Referential transparency on non-deterministic functions

[AnnelineD]

Describe the bug
Non-deterministic functions are not referentially transparent.

To Reproduce

> !(subtraction (superpose (a b d)) (intersection (superpose (a b c)) (superpose (a b d))))
[d]
> (= (test6 $set $subset) (subtraction $subset (intersection $set $subset)))
> !(test6 (superpose (a b c)) (superpose (a b d)))
[b, d, a, d, a, b, d]

Expected behavior
The wrapped function test6 produces the same result as the top-level test.

Actual behavior
test6 gets applied element-wise.

Additional context

• MeTTa version: 0.1.10

[vsbogd]

The issue is that test6 is not typed and thus its arguments are evaluated before function call. Thus test6 is called in the pairs. First item of the pair is an element of the (a b c), second item is an element of the (a b d). It works according semantics of the non-determinism in MeTTa. One can make work this example by adding type for the test6:

(: test6 (-> Atom Atom Atom))
(= (test6 $set $subset) (subtraction $subset (intersection $set $subset)))
!(test6 (superpose (a b c)) (superpose (a b d)))
[d]