On Tue, 30 Mar 2021 at 13:51, Ilija Tovilo <tovilo.il...@gmail.com> wrote:

> Hi Matthew
>
> > I like the proposal. I also support the covariance.
> >
> > One question though.
> > The RFC mentions possible future use of "nothing" as a bottom type for
> > parameters in generics. This makes a lot of sense to me. Even without
> > generics, it could be used for parameters in abstract methods.
> >
> > So why not already introduce "nothing" as the universal bottom type, and
> > use it instead of "noreturn"?
> > Can "noreturn" do anything that "nothing" can't?
>
> I'm also a little confused by this statement. The exact wording from the
> RFC is:
>
> > Arguments for never: ... It's a full-fledged type, rather than a keyword
> used in a specific situation. A far-in-the-future generics proposal could
> use never as a placeholder inside contravariant generic types.
>
> From what I understand, in Hack noreturn and never are both
> full-fledged types. Nothing is a bottom type while noreturn is not.
> Since in your proposal noreturn would also be a bottom type there's no
> reason why it couldn't be used in covariant or contravariant generic
> parameters.
>
> Please correct me if I'm missing something.
>
> Ilija
>
> --
> PHP Internals - PHP Runtime Development Mailing List
> To unsubscribe, visit: https://www.php.net/unsub.php
>
>
In Hack it's a bit confusing. I think the Hack team ultimately decided that
"nothing" was a better name for the bottom type than "noreturn", and
confusingly introduced one while the other was still active.

They have a long-term plan to deprecate "noreturn" in favour of "nothing"
and stop supporting the former name, though it's unclear when that might
happen.

Reply via email to