srand not implemented, fails silently

[wmnnd]

Random::srand seems to be not implemented, instead it fails silently.
I find this problematic because it creates inconsistent results with other Ruby implementations. I have tried this code sample in both MRI and JRuby and received the same results.

On a side note, the methods of Random can only be accessed as Object::method in Opal and not as Random::method.

srand 123
a = [rand(100), rand(100), rand(100)]

srand 123
b = [rand(100), rand(100), rand(100)]

puts "#{a} == #{b}: #{a==b}"
#Expected: [66, 92, 98] == [66, 92, 98]: true

P.S.:
I know that JavaScript's own pseudorandom number generator doesn't support seeding. Would there be any other way to fix this?

[mieko]

Considering it'd require a PRNG implementation anyway to allow for srand, it'd be awesome if it matched MRI's output given the same seed value. Kernel#rand and #srand are implemented in terms of an instance of Random (Random::DEFAULT) in MRI, which we don't have an implementation of in Opal.

Dealing with the performance tradeoff, there could exist a Random::ECMA that ignores the seed value and acts as the current Kernel#rand does.

Edit: Here's a Javascript port of the same Mersenne Twister implementation MRI is based on, mt19937ar: https://gist.github.com/banksean/300494

[wmnnd]

@mieko I was just gonna post the link to that Gist too, because I hadn't seen your edit yet ;-)
The genrand_res53() of this implementation is actually compatible with MRI's implementation of rand and returns the same pseudorandom values if the same seed is used (except for using rand with a parameter which is not supported here).

[mieko]

Ah, cool to know they're compatible, @wmnnd.

Here's a sketch of Random, if anyone wants to play with it. It has not been tested or even ran (I don't have the connectivity to get a full Opal stack set up this weekend). Otherwise I'll get back to it later.

[daveheitzman]

I'm interested in implementing this.

[wmnnd]

I have not tried the implementation by @mieko and just used a simple wrapper for the Mersenne twister JavaScript implementation. But @mieko's code seems a good place to start.

[daveheitzman]

I did use mieko's code in my implementation

[daveheitzman]

to be clear, I used this js in my implementation: https://gist.github.com/banksean/300494
I didn't use Mieko's sketch.

[iliabylich]

Fixed in #1540