[rdfweb-dev] The toolset issue

[Julian Bond]

I did some digging around in RAP and Morten's code last night. If we're 
looking at including it in something like Drupal then we can chuck out 
RDQL support, all the database stuff, all the N3 stuff and the RDF to 
XML serialisers. That gets you down to 13 files and 144Kb. It's a matter 
of minutes to combine what's left into two files. A few more minutes to 
remove all the copious comments and it's two files and 120Kb. That's not 
so bad. RAP's only getting version updates every 6 months or so, so 
staying in step wouldn't be hard.

In this case, the important timings in Morten's work are 25 secs for 
full RAP and 17 secs for Morten's cut down version. The 25% difference 
is all in replacing an array of triples as objects against a simple 
array of triples as arrays. But in the process you also chuck out the 
RAP API which means considerably more code to actually do useful work 
with the results. The remaining 17 secs vs 2.7 secs for Redland is 
entirely basic RDF parsing in PHP vs C. At this point, we're stuck with 
either trying to completely re-write the RDF parser, switching to 
Redland or just putting up with the performance. There maybe some 
performance to be had in optimising RAP's basic parser but that's beyond 
me.

I've got some big FOAF files so I'll try and get some real world timing 
on how long RAP takes to build the triple array and index it.

-- 
Julian Bond Email&MSM: julian.bond at voidstar.com
Webmaster:                 http://www.ecademy.com/
Personal WebLog:          http://www.voidstar.com/
M: +44 (0)77 5907 2173      T: +44 (0)192 0412 433