Sild is a lisp dialect

June 5, 2016

Today, I’m releasing Sild, a tiny little intepreted lisp that I wrote in C.

I’ve been interested in trying to learn about language design and implementation for a while now. I’ve also been interested in Lisp as a concept, and I had also been wanting to learn C so that I could start wrapping my head around systems programming. This project brought all of those objectives together in a really natural way!

Sild is not conformant to any existing spec. It’s not really a Scheme and it is definitely not a version of Common Lisp. It is simply my attempt to build a minimal lispy language in a semi-vacuum from first principles.

Though I’m proud of the final result, that’s not what I’m most excited about sharing. I also wrote blog posts as I went along. A lot of them. Just about 40,000 words’ worth! I’m going to post them here in the coming weeks. It’s my hope that other people might be able to learn some of the things I learned!

I’ve tried to write those posts so that they will be understandable to someone who can program in some language, and who maybe knows a little about C, and maybe knows a little about lisp.

I am open to suggestions and criticisms. The posts might seem like they are written from a position of authority, but do not be fooled. I was figuring everything out as I went along, then writing them down just after I had figured them out, and I most certainly did not nail everything! I am very interested to hear where what I tried to do diverged from historical attempts, or where I reinvented some wheel or another.


I was inspired to try this project by a variety of things.

At some point during a long flight I tried coding most of this post in Ruby. I don’t remember what I did with that repo, and I don’t think it even really worked all the way, but it gave me the idea that a lisp might be a lot easier to implement than I thought.

This is a very sticky essay. Graham’s proceduralness possesses a certain kind of clarity that left me with the impression that it might be worth trying to write the simple lisp he describes, and indeed that’s pretty much what I ended up with. Sild, as it is right now, has no types (like strings or numbers), and has no I/O save a basic display that prints to stdout. It has no real standard library to speak of, either, but it does have the basic operations that Graham describes as necessary to implement eval.

Google “build your own lisp”, and you’ll get an entire first page of results referencing this site. I gave it a good go, making it to chapter 9 or so, and I have to say that it is incredibly well thought out and well expressed and I totally learned a lot from it! But somewhere along the way I started feeling like I was just typing things in until they worked without really understanding why, which is my personal canary in the coal mine for impending this-isn’t-working-for-me-anymore-ness. Then I read the URL and realized what appealed to me so much about the site is the your in “build your own lisp”, so I stopped working through the tutorial and decided to try that for real! From there, I circled back to the Paul Graham essay and used it as a guide.


There were a variety of other things I ran across while writing this language that I read parts of, with the intention of revisiting them with a more context after I was done with the project.

I had “read” this a few times, but always with the feeling I wasn’t really getting it. Implementing what amounts to the lisp described in this paper (via the Paul Graham essay), has made this seminal work a lot easier to comprehend!

I consulted this a few times when I ran into specific questions, but again, I didn’t really grok it until after I had finished the whole thing. The finer points of closures and mutability and garbage collection were all lost on me until I banged my own head against them all for a while. The first time I read this paper it was a “I know these words” sort of situation. Now, it’s more of a “YES OMG YES SO TRUE” kind of deal.

Again, I had read a lot of this before, and had gotten real value out of it. But now when I go back, it just reads a lot richer and more nuanced, and I find myself considering intricacies that simply wouldn’t have occured to me before.

Thanks

Big thanks to Andrew Kelley, whose brilliant code reviews levelled up my C on multiple occasions. Also to Darius Bacon, whose feedback had a fantastic twinge of “I have trod where you tread” that helped me appreciate all the things I must have missed. I look forward to finding them out in more detail!

Also to Pam Selle and Nick Drozd for reading early posts, and Bert Muthalany for his enthusiasm.


Working on this language was immensely rewarding, and I learned an enormous amount from it. While I intend to work on it more, adding some of the TODO’s in the README, and hope to end up with a useful scripting language all to myself, the original motivation of the project has been satisfied. I hope you find it as interesting as I did!