|
Posted by Mumia W. on August 26, 2006, 12:00 am
Please log in for more thread options
On 08/25/2006 06:35 PM, Michael Küper wrote:
> thanks for your reply ...
>
> Mumia W. wrote:
>> I've haven't used CGI::Session, but I don't think it'll work for two
>> reasons:
>> 1) serialization and deserialization breaks the internal connection
>> between hashes and their associated classes, and
> What "internal connection" are you refering to? As far as i know, an object
> is nothing more than a blessed hash reference!?
Again, my entire previous post was a Wild A$$ Guess (WAG);
take it with some grains of salt.
I imagined that invisible Perl bytecode establishes the
connection between a hash reference and it's class. I was
wrong, because I tried it with Data::Dumper and FreezeThaw,
and it worked.
Note, I didn't use CGI::Session but instead stored the data in
the __DATA__ section.
> So the "connection" between
> hashref (object) and package (class) is the blessing (the packagename stored
> with the hashref), or am I missing something here? If there IS actually
> something more, this would explain the behaviour ...
>
>> 2) after deserialization, objects referred to by references may be in
>> different locations.
> This is not a problem (at least in my case). The serialized structure looks
> good (I had a look at the session text file), references are replaced by
> their serialized target, no memory address pointers inside ...
>
>> I have three bits of advice: 1) bless the hashes back into their
>> respective classes after deserialization,
> they do have the correct class! (since ref($hashref) returns the correct
> name).
>
It should work, even without re-blessing.
>> 2) use custom serialization and deserialization methods to reconstruct the
>> object tree for each request, and
> would be an option ...
> but I still hope to find a way to use the default session deserializer.
>
>> 3) make sure the objects' package (class) files are loaded before
>> deserializing.
> This is something I thought of too ...
> I will check, if this helps. But anyway: if the packages would not have been
> loaded, the "eval-way" of calling the method could not have worked, could
> it?
>
At least in PHP, storing objects in sessions doesn't work if
the class definitions aren't loaded when the session data is
loaded by the interpreter (AFAIR). Although Perl is a
completely different language, the problem is (probably) the
same. How can you associate an object's data with a class if
the class' definition isn't loaded?
>> This post is a WAG so forgive me if I'm totally off the wall.
> I don't know what WAG stands for, but i forgive you in any case and thank
> you for your time :-)
>
>
>> PS.
>> The ref() function returns a string of text and, by itself, is no
>> indication that the internal bytecode that connects a hash reference to a
>> class is still there.
> If there is some "internal bytecode" could you please give me a link to an
> explanation? Cause I didn't find anything like this in my perl book.
>
> kind regards
> Michael
>
>
I can't test your program because you didn't post one.
Why don't you post a minimal but complete Perl script that
demonstrates your problem, so you can get some responses that
are better than WAGs :-)
| 
XML Sitemap