Author: Nicola Prezza ([email protected]) Joint work with Travis Gagie and Gonzalo Navarro
cite as:
Gagie T, Navarro G, Prezza N. Optimal-time text indexing in BWT-runs bounded space. In Proceedings of the Twenty-Ninth Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2018, New Orleans, NA, USA, January 7-10 2017.
Travis Gagie, Gonzalo Navarro, and Nicola Prezza. Fully Functional Suffix Trees and Optimal Text Searching in BWT-Runs Bounded Space. J. ACM 67, 1, Article 2 (April 2020)
The r-index is the first full-text index of size O(r), r being the number of BWT runs of the input text (of size n), supporting fast (almost optimal) locate of pattern occurrences. The r-index employs a novel suffix array sampling of size 2r; in classical FM-indexes, this sampling would result in a locate time of Omega(n/r) per occurrence. The r-index, on the other hand, reduces this time to O(log(n/r)).
Let s be the alphabet size and fix a constant eps>0. The r-index offers the following tradeoffs:
- Space: r * ( log s + (1+eps)log(n/r) + 2log n ) bits
- Count time: O( (m/eps) * (log (n/r) + log s) )
- Locate time: After count, O( log(n/r) ) time per occurrence
On very repetitive datasets, the r-index locates orders of magnitude faster than the RLCSA (with a sampling rate resulting in the same size for the two indexes).
NEWS: refactored locate strategy. Let (l,r) be the SA range. Now, the index first finds SA[r] and then applies function Phi to locate SA[r-1], SA[r-2], ..., SA[l]. This is both faster and more space efficient than the strategy originally implemented and described in the paper.
To clone the repository, run:
The library has been tested under linux using gcc 6.2.0. You need the SDSL library installed on your system (https://github.com/simongog/sdsl-lite).
We use cmake to generate the Makefile. Create a build folder in the main r-index folder:
mkdir build
run cmake:
cd build; cmake ..
and compile:
make
After compiling, run
ri-build input
This command will create the r-index of the text file "input" and will store it as "input.ri". Use option -o to specify a different basename for the index file.
Run
ri-count index.ri patterns
to count number of occurrences of the patterns, where is a file containing the patterns in pizza&chili format (http://pizzachili.dcc.uchile.cl/experiments.html). To generate pattern files, use the tool http://pizzachili.dcc.uchile.cl/utils/genpatterns.c. Note: the Pizza&chili format consists of a header followed by newline followed by all the patterns concatenated (without any separator). The patterns must all be of the same length.
Run
ri-locate index.ri patterns
to locate all occurrences of the patterns.
Be aware that the above executables are just benchmarking tools: no output is generated (pattern occurrences are deleted after being extracted and not printed to output).
Nicola Prezza has been supported by the project Italian MIUR-SIR CMACBioSeq ("Combinatorial methods for analysis and compression of biological sequences") grant n.~RBSI146R5L, PI: Giovanna Rosone. Link: http://pages.di.unipi.it/rosone/CMACBioSeq.html