Model-driven engineering (MDE) is a new software development paradigm that intends to improve software construction by raising the abstraction level through the use of models and transformations. Recently, MDE has also been used for automatic change evolution and in particular for tailoring software processes. Even though there are some proposals for automatically generating part of the transformation models, they are not easily applicable in the software industry because the potential users usually do not have the capabilities of developing or maintaining transformations. Proposals trying to address this problem should balance the formality required by MDE and the usability needed by the users. This research applies higher-order transformations (HOTs) and decision models for generating transformations to tailor software process models with no direct user interaction with the code. To that end, a graphical interface was de ned with such a purpose, so that transformation rules are interactively de ned using a domain speci c languaje (DSL), and the tailoring transformation is automatically generated and applied using a domain speci c transformation language (DSTL).
Model-driven engineering (MDE) [23] looks forward to improving development
productivity and software quality by reducing the semantic gap between the
problem domain and its solution, using models and transformations that raise
the abstraction level of the managed concepts [
As a means to address such a problem, this research proposes the
automatic generation of tailoring transformations through the interactive de nition
of tailoring rules, that improves the usability and hides the complexity of MDE
components. In our research, a graphical environment allows rule de nition as
a variation decision model (VDM). We developed a HOT [30] that takes the
VDM as input and generates a tailoring transformation written in a DSTL (in
this case, a subset of ATL). De ning the transformation rules using our DSL is
easier than writing them directly in a transformation language such as ATL [
Tailoring transformations implementation is di cult because it requires high expertise about MDE concepts and also about the application domain, and these two kinds of knowledge are almost never mastered by the same person. MDE knowledge can be obtained through training, but highly skilled people are still required, and these people are not generally in charge of the process in SSEs.
MDE solutions are powerful and suitable for di erent application domains, but they are almost always complex. There are social and organizational factors that threaten industrial adoption [16] (e.g., technological factors, resistance to change). Probably, if we could lower complexity, then adoption would be higher. In this sense, if SSEs would like to reuse their processes through model-based tailoring by de ning and applying the transformation rules, they should not require a direct interaction with the source code of any model transformation. 3
There have been several proposals to deal with the di culty of writing
transformations. Some high abstraction level transformation languages speci cally
created for this purpose have been proposed, such as ATL [
Varro and Balogh, through the VIATRA framework [31], provide an integrated environment for building and executing transformations at an even higher abstraction level. It includes its own language that intends to hide the complexity of transformations and is supported by Eclipse. However it does not provide an easy-to-use environment for process engineers.
Provided that transformations can also be considered as models [
The Atlas Model Weaver (AMW) [
There are other MDE proposals that intend to address usability. MOLA [17]
and GREaT [
There are some recent proposals such as MTBE (Model Transformations By Example) [33] and MTBD (Model Transformation By Demonstration) [28] that present innovative solutions for simplifying the implementation of model transformations using visual strategies and patterns. These strategies generate part of the code of the model transformations; however, it is still needed to complete such code. Therefore, this represents a semi-automatic approach to generate model transformations that is still not enough for process engineers. 4
The proposed solution consists of applying diverse MDE concepts to allow automatic transformation generation in a usable way for nal users. The generated transformations should be expressive enough for its purpose, i.e. for tailoring software process, the interactive de nition of transformations should be usable for target users, i.e. process engineers, and the complete generation process should be transparent, and no direct interaction with the code should be necessary. This proposal focuses particularly on the need of SSEs, which have to adapt their software processes for e cient development, but they generally do not count on highly specialized people for manually tailoring their processes. The proposed solution is called Architect of Tailoring Rules (ATR). ATR is a model-based tool that allows process engineers to interactively de ne the transformation rules using a graphical environment. The output of the tool is the tailoring transformation that can be used to adapt the organizational software process.
The proposed solution has four major components: { The graphical environment rules de nition requires two source models. The rst input is the organizational software process model de ned by the SSE for guiding its software developments (component 1 in Fig. 1). This model contains all activities, roles, work products and tasks involved in the software process. In particular, it identi es the variable process elements whose inclusion is de ned by the rules [27]. The second input is the organizational projects context model (component 2 in Fig. 1), where the attributes that characterize projects along with their set of potential values are de ned. The graphical environment itself (component 3 in Fig. 1) includes two elements: an interactive rule de nition interface for de ning transformation rules and the Text-to-Model transformation for generating a VDM. { The variation decision model (VDM) for formally representing transformation rules. We have de ned the VDM using our DSL (component 4 in Fig. 1) that is a high-level representation of the transformation rules using an abstract syntax (metamodel) and a concrete syntax (textual representation). The VDM is inspired by Decision Models [32] and Semantics of Business Vocabulary and Business Rules (SMVB) used for building decision rules [19].
In this work, we also de ned a Variation Decision Metamodel (VDMM). { The tailoring transformation generation (HOT) (component 5 in Fig. 1).
This HOT has one input model, the VDM, and follows a transformation synthesis pattern for generating the tailoring transformation model, that conforms to the ATL metamodel. This HOT is an exogenous transformation because the models are expressed using di erent languages (decision and transformation models). Then, we apply an ATL extractor for generating the tailoring transformation. In this work, we applied a vertical transformation because the source and target model of the HOT reside at di erent abstraction levels. { The tailoring transformation. After applying the HOT, we obtain a tailoring transformation (component 6 in Fig. 1) using an ATL subset. The DSTL uses only some structures of the ATL language that are necessary in this domain (e.g., lazy rules and called rules are not used). The tailoring transformation is generated only once and it can be used for tailoring all concrete contexts. The adapted software process model (component 8 in Fig. 1) is the output of applying such transformation using the organizational software process and the concrete project context as input. Such a process includes the process elements that are needed for the speci ed context.
This research has two speci c goals: (1) automatically generating tailoring transformations using our DSL, and (2) applying the generated tailoring transformation for obtaining an adapted software processes. The results of this work will be applied in organizational software process of Chilean SSE, so that they can be adapted to di erent project contexts, and thus the development productivity and product quality are improved. The expressiveness of the DSTL and the usability of the graphical environment will be rigorously evaluated. 5
We have studied several alternatives for implementing the preliminary solution of the proposal, and we have analyzed the feasibility and the e ort required to implement the ATR components.
The rst step was to explore the feasibility to implement HOTs using two input models. We have successfully addressed this issue, in our case the organizational software process model and the organizational context model. We tried using some of the most promising techniques and tools for developing HOTs such as AMW and MOFScript but we found limitations for dealing with two input models [25]. We then decided to divide the HOT in two parts: rule de nition and transformation generation. We de ned one input model for the HOT and we called it VDM. We found researches about business rules [19] and decision models [32], we used them as inspiration and we formalized the DSL for our application. The VDM is a formal representation of transformation rules.
The second step consisted of building a proof of concept and an initial implementation of the ATR. For this step, we were able to build the infrastructure of the ATR and apply it through a running example. The initial ATR implementation has a graphical interface, a preliminary VDM (Fig. 2), a HOT and an initial DSTL. For the rst implementation of ATR, we de ned the models using Eclipse Modeling Tools (EMT) and implemented the transformations using Java. The running example was used to generate a tailoring transformation for a Chilean SSE [26]. The output of the Model-Based tool is an ATL le that is compilable in EMT for software process tailoring.
The third step encapsulated ATR in a megamodel application. We have
proposed a megamodel to improve the adoption and evolution of the MDE-based
tailoring approach in industry [
We have applied ATR to an exploratory case study in a Chilean company, and we will apply a con rmatory multi-case study in two or three other companies.
The exploratory case study addressed the evaluation of correctness and
productivity. We compared the productivity of adapted software processes obtained
by using two di erent approaches: template-based tailoring [
In the multi-case study, we will incorporate the evaluation of usability and
expressiveness. We will evaluate the usability by applying structured
questionnaires to process engineers [
We expect that SSEs will improve their competitiveness using the proposed approach because they will be able to apply software processes that are specifically adjusted to the needs of their particular projects. Using the solution will probably also enhance the quality of the software products, although this evaluation is out of the scope of this work.
Acknowledgments. This work has been partly funded by Project Fondef GEMS IT13I20010 and Luis Silvestre was also supported by PhD Scholarship Program of Conicyt, Chile (CONICYT-PCHA/2013-63130130). 13. J. Gray, Y. Lin, and J. Zhang. Automating change evolution in model-driven engineering. Computer, 39(2):51{58, 2006. 14. J. A. Hurtado, M. C. Bastarrica, A. Quispe, and S. F. Ochoa. An MDE approach to software process tailoring. In D. Ra o, D. Pfahl, and L. Zhang, editors, ICSSP, pages 43{52, Honolulu, HI, USA, 2011. ACM. 15. J. A. Hurtado, M. C. Bastarrica, A. Quispe, and S. F. Ochoa. MDE-based process tailoring strategy. Journal of Software: Evolution and Process, 26(4):386{403, 2014. 16. J. Hutchinson, J. Whittle, and M. Rounce eld. Model-driven engineering practices in industry: Social, organizational and managerial factors that lead to success or failure. Science of Computer Programming, 89:144{161, 2014. 17. A. Kalnins, J. Barzdins, and E. Celms. Model Transformation Language MOLA.
In U. A mann, M. Aksit, and A. Rensink, editors, MDAFA, volume 3599 of LNCS, pages 62{76, Twente, The Netherlands, 2004. Springer. 18. M. Kuhrmann. You can't tailor what you haven't modeled. In H. Zhang, L. Huang, and I. Richardson, editors, ICSSP, pages 189{190. ACM, 2014. 19. OMG. Semantics of Business Vocabulary and Business Rules (SBVR) Version 1.0.
Technical Report 2008-04-01, 2008. http://www.omg.org/spec/SBVR/1.0/. 20. OMG. Software & Systems Process Engineering Metamodel Speci cation (SPEM).
Technical Report 2008-04-01, 2008. http://www.omg.org/spec/SPEM/2.0/PDF/. 21. OMG. Meta Object Facility (MOF) 2.0 Query/View/Transformation Speci cation.
Technical report, January 2011. http://www.omg.org/spec/QVT/1.1/PDF/. 22. H. Padda. QUIM: A Model for Usability/Quality in Use Measurement. LAP
Lambert Academic Publishing, Germany, 2009. 23. D. C. Schmidt. Guest Editor's Introduction: Model-Driven Engineering. IEEE
Computer, 39(2):25{31, 2006. 24. C. B. Seaman. Qualitative Methods in Empirical Studies of Software Engineering.
IEEE Trans. Software Eng., 25(4):557{572, 1999. 25. L. Silvestre, M. C. Bastarrica, and S. F. Ochoa. HOTs for Generating Transformations with Two Input Models. In proc. SCCC, 2013. 26. L. Silvestre, M. C. Bastarrica, and S. F. Ochoa. A model-based tool for generating software process model tailoring transformations. In L. F. Pires, S. Hammoudi, and J. Filipe, editors, MODELSWARD, pages 533{540. SciTePress, 2014. 27. J. Simmonds, M. Bastarrica, L. Silvestre, and A. Quispe. Variability in software process models: Requirements for adoption in industrial settings. In PLEASE, pages 33{36, May 2013. 28. Y. Sun, J. White, and J. Gray. Model Transformation by Demonstration. In A. Schurr and B. Selic, editors, MoDELS, volume 5795 of LNCS, pages 712{726, Denver, CO, USA,, 2009. Springer. 29. M. Tisi, J. Cabot, and F. Jouault. Improving Higher-Order Transformations Support in ATL. In ICMT, volume 6142 of LNCS, pages 215{229, Malaga, Spain, 2010. Springer. 30. M. Tisi, F. Jouault, P. Fraternali, S. Ceri, and J. Bezivin. On the Use of HigherOrder Model Transformations. In ECMDA-FA, volume 5562 of LNCS, pages 18{33, Enschede, The Netherlands, 2009. Springer. 31. D. Varro and A. Balogh. The model transformation language of the VIATRA2 framework. Science of Computer Programming, 68(3):214{234, 2007. 32. D. Weiss, J. Li, H. Slye, T. Dinh-Trong, and S. Hongyu. Decision-Model-Based
Code Generation for SPLE. In SPLC, pages 129{138, Sept 2008. 33. M. Wimmer, M. Strommer, H. Kargl, and G. Kramler. Towards Model Transformation Generation By-Example. In HICSS, page 285, Big Island, HI, USA, 2007. IEEE Computer.