[Bf-committers] CVS commit: blender/source/blender/blenkernel BKE_booleanops.h blender/source/blender/src booleanops.c

[Laurence Bourn]

Umm, no.
However I have some ideas on how the current implementation could be coerced
into something similar.

Basically the algorithm falls in 3 major pieces.
1) Efficiently find intersecting polygons or (so a rough set including all
the intersecting polygons is a good start)
2) Split all such polygons with their spplitters from the other object (The
crucial part here is you can do this in any order) 
3) classify all the polygons in the new split object as either in or out the
other object. Using the fact that they are either in or out and not
straddling. 

This differs from the current implementation only in step 2. Where we used a
BSP tree. We can still use a BSP tree for parts 1 and 3 (infact the current
code does exactly the first step). Step 3 is pretty easy to. Just a ray BSP
tree intersection test and a cunning neighbourhood flood fill to speed
things up a bit. 

In fact this is probably a bit easier todo than the current implementation
and you are guarenteed only to split polygons only when it's absolutely
necessary (well give or take a few cases). 

Why I didn't do it this way? I must have been mentally confused by the
amount of cool Jazz Strubi used to play on his NaN office piano!

Unfortunately I don't have much time at all for this kind of stuff but it
would be nice to finally slay this beast once and for all.

Cheers,
Laurence.



-----Original Message-----
From: LarstiQ [mailto:larstiq@larstiq.dyndns.org]
Sent: maandag 24 maart 2003 17:33
To: 'bf-committers@blender.org'
Subject: Re: [Bf-committers] CVS commit:
blender/source/blender/blenkernel BKE_booleanops.h
blender/source/blender/src booleanops.c


On Mon, Mar 24, 2003 at 05:17:59PM +0100, Laurence Bourn wrote:
> Hiya Daniel,
> Good to see you working on the Sauces hombre!
> 
> Vertex normals on boolean mangled meshes look bad coz the polygons in the
> neighbourhood of the intersection between the objects are of poor quality
> (long thin etc). It is much better to make the boolean operation itself
> compute a new vertex normal each time it splits a face. It shouldn't be
that
> hard to convince the boolean op stuff in intern to do this. Maybe I should
> do it? All that BSP code was such a bad idea for doing boolean ops much
> better to use the algorithm here:
> 
>
<http://www.flipcode.com/cgi-bin/msg.cgi?showThread=13November2000-Construct
> iveSolidGeometry&forum=askmid&id=-1>

That looks good to me, also a bit of work tho :) Do you want to
implement all that?

LarstiQ
_______________________________________________
Bf-committers mailing list
Bf-committers@blender.org
http://www.blender.org/mailman/listinfo/bf-committers