app/controllers/search_methods/sfx4.rb in umlaut-3.1.0 vs app/controllers/search_methods/sfx4.rb in umlaut-3.1.1

@@ -38,22 +38,26 @@
     def sfx4_db_connection
       az_title_klass.connection
     end
 
     # Needs to return ContextObjects
-    def find_by_title                  
+    def find_by_title
       connection = sfx4_db_connection
       query_match_clause = case search_type_param
         when "contains"
           terms = title_query_param.split(" ")
           #SFX4 seems to ignore 'the' or 'a' on the front, so we will too. 
           if (["the", "a"].include? terms[0])
             terms = terms.slice(1..-1)
           end
           # Then make each term required, but stemmed. Seems to match SFX4, 
           # and more importantly give us decent results. 
+          #
+          # For reasons we can't entirely tell, the wildcard "*" on terms of less
+          # than 2 causes false negatives. Otherwise we use it to be consistent
+          # with SFX. This reverse-engineering is full of pitfalls.
           query = terms.collect do |term|
-            "+" + connection.quote_string(term) + "*"
+            "+" + connection.quote_string(term) + (term.length > 2 ? "*" : "")
           end.join(" ")
           "MATCH (TS.TITLE_SEARCH) AGAINST ('#{query}' IN BOOLEAN MODE)"
         when "begins"
           # For 'begins', searching against TITLE itself rather than TITLE_SEARCH gives us 
           # results more like SFX4 native, without so many 'also known as' titles confusing