Eric Kenseth History Terminology Aspect Oriented Programming Analysis Design Patterns Testing Maintenance Conclusion Early 90s, U of Twente, Netherlands Subject-Oriented 1993, IBM Researched similar subjects Feature-Oriented 1997-2000, Xerox PARC develops AspectJ True birth of Aspect-Oriented Concern Anything any stakeholder wants the system to do. Anything, any stakeholder Cross-cutting Concern Far-reaching concern Applies to many subsections of the system The reason for Aspect Oriented to exist Separation of Concerns Removal of sections in a module that deals with other concerns Joinpoint Location where aspects can take over execution Depends on language as to where is supported Method Calls Method Returns Exception Throws Reading/Writing to a field • Advice – The code of a cross-cutting concern – Needs to be executed in many places • Pointcut – Points in the program where the advice needs to be executed • Joinpoints • Inter-Type declaration • Declared member of another class in an aspect • Visible only to classed declared to have access • Aspect – A class-like structure to encapsulate cross-cut concerns – Can be static or dynamic – Can have fields and methods as members – Can be abstract or not – Can be instantiated – Can have pointcuts, advice, and inter-type declairations – May be ‘privileged’ to access private members of other types Pointcut defines where to execute the Advice Advice may modify program flow Advice may access values in the context of the cut Aspects encapsulate a cross-cut concern Why? Modularization Tangling Scattering How? • What are the concerns? • Which of these are core concerns? • Which are cross-cutting concerns? – – – – – – – – – – Logging Caching Security concerns Error detection Error correction Memory management Real-time issues Synchronization Mutual exclusion Complex rule monitoring Aspect use cases Generalized in situation Detailed in steps Not useful for the client, only the designers Other use cases include the aspect cases a lot Not well supported by UML For now no formally accepted answer Several extensions proposed Some suggest separating base code from aspects in design Very useful when designing with aspects Demonstrates when aspects are invoked Can help spot where pointcuts need to be set Spectator Pattern ‘Harmless advice’ Doesn’t modify the other concerns, just watches it Loggers, Tracers, Profilers Regulator Pattern Enforces requirements on the system state May control the flow if errors occur Security Authorization, System Constraints Patch Pattern Enhances or Modifies features being Reused Aspects act as methods for a class Add new methods to classes you don’t have access to rewrite Modify methods Allows updating the reused asset without re-modification Must be careful about breaking assertions Extension Pattern Making an extendable system is problematic Aspect solution: Create extension points Empty method calls where the system will be extended Extension aspects use these points to add to the system Independent evolution of base code vs. extensions Language Specialty language needed. Compiler/Interpreter needs to support aspects Most languages only have minor changes to them AspectJ Dozens of other Java-based C/C++/C# based languages Many other languages have Aspect-supporting versions of them Even Cobol Development aspects can help test and debug base code Logging Tracing Profiling Performance Monitoring These aspects have several advantages Easier to add/remove to the system as aspects Aspects require different testing means Lack independent identity May be coupled to context New ways for aspects to cause faults Incorrect pointcut patters Incorrect aspect ordering Incorrect context checking in pointcuts Systematic testing means are lacking Add new features without disturbing code. Separation makes modifying cross-cut concerns much easier Untangling makes modifying primary concerns safer Unlikely to break other concerns while modifying. Modifying Joinpoints may be dangerous still. Code is modular No tangling with cross-cut concerns in components No scattering of cross-cut concerns in other concerns Many aspects are features useful in almost all systems Logging Tracing Profiling New and emerging Still needs refining in ways Designing Testing Shows promise in certain situations Big, Complex, with multiple user-roles Patching and Extending [1] O. Aldawud, T. Elrad, and A. Bader. (2003) UML profile for aspect-oriented software development. In Proceedings of the AOM workshop at AOSD, 2003 [2] Hachani, O., Bardou, D. (2002) Using Aspect-Oriented Programming for Design Patterns Implementation. In: OOIS 2002 Workshop on Reuse in Object-Oriented Information Systems Design [3] Blair, G., Blair, L., Rashid, A., Moreira, A., Araujo, J., Chitchyan, R. (2005). Engineering Aspect-Oriented Systems. In Fillman, Elrad, Clark, Aksit (Eds.), AspectOriented Software Development (pp.380-398). Boston: Addison-Wesley. [4] James Noble, Arno Schmidmier, David J Pearce, Andrew P Black (2007) Patterns of Aspect-Oriented Design. In Proceedings of European Conference on Pattern Languages of Programs. Irsee [5] Filman, R. & Friedman, D. (2005). Aspect-Oriented Programming Is Quantification and Obliviousness. In Fillman, Elrad, Clark, Aksit (Eds.), Aspect-Oriented Software Development (pp.1-7). Boston: Addison-Wesley. [6] R. T. Alexander, J. M. Bieman, and A. A. Andrews. (2004) Towards the systematic testing of aspect-oriented programs. Technical Report CS-4-105, Department of Computer Science, Colorado State University, Fort Collins, Colorado. [7] AspectJ. (n.d.). Retrieved October 31, 2009 from Eclipse website, www.eclipse.org/aspectJ Questions?