Configuring Vim for SICP

Posted Aug 26, 2014

For the past few months, I’ve been chipping away at Structure and Interpretation of Computer Programs, MIT’s classic textbook on the principles of computer programming. It’s not always an easy read, but it is rewarding you’re willing to put the time in, and especially if you work through the exercises to reinforce your understanding. Having studied maths to a reasonable level, and having done a reasonable amount of practical programming, I like the way the book makes explicit the links between these disciplines.

However, other people more eloquent than I have written about the virtues of SICP and why reading it and completing the exercises is a worthwhile use of your time. Assuming then, that you’ve already decided to start your own SICP quest, I hope maybe I can save some fellow Vim users a little time by describing how I’ve setup my environment to make working through the exercises as easy as possible. You do use Vim, right?

In a nutshell, I’m going to describe where to get SICP, which Scheme interpreter to use, and how to send snippets of code from Vim to a REPL running in a different tmux pane using tslime. For what it’s worth, I’m using OS X 10.9.4, but I think things should work equally well on Linux. Windows folks, good luck!

Getting SICP

So first of all, you’re going to need a copy of SICP, and you’re in luck, because it’s not too hard to find. It’s still in print, so if you want a paper copy you should be able to pick one up on Amazon, or directly from MIT Press.

Alternatively, if an ebook is more to your liking, the full text is available for free on MIT’s support page for the book. However, I would highly recommend Andres Raba’s unofficial ebook version, which updates the original with new typesetting, nice fonts, and SVG diagrams. It’s available in EPUB3 and PDF vserions here, or online in HTML5 here. The HTML5 version is a very lovely thing indeed, and it’s this version I’ve been working from, despite owning a copy of the dead tree version.

Choosing a Scheme implementation

To get the most out of SICP, you’re going to need to do the exercises, and to do the exercises, you’re going to need a Scheme interpreter. Scheme being a variant of the Lisp programming language invented by Gerald Sussman (along with Guy Steele), one of the authors of SICP. There are several options available:


I will go ahead and spoil the surprise, and recommend that you download and install Racket. It’s easy to install, includes a nice REPL with good readline support, and includes a third-package to support users working through SICP. The support package adds procedures like inc and dec, and variables like true, false, and nil, which are assumed by the book, but not actually included by default in most Scheme implementations. None of these are too difficult to add yourself, but it’s nice to not have to.

Once you’ve downloaded and installed Racket, add the application bin directory to your path, and then start the REPL with the following command.

racket -i -p neil/sicp -l xrepl

-i enables interactive mode. That is, it tells Racket to start a REPL and wait for input.

-p neil/sicp enables use of the SICP package, downloading it from the Racket PLaneT package repository if it is not already installed. I get an error relating to internationalisation when this package is first installed, but it doesn’t seem to make any difference so far as I can tell. The package works fine and there is no error on subsequent invocations.

Finally, -l xrepl, enables Racket’s extended REPL mode, which amongst other things, enables readline support. This means you get command history, tab completion, and emacs/bash style key bindings.

If Racket is not to your taste for any reason, you have a few other options available, which I have described below in brief.

Petite Chez Scheme

Petite Chez Scheme is another good option for SICP, as it is fast, similarly easy to install as Racket, and has good readline support in the REPL. In many ways the REPL is nicer, with better indentation and visual highlighting of matched parentheses. The only thing that bugs me about it is the way to command line history works on multiline expressions, where it shows the first line of an expression when navigating backwards through the history, but all lines when navigating forwards. It’s possible this is configurable through .inputrc, but I have yet to find the appropriate option. If you can’t use Racket for any reason, Petite Chez Scheme is a decent alternative.

The rest

MIT Scheme was the first interpreter I tried, figuring that it would be the best option for working on exercises in an MIT textbook. I believe it’s the version of Scheme Abelman and Sussman used in their lectures. However, while there’s nothing especially wrong with it, there’s not much to recommend it either. At least in the homebrew version I’m using, I was unable to compile in the readline support I wanted, so I promptly moved on to something else.

I also tried Chicken Scheme, and GNU Guile, but again, I didn’t really find anything that made them stand out in comparison to Racket. I don’t want to be dismissive, and it may well be that these interpreters have strengths and weaknesses which are beyond the scope of my limited experience with Scheme. For my narrow focus, however, as a beginner starting out with SICP, I found Racket to be the easiest option to get up and running quickly.


You need to use tmux for this guide to work, but beyond that it doesn’t really matter much how you configure it. Minimally, you should familiarise yourself with the ideas of sessions, windows, and panes, and the associated key bindings to switch between them. This is a decent overview for beginners if you’re looking to learn.

I tend to use two panes side by side in a window with Vim running in one and the REPL running in another, but you could just as easily put them in separate windows or different sessions altogether.

Vim configuration

Vim has pretty good support for Scheme out of the box. The syntax highlighting works, and the automatic indentation is not too bad. Not too bad, but not quite perfect. There are a few edge cases where it gets caught out and fails to indent a line correctly according to common convention for Lisp files. Dorai Sitaram has an excellent article describing the problem, and a nice workaround. He provides a small Racket script, scmindent.rkt, that can be used to supplant Vim’s default indentation.

Download the script and put it somewhere in your path (I use ~/bin), then add the following to your .vimrc:

autocmd filetype lisp,scheme,art setlocal equalprg=scmindent.rkt

The script is configurable through the presence of a file called .lispwords in your home directory. I add (3 if) to that file to make the if conditional’s then and else expressions align with the predicate.

Parentheses matching and highlighting

There’s no escaping that Scheme programming involves typing a lot of parentheses. Unless you are careful, it can be easy to get lost and lose track of which opening and closing parentheses correspond to one another. Correct indentation helps, but there are also a couple of vim plugins that make Scheme files easier to work with.

The first of these is a rainbow parentheses plugin, which colourises matching pairs of parentheses, brackets, and braces so it’s easier to tell immediately which pairs correspond. There are a few versions of this plugin that do more or less the same thing, but the one I’m using is vim-niji. It works pretty well out of the box, but I change the default colours to something slightly more psychedelic for better contrast. You can do the same with something like this in your .vimrc:

let g:niji_dark_colours = [
    \ [ '81', '#5fd7ff'],
    \ [ '99', '#875fff'],
    \ [ '1',  '#dc322f'],
    \ [ '76', '#5fd700'],
    \ [ '3',  '#b58900'],
    \ [ '2',  '#859900'],
    \ [ '6',  '#2aa198'],
    \ [ '4',  '#268bd2'],
    \ ]¬

The second plugin is paredit.vim, which keeps matched characters (such as parentheses, quotes, braces, etc) balanced by automatically adding them in pairs. When you type (, Paredit will automatically add a ) and position the cursor between the pair ready for further input. This is, at least, the most obvious effect for the beginning Scheme programmer.

You might find it a little strange at first because it prevents you from deleting characters that would unbalance an expression, but in a way, this is Paredit’s core feature in disguise. Paredit recognises the nested list structure of Scheme (and other Lisp-ish) programs, and allows you to manipulate that structure directly rather than treating it as freeform text. For example, you can join adjacent lists, or splice the contents of a list into a parent list. It’s powerful, but you don’t need to understand all of the features to make use of it straight away, and the documentation is good when you’re ready to learn more.


The last piece of the puzzle is tslime, a simple but clever plugin that allows you to send snippets of text from a Vim buffer into a separate tmux pane. It’s perfect for working on Scheme code in Vim, because you can quickly highlight and dispatch sections for evaluation in a REPL running in a separate window. As I mentioned, I like to open two panes side by side with Vim in one and my REPL in the other. I find this is the best way to get immediate feedback on a piece of code without needing to shift focus significantly by switching windows or anything like that. It’s flexible though. You can use any tmux configuration you like, and it works just fine if you’re using a non-terminal Vim like gvim or MacVim.

There are a few version of tslime floating around, but I would I recommend using Steve Losh’s version, which has the best features and by far the best documentation. By default, it uses in normal or visual mode to send text to a tmux pane, but I map this to <localleader>t, which is \t for my setup. Using ctrl-C just doesn’t feel right to me given that it usually kills shell programs. Whatever key bindngs you choose, the first time you use it in a session, you will be prompted for the tmux session, window, and pane you’d like to target. You can hit tab for each prompt to get a list of possible values.

In visual mode tslime sends the currently selected text into the specified tmux pane. In normal mode, it sends the current paragraph, i.e. the contents of vip. When using tslime with a REPL, you will likely also want to set g:tslime_ensure_trailing_newlines = 1 to ensure that at least one newline is present at the end of any text sent to tmux. Without this, code will be sent to the REPL, but not executed until you switch to the REPL pane and hit enter.

For reference, here’s the tslime section from my .vimrc:

" tslime {{{
let g:tslime_ensure_trailing_newlines = 1
let g:tslime_normal_mapping = '<localleader>t'
let g:tslime_visual_mapping = '<localleader>t'
let g:tslime_vars_mapping = '<localleader>T'
" }}}

Putting it all together

So where am I going with this? The screencast below gives a simple example of it all in action:

  • Starting tmux and splitting the window into two panes.
  • Starting the Racket REPL in the right pane, then Vim in the left pane.
  • Writing a simple factorial function demonstrating Niji and Paredit.
  • Reformatting the code with scmindent.rkt
  • Sending the code to the REPL with tslime, using tab completion to select the session, window, and pane.
  • Calling the function to demonstrate that it works.


And that’s that. I hope some of this might be useful to others working on SICP, or thinking of doing so. If you’ve got any questions about my setup, or suggestions as to how it could be improved, I’d love to hear from you. Send me an email, or hit me up on Twitter.