I need to write some asynchronous code that essentially attempts to repeatedly talk to and initialise a database. Quite often the first attempt will fail hence the requirem
You don't really need WaitItForWork
method, just await for a database initialization task:
async Task Run()
{
await InitializeDatabase();
// Do what you need after database is initialized
}
async Task InitializeDatabase()
{
// Perform database initialization here
}
If you have multiple pieces of code that call to WaitForItToWork
then you need to wrap database initialization into a Task
and await it in all workers, for example:
readonly Task _initializeDatabaseTask = InitializeDatabase();
async Task Worker1()
{
await _initializeDatabaseTask;
// Do what you need after database is initialized
}
async Task Worker2()
{
await _initializeDatabaseTask;
// Do what you need after database is initialized
}
static async Task InitializeDatabase()
{
// Initialize your database here
}
Just provide another solution
public static void WaitForCondition(Func<bool> predict)
{
Task.Delay(TimeSpan.FromMilliseconds(1000)).ContinueWith(_ =>
{
var result = predict();
// the condition result is false, and we need to wait again.
if (result == false)
{
WaitForCondition(predict);
}
});
}
If the task is asynchronous you can try with:
async Task WaitForItToWork()
{
await Task.Run(() =>
{
bool succeeded = false;
while (!succeeded)
{
// do work
succeeded = outcome; // if it worked, make as succeeded, else retry
System.Threading.Thread.Sleep(1000); // arbitrary sleep
}
});
}
See http://msdn.microsoft.com/en-us/library/hh195051.aspx.
You could refactor that fragment like this:
async Task<bool> WaitForItToWork()
{
bool succeeded = false;
while (!succeeded)
{
// do work
succeeded = outcome; // if it worked, make as succeeded, else retry
await Task.Delay(1000); // arbitrary delay
}
return succeeded;
}
Apparently, the only benefit it would give you is more efficient use of thread pool, because it doesn't always take a whole thread to make the delay happen.
Depending on how you obtain outcome
, there may be much more efficient ways to get this job done using async/await
. Often you may have something like GetOutcomeAsync()
which would make a web service, database or socket call asynchronously in a natural way, so you'd just do var outcome = await GetOutcomeAsync()
.
It's important to take into account that WaitForItToWork
will be split into parts by compiler and the part from await
line will be continued asynchronously. Here's perhaps the best explanation on how it's done internally. The thing is, usually at some point of your code you'd need to synchronize on the result of the async task. E.g.:
private void Form1_Load(object sender, EventArgs e)
{
Task<bool> task = WaitForItToWork();
task.ContinueWith(_ => {
MessageBox.Show("WaitForItToWork done:" + task.Result.toString()); // true or false
}, TaskScheduler.FromCurrentSynchronizationContext());
}
You could have simply done this:
private async void Form1_Load(object sender, EventArgs e)
{
bool result = await WaitForItToWork();
MessageBox.Show("WaitForItToWork done:" + result.toString()); // true or false
}
That would however make Form1_Load
an async method too.
[UPDATE]
Below is my attempt to to illustrate what async/await
actually does in this case. I created two versions of the same logic, WaitForItToWorkAsync
(using async/await
) and WaitForItToWorkAsyncTap
(using TAP pattern without async/await
). The frist version is quite trivial, unlike the second one. Thus, while async/await
is largely the compiler's syntactic sugar, it makes asynchronous code much easier to write and understand.
// fake outcome() method for testing
bool outcome() { return new Random().Next(0, 99) > 50; }
// with async/await
async Task<bool> WaitForItToWorkAsync()
{
var succeeded = false;
while (!succeeded)
{
succeeded = outcome(); // if it worked, make as succeeded, else retry
await Task.Delay(1000);
}
return succeeded;
}
// without async/await
Task<bool> WaitForItToWorkAsyncTap()
{
var context = TaskScheduler.FromCurrentSynchronizationContext();
var tcs = new TaskCompletionSource<bool>();
var succeeded = false;
Action closure = null;
closure = delegate
{
succeeded = outcome(); // if it worked, make as succeeded, else retry
Task.Delay(1000).ContinueWith(delegate
{
if (succeeded)
tcs.SetResult(succeeded);
else
closure();
}, context);
};
// start the task logic synchronously
// it could end synchronously too! (e.g, if we used 'Task.Delay(0)')
closure();
return tcs.Task;
}
// start both tasks and handle the completion of each asynchronously
private void StartWaitForItToWork()
{
WaitForItToWorkAsync().ContinueWith((t) =>
{
MessageBox.Show("WaitForItToWorkAsync complete: " + t.Result.ToString());
}, TaskScheduler.FromCurrentSynchronizationContext());
WaitForItToWorkAsyncTap().ContinueWith((t) =>
{
MessageBox.Show("WaitForItToWorkAsyncTap complete: " + t.Result.ToString());
}, TaskScheduler.FromCurrentSynchronizationContext());
}
// await for each tasks (StartWaitForItToWorkAsync itself is async)
private async Task StartWaitForItToWorkAsync()
{
bool result = await WaitForItToWorkAsync();
MessageBox.Show("WaitForItToWorkAsync complete: " + result.ToString());
result = await WaitForItToWorkAsyncTap();
MessageBox.Show("WaitForItToWorkAsyncTap complete: " + result.ToString());
}
A few words on threading. There is no additional threads explicitly created here. Internally, Task.Delay()
implementation may use pool threads (I suspect they use Timer Queues), but in this particular example (a WinForms app), the continuation after await
will happen on the same UI thread. In other execution environments (e.g. a console app), it might continue on a different thread. IMO, this article by Stephen Cleary is a must-read to understand async/await
threading concepts.