[Xapian-discuss] Xapian performance

Olly Betts olly at survex.com
Mon Oct 30 02:01:13 GMT 2006


On Sun, Oct 29, 2006 at 03:30:37PM -0300, Fernando Nemec wrote:
> Is there any performance increase if I gzip a document's data before I
> index it?

Assuming you're doing the standard "do a search and show 10 or so
mathches at a time", it's likely to reduce disk cache pressure a little
so it'll be more likely that the postlists and values you want are
cached and won't have to be read from disk so it should speed things
up a little.

There's actually a patch to do this already - look for the "zlib"
patches here (one for flint, one for quartz):

http://www.oligarchy.co.uk/xapian/patches/

These allow you to compress tags in any table, but record is probably
the most useful.  Instructions are here:

http://article.gmane.org/gmane.comp.search.xapian.general/1323

Cheers,
    Olly



More information about the Xapian-discuss mailing list