We propose an optimization method for queries over XML views, especially document views. We focus on XML document views, in particular, constructed by an aggregation peculiar to structured documents. The prominent feature of the aggregation leads to developing a novel optimization technique for the query.The point of our optimization is that a traditional optimization method of database queries, called “pushing selections”, is applied into a predicate peculiar to text retrieval. When pushing the predicate is applied to optimizing a query over XML document views constructed by the aggregation, the feature of the aggregation requires rewriting not only the predicate but also the whole query. We will show how input queries over the XML views are rewritten.