Poor performance of CMP entity bean finder that uses unicode notation N''

Question

I am using Weblogic 6.1 with SQL Server 2000. I have an entity bean
with a finder method that does a lookup based on a (indexed) varchar
field. I noticed during testing that the performance of this finder
method was much worse than other finders. I did some profiling of the
SQL calls going from the Weblogic container to the database, and
noticed that the query generated for the finder call used the unicode
notation, which seems to result in very poor performance. If you run
the following two queries through SQL Analyzer (or SQL Profiler),
you'll notice a significant performance hit on the query that uses the
unicode notation:
example of generated SQL:
SELECT WL0.field1, WL0.field2 ... FROM tblAccount WL0 WHERE
WL0.accountId = N'40879'
example of desired SQL:
SELECT WL0.field1, WL0.field2 ... FROM tblAccount WL0 WHERE
WL0.accountId = '40879'
Is there any way to force it to generate the SQL without using the
notation with the unicode 'N'?
Thanks for any help,
Don

Answer 1

Don wrote:
That worked (we use the Weblogic JDBC driver). The performance is MUCH
better! Glad to help.
Is there some place in the documentation where I should have
seen this (ie. where can I find the documentation on other possibly
useful connection properties)?It should be documented in the jdbc driver documents...
Joe
>
Thanks,
Don Tranquillo
Joe Weinstein wrote:
Hi. The driver does this to ensure that any 16-bit character string
that Java has,
is absorbed unaltered into the DBMS. There is no way to know a-priori
whether
the DBMS needs/wants an nvarchar datum or an 8-bit varchar datum. As
you see,
there is a big performance issue. You can tell the jdbc driver to
send data
as 8-bit (without the "N'" prefix) by setting a connection property:
If you are using the weblogic mssqlserver driver, add the
useVarChars=true
property. If you are using the MS free driver or the DataDirect driver,
add a property sendStringParametersAsUnicode=false.
Joe
Don wrote:
I am using Weblogic 6.1 with SQL Server 2000. I have an entity bean
with a finder method that does a lookup based on a (indexed) varchar
field. I noticed during testing that the performance of this finder
method was much worse than other finders. I did some profiling of the
SQL calls going from the Weblogic container to the database, and
noticed that the query generated for the finder call used the unicode
notation, which seems to result in very poor performance. If you run
the following two queries through SQL Analyzer (or SQL Profiler),
you'll notice a significant performance hit on the query that uses the
unicode notation:
example of generated SQL:
SELECT WL0.field1, WL0.field2 ... FROM tblAccount WL0 WHERE
WL0.accountId = N'40879'
example of desired SQL:
SELECT WL0.field1, WL0.field2 ... FROM tblAccount WL0 WHERE
WL0.accountId = '40879'
Is there any way to force it to generate the SQL without using the
notation with the unicode 'N'?
Thanks for any help,
Don