5

Type captures are a cool thing to use for even some simple stuff:

sub assert-same(::T $a, T $b) { ; }

assert-same 1, 2;   # silent 
assert-same 1, "a"; # type check error

However, the result is non-intuitive for positionals. If I have the following signature

sub foo(Str @bar, Str @xyz) { ; }

Then @bar is a positional whose elements must be Str, as is @xyz. But if I use a type capture, things go weird: 

sub assert-same(::T @a, T @b) { ; }

my Str @x = <i j>;
my Str @y = <x y>;

assert-same @x, @y; 
# Type check failed in binding to parameter '@b';
# expected Positional[Array[Str]] but got Array[Str] (Array[Str].new("x", "y"))

It seems that the first type capture is capturing via .WHAT (which makes sense for scalars) rather than .of which to me is the intuitive sense for positionals, given that the immediate reuse of the capture would only work if .of had been originally used.

This smells of a bug, but if it's by design, is there a way to enforce via type capture that two typed positionals have elements that are of the same type? In the meantime, I can use

sub assert-same(::T @a, @b where .all ~~ @a.of ) { ; }

my Str @x = <a b c>;
my Str @y = <x y z>;
my Int @i = 1,2,3;
assert-same @x, @y; # silent 
assert-same @x, @i; # type check

But that seems a bit silly.

user0721090601
  • 5,276
  • 24
  • 41

1 Answers1

5

I think there are many bugs lurking in this area. I posted these a couple weeks ago:

I searched RT and the GH rakudo issues queue before posting them. In the latter I mention "cf #2595 and other type capture bugs". Maybe it's already in there; if not, please add an issue. TIA.

raiph
  • 31,607
  • 3
  • 62
  • 111
  • 1
    Noted. Those seem to both capture on the scalar, so this would appear to be heavily related but slightly distinct. I'll get some more research done in issues and write one up if it indeed ends up being distinct. – user0721090601 Aug 20 '19 at 14:39
  • I now recall that I bumped into this using a positional, then found it was also wrong for a scalar and concluded I'd post the scalar example. That said, your bug looks fundamentally different (expected a[b[c]] got b[c]). Anyhoo, if you don't find essentially an exact match among the issues, but also decide it's not sufficiently distinct to warrant its own issue, it still makes sense to me to add a link to this SO to one of the existing issues. – raiph Aug 21 '19 at 12:49