At 08:25 PM 2/09/2007, you wrote:

>Some days ago, I downloaded the IBObjects components for testing.
>
>I have one problem with Interbase/Firebird databases and I want to
>know if it is possible to solve it with IBObjects.
>
>I'm working with a remote connection to databases and I have a table
>with lots of records, when I use the Last()method to go to the last
>record in table (I'm in a grid) I have to wait lots of time.
>
> How Can I speed up this situation?

Don't use table components (or unlimited queries) for a client/server
database. Use a query with a WHERE clause to limit the rows to only
the ones that the user needs. And avoid (or minimise use of)
workflows that have to step through large datasets in a loop.

> The table has a primary key and I tried to create a second index
>that is the same that primary key but in reverse order (with desc),
>but the speed is the same.

That won't hurt but it won't help, either. The client side doesn't
use indexes in the ways that the VCL enables for local or file-served
databases like Paradox and Access, where indexes are physical objects
that can be navigated. A descending index can be helpful for
descending-order queries, greater-than searches and high-end
aggregations but don't create them unless you actually have a use for them.

The slow part is in your application having to drag all of the
records across the interface. With a Last() call, the dataset has to
keep dragging records until the server's buffer is empty. There is
no way to make this fast enough across a network to make sense with
whole-table sets on large tables. It belongs to desktop databases,
where a "table set" is a local physical file. In client/server,
every search involves round trips between the client application and
databases on remote machines. You need to work zealously on
designing datasets and operations that use client/server features
efficiently. You could pick up the TechInfo sheet "Moving to
Client/Server" from http://www.ibobjects.com/TechInfo.html#ti_Moving_to_CS

That said, a lot of work has been done in IBO over the years to
refine implementation of those desktop-style navigation methods on
bi-directional Fb/IB datasets. Particularly, it improves Locate()
for those who believe they can't live without it and it does some
useful tricks with scrolling. But designing interfaces that scroll
to the end of huge datasets, or depend on a grid interface for
accessing such datasets, is for desktop databases and
spreadsheets. It is not a sensible thing to do in client/server if
good performance is important.

Helen