Not just [0,1), but also random int

[Artoria2e5]

In #19 (comment) I talked about how it might be useful to have a "correct dice roll" as a tangential. I still think that's important (people mess that up more commonly do you think), so I'm making a new issue.

Beyond just [0,1), I see potential use for a method for a random integer.

The .randomInt() method

To obtain a random integer number from a PRNG object, the object has a .randomInt() method. On each invocation, it will output an appropriate pseudo-random number based on its seed, and then update its seed for the next invocation.

• When .randomInt() is invoked with zero arguments, it returns a random unsigned 32-bit integer in the range [0, 2^32-1].
• When .randomInt() is invoked with one argument x, it returns a random integer in the range [0, h). x must be in the range [1, Number.MAX_SAFE_INTEGER].
• When .randomInt() is invoked with two argument l, h, it returns a random integer in the range [l, h). h-l must be in the range [1, Number.MAX_SAFE_INTEGER].

Recommended implementation

See swiftlang/swift#39143.

Open questions

• Do we return a number constrained to uint53 (MAX_SAFE_INTEGER), a 64-bit-or-whatever BigInt, or a number constrained to uint32?

[bakkot]

See https://github.com/tc39-transfer/proposal-random-functions.

[michaelficarra]

Also duplicate of #21.

[Artoria2e5]

Aye, will close.

[tabatkins]

Yup, the current intention is that the seeded PRNG will support .random(), acting identically to Math.random() (but seeded), and then additionally support any and all additional randomness methods the language grows. I don't intend to innovate in that regard in this proposal.

[Artoria2e5]

My original motivation is that "64 bits of random" is more fundamental to an RNG than "[0,1) double", but I guess that alone doesn't mean it has to be in the same spec. After all everyone knows that.

The consideration around MAX_SAFE_INTEGER should maybe also get into your separate proposal. Or maybe we get a BigInt output, I still can't decide.