[Bf-committers] Patch for bug 1355

[Ton Roosendaal]

Hi,

If you check on the demo file in that bug, you see it has a concave  
part too, that's where the original code failed... convex is much  
simpler!

-Ton-

On Sunday, Jun 13, 2004, at 15:14 Europe/Amsterdam,  
jeremy@marzhillstudios.com wrote:

> Hi,
>
> While doing some googling I found this method of determining whether
> a normal points in or out for convex objects.
>
> http://www.doc.ic.ac.uk/~dfg/graphics/GraphicsSlides06.pdf
>
>  It's similar to my adjacent face method except I think this one
> might work for all convex meshes and with 4 or 3 point faces. I'm
> going to see about implementing it later on today or on tuesday if I
> don't have the time today.
> On 13 Jun 2004 at 12:26, Ton Roosendaal wrote:
>
>> Hi,
>>
>> Although the method used is very simple, it has the advantage of the
>> simplicity too. Carefully examining the sample file (in bug report)  
>> and
>> what the code does, some aspects pop up:
>>
>> - the Mesh is highly irregular (face centroids at unreliable places)
>> - the Mesh has non-flat quads
>>
>> and:
>>
>> - the code doesn't take quads into account at all
>> - the code uses (cpu cheapness) manhattan distance, no real distance.
>>
>> You could have checked that converting the error file to triangles
>> already worked, but not in all cases... for the latter the manhattan
>> distance solves it.
>>
>> Most important was just adding code to evaluate quads properly,  
>> meaning
>> treating them as 2 individual triangles, since the quads can be
>> non-flat. Check the commit I just did.
>>
>> Still, the used system is simple, and I bet you can find some
>> situations it won't work for. An improved system only can work when  
>> it:
>>
>> - separates parts first with a sortof select-connected
>> - from each part, find the face that has least intersections with  
>> other
>> faces, with one of its possible normals (both face sides).
>>
>> Which all will be quite costy, and I propose we could wait for problem
>> files & evidence first. :)
>>
>> -Ton-
>>
>>
>> On Saturday, Jun 12, 2004, at 21:57 Europe/Amsterdam,
>> jeremy@marzhillstudios.com wrote:
>>
>>> On 12 Jun 2004 at 14:12, Jiri Hnidek wrote:
>>>
>>>> Hi,
>>>> it works for me, but can you tell me: "What is meaning of variable
>>>> dotcent?" ;-)
>>>
>>> I was fiddling with the feature some more today and noticed it still
>>> doesn't work for all meshes, particularly a uvsphere mesh.
>>>
>>> I can make it work for the sphere mesh but then it won't work for the
>>> bug blend. Basically I had two possible ways to calculate whether a
>>> normal was facing in or out. One used the mesh centroid and the other
>>> used the centroid of an adjacent face and the test face.
>>>
>>> The adjacent face method worked for sphere but not test blend. The
>>> mesh centroid worked for test blend but not sphere. Anyone got a
>>> suggestion as to how to proceed from here? Its possible we migh need
>>> to offer a selection of algorithms for this function. Different types
>>> of meshes need different types of methods.
>>>
>>> Jeremy Wall
>>> _______________________________________________
>>> Bf-committers mailing list
>>> Bf-committers@blender.org
>>> http://www.blender.org/mailman/listinfo/bf-committers
>>>
>>>
>> ---------------------------------------------------------------------- 
>> --
>> --
>> Ton Roosendaal  Blender Foundation ton@blender.org
>> http://www.blender.org
>>
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers@blender.org
>> http://www.blender.org/mailman/listinfo/bf-committers
>>
>>
>>
>
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers@blender.org
> http://www.blender.org/mailman/listinfo/bf-committers
>
>
------------------------------------------------------------------------ 
--
Ton Roosendaal  Blender Foundation ton@blender.org  
http://www.blender.org