url - Universität zu Lübeck

Chapter 7
The XML Index Update Problem
In contrast to relational DBMS, where indexes and index structures are wellknown
since decades, indexes in XDBMS are still an active field of research with
no standards established yet. A lot of approaches have been introduced in recent
years dealing with indexes for querying XML data. The problem of updating an
index is a minor focus of most publications. If at all, the authors describe how
their data structure can be updated from a technical point of view. To the best
of our knowledge, the problem whether an XML index i is affected by a modifying
operation o has never been faced before. We call this problem the XML Index Update
Problem (XIUP).
In this chapter we give an algorithm that is based on finite automata theory and
decides whether an XPath-based database operation affects an index that is defined
universally upon keys, qualifiers, and a return value of an XPath expression.
Although we focus on the KeyX indexing approach the general idea and the algorithms
are transferable to other selective indexing approaches for semistructured
data. If an approach is non-selective (it covers all data) it is affected by any modification.
We present an efficient intersection algorithm for the XPath fragment XP {[],∗,//}
containing path expressions without the NOT operator. The algorithm is based
on finite automata. For the XPath fragment XP {[],∗,//,NOT } containing path expressions
with the NOT operator the intersection problem becomes NP -complete
leading to exponential computations in general. With an average case simulation
we show that the NP -completeness is no significant limitation for most real-world
database operations.
In addition, we provide algorithms for updating our KeyX indexes efficiently if
they are affected by a modification. The Index Update Problem is relevant for all
applications that use a secondary XML data representation (e.g. indexes, caches,
XML replication/synchronization services) where updates must be identified and
realized.