Re: JDBC / Oracle / MS SQL Server performance anomalies

"Niall Litchfield" <n-litchfield_at_audit-commission.gov.uk> wrote in message news:<3e35444a$0$247$ed9e5944_at_reading.news.pipex.net>...
> I did a similar test to you (code shown below) tab1 has a number,date and
> varchar column for the original tests, I added an NVARCHAR2 after rereading.
> Results were
>
> using a statement 30,000 inserts = 113s, delete 4s.
> using a prepared statement 30,000 inserts 39s,delete 4s. adding the
> nvarchar2 column 30,000 inserts 41s. I didn't bother batching inserts but
> would expect a speed increase from that as well. In other words thats over
> twice as fast.

Yes thats almost identical to the results from my tests of a similar nature.

> You'll note that I didn't hit any Oracle bugs or use any Oracle objects.
> What bug am I missing?

Look at this link:

http://download-west.oracle.com/docs/cd/A91202_01/901_doc/java.901/a90211/oraint.htm#1058698

Scan forward for the notes surrounding setFormOfUse().

*However*, when I went hunting for the docs I found this in the "fixed bugs" section of the README for the Oracle 9.2.0.1 drivers:

BUG-1730903 (since 9.0.0)
setFormOfUse() method must be put before blinding/definding SQL NCHAR datatypes. For example, setString() and registerOutParameter() has to be called after setFormOfUse(). Otherwise, unpredictable results will be received. This restriction applies to all JDBC drivers.

I should have seen this before. Thanks for making me look for it.

Omi Received on Mon Jan 27 2003 - 15:55:44 CST