Op 9 sep 2013, om 10:06 heeft Harmen de Jong - CoachR Group B.V. het
volgende geschreven:
Our table and column names are not too long either as E.Pasma
suggests.
Ha, I did not mean the length of the names but the length of the hash
table (NL: klutstabel), That is the number of buckets over which the
hash values are distributed. I looked some further in the code and now
believe that this is 1024. That is stilll generous for a normal
database, But with 10.000 tables you will have about 10 table names in
each bucket. So the enigine wiill need to do a lot of searching within
each bucket. That involves checking the name of each item in the
bucket with the searched name.
With regards to the number buckets I found a meaningful comment in the
code:
/* The inability to allocates space for a larger hash table is
** a performance hit but it is not a fatal error. So mark the
** allocation as a benign.
*/
Another generousity is that the code uses a constant for the length of
the hash tables so you can change that by a single edit. There are of
course advantages and disadvantages of doing that,.
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
