← Back to team overview

yade-users team mailing list archive

Re: Arbitrary aligned box or plane

 

Hi Václav!

> > Well, use pack.regularHexa(...) with pack.inAlignedBox. Is there a way to
> > create such packings with pack.SpherePack()? Or how can I convert it in a
> > SpherePack()? I tried to pass the regularHexa packing to the SpherePack,
> > dosen't work. If this is going to work somehow it would be enough for my
> > purposes.
> 
> Oh, ding it -- I thought pack's packers (regularHexa etc) were returning
> SpherePack instances, but in reality they return list of bodies already
> (I should know, since it is my code, sorry)!
> 
> Now, I think it would be more logical for them to return SpherePack's,
> since they are just centers+positions, what one would expect (perhaps
> there are some issues why I decided to return bodies directly that I
> don't see now), and then one might to call SpherePack.toSimulation() to
> add them to O.bodies (or a similar function). (I would add some
> additional argument, to make it backwards-compatible for some time, so
> that old code would work as it did and give warning about future break).
> Are you using trunk? Would it be useful for you?

I am using the 0.5 release right now. But I think that I will move to the 
trunk soon. Well, sure it would be useful for me, but there is no hurry since 
what you proposed below is working for now.

> Meanwhile, you can change spheres' positions by hand in a loop, like this:
> 
> from yade import pack
> pred=pack.inAlignedBox((10.,0.,0.),(20,20,1.5))
> hexapack=pack.regularHexa(pred,radius=0.5,gap=0.0,color=(0,1,0))
> # define around which point to rotate, and the rotation itself
> p0=Vector3(10,10,10); rot=Quaternion((0,1,0),-pi/4)
> # apply to all spheres in the packing
> for b in hexapack: b.state.pos=p0+rot*(b.state.pos-p0)
> # add to the simulation (it does not matter if you add particles to
> simulation first or rotate them first, really)
> O.bodies.append(hexapack)

This is working!

> >> The reason I did not create general box orientation is that checking
> >> whether particle is inside or outside is more complicated; but now I
> >> have the idea that it could be actually done by keeping corners and box
> >> transformation in the predicate, then transforming points to be checked,
> >> rather than checking points against true (non-aligned) boundaries.
> > 
> > Hm, I am a newbie and I still have to understand the difference between
> > the two packing options. By the way, what's the difference? Can't you
> > just use predicates?
> 
> The two have orthogonal meanings. Packers defines the pattern (texture,
> sphere configuration, ... -- however you would like to call it) whereas
> predicate defines the volume where it will be created. E.g. hexagonal
> packer will create hexagonal configuration in volume given by
> inAlignedBox, but it can be created in a cylinder, in a volume defined
> as intersection of box with sphere, and so on.

Thanks.

> (That makes me think that I could also add something like inHalfSpace
> (oriented clipping plane) predicate, then you could create e.g. box and
> clip it by 2 arbitrary planes from each side; or define a box-like
> volume with 6 of such planes.)

Good idea!



References