From my reading of that statement it says it is going to use the index
to match company_id=1000 and then fetch the row and compare
product_desc. Where as in the the case without the like it is using
product description in the Index Condition.
The speed of the queries certainly seems to bare it out .. The query
where the product_desc is in the index condition, returns almost
immediately where as the other takes about 10 seconds before the first
row is returned.
On Fri, 2005-02-25 at 11:44, Peter Eisentraut wrote:
> Am Freitag, 25. Februar 2005 17:31 schrieb Dave Smith:
> > I am using 7.4.5 and trying to use the like clause (Local C) but
> > postgres does not seem to want to use the index.
> >
> > explain
> > declare t scroll cursor for
> > select * from product where company_id=1000 and product_desc like 'J%'
> > order by company_id,product_desc;
> >
> > QUERY
> > PLAN
> > ---------------------------------------------------------------------------
> >------ Index Scan using product_4 on product (cost=0.00..12306.67 rows=881
> > width=181)
> > Index Cond: (company_id = 1000)
> > Filter: ((product_desc)::text ~~ 'J%'::text)
>
> What part of "Index Scan" are you misunderstanding?
--
Dave Smith
CANdata Systems Ltd
416-493-9020