Re: [Jmol-developers] space partitioning & auto-bonding distances

[E.L. W. (Egon) <eg...@sc...>]

On Friday 23 May 2003 09:02, Miguel wrote:
> To solve the "rebonding" problem I built a prototype of a binary space
> partitioning tree to hold the points. This is an effecient data structure
> which provides access to points which are "near" other points.
>
> It has not yet been integrated into cdk/jmol, but is a standalone
> prototype.

This should go into CDK.

> In our case, the BSP tree divides the 3-space into little blocks. You can
> then use an iterator to enumerate all the points within those blocks.
>
> The API is currently
>   Enumeration enumDelta(Point3D center, double delta)
>
> This enumerator is returns all the points that are within the cube defined
> by:
>   center.x +/- delta
>   center.y +/- delta
>   center.z +/- delta
>
> Two points
>  1) this is a cube, not a sphere
>  2) because of way the data structure works, some points outside
>     this cube are returned also ... basically the contents of all
>     little boxes defined by the bsp tree.
>
> It would be easy for me to make another variant of this iterator which
> restricts to all the points that are within the sphere. But this may not
> be worth it if we are going to turn around and do some more distance
> calculations with the result.

Blocks should be fine. One question though... what about two neighboring atoms 
that fall into two different block, because the are distributed just around 
the edge?

> Questions:
> Does the auto-bonding distance depend upon the types of the atoms involved?
> *Should* it depend upon the types of the atoms involved?

Yes, it should (and does?) depend on the Vanderwaals radii...

Egon