needed to resolve the conflict.

  #384 [Michael Koziarski]

* Added dynamic attribute-based finders as a cleaner way of getting objects by simple queries without turning to SQL. 
  They work by appending the name of an attribute to <tt>find_by_</tt>, so you get finders like <tt>Person.find_by_user_name,
  Payment.find_by_transaction_id</tt>. So instead of writing <tt>Person.find_first(["user_name = ?", user_name])</tt>, you just do
  <tt>Person.find_by_user_name(user_name)</tt>.
  
  It's also possible to use multiple attributes in the same find by separating them with "_and_", so you get finders like
  <tt>Person.find_by_user_name_and_password</tt> or even <tt>Payment.find_by_purchaser_and_state_and_country</tt>. So instead of writing
  <tt>Person.find_first(["user_name = ? AND password = ?", user_name, password])</tt>, you just do 
  <tt>Person.find_by_user_name_and_password(user_name, password)</tt>.

  While primarily a construct for easier find_firsts, it can also be used as a construct for find_all by using calls like 
  <tt>Payment.find_all_by_amount(50)</tt> that is turned into <tt>Payment.find_all(["amount = ?", 50])</tt>. This is something not as equally useful,
  though, as it's not possible to specify the order in which the objects are returned.

* Added block-style for callbacks #332 [bitsweat].

    Before:
      before_destroy(Proc.new{ |record| Person.destroy_all "firm_id = #{record.id}" })
    
    After:
      before_destroy { |record| Person.destroy_all "firm_id = #{record.id}" }

* Added :counter_cache option to acts_as_tree that works just like the one you can define on belongs_to #371 [Josh]

* Added Base.default_timezone accessor that determines whether to use Time.local (using :local) or Time.utc (using :utc) when pulling dates 
  and times from the database. This is set to :local by default.

* 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 the possibility of having objects with acts_as_list created before their scope is available or...

* Added a db2 adapter that only depends on the Ruby/DB2 bindings (http://raa.ruby-lang.org/project/ruby-db2/) #386 [Maik Schmidt]

* Added the final touches to the Microsoft SQL Server adapter by Joey Gibson that makes it suitable for actual use #394 [DeLynn Barry]

* Added that Base#find takes an optional options hash, including :conditions. Base#find_on_conditions deprecated in favor of #find with :conditions #407 [bitsweat]

* Added HasManyAssociation#count that works like Base#count #413 [intinig]

* Fixed handling of binary content in blobs and similar fields for Ruby/MySQL and SQLite #409 [xal]

* Fixed a bug in the Ruby/MySQL that caused binary content to be escaped badly and come back mangled #405 [Tobias Luetke]

* Fixed that the const_missing autoload assumes the requested constant is set by require_association and calls const_get to retrieve it. 
  If require_association did not set the constant then const_get will call const_missing, resulting in an infinite loop #380 [bitsweat]

* Fixed broken transactions that were actually only running object-level and not db level transactions [andreas]

* Fixed that validates_uniqueness_of used 'id' instead of defined primary key #406

* Fixed that the overwritten respond_to? method didn't take two parameters like the original #391

* Fixed quoting in validates_format_of that would allow some rules to pass regardless of input #390 [Dmitry V. Sabanin]


*1.3.0* (December 23, 2004)

* 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 that Base.update_all and Base.delete_all return an integer of the number of affected rows #341

* Added scope option to validation_uniqueness #349 [Kent Sibilev]

* Added respondence to *_before_type_cast for all attributes to return their string-state before they were type casted by the column type.
  This is helpful for getting "100,000" back on a integer-based validation where the value would normally be "100".

* Added allow_nil options to validates_inclusion_of so that validation is only triggered if the attribute is not nil [what-a-day]

* Added work-around for PostgreSQL and the problem of getting fixtures to be created from id 1 on each test case.
  This only works for auto-incrementing primary keys called "id" for now #359 [Scott Baron]

* Added Base#clear_association_cache to empty all the cached associations #347 [Tobias Luetke]

* Added more informative exceptions in establish_connection #356 [bitsweat]

* Added Base#update_attributes that'll accept a hash of attributes and save the record (returning true if it passed validation, false otherwise). 

    Before:
      person.attributes = @params["person"]
      person.save
    
    Now:
      person.update_attributes(@params["person"])

* Added Base.destroy and Base.delete to remove records without holding a reference to them first.

* Added that query benchmarking will only happen if its going to be logged anyway #344

* Added higher_item and lower_item as public methods for acts_as_list #342 [Tobias Luetke]

* Fixed that options[:counter_sql] was overwritten with interpolated sql rather than original sql #355 [bitsweat]

* Fixed that overriding an attribute's accessor would be disregarded by add_on_empty and add_on_boundary_breaking because they simply used 
  the attributes[] hash instead of checking for @base.respond_to?(attr.to_s). [Marten]

* Fixed that Base.table_name would expect a parameter when used in has_and_belongs_to_many joins [Anna Lissa Cruz]

* Fixed that nested transactions now work by letting the outer most transaction have the responsibilty of starting and rolling back the transaction.
  If any of the inner transactions swallow the exception raised, though, the transaction will not be rolled back. So always let the transaction
  bubble up even when you've dealt with local issues. Closes #231 and #340.

* Fixed validates_{confirmation,acceptance}_of to only happen when the virtual attributes are not nil #348 [dpiddy@gmail.com]

* Changed the interface on AbstractAdapter to require that adapters return the number of affected rows on delete and update operations.

* Fixed the automated timestamping feature when running under Rails' development environment that resets the inheritable attributes on each request.



*1.2.0*

* Added Base.validates_inclusion_of that validates whether the value of the specified attribute is available in a particular enumerable
  object. [what-a-day]

    class Person < ActiveRecord::Base
      validates_inclusion_of :gender, :in=>%w( m f ), :message=>"woah! what are you then!??!!"
      validates_inclusion_of :age, :in=>0..99
    end

* Added acts_as_list that can decorates an existing class with methods like move_higher/lower, move_to_top/bottom. [Tobias Luetke] Example:

    class TodoItem < ActiveRecord::Base
      acts_as_list :scope => :todo_list_id
      belongs_to :todo_list
    end

* Added acts_as_tree that can decorates an existing class with a many to many relationship with itself. Perfect for categories in 
  categories and the likes. [Tobias Luetke]

* Added that Active Records will automatically record creation and/or update timestamps of database objects if fields of the names 
  created_at/created_on or updated_at/updated_on are present. [Tobias Luetke]

* Added Base.default_error_messages as a hash of all the error messages used in the validates_*_of so they can be changed in one place [Tobias Luetke]

* Added automatic transaction block around AssociationCollection.<<, AssociationCollection.delete, and AssociationCollection.destroy_all

* Fixed that Base#find will return an array if given an array -- regardless of the number of elements #270 [Marten]

* Fixed that has_and_belongs_to_many would generate bad sql when naming conventions differed from using vanilla "id" everywhere [RedTerror]

* Added a better exception for when a type column is used in a table without the intention of triggering single-table inheritance. Example:

    ActiveRecord::SubclassNotFound: The single-table inheritance mechanism failed to locate the subclass: 'bad_class!'.
    This error is raised because the column 'type' is reserved for storing the class in case of inheritance. 
    Please rename this column if you didn't intend it to be used for storing the inheritance class or 
    overwrite Company.inheritance_column to use another column for that information.

* Added that single-table inheritance will only kick in if the inheritance_column (by default "type") is present. Otherwise, inheritance won't
  have any magic side effects.

* 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] 

* Fixed Base.errors to be indifferent as to whether strings or symbols are used. Examples:

    Before:
      errors.add(:name, "must be shorter") if name.size > 10
      errors.on(:name)  # => "must be shorter"
      errors.on("name") # => nil

    After:
      errors.add(:name, "must be shorter") if name.size > 10
      errors.on(:name)  # => "must be shorter"
      errors.on("name") # => "must be shorter"

* Added Base.validates_format_of that Validates whether the value of the specified attribute is of the correct form by matching 
  it against the regular expression provided. [Marcel]

    class Person < ActiveRecord::Base
      validates_format_of :email, :with => /^([^@\s]+)@((?:[-a-z0-9]+\.)+[a-z]{2,})$/, :on => :create
    end

* Added Base.validates_length_of that delegates to add_on_boundary_breaking #312 [Tobias Luetke]. Example:

    Validates that the specified attribute matches the length restrictions supplied in either:
    
      - configuration[:minimum]
      - configuration[:maximum]
      - configuration[:is]
      - configuration[:within] (aka. configuration[:in])
    
    Only one option can be used at a time.
    
      class Person < ActiveRecord::Base
        validates_length_of :first_name, :maximum=>30
        validates_length_of :last_name, :maximum=>30, :message=>"less than %d if you don't mind"
        validates_length_of :user_name, :within => 6..20, :too_long => "pick a shorter name", :too_short => "pick a longer name"
        validates_length_of :fav_bra_size, :minimum=>1, :too_short=>"please enter at least %d character"
        validates_length_of :smurf_leader, :is=>4, :message=>"papa is spelled with %d characters... don't play me."
      end
    
* Added Base.validate_presence as an alternative to implementing validate and doing errors.add_on_empty yourself.

* Added Base.validates_uniqueness_of that alidates whether the value of the specified attributes are unique across the system. 
  Useful for making sure that only one user can be named "davidhh".
  
    class Person < ActiveRecord::Base
      validates_uniqueness_of :user_name
    end
  
  When the record is created, a check is performed to make sure that no record exist in the database with the given value for the specified
  attribute (that maps to a column). When the record is updated, the same check is made but disregarding the record itself.


* Added Base.validates_confirmation_of that encapsulates the pattern of wanting to validate a password or email address field with a confirmation. Example:
 
     Model:
       class Person < ActiveRecord::Base
         validates_confirmation_of :password
       end
  
     View:
       <%= password_field "person", "password" %>
       <%= password_field "person", "password_confirmation" %>
  
   The person has to already have a password attribute (a column in the people table), but the password_confirmation is virtual.
   It exists only as an in-memory variable for validating the password. This check is performed both on create and update.


* 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:
  
   class Person < ActiveRecord::Base
     validates_acceptance_of :terms_of_service
   end
  
  The terms_of_service attribute is entirely virtual. No database column is needed. This check is performed both on create and update.

  NOTE: The agreement is considered valid if it's set to the string "1". This makes it easy to relate it to an HTML checkbox.

  
* Added validation macros to make the stackable just like the lifecycle callbacks. Examples:

    class Person < ActiveRecord::Base
      validate { |record| record.errors.add("name", "too short") unless name.size > 10 }
      validate { |record| record.errors.add("name", "too long")  unless name.size < 20 }
      validate_on_create :validate_password
      
      private
        def validate_password
          errors.add("password", "too short") unless password.size > 6
        end
    end

* Added the option for sanitizing find_by_sql and the offset parts in regular finds [Sam Stephenson]. Examples:

    Project.find_all ["category = ?", category_name], "created ASC", ["? OFFSET ?", 15, 20]
    Post.find_by_sql ["SELECT * FROM posts WHERE author = ? AND created > ?", author_id, start_date]

* Fixed value quoting in all generated SQL statements, so that integers are not surrounded in quotes and that all sanitation are happening
  through the database's own quoting routine. This should hopefully make it lots easier for new adapters that doesn't accept '1' for integer
  columns.

* Fixed has_and_belongs_to_many guessing of foreign key so that keys are generated correctly for models like SomeVerySpecialClient 
  [Florian Weber]

* Added counter_sql option for has_many associations [bitsweat]. Documentation:

    <tt>:counter_sql</tt> - specify a complete SQL statement to fetch the size of the association. If +:finder_sql+ is
    specified but +:counter_sql+, +:counter_sql+ will be generated by replacing SELECT ... FROM with SELECT COUNT(*) FROM.

* Fixed that methods wrapped in callbacks still return their original result #260 [bitsweat]

* Fixed the Inflector to handle the movie/movies pair correctly #261 [Scott Baron]

* Added named bind-style variable interpolation #281 [Michael Koziarski]. Example:

    Person.find(["id = :id and first_name = :first_name", { :id => 5, :first_name = "bob' or 1=1" }])

* Added bind-style variable interpolation for the condition arrays that uses the adapter's quote method [Michael Koziarski]

  Before:
    find_first([ "user_name = '%s' AND password = '%s'", user_name, password ])]
    find_first([ "firm_id = %s", firm_id ])] # unsafe!

  After:
    find_first([ "user_name = ? AND password = ?", user_name, password ])]
    find_first([ "firm_id = ?", firm_id ])]

* Added CSV format for fixtures #272 [what-a-day]. (See the new and expanded documentation on fixtures for more information)

* Fixed fixtures using primary key fields called something else than "id" [dave]

* Added proper handling of time fields that are turned into Time objects with the dummy date of 2000/1/1 [HariSeldon]

* Added reverse order of deleting fixtures, so referential keys can be maintained #247 [Tim Bates]

* Added relative path search for sqlite dbfiles in database.yml (if RAILS_ROOT is defined) #233 [bitsweat]

* Added option to establish_connection where you'll be able to leave out the parameter to have it use the RAILS_ENV environment variable

* 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.


*1.1.0* (34)

* Added automatic fixture setup and instance variable availability. Fixtures can also be automatically 
  instantiated in instance variables relating to their names using the following style:

    class FixturesTest < Test::Unit::TestCase
      fixtures :developers # you can add more with comma separation

      def test_developers
        assert_equal 3, @developers.size # the container for all the fixtures is automatically set
        assert_kind_of Developer, @david # works like @developers["david"].find
        assert_equal "David Heinemeier Hansson", @david.name
      end
    end

* 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
  approach:
  
    post.categories.push_with_attributes(category, :added_on => Date.today)
    post.categories.first.added_on # => Date.today
    
  NOTE: The categories table doesn't have a added_on column, it's the categories_post join table that does!

* Fixed that :exclusively_dependent and :dependent can't be activated at the same time on has_many associations [bitsweat]

* Fixed that database passwords couldn't be all numeric [bitsweat]

* Fixed that calling id would create the instance variable for new_records preventing them from being saved correctly [bitsweat]

* Added sanitization feature to HasManyAssociation#find_all so it works just like Base.find_all [Sam Stephenson/bitsweat]

* Added that you can pass overlapping ids to find without getting duplicated records back [bitsweat]

* Added that Base.benchmark returns the result of the block [bitsweat]

* Fixed problem with unit tests on Windows with SQLite [paterno]

* Fixed that quotes would break regular non-yaml fixtures [Dmitry Sabanin/daft]

* Fixed fixtures on windows with line endings cause problems under unix / mac [Tobias Luetke]

* 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:

    [ Person.find(1), Person.find(2), Person.find(3) ] & [ Person.find(1), Person.find(4) ] # => [ Person.find(1) ]

* Added :uniq as an option to has_and_belongs_to_many which will automatically ensure that AssociateCollection#uniq is called
  before pulling records out of the association. This is especially useful for three-way (and above) has_and_belongs_to_many associations.

* Added AssociateCollection#uniq which is especially useful for has_and_belongs_to_many associations that can include duplicates,
  which is common on associations that also use metadata. Usage: post.categories.uniq

* Fixed respond_to? to use a subclass specific hash instead of an Active Record-wide one

* Fixed has_and_belongs_to_many to treat associations between classes in modules properly [Florian Weber]

* Added a NoMethod exception to be raised when query and writer methods are called for attributes that doesn't exist [geech]

* Added a more robust version of Fixtures that throws meaningful errors when on formatting issues [geech]

* Added Base#transaction as a compliment to Base.transaction for prettier use in instance methods [geech]

* Improved the speed of respond_to? by placing the dynamic methods lookup table in a hash [geech]

* Added that any additional fields added to the join table in a has_and_belongs_to_many association 
  will be placed as attributes when pulling records out through has_and_belongs_to_many associations. 
  This is helpful when have information about the association itself that you want available on retrival.

* Added better loading exception catching and RubyGems retries to the database adapters [alexeyv]

* Fixed bug with per-model transactions [daniel]

* Fixed Base#transaction so that it returns the result of the last expression in the transaction block [alexeyv]

* Added Fixture#find to find the record corresponding to the fixture id. The record 
  class name is guessed by using Inflector#classify (also new) on the fixture directory name.
  
    Before: Document.find(@documents["first"]["id"])
    After : @documents["first"].find

* Fixed that the table name part of column names ("TABLE.COLUMN") wasn't removed properly [Andreas Schwarz]

* Fixed a bug with Base#size when a finder_sql was used that didn't capitalize SELECT and FROM [geech]

* Fixed quoting problems on SQLite by adding quote_string to the AbstractAdapter that can be overwritten by the concrete
  adapters for a call to the dbm. [Andreas Schwarz]
  
* Removed RubyGems backup strategy for requiring SQLite-adapter -- if people want to use gems, they're already doing it with AR.


*1.0.0 (35)*

* Added OO-style associations methods [Florian Weber]. Examples:

    Project#milestones_count       => Project#milestones.size
    Project#build_to_milestones    => Project#milestones.build
    Project#create_for_milestones  => Project#milestones.create
    Project#find_in_milestones     => Project#milestones.find
    Project#find_all_in_milestones => Project#milestones.find_all

* Added serialize as a new class method to control when text attributes should be YAMLized or not. This means that automated
  serialization of hashes, arrays, and so on WILL NO LONGER HAPPEN (#10). You need to do something like this:
  
    class User < ActiveRecord::Base
      serialize :settings
    end
  
  This will assume that settings is a text column and will now YAMLize any object put in that attribute. You can also specify
  an optional :class_name option that'll raise an exception if a serialized object is retrieved as a descendent of a class not in
  the hierarchy. Example:
  
    class User < ActiveRecord::Base
      serialize :settings, :class_name => "Hash"
    end
  
    user = User.create("settings" => %w( one two three ))
    User.find(user.id).settings # => raises SerializationTypeMismatch

* Added the option to connect to a different database for one model at a time. Just call establish_connection on the class
  you want to have connected to another database than Base. This will automatically also connect decendents of that class
  to the different database [Renald Buter].

* Added transactional protection for Base#save. Validations can now check for values knowing that it happens in a transaction and callbacks
  can raise exceptions knowing that the save will be rolled back. [Suggested by Alexey Verkhovsky]

* Added column name quoting so reserved words, such as "references", can be used as column names [Ryan Platte]

* Added the possibility to chain the return of what happened inside a logged block [geech]:

    This now works: 
      log { ... }.map { ... }

    Instead of doing:
      result = []
      log { result = ... }
      result.map { ... }

* Added "socket" option for the MySQL adapter, so you can change it to something else than "/tmp/mysql.sock" [Anna Lissa Cruz]

* Added respond_to? answers for all the attribute methods. So if Person has a name attribute retrieved from the table schema, 
  person.respond_to? "name" will return true.

* Added Base.benchmark which can be used to aggregate logging and benchmark, so you can measure and represent multiple statements in a single block.
  Usage (hides all the SQL calls for the individual actions and calculates total runtime for them all):

    Project.benchmark("Creating project") do
      project = Project.create("name" => "stuff")
      project.create_manager("name" => "David")
      project.milestones << Milestone.find_all
    end

* Added logging of invalid SQL statements [Suggested by Daniel Von Fange]

* Added alias Errors#[] for Errors#on, so you can now say person.errors["name"] to retrieve the errors for name [Andreas Schwarz]

* Added RubyGems require attempt if sqlite-ruby is not available through regular methods.

* Added compatibility with 2.x series of sqlite-ruby drivers. [Jamis Buck]

* Added type safety for association assignments, so a ActiveRecord::AssociationTypeMismatch will be raised if you attempt to
  assign an object that's not of the associated class. This cures the problem with nil giving id = 4 and fixnums giving id = 1 on 
  mistaken association assignments. [Reported by Andreas Schwarz]

* Added the option to keep many fixtures in one single YAML document [what-a-day]

* Added the class method "inheritance_column" that can be overwritten to return the name of an alternative column than "type" for storing
  the type for inheritance hierarchies. [Dave Steinberg]

* Added [] and []= as an alternative way to access attributes when the regular methods have been overwritten [Dave Steinberg]

* Added the option to observer more than one class at the time by specifying observed_class as an array

* Added auto-id propagation support for tables with arbitrary primary keys that have autogenerated sequences associated with them 
  on PostgreSQL. [Dave Steinberg]

* Changed that integer and floats set to "" through attributes= remain as NULL. This was especially a problem for scaffolding and postgresql. (#49)

* Changed the MySQL Adapter to rely on MySQL for its defaults for socket, host, and port [Andreas Schwarz]

* Changed ActionControllerError to decent from StandardError instead of Exception. It can now be caught by a generic rescue.

* Changed class inheritable attributes to not use eval [Caio Chassot]

* Changed Errors#add to now use "invalid" as the default message instead of true, which means full_messages work with those [Marcel Molina Jr]

* Fixed spelling on Base#add_on_boundry_breaking to Base#add_on_boundary_breaking (old naming still works) [Marcel Molina Jr.]

* Fixed that entries in the has_and_belongs_to_many join table didn't get removed when an associated object was destroyed.

* Fixed unnecessary calls to SET AUTOCOMMIT=0/1 for MySQL adapter [Andreas Schwarz]

* Fixed PostgreSQL defaults are now handled gracefully [Dave Steinberg]

* Fixed increment/decrement_counter are now atomic updates [Andreas Schwarz]

* Fixed the problems the Inflector had turning Attachment into attuchments and Cases into Casis [radsaq/Florian Gross]

* Fixed that cloned records would point attribute references on the parent object [Andreas Schwarz]

* Fixed SQL for type call on inheritance hierarchies [Caio Chassot]

* Fixed bug with typed inheritance [Florian Weber]

* Fixed a bug where has_many collection_count wouldn't use the conditions specified for that association


*0.9.5*

* Expanded the table_name guessing rules immensely [Florian Green]. Documentation:

    Guesses the table name (in forced lower-case) based on the name of the class in the inheritance hierarchy descending
    directly from ActiveRecord. So if the hierarchy looks like: Reply < Message < ActiveRecord, then Message is used
    to guess the table name from even when called on Reply. The guessing rules are as follows:
    * Class name ends in "x", "ch" or "ss": "es" is appended, so a Search class becomes a searches table.
    * Class name ends in "y" preceded by a consonant or "qu": The "y" is replaced with "ies", 
      so a Category class becomes a categories table. 
    * Class name ends in "fe": The "fe" is replaced with "ves", so a Wife class becomes a wives table.
    * Class name ends in "lf" or "rf": The "f" is replaced with "ves", so a Half class becomes a halves table.
    * Class name ends in "person": The "person" is replaced with "people", so a Salesperson class becomes a salespeople table.
    * Class name ends in "man": The "man" is replaced with "men", so a Spokesman class becomes a spokesmen table.
    * Class name ends in "sis": The "i" is replaced with an "e", so a Basis class becomes a bases table.
    * Class name ends in "tum" or "ium": The "um" is replaced with an "a", so a Datum class becomes a data table.
    * Class name ends in "child": The "child" is replaced with "children", so a NodeChild class becomes a node_children table.
    * Class name ends in an "s": No additional characters are added or removed.
    * Class name doesn't end in "s": An "s" is appended, so a Comment class becomes a comments table.
    * Class name with word compositions: Compositions are underscored, so CreditCard class becomes a credit_cards table.
    Additionally, the class-level table_name_prefix is prepended to the table_name and the table_name_suffix is appended.
    So if you have "myapp_" as a prefix, the table name guess for an Account class becomes "myapp_accounts".
    
    You can also overwrite this class method to allow for unguessable links, such as a Mouse class with a link to a
    "mice" table. Example:
    
      class Mouse < ActiveRecord::Base
         def self.table_name() "mice" end
      end
  
  This conversion is now done through an external class called Inflector residing in lib/active_record/support/inflector.rb.

* Added find_all_in_collection to has_many defined collections. Works like this:

    class Firm < ActiveRecord::Base
      has_many :clients
    end
    
    firm.id # => 1
    firm.find_all_in_clients "revenue > 1000" # SELECT * FROM clients WHERE firm_id = 1 AND revenue > 1000

  [Requested by Dave Thomas]

* Fixed finders for inheritance hierarchies deeper than one level [Florian Weber]

* Added add_on_boundry_breaking to errors to accompany add_on_empty as a default validation method. It's used like this:

    class Person < ActiveRecord::Base
      protected
        def validation
          errors.add_on_boundry_breaking "password", 3..20
        end
    end
    
  This will add an error to the tune of "is too short (min is 3 characters)" or "is too long (min is 20 characters)" if
  the password is outside the boundry. The messages can be changed by passing a third and forth parameter as message strings.

* Implemented a clone method that works properly with AR. It returns a clone of the record that 
  hasn't been assigned an id yet and is treated as a new record.

* Allow for domain sockets in PostgreSQL by not assuming localhost when no host is specified [Scott Barron]

* Fixed that bignums are saved properly instead of attempted to be YAMLized [Andreas Schwartz]

* Fixed a bug in the GEM where the rdoc options weren't being passed according to spec [Chad Fowler]

* Fixed a bug with the exclusively_dependent option for has_many


*0.9.4*

* Correctly guesses the primary key when the class is inside a module [Dave Steinberg].

* Added [] and []= as alternatives to read_attribute and write_attribute [Dave Steinberg]

* has_and_belongs_to_many now accepts an :order key to determine in which order the collection is returned [radsaq].

* The ids passed to find and find_on_conditions are now automatically sanitized.

* Added escaping of plings in YAML content.

* Multi-parameter assigns where all the parameters are empty will now be set to nil instead of a new instance of their class.

* Proper type within an inheritance hierarchy is now ensured already at object initialization (instead of first at create)


*0.9.3*

* Fixed bug with using a different primary key name together with has_and_belongs_to_many [Investigation by Scott] 

* Added :exclusively_dependent option to the has_many association macro. The doc reads:

    If set to true all the associated object are deleted in one SQL statement without having their
    before_destroy callback run. This should only be used on associations that depend solely on 
    this class and don't need to do any clean-up in before_destroy. The upside is that it's much
    faster, especially if there's a counter_cache involved.

* Added :port key to connection options, so the PostgreSQL and MySQL adapters can connect to a database server
  running on another port than the default.

* Converted the new natural singleton methods that prevented AR objects from being saved by PStore
  (and hence be placed in a Rails session) to a module. [Florian Weber]

* Fixed the use of floats (was broken since 0.9.0+)

* Fixed PostgreSQL adapter so default values are displayed properly when used in conjunction with 
  Action Pack scaffolding.

* Fixed booleans support for PostgreSQL (use real true/false on boolean fields instead of 0/1 on tinyints) [radsaq]


*0.9.2*

* Added static method for instantly updating a record

* Treat decimal and numeric as Ruby floats [Andreas Schwartz]

* Treat chars as Ruby strings (fixes problem for Action Pack form helpers too)

* Removed debugging output accidently left in (which would screw web applications)


*0.9.1*

* Added MIT license

* Added natural object-style assignment for has_and_belongs_to_many associations. Consider the following model:

    class Event < ActiveRecord::Base
      has_one_and_belongs_to_many :sponsors
    end
    
    class Sponsor < ActiveRecord::Base
      has_one_and_belongs_to_many :sponsors
    end

  Earlier, you'd have to use synthetic methods for creating associations between two objects of the above class:
  
    roskilde_festival.add_to_sponsors(carlsberg)
    roskilde_festival.remove_from_sponsors(carlsberg)

    nike.add_to_events(world_cup)
    nike.remove_from_events(world_cup)
    
  Now you can use regular array-styled methods:
  
    roskilde_festival.sponsors << carlsberg
    roskilde_festival.sponsors.delete(carlsberg)

    nike.events << world_cup
    nike.events.delete(world_cup)

* Added delete method for has_many associations. Using this will nullify an association between the has_many and the belonging
  object by setting the foreign key to null. Consider this model:
  
    class Post < ActiveRecord::Base
      has_many :comments
    end

    class Comment < ActiveRecord::Base
      belongs_to :post
    end

  You could do something like:

    funny_comment.has_post? # => true
    announcement.comments.delete(funny_comment)
    funny_comment.has_post? # => false


*0.9.0*

* Active Record is now thread safe! (So you can use it with Cerise and WEBrick applications)
  [Implementation idea by Michael Neumann, debugging assistance by Jamis Buck]

* Improved performance by roughly 400% on a basic test case of pulling 100 records and querying one attribute. 
  This brings the tax for using Active Record instead of "riding on the metal" (using MySQL-ruby C-driver directly) down to ~50%.
  Done by doing lazy type conversions and caching column information on the class-level.

* Added callback objects and procs as options for implementing the target for callback macros.

* Added "counter_cache" option to belongs_to that automates the usage of increment_counter and decrement_counter. Consider:

    class Post < ActiveRecord::Base
      has_many :comments
    end

    class Comment < ActiveRecord::Base
      belongs_to :post
    end

  Iterating over 100 posts like this:
  
    <% for post in @posts %>
      <%= post.title %> has <%= post.comments_count %> comments
    <% end %>
    
  Will generate 100 SQL count queries -- one for each call to post.comments_count. If you instead add a "comments_count" int column
  to the posts table and rewrite the comments association macro with:

    class Comment < ActiveRecord::Base
      belongs_to :post, :counter_cache => true
    end
  
  Those 100 SQL count queries will be reduced to zero. Beware that counter caching is only appropriate for objects that begin life
  with the object it's specified to belong with and is destroyed like that as well. Typically objects where you would also specify
  :dependent => true. If your objects switch from one belonging to another (like a post that can be move from one category to another),
  you'll have to manage the counter yourself. 

* Added natural object-style assignment for has_one and belongs_to associations. Consider the following model:

    class Project < ActiveRecord::Base
      has_one :manager
    end
    
    class Manager < ActiveRecord::Base
      belongs_to :project
    end
  
  Earlier, assignments would work like following regardless of which way the assignment told the best story:
  
    active_record.manager_id = david.id
  
  Now you can do it either from the belonging side:

    david.project = active_record
  
  ...or from the having side:
  
    active_record.manager = david
  
  If the assignment happens from the having side, the assigned object is automatically saved. So in the example above, the 
  project_id attribute on david would be set to the id of active_record, then david would be saved.

* Added natural object-style assignment for has_many associations [Florian Weber]. Consider the following model:

    class Project < ActiveRecord::Base
      has_many :milestones
    end
    
    class Milestone < ActiveRecord::Base
      belongs_to :project
    end
  
  Earlier, assignments would work like following regardless of which way the assignment told the best story:
  
    deadline.project_id = active_record.id
  
  Now you can do it either from the belonging side:

    deadline.project = active_record
  
  ...or from the having side:
  
    active_record.milestones << deadline
  
  The milestone is automatically saved with the new foreign key.

* API CHANGE: Attributes for text (or blob or similar) columns will now have unknown classes stored using YAML instead of using
  to_s. (Known classes that won't be yamelized are: String, NilClass, TrueClass, FalseClass, Fixnum, Date, and Time).
  Likewise, data pulled out of text-based attributes will be attempted converged using Yaml if they have the "--- " header.
  This was primarily done to be enable the storage of hashes and arrays without wrapping them in aggregations, so now you can do:
  
    user = User.find(1)
    user.preferences = { "background" => "black", "display" => large }
    user.save
    
    User.find(1).preferences # => { "background" => "black", "display" => large }
  
  Please note that this method should only be used when you don't care about representing the object in proper columns in
  the database. A money object consisting of an amount and a currency is still a much better fit for a value object done through
  aggregations than this new option.

* POSSIBLE CODE BREAKAGE: As a consequence of the lazy type conversions, it's a bad idea to reference the @attributes hash
  directly (it always was, but now it's paramount that you don't). If you do, you won't get the type conversion. So to implement
  new accessors for existing attributes, use read_attribute(attr_name) and write_attribute(attr_name, value) instead. Like this:
  
    class Song < ActiveRecord::Base
      # Uses an integer of seconds to hold the length of the song
      
      def length=(minutes)
        write_attribute("length", minutes * 60)
      end
      
      def length
        read_attribute("length") / 60
      end
    end

  The clever kid will notice that this opens a door to sidestep the automated type conversion by using @attributes directly.
  This is not recommended as read/write_attribute may be granted additional responsibilities in the future, but if you think
  you know what you're doing and aren't afraid of future consequences, this is an option.

* Applied a few minor bug fixes reported by Daniel Von Fange.


*0.8.4*

_Reflection_

* Added ActiveRecord::Reflection with a bunch of methods and classes for reflecting in aggregations and associations.

* Added Base.columns and Base.content_columns which returns arrays of column description (type, default, etc) objects.

* Added Base#attribute_names which returns an array of names for the attributes available on the object.

* Added Base#column_for_attribute(name) which returns the column description object for the named attribute.


_Misc_

* Added multi-parameter assignment:

    # Instantiate objects for all attribute classes that needs more than one constructor parameter. This is done
    # by calling new on the column type or aggregation type (through composed_of) object with these parameters.
    # So having the pairs written_on(1) = "2004", written_on(2) = "6", written_on(3) = "24", will instantiate
    # written_on (a date type) with Date.new("2004", "6", "24"). You can also specify a typecast character in the
    # parenteses to have the parameters typecasted before they're used in the constructor. Use i for Fixnum, f for Float,
    # s for String, and a for Array.
  
  This is incredibly useful for assigning dates from HTML drop-downs of month, year, and day.

* Fixed bug with custom primary key column name and Base.find on multiple parameters.

* Fixed bug with dependent option on has_one associations if there was no associated object.


*0.8.3*

_Transactions_

* Added transactional protection for destroy (important for the new :dependent option) [Suggested by Carl Youngblood]

* Fixed so transactions are ignored on MyISAM tables for MySQL (use InnoDB to get transactions)

* Changed transactions so only exceptions will cause a rollback, not returned false.


_Mapping_

* Added support for non-integer primary keys [Aredridel/earlier work by Michael Neumann]
  
    User.find "jdoe"
    Product.find "PDKEY-INT-12"

* Added option to specify naming method for primary key column. ActiveRecord::Base.primary_key_prefix_type can either
  be set to nil, :table_name, or :table_name_with_underscore. :table_name will assume that Product class has a primary key
  of "productid" and :table_name_with_underscore will assume "product_id". The default nil will just give "id".
    
* Added an overwriteable primary_key method that'll instruct AR to the name of the 
  id column [Aredridele/earlier work by Guan Yang]
    
    class Project < ActiveRecord::Base
      def self.primary_key() "project_id" end
    end

* Fixed that Active Records can safely associate inside and out of modules.

    class MyApplication::Account < ActiveRecord::Base
      has_many :clients # will look for MyApplication::Client
      has_many :interests, :class_name => "Business::Interest" # will look for Business::Interest
    end

* Fixed that Active Records can safely live inside modules [Aredridel]

    class MyApplication::Account < ActiveRecord::Base
    end


_Misc_

* Added freeze call to value object assignments to ensure they remain immutable [Spotted by Gavin Sinclair]
 
* Changed interface for specifying observed class in observers. Was OBSERVED_CLASS constant, now is 
  observed_class() class method. This is more consistant with things like self.table_name(). Works like this:

    class AuditObserver < ActiveRecord::Observer
      def self.observed_class() Account end
      def after_update(account)
        AuditTrail.new(account, "UPDATED")
      end
    end

  [Suggested by Gavin Sinclair]

* Create new Active Record objects by setting the attributes through a block. Like this:

    person = Person.new do |p|
      p.name = 'Freddy'
      p.age  = 19
    end

  [Suggested by Gavin Sinclair]


*0.8.2*

* Added inheritable callback queues that can ensure that certain callback methods or inline fragments are
  run throughout the entire inheritance hierarchy. Regardless of whether a descendent overwrites the callback
  method:
  
    class Topic < ActiveRecord::Base
      before_destroy :destroy_author, 'puts "I'm an inline fragment"'
    end
  
  Learn more in link:classes/ActiveRecord/Callbacks.html

* Added :dependent option to has_many and has_one, which will automatically destroy associated objects when 
  the holder is destroyed:
  
    class Album < ActiveRecord::Base
      has_many :tracks, :dependent => true
    end
    
  All the associated tracks are destroyed when the album is.

* Added Base.create as a factory that'll create, save, and return a new object in one step.

* Automatically convert strings in config hashes to symbols for the _connection methods. This allows you
  to pass the argument hashes directly from yaml. (Luke)

* Fixed the install.rb to include simple.rb [Spotted by Kevin Bullock]

* Modified block syntax to better follow our code standards outlined in 
  http://www.rubyonrails.org/CodingStandards


*0.8.1*

* Added object-level transactions [Thanks to Austin Ziegler for Transaction::Simple]

* Changed adapter-specific connection methods to use centralized ActiveRecord::Base.establish_connection,
  which is parametized through a config hash with symbol keys instead of a regular parameter list.
  This will allow for database connections to be opened in a more generic fashion. (Luke)
  
  NOTE: This requires all *_connections to be updated! Read more in:
  http://ar.rubyonrails.org/classes/ActiveRecord/Base.html#M000081

* Fixed SQLite adapter so objects fetched from has_and_belongs_to_many have proper attributes
  (t.name is now name). [Spotted by Garrett Rooney]

* Fixed SQLite adapter so dates are returned as Date objects, not Time objects [Spotted by Gavin Sinclair]

* Fixed requirement of date class, so date conversions are succesful regardless of whether you 
  manually require date or not.


*0.8.0*

* Added transactions

* Changed Base.find to also accept either a list (1, 5, 6) or an array of ids ([5, 7]) 
  as parameter and then return an array of objects instead of just an object

* Fixed method has_collection? for has_and_belongs_to_many macro to behave as a 
  collection, not an association

* Fixed SQLite adapter so empty or nil values in columns of datetime, date, or time type
  aren't treated as current time [Spotted by Gavin Sinclair]


*0.7.6*

* Fixed the install.rb to create the lib/active_record/support directory [Spotted by Gavin Sinclair]
* Fixed that has_association? would always return true [Spotted by Daniel Von Fange]