__clone() vs unserialize(serialize())?

With t开发者_运维百科he reference of this question, I got an another solution is use

$obj2 = unserialize(serialize($obj1));

instead of

$obj2 = clone $obj1;

Which one is better to use?

tl;dr version: Use clone for simple objects and trees, unserialize(serialize()) for complicated graphs of objects.

Longer explanation: Unless $obj1 implements __clone(), the expression clone $obj1 will return a shallow copy of $obj1, but sharing the contents of objects pointed to by $obj1. Even if __clone() is implemented to perform a deep copy by recursive clone of members, it will only work safely if the object graph is a tree. If the object graph contains cycles, it will recurse indefinitely and... well... this is Stack Overflow for a reason. :-) If it is a directed acyclic graph but not a tree, any object referenced multiple times will have those multiple references replaced by copies, which may not be what you want.

unserialize(serialize($obj1)), on the other hand, will deal with cycles in the object graph, but is more expensive in terms of both CPU time and memory.

Sure, that creates a clone, but with incredible overhead and without the ability to actually define any settings or behaviors using the __clone magic method.

I would use the clone keyword and if you're in need of further research see the comments section at http://php.net/manual/en/language.oop5.cloning.php for proofs that cloning doesn't require that a change to one object persists to its clone.