Re:A Lisp in 99LOC
by eqvinox on Saturday August 16, 2025 @09:37PM
(#44928240)
Holy cow this is —structurally, not just expression— some of the worst C code I have ever seen, with the abuse of the 'double' type, 'T' cast that looks like a declaration, endian dependency, and strict aliasing violations galore… does this even work on a modern compiler? o.O
[ Parent
| on HN ]
Yeah. It's munged to fit in 99 lines.
That's besides my point, which is why I said "structure, not just expression".
It could've used a struct rather than wedging tags into a double's first byte and still be 99 lines.
If that's the trick you object to then you will be sad to hear that Ruby uses it.
That's a standard technique in interpreters. All non-toy Javascript engines use it, for example.
Re:A Lisp in 99LOC
by cardiffspaceman on Sunday August 17, 2025 @04:07PM
(#44934506)
Nan-boxing is awesome.
It does not, because there is a syntax error on line 81 (extra close parenthesis):
https://github.com/Robert-van-Engelen/tinylisp/blob/2d0fb35b...
Brought to you by this marvellous commit with the message "update",
https://github.com/Robert-van-Engelen/tinylisp/commit/40c6c0...
Surpringly readable though, despite all that, if you've ever implanted a language in similar constraints.
*implemented.
Too late to edit now.
Certainly not the worst I have seen, by far; but yes, not pretty. IMHO “Just for making it shorter“. I would very much prefer 200 lines of actually readable nice code.
For reading, I enjoyed fe, which was very clear: https://news.ycombinator.com/item?id=36239175
Yes! Exactly what I meant. 700 lines, but of code that can be understood, and looks clearly as C. Also btw, the general file structure, the documentation, I prefer fe any day of the week. Thanks for pointing that out, I will take a look at it.