Re: [GDL2] Question about relation definition and faulting

[Stéphane Corthésy]

On Tuesday, Feb 18, 2003, at 21:03 Europe/Zurich, Manuel Guesdon wrote:

> Hi,
>
> Question for EOF/GDL2 gurus :-)
>
> I have 4 entities A, B, C, D, with relation A<->B, B<->C, C<->D
> I also have a relation  to 'jump' from A to C with a definition named 
> aToC like
>         aToB.bToC
> If I de-fault this relation and get (for example) an empty array 
> (matching B record exists but there's no B related C
> record) and next add D and C records with C matching the B record, As 
> aToC, is already de-faulted, the array is still
> empty.
>
> Do you think it's 'normal' so I need to refault aToC or explicitely 
> add object to both side of a relation named cToA
> (i.e. GDL2 is right).


I can't say it's "normal", but it's the way it's implemented in EOF. 
Flatten relationships have nasty "side-effects", and should be avoided, 
IMHO, except for readOnly (final) properties, as flatten 
attributes/relationships are not updated automatically, or when 
intermediate entity has no class-properties. BTW, flattening a 
relationship has quite no sense when the intermediate entity has 
class-properties: it has full sense though when building many-to-many 
relationships with an intermediate entity that no class-properties at 
all.


> Or do you think aToC relation shouldn't be a fault array but a call to 
> something like valueForKeyPath:@"aToB.bToC" (i.e.
> dynamically trying to find matching values) ? And if yes, is it a 
> problem of GDL2 or is it a wrong model ?


Seems current implementation is like EOF's. I'd say you should not 
change it, but document the way flattening works, and when it can be 
used "safely".

Stéphane


> Thank you !!
>
> Manuel
>
>
> --
> ______________________________________________________________________
> Manuel Guesdon - OXYMIUM <address@hidden>
> 14 rue Jean-Baptiste Clement  -  93200 Saint-Denis  -  France
> Tel: +33 1 4940 0999  -  Fax: +33 1 4940 0998
>
>
>
> _______________________________________________
> Gnustep-dev mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gnustep-dev