Constant Values Distribution Investigation in the C Programs Source Code

Currently, software engineering plays a key role in software development, one of the criteria for the development of which is the investigation of its factology and various scientific and practical patterns. An important aspect of this area is the logic of program execution, operating with internal data, and, in particular, constant values, the identification of patterns in which actualizes this research. The main applications of this pattern include obtaining fundamental knowledge about algorithms, creating new and expanding existing metrics for evaluating and comparing program code, developing methods for its optimization, using it in genetic programming, etc. The purpose of this article is to obtain the frequency distribution of constant values in the source code of programs in the C programming language. The essence of the presented approach is to create a method for statistical analysis of the text of the source codes of programs contained in the ExeBench dataset (which consists of a huge amount of source code of functions in the C programming language, their assembler code for various processor architectures, compilation errors and other information). The proposed method is based on the use of algorithms for lexical and syntactic analysis of source code functions, semantic definition of constant types, and conversion of the recording of programming language symbols into the corresponding numeric or string values. The method has an implementation in the form of a software tool in the Python programming language, given in the form of an intuitive pseudocode. Experiments using this prototype allowed us to obtain the desired distribution of constant values for the source code of programs in the C programming language. Analysis of the obtained results allowed us to make a number of important theoretical and practical conclusions regarding the most frequently used constants, the correspondence of the obtained distribution to the Zipf law and its proximity to the exponential function, the anomalous appearance of a number of constants in the Top 50, etc. The scientific novelty of the proposed approach lies in the fact that the distribution of constant values for the source code of programs in the C programming language is obtained for the first time. The theoretical significance consists in obtaining new fundamental knowledge regarding the features and patterns of source code constructions, which can be extended to other programming languages. The practical significance consists in applying the distribution to a wide range of tasks, including the author's genetic reverse engineering, which in itself is a qualitatively new direction.