How does the scope effect EVAL of an infix:<> sub?

Question

This code works as expected:

sub infix:<mean>(*@a) {
    @a.sum / @a.elems
}
sub Mean (*@a) {
    @a.sum  / @a.elems
}

say EVAL 'Mean 2, 6, 4';     # Output: 4
say EVAL '2 mean 6 mean 4';  # Output: 4

It works as expected when line 7 is in its own scope:

{say EVAL 'Mean 2, 6, 4';}   # Output: 4

But when line 8 is in its own scope:

{say EVAL '2 mean 6 mean 4';} 

===SORRY!=== Error while compiling .../EVAL_1
Two terms in a row
at .../EVAL_1:1
------> 2⏏ mean 6 mean 4
    expecting any of:
        infix
        infix stopper
        statement end
        statement modifier
        statement modifier loop

Why are the two subs treated differently?

Answer 1

This is a known issue, also affecting the REPL.

Problem is that doing a string EVAL will "see" the surrounding scope with all of its additions, so:

say EVAL 'infix:<mean>(infix:<mean>(2, 6), 4)';

will produce 4, because the sub &infix:<mean>(*@a) is known.

But the grammar changes to allow it to work as an infix, are currently not seen inside the EVAL. And that's why you see the compile time error.

I have good hopes we will be able to fix this with the new Raku grammar, based on RakuAST.