Added documentation of the url attribute and how to use attributes with the entity syntax. Also clarified use of the fmt attribute.

finish_starttag():  Make sure we do not close too many levels; this was
    possible with some interactions between nested tables.

Merged in Jim's big fix from 2.3 branch dealing with returning the lexicon attributes of a Vocabulary object.

Test cases for TAL.HTMLParser module.

Add some helper functions for use in test cases.

Rename nsgmllib to HTMLParser; there really is not any SGML capability
beyond the HTML support at this point.  Rename all references.

New catalogs have 'lexicon' attributes that are lexicon names, but
old catalogs store their lexicons directly. This makes it hard
to manage text indexes correctly. Added some code to assure that
when a text index is added, we always pass the actual lexicon.

When converting BTrees, convert indexes in context. This is necessary
to help UnTextIndexes get their (previously) acquired lexicons.

UnTextIndexes now refer directly to their lexicons.

Added logic in conversion code to implement the rule above.

Took out some broken code to unscrew some hysterical breakage.

some old broken lexicon settings.

would trigger the error "replace structure with attribute replacements
not yet implemented".  We were right that it needed testing --
TALError was not imported.

Fixing this in TALInterpreter.py, and adding code to TALGenerator.py
that passes the "repldict" variable from emitStartElement() to
emitEndElement(), through the "todo" variable, when a "replace"
command is present.

hence were unable to cope with the newlines that may occur in HTML
attributes.  This caused sustitution of "str:foo\nbar" to insert
"foo", omitting "\nbar".

None (for auto), use a variable mode which can be "html", "xml" or
None.

METALError exception (and slight rewording of message).  (This was
Tim's idea.)

mixed with old.

Need to copy length value to __len__.

storage.  Some of the internal operations have changed, coding style
is a bit less compact.  Other substantive changes include:

- The implementation now uses the CommitLog.PacklessLog temporary
  transaction log class.  This avoids the problem with tempfiles being
  unlinked as soon as they are created (and then lost if the process
  exits abnormally).

- Minimal is now derived from BerkeleyBase instead of Base.

- Some updates to use the latest PyBSDDB API, e.g. get() instead of
  has_key().

- The table names have changed to 'serials' and 'pickles'.

- tpc_vote() is now supported.  BerkeleyBase.tpc_vote() calls
  Minimal._vote() and this is where the commit log is placed into the
  promised state.

- The semantics are now that when the transaction successfully
  commits, the commit log file is deleted.  This may need to be
  changed to either a truncation-on-close, or scribble-zero-in-header
  approach for performance reasons.

and Minimal Berkeley storages.  It derives from
ZODB.BaseStorage.BaseStorage and supports the tpc_vote() method
wrapper calling _vote() on the derived class.

CommitLog.py -- contains the temporary transaction commit log file
classes used to hold transaction changes until they can be committed
to Berkeley.

test_commitlog.py tests creation, deletion, state transitions,
low-level and high-level APIs of the CommitLog and derived classes.

test_create.py tests creation of the Berkeley files when a storage is
created.

test_virgin.py tests that a newly minted Berkeley storage is empty,
and tests the first commits to such a new storage.

Correct an error in the output relating to implicit closure of <dt>
elements.

Revert an accidental re-indenting of a line that occurred in revision 1.8
that introduced a bug in the implicit closing of list- and table-related
elements.

Move all the test input/output files around so the test/ directory is
available for Zope-style regression tests; these will appear later.

the position info yet, though.

for consistency with nsgmllib and expat.

break the use from TALParser.

Repaired test for ""at most one of content, replace, repeat" TAL error
and improved the msg.

Update to reflect the new support for SGML declaration syntax.

.handle_decl():  Push SGML declarations into the instruction stream.

Change RuntimeError to SGMLParseError, which carries line number and
offset information.

Add support for SGML declaration syntax (<!....>) to some reasonable
degree.  This does not support everything allowed in SGML, but should
work with "real" HTML (internal subset in a DOCTYPE is not handled).
The content of the declaration is passed to the .handle_decl() method,
which can be overridden by subclasses.