Empirical investigation of COBOL features

摘要

This study investigated programming activity in COBOL. Attempts were made to identify problem areas so that improvements can be made in COBOL compilers and in the manner in which COBOL is taught. Identification of problem areas was achieved through examining program changes made by student programmers during the development of four different programs written in COBOL. The data, which was collected from a COBOL course at Purdue University, consisted of all versions of all programs submitted for compilation by each student. Thus, the data represented a complete history of each subject's program development process beginning with the initial version compiled and ending with the final version submitted for grading. All program changes made between two successive versions were classified into four categories: COBOL-related, algorithmic, cosmetic and report-generation-related. A significant number of changes are related to report generation which suggests a need for support in this area. COBOL-related changes were delineated into 104 error categories which suggested that there are several problem areas in COBOL. Finally, the four categories of program changes were observed with respect to various points in the program development process. Most COBOL-related changes occur before the midpoint of the development process whereas most cosmetic changes occur late in the process.