PROJECT: Technology Transfer from C/C++

Overview

This project is part of the requirements for the course COM3360 being taught by Prof. Karl Lieberher at Northeastern University. This course is taken through the NTU, National Technological University, program. The purpose of this project was to develop a Demeter/Java program from a C++ program and demonstrate whether there are benefits (less code, shorter development time, more reuseability) resulting from this conversion. This project helped in developing my own understanding of how to use Demeter/Java successfully in place of C++.

The program chosen was a software engineering project that was developed for another course, first in simple C and then in C++. The progression to Demeter/Java seemed a natural next step. Since Java and C++ are very similar in syntax and structure it was unnecessary to convert from C++ to Java code before implementing the Demeter/Java. This turned out to be a valid decision. Converting straight from C++ to Demeter/Java was fairly easy. Actually, the C++ program did not convert directly to Demeter/Java - I just used the logic of the original program and developed different methods of doing the same tasks using traversals and visitors.

Data is read in from three input files containing information about insurance agents, customers and what policies they had purchased and transactions to perform on this data. The format of the input files was specific and the files used are agent.input, insuree.input and trans.input. More details on the format of these files plus other information can be found in the README file for this project. The class dictionary is found in project.cd and the behavior code is found in project.beh.

Project Author

Karen Yahle email: kareny@lexis-nexis.com
Senior Software Engineer, Lexis Nexis, Dayton, Ohio.

Tasks

Use Cases

The use cases for the Insurance Database program.

Growth Plan

Start with simple input/output tasks, verify correctness and slowly build upon it, adding progressively more complex transactions one at a time.

Summary

The size of the code for Demeter/Java implementation was significantly less than for the "C" implementation and the "C++" implementation. This can be seen in the table below. Development time is estimated at about 25 hours and with more experience that time would probably have been cut in half, since each incremental change required recompilation and retesting, often with an error popping up. The strategies I used in the code of defining separate traversals and visitors for each different transaction could probably be refined by using inheritance and in-lining. The end product works as intended but feels a bit cumbersome - I think the code could be made more elegant and reuse implemented more often. Overall, though, I felt that the program lent itself well to the use of Demeter/Java. As mentioned, the original program required a number of parsing functions to be developed, which were taken care of by Demeter/Java for me. That was very nice!

Comparison