Faster Path Indexes for Search in XML Data

This article describes how to implement efficient memory resident path indexes for semi-structured data. Two techniques are introduced, and they are shown to be significantly faster than previous methods when facing path queries using the descendant axis and wild-cards. The first is conceptually simple and combines inverted lists, selectivity estimation, hit expansion and brute force search. The second uses suffix trees with additional statistics and multiple entry points into the query. The entry points are partially evaluated in an order based on estimated cost until one of them is complete. Many path index implementations are tested, using paths generated both from statistical models and DTDs.

• Type: Conference
• Year: 2008
• Where: ADC
• Authors: Nils Grimsmo
• Keywords: ADC 2008, Database, Entry Points, Path Queries, Resident Path Indexes