| Can all Classifier take part in an Association? [message #691134] |
Thu, 30 June 2011 15:36  |
Eclipse User |
|
|
|
hi to all
i don't unterstand if really all classifier can take part in an association
up to now i thougth, that is only possible for classes and interfaces but i found no constraint or something like that.
that would mean, a datatype, a enumeration or even an association can take part in an association, that would be very strange
is there no constraint like that or am i just blind?
best regards
|
|
|
|
|
| Re: (no subject) [message #691756 is a reply to message #691456] |
Sat, 02 July 2011 04:51  |
| Eclipse User |
|
|
|
hi ed
thanks for your reply.
i'm not sure if i understood all now(because i'm still thinking that way, that an association is between classes):
an association is between its associationsends, which are properties, the properties has an type, and this type is the part of the association (that would mean the owner of the attribute has no influence)?
for example three classes A,B,C and one association (like in the picture):
class A has an property with type C and Class B has an property with type C, the associationsends of the association are the properties of the classes A and B. Now when the type of the property is deciding, then that would mean we have an reflective Association from the class C, although the owners of the properties are Class A and B?
that seems a little bit strange to me
Attachment: ClassABC.png
(Size: 228.36KB, Downloaded 290 times)
|
|
|
| Re: (no subject) [message #691777 is a reply to message #691756] |
Sat, 02 July 2011 05:23 |
| Eclipse User |
|
|
|
Hi
It certainly seems strange to me.
Just because you have N information fields that you can enter, it does
not mean that you have a free choice for all of them.
Have you validated your example?
The usual case for a bidirectional association is that each end has the
opposite container as its type; two type choices are therefore constrained.
The extra UML specification complexity seems to arise from an ability to
have an association to a type that cannot own associationEnds/properties
so the association has to own the associationEnd/property on behalf of
that type.
I recommend ignoring the complex case and concentrate on the more
obvious bidirectional use case.
Regards
Ed Willink
On 02/07/2011 10:51, stoehm wrote:
> hi ed
> thanks for your reply.
> i'm not sure if i understood all now(because i'm still thinking that way, that an association is between classes):
> an association is between its associationsends, which are properties, the properties has an type, and this type is the part of the association (that would mean the owner of the attribute has no influence)?
> for example three classes A,B,C and one association (like in the picture):
> class A has an property with type C and Class B has an property with type C, the associationsends of the association are the properties of the classes A and B. Now when the type of the property is deciding, then that would mean we have an reflective Association from the class C, although the owners of the properties are Class A and B?
> that seems a little bit strange to me
>
|
|
|
|
|
| Re: (no subject) [message #692340 is a reply to message #691910] |
Mon, 04 July 2011 05:02 |
| Eclipse User |
|
|
|
hi ed
thanks for your reply
you are right, there are no real problems when a enumeration or a datatype take parts in an association, but i don't think there is a need for that. a normal attribute should be enough for that case. for me it is a little problem, because in my codegeneration i have take care for the referential integrity, and i thougth during runtime it makes no sense to take care of referential integrity in case of datatypes or something like this.
so far i thank so for helping me to reflect the problem
best regards
stoehm
|
|
|
Powered by
FUDForum. Page generated in 0.03441 seconds
Home