Perl is my first choice go-to language, but I'm not finding the CPAN advantage as big as it used to be. Increasingly I'm having to use languages I don't prefer (not out of any hostility, just not as comfortable in them) because they have libraries that CPAN either doesn't include, or doesn't include with the same polish/functionality. For example, there doesn't seem to be anything in the Perl world as good as the Python Imaging Library (PIL); PerlMagick is featureful but not very fun to use. The data-visualization libraries are also way behind Python's matplotlib. NLP is a closer call, but overall I find NLTK better integrated and maintained than the stuff in CPAN, though CPAN does have modules for most things you might want to do. I'm currently using Perl for some simple GIS stuff, and the GIS packages also are pretty hit-or-miss.

Admittedly I might just not be good at finding the right packages. Python seems to have more of a culture of headline projects, which register their own domain names, like NLTK, PIL, numpy, etc., while Perl mostly has a big pile of CPAN modules, which range from excellent and actively maintained, to a script someone once wrote that's incomplete and hasn't been touched in 12 years.