develooper Front page | perl.perl5.porters | Postings from August 2023

Re: PPC Elevator Pitch for Perl::Types

Thread Previous | Thread Next
Chris Prather
August 21, 2023 09:21
Re: PPC Elevator Pitch for Perl::Types
Message ID:
On Mon, Aug 21, 2023 at 2:38 AM Martijn Lievaart via perl5-porters
<> wrote:
>[...] But I may miss something obvious, or something very deep. So can anyone explain why Perl::Types can work at all, and what changes to the implementation of the language are needed for that?

For anything like Perl::Types to work in we're going to at a minimum
need to expand type annotation support, especially around subroutine

Any Type system is going to want three things:

1) A way to annotate code with the types (e.g. `my Dog $fido`)
2) A way to assert the type (e.g. `my Dog $fido = Cat->new(); # dies
because Cats aren't Dogs `)
3) A way to define new types (e.g. `type Dog { $_ isa Dog }`)

#3 depends on #2 and #1, with Perl::Types you can just create a class
… how that mechanism works outside of RPerl is currently undefined. In
theory #2 is taken care of if all they're doing is exposing the
underlying SV subtypes. In practice people are arguing that's a
nonsensical thing to do, and ultimately there may need to be some work
there (there definitely is for Oshun style checks).

#1 however can't work without better support from Perl guts because
the mechanism that Perl::Types is currently using[1] will, as best
I've seen, only work on scalars (`my Child @children` doesn't trigger)
and will only work with explicit syntax like `(my Dog $fido, my Cat
$kitty) = @_` which is IMO painful enough I'd avoid using it. There is
also no support (yet) for subroutine signatures `sub (Dog $fido, Cat
$kitty, Child @children) { ... }`.[2]


[1] Technically the mechanism that Perl::Types is _currently_ using
won't work at all. The fields pragma only kinda sorta does a check and
Perl::Types isn't even using that. The above is based on work I've
seen from others and replicated myself using Lexical::Types, which
exposes underlying XS APIs … but as far as I can tell those APIs only
exist for scalars, and it doesn't change the fact that `my ( Dog
$fido, Cat $boomer)  = @_;` and `sub (Dog $fido, Cat $boomer) {...}`
are syntax errors.

[2] You can currently annotate a return type using a subroutine
attribute `sub adopt :returns(Dog) ($id) { ... }`. Implementing #2 on
subroutines is left as a lemma for the reader.

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About