Nuclex.Support/Source/Tracking/FailableProgression.cs

110 lines
4.4 KiB
C#

using System;
using System.Collections.Generic;
namespace Nuclex.Support.Tracking {
/// <summary>Extended type of progression that is able to fail</summary>
/// <remarks>
/// <para>
/// If the background process fails, the exception that caused it to fail is
/// communicated to all parties waiting on the progression through the
/// Exception property. Implementers should place their code in try..catch
/// blocks and call SetException() to temporarily store the exception for
/// retrieval by the caller(s).
/// </para>
/// <para>
/// As with all progressions, the interface contract still requires you to call
/// OnAsyncEnded(), no matter what the outcome of your background operation is.
/// </para>
/// </remarks>
public class FailableProgression : Progression {
#region class EndedDummyProgression
/// <summary>Dummy progression that is always in the ended state</summary>
private class EndedDummyProgression : FailableProgression {
/// <summary>Creates a new successfully completed dummy progression</summary>
public EndedDummyProgression() : this(null) { }
/// <summary>Creates a new failed dummy progression</summary>
/// <param name="exception">Exception that caused the dummy to fail</param>
public EndedDummyProgression(Exception exception) {
OnAsyncEnded();
// Only call SetException() if an actual exception was provided. Who knows what
// evil code might be inside SetException() after all ;)
if(exception != null)
SetException(exception);
}
}
#endregion // EndedDummyProgression
/// <summary>Creates a new failed dummy progression</summary>
/// <param name="error">
/// Exception that supposedly caused the progression to fail
/// </param>
/// <returns>
/// A failed progression that reports the provided exception as cause for its failure
/// </returns>
public static FailableProgression CreateFailedDummyProgression(Exception exception) {
return new EndedDummyProgression(exception);
}
/// <summary>Waits for the background operation to end</summary>
/// <remarks>
/// Any exceptions raised in the background operation will be thrown
/// in this method. If you decide to override this method, you should
/// call End() first (and let any possible exception through to your
/// caller).
/// </remarks>
public virtual void Join() {
// By design, end can only be called once!
lock(this) {
if(this.endCalled)
throw new InvalidOperationException("End() has already been called");
this.endCalled = true;
}
// If the progression itself hasn't ended yet, block the caller until it has.
if(!Ended)
WaitHandle.WaitOne();
// If an exception occured during the background execution
if(this.occuredException != null)
throw this.occuredException;
}
/// <summary>Exception that occured while the operation was executing</summary>
/// <remarks>
/// If this field is null, it is assumed that no exception has occured
/// in the background process. If it is set, however, the End() method will
/// re-raise the exception to the calling thread when it is called.
/// </remarks>
public Exception OccuredException {
get { return this.occuredException; }
}
/// <summary>Sets the exception to raise to the caller of the End() method</summary>
/// <param name="exception">Exception to raise to the caller of the End() method</param>
protected void SetException(Exception exception) {
// We allow the caller to set the exception multiple times. While I certainly
// can't think of a scenario where this would happen, throwing an exception
// in that case seems worse. The caller might just be executing an exception
// handling block and locking + throwing here could cause all kinds of problems.
this.occuredException = exception;
}
/// <summary>Exception that occured while the operation was executing</summary>
private volatile Exception occuredException;
/// <summary>Whether the End() method has been called already</summary>
private volatile bool endCalled;
}
} // namespace Nuclex.Support.Tracking