JAX-WS or JAXB working with the any type

Let's say you are working with a Schema that uses an "any" type like in the ws-security schema: (http://docs.oasis-open.org/wss/2004/01/oasis-200401-wss-wssecurity-secext-1.0.xsd):

<xsd:complextype name="SecurityHeaderType">
 <xsd:annotation>
  <xsd:documentation>This complexType defines header block to use for security-relevant data directed at a specific SOAP actor.</xsd:documentation>
 </xsd:annotation>
 <xsd:sequence>
          <xsd:any maxoccurs="unbounded" minoccurs="0" processcontents="lax">
   <xsd:annotation>
    <xsd:documentation>The use of "any" is to allow extensibility and different forms of security data.</xsd:documentation>
   </xsd:annotation>
  </xsd:any>
 </xsd:sequence>
 <xsd:anyattribute namespace="##other" processcontents="lax">
</xsd:anyattribute></xsd:complextype>

You generate the Java code using xjc or wsimport and then are wondering how to use the generated SecurityHeaderType object to put something like a username and password inside of it.

You may try to do something like this:

                // build the security header
  SecurityHeaderType headerWSSE = new SecurityHeaderType();
  
  // Create the userName and password that will go in the security header
  UsernameTokenType userNameToken = new UsernameTokenType();
  AttributedString userNameAttribute = new AttributedString();
  userNameAttribute.setValue("joebob");
  userNameToken.setUsername(userNameAttribute );
  
  PasswordString passwordString = new PasswordString();
  passwordString.setValue("secret");
  userNameToken.getAny().add(passwordString);
  headerWSSE.getAny().add(userNameToken);

(NOTE: UsernameTokenType and PasswordString types are both defined in the ws-security schema found here.)
If you run with this code, you will most likely get an error like this when it attempts to create the XML:

javax.xml.ws.WebServiceException: com.sun.istack.internal.XMLStreamException2: javax.xml.bind.MarshalException
 - with linked exception:
[com.sun.istack.internal.SAXException2: unable to marshal type "org.oasis_open.docs.wss._2004._01.oasis_200401_wss_wssecurity_secext_1_0.UsernameTokenType" as an element because it is missing an @XmlRootElement annotation]

The reason is that when xjc or wsimport generated the JAXB annotated classes it didn't create XmlRootElement annotations for UsernameTokenType or PasswordString.  So, it doesn't know how to handle them properly when they are added to the "any" collection.

So, one fix, is to manually add the XmlRootElement annotations to the UsernameTokenType and PasswordString classes.  However, I generally do not like the idea of manually modifying generated code.  It should be throw away to a certain degree.  That is if the schema changes and I need to re-generate I'd like to be able to just wipe the previously generated code and re-generate the new code.

Another option, is to use a plug-in like the annotate plugin to allow you to create a custom JAXB binding file that will add the XmlRootElement annotation as part of the code generation process.
You can find the annotate plugin here. http://confluence.highsource.org/display/J2B/Annotate+Plugin

The third option you have is to write your code in such a way that it wraps the UsernameTokenType and PasswordString objects in JAXBElement objects that can tell it how to create the root element names.  Here's the code for that:

                // build the security header
  SecurityHeaderType headerWSSE = new SecurityHeaderType();
  
  // Create the userName and password that will go in the security header
  UsernameTokenType userNameToken = new UsernameTokenType();
  AttributedString userNameAttribute = new AttributedString();
  userNameAttribute.setValue("joebob");
  userNameToken.setUsername(userNameAttribute );
  
  PasswordString passwordString = new PasswordString();
  passwordString.setValue("secret");
  
  // Since SecurityHeaderType takes "any" type, we must create JAXBElements to properly set the root elements... stupid "any" type!
  QName qnamePassword = new QName("http://docs.oasis-open.org/wss/2004/01/oasis-200401-wss-wssecurity-secext-1.0.xsd", "Password");
  JAXBElement jaxbPassword = new JAXBElement(qnamePassword ,PasswordString.class,passwordString);
  userNameToken.getAny().add(jaxbPassword);
  
  QName qname = new QName("http://docs.oasis-open.org/wss/2004/01/oasis-200401-wss-wssecurity-secext-1.0.xsd", "UsernameToken");
  JAXBElement jaxbUserName = new JAXBElement(qname ,UsernameTokenType.class,userNameToken);
   
  // Add our JAXBElements to the headers "any" list.
  headerWSSE.getAny().add(jaxbUserName);

This final approach requires you to create a QName instance to set the namespace and the root element name. Then you can create a JAXBElement that wraps your actual PasswordString and UsernameTokenType objects.  In my case, I wanted the PasswordString inside the UsernameToken so that's why you see that I added PasswordString into the usernameToken's any collection.

The code above produces the following XML:

<ns3:Security xmlns:ns3="http://docs.oasis-open.org/wss/2004/01/oasis-200401-wss-wssecurity-secext-1.0.xsd" >
 <ns3:UsernameToken>
  <ns3:Username>joebob</ns3:Username>
  <ns3:Password>secret</ns3:Password>
 </ns3:UsernameToken>
</ns3:Security>

Your namespace pre-fix may vary of course and I removed some of the extra namespace entries that were declared as part of the Security element since they weren't used in this node.

Hope this helps!  If you have any questions, feel free to post them below.

JAX-WS overriding Parameter name when generating code

When generating JAX-WS from code, you may at times need to override the default parameter names of the generated code.  Actually, sometimes it's required if you are having parameter name conflicts.  That happened to me recently.  I had a wsdl that defined an input message and output message that both had a header part like this:

<wsdl:message name="Ping">
<wsdl:part name="headerWSSE" element="wsse:Security"/>
<wsdl:part name="headerSYS" element="tnv:SystemInfo_Request"/>
<wsdl:part name="ping" element="tnv:ping"/>
</wsdl:message>
<wsdl:message name="PingResponse">
<wsdl:part name="headerSYS" element="tnv:SystemInfo_Response"/>
<wsdl:part name="pingResponse" element="tnv:pingResponse"/>
</wsdl:message>

Notice that both the Ping and PingResponse message define a part with the name="headerSYS".

The issue here is that when the code generator tries to create the ping() method in Java, it tries to create a method with 5 parameters... 3 for the input message parts and 2 for the output message parts. However, it would try to use the part name "headerSYS" as a parameter name for one of the input and one of the ouput parameters and it would bomb out because of two parameters having the same name in the same method signature.

Therefore, one fairly simple solution is to just create a custom binding file that tells ws-import how to handle these parameter names.  Here's what the custom binding file would look like:

<jaxws:bindings wsdlLocation="TelephoneNumberVerification_1_0_Logical.wsdl"
                 xmlns:xs="http://www.w3.org/2001/XMLSchema"
                 xmlns:wsdl="http://schemas.xmlsoap.org/wsdl/"
                 xmlns:jaxws="http://java.sun.com/xml/ns/jaxws"
                 xmlns:jaxb="http://java.sun.com/xml/ns/jaxb"
                 xmlns:tnv="http://barry.com/schemas/customerRelationship/Telephony">
       <jaxws:bindings node="wsdl:definitions/wsdl:portType[@name='TelephoneNumberVerification']/wsdl:operation[@name='ping']">
         <jaxws:parameter
                 part="wsdl:definitions/wsdl:message[@name='PingResponse']/wsdl:part[@name='headerSYS']"
                 childElementName="tnv:SystemInfo_Response" name="headerSysResponse"/>
     </jaxws:bindings>
     
 </jaxws:bindings>

I've attempted to hi-lite any thing that is custom for my wsdl that someone would have to change if they copied it and used it as a template.

By pointing ws-import to this custom binding, I was able to generate the code successfully and ended up with a method signature like this:

  public void ping(SecurityHeaderType headerWSSE, SystemInfoRequest headerSYS, boolean ping, Holder<SystemInfoResponse> headerSysResponse, Holder<Boolean> pingResponse)

Hope this helps!

Dozer Deep Nesting...A different approach. (No Setters, No Problem.)

Have you ever had the issue of trying to do a deep nested mapping using Dozer, but you keep running into issues because the destination object doesn't have a set method for the property you need to populate?

We see this a lot when dealing with generated JAX-WS code.  The JAXB classes generated from the schema typically do not create set methods for collections.  So, if you are trying to create a Dozer mapping to map from the JAX-WS/JAXB generated code to and from your application domain classes, you can run into some issues if you start trying to do deep mappings.

Here are the classes I want to populate via Dozer (made up for an example... all in package com.barry.to).  Starting with the first one and going down each class contains the class below it.  Usually in a List.

public class StartDestObject {
 
 private HaveNestedListNoSetter haveNestedListNoSetter;

 public HaveNestedListNoSetter getHaveNestedListNoSetter() {
  return haveNestedListNoSetter;
 }

 public void setHaveNestedListNoSetter(HaveNestedListNoSetter haveNestedListNoSetter) {
  this.haveNestedListNoSetter = haveNestedListNoSetter;
 }

}

public class HaveNestedListNoSetter {
 
 private List<HaveListNoSetter> haveListNoSetterList = new ArrayList<HaveListNoSetter>();

 public List<HaveListNoSetter> getHaveListNoSetterList() {
  return haveListNoSetterList;
 }
}

public class HaveListNoSetter {
 
 private List<Thing> things = new ArrayList<Thing>();

 public List<Thing> getThings() {
  return things;
 } 
}

public class Thing {

 private String name;
 private String value;
 public String getName() {
  return name;
 }
 public void setName(String name) {
  this.name = name;
 }
 public String getValue() {
  return value;
 }
 public void setValue(String value) {
  this.value = value;
 }
 
}

Notice that none of my classes that contain collections have any set methods for the List.

So, the idea is that I want to map a much simpler object into this object tree...  Here's my much simpler object structure:

package com.barry.from;

import java.util.List;

public class ItemHolder {
 
 private List<Item> items;

 public List<Item> getItems() {
  return items;
 }

 public void setItems(List<Item> items) {
  this.items = items;
 }
}
public class Item {
 
 private String itemName;
 private String itemValue;
 
 public Item getItem(){
  return this;
 }
 public String getItemName() {
  return itemName;
 }
 public void setItemName(String name) {
  this.itemName = name;
 }
 public String getItemValue() {
  return itemValue;
 }
 public void setItemValue(String value) {
  this.itemValue = value;
 }
}

Here's my test to validate it's working:

public class DozerTest {

 Mapper mapper;
 
 Item item = new Item();
 
 @Before
 public void setup(){
  
  List<String> files = new ArrayList<String>();
  files.add("dozer-bean-mappings.xml");
  mapper = new DozerBeanMapper(files);
  
  item = new Item();
  
  item.setItemName("gear");
  item.setItemValue("Value");
 }
 
 @Test
 public void testEvenMoreNestedMapping(){
  
  ItemHolder itemHolder = new ItemHolder();
  
  itemHolder.setItems(new ArrayList<Item>());
  
  itemHolder.getItems().add(item);
  
  StartDestObject startDestObject = mapper.map(itemHolder, StartDestObject.class);
  
  assertNotNull(startDestObject);
  
  assertEquals(item.getItemName(), startDestObject.getHaveNestedListNoSetter().getHaveListNoSetterList().get(0).getThings().get(0).getName());
  assertEquals(item.getItemValue(), startDestObject.getHaveNestedListNoSetter().getHaveListNoSetterList().get(0).getThings().get(0).getValue());
 }

I'm creating a single Item instance and then set that in the ItemHolder's list.  Then I want to map that Item instance into the Thing class that is deeply nested down at the bottom of my StartDestObject tree.

The trick to allow this deep mapping to work with nested collections that have no set methods is to Not do a deep mapping.  Instead, you break the mappings up into smaller chunks.  You create a mapping that goes from the ItemHolder's item to the StartDestObject's haveNestedListNoSetter property.  That gets us to the first step.  Then Dozer would try to figure out how to map from Item to HaveNestedListNoSetter class.  So, you create a mapping for this which actually just passes the Item (via "this") onto the first item in the HaveNestedListNoSetter's list.  Then you just keep this pattern up, utilizing the "this" to keep passing the Item instance along and then use the is-accessible="true" attribute to access the lists with no set method.

If you're not following my description, here's the dozer mapping to make it all work:

<?xml version="1.0" encoding="UTF-8"?>
<mappings xmlns="http://dozer.sourceforge.net"
          xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
          xsi:schemaLocation="http://dozer.sourceforge.net
          http://dozer.sourceforge.net/schema/beanmapping.xsd">

  <configuration>
    <stop-on-errors>true</stop-on-errors>
    <wildcard>true</wildcard>
  </configuration>

  <mapping>
    <class-a>com.barry.from.ItemHolder</class-a>
    <class-b>com.barry.to.StartDestObject</class-b>
       <field> 
        <a>items[0]</a>
        <b>haveNestedListNoSetter</b>
        <a-hint>com.barry.from.Item</a-hint> 
      </field>
  </mapping>

   <mapping>
    <class-a>com.barry.from.Item</class-a>
    <class-b>com.barry.to.HaveNestedListNoSetter</class-b>
       <field> 
        <a>this</a>
        <b is-accessible="true">haveListNoSetterList[0]</b>
        <a-hint>com.barry.from.Item</a-hint> 
        <b-hint>com.barry.to.HaveListNoSetter</b-hint>
      </field>
  </mapping>

  <mapping>
    <class-a>com.barry.from.Item</class-a>
    <class-b>com.barry.to.HaveListNoSetter</class-b>
       <field> 
        <a>this</a>
        <b is-accessible="true">things[0]</b>
      <b-hint>com.barry.to.Thing</b-hint>
      </field>
  </mapping>
 
  <mapping>
    <class-a>com.barry.from.Item</class-a>
    <class-b>com.barry.to.Thing</class-b>
    <field> 
        <a>itemName</a>
        <b>name</b>
      </field>
      <field> 
        <a>itemValue</a>
        <b>value</b>
      </field>
  </mapping>
                   
</mappings>

This may make for more mappings, but I don't know of a better solution without creating custom mapping classes or adding setters to the generated code (which sometimes you don't have access to).

If this isn't perfectly clear or you have a better way, please post in the comments.

Thanks!

Overriding WS-Addressing headers in JAX-WS

This is going to be quick and to the point, let's suppose you have a web service you have to call that requires WS-Addressing headers.  Usually, if the WSDL is created with the <wsaw:UsingAddressing wsdl:required="true"/> under the binding section, then the generated JAX-WS client will automatically add the WS-Addressing headers for you.

The problem I've had is that with the IBM Websphere runtime, you have no easy way to override the default parameters.  I've tried many different ways, but was only able to get one way to work (which I'll share in just a minute).  I originally thought it would be pretty simple... just create a handler that intercepts the message as it's going out and modify the appropriate SOAP Header element.  But the problem with this approach is that (at least in the IBM runtime) the WS-Addressing header information isn't created and added to the soap request until after it goes through the handler chains.  So, in a handler, you won't see any of the WS-Addressing headers because they haven't been added yet.  However, I did find that you if you add your own WS-Addressing header in a handler, the JAX-WS runtime won't try to create it.  So, I ended up creating a handler and just creating the WS-Addressing header I cared about overriding and set it there.

Here's the handler code:

 @Override  
      public boolean handleMessage(SOAPMessageContext smc) {  
           Boolean outBoundProperty = (Boolean) smc.get(MessageContext.MESSAGE_OUTBOUND_PROPERTY);  
           if (outBoundProperty.booleanValue()) {  
                SOAPFactory factory = null;  
                javax.xml.soap.SOAPElement element = null;  
                try {  
                     factory = SOAPFactory.newInstance();  
                     element = factory.createElement(new QName("http://www.w3.org/2005/08/addressing", "To"));  
                     element.setTextContent("IT");  
                     // check if there's a SOAP Header  
                     SOAPHeader header = smc.getMessage().getSOAPPart().getEnvelope().getHeader();  
                     if (header == null) {  
                          // create header since there isn't one  
                          header = smc.getMessage().getSOAPPart().getEnvelope().addHeader();  
                     }  
                     header.addChildElement(element);  
                } catch (SOAPException e) {  
                     // TODO: Do better error handling and logging  
                     e.printStackTrace();  
                }  
           }  
           return true;  
      }  

This handler creates a "To" element then adds it to the header.  It makes sure there's a header to add it to first.

This will produce a header like this:

   <soapenv:Header xmlns:wsa="http://www.w3.org/2005/08/addressing">
      <To xmlns="http://www.w3.org/2005/08/addressing">IT</To>
      <wsa:MessageID>urn:uuid:e180d2ff-7254-45eb-8bd8-f0058bb61ec7</wsa:MessageID>
      <wsa:Action>http://ws.cdyne.com/CheckTextBody</wsa:Action>
   </soapenv:Header>

With a handler like this you can easily override the parameters.  I'd be interested to hear of other ways.  Perhaps easier ways of doing the same.  Or let me know what you find out.  My experience is with Websphere 8.

Deploying older JAX-WS applications on WAS 7 or higher (Web Service not recognized)

At my company, we've recently been making a large move from Websphere Application Server(WAS) V6.1 to v8.0.  Some of our web service applications have noticed that if they were exposing JAX-WS service in their app and they simply deployed their app to WAS v8.0, their JAX-WS service was not recognized.  That is there are no errors, but the service is not exposed in anyway.  You can't reach it.

This turns out to be due to the way WAS does the scanning for the JAX-WS annotations.  This article from IBM helps explain it, but I'll give my abbreviated version as well...

Turns out that on WAS 6.1 with the WS Feature pack, WAS would scan your web app for JAX-WS annotations and if a Web Service was found it would expose it properly.  However, as of WAS 7, only Java EE 5 compliant modules are scanned.  That means if you created your app for WAS 6.1, then your web project would have been Java EE 1.4 compatible.  Check out wikipedia for a nice chart of WAS versions and their version support.

Java EE 1.4 means your web module will be version 2.4 where in Java EE 5, your web module would be 2.5.  So, if your web module (war) is not 2.5, then WAS 7 and beyond will not scan it for JAX-WS annotations and therefore your JAX-WS web services will never be exposed.

There are two fixes for this.

1) Upgrade your web module to 2.5 to be Java EE 5 compliant and then WAS 7 and higher will scan it for JAX-WS services.

2) Add a special IBM specific flag in your web applciations META-INF/MANIFEST.MF file like this:

Manifest-Version: 1.0
UseWSFEP61ScanPolicy: true

In the second option, this tell WAS to scan for the JAX-WS annotations even though its an older web module.


If you are using maven you may need to modify your POM file if maven is creating your MANIFEST.MF files.  This stackoverflow post helped us.

Here's what we ended up with:

<build>
<plugins>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-war-plugin</artifactId>
<configuration>
<warSourceDirectory>webapp</warSourceDirectory>
<webXml>webapp/WEB-INF/web.xml</webXml>
<archive>
<manifest>
<addClasspath>true</addClasspath>
</manifest>
<manifestEntries>
        <UseWSFEP61ScanPolicy>true</UseWSFEP61ScanPolicy>
      </manifestEntries>
</archive>
</configuration>
</plugin>
...



With this configuration we were able to have maven build our META-INF/MANIFEST.MF file in the war and add the special flag "UseWSFEP61ScanPolicy"

Hope this helps.

Building JAX-WS application with Maven

I was having issues building a JAX-WS application using Maven(NOTE: I'm using maven 2.0.11), when I generated my JAX-WS code using JAX-WS 2.2.  I saw errors like this:

cannot find symbol
symbol  : method required()
location: @interface javax.xml.bind.annotation.XmlElementRef

or

cannot find symbol
symbol  : constructor Service(java.net.URL,javax.xml.namespace.QName,javax.xml.ws.WebServiceFeature[])
location: class javax.xml.ws.Service

This blog really helped me find my solution.

I'll summarize briefly what the issue was.

Java comes with several specifications APIs packaged into it, like JAX-WS & JAXB. However, the version of these specifications are not always the latest.   That is, newer versions of JAX-WS and JAXB may be released, but are not in the version of Java you may be using to compile with.  Even if the newer version is available on your server runtime, they may not be available in your JDK.  And even if you try to explicitly try to add the newer versions of the JAX-WS and JAXB jars to your maven dependency section, it still will not work.

In order to override the built in versions of the JAX-WS and JAXB, you have to utilize the "endorsed" folder.   Follow this link to learn more about the "endorsed" folder.

In short, adding newer versions of the jars to an endorsed folder in your JDK allow them to override any embedded versions.

So, the big question then is how to do this with maven.  Here's the answer..

You use the maven-dependency-plugin to create the "endorsed" folder and copy the newer version of the jars into it.  Then you update the maven-compiler-plugin to refer to this endorsed folder.

Here's a sample POM.


<build>
<plugins>
<plugin>
<artifactId>maven-compiler-plugin</artifactId>
<version>2.3.2</version>
<configuration>
<source>1.6</source>
<target>1.6</target>
<compilerArguments>
<endorseddirs>${project.build.directory}/endorsed</endorseddirs>
</compilerArguments>
</configuration>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-dependency-plugin</artifactId>
<version>2.1</version>
<executions>
<execution>
<phase>validate</phase>
<goals>
<goal>copy</goal>
</goals>
<configuration>
<outputDirectory>${project.build.directory}/endorsed</outputDirectory>
<silent>true</silent>
<artifactItems>
<artifactItem>
<groupId>javax.xml.bind</groupId>
<artifactId>jaxb-api</artifactId>
<version>2.2</version>
<type>jar</type>
</artifactItem>
<artifactItem>
<groupId>javax.xml.ws</groupId>
<artifactId>jaxws-api</artifactId>
<version>2.2</version>
<type>jar</type>
</artifactItem>
</artifactItems>
</configuration>
</execution>
</executions>
</plugin>
</plugins>
</build>


With these settings we were able to successfully build our JAX-WS 2.2 code with Maven.

Hope this helps!

Changing Classloader policy in RAD for Websphere Applications

Changing Classloader policy in RAD for Websphere Applications

Sometimes the need arises to make sure the jars in your application get used over class definitions that may be part of the server runtime.  This is true when using Axis2 as your web service engine in Websphere.

To do this follow these instructions below:

In IBM’s RAD, go to your EAR project and open the application.xml file.   Be sure to do this in a JEE like perspective so that you get to use RAD’s special Application Deployment Descriptor UI.

 

At the bottom of the “Application Deployment Descriptor” UI, go to the Deployment tab. And change the “Classloader Mode” to Parent_Last.  I usually do this for both the EAR and the WAR.

This is all you have to do for local testing and development.  Now if you publish to the local WAS test environment, your application should first look for class definitions from the jars in your WEB-INF/lib before looking at the server runtime…(you  may need to restart the server)

Maven build changes.

When you follow the directions above to change the classloader policy, it will create some extra files under the META-INF folder under the ear project.... to get Maven to package these files with your app you have to add a resources section to your ear pom.

<modelVersion>4.0.0</modelVersion>

<artifactId>mobius-ear</artifactId>

<name>Mobius EAR</name>

<packaging>ear</packaging>

<build>

<resources>

<resource>

<directory>META-INF</directory>

<targetPath> ../${project.artifactId}-${project.version}/META-INF</targetPath>

<includes>

<include>**/*</include>

</includes>

</resource>

</resources>

<plugins>

<plugin>

<artifactId>maven-ear-plugin</artifactId>

<configuration>

<modules>

<webModule>

<groupId>${project.groupId}</groupId>

<artifactId>mobius-ws</artifactId>

<bundleFileName>mobius-ws.war</bundleFileName>

</webModule>

</modules>

</configuration>

</plugin>

</plugins>

</build>  

Isn’t there another way…

You can also change the classloader policy via the Websphere Admin console, but depending your Enterprise environments, you as a developer may not have access to Admin Console… so following the steps above, will automatically change the class loader policy when you application is deployed without requiring changing the classloader policy through the admin console.