Hi,

At August 6, 2003, 08:08, Paul Vinkenoog wrote:

> Daniel Rail wrote:

>> > Can anyone address to this issue?
>>
>> What about simply using the following SQL statement:
>> UPDATE TABLE SET FIELDA=FIELDA+100
>>
>> This should be a lot quicker.
>>
>> Accessing a RDBMS in the same way you would Paradox, is not always
>> the most efficient way. Sometimes, you have to learn what can be
>> done natively with the RDBMS.

> But what James did is still interesting as a speed test, and he found
> IBX 2.5 - 4 times as fast as IBO (in these particular tests).

I do agree. It is interesting as a speed test.

> Maybe this is unavoidable, as IBO is a lot richer and therefore more
> complicated than IBX.

This could be due to how the components are setup, i.e. PreparedEdits
vs non-prepared edits. And, how fast the TCP/IP connection is. If it's
performed with non-prepared edits and no metadata cache, then I can
see how the speed difference can be, and it's amplified with slower
connection speeds.

> James, what happens if you don't attach any DataSource to the queries,
> and run the tests again?

And, what happens when using PreparedEdits=True and/or the metadata
cache enabled? It's just that when IBO prepares a statement, it parses
it and fetches metadata from the server to try to validate the
statement before sending it to the server. So, if you use
PreparedEdits, the prepare stage is performed once, and not every time
the application sends an UPDATE statement.

--
Best regards,
Daniel Rail
Senior System Engineer
ACCRA Group Inc. (www.accra.ca)
ACCRA Med Software Inc. (www.filopto.com)