1 package org.wikimedia.search.extra.regex.ngram;
2
3 import org.apache.lucene.analysis.Analyzer;
4 import org.apache.lucene.util.automaton.Automaton;
5 import org.wikimedia.search.extra.regex.expression.Expression;
6 import org.wikimedia.search.extra.regex.expression.True;
7
8 /**
9 * Extracts ngrams from automatons.
10 */
11 public class NGramExtractor {
12 private final int gramSize;
13 private final int maxExpand;
14 private final int maxStatesTraced;
15 private final int maxNgrams;
16 private final Analyzer ngramAnalyzer;
17
18 /**
19 * Build it.
20 *
21 * @param gramSize size of the ngram. The "n" in ngram.
22 * @param maxExpand Maximum size of range transitions to expand into single
23 * transitions. Its roughly analogous to the number of characters
24 * in a character class before it is considered a wildcard for
25 * optimization purposes.
26 * @param maxStatesTraced maximum number of states traced during automaton
27 * functions. Higher number allow more complex automata to be
28 * converted to ngram expressions at the cost of more time.
29 * @param maxNgrams the maximum number of ngrams extracted from the regex.
30 * If more could be exracted from the regex they are ignored.
31 * @param ngramAnalyzer the analyzer used to generate indexed ngrams
32 */
33 public NGramExtractor(int gramSize, int maxExpand, int maxStatesTraced, int maxNgrams, Analyzer ngramAnalyzer) {
34 this.gramSize = gramSize;
35 this.maxExpand = maxExpand;
36 this.maxStatesTraced = maxStatesTraced;
37 this.maxNgrams = maxNgrams;
38 this.ngramAnalyzer = ngramAnalyzer;
39 }
40
41 /**
42 * Extract an Expression containing ngrams from an automaton.
43 */
44 public Expression<String> extract(Automaton automaton) {
45 if (automaton.isAccept(0)) {
46 return True.instance();
47 }
48 return new NGramAutomaton(automaton, gramSize, maxExpand, maxStatesTraced, maxNgrams, ngramAnalyzer).expression().simplify();
49 }
50 }