The first one that I thought was pretty cool was something that took a memory model (in tabular format) and generated the necessary CREATE TABLE commands to build the table, complete with primary key and foreign key constraints. Then it took that output (in the Unix pipe-lining philosophy) and checked that each of the tables primary key types matched the foreign key types when they appeared in other tables; another script ordered the commands, so that the independent tables appeared first, followed by the dependent tables.

It impressed the hell out of the senior consultant on the contract, which didn't hurt. It definitely sold me on the value of Perl as an Open Source tool. I sure wish I'd made copies of those scripts. That was in about '98, and (I think) Perl 5 had just come out. Good times.

PS Oh yeah, all built without actual access to an SQL database .. I just looked up the appropriate syntax on-line. Such a geek.