1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
|
// Copyright 2019 Oath Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include "nearest_neighbor_iterator.h"
#include <vespa/searchlib/common/bitvector.h>
using search::tensor::DenseTensorAttribute;
using vespalib::ConstArrayRef;
using vespalib::tensor::DenseTensorView;
using vespalib::tensor::MutableDenseTensorView;
using vespalib::tensor::TypedCells;
using CellType = vespalib::eval::ValueType::CellType;
namespace search::queryeval {
namespace {
bool
is_compatible(const vespalib::eval::ValueType& lhs,
const vespalib::eval::ValueType& rhs)
{
return (lhs == rhs);
}
}
/**
* Search iterator for K nearest neighbor matching.
* Uses unpack() as feedback mechanism to track which matches actually became hits.
* Keeps a heap of the K best hit distances.
* Currently always does brute-force scanning, which is very expensive.
**/
template <bool strict, bool has_filter>
class NearestNeighborImpl : public NearestNeighborIterator
{
public:
NearestNeighborImpl(Params params_in)
: NearestNeighborIterator(params_in),
_lhs(params().queryTensor.cellsRef()),
_fieldTensor(params().tensorAttribute.getTensorType()),
_lastScore(0.0)
{
assert(is_compatible(_fieldTensor.fast_type(), params().queryTensor.fast_type()));
}
~NearestNeighborImpl();
void doSeek(uint32_t docId) override {
double distanceLimit = params().distanceHeap.distanceLimit();
while (__builtin_expect((docId < getEndId()), true)) {
if ((!has_filter) || params().filter->testBit(docId)) {
double d = computeDistance(docId, distanceLimit);
if (d <= distanceLimit) {
_lastScore = d;
setDocId(docId);
return;
}
}
if (strict) {
++docId;
} else {
return;
}
}
setAtEnd();
}
void doUnpack(uint32_t docId) override {
double score = params().distanceFunction->to_rawscore(_lastScore);
params().tfmd.setRawScore(docId, score);
params().distanceHeap.used(_lastScore);
}
Trinary is_strict() const override { return strict ? Trinary::True : Trinary::False ; }
private:
double computeDistance(uint32_t docId, double limit) {
params().tensorAttribute.getTensor(docId, _fieldTensor);
auto rhs = _fieldTensor.cellsRef();
return params().distanceFunction->calc_with_limit(_lhs, rhs, limit);
}
TypedCells _lhs;
MutableDenseTensorView _fieldTensor;
double _lastScore;
};
template <bool strict, bool has_filter>
NearestNeighborImpl<strict, has_filter>::~NearestNeighborImpl() = default;
namespace {
template <bool has_filter>
std::unique_ptr<NearestNeighborIterator>
resolve_strict(bool strict, const NearestNeighborIterator::Params ¶ms)
{
CellType lct = params.queryTensor.fast_type().cell_type();
CellType rct = params.tensorAttribute.getTensorType().cell_type();
if (lct != rct) abort();
if (strict) {
using NNI = NearestNeighborImpl<true, has_filter>;
return std::make_unique<NNI>(params);
} else {
using NNI = NearestNeighborImpl<false, has_filter>;
return std::make_unique<NNI>(params);
}
}
} // namespace <unnamed>
std::unique_ptr<NearestNeighborIterator>
NearestNeighborIterator::create(
bool strict,
fef::TermFieldMatchData &tfmd,
const vespalib::tensor::DenseTensorView &queryTensor,
const search::tensor::DenseTensorAttribute &tensorAttribute,
NearestNeighborDistanceHeap &distanceHeap,
const search::BitVector *filter,
const search::tensor::DistanceFunction *dist_fun)
{
Params params(tfmd, queryTensor, tensorAttribute, distanceHeap, filter, dist_fun);
if (filter) {
return resolve_strict<true>(strict, params);
} else {
return resolve_strict<false>(strict, params);
}
}
} // namespace
|
