12. Finding Components in Component Repositoriesst.inf.tu-dresden.de/.../teaching/ss15/cbse/slides/... · Prof. U. Aßmann, CBSE 23 Example: Service Facets in a UNIX System To describe
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
• http://simile.mit.edu/wiki/Longwell • http://simile.mit.edu/exhibit • http://flamenco.berkeley.edu • http://search.express.ebay.com • http://base.google.com ► FacetMap: Greg Smith, Mary Czerwinski, Brian Meyers, Daniel Robbins,
George Robertson, Desney S. Tan. FacetMap: A Scalable Search and Browse Visualization. IEEE Transactions on visualization and computer graphics, vol.12 , No. 5, september/october 2006.
► Thorsten Teschke. Semantische Komponentensuche auf Basis von Geschäftsprozessmodellen. Dissertation. Universität Oldenburg, 2003.
► Facet-based search of computer science literature in DBLP repository ► http://dblp.l3s.de/?q=&newQuery=yes&resTableName=query_resultOsC5mC
§ Luca de Alfaro and Thomas A. Henzinger, Interface automata, ACM SIGSOFT FSE/ESEC, 2001
It should be as easy to find good quality reusable software assets as it is to find a book on the internet
http://softwarereuse.nasa.gov/
Prof. U. Aßmann, CBSE 5
Component Repositories
§ Components must be stored in component repositories with metadata (markup, attributes) to find them again
§ Description by Metadata: • Attributes: Keywords, Author data • Contracts (Usage protocols, behavioral specifications)
State machines Sequence diagrams Contracts (pre/post/invariants)
§ Examples of Component Repositories • CORBA
implementation registry interface registry
• COM+ registry • Commercial Component Stores www.componentsource.com • Debian Linux Component System (apt, dpkg) • CTAN TeX Archive
Prof. U. Aßmann, CBSE 6
Why Searching Components?
§ A public component repository is called a market, managed by a trader (broker)
• Distributing or selling components • Companies can register components at the the trader • Customers can search components in the markets and buy or rent them
§ Searching for functionality (interface, contract, protocol) • Reuse instead of build • Searching for components to replace own ones • Semantic substituability (CM-S) should be ensured
§ Searching for quality features • Performance, energy consumption, reliability
12.2 Searching and Browsing with Faceted Classifications
(thanks to Jan Polowinski)
Prof. U. Aßmann, CBSE 8
Faceted Classification for Better Matchmaking
► A facet is a dimension of a classification ■ Facets simplify search: Facet classification has been invented in library science to
simplify the description and search for books [Ranganathan]. ■ A component (or service) is described in several facets, dimensions, which are
orthogonal to each other
► Matchmaking engines can look up a service by stating the desired properties for all facets.
► Classifications can be arranged in facets if several partitions of a group of objects exist that are orthogonal ■ In domain modelling, this is often the case ■ Without facets, multiple inheritance hierarchies have to be specified, which are
often clumsy and error-prone
► Idea: use facets for better matchmaking
Prof. U. Aßmann, CBSE 9
Comparison
Standard Classification ► V Vögel
■ V1 Atmung der Vögel ■ V2 Fortpflanzung der Vögel
► F Fische ■ F1 Atmung der Fische ■ F2 Fortpflanzung der Fische
► S Säugetiere ■ S1 Atmung der Säugetiere ■ S2 Fortpflanzung der Säugetiere
► I Insekten ■ I1 Atmung der Insekten ■ I2 Fortpflanzung der Insekten
12.3 Faceted Metadata for Search in Component Repositories
Prof. U. Aßmann, CBSE 23
Example: Service Facets in a UNIX System
► To describe the services of a UNIX system, [Prieto-Diaz] employed a 4-faceted scheme ■ function ■ logical object ■ implementation object ■ tool
► UNIX services can be described with appropriate facet values and looked up in a repository
► Example: “append a line to a file with a text editor” ■ (function = append, logical class = line, implementation class = file, tool = text
editor):
Function Logical class
Implementation Class
Tool
edit Line File Text editor
Prof. U. Aßmann, CBSE 24
Example: Services in a UNIX System
► [Prieto-Diaz] already suggested to use controlled vocabulary (domain ontologies) to improve the effectiveness of the search: ■ If every facet is described by an ontology, the service descriptions are
standardized for a user group and improve understanding of service semantics.
► Facets simplified the description of the components, improved the understanding of their domain, and facilitated the search in component libraries.
Prof. U. Aßmann, CBSE 25
And for Components?
Prof. U. Aßmann, CBSE 26
And for Components?
Prof. U. Aßmann, CBSE 27
Other Advantages
► The facet classification is rather immune to extensions ■ Extending one facet leaves all others invariant ■ Example: If Europe is extended with a new member state, the matchmaking
algorithm can deliver new courses from the new member state, without affecting the rest of the semantic specifications at all
► The accuracy can be improved by synonym lists (thesauri) ■ Synonyms increase the chances for a match ■ They permit to search not only for keywords, but also for their synonyms
(assembled in a thesaurus) ■ Beyond synonyms other refinement relations of concepts can be used to improve
the search
■ Example: Great Britain is used as a synonym for England, Scotland, and Wales. Synonyms allows for matchmaking on any of the keywords, so that students looking for a course need not bother about geographic and political details.
Prof. U. Aßmann, CBSE 28
The Use of Ontologies in Faceted Matchmaking
► Ontologies simplify matchmaking by standardization ■ Since they provide standardized terminology and standardized
ontological relations between the terms, queries can specify . keywords with a precise, shared, and standardized meaning (semantic
search), . contextual information for search in context, where the context is defined by
the ontological relations of the terms.
► Example: ■ A web course on IT basics can be queried by the standardized word IT-
basics (being semantic search) ■ also in context, by relating it to courses such as IT-advanced or IT-
preparatory (contextual search) . “find me an IT basics course, which has a preceding preparatory IT course
and has a follow-up advanced IT course“
Prof. U. Aßmann, CBSE 29
Putting up a Component Repository for Your Company
► Define facets for component metadata ■ If possible, reuse an ontology for a facet ■ Form a thesaurus for synonyms ■ Store the metadata as a tuple in the database
► Realize a search algorithm that uses facets together with thesauri ► Or use a faceted browser with the metadata
§ An object with a natural type (entity type) lives on its own and exists independent of context and collaborators (see DPF) ■ The type does not depend on other types (independent type) ■ Types that depend on others are called dependent types. ■ Role types, facet types, part types are dependent types.
. Hotel vs. HotelRoom
. Car vs. Screw or Motor
. Book vs. Reader
§ A big object (bob) is complex, hierarchical object with a natural type
• Usually, it has subobjects with dependent types, role types and others.
► A business object (domain object) is a bob with a natural type of the domain model (business model) ► Usually, business objects (domain objects) are large hierarchical objects ■ They can consist of thousands of smaller objects of dependent types (part-of
relation) ■ They can play many roles with context-based types
Prof. U. Aßmann, CBSE 32
Component Specification with UML Components
• A UML component is a hierarchical class for big objects with provided and required interfaces (roles)
• Provided interfaces (provided roles) use „lollipop“ notation • Required interfaces (required roles) use „plug“ notation
• UML components can specify big objects with one natural core object and many dependent subobjects Some components are required to use specific other interfaces
<<comp spec>> CompanyMgr ICompanyMgt
<<comp spec>> CompanyMgr
ICompanyMgt
IAddressMgt
<<comp spec>> AdressMgr
Prof. U. Aßmann, CBSE 33
Ports of UML Components
§ A port is a connection point of a UML component. • A port has a set of roles (interfaces) • It may be represented by a port object (gate)
System
Port Provided interfaces Required
interfaces
Prof. U. Aßmann, CBSE 34
Lollipops und Plugs (Balls and Sockets)
► For a UML component, provided and required interfaces can be distinguished
n A required interface specifies what the current class needs to execute.
<<provided>> Addresses
<<required>> Text AddressManager
listAdresses() listAdresses() sort()
Adresses
Text
Prof. U. Aßmann, CBSE 35
Ports
► Ports consist of port classes with interfaces and behavior in form of interface automata
n provided: normal, offered interface n required: used, necessary interface
Component
<<provided>> Port class
<<required>> Port class
Component
Port
Prof. U. Aßmann, CBSE 36
Nesting of UML Components
► UML components n Ports are connected by links (connections) n Delegation link: links outer and inner port
DocumentSystem Link/connection Delegator
Text Manager
Address Manager Adresses
email
email Manager
Text
Forms
Buffer
Lines TextRep
IText
IForm
Prof. U. Aßmann, CBSE 37
Refinement of UML Components
► UML components are nested, i.e., are bobs. ► Nesting is indicated by aggregation and part-of relationship. ► Nesting is introduced by an encapsulation operator encapsulate.
Document System
Document System
Text Manager
Address Manager Adresses
email email
Manager
Text
Forms
Buffer
Lines TextRep
IText
IForm
encapsulate
decompose
Prof. U. Aßmann, CBSE 38
Encapsulation means Aggregation
► Nesting means Aggregation n A UML component is a package and a façade for all subcomponents
12.5 Searching in Component Repositories by Contract Conformance
Contract Conformance means semantic substituability
Prof. U. Aßmann, CBSE 40
Ports can be Equipped with Interface Automata Contracts
► Ports consist of port classes with interfaces and behavior in form of interface automata (port automata, protocol automata)
n provided: normal, offered interface n required: used, necessary interface
Component
<<provided>> Port class
<<required>> Port class
Component Port
Interface automaton
Interface automaton
Prof. U. Aßmann, CBSE 41
Component Protocols with Operational Contracts
§ The port protocol automata can be composed to a component protocol automaton § Components may have a protocol automaton in which their ports, services,
procedures should be called, invoked, or signalled • The provided protocol specifies in which order the services can be invoked (given by a provided interface
automaton)
• The required protocol specifies in which order the services can be invoked (given by a requried interface automaton)
§ The order of component invocation can be specified by a language over the alphabet of the ports, services, procedures (state-based protocol contract, operational contract)
• Finite state automaton (regular language) UML state chart (Hierarchical finite state machine, prococol machines)
Data flow diagram • Stack machine (context-free language)
• Petri net (regular dialects, context-free and context-sensitive dialects)
§ The contract provides an abstraction of the implementation of the component • Implementations must be proven to be conformant to the procotol
§ The conformance checking is decidable if the protocol language is decidable § Sets of paths over states (words over state and edge alphabet)
Prof. U. Aßmann, CBSE 42
Subsumption of Interface/Port Protocols
§ A provided-port protocol Pr(P1) is-contained-in (is smaller than, is-weaker-than) a provided-port protocol Pr(P2), if Pr(P1) <= Pr(P1)
§ A port protocol Pr(P1) subsumes (contains, is-stronger-than) a port protocol Pr(P2), if Pr(P1) > Pr(P2), i.e., if the set of guarantees of P1 is larger than that of P2, i.e., it delivers more
§ Then, port P1 is conformant to P2 and P1 can substitute P2 in connections, because P1 delivers less paths of states than P2
§ Subsumption checking of ports, and thus, conformance checking, should be decidable (protocol language should be decidable)
Pr(P1) = a b b* c <= Pr(P2) = a b* c
Component
<<provided>> Port Class
P1 a b
c b
Component
<<provided>> Port Class
P2 a b c
<=
Prof. U. Aßmann, CBSE 43
D:Component
r r r s t
<<provided>> Port Class 2
Subsumption of Guarantees (Provided Protocols)
§ A component‘s guarantee (provided protocol) G(C1) is-contained-in (is-weaker-than, offers-less) a second component‘s guarantee G(C2), if forall ports P in C1.providedPorts Exists Q in C2.providedPorts: Pr(P) <= Pr(Q)
• C1 offers/delivers less than C2
§ A component‘s guarantee G(C1) subsumes (is-stronger-than, offers-more) a second component‘s guarantee G(C2), if for all ports P in C1.providedPorts exists Q in C2.providedPorts: Pr(P) > Pr(Q)
§ Then, C1 is guarantee-conformant to C2 and can substitute it in output connections
a b b* c
<<provided>> Port Class 1
<= E:Component
r* s t
<<provided>> Port Class 2
a b* c
<<provided>> Port Class 1
Pr(D.PortClass1) = a b b* c <= Pr(E.PortClass1) = a b* c Pr(D.PortClass2) = r r r s t <= Pr(E.PortClass2) = r* s t
Prof. U. Aßmann, CBSE 44
D:Component
r* s t
<<required>> Port Class 2
Subsumption of Assumptions (Required Protocols) § A component‘s assumption (required protocol) A(C1) contains (is-stronger-than, demands-
more) a second component‘s assumption A(C2), if for all ports P in C1.requiredPorts holds: • Exists Q in C2.requiredPorts: Pr(P) > Pr(Q) (required ports of C2 have smaller or less assumptions)
• Q‘s required ports require less assumptions than P‘s required ports
§ A component‘s assumption A(C1) subsumes (is-contained-in, is-weaker-than, demands-less) a second component‘s assumption A(C2), if for all ports P in C1.requiredPorts holds:
• Exists Q in C2.requiredPorts: Pr(P) <= Pr(P) (required ports of C2 have larger and more assumptions) § Then, C1 is assumption-conformant to C2 and can substitute it in input connections
a b* c
<<required>> Port Class 1
> E:Component
r s t
<<required>> Port Class 2
a b c
<<required>> Port Class 1
A(E.ReqPortClass1) = a b c <= G(D.ReqPortClass1) = a b* c A(E.ReqPortClass2) = r s t <= G(D.ReqPortClass2) = r* s t
Prof. U. Aßmann, CBSE 45
Subsumption of Guarantees and Counterwise Subsumption of Assumptions
§ A component protocol Prot(C1) can subsume a component protocol Prot(C2) if G(C1) <= G(C2) and A(C2) <= A(C1)
• If C1‘s assumptions are weaker than C2‘s • And if C1‘s guarantees are stronger than C2‘s
§ Then, C1 is conformant to C2 and C1 can substitute C2 in input and output connections
§ C1 can be substituted for C2, if • C1 delivers more or equal than C2 and (more performant) • C1 demands less or equal than C2 (less demanding)
Prof. U. Aßmann, CBSE 46
Component
Subsumption of Protocols
§ A component protocol P(C1) can subsume a component protocol P(C2)
• P(C1) <= P(C2) • For finite automata: P(C1) P(C2)
§ Then, C1 is conformant to C2 and C1 can substitute C2 § Subsumption checking and thus, conformance checking, should
be decidable (protocol language should be decidable)
!
<<provided>> Port class
<<required>> Port class
Prof. U. Aßmann, CBSE 47
Searching by Protocol
§ A component C can be found in a repository, if a query protocol Q is given with Q <= P(C)
§ Search consists of subsumption checking with all component protocols in the repository
§ Query protocols can be: • Metadata about the component • Provided protocols • Required protocols • Provided and required protocols
Prof. U. Aßmann, CBSE 48
Declarative Protocols
§ A protocol can also be specified as predicates over the states of a component (declarative contract)
§ Subsumption checking of protocols and conformance can be done by reasoning
• E.g., by subsumption checking of an OWL class hierarchy
Prof. U. Aßmann, CBSE 49
The End - Acknowledgements
§ Faceted browsing slides are courtesy to Jan Polowinski.
Prof. U. Aßmann, CBSE 50
Example: Finding Courses in Europe based on Ontologies
► A course in the unified Bologna world of European education can be described by several facets: ■ topic area (computer science, music, literature, etc.), ■ level of advancement (undergraduate, graduate), ■ cost (free, non-free), ■ country (Germany, Italy, WesternEurope, EasternEurope, etc.)
► Every facet can be described by an ontology, in this case on ■ topic area ■ level ■ cost ■ country
► A semantic description of a course selects one value for each facet and forms a tuple ■ A free undergraduate music course could be described by the tuple (topic area =
music, advancement = undergraduate, cost = free, country = WesternEurope).
Prof. U. Aßmann, CBSE 51
Ex. Finding Courses in Europe
► Searching a course throughout the course databases in Europe consists of comparing the tuple point-wise to database entries.
► The values need not match exactly, ■ Subsumption (inheritance) in the facet ontologies can be used to deliver
refinement of matchings. ■ Example: if free-course is subsumed by non-free-course, the matcher can yield a
free course, even if the client desired a non-free one. ■ Example: a matchmaker can return a (music, undergraduate, non-free, Germany)-