Inconsistent type inference between code typed in the REPL and imported as module.

[aanastasiou]

Describe the bug

The same set of definitions typed step-by-step in the REPL and imported as a module produce different behaviours.

To Reproduce

1. Suppose the following definitions:

    data s_cell = push(int a);
    alias s_stack = list[s_cell];
    data s_sym = push(int a)|add();
    alias s_prog = list[s_sym];
   

2. And the following minimal "test":

    for([add(), *s_cell r]:=[add(), push(0), push(0), push(0)]) print(r);
   

If the type definitions in (1) are typed into the REPL, then (2) works. If they are put in a module and imported then (2) does not return the same result. Specifically:

If block 1 is typed in the REPL:

This is inferred as a list[s_cell] (Expected).

[push(0), push(0)]

This is inferred as a list[node] (Kind of expected, it sees symbols from two different types, it's fine to report this as a list of the general common ancestor node)

[add(), push()]

This works (Expected)

import IO;

for ([add(), *s_cell q]:=[add(), push(0), push(0), push(0)]) print(q);
//Returns [push(0), push(0), push(0)]

If block 1 is put in a module and imported, then:

This is inferred as a list[s_sym]. (...Not expected)

[push(0), push(0)]

This is inferred as a list[s_sym] (Expected...but puzzling)

[add(), push()]

This fails (Not expected)

import IO;

for ([add(), *s_cell q]:=[add(), push(0), push(0), push(0)]) print(q);
//Returns []

Expected behavior

I think that the REPL behaviour should be the expected one.

Desktop (please complete the following information):

• Context: Commandline REPL
• Rascal Version 0.28.11

Additional context I am working along a specific example here, which is not posted in its entirety. However, the point of raising this is not this specific example but the behaviour it causes.