Hands-On Objective-C 2.0: Blocks
Blocks provide a way to package up some executable code and a context (various variables) as a single entity so they can be handed off for execution at a later time or on a different thread. In other languages, blocks or similar constructs are sometimes called closures or anonymous functions. Blocks are an Apple-supplied extension to C, Objective-C 2.0, and C++. Apple has submitted blocks to the C standards working group as a proposed extension to C. At the time of this writing, blocks are only available on Mac OS X Snow Leopard (v 10.6 and on iOS 4). They are not available on earlier versions of Mac OS X or iPhone iOS.
Handing off a package of work is useful in many situations, but one of the main driving forces behind the adoption of blocks is Apple's new Grand Central Dispatch (GCD) feature. GCD is designed to make concurrency easier to program and more efficient to execute. Essentially, GCD is a thread pool that is managed for you by the operating system. The idea behind GCD is that the operating system has a global view of all the processes running on your Mac, and allocates resources (CPU, GPU, and RAM) as needed to make things run more efficiently. GCD can make better decisions than a user space program can about the number of threads to use and when to schedule them for execution. You use blocks to submit units of work for GCD to execute.
This chapter is an introduction to blocks. You will learn how to define a block, how a block has access to variables in its surrounding context, how to use a block in your own code, and about the somewhat tricky topic of memory management for blocks. The chapter also explores some pitfalls that can befall an unwary user of blocks.
Before looking at blocks in detail, the chapter takes a pair of detours and looks at two earlier ways of packaging up functionality: function pointers and the Foundation class NSInvocation.
Function Pointers
When the compiler encounters a function call, it inserts a jump instruction to the code that performs the function. (A jump instruction causes the program execution to jump to the specified code instead of executing the line of code directly after the jump instruction.) To return, the function executes a jump instruction back to the line of code following the original function call. In a normal function call, the landing point of the jump instruction (and hence the function that is called) is static. It is determined at compile time. But a function call can be made dynamic through the use of a function pointer.
The following line declares myFunctionPtr as a pointer to a function that takes two ints as arguments and returns an int:
int (*myFunctionPtr) (int, int);
Figure 16.1 shows the anatomy of a function pointer.
Figure 16.1 The anatomy of a function pointer
The general form of a function pointer is:
return_type (*name)(list of argument types);
Function pointers are a low point in C syntax. Instead of reading left-to-right or right-to-left, they read from the inside out. More complicated function pointer declarations can quickly turn into puzzles, as you will see in Exercise 16.1.
You can also declare arrays of function pointers. The following line declares fpArray as an array of 10 pointers to functions. Each function takes a single argument, a pointer to a float, and returns void:
void (*fpArray[10])(float*);
A function pointer can point to a function that has another function pointer as an argument or a return value:
void (*(*complicatedFunctionPointer)(void))(void);
complicatedFunctionPointer is a pointer to a function that takes no arguments and returns a pointer to a function that takes no arguments and returns void.
Declarations like the preceding one are ugly, but you can make your code cleaner by hiding the ugliness with a typedef:
typedef void (*(*complicatedFunctionPointer)(void))(void); complicatedFunctionPointer fp;
Calling a Function with a Function Pointer
The following example shows how to assign a function to a function pointer and how to call the function using the function pointer:
void logInt( int n ) { NSLog("The integer is: %d", n); } void (*myFunctionPtr)(int); // Declare a function pointer myFunctionPtr = logInt; // Make it point to logInt myFunctionPtr( 5 ); // Execute the function through the pointer
To make the function pointer refer to a function, you simply assign it the name of the function. The function must be defined or visible by a forward declaration at the point it is assigned.
To call a function through a function pointer, you simply add the arguments, encased in parentheses, to the function pointer. A function call through a function pointer is just like a normal function call except that you use the name of the function pointer variable instead of the function name, as shown in the previous code snippet.
Using Function Pointers
One of the primary uses of function pointers is for callbacks. Callbacks are used in situations where you have a function or method that is going to do some work for you, but you would like the opportunity to insert your own code somewhere in the process. To do this, you pass the working function or method a pointer to a function containing the code you want executed. At the appropriate time, the working function or method will call your function for you.
For example, NSMutableArray provides the following method for use in custom sorting:
- (void)sortUsingFunction: (NSInteger (*)(id, id, void *))compare context:(void *)context
When you invoke sortUsingFunction:context:, the method sorts the contents of the receiver. To perform the sort, sortUsingFunction:context: must examine pairs of array elements and decide how they are ordered. To make these decisions, sortUsingFunction:context: calls the compare function that you passed in by pointer when the method was invoked.
The compare function must look at the two objects it receives and decide (based on whatever criterion you require) whether they are ordered NSOrderedAscending, NSOrderedSame, or NSOrderedDescending.
sortUsingFunction:context: also passes compare the void* pointer that it received as its context argument. This is a pure pass-through; sortUsingFunction:context: doesn't look at or modify context. context may be NULL if compare doesn't require any additional information.
Listing 16.1 sorts an array containing some NSNumber objects. The address of a BOOL is passed in to control the direction of a numerical sort.
Example 16.1. ArraySortWithFunctionPointer.m
#import <Foundation/Foundation.h> NSInteger numericalSortFn( id obj1, id obj2, void* ascendingFlag ) { int value1 = [obj1 intValue]; int value2 = [obj2 intValue]; if ( value1 == value2 ) return NSOrderedSame; if ( *(BOOL*) ascendingFlag ) { return ( value1 < value2 ) ? NSOrderedAscending : NSOrderedDescending; } else { return ( value1 < value2 ) ? NSOrderedDescending : NSOrderedAscending; } } int main (int argc, const char * argv[]) { NSAutoreleasePool * pool = [[NSAutoreleasePool alloc] init]; // Put some number NSNumber objects in an array NSMutableArray *numberArray = [[NSMutableArray alloc] initWithCapacity: 5]; [numberArray addObject: [NSNumber numberWithInt: 77]]; [numberArray addObject: [NSNumber numberWithInt: 59]]; [numberArray addObject: [NSNumber numberWithInt: 86]]; [numberArray addObject: [NSNumber numberWithInt: 68]]; [numberArray addObject: [NSNumber numberWithInt: 51]]; NSLog( @"Before sort: %@", [numberArray description] ); // This flag controls the sort direction. // Change it to NO to sort in descending order. BOOL ascending = YES; // Sort the array [numberArray sortUsingFunction: numericalSortFn context: &ascending]; NSLog( @"After sort: %@", [numberArray description] ); [numberArray release]; [pool drain]; return 0; }
Notice:
- ascendingFlag is passed in as void*. It must be cast as BOOL* before it can be dereferenced to get the BOOL value.
- The name of a function, in this case numericalSortFn, can serve as a properly typed pointer to that function. Here, it is used as the argument when invoking sortUsingFunction:context: without defining a separate function pointer variable.