Hi Erik,

On Jul 21, 2005, at 2:10 AM, erik@jpox.org wrote:

See inline

Erik Bengtson
-----Original Message-----
From: Craig Russell [mailto:Craig.Russell@Sun.COM] 
Sent: Thursday, July 21, 2005 12:26 AM
To: jdo-dev@db.apache.org
Subject: Re: Binding a char to PreparedStatement


I was thinking that most of the value in the jdbc-type would be in the
java-to-database schema creation, where the user could rely on the jdo
impl to create appropriate schema. In the case of char or >Character,
that would be CHAR(1). In the case of double or Double, with a length of
20 and a scale of 12, that would be DOUBLE if the database supported it,
or DECIMAL(20, 12) if not.

CHAR(1) will not allow Unicode storage, you need either to use NCHAR(1),
CHAR(4) or another type that allows you to store 2 bytes.

Correct. The point is what should be the default. I believe that if you are using a single-byte character encoding like Latin or well, some others, that this is sufficient. And columns that need to accommodate national use characters are usually well-known by the application. I believe if the application has to store wide characters, it's ok for the user to have to tell the JDO metadata that you want an INTEGER to store these single characters.

We use INTEGER
data type by default which works in all databases, and more, it allows
operators like > < on SQL queries without having to convert it using
ASCII functions.

Unfortunately, using < and > on SQL queries of characters converted from Unicode to numbers using a trivial translation is meaningless due to various ordering schemes that are often locale-dependent.

The origin of the problem Michael reported is down to the lack of a
jdbc-type 
specification in the metadata, and the fact that JPOX's current
"default" for 
char is to store it as INTEGER (for whatever reason) - hence why he got
the 
problem.

I'd say it is the lack of a default in the specification as opposed to the lack of a jdbc-type in the metadata. The usage for single character columns is often trivial: M/F for gender, A/B/C/D/E for answer, 0/1 for binary data where the database doesn't understand boolean, etc.

To summarize, I'd prefer that the default be whatever the JDO experts agree is the default, and we adapt the TCK to specify the jdbc-type in case the default turns out to be something other than CHAR.

Ok? I'll see you on the JDO expert alias. :-)

Craig

Ok. I thought that you would agree that this is a bug in JPOX, since the
"obvious default" for a char column is CHAR.


It is not that obvious when you have to support a wide range of
databases and data types. The most obvious choice to me is the most
common type among these databases.




Craig Russell

Architect, Sun Java Enterprise System http://java.sun.com/products/jdo

408 276-5638 mailto:Craig.Russell@sun.com

P.S. A good JDO? O, Gasp!