EJB CMP/CMR example with JBoss+Xdoclet

admin 14
0 0
Read Time:12 Minute, 18 Second

An abstract example

As said before, using Xdoclet, we merely have to focus on the implementation of our beans, and let Xdoclet generate all the necessary interface classes and deployment descriptors for us.

Consequently, the files involved in a (unidirectional) one-to-one CMR are Table1Bean.java andTable2Bean.java. In the concrete example section, we will also take a closer look at the generated jboss.xml, ejb-jar.xml, jbosscmp-jdbc.xml though.

(The ejb-jar.xml file contains elements that describe relationships and navigability. The jbosscmp-jdbc.xml contains the details of mapping to a database, hence are application server (container) specific.)

I’ll omit discussing the build.xml which is supposed to deploy your beans and the xdoclet.xml, which invokes Xdoclet to generate the necessary sources from the source+Xdoclet tags. (also known as ejb-generate.xml, see this post). In my case these are taken care of by Eclipse, but you’re supposed to be acquainted enough with these files for your own favourite IDE or command line interface, otherwise this long post would be even longer.

Let’s get down to business. Suppose Table1Bean wants to (unidirectionally one-to-one) refer to Table2Bean.

To make this possible

  • Table1Bean must have at least a primary key field table1ID with associated database column table1_id VARCHAR (of course, any simple (not compound in CMR fields!!) type primary key definition will do), and
  • Table2Bean must at least have a foreign key field table1IDfk with associated database column table1_id_fk VARCHAR.

I’ll assume you can set-up the entity beans without relations (if not, I’ll recommed this tutorial once more).

Now you have to add Xdoclet tags for establishing the (unidirectional one-to-one) relation for

  • the bean deployment descriptor, ejb-jar.xml, using the @ejb.relation tag and
  • the (JBoss) application server, using the and @jboss.target-relation tag

You have to add these tags to the comment section of the definition of the getter of your CMR field, so in Table1Bean.java

    /**
     * ...omitted normal method comments with @return stuff etc...
     *
     * @ejb.interface-method
     *
     * @ejb.relation
     *    name="table1-refers-to-table2"
     *    role-name="Table1ToTable2"
     *    cascade-delete="yes"
     *    target-cascade-delete="yes"
     *    target-role-name="Table2ToTable1"
     *    target-ejb="Table2"
     *
     * @jboss.target-relation
     *    related-pk-field="table1ID"
     *    fk-column="table1_id_fk"
     */
    public abstract ArticleLocal getArticle();

It is not necessary to use the @ejb.ejb-ref class scope tags, you use these in e.g. your facade stateless session beans.

Now look very good what’s happening here. It took me a lot of time to figure out! Since we have unidirectionality here (the “blind side” in the EJB Xdoclet tag documentation is the side that is referred to in the unidirectional case), we have to specify the target side with the target-rolename field. Not complying with this rule will result in deployment errors. The target-ejb is compulsory (of course).

The @jboss.target-relation has a field designating the primary key field of the originating bean (related-pk-field) and a field designating the foreign-key database column of the referred to bean, Table2Bean (fk-column). Finally the cascade-delete means that the associated bean is deleted, once the main bean is. This ensures referential integrity.

In our concrete example we’ll look at the resulting ejb-jar.xml, jboss.xml and jbosscmp-jdbc.xml.

Now upon creation of a Table1Bean item (consisting of fields in Table1bean and Table2Bean), we must

  1. Create an instance of Table1Bean
  2. Create an instance of Table2Bean
  3. Make a link/reference from Table1Bean to Table2Bean using the CMR field.

According to the EJB 2.x spec, the CMR field can only be set in the ejbPostCreate() method! This method must have the same signature as the ejbCreate() method, of course.

Using the CMR field associated get-method, you are able to retrieve a reference to Table2Bean. Looking ahead to one-to-many relation ships, we can already see that this get-method will return a Collection, instead of a reference to a single bean.

With these ingredients, you should be able to finish your application.

Finally, I want to stress once more that besides the JBoss tags in Xdoclet, there are corresponding tags for all kinds of application servers, such as JOnAS, Bea’s Weblogic and many more.

Happy
Happy
0 %
Sad
Sad
0 %
Excited
Excited
0 %
Sleepy
Sleepy
0 %
Angry
Angry
0 %
Surprise
Surprise
0 %

Average Rating

5 Star
0%
4 Star
0%
3 Star
0%
2 Star
0%
1 Star
0%

14 thoughts on “EJB CMP/CMR example with JBoss+Xdoclet

  1. Did it and worked. In fact the CMR element -Container Managed Relations- is an obvious missing point. Do you know where I can get some good sources for CMR?
    Thanks!

  2. Hi everyone:

    I would like to know if there’s possible to generate querys in the .java’s bean with xdoclect, something like Jboss IDE or hibernate. I must define relationships and create new querys and can’t do enything of that with my Lomboz Eclipse plugin and Jboss server.

    I would be greateful at any kind of help

  3. Regarding Comment 4, that’s because it’s after the “@jboss.target-relation” tag, which refers to the blind side of the relation. In the case which you mentioned, one would use the “@jboss.relation” tag, which makes it work from the other side.

    It took me very long to figure this out…somehow most tutorials on the net use that kind of example with the CMR Relation on the blind side, which i personally find not very intuitive for a first example. When trying it with the tags on the side with the foreign key column this produces “unknown column in where query” exceptions.

  4. I did it and it works correctly (after some hacks :o)).
    Now i want to make tests around it. need i use cactus? is there another solution?
    I dont want to kludge without another ejb. 🙁
    Thank you
    Breno Leitão.
    http://lcr.icmc.usp.br

  5. I got it!!!!
    “ss� and me have the following problem:
    XDoclet generates an empty {field-name} and {column -name}.
    Here is the reason for that behaviour:
    We got the source via copy and paste from this website. This leads us to the following line:

    related-pk-field="table1ID�

    The closing quotation mark is non ASCII! This confuses XDoclet. Just replace it with the correct one (“) and everything works properly.

    Yours
    Arne

  6. This example is part of quite a large project in Eclipse, hence it is not possible to compose such a JAR instantly. However, I think it is a very good idea. Moreover, the statistics show that many people are struggling with these issues and are consulting my post.

    Although I cannot promise to deliver the JAR file tomorrow, I’ll do my best to make it available soon. If so, I’ll make a new post, instead of changing this one. Of course, I’ll leave a comment referring to the new post.

    If anyone else already has such a JAR file (preferrably based on my example) and is willing to share his/her efforts with the rest of the world, please let me know.

    Regards,

  7. It would be *very* nice if you can zip a complete working example (inclusive build.xml) and give it for download. Please.
    Yours
    Arne
    Btw.: Your examplanations are very understandable.

  8. I got the same problems as “ss” has.
    It would be *very* nice if you can zip a complete working example (inclusive build.xml) and give it for download. >> Please < <
    Yours
    Arne
    Btw.: Your examplanations are very understandable.

  9. From your comment I cannot see what you find confusing. I did not post the database schema, since the intention was to let the container create the schema for me. But now that the schema has been created, I can post it, of course!

    An article is created in the Hypersonic database with this SQL statement:

    CREATE TABLE ARTICLE(PUBLICATION_ID_FK VARCHAR(256) NOT NULL,TITLE VARCHAR(256),JOURNAL VARCHAR(256),YEAR INTEGER,VOLUME VARCHAR(256),NUMBER INTEGER,PAGES VARCHAR(256),MONTH VARCHAR(256),NOTE VARCHAR(256),CONSTRAINT PK_ARTICLE PRIMARY KEY(PUBLICATION_ID_FK))
    

    The PUBLICATION_ID_FK is primary key and foreign key at the same time for the article table. This is sound, since we are dealing with a one-to-one relationship here.

    A bibliography is created with:

    CREATE TABLE BIBLIOGRAPHY(PUBLICATION_ID VARCHAR(256) NOT NULL,PUBLICATION_KEY VARCHAR(256),PUBLICATION_TYPE VARCHAR(256),ANNOTATE VARCHAR(256),CROSSREF VARCHAR(256),ABSTRACT VARCHAR(256),AFFILIATION VARCHAR(256),CONTENTS VARCHAR(256),COPYRIGHT VARCHAR(256),ISBN_ISSN VARCHAR(256),KEYWORDS VARCHAR(256),LANGUAGE VARCHAR(256),LCCN VARCHAR(256),LOCATION VARCHAR(256),SIZE INTEGER,MRNUMBER INTEGER,PRICE DECIMAL,URL VARCHAR(256),CATEGORY VARCHAR(256),CONSTRAINT PK_BIBLIOGRAPHY PRIMARY KEY(PUBLICATION_ID))
    

    I hope this helped!

  10. Looks like you may have things the wrong way around? Isn’t Table 2 actually referring to Table 1? Rather confusing. Also I find the Concrete example even more confusing! It would be better accompanied by a database schema to refer to.

  11. Thanks a lot for the tutorial. however,

    i used the following in my ant build script, but i had problems generating the <key -fields/> in jbosscmp-jdbc.xml.
    And hence i had problems deploying my bean. It threw the following exception.

    org.jboss.deployment.DeploymentException: Atleast one role of a foreign-key mapped relationship must have key fields (or <primkey -field> is missing from ejb-jar.xml): ejb-relation-name=FolderMayContainDevices

    Can you please help.

      <relationships>
        <ejb -relation>
          <ejb -relation-name>FolderMayContainDevices
          <relation -table-mapping>
          </relation>
    
          <ejb -relationship-role>
              <ejb -relationship-role-name>FolderToDevices
          <ejb -relationship-role><ejb -relationship-role-name>DeviceToFolder</ejb>
              <key -fields>
                 <key -field>
                   <field -name>
                   <column -name>
                 </key>
              </key>
    
          </ejb>
        </ejb>
        <ejb -relation>
          <ejb -relation-name>FolderMayContainFoldersFolderToSubFolders
    		  <key -fields/>
    
          </key>
          <ejb -relationship-role><ejb -relationship-role-name>FolderToParentFolder</ejb>
              <key -fields>
                 <key -field>
                   <field -name></field>
                   <column -name></column>
                 </key>
              </key>
    
          </ejb>
    
         &lt!--
           To add jboss relationships for beans not managed by XDoclet, add
           a file to your XDoclet merge directory called jbosscmp-jdbc-relationships.xml that contains
           the  markups for those beans.
         -->
      </>
    

    the build.xml was as follows:

          <taskdef name = "ejbdoclettask"  classname = "xdoclet.modules.ejb.EjbDocletTask"  classpathref="xdoclet.class.path"/>ejbdoclettask destdir="${doclet.gensrc.dir}"  mergedir="parent-fake-to-debug"
                excludedtags="@version,@author,@todo"  addedtags = "@xdoclet-generated at ${TODAY},@copyright The XDoclet T
    eam, @author XDoclet,@version ${version}"   ejbspec="2.0" force="false" verbose="false">
                            <fileset dir = "${source.dir}">
                                    <include name = "**/*.java"/>
                            </include>
    
                            < !-- Comment the next two lines, if you do not want to e
    xpose the Beans via the Remote Interface -->      
  12. It was the first tutorial I found, which really helped me, to understand, how to make cmr with Xdoclet.
    Thanks for it.
    Janos

Comments are closed.

Next Post

Summary "Expert one-on-one J2EE design and development"

Chapter 1: J2EE architectures One should apply J2EE to realize OO design, not let J2EE technologies dictate object design. Goals of an Enterprise Architecture Be robust Be performant and scalable Take advantage of OO design principles Avoid unnecessary complexity Be maintainable and extensible Be delivered on time Be easy to […]
%d bloggers like this: