it.unimi.dsi.util.TernaryIntervalSearchTree Maven / Gradle / Ivy
Show all versions of dsi-utils Show documentation
package it.unimi.dsi.util;
/*
* DSI utilities
*
* Copyright (C) 2005-2009 Sebastiano Vigna
*
* This library is free software; you can redistribute it and/or modify it
* under the terms of the GNU Lesser General Public License as published by the Free
* Software Foundation; either version 2.1 of the License, or (at your option)
* any later version.
*
* This library is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
* or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License
* for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
*/
import it.unimi.dsi.fastutil.io.BinIO;
import it.unimi.dsi.io.FastBufferedReader;
import it.unimi.dsi.lang.MutableString;
import it.unimi.dsi.logging.ProgressLogger;
import java.io.IOException;
import java.io.InputStreamReader;
import java.io.Serializable;
import java.nio.charset.Charset;
import java.util.Collection;
import java.util.Iterator;
import com.martiansoftware.jsap.FlaggedOption;
import com.martiansoftware.jsap.JSAP;
import com.martiansoftware.jsap.JSAPException;
import com.martiansoftware.jsap.JSAPResult;
import com.martiansoftware.jsap.Parameter;
import com.martiansoftware.jsap.SimpleJSAP;
import com.martiansoftware.jsap.UnflaggedOption;
import com.martiansoftware.jsap.stringparsers.ForNameStringParser;
/** Ternary interval search trees.
*
* Ternary search trees are a data structure used to store words over an alphabet; they are
* a useful alternatives to tries when the alphabet is large.
*
*
Ternary interval search trees have the additional properties of being able
* to locate quickly intervals of words extending a given prefix (where “quickly” means
* that no more successful character comparisons than the prefix length are performed). They do so
* by storing at each node the number of words covered by that node.
*
*
This implementation exposes a number of interfaces: in particular, the set of words is
* seen as a lexicographically ordered {@link it.unimi.dsi.fastutil.objects.ObjectList}.
*
*
This class is mutable, but for the time it implements only {@link #add(CharSequence)}. Words cannot
* be removed.
*/
public class TernaryIntervalSearchTree extends AbstractPrefixMap implements Serializable {
private static final long serialVersionUID = 1L;
/** A node of the tree. */
private final static class Node implements Serializable {
private static final long serialVersionUID = 1L;
/** A pointer to the left subtree. */
public Node left;
/** A pointer to the middle subtree. */
public Node middle;
/** A pointer to the right subtree. */
public Node right;
/** The nonempty path compressed at this node. */
public char[] path;
/** Whether this node represents a word. */
public boolean isWord;
/** The number of words covered by this node (including the word possibly represented by this node). */
public int numNodes;
/** Creates a new node containing a path specified by a character-sequence fragment.
*
* @param s a character sequence contaning the path of the node.
* @param offset the starting character of the path.
* @param length the length of the path.
* @param isWord whether this node represents a word.
* @param numNodes the number of words covered by this node.
*/
public Node( final CharSequence s, final int offset, final int length, final boolean isWord, final int numNodes ) {
path = new char[ length ];
MutableString.getChars( s, offset, offset + length, path, 0 );
this.isWord = isWord;
this.numNodes = numNodes;
}
/** Creates a new node containing a path specified by a character-array fragment.
*
* @param a a character array contaning the path of the node.
* @param offset the starting character of the path.
* @param length the length of the path.
* @param isWord whether this node represents a word.
* @param numNodes the number of words covered by this node.
*/
public Node( final char[] a, final int offset, final int length, final boolean isWord, final int numNodes ) {
path = new char[ length ];
System.arraycopy( a, offset, path, 0, length );
this.isWord = isWord;
this.numNodes = numNodes;
}
/** Removes a prefix from the path of this node.
*
* @param length the length of the prefix to be removed
*/
public void removePathPrefix( final int length ) {
final char[] a = new char[ path.length - length ];
System.arraycopy( path, length, a, 0, a.length );
path = a;
}
}
/** The root of the tree. */
private Node root;
/** The number of nodes in the tree. */
private int size;
/** Creates a new empty ternary search tree. */
public TernaryIntervalSearchTree() {
defRetValue = -1;
}
/** Creates a new empty ternary search tree and populates it with a given collection of character sequences.
*
* @param c a collection of character sequences.
* */
public TernaryIntervalSearchTree( final Collection c ) {
int n = c.size();
final Iterator i = c.iterator();
while( n-- != 0 ) add( i.next() );
defRetValue = -1;
}
protected Interval getInterval( final CharSequence s ) {
final int l = s.length();
Node e = root;
int i;
int offset = 0;
int wordsAtLeft = 0;
char c;
char[] path;
while( e != null ) {
path = e.path;
for( i = 0; i < path.length - 1 && offset + i < l && s.charAt( offset + i ) == path[ i ]; i++ );
if ( offset + i == l ) return Interval.valueOf( wordsAtLeft, wordsAtLeft + e.numNodes - 1 );
if ( i < path.length - 1 ) return Intervals.EMPTY_INTERVAL;
offset += i;
c = s.charAt( offset );
if ( c < path[ i ] ) e = e.left;
else if ( c > path[ i ] ) {
if ( e.left != null ) wordsAtLeft += e.left.numNodes;
if ( e.middle != null ) wordsAtLeft += e.middle.numNodes;
if ( e.isWord ) wordsAtLeft++;
e = e.right;
}
else {
offset++;
if ( e.left != null ) wordsAtLeft += e.left.numNodes;
if ( offset == l ) return Interval.valueOf( wordsAtLeft, wordsAtLeft + ( e.isWord ? 1 : 0 ) + ( e.middle == null ? 0 : e.middle.numNodes ) - 1 );
if ( e.isWord ) wordsAtLeft++;
e = e.middle;
}
}
return Intervals.EMPTY_INTERVAL;
}
public Interval getApproximatedInterval( final CharSequence s ) {
final int l = s.length();
Node e = root;
int i;
int offset = 0;
int wordsAtLeft = 0;
char c;
char[] path;
while( e != null ) {
path = e.path;
for( i = 0; i < path.length - 1 && offset + i < l && s.charAt( offset + i ) == path[ i ]; i++ );
if ( offset + i == l ) {
// Our sequence is a proper prefix of path.
return wordsAtLeft > 0 ? Interval.valueOf( wordsAtLeft - 1, wordsAtLeft + e.numNodes - 1 ) : Interval.valueOf( wordsAtLeft, wordsAtLeft + e.numNodes - 1 );
}
if ( i < path.length - 1 ) {
// We stopped the loop prematurely.
if ( s.charAt( offset + i ) < path[ i ] ) return wordsAtLeft > 0 ? Interval.valueOf( wordsAtLeft -1 ) : Intervals.EMPTY_INTERVAL;
else return Interval.valueOf( wordsAtLeft + e.numNodes - 1 );
}
offset += i;
c = s.charAt( offset );
if ( c < path[ i ] ) e = e.left;
else if ( c > path[ i ] ) {
if ( e.left != null ) wordsAtLeft += e.left.numNodes;
if ( e.middle != null ) wordsAtLeft += e.middle.numNodes;
if ( e.isWord ) wordsAtLeft++;
e = e.right;
}
else {
offset++;
if ( e.left != null ) wordsAtLeft += e.left.numNodes;
if ( offset == l ) return Interval.valueOf( wordsAtLeft - ( e.isWord ? 0 : 1 ), wordsAtLeft + ( e.isWord ? 1 : 0 ) + ( e.middle == null ? 0 : e.middle.numNodes ) - 1 );
if ( e.isWord ) wordsAtLeft++;
e = e.middle;
}
}
return wordsAtLeft > 0 ? Interval.valueOf( wordsAtLeft - 1 ) : Intervals.EMPTY_INTERVAL;
}
protected MutableString getTerm( int index, final MutableString s ) {
Node e = root;
for( ;; ) {
if ( e.left != null ) {
if ( index < e.left.numNodes ) {
s.append( e.path, 0, e.path.length - 1 );
e = e.left;
continue;
}
index -= e.left.numNodes;
}
if ( e.isWord ) {
if ( index == 0 ) return s.append( e.path ).compact();
index--;
}
if ( e.middle != null ) {
if ( index < e.middle.numNodes ) {
s.append( e.path );
e = e.middle;
continue;
}
index -= e.middle.numNodes;
}
s.append( e.path, 0, e.path.length - 1 );
e = e.right;
}
}
protected long getIndex( final CharSequence s ) {
final int l = s.length();
Node e = root;
int i;
int offset = 0;
int wordsAtLeft = 0;
char c;
char[] path;
while( e != null ) {
path = e.path;
for( i = 0; i < path.length - 1; i++ )
if ( offset + i == l || s.charAt( offset + i ) != path[ i ] ) return -1;
offset += i;
if ( offset == l ) return -1;
c = s.charAt( offset );
if ( c < e.path[ i ] ) e = e.left;
else if ( c > e.path[ i ] ) {
if ( e.left != null ) wordsAtLeft += e.left.numNodes;
if ( e.middle != null ) wordsAtLeft += e.middle.numNodes;
if ( e.isWord ) wordsAtLeft++;
e = e.right;
}
else {
offset++;
if ( e.left != null ) wordsAtLeft += e.left.numNodes;
if ( offset == l ) return e.isWord ? wordsAtLeft : -1;
if ( e.isWord ) wordsAtLeft++;
e = e.middle;
}
}
return -1;
}
public boolean containsKey( Object o ) {
return getIndex( (CharSequence)o ) != -1;
}
public long getLong( final Object o ) {
final CharSequence s = (CharSequence)o;
final long result = getIndex( s );
return result == -1 ? defRetValue : result;
}
/** True if the last {@link #add(CharSequence)} modified the tree. */
private boolean modified;
public boolean add( final CharSequence s ) {
modified = false;
root = addRec( s, 0, s.length(), root );
return modified;
}
/** Inserts the given character sequence, starting at the given position, in the given subtree.
*
* @param s the character sequence containing the characters to be inserted.
* @param offset the first character to be inserted.
* @param length the number of characters to be inserted.
* @param e the subtree in which the characters should be inserted, or null if
* a new node should be created.
* @return the new node at the top of the subtree.
*/
private Node addRec( final CharSequence s, final int offset, final int length, final Node e ) {
if ( e == null ) {
// We create a new node containing all the characters and return it.
modified = true;
size++;
return new Node( s, offset, length, true, 1 );
}
/* We start scanning the path contained in the current node, up to
* the last character excluded. If we find a mismatch, or if we exhaust our
* characters, we must fork this node. */
char c;
int i;
Node n = null;
final char[] path = e.path;
for ( i = 0; i < path.length - 1; i++ ) {
c = s.charAt( offset + i );
if ( c < path[ i ] ) {
/* We fork on the left, keeping just the first i + 1 characters (this is necessary
* as at least one character must be present in every node). The new
* node will cover one word more than e.
*/
n = new Node( path, 0, i + 1, false, e.numNodes + 1 );
n.middle = e;
e.removePathPrefix( i + 1 );
n.left = addRec( s, offset + i, length - i, null );
break;
}
else if ( c > path[ i ] ) {
// As before, but on the right.
n = new Node( path, 0, i + 1, false, e.numNodes + 1 );
n.middle = e;
e.removePathPrefix( i + 1 );
n.right = addRec( s, offset + i, length - i, null );
break;
}
else {
if ( i == length - 1 ) {
/* We exhausted the character sequence. We fork in the middle,
* keeping length characters and marking the new node as
* containing one work. Again, the new code will cover one word
* more than e. */
n = new Node( s, offset, length, true, e.numNodes + 1 );
n.middle = e;
e.removePathPrefix( length );
size++;
modified = true;
break;
}
}
}
if ( i < path.length - 1 ) return n;
/* We are positioned on the last character of the path. In this case our
* behaviour is different, as if we must fork we must not perform any
* splitting. Moreover, if we exhaust the characters we either found
* the new sequence in the tree, or we just have to mark the node. */
c = s.charAt( offset + i );
if ( c < path[ i ] ) {
/** We fork on the left. The number of words under this node will
* increase only if the structure is modified. */
e.left = addRec( s, offset + i, length - i, e.left );
if ( modified ) e.numNodes++;
}
else if ( c > path[ i ] ) {
e.right = addRec( s, offset + i, length - i, e.right );
if ( modified ) e.numNodes++;
}
else {
if ( i == length - 1 ) {
// This is the node.
if ( modified = !e.isWord ) {
e.numNodes++;
size++;
}
e.isWord = true;
}
else {
// We add a node in the middle, completing the sequence.
e.middle = addRec( s, offset + i + 1, length - i - 1, e.middle );
if ( modified ) e.numNodes++;
}
}
return e;
}
public int size() {
return size;
}
public static void main( final String[] arg ) throws IOException, JSAPException, NoSuchMethodException {
final SimpleJSAP jsap = new SimpleJSAP( TernaryIntervalSearchTree.class.getName(), "Builds a ternary interval search tree reading from standard input a newline-separated list of terms.",
new Parameter[] {
new FlaggedOption( "bufferSize", JSAP.INTSIZE_PARSER, "64Ki", JSAP.NOT_REQUIRED, 'b', "buffer-size", "The size of the I/O buffer used to read terms." ),
new FlaggedOption( "encoding", ForNameStringParser.getParser( Charset.class ), "UTF-8", JSAP.NOT_REQUIRED, 'e', "encoding", "The term file encoding." ),
new UnflaggedOption( "tree", JSAP.STRING_PARSER, JSAP.NO_DEFAULT, JSAP.REQUIRED, JSAP.NOT_GREEDY, "The filename for the serialised tree." )
});
JSAPResult jsapResult = jsap.parse( arg );
if ( jsap.messagePrinted() ) return;
final TernaryIntervalSearchTree tree = new TernaryIntervalSearchTree();
MutableString term = new MutableString();
final ProgressLogger pl = new ProgressLogger();
pl.itemsName = "terms";
final FastBufferedReader terms = new FastBufferedReader( new InputStreamReader( System.in, (Charset)jsapResult.getObject( "encoding" ) ), jsapResult.getInt( "bufferSize" ) );
pl.start( "Reading terms..." );
while( terms.readLine( term ) != null ) {
pl.update();
tree.add( term );
}
pl.done();
BinIO.storeObject( tree, jsapResult.getString( "tree" ) );
}
}