It would be pretty interesting to analyze most popular lexems in code and compare them? Use clang diagnostics to watch only USER identifiers?