Extensible Markup Language (XML) path queries are fundamental to many XML query languages, like XQuery and XPath, which are often used for complex filtering of XML data. Most XML path query processing techniques use either navigation-based algorithms, which compute results by analyzing an input document one tag at a time, or index-based algorithms, which take advantage of pre-computed numbering schemes over the input XML document. Since XML data is highly irregular and complex, the use of indexes is often preferred.

There are many variations of indexing techniques and algorithms available today, but due to the nature of XML data, which is very rich in structure, the authors of this paper propose an advanced indexing technique that could further improve query processing. The proposed method is called APEX, short for “adaptive path index for XML.” APEX is different from other traditional methods because it only stores paths of length two, plus frequent query paths as measured by the query workload. Traditional methods store all paths from root to leaves, which is very inefficient for complex queries. APEX breaks down XML data into the subunits of element, attribute, name, value, and structure; this introduces automatic detection and updates of frequent query paths. Additionally, APEX supports dynamic data insertion, deletion, and structural changes.

In this paper, the authors compare APEX to several other techniques, such as DataGuides and Index Fabric. They discuss the scenarios under which the APEX technique is superior to the others. In particular, they show that APEX can be used to evaluate queries faster. They also present experimental results over real and synthetic XML documents that validate their claims.