JavaScript伪随机序列生成器 [英] JavaScript pseudo-random sequence generator

问题描述

我需要生成一个确定性（即可重复）的伪随机数字序列，并给出初始种子和选择该序列中的第n个项目。

如果JavaScript的随机函数是可植入的，我可以这样做：

 函数randomNth（seed，seq）
 {
 var r; 
 Math.randomSeed（seed）; 
 for（var i = 0; i ++< seq; i ++）
 {
 r = Math.random（）; 
} 
 return r; 
} 
  

然而，这不是也是另一种可选择的PRNG看起来有点慢;要求第250个号码会很贵。

我认为哈希是我想要的，也许类似于 md5（seed + seq）％最大但JavaScript没有md5（），如果我在代码中做的话，可能会有更好的选择。

我'd like a function where

x = randomNth（seed，seq，maxVal）// x是int&& x> = 0&& x < maxVal

或者理想情况下

x = randomNth （种子，seq）// x> = 0&& x < 1，与Math.random（）一样

其他要求：

• 必须在node.js和浏览器中运行


• 数字应该是统计上随机的（或者足够接近，因为周期很小）
  
• 应该是O（1）并且性能合理

解决方案

使用了（非SO）朋友的建议。我去了CRC32（），因为这是非常快的，并给出了体面随机值。

return crc32（seq + seed）％maxVal; code>

800万的运行产生了maxVal = 8的以下分配：

0 999998

1 999998

2
1000007

3
1000003

<4>
1000001

5
1000003

6
999992

7
999998

我也跑了，其结果如下： CRC32（）适用于随机数字Diehard结果。简短的版本是它失败了（对于如此少量的测试数据），但它仍然足以满足我在小范围内产生数字的需求。

I need to generate a deterministic (i.e. repeatable) sequence of pseudo-random numbers given an initial seed and select the nth item from that sequence.

If JavaScript's random function was seedable, I could just do:

function randomNth(seed, seq)
{
    var r;
    Math.randomSeed(seed);
    for (var i = 0; i++ < seq; i++)
    {
        r = Math.random();
    }
    return r;
}

However, it's not, and alternative, seedable PRNGs look to be a little slow; asking for the 250th number would be expensive.

I think a hash is what I want here, perhaps something like md5(seed + seq) % max but JavaScript doesn't have md5() and if I'm doing it in code there's probably a better choice of hash.

I'd like a function where

x = randomNth(seed, seq, maxVal) // x is int && x >= 0 && x < maxVal

or, ideally

x = randomNth(seed, seq) // x >= 0 && x < 1, same as Math.random()

Other requirements:

• must run in node.js and in a browser
• numbers should be statistically random (or close enough as the period will be small)
• should be O(1) and reasonably performant

解决方案

In the end I used a suggestion from a (non-SO) friend. I went with CRC32() as this is extremely fast and gives decently random values.

return crc32(seq + seed) % maxVal;

A run of eight million produced the following distribution for maxVal = 8:

0 999998

1 999998

2 1000007

3 1000003

4 1000001

5 1000003

6 999992

7 999998

I also ran "Marsaglia's famous "Die Hard" battery of tests" mentioned in the Donald Knuth page Hans mentioned, the results of which are here: CRC32() for random numbers Diehard results. The short version is that it fails miserably (for such a small amount of test data), but it's still good enough for my needs where it is generating numbers in a small range.

这篇关于JavaScript伪随机序列生成器的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！