/
dictionary_merger.h
314 lines (250 loc) · 9.26 KB
/
dictionary_merger.h
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
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
/* * keyvi - A key value store.
*
* Copyright 2015, 2016 Hendrik Muhs<hendrik.muhs@gmail.com>
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* dictionary_merger.h
*
* Created on: Feb 27, 2016
* Author: hendrik
*/
#ifndef KEYVI_DICTIONARY_DICTIONARY_MERGER_H_
#define KEYVI_DICTIONARY_DICTIONARY_MERGER_H_
#include <algorithm>
#include <fstream>
#include <functional>
#include <memory>
#include <queue>
#include <string>
#include <vector>
#include <boost/filesystem.hpp>
#include <msgpack.hpp>
#include "dictionary/fsa/automata.h"
#include "dictionary/fsa/entry_iterator.h"
#include "dictionary/fsa/generator_adapter.h"
#include "dictionary/fsa/internal/constants.h"
#include "util/configuration.h"
// #define ENABLE_TRACING
#include "dictionary/util/trace.h"
namespace keyvi {
namespace dictionary {
/**
* Exception class for dictionary merger
*/
struct merger_exception : public std::runtime_error {
using std::runtime_error::runtime_error;
};
struct MergeStats {
size_t number_of_keys_ = 0;
size_t deleted_keys_ = 0;
size_t updated_keys_ = 0;
};
template <class PersistenceT, class ValueStoreT = fsa::internal::NullValueStore>
class DictionaryMerger final {
using GeneratorAdapter = fsa::GeneratorAdapterInterface<PersistenceT, ValueStoreT>;
private:
class SegmentIterator {
using EntryIteratorPtr = std::shared_ptr<fsa::EntryIterator>;
public:
/**
*
* @param segment_index, merge segment index also used as a priority
* indicator
* when comparing two keys with the same value.
*/
SegmentIterator(const fsa::EntryIterator& e, size_t segment_index)
: entry_iterator_ptr_(std::make_shared<fsa::EntryIterator>(e)), segment_index_(segment_index) {}
bool operator<(const SegmentIterator& rhs) const {
// very important difference in semantics: we have to ensure that in case
// of equal key,
// the iterator with the higher index (priority) is taken
if (segment_index_ < rhs.segment_index_) {
return entryIterator() > rhs.entryIterator();
}
return rhs.entryIterator() < entryIterator();
}
operator bool() const { return entryIterator() != endIterator(); }
SegmentIterator& operator++() {
++(*entry_iterator_ptr_);
return *this;
}
const fsa::EntryIterator& entryIterator() const { return *entry_iterator_ptr_; }
const size_t segmentIndex() const { return segment_index_; }
private:
static const fsa::EntryIterator& endIterator() {
static fsa::EntryIterator end_it;
return end_it;
}
private:
EntryIteratorPtr entry_iterator_ptr_;
size_t segment_index_;
};
public:
/**
* Instantiate a dictionary merger.
*
* @params params merger parameters
*/
explicit DictionaryMerger(const keyvi::util::parameters_t& params = keyvi::util::parameters_t())
: dicts_to_merge_(), params_(params), stats_() {
params_[TEMPORARY_PATH_KEY] = keyvi::util::mapGetTemporaryPath(params);
append_merge_ = MERGE_APPEND == keyvi::util::mapGet<std::string>(params_, MERGE_MODE, "");
}
void Add(const std::string& filename) {
if (std::count(inputFiles_.begin(), inputFiles_.end(), filename)) {
throw std::invalid_argument("File is added already: " + filename);
}
fsa::automata_t fsa;
if (append_merge_) {
fsa.reset(new fsa::Automata(filename, loading_strategy_types::lazy, false));
} else {
fsa.reset(new fsa::Automata(filename));
}
if (fsa->GetValueStoreType() != ValueStoreT::GetValueStoreType()) {
throw std::invalid_argument("Dictionaries must have the same type.");
}
// check whether dictionary is completely empty
const auto segment_iterator = SegmentIterator(fsa::EntryIterator(fsa), segments_pqueue_.size());
if (!segment_iterator) {
return;
}
// push back deleted keys (list might be empty)
deleted_keys_.push_back(TryLoadDeletedKeys(filename));
segments_pqueue_.push(segment_iterator);
inputFiles_.push_back(filename);
dicts_to_merge_.push_back(fsa);
}
/**
* Set a custom manifest to be embedded into the index file.
*
* @param manifest as JSON string
*/
void SetManifestFromString(const std::string& manifest) { manifest_ = manifest; }
void Merge(const std::string& filename) {
Merge();
generator_->SetManifestFromString(manifest_);
generator_->WriteToFile(filename);
}
void Merge() {
size_t sparse_array_size_sum = 0;
for (auto fsa : dicts_to_merge_) {
sparse_array_size_sum += fsa->SparseArraySize();
}
ValueStoreT* value_store = append_merge_ ? new ValueStoreT(inputFiles_) : new ValueStoreT(params_);
generator_ = GeneratorAdapter::CreateGenerator(sparse_array_size_sum, params_, value_store);
std::string top_key;
while (!segments_pqueue_.empty()) {
auto segment_it = segments_pqueue_.top();
segments_pqueue_.pop();
top_key = segment_it.entryIterator().GetKey();
// check for same keys and merge only the most recent one
while (!segments_pqueue_.empty() && segments_pqueue_.top().entryIterator().operator==(top_key)) {
++stats_.updated_keys_;
auto to_inc = segments_pqueue_.top();
segments_pqueue_.pop();
if (++to_inc) {
TRACE("push iterator");
segments_pqueue_.push(to_inc);
}
}
if (!deleted_keys_[segment_it.segmentIndex()].empty() &&
top_key == deleted_keys_[segment_it.segmentIndex()].back()) {
deleted_keys_[segment_it.segmentIndex()].pop_back();
++stats_.deleted_keys_;
// check the other deleted_keys for duplicates
for (auto& deleted_keys : deleted_keys_) {
if (!deleted_keys.empty() && top_key == deleted_keys.back()) {
++stats_.deleted_keys_;
deleted_keys.pop_back();
}
}
} else {
fsa::ValueHandle handle;
handle.no_minimization = false;
// get the weight value, for now simple: does not require access to the
// value store itself
handle.weight = value_store->GetMergeWeight(segment_it.entryIterator().GetValueId());
if (append_merge_) {
handle.value_idx =
value_store->GetMergeValueId(segment_it.segmentIndex(), segment_it.entryIterator().GetValueId());
} else {
handle.value_idx =
value_store->GetValue(segment_it.entryIterator().GetFsa()->GetValueStore()->GetValueStorePayload(),
segment_it.entryIterator().GetValueId(), &handle.no_minimization);
}
TRACE("Add key: %s", top_key.c_str());
++stats_.number_of_keys_;
generator_->Add(std::move(top_key), handle);
}
if (++segment_it) {
segments_pqueue_.push(segment_it);
}
}
dicts_to_merge_.clear();
TRACE("finished iterating, do final compile.");
generator_->CloseFeeding();
}
void Write(std::ostream& stream) {
if (!generator_) {
throw merger_exception("not merged yet");
}
generator_->Write(stream);
}
void WriteToFile(const std::string& filename) {
if (!generator_) {
throw merger_exception("not merged yet");
}
generator_->WriteToFile(filename);
}
const MergeStats& GetStats() const { return stats_; }
private:
typename GeneratorAdapter::AdapterPtr generator_;
bool append_merge_ = false;
std::vector<fsa::automata_t> dicts_to_merge_;
std::vector<std::vector<std::string>> deleted_keys_;
std::vector<std::string> inputFiles_;
std::priority_queue<SegmentIterator> segments_pqueue_;
keyvi::util::parameters_t params_;
std::string manifest_ = std::string();
MergeStats stats_;
/**
* Load a file with deleted keys if it exists
*/
std::vector<std::string> TryLoadDeletedKeys(const std::string& filename) {
std::vector<std::string> deleted_keys;
boost::filesystem::path deleted_keys_file{filename};
deleted_keys_file += ".dk";
TRACE("check for deleted keys file: %s", deleted_keys_file.string().c_str());
std::ifstream deleted_keys_stream(deleted_keys_file.string(), std::ios::binary);
if (deleted_keys_stream.good()) {
TRACE("found deleted keys file");
{
// reads the buffer as 1 big chunk, could be improved
// msgpack v2.x provides a better interface (visitor)
std::stringstream buffer;
buffer << deleted_keys_stream.rdbuf();
msgpack::unpacked unpacked_object;
msgpack::unpack(unpacked_object, buffer.str().data(), buffer.str().size());
unpacked_object.get().convert(deleted_keys);
}
// sort in reverse order
std::sort(deleted_keys.begin(), deleted_keys.end(), std::greater<std::string>());
}
return deleted_keys;
}
};
} /* namespace dictionary */
} /* namespace keyvi */
#endif // KEYVI_DICTIONARY_DICTIONARY_MERGER_H_