DB name shortening algorithm probably needs more sophistication

Question

Hi,
I think it might be wise to come up with a more sophisticated way to
shorten the names when the class names and field names are too long. I
ran into the following situation with a class called
"AttributeCollection" with a HashSet field called "attributes". The
elements of attributes are instances of "Attribute" a simple persistent
class with a "name" and "value" string. Changing the field name to
"set" fixed it. The only reason that I found the problem relatively
easily is that I remembered observing the shortening algorithm (snip) in
another harmless case.
David Ezzio
Yankee Software
E:\MyCode>kodoschematool -action refresh
Refreshing schema for type "class
com.ysoft.jdo.attributes.AttributeCollection" and known subclasses...
CREATE TABLE ATTRIBUTECOLLECTION_ATTRIBUTEX (ATTRIBUTESX NUMBER, JDOIDX
NUMBER)
CREATE INDEX ATTRIBUTECOLLECTION_ATTRIBUTEX ON
ATTRIBUTECOLLECTION_ATTRIBUTEX (ATTRIBUTESX)
CREATE INDEX ATTRIBUTECOLLECTION_ATTRIBUTEX ON
ATTRIBUTECOLLECTION_ATTRIBUTEX (JDOIDX)
java.sql.SQLException: ORA-00955: name is already used by an existing
object
E:\MyCode>kodoschematool -action refresh
Refreshing schema for type "class
com.ysoft.jdo.attributes.AttributeCollection" and known subclasses...
CREATE TABLE ATTRIBUTECOLLECTION_SETX (SETX NUMBER, JDOIDX NUMBER)
CREATE INDEX ATTRIBUTECOLLECTION_SETXSETX ON ATTRIBUTECOLLECTION_SETX
(SETX)
CREATE INDEX ATTRIBUTECOLLECTION_SETXJDOIDX ON ATTRIBUTECOLLECTION_SETX
(JDOIDX)
CREATE TABLE ATTRIBUTECOLLECTIONX (JDOIDX NUMBER NOT NULL, JDOLOCKX
NUMBER, JDOCLASSX VARCHAR2(255), PRIMARY KEY (JDOIDX))
Refreshing schema for type "class com.ysoft.jdo.attributes.Attribute"
and known subclasses...
CREATE TABLE ATTRIBUTEX (JDOIDX NUMBER NOT NULL, JDOLOCKX NUMBER,
JDOCLASSX VARCHAR2(255), NAMEX VARCHAR2(255), VALUEX VARCHAR2(255),
PRIMARY KEY (JDOIDX))

Answer 1

Hi,
I had the same problem about a month ago. Patrick Linskey suggested in his post from 12.09.01 "Re: Schematool problem" to "use the 'table' class-level extension in
your system.jdo file". But this didn't seem to work, so I shortened the name of the class to fix it.
Here is Patricks original post:
Martin,
As Joe Mannerling pointed out earlier, the problem that you are seeing
is related to a bug in Kodo's table name generation algorithm. In
order to keep table and column names below database limits (Oracle,
for example, has a 30 character limit), Kodo truncates names to this
length. Unfortunately, it performs this truncation after assembling
the name, so if a class name is over a certain length, then name
conflicts such as those that you are seeing will arise.
We are working on a better name generation algorithm to resolve this
problem. In the mean time, use the 'table' class-level extension in
your system.jdo file:
<?xml version="1.0"?>
<jdo>
     <package name="de.mediateam.quizmaster.data">
          <... snip ...>
          <class name="Question">
               <field name="prerequisites">
                    <collection element-type="Question" />
               </field>
          </class>
          <class name="TextQuestion" persistence-capable-superclass="Question" />
<!-- change this section -->
          <class name="MultipleChoiceQuestion"
persistence-capable-superclass="Question" >
               <extension key="table" value="MCQuestionX"/>
</class>
<!-- end changes -->
          <class name="BooleanQuestion" persistence-capable-superclass="Question" />
          <... snip ...>
     </package>
</jdo>
Thanks, and good luck.
Cheers
Martin
On Thu, 18 Oct 2001 08:44:11 -0400, David Ezzio <[email protected]> wrote:
Hi,
I think it might be wise to come up with a more sophisticated way to
shorten the names when the class names and field names are too long. I
ran into the following situation with a class called
"AttributeCollection" with a HashSet field called "attributes". The
elements of attributes are instances of "Attribute" a simple persistent
class with a "name" and "value" string. Changing the field name to
"set" fixed it. The only reason that I found the problem relatively
easily is that I remembered observing the shortening algorithm (snip) in
another harmless case.
David Ezzio
Yankee Software
E:\MyCode>kodoschematool -action refresh
Refreshing schema for type "class
com.ysoft.jdo.attributes.AttributeCollection" and known subclasses...
CREATE TABLE ATTRIBUTECOLLECTION_ATTRIBUTEX (ATTRIBUTESX NUMBER, JDOIDX
NUMBER)
CREATE INDEX ATTRIBUTECOLLECTION_ATTRIBUTEX ON
ATTRIBUTECOLLECTION_ATTRIBUTEX (ATTRIBUTESX)
CREATE INDEX ATTRIBUTECOLLECTION_ATTRIBUTEX ON
ATTRIBUTECOLLECTION_ATTRIBUTEX (JDOIDX)
java.sql.SQLException: ORA-00955: name is already used by an existing
object
E:\MyCode>kodoschematool -action refresh
Refreshing schema for type "class
com.ysoft.jdo.attributes.AttributeCollection" and known subclasses...
CREATE TABLE ATTRIBUTECOLLECTION_SETX (SETX NUMBER, JDOIDX NUMBER)
CREATE INDEX ATTRIBUTECOLLECTION_SETXSETX ON ATTRIBUTECOLLECTION_SETX
(SETX)
CREATE INDEX ATTRIBUTECOLLECTION_SETXJDOIDX ON ATTRIBUTECOLLECTION_SETX
(JDOIDX)
CREATE TABLE ATTRIBUTECOLLECTIONX (JDOIDX NUMBER NOT NULL, JDOLOCKX
NUMBER, JDOCLASSX VARCHAR2(255), PRIMARY KEY (JDOIDX))
Refreshing schema for type "class com.ysoft.jdo.attributes.Attribute"
and known subclasses...
CREATE TABLE ATTRIBUTEX (JDOIDX NUMBER NOT NULL, JDOLOCKX NUMBER,
JDOCLASSX VARCHAR2(255), NAMEX VARCHAR2(255), VALUEX VARCHAR2(255),
PRIMARY KEY (JDOIDX))