Efficient XML - to - SQL Query Translation : Where to Add the Intelligence ? ( Extended

Exporting XML views of relational data gives rise to the problem of translating XML queries into SQL. To date, the focus of most of the work in the published literature [9, 14, 20] has been on mechanisms for correctly translating complex XML queries into SQL queries, with less emphasis on evaluating the quality of the resulting SQL queries. The efficiency of the SQL queries generated by the translation process is the focus in this paper. Translating XML queries to SQL involves translating queries over hierarchical schemas into queries over a flat relational schemas. This turns out to be problematic — a closer look at the queries generated by the published translation algorithms shows that the hierarchical nature of the exported XML schema is often blindly reflected in the generated SQL query, even when this is clearly not necessary. As a result, in many cases even simple path expression queries result in unnecessarily complex SQL queries. This problem is aggravated when the input XML query includes a traversal of the descendant axis (//), because it does not have a simple equivalent in SQL. A natural question to ask next is whether the phenomenon of large, complex SQL queries arising from simple XML queries is avoidable, or if it is intrinsic due to the mismatch in data models. We show by example in Section 2 that in many