Fix unique

[vsbogd]

Fixes #760 Also removes interpreter from the functions implementations.

[vsbogd]

@ngeiswei could you please check if it what you needed?

[ngeiswei]

Works like a charm, thanks!

[ngeiswei]

Unfortunately I spoke too fast :-(, it does not work in my case, but it is a complex one, I want to simplify it a bunch before pasting it here. But to give you an idea, the following

!(unique (bc &kb
             (fromNumber 5)
             (: $prg (-> (: $x (SongIn "English")) (SongIn "Chinese")))))

does not work, while the following

!(let* (($results (collapse
                   (bc &kb
                       (fromNumber 5)
                       (: $prg (-> (: $x (SongIn "English")) (SongIn "Chinese"))))))
        ($uniq-results (unique (superpose $results))))
   $uniq-results)

does. Meaning that I need to explicitly collapse the results, then call unique on their superposition, and then it works.

I am not sure if it is a bug or a feature, I can still use that, but of course it is not as elegant as just calling unique on my non-deterministic function call.

[ngeiswei]

Please find a simplified example

(= (f) a)
(= (f) a)

!(unique (f))
!(let $y (collapse (f)) (unique (superpose $y)))

[vsbogd]

Thanks @ngeiswei. Seems like it is a bug which prevented me to implement it last time. unique is implemented in stdlib and thus when collapse is called from implementation it uses stdlib space as a context. As consequence (f) is not evaluated and stays (f) because stdlib doesn't have definition of (f). It is the reason why unique returns (f) and after collapse it is evaluated in two instances of a.

[ngeiswei]

This is better than nothing, I can still use it by collapsing and superposing, so I approve merging if that's the best we can get for now.

[vsbogd]

Well, I think we can merge it, because it is better then previous state. Then I will think how to fix it properly and raise another PR.

[vsbogd]

Btw, @ngeiswei , the following code works:

(= (f) a)
(= (f) a)

!(unique (capture (f)))