A step toward code granulation space

Abstract

This paper proposes a code granulation space which can be applied to program construction and processing. Each granule in the space is defined as an improvement to the seed program of the space. The class tree of the seed program is partitioned into forests with context variables dependable. The boundary of granules can be specified according to both trees in the forest and shadow classes defined for improvement. Granules are organized in hierarchical style where ones at upper level are more generous than ones at low level. Granule similarity is also defined according to granule hierarchies. This Space can represent all improvements to given seed program, and can explain the fitness of programs as satisfying of granule's context within these programs. In this paper we describe details of the code granulation space, and show its effectiveness with examples.