Lossless R-tree compression using variable-length codes

The R-tree is one of the most popular multidimensional data structure. This data structure bounds spatially near points in multidimensional rectangles and supports various types of queries, e.g. point and range queries. When a compression of the data structure is considered, we follow two objectives. The first objective is a smaller index file and the second one is a reduction of the query processing time. In this paper, we introduce a lossless R-tree compression using variable-length codes. Although variable-length codes are well known in the area of data compression, they have not been yet successfully applied in the case of the data structure compression. The main reasons of this fact are inefficient decoding/encoding algorithms. In this paper, we apply recently introduced fast decoding algorithms and we show that these codes provides more efficient query processing time than the lossless RLE or lossy quantization compressions. Moreover, we can utilize some features of variable length codes for the compression. The proposed compression method saves 84% of the index file's size compared to the uncompressed R-tree.