On Sun, Apr 5, 2009 at 11:33 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> (It's also worth pointing out that the calculations we do with
>> ndistinct are pretty approximations anyway. If the difference between
>> stadistinct = -1 x 10^-6 and stadistinct = -1.4^10-6 is the thing
>> that's determining whether the planner is picking the correct plan on
>> your 4-billion-row table,
>
> No, it's the loss of ability to set stadistinct to -1e-9 or -1e-12 or
> -1e-15 or so that is bothering me. In a table with billions of rows
> that could become important.
>
> Or maybe not; but the real bottom line here is that it is 100% silly to
> use a different representation in this column than is used in the
> underlying stadistinct column. All you accomplish by that is to impose
> on the user the intersection of the accuracy/range limits of the two
> different representations.
Well, I think I was pretty clear about what I was trying to
accomplish. I think there are more people who care about pg_dump
output being diffable than there are who need to set ndistinct more
accurately than 1 ppm and yet not as an integer. Perhaps if any of
those people are reading this thread they could chime in. Otherwise,
I will implement as you propose.
...Robert
