> On Tue, 28 Aug 2001, Reinoud van Leeuwen wrote:
>
>> Can somebody explain to me:
>>
>> > radius=# explain select count (radiuspk) from radius ;
>> > NOTICE: QUERY PLAN:
>> >
>> > Aggregate (cost=12839.79..12839.79 rows=1 width=8)
>> > -> Seq Scan on radius (cost=0.00..11843.43 rows=398543 width=8)
>> >
>> > EXPLAIN
>>
>>
>> This query answers me *instantly* after hitting return
>>
>> > radius=# select count (radiuspk) from radius ;
>> > count
>> > --------
>> > 398543
>> > (1 row)
>>
>> This query takes about 3 seconds. But the query plan *already* knows
>> the number of rows ("rows=398543"). So why does it take 3 seconds. Is
>> my assumption correct that the optimiser still can be optimized a
>> little? :-)
>
> Not in this case. The row numbers from explain are just estimates
> from the last vacuum. As you modify the table, the estimated rows will
> be off.
Yes, I just found out that somebody else is running a script on our test
server that vacuums all databases each night. That explains a lot.
Thanx for thinking with me
Reinoud