How to Write a Lisp Interpreter in Python (2010) | Hacker News Reader

How to Write a Lisp Interpreter in Python (2010)

norvig.com

130 pointsxorbox28 comments

I'm always impressed how the magic of interpreters and compilers can be explained so succinctly.

Inspired by Jonesforth, I wrote a complete Lisp interpreter in a single ARM assembly file: https://github.com/marcpaq/arpilisp

How long did it take you? The documentation is A+ (it reminds me of this compiler: https://github.com/thejameskyle/the-super-tiny-compiler ).

I estimate it took 60-70 hours, of which 2/3 of that was for the documentation.

Thanks for the compliment and the link to TSTC. I'll add it to my personal list.

Very good quality congratulations!

While quickly reading through it, my first idea was getting the PI to boot directly to it. :)

This is brilliant!

I've failed more than once to wrap my stupid brain around lisp but after reading your code, something clicked. I dare say I think I get it! https://github.com/marcpaq/arpilisp/blob/master/arpilisp.s

It's all extraordinarily clean and clear (and must set a new world record for most thorough comment --1120 lines!)

You are my new hero.

I've become interested in Lisp every couple of years. This fall I even learned Clojure and implemented a Lisp interpreter in Python. I searched a lot for any open-source Lisp implementation in assembly, but found none. Here it is!

I thought a lot about bootstrapping, too, since apparently almost every compiler is written in C these days. Somewhere I got the idea that in the 1970s someone wrote a compiler in assembly and that was used to bootstrap the first C compiler. Guess I'm a romantic, but for me it's important that we know how to get where we are now. Having a high-level language implemented in assembly is quite essential to that, IMO.

I think there's a separate Lisp lineage anyway, wherein someone in the 1960s wrote a minimal machine-code interpreter, and ever since lispers have been writing interpreters on top of other Lisps and revising them to emit compiled code along the way. We would still have Lisp compilers if C never existed.

Obligatory link to today's best solution, using the Python AST and running straight off bytecode: http://hylang.org

Also, https://github.com/pixie-lang/pixie

I've always found Hy interesting, but cannot picture anyone using it in production. Are there any real live examples?

The maintainers quoted having a few customers using it, but here's my stuff: http://github.com/rcarmo/sushy

That code is more readable than I expected. :) May you share more about the experience of using it? Do you keep a blog?

Erm. My blog runs on it, it's the demo site :)

Well, ok. :)

I tweeted the repo link to the HyLang account.

Original HN Submission + discussion here:

https://news.ycombinator.com/item?id=1745322

(and many re-submissions after that)

Follow up:

https://news.ycombinator.com/item?id=1746916

For anyone wanting to try something similar yourself, I highly recommend MAL. It's a Lisp dialect designed to be a learning tool: there's a nice guide for implementing the interpreter. It doesn't take much away from the learning, but it really lessens the possible frustration a lot. https://github.com/kanaka/mal/blob/master/process/guide.md

Just in case anyone is interested, here's my implementation in Python 3.5+. I'm guessing it took me about 15 hours to write. https://github.com/dancek/mal/tree/python3-dancek/python3-da...

Of course, Peter Norvig's followup to the linked article may also interest anyone who read this far: http://norvig.com/lispy2.html

Is LISP still relevant? ;-)

I'm curious whether folks think the ideas of LISP are still resonant and powerful in the startup world and wider industry.

This is different from the question of inspiration. Yes, I admire.

Does PG, for instance, still think that startups can have 'secret' technical advantages and use strategies like 'running upstairs', i.e. choosing to do hard things as a competitive advantage.

By ideas of LISP I mean things like programs-as-data and a small kernel from which the system is bootstrapped.

He he, I wrote a (primitive) Lisp interpreter in Python once. It was crappy and really slow, but it had lexical scoping and a (naive) macro system. Fun times... ;-)

Along the same lines may I say buildyourownlisp.com somehow manages to be a great intro text to C as well as lisp. Highly recommended

Here is my version of a lisp interpreter in python:

https://github.com/keithgabryelski/plisp

it's not meant to be pythonic -- it's meant to readable and organized

my inspiration was Interpreting LISP by Gary D. Knott written in 1997 (the pdf of it is included in the repository)

This is currently the final project of CS61A, the introductory CS course taught at UC Berkeley.

https://inst.eecs.berkeley.edu/~cs61a/fa14/proj/scheme/ (no link for the current semester is up yet)

Surely something more like

(how (in 'python (write (interpreter 'lisp))

:-P

More like:

  (howp (write (interpreter :of :lisp
                            :in :python))))

'social-problems-of-lisp

> howp

“how?” is not a predicate (i.e. something that returns a Boolean value), so it wouldn't be named howp.

Somebody read the Jargon File...

Please read the post title again.

A minimal version would be more like:

# REPL

def interpret_lisp():

    while not control_C_pressed():

        print(eval(read()))

where those functions would have to be defined, of course.

This is from 2010. Please mark in the title.

Reply on news.ycombinator.com