I l@ve RuBoard Previous Section Next Section

1.15 Optimizing Method Calls

Objective-C's message-passing implementation of method calls is simple and flexible. What it sacrifices is the speed of a C++-style method call, or of a direct function call.

If you are writing time-critical code, you may want to relinquish the dynamic nature of the Objective-C method call for some extra speed. For example, if your code calls the same method on an object many times in a loop, you may not want to send a dynamic message each time you invoke the method. Objective-C provides a way for you to get a pointer to the function implementing the method, and then call the method via the pointer, bypassing the Objective-C method dispatch process.

If you only call a method once, you should use a standard Objective-C method call. This optimization's gain in efficiency is directly related to the number of times you invoke the method.

For example, suppose you want to send the following message to invoke a method that takes an integer and has no return value: 

[obj methodName:anInt];

You can replace this ordinary Objective-C method call with the following code: 

1 SEL  sel  = @selector (methodName :);
2 typedef void  (*MpType ) (id , SEL , int );
3 MpType  mptr  = (MpType ) [obj  methodFor:sel ];
4 ...
5 mptr (obj , sel , anInt );

Line 1. Get the selector for the method you want to call. You will use this both to acquire the method pointer and to use it.

Line 2. Define the type of your method pointer. This makes line 3 easier to read. Recall that methods are just C functions with two hidden parameters: the receiver and the selector. Here, the type of mptr takes these into account, as well as the return (void) and parameter (int) types of the method.

Line 3. Get the function pointer. If you already have the receiver at hand (as in this example) you can use -methodFor: (for descendants of Object) or -methodForSelector: (for descendants of NSObject). If you don't have the receiver or an object of the same type, you can call the class methods +instanceMethodFor: (for subclasses of Object) or +instanceMethodForSelector: (for subclasses of NSObject) on the class of the object whose method you will call.

Since these methods are declared to return an IMP, which is a pointer to a function whose signature is (id, SEL, ...) you need to cast to your own type the value it returns, to prevent compiler warnings at line 5.

Line 5. Use the function pointer as in C. Pass in the receiver and selector as the first two arguments. After you execute lines 1-3, line 5 will have the same effect as calling [obj methodName:anInt], but will execute faster. Typically this line will be in a loop.

Optimizing a method call in this manner only makes sense when you are going to invoke the method repeatedly, such as within a loop. Lines 1 through 3 represent a one-time setup that should be done prior to entering the loop. Line 5 is the optimized method call that would be used inside the loop.

    I l@ve RuBoard Previous Section Next Section