Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
Description
As noted by ldavid2020 on the solr-user mailing list (Tue, 20 Dec 2011) the behavior of edismax when the pf param is used and the query string contains quotes arround part of the query makes no sense at all...
For the same query: 2012 "japan airlines" flight status
dismax...
outputs:
+((DisjunctionMaxQuery((TTL:2012)~0.1) DisjunctionMaxQuery((TTL:"japan airlin"~3)~0.1) DisjunctionMaxQuery((TTL:flight)~0.1) DisjunctionMaxQuery((TTL:status)~0.1) )~3) DisjunctionMaxQuery((TTL:"2012 japan airlin flight status"~3)~0.1)The parsedquery has DisjunctionMaxQuery((TTL:"2012 japan airlin flight status"~3)~0.1).
While edismax...
outputs:
+((DisjunctionMaxQuery((TTL:2012)~0.1) DisjunctionMaxQuery((TTL:"japan airlin"~3)~0.1) DisjunctionMaxQuery((TTL:flight)~0.1) DisjunctionMaxQuery((TTL:status)~0.1) )~3) DisjunctionMaxQuery((TTL:"2012 flight status"~3)~0.1)The parsedquery has DisjunctionMaxQuery((TTL:"2012 flight status"~3)~0.1).
...
So it seems edismax ignores "japan airlines" for the pf matching. This could cause some issues, in that a document with exactly phrase "2012 japan airlines flight status" will have the same relevancy score with another one with two phrases "japan airlines", "2012 flight status" far away.