Model-Driven Development (MDD) is a term used to describe the systematic use of software abstractions (models) as primary artifacts during a Software Engineering (SE) process. However, most of the exiting MDD methods are dificult to apply, which leads to no wide use of MDD in industry. This project aims to define an agile method to deal with MDD from requirements to code. By agile, we mean three characteristics: 1) Modelers can involve end-users in the development; (2) End-users can validate small parts of the system (Sprints); (3) Code implementation is free of errors since it depends on automatic transformations. The project aims to define a method to transform strategy models into business models, a method to generate business process models from strategies, and a method to generate code from business processes. The project also aims to ensure the understandability of the models and automate the transformation process as much as possible. The project is scheduled for 3 years and started in September 2022. It is financed by the Spanish Ministry of Science and Innovation.
Agility [1] has become the most important attribute for organisations to embrace a continuously
changing environment. Initially proposed for aligning software projects with business
requirements, currently it is also considered for defining an adaptive business strategy in response to
market changes. Agile practices and Model-Driven Development [2] are aligned in terms of their
emphasis on integrating business people and developers, quick feedback loops on requirement
changes, frequent software delivery, reducing development efort, among many others [
3]. Up
to now, agile development has been mainly undertaken with traditional programming and not
with conceptual model programming. Some of the reasons that appear in previous works to
justify the poor adoption of MDD in real developments are the lack of integration [4] [5] [6] (
The current state shows that there are existing holistic MDD methods that allow to generate code from models at diferent abstraction layers. Shamsujjoha et al. [ 7] have conducted a systematic literature review to investigate how MDD techniques are used in the development of mobile apps. The main benefits of MDD in these developments are an enhancement in productivity, scalability, and reliability. The authors also identify some challenges such as poor graphical user interface generation and unsupported requirements from the early stages of the software development process. Ameller et al. [ 8] proposed a survey about how existing MDD methods deal with non-functional requirements. The survey is based on interviews done in 18 diferent companies from 6 European countries. Results yield that those practitioners perceive MDD adoption as a complex process with little to no tool support for non-functional requirements. When non-functional requirements are not supported, the generated code is changed manually. Magalhaes et al. [9] proposed a framework to support the development of model transformation chains using MDD. The framework is comprised of a process to guide on transformation development tasks.
Taking these works into consideration, there is a precise motivation in this project that aims to
go far beyond optimizing the MDD paradigm from an agile development perspective. Our vision
is that an MDD approach must ofer a simple yet complete modelling language and modelling
process to integrate business strategy, business process, and information system domains into
the development process, as well as integration mechanism between diferent domains enabled
by semantically meaning connections and as automatic as possible transformations. We aim to
automate as much as possible model-to-model and model-to-code transformations from a holistic
point of view. All system features must be represented in the model. This way, developers can
be seen as modelers where no code on a programming language has to be implemented, making
true the statement that “the model is the code” (instead of the conventional one ”the code is
the model”). This is aligned with the concept of agile in next terms: (
The hypothesis we want to contrast in this project is: “It is possible to build a holistic method from an agile point of view in such a way that we can generate a system from requirements to code automatizing the process as much as possible”.
In order to answer this hypothesis, the project has to face 4 challenges.
Challenge 1 is the integration of business requirements with an agile approach into a
holistic model-driven development process. Organisational goals and the high-level definitions to
achieve them have a great impact on business processes and over the requirements and design
of the information systems that support the organisation’s operation. Business strategy deals
with what organisations define as their ends and the means to achieve them. Capturing this
knowledge using conceptual models has been addressed mainly by enterprise architecture (EA)
(https://pubs.opengroup.org/architecture/archimate2-doc/), goal modelling (GM) frameworks
[10], and i* [11], as an intent to align information systems with the strategical definitions
that motivate their design and evolution. Even though the use of these conceptual models is
significant, there are some issues that make dificult its use in an automatic way. Some of these
problems are: (
Challenge 2 is to ensure the quality of the models since modelling is a complex activity not free of errors. An important quality criterion is understandability, that can be defined as how easy is to understand a model by its stakeholders and any readers. A factor is defined as a condition or characteristic that actively contributes to the quality of the software. So, a factor is a characteristic that can afect -positively or negatively- the models understandability. It is important to identify such factors to know what elements of the models afect the understandability.
Challenge 3 is the design of usable end-user interfaces, since existing holistic MDD methods lack of models to represent interface details to get customizable interfaces. Even though there are MDD tools that face this issue (such as WebRatio (https://www.webratio.com/) or INTEGRANOVA M.E.S. (Model Execution System) (https://www.integranova.com/)), the features to adapt interfaces to end-users’ preferences is very tight. Moreover, there is no automatic process to generate usable interfaces from requirements models. The interfaces are usually derived from requirements models manually, without any kind of procedure, only relying on the analyst experience. This means that the efort made in building the requirements model is not useful when designing the interfaces at the end. In an agile development that aims to reduce the analyst efort, GUI generation must be part of the holistic development process.
Challenge 4 is the automated software testing since we must ensure that every change in the models do not imply errors in previously specified system features. The continuous testing is a characteristic of the agile development. Only tested functionalities should be included in production software. The process to design and run tests cases should be done as automatically as possible considering the business process that guide the interactions and the robustness of the generated software products.
This project has the code SREC (PID2021-123824OB-I00) and is financed by the Spanish Ministry of Science and Innovation. It started in September 2022 and finishes in September
The main objective of the project is the design of a MDD method for the development of strategically aligned information systems from an agile perspective. We aim to help the stakeholders of the method – agile software development teams - to focus on the design of information systems in consistency with constantly changing business strategies and process by: 1) maintaining traceability and consistency from organisational strategy to information system code through a holistic model-driven software development method, 2) enforcing the understanding of the business, processes, and systems via conceptual models by improving model comprehension using an automatic model understandability assessing method, 3) designing user interfaces which are usable and compliant with the system requirements without focusing on the constantly changing front-end technologies using the automatic generation of interaction models from business process models, and 4) avoiding exhaustive tasks of programming tests cases through the design and generation of executable test cases from the information system’s conceptual model. With these contributions, we aim to reduce developers’ efort, focusing all the development on building conceptual models that abstractly represent the system, and ensuring full traceability between organizational goals and their software representation. In Figure 1, we depict the project’s specific goals, which are detailed below.
G1. Design of an Agile Business Strategy Modelling Method. This goal deals with
designing a business strategy modelling method, that considers: (
G2. Design of an MDD Method to Support Agile Software Development . An agile
implementation of business strategy requires an agile method for designing business processes
and the information systems that support their operation.,This goal addresses the design of
a business process modelling method that supports the requirements specification and the
automatic generation of the conceptual model that represents the information system. The
method deals with (
G3. Automatic Assessing of Understandability in Conceptual Models. The model
understandability is a relevant criterion to measure the quality of any conceptual model. This
criterion is afected by multiple factors: model factors and personal factors. However, existing
studies about model understandability do not deal with all possible factors, which prevents
providing a larger explanatory power about how to enhance model understandability. This goal
focuses on designing a method to assess the understandability in conceptual models by applying
machine learning techniques. We focus on supervised learning techniques like classification
and regression in order to explore data and to elaborate predictive modelling. The method
provides: (
G4. User Interaction Generation The efort spent in building business models must be used to generate Graphical User Interfaces (GUI) automatically. We propose identifying transformation rules from business model patterns in such a way that we can generalize these transformations for any model. In case one business model primitive can generate more than one widget in the GUI, we can add stereotypes at the business model to express unequivocally what widget must be generated to support that part of the model. The approach must also suggest what stereotypes may improve the usability according to standard usability guidelines.
G5. Automatic Test Case Generation. We aim to automate the testing process as much
as possible through three lines: (
Participants in this project are researchers of the Universitat Politècnica de València within the Valencian Research Institute for Artificial Intelligence. Roles involved in this project are: method engineering, requirements engineering and empirical software engineering. The two ifrst authors of the paper are experts at requirements and model transformations. The last two authors of the paper are responsible of the project; they are experts at conceptual modelling and experiment design.
The results of the project will be made available to the community through research journals and conferences, open-source tools, and open repositories such as Zenodo. The main outcome of this project is the design of an agile method to develop systems from business strategy models to code. The first outcome of this project is the design of an agile business strategy modelling method. This outcome aims to design a light-weight modelling method for representing business strategy in a systematic way. The method will provide a language suitable for business strategy, a modelling procedure based on business agility practices, and guidelines for mapping strategic information into business process models.
The second outcome of the project is the design of an MDD method for supporting agile software development. We propose a business process modelling method that supports the specification of requirements for the information system model, and the automatic generation of the conceptual model of the information system. In a last stage, we propose the automatic code generation from the conceptual models including its GUIs.
The third outcome is the definition of a method to ensure the quality assessment throughout the whole process from strategy to code. This outcome aims to assess models and generated systems quality. For assessing models, an assisted modelling method for improving understandability is designed. For assessing generated information system quality, test cases will be automatically designed from business and interaction models and run, resulting as a report with the failures and the coverage of testing process. This report is a fundamental input for the continuous improvement of the system.
The fourth outcome is the report of the diferent validations conducted in the proposed methods. Applying a technical action research approach, we will co-design and execute with industrial partners the methodology of the case studies for the validation of the whole MDD method. With industrial partners that are software developers, we will deploy SREC method to analyse pros and cons. We will use the industrial partners that works on clinical and genomic ifeld to test the applicability in a real software development focused on a specific domain.
The project started in September 2022, so at the moment we have conducted 8 months. Next, we list the tasks done completely.
1. Analysis of challenges and gaps for integrating business strategy into MDD methods.
This study has identified : (
This project shares several topics of the CAiSE conference, such as: Conceptual modeling, languages and design; Requirements engineering; Process modeling, analysis and improvement; and Decision models and business intelligence. The proposal of a strategy model requires the use of requirements elicitation techniques. How this strategy model is draw depends on conceptual models. Moreover, how to transform these strategy models to business process models involves model to model transformations, also related with CAiSE. Drawing the business models requires process modeling and business intelligence.
The result of SREC will help in the process of requirements elicitation, simplifying the method to represent abstractly business processes. Our approach aims to guide the interviews with end-users to capture requirements fluently. Once requirements elicitation is over, all the efort spent on building the business model results in analysis and design models that automatically can generate fully functional systems. So, starting from requirements models we can generate quick code, which results in an agile development where end-users can validate the generation. Improvements to the system can be made just modifying conceptual models and regenerating the code quickly. Another advantage for analysts is that our approach suggests understandability improvements while the models are built. We plan to disseminate the results in the main track of the next CAiSE conferences. This work was supported by the Spanish State Research Agency through SREC (PID2021123824OB-I00) project, and by the Generalitat Valenciana through the CoMoDiD project (CIPROM/2021/023), and co-financed with Santiago Grisolía fellowship (GRISOLIAP/2020/096).