[redland-dev] Raptor 2 GIT work

[Dave Beckett]

I just posted a blog entry at
http://journal.dajobe.org/journal/posts/2010/01/30/raptor-1421-raptor2-git/

and I don't expect you are all following me so I'll repeat some of it here
since I'm interested in feedback from redland users / devs about the raptor
2 decisions still left.


The ongoing Raptor 2 ABI/API updating is cleaning up 10 years of API and
internal cruft.  GIT is really helping speed up the ease of this work with
the branching, staging/index and stash concepts it supports allowing false
paths to be managed.  The results can be seen on branch ‘master’ of raptor.

The updating is going well in the sense that make distcheck test suite
passes, but there are still things to decide including:
    * Rename all raptor_CLASS_copy copy constructors to something else
wither raptor_new_CLASS_from_CLASS (also used in raptor – Doh!) or to
raptor_CLASS_addref which signifies better that it just adds a reference to
the object, it’s a shallow copy, not a deep one.
    * Unify raptor_world, rasqal_world and librdf_world – which might help
share classes between the libraries. Not sure if this is a good idea yet.
    * Add a graph term to the (subject, predicate, object) triple returned
from parsing. I am probably going to do this.
    * Turn the raptor_locator object into a more of a log (like librdf_log)
or exception object, with inner log/exceptions.
    * Improve the callback interface that passes error, warning etc.
messages to user code.

I need to decide at what point to roll out an alpha release of Raptor 2,
which will probably be numbered 1.9.0. Some of the above possibilities might
be worth putting in a later alpha release.

This can all be seen in the GIT repository which includes instructions for
checkout at http://git.librdf.org/

Dave