Re: Why does query planner choose slower BitmapAnd ?

On Mon, Feb 22, 2016, at 02:30 PM, Jeff Janes wrote:
> It charges 0.1 CPU_operator_cost, while reality seemed to be more like 6 CPU_operator_cost.

fdy=> select name, setting, boot_val from pg_settings where name ~
'cpu';
         name         | setting | boot_val
----------------------+---------+----------
 cpu_index_tuple_cost | 0.005   | 0.005
 cpu_operator_cost    | 0.0025  | 0.0025
 cpu_tuple_cost       | 0.01    | 0.01
(3 rows)

Inspired, I changed cpu_index_tuple_cost to 0.1 (default: 0.005). It
"fixed" my problem by preventing the BitmapAnd.

Is this dangerous?