An Optimized Two-Step Solution for Updating XML Views

Abstract

View updating is a long standing difficult problem. Given a view defined over base data sources and a view update, there are several different updates over the base data sources, called translations, that perform the update. A translation is said to be correct if it performs the update and at the same time does not update any portion of the view not specified in the update (no view side-effects). The view update problem finds a correct translation for a given view update if one exists. In the relational scenario, previous research has attempted to study the view update problem either by utilizing only the schema knowledge, or by directly examining the base data. While utilizing only the schema knowledge is very efficient, we are not guaranteed to find a correct translation even if one exists. On the other hand, examining the base data is guaranteed to find a correct translation if one exists, but is very time-consuming. The view update problem is even more complex in the XML context due to the nested hierarchical structure of XML and the restructuring capabilities of the XQUERY view specification. In this paper we propose a schema-centric framework, named HUX, for efficiently updating XML views specified over relational databases. HUX is complete (always finds a correct translation if one exists) and is efficient. The efficiency of HUX is achieved as follows. Given a view update, HUX first exploits the schema to determine whether there will never be a correct translation, or there will always be a correct translation. Only if the update cannot be classified using the schema, HUX will examine the base data to determine if there is a correct translation. This data-level checking is further optimized in HUX, by exploiting the schema knowledge extracted in the first step to significantly prune the space of potential translations that is explored. Experiments illustrate the performance benefits of HUX over previous solutions.