lindera_dictionary/
dictionary.rs1pub mod character_definition;
2pub mod connection_cost_matrix;
3pub mod context_id_map;
4pub mod metadata;
5pub mod prefix_dictionary;
6pub mod schema;
7pub mod unknown_dictionary;
8
9use std::fs;
10use std::path::Path;
11use std::str;
12use std::sync::Arc;
13
14use byteorder::{ByteOrder, LittleEndian};
15use once_cell::sync::Lazy;
16use rkyv::{Archive, Deserialize as RkyvDeserialize, Serialize as RkyvSerialize};
17use serde::{Deserialize, Serialize};
18
19use crate::LinderaResult;
20use crate::dictionary::character_definition::CharacterDefinition;
21use crate::dictionary::connection_cost_matrix::ConnectionCostMatrix;
22use crate::dictionary::context_id_map::ContextIdMap;
23use crate::dictionary::metadata::Metadata;
24use crate::dictionary::prefix_dictionary::PrefixDictionary;
25use crate::dictionary::unknown_dictionary::UnknownDictionary;
26use crate::error::LinderaErrorKind;
27use crate::loader::character_definition::CharacterDefinitionLoader;
28use crate::loader::connection_cost_matrix::ConnectionCostMatrixLoader;
29use crate::loader::metadata::MetadataLoader;
30use crate::loader::prefix_dictionary::PrefixDictionaryLoader;
31use crate::loader::unknown_dictionary::UnknownDictionaryLoader;
32use crate::util::Data;
33use crate::viterbi::WordEntry;
34
35pub static UNK: Lazy<Vec<&str>> = Lazy::new(|| vec!["UNK"]);
36
37#[derive(Clone)]
43pub struct Dictionary {
44 pub prefix_dictionary: Arc<PrefixDictionary>,
45 pub connection_cost_matrix: Arc<ConnectionCostMatrix>,
46 pub character_definition: Arc<CharacterDefinition>,
47 pub unknown_dictionary: Arc<UnknownDictionary>,
48 pub metadata: Arc<Metadata>,
49}
50
51impl Dictionary {
52 pub fn unknown_word_details(&self, word_id: usize) -> Vec<&str> {
54 match self.unknown_dictionary.word_details(word_id as u32) {
55 Some(details) => details,
56 None => UNK.to_vec(),
57 }
58 }
59
60 pub fn word_details(&self, word_id: usize) -> Vec<&str> {
61 if 4 * word_id >= self.prefix_dictionary.words_idx_data.len() {
62 return vec![];
63 }
64
65 let idx: usize = match LittleEndian::read_u32(
66 &self.prefix_dictionary.words_idx_data[4 * word_id..][..4],
67 )
68 .try_into()
69 {
70 Ok(value) => value,
71 Err(_) => return UNK.to_vec(), };
73 let data = &self.prefix_dictionary.words_data[idx..];
74 let joined_details_len: usize = match LittleEndian::read_u32(data).try_into() {
75 Ok(value) => value,
76 Err(_) => return UNK.to_vec(), };
78 let joined_details_bytes: &[u8] =
79 &self.prefix_dictionary.words_data[idx + 4..idx + 4 + joined_details_len];
80
81 let mut details = Vec::new();
82 for bytes in joined_details_bytes.split(|&b| b == 0) {
83 let detail = match str::from_utf8(bytes) {
84 Ok(s) => s,
85 Err(_) => return UNK.to_vec(), };
87 details.push(detail);
88 }
89 details
90 }
91
92 pub fn load_from_path(dict_path: &Path) -> LinderaResult<Self> {
99 Self::load_from_path_with_options(dict_path, cfg!(feature = "mmap"))
100 }
101
102 pub fn load_from_path_with_options(dict_path: &Path, use_mmap: bool) -> LinderaResult<Self> {
120 if !dict_path.exists() {
122 return Err(LinderaErrorKind::Io.with_error(anyhow::anyhow!(
123 "Dictionary path does not exist: {}",
124 dict_path.display()
125 )));
126 }
127
128 if !dict_path.is_dir() {
129 return Err(LinderaErrorKind::Io.with_error(anyhow::anyhow!(
130 "Dictionary path is not a directory: {}",
131 dict_path.display()
132 )));
133 }
134
135 let metadata = MetadataLoader::load(dict_path)?;
137 let character_definition = CharacterDefinitionLoader::load(dict_path)?;
138
139 let connection_cost_matrix = {
140 #[cfg(feature = "mmap")]
141 if use_mmap {
142 ConnectionCostMatrixLoader::load_mmap(dict_path)?
143 } else {
144 ConnectionCostMatrixLoader::load(dict_path)?
145 }
146 #[cfg(not(feature = "mmap"))]
147 ConnectionCostMatrixLoader::load(dict_path)?
148 };
149
150 let prefix_dictionary = {
151 #[cfg(feature = "mmap")]
152 if use_mmap {
153 PrefixDictionaryLoader::load_mmap(dict_path)?
154 } else {
155 PrefixDictionaryLoader::load(dict_path)?
156 }
157 #[cfg(not(feature = "mmap"))]
158 PrefixDictionaryLoader::load(dict_path)?
159 };
160
161 let unknown_dictionary = UnknownDictionaryLoader::load(dict_path)?;
162
163 Ok(Dictionary {
164 prefix_dictionary: Arc::new(prefix_dictionary),
165 connection_cost_matrix: Arc::new(connection_cost_matrix),
166 character_definition: Arc::new(character_definition),
167 unknown_dictionary: Arc::new(unknown_dictionary),
168 metadata: Arc::new(metadata),
169 })
170 }
171
172 pub fn save_to_path(&self, dict_path: &Path) -> LinderaResult<()> {
174 fs::create_dir_all(dict_path)
176 .map_err(|err| LinderaErrorKind::Io.with_error(anyhow::anyhow!(err)))?;
177
178 todo!("Dictionary saving will be implemented when needed")
181 }
182}
183
184#[derive(Clone, Serialize, Deserialize, Archive, RkyvSerialize, RkyvDeserialize)]
185
186pub struct UserDictionary {
187 pub dict: PrefixDictionary,
188}
189
190impl UserDictionary {
191 pub fn remap_context_ids(&mut self, map: &ContextIdMap) {
209 const LEFT_ID_OFFSET: usize = 6;
210 const RIGHT_ID_OFFSET: usize = 8;
211
212 let mut vals = self.dict.vals_data.to_vec();
213 for entry in vals.chunks_exact_mut(WordEntry::SERIALIZED_LEN) {
214 let left = LittleEndian::read_u16(&entry[LEFT_ID_OFFSET..][..2]);
215 let right = LittleEndian::read_u16(&entry[RIGHT_ID_OFFSET..][..2]);
216 LittleEndian::write_u16(&mut entry[LEFT_ID_OFFSET..][..2], map.map_left(left));
217 LittleEndian::write_u16(&mut entry[RIGHT_ID_OFFSET..][..2], map.map_right(right));
218 }
219 self.dict.vals_data = Data::Vec(vals);
220 }
221
222 pub fn load(user_dict_data: &[u8]) -> LinderaResult<UserDictionary> {
223 let mut aligned = rkyv::util::AlignedVec::<16>::new();
224 aligned.extend_from_slice(user_dict_data);
225 rkyv::from_bytes::<UserDictionary, rkyv::rancor::Error>(&aligned).map_err(|err| {
226 LinderaErrorKind::Deserialize.with_error(anyhow::anyhow!(err.to_string()))
227 })
228 }
229
230 pub fn word_details(&self, word_id: usize) -> Vec<&str> {
231 if 4 * word_id >= self.dict.words_idx_data.len() {
232 return UNK.to_vec(); }
234 let idx = LittleEndian::read_u32(&self.dict.words_idx_data[4 * word_id..][..4]);
235 let data = &self.dict.words_data[idx as usize..];
236
237 let joined_details_len: usize = match LittleEndian::read_u32(data).try_into() {
239 Ok(value) => value,
240 Err(_) => return UNK.to_vec(), };
242 let joined_details_bytes: &[u8] =
243 &self.dict.words_data[idx as usize + 4..idx as usize + 4 + joined_details_len];
244
245 let mut details = Vec::new();
246 for bytes in joined_details_bytes.split(|&b| b == 0) {
247 let detail = match str::from_utf8(bytes) {
248 Ok(s) => s,
249 Err(_) => return UNK.to_vec(), };
251 details.push(detail);
252 }
253 details
254 }
255}