Most significantly, I have found the cause of a problem I've been having (this has to be the most ambiguous line ever). The problem was graph layouts, not quite matching the shape they should have, which without edges is quite difficult to gauge in large graphs. The edge drawing mechanism was to help with this, but due to time and other priorities, completion of this has been postponed. A temporary line drawing mechanism (similar to that used when i first started) has been implemented, allowing me to see what's actually going on... and good god its a mess.
It appears the layout gets very confused, and starts "flicking" out in the wrong direction causing crossed edges (which clearly shouldn't be happening). This has only occurred since my attempt at moduli sing the work, and has since been blamed on a few changes to my implementation of eades algorithm (which worked before). The edges now being drawn show that the problem may also be down to incorrect edges being drawn, and possibly a result of a change in my data structure. Nothing has been been shown to be the cause yet, but when it is found, I will mention it in another blog post.
There is not much else to report, other than tens of pages of notes and a busy looking NetBeans, most of the time has been spent analysing and attempting to resolve the issue with the graphs. I will update soon. I promise. Please don't leave me.
Below is a picture showing some more output, specifically, the difference of new algorithm layout vs old algorithm output and what this problem is.
The left image is of the old implementation, updated to include edges and animation, whereas the new implementation on the right, which only differs in that the data structure is different. Both are of a cube, can you tell which has the better layout?
No comments:
Post a Comment