As far as our app is very dependant on locate method (there is a need to
show open dataset of ~2mil rows, and sometimes locate data), i have
done some research among different borland components
(ado,bde,zeos,ibo,ibx,fib), and here are conclusions : locate is fastest
on indexed columns with bde ttable, but on various queries it seems
fastest with ibo components (something like tquery descendants).

Adomas



David Johnson wrote:

>In an SQL paradigm, locate has no purpose.
>
>It would not be likely to use an index unless you specify an "order by"
>column that is indexed, because conceptually it is a table space scan on
>the entire table.
>
>Instead of using a locate, you would use "select columns from table
>where column = value". Your local dataset would not contain anything
>but the rows you wanted.
>
>On Fri, 2005-05-27 at 15:01 +0200, Thomas Besand wrote:
>
>
>>Hej set and everybody,
>>
>>thanks a lot for your answer. You're definitely right, what you propose is much easier than what what I was planning to do. Seems like it's time to have a break.
>>
>>However, what about the Locate method? Does it use an index if one is available for the fields in question?
>>
>>Greetings,
>>
>>
>>
>
>
>

[Non-text portions of this message have been removed]