Version control for graphbased models

1
Version control for graph-based models

• Z. Protic
• M. F. van Amstel
• M.G.J. van den Brand

2
Outline

• Goals
• Software versioning
• Model versioning
• Graphs as basis for models
• Versioning graphs
• Versioning of graph-based models
• Tool architecture

3
Goals

• Create a system that is able to put multiple
  types of models under version control
• The models in place are amongst others labeled
  transition systems, UML diagrams, Petri nets,
• This system should be able to work with tooling
  available at the market

4
Software versioning

• The process of assigning unique identifiers to
  different states of software
• Allows for temporal and logical relation of
  states of computer software (version 1.0 precedes
  version 2.0 and is less advanced)
• Software versioning and revision control form the
  basis for software configuration management
  (SCM).
• Quite suitable for text-based files (CVS, SVN).
• Used in most large software projects

5
Software versioningRevision of SCM features

• SCM basic features
• Adding documents to a repository
• Retrieving documents from a repository
• Committing documents into a repository
• Resolving conflicts that occur during the process
  of committing

6
Model versioning

• Tries to solve the inability of existing software
  versioning systems to deal with models of
  software artifacts

Text file
Java program
UML XMI file
Document
Package
Model
V
V
Paragraph
Class
Package
V
C
Phrase
Method
Class
Word
Statement
Operation
C
V Unit of versioning C Unit of comparison
Odysey-VCS a Flexible Version Control System
for UML Model Elements, H. Oliveiar et. al., SCM
2005
7
Model versioning

• Challenges
• Represent models (in enough detail)
• Store models
• Calculate differences (deltas) between models
• Describe and resolve conflicts while
  synchronizing models

8
Model versioningRepresentation of differences

• State based
• Models are represented as sets of entities and
  relations.
• Differences are expressed in terms of states of
  entities and relations in initial and final model
  (present or absent).
• Operation based
• Models are represented as operations required to
  produce a model.
• Difference are expressed in terms of operations
  required to come from initial to final model

9
Versioning graphs

• Graphs consist of
• Labeled nodes
• Labeled edges

Example graph
10
Versioning graphs

• Graphs can be transformed to structured text.
• Text can be versioned by SVN (or ).
• Differences expressed in terms of added/removed
  nodes and edges (state based).

O1 O2 O3 O4 O1, O2, E1 O1, O4, E2
Bi-directional transformation
Example graph
Textual representation
11
Graphs as basis for models

• Claim Most models describing software artifacts
  can be transformed into graphs.
• However the transformation is different for every
  model type (for every meta-model).

12
Graphs as basis for models

• Example simple UML class diagram

ltclassgt
ltattributegt
lttypegt
A
ID1
ID2
ID3
B Int
A
B
Int
13
Graphs as basis for modelsGraphs for more
complex model types
14
Graphs as basis for models

• Generic approach
• Models are represented as graphs.
• Differences are calculated on graphs, and hence
  are state based.
• Unit of versioning is the graph representation of
  the model.
• Unit of comparison is defined by the user.

15
Graph-based model versioning
16
Graph-based model versioning tool architecture
17
Use-case Application of a model version tool for
versioning of UML class diagrams
18
Use-case Application of a model version tool for
versioning of UML class diagrams
19
Use-case Application of a model version tool for
versioning of UML class diagrams
20
Use-case Application of a model version tool for
versioning of UML class diagrams