| Summary: | Functional purity is a fundamental part of the functional programming paradigm. A function is functionally pure if it is side-effect free and deterministic. Pure functions provide many benefits compared to impure ones, including guaranteed thread-safety as well as easier testing, debugging and maintenance. But how can functional purity be measured? This thesis develops a method for statically measuring the level of functional purity in any given C# program. It also investigates problems with determining purity in object-oriented languages, with a focus on C#. Moreover, a prototype of the method is implemented in order to evaluate the method using a benchmark consisting of 11 open source repositories that use C#'s [Pure] attribute. The [Pure] attribute can be placed in front of a method declaration to indicate that it is side-effect free. Due to a number of limitations to the implementation as well as to [Pure]'s definition of functional purity, which excludes determinism, the results of the evaluation appear relatively poor. After normalizing the implementation's classification distribution for each repository, its classification of pure functions has a precision of 65% and recall 17%, and its classification of impure functions has 54% precision and 69% recall. Nevertheless, the prototype still shows the potential of the full analysis method. A complete implementation of the analysis method could potentially yield a fully working system for measuring any C# program's level of functional purity.
|
|---|
