Maintaining materialized xpath views over dynamic xml data

Materializing views (query results) is a longtime tradition in data management systems. Materialized views are typically used to answer new queries more efficiently than answering them using the base data. When the base data is updated, the materialized views may become inconsistent with it. Maintaining such views is the process of refreshing them according to the base updates. Efficiency and scalability requirements typically mandate that this process be done incrementally rather than by full re-computation. This dissertation addresses the problem of incrementally maintaining views specified using XPath queries over dynamic XML base data. We identify two main scenarios of the problem, and provide solutions for them. The two scenarios are distinguished based on the degree of coupling between the base data system and the view system. If the systems are tightly coupled, then they can exchange non-standard forms of information—e.g. intermediate query results and internal node identifiers—which is specifically customized to support the maintenance process. On the other hand, if the systems are loosely coupled, then they can only exchange standard forms of information, e.g. standard XPath queries and their standard XML results. The tight-coupling scenario typically occurs when the base data system itself materializes some views—transparent to its clients—to improve the query performance. In this case, the base system and the view system are actually the same. The loose-coupling scenario is typical in the context of SOA (Service Oriented Architecture) systems, e.g. data-centric web services. Our solution for this scenario comprises two main steps: (1) The base system publishes standard XML updates, (2) Each client (view system) uses the published updates to refresh its views. This solution can be viewed as an XML syndication scheme which generalizes the idea of modern web syndication schemes, such as RSS (Really Simple Syndication). Many Data-centric web services use efficient back-end relational databases, while interfacing the outside world through standard XML front-ends. In this case, the publishing step of the syndication scheme requires publishing standard XML updates from relational SQL updates. We discuss a solution for this problem using triggers defined on the back-end relational database.