Reproducible Construction of Interconnected Technology Models for EMF Code Generation.

Abstract

Software technologies such as the Eclipse Modeling Framework (EMF ) involve complex usage scenarios that need to be understood or communicated by newcomers, developers, teachers, contributors, and others. Such different stakeholders consult scattered resources that provide textual explanations and code examples that cover different facets of a technology. More specifically, textual explanations can be located in i) developer literature that describes idiomatic usage of a technology, and in ii) scientific literature that explains technology from a formal and abstract perspective. Code examples can be located in iii) demo projects that exemplify intended usage, and in iv) wild projects that provide complex code examples in actual applications. In this paper, we propose annotated megamodels of technology usage as macroscopic summaries, i.e., technology models; these models interconnect the scattered textual explanations and code examples. We present a methodology for the incremental construction of technology models in a reproducible manner. It relies on the systematic reduction of a corpus to 'useful' evidence for each increment. The manual effort of searching for representative links is reduced by dedicated queries. We exemplify the application of the methodology for technology models that summarize EMF code generation.