iOS 多线程 GCD part3:API

作者: 破弓 | 来源:发表于2017-03-05 22:54 被阅读90次

0. 预备知识

GCD对时间的描述有些新奇

#define NSEC_PER_SEC 1000000000ull
#define NSEC_PER_MSEC 1000000ull
#define USEC_PER_SEC 1000000ull
#define NSEC_PER_USEC 1000ull

MSEC：毫秒
USEC：微秒
NSEC：纳秒
SEC：秒

NSEC_PER_SEC，每秒有多少纳秒
USEC_PER_SEC，每秒有多少毫秒

#define NSEC_PER_SEC 1000000000ull //GCD最常用的时间描述

打现在开始往后5秒

double delayInSeconds = 5.0;
dispatch_time_t popTime = dispatch_time(DISPATCH_TIME_NOW, (int64_t) (delayInSeconds * NSEC_PER_SEC));1->开始时间,2->延后时间

1. dispatch_group

1.1 dispatch_group_notify

当多个任务在并行队列内无序的进行,需要在多个任务全部完成后立刻开启新任务,那么这时就是需要用到
dispatch_group_async与dispatch_group_notify的组合

    dispatch_queue_t coucurrent_queue = dispatch_queue_create("com.pogong.www", DISPATCH_QUEUE_CONCURRENT);
    
    dispatch_group_t group = dispatch_group_create();
    
    dispatch_group_async(group, coucurrent_queue, ^{
        for (int i = 0; i<10; i++) {
            NSLog(@"group-01 - %@", [NSThread currentThread]);
        }
    });
    
    dispatch_group_async(group, coucurrent_queue, ^{
        for (int i = 0; i<10; i++) {
            NSLog(@"group-02 - %@", [NSThread currentThread]);
        }
    });
    
    dispatch_group_notify(group, coucurrent_queue,^{
        NSLog(@"op finish,start new op");
    });

1.2 dispatch_group_wait

还是1.1同样的需求:当任务在并行队列内无序的进行,需要在多个任务结束后立刻开启新任务
也可以用dispatch_group_async与dispatch_group_wait的组合来完成

    dispatch_queue_t coucurrent_queue = dispatch_queue_create("com.pogong.www", DISPATCH_QUEUE_CONCURRENT);
    
    dispatch_group_t group = dispatch_group_create();
    
    dispatch_group_async(group, coucurrent_queue, ^{
        for (int i = 0; i<10; i++) {
            NSLog(@"group-01 - %@", [NSThread currentThread]);
        }
    });
    
    dispatch_group_async(group, coucurrent_queue, ^{
        for (int i = 0; i<10; i++) {
            NSLog(@"group-02 - %@", [NSThread currentThread]);
        }
    });
    
    dispatch_group_wait(group, DISPATCH_TIME_FOREVER);
    NSLog(@"op finish");

只是需要注意的是dispatch_group_wait会阻塞当前线程,或者说GCD的所有带_wait都会阻塞当前线程
dispatch_group_wait的第二参数为超时时间,DISPATCH_TIME_FOREVER代表一直等下去.

也可以将以上dispatch_group_wait代码换成:

int delayInSeconds = 20;
dispatch_time_t cheak_Time = dispatch_time(DISPATCH_TIME_NOW, (int64_t) (delayInSeconds * NSEC_PER_SEC));
long res = dispatch_group_wait(group, cheak_Time);
if (res == 0) {
NSLog(@"zc done");
}else{
NSLog(@"zc ing");
}

设置超时时间为20s,
1.group任务提前完成,提前返回
2.按指定时间返回

dispatch_group_wait的返回值为0代表group的任务全部完成,否则就代表任务还在进行中

1.3 dispatch_async+dispatch_group_enter+dispatch_group_leave

    dispatch_group_t group = dispatch_group_create();
    NSMutableArray * array = [NSMutableArray array];
    for (int i=0; i<5000; i++) {
        dispatch_group_enter(group);//enter和leave成对出现
        dispatch_async(dispatch_get_global_queue(0, 0), ^{
            
            NSLog(@"add in %@", [NSThread currentThread]);
            
            [array addObject:[NSNumber numberWithInt:i]];
            dispatch_group_leave(group);//enter和leave成对出现
            
        });
    }
    NSLog(@"before wait %@", [NSThread currentThread]);
    dispatch_group_wait(group, DISPATCH_TIME_FOREVER);

还是1.1同样的需求:当任务在并行队列内无序的进行,需要在多个任务结束后立刻开启新任务
也可以用dispatch_async+dispatch_group_enter+dispatch_group_leaver的组合来完成

2.dispatch_after

double delayInSeconds = 5.0;
dispatch_time_t popTime = dispatch_time(DISPATCH_TIME_NOW, (int64_t) (delayInSeconds * NSEC_PER_SEC));
dispatch_after(popTime, dispatch_get_main_queue(), ^{
    NSLog(@"zc after");
});

dispatch_after的用法如上方代码,就不多说了

3.信号量dispatch_semaphore

首先介绍一下信号量(semaphore)的概念。信号量是持有计数的信号，不过这么解释等于没解释。我们举个生活中的例子来看看。
假设有一个房子，它对应进程的概念，房子里的人就对应着线程。一个进程可以包括多个线程。这个房子(进程)有很多资源，比如花园、客厅等，是所有人(线程)共享的。
但是有些地方，比如卧室，最多只有两个人能进去睡觉。怎么办呢，在卧室门口挂上两把钥匙。进去的人(线程)拿着钥匙进去，没有钥匙就不能进去，出来的时候把钥匙放回门口。
这时候，门口的钥匙数量就称为信号量(Semaphore)。很明显，信号量为0时需要等待，信号量不为零时，减去1而且不等待。
此段描述完全摘自:bestswifer iOS多线程编程——GCD与NSOperation总结

3.1 多线程资源单一占用

    dispatch_queue_t q = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_HIGH, 0);
    dispatch_semaphore_t sem = dispatch_semaphore_create(1);
    
    NSMutableArray * muArr = [NSMutableArray array];
    
    for (int i = 0; i< 1000; i++) {
        dispatch_async(q, ^{
            
            dispatch_semaphore_wait(sem, DISPATCH_TIME_FOREVER);
            
            [muArr addObject:@(i)];
            NSLog(@"%d",i);
            
            dispatch_semaphore_signal(sem);
            
        });
    }

dispatch_semaphore_create,创建一个信号量为1的信号
dispatch_semaphore_wait,只有信号量大于0才能通过dispatch_semaphore_wait,通过的同时信号量减1
dispatch_semaphore_signal,信号量加1

3.2 超时显示

- (void)downloadFile
{
    _semaphore = dispatch_semaphore_create(0);
    NSURLRequest *request = [NSURLRequest requestWithURL:[NSURL URLWithString:@"http://baobab.wdjcdn.com/14525705791193.mp4"] cachePolicy:1 timeoutInterval:30];
    [[self.session downloadTaskWithRequest:request] resume];
    
    double delayInSeconds = 5.0;
    dispatch_time_t popTime = dispatch_time(DISPATCH_TIME_NOW, (int64_t) (delayInSeconds * NSEC_PER_SEC));
    
    NSLog(@"zc wait");
    long res = dispatch_semaphore_wait(_semaphore, popTime);
    if (res) {
        NSLog(@"zc timed out");
    }else{
        NSLog(@"zc timed in");
    }
}

- (void)URLSession:(NSURLSession *)session downloadTask:(NSURLSessionDownloadTask *)downloadTask didFinishDownloadingToURL:(NSURL *)location
{
    dispatch_semaphore_signal(_semaphore);
    NSLog(@"zc done");
}

还是那句所有GCD带_wait函数都会阻塞当前线程

开始下载创建信号,
下载成功,增加信号量,

任务提前完成,提前返回
按指定时间返回
dispatch_semaphore_wait的返回值为0代表任务完成,否则就代表任务还在进行中

3.3 dispatch_semaphore_wait超时时间的理解

overTime不是调用dispatch_semaphore_wait后等待的时间,而是信号量创建后的时间

-(void)overTimeTest{
    dispatch_semaphore_t signal = dispatch_semaphore_create(0);
    double delayInSeconds = 5.0;
    dispatch_time_t overTime = dispatch_time(DISPATCH_TIME_NOW, (int64_t) (delayInSeconds * NSEC_PER_SEC));
    
    dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^{
        NSLog(@"1 start wait");
        dispatch_semaphore_wait(signal, overTime);
        NSLog(@"需要线程同步的操作1 开始");
        dispatch_semaphore_signal(signal);
    });
    
    dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^{
        sleep(3);
        NSLog(@"2 start wait");
        dispatch_semaphore_wait(signal, overTime);
        NSLog(@"需要线程同步的操作2");
        dispatch_semaphore_signal(signal);
    });
}

以上两个block内的dispatch_semaphore_wait调用相差3秒,但执行dispatch_semaphore_wait却是同时的.

4.dispatch_barrier_async

多个线程对内存中的数组或是字典进行读的操作是OK,
但如果多个线程对内存中的数组或是字典进行读+写的操作,就会有问题

    NSMutableArray * muArr = [NSMutableArray array];
    [muArr addObject:@"1"];
    [muArr addObject:@"2"];
    [muArr addObject:@"3"];
    dispatch_queue_t con_queue = dispatch_queue_create("com.pogong.www", DISPATCH_QUEUE_CONCURRENT);
    
    dispatch_async(con_queue, ^{
        NSLog(@"%@",muArr);
    });
    
    dispatch_async(con_queue, ^{
        NSLog(@"%@",muArr);
    });
    
    dispatch_barrier_async(con_queue, ^{
        NSLog(@"add add add");
        [muArr addObject:@"4"];
    });
    
    dispatch_async(con_queue, ^{
        NSLog(@"%@",muArr);
    });
    
    dispatch_async(con_queue, ^{
        NSLog(@"%@",muArr);
    });

dispatch_barrier_async很好解决了多线程读写安全进行的问题,
在原有的多线程执行时,加入一个dispatch_barrier_async,
会像有栅栏一样挡住多线程执行,而先执行dispatch_barrier_async,
在执行完dispatch_barrier_async之后,再继续多线程的操作

5. dispatch_source

Dispatch Source用于监听系统的底层对象，比如文件描述符，Mach端口，信号量等。主要处理的事件如下表

DISPATCH_SOURCE_TYPE_DATA_ADD   数据增加
DISPATCH_SOURCE_TYPE_DATA_OR    数据OR
DISPATCH_SOURCE_TYPE_MACH_SEND  Mach端口发送
DISPATCH_SOURCE_TYPE_MACH_RECV  Mach端口接收
DISPATCH_SOURCE_TYPE_MEMORYPRESSURE 内存情况
DISPATCH_SOURCE_TYPE_PROC   进程事件
DISPATCH_SOURCE_TYPE_READ   读数据
DISPATCH_SOURCE_TYPE_SIGNAL 信号
DISPATCH_SOURCE_TYPE_TIMER  定时器
DISPATCH_SOURCE_TYPE_VNODE  文件系统变化
DISPATCH_SOURCE_TYPE_WRITE  文件写入

方法
dispatch_source_create：创建dispatch source，创建后会处于挂起状态进行事件接收，需要设置事件处理handler进行事件处理。
dispatch_source_set_event_handler：设置事件处理handler
dispatch_source_set_cancel_handler：事件取消handler，就是在dispatch source释放前做些清理的事。
dispatch_source_cancel：关闭dispatch source，这样后续触发的事件时不去调用对应的事件处理handler，但已经在执行的handler不会被取消.

5.1 dispatch_source_set_timer

UITableView在被拖拽时NSTimer就不起作用了
必须加[[NSRunLoop currentRunLoop] addTimer:anyTimer forMode:NSRunLoopCommonModes];
而dispatch source timer与runloop是没有关系的,所以可以放心使用

dispatch_source_t source = dispatch_source_create(DISPATCH_SOURCE_TYPE_TIMER,0, 0, DISPATCH_TARGET_QUEUE_DEFAULT);
 dispatch_source_set_event_handler(source, ^(){
  NSLog(@"sourceTimer Time log");
 });
 dispatch_source_set_timer(source, DISPATCH_TIME_NOW,5*NSEC_PER_SEC,1*NSEC_PER_SEC);//1->源,2->开始时间,3->间隔时间,4->误差秒数
 _source = source;
 dispatch_resume(_source);

5.2 文件监听

监视文件夹内文件变化

NSURL * directoryURL = [NSURL URLWithString:_path]; // assume this is set to a directory
 int const fd = open([[directoryURL path] fileSystemRepresentation], O_EVTONLY);
 if (fd < 0) {
  char buffer[80];
  strerror_r(errno, buffer, sizeof(buffer));
  NSLog(@"Unable to open \"%@\": %s (%d)", [directoryURL path], buffer, errno);
  return;
 }
 dispatch_source_t source = dispatch_source_create(DISPATCH_SOURCE_TYPE_VNODE, fd,
               DISPATCH_VNODE_WRITE | DISPATCH_VNODE_DELETE, DISPATCH_TARGET_QUEUE_DEFAULT);
 dispatch_source_set_event_handler(source, ^(){
  unsigned long const data = dispatch_source_get_data(source);
  if (data & DISPATCH_VNODE_WRITE) {
   NSLog(@"The directory changed.");
  }
  if (data & DISPATCH_VNODE_DELETE) {
   NSLog(@"The directory has been deleted.");
  }
 });
 dispatch_source_set_cancel_handler(source, ^(){
  close(fd);
 });
 _source = source;
 dispatch_resume(_source);

还要注意需要用DISPATCH_VNODE_DELETE去检查监视的文件或文件夹是否被删除，如果删除了就停止监听

6. dispathc_once

dispathc_once函数可以确保某个 block 在应用程序执行的过程中只被处理一次，而且它是线程安全的。所以单例模式可以很简单的实现,写单例必备

+ (Manager *)sharedInstance {
    static Manager *sharedManagerInstance = nil;
    static dispatch_once_t once;

    dispatch_once($once, ^{
        sharedManagerInstance = [[Manager alloc] init];
    });

    return sharedManagerInstance;
}

7. dispathc_apply

7.1 并行队列完成任务,开启新任务

dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
dispatch_apply(10, queue, ^(size_t index) {
    NSLog(@"%d\n",index);
});
NSLog(@"done");

打印结果为:

2017-03-04 21:40:43.454 GCDTrain[1289:42894] 0
2017-03-04 21:40:43.454 GCDTrain[1289:42937] 3
2017-03-04 21:40:43.454 GCDTrain[1289:42940] 2
2017-03-04 21:40:43.454 GCDTrain[1289:42894] 4
2017-03-04 21:40:43.454 GCDTrain[1289:42894] 5
2017-03-04 21:40:43.454 GCDTrain[1289:42938] 1
2017-03-04 21:40:43.455 GCDTrain[1289:42894] 8
2017-03-04 21:40:43.454 GCDTrain[1289:42940] 6
2017-03-04 21:40:43.455 GCDTrain[1289:42937] 7
2017-03-04 21:40:43.455 GCDTrain[1289:42938] 9
2017-03-04 21:40:43.455 GCDTrain[1289:42894] done

所以这又是一种对多线程无序的限定的API,dispatch_applyblock内的所有任务被执行完之后才会执行后面的代码,当然dispatch_apply与前面提到的所有带_wait的API一样都是阻塞当前线程的
dispatch_apply与带_wait的API也有不同,dispatch_apply比较适合做些重复的+执行次数确定的任务

7.2 防止开启线程过多

dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
for (int i = 0; i < 999; i++){
      dispatch_async(queue, ^{
         NSLog(@"%d,%@",i,[NSThread currentThread]);
      });
}

dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
dispatch_apply(999, queue, ^(size_t i){
     NSLog(@"%d,%@",i,[NSThread currentThread]);
});

看两份代码的打印结果可知道,不用dispatch_apply的并行+异步会开启许多线程,而我们已经知道:「使用太多线程会导致消耗大量内存」,所以在这种场景下应该使用dispatch_apply

8.dispatch_set_target_queue

8.1dispatch_set_target_queue可以设置queue的优先级

dispatch_queue_create创建队列的优先级跟global dispatch queue的默认优先级一样，假如我们需要设置队列的优先级，可以通过dispatch_set_target_queue方法

dispatch_queue_t serialQueue = dispatch_queue_create("com.pogong.www", DISPATCH_QUEUE_SERIAL);  
dispatch_queue_t globalQueue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_HIGH,0);  
dispatch_set_target_queue(serialQueue, globalQueue);
//这样serialQueue的优先级和globalQueue的优先级一样

8.2 定义队列层级关系

将多个队列用dispatch_set_target_queue设置为某个串行队列的下属队列,可以防止并行执行

加队列层级关系

    dispatch_queue_t targetQueue = dispatch_queue_create("target_queue", DISPATCH_QUEUE_SERIAL);
    dispatch_queue_t queue1 = dispatch_queue_create("queue1", DISPATCH_QUEUE_SERIAL);
    dispatch_queue_t queue2 = dispatch_queue_create("queue2", DISPATCH_QUEUE_CONCURRENT);
    dispatch_set_target_queue(queue1, targetQueue);//加层级关系
    dispatch_set_target_queue(queue2, targetQueue);//加层级关系
    dispatch_async(queue1, ^{
        NSLog(@"do job1");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job2");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job3");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job4");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job5");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job6");
    });

2017-03-04 22:07:40.723 GCDTrain[1400:51089] do job1
2017-03-04 22:07:40.723 GCDTrain[1400:51089] do job2
2017-03-04 22:07:40.724 GCDTrain[1400:51089] do job3
2017-03-04 22:07:40.724 GCDTrain[1400:51089] do job4
2017-03-04 22:07:40.724 GCDTrain[1400:51089] do job5
2017-03-04 22:07:40.725 GCDTrain[1400:51089] do job6

不加队列层级关系

    dispatch_queue_t queue1 = dispatch_queue_create("queue1", DISPATCH_QUEUE_SERIAL);
    dispatch_queue_t queue2 = dispatch_queue_create("queue2", DISPATCH_QUEUE_CONCURRENT);
    dispatch_async(queue1, ^{
        NSLog(@"do job1");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job2");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job3");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job4");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job5");
    });
    dispatch_async(queue2, ^{
        NSLog(@"do job6");
    });

2017-03-04 22:01:04.781 GCDTrain[1365:48597] do job2
2017-03-04 22:01:04.781 GCDTrain[1365:48598] do job1
2017-03-04 22:01:04.781 GCDTrain[1365:48599] do job3
2017-03-04 22:01:04.781 GCDTrain[1365:48609] do job5
2017-03-04 22:01:04.781 GCDTrain[1365:48608] do job4
2017-03-04 22:01:04.782 GCDTrain[1365:48597] do job6

打印结果一目了然

9. dispatch IO

在书上,在别人的文章里对dispatch IO都是一段引用苹果官方的代码,跑不起来,自己真会用了再说

10. 未完待续

还有很多API并未见过介绍,所以.....

文章参考:
Objective-C高级编程:iOS与OS X多线程和内存管理
 bestswifer iOS多线程编程——GCD与NSOperation总结
 戴铭细说GCD如何用