The application of space-filling curves to the storage and retrieval of multi-dimensional data

Indexing of multi-dimensional data has been the focus of a considerable amount of research e ort over many years but no generally agreed paradigm has emerged to compare with the impact of the B-Tree, for example, on the indexing of one-dimensional data. At the same time, the need for e cient methods is ever more important in an environment where databases become larger and more complex in their structures. Mapping multi-dimensional data to one dimension, thus enabling one-dimensional access methods to be exploited, has been suggested in the literature but for the most part interest has been con ned to the Z-order curve. The possibility of using other curves, such as the Hilbert and Gray-code curves, whose characteristics di er from those of the Z-order curve, has also been suggested. In this thesis we design and implement a working le store which is underpinned by the principle of mapping multi-dimensional data to one of a variety of spacelling curves and their variants. Data is then indexed using a B+ Tree which remains compact, regardless of the volume and number of dimensions. The implementation has entailed developing algorithms for mapping data to one dimension and, most importantly, developing algorithms to facilitate the querying of data in a exible way. We focus on the Hilbert curve but also consider other curves and propose new alternative algorithms for querying data mapped to the Z-order curve. The current implementation accommodates data in up to sixteen dimensions but the approach is generic and not limited to this number. We report on preliminary testing of the implemetation, which provides very encouraging results. We also undertake a brief exploration of the application of spacelling curves to the indexing of spatial data.