yEd-to-SBGN: Left/right position of the connected glyphs should be considered for orientation of ports/operators

[ibalaur]

In case the connected glyphs are displayed in both left/right sides, the process/ operator may be wrongly oriented.
For example, in the attached, in a) the process is not well orientated because the bend points is too close to the process, it is "absorbed" by the process' port. However, in b) the process is well oriented given that the bend point "coordinated" the process orientation.
Bend-point-absorbed-by-the-process-port.zip
Input:

Output:

[amazein]

Maybe, for a more sophisticated algorithm, two sides of the equations should be seen as kind of two sums of the vectors or using a similar logic., not just focusing on one connector per time? I do not know if it is going to be heavy on calculations for many cases, probably it could slow down the conversion...

[amazein]

It is also possible that is not always possible to predict the satisfying orientation...

[amazein]

One more thought: we do not like orientations like that when we see sharp angles near the ports. Probably this is the key? Too many sharp angles, consider the alternative 90 degree position, compare the sharpness in two case? Might be computationally expensive too.
It is an interesting problem, but, in general, the layout is satisfying is most cases.
I am also forgetting that we do not need to guess the position: maybe we should focus on how to get information from the original graph where we already have the orientation.