Redland Blog

Download: raptor-1.4.17.tar.gz

• Added two new JSON serializers: resource-centric ‘json’
  (Talis RDF/JSON)
  and triple-centric ‘json-triples’
• Added a new public XML SAX2 API class
• Added a new error handling structure
• Made the I/O Stream class support reading
• Added several new API methods.
• Made several fixes, portability and resilience improvements.
• Fixed reported issues: 0000252 and 0000245.

See the Raptor 1.4.17 Release Notes
for the full details of the changes.

I’ve been updating the new JSON serializers (resource based on the Talis RDF/JSON and triples-centric) in Subversion so that a callback parameter can be set. This is what was formerly available in Triplr when it was using the perl JSON module, before I switched it to the native C JSON serializers. That should restore that useful feature again, once I add the necessary feature setting into Triplr.

As usual I made tests work – ‘make distcheck’ has to pass in several ways:

• On OSX 10.4.x with gcc 4.0.1, MacPorts curl and libxml (without distcheck, it’s just too darn slow on ibook)
• Linux Debian gcc 4.2.3 (aka newest) with curl and libxml
• Linux Debian gcc 4.2.3 with curl and expat
• Linux Debian g++4.2.3 and configure option --enable-release and tweaked to use compile options “-O3 -Wall” with curl and libxml. This uses g++ as a “better C compiler” and usually causes some tweaks such as explicit casts and some portability bits.

I also run some tests on Linux inside valgrind as a leak check.

I probably could and should automate this but never found the time so far. It should probably produce pretty red/green failure graphs and all that jazz.

Most of the time is not code changes above but in turning the ChangeLog commits into some form of coherent draft release notes. The major items from those will turn into the summary of the release, the draft news nearby.

I was updating raptor SVN to fix bug 245 that said there were duplicate triples in the turtle output(i.e. generated by the internal raptor_abbrev system), by the dumbest functional way to do it – scan all the raptor_sequence of triples gathered. This, as I expected, turns out to make an already quite slow serializer (since it’s using raptor_sequence) into one running incredibly slow.

So, I’ve just finished updating the internal raptor_avltree to gain a cursor. This allows storing ordered items in an AVL Tree, which has O(log(N)) for add, delete and lookup, compared to raptor_sequence‘s O(N). Yay!

Next step is to try putting it into the raptor_abbrev subsystem.