Hashtable for UID lookups is a bad choice. Eg. with UID FETCH 100:200 you may have only a few of the UIDs in that range.
Good point. I'll go with binary search then.
My server keeps indexes in disk, so I use the kind of b-tree you
described. It's C and LGPL, so go ahead and see if it helps. http://dovecot.procontrol.fi/, src/lib-index/mail-tree*. That could
probably still be done faster, at least building the tree first
time feels too slow.
Thanks! I'll share the Java version after I've created it.
Cheers,
- Joe
