XML data binding refers to a means of representing information in an XML document as a business object in computer memory. This allows applications to access the data in the XML from the object, rather than using the DOM or SAX to retrieve the data from a direct representation of the XML itself.

It makes it possible to read and write XML data using a programming language class library (e.g. C++, C#, Java), specifically created for a given XML data format.^[1] Whilst it is possible to manually write a computer program to achieve this, XML data binding tools generate the source code to perform these tasks.

An XML data binder accomplishes this by automatically creating a mapping between elements of the XML schema of the document we wish to bind and members of a class to be represented in memory.

When this process is applied to convert an XML document to an object, it is called unmarshalling (also called deserialization). The reverse process, to serialize an object as XML, is called marshalling.

Approaches to data binding can be distinguished as follows:

• XML schema based: Based on an existing XML schema, classes that correspond to the schema are generated.
• Class based: Based on a set of classes to be serialized, a corresponding XML schema is generated.
• Mapping-based: A mapping description, usually itself an XML document, describes how an existing XML schema maps to a set of classes, and vice versa.

Since XML is a document-oriented format and objects are (usually) not document-oriented, simple XML data binding mappings may ignore some of the structural information embedded in an XML document. Specifically, information such as comments, XML entity references, and sibling order may not be preserved in the object representation created by the binding application. However, this is not always the case; sufficiently powerful XML data binding tools are capable of preserving 100% of the information stored in an XML document.

Similarly, since objects residing in computer memory are not inherently sequentially stored, and may include links to other objects (including self-referential links), simple XML data binding mappings may not be capable of preserving all the information about an object when it is marshalled to XML. However, sufficiently powerful data binding tools perform graph structure analysis on objects residing in memory to marshall (cyclic) object graph structures in XML by utilizing standard XML reference attributes.

An alternative approach to automatic data binding relies instead on hand-crafted XPath expressions that extract data from XML. This approach has some benefits but also has some drawbacks. First, the approach only needs proximate knowledge (e.g., topology, tag names, etc.) of the XML tree structure, which developers can determine by looking at the XML data. Furthermore, XPath allows the application to bind the relevant data items and filter out everything else, avoiding the unnecessary processing that would be required to completely unmarshall the entire XML document. The drawback of this approach is the lack of automation in implementing the object model and XPath expressions. Instead, the application developers have to create these artifacts manually, which is time-consuming, potentially error-prone, and hampers application maintenance when XML schemas and XML content models are updated. Another drawback is the lack of XML schema verification, which XML data bindings typically apply automatically during unmarshalling. Schema validity is typically required in secure applications.

One of XML data binding's strengths is the ability to deserialize objects across programs, languages, and platforms.^[2] You can dump a time series of structured objects from a datalogger written in C (programming language) on an embedded processor, bring it across the network to process in Perl and finally visualize in Octave. The structure and the data remain consistent and coherent throughout the journey, and no custom formats or parsing is required. This is not unique to XML. YAML, for example, is emerging as a powerful data-binding alternative to XML. JSON (which can be regarded as a subset of YAML) is often suitable for lightweight or restricted applications.

• Bound control
• Data structure
• JSON
• LDX+
• Serialization
• YAML

• XML Data Binding Resources, by Ronald Bourret
• XML Schema Patterns for Databinding Working Group