REUSE AND REVERSE ENGINEERING FOR DATA-INTENSIVE BUSINESS PROGRAMS

Abstract

The development and maintenance of data-intensive business systems still account for a large proportion of the world's software investment. In this thesis, the characteristics of data-intensive business programs are studied. Basing on the characteristics, an object-oriented architecture, called the functional program model, is proposed as an unified architecture for the specification and representation of data-intensive business programs throughout their entire development lifecycle from the conceptual design stage to the implementation design stage. The model specifies/represents a program by composing program factors which provide a set of tightly coupled functionalities each. With the use of the model, approaches are proposed to address some current problems in software reuse and reverse engineering for data-intensive business programs. Similar program factors are repeatedly designed and developed for the design and development of data-intensive business programs, especially for business transactions. The main reusable factors for data-intensive business programs can be classified into four types: parameterized computation factor(e.g., compute average), parameterized database factor (e.g., commonly used database retrievals), user interface, and enforcement of constraint in factor interface. The main problem is that current techniques have difficulty in using reusable factors to design a program. In order to use a reusable factor to design a program, details which are not intrinsic properties of the factor are required to be added into the factor. This affects the reusability of the factor adversely. Using the functional program model and the reuse techniques, parameterization and generation, a reuse-oriented approach is proposed for the design and development of data-intensive business programs to address the problem. In the proposed approach, reusable factors are used to design a program without the need of adding details which are not their intrinsic properties into them. As a result, in the design of any program including any business transaction, commonly used parameterized computation factors and parameterized database factors are reused through generation (for simple cases) or library approach. Most of the user interfaces are generated based on the data structures of the inputs and outputs involved. The enforcement of the constraints in factor interfaces is generated based on the types of the constraints. Recovery of designs and reusable factors is an important issue in reverse engineering. Input/output statements provide the external interfaces in data-intensive business programs, however, most of the approaches proposed to address this issue do not put much emphasis on these statements. In this thesis, an approach is proposed for the recovery of object-oriented designs and reusable factors from procedural data-intensive business programs. The recovery is through transforming a program into a representation built on the functional program model to represent the program. The representation is composed from program factors which provide a set of tightly coupled functionalities each. Most of the time, the input/output statements which perform the same type of operation (read, write, rewrite and delete) on the same record type in a data-intensive business program, perform a functionality of the program. The approach identifies the functionalities of a program based on this characteristic. Reusable factors are extracted from the program based on the above mentioned types of main reusable factors. Majority of the tasks in the approach can be automated. Algorithms for these tasks are presented.