News for hackers. Stuff that matters.

A comment on Unification in Elixir. Parent comment

Re:Unification in Elixir by ReleaseCandidat on Monday July 01, 2024 @08:58AM (#40845423)
Well, I'd do that with "real" interpreters (using bytecode) too. But on the other hand also _especially_ for "toy" tree walking interpreters, to learn how to build the pattern matching tree -- as unification is needed for the type checker anyways ;). Although I would start by reading Wadler's version (that I linked in the other post), it's IMHO a bit more accessible than Peterson's https://www.classes.cs.uchicago.edu/archive/2011/spring/2262...

But please don't get me wrong: I'm just saying that there exist other, specialised and (well, generally) more performant algorithms for pattern matching and people should have at least heard about them. If somebody decides to use unification for pattern matching, that still could be changed if performance or better error messages or ... matters.

[ Parent | on HN ]
Re:Unification in Elixir by contificate on Monday July 01, 2024 @09:37AM (#40845758)
Ah, I thought your username was familiar: you recommended Wadler's approach on a previous HN thread concerning my blog post about Pettersson/Maranget's algorithm (https://news.ycombinator.com/item?id=39241776). Yeah, I admit that Pettersson's telling of the algorithm is a bit turgid (probably why Maranget's paper - which is the same technique as its core, at least to my understanding - is cited more often).
[ Parent | on HN ]