Practical Rank/Select Queries over Arbitrary Sequences
Practical Rank/Select Queries over Arbitrary Sequences
复制标题
对任意序列的实用排序/选择查询
DOI:
10.1007/978-3-540-89097-3_18
复制
发表时间:
2008
影响因子:
0.5
通讯作者:
G. Navarro
中科院分区:
文献类型:
--
作者:
Francisco Claude;G. Navarro
We present a practical study on the compact representation of sequences supporting rank , select , and access queries. While there are several theoretical solutions to the problem, only a few have been tried out, and there is little idea on how the others would perform, especially in the case of sequences with very large alphabets. We first present a new practical implementation of the compressed representation for bit sequences proposed by Raman, Raman, and Rao [SODA 2002], that is competitive with the existing ones when the sequences are not too compressible. It also has nice local compression properties, and we show that this makes it an excellent tool for compressed text indexing in combination with the Burrows-Wheeler transform. This shows the practicality of a recent theoretical proposal [Makinen and Navarro, SPIRE 2007], achieving spaces never seen before. Second, for general sequences, we tune wavelet trees for the case of very large alphabets, by removing their pointer information. We show that this gives an excellent solution for representing a sequence within zero-order entropy space, in cases where the large alphabet poses a serious challenge to typical encoding methods. We also present the first implementation of Golynski et al.'s representation [SODA 2006], which offers another interesting time/space trade-off.