Improving ECC transition routing

[vikash1703]

Hi @azoitl and 4diac-ide Community

As suggested to there is a problem in transition when drawn between two states it bends ,The currently applied spline has a strange trajectory at the start and end which need improvement and Also source and end location calculation will need some improvements.

I have been looking into the transition routing issue. I found two problems:
First in CreateTransitionCommand the sourceLocation and destLocation are set using toScreenPoint() which gives the top-left corner of the state instead of the center. I think this bendpoint is calculated from the wrong position. and the second is
in SplineConnection.getPath(), the control points for the cubic spline are calculated using startMid and midEnd vectors. When the destination state are drawn directly below these vectors cause a strange hook at the start of the transition.

For the location fix, I think the center can be obtained from ECStateEditPart using nameLabel.getBounds().getCenter(). But the viewer is only available in ECCEditorEditDomain, not in CreateTransitionCommand. Would it be correct to pass the viewer to CreateTransitionCommand to get the figure bounds?
For the spline fix, I am not sure about the best approach. Could you give some guidance?"

This is the finding which on which i start work when the existing PR's will get merged.

[vikash1703]

This is the 200% zoomed image I hope it will help to understand

[azoitl]

@vikash1703 there are two things that I want to point out in your investigation on this.

1. the source and destination location are only used to get the start and end state of the transition. It think that we are using here positions is just wrong and the code should be changed that it is just using ECStaes for the start and end. This will make the commands reusable and more independent from ui code.
2. For the real start and endpoints of the spline that is drawn GEF Classic uses ConnectionAnchors. An EditPart that has target or source connection endpoints provides these via the getTargetConnectionAnchor() and getSourceConnectionAnchor() method of the EditPart for the ECC Editor this is done in the ECStateEditPart. More information on the connection routing behavior can be found in the Eclipse GEF Classic Wiki.

[vikash1703]

@azoitl

Very thanks for this explaination now it get clear to me i will do as it is and update you.

[vikash1703]

After reading a second time your question I think I forgot one thing. In SplineConnection there shouldn't be any point calculations anymore. There based on the given seven points (the 2 green and the 5 orange) the Path for drawing should be created. I think with such a design the overall code will simplify a lot.

What you can also do to get better feedback is that in SplineConnection you draw in addition to the Path also the points. This can help you to find issues in the calculations of the points.

@azoitl

Hii today i tried as per your instruction i know what you want ,How spline and transition would like to draw i follow the approach i taken the refrence from the 3 wiki that you have provided what problem i face when i give the points it does not form the curved D line shape that you like and that must be formed when the user drag transition it somewhat form a sharp shape in one image you can see that the spline ki curved forming D line shape but when another transition drawn from the same state it refresh now i have understand that i need to calculate at what pixel my 7 points are to be placed sorry for this much disturbance,and big thanks for this much patience.

[azoitl]

@vikash1703 I expected that this will not be just straight forward. I like the screenshots with the dots. At lest for me they explain a lot. Also I think fixing the spline should be done after PR #2369. Because otherwise some control points will be hard to determine, or?

[vikash1703]

@vikash1703 I expected that this will not be just straight forward. I like the screenshots with the dots. At lest for me they explain a lot. Also I think fixing the spline should be done after PR #2369. Because otherwise some control points will be hard to determine, or?

@azoitl
Ooh. Then I will do it by tomorrow and after that I will start doing spline part I also worried about that that's why I am not touch my pr that I have raised

[vikash1703]

@vikash1703 I expected that this will not be just straight forward. I like the screenshots with the dots. At lest for me they explain a lot. Also I think fixing the spline should be done after PR #2369. Because otherwise some control points will be hard to determine, or?

@azoitl

Any suggestion for me !

[vikash1703]

@azoitl

I have a hakathon and and interview at infosys on 21 april thats why i am busy in that i will continue after 21 april.

[vikash1703]

@azoitl

Sorry for that much time I was gone I was busy in my college activities I have something on my exploration on spline I will discuss in chat tomorrow what we have arrived so far and what we can do and I will not take more time on this try to finish as fast as possible.

[azoitl]

@vikash1703 thx for the update. I know that you have other things on your plate. Looking forward to your next steps.

[vikash1703]

@azoitl

This is the whole story we are currently facing we have to make our p3,p4,p5 collinear but after doing this as well the curve taht we get is not that we want its sharp '>' shaped transition after dragging and where i am facing problem is the implementation part i have reached our transition become straigth. The important thing is that the logic is clear to me if i am not wrong but the implementation i am failing. What you think.

[azoitl]

@vikash1703 I'm not fully sure if I can follow you from the second screenshot I can say the control points 1, 2, 6 and 7 look very good. I'm also not sure about 3 and 5. However what you have shown in your second screenshot is a rather extreme example. IF you move the control point nearer to the states it should in general look better. But I guess this is not the full problem. Because we need some rule for 3 and 5. Maybe the first thing that we need to discuss is what should be the gradient of the slope through point 4. If we can come up with a good rule for that I think we are one step further

[azoitl]

As I was curios what my UX helpers would say to the problem I asked ChatGPT and Gemini. Both came to nearly the same solution.

• For the distance of the points 2 and 6 they suggest to use 0.3 the distance from 1 to 4 and 7 to 4, respectively
• For the gradient they suggest to use the average of the two vectors 1 -> 4 and 4 -> 7, for the average it will be interesting to use the correct direction.
• as distance for the points 3 and 5 the again suggest 0.3 the distance from 1 to 4 adn 7 to 4.

@vikash1703 what do you think of this proposal.