1
0
Fork 0
easejs/doc/classes.texi

1737 lines
63 KiB
Plaintext

@c This document is part of the ease.js manual
@c Copyright (c) 2011 Mike Gerwitz
@c Permission is granted to copy, distribute and/or modify this document
@c under the terms of the GNU Free Documentation License, Version 1.3
@c or any later version published by the Free Software Foundation;
@c with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
@c Texts. A copy of the license is included in the section entitled ``GNU
@c Free Documentation License''.
@node Classes
@chapter Working With Classes
In Object-Oriented programming, the most common term you are likely to encounter
is ``Class''. A @dfn{class} is like a blueprint for creating an @dfn{object},
which is an @dfn{instance} of that class. Classes contain @dfn{members}, which
include primarily properties and methods. A @dfn{property} is a value, much like
a variable, that a class ``owns''. A @dfn{method}, when comparing with
JavaScript, is a function that is ``owned'' by a class. As a consequence,
properties and methods are not part of the global scope.
JavaScript does not support classes in the manner traditionally understood by
Object-Oriented programmers. This is because JavaScript follows a different
model which instead uses prototypes. Using this model, JavaScript supports
basic instantiation and inheritance. Rather than instantiating classes,
JavaScript instantiates constructors, which are functions. The following example
illustrates how you would typically create a class-like object in JavaScript:
@float Figure, f:class-js
@verbatim
/**
* Declaring "classes" WITHOUT ease.js
*/
// our "class"
var MyClass = function()
{
this.prop = 'foobar';
}
// a class method
MyClass.prototype.getProp = function()
{
return this.prop;
};
// create a new instance of the class and execute doStuff()
var foo = new MyClass();
console.log( foo.getProp() ); // outputs "foobar"
@end verbatim
@caption{Basic ``Class'' in JavaScript @emph{without} using ease.js}
@end float
This gets the job done, but the prototypal paradigm has a number of limitations
amongst its incredible flexibility. For Object-Oriented programmers, it's both
alien and inadequate. That is not to say that it is not useful. In fact, it is
so flexible that an entire Object-Oriented framework was able to be built atop
of it.
ease.js aims to address the limitations of the prototype model and provide a
familiar environment for Object-Oriented developers. Developers should not have
to worry about @emph{how} classes are implemented in JavaScript (indeed, those
details should be encapsulated). You, as a developer, should be concerned with
only how to declare and use the classes. If you do not understand what a
prototype is, that should be perfectly fine. You shouldn't need to understand it
in order to use the library (though, it's always good to understand what a
prototype is when working with JavaScript).
In this chapter and those that follow, we will see the limitations that ease.js
addresses. We will also see how to declare the classes using both prototypes and
ease.js, until such a point where prototypes are no longer adequate.
@menu
* Defining Classes:: Learn how to define a class with ease.js
* Inheritance:: Extending classes from another
* Static Members:: Members whose use do not require instantiation
* Abstract Members:: Declare members, deferring their definition to subtypes
@end menu
@node Defining Classes
@section Defining Classes
@table @code
@item C = Class( string @var{name}, Object @var{dfn} )
Define named class @var{C} identified by @var{name} described by @var{dfn}.
@item C = Class( string @var{name} ).extend( Object @var{dfn} )
Define named class @var{C} identified by @var{name} described by @var{dfn}.
@item C = Class( Object @var{dfn} )
Define anonymous class @var{C} as described by @var{dfn}.
@item C = Class.extend( Object @var{dfn } )
Define anonymous class @var{C} as described by @var{dfn}.
@end table
Class @var{C} can be defined in a number of manners, as listed above, provided
a @dfn{definition object} @var{dfn} containing the class members and options. An
optional string @var{name} may be provided to set an internal identifier for
@var{C}, which may be used for reflection and error messages. If @var{name} is
omitted, @var{C} will be declared anonymous.
@code{Class} must be imported (@pxref{Including}) from @code{easejs.Class}; it
is not available in the global scope.
@anchor{dfnobj}
@subsection Definition Object
@table @code
@item dfn = @{ '[@var{keywords}] @var{name}': @var{value}[, ...] @}
Define definition object @var{dfn} containing a member identified by @var{name},
described by optional @var{keywords} with the value of @var{value}. The member
type is determined by @code{typeof} @var{value}. Multiple members may be
provided in a single definition object.
@end table
The definition object @var{dfn} has the following properties:
@enumerate
@item
The keys represent the @dfn{member declaration}, which may optionally contain
one or more @var{keywords} delimited by spaces. A space must delimit the final
keyword and @var{name}.
@enumerate
@item
@var{keywords} must consist only of recognized tokens, delimited by spaces.
@item
Each token in @var{keywords} must be unique per @var{name}.
@end enumerate
@item
The @var{value} represents the @dfn{member definition}, the type of which
determines what type of member will be declared.
@enumerate
@item
A @var{value} of type @code{function} will define a @dfn{method}, which is an
invokable member whose context is assigned to the class or class instance
depending on @var{keywords}.
@item
All other types of @var{value} will define a @dfn{property} - a mutable
value equal to @var{value}, assigned to a class or instance depending on
@var{keywords}. Properties may be made immutable using @var{keywords}.
@item
Getters/setters may be defined in an ECMAScript 5 or greater environment.
Getters/setters must share the same value for @var{keywords}.
@end enumerate
@item
@var{name} must be unique across all members of @var{dfn}.
@end enumerate
@subsection Member Validations
For any member @var{name}:
@itemize
@item
@var{keywords} of member @var{name} may contain only one access modifier
(@pxref{Access Modifiers}).
@item
See @ref{Member Keywords,,Member Keywords} for @var{keywords} restrictions.
@end itemize
For any member @var{name} declared as a @emph{method}, the following must hold
true:
@itemize
@item
@var{keywords} of member @var{name} may not contain
@ref{Member Keywords,,@code{override}} without a super method of the same
@var{name} (@pxref{Inheritance}).
@item
@var{keywords} of member @var{name} may contain both
@ref{Member Keywords,,@code{static}} and @ref{Member Keywords,,@code{virtual}}
keywords (@pxref{Static Members} and @ref{Inheritance}).
@item
@var{keywords} of member @var{name} may not contain the
@ref{Member Keywords,,@code{const}} keyword.
@item
For any member @var{name} that contains the keyword
@ref{t:keywords,,@code{abstract}} in @var{keywords}, class @var{C} must instead
be declared as an @code{AbstractClass} (@pxref{Abstract Classes}).
@end itemize
@subsection Discussion
In @ref{f:class-js}, we saw how one would conventionally declare a class-like
object (a prototype) in JavaScript. This method is preferred for many
developers, but it is important to recognize that there is a distinct difference
between Prototypal and Classical Object-Oriented development models. Prototypes
lack many of the conveniences and features that are provided by Classical
languages, but they can be emulated with prototypes. As an Object-Oriented
developer, you shouldn't concern yourself with @emph{how} a class is declared in
JavaScript. In true OO fashion, that behavior should be encapsulated. With
ease.js, it is.
Let's take a look at how to declare that exact same class using ease.js:
@float Figure, f:class-easejs
@verbatim
var Class = require( 'easejs' ).Class;
var MyClass = Class(
{
'public prop': 'foobar',
'public getProp': function()
{
return this.prop;
}
} );
// create a new instance of the class and execute doStuff()
var foo = MyClass();
console.log( foo.getProp() ); // outputs "foobar"
@end verbatim
@caption{Basic anonymous class declaration using ease.js}
@end float
That should look much more familiar to Object-Oriented developers. There are a
couple important notes before we continue evaluating this example:
@itemize
@item
The first thing you will likely notice is our use of the @code{public} keyword.
This is optional (the default visibility is public), but always recommended.
Future versions of ease.js may provide warnings when the visibility is omitted.
We will get more into visibility later on.
@item
Unlike @ref{f:class-js,}, we do not use the @code{new} keyword in order to
instantiate our class. You are more than welcome to use the @code{new} keyword
if you wish, but it is optional when using ease.js. This is mainly because
without this feature, if the keyword is omitted, the constructor is called as a
normal function, which could have highly negative consequences. This style of
instantiation also has its benefits, which will be discussed later on.
@item
ease.js's class module is imported using @code{require()} in the above example.
If using ease.js client-side (@pxref{Client-Side Include}), you can instead use
@samp{var Class = easejs.Class}. From this point on, importing the module will
not be included in examples.
@end itemize
The above example declares an anonymous class, which is stored in the
variable @var{MyClass}. By convention, we use CamelCase, with the first letter
capital, for class names (and nothing else).
@menu
* Anonymous vs. Named Classes::
* Constructors:: How to declare a constructor
* Temporary Classes:: Throwaway classes that only need to be used once
* Temporary Instances:: Throwaway instances that only need to be used once
@end menu
@node Anonymous vs. Named Classes
@subsection Anonymous vs. Named Classes
We state that @ref{f:class-easejs,} declared an @dfn{anyonmous class} because
the class was not given a name. Rather, it was simply assigned to a variable,
which itself has a name. To help keep this idea straight, consider the common
act of creating anonymous functions in JavaScript:
@float Figure, f:anon-func
@verbatim
// anonymous
var myFunc = function() {};
// named
function myNamedFunc() {};
@end verbatim
@caption{Anonymous functions in JavaScript}
@end float
If the function itself is not given a name, it is considered to be anonymous,
even though it is stored within a variable. Just as the engine has no idea what
that function is named, ease.js has no idea what the class is named because it
does not have access to the name of the variable to which it was assigned.
Names are not required for classes, but they are recommended. For example,
consider what may happen when your class is output in an error message.
@float Figure, f:anon-err
@verbatim
// call non-existent method
foo.baz();
// TypeError: Object #<anonymous> has no method 'baz'
@end verbatim
@caption{Anonymous classes do not make for useful error messages}
@end float
If you have more than a couple classes in your software, that error message is
not too much help. You are left relying on the stack trace to track down the
error. This same output applies to converting a class to a string or viewing it
in a debugger. It is simply not helpful. If anything, it is confusing. If you've
debugged large JS applications that make liberal use of anonymous functions, you
might be able to understand that frustration.
Fortunately, ease.js permits you to declare a named class. A @dfn{named class}
is simply a class that is assigned a string for its name, so that error
messages, debuggers, etc provide more useful information. @emph{There is
functionally no difference between named and anonymous classes.}
@float Figure, f:class-named
@verbatim
var MyFoo = Class( 'MyFoo', {} ),
foo = MyFoo();
// call non-existent method
foo.baz();
// TypeError: Object #<MyFoo> has no method 'baz'
@end verbatim
@caption{Declaring an empty @emph{named} class}
@end float
Much better! We now have a useful error message and immediately know which class
is causing the issue.
@node Constructors
@subsection Constructors
In JavaScript, we are used to functions themselves being a constructor because
only constructors can be instantiated. With ease.js, constructors are handled in
a manner similar to most other languages, by providing a separate method. The
implementation ease.js chose is very similar to that of PHP's
(@pxref{Constructor Implementation}).
@float Figure, f:constructor
@verbatim
var Foo = Class( 'Foo',
{
'public __construct': function( name )
{
console.log( 'Hello, ' + name + '!' );
}
} );
// instantiate the class, invoking the constructor
Foo( 'World' );
// Output:
// Hello, World!
@end verbatim
@caption{Declaring constructors using ease.js}
@end float
When the class is instantiated, the constructor is invoked, permitting you do to
any necessary initialization tasks before the class can be used. The constructor
operates exactly how you would expect a constructor to in JavaScript, with one
major difference. Returning an object in the constructor does @emph{not} return
that object instead of the new class instance, since this does not make sense in
a Class-based model.
If you wish to prevent a class from being instantiated, simply throw an
exception within the constructor. This is useful if the class is intended to
provide only static methods, or if you wish to enforce a single instance
(one means of achieving a Singleton).
@float Figure, f:constructor-prevent
@verbatim
var Foo = Class( 'Foo',
{
'public __construct': function( name )
{
throw Error( "Cannot instantiate class Foo" );
}
} );
@end verbatim
@caption{Prevent class from being instantiated}
@end float
Constructors are optional. By default, nothing is done after the class is
instantiated.
@node Temporary Classes
@subsection Temporary Classes
In @ref{f:class-easejs,}, we saw that the @code{new} keyword was unnecessary
when instantiating classes. This permits a form of shorthand that is very useful
for creating @dfn{temporary classes}, or ``throwaway`` classes which are used
only once.
Consider the following example:
@float Figure, f:class-tmp
@verbatim
// new instance of anonymous class
var foo = Class(
{
'public bar': function()
{
return 'baz';
}
} )();
foo.bar(); // returns 'baz'
@end verbatim
@caption{Declaring a temporary (throwaway) class}
@end float
In @ref{f:class-tmp,} above, rather than declaring a class, storing that in a
variable, then instantiating it separately, we are doing it in a single command.
Notice the parenthesis at the end of the statement. This invokes the
constructor. Since the @code{new} keyword is unnecessary, a new instance of the
class is stored in the variable @var{foo}.
We call this a temporary class because it is used only to create a single
instance. The class is then never referenced again. Therefore, we needn't even
store it - it's throwaway.
The downside of this feature is that it is difficult to notice unless the reader
is paying very close attention. There is no keyword to tip them off. Therefore,
it is very important to clearly document that you are storing an instance in the
variable rather than an actual class definition. If you follow the CamelCase
convention for class names, then simply do not capitalize the first letter of
the destination variable for the instance.
@node Temporary Instances
@subsection Temporary Instances
Similar to @ref{Temporary Classes,}, you may wish to use an @emph{instance}
temporarily to invoke a method or chain of methods. @dfn{Temporary instances}
are instances that are instantiated in order to invoke a method or chain of
methods, then are immediately discarded.
@float Figure, f:inst-tmp
@verbatim
// retrieve the name from an instance of Foo
var name = Foo().getName();
// method chaining
var car = VehicleFactory().createBody().addWheel( 4 ).addDoor( 2 ).build();
// temporary class with callback
HttpRequest( host, port ).get( path, function( data )
{
console.log( data );
} );
// Conventionally (without ease.js), you'd accomplish the above using the
// 'new' keyword. You may still do this with ease.js, though it is less
// clean looking.
( new Foo() ).someMethod();
@end verbatim
@caption{Declaring a temporary (throwaway) class}
@end float
Rather than storing the class instance, we are using it simply to invoke
methods. The results of those methods are stored in the variable rather than the
class instance. The instance is immediately discarded, since it is no longer
able to be referenced, and is as such a temporary instance.
In order for method chaining to work, each method must return itself.
This pattern is useful for when a class requires instantiation in order to
invoke a method. Classes that intend to be frequently used in this manner should
declare static methods so that they may be accessed without the overhead of
creating a new class instance.
@node Inheritance
@section Inheritance
@table @code
@item C' = Class( string @var{name} ).extend( Object @var{base}, Object @var{dfn} )
Define named class @var{C'} identified by @var{name} as a subtype of @var{base},
described by @var{dfn}. @var{base} may be of type @code{Class} or may be any
enumerable object.
@item C' = C.extend( Object @var{dfn} )
Define anonymous class @var{C'} as a subtype of class @var{C}, described by
@var{dfn}.
@item C' = Class.extend( Object @var{base}, Object @var{dfn} )
Define anonymous class @var{C'} as a subtype of @var{base}, described by
@var{dfn}. @var{base} may be of type @code{Class} or may be any enumerable
object.
@end table
@var{C} is a class as defined in @ref{Defining Classes}. @var{base} may be any
class or object containing enumerable members. @var{dfn} is to be a definition
object as defined in @ref{dfnobj,,Definition Object}.
Provided non-final @var{C} or @var{base} to satisfy requirements of @var{C},
class @var{C'} will be defined as a @dfn{subtype} (child) of @dfn{supertype}
(parent) class @var{C}. Provided @var{base} that does @emph{not} satisfy
requirements of @var{C}, @var{C'} will be functionally equivalent to a subtype
of anonymous class @var{B} as defined by @var{B} = Class( @var{base} ).
@subsection Member Inheritance
Let @var{dfn\_n\^c} denote a member of @var{dfn} in regards to class @var{c}
that matches (case-sensitive) name @var{n}. Let @var{o\_n} denote an override,
represented as boolean value that is true under the condition that both
@var{dfn\_n\^C'} and @var{dfn\_n\^C} are defined values.
@var{C'} will @dfn{inherit} all public and protected members of supertype
@var{C} such that @var{dfn\_n\^C'} = @var{dfn\_n\^C} for each @var{dfn\^C}. For
any positive condition @var{o\_n}, member @var{dfn\_n\^C'} will be said to
@dfn{override} member @var{dfn\_n\^C}, provided that overriding member @var{n}
passes all validation rules associated with the operation. A @code{protected}
member may be @dfn{escalated} to @code{public}, but the reverse is untrue.
@code{private} members are invisible to subtypes.@footnote{This is true
conceptually, but untrue in pre-ES5 environments where ease.js is forced to fall
back (@pxref{Private Member Dilemma}). As such, one should always develop in an
ES5 or later environment to ensure visibility restrictions are properly
enforced.}
For any positive condition @var{o\_n} where member @var{n} is defined as a
@emph{method}:
@itemize
@item
One of the following conditions must always be true:
@itemize
@item
@var{dfn\_n\^C} is declared with the @ref{Member Keywords,,@code{virtual}}
keyword and @var{dfn\_n\^C'} is declared with the
@ref{Member Keywords,,@code{override}} keyword.
@itemize
@item
Note that @var{dfn\_n\^C'} will not become @ref{Member
Keywords,,@code{virtual}} by default (unlike languages such as C++); they
must be explicitly declared as such.
@end itemize
@item
@var{dfn\_n\^C} is declared with the @ref{Member Keywords,,@code{abstract}}
keyword and @var{dfn\_n\^C'} omits the @ref{Member Keywords,,@code{override}}
keywords.
@end itemize
@item
The argument count of method @var{dfn\_n\^C'} must be >= the argument count of
method @var{dfn\_n\^C} to permit polymorphism.
@item
A reference to super method @var{dfn\_n\^C} will be preserved and assigned to
@samp{this.__super} within context of method @var{dfn\_n\^C'}.
@item
A method is said to be @dfn{concrete} when it provides a definition and
@dfn{abstract} when it provides only a declaration
(@pxref{dfnobj,,Definition Object}).
@itemize
@item
Any method @var{n} such that @var{dfn\_n\^C} is declared @code{abstract} may
be overridden by a concrete or abstract method @var{dfn\_n\^C'}.
@item
A method @var{n} may @emph{not} be declared @ref{Member
Keywords,,@code{abstract}} if @var{dfn\_n\^C} is concrete.
@end itemize
@item
Member @var{dfn\_n\^C'} must be a method.
@item
Member @var{dfn\_n\^C} must not have been declared @ref{Member
Keywords,,@code{private}} (@pxref{Private Member Dilemma}).
@end itemize
Members that have been declared @code{static} cannot be overridden
(@pxref{Static Members}).
@subsection Discussion
Inheritance can be a touchy subject among many Object-Oriented developers due to
encapsulation concerns and design considerations over method overrides. The
decision of whether or not inheritance is an appropriate choice over composition
is left to the developer; ease.js provides the facilities for achieving
classical inheritance where it is desired.
@float Figure, f:inheritance-ex
@image{img/inheritance-ex}
@caption{Basic inheritance example}
@end float
In the above example, we would say that @var{LazyDog} and @var{TwoLeggedDog}
are @emph{subtypes} of @var{Dog}, and that @var{Dog} is the @emph{supertype}
of the two. We describe inheritance as an ``is a'' relationship. That is:
@itemize
@item
@var{LazyDog} is a @var{Dog}.
@item
@var{TwoLeggedDog} is also a @var{Dog}.
@item
@var{Dog} is @emph{not} a @var{LazyDog} or a @var{TwoLeggedDog}.
@end itemize
Subtypes @dfn{inherit} all public and protected members of their supertypes
(@pxref{Access Modifiers}). This means that, in the case of our above example,
the @code{walk()} and @code{bark()} methods would be available to our subtypes.
If the subtype also defines a method of the same name, as was done above, it
will @dfn{override} the parent functionality. For now, we will limit our
discussion to public members. How would we represent these classes using
ease.js?
@float Figure, f:inheritance
@verbatim
// our parent class (supertype)
var Dog = Class( 'Dog',
{
'virtual public walk': function()
{
console.log( 'Walking the dog' );
},
'public bark': function()
{
console.log( 'Woof!' );
}
} );
// subclass (child), as a named class
var LazyDog = Class( 'LazyDog' ).extend( Dog,
{
'override public walk': function()
{
console.log( 'Lazy dog refuses to walk.' );
}
} );
// subclass (child), as an anonymous class
var TwoLeggedDog = Dog.extend(
{
'override public walk': function()
{
console.log( 'Walking the dog on two feet' );
}
} );
@end verbatim
@caption{Inheritance in ease.js}
@end float
You should already understand how to define a class (@pxref{Defining Classes}).
The above example introduced two means of @dfn{extending} classes -- defining a
new class that inherits from a parent:
@table @strong
@item Named Subclasses
@var{LazyDog} is defined as a @emph{named} subclass
(@pxref{Anonymous vs. Named Classes}). This syntax requires the use of
@samp{Class( 'Name' )}. The @code{extend()} method then allows you to extend
from an existing class by passing the class reference in as the first argument.
@item Anonymous Subclasses
@var{TwoLeggedDog} was declared as an @emph{anonymous} subclass. The syntax for
this declaration is a bit more concise, but you forfeit the benefits of named
classes (@pxref{Anonymous vs. Named Classes}). In this case, you can simply call
the supertype's @code{extend()} method. Alternatively, you can use the
@samp{Class.extend( Base, @{@} )} syntax, as was used with the named subclass
@var{LazyDog}.
@end table
You are @emph{always} recommended to use the named syntax when declaring classes
in order to provide more useful error messages. If you are willing to deal with
the less helpful error messages, feel free to use anonymous classes for their
conciseness.
@menu
* Understanding Member Inheritance:: How to work with inherited members
* Overriding Methods:: Overriding inherited methods
* Type Checks and Polymorphism:: Substituting similar classes for
one-another
* Visibility Escalation:: Increasing visibility of inherited members
* Final Classes:: Classes that cannot be inherited from
@end menu
@node Understanding Member Inheritance
@subsection Understanding Member Inheritance
In @ref{f:inheritance}, we took a look at how to inherit from a parent class.
What does it mean when we ``inherit'' from a parent? What are we inheriting? The
answer is: the API.
There are two types of APIs that subtypes can inherit from their parents:
@table @emph
@item Public API
This is the API that is accessible to everyone using your class. It contains all
public members. We will be focusing on public members in this chapter.
@item Protected API
Protected members make up a protected API, which is an API available to
subclasses but @emph{not} the outside world. This is discussed more in the
Access Modifiers section (@pxref{Access Modifiers}), so we're going to leave
this untouched for now.
@end table
When a subtype inherits a member from its parent, it acts almost as if that
member was defined in the class itself@footnote{This statement is not to imply
that inheritance is a case of copy-and-paste. There are slight variations, which
are discussed in more detail in the Access Modifiers section (@pxref{Access
Modifiers}).}. This means that the subtype can use the inherited members as if
they were its own (keep in mind that members also include properties). This
means that we @emph{do not} have to redefine the members in order to use them
ourselves.
@var{LazyDog} and @var{TwoLeggedDog} both inherit the @code{walk()} and
@code{bark()} methods from the @var{Dog} supertype. Using @var{LazyDog} as an
example, let's see what happens when we attempt to use the @code{bark()} method
inherited from the parent.
@float Figure, f:using-inherited-members
@verbatim
var LazyDog = Class( 'LazyDog' ).extend( Dog,
{
/**
* Bark when we're poked
*/
'public poke': function()
{
this.bark();
}
} );
// poke() a new instance of LazyDog
LazyDog().poke();
// Output:
// Woof!
@end verbatim
@caption{Using inherited members}
@end float
In @ref{f:using-inherited-members} above, we added a @code{poke()} method to our
@var{LazyDog} class. This method will call the @code{bark()} method that was
inherited from @var{Dog}. If we actually run the example, you will notice that
the dog does indeed bark, showing that we are able to call our parent's method
even though we did not define it ourselves.
@node Overriding Methods
@subsection Overriding Methods
When a method is inherited, you have the option of either keeping the parent's
implementation or overriding it to provide your own. When you override a method,
you replace whatever functionality was defined by the parent. This concept was
used to make our @var{LazyDog} lazy and our @var{TwoLeggedDog} walk on two legs
in @ref{f:inheritance}.
After overriding a method, you may still want to invoke the parent's method.
This allows you to @emph{augment} the functionality rather than replacing it
entirely. ease.js provides a magic @code{__super()} method to do this. This
method is defined only for the overriding methods and calls the parent method
that was overridden.
In order to demonstrate this, let's add an additional subtype to our hierarchy.
@var{AngryDog} will be a subtype of @var{LazyDog}. Not only is this dog lazy,
but he's rather moody.
@float Figure, f:super-method
@verbatim
var AngryDog = Class( 'AngryDog' ).extend( LazyDog,
{
'public poke': function()
{
// augment the parent method
console.log( 'Grrrrrr...' );
// call the overridden method
this.__super();
}
} );
// poke a new AngryDog instance
AngryDog().poke();
// Output:
// Grrrrrr...
// Woof!
@end verbatim
@caption{Using @code{__super()} method}
@end float
If you remember from @ref{f:using-inherited-members}, we added a @code{poke()}
method to @var{LazyDog}. In @ref{f:super-method} above, we are overriding this
method so that @var{AngryDog} growls when you poke him. However, we still want
to invoke @var{LazyDog}'s default behavior when he's poked, so we also call the
@code{__super()} method. This will also make @var{AngryDog} bark like
@var{LazyDog}.
It is important to note that @code{__super()} must be invoked like any other
method. That is, if the overridden method requires arguments, you must pass them
to @code{__super()}. This allows you to modify the argument list before it is
sent to the overridden method.
@node Type Checks and Polymorphism
@subsection Type Checks and Polymorphism
The fact that the API of the parent is inherited is a very important detail. If
the API of subtypes is guaranteed to be @emph{at least} that of the parent, then
this means that a function expecting a certain type can also work with any
subtypes. This concept is referred to as @dfn{polymorphism}, and is a very
powerful aspect of Object-Oriented programming.
Let's consider a dog trainer. A dog trainer can generally train any type of dog
(technicalities aside), so it would stand to reason that we would want our dog
trainer to be able to train @var{LazyDog}, @var{AngryDog}, @var{TwoLeggedDog},
or any other type of @var{Dog} that we may throw at him/her.
@float Figure, f:polymorphism-uml
@image{img/composition-uml}
@caption{Class structure to demonstrate polymorphism}
@end float
Type checks are traditionally performed in JavaScript using the
@code{instanceOf} operator. While this can be used in most inheritance cases
with ease.js, it is not recommended. Rather, you are encouraged to use ease.js's
own methods for determining instance type@footnote{The reason for this will
become clear in future chapters. ease.js's own methods permit checking for
additional types, such as Interfaces.}. Support for the @code{instanceOf}
operator is not guaranteed.
Instead, you have two choices with ease.js:
@table @code
@item Class.isInstanceOf( type, instance );
Returns @code{true} if @var{instance} is of type @var{type}. Otherwise, returns
@code{false}.
@item Class.isA( type, instance );
Alias for @code{Class.isInstanceOf()}. Permits code that may read better
depending on circumstance and helps to convey the ``is a'' relationship that
inheritance creates.
@end table
For example:
@float Figure, f:instanceof-ex
@verbatim
var dog = Dog()
lazy = LazyDog(),
angry = AngryDog();
Class.isInstanceOf( Dog, dog ); // true
Class.isA( Dog, dog ); // true
Class.isA( LazyDog, dog ); // false
Class.isA( Dog, lazy ); // true
Class.isA( Dog, angry ); // true
// we must check an instance
Class.isA( Dog, LazyDog ); // false; instance expected, class given
@end verbatim
@caption{Using ease.js to determine instance type}
@end float
It is important to note that, as demonstrated in @ref{f:instanceof-ex} above, an
@emph{instance} must be passed as a second argument, not a class.
Using this method, we can ensure that the @var{DogTrainer} may only be used with
an instance of @var{Dog}. It doesn't matter what instance of @var{Dog} - be it a
@var{LazyDog} or otherwise. All that matters is that we are given a @var{Dog}.
@float Figure, f:polymorphism-easejs
@verbatim
var DogTrainer = Class( 'DogTrainer',
{
'public __construct': function( dog )
{
// ensure that we are given an instance of Dog
if ( Class.isA( Dog, dog ) === false )
{
throw TypeError( "Expected instance of Dog" );
}
}
} );
// these are all fine
DogTrainer( Dog() );
DogTrainer( LazyDog() );
DogTrainer( AngryDog() );
DogTrainer( TwoLeggedDog() );
// this is not fine; we're passing the class itself
DogTrainer( LazyDog );
// nor is this fine, as it is not a dog
DogTrainer( {} );
@end verbatim
@caption{Polymorphism in ease.js}
@end float
It is very important that you use @emph{only} the API of the type that you are
expecting. For example, only @var{LazyDog} and @var{AngryDog} implement a
@code{poke()} method. It is @emph{not} a part of @var{Dog}'s API. Therefore, it
should not be used in the @var{DogTrainer} class. Instead, if you wished to use
the @code{poke()} method, you should require that an instance of @var{LazyDog}
be passed in, which would also permit @var{AngryDog} (since it is a subtype of
@var{LazyDog}).
Currently, it is necessary to perform this type check yourself. In future
versions, ease.js will allow for argument type hinting/strict typing, which will
automate this check for you.
@node Visibility Escalation
@subsection Visibility Escalation
Let @var{a\_n} denote a numeric level of visibility for @var{dfn\_n\^C} such
that the access modifiers (@pxref{Access Modifiers}) @code{private},
@code{protected} and @code{public} are associated with the values @code{1},
@code{2} and @code{3} respectively. Let @var{a'} represent @var{a} in regards to
@var{C'} (@pxref{Inheritance}).
For any member @var{n} of @var{dfn}, the following must be true:
@itemize
@item
@var{a'\_n} >= @var{a\_n}.
@item
@var{dfn\_n\^C'} cannot be redeclared without providing a new definition
(@var{value}).
@end itemize
@subsubsection Discussion
@dfn{Visibility escalation} is the act of increasing the visibility of a member.
Since private members cannot be inherited, this would then imply that the only
act to be considered "escallation" would be increasing the level of visibility
from @code{protected} to @code{private}.
Many follow the convention of prefixing private members with an underscore but
leaving omitting such a prefix from protected members. This is to permit
visibility escalation without renaming the member. Alternatively, a new member
can be defined without the prefix that will simply call the overridden member
(although this would then not be considered an escalation, since the member name
varies).
In order to increase the visibility, you must override the member; you cannot
simply redeclare it, leaving the parent definition in tact. For properties, this
has no discernible effect unless the @var{value} changes, as you are simply
redefining it. For methods, this means that you are overriding the entire
@var{value}. Therefore, you will either have to provide an alternate
implementation or call @samp{this.__super()} to invoke the original method.
Note that @emph{you cannot de-escalate from public to protected}; this will
result in an error. This ensures that once a class defines an API, subclasses
cannot alter it. That API must be forever for all subtypes to ensure that it
remains polymorphic.
Let's take a look at an example.
@float Figure, f:vis-esc
@verbatim
var Foo = Class(
{
'protected canEscalate': 'baz',
'protected escalateMe': function( arg )
{
console.log( 'In escalateMe' );
},
'public cannotMakeProtected': function()
{
}
} ),
SubFoo = Foo.extend(
{
/**
* Escalating a property means redefining it
*/
'public canEscalate': 'baz',
/**
* We can go protected -> public
*/
'public escalateMe': function( arg )
{
// simply call the parent method
this.__super( arg );
}
} );
@end verbatim
@caption{Visibility can be escalated}
@end float
Note that, in the above example, making the public @var{cannotMakeProtected}
method protected would throw an error.
@node Final Classes
@subsection Final Classes
@table @code
@item F = FinalClass( string @var{name}, Object @var{dfn} )
Define final named class @var{C} identified by @var{name} described by
@var{dfn}.
@item F = FinalClass( string @var{name} ).extend( Object @var{dfn} )
Define final named class @var{C} identified by @var{name} described by
@var{dfn}.
@item F = FinalClass( Object @var{dfn} )
Define anonymous final class @var{C} as described by @var{dfn}.
@item F = FinalClass.extend( Object @var{dfn } )
Define anonymous final class @var{C} as described by @var{dfn}.
@end table
Final classes operate exactly as ``normal'' classes do (@pxref{Defining
Classes}), with the exception that they cannot be inherited from.
@node Static Members
@section Static Members
@dfn{Static members} do not require instantiation of the containing class in
order to be used, but may also be called by instances. They are attached to the
class itself rather than an instance. Static members provide convenience under
certain circumstances where class instantiation is unnecessary and permit
sharing data between instances of a class. However, static members, when used
improperly, can produce poorly designed classes and tightly coupled code that is
also difficult to test. Static properties also introduce problems very similar
to global variables.
Let us consider an implementation of the factory pattern. Class @var{BigBang}
will declare two static methods in order to satisfy different means of
instantiation: @code{fromBraneCollision()} and @code{fromBigCrunch()} (for the
sake of the example, we're not going to address every theory). Let us also
consider that we want to keep track of the number of big bangs in our universe
(perhaps to study whether or not a "Big Crunch" could have potentially
happened in the past) by incrementing a counter each time a new big bang occurs.
Because we are using a static method, we cannot use a property of an instance in
order to store this data. Therefore, we will use a static property of class
@var{BigBang}.
@float Figure, f:static-ex
@verbatim
var BigBang = Class( 'BigBang',
{
/**
* Number of big bangs that has occurred
* @type {number}
*/
'private static _count': 0,
/**
* String representing the type of big bang
* @type {string}
*/
'private _type': '',
/**
* Create a new big bang from the collision of two membranes
*
* @return {BraneSet} the set of branes that collided
*
* @return {BigBang} new big bang
*/
'public static fromBraneCollision': function( brane_set )
{
// do initialization tasks...
return BigBang( 'brane', brane_set.getData() );
},
/**
* Create a new big bang following a "Big Crunch"
*
* @param {BigCrunch} prior crunch
*
* @return {BigBang} new big bang
*/
'public static fromBigCrunch': function( crunch )
{
// do initialization tasks...
return BigBang( 'crunch', crunch.getData() );
},
/**
* Returns the total number of big bangs that have occurred
*
* @return {number} total number of big bangs
*/
'public static getTotalCount': function()
{
return this.$('_count');
}
/**
* Construct a new big bang
*
* @param {string} type big bang type
* @param {object} data initialization data
*
* @return {undefined}
*/
'public __construct': function( type, data )
{
this._type = type;
// do complicated stuff with data
// increment big bang count
this.__self.$( '_count',
this.__self.$('count') + 1
);
},
} );
// create one of each
var brane_bang = BigBang.fromBraneCollision( branes ),
crunch_bang = BigBang.fromBigCrunch( crunch_incident );
console.log( "Total number of big bangs: %d", BigBang.getTotalCount() );
// Total number of big bangs: 2
@end verbatim
@caption{Static member example using the factory pattern}
@end float
Due to limitations of pre-ECMAScript 5 implementations, ease.js's static
implementation must be broken into two separate parts: properties and methods.
@menu
* Static Methods::
* Static Properties::
* Constants:: Immutable static properties
@end menu
@node Static Methods
@subsection Static Methods
In @ref{f:static-ex}, we implemented three static methods: two factory methods,
@code{fromBraneCollision()} and @code{FromBigCrunch()}, and one getter method to
retrieve the total number of big bangs, @code{getTotalCount()}. These methods are
very similar to instance methods we are already used to, with a few important
differences:
@enumerate
@item
Static methods are declared with the @code{static} keyword.
@item
In the body, @code{this} is bound to the class itself, rather than the instance.
@item
Static methods cannot call any non-static methods of the same class without
first instantiating it.
@end enumerate
The final rule above is not true when the situation is reversed. Non-static
methods @emph{can} call static methods through use of the @var{__self} object,
which is a reference to the class itself. That is, @var{this} in a static method
is the same object as @var{this.__self} in a non-static method. This is
demonstrated by @code{getTotalCount()}
@verbatim
this.$('_count')
@end verbatim
and @code{__construct()}.
@verbatim
this.__self.$('_count')
@end verbatim
To help remember @var{__self}, consider what the name states. A class is a
definition used to create an object. The body of a method is a definition, which
is defined on the class. Therefore, even though the body of a method may be
called in the context of an instance, it is still part of the class. As such,
@var{__self} refers to the class.
@node Static Properties
@subsection Static Properties
You have likely noticed by now that static properties are handled a bit
differently than both static methods and non-static properties. This difference
is due to pre-ECMAScript 5 limitations and is discussed at length in the
@ref{Static Implementation} section.
Static properties are read from and written to using the @dfn{static accessor
method} @code{$()}. This method name was chosen because the @code{$} prefix is
common in scripting languages such as BASH, Perl (for scalars) and PHP. The
accessor method accepts two arguments, the second being optional. If only the
first argument is provided, the accessor method acts as a getter, as in
@ref{f:static-ex}'s @code{getTotalCount()}:
@verbatim
return this.$('_count');
@end verbatim
If the second argument is provided, it acts as a setter, as in
@code{__construct()}:
@verbatim
this.__self.$( '_count',
this.__self.$('count') + 1
);
@end verbatim
Setting @code{undefined} values is supported. The @code{delete} operator is not
supported, as its use is both restricted by the language itself and doesn't make
sense to use in this context. As hinted by the example above, the increment and
decrement operators (@code{++} and @code{--}) are not supported because
JavaScript does not permit returning values by reference.
It is important to understand that, currently, the accessor method cannot be
omitted. Consider the following example:
@float Figure, f:static-accessor
@verbatim
var Foo = Class( 'Foo',
{
'public static bar': 'baz',
},
SubFoo = Class( 'SubFoo' ).extend( Foo, {} )
;
// correct
Foo.$( 'bar, 'baz2' );
Foo.$('bar'); // baz2
SubFoo.$('bar'); // baz2
SubFoo.$( 'bar', 'baz3' );
Foo.$('bar'); // baz3
// INCORRECT
Foo.bar = 'baz2';
Foo.bar; // baz2
SubFoo.bar; // undefined
@end verbatim
@caption{Static accessor method cannot be omitted}
@end float
@node Constants
@subsection Constants
@dfn{Constants}, in terms of classes, are immutable static properties. This
means that, once defined, a constant cannot be modified. Since the value is
immutable, it does not make sense to create instances of the property. As such,
constant values are implicitly static. This ensures that each instance, as well
as any static access, references the exact same value. This is especially
important for objects and arrays.
One important difference between other languages, such as PHP, is that ease.js
supports the @ref{Access Modifiers, visibility modifiers} in conjunction with
the @code{const} keyword. That is, you can have public, protected and private
constants. Constants are public by default, like every other type of member.
This feature permits encapsulating constant values, which is important if you
want an immutable value that shouldn't be exposed to the rest of the world (e.g.
a service URL, file path, etc). Consider the following example in which we have
a class responsible for reading mount mounts from @file{/etc/fstab}:
@float Figure, f:const-ex
@verbatim
Class( 'MountPointIterator',
{
'private const _PATH': '/etc/fstab',
'private _mountPoints': [],
'public __construct': function()
{
var data = fs.readFileSync( this.$('_PATH') );
this._parseMountPoints( data );
},
// ...
} );
@end verbatim
@caption{Using the @code{const} keyword}
@end float
In the above example, attempting to access the @var{_PATH} constant from outside
the class would return @code{undefined}. Had the constant been declared as
public, or had the visibility modifier omitted, it could have been accessed just
like any other static property:
@verbatim
// if PATH were a public constant value
MountPointIterator.$('PATH');
@end verbatim
Any attempts to modify the value of a constant will result in an exception. This
will also work in pre-ES5 engines due to use of the @ref{Static Properties,
static accessor method} (@code{$()}).
It is important to note that constants prevent the @emph{value of the property}
from being reassigned. It @emph{does not} prevent modification of the value that
is @emph{referenced} by the property. For example, if we had a constant
@var{foo}, which references an object, such that
@verbatim
'const foo': { a: 'b' }
@end verbatim
it is perfectly legal to alter the object:
@verbatim
MyClass.$('foo').a = 'c';
@end verbatim
@node Abstract Members
@section Abstract Members
@table @code
@item 'abstract [@var{keywords}] @var{name}': @var{args}
Declare an abstract method @var{name} as having @var{args} arguments, having
optional additional keywords
@var{keywords}.
@end table
Abstract members permit defining an API, deferring the implementation to a
subtype. Abstract methods are declared as an array of string argument names
@var{args}.
@verbatim
// declares abstract method 'connect' expecting the two arguments, 'host'
// and 'path'
{ 'abstract connect': [ 'host', 'path' ] }
@end verbatim
@itemize
@item
Abstract members are defined using the @ref{t:keywords,,@code{abstract}}
keyword.
@itemize
@item
Except in interfaces (@pxref{Interfaces}), where the
@ref{t:keywords,,@code{abstract}} keyword is implicit.
@end itemize
@item
Currently, only methods may be declared abstract.
@item
The subtype must implement at least the number of arguments declared in
@var{args}, but the names needn't match.
@itemize
@item
Each name in @var{args} must be a valid variable name, as satisfied by the
regular expression @code{/^[a-z][a-z0-9]*$/i}.
@item
The names are use purely for documentation and are not semantic.
@end itemize
@end itemize
Abstract members may only be a part of one of the following:
@menu
* Interfaces::
* Abstract Classes::
@end menu
@node Interfaces
@subsection Interfaces
@table @code
@item I = Interface( string @var{name}, Object @var{dfn} )
Define named interface @var{I} identified by @var{name} described by @var{dfn}.
@item I = Interface( string @var{name} ).extend( Object @var{dfn} )
Define named interface @var{I} identified by @var{name} described by @var{dfn}.
@item I = Interface( Object @var{dfn} )
Define anonymous interface @var{I} as described by @var{dfn}.
@item I = Interface.extend( Object @var{dfn } )
Define anonymous interface @var{I} as described by @var{dfn}.
@end table
Interfaces are defined with a syntax much like classes (@pxref{Defining
Classes}) with the following properties:
@itemize
@item
Interface @var{I} cannot be instantiated.
@item
Every member of @var{dfn} of @var{I} is implicitly
@ref{t:keywords,,@code{abstract}}.
@itemize
@item
Consequently, @var{dfn} of @var{I} may contain only abstract methods.
@end itemize
@item
Interfaces may only extend other interfaces (@pxref{Inheritance}).
@end itemize
@code{Interface} must be imported (@pxref{Including}) from
@code{easejs.Interface}; it is not available in the global scope.
@subsubsection Implementing Interfaces
@table @code
@item C = Class( @var{name} ).implement( @var{I\_0}[, ...@var{I\_n}] ).extend( @var{dfn} )
Define named class @var{C} identified by @var{name} implementing all interfaces
@var{I}, described by @var{dfn}.
@item C = Class.implement( @var{I\_0}[, ...@var{I\_n} ).extend( @var{dfn} )
Define anonymous class @var{C} implementing all interfaces @var{I}, described
by @var{dfn}.
@end table
Any class @var{C} may implement any interface @var{I}, inheriting its API.
Unlike class inheritance, any class @var{C} may implement one or more
interfaces.
@itemize
@item
Class @var{C} implementing interfaces @var{I} will be considered a subtype of
every @var{I}.
@item
Class @var{C} must either:
@itemize
@item Provide a concrete definition for every member of @var{dfn} of @var{I},
@item or be declared as an @code{AbstractClass} (@pxref{Abstract Classes})
@itemize
@item
@var{C} may be declared as an @code{AbstractClass} while still providing a
concrete definition for some of @var{dfn} of @var{I}.
@end itemize
@end itemize
@end itemize
@subsubsection Discussion
Consider a library that provides a websocket abstraction. Not all environments
support web sockets, so an implementation may need to fall back on long polling
via AJAX, Flash sockets, etc. If websocket support @emph{is} available, one
would want to use that. Furthermore, an environment may provide its own type of
socket that our library does not include support for. Therefore, we would want
to provide developers for that environment the ability to define their own type
of socket implementation to be used in our library.
This type of abstraction can be solved simply by providing a generic API that
any operation on websockets may use. For example, this API may provide
@code{connect()}, @code{onReceive()} and @code{send()} operations, among others. We
could define this API in a @code{Socket} interface:
@float Figure, f:interface-def
@verbatim
var Socket = Interface( 'Socket',
{
'public connect': [ 'host', 'port' ],
'public send': [ 'data' ],
'public onReceive': [ 'callback' ],
'public close': [],
} );
@end verbatim
@caption{Defining an interface}
@end float
We can then provide any number of @code{Socket} implementations:
@float Figure f:interface-impl
@verbatim
var WebSocket = Class( 'WebSocket' ).implement( Socket ).extend(
{
'public connect': function( host, port )
{
// ...
},
// ...
} ),
SomeCustomSocket = Class.implement( Socket ).extend(
{
// ...
} );
@end verbatim
@caption{Implementing an interface}
@end float
Anything wishing to use sockets can work with this interface polymorphically:
@float Figure, f:interface-poly
@verbatim
var ChatClient = Class(
{
'private _socket': null,
__construct: function( socket )
{
// only allow sockets
if ( !( Class.isA( Socket, socket ) ) )
{
throw TypeError( 'Expected socket' );
}
this._socket = socket;
},
'public sendMessage': function( channel, message )
{
this._socket.send( {
channel: channel,
message: message,
} );
},
} );
@end verbatim
@caption{Polymorphism with interfaces}
@end float
We could now use @code{ChatClient} with any of our @code{Socket}
implementations:
@float Figure, f:interface-poly-use
@verbatim
ChatClient( WebSocket() ).sendMessage( '#lobby', "Sweet! WebSockets!" );
ChatClient( SomeCustomSocket() ).sendMessage( '#lobby', "I can chat too!" );
@end verbatim
@caption{Obtaining flexibility via dependency injection}
@end float
The use of the @code{Socket} interface allowed us to create a powerful
abstraction that will allow our library to work across any range of systems. The
use of an interface allows us to define a common API through which all of our
various components may interact without having to worry about the implementation
details - something we couldn't worry about even if we tried, due to the fact
that we want developers to support whatever environment they are developing for.
Let's make a further consideration. Above, we defined a @code{onReceive()}
method which accepts a callback to be called when data is received. What if our
library wished to use an @code{Event} interface as well, which would allow us to
do something like @samp{some_socket.on( 'receive', function() @{@} )}?
@float Figure, f:interface-impl-multi
@verbatim
var AnotherSocket = Class.implement( Socket, Event ).extend(
{
'public connect': // ...
'public on': // ... part of Event
} );
@end verbatim
@caption{Implementing multiple interfaces}
@end float
Any class may implement any number of interfaces. In the above example,
@code{AnotherSocket} implemented both @code{Socket} and @code{Event}, allowing
it to be used wherever either type is expected. Let's take a look:
@float Figure, f:interface-multi-isa
@verbatim
Class.isA( Socket, AnotherSocket() ); // true
Class.isA( Event, AnotherSocket() ); // true
@end verbatim
@caption{Implementors of interfaces are considered subtypes of each implemented
interface}
@end float
Interfaces do not suffer from the same problems as multiple inheritance, because
we are not providing any sort of implementation that may cause conflicts.
One might then ask - why interfaces instead of abstract classes (@pxref{Abstract
Classes})? Abstract classes require subclassing, which tightly couples the
subtype with its parent. One may also only inherit from a single supertype
(@pxref{Inheritance}), which may cause a problem in our library if we used an
abstract class for @code{Socket}, but a developer had to inherit from another
class and still have that subtype act as a @code{Socket}.
Interfaces have no such problem. Implementors are free to use interfaces
wherever they wish and use as many as they wish; they needn't worry that they
may be unable to use the interface due to inheritance or coupling issues.
However, although interfaces facilitate API reuse, they do not aid in code
reuse as abstract classes do@footnote{This is a problem that will eventually be
solved by the introduction of traits/mixins.}.
@node Abstract Classes
@subsection Abstract Classes
@table @code
@item A = AbstractClass( string @var{name}, Object @var{dfn} )
Define named abstract class @var{A} identified by @var{name} described by
@var{dfn}.
@item A = AbstractClass( string @var{name} ).extend( Object @var{dfn} )
Define named abstract class @var{A} identified by @var{name} described by
@var{dfn}.
@item A = AbstractClass( Object @var{dfn} )
Define anonymous abstract class @var{A} as described by @var{dfn}.
@item A = AbstractClass.extend( Object @var{dfn } )
Define anonymous abstract class @var{A} as described by @var{dfn}.
@end table
Abstract classes are defined with a syntax much like classes (@pxref{Defining
Classes}). They act just as classes do, except with the following additional
properties:
@itemize
@item
Abstract class @var{A} cannot be instantiated.
@item
Abstract class @var{A} must contain at least one member of @var{dfn} that is
explicitly declared as @ref{t:keywords,,@code{abstract}}.
@item
Abstract classes may extend both concrete and abstract classes
@end itemize
An abstract class @emph{must} be used if any member of @var{dfn} is declared as
abstract. This serves as a form of self-documenting code, as it would otherwise
not be immediately clear whether or not a class was abstract (one would have to
look through every member of @var{dfn} to make that determination).
@code{AbstractClass} must be imported (@pxref{Including}) from
@code{easejs.AbstractClass}; it is not available in the global scope.
@subsubsection Discussion
Abstract classes allow the partial implementation of an API, deferring portions
of the implementation to subtypes (@pxref{Inheritance}). As an example, let's
consider an implementation of the @dfn{Abstract Factory} pattern@footnote{See
Abstract Factory, GoF}) which is responsible for the instantiation and
initialization of an object without knowing its concrete type.
Our hypothetical library will be a widget abstraction. For this example, let us
consider that we need a system that will work with any number of frameworks,
including jQuery UI, Dojo, YUI and others. A particular dialog needs to render a
simple @code{Button} widget so that the user may click "OK" when they have
finished reading. We cannot instantiate the widget from within the dialog
itself, as that would tightly couple the chosen widget subsystem (jQuery UI,
etc) to the dialog, preventing us from changing it in the future. Alternatively,
we could have something akin to a switch statement in order to choose which type
of widget to instantiate, but that would drastically inflate maintenance costs
should we ever need to add or remove support for other widget system in the
future.
We can solve this problem by allowing another object, a @code{WidgetFactory}, to
perform that instantiation for us. The dialog could accept the factory in its
constructor, like so:
@float Figure, f:abstract-factory-use
@verbatim
Class( 'Dialog',
{
'private _factory': null,
__construct: function( factory )
{
if ( !( Class.isA( WidgetFactory, factory ) ) )
{
throw TypeError( 'Expected WidgetFactory' );
}
this._factory = factory;
},
'public open': function()
{
// before we open the dialog, we need to create and add the widgets
var btn = this._factory.createButtonWidget( 'btn_ok', "OK" );
// ...
},
} );
@end verbatim
@caption{Hypothetical use case for our Abstract Factory}
@end float
We now have some other important considerations. As was previously mentioned,
@code{Dialog} itself could have determined which widget to instantiate. By using
a factory instead, we are moving that logic to the factory, but we are now
presented with a similar issue. If we use something like a switch statement to
decide what class should be instantiated, we are stuck with modifying the
factory each and every time we add or remove support for another widget library.
This is where an abstract class could be of some benefit. Let's consider the
above call to @code{createButtonWidget()}, which accepted two arguments: an id
for the generated DOM element and a label for the button. Clearly, there is some
common initialization logic that can occur between each of the widgets. However,
we do not want to muddy the factory up with log to determine what widget can be
instantiated. The solution is to define the common logic, but defer the actual
instantiation of the @code{Widget} to subtypes:
@float Figure, f:abstract-factory-define
@verbatim
AbstractClass( 'WidgetFactory',
{
'public createButtonWidget': function( id, label )
{
// note that this is a call to an abstract method; the implementation is
// not yet defined
var widget = this.getNewButtonWidget();
// perform common initialization tasks
widget.setId( id );
widget.setLabel( label );
// return the completed widget
return widget;
},
// declared with an empty array because it has no parameters
'abstract protected getNewButtonWidget': [],
} );
@end verbatim
@caption{Defining our Abstract Factory}
@end float
As demonstrated in @ref{f:abstract-factory-define} above, we can see a very
interesting aspect of abstract classes: we are making a call to a method that is
not yet defined (@code{getNewButtonWidget()}@footnote{Note that we declared this
method as @ref{t:keywords,,@code{protected}} in order to encapsulate which
the widget creation logic (@pxref{Access Modifiers Discussion}). Users of the
class should not be concerned with how we accomplish our job. Indeed, they
should be concerned only with the fact that we save them the trouble of
determining which classes need to be instantiated by providing them with a
convenient API.}). Instead, by declaring it
@ref{t:keywords,,@code{abstract}}, we are stating that we want to call this
method, but it is up to a subtype to actually define it. It is for this reason
that abstract classes cannot be instantiated - they cannot be used until each of
the abstract methods have a defined implementation.
We can now define a concrete widget factory (@pxref{Inheritance}) for each of
the available widget libraries@footnote{Of course, the @code{Widget} itself
would be its own abstraction, which may be best accomplished by the Adapter
pattern.}:
@float Figure, f:concrete-abstract-factory
@verbatim
Class( 'JqueryUiWidgetFactory' )
.extend( WidgetFactory,
{
// concrete method
'protected getNewButtonWidget': function()
{
// ...
},
} );
Class( 'DojoWidgetFactory' )
.extend( WidgetFactory,
{
// ...
} );
// ...
@end verbatim
@caption{Defining our concrete factories}
@end float
With that, we have solved our problem. Rather than using a simple switch
statement, we opted for a polymorphic solution:
@float Figure, f:abstract-factory-inject
@verbatim
// we can use whatever widget library we wish by injecting it into Dialog
Dialog( JqueryUiWidgetFactory() ).show();
Dialog( DojoWidgetFactory() ).show();
Dialog( YuiWidgetFactory() ).show();
@end verbatim
@caption{Using our abstract factory @code{WidgetFactory} via dependency
injection}
@end float
Now, adding or removing libraries is as simple as defining or removing a
@code{WidgetFactory} class.
Another noteworthy mention is that this solution could have just as easily used
an interface instead of an abstract class (@pxref{Interfaces}). The reason we
opted for an abstract class in this scenario is due to code reuse (the common
initialization code), but in doing so, we have tightly coupled each subtype with
the supertype @code{WidgetFactory}. There are a number of trade-offs with each
implementation; choose the one that best fits your particular problem.