Collision Detection (Reposition Point) Code not working

[RyanJP]

Hi I wonder if someone can explain why the following code won't work? Basically just trying to take an emitted particle and reposition it if it overlaps with existing geo. The code runs fine but doesn't reposition the point at all, wondering why. Thanks!

v@ptemp = @P;
int npts[] = nearpoints(1, @ptemp, @pscale);
int lent = len(npts);
float seed = 1;

while(lent){
    @Cd = {1, 0, 0};
    @ptemp = set(random(seed), random(seed+1), random(seed+2));
    seed++;
}

@P = @ptemp;

 

[DonRomano]

I'm not sure that this is the right method to achieve what you want. First, you have an infinite loop (and I believe the vex compiler detects it and erases it from the code, that might be why it does nothing). Second, you never check the nearpoints again, and moreover, nearpoints can return the current point too, so you should be careful with that. Third, your @ptemp will always be in the range (0.0f, 1.0f) because you set it to be a random number, and not your initial position plus some random value on each axis.

You might want to write something like this :

vector pTemp = v@P;

int npts[] = nearpoints(0, pTemp, f@pscale);
int lenNpts = len(npts);

int seed = 0;

while(lenNpts > 0):
	vector randToAdd = fit01(vector(rand(seed)), vector(set(-1.0f)), vector(set(1.0f)));
	pTemp = pTemp + randToAdd;
	
	npts = nearpoints(0, pTemp, f@pscale);
	lenNpts = len(npts);

	seed++;

v@P = pTemp;

Hope this can help.

 

Cheers,

[RyanJP]

12 hours ago, DonRomano said:

 

vector pTemp = v@P;

int npts[] = nearpoints(0, pTemp, f@pscale);
int lenNpts = len(npts);

int seed = 0;

while(lenNpts > 0):
	vector randToAdd = fit01(vector(rand(seed)), vector(set(-1.0f)), vector(set(1.0f)));
	pTemp = pTemp + randToAdd;
	
	npts = nearpoints(0, pTemp, f@pscale);
	lenNpts = len(npts);

	seed++;

v@P = pTemp;

 

Hi thanks for your reply! Yeah my random vector was just for testing purposes. I am curious about your "randToAdd" variable though- I understand the intention but what is the vector() function? I can't find it in the documentation. And why run a set() on the y and z?

Overall though this really helps me understand the code better, thank you again. Curious as to how you would approach it? I'm trying to learn more Vex functions and also more of "how to think like a programmer" so any advice is very welcome

[DonRomano]

The vector() is just a cast, i.e in this case helping the compiler get the right function overload : because rand has multiple overloads as it can return either a float, a vector2..., so it helps the compiler finding which overload I need, here the vector one, and same for the set() function (as it has overloads for vector2, arrays, matrices...).

I remap the random vector because the rand() function always return random numbers in the range 0.0,1.0 and you want to have a more random position by having either a positive or negative jitter to add, it's better because if you only add positive random numbers to your position, it will end up going in the same direction.

I've made a one liner for the randToAdd but it can be decomposed in that way to help understand what it does :

while(lenNpts > 0):
	vector random = vector(rand(seed)); 		// create a random vector in the range (0.0f, 1.0f); ex {0.34f, 0.85f, 0.14f}
	vector min = set(-1.0f); 					// set a minimum vector; similar to {-1.0f, -1.0f, -1.0f}
	vector max = set(1.0f);	 					// set a maximum vector; similar to {1.0f, 1.0f, 1.0f}
	vector randToAdd = fit01(random, min, max); // remap our random vector to the min and max range
	pTemp = pTemp + randToAdd;					// add our new remaped random vector to the position

The best I could recommand is learning algorithms and data structures, I have a few books that helped me on this side, but you might find this channel useful : https://www.youtube.com/c/JunichiroHorikawa

The language is just a tool (here vex, the same can be done with python, just slower), the most important is the algorithms you use and design, and the more you know the more tools you'll have to solve problems.

Edited June 3, 2022 by DonRomano

[RyanJP]

Ahhhh vector1 got it got it, never would think to use that. I actually am considering doing some computer science classes in the fall to learn algorithmic/systemic principles. Could I ask the names of those books? Maybe they can save me some $$$ on school. I do watch Junichiro also but I feel like I need more info into what he talks about, also the videos are very long.

Thanks for your time and I really appreciate the breakdown, that does clarify it for me, and makes perfect sense.

[DonRomano]

There you go for maths and algorithms : 

- https://www.amazon.com/Introduction-Algorithms-Leiserson-Charles-Clifford/dp/B0839JW93F

- https://www.amazon.com/Mathematics-Computer-Graphics-Undergraduate-Science/dp/1447173341

You can have a look at gpu gems from nvidia (edition 1 is there : https://www.nvidia.com/en-us/drivers/gpu-gems-home/), there are out of date but there are some very nice algorithms and techniques you can try to implement in Houdini to practice.

 

Cheers and good luck in your learning path !