Xapian discuss - Jul 2011 - Phrase search problem

Hi,

I'm experiencing problems when doing phrase searches with adjacent
repeated terms. Example:

if I search for 'curtain curtain' and there are documents that matches
the query, they aren't returned. But, if I search for 'curtain nice
curtain' and there are documents that matches this query, it works ok.

attached there is a python program that shows the problem. I tried
this program with python 2.5 and python 2.6, the xapian version I'm
using is the one used together with xappy (a modified version of
1.2.0, according to xapian.version_string()). I tried running this
same program in the xapian that comes with Ubuntu (1.0.20) and it
doesn't work since chert is not available.

can someone help me with this?

-- 
Bruno
-------------- next part --------------
A non-text attachment was scrubbed...
Name: phrasesearch.py
Type: text/x-python
Size: 1467 bytes
Desc: not available
URL:
<http://lists.xapian.org/pipermail/xapian-discuss/attachments/20110720/d72760cb/attachment.py>

On Wed, Jul 20, 2011 at 09:14:03AM -0300, Bruno Rezende
wrote:
> if I search for 'curtain curtain' and there are documents that
matches
> the query, they aren't returned. But, if I search for 'curtain nice
> curtain' and there are documents that matches this query, it works ok.
If you look at the positional information for curtain in document 1 with
delve, you'll notice it only occurs once:

$ delve /tmp/tmplLEyRQ/ -r1 -tcurtain
Position List for term `curtain', record #1: 1

Your issue is that you are trying to add curtain at positions 0 and 1,
but position 0 isn't getting set, so you only actually add it at
position 1.  Simply adding one to tpos seems to avoid this problem.

This is pretty clearly a bug, but I'm not certain where it goes wrong
yet.  It is an indexing issue though.

(Incidentally, for more tips on debugging this sort of situation, see:
http://trac.xapian.org/wiki/FAQ/NoMatches )
> I tried running this
> same program in the xapian that comes with Ubuntu (1.0.20) and it
> doesn't work since chert is not available.
You can just change "xapian.chert_open" to
"xapian.WritableDatabase" -
unless there's a reason why you wanted chert, it's better to use the
non-specific form.

Cheers,
    Olly