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

use aho_corasick::MatchKind;
use tantivy::tokenizer::{Token, TokenStream, Tokenizer};

/// Tokenize the text by splitting on whitespaces and punctuation.
#[derive(Clone)]
pub struct DictTokenizer {
    ac: aho_corasick::AhoCorasick,
    words: Vec<String>,
    dict: Vec<usize>,
}

impl DictTokenizer {
    pub fn new() -> DictTokenizer {
        let mut synsets = vec![];
        let mut csv_reader = csv::ReaderBuilder::new()
            .has_headers(false)
            .flexible(true)
            .from_reader(include_bytes!("../../../resources/drugs.csv").as_slice());
        for record in csv_reader.records() {
            let mut synset = vec![];
            for word in record.expect("dictionary is broken").iter() {
                synset.push(word.to_string())
            }
            synsets.push(synset);
        }

        let mut base_offset = 0;
        let mut dict = vec![];
        let words: Vec<String> = synsets
            .into_iter()
            .flat_map(|synset| {
                dict.extend(std::iter::repeat(base_offset).take(synset.len()));
                base_offset += synset.len();
                synset
            })
            .collect();
        let ac = aho_corasick::AhoCorasickBuilder::new()
            .ascii_case_insensitive(true)
            .match_kind(MatchKind::LeftmostLongest)
            .build(words.iter())
            .expect("internal error");
        DictTokenizer { ac, words, dict }
    }
}

impl Default for DictTokenizer {
    fn default() -> Self {
        DictTokenizer::new()
    }
}

pub struct DictTokenStream<'a> {
    text: &'a str,
    words: &'a Vec<String>,
    dict: &'a Vec<usize>,
    ah_iter: aho_corasick::FindIter<'a, 'a>,
    token: Token,
}

impl<'a> DictTokenStream<'a> {
    pub fn new(text: &'a str, words: &'a Vec<String>, dict: &'a Vec<usize>, ac: &'a aho_corasick::AhoCorasick) -> DictTokenStream<'a> {
        DictTokenStream {
            text,
            words,
            dict,
            ah_iter: ac.find_iter(text),
            token: Token::default(),
        }
    }
}

impl Tokenizer for DictTokenizer {
    type TokenStream<'a> = DictTokenStream<'a>;

    fn token_stream<'a>(&'a mut self, text: &'a str) -> DictTokenStream<'a> {
        DictTokenStream::new(text, &self.words, &self.dict, &self.ac)
    }
}

impl<'a> TokenStream for DictTokenStream<'a> {
    fn advance(&mut self) -> bool {
        self.token.text.clear();
        self.token.position = self.token.position.wrapping_add(1);
        for pattern in self.ah_iter.by_ref() {
            let properly_beginning = pattern.start() == 0
                || self.text.as_bytes()[pattern.start() - 1].is_ascii_punctuation()
                || self.text.as_bytes()[pattern.start() - 1].is_ascii_whitespace();
            let properly_ending = pattern.end() == self.text.len()
                || self.text.as_bytes()[pattern.end()].is_ascii_punctuation()
                || self.text.as_bytes()[pattern.end()].is_ascii_whitespace();
            if properly_beginning && properly_ending {
                self.token.offset_from = pattern.start();
                self.token.offset_to = pattern.end();
                self.token.text.push_str(&self.words[self.dict[pattern.pattern()]]);
                return true;
            }
        }
        false
    }

    fn token(&self) -> &Token {
        &self.token
    }

    fn token_mut(&mut self) -> &mut Token {
        &mut self.token
    }
}

#[cfg(test)]
pub mod tests {
    use tantivy::tokenizer::{TextAnalyzer, Token, TokenizerManager};

    use super::DictTokenizer;

    pub fn assert_token(token: &Token, position: usize, text: &str, from: usize, to: usize) {
        assert_eq!(token.position, position, "expected position {} but {:?}", position, token);
        assert_eq!(token.text, text, "expected text {} but {:?}", text, token);
        assert_eq!(token.offset_from, from, "expected offset_from {} but {:?}", from, token);
        assert_eq!(token.offset_to, to, "expected offset_to {} but {:?}", to, token);
    }

    #[test]
    fn test_dict_tokenizer() {
        let tokenizer_manager = TokenizerManager::default();
        tokenizer_manager.register("tokenizer", TextAnalyzer::builder(DictTokenizer::new()).build());
        let mut tokenizer = tokenizer_manager.get("tokenizer").unwrap();
        let mut tokens: Vec<Token> = vec![];
        {
            let mut add_token = |token: &Token| {
                tokens.push(token.clone());
            };
            tokenizer
                .token_stream("FOXP2 gene (not FOXP21) can be correlated with autism spectrum disorder or just autismo")
                .process(&mut add_token);
        }

        assert_eq!(tokens.len(), 1);
        assert_token(&tokens[0], 0, "foxp2", 0, 5);

        let mut tokens: Vec<Token> = vec![];
        {
            let mut add_token = |token: &Token| {
                tokens.push(token.clone());
            };
            tokenizer.token_stream("FOXP2ген связан с аутизмом").process(&mut add_token);
        }

        assert_eq!(tokens.len(), 0);
    }
}