A Data Set of OCL Expressions on GitHub

Abstract

In model driven engineering (MDE), meta-models are the central artifacts. As a complement, the Object Constraint Language (OCL) is a language used to express constraints and operations on meta-models. The Eclipse Modeling Framework (EMF) provides an implementation of OCL, enabling OCL annotated meta-models. Existing empirical studies of the OCL have been conducted on small collections of data. To facilitate empirical research into the OCL on a larger scale, we present the first publicly available data set of OCL expressions. The data set contains 9188 OCL expressions originating from 504 EMF meta-models in 245 systematically selected GitHub repositories. Both the original meta models and the generated abstract syntax trees are included, allowing for a variety of empirical studies of the OCL. To illustrate the applicability of this data set in practice, we performed three case studies

Empowering OCL research: a large-scale corpus of open-source data from GitHub

“…The current submission is based on and extends two of our earlier publications (Noten et al 2017a;Mengerink et al 2017a). In Noten et al (2017a) we have described the data collection process, presented the dataset of OCL expressions derived from the Ecore meta-models, replicated a study of Cadavid et al (2015) and evaluated the assumptions made by a previous work of Anastasakis et al (2008). In the follow-up work we have compared complexity of OCL expressions derived from open-source and industrial meta-models.…”

“…4 To ensure persistence and permanent identification of the data we have registered it with the 4TU federation through DataCite Netherlands (Noten et al 2017b). …”

“…The extended dataset, in addition to the scripts written to compose it, is publicly available on GitHub. 2 To ensure persistence and permanent identification of the data we have registered it with the 4TU federation through DataCite Netherlands (Noten et al 2017b). We show that, using the extended dataset, one is able to:…”

“…This work builds on our previously published dataset (Noten et al 2017a) of 9173 OCL expressions derived from 504 open-source meta-models, and earlier experimentation by Mengerink et al (2017a). We build upon both works by extending the dataset with model-totext transformations and feature several experiments that use the 94089 expressions derived from 2634 open-source model-to-text transformations available on GitHub.…”

“…provide a complementary perspective on the earlier results obtained through controlled experiments (Correa et al 2007;Reynoso et al 2006); 2. replicate corpus-based studies conducted in the context of traditional software engineering, in a MDE context; Our dataset (Noten et al 2017b) will also allow companies specializing in software quality to benchmark OCL expressions from systems under investigation against a larger collection of OCL expressions, and thus derive conclusions about the system quality (cf. similar work for non-MDE software Heitlager et al 2007;Oliveira et al 2014).…”