Re: [Firebird-devel] Count(*) faster without index ?

[Dmitry Y. <fir...@ya...>]

Derryck welas wrote:
> 
> I always assumed using an index works faster

Generally, this is a false assumption. Index is useful only if it's 
selective enough.

> select count(*) as count_order
> from   lineitem
> where  l_shipdate <= date '1998-12-01' - 90;

How many rows is in lineitem: (a) totally and (b) corresponding your 
predicate? I suppose the difference between these two numbers is not 
very big.

> Can i put this in the tracker ?

Depending on what do you want us to improve :-) Do you want the 
optimizer to ignore that index in your query? But this is impossible 
without value distribution histograms in table/index statistics.


Dmitry