[Xapian-discuss] Xapian performance on gmane.org compared

Olly Betts olly at survex.com
Fri Aug 28 12:26:49 BST 2009

On Fri, Aug 28, 2009 at 08:57:28AM +0200, Henry wrote:
> So what's the bottom line?  I'm going to spend some time now building an 
> equivalent sized flint DB to see whether it's best to revert; but it 
> would be interesting to know what your gut is telling you (I can't parse 
> exactly what your final judgement was in your post):  flint or chert for 
> phrase searching?

I'd guess chert is as good or better.

I can't see #326 making things significantly worse in slow phrase cases,
and the reduction in postlist table size should help a bit.  Also,
chert tracks statistics which allow tighter bounds on weights, which
may help some cases.

It would be interesting to hear how they actually compare if you want to
run tests with flint too.


More information about the Xapian-discuss mailing list