[CDBI] Re: Class::DBI internal column swapping wierdness ?

William Ross will at spanner.org
Thu Sep 29 16:53:09 BST 2005

On 29 Sep 2005, at 16:41, Richard.Foley at t-online.de wrote:

> My base class simply sets up the DB connection, and has a few triggers
> for creation, update and deletion, and that's it.  There are no column
> or group relationships set up or assumed in any way.  All the  
> behaviour
> (should) be defined in the child classes, as described.  I'm really  
> not
> doing anything fancy here, but something is definitely biting.

Hm. It might possibly be to do with the triggers, I suppose, but  
you've no trigger declarations in Att::Task, or anyway not in the  
sample code.

I ask because there's a bug in cdbi that results in shared meta_info  
if you declare similar class-level properties at more than one level  
in the inheritance hierarchy. Relationships in later versions of cdbi  
are essentially dispatched methods, and from your symptoms it sound  
like your meta_info hash is getting stamped on, resulting in the  
wrong call during retrieval of the related object.

I'm intrigued to see that a newly-retrieved object is well formed.  
Can you identify a short sequence of operations that gives you a  
broken one?


More information about the ClassDBI mailing list