4 ms·

A Lisp interpreter in TypeScript

58 points by suzuki 11y ago

suzuki 11y ago
The Lisp in 1,200 lines of Dart, which has macros, quasi-quotation and tail-call optimization (https://news.ycombinator.com/item?id=9903332 https://news.ycombinator.com/item?id=9903332), was translated into 1,390 lines of TypeScript. It runs on browsers.
lomnakkus 11y ago
Thus, yet again, sort-of-validating Greenspun's 10th Rule. I'm not complaining... just observing :).
- gadtfly 11y ago
  As much as I love a good Greenspun invocation, I don't think this really counts: > Any sufficiently complicated C or Fortran program contains an ad hoc, informally-specified, bug-ridden, slow implementation of half of Common Lisp. isn't about the inevitability of literally implementing a Lisp interpreter. It's about something more abstract. see: http://c2.com/cgi-bin/wiki?LevelsOfGreenspunning http://c2.com/cgi-bin/wiki?LevelsOfGreenspunning
  - lomnakkus 11y ago
    Mostly agreed. Which is why I said "sort-of-validating". ... so that I would be technically correct. Which, as we all know, is the best kind of correct. (Source: Futurama)
  - DonHopkins 11y ago
    It's still got a good shot at validating Zawinski's Law of Software Envelopment. > Every program attempts to expand until it can read mail. Those programs which cannot so expand are replaced by ones which can.
agumonkey 11y ago
I will now write all my comment in Kanjis. It's too beautiful.
eatonphil 11y ago
Why does the foldl function use a while loop?
- xyzzy123 11y ago
  I think it's just manual tail-call optimization?
- junke 11y ago
  Why not? map, fold/reduce are not necessarily implemented as the textbook says.
  - suzuki 11y ago
    Certainly not. The foldl is a part of the TypeScript program that implements the Lisp. By the way, as for "while" in the implemented Lisp, it is defined with recursion and will be executed with tail-call optimization. See (defmacro while (test &rest body) (let ((loop (gensym))) `(letrec ((,loop (lambda () (cond (,test ,@body (,loop)))))) (,loop)))) in the const prelude of lisp.ts (http://www.oki-osk.jp/esc/typescript/lisp/lisp.ts.html http://www.oki-osk.jp/esc/typescript/lisp/lisp.ts.html).
    - junke 11y ago
      Thanks for this information. My point was more that map and reduce do not inherently require a recursive implementation and can be implemented directly as loops. I don't follow closely TypeScript: is tail-call merging done when compiling to Javascript or is the Javascript runtime responsible for this behavior?
      - suzuki 11y ago
        No, neither. That behavior is implemented in the same way as in the Lisp in Dart (http://www.oki-osk.jp/esc/dart/lisp-en.html#4 http://www.oki-osk.jp/esc/dart/lisp-en.html#4).
hydandata 11y ago
If you are interested in the future of Lisp in the context of browsers or Nodejs, take a look at this: http://lispblog.xach.com/post/138667931963/lisp-and-webassembly http://lispblog.xach.com/post/138667931963/lisp-and-webassem...

58 points by suzuki 11y ago

The Lisp in 1,200 lines of Dart, which has macros, quasi-quotation and tail-call optimization (https://news.ycombinator.com/item?id=9903332 https://news.ycombinator.com/item?id=9903332), was translated into 1,390 lines of TypeScript. It runs on browsers.

lomnakkus 11y ago

Thus, yet again, sort-of-validating Greenspun's 10th Rule. I'm not complaining... just observing :).

gadtfly 11y ago
As much as I love a good Greenspun invocation, I don't think this really counts: > Any sufficiently complicated C or Fortran program contains an ad hoc, informally-specified, bug-ridden, slow implementation of half of Common Lisp. isn't about the inevitability of literally implementing a Lisp interpreter. It's about something more abstract. see: http://c2.com/cgi-bin/wiki?LevelsOfGreenspunning http://c2.com/cgi-bin/wiki?LevelsOfGreenspunning
- lomnakkus 11y ago
  Mostly agreed. Which is why I said "sort-of-validating". ... so that I would be technically correct. Which, as we all know, is the best kind of correct. (Source: Futurama)
- DonHopkins 11y ago
  It's still got a good shot at validating Zawinski's Law of Software Envelopment. > Every program attempts to expand until it can read mail. Those programs which cannot so expand are replaced by ones which can.

agumonkey 11y ago

I will now write all my comment in Kanjis. It's too beautiful.

eatonphil 11y ago

Why does the foldl function use a while loop?

xyzzy123 11y ago
I think it's just manual tail-call optimization?
junke 11y ago
Why not? map, fold/reduce are not necessarily implemented as the textbook says.
- suzuki 11y ago
  Certainly not. The foldl is a part of the TypeScript program that implements the Lisp. By the way, as for "while" in the implemented Lisp, it is defined with recursion and will be executed with tail-call optimization. See (defmacro while (test &rest body) (let ((loop (gensym))) `(letrec ((,loop (lambda () (cond (,test ,@body (,loop)))))) (,loop)))) in the const prelude of lisp.ts (http://www.oki-osk.jp/esc/typescript/lisp/lisp.ts.html http://www.oki-osk.jp/esc/typescript/lisp/lisp.ts.html).
  - junke 11y ago
    Thanks for this information. My point was more that map and reduce do not inherently require a recursive implementation and can be implemented directly as loops. I don't follow closely TypeScript: is tail-call merging done when compiling to Javascript or is the Javascript runtime responsible for this behavior?
    - suzuki 11y ago
      No, neither. That behavior is implemented in the same way as in the Lisp in Dart (http://www.oki-osk.jp/esc/dart/lisp-en.html#4 http://www.oki-osk.jp/esc/dart/lisp-en.html#4).

hydandata 11y ago

If you are interested in the future of Lisp in the context of browsers or Nodejs, take a look at this: http://lispblog.xach.com/post/138667931963/lisp-and-webassembly http://lispblog.xach.com/post/138667931963/lisp-and-webassem...