Notes by a software developer

Walking through a basic Racket web service

December 29, 2016 by Phil Eaton


Racket is an impressive language and ecosystem. Compared to Python, Racket (an evolution of Scheme R5RS) is three years younger. It is as concise and expressive as Python but with much more reasonable syntax and semantics. Racket is also faster in many cases due in part to:

Furthermore, the built-in web server libraries and database drivers for MySQL and PostgreSQL are fully asynchronous. This last bit drove me here from Play / Akka. (But strong reservations about the complexity of Scala and the ugliness of Play in Java helped too.)

With this motivation in mind, I'm going to break down the simple web service example provided in the Racket manuals. If you don't see the following code in the linked page immediately, scroll down a bit.

      
#lang web-server

(require web-server/http)

(provide interface-version stuffer start)

(define interface-version 'stateless)

(define stuffer
  (stuffer-chain
   serialize-stuffer
   (md5-stuffer (build-path (find-system-path 'home-dir) ".urls"))))

(define (start req)
  (response/xexpr
   `(html (body (h2 "Look ma, no state!")))))
      
    

First we notice the #lang declaration. Racket libraries love to make new "languages". These languages can include some entirely new syntax (like the Algol language implementation) or can simply include a summary collection of libraries and alternative program entrypoints (such as this web-server language provides). So the first thing we'll do to really understand this code is to throw out the custom language. And while we're at it, we'll throw out all typical imports provided by the default racket language and use the racket/base language instead. This will help us get a better understanding of the Racket libraries and the functions we're using from these libraries.

While we're throwing the language away, we notice the paragraphs just below that original example in the manual. It mentions that the web-server language also imports a bunch of modules. We can discover which of these modules we actually need by searching in the Racket manual for functions we've used. For instance, searching for "response/xexpr" tells us it's in the web-server/http/xexpr module. We'll import the modules we need using the "prefix-in" form to make function-module connections explicit.

      
#lang racket/base

(require (prefix-in xexpr: web-server/http/xexpr)
         (prefix-in hash: web-server/stuffers/hash)
         (prefix-in stuffer: web-server/stuffers/stuffer)
         (prefix-in serialize: web-server/stuffers/serialize))

(provide interface-version stuffer start)

(define interface-version 'stateless)

(define stuffer
  (stuffer:stuffer-chain
   serialize:serialize-stuffer
   (hash:md5-stuffer (build-path (find-system-path 'home-dir) ".urls"))))

(define (start req)
  (xexpr:response/xexpr
   `(html (body (h2 "Look ma, no state!")))))
      
    

Now we've got something that is a little less magical. We can run this file by calling it: "racket server.rkt". But nothing happens. This is because the web-server language would start the service itself using the exported variables we provided. So we're going to have to figure out what underlying function calls "start" and call it ourselves. Unfortunately searching for "start" in the manual search field yields nothing relevant. So we Google "racket web server start". Down the page on the second search result we notice an example using the serve/servlet function to register the start function. This is our in.

      
#lang racket/base

(require (prefix-in xexpr: web-server/http/xexpr)
         (prefix-in hash: web-server/stuffers/hash)
         (prefix-in stuffer: web-server/stuffers/stuffer)
         (prefix-in serialize: web-server/stuffers/serialize)
         (prefix-in servlet-env: web-server/servlet-env))

(provide interface-version stuffer start)

(define interface-version 'stateless)

(define stuffer
  (stuffer:stuffer-chain
   serialize:serialize-stuffer
   (hash:md5-stuffer (build-path (find-system-path 'home-dir) ".urls"))))

(define (start req)
  (xexpr:response/xexpr
   `(html (body (h2 "Look ma, no state!")))))

(servlet-env:serve/servlet start)
      
    

Run this version and it works! We are directed to a browser with our HTML. But we should clean this code up a bit. We no longer need to export anything so we'll drop the provide line. We aren't even using the interface-version and stuffer code. Things seem to be fine without them, so we'll drop those too. Also, looking at the serve/servlet documentation we notice some other nice arguments we can tack on.

      
#lang racket/base

(require (prefix-in xexpr: web-server/http/xexpr)
         (prefix-in servlet-env: web-server/servlet-env))

(define (start req)
  (xexpr:response/xexpr
   `(html (body (h2 "Look ma, no state!")))))

(servlet-env:serve/servlet
 start
 #:servlet-path "/"
 #:servlet-regexp rx""
 #:stateless? #t)
      
    

Ah, that's much cleaner. When you run this code, you will no longer be directed to the /servlets/standalone.rkt path but to the site root -- set by the #:servlet-path optional variable. Also, every other path you try to reach such as /foobar will successfully map to the start function -- set by the #:servlet-regexp optional variable. Finally, we also found the configuration to set the servlet stateless -- set by the optional variable #:stateless?.

But this is missing two things we could really use out of a simple web service. The first is routing. We do that by looking up the documentation for the web-server/dispatch module. We'll use this module to define some routes -- adding a 404 route to demonstrate the usage.

      
#lang racket/base

(require (prefix-in dispatch: web-server/dispatch)
         (prefix-in xexpr: web-server/http/xexpr)
         (prefix-in servlet: web-server/servlet-env))

(define (not-found-route request)
  (xexpr:response/xexpr
   `(html (body (h2 "Uh-oh! Page not found.")))))

(define (home-route request)
  (xexpr:response/xexpr
   `(html (body (h2 "Look ma, no state!!!!!!!!!")))))

(define-values (route-dispatch route-url)
  (dispatch:dispatch-rules
   [("") home-route]
   [else not-found-route]))

(servlet:serve/servlet
 route-dispatch
 #:servlet-path "/"
 #:servlet-regexp #rx""
 #:stateless? #t)
      
    

Run this version and check out the server root. Then try any other path. Looks good. The final missing piece to this simple web service is logging. Thankfully, the web-server/dispatch-log module has us covered with some request formatting functions. So we'll wrap the route-dispatch function and we'll print out the formatted request.

      
#lang racket/base

(require (prefix-in dispatch: web-server/dispatch)
         (prefix-in dispatch-log: web-server/dispatchers/dispatch-log)
         (prefix-in xexpr: web-server/http/xexpr)
         (prefix-in servlet: web-server/servlet-env))

(define (not-found-route request)
  (xexpr:response/xexpr
   `(html (body (h2 "Uh-oh! Page not found.")))))

(define (home-route request)
  (xexpr:response/xexpr
   `(html (body (h2 "Look ma, no state!!!!!!!!!")))))

(define-values (route-dispatch route-url)
  (dispatch:dispatch-rules
   [("") home-route]
   [else not-found-route]))

(define (route-dispatch/log-middleware req)
  (display (dispatch-log:apache-default-format req))
  (flush-output)
  (route-dispatch req))

(servlet:serve/servlet
 route-dispatch/log-middleware
 #:servlet-path "/"
 #:servlet-regexp #rx""
 #:stateless? #t)
      
    

Run this version and notice the logs displayed for each request. Now you've got a simple web service with routing and logging! I hope this gives you a taste for how easy it is to build simple web services in Racket without downloading any third-party libraries. Database drivers and HTML template libraries are also included and similarly well-documented. In the future I hope to add an example of a slightly more advanced web service.

Aside: I have had huge difficulty discovering the source of Racket libraries. These library sources are nearly impossible to Google and search on Github is insane. Best scenario, the official racket.org docs would link directly to the source of a function when the function is documented. Of course I could just download the Racket source and start grepping... but I'm only so interested.