Entity-relationship Modeling Tools and DSLs: Is It Still Possible to

Advance the State of the Art from Observations in Practice?

Jonnathan Lopes

, Maicon Bernardino

, Fábio Basso

and Elder Rodrigues

Postgraduate Program in Software Engineering (PPGES),

Laboratory of Empirical Studies in Software Engineering (LESSE),

Federal University of Pampa (UNIPAMPA), Av. Tiarajú, 810, Ibirapuitã, CEP 97546-550, Alegrete, RS, Brazil

Keywords:

Multivocal Literature Mapping, Domain Speciﬁc Language, Conceptual Data Model, Conceptual Model,

Database Modeling.

Abstract:

The variety of database system technologies that became available in recent years makes difﬁcult the selection

of tools for modeling entity-relationship (ER). The published mapping studies on this topic date back to 2000,

thus outdated and limited to guide designers towards the recent innovations selection for the design and im-

plementation of databases. In this sense, we contribute with an overview of the recent innovations through a

systematic literature mapping complemented by research in the gray literature. This paper scopes ten (10) pri-

mary studies focused on Domain-Speciﬁc Languages (DSL) and identiﬁes ﬁfty-ﬁve (55) tools already applied

in industry and academia for ER modeling at the conceptual, logical, and physical level. Hence, as a signif-

icant increment to existing mapping studies, this presents the state-of-the-art and practice for ER modeling,

including its characterization and research gaps.

1 INTRODUCTION

In the software development process, data models can

undergo different transformations. The three-level

modeling approach to database (DB) design (ANSI,

1975) suggests separating data models into different

degrees of abstraction, known as conceptual, logical,

and physical models. The conceptual modeling of

database systems requires the highest level of abstrac-

tion, and the most widespread technique for its devel-

opment is the conceptual data modeling proposal con-

ceived by Chen (1976). This proposal, inﬂuenced by

the three-level approach, was so well accepted that it

was considered a deﬁnitive reference for DB model-

ing (Cougo, 2013).

DB modeling is a relevant part of the software de-

velopment lifecycle, which generally some Domain-

Speciﬁc Language (DSL) supported (Voelter, 2009).

According to Van Deursen et al. (2000), a DSL is

an executable programming language or speciﬁca-

tion language that offers expressive power focused,

through appropriate notations and abstractions, usu-

https://orcid.org/0000-0001-9402-801X

https://orcid.org/0000-0003-2776-8020

https://orcid.org/0000-0003-4275-0638

https://orcid.org/0000-0003-3431-2814

ally restricted to a domain-speciﬁc problem. Like

other languages, a DSL must present a set of sen-

tences well deﬁned by syntax and semantics them-

selves. For Fowler (2010), a DSL is a computer pro-

gramming language with limited expressiveness and

focused on a particular domain. In addition to code-

ﬁrst approaches, the adoption of DSLs in software

processes promotes a signiﬁcant increase in produc-

tivity, quality, ease of use, and ﬂexibility (Mernik

et al., 2005; Vara et al., 2014). The great advantage

of using DSLs is an abstraction, which takes a less

complex level of knowledge needed for their devel-

opment. In this way, domain experts can understand,

validate and modify the code through model transfor-

mations and reﬁnements (Van Deursen et al., 2000).

According to Mernik et al. (2005), we can clas-

sify DSLs into three dimensions: origin (internal or

external), appearance (textual, graphical, tabular, or

symbolic), and implementation (run mode). In gen-

eral, the principal consideration for creating a DSL

should be where it comes from, as each approach

has speciﬁc advantages and disadvantages (Fowler,

2010). On the one hand, we can design an inter-

nal DSL around the grammatical rules of an existing

language, which can be a General Purpose Language

(GPL) or another DSL. On the other hand, an exter-

Lopes, J., Bernardino, M., Basso, F. and Rodrigues, E.

Entity-relationship Modeling Tools and DSLs: Is It Still Possible to Advance the State of the Art from Observations in Practice?.

DOI: 10.5220/0011044500003179

In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 1, pages 179-186

ISBN: 978-989-758-569-2; ISSN: 2184-4992

179

nal DSL is a language with syntax owned and depends

on its infrastructure for lexical, syntactic, and seman-

tic analysis, interpretation, compilation, optimization,

and code generation (Mernik et al., 2005). As for ex-

ecution, conceptually, this dimension ranges from the

speciﬁcation of graphic and textual notations of lan-

guages applied on speciﬁc domains to implemented

solutions.

The objective of this study is to map existing

DSLs to DB modeling at any of the three model lev-

els. To meet this purpose, we carried out a Multivocal

Literature Mapping (MLM) to ﬁnd primary studies

and tools that present approaches, methods, or tech-

niques for the DB models creation.

We organized this study as follows. Section 2

presents the discussion of related works. The proto-

col used we detailed in Section 3. Its execution is

described in Section 4, while the obtained results are

analyzed and discussed in Section 5. Section 6 de-

scribes the threats to the study’s validity. Section 7

points out possible gaps for future research and, ﬁ-

nally, Section 8 presents the ﬁnal remarks.

2 RELATED WORK

Before performing this study, we performed an ex-

ploratory ad hoc

search in the literature for studies

that proposed research related to approaches to mod-

eling entity-relationship (ER) in DB, with a focus on

primary studies and mappings.

Gregersen and Jensen (1999) performed a survey

and analyzed the design properties applied to spec-

ify temporal attributes in ER modeling approaches.

The aiming study presents the new concepts used to

facilitate temporal modeling in 19 different propos-

als. These proposals are based on models, namely:

ER, EER, OO, and ERC+. As a result, the differ-

ent mapping models for relational models are high-

lighted, used in the database implementation plat-

form, which has algorithms that optimize the deﬁ-

nition of attributes with temporal value. In addition,

we evaluated ten (10) models against various design

properties to verify their completeness. The authors

conclude that while the models collectively have ade-

quate design coverage, none satisfy all aspects of the

evaluated modeling. It denotes that the temporal as-

pects modeling in data models is a motivating activity

and a promising research area.

Siau (2012) conducted a literature review on re-

lational and semantic approaches to ER modeling.

Ad hoc search is a customized search, using a tailor-

made method, and structured to meet speciﬁc needs without

necessarily a high degree of scientiﬁc rigor.

He shows that most studies point out that semantic

models are better than relational models and that, to

a lesser degree, other studies do just the opposite.

Early studies argue that, in general, semantic mod-

els are easier to use for end-users as they offer no-

tations with a higher level of abstraction for domain

modeling. More recent studies that oppose the ﬁrst

observation, on the other hand, commonly assert that

the concepts of relationships between entities in re-

lational models are more understandable in making

their relationship identiﬁers explicit. In addition, it

discusses some works that claim there are no signiﬁ-

cant differences between the two approaches studied,

citing, for example, the author carried out a controlled

experiment using the same SQL statement, i.e. the

subjects did not show any difference signiﬁcant when

they performed their modeling using the different ap-

proaches.

We can mention three other relevant non-

systematic mappings of ER approaches. The un-

structured literature review presented in Ma and Yan

(2010) searches contributions to uncertain informa-

tion in conceptual modeling, encompassing complex

objects and inaccurate data. In Lorenz et al. (2016),

the authors present an unstructured mapping litera-

ture study comprehending industry and academic best

practices in ER projects.

In Torres et al. (2017), the authors describe a map-

ping study to characterize nine (9) platforms that sup-

port Object-Relational Mapping Solutions (ORMS)

that includes examples in UML and Barker notations.

These are complementary studies and do not overlap

with the research questions proposed in this research.

Furthermore, unlike ours, the studies aforementioned

are awkward to replicate due to the lack of a protocol.

3 MULTIVOCAL LITERATURE

MAPPING

A Multivocal Literature Mapping (MLM) is a form of

Systematic Literature Mapping (SLM) that includes

gray literature. MLMs are helpful for researchers and

practitioners as they provide comprehensive insights

into the state-of-the-art and practice in a particular

area. In this study, we conducted an MLM

using the

SLM process deﬁned by Petersen et al. (2008) and the

guidelines proposed by Garousi, V. and Felderer, M.

and Mäntylä (2019) to search the gray literature.

To this end, we deﬁned the following Research

Questions (RQs): RQ1. What is the state-of-the-

Data from the experimental package available in the

Zenodo DOI repository: 10.5281/zenodo.5724135

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

180

art in developing DSLs for transforming ER models?

e.g. modeling or manipulation, and transformation.

RQ1.1. What are reﬁnement methodologies, tech-

niques, and proposals (automated design) based on

data models presented in the primary studies? e.g.

normalization. RQ1.2. What tools support the devel-

opment of these DSLs? RQ2 What evaluation meth-

ods are used in primary studies? RQ2.1. What are

the positive and negative points observed in carrying

out the studies? RQ2.2. What challenges do the pri-

mary studies point out? RQ3. What are the tools for

conceptual modeling of databases? RQ3.1. What no-

tations or languages do these tools use? RQ3.2. What

modeling levels (Conceptual, Logical, Physical) do

these tools support?

Some of the RQs (RQ1, ..., RQ2.2) were essen-

tially answered through academic studies, thus seek-

ing to understand the state-of-the-art, while others

(RQ3, ..., RQ3.2) reﬂected the investigation of the

state-of-the-practice and therefore, we used gray lit-

erature for a better conclusion.

For the SLM, it was necessary to access Digi-

tal Libraries (DLs) to obtain scientiﬁc studies. For

the research conducted, the DLs consulted should in-

clude three mandatory requirements: have a web-

based search engine; have a keyword search engine,

and; cover primary studies in the ﬁeld of Computer

Science. Therefore, we carried out the search in ﬁve

DLs: (i) ACM DL; (ii) IEEE Xplore; (iii) ScienceDi-

rect; (iv) Scopus; (v) SpringerLink. Afterward, we

established the most expressive terms and synonyms

for the research object. The combination generated

the generic string in Figure 1.

(DSL OR Domain Specific Language OR Domain-Specific Language OR

Domain-Specific-Language OR DSML OR Domain Specific Modeling

Language OR Domain-Specific Modeling Language OR

Domain-Specific-Modeling-Language OR Query Language) AND (ER OR

Entity-Relationship OR Enhanced Entity-Relationship OR Extended

Entity-Relationship OR Database)

Figure 1: Search string.

We elaborated selection criteria for primary stud-

ies to include the most relevant works. We divided the

criteria into:

(i) Inclusion Criteria (IC): (IC1.) The study pro-

poses some technique, method, approach or tool for

the representation and transformation of DB mod-

els using DSL. (ii) Exclusion Criteria (EC): (EC1.)

Study with less than four pages; (EC2.) Study not

written in English; (EC3.) Duplicate study; (EC4.)

Study not provide full access to its content; (EC5.)

Study not meet IC1.

In addition, we established a set of seven (7) qual-

ity assessment criteria (QCs). The purpose of the

quality assessment was both to qualify the relevance

of the studies and assist in the interpretation of the

results, as well as to limit the SLM bias (Dybå and

Dingsøyr, 2008). In addition, we gave a score to the

QCs: (i) T:Total, fully considering the evaluated QC;

(ii) P:Partial, partially considering the evaluated QC

and according to its weight; (iii) N:Negative, does not

include the evaluated QC at all.

The maximum possible score, evaluating all cri-

teria, was ten (10.0) and the minimum zero (0). Each

QC had a speciﬁc weight (1 → 1.5 → 2) depending on

its importance considered for this study. We consid-

ered aspects relevant to the research these QCs based

on, namely report (QC1, QC4, QC5), rigor (QC2,

QC3), credibility (QC2, QC3 ) and relevance (QC1,

QC6, QC7). Table 1 presents the list of QCs and their

respective weights.

Table 1: Quality Assessment Criteria

QC Description Weight

QC1 Does the study make any contribution to the ﬁeld of DB mod-

eling?

1.5

QC2 Does the study present techniques or proposals for reﬁne-

ment based on data models ?

1.5

QC3 Does the study apply any form of empirical evaluation? 1.5

QC4 Does the study describe the characteristics of the DSL cre-

ation process?

1.5

QC5 Does the study characterize the model transformation activi-

ties for different DB technologies?

2.0

CQ6 Does the study have positive and negative points observed in

its execution?

1.0

QC7 Does the study point out challenges arising from its imple-

mentation?

1.0

Once we carry out the quality assessment, it is

necessary to perform the data extraction step to in-

terpret the results. Thus, a data extraction form was

created, consisting of a series of items, such as the

origin of the solution, the year of publication, the ob-

jective, the evaluation of the proposal, the positive and

negative points observed, among other data.

We used the Google search engine to seek

database modeling tools in gray literature. For this,

we deﬁned combinations of keywords from the search

string and adapted them to the context of the search on

the web: (i) ERD Modelling Tool; (ii) ERD Design

Tool; (iii) Conceptual Design of Database; (iv) Con-

ceptual Modelling of Database; (v) Conceptual Mod-

elling; (vi) Database Modelling Tool; (vii) Database

Design Tool. The veriﬁcation of results was limited

to 10 (ten) pages in the search engine for each key-

word.

The acceptance of tools into MLM was subject

to a selection process based on certain requirements.

First, the tool should provide access for some form of

use (e.g. trial version), it would need to support some

level of DB modeling and it would need to have inter-

face in English or Portuguese (our native language).

Afterwards, the included tools should have the nota-

tions and modeling levels extracted, along with other

relevant information, and then categorized.

Entity-relationship Modeling Tools and DSLs: Is It Still Possible to Advance the State of the Art from Observations in Practice?

181

4 MLM EXECUTION

For the execution of the SLM, it was necessary to

adapt the syntax of the generic string to generate other

versions, seeking to adapt the peculiarities of param-

eterization of the different databases used. Then, we

performed a search for studies in the databases. We

obtained the number of primary studies returned by

the digital library as follows: (i) IEEE Xplore: 513;

(ii) Scopus: 465; (iii) ACM: 1240; (iv) Springer Link:

683; (v) ScienceDirect: 826.

With the initial set of 3727 primary studies identi-

ﬁed, we deﬁned ﬁve (5) selection cycles. In these it-

erations, duplicate studies were excluded (with 3513

remainings), selection of studies based on title and ab-

stract (with 34 remainings), selection based on full

text (with 18 remainings), and selection based on

quality assessment (with 10 remainings). Each itera-

tion had the objective of eliminating studies that were

outside the scope of the research or considered not rel-

evant. In the last iteration, 18 studies had their quality

analyzed. We established that only studies with scores

above ﬁve (5) would be accepted.

Thus, after applying the QCs, eight (8) studies

were excluded. The ﬁnal set of ten (10) studies ap-

proved by the SLM proceeded to the data extraction

stage. With academic works, the lowest score was

5.3, while the highest reached ten (10). Table 2 sum-

marizes the results obtained in the quality assessment.

Table 2: Quality Assessment Results.

Primary Studies Quality Criteria

Reference

CQ1

CQ2

CQ3

CQ4

CQ5

CQ6

CQ7

Total

Ayadi et al. (2016) P P T P N T P 5.7

Celikovic et al. (2014) T T N T P P T 7.0

Dimitrieski et al. (2015) T T T T T T T 10.0

Hammer and Mc Leod (1981) T T N T N P P 5.5

Jagannathan et al. (1988) T T N T P N N 5.5

Kersten et al. (2011) P T T P N N P 5.3

Litwin et al. (1989) P T N T N P T 5.4

Mazairac and Beetz (2013) T T P T N N P 5.9

Shipman (1981) T T N T P N P 6.0

Tian et al. (2006) T P N T P P T 6.4

Caption: T = Total, P = Partial, N = Negative

The execution of the search protocol in the gray

literature returned a total of 132 tools. After sum-

marizing, duplicates were excluded (with 67 remain-

ings). During this process, we did not evaluate twelve

(12) tools due to some unavailability type, e.g. the im-

possibility of installation or incompatibility with the

environment used. In the end, it was possible to run

usage tests on 55 tools, and consequently, extract rel-

evant data for the study. During the use of the tools,

we created simple database models, in which we tried

to observe the support for modeling levels and the no-

tations and languages used.

5 MLM RESULTS

Regarding the state-of-the-art (RQ1) in the develop-

ment of DSLs applied to database modeling, a study

was identiﬁed that presents the Multi-paradigm Infor-

mation System Modeling Tool (MIST) (Dimitrieski

et al., 2015) , which uses a bidirectional DSL called

EERDSL, for conceptual modeling using the Ex-

tended Entity-Relationship (EER) approach. Bidirec-

tional modeling and transformation is a model-driven

engineering paradigm concept. In this context, we

are referring to a DSL capable of transforming mod-

els to code, and from code to models. As for the

reﬁnement methodologies, techniques, or proposals

(RQ1.1) based on data models, only the two stud-

ies mentioned applying concepts for reﬁnement, sup-

ported by database normalization to help developers

use their solutions.

With regard to the technologies used to support

DSL development (RQ1.2), we registered the fol-

lowing: Xtext, Xtend, Sirius and Eugenia (Celikovic

et al., 2014; Dimitrieski et al., 2015), StarUML

(Ayadi et al., 2016), IfcDoc Tool and ViewEdit Tool

(Mazairac and Beetz, 2013), MonetDB (Kersten et al.,

2011), Java, JFlex, and JCup (Tian et al., 2006). How-

ever, older studies were generally DSL speciﬁcations,

not showing any form of implementation or tool used

(Jagannathan et al., 1988; Litwin et al., 1989; Ship-

man, 1981).

The database representations adopted (RQ1.3)

have Tables and Functions in all primary studies

analyzed. There are also explicit references to the

deﬁnition of Stored Procedures, Triggers, and

Views in other studies. Table 3 summarizes the data

retrieved from each of the primary studies.

Table 3: Represented Database Objects.

Primary Studies DB Objects

Reference DSL TB SP F T V

Ayadi et al. (2016) Ayadi’s Notation X

Celikovic et al. (2014) EERDSL v.1 X X X

Dimitrieski et al. (2015) ERRDSL v.2 X X X X X

Hammer and Mc Leod (1981) SDM X X

Jagannathan et al. (1988) SDM X X

Kersten et al. (2011) SciSQL X X

Litwin et al. (1989) MSQL X X X X X

Mazairac and Beetz (2013) BIMQL X X X

Shipman (1981) DAPLEX X X

Tian et al. (2006) NeuroQL X X

Caption: TB = Tables, SP = Stored Procedures, F = Funtions, T = Triggers, V = Views

On the methods used to evaluate DSL (RQ2),

there is only one preliminary study that presents the

validation of the proposal (Dimitrieski et al., 2015)

using 16 participants, two (2) experts in IHC, three

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

182

(3) experts in systems modeling and 11 students (6

masters in the database area and ﬁve (5) doctoral stu-

dents with experience in modeling). In general, the

other studies indicate the lack of an evaluation of their

propositions as possible future work.

Among the positive and negative aspects observed

(RQ2.1), the positive aspects of being easy to under-

stand and with intuitive modeling and platform in-

dependence (Mazairac and Beetz, 2013; Tian et al.,

2006) stand out. The negatives were the lack of auto-

matic SQL code generation (Ayadi et al., 2016; Dim-

itrieski et al., 2015) and the lack of actual implemen-

tation of DSLs, with only speciﬁcations among the

works (Ayadi et al., 2016; Hammer and Mc Leod,

1981; Jagannathan et al., 1988; Kersten et al., 2011;

Tian et al., 2006).

In general, the main challenges identiﬁed by the

studies (RQ2.2) are the evaluations of the approaches,

as well as the evolution and/or simpliﬁcation of the

proposals. Table 4 presents the DSLs concerning their

type. However, it is worth noting that the studies that

mark the column bidirectional (Celikovic et al.,

2014; Dimitrieski et al., 2015) are different versions

of the same DSL implementation, while (Hammer

and Mc Leod, 1981; Jagannathan et al., 1988) are a

speciﬁcation and implementation of DSL based on

this speciﬁcation, respectively.

Table 4: Categorization of DSLs Proposals.

Primary Studies DSL Type

Reference DSL T G B

Ayadi et al. (2016) Ayadi’s Notation X

Celikovic et al. (2014) EERDSL v.1 X

Dimitrieski et al. (2015) EERDSL v.2 X

Hammer and Mc Leod (1981) SDM X

Jagannathan et al. (1988) SDM X

Kersten et al. (2011) SciQL X

Litwin et al. (1989) MSQL X

Mazairac and Beetz (2013) BIMQL X

Shipman (1981) DAPLEX X

Tian et al. (2006) NeuroQL X

Caption: T = Textual, G = Graphical, B = Bidirectional.

As for the state of practice (RQ3) of the tools used

in the modeling of DBs, we mapped 55 tools. Thus,

there was the classiﬁcation as to its type, being 29 ex-

clusive modeling DBs (Data Modeling), 13 of model-

ing with connection to integrated DBs and execution

of queries (Full IDE), 10 with support for diagram-

ming of different types of models (Diagramming) and

three (3) tools designed for large companies (Enter-

prise Modeling).

Regarding the notations used in the tools (RQ3.1),

we identiﬁed 18 notations, where the Crow’s Foot no-

tation highlights 35 occurrences and the IDEF1X no-

tation with 23 records. Table 5 listed these and other

data. And ﬁnally, concerning the models supported by

the tools (RQ3.2) it was found that, individually, 26

tools support conceptual modeling, 48 tools support

logical modeling and 37 support physical modeling.

Figure 2 presents the set representing the intersections

of the model support and the tools is presented in the

Venn diagram. The numbers inside the diagram rep-

resent the tool IDs, also displayed in Table 5.

7, 21, 41

10, 11, 39

15, 45, 50

1, 3, 4,

8, 9, 16, 18, 26,

31, 40, 53, 55

5, 12, 20, 23,

33, 34, 35

37, 38, 51

2, 13, 14, 17, 19,

22, 24, 25, 27, 28,

29, 30, 32, 36, 42,

43, 44, 46, 47, 48,

49, 52, 54

Logical

Physical

Conceptual

Figure 2: Venn diagram of the models supported in the

tools.

6 THREATS TO VALIDITY

We identiﬁed threats to the study result in the MLM

performed and then categorized it into the following

types: construct validity, internal validity, external va-

lidity, and conclusion validity, according to the guide-

lines of Cook and Campbell (1979); Wohlin et al.

(2012).

Construct Validity: Addresses the possibility that

the RQs or search terms that structure the search

string are inappropriate or incomplete. To mitigate

these threats, we consulted DSL and researchers of

data modeling. In addition, we perform a pilot survey

to assess the consistency of our search string. An-

other threat is the quality of published material that

has been collected in gray literature.

Internal Validity: Some possible threats are the im-

plementation of incorrect search methods, which can

lead to the exclusion of relevant studies, an applica-

tion of poor data extraction strategy, the occurrence

of biases in the selection or content of primary stud-

ies. In an attempt to mitigate these risks, we deﬁned a

protocol based on well-established reference models

Entity-relationship Modeling Tools and DSLs: Is It Still Possible to Advance the State of the Art from Observations in Practice?

183

Table 5: Database Modeling Tools.

Type Models Supported Notations Plataform License

ID Ferramenta DM FIDE DG EM C L F CF IDEF1x CN MN UML BN AN ON D W C FC

1 AnalyseSI X X X X X X X

2 Aqua Data Studio ER Modeler X X X X X X X

3 Astah X X X X X X X

4 brModelo X X X X X X X

5 Creately X X X X X X

6 Database Deployment Manager X X X X X X X

7 Database Workbench X X X X X X

8 DB Designer X X X X X

9 DB-Main X X X X X X

10 DBDesigner 4 X X X X X X X X

11 DBDesigner.net X X X X X

12 dbdiagram.io X X X X X

13 dbDiffo X X X X X X X X

14 dbForge Studio for MySQL X X X X X X X

15 DBSchema X X X X X X X X

16 DBVisualizer X X X X X

17 DbWrench X X X X X X

18 DeZign for Databases X X X X X X X

19 Dia X X X X X X X X X X X

20 dModelAid X X X X X

21 Enterprise Architect X X X X X X X X

22 ER-Assistant X X X X X

23 ER/Builder X X X X X X

24 ER/Studio Data Architect X X X X X X X X

25 ERD Concepts X X X X X X X

26 ERDesigner NG X X X X X X

27 ERDPlus X X X X X X

28 Erwin Data Modeler X X X X X X X

29 GenMyModel RDS X X X X X X

30 InfoSphere Data Architect X X X X X X X

31 Jeddict X X X X X X

32 ModelRight X X X X X X

33 MySQL Workbench X X X X X X X X X

34 N Data Modeler X X X X X X X X

35 Navicat Data Modeler X X X X X X X X X

36 Open ModelSphere X X X X X X X X X

37 Oracle SQL Developer Data Modeler X X X X X X X

38 pgModeler X X X X X X X

39 PowerDesigner X X X X X X X X X X X

40 QuickDBD X X X X X

41 RISE X X X X X X

42 Software Ideas Modeler X X X X X X X

43 SQL Database Modeler X X X X X X

44 SQL Maestro X X X X X X

45 SQL Power Architect X X X X X X

46 SQL Server Management Studio X X X X X

47 SQLDBM X X X X X X

48 SQLyog X X X X X X

49 Toad Data Modeler X X X X X X X

50 Valentina Studio X X X X X X

51 Vertabelo X X X X X

52 Visual Paradigm X X X X X X X

53 Win A&D X X X X X X

54 WWW SQL Designer X X X X X X

55 xCase X X X X X X

Caption: DM (Data Modeling), FIDE (Full IDE), DG (Diagramming), EM (Enterprise Modeling) | C (Conceptual), L (Logical) F (Physical) | CF (Crow’s Foot), CN (Chen’s Notation),

MN (Merise Notation), BN (Barker’s Notation), AN (Arrow Notation), ON (Other Notation) | D (Desktop), W (Web), C (Comercial), FC (Free of Charge)

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

184

in the literature.

External Validity: External threats often address

whether ﬁndings from one study can be generalized to

another domain. One reason for this threat would be

the occurrence of primary studies selection containing

incomplete information. However, it is likely that, as

this is an area of intersection between data modeling

and DSLs, the results cannot be generalized to other

research topics, thus reducing this risk naturally.

Conclusion Validity: One possible threat is bias in

data extraction, which leads to completion errors. To

alleviate this problem, we performed a careful read-

ing and tests of the use of the tools, and there was

the synthesis of data in electronic spreadsheets for a

better analysis.

7 RESEARCH PERSPECTIVES

This section presents the main research perspectives

on still open questions that researchers can explore in

the future.

Lack of Previous Research on the Topic: We easily

found in the literature different secondary studies of

domains in data modeling or DSLs. Notwithstanding,

before the execution of this MLM, we carried out an

ad hoc search for other research initiatives similar to

ours, but no results with exactly the same objectives

were obtained.

Absence or Limitation in the Evaluation of Study

Proposals: There is a consensus that software tech-

nologies need to be investigated and empirically eval-

uated and that primary studies should generally re-

port how their proposal is validated. However, the

lack of evaluation descriptions in the studies seen in

this MLM was frequent during the analysis studies.

It demonstrates a weakness in the methodology im-

plemented to carry out the studies that highlight the

great concern with development. It is also possible

to understand that activities inherent to Software En-

gineering that could add value to the quality of the

study proposal are also not discussed, including do-

main analysis for the creation of DSLs or their main-

tenance.

Automatic SQL Code Generation in Academic

Studies: The lack of emphasis on approaches to

transform data models is notable as opposed to tools

applied in the industry, although at least two stud-

ies deal with automatic SQL code generation for

databases. Although most of the identiﬁed DSLs

have good speciﬁcations, we did not ﬁnd initiatives

to model databases using textual DSLs. This fact may

indicate the odds this potential ﬁeld to become a rele-

vant research area.

8 FINAL REMARKS

Every year, several contributions to modeling (ER)

are published. Database modeling is a fundamental

area in Software Engineering, and DSLs that support

this activity are did not trivially ﬁnd in the literature.

Alternatives research for database modeling is nec-

essary to follow the evolution and trends of various

Database Management Systems (DBMSs). This pa-

per aims to provide an overview of the DSLs used

by ER modeling and the tools used in the industry

through an MLM. The principal motivation for this

research is that we believe that the usability of DSLs,

with their high power of domain expertise, is an el-

ement that can sharp help in modeling activities in

DBMSs, especially at a conceptual level.

This study covered 3513 distinct studies with the

intention of investigating primary studies that pro-

posed DSLs for modeling DBs. Likewise, we raised

a set of 67 different tools that support database mod-

eling, seeking to map their notations and supported

models. We detailed the MLM protocol, its conduc-

tion and subsequent, analysis of the results obtained.

In the end, we selected ten (10) primary studies and

55 tools to be analyzed quantitatively and qualita-

tively. As a result, only DSLs currently used to sup-

port ER modeling and tools used in database design

were rated. Among the results, the study by Dim-

itrieski et al. (2015) stands out, which presents a bidi-

rectional modeling tool that applies its DSL based on

the EER approach.

Finally, our research provides shreds of evidence

that each year a signiﬁcant of papers presenting dif-

ferent notations types are published, as seen in the

number of publications analyzed from the result re-

turned from the base string. It is somewhat surpris-

ing, given that the ER notations used today by indus-

try and academia, such as those by Peter Chen and

Barker, are not recent proposals. Therefore, we con-

cluded that ER modeling remains a wide ﬁeld of re-

search with a few gaps, e.g. the loss of tools that sup-

port the design in many lifecycles through automated

design techniques.

REFERENCES

ANSI, A. N. S. I. (1975). Interim Report: ANSI/X3/SPARC

Study Group on DBMSs 75-02-08. ACM SIG on Man-

agement of Data.

Ayadi, M. G., Bouslimi, R., and Akaichi, J. (2016). A

framework for medical and health care databases and

data warehouses conceptual modeling support. Net-

work Modeling Analysis in Health Informatics and

Bioinformatics.

Entity-relationship Modeling Tools and DSLs: Is It Still Possible to Advance the State of the Art from Observations in Practice?

185

Celikovic, M., Dimitrieski, V., Aleksic, S., Ristic, S., and

Lukovic, I. (2014). A DSL for EER Data Model Spec-

iﬁcation. In 23rd Int. Conf. on Information Systems

Development, pages 290–297.

Chen, P. P.-S. (1976). The Entity-relationship Model - To-

ward a Uniﬁed View of Data. ACM Transactions on

Database Systems, 1(1):9–36.

Cook, T. and Campbell, D. (1979). Quasi-Experimentation:

Design and Analysis Issues for Field Settings.

Houghton Mifﬂin.

Cougo, P. (2013). Modelagem conceitual e projeto de banco

de dados. Elsevier, Rio de Janeiro, RJ, Brasil.

Dimitrieski, V.,

Celikovi

c, M., Aleksi

c, S., Risti

c, S.,

Alargt, A., and Lukovi

c, I. (2015). Concepts and Eval-

uation of the Extended Entity-relationship Approach

to Database Design in a Multi-paradigm Information

System Modeling Tool. Computer Languages, Sys-

tems & Structures, 44(Part C):299–318.

Dybå, T. and Dingsøyr, T. (2008). Strength of evidence in

systematic reviews in software engineering. In 2nd

ACM-IEEE Int. Symp. on Empirical Software Engi-

neering and Measurement, pages 178–187. ACM.

Fowler, M. (2010). Domain Speciﬁc Languages. Addison-

Wesley, 1st edition.

Garousi, V. and Felderer, M. and Mäntylä, M. V. (2019).

Guidelines for including grey literature and conduct-

ing multivocal literature reviews in software engineer-

ing. Information & Software Technology, 106:101–

121.

Gregersen, H. and Jensen, C. S. (1999). Temporal entity-

relationship models-a survey. IEEE Transactions on

Knowledge and Data Engineering, 11(3):464–497.

Hammer, M. and Mc Leod, D. (1981). Database Descrip-

tion with SDM: A Semantic Database Model. ACM

Transactions on Database Systems, 6(3):351–386.

Jagannathan, D., Fritchman, B. L., Guck, R. L., Thompson,

J. P., and Tolbert, D. M. (1988). SIM: A Database

System Based on the Semantic Data Model. ACM

SIGMOD Record, 17(3):46–55.

Kersten, M., Zhang, Y., Ivanova, M., and Nes, N. (2011).

SciQL, a Query Language for Science Applications.

In Work. on Array Databases, pages 1–12, New York,

NY, USA. ACM.

Litwin, W., Abdellatif, A., Zeroual, A., Nicolas, B., and

Vigier, P. (1989). MSQL: A Multidatabase Language.

Information Sciences, 49:59–101.

Lorenz, M., Hesse, G., and Rudolph, J.-P. (2016). Object-

relational Mapping Revised - A Guideline Review and

Consolidation. In 11th International Joint Conference

on Software Technologies, pages 157–168.

Ma, Z. and Yan, L. (2010). A Literature Overview of Fuzzy

Conceptual Data Modeling. Information Science and

Engineering, 26:427–441.

Mazairac, W. and Beetz, J. (2013). BIMQL - an open query

language for building information models. Advanced

Engineering Informatics, 27(4):444–456.

Mernik, M., Heering, J., and Sloane, A. M. (2005). When

and How to Develop Domain-speciﬁc Languages.

ACM Computer Survey, 37(4):316–344.

Petersen, K., Feldt, R., Mujtaba, S., and Mattsson, M.

(2008). Systematic Mapping Studies in Software En-

gineering. In 12th Int. Conf. on Evaluation and As-

sessment in Software Engineering, pages 68–77. BCS

Learning & Dev. Ltd.

Shipman, D. W. (1981). The Functional Data Model and

the Data Languages DAPLEX. ACM Transactions

Database Systems, 6(1):140–173.

Siau, K. (2012). Innovations in Database Design, Web Ap-

plications, and Information Systems Management. In-

formation Science Reference.

Tian, H., Sunderraman, R., Calin-Jageman, R., Yang, H.,

Zhu, Y., and Katz, P. (2006). NeuroQL: A Domain-

speciﬁc Query Language for Neuroscience Data. In

Int. Conf. on Current Trends in Database Technology,

pages 613–624, Berlin, Heidelberg. Springer.

Torres, A., Galante, R., Pimenta, M. S., and Martins, A.

J. B. (2017). Twenty years of object-relational map-

ping: A survey on patterns, solutions, and their impli-

cations on application design. Information and Soft-

ware Technology.

Van Deursen, A., Klint, P., and Visser, J. (2000). Domain-

speciﬁc languages: An annotated bibliography. ACM

SIGPLAN Notices, 35(6):26–36.

Vara, J., Bollati, V., Jiménez, A., and Marcos, E. (2014).

Dealing with Traceability in the MDD of Model

Transformations. Trans. on Soft. Eng., 40(6):555–583.

Voelter, M. (2009). Best Practices for DSLs and Model-

Driven Development. Journal of Object Technology,

8(6):79–102.

Wohlin, C., Runeson, P., Hst, M., Ohlsson, M. C., Reg-

nell, B., and Wessln, A. (2012). Experimentation in

Software Engineering. Springer Publishing Company,

Incorporated.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

186