[Xapian-tickets] [Xapian] #653: OP_PHRASE seems to sometimes match terms in wrong order
Xapian
nobody at xapian.org
Sat Aug 30 06:48:37 BST 2014
#653: OP_PHRASE seems to sometimes match terms in wrong order
----------------------------------+------------------
Reporter: medoc | Owner: olly
Type: defect | Status: new
Priority: normal | Milestone:
Component: Backend-Chert | Version:
Severity: normal | Keywords:
Blocked By: | Blocking:
Operating System: Linux |
----------------------------------+------------------
\
\
It seems that OP_PHRASE will sometimes match terms in the wrong order.
This was reported by a Recoll user on Xapian 1.2.12, and I seem to be able
to reproduce it on Xapian 1.2.16:
{{{
jfmac$ delve -t hurricane xapiandb/
Posting List for term `hurricane' (termfreq 1, collfreq 2, wdf_max 2): 4
jfmac$ delve -t katrina xapiandb/
Posting List for term `katrina' (termfreq 1, collfreq 3, wdf_max 3): 4
jfmac$ delve -t hurricane -r 4 xapiandb/
Position List for term `hurricane', record #4: 199881 203084
jfmac$ delve -t katrina -r 4 xapiandb/
Position List for term `katrina', record #4: 199882 202473 203085
jfmac$ xadump -q katrina hurricane
q argc 2
DB: ndocs 8 lastdocid 8 avglength 135233
DB: terms are stripped
Performing query `Xapian::Query((katrina PHRASE 5 hurricane))'
Estimated results: 1
Document ID 4 100% [url=file:///home/dockes/Downloads/[Mehdi_Khosrow-
Pour,_Mehdi_Khosrow-Pour]_Encyclope(BookZZ.org).txt
...
}}}
(xadump is a tool in the recoll source which performs low level Xapian
operations much like delve. Unlike delve it has a primitive query
capability, normally performing AND queries, here modified to do
OP_PHRASE, which a compiled-in phrase window (here 5). The window has to
be at least 4 for the problem to appear (no match with a window of 2 or
3).
The index is not that big (3.6 MB compressed), I'll try to attach it to
the ticket. If this fails, I can upload it wherever convenient.
\
\
\
--
Ticket URL: <http://trac.xapian.org/ticket/653>
Xapian <http://xapian.org/>
Xapian
More information about the Xapian-tickets
mailing list