Introduction

XML has emerged as the de facto standard for data integration; it has become an integral part of diverse application domains such as databases, messaging systems, Web Services, etc. Despite the popularity of XML, current mechanisms for developing XML-based applications in Java are low level. For example, APIs such as W3C Document Object Module (DOM) [World Wide Web Consortium 2000] or Simple API for XML (SAX) [SAX ], provide minimal support for ensuring that programs are correct with respect to the XML Schemas [Thompson et al. 2004] governing the XML data. With data-binding approaches such as Java Architecture for XML Binding (JAXB) [Vajjhala and Fialli 2004], a programmer must understand how the specification maps XML Schemas to Java™ classes and cannot program purely in terms of the XML data. Furthermore, there is no support for ensuring the correctness of XPath expressions [Clark and DeRose 1999] evaluated against documents. In these systems, a mistyped XPath expression (where one of the element names is misspelled) would not raise an error at run time -- it would silently return no results. Finally, runtime library approaches to XML, such as DOM, cannot take advantage of compiler techniques in order to optimize the performance of XML processing. The high cost of processing XML is a common complaint.

XML Enhancements to Java™, or XJ, is a research language designed to facilitate the development of applications that process XML data. It is a strict extension to Java 1.4 -- programs that are valid according to the Java Language Specification 1.4 [Gosling et al. 2000] are valid XJ programs; they have identical behavior in both languages. XJ adds a few constructs to Java to facilitate XML processing:

• The semantics of the import statement is extended to allow for the import of XML Schemas, in addition to packages and types. Element declarations in an imported XML Schema are available to a programmer as classes.
• Programmers may write XPath expressions inline to navigate and extract information from XML data. The XJ compiler type checks XPath expressions with respect to appropriate XML Schema types to detect errors.
• Programmers may construct XML data by writing literal XML inline. Again, the compiler ensures that the constructed XML is valid with respect to appropriate XML types.
• The result of compiling an XJ class is standard bytecode that can be executed on any Java Virtual Machine.

In subsequent sections, we describe how to install the XJ system, compile and execute programs (§2). We provide an overview of the constructs of the XJ programming language (§3), and discuss the limitations of the current system (§4). XJ is an evolving language and we will be adding features that address some of the current limitations (§5). We welcome any suggestions for improvement both in terms of syntax as well as language semantics.

Document Conventions

All examples in this manual refer to the Totals.xj sample which is provided with the XJ distribution. The XML Schema used by the sample, the Totals.xj XJ class, and input data are provided for reference in Appendix A.

The "sample code" typesetting is used to highlight sample XJ code, and the "schema construct" typesetting is used to highlight XML Schema constructs.

Online Information

To get the latest XJ information (including the latest version of this manual) on-line, access the IBM external XJ project page at http://www.research.ibm.com/xj/.

Reporting Bugs

Please either use the XJ forum on alphaWorks (http://alphaworks.ibm.com/tech/xj) or send mail to xj@watson.ibm.com.