melisa 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- data/README.md +11 -0
- data/ext/marisa/bindings/marisa-swig.cxx +253 -0
- data/ext/marisa/bindings/marisa-swig.h +183 -0
- data/ext/marisa/bindings/perl/marisa-swig.cxx +253 -0
- data/ext/marisa/bindings/perl/marisa-swig.h +183 -0
- data/ext/marisa/bindings/perl/marisa-swig_wrap.cxx +5160 -0
- data/ext/marisa/bindings/python/marisa-swig.cxx +253 -0
- data/ext/marisa/bindings/python/marisa-swig.h +183 -0
- data/ext/marisa/bindings/python/marisa-swig_wrap.cxx +6090 -0
- data/ext/marisa/bindings/ruby/extconf.rb +5 -0
- data/ext/marisa/bindings/ruby/marisa-swig.cxx +253 -0
- data/ext/marisa/bindings/ruby/marisa-swig.h +183 -0
- data/ext/marisa/bindings/ruby/marisa-swig_wrap.cxx +4708 -0
- data/ext/marisa/lib/marisa.h +14 -0
- data/ext/marisa/lib/marisa/agent.cc +51 -0
- data/ext/marisa/lib/marisa/agent.h +73 -0
- data/ext/marisa/lib/marisa/base.h +193 -0
- data/ext/marisa/lib/marisa/exception.h +82 -0
- data/ext/marisa/lib/marisa/grimoire/algorithm.h +26 -0
- data/ext/marisa/lib/marisa/grimoire/algorithm/sort.h +196 -0
- data/ext/marisa/lib/marisa/grimoire/intrin.h +115 -0
- data/ext/marisa/lib/marisa/grimoire/io.h +18 -0
- data/ext/marisa/lib/marisa/grimoire/io/mapper.cc +163 -0
- data/ext/marisa/lib/marisa/grimoire/io/mapper.h +67 -0
- data/ext/marisa/lib/marisa/grimoire/io/reader.cc +147 -0
- data/ext/marisa/lib/marisa/grimoire/io/reader.h +66 -0
- data/ext/marisa/lib/marisa/grimoire/io/writer.cc +148 -0
- data/ext/marisa/lib/marisa/grimoire/io/writer.h +65 -0
- data/ext/marisa/lib/marisa/grimoire/trie.h +16 -0
- data/ext/marisa/lib/marisa/grimoire/trie/cache.h +81 -0
- data/ext/marisa/lib/marisa/grimoire/trie/config.h +155 -0
- data/ext/marisa/lib/marisa/grimoire/trie/entry.h +82 -0
- data/ext/marisa/lib/marisa/grimoire/trie/header.h +61 -0
- data/ext/marisa/lib/marisa/grimoire/trie/history.h +65 -0
- data/ext/marisa/lib/marisa/grimoire/trie/key.h +228 -0
- data/ext/marisa/lib/marisa/grimoire/trie/louds-trie.cc +876 -0
- data/ext/marisa/lib/marisa/grimoire/trie/louds-trie.h +134 -0
- data/ext/marisa/lib/marisa/grimoire/trie/range.h +115 -0
- data/ext/marisa/lib/marisa/grimoire/trie/state.h +117 -0
- data/ext/marisa/lib/marisa/grimoire/trie/tail.cc +218 -0
- data/ext/marisa/lib/marisa/grimoire/trie/tail.h +72 -0
- data/ext/marisa/lib/marisa/grimoire/vector.h +18 -0
- data/ext/marisa/lib/marisa/grimoire/vector/bit-vector.cc +826 -0
- data/ext/marisa/lib/marisa/grimoire/vector/bit-vector.h +179 -0
- data/ext/marisa/lib/marisa/grimoire/vector/flat-vector.h +205 -0
- data/ext/marisa/lib/marisa/grimoire/vector/pop-count.h +110 -0
- data/ext/marisa/lib/marisa/grimoire/vector/rank-index.h +82 -0
- data/ext/marisa/lib/marisa/grimoire/vector/vector.h +256 -0
- data/ext/marisa/lib/marisa/iostream.h +18 -0
- data/ext/marisa/lib/marisa/key.h +85 -0
- data/ext/marisa/lib/marisa/keyset.cc +181 -0
- data/ext/marisa/lib/marisa/keyset.h +80 -0
- data/ext/marisa/lib/marisa/query.h +71 -0
- data/ext/marisa/lib/marisa/scoped-array.h +48 -0
- data/ext/marisa/lib/marisa/scoped-ptr.h +52 -0
- data/ext/marisa/lib/marisa/stdio.h +15 -0
- data/ext/marisa/lib/marisa/trie.cc +249 -0
- data/ext/marisa/lib/marisa/trie.h +64 -0
- data/ext/marisa/tests/base-test.cc +309 -0
- data/ext/marisa/tests/io-test.cc +252 -0
- data/ext/marisa/tests/marisa-assert.h +26 -0
- data/ext/marisa/tests/marisa-test.cc +388 -0
- data/ext/marisa/tests/trie-test.cc +507 -0
- data/ext/marisa/tests/vector-test.cc +466 -0
- data/ext/marisa/tools/cmdopt.cc +298 -0
- data/ext/marisa/tools/cmdopt.h +58 -0
- data/ext/marisa/tools/marisa-benchmark.cc +418 -0
- data/ext/marisa/tools/marisa-build.cc +206 -0
- data/ext/marisa/tools/marisa-common-prefix-search.cc +143 -0
- data/ext/marisa/tools/marisa-dump.cc +151 -0
- data/ext/marisa/tools/marisa-lookup.cc +110 -0
- data/ext/marisa/tools/marisa-predictive-search.cc +143 -0
- data/ext/marisa/tools/marisa-reverse-lookup.cc +110 -0
- data/lib/melisa.rb +7 -0
- data/lib/melisa/base_config_flags.rb +76 -0
- data/lib/melisa/bytes_trie.rb +55 -0
- data/lib/melisa/int_trie.rb +14 -0
- data/lib/melisa/search.rb +55 -0
- data/lib/melisa/trie.rb +96 -0
- data/lib/melisa/version.rb +3 -0
- data/melisa.gemspec +36 -0
- data/spec/base_config_flags_spec.rb +73 -0
- data/spec/bytes_trie_spec.rb +16 -0
- data/spec/int_trie_spec.rb +16 -0
- data/spec/search_spec.rb +29 -0
- data/spec/spec_helper.rb +1 -0
- data/spec/trie_spec.rb +30 -0
- metadata +207 -0
@@ -0,0 +1,82 @@
|
|
1
|
+
#ifndef MARISA_GRIMOIRE_TRIE_ENTRY_H_
|
2
|
+
#define MARISA_GRIMOIRE_TRIE_ENTRY_H_
|
3
|
+
|
4
|
+
#include "marisa/base.h"
|
5
|
+
|
6
|
+
namespace marisa {
|
7
|
+
namespace grimoire {
|
8
|
+
namespace trie {
|
9
|
+
|
10
|
+
class Entry {
|
11
|
+
public:
|
12
|
+
Entry()
|
13
|
+
: ptr_(static_cast<const char *>(NULL) - 1), length_(0), id_(0) {}
|
14
|
+
Entry(const Entry &entry)
|
15
|
+
: ptr_(entry.ptr_), length_(entry.length_), id_(entry.id_) {}
|
16
|
+
|
17
|
+
Entry &operator=(const Entry &entry) {
|
18
|
+
ptr_ = entry.ptr_;
|
19
|
+
length_ = entry.length_;
|
20
|
+
id_ = entry.id_;
|
21
|
+
return *this;
|
22
|
+
}
|
23
|
+
|
24
|
+
char operator[](std::size_t i) const {
|
25
|
+
MARISA_DEBUG_IF(i >= length_, MARISA_BOUND_ERROR);
|
26
|
+
return *(ptr_ - i);
|
27
|
+
}
|
28
|
+
|
29
|
+
void set_str(const char *ptr, std::size_t length) {
|
30
|
+
MARISA_DEBUG_IF((ptr == NULL) && (length != 0), MARISA_NULL_ERROR);
|
31
|
+
MARISA_DEBUG_IF(length > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
32
|
+
ptr_ = ptr + length - 1;
|
33
|
+
length_ = (UInt32)length;
|
34
|
+
}
|
35
|
+
void set_id(std::size_t id) {
|
36
|
+
MARISA_DEBUG_IF(id > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
37
|
+
id_ = (UInt32)id;
|
38
|
+
}
|
39
|
+
|
40
|
+
const char *ptr() const {
|
41
|
+
return ptr_ - length_ + 1;
|
42
|
+
}
|
43
|
+
std::size_t length() const {
|
44
|
+
return length_;
|
45
|
+
}
|
46
|
+
std::size_t id() const {
|
47
|
+
return id_;
|
48
|
+
}
|
49
|
+
|
50
|
+
class StringComparer {
|
51
|
+
public:
|
52
|
+
bool operator()(const Entry &lhs, const Entry &rhs) const {
|
53
|
+
for (std::size_t i = 0; i < lhs.length(); ++i) {
|
54
|
+
if (i == rhs.length()) {
|
55
|
+
return true;
|
56
|
+
}
|
57
|
+
if (lhs[i] != rhs[i]) {
|
58
|
+
return (UInt8)lhs[i] > (UInt8)rhs[i];
|
59
|
+
}
|
60
|
+
}
|
61
|
+
return lhs.length() > rhs.length();
|
62
|
+
}
|
63
|
+
};
|
64
|
+
|
65
|
+
class IDComparer {
|
66
|
+
public:
|
67
|
+
bool operator()(const Entry &lhs, const Entry &rhs) const {
|
68
|
+
return lhs.id_ < rhs.id_;
|
69
|
+
}
|
70
|
+
};
|
71
|
+
|
72
|
+
private:
|
73
|
+
const char *ptr_;
|
74
|
+
UInt32 length_;
|
75
|
+
UInt32 id_;
|
76
|
+
};
|
77
|
+
|
78
|
+
} // namespace trie
|
79
|
+
} // namespace grimoire
|
80
|
+
} // namespace marisa
|
81
|
+
|
82
|
+
#endif // MARISA_GRIMOIRE_TRIE_ENTRY_H_
|
@@ -0,0 +1,61 @@
|
|
1
|
+
#ifndef MARISA_GRIMOIRE_TRIE_HEADER_H_
|
2
|
+
#define MARISA_GRIMOIRE_TRIE_HEADER_H_
|
3
|
+
|
4
|
+
#include "marisa/grimoire/io.h"
|
5
|
+
|
6
|
+
namespace marisa {
|
7
|
+
namespace grimoire {
|
8
|
+
namespace trie {
|
9
|
+
|
10
|
+
class Header {
|
11
|
+
public:
|
12
|
+
enum {
|
13
|
+
HEADER_SIZE = 16
|
14
|
+
};
|
15
|
+
|
16
|
+
Header() {}
|
17
|
+
|
18
|
+
void map(Mapper &mapper) {
|
19
|
+
const char *ptr;
|
20
|
+
mapper.map(&ptr, HEADER_SIZE);
|
21
|
+
MARISA_THROW_IF(!test_header(ptr), MARISA_FORMAT_ERROR);
|
22
|
+
}
|
23
|
+
void read(Reader &reader) {
|
24
|
+
char buf[HEADER_SIZE];
|
25
|
+
reader.read(buf, HEADER_SIZE);
|
26
|
+
MARISA_THROW_IF(!test_header(buf), MARISA_FORMAT_ERROR);
|
27
|
+
}
|
28
|
+
void write(Writer &writer) const {
|
29
|
+
writer.write(get_header(), HEADER_SIZE);
|
30
|
+
}
|
31
|
+
|
32
|
+
std::size_t io_size() const {
|
33
|
+
return HEADER_SIZE;
|
34
|
+
}
|
35
|
+
|
36
|
+
private:
|
37
|
+
|
38
|
+
static const char *get_header() {
|
39
|
+
static const char buf[HEADER_SIZE] = "We love Marisa.";
|
40
|
+
return buf;
|
41
|
+
}
|
42
|
+
|
43
|
+
static bool test_header(const char *ptr) {
|
44
|
+
for (std::size_t i = 0; i < HEADER_SIZE; ++i) {
|
45
|
+
if (ptr[i] != get_header()[i]) {
|
46
|
+
return false;
|
47
|
+
}
|
48
|
+
}
|
49
|
+
return true;
|
50
|
+
}
|
51
|
+
|
52
|
+
// Disallows copy and assignment.
|
53
|
+
Header(const Header &);
|
54
|
+
Header &operator=(const Header &);
|
55
|
+
};
|
56
|
+
|
57
|
+
} // namespace trie
|
58
|
+
} // namespace marisa
|
59
|
+
} // namespace grimoire
|
60
|
+
|
61
|
+
#endif // MARISA_GRIMOIRE_TRIE_HEADER_H_
|
@@ -0,0 +1,65 @@
|
|
1
|
+
#ifndef MARISA_GRIMOIRE_TRIE_STATE_HISTORY_H_
|
2
|
+
#define MARISA_GRIMOIRE_TRIE_STATE_HISTORY_H_
|
3
|
+
|
4
|
+
#include "marisa/base.h"
|
5
|
+
|
6
|
+
namespace marisa {
|
7
|
+
namespace grimoire {
|
8
|
+
namespace trie {
|
9
|
+
|
10
|
+
class History {
|
11
|
+
public:
|
12
|
+
History()
|
13
|
+
: node_id_(0), louds_pos_(0), key_pos_(0),
|
14
|
+
link_id_(MARISA_INVALID_LINK_ID), key_id_(MARISA_INVALID_KEY_ID) {}
|
15
|
+
|
16
|
+
void set_node_id(std::size_t node_id) {
|
17
|
+
MARISA_DEBUG_IF(node_id > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
18
|
+
node_id_ = (UInt32)node_id;
|
19
|
+
}
|
20
|
+
void set_louds_pos(std::size_t louds_pos) {
|
21
|
+
MARISA_DEBUG_IF(louds_pos > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
22
|
+
louds_pos_ = (UInt32)louds_pos;
|
23
|
+
}
|
24
|
+
void set_key_pos(std::size_t key_pos) {
|
25
|
+
MARISA_DEBUG_IF(key_pos > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
26
|
+
key_pos_ = (UInt32)key_pos;
|
27
|
+
}
|
28
|
+
void set_link_id(std::size_t link_id) {
|
29
|
+
MARISA_DEBUG_IF(link_id > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
30
|
+
link_id_ = (UInt32)link_id;
|
31
|
+
}
|
32
|
+
void set_key_id(std::size_t key_id) {
|
33
|
+
MARISA_DEBUG_IF(key_id > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
34
|
+
key_id_ = (UInt32)key_id;
|
35
|
+
}
|
36
|
+
|
37
|
+
std::size_t node_id() const {
|
38
|
+
return node_id_;
|
39
|
+
}
|
40
|
+
std::size_t louds_pos() const {
|
41
|
+
return louds_pos_;
|
42
|
+
}
|
43
|
+
std::size_t key_pos() const {
|
44
|
+
return key_pos_;
|
45
|
+
}
|
46
|
+
std::size_t link_id() const {
|
47
|
+
return link_id_;
|
48
|
+
}
|
49
|
+
std::size_t key_id() const {
|
50
|
+
return key_id_;
|
51
|
+
}
|
52
|
+
|
53
|
+
private:
|
54
|
+
UInt32 node_id_;
|
55
|
+
UInt32 louds_pos_;
|
56
|
+
UInt32 key_pos_;
|
57
|
+
UInt32 link_id_;
|
58
|
+
UInt32 key_id_;
|
59
|
+
};
|
60
|
+
|
61
|
+
} // namespace trie
|
62
|
+
} // namespace grimoire
|
63
|
+
} // namespace marisa
|
64
|
+
|
65
|
+
#endif // MARISA_GRIMOIRE_TRIE_STATE_HISTORY_H_
|
@@ -0,0 +1,228 @@
|
|
1
|
+
#ifndef MARISA_GRIMOIRE_TRIE_KEY_H_
|
2
|
+
#define MARISA_GRIMOIRE_TRIE_KEY_H_
|
3
|
+
|
4
|
+
#include "marisa/base.h"
|
5
|
+
|
6
|
+
namespace marisa {
|
7
|
+
namespace grimoire {
|
8
|
+
namespace trie {
|
9
|
+
|
10
|
+
class Key {
|
11
|
+
public:
|
12
|
+
Key() : ptr_(NULL), length_(0), union_(), id_(0) {
|
13
|
+
union_.terminal = 0;
|
14
|
+
}
|
15
|
+
Key(const Key &entry)
|
16
|
+
: ptr_(entry.ptr_), length_(entry.length_),
|
17
|
+
union_(entry.union_), id_(entry.id_) {}
|
18
|
+
|
19
|
+
Key &operator=(const Key &entry) {
|
20
|
+
ptr_ = entry.ptr_;
|
21
|
+
length_ = entry.length_;
|
22
|
+
union_ = entry.union_;
|
23
|
+
id_ = entry.id_;
|
24
|
+
return *this;
|
25
|
+
}
|
26
|
+
|
27
|
+
char operator[](std::size_t i) const {
|
28
|
+
MARISA_DEBUG_IF(i >= length_, MARISA_BOUND_ERROR);
|
29
|
+
return ptr_[i];
|
30
|
+
}
|
31
|
+
|
32
|
+
void substr(std::size_t pos, std::size_t length) {
|
33
|
+
MARISA_DEBUG_IF(pos > length_, MARISA_BOUND_ERROR);
|
34
|
+
MARISA_DEBUG_IF(length > length_, MARISA_BOUND_ERROR);
|
35
|
+
MARISA_DEBUG_IF(pos > (length_ - length), MARISA_BOUND_ERROR);
|
36
|
+
ptr_ += pos;
|
37
|
+
length_ = (UInt32)length;
|
38
|
+
}
|
39
|
+
|
40
|
+
void set_str(const char *ptr, std::size_t length) {
|
41
|
+
MARISA_DEBUG_IF((ptr == NULL) && (length != 0), MARISA_NULL_ERROR);
|
42
|
+
MARISA_DEBUG_IF(length > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
43
|
+
ptr_ = ptr;
|
44
|
+
length_ = (UInt32)length;
|
45
|
+
}
|
46
|
+
void set_weight(float weight) {
|
47
|
+
union_.weight = weight;
|
48
|
+
}
|
49
|
+
void set_terminal(std::size_t terminal) {
|
50
|
+
MARISA_DEBUG_IF(terminal > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
51
|
+
union_.terminal = (UInt32)terminal;
|
52
|
+
}
|
53
|
+
void set_id(std::size_t id) {
|
54
|
+
MARISA_DEBUG_IF(id > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
55
|
+
id_ = (UInt32)id;
|
56
|
+
}
|
57
|
+
|
58
|
+
const char *ptr() const {
|
59
|
+
return ptr_;
|
60
|
+
}
|
61
|
+
std::size_t length() const {
|
62
|
+
return length_;
|
63
|
+
}
|
64
|
+
float weight() const {
|
65
|
+
return union_.weight;
|
66
|
+
}
|
67
|
+
std::size_t terminal() const {
|
68
|
+
return union_.terminal;
|
69
|
+
}
|
70
|
+
std::size_t id() const {
|
71
|
+
return id_;
|
72
|
+
}
|
73
|
+
|
74
|
+
private:
|
75
|
+
const char *ptr_;
|
76
|
+
UInt32 length_;
|
77
|
+
union Union {
|
78
|
+
float weight;
|
79
|
+
UInt32 terminal;
|
80
|
+
} union_;
|
81
|
+
UInt32 id_;
|
82
|
+
};
|
83
|
+
|
84
|
+
inline bool operator==(const Key &lhs, const Key &rhs) {
|
85
|
+
if (lhs.length() != rhs.length()) {
|
86
|
+
return false;
|
87
|
+
}
|
88
|
+
for (std::size_t i = 0; i < lhs.length(); ++i) {
|
89
|
+
if (lhs[i] != rhs[i]) {
|
90
|
+
return false;
|
91
|
+
}
|
92
|
+
}
|
93
|
+
return true;
|
94
|
+
}
|
95
|
+
|
96
|
+
inline bool operator!=(const Key &lhs, const Key &rhs) {
|
97
|
+
return !(lhs == rhs);
|
98
|
+
}
|
99
|
+
|
100
|
+
inline bool operator<(const Key &lhs, const Key &rhs) {
|
101
|
+
for (std::size_t i = 0; i < lhs.length(); ++i) {
|
102
|
+
if (i == rhs.length()) {
|
103
|
+
return false;
|
104
|
+
}
|
105
|
+
if (lhs[i] != rhs[i]) {
|
106
|
+
return (UInt8)lhs[i] < (UInt8)rhs[i];
|
107
|
+
}
|
108
|
+
}
|
109
|
+
return lhs.length() < rhs.length();
|
110
|
+
}
|
111
|
+
|
112
|
+
inline bool operator>(const Key &lhs, const Key &rhs) {
|
113
|
+
return rhs < lhs;
|
114
|
+
}
|
115
|
+
|
116
|
+
class ReverseKey {
|
117
|
+
public:
|
118
|
+
ReverseKey()
|
119
|
+
: ptr_(static_cast<const char *>(NULL) - 1),
|
120
|
+
length_(0), union_(), id_(0) {
|
121
|
+
union_.terminal = 0;
|
122
|
+
}
|
123
|
+
ReverseKey(const ReverseKey &entry)
|
124
|
+
: ptr_(entry.ptr_), length_(entry.length_),
|
125
|
+
union_(entry.union_), id_(entry.id_) {}
|
126
|
+
|
127
|
+
ReverseKey &operator=(const ReverseKey &entry) {
|
128
|
+
ptr_ = entry.ptr_;
|
129
|
+
length_ = entry.length_;
|
130
|
+
union_ = entry.union_;
|
131
|
+
id_ = entry.id_;
|
132
|
+
return *this;
|
133
|
+
}
|
134
|
+
|
135
|
+
char operator[](std::size_t i) const {
|
136
|
+
MARISA_DEBUG_IF(i >= length_, MARISA_BOUND_ERROR);
|
137
|
+
return *(ptr_ - i);
|
138
|
+
}
|
139
|
+
|
140
|
+
void substr(std::size_t pos, std::size_t length) {
|
141
|
+
MARISA_DEBUG_IF(pos > length_, MARISA_BOUND_ERROR);
|
142
|
+
MARISA_DEBUG_IF(length > length_, MARISA_BOUND_ERROR);
|
143
|
+
MARISA_DEBUG_IF(pos > (length_ - length), MARISA_BOUND_ERROR);
|
144
|
+
ptr_ -= pos;
|
145
|
+
length_ = (UInt32)length;
|
146
|
+
}
|
147
|
+
|
148
|
+
void set_str(const char *ptr, std::size_t length) {
|
149
|
+
MARISA_DEBUG_IF((ptr == NULL) && (length != 0), MARISA_NULL_ERROR);
|
150
|
+
MARISA_DEBUG_IF(length > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
151
|
+
ptr_ = ptr + length - 1;
|
152
|
+
length_ = (UInt32)length;
|
153
|
+
}
|
154
|
+
void set_weight(float weight) {
|
155
|
+
union_.weight = weight;
|
156
|
+
}
|
157
|
+
void set_terminal(std::size_t terminal) {
|
158
|
+
MARISA_DEBUG_IF(terminal > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
159
|
+
union_.terminal = (UInt32)terminal;
|
160
|
+
}
|
161
|
+
void set_id(std::size_t id) {
|
162
|
+
MARISA_DEBUG_IF(id > MARISA_UINT32_MAX, MARISA_SIZE_ERROR);
|
163
|
+
id_ = (UInt32)id;
|
164
|
+
}
|
165
|
+
|
166
|
+
const char *ptr() const {
|
167
|
+
return ptr_ - length_ + 1;
|
168
|
+
}
|
169
|
+
std::size_t length() const {
|
170
|
+
return length_;
|
171
|
+
}
|
172
|
+
float weight() const {
|
173
|
+
return union_.weight;
|
174
|
+
}
|
175
|
+
std::size_t terminal() const {
|
176
|
+
return union_.terminal;
|
177
|
+
}
|
178
|
+
std::size_t id() const {
|
179
|
+
return id_;
|
180
|
+
}
|
181
|
+
|
182
|
+
private:
|
183
|
+
const char *ptr_;
|
184
|
+
UInt32 length_;
|
185
|
+
union Union {
|
186
|
+
float weight;
|
187
|
+
UInt32 terminal;
|
188
|
+
} union_;
|
189
|
+
UInt32 id_;
|
190
|
+
};
|
191
|
+
|
192
|
+
inline bool operator==(const ReverseKey &lhs, const ReverseKey &rhs) {
|
193
|
+
if (lhs.length() != rhs.length()) {
|
194
|
+
return false;
|
195
|
+
}
|
196
|
+
for (std::size_t i = 0; i < lhs.length(); ++i) {
|
197
|
+
if (lhs[i] != rhs[i]) {
|
198
|
+
return false;
|
199
|
+
}
|
200
|
+
}
|
201
|
+
return true;
|
202
|
+
}
|
203
|
+
|
204
|
+
inline bool operator!=(const ReverseKey &lhs, const ReverseKey &rhs) {
|
205
|
+
return !(lhs == rhs);
|
206
|
+
}
|
207
|
+
|
208
|
+
inline bool operator<(const ReverseKey &lhs, const ReverseKey &rhs) {
|
209
|
+
for (std::size_t i = 0; i < lhs.length(); ++i) {
|
210
|
+
if (i == rhs.length()) {
|
211
|
+
return false;
|
212
|
+
}
|
213
|
+
if (lhs[i] != rhs[i]) {
|
214
|
+
return (UInt8)lhs[i] < (UInt8)rhs[i];
|
215
|
+
}
|
216
|
+
}
|
217
|
+
return lhs.length() < rhs.length();
|
218
|
+
}
|
219
|
+
|
220
|
+
inline bool operator>(const ReverseKey &lhs, const ReverseKey &rhs) {
|
221
|
+
return rhs < lhs;
|
222
|
+
}
|
223
|
+
|
224
|
+
} // namespace trie
|
225
|
+
} // namespace grimoire
|
226
|
+
} // namespace marisa
|
227
|
+
|
228
|
+
#endif // MARISA_GRIMOIRE_TRIE_KEY_H_
|
@@ -0,0 +1,876 @@
|
|
1
|
+
#include <algorithm>
|
2
|
+
#include <queue>
|
3
|
+
|
4
|
+
#include "marisa/grimoire/algorithm.h"
|
5
|
+
#include "marisa/grimoire/trie/header.h"
|
6
|
+
#include "marisa/grimoire/trie/range.h"
|
7
|
+
#include "marisa/grimoire/trie/state.h"
|
8
|
+
#include "marisa/grimoire/trie/louds-trie.h"
|
9
|
+
|
10
|
+
namespace marisa {
|
11
|
+
namespace grimoire {
|
12
|
+
namespace trie {
|
13
|
+
|
14
|
+
LoudsTrie::LoudsTrie()
|
15
|
+
: louds_(), terminal_flags_(), link_flags_(), bases_(), extras_(),
|
16
|
+
tail_(), next_trie_(), cache_(), cache_mask_(0), num_l1_nodes_(0),
|
17
|
+
config_(), mapper_() {}
|
18
|
+
|
19
|
+
LoudsTrie::~LoudsTrie() {}
|
20
|
+
|
21
|
+
void LoudsTrie::build(Keyset &keyset, int flags) {
|
22
|
+
Config config;
|
23
|
+
config.parse(flags);
|
24
|
+
|
25
|
+
LoudsTrie temp;
|
26
|
+
temp.build_(keyset, config);
|
27
|
+
swap(temp);
|
28
|
+
}
|
29
|
+
|
30
|
+
void LoudsTrie::map(Mapper &mapper) {
|
31
|
+
Header().map(mapper);
|
32
|
+
|
33
|
+
LoudsTrie temp;
|
34
|
+
temp.map_(mapper);
|
35
|
+
temp.mapper_.swap(mapper);
|
36
|
+
swap(temp);
|
37
|
+
}
|
38
|
+
|
39
|
+
void LoudsTrie::read(Reader &reader) {
|
40
|
+
Header().read(reader);
|
41
|
+
|
42
|
+
LoudsTrie temp;
|
43
|
+
temp.read_(reader);
|
44
|
+
swap(temp);
|
45
|
+
}
|
46
|
+
|
47
|
+
void LoudsTrie::write(Writer &writer) const {
|
48
|
+
Header().write(writer);
|
49
|
+
|
50
|
+
write_(writer);
|
51
|
+
}
|
52
|
+
|
53
|
+
bool LoudsTrie::lookup(Agent &agent) const {
|
54
|
+
MARISA_DEBUG_IF(!agent.has_state(), MARISA_STATE_ERROR);
|
55
|
+
|
56
|
+
State &state = agent.state();
|
57
|
+
state.lookup_init();
|
58
|
+
while (state.query_pos() < agent.query().length()) {
|
59
|
+
if (!find_child(agent)) {
|
60
|
+
return false;
|
61
|
+
}
|
62
|
+
}
|
63
|
+
if (!terminal_flags_[state.node_id()]) {
|
64
|
+
return false;
|
65
|
+
}
|
66
|
+
agent.set_key(agent.query().ptr(), agent.query().length());
|
67
|
+
agent.set_key(terminal_flags_.rank1(state.node_id()));
|
68
|
+
return true;
|
69
|
+
}
|
70
|
+
|
71
|
+
void LoudsTrie::reverse_lookup(Agent &agent) const {
|
72
|
+
MARISA_DEBUG_IF(!agent.has_state(), MARISA_STATE_ERROR);
|
73
|
+
MARISA_THROW_IF(agent.query().id() >= size(), MARISA_BOUND_ERROR);
|
74
|
+
|
75
|
+
State &state = agent.state();
|
76
|
+
state.reverse_lookup_init();
|
77
|
+
|
78
|
+
state.set_node_id(terminal_flags_.select1(agent.query().id()));
|
79
|
+
if (state.node_id() == 0) {
|
80
|
+
agent.set_key(state.key_buf().begin(), state.key_buf().size());
|
81
|
+
agent.set_key(agent.query().id());
|
82
|
+
return;
|
83
|
+
}
|
84
|
+
for ( ; ; ) {
|
85
|
+
if (link_flags_[state.node_id()]) {
|
86
|
+
const std::size_t prev_key_pos = state.key_buf().size();
|
87
|
+
restore(agent, get_link(state.node_id()));
|
88
|
+
std::reverse(state.key_buf().begin() + prev_key_pos,
|
89
|
+
state.key_buf().end());
|
90
|
+
} else {
|
91
|
+
state.key_buf().push_back((char)bases_[state.node_id()]);
|
92
|
+
}
|
93
|
+
|
94
|
+
if (state.node_id() <= num_l1_nodes_) {
|
95
|
+
std::reverse(state.key_buf().begin(), state.key_buf().end());
|
96
|
+
agent.set_key(state.key_buf().begin(), state.key_buf().size());
|
97
|
+
agent.set_key(agent.query().id());
|
98
|
+
return;
|
99
|
+
}
|
100
|
+
state.set_node_id(louds_.select1(state.node_id()) - state.node_id() - 1);
|
101
|
+
}
|
102
|
+
}
|
103
|
+
|
104
|
+
bool LoudsTrie::common_prefix_search(Agent &agent) const {
|
105
|
+
MARISA_DEBUG_IF(!agent.has_state(), MARISA_STATE_ERROR);
|
106
|
+
|
107
|
+
State &state = agent.state();
|
108
|
+
if (state.status_code() == MARISA_END_OF_COMMON_PREFIX_SEARCH) {
|
109
|
+
return false;
|
110
|
+
}
|
111
|
+
|
112
|
+
if (state.status_code() != MARISA_READY_TO_COMMON_PREFIX_SEARCH) {
|
113
|
+
state.common_prefix_search_init();
|
114
|
+
if (terminal_flags_[state.node_id()]) {
|
115
|
+
agent.set_key(agent.query().ptr(), state.query_pos());
|
116
|
+
agent.set_key(terminal_flags_.rank1(state.node_id()));
|
117
|
+
return true;
|
118
|
+
}
|
119
|
+
}
|
120
|
+
|
121
|
+
while (state.query_pos() < agent.query().length()) {
|
122
|
+
if (!find_child(agent)) {
|
123
|
+
state.set_status_code(MARISA_END_OF_COMMON_PREFIX_SEARCH);
|
124
|
+
return false;
|
125
|
+
} else if (terminal_flags_[state.node_id()]) {
|
126
|
+
agent.set_key(agent.query().ptr(), state.query_pos());
|
127
|
+
agent.set_key(terminal_flags_.rank1(state.node_id()));
|
128
|
+
return true;
|
129
|
+
}
|
130
|
+
}
|
131
|
+
state.set_status_code(MARISA_END_OF_COMMON_PREFIX_SEARCH);
|
132
|
+
return false;
|
133
|
+
}
|
134
|
+
|
135
|
+
bool LoudsTrie::predictive_search(Agent &agent) const {
|
136
|
+
MARISA_DEBUG_IF(!agent.has_state(), MARISA_STATE_ERROR);
|
137
|
+
|
138
|
+
State &state = agent.state();
|
139
|
+
if (state.status_code() == MARISA_END_OF_PREDICTIVE_SEARCH) {
|
140
|
+
return false;
|
141
|
+
}
|
142
|
+
|
143
|
+
if (state.status_code() != MARISA_READY_TO_PREDICTIVE_SEARCH) {
|
144
|
+
state.predictive_search_init();
|
145
|
+
while (state.query_pos() < agent.query().length()) {
|
146
|
+
if (!predictive_find_child(agent)) {
|
147
|
+
state.set_status_code(MARISA_END_OF_PREDICTIVE_SEARCH);
|
148
|
+
return false;
|
149
|
+
}
|
150
|
+
}
|
151
|
+
|
152
|
+
History history;
|
153
|
+
history.set_node_id(state.node_id());
|
154
|
+
history.set_key_pos(state.key_buf().size());
|
155
|
+
state.history().push_back(history);
|
156
|
+
state.set_history_pos(1);
|
157
|
+
|
158
|
+
if (terminal_flags_[state.node_id()]) {
|
159
|
+
agent.set_key(state.key_buf().begin(), state.key_buf().size());
|
160
|
+
agent.set_key(terminal_flags_.rank1(state.node_id()));
|
161
|
+
return true;
|
162
|
+
}
|
163
|
+
}
|
164
|
+
|
165
|
+
for ( ; ; ) {
|
166
|
+
if (state.history_pos() == state.history().size()) {
|
167
|
+
const History ¤t = state.history().back();
|
168
|
+
History next;
|
169
|
+
next.set_louds_pos(louds_.select0(current.node_id()) + 1);
|
170
|
+
next.set_node_id(next.louds_pos() - current.node_id() - 1);
|
171
|
+
state.history().push_back(next);
|
172
|
+
}
|
173
|
+
|
174
|
+
History &next = state.history()[state.history_pos()];
|
175
|
+
const bool link_flag = louds_[next.louds_pos()];
|
176
|
+
next.set_louds_pos(next.louds_pos() + 1);
|
177
|
+
if (link_flag) {
|
178
|
+
state.set_history_pos(state.history_pos() + 1);
|
179
|
+
if (link_flags_[next.node_id()]) {
|
180
|
+
next.set_link_id(update_link_id(next.link_id(), next.node_id()));
|
181
|
+
restore(agent, get_link(next.node_id(), next.link_id()));
|
182
|
+
} else {
|
183
|
+
state.key_buf().push_back((char)bases_[next.node_id()]);
|
184
|
+
}
|
185
|
+
next.set_key_pos(state.key_buf().size());
|
186
|
+
|
187
|
+
if (terminal_flags_[next.node_id()]) {
|
188
|
+
if (next.key_id() == MARISA_INVALID_KEY_ID) {
|
189
|
+
next.set_key_id(terminal_flags_.rank1(next.node_id()));
|
190
|
+
} else {
|
191
|
+
next.set_key_id(next.key_id() + 1);
|
192
|
+
}
|
193
|
+
agent.set_key(state.key_buf().begin(), state.key_buf().size());
|
194
|
+
agent.set_key(next.key_id());
|
195
|
+
return true;
|
196
|
+
}
|
197
|
+
} else if (state.history_pos() != 1) {
|
198
|
+
History ¤t = state.history()[state.history_pos() - 1];
|
199
|
+
current.set_node_id(current.node_id() + 1);
|
200
|
+
const History &prev =
|
201
|
+
state.history()[state.history_pos() - 2];
|
202
|
+
state.key_buf().resize(prev.key_pos());
|
203
|
+
state.set_history_pos(state.history_pos() - 1);
|
204
|
+
} else {
|
205
|
+
state.set_status_code(MARISA_END_OF_PREDICTIVE_SEARCH);
|
206
|
+
return false;
|
207
|
+
}
|
208
|
+
}
|
209
|
+
}
|
210
|
+
|
211
|
+
std::size_t LoudsTrie::total_size() const {
|
212
|
+
return louds_.total_size() + terminal_flags_.total_size()
|
213
|
+
+ link_flags_.total_size() + bases_.total_size()
|
214
|
+
+ extras_.total_size() + tail_.total_size()
|
215
|
+
+ ((next_trie_.get() != NULL) ? next_trie_->total_size() : 0)
|
216
|
+
+ cache_.total_size();
|
217
|
+
}
|
218
|
+
|
219
|
+
std::size_t LoudsTrie::io_size() const {
|
220
|
+
return Header().io_size() + louds_.io_size()
|
221
|
+
+ terminal_flags_.io_size() + link_flags_.io_size()
|
222
|
+
+ bases_.io_size() + extras_.io_size() + tail_.io_size()
|
223
|
+
+ ((next_trie_.get() != NULL) ?
|
224
|
+
(next_trie_->io_size() - Header().io_size()) : 0)
|
225
|
+
+ cache_.io_size() + (sizeof(UInt32) * 2);
|
226
|
+
}
|
227
|
+
|
228
|
+
void LoudsTrie::clear() {
|
229
|
+
LoudsTrie().swap(*this);
|
230
|
+
}
|
231
|
+
|
232
|
+
void LoudsTrie::swap(LoudsTrie &rhs) {
|
233
|
+
louds_.swap(rhs.louds_);
|
234
|
+
terminal_flags_.swap(rhs.terminal_flags_);
|
235
|
+
link_flags_.swap(rhs.link_flags_);
|
236
|
+
bases_.swap(rhs.bases_);
|
237
|
+
extras_.swap(rhs.extras_);
|
238
|
+
tail_.swap(rhs.tail_);
|
239
|
+
next_trie_.swap(rhs.next_trie_);
|
240
|
+
cache_.swap(rhs.cache_);
|
241
|
+
marisa::swap(cache_mask_, rhs.cache_mask_);
|
242
|
+
marisa::swap(num_l1_nodes_, rhs.num_l1_nodes_);
|
243
|
+
config_.swap(rhs.config_);
|
244
|
+
mapper_.swap(rhs.mapper_);
|
245
|
+
}
|
246
|
+
|
247
|
+
void LoudsTrie::build_(Keyset &keyset, const Config &config) {
|
248
|
+
Vector<Key> keys;
|
249
|
+
keys.resize(keyset.size());
|
250
|
+
for (std::size_t i = 0; i < keyset.size(); ++i) {
|
251
|
+
keys[i].set_str(keyset[i].ptr(), keyset[i].length());
|
252
|
+
keys[i].set_weight(keyset[i].weight());
|
253
|
+
}
|
254
|
+
|
255
|
+
Vector<UInt32> terminals;
|
256
|
+
build_trie(keys, &terminals, config, 1);
|
257
|
+
|
258
|
+
typedef std::pair<UInt32, UInt32> TerminalIdPair;
|
259
|
+
|
260
|
+
Vector<TerminalIdPair> pairs;
|
261
|
+
pairs.resize(terminals.size());
|
262
|
+
for (std::size_t i = 0; i < pairs.size(); ++i) {
|
263
|
+
pairs[i].first = terminals[i];
|
264
|
+
pairs[i].second = (UInt32)i;
|
265
|
+
}
|
266
|
+
terminals.clear();
|
267
|
+
std::sort(pairs.begin(), pairs.end());
|
268
|
+
|
269
|
+
std::size_t node_id = 0;
|
270
|
+
for (std::size_t i = 0; i < pairs.size(); ++i) {
|
271
|
+
while (node_id < pairs[i].first) {
|
272
|
+
terminal_flags_.push_back(false);
|
273
|
+
++node_id;
|
274
|
+
}
|
275
|
+
if (node_id == pairs[i].first) {
|
276
|
+
terminal_flags_.push_back(true);
|
277
|
+
++node_id;
|
278
|
+
}
|
279
|
+
}
|
280
|
+
while (node_id < bases_.size()) {
|
281
|
+
terminal_flags_.push_back(false);
|
282
|
+
++node_id;
|
283
|
+
}
|
284
|
+
terminal_flags_.push_back(false);
|
285
|
+
terminal_flags_.build(false, true);
|
286
|
+
|
287
|
+
for (std::size_t i = 0; i < keyset.size(); ++i) {
|
288
|
+
keyset[pairs[i].second].set_id(terminal_flags_.rank1(pairs[i].first));
|
289
|
+
}
|
290
|
+
}
|
291
|
+
|
292
|
+
template <typename T>
|
293
|
+
void LoudsTrie::build_trie(Vector<T> &keys,
|
294
|
+
Vector<UInt32> *terminals, const Config &config, std::size_t trie_id) {
|
295
|
+
build_current_trie(keys, terminals, config, trie_id);
|
296
|
+
|
297
|
+
Vector<UInt32> next_terminals;
|
298
|
+
if (!keys.empty()) {
|
299
|
+
build_next_trie(keys, &next_terminals, config, trie_id);
|
300
|
+
}
|
301
|
+
|
302
|
+
if (next_trie_.get() != NULL) {
|
303
|
+
config_.parse((next_trie_->num_tries() + 1) |
|
304
|
+
next_trie_->tail_mode() | next_trie_->node_order());
|
305
|
+
} else {
|
306
|
+
config_.parse(1 | tail_.mode() | config.node_order() |
|
307
|
+
config.cache_level());
|
308
|
+
}
|
309
|
+
|
310
|
+
link_flags_.build(false, false);
|
311
|
+
std::size_t node_id = 0;
|
312
|
+
for (std::size_t i = 0; i < next_terminals.size(); ++i) {
|
313
|
+
while (!link_flags_[node_id]) {
|
314
|
+
++node_id;
|
315
|
+
}
|
316
|
+
bases_[node_id] = (UInt8)(next_terminals[i] % 256);
|
317
|
+
next_terminals[i] /= 256;
|
318
|
+
++node_id;
|
319
|
+
}
|
320
|
+
extras_.build(next_terminals);
|
321
|
+
fill_cache();
|
322
|
+
}
|
323
|
+
|
324
|
+
template <typename T>
|
325
|
+
void LoudsTrie::build_current_trie(Vector<T> &keys,
|
326
|
+
Vector<UInt32> *terminals, const Config &config,
|
327
|
+
std::size_t trie_id) try {
|
328
|
+
for (std::size_t i = 0; i < keys.size(); ++i) {
|
329
|
+
keys[i].set_id(i);
|
330
|
+
}
|
331
|
+
const std::size_t num_keys = Algorithm().sort(keys.begin(), keys.end());
|
332
|
+
reserve_cache(config, trie_id, num_keys);
|
333
|
+
|
334
|
+
louds_.push_back(true);
|
335
|
+
louds_.push_back(false);
|
336
|
+
bases_.push_back('\0');
|
337
|
+
link_flags_.push_back(false);
|
338
|
+
|
339
|
+
Vector<T> next_keys;
|
340
|
+
std::queue<Range> queue;
|
341
|
+
Vector<WeightedRange> w_ranges;
|
342
|
+
|
343
|
+
queue.push(make_range(0, keys.size(), 0));
|
344
|
+
while (!queue.empty()) {
|
345
|
+
const std::size_t node_id = link_flags_.size() - queue.size();
|
346
|
+
|
347
|
+
Range range = queue.front();
|
348
|
+
queue.pop();
|
349
|
+
|
350
|
+
while ((range.begin() < range.end()) &&
|
351
|
+
(keys[range.begin()].length() == range.key_pos())) {
|
352
|
+
keys[range.begin()].set_terminal(node_id);
|
353
|
+
range.set_begin(range.begin() + 1);
|
354
|
+
}
|
355
|
+
|
356
|
+
if (range.begin() == range.end()) {
|
357
|
+
louds_.push_back(false);
|
358
|
+
continue;
|
359
|
+
}
|
360
|
+
|
361
|
+
w_ranges.clear();
|
362
|
+
double weight = keys[range.begin()].weight();
|
363
|
+
for (std::size_t i = range.begin() + 1; i < range.end(); ++i) {
|
364
|
+
if (keys[i - 1][range.key_pos()] != keys[i][range.key_pos()]) {
|
365
|
+
w_ranges.push_back(make_weighted_range(
|
366
|
+
range.begin(), i, range.key_pos(), (float)weight));
|
367
|
+
range.set_begin(i);
|
368
|
+
weight = 0.0;
|
369
|
+
}
|
370
|
+
weight += keys[i].weight();
|
371
|
+
}
|
372
|
+
w_ranges.push_back(make_weighted_range(
|
373
|
+
range.begin(), range.end(), range.key_pos(), (float)weight));
|
374
|
+
if (config.node_order() == MARISA_WEIGHT_ORDER) {
|
375
|
+
std::stable_sort(w_ranges.begin(), w_ranges.end(),
|
376
|
+
std::greater<WeightedRange>());
|
377
|
+
}
|
378
|
+
|
379
|
+
if (node_id == 0) {
|
380
|
+
num_l1_nodes_ = w_ranges.size();
|
381
|
+
}
|
382
|
+
|
383
|
+
for (std::size_t i = 0; i < w_ranges.size(); ++i) {
|
384
|
+
WeightedRange &w_range = w_ranges[i];
|
385
|
+
std::size_t key_pos = w_range.key_pos() + 1;
|
386
|
+
while (key_pos < keys[w_range.begin()].length()) {
|
387
|
+
std::size_t j;
|
388
|
+
for (j = w_range.begin() + 1; j < w_range.end(); ++j) {
|
389
|
+
if (keys[j - 1][key_pos] != keys[j][key_pos]) {
|
390
|
+
break;
|
391
|
+
}
|
392
|
+
}
|
393
|
+
if (j < w_range.end()) {
|
394
|
+
break;
|
395
|
+
}
|
396
|
+
++key_pos;
|
397
|
+
}
|
398
|
+
cache<T>(node_id, bases_.size(), w_range.weight(),
|
399
|
+
keys[w_range.begin()][w_range.key_pos()]);
|
400
|
+
|
401
|
+
if (key_pos == w_range.key_pos() + 1) {
|
402
|
+
bases_.push_back(keys[w_range.begin()][w_range.key_pos()]);
|
403
|
+
link_flags_.push_back(false);
|
404
|
+
} else {
|
405
|
+
bases_.push_back('\0');
|
406
|
+
link_flags_.push_back(true);
|
407
|
+
T next_key;
|
408
|
+
next_key.set_str(keys[w_range.begin()].ptr(),
|
409
|
+
keys[w_range.begin()].length());
|
410
|
+
next_key.substr(w_range.key_pos(), key_pos - w_range.key_pos());
|
411
|
+
next_key.set_weight(w_range.weight());
|
412
|
+
next_keys.push_back(next_key);
|
413
|
+
}
|
414
|
+
w_range.set_key_pos(key_pos);
|
415
|
+
queue.push(w_range.range());
|
416
|
+
louds_.push_back(true);
|
417
|
+
}
|
418
|
+
louds_.push_back(false);
|
419
|
+
}
|
420
|
+
|
421
|
+
louds_.push_back(false);
|
422
|
+
louds_.build(trie_id == 1, true);
|
423
|
+
bases_.shrink();
|
424
|
+
|
425
|
+
build_terminals(keys, terminals);
|
426
|
+
keys.swap(next_keys);
|
427
|
+
} catch (const std::bad_alloc &) {
|
428
|
+
MARISA_THROW(MARISA_MEMORY_ERROR, "std::bad_alloc");
|
429
|
+
}
|
430
|
+
|
431
|
+
template <>
|
432
|
+
void LoudsTrie::build_next_trie(Vector<Key> &keys,
|
433
|
+
Vector<UInt32> *terminals, const Config &config, std::size_t trie_id) {
|
434
|
+
if (trie_id == config.num_tries()) {
|
435
|
+
Vector<Entry> entries;
|
436
|
+
entries.resize(keys.size());
|
437
|
+
for (std::size_t i = 0; i < keys.size(); ++i) {
|
438
|
+
entries[i].set_str(keys[i].ptr(), keys[i].length());
|
439
|
+
}
|
440
|
+
tail_.build(entries, terminals, config.tail_mode());
|
441
|
+
return;
|
442
|
+
}
|
443
|
+
Vector<ReverseKey> reverse_keys;
|
444
|
+
reverse_keys.resize(keys.size());
|
445
|
+
for (std::size_t i = 0; i < keys.size(); ++i) {
|
446
|
+
reverse_keys[i].set_str(keys[i].ptr(), keys[i].length());
|
447
|
+
reverse_keys[i].set_weight(keys[i].weight());
|
448
|
+
}
|
449
|
+
keys.clear();
|
450
|
+
next_trie_.reset(new (std::nothrow) LoudsTrie);
|
451
|
+
MARISA_THROW_IF(next_trie_.get() == NULL, MARISA_MEMORY_ERROR);
|
452
|
+
next_trie_->build_trie(reverse_keys, terminals, config, trie_id + 1);
|
453
|
+
}
|
454
|
+
|
455
|
+
template <>
|
456
|
+
void LoudsTrie::build_next_trie(Vector<ReverseKey> &keys,
|
457
|
+
Vector<UInt32> *terminals, const Config &config, std::size_t trie_id) {
|
458
|
+
if (trie_id == config.num_tries()) {
|
459
|
+
Vector<Entry> entries;
|
460
|
+
entries.resize(keys.size());
|
461
|
+
for (std::size_t i = 0; i < keys.size(); ++i) {
|
462
|
+
entries[i].set_str(keys[i].ptr(), keys[i].length());
|
463
|
+
}
|
464
|
+
tail_.build(entries, terminals, config.tail_mode());
|
465
|
+
return;
|
466
|
+
}
|
467
|
+
next_trie_.reset(new (std::nothrow) LoudsTrie);
|
468
|
+
MARISA_THROW_IF(next_trie_.get() == NULL, MARISA_MEMORY_ERROR);
|
469
|
+
next_trie_->build_trie(keys, terminals, config, trie_id + 1);
|
470
|
+
}
|
471
|
+
|
472
|
+
template <typename T>
|
473
|
+
void LoudsTrie::build_terminals(const Vector<T> &keys,
|
474
|
+
Vector<UInt32> *terminals) const {
|
475
|
+
Vector<UInt32> temp;
|
476
|
+
temp.resize(keys.size());
|
477
|
+
for (std::size_t i = 0; i < keys.size(); ++i) {
|
478
|
+
temp[keys[i].id()] = (UInt32)keys[i].terminal();
|
479
|
+
}
|
480
|
+
terminals->swap(temp);
|
481
|
+
}
|
482
|
+
|
483
|
+
template <>
|
484
|
+
void LoudsTrie::cache<Key>(std::size_t parent, std::size_t child,
|
485
|
+
float weight, char label) {
|
486
|
+
MARISA_DEBUG_IF(parent >= child, MARISA_RANGE_ERROR);
|
487
|
+
|
488
|
+
const std::size_t cache_id = get_cache_id(parent, label);
|
489
|
+
if (weight > cache_[cache_id].weight()) {
|
490
|
+
cache_[cache_id].set_parent(parent);
|
491
|
+
cache_[cache_id].set_child(child);
|
492
|
+
cache_[cache_id].set_weight(weight);
|
493
|
+
}
|
494
|
+
}
|
495
|
+
|
496
|
+
void LoudsTrie::reserve_cache(const Config &config, std::size_t trie_id,
|
497
|
+
std::size_t num_keys) {
|
498
|
+
std::size_t cache_size = (trie_id == 1) ? 256 : 1;
|
499
|
+
while (cache_size < (num_keys / config.cache_level())) {
|
500
|
+
cache_size *= 2;
|
501
|
+
}
|
502
|
+
cache_.resize(cache_size);
|
503
|
+
cache_mask_ = cache_size - 1;
|
504
|
+
}
|
505
|
+
|
506
|
+
template <>
|
507
|
+
void LoudsTrie::cache<ReverseKey>(std::size_t parent, std::size_t child,
|
508
|
+
float weight, char) {
|
509
|
+
MARISA_DEBUG_IF(parent >= child, MARISA_RANGE_ERROR);
|
510
|
+
|
511
|
+
const std::size_t cache_id = get_cache_id(child);
|
512
|
+
if (weight > cache_[cache_id].weight()) {
|
513
|
+
cache_[cache_id].set_parent(parent);
|
514
|
+
cache_[cache_id].set_child(child);
|
515
|
+
cache_[cache_id].set_weight(weight);
|
516
|
+
}
|
517
|
+
}
|
518
|
+
|
519
|
+
void LoudsTrie::fill_cache() {
|
520
|
+
for (std::size_t i = 0; i < cache_.size(); ++i) {
|
521
|
+
const std::size_t node_id = cache_[i].child();
|
522
|
+
if (node_id != 0) {
|
523
|
+
cache_[i].set_base(bases_[node_id]);
|
524
|
+
cache_[i].set_extra(!link_flags_[node_id] ?
|
525
|
+
MARISA_INVALID_EXTRA : extras_[link_flags_.rank1(node_id)]);
|
526
|
+
} else {
|
527
|
+
cache_[i].set_parent(MARISA_UINT32_MAX);
|
528
|
+
cache_[i].set_child(MARISA_UINT32_MAX);
|
529
|
+
}
|
530
|
+
}
|
531
|
+
}
|
532
|
+
|
533
|
+
void LoudsTrie::map_(Mapper &mapper) {
|
534
|
+
louds_.map(mapper);
|
535
|
+
terminal_flags_.map(mapper);
|
536
|
+
link_flags_.map(mapper);
|
537
|
+
bases_.map(mapper);
|
538
|
+
extras_.map(mapper);
|
539
|
+
tail_.map(mapper);
|
540
|
+
if ((link_flags_.num_1s() != 0) && tail_.empty()) {
|
541
|
+
next_trie_.reset(new (std::nothrow) LoudsTrie);
|
542
|
+
MARISA_THROW_IF(next_trie_.get() == NULL, MARISA_MEMORY_ERROR);
|
543
|
+
next_trie_->map_(mapper);
|
544
|
+
}
|
545
|
+
cache_.map(mapper);
|
546
|
+
cache_mask_ = cache_.size() - 1;
|
547
|
+
{
|
548
|
+
UInt32 temp_num_l1_nodes;
|
549
|
+
mapper.map(&temp_num_l1_nodes);
|
550
|
+
num_l1_nodes_ = temp_num_l1_nodes;
|
551
|
+
}
|
552
|
+
{
|
553
|
+
UInt32 temp_config_flags;
|
554
|
+
mapper.map(&temp_config_flags);
|
555
|
+
config_.parse((int)temp_config_flags);
|
556
|
+
}
|
557
|
+
}
|
558
|
+
|
559
|
+
void LoudsTrie::read_(Reader &reader) {
|
560
|
+
louds_.read(reader);
|
561
|
+
terminal_flags_.read(reader);
|
562
|
+
link_flags_.read(reader);
|
563
|
+
bases_.read(reader);
|
564
|
+
extras_.read(reader);
|
565
|
+
tail_.read(reader);
|
566
|
+
if ((link_flags_.num_1s() != 0) && tail_.empty()) {
|
567
|
+
next_trie_.reset(new (std::nothrow) LoudsTrie);
|
568
|
+
MARISA_THROW_IF(next_trie_.get() == NULL, MARISA_MEMORY_ERROR);
|
569
|
+
next_trie_->read_(reader);
|
570
|
+
}
|
571
|
+
cache_.read(reader);
|
572
|
+
cache_mask_ = cache_.size() - 1;
|
573
|
+
{
|
574
|
+
UInt32 temp_num_l1_nodes;
|
575
|
+
reader.read(&temp_num_l1_nodes);
|
576
|
+
num_l1_nodes_ = temp_num_l1_nodes;
|
577
|
+
}
|
578
|
+
{
|
579
|
+
UInt32 temp_config_flags;
|
580
|
+
reader.read(&temp_config_flags);
|
581
|
+
config_.parse((int)temp_config_flags);
|
582
|
+
}
|
583
|
+
}
|
584
|
+
|
585
|
+
void LoudsTrie::write_(Writer &writer) const {
|
586
|
+
louds_.write(writer);
|
587
|
+
terminal_flags_.write(writer);
|
588
|
+
link_flags_.write(writer);
|
589
|
+
bases_.write(writer);
|
590
|
+
extras_.write(writer);
|
591
|
+
tail_.write(writer);
|
592
|
+
if (next_trie_.get() != NULL) {
|
593
|
+
next_trie_->write_(writer);
|
594
|
+
}
|
595
|
+
cache_.write(writer);
|
596
|
+
writer.write((UInt32)num_l1_nodes_);
|
597
|
+
writer.write((UInt32)config_.flags());
|
598
|
+
}
|
599
|
+
|
600
|
+
bool LoudsTrie::find_child(Agent &agent) const {
|
601
|
+
MARISA_DEBUG_IF(agent.state().query_pos() >= agent.query().length(),
|
602
|
+
MARISA_BOUND_ERROR);
|
603
|
+
|
604
|
+
State &state = agent.state();
|
605
|
+
const std::size_t cache_id = get_cache_id(state.node_id(),
|
606
|
+
agent.query()[state.query_pos()]);
|
607
|
+
if (state.node_id() == cache_[cache_id].parent()) {
|
608
|
+
if (cache_[cache_id].extra() != MARISA_INVALID_EXTRA) {
|
609
|
+
if (!match(agent, cache_[cache_id].link())) {
|
610
|
+
return false;
|
611
|
+
}
|
612
|
+
} else {
|
613
|
+
state.set_query_pos(state.query_pos() + 1);
|
614
|
+
}
|
615
|
+
state.set_node_id(cache_[cache_id].child());
|
616
|
+
return true;
|
617
|
+
}
|
618
|
+
|
619
|
+
std::size_t louds_pos = louds_.select0(state.node_id()) + 1;
|
620
|
+
if (!louds_[louds_pos]) {
|
621
|
+
return false;
|
622
|
+
}
|
623
|
+
state.set_node_id(louds_pos - state.node_id() - 1);
|
624
|
+
std::size_t link_id = MARISA_INVALID_LINK_ID;
|
625
|
+
do {
|
626
|
+
if (link_flags_[state.node_id()]) {
|
627
|
+
link_id = update_link_id(link_id, state.node_id());
|
628
|
+
const std::size_t prev_query_pos = state.query_pos();
|
629
|
+
if (match(agent, get_link(state.node_id(), link_id))) {
|
630
|
+
return true;
|
631
|
+
} else if (state.query_pos() != prev_query_pos) {
|
632
|
+
return false;
|
633
|
+
}
|
634
|
+
} else if (bases_[state.node_id()] ==
|
635
|
+
(UInt8)agent.query()[state.query_pos()]) {
|
636
|
+
state.set_query_pos(state.query_pos() + 1);
|
637
|
+
return true;
|
638
|
+
}
|
639
|
+
state.set_node_id(state.node_id() + 1);
|
640
|
+
++louds_pos;
|
641
|
+
} while (louds_[louds_pos]);
|
642
|
+
return false;
|
643
|
+
}
|
644
|
+
|
645
|
+
bool LoudsTrie::predictive_find_child(Agent &agent) const {
|
646
|
+
MARISA_DEBUG_IF(agent.state().query_pos() >= agent.query().length(),
|
647
|
+
MARISA_BOUND_ERROR);
|
648
|
+
|
649
|
+
State &state = agent.state();
|
650
|
+
const std::size_t cache_id = get_cache_id(state.node_id(),
|
651
|
+
agent.query()[state.query_pos()]);
|
652
|
+
if (state.node_id() == cache_[cache_id].parent()) {
|
653
|
+
if (cache_[cache_id].extra() != MARISA_INVALID_EXTRA) {
|
654
|
+
if (!prefix_match(agent, cache_[cache_id].link())) {
|
655
|
+
return false;
|
656
|
+
}
|
657
|
+
} else {
|
658
|
+
state.key_buf().push_back(cache_[cache_id].label());
|
659
|
+
state.set_query_pos(state.query_pos() + 1);
|
660
|
+
}
|
661
|
+
state.set_node_id(cache_[cache_id].child());
|
662
|
+
return true;
|
663
|
+
}
|
664
|
+
|
665
|
+
std::size_t louds_pos = louds_.select0(state.node_id()) + 1;
|
666
|
+
if (!louds_[louds_pos]) {
|
667
|
+
return false;
|
668
|
+
}
|
669
|
+
state.set_node_id(louds_pos - state.node_id() - 1);
|
670
|
+
std::size_t link_id = MARISA_INVALID_LINK_ID;
|
671
|
+
do {
|
672
|
+
if (link_flags_[state.node_id()]) {
|
673
|
+
link_id = update_link_id(link_id, state.node_id());
|
674
|
+
const std::size_t prev_query_pos = state.query_pos();
|
675
|
+
if (prefix_match(agent, get_link(state.node_id(), link_id))) {
|
676
|
+
return true;
|
677
|
+
} else if (state.query_pos() != prev_query_pos) {
|
678
|
+
return false;
|
679
|
+
}
|
680
|
+
} else if (bases_[state.node_id()] ==
|
681
|
+
(UInt8)agent.query()[state.query_pos()]) {
|
682
|
+
state.key_buf().push_back((char)bases_[state.node_id()]);
|
683
|
+
state.set_query_pos(state.query_pos() + 1);
|
684
|
+
return true;
|
685
|
+
}
|
686
|
+
state.set_node_id(state.node_id() + 1);
|
687
|
+
++louds_pos;
|
688
|
+
} while (louds_[louds_pos]);
|
689
|
+
return false;
|
690
|
+
}
|
691
|
+
|
692
|
+
void LoudsTrie::restore(Agent &agent, std::size_t link) const {
|
693
|
+
if (next_trie_.get() != NULL) {
|
694
|
+
next_trie_->restore_(agent, link);
|
695
|
+
} else {
|
696
|
+
tail_.restore(agent, link);
|
697
|
+
}
|
698
|
+
}
|
699
|
+
|
700
|
+
bool LoudsTrie::match(Agent &agent, std::size_t link) const {
|
701
|
+
if (next_trie_.get() != NULL) {
|
702
|
+
return next_trie_->match_(agent, link);
|
703
|
+
} else {
|
704
|
+
return tail_.match(agent, link);
|
705
|
+
}
|
706
|
+
}
|
707
|
+
|
708
|
+
bool LoudsTrie::prefix_match(Agent &agent, std::size_t link) const {
|
709
|
+
if (next_trie_.get() != NULL) {
|
710
|
+
return next_trie_->prefix_match_(agent, link);
|
711
|
+
} else {
|
712
|
+
return tail_.prefix_match(agent, link);
|
713
|
+
}
|
714
|
+
}
|
715
|
+
|
716
|
+
void LoudsTrie::restore_(Agent &agent, std::size_t node_id) const {
|
717
|
+
MARISA_DEBUG_IF(node_id == 0, MARISA_RANGE_ERROR);
|
718
|
+
|
719
|
+
State &state = agent.state();
|
720
|
+
for ( ; ; ) {
|
721
|
+
const std::size_t cache_id = get_cache_id(node_id);
|
722
|
+
if (node_id == cache_[cache_id].child()) {
|
723
|
+
if (cache_[cache_id].extra() != MARISA_INVALID_EXTRA) {
|
724
|
+
restore(agent, cache_[cache_id].link());
|
725
|
+
} else {
|
726
|
+
state.key_buf().push_back(cache_[cache_id].label());
|
727
|
+
}
|
728
|
+
|
729
|
+
node_id = cache_[cache_id].parent();
|
730
|
+
if (node_id == 0) {
|
731
|
+
return;
|
732
|
+
}
|
733
|
+
continue;
|
734
|
+
}
|
735
|
+
|
736
|
+
if (link_flags_[node_id]) {
|
737
|
+
restore(agent, get_link(node_id));
|
738
|
+
} else {
|
739
|
+
state.key_buf().push_back((char)bases_[node_id]);
|
740
|
+
}
|
741
|
+
|
742
|
+
if (node_id <= num_l1_nodes_) {
|
743
|
+
return;
|
744
|
+
}
|
745
|
+
node_id = louds_.select1(node_id) - node_id - 1;
|
746
|
+
}
|
747
|
+
}
|
748
|
+
|
749
|
+
bool LoudsTrie::match_(Agent &agent, std::size_t node_id) const {
|
750
|
+
MARISA_DEBUG_IF(agent.state().query_pos() >= agent.query().length(),
|
751
|
+
MARISA_BOUND_ERROR);
|
752
|
+
MARISA_DEBUG_IF(node_id == 0, MARISA_RANGE_ERROR);
|
753
|
+
|
754
|
+
State &state = agent.state();
|
755
|
+
for ( ; ; ) {
|
756
|
+
const std::size_t cache_id = get_cache_id(node_id);
|
757
|
+
if (node_id == cache_[cache_id].child()) {
|
758
|
+
if (cache_[cache_id].extra() != MARISA_INVALID_EXTRA) {
|
759
|
+
if (!match(agent, cache_[cache_id].link())) {
|
760
|
+
return false;
|
761
|
+
}
|
762
|
+
} else if (cache_[cache_id].label() ==
|
763
|
+
agent.query()[state.query_pos()]) {
|
764
|
+
state.set_query_pos(state.query_pos() + 1);
|
765
|
+
} else {
|
766
|
+
return false;
|
767
|
+
}
|
768
|
+
|
769
|
+
node_id = cache_[cache_id].parent();
|
770
|
+
if (node_id == 0) {
|
771
|
+
return true;
|
772
|
+
} else if (state.query_pos() >= agent.query().length()) {
|
773
|
+
return false;
|
774
|
+
}
|
775
|
+
continue;
|
776
|
+
}
|
777
|
+
|
778
|
+
if (link_flags_[node_id]) {
|
779
|
+
if (next_trie_.get() != NULL) {
|
780
|
+
if (!match(agent, get_link(node_id))) {
|
781
|
+
return false;
|
782
|
+
}
|
783
|
+
} else if (!tail_.match(agent, get_link(node_id))) {
|
784
|
+
return false;
|
785
|
+
}
|
786
|
+
} else if (bases_[node_id] == (UInt8)agent.query()[state.query_pos()]) {
|
787
|
+
state.set_query_pos(state.query_pos() + 1);
|
788
|
+
} else {
|
789
|
+
return false;
|
790
|
+
}
|
791
|
+
|
792
|
+
if (node_id <= num_l1_nodes_) {
|
793
|
+
return true;
|
794
|
+
} else if (state.query_pos() >= agent.query().length()) {
|
795
|
+
return false;
|
796
|
+
}
|
797
|
+
node_id = louds_.select1(node_id) - node_id - 1;
|
798
|
+
}
|
799
|
+
}
|
800
|
+
|
801
|
+
bool LoudsTrie::prefix_match_(Agent &agent, std::size_t node_id) const {
|
802
|
+
MARISA_DEBUG_IF(agent.state().query_pos() >= agent.query().length(),
|
803
|
+
MARISA_BOUND_ERROR);
|
804
|
+
MARISA_DEBUG_IF(node_id == 0, MARISA_RANGE_ERROR);
|
805
|
+
|
806
|
+
State &state = agent.state();
|
807
|
+
for ( ; ; ) {
|
808
|
+
const std::size_t cache_id = get_cache_id(node_id);
|
809
|
+
if (node_id == cache_[cache_id].child()) {
|
810
|
+
if (cache_[cache_id].extra() != MARISA_INVALID_EXTRA) {
|
811
|
+
if (!prefix_match(agent, cache_[cache_id].link())) {
|
812
|
+
return false;
|
813
|
+
}
|
814
|
+
} else if (cache_[cache_id].label() ==
|
815
|
+
agent.query()[state.query_pos()]) {
|
816
|
+
state.key_buf().push_back(cache_[cache_id].label());
|
817
|
+
state.set_query_pos(state.query_pos() + 1);
|
818
|
+
} else {
|
819
|
+
return false;
|
820
|
+
}
|
821
|
+
|
822
|
+
node_id = cache_[cache_id].parent();
|
823
|
+
if (node_id == 0) {
|
824
|
+
return true;
|
825
|
+
}
|
826
|
+
} else {
|
827
|
+
if (link_flags_[node_id]) {
|
828
|
+
if (!prefix_match(agent, get_link(node_id))) {
|
829
|
+
return false;
|
830
|
+
}
|
831
|
+
} else if (bases_[node_id] == (UInt8)agent.query()[state.query_pos()]) {
|
832
|
+
state.key_buf().push_back((char)bases_[node_id]);
|
833
|
+
state.set_query_pos(state.query_pos() + 1);
|
834
|
+
} else {
|
835
|
+
return false;
|
836
|
+
}
|
837
|
+
|
838
|
+
if (node_id <= num_l1_nodes_) {
|
839
|
+
return true;
|
840
|
+
}
|
841
|
+
node_id = louds_.select1(node_id) - node_id - 1;
|
842
|
+
}
|
843
|
+
|
844
|
+
if (state.query_pos() >= agent.query().length()) {
|
845
|
+
restore_(agent, node_id);
|
846
|
+
return true;
|
847
|
+
}
|
848
|
+
}
|
849
|
+
}
|
850
|
+
|
851
|
+
std::size_t LoudsTrie::get_cache_id(std::size_t node_id, char label) const {
|
852
|
+
return (node_id ^ (node_id << 5) ^ (UInt8)label) & cache_mask_;
|
853
|
+
}
|
854
|
+
|
855
|
+
std::size_t LoudsTrie::get_cache_id(std::size_t node_id) const {
|
856
|
+
return node_id & cache_mask_;
|
857
|
+
}
|
858
|
+
|
859
|
+
std::size_t LoudsTrie::get_link(std::size_t node_id) const {
|
860
|
+
return bases_[node_id] | (extras_[link_flags_.rank1(node_id)] * 256);
|
861
|
+
}
|
862
|
+
|
863
|
+
std::size_t LoudsTrie::get_link(std::size_t node_id,
|
864
|
+
std::size_t link_id) const {
|
865
|
+
return bases_[node_id] | (extras_[link_id] * 256);
|
866
|
+
}
|
867
|
+
|
868
|
+
std::size_t LoudsTrie::update_link_id(std::size_t link_id,
|
869
|
+
std::size_t node_id) const {
|
870
|
+
return (link_id == MARISA_INVALID_LINK_ID) ?
|
871
|
+
link_flags_.rank1(node_id) : (link_id + 1);
|
872
|
+
}
|
873
|
+
|
874
|
+
} // namespace trie
|
875
|
+
} // namespace grimoire
|
876
|
+
} // namespace marisa
|