xml object database1 a logic programming approach to supporting the entries of xml documents in an...

XML Object Database 1

A Logic Programming Approach to A Logic Programming Approach to Supporting the Entries of XML Documents in Supporting the Entries of XML Documents in

an Object Databasean Object Database

Ching-Long Yeh 葉慶隆Department of Computer Science and Engineering

Tatung University

Taipei 104, [email protected]


IntroductionIntroduction

• XML improves upon HTML in – capturing the meaning of a document and – extending the tag set.

• At the same time, it also reduces the complexity of SGML.

• It is believed that XML will soon be the standard of data exchanges on the Web.



• Due to lack of indices in files, we are not able to make full use of the meaning (or metadata) in an XML document, if it is stored in a file.

• Since an XML document can be easily viewed according to the object-oriented model, a promising solution is to employ object database technology to manage the access of XML documents.



• In this talk, I will present our research in– the design and implementation of an XML object

DB, and– an extensible template-based query interface to

accessing to XML object database


The Remainder of the TalkThe Remainder of the Talk

• An Introduction to XML• Design and Implementation of an XML Object

Database• An Extensible Template-based Interface


An Introduction to XMLAn Introduction to XML


HyperText Markup LanguageHyperText Markup Language

• HTML is a language used to create hyperlink text in the WWW.

• The text is presented according to a set of predefined tags.

• The definition of tags is based on the Document Type Definition (DTD) of SGML.

• In other words, HTML is an application of SGML in the WWW.


• Central to SGML is the concept that documents have structurestructure, contentcontent, and formatformat.

• These three ingredients combine to form a document.

Standard Generalized Markup LanguageStandard Generalized Markup Language


What is Content?What is Content?

• Content is the actual data within a document.• The words and illustrations that make up a bicycle

assembly manual are its contents.


What is Format?What is Format?

• Format consists of how the words, sentences, and paragraphs are visually presentedvisually presented and distinguished from one another within a document.– Boldface for title, italics for special terms, and blank lines

between sections are examples of document formats.

• People often confuse format with structure.People often confuse format with structure.


What is Structure?What is Structure?

Coconut Pudding

12 ounces coconut milk

4 to 6 tablespoons sugar

4 to 6 tablespoons cornstarch

3/4 cup water

Pour coconut milk into saucepan.

Combine sugar and cornstarch; stir in waterand blend well.

Stir sugar mixture into coconut milk; cook and stir over low heat until thickened.

Recipe

Title

IngredientList

Ingredient

InstructionList

Step


Document Type DefinitionDocument Type Definition

• Defining the structures in XML/SGML– The structure of a document its type is defined by a

document type definition, or DTD.– The DTD lays out the rules for a document through the use

of elements, attributes, and entities.


Document Type DefinitionDocument Type Definition

<!ELEMENT recipe -- ( title, ingredientList, instructionList)><!ELEMENT title -- (#PCDATA)><!ELEMENT ingredientList -- (ingredient*)><!ELEMENT instructionList -- (step*)><!ELEMENT ingredient -- (#PCDATA) ><!ELEMENT step -- (#PCDATA)>

• A DTD looks like


Document InstanceDocument Instance

<!DOCTYPE RECIPE PUBLIC ”recipe" ”recipe"><RECIPE><TITLE>Coconut Pudding</TITLE><INGREDIENTLIST> <INGREDIENT> 12 ounces coconut milk</INGREDIENT> <INGREDIENT> 4 to 6 tablespoons sugar </INGREDIENT> <INGREDIENT> 4 to 6 tablespoons cornstarch </INGREDIENT> <INGREDIENT> 3/4 cup water </INGREDIENT><INGREDIENTLIST><INSTRUCTIONLIST>` <STEP> Pour coconut milk into saucepan. </STEP> <STEP>Combine sugar and cornstarch; stir in water and blend well. </STEP> <STEP>Stir sugar mixture into coconut milk; cook and stir over low heat until thickened. </STEP> …</INSTRUCTIONLIST></RECIPE>


HTML, SGML, XMLHTML, SGML, XML

• HTML helped establish the Internet by providing a universal way to present information.

• However, HTML only addresses the presentation of data.

• Using SGML, user can add structure along with the content of a document.

• However, SGML has proven too heavy-weight for the Internet.


Extensible Markup LanguageExtensible Markup Language

• The XML is a simple dialect of SGML.• HTML is sufficient for sending web pages that are

viewed by human beings.• XML, however, adds the tags that enable computers

to understand, act on or process the information.• XML has been designed for ease of implementation

and for interoperability with both SGML and HTML.


XML Application ProfileXML Application Profile

• Electronic commerce• Electronic data interchange (EDI)• Fine-grain content publishing• Internet search engines• Distributed application design• etc.


Data Type Requirements of DocumentsData Type Requirements of Documents

• HTML– One file per page– Simple uni-directional linking

• XML– Tens, hundreds or even thousands of objects per page– Multiple DTDs– Hierarchical structure and rich linking– Query and navigation capabilities required– Agents and business rules interact with the data


Data Types of Storage Data Types of Storage

• File system– Store monolithic stuff.– Folder system on top of them– Good at storing multimedia data


Data Types of StorageData Types of Storage

• Relational database– Tabular in nature– Good at storing rows and columns of data like

spreadsheets and data from forms like invoices.


Data Types of StorageData Types of Storage

• Object-oriented database– Good at managing structured, hierarchical rich

linked information.– That’s exactly what XML is.– XML is the object representation of data.


Design and Implementation of an Design and Implementation of an XML Object DatabaseXML Object Database


Basic IdeaBasic Idea

• The arrangement of elements in an XML document is governed by the element and attribute list declarations in document type definition.

• The creation of DTD in a sense is closely related to defining new data types and hierarchical relationship in an object database.

• Thus, to enter an XML document into an object database, at first a new schema corresponding to a DTD is generated in the object database, and then the document conforming to that DTD is fragmented into objects and entered into the database.



• Both the tasks of creating a schema in object database for a DTD and fragmenting XML documents into objects can be divided into two parts: analysis and generation. – For the former task, an input DTD is analyzed according to

the formation rules specified in the XML recommendation, and the schema definitions are produced for the structures found in the analysis of DTD.

– The other task is to analyze XML document instances and produce object definitions for the elements found in them.



• We employ the definite clause grammar (DCG) in Prolog as a tool to implement the analysis and generation tasks.

• The basic idea is to encode the analysis task in the context-free rule part and the generation task in the action part of the DCG rules.


Strucuture Document DatabaseStrucuture Document Database

• Combine structured document with OODB technology:– VERSO project at INRIA– News-On-Demand Application – Document Database from GMD-IPSI

• XML document database products:– The Poet XML Repository– eXcelon, ODI– Ardent Sofiware, Inc


DTDparser

Schemagenerator

DI parsergenerator

DI parserand objectgenerator

DTD

DI

Schema definitions

Object definitions

XML repository

DBlanguageprocessor

produces

Schemageneration

rules

XMLrules

Parsergeneration

rules

structure

Userinterface

System ArchitectureSystem Architecture


DTD ParserDTD Parser

elementdecl::= ’<!ELEMENT S Name S contentspec S? ‘>’

elementdecl(contentModel(N,C))--> elementPrefix, name(N), contentSpec(C), rightAngle.

contentspec::=‘EMPTY’| ‘ANY’| Mixed | children

contentSpec(C)-->empty,{C=’EMPTY’};any,{C={ANY’};mixed(C);children(C).


Parsing ResultParsing Result<!ELEMENT top (p,spec,div1)><!ELEMENT p (#PCDATA|a|ul|b|i|em)*><!ELEMENT spec (front,body, back?)*><!ELEMENT div1 (head,(p|list1 |note)*, div2*)><!ELEMENT name (#PCDATA)><!ELEMENT a (#PCDATA)><!ELEMENT ul (#PCDATA)><!ELEMENT b (#PCDATA)><!ELEMENT i (#PCDATA)><!ELEMENT em (#PCDATA)><!ELEMENT front (#PCDATA)><!ELEMENT body (#PCDATA)><!ELEMENT back (#PCDATA)><!ELEMENT head (#PCDATA)><!ELEMENT list1 (#PCDATA)><!ELEMENT note (#PCDATA)><!ELEMENT div2 (#PCDATA)>

[contentModel(top,seq([p/null,spec /null,div1/null])/null),

contentModel(p,mixed([pcdata,a,ul,b,i,em])),

contentModel(spec,seq([front/null,body/null,back/question])/star),

contentModel(div1,seq([head/null,alt([p/null,list1/null,note/null]) /star,div2/star])/null),

contentModel(name,pcdata), contentModel(a,pcdata), contentModel(ul,pcdata), contentModel(b,pcdata), contentModel(i,pcdata), contentModel(em,pcdata), contentModel(front,pcdata), contentModel(body,pcdata), contentModel(back,pcdata), contentModel(head,pcdata), contentModel(list1,pcdata), contentModel(note,pcdata), contentModel(div2,pcdata)]


Schema GenerationSchema Generation

defineClass 'Top' super: SingleSeq{ instance: 'P' 'p'; 'Spec' 'spec'; 'Div1' 'div1';};defineClass 'P' super: Mixed{ instance: List<Mixedp> mixedp;};defineClass Mixedp super: SingleAlt{ instance: String pcdata; 'A' 'a'; 'Ul' 'ul'; 'B' 'b'; 'I' 'i'; 'Em' 'em';};defineClass 'Spec' super: MultiSeq{ instance: List<Seqspec> seqspec;};

defineClass 'Seqspec' super: SingleSeq{ instance: 'Front' 'front'; 'Body' 'body'; 'Back' 'back';};defineClass 'Div1' super: SingleSeq{ instance: 'Head' 'head'; List<Alt1> 'alt1'; List<Div2> 'div2';};defineClass 'Alt1' super: SingleAlt{ instance: 'P' 'p'; 'List1' 'list1'; 'Note' 'note';};defineClass 'Name' super: Unstructured{ instance: String pcdata;}; ...


DI ParserDI Parsertop(V) --> stg(top), p(P),spec(Spec),div1(Div1), etg(top).p(V) --> stg(p), mixedp(Mixedp),etg(p).mixedp(V) --> (pcdata(Pcdata); a(A);ul(Ul);b(B); I(I); em(Em);{false}), mixedp(_); [].spec(V) --> stg(spec), spec1(Spec), etg(spec).spec1(V) --> front(Front), body(Body), (back(Back);[ ]), spec1(_); [].

div1(V) --> stg(div1), head(Head), alt1(Alt1), div21(Div21), etg(div1).alt1(V) --> (p(P); list1(List1); note(Note); {false}), alt1(_); [].div21(V) --> div2(Div2), div21(_) ; [].name(V) --> stg(name), pcdata(Pcdata), etg(name).


DI Parser GenerationDI Parser Generation

Rule_Head --> Start_Tag, Rule_Body, End_Tag, {Semantic Actions}.

for each contentModel(ElementName,ContentStructure) do generate the rule head for ElementName; generate the start tag for ElementName; generate the rule body for ContentStructure; generate the end tag for ElementName; generate the semantic action;


ImplementationImplementation

• We have built a prototype of the system using LPA Win-Prolog V3.5 on personal computer.

• It consists of a DTD parser, Schema generator and DI parser generator.

• After creating the physical store and class family for XML documents, we can proceed to build the database schema for DTD by executing the ODQL codes generated by the DTD schema generator.


An Extensible Query-By-Template Interface An Extensible Query-By-Template Interface to Accessing XML Document Databaseto Accessing XML Document Database


MotivationMotivation

• Vastness of search results on current WWW search engines

• Textual-based query language with a simple English-like syntax is inconvenient for the user.

• Current user interfaces primarily use form-based queries.


GoalGoal

• The goal is to design a convenient interface for user to access XML document without knowing the knowledge of the document types.

• The interface will relieve user from typing complex query language.

• The interface should be web-based and platform-independent.


System ArchitectureSystem Architecture

Visual Query Interface


Visual Query FacilityVisual Query Facility

• Query By Example (QBE)– The interface is composed of tabular skeletons representing

tables in the database.

• Query By Forms (QBF)– The interface is presented with a list of searchable fields,

each with an entry area that can be used to indicate the search string.

• Query By Template (QBT)– The interface is displayed a template for a representative

entry of the database. User express their queries by indicating the search keywords in the appropriate regions of the template.


Example of Image-based QBTExample of Image-based QBT


Limits of Image-based QBTLimits of Image-based QBT

• The image template is divided into regions, each of which corresponds to an element in the document structure.

• Associated with each regions is the query action. • Its significant drawback is the lack of flexibility in the

template creation.• It is difficult to automate the task of reconfiguration of

query action associate with the new template.• A single interface template for all types of document

is probably not a good idea.


Concept of eXtensible QBT (XQBT)Concept of eXtensible QBT (XQBT)

• The environment provides a template creator which consists of a DTD schema browser and a scene for presentation design.

• The environment aims at providing automatic configuration of query actions associated with presentation of template.

• The design of the template presentation must be tightly coupled with the arrangement of document data stored in the repository.

• The component in the design of presentation must be properly associated with corresponding nodes in the object database schema.


Environment for XQBTEnvironment for XQBT


Template CreatorTemplate Creator

• The template creator consists of a DTD schema browser a scene for template draft, and functional area.

• The template creator in mainly relied on a DTD schema browser, which corresponds to the database schema.

• The scene is a visual display area where the designer can organize a template draft for certain purpose.

• The content of template draft is exported to a file, which contains the template presentation and additional information.



Functional areaFunctional Area


Exported FileExported File

• The file contains the information about the template presentation property associate with each element.

• Each element is appended with the path information in the database schema, in order that the template executor, which can make use of the information to carry out query actions.


Template ExecutorTemplate Executor

• The template executor loads the exported file and presents the template as was originally designed in the template creator.

• The path of each node in the DTD schema browser is used to carry out the query action required by the user.


XQBTQBT

Comparison between Image-based QBT and XQBTComparison between Image-based QBT and XQBT

• The template is an image by taking a photograph or by scanning from existing pages.

• The query action associate with each region is hand-coded.

• Either planar or nested template is limited to region level that is not very deep.

• The template is generated for a representative document.

• The associated query action can be generated automatically for the interface program.

• The designer can change the template to meet the requirement of various region level.


ImplementationImplementation

• Java Proxies (Jp) for Jasmine– Jp allows developer to build their application in J-API, and ta

ke advantage of Jasmine class libraries.


The interface for our XML document databaseThe interface for our XML document database

ingredient

Ingredientname

Ingredientstep


Query FormulationQuery Formulation

– Such searches are performed by simply entering the search string in the corresponding region of the template.


Query Formulation (cont.)Query Formulation (cont.)


Query Formulation (cont.)Query Formulation (cont.)

• The multiple condition are specified in different regions which are combined using logical conjunctions(such as AND, OR, NOT).

• The approach used to derive the logical expression

from its graphical representation is using the default precedence.

• User can insert parentheses as necessary in the

condition box, which used in QBE interface.


The results of the query formulationThe results of the query formulation


Template ExecutorTemplate Executor


ConclusionConclusion

• We employ the DCG in Prolog to translate XML documents into the schema and object definitions of an object database.

• The features of backtracking in Prolog and the CFG formalism in the DCG are useful that we can construct the parsers easily by expressing the rules in the XML specification in DCG.

• We need not worry about providing information to choose which production rule to use as the recursive-descent parser does.

• Similarly, because of the features we can easily generate the DCG rules for the DI parser as the result of the parsing process.


Future WorkFuture Work

• Looking for possible applications of the XML database– Electronic commerce– Intelligent multiagent system– Knowledge management system

• Performance evaluation• Efficient query interfaces