randomSample with unknown length

Magnus Lie Hetland magnus at hetland.org
Wed Feb 2 12:06:44 PST 2011


On 2011-02-02 16:32:25 +0100, Andrei Alexandrescu said:

> randomSample is designed to subsample a large stream in constant space 
> and without needing to scan the entire stream in order to output the 
> first element.

Sure. I was just thinking that you could have a version for the cases 
where there was no end in sight :)

> I used in in my dissertation where e.g. I had to select 100K samples 
> from a stream of many millions.

Cool.

> Having a reservoir sample available would be nice. I'd be thrilled if 
> you coded up a reservoirSample(r, n) function for addition to 
> std.random.

Seems Simen beat me to it :)

-- 
Magnus Lie Hetland
http://hetland.org



More information about the Digitalmars-d mailing list