Bounding the Average Move Structure Query for Faster and Smaller RLBWT Permutations

TL;DR

This paper introduces a simplified interval truncation method for move structures in compressed text indexes, achieving optimal average query time, reduced space, and faster construction, especially beneficial for genomics data.

Contribution

It proposes a length capping splitting scheme that simplifies move structure construction, improves query and construction times, and reduces space in RLBWT permutations.

Findings

Length capping bounds average move structure query time to optimal.

The method reduces overall representation size by O(r log r) bits.

Experiments show faster construction, lower memory usage, and significant space reduction in genomic data.

Abstract

The move structure represents permutations with long contiguously permuted intervals in compressed space with optimal query time. They have become an important feature of compressed text indexes using space proportional to the number of Burrows-Wheeler Transform (BWT) runs, often applied in genomics. This is in thanks not only to theoretical improvements over past approaches, but great cache efficiency and average case query time in practice. This is true even without using the worst case guarantees provided by the interval splitting balancing of the original result. In this paper, we show that an even simpler type of splitting, length capping by truncating long intervals, bounds the average move structure query time to optimal whilst obtaining a superior construction time than the traditional approach. This also proves constant query time when amortized over a full traversal of a single cycle permutation from an arbitrary starting position. Such a scheme has surprising benefits both in theory and practice. For a move structure with $r$ runs over a domain $n$, we replace all $O(r \log n)$-bit components to reduce the overall representation by $O(r \log r)$-bits. The worst case query time is also improved to $O(\log \frac{n}{r})$ without balancing. An $O(r)$-time and $O(r)$-space construction lets us apply the method to run-length encoded BWT (RLBWT) permutations such as LF and $\phi$ to obtain optimal-time algorithms for BWT inversion and suffix array (SA) enumeration in $O(r)$ additional working space. Finally, we introduce the Orbit library for move structure support, and use it to evaluate our splitting approach. Experiments find length capping construction is faster and uses less memory than balancing, with faster queries. We also see a space reduction in practice, with at least a $\sim 40\%$ disk size decrease for LF across large repetitive genomic collections.