Hi Sebastian,
Can you apply the attached patch to your libsrc/Wi directory and try
again please. This should make your insert run in much better
time, even
without large number of buffers.
Impressive. Even with 5000 buffers it only takes 3 seconds or less
(I timed
everything including startup time). Nice!
Will you put that in the next snapshot?
Yes, this patch will be part of the upcoming releases of both V5 and V6.
You may want to concider using the TTLP_MT stored procedure to load
either large number of triples or large triples, since this is
much more
efficient. You can call this with:
TTLP_MT( ?, 'graphname', 'graphname', 255);
and bind with a blob parameter that holds the triples.
Is this faster for all kinds of triples or only large ones?
This is possibly faster for all kinds of triples.
Patrick
