Configuring Friendly URLs

Earlier versions of Tapestry have had a long-standing tradition of really ugly URLs . Because the framework generates the URLs and is also responsible for parsing and dispatching on them in later requests, it was not seen as an issue.

In fact, the ugly URLs do cause some problems:

  • Since all requests are routed through a single servlet (typically mapped to /app ), J2EE declarative security, which is path-based, is defeated.
  • Ugly URLs tend to be longer than friendly URLs, which can make a difference when creating a WML application.
  • A single directory may contain all the artifacts (HTML templates, specifications, properties files) for all the pages in an entire application. There isn't a sanctioned approach to organizing things into subdirectories.
  • The reliance on query parameters means that common search engines will only see a tiny fraction of the application.
Starting with 4.0, friendly URLs are integrated directly into framework (in 3.0 an ambitious, but more limited, patch was required). Warning: For security purposes, enabling friendly URLs implies that pages are no longer accessible via their ugly URL counterpart. This is not the case. If a malevolent user can either guess - or via cookies identify - your servlet path, they can construct an ugly URL to a resource that is protected via security and gain access to the protected resource. Friendly URLs are divided into two concerns:
  • Converting information normally stored as a query parameter into part of the URL path.
  • Parsing the path to restore the information previously encoded.
For example, the ugly URL /app?page=news/Thread&service=page may be converted into the friendly URL /news/Threads.html . In this case, the page=news/Thread query parameter became the news/Thread portion of the URL, and the service=page query parameter became the .html extension to the URL.

Understanding Tapestry URLs

To understand how to get friendly URLs, you must understand a little about what information Tapestry packs into URLs. Every request to a Tapestry application is mapped to an engine service . An engine service is something like a servlet, embedded within Tapestry. The service query parameter is used to select an engine service by name. A number of services are provided with the framework, the most common of which are:
page
Activates and renders a specific page.
direct
Used with the DirectLink and Form components.
home
Default service used when the service parameter is not specified (such as when first accessing the application); activates and renders the Home page.

Each service is responsible for creating URLs with the correct query parameters. By default, the URL path is always /app and any additional information comes out of the query parameters. The most common parameters are:

page
The name of a page to activate.
service
The service responsible for the request.
component
The nested component id of a component.
sp
Stores listener parameters passed in the URL (used by DirectLink and passed into listener method s, the "sp" is a holdover from 3.0).

Following this scheme, a typical URL might be /app?component=border.logout&page=news/Thread&service=direct . Yep, that's UGLY.

Enabling Friendly URLs

To use ordinary ugly URLs, Tapestry requires only a small amount of configuration in web.xml . Enabling friendly URLs requires adding more configuration to web.xml, and to your HiveMind module deployment descriptor .

Friendly URLs are controlled by ServiceEncoder s. Getting Tapestry to output friendly URLs is a matter of plugging encoders into the correct pipeline ... this is all done using HiveMind.

page-service-encoder

The most common type of encoder is the page-service-encoder , which encodes the page and service parameters. In your hivemodule.xml:

                        <contribution configuration-id="tapestry.url.ServiceEncoders">
                        <page-service-encoder id="page" extension="html" service="page"/>
                        </contribution>

This contribution to the tapestry.url.ServiceEncoders configuration point creates a ServiceEncoder that maps the .html extension (on the URL path) to the page service. The id attribute must be unique for all contributed encoders.

For Tapestry to recognize the URLs, you must inform the servlet container to send them to the Tapestry application servlet, by adding a mapping to web.xml:

                        <servlet-mapping>
                        <servlet-name>myapp</servlet-name>
                        <url-pattern>*.html</url-pattern>
                        </servlet-mapping>
Note:

This means that even static HTML pages that are part of your web application will be treated as Tapestry pages; any incoming request that ends with .html will be routed into the Tapestry application. Page specifications are optional, so Tapestry will treat the HTML pages are if they were HTML page templates. If you want to allow ordinary static content, then you should use another extension such as ".page" or ".tap" (the choice is arbitrary).

direct-service-encoder

A specialized encoder used exclusively with the direct service. Encodes the page name into the servlet path, then a comma, then the nested id for the component. One of two extensions is used, depending on whether the URL is stateful (an HttpSession existed when the link was rendered), or stateless.

A typical URL might be: /admin/Menu,border.link.direct . This indicates a page name of admin/Menu and a component id of border.link . By convention, the ".direct" extension is for stateless URLs.

The hivemodule.xml contribution:

                        <contribution configuration-id="tapestry.url.ServiceEncoders">
                        <direct-service-encoder id="direct" stateless-extension="direct" stateful-extension="sdirect"/>
                        </contribution>
                    

In addition, the *.direct and *.sdirect mappings must be added to web.xml:

                        <servlet-mapping>
                        <servlet-name>myapp</servlet-name>
                        <url-pattern>*.direct</url-pattern>
                        </servlet-mapping>

                        <servlet-mapping>
                        <servlet-name>myapp</servlet-name>
                        <url-pattern>*.sdirect</url-pattern>
                        </servlet-mapping>  

asset-encoder

The asset-encoder is for use with the asset service. The asset service exposes assets stored on the classpath (i.e., inside JARs) to the client web browser. The asset service receives a request with a resource path, and writes back a binary stream of that resources content.

In addition, each request includes a message digest , a string generated from the bytes of the the resource. This message digest acts as a credential , assuring that only classpath resources explicitly exposed by the application are accessible by the client (this prevents devious users from obtaining Java class files, for example). The message digest can only be computed by the server, using the full content of the actual file.

To enable friendly URLs for the asset service, add the following to your hivemodule.xml:

                        <contribution configuration-id="tapestry.url.ServiceEncoders">
                        <asset-encoder id="asset" path="/assets"/>
                        </contribution>

This contribution will encode asset URLs using the given path. The provided path, /assets comes first, then the digest string, then the path for the URL. An example URI would be /assets/91ab6d51232df0384663312f405babbe/org/apache/tapestry/contrib/palette/select_right.gif .

In addition you must add a mapping to web.xml:

                        <servlet-mapping>
                        <servlet-name>myapp</servlet-name>
                        <url-pattern>/assets/*</url-pattern>
                        </servlet-mapping>

If you choose a different folder than /assets/ then be sure to make corresponding changes in both hivemodule.xml and web.xml.

extension-encoder

The extension-encoder is used to encode just the service query parameter. The output URL is the service name with a fixed extension (typically, ".svc"), i.e., /home.svc or /restart.svc .

In your hivemodule.xml:

                        <contribution configuration-id="tapestry.url.ServiceEncoders">
                        <extension-encoder id="extension" extension="svc" after="*"/>
                        </contribution>

The use of the after attribute ensures that this encoder is always executed after any other encoders. Order is important!

For this example, another mapping is required in the web.xml:

                        <servlet-mapping>
                        <servlet-name>myapp</servlet-name>
                        <url-pattern>*.svc</url-pattern>
                        </servlet-mapping>

encoder

Finally, when one of the pre-defined encoders is insufficient, you can define your own. The <encoder> element allows an arbitrary object that implements the ServiceEncoder interface to be plugged into the pipeline. The <encoder> element supports the (required) id attribute, and the optional before and after attributes.

From the Virtual Library example, a custom encoder implementation is used as a special way to reference the ViewBook and ViewPerson pages using the external service (see the ExternalLink component for more information about using this engine service). The end result is that the URLs for these two pages look like /vlib/book/2096 rather than /vlib/ViewBook.external?sp=2096 or /vlib/app?page=ViewBook&service=external&sp=2096 . Certainly the first option is by far the prettiest.

These encoders are configured in hivemodule.xml as follows:

                        <encoder id="viewbook" before="external" object="instance:ViewPageEncoder,pageName=ViewBook,url=/book"/>
                        <encoder id="viewperson" before="external" object="instance:ViewPageEncoder,pageName=ViewPerson,url=/person"/>
                        <page-service-encoder id="external" extension="external" service="external"/>
                    

The order of the encoders in the pipline is very important, so the use of the before attribute ensures that the specialized encoders for these two pages are allowed to operate before the general purpose external service encoder.

The two special pages are mapped in web.xml using their custom URLs:

  <servlet-mapping>
                        <servlet-name>vlib</servlet-name>
                        <url-pattern>/book/*</url-pattern>
                        </servlet-mapping>
                        <servlet-mapping>
                        <servlet-name>vlib</servlet-name>
                        <url-pattern>/person/*</url-pattern>
                        </servlet-mapping>
                    

The implementation of the ViewPageEncoder class is all about an encode() method and a matching decode() method.

The encode() method must check to see if the link being generated is the right page name and the right service, returning (without doing anything) if not. The link being constructed is represented as an instance of ServiceEncoding :

                        public void encode(ServiceEncoding encoding)
                        {
                        if (!isExternalService(encoding))
                        return;

                        String pageName = encoding.getParameterValue(ServiceConstants.PAGE);

                        if (!pageName.equals(_pageName))
                        return;

                        StringBuilder builder = new StringBuilder(_url);

                        String[] params = encoding.getParameterValues(ServiceConstants.PARAMETER);

                        // params will not be null; in fact, pretty sure it will consist
                        // of just one element (an integer).

                        for (String param : params)
                        {
                        builder.append("/");
                        builder.append(param);
                        }

                        encoding.setServletPath(builder.toString());

                        encoding.setParameterValue(ServiceConstants.SERVICE, null);
                        encoding.setParameterValue(ServiceConstants.PAGE, null);
                        encoding.setParameterValue(ServiceConstants.PARAMETER, null);
                        }

                        private boolean isExternalService(ServiceEncoding encoding)
                        {
                        String service = encoding.getParameterValue(ServiceConstants.SERVICE);

                        return service.equals(Tapestry.EXTERNAL_SERVICE);
                        }  

We cheat just a bit here because we know that the service parameters will be a single numeric string. You can see exactly how encoder works, by building a new servlet path that encodes information that was stored as query parameters, the setting those query parameters to null

The flip side is the decode() method, which works by recognizing the URL generated by the encode() method and restoring the query parameters by parsing the URL:

  public void decode(ServiceEncoding encoding)
                        {
                        String servletPath = encoding.getServletPath();

                        if (!servletPath.equals(_url))
                        return;

                        String pathInfo = encoding.getPathInfo();

                        String[] params = TapestryUtils.split(pathInfo.substring(1), '/');

                        encoding.setParameterValue(ServiceConstants.SERVICE, Tapestry.EXTERNAL_SERVICE);
                        encoding.setParameterValue(ServiceConstants.PAGE, _pageName);
                        encoding.setParameterValues(ServiceConstants.PARAMETER, params);
                        } 

When constructing this style of encoder, it is important to remember that the servlet path does not end with a slash, but tthe path info, if non-null, will start with a slash.