use of org.opensearch.search.suggest.Suggest.Suggestion.Entry in project OpenSearch by opensearch-project.
the class Suggest method reduce.
public static List<Suggestion<? extends Entry<? extends Option>>> reduce(Map<String, List<Suggest.Suggestion>> groupedSuggestions) {
List<Suggestion<? extends Entry<? extends Option>>> reduced = new ArrayList<>(groupedSuggestions.size());
for (Map.Entry<String, List<Suggestion>> unmergedResults : groupedSuggestions.entrySet()) {
List<Suggestion> value = unmergedResults.getValue();
Class<? extends Suggestion> suggestionClass = null;
for (Suggestion suggestion : value) {
if (suggestionClass == null) {
suggestionClass = suggestion.getClass();
} else if (suggestionClass != suggestion.getClass()) {
throw new IllegalArgumentException("detected mixed suggestion results, due to querying on old and new completion suggester," + " query on a single completion suggester version");
}
}
Suggestion reduce = value.get(0).reduce(value);
reduce.trim();
reduced.add(reduce);
}
return reduced;
}
use of org.opensearch.search.suggest.Suggest.Suggestion.Entry in project OpenSearch by opensearch-project.
the class SuggestionTests method doTestFromXContent.
@SuppressWarnings({ "rawtypes" })
private void doTestFromXContent(boolean addRandomFields) throws IOException {
ToXContent.Params params = new ToXContent.MapParams(Collections.singletonMap(RestSearchAction.TYPED_KEYS_PARAM, "true"));
for (Class<Suggestion<? extends Entry<? extends Option>>> type : SUGGESTION_TYPES) {
Suggestion suggestion = createTestItem(type);
XContentType xContentType = randomFrom(XContentType.values());
boolean humanReadable = randomBoolean();
BytesReference originalBytes = toShuffledXContent(suggestion, xContentType, params, humanReadable);
BytesReference mutated;
if (addRandomFields) {
// - "contexts" is an object consisting of key/array pairs, we shouldn't add anything random there
// - there can be inner search hits fields inside this option where we cannot add random stuff
// - the root object should be excluded since it contains the named suggestion arrays
// We also exclude options that contain SearchHits, as all unknown fields
// on a root level of SearchHit are interpreted as meta-fields and will be kept.
Predicate<String> excludeFilter = path -> path.isEmpty() || path.endsWith(CompletionSuggestion.Entry.Option.CONTEXTS.getPreferredName()) || path.endsWith("highlight") || path.contains("fields") || path.contains("_source") || path.contains("inner_hits") || path.contains("options");
mutated = insertRandomFields(xContentType, originalBytes, excludeFilter, random());
} else {
mutated = originalBytes;
}
Suggestion parsed;
try (XContentParser parser = createParser(xContentType.xContent(), mutated)) {
ensureExpectedToken(XContentParser.Token.START_OBJECT, parser.nextToken(), parser);
ensureExpectedToken(XContentParser.Token.FIELD_NAME, parser.nextToken(), parser);
ensureExpectedToken(XContentParser.Token.START_ARRAY, parser.nextToken(), parser);
parsed = Suggestion.fromXContent(parser);
assertEquals(XContentParser.Token.END_OBJECT, parser.nextToken());
assertNull(parser.nextToken());
}
assertEquals(suggestion.getName(), parsed.getName());
assertEquals(suggestion.getEntries().size(), parsed.getEntries().size());
// We don't parse size via xContent, instead we set it to -1 on the client side
assertEquals(-1, parsed.getSize());
assertToXContentEquivalent(originalBytes, toXContent(parsed, xContentType, params, humanReadable), xContentType);
}
}
use of org.opensearch.search.suggest.Suggest.Suggestion.Entry in project OpenSearch by opensearch-project.
the class SuggestionTests method createTestItem.
@SuppressWarnings({ "unchecked", "rawtypes" })
public static Suggestion<? extends Entry<? extends Option>> createTestItem(Class<? extends Suggestion> type) {
String name = randomAlphaOfLengthBetween(5, 10);
// note: size will not be rendered via "toXContent", only passed on internally on transport layer
int size = randomInt();
Supplier<Entry> entrySupplier;
Suggestion suggestion;
if (type == TermSuggestion.class) {
suggestion = new TermSuggestion(name, size, randomFrom(SortBy.values()));
entrySupplier = () -> SuggestionEntryTests.createTestItem(TermSuggestion.Entry.class);
} else if (type == PhraseSuggestion.class) {
suggestion = new PhraseSuggestion(name, size);
entrySupplier = () -> SuggestionEntryTests.createTestItem(PhraseSuggestion.Entry.class);
} else if (type == CompletionSuggestion.class) {
suggestion = new CompletionSuggestion(name, size, randomBoolean());
entrySupplier = () -> SuggestionEntryTests.createTestItem(CompletionSuggestion.Entry.class);
} else {
throw new UnsupportedOperationException("type not supported [" + type + "]");
}
int numEntries;
if (frequently()) {
if (type == CompletionSuggestion.class) {
// CompletionSuggestion can have max. one entry
numEntries = 1;
} else {
numEntries = randomIntBetween(1, 5);
}
} else {
// also occasionally test zero entries
numEntries = 0;
}
for (int i = 0; i < numEntries; i++) {
suggestion.addTerm(entrySupplier.get());
}
return suggestion;
}
use of org.opensearch.search.suggest.Suggest.Suggestion.Entry in project OpenSearch by opensearch-project.
the class PhraseSuggester method innerExecute.
/*
* More Ideas:
* - add ability to find whitespace problems -> we can build a poor mans decompounder with our index based on a automaton?
* - add ability to build different error models maybe based on a confusion matrix?
* - try to combine a token with its subsequent token to find / detect word splits (optional)
* - for this to work we need some way to defined the position length of a candidate
* - phonetic filters could be interesting here too for candidate selection
*/
@Override
public Suggestion<? extends Entry<? extends Option>> innerExecute(String name, PhraseSuggestionContext suggestion, IndexSearcher searcher, CharsRefBuilder spare) throws IOException {
double realWordErrorLikelihood = suggestion.realworldErrorLikelihood();
final PhraseSuggestion response = new PhraseSuggestion(name, suggestion.getSize());
final IndexReader indexReader = searcher.getIndexReader();
List<PhraseSuggestionContext.DirectCandidateGenerator> generators = suggestion.generators();
final int numGenerators = generators.size();
final List<CandidateGenerator> gens = new ArrayList<>(generators.size());
for (int i = 0; i < numGenerators; i++) {
PhraseSuggestionContext.DirectCandidateGenerator generator = generators.get(i);
DirectSpellChecker directSpellChecker = generator.createDirectSpellChecker();
Terms terms = MultiTerms.getTerms(indexReader, generator.field());
if (terms != null) {
gens.add(new DirectCandidateGenerator(directSpellChecker, generator.field(), generator.suggestMode(), indexReader, realWordErrorLikelihood, generator.size(), generator.preFilter(), generator.postFilter(), terms));
}
}
final String suggestField = suggestion.getField();
final Terms suggestTerms = MultiTerms.getTerms(indexReader, suggestField);
if (gens.size() > 0 && suggestTerms != null) {
final NoisyChannelSpellChecker checker = new NoisyChannelSpellChecker(realWordErrorLikelihood, suggestion.getRequireUnigram(), suggestion.getTokenLimit());
final BytesRef separator = suggestion.separator();
WordScorer wordScorer = suggestion.model().newScorer(indexReader, suggestTerms, suggestField, realWordErrorLikelihood, separator);
Result checkerResult;
try (TokenStream stream = tokenStream(suggestion.getAnalyzer(), suggestion.getText(), spare, suggestion.getField())) {
checkerResult = checker.getCorrections(stream, new MultiCandidateGeneratorWrapper(suggestion.getShardSize(), gens.toArray(new CandidateGenerator[gens.size()])), suggestion.maxErrors(), suggestion.getShardSize(), wordScorer, suggestion.confidence(), suggestion.gramSize());
}
PhraseSuggestion.Entry resultEntry = buildResultEntry(suggestion, spare, checkerResult.cutoffScore);
response.addTerm(resultEntry);
final BytesRefBuilder byteSpare = new BytesRefBuilder();
final TemplateScript.Factory scriptFactory = suggestion.getCollateQueryScript();
final boolean collatePrune = (scriptFactory != null) && suggestion.collatePrune();
for (int i = 0; i < checkerResult.corrections.length; i++) {
Correction correction = checkerResult.corrections[i];
spare.copyUTF8Bytes(correction.join(SEPARATOR, byteSpare, null, null));
boolean collateMatch = true;
if (scriptFactory != null) {
// Checks if the template query collateScript yields any documents
// from the index for a correction, collateMatch is updated
final Map<String, Object> vars = suggestion.getCollateScriptParams();
vars.put(SUGGESTION_TEMPLATE_VAR_NAME, spare.toString());
QueryShardContext shardContext = suggestion.getShardContext();
final String querySource = scriptFactory.newInstance(vars).execute();
try (XContentParser parser = XContentFactory.xContent(querySource).createParser(shardContext.getXContentRegistry(), LoggingDeprecationHandler.INSTANCE, querySource)) {
QueryBuilder innerQueryBuilder = AbstractQueryBuilder.parseInnerQueryBuilder(parser);
final ParsedQuery parsedQuery = shardContext.toQuery(innerQueryBuilder);
collateMatch = Lucene.exists(searcher, parsedQuery.query());
}
}
if (!collateMatch && !collatePrune) {
continue;
}
Text phrase = new Text(spare.toString());
Text highlighted = null;
if (suggestion.getPreTag() != null) {
spare.copyUTF8Bytes(correction.join(SEPARATOR, byteSpare, suggestion.getPreTag(), suggestion.getPostTag()));
highlighted = new Text(spare.toString());
}
if (collatePrune) {
resultEntry.addOption(new PhraseSuggestion.Entry.Option(phrase, highlighted, (float) (correction.score), collateMatch));
} else {
resultEntry.addOption(new PhraseSuggestion.Entry.Option(phrase, highlighted, (float) (correction.score)));
}
}
} else {
response.addTerm(buildResultEntry(suggestion, spare, Double.MIN_VALUE));
}
return response;
}
use of org.opensearch.search.suggest.Suggest.Suggestion.Entry in project OpenSearch by opensearch-project.
the class SuggestPhase method execute.
public void execute(SearchContext context) {
final SuggestionSearchContext suggest = context.suggest();
if (suggest == null) {
return;
}
try {
CharsRefBuilder spare = new CharsRefBuilder();
final List<Suggestion<? extends Entry<? extends Option>>> suggestions = new ArrayList<>(suggest.suggestions().size());
for (Map.Entry<String, SuggestionSearchContext.SuggestionContext> entry : suggest.suggestions().entrySet()) {
SuggestionSearchContext.SuggestionContext suggestion = entry.getValue();
Suggester<SuggestionContext> suggester = suggestion.getSuggester();
Suggestion<? extends Entry<? extends Option>> result = suggester.execute(entry.getKey(), suggestion, context.searcher(), spare);
if (result != null) {
assert entry.getKey().equals(result.name);
suggestions.add(result);
}
}
context.queryResult().suggest(new Suggest(suggestions));
} catch (IOException e) {
throw new OpenSearchException("I/O exception during suggest phase", e);
}
}
Aggregations