Models from code or code as a model?

Antonio Garcia-Dominguez, Dimitrios S. Kolovos

Research output: Unpublished contribution to conferenceUnpublished Conference Paperpeer-review


Many model-driven engineering workflows take the existing
code of a system as an input. Some of these include validation and verification,
software modernisation or knowledge extraction. Current approaches
take the code and produce a standalone model, instead of treating
the code itself as a model and navigating through it on demand. For
very large codebases this can be quite expensive to produce, and the
resulting model can be too large to suit the simplest monolithic filebased
stores. In this work we propose reusing a model that is already
incrementally built for us and supports fast lookups: the internal indices
maintained by our integrated development environments for code analysis
and refactoring. In particular, we show an Epsilon driver that exposes
the Eclipse JDT indices of one or more Java projects as a model, and
compare its performance and usability against MoDisco for validating
Java code with regards to a partial UML model.
Original languageEnglish
Publication statusPublished - 2 Oct 2016
Event16th International Workshop in OCL and Textual Modeling, - Saint Malo, France
Duration: 2 Oct 20162 Oct 2016


Workshop16th International Workshop in OCL and Textual Modeling,
Abbreviated titleOCL
CitySaint Malo
Internet address


Dive into the research topics of 'Models from code or code as a model?'. Together they form a unique fingerprint.

Cite this