can't use rmultinom with n = 1

[bluelovers]

can't use libRmath.Multinomial rmultinom with n = 1, only when n > 1, it will not throw error

  const result = sequence(n).map(() => _rmultinom(size, prob, rng));
                             ^
TypeError: sequence(...).map is not a function

[jacobbogers]

Thanks, I will take a look!

[jacobbogers]

Hi,
the n=1 returns a scalar, this was based on my wrong understanding of R in edge cases were n=1 or n=0. aka runinf(0) or runif(1) for example, I thought they returned scalars but they were arrays

For now you can use c to always make it an array c(sequece(1)).map(.. as a workaround

The new version will seperate out the pure math functions and helper functions. the work is already done for all functions, but i need to update the docs)

[bluelovers]

@jacobbogers thx for reply, but libRmath.Multinomial(rng).rmultinom

what is c ?

[jacobbogers]

c is an R function that is ported to javascript aswell https://github.com/R-js/libRmath.js#flatten-or-c-alias
R counterpart https://www.rdocumentation.org/packages/base/versions/3.5.1/topics/c

[jacobbogers]

So
c( sequence(1) ) <- an array
c(1,2,3) <- an array
c(1,2, c(4,5)) <- an array [1,2,4,5]

[bluelovers]

another issues in npm v1.0.83

lib-r-math.js\dist\lib\webpack:\src\lib\multinom\rmultinom.ts:31
  const result = sequence(n).map(() => _rmultinom(size, prob, rng));
                             ^
TypeError: sequence(...).map is not a function

but in

libRmath.js/src/lib/multinom/rmultinom.ts

Lines 30 to 37 in f775ba8

	export function rmultinom(
	n: number,
	size: number,
	prob: number | number[],
	rng: IRNG
	): number[] {
	return randomGenHelper(n, rmultinomOne, size, prob, rng)
	}

is use randomGenHelper, but i see blame show it change at 2 month ago
does this also not yet push to npm too?

export function rmultinom(
  n: number,
  size: number,
  prob: number | number[],
  rng: IRNG
): number[] {
  return randomGenHelper(n, rmultinomOne, size, prob, rng)
}

[jacobbogers]

Hi Bluelovers
Thank you for your interest, awesome

btw: Did the 'c' function work for you?

Yes, the source code on git is already adjusted for the new situation, but documentation not adjusted yet)) There are some breaking changes, but needs be crearly documented. I need to swing around from my other projects to do this,

Pull request welcome)))

[bluelovers]

@jacobbogers

im use libRmath.Multinomial(rng).rmultinom in current lib version at npm

there is no way import _rmultinom and re-pack to c(sequece(n)).map(() => _rmultinom(size, prob, rng))

so i can't use c for rmultinom

i think pr is no need, ur source code is use randomGenHelper, but just not yet publish to npm

[jacobbogers]

Ah i understand what you are trying to do i think, you can wrap things is c it will stay forward compatible.

This should work
c( sequece(n) ).map(() => c( rmultinom(size, prob, rng) ) )

If you want, you can post your code to jsfiddle, and I can take a look

[bluelovers]

demo code about what i saying lol

import libRmath = require('lib-r-math.js');

let rmultinomFn = libRmath.Multinomial(fakeLibRmathRng(Math.random)).rmultinom;

/**
 * n can't be 1 at current lib-r-math.js version (publish in npm)
 * 
 * and source code already change at github,
 * but not yet publish to npm
 */
let n = 1
let size = 21;

let ret = rmultinomFn(n, size, [1, 2]) // here will throw

console.log(ret);

export function fakeLibRmathRng(fn: () => number)
{
	return {
		unif_rand(n?: number)
		{
			if (n > 1)
			{
				let a = [];
				while (n--)
				{
					a[n] = fn()
				}
				return a;
			}

			return fn()
		},
	} as libRmath.IRNG
}

[jacobbogers]

hi,
This is most certainly a bug!, the current npm version needs to be patched and released

thanks for finding it

const libR = require('lib-r-math.js')

const { rmultinom } = libR.Multinomial(); // use default PRNG

const relativeProbabilityWeights = [167, 500, 167, 167]

const data = rmultinom(1, 21, relativeProbabilityWeights); --> error

[jacobbogers]

can you re-open this issue?) thanks (you have closed it)

[bluelovers]

I thought u are already fix it, just waiting some other job done, so didn't publish lol