Is it possible to define a lookup relation for map keys, similar to (geto k m v) ? I saw that maps unify only over values, so this might be hard.

Okay,

(defn geto [m k v]
  (featurec m {k v}))

it is very tricky

if k is a key in m, and v is a value in m, and q unifies with m, what values does q have?

(values in the sense of, what are the possible values of q)

Hmm, I'd assume all values that (featurec m {k v}) produces? 😉

it is trivial to determine the values when keys are ground

What do you mean by ground?

a concrete value

a known keyword

Ok

(as opposed to a logic variable with an unknown value)

Yeah, got it

basically you can't use just a clojure map to represent that

you need some more complicated datastructure that understands unification and can do things like merge keys that are unified

like if you say (get m k v) (get m y x) (== y k)

uh, geto

Hm, so it would be possible, but only if I converted my maps to that richer datastructure before?

And I'd have to make it extend the unification protocol?

I am not sure

I feel like I was reading a paper on unifying maps recently

Really? Nice coincidence then 🙂

oh, no, it was clp(set)

Ah, ok

but I think that would be really similar

e.g. you would represent the map as a set of k/v pairs

https://github.com/clojure/core.logic/wiki/CLP(Set)

there is a scheme implementation somewhere that needs porting

Hm yeah, it looks like this would do it, if it were implemented, which it doesn't seem to be 😄

https://github.com/namin/clpset-miniKanren

(the link to the paper is in the readme on that repo)

Hm, that looks more like a weekend-y kind of thing 😉

I wouldn't mind a simpler way to achieve my goal, which is basically have an interpreter for a very simple language

I think the scheme guys just use a list of pairs

Yeah, that might be the way to go for now

Thanks, I'll try that

Also read the paper when I have a moment, which might be a while 😉