next »

[gimymblert]

Come to the dark side and try a linear congruential generator! Who needs any semblance of quality when you've got raw speeed!

Code:

static unsigned int seed;

void seed(int newseed)
{
	seed = newseed;
}

int roll(void)
{
	seed = (seed * 214013 + 2531011) % 2147483648;
	return seed >> 16;
}

Still a pattern with successive seed sorry! (it goes 3 by 3)

Randomness is surprisingly serious business

I need randomness across successive SEED ...

[surt]

You can combine multiple different LCGs to increase apparent randomness quality.

[gimymblert]

That won't solve the original problem though, it's that seed must be isolate from each other so that there is no relation between consecutive seed. Combining LCG don't help with that. IN general, pcg proposed are sequenced randomness they don't allow random "access". You can't compute the nth random number without computing the nth-1.

[gimymblert]

After looking at a lot of stuff I decided to go with this (if anyone is interested) for now:

(nested loop of i and j, basically x and y)
width = (i*10000) + (j*100)
Seed (1 + (i*width) + (j*width*width) )

Not sure if it's optimal but it work for the prototype I'm doing

[gimymblert]

I'm trying to make a small code that hash the space, generate 1 to 4 points inside, and link point inside a cell to the nearest point in the 8 neighbors ... BUT I have a small bug ... it doesn't seem to take obvious close point in neighbor's cell ...

Code:

;loop all points to compute distance
				; loop relative positions of neighbor cell
				For g = -1 To 1
					For h = -1 To 1
						; if central cell, skip (don't link point within that cell)
						If g = 0 And h = 0 Then Goto label

						; bound test
						If i+g < 0 Or j+h < 0 Then Goto label
						
						;setting the current neighbor
						looped.cell = unit.cell(i+g,j+h)

						;looping through all points of the current neighbor
						For l = 0 To looped\number
						
							loopt.point = cellpoint(looped,l)

							;computing the square magnitude to compare
							lsx = Abs (loopt\x - pos\x)
							lsy = Abs (loopt\y - pos\y)
							distanceSqr = lsx^2 + lsy^2

							;check for the lowest magnitude for this point
							If distanceSqr < distanceMin Then
								distanceMin = distanceSqr
								pointMin = loopt
								
								pointMin\i = g
								pointMin\j = h
							EndIf 
							
						Next
						
						.label
					Next
				Next

It's blitzcode:
x.type is how you declare custom type (basically struct)
x\var is how you access member field of the struct
";" is to comment stuff

[gimymblert]

NEVER MIND


 



I WAS COMPUTING DISTANCE USING THE RELATIVE LOCAL CELL POSITION (as if they were inside the same cell) INSTEAD OF USING ABSOLUTE POSITION

 

WRONG!!!!!!!! 
lsx = Abs(loopt\x - pos\x)
lsy = Abs(loopt\y - pos\y)

CORRECT!!! 
lsx = Abs ( (i*cx + g*cx + loopt\x) - (i*cx + pos\x) )
lsy = Abs ( (j*cy + h*cy + loopt\y) - (j*cy + pos\y) )
                     
                     
                     distanceSqr = lsx^2 + lsy^2

[gimymblert]

Okay, now do someone know how to make a staircase function where each step is 2^x in size?

Generally to make a stair case linear function you do something like int i = x/step, but I don't know what step should be to have power of 2 size of step :/

[BorisTheBrave]

Code:

// Which step you are on.
int i = floor( log(x) / log(2) )
// Height of that step
float h = pow(2, i)

[gimymblert]

Oh great! thanks a lot!

[CheeseCakeMan]

I've been reading a lot about how to timestep games and the often mentioned Fix Your Timestep! on Gaffer on Games has worked well for me. However I'm currently working on something that has a lot of production buildings and spawners as well as potentially low tickrates for entities far in the distance and I don't quite understand how I would apply this to to things which can only come into existence as a whole.

Stepping physics twice if necessary is fine, we just move half the distance. But...

If I have a spawner with a cooldown of 2 seconds, but I step every 1.1 seconds (exaggerated) I would run the spawner logic after 2.2 seconds for the second time (the time the spawner would answer yes to "Can I spawn?"). To have it spawn every 2 seconds on average I would need to accumulate this time and spawn twice whenever it reaches the cooldown. In the long run it would spawn every 2 seconds, however in this example it would take several seconds for the accumulator to fill and trigger a second spawn. It seems rather odd that all of a sudden two entities appear at the same time, at least from a players perspective.

With amounts this seems doable. A machine could produce 6.4858 kg iron instead of 5 from time to time if the steps don't align. But I can't spawn half a goblin!

[ThemsAllTook]

Sounds like your spawner entity's tick rate would need to match (or be a multiple of) its cooldown interval. Some time travel might be necessary for things like sounds, which would be more noticeably aliased.

[gimymblert]

going from cube to sphere without the pinched effect around the pole ...


http://mathproofs.blogspot.nl/2005/07/mapping-cube-to-sphere.html
http://catlikecoding.com/unity/tutorials/cube-sphere/
http://math.stackexchange.com/questions/118760/can-someone-please-explain-the-cube-to-sphere-mapping-formula-to-me

I tried to implement this, Was very excited  as it's a proven method, it should have worked flawlessly 

... 

 

Code:

x# = VertexX (surface,i) ;*2/grid
	y# = VertexY (surface,i) ;*2/grid
	z# = VertexZ (surface,i) ;*2/grid
	
	
	x2# = x*x
	y2# = y*y
	z2# = z*z
	
	sx# = x * Sqr(1 - (y2 / 2) - (z2 / 2) + ((y2 * z2) / 3))
	sy# = y * Sqr(1 - (z2 / 2) - (x2 / 2) + ((x2 * z2) / 3))
	sz# = z * Sqr(1 - (x2 / 2) - (y2 / 2) + ((x2 * y2) / 3))
	
	sx# = sx*radius
	sy# = sy*radius
	sz# = sz*radius
	
	VertexCoords surface, i, sx,sy,sz

why it does not work?  it's the exact same things! (yes I know about the vector.one and grid, I tried)


Tell me I'm dumb and I forgot something obvious  I have been staring at this for hours 



It's fat but not round


Yes I tested the naive version with simply normalizing vertices, it works, BUT I need this smooth non pinched curved.

[Sik]

Looked around at the code I used to generate the globe look-up table in Miniplanets and I think this is the relevant piece of code:

Code:

         float sinx = sin(basex * 3.1415926f * 0.5f);
         float siny = sin(basey * 3.1415926f * 0.5f);
         float circx = sinx * sqrt(1.0f - siny * siny * 0.5f);
         float circy = siny * sqrt(1.0f - sinx * sinx * 0.5f);

basex and basey are the original coordinates, circx and circy are the coordinates turned into a circle (values in both cases range from -1 to 1). Shouldn't be hard to add the extra dimension I guess.

No guarantees that this is a perfect circle though (and no guarantees that it'll work as intended beyond 90º, but I guess you could just handle it per face instead?), it just happened to work well enough for me. And of course remember to scale the results to whatever range you actually need =P

EDIT: in hindsight this does half a sphere when you only want quarter... I guess you should work with coordinates on the -0.5 to +0.5 range then. Take that into account if something goes wrong.

[gimymblert]

Okay I solved it
For some reason, I must multiply by two THEN divide by two
Don't ask the logic, trial and error for the win ...

[Thorves]

In JavaScript: say I have array A and array B. Some of the contents of these arrays may overlap. (e.g. A[0] might be the same object as B[4])

What would be the quickest way to get an array containing all elements in A that are not in B and an array with all elements of B that are not in A?

[JohnsyJohnsy]

thorves:
like so: http://stackoverflow.com/questions/14930516/compare-two-javascript-arrays-and-remove-duplicates

adapted slightly so you dont overwrite your original arrays like this:

Code:

let arrayA, arrayB, arrayA2, arrayB2

arrayA2 = arrayA.filter(function(val) {
  return arrayB.indexOf(val) == -1;
});
arrayB2 = arrayB.filter(function(val) {
  return arrayA.indexOf(val) == -1;
});

or the es6 way

Code:

let arrayA, arrayB, arrayA2, arrayB2
arrayA2 = arrayA.filter(val => !arrayB.includes(val));
arrayB2 = arrayB.filter(val => !arrayA.includes(val));

[_glitch]

What is the best way to store strings for e.g. dialogs?

In a separate class as properties

Code:

 internal static string DiagString { get { return "Some String"; } } 

Or as var

Code:

string DiagString = "Some String";

Or directly in the code

Code:

SpriteBatch.DrawString(Font, "Some String", Position, Color);

[Sik]

Loaded from a file so it can be easily translated later without even having to touch the code, which means you'll want a different approach altogether (as you'll want a place to keep all of them after loading and the means to fetch them from other parts in the code).

[_glitch]

Thank you, I had something like this in mind before.

[_glitch]

And what would be the best way to achieve that the right string is read in the file? With using a stream reader? I'm using C#, by the way.