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
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
use itertools::*;
use lazy_static::*;
use rand::distributions::*;
use rand::{CryptoRng, Rng};
use sha2::{Digest, Sha256};
use std::collections::HashSet;
use std::ops::DerefMut;
use std::sync::{Mutex, MutexGuard};
use unicode_segmentation::UnicodeSegmentation;
use unidecode::unidecode_char;
use Result::{Err, Ok};

const FILTERED_CHARS: [char; 31] = [
    '!', '@', '#', '$', '%', '^', '&', '*', '(', ')', '{', '}', '_', '<', '>', ':', ';', ',', '.',
    '"', '\'', '`', '|', '+', '=', '/', '~', '[', ']', '\\', '-',
];

///True if we should keep the character in the string.
fn should_keep_char(c: &char) -> bool {
    !FILTERED_CHARS.contains(c)
}
lazy_static! {
    ///Special chars that should be filtered out.
    static ref ALL_U32: Uniform<u32> = Uniform::new_inclusive(0u32, u32::max_value());
    //We use this so we don't have to generate the floating numbers and do comparisons on them. It allows us to do 1/2 percent level scaling.
    static ref ONE_TO_TWO_HUNDRED: Uniform<u8> = Uniform::new_inclusive(1, 200);
}

///Something over 200 chars isn't really suitable for this approach, so we won't accept it.
const MAX_STRING_LEN: usize = 200;

/// Make an index, for the string s considering all tri-grams.
/// The string will be latinised, lowercased and stripped of special chars before being broken into tri-grams.
/// The values will be prefixed with partition_id and salt before being hashed.
/// Each entry in the HashSet will be truncated to 32 bits and will be encoded as a big endian number.
/// This function will also add some random entries to the HashSet to not expose how many tri-grams were actually found.
pub fn generate_hashes_for_string_with_padding<R: Rng + CryptoRng>(
    s: &str,
    partition_id: Option<&str>,
    salt: &[u8],
    rng: &Mutex<R>,
) -> Result<HashSet<u32>, String> {
    let mut hashes = generate_hashes_for_string(s, partition_id, salt)?;

    let prob = take_lock(&rng).deref_mut().sample(*ONE_TO_TWO_HUNDRED);
    let to_add: u8 = {
        //Just take the lock once because we need it in all cases and it makes the code look better.
        let r = &mut *take_lock(&rng);
        if prob <= 1 {
            r.gen_range(1, 200)
        } else if prob <= 5 {
            r.gen_range(1, 30)
        } else if prob <= 50 {
            r.gen_range(1, 10)
        } else {
            r.gen_range(1, 5)
        }
    };
    //This will never be negative because generate_hashes_for_string would error if hashes was going to be larger than and will never be larger than MAX_STRING_LEN.
    //This also ensures we're able to pad by at least 2 since the maximum trigram length is always 2 less than the max string length.
    let pad_len = std::cmp::min(MAX_STRING_LEN - hashes.len(), to_add as usize);
    hashes.extend(
        take_lock(&rng)
            .deref_mut()
            .sample_iter(*ALL_U32)
            .take(pad_len),
    );
    Ok(hashes)
}

/// Make an index, for the string s considering all tri-grams.
/// The string will be latinised, lowercased and stripped of special chars before being broken into tri-grams.
/// The values will be prefixed with partition_id and salt before being hashed.
/// Each entry in the HasheSet will be truncated to 32 bits and will be encoded as a big endian number.
/// If the string is longer than 200 characters, this will return an error.
pub fn generate_hashes_for_string(
    s: &str,
    partition_id: Option<&str>,
    salt: &[u8],
) -> Result<HashSet<u32>, String> {
    if s.len() > MAX_STRING_LEN {
        Err(format!("The input string is too long. This function only supports strings that are no longer than {} chars.", MAX_STRING_LEN))
    } else {
        //Compute a partial sha256 with the partition_id and the salt - We can reuse this for each word
        let partial_sha256 = partition_id
            .map(|k| k.as_bytes())
            .iter()
            .chain([salt].iter())
            .fold(Sha256::new(), |hasher, k| hasher.chain(k));

        let short_hash = |word: &[u8]| -> u32 {
            let sha256_hash = partial_sha256.clone().chain(word);
            as_u32_be(&sha256_hash.result().into())
        };

        let result: HashSet<_> = make_tri_grams(s)
            .iter()
            .map(|tri_gram| short_hash(tri_gram.as_bytes()))
            .collect();
        Ok(result)
    }
}

/// Generate a version of the input string where each character has been latinized using the
/// same function as our tokenization routines.
pub fn transliterate_string(s: &str) -> String {
    s.chars()
        .filter(should_keep_char)
        .map(char_to_trans)
        .collect()
}

/// If s is empty, the resulting set will also be empty.
/// If s is shorter than 3, '-' padding will be added to the end.
/// All Strings inside of the resulting set will always be of size 3.
fn make_tri_grams(s: &str) -> HashSet<String> {
    let converted_string = transliterate_string(s);
    converted_string
        .unicode_words()
        .into_iter()
        .map(|short_word| {
            let short_word_len = short_word.chars().count();
            if short_word_len < 3 {
                //Pad the short_word with
                format!("{:-<3}", short_word)
            } else {
                short_word.to_string()
            }
        })
        .flat_map(|word| word_to_trigrams(&word))
        .collect()
}

fn word_to_trigrams(s: &str) -> HashSet<String> {
    s.chars()
        .tuple_windows()
        .map(|(c1, c2, c3)| format!("{}{}{}", c1, c2, c3))
        .collect()
}

///Convert the char if we can, if we can't just create a string out of the character.
fn char_to_trans(c: char) -> String {
    let trans_string = unidecode_char(c);
    if trans_string == "" {
        format!("{}", c)
    } else {
        trans_string.to_lowercase()
    }
}

///Interpret the most significant 4 bytes as a bigendian u32
#[inline]
fn as_u32_be(slice: &[u8; 32]) -> u32 {
    ((slice[0] as u32) << 24)
        + ((slice[1] as u32) << 16)
        + ((slice[2] as u32) << 8)
        + ((slice[3] as u32) << 0)
}

/// Acquire mutex in a blocking fashion. If the Mutex is or becomes poisoned, panic.
///
/// The lock is released when the returned MutexGuard falls out of scope.
///
/// # Usage:
/// single statement (mut)
/// `let result = take_lock(&t).deref_mut().call_method_on_t();`
///
/// multi-statement (mut)
///
/// ```ignore
/// let t = T {};
/// let result = {
///     let g = &mut *take_lock(&t);
///     g.call_method_on_t()
/// }; // lock released here
/// ```
///
fn take_lock<T>(m: &Mutex<T>) -> MutexGuard<T> {
    m.lock().unwrap_or_else(|e| {
        let error = format!("Error when acquiring lock: {}", e);
        panic!(error);
    })
}

#[cfg(test)]
mod tests {
    use super::*;
    use rand::rngs::ThreadRng;

    fn make_set(array: &[&str]) -> HashSet<String> {
        array
            .into_iter()
            .map(|&s| From::from(s))
            .collect::<HashSet<_>>()
    }

    #[test]
    fn as_u32_be_known_result() {
        let known_result = 16909060u32; //16777216 + 131072 + 768 + 4
        let mut input = [0u8; 32];
        input[0] = 1;
        input[1] = 2;
        input[2] = 3;
        input[3] = 4;
        let result = as_u32_be(&input);
        assert_eq!(result, known_result);
    }

    #[test]
    fn string_transliterated() {
        assert_eq!(transliterate_string("Gumby, dammit!"), "gumby dammit");
        assert_eq!(transliterate_string("北亰"), "bei jing ");
        assert_eq!(transliterate_string("Æneid"), "aeneid");
    }

    #[test]
    fn word_to_trigrams_known() {
        let result = word_to_trigrams("five");
        assert_eq!(result, make_set(&["fiv", "ive"]));
    }

    #[test]
    fn make_tri_grams_works_multi_word() {
        assert_eq!(
            make_tri_grams("123 José  Núñez 812-111-7654"),
            make_set(&[
                "123", "jos", "ose", "nun", "une", "nez", "812", "121", "211", "111", "117", "176",
                "765", "654",
            ])
        );
    }

    #[test]
    fn make_tri_grams_works_non_ascii() {
        assert_eq!(
            make_tri_grams("TİRYAKİ"),
            make_set(&["tir", "iry", "rya", "yak", "aki"])
        );
    }

    #[test]
    fn make_tri_grams_eliminates_duplicates() {
        assert_eq!(
            make_tri_grams("TİRYAKİ TİRYAKİ"),
            make_set(&["tir", "iry", "rya", "yak", "aki"])
        );
    }

    #[test]
    fn make_tri_grams_works_short_non_ascii() {
        assert_eq!(make_tri_grams("Tİ"), make_set(&["ti-"]));
    }

    #[test]
    fn make_tri_grams_works_multichar_translate() {
        assert_eq!(
            make_tri_grams("志    豪 İ"),
            make_set(&["zhi", "hao", "i--"])
        );
    }

    #[test]
    fn make_tri_grams_works_arabic() {
        assert_eq!(
            make_tri_grams("شريط فو"),
            make_set(&["shr", "hry", "ryt", "fw-"])
        );
    }
    #[test]
    fn make_tri_grams_works_short_multibyte() {
        assert_eq!(
            make_tri_grams("\u{102AE}\u{102AF}"),
            make_set(&["\u{102AE}\u{102AF}-"])
        );
    }

    #[test]
    fn char_to_trans_latinizable() {
        assert_eq!(char_to_trans('İ'), "i")
    }

    #[test]
    fn char_to_trans_not_latinizable() {
        let c = "\u{102AE}".chars().nth(0).unwrap();
        assert_eq!(char_to_trans(c), "\u{102AE}")
    }
    #[test]
    fn generate_hashes_for_string_compute_known_value() -> Result<(), String> {
        let result = generate_hashes_for_string("123", Some("foo"), &[0u8; 1])?;
        //We compute this to catch cases where this computation might change.
        let expected_result = {
            let mut hasher = Sha256::new();
            hasher.input("foo".as_bytes());
            hasher.input([0u8; 1]);
            hasher.input("123");
            as_u32_be(&(hasher.result().into()))
        };
        assert_eq!(result, [expected_result].iter().map(|x| *x).collect());
        Ok(())
    }

    #[test]
    fn generate_hashes_for_string_with_padding_adds_at_least_one() -> Result<(), String> {
        let rng = Mutex::new(ThreadRng::default());
        let result = generate_hashes_for_string_with_padding("123", Some("foo"), &[0u8; 1], &rng)?;
        assert!(result.len() > 1);
        Ok(())
    }

    #[test]
    fn generate_hashes_for_string_with_padding_empty_string() -> Result<(), String> {
        let rng = Mutex::new(ThreadRng::default());
        let result = generate_hashes_for_string_with_padding("", Some("foo"), &[0u8; 1], &rng)?;
        assert!(result.len() >= 1);
        Ok(())
    }

    #[test]
    fn generate_hashes_for_string_too_long_errors() -> Result<(), String> {
        let rng = ThreadRng::default();
        let input: String = rng
            .sample_iter(rand::distributions::Alphanumeric)
            .take(201)
            .collect();
        generate_hashes_for_string(&input, Some("foo"), &[0u8; 1]).unwrap_err();
        Ok(())
    }
}