关于Block

下文所有的讨论都基于ARC

如有错误或不严谨,请多多赐教.感激不尽

1.什么是Block?

block是"带有自动变量值的匿名函数"

2.ARC下三种block都是什么?

__NSGlobalBlock__ -- 类似函数，位于text段
__NSMallocBlock__ -- 堆区
__NSStackBlock__ -- 栈区

3.三种block分别在什么情况下出现?

NSGlobalBlock(遇见情况较少)
没有引用外部变量的block,此时的block为NSGlobalBlock

NSStackBlock

情况1

assign修饰,并且引用了外部变量

@interface ViewController ()
@property (assign, nonatomic) void(^myBlock1)();
@end
@implementation ViewController

- (void)viewDidLoad {
    [super viewDidLoad];
    int a = 10;
    self.myBlock1 = ^{
        NSLog(@"%d",a);
    };
}
@end

情况2

block做方法参数

- (void)demo:(void(^)())block {   
    NSLog(@"%@",block);
}
- (void)viewDidLoad {
    [super viewDidLoad];
    [self demo:^{
        NSLog(@"%d",a);
    }];
}

一定是在调用方法传递参数时实现block才是NSStackBlock,如果事先实现好的block,如
[self demo:self.myBlock1];
这样传递来的block是和self.myBlock的block是同类型的

NSMallocBlock(最常见最常用的)

使用copy修饰并引用了外部变量,或直接声明并引用了外部变量

@property (copy, nonatomic) void(^myBlock2)();
    self.myBlock2 = ^{
        
        NSLog(@"%d",a);
    };
    或
    void(^myBlock3)() = ^{
        
        NSLog(@"%@",arr);
    };

4.__block的原理是什么?

结论: 把原本以值传递传递到block中的变量换成引用传递传递到block内部.
结论怎么来的,下面开始解释.
第一个测试代码

    int myNumA = 10;
    self.myBlock1 = ^{     
        myNumA++;
    };

这种写法是会报错的, 为什么报错?

struct __DemoClass__demo_block_impl_0 {
    struct __block_impl impl;
    struct __DemoClass__demo_block_desc_0* Desc;
    int myNumA;
    __DemoClass__demo_block_impl_0(void *fp, struct __DemoClass__demo_block_desc_0 *desc, int _myNumA, int flags=0) : myNumA(_myNumA) {
        impl.isa = &_NSConcreteStackBlock;
        impl.Flags = flags;
        impl.FuncPtr = fp;
        Desc = desc;
    }
};
static void __DemoClass__demo_block_func_0(struct __DemoClass__demo_block_impl_0 *__cself) {
    int myNumA = __cself->myNumA; // bound by copy
    
    
    NSLog((NSString *)&__NSConstantStringImpl__var_folders_97_s25cw5dj7ss2rf1q7_bchd7w0000gn_T_DemoClass_8ec7ba_mi_0,myNumA);
}

static struct __DemoClass__demo_block_desc_0 {
    size_t reserved;
    size_t Block_size;
} __DemoClass__demo_block_desc_0_DATA = { 0, sizeof(struct __DemoClass__demo_block_impl_0)};

static void _I_DemoClass_demo(DemoClass * self, SEL _cmd) {
    
    int myNumA = 10;
    
    void(*myBlock)() = ((void (*)())&__DemoClass__demo_block_impl_0((void *)__DemoClass__demo_block_func_0, &__DemoClass__demo_block_desc_0_DATA, myNumA));
}

这是编译成的C++代码, 所有myNumA都是值传递.所以,没法修改是很显然的.

    __block int myNumA = 10;
    self.myBlock1 = ^{     
        myNumA++;
    };

编译之后,首先,对myNumA进行一个处理

struct __Block_byref_myNumA_0 {
    void *__isa;
    __Block_byref_myNumA_0 *__forwarding;
    int __flags;
    int __size;
    int myNumA;
};

其中有一个__Block_byref_myNumA_0 *__forwarding;指针.
接下来

struct __DemoClass__demo_block_impl_0 {
    struct __block_impl impl;
    struct __DemoClass__demo_block_desc_0* Desc;
    __Block_byref_myNumA_0 *myNumA; // by ref
    __DemoClass__demo_block_impl_0(void *fp, struct __DemoClass__demo_block_desc_0 *desc, __Block_byref_myNumA_0 *_myNumA, int flags=0) : myNumA(_myNumA->__forwarding) {
        impl.isa = &_NSConcreteStackBlock;
        impl.Flags = flags;
        impl.FuncPtr = fp;
        Desc = desc;
    }
};
static void __DemoClass__demo_block_func_0(struct __DemoClass__demo_block_impl_0 *__cself) {
    __Block_byref_myNumA_0 *myNumA = __cself->myNumA; // bound by ref
    
    
    NSLog((NSString *)&__NSConstantStringImpl__var_folders_97_s25cw5dj7ss2rf1q7_bchd7w0000gn_T_DemoClass_29df03_mi_0,(myNumA->__forwarding->myNumA)++);
}
static void __DemoClass__demo_block_copy_0(struct __DemoClass__demo_block_impl_0*dst, struct __DemoClass__demo_block_impl_0*src) {_Block_object_assign((void*)&dst->myNumA, (void*)src->myNumA, 8/*BLOCK_FIELD_IS_BYREF*/);}

static void __DemoClass__demo_block_dispose_0(struct __DemoClass__demo_block_impl_0*src) {_Block_object_dispose((void*)src->myNumA, 8/*BLOCK_FIELD_IS_BYREF*/);}

static struct __DemoClass__demo_block_desc_0 {
    size_t reserved;
    size_t Block_size;
    void (*copy)(struct __DemoClass__demo_block_impl_0*, struct __DemoClass__demo_block_impl_0*);
    void (*dispose)(struct __DemoClass__demo_block_impl_0*);
} __DemoClass__demo_block_desc_0_DATA = { 0, sizeof(struct __DemoClass__demo_block_impl_0), __DemoClass__demo_block_copy_0, __DemoClass__demo_block_dispose_0};

static void _I_DemoClass_demo(DemoClass * self, SEL _cmd) {
    
    __attribute__((__blocks__(byref))) __Block_byref_myNumA_0 myNumA = {(void*)0,(__Block_byref_myNumA_0 *)&myNumA, 0, sizeof(__Block_byref_myNumA_0), 10};
    
    void(*myBlock)() = ((void (*)())&__DemoClass__demo_block_impl_0((void *)__DemoClass__demo_block_func_0, &__DemoClass__demo_block_desc_0_DATA, (__Block_byref_myNumA_0 *)&myNumA, 570425344));
}

所有对myNumA的引用都变成了__Block_byref_myNumA_0 *myNumA.
在最后两行,我们声明的int myNumA = 10变成了__Block_byref_myNumA_0 myNumA = {(void*)0,(__Block_byref_myNumA_0 *)&myNumA, 0, sizeof(__Block_byref_myNumA_0), 10};,传入进去的是(__Block_byref_myNumA_0 *)&myNumA,很明显,传递的是myNumA的内存地址, 所以可以改变变量的值.

5.修改什么样的变量需要使用__block,什么样的不需要?

先说结论:
需要加__block修饰的只有局部变量且非指针变量.
成员变量,属性,static修饰过得局部变量,局部指针变量都不需要加__block修饰.

解释原因

1.属性和成员变量不需要加`__block`

测试代码:

@interface DemoClass ()
@property (assign, nonatomic) int myNumA;
@end

@implementation DemoClass {
    int myNumB;
}

- (void)demo {
    self.myNumA = 10;
    myNumB = 10;
    void(^myBlock)() = ^{
        self.myNumA = self.myNumA + 1;
        myNumB = myNumB + 1;
    };
    myBlock();
}

编译后的C++代码:

1.static void __DemoClass__demo_block_func_0(struct __DemoClass__demo_block_impl_0 *__cself) {
        DemoClass *self = __cself->self; // bound by copy
       ① ((void (*)(id, SEL, int))(void *)objc_msgSend)((id)self, sel_registerName("setMyNumA:"), ((int (*)(id, SEL))(void *)objc_msgSend)((id)self, sel_registerName("myNumA")) + 1);
       ②(*(int *)((char *)self + OBJC_IVAR_$_DemoClass$myNumB)) = (*(int *)((char *)self + OBJC_IVAR_$_DemoClass$myNumB)) + 1;
}

2.static void _I_DemoClass_demo(DemoClass * self, SEL _cmd) {

    ((void (*)(id, SEL, int))(void *)objc_msgSend)((id)self, sel_registerName("setMyNumA:"), 10);
    ③ (*(int *)((char *)self + OBJC_IVAR_$_DemoClass$myNumB)) = 10;
    void(*myBlock)() = ((void (*)())&__DemoClass__demo_block_impl_0((void *)__DemoClass__demo_block_func_0, &__DemoClass__demo_block_desc_0_DATA, self, 570425344));
    ((void (*)(__block_impl *))((__block_impl *)myBlock)->FuncPtr)((__block_impl *)myBlock);
}

方法"1."是block的回调函数.由函数参数可以看出来,即使是没有参数的block,实际运行的时候也会接受一个__DemoClass__demo_block_impl_0的结构体,结构体代码如下:

struct __DemoClass__demo_block_impl_0 {
  struct __block_impl impl;
  struct __DemoClass__demo_block_desc_0* Desc;
  DemoClass *self;
  __DemoClass__demo_block_impl_0(void *fp, struct __DemoClass__demo_block_desc_0 *desc, DemoClass *_self, int flags=0) : self(_self) {
    impl.isa = &_NSConcreteStackBlock;
    impl.Flags = flags;
    impl.FuncPtr = fp;
    Desc = desc;
  }
};

从结构体中可以看出,此时的block内部可以获取到self
如①所示,之所以修改属性不需要__block是因为block中可以获取到self,而通过self和objc_msgSend就可以调用该属性的set和get方法去赋值.
成员变量呢?
如③所示,myNumB = 10被编译成(*(int *)((char *)self + OBJC_IVAR_$_DemoClass$myNumB)) = 10.
我们可以近似的看成(*(int *))myNumB = 10.可以看出,myNumB的值也是通过修改其内存地址里面的内容赋值的,而赋值也需要self.
结合②③很明显可以看出,成员变量的赋值和访问也是通过self去赋值的,只不过没有set和get方法.
2.经过static修饰过的局部变量
测试代码如下:

- (void)demo {
    static int myNumA = 10;
    void(^myBlock)() = ^{
        myNumA = myNumA + 1;
    };
    myBlock();
}

编译后的C++代码如下:

struct __DemoClass__demo_block_impl_0 {
  struct __block_impl impl;
  struct __DemoClass__demo_block_desc_0* Desc;
  int *myNumA;
  __DemoClass__demo_block_impl_0(void *fp, struct __DemoClass__demo_block_desc_0 *desc, int *_myNumA, int flags=0) : myNumA(_myNumA) {
    impl.isa = &_NSConcreteStackBlock;
    impl.Flags = flags;
    impl.FuncPtr = fp;
    Desc = desc;
  }
};
static void __DemoClass__demo_block_func_0(struct __DemoClass__demo_block_impl_0 *__cself) {
  int *myNumA = __cself->myNumA; // bound by copy

        (*myNumA) = (*myNumA) + 1;
}

此时的结构体中获取到的是myNumA的内存地址,所以不需要__block修饰.

3.局部指针变量

测试代码如下:

- (void)demo {
    int a = 10;
    int *myNumA = &a;
    NSMutableArray *arrM = [NSMutableArray arrayWithObjects:@1, @2, nil];
    void(^myBlock)() = ^{
        *myNumA = *myNumA + 1;
        [arrM addObject:@3];
    };
    myBlock();
}