Now that PB is stable enough to run in deerpark (and I can finally ditch
safari, fast and pretty but plain as a piece of toast without nothing
on), I moved over the old piggy-bank directory from the old profile to
the new one and restarted.
And piggy-bank took litterarely a few minutes to start.
First I thought it was doing some reinitialization or reindexing of some
sort... but now that I know the code well I knew that nothing like that
was going on, it was simply loading the triple store.
I looked into the folder again and the file that contains the triples is
a 7Mb file. Hmmm, considering that I have roughly 60 objects, for maybe
a total of 600 triples in there, 7mb seemed a little bit excessive to me.
Given that run-away loops are not unlikely to happen around here ;-) I
thought there might be something wrong with the way the triple store is
handled. So I exported all my data out of the triple store, moved the
piggy-bank folder out of the way, went in the about:config box, search
for piggy-bank.initialize, turned it to true and restarted deerpark.
As expected, the new "my piggy bank" is now starting up in a few
seconds. Empty.
So I loaded back the dumped triples (simply drag your dump.n3 file back
into the window, click the data icon, then do a 'save all') and restarted.
To my surprise, the loaded "my piggy bank" still starts up in a few
seconds... a 20-fold performance improvement in starting up.
NOTE: this shows that there is something deteriorating in the way piggy
bank handles the triple store or something wrong in the triple store itself.
Jeen, did you guys experienced anything like this before?
NOTE2: the resulting triple store file is still 7mb so I gather that
this is probably something that has to do with hashing performance and
the file is treated as a memory disk. I really think we should ditch
this native store and more over to the memory one... but I'll have to
move to Sesame 1.2.x for that. For sure the use of the native store for
temporary profiles is silly.
WARNING: do not try this at home on your own piggy-bank because I found
a bug in the data import that destroys all your tags (basically the
base64 encoded tags are re-base64-encoded before being saved... don't
don't why this is happening, it's right next on my list).
--
Stefano Mazzocchi
Research Scientist Digital Libraries Research Group
Massachusetts Institute of Technology location: E25-131C
77 Massachusetts Ave telephone: +1 (617) 253-1096
Cambridge, MA 02139-4307 email: stefanom at mit . edu
-------------------------------------------------------------------
Received on Fri Aug 19 2005 - 22:52:57 EDT