Filter transform is run, the only type of component which has not yet
been used *and* has the notion of a "format" is the writer. I (Felix)
cannot imagine a use case where components.Filter would filter, say,
parser-specific nodes. Thus, it's possible to make components.Filter
specific to writers without loss of functionality.
Added FormatSpecific node. Made "meta" and "raw" subclasses of
nodes.FormatSpecific.
nodes.Node.traverse: When a class is passed as first argument, do not
check if it is a subclass of nodes.Node. This is necessary because
some node classes (like nodes.FormatSpecific) aren't subclasses of
nodes.Node.
directives.meta: Do not insert pending nodes but rather the meta node
directly.
components.WriterFilter does no longer work on pending nodes but
traverses the document tree looking for FormatSpecific nodes.
Removed format-handling code for raw node from writers. This is now
done by the WriterFilter transform.
Removed test/test_transforms/test_filter.py; that test isn't usable
for testing the WriterFilter transform. components.WriterFilter is
automatically tested in the functional tests; that's enough for now.
Updated docs/ref/transforms.txt.
Partially updated docs/peps/pep-0258.txt.
git-svn-id: http://svn.code.sf.net/p/docutils/code/branches/transforms@3825 929543f6-e4f2-0310-98a6-ba3bd3dd1d04