__sleep() function in TActiveRecord

[short]

Hi,

there is an inconsistency in PHP's calling of the _ _sleep function on objects, which results in incorrectly omitting serialization of private variables from ancestor classes. Details can be found in this thread in the Prado Forum.

To summarize, there are two problems in TActiveRecord:

• The variable _readOnly is private, and there is a _ _sleep() function present, which results in PHP not serializing the variable when a class derived from TActiveRecord is serialized. It also throws a notice in some cases.
• The _ _wake() function should be called _ _wakeup().

Just removing the _ _sleep() function was remedy for the unwanted notice PHP threw.

[donkee]

The function __sleep() shouldn't de deleted but fixed. This method permit serialization of object for be saved between different request of the user.

The function should don't return a table which contain name of all members of class. He should only have the public one.

The function __wake() should effectivelly renamed as __wakeup() this one permit to finish initialization of object after serialization.

But i think that this function is bugged too. Because the private var _recordState stay at 0 which will maybe an inconsistant state for TActiveRecord.

[short]

I agree that it should better be fixed, however, since it returns all object variables anyway, there would be no change in simply deleting it. I think fixing it would be impossible with the current state of PHP, since there is the problem with calling __sleep() on an inherited class I described on the thread. This bug with serialization of private variables leads to _readOnly not being serialized.

Why should it return only public members?

[donkee]

You work with the Prado v3.1 which is revision 2290 in SVN repository. If take care about reading the last revision, you can see that the problem of serialization has allready been resolved.

But for the problem of unserialization is still here. And function __wake() must be renamed as __wakeup()

[xue]

Thanks for the discussion. I will change _recordState to protected to solve this problem. It's not pretty, but I don't have any other solution without big change.

[jake]

This problem is still an issue! I downloaded the latest head and hoped it was fixed but I still get that problem on php 5.2.5!

I even made a thread about it with all the errors I got...

http://www.pradosoft.com/forum/index.php/topic,9780.0.html

[jake]

I'm sorry this ticket can be closed again. I reopened that ticket because I was experiencing the same error and had the same symptoms of this ticket so I thought it was the same thing. Later when I resolved my issue I learned that it was something else. Sorry there is another ticket created for the issue I had.

Thanks, - Jake

[schmunk]

Although this bug is closed, I think these two links are pretty interesting. They explain how to treat private and protected variables with sleep() functions: http://www.php.net/manual/en/language.oop.serialization.php#81642 http://us.php.net/manual/en/language.oop5.magic.php#81492

[sergeymorkovkin]

Not fixed properly. See this topic: http://www.pradosoft.com/forum/index.php/topic,10242.0.html

[knut]

I'm closing this one since the the original problem related to private $_recordState is solved. For private $_connection see #855.