On Monday, 25 January 2016 at 23:37:25 UTC, Andrei Alexandrescu
wrote:
On 01/25/2016 05:05 PM, Joseph Rushton Wakeling wrote:
One option would be to implement the basic RNG data structor à
la C++,
as a functor
That's semantically the same as an input range. -- Andrei
“Yes, but...” :-P
There are actually some interesting subtleties required for the
input range design, not just for RNGs but for ANY range whose
elements are generated randomly.
I will try and write this up properly later, but the TL;DR is, it
involves doing extra work to ensure every element is _truly_
lazily evaluated.
To some extent, splitting into functor and range wrapper can help
clarify the code design there (even if the functor stays hidden
as an implementation detail).
