Categories and Subject Descriptors H.2 [Database Management]: Miscellaneous General Terms Algorithms

The problem of updating XML views is more complex than in the relational scenario due to its nested hierarchical structure. While several research projects [1, 4] began to explore this XML view updating problem, they typically provide no guarantee for avoiding view side eects. An update translatability analysis can be employed to reason about the potential view side eects before performing the update. Intuitively, such analysis could examine the ac-tual base data [2, 3]. However, as indicated by [2], this data-centric search for a translation tends to be very ex-pensive, even for relational view updates. Instead, in our HUX project, we have designed a comprehensive solution of exploring schema-knowledge to optimize this analysis. Fig. 1(a) shows a running example of a relational database for a course registration system. A virtual XML view in Fig. 1(c) is dened by the view query in Fig. 1(b). The following examples illustrate cases of classifying updates as translatable or not translatable. Here we only use a delete primitive with the format (delete nodeID), where nodeID is the abbreviated identier of the element to be deleted. For example, CI1.PS1 represents the rst Professor-Student element of the rst ClassInfo element. Example 1. Update u1 =fdelete CI1.PS1.S2g over the XML view (Fig. 1) deletes the student `Mike Fisher". We can delete Student.t2 to achieve this without causing any view side eect. This can be concluded by looking at the schema of the view. From the view query (Fig. 1(b)), each student can only appear once in the view, namely, in the ClassInfo element that represents its course-professor-student relationship. Deleting any student element in the view can Copyright is held by the author/owner(s)