Tamino a DBMS Designed for XML Dr Harald

Tamino – a DBMS Designed for XML Dr. Harald Schoning Presenter: Wenhui Li University of Ottawa Instructed by: Dr. Mengchi Liu Carleton University

Abstract n n n Who? - Software AG What? - XML database management system When? n n n 1999 the first time unveiled 2004 June Tamino XML Server 4. 2 Why? n n management and transfer of structured and unstructured data completely designed for XML

Industry Background n n XML is becoming prevailing for data processing in the internet. Early goals of Tamino n n Easy data exchanging Evolution trend n n Storing, managing, publishing and exchanging XML documents Business modeling

Industry Background cont’ XML support in databases n n n Oracle XML Developer’s Kit SQL Server 2000 DB 2 XML Extender

Limitations of XML support via traditional RDBMS or ORDB n n XML is not well-structured like RDB, ORDB or OODB Storing and querying XML is possible but not feasible in these DB systems

Two Modeling approaches n Data-centric documents n n Regular structure Order does not matter No mixed content Document-centric documents n n n less regular structure significance of the order mixed content

Why don’t use relational DB n n XML documents can have schematic information (DTD), but they are not required to. classical database handling objects of a predefined type, cannot be applied in XML

Why doesn’t use XML itself? n n XML is just a markup language, it does not contain processing facilities on its own querying a set of XML documents is outside the scope of the XML recommendation Therefore, comes the Tamino!

What does Tamino do? n n n What’s Tamino (the 1 st slide) Store XML documents, HTML files and GIF images, etc. Retrieve them in a set-oriented manner, with sophisticated query facilities

Tamino’s architecture

The schema of XML documents n n XML support schematic information, but it differs from the classical databases DTD have a couple of deficiencies (e. g. data type) W 3 C working group is developing an XML schema description language However, DTD is the only standard schema at present

XML schema vs. RDB and OODB schema n n In RDB or OODB, the schema is created before the instances can be stored Instances must conform to the declared schema In XML database, each instance declares a schema on its own. for XML documents, grouping of objects of homogeneous structure into (pre-defined) tables or classes doesn’t work

Query and Index of XML schema n n Queries operate on sets Indexes are defined on the basis of a common schema For the purpose of querying, arbitrary objects could be grouped to sets Index definition also requires at least a common subset in the structure

Schema handling in Tamino n n n Grouping documents by open content model + user-directed document grouping Documents grouped into collections Within a collection, declare several document types For each document type define a common schema (open content model) For each document, Tamino assigns one of the document type

Type Assignment n n Assignment is based on the root element type Document must match the schema of the document type assigned, but might have additional elements/attributes In a document type, documents might differ considerably If no appropriate document type, document is stored without any schema checking

Tamino schema example

Document accepted by Tamino <City Inhabitants=” 138000”> <Name>Darmstart</Name> <Addition>The city of art nouveaud</Addtion> <Monument Height=” 39 m”> <Name>Langer Ludwig</Name> <Location> <Name>Luisenplatz<Name> <Map. Index>M 5</Map. Index> </Location> </Monument> </City>

Is an element/attribute should be modeled? n n n an index will be defined on this element/attribute the element/attribute is to be mapped to an external data source or to a server extension dedicated access rights will be defined on the element/attribute the presence / multiplicity of the element is to be enforced one of the above conditions hold for a child of the element

Indexing of Tamino n value-based indexes n n well known from traditional database systems used to accelerate the search exactly address the data object names need not be unique within a DTD

Example of value-based index n value-based indexes data-centric view <!ELEMENT City (Name, Inhabitants, Monument+)> <!ELEMENT Monument (Name, Description)> <!ELEMENT Inhabitants (#PCDATA)> <!ELEMENT Name (#PCDATA)> <!ELEMENT Description (#PCDATA)> n

Indexing of Tamino (cont’) n text indexing n n n document-centric view limit the scope to a specific part of the document the scope might span element content

Example of text index n text indexing document-centric view <statement> <author> <firstname>Harald></firstname> <lastname>Schoning</lastname> </author> <text> X<italic>M</italic>L and X<italic>S</italic>L are <stressed>very</stressed> important </text> </speech> n

Indexing of Tamino (cont’) n structural index n n n If multiplicity permits the omission of elements or if no DTD is known Example n n in a database of all European cities search all those cities which have an element called “beach”

Querying XML documents n n n Currently, there is no standardized query language XPath allows positioning within a single document XPath fits well the needs of retrieval in datacentric environments document-centric environments need a more content-based retrieval facility Tamino also supports full text search

Expectation for XML processor n n n W 3 C: XML recommendation specifies the handling of entities, comments and processing instructions. User: Tamino, leave comments intact, no processing instruction evaluated, leave entity references unresolved. User: the output of a Tamino query should match the specification of an XML processor.

Why don’t leave entities unresolved? n n In case result is a set of (parts of) matching documents This result DTD must include all different entity declarations of the original document Definition of the entity might differ from document to document So, for the same entity name, entities are renamed, and the entity references are changed accordingly.

problems of external entities n n n These entities can change without the database system knowing about this Thus, the values of external entities must not be included in indexes Example: <!ENTITY &mysubject SYSTEM “http: //www. softwareag. com/hottopic. xml”>. . . <ticker>Todays hot topic: &mysubject</ticker> n Checking the current contents of the external entity lead to unacceptable response times.

Relational Databases and XML n n major (object-) relation database systems include some forms of XML support The simplest form is to generate XML documents for existing relational data. But, real database handling of XML requires that XML data can be stored and retrieved Two approaches

XML support approach(1) n n n Map the XML document is to relational tables and their columns Markup is ignored on storage, and reconstructed on retrieval advantage of this approach: n the contents of an XML document can be handled with traditional SQL

XML support approach(1) cont’ n Shortcomings: n The sequence information lost <Order Customer. Id=” 567” Date=” 12 -2000”> <Item Product. ID=” 17” Quantity=” 2”/> <Item Product. ID=”l 6” Quantity=” 9”/> <Item Product. ID=“ 19 ” Quantity=“ 8”/> </Order> The retrieval of the order: <Order Customer. Id=” 567” Date=” 12 -12 -2000”> <Item Product. ID=” 16” Quantity=” 9’/> <Item Product. ID=” 17” Quantity=” 2”/> <Item Product. ID=” 19” Quantity=” 8”/> </Order>

XML support approach(1) cont’ n n n Data-centric documents sequence might not matter, it does for document-centric this approach loses all comments and processing instructions mixed content cannot be stored easily in this model

XML support approach(2) n n Leaves the XML document intact and stores it in a large text field (“BLOB”) Or even outside the database Text search is possible Can limit a certain text-based condition

XML support approach(2) cont’ n Limitations: n n no structure-aware combinations are possible Value-based search is not supported on these text fields n n n IBM solution: side tables But, direct manipulation of side tables destroys the consistency of the database Security can be defined on document level only, but not on elements or attributes

Summary n n n Tamino was designed with particular attention to the XML Schema handling for XML is different from relational databases does In Schema handling, external entities cause conceptual problems value-based indexes are useful for XML, as well as text index and structural index Comments and processing instructions should be preserved when documents are stored The result of a query against an XML database should be XML

Q&A Thanks!