21 Accessing Oracle XML DB Repository Data

Names of Generated Tables

The names of the generated tables are assigned by Oracle XML DB, and can be obtained by finding the xdb:defaultTable attribute in your XML schema document (or in the default XML schema document). When you register an XML schema, you can alternatively provide your own table name, instead of using the default name supplied by Oracle XML DB. If the table specifies binary XML storage, then a document is encoded in binary XML format before storing it in the table.

Defining Structured Storage for Resources

Applications that need to define structured storage for resources can do so in one of these ways:

• Subclass the Oracle XML DB resource type. Subclassing Oracle XML DB resources requires privileges on the table XDB$RESOURCE.

• Store data that conforms to a visible, registered XML schema.

  See Also:

  Chapter 6, "XML Schema Storage and Query: Basic"

ASM Virtual Folder

The ASM virtual folder, /sys/asm, is an exception to the description of the previous sections: its contents are ASM files and folders that are managed automatically by Oracle Automatic Storage Management (ASM).

Repository and Document Links

In addition to containing resources, a folder resource can contain links to other resources (files or folders). These repository links, sometimes called folder links, are not to be confused with document links, which correspond to the links provided by the XLink and XInclude standards, and which are also supported by Oracle XML DB. Repository links are navigational, folder–child links among repository resources; document links are arbitrary links among documents that are not necessarily repository resources.

Repository links represent repository hierarchical relationships; document links represent arbitrary relationships whose semantics derive from the applications that use them. Because they represent repository hierarchical relationships, repository links can be navigated using file system-related protocols; this is not true of document links. Because document links can represent arbitrary relationships, they can also represent repository relationships. When document links thus target resources, they can also be hard or weak.

Hard Links and Weak Links

Links that target repository resources can be hard links or weak links. Both hard and weak links are references, or pointers, to physical data — (internal) repository resource identifiers; they do not point to symbolic names or paths to other links. Their targets are resolved at the time of link creation. Because they point directly to resource identifiers, hard and weak links cannot dangle: they remain valid even when their targets are renamed or moved. You need the same privileges to create or delete hard and weak links.

The difference between hard and weak links lies in their relationship to target resource deletion. A target resource is dependent on its hard links, in the sense that it cannot be deleted as long as it remains the target of a hard link. Deletion of a hard link also deletes the resource targeted by the link, if the following are both true:

• The resource is not versioned.

• The hard link that was deleted was the last (that is, the only) hard link to the resource.

A weak link has no such hold on a resource: you can delete a resource, even if it is the target of a weak link (as long as it is not the target of a hard link). Because of this, weak links can be used as shortcuts to frequently accessed resources, without impacting deletion of those resources.

There is a dependency in the other direction, however: If you delete a resource that is the target of one or more weak links, then those links are automatically deleted, as well. In this sense, too, weak links cannot dangle. Both hard and weak links provide referential integrity: if a link exists, then so does its target.

Another difference between hard and weak links is this: Hard links to ancestor folders are not permitted, because they introduce cycles. There is no such restriction for weak links: a weak link can target any folder, possibly creating a cycle. It is the set of hard links that define the (acyclic) structure of Oracle XML DB Repository. Weak links represent an additional mapping on top of that basic structure.

You can query the repository path view, PATH_VIEW, to determine the type of a repository link: the link information contains the link type. XMLType column LINK of PATH_VIEW contains this information in element LinkType, which is a child of the root element, LINK. Example 21-1 illustrates this. You can also determine the type of a repository link by using the getLink() callback function on an event handler (LinkIn, LinkTo, UnlinkIn, or UnlinkFrom).

SELECT RESID, extractValue(LINK, '/LINK/LinkType') link_type
  FROM PATH_VIEW WHERE equals_path(RES, '/home/QUINE/purchaseOrder.xml') = 1;

RESID                              LINK_TYPE
--------------------------------   ---------
DF9856CF2FE0829EE030578CCE0639C5   Weak

Creating a Weak Link Without Knowledge of Folder Hierarchy

Suppose that you want to read a file resource that belongs to one of your colleagues. You cannot create a hard link to that resource, to make it accessible for your use, unless you have the privilege <xdb:resolve> on all of the ancestor folders of that file. Having that privilege would mean that you could see all of your colleague's folder names and the structure of the hierarchy down to the target resource.

However, because weak links represent, essentially, a mapping on top of the real repository structure, which structure is determined by the set of hard links, you can create a weak link to a resource using just its OID, rather than its full, named path (URL). This means that your colleague can determine the OID path to the file, send you that instead of the named path, and you can create a weak link to the document using that OID path. Example 21-2 and Example 21-3 illustrate this.

DECLARE
  resoid  RAW(16);
  oidpath VARCHAR2(100);
BEGIN
  SELECT RESID INTO :resoid FROM RESOURCE_VIEW
    WHERE equals_path(RES, '/home/QUINE/purchaseOrder.xml') = 1;
    oidpath := DBMS_XDB.createOIDPath(resoid)
  DBMS_OUTPUT.put_line(oidpath);
END;

CALL DBMS_XDB.Link(/sys/oid/1BDCB46477B59C20E040578CCE0623D3
                   '/home/CURRY', 'quinePurchaseOrder.xml',
                   DBMS_XDB.LINK_TYPE_WEAK);

Restricting Multiple Hard Links

Sometimes, it is useful to restrict the creation of hard links, disallowing multiple hard links to folders or files (or both). In particular, allowing multiple hard links to file resources, but disallowing multiple hard links to folder resources, provides behavior that is similar to that for some file systems, including Unix and Linux. This can simplify application design, by, in effect, ensuring that each file resource has a unique, canonical hard-link path to it. In addition, preventing multiple hard links to a resource can lead to query performance improvements.

You can configure the prevention of multiple hard links using the following Boolean parameters in configuration file xdbconfig.xml. The default value of each parameter is true, meaning that multiple hard links can be created.

• folder-hard-links – Prevent the creation of multiple hard links to a folder resource, if false.

• non-folder-hard-links – Prevent the creation of multiple hard links to a file resource, if false.

Where You Can Use Oracle XML DB Protocol Access

Oracle Net Services is optimized for record-oriented data. Internet protocols are designed for stream-oriented data, such as binary files or XML text documents. Oracle XML DB protocol access is a valuable alternative to Net Services in the following scenarios:

• Direct database access from file-oriented applications using the database like a file system

• Heterogeneous application server environments that require a uniform data access method (such as XML over HTTP, which is supported by most data servers, including MS SQL Server, Exchange, Notes, many XML databases, stock quote services and news feeds)

• Application server environments that require data in the form of XML text

• Web applications that use client-side XSL to format datagrams that do not need much application processing

• Web applications that use Java servlets that run inside the database

• Web access to XML-oriented stored procedures

Using Protocol Access

Accessing Oracle XML DB using a protocol proceeds as follows:

1. A connection object is established, and the protocol might read part of the request.

2. The protocol decides whether the user is already authenticated and wants to reuse an existing session or the connection must be re-authenticated (the latter is more common).

3. An existing session is pulled from the session pool, or else a new one is created.

4. If authentication has not been provided, and the request is HTTP get or head, then the session is run as the ANONYMOUS user. If the session has already been authenticated as the ANONYMOUS user, then there is no cost to reuse the existing session. If authentication has been provided, then the database re-authentication routines are used to authenticate the connection.

5. The request is parsed.

6. (HTTP only) If the requested path name maps to a servlet, then the servlet is invoked using Java Virtual Machine (VM). The servlet code writes the response to a response stream or asks XMLType instances to do so.

Retrieving Oracle XML DB Resources

When the protocol indicates that a resource is to be retrieved, the path name to the resource is resolved. Resources being fetched are always streamed out as XML, with the exception of resources containing the XDBBinary element, an element defined to be the XML binary data type, which have their contents streamed out in RAW form.

Storing Oracle XML DB Resources

When the protocol indicates that a resource must be stored, Oracle XML DB checks the document file name extension for .xml, .xsl, .xsd, and so on. If the document is XML, then a pre-parse step is done, whereby enough of the resource is read to determine the XML schemaLocation and namespace of the root element in the document. If a registered schema is located at the schemaLocation URL, and it has a definition for the root element of the current document, then the default table specified for that root element is used to store the contents of the resource.

Using Internet Protocols and XMLType: XMLType Direct Stream Write

Oracle XML DB supports Internet protocols at the XMLType level by using the writeToStream() Java method on XMLType. This method is natively implemented, and writes XMLType data directly to the protocol request stream. This avoids Java VM execution costs and the overhead of converting database data through Java data types and creating Java objects, resulting in significantly higher performance. Performance is further enhanced if the Java code deals only with XML element trees that are close to the root, and does not traverse too many of the leaf elements, so that relatively few Java objects are created.