Add the autoGeneratePhraseQueries=true attribute to your text field type
since it now defaults to false.
Also, change your query-time WDF to preserveOriginal="0". I think there were
some changes or bug fixes about the position of the original vs. the
generated parts and you only need the generated parts at query time anyway.
-- Jack Krupansky
-----Original Message-----
From: Shawn Heisey
Sent: Friday, December 07, 2012 8:25 PM
To: solr-user@lucene.apache.org
Subject: parsed query differences between 3.5 and 4.1 snapshot
While investigating differences in query results between Solr 3.5 and a
branch_4x snapshot with a slightly different schema, I came across some
fairly radical differences in how a particular query is parsed. My
default operator on both versions is AND. I am using the lucene query
parser.
The first part of the query gets split into multiple pieces by
WordDelimeterFilter, but solr 3.5 turns it into a phrase query and 4.1
just makes every single variation mandatory. The entire query (which
includes a complex filter) returns 11 results on 3.5 and only 2 on 4.1.
Is there a bug here?
query string:
POLAND-COFFIN CALENDAR
parsed query for Solr 3.5:
+catchall:"(poland-coffin poland) coffin" +catchall:calendar
parsed query for Solr 4.1 snapshot:
+(+catchall:poland-coffin +catchall:poland +catchall:coffin)
+catchall:calendar
Schema for 3.5:
http://fpaste.org/AuZj/
Schema for 4.1 snapshot:
http://fpaste.org/Maaz/
