Presentation on theme: "Research and Technology Object Oriented Defect Detection Frank HoudekForrest Shull DaimlerChrysler AGFraunhofer Center - Maryland Research and Technology."— Presentation transcript:
Research and Technology Object Oriented Defect Detection Frank HoudekForrest Shull DaimlerChrysler AGFraunhofer Center - Maryland Research and Technology
Agenda Overview, Goals of the Session Murray Wood: Challenges in Object-Oriented Code Inspection ISERN Survey Outcome Classification Schemes Thilo Schwinn: Quality Gate Driven Definition of Classification Schemes Reading Techniques and Strategies Forrest Shull: A Set of OO Design Reading Techniques Stefan Biffl: Comparison of Checklists to Scenario-Based Reading Partitioning of Artifacts Andreas Birk: PBR applied to OO designs Discussion and Planning of Future Steps
Research and Technology OO Inspections - A Multi-Faceted Selection Problem Input: Artifact type Properties - Size - Language - Standards used Domain Inspectors Inspection Goals Effectiveness (%defects found) Efficiency (#defects/time) Focussed follow-up activities Learning/Training Inspection Method Classification Scheme Reading technique and strategy Partitioning of artifacts Auxiliary Material Meeting & Documentation techniques Outcome Effectiveness Effort Efficiency Focussed follow-up Learning/Training Statisfaction Selection
Research and Technology Open Issues Empirical results for many combinations of Input/Goal/Method Consistent scheme for reporting results ( Framework) Evolution and tailoring of existing methods Emphasis in this track: Classification schemes Reading techniques and strategies Partitioning of artifacts
Research and Technology Potential Outcome of the Session Building a common framework for OO-DD experiments (e.g. by using the proposed multi-faceted selection problem framework) Building a repository of knowledge, e.g. a common repository of OO-DD experiment descriptions (using a unique form?) Post-mortem analysis of already performed experiments First version of a selection mechanism for (and especially identifying most important influence factors) Classification schemes Reading techniques and strategies Partitioning of artifacts
Research and Technology Discussion Question 0: The Right Taxonomy? Is the proposed framework for inspection methods (slide 3) a useful one for organizing research? Is some aspect missing? Is it a logical way to organize the work?
Research and Technology Discussion Question 1: Building a body of knowledge? Are there common trends in the results reported here? E.g.: OO design inspections seem feasible… OO design inspections seem effective… OO reading is an effective way to perform inspections… OO reading is an effective way to perform inspections for certain users… What re-analysis can be done to further support these hypotheses? Who’s going to do it? We need names! Output: Can we get a joint paper that summarizes the results of our independent studies and draws some conclusions?
Research and Technology Discussion Question 2: Component pieces of a process? Can we aggregate the results into an approach to doing OO inspections? E.g. We have discussed ways to do defect classification reading partitioning of documents etc. Can they work together? Do we know when to use the different approaches?
Research and Technology Discussion Question 3: An experience repository? ISERNers often talk about experience/data repositories… (Have any ever really gotten going?) Imagine a “lightweight” repository for OO inspections. What would you want to get out of it? What would make it worth the effort? How much effort / What kind of contribution is reasonable to expect from participants? Who will contribute? Who will manage? We need names!
Research and Technology Survey: Object Oriented Defect Detection Experience in ISERN (1) Sent out to all ISERN members Returns: 9 (6 filled out questionnaires [2 Industry, 4 University] from 5 partners, 4 references or ‘no contribution’) Questions: Artifacts Scope and size in one inspection (partitioning) Mechanism of walking through the artifacts and identify objects Used classification scheme Findings
Research and Technology Survey: Object Oriented Defect Detection Experience in ISERN (2) SourceEricssonTU ViennaIESEDCUniv. UlmUMD ArtifactsUse casesRequirementsUML designRequirementsOctopus OOARequirements Sequence diag.Use casesUse cases State diags.HL designUML diags. UnitArround classesAllLogical entitiesWholeWholePairs of documentdocumentdocuments Artifact size 1-2 classes20-30 pages30 pages40-80 pages10-20 pages6 classes (in inspection) Partitioning--- ---n.a.n.a.n.a. criteria Naming elements---------line no.page, line--- in the meeting ClassificationError noneNon-critical Ommision schemeSuperfluous Important Incorrect fact Improvement CriticalInconsistency MissingOptimization Ambiguity QuestionExtra Information vertical/horicontal1 Difference toEasiern.a.NoneNonen.a. structured artifacts Defect classificationOkFine enough fast, stableno problemsreading pre- for analysis, dominates defects but robust