Search performance optimization

include/lucene++/AbstractAllTermDocs.h

@@ -32,7 +32,7 @@ class LPPAPI AbstractAllTermDocs : public TermDocs, public LuceneObject {
     virtual int32_t doc();
     virtual int32_t freq();
     virtual bool next();
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
     virtual bool skipTo(int32_t target);
     virtual void close();
     virtual bool isDeleted(int32_t doc) = 0;

include/lucene++/Array.h

@@ -100,7 +100,6 @@ class Array {
     }
 
     TYPE& operator[] (int32_t i) const {
-        BOOST_ASSERT(i >= 0 && i < array->size);
         return array->data[i];
     }
 

include/lucene++/BooleanScorer.h

@@ -46,6 +46,7 @@ class LPPAPI BooleanScorer : public Scorer {
     int32_t minNrShouldMatch;
     int32_t end;
     BucketPtr current;
+    Bucket* __current = nullptr;
     int32_t doc;
 
 protected:
@@ -71,8 +72,10 @@ class BooleanScorerCollector : public Collector {
 
 protected:
     BucketTableWeakPtr _bucketTable;
+    BucketTable* __bucketTable = nullptr;
     int32_t mask;
     ScorerWeakPtr _scorer;
+    Scorer* __scorer = nullptr;
 
 public:
     virtual void collect(int32_t doc);
@@ -121,6 +124,7 @@ class Bucket : public LuceneObject {
     int32_t bits; // used for bool constraints
     int32_t coord; // count of terms in score
     BucketWeakPtr _next; // next valid bucket
+    Bucket* __next = nullptr; // next valid bucket
 };
 
 /// A simple hash table of document scores within a range.
@@ -137,6 +141,7 @@ class BucketTable : public LuceneObject {
 
     Collection<BucketPtr> buckets;
     BucketPtr first; // head of valid list
+    Bucket* __first = nullptr; // head of valid list
 
 public:
     CollectorPtr newCollector(int32_t mask);

include/lucene++/BufferedIndexInput.h

@@ -30,12 +30,18 @@ class LPPAPI BufferedIndexInput : public IndexInput {
     int32_t bufferLength; // end of valid bytes
     int32_t bufferPosition; // next byte to read
     ByteArray buffer;
+    decltype(buffer.get()) __buffer;
 
 public:
     /// Reads and returns a single byte.
     /// @see IndexOutput#writeByte(uint8_t)
     virtual uint8_t readByte();
 
+    /// Reads an int stored in variable-length format.  Reads between one and five
+    /// bytes.  Smaller values take fewer bytes.  Negative numbers are not supported.
+    /// @see IndexOutput#writeVInt(int32_t)
+    virtual int32_t readVInt();
+
     /// Change the buffer size used by this IndexInput.
     void setBufferSize(int32_t newSize);
 

include/lucene++/Collection.h

@@ -193,6 +193,10 @@ class Collection : public LuceneSync {
     bool operator!= (const this_type& other) {
         return (container != other.container);
     }
+
+    collection_type* get() {
+        return container.get();
+    }
 };
 
 template <typename TYPE>

include/lucene++/DirectoryReader.h

@@ -262,7 +262,7 @@ class MultiTermDocs : public TermPositions, public LuceneObject {
 
     /// Attempts to read multiple entries from the enumeration, up to length of docs.
     /// Optimized implementation.
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
 
     /// Skips entries to the first beyond the current whose document number is greater than or equal to target.
     virtual bool skipTo(int32_t target);

include/lucene++/FilterIndexReader.h

@@ -93,7 +93,7 @@ class LPPAPI FilterTermDocs : public TermPositions, public LuceneObject {
     virtual int32_t doc();
     virtual int32_t freq();
     virtual bool next();
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
     virtual bool skipTo(int32_t target);
     virtual void close();
 };

include/lucene++/LuceneFactory.h

@@ -104,7 +104,11 @@ boost::shared_ptr<T> newInstance(A1 const& a1, A2 const& a2, A3 const& a3, A4 co
 
 template <class T>
 boost::shared_ptr<T> newLucene() {
-    boost::shared_ptr<T> instance(newInstance<T>());
+#if BOOST_VERSION <= 103800
+    boost::shared_ptr<T> instance = boost::shared_ptr<T>(new T);
+#else
+    boost::shared_ptr<T> instance = boost::make_shared<T>();
+#endif
     instance->initialize();
     return instance;
 }

include/lucene++/MiscUtils.h

@@ -132,6 +132,14 @@ class LPPAPI MiscUtils {
     static int32_t unsignedShift(int32_t num, int32_t shift);
 };
 
+inline int64_t MiscUtils::unsignedShift(int64_t num, int64_t shift) {
+    return (shift & 0x3f) == 0 ? num : (((uint64_t)num >> 1) & 0x7fffffffffffffffLL) >> ((shift & 0x3f) - 1);
+}
+
+inline int32_t MiscUtils::unsignedShift(int32_t num, int32_t shift) {
+    return (shift & 0x1f) == 0 ? num : (((uint32_t)num >> 1) & 0x7fffffff) >> ((shift & 0x1f) - 1);
+}
+
 }
 
 #endif

include/lucene++/MultipleTermPositions.h

@@ -41,7 +41,7 @@ class LPPAPI MultipleTermPositions : public TermPositions, public LuceneObject {
     virtual void seek(const TermEnumPtr& termEnum);
 
     /// Not implemented.
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
 
     /// Not implemented.
     virtual ByteArray getPayload(ByteArray data, int32_t offset);

include/lucene++/PhrasePositions.h

@@ -25,7 +25,7 @@ class PhrasePositions : public LuceneObject {
     int32_t count; // remaining pos in this doc
     int32_t offset; // position in phrase
     TermPositionsPtr tp; // stream of positions
-    PhrasePositionsPtr _next; // used to make lists
+    PhrasePositions* __next = nullptr; // used to make lists
     bool repeats; // there's other pp for same term (eg. query="1st word 2nd word"~1)
 
 public:

include/lucene++/PhraseQueue.h

@@ -10,16 +10,19 @@
 #include "PriorityQueue.h"
 
 namespace Lucene {
+// raw pointer 
+typedef PhrasePositions* PhrasePositionsStar;
 
-class PhraseQueue : public PriorityQueue<PhrasePositionsPtr> {
+class PhraseQueue : public PriorityQueue<PhrasePositionsStar> {
 public:
     PhraseQueue(int32_t size);
     virtual ~PhraseQueue();
 
     LUCENE_CLASS(PhraseQueue);
 
 protected:
-    virtual bool lessThan(const PhrasePositionsPtr& first, const PhrasePositionsPtr& second);
+    virtual bool lessThan(const PhrasePositionsStar& first, const PhrasePositionsStar& second);
+
 };
 
 }

include/lucene++/PhraseScorer.h

@@ -8,6 +8,7 @@
 #define PHRASESCORER_H
 
 #include "Scorer.h"
+#include <vector>
 
 namespace Lucene {
 
@@ -27,14 +28,16 @@ class PhraseScorer : public Scorer {
 
 protected:
     WeightPtr weight;
+    Weight* __weight = nullptr;
     ByteArray norms;
     double value;
 
     bool firstTime;
     bool more;
     PhraseQueuePtr pq;
-    PhrasePositionsPtr first;
-    PhrasePositionsPtr last;
+    std::vector<PhrasePositionsPtr> _holds;
+    PhrasePositions* __first = nullptr;
+    PhrasePositions* __last = nullptr;
 
     double freq; // phrase frequency in current doc as computed by phraseFreq().
 

include/lucene++/SegmentTermDocs.h

@@ -20,10 +20,13 @@ class LPPAPI SegmentTermDocs : public TermPositions, public LuceneObject {
 
 protected:
     SegmentReaderWeakPtr _parent;
+    SegmentReader* __parent;
     IndexInputPtr _freqStream;
+    IndexInput* __freqStream;
     int32_t count;
     int32_t df;
     BitVectorPtr deletedDocs;
+    BitVector* __deletedDocs;
     int32_t _doc;
     int32_t _freq;
 
@@ -61,7 +64,7 @@ class LPPAPI SegmentTermDocs : public TermPositions, public LuceneObject {
     virtual bool next();
 
     /// Optimized implementation.
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
 
     /// Optimized implementation.
     virtual bool skipTo(int32_t target);
@@ -72,7 +75,7 @@ class LPPAPI SegmentTermDocs : public TermPositions, public LuceneObject {
 
 protected:
     virtual void skippingDoc();
-    virtual int32_t readNoTf(Collection<int32_t> docs, Collection<int32_t> freqs, int32_t length);
+    virtual int32_t readNoTf(Collection<int32_t>& docs, Collection<int32_t>& freqs, int32_t length);
 
     /// Overridden by SegmentTermPositions to skip in prox stream.
     virtual void skipProx(int64_t proxPointer, int32_t payloadLength);

include/lucene++/SegmentTermPositions.h

@@ -46,7 +46,7 @@ class LPPAPI SegmentTermPositions : public SegmentTermDocs {
     virtual bool next();
 
     /// Not supported
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
 
     /// Returns the length of the payload at the current term position.
     virtual int32_t getPayloadLength();

include/lucene++/Similarity.h

@@ -434,8 +434,8 @@ class LPPAPI Similarity : public LuceneObject {
 protected:
     static const int32_t NO_DOC_ID_PROVIDED;
 
-protected:
-    static const Collection<double> NORM_TABLE();
+public:
+    static const Collection<double> NORM_TABLE;
 
 public:
     /// Return the default Similarity implementation used by indexing and search code.
@@ -450,7 +450,7 @@ class LPPAPI Similarity : public LuceneObject {
 
     /// Returns a table for decoding normalization bytes.
     /// @see #encodeNorm(double)
-    static const Collection<double> getNormDecoder();
+    static const Collection<double>& getNormDecoder();
 
     /// Compute the normalization value for a field, given the accumulated state of term processing for this
     /// field (see {@link FieldInvertState}).

include/lucene++/SloppyPhraseScorer.h

@@ -20,8 +20,8 @@ class SloppyPhraseScorer : public PhraseScorer {
 
 protected:
     int32_t slop;
-    Collection<PhrasePositionsPtr> repeats;
-    Collection<PhrasePositionsPtr> tmpPos; // for flipping repeating pps
+    Collection<PhrasePositions*> repeats;
+    Collection<PhrasePositions*> tmpPos; // for flipping repeating pps
     bool checkedRepeats;
 
 public:
@@ -42,7 +42,7 @@ class SloppyPhraseScorer : public PhraseScorer {
 protected:
     /// Flip pp2 and pp in the queue: pop until finding pp2, insert back all but pp2, insert pp back.
     /// Assumes: pp!=pp2, pp2 in pq, pp not in pq.  Called only when there are repeating pps.
-    PhrasePositionsPtr flip(const PhrasePositionsPtr& pp, const PhrasePositionsPtr& pp2);
+    PhrasePositions* flip(PhrasePositions* pp, PhrasePositions* pp2);
 
     /// Init PhrasePositions in place.
     /// There is a one time initialization for this scorer:
@@ -61,7 +61,7 @@ class SloppyPhraseScorer : public PhraseScorer {
     /// of the same word would go elsewhere in the matched doc.
     /// @return null if differ (i.e. valid) otherwise return the higher offset PhrasePositions out of the first
     /// two PPs found to not differ.
-    PhrasePositionsPtr termPositionsDiffer(const PhrasePositionsPtr& pp);
+    PhrasePositions* termPositionsDiffer(PhrasePositions* pp);
 };
 
 }

include/lucene++/TermDocs.h

@@ -44,7 +44,7 @@ class LPPAPI TermDocs {
     /// Attempts to read multiple entries from the enumeration, up to length of docs.  Document numbers are stored
     /// in docs, and term frequencies are stored in freqs.  Returns the number of entries read.  Zero is only
     /// returned when the stream has been exhausted.
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs) = 0;
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs) = 0;
 
     /// Skips entries to the first beyond the current whose document number is greater than or equal to target.
     /// Returns true if there is such an entry.

include/lucene++/TermScorer.h

@@ -27,13 +27,16 @@ class LPPAPI TermScorer : public Scorer {
 
 protected:
     WeightPtr weight;
-    TermDocsPtr termDocs;
+    TermDocsPtr termDocs; // for malloc and free
+    TermDocs* __termDocs; // for work,   
     ByteArray norms;
     double weightValue;
     int32_t doc;
 
     Collection<int32_t> docs; // buffered doc numbers
+    decltype(docs.get()) __docs; // 
     Collection<int32_t> freqs; // buffered term freqs
+    decltype(freqs.get()) __freqs; // 
 
     int32_t freq;
     int32_t pointer;
@@ -70,7 +73,7 @@ class LPPAPI TermScorer : public Scorer {
     }
 
 protected:
-    static const Collection<double> SIM_NORM_DECODER();
+    static const Collection<double>& SIM_NORM_DECODER();
 
     virtual bool score(const CollectorPtr& collector, int32_t max, int32_t firstDocID);
 };

include/lucene++/TopScoreDocCollector.h

@@ -29,6 +29,7 @@ class LPPAPI TopScoreDocCollector : public TopDocsCollector {
     ScoreDocPtr pqTop;
     int32_t docBase;
     ScorerWeakPtr _scorer;
+    Scorer* __scorer;
 
 public:
     /// Creates a new {@link TopScoreDocCollector} given the number of hits to collect and whether documents

src/contrib/include/MemoryIndex.h

@@ -323,7 +323,7 @@ class LPPCONTRIBAPI MemoryIndexTermPositions : public TermPositions, public Luce
     virtual int32_t doc();
     virtual int32_t freq();
     virtual bool next();
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
     virtual bool skipTo(int32_t target);
     virtual void close();
 

src/contrib/memory/MemoryIndex.cpp

@@ -569,7 +569,7 @@ bool MemoryIndexTermPositions::next() {
     return _next;
 }
 
-int32_t MemoryIndexTermPositions::read(Collection<int32_t> docs, Collection<int32_t> freqs) {
+int32_t MemoryIndexTermPositions::read(Collection<int32_t>& docs, Collection<int32_t>& freqs) {
     if (!hasNext) {
         return 0;
     }

src/core/include/_ParallelReader.h

@@ -61,7 +61,7 @@ class ParallelTermDocs : public TermPositions, public LuceneObject {
     virtual void seek(const TermPtr& term);
     virtual void seek(const TermEnumPtr& termEnum);
     virtual bool next();
-    virtual int32_t read(Collection<int32_t> docs, Collection<int32_t> freqs);
+    virtual int32_t read(Collection<int32_t>& docs, Collection<int32_t>& freqs);
     virtual bool skipTo(int32_t target);
     virtual void close();
 };

src/core/index/AbstractAllTermDocs.cpp

@@ -41,7 +41,7 @@ bool AbstractAllTermDocs::next() {
     return skipTo(_doc + 1);
 }
 
-int32_t AbstractAllTermDocs::read(Collection<int32_t> docs, Collection<int32_t> freqs) {
+int32_t AbstractAllTermDocs::read(Collection<int32_t>& docs, Collection<int32_t>& freqs) {
     int32_t length = docs.size();
     int32_t i = 0;
     while (i < length && _doc < maxDoc) {

src/core/index/DirectoryReader.cpp

@@ -984,7 +984,7 @@ bool MultiTermDocs::next() {
     }
 }
 
-int32_t MultiTermDocs::read(Collection<int32_t> docs, Collection<int32_t> freqs) {
+int32_t MultiTermDocs::read(Collection<int32_t>& docs, Collection<int32_t>& freqs) {
     while (true) {
         while (!current) {
             if (pointer < readers.size()) { // try next segment

src/core/index/FilterIndexReader.cpp

@@ -194,7 +194,7 @@ bool FilterTermDocs::next() {
     return in->next();
 }
 
-int32_t FilterTermDocs::read(Collection<int32_t> docs, Collection<int32_t> freqs) {
+int32_t FilterTermDocs::read(Collection<int32_t>& docs, Collection<int32_t>& freqs) {
     return in->read(docs, freqs);
 }
 

src/core/index/MultipleTermPositions.cpp

@@ -98,7 +98,7 @@ void MultipleTermPositions::seek(const TermEnumPtr& termEnum) {
     boost::throw_exception(UnsupportedOperationException());
 }
 
-int32_t MultipleTermPositions::read(Collection<int32_t> docs, Collection<int32_t> freqs) {
+int32_t MultipleTermPositions::read(Collection<int32_t>& docs, Collection<int32_t>& freqs) {
     boost::throw_exception(UnsupportedOperationException());
     return 0;
 }