UML View Integration

Transcription

1 UML View Integration lexander Egyed nnual Research Review February 89, 1999 Iclsl~I Outline Problem Solving in UML Examples of Mismatches Integration ctivities Goals and Limitations

2 Iclsll rchitecting and UML n a l y s i s rchitecting and HighLevel Design LowLevel Design rchitecting and ;... HighLevel Design1 b l ' :.... LowLevel Desien I I ';,,*%. lclsll rchitecting and UML Software Development seems to have a diagram (view) centric problem solving approach. lthough, these views are very useful on their own, there is only little which keeps them together. This is a problem because: => they are standalonelindependent => they hardly share modeling elements => they are for different audienceslstakeholders (different interpretations) => they are often used concurrently

3 lclsll The View Integration Problem That means that... => Samelsimilar information is entered multiple times => Related information must be kept consistent manually Problem is that... => often not apparent what information is same/similar => information often cannot easily be 'translated' This work is about integrating architectural views in UML so that it provides more than just structural assistance and allows model information to be shared among views. IclsW Integrating... what and why? JSC Center for Software Engineering Why rchitecture? => Still 'highlevel' enough for defects to be less 'catastrophic'. => lready 'lowlevel' enough to be less ambiguous. Why OOIUML? => Because both dominate the marketlstandardized. => UML is used even beyond 00. => Because their views are commonly understood and used. => UML Notation is extensible. => Some progress made by others.

4 Iclsla Outline JSC Center for Software Engineering Problem Solving in UML Examples of Mismatches Integration ctivities Goals and Limitations 3 Mismatch Example 1 I._......_.... '.. Model Elements of Layer 1. Flight Controller, :. Mechanic $., +...,..... dependency of Flight to Fliaht Controllernot L reflected in lower level viewy "1 I / U. 7 I i ircraft ' /, [ Flight uthorization 1 1 Model Elements of Laver 2 r, I3 oe ing 747 a Pibl Flight plan?... <: '\

5 ID invalid Enter messed patient not,found a / input visiting record validated 1 created yg@g +!2!3 &GQ@ 6: add visiting record Iclsl~I Outline. ; Problem Solving in UML Examples of Mismatches Integration ctivities Goals and Limitations

6 Iclslfl Transformation SCED JSC Center for Software Engineering 1 transform \\ : // Scenarios het ~atieht I into State (?pond Diagrams. i create vgiting recc NOTE: here /Gg)ecorb patient is \ / j created by screen created i i ; ~eate Witing recdrd date htient (Gg)eM* createdl Work from Koskimies, Systa, Tuami, and Mamist6 University of Tampere, Finland & I ~clslfl Transformation SCED I Derived (update State Diagram using SCED t patient State Diagram for Screen [patient found] do: create check for patient]

7 Iclslfl Comparison JSC Center for Software Engineering Derived State Diagram Compare derived state diagram with original one reveals potential design mismatches do: create I Original State Diagram 1... I...,_ 1 rcheckpr patient]... ID invalid... input ID valid visiting record validated checked found created Iclslfl Transformation RoseIrc hitect Fli ht *.. ~ircraft 1 :. L '.'., ( Flight uthorization Using RoseIrchitect to abstract class diagrams I Boeing 747 : Mechanic / I Use Rule 4,.) Boeing 747 I......:...,.,......

8 I 1 Derived (abstracted) Class Diagram I Compare derived state..... diagram with.... original one reveals original class Diagram] incompleteness..... * , Flight pilot_/ _ I lclsld View Integration ctivities View Synthesis Mapping (CrossReferencing) names View nalysis traces f \ L J Differentiation (Comparison) check for consistency and completeness Gransformation (Extraction) CollD to StateD Roserchitect Merging and Splitting '& J

9 Iclsld Outline Problem Solving in UML Examples of Mismatches Integration ctivities ra"".. Goals and Limitations 6 IclslH Goals and Limitations usc Center for Software Engineering Issues: What other Techniques? State Explosion Problem? Mismatch Identification vs. Resolution? Scalability? Great Benefits: => There RE automated ways of identifying mismatches between views. => Computer is more efficient in comparing views. => Mismatches may be identified as early on as they are created (e.g. agents).