Re: Tuple sampling

On Sun, 23 May 2004 17:32:36 -0400, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>I took out the TupleCount typedef and went
>back to using doubles for the tuple counts; this is more consistent with
>the coding style used elsewhere, and I really doubt that it's any
>slower.

Performance was not the primary motivation. I found it confusing to
have doubles everywhere and not to know whether a variable is declared
as double, because
. we need the fractional part (e.g. a probability)
. or it should be able to hold an integral value of more than 32 bits.
So I just invented my own datatype for huge integers. Long long would
have been a natural choice, but AFAIK its not available on all
platforms.

>I was initially convinced that your implementation of Knuth's algorithm
>S was all wet, so now there's a bunch of comments explaining why it's
>actually correct...

Thanks. I like your explanation. My justification for that change was
a lot more complicated.

Servus
Manfred