Morloc definitions should be able to use future implementations

[arendsee]

Suppose we have this base module:

module low (*)

fst a b :: (a, b) -> a
snd a b :: (a, b) -> b

const a b :: a -> b -> a
const a b = a

uncurry a b c :: (a -> b -> c) -> (a, b) -> c
uncurry fabc ab = fabc (fst ab) (snd ab) 

And a higher module that imports it and provides definitions for fst and snd:

module high (foo)

import low

type Py => Int = "int"

source Py from "high.py" ("fst", "snd")

foo :: (Int, Int) -> Int
foo = uncurry const

Now when uncurry runs, here I want it to access the fst and snd definitions provided here. But currently it does not, since only the local scope in the module low is searched for definitions.

This was the intended behavior. However, when I write the language-independent code of the base modules, it would be very useful to define new terms in the abstract.

[arendsee]

I've realized that Morloc modules should form an undirected graph, not a directed acyclic graph. Cycles are allowed. But most importantly, all connections between modules should be symmetric and bidirectional. Any information about a term that is defined in one module is valid in all connected modules.