Why off-the-shelf RDBMSs are better at XPath than you might expect

摘要: To compensate for the inherent impedance mismatch between relational data model (tables of tuples) and XML (ordered, unranked trees), tree join algorithms have become prevalent means to process in databases, most notably TwigStack[6], structural join[1], staircase join[13] algorithms. However, addition these existing systems depends on a significant invasion underlying database kernel, an option intolerable vendors. Here, we demonstrate that can achieve comparable XPath performance without touching heart system. We carefully exploit functionality accelerate navigation by purely means: partitioned B-trees bring access costs secondary storage minimum, while aggregation functions avoid expensive computation removal duplicate result nodes comply with semantics. Experiments carried out IBM DB2 confirm our approach turn off-the-shelf into efficient processors.