v3.0.0

BREAKING CHANGES

• IDispatchTracker<TAction> is not longer taking a TAction and the service is no longer register per action but instead as a scope service IDispatchTracker.
• IDispatcher.Prepare<TAction>().DispatchAsync(CancellationToken ct = default) no longer returns Guid as it was unused artifacts of a idea of a system for cancellations.
• IDispatcher.Prepare<TAction>().DispatchAsync(bool asSafe = false, CancellationToken ct = default) no longer takes bool as first param instead use fluent api .AsSafe() to achieve the same result.

WARNING CHANGES

• IDispatcherMiddleware.OnDispatchFailure(Exception exception, object action); renamed arguments to IDispatcherMiddleware.OnDispatchFailure(Exception ex, object action); which could trigger compiler warnings.

New Features

• Dispatch now supports Cancellation tokens.
• Token are passed down to all subsequent chain until you define a different one DispatchAsync(false,new CT);.
• .DoNotAwait() is now available, design for effect sub-dispatches when specific dispatch should not be awaited in case the origin call awaits the full pipeline... some tasks should not be awaited? that's your method.
• Access Token from effects using dispatcher.CancelToken which is default or correspond to upper chain origin.
• You can check if you shouldd stop execution of effect using dispatcher.IsCancellationRequested.
• dispatcher.IsCancellationRequested include now if the anti-duplication chain that was cancelled but slipped through the cracks if any you can now catch it in the effects.
• dispatcer.AsSafe() replaces the first param IDispatcher.Prepare<TAction>().DispatchAsync(bool asSafe = false, CancellationToken ct = default) to execute the context with race condition resistance.
• dispatcer.AsUnSafe() turn off the the rc resistance, it is useful to disconnect a usually safe flow to prevent cancellation.

Performance

• Improved Dispatcher Performance up to 24% by removing class creation via reflection at every dispatch.
• Improved Dispatch Performance up to 20% by only performing allocation for the middleware class when middlewares are actually present. (Note: the second you start adding middleware slight performance drop is expected).
• Improved Reducer Performance, we initially tested every single state registered to compare if the running action had reducers... now we get reducers by Action or by State directly.

Major Fixes

• Await feature did not await the full pipeline, subsequent dispatches were set as normal fire-and-yield unless specified. Now the await feature passes down the await flag thus fully awaiting as the intended feature was design for.