A couple years ago I hacked on a Python type inferencer someone wrote in Prolog. I wasn't enlightened, despite expecting to be, from a bunch of HN posts like this.

https://github.com/andychu/hatlog

For example, can someone add good error messages to this? It didn't really seem practical. I'm sure I am missing something, but there also seemed to be a lot of deficiencies.

In fact I think I learned the opposite lesson. I have to dig up the HN post, but I think the point was "Prolog is NOT logic". It's not programming and it's not math.

(Someone said the same thing about Project Euler and so forth, and I really liked that criticism. https://lobste.rs/s/bqnhbo/book_review_elements_programming )

Related thread but I think there was a pithy blog post too: https://news.ycombinator.com/item?id=18373401 (Prolog Under the Hood)

Yeah this is the quote and a bunch of the HN comments backed it up to a degree:

Although Prolog's original intent was to allow programmers to specify programs in a syntax close to logic, that is not how Prolog works. In fact, a conceptual understanding of logic is not that useful for understanding Prolog.

I have programmed in many languages, and I at least have a decent understanding of math. In fact I just wrote about the difference between programming and math with regards to parsing here:

http://www.oilshell.org/blog/2021/01/comments-parsing.html#w...

But I had a bad experience with Prolog. Even if you understand programming and math, you don't understand Prolog.

I'm not a fan of the computational complexity problem either; that makes it unsuitable for production use.