Rendering of XML documents in XPCE


Based on graphicals:
--------------------

Map each element to an object.  Element attributes are mapped to object
methods.  There is a uniform mechanism for addressing the contents of
an element.

--> Easy to define new renderings and map and XML document onto an
    XPCE object structure and visa-versa.

--> Easy to handle general graphical objects

--> Difficult to edit plain text.  


Based on editor
---------------

The document is represented by an editor. Each element is mapped to a
fragment. Fragments can hold a graphical object (embedded graphical),
which can be converted from an XML document using the first method
described.

--> Requires many extensions to class editor and fragment:

	+ Well supported word-wrapping for long paragraphs.
	+ Well supported embedding of graphicals.
	+ Many more options to fragments and styles.
	+ Nested fragments might be needed for clarity and
	  performance handling long documents.  They are also
	  a nice representation of the XML content hierarchy!

--> Embedding graphicals:

	+ Represent using fragments holding 1 character.

--> Class text_image:

	+ Turn into a device with height/width equal to the rendered
	  text. Scrolling implies scrolling this device. Caching ensures
	  only visible parts are actually painted.

--> Fragments as `stand-alone' objects:

	+ We need some way to represent and element with its content
	  model as a single object.

	+ Allow fragments on strings?  Or use text-buffer nevertheless.
	  The overhead of text-buffers is very modest, as long as we
	  do not use the `undo buffer'.

Summarising:
------------

	+ text_buffer fragment list is a linear list of non-overlapping
	  fragments.
	
	+ Fragments can hold fragments.  Positions of sub-fragments are
	  relative to the embedded fragment.

	+ Graphicals are represented using a fragment holding a single
          character.  Such fragments should be `read-only'

	+ A text_image is a graphical device rendering an entire
	  document.  Scrolling is achieved using by moving this
	  object in a window (also realises horizontal scrolling).

Text-image caching:
-------------------

	+ Maintain a list holding character-start indices of each
	  physical line as well as the height and baseline of the
	  line.  Editing updates this list from the first change
	  to the end of the visible-area or end of the change
	  (whichever comes first).

	+ For the visible area, maintain a data-structure for each
	  line, specifying width of each character.


Placing graphicals:
-------------------

	+ If a graphical is encountered in the running text

		- If the current line is empty place it on the
	 	  current line

		- Else place it on the queue

		- Update physical line descriptions (start/end)