--- In firebird-support@yahoogroups.com, Helen Borrie <helebor@...> wrote:

>
> At 09:14 PM 10/12/2008, you wrote:
> >Hi Folks:
> >
> > I had a bug in a function that is passed a table
> >name and a where clause, and returns a count of the
> >records that qualify for the where clause, if one is
> >passed in.
> >
> > The function builds an SQL of
> >
> > SELECT
> > COUNT(*) FROM
> > <Table Name>
> > WHERE <--------- If where clause passed in
> > <Where Clause>;
> >
> > While tracking down the problem I read in Helen's
> >book that SELECT COUNT(*) is inefficient.
>
> It's an inefficient (and not very safe) way to perform an existence

test.

>
>
> > Should I re-think the use of COUNT(*) in this function?
>
> As long as you *need* it to COUNT things, no, there's no other way

to count things. But if the idea is to count all the records so you
can know ahead of time how many records there are going to be - as
e.g. for a progress bar - then that's inefficient. But there are
still a lot of people out there writing apps like that....and
complaining about Firebird's performance... ;-)

Thanks Martijn, Dimitry and Helen:

This is, in fact, about displaying a progress window.

I start a transaction, then get the count of records to
be read, and display that. Then I run the select, updating
the display of records read after every 100 fetches, and
then I commit the transaction.

This application is currently using the embedded server,
so at this time the database won't be changed. And if it
is, so what? We have to start a transaction at some point,
and if the

So I am using this count for exactly the application you
say is inefficient. How would you do this?

Thanks
Larry