textprep 0.1.5

Text preprocessing primitives: normalization, tokenization, and fast keyword matching.
Documentation 
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
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269

use proptest::prelude::*;

const ZWSP: char = '\u{200B}'; // ZERO WIDTH SPACE
const ZWNJ: char = '\u{200C}'; // ZERO WIDTH NON-JOINER
const ZWJ: char = '\u{200D}'; // ZERO WIDTH JOINER
const WJ: char = '\u{2060}'; // WORD JOINER
const BOM: char = '\u{FEFF}'; // ZERO WIDTH NO-BREAK SPACE (BOM)

const LRE: char = '\u{202A}'; // LEFT-TO-RIGHT EMBEDDING
const RLE: char = '\u{202B}'; // RIGHT-TO-LEFT EMBEDDING
const PDF: char = '\u{202C}'; // POP DIRECTIONAL FORMATTING
const LRO: char = '\u{202D}'; // LEFT-TO-RIGHT OVERRIDE
const RLO: char = '\u{202E}'; // RIGHT-TO-LEFT OVERRIDE
const LRI: char = '\u{2066}'; // LEFT-TO-RIGHT ISOLATE
const RLI: char = '\u{2067}'; // RIGHT-TO-LEFT ISOLATE
const FSI: char = '\u{2068}'; // FIRST-STRONG ISOLATE
const PDI: char = '\u{2069}'; // POP DIRECTIONAL ISOLATE
const LRM: char = '\u{200E}'; // LEFT-TO-RIGHT MARK
const RLM: char = '\u{200F}'; // RIGHT-TO-LEFT MARK
const ALM: char = '\u{061C}'; // ARABIC LETTER MARK

fn any_reasonable_string() -> impl Strategy<Value = String> {
    // Keep it bounded to avoid slow quadratic behavior in tests.
    // Includes full Unicode scalar range (including control chars).
    proptest::collection::vec(any::<char>(), 0..200).prop_map(|cs| cs.into_iter().collect())
}

fn slice_by_char_range(text: &str, start: usize, end: usize) -> String {
    text.chars()
        .skip(start)
        .take(end.saturating_sub(start))
        .collect()
}

fn assert_flash_matches_sane(
    text: &str,
    matches: &[textprep::KeywordMatch],
) -> Result<(), TestCaseError> {
    let char_count = text.chars().count();

    let mut last_end = 0usize;
    for m in matches {
        prop_assert!(m.start <= m.end);
        prop_assert!(m.end <= char_count);
        prop_assert!(last_end <= m.start);

        let extracted = slice_by_char_range(text, m.start, m.end);
        // `FlashText` uses ASCII case-insensitive matching (configurable internally),
        // so the matched substring may differ in ASCII casing from the pattern.
        prop_assert_eq!(
            extracted.to_ascii_lowercase(),
            m.keyword.to_ascii_lowercase()
        );

        last_end = m.end;
    }

    Ok(())
}

proptest! {
    #[test]
    fn collapse_whitespace_is_trimmed_and_single_spaced(s in any_reasonable_string()) {
        let out = textprep::unicode::collapse_whitespace(&s);

        prop_assert!(!out.starts_with(' '));
        prop_assert!(!out.ends_with(' '));
        prop_assert!(!out.contains("  "));

        // Only ASCII spaces may remain as whitespace.
        prop_assert!(out.chars().all(|c| !c.is_whitespace() || c == ' '));
        prop_assert!(!out.contains('\t'));
        prop_assert!(!out.contains('\n'));
        prop_assert!(!out.contains('\r'));
    }

    #[test]
    fn remove_zero_width_is_idempotent_and_removes_targets(s in any_reasonable_string()) {
        let out1 = textprep::unicode::remove_zero_width(&s);
        let out2 = textprep::unicode::remove_zero_width(&out1);
        prop_assert_eq!(out1.as_str(), out2.as_str());

        // Avoid `\u{...}` escapes in `prop_assert!` (they contain `{}` and break formatting).
        prop_assert!(!out1.contains(ZWSP));
        prop_assert!(!out1.contains(ZWNJ));
        prop_assert!(!out1.contains(ZWJ));
        prop_assert!(!out1.contains(WJ));
        prop_assert!(!out1.contains(BOM));
    }

    #[test]
    fn zero_width_offsets_roundtrip(s in any_reasonable_string()) {
        let hits = textprep::unicode::zero_width_with_offsets(&s);
        prop_assert_eq!(textprep::unicode::contains_zero_width(&s), !hits.is_empty());
        for (i, c) in &hits {
            prop_assert_eq!(s.chars().nth(*i), Some(*c));
            prop_assert!(matches!(*c, ZWSP | ZWNJ | ZWJ | WJ | BOM));
        }

        let out = textprep::unicode::remove_zero_width(&s);
        prop_assert!(!textprep::unicode::contains_zero_width(&out));
        prop_assert!(textprep::unicode::zero_width_with_offsets(&out).is_empty());

        // remove_* is a pure deletion of these codepoints.
        prop_assert_eq!(out.chars().count() + hits.len(), s.chars().count());
    }

    #[test]
    fn bidi_offsets_roundtrip(s in any_reasonable_string()) {
        let hits = textprep::unicode::bidi_controls_with_offsets(&s);
        prop_assert_eq!(textprep::unicode::contains_bidi_controls(&s), !hits.is_empty());
        for (i, c) in &hits {
            prop_assert_eq!(s.chars().nth(*i), Some(*c));
            prop_assert!(matches!(
                *c,
                LRE | RLE | PDF | LRO | RLO | LRI | RLI | FSI | PDI | LRM | RLM | ALM
            ));
        }

        let out = textprep::unicode::remove_bidi_controls(&s);
        prop_assert!(!textprep::unicode::contains_bidi_controls(&out));
        prop_assert!(textprep::unicode::bidi_controls_with_offsets(&out).is_empty());

        prop_assert_eq!(out.chars().count() + hits.len(), s.chars().count());
    }

    #[test]
    fn normalize_newlines_removes_cr(s in any_reasonable_string()) {
        let out = textprep::unicode::normalize_newlines(&s);
        prop_assert!(!out.contains('\r'));
    }

    #[test]
    fn tokenize_with_offsets_produces_valid_spans(s in any_reasonable_string()) {
        let tokens = textprep::tokenize::tokenize_with_offsets(&s);
        let char_count = s.chars().count();

        for t in &tokens {
            prop_assert!(t.start <= t.end);
            prop_assert!(t.end <= char_count);
            prop_assert!(!t.text.is_empty());
            prop_assert!(t.text.chars().all(|c| !c.is_whitespace()));

            let extracted = slice_by_char_range(&s, t.start, t.end);
            prop_assert_eq!(extracted.as_str(), t.text.as_str());
        }

        for w in tokens.windows(2) {
            prop_assert!(w[0].start <= w[1].start);
            prop_assert!(w[0].end <= w[1].end);
            prop_assert!(w[0].end <= w[1].start);
        }
    }

    #[test]
    fn similarity_metrics_are_symmetric_and_bounded(
        a in any_reasonable_string(),
        b in any_reasonable_string(),
        n in 1usize..6usize,
    ) {
        let w1 = textprep::similarity::word_jaccard(&a, &b);
        let w2 = textprep::similarity::word_jaccard(&b, &a);
        prop_assert!((0.0..=1.0).contains(&w1));
        prop_assert!((w1 - w2).abs() < 1e-12);

        let c1 = textprep::similarity::char_ngram_jaccard(&a, &b, n);
        let c2 = textprep::similarity::char_ngram_jaccard(&b, &a, n);
        prop_assert!((0.0..=1.0).contains(&c1));
        prop_assert!((c1 - c2).abs() < 1e-12);
    }

    #[test]
    fn weighted_jaccard_is_bounded_when_weights_sum_to_one(
        a in any_reasonable_string(),
        b in any_reasonable_string(),
        n in 1usize..6usize,
        w in 0.0f64..=1.0f64,
    ) {
        let s = textprep::similarity::weighted_word_char_ngram_jaccard(&a, &b, n, w, 1.0 - w);
        prop_assert!((0.0..=1.0).contains(&s));
    }

    #[test]
    fn token_ngrams_match_windows(
        words in proptest::collection::vec(
            proptest::collection::vec(any::<char>(), 0..16).prop_map(|cs| cs.into_iter().collect::<String>()),
            0..20
        ),
        n in 1usize..6usize,
    ) {
        let refs: Vec<&str> = words.iter().map(|s| s.as_str()).collect();
        let out = textprep::ngram::token_ngrams(&refs, n);

        if refs.len() < n {
            prop_assert!(out.is_empty());
        } else {
            prop_assert_eq!(out.len(), refs.len() - n + 1);
            for (i, gram) in out.iter().enumerate() {
                prop_assert_eq!(gram.len(), n);
                for j in 0..n {
                    prop_assert_eq!(gram[j], refs[i + j]);
                }
            }
        }
    }

    #[test]
    fn flashtext_find_is_well_formed_and_matches_find_into(s in any_reasonable_string()) {
        let mut ft = textprep::FlashText::new();
        ft.add_keyword("François", "francois");
        ft.add_keyword("Müller", "muller");
        ft.add_keyword("北京", "beijing");
        ft.add_keyword("hello", "hello");

        let matches = ft.find(&s);
        assert_flash_matches_sane(&s, &matches)?;

        let mut out = Vec::new();
        ft.find_into(&s, &mut out);
        // Compare by reference so we can keep using `out` below.
        prop_assert_eq!(&out, &matches);
        assert_flash_matches_sane(&s, &out)?;
    }

    #[test]
    fn flashtext_finds_embedded_keyword(
        prefix in any_reasonable_string(),
        suffix in any_reasonable_string(),
    ) {
        // Keep the constructed string modest so tests stay fast.
        let prefix: String = prefix.chars().take(80).collect();
        let suffix: String = suffix.chars().take(80).collect();

        let text = format!("{prefix} HeLLo {suffix}");

        let mut ft = textprep::FlashText::new();
        ft.add_keyword("hello", "hello");

        let matches = ft.find(&text);
        assert_flash_matches_sane(&text, &matches)?;
        prop_assert!(matches.iter().any(|m| m.keyword == "hello"));
    }

    #[test]
    fn scrub_search_key_is_idempotent(s in any_reasonable_string()) {
        let cfg = textprep::ScrubConfig::search_key();
        let out1 = textprep::scrub_with(&s, &cfg);
        let out2 = textprep::scrub_with(&out1, &cfg);
        prop_assert_eq!(out1, out2);
    }

    #[test]
    fn scrub_search_key_is_trimmed_and_single_spaced(s in any_reasonable_string()) {
        let cfg = textprep::ScrubConfig::search_key();
        let out = textprep::scrub_with(&s, &cfg);
        prop_assert!(!out.starts_with(' '));
        prop_assert!(!out.ends_with(' '));
        prop_assert!(!out.contains("  "));
        prop_assert!(out.chars().all(|c| !c.is_whitespace() || c == ' '));
    }

    #[test]
    fn scrub_search_key_strict_is_idempotent(s in any_reasonable_string()) {
        let cfg = textprep::ScrubConfig::search_key_strict_invisibles();
        let out1 = textprep::scrub_with(&s, &cfg);
        let out2 = textprep::scrub_with(&out1, &cfg);
        prop_assert_eq!(out1, out2);
    }
}