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
|
#include "quering.hh"
unsigned char * read_binary_file(const char * filename, size_t filesize)
{
//Get filesize
int fd;
unsigned char * map;
fd = open(filename, O_RDONLY);
if (fd == -1) {
perror("Error opening file for reading");
exit(EXIT_FAILURE);
}
map = (unsigned char *)mmap(0, filesize, PROT_READ, MAP_SHARED, fd, 0);
if (map == MAP_FAILED) {
close(fd);
perror("Error mmapping the file");
exit(EXIT_FAILURE);
}
return map;
}
QueryEngine::QueryEngine(const char * filepath) : decoder(filepath)
{
//Create filepaths
std::string basepath(filepath);
std::string path_to_hashtable = basepath + "/probing_hash.dat";
std::string path_to_data_bin = basepath + "/binfile.dat";
std::string path_to_source_vocabid = basepath + "/source_vocabids";
///Source phrase vocabids
read_map(&source_vocabids, path_to_source_vocabid.c_str());
//Target phrase vocabIDs
vocabids = decoder.get_target_lookup_map();
//Read config file
std::string line;
std::ifstream config ((basepath + "/config").c_str());
//Check API version:
getline(config, line);
if (atoi(line.c_str()) != API_VERSION) {
std::cerr << "The ProbingPT API has changed, please rebinarize your phrase tables." << std::endl;
exit(EXIT_FAILURE);
}
//Get tablesize.
getline(config, line);
int tablesize = atoi(line.c_str());
//Number of scores
getline(config, line);
num_scores = atoi(line.c_str());
//do we have a reordering table
getline(config, line);
std::transform(line.begin(), line.end(), line.begin(), ::tolower); //Get the boolean in lowercase
is_reordering = false;
if (line == "true") {
is_reordering = true;
std::cerr << "WARNING. REORDERING TABLES NOT SUPPORTED YET." << std::endl;
}
config.close();
//Mmap binary table
struct stat filestatus;
stat(path_to_data_bin.c_str(), &filestatus);
binary_filesize = filestatus.st_size;
binary_mmaped = read_binary_file(path_to_data_bin.c_str(), binary_filesize);
//Read hashtable
table_filesize = Table::Size(tablesize, 1.2);
mem = readTable(path_to_hashtable.c_str(), table_filesize);
Table table_init(mem, table_filesize);
table = table_init;
std::cerr << "Initialized successfully! " << std::endl;
}
QueryEngine::~QueryEngine()
{
//Clear mmap content from memory.
munmap(binary_mmaped, binary_filesize);
munmap(mem, table_filesize);
}
std::pair<bool, std::vector<target_text> > QueryEngine::query(std::vector<uint64_t> source_phrase)
{
bool found;
std::vector<target_text> translation_entries;
const Entry * entry;
//TOO SLOW
//uint64_t key = util::MurmurHashNative(&source_phrase[0], source_phrase.size());
uint64_t key = 0;
for (int i = 0; i < source_phrase.size(); i++) {
key += (source_phrase[i] << i);
}
found = table.Find(key, entry);
if (found) {
//The phrase that was searched for was found! We need to get the translation entries.
//We will read the largest entry in bytes and then filter the unnecesarry with functions
//from line_splitter
uint64_t initial_index = entry -> GetValue();
unsigned int bytes_toread = entry -> bytes_toread;
//ASK HIEU FOR MORE EFFICIENT WAY TO DO THIS!
std::vector<unsigned char> encoded_text; //Assign to the vector the relevant portion of the array.
encoded_text.reserve(bytes_toread);
for (int i = 0; i < bytes_toread; i++) {
encoded_text.push_back(binary_mmaped[i+initial_index]);
}
//Get only the translation entries necessary
translation_entries = decoder.full_decode_line(encoded_text, num_scores);
}
std::pair<bool, std::vector<target_text> > output (found, translation_entries);
return output;
}
std::pair<bool, std::vector<target_text> > QueryEngine::query(StringPiece source_phrase)
{
bool found;
std::vector<target_text> translation_entries;
const Entry * entry;
//Convert source frase to VID
std::vector<uint64_t> source_phrase_vid = getVocabIDs(source_phrase);
//TOO SLOW
//uint64_t key = util::MurmurHashNative(&source_phrase_vid[0], source_phrase_vid.size());
uint64_t key = 0;
for (int i = 0; i < source_phrase_vid.size(); i++) {
key += (source_phrase_vid[i] << i);
}
found = table.Find(key, entry);
if (found) {
//The phrase that was searched for was found! We need to get the translation entries.
//We will read the largest entry in bytes and then filter the unnecesarry with functions
//from line_splitter
uint64_t initial_index = entry -> GetValue();
unsigned int bytes_toread = entry -> bytes_toread;
//At the end of the file we can't readd + largest_entry cause we get a segfault.
std::cerr << "Entry size is bytes is: " << bytes_toread << std::endl;
//ASK HIEU FOR MORE EFFICIENT WAY TO DO THIS!
std::vector<unsigned char> encoded_text; //Assign to the vector the relevant portion of the array.
encoded_text.reserve(bytes_toread);
for (int i = 0; i < bytes_toread; i++) {
encoded_text.push_back(binary_mmaped[i+initial_index]);
}
//Get only the translation entries necessary
translation_entries = decoder.full_decode_line(encoded_text, num_scores);
}
std::pair<bool, std::vector<target_text> > output (found, translation_entries);
return output;
}
void QueryEngine::printTargetInfo(std::vector<target_text> target_phrases)
{
int entries = target_phrases.size();
for (int i = 0; i<entries; i++) {
std::cout << "Entry " << i+1 << " of " << entries << ":" << std::endl;
//Print text
std::cout << getTargetWordsFromIDs(target_phrases[i].target_phrase, &vocabids) << "\t";
//Print probabilities:
for (int j = 0; j<target_phrases[i].prob.size(); j++) {
std::cout << target_phrases[i].prob[j] << " ";
}
std::cout << "\t";
//Print word_all1
for (int j = 0; j<target_phrases[i].word_all1.size(); j++) {
if (j%2 == 0) {
std::cout << (short)target_phrases[i].word_all1[j] << "-";
} else {
std::cout << (short)target_phrases[i].word_all1[j] << " ";
}
}
std::cout << std::endl;
}
}
|
