Get two programmers in a room and the conversation, or argument soon turns to which language is best. No - that's not true you only need one programmer in a room...

So which language is best, most popular, etc?

The interesting thing is that there isn't much in the way of hard facts about who uses what for what. This of course allows the argument to run and run. Now two members of the Dataists website have invented a clever way of using public data to at least give an indication of how languages relate to one another in their real world use.

What they did was collect data on the number of StackOverflow questions that were tagged with the name of a language and on the number of projects using the language in Github.  The number of questions is hopefully a measure of how many people are using the language - but it could also be proportional to how difficult the language is. The number of projects is also a good measure of popularity, but it too has its problems - not all language users view gitHub with the same enthusiasm for example.

One way to check the goodness of the measures is to see how correlated they are.

The result?

The two measures are correlated at the 0.8 level and this means that they are mostly measuring the same thing. What is more interesting is that the regression plot shows that languages do fall into three categories - popular C#, Java, Python, Ruby etc; not so popular - Scheme, Visual Basic, Fortran and the fairly uncommon - Eiffel, Boo, Ada and so on.

Click for larger image - chart from dataists.

In the main the chart fits in with most people's (i.e. my own) prejudices about languages so it at least has face validity. On the other hand what is Visual Basic doing so low and next to assembler? Is this real? And why are Delphi and F# such outliers? I'm sure you can invent your own mystery questions.

For more information on how the survey was carried out, the raw data, comments and more visit:

http://www.dataists.com/2010/12/ranking-the-popularity-of-programming-langauges/