Sebastian Gebski

I have a weak spot for the policy design pattern since ol' good C++ times: I think it has all started when I've read one of Sutter's / Alexandrescu's books with some great examples of policies injected via template parameters. This way you could get a variable behavior without complex inheritance and excessive coupling. Today, the times of templates are long gone (trust me: C# generics are 5% of what was possible with C++ templates ...), but it doesn't mean you can't use strategy / policy design patterns anymore.

Back to business:

The star of the day is named "Polly".

Polly is an exception handling policy library for .NET (3.5+; 4.0+) with a sweet (due to a proper dose of syntactic sugar), fluent syntax. What does it really mean?

  1. You declaratively describe the behavior related to particular exception "class" handling. By class I mean subset of exceptions that meet some of the following criteria (1 or more):

    • of a given type
    • with an exception property that meets a given condition
  2. Behavior may mean:

  3. Such a policy may be applied to one or more lambdas of your choice. In case of applying a retry, you're the one to ensure that a re-execution of lambda make sense.

What's the benefit?

Basically: a very clean & maintainable code. Especially if you intend to re-use a particulr policy (same behavior) in many situations - you're abstracting (encapsulating & isolatings) your error-handling policies in a way that doesn't interfere with the actual business logic. Neat.

What is more, you're hiding the painful context (retrying, circuit breaking for a given time) - now it's hidden within a poliy, handled automatically & adjusting the behavior is as easy as it could only be.

If you want to have a look at the actual samples of code, just follow the link I gave above: The code samples on the official GitHub page are much more than sufficient to present the capabilities of the library. Here's a sample, just to prove that the well-thought, fluent syntax doesn't need additional comments:

// first, create a policy that handles 2 types of exceptions &
// retries (whatever's to be retried) few times
var policy = Policy.Handle<SampleEx>(ex => true == ex.NeedsToBeHandled)  
                   .WaitAndRetry(new[] {
                   }, (ex, timeSpan, ctx) => {
                      // do something after each retry, like logging   

// now, execute the policy with your action & some neat context
    () => DoYourAction(),
    new Dictionary<string, object>() {
        { "ctxKeySample", "ctxValueSample" }