Minimize Database Calls and Perform Lazy-Loading
One of the nastiest things about the object/relational mismatch is the fact that proper object-oriented coding implies that you should make an inordinate number of database calls. The canonical object-oriented search looks like this:
Get all the primary keys from the database that match search criteria.
For each match, go to the database to load its primary attributes.
Also for each match, go to the database once for each object attribute.
This degree of database access is not only unnecessary, but it will also render your application totally unusable. Therefore, this area is where compromising on object-oriented principles is not only acceptable, but also a requirement.
As you’ve probably guessed, your Java application should do whatever it can to minimize the number of trips it makes to the database to support a single logical operation. Selecting the columns necessary to instantiate objects as part of the search will reduce the number of calls immensely. In the true object-oriented way, a search returning 100 rows results in at least 101 calls to the database. By selecting the necessary columns in your search, you can potentially drop this number to just the single search query.
Of course, selecting all the columns when you are probably going to use only one or two can itself be a performance problem. Your application needs to balance the need to minimize calls to the database with the reality of how many rows you will actually use. ...