一、Job生命周期
通过
1.Job与async
CoroutineScope的扩展函数launch返回一个Job对象。
public fun CoroutineScope.launch(
context: CoroutineContext = EmptyCoroutineContext,
start: CoroutineStart = CoroutineStart.DEFAULT,
block: suspend CoroutineScope.() -> Unit
): Job {
val newContext = newCoroutineContext(context)
val coroutine = if (start.isLazy)
LazyStandaloneCoroutine(newContext, block) else
StandaloneCoroutine(newContext, active = true)
coroutine.start(start, coroutine, block)
return coroutine
}
public interface Job : CoroutineContext.Element {
public companion object Key : CoroutineContext.Key<Job>
public val isActive: Boolean
......
}
CoroutineScope的扩展函数async返回一个Deferred对象,Deferred 继承 Job。
public fun <T> CoroutineScope.async(
context: CoroutineContext = EmptyCoroutineContext,
start: CoroutineStart = CoroutineStart.DEFAULT,
block: suspend CoroutineScope.() -> T
): Deferred<T> {
val newContext = newCoroutineContext(context)
val coroutine = if (start.isLazy)
LazyDeferredCoroutine(newContext, block) else
DeferredCoroutine<T>(newContext, active = true)
coroutine.start(start, coroutine, block)
return coroutine
}
Deferred 继承 Job,同样是一个接口,Deferred 有一个泛型参数T,并且有一个关键的await()方法,其返回值类型是T。
public interface Deferred<out T> : Job {
public suspend fun await(): T
......
}
2.job.cancel()
fun main() {
runBlocking {
val job = launch {
println("launch start")
delay(2000L)
println("launch end")
}
job.printlog()
job.cancel()
job.printlog()
}
}
fun Job.printlog() {
printCoroutine("isActive:$isActive, isCancelled:$isCancelled, isCompleted:$isCompleted")
}
fun printCoroutine(any: Any?) {
println("" + any + ";Thread:" + Thread.currentThread().name)
}
输出log 如下:
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
isActive:false, isCancelled:true, isCompleted:false;Thread:main @coroutine#1
上面的代码中,通过job.cancel()取消协程任务,然后通过job.isCancelled查看状态为true。代表了协程任务处于取消状态。所以launch中的协程任务没有执行就取消了。
对上面的代码做简单修改:在job执行cancel前delay 1000ms,可以发现launch任务执行过程中被取消了。
runBlocking {
val job = launch {
println("launch start")
delay(2000L)
println("launch end")
}
job.printlog()
delay(1000L)
job.cancel()
job.printlog()
}
输出log:
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
launch start
isActive:false, isCancelled:true, isCompleted:false;Thread:main @coroutine#1
如果在上述代码中delay(2000L)后再次执行 job.printlog()。可以发现此时:isCompleted:true。
runBlocking {
val job = launch {
println("launch start")
delay(2000L)
println("launch end")
}
job.printlog()
delay(1000L)
job.cancel()
job.printlog()
delay(2000)
job.printlog()
}
Log输出如下:
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
launch start
isActive:false, isCancelled:true, isCompleted:false;Thread:main @coroutine#1
isActive:false, isCancelled:true, isCompleted:true;Thread:main @coroutine#1
3.job.start()
fun main() {
runBlocking {
val job = launch(start = CoroutineStart.LAZY) {
printCoroutine("Coroutinue start!")
delay(2000L)
printCoroutine("Coroutinue end!")
}
delay(1000L)
job.printlog()
job.start()
job.printlog()
delay(1000L)
job.cancel()
delay(1000L)
job.printlog()
}
}
结果:
isActive:false, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
Coroutinue start!;Thread:main @coroutine#2
isActive:false, isCancelled:true, isCompleted:true;Thread:main @coroutine#1
将启动模式设置为Coroutinue.LAZY,协程任务被launch以后,可以发现isActive为false,不会立即执行,当执行job.start()之后,isActive为true,所以Coroutinue.LAZY是一种懒加载的行为模式。所以通过Log可以发现只有调用了job.start()以后,job的状态才变成了Active。调用了cancel以后,job的状态才会变成isCancelled、isCompleted。
协程的Job有两种初始状态,如果Job是以懒加载的方式创建,那么它的初始状态是New;如果协程以非懒加载的方式创建,它的初始状态是Active。
注意:当调用cancel以后,isCancelled = true,然后isCompleted = true,说明对于协程来说,取消协程,也是一种“结束状态”。
New、Active、Completing、Cancelling、Completed、Cancelled 这些状态,都是 Job 内部私有的状态。而 Job 对外暴露出的 isCompleted 并不是与其一一对应的。Job 内部私有的 Completed、Cancelled 状态,都会认为是外部的 isCompleted 状态。
4.协程正常结束的生命周期
fun main() {
runBlocking {
val job = launch(start = CoroutineStart.LAZY) {
printCoroutine("Coroutinue start!")
delay(1000L)
printCoroutine("Coroutinue end!")
}
delay(500L)
job.printlog()
job.start()
job.printlog()
delay(2000L)
job.printlog()
delay(2000L)
printCoroutine("Process end!")
}
}
输出:
isActive:false, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
Coroutinue start!;Thread:main @coroutine#2
Coroutinue end!;Thread:main @coroutine#2
isActive:false, isCancelled:false, isCompleted:true;Thread:main @coroutine#1
Process end!;Thread:main @coroutine#1
从Log可以发现:isActive从false到true;isCompleted从false到true。
5.Job内部delay大于外部delay
runBlocking {
val job = launch(start = CoroutineStart.LAZY) {
printCoroutine("Coroutinue start!")
delay(6000L)
printCoroutine("Coroutinue end!")
}
delay(500L)
job.printlog()
job.start()
job.printlog()
delay(1000L)
job.printlog()
delay(1000L)
printCoroutine("Process end!")
}
输出:
isActive:false, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
Coroutinue start!;Thread:main @coroutine#2
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
Process end!;Thread:main @coroutine#1
Coroutinue end!;Thread:main @coroutine#2
在上面的代码中,将job内部的delay改为6000L后,runBlocking会一直阻塞,等到job任务执行完毕以后才能真正退出。
6.等待和监听协程的结束事件
runBlocking {
suspend fun download() {
printCoroutine("Start Download...")
delay(4000L)
printCoroutine("End Download...")
}
val job = launch(start = CoroutineStart.LAZY) {
download()
}
delay(200L)
job.printlog()
job.start()
job.printlog()
job.invokeOnCompletion {
job.printlog()
}
job.join()
printCoroutine("Process End!")
}
Log:
isActive:false, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
isActive:true, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
Start Download...;Thread:main @coroutine#2
End Download...;Thread:main @coroutine#2
isActive:false, isCancelled:false, isCompleted:true;Thread:main @coroutine#2
Process End!;Thread:main @coroutine#1
通过Log可以发现,invokeOnCompletion{}可以用来监听协程结束的事件,即使job取消,invokeOnCompletion{}仍然会被调用。
在上面代码中job.start()执行以后执行job.cancel()。可以发现:即使job取消,invokeOnCompletion{}仍然会被调用。
runBlocking {
suspend fun download() {
printCoroutine("Start Download...")
delay(4000L)
printCoroutine("End Download...")
}
val job = launch(start = CoroutineStart.LAZY) {
download()
}
delay(200L)
job.printlog()
job.start()
job.cancel()
job.printlog()
job.invokeOnCompletion {
job.printlog()
}
job.join()
printCoroutine("Process End!")
}
isActive:false, isCancelled:false, isCompleted:false;Thread:main @coroutine#1
isActive:false, isCancelled:true, isCompleted:false;Thread:main @coroutine#1
isActive:false, isCancelled:true, isCompleted:true;Thread:main @coroutine#2
Process End!;Thread:main @coroutine#1
job.join() 其实是一个“挂起函数”,它的作用就是:挂起当前的程序执行流程,等待 job 当中的协程任务执行完毕,然后再恢复当前的程序执行流程。
7.Job源码
public interface Job : CoroutineContext.Element {
// ------------ 状态查询API ------------
public val isActive: Boolean
public val isCompleted: Boolean
public val isCancelled: Boolean
public fun getCancellationException(): CancellationException
// ------------ 操控状态API ------------
public fun start(): Boolean
public fun cancel(cause: CancellationException? = null)
public fun cancel(): Unit = cancel(null)
public fun cancel(cause: Throwable? = null): Boolean
// ------------ 等待状态API ------------
public suspend fun join()
public val onJoin: SelectClause0
// ------------ 完成状态回调API ------------
public fun invokeOnCompletion(handler: CompletionHandler): DisposableHandle
public fun invokeOnCompletion(
onCancelling: Boolean = false,
invokeImmediately: Boolean = true,
handler: CompletionHandler): DisposableHandle
}
8.Deferred
await()方法
runBlocking {
val deferred = async {
printCoroutine("Coroutinue Start!")
delay(1000L)
printCoroutine("Coroutinue End!")
"result"
}
val result = deferred.await()
println("Result:$result")
printCoroutine("Process End!")
}
Coroutinue Start!;Thread:main @coroutine#2
Coroutinue End!;Thread:main @coroutine#2
Result:result
Process End!;Thread:main @coroutine#1
deferred.await() 可以获取协程的执行结果,它还会阻塞当前协程的执行流程,直到协程任务执行完毕。在这一点的行为上,await() 和 join() 是类似的。
public interface Deferred<out T> : Job {
public suspend fun await(): T
... ...
}
await() 是一个挂起函数,拥有挂起和恢复的能力。如果当前的 Deferred 任务还没执行完毕,那么,await() 就会挂起当前的协程执行流程,等待 Deferred 任务执行完毕,再恢复执行后面剩下的代码。而job.join() 在协程执行完毕之前,后面的协程代码都被暂时挂起了,等到协程执行完毕,才有机会继续执行。
二、Job 与结构化并发
runBlocking {
val parentJob: Job
var job1: Job? = null
var job2: Job? = null
var job3: Job? = null
parentJob = launch {
job1 = launch {
delay(1000L)
}
job2 = launch {
delay(2000L)
}
job3 = launch {
delay(3000L)
}
}
delay(500L)
parentJob.children.forEachIndexed { index, job ->
when (index) {
0 -> println("job1 === job is ${job1 === job}")
1 -> println("job2 === job is ${job2 === job}")
2 -> println("job2 === job is ${job3 === job}")
}
}
parentJob.join()
printCoroutine("Process End!")
}
job1 === job is true
job2 === job is true
job2 === job is true
Process End!;Thread:main @coroutine#1
parentJob 是最外层的 launch 返回的对象,在这个 launch 的内部,嵌套了三个 launch,它们的 Job 对象分别赋值给了 job1、job2、job3。在parentJob.children遍历过程中,发现job1、job2、job3 其实就是 parentJob 的 children。也就是说,我们使用 launch 创建出来的协程,是存在父子关系的。
public interface Job : CoroutineContext.Element {
// ------------ parent-child ------------
public val children: Sequence<Job>
@InternalCoroutinesApi
public fun attachChild(child: ChildJob): ChildHandle
}
每个 Job 对象,都会有一个Sequence类型的 children 属性,是一个惰性的集合,我们可以对它进行遍历。而 attachChild() 则是一个协程内部的 API,用于绑定 ChildJob 。
通过调用parentJob 的 join() 方法,会等待其内部的 job1、job2、job3 全部执行完毕,才会恢复执行。
runBlocking {
val parentJob: Job
var job1: Job? = null
var job2: Job? = null
var job3: Job? = null
parentJob = launch {
job1 = launch {
printCoroutine("job1 start")
delay(1000L)
printCoroutine("job1 end")
}
job2 = launch {
printCoroutine("job2 start")
delay(2000L)
printCoroutine("job2 end")
}
job3 = launch {
printCoroutine("job3 start")
delay(3000L)
printCoroutine("job3 end")
}
}
delay(500L)
parentJob.children.forEachIndexed { index, job ->
when (index) {
0 -> println("job1 === job is ${job1 === job}")
1 -> println("job2 === job is ${job2 === job}")
2 -> println("job2 === job is ${job3 === job}")
}
}
parentJob.cancel()
printCoroutine("Process End!")
}
job1 start;Thread:main @coroutine#3
job2 start;Thread:main @coroutine#4
job3 start;Thread:main @coroutine#5
job1 === job is true
job2 === job is true
job2 === job is true
Process End!;Thread:main @coroutine#1
parentJob 调用cancel() 方法,job1、job2、job3的协程任务也全都被取消了。
三、async使用场景
runBlocking {
suspend fun getResult1(): String {
delay(1000L)
return "Result1"
}
suspend fun getResult2(): String {
delay(1000L)
return "Result2"
}
suspend fun getResult3(): String {
delay(1000L)
return "Result3"
}
val results = mutableListOf<String>()
val time = measureTimeMillis {
results.add(getResult1())
results.add(getResult2())
results.add(getResult3())
}
println("Time:$time")
println(results)
}
Time:3052
[Result1, Result2, Result3]
函数总耗时未3052毫秒,可以使用async进行优化。
runBlocking {
suspend fun getResult1(): String {
delay(1000L)
return "Result1"
}
suspend fun getResult2(): String {
delay(1000L)
return "Result2"
}
suspend fun getResult3(): String {
delay(1000L)
return "Result3"
}
val results = mutableListOf<String>()
val time = measureTimeMillis {
val result1 = async {
getResult1()
}
val result2 = async {
getResult1()
}
val result3 = async {
getResult1()
}
results.add(result1.await())
results.add(result2.await())
results.add(result3.await())
}
println("Time:$time")
println(results)
}
Time:1029
[Result1, Result1, Result1]
可以看到 async可以用来优化一些并发操作。