Compact Data Structures for Efficient Processing of Distance-Based Join Queries

Compact data structures can represent data with usually a much smaller memory footprint than its plain representation. In addition to maintaining the data in a form that uses less space, they allow us to efficiently access and query the data in its compact form. The $$k^2$$ -tree is a self-indexed, compact data structure used to represent binary matrices, that can also be used to represent points in a spatial dataset. Efficient processing of the Distance-based Join Queries (DJQs) is of great importance in spatial databases due to its wide area of application. Two of the most representative and known DJQs are the K Closest Pairs Query (KCPQ) and the $$\varepsilon $$ Distance Join Query ( $$\varepsilon $$ DJQ). These types of join queries are executed over two spatial datasets and can be solved by plane-sweep algorithms, which are efficient but with great requirements of RAM, to be able to fit the whole datasets into main memory. In this work, we present new and efficient algorithms to implement DJQs over the $$k^2$$ -tree representation of the spatial datasets, experimentally showing that these algorithms are competitive in query times, with much lower memory requirements.