CHANGES.txt

ParsedXML changes

Bugfix/performance improvement. Do not rely on getPersistentDocument() but instead use acquisition parent. This fixes a memory leak triggered when doing document.documentElement, and also likely improves performance when accessing the DOM through the ManageableDOM wrappers.
The Zope Find tab should now not give an error anymore when searching with ids.
Use the One True Way to import expat now (from xml.parsers import expat).

All element nodes now have a _element_id id. This id is guaranteed to be unique in the document, though which id an element has may change in a reparse.
NodePath system for creating various paths to nodes. This can be based on child (child node index), element_id, or robust, which is not very robust as yet in some ways, but should be resistent to quite a few changes to the document.
added pretty printing feature. A pretty print button renders the document in pretty printed form, but does not save this changed version (you can do so yourself).

Removed ParsedXML's Expat, introduced dependency on PyXML's pyexpat instead (or just compile your own). This gets rid of lots of install hassles, especially on Windows. Just install PyXML.
An ongoing attempt to bring sanity to the unit test story.
Avoid XML-garbling bug in Printer by using PrettyPrinter.
various bugfixes in the DOM.

Tests are now more conformant with Zope unit testing guidelines.
Should work with Python 2.1/Zope 2.4, but not all the way there yet (parser segfault..)
new version of PyExpat

Fixed an ExpatBuilder bug that caused a DOM reference to be leaked when parsing occured at the document. There is still a leak for fragment parsing that we're looking into.

The value returned by get_size() is cached, which will often speed up the management view of an instance's container.

Problems with Attr Node manipulation not being reflected by the getAttribute methods of their Elements and vice versa fixed.
Erroneous position information for parse error output on subnodes fixed.
Default attributes are noticed by the parser and printer, and the relevent DOM methods work.

ManageableDOM Nodes can find the persistent Document wrapper when it has been installed in a Zope ObjectManager. This object, rather than a newly created ManageableDocument wrapper, is returned when available by OwnerDocument calls. This allows Zopish navigation and discovery out of the Document, helps shorten acquisition paths, and fixes some bugs with manipulation at the Document.
ManageableDOM's usage of namespaces for parsing is now optional and settable.
The DOM 2 Traversal interface has been fleshed out, although support for visiblity and roots is not complete.

Many bugs found and fixed as we hammered out new DOM tests, especially in namespace usage, attribute printing, and attribute children.