Mesh Fill Holes - Best strategy?

:slight_smile: I’ve tried them all (the mesh I posted was initialy made isotropic with OpenFlipper, which is a remarkable good tool. Some of its features are outstanding) . In this particular case however I need something quick and lightweight, something moped-like which does the job. :wink:

But while at it; A year ago I actually made a GH componet wrapping MeshLab (server), but the meshlabserver.exe binary has a fatal bug, and I don’t speak C++ and… but never mind. For this reason it’s currently not useful, which I regret. Meshlab is a powerful tool for mesh repair. Same irritating thing (bugs) in several other tool-binaries as well renders my wrappers useless in GH, Instant Meshes works though (screenshot from my grasshopper mesh tools menu):

If they were functional and well tested I would have shared them on Food4Rhino long ago.

Thank you for the links on mesh algorithms. Added to the collection. :slight_smile: I’m actually familiar with the basic concepts involved (like half-edge meshes etc), I just have not yet done some programming using it, not yet strated walking the walk, but that day is coming closer. So far it has only been a matter of priority (I have an ongoing project and so far existing tools has been a priority to get things done). But I will definitely return to this subject in due time.

BTW, my son defended his thesis and got his title.

// Rolf

Title (congrats) means something … but a proper lethal Panigale means everything


Encouraged from what Lewis did in Russia (well … not exactly good old racing … but we live in plastic is fantastic Times) spend 30 minutes on the fill concept briefly out lined in the first paper. It “works” (kinda)… but for make it handle all situations one needs at least 3 whole days (min) + a zillion of test cases + Karma + several decent cigars. But truth is that I hate meshes (but this may serve in truss design as well … meaning that IF Lewis wins in Japan next w/e … expect the real thing soon).

Here’s some tests on that WIP matter (but what is the midFactor var? Answers: The Lord, District 666, North Pole):

1 Like

And here’s why one needs 3 days instead of 30 minutes to manage the thing in all situations (read the Method in the paper to understand what’s happeneing) :

Here’s a step by step test (but Long is the Path (and hilly)):

1 Like

The same test poly but with a different midFactor value: controls how the new faces (per loop) are made in relation to a percentage of the adjacent angle edges. The challenge is to make the recursion smart meaning: fill the hole with a mesh as much “the same” as possible VS the OEM mesh (and/or change the OEM mesh as well).

Moral: Go Lewis, go … only 5 races left.


Great progress: ~1ms for solving any “reasonable” (hope dies last) complexity polyline hole + 66.6 bugs fixed + far more compact code (666 lines become 665) + only 3 remaining divisions by zero (but is this a big deal?).


Testing against the Ultimate (i.e. The Lord [the Merciless King of Sardineland]).

Here’s the 1M millidollar dilemma: should we do an always doable - but a “bit” ugly - mesh (midFactor ~1.0) and then write some lines of code for relaxing the mesh via K2 (Daniel had exposed the K2 Methods you know) … or … mastermind some smart (???) stuff more and finish the task without K2?

This is a proper challenge for my top IT guru I confess.


Top Guru said: you idiot Lord > you use Dot, Cross and the likes in order to manage what’s inside (and convex, concave stuff) instead of sorted MTV Indices. See a totally new WIP approach that goes from 2 ms to 1.3 > take that you idiot.

Lord said: you idiot > I did that having polylines and MERO trusses in mind (because I hate meshes and holes and no holes and whatever). Take that you idiot .