In modern software projects, developers often use a combination of multiple programming languages instead of a single language. This study will provide empirical evidence of the usage of programming languages on various open source software projects provided from SourceForge. The results will also be compared to the findings from a similar study, conducted with data from Github.

The analyzed projects were selected from all stable marked projects on SourceForge. Initially all projects were downloaded and then a smaller, diverse, sample of 439 projects has been selected to perform the final analyses. The projects have been investigated for a) statistical data regarding the number, type and size of the languages; b) associations between the number of languages and the covariates size, age, number of contributors and number of commits using a Poisson regression model, and c) association patterns found between the different languages using frequent item set mining.

The study found that a) projects consist of a mean of 5.8 different languages with ~75% written in the main GPL; b) the size and number of commits had a significant influence on the number of languages and c) there are clear association patterns regarding the usage of particular languages which can be grouped into three ecosystems based around the DSLs XML, HTML/CSS and Shell/Make.

It was also found that the results from Github and SourceForge are very similar even though the two platforms are focused on very different types of software projects.