Updating XML views
Abstract (Summary)
Update operations over XML views are essential for applications using
XML views. In this dissertation work, we provide scalable solutions to
support updating through XML views defined over relational databases.
Especially we focus on the update-public semantic, where updates are always
public (made to the public database), and the update-local semantic,
where update effects are first kept local and then made public as and when
required.
Towards this, we propose the clean extended-source theory for determining
whether a correct view update translation exists, which then serves
as a theoretical foundation for us to design practical XML view updating
algorithms.
Under update-public semantic, state-of-the-art view updating work focus
on identifying the correct update translation purely on the data. We
instead take a schema-centric solution, which utilizes the schema of the
underlying source to effectively prune updates that are guaranteed to be
not translatable and pass updates that are guaranteed to be translatable directly
to the SQL engine. Only those updates that could not be classified
ii
using schema knowledge are finally analyzed by examining the data. This
required data-level check is further optimized under schema guidance to
prune the search space for finding a correct translation.
As the first work addressing the update-local semantic, we propose a
practical framework, called LoGo. LoGo Localizes the view update translation,
while preserves the properties of views being side-effect free and
updates being always updatable. LoGo also supports on-demand merging
of the local database of the subject view into the public database (also called
global database), while still guaranteeing the subject view being free of side
effects. A flexible synchronization service is provided in LoGo that enables
all other views defined over the same public database to be refreshed, i.e.,
synchronized with the publically committed changes, if so desired.
Further, given that XML is an ordered data model, we propose an ordersensitive
solution named O-HUX to support XML view updating with order.
We have implemented the algorithms, along with respective optimization
techniques. Experimental results confirm the effectiveness of the proposed
services, and highlight its performance characteristics.
iii
Bibliographical Information:
Advisor:
School:Worcester Polytechnic Institute
School Location:USA - Massachusetts
Source Type:Master's Thesis
Keywords:xml document markup language querying computer science
ISBN:
Date of Publication: