Re: Serializing references - part I

On Tue, 24 Oct 2000, Stanislav Malyshev wrote: > SS>> HashTable var_hash; > SS>> > SS>> zend_hash_init(..); > SS>> php_var_(un)serialize(..); > SS>> zend_hash_destroy(..); > SS>> > SS>> we would use > SS>> > SS>> php_var_hash_t var_hash; > SS>> > SS>> var_hash = var_hash_init(); > SS>> php_var_(un)serialize(..); > SS>> var_hash_destroy(var_hash); > > In fact, I do not see any difference. All the custom (for different > serializers) code will go after hash init anyway. But I can make it, if > you like, I only see no point in it. Every instance of code calling php_var_(un)serialize will have to initialize and destroy data structures. Let us assume that a faster/better/whatever algorithm for this references stuff would appear. The algorithm would need to store additional data. Would you then want to add more init/destroy stuff all over our codebase? The point of the interface is to hide the implementation details (aka abstraction). > SS>> print serialize(&$c)."\n"; > > Why call-time reference? I'm even not sure what it does, but you should > not use it. Why should not it work with your code? > SS>> O:3:"foo":2:{s:2:"ok";s:3:"yes";} > > This is wrong result, for me it gives: > > > O:3:"foo":2:{s:2:"ok";s:3:"yes";s:1:"n";O:3:"foo":2:{s:2:"ok";R:2;}} > > Which, btw, is wrong too - should be R:1. More work is needed on ot. - Sascha