Generic and Meta-Transformations for Model Transformation Engineering

Generic and Meta-Transformations
for Model Transformation Engineering
Dániel Varró, András Pataricza
Budapest University of Technology and Economics
Most Influential Paper Presentation@MODELS 2014
Valencia, Spain, October 2nd, 2014
Budapest University of Technology and Economics
Department of Measurement and Information Systems

Contributions of our UML 2004 paper
Generic&Meta
40%
VIATRA
27%
MT Engineering
10%
Other
6%
None
9%
Self
8%
Citations
 Novel MT Concepts
o Generic and
o Meta-Transformations
 Highlight the software
engineering aspects of MT
o Reusability
o Maintainability
o Performance
o Compactness
 1st prototype of VIATRA2
transformation framework
 Impact (Citation) Analysis
o Total: 137 (Google Scholar)
2

Overview
UML 2004 talk
Outlook on Generic and Meta-Transformations
VIATRA History
3

(Slides with this design are taken from my UML 2004 talk)
Generic and
Meta-Transformations for
Model Transformation Engineering
Dániel Varró
András Pataricza
Budapest Univ. of Technology and Economics
Dept. of Measurement and Information Systems

The Model Transformation Problem
UML DB
Language A Language B
Metamodel A
How to specify?
conforms to conforms to
How to execute?
October 14th, 2004 UML 2004
Metamodel B
Model a Model b
Platform Indep. Trans. (PIT)
J. Bézivin, N. Farcet, J.-M. Jézéquel, B. Langlois, and D. Pollet.
Reflective model driven engineering. In Proc. UML 2003:
6th Int. Conf. on the Unified Modeling Language, LNCS 2863,
pp. 175–189. Springer, San Francisco, CA, USA, 2003

Solution 1: Manually written PSTs
UML DB
Metamodel A
Metamodel B
Model a Model b
Platform Indep. Trans. (PIT)
Platform Spec. Trans. (PST)
XSLT spec. XT
Schema SA Schema SA
Document da Document db
XSLT exec. xt

Solution 1: Manually written PSTs
UML DB
Advantages:
+ standard model format (XMI) and transform. language (XSLT)
Metamodel A
Metamodel B
+ extensive tool support (XSLT engines)
Disadvantages:
– knowledge is required Model a in another technological space Model (XSLT)
b
– hard to debug/maintain/understand complex XSLT scripts
– XSLT performance problems (for graph-like structures, e.g. XMI)
MOF technological space
Platform Spec. Trans. (PST)
XSLT spec. XT
Schema SA Schema SA
XSLT exec. xt

Solution 2: PIT Specification
UML DB
Metamodel A
Transf. Spec. T
Metamodel B
QVT
Model a Model b

Solution 2: PIT specification + engine
UML DB
Metamodel A
Transf. Spec. T
Metamodel B
Model a Model b
Transf. Exec. t

QVT example
QVT rule Relation
C:Class
:attrs
Pattern
:columns
A:Attribute
name=NM
type=TP
B:Column
name=NM
type=TP
Cls2Tbl(C,T)
T:Table
Attr2Col(A,B)
Attr2ColumnR
when
Condition
Variable

Graph transformation example
LHS
C:Class
:attrs
A:Attribute
name=NM
type=TP
T:Table
Cls2Tbl(C,T)
:attrs
C:Class
T:Table
Cls2Tbl(C,T)
:columns
A:Attribute
name=NM
type=TP
B:Column
name=NM
type=TP
RHS
Att2Col(A,B)

Solution 2: PIT specification + engine
UML DB
Metamodel A
Transf. Spec. T
Metamodel B
Model a Model b
Transf. Exec. t
Advantages:
+ pattern-based transformation specification in UML/MOF
(upcoming QVT standard)
+ increased performance for complex graph models
Disadvantages:
– lack of (industrial) model transformation tools

Solution 3: Model-driven transformation of
PSMs
UML DB
Metamodel A
Transf. Spec. T
Transf. Exec. t
Metamodel B
Model a Model b
Schema SA Schema SA

Solution 3: Model-driven transformation of
PSMs
UML DB
Metamodel A
Transf. Spec. T
Transf. Exec. t
Advantages: (as before)
+ pattern-based transformation XSLT spec. specification XT
in UML/MOF
(upcoming QVT standard)
+ increased performance for complex graph conforms models to
(vs. XSLT)
conforms to Disadvantages:
– HOW TO INTEGRATE INTO EXISTING TOOLS?
Metamodel B
Model a Model b
Schema SA Schema SA
XSLT exec. xt

Solution 4: Automatically generated PST
UML DB
Metamodel A
Metamodel B
Transf. Spec. T
Model a Model b
Meta-transformation
XSLT spec. XT
Schema SA Schema SA
XSLT exec. xt

Solution 4: Automatically generated TST
UML DB
Advantages:
Transf. Spec. T
+ transformation specification in QVT/MOF/UML
Metamodel A
Metamodel B
+ automatically generated XSLT scripts
+ platform specific transformers
(integrated into the target technological space)
Disadvantages:
– PIT2PST is a complex meta-transformation (e.g. QVT2XSLT)
– we have to do it ONCE / we HAVE TO do it once
Model a Model b
XSLT spec. XT
Schema SA Schema SA
XSLT exec. xt

Transformation engineering in MDA
 Transformation design will be an
engineering principle (if MDA succeeds)
 Questions to answer:
– How to specify and execute MTs?
– Compactness / Intuitiveness of MT Specs?
– Automation? - Performance?
– Analysis? - Correctness of transformations?
– Reusability? - Maintainability?

Objectives
 How to tackle (some) transware problems?
– By storing/handling transformations
as models
 Generic (higher-order) transformations:
– Type / class variables in rules
 Meta transformations:
– Consumes (produces) transformation rules as
input (output)
Higher-Order
Transformations

Example: XMI export
 Problem:
– generate XMI documents
– from MOF-based models
– when the metamodel is
a parameter
n1 : Node
name = First
isStart = true
n2 : Node
name = Second
isStart = false
outgoing
e1 : Edge
weight = 1
incoming
Node
subnodes
name : String
isStart : Boolean
Edge
outgoing
weight : Integer
incoming
Graph
<Node xmi.id = "n1">
<Node.name>FirstNode</Node.name>
<Node.isStart xmi.value="true" />
<Node.outgoing>
<Edge xmi.idref = "e1"/>
</Node.outgoing>
<Node.subnodes>
<Node xmi.id = "n2"> ... </Node>
</Node.subnodes>
</Node>

Overview of
generic transformations
 Advantages
– General + Powerful
– Succinct
Generic
rules
 Disadvantages
– Performance?
– Implementation
as PSTs?
Metamodel
(Graphs)
Graph
model
Generic
Trans. Eng.
XMI
document
Generic transformation = Rules with type variables

Example: A generic rule
x:Class
name = n
o:x
identifier = id
e:XMLElem
name = n
:attributes
a:XMLAttr
name = "xmi.id"
value = id
Obj2Elem(o,e)
{new}
{new}
{new}
{new}
Obj2ElemR
Type
parameter

Overview of
meta-transformations
 Advantage:
Meta
rules
– Executed by traditional
transformation engines
 Prerequisite:
– Rules stored as models
Metamodel
(Graphs)
Graph
model
“Ordinary”
Trans. Eng.
XMI
document
“Ordinary”
rules
“Ordinary”
Trans. Eng.
Meta-Transformation = Rules that consume/generate rules
Transformation
Models

Example: A meta rule
x:Class
name = cl
as:attributes
att:Attribute
name = at
:type
y:Class
isEnum = "true"
EnumAttr2ElemR(x,att)
o:x
a:att
value = v
:as
e1:XMLElem
{new} :contains
{new}
e2:XMLElem
name = cl+"."+at
:attributes
a2:XMLAttr
name = "xmi.value"
value = v
Obj2Elem(o,e1)
Attr2Elem(a,e2)
{new}
{new}
{new}
Attr2Rule
Attr2RuleR
Type variables are data in the RHS resolved prior to main transformation

Result of applying a meta rule
EnumAttr2ElemR(“Edge”,”weight”)
o:Edge
:as
a:weight
value = v
Obj2Elem(o,e1)
e1:XMLElem
{new} :contains
{new}
e2:XMLElem
name = "Edge.weight"
{new} :attributes
{new}
a2:XMLAttr
name = "xmi.value"
value = v
Attr2Elem(a,e2)
{new}
Ordinary (first order) transformation rule without type variables

Generic rules for model maintenance
a:att:Attribute
e1:XMLElem
e2.XMLElem
:attributes
a1:XMLAttr
:contains
name = n
name = "xmi.value"
value = v
Attr2Elem(a,e2)
{delete}
:contains
a2:XMLAttr
name = n
value = v
Attr2Attr(a,a2)
{new}
{delete}
{delete}
{delete}
{delete}
{new}
{new}
ChangeXMIAttributesR
Migration from XMI 1.0 to XMI 1.1

Another scenario of
generic and meta-transformations
 Advantage:
uniform management of
– models
– transformations
“Ordinary”
rules
Graph
model
Generic &
Meta rules
Generic
Trans. Eng.
XMI
document
“Ordinary”
rules
“Ordinary”
Trans. Eng.
Meta-Transformation = Rules that consume/generate rules
Generic Transformation = Rules with type variables

The VIATRA2 Approach
 Model management:
o Model space: Unified, global view of
models, metamodels and transformations
• Hierarchical graph model
• Complex type hierarchy
• Multilevel metamodeling
 Model manipulation and transformations:
integration of two mathematically precise,
rule and pattern-based formalisms
o Graph patterns (GP): structural conditions
o Graph transformation (GT): elementary transformation steps
o Abstract state machines (ASM): complex transformation programs
 Code generation:
o Special model transformations with
o Code templates and code formatters

Generic GT rules in VIATRA
precondition pattern lhs(Par,Child,P1,ClsE,ParR,AncR) =
{
entity(ClsE);
relation(ParR,ClsE,ClsE);
relation(AncR,ClsE,ClsE);
entity(Par);
entity(Child);
relation (P1, Child,Par);
instanceOf(Par, ClsE);
instanceOf(Child, ClsE);
instanceOf(P1, ParR);
} …
 Explicit (dynamic)
instanceOf relations
 Variables matched
to metamodel
elements (types)
gtrule parentIsAncR(Par,Child,ClsE,ParR,AncR)
precondition lhs(Par, Child, P1
ClsE,ParR,AncR)
Par:entity
ClsE:entity
P1:relation
AncR:
relation
Child:entity
ParR:
relation
postcondition rhs(Par, Child, P1
ClsE,ParR,AncR)
Par:entity
ClsE:entity
P1:relation
AncR:
relation
Child:entity
ParR:
relation
E:relation

Meta-transformations in VIATRA
 Main ideas:
o Transformations stored as
models
o Processed by an interpreter
(GT + ASM)
o Executed over models
stored in the model space
 Unusual consequences:
o Self-modifying meta-transformations
o Pattern elements appearing
in match sets
Scopes of MT rules required
EMF transformation model
 Quiz: Sample model space
o Graph pattern: chase
o Instance model
How many matches does pattern
chase have in the model space? 6
29

Higher-Order Transformation
Approaches in Tools: An Outlook
With direct contributions from M. Tisi, C. Krause, E. Syriani,
H. Vangheluwe, D. Kolovos

Our Influencers (in 2004)
• Introduced for the definition of
context-sensitive language grammars
Two-level
graph
grammars
• Meta-theoretical foundations in Maude
• Arbitrary models can be transformed
into their meta-representation
Rewriting
logics
• Reflective OO programming (Smalltalk)
• Functional programming
Programming
languages
31

Categorization of HOTs
MT analysis
• Process transformations to
extract meaningful data
MT synthesis
• Create new transformations from
data modeled in other forms
MT modification
• Manipulate the logic of an input
transformation
MT (de)composition
• Merge /split other MTs according
to a (de)composition criterion
M. Tisi, F. Jouault, P. Fraternali, S. Ceri, and J. Bézivin. On the Use of Higher-Order Model
Transformations. In Proc. of the 5thEuropean Conference on Model-Driven Architecture
Foundations and Applications (ECMDA2009), pp 18–32
33. Springer, 2009

ATL
Transformation
injection
• Textual
transformation is
translated into a
transformation
model
Higher-order
transformation
• Transformation
models for
• Input+output+HOT
Transformation
extraction
• Output model is
serialized into a
textual
transformation
33

Henshin
 Generic, dynamically typed MT
o Transform metamodels and
instance models
at the same time
o Use special wrapper objects
 Meta/Higher-order MT
o Henshin allows to mix
normal and HO
transformations
o E.g. Ecore2GenModel
(customization)
C. Krause, J. Dyck, H. Giese: Metamodel-Specific Coupled Evolution Based on Dynamically Typed
Graph Transformations. In Proc. of the 6th Int. Conf. on Theory and Practice of Model
Transformations (ICMT2013), pp 79–91. Springer, 34
2013
https://www.eclipse.org/henshin/examples.php

Transformation models in
 Ongoing work on
o parsing Epsilon programs
o into abstract syntax models
o that conform to Ecore metamodels
o All task-specific language metamodels extend the core
(EOL) language metamodel
 Abstract syntax models used to perform
o static type checking,
o dependency analysis,
o sub-optimal code detection
Ran Wei, Dimitris S. Kolovos: Automated Analysis, Validation and Suboptimal Code Detection in
Model Management Programs. In Proc. of the 2nd BigMDE Workshop, York, UK. 2014
35

Ramification in TCore
 Goal: Customize generic
pattern languages to
domain-specific use
o Generate modeling
environment
quasi-automatically
o Implemented by HOT
o Relax – Augment - Modify
 Usage: embedding MTs
into COTS tools
o MetaEdit+
o Matlab Simulink
36
Kühne, T.; Mezei, G.; Syriani, E.; Vangheluwe,
H. & Wimmer, M. Systematic Transformation
Development. Journal of the ECEASST: 21,
Multi-Paradigm Modeling, Denver (2009).
Input Meta-Model Output Meta-Model
Relax Augment Modify
Customized Pattern Meta-Model

MOFLON
 Generic rewrite rules
o via JMI model repository
o Example: Generic checks
• attribute to be checked
passed as argument
• names of classes to be
checked
 Reflective rewrite rules
o Examine metamodel first
and extract properties
o Instantiate generic rules on
the instance level
 Application:
o Checking modeling
guidelines for Simulink
Carsten Amelunxen, Elodie Legros, Andy Schürr: Generic and reflective graph transformations
for the checking and enforcement of modeling guidelines. 37
VL/HCC 2008: pp. 211-218

Further approaches for HOT
MOTMOT (2006- )
 Add language constructs
to Story Diagrams
 Main Usage Scenario
o Copying + Adaptation
o Business Process Modeling
Medini QVT
 Generate copy rules from
Ecore metamodels
 Refinement
transformations
o Generic copy rules
o Exception rules
38
Pieter Van Gorp, H. Schippers, D. Janssens
Copying Subgraphs within Model Repositories.
5th Int. Workshop on Graph Transformation
and Visual Modeling Techniques, ENTCS, pp.
127-139, Vienna, Austria, 1 April 2006. Elsevier.
Thomas Goldschmidt, Guido Wachsmuth:
Refinement Transformation Support for QVT
Relational Transformations. 3rd Workshop on
Model Driven Software Engineering (MDSE
2008)

Evolution of the VIATRA Tool Family
VIATRA VIATRA2
VIATRA2
R3.2
„VIATRA3”
IncQuery
EMF-IncQuery
IncQuery-D
http://www.eclipse.org/incquery/
40
http://www.eclipse.org/viatra2/

VIATRA
 Main facts & features:
o Written in Prolog
o XMI import/export for
arbitrary metamodels
o UML as rule language
o Auto-generated Prolog code
 Main applications:
o SC2Promela: model checking
UML statecharts
o UML2DF: fault propagation
analysis
o SC2SPN: reliability analysis
 Major projects:
o HIDE + National
 Main publications:
o ASE 2002, SCP 2002
• XMI export/import
• Metamodel
Start:
Sep 2000
• Auto-generated MT
code from UML rules
• First complex MT
M1:
Summer
2001
• UML2DF transformation
• SC2SPN transformation
M2:
Summer
2002
• Development of new
Java/Eclipse version
End:
Early
2004
41

VIATRA2
 Main Features
o Unified storage of
• Models + Metamodels +
Transformations
o Language:
• Graph transformation (GT)
• Abstract state machines (ASM)
o Hybrid pattern matching
• Local search + Incremental
o Live MT
• Event-driven
o VIATRA DSM support
o Design-space exploration
 Main Projects:
o DECOS, SENSORIA, DIANA,
MOGENTES
 Main Publications:
o SCP 2007, ICMT 08-09, MODELS09,
ASE2011, 3xSOSYM, STTT, AUSE
• Initial prototype: Spring 2004
• Eclipse start: @ GMT Sept 1st, 2005
Initial prototype:
Summer 2004
•Model space + GT + ASM
• 1 PhD + 3 MSc students
R1:
Oct 21, 2005
• GUI improvements & documentation
• 2 PhD student + 8 MSc students
R2:
Oct. 10, 2006
• EMF-based MT model
• 3 PhD students + 1 developer + *MSc
R3:
June 2, 2008
• Incremental & live MT, MT workflow
• 5 PhD students + * MSc
M3.1:
March 23, 2010
• EMF support + Visualization
• UML2, BPMN, BPEL importers
M3.2:
April 11, 2011
• Design-space exploration
• Incremental transitive closure
M3.3:
August, 2013
• Event-driven execution platform
• Complex Event Processing
VIATRA3
Video: http://vimeo.com/741248271

Selected Applications of VIATRA2
Toolchain for critical embedded systems
• Model-based tool integration
• Workflow-driven interactive design tools
• Ontology based consistency checks
• Model-driven test generation
Service-oriented applications
• Formal analysis and back-annotation of BPEL processes
• Model based performability analysis
• Model-based service deployment
• Incremental generation of service configurations
GRaTS
• with R. Heckel (Leicester) and P. Torrini
• Stochastic simulator for graph transformation systems
43

EMF-IncQuery
 Main features:
o Define model queries
declaratively (with Xtext editors)
o Execute queries efficiently and
incrementally for EMF models
o Integrate queries into EMF apps
with no coding (validations,
viewers, derived features)
 Contributors
o BME
o IncQuery Labs (Support)
o Itemis
o MODELS10, ICMT11, MODELS12,
ECMFA12, SOSYM14, SCP2014
o Demos at MODELS, ASE, ECMFA,
TOOLS, EclipseCon, …
• Initial version detached from VIATRA
• First performance benchmark
• 1.5 million model elements
1st internal
release: Spring
2010
• Project lead: István Ráth
• Team: Z. Ujhely, Á. Horváth, Á. Hegedüs,
G. Bergmann, T. Szabó, (D. Varró)
Eclipse start:
April 4, 2011
• IncQuery base: Efficient EMF navigation
• Inverse, allInstances, trans. closure
• Incremental
V0.4.0:
Nov 2011
• Xtext based query editor
• Derived features
• Validation framework
V0.6.5:
Oct 2012
• Event-driven VM
• Core viewers
• Xbase
V0.7.0:
July, 2013
• Query libraries + Xcore
• IQ Viewers
• IQ Debugger
V0.8.0:
July, 2014
44
CERTIMOT

Selected Applications (EMF-IncQuery)
• Complex traceability
• Query driven views
• Abstract models by
derived objects
Toolchain for
IMA configs
• Connect to Matlab
Simulink model
• Export: Matlab2EMF
• Change model in EMF
• Re-import:
EMF2Matlab
MATLAB-EMF
Bridge
• Live models
(refreshed 25
frame/s)
• Complex event
processing
Gesture
recognition
• Experiments on open
source Java projects
• Local search vs.
Incremental vs.
Native Java code
Detection of
bad code smells
• Rules for operations
• Complex structural
constraints (as GP)
• Hints and guidance
• Potentially infinite
state space
Design Space
Exploration
• Itemis (developer)
• Embraer
• Thales
• ThyssenKrupp
• CERN
Known Users

IncQuery-D
 Main features:
o Distributed
• Storages
• Indexers
• Incremental Queries
o Deployed over cloud
infrastructure
o Goal: Scalability along
• # of machines
• Memory/CPU
• #of concurrent users
 Contributors
o BME + IncQuery Labs
o Friday@MODELS 2014
o Invited talks at CloudMDE14,
GT-VMT14
• Proof of concept
• Initial architecture
• First performance
benchmark
V0.1
April 2013
• Multiple storage back-ends
• Manual performance
optimizations
V0.2
December
2013
• RDF-IncQuery language
• Monitoring +
Optimized allocation
• First external users
(students)
V0.3
September
2014
46

Acknowledgements
 Current Contributors
o István Ráth, PhD
o Ákos Horváth, PhD
o Gábor Bergmann, PhD
o Ábel Hegedüs, PhD candidate,
(IncQuery Labs)
o Zoltán Ujhelyi, PhD candidate
o Dénes Harmath (IncQuery Labs)
o Tamás Szabó (Itemis)
o Benedek Izsó
o Gábor Szárnyas
o Oszkár Semeráth
o Csaba Debreceni
o István Dávid (Uni-Antwerp)
o András Szabolcs Nagy
o Márton Búr
o József Makai
o Dániel Stein
 Past Contributors
o Prof. András Pataricza
o Gergely Varró, PhD
(TU Darmstadt)
o András Balogh, PhD
(ThyssenKruppPresta)
o András Schmidt (QGears)
o Dávid Vágó (Google)
o Balázs Grill (ThyssenKruppPresta)
o Zoltán Balogh
o András Ökrös
o Balázs Polgár, PhD
o Márk Czotter
o Dániel Tóth (NSN)
o Péter Pásztor
o Gergely Nyilas (Lufthansa Systems)
o Zsolt Sándor (Siemens PSE)
http://www.eclipse.org/viatra2/ 4h7ttp://www.eclipse.org/incquery/

Summary
Metamodel
(Graphs)
Graph
model
Meta
rules
Generic
rules
“Ordinary”
Trans. Eng.
Generic Transformations
Meta/Higher-Order Transformations
XMI
document
“Ordinary”
rules
“Ordinary”
Trans. Eng.
Metamodel
(Graphs)
Graph
model
Generic
Trans. Eng.
XMI
document

Generic and Meta-Transformations for Model Transformation Engineering

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Generic and Meta-Transformations for Model Transformation Engineering

Similar to Generic and Meta-Transformations for Model Transformation Engineering (20)

Recently uploaded

Recently uploaded (20)

Generic and Meta-Transformations for Model Transformation Engineering