PEREIRA, S. A.; PEREIRA, Saulo de Araújo.
Résumé:
The reasoned use of metrics (measurement standards) is essential for the execution of measurable software projects whose development can be predicted, monitored and improved. Among the metrics for the coding phase presented in the literature, we highlight the cyclomatic number and Software Science metrics, for their theoretical foundation and empirical support, and the code line metrics and number of routines, for their popularity and simplicity. . This work verifies the validity of these metrics applied to the Pascal and C languages. For this, a tool for the automatic obtaining of the metrics is developed. The analyzes, performed on a sample of 1,670 routines, use regression techniques and are based on five statistical parameters. Some of the metrics studied are valid for program size estimation (N and Nj length estimator metrics) and program development time estimation (Software Science length and volume metrics, code lines, and cyclomatic number). . Metrics for estimating the size and timing of routine development and metrics for other applications, such as language level assessment and routine information content, are inaccurate or inconsistent.