models and meanings

Assigning Meanings to Models (Using Model Transformations)

Antonio Vallecillo[Francisco Durán, José E. Rivera, J. Troya]

Atenea Research Group

Málaga, 2011

Assigning meanings to Models

… When we consider MDE with this unified vision, many well-known situations may be integrated into a more general consideration. To take one additional example, the seminal work of R. Floyd ("Assigning meanings to programs", [12]) that founded the research field of programming axiomatics may be viewed as "decorating" a flowchart model with an axioms model. This may lead us first to consider decorations as models, then to understand the nature of the binding between the flowchart model and the axioms model, this binding being itself some kind of correspondence model. Finally, these considerations may lead us to generalize R. Floyd’s approach and to add on the research agenda a new item on "Assigning meaning to models". Model weaving and model transformation will be essential to the study of this subject.

[Jean Bézivin, “On the Unification Power of Models”, 2005]

A. Vallecillo: "Assigning Meaning to Models" 2

Models


Model

“A description or specification of a system and its environment for some certain purpose. A model is often presented as a combination of drawings and text. The text may be in a modeling language or in a natural language.“ [MDA guide (V1 and 2, ab/2003-01-03, 23 January, 2003)]"A model represents some concrete or abstract thing of interest, and does so with a specific purpose. A model is related to the thing it represents by explicit or implicit correspondences between the elements of that model and the parts of that thing. This correspondence enables understanding the intended meaning of that model.” [MDA Guide (V3, ormsc/05-11-03, 30 November 2005)]"A model captures a view of a physical system. It is an abstraction of the physical system, with a certain purpose. This purpose determines what is to be included in the model and what is irrelevant. Thus the model completely describes those aspects of the physical system that are relevant to the purpose of the model, at the appropriate level of detail." [UML Superstructure 2.1.1 (formal/2007-02-05)]"A description of (part of) a system written in a well-defined language." (NOTE: Equivalent to specification.) [Kleppe, 2003]"A representation of a part of the function, structure and/or behavior of a system” [Model Driven Architecture (MDA) ormsc/2001-07-01]"A set of statements about the system.“ (Statement: expression about the system that can be true or false.) [Seidewitz, 2003] "M is a model of S if M can be used to answer questions about S" [D.T. Ross and M. Minsky, 1960]


Model ~ Description

Model

1: obsolete : a set of plans for a building2: dialect British : copy, image3: structural design <a home on the model of an old farmhouse>4: a usually miniature representation of something ; also : a pattern of something to be made5: an example for imitation or emulation6: a person or thing that serves as a pattern for an artist ; especially : one who poses for an artist7: archetype8: an organism whose appearance a mimic imitates9: one who is employed to display clothes or other merchandise10 a: a type or design of clothing b: a type or design of product (as a car)

11: a description or analogy used to help visualize something (as an atom) that cannot be directly observed

12: a system of postulates, data, and inferences presented as a mathematical description of an entity or state of affairs; also: a computer simulation based on such a system <climate models>

13: version14: animal model


http://www.merriam-webster.com/dictionary/model

Some definitions

Model (of a <X>): A representation or specification of a <X> from a given point of view and with a particular purpose

Prototype model: A functional model of a <X>, where the emphasis in on testing – e.g., to verify the design

Domain Specific Model: A model written in a domain specific language

Domain Specific Language: A language which offers concepts and notations closer to the domain experts, at an appropriate level of abstraction, and with a particular purpose

Model Transformation: An algorithmic specification (declarative or operational) of the relationship between models


Meaning

1 a: the thing one intends to convey especially by language: purport 1 b: the thing that is conveyed especially by language: import

2: something meant or intended: aim <a mischievous meaning was apparent>

3: significant quality; especially: implication of a hidden or special significance <a glance full of meaning>

4 a: the logical connotation of a word or phrase 4 b: the logical denotation or extension of a word or phrase


http://www.merriam-webster.com/dictionary/meaning

Semantics

1: the study of meanings [Merrian-Webster]a: the historical and psychological study and the classification of changes in the signification of words or forms viewed as factors in linguistic development b (1): semiotic (2): a branch of semiotic dealing with the relations between signs and what they refer to and including theories of denotation, extension, naming, and truth

Formal semanticists are concerned with the modeling of meaning in terms of the semantics of logic. In computer science, where it is considered as an application of mathematical logic, semantics reflects the meaning of programs or functions [wikipedia]


The Meaning of Models

“ There are only 10 types of people in

the world: Those who understand binary, and those who don't

10A. Vallecillo: "Assigning Meaning to Models"

Sure?

“There are only 10 types of people in

the world: Those who understand binary, and those who don't”


Same model for different concepts

“ There are only 10 types of people in

the world: Those who understand binary, and those who don't


210

Your meaning

goes here…

Different models for the same concept

10 X 1010 0A


10101010

What does this model mean?


What does this model means?


[[M]]BPM

N

[[M]]UML

A

C

B

Exclusivesemantics

Inclusivesemantics


SM

[[M]]

{{S}}

purportimport

Why do I need to assign meanings to models?

What do I need models for?Describe the system

Structure, behaviour, ...Separate concepts at different conceptual levelsCommunicate with stakeholders

Understand the systemIf existing (legacy applications)

Validate the modelDetect errors and omissions in design ASAP

Mistakes are cheaper at this stage

Prototype the system (execution of the model)Formal analysis of system properties

Drive implementationCode skeleton and templates, complete programs (?)


[Bran Selic, 2003]

Impossible without

precise meanings!

How do we assign meaning?

How do we express the meaning ofStructure?Behavior?Time-dependent functionality?QoS properties?…

Which is the best notation for each of those aspects?It depends on the purpose of the model……and must have a precise meaning


Domain Specific Modeling Languages


Domain Specific Modeling Languages (DSML)

Languages for representing different views of a system in terms of models

Higher-level abstraction than general purpose languages

Closer to the problem domain than to the implementation domain

Closer to the domain experts, allowing modelers to perceive themselves as working directly with domain concepts

Domain rules can be included into the language as constraints, disallowing the specification of illegal or incorrect models.


An example of a DSM

The design for a conference application intended to run on a Symbian/S60 phone.

[Borrowed fromJuha-Pekka Tolvanen,

“Domain-specific Modeling: Making Code Generation Complete”

http://www.devx.com.]

http://www.devx.com/

Visual DSMLs

VDSMLs tend to offer substantial gains over conventional textual languages

Formal studies show significant benefits for novicesIncreasing number of VDSMLs being defined

But not a panacea: every notation has advantages and disadvantages

[Not the subject of this talk]


[Thomas R.G. Green & Marian Petre]

[T. Green, M. Petre “Usability Analysis of Visual Programming Environments: a ‘cognitive dimensions’ framework” JVLC, 1996]

[Kirsten whitley “Visual programming languages and the empirical evidence for and against”, JVLC 1997]

[R. Navarro-Prieto, J. Cañas “Are visual programming languages better? The role of imagery in program comprehension”, IJHCS 2001]

“The impossible equation”

USA (estimates for 2012):

• 90M computer users

• 55M Spreadsheet & DB users

• 13M self-described as programmers

• 2.5M professional programmers[Source:US Bureau of Labour Statistics, 2006]


[J. Bezivin, Keynote at JISBD 2009]

End-user Programming[Modeling]

Most software creators are not software professionals

End users are participants and developers, not passive consumersThey do not reason about software like professionals[Mary Shaw, The Challenge of Pervasive Software to the Conventional Wisdom of Software Engineering, ESEC-FSE09]

End users are not “casual,” “novice” or “naive” users; they are people such as chemists, librarians, teachers, architects, and accountants, who have computational needs and want to make serious use of computers, but who are not interested in becoming professional programmers.

[Bonnie A. Nardi, A Small Matter of Programming. MIT Press, 1993]


End-user (Visual) Modeling


Production systems


DSMLs are starting to proliferate

They allow users to model their systems at an appropriate level of abstraction

Some of them allow more than “documentation”Code generationAnimationSimulation…

Very few allow specification and analysis of the Quality Properties (NFPs) of modeled systems

QoS usage and management constraints: performance, reliability, resource consumption and allocation, etc.


Requirements for PA of DSM

Notations for describing systems must be: Simple and intuitive Close to the problem domainClose to the domain experts’ language

Models must be: Abstract, yet preciseExecutable (to, at least, prototype systems)

QoS notations must be:Simple and precise, yet expressive

Analysis results and feedback must be:Understandable and easy to manage


“Being abstract is something profoundly different from being vague... The purpose of abstraction is not to be vague, but to create a new semantic level in which one can be absolutely precise.”

Edsger Dijkstra


Notations for describing systems must be normally are: Simple and intuitive Complex Close to the problem domainClose to the solution domainClose to the domain experts’ language General purpose

Models must be normally are: Abstract, yet precise Too detailed, impreciseExecutable Non-executable

QoS notations must be normally are: Simple and precise, yet expressive Too complex and low level (more than needed for most end-user DSMLs)

Analysis results and feedback must be normally are: Understandable and easy to manage Tough to deal with!

Current notations for DSMs


“Close to the problem domain”

Once upon a time, there was a team leader that was going on holidays. Before leaving, she made the last recommendation to her small team of three young engineers: “For the ongoing project, do not start coding in Java before the UML model is completely finished and you all agree on the model.”

On the Monday morning, as soon as she left, one of the engineers told the others about a wonderful discovery he made while twittering in the weekend: a very powerful tool that generates UML diagrams from code. The decision was rapidly taken and all three started coding the problem in Java.

Some days before the end of the leader’s holidays, all the Java code was used to generate UML diagrams and both the code and the UML diagrams were handled to the group leader.

She was quite impressed about the level of detail of the UML model and the narrow correspondence between the code and the model.


[Borrowed from J. Bezivin]


The resulting model

Each notation is more apt for a task


MCMLXVII+ DLXXIX ???

1.967+ 579



MCMLXVII+ DLXXIX ???

1.967+ 579 2.546



MCMLXVII+ DLXXIX MMDXLVI

1.967+ 579 2.546

How do you solve this problem?

A 40-years-old man has a daughter and a son. If the difference of age between the kids is 4 years, and the sum of their ages is half of the age of the father, how old are they?


x – y = 4x + y = 20

x = 12y = 8

Solution: the older is 12 and the younger is 8

2x = 24

+

Problems, Notations, Solutions

An invariant through the history of mature disciplines is the search for notations that allow formulating problems in a language that allows their easy solution

http://en.wikipedia.org/wiki/History_of_mathematical_notation http://en.wikipedia.org/wiki/Temporal_logic


Our current software modeling notations


The UML way…

Sauron’s approach to metamodeling(e.g., OMG’s UML metamodel)

The lord of the Metamodels (obviously, adapted)

Three notations for the Structure modelers under the sky,Seven for the Behavior modelers in their halls of stone,

Tree for mortal Packagers doomed to die,One for the Designer of the Whole system on his dark

throneIn the Land of Mof where the shadows lie.

One Metamodel to rule them all, One Metamodel to find them,

One Metamodel to bring them all and in the darkness bind them

In the Land of Mof where the shadows lie.


No general purpose language can express all different semantics without becoming a monster

Especially under the presence of antagonist semantics (Discrete & continuous; synchronous & asynchronous;...)


“More general does not mean Better. Heterogeneity may be better than generality. ...Useful semantics imply constraints on designers.” Edward A. Lee

How can we specify DSMs?

How do we express in a precise and abstract manner:

StructureBehaviorTime-dependent functionalityQuality properties (QoS,…)

Which is the best notation for each of those aspects?It depends on the purpose of the model……must have a precise meaning…and must allow the analysis of the models


Visual DSMLs


Anatomy of a DSML (I)


We need more than syntax…

Describe meaningful modelsAt the appropriate level of abstractionIn a correct, complete and accurate mannerUsing a notation natural to the target domain engineer

Animate models Explicitly define behavioral semantics of DSLs so that models can be understood, manipulated and maintained by both users and machinesAdd Non-Functional Properties (Time, QoS,…) to DSLsMake models amenable to simulation

Analyse modelsConnect DSLs to Analysis tools


Anatomy of a DSML (II)


The village metaphor


Semantic (or “Meaningful”) Domains

The meaning of a model M is defined by its interpretation in a meaningful semantic domain D.

Each Semantic Domain has

Precise semanticsA set of (equivalent) notationsA set of analysis toolsUnderlying logic

Semantic Bridges connect Semantic Domains


The Prolog villageThe QNM villageThe Petri Net villageThe Coq villageThe Process Alg villageThe Maude villageThe Z villageThe Modellica villageThe B village…

(vs. “meaningless” Models)

Expressing semantic bridges

As Model Transformations

Possible if correspondences can be expressed as functionsPairwise consistency can be formally studied

One form of consistency involves a set of correspondence rules to steer a transformation from one language to another. Thus given a specification S1 in viewpoint language L1 and specification S2 in viewpoint language L2, a transformation T can be applied to S1 resulting in a new specification T(S1) in viewpoint language L2 which can be compared directly to S2 to check, for example, for behavioral compatibility between allegedly equivalent objects or configurations of objects [RM-ODP, Part 3]

As Weaving ModelsPossible if correspondences are just mappings


Semantic Mappings as Model Transformations

TypesDomesticHorizontalVerticalAbstractingRefiningPruningForgetful…


The relationship between domains D and D’ is defined by a model transformation T:D->D’.

[[M]]D’ := [[T(M)]]D’

How do we analyse models?

Crossing the bridges!!!


Models to connect (analysis) tools!

Material Model ABB:

Continuum ABBs:

E

One D LinearElastic Model

T

G

e

t

material model

polar moment of inertia, J

radius, r

undeformed length, Lo

twist,

theta start, 1

theta end, 2

r1

12

r3

0L

r

J

rTr

torque, Tr

x

TT

G, r, , ,J

Lo

y

material model

temperature, T

reference temperature, To

force, F

area, A

undeformed length, Lo

total elongation,L

length, L

start, x1

end, x2

E

One D LinearElastic Model

(no shear)

T

e

t

r1

12 xxL

r2

oLLL

r4

A

F

edb.r1

oTTT

r3

L

L

x

FF

E, A,

LLo

T, ,

yL

Torsional Rod

Extensional Rod

temperature change,T

cte,

youngs modulus, E

stress,

shear modulus, G

poissons ratio,

shear stress, shear strain,

thermal strain, t

elastic strain, e

strain,

r2

r1)1(2

EG

r3

r4Tt

Ee

r5

G

te

1D Linear Elastic Model

material

effective length, Leff

linear elastic model

Lo

Extensional Rod(isothermal)

F

L

A

L

E

x2

x1

youngs modulus, E

cross section area, A

al1

al3

al2

linkage

mode: shaft tension

condition reaction

allowable stress

stress mos model

Margin of Safety(> case)

allowable

actual

MS

Analysis Templatesof Diverse Behavior & Fidelity

(CBAMs)MCAD Tools

Materials LibrariesIn-House, ...

FEAAnsys

Abaqus*

CATIA Elfini*

MSC Nastran*

MSC Patran*

NX Nastran*

...

General MathMathematica

Matlab*

MathCAD*

...

Analyzable Product Model(APM)

Extension

Torsion

1D

1D

Analysis Building Blocks(ABBs)

CATIA, NX,Pro/E*, ...

Analysis Solvers(via SMMs)

Design Tools

2D

flap_link

critical_section

critical_simple

t2f

wf

tw

hw

t1f

area

effective_length

critical_detailed

stress_strain_model linear_elastic

E

cte area

wf

tw

hw

tf

sleeve_1

b

h

t

b

h

t

sleeve_2

shaft

rib_1

material

rib_2

w

t

r

x

name

t2f

wf

tw

t1f

cross_section

w

t

r

x

R3

R2

R1

R8

R9

R10

6R

R7

R12

11R

1R

2

3

4

5

R

R

R

R

name

linear_elastic_model

wf

tw

tf

inter_axis_length

sleeve_2

shaft

material

linkage

sleeve_1

w

t

r

E

cross_section:basic

w

t

rL

ws1

ts1

rs2

ws2

ts2

rs2

wf

tw

tf

E

deformation model

x,max

ParameterizedFEA Model

stress mos model


allowable

actual

MS

ux mos model


allowable

actual

MS

mode: tensionux,max

Fcondition reaction

allowable inter axis length change

allowable stress

ts1

B

sleeve1

B ts2

ds2

ds1

sleeve2

L

shaft

Leff

s

rib1 rib2

material

effective length, Leff

deformation model

linear elastic model

Lo

Torsional Rod

G

J

r

2

1

shear modulus, G

cross section:effective ring polar moment of inertia, J

al1

al3

al2a

linkage

mode: shaft torsion

condition reactionT

outer radius, ro al2b

stress mos model

allowable stress

twist mos model


allowable

actual

MS


allowable

actual

MS

allowabletwist

Linkage Extensional Model

Linkage Plane Stress Model

Linkage Torsional Model* = Item not yet available in toolkit—all others have working examples 2007-04

Parts LibrariesIn-House*, ...

LegendTool AssociativityObject Re-use

[Borrowed from Russell Peak presentation at OMG, 2007]


Multiple views and coordinated DSLs

Each view is expressed in a given domain languageVocabularies of different corporations are differentHowever they allow talking about a common building


Assigning Meanings to Models

(Using model transformations)

Our proposal

Describe the structure of Models/Metamodels with usual modeling notation and tools (Ecore, EMF, GMF, …)Describe the behavior of models using visual languages (including Time and QoS aspects) Specify the structure of models and metamodels using a formal system, i.e., a precise Semantic Domain (e.g., Maude)Specify the behavior of models using a formal system, i.e., in a Formal Semantic Domain (e.g., Maude)Define mappings between the visual and formal notations (the latter provides the meaning for the former)Make use of the analysis tools in the target domain to reason about the visual models


A Production System Example


Specifying dynamic behavior

Use of In-place Transformation Rules (e.g., graph transformations)Abstract and concrete syntaxes: Independent from the underlying semantic framework (e.g., Maude)

l:[NAC] x LHS → RHS


Adding time to behavioral specifications

Part of the e-Motions modeling notationRule durationPeriodicity, soft schedulingOngoing rulesAccess to the Global Time Elapse

Time stamps, scheduled actions


Carry [5]

c t

p

c t

p xPos = t.xPosyPos = t.yPos

l:[NAC] x LHS → RHSt

Adding action executions

Specification of action properties Withouth the need of unnaturally modifying the metamodel


Precise Semantics

Defined by a Semantic Mapping to Real-Time Maude

This makes models amenable to simulation and to formal analysis using the Real-Time toolkit!

Implementable (by a set of ATL model transformations)


(65)

ProductionSystem {< 'p : Plant | els : 'heg 'hag 'c1 'c2 't1 'a 'c3 't2 'u >< 'hag : HandleGen | in : null, out : 'c2, xPos : 1, yPos : 1 >< 'heg : HeadGen | in : null, out : 'c1, xPos : 1, yPos : 3 >< 'c1 : Conveyor | outParts : nil, out : 't1, xPos : 2, yPos : 3 >< 'c2 : Conveyor | outParts : nil, out : 't1, xPos : 2, yPos : 1 >< 't1 : Tray | parts : nil, capacity : 4, xPos : 3, yPos : 2 >< 'a : Assembler | in : 't1, out : 'c3, xPos : 4, yPos : 2 >< 'c3 : Conveyor | outParts : nil, out : 't2, xPos : 5, yPos : 2 >< 't2 : Tray | parts : nil, capacity : 4, xPos : 6, yPos : 2 >< 'u : User | parts : nil, xPos : 6, yPos : 3 >}

Representing Models with Maude

A. Vallecillo: "Assigning Meaning to Models"

(66)

op ProductionSystem : -> @Metamodel .

op PS : -> @Package .

sort PositionedEl .subsort PositionedEl < @Class .op PositionedEl : -> PositionedEl . op xPos : -> @Attribute . op yPos : -> @Attribue .

sort Container .subsort Container < PositionedEl .op Container : -> Container . op parts : -> @Reference .

sort Machine .subsort Machine < PositionedEl .op Machine : -> Machine . op in : -> @Reference . op out : -> @Reference .

…

Representing Metamodels with Maude

eq isAbstract(Machine) = true ....eq type(in) = Tray .eq lowerBound(in) = 0 .eq upperBound(in) = 1 ....eq type(out) = Conveyor .eq opposite(out) = null .eq lowerBound(out) = 1 .eq upperBound(out) = 1 .


(67)

Representing Behavior with Maude

rl [Transfer] : ProductionSystem { < p : P:Part | xPos : XPOS, yPos : YPOS, SFS > < c : Conveyor | OutParts : (p PARTS), out : t, SFS’ > < t : Tray | xPos : XPOS’, yPos : YPOS’, parts : PARTS’, SFS’’ > OBJSET }=> ProductionSystem{ < p : P:Part | xPos : XPOS’, yPos : YPOS’,SFS > < c : Conveyor | outParts : PARTS, out : t, SFS’ > < t : Tray | xPos : XPOS’, yPos : YPOS’, parts : (p PARTS’), SFS’’ > OBJSET }.


e-Motions


http://atenea.lcc.uma.es/e-Motions

(69)

Model Simulation and Analysis

Simulation/Execution of specifications

Reachability AnalysisDeadlock

Invariants

Others

LTL Model checkingLiveness properties

search [10] initModel =>! ProductionSystem { < ’t2 : Tray | parts : empty, SFS > OBJSET } .

(trew initModel in time <= 20 .)

reduce modelCheck(initModel, [](exist(’he10) -> ~exist(’he10)) .

search initModel =>* ProductionSystem { < O : Tray | capacity : CAP, parts : PARTS, SFS > OBJSET }such that | PARTS | > CAP .

(mc {initModel} |=t [](ensembled(’he10.ha10) -> collected(’he10.ha10)) in time <= 100 .)

(find earliest {initModel} =>* {ProductionSystem {< T : ActionExec | rule : "Collect", value : null, SFS@T > OBJSET }} .)


[Simulation 2009]

Adding NFPs to DSMLs


Use of “observers”




Observers capture the state of the NFPs and monitor their progress


Making use of the Observers

The system can self-adapt under certain conditions


Making use of the Observers

With the new configuration, the system transmits sounds in a faster way


A more complex example


[TAP 2011, http://atenea.lcc.uma.es/e-Motions]

Analysing the results


http://atenea.lcc.uma.es/index.php/Main_Page/Resources/E-motions/PacketSwitchingExample/Results

Pros and Cons

AdvantagesAddition of observers independently of the systemSimple modelling of QoS propertiesAbility to monitor QoS propertiesResults obtained in easy-to-manipulate formatMore expressive than other notations (SPA, SPN, QNM,…)(generalized distributions, OCL expressiveness, dynamic topologies, action executions as 1st class citizens,…)

LimitationsEfficiency of simulationsDifficult to debugNot for every problem or domainMore expressive than other notations (SPA, SPN, QNM,…)(Difficulties for defining semantic bridges due to large gaps/chasms)


We are not alone…


[Pamela Zave, keynote talk at MODELS 2010]

Recap

One challenge for software engineers now is to provide end-users with Modeling Languages (and associated tools) that allow them to model their systems in a cheap, quick and useful way, and to analyse them using push-button approaches.

Current widely-used general-purpose modeling notations (especially behavioural and QoS) do not seem to be really up to the job

Integrating heterogeneous notations and their associated tools using model transformations seems to be one promising way to go


Challenges

Definition of new languages for behaviouraldescriptions, which allow easy specificationof Quality Properties and their analysis

Improved languages for QoS specification

Semantic bridges to other domainsBetter connection with analysis tools

Improved traceability mechanisms Improve understandability of results

Better feedback to usersE.g., Performance anti-patterns (!)


models and meanings

Documents

flowchart model

axioms model

animal model

functional model

model weaving

model of s

assigning meaning

kind of correspondence