[CDO] Wiki Performances tips [message #1760325] |
Tue, 25 April 2017 20:11 |
Stephane fournier Messages: 340 Registered: July 2009 |
Senior Member |
|
|
Hi,
Reading the CDO Performances wiki, regarding the model definition, I cannot test what it is mentioned : https://wiki.eclipse.org/CDO/Tweaking_Performance#Defining_a_Model
Indeed, in this section, one hint is to set to false the unique property for containment reference, in the Ecore model, if we know 'a priori' that our algorithm does never violate a uniqueness constraint.
I tried to do so for tests purpose, but I cannot validate my EMF model.
I get this validation error : A containment or bidirectional reference must be unique if its upper bound is different from 1
I can only switch to false for non-containment reference and for many-valued attribute based on EDataType.
Does anyone know if it is a mistake in the CDO wiki or a change since the page was written ?
If CDO do not use EMF proxies, could you please detail why switching resolveProxies to false can help performances since it is not used in CDO implementation ?
Thanks for your help,
Stéphane
[Updated on: Tue, 25 April 2017 20:12] Report message to a moderator
|
|
|
Re: [CDO] Wiki Performances tips [message #1760339 is a reply to message #1760325] |
Wed, 26 April 2017 05:34 |
Ed Merks Messages: 33113 Registered: July 2009 |
Senior Member |
|
|
It advice looks bogus to me. Containment and bidirectional references must be unique and while a non-containment, unidirectional references can be non-unique, the implementation classes always enforce uniqueness for all many-valued references and that's always been the case. I don't understand the point about proxies either; like you say, if CDO doesn't use them why does it matter at all.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
Powered by
FUDForum. Page generated in 0.02446 seconds