/frank

A combinator library for building web apis using System.Net.Http.

Primary LanguageF#OtherNOASSERTION

Frank

![Gitter](https://badges.gitter.im/Join Chat.svg)

Build status Build Status

Frank is resource-oriented wrapper library for working with the Web API and is developed in F#.

Goals

  1. Provide a simple, to-the-metal framework for quickly building HTTP applications and services without a lot of hassle. In this case, "to-the-metal" means utilizing the new HTTP stack coming in .NET 4.5 and available now in the HttpClient library available with Web API.
  2. Focus on composition over everything else. Based in F#, composition comes fairly naturally.
  3. Run on a number of server platforms. Frank currently uses the Web API platform to allow it to run either atop ASP.NET or self-hosted using the HttpSelfHostServer. Other hosting options will be made available soon.

Usage

Define an Application

One may define a web application interface using a large variety of signatures. Indeed, if you search the web, you're likely to find a large number of approaches. When starting with Frank, I wanted to try to find a way to define an HTTP application using pure functions and function composition. The closest I found was the following:

type HttpApplication = HttpRequestMessage -> Async<HttpResponseMessage>

let orElse left right = fun request -> Option.orElse (left request) (right request)
let inline (<|>) left right = orElse left right

The last of these was a means for merging multiple applications together into a single application. This allowed for a nice symmetry and elegance in that everything you composed would always have the same signature. Additional functions would allow you to map applications to specific methods or uri patterns.

A "Hello, world!" application using these signatures would look like the following:

let helloWorld request =
    OK ignore <| Str "Hello, world!"
    |> async.Return

A simple echo handler that returns the same input as it received might look like the following:

let echo (request: HttpRequestMessage) = async {
    let! content = request.Content.AsyncReadAsString()
    return respond HttpStatusCode.OK
           <| ``Content-Type`` "text/plain"
           <| new StringContent(content)
}

or just:

let echo (request: HttpRequestMessage) =
    OK <| ``Content-Type`` "text/plain" <| request.Content.AsyncReadAsString()

If you want to provide content negotiation, use:

let echo = runConneg formatters <| fun request -> request.Content.AsyncReadAsString()

Define an HTTP Resource

Alas, this approach works only so well. HTTP is a rich communication specification. The simplicity and elegance of a purely functional approach quickly loses the ability to communicate back options to the client. For instance, given the above, how do you return a meaningful 405 Method Not Allowed response? The HTTP specification requires that you list the allowed methods, but if you merge all the logic for selecting an application into the functions, there is no easy way to recall all the allowed methods, short of trying them all. You could require that the developer add the list of used methods, but that, too, misses the point that the application should be collecting this and helping the developer by taking care of all of the nuts and bolts items.

The next approach I tried involved using a tuple of a list of allowed HTTP methods and the application handler, which used the merged function approach described above for actually executing the application. However, once again, there are limitations. This structure accurately represents a resource, but it does not allow for multiple resources to coexist side-by-side. Another tuple of uri pattern matching expressions could wrap a list of these method * handler tuples, but at this point I realized I would be better served by using real types and added an HttpResource type to ease the type burden.

HTTP resources expose an resource handler function at a given uri. In the common MVC-style frameworks, this would roughly correspond to a Controller. Resources should represent a single entity type, and it is important to note that a Foo is not the same entity type as a Foo list, which is where the typical MVC approach goes wrong.

The 405 Method Not Allowed function allows a resource to correctly respond to messages. Therefore, we extend the HttpResource with an Invoke method. Also note that the methods will always be looked up using the latest set. This could probably be memoized so as to save a bit of time, but it allows us to ensure that all available methods are reported.

Compose Applications and Resources into Applications

A compositional approach to type mapping and handler design. Here, the actual function shows clearly that we are really using the id function to return the very same result.

let echo2Transform = id

The echo2ReadRequest maps the incoming request to a value that can be used within the actual computation, or echo2Transform in this example.

let echo2ReadRequest (request: HttpRequestMessage) =
    request.Content.AsyncReadAsString()

The echo2Respond maps the outgoing message body to an HTTP response.

let echo2Respond body =
    respond <| ``Content-Type`` "text/plain" <| body

This echo2 is the same in principle as echo above, except that the logic for the message transform deals only with the concrete types about which it cares and isn't bothered by the transformations.

let echo2 request = async {
    let! content = echo2ReadRequest request
    let body = echo2Transform content
    return echo2Respond <| new StringContent(body)
}

Create a HttpResource instance at the root of the site that responds to POST.

let resource = route "/" <| post echo2

Other combinators are available to handle other scenarios, such as:

  1. Content Negotiation
  2. Building Responses
  3. Combining applications into resources
  4. Combining resources into applications

Check the samples for more examples of these combinators.

Define a Middleware

Middlewares follow a Russian-doll model for wrapping resource handlers with additional functionality. Frank middlewares take an HttpApplication and return an HttpApplication. The Frank.Middleware module defines several, simple middlewares, such as the log middleware that intercepts logs incoming requests and the time taken to respond:

let log app = fun (request : HttpRequestMessage) -> async {
    let sw = System.Diagnostics.Stopwatch.StartNew()
    let! response = app request
    printfn "Received a %A request from %A. Responded in %i ms."
            request.Method.Method
            request.RequestUri.PathAndQuery
            sw.ElapsedMilliseconds
    sw.Reset()
    return response
}

The most likely place to insert middlewares is the outer edge of your application. However, since middlewares are themselves just HttpApplications, you can compose them into a Frank application at any level. Want to support logging only on one troublesome resource? No problem. Expose the resource as an application, wrap it in the log middleware, and insert it into the larger application as you did before.

Hosting

Frank will run on any hosting platform that supports the Web API library. To hook up your Frank application, use the register function, passing in the resources and the instance of HttpConfiguration.

register [resource] config

This extension adds a default route to your HttpConfiguration instance and adds a DelegatingHandler instance to the route's HttpConfiguration.MessageHandlers collection.

TODO

  • Loads of documentation
  • Integrate with Dyfrig

Team