2 Getting Started with Oracle XML DB

A. Data

Will your data be highly structured (mostly XML), semistructured, or mostly unstructured? If highly structured, will your tables be XML schema-based or non-schema-based? See "Oracle XML DB Application Design: A. How Structured Is Your Data?" and Chapter 3, "Using Oracle XML DB".

B. Access

How will other applications and users access your XML and other data? How secure must the access be? Do you need versioning? See "Oracle XML DB Application Design: B. Access Models".

C. Application Language

In which language(s) will you be programming your application? See "Oracle XML DB Application Design: C. Application Language".

D. Processing

Will you need to generate XML? See Chapter 16, "Generating XML Data from the Database".

How often will XML documents be accessed, updated, and manipulated? Will you need to update fragments or the whole document?

Will you need to transform the XML to HTML, WML, or other languages, and how will your application transform the XML? See Chapter 9, "Transforming and Validating XMLType Data".

Does your application need to be primarily database resident or work in both database and middle tier?

Is your application data-centric, document- and content-centric, or integrated (is both data- and document-centric). "Oracle XML DB Application Design: D. Processing Models".

Will you be exchanging XML data with other applications, across gateways? Will you need Advanced Queuing (AQ) or SOAP compliance? See Chapter 31, "Exchanging XML Data with Oracle Streams AQ".

E. Storage

How and where will you store the data, XML data, XML schema, and so on? See "Oracle XML DB Application Design: F. Storage Models".

Figure 2-1 Oracle XML DB Design Options

Description of "Figure 2-1 Oracle XML DB Design Options"

XML Schema-Based or Non-Schema-Based

Also consider the following data modeling questions:

• If your application is XML schema-based:

  • For structured data, you can use either Character Large Object (CLOB) or structured storage.

  • For semistructured data, you can use either CLOB, structured, or hybrid storage. Here your XML schema can be more loosely coupled. See also "Oracle XML DB Application Design: F. Storage Models".

  • For unstructured data, an XML schema design is not applicable.

• If your application is non-schema-based. For structured, semi/ pseudo-structured, and unstructured data, you can store your data in either CLOB values in XMLType tables or views or in files in Oracle XML DB Repository folders. With this design you have many access options including path- and query-based access through resource views.

Figure 2-2 Data Storage Models: How Structured Is Your Data?

Description of "Figure 2-2 Data Storage Models: How Structured Is Your Data?"

Messaging Options

Advanced Queuing (AQ) supports XML and XMLType applications. You can create queues with payloads that contain XMLType attributes. These can be used for transmitting and storing messages that contain XML documents. By defining Oracle Database objects with XMLType attributes, you can do the following:

• Store more than one type of XML document in the same queue. The documents are stored internally as CLOB values.

• Selectively dequeue messages with XMLType attributes using SQL functions such as existsNode and extract.

• Define rule-based subscribers that query message content using SQL functions such as existsNode and extract.

• Define transformations to convert Oracle Database objects to XMLType.

  
  

  See Also: Chapter 31, "Exchanging XML Data with Oracle Streams AQ" Oracle Streams Advanced Queuing User's Guide and Reference

  
  

Using XMLType Tables

If you are using XMLType tables you can store your data in:

• CLOB (unstructured) storage

• Structured storage

• Semistructured storage

Using XMLType Views

Use XMLType views if you have existing relational data. You can use the following options to define the XMLType views:

• SQL/XML SQL functions. Using these functions you can store the data in relational tables and also generate/regenerate the XML. See Chapter 16, "Generating XML Data from the Database".

• Object Types:

  • Object tables

  • Object constructors. You can store the data in relational tables using object constructors.

  • Object views

Figure 2-4 Structured Storage Options

Description of "Figure 2-4 Structured Storage Options"

XML Storage Requirements

Surveys show that data represented in XML and stored in a text file is three times the size of the same data in a Java object or in relational tables. There are two main reasons for this:

• Tag names (metadata describing the data) and white space (formatting characters) take up a significant amount of space in the document, particularly for highly structured, data-centric XML.

• All data in an XML file is represented in human readable (string) format.

Storing Structured Documents in Oracle XML DB Saves Space

The string representation of a numeric value needs about twice as many bytes as the native (binary) representation When XML documents are stored in Oracle XML DB using the structured storage option, the shredding process discards all tags and white space in the document.

The amount of space saved by this optimization depends on the ratio of tag names to data, and the number of collections in the document. For highly-structured, data-centric XML the savings can be significant. When a document is printed, or when node-based operations such as XPath evaluations take place, Oracle XML DB uses the information contained in the associated XML schema to dynamically reconstruct any necessary tag information.

XML Memory Management

Document Object Model (DOM) is the dominant programming model for XML documents. DOM APIs are easy to use but the DOM Tree that underpins them is expensive to generate, in terms of memory. A typical DOM implementation maintains approximately 80 to 120 bytes of system overhead for each node in the DOM tree. This means that for highly structured data, the DOM tree can require 10 to 20 times more memory than the document on which it is based.

A conventional DOM implementation requires the entire contents of an XML document to be loaded into the DOM tree before any operations can take place. If an application only needs to process a small percentage of the nodes in the document, this is extremely inefficient in terms of memory and processing overhead. The alternative SAX approach reduces the amount of memory required to process an XML document, but its disadvantage is that it only allows linear processing of nodes in the XML Document.

Oracle XML DB Reduces Memory Overhead for XML Schema-Based Documents by Using XML Objects (XOBs)

Oracle XML DB reduces memory overhead associated with DOM programming by managing XML schema-based XML documents using an internal in-memory structure called an XML Object (XOB). A XOB is much smaller than the equivalent DOM since it does not duplicate information like tag names and node types, that can easily be obtained from the associated XML schema. Oracle XML DB automatically uses a XOB whenever an application works with the contents of a schema-based XMLType. The use of the XOB is transparent to you. It is hidden behind the XMLType datatype and the C, PL/SQL, and Java APIs.

XOB Uses Lazily-Loaded Virtual DOM

The XOB can also reduce the amount of memory required to work with an XML document using the Lazily-Loaded Virtual DOM feature. This allows Oracle XML DB to defer loading in-memory representation of nodes that are part of sub-elements or collection until methods attempt to operate on a node in that object. Consequently, if an application only operates on a few nodes in a document, only those nodes and their immediate siblings are loaded into memory.The XOB can only used when an XML document is based on an XML schema. If the contents of the XML document are not based on an XML schema, a traditional DOM is used instead of the XOB.

XML Parsing Optimizations

To populate a DOM tree the application must parse the XML document. The process of creating a DOM tree from an XML file is very CPU- intensive. In a typical DOM-based application, where the XML documents are stored as text, every document has to be parsed and loaded into the DOM tree before the application can work with it. If the contents of the DOM tree are updated the whole tree has to be serialized back into a text format and written out to disk.

With Oracle XML DB No Re-Parsing is Needed

Oracle XML DB eliminates the need to keep re-parsing documents. Once an XML document has been stored using structured storage techniques no further parsing is required when the document is loaded from disk into memory. Oracle XML DB is able to map directly between the on disk format and in-memory format using information derived from the associated XML schema. When changes are made to the contents of a schema-based XMLType, Oracle XML DB is able to write just the updated data back to disk.Again, when the contents of the XMLType are not based on an XML schema a traditional DOM is used instead.

Node-Searching Optimizations

Most DOM implementations use string comparisons when searching for a particular node in the DOM tree. Even a simple search of a DOM tree can require hundreds or thousands of instruction cycles. Searching for a node in a XOB is much more efficient than searching for a node in a DOM. A XOB is based on a computed offset model, similar to a C/C++ object, and uses dynamic hashtables rather than string comparisons to perform node searches.

XML Schema Optimizations

Making use of the powerful features associated with XML schema in a conventional XML application can generate significant amounts of additional overhead. For example, before an XML document can be validated against an XML schema, the schema itself must be located, parsed, and validated.

Minimizing XML Schema Overhead After a Schema Is Registered

Oracle XML DB minimizes the overhead associated with using XML schema. When an XML schema is registered with the database it is loaded in the Oracle XML DB schema cache, along with all of the metadata required to map between the XML, XOB and on disk representations of the data. This means that once the XML schema has been registered with the database, no additional parsing or validation of the XML schema is required before it can be used. The schema cache is shared by all users of the database. Whenever an Oracle XML DB operation requires information contained in the XML schema it can access the required information directly from the cache.

Load Balancing Through Cached XML Schema

Some operations, such as performing a full schema validation, or serializing an XML document back into text form can still require significant memory and CPU resources. Oracle XML DB allows these operations to be off-loaded to the client or middle tier processor. Oracle Call Interface (OCI) interface and thick Java Database Connectivity (JDBC) driver both allow the XOB to be managed by the client.The cached representation of the XML schema can also be downloaded to the client. This allows operations such as XML printing, and XML schema validation to be performed using client or middle tier resources, rather than server resources.

Reduced Bottlenecks From Code That Is Not Native

Another bottleneck for XML-based Java applications happens when parsing an XML file. Even natively compiled or JIT compiled Java performs XML parsing operations twice as slowly compared to using native C language. One of the major performance bottlenecks in implementing XML applications is the cost of transforming data in an XML document between text, Java, and native server representations. The cost of performing these transformations is proportional to the size and complexity of the XML file and becomes severe even in moderately sized files.

Oracle XML DB Implements Java and PL/SQL APIs Over Native C

Oracle XML DB addresses these issues by implementing all of the Java and PL/SQL interfaces as very thin facades over a native 'C' implementation. This provides for language-neutral XML support (Java, C, PL/SQL, and SQL all use the same underlying implementation), as well as the higher performance XML parsing and DOM processing.

Reduced Java Type Conversion Bottlenecks

One of the biggest bottlenecks when using Java and XML is with type conversions. Internally Java uses UCS-2 to represent character data. Most XML files and databases do not contain UCS-2 encoded data. This means that all data contained in an XML file has to be converted from 8-Bit or UTF-8 encoding to UCS-2 encoding before it can be manipulated in a Java program.

Oracle XML DB Uses Lazy Type Conversion to Avoid Unneeded Type Conversions

Oracle XML DB addresses these problems with lazy type conversions. With lazy type conversions the content of a node is not converted into the format required by Java until the application attempts to access the contents of the node. Data remains in the internal representation till the last moment. Avoiding unnecessary type conversions can result in significant performance improvements when an application only needs to access a few nodes in an XML document.

Consider a JSP that loads a name from the Oracle Database and prints it out in the generated HTML output. Typical JSP implementations read the name from the database (that probably contains data in the ASCII or ISO8859 character sets) convert the data to UCS-2, and return it to Java as a string. The JSP would not look at the string content, but only print it out after printing the enclosing HTML, probably converting back to the same ASCII or ISO8859 for the client browser. Oracle XML DB provides a write interface on XMLType so that any element can write itself directly to a stream (such as a ServletOutputStream) without conversion through Java character sets. Figure 2-5 shows the Oracle XML DB Application Program Interface (API) stack.

Figure 2-5 Oracle XML DB Application Program Interface (API) Stack

Description of "Figure 2-5 Oracle XML DB Application Program Interface (API) Stack"