[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: ldbm database usage

"Kurt D. Zeilenga" wrote:

> I should also note that the IDList (ID_BLOCK) code is
> terribly inefficient.  It uses for loops to move blocks
> of data AND uses linear searches to find items in
> sorted arrays.

It's also extremely hard to understand in places.

I found that changing to a binary search for
ID insertions speeded up some operations significantly.

FYI the plan is to kill off ID lists completely
in Netscape Directory Server 5.0. There's new 
stuff in BerkeleyDB 2.6 which was added for this
particular purpose. Basically the maintenance of the 
secondard index structures is pushed down into
BDB, where it plays better with transactions.