The use of Model-Driven Engineering in software development is increasingly growing in industrial applications as the technologies are becoming more mature. In particular, domain-speci c languages bring to end-users simplicity of use and productivity by means of various artifacts generators. However, end-users still need to cope with heavy modeling infrastructures and complex deployment procedures, before being able to work on models. In this paper, we propose a centralized lightweight approach for performing textual modeling through web browsers. DSLFORGE is a generator of online text editors. Given a language grammar, the tool allows to generate lightweight web editors, supporting syntax highlighting, syntax validation, scoping, and code completion. DSLFORGE allows also automatic integration of existing code generators into the generated web editor providing a complete online modeling user experience. Demo: http://youtu.be/KN6cneWhhKY
Domain-Speci c Languages (DSL) [
However, end-users still need to cope with heavy development
infrastructures and complex deployment procedures when it comes to using modeling
tools. This is because most of the existing tools are based on general purpose
Integrated Development Environments (IDEs) such as Eclipse or MPS, or
standalone proprietary applications as MetaEdit+ [
In this paper, we present the DSLFORGE tool, as support of a lightweight centralized approach that allows end-users to edit and process textual models through web browsers. The paper is organized as follows: Section 2 describes current web-based approaches and Section 3 presents the DSLFORGE tool, its methodology of use, two functional examples and its internal architecture. 2
There are still technological challenges to achieve modeling on the web.
Nevertheless, some initiatives targeting graphical languages are worth to notice.
GenMyModel [
Few tools provide exible means to de ne textual languages and they come
with in-house formalisms. For example, in Concrete Editor [
DSLFORGE is a framework for the generation of web textual DSL editors. The generated editors are packaged into workbench web applications which let users create, edit, and launch transformations from models online, making it possible to work simultaneously with partners or colleagues on the same resources. These online editors are also easily customizable and extensible. DSLFORGE is proposed to two main categories of users: (i) DSL developers, and (ii) DSL users. The former is given a technology to build and publish online editors on the web. The latter uses the DSL editor through lightweight applications (enterprise intranet, mobile devices, tablets, etc.). 3.1
The framework is packaged into two features: Tooling and Runtime. DSL developers use the Tooling which contains all the needed components to generate and deploy the editor on application servers. The steps below are followed iteratively by DSL developers to get an operational online editor: 1. Design/enhance the DSL 2. If applicable, design/enhance transformations to generate some kind of artifacts from DSL instances 3. Automatically generate from the grammar the RCP and the RAP editor 4. If needed, enhance the web editor generated code. Third party plugins could provide extra functionalities. 5. Automatically package and deploy the editor.
DSL users use the deployed editor to edit and process models online, namely they can: create, edit and save models, trigger code generation from a model, and execute other actions or functionalities if the web editor was extended by third party plugins. 3.2
Martin Fowler's state machine example provided by Xtext. The example comes with a grammar allowing textual speci cation of state machines. Our objective is to provide a web editor that allows modeling state machines and generating java test classes from these state machines on the server using the browser. An extra functionality should be integrated for compiling and executing the test classes through the browser too.
Using DSLFORGE, as shown in Figure 1, we select the Xtext grammar to generate its online editor. At this moment we are able to select the existing transformations that will be available online for end-users. In this case, we select the existing state-machine to java transformation. The online editor is automatically packaged into a web application, together with a workspace navigator.
To execute the web application we launch the browser. Figure 2 shows how the editor handles syntax validation, content assist and how the web application is enhanced with a code generation action contributed to the Tools menu. Semantic highlighting, scoping, history management (undo/redo), key bindings and folding are also supported by default. In addition, a custom action is integrated to the web application for compiling and executing Java classes on the server, to show an example of the online editor extensibility.
A web editor for the speci cation of conference websites: As a second example, we use DSLFORGE in the context of a conference DSL. This DSL allows end-users, who are not necessarily HTML/JavaScript experts, to generate and update on the server the conference or workshop website. This way, they can update the conference website wherever they are and without knowing the website technical details. 3.3
The Tooling feature is shipped with Xtext, EMF, and ANTLR development features, and contains the editor generator and other contributed plugins which may be integrated with the editor into online workbenches (e.g. workspace navigator, authentication, le uploader). The Runtime contains the standard RAP target platform together with EMF target components, Xtext runtime, and additional plugins used during the execution of the editor. The editor can be integrated with any third party RAP-compatible plugin.
DSLFORGE follows the RAP standard architecture and lifecycle. Indeed, each generated editor widget has its counterpart in JavaScript. Resource processing is dispatched between the server and the client. The communication between the client and the server uses the JSON format. Multi-threading is used on the client side to handle the computationally expensive resource processing. Indeed, dedicated workers parse the user input and feed-back the UI with annotations. Shared workers are used to manage the global index which maintains available the cross references to all editors opened within the same user session. On the server side, parsing and validation is triggered when saving the resource. Asynchronous server pushes are used to notify the client about workspace change events and server-side resource validation. The generated web applications have been successfully deployed on two application servers (Tomcat and Jetty). Future work will include securing the communication between the client and the server and enhancing the management of resource access privileges.