Hello Jean-Claude,

>> Geez, Brian, you're a wizard!

JCW> I agree 100%.

Err, I have to admit I made this mistake - I have many versions of
this data file and in the last MK test I used the one that was half as
big as previously. Still, Brian's version of my Python/MK script has
came ahead of BSDDB with file containing 1.2 million words: 53 seconds
vs 63 seconds on average for BSDDB.

>> After syncing: 23.08
JCW> [...]
>> Thanks for the help, Brian, now I have to go away to munch on
>> all that.

JCW> Now that you have these results: what file sizes do you see across the
JCW> different DB's?

Yeah, I've noticed that before - what puzzled me were sizes of various
DBs:

MetaKit: 54.5 MB
SQLite: 61.3 MB
BSDDB: 146.5 MB !!!

The raw data size itself is ~45.8 MB. So MK has little of the overhead
when compared to alternatives.

Benchmarks are fun. :-)

I guess minimizing disk access was among your goals, right?

It's nice if I can read a few values from BSDDB quickly, but if the
program has to read substantial portions of db, disk access must quickly
become the bottleneck.


JCW> (It might also be interesting to compare int-field performances &
JCW> sizes, BTW)

As I suspected.


-- 
Best regards,
 Marcin                            mailto:[EMAIL PROTECTED]

_____________________________________________
Metakit mailing list  -  [email protected]
http://www.equi4.com/mailman/listinfo/metakit

Reply via email to