Component Search and Retrieval Advanced Reuse Seminars Eduardo Cruz
Information Retrieval Structured Documents Unstructured Documents No software documentation standard Semi-Structured Documents Calvin Northrup Mooers
Mooers' Law: “An information retrieval system will tend not to be used whenever it is more painful and troublesome for a customer to have information than for him not to have it,” 1959 Calvin Northrup Mooers
Mass Production Software components [Mcllroy, 1968]
“software industry is weakly founded, and that one aspect of this weakness is the absence of a software components subindustry” [McIlroy, 1968]
“The storage and retrieval of software assets is nothing but a specialized form of information storage and retrieval” [Mili, 1998]
Software Library Browsing – Inspecting without a predefined criterion Retrieval – Satisfy a predefined matching criterion
Classification Scheme Facet-based Better than hierarchical classification Manual classification different facets Automatic classification Controlled Vocabulary Semantic information Uncontrolled Vocabulary Big software libraries Little or no descriptors
Recall and Precision High Precision – Most retrieved elements are relevant High Recall – Few elements left behind Spreading Activation (Relaxed Search) – Related matches are retrieved Coverage – The average number of assets that are visited over the total size of the library
Asset Representation Library representation is made in full knowledge of the artifact. User representation is made in ignorance of the artifact Asset representation is purposefully abstract to capture important features while overlooking miner or irrelevant details Asset's surrogate is used in retrieval literature
Asset retrieval Goals Exact retrieval – Black box reuse Approximate retrieval – White box reuse Generative modification – Reusing the design Compositional modification – using building blocks of the retrieved asset
Usually non included information Interface description Non-functional requirements Interoperability
Situational Model x System Model Component retrieval model [Lucrédio et. al, 2004]
“Repository representation is made in full knowledge of the artifact at hand” “User representation is made in ignorance of the artifact” [Mili, 1998]
Scott Henninger
Tools
Component Search Tools Web Delphi Search Engine Ispey CSourceSearch.net (2004) Gonzui SourceBank Koders (2004) Codase (2005) Aplications Agora (1998) Codebroker (2002) Koders Enterprise (2004) Maracatu (2005)
Delphi Search Engine
Ispey.com
SPARS-J – (2003) Filter
SourceBank Filter
CSourceSearch.Net – (2004)
Koders.com – (2004)
CODASE – Launched Sep 9, 2005 Example Searches Browsing Multiple Search Options “…based on the number of people in your company, starting from $5,000 USD”
CODASE - Browsing
Other Tools
AGORA - Location and Indexing (1998) INTERNET JavaBeans Agent JavaBeans Introspector JavaBeans Agent JavaBeans Introspector JavaBeans Agent JavaBeans Introspector AltaVista Search Index Server Filter INDEX AltaVista Query Server Web Server
Component Rank (1998) V1V1 V3V3 V2V D12 = 0.5 D13 = 0.5 D23 = 1 D31 = 1 Nodes v Edges e Graph G Weight w Distribution Ratio d
“Classes defining data structures and their containers are highly ranked”
Clustered Component Graph V3 V2 V1 V1 ≡ V4, V2 ≡ V6 V7 V6 V4V5 V7 V’26 V’14V’5 V’3
NO MORE MULTIPLE DISCONNECTED COMPONENTS V3 V2 V1 V7 V6 V4V5
Component Rank System Architecture.java file ≡ component (1) Similarity Measurement (2) Clustering (3) Use Relation Extraction (4) Component Graph Construction (5) Component Rank Computation by Repetition (6) De-Clustering to Original Component Graph INPUT OUTPUT Order of Weights ≡ Component Rank of.java files
Simple Copied Components A B A B X Y Copied Components Other Components Non-clustered component Graph A’ B’ X’ Y’ 1/4 Clustering Before Weight Computation 1/4 A’ B’ X’ Y’ 1/3 Clustering After Weight Computation 1/3 1/6
DO NOT COUNT SIMPLY DUPLICATED COMPONENTS
Copied AND MODIFIED Components A B A C X Y Copied and Modified Components Other Components Non-clustered component Graph X’ Y’ Clustering Before Weight Computation 1/5 Original Components A B’C’ 2/5 1/5 X’ Y’ Clustering Before Weight Computation 1/5 1/6 A’ B’C’ 1/3 1/6
Beyond Searching and Browsing Searching and browsing Require users to initiate the information seeking process Information access and Information Delivery
CodeBroker – (2001) Components repositories are often so large that software developers cannot learn about all of the components Component repositories are not static New components added Old components updated Context-Aware browsing
May not have suficient knowledge about the reuse repository May perceive that reuse costs more than developing from scratch May not be able to use the repository by formulating a proper query May not be able to understand the found components
Belief Vaguely Known Information Islands Well Known L4: Entire Information Space Unknown components
L3: Belief L2: Vaguely Known CodeBroker L1: Well Known L4: Entire Information Space Information Use: L1 – Use by Memory L2 – Use by Recall L3 – Use by Anticipation L4 – Use by Delivery Already Known Components Irrelevant Components Task Relevant Information
Program Aspects Concept Formal Informal Indentation, comments, identifier names (semantic) Executability Code Constraint environment Signature
Information delivery Feedback After execution of the action Feedforward Affects the execution of the action
Information delivery Interruptive Noninterruptive
Latent Semantic Analysis (LSA) Synonymy Polysemy “Text documents and queries are represented as vectors in the semantic space, based on the words contained and the similarity between a query and a document is determined by the distance of their respective vectors”
Comments signature Discourse model User model
Koders Enterprise – (2004)
M.A.R.A.C.A.T.U. – Modern Architecture for Retrieving All Components At The Universe (2005)
Using Structural Context to Recommend Source Code Examples Reid Holmes and Gail C. Murphy University of British Columbia Software Practices Lab
The Problem: A Concrete Example Frameworks can improve developer productivity. But developers can become stuck trying to use the APIs Imagine trying to use the Eclipse APIs to place text in the status line of the Eclipse IDE Eclipse has 38,000 public methods
Structural Context Project Repository Development Environment Examples Using Structural Context to Recommend Source Code Examples - Reid Holmes and Gail C. Murphy
Strathcona: Extract Structural Context ViewPart SampleView setMessage(String) IStatusLineManager setMessage(String)
Visual representation Highlights key relationships between example and query Multiple examples can be quickly viewed Strathcona: Example Navigation
Strathcona: Viewing Example Source Code view Example shows how to get a status line manager Example is not a perfect match, but good enough to help
Conclusion Information Delivery Similarity Analyser Ranking – Metrics Context Automatic Facet Classification Uncontrolled vocabulary + additional terms
References [McIlroy, 1968] M. D. McIlroy, Mass Produced Software Components, NATO Software Engineering Conference Report, Garmisch, Germany, October, 1968, pp [Mili, 1998] A. Mili, R. Mili, R. T. Mittermeir, A survey of software reuse libraries, Annals of Software Engineering, Vol. 5, 1998, pp [Seacord, 1998] Robert C. Seacord, Scott A. Hissam, Kurt C. Wallnau. "Agora: A Search Engine for Software Components," IEEE Internet Computing, vol. 02, no. 6, pp , November/December, 1998 [Szyperski, 1999] Szyperski C., “Component Software: Beyond Object-Oriented Programming”. Addison Wesley, 1999 [Dey, 2001] Dey, A.. Understanding and Using Context. Personal Ubiquitous Comput. 5, 1 (Jan. 2001) [Greengrass, 2001] Greengrass, Ed. Information retrieval: A survey. DOD Technical Report TR-R , 2001 [Ye, 2001] Ye, Y. and Fischer, G. Context-Aware Browsing of Large Component Repositories. In Proceedings of the 16th IEEE international Conference on Automated Software Engineering (November , 2001). ASE. IEEE Computer Society, Washington, DC, 99. [Ye, 2002] Y. Yunwen and G. Fischer. Information delivery in support of learning reusable software components on demand. In Proceedings of the 7th international conference on Intelligent user interfaces, California, USA [Ye, 2002] Ye, Y. and Fischer, G. Supporting Reuse by Delivering Task Relevant and Personalized Information. In Proceedings of the 24th International Conference on Software Engineering. p , Orlando, Florida, May, 2002
Bibliography [Inoue, 2003] K. Inoue et al.: "Component Rank: Relative Significance Rank for Software Component Search", Proceedings of ICSE 2003 [Maxville, 2003] Valerie Maxville, Chiou Peng Lam, Jocelyn Armarego. "Selecting Components: a Process for Context-Driven Evaluation," apsec, p. 456, 10th Asia- Pacific Software Engineering Conference (APSEC'03), 2003 [Maxville, 2004] Valerie Maxville, Jocelyn Armarego, Chiou Peng Lam. "Intelligent Component Selection," compsac, pp , 28th Annual International Computer Software and Applications Conference (COMPSAC'04), [Prado, 2004] Lucrédio, D.; Almeida, E, S.; Prado, A, F. A Survey on Software Components Search and Retrieval, In the 30th IEEE EUROMICRO Conference, Component-Based Software Engineering Track, 2004, Rennes - France. IEEE Press,2004 [Holmes, 2005] Holmes, R. and Murphy, G. C Using structural context to recommend source code examples. In Proceedings of the 27th international Conference on Software Engineering (St. Louis, MO, USA, May , 2005). ICSE '05
“Imperfect technology in a working market is sustainable; perfect technology without any market will vanish” [Szyperski, 1999]