How can you set seed type long?

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

Awaiting Schedule

These updates are awaiting their schedule. Click on a checkbox to get an update now.

• chore(deps): Update patch dependency changes to v5.10.3 (org.junit.jupiter:junit-jupiter-engine, org.junit.jupiter:junit-jupiter-api)
• chore(deps): Update dependency net.ltgt.errorprone to v4

Open

These updates have all been created already. Click a checkbox below to force a retry/rebase of any.

• chore(deps): Update dependency gradle to v8.8
• chore(deps): Update error-prone monorepo to v2.28.0 (com.google.errorprone:error_prone_annotations, com.google.errorprone:error_prone_core)

Detected dependencies

• Check this box to trigger a request for Renovate to run again on this repository

With PR #5, flow noise will look less and less similar its parent, jlibnoise. The only similarity would be several commits early in history, meaning that the "forked from" label is not really necessary. The last commit to jlibnoise was in August 2013, meaning that flow noise has become its spritual successor, and its own project.

Removing the "fork" relationship to jlibnoise allows a few other perks, like searching code.

This can be done, as explained by Github:

To detach the fork and turn it into a standalone repository on GitHub, contact GitHub support.

Hey. Thanks for working on the library still and publishing new changes. That's greatly appreciated.

But wtf were you thinking pushing breaking changes to the repository under the same version number? It just broke our whole world generation project. We can't even revert to an older version because of this.

Please use the version number what it's intended for and bump it on uploads. Thanks!

In modifiers/Curve.getValue(), there's a divide by zero error:

    // Compute the alpha value used for cubic interpolation.
    double input0 = controlPoints.get(indexPos).inputValue;
    double input1 = controlPoints.get(indexPos).inputValue;
    double alpha = (sourceModuleValue - input0) / (input1 - input0);

Because input0 = input1 since they're getting the same indexPos, input1 - input0 = 0, thus the divide by zero error.

It needs to be replaced with the following:

    // Compute the alpha value used for cubic interpolation.
    double input0 = controlPoints.get(index1).inputValue;
    double input1 = controlPoints.get(index2).inputValue;
    double alpha = (sourceModuleValue - input0) / (input1 - input0);

Your website appears to be down and your quickstart/usage guide doesnt work..

Why flow noise has value range 0-1? Why not leave it as -1 to 1?

This value range is very counterintuitive when using it for terrain generation, and annoying to work with, where the average value specifies average terrain height (especially when used as 3d noise):

• Scaling the noise generator output changes average value
• Changing almost any parameters when using multiple octaves (except frequency) changes average value

With value range 0-1, there is a very simple way to make 3d noise based terrain generator: multiply the noise output by desired height variation, add average height, and use the output as "heightmap" that also depends on height.

The only situation I see where values 0 to 1 are simpler is when outputting to image, which is done only sometimes for testing. And when using noise output for interpolation, where it can be easily changed into 0-1 range.

Well, technically it is. But technically it's also between -infinity and +infinity. After testing with 64 million samples (not random, some predefined area) I got min=0.30759661550127604, max=0.692403384498724. It seems safe to say that the actual value range that happens in practice is 0.3 to 0.7. When eriting terrain generator using it with the assumption that the value range will be 0 to 1, the terrain will never reach values even halfway through to the intended max height and depth.

I also did an experiment replacing the RANDOM_VECTORS array (I currently use that "hack" in my code) with the values Minecarft uses internally - which is all possibilities with 2 coordinates being +/-1 and one coordinate being zero - prescaled down by 0.5 to make halve the value range so that after adding 0.5 - the resulting range is 0-1. After testing with the same 64 million samples, I got the exact value range of 0 to 1. Not 0.00000001 to 0.999999999, exactly 0.0 to 1.0. There are actually 12 such vectors, so I choose them randomly to fill the whole array.

And some images (this is the frequency I used for sampling, just 8k by 8k image, pixel color intensity is clamp(floor(value * 256), 0, 1):
Current values:

"fixed" values:

Unfortunately, fixing this will break compatibility with existing code, so maybe another option would be updating documentation to say that?

Your Noise.gradientCoherentNoise3D is actually Perlin noise and your Perlin noise is actually fBM noise. libnoise made this mistake and since everyone else is either porting that library or copy/pasting from it this is a pretty widespread mix up.

https://en.wikipedia.org/wiki/Perlin_noise
https://en.wikipedia.org/wiki/Fractional_Brownian_motion
https://iquilezles.org/www/articles/fbm/fbm.htm

The link to the "Examples and code snippets" in the Getting Started section (https://github.com/flow/noise#getting-started) takes me to the GitHub 404 page.

Hello. I like to use noise for my game. I was under the impression that this a re-implementation of the noise lib in Java. I was trying to follow this tutorial: http://libnoise.sourceforge.net/tutorials/tutorial3.html

But you didn't implemented NoiseMapBuilderPlane and other classes from utils?

How can you then do tiling terrain height maps? Or is this not in the scope?

If you don't like to implement utils then I could implement it.

Unfortunately, Java can only expose mutable arrays at the moment, and we don't want to allow our internal state to be exposed like this.

Alternatives are:

• provide a DoubleStream
• add Terrace.controlPoint(index)
• maybe both?

[20:36:56] [Sponge-AsyncScheduler-0/INFO] [rustmesponge]: Initializing ProceduralGenerator with following settings: 
[20:36:56] [Sponge-AsyncScheduler-0/INFO] [rustmesponge]: Frequency: 0.00390625, OctaveCount: 5, Persistence: 1.0, Seed: 123
[20:36:56] [Sponge-AsyncScheduler-0/INFO] [rustmesponge]: Initializing height map task from (-100, 0, -100) to (100, 255, 100)
[20:36:56] [Sponge-AsyncScheduler-0/WARN] [rustmesponge]: Max Value: NaN
[20:36:56] [Sponge-AsyncScheduler-0/INFO] [rustmesponge]: Found height: 310.84143196993296. Noise: 2.428448687265101

This is how I initialize Perlin module:

        var perlinModule = new Perlin();
        perlinModule.setFrequency(frequency);
        perlinModule.setOctaveCount(octaveCount);
        perlinModule.setPersistence(persistence);
        perlinModule.setSeed(seed);

        this.MAX_VALUE = perlinModule.getMaxValue();
        this.proceduralModel = new Plane(perlinModule);

Later I get the MAX_VALUE and it turns out to be NaN.
Full code:
Initialization https://ptero.co/apezygamas.java
Runnable I use to retrieve values https://ptero.co/ujuvamadub.java

This has been found by DaPorkchop_ while working on another project

This method:

has been originally ported from the following C code in libnoise:

  inline double MakeInt32Range (double n)
  {
    if (n >= 1073741824.0) {
      return (2.0 * fmod (n, 1073741824.0)) - 1073741824.0;
    } else if (n <= -1073741824.0) {
      return (2.0 * fmod (n, 1073741824.0)) + 1073741824.0;
    } else {
      return n;
    }
  }

But because of operator precedence in java, the java code is actually equivalent to this

    public static double makeInt32Range(double n) {
        if (n >= 1073741824.0) {
            return ((2.0 * n) % 1073741824.0) - 1073741824.0;
        } else if (n <= -1073741824.0) {
            return ((2.0 * n) % 1073741824.0) + 1073741824.0;
        } else {
            return n;
        }
    }

Where in the original C code, modulo should take precedence.

For values of n between k * 1073741824 and k * 1073741824 + 536870912 for integer k >= 1 (and similar for negative range) this just happens to output the same values, but otherwise for n > 1073741824 the result will be always negative (and similarly for n < -1073741824 the result will be always positive) instead of repeating the entire range from -1073741824 to 1073741824.

Changing this now may be considered a breaking change as there could be terrain generators that would generate different terrain with this fixed.

Can't find any version of the noise library in Maven:

https://mvnrepository.com/artifact/org.spongepowered

Here is snippet of noise value examples from 10x10x10 cuboid:

[16:05:17] [Server thread/INFO] [ens3]: Value for [8, -4, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, -3, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, -2, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, -1, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 0, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 1, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 2, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 3, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 4, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 5, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 6, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 7, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 8, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, 9, 0] is 1.6301681263765515
[16:05:17] [Server thread/INFO] [ens3]: Value for [8, -10, 1] is 1.6301681263765515

Code I use for initializing Perlin noise module:

public class ProceduralGenerator {
    private final Module heightGenerator;

    public ProceduralGenerator() {
        var perlinModule = new Perlin();
        perlinModule.setPersistence(0.714);
        perlinModule.setOctaveCount(8);
        heightGenerator = perlinModule;

    }

    public double applyCoordinates(Vector3d position) {
        return heightGenerator.getValue(position.x(), position.y(), position.z());
    }
}

As pointed out by Diggy in SpongePowered Discord (https://discord.com/channels/142425412096491520/142425521391665153/893802421703278622) issue also seems to affect billow and ridged multi.