Lucene Queries -- joining documents and maintaining relevancy

Question

I am trying to create a Lucene search using school name and player name to return videos. I am trying to decide between two methods.

Method A is to index the school name and player name on the video document and use a boolean query to search on these fields.

Method B is to create separate document types and make 3 unique queries.

Documents:

• school document - stores a school_id and indexes the school name
• player document - stores a school_id and sport_id, and indexes the player name

The 3 queries:

1. Search for all school documents with school name
2. Search for all player documents with player name
3. Search the videos for all content with school_id and sport_id from the first two queries.

What are the pros/cons of both methods?

Answer 1

You almost certainly want to go with method A. In order to combine relevance scores from two indexes you essentially have to reinvent Lucene.

The downside is that if a school/player changes their name, you have to reindex. That seems pretty unusual though.

Keep in mind here that Lucene is really only good if you have a large amount of free-text to search. If it's just a few words (like the name of a school) using the free-text capabilities of MySQL or your other favorite rdbms will probably be just as fast and will be a lot easier to implement. You won't have the issue with reindexing on rename, for example.