Andrew Gierth writes:
> Just got a report on IRC of a bug in the array version of
> percentile_cont; if two of the requested percentiles were between the
> same pair of input rows, the result could be wrong or an error would
> be generated.
Oooh, good catch.
> Proposed patch (against current mas
Just got a report on IRC of a bug in the array version of
percentile_cont; if two of the requested percentiles were between the
same pair of input rows, the result could be wrong or an error would
be generated.
e.g.
select percentile_cont(array[0.4,0.6]) within group (order by gs)
from generate
