Top Ten Reasons Cocoa is Great Because of Objective-C (and Not In Spite of It)
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
10. No need for external object models: COM, SOM, DCOM, or CORBA
Once upon a time, a bunch of software companies heard a rumor that objects, so-called reusable components, were going to be really big. They might have even heard about the seminal book, Object Oriented Programming: An Evolutionary Approach by Dr. Brad J. Cox, who invented Objective-C. Nobody wanted to miss out on the coming software industrial revolution that would breed a vibrant marketplace for software components.
Microsoft and IBM were particularly interested in software components. Microsoft invented the Component Object Model (COM), and IBM invented System Object Model (SOM). An international computer industry consortium invented Common Object Request Broker Architecture (CORBA). All of these systems were intended to enable the use of reusable software components with multiple programming languages. All of these systems attempted to add dynamic late binding and introspection to programming languages and environments that emphasized static early binding.
In the interest of brevity, focus on just COM: In order to let programs discover what reusable components are available and what those components can do, COM provides the IUnknow "interface". IUnknown contains exactly three "methods:" AddRef, Release, and QueryInterface. If you have ever programmed with Cocoa or Apple's Core Foundation, you will recognize AddRef as -retain or CFRetain(). You'll recognize Release as -release or CFRelease(). So what's QueryInterface? It's a lot like Cocoa's NSObject -conformsToProtocol: method. It tells you whether or not an anonymous object provides a group of methods.
SOM and CORBA are now nearly non-existent. COM was rebranded Object Linking and Embedding (OLE) and then Active-X and is not currently hyped very much even though it's ubiquitous in Windows and necessary for using Microsoft technologies like DirectX. Curiously, Steve Jobs stood on stage in 1995 (start 3 minutes into http://www.youtube.com/watch?v=Hu-jvAWTZ9o) and demonstrated an implementation of Distributed-COM (DCOM) years before Microsoft delivered it. Of course, Steve was using Objective-C and a compatibility adapter on the Windows side of a network connection.
Bottom line: Objective-C contains (as a standard feature) the ability to use methods of anonymous objects and determine what methods are available. There is no need to glue on COM, SOM, DCOM, or CORBA even when sending messages to objects over network connections. Objective-C was created by the man who espoused reusable software components, and Cocoa is the highest example of object reuse. Java and C# are following Objective-C's lead on this technology.
9. Objective-C is C
For better or worse, the deluxe macro-assembler that we call the C programming language is the lingua franca of programming. Billions of lines of C code execute daily. Most operating systems are written in C. Device drivers are written in C. Libraries ranging from OpenGL to POSIX are written in C and/or provide C interfaces. A language that can't reuse C libraries is pretty much dead on arrival; that's why they all can. It just comes down to the degree of pain programmers must suffer because of the almost inevitable impedance mismatch between any other language and C. Even C++ programmers sometimes have difficulty reusing C libraries.
Objective-C is C. Unlike C++, any legal C program will compile with an Objective-C compiler. There is no impedance mismatch for Objective-C.
8. Objective-C++ is C++
Any valid C++ program will compile with an Objective-C++ compiler. Just like Objective C seamlessly mixes Smalltalk-style objects with C, Objective-C++ mixes Smalltalk style objects with C++. There is a bit of an impedance mismatch with C++ because it contains its own impoverished model of "objects" which is not really compatible with the real thing, but for the most part you can mix and match Objective-C objects with C++ objects to meet your needs.
7. Categories
You use Categories to add methods to existing classes, particularly when you don't have source code for the classes. The added methods are inherited by all subclasses of the extended class and are indistinguishable at run-time from methods originally compiled into the class.
Cocoa's NSObject base class has at least 69 categories within the Cocoa frameworks and countless others in third party code. If you add methods to the base class, you can confidently invoke those methods with any anonymous object you come across because, if nothing else, the objects will have the method implementations inherited from your category of NSObject.
Categories are preferred over subclasses in many situations. In particular, categories circumvent the fact that it isn't always possible to force an existing framework to use your subclass instead of whatever class it was compiled to use. If you use a Category to add methods to Cocoa's NSString class, those methods are available from every string returned from every method in every linked framework just as if your methods were originally compiled into NSString.
6. Introspection: -performSelector:, -respondsToSelector:, Key Value Coding
This little feature is essential for effective use of duck typing (reason 5 that Cocoa is great because of Objective-C). Any object can serve as a delegate because Cocoa classes ask each delegate if it responds to each message before sending the message. Cocoa's Responder Chain is implemented with -respondsToSelector: and -performSelector:.
Key Value Coding lets you ask any object for information about any property. The neat thing about both -performSelector: and Ket Value Coding is that they postpone decisions until runtime. They can use strings which might not be known until runtime. They add to the dynamism of Cocoa. They decouple objects from each other.
5. Duck Typing (If it walks like a duck and quacks like a duck, I would call it a duck)
http://en.wikipedia.org/wiki/Duck_typing
The idea is that the type of my objects doesn't actually matter. What's important is whether or not the object can do what I ask it to do. When you take advantage duck typing, you start perceiving static typing as the straight-jacket it is. Don't worry; you'll find any inadvertent type errors during the testing that you should be performing anyway. There's a nice rant on the subject by Bruce Eckel at http://www.mindview.net/WebLog/log-0052.
Without the dynamism of Objective-C's duck typing, you wouldn't have Cocoa's Target/Action paradigm; you wouldn't have Delegates; you wouldn't have the easy Undo/Redo. You wouldn't have heterogeneous container classes. You wouldn't have Proxies or Faults. The Responder Chain would require a much more complicated implementation, etc.
4. Sending messages to nil without catastrophe
Code is read many times more often than it's written. Which is easier to read?
- (IBAction)takeSelectedPluginFrom:(id)sender {
[pluginIconView setImage:
[NSImage imageNamed:[[sender object] iconImageName]]];
}
- or -
- (IBAction)takeSelectedPluginFrom:(id)sender
{
if(nil != pluginIconView) {
if(nil != sender) {
id selectedObject = [sender object];
if(nil != selectedObject) {
NSString *iconImageName = [selectedObject iconImageName];
if(nil != iconImageName) {
NSImage *iconImage = [NSImage imageNamed:iconImageName];
if(nil != iconImage)
{
[pluginIconView setImage:iconImage];
}
}
}
}
}
}
3. The syntax for sending messages
Code is read many times more often than it's written. Which code is easier to read? What is the purpose for each argument in the following? (Don't cheat by looking at the Objective-C version!)
editorWindow.dragImage(self.imageOfSelection(),
interestingPoint,
self.selectionOrigin(),
self.eventThatStartedDrag(),
NSPasteboard.pasteboardWithName(NSDragPboard),
self,
!self.isOperationDelete());
- or -
[editorWindow dragImage:[self imageOfSelection]
at:interestingPoint
offset:[self selectionOrigin]
event:[self eventThatStartedDrag]
pasteboard:[NSPasteboard pasteboardWithName:NSDragPboard]
source:self
slideBack:![self isOperationDelete]];
2. Memory Management Conventions
I like automatic memory garbage collection a lot. Really! There's just one problem: languages that have mandatory automatic memory garbage collection exclude themselves from critical programming domains. Who will write an operating system in a language that has mandatory automatic memory garbage collection? Who will write an automatic garbage collector in a language that has mandatory automatic memory garbage collection? I'll grant that not very many people need to write operating systems or device drivers or garbage collectors very often. The genius of Objective-C is that you can productively write all of those things and embedded cell phone applications and awesome desktop applications.
Cocoa's manual memory management conventions are a near perfect compromise. If you do it right, you never see manual memory management. It’s fast and deterministic. Best of all, you can enjoy the automatic memory garbage collector as long as all the frameworks you use work with it and can tolerate the resulting slight impedance mismatch with C.
1. Sending messages
Alan Kay, the creator of Smalltalk and the person who coined the term "object oriented," wrote, "I thought of objects being like biological cells and/or individual computers on a network, only able to communicate with messages (so messaging came at the very beginning — it took a while to see how to do messaging in a programming language efficiently enough to be useful)." (http://userpage.fu-berlin.de/~ram/pub/pub_jf47ht81Ht/doc_kay_oop_en)
Objective-C is a mix of Smalltalk and C. The best way to understand Objective-C is to start with the idea of a message. A message is a way to communicate information or make a request. For example, the boss calls out from her office, "someone get me last month's financial report." The boss doesn't care who satisfies her request or how the request is satisfied. She just wants the report. Sending messages in Objective-C works the same way. You can send any message to any object. If the receiver of the message can't satisfy the request, the receiver can pass the message on to another object or say, "Boss, I don't understand what you want."
The best definition for a Smalltalk or Objective-C "object" is "something that can respond to messages." This one basic idea enables everything else that you love about Cocoa. This is how you get duck typing, categories, sending messages to nil, and introspection. Messaging allows us to avoid COM and SOM and other add-on nonsense. Messaging is so powerful and yet so simple that it's easy to add to base languages like C and C++. Messaging also enables straightforward language bridges to dynamic languages like Python and Ruby.
It's all about messaging!