@aphyr Ah yeah, that's true. The implementations can do *some* optimizations to avoid extra choice-points, but it's pretty finicky & pretty ugly to take advantage of; that makes sense.

Yeah, reif is very cool -- I used it to write grammars for HTTP/2 that can run in either direction (i.e. parse or serialize) but it was very complicated to make it work (https://github.com/jamesnvc/http2_client/blob/master/prolog/frames.pl).

Very cool, in any case.

@aphyr Wait, so it's going Prolog -> Lisp -> miniKanren? That is beautiful, I love it!
I also like your comment on line 91 of that gist; logic variables are amazing!

For the code itself, any particular reason why all the semi-colons instead of defining more clauses for the predicate? I'd usually see like

foo(X) :- a.
foo(X) :- b.
...

vs

foo(X) :- ( a ); ( b ) ; ....

@aphyr haha, I think the Prolog community tends to enjoy wild experiments with the language itself :grin:

Also, wrt to unicode characters, the semantics of what's a variable/atom should be defined by these rules: https://www.swi-prolog.org/pldoc/man?section=syntax#sec:2.16.1.9 I'm sure Jan would like to update either the language or the docs if there are inconsistencies!