Home » Modeling » GMF (Graphical Modeling Framework) » ER like connections
|
| Re: ER like connections [message #642137 is a reply to message #641843] |
Tue, 30 November 2010 03:42   |
Eclipse User |
|
|
|
Originally posted by: christian.busch2.gmx.de
Hi,
you could try using an edit helper in combination with an instead
advice. This way you should be able to create an "ER node" and the
corresponding links if you use a link (to get start and end node) as
object for the tooling/mapping and overwrite its creation process as
stated above.
Otherwise I think you would have to modify generated code, and this is
neither funny nor easy to maintain. ;)
If I have missed a better way, please someone correct me.
Best regards,
Christian
Am 28.11.2010 18:18, schrieb ralph.:
> Hi,
>
> I want to have a link that can be added like a link but looks like a
> node with two links. To achive this I currently have a node and a link
> with the target attribute set to an attribute of the node. So I need to
> add first the node and then the links.
> What I want is to do is add a link and get a node with two links. I need
> something looking similar to the relations in an ER diagram
> ( http://en.wikipedia.org/wiki/File:Erd-entity-relationship-ex ample1.svg)
> but with the ability to add attributes etc. to the relation not only a
> label.
>
> Thanks in advance
|
|
| | |
| Re: ER like connections [message #642231 is a reply to message #642202] |
Tue, 30 November 2010 13:26 |
| Eclipse User |
|
|
|
Originally posted by: christian.busch2.gmx.de
Hey.
Am 30.11.2010 12:51, schrieb ralph.:
> Hi,
>
> first thank you for the quick reply. I did play around with edit helpers
> and advices. Unfortunately I could not to get it work.
> The only method that gets executed in my advice is the
> getBeforeConfigureCommand command. All others like
> getBeforeCreateRelationshipCommand do not get executed. Do you have a
> idea why this methods do not get called?
>
> In the edit helper the methods like getCreateRelationshipCommand or
> getCreateCommand are not executed either. Only createCommand is
> executed. Do you know what can cause this?
As far as I can tell from the javadoc getInsteadCommand has to delegate
to the other more specific classes in the editHelper. Chances are that
same applies for getBeforeConfigureCommand, but to be honest this is
just a guess.
> If I understood right the xxxcreateRelationshipCommands schould be
> called in my case as I am adding a link from the tool box to the diagthe IEditCommandRequestram.
Same as above.
> Another problem is that I cannot modify the ElementsToEdit list of the
> request to add/remove elements I want to have added to my diagram.
I am not quite shure, but the request may be immutable. But even if
that's the case there shouldn't be a problem with this as you build the
command on your own, where you don't have to use anything out of the
request if not needed.
> I could not find a very detailed description of all those techniques I
> used. Do you know where to find a good documentation. For example I
> could find nothing about instead advices.
Yes it is actually really hard to find good
documentation/literature/tutorials (that are not outdated) for many
things concerning gmf. I'm sorry but I don't know a single good
documentation either. Most of what I think to know I learned from others
and even that may be wrong, bad practices or outdated. :/ The instead
advice was mentioned in this german wiki:
https://sdqweb.ipd.uka.de/wiki/GMF_Tutorial:_Elemente_konfig urieren_w%C3%A4hrend_sie_angelegt_werden
Greetings,
Christian
|
|
| | |
| Re: ER like connections [message #642622 is a reply to message #642562] |
Thu, 02 December 2010 03:05 |
 Michael Golubev
Messages: 383
Registered: July 2009 |
Senior Member |
|
|
Hello,
I have posted detailed explanations, and the links to the sample code at Google docs, at https://docs.google.com/document/d/11hWszd3fyE2tRGIe08NQFXEj NraZc3r5ztBC4ruMr-U/edit?hl=en#.
Just to enable searching, here is the copy:
As you probably got by now there are no one-line solutions to this use case. The only good news is that I am going to try to push this solution to GMF-tooling, and with a bit of luck it will be possible to define and generate the code below in Indigo. To be honest, it is one of the reason for me to describe this somewhat long story here.
The complete code is available in the public SVN repoistory at http://www.assembla.com/code/gmf-association-class-links/sub version/nodes
There are 4 plugins, 2 of them are just pure EMF plugins (inthemiddle and inthemiddle.edit, nothing interesting there), + supporting .gmf plugin containing helper code, GMF artifacts and custom templates, and finally the resulting .diagram plugin.
There is very simple diagram definition in the *.gmf/gmf/def/ folder. There is nothing special here, except that we set the incoming link creations to true for both GenLink's in the gmfgen model, to allow user to start drawing of the link from the Rectangle instead of not-yet-created Rhomb.
Also I was somewhat lazy to play with gmfgraph today, so don't be surprised that Rhomb model elements are rendered as rounded rectangles at the diagram  .
Proposed solution consists of 2 parts:
a) modifying the GraphicalNodeEditPolicy (GNEP) to collect and pass the diagram information into the semantic creation command and
b) implementing the composite creation command itself
a) So we are going to replace org.eclipse.gmf.runtime.diagram.ui.editpolicies.GraphicalNod eEditPolicy with tweaked ITMGraphicalNodeEditPolicy from the *.gmf plugin. The changes made are as follows:
- Default GNEP packs the CompositeTransactionalCommand and then uses the ordering to access the different parts of the pack. E.g, lines 508-515 in the default GNEP looks like:
| Quote: |
CompositeCommand cc = (CompositeCommand) proxy.getICommand();
//....
Iterator commandItr = cc.iterator();
commandItr.next(); //0
SetConnectionEndsCommand sceCommand = (SetConnectionEndsCommand) commandItr.next(); //1
sceCommand.setNewTargetAdaptor(new EObjectAdapter(((IGraphicalEditPart) targetEP).getNotationView()));
SetConnectionAnchorsCommand scaCommand = (SetConnectionAnchorsCommand) commandItr.next(); //2
//....
|
This is clearly bad, so the changed code introduces ITMCreateLinkCommand with explicit access to its parts. It also has a code to register the results pack in the parameters of original request, so our semantic command will be able to access the whole pack at the execution time.
The changed code also computes an additional ITMCreateParameters describing the diagram information about the link source and target, but, while useful in general, it may be out of scope of this particular use case.
To register all the ITM* staff above in the generated diagram I used the following custom templates (*.gmf plugin, /gmf/templates/...)
itm::graphicalnodeep::Guard.qvto
it is the guard that prevents generation of the new code for unrelated diagrams (I had more than 1):
| Quote: |
helper requiresSpecificGraphicalNodeEditPolicy(diagram : genModel::GenDiagram) : Boolean {
return
'your.diagram.plugin.id = diagram.editorGen.plugin.iD
}
|
This guard allows every other affected template to have something like:
| Quote: |
«EXTENSION itm::graphicalnodeep::Guard»
...
«IF requiresSpecificGraphicalNodeEditPolicy(self.getDiagram())-
|
|
|
| Re: ER like connections [message #643144 is a reply to message #641843] |
Sun, 05 December 2010 09:15 |
Ralph Gerbig
Messages: 702
Registered: November 2009 |
Senior Member |
|
|
Hi,
this does exactly the thing I need. However, it is a lot of source code and template which make the source code really hard to understand. I managed to get I work by executing the dynamic templates, edit the guard expression and modify Explicit Rhomb and my xxxSemanticItemPolicy as describe in the article in you google docs This last part from google docs is missing here. I needed a few days to get this work and could not completely understand what you are doing, because I am completely new to all of this. My main questions have a main foundational character.
1) What is the NodeEditingPolicy for?
2) Is all that code needed to solve the problem or could it be done in a shorter but less sophisticated way?
Maybe you should think about posting into the GMF Tips Wiki (http://wiki.eclipse.org/GMF_Tips) and submit you code to the Eclipse CVS or publish it at http://code.google.com/p/gmftools/.
Thank you
Ralph
[Updated on: Sun, 05 December 2010 09:19] Report message to a moderator |
|
| | | | | | | | |
| Re: ER like connections [message #643782 is a reply to message #643662] |
Wed, 08 December 2010 09:40 |
Ralph Gerbig
Messages: 702
Registered: November 2009 |
Senior Member |
|
|
Hi,
I found the errors when comparing my diagram xmi to the diagram xmi you plug in produces. I marked the differences in bold:
Your XMI:
<edges xmi:type="notation:Connector" xmi:id="_gP7u8AFPEeClrNQa80wp6g" type="4002" source="_gPx98AFPEeClrNQa80wp6g" target="_dPkVQAFPEeClrNQa80wp6g">
<styles xmi:type="notation:FontStyle" xmi:id="_gP7u8QFPEeClrNQa80wp6g" fontName="Segoe UI"/>
<element xsi:nil="true"/>
<bendpoints xmi:type="notation:RelativeBendpoints" xmi:id="_gP7u8gFPEeClrNQa80wp6g" points="[0, 0, 0, 0]$[0, 0, 0, 0]"/>
</edges>
<edges xmi:type="notation:Connector" xmi:id="_gQrV0AFPEeClrNQa80wp6g" type="4001" source="_gPx98AFPEeClrNQa80wp6g" target="_c_M-AAFPEeClrNQa80wp6g">
<styles xmi:type="notation:FontStyle" xmi:id="_gQrV0QFPEeClrNQa80wp6g" fontName="Segoe UI"/>
<element xsi:nil="true"/>
<bendpoints xmi:type="notation:RelativeBendpoints" xmi:id="_gQrV0gFPEeClrNQa80wp6g" points="[0, 0, -372, -173]$[352, 163, -20, -10]"/>
</edges>
My Code:
<edges xmi:type="notation:Connector" xmi:id="_AAXpIAKtEeC4ouyRqPLhGA" type="4005" source="_GehNwAKsEeC4ouyRqPLhGA" target="_JyXtIAJJEeCucI_oimNlnA">
<styles xmi:type="notation:FontStyle" xmi:id="_AAXpIQKtEeC4ouyRqPLhGA" fontName="Segoe UI"/>
<element xsi:nil="true"/>
<bendpoints xmi:type="notation:RelativeBendpoints" xmi:id="_AAXpIgKtEeC4ouyRqPLhGA" points="[0, 0, 0, 0]$[0, 0, 0, 0]"/>
</edges>
<edges xmi:type="notation:Connector" xmi:id="_O6dC0AKtEeC4ouyRqPLhGA" type="4005" source="_Jhjp8AJJEeCucI_oimNlnA" target="_GehNwAKsEeC4ouyRqPLhGA">
<styles xmi:type="notation:FontStyle" xmi:id="_O6dC0QKtEeC4ouyRqPLhGA" fontName="Segoe UI"/>
<element xsi:nil="true"/>
<bendpoints xmi:type="notation:RelativeBendpoints" xmi:id="_O6dC0gKtEeC4ouyRqPLhGA" points="[52, 1, -342, -9]$[340, -12, -54, -22]"/>
<sourceAnchor xmi:type="notation:IdentityAnchor" xmi:id="_O6mMwAKtEeC4ouyRqPLhGA" id="(0.8306188925081434,0.1702127659574468)"/>
<targetAnchor xmi:type="notation:IdentityAnchor" xmi:id="_O6mMwQKtEeC4ouyRqPLhGA" id="(0.30793650793650795,0.23404255319148937)"/>
</edges>
It looks like in my case something goes fundamentally wrong but I cannot find the place when debugging everythin looks just fine. However, it creates only the target shapes (rhombRight) instead of a target and a source shape (rhombLeft). Secondly my second generated link has the source set in the target attribute. Do you know where this can go wrong in the code?
Thanks in advance,
Ralph
|
|
| | | | | | |
Goto Forum:
Current Time: Wed Apr 01 22:13:45 GMT 2020
Powered by FUDForum. Page generated in 0.06727 seconds |
Back to the top
|
Search
Help
Register
Login
Home
