Re: Introspection for references

On Tue, Jul 10, 2018 at 12:15 PM, Nicolas Grekas < nicolas.grekas+php@gmail.com> wrote: > If we're seeking for a benefit, it would be by turning this to a >>> ReflectionZval insteaf of ReflectionReference. There would be isRef, but >>> also getType, etc. But honeslty I'm with you on this, only refs make sense. >>> >> >> I'm wondering if there is any legitimate use for something like this. The >> only thing that comes to mind is comparing arrays by identity. E.g. >> https://github.com/symfony/symfony/blob/master/src/Symfony/ >> Component/VarDumper/Cloner/VarCloner.php#L153 could check directly check >> whether an array is the $GLOBALS array. It would also allow direct >> recursion detection on arrays (ref >> https://stackoverflow.com/ques >> tions/9105816/is-there-a-way-to-detect-circular-arrays-in-pure-php). >> Doing the same on just references is nearly the same, but doesn't handle >> the $GLOBALS case correctly. >> > > Note that AFAIK, not only $GLOBALS has a special behavior, but also any > "scope" array (the deprecated $context argument of error handlers at least, > maybe what get_defined_vars() also? > > If we can return this info, I would definitely leverage it. > If not, that's no blocker, I'll let you decide. > > > >> There is a 4th possibility: have a "ReflectionReference::getHash" method, >>> that would return a unique string/id per reference (what >>> symfony_zval_debug() does.) >>> >> >> Is this the same as my 3rd suggestion with a different name >> > > No, it's the same, I read too fast. > > > >> we could return the address hashed with a per-process key. This would >> make the function somewhat slower though. >> > > That's exactly what we do in our extension, inspired by spl_object_hash(). > On the perf side, a XOR should be OK :) > Unfortunately XOR is about as good as just returning the address directly. XORing the value is basically doing a one-time-pad without the "one-time" part, aka completely broken. If we wanted to actually hide the address, we'd have to run it through a proper hash. (Then again, you are right that spl_object_hash used to do this, so maybe we just don't care...) Nikita