PurposeThe purpose of this paper is to reduce the number of join operations for retrieving Extensible Markup Language (XML) data from a relational database.Design/methodology/approachThe paper proposes a new approach to eliminate the join operations for parent‐child traversing and/or sibling searching such that the performance of query processing could be improved. The rationale behind the design of the proposed approach is to distribute the structural information into relational databases.FindingsThe paper finds that the number of join operations which are needed for processing parent‐child traversal and sibling search can be bounded under the proposed approach. It also verifies the capability of the proposed approach by a series of experiments based on the XMark benchmark, for which it has encouraging results.Research limitations/implicationsCompared with previous approaches based on the structure encoding method, the proposed approach needs more space to store additional immediate predecessor's IDs. However, the approach has similar performance to others and it is much easier to implement.Practical implicationsThe experimental results show that the performance of the proposed approach is less than 3 per cent of the well‐known MonetDB approach for processing benchmark queries. Moreover, its bulkloading time is much less than that for the MonetDB. There is no doubt that the approach is efficient for accessing XML data with acceptable overheads.Originality/valueThis paper contributes to the implementations of XML database systems.