> Surely the most important point in the OP was that ineqsel does not
> correctly binary search in the presence of duplicates.
>
It would have been if I were correct :-( .
Looking at it again, that was from a bug in my code. Thanks for your
time, and sorry about the noise.
-Nathan
--
Sent via
On Tue, 2009-01-06 at 18:40 -0500, Tom Lane wrote:
> "Nathan Boley" writes:
> >> I don't think this is a bug.
>
> > hmmm... Well, I assumed it was a bug from a comment in analyze.
>
> > From ( near ) line 2130 in analyze.c
>
> > * least 2 instances in the sample. Also, we won't suppress valu
"Nathan Boley" writes:
>> I don't think this is a bug.
> hmmm... Well, I assumed it was a bug from a comment in analyze.
> From ( near ) line 2130 in analyze.c
> * least 2 instances in the sample. Also, we won't suppress values
> * that have a frequency of at least 1/K where K is the intende
>> For heavy tailed distributions, it is possible for analyze to
>> duplicate histogram boundaries.
>
> I don't think this is a bug.
hmmm... Well, I assumed it was a bug from a comment in analyze.
>From ( near ) line 2130 in analyze.c
* least 2 instances in the sample. Also, we won't suppress
"Nathan Boley" writes:
> For heavy tailed distributions, it is possible for analyze to
> duplicate histogram boundaries.
I don't think this is a bug. You've got values that didn't make it into
the MCV list, but nonetheless occupy multiple buckets' worth of space in
the remainder of the distribut
