[Bf-committers] Re: Re: Re: [Bf-blender-cvs] CVS commit: blender/source/blender/blenkernel BKE_bad_level_calls.h BKE_global.h blender/source/blender/blenkernel/bad_level_call_stubs stubs.c blender/source/blender/blenkernel/intern blender.c exotic.c scene.c blender/source/blender/blenlib BLI_editVert.h blender/source/blender/blenloader/intern readfile.c ...

GSR - FR famrom at infernal-iceberg.com
Sun Sep 26 22:47:54 CEST 2004

theeth at yahoo.com (2004-09-26 at 1223.00 -0700):
> I haven't thought of PET much so I'm open for
> suggestions.

Well, pure distance based would check non selected vertices distance
to nearest selected as decisor for how much to be affected. It can be
speeded up from the brute force "test every unselect against all
selected" using bounding box first to discard far away vertices and
then doing the test in such way that before doing the slow vector
calc, it has to pass 3 absolute value test (if any coord pair delta,
X1 X2, Y1 Y2 or Z1 Z2, is bigger than the effect radius, you can be
sure vertex 1 is not going to be in effect radius of vertex 2 at all).

The linked plus distance method would be another way to speed up the
decisions, it would discard based if they not they are directly linked
(or by N jumps), like grow selection, and use the distance of the link
(edge). Though anyone can see the results will be really different,
think about a flatteded thing, modifying one side will hardly affect
the other, even if distance would say it has to.

I point this cos if you do a 32*32 grid and select four vertices, one
each near the middle of each side (but not in the edge) and them try
to apply PET, it is far from what one would expects after reading
"magnetism" (IIRC that is the word used in the docs). I think it was
also pointed with an example in which rotation respect to axis
affected the result (both are the same problem, current PET is done
with a bound box of the selection).

Probably a math expert knows better ways to compute min distance from
one vertex to a group of other vertex, or some other method less rough
and problematic than the bound box one.


More information about the Bf-committers mailing list