use of org.apache.lucene.util.automaton.LevenshteinAutomata in project lucene-solr by apache.
the class FuzzySuggester method toLevenshteinAutomata.
Automaton toLevenshteinAutomata(Automaton automaton) {
List<Automaton> subs = new ArrayList<>();
FiniteStringsIterator finiteStrings = new FiniteStringsIterator(automaton);
for (IntsRef string; (string = finiteStrings.next()) != null; ) {
if (string.length <= nonFuzzyPrefix || string.length < minFuzzyLength) {
subs.add(Automata.makeString(string.ints, string.offset, string.length));
} else {
int[] ints = new int[string.length - nonFuzzyPrefix];
System.arraycopy(string.ints, string.offset + nonFuzzyPrefix, ints, 0, ints.length);
// TODO: maybe add alphaMin to LevenshteinAutomata,
// and pass 1 instead of 0? We probably don't want
// to allow the trailing dedup bytes to be
// edited... but then 0 byte is "in general" allowed
// on input (but not in UTF8).
LevenshteinAutomata lev = new LevenshteinAutomata(ints, unicodeAware ? Character.MAX_CODE_POINT : 255, transpositions);
subs.add(lev.toAutomaton(maxEdits, UnicodeUtil.newString(string.ints, string.offset, nonFuzzyPrefix)));
}
}
if (subs.isEmpty()) {
// matches nothing
return Automata.makeEmpty();
} else if (subs.size() == 1) {
// no synonyms or anything: just a single path through the tokenstream
return subs.get(0);
} else {
// multiple paths: this is really scary! is it slow?
// maybe we should not do this and throw UOE?
Automaton a = Operations.union(subs);
// this only happens if you have multiple paths anyway (e.g. synonyms)
return Operations.determinize(a, DEFAULT_MAX_DETERMINIZED_STATES);
}
}
use of org.apache.lucene.util.automaton.LevenshteinAutomata in project lucene-solr by apache.
the class FuzzyCompletionQuery method toLevenshteinAutomata.
private Automaton toLevenshteinAutomata(Automaton automaton, Set<IntsRef> refs) {
List<Automaton> subs = new ArrayList<>();
FiniteStringsIterator finiteStrings = new FiniteStringsIterator(automaton);
for (IntsRef string; (string = finiteStrings.next()) != null; ) {
refs.add(IntsRef.deepCopyOf(string));
if (string.length <= nonFuzzyPrefix || string.length < minFuzzyLength) {
subs.add(Automata.makeString(string.ints, string.offset, string.length));
} else {
int[] ints = new int[string.length - nonFuzzyPrefix];
System.arraycopy(string.ints, string.offset + nonFuzzyPrefix, ints, 0, ints.length);
// TODO: maybe add alphaMin to LevenshteinAutomata,
// and pass 1 instead of 0? We probably don't want
// to allow the trailing dedup bytes to be
// edited... but then 0 byte is "in general" allowed
// on input (but not in UTF8).
LevenshteinAutomata lev = new LevenshteinAutomata(ints, unicodeAware ? Character.MAX_CODE_POINT : 255, transpositions);
subs.add(lev.toAutomaton(maxEdits, UnicodeUtil.newString(string.ints, string.offset, nonFuzzyPrefix)));
}
}
if (subs.isEmpty()) {
// matches nothing
return Automata.makeEmpty();
} else if (subs.size() == 1) {
// no synonyms or anything: just a single path through the tokenstream
return subs.get(0);
} else {
// multiple paths: this is really scary! is it slow?
// maybe we should not do this and throw UOE?
Automaton a = Operations.union(subs);
// this only happens if you have multiple paths anyway (e.g. synonyms)
return Operations.determinize(a, maxDeterminizedStates);
}
}
use of org.apache.lucene.util.automaton.LevenshteinAutomata in project stargate-core by tuplejump.
the class FuzzyCondition method getAutomaton.
@Override
public Automaton getAutomaton(Options schema) {
Properties properties = schema.getProperties(field);
String message;
Type fieldType = properties != null ? properties.getType() : Type.text;
if (fieldType == Type.string || fieldType == Type.text) {
String analyzedValue = analyze(field, value, schema.analyzer);
LevenshteinAutomata levenshteinAutomata = new LevenshteinAutomata(analyzedValue, transpositions);
return levenshteinAutomata.toAutomaton(maxEdits);
}
message = String.format("Fuzzy queries cannot be supported for field type %s", fieldType);
throw new UnsupportedOperationException(message);
}
use of org.apache.lucene.util.automaton.LevenshteinAutomata in project elasticsearch by elastic.
the class XFuzzySuggester method toLevenshteinAutomata.
Automaton toLevenshteinAutomata(Automaton automaton) {
List<Automaton> subs = new ArrayList<>();
FiniteStringsIterator finiteStrings = new FiniteStringsIterator(automaton);
for (IntsRef string; (string = finiteStrings.next()) != null; ) {
if (string.length <= nonFuzzyPrefix || string.length < minFuzzyLength) {
subs.add(Automata.makeString(string.ints, string.offset, string.length));
} else {
int[] ints = new int[string.length - nonFuzzyPrefix];
System.arraycopy(string.ints, string.offset + nonFuzzyPrefix, ints, 0, ints.length);
// TODO: maybe add alphaMin to LevenshteinAutomata,
// and pass 1 instead of 0? We probably don't want
// to allow the trailing dedup bytes to be
// edited... but then 0 byte is "in general" allowed
// on input (but not in UTF8).
LevenshteinAutomata lev = new LevenshteinAutomata(ints, unicodeAware ? Character.MAX_CODE_POINT : 255, transpositions);
subs.add(lev.toAutomaton(maxEdits, UnicodeUtil.newString(string.ints, string.offset, nonFuzzyPrefix)));
}
}
if (subs.isEmpty()) {
// matches nothing
return Automata.makeEmpty();
} else if (subs.size() == 1) {
// no synonyms or anything: just a single path through the tokenstream
return subs.get(0);
} else {
// multiple paths: this is really scary! is it slow?
// maybe we should not do this and throw UOE?
Automaton a = Operations.union(subs);
// this only happens if you have multiple paths anyway (e.g. synonyms)
return Operations.determinize(a, DEFAULT_MAX_DETERMINIZED_STATES);
}
}
use of org.apache.lucene.util.automaton.LevenshteinAutomata in project lucene-solr by apache.
the class MultiTermHighlighting method extractAutomata.
/**
* Extracts MultiTermQueries that match the provided field predicate.
* Returns equivalent automata that will match terms.
*/
public static CharacterRunAutomaton[] extractAutomata(Query query, Predicate<String> fieldMatcher, boolean lookInSpan, Function<Query, Collection<Query>> preRewriteFunc) {
// TODO Lucene needs a Query visitor API! LUCENE-3041
List<CharacterRunAutomaton> list = new ArrayList<>();
Collection<Query> customSubQueries = preRewriteFunc.apply(query);
if (customSubQueries != null) {
for (Query sub : customSubQueries) {
list.addAll(Arrays.asList(extractAutomata(sub, fieldMatcher, lookInSpan, preRewriteFunc)));
}
} else if (query instanceof BooleanQuery) {
for (BooleanClause clause : (BooleanQuery) query) {
if (!clause.isProhibited()) {
list.addAll(Arrays.asList(extractAutomata(clause.getQuery(), fieldMatcher, lookInSpan, preRewriteFunc)));
}
}
} else if (query instanceof ConstantScoreQuery) {
list.addAll(Arrays.asList(extractAutomata(((ConstantScoreQuery) query).getQuery(), fieldMatcher, lookInSpan, preRewriteFunc)));
} else if (query instanceof BoostQuery) {
list.addAll(Arrays.asList(extractAutomata(((BoostQuery) query).getQuery(), fieldMatcher, lookInSpan, preRewriteFunc)));
} else if (query instanceof DisjunctionMaxQuery) {
for (Query sub : ((DisjunctionMaxQuery) query).getDisjuncts()) {
list.addAll(Arrays.asList(extractAutomata(sub, fieldMatcher, lookInSpan, preRewriteFunc)));
}
} else if (lookInSpan && query instanceof SpanOrQuery) {
for (Query sub : ((SpanOrQuery) query).getClauses()) {
list.addAll(Arrays.asList(extractAutomata(sub, fieldMatcher, lookInSpan, preRewriteFunc)));
}
} else if (lookInSpan && query instanceof SpanNearQuery) {
for (Query sub : ((SpanNearQuery) query).getClauses()) {
list.addAll(Arrays.asList(extractAutomata(sub, fieldMatcher, lookInSpan, preRewriteFunc)));
}
} else if (lookInSpan && query instanceof SpanNotQuery) {
list.addAll(Arrays.asList(extractAutomata(((SpanNotQuery) query).getInclude(), fieldMatcher, lookInSpan, preRewriteFunc)));
} else if (lookInSpan && query instanceof SpanPositionCheckQuery) {
list.addAll(Arrays.asList(extractAutomata(((SpanPositionCheckQuery) query).getMatch(), fieldMatcher, lookInSpan, preRewriteFunc)));
} else if (lookInSpan && query instanceof SpanBoostQuery) {
list.addAll(Arrays.asList(extractAutomata(((SpanBoostQuery) query).getQuery(), fieldMatcher, lookInSpan, preRewriteFunc)));
} else if (lookInSpan && query instanceof SpanMultiTermQueryWrapper) {
list.addAll(Arrays.asList(extractAutomata(((SpanMultiTermQueryWrapper<?>) query).getWrappedQuery(), fieldMatcher, lookInSpan, preRewriteFunc)));
} else if (query instanceof PrefixQuery) {
final PrefixQuery pq = (PrefixQuery) query;
Term prefix = pq.getPrefix();
if (fieldMatcher.test(prefix.field())) {
list.add(new CharacterRunAutomaton(Operations.concatenate(Automata.makeString(prefix.text()), Automata.makeAnyString())) {
@Override
public String toString() {
return pq.toString();
}
});
}
} else if (query instanceof FuzzyQuery) {
final FuzzyQuery fq = (FuzzyQuery) query;
if (fieldMatcher.test(fq.getField())) {
String utf16 = fq.getTerm().text();
int[] termText = new int[utf16.codePointCount(0, utf16.length())];
for (int cp, i = 0, j = 0; i < utf16.length(); i += Character.charCount(cp)) {
termText[j++] = cp = utf16.codePointAt(i);
}
int termLength = termText.length;
int prefixLength = Math.min(fq.getPrefixLength(), termLength);
String suffix = UnicodeUtil.newString(termText, prefixLength, termText.length - prefixLength);
LevenshteinAutomata builder = new LevenshteinAutomata(suffix, fq.getTranspositions());
String prefix = UnicodeUtil.newString(termText, 0, prefixLength);
Automaton automaton = builder.toAutomaton(fq.getMaxEdits(), prefix);
list.add(new CharacterRunAutomaton(automaton) {
@Override
public String toString() {
return fq.toString();
}
});
}
} else if (query instanceof TermRangeQuery) {
final TermRangeQuery tq = (TermRangeQuery) query;
if (fieldMatcher.test(tq.getField())) {
final CharsRef lowerBound;
if (tq.getLowerTerm() == null) {
lowerBound = null;
} else {
lowerBound = new CharsRef(tq.getLowerTerm().utf8ToString());
}
final CharsRef upperBound;
if (tq.getUpperTerm() == null) {
upperBound = null;
} else {
upperBound = new CharsRef(tq.getUpperTerm().utf8ToString());
}
final boolean includeLower = tq.includesLower();
final boolean includeUpper = tq.includesUpper();
final CharsRef scratch = new CharsRef();
@SuppressWarnings("deprecation") final Comparator<CharsRef> comparator = CharsRef.getUTF16SortedAsUTF8Comparator();
// this is *not* an automaton, but its very simple
list.add(new CharacterRunAutomaton(Automata.makeEmpty()) {
@Override
public boolean run(char[] s, int offset, int length) {
scratch.chars = s;
scratch.offset = offset;
scratch.length = length;
if (lowerBound != null) {
int cmp = comparator.compare(scratch, lowerBound);
if (cmp < 0 || (!includeLower && cmp == 0)) {
return false;
}
}
if (upperBound != null) {
int cmp = comparator.compare(scratch, upperBound);
if (cmp > 0 || (!includeUpper && cmp == 0)) {
return false;
}
}
return true;
}
@Override
public String toString() {
return tq.toString();
}
});
}
} else if (query instanceof AutomatonQuery) {
final AutomatonQuery aq = (AutomatonQuery) query;
if (fieldMatcher.test(aq.getField())) {
list.add(new CharacterRunAutomaton(aq.getAutomaton()) {
@Override
public String toString() {
return aq.toString();
}
});
}
}
return list.toArray(new CharacterRunAutomaton[list.size()]);
}
Aggregations