Hibernate oracle unicode support

Myk · **Joined:** Sun Jan 04, 2004 2:46 pm **Posts:** 147

This easiest way is to specify

Code:

hibernate.connection.defaultNChar true

in your hibernate properties file. This will cause ALL string parameters to be treated as unicode. This might be 10g only tho ( not sure ) and you'll need the latest JDBC drivers from the oracle website.

Using DBCP you can unwrap the prepared statement by doing.

Code:

//                DelegatingPreparedStatement ps = (DelegatingPreparedStatement)preparedStatement;
//                OraclePreparedStatement ops = (OraclePreparedStatement)ps.getInnermostDelegate();

and then the setFormOfUse thing. C3P0 I don't think this is possible ( cast to NewProxyPreparedStatement OK, but theres no public getter for the inner PreparedStatement ). Proxool I didn't try.

michael · **Posted:** Mon Sep 20, 2004 9:30 am

With Proxool you should be able to just cast to OraclePreparedStatement AFAIK

bertrand · **Posted:** Mon Sep 20, 2004 10:03 am

michael wrote:

With Proxool you should be able to just cast to OraclePreparedStatement AFAIK

The version of Proxool supporting this feature is not yet released - please grab a copy from CVS in the meantime.

michael · **Posted:** Mon Sep 20, 2004 11:13 am

Thanks Bertrand, proxool guru :) Thats a nice feature by the way, I really wonder why not more pools are supporting this - with the wide spread of dynamic proxies/cglib now have.

michael · **Posted:** Mon Sep 20, 2004 11:13 am

Ah, english grammar no good :)

bertrand · **Posted:** Tue Sep 21, 2004 3:29 am

michael wrote:

Thanks Bertrand, proxool guru :) Thats a nice feature by the way, I really wonder why not more pools are supporting this - with the wide spread of dynamic proxies/cglib now have.

Yes indeed... unfortunately, this feature is in 'early beta' for now and hasn't been intensively tested yet (no more free time to work on the project :(

joe_the_quick · **Joined:** Fri Mar 26, 2004 8:44 am **Posts:** 10

I'll test this using Oracle 9i and Oracle 10g and let you know.

BTW - this workaround is only necessary if the database character set is NOT UTF8(16) (if it is UTF8, nvarchar and varchar are effectively the same).

If the Oracle database is non-UTF8, setFormOfUse is necessary for any correct updates using Unicode-Characters.

johannes

bertrand · **Posted:** Tue Sep 21, 2004 3:27 pm

joe_the_quick wrote:

I'll test this using Oracle 9i and Oracle 10g and let you know.

Make sure you are using oracle.jdbc.OracleDriver and not the old *deprecated* oracle.jdbc.driver.OracleDriver (as exlained in the Oracle release notes).

This new version of the jdbc driver is made of interfaces instead of concrete classes (as the old one) - making it possible for Proxool to implement them as well...

christian · **Posted:** Tue Sep 21, 2004 3:28 pm

Again, you have to applaud Oracle for the choice of names. Well done... not.

bertrand · **Posted:** Tue Sep 21, 2004 4:04 pm

christian wrote:

Again, you have to applaud Oracle for the choice of names. Well done... not.

Indeed...
To be honest, I'm quite new to Oracle and I'm very disapointed by their JDBC driver (and their support!).

We already spent lots of hours tracking bugs in our applications, thinking we made a mistake somewhere... We never thought it could be a problem in the JDBC driver... Now it's becoming the first thing we look at :(

Myk · **Joined:** Sun Jan 04, 2004 2:46 pm **Posts:** 147

Would it be possible to post a quick list of the bugs you've found in the driver? I've just started working with Oracle so it would be very useful for me and probably others. :)

bertrand · **Posted:** Wed Sep 22, 2004 12:38 pm

Don't remember them all, but I'm pretty sure you can ask Google for more info ;)

From what I remember:

- problems with types read from the ResultSet using a Stream (BLOB's, and other large types like the byte[] - maybe others): they must be fetch the latest otherwise the resultset complains with a 'stream closed' exception. This means these types cannot be used in conjunction with a 'join fetch';

- fields must be retrieved in the order they appear in the select clause (not sure it is still applicable for the very latest versions);

- this one is not a bug but a feature (very annoying feature): empty strings are replaced by NULL values. Very confusing when you persist your classes with an empty string and you get it back with a null value. Even more annoying when you need to make the distinction between null and empty - which ARE two different things;

That's all I remember from now... maybe we should add some wiki pages with the known issues with regards databases (not sure Gavin wants to turn his site into an Oracle bug tracking though ;)

owenfellows · **Joined:** Wed Sep 20, 2006 4:09 am **Posts:** 1

If you are using weblogic you can retrieve the actuall connection from the proxy like this. It is not very nice but it seems to work.

Code:

  Field declaredField = ps.getClass().getField("vendorObj");
  OraclePreparedStatement preparedStatement = (OraclePreparedStatement) declaredField.get(ps);
  preparedStatement.setFormOfUse(1, OraclePreparedStatement.FORM_NCHAR);

Not sure if this helps.[/code]

keneida · **Joined:** Mon Nov 26, 2007 10:37 am **Posts:** 2

i have similar problem.
I have 6 oracle 9.2 databases and on 4 of them the system works nice.
I want to store xml inside xml and it does so. But those 2 broken DB put some rubish characters inside.
When i use regular jdbc it works good. I do not know what the problem might be.

quangloc · **Joined:** Tue Mar 11, 2008 3:58 pm **Posts:** 1

Since neither of setting the defaultNChar or setFormOfUse is an option for me, I'm converting the data into a string of Unicode hex notations ('\xxxx') and insert it into the nvarchar2 column using the SQL unistr(hex notations).

I'm wondering if there is a setting in Hibernate to generate the unistr() SQL automatically for a property.