Efficient processing of XML containment queries using partition-based schemes

XML query languages provide facilities to query XML data both on their value as well as their structure. A basic operation in processing and optimizing such queries is the containment join, which takes two sets of elements and returns pairs of elements where one is the ancestor (or descendant) of the other. Most of the techniques proposed so far assume that the two sets are already sorted or utilize preexisting indexing schemes. In contrast, a partition-based technique does not require indexing or sorting. Instead, the containment join is processed by dividing the input sets into smaller partitions. In this paper, we present a new partition-based scheme that gracefully adapts to different document sizes. The advantages of our approach are validated through an experimental comparison with previous work. Moreover, the experiments demonstrate that our solution provides a viable alternative to non-partition join algorithms when the input data is neither sorted nor indexed.