Neil Conway wrote:
> Clearly, the current approach is fine when the array is small -- perhaps
> for arrays above a certain number of elements, we could switch to
> randomly sampling array elements, estimating their selectivities, and
> then using that information to infer the estimated selectivity of the
> entire array expression. That seems fairly crude, though: does anyone
> have any better ideas?
Optimizer hints in SQL
/me ducks and runs for cover.
regards,
Lukas