Preserving a path when deleting a point



  • RF could do better with preserving a path's curvature while deleting a point. Deleting a curve point with two BCPs on a lengthy outline does quite well. But when I've got a corner point close to a tangent point (with one BCP) and I remove that tangent point, RF fails to follow the former path (FL does a much more accurate job), as BCP drops away and is not glued into a corner point (standard FL behavior). I would love to have this included in the next version of RF. Any chance?



  • These images probably explain this better: erasing a tangent point in #1 gives #2 -> BCP gets killed while it should stay alive!



  • Am I alone with this bug? Where are my soul mates?


  • admin

    he,

    you're not alone 🙂



  • Frederik, did something change in this behavior from RF 1.2 to RF 1.3? I have a feeling it got worse, but maybe I'm wrong. Here you can compare the difference between FL 5.1.2 and RF 1.3, same PS curves on both cases. Radical...


  • admin

    Some values get limited to prevent extreme results, like jumping bcp with more then 5000 units.

    enjoy!



  • But RF could do a much better job in preserving paths, like in the example above. That example doesn't contain any extreme values. What's the reason RF's paths are so far off the original path? And which value get limited?



  • I don't know anything about bezier maths, but there must be a code available somewhere (in a pocket of Adam T?) to improve this behavior.


  • admin

    I wish there was such a pocket....

    The examples shows a point that should be removed that is rather close to one side of the newly generated curve. In the next version there will be no limit if the removed point has smooth bcp's, as in your example.



  • I agree with you guys. My way of drawing involve a lot of "add extreme points + cleanup" method, and it is… painful to work like this in RF. Just a suggestion: on removing a point, maybe the opposite handle should be kept as is, at least? If recalculation of curve is hard, I don't see a reason why both handlelength should be affected. Am I clear?
    (and yes, FLS' "optimize" algorithm performs well)


Log in to reply
 

Looks like your connection to RoboFont ● Forum was lost, please wait while we try to reconnect.