Extracting typed values from XML data

Values of existing typed programming languages are increas- ingly generated and manipulated outside the language juris- diction. Instead, they often occur as fragments of XML docu- ments, where they are uniformly interpreted as labelled trees in spite of their domain-specific semantics. In particular, the values are divorced from the high-level type with which they are conveniently, safely, and efficiently manipulated within the language. We propose language-specific mechanisms which extract language values from arbitrary XML documents and inject them in the language. In particular, we provide a general framework for the formal interpretation of extraction mecha- nisms and then instantiate it to the definition of a mechanism for a sample language core L. We prove that such mechanism can be built by giving a sound and complete algorithm that implements it. The values, types, and type semantics of L are sufficiently general to show that extraction mechanisms can be defined for many existing typed languages, including object-oriented languages. In fact, extraction mechanisms for a large class of existing languages can be directly derived from L's. As a proof of this, we introduce the SNAQue prototype system, which transforms XML fragments into CORBA objects and exposes them across the ORB framework to any CORBA-compliant language.