Some thoughts about product line engineering using UML

Some thoughts about product line engineering: using UML, B, and refinement emerging from CS 2/ATEC A research programme arising from EU project IST 511599 - RODIN Mike Poppleton, Colin Snook (ECS, University of Southampton) {mrp, cfs}@ecs. soton. ac. uk Ian Johnson (AT Engine Controls (ATEC)) IJohnson@atenginecontrols. com July 2005 REFT workshop, Newcastle 1

ATEC case study – state of play • Requirements document = “First-cut” generic model + instance model exists as RODIN deliverable • Validated generic static model exists Work in progress/ planned: • Construction of validated generic & instance dynamic models – via UML-B, refinement • Proposal of refinement (and abstraction ? ) patterns for FM domain in product line engineering (PLE) PLE: The engineering, from a generic set of requirements for a given domain, of product instances configured to specific requirements. July 2005 REFT workshop, Newcastle 2

One of many issues: feature modelling … from Muthig* July 2005 REFT workshop, Newcastle 3

Issues • Methodological investigations: in PLE – much emphasis on domain analysis/ engineering – the quest for good generic abstractions modelling variability and commonality • and consequent instantiation transformations – (not) surprising degree of difference between PLE domains … July 2005 REFT workshop, Newcastle 4

Product instantiation - ATEC + generic instance July 2005 + 120% 0 % 100%/ sec Engine Stood 120% 100%/ sec Engine Starting 120% 50 % 100%/ sec Engine Running REFT workshop, Newcastle product instance 5

Product line issues - ATEC variability ? + generic instance data management ? instance July 2005 + 12 0% 0 % 100%/ sec Engine Stood 12 0% 100%/ sec Engine Starting 12 0% 50 % 100%/ sec Engine Running REFT workshop, Newcastle product instance 6

Variability ? product instantiation decision model + July 2005 REFT workshop, Newcastle 7

Instance data management ? product line feature data 120 % 0% 100%/s ec Engine Stood 120 % 100%/s ec Engine Starting 120 % 50 % 100%/s ec Engine Running 12 0% 0 % 100%/ sec Engine Stood 12 0% 100%/ sec Engine Starting 12 0% 50 % 100%/ sec Engine Running July 2005 product instantiation decision model product instance feature data + REFT workshop, Newcastle 12 0% 0 % 100%/ sec Engine Stood 12 0% 100%/ sec Engine Starting 12 0% 50 % 100%/ sec Engine Running 8

Links to some related work … • • “preprocessor”-style specification of variability as per Muthig*, applied to UML-B ? a more sophisticated analog of MDA/metamodel approach of Muthig**, i. e. MOF + UML-B ? ? ? Muthig* Go. Phone - A Software Product Line in the Mobile Phone Domain Dirk Muthig et al, IESE-Report No. 025. 04/E (Fraunhofer Institut Experimentelles Software Engineering Muthig** Model-Driven Product Line Architectures, SPLC 2002, LNCS 2379 July 2005 REFT workshop, Newcastle 9

More links to related work … Notions of refinement … • Wasowski*: refinement (by trace inclusion ) by model restriction from the maximal feature model. A slightly odd notion of refinement • Batory*: “feature refinement”: algebraic approache to scaling stepwise refinement: facets, aspects, mixin layers, gluons and origami … A very odd notion of refinement – ‘The definition of “refinement” that seems closest to our intended meaning is “the act of making improvement by introducing subtleties or distinctions” (Merriam. Webster’s Dictionary). . At the implementation level, however, a refinement can yield dramatic changes: both the exported functionality (semantics of operations) and the exported interface (signatures of operations) may change. Thus, unlike the use of “refinement” in formal approaches to programming, the set of allowed behaviors of our “refined” program might not be a subset of the behaviors of the “unrefined” program. ’ Wasowski* Automatic generation of Program Families by Model Restrictions, SPLC 2004, LNCS 3154 Batory* many papers; see http: //www. cs. utexas. edu/users/schwartz/pub. htm July 2005 REFT workshop, Newcastle 10

More links to related work … Notions of refinement … • Wasowski*: refinement (by trace inclusion ) by model restriction from the maximal feature model. A slightly odd notion of refinement • Batory*: “feature refinement”: algebraic approache to scaling stepwise refinement: facets, aspects, mixin layers, gluons and origami … A very odd notion of refinement – ‘The definition of “refinement” that seems closest to our intended meaning is “the act of making improvement by introducing subtleties or distinctions” (Merriam. Webster’s Dictionary). . At the implementation level, however, a refinement can yield dramatic changes: both the exported functionality (semantics of operations) and the exported interface (signatures of operations) may change. Thus, unlike the use of “refinement” in formal approaches to programming, the set of allowed behaviors of our “refined” program might not be a subset of the behaviors of the “unrefined” program. ’ – So isn’t this just … retrenchment ? (e. g. Banach, Poppleton 2003) Wasowski* Automatic generation of Program Families by Model Restrictions, SPLC 2004, LNCS 3154 Batory* many papers; see http: //www. cs. utexas. edu/users/schwartz/pub. htm R. Banach, M. Poppleton Retrenching Partial Requirements into System Definitions: A Simple Feature Interaction Case Study, 2003, Req. Eng. J. , 8 (4) July 2005 REFT workshop, Newcastle 11

This is RODIN-CS 2 Looking forward to explorations in … “Product Line Refinement” July 2005 REFT workshop, Newcastle 12

Decomposing the example Work July 2005 REFT workshop, Newcastle 13

Decomposing the example Work July 2005 REFT workshop, Newcastle 14

Decomposing the example Work July 2005 REFT workshop, Newcastle 15

Decomposing the example Work July 2005 REFT workshop, Newcastle 16