Some fundamentals of meta-modelling with applications to measurement of software artefact reuse

摘要

This thesis proposes a meta-model based framework to investigate measurement ofudthe amount of reuse for a variety of software models. The framework was assessedudusing a prototype tool.udThe evaluation of the framework consisted of four phases that constitute theudexperiments conducted in the study. The phases are:ud• Software Model Type Classification. The framework is assessed to see if it canudclassify different kinds of software models.ud• Software Model Classification. The framework is assessed to see if it can classifyudand measure size related to the amount of reuse for different software models.ud• Measurement Testing. The framework is assessed to see if it can measure theudamount of reuse using a series of software models as test cases.ud• Automation assessment. The framework is assessed to identify the limits of itsudapplication for measurement of the amount of reuse.udFindings from this study indicated that most software model types selected could beudmeasured for the amount of reuse for different software models, provided that eachudsoftware model type is classified prior to measurement of reuse. This did not requireudeither additional programming or different software components in the prototype tooludfor different software models. Nor did the framework itself have to be modified atudmeta-level two, three, or four. Measurement of the amount of reuse using generationudprovided a reasonably accurate reflection of the actual contribution made using thisudreuse approach. Importing of software models using data entry and imported text file did work butudwas labour intensive. To automate this, tools specific to each software model typeudneed to be made to translate the software model into the text file format defined byudthe prototype tool. Measurement of reuse using composition for analysis and designudmodels was more realistic in its measurement than for source code models,udparticularly if the CASE tool used an import/export approach to the reuse of softwareudmodels. Measurement of the amount of reuse using composition for source code wasudinaccurate for practical purposes.udRecommendations for further research include the need to streamline theuddevelopment of tools for translation of software models, refine the framework toudinclude the concept of references for measuring internal reuse using composition, andudrefine the measurement of reuse for source code and repository-based reuse in CASEudtechnology.