actor 是线程安全的么?

actor 是线程安全的么?

The Actor Reentrancy Problem in Swift

不是的。考虑reentrance的情况

suspension point是否在actor里,如果有的话,那么在并发的两个task,就有可能会发生重入(Reentrance)的情况。

举个例子:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
swift复制代码actor BankAccount {
   
   private var balance = 1000
   
   func withdraw(_ amount: Int) async {
       
       print("🤓 Check balance for withdrawal: (amount)")
       
       guard canWithdraw(amount) else {
           print("🚫 Not enough balance to withdraw: (amount)")
           return
       }
       
       guard await authorizeTransaction() else {
           return
       }
       
       print("✅ Transaction authorized: (amount)")
       
       balance -= amount
       
       print("💰 Account balance: (balance)")
   }
   
   private func canWithdraw(_ amount: Int) -> Bool {
       return amount <= balance
   }
   
   private func authorizeTransaction() async -> Bool {
       
       // Wait for 1 second
       try? await Task.sleep(nanoseconds: 1 * 1000000000)
       
       return true
   }
}

let account = BankAccount()

Task {
   await account.withdraw(800)
}

Task {
   await account.withdraw(500)
}

我们可以想象一下,这个会输出什么结果?3,2,1

乍一看,我们会想,先执行完await account.withdraw(800),然后剩余200,然后执行await account.withdraw(500),发现余额不足,不执行。

但是结果如下:

1
2
3
4
5
6
yaml复制代码🤓 Check balance for withdrawal: 800
🤓 Check balance for withdrawal: 500
✅ Transaction authorized: 800
💰 Account balance: 200
✅ Transaction authorized: 500
💰 Account balance: -300

你看,余额变成了-300,这明显与我们的认知不符合。那么为什么会出现的结果呢,其实是因为,actor只是能保证不出现数据竞争,但是我们的withdraw方法都会在authorizeTransaction挂起,那么也就是意味着,我们并没有在一个task执行完,才去执行另一个。也就是说,它并不会去判断挂起的时候,可变的状态还是保持你刚进入task的状态,可能有点拗口,那拿这个例子举例,就是actor不会保证,balance会自动修改为task1执行完后的值,也就是200。所以我们在做balance -= amount操作的时候,需要再判断一下,余额是否满足条件,这也是apple建议我们做的,即始终保持actor的状态变更是同步的。

即:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
swift复制代码func withdraw(_ amount: Int) async {
   
   // Perform authorization before check balance
   guard await authorizeTransaction() else {
       return
   }
   print("✅ Transaction authorized: (amount)")
   
   print("🤓 Check balance for withdrawal: (amount)")
   guard canWithdraw(amount) else {
       print("🚫 Not enough balance to withdraw: (amount)")
       return
   }
   
   balance -= amount
   
   print("💰 Account balance: (balance)")
   
}

但是,假设authorizeTransaction方法非常耗时,而余额很明显是不满足扣款条件的,就是一种资源的浪费。所以我们可以:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
swift复制代码func withdraw(_ amount: Int) async {
   
   print("🤓 Check balance for withdrawal: (amount)")
   guard canWithdraw(amount) else {
       print("🚫 Not enough balance to withdraw: (amount)")
       return
   }
   
   guard await authorizeTransaction() else {
       return
   }
   print("✅ Transaction authorized: (amount)")
   
   // Check balance again after the authorization process
   guard canWithdraw(amount) else {
       print("⛔️ Not enough balance to withdraw: (amount) (authorized)")
       return
   }

   balance -= amount
   
   print("💰 Account balance: (balance)")
   
}

虽然多做了一次canWithdraw,但是可以提升效率,且避免了重入的问题。

本文转载自: 掘金

开发者博客 – 和开发相关的 这里全都有

0%