Numbers and ranges

ransu.random(); // [0, 1)
ransu.float(-1, 1); // [min, max)
ransu.integer(1, 6); // [min, max] — inclusive
ransu.below(arr.length); // [0, n)     — what indices want
ransu.range(0, 100, 5); // Python's randrange
ransu.bigint(0n, 2n ** 128n);
ransu.bool(); // true or false, evenly
ransu.chance(0.25); // true a quarter of the time
ransu.oneIn(20);
ransu.sign(); // -1 or 1
ransu.bits(12); // Python's getrandbits
ransu.bytes(32); // Uint8Array
ransu.floats(1000); // Float64Array
ransu.integers(1000, 1, 6); // bounds validated once, not per element

Which end is included

This is the one thing worth memorising.

Call	Range
integer(min, max)	[min, max] — both ends
bigint(min, max)	[min, max] — both ends
below(n)	[0, n)
range(start, stop, step)	[start, stop)
float(min, max)	[min, max)
date(from, to)	[from, to)

integer matches Python’s randint and a die: integer(1, 6) can return 6. For an array index you want below(arr.length) or pickIndex(arr) — writing integer(0, arr.length) is the classic off-by-one, and no type can catch it.

No modulo bias

Ranges up to 2^32 use Lemire’s nearly-divisionless method; larger ones use rejection over two words. Anything wider than 2^53 throws RANGE_TOO_LARGE and points you at bigint, rather than silently rounding.

ransu.integer(0, Number.MAX_SAFE_INTEGER);
// RansuError: integer(...) spans more than 2^53 values … Use bigint(min, max)

Bulk draws

floats(n) and integers(n, min, max) return typed arrays and validate their bounds once instead of per element. Use them for simulations.

const noise = ransu.floats(1_000_000);
const rolls = ransu.integers(1_000_000, 1, 6);