XML stream transformer generation through program composition and dependency analysis

XML stream transformation, which sequentially processes the input XML data on the fly, makes it possible to process large sized data within a limited amount of memory. Though being efficient in memory-use, stream transformation requires stateful programming, which is error-prone and hard to manage. This paper proposes a scheme for generating XML stream transformers. Given an attribute grammar definition of transformation over an XML tree structure, we systematically derive a stream transformer in two steps. First, an attribute grammar definition of the XML stream transformation is inferred by applying a program composition method. Second, a finite state transition machine is constructed through a dependency analysis. Due to the closure property of the program composition method, our scheme also allows modular construction of XML stream transformers. We have implemented a prototype XML stream transformer generator, called altSAX. The experimental results show that the generated transformers are efficient in memory consumption as well as in execution time.