Would you argue that 1/6, 3/6 and 5/6 are not equidistant? What's the compelling use case to add this?

They are not equidistant in the sense that the line is not broken into equal length segments. If one only focuses on the distance between the intermediate points then one can argue that 1/10, 5/10 and 9/10 are just as equidistant as 1/6, 3/6 and 5/6. We are trying to use the function to reconstruct the position of individuals at equal intervals as they move from point A to point B. We know where they are at the start and end and want to know where they are when they are 1/4, 2/4 and 3/4 of the way along.

Don't you want 0 and 1 included, then?

You can argue it either way - since they are easy to include or exclude. I would prefer 0 and 1 be included.

the only trouble is if one sets n = 1 it is not possible to include 0 and 1.

I guess one could go
n = 1: 1/2
n = 2 0/2 and 2/2
n = 3 0/2 and 1/2 and 2/2

How about adding an argument, e.g. sample, that should take numbers between 0 and 1 (inclusive) that are then being used instead of fn(n[i])?

I like that idea. Do you want me to implement? and do you think sample should replace the arguments n, density and type?

Yes; no, but it shall overrule them.

Roger that - await my pull request!

Seems to me little point in passing a function unless it is going to use n i.e. unless it takes one or more arguments all it is going to do is return a vector of numbers between 0 and 1. Would you like sample to be a vector of numbers between 0 and 1 or would you like sample to take n (or another argument)?

a vector with numbers between 0 and 1 inclusive.

Roger that

pull request
#300