> That is a bit problematic because it necessitates knowing the number
> of rows total, and slow counting is an idiosyncrasy of postgres.
>
> http://wiki.postgresql.org/wiki/Slow_Counting
>
> To get the top 10%:
>
> SELECT * FROM table LIMIT(SELECT (COUNT(*) * 0.1)::integer FROM table)
I think I wasn't making myself clear. I don't want the top 10% of the
rows. I want the rows with the top 10% of the values in a column.
In my case there is a very non linear set of values. The lowest value
is 1 and the vast majority of records have a 1 in the column. The
highest value might be in the tens of thousands. I want to pull up
the records that have the top ten percent values.