The XR data model allows capturing rich semistructured corpora of documents (XML) with annotations (RDF triples). A simple query language for XR was presented in previous work, which allowed to retrieve results from from XR databases but not to build new XR content. This work makes two contributions. First, we present an extension of the query language giving powerful content structuring and grouping capabilities which enables flexible generation of new XR results. Second, we provide a query-view composition algorithm which enables the efficient evaluation of XR queries over previously defined XR views.
