Better random function for C#

Question

I'm generating many random numbers and I need a good function, since this doesn't help much:

public static class Randomizer
{
    static Random random = new Random((int)DateTime.Now.Ticks);

    public static int RandomInteger(int minimum, int maximum)
    {
        return random.Next(minimum, maximum + 1);
    }

    public static double RandomDouble()
    {
        return random.NextDouble();
    }
}

When I use this class, my numbers are very often the same. Do you have any simple idea how I can improve performance of the randomizer?

Thanks, Ivan

Solution

See System.Security.Cryptography.RandomNumberGenerator

Comments

[Arpit Arya]

Since the random field in the class is static, it's only instantiated once.

[abatishchev]

He don't create new Random() every call -- or he edited the question

[BrokenGlass]

No he's not - static Random random is only initialized once - where is he re-creating it?

[dampee]

True, or initiate "Random" in the methods like this: int randomInt = (new Random()).Next(min, max);

[user541686]

Click the link again. :) (Ever noticed the "Other Versions" dropdown?)

[BrokenGlass]

@dampee it should be only initialized once, it's a common mistake to do exactly what you just suggested, in that case you will get duplicate numbers.

[Ivan]

Thanks a lot. Can you just tell me how to generate then a random number? I don't really manage to get what I want..

[user541686]

Use RandomNumberGenerator.Create to instantiate it, and then call GetBytes on the instance each time to fill an array of bytes. Then try using BitConverter's methods to convert it to an integer.

[dampee]

@BrokenGlass. I posted my previous comment at the same time as yours. You are completely right! Looked it up on MSDN. (msdn.microsoft.com/en-us/library/system.random.aspx) Quote: Because the clock has finite resolution, using the parameterless constructor to create different Random objects in close succession creates random number generators that produce identical sequences of random numbers.

[Henk Holterman]

This will make it a lot slower, and more secure. But will it give better 'randomness'?

[David Waters]

The only noticeable difference from the OP is locking for thread safety assuming OP is in a single threaded environment (big assumption) there will not be any difference.

[user541686]

It's "secure" precisely because of better randomness.

[Henk Holterman]

Not when you use it wrong. I just meant the OP should have answered some of the comments instead of accepting this.

[user541686]

@Henk: I agree -- I think the problem lied somewhere else. I answered his question, but I'm not sure if it's the solution that actually solved his problem.