What's That Racket?

For my third language of the month, I figured I would cover a language of venerable heritage: Lisp. Part of Lisp's fame centers around the supposed enlightenment one achieves when one finally "gets" Lisp.

Now, the trick with learning Lisp is first you need to pick one. There is more than one flavor of Lisp around, and even among the various flavors, there is a cornucopia of implementations. The biggest "flavors" of Lisp around are Common Lisp, Scheme, and Clojure. Clojure's claim to fame is that it's backed by the Java Virtual Machine, and Common Lisp and Scheme are two sides of the same coin. Common Lisp has more of a "kitchen sink" mentality, which resulted in a large language; whereas Scheme has much more minimalistic ideals ^{Another interesting difference is whether functions belong in a namespace all their own or in the same namespace as variables; this is what's known as the Lisp-1 vs Lisp-2 distinction.} Another interesting difference is whether functions belong in a namespace all their own or in the same namespace as variables; this is what's known as the Lisp-1 vs Lisp-2 distinction. . In fact, Scheme is small enough that creating an implementation is a weekend project for some.

Scheme's "functions and variables in the same namespace" choice, in addition to its minimalism, appeals to me (I know, ironic, given my love for Perl!), so I decided that learning a Scheme would be most appealing. Now...which Scheme?

Since I wanted to be productive in my new language without having to hunt down libraries, let alone figure out how to install them, I figured I would pick a Scheme that has the batteries included. A dialect of Scheme that I've known of for some time came to mind: Racket.

Learning the Language

Racket's documentation page has a few tutorials on it; I rather like the "quick" intro to Racket, as it introduces you to a picture data type that you can play with and operate on in DrRacket, the Racket IDE. This introduction to Racket is a lot more visual than other languages' intros, which is defintely a point in Racket's favor for teaching beginners.

I've heard that Structure and Interpretation of Computer Programs (SICP) is a must read for not only wannabe Lispers, but any programmer, but I'm afraid I didn't find the time to really sink my teeth into it; I think I'll be returning to it and Lisp in the future.

In my search for learning materials, I tried to find blog posts by other developers about Racket, but I came up short. I'm interested in following up with Racket in the future, so if you know of some good posts or blogs about Racket, please let me know!

Pros of Lisp, and Racket in Particular

Destructuring bind via match

In the program I wrote for this month, I dealt with two-dimensional coordinates a lot, which I simply represented as a two-element list. So getting at the component coordinates can be done like this:

(define (my-coordinate-function coords)
  (let [(row (first coords))
        (column (second coords))]
    (do-stuff x y)))

However, manually deconstructing a list via (first coords) and (second coords) ^{If you're familiar with Lisp, you may wonder what's up with first/second instead of car/cadr. Racket realizes that car and friends aren't particularly memorable, so they provide handy aliases.} If you're familiar with Lisp, you may wonder what's up with first/second instead of car/cadr. Racket realizes that car and friends aren't particularly memorable, so they provide handy aliases. to get at coordinates feels so clunky to me. Fortunately, Racket provides destructuring binding via the racket/match module. So we can rewrite the above like so:

(define (my-coordinate-function coords)
  (match coords
    [(list row column) (do-stuff row column)]))

This cleans up the code quite a bit, and has the added benefit of allowing you to specify multiple patterns, and run the associated code on the first one that matches.

Build your own syntax via macros

Let's have another look at my function from above:

(define (my-coordinate-function coords)
  (match coords
    [(list row column) (do-stuff row column)]))

Since two-dimensional coordinates were at the heart of my application, I ended up writing this same boilerplate a lot. They all looked the same, except for the function name and the do-stuff logic inside.

So what if I wanted some pretty syntax to define a function that takes a coordinate pair? This is Lisp's true strength; it allows you to define your own syntax via macros:

(define-syntax-rule (define-pair (name a b) body)
  (define (name coords)
    (match coords
      [(list a b) body])))

(define-pair (my-coordinate-function row column)
  (do-stuff row column))

It might not be immediately obvious, but macros are just Lisp functions; they're code that generates new code. It's all very meta.

This "function takes a coordinate" shorthand is useful for my program, but I can imagine many other uses for custom syntax. For example, let's say you want something like a low-level SQL interface, but you also want the composability feature of ORMs. You could easily write a SQL macro that would allow you to write code like the following:

(define (new-users-sql #:active-only [active-only #f])
  (define base-sql (sql select '(id username) from users
    (where (>= date_added '2016-01-01'))))

  (if active-only
    (sql-compose base-sql '(where (= active 1)))
    base-sql))

Another idea could be an intelligent assertion function that could detect more complex assertions and tell you why they failed, which could look something like this:

(assert (and (>= x 0)
             (>= y 0)
             (<= (+ x width) total-width)
             (<= (+ y height) total-height)))
; example failure message: (>= x 0) was false (x is -1)

A full introduction to macros is outside of the scope of this post, so I'll refer you to the excellent Fear of Macros. Racket also allows you to implement custom languages through what's called a reader; you can read more about that at Beautiful Racket.

...and much, much more!

There are a number of features in Racket that I find novel and/or useful, but sadly I don't have the space to cover them here. If I continue learning it, I may cover them in future posts. Here are some of those features if you want to look into them yourself:

• X-expressions
• Dynamic binding
• Continuations
• Contracts

(cons) of Racket

You can have any syntax you like, as long as it's sexprs

As I just demonstrated, Racket (and Lisps in general) allow you to define your own syntax; you can do anything you want, but you're restricted to using s-expressions ^{The reader facility I mentioned above allows you to get around this, but I would argue at this point, you're escaping Lisp-land.} The reader facility I mentioned above allows you to get around this, but I would argue at this point, you're escaping Lisp-land. . Lispers argue that Lisp has no real syntax and that after a while, the parentheses just fall away; I can see where they're coming from, but it seems to me that this very lack of syntax makes it kind of hard to determine what good-looking Racket looks like, and that structuring your code is very important for readability. (Then again, which language doesn't require good structure for readability?)

Language size

Going through the entire Racket reference, as I did for R and Elm, is...hard. The Racket Guide and Reference are huge! Is a much bigger language than I expected, especially considering it's a Scheme derivative.

Function names

Quite often, trying to find a function that did what I needed was a challenge. For example, I was looking for a function to create an list of length N, where each element shared the same value (known as repeat in Elm and Haskell). I looked for "repeat", I skimmed the racket/list module documentation, but came up with nothing. I ended up writing my own implementation. It was during a third or fourth pass of racket/list looking for a completely different function that I stumbled upon make-list, which does exactly what I need. That's not to mention that there's a build-list function that does something very similar, but still different, from make-list. I had other experiences, but some of this might come down to thinking about other names for functions that I know from other languages.

Enlightenment Attained?

I don't think a month learning Racket on-and-off is enough to really grok Lisp; I definitely see some of the "code is data" benefits, but I think I have much more to learn. I think I'll be returning Racket at sometime in the future.

Join me next time for a discussion of the program I wrote in Racket!