Projector – a Partially Typed Language for Querying XML

We describe Projector, a language that can be used to perform a mixture of typed and untyped computation against data represented in XML. For some problems, notably when the data is unstructured or semistructured, the most desirable programming model is against the tree structure underlying the document. When this tree structure has been used to model regular data structures, then these regular structures themselves are a more desirable programming model. The language Projector, described here in outline, gives both models within a single partially typed algebra and is well suited for hybrid applications, for example when fragments of a known structure are embedded in a document whose overall structure is unknown. Projector is an extension of ECMA-262 (aka JavaScript), and therefore inherits an untyped DOM interface. To this has been added some static typing and a dynamic projection primitive, which can be used to assert the presence of a regular structure modelled within the XML. If this structure does exist, the data is extracted and presented as a typed value within the programming language.