Re: What is the right way to handle a hierarchy of same-class objects?

[prev] [thread] [next] [Date index for 2005/02/22]

On 22 Feb 2005, at 02:25, Karjala wrote:

> However it would be nice to get undef as an answer.
>
> I have the following scenario on my sites:
>
> Class A objects get created and deleted all the time.
>
> Class B objects (a history log) have a field that refers to a class A 
> object (which may or may not exist at the present time). Class B 
> objects never get deleted.
>
> I need to check with a short piece of code if given a B object, the 
> corresponding A object exists, in order to display a link on the 
> website if it does exist.
>
> $b->a would be enough, if it would return null instead of an object 
> reference.

$b->a will return undef if the 'a' column is NULL. Or it will in mysql 
and SQLite, anyway: I can't speak for the grown-up databases. That's 
what I meant by 'it should work as you expect'. Sorry: wasn't very 
clear.

If the 'a' column contains a defined value - even 0 - then cdbi will 
try to inflate it. With recent versions, if you try to inflate from a 
broken foreign key you get a fully-formed but empty cdbi object, 
because CDBI::Relationship::HasA is lazy (in the good sense) and calls 
_init($fkey) without ever touching the foreign table.

I suppose ideally you'd get undef back at that point, but the integrity 
check is presumably omitted for good reasons, and I'm keeping quiet 
because it's all such a great improvement over the old behaviour, in 
which cdbi would _croak if the foreign object didn't exist or the 
column was empty...

and as long as your relationship column defaults to NULL everything 
works nicely.

best

will