Although the original post is a few years old, I feel compelled to respond to it. Recently I had to develop a few dynamic PDFs from data in a database. After looking at the avaliable options (including writing out the pdf by hand - yuck), I found PDF::API2 to be the most useful.

The biggest drawback I found to PDF::API2 was that the perldoc left much to be desired. Although it very much has the capabilities to go beyond "simple text" figuring out how to draw curves or use barcodes from the perldoc was difficult at best. If you use PDF::API2 I recommend using http://pdfapi2.sourceforge.net/twiki/ for documentation.

Other than the documentation drawback, expect to spend about a half a day learning the API. Once you get the hang of it, its pretty easy and well done.