On Dec 30, 2013, at 10:54 AM, Maggie X <[email protected]> wrote: > > - Worst thing(s) about PDL > > Not enough adoption in the industry. > > - I would use PDL more (or at all) if only ... > > More people use it for data science in the industry. Currently people mostly > use python (numpy, scipy), with a few trying out julia and even go. > > - Any suggestions for PDL development? > > Better packaging and dependency management (indestructible core-installation > plus different packages for different use cases). > Easier graphics (documentation and examples; default go-to option, whether > it's PLplot or Prima or Gnuplot). > Better marketing.
Yes, these are the perennial issues. A decade ago I marketed PDL pretty hard in the solar physics community but the installation process at that time was pretty difficult and that limited adoption. More recently, people are migrating to Python within that community. Numpy seems to have achieved critical mass early on, in part because of the faster learning curve for the underlying language. (That's unfortunate since, as far as I can see, it is *still* technically inferior to PDL...) As near as I can tell the battle for widespread adoption/dominance has been lost ever since Python became the de facto standard for introductory programming courses -- there's just too much inertia (and too much anti-Perl propaganda) steering new graduates in the direction of Numpy and Python. That's not to say PDL is doomed -- It is technically superior and I continue to exploit that routinely to get great science done. And we do seem to have a large and still-growing user base. But I believe that PDL will remain a market underdog for the foreseeable future, at least for interactive scientific computing.
_______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
