Principles

What is Apache Tapestry?

Apache Tapestry is an open-source framework for creating dynamic, robust, highly scalable web applications in Java. Tapestry complements and builds upon the standard Java Servlet API, and so it works in any servlet container or application server.

Tapestry divides a web application into a set of pages, each constructed from components. This provides a consistent structure, allowing the Tapestry framework to assume responsibility for key concerns such as URL construction and dispatch, persistent state storage on the client or on the server, user input validation, localization/internationalization, and exception reporting. Developing Tapestry applications involves creating HTML templates using plain HTML, and combining the templates with small amounts of Java code. In Tapestry, you create your application in terms of objects, and the methods and properties of those objects – and specifically not in terms of URLs and query parameters. Tapestry brings true object oriented development to Java web applications.

Tapestry is specifically designed to make creating new components very easy, as this is a routine approach when building applications.

Tapestry is architected to scale from tiny, single-page applications all the way up to massive applications consisting of hundreds of individual pages, developed by large, diverse teams. Tapestry easily integrates with any kind of backend, including JEE, Spring and Hibernate.

Tapestry is designed to be extremely scalable in several dimensions:

• Tapestry applications may contain large numbers of pages and many custom components.
• Tapestry applications may contain very complex functionality.
• Tapestry applications may be created by large, diverse teams.
• Tapestry applications can service large numbers of concurrent users.

It's more than what you can do with Tapestry ... it's also how you do it! Tapestry is a vastly productive environment. Java developers love it because they can make Java code changes and see them immediately ... no redeploy, no restart! And it's blazingly fast to boot (even when files change). Designers love it because Tapestry templates are so close to ordinary HTML, without all the cruft and confusion seen in JavaServer Pages. Managers love it because it makes it easy for large teams to work together, and because they know important features (including localization) are baked right in. Once you work in Tapestry there's no going back!

Tapestry is released under the Apache Software Licence 2.0.

Principle 1 – Static Structure, Dynamic Behavior

The concept of "Dynamic Behavior" should be pretty obvious when you are building a web application; things should look different for different users/situations. But what does it mean that Tapestry has "Static Structure?" Static structure implies that when you build a page in Tapestry you are going to define all of the components that are used within that page. Under no circumstance during the rendering or event processing of the page will you be able to dynamically create a new component and place that into the component tree.

Why use static structure? Doesn't this hamper my ability to build software they way I want?

Second part first, static structure is not as limiting as you may think. You have plenty of options for getting dynamic behavior out of your static structure; from the simple conditional and looping components to the more advanced implementations of Tapestry's BeanEditor or Grid components. It's open source, peel back the covers and see how it works!

Why did Tapestry choose static structure as a core principle? It's really a matter of meeting the requirements of agility and scalability.

Agility
Tapestry is designed to be an agile working environment; "Code less, deliver more". To support you writing less code Tapestry does a lot of work on your POJO page/component before processing it's first request. For complex pages this work can eat up CPU cycles as well as resources. Using a Static Structure means that Tapestry can build this page model once and reuse it to handle each request, whether it's a link or a form post.

Scalability
When building large scale systems it's important to consider how your resources are going to be used on each deployed server, and how that information is going to be shared between servers. Static Structure means that page instances do not need to be stored inside the HttpSession and simple browsing users do not require extra system resources. This lean use of the HttpSession is key to Tapestry's very high scalability, especially in a clustered configuration.

Principle 2 – Adaptive API

A key feature of Tapestry 5 is adaptive API.

In traditional Java frameworks, including Tapestry 4, user code is expected to conform to the framework. You create classes that extend from framework-provided base classes, or implement framework-provided interfaces.

This works well until you upgrade to the next release of the framework: with the new features of the upgrade, you will more often than not experience breaks in backwards compatibility. Interfaces or base classes will have changed and your existing code will need to be changed to match.

In Tapestry 5, the framework adapts to your code. You have control over the names of the methods, the parameters they take, and the value that is returned. This is driven by annotations, which tell Tapestry under what circumstances your methods are to be invoked.

For example, you may have a login form and have a method that gets invoked when the form is submitted:

public class Login
{
@Persist
@Property
private String userId;

@Property
private String password;

@Component
private Form form;

@Inject
private LoginAuthenticator authenticator;

void onValidateForm()
{
if (! authenticator.isValidLogin(userId, password))
{
form.recordError("Invalid user name or password.");
}
}

Object onSuccess()
{
return PostLogin.class;
}
}

This short snippet demonstrates a bit about how Tapestry operates. Pages and services within the application are injected with the @Inject annotation. The method names, onValidateForm() and onSuccess(), inform Tapestry about when the method is to be invoked. The two events validateForm and success occur when a form is submitted; "validateForm" is triggered to perform cross-field validations, and "success" is only triggered when there are no validation errors. The onSuccess() method's return value directs Tapestry on what to do next: jump to another page within the application (here identified as the class for the page, but many other options exist). When there are exceptions, the page will be redisplayed to the user.

This also represents a distinct change from Tapestry 4. In earlier versions of Tapestry, the Form component's listener parameter would be bound to the method to invoke, by name. Further, the listener method had to be public. This new approach not only support multiple listeners, but provides an improved separation of view concerns (inside the page's HTML template) and logic concerns, inside the Java class.

In many cases, additional information about the event is available, and can be passed into the method by adding parameters to the method. Again, Tapestry will adapt to your parameters, in whatever order you supply them.

Tapestry also saves you effort: the @Property annotation marks a field as readable and writable; Tapestry will provide the accessor methods automatically.

Finally, Tapestry 5 explicitly separates actions (requests that change things) and rendering (requests that render pages) into two separate requests. Performing an action, such as clicking a link or submitting a form, results in a client side redirect to the new page. This is often called "redirect after post". This helps ensure that URLs in the browser are book-markable ... but also requires that a bit more information be stored in the session between requests (using the @Persist annotation).

Principle 3 – Public vs. Internal

An issue plaguing previous versions of Tapestry 4 (and earlier) was the lack of a clear delineator between private, internal APIs and public, external APIs. The fact that your code would extend from base objects but that many of the methods on those base objects were "off limits" further confused the issue. This has been identified as a key factor in the "steep learning curve of Tapestry" meme.

With the clean slate of Tapestry 5, we are being much more ruthless about internal vs. external.

First of all, anything inside the org.apache.tapestry5.internal package is internal. It is part of the implementation of Tapestry. It is the man behind the curtain. You should not ever need to directly use this code. It is a bad idea to do so, because internal code may change from one release to the next without concern for backwards compatibility.

Principle 4 – Backwards Compatibility

Tapestry has been plagued by backwards compatibility problems with every major release. Tapestry 5 does not even attempt to be backwards compatible to Tapestry 4. Instead, it lays the ground work for true backwards compatibility going forwards.

Tapestry 5's API is based almost entirely on naming conventions and annotations. Your components are just ordinary Java classes; you will annotate fields to allow Tapestry to maintain their state or to allow Tapestry to inject resources, and you will name (or annotate) methods to tell Tapestry under what circumstances a method should be invoked.

Tapestry will adapt to your classes. It will call your methods, passing in values via method parameters. Instead of the rigidness of a fixed interface to implement, Tapestry will simply adapt to your classes, using the hints provided by annotations and simple naming conventions.

Because of this, Tapestry will be able to change internally to a great degree without it affecting any of the application code you write. This should finally crack the backwards compatibility nut, allowing you to have great assurance that you can upgrade to future releases of Tapestry without breaking your existing applications.

This is already evident in Tapestry 5.1, where many new features and improvements have occurred, but is still 100% backwards compatible to Tapestry 5.0, as long as you've avoided the temptation to make use of internal APIs.