Reference attributed grammars (RAGs) extend Knuth's attribute grammars with references. These references can be used to extend the abstract syntax tree to a graph. We investigate how RAGs can be used for implementing tools for visual languages. Programs in those languages can often be expressed as graphs.
For certain applications domain-speci c languages (DSLs) play an important
role, since they can make it easier to solve problems in the domain in a more
concise way than a general purpose language (GPL) [
A DSL user may want di erent tools, such as a batch compiler and an
integrated development environment (IDE). Such tools often share analysis, and
reusing speci cation between them can reduce the implementation e ort and
help to keep the tools consistent with each other. One tool aiming for reuse of
language and compiler speci cation is JastAdd [
Often, new insights and knowledge about a particular domain arise over time,
leading to new abstractions in the DSL. In such cases, the language developer
typically wishes to extend the existing DSL in a backward compatible way.
JastAdd has proven to be useful for extending languages; compilers for complex
languages like Java and Modelica have successfully been implemented and
extended with new language constructs [
In several domains, a visual notation is preferable to a textual notation, which
can increase the accessibility for non-programmers. Such languages are called
domain-speci c visual languages (DSVLs). Some languages, such as Modelica [
We want to investigate how the metacompilation system JastAdd, and its
formalism reference attributed grammars [
There are several ways for de ning the semantics, that is the meaning, of a
programming language. Examples include denotational semantics [
In this work, we will use reference attributed grammars (RAGs), which is
an extension to Knuth's attribute grammars [
For visual languages, there are several ways to de ne the syntax of a language.
There is, however, no consensus on what formalism to use, in contrast to textual
languages, where Chomsky's context-free grammars are widely used. Examples of
formalisms are graph grammars [
Erwig [
Another related work is reusable visual patterns by Schmidt et al. [
We want to investigate how well suited RAGs are for de ning the semantics
for a visual language, and how to use RAGs for semantics-dependent editor
interaction. In particular, we have the following research goals:
{ Declarative semantic speci cation.
{ Reuse of semantic speci ation.
{ Semantic support in visual editor, for example, showing computed properties.
{ Support languages with both a textual and visual syntax.
{ Automatic incremental evaluation, using work by Soderberg et al. [
We have identi ed the following research challenges, based on experience from the preliminary work described below.
De ning visual syntax. When de ning a visual language, there should be a
way to de ne its syntax and the visual representation, that is, properties such
as shape and color. One way is to have a DSL supporting that and which
includes concepts such as nodes, connections and containment. To specify
how nodes can connect, constraints can be used. Attributes are one way to
specify such constraints. Another way is to use OCL as Bottoni et al. [
Another approach for de ning the syntax is by graph grammars [
We want to nd and choose one approach and connect it to RAGs. Model consistency. In the visual editor, we have both the view and the AST model, and a challenge is to keep these models consistent with each other. One way is to let the visual editor update the AST, which then informs the view about the change and is updated accordingly. If the editor supports both textual and visual editing, is this technique feasible? Another way is to support bidirectional transformations [29].
Automatic layout. We want to nd out how to handle layout that can be sensitive to semantics. A language from ABB has this property, as described below. For these languages, automatic layout must take this semantics into consideration, in order to not inadvertently change the meaning of a diagram. Erroneous diagrams. It can be helpful for the language user if it is possible to temporarily have diagrams that contain errors. For example, when two edges are being switched and it is forbidden to have more than one edge to a node, then it may be useful to temporarily have one edge pointing to nothing.
Undo / redo functionality. To increase the usability of editors for visual languages, functionality like undo and redo is often desirable. We would like to generate edit operations for the AST that automatically provide support for undo and redo functionality.
Refactoring. Another challenge is to support refactoring in visual languages with RAGs. Is it possible to extend the work done by Schafer et al. [30]? 3.2
The research methodology we use is similar to design science research in information systems. Instead of trying to understand the reality, as natural science does, design science tries to create new innovative artifacts that have utility [31]. We develop prototypes that are based on practical applications, as the preliminary work with ABB demonstrates. The research contribution is the technical artifact itself and the development of the foundations of RAGs. Design science consists of two activities, build and evaluate, and is an iterative process; the artifact is continuously evaluated to provide feedback to the building activity [32]. The evaluation will include performance measurements, for example, the response time for standard editing operations, as well as evaluation of speci cation size and modularity. We develop a prototype visual language together with ABB using RAGs, and it would be interesting to see if the engineers at ABB successfully can experiment with new language constructs for this language in a modular way. 3.3
We have started a collaboration with ABB, and designed a control language based on a industrial language from ABB [33]. This language has both a textual and a visual syntax. A diagram in the language contains blocks that are executed periodically and connections between them that specify the data ow. One interesting aspect of the language is that the semantics are in uenced by the visual placements of the blocks. The execution order of the blocks is primarily de ned by the data ow, and secondary by visual placements. To de ne the semantics without visual coordinates, the visual placements are re ected in the declaration order of the blocks in the textual syntax. We have implemented a visual editor for this language using RAGs and the Graphical Editing Framework (GEF), an Eclipse-based framework. A screenshot of the editor can be seen in Fig. 1, with the corresponding textual syntax.
(a) Visual syntax diagramtype tank-regulator(
Int refLevel, Int tolerance => StatusData status) f StatusMonitor level; Sub Sub_1; ... connect(refLevel, Sub_1.in1); ... g (b) Textual syntax We have in this paper described the opportunity to use RAGs for implementing tools for visual languages. Attribute gammars is a proven and convenient formalism to specify the semantics of a language. RAGs extends AGs with references, and makes it possibly to superimpose graphs on an AST, which can be used to represent programs of visual languages. Expected bene ts of using RAGs include reuse of compiler speci cation between di erent tools and the possibility to experiment with new language constructs in a modular way. The expected contributions are the prototype system and development of the foundations of RAGs to support visual languages. Research goals have been described and research challenges identi ed. Finally, we mention design science research as our research methodology and describe the preliminary work with ABB that serves as a case study.
We thank the anonymous reviewers for interesting and useful comments on a earlier draft of this paper. 27. Esser, R., Janneck, J.: A framework for de ning domain-speci c visual languages.
Workshop on Domain Speci c Visual Languages, ACM Conference on ObjectOriented Programming, Systems, Languages and Applications (OOPSLA-2001) (2001) 28. Tveit, M.: Speci cation of graphical representations-using hypergraphs or metamodels? Norsk informatikkonferanse NIK (2008) 39{50 29. Antkiewicz, M., Czarnecki, K.: Design space of heterogeneous synchronization. In: Generative and Transformational Techniques in Software Engineering II. LNCS, Springer (2008) 3{46 30. Schafer, M., Ekman, T., de Moor, O.: Sound and Extensible Renaming for Java.
In Kiczales, G., ed.: 23rd Annual ACM SIGPLAN Conference on Object-Oriented Programming, Systems, Languages, and Applications (OOPSLA 2008), ACM Press (2008) 31. March, S.T., Smith, G.F.: Design and natural science research on information technology. Decis. Support Syst. 15(4) (December 1995) 251{266 32. Hevner, A.R., March, S.T., Park, J., Ram, S.: Design science in information systems research. MIS Q. 28(1) (March 2004) 75{105 33. Fors, N., Hedin, G.: Handling of layout-sensitive semantics in a visual control language. In: Proceedings of the 2012 IEEE Symposium on Visual Languages and Human-Centric Computing, IEEE (2012) poster paper to appear.