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
//! Myers' bit-vector algorithm for approximate string matching.
//!
//! Reference: "A fast bit-vector algorithm for approximate string matching
//! based on dynamic programming" by Gene Myers (1999/JACM).
#![allow(clippy::needless_range_loop)]
use super::damlev::{DamLevMatch, EditLimits};
/// Myers bit-vector matcher - uses bit-parallel DP for O(n) performance.
#[derive(Debug)]
pub struct MyersMatcher {
pattern_len: usize,
edit_limits: EditLimits,
#[allow(dead_code)]
pub masks: Vec<u64>,
#[allow(dead_code)]
accept_mask: u64,
}
impl MyersMatcher {
/// Create a new Myers matcher.
#[must_use]
pub fn new(pattern: &str, edit_limits: EditLimits, case_insensitive: bool) -> Option<Self> {
if case_insensitive {
return None;
}
let pattern_bytes = pattern.as_bytes();
let m = pattern_bytes.len();
if m == 0 || m > 63 {
return None;
}
// Build character masks
let mut masks = vec![0u64; 256];
for (i, &byte) in pattern_bytes.iter().enumerate() {
masks[byte as usize] |= 1u64 << i;
}
Some(MyersMatcher {
pattern_len: m,
edit_limits,
masks,
accept_mask: 1u64 << (m - 1),
})
}
/// Calculate edit distance from current state.
#[allow(dead_code)]
#[inline]
#[allow(clippy::unused_self)]
fn get_score(&self, vp: u64, vn: u64) -> usize {
let score_mask = !vp & vn;
if score_mask == 0 {
return 0;
}
score_mask.trailing_zeros() as usize
}
/// Find first match using Myers' algorithm.
/// Uses dynamic programming to track minimum score at each position.
#[inline]
#[must_use]
pub fn find_first(&self, text: &str, threshold: f32) -> Option<DamLevMatch> {
let max_edits = self.edit_limits.max_edits as usize;
let m = self.pattern_len;
if m == 0 {
return Some(DamLevMatch {
start: 0,
end: 0,
insertions: 0,
deletions: 0,
substitutions: 0,
swaps: 0,
similarity: 1.0,
});
}
let text = text.as_bytes();
let n = text.len();
if n == 0 {
if m <= max_edits {
let sim = 1.0 - (m as f32 / (m + max_edits) as f32);
return Some(DamLevMatch {
start: 0,
end: 0,
insertions: 0,
deletions: m as u8,
substitutions: 0,
swaps: 0,
similarity: sim,
});
}
return None;
}
// Track minimum score at each position
let mut min_score = m + 1;
let mut min_pos = 0;
// Run Myers through text
let mut vp: u64 = !0u64;
let mut vn: u64 = 0u64;
let mut score = m;
for i in 0..n {
let eq = self.masks[text[i] as usize];
let x = eq | vn;
let y = (vp & x).wrapping_add(vp);
let d0 = (y ^ vp) | x;
let hp = vn | !d0;
let hn = vp & d0;
if hp & (1u64 << (m - 1)) != 0 {
score = score.saturating_add(1);
}
if hn & (1u64 << (m - 1)) != 0 {
score = score.saturating_sub(1);
}
let hp_shift = (hp << 1) | 1;
let vn_shift = hn << 1;
vp = vn_shift | !(x | hp_shift);
vn = hp_shift & x;
// After at least m chars processed, check for match
// The match could end at position i, starting at i-m+1
if i >= m - 1 && score <= max_edits && score < min_score {
min_score = score;
min_pos = i + 1;
}
}
if min_score <= max_edits {
let sim = 1.0 - (min_score as f32 / (m + max_edits) as f32);
if sim >= threshold {
let start = min_pos.saturating_sub(m);
return Some(DamLevMatch {
start,
end: min_pos,
insertions: 0,
deletions: min_score as u8,
substitutions: 0,
swaps: 0,
similarity: sim,
});
}
}
None
}
/// Find all matches.
#[must_use]
pub fn find_all(&self, text: &str, threshold: f32) -> Vec<DamLevMatch> {
let max_edits = self.edit_limits.max_edits as usize;
let m = self.pattern_len;
let mut matches = Vec::new();
if m == 0 {
return vec![DamLevMatch {
start: 0,
end: 0,
insertions: 0,
deletions: 0,
substitutions: 0,
swaps: 0,
similarity: 1.0,
}];
}
let text = text.as_bytes();
let n = text.len();
if n == 0 {
if m <= max_edits {
let sim = 1.0 - (m as f32 / (m + max_edits) as f32);
matches.push(DamLevMatch {
start: 0,
end: 0,
insertions: 0,
deletions: m as u8,
substitutions: 0,
swaps: 0,
similarity: sim,
});
}
return matches;
}
let mut vp: u64 = !0u64;
let mut vn: u64 = 0u64;
let mut score = m;
// For find_all, we need to track matches at each position
// since the minimum score could be at any position
for i in 0..n {
let eq = self.masks[text[i] as usize];
let x = eq | vn;
let y = (vp & x).wrapping_add(vp);
let d0 = (y ^ vp) | x;
let hp = vn | !d0;
let hn = vp & d0;
if hp & (1u64 << (m - 1)) != 0 {
score = score.saturating_add(1);
}
if hn & (1u64 << (m - 1)) != 0 {
score = score.saturating_sub(1);
}
let hp_shift = (hp << 1) | 1;
let vn_shift = hn << 1;
vp = vn_shift | !(x | hp_shift);
vn = hp_shift & x;
// Check for match after processing at least m characters
if i >= m - 1 && score <= max_edits {
let sim = 1.0 - (score as f32 / (m + max_edits) as f32);
if sim >= threshold {
let end = i + 1;
let start = end.saturating_sub(m);
matches.push(DamLevMatch {
start,
end,
insertions: 0,
deletions: score as u8,
substitutions: 0,
swaps: 0,
similarity: sim,
});
}
}
}
matches
}
}