I'm trying to create a prodecural rollercoaster. I managed to include the pillars, but I want that, in some parts, the pillars don't show up, like in a loop:
How can I do this?
A side problem that I have is that my pillar index are based in the spline parameter, but would be better if were based in the "planar lenght" off my rollercoaster, so that the inclined sections do not have more pillars than the straight sections. Is it possible?
My nodes for the pillar:
I would solve it like this:
First I create an offset of the curve to get a direction vector from the original positions and the new positions.
Then I check if the angle of this direction vector compared to the vector $0, 0, -1$ is in the range of $\frac{\pi}2$.
If it does, then it is obviously a loop. In addition, I check to see if the positions of the curve points are above a certain value to the ground. This allows me not only to detect loops, but also to avoid placing pylons where the track rests on the bottom.
Next, I separate the curve segments and place their points on the ground, because this allows me to achieve a regular spacing of the pylons.
The line created in this way is then converted back into a curve, and divided into segments of a certain length. With this I have once determined the positions of the pylons.
Next, I create the pylons, as well as the track.
Lastly, I instantiate the pylons and create a raycast from the points up to the track to capture the distance/scaling of the pylons.
This example assumes the plan view of the curve-path for the roller coaster is constructed in XY, and raised to its various heights in Z. The incoming curve has its 'Twist' set to 'Minimum'.
This approach constructs another curve from the supplied path, adapted to be a base for your pillars.
.. discriminating by asking how aligned the curve's Normal x Tangent is aligned to object Z. If you delete/separate points from a curve, in GN, it just dissolves them. We want to delete them. That's why the curve is converted to a mesh, and back again. The mesh is deleted according to the curve's captured normal.
.. we store the curve's Z for future use, and then scale it to 0 in Z. We sample lengths from the flat version. We can then instance the pylons on those samples, and use the stored Z, one way or another, to raise them to the correct height.
In this case, I've put simple extruded-by-0 grids on the sample points, and then ray-cast up from them, 'shrink-wrapping' the tops of the extrusions to the underside of the track, so I haven't used the stored Z...
You might use the stored Z. Perhaps to locate the tops of your columns, as instances, or by raising a selection-by-Z of your column-template to an offset of the stored value.
