Stoyan Stefanov has posted some of his tips to his blog today. Specifically, they deal with the PEAR::DB class, demonstrating some of the optimization of the package he's discovered in his coding experience.
If you use PEAR::MDB2, you can set a custom debug handler and collect all the queries you execute for debugging and performance tunning purposes, as shown before. But what if you're using PEAR::DB? Well, since PEAR::DB doesn't allow you such a functionality out of the box, you can hack it a bit to get similar results.
He creates a simple app to help with the illustration - a number of select queries to grab zipcode information from the database. As it stands, the PEAR::DB package doesn't handle the debugging well, so he adds in a few more lines to buffer the connection and some reporting code to check the resulting output (as well as some of his sample reports).