🚀 go-pugleaf

I cannot confirm that it will be faster in all cases.

But _generally_, "sequential scan and in-memory sorting is faster than
several index scans". (C) Someone from FB dev team.
And I can confirm that in most cases it's so.
Especially if your indicies are not very selective.

So it all depends.

Probably the ultimate way would be:

   declare variable v1 integer;
   declare variable v2 integer;

   v1 = 0;
   v2 = 0;

   select field_a, field_b
   from t
   into :v_a, :v_b
 do begin
   if (v_a = 1) then v1 = v1 + v_b;
   else if (v_a = 2) then v2 = v2 + v_b;
   /* and more if's if ther are more distinct values in field_a */
 end

But here you again trade code clarity and "declarativeness" for performance.

Just test it.

But if there are many different values in field_a and you count only sums
for subset of them than several indexed scans would be faster of course.


Maxim.


"Isaac" <no@no.com> ???????/???????? ? ???????? ?????????:
news:483c12b2@newsgroups.borland.com...
> Maxim Shiryaev wrote:
> > Hi.
> >
> > Try:
> >
> > for
> >   select field_a, sum(field_b)
> >   from t
> >   group by field_a
> >   into :v_a, :v_sum_b
> > do begin
> >   if (v_a) = 1 then some_var_1 = v_sum_b;
> >   else if (v_a) = 2 then some_var_2 = v_sum_b;
> >   /* and more if's if ther are more distinct values in field_a */
> > end
> Is this faster than using Two Selects
>
> Select sum(f2) from table_1 where f1 = 1
> Select sum(f2) from table_2 where f1 = 0
>
> Thanks