Document Object Model

Tapestry 5 takes a very different approach to markup generation than Tapestry 4, or most other frameworks.

Tapestry 4 Approach

In Tapestry 4, markup generation was based on generating a character stream. At the lowest level, the fact that the output was in a markup format such as HTML, XHTML or WML was not known. Higher levels, such as the IMarkupWriter interface (and its implementations) provide the concept of markup generation: elements, attributes, start tags and end tags.

Often, components would work together to generate markup. A common example would be a Form component and the many form control components it contains (such as TextField, Checkbox, etc.). The Form could not fully render until all such enclosed components had rendered first.

IMarkupWriter supported nested writers. Nested writers are a way to buffer output until needed. A Form component would render its body using a nested writer, then write out its <form> and <input type="hidden"> elements, as well as the nested, buffered body.

This technique breaks down when two elements are peers, and not in a parent/child relationship. For example, the rendering of a FieldLabel component is affected by its companion TextField component. Handling these cases in Tapestry 4 required a number of kludges and special cases.

Tapestry 5 Approach

Tapestry 5 components render out a DOM, a Document Object Model. This is a tree of nodes representing elements, attributes and text within a document.

The DOM may ultimately be operated upon in a random access manner, rather than the serial (or buffered) approach used in Tapestry 4.

A new MarkupWriter interface allows the majority of code to treat the generation of output as a stream. In fact, MarkupWriter is more like a cursor into the DOM tree.

Once all rendering is complete, the DOM tree is streamed to the client.

DOM Classes

The implementation of this DOM is part of Tapestry, despite the fact that several third-party alternatives exist. This represents a desire to limit dependencies for the framework, but also the Tapestry DOM is streamlined for initial creation, and a limited amount of subsequent modification. Most DOM implemenations are more sophisticated, with greater support for querying (often using XPath) and manipulation.

Once the Document object is created, you don't directly create new DOM objects; instead, each DOM object includes methods that create new sub-objects. This primarily applies to the Element class, which can be a container of text, comments and other elements.

Document

The Document Object represents the an entire document, which is to say, an entire response to be sent to the client.

Documents will have a single root element. The newRootElement() method is used to create the root element for the document.

TODO: Support for doctypes, content type, processing instructions, and top-level comments.

Element

An Element of the document. Elements may have attributes, and they may themselves contain other elements, as well as text and comments.

The addAttribute() method adds a new attribute/value pair to the Element. If an existing attribute with the specified name already exists, then then the new value is ignored. This has implications when different pieces of code try to add attributes to an Element ... the first to add an attribute will "win"

Not yet implemented are some basic methods for manipulating the DOM after it is built. Plans are to add a few methods for re-parenting DOM nodes into new elements. In addition, some searching methods may be added.

MarkupWriter

The MarkupWriter interface allows the structure of the document to be built while maintaining a streaming metaphor.

element() and end()

Calls to element() create a new element within the tree, and may provide attributes for the new element as well. Calls to write(), writeln() and writef() write text nodes within the current element. Every call to element() should be matched with a call to end(), which is used to move the current node up one level.

  writer.element("img", "src", "icon.png", "width", 20, "height", 20, alt, "*");
  writer.end();

Note that end() must be called here, even though the <img> element is empty (has no body). If the call to end() is omitted, then later elements created by calls to element() will be nested inside the <img> element, which is not desired.

Again, every call to element() must be matched with a call to end():

  writer.element("select", "name", "choice");
  
  for (String name : optionsNames)
  {
    writer.element("option");
    writer.write(name);
    writer.end();
  }
  
  writer.end();

attributes()

Adds additional name/value pairs to the current element.

When a value is null, no attribute is added.

When a new name conflicts with an existing name, the new value is ignored. This gives precedence to the first value specified for an attribute over any subsequent value.

write()

The write() method writes text inside the current element. It scans the provided text for XML control characters ('<', '>', and '&') and converts them to their XML entity equivalents ('<', '>', and '&'). The result is correct, safe, HTML/XML output even when the content (which may come from a template, or from an external source such as a database) contains such problematic characters.

writef()

The writef() method formats an number of arguments. It uses a java.util.Formatter. It is a convience for formatting that ultimately invokes write().

writeRaw()

The writeRaw() method writes unfiltered text into the DOM. When the DOM is rendered to markup, the provided string is written to the output stream exactly as-is. Care should be taken, as this can easily result invalid markup, or even markup that is not well formed.

comment()

Adds an XML comment. The comment delimiters will be supplied by Tapestry:

  writer.comment("Start of JS Menu code");
  

DOM Manipulation

A powerful feature of Tapestry 5 is the ability to manipulate the DOM after it has been rendered.

Methods on Node allow an existing node to be moved relative to an Element. Nodes may be moved before or after the Element, or may be moved inside an Element at the top (the first child) or the bottom (the last child).

In addition, the children of an Element may be removed or a Node (and all of its children) removed entirely.

Finally, an Element may "pop": the Element is removed and replaced with its children.