Towards a Generic Traceability Framework for

Model-driven Software Engineering

Birgit Grammel

SAP Research CEC Dresden, Chemnitzer Str. 48

01187 Dresden, Germany

Abstract. With the inception of Model-Driven Software Engineering (MDSD)

the need for traceability is raised to understand the complexity of model trans-

formations and overall to improve the quality of MDSD. Using the advantage of

generating traceability information automatically in MDSD, eases the problem of

creating and maintaining trace links, which is a labor intensive task, when done

manually. Yet, there is still a wide range of open challenges in existing trace-

ability solutions and a need to consolidate traceability domain knowledge. This

paper proposes a generic framework for augmenting arbitrary model transforma-

tion approaches with a traceability mechanism. Essentially, this augmentation is

based on a domain-speciﬁc language for traceability providing the formalization

on integration conditions needed for implementing traceability. The paper is of

positional nature and outlines work currently in progress.

1 Introduction

In the Standard Glossary of Software Engineering Terminology

traceability is deﬁned

as: The degree to which a relationship can be established between two or more prod-

ucts of the development process, especially products having a predecessor-successor or

master-subordinate relationship to one another. Traceability information in MDSD can

be understood as the runtime footprint of model transformations. Essentially, trace links

provide this kind of information by associating source and target model elements w.r.t.

the execution of a certain model transformation. The corresponding models may con-

form to the same or different metamodels. Trace links have a manifold application do-

main [1], [2]: System analysis to understand system complexity by navigating via trace

links along model transformation chains; Coverage analysis to determine whether all

requirements were covered by test cases in the development life cycle; Change impact

analysis to analyze how changing one model would effect other related models; Orphan

analysis to ﬁnd orphaned model elements with respect to a speciﬁed trace link; Model-

based debugging, i.e. mapping the stepwise execution of an implementation back to its

high-level model and Model transformation debugging.

In this paper, we propose a generic traceability framework for augmenting arbitrary

model transformation approaches with a traceability mechanism. To achieve this goal,

we advocate two research directions: a) the augmentation of the generator logic, for the

Std 610.12-1990

Grammel B. (2009).

Towards a Generic Traceability Framework for Model-driven Software Engineering.

In Proceedings of the 1st International Workshop on Future Trends of Model-Driven Development, pages 44-47

DOI: 10.5220/0002185400440047

 SciTePress

sake of trace link generation b) the augmentation of input models with model mark-

ers, serving as a reference for trace link generation. Both approaches are based on a

domain-speciﬁc language for traceability, presenting the formalization on all adequate

conditions and measures needed for implementing traceability in the context of MDSD.

The content of the following sections are structured as follows. First, we reﬂect on

challenges for traceability and give an overview on related work. In section 3 an outline

of our approach is proposed.Finally, section 4 concludes this paper and gives an outlook

on future work.

2 Challenges for Traceability in MDSD

MDSD raises the need for traceability to understand model transformation complexity

and overall to improve the quality of MDSD. Introducing model transformations, to

generate trace links automatically, eases the task of creating and maintaining trace links,

as opposed to when done manually, entailing high development costs. [3]

According to [2], [4] transformation approaches either generate trace links implic-

itly or explicitly, that is, in the former case, either provide an integrated support for

traceability or in the latter one, rely on a developer to encode traceability as a regu-

lar output model. The main advantage of implicit generation (e.g. QVT

, MOFScript

)

is, that no additional effort is necessary to generate trace links, as this is done auto-

matically in parallel to the actual transformation. A disadvantage is, that the traceabil-

ity metamodel is ﬁxed. Since most transformation approaches have differently deﬁned

metamodels, standardization among different approaches is aggravated. Furthermore,

the level of granularity of trace links may differ form one traceability scenario to an-

other. Setting the granularity of trace links has been identiﬁed as a challenge [5]. When

tracing all model element references, this might lead to the following issues: incompre-

hensibility of trace link data and hence be less useful to developers; performance, when

handling large and complex transformations and security, when not all model informa-

tion is allowed to be traced for security reasons, mandated for instance by customer

needs.

Alternatively, the case of explicit generation necessitates the incorporation of addi-

tional transformation rules to generate trace links. (e.g. ATL

, oAW

) The metamodel

deﬁnition is not predeﬁned and transformation engine independent. Hence, the trace

link granularity is adaptable. Yet, the drawback is that additional effort is required to

add traceability-speciﬁc transformation rules, which may also pollute the implementa-

tion. An approach that partly solves these issues in ATL by automatically generating

traceability-speciﬁc transformation rules was proposed in [6].

Another challenge concerns the semantics of trace links [5]. It is often necessary to

distinguish between different kinds of links, e.g. a link between a textual requirement

and a model element has a different semantic, than a reﬁnement relationship within a

model. Fixing the kinds of semantic links, has the consequence of less ﬂexibility for

Query View Transformation, http://www.omg.org/docs/ptc/07-07-07.pdf

http://www.modelbased.net/mofscript/

ATLAS Transformation Language, http://www.eclipse.org/m2m/atl/

openArchitectureWare, http://www.openarchitectureware.org/

user-deﬁned links that might be necessary to meet different project needs [7]. On the

other hand, since the choice of semantic, attributed to a link, is guided by the reasoning

about what the user will perform with the link [1], not predeﬁning the link semantics,

may result in failure of reasoning, due to misuse of the semantics.

3 Traceability Approach

The main idea of our approach is to rely on work and solutions already available and

not to implement yet another transformation language. Nevertheless, is the aim to fo-

cus on current traceability solutions, consolidate beneﬁts of implicit and explicit trace

link generation and tackle the above-mentioned challenges to derive the necessary re-

quirements for a generic traceability framework. In essence, our approach follows two

research directions both based on a commonly used conceptional layer. The main moti-

vation for such a conceptional layer is based on the idea, proposed in [7], that the kind

of traceability data to be collected is dependent on the actual traceability goal, that is,

which traceability question is expected to be answered. To formalize this traceability

concern, a system of rules and regulations is necessary, on when to trace, what kind of

traceability data; additionally, to specify, the granularity level and what kind of trace

links are allowed between certain artefact types, given a certain traceability goal. This

formalization will be deﬁned w.r.t a domain speciﬁc language for traceability, which

is called Trace-DSL in the following. The Trace-DSL logic encompasses all heuris-

tics on how to manifest traceability-speciﬁc rules for arbitrary model transformation

approaches, providing explicit knowledge to the developer on their integration.

To create a traceability model we discussed two classes of transformationapproaches

in section 2. Explicit trace link generation requires the incorporation of additional trans-

formation rules to generate trace links. As this task is generally done manually, it is

likely to be error-prone and time consuming. Although work has been done on the au-

tomation, e.g. [6], still every transformation template needs to be adapted individually,

for the explicit generation. For generators, that already provide a dedicated traceability

support, e.g. QVT, this effort would theoretically not be necessary, yet there is a need

to tune the level of granularity of trace links, as motivated in section 2. Native tracing

without a clear rule system on the conditions and measures of tracing, may lead to an

unnecessary overkill of traceability information, performance issues and violation of

security measures. The guidance, to trace only relevant information is provided by the

Trace-DSL.

Hence, the following two research directions are investigated, showing different

approaches on how to achieve the population of a traceability metamodel. The ﬁrst re-

search direction (Fig.1a) is proposedto enhance the generatorlogic itself for the purpose

of traceability functionality. a) Augmentation of the generator logic: One way to pop-

ulate a traceability metamodel is to provide a generic interface and trace engine for arbi-

trary generators. In this case the interface supplies the engineer with an API to connect

his generator to the trace engine. As a result, the generator is featured with traceability

functionality. The interface is based on the Trace-DSL, to account for a formalized rule

system on traceability. The second approach (Fig.1b) to be examined follows the idea

of high order transformations and in contrast to the ﬁrst approach does not focus on the

level of the generator logic. b) Augmentation of input and output models: Prior to the

actual transformation the input model is augmented via model transformations with cer-

tain markers such as identiﬁers or properties. This enhanced model is then transformed

by the generator into an enhanced output model. In essence, both models are compared

automatically w.r.t. the correlation of before and after markers. Querying of traceability

information is then coupled to these methods of auto-correlation. The Trace-DSL or-

chestrates the model augmentation to again only trace relevant information for a certain

underlying traceability question.

Output Models

Input Models

Input

Models

Generic

Interface

Trace Engine

Trace-DSL

Output

Models

Transformation Engine

Model Marker Correlation

Transformation Engine

)

Fig.1. Traceability research approaches.

4 Conclusions and Future Work

The two research directions run orthogonally to each other w.r.t. their target points,

being the generator logic level versus the model level. By comparing their complexities,

advantages and disadvantages, we wish to develop a generic traceability framework for

arbitrary transformation approaches.

Future work is directed at the generation of traceability-speciﬁc rules based on the

principles of metamodel matching [8]. It is to be investigated, how to use the resulting

alignment of metamodel matching, to generate the code of traceability-speciﬁc rules,

automatically.

References

1. Ramesh, B., Jarke, M.: Towards reference models for requirements traceability. In: IEEE

Transactions on Software Engineering. Volume 21. (2001)

2. Czarnecki, K., Helsen, S.: Feature-based survey of model transformation approaches. IBM

Systems Journal, Vol 45, No 3 (2006)

3. Aizenbud-Reshef, N., et al.: Model traceability. IBM Systems Journal, Vol 45, No 3 (2006)

4. Vanhooff, B., Baelen, S.V., Joosen, W., Berbers, Y.: Traceability as input for model transfor-

mations. In: Proceedings of the ECMDA Traceability Workshop. (2007)

5. Antoniol, G., et al.: Problem statement and grand challenges in traceability. Center of Excel-

lence for Traceability (2006)

6. Jouault, F., Kurtev, I.: Transforming models with ATL. In: Proceedings of the Model Trans-

formations in Practice Workshop at MoDELS 2005. (2005)

7. Knethen, A., Paech, B.: A survey on tracing approaches in practice and research. Technical

report, Frauenhofer IESE (2002)

8. Falleri, J., Huchard, M., Lafourcade, M., Nebut, C.: Metamodel matching for automatic model

transformation generation. In: Proceedings of MoDELS. (2008)