Rails 2.2 delivers a number of new and improved features. This list covers the major upgrades, but doesn't include every little bug fix and change. If you want to see everything, check out the [list of commits](http://github.com/rails/rails/commits/master) in the main Rails repository on GitHub.
Along with Rails, 2.2 marks the launch of the [Ruby on Rails Guides](http://guides.rubyonrails.org/), the first results of the ongoing [Rails Guides hackfest](http://hackfest.rubyonrails.org/guide). This site will deliver high-quality documentation of the major features of Rails.
Along with thread safety, a lot of work has been done to make Rails work well with JRuby and the upcoming Ruby 1.9. With Ruby 1.9 being a moving target, running edge Rails on edge Ruby is still a hit-or-miss proposition, but Rails is ready to make the transition to Ruby 1.9 when the latter is released.
The internal documentation of Rails, in the form of code comments, has been improved in numerous places. In addition, the [Ruby on Rails Guides](http://guides.rubyonrails.org/) project is the definitive source for information on major Rails components. In its first official release, the Guides page includes:
This will put the guides inside `Rails.root/doc/guides` and you may start surfing straight away by opening `Rails.root/doc/guides/index.html` in your favourite browser.
Supporting the etag and last modified timestamp in HTTP headers means that Rails can now send back an empty response if it gets a request for a resource that hasn't been modified lately. This allows you to check whether a response needs to be sent at all.
The work done to make Rails thread-safe is rolling out in Rails 2.2. Depending on your web server infrastructure, this means you can handle more requests with fewer copies of Rails in memory, leading to better server performance and higher utilization of multiple cores.
There are two big additions to talk about here: transactional migrations and pooled database transactions. There's also a new (and cleaner) syntax for join table conditions, as well as a number of smaller improvements.
Historically, multiple-step Rails migrations have been a source of trouble. If something went wrong during a migration, everything before the error changed the database and everything after the error wasn't applied. Also, the migration version was stored as having been executed, which means that it couldn't be simply rerun by `rake db:migrate:redo` after you fix the problem. Transactional migrations change this by wrapping migration steps in a DDL transaction, so that if any of them fail, the entire migration is undone. In Rails 2.2, transactional migrations are supported on PostgreSQL out of the box. The code is extensible to other database types in the future - and IBM has already extended it to support the DB2 adapter.
Connection pooling lets Rails distribute database requests across a pool of database connections that will grow to a maximum size (by default 5, but you can add a `pool` key to your `database.yml` to adjust this). This helps remove bottlenecks in applications that support many concurrent users. There's also a `wait_timeout` that defaults to 5 seconds before giving up. `ActiveRecord::Base.connection_pool` gives you direct access to the pool if you need it.
The new bang! version of `find_by_attribute!` is equivalent to `Model.first(:conditions => {:attribute => value}) || raise ActiveRecord::RecordNotFound` Instead of returning `nil` if it can't find a matching record, this method will raise an exception if it cannot find a match.
Active Record association proxies now respect the scope of methods on the proxied object. Previously (given User has_one :account) `@user.account.private_method` would call the private method on the associated Account object. That fails in Rails 2.2; if you need this functionality, you should use `@user.account.send(:private_method)` (or make the method public instead of private or protected). Please note that if you're overriding `method_missing`, you should also override `respond_to` to match the behavior in order for associations to function normally.
* [Rails 2.2 Change: Private Methods on Association Proxies are Private](http://afreshcup.com/2008/10/24/rails-22-change-private-methods-on-association-proxies-are-private/)
On the controller side, there are several changes that will help tidy up your routes. There are also some internal changes in the routing engine to lower memory usage on complex applications.
Shallow route nesting provides a solution to the well-known difficulty of using deeply-nested resources. With shallow nesting, you need only supply enough information to uniquely identify the resource that you want to work with.
You can now supply an array of methods for new member or collection routes. This removes the annoyance of having to define a route as accepting any verb as soon as you need it to handle more than one. With Rails 2.2, this is a legitimate route declaration:
By default, when you use `map.resources` to create a route, Rails generates routes for seven default actions (index, show, create, new, edit, update, and destroy). But each of these routes takes up memory in your application, and causes Rails to generate additional routing logic. Now you can use the `:only` and `:except` options to fine-tune the routes that Rails will generate for resources. You can supply a single action, an array of actions, or the special `:all` or `:none` options. These options are inherited by nested resources.
* You can now easily [show a custom error page](http://m.onkey.org/2008/7/20/rescue-from-dispatching) for exceptions raised while routing a request.
* The HTTP Accept header is disabled by default now. You should prefer the use of formatted URLs (such as `/customers/1.xml`) to indicate the format that you want. If you need the Accept headers, you can turn them back on with `config.action_controller.use_accept_header = true`.
* Polymorphic URLs behave more sensibly if a passed parameter is nil. For example, calling `polymorphic_path([@project, @date, @area])` with a nil date will give you `project_area_path`.
*`javascript_include_tag` and `stylesheet_link_tag` support a new `:recursive` option to be used along with `:all`, so that you can load an entire tree of files with a single line of code.
Action Mailer now supports mailer layouts. You can make your HTML emails as pretty as your in-browser views by supplying an appropriately-named layout - for example, the `CustomerMailer` class expects to use `layouts/customer_mailer.html.erb`.
Active Support now offers built-in memoization for Rails applications, the `each_with_object` method, prefix support on delegates, and various other new utility methods.
Memoization is a pattern of initializing a method once and then stashing its value away for repeat use. You've probably used this pattern in your own applications:
The `each_with_object` method provides an alternative to `inject`, using a method backported from Ruby 1.9. It iterates over a collection, passing the current element and the memo into the block.
To avoid deployment issues and make Rails applications more self-contained, it's possible to place copies of all of the gems that your Rails application requires in `/vendor/gems`. This capability first appeared in Rails 2.1, but it's much more flexible and robust in Rails 2.2, handling complicated dependencies between gems. Gem management in Rails includes these commands:
*`config.gem _gem_name_` in your `config/environment.rb` file
*`rake gems` to list all configured gems, as well as whether they (and their dependencies) are installed, frozen, or framework (framework gems are those loaded by Rails before the gem dependency code is executed; such gems cannot be frozen)
*`rake gems:install` to install missing gems to the computer
*`rake gems:unpack` to place a copy of the required gems into `/vendor/gems`
*`rake gems:unpack:dependencies` to get copies of the required gems and their dependencies into `/vendor/gems`
*`rake gems:build` to build any missing native extensions
*`rake gems:refresh_specs` to bring vendored gems created with Rails 2.1 into alignment with the Rails 2.2 way of storing them
* If you're a fan of the [Thin](http://code.macournoyer.com/thin/) web server, you'll be happy to know that `script/server` now supports Thin directly.
*`script/plugin install <plugin> -r <revision>` now works with git-based as well as svn-based plugins.
*`script/console` now supports a `--debugger` option
*`Rails::SecretKeyGenerator` has been replaced by `ActiveSupport::SecureRandom`
*`render_component` is deprecated. There's a [render_components plugin](http://github.com/rails/render_component/tree/master) available if you need this functionality.
Previously the above code made available a local variable called `customer` inside the partial 'customer'. You should explicitly pass all the variables via :locals hash now.
*`country_select` has been removed. See the [deprecation page](http://www.rubyonrails.org/deprecation/list-of-countries) for more information and a plugin replacement.
*`ActiveRecord::Base.allow_concurrency` no longer has any effect.
*`ActiveRecord::Errors.default_error_messages` has been deprecated in favor of `I18n.translate('activerecord.errors.messages')`
* The `%s` and `%d` interpolation syntax for internationalization is deprecated.
*`String#chars` has been deprecated in favor of `String#mb_chars`.
* Durations of fractional months or fractional years are deprecated. Use Ruby's core `Date` and `Time` class arithmetic instead.
*`Request#relative_url_root` is deprecated. Use `ActionController::Base.relative_url_root` instead.
