If you have data in a CSV format or if it is more convenient for you to get CSV than XML or JSON, then you may prefer the CSV option. Solr's CSV support is fairly flexible. You won't be able to specify an index-time boost but that's an uncommon need.
Tip
CSV is uniquely the only format that Solr supports for round-tripping data. As such, you can query for CSV-formatted data that is suitable to be added right back into Solr (for stored fields only, of course). The XML and JSON query output formats are structured differently than their input formats, so they don't count.
To get CSV data out of a local PostgreSQL database for the MusicBrainz tracks, we ran this command:
psql -U postgres -d musicbrainz_db -c "COPY (\ select 'Track:' || t.id as id, 'Track' as type, t.name as t_name, t.length/1000 as t_duration, a.id as t_a_id, a.name as t_a_name, albumjoin.sequence as t_num, r.id as t_r_id, r.name as t_r_name, array_to_string(r.attributes,' ') as t_r_attributes...