<img src="https://promisesaplus.com/assets/logo-small.png" alt="Promises/A+ logo" title="Promises/A+ 1.0 compliant" align="right" />
Promises library for C# for management of asynchronous operations.
Inspired by JavaScript promises, but slightly different.
Used by Real Serious Games for building serious games and simulations on Unity3d.
If you are interested in using promises for game development and Unity please see this article.
Table of Contents generated with DocToc
To learn about promises:
This promise library conforms to the Promises/A+ Spec (at least, as far as is possible with C#):
The DLL can be installed via nuget. Use the Package Manager UI or console in Visual Studio or use nuget from the command line.
See here for instructions on installing a package via nuget: http://docs.nuget.org/docs/start-here/using-the-package-manager-console
The package to search for is RSG.Promise.
You can get the code by cloning the github repository. You can do this in a UI like SourceTree or you can do it from the command line as follows:
git clone https://github.com/Real-Serious-Games/C-Sharp-Promise.git
Alternately, to contribute please fork the project in github.
Reference the DLL and import the namespace:
using RSG;
Create a promise before you start the async operation:
var promise = new Promise<string>();
The type of the promise should reflect the result of the async op.
Then initiate your async operation and return the promise to the caller.
Upon completion of the async op the promise is resolved:
promise.Resolve(myValue);
The promise is rejected on error/exception:
promise.Reject(myException);
To see it in context, here is an example function that downloads text from a URL. The promise is resolved when the download completes. If there is an error during download, say unresolved domain name, then the promise is rejected:
public IPromise<string> Download(string url)
{
var promise = new Promise<string>(); // Create promise.
using (var client = new WebClient())
{
client.DownloadStringCompleted += // Monitor event for download completed.
(s, ev) =>
{
if (ev.Error != null)
{
promise.Reject(ev.Error); // Error during download, reject the promise.
}
else
{
promise.Resolve(ev.Result); // Downloaded completed successfully, resolve the promise.
}
};
client.DownloadStringAsync(new Uri(url), null); // Initiate async op.
}
return promise; // Return the promise so the caller can await resolution (or error).
}
There is another way to create a promise that replicates the JavaScript convention of passing a resolver function into the constructor. The resolver function is passed functions that resolve or reject the promise. This allows you to express the previous example like this:
var promise = new Promise<string>((resolve, reject) =>
{
using (var client = new WebClient())
{
client.DownloadStringCompleted += // Monitor event for download completed.
(s, ev) =>
{
if (ev.Error != null)
{
reject(ev.Error); // Error during download, reject the promise.
}
else
{
resolve(ev.Result); // Downloaded completed successfully, resolve the promise.
}
};
client.DownloadStringAsync(new Uri(url), null); // Initiate async op.
}
});
The simplest usage is to register a completion handler to be invoked on completion of the async op:
Download("http://www.google.com")
.Then(html =>
Console.WriteLine(html)
);
This snippet downloads the front page from Google and prints it to the console.
For all but the most trivial applications you will also want to register an error hander:
Download("http://www.google.com")
.Then(html =>
Console.WriteLine(html)
)
.Catch(exception =>
Console.WriteLine("An exception occured while downloading!")
);
The chain of processing for a promise ends as soon as an error/exception occurs. In this case when an error occurs the Catch handler would be called, but not the Done handler. If there is no error, then only Done is called.
Multiple async operations can be chained one after the other using Then:
Download("http://www.google.com")
.Then(html =>
return Download(ExtractFirstLink(html)) // Extract the first link and download it.
)
.Then(firstLinkHtml =>
Console.WriteLine(firstLinkHtml)
)
.Catch(exception =>
Console.WriteLine("An exception occured while downloading!")
);
Here we are chaining another download onto the end of the first download. The first link in the html is extracted and we then download that. Then expects the return value to be another promise. The chained promise can have a different result type.
Sometimes you will want to simply transform or modify the resulting value without chaining another async operation.
Download("http://www.google.com")
.Then(html =>
return ExtractAllLinks(html)) // Extract all links and return an array of strings.
)
.Then(links => // The input here is an array of strings.
foreach (var link in links)
{
Console.WriteLine(link);
}
);
As is demonstrated the type of the value can also be changed during transformation. In the previous snippet a Promise<string>
is transformed to a Promise<string[]>
.
An error raised in a callback aborts the function and all subsequent callbacks in the chain:
promise.Then(v => Something()) // <--- An error here aborts all subsequent callbacks...
.Then(v => SomethingElse())
.Then(v => AnotherThing())
.Catch(e => HandleError(e)) // <--- Until the error handler is invoked here.
When Catch
is omitted exceptions go silently unhandled. This is an acknowledged issue with the Promises pattern.
We handle this in a similar way to the JavaScript Q library. The Done
method is used to terminate a chain, it registers a default catch handler that propagates unhandled exceptions to a default error handling mechanism that can be hooked into by the user.
Terminating a Promise chain using Done
:
promise.Then(v => Something())
.Then(v => SomethingElse())
.Then(v => AnotherThing())
.Done(); // <--- Terminate the pipeline and propagate unhandled exceptions.
To use the Done
you must apply the following rule: When you get to the end of a chain of promises, you should either return the last promise or end the chain by calling Done
.
To hook into the unhandled exception stream:
Promise.UnhandledException += Promise_UnhandledException;
Then forward the exceptions to your own logging system:
private void Promise_UnhandledException(object sender, ExceptionEventArgs e)
{
Log.Error(e.Exception, "An unhandled promises exception occured!");
}
Promises can additionally report their progress towards completion, allowing the implementor to give the user feedback on the asynchronous operation. The general convention is to report progress as a value from 0 to 1.
For this, you can either call Progress
in the promise definition chain or add a third parameter to the Then
method.
Listening for progress reporting from a promise using Progress
:
var promise = new Promise();
promise.Progress((progress) => Log.Info("Current progress is " + (100f * progress) + "%"));
Listening for progress on a Then
call:
var promiseA = new Promise();
var promiseB = new Promise();
promise
.Then(() => promiseB, null, (progress) => Log.Info("promiseA made progress: " + progress))
.Progress(progress => Log.Info("promiseB made progress: " + progress));
In order to report progress for a promise, you need to call the ReportProgress
method:
var promise = new Promise();
promise.ReportProgress(0.5f); // Report a 50% completion
For convenience or testing you will at some point need to create a promise that starts out in the resolved or rejected state. This is easy to achieve using Resolved and Rejected functions:
var resolvedPromise = Promise<string>.Resolved("some result");
var rejectedPromise = Promise<string>.Rejected(someException);
The class Promise
IPromise<T>
Interface to await promise resolution.IPendingPromise<T>
Interface that can resolve or reject the promise.The All function combines multiple async operations to run in parallel. It converts a collection of promises or a variable length parameter list of promises into a single promise that yields a collection.
Say that each promise yields a value of type T, the resulting promise then yields a collection with values of type T.
Here is an example that extracts links from multiple pages and merges the results:
var urls = new List<string>();
urls.Add("www.google.com");
urls.Add("www.yahoo.com");
Promise<string[]>
.All(urls.Select(url => Download(url))) // Download each URL.
.Then(pages => // Receives collection of downloaded pages.
pages.SelectMany(
page => ExtractAllLinks(page) // Extract links from all pages then flatten to single collection of links.
)
)
.Done(links => // Receives the flattened collection of links from all pages at once.
{
foreach (var link in links)
{
Console.WriteLine(link);
}
});
When listening for progress events in an All operation, the progress that you will receive will be the average of all the progress values reported by all the given promises.
The ThenAll function is a convenient way of chaining multiple promise onto an existing promise:
promise
.Then(result => SomeAsyncOperation(result)) // Chain a single async operation
.ThenAll(result => // Chain multiple async operations.
new IPromise<string>[] // Return an enumerable of promises.
{
SomeAsyncOperation1(result),
SomeAsyncOperation2(result),
SomeAsyncOperation3(result)
}
)
.Done(collection => ...); // Final promise resolves
// with a collection of values
// when all operations have completed.
The Race and ThenRace functions are similar to the All and ThenAll functions, but it is the first async operation that completes that wins the race and it's value resolves the promise.
promise
.Then(result => SomeAsyncOperation(result)) // Chain an async operation.
.ThenRace(result => // Race multiple async operations.
new IPromise<string>[] // Return an enumerable of promises.
{
SomeAsyncOperation1(result),
SomeAsyncOperation2(result),
SomeAsyncOperation3(result)
}
)
.Done(result => ...); // The result has come from whichever of
// the async operations completed first.
When listening for progress events in a race operation, the progress that you will receive will be the maximum of those reported by all the given promises.
The Then function can be used to chain synchronous operations that yield no result.
var promise = ...
promise
.Then(result => SomeAsyncOperation(result)) // Chain an async operation.
.Then(result => Console.WriteLine(result)) // Chain a sync operation that yields no result.
.Done(() => ...); // No result is passed because the previous operation returned nothing.
What about a promise that has no result? This represents an asynchronous operation that promises only to complete, it doesn't promise to yield any value as a result. I call this a non-value promise, as opposed to a value promise, which is a promise that does yield a value. This might seem like a curiousity but it is actually very useful for sequencing visual effects.
Promise
is very similar to Promise<T>
and implements the similar interfaces: IPromise
and IPendingPromise
.
Promise<T>
functions that affect the resulting value have no relevance for the non-value promise and have been removed.
As an example consider the chaining of animation and sound effects as we often need to do in game development:
RunAnimation("Foo") // RunAnimation returns a promise that
.Then(() => RunAnimation("Bar")) // is resolved when the animation is complete.
.Then(() => PlaySound("AnimComplete"));
From time to time you might want to convert a value promise to a non-value promise or vice versa. Both Promise
and Promise<T>
have overloads of Then
and ThenAll
that do this conversion. You just need to return the appropriate type of promise (for Then
) or enumerable of promises (for ThenAll
).
As an example consider a recursive link extractor and file downloader function:
public IPromise DownloadAll(string url)
{
return DownloadURL(url) // Yields a value, the HTML text downloaded.
.Then(html => ExtractLinks(html)) // Convert HTML into an enumerable of links.
.ThenAll(links => // Process each link.
{
// Determine links that should be followed, then follow them.
var linksToFollow = links.Where(link => IsLinkToFollow(link));
var linksFollowing = linksToFollow.Select(link => DownloadAll(link));
// Determine links that are files to be downloaded, then download them.
var linksToDownload = links.Where(link => IsLinkToDownload(link));
var linksDownloading = linksToDownload.Select(link => DownloadFile(link));
// Return an enumerable of promises.
// This combines the recursive link following and any files we want to download.
// Because we are returning an enumerable of non-value promises, the resulting
// chained promises is also non-value.
return linksToFollow.Concat(linksDownloading);
});
}
Usage:
DownloadAll("www.somewhere.com")
.Done(() =>
Console.WriteLine("Recursive download completed.");
);
The Sequence
and ThenSequence
functions build a single promise that wraps multiple sequential operations that will be invoked one after the other.
Multiple promise-yielding functions are provided as input, these are chained one after the other and wrapped in a single promise that is resolved once the sequence has completed.
var sequence = Promise.Sequence(
() => RunAnimation("Foo"),
() => RunAnimation("Bar"),
() => PlaySound("AnimComplete")
);
The inputs can also be passed as a collection:
var operations = ...
var sequence = Promise.Sequence(operations);
This might be used, for example, to play a variable length collection of animations based on data:
var animationNames = ... variable length array of animation names loaded from data...
var animations = animationNames.Select(animName => (Func<IPromise>)(() => RunAnimation(animName)));
var sequence = Promise.Sequence(animations);
sequence
.Done(() =>
{
// All animations have completed in sequence.
});
Unfortunately we find that we have reached the limits of what is possible with C# type inference, hence the use of the ugly cast (Func<IPromise>)
.
The cast can easily be removed by converting the inner anonymous function to an actual function which I'll call PrepAnimation
:
private Func<IPromise> PrepAnimation(string animName)
{
return () => RunAnimation(animName);
}
var animations = animationNames.Select(animName => PrepAnimation(animName));
var sequence = Promise.Sequence(animations);
sequence
.Done(() =>
{
// All animations have completed in sequence.
});
Holy cow... we've just careened into functional programming territory, herein lies very powerful and expressive programming techniques.
We can easily combine sequential and parallel operations to build very expressive logic.
Promise.Sequence( // Play operations 1 and 2 sequently.
() => Promise.All( // Operation 1: Play animation and sound at same time.
RunAnimation("Foo"),
PlaySound("Bar")
),
() => Promise.All(
RunAnimation("One"), // Operation 2: Play animation and sound at same time.
PlaySound("Two")
)
);
I'm starting to feel like we are defining behavior trees.
If you have a promise that comprises a sequence of other promises, you might want to report the total progress for these, and even give more weight to the progress of some promise over another. In this example, we are first downloading an asset from some URL and then we are loading the downloaded asset into memory. We consider that the time it takes to download the asset will be an 80% of the total time, while the time to load it into memory is a 20%:
var promise = new Promise();
Download(url)
.Progress((downloadProgress) => promise.ReportProgress(0.8f * downloadProgress))
.Then((asset) => LoadAssetIntoMemory(asset))
.Progress((loadProgress) => promise.ReportProgress(0.8f + 0.2f * loadProgress))
.Then(() => promise.Resolve())
.Catch((ex) => promise.Reject(ex));
return promise;
The promise timer is not part of the Promises/A+ standard but is a utility that makes it possible to create promises that check if a condition is met each time the promise timer is updated. A common usage of this is in games where the promise timer is updated each frame.
To use it, create an instance of the promise timer and call its Update
method in your main loop:
class Example
{
private IPromiseTimer promiseTimer;
Example()
{
promiseTimer = new PromiseTimer();
}
// Run once for every frame - equivilant to Update() in Unity
void MainLoop()
{
// deltaTime is equal to the time since the last MainLoop
promiseTimer.Update(Time.deltaTime);
// Process your other logic here
}
}
Note that usually it is best to call PromiseTimer.Update
before your other logic, otherwise you may have unintended behaviour such as promises that are supposed to take a very short time resolving in the same update loop as they were created in.
This method creates a promise that resolves after the specified amount of time in seconds has passed. Time is calculated as the sum of the delta values passed into PromiseTimer.Update
IPromise LogAfterFourSeconds()
{
return promiseTimer.WaitFor(4f)
.Then(() => Console.Log("4 seconds have passed!"));
}
WaitUntil takes a predicate to check each update and resolves once the predicate returns true. This predicate function is passed a TimeData
object, which just contains the most recent frame's deltaTime
and elapsedTime
which is the total amount of time since the promise was created.
IPromise FadeOut(float duration)
{
return promiseTimer.WaitUntil(timeData =>
{
// Here we are using the amount of elapsed time to calculate what the current
// fade value should be (between 0 and 1).
// Since we're fading our we should be going from 0 (not faded) to 1 (full)
var fadeAmount = Mathf.Clamp01(timeData.elapsedTime / duration);
SetFadeValue(fadeAmount);
// Resolve the promsie once we've finished.
return fadeAmount >= 1f;
});
}
WaitWhile is exactly the same as WaitUntil except that it resolves when its predicate function returns false. Think of WaitUntil as running until its predicate returns true, and WaitWhile as running while its predicate returns true, stopping when it is false.
TimeData is passed to you as a paramter when using either PromiseTimer.WaitUntil or PromiseTimer.WaitWhile. It contains the following public fields:
All
function.