|
Based on my experience, I think the important concerns: database drivers, bridges to linear algebra and stat/prob libs i mentioned, C extensions and java libraries. I work in python and ruby where appropriate, and forgot all the perl i know, except when I maintain other people's code. The database driver issue, is moot, all three languages have solid packages for mysql, SQLite or any non-obscure RDBMS and probably hook into couchDB, or mnesia, or any of the non-SQL databases. The hooks to MATLAB, R, gsl or octave or linpack or whatever, not hard to google. Worst case, you pipe / tee flat files between apps (I'm assuming you're using linux or FreeBSD or solaris). C extensions: Pretty straightforward in python or ruby. In face, ruby-inline, pyrex/cython make it about as straightforward as could be, assuming you know to look for memory leaks and clean up after yourself. And Jruby is produciton-ready, you can be pretty confident you can hook into whatever java libs you need. I suggest you look at what Jane St Capital is writing about OCaml analytics. |