eu.interedition.collatex.suffixarray.SuffixArrays Maven / Gradle / Ivy
package eu.interedition.collatex.suffixarray;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.List;
/*
* TODO: ultimately, this class should be "intelligent" enough to pick the best
* algorithm, depending on the distribution and properties of the input (alphabet size,
* symbols distribution, etc.).
*/
/**
*
* Factory-like methods for constructing suffix arrays for various data types. Whenever
* defaults are provided, they aim to be sensible, "best guess" values for the given data
* type.
*
* Note the following important aspects that apply to nearly all methods in this class:
*
* - In nearly all cases, the returned suffix array will not be length-equal to the
* input sequence (will be slightly larger). It is so because most algorithms use extra
* space for end of sequence delimiters and it makes little sense to temporary duplicate
* memory consumption just to have exact length counts.
*
*
* @author Michał Nowak (Carrot Search)
* @author Dawid Weiss (Carrot Search)
* @author Anton Olsson for friprogramvarusyndikatet.se
*/
public final class SuffixArrays {
/**
* Maximum required trailing space in the input array (certain algorithms need it).
*/
final static int MAX_EXTRA_TRAILING_SPACE = DeepShallow.OVERSHOOT;
/*
*
*/
private SuffixArrays() {
// no instances.
}
/**
* Create a suffix array for a given character sequence with the default algorithm.
*/
public static int[] create(CharSequence s) {
return create(s, defaultAlgorithm());
}
/**
* Create a suffix array for a given character sequence, using the provided suffix
* array building strategy.
*/
public static int[] create(CharSequence s, ISuffixArrayBuilder builder) {
return new CharSequenceAdapter(builder).buildSuffixArray(s);
}
/**
* Create a suffix array and an LCP array for a given character sequence.
*
* @see #computeLCP(int[], int, int, int[])
*/
public static SuffixData createWithLCP(CharSequence s) {
return createWithLCP(s, defaultAlgorithm());
}
/**
* Create a suffix array and an LCP array for a given character sequence, use the
* given algorithm for building the suffix array.
*
* @see #computeLCP(int[], int, int, int[])
*/
public static SuffixData createWithLCP(CharSequence s, ISuffixArrayBuilder builder) {
final CharSequenceAdapter adapter = new CharSequenceAdapter(builder);
final int[] sa = adapter.buildSuffixArray(s);
final int[] lcp = computeLCP(adapter.input, 0, s.length(), sa);
return new SuffixData(sa, lcp);
}
/**
* Create a suffix array and an LCP array for a given input sequence of symbols.
*/
public static SuffixData createWithLCP(int[] input, int start, int length) {
final ISuffixArrayBuilder builder = new DensePositiveDecorator(
new ExtraTrailingCellsDecorator(defaultAlgorithm(), 3));
return createWithLCP(input, start, length, builder);
}
/**
* Create a suffix array and an LCP array for a given input sequence of symbols and a
* custom suffix array building strategy.
*/
public static SuffixData createWithLCP(int[] input, int start, int length,
ISuffixArrayBuilder builder) {
final int[] sa = builder.buildSuffixArray(input, start, length);
final int[] lcp = computeLCP(input, start, length, sa);
return new SuffixData(sa, lcp);
}
/**
* Create a suffix array and an LCP array for a given generic array and a
* custom suffix array building strategy, using the given T object
* comparator.
*/
public static SuffixData createWithLCP(T[] input, ISuffixArrayBuilder builder, Comparator comparator) {
final GenericArrayAdapter adapter = new GenericArrayAdapter(builder, comparator);
final int[] sa = adapter.buildSuffixArray(input);
final int[] lcp = computeLCP(adapter.input, 0, input.length, sa);
return new SuffixData(sa, lcp);
}
/**
* Calculate longest prefix (LCP) array for an existing suffix array and input. Index
* i of the returned array indicates the length of the common prefix
* between suffix i and i-1. The 0-th
* index has a constant value of -1.
*
* The algorithm used to compute the LCP comes from
* T. Kasai, G. Lee, H. Arimura, S. Arikawa, and K. Park. Linear-time longest-common-prefix
* computation in suffix arrays and its applications. In Proc. 12th Symposium on Combinatorial
* Pattern Matching (CPM ’01), pages 181–192. Springer-Verlag LNCS n. 2089, 2001.
*/
public static int[] computeLCP(int[] input, final int start, final int length,
int[] sa) {
final int[] rank = new int[length];
for (int i = 0; i < length; i++)
rank[sa[i]] = i;
int h = 0;
final int[] lcp = new int[length];
for (int i = 0; i < length; i++) {
int k = rank[i];
if (k == 0) {
lcp[k] = -1;
} else {
final int j = sa[k - 1];
while (i + h < length && j + h < length
&& input[start + i + h] == input[start + j + h]) {
h++;
}
lcp[k] = h;
}
if (h > 0) h--;
}
return lcp;
}
/**
* @return Return a new instance of the default algorithm for use in other methods. At
* the moment {@link QSufSort} is used.
*/
private static ISuffixArrayBuilder defaultAlgorithm() {
return new QSufSort();
}
/**
* Utility method converting all suffixes of a given sequence to a list of strings.
*/
public static List toString(CharSequence input, int[] suffixes) {
final String full = input.toString();
final ArrayList result = new ArrayList<>();
for (int i = 0; i < input.length(); i++) {
result.add(full.subSequence(suffixes[i], full.length()));
}
return result;
}
}