On the server side we attempted to use different approaches to identify which methods exist in each commit. We attempted to do so by performing textual mining on the JSON file used as the definition of the web API as well as mining a C header file which also contains all the web API methods. Both approaches yielded no usable results. The JSON file is large (142 methods, 2221 LOC) and encompasses many different web API methods. It
also relies on a separate file where the data types are declared (1610 LOC). This makes manual inspection of these files across the many thousands of commits infeasible. Automatic diffing of this file in pairs of consecutive commits also resulted in many false positives which would also be too time consuming to analyze manually