GCD的一些使用
2018-11-27 本文已影响7人
wwwwwwdi
本片文字意在记录gcd的一些用法
1. dispatch_semaphore
相关方法：
dispatch_semaphore_create() //创建信号量，设置能同时执行的最大线程数
dispatch_semaphore_wait() //信号量-1
dispatch_semaphore_signal() //信号量+1
dispatch_semaphore_wait方法：
Decrement the counting semaphore. If the resulting value is less than zero, this function waits for a signal to occur before returning.
信号量减1，如果结果小于0，会阻塞当前线程等待新信号的出现。
如果为DISPATCH_TIME_FOREVER，则会一直等待
dispatch_semaphore_signal方法：
Increment the counting semaphore. If the previous value was less than zero, this function wakes a thread currently waiting in dispatch_semaphore_wait
信号量加1，如果之前的信号量小于0，则会唤醒dispatch_semaphore_wait方法中等待着的线程
使用的方法：先降后升
先创建信号量，设置能同时执行的最大线程数。如果为1，则同时可执行的线程就是1个，等进入到耗时操作1的时候，先让信号量降低，然后处理完成之后，再调用signal方法使信号量升高，在别的耗时操作中，也是一样的逻辑。
那么，在耗时操作2开始的时候，由于耗时操作1中信号量为0，且等待时间为DISPATCH_TIME_FOREVER，此时则会一直等待。
常规操作如下：
/**
* 常规dispatch_semaphore使用方法（多个网络请求，限制最大并发数量）
*/
- (void)dispatchSemaphoreTest1 {
dispatch_queue_t queue = dispatch_queue_create(0, 0);
dispatch_semaphore_t sema = dispatch_semaphore_create(2);
NSLog(@"--------------------------dispatch begin-----------------");
dispatch_async(queue, ^{
dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(4 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
NSString *imgUrl = @"https://www.baidu.com/img/bd_logo1.png";
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success1-----------------");
dispatch_semaphore_signal(sema);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail1-----------------");
dispatch_semaphore_signal(sema);
}];
});
NSLog(@"--------------------------dispatch async complete1-----------------");
});
dispatch_async(queue, ^{
//Thread4: EXC_BAD_INSTRUCTION(code=EXC_I386_INVOP,subcode=0x0)
//https://stackoverflow.com/questions/24337791/exc-bad-instruction-code-exc-i386-invop-subcode-0x0-on-dispatch-semaphore-dis
dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(8 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
NSString *imgUrl = @"https://www.baidu.com/img/bd_logo1.png";
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success2-----------------");
dispatch_semaphore_signal(sema);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail2-----------------");
dispatch_semaphore_signal(sema);
}];
});
NSLog(@"--------------------------dispatch async complete2-----------------");
});
dispatch_async(queue, ^{
dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
NSString *imgUrl = @"https://www.baidu.com/img/bd_logo1.png";
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success3-----------------");
dispatch_semaphore_signal(sema);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail3-----------------");
dispatch_semaphore_signal(sema);
}];
NSLog(@"--------------------------dispatch async complete3-----------------");
});
NSLog(@"--------------------------dispatch complete-----------------");
}
运行之后，竟然崩溃了！！
报错信息：
EXC_BAD_INSTRUCTION(code=EXC_I386_INVOP,subcode=0x0)
查询了stackoverflow之后，发现此篇文章
此问题是使用了dispatch_group_enter,dispatch_group_leave和dispatch_group_wait三个方法的途中，遇到的这个问题，他的写法如下：
dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_HIGH, 0);
dispatch_group_t group = dispatch_group_create();
for (...) {
if (...) {
dispatch_group_enter(group);
dispatch_async(queue, ^{
[self webserviceCall:url onCompletion:^{
dispatch_group_leave(group);
}];
});
}
}
dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^{
dispatch_group_wait(group, dispatch_time(DISPATCH_TIME_NOW, (int64_t)(2.0 * NSEC_PER_SEC)));
dispatch_sync(queue, ^{
// call completion handler passed in by caller
});
});
其中回答说：
是在使用的过程中，调用已经释放了的group才导致的这个问题。
受此文章启发，我考虑到上面的写法是否也是由于这个原因呢？
然后我把dispatch_semaphore_t作为当前类的属性，强引用一下，发现解决了这个问题。
//只需要在前面创建semaphore和queue的时候，强引用一下，然后再后续的调用中，都修改为self.xxx即可
self.sema = dispatch_semaphore_create(2);
self.queue = dispatch_queue_create(0, 0);
2018年11月28日
今天我又试了一下：发现竟然又不崩溃了？？？？
而且也无法复现那个问题了。。。excuse me ???
异步变同步
下面的实现是一种取巧的方法，达到了同步的效果，
- (void)testDispatchSemaphore {
NSLog(@"--------------------------begin-----------------");
dispatch_semaphore_t sema = dispatch_semaphore_create(0);
dispatch_queue_t que = dispatch_get_global_queue(0, 0);
dispatch_async(que, ^{
NSLog(@"--------------------------async begin-----------------");
[NSThread sleepForTimeInterval:3];
dispatch_semaphore_signal(sema);
NSLog(@"--------------------------async complete-----------------");
});
dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
NSLog(@"--------------------------complete-----------------");
}
//log如下：
--------------------------begin-----------------
--------------------------async begin-----------------
--------------------------async complete-----------------
--------------------------complete-----------------
这里有一个问题点：创建semaphore的时候的初始值问题
我们都知道，此semaphore值，就是用来限制最大并行线程的，也就是说如果传1，则运行的时候，同一时间只能处理一个线程。如果传4，则同时处理4个。那有没有考虑过，如果传0呢？
这里给出了苹果文档上的解释
Passing zero for the value is useful for when two threads need to reconcile the completion of a particular event.
Passing a value greater than zero is useful for managing a finite pool of resources, where the pool size is equal to the value.
大意好像是：
如果传0进去，则会使两个线程同时处理完成回调。（当两个线程需要使某个事件完成一致）
如果传大于0的值进去，会使队列资源池变为有限的，从而限制最大并发数
关于队列资源池：
详见Objective-C高级编程 iOS与OS X多线程和内存管理第三章
2. 多个异步操作结束之后，统一操作(dispatch_group)
用到的方法：
dispatch_group_enter()
dispatch_group_leave()
dispatch_group_notify()
注意：这里enter和leave要成对出现
- (void)dispatchGroupTest {
dispatch_group_t group = dispatch_group_create();
dispatch_queue_t queue = dispatch_queue_create("com.fxeyeQueue.www", DISPATCH_QUEUE_CONCURRENT);
dispatch_group_enter(group);
dispatch_group_enter(group);
dispatch_group_async(group, queue, ^{
dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(3 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
NSString *imgUrl = @"https://www.baidu.com/img/bd_logo1.png";
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success1-----------------");
dispatch_group_leave(group);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail1-----------------");
dispatch_group_leave(group);
}];
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success2-----------------");
dispatch_group_leave(group);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail2-----------------");
dispatch_group_leave(group);
}];
});
});
dispatch_group_notify(group, queue, ^{
NSLog(@"--------------------------notify-----------------");
});
NSLog(@"--------------------------complete-----------------");
}
//log如下：
14:08:58.014812+0800 testOperation[6098:474448] --------------------------complete-----------------
14:09:01.035264+0800 testOperation[6098:474448] --------------------------fail2-----------------
14:09:01.035430+0800 testOperation[6098:474448] --------------------------fail1-----------------
14:09:01.035564+0800 testOperation[6098:474778] --------------------------notify-----------------
也有另外一种写法：
把两个请求分别放到两个异步group操作中，即：
- (void)dispatchGroupTest {
dispatch_group_t group = dispatch_group_create();
dispatch_queue_t queue = dispatch_queue_create("com.fxeyeQueue.www", DISPATCH_QUEUE_CONCURRENT);
dispatch_group_async(group, queue, ^{
NSString *imgUrl = @"https://www.baidu.com/img/bd_logo1.png";
dispatch_group_enter(group);
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success1-----------------");
dispatch_group_leave(group);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail1-----------------");
dispatch_group_leave(group);
}];
});
dispatch_group_async(group, queue, ^{
NSString *imgUrl = @"https://www.baidu.com/img/bd_logo1.png";
dispatch_group_enter(group);
[[AFHTTPSessionManager manager] GET:imgUrl parameters:nil progress:^(NSProgress * _Nonnull downloadProgress) {
} success:^(NSURLSessionDataTask * _Nonnull task, id _Nullable responseObject) {
NSLog(@"--------------------------success2-----------------");
dispatch_group_leave(group);
} failure:^(NSURLSessionDataTask * _Nullable task, NSError * _Nonnull error) {
NSLog(@"--------------------------fail2-----------------");
dispatch_group_leave(group);
}];
});
dispatch_group_notify(group, queue, ^{
NSLog(@"--------------------------notify-----------------");
});
NSLog(@"--------------------------complete-----------------");
}
从log上看，也是可以达到效果的，
但是！此种写法在时机使用中，遇到了一个问题，会造成整个APP卡死。。
然后Bugly报错说max operation count reached，思来想去也没找到问题。还请知道解决方法的大佬给予明示！thx~
今天又接着Google，终于找到了问题所在。受此文启发。
项目中代码是在WKWebview的didFinishNavigation方法中调用请求方法。
然后
/**
* 两个请求结束之后，统一操作
*/
- (void)groupRequest {
dispatch_group_t group = dispatch_group_create();
self.group = group;
dispatch_queue_t queue = dispatch_get_global_queue(0, 0);
dispatch_group_enter(self.group);
dispatch_group_enter(self.group);
dispatch_group_async(self.group, queue, ^{
[某网络请求方法 {
dispatch_group_leave(self.group);
}];
});
dispatch_group_async(self.group, queue, ^{
[某网络请求方法 {
dispatch_group_leave(self.group);
}];
});
dispatch_group_notify(self.group, queue, ^{
//获取到之后，刷新页面
[self reloadViewWithModelData];
});
}
由于webview的加载过程不可控，所以可能存在多次调用此方法的情况，然后就会创建很多的group，造成Bugly报出的max operation count reached问题，因此懒加载group理论上就能解决此问题。上面代码改为：
- (void)groupRequest {
if (self.group == nil) {
dispatch_group_t group = dispatch_group_create();
self.group = group;
}
dispatch_queue_t queue = dispatch_get_global_queue(0, 0);
......
}
3. dispatch_barrier
其实添加依赖之类的操作，除了使用上面的两个， 还可以使用dispatch_barrier
notes:
但是有一点要先说明：
dispatch_barrier必须用在并发队列中,而且不能是获取的系统的全局队列，得是通过dispatch_queue_create自己创建的并发队列才可以
/**
* dispatch_barrier,只在并发队列中才有用，而且不能是获取的系统的全局队列，得是通过dispatch_queue_create自己创建的并发队列才可以
*/
- (void)dispatchBarrierFunction {
dispatch_queue_t queue = dispatch_queue_create("com.fxxxxxx.www", DISPATCH_QUEUE_CONCURRENT);
dispatch_async(queue, ^{
NSLog(@"task1 -%@",[NSThread currentThread]);
});
dispatch_async(queue, ^{
NSLog(@"task2 -%@",[NSThread currentThread]);
});
dispatch_async(queue, ^{
NSLog(@"task3 -%@",[NSThread currentThread]);
});
dispatch_barrier_async(queue, ^{
[NSThread sleepForTimeInterval:5];
NSLog(@"barrier");
});
NSLog(@"asynchronous");
dispatch_async(queue, ^{
NSLog(@"task4 -%@",[NSThread currentThread]);
});
dispatch_async(queue, ^{
NSLog(@"task5 -%@",[NSThread currentThread]);
});
dispatch_async(queue, ^{
NSLog(@"task6 -%@",[NSThread currentThread]);
});
}
log如下：
2018-11-28 17:41:56.986682+0800 asynchronous
2018-11-28 17:41:56.986849+0800 task1 -<NSThread: 0x600002ef63c0>{number = 7, name = (null)}
2018-11-28 17:41:56.986849+0800 task2 -<NSThread: 0x600002efd7c0>{number = 8, name = (null)}
2018-11-28 17:41:56.986885+0800 task3 -<NSThread: 0x600002e20e00>{number = 3, name = (null)}
2018-11-28 17:42:01.991930+0800 barrier
2018-11-28 17:42:01.992341+0800 task6 -<NSThread: 0x600002ef63c0>{number = 7, name = (null)}
2018-11-28 17:42:01.992397+0800 task5 -<NSThread: 0x600002e20e00>{number = 3, name = (null)}
2018-11-28 17:42:01.992341+0800 task4 -<NSThread: 0x600002efd7c0>{number = 8, name = (null)}
可以看到，barrier方法可以在中间设置一个阻拦，然后等上面的处理完成之后，再进行下面的操作！
但是！！！！如果操作中有网络请求等耗时操作呢？
/**
* dispatch_barrier,只在并发队列中才有用，而且不能是获取的系统的全局队列，得是通过dispatch_queue_create自己创建的并发队列才可以
* 可以看到，带有网络请求异步回调的情况，并不适用，可以使用dispatch_group_enter/dispatch_group_leave或者dispatch_semaphore_wait来实现
*/
- (void)dispatchBarrierTest {
dispatch_queue_t queue = dispatch_queue_create("com.fxxxxxx.wwww", DISPATCH_QUEUE_CONCURRENT);
dispatch_async(queue, ^{
[self afnRequestWithTag:1 delay:2 complete:^{
}];
});
dispatch_async(queue, ^{
[self afnRequestWithTag:2 delay:2 complete:^{
}];
});
dispatch_async(queue, ^{
[self afnRequestWithTag:3 delay:2 complete:^{
}];
});
dispatch_barrier_async(queue, ^{
NSLog(@"barrier");
});
dispatch_async(queue, ^{
[self afnRequestWithTag:4 delay:2 complete:^{
}];
});
dispatch_async(queue, ^{
[self afnRequestWithTag:5 delay:2 complete:^{
}];
});
NSLog(@"--------------------------barrier complete-----------------");
}
log:
2018-11-28 17:44:43.410024+0800 --------------------------barrier complete-----------------
2018-11-28 17:44:43.410079+0800 barrier
2018-11-28 17:44:45.425016+0800 --------------------------fail3-----------------
2018-11-28 17:44:45.425200+0800 --------------------------fail2-----------------
2018-11-28 17:44:45.425346+0800 --------------------------fail1-----------------
2018-11-28 17:44:45.425454+0800 --------------------------fail4-----------------
2018-11-28 17:44:45.425556+0800 --------------------------fail5-----------------
可以看出来，并不管用了。。
所以，如果要在多个操作中请求接口，还是要用上面说的dispatch_group_enter/dispatch_group_leave或者dispatch_semaphore_wait来实现
