rmmseg-cpp-traditional 0.0.2
Sign up to get free protection for your applications and to get access to all the features.
- data/.gitignore +17 -0
- data/Gemfile +4 -0
- data/History.txt +21 -0
- data/LICENSE.txt +22 -0
- data/Manifest.txt +43 -0
- data/README +111 -0
- data/README.md +29 -0
- data/Rakefile +19 -0
- data/bin/rmmseg +63 -0
- data/data/chars.dic +12638 -0
- data/data/words.dic +120308 -0
- data/ext/rmmseg/algor.cpp +222 -0
- data/ext/rmmseg/algor.h +80 -0
- data/ext/rmmseg/chunk.h +59 -0
- data/ext/rmmseg/dict.cpp +230 -0
- data/ext/rmmseg/dict.h +34 -0
- data/ext/rmmseg/extconf.rb +17 -0
- data/ext/rmmseg/memory.cpp +9 -0
- data/ext/rmmseg/memory.h +43 -0
- data/ext/rmmseg/rmmseg.cpp +263 -0
- data/ext/rmmseg/rules.h +86 -0
- data/ext/rmmseg/token.h +19 -0
- data/ext/rmmseg/word.h +44 -0
- data/lib/rmmseg/dictionary.rb +59 -0
- data/lib/rmmseg/ferret.rb +64 -0
- data/lib/rmmseg-cpp-traditional/version.rb +7 -0
- data/lib/rmmseg-cpp-traditional.rb +9 -0
- data/lib/rmmseg.rb +3 -0
- data/misc/convert.rb +114 -0
- data/misc/ferret_example.rb +59 -0
- data/misc/homepage.erb +196 -0
- data/misc/homepage.html +1212 -0
- data/rmmseg-cpp-traditional.gemspec +19 -0
- data/spec/rmmseg_spec.rb +8 -0
- data/spec/spec_helper.rb +17 -0
- data/tasks/ann.rake +81 -0
- data/tasks/bones.rake +21 -0
- data/tasks/gem.rake +126 -0
- data/tasks/git.rake +41 -0
- data/tasks/homepage.rake +15 -0
- data/tasks/manifest.rake +49 -0
- data/tasks/notes.rake +28 -0
- data/tasks/post_load.rake +39 -0
- data/tasks/rdoc.rake +51 -0
- data/tasks/rubyforge.rake +58 -0
- data/tasks/setup.rb +268 -0
- data/tasks/spec.rake +55 -0
- data/tasks/svn.rake +48 -0
- data/tasks/test.rake +38 -0
- data/test/test_rmmseg.rb +0 -0
- metadata +116 -0
@@ -0,0 +1,222 @@
|
|
1
|
+
#include <cctype>
|
2
|
+
#include <cassert>
|
3
|
+
|
4
|
+
#include "rules.h"
|
5
|
+
#include "algor.h"
|
6
|
+
|
7
|
+
using namespace std;
|
8
|
+
|
9
|
+
namespace rmmseg
|
10
|
+
{
|
11
|
+
Token Algorithm::next_token()
|
12
|
+
{
|
13
|
+
do
|
14
|
+
{
|
15
|
+
if (m_pos >= m_text_length)
|
16
|
+
return Token(NULL, 0);
|
17
|
+
|
18
|
+
Token tk(NULL, 0);
|
19
|
+
int len = next_char();
|
20
|
+
if (len == 1)
|
21
|
+
tk = get_basic_latin_word();
|
22
|
+
else
|
23
|
+
tk = get_cjk_word(len);
|
24
|
+
if (tk.length > 0)
|
25
|
+
return tk;
|
26
|
+
}
|
27
|
+
while (true);
|
28
|
+
}
|
29
|
+
|
30
|
+
Token Algorithm::get_basic_latin_word()
|
31
|
+
{
|
32
|
+
int len = 1;
|
33
|
+
int start, end;
|
34
|
+
|
35
|
+
// Skip pre-word whitespaces and punctuations
|
36
|
+
while (m_pos < m_text_length)
|
37
|
+
{
|
38
|
+
if (len > 1)
|
39
|
+
break;
|
40
|
+
if (isalnum(m_text[m_pos]))
|
41
|
+
break;
|
42
|
+
m_pos++;
|
43
|
+
len = next_char();
|
44
|
+
}
|
45
|
+
|
46
|
+
start = m_pos;
|
47
|
+
while (m_pos < m_text_length)
|
48
|
+
{
|
49
|
+
if (len > 1)
|
50
|
+
break;
|
51
|
+
if (!isalnum(m_text[m_pos]))
|
52
|
+
break;
|
53
|
+
m_pos++;
|
54
|
+
len = next_char();
|
55
|
+
}
|
56
|
+
end = m_pos;
|
57
|
+
|
58
|
+
// Skip post-word whitespaces and punctuations
|
59
|
+
while (m_pos < m_text_length)
|
60
|
+
{
|
61
|
+
if (len > 1)
|
62
|
+
break;
|
63
|
+
if (isalnum(m_text[m_pos]))
|
64
|
+
break;
|
65
|
+
m_pos++;
|
66
|
+
len = next_char();
|
67
|
+
}
|
68
|
+
|
69
|
+
return Token(m_text+start, end-start);
|
70
|
+
}
|
71
|
+
|
72
|
+
Token Algorithm::get_cjk_word(int len)
|
73
|
+
{
|
74
|
+
vector<Chunk> chunks = create_chunks();
|
75
|
+
|
76
|
+
if (chunks.size() > 1)
|
77
|
+
mm_filter(chunks);
|
78
|
+
if (chunks.size() > 1)
|
79
|
+
lawl_filter(chunks);
|
80
|
+
if (chunks.size() > 1)
|
81
|
+
svwl_filter(chunks);
|
82
|
+
if (chunks.size() > 1)
|
83
|
+
lsdmfocw_filter(chunks);
|
84
|
+
|
85
|
+
if (chunks.size() < 1)
|
86
|
+
return Token(NULL, 0);
|
87
|
+
|
88
|
+
Token token(m_text+m_pos, chunks[0].words[0]->nbytes);
|
89
|
+
m_pos += chunks[0].words[0]->nbytes;
|
90
|
+
return token;
|
91
|
+
}
|
92
|
+
|
93
|
+
vector<Chunk> Algorithm::create_chunks()
|
94
|
+
{
|
95
|
+
vector<Chunk> chunks;
|
96
|
+
Chunk chunk;
|
97
|
+
Word *w1, *w2, *w3;
|
98
|
+
|
99
|
+
int orig_pos = m_pos;
|
100
|
+
typedef vector<Word *> vec_t;
|
101
|
+
typedef vec_t::iterator it_t;
|
102
|
+
|
103
|
+
vec_t words1 = find_match_words();
|
104
|
+
for (it_t i1 = words1.begin();
|
105
|
+
i1 != words1.end();
|
106
|
+
++i1)
|
107
|
+
{
|
108
|
+
w1 = *i1;
|
109
|
+
chunk.words[0] = w1;
|
110
|
+
m_pos += w1->nbytes;
|
111
|
+
if (m_pos < m_text_length)
|
112
|
+
{
|
113
|
+
vec_t words2 = find_match_words();
|
114
|
+
for (it_t i2 = words2.begin();
|
115
|
+
i2 != words2.end();
|
116
|
+
++i2)
|
117
|
+
{
|
118
|
+
w2 = *i2;
|
119
|
+
chunk.words[1] = w2;
|
120
|
+
m_pos += w2->nbytes;
|
121
|
+
if (m_pos < m_text_length)
|
122
|
+
{
|
123
|
+
vec_t words3 = find_match_words();
|
124
|
+
for (it_t i3 = words3.begin();
|
125
|
+
i3 != words3.end();
|
126
|
+
++i3)
|
127
|
+
{
|
128
|
+
w3 = *i3;
|
129
|
+
if (w3->length == -1) // tmp word
|
130
|
+
{
|
131
|
+
chunk.n = 2;
|
132
|
+
}
|
133
|
+
else
|
134
|
+
{
|
135
|
+
chunk.n = 3;
|
136
|
+
chunk.words[2] = w3;
|
137
|
+
}
|
138
|
+
chunks.push_back(chunk);
|
139
|
+
}
|
140
|
+
}
|
141
|
+
else if (m_pos == m_text_length)
|
142
|
+
{
|
143
|
+
chunk.n = 2;
|
144
|
+
chunks.push_back(chunk);
|
145
|
+
}
|
146
|
+
m_pos -= w2->nbytes;
|
147
|
+
}
|
148
|
+
}
|
149
|
+
else if (m_pos == m_text_length)
|
150
|
+
{
|
151
|
+
chunk.n = 1;
|
152
|
+
chunks.push_back(chunk);
|
153
|
+
}
|
154
|
+
m_pos -= w1->nbytes;
|
155
|
+
}
|
156
|
+
|
157
|
+
m_pos = orig_pos;
|
158
|
+
return chunks;
|
159
|
+
}
|
160
|
+
|
161
|
+
int Algorithm::next_char()
|
162
|
+
{
|
163
|
+
// ONLY for UTF-8
|
164
|
+
unsigned char ch = m_text[m_pos];
|
165
|
+
if (ch >= 0xC0 && ch <= 0xDF)
|
166
|
+
{
|
167
|
+
return min(2, m_text_length-m_pos);
|
168
|
+
}
|
169
|
+
if (ch >= 0xE0 && ch <= 0xEF)
|
170
|
+
{
|
171
|
+
return min(3, m_text_length-m_pos);
|
172
|
+
}
|
173
|
+
return 1;
|
174
|
+
}
|
175
|
+
|
176
|
+
vector<Word *> Algorithm::find_match_words()
|
177
|
+
{
|
178
|
+
for (int i = 0; i < match_cache_size; ++i)
|
179
|
+
if (m_match_cache[i].first == m_pos)
|
180
|
+
return m_match_cache[i].second;
|
181
|
+
|
182
|
+
vector<Word *> words;
|
183
|
+
Word *word;
|
184
|
+
int orig_pos = m_pos;
|
185
|
+
int n = 0, len;
|
186
|
+
|
187
|
+
while (m_pos < m_text_length)
|
188
|
+
{
|
189
|
+
if (n >= max_word_length())
|
190
|
+
break;
|
191
|
+
len = next_char();
|
192
|
+
if (len <= 1)
|
193
|
+
break;
|
194
|
+
|
195
|
+
m_pos += len;
|
196
|
+
n++;
|
197
|
+
|
198
|
+
word = dict::get(m_text+orig_pos, m_pos-orig_pos);
|
199
|
+
if (word)
|
200
|
+
words.push_back(word);
|
201
|
+
}
|
202
|
+
|
203
|
+
m_pos = orig_pos;
|
204
|
+
|
205
|
+
if (words.empty())
|
206
|
+
{
|
207
|
+
word = get_tmp_word();
|
208
|
+
word->nbytes = next_char();
|
209
|
+
word->length = -1;
|
210
|
+
strncpy(word->text, m_text+m_pos, word->nbytes);
|
211
|
+
word->text[word->nbytes] = '\0';
|
212
|
+
words.push_back(word);
|
213
|
+
}
|
214
|
+
|
215
|
+
m_match_cache[m_match_cache_i] = make_pair(m_pos, words);
|
216
|
+
m_match_cache_i++;
|
217
|
+
if (m_match_cache_i >= match_cache_size)
|
218
|
+
m_match_cache_i = 0;
|
219
|
+
|
220
|
+
return words;
|
221
|
+
}
|
222
|
+
}
|
data/ext/rmmseg/algor.h
ADDED
@@ -0,0 +1,80 @@
|
|
1
|
+
#ifndef _ALGORITHM_H_
|
2
|
+
#define _ALGORITHM_H_
|
3
|
+
|
4
|
+
#include <vector>
|
5
|
+
|
6
|
+
#include "chunk.h"
|
7
|
+
#include "token.h"
|
8
|
+
#include "dict.h"
|
9
|
+
|
10
|
+
/**
|
11
|
+
* The Algorithm of MMSeg use four rules:
|
12
|
+
* - Maximum matching rule
|
13
|
+
* - Largest average word length rule
|
14
|
+
* - Smallest variance of word length rule
|
15
|
+
* - Largest sum of degree of morphemic freedom of one-character
|
16
|
+
* words rule
|
17
|
+
*/
|
18
|
+
|
19
|
+
namespace rmmseg
|
20
|
+
{
|
21
|
+
class Algorithm
|
22
|
+
{
|
23
|
+
public:
|
24
|
+
Algorithm(const char *text, int length)
|
25
|
+
:m_text(text), m_pos(0),
|
26
|
+
m_text_length(length),
|
27
|
+
m_tmp_words_i(0),
|
28
|
+
m_match_cache_i(0)
|
29
|
+
{
|
30
|
+
for (int i = 0; i < match_cache_size; ++i)
|
31
|
+
m_match_cache[i].first = -1;
|
32
|
+
}
|
33
|
+
|
34
|
+
Token next_token();
|
35
|
+
|
36
|
+
const char *get_text() const
|
37
|
+
{
|
38
|
+
return m_text;
|
39
|
+
}
|
40
|
+
|
41
|
+
private:
|
42
|
+
Token get_basic_latin_word();
|
43
|
+
Token get_cjk_word(int);
|
44
|
+
|
45
|
+
std::vector<Chunk> create_chunks();
|
46
|
+
int next_word();
|
47
|
+
int next_char();
|
48
|
+
std::vector<Word *> find_match_words();
|
49
|
+
int max_word_length() { return 4; }
|
50
|
+
|
51
|
+
|
52
|
+
const char *m_text;
|
53
|
+
int m_pos;
|
54
|
+
int m_text_length;
|
55
|
+
|
56
|
+
/* tmp words are only for 1-char words which
|
57
|
+
* are not exist in the dictionary. It's length
|
58
|
+
* value will be set to -1 to indicate it is
|
59
|
+
* a tmp word. */
|
60
|
+
Word *get_tmp_word()
|
61
|
+
{
|
62
|
+
if (m_tmp_words_i >= max_tmp_words)
|
63
|
+
m_tmp_words_i = 0; // round wrap
|
64
|
+
return &m_tmp_words[m_tmp_words_i++];
|
65
|
+
}
|
66
|
+
|
67
|
+
/* related to max_word_length and match_cache_size */
|
68
|
+
static const int max_tmp_words = 64;
|
69
|
+
Word m_tmp_words[max_tmp_words];
|
70
|
+
int m_tmp_words_i;
|
71
|
+
|
72
|
+
/* match word caches */
|
73
|
+
static const int match_cache_size = 3;
|
74
|
+
typedef std::pair<int, std::vector<Word *> > match_cache_t;
|
75
|
+
match_cache_t m_match_cache[match_cache_size];
|
76
|
+
int m_match_cache_i;
|
77
|
+
};
|
78
|
+
}
|
79
|
+
|
80
|
+
#endif /* _ALGORITHM_H_ */
|
data/ext/rmmseg/chunk.h
ADDED
@@ -0,0 +1,59 @@
|
|
1
|
+
#ifndef _CHUNK_H_
|
2
|
+
#define _CHUNK_H_
|
3
|
+
|
4
|
+
#include <cmath>
|
5
|
+
|
6
|
+
#include "word.h"
|
7
|
+
|
8
|
+
namespace rmmseg
|
9
|
+
{
|
10
|
+
/**
|
11
|
+
* A chunk stores 3 (or less) successive words.
|
12
|
+
*/
|
13
|
+
struct Chunk
|
14
|
+
{
|
15
|
+
int total_length() const
|
16
|
+
{
|
17
|
+
int len = 0;
|
18
|
+
for (int i = 0; i < n; ++i)
|
19
|
+
len += std::abs(words[i]->length);
|
20
|
+
//if (words[i]->length == -1) /* tmp word */
|
21
|
+
// len += 1;
|
22
|
+
//else
|
23
|
+
// len += words[i]->length;
|
24
|
+
return len;
|
25
|
+
}
|
26
|
+
double average_length() const
|
27
|
+
{
|
28
|
+
return ((double)total_length())/n;
|
29
|
+
}
|
30
|
+
double variance() const
|
31
|
+
{
|
32
|
+
double avg = average_length();
|
33
|
+
double sqr_sum = 0;
|
34
|
+
double tmp;
|
35
|
+
for (int i = 0; i < n; ++i)
|
36
|
+
{
|
37
|
+
tmp = std::abs(words[i]->length);
|
38
|
+
//if (tmp == -1)
|
39
|
+
// tmp = 1;
|
40
|
+
tmp = tmp-avg;
|
41
|
+
sqr_sum += tmp*tmp;
|
42
|
+
}
|
43
|
+
return std::sqrt(sqr_sum);
|
44
|
+
}
|
45
|
+
int degree_of_morphemic_freedom() const
|
46
|
+
{
|
47
|
+
int sum = 0;
|
48
|
+
for (int i = 0; i < n; ++i)
|
49
|
+
sum += words[i]->freq;
|
50
|
+
return sum;
|
51
|
+
}
|
52
|
+
|
53
|
+
|
54
|
+
int n;
|
55
|
+
Word *words[3];
|
56
|
+
};
|
57
|
+
}
|
58
|
+
|
59
|
+
#endif /* _CHUNK_H_ */
|
data/ext/rmmseg/dict.cpp
ADDED
@@ -0,0 +1,230 @@
|
|
1
|
+
#include <cstdio>
|
2
|
+
|
3
|
+
#include "dict.h"
|
4
|
+
|
5
|
+
using namespace std;
|
6
|
+
|
7
|
+
namespace rmmseg
|
8
|
+
{
|
9
|
+
struct Entry
|
10
|
+
{
|
11
|
+
Word *word;
|
12
|
+
Entry *next;
|
13
|
+
};
|
14
|
+
|
15
|
+
const size_t init_size = 262147;
|
16
|
+
const size_t max_density = 5;
|
17
|
+
/*
|
18
|
+
Table of prime numbers 2^n+a, 2<=n<=30.
|
19
|
+
*/
|
20
|
+
static size_t primes[] = {
|
21
|
+
524288 + 21,
|
22
|
+
1048576 + 7,
|
23
|
+
2097152 + 17,
|
24
|
+
4194304 + 15,
|
25
|
+
8388608 + 9,
|
26
|
+
16777216 + 43,
|
27
|
+
33554432 + 35,
|
28
|
+
67108864 + 15,
|
29
|
+
134217728 + 29,
|
30
|
+
268435456 + 3,
|
31
|
+
536870912 + 11,
|
32
|
+
1073741824 + 85,
|
33
|
+
};
|
34
|
+
|
35
|
+
|
36
|
+
static size_t n_bins = init_size;
|
37
|
+
static size_t n_entries = 0;
|
38
|
+
static Entry **bins = static_cast<Entry **>(std::calloc(init_size,
|
39
|
+
sizeof(Entry *)));
|
40
|
+
|
41
|
+
static size_t new_size()
|
42
|
+
{
|
43
|
+
for (size_t i = 0;
|
44
|
+
i < sizeof(primes)/sizeof(primes[0]);
|
45
|
+
++i)
|
46
|
+
{
|
47
|
+
if (primes[i] > n_bins)
|
48
|
+
{
|
49
|
+
return primes[i];
|
50
|
+
}
|
51
|
+
}
|
52
|
+
// TODO: raise exception here
|
53
|
+
return n_bins;
|
54
|
+
}
|
55
|
+
|
56
|
+
static unsigned int hash(const char *str, int len)
|
57
|
+
{
|
58
|
+
unsigned int key = 0;
|
59
|
+
while (len--)
|
60
|
+
{
|
61
|
+
key += *str++;
|
62
|
+
key += (key << 10);
|
63
|
+
key ^= (key >> 6);
|
64
|
+
}
|
65
|
+
key += (key << 3);
|
66
|
+
key ^= (key >> 11);
|
67
|
+
key += (key << 15);
|
68
|
+
return key;
|
69
|
+
}
|
70
|
+
|
71
|
+
static void rehash()
|
72
|
+
{
|
73
|
+
int new_n_bins = new_size();
|
74
|
+
Entry **new_bins = static_cast<Entry **>(calloc(new_n_bins,
|
75
|
+
sizeof(Entry *)));
|
76
|
+
Entry *entry, *next;
|
77
|
+
unsigned int hash_val;
|
78
|
+
|
79
|
+
for (size_t i = 0; i < n_bins; ++i)
|
80
|
+
{
|
81
|
+
entry = bins[i];
|
82
|
+
while (entry)
|
83
|
+
{
|
84
|
+
next = entry->next;
|
85
|
+
hash_val = hash(entry->word->text,
|
86
|
+
entry->word->nbytes) % new_n_bins;
|
87
|
+
entry->next = new_bins[hash_val];
|
88
|
+
new_bins[hash_val] = entry;
|
89
|
+
entry = next;
|
90
|
+
}
|
91
|
+
}
|
92
|
+
free(bins);
|
93
|
+
n_bins = new_n_bins;
|
94
|
+
bins = new_bins;
|
95
|
+
}
|
96
|
+
|
97
|
+
namespace dict
|
98
|
+
{
|
99
|
+
|
100
|
+
/**
|
101
|
+
* str: the base of the string
|
102
|
+
* len: length of the string (in bytes)
|
103
|
+
*
|
104
|
+
* str may be a substring of a big chunk of text thus not nul-terminated,
|
105
|
+
* so len is necessary here.
|
106
|
+
*/
|
107
|
+
Word *get(const char *str, int len)
|
108
|
+
{
|
109
|
+
unsigned int h = hash(str, len) % n_bins;
|
110
|
+
Entry *entry = bins[h];
|
111
|
+
if (!entry)
|
112
|
+
return NULL;
|
113
|
+
do
|
114
|
+
{
|
115
|
+
if (len == entry->word->nbytes &&
|
116
|
+
strncmp(str, entry->word->text, len) == 0)
|
117
|
+
return entry->word;
|
118
|
+
entry = entry->next;
|
119
|
+
}
|
120
|
+
while (entry);
|
121
|
+
|
122
|
+
return NULL;
|
123
|
+
}
|
124
|
+
|
125
|
+
void add(Word *word)
|
126
|
+
{
|
127
|
+
unsigned int hash_val = hash(word->text, word->nbytes);
|
128
|
+
unsigned int h = hash_val % n_bins;
|
129
|
+
Entry *entry = bins[h];
|
130
|
+
if (!entry)
|
131
|
+
{
|
132
|
+
if (n_entries/n_bins > max_density)
|
133
|
+
{
|
134
|
+
rehash();
|
135
|
+
h = hash_val % n_bins;
|
136
|
+
}
|
137
|
+
|
138
|
+
entry = static_cast<Entry *>(pool_alloc(sizeof(Entry)));
|
139
|
+
entry->word = word;
|
140
|
+
entry->next = NULL;
|
141
|
+
bins[h] = entry;
|
142
|
+
n_entries++;
|
143
|
+
return;
|
144
|
+
}
|
145
|
+
|
146
|
+
bool done = false;
|
147
|
+
do
|
148
|
+
{
|
149
|
+
if (word->nbytes == entry->word->nbytes &&
|
150
|
+
strncmp(word->text, entry->word->text, word->nbytes) == 0)
|
151
|
+
{
|
152
|
+
/* Overwriting. WARNING: the original Word object is
|
153
|
+
* permanently lost. This IS a memory leak, because
|
154
|
+
* the memory is allocated by pool_alloc. Instead of
|
155
|
+
* fixing this, tuning the dictionary file is a better
|
156
|
+
* idea
|
157
|
+
*/
|
158
|
+
entry->word = word;
|
159
|
+
done = true;
|
160
|
+
break;
|
161
|
+
}
|
162
|
+
entry = entry->next;
|
163
|
+
}
|
164
|
+
while (entry);
|
165
|
+
|
166
|
+
if (!done)
|
167
|
+
{
|
168
|
+
entry = static_cast<Entry *>(pool_alloc(sizeof(Entry)));
|
169
|
+
entry->word = word;
|
170
|
+
entry->next = bins[h];
|
171
|
+
bins[h] = entry;
|
172
|
+
n_entries++;
|
173
|
+
}
|
174
|
+
}
|
175
|
+
|
176
|
+
bool load_chars(const char *filename)
|
177
|
+
{
|
178
|
+
FILE *fp = fopen(filename, "r");
|
179
|
+
if (!fp)
|
180
|
+
{
|
181
|
+
return false;
|
182
|
+
}
|
183
|
+
|
184
|
+
const size_t buf_len = 24;
|
185
|
+
char buf[buf_len];
|
186
|
+
char *ptr;
|
187
|
+
|
188
|
+
while(fgets(buf, buf_len, fp))
|
189
|
+
{
|
190
|
+
// NOTE: there SHOULD be a newline at the end of the file
|
191
|
+
buf[strlen(buf)-1] = '\0'; // truncate the newline
|
192
|
+
ptr = strchr(buf, ' ');
|
193
|
+
if (!ptr)
|
194
|
+
continue; // illegal input
|
195
|
+
*ptr = '\0';
|
196
|
+
add(make_word(ptr+1, 1, atoi(buf)));
|
197
|
+
}
|
198
|
+
|
199
|
+
fclose(fp);
|
200
|
+
return true;
|
201
|
+
}
|
202
|
+
|
203
|
+
bool load_words(const char *filename)
|
204
|
+
{
|
205
|
+
FILE *fp = fopen(filename, "r");
|
206
|
+
if (!fp)
|
207
|
+
{
|
208
|
+
return false;
|
209
|
+
}
|
210
|
+
|
211
|
+
const int buf_len = 48;
|
212
|
+
char buf[buf_len];
|
213
|
+
char *ptr;
|
214
|
+
|
215
|
+
while(fgets(buf, buf_len, fp))
|
216
|
+
{
|
217
|
+
// NOTE: there SHOULD be a newline at the end of the file
|
218
|
+
buf[strlen(buf)-1] = '\0'; // truncate the newline
|
219
|
+
ptr = strchr(buf, ' ');
|
220
|
+
if (!ptr)
|
221
|
+
continue; // illegal input
|
222
|
+
*ptr = '\0';
|
223
|
+
add(make_word(ptr+1, atoi(buf), 0));
|
224
|
+
}
|
225
|
+
|
226
|
+
fclose(fp);
|
227
|
+
return true;
|
228
|
+
}
|
229
|
+
}
|
230
|
+
}
|
data/ext/rmmseg/dict.h
ADDED
@@ -0,0 +1,34 @@
|
|
1
|
+
#ifndef _DICT_H_
|
2
|
+
#define _DICT_H_
|
3
|
+
|
4
|
+
#include "word.h"
|
5
|
+
|
6
|
+
/**
|
7
|
+
* A dictionary is a hash table of
|
8
|
+
* - key: string
|
9
|
+
* - value: word
|
10
|
+
*
|
11
|
+
* Dictionary data can be loaded from files. Two type of dictionary
|
12
|
+
* files are supported:
|
13
|
+
* - character file: Each line contains a number and a character,
|
14
|
+
* the number is the frequency of the character.
|
15
|
+
* The frequency should NOT exceeds 65535.
|
16
|
+
* - word file: Each line contains a number and a word, the
|
17
|
+
* number is the character count of the word.
|
18
|
+
*/
|
19
|
+
|
20
|
+
namespace rmmseg
|
21
|
+
{
|
22
|
+
/* Instead of making a class with only one instance, i'll not
|
23
|
+
* bother to make it a class here. */
|
24
|
+
|
25
|
+
namespace dict
|
26
|
+
{
|
27
|
+
void add(Word *word);
|
28
|
+
bool load_chars(const char *filename);
|
29
|
+
bool load_words(const char *filename);
|
30
|
+
Word *get(const char *str, int len);
|
31
|
+
}
|
32
|
+
}
|
33
|
+
|
34
|
+
#endif /* _DICT_H_ */
|
@@ -0,0 +1,17 @@
|
|
1
|
+
require 'mkmf'
|
2
|
+
|
3
|
+
CONFIG['LDSHARED'] = CONFIG['LDSHARED'].sub(/^\$\(CC\)/, 'g++')
|
4
|
+
|
5
|
+
# if RUBY_PLATFORM =~ /darwin/
|
6
|
+
# # CONFIG['LDSHARED'] = 'g++ --dynamiclib -flat_namespace -undefined suppress'
|
7
|
+
# CONFIG['LDSHARED'] = 'g++ --dynamiclib'
|
8
|
+
# elsif RUBY_PLATFORM =~ /linux/
|
9
|
+
# CONFIG['LDSHARED'] = 'g++ -shared'
|
10
|
+
# end
|
11
|
+
|
12
|
+
if RUBY_PLATFORM =~ /darwin/
|
13
|
+
CONFIG['LDSHARED'] = 'g++ -dynamiclib -single_module -flat_namespace -undefined suppress'
|
14
|
+
end
|
15
|
+
|
16
|
+
$objs = ['algor.o', 'dict.o', 'memory.o', 'rmmseg.o']
|
17
|
+
create_makefile('rmmseg')
|
data/ext/rmmseg/memory.h
ADDED
@@ -0,0 +1,43 @@
|
|
1
|
+
#ifndef _MEMORY_H_
|
2
|
+
#define _MEMORY_H_
|
3
|
+
|
4
|
+
#include <cstdlib>
|
5
|
+
|
6
|
+
/**
|
7
|
+
* Pre-allocate a chunk of memory and allocate them in small pieces.
|
8
|
+
* Those memory are never freed after allocation. Used for persist
|
9
|
+
* data like dictionary contents that will never be destroyed unless
|
10
|
+
* the application exited.
|
11
|
+
*/
|
12
|
+
|
13
|
+
namespace rmmseg
|
14
|
+
{
|
15
|
+
const size_t REALLOC_SIZE = 2048; /* 2KB */
|
16
|
+
|
17
|
+
extern size_t _pool_size;
|
18
|
+
extern char *_pool_base;
|
19
|
+
|
20
|
+
inline void *pool_alloc(size_t len)
|
21
|
+
{
|
22
|
+
void *mem = _pool_base;
|
23
|
+
|
24
|
+
if (len <= _pool_size)
|
25
|
+
{
|
26
|
+
_pool_size -= len;
|
27
|
+
_pool_base += len;
|
28
|
+
return mem;
|
29
|
+
}
|
30
|
+
|
31
|
+
/* NOTE: the remaining memory is simply discard, which WILL
|
32
|
+
* cause memory leak. However, this function is not for allocating
|
33
|
+
* large object. Larger pre-alloc chunk size will also reduce the
|
34
|
+
* impact of this leak. So this is generally not a problem. */
|
35
|
+
_pool_base = static_cast<char *>(std::malloc(REALLOC_SIZE));
|
36
|
+
mem = _pool_base;
|
37
|
+
_pool_base += len;
|
38
|
+
_pool_size = REALLOC_SIZE - len;
|
39
|
+
return mem;
|
40
|
+
}
|
41
|
+
}
|
42
|
+
|
43
|
+
#endif /* _MEMORY_H_ */
|