[redland-dev] Dydra's use of libraptor2

[Arto Bendiken]

Hello Dave,

As an addition to <http://librdf.org/using.html>, please feel free to
mention Dydra's use of Raptor.

Following our platform upgrade yesterday [1], we are now using
libraptor2 to perform all parsing of data import operations (in any
serialization format supported by Raptor) for our users, as well as
the serving of downloadable export dumps (in the usual RDF formats) of
users' repositories.

Internally, our data flows are based mostly on N-Triples/N-Quads, for
which we have rolled our own custom C library that churns through
triples a little faster than libraptor2. (Some 40+% faster, due to an
uber-optimized, hand-rolled parser, plus no dynamic memory
allocation.)

Externally at the interfaces, however, it's now all libraptor2 --
wholly replacing the earlier Ruby-based parsing/serialization
pipelines that simply proved too slow for the demands of our
customers. Particularly praiseworthy are Raptor's capabilities for
autodetecting input serialization formats, which have considerably
simplified the implementation of our data imports.

Our GitHub fork of libraptor2 is located at
<https://github.com/datagraph/raptor>. Time permitting, and provided
it'd be welcome, we'd like to eventually contribute back our N-Quads
parser implementation, as that would enable us to simply rely on
libraptor2 throughout the stack.

Cheers,
Arto

-- 
Arto Bendiken | arto at dydra.com | http://dydra.com

[1] http://blog.dydra.com/2011/09/07/sparql-11