Active Support is the Rails component responsible for providing Ruby language extensions, utilities, and other transversal stuff. It offers a richer bottom-line at the language level, targeted both at the development of Rails applications, and at the development of Rails itself.
By referring to this guide you will learn:
- The extensions to the Ruby core modules and classes provided by Rails.
- The rest of fundamental libraries available in Rails.
endprologue.
The following values are considered to be blank in a Rails application:
- nil and false,
- strings composed only of whitespace, i.e. matching /\A\s*\z/,
- empty arrays and hashes, and
- any other object that responds to empty? and it is empty.
WARNING: Note that numbers are not mentioned, in particular 0 and 0.0 are not blank.
For example, this method from ActionDispatch::Response uses blank? to easily be robust to nil and whitespace strings in one shot:
def charset
charset = String(headers[“Content-Type”] || headers[“type”]).split(“;”)1
charset.blank? ? nil : charset.strip.split(“=”)1
end
That’s a typical use case for blank?.
Here, the method Rails runs to instantiate observers upon initialization has nothing to do if there are none:
def instantiate_observers
return if @observers.blank?
- …
end
The method present? is equivalent to !blank?:
assert response.body.present? # same as !response.body.blank?
A few fundamental objects in Ruby are singletons. For example, in the whole live of a program the integer 1 refers always to the same instance:
1.object_id # => 3
Math.cos(0).to_i.object_id # => 3
Hence, there’s no way these objects can be duplicated through dup or clone:
true.dup # => TypeError: can’t dup TrueClass
Some numbers which are not singletons are not duplicable either:
0.0.clone # => allocator undefined for Float
(2**1024).clone # => allocator undefined for Bignum
Active Support provides duplicable? to programmatically query an object about this property:
"".duplicable? # => true
false.duplicable? # => false
By definition all objects are duplicable? except nil, false, true, symbols, numbers, and class objects.
WARNING. Using duplicable? is discouraged because it depends on a hard-coded list. Classes have means to disallow duplication like removing dup and clone or raising exceptions from them, only rescue can tell.
The method returning yields its argument to a block and returns it. You tipically use it with a mutable object that gets modified in the block:
def html_options_for_form(url_for_options, options, *parameters_for_url)
returning options.stringify_keys do |html_options|
html_options[“enctype”] = “multipart/form-data” if html_options.delete(“multipart”)
html_options[“action”] = url_for(url_for_options, *parameters_for_url)
end
end
See also Object#tap.
Object#tap exists in Ruby 1.8.7 and 1.9, and it is defined by Active Support for previous versions. This method yields its receiver to a block and returns it.
For example, the following class method from ActionDispatch::TestResponse creates, initializes, and returns a new test response using tap:
def self.from_response(response)
new.tap do |resp|
resp.status = response.status
resp.headers = response.headers
resp.body = response.body
end
end
See also Object#returning.
Sometimes you want to call a method provided the receiver object is not nil, which is something you usually check first.
For instance, note how this method of ActiveRecord::ConnectionAdapters::AbstractAdapter checks if there’s a @logger:
def log_info(sql, name, ms)
if @logger && @logger.debug?
name = ‘s (.1fms)’ % [name || ‘SQL’, ms]
@logger.debug(format_log_entry(name, sql.squeeze(’ ’)))
end
end
You can shorten that using Object#try. This method is a synonim for Object#send except that it returns nil if sent to nil. The previous example could then be rewritten as:
def log_info(sql, name, ms)
if @logger.try(:debug?)
name = ‘s (.1fms)’ % [name || ‘SQL’, ms]
@logger.debug(format_log_entry(name, sql.squeeze(’ ’)))
end
end
The method metaclass returns the singleton class on any object:
String.metaclass # => #
String.new.metaclass # => #<Class:#>
You can evaluate code in the context of any object’s singleton class using class_eval:
class Proc
def bind(object)
block, time = self, Time.now
object.class_eval do
method_name = “__bind_#{time.to_i}_#{time.usec}”
define_method(method_name, &block)
method = instance_method(method_name)
remove_method(method_name)
method
end.bind(object)
end
end
The method acts_like provides a way to check whether some class acts like some other class based on a simple convention: a class that provides the same interface as String defines
def acts_like_string?
end
which is only a marker, its body or return value are irrelevant. Then, client code can query for duck-type-safeness this way:
some_klass.acts_like?(:string)
Rails has classes that act like Date or Time and follow this contract.
All objects in Rails respond to the method to_param, which is meant to return something that represents them as values in a query string, or as a URL fragments.
By default to_param just calls to_s:
7.to_param # => “7”
The return value of to_param should not be escaped:
“Tom & Jerry”.to_param # => “Tom & Jerry”
Several classes in Rails overwrite this method.
For example nil, true, and false return themselves. Array#to_param calls to_param on the elements and joins the result with “/”:
[0, true, String].to_param # => “0/true/String”
Notably, the Rails routing system calls to_param on models to get a value for the :id placeholder. ActiveRecord::Base#to_param returns the id of a model, but you can redefine that method in your models. For example, given
class User
def to_param
“#{id}-#{name.parameterize}”
end
end
we get:
user_path(@user) # => “/users/357-john-smith”
WARNING. Controllers need to be aware of any redifinition of to_param because when a request like that comes in “357-john-smith” is the value of params[:id].
Except for hashes, given an unescaped key this method constructs the part of a query string that would map such key to what to_param returns. For example, given
class User
def to_param
“#{id}-#{name.parameterize}”
end
end
we get:
current_user.to_query(‘user’) # => user=357-john-smith
This method escapes whatever is needed, both for the key and the value:
account.to_query(‘company[name]’)
- => “company%5Bname%5D=Johnson+%26+Johnson”
so its output is ready to be used in a query string.
Arrays return the result of applying to_query to each element with key[] as key, and join the result with “/”:
[3.4, -45.6].to_query(‘sample’)
- => “sample%5B%5D=3.4&sample%5B%5D=-45.6”
Hashes also respond to to_query but with a different signature. If no argument is passed a call generates a sorted series of key/value assigments calling to_query(key) on its values. Then it joins the result with “&”:
{:c => 3, :b => 2, :a => 1}.to_query # => “a=1&b=2&c=3”
The method Hash#to_query accepts an optional namespace for the keys:
{:id => 89, :name => “John Smith”}.to_query(‘user’)
- => “user%5Bid%5D=89&user%5Bname%5D=John+Smith”
The method with_options provides a way to factor out common options in a series of method calls.
Given a default options hash, with_options yields a proxy object to a block. Within the block, methods called on the proxy are forwarded to the receiver with their options merged. For example, you get rid of the duplication in:
class Account < ActiveRecord::Base
has_many :customers, :dependent => :destroy
has_many :products, :dependent => :destroy
has_many :invoices, :dependent => :destroy
has_many :expenses, :dependent => :destroy
end
this way:
class Account < ActiveRecord::Base
with_options :dependent => :destroy do |assoc|
assoc.has_many :customers
assoc.has_many :products
assoc.has_many :invoices
assoc.has_many :expenses
end
end
That idiom may convey grouping to the reader as well. For example, say you want to send a newsletter whose language depends on the user. Somewhere in the mailer you could group locale-dependent bits like this:
I18n.with_options :locale => user.locale, :scope => “newsletter” do |i18n|
subject i18n.t :subject
body i18n.t :body, :user_name => user.name
end
TIP: Since with_options forwards calls to its receiver they can be nested. Each nesting level will merge inherited defaults in addition to their own.
Active Support provides several methods to ease access to instance variables.
The method instance_variable_defined? exists in Ruby 1.8.6 and later, and it is defined for previous versions anyway:
class C
def initialize
@a = 1
end
end
c = C.new
c.instance_variable_defined?(“@a”) # => true
c.instance_variable_defined?(:@a) # => true
c.instance_variable_defined?(“a”) # => NameError: `a’ is not allowed as an instance variable name
c.instance_variable_defined?(“@b”) # => false
c.m
c.instance_variable_defined?(“@b”) # => true
Ruby 1.8 and 1.9 have a method called instance_variables that returns the names of the defined instance variables. But they behave differently, in 1.8 it returns strings whereas in 1.9 it returns symbols. Active Support defines instance_variable_names as a portable way to obtain them as strings:
class C
def initialize(x, y)
@x, @y = x, y
end
end
C.new(0, 1).instance_variable_names # => [“@y”, “@x”]
WARNING: The order in which the names are returned is unespecified, and it indeed depends on the version of the interpreter.
The method instance_values returns a hash that maps instance variable names without “@” to their
corresponding values. Keys are strings both in Ruby 1.8 and 1.9:
class C
def initialize(x, y)
@x, @y = x, y
end
end
C.new(0, 1).instance_values # => {"x" => 0, “y” => 1}
Copies the instance variables of object into self.
Instance variable names in the exclude array are ignored. If object
responds to protected_instance_variables the ones returned are
also ignored. For example, Rails controllers implement that method.
In both arrays strings and symbols are understood, and they have to include
the at sign.
class C
def initialize(x, y, z)
@x, @y, @z = x, y, z
end
end
a = C.new(0, 1, 2)
b = C.new(3, 4, 5)
a.copy_instance_variables_from(b, [:@y])
- a is now: @x = 3, @y = 1, @z = 2
In the example object and self are of the same type, but they don’t need to.
The methods silence_warnings and enable_warnings change the value of $VERBOSE accordingly for the duration of their block, and reset it afterwards:
silence_warnings { Object.const_set “RAILS_DEFAULT_LOGGER”, logger }
You can silence any stream while a block runs with silence_stream:
silence_stream(STDOUT) do
- STDOUT is silent here
end
Silencing exceptions is also possible with suppress. This method receives an arbitrary number of exception classes. If an exception is raised during the execution of the block and is kind_of? any of the arguments, suppress captures it and returns silently. Otherwise the exception is reraised:
- If the user is locked the increment is lost, no big deal.
suppress(ActiveRecord::StaleObjectError) do
current_user.increment! :visits
end
…
The macros cattr_reader, cattr_writer, and cattr_accessor are analogous to their attr_* counterparts but for classes. They initialize a class variable to nil unless it already exists, and generate the corresponding class methods to access it:
class MysqlAdapter < AbstractAdapter
- Generates class methods to access @@emulate_booleans.
cattr_accessor :emulate_booleans
self.emulate_booleans = true
end
Instance methods are created as well for convenience. For example given
module ActionController
class Base
cattr_accessor :logger
end
end
we can access logger in actions. The generation of the writer instance method can be prevented setting :instance_writer to false (not any false value, but exactly false):
module ActiveRecord
class Base
- No pluralize_table_names= instance writer is generated.
cattr_accessor :pluralize_table_names, :instance_writer => false
end
end
Class variables are shared down the inheritance tree. Class instance variables are not shared, but they are not inherited either. The macros class_inheritable_reader, class_inheritable_writer, and class_inheritable_accessor provide accesors for class-level data which is inherited but not shared with children:
module ActionController
class Base
- FIXME: REVISE/SIMPLIFY THIS COMMENT.
- The value of allow_forgery_protection is inherited,
- but its value in a particular class does not affect
- the value in the rest of the controllers hierarchy.
class_inheritable_accessor :allow_forgery_protection
end
end
They accomplish this with class instance variables and cloning on subclassing, there are no class variables involved. Cloning is performed with dup as long as the value is duplicable.
There are some variants specialised in arrays and hashes:
class_inheritable_array
class_inheritable_hash
Those writers take any inherited array or hash into account and extend them rather than overwrite them.
As with vanilla class attribute accessors these macros create convenience instance methods for reading and writing. The generation of the writer instance method can be prevented setting :instance_writer to false (not any false value, but exactly false):
module ActiveRecord
class Base
class_inheritable_accessor :default_scoping, :instance_writer => false
end
end
Since values are copied when a subclass is defined, if the base class changes the attribute after that, the subclass does not see the new value. That’s the point.
There’s a related macro called superclass_delegating_accessor, however, that does not copy the value when the base class is subclassed. Instead, it delegates reading to the superclass as long as the attribute is not set via its own writer. For example, ActionMailer::Base defines delivery_method this way:
module ActionMailer
class Base
superclass_delegating_accessor :delivery_method
self.delivery_method = :smtp
end
end
If for whatever reason an application loads the definition of a mailer class and after that sets ActionMailer::Base.delivery_method, the mailer class will still see the new value. In addition, the mailer class is able to change the delivery_method without affecting the value in the parent using its own inherited class attribute writer.
The subclasses method returns the names of all subclasses of a given class as an array of strings. That comprises not only direct subclasses, but all descendants down the hierarchy:
class C; end
C.subclasses # => []
Integer.subclasses # => [“Bignum”, “Fixnum”]
module M
class A; end
class B1 < A; end
class B2 < A; end
end
module N
class C < M::B1; end
end
M::A.subclasses # => [“N::C”, “M::B2”, “M::B1”]
The order in which these class names are returned is unspecified.
See also Object#subclasses_of in Extensions to All Objects FIX THIS LINK.
Roughly speaking, the remove_class method removes the class objects passed as arguments:
Class.remove_class(Hash, Dir) # => [Hash, Dir]
Hash # => NameError: uninitialized constant Hash
Dir # => NameError: uninitialized constant Dir
More specifically, remove_class attempts to remove constants with the same name as the passed class objects from their parent modules. So technically this method does not guarantee the class objects themselves are not still valid and alive somewhere after the method call:
module M
class A; end
class B < A; end
end
A2 = M::A
M::A.object_id # => 13053950
Class.remove_class(M::A)
M::B.superclass.object_id # => 13053950 (same object as before)
A2.name # => “M::A” (name is hard-coded in object)
WARNING: Removing fundamental classes like String can result in really funky behaviour.
The method remove_subclasses provides a shortcut for removing all descendants of a given class, where “removing” has the meaning explained above:
class A; end
class B1 < A; end
class B2 < A; end
class C < A; end
A.subclasses # => [“C”, “B2”, “B1”]
A.remove_subclasses
A.subclasses # => []
C # => NameError: uninitialized constant C
See also Object#remove_subclasses_of in Extensions to All Objects FIX THIS LINK.
…
…
…
…
…
…
…
…
…
…
Active Support augments the API of arrays to ease certain ways of accessing them. For example, to returns the subarray of elements up to the one at the passed index:
%w(a b c d).to(2) # => %w(a b c)
[].to(7) # => []
Similarly, from returns the tail from the element at the passed index on:
%w(a b c d).from(2) # => %w(c d)
%w(a b c d).from(10) # => nil
[].from(0) # => nil
The methods second, third, fourth, and fifth return the corresponding element (first is builtin). Thanks to social wisdom and positive constructiveness all around, forty_two is also available.
You can pick a random element with rand:
shape_type = [Circle, Square, Triangle].rand
The method in_groups_of splits an array into consecutive groups of a certain size. It returns an array with the groups:
[1, 2, 3].in_groups_of(2) # => [[1, 2], [3, nil]]
or yields them in turn if a block is passed:
<% sample.in_groups_of(3) do |a, b, c| >
<=h a >
<=h b >
<=h c %>
<% end %>
The first example shows in_groups_of fills the last group with as many nil elements as needed to have the requested size. You can change this padding value using the second optional argument:
[1, 2, 3].in_groups_of(2, 0) # => [[1, 2], [3, 0]]
And you can tell the method not to fill the last group passing false:
[1, 2, 3].in_groups_of(2, false) # => [[1, 2], 3]
As a consequence false can’t be a used as a padding value.
The method in_groups splits an array into a certain number of groups. The method returns and array with the groups:
%w(1 2 3 4 5 6 7).in_groups(3)
- => [[“1”, “2”, “3”], [“4”, “5”, nil], [“6”, “7”, nil]]
or yields them in turn if a block is passed:
%w(1 2 3 4 5 6 7).in_groups(3) {|group| p group}
[“1”, “2”, “3”]
[“4”, “5”, nil]
[“6”, “7”, nil]
The examples above show that in_groups fills some groups with a trailing nil element as needed. A group can get at most one of these extra elements, the rightmost one if any. And the groups that have them are always the last ones.
You can change this padding value using the second optional argument:
%w(1 2 3 4 5 6 7).in_groups(3, “0”)
- => [[“1”, “2”, “3”], [“4”, “5”, “0”], [“6”, “7”, “0”]]
And you can tell the method not to fill the smaller groups passing false:
%w(1 2 3 4 5 6 7).in_groups(3, false)
- => [[“1”, “2”, “3”], [“4”, “5”], [“6”, “7”]]
As a consequence false can’t be a used as a padding value.
The method split divides an array by a separator and returns the resulting chunks.
If a block is passed the separators are those elements of the array for which the block returns true:
(-5..5).to_a.split { |i| i.multiple_of?(4) }
- => [[-5], [-3, -2, -1], [1, 2, 3], 5]
Otherwise, the value received as argument, which defaults to nil, is the separator:
[0, 1, -5, 1, 1, “foo”, “bar”].split(1)
- => [0, [-5], [], [“foo”, “bar”]]
NOTE: Observe in the previous example that consecutive separators result in empty arrays.
…
…
…
…
…
…
…
…
…
…
…
…
…
…
- April 18, 2009: Initial version by Xavier Noria