Generative Patterns

We use the concept of generative patterns to refer to patterns from which object-oriented programs can be automatically obtained. Two kinds of generative patterns are presented: propagation patterns and synchronization patterns, the former being the basis for automatic program generation, and the later being an extension for the purpose of concurrent programming. 1 What is a pattern? From the Webster dictionary, we get the following information: Word: pattern Cross references: model pattern: [...] something designed or used as a model for making things. [...] In the context of Software Engineering, we use this word to designatemodels for making objectoriented programs. These models, or patterns, exist in a programming level above object-orientation which we call adaptive programming. We call the patterns generative because they are used to generate the low level object-oriented code. Moreover, this generation follows precise rules and, therefore, can be automated by a compiler. A pattern consists of an algorithm and an applicability range. The applicability range de nes the contexts in which the generative pattern may be applied, and is determined by constraints such as paths in graphs, signature of operations and cardinality constraints. An adaptive program consists of a set of patterns with non-contradicting applicability ranges. Object-oriented programs are created from adaptive programs by a set of rules that de ne how to construct such programs from the patterns. The notation we use for our generative patterns is an extension of C++. The construction rules are given by a class dictionary (see Communications of the ACM [6]). Generative patterns are executable This research was supported in part by the National Science Foundation grant CCR-9102578 and CDA-9015692 (Research Instrumentation). Supported by the Portuguese Foundation for Research, Science and Technology (JNICT). and have a formal semantics ([5, 9]). Their main advantage to software development is the fact that they make object-oriented programs more reusable, in the sense that the algorithms have less dependencies on the speci c objects over which they operate. In addition, generative patterns are also useful as a formal language to express other kinds of patterns (namely, design patterns) de ned in the patterns community ([3, 4, 8, 2, 1]). 2 Propagation Patterns A propagation pattern de nes an operation at the adaptive programming level. The applicability range consists of a set of collaborating classes, and is speci ed by a succinct and generic subgraph specication. As an example, we present two propagation patterns in the context of Operating Systems. The rst propagation pattern returns the number of ready processes, and the second one removes the oldest ready process from the queue. We assume that the Operating System contains processes, and that these are organized in queues: OPERATION int ready_processes ( ) INIT {0;} TRAVERSE FROM OperatingSystem VIA Process TO Ready WRAPPER Ready PREFIX { return_value++; } OPERATION Process *Dispatch () TRAVERSE FROM OperatingSystem THROUGH -> *, ready_processes, * TO ReadyQueue WRAPPER ReadyQueue PREFIX { ret_value = this->get_first(); this->remove_first (); } These patterns can now be customized by speci c class dictionaries representing the Operating System structure. The following 3 class dictionaries could be used: