Toby's Random Thought #1 - overloading

[tobyink]

It is reasonable for someone who has declared:

my $x :of(INT) = shift @values;

To assume that my $y = $x * 1 will then be safe code, and they don't need to wrap it in try, or add no warnings "numeric".

However, as things currently are, it can throw an exception.

Why? $x could be an object which overloads 0+ but doesn't overload *.

https://toby.ink/blog/2023/04/02/experiments-in-overloading/

Accepting overloaded objects by default is dangerous. However, rejecting them by default pushes overloaded objects further into the realms of being "second-class citizens". I don't have a solid solution for this, but it's something to consider.

(A possible solution might be for Perl itself to always enable numeric fallbacks when 0+ is overloaded, always enable string fallbacks when "" is overloaded, and always enable logic fallbacks when bool is overloaded, regardless of whether the class declares fallback => 1.)

[Ovid]

I think you're correct that this is a problem, but I don't think Data::Checks is the right level to fix it.

use Math::BigInt;
my $x = Math::BigInt->new(12);
my $y = Math::BigInt->new(5);
say $x / $y;
say $x + $y;

Those are overloaded objects and they do the right thing. By your suggestion, this would break:

use Math::BigInt;
my $x :of(INT) = Math::BigInt->new(12);
my $y :of(INT) = Math::BigInt->new(5);
say $x / $y;
say $x + $y;

That's not going to match people's expectations. If the overload is broken, fix the class. I think Data::Checks work here (though perhaps Math::BigInt isn't the best example here).

[zmughal]

In some cases you might want such things to throw an exception / be invalid operations. For example, this is what statistical tools do with https://en.wikipedia.org/wiki/Level_of_measurement. That's what you'd want when dealing with certain units of measure, dates, etc. More here: https://pubmed.ncbi.nlm.nih.gov/24987515/.