public void testSimpleSkip() throws IOException {
Directory dir = new CountingRAMDirectory(new RAMDirectory());
IndexWriter writer =
new IndexWriter(
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new PayloadAnalyzer())
.setCodec(_TestUtil.alwaysPostingsFormat(new Lucene40PostingsFormat()))
.setMergePolicy(newLogMergePolicy()));
Term term = new Term("test", "a");
for (int i = 0; i < 5000; i++) {
Document d1 = new Document();
d1.add(newTextField(term.field(), term.text(), Field.Store.NO));
writer.addDocument(d1);
}
writer.commit();
writer.forceMerge(1);
writer.close();
AtomicReader reader = getOnlySegmentReader(DirectoryReader.open(dir));
for (int i = 0; i < 2; i++) {
counter = 0;
DocsAndPositionsEnum tp = reader.termPositionsEnum(term);
checkSkipTo(tp, 14, 185); // no skips
checkSkipTo(tp, 17, 190); // one skip on level 0
checkSkipTo(tp, 287, 200); // one skip on level 1, two on level 0
// this test would fail if we had only one skip level,
// because than more bytes would be read from the freqStream
checkSkipTo(tp, 4800, 250); // one skip on level 2
}
}
public void testDocsWithField() throws Exception {
Directory dir = newDirectory();
IndexWriterConfig conf = newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()));
IndexWriter writer = new IndexWriter(dir, conf);
Document doc = new Document();
doc.add(new NumericDocValuesField("dv", 0L));
writer.addDocument(doc);
doc = new Document();
doc.add(new TextField("dv", "some text", Field.Store.NO));
doc.add(new NumericDocValuesField("dv", 0L));
writer.addDocument(doc);
DirectoryReader r = writer.getReader();
writer.close();
AtomicReader subR = r.leaves().get(0).reader();
assertEquals(2, subR.numDocs());
Bits bits = FieldCache.DEFAULT.getDocsWithField(subR, "dv");
assertTrue(bits.get(0));
assertTrue(bits.get(1));
r.close();
dir.close();
}
/** Creates a {@link DocMap} instance appropriate for this reader. */
public static DocMap build(AtomicReader reader) {
final int maxDoc = reader.maxDoc();
if (!reader.hasDeletions()) {
return new NoDelDocMap(maxDoc);
}
final Bits liveDocs = reader.getLiveDocs();
return build(maxDoc, liveDocs);
}
protected AtomicReaderContext refreshReader() throws Exception {
if (readerContext != null) {
readerContext.reader().close();
}
AtomicReader reader =
SlowCompositeReaderWrapper.wrap(topLevelReader = DirectoryReader.open(writer, true));
readerContext = reader.getContext();
return readerContext;
}
// indexes Integer.MAX_VALUE docs with a fixed binary field
public void testFixedSorted() throws Exception {
BaseDirectoryWrapper dir = newFSDirectory(_TestUtil.getTempDir("2BFixedSorted"));
if (dir instanceof MockDirectoryWrapper) {
((MockDirectoryWrapper) dir).setThrottling(MockDirectoryWrapper.Throttling.NEVER);
}
IndexWriter w =
new IndexWriter(
dir,
new IndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setMaxBufferedDocs(IndexWriterConfig.DISABLE_AUTO_FLUSH)
.setRAMBufferSizeMB(256.0)
.setMergeScheduler(new ConcurrentMergeScheduler())
.setMergePolicy(newLogMergePolicy(false, 10))
.setOpenMode(IndexWriterConfig.OpenMode.CREATE));
Document doc = new Document();
byte bytes[] = new byte[2];
BytesRef data = new BytesRef(bytes);
SortedDocValuesField dvField = new SortedDocValuesField("dv", data);
doc.add(dvField);
for (int i = 0; i < Integer.MAX_VALUE; i++) {
bytes[0] = (byte) (i >> 8);
bytes[1] = (byte) i;
w.addDocument(doc);
if (i % 100000 == 0) {
System.out.println("indexed: " + i);
System.out.flush();
}
}
w.forceMerge(1);
w.close();
System.out.println("verifying...");
System.out.flush();
DirectoryReader r = DirectoryReader.open(dir);
int expectedValue = 0;
for (AtomicReaderContext context : r.leaves()) {
AtomicReader reader = context.reader();
BytesRef scratch = new BytesRef();
BinaryDocValues dv = reader.getSortedDocValues("dv");
for (int i = 0; i < reader.maxDoc(); i++) {
bytes[0] = (byte) (expectedValue >> 8);
bytes[1] = (byte) expectedValue;
dv.get(i, scratch);
assertEquals(data, scratch);
expectedValue++;
}
}
r.close();
dir.close();
}
@Test
public void testNoTerms() throws Exception {
FieldDataTermsFilter hFilterBytes =
FieldDataTermsFilter.newBytes(getFieldData(strMapper), new ObjectOpenHashSet<BytesRef>());
FieldDataTermsFilter hFilterLongs =
FieldDataTermsFilter.newLongs(getFieldData(lngMapper), new LongOpenHashSet());
FieldDataTermsFilter hFilterDoubles =
FieldDataTermsFilter.newDoubles(getFieldData(dblMapper), new DoubleOpenHashSet());
assertNull(hFilterBytes.getDocIdSet(reader.getContext(), reader.getLiveDocs()));
assertNull(hFilterLongs.getDocIdSet(reader.getContext(), reader.getLiveDocs()));
assertNull(hFilterDoubles.getDocIdSet(reader.getContext(), reader.getLiveDocs()));
}
/**
* Expert: Get the list of readers to merge. Note that this list does not necessarily match the
* list of segments to merge and should only be used to feed SegmentMerger to initialize a
* merge. When a {@link OneMerge} reorders doc IDs, it must override {@link #getDocMap} too so
* that deletes that happened during the merge can be applied to the newly merged segment.
*/
public List<AtomicReader> getMergeReaders() throws IOException {
if (readers == null) {
throw new IllegalStateException(
"IndexWriter has not initialized readers from the segment infos yet");
}
final List<AtomicReader> readers = new ArrayList<AtomicReader>(this.readers.size());
for (AtomicReader reader : this.readers) {
if (reader.numDocs() > 0) {
readers.add(reader);
}
}
return Collections.unmodifiableList(readers);
}
private static void printDocs(DirectoryReader r) throws Throwable {
for (AtomicReaderContext ctx : r.leaves()) {
// TODO: improve this
AtomicReader sub = ctx.reader();
Bits liveDocs = sub.getLiveDocs();
System.out.println(" " + ((SegmentReader) sub).getSegmentInfo());
for (int docID = 0; docID < sub.maxDoc(); docID++) {
StoredDocument doc = sub.document(docID);
if (liveDocs == null || liveDocs.get(docID)) {
System.out.println(" docID=" + docID + " id:" + doc.get("id"));
} else {
System.out.println(" DEL docID=" + docID + " id:" + doc.get("id"));
}
}
}
}
public void testSimple() throws Exception {
Directory dir = newDirectory();
final RandomIndexWriter w =
new RandomIndexWriter(
random(),
dir,
newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random()))
.setMergePolicy(newLogMergePolicy()));
Document doc = new Document();
Field field = newTextField("field", "", Field.Store.NO);
doc.add(field);
field.setStringValue("a b c");
w.addDocument(doc);
field.setStringValue("d e f");
w.addDocument(doc);
field.setStringValue("a f");
w.addDocument(doc);
final IndexReader r = w.getReader();
w.close();
final AtomicReader ar = SlowCompositeReaderWrapper.wrap(r);
final DocTermOrds dto = new DocTermOrds(ar, ar.getLiveDocs(), "field");
SortedSetDocValues iter = dto.iterator(ar);
iter.setDocument(0);
assertEquals(0, iter.nextOrd());
assertEquals(1, iter.nextOrd());
assertEquals(2, iter.nextOrd());
assertEquals(SortedSetDocValues.NO_MORE_ORDS, iter.nextOrd());
iter.setDocument(1);
assertEquals(3, iter.nextOrd());
assertEquals(4, iter.nextOrd());
assertEquals(5, iter.nextOrd());
assertEquals(SortedSetDocValues.NO_MORE_ORDS, iter.nextOrd());
iter.setDocument(2);
assertEquals(0, iter.nextOrd());
assertEquals(5, iter.nextOrd());
assertEquals(SortedSetDocValues.NO_MORE_ORDS, iter.nextOrd());
r.close();
dir.close();
}
@After
public void tearDown() throws Exception {
super.tearDown();
reader.close();
writer.close();
ifdService.clear();
SearchContext.removeCurrent();
}
public void testAddIndexes() throws IOException {
Directory d1 = newDirectory();
RandomIndexWriter w = new RandomIndexWriter(random(), d1);
Document doc = new Document();
doc.add(newStringField("id", "1", Field.Store.YES));
doc.add(new NumericDocValuesField("dv", 1));
w.addDocument(doc);
IndexReader r1 = w.getReader();
w.close();
Directory d2 = newDirectory();
w = new RandomIndexWriter(random(), d2);
doc = new Document();
doc.add(newStringField("id", "2", Field.Store.YES));
doc.add(new NumericDocValuesField("dv", 2));
w.addDocument(doc);
IndexReader r2 = w.getReader();
w.close();
Directory d3 = newDirectory();
w = new RandomIndexWriter(random(), d3);
w.addIndexes(SlowCompositeReaderWrapper.wrap(r1), SlowCompositeReaderWrapper.wrap(r2));
r1.close();
d1.close();
r2.close();
d2.close();
w.forceMerge(1);
DirectoryReader r3 = w.getReader();
w.close();
AtomicReader sr = getOnlySegmentReader(r3);
assertEquals(2, sr.numDocs());
NumericDocValues docValues = sr.getNumericDocValues("dv");
assertNotNull(docValues);
r3.close();
d3.close();
}
@Test
public void testDoubles() throws Exception {
List<Integer> docs = Arrays.asList(1, 5, 7);
DoubleOpenHashSet hTerms = new DoubleOpenHashSet();
List<Double> cTerms = new ArrayList<Double>(docs.size());
for (int i = 0; i < docs.size(); i++) {
double term = Double.valueOf(docs.get(i));
hTerms.add(term);
cTerms.add(term);
}
FieldDataTermsFilter hFilter = FieldDataTermsFilter.newDoubles(getFieldData(dblMapper), hTerms);
int size = reader.maxDoc();
FixedBitSet result = new FixedBitSet(size);
result.clear(0, size);
assertThat(result.cardinality(), equalTo(0));
result.or(hFilter.getDocIdSet(reader.getContext(), reader.getLiveDocs()).iterator());
assertThat(result.cardinality(), equalTo(docs.size()));
for (int i = 0; i < reader.maxDoc(); i++) {
assertThat(result.get(i), equalTo(docs.contains(i)));
}
// filter from mapper
result.clear(0, size);
assertThat(result.cardinality(), equalTo(0));
result.or(
dblMapper
.termsFilter(ifdService, cTerms, null)
.getDocIdSet(reader.getContext(), reader.getLiveDocs())
.iterator());
assertThat(result.cardinality(), equalTo(docs.size()));
for (int i = 0; i < reader.maxDoc(); i++) {
assertThat(result.get(i), equalTo(docs.contains(i)));
}
hFilter = FieldDataTermsFilter.newDoubles(getFieldData(lngMapper), hTerms);
assertNull(hFilter.getDocIdSet(reader.getContext(), reader.getLiveDocs()));
}
public static TermsEnum filter(
TermsEnum toFilter, Terms terms, AtomicReader reader, Settings settings)
throws IOException {
int docCount = terms.getDocCount();
if (docCount == -1) {
docCount = reader.maxDoc();
}
final double minFrequency = settings.getAsDouble("min", 0d);
final double maxFrequency = settings.getAsDouble("max", docCount + 1d);
final double minSegmentSize = settings.getAsInt("min_segment_size", 0);
if (minSegmentSize < docCount) {
final int minFreq =
minFrequency >= 1.0 ? (int) minFrequency : (int) (docCount * minFrequency);
final int maxFreq =
maxFrequency >= 1.0 ? (int) maxFrequency : (int) (docCount * maxFrequency);
assert minFreq < maxFreq;
return new FrequencyFilter(toFilter, minFreq, maxFreq);
}
return toFilter;
}
@Override
public DoubleArrayAtomicFieldData loadDirect(AtomicReaderContext context) throws Exception {
AtomicReader reader = context.reader();
Terms terms = reader.terms(getFieldNames().indexName());
if (terms == null) {
return DoubleArrayAtomicFieldData.EMPTY;
}
// TODO: how can we guess the number of terms? numerics end up creating more terms per value...
final TDoubleArrayList values = new TDoubleArrayList();
ArrayList<int[]> ordinals = new ArrayList<int[]>();
int[] idx = new int[reader.maxDoc()];
ordinals.add(new int[reader.maxDoc()]);
values.add(0); // first "t" indicates null value
int termOrd = 1; // current term number
TermsEnum termsEnum = terms.iterator(null);
try {
DocsEnum docsEnum = null;
for (BytesRef term = termsEnum.next(); term != null; term = termsEnum.next()) {
values.add(FieldCache.NUMERIC_UTILS_DOUBLE_PARSER.parseDouble(term));
docsEnum = termsEnum.docs(reader.getLiveDocs(), docsEnum, 0);
for (int docId = docsEnum.nextDoc();
docId != DocsEnum.NO_MORE_DOCS;
docId = docsEnum.nextDoc()) {
int[] ordinal;
if (idx[docId] >= ordinals.size()) {
ordinal = new int[reader.maxDoc()];
ordinals.add(ordinal);
} else {
ordinal = ordinals.get(idx[docId]);
}
ordinal[docId] = termOrd;
idx[docId]++;
}
termOrd++;
}
} catch (RuntimeException e) {
if (e.getClass().getName().endsWith("StopFillCacheException")) {
// all is well, in case numeric parsers are used.
} else {
throw e;
}
}
if (ordinals.size() == 1) {
int[] nativeOrdinals = ordinals.get(0);
FixedBitSet set = new FixedBitSet(reader.maxDoc());
double[] sValues = new double[reader.maxDoc()];
boolean allHaveValue = true;
for (int i = 0; i < nativeOrdinals.length; i++) {
int nativeOrdinal = nativeOrdinals[i];
if (nativeOrdinal == 0) {
allHaveValue = false;
} else {
set.set(i);
sValues[i] = values.get(nativeOrdinal);
}
}
if (allHaveValue) {
return new DoubleArrayAtomicFieldData.Single(sValues, reader.maxDoc());
} else {
return new DoubleArrayAtomicFieldData.SingleFixedSet(sValues, reader.maxDoc(), set);
}
} else {
int[][] nativeOrdinals = new int[ordinals.size()][];
for (int i = 0; i < nativeOrdinals.length; i++) {
nativeOrdinals[i] = ordinals.get(i);
}
return new DoubleArrayAtomicFieldData.WithOrdinals(
values.toArray(new double[values.size()]),
reader.maxDoc(),
Ordinals.Factories.createFromFlatOrdinals(
nativeOrdinals, termOrd, fieldDataType.getSettings()));
}
}
@Test
public void testBytes() throws Exception {
List<Integer> docs = Arrays.asList(1, 5, 7);
ObjectOpenHashSet<BytesRef> hTerms = new ObjectOpenHashSet<BytesRef>();
List<BytesRef> cTerms = new ArrayList<BytesRef>(docs.size());
for (int i = 0; i < docs.size(); i++) {
BytesRef term = new BytesRef("str" + docs.get(i));
hTerms.add(term);
cTerms.add(term);
}
FieldDataTermsFilter hFilter = FieldDataTermsFilter.newBytes(getFieldData(strMapper), hTerms);
int size = reader.maxDoc();
FixedBitSet result = new FixedBitSet(size);
result.clear(0, size);
assertThat(result.cardinality(), equalTo(0));
result.or(hFilter.getDocIdSet(reader.getContext(), reader.getLiveDocs()).iterator());
assertThat(result.cardinality(), equalTo(docs.size()));
for (int i = 0; i < reader.maxDoc(); i++) {
assertThat(result.get(i), equalTo(docs.contains(i)));
}
// filter from mapper
result.clear(0, size);
assertThat(result.cardinality(), equalTo(0));
result.or(
strMapper
.termsFilter(ifdService, cTerms, null)
.getDocIdSet(reader.getContext(), reader.getLiveDocs())
.iterator());
assertThat(result.cardinality(), equalTo(docs.size()));
for (int i = 0; i < reader.maxDoc(); i++) {
assertThat(result.get(i), equalTo(docs.contains(i)));
}
result.clear(0, size);
assertThat(result.cardinality(), equalTo(0));
// filter on a numeric field using BytesRef terms
// should not match any docs
hFilter = FieldDataTermsFilter.newBytes(getFieldData(lngMapper), hTerms);
result.or(hFilter.getDocIdSet(reader.getContext(), reader.getLiveDocs()).iterator());
assertThat(result.cardinality(), equalTo(0));
// filter on a numeric field using BytesRef terms
// should not match any docs
hFilter = FieldDataTermsFilter.newBytes(getFieldData(dblMapper), hTerms);
result.or(hFilter.getDocIdSet(reader.getContext(), reader.getLiveDocs()).iterator());
assertThat(result.cardinality(), equalTo(0));
}
private void verify(AtomicReader r, int[][] idToOrds, BytesRef[] termsArray, BytesRef prefixRef)
throws Exception {
final DocTermOrds dto =
new DocTermOrds(
r,
r.getLiveDocs(),
"field",
prefixRef,
Integer.MAX_VALUE,
_TestUtil.nextInt(random(), 2, 10));
final FieldCache.Ints docIDToID = FieldCache.DEFAULT.getInts(r, "id", false);
/*
for(int docID=0;docID<subR.maxDoc();docID++) {
System.out.println(" docID=" + docID + " id=" + docIDToID[docID]);
}
*/
if (VERBOSE) {
System.out.println(
"TEST: verify prefix=" + (prefixRef == null ? "null" : prefixRef.utf8ToString()));
System.out.println("TEST: all TERMS:");
TermsEnum allTE = MultiFields.getTerms(r, "field").iterator(null);
int ord = 0;
while (allTE.next() != null) {
System.out.println(" ord=" + (ord++) + " term=" + allTE.term().utf8ToString());
}
}
// final TermsEnum te = subR.fields().terms("field").iterator();
final TermsEnum te = dto.getOrdTermsEnum(r);
if (dto.numTerms() == 0) {
if (prefixRef == null) {
assertNull(MultiFields.getTerms(r, "field"));
} else {
Terms terms = MultiFields.getTerms(r, "field");
if (terms != null) {
TermsEnum termsEnum = terms.iterator(null);
TermsEnum.SeekStatus result = termsEnum.seekCeil(prefixRef, false);
if (result != TermsEnum.SeekStatus.END) {
assertFalse(
"term="
+ termsEnum.term().utf8ToString()
+ " matches prefix="
+ prefixRef.utf8ToString(),
StringHelper.startsWith(termsEnum.term(), prefixRef));
} else {
// ok
}
} else {
// ok
}
}
return;
}
if (VERBOSE) {
System.out.println("TEST: TERMS:");
te.seekExact(0);
while (true) {
System.out.println(" ord=" + te.ord() + " term=" + te.term().utf8ToString());
if (te.next() == null) {
break;
}
}
}
SortedSetDocValues iter = dto.iterator(r);
for (int docID = 0; docID < r.maxDoc(); docID++) {
if (VERBOSE) {
System.out.println(
"TEST: docID=" + docID + " of " + r.maxDoc() + " (id=" + docIDToID.get(docID) + ")");
}
iter.setDocument(docID);
final int[] answers = idToOrds[docIDToID.get(docID)];
int upto = 0;
long ord;
while ((ord = iter.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
te.seekExact(ord);
final BytesRef expected = termsArray[answers[upto++]];
if (VERBOSE) {
System.out.println(
" exp=" + expected.utf8ToString() + " actual=" + te.term().utf8ToString());
}
assertEquals(
"expected="
+ expected.utf8ToString()
+ " actual="
+ te.term().utf8ToString()
+ " ord="
+ ord,
expected,
te.term());
}
assertEquals(answers.length, upto);
}
}
@Override
public void process(ResponseBuilder rb) throws IOException {
SolrParams params = rb.req.getParams();
if (!params.getBool(TermsParams.TERMS, false)) return;
String[] fields = params.getParams(TermsParams.TERMS_FIELD);
NamedList<Object> termsResult = new SimpleOrderedMap<>();
rb.rsp.add("terms", termsResult);
if (fields == null || fields.length == 0) return;
int limit = params.getInt(TermsParams.TERMS_LIMIT, 10);
if (limit < 0) {
limit = Integer.MAX_VALUE;
}
String lowerStr = params.get(TermsParams.TERMS_LOWER);
String upperStr = params.get(TermsParams.TERMS_UPPER);
boolean upperIncl = params.getBool(TermsParams.TERMS_UPPER_INCLUSIVE, false);
boolean lowerIncl = params.getBool(TermsParams.TERMS_LOWER_INCLUSIVE, true);
boolean sort =
!TermsParams.TERMS_SORT_INDEX.equals(
params.get(TermsParams.TERMS_SORT, TermsParams.TERMS_SORT_COUNT));
int freqmin = params.getInt(TermsParams.TERMS_MINCOUNT, 1);
int freqmax = params.getInt(TermsParams.TERMS_MAXCOUNT, UNLIMITED_MAX_COUNT);
if (freqmax < 0) {
freqmax = Integer.MAX_VALUE;
}
String prefix = params.get(TermsParams.TERMS_PREFIX_STR);
String regexp = params.get(TermsParams.TERMS_REGEXP_STR);
Pattern pattern = regexp != null ? Pattern.compile(regexp, resolveRegexpFlags(params)) : null;
boolean raw = params.getBool(TermsParams.TERMS_RAW, false);
final AtomicReader indexReader = rb.req.getSearcher().getAtomicReader();
Fields lfields = indexReader.fields();
for (String field : fields) {
NamedList<Integer> fieldTerms = new NamedList<>();
termsResult.add(field, fieldTerms);
Terms terms = lfields == null ? null : lfields.terms(field);
if (terms == null) {
// no terms for this field
continue;
}
FieldType ft = raw ? null : rb.req.getSchema().getFieldTypeNoEx(field);
if (ft == null) ft = new StrField();
// prefix must currently be text
BytesRef prefixBytes = prefix == null ? null : new BytesRef(prefix);
BytesRef upperBytes = null;
if (upperStr != null) {
upperBytes = new BytesRef();
ft.readableToIndexed(upperStr, upperBytes);
}
BytesRef lowerBytes;
if (lowerStr == null) {
// If no lower bound was specified, use the prefix
lowerBytes = prefixBytes;
} else {
lowerBytes = new BytesRef();
if (raw) {
// TODO: how to handle binary? perhaps we don't for "raw"... or if the field exists
// perhaps we detect if the FieldType is non-character and expect hex if so?
lowerBytes = new BytesRef(lowerStr);
} else {
lowerBytes = new BytesRef();
ft.readableToIndexed(lowerStr, lowerBytes);
}
}
TermsEnum termsEnum = terms.iterator(null);
BytesRef term = null;
if (lowerBytes != null) {
if (termsEnum.seekCeil(lowerBytes) == TermsEnum.SeekStatus.END) {
termsEnum = null;
} else {
term = termsEnum.term();
// Only advance the enum if we are excluding the lower bound and the lower Term actually
// matches
if (lowerIncl == false && term.equals(lowerBytes)) {
term = termsEnum.next();
}
}
} else {
// position termsEnum on first term
term = termsEnum.next();
}
int i = 0;
BoundedTreeSet<CountPair<BytesRef, Integer>> queue =
(sort ? new BoundedTreeSet<CountPair<BytesRef, Integer>>(limit) : null);
CharsRef external = new CharsRef();
while (term != null && (i < limit || sort)) {
boolean externalized = false; // did we fill in "external" yet for this term?
// stop if the prefix doesn't match
if (prefixBytes != null && !StringHelper.startsWith(term, prefixBytes)) break;
if (pattern != null) {
// indexed text or external text?
// TODO: support "raw" mode?
ft.indexedToReadable(term, external);
externalized = true;
if (!pattern.matcher(external).matches()) {
term = termsEnum.next();
continue;
}
}
if (upperBytes != null) {
int upperCmp = term.compareTo(upperBytes);
// if we are past the upper term, or equal to it (when don't include upper) then stop.
if (upperCmp > 0 || (upperCmp == 0 && !upperIncl)) break;
}
// This is a good term in the range. Check if mincount/maxcount conditions are satisfied.
int docFreq = termsEnum.docFreq();
if (docFreq >= freqmin && docFreq <= freqmax) {
// add the term to the list
if (sort) {
queue.add(new CountPair<>(BytesRef.deepCopyOf(term), docFreq));
} else {
// TODO: handle raw somehow
if (!externalized) {
ft.indexedToReadable(term, external);
}
fieldTerms.add(external.toString(), docFreq);
i++;
}
}
term = termsEnum.next();
}
if (sort) {
for (CountPair<BytesRef, Integer> item : queue) {
if (i >= limit) break;
ft.indexedToReadable(item.key, external);
fieldTerms.add(external.toString(), item.val);
i++;
}
}
}
}
public void testNumericField() throws Exception {
Directory dir = newDirectory();
RandomIndexWriter w = new RandomIndexWriter(random(), dir);
final int numDocs = atLeast(500);
final Number[] answers = new Number[numDocs];
final NumericType[] typeAnswers = new NumericType[numDocs];
for (int id = 0; id < numDocs; id++) {
Document doc = new Document();
final Field nf;
final Field sf;
final Number answer;
final NumericType typeAnswer;
if (random().nextBoolean()) {
// float/double
if (random().nextBoolean()) {
final float f = random().nextFloat();
answer = Float.valueOf(f);
nf = new FloatField("nf", f, Field.Store.NO);
sf = new StoredField("nf", f);
typeAnswer = NumericType.FLOAT;
} else {
final double d = random().nextDouble();
answer = Double.valueOf(d);
nf = new DoubleField("nf", d, Field.Store.NO);
sf = new StoredField("nf", d);
typeAnswer = NumericType.DOUBLE;
}
} else {
// int/long
if (random().nextBoolean()) {
final int i = random().nextInt();
answer = Integer.valueOf(i);
nf = new IntField("nf", i, Field.Store.NO);
sf = new StoredField("nf", i);
typeAnswer = NumericType.INT;
} else {
final long l = random().nextLong();
answer = Long.valueOf(l);
nf = new LongField("nf", l, Field.Store.NO);
sf = new StoredField("nf", l);
typeAnswer = NumericType.LONG;
}
}
doc.add(nf);
doc.add(sf);
answers[id] = answer;
typeAnswers[id] = typeAnswer;
FieldType ft = new FieldType(IntField.TYPE_STORED);
ft.setNumericPrecisionStep(Integer.MAX_VALUE);
doc.add(new IntField("id", id, ft));
w.addDocument(doc);
}
final DirectoryReader r = w.getReader();
w.close();
assertEquals(numDocs, r.numDocs());
for (AtomicReaderContext ctx : r.leaves()) {
final AtomicReader sub = ctx.reader();
final FieldCache.Ints ids = FieldCache.DEFAULT.getInts(sub, "id", false);
for (int docID = 0; docID < sub.numDocs(); docID++) {
final Document doc = sub.document(docID);
final Field f = (Field) doc.getField("nf");
assertTrue("got f=" + f, f instanceof StoredField);
assertEquals(answers[ids.get(docID)], f.numericValue());
}
}
r.close();
dir.close();
}
int resolveParentDocuments(
TopDocs topDocs,
SearchContext context,
Recycler.V<ObjectObjectOpenHashMap<Object, ParentDoc[]>> parentDocs) {
int parentHitsResolved = 0;
Recycler.V<ObjectObjectOpenHashMap<Object, Recycler.V<IntObjectOpenHashMap<ParentDoc>>>>
parentDocsPerReader =
cacheRecycler.hashMap(context.searcher().getIndexReader().leaves().size());
for (ScoreDoc scoreDoc : topDocs.scoreDocs) {
int readerIndex =
ReaderUtil.subIndex(scoreDoc.doc, context.searcher().getIndexReader().leaves());
AtomicReaderContext subContext =
context.searcher().getIndexReader().leaves().get(readerIndex);
int subDoc = scoreDoc.doc - subContext.docBase;
// find the parent id
HashedBytesArray parentId =
context.idCache().reader(subContext.reader()).parentIdByDoc(parentType, subDoc);
if (parentId == null) {
// no parent found
continue;
}
// now go over and find the parent doc Id and reader tuple
for (AtomicReaderContext atomicReaderContext : context.searcher().getIndexReader().leaves()) {
AtomicReader indexReader = atomicReaderContext.reader();
int parentDocId = context.idCache().reader(indexReader).docById(parentType, parentId);
Bits liveDocs = indexReader.getLiveDocs();
if (parentDocId != -1 && (liveDocs == null || liveDocs.get(parentDocId))) {
// we found a match, add it and break
Recycler.V<IntObjectOpenHashMap<ParentDoc>> readerParentDocs =
parentDocsPerReader.v().get(indexReader.getCoreCacheKey());
if (readerParentDocs == null) {
readerParentDocs = cacheRecycler.intObjectMap(indexReader.maxDoc());
parentDocsPerReader.v().put(indexReader.getCoreCacheKey(), readerParentDocs);
}
ParentDoc parentDoc = readerParentDocs.v().get(parentDocId);
if (parentDoc == null) {
parentHitsResolved++; // we have a hit on a parent
parentDoc = new ParentDoc();
parentDoc.docId = parentDocId;
parentDoc.count = 1;
parentDoc.maxScore = scoreDoc.score;
parentDoc.sumScores = scoreDoc.score;
readerParentDocs.v().put(parentDocId, parentDoc);
} else {
parentDoc.count++;
parentDoc.sumScores += scoreDoc.score;
if (scoreDoc.score > parentDoc.maxScore) {
parentDoc.maxScore = scoreDoc.score;
}
}
}
}
}
boolean[] states = parentDocsPerReader.v().allocated;
Object[] keys = parentDocsPerReader.v().keys;
Object[] values = parentDocsPerReader.v().values;
for (int i = 0; i < states.length; i++) {
if (states[i]) {
Recycler.V<IntObjectOpenHashMap<ParentDoc>> value =
(Recycler.V<IntObjectOpenHashMap<ParentDoc>>) values[i];
ParentDoc[] _parentDocs = value.v().values().toArray(ParentDoc.class);
Arrays.sort(_parentDocs, PARENT_DOC_COMP);
parentDocs.v().put(keys[i], _parentDocs);
Releasables.release(value);
}
}
Releasables.release(parentDocsPerReader);
return parentHitsResolved;
}
@Test
public void testRandom() throws Exception {
Directory directory = newDirectory();
RandomIndexWriter indexWriter = new RandomIndexWriter(random(), directory);
int numUniqueChildValues = 1 + random().nextInt(TEST_NIGHTLY ? 10000 : 1000);
String[] childValues = new String[numUniqueChildValues];
for (int i = 0; i < numUniqueChildValues; i++) {
childValues[i] = Integer.toString(i);
}
IntOpenHashSet filteredOrDeletedDocs = new IntOpenHashSet();
int childDocId = 0;
int numParentDocs = 1 + random().nextInt(TEST_NIGHTLY ? 20000 : 1000);
ObjectObjectOpenHashMap<String, NavigableSet<String>> childValueToParentIds =
new ObjectObjectOpenHashMap<String, NavigableSet<String>>();
for (int parentDocId = 0; parentDocId < numParentDocs; parentDocId++) {
boolean markParentAsDeleted = rarely();
boolean filterMe = rarely();
String parent = Integer.toString(parentDocId);
Document document = new Document();
document.add(
new StringField(UidFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.YES));
document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
if (markParentAsDeleted) {
filteredOrDeletedDocs.add(parentDocId);
document.add(new StringField("delete", "me", Field.Store.NO));
}
if (filterMe) {
filteredOrDeletedDocs.add(parentDocId);
document.add(new StringField("filter", "me", Field.Store.NO));
}
indexWriter.addDocument(document);
int numChildDocs;
if (rarely()) {
numChildDocs = random().nextInt(TEST_NIGHTLY ? 100 : 25);
} else {
numChildDocs = random().nextInt(TEST_NIGHTLY ? 40 : 10);
}
for (int i = 0; i < numChildDocs; i++) {
boolean markChildAsDeleted = rarely();
String childValue = childValues[random().nextInt(childValues.length)];
document = new Document();
document.add(
new StringField(
UidFieldMapper.NAME,
Uid.createUid("child", Integer.toString(childDocId)),
Field.Store.NO));
document.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
document.add(
new StringField(
ParentFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.NO));
document.add(new StringField("field1", childValue, Field.Store.NO));
if (markChildAsDeleted) {
document.add(new StringField("delete", "me", Field.Store.NO));
}
indexWriter.addDocument(document);
if (!markChildAsDeleted) {
NavigableSet<String> parentIds;
if (childValueToParentIds.containsKey(childValue)) {
parentIds = childValueToParentIds.lget();
} else {
childValueToParentIds.put(childValue, parentIds = new TreeSet<String>());
}
if (!markParentAsDeleted && !filterMe) {
parentIds.add(parent);
}
}
}
}
// Delete docs that are marked to be deleted.
indexWriter.deleteDocuments(new Term("delete", "me"));
indexWriter.commit();
IndexReader indexReader = DirectoryReader.open(directory);
IndexSearcher searcher = new IndexSearcher(indexReader);
Engine.Searcher engineSearcher =
new Engine.SimpleSearcher(ChildrenConstantScoreQueryTests.class.getSimpleName(), searcher);
((TestSearchContext) SearchContext.current())
.setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));
Filter rawParentFilter = new TermFilter(new Term(TypeFieldMapper.NAME, "parent"));
Filter rawFilterMe = new NotFilter(new TermFilter(new Term("filter", "me")));
int max = numUniqueChildValues / 4;
for (int i = 0; i < max; i++) {
// Randomly pick a cached version: there is specific logic inside ChildrenQuery that deals
// with the fact
// that deletes are applied at the top level when filters are cached.
Filter parentFilter;
if (random().nextBoolean()) {
parentFilter = SearchContext.current().filterCache().cache(rawParentFilter);
} else {
parentFilter = rawParentFilter;
}
// Using this in FQ, will invoke / test the Scorer#advance(..) and also let the Weight#scorer
// not get live docs as acceptedDocs
Filter filterMe;
if (random().nextBoolean()) {
filterMe = SearchContext.current().filterCache().cache(rawFilterMe);
} else {
filterMe = rawFilterMe;
}
// Simulate a parent update
if (random().nextBoolean()) {
int numberOfUpdates = 1 + random().nextInt(TEST_NIGHTLY ? 25 : 5);
for (int j = 0; j < numberOfUpdates; j++) {
int parentId;
do {
parentId = random().nextInt(numParentDocs);
} while (filteredOrDeletedDocs.contains(parentId));
String parentUid = Uid.createUid("parent", Integer.toString(parentId));
indexWriter.deleteDocuments(new Term(UidFieldMapper.NAME, parentUid));
Document document = new Document();
document.add(new StringField(UidFieldMapper.NAME, parentUid, Field.Store.YES));
document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
indexWriter.addDocument(document);
}
indexReader.close();
indexReader = DirectoryReader.open(indexWriter.w, true);
searcher = new IndexSearcher(indexReader);
engineSearcher =
new Engine.SimpleSearcher(
ChildrenConstantScoreQueryTests.class.getSimpleName(), searcher);
((TestSearchContext) SearchContext.current())
.setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));
}
String childValue = childValues[random().nextInt(numUniqueChildValues)];
TermQuery childQuery = new TermQuery(new Term("field1", childValue));
int shortCircuitParentDocSet = random().nextInt(numParentDocs);
Filter nonNestedDocsFilter = random().nextBoolean() ? NonNestedDocsFilter.INSTANCE : null;
Query query;
if (random().nextBoolean()) {
// Usage in HasChildQueryParser
query =
new ChildrenConstantScoreQuery(
childQuery,
"parent",
"child",
parentFilter,
shortCircuitParentDocSet,
nonNestedDocsFilter);
} else {
// Usage in HasChildFilterParser
query =
new XConstantScoreQuery(
new CustomQueryWrappingFilter(
new ChildrenConstantScoreQuery(
childQuery,
"parent",
"child",
parentFilter,
shortCircuitParentDocSet,
nonNestedDocsFilter)));
}
query = new XFilteredQuery(query, filterMe);
BitSetCollector collector = new BitSetCollector(indexReader.maxDoc());
searcher.search(query, collector);
FixedBitSet actualResult = collector.getResult();
FixedBitSet expectedResult = new FixedBitSet(indexReader.maxDoc());
if (childValueToParentIds.containsKey(childValue)) {
AtomicReader slowAtomicReader = SlowCompositeReaderWrapper.wrap(indexReader);
Terms terms = slowAtomicReader.terms(UidFieldMapper.NAME);
if (terms != null) {
NavigableSet<String> parentIds = childValueToParentIds.lget();
TermsEnum termsEnum = terms.iterator(null);
DocsEnum docsEnum = null;
for (String id : parentIds) {
TermsEnum.SeekStatus seekStatus =
termsEnum.seekCeil(Uid.createUidAsBytes("parent", id));
if (seekStatus == TermsEnum.SeekStatus.FOUND) {
docsEnum =
termsEnum.docs(slowAtomicReader.getLiveDocs(), docsEnum, DocsEnum.FLAG_NONE);
expectedResult.set(docsEnum.nextDoc());
} else if (seekStatus == TermsEnum.SeekStatus.END) {
break;
}
}
}
}
assertBitSet(actualResult, expectedResult, searcher);
}
indexWriter.close();
indexReader.close();
directory.close();
}
@Test
public void testRandom() throws Exception {
Directory directory = newDirectory();
RandomIndexWriter indexWriter = new RandomIndexWriter(random(), directory);
int numUniqueChildValues = 1 + random().nextInt(TEST_NIGHTLY ? 6000 : 600);
String[] childValues = new String[numUniqueChildValues];
for (int i = 0; i < numUniqueChildValues; i++) {
childValues[i] = Integer.toString(i);
}
int childDocId = 0;
int numParentDocs = 1 + random().nextInt(TEST_NIGHTLY ? 20000 : 1000);
ObjectObjectOpenHashMap<String, NavigableMap<String, FloatArrayList>> childValueToParentIds =
new ObjectObjectOpenHashMap<String, NavigableMap<String, FloatArrayList>>();
for (int parentDocId = 0; parentDocId < numParentDocs; parentDocId++) {
boolean markParentAsDeleted = rarely();
String parent = Integer.toString(parentDocId);
Document document = new Document();
document.add(
new StringField(UidFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.YES));
document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
if (markParentAsDeleted) {
document.add(new StringField("delete", "me", Field.Store.NO));
}
indexWriter.addDocument(document);
int numChildDocs = random().nextInt(TEST_NIGHTLY ? 100 : 25);
for (int i = 0; i < numChildDocs; i++) {
boolean markChildAsDeleted = rarely();
String childValue = childValues[random().nextInt(childValues.length)];
document = new Document();
document.add(
new StringField(
UidFieldMapper.NAME,
Uid.createUid("child", Integer.toString(childDocId)),
Field.Store.NO));
document.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
document.add(
new StringField(
ParentFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.NO));
document.add(new StringField("field1", childValue, Field.Store.NO));
if (markChildAsDeleted) {
document.add(new StringField("delete", "me", Field.Store.NO));
}
indexWriter.addDocument(document);
if (!markChildAsDeleted) {
NavigableMap<String, FloatArrayList> parentIdToChildScores;
if (childValueToParentIds.containsKey(childValue)) {
parentIdToChildScores = childValueToParentIds.lget();
} else {
childValueToParentIds.put(
childValue, parentIdToChildScores = new TreeMap<String, FloatArrayList>());
}
if (!markParentAsDeleted) {
FloatArrayList childScores = parentIdToChildScores.get(parent);
if (childScores == null) {
parentIdToChildScores.put(parent, childScores = new FloatArrayList());
}
childScores.add(1f);
}
}
}
}
// Delete docs that are marked to be deleted.
indexWriter.deleteDocuments(new Term("delete", "me"));
indexWriter.close();
IndexReader indexReader = DirectoryReader.open(directory);
IndexSearcher searcher = new IndexSearcher(indexReader);
Engine.Searcher engineSearcher =
new Engine.SimpleSearcher(ChildrenQueryTests.class.getSimpleName(), searcher);
((TestSearchContext) SearchContext.current())
.setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));
TermFilter parentFilter = new TermFilter(new Term(TypeFieldMapper.NAME, "parent"));
for (String childValue : childValues) {
Query childQuery = new ConstantScoreQuery(new TermQuery(new Term("field1", childValue)));
int shortCircuitParentDocSet = random().nextInt(numParentDocs);
ScoreType scoreType = ScoreType.values()[random().nextInt(ScoreType.values().length)];
Query query =
new ChildrenQuery(
"parent", "child", parentFilter, childQuery, scoreType, shortCircuitParentDocSet);
BitSetCollector collector = new BitSetCollector(indexReader.maxDoc());
int numHits = 1 + random().nextInt(25);
TopScoreDocCollector actualTopDocsCollector = TopScoreDocCollector.create(numHits, false);
searcher.search(query, MultiCollector.wrap(collector, actualTopDocsCollector));
FixedBitSet actualResult = collector.getResult();
FixedBitSet expectedResult = new FixedBitSet(indexReader.maxDoc());
MockScorer mockScorer = new MockScorer(scoreType);
TopScoreDocCollector expectedTopDocsCollector = TopScoreDocCollector.create(numHits, false);
expectedTopDocsCollector.setScorer(mockScorer);
if (childValueToParentIds.containsKey(childValue)) {
AtomicReader slowAtomicReader = SlowCompositeReaderWrapper.wrap(indexReader);
Terms terms = slowAtomicReader.terms(UidFieldMapper.NAME);
if (terms != null) {
NavigableMap<String, FloatArrayList> parentIdToChildScores = childValueToParentIds.lget();
TermsEnum termsEnum = terms.iterator(null);
DocsEnum docsEnum = null;
for (Map.Entry<String, FloatArrayList> entry : parentIdToChildScores.entrySet()) {
TermsEnum.SeekStatus seekStatus =
termsEnum.seekCeil(Uid.createUidAsBytes("parent", entry.getKey()));
if (seekStatus == TermsEnum.SeekStatus.FOUND) {
docsEnum =
termsEnum.docs(slowAtomicReader.getLiveDocs(), docsEnum, DocsEnum.FLAG_NONE);
expectedResult.set(docsEnum.nextDoc());
mockScorer.scores = entry.getValue();
expectedTopDocsCollector.collect(docsEnum.docID());
} else if (seekStatus == TermsEnum.SeekStatus.END) {
break;
}
}
}
}
assertBitSet(actualResult, expectedResult, searcher);
assertTopDocs(actualTopDocsCollector.topDocs(), expectedTopDocsCollector.topDocs());
}
indexReader.close();
directory.close();
}
