TRACK : A Novel XML Join Algorithm for Efficient Processing Twig Queries

In order to find all occurrences of a tree/twig pattern in an XML database, a number of holistic twig join algorithms have been proposed. However, most of these algorithms focus on identifying a larger query class or using a novel label scheme to reduce I/O operations, and ignore the deficiency of the root-to-leaf strategy. In this paper, we propose a novel twig join algorithm called Track, which adopts the opposite leaf-to-root strategy to process queries. It brings us two benefits: (i) avoiding too much time checking the element index to make sure all branches are satisfied before a new element comes, (ii)using the tree structure to encode final tree matches so as to avoid the merging process. Further experiments on diverse data sets show that our algorithm is indeed superior to current algorithms in terms of query processing performance.