Predicting Source Code Effectiveness of Prediction based Source Code Auto Completion

Auto Completion is the facility provided by most modern Integrated Development Environments and source code editors for word completion when editing source code. All auto completion mechanisms that we know of use syntactic knowledge of the programming language to provide this feature. We investigate the use of programming language agnostic prediction models to provide auto completion. We implement prediction based auto completion by using a variant of the popular compression scheme called Prediction by Partial Matching. We compare the results to an implementation of Syntax based Auto Completion. The results include a measurement of Key Strokes Per Character as well as subjective evaluation. We find that the prediction based auto completion performs remarkably well and consistently performs better than the syntax based completion method implemented.