2007/04/??
        RELEASE 2.4.0

        Improved Literal comparison / equality

        Sparql cleanup.

        getLiteralValue now returns the Literal object instead of the
        result of toPython().  Now that Literals override a good
        coverage of comparison operators, they should be passed around
        as first class objects in the SPARQL evaluation engine.

        Added support for session bnodes re: sparql

        Fixed prolog reduce/reduce conflict.  Added Py_None IncRefs
        where they were being passed into Python method invokations
        (per drewp's patch)

        Fixed sparql queries involving empty namespace prefix.

        Fixed the selected variables sparql issue

        Fixed <BASE> support in SPARQL queries.

        Fixed involving multiple unions and queries are nested more
        than one level (bug in _getAllVariables causing failure when
        parent.top is None)

        Fixed test_sparql_equals.py.

        Fixed sparql json result comma errors issue.

        Fixed test_sparql_json_results.py (SELECT * variables out of
        order)

        Added a 4Suite-based SPARQL XML Writer implementation.  If
        4Suite is not installed, the fallback python saxutils is used
        instead

        applied patch from
        http://rdflib.net/issues/2007/02/23/bugs_in_rdflib.sparql.queryresult/issue

        The restriction on GRAPH patterns with variables has been
        relieved a bit to allow such usage when the variable is
        provided as an initial binding

        Fix for OPTIONAL patterns.  P1 OPT P2, where P1 and P2 shared
        variables which were bound to BNodes were not unifying on
        these BNode variable efficiently / correctly.  The fix was to
        add bindings for 'stored' BNodes so they aren't confused for
        wildcards




        Added support to n3 parser for retaining namespace bindings.

        Fixed several RDFaParser bugs.

        Added serializer specific argument support.

        Fixed a few PrettyXMLSerializer issues and added a max_depth
        option.

        Fixed some TurtleSerializer issues.

        Fixed some N3Serializer issues.



        Added support easy_install

        added link to long_descriptin for easy_install -U rdflib==dev
        to work; added download_url back

        added continuous-releases-using-subversion bit



        Added rdflib_tools package
          Added rdfpipe
          Added initial EARLPluging



        Improved test running... using nose... added tests

        Exposed generated test cases for nose to find.
        added bit to configure 'setup.py nosetests' to run doc tests

        added nose test bits



        Added md5_term_hash method to terms.

        Added commit_pending_transaction argument to Graph's close
        method.

        Added DeprecationWarning to rdflib.constants

        Added a NamespaceDict class for those who want to avoid the
        Namespace as subclass of URIRef issues

        Added bind function

        Fixed type of Namespace re: URIRef vs. unicode

        Improved ValueError message

        Changed value method's any argument to default to True

        Changed __repr__ to always reflect that it's an rdf.Literal --
        as this is the case even though we now have it acting like the
        corresponding type in some casses

        A DISTINCT was added to the SELECT clause to ensure duplicate
        triples are not returned (an RDF graph is a set of triples) -
        which can happen for certain join expressions.

        Support for ConditionalAndExpressionList and
        RelationalExpressionList (|| and && operators in FILTER)

        Fixed context column comparison.  The hash integer was being
        compared with 'F' causing a warning:Warning: Truncated
        incorrect DOUBLE value: 'F'

        applied patch in
        http://rdflib.net/issues/2006/12/13/typos_in_abstractsqlstore.py/issue

        fix for
        http://rdflib.net/issues/2006/12/07/problems_with_graph.seq()_when_sequences_contain_more_than_9_items./issue





        General code cleanup (removing redundant imports, changing
        relative imports to absolute imports etc)

        Removed usage of deprecated bits.

        Added a number of test cases.

        Added DeprecationWarning for save method

        refactoring of GraphPattern

        ReadOnlyGraphAggregate uses Graph constructor properly to
        setup (optionally) a common store


        Fixed bug with . (fullstop) in localname parts.

        Changed Graph's value method to return None instead of raising
        an AssertionError.

        Fixed conversion of (exiplicit) MySQL ports to integers.

        Fixed MySQL store so it properly calculates __len__ of
        individual Graphs

        Aligned with how Sleepycat is generating events (remove events
        are expressed in terms of interned strings)

        Added code to catch unpickling related exceptions

        Added BerkeleyDB store implementation.

        Merged TextIndex from michel-events branch.


2006/10/15
        RELEASE 2.3.3

        Added TriXParser, N3Serializer and TurtleSerializer.

        Added events to store interface: StoreCreated, TripleAdded and
        TripleRemoved.

        Added Journal Reader and Writer.

        Removed Sleepycat level journaling.

        Added support for triple quoted Literal's.

        Fixed some corner cases with Literal comparison.

        Fixed PatternResolution for patterns that return contexts only.

        Fixed NodePickler not to choke on unhashable objects.

        Fixed Namespace's __getattr__ hack to ignore names starting
        with __

        Added SPARQL != operator.

        Fixed query result __len__ (more efficient).

        Fixed and improved RDFa parser.

        redland patches from
        http://rdflib.net/pipermail/dev/2006-September/000069.html

        various patches for the testsuite -
        http://rdflib.net/pipermail/dev/2006-September/000069.html


2006/08/01
        RELEASE 2.3.2

        Added SPARQL query support.

        Added XSD to/from Python datatype support to Literals.

        Fixed ConjunctiveGraph so that it is a proper subclass of Graph. 

        Added Deprecation Warning when BackwardCompatGraph gets used.

        Added RDFa parser.

        Added Collection Class for working with RDF Collections.

        Added method to Graph for testing connectedness

        Fixed bug in N3 parser where identical BNodes were not being combined.

        Fixed literal quoting in N3 serializer.

        Fixed RDF/XML serializer to skip over N3 bits.

        Changed Literal and URIRef instanciation to catch
        UnicodeDecodeErrors - which were being thrown when the default
        decoding method (ascii) was hitting certain characters.

        Changed Graph's bind method to also override the binding in
        the case of an existing generated bindings.
        
        Added FOPLRelationalModel - a set of utility classes that
        implement a minimal Relational Model of FOPL implemented as a
        SQL database (uses identifier/value interning and integer
        half-md5-hashes for space and index efficiency). 

        Changed MySQL store to use FOPLRelationalModel plus fixes and
        improvements.

        Added more test cases.

        Cleaned up source code to follow pep8 / pep257.


2006/02/27
        RELEASE 2.3.1

        Added save method to BackwardCompatibleGraph so that
        example.py etc work again.

        Applied patch from Drew Perttula to add local_time_zone
        argument to util's date_time method.

        Fixed a relativize bug in the rdf/xml serializer.

        Fixed NameError: global name 'URIRef' is not defined error in
        Sleepycat.py by adding missing import.

        Applied patch for Seq to sort list by integer, added by Drew
        Hess.

        Added a preserve_bnode_ids option to rdf/xml parser.

        Applied assorted patches for tests (see
        http://tracker.asemantics.com/rdflib/ticket/8 )

        Applied redland.diff (see
        http://tracker.asemantics.com/rdflib/ticket/9 )

        Applied changes specified
        http://tracker.asemantics.com/rdflib/ticket/7

        Added a set method to Graph.

        Fixed RDF/XML serializer so that it does not choke on n3 bits
        (rather it'll just ignore them)


2005/12/23
        RELEASE 2.3.0

        See http://rdflib.net/2.3.0/ for most up-to-date release notes

        Added N3 support to Graph and Store.

        Added Sean's n3p parser, and ntriples parser.

        Sleepycat implementation has been revamped in the process of
        expanding it to support the new requirements n3
        requirements. It also now persists a journal -- more to come.

        detabified source files.

        Literal and parsers now distinguish between datatype of None and datatype of "".

        Store-agnostic 'fallback' implementation of REGEX matching
        (inefficient but provides the capability to stores that don't
        support it natively). Implemented as a 'wrapper' around any
        Store which replaces REGEX terms with None (before dispatching
        to the store) and whittles out results that don't match the
        given REGEX term expression(s).

        Store-agnostic 'fallback' implementation of transactional
        rollbacks (also inefficient but provides the capablity to
        stores that don't support it natively). Implemented as a
        wrapper that tracks a 'thread-safe' list of reversal
        operations (for every add, track the remove call that reverts
        the store, and vice versa). Upon store.rollback(), execute the
        reverse operations. However, this doesn't guarantee
        durability, since if the system fails before the rollbacks are
        all executed, the store will remain in an invalid state, but
        it provides Atomicity in the best case scenario.


2005/10/10
        RELEASE 2.2.3

        Fixed Sleepycat backend to commit after an add and
        remove. This should help just a bit with those unclean
        shutdowns ;)

        Fixed use of logging so that it does not mess with the root
        logger. Thank you, Arve, for pointing this one out.

        Fixed Graph's value method to have default for subject in
        addition to predicate and object.

        Fixed Fourthought backend to be consistent with interface. It
        now supports an empty constructor and an open method that
        takes a configuration string.


2005/09/10
        RELEASE 2.2.2

        Applied patch from inkel to add encoding argument to all
        serialization related methods.

        Fixed XMLSerializer bug regarding default namespace bindings.

        Fixed namespace binding bug involving binding a second default
        namespace.

        Applied patch from Gunnar AAstrand Grimnes to add context
        support to __iadd__ on Graph. (Am considering the lack of
        context support a bug. Any users currently using __iadd__, let
        me know if this breaks any of your code.)

        Added Fourthought backend contributed by Chimezie Ogbuji.

	Fixed a RDF/XML parser bug relating to XMLLiteral and
	escaping.

        Fixed setup.py so that install does not try to uninstall
        (rename_old) before installing; there's now an uninstall
        command if one needs to uninstall.


2005/08/25
        RELEASE 2.2.1

        Fixed issue regarding Python2.3 compatibility.

        Fixed minor issue with URIRef's absolute method.


2005/08/12
        RELEASE 2.1.4

        Added optional base argument to URIRef.

        Fixed bug where load and parse had inconsistent behavior.

        Added a FileInputSource.

        Added skeleton sparql parser and test framework.

        Included pyparsing (pyparsing.sourceforge.net) for sparql parsing.

        Added attribute support to namespaces.


2005/06/28
        RELEASE 2.1.3

        Added Ivan's sparql-p implementation.

        Literal is now picklable.

        Added optional base argument to serialize methods about which to relativize.

        Applied patch to remove some dependencies on Python 2.4
        features.

        Fixed BNode's n3 serialization bug (recently introduced).

        Fixed a collections related bug.


2005/05/13
        RELEASE 2.1.2

        Added patch from Sidnei da Silva that adds a sqlobject based backend.

        Fixed bug in PrettyXMLSerializer (rdf prefix decl was missing sometimes)

        Fixed bug in RDF/XML parser where empty collections where
        causing exceptions.


2005/05/01
        RELEASE 2.1.1

        Fixed a number of bugs relating to 2.0 backward compatibility.

        Fixed split_uri to handle URIs with _ in them properly.

        Fixed bug in RDF/XML handler's absolutize that would cause some URIRefs to end in ##

        Added check_context to Graph.

        Added patch the improves IOMemory implementation.


2005/04/12
        RELEASE 2.1.0

        Merged TripleStore and InformationStore into Graph.

        Added plugin support (or at least cleaned up, made consistent the
        plugin support that existed).

        Added value and seq methods to Graph.

        Renamed prefix_mapping to bind.

        Added namespaces method that is a generator over all prefix,
        namespace bindings.

        Added notion of NamespaceManager.

        Added couple new backends, IOMemory and ZODB.


2005/03/19
        RELEASE 2.0.6

        Added pretty-xml serializer (inlines BNodes where possible,
        typed nodes, Collections).

        Fixed bug in NTParser and n3 methods where not all characters
        where being escaped.

        Changed label and comment methods to return default passed in
        when there is no label or comment. Moved methods to Store
        Class. Store no longer inherits from Schema.

        Fixed bug involving a case with rdf:about='#'

        Changed InMemoryBackend to update third index in the same style it
        does the first two.


2005/01/08
        RELEASE 2.0.5

        Added publicID argument to Store's load method.

        Added RDF and RDFS to top level rdflib package.


2004/10/14
        RELEASE 2.0.4

        Removed unfinished functionality.

        Fixed bug where another prefix other than rdf was getting
        defined for the rdf namespace (causing an assertion to fail).

        Fixed bug in serializer where nodeIDs were not valid NCNames.


2004/04/21
        RELEASE 2.0.3

        Added missing "from __future__ import generators" statement to
        InformationStore.

        Simplified RDF/XML serializer fixing a few bugs involving
        BNodes.

        Added a reset method to RDF/XML parser.

        Changed 'if foo' to "if foo is not None" in a few places in
        the RDF/XML parser.

        Fully qualified imports in rdflib.syntax {parser, serializer}.

        Context now goes through InformationStore (was bypassing it
        going directly to backend).


2004/03/22
        RELEASE 2.0.2

        Improved performance of Identifier equality tests.

        Added missing "from __future__ import generators" statements
        needed to run on Python2.2.

        Added alternative to shlib.move() if it isn't present.

        Fixed bug that occured when specifying a backend to
        InformationStore's constructor.

        Fixed bug recently introduced into InformationStore's remove
        method.


2004/03/15
        RELEASE 2.0.1

        Fixed a bug in the SleepyCatBackend multi threaded concurrency
        support. (Tested fairly extensively under the following
        conditions: multi threaded, multi process, and both).  NOTE:
        fix involved change to database format -- so 2.0.1 will not be
        able to open databases created with 2.0.0

        Removed the use of the Concurrent wrapper around
        InMemoryBackend and modified InMemoryBackend to handle
        concurrent requests. (Motivated by Concurrent's poor
        performance on bigger TripleStores.)

        Improved the speed of len(store) by making backends
        responsible for implementing __len__.

        Context objects now have a identifier property.


2004/03/10
        RELEASE 2.0.0

        Fixed a few bugs in the SleepyCatBackend multi process
        concurrency support.

        Removed rdflib.Resource

        Changed remove to now take a triple pattern and removed
        remove_triples method.

        Added __iadd__ method to Store in support of store +=
        another_store.


2004/01/04
        RELEASE 1.3.2

        Added a serialization dispatcher.

        Added format arg to save method.

        Store now remembers prefix/namespace bindings.

        Backends are now more pluggable

        ...

2003/10/14
        RELEASE 1.3.1

        Fixed bug in serializer where triples where only getting
        serialized the first time.

        Added type checking for contexts.

        Fixed bug that caused comparisons with a Literal to fail when
        the right hand side was not a string.

        Added DB_INIT_CDB flag to SCBacked for supporting multiple
        reader/single writer access

        Changed rdf:RDF to be optional to conform with latest spec.

        Fixed handling of XMLLiterals


2003/04/40
        RELEASE 1.3.0

        Removed bag_id support and added it to OLD_TERMS.

        Added a double hash for keys in SCBacked.

        Fixed _HTTPClient so that it no longer removes metadata about
        a context right after it adds it.

        Added a KDTreeStore and RedlandStore backends.

        Added a StoreTester.


2003/02/28
        RELEASE 1.2.4

        Fixed bug in SCBackend where language and datatype information
        where being ignored.

        Fixed bug in transitive_subjects.

        Updated some of the test cases that where not up to date.

        async_load now adds more http header and error information to
        the InformationStore.


2003/02/11
        RELEASE 1.2.3

        Fixed bug in load methods where relative URLs where not being
        absolutized correctly on Windows.

        Fixed serializer so that it throws an exception when trying to
        serialize a graph with a predicate that can not be split.


2003/02/07
        RELEASE 1.2.2

        Added an exists method to the BackwardCompatibility mixin.

        Added versions of remove, remove_triples and triples methods
        to the BackwardCompatility mixin for TripleStores that take an
        s, p, o as opposed to an (s, p, o).


2003/02/03
        RELEASE 1.2.1

        Added support for parsing XMLLiterals.

        Added support for proper charmod checking (only works in
        Python2.3).

        Fixed remaining rdfcore test cases that where not passing.

        Fixed windows bug in AbstractInformationStore's run method.


2003/01/02
        RELEASE 1.2.0

        Added systemID, line #, and column # to error messages.

        BNode prefix is now composed of ascii_letters instead of letters.

        Added a bsddb backed InformationStore.

        Added an asyncronous load method, methods for scheduling context
        updates, and a run method.


2002/12/16
        RELEASE 1.1.5

        Introduction of InformationStore, a TripleStore with the
        addition of context support.

        Resource __getitem__ now returns object (no longer returns a
        Resource for the object).

        Fixed bug in parser that was introduced in last release
        regaurding unqualified names.


2002/12/10
        RELEASE 1.1.4

        Interface realigned with last stable release.

        Serializer now uses more of the abbreviated forms where
        possible.

        Parser optimized and cleaned up.

        Added third index to InMemoryStore.

        The load and parse methods now take a single argument.

        Added a StringInputSource for to support parsing from strings.

        Renamed rdflib.BTreeTripleStore.TripleStore to
        rdflib.BTreeTripleStore.BTreeTripleStore.

        Minor reorganization of mix-in classes.


2002/12/03
        RELEASE 1.1.3

        BNodes now created with a more unique identifier so BNodes
        from different sessions do not collide.

        Added initial support for XML Literals (for now they are
        parsed into Literals).

        Resource is no longer a special kind of URIRef.

        Resource no longer looks at range to determine default return
        type for __getitem__. Instead there is now a get(predicate, default)
        method.


2002/11/21
        RELEASE 1.1.2

        Fixed Literal's __eq__ method so that Literal('foo')=='foo' etc.

        Fixed Resource's __setitem__ method so that it does not raise
        a dictionary changed size while iterating exception.


2002/11/09
        RELEASE 1.1.1

        Resource is now a special kind of URIRef

        Resource's __getitem__ now looks at rdfs:range to determine
        return type in default case.



2002/11/05
        RELEASE 1.1.0 # A new development branch

        Cleaned up interface and promoted it to SIR: Simple Interface
        for RDF.

        Updated parser to use SAX2 interfaces instead of using expat directly.

        Added BTreeTripleStore, a ZODB BTree TripleStore backend. And
        a default pre-mixed TripleStore that uses it.

        Synced with latest (Editor's draft) RDF/XML spec.

        Added datatype support.

        Cleaned up interfaces for load/parse: removed generate_path
        from loadsave andrenamed parse_URI to parse.


2002/10/08
        RELEASE 0.9.6 # The end of a development brant

        BNode can now be created with specified value.

        Literal now has a language attribute.

        Parser now creates Literals with language attribute set
        appropriately as determined by xml:lang attributes.


        TODO: Serializer-Literals-language attribute

        TODO: Change __eq__ so that Literal("foo")=="foo" etc

        TripleStores now support "in" operator.
        For example: if (s, p, o) in store: print "Found ", s, p, o

        Added APIs/object for working at level of a Resource. NOTE:
        This functionality is still experimental

        Consecutive Collections now parse correctly.

2002/08/06
        RELEASE 0.9.5

        Added support for rdf:parseType="Collection"

        Added items generator for getting items in a Collection

        Renamed rdflib.triple_store to rdflib.TripleStore to better follow
        python style conventions.

        Added an Identifier Class

        Moved each node into its own Python module.

        Added rdflib.util with a first and uniq function.

        Added a little more to example.py

        Removed generate_uri since we have BNodes now.


2002/07/29
        RELEASE 0.9.4

        Added support for proposed rdf:nodeID to both the parser and
        serializer.

        Reimplemented serializer which now nests things where
        possible.

        Added partial support for XML Literal parseTypes.


2002/07/16
        RELEASE 0.9.3

        Fixed bug where bNodes where being created for nested property
        elements when they where not supposed to be.

        Added lax mode that will convert rdf/xml files that contain bare
        IDs etc. Also, lax mode will only report parse errors instead of
        raising exceptions.

        Added missing check for valid attribute names in the case of
        production 5.18 of latest WD spec.


2002/07/05
        RELEASE 0.9.2

        Added missing constants for SUBPROPERTYOF, ISDEFINEDBY.

        Added test case for running all of the rdf/xml test cases.

        Reimplemented rdf/xml parser to conform to latest WD.


2002/06/10
        RELEASE 0.9.1

        There is now a remove and a remove_triples (no more overloaded
        remove).

        Layer 2 has been merged with layer 1 since there is no longer a
        need for them to be separate layers.

        The generate_uri method has moved to LoadSave since triple stores
        do not have a notion of a uri. [Also, with proper bNode support on
        its way the need for a generate_uri might not be as high.]

        Fixed bug in node's n3 function: URI -> URIRef.

        Replaced string based exceptions with class based exceptions.

        Added PyUnit TestCase for parser.py

        Added N-Triples parser.

        Added __len__ and __eq__ methods to store interface.


2002/06/04
        RELEASE 0.9.0

        Initial release after being split from redfootlib.


