Drew Conway and John Myles White of the website Dataists decided to try ranking programming languages using a new system: the size of the community and the number of projects. In Conway’s blog post about the results, he admits that there’s no perfect way to find data about either measure. He and Myles settled on using the number of questions in StackOverflow as a measure of the community and the number of projects in Github as a measure of the number of projects.

Notably, the ranks from the two measures were highly correlated. There were some strange outliers – Delphi had far more questions than projects, and Viml had far more projects than questions. And one commenter noted that JavaScript was probably overweighted because many projects in Github use JavaScript but aren’t really JavaScript projects. The data is interesting none-the-less. Conway speculates that StackOverflow questions may actually indicate opacity of a language and not community.

In addition to being correlated, the results were tiered:

  • Tier One: C#, Java, PHP, C++, Objective C, JavaScript, Python, C, Ruby, Perl and Shell
  • Tier Two: Scala, Actionscript, Haskell
  • Tier Three: R, Assembly, Visual Basic, Coldfusion, F#, Groovy, ASP, Clojure, Erlang, Lua, Scheme, Common Lisp, Emacs Lisp
  • Tier Four: D, Fortran, OCaml, Tcl, Go, Smalltalk, Viml
  • Tier Five: IO, Racket, VHDL, Verilog, Haxe, Objective-J, Boo, Vala, Coffeescript, Supercollider, Ada, Self, Arc, Eiffel, Factor, Nu, Sclang, PureData, Duby, ooc

klint finley