You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 28 Next »

This page will contain the accumulated thoughts on what the wicket best practices are. It will also list situations that cause erroneus behaviours and are hard to track down.

Table of contents

Best Practices

Html Template Declaration

<!DOCTYPE html>
 <html
      xmlns="http://www.w3.org/1999/xhtml"
      xmlns:wicket="http://wicket.apache.org/"
      xml:lang="en"
      lang="en">

Wicket Filter Mapping

Wicket 1.4.x and upwards {}

On these versions, you should prefer to use WicketFilter instead of WicketServlet. If you want, it's possible to map /* instead of /app/*. But remember to configure the ignorePaths parameter to ignore static resources and leave them to be handled by the container.

Ignoring paths

If you want to leave static resources to be handled by the container, configure the ignorePaths init parameter as follows:

<filter>
	<filter-name>wicket.app</filter-name>
	<filter-class>org.apache.wicket.protocol.http.WicketFilter</filter-class>
	<init-param>
		<param-name>applicationClassName</param-name>
		<param-value>com.myproject.MyApplication</param-value>
	</init-param>
	<init-param>
		<param-name>filterMappingUrlPattern</param-name>
		<param-value>/*</param-value>
	</init-param>
	<init-param>
		<param-name>ignorePaths</param-name>
		<param-value>/css,/js,/images,/icons</param-value>
	</init-param>
</filter>

Otherwise, there's a risk the wicketservlet will handle the job to serve static resources, instead of the container.
For example,let's say you have /myapp/images/foo.gif and you map your wicket servlet to /* and dpn't have set ignorePaths.

Antipatterns

Anti-Pattern #1.

public FooPanel(String id,IModel<User> userMdl) {
        super(id,userMdl);
        final User user = getModelObject();
        IModel<List<User>> model = new LoadableDetachableModel<List<User>>() {
                private static final long serialVersionUID = 1L;

                @Override
                protected List<User> load() {
                        // A Stale User entity is now serialized with each page view
                        // This can consume lots of session space if the user entity is large.
                        return getWebSession().getUserService().getUserFriends(user);
                }
        };
}

Here is the correct way to do this
//IModel needs to be a LoadableDetachableModel!
public FooPanel(String id,IModel<User> userMdl) {
        super(id,userMdl);
        final User user = getModelObject();
        IModel<List<User>> model = new LoadableDetachableModel<List<User>>() {
                private static final long serialVersionUID = 1L;
                @Override
                protected List<User> load() {
                        User u = FooPanel.this.getModelObject();
                        return getWebSession().getUserService().getUserFriends(u);
                }
        };
}

The entity is fetched each page view, and NOT serialized with the page. The only thing that gets serialized is the fetching model.

Anti-Pattern #2.

public FooPanel(String id,IModel<User> userMdl) {
        super(id.userMdl);
        User user = getModelObject();
        // Doh! user is not serialized with each page view, the PropertyModel holds a reference!
        add(new Label("name",new PropertyModel(user,"screenName")));
}

Here is the correct way to do this
public FooPanel(String id,IModel<User> userMdl) {
        super(id.userMdl);
        add(new Label("name",new PropertyModel(userMdl,"screenName")));
}

The PropertyModel holds a reference to the Model which fetches the User on each page view. That way the User object is always fresh and is not serialized with the page.

Anti-Pattern #3.

public FooPanel(String id,User user) {
        // The stale User object will be in your session
        // for the life span of the stateful page.
        super(id.new Model(user));
}

Here is the correct way to do this, though not very elegant.
public FooPanel(String id,User user) {
        super(id);
        final int id = user.getId();
        setModel(new LoadableDetachableModel<List<User>>() {
                @Override
                protected List<User> load() {
                        return getUserDao().findById(id);
                }

antipattern 1-3 from blog letsgetdugg, thanks

Anti-pattern #4. - Anonymous Inner classes

Don't do this:

class MyPage extends WebPage {
    private MyVeryLargeObject subject;
    // ....
    public MyPage() {
        // get my very large object from database
        subject = MyVeryLargeObjectDao.get();
        // ...
        form.add(new TextField("name", new PropertyModel(MyPage.this, "some.very.large.navigational.structure.name"));
    }
}

The 'subject' instance of the MyVeryLargeObjectDao class will be serialized into the session with each page version. This can get out of hand, because with some business object models, the attached object can become very large. For this we introduced DetachableModels, which will retrieve the data from the database when needed, and will clean up when the data is not needed (at the end of the request for instance).

The other thing to avoid is anonymous or nested instances of IModel. Usually you share an instance of a model between two page instances. If you create an anonymous or nested instance of IModel, then you automatically get a 'this' reference to the class that surrounds it. This will usually be the page, but can also be the form or a listview. Anyway, because the reference is /final/, you will copy that reference to the old page, with the model to the new page, thus duplicating the component tree (it gets versioned etc.). This will eventually lead to OutOfMemoryError.

Search in the mailinglist for outofmemoryerror for other descriptions of this behaviour. I doubt that I have done the subject justice.

''Martijn Dashorst - Extracted from wicket-user mailing list

Debugging NotSerializableException

If you get a NotSerializableException on a deeply nested class like MyClass$2$1$1, it can be difficult to figure out which part of your code is causing the problem.

Solution: use javap, but be sure to escape the $'s: javap mypackage.MyClass\$2\$1\$1

Gotchas

Empty URL Parameter in CSS Style ( background-image: url(""); )

I ran into this problem where clicking on links in DataTable's generated an internal error in wicket and an error log indicating that the requested component was not found. Strangely, the case was only with Firefox and I had no problem with IE.
After debugging and digging everywhere, I found the cause was an empty image style attribute in my page causing firefox to re-generate a request to the page after the page has been shown, which would in turn re-render the component tree and invalidate the previous components:

    <td style="background-image: url(''); ">

This kind of problem is tricky to detect and even if it isn't causing any errors it would generate an extra load on servers.

--Iman Rahmatizadeh

BODY:onLoad in Panel not added without wicket:head

The onLoad attribute will reference some javascript and that must be added to a header. And because the Panel requires the onLoad AND the javascript it is natural to add the javascript to the Panels header. Thus you are able to create completely independent components which have all the information required to properly execute. No need to remember that you have to add a javascrpt reference to your Page. The Panel will handle it for you.

Taken from the mailing list. Question by Joshua Lim, answer by Juergen Donnerstag

Adding wicket:head to a Page

The <wicket:head> pair can be used in Panels, Borders and Fragments, and is used for doing a 'header contribution', meaning that the contents will be written to the <head> section of the page the component is placed in. If you are authoring a normal page, you don't need <wicket:head> tags but instead put it in the page's head section directly. The exception to this is when you use markup inheritance, and one of the pages to extend wants to contribute to the header of some extending page, and the header is not part of the <wicket:extend> region. In that case, you can use <wicket:head>.

Starting download after form submission (Wicket 1.1)

If you want to respond with a download to a form submission, you have to set the response page to null after writing your data into the response.

protected void onSubmit() {
   WebResponse r = (WebResponse)getRequestCycle().getResponse();
   r.setContentType("application/octet-stream");
   r.setHeader("Content-Disposition", "attachment; filename=\"data.abc\"");
   r.write("data");
   getRequestCycle().setResponsePage((WebPage)null);
}

See wicket.examples.displaytag.export.Export for a more complete example.

Starting download after form submission (Wicket 1.2)

The same as for 2.0, except that target.setFileName is from 1.2.3. You can override configure, cast to WebResponse, and call setAttachmentHeader on the response like this:

ResourceStreamRequestTarget target = new ResourceStreamRequestTarget(new FileResourceStream(f), d.getContentType())
{
	@Override
	protected void configure(Response arg0, IResourceStream arg1)
	{
		super.configure(arg0, arg1);
		WebResponse response = (WebResponse) arg0;
		response.setAttachmentHeader(filename);
	}
};

or do it like this:

 WebResource export = new WebResource() {
       @Override
       public IResourceStream getResourceStream() {
         CharSequence discounts = DataBase.getInstance()
             .exportDiscounts();
         return new StringResourceStream(discounts, "text/plain");
       }

       @Override
       protected void setHeaders(WebResponse response) {
         super.setHeaders(response);
         response.setAttachmentHeader("discounts.csv");
       }
     };
     export.setCacheable(false);

     new ResourceLink(this, "exportLink", export);

Starting download after form submission (Wicket 2.0)

This can be achieved using code like this (2.0):

Form form = new Form(this, "exportForm");
new Button<String>(form, "exportButton", new Model<String>(
    "export")) {

  @Override
  protected void onSubmit() {
    CharSequence export = DataBase.getInstance()
        .exportDiscounts();
    ResourceStreamRequestTarget target = new ResourceStreamRequestTarget(
        new StringResourceStream(export, "text/plain"));
    target.setFileName("discounts.csv");
    RequestCycle.get().setRequestTarget(target);
  }
};

PackageResources and relative paths

Assuming you have the following directory structure:

/com
/com/A.class
/com/sun/B.class
/com/sun/resource.txt

with the following Application.mountBookmarkablePage() rules:

mount("/A", A.class);
mount("/A/B", B.class);

Keep in mind that the path argument in the PackageResource methods ignores any Application mount() rules you might have applied. So it is legal to write:

PackageResource.get(com.sun.B.class, "resource.txt")

but it is not legal to write:

PackageResource.get(com.A.class, "B/resource.txt")

PBEWithMD5AndDES not found

[For the moment, I'll leave this, but I'm not convinced this actually belongs here at all - [Gwyn] 13:56, 23 Aug 2006 (BST)]

Not entirely clear what causes this.

INFO  - ClassCryptFactory          - using encryption/decryption object wicket.util.crypt.SunJceCrypt@17d26fc
ERROR - AbstractCrypt              - Unable to encrypt text ''
java.security.NoSuchAlgorithmException: Algorithm PBEWithMD5AndDES not available
	at javax.crypto.SunJCE_b.a(DashoA12275)
	at javax.crypto.SecretKeyFactory.getInstance(DashoA12275)
	at wicket.util.crypt.SunJceCrypt.generateSecretKey(SunJceCrypt.java:119)
	at wicket.util.crypt.SunJceCrypt.crypt(SunJceCrypt.java:95)
	at wicket.util.crypt.AbstractCrypt.encryptStringToByteArray(AbstractCrypt.java:204)
	at wicket.util.crypt.AbstractCrypt.encrypt(AbstractCrypt.java:107)
	at wicket.markup.html.form.PasswordTextField.getModelValue(PasswordTextField.java:97)
	at wicket.markup.html.form.FormComponent.getValue(FormComponent.java:387)
	at wicket.markup.html.form.TextField.onComponentTag(TextField.java:102)

A possible fix is to add another security provider like this:

[Comments on the mailing list suggest this has been seen twice only, and may be due to misconfigured/corrupted JDK - Suggest reinstalling the JDK before adding BouncyCastle... [Gwyn] 13:56, 23 Aug 2006 (BST)]

I've only seen this under the SysDeo Tomcat plug-in for Eclipse, when the boot classpath wasn't configured properly to include jce.jar, etc. – Al Maw

I ran into a similar problem with Red Hat Enterprise Linux because the default JDK is from IBM, and I had to replace wicket's SunJceCrypt and CachingSunJceCryptFactory with my own IBM versions. – Jim Hunziker

Adding id attribute to a tag

Do not add an explicit id to a tag (div, input, select,...); this will break the javascript code injected by Wicket (Ajax behavior,...). This is because the injected javascript uses the wicket generated id. – David Bernard

This is due to a hard to fix issue. As long as that issue is open, people should avoid setting id attributes on tags that are coupled to components that use rely on the markup id (setOutputMarkupId true). See http://issues.apache.org/jira/browse/WICKET-694 – Eelco

Avoid using empty elements (i.e. <textarea />)

Wicket will not automatically expand an empty element (i.e., an element that is expressed as a single tag ending in "/>") during the rendering phase. You should always include both an open tag and a close tag in your markup when you want Wicket to include content in the tag body.

Don't do this
<textarea wicket:id="mytext" />
Do this Instead
<textarea wicket:id="mytext">Content to be replaced by Wicket</textarea>

--Jay

  • No labels