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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
|
// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include "document_inverter_context.h"
#include <vespa/document/datatype/documenttype.h>
#include <vespa/document/fieldvalue/document.h>
#include <cassert>
#include <vespa/log/log.h>
LOG_SETUP(".memoryindex.document_inverter_context");
namespace search::memoryindex {
using document::DataType;
using document::Document;
using document::DocumentType;
using document::Field;
using vespalib::ISequencedTaskExecutor;
using index::SchemaIndexFields;
namespace {
template <typename Context>
void make_contexts(const SchemaIndexFields& schema_index_fields, ISequencedTaskExecutor& executor, std::vector<Context>& contexts)
{
using ExecutorId = ISequencedTaskExecutor::ExecutorId;
using IdMapping = std::vector<std::tuple<ExecutorId, bool, uint32_t>>;
IdMapping map;
for (uint32_t field_id : schema_index_fields._textFields) {
// TODO: Add bias when sharing sequenced task executor between document types
map.emplace_back(executor.getExecutorId(field_id), false, field_id);
}
uint32_t uri_field_id = 0;
for (auto& uri_field : schema_index_fields._uriFields) {
// TODO: Add bias when sharing sequenced task executor between document types
map.emplace_back(executor.getExecutorId(uri_field._all), true, uri_field_id);
++uri_field_id;
}
std::sort(map.begin(), map.end());
std::optional<ExecutorId> prev_id;
for (auto& entry : map) {
if (!prev_id.has_value() || prev_id.value() != std::get<0>(entry)) {
contexts.emplace_back(std::get<0>(entry));
prev_id = std::get<0>(entry);
}
if (std::get<1>(entry)) {
contexts.back().add_uri_field(std::get<2>(entry));
} else {
contexts.back().add_field(std::get<2>(entry));
}
}
}
class PusherMapping {
std::vector<std::optional<uint32_t>> _pushers;
public:
PusherMapping(size_t size);
~PusherMapping();
void add_mapping(const std::vector<uint32_t>& fields, uint32_t pusher_id) {
for (auto field_id : fields) {
assert(field_id < _pushers.size());
auto& opt_pusher = _pushers[field_id];
assert(!opt_pusher.has_value());
opt_pusher = pusher_id;
}
}
void use_mapping(const std::vector<uint32_t>& fields, std::vector<uint32_t>& pushers) {
for (auto field_id : fields) {
assert(field_id < _pushers.size());
auto& opt_pusher = _pushers[field_id];
assert(opt_pusher.has_value());
pushers.emplace_back(opt_pusher.value());
}
}
};
PusherMapping::PusherMapping(size_t size)
: _pushers(size)
{
}
PusherMapping::~PusherMapping() = default;
/*
* Connect contexts for inverting to contexts for pushing. If we use
* different sequenced task executors or adds different biases to the
* getExecutorId() argument (to enable double buffering) then contexts
* for inverting and contexts for pushing will bundle different sets
* of fields, preventing a 1:1 mapping. If we use the same sequenced
* task executor and drop double buffering then we can simplify this
* to a 1:1 mapping.
*/
void connect_contexts(std::vector<InvertContext>& invert_contexts,
const std::vector<PushContext>& push_contexts,
uint32_t num_fields,
uint32_t num_uri_fields)
{
PusherMapping field_to_pusher(num_fields);
PusherMapping uri_field_to_pusher(num_uri_fields);
uint32_t pusher_id = 0;
for (auto& push_context : push_contexts) {
field_to_pusher.add_mapping(push_context.get_fields(), pusher_id);
uri_field_to_pusher.add_mapping(push_context.get_uri_fields(), pusher_id);
++pusher_id;
}
std::vector<uint32_t> pushers;
for (auto& invert_context : invert_contexts) {
pushers.clear();
field_to_pusher.use_mapping(invert_context.get_fields(), pushers);
uri_field_to_pusher.use_mapping(invert_context.get_uri_fields(), pushers);
std::sort(pushers.begin(), pushers.end());
auto last = std::unique(pushers.begin(), pushers.end());
pushers.erase(last, pushers.end());
for (auto pusher : pushers) {
invert_context.add_pusher(pusher);
}
}
}
}
void
DocumentInverterContext::add_field(const DocumentType& doc_type, uint32_t fieldId)
{
assert(fieldId < _indexed_fields.size());
std::unique_ptr<Field> fp;
if ( ! doc_type.hasField(_schema.getIndexField(fieldId).getName())) {
LOG(error,
"Mismatch between documentdefinition and schema. "
"No field named '%s' from schema in document type '%s'",
_schema.getIndexField(fieldId).getName().c_str(),
doc_type.getName().c_str());
} else {
fp = std::make_unique<Field>(doc_type.getField(_schema.getIndexField(fieldId).getName()));
}
_indexed_fields[fieldId] = std::move(fp);
}
void
DocumentInverterContext::build_fields(const DocumentType& doc_type, const DataType *data_type)
{
_indexed_fields.clear();
_indexed_fields.resize(_schema.getNumIndexFields());
for (const auto & fi : _schema_index_fields._textFields) {
add_field(doc_type, fi);
}
for (const auto & fi : _schema_index_fields._uriFields) {
add_field(doc_type, fi._all);
}
_data_type = data_type;
}
DocumentInverterContext::DocumentInverterContext(const index::Schema& schema,
ISequencedTaskExecutor &invert_threads,
ISequencedTaskExecutor &push_threads,
IFieldIndexCollection& field_indexes)
: _schema(schema),
_indexed_fields(),
_data_type(nullptr),
_schema_index_fields(),
_invert_threads(invert_threads),
_push_threads(push_threads),
_field_indexes(field_indexes),
_invert_contexts(),
_push_contexts()
{
_schema_index_fields.setup(schema);
setup_contexts();
}
DocumentInverterContext::~DocumentInverterContext() = default;
void
DocumentInverterContext::set_data_type(const Document& doc)
{
const DataType *data_type(doc.getDataType());
if (_indexed_fields.empty() || _data_type != data_type) {
build_fields(doc.getType(), data_type);
}
}
std::unique_ptr<document::FieldValue>
DocumentInverterContext::get_field_value(const Document& doc, uint32_t field_id) const
{
const Field *const field(_indexed_fields[field_id].get());
if (field != nullptr) {
return doc.getValue(*field);
}
return {};
}
void
DocumentInverterContext::setup_contexts()
{
make_contexts(_schema_index_fields, _invert_threads, _invert_contexts);
make_contexts(_schema_index_fields, _push_threads, _push_contexts);
connect_contexts(_invert_contexts, _push_contexts, _schema.getNumIndexFields(), _schema_index_fields._uriFields.size());
}
}
|
