actually, i'd worry plenty about memory usage and caching result sets if i were Kozz. as was remarked, the fetchrow_ methods do precisely that: fetch a single result record. however the fetchall_ methods cache entire result sets locally; that can get pretty massive if you're querying against credit bureau data or some 120 GB instance.

it's usually a good idea to let the database server do as much work as possible, including summarization and grouping operations. sometimes you do need all query results before you can proceed, but more often than not, you can just process the results iteratively. and to reiterate Higle's comment, the dbms is usually much more conscientious about memory optimization than you'll be on the client side.

cheers.