Querying virtual hierarchies using virtual prefix-based numbers

Abstract

Prefix-based numbering (PBN) is a popular method for numbering nodes in a hierarchy. But PBN breaks down when a node's location within a hierarchy changes, such as when XML data is queried after being transformed by an XSLT program or when data is reformatted in the return clause of an inner FLWR expression in a nested XQuery program. A query on transformed data cannot be evaluated as efficiently since the extant PBN numbers cannot be used (unless the data is materialized and then renumbered, which can be expensive). In this paper we present a novel strategy to virtually transform the data without instantiating and renumbering. Our method, which we call virtual prefix-based numbering, couples each PBN number with a level array that locates the node in the numbering space of the virtual hierarchy. The virtual numbering space preserves the property that location-based relationships between nodes can be determined by comparing (virtual) numbers.

Virtual eXist-db

“…There are two important parts to evaluating in the transformed space [6]. First, the relationships between nodes potentially change.…”

“…Step 4: Virtual eXist-db is efficient: Next, we will demostrate that the virtual transformation happens only for data used in the query, i.e., it has an efficiency similar to view transformations (see [6] for more details). We increase the non-Addison Wesley size of the document to 1GB and rerun the query with the virtualDoc() to show that the cost remains flat.…”

“…Reintepret the data -Virtual eXist-db uses this approach. We introduced a numbering system called virtual prefix-based numbering (vPBN) and showed that a vPBN number can be moved to a new location within a hierarchy, yet be used just like a PBN number to determine location-based relationships in the context of the new location [6]. In effect, vPBN virtually rather than physically transforms data and supports query evaluation in the transformed data space.…”