Modern software development processes often use domainspeci c modeling languages (DSMLs) combined with custom code generators and/or interpreters. Especially textual DSMLs as provided by Eclipse Xtext are becoming more and more popular. As a consequence, software quality assurance frequently leads back to quality assurance of the involved textual models. Here, various quality aspects have to be considered depending on the modeling purpose and domain. In this paper, we present a quality assurance tool set for textual models in Eclipse using several interrelated components like Xtext, EMF Refactor, Henshin and the OCL tools which are all based on the Eclipse Modeling Framework (EMF). The practicability and exibility of this tool set are demonstrated by the design and implementation of a case study that is based on a textual modeling language for simple web applications named SWM (Simple Web Modeling Language).
The use of models in modern software development processes is becoming more
and more popular. Model-based software development (MBSD) lifts software
models to be the primary artifacts in the software development process. This
is especially true in model-driven software development (MDSD) where
models are nally used for code generation purposes. Moreover, the use of (often
textual) domain-speci c modeling languages (DSMLs) is a promising trend in
modern software development processes to overcome the drawbacks concerned
with the universality and the broad scope of general-purpose languages like the
Uni ed Modeling Language (UML) [
Often, a DSML comes along with a code generator and/or interpreter to
provide functionality that should be hidden from the domain expert. In the
generator case, high code quality can be reached only if the quality of input
models is already high. Typical quality assurance techniques considering the
model syntax are model metrics, model smells, and model refactorings. In [
This paper contributes a exible quality assurance tool set for textual
models supporting quality assurance techniques like model metrics, smells, and
refactorings integrated in textual model editors within the Eclipse IDE [
To demonstrate the practicability and exibility of this tool set we present a case study for quality assurance of textual models. As example language we take a DSML called Simple Web Model (SWM) for de ning a speci c kind of web applications in a platform-independent way1. In the case study, we concentrate on quality aspect completeness, i.e., we analyze whether SWM models are ready for code generation and improve model parts using domain-speci c refactorings. The case study shows that Eclipse is particularly valuable for implementing model quality assurance tools. On the one hand, the plugin technology provides a exible choice of concrete speci cation languages. On the other hand, the use of the abstract syntax (provided by EMF) for implementing quality assurance techniques provides a exible use for visual and textual models in Eclipse.
The paper is structured as follows: In Section 2, we re ect the used model quality assurance process. Section 3 presents the textual SWM language and an example SWM instance model being used in Section 4 to discuss and present techniques and tool support for quality assurance of SWM models. Finally, we conclude with related work in Section 5 and a summary in Section 6. 2
The increasing use of model-based or model-driven software development
processes induces the need for high-quality software models. In [
In our approach, we de ne a process for specifying new quality assurance
techniques as shown in the left-hand side of Figure 1. After having identi ed the
1 Several variations of SWM are used in literature, for example in [
In the next step, static syntax checks for these quality aspects are de ned. This is done by formulating questions that should lead to so-called model smells hinting to model parts that might violate a speci c model quality aspect. Some of these answers can be based on metrics. Other questions may be better answered by considering speci c patterns which can be formulated on the abstract syntax of the model. A speci ed smell serves as precondition of at least one model refactoring that can be used to restructure models in order to improve model quality aspects but appreciably do not in uence the semantics of the model. 3
In the case study, we assume the following scenario (taken from [
The SWM language is de ned as follows. A WebModel consists of two parts: a DataLayer for modeling entities which should be persisted in the database, and a HypertextLayer presenting the web pages of the application. An Entity owns several Attributes (each having a SimpleType) and can be related to several other entities. A Page is either a StaticPage having a static content or a DynamicPage (IndexPage or DataPage) having a dynamic content depending on a referenced entity type. An IndexPage lists objects of this entity whereas a DataPage shows concrete information on a speci c entity like its name, attributes, and references. Finally, pages are connected by Links.
A prominent environment for textual modeling in Eclipse is Xtext [
Listing 1. Part of the constructive Xtext grammar of the SWM language . . .
E n t i t y : ' e n t i t y ' name=ID ' f ' a t t r i b u t e s+=A t t r i b u t e r e f e r e n c e s+=R e f e r e n c e ' g ' ; A t t r i b u t e : ' a t t ' name=ID ' : ' t y p e=SimpleType ; . . .
IndexPage : ' i n d e x page ' name=ID ( ' shows e n t i t y ' e n t i t y =[ E n t i t y ] ) ? ' f ' l i n k s+=Link ' g ' ; . . . 4
In this section, we use the structured model quality assurance process presented
in Section 2 for textual models of the SWM language. First, we de ne concrete
quality assurance techniques for SWM. Then, we demonstrate how these
techniques are applied on a concrete SWM instance model. Finally, we show concrete
speci cations using di erent speci cation languages.
2 The complete grammar can be found in Appendix A and is partially taken from [
Quality assurance techniques for SWM models
Since platform-speci c models should be derived from SWM models and should
be used to generate the Java-based implementations, the major quality aspect to
be ful lled on SWM models is Completeness. A model is complete if it contains
all relevant information, and if it is detailed enough according to the modeling
purpose [
Furthermore, several metrics can be used to analyze completeness of SWM models. For example, metrics Number of Entities in the Model (NEM) and Number of Dynamic Pages in the Model (NDPM) can be used to get a rst overview on the model structure. Here, a ratio between the values of these metrics less than 1 : 2 might be a hint for missing dynamic pages 3. Similarly, metrics Average number of Attributes (resp. References) in Entities of the Model (AvNAE resp. AvNRE) are useful to detect missing information in the data layer.
After having speci ed appropriate model smells, suitable refactorings have to be de ned in order to support the handling of smelly SWM models. Smells No Dynamic Page and Unused Entity can be eliminated by a refactoring which inserts both an index page and a data page referencing the corresponding entity to the hypertext model (refactoring Insert Dynamic Pages). For eliminating smell Missing Link an appropriate refactoring Update Links to Index Pages can be used that ensures that the start page owns links to all index pages of the model. Finally, there is no adequate refactoring to eliminate smell Empty Entity. Here, manual model changes should be performed. 4.2
Application of quality assurance techniques to SWM models We now assume that the software company has to develop a web application for the rental system of a vehicle rental company. Listing 2 shows a rst SWM model being developed in an early stage of the development process.
For the rst overview on a model, a report on project-speci c model metrics might be helpful. Calculated metric values are presented in a speci c view within the Eclipse workbench. For reporting purposes, the results can be exported using 3 I.e., one entity should be referenced by both an index page and a data page. several output formats (like PDF, HTML, or widely used MS O ce formats) and designs (like simple lists or tube diagrams). In our concrete example model, metrics NEM and NDPM (see Section 4.1) are calculated to 4 and 3, respectively. This means that there are more entities in the web model than dynamic pages hinting to potentially missing dynamic pages.
Listing 2. Example SWM instance model before model review webmodel VehicleRentalCompany f data f e n t i t y Customer f a t t name : String a t t e m a i l : Email r e f a d d r e s s : A d d r e s s g e n t i t y A d d r e s s f a t t s t r e e t : String a t t c i t y : String g e n t i t y Car f
a t t t y p e : String g e n t i t y Agency f g g hypertext f index page c a r i n d e x shows e n t i t y Car f
l i n k to page c a r d a t a g data page c a r d a t a shows e n t i t y Car f g index page a g e n c y i n d e x shows e n t i t y Agency f s t a t i c page i n d e x p a g e f l i n k to page c a r i n d e x l i n k to page a g e n c y i n d e x g s t a r t page i s i n d e x p a g e g g g
To make this problems more explicit (and thus more obvious), EMF Refactor supports analysis functionality with respect to so-called model smells representing model parts to be improved. As for model metrics, our tool environment provides a con guration of speci c model smells being relevant within the current project. Similarly to the metrics calculation process, a smell analysis can be triggered from an an element shown in the textual model editor. The results of a smell analysis are presented in a speci c view within the Eclipse workbench. The bottom part of Figure 2 shows two entities being not referenced by a dynamic page (smell No Dynamic Page on entities Customer and Address) and one occurrence of smells Empty Entity and Unused Entity each. After selecting a concrete smell occurrence in the tree-based view the involved element is highlighted in the textual editor (see top part of Figure 2).
Besides manually changing the model, refactoring is the technique of choice to eliminate occurring smells. In our example, we can use refactoring Insert Dynamic Pages to eliminate smell No Dynamic Page on entity Customer 4. 4 Note that we do not eliminate smell No Dynamic Page on entity Address since this entity is referenced by entity Customer, i.e. it is part of this entity.
The refactoring is triggered from the context menu of entity Customer (see highlighted part in Figure 2). Then, the tool set provides two previews: the rst for visualizing model changes performed by the refactoring, the second for a concrete overview on smell occurrence changes when applying the refactoring. The result of refactoring Insert Dynamic Pages is shown in Figure 3.
Two dynamic pages (an index page and a data page) referencing entity Customer are in- Fig. 2. Report of concrete smell occurrences in serted into the hypertext layer our example SWM model and highlighting of inof the model. Furthermore, the tvhoelvteedxteuleamleXnttexint semdietlolrNo Dynamic Page within inserted data page is linked by the index page which is in turn linked by the static page named indexpage being the starting page of the hypertext layer (see Listing 2).
Due to space limitations,
we have to skip further model
analysis and refactoring steps
here. However, we think that
the application of the
quality assurance process and the
handling of the supporting
Fig. 3. Inserted and changed model elements after tools are su ciently and
plauapplying refactoring Insert Dynamic Pages sibly presented. For more
detailed discussions on process
and tooling we refer to [
Speci cation of quality assurance techniques for SWM models EMF Refactor provides a wizard-based speci cation process for quality assurance techniques as well as basic code generation facilities. This has the advantage that the designer can concentrate on the essential speci cation part only.
Listing 3. OCL speci cation of SWM smell No Dynamic Page context WebModel def : noDynamicPages ( ) : Set ( E n t i t y ) =
E n t i t y . a l l I n s t a n c e s ( ) > e x c l u d e s A l l ( DynamicPage . a l l I n s t a n c e s ( ) > c o l l e c t ( e n t i t y ) )
OCL has been proven to be well-suited to specify metrics. For example, metric
NEM is simply de ned using OCL expression self.dataLayer -> size() on
context element WebModel. Listing 3 shows the OCL speci cation of SWM model
smell No Dynamic Pages. OCL operation noDynamicPages() returns the set of
entities being not referenced by any dynamic page (line 2): Starting from all
instances of type Entity within the web model (line 3), we exclude those which
are referenced by at least one dynamic page (line 4).
For refactoring speci cation we use Java and the EMF model transformation
language Henshin [
All the sources of the case study (code, models, quality assurance techniques,
etc.) can be found in the download section of the EMF Refactor web site [
In this section, we give an overview on quality assurance tools within the EMF
world, especially in the eld of textual modeling using Xtext. For a
comprehensive overview also considering quality assurance of UML models we refer to [
To the best of our knowledge, explicit tool support for metrics calculation on
EMF models is not yet available (besides the EMF Refactor tooling). The EMF
Query Framework [
The Epsilon language family [
In another approach, the authors propose the de nition of EMF-based
refactoring in a generic way [
Xtext has outstanding support for static model analysis and validation. Custom constraints and quick xes can be de ned to tackle errors and warnings instantaneously. However, the main purpose of these constraints and quick xes is to address model consistency. They are not especially dedicated to quality assurance in a common sense. Moreover, there is no support for custom congurations of validation suites. Furthermore, Xtext provides basic refactoring functionality for generic renaming of arbitrary model elements. Support for custom refactorings (for example, for custom DSMLs like SWM) is not provided. 6
In this paper, we present a exible tool set for quality assurance of textual models within Eclipse. The tool set integrates a number of tools which are all built atop EMF (Xtext, EMF Refactor, OCL/MDT, and Henshin). In a case study, we use a DSML for de ning a speci c kind of web applications and concentrate on the quality aspect completeness. We use model metrics and model smells for static analysis and model refactoring for improving the structure of the models.
The implementation shows that (1) the structured model quality assurance
process presented in [
Future work on the topic presented in this paper is separated into two directions. On the one hand, we want to integrate further speci cation languages into the EMF Refactor infrastructure. Here, we are currently working on the integration of EMF Query. On the other hand, it would be useful to provide a suite of prede ned metrics, smells, and refactorings for speci c Xtext grammars in order to analyze and improve the structure of the corresponding language. A Complete Xtext grammar of the SWM language grammar org . e c l i p s e . emf . r e f a c t o r . examples . SimpleWebModel with org . e c l i p s e . x t e x t . common . Terminals generate simpleWebModel " http : / /www. e c l i p s e . org /SWM/ 1 . 0 " WebModel : ' webmodel ' name=ID ' f ' dataLayer=DataLayer hypertextLayer=HypertextLayer ' g ' ; DataLayer : ' data f ' f DataLayer g
e n t i t i e s+=Entity ' g ' ; Entity :
' e n t i t y ' name=ID ' f ' a t t r i b u t e s+=A t t r i b u t e r e f e r e n c e s+=R e f e r e n c e ' g ' ; A t t r i b u t e : ' att ' name=ID ' : ' type=SimpleType
; enum SimpleType : Boolean j Email j I n t e g e r j S t r i n g ; R e f e r e n c e : ' r e f ' name=ID ' : ' type =[ Entity ]
; HypertextLayer : ' h y p e r t e x t f ' pages+=Page+ ' s t a r t page i s ' s t a r t P a g e =[ S t a t i c P a g e ] ' g ' ; Page : S t a t i c P a g e j DynamicPage
; S t a t i c P a g e : ' s t a t i c page ' name=ID ' f ' l i n k s+=Link ' g ' ; Link : ' l i n k to page ' t a r g e t =[Page ]
; DynamicPage : IndexPage j DataPage
; IndexPage : ' index page ' name=ID ( ' shows e n t i t y ' e n t i t y =[ Entity ] ) ? ' f ' l i n k s+=Link ' g ' ; DataPage : ' data page ' name=ID ( ' shows e n t i t y ' e n t i t y =[ Entity ] ) ? ' f ' l i n k s+=Link ' g ' ;
Listing 4. Complete Xtext grammar of the SWM language webmodel VehicleRentalCompany f data f entity Customer f att name : String att email : Email ref a d d r e s s : Address ref account : BankAccount g entity Address f att s t r e e t : String att postalCode : I n t e g e r att c i t y : String g entity BankAccount f att number : I n t e g e r att bankCode : String att bankName : String g entity Car f att manufacturer : String att type : String att power : I n t e g e r g entity Agency f
ref a d d r e s s : Address g g hypertext f index page c a r i n d e x shows entity Car f
link to page cardata g data page cardata shows entity Car f g index page agencyindex shows entity Agency f
link to page agencydata g data page agencydata shows entity Agency f g index page customerindex shows entity Customer f link to page customerdata g data page customerdata shows entity Customer f g s t a t i c page indexpage f link to page agencyindex link to page c a r i n d e x link to page customerindex g start page i s indexpage