On Tue, Nov 16, 2021 at 3:59 PM James Gilliland <neclim...@gmail.com> wrote:
> On Tue, Nov 16, 2021 at 4:23 AM Rowan Tommins <rowan.coll...@gmail.com> > wrote: > > > On 16/11/2021 09:27, Andreas Heigl wrote: > > > > > >> I see, yes, code that is 100% perfectly tested can get away without > > >> the language performing any error checking at all - the behaviour is > > >> all guaranteed by the tests. I would be very surprised if even 1% of > > >> PHP applications can claim such comprehensive tests. > > > > > > The topic here was that new code can verify the declaration of a > > > property by using tests. That does not need to happen on the language > > > level. I was never talking about adding tests for existing code. > > > > > > > > For most code bases, even new ones being written from scratch in PHP > > 8.0, that level of testing simply doesn't exist, and having the language > > tell you "hey, you wrote $this->loger instead of $this->logger" is a > > useful feature. And, in a lot of cases, more useful than having the > > language say "OK, I've created your dynamic $loger property for you", > > which is what currently happens. > > > > What you described there sounds like a warning and not a fatal error. Maybe > that's where some of the trepidation is coming from. I know I'm less > worried about the deprecation notice and more worried about what happens in > PHP 9 when it's a fatal error. > I can't say that this line of reasoning has its merits, but then there's no benefit to the engine itself. Issuing a warning and carry on materializing dynamic properties will never bring the original performance improvement that was part of the original state of the RFC. -- Marco Aurélio Deleu
