代码:
async fn doAsyncPrint(v:u32) {
println!("start doAsyncPrint,v is {},tid is {:?}",v,system::myTid());
//thread::sleep(Duration::from_secs(1));
time::sleep(Duration::from_secs(10)).await;
println!("end,v is {},tid is {:?}",v,system::myTid());
}
fn testCoroutine3() {
println!("testCoroutine3 start");
let array = Arc::new(BlockingQueue::<u32>::new());
let arrayclosure1 = array.clone();
thread::spawn(move|| {
let mut count:u32 = 0;
loop {
//println!("thread start,count is {}",count);
arrayclosure1.putLast(count);
count += 1;
thread::sleep(Duration::from_secs(1));
}
});
let arrayclosure2 = array.clone();
thread::spawn(move ||{
let rt = tokio::runtime::Builder::new_current_thread().enable_all().build().unwrap();
let guard1 = rt.enter();
println!("coroutine start,tid is {:?}", system::myTid());
loop {
let v = arrayclosure2.takeFirst();
//println!("coroutine trace1,v is {}",v);
tokio::spawn(doAsyncPrint(v));
}
}).join();
}
代码运行时会直接卡住,这个和协程的原理有关:
1.let rt = tokio::runtime::Builder::new_current_thread().enable_all().build().unwrap();
这里创建一个协程的runtime,注意:这里协程是跑在当前的线程中
2.let v = arrayclosure2.takeFirst();
从blocking队列中获取数据,注意这个blocking队列使用了mutex和condvar,所以当condvar卡住的时候,实际上是整个线程直接卡住,由于协程只是在线程中来回切换调用栈,所以协程中所有的等待处理实际上都是切换调用栈,但是这个condvar不是切换调用栈,所以就会卡住。
3.tokio::spawn(doAsyncPrint(v));
doAsyncPrint一直没有被调用
方案:
将new_current_thread改成new_multi_thread
代码如下:
async fn doAsyncPrint(v:u32) {
println!("start doAsyncPrint,v is {},tid is {:?}",v,system::myTid());
//thread::sleep(Duration::from_secs(1));
time::sleep(Duration::from_secs(10)).await;
println!("end,v is {},tid is {:?}",v,system::myTid());
}
thread::spawn(move ||{
let rt = tokio::runtime::Builder::new_multi_thread().worker_threads(1).enable_all().build().unwrap();
let guard1 = rt.enter();
println!("coroutine start,tid is {:?}", system::myTid());
loop {
let v = arrayclosure2.takeFirst();
//println!("coroutine trace1,v is {}",v);
tokio::spawn(doAsyncPrint(v));
}
}).join();
这样实际上协程的runtime会创建线程,让协程跑在这些线程上,如上述代码,我们通过worker_threads创建了一个线程,所有的协程都会跑在这个线程上。
我们看一下运行结果:
你会发现,所有的协程都跑在了thread4上,我们看一下代码:
async fn doAsyncPrint(v:u32) {
println!("start doAsyncPrint,v is {},tid is {:?}",v,system::myTid());
//thread::sleep(Duration::from_secs(1));
time::sleep(Duration::from_secs(10)).await;
println!("end,v is {},tid is {:?}",v,system::myTid());
}
发现没有,一开始的时候连续打印了多个“start doAsyncPrint”,但是没有打印“end,v is ....”这个是为啥咧?原因就在time::sleep这个函数(tokio::time::sleep),实际上,这个sleep没有真让线程sleep,而且直接切换调用栈,切换到下一个函数上,所以一开始我们可以看到连续的“start doAsyncPrint”。
协程的使用还是比较要当心的,所有操作io操作啥估计都要使用tokio的接口,否者就会导致卡住。例如上面的代码,如果改成thread::sleep的话,就会直接卡住线程,输出结果就如下: