Class attribute constraints applied on elements of array attribute, not the attribute itself

[jkramer]

The Problem

When adding a where constraint on a class attribute of type Array (others might be affected as well though), the constraint is applied to the elements of the array instead of the array attribute itself as one would expect.

Expected Behavior

The constraint checks should be performed on the attribute, not its array elements.

Actual Behavior

It seems like the constraint is applied to the elements of the array attribute.

Steps to Reproduce

# This crashes:
class Foo {
  has @.three-things where *.elems == 3;
}.new(three-things => <three excellent things>);

# This works, as apparently the where is applied to each of the three strings
# instead of @.three-things itself:
class Bar {
  has @.three-things where *.chars > 3;
}.new(three-things => <three excellent things>);

# This however works as expected:
sub foo(@three-things where *.elems == 3) {
  say "Got three things! {@three-things.gist}";
}(<exactly three things>);

Environment

• Linux 4.11.2-1-ARCH cc1: error: unrecognized command line option "-Wjump-missed-init" #1 SMP PREEMPT Mon May 22 06:53:49 CEST 2017 x86_64 GNU/Linux
• This is Rakudo version 2017.07-105-gc8c5c7129 built on MoarVM version 2017.07-266-ged84a632 implementing Perl 6.c.

[lizmat]

I don’t think this is a bug, although maybe a bug in documentation. The where clause is applied to the descriptor of the array, which governs the elements of the array. If you only want 3 element lists, you should probably used shaped arrays: class A { has @.a[3] } dd A.new(a => [1,2,3,4]); # Index 3 for dimension 1 out of range (must be 0..2) dd A.new(a => [1,2,3]); # A.new(a => Array.new(:shape(3,), [1, 2, 3])) HTH

…

On 16 Jan 2018, at 17:12, Jonas Kramer ***@***.***> wrote: The Problem When adding a where constraint on a class attribute of type Array (others might be affected as well though), the constraint is applied to the elements of the array instead of the array attribute itself as one would expect. Expected Behavior The constraint checks should be performed on the attribute, not its array elements. Actual Behavior It seems like the constraint is applied to the elements of the array attribute. Steps to Reproduce # This crashes: class Foo { has @.three-things where *.elems == 3; }.new(three-things => <three excellent things>); # This works, as apparently the where is applied to each of the three strings # instead of @.three-things itself: class Bar { has @.three-things where *.chars > 3; }.new(three-things => <three excellent things>); # This however works as expected: sub ***@***.*** where *.elems == 3) { say "Got three things! ***@***.***}"; }(<exactly three things>); Environment • Linux 4.11.2-1-ARCH #1 SMP PREEMPT Mon May 22 06:53:49 CEST 2017 x86_64 GNU/Linux • This is Rakudo version 2017.07-105-gc8c5c7129 built on MoarVM version 2017.07-266-ged84a632 implementing Perl 6.c. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[zoffixznet]

The where clause is applied to the descriptor of the array, which governs the elements of the array.

How come it applies to the Array itself when it's part of the routine's parameters?

[lizmat]

A very good question. I think TimToady / jnthn should have a say about this.

…

On 16 Jan 2018, at 17:28, Zoffix Znet ***@***.***> wrote: The where clause is applied to the descriptor of the array, which governs the elements of the array. How come it applies to the Array itself when it's part of the routine's parameters? — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[zoffixznet]

https://irclog.perlgeek.de/perl6/2018-01-16#i_15697385

<TimToady> someone probably just implemented the one as an anonymous subset,
       so it distributes like a subset type
<TimToady> we should probably make them consistent 

(there's more discussion in the chat)

P.S.: &-sigils are similarly affected

P.P.S.: there appear to be some glitches with Subset @foo parameter constraints; see here and here

[raiph]

I'm thinking this issue's title should be changed to "Problems with subset/where constraints applied to variables/parameters/attributes with anything other than $ sigil". Any objections?

As background info, I found this bug to be the nearest match for a scenario I just encountered:

my &foo where { True } = sub {}

yields

Type check failed in assignment to &foo; expected Callable[<anon>] but got Sub

In a comment above @zoffixznet has provided links to some IRC chat exchanges and added "P.S.: &-sigils are similarly affected". That reinforces my sense that all such problems belong in one issue until/unless it becomes clear a particular variation needs to be broken out into a separate one.

The IRC links in Zoffix's comment are broken; here's what I think they are supposed to be pointed to:

• TimToady: someone probably just ...

• the sigil makes all the difference.

• Looks like there's additional issues with Subset @foo parameters.