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
|
/*
Copyright (c) 2009 Yahoo! Inc. All rights reserved. The copyrights
embodied in the content of this file are licensed under the BSD
(revised) open source license
*/
#include <math.h>
#include <ctype.h>
#include "parse_example.h"
#include "hash.h"
#include "cache.h"
#include "unique_sort.h"
size_t hashstring (substring s, unsigned long h)
{
size_t ret = 0;
//trim leading whitespace
for(; *(s.start) <= 0x20 && s.start < s.end; s.start++);
//trim trailing white space
for(; *(s.end-1) <= 0x20 && s.end > s.start; s.end--);
char *p = s.start;
while (p != s.end)
if (isdigit(*p))
ret = 10*ret + *(p++) - '0';
else
return uniform_hash((unsigned char *)s.start, s.end - s.start, h);
return ret + h;
}
size_t hashall (substring s, unsigned long h)
{
return uniform_hash((unsigned char *)s.start, s.end - s.start, h);
}
hash_func_t getHasher(const string& s){
if (s=="strings")
return hashstring;
else if(s=="all")
return hashall;
else{
cerr << "Unknown hash function: " << s << ". Exiting " << endl;
exit(1);
}
}
void feature_value(substring &s, v_array<substring>& name, float &v)
{
tokenize(':', s, name);
switch (name.index()) {
case 0:
case 1:
v = 1.;
break;
case 2:
v = float_of_substring(name[1]);
if (isnan(v))
{
cerr << "error NaN value for feature: ";
cerr.write(name[0].start, name[0].end - name[0].start);
cerr << " terminating." << endl;
exit(1);
}
break;
default:
cerr << "example with a wierd name. What is ";
cerr.write(s.start, s.end - s.start);
cerr << "\n";
}
}
// new should be empty. --Alex
template<class T> void copy_v_array(v_array<T>& old, v_array<T>& new_va)
{
// allocate new memory for new.
new_va.begin = (T *) malloc(sizeof(T)*old.index());
// copy the old to the new.
memcpy(new_va.begin,old.begin,sizeof(T)*old.index());
// update pointers
new_va.end = new_va.begin + old.index();
new_va.end_array = new_va.end;
}
char* c_string_of_substring(substring s)
{
size_t len = s.end - s.start+1;
char* ret = (char *)calloc(len,sizeof(char));
memcpy(ret,s.start,len-1);
return ret;
}
char* copy(char* base)
{
size_t len = 0;
while (base[len++] != '\0');
char* ret = (char *)calloc(len,sizeof(char));
memcpy(ret,base,len);
return ret;
}
int read_features(parser* p, void* ex)
{
example* ae = (example*)ex;
char *line=NULL;
int num_chars = readto(*(p->input), line, '\n');
if (num_chars == 0)
return num_chars;
substring example = {line, line + num_chars};
tokenize('|', example, p->channels);
p->lp->default_label(ae->ld);
substring* feature_start = &(p->channels[1]);
if (*line == '|')
feature_start = &(p->channels[0]);
else
{
substring label_space = p->channels[0];
char* tab_location = safe_index(label_space.start, '\t', label_space.end);
if (tab_location != label_space.end)
label_space.start = tab_location+1;
tokenize(' ',label_space,p->words);
if (p->words.index() > 0 && p->words.last().end == label_space.end) //The last field is a tag, so record and strip it off
{
substring tag = p->words.pop();
push_many(ae->tag, tag.start, tag.end - tag.start);
}
p->lp->parse_label(ae->ld, p->words);
}
size_t mask = global.mask;
bool audit = global.audit;
for (substring* i = feature_start; i != p->channels.end; i++) {
substring channel = *i;
tokenize(' ',channel, p->words);
if (p->words.begin == p->words.end)
continue;
float channel_v = 1.;
size_t channel_hash;
char* base=NULL;
size_t index = 0;
bool new_index = false;
size_t feature_offset = 0;
if (channel.start[0] != ' ')//we have a nonanonymous namespace
{
feature_offset++;
feature_value(p->words[0], p->name, channel_v);
if (p->name.index() > 0) {
index = (unsigned char)(*p->name[0].start);
if (ae->atomics[index].begin == ae->atomics[index].end)
new_index = true;
}
if (audit)
base = c_string_of_substring(p->name[0]);
channel_hash = p->hasher(p->name[0], hash_base);
}
else
{
index = (unsigned char)' ';
if (ae->atomics[index].begin == ae->atomics[index].end)
new_index = true;
if (audit)
{
base = (char *)calloc(2,sizeof(char));
base[0]=' ';
base[1]='\0';
}
channel_hash = 0;
}
for (substring* i = p->words.begin+feature_offset; i != p->words.end; i++) {
float v;
feature_value(*i, p->name, v);
v *= channel_v;
size_t word_hash = (p->hasher(p->name[0], channel_hash)) & mask;
feature f = {v,word_hash};
ae->sum_feat_sq[index] += v*v;
push(ae->atomics[index], f);
}
if (new_index && ae->atomics[index].begin != ae->atomics[index].end)
push(ae->indices, index);
if (audit)
{
for (substring* i = p->words.begin+feature_offset; i != p->words.end; i++) {
float v;
feature_value(*i, p->name, v);
v *= channel_v;
size_t word_hash = (p->hasher(p->name[0], channel_hash)) & mask;
char* feature = c_string_of_substring(p->name[0]);
audit_data ad = {copy(base), feature, word_hash, v, true};
push(ae->audit_features[index], ad);
}
free(base);
}
}
return num_chars;
}
|
