Re: [RFC] Patch for Union and Intersection Types

2016-04-26 19:33 GMT+02:00 Bob Weinand <bobwei9@hotmail.com>: > > Am 26.04.2016 um 19:00 schrieb Niklas Keller <me@kelunik.com>: > > > > 2016-04-26 16:58 GMT+02:00 Bob Weinand <bobwei9@hotmail.com>: > > > >> Yeah, I'd like to not allow ?Foo in any case if union types pass. > >> > > > > What's the reason for that? To me, null is neither a type nor should it > be. > > I don't want to argue that now, but null most definitely IS a type. It is > the null type with a single value, null, whose semantics is to represent > the absence of any other value. > > There's also plenty of evidence in PHP that null is considered a type: > get_type(), https://php.net/null, has it's own is_* > function etc. > At least in the PHP ecosystem it is considered its own type. Other type > systems may not, but the one of PHP does. > > Thus, we are definitely considering it as a type - a question for you: > What's the type of the value null then, if not null? > I'd say it's special and doesn't have a type at all. No value, no type. > If we wouldn't make a distinction between null and an object (i.e. allow > null to be passed to a function(Foo $foo) {}), it could be considered a > supervalue of type object. But that's not the case. If null doesn't have a type, it's clear that it can't be passed to a function(Foo $foo) {}.