Written in Racket

:: Racket

This is an overview of things I’ve created using Racket. Two motivations for writing this now:

  1. Over the last week I was at three conferences (whew!) where, when meeting or catching up with someone, I had to explain what I’ve been doing. I mentioned my current projects or projects I guessed they’d relate to. But that’s not necessarily representative of all that I’ve been able to do with Racket. I wish I’d been able to give a better overview. I have quite a few repos on GitHub, but that’s just a big list with no structure.

  2. In about a week I start my batch at Hacker School. I’ll likely spend less time with Racket, because the whole point is to learn new things. Now is a good time to take inventory. And I’ll be better prepared to talk about Racket there.

As a result, here’s an inventory, grouped into categories.

Web services

I wrote a wrapper for most of the Amazon AWS web services. Working with AWS directly at the HTTP level is really fun and educational. Well. It’s fun when writing a wrapper library. Not so fun when writing an app. Hence a wrapper library.

To support aws, I made a small http library to help with things like 100-continue responses, chunked transfer decoding, and so on.

I made an interface to Google API Discovery Service. This was fun because I didn’t hand write wrapper code. Instead, I query this Google web service that describes their web services. Racket macros use that response data to create wrapper functions. I even generate Scribble documentation.

More generally, webapi-markdown is a way I came up with to do “literate” (in the sense of “literate programming”) web API specifications. A web service is both documented and specified using a markdown file containing parameterized HTTP request and response templates. wffi (think “web foreign function interface”) is an implementation for Racket.

When I was grumpy with Google for killing Google Reader, I wrote feeds2gmail, which is similar to the venerable rss2email, but:

  • Creates emails directly in your mailbox using IMAP APPEND (faster and no delivery issues).

  • Uses two IMAP mailboxes — Gmail “labels” — to be a bit closer to the GReader experience.

Static blogging and markdown parsing

My most-starred project on GitHub is Frog. Making a static blog generator isn’t very special. But Frog is special among my Racket projects because it’s an application, not a library or programming tool. I wanted something that was much easier and simpler to install and use than Octopress. To the extent that succeeded, it’s largely due to the choice of using Racket.

As I blogged about, the hardest part of that project was parsing markdown. I ended up using a monadic parser combinator approach, very similar to how Pandoc does this. (Later I used the same approach to whip up a toml parser.)

A racket mode for Emacs

Although a big chunk of racket-mode is obviously Emacs Elisp, most of the more challenging and interesting bits are actually written in Racket. Such as:

  • Given a symbol, how do you find the definition site? How do you extract its “signature” and/or contract and/or Typed Racket type? How do you find its installed documentation, if any?

  • How do you let people write programs using racket/gui, but not make everyone pay the cost of loading it and popping up a racket frame window all the time? In other words, how do you do an on-demand, one-time initializiation of racket/gui?

  • And so on.

I hope I have time to blog more about this.

Some Clojure idioms for Racket

I have a few “armchair” languages. I’ve read a lot about them, read quite a bit of code, but not yet written much myself. At the moment these include Haskell and Clojure. During my time at Hacker School I expect to move these out of the armchair category.

Meanwhile, I learned enough about Clojure to like a few of its ideas or idioms and want to use them in Racket. The result, with contributions from some other people, is rackjure.

A few features — such as applicative dictionaries, curly brace dictonary literals, and function literals — must be used as a Racket language: #lang rackjure. The remaining features can be required as needed.

Things not on GitHub

I did a few things that I didn’t put on GitHub. Yeah, I know. Code or it didn’t happen. Regardless….

Signal processing

I experimented with signal processing combinators (including but not limited to audio signal processing).

Why isn’t this on GitHub?

  1. It’s just proof-of-concept, playing with signals being functions from time to a value. (Pleasant surprise: This turned out to be much more performant for audio processing than I expected.)

  2. Eventually I realized I was starting to reinvent aspects of FRP; maybe I should back up and learn from others.

  3. It was a chance for me to try Typed Racket. With great results: Typed Racket made my code less buggy and faster, both. I could specify specific vs. generic signals — e.g. (Signal Float) vs. (Signal a).

  4. My background is the music tech industry. I’m leery of people’s expectations. (“Isn’t it cute, the monkey thinks it can write professional audio DSP code.”)

Web app middleware framework

“Lob” is a low-level library for web applications, in the spirit of Python’s WSGI, Ruby’s Rack, and Clojure’s Ring. Like those, Lob is intended to be the glue between a web server and a higher-level web application framework.

Although I was pretty happy with how this turned out, I wasn’t using it actively in a real world project. It’s one thing to dogfood a project and support other people using it for similar use cases. But it’s another thing to, well, not do that. So… not on GitHub.

Writing about Racket

I wrote Fear of Macros, which is a sort of autobiographical tutorial for Racket macros. Although it’s probably not the best introduction for everyone, it might help people like me.1

I’ve written some blog posts.

I’ve tried to pull my weight on the Racket user mailing list and on Stack Overflow — it’s fun as you gradually gain confidence to help others the way you were helped. I find it’s also a great test whether I really understand something as well as I imagined.

All other

I think that sums up nearly everything I’ve made into a repo on GitHub, or even an actual Racket package for others to use.

Of course I’ve tried many more things in Racket, that I have sitting around here and there. Web crawling and scraping. Alternative definition forms. And more. All have been fun to do, and most have taught me more about Racket and/or programming in general.

  1. Macros might be one of those subjects — like continuations or monads? — where people need to read a few introductions until one “clicks” for them. I suppose the common theme among these things is that they tend to get mythologized and jargonized. Some of us need to take a deep breath, see how the underlying mechanism is actually quite simple, and build up from there.