You've already been told about the eval/require dance, which I won't repeat. Instead, I'll tell you about a similar problem I had recently and how I solved it. It involves an app I'm writing for the Palm, which needs a lot of data to work with. I'm writing my app in C, but fetching the data and mangling it into an appropriate format using perl.

I did briefly consider using my perl script to mangle the data into C source code and then #include it into my C program. That would certainly make the task easier as I wouldn't have to mess around with databases. However, it would mean that users would have to rebuild the program to update the data, and would also lead to resource "issues". Instead, I did The Right Thing, and my perl script instead spews out a few PDBs using Palm::Raw (which whilst it is listed on CPAN and downloadable through a browser, CPAN.pm didn't want to install for me, but that's irrelevant here). And my C code opens those databases and uses them.

Of course, most of the reasons for me using PDBs are irrelevant to your perl program, but you might like to consider using something like dbm files instead. It seems somehow more elegant :-)