@Suever One of these days we need to assign & for r. Possibilities: (a) 2$ (b) 1$ (c) 1$ and make r with 1 input behave with an implicit leading 1, so 3&r would be rand([1,3]))
(a) If & is defined as 2$ (and no leading zero for single scalar input; that would be useless): rand(1,3): l3&r rand(3): 3t&r rand(3,4): 3K&r
(b) If & is defined as 1$ (and no leading zero for single scalar input): rand(1,3): l3h&r or l3H$r rand(3): 3&r rand(3,4): 3Kh&r or 3KH$r
(c) If & is defined as 1$ and leading 1 is implicitly added to single scalar input: rand(1,3): 3&r rand(3): 3th&r or 3t2$r rand(3,4): 3Kh&r or 3K2$r
So its (a) 4,4,4 bytes versus (b) 5,3,5 versus (c) 3,5,5. Option (b) doesn't seem useful; the square case is not that common. Between (a) and (b), the question is, is the rand(1,n) case so much more common to warrant saving 1 byte at the cost of the other cases gaining 1 byte?
(a) is like l and O. (c) is like Yr
Hm no, Yr is different, because its default is 1$, not 0$