public DocumentFilteredLeafIndexReader(
LeafReaderContext context, Filter preserveFilter, boolean negateFilter) throws IOException {
super(context.reader());
final int maxDoc = in.maxDoc();
final FixedBitSet bits = new FixedBitSet(maxDoc);
// ignore livedocs here, as we filter them later:
final DocIdSet docs = preserveFilter.getDocIdSet(context, null);
if (docs != null) {
final DocIdSetIterator it = docs.iterator();
if (it != null) {
bits.or(it);
}
}
if (negateFilter) {
bits.flip(0, maxDoc);
}
if (in.hasDeletions()) {
final Bits oldLiveDocs = in.getLiveDocs();
assert oldLiveDocs != null;
final DocIdSetIterator it = new BitSetIterator(bits, 0L); // the cost is not useful here
for (int i = it.nextDoc(); i < maxDoc; i = it.nextDoc()) {
if (!oldLiveDocs.get(i)) {
// we can safely modify the current bit, as the iterator already stepped over it:
bits.clear(i);
}
}
}
this.liveDocs = bits;
this.numDocs = bits.cardinality();
}
@Override
public Scorer filteredScorer(LeafReaderContext context, Weight weight, DocIdSet docIdSet)
throws IOException {
// CHANGE: If threshold is 0, always pass down the accept docs, don't pay the price of calling
// nextDoc even...
final Bits filterAcceptDocs = docIdSet.bits();
if (threshold == 0) {
if (filterAcceptDocs != null) {
return weight.scorer(context, filterAcceptDocs);
} else {
return FilteredQuery.LEAP_FROG_QUERY_FIRST_STRATEGY.filteredScorer(
context, weight, docIdSet);
}
}
// CHANGE: handle "default" value
if (threshold == -1) {
// default value, don't iterate on only apply filter after query if its not a "fast"
// docIdSet
// TODO: is there a way we could avoid creating an iterator here?
if (filterAcceptDocs != null && DocIdSets.isBroken(docIdSet.iterator())) {
return FilteredQuery.QUERY_FIRST_FILTER_STRATEGY.filteredScorer(
context, weight, docIdSet);
}
}
return super.filteredScorer(context, weight, docIdSet);
}
@Override
public BulkScorer bulkScorer(LeafReaderContext context) throws IOException {
if (used.compareAndSet(false, true)) {
policy.onUse(getQuery());
}
DocIdSet docIdSet = get(in.getQuery(), context);
if (docIdSet == null) {
if (shouldCache(context)) {
docIdSet = cache(context);
putIfAbsent(in.getQuery(), context, docIdSet);
} else {
return in.bulkScorer(context);
}
}
assert docIdSet != null;
if (docIdSet == DocIdSet.EMPTY) {
return null;
}
final DocIdSetIterator disi = docIdSet.iterator();
if (disi == null) {
return null;
}
return new DefaultBulkScorer(new ConstantScoreScorer(this, 0f, disi));
}
// Delete by query
private static long applyQueryDeletes(
Iterable<QueryAndLimit> queriesIter, ReadersAndUpdates rld, final SegmentReader reader)
throws IOException {
long delCount = 0;
final AtomicReaderContext readerContext = reader.getContext();
boolean any = false;
for (QueryAndLimit ent : queriesIter) {
Query query = ent.query;
int limit = ent.limit;
final DocIdSet docs =
new QueryWrapperFilter(query).getDocIdSet(readerContext, reader.getLiveDocs());
if (docs != null) {
final DocIdSetIterator it = docs.iterator();
if (it != null) {
while (true) {
int doc = it.nextDoc();
if (doc >= limit) {
break;
}
if (!any) {
rld.initWritableLiveDocs();
any = true;
}
if (rld.delete(doc)) {
delCount++;
}
}
}
}
}
return delCount;
}
@Override
public DocIdSetIterator iterator() throws IOException {
// we try and be smart here, if we can iterate through docsets quickly, prefer to iterate
// over them as much as possible, before actually going to "bits" based ones to check
List<DocIdSet> iterators = new ArrayList<DocIdSet>(sets.length);
List<Bits> bits = new ArrayList<Bits>(sets.length);
for (DocIdSet set : sets) {
if (DocIdSets.isFastIterator(set)) {
iterators.add(set);
} else {
Bits bit = set.bits();
if (bit != null) {
bits.add(bit);
} else {
iterators.add(set);
}
}
}
if (bits.isEmpty()) {
return new IteratorBasedIterator(iterators.toArray(new DocIdSet[iterators.size()]));
}
if (iterators.isEmpty()) {
return new BitsDocIdSetIterator(new AndBits(bits.toArray(new Bits[bits.size()])));
}
// combination of both..., first iterating over the "fast" ones, and then checking on the more
// expensive ones
return new BitsDocIdSetIterator.FilteredIterator(
new IteratorBasedIterator(iterators.toArray(new DocIdSet[iterators.size()])),
new AndBits(bits.toArray(new Bits[bits.size()])));
}
/**
* Returns the best nested {@link ObjectMapper} instances that is in the scope of the specified
* nested docId.
*/
public ObjectMapper findNestedObjectMapper(
int nestedDocId, SearchContext sc, LeafReaderContext context) throws IOException {
ObjectMapper nestedObjectMapper = null;
for (ObjectMapper objectMapper : objectMappers().values()) {
if (!objectMapper.nested().isNested()) {
continue;
}
Filter filter = objectMapper.nestedTypeFilter();
if (filter == null) {
continue;
}
// We can pass down 'null' as acceptedDocs, because nestedDocId is a doc to be fetched and
// therefor is guaranteed to be a live doc.
DocIdSet nestedTypeSet = filter.getDocIdSet(context, null);
if (nestedTypeSet == null) {
continue;
}
DocIdSetIterator iterator = nestedTypeSet.iterator();
if (iterator == null) {
continue;
}
if (iterator.advance(nestedDocId) == nestedDocId) {
if (nestedObjectMapper == null) {
nestedObjectMapper = objectMapper;
} else {
if (nestedObjectMapper.fullPath().length() < objectMapper.fullPath().length()) {
nestedObjectMapper = objectMapper;
}
}
}
}
return nestedObjectMapper;
}
@Override
public Scorer scorer(
AtomicReaderContext context,
boolean scoreDocsInOrder,
boolean topScorer,
final Bits acceptDocs)
throws IOException {
final DocIdSetIterator disi;
if (filter != null) {
assert query == null;
final DocIdSet dis = filter.getDocIdSet(context, acceptDocs);
if (dis == null) {
return null;
}
disi = dis.iterator();
} else {
assert query != null && innerWeight != null;
disi = innerWeight.scorer(context, scoreDocsInOrder, topScorer, acceptDocs);
}
if (disi == null) {
return null;
}
return new ConstantScorer(disi, this, queryWeight);
}
void remove(Query query) {
assert query instanceof BoostQuery == false;
assert query instanceof ConstantScoreQuery == false;
DocIdSet removed = cache.remove(query);
if (removed != null) {
onDocIdSetEviction(HASHTABLE_RAM_BYTES_PER_ENTRY + removed.ramBytesUsed());
}
}
@Test
public void testStream_docIdSetIterator_empty() throws Exception {
final Builder disBuilder = new Builder(10);
final DocIdSet dis = disBuilder.build();
Assert.assertEquals(
"Too much document ids streamed.", 0L, StreamUtils.stream(dis.iterator()).count());
}
@Override
public boolean isCacheable() {
for (DocIdSet set : sets) {
if (!set.isCacheable()) {
return false;
}
}
return true;
}
private RoaringDocIdSet(DocIdSet[] docIdSets, int cardinality) {
this.docIdSets = docIdSets;
long ramBytesUsed = BASE_RAM_BYTES_USED + RamUsageEstimator.shallowSizeOf(docIdSets);
for (DocIdSet set : this.docIdSets) {
if (set != null) {
ramBytesUsed += set.ramBytesUsed();
}
}
this.ramBytesUsed = ramBytesUsed;
this.cardinality = cardinality;
}
private void tstFilterCard(String mes, int expected, Filter filt) throws Exception {
final DocIdSet docIdSet = filt.getDocIdSet(reader.getContext(), reader.getLiveDocs());
int actual = 0;
if (docIdSet != null) {
DocIdSetIterator disi = docIdSet.iterator();
while (disi.nextDoc() != DocIdSetIterator.NO_MORE_DOCS) {
actual++;
}
}
assertEquals(mes, expected, actual);
}
@Override
public Scorer scorer(
AtomicReaderContext context, boolean scoreDocsInOrder, boolean topScorer, Bits acceptDocs)
throws IOException {
DocIdSet childrenDocSet = childrenFilter.getDocIdSet(context, acceptDocs);
if (childrenDocSet == null || childrenDocSet == DocIdSet.EMPTY_DOCIDSET) {
return null;
}
IdReaderTypeCache idTypeCache =
searchContext.idCache().reader(context.reader()).type(parentType);
return new ChildScorer(this, uidToScore, childrenDocSet.iterator(), idTypeCache);
}
private static void assertDocIdSetCacheable(
IndexReader reader, Filter filter, boolean shouldCacheable) throws IOException {
assertTrue(reader.getContext() instanceof AtomicReaderContext);
AtomicReaderContext context = (AtomicReaderContext) reader.getContext();
final CachingWrapperFilter cacher = new CachingWrapperFilter(filter);
final DocIdSet originalSet = filter.getDocIdSet(context, context.reader().getLiveDocs());
final DocIdSet cachedSet = cacher.getDocIdSet(context, context.reader().getLiveDocs());
if (originalSet == null) {
assertNull(cachedSet);
}
if (cachedSet == null) {
assertTrue(originalSet == null || originalSet.iterator() == null);
} else {
assertTrue(cachedSet.isCacheable());
assertEquals(shouldCacheable, originalSet.isCacheable());
// System.out.println("Original: "+originalSet.getClass().getName()+" -- cached:
// "+cachedSet.getClass().getName());
if (originalSet.isCacheable()) {
assertEquals(
"Cached DocIdSet must be of same class like uncached, if cacheable",
originalSet.getClass(),
cachedSet.getClass());
} else {
assertTrue(
"Cached DocIdSet must be an FixedBitSet if the original one was not cacheable",
cachedSet instanceof FixedBitSet || cachedSet == null);
}
}
}
@Override
public Scorer scorer(LeafReaderContext context) throws IOException {
DocIdSet docIdSet =
filter instanceof SolrFilter
? ((SolrFilter) filter).getDocIdSet(this.context, context, null)
: filter.getDocIdSet(context, null);
if (docIdSet == null) {
return null;
}
DocIdSetIterator iterator = docIdSet.iterator();
if (iterator == null) {
return null;
}
return new ConstantScoreScorer(this, score(), iterator);
}
@Override
public Scorer scorer(IndexReader reader, boolean scoreDocsInOrder, boolean topScorer)
throws IOException {
final DocIdSetIterator disi;
if (filter != null) {
assert query == null;
final DocIdSet dis = filter.getDocIdSet(reader);
if (dis == null) return null;
disi = dis.iterator();
} else {
assert query != null && innerWeight != null;
disi = innerWeight.scorer(reader, scoreDocsInOrder, topScorer);
}
if (disi == null) return null;
return new ConstantScorer(similarity, disi, this);
}
private void assertEquals(DocIdSet d1, DocIdSet d2) throws IOException {
if (d1 == null) {
if (d2 != null) {
assertEquals(DocIdSetIterator.NO_MORE_DOCS, d2.iterator().nextDoc());
}
} else if (d2 == null) {
assertEquals(DocIdSetIterator.NO_MORE_DOCS, d1.iterator().nextDoc());
} else {
DocIdSetIterator i1 = d1.iterator();
DocIdSetIterator i2 = d2.iterator();
for (int doc = i1.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = i1.nextDoc()) {
assertEquals(doc, i2.nextDoc());
}
assertEquals(DocIdSetIterator.NO_MORE_DOCS, i2.nextDoc());
}
}
public ConstantScorer(Similarity similarity, IndexReader reader, Weight w) throws IOException {
super(similarity);
this.reader = reader;
theScore = w.getValue();
DocIdSet docIdSet = filter.getDocIdSet(reader);
if (docIdSet == null) {
_innerIter = DocIdSet.EMPTY_DOCIDSET.iterator();
} else {
DocIdSetIterator iter = docIdSet.iterator();
if (iter == null) {
_innerIter = DocIdSet.EMPTY_DOCIDSET.iterator();
} else {
_innerIter = iter;
}
}
}
// pkg-private for testing
synchronized void assertConsistent() {
if (requiresEviction()) {
throw new AssertionError(
"requires evictions: size="
+ mostRecentlyUsedQueries.size()
+ ", maxSize="
+ maxSize
+ ", ramBytesUsed="
+ ramBytesUsed()
+ ", maxRamBytesUsed="
+ maxRamBytesUsed);
}
for (LeafCache leafCache : cache.values()) {
Set<Query> keys = Collections.newSetFromMap(new IdentityHashMap<>());
keys.addAll(leafCache.cache.keySet());
keys.removeAll(mostRecentlyUsedQueries);
if (!keys.isEmpty()) {
throw new AssertionError(
"One leaf cache contains more keys than the top-level cache: " + keys);
}
}
long recomputedRamBytesUsed =
HASHTABLE_RAM_BYTES_PER_ENTRY * cache.size()
+ LINKED_HASHTABLE_RAM_BYTES_PER_ENTRY * uniqueQueries.size();
for (Query query : mostRecentlyUsedQueries) {
recomputedRamBytesUsed += ramBytesUsed(query);
}
for (LeafCache leafCache : cache.values()) {
recomputedRamBytesUsed += HASHTABLE_RAM_BYTES_PER_ENTRY * leafCache.cache.size();
for (DocIdSet set : leafCache.cache.values()) {
recomputedRamBytesUsed += set.ramBytesUsed();
}
}
if (recomputedRamBytesUsed != ramBytesUsed) {
throw new AssertionError(
"ramBytesUsed mismatch : " + ramBytesUsed + " != " + recomputedRamBytesUsed);
}
long recomputedCacheSize = 0;
for (LeafCache leafCache : cache.values()) {
recomputedCacheSize += leafCache.cache.size();
}
if (recomputedCacheSize != getCacheSize()) {
throw new AssertionError(
"cacheSize mismatch : " + getCacheSize() + " != " + recomputedCacheSize);
}
}
public SpansFiltered(Spans spans, DocIdSet filterDocs) throws IOException {
this.spans = BLSpansWrapper.optWrap(spans);
docIdSetIter = filterDocs.iterator();
more = false;
if (docIdSetIter != null) {
more = (docIdSetIter.nextDoc() != DocIdSetIterator.NO_MORE_DOCS);
}
}
void putIfAbsent(Query query, DocIdSet set) {
assert query instanceof BoostQuery == false;
assert query instanceof ConstantScoreQuery == false;
if (cache.putIfAbsent(query, set) == null) {
// the set was actually put
onDocIdSetCache(HASHTABLE_RAM_BYTES_PER_ENTRY + set.ramBytesUsed());
}
}
@Override
public void compute(OpenBitSetDISI accumulator, DocIdSet bitset) throws IOException {
if (bitset instanceof OpenBitSet) {
accumulator.and((OpenBitSet) bitset);
} else if (bitset instanceof SortedVIntList) {
accumulator.inPlaceAnd(bitset.iterator());
} else {
throw new IllegalArgumentException("Not supported:" + bitset);
}
}
public void doTestIteratorEqual(DocIdSet a, DocIdSet b) throws IOException {
DocIdSetIterator ia = a.iterator();
DocIdSetIterator ib = b.iterator();
// test for next() equivalence
for (; ; ) {
int da = ia.nextDoc();
int db = ib.nextDoc();
assertEquals(da, db);
assertEquals(ia.docID(), ib.docID());
if (da == DocIdSetIterator.NO_MORE_DOCS) break;
}
for (int i = 0; i < 10; i++) {
// test random skipTo() and next()
ia = a.iterator();
ib = b.iterator();
int doc = -1;
for (; ; ) {
int da, db;
if (rand.nextBoolean()) {
da = ia.nextDoc();
db = ib.nextDoc();
} else {
int target =
doc
+ rand.nextInt(10)
+ 1; // keep in mind future edge cases like probing (increase if necessary)
da = ia.advance(target);
db = ib.advance(target);
}
assertEquals(da, db);
assertEquals(ia.docID(), ib.docID());
if (da == DocIdSetIterator.NO_MORE_DOCS) break;
doc = da;
}
}
}
private List<Document> lookupDocs(Term term, final LoadFieldCallback lfc) throws IOException {
final List<Document> documents = new ArrayList<Document>();
final TermFilter tf = new TermFilter(term);
try {
for (AtomicReaderContext arc : searcher.getIndexReader().leaves()) {
AtomicReader ar = arc.reader();
Bits liveDocs = ar.getLiveDocs();
DocIdSet docSet = tf.getDocIdSet(arc, liveDocs);
if (docSet != null) {
DocIdSetIterator disi = docSet.iterator();
if (disi != null) {
int docId;
while ((docId = disi.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) {
DocumentStoredFieldVisitor fv =
new DocumentStoredFieldVisitor() {
@Override
public StoredFieldVisitor.Status needsField(FieldInfo fieldInfo)
throws IOException {
if (lfc == null || lfc.loadField(fieldInfo.name)) {
return StoredFieldVisitor.Status.YES;
}
return StoredFieldVisitor.Status.NO;
}
};
ar.document(docId, fv);
documents.add(fv.getDocument());
}
}
}
}
} catch (IOException io) {
throw new IndexException(io);
}
return documents;
}
@Test
public void testStream_docIdSetIterator() throws Exception {
final Builder disBuilder = new Builder(10);
disBuilder.add(1).add(3).add(6).add(7).add(8).add(10);
final DocIdSet dis = disBuilder.build();
Assert.assertEquals(
"Not all document ids streamed.", 6L, StreamUtils.stream(dis.iterator()).count());
Assert.assertEquals(
"Document id count mismatch.",
1L,
StreamUtils.stream(dis.iterator()).filter(id -> id == 1).count());
Assert.assertEquals(
"Document id count mismatch.",
1L,
StreamUtils.stream(dis.iterator()).filter(id -> id == 3).count());
Assert.assertEquals(
"Document id count mismatch.",
1L,
StreamUtils.stream(dis.iterator()).filter(id -> id == 6).count());
Assert.assertEquals(
"Document id count mismatch.",
1L,
StreamUtils.stream(dis.iterator()).filter(id -> id == 7).count());
Assert.assertEquals(
"Document id count mismatch.",
1L,
StreamUtils.stream(dis.iterator()).filter(id -> id == 8).count());
Assert.assertEquals(
"Document id count mismatch.",
1L,
StreamUtils.stream(dis.iterator()).filter(id -> id == 10).count());
Assert.assertEquals(
"Unknown document id found.",
0L,
StreamUtils.stream(dis.iterator())
.filter(id -> id != 1 && id != 3 && id != 6 && id != 7 && id != 8 && id != 10)
.count());
}
@Override
public Scorer scorer(LeafReaderContext context, Bits acceptDocs) throws IOException {
if (remaining == 0) {
return null;
}
if (shortCircuitFilter != null) {
DocIdSet docIdSet = shortCircuitFilter.getDocIdSet(context, acceptDocs);
if (!DocIdSets.isEmpty(docIdSet)) {
DocIdSetIterator iterator = docIdSet.iterator();
if (iterator != null) {
return ConstantScorer.create(iterator, this, queryWeight);
}
}
return null;
}
DocIdSet parentDocIdSet = this.parentFilter.getDocIdSet(context, acceptDocs);
if (!DocIdSets.isEmpty(parentDocIdSet)) {
// We can't be sure of the fact that liveDocs have been applied, so we apply it here. The
// "remaining"
// count down (short circuit) logic will then work as expected.
parentDocIdSet = BitsFilteredDocIdSet.wrap(parentDocIdSet, context.reader().getLiveDocs());
DocIdSetIterator innerIterator = parentDocIdSet.iterator();
if (innerIterator != null) {
LongBitSet parentOrds = collector.parentOrds;
SortedDocValues globalValues = globalIfd.load(context).getOrdinalsValues(parentType);
if (globalValues != null) {
DocIdSetIterator parentIdIterator =
new ParentOrdIterator(innerIterator, parentOrds, globalValues, this);
return ConstantScorer.create(parentIdIterator, this, queryWeight);
}
}
}
return null;
}
IteratorBasedIterator(DocIdSet[] sets) throws IOException {
iterators = new DocIdSetIterator[sets.length];
int j = 0;
long cost = Integer.MAX_VALUE;
for (DocIdSet set : sets) {
if (set == null) {
lastReturn = DocIdSetIterator.NO_MORE_DOCS; // non matching
break;
} else {
DocIdSetIterator dcit = set.iterator();
if (dcit == null) {
lastReturn = DocIdSetIterator.NO_MORE_DOCS; // non matching
break;
}
iterators[j++] = dcit;
cost = Math.min(cost, dcit.cost());
}
}
this.cost = cost;
if (lastReturn != DocIdSetIterator.NO_MORE_DOCS) {
lastReturn = (iterators.length > 0 ? -1 : DocIdSetIterator.NO_MORE_DOCS);
}
}
@Override
public DocIdSet getDocIdSet(IndexReader subReader) throws IOException {
if (bases == null || !bases.containsKey(subReader)) {
return docs;
}
int docBase = bases.get(subReader);
int readerSize = subReader.maxDoc();
OpenBitSet filter = new OpenBitSet(readerSize);
DocIdSetIterator iterator = docs.iterator();
int doc = iterator.advance(docBase);
while (doc < docBase + readerSize) {
filter.set(doc - docBase);
doc = iterator.nextDoc();
}
return filter;
}
private InternalSearchHit.InternalNestedIdentity getInternalNestedIdentity(
SearchContext context,
int nestedSubDocId,
LeafReaderContext subReaderContext,
DocumentMapper documentMapper,
ObjectMapper nestedObjectMapper)
throws IOException {
int currentParent = nestedSubDocId;
ObjectMapper nestedParentObjectMapper;
StringBuilder field = new StringBuilder();
ObjectMapper current = nestedObjectMapper;
InternalSearchHit.InternalNestedIdentity nestedIdentity = null;
do {
Filter parentFilter;
nestedParentObjectMapper = documentMapper.findParentObjectMapper(current);
if (field.length() != 0) {
field.insert(0, '.');
}
field.insert(0, current.name());
if (nestedParentObjectMapper != null) {
if (nestedParentObjectMapper.nested().isNested() == false) {
current = nestedParentObjectMapper;
continue;
}
parentFilter = nestedParentObjectMapper.nestedTypeFilter();
} else {
parentFilter = Queries.newNonNestedFilter();
}
Filter childFilter = nestedObjectMapper.nestedTypeFilter();
if (childFilter == null) {
current = nestedParentObjectMapper;
continue;
}
// We can pass down 'null' as acceptedDocs, because we're fetching matched docId that matched
// in the query phase.
DocIdSet childDocSet = childFilter.getDocIdSet(subReaderContext, null);
if (childDocSet == null) {
current = nestedParentObjectMapper;
continue;
}
DocIdSetIterator childIter = childDocSet.iterator();
if (childIter == null) {
current = nestedParentObjectMapper;
continue;
}
BitDocIdSet parentBitSet =
context
.bitsetFilterCache()
.getBitDocIdSetFilter(parentFilter)
.getDocIdSet(subReaderContext);
BitSet parentBits = parentBitSet.bits();
int offset = 0;
int nextParent = parentBits.nextSetBit(currentParent);
for (int docId = childIter.advance(currentParent + 1);
docId < nextParent && docId != DocIdSetIterator.NO_MORE_DOCS;
docId = childIter.nextDoc()) {
offset++;
}
currentParent = nextParent;
current = nestedObjectMapper = nestedParentObjectMapper;
nestedIdentity =
new InternalSearchHit.InternalNestedIdentity(field.toString(), offset, nestedIdentity);
field = new StringBuilder();
} while (current != null);
return nestedIdentity;
}
@Override
public OpenBitSetDISI newAccumulator(int bitsetSize, DocIdSet b) throws IOException {
return new OpenBitSetDISI(b.iterator(), bitsetSize);
}
