Updates over virtual XML views that wrap relational data are not well supported by XML data management systems. This paper studies the problem of whether a correct relational update translation can be found for a given update over an XML view. For this, we propose a lightweight update checking framework named U-Filter. It first performs two steps of schema-level (and thus very inexpensive) checks, which is based on the result of an update translatability analysis for the view query. Only when necessary, a third more expensive checking step, requiring base data access, is employed. For the latter, we design both the DBMS internal as well as external strategies. This three-step check process filters out XML updates that cannot be translated. Others are fed to the update translation engine, which generates the corresponding SQL update statements. We report on several experiments that illustrate the usefulness of U-Filter and the performance inpact achievable by the proposed algorithm.
Ling Wang, Elke A. Rundensteiner, Murali Mani