I have a question regarding performing a lucene query involving permutation.
Say I have two fields: “name” and “keyword” and the user searches for “joes pizza restaurant”. I want some part of that search to match the full contents of the “name” field and some part to match the full content of the keyword field. It should match all the the supplied terms and should match the entire contents of the fields. For example it could match:
1) name:”joes restaurant” keyword:”pizza”
2) name:”joes pizza” keyword:”restaurant”
3) name:”pizza restaurant” keyword:”joes”
4) name:”pizza” keyword:”joes restaurant”
5) name:”pizza joes” keyword:”restaurant”
but it would not match
6) name:”big joes restaurant” keyword:”pizza” – because it’s not a match on the full field
7) name:”joes pizza restaurant” keyword:”nomatch” – because at least one of the terms should match to the keyword field
I’ve thought about possible ways to implement this by calculating all the permutations of the fields and using boolean queries however this doesn’t scale very well as the number of terms increases. Anyone have any clues how to implement this sort of query efficiently?
Lucene docs recommend using separate field which is concatenation of ‘name’ and ‘keyword’ fields for queries spanning multiple fields. Do the search on this field.