No! You misunderstood me.

Declare the PK's, FK's etc in the database. Do not abandon good database practices for OJB -- particularly as never know what programs will access the DB directly in the future.

The DB managed auto-increment is the thing that is causing OJB to hiccup on you, I suspect. If you do an insert specifying the PK will DB2 override your specified one and replace it with its generated one, or will it accept your specified one?

A good description of the High-Low sequence algorithm is in the SDMagazine issue ( http://www.sdmagazine.com/articles/1999/9912/ ) but it is non free. You can google for other examples =) It is an efficient algorithm for managing sequences, basically.

If you have some values inserted by OJB's hi-low implementation, and some by the database using auto-generation you risk a collision. If you need the database auto-generation of keys (for batch inserts via a different tool, or other applications for example) having the database pull values from a database sequence that you also have OJB pull values from for OJB auto increment fields is the most reliable way to do this.

http://db.apache.org/ojb/api/org/apache/ojb/broker/util/sequence/ SequenceManagerNextValImpl.html

Is the base implementation for OJB that works (I am told, I don't use DB2) with DB2 sequences.

-Brian

On Jan 15, 2004, at 11:42 AM, Dirk Manske (Service Respond) wrote:

hm. our db2 database tables are configured with an autoincrement for the
primary keys by default. (for exampl: activityId INTEGER NOT NULL GENERATED
BY DEFAULT AS IDENTITY (STARTS WITH 1, INCREMENT BY 1, NO CACHE, PRIMARY
KEY). this is my first time using ojb. Meanwhile I get aware of what ojb
can all do for me. In the repository-user.xml I set every pk column with
primarykey="true" and autoincrement="true" for "declarative reasons",
because I thought OJB must get to know which columns in db2 are set to
pk's... Now I am wondering if the best practice would be to set up the
physical database with no pk's, fk's etc at all and declare everything in
the repository and let ojb control it?


Furthermore I already declared a high/low sequence manager in the
repository without actually knowing what I was doing (I copied it from
somewhere). So please could you tell me what happend behind the scenes? I
guess OJB generated an id and tried to store it with the object but then,
db2 overwrote the id via autoincrement. After the second store operation,
OJB looks for its generated id, cannot find it and stores it as a new
activity instead of updating it. hm. what would be the best advised? I
haven't asked yet but I am pretty sure that I am not allowed to set
autoincrement off in db2, because other apps might use these tables as
well...


-----Ursprüngliche Nachricht-----
Von: Brian McCallister [mailto:[EMAIL PROTECTED]
Gesendet: Donnerstag, 15. Januar 2004 16:44
An: OJB Users List
Betreff: Re: AW: AW: AW: Two 1:N mappings to one table

Ah hah! You are allowing DB2 to add the ID via a trigger, aren't you?
OJB doesn't play super nicely with the database doing that for it behind the
scenes, unfortunately. If you are allowed, try using one of the OJB auto
increment modules. I suggest the one that pulls from a db sequence if that
is what you are using in DB2 now. Basically it will have OJB pull from the
sequence and populate the id prior to the insert so that OJB knows the id.


-Brian

On Jan 15, 2004, at 10:25 AM, Dirk Manske (Service Respond) wrote:

Unfortunately I cannot provide a testcase for this. But in general,
after I stored the activity with pm.store(venture), the second
operation
pm.store(person) should automagically update activity and not save it
as a new entry, right? Or will it be only updated, if the activityId
is known?
Because I use DB2 with autoincrement for pk's and therefore on the
first store operation I have no id yet. hm. any idea?

Dirk

-----Ursprüngliche Nachricht-----
Von: Brian McCallister [mailto:[EMAIL PROTECTED]
Gesendet: Donnerstag, 15. Januar 2004 04:28
An: OJB Users List
Betreff: Re: AW: AW: Two 1:N mappings to one table

Hmm, fun one. I suspect you may have run into the weird edge case.

I don't have an immediate answer on this -- need to play and dig.

Anyone else hit this before? Any chance you can create a unit test
that demonstrates the behavior? If so send it to me and I'll check it
in and dig in (or start prodding).

-Brian

On Jan 14, 2004, at 1:21 PM, Dirk Manske (Service Respond) wrote:

Hi Brian,

exactly. Due to copy & paste I assigned a wrong column to the
attribute activityId in class Person. But know I experience the
problem, that each created activity is stored twice (with different
id's). I mean, I know why, I add activity once to the 'person
collection' and to the 'venture collection' and do then
pm.store(venture) and pm.store(person). I guess I have to follow a
different approach?! Say, store first venture, then retrieve activity
to store it with person? or how does it work?

thx,
Dirk

-----Ursprüngliche Nachricht-----
Von: Brian McCallister [mailto:[EMAIL PROTECTED]
Gesendet: Mittwoch, 14. Januar 2004 18:33
An: OJB Users List
Betreff: Re: AW: Two 1:N mappings to one table

It sounds like there is actually a problem in the mapping from class
-> table where the activityId is listed.

Any chance you could post the full mapping for these classes, along
with the database schema?

-Brian

On Jan 14, 2004, at 11:33 AM, Dirk Manske (Service Respond) wrote:

Hi,

yes, i am using pb.  Setting auto-retrieve="true" in the
collection-descriptor has not the desired effect, I still get the
following two strange error messages:

within browser it shows a null pointer execption for Venture when it
tried to retrieve all venture objects associated with activity and
in the tomcat console I get a SQL Exception for class PERSON, state
that activityId is not valid in the context where it is used (db2
database error message).

When I set auto-retrieve="false" for PERSON in the
reference-descriptor of class ACTIVITY, the venture objects are
retrieved. no error message.
everything works fine, except - of course I have no person objects
then.

any idea?

Dirk




-----Ursprüngliche Nachricht----- Von: Brian McCallister [mailto:[EMAIL PROTECTED] Gesendet: Mittwoch, 14. Januar 2004 15:13 An: OJB Users List Betreff: Re: Two 1:N mappings to one table

Mapping to 1:N's to the same table is fine, heack mapping 1:N's form
one class to itself multiple times works fine.

 From looking at your configs I will venture that you are using the
PB API and one of the collections isn't loading -- try setting
auto-retrieve on the collections, or (these are exclusive) try using
the ODMG API (which will retrieve them automagically)

-Brian

On Jan 13, 2004, at 7:58 PM, Dirk Manske wrote:

Hi,

how can I map two tables with each having a 1:N relation to the
same table?
I always get a null pointer exception for class Person.

I have three tables:

PERSON with a 1:N relation to ACTIVITY

and

VENTURE with a 1:N relation to ACTIVITY

Table PERSON and VENTURE both references activityId from table
ACTIVITY. In the repository-user.xml for each table which relates
to ACTIVITY the collection-descriptor is setup

for table VENTURE:

<collection-descriptor
name="allAssignedActivities"
element-class-ref="de.plato.Activity"
orderBy="activityId"
sort="DESC"
auto-update="true" >
<inverse-foreignkey field-ref="ventureId" />
</collection-descriptor>

and for table PERSON:

<collection-descriptor
name="allAssignedActivities"
element-class-ref="de.plato.Activity"
orderBy="activityId"
sort="DESC"
auto-update="true" >
<inverse-foreignkey field-ref="personId" />
</collection-descriptor>

for ACTIVITY, both tables are referenced:

<reference-descriptor name="person"> class-ref="de.plato.Person"
refresh="true"
auto-retrieve="true">
<foreign-key field-ref="personId" />

and

<reference-descriptor name="venture"> class-ref="de.plato.Venture"
refresh="true"
auto-retrieve="true">
<foreign-key field-ref="ventureId" />

Unfortunately, this does not work. What do I have to do?

thx,
Dirk


------------------------------------------------------------------- - - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]





-------------------------------------------------------------------- - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]


-------------------------------------------------------------------- - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]





--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]


--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]





--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]


--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]





--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]


--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]





--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to