...
Run Format

Source file src/index/suffixarray/qsufsort.go

Documentation: index/suffixarray

  // Copyright 2011 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  
  // This algorithm is based on "Faster Suffix Sorting"
  //   by N. Jesper Larsson and Kunihiko Sadakane
  // paper: http://www.larsson.dogma.net/ssrev-tr.pdf
  // code:  http://www.larsson.dogma.net/qsufsort.c
  
  // This algorithm computes the suffix array sa by computing its inverse.
  // Consecutive groups of suffixes in sa are labeled as sorted groups or
  // unsorted groups. For a given pass of the sorter, all suffixes are ordered
  // up to their first h characters, and sa is h-ordered. Suffixes in their
  // final positions and unambiguously sorted in h-order are in a sorted group.
  // Consecutive groups of suffixes with identical first h characters are an
  // unsorted group. In each pass of the algorithm, unsorted groups are sorted
  // according to the group number of their following suffix.
  
  // In the implementation, if sa[i] is negative, it indicates that i is
  // the first element of a sorted group of length -sa[i], and can be skipped.
  // An unsorted group sa[i:k] is given the group number of the index of its
  // last element, k-1. The group numbers are stored in the inverse slice (inv),
  // and when all groups are sorted, this slice is the inverse suffix array.
  
  package suffixarray
  
  import "sort"
  
  func qsufsort(data []byte) []int {
  	// initial sorting by first byte of suffix
  	sa := sortedByFirstByte(data)
  	if len(sa) < 2 {
  		return sa
  	}
  	// initialize the group lookup table
  	// this becomes the inverse of the suffix array when all groups are sorted
  	inv := initGroups(sa, data)
  
  	// the index starts 1-ordered
  	sufSortable := &suffixSortable{sa: sa, inv: inv, h: 1}
  
  	for sa[0] > -len(sa) { // until all suffixes are one big sorted group
  		// The suffixes are h-ordered, make them 2*h-ordered
  		pi := 0 // pi is first position of first group
  		sl := 0 // sl is negated length of sorted groups
  		for pi < len(sa) {
  			if s := sa[pi]; s < 0 { // if pi starts sorted group
  				pi -= s // skip over sorted group
  				sl += s // add negated length to sl
  			} else { // if pi starts unsorted group
  				if sl != 0 {
  					sa[pi+sl] = sl // combine sorted groups before pi
  					sl = 0
  				}
  				pk := inv[s] + 1 // pk-1 is last position of unsorted group
  				sufSortable.sa = sa[pi:pk]
  				sort.Sort(sufSortable)
  				sufSortable.updateGroups(pi)
  				pi = pk // next group
  			}
  		}
  		if sl != 0 { // if the array ends with a sorted group
  			sa[pi+sl] = sl // combine sorted groups at end of sa
  		}
  
  		sufSortable.h *= 2 // double sorted depth
  	}
  
  	for i := range sa { // reconstruct suffix array from inverse
  		sa[inv[i]] = i
  	}
  	return sa
  }
  
  func sortedByFirstByte(data []byte) []int {
  	// total byte counts
  	var count [256]int
  	for _, b := range data {
  		count[b]++
  	}
  	// make count[b] equal index of first occurrence of b in sorted array
  	sum := 0
  	for b := range count {
  		count[b], sum = sum, count[b]+sum
  	}
  	// iterate through bytes, placing index into the correct spot in sa
  	sa := make([]int, len(data))
  	for i, b := range data {
  		sa[count[b]] = i
  		count[b]++
  	}
  	return sa
  }
  
  func initGroups(sa []int, data []byte) []int {
  	// label contiguous same-letter groups with the same group number
  	inv := make([]int, len(data))
  	prevGroup := len(sa) - 1
  	groupByte := data[sa[prevGroup]]
  	for i := len(sa) - 1; i >= 0; i-- {
  		if b := data[sa[i]]; b < groupByte {
  			if prevGroup == i+1 {
  				sa[i+1] = -1
  			}
  			groupByte = b
  			prevGroup = i
  		}
  		inv[sa[i]] = prevGroup
  		if prevGroup == 0 {
  			sa[0] = -1
  		}
  	}
  	// Separate out the final suffix to the start of its group.
  	// This is necessary to ensure the suffix "a" is before "aba"
  	// when using a potentially unstable sort.
  	lastByte := data[len(data)-1]
  	s := -1
  	for i := range sa {
  		if sa[i] >= 0 {
  			if data[sa[i]] == lastByte && s == -1 {
  				s = i
  			}
  			if sa[i] == len(sa)-1 {
  				sa[i], sa[s] = sa[s], sa[i]
  				inv[sa[s]] = s
  				sa[s] = -1 // mark it as an isolated sorted group
  				break
  			}
  		}
  	}
  	return inv
  }
  
  type suffixSortable struct {
  	sa  []int
  	inv []int
  	h   int
  	buf []int // common scratch space
  }
  
  func (x *suffixSortable) Len() int           { return len(x.sa) }
  func (x *suffixSortable) Less(i, j int) bool { return x.inv[x.sa[i]+x.h] < x.inv[x.sa[j]+x.h] }
  func (x *suffixSortable) Swap(i, j int)      { x.sa[i], x.sa[j] = x.sa[j], x.sa[i] }
  
  func (x *suffixSortable) updateGroups(offset int) {
  	bounds := x.buf[0:0]
  	group := x.inv[x.sa[0]+x.h]
  	for i := 1; i < len(x.sa); i++ {
  		if g := x.inv[x.sa[i]+x.h]; g > group {
  			bounds = append(bounds, i)
  			group = g
  		}
  	}
  	bounds = append(bounds, len(x.sa))
  	x.buf = bounds
  
  	// update the group numberings after all new groups are determined
  	prev := 0
  	for _, b := range bounds {
  		for i := prev; i < b; i++ {
  			x.inv[x.sa[i]] = offset + b - 1
  		}
  		if b-prev == 1 {
  			x.sa[prev] = -1
  		}
  		prev = b
  	}
  }
  

View as plain text