Faces

[dragoonreas]

Code: [Select]

t = ((i * (Px1 - Lx1)) + (j * (Py1 - Ly1)) + (k * (Pz1 - Lz1))) / ((i * (Lx2 - Lx1)) + (j * (Ly2 - Ly1)) + (k * (Lz2 - Lz1)))

Perhaps this?

Where did tx, ty and tz go?

Code: [Select]

t = ((i * (Px1 - Lx1)) + (j * (Py1 - Ly1)) + (k * (Pz1 - Lz1))) / ((i * tx) + (j * ty) + (k * tz))That's the answer I got using this site.

(click to show/hide)

Code: [Select]

0 = (i * (((t * tx) + Lx1) - Px1)) + (j * (((t * ty) + Ly1) - Py1)) + (k * (((t * tz) + Lz1) - Pz1))
u=tx
v=ty
w=tz
m=Lx1
n=Ly1
o=Lz1
q=Px1
r=Py1
s=Pz1
0 = (i * (((t * u) + m) - q)) + (j * (((t * v) + n) - r)) + (k * (((t * w) + o) - s))                    <-- Input for the website
t = (-j * n + j * r - k * o + k * s - i * m + i * q) / (j * v + k * w + i * u)                           <-- Output from the website
t = (-j * Ly1 + j * Py1 - k * Lz1 + k * Pz1 - i * Lx1 + i * Px1) / (j * ty + k * tz + i * tx)
t = ((i * (Px1 - Lx1)) + (j * (Py1 - Ly1)) + (k * (Pz1 - Lz1))) / ((i * tx) + (j * ty) + (k * tz))

[annikk.exe]

tx = Lx2 - Lx1

So our answers are actually the same.

It seems like we are dividing one dot product by another.  That's how a lot of people seem to be doing it..

[annikk.exe]

Lets have a look at ALL the code that governs this:

Code: [Select]

-- get plane coordinates
Px1 = vertex3dXtransformed[firstvertex]
Py1 = vertex3dYtransformed[firstvertex]
Pz1 = vertex3dZtransformed[firstvertex]

Px2 = vertex3dXtransformed[secondvertex]
Py2 = vertex3dYtransformed[secondvertex]
Pz2 = vertex3dZtransformed[secondvertex]

Px3 = vertex3dXtransformed[thirdvertex]
Py3 = vertex3dYtransformed[thirdvertex]
Pz3 = vertex3dZtransformed[thirdvertex]


-- get line coordinates
Lx1 = GetCameraX()
Ly1 = GetCameraY()
Lz1 = CameraZ

Lx2 = vertex3dXtransformed[vertexID]
Ly2 = vertex3dYtransformed[vertexID]
Lz2 = vertex3dZtransformed[vertexID]


-- get 2 vectors along the plane
-- PV1x stands for Plane Vector #1, X-coordinate

PV1x = Px2 - Px1
PV1y = Py2 - Py1
PV1z = Pz2 - Pz1

PV2x = Px3 - Px1
PV2y = Py3 - Py1
PV2z = Pz3 - Pz1

-- now find the cross product, which is a line perpendicular to the plane
-- PV1 X PV2 = normal vector!
i = (PV1y * PV2z) - (PV1z * PV2y)
j = (PV1z * PV2x) - (PV1x * PV2z)
k = (PV1x * PV2y) - (PV1y * PV2x)

-- for the line, calculate tx, ty, tz
tx = Lx2 - Lx1
ty = Ly2 - Ly1
tz = Lz2 - Lz1

t = ((i * (Lx1 - Px1)) + (j * (Ly1 - Py1)) + (k * (Lz1 - Pz1))) / ((i * (Lx2 - Lx1)) + (j * (Ly2 - Ly1)) + (k * (Lz2 - Lz1)))

intersectX = (t * tx) + Lx1
intersectY = (t * ty) + Ly1
intersectZ = (t * tz) + Lz1


-- calculate distance from intersection
intersectdistance2d = (GetCameraX() - intersectX)^2 + (GetCameraY() - intersectY)^2
intersectdistance3d = math.sqrt((CameraZ - intersectZ)^2 + (intersectdistance2d))


if intersectdistance3d > distance3d then
-- return false if the vertex is in front of the face
return false
else
-- return true if the plane is closer than the vertex
return true
end

[annikk.exe]

I've tried 2 different versions of t.
One like this:

Code: [Select]

t = ((i * (Lx1 - Px1)) + (j * (Ly1 - Py1)) + (k * (Lz1 - Pz1))) / ((i * (Lx2 - Lx1)) + (j * (Ly2 - Ly1)) + (k * (Lz2 - Lz1)))

And one like this:

Code: [Select]

t = ((i * (Px1 - Lx1)) + (j * (Py1 - Ly1)) + (k * (Pz1 - Lz1))) / ((i * (Lx2 - Lx1)) + (j * (Ly2 - Ly1)) + (k * (Lz2 - Lz1)))
Neither is working :<  But we're getting close now, I think...

[Bonobo]

Ouf … http://www.myalgebra.com/algebra_solver.aspx shows two different solutions w/ mighty big square roots




<edit>

Dang. I missed a whole page of posts …

</edit>

[Pilchard123]

Are tx, ty and tz their own thing, or t*x, t*y, t*z? It has a _*massive*_ effect on the outcome.

To me, the t*tx is the same as (t^2)x, and that is what most other people would say as well.

What are you even trying to do?

[annikk.exe]

"tx" is NOT "t * x".
"tx" is a unique variable.  It represents the x-component of the vector of the line from L1 to L2.  I gave it a short name instead of something more descriptive to keep the calculations reasonably short.


I've had a breakthrough, though not where I was expecting.
I found a problem with my plane-overlapping-in-2D code.  It doesn't work well with horizontal or vertical edges.  Testing and writing code to trap for it now...  Seems like the only reason I thought it was working correctly before is because I had my test cube rotating constantly.

[annikk.exe]

I've almost cracked it.  Trapping for horizontal and vertical seems to work fine, and the vertices of static objects behave correctly now.

There's still a really bizarre problem, though.  The vertices on the right hand side of each static cube flicker.  I have no idea why.  Moving or zooming the camera seems to "excite" them to flicker more.

[Aino]

Good to hear

I hope to make some mor emodels, it's very fun, maybe I'll even make an editor for it when you're done, who knows...

[annikk.exe]

A girl from another forum I post on has taken a look at the maths for the line-through-plane check and confirmed this is what the calculation for t should be:

Code: [Select]

t = ((i * (Px1 - Lx1)) + (j * (Py1 - Ly1)) + (k * (Pz1 - Lz1))) / ((i * (Lx2 - Lx1)) + (j * (Ly2 - Ly1)) + (k * (Lz2 - Lz1)))
She studies physics and this stuff is apparently quite simple for her so hopefully it's gonna work this time.  :>

So once I get this flickering issue sorted out, I'll have another crack at that and see if I can get it to work with expected behaviour.  :>

Now, what is up with this flickering, hmm?  It's bizarro.

[Aino]

Ohh, wow

How many more cracks are there?

[annikk.exe]

If by cracks you mean attempts, that number is undefined.. :p

[Aino]

Sounds scary...

Gonna do a little more relaxing things for now: creating a risk map, a proper risk, not the eufloria map I released

[annikk.exe]

This bug is really annoying.  I can't figure out why it produces this behaviour.  Very strange....

[Aino]

I know that feeling :/

Adventure Map is so annoying with that one bug... If I kept the Maps internaly in the file, it would get alot simplier, but harder to create maps...