cringe

like omg so random

but like

with gdx

theres like soooooo many methods

its p. cool

plus if u use this

u dont need 2 understand how this works

    public float nextExclusiveFloat () {
        final long bits = nextLong();
        return NumberUtils.intBitsToFloat(126 - Compatibility.countTrailingZeros(bits) << 23 | (int)(bits >>> 41));
}

it can just be some cool magic code or smth

u use it when u want a random float but not 0 or 1 lol

theres other cool stuff 2

shuffle() and randomElement() these work with libgdx types

all the stuff from mathutils is here 2 but like instance methods on a gdxrandom

stringSerialize() and stringDeserialize() store random stuff in like a string and totes load it later

read() and write() do that but like with json and the json stuff does it like auto

u get the picture

o yeh

theres this other stuff now

u can make a GapShuffler out of a randommy thingy and like a lot of stuff u dont wanna repeat

that one will go on like foreeeeeeverrrrrr so u gotta only get a lil bit k

theres this like WeightedTable and it gets u ints but like each one has its own weight

the big chonky ones get picked more. like my cat lol

um theres like a new one called UniqueIdentifier and it just like makes unique things when u call next()

its just like UUID but it works on GWT i guess and its like Json.Serializable which is p. cool

o and theres Scramblers 2 take a number and make it like wiggly jiggly or smth

o wate moar!!!

now there like all these like um RawNoise types

each 1 has liek a diff flavor i guess?

PerlinNoise is all squarey but also liek blobby

SimplexNoise is liek triagnles n stuffs

ValueNoise is real dum squarezzz so u better use it wit somethin else

HoneyNoise mixiz it up wit SimplexNoise and ValueNoise all hot n hevay

FoamNoise is super orgnanic an natrual lookin

CyclicNoise uses som liek rotaiton stuf an looks real diff

SorbetNoise tastes niiiice but its also liek CyclicNoise and bettr in 2D i think?

PerlueNoise is liek Perlin but aslo makin Value at teh saem tiem liek, an then woop woop mixes em up!

OpenSimplex2FastNoise is tht fancy OpenSimplex2 thgin in fassssst moddddde

OpenSimplex2SmoothNoise is tht fancy OpenSimplex2 thgin in smooooth moddddde

an those 2 use liek RotationTools but u prolly dont have 2 lololol

then u can give 1 of thoes 2 ContinuousNoise n get liek rideged mode n octaevs n stuf

these r all liek Json.Serializable so u can liek saev and laod stuf

LineWobble is kinda liek those but jus 4 1D noise

even moar!!!!!!11!!!

Compatibility maeks ur bit stuff work on the googles web tool kit thingy

ColorSupport supports ur colors an stuff liek w hue an satruation an all that

PoissonDiskDiscrete maeks poitns placed far aways from each ohter, but liek on a liek grid

PoissonDiskSmooth maeks poitns placed far aways from each ohter, but liek not on um a grid

PointHasher is liek Scrambler but 4 a bunch of numburs insted of liek 1

EncryptedFileHandle is liek lamebad encarption 4 ur filez but it does liek hide ur images and scriptz

teh encarption is not 4 web tho :(

RoughMath is omg so itchy n rough but has superfast math if u can handle a lil mess >:)

but i think this new stuf is p. cool

an then theres liek anothr i guess

PointSequence has liek diff classes insied 4 maekin a buncha poitns but thye wont be nreaby in liek forever

PointSequence.Halton2 is a 2D one an u can cnofgiure its bases or smth

PointSequence.R2 what evne is an R liek omg. but it also maeks a buncha 2D poitns an thye wont be cloes

theres like a lot more Ds 4 each of thees, 3D an 4D an 5D an 6D wowowow

Vector5 an Vector6 r xtendorizatioins on libgdx Vector2, Vector3, an Vector4, 4 biggar D lololol

cereal

u can liek cerealize an decerealize stuffs in here

with like

Json from libgdx, out of teh box!!!

or Kryo 4 bianry storaeg usign kryo-more

or Fury 4 moar bianry storaeg from teh apache foudnatoin, but also liek, out of teh box!!!

or sometimes stuff has codes that liek wriets 2 a String or reads frm 1

get

The easiest way to use this library was to copy files from the source code. ... But, that was only in the earliest versions. In 0.1.0 and newer, you should probably use Gradle (or Maven or whatever your chosen build tool is) to depend on cringe.

The three GdxRandom implementations: (RandomAce320, RandomDistinct64, and RandomXMX256) can be used as drop-in replacements for java.util.Random or an almost-identical replacement for libGDX's RandomXS128.

The different GdxRandom implementations differ in various properties. All are high-quality statistically, passing the PractRand battery of tests to 64TB of tested data. They each have different "periods," meaning the number of random results that can be produced before the sequence repeats. In practice, a game will probably never have to worry about exhausting an entire period and actually seeing the sequence repeat. However, far enough along through the actual period of a generator, it becomes possible for an adversary (typically a player) to notice and exploit patterns. The shortest I believe this could happen is in RandomDistinct64, after over 4 billion random numbers are generated (2 to the 32). For RandomXMX256, the earliest point should be 2 to the 128, which just means it isn't going to happen. AceRandom320 has many different possible cycles, with the shortest possible one at least as long as RandomDistinct64's period.

RandomAce320 is very fast. It uses very few types of instruction, and only performs one math instruction on each state to get a random result. It also can perform those 5 math instructions using instruction-level parallelism, which the JVM has been good at exploiting to improve performance since Java 16 or so. The minimum cycle length is 2 to the 64, and the maximum is over 2 to the 300 (but it's impractical to say exactly). Most random seeds will be in one of the largest cycles, and even in the minuscule chance that you randomly start in the shortest cycle, that's still almost always going to be good enough for any app, and effectively always enough for any game. It has five longs of state, each of which can have any value. This generator relies entirely on its state transition to appear random.

RandomDistinct64 produces each long from nextLong() exactly once, which makes it good for producing random UUIDs. It's also rather fast on most JVMs, and might be much faster on GraalVM (version 20 and later). I need to verify some unusual benchmarks using GraalVM 20, but they appear to show RandomDistinct64 producing over 4 billion long values per second, which is over twice as fast as RandomAce320. This could be the benchmark code eliminating the tested loop entirely, so take it with a large grain of salt. It has one long of state, which can have any value. This generator relies entirely on its output function to appear random, and doesn't have a random-like state transition at all.

RandomXMX256 is a fortified version of Xoshiro256**, which is present in Java 17, but removes its ** or StarStar "scrambler" and replaces it with the much more robust MX3 unary hash, which is also used in our Scramblers class. This generator produces each sequence of four numbers exactly once, except for 0, 0, 0, 0, which never appears. RandomXMX256 is slower than the other two generators, but should be much more resilient against statistical quality issues, probably to the point of being overkill. That's still only overkill for the most common usages, with either one generator producing many random numbers, or many generators with random initial states -- if the initial states are extremely similar, patterns may become observable over time. It has 4 longs of state, each of which can have any value unless all states are 0 (which is not permitted). This generator has a mildly random state transition and an extremely thorough output function (also called a scrambler).

There are also a few files that use random number generators in some way:

• GapShuffler is an infinite Iterable that will never produce the same item twice in a row, as long as more than one item was given to it and all items are unique. It has a different way of shuffling its input to do this.
• Scramblers is purely static and provides methods to "scramble" long input into different formats.
• WeightedTable takes an array of weights and produces indices with its random() method with frequency adjusted by weight.

You don't need any of these 3 files to use the library, but they may come in handy.

UniqueIdentifier is also here, and doesn't use any GdxRandom. Instead, it produces a sequence of UUID-like values that are guaranteed to be unique unless an impossibly large amount of identifiers have been produced (2 to the 128 minus 1). Unlike the JDK class UUID, UniqueIdentifier is compatible with GWT. Random UniqueIdentifier values are less likely to collide than random UUID values, though in both cases the chance of a collision is mind-bogglingly rare.

In more recent versions, cringe has gotten larger, and now has code for continuous noise, which is fittingly random-like, and EncryptedFileHandles, which are akin to randomized, but reversible, scrambling of files. The noise classes are almost always used with a ContinuousNoise object wrapping some RawNoise implementation. The ContinuousNoise provides the most useful features, but the RawNoise provides the underlying algorithm. The noise classes are more inter-dependent; most RawNoise types use LineWobble to provide 1D noise, and classes like FoamNoise and HoneyNoise call ValueNoise methods directly. Meanwhile, SorbetNoise inherits from CyclicNoise, and there's lots of other connections. More noise classes have been added more recently. To use the noise, you should probably have a dependency on all of cringe.

This also adds Vector5 and Vector6 to the Vector2 through Vector4 libGDX provides, and the PointSequence class' implementations can generate the appropriate type of Vector for a given dimension. PointSequence produces either Halton sequences or R2-like sequences, and can be 2D through 6D. The Distributor class handles conversion from double, int, or long values to normal-distributed ones. RoughMath provides approximations to various math functions. There was briefly a Ziggurat class, but it has been merged entirely into Distributor.

You can depend on the library as a whole, using it as a normal Gradle or Maven dependency.

api "com.github.tommyettinger:cringe:0.2.0"

If you use GWT, then your GWT module needs to depend on:

implementation "com.github.tommyettinger:cringe:0.2.0:sources"

GWT also needs this inherits line added to your GdxDefinition.gwt.xml file, with the other inherits lines:

<inherits name="com.github.tommyettinger.cringe" />

You don't need the inherits line if you just copy a few classes into your sources.

license

apache 2.0

thanks 2 paul mineiro 4 makin these liek blazign fast apropximations w liek bit haxx

those apropximations are new bsd license which should be cool i think???

also thanks 2 rafa skoberg for like givin me a ton of work 2 do lololol jk

n thanks 2 raeleus since ur liek the king of cringe

yeah

now put on ur robe and wizard hat

its time for number magic