jasonkuo41 / CellWars.Threading.AsyncLock

A high performance and re-entrant async lock for C#.

CellWars.Threading.AsyncLock

A high performance and re-entrant async lock for C#.

No dead-lock if re-acquired on different thread, it's just works like your plain old lock.

Supports netstandard 1.3 (.Net Framework 4.6, .Net Core 1.0 and above)

AsyncLock

This is an async-compatible counterpart of lock in C#.

It not only provides the ability to lock an async-await block but also provides features like re-entrant; meaning you can re-acquire and lock the same lock again without suffering dead-lock or any other consequences, not even performance degradation.

Note : We still recommend using pure lock if your code does not involve any asynchronous calls, as it's proven to be working wonderfully and is faster then using AsyncLock or any other implementation

You can read the #Theory for further understanding of how this is achieved.

Quick Start

The using of AsyncLock is extremely straight forward, it takes advantage of using and unlocks the mutex when Dispose() is called.

public class ExampleClass {
    private readonly AsyncLock _mutex = new AsyncLock();
    public async Task LockContentAsync() {
        // This is equal to lock(_mutex) {}
        using(await _mutex.LockAsync()) {
            // It is safe to await when AsyncLock is locked
            // This garuntees only one thread is allow to access this block one at a time
            await Task.Delay(1000);
        }
    }
}

We highly discourage users to use the synchronized version of the AsyncLock, unless needed in special circumstances. It is best to combine both AsyncLock and lock together. AsyncLock for async content and lock for synchronize content.

For Example :

public async Task FooAAsync() {
    using (await _mutex.LockAsync()) {
		lock (syncObject) {
			// Access sychronized content, for example get a state of your current class
		}
		// Apply this state to some async functions
	}
}

public void FooB() {
	lock(syncObject) {
		// Do some synchronized stuff
	}
}

Since the biggest feature of AsyncLock is that it behaves like lock which is re-entrant, meaning that the code below would not end up in a dead-lock (like SempahoreSlim.WaitAsync and other commonly used async lock provided by others).

public async Task ReentrantLockAsync() {
    using (await _mutex.LockAsync()) {
        using (await _mutex.LockAsync()) {
            // Does not dead lock
        }
    }
}

Notes : Remember to Dispose AsyncLock when it's belonging class no longer uses it; although it is pretty safe not disposing it in the current implementation, we still recommend doing a check.

Timeout and Cancellation

AsyncLock also provides a timeout and CancellationToken parameter for its interface. When timeout is provided, the lock will only wait for that timespan before throwing TimoutException; If CancellationToken is canceled, this would cause OperationCancelledException to be thrown.

If the mutex is already acquired, then cancellation from CancellationToken would not be effective. If timeout is not specified, it would fallback to wait for the default timeout specified in the constructor of AsyncLock, if that is also not specified, then it would wait forever.

Theory

This section will explain how AsyncLock manages to allow re-entrant and it's high performance.

AsyncLock utilizes two main classes SemphoreSlim and AsyncLocal; the actual lock is maintained by SemphoreSlim and the re-entrant check is done by using AsyncLocal.

When locking with AsyncLock, it returns an object implementing ILockHandle, which can either contain actual content that does the unlocking when Dispose() is called, or does absolutely nothing. The former one is acquired when the current Task has not acquired AsyncLock, the latter one is returned when AsyncLock detects when it already acquired the lock.

The detection of the current lock acquired by the current task is achieved through AsyncLocal, the underlying value of AsyncLocal is (almost) different when it's in different Task. When it's underlying value is null we know that the current Task hasn't acquired the lock and is put into the semaphore; if it's not null then we know the current Task has already acquired a lock given by this current AsyncLock instance, and would not await the semaphore and returns an empty ILockHandle.

If you like to know how AsyncLocal can distinguish between different Task's, you are welcome to read up the documents provided by MSDN and this SO post.

Warning : Although it's unlikely to happen, you shall not use AsyncLock inside any blocks of where ExecutionContext.SuppressFlow() is present. This would effectively break the re-entrant check with AsyncLocal