C#•2y ago

❔ ✅ Dice Roll with Generic Math

Has someone use non-generic funcs with generic math... The solution that I have found is a bit hack. There would be solutions that dont involves a bunch of conditionals?

public readonly struct Dice
{
    public readonly int sides;

    public Dice(int sides) => this.sides = sides;

    public override string ToString() => sides.ToString();

    public double GetAverage() => (sides + 1) / 2.0;

    public int RollInt(Random rng) => rng.Next(1, sides + 1);
    public long RollLong(Random rng) => rng.NextInt64() * sides + 1;
    public float RollFloat(Random rng) => rng.NextSingle() * sides + 1;
    public double RollDouble(Random rng) => rng.NextDouble() * sides + 1;

    public T Roll<T>(Random rng) where T : INumber<T> 
    {
        var type = typeof(T);
        if (type == typeof(int))
            return (T)(object)RollInt(rng);
        if (type == typeof(long))
            return (T)(object)RollLong(rng);
        if (type == typeof(float))
            return (T)(object)RollFloat(rng);
        if (type == typeof(double))
            return (T)(object)RollDouble(rng);
        return T.One;
    }
}

public readonly struct Dice
{
    public readonly int sides;

    public Dice(int sides) => this.sides = sides;

    public override string ToString() => sides.ToString();

    public double GetAverage() => (sides + 1) / 2.0;

    public int RollInt(Random rng) => rng.Next(1, sides + 1);
    public long RollLong(Random rng) => rng.NextInt64() * sides + 1;
    public float RollFloat(Random rng) => rng.NextSingle() * sides + 1;
    public double RollDouble(Random rng) => rng.NextDouble() * sides + 1;

    public T Roll<T>(Random rng) where T : INumber<T> 
    {
        var type = typeof(T);
        if (type == typeof(int))
            return (T)(object)RollInt(rng);
        if (type == typeof(long))
            return (T)(object)RollLong(rng);
        if (type == typeof(float))
            return (T)(object)RollFloat(rng);
        if (type == typeof(double))
            return (T)(object)RollDouble(rng);
        return T.One;
    }
}

30 Replies

Angius•2y ago

Unless Random implements a generic method that uses INumber, not really I guess maybe a switch expression would shorten that code But that's about it

Aaron•2y ago

why not just

public T Roll<T>(Random rng) where T : INumber<T>
{
    return T.CreateChecked(rng.Next(1, sides + 1));
}

public T Roll<T>(Random rng) where T : INumber<T>
{
    return T.CreateChecked(rng.Next(1, sides + 1));
}

MODiX•2y ago

Windows10CE#8553

REPL Result: Success

int sides = 6;
public T Roll<T>(Random rng) where T : INumber<T>
{
    return T.CreateChecked(rng.Next(1, sides + 1));
}
Roll<double>(Random.Shared)

int sides = 6;
public T Roll<T>(Random rng) where T : INumber<T>
{
    return T.CreateChecked(rng.Next(1, sides + 1));
}
Roll<double>(Random.Shared)

Result: double

Compile: 560.954ms | Execution: 40.180ms | React with ❌ to remove this embed.

SpectraOP•2y ago

Well it's not exactly the same thing, but it's valid solution, if you don't care about lost in conversion. Well you can also use the double one so there would be no data loss, but less efficient. Well it's some kind of solution, good to know thx.

Aaron•2y ago

oh yeah, I didn't realize you had the the in-betweens as valid states yeah just swap Next for what you do in RollDouble

SpectraOP•2y ago

probaly is more efficient doing double rather than a bunch of ifs

Aaron•2y ago

CreateChecked is the same as checked((T) input)

DaVinki•2y ago

If you would rather not lose anything in a conversion and support integers above 64-bit in size, you can: Stackalloc a byte span Use the random object to fill it with NexyBytes Use Unsafe.As

Aaron•2y ago

that isnt valid

DaVinki•2y ago

What do you mean? If the type is unmanaged, it is So it would just add support for 128-bit but with a little more tampering you could absolutely do it for arbitrarily large types

Aaron•2y ago

it would break floats

MODiX•2y ago

Windows10CE#8553

REPL Result: Failure

uint u = uint.MaxValue;
Unsafe.As<uint, float>(ref u)

uint u = uint.MaxValue;
Unsafe.As<uint, float>(ref u)

Exception: ArgumentException

- An exception occurred when serializing the response: ArgumentException: .NET number values such as positive and negative infinity cannot be written as valid JSON. To make it work when using 'JsonSerializer', consider specifying 'JsonNumberHandling.AllowNamedFloatingPointLiterals' (see https://docs.microsoft.com/dotnet/api/system.text.json.serialization.jsonnumberhandling).

- An exception occurred when serializing the response: ArgumentException: .NET number values such as positive and negative infinity cannot be written as valid JSON. To make it work when using 'JsonSerializer', consider specifying 'JsonNumberHandling.AllowNamedFloatingPointLiterals' (see https://docs.microsoft.com/dotnet/api/system.text.json.serialization.jsonnumberhandling).

Compile: 475.020ms | Execution: 31.323ms | React with ❌ to remove this embed.

DaVinki•2y ago

Why did you test that with uint

Aaron•2y ago

easiest way to get four bytes of 0xFF its the same as doing

MODiX•2y ago

Windows10CE#8553

REPL Result: Failure

unsafe
{
Span<byte> bytes = stackalloc byte[4];
bytes.Fill(0xFF);
Console.WriteLine(Unsafe.As(ref bytes[0]));
}

unsafe
{
Span<byte> bytes = stackalloc byte[4];
bytes.Fill(0xFF);
Console.WriteLine(Unsafe.As(ref bytes[0]));
}

Exception: CompilationErrorException

- Parameters or locals of type 'Span<byte>' cannot be declared in async methods or async lambda expressions.
- The type arguments for method 'Unsafe.As<T>(object?)' cannot be inferred from the usage. Try specifying the type arguments explicitly.

- Parameters or locals of type 'Span<byte>' cannot be declared in async methods or async lambda expressions.
- The type arguments for method 'Unsafe.As<T>(object?)' cannot be inferred from the usage. Try specifying the type arguments explicitly.

Compile: 603.036ms | Execution: 0.000ms | React with ❌ to remove this embed.

Aaron•2y ago

but easier also oof

MODiX•2y ago

Windows10CE#8553

REPL Result: Success

static unsafe void Main()
{
Span<byte> bytes = stackalloc byte[4];
bytes.Fill(0xFF);
Console.WriteLine(Unsafe.As<byte, float>(ref bytes[0]));
}
Main();

static unsafe void Main()
{
Span<byte> bytes = stackalloc byte[4];
bytes.Fill(0xFF);
Console.WriteLine(Unsafe.As<byte, float>(ref bytes[0]));
}
Main();

Console Output

NaN

NaN

Compile: 607.462ms | Execution: 37.521ms | React with ❌ to remove this embed.

Aaron•2y ago

i really shouldnt write unsafe code on discord but yeah, converting arbitrary bytes to a float is a great way to get random NaNs and Infs (but yes, if floats weren't involved, and you didn't care about bounding the output, this works fine)

Anton•2y ago

those ifs don't cost you anything, they resolve as constants at jit time and the dead branches get eliminated. the function then will likely get inlined. I'm only not sure about (object) conversions, that's a thing I have wondered previously, but the boxing should also get eliminated at jit time but check out the perf yourself the above is what I've heard from other people

Aaron•2y ago

all that is correct as well, if you're fine with what you have now i find return T.CreateChecked(rng.NextDouble() * sides + 1); to be much easier to read though though if you want the range to be [1, sides] instead of [1, sides + 1) you have to do

return T.CreateChecked(double.Min(rng.NextDouble() * sides + 1, sides));

return T.CreateChecked(double.Min(rng.NextDouble() * sides + 1, sides));

Anton•2y ago

I think you've mentioned that, but double has less precision than e.g. long for integers, so it's not the same implicit conversion of the compiler from long to double or from int to float is a lie made for convenience

Aaron•2y ago

as long as sides isn't very large it doesn't matter much actually i think this weights to die on sides, kekw

actually

return T.CreateChecked(rng.NextDouble() * (sides - 1) + 1);

return T.CreateChecked(rng.NextDouble() * (sides - 1) + 1);

hm no that doesnt work either because for integers it will be very unlikely to get sides honestly

if (default(T) is IFloatingPoint<T>)
{
    return T.CreateChecked(rng.NextDouble() * (sides - 1) + 1);
}
else
{
    return T.CreateChecked(rng.NextInt64(1, sides + 1));
}

if (default(T) is IFloatingPoint<T>)
{
    return T.CreateChecked(rng.NextDouble() * (sides - 1) + 1);
}
else
{
    return T.CreateChecked(rng.NextInt64(1, sides + 1));
}

i think that works the best? for both integers and floats that will get you the range [1, sides] with equal distribution (assuming System.Random has equal distribution, of course)

Anton•2y ago

your double can output 0

Aaron•2y ago

no it can't +1 on the end

Anton•2y ago

ah yeah you're right

Aaron•2y ago

aaaaaaa that is check doesnt compile i'm back with a vengence

if (T.One / (T.One + T.One) == T.Zero)
{
    return T.CreateChecked(rng.NextInt64(1, sides + 1));
}
else
{
    return T.CreateChecked(rng.NextDouble() * (sides - 1) + 1);
}

if (T.One / (T.One + T.One) == T.Zero)
{
    return T.CreateChecked(rng.NextInt64(1, sides + 1));
}
else
{
    return T.CreateChecked(rng.NextDouble() * (sides - 1) + 1);
}

Accord•2y ago

Was this issue resolved? If so, run /close - otherwise I will mark this as stale and this post will be archived until there is new activity.

SpectraOP•2y ago

Have done some benchmarks, there almost no difference it completely depends on the types that is been casted. And there is a problem with (T)(object) which the cast can throw exception while the others don't.

SpectraOP•2y ago

I think this can be closed so...

Accord•2y ago

Was this issue resolved? If so, run /close - otherwise I will mark this as stale and this post will be archived until there is new activity.

Gaming

Programming

❔ ✅ Dice Roll with Generic Math