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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
|
// Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include "parallel_weak_and_blueprint.h"
#include "wand_parts.h"
#include "parallel_weak_and_search.h"
#include <vespa/searchlib/queryeval/field_spec.hpp>
#include <vespa/searchlib/queryeval/searchiterator.h>
#include <vespa/searchlib/queryeval/flow_tuning.h>
#include <vespa/searchlib/fef/termfieldmatchdata.h>
#include <vespa/vespalib/objects/visit.hpp>
#include <algorithm>
namespace search::queryeval {
ParallelWeakAndBlueprint::ParallelWeakAndBlueprint(FieldSpecBase field,
uint32_t scoresToTrack,
score_t scoreThreshold,
double thresholdBoostFactor)
: ComplexLeafBlueprint(field),
_scores(scoresToTrack),
_scoreThreshold(scoreThreshold),
_thresholdBoostFactor(thresholdBoostFactor),
_scoresAdjustFrequency(DEFAULT_PARALLEL_WAND_SCORES_ADJUST_FREQUENCY),
_layout(),
_weights(),
_terms()
{
}
ParallelWeakAndBlueprint::ParallelWeakAndBlueprint(FieldSpecBase field,
uint32_t scoresToTrack,
score_t scoreThreshold,
double thresholdBoostFactor,
uint32_t scoresAdjustFrequency)
: ComplexLeafBlueprint(field),
_scores(scoresToTrack),
_scoreThreshold(scoreThreshold),
_thresholdBoostFactor(thresholdBoostFactor),
_scoresAdjustFrequency(scoresAdjustFrequency),
_layout(),
_weights(),
_terms()
{
}
ParallelWeakAndBlueprint::~ParallelWeakAndBlueprint() = default;
void
ParallelWeakAndBlueprint::reserve(size_t num_children) {
_weights.reserve(num_children);
_terms.reserve(num_children);
}
void
ParallelWeakAndBlueprint::addTerm(Blueprint::UP term, int32_t weight, HitEstimate & estimate)
{
HitEstimate childEst = term->getState().estimate();
if (!childEst.empty) {
if (estimate.empty) {
estimate = childEst;
} else {
estimate.estHits += childEst.estHits;
}
}
_weights.push_back(weight);
_terms.push_back(std::move(term));
}
void
ParallelWeakAndBlueprint::sort(InFlow in_flow)
{
strict(in_flow.strict());
auto flow = OrFlow(in_flow);
for (auto &term: _terms) {
term->sort(InFlow(flow.strict(), flow.flow()));
flow.add(term->estimate());
}
}
FlowStats
ParallelWeakAndBlueprint::calculate_flow_stats(uint32_t docid_limit) const
{
for (auto &term: _terms) {
term->update_flow_stats(docid_limit);
}
double child_est = OrFlow::estimate_of(_terms);
double my_est = abs_to_rel_est(_scores.getScoresToTrack(), docid_limit);
double est = (child_est + my_est) / 2.0;
return {est, OrFlow::cost_of(_terms, false),
OrFlow::cost_of(_terms, true) + flow::heap_cost(est, _terms.size())};
}
SearchIterator::UP
ParallelWeakAndBlueprint::createLeafSearch(const search::fef::TermFieldMatchDataArray &tfmda) const
{
assert(tfmda.size() == 1);
fef::MatchData::UP childrenMatchData = _layout.createMatchData();
wand::Terms terms;
terms.reserve(_terms.size());
for (size_t i = 0; i < _terms.size(); ++i) {
const State &childState = _terms[i]->getState();
assert(childState.numFields() == 1);
// TODO: pass ownership with unique_ptr
terms.emplace_back(_terms[i]->createSearch(*childrenMatchData).release(),
_weights[i],
childState.estimate().estHits,
childState.field(0).resolve(*childrenMatchData));
}
return SearchIterator::UP
(ParallelWeakAndSearch::create(terms,
ParallelWeakAndSearch::MatchParams(_scores,
_scoreThreshold,
_thresholdBoostFactor,
_scoresAdjustFrequency).setDocIdLimit(get_docid_limit()),
ParallelWeakAndSearch::RankParams(*tfmda[0],
std::move(childrenMatchData)), strict()));
}
std::unique_ptr<SearchIterator>
ParallelWeakAndBlueprint::createFilterSearch(FilterConstraint constraint) const
{
return create_atmost_or_filter(_terms, strict(), constraint);
}
void
ParallelWeakAndBlueprint::fetchPostings(const ExecuteInfo & execInfo)
{
for (const auto & _term : _terms) {
_term->fetchPostings(execInfo);
}
}
bool
ParallelWeakAndBlueprint::always_needs_unpack() const
{
return true;
}
void
ParallelWeakAndBlueprint::visitMembers(vespalib::ObjectVisitor &visitor) const
{
LeafBlueprint::visitMembers(visitor);
visit(visitor, "_weights", _weights);
visit(visitor, "_terms", _terms);
}
}
|
