Re: extend pgbench expressions with functions

>>> 1. It's really not appropriate to fold the documentation changes
>>> raised on the other thread into this patch. I'm not going to commit
>>> something where the commit message is a laundry list of unrelated
>>> changes. Please separate out the documentation changes as a separate
>>> patch. Let's do that first, and then make this patch apply on top of
>>> those changes. The related changes in getGaussianRand etc. should
>>> also be part of that patch, not this one.
>>
>> Hmmm. Attached is a two-part v16.
>
> Thanks. Part 1 looks, on the whole, fine to me, although I think the
> changes to use less whitespace and removing decimal places in the
> documentation are going in the wrong direction. That is:
>
> - About 67% of values are drawn from the middle <literal>1.0 / threshold</>
> + About 67% of values are drawn from the middle <literal>1/param</>,
>
> I would say 1.0 / param, just as we used to say 1.0 / threshold. Any
> reason why not?

For the 1.0 -> 1, this because in the example afterwards I set param to
2.0 and I wanted it clear where the one half was coming from, and ISTM
that the 2.0 stands out more with "1 / 2.0" than with "1.0 / 2.0".

For the spaces, this is because with just "1/" the space seemed less
necessary for clarity, but it seemed necessary with "1.0 /"

Now it is easy to backtrack.

> That's easier to read IMHO and makes it more clear that it's integer
> division.

It is not. One benefit of "1.0" makes it clear that this is a double
division.

> I'm copying Tomas Vondra on this email since he was the one who kicked
> off the other thread where this was previously being discussed.

Fine.

--
Fabien.