Assessment of EMF Model to Text Generation Strategies and Libraries
in an Industrial Context
Christophe Ponsard, Denis Darquennes, Valery Ramon and Jean-Christophe Deprez
CETIC Research Centre, Charleroi, Belgium
Keywords:
Model-based System Engineering, Model to Document, Tool Support, Industry Transfer, Eclipse Modelling
Framework.
Abstract:
Model-Based System Engineering is increasingly adopted. However it is centred on the notion of model while
the industry is still strongly document-based. In order to enable a smooth evolution between those paradigms,
the interplay between documents and models need to be managed, especially models can be used to efficiently
derive up-to-date documents matching standard templates from initial requirements to final certification. The
purpose of this paper is to review and to assess current strategies to manage model for document generators in
order to meet industrial requirements like document complexity, document/model scalability, quick generation
time, maintainability and evolvability of document templates. After exploring different generation strategies,
our work focuses on toolchains based on the Eclipse Modelling Framework and compares a few shortlisted
libraries in the light of our requirements and using document derived from industry cases.
1 INTRODUCTION
Because of its capability of abstraction/reasoning on
real world artefacts before building them, modelling
has been widely used across many engineering dis-
ciplines like construction, electronics or aeronau-
tics. Model-Based System Engineering (MBSE) is a
strong trend to apply modelling at system level and
across disciplines. For the industry, it is however a
paradigm shift because domain models become the
central part for the exchange between engineers rather
than documents. Model-Driven Engineering (MDE)
is a similar trend focusing only on the software devel-
opment process (Schmidt, 2006). Such approaches
can rely on standardised and widely adopted mod-
elling languages like SysML (OMG, 2005) at system
level, UML (OMG, 1997) for software and increas-
ingly Domain Specific Languages (DSLs). They pro-
vide a visual syntax enabling the design and commu-
nication activities of the engineers but also have pre-
cise semantics to enable automation of parts of the
System Development Life Cycle (SDLC).
Being able to synchronise model and text is a key
element for the industrial adoption of model-based
toolchains because documents are and will certainly
remain central to industrial development and valida-
tion workflows for a long time, even though the goal
is to progressively reduce them. Documents will stay
mandatory for the identification of requirements, re-
view and validation at various SDLC steps and for
specific activities like certification. This synchroni-
sation involves different interactions between models
and documents, like identifying concepts from docu-
ments, generating documents from models or keeping
both artefacts synchronised over the system evolution.
In our previous work, we have investigated different
kinds of such interplays using diagrams, tables and
texts (Ponsard et al., 2015) and for re synchronising
models from documents (thus achieving rountrip with
document generation) (Ponsard and Darimont, 2017).
This work focuses on model to text document
generation. This may seem easy to achieve at first
sight but requires to fulfil a number of key require-
ments, both functional and non-functional, in order
to meet industry standards. Based on our experience
with industrial customers, we could identify such re-
quirements and also evaluate how currently available
technologies fulfil them. Then, we review different
document generation strategies with different possi-
ble trade-offs, e.g. favouring multiple output formats
through an intermediary format/API or directed to-
wards a specific format with more features and per-
formance. Our work is more specifically addressing
modelling solution based on the Eclipse Modelling
Framework (EMF) because it is an open industrial
standard and also because it fitted our industrial con-
Ponsard, C., Darquennes, D., Ramon, V. and Deprez, J.
Assessment of EMF Model to Text Generation Strategies and Libraries in an Industrial Context.
DOI: 10.5220/0009155804330440
In Proceedings of the 8th International Conference on Model-Driven Engineering and Software Development (MODELSWARD 2020), pages 433-440
ISBN: 978-989-758-400-8; ISSN: 2184-4348
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
433
text. However, the same requirements and evaluation
process may be reused for other modelling technolo-
gies, including closed ones. The last part of our work
further investigates and compares a few solutions. It
benchmarks prototypes against a test suite covering
the target requirements and relying on real-world doc-
uments produced by industrial partners in two real-
world and complementary toolchains, one Eclipse-
based (Capella) (Polarsys/CIS, 2015) and the other
web-based with a SaaS model (Respect-IT, 2005; Da-
rimont et al., 2016). Note however that for confiden-
tiality purposes excerpts presented are based on the
public in-flight entertainment model (Bonnet, 2015).
This paper is structured as follows. Section 2 re-
minds about key industry requirements. Then Sec-
tion 3 surveys possible generation strategies based on
both literature and industry practices. Section 4 de-
tails different candidate libraries focusing on direct
generation for an EMF toolchain. Section 5 presents
our benchmarking process, test data and raw results
which are analysed in Section 6. Finally Section 7
concludes and discusses some complementary work.
2 INDUSTRY REQUIREMENTS
FOR DOCUMENT
GENERATION
In order to base our comparison on sound crite-
ria, key industrial requirements have been gathered
and grouped in two categories: functional and non-
functional. Criteria identified in bold are evaluated
later in this paper and are independent of any specific
context (e.g. EMF).
2.1 Functional Requirements (FRs)
Industrial specifications are complex in nature and
are characterised by a well-defined structure usually
based on templates. An elegant way to introduce
MBSE is to use enriched templates specifying how
to populate the document, in the same format of ex-
isting corporate templates. The document generator
needs to support a variety of standard formats used
in the industry, typically Microsoft (docx) or LibreOf-
fice (odt). Non editable formats like PDF and HTML
may be useful too although easy to derive. Other tar-
get formats may also be interesting, e.g. spreadsheets
or presentations. On the model side, our focus is the
support of Eclipse Modelling Framework standard.
As shown in Figure 1, such documents also usu-
ally combine
Figure 1: Excerpt of a complex specification document mix-
ing text, diagram and tables.
text descriptions, e.g. for requirements or func-
tion descriptions with specific writing conven-
tions. This typically requires the use of rich text
(as bold, italic, underline, bullets...).
figures, where especially the inclusion of dia-
grams from the model should be easy, using stan-
dard formats (e.g. png, svg) and with a good qual-
ity.
tables, to more systematically present well-
structured concepts (e.g. requirements tables with
identifier, definition, priority, effort...), system de-
composition (subcomponents, interface compo-
nents) or some specific view (e.g. traceability
matrix). Although tables can be managed using
spreadsheets, the direct generation of tables is
mandatory.
The overall structure and consistency requires auto-
mated section numbering and styles. Other impor-
tant features are the generation of bookmarks to ease
the navigation in large documents. Table of contents
are already managed by standard templates.
2.2 Non-Functional Requirements
(NFRs)
The following classification is inspired from the
SQuaRE classification (ISO, 2014) used as checklist,
key NFRs are:
Costs relates to a number of attributes. About ac-
quisition costs, the library maybe Free of charge
or even Open Source. However more important
cost factors relate to the adaptation and mainte-
nance of the solution which may require long term
support. This can be evaluated by the global Ma-
turity of the product (known bugs, documenta-
MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development
434
tion, tutorials...) and its support, either commer-
cial or through a community.
Scalability is a mandatory requirement. The size
of industrial system means large model and thus
the ability to generate large documents gener-
ally also based on complex templates.
Efficiency is less critical in case of batch/server
based generation. However, for laptop generation
or when frequently update of a work document or
developing a template, good speed of generation
(i.e. under a minute) is important. Reducing the
generation scope may also be used to accelerate
modelling/generation cycles.
Usability is important for industrial adoption. We
should not assume the developer/maintainer of
a generator template has a programming back-
ground and favour declarative and easy to learn
tools. Two complementary levels need to be
easy to use: how to retrieve information from the
model and how to present that information in the
document under the adequate form (see FR). This
requires a good learning curve of the associated
language (preferably standard like OCL), good
writing support for both of the above levels and
good readability, i.e. to maintain and evolve the
generator template, preferably with some mod-
ularity feature. In order to invoke the tool, a
good user interface integration with the mod-
elling tool is required while a portable API (e.g.
Java) is also useful for non UI invocation, e.g. for
batch/nightly document regeneration.
3 MODEL TO DOCUMENT
GENERATION STRATEGIES
Different strategies may be considered to produce a
document from a model. In this section, we review
the available design space and give some high-level
guidelines that can help drive the choice based on typ-
ical trade-offs that need to be considered.
Template-based Generation. The most obvious
choice is to use a single transformation tool able to
convert a model into the target document format based
on a transformation specified through some template
which can take the form of a specific format (e.g.
XML with a specific schema) or a Domain Specific
Language (DSL). This simple design helps to give
good control on the available features and perfor-
mance. However, restricted choices may be available
for the model source/target document format combi-
nation. If multiple document formats need to be con-
sidered, developing independent converters for each
format is not a good idea because the template design
will not be reused. An additional converter might be
chained to transform the target document into other
formats if the converter is reliable enough (e.g. to
produce a PDF from most editable formats). In this
context, we obviously restrict to Model-to-Text gen-
eration tools. This transformation can be managed by
an independent library, e.g. Gendoc (Eclipse, 2018)
or M2Doc (Obeo, 2019) for EMF) or directly inside a
Word processor through internal macro/scripting. The
latter is not recommended because the scripting lan-
guage is not effective and maintainable for such trans-
formation and for accessing a model repository. An
editor is also not optimised for generation and should
not be required at this step.
Generic Reporting Tools. Efficient reporting tools
exist and are able to generate many different tar-
get document formats either editable (office docu-
ments but also spreadsheets or even presentations) or
read-only (PDF). Birt (Eclipse Foundation, 2005) and
Jasper (Jaspersoft, 2005) are examples of such plat-
forms. They usually rely on an internal report descrip-
tion format that is populated from one or several data
sources. The usual scenario is rather to generate busi-
ness reports from spreadsheets or relational databases.
However, data sources can be extended through ad-
ditional drivers through a protocol such as the Open
Data Access (ODA) which includes support for con-
necting, accessing meta-data, and executing queries.
It is not restricted to classical relational databases and
can also be used for model-based repositories. For ex-
ample, an EMF connector is available and can be used
together with Birt (Eclipse Foundation, 2008) with
queries in OCL. The developed template can then be
run in server mode. The performance might be im-
pacted by the efficiency of the query engine. Alter-
natively, an intermediate data file in a specific format
might also be produced from the model and fed more
efficiently in the reporting tool.
Existing Modelling Platforms. Some modelling
platforms integrate reporting capabilities and some
also support model interchange, e.g. the MagicDraw
tool can process EMF and also generate files us-
ing Velocity Template Library (Apache Foundation,
2006). However, this setting is hard to automate and
reuse outside of this specific toolchain even though
some parts might be open source.
Relying on an Intermediary Format. Is an inter-
esting option because such a format can stay abstract
enough to capture the essence (semantics) of the re-
Assessment of EMF Model to Text Generation Strategies and Libraries in an Industrial Context
435
port without focusing on the concrete presentation.
From that format, different target formats may be gen-
erated. Reporting tool template follows that princi-
ple but other options are possible and an open for-
mat should be favoured. For example, XML Doc-
Book (Walsh and Hamilton, 2010) format can be
processed using W3C Formatting Object (XSL-FO)
and then to standard format such as PDF and RTF
to some extend. Such an approach was applied by
JPL on MagicDraw (Delp et al., 2013). Another
more lightweight option is to use a variant of mark-
down language relying on plain text formatting syn-
tax (Gruber, 2004). Based on this format, many
concrete document formats can be generated using
universal transformers such as Pandoc (MacFarlane,
2006). Of course, the final quality should be care-
fully assessed w.r.t. all required features described in
Section 2. An interesting variant here is also to intro-
duce a meta-model for the document itself which en-
ables to use generally more powerful Model-to-Model
transformation like the QVT Query/View/Transform
standard (OMG, 2011) for dealing with the transfor-
mation with an additional serialisation to the textual
output format (Willink, 2018). A more elaborated
form of intermediary is to use a Domain Specific Lan-
guage directly embedding the model transformation
language. pxDoc detailed later is based on this ap-
proach (Pragmatic Modeling, 2019).
An open discussion point is whether the target for-
mat should be editable or not. While the latter ensures
all changes are made inside the model, most of the
time an editable format is preferred to enable integra-
tion, style adaptations and unfortunately some correc-
tions which are then lost unless some synchronisation
mechanism is used (Ponsard and Darimont, 2017).
Finally note that although we tried to provide a
consistent classification, the above categories are not
mutually exclusive. For example, generic reporting
tools also rely on a template. Exiting tools may use an
intermediary format or use an external generation li-
brary. This cannot always be assessed (e.g. for closed
tool). A key point guiding the classification design
was the kind of interface and integration capability.
4 TEMPLATE-BASED
GENERATION LIBRARIES
The rest of this document will focus on the single step
generation strategy. Although it is less able to cope
with multiple formats, it was selected for further study
in both of our industrial cases. After discarding low-
level generations libraries lacking any modelling sup-
port such as POI or VTL (Apache Foundation, 2006),
we shortlisted three candidates which are presented
hereafter. All of them rely on OCL (or variant) as
model query language.
4.1 Gendoc
Gendoc (Eclipse, 2018) is an Open Source library de-
veloped by the Eclipse Foundation under its EPL li-
cense. It supports the generation of documents based
on EMF models in different formats from Microsoft
Office (docx, xslx, pptx) and Libre Office (.odt). Gen-
doc relies on existing templates for the general struc-
ture and styling while the content is specified using
XML tags embedding commands for generating rich
text, diagrams and tables. It strongly relies on the
Acceleo Query Language (AQL), an OCL variant to
dynamically retrieve content from the model. It is
also compatible with the Papyrus (Eclipse, 2010) and
Capella (Polarsys/CIS, 2015) modellers.
Figure 2: Fragment of Gendoc template.
Figure 2 shows an example of template built using
Gendoc. It is divided in two parts respectively marked
by the context and gendoc tags. The context con-
tains the absolute path to the model, the searchMeta-
model and importedBundles information. The gen-
doc section contains a static text ”List of entities”,
followed by a variable self.name which is the name
of the model. A for loop lists inside the model the
names of the entities resulting from the AQL request
allInstances()->sortedBy(name)”. Figure 3 shows
the result generated from this fragment.
Figure 3: Result of the fragment of Gendoc template.
MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development
436
4.2 M2Doc
M2Doc (Obeo, 2019) is an Open Source library (EPL)
developed by Obeo to generate documents from EMF
models. Like Gendoc, it relies on the combination of
a document template with AQL. The main differences
are the limitation to Word documents (.docx) and the
use of special fields of the form { m: <content>
} rather than XML tags for retrieving the model ele-
ments, resulting in a better readability. M2Doc relies
on ApachePOI for creating docx files. Figure 4 shows
an example of fragment of M2Doc template function-
ally equivalent to the Gendoc fragment presented in
Figure 2 and producing the same result as in Figure 3.
The elaborated document presented in Figure 1 was
also produced using M2Doc from the public in-flight
entertainment model (Bonnet, 2015).
Figure 4: Fragment of M2Doc template.
4.3 pxDoc
pxDoc (Pragmatic Modeling, 2019) is a commercial
library used to generate Microsoft Word documents
from EMF models. It requires an end-user license.
Contrary to Gendoc and M2Doc, it does not extend
an existing document template with queries to retrieve
model content. Instead, a textual Domain Specific
Language (DSL) is used for transforming the model
into text and applying styles referenced from an ex-
isting and unmodified template. Figure 5 shows how
to retrieve the same content than in the Gendoc and
M2Doc sections, by using model navigation expres-
sion in a format closed to OCL and AQL queries.
Styling is applied using styling reference like #Ti-
tle1 and specific command like numbering to control
Figure 5: Fragment of pxDoc template.
the numbering level. The DSL is implemented with
EMF and XText technologies that are completely in-
tegrated with the Eclipse development environment
and Eclipse-based modellers like Papyrus (Eclipse,
2010) and Capella (Polarsys/CIS, 2015).
Compared to GenDoc and M2Doc, pxDoc is
closer to coding and has a direct integration with Java
and Eclipse. The DSL looks like a simplified form
of Java and is not complex to learn. The DSL edi-
tor provides productivity features like syntax check-
ing and content assist, including to navigate into the
target model when writing a query. This greatly ac-
celerates the development of a document generator. It
also does not require Word in this phase.
5 ASSESSMENT OF SELECTED
GENERATION LIBRARIES
This section presents our assessment of the libraries
shortlisted in Section 4. Our assessment was based
on prototyping generation of realistic industrial doc-
uments without prior knowledge of the libraries. For
the models and templates, we used the publicly avail-
able in-flight entertainment system (Bonnet, 2015)
and models provided by two industrial partners: one
producing large design documents for railways sys-
tems and the other producing very elaborated require-
ments documents for large call for tenders.
Table 1 and 2 present the assessment respectively
for the functional and non-functional criteria defined
in Section 2. Criteria are ranked either on a binary
Table 1: Comparison of selected libraries for the functional criteria.
Gendoc M2Doc pxDoc
EMF support x x x
Standard formats supported docx, odt, docx docx
xlsx, pptx
Use of enriched templates x x -
Rich text inclusion x - x
Diagram inclusion x x x
Table generation x x x
Section numbering x x x
Style support x x x
Bookmark/hyperlink generation x x x
Assessment of EMF Model to Text Generation Strategies and Libraries in an Industrial Context
437
Table 2: Comparison of selected libraries for the non-functional criteria.
Gendoc M2Doc pxDoc
Cost
Free of charge x x -
Open Source x x x
Maturity ++ + ++
Scalability + ++ +++
Efficiency
Large document + ++ +++
Complex template + ++ +++
Usability
Learning curve + + -
Writing support + ++ ++
Readability + +++ ++
Interface integration ++ + ++
Figure 6: Comparison of generation time for pure text document (left) and documents with figure (right).
scale (fulfilled=”x” and not fulfilled=”-”) or on a qual-
itative scale (from ”—” for very poor to ”+++” for
very good). For the scalability criteria, quantitative
benchmarks were run but are reported on a qualitative
scale with some additional comments. Note that due
to the preselection, the global ranking is rather good.
Table 1 shows that three preselected libraries pro-
vide good coverage of all functional criteria. A
weaker point for M2Doc is the current lack of support
for rich text import. Gendoc is the only to support the
Open Source odt format. Table 2 shows more con-
trasted results for the non-functional criteria that are
further discussed in Section 6.
About efficiency, our prototypes were bench-
marked using the same models to produce large doc-
uments (up to 1000 pages) on a standard laptop
(Core I7 2.8GHz with 16GB RAM running Windows
10). We considered simple templates (mostly sec-
tion, subsections and lists) and more complex tem-
plates (including figures and tables resulting from
complex queries). Figure 6 summarises our compar-
ison for Gendoc (V0.7.2), M2DOC (V2.0.2) and px-
Doc (V1.1.1):
for pure text documents, the generation time re-
mains acceptable in all cases: less than 1 minute
globally pxDoc is faster, probably due to the over-
head for interpreting AQL in M2Doc and Gendoc.
Gendoc is slower, especially for generating large
documents with a complex template. For 1000
pages, it did not crash but took 15 minutes. For
M2Doc and pxDoc template complexity has a
limited impact resulting in the same curves.
when considering figures, the results are slower
and more balanced: between 3 and 4 minutes for
500 pages. This is probably due to the similar fig-
ure transfer time from Capella to the document.
Gendoc is slightly better than M2Doc here.
MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development
438
6 DISCUSSION: WHAT TO USE
AND IN WHICH CONTEXT ?
We provide here some guidelines to help in the se-
lection of libraries within the scope of the libraries
studied in Section 3. Section 2 gives more general
guidelines about other possible strategies but without
digging in concrete solutions. More detailed guide-
lines are out of the scope of this paper but part of our
future work.
Globally, pxdoc is quite different because it does
not extend existing templates but it requires a licence
and to learn a DSL. However it provides very effec-
tive tool support for writing templates. The solution
can be suitable for a big company which can afford a
license (e.g. a few floating might be enough) and with
usually quite stable templates and a dedicated trained
person in charge of the corporate templates.
Gendoc is a good solution for not too overly large
documents. However, XML tags are not easy to read
and also reduce the maintainability. The direct sup-
port for odt is a strong point if this is a standard for-
mat adopted by the company. It allows a fully Open
Source toolchain. The long term support needs to be
monitored as the project does not seem very active.
M2Doc is a more recent solution which still lack
some features like rich text and adapted if the focus
is on the docx format only. It is quite efficient and
easy to learn and maintain due to the lighter form
of embedding. The use of a docx template requires
careful writing and a few cycle of debugging which is
decently supported by errors reported in a validation
document or directly inside the generated document.
The library is still actively evolving, although the last
version is quite behind schedule.
About the dependency over Eclipse, M2Doc is
quite easy to isolate and run outside of Eclipse, e.g.
to develop a web-based tool. On the contrary, Gendoc
has intricate Eclipse dependencies difficult to isolate.
A workaround lacking efficiency and robustness is to
run Gendoc as console-mode Eclipse service.
7 CONCLUSION AND FUTURE
WORK
In this paper, we stated key industrial criteria for
model-based document generation. Based on those
criteria, we first identified and discussed possible doc-
ument generation strategies with different benefits
and drawbacks. Then, we focused on a direct gener-
ation strategy and assessed the best libraries for this
purpose and an EMF modelling context. As a re-
sult, we produced guidelines that we believe useful
for driving the selection of a generation solution in an
EMF context and, more generally, based on the iden-
tified criteria and assessment process.
In our future work, we plan to extend our bench-
marking to compare different strategies especially
with a generic reporting tool and an intermediary
markdown format. Different quality attributes will be
quantified in order to help selecting the best approach
given a specific industrial context. We will also keep
monitoring the evolution of the identified libraries and
also study new ones like Papyrus Model2Doc which
is announced (Lorenzo, 2019). Based on this we
could provide more extensive guidelines to help in the
decision process of a document generation solution in
the context of a set of specific requirements among
those we identified.
ACKNOWLEDGEMENTS
This research was partly supported by the IDEES Co-
innovation project (nr. ETR121200001379) of the
Walloon Region. We thank Respect-IT for its indus-
trial cases and feedback.
REFERENCES
Apache Foundation (2006). Velocity Template Library.
https://velocity.apache.org/engine/1.7/vtl-reference.
html.
Bonnet, S. (2015). In-flight entertainment system model.
https://wiki.polarsys.org/Capella/Samples/IFE.
Darimont, R., Zhao, W., Ponsard, C., and Michot, A.
(2016). A modular requirements engineering frame-
work for web-based toolchain integration. In 24th
IEEE Int. Req. Eng. Conf., Beijing, China, Sept.
Delp, C., Lam, D., Fosse, E., and Lee, C.-Y. (2013). Model
based document and report generation for systems en-
gineering. In IEEE Aerospace Conference.
Eclipse (2010). Papyrus Modeler. https://www.eclipse.org/
papyrus.
Eclipse (2018). Gendoc - Generation of Documentation
from EMF models using document templates. https:
//www.eclipse.org/gendoc.
Eclipse Foundation (2005). Business Intelligence and Re-
porting Tools. https://www.eclipse.org/birt.
Eclipse Foundation (2008). Open data access driver. https:
//wiki.eclipse.org/Ecore ODA Driver.
Gruber, M. (2004). Markdown. https://daringfireball.net/
projects/markdown.
ISO (2014). 25000 systems and software quality require-
ments and evaluation (square). https://www.iso.org/
standard/64764.html.
Assessment of EMF Model to Text Generation Strategies and Libraries in an Industrial Context
439
Jaspersoft (2005). Jasper Reports. https://community.
jaspersoft.com/project/jasperreports-library.
Lorenzo, V. (2019). EclipseCON, Munich (accepted).
MacFarlane, J. (2006). Pandoc - a universal document con-
verter. https://pandoc.org.
Obeo (2019). M2Doc - generation of Office Open XML
documents from EMF models and document template.
http://www.m2doc.org/ref-doc/2.0.2/.
OMG (1997). Unified modeling language. http://www.omg.
org/spec/UML.
OMG (2005). System modeling language. http://www.omg.
org/spec/SysML.
OMG (2011). MOF Query/View/Transformation. https://
www.omg.org/spec/QVT/1.1.
Polarsys/CIS (2015). Capella - open source solution
for model-based systems engineering. https://www.
polarsys.org/capella.
Ponsard, C. and Darimont, R. (2017). Improving Re-
quirements Engineering through Goal-oriented Mod-
els and Tools: Feedback from a Large Industrial De-
ployment. In Proc. 12th Int. Conf. on Soft. Tech., IC-
SOFT, Madrid, Spain, July 24-26.
Ponsard, C., Darimont, R., and Michot, A. (2015). Com-
bining Models, Diagrams and Tables for Efficient Re-
quirements Engineering: Lessons Learned from the
Industry. In INFORSID 2015, Biarritz, France.
Pragmatic Modeling (2019). pxDoc - generate Word docu-
ments from Java. https://www.pxdoc.fr.
Respect-IT (2005). The Objectiver Goal-Oriented Require-
ments Engineering Tool. http://www.objectiver.com.
Schmidt, D. C. (2006). Guest editor’s introduction: Model-
driven engineering. Computer, 39(2):25–31.
Walsh, N. and Hamilton, R. L. (2010). DocBook 5: The
Definitive Guide. O’Reilly, Sebastopol, CA.
Willink, E. D. (2018). A text model - use your favourite
M2M for M2T. In Proceedings of MODELS 2018
Workshops, Copenhagen, Denmark, October.
MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development
440