olly at survex.com
Mon Oct 6 12:59:00 BST 2008
On Mon, Oct 06, 2008 at 12:22:13PM +0100, Olly Betts wrote:
> Thankyou! I can reproduce it with these - investigating...
OK, I've found the cause. The document lengths mean that we get weights,
but if you test on a large messy database with a lot of documents with
length 1, then the top 10 all have the same high weight and you get all
100% entries in the MSet (which is why neither Richard nor I could
reproduce when we tried).
The attached patch fixes this for me. I'll distill your example into
a testcase and commit it.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 445 bytes
Desc: not available
Url : http://lists.xapian.org/pipermail/xapian-discuss/attachments/20081006/7f59bb9d/attachment.patch
More information about the Xapian-discuss