a survey on incremental model transformation...

1 Funding: BRIDGE 832160

www.model-evolution.net

Dept. of Cooperative Information Systems Johannes Kepler University Linz

http://www.cis.jku.at

Business Informatics Group Vienna University of Technology

http://www.big.tuwien.ac.at

LieberLieber Software GmbH http://www.lieberlieber.com

A Survey on Incremental Model Transformation Approaches

Models and Evolution Workshop, Miami, Florida, Sept. 30th 2013

A. Kusel, Jürgen Etzlstorfer, E. Kapsammer, P. Langer, W. Retschitzegger, J. Schönböck,

W. Schwinger, M. Wimmer

Motivation n Language Coverage n Execution Phases n Overhead n Lessons Learned

2 www.model-evolution.net

•  Model transformations allow to generate models out of other models

•  Evolution entails re-execution of model transformations •  Batch transformations generate complete target models

each time the transformation is (re-)executed •  In contrast, incremental model transformations propagate

changes in the source model to the target model, only –  Rather depend on size of changes than on size of the source model –  Especially useful for frequent changes in large source models

Need for Incremental Model Transformations

Source Metamodel Target Metamodel Transformation Definition

Source Model Target Model Transformation Engine

refers to

reads

refers to

writes conforms to conforms to executes

Batch Transformation Incremental Transformation

Recent approaches for incremental model transformations

differ substantially



Execution Engine

Motivating Example

NamedElement name : String

Classifier

DataType

Class persistent:boolean

Attribute multivalued:boolean

Named name : String

Table

Column

Type type

col

type

attr

c1: Class name="Person" persistent=true

a1 : Attribute name="Name"

multivalued=false

a2 : Attribute name="Address" multivalued=false

d1 : DataType name="String"

t1: Table name="Person"

c1 : Column name="Name"

c2 : Column name="Address"

d1 : Type name="String"

Source Model

Target Model

Target Metamodel

Source Metamodel

conforms to conforms to

abstract rule NamedEl2Named { from n : ClassD!NamedElement to t : Relational!Named ( name <- n.name )

} rule Class2Table extends NamedEl2Named {

from c : ClassD!Class (c.persistent) to t: Relational!Table ( col <- c.attr->select(e | not e.multiValued) ->subSequence(1,2) )

} rule Attribute2Column extends NamedEl2Named {

from a : ClassD!Attribute (not a.multivalued) to t: Relational!Column ( type <- a.type )

}

TraceLink

TraceLink

TraceLink

TraceLink

"Fullname"

"Fullname"

How is incrementality achieved?

executes

Model2Model Trace

Change in the source model

Change is propagated to target model

Transformation Specification

(ATL Code) Which language parts are supported for incremental

transformation?

How much overhead is imposed by incrementality?



Evaluation Framework Overview

Evaluation Framework

Language Coverage

Execution Phases Overhead

•  Declarative Language Parts •  Imperative Language Parts

•  Change Detection •  Impact Analysis •  Change Propagation

•  Specification •  Run-time •  Memory

8 approaches have been examined



Approach Giese et al., 2008

Lauder et al., 2012

Hearnden et al., 2006

Jouault and Tisi, 2010

Rath et al., 2008

Bergmann et al., 2012a

Bergmann et al., 2012b

Razavi et al., 2009

Language TGG TGG Te*at ATL Viatra2 Viatra2 Viatra2 Arbitrary

How is incrementality achieved?

Which language parts are supported for incremental

transformation?

How much overhead is imposed by incrementality?



rule Attribute2Column extends NamedEl2Named { from a : ClassD!Attribute (not a.multiValued) to t: Relational!Column ( type <- a.type )

}

Execution Engine

Language Coverage

c1: Class name=Person persistent=true

a1 : Attribute name=Fullname

multivalued=false

a2 : Attribute name=Address

multivalued=false a3 : Attribute name=Age

multivalued=false

t1: Table name=Person

c1 : Column name=Fullname

c2 : Column name=Address

c3 : Column name=Age

true

TraceLink

TraceLink

TraceLink

TraceLink

helper def : id : Integer = 0; abstract rule NamedEl2Named {

from n : ClassD!NamedElement to t : Relational!Named ( name <- n.name )


from c : ClassD!Class (c.persistent) to t: Relational!Table ( col <- c.attr->select(e | not e.multiValued) ->subSequence(1,2) ) do { thisModule.id <- thisModule.id + 1; }

} ATL Code

Source Model Target Model Trace Model

DeclaraFve Language Parts •  Number of Input Elements •  Number of Output Elements •  Assignments •  Rule Inheritance •  CondiBons ImperaFve Language Parts

Giese, 2008

Lauder, 2012

Hearnden, 2006

Jouault, 2010

Rath, 2008

Bergmann, 2012a

Bergmann, 2012b

Razavi, 2009

TGG TGG Te*at ATL Viatra2 Viatra2 Viatra2 Arbitrary

1..* 1..* 1..* 1 1..* 1..* 1..* 1

1..* 1..* 1..* 1..* 1..* 1..* 1..* 1 ü ü ü ü ü ü ü ü -‐ -‐ -‐ -‐ n.a. n.a. n.a. ü ü ü ü ü ü ü ü -‐

n.a. n.a. n.a. -‐ ü ü ü ~

Insufficient Support for Imperative

Language Parts





Language Coverage







Change Detection

Change DetecFon •  Source Minimality •  Type & Number of Changes

•  Insert •  Delete •  Update •  Move

•  Granularity •  Object •  Value •  Link

•  Change Log OpBmizaBon


a1 : Attribute name="Fullname" multivalued=false


a3 : Attribute name="Age"

multivalued=false

Giese, 2008

Lauder, 2012

Hearnden, 2006

Jouault, 2010

Rath, 2008

Bergmann, 2012a

Bergmann, 2012b

Razavi, 2009

TGG TGG Te*at ATL Viatra2 Viatra2 Viatra2 Arbitrary ü ü ü ü ü ü ü ü

1..* 1..* 1 1 1..* 1..* 1 1 1..* 1..* 1 1 1..* 1..* 1 1 1..* -‐ -‐ 1 1..* 1..* 1 1 -‐ -‐ -‐ -‐ -‐ 1..* -‐ -‐

ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü ü -‐ n.a. n.a. ü -‐ n.a. n.a.

Execution Engine


c1 : Column name="Fullname"


d1 : Type name=String

Target Model

(i) Detect single changes (ii) Detect mul8ple changes

(i) Detect on course-‐grained level (ii) Detect on fine-‐grained level

Source Model

(i) Compare to previous version of model to detect changes (ii) Listen for changes

Focus on Basic Change Opera8ons



abstract rule NamedEl2Named { from n : ClassD!NamedElement to t : Relational!Named ( name <- n.name )


from c : ClassD!Class (c.persistent) to t: Relational!Table ( col <- c.attr->select(e | not e.multiValued) ->subSequence(1,2) )

} rule Attribute2Column extends NamedEl2Named {

from a : ClassD!Attribute (not a.multivalued) to t: Relational!Column ( type <- a.type )

}

Execution Engine

Impact Analysis









Source Model Target Model

TraceLink TraceLink TraceLink TraceLink

Impact Analysis •  Required Knowledge of TransformaBon SpecificaBon •  Black Box •  White Box

•  Trace •  Model2Model •  Model2TransformaBon

•  Compile-‐Time •  Run-‐Time

•  Auxiliary InformaBon

TraceLink TraceLink TraceLink

-‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐

Auxiliary Information

Model2Model Trace

Model2Transformation Trace

(i) Black Box (ii) White Box

ATL Code

e.g. o) Compiler-‐generated rules o) Database tables

compiler generated rule: Relation!Column. set_Attribute2Column_t_name(a1)

Giese, 2008

Lauder, 2012

Hearnden, 2006

Jouault, 2010

Rath, 2008

Bergmann, 2012a

Bergmann, 2012b

Razavi, 2009


-‐ -‐ -‐ -‐ -‐ -‐ -‐ ü ü ü ü ü ü ü ü -‐

ü ü ü ü ü ü ü ü

-‐ -‐ -‐ ü -‐ -‐ ü -‐ ü ü ü ü ü ü ü -‐ -‐ -‐ SLDNF

tree Compiler Generated Rules

-‐ Change History Models

Database Tables

Concept Pool

Trace Informa8on is Mandatory







Change Propagation

Change PropagaFon •  Time

•  Eager (Synchron) •  Lazy (Asynchron)

•  Coverage •  ParBal •  Complete

•  Granularity •  Coarse (Rule) •  Fine (Binding)

•  DirecBon •  UnidirecBonal •  BidirecBonal

•  Strategy

Execution Engine






abstract rule NamedEl2Named { // ...


// ... } rule Attribute2Column

extends NamedEl2Named { from a : ClassD!Attribute (not a.multivalued) to t: Relational!Column ( name <- a.name )

}

(i) Propagate changes when they occur (ii) Propagate changes when affected model elements are accessed

(i) Propagate parts of the changes (ii) Propagate all changes

(i) Propagate from source to target (ii) Propagate in both direc8ons (i) Support different strategies

(ii) Automa8cally select strategies

ATL Code

Giese, 2008

Lauder, 2012

Hearnden, 2006

Jouault, 2010

Rath, 2008

Bergmann, 2012a

Bergmann, 2012b

Razavi, 2009


ü ü ü ü ü -‐ ü ü -‐ -‐ -‐ -‐ -‐ ü -‐ -‐

-‐ -‐ -‐ -‐ -‐ -‐ ü ü ü ü ü ü ü ü ü

ü ü -‐ -‐ ü ü ü n.a. -‐ -‐ ü ü -‐ -‐ -‐ n.a.

-‐ -‐ ü ü ü ü ü -‐ ü ü -‐ -‐ -‐ -‐ -‐ ü 1 1 1 1 2 (manual) 1 1 1

Lack of Appropriate Propaga8on Strategies

(i) Re-‐execute complete rule (ii) Re-‐execute bindings only





Language Coverage







cdrule propagateUpdate (A) = { guard change pattern (); postcondition pattern newAttr () = { // ... }

} Viatra2 Code t1: Table

name="Person"








Model2 Transformation

Trace

-‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐

Auxiliary Information

Overhead

Overhead •  No SpecificaBon Overhead •  No Run-‐Bme Overhead •  No Memory Overhead

rule Class2Table { // ...

} rule Attribute2Column {

// ... } rule DataType2Type {

// ... } @Trigger(priority=’10’, mode=’always’, sensitivity=’rise ’) gtrule Attribute2Column() = { // ... }

ATL Code

Viatra2 Code

Execution Engine


Model2 Model Trace

(i) Genera8on of trace informa8on (ii) Maintaining trace informa8on

Giese, 2008

Lauder, 2012

Hearnden, 2006

Jouault, 2010

Rath, 2008

Bergmann, 2012a

Bergmann, 2012b

Razavi, 2009

TGG TGG Te*at ATL Viatra2 Viatra2 Viatra2 Arbitrary ü ü ü ü ü -‐ -‐ ü ü ü unknown unknown unknown ü unknown unknown -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐

(i) Model2Model Trace (ii) Model2Transforma8on Trace (iii) Auxiliary Informa8on



(i) No Specifica8on Overhead (ii) Annota8ons (iii) New Specifica8on



Trade-‐off between space and 8me



Lessons Learned


Language Coverage





Insufficient Support for Impera8ve Parts

Focus on Basic Change Opera8ons

Trace Informa8on is Mandatory

Lack of Appropriate Propaga8on Strategies

Lack of Proof of Correctness

13 Funding: BRIDGE 832160

www.model-evolution.net

Dept. of Cooperative Information Systems Johannes Kepler University Linz

http://www.cis.jku.at

Business Informatics Group Vienna University of Technology

http://www.big.tuwien.ac.at

LieberLieber Software GmbH http://www.lieberlieber.com

Thank you!

Jürgen Etzlstorfer [email protected]


•  [Giese2008] Giese, H., Hildebrandt, S.: Incremental Model Synchronization for Multiple Updates. In: 3rd Int. Workshop on Graph and Model Transformations. ACM (2008)

•  [Lauder2012] Lauder, M., Anjorin, A., Varró, G., Schürr, A.: Efficient Model Synchronization with Precedence Triple Graph Grammars. In: 6th Int. Conf. on Graph Transformations. Springer (2012)

•  [Hearnden2006] Hearnden, D., Lawley, M., Raymond, K.: Incremental Model Transformation for the Evolution of Model-Driven Systems. In: 9th Int. Conf. on Model Driven Engineering Languages and Systems. Springer (2006)

•  [Jouault2010] Jouault, F., Tisi, M.: Towards Incremental Execution of ATL Transformations. In: 3rd Int. Conf. on Theory and Practice of Model Transformations. Springer (2010)

•  [Rath2008] Rath, I., Bergmann, G., Okrös, A., Varr´ o, D.: Live Model Transformations Driven by Incremental Pattern Matching. In: 1st Int. Conf. on Theory and Practice of Model Transformations. Springer (2008)

•  [Bergmann2012a] Bergmann, G., Ráth, I., Varró, G., Varró, D.: Change-driven model transformations. SoSym 11(3) (Jul 2012)

•  [Bergmann2012b] Bergmann, G., Horváth, D., Horv´ ath, A.: Applying Incremental Graph Transformation to Existing Models in Relational Databases. In: 6th Int. Conf. on Graph Transformations. Springer (2012)

•  [Razavi2009] Razavi, A., Kontogiannis, K., Brealey, C., Nigul, L.: Incremental Model Synchronization in Model Driven Development Environments. In: Conf. of the Center f. Adv. Studies on Collab. Research. IBM (2009)

References

a survey on incremental model transformation...

Documents