Fast and flexible string matching by combining bit-parallelism and suffix automata [28 citations — 11 self]
|
Popular Tags
No tags have been applied to this document.
BibTeX | Add To MetaCart
@ARTICLE{Navarro_fastand,author = {Gonzalo Navarro},
title = {Fast and flexible string matching by combining bit-parallelism and suffix automata},
journal = {ACM Journal of Experimental Algorithmics (JEA},
year = {},
volume = {5},
pages = {2000}
}
Years of Citing Articles
Abstract:
Several string matching algorithms exist, the most famous are Knuth-Morris-Pratt (KMP), BoyerMoore (BM), and some variations on BM, like Hoorspool and Sunday. Most of these algorithms rely on different kinds of automata to speed up the search, which were traditionally made deterministic. After 1990, two new approaches have been studied separately. The first one simulates the automata in their nondeterministic form by using bits and exploiting the intrinsic parallelism inside the computer word, e.g. Shift-Or. Those algorithms are extended to handle classes of characters and errors in the pattern and/or in the text, their drawback being their inability to skip characters. The second one uses "suffix automata " to design new optimal string matching algorithms, e.g. BDM and Turbo BDM. In this paper we merge both approaches to obtain a new algorithm, called BNDM, which uses a nondeterministic suffix automaton simulated using bit-parallelism. This algorithm is 20%-25 % faster than BDM, uses less memory, it is 2-3 times faster than Shift-Or, it is 10%-40 % faster than all the BM family, and it is very simple to implement. The algorithm becomes the fastest in all cases, except for extremely short or extremely long patterns (e.g. on English we are the fastest between 2 and 110 characters). Moreover, the algorithm inherits all the flexibility of the bit-parallel paradigm: we show that all the extensions devised for Shift-Or to handle classes of characters, multiple patterns and even errors can be speeded-up with the technique to skip characters. We obtain faster, very competitive algorithms for all these extensions. In particular ours is by far the fastest technique to deal with classes of characters. As a theoretical development related with this, we introduce a new automaton to recognize suffixes of patterns with classes of characters. To the best of our knowledge, this automaton has not been studied before.
