Objective C and C++ verbosities compared

In my previous post I mentioned Steve Job's choice of Objective C over C++ for development, based on his assertion that he wanted "to eliminate 80% of the code you have to write for you app". To which Chris commented that in this older post, a piece of C++ code that I had converted to Objective C was actually twice as long as the original.

I probably should have gone into a bit more details in that initial remark on Objective C vs. C++, but that would have been rather off topic. Chris's comment calls for more discussion, though.

The quick reply :

the C++ code I converted uses a simple STL-based data structure. The Objective C version uses Core Data.

There's no question that, for simple operations, Core Data is more verbose than the STL equivalent. In the latter case you're iterating over a simple container, in the former you're actually querying a DB (an SQLite db, btw - yet another example of a well-reused piece of technology).

So, even though both codes do the same thing conceptually, the underlying technology is completely different. Yes, the Objc version is much longer, however... in the C++ case, there's the whole definition of the STL data structure which is not shown in the example, and that you have to write. In the case of Core Data, well, there isn't. You simply design the data model with Xcode's Core Data builder :-). So the complete number of lines of code is actually smaller in the ObjC case.

The longer reply :

Core Data didn't exist yet in at the time Jobs chose Objective C as the base language for NeXTStep, and Objective C does tend to be more verbose in its APIs than C++. The best way to demonstrate this is through a basic example of an array of ints :

C++

// declare an array of ints
std::vector arrayOfInts;

// add one element
arrayOfInts.push_back(1);

// get the element's value
int i = arrayOfInts[0];

Objective C

// no array of ints, only arrays of NSObject* so :
NSMutableArray* arrayOfInts = [NSMutableArray array];

// add one element
[arrayOfInts addObject:[NSNumber numberWithInt:1]];

int i = [[arrayOfInts objectAtIndex:0] intValue];

If that reminds you of Java, you're right. And thankfully, autoboxing is being added in Objective C (better late than never).

Edit Jan. 12th, 2013 : the above code would now be written as follows :

// no array of ints, only arrays of NSObject* so :
NSMutableArray* arrayOfInts = [NSMutableArray array];

// add one element
[arrayOfInts addObject:@1];

int i = arrayOfInts[0].intValue;

// or, even simpler :

NSArray* arrayOfInts = @[ @1 ];

int i = arrayOfInts[0].intValue;

end Edit

So why choose Objective C over C++ at the time ?

Because Objective C, as verbose are its basic APIs, is actually closer to Python than to C++. The object model is much more elaborate than in C++ (in which you have nothing else but virtual methods) : classes are first class objects, an object can be asked if it handles a method or not, you can add methods to a class at runtime and without having to derive it, and the language's position toward type safety is much more relaxed - it's essentially duck typing. You don't have to declare a method in the interface to implement it, which is very convenient for internal methods. You do not have to declare a method that you override either, again simply implementing your classes's own version is enough.

@interface MyClass : NSObject

// nothing

@end

@implementation MyClass

// this overrides NSObject:init
- (id)init
{
// some code
[self moreInit]; // moreInit not declared in interface
}

- (void)moreInit
{
// some more code
}

@end

In C++ that would be :

class MyClass : public Object
{
virtual void init();
virtual void moreInit();
}

void MyClass::init()
{
// some code
moreInit();
}

void MyClass::moreInit()
{
// some more code
}

All this makes of Objective C almost a scripted language in disguise. While in C++ you will spend a lot of time getting your types right, and any non-trivial refactoring will take a whole lot of time, Objective C lets you code much more freely.