Cannot assign to an immutable value

[zostay]

This may be a transient error in testing. I didn't notice this one until after I fixed the other initializer test failures. It may have been layered on top of previous failures that I was just ignoring while I had a known stupidity failure in the way I wrote the original tests.

[zostay]

This proof of concept demonstrates the problem.

use v6;
use ArrayHash;

my $b = 2;
my %hash := ArrayHash.new('a' => 1, 'b' => $b, 'c' => 3);
%hash<a> = 4;

This appears to be a regression that started with 2018.10.

[zostay]

Some investigation shows that the problem is basically that before 2018.10, a Pair created like 'a' => 1 contained a Scalar container for the value. At some point between 2018.06 and 2018.10 rakudo releases, this was changed so that 'a' => 1 lacks a Scalar and points to a bare Int (i.e., is immutable). The hash version of this assignment does provide a Scalar in every case. We can see this in the following one-liners:

% perl -e 'my @array-of-pairs = "a" => 1; @array-of-pairs[0].value = 2; dd @array-of-pairs'
Cannot modify an immutable Int (1)
  in block <unit> at -e line 1

and

% perl -e 'my %hash = "a" => 1; %hash<a> = 2; dd %hash'
Hash %hash = {:a(2)}

I'm not sure what the justification for this difference in behavior is, but I'm assuming to this point this change is intentional and has a good justification. In any case, this appears to be a feature, not a bug. The test needs to be fixed, not the module.

Weirdly, I would have expected the following modification to the @array-of-pairs one-liner to work:

% perl -e 'my @array-of-pairs = "a" => $(1); @array-of-pairs[0].value = 2; dd @array-of-pairs'
Cannot modify an immutable Int (1)
  in block <unit> at -e line 1

However, it most certainly does not. The following modification does work, though:

% perl6 -e 'my @array-hash = "a" => $=1; @array-hash[0].value = 2; dd @array-hash'
Array @array-hash = [:a(2)]

I propose to fix this by adding anonymous scalars into the initializers for array-hash.t and multi-hash.t. Bleh.

[zostay]

Upon further reflection, I think I want to lean towards the hash treatment. This is an ArrayHash after all. I need to think more about how to make the interface consistent, but not make this unexpected distinction between positional pairs and named pairs.