similari 0.2.5

Vector similarity search engine library
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

use anyhow::Result;
use itertools::Itertools;

/// Trait to implement distance voting engines.
///
/// Distance voting engine is used to select winning tracks among distances
/// resulted from the distance calculation.
///
pub trait Voting<R> {
    /// Method that selects winning tracks
    ///
    ///
    /// # Arguments
    /// * `distances` - distances resulted from the distance calculation.
    ///   * `.0` is the track_id
    ///   * `.1` is the distance
    ///
    fn winners(&self, distances: &[(u64, Result<f32>)]) -> Vec<R>;
}

/// TopN winners voting engine that selects Top N vectors with most close distances.
///
/// It calculates winners as:
/// 1. removes all distances that are greater than threshold
/// 2. sorts remaining tracks according to their IDs
/// 3. counts tracks by their ID's
/// 4. sorts groups by frequency decreasingly
/// 5. returns TopN
///
pub struct TopNVoting {
    topn: usize,
    max_distance: f32,
    min_votes: usize,
}

impl TopNVoting {
    /// Constructs new engine
    ///
    /// # Arguments
    /// * `topn` - top winners
    /// * `max_distance` - max distance permitted to participate
    /// * `min_votes` - minimal amount of votes required the track to participate
    ///
    pub fn new(topn: usize, max_distance: f32, min_votes: usize) -> Self {
        Self {
            topn,
            max_distance,
            min_votes,
        }
    }
}

/// Return type fot TopN voting engine
///
#[derive(Default, Debug, PartialEq)]
pub struct TopNVotingElt {
    /// winning track
    pub track_id: u64,
    /// number of votes it gathered
    pub votes: usize,
}

impl TopNVotingElt {
    pub fn new(track_id: u64, votes: usize) -> Self {
        Self { track_id, votes }
    }
}

impl Voting<TopNVotingElt> for TopNVoting {
    fn winners(&self, distances: &[(u64, Result<f32>)]) -> Vec<TopNVotingElt> {
        let mut tracks: Vec<_> = distances
            .iter()
            .filter(|(_, e)| match e {
                Ok(e) => *e <= self.max_distance,
                _ => false,
            })
            .map(|(track, _)| track)
            .collect();
        tracks.sort_unstable();
        let mut counts = tracks
            .into_iter()
            .counts()
            .into_iter()
            .filter(|(_, count)| *count >= self.min_votes)
            .map(|(e, c)| TopNVotingElt {
                track_id: *e,
                votes: c,
            })
            .collect::<Vec<_>>();
        counts.sort_by(|l, r| r.votes.partial_cmp(&l.votes).unwrap());
        counts.truncate(self.topn);
        counts
    }
}

#[cfg(test)]
mod tests {
    use crate::track::voting::{TopNVoting, TopNVotingElt, Voting};

    #[test]
    fn default_voting() {
        let v = TopNVoting {
            topn: 5,
            max_distance: 0.32,
            min_votes: 1,
        };

        let candidates = v.winners(&vec![(1, Ok(0.2))]);
        assert_eq!(candidates, vec![TopNVotingElt::new(1, 1)]);

        let candidates = v.winners(&vec![(1, Ok(0.2)), (1, Ok(0.3))]);
        assert_eq!(candidates, vec![TopNVotingElt::new(1, 2)]);

        let candidates = v.winners(&vec![(1, Ok(0.2)), (1, Ok(0.4))]);
        assert_eq!(candidates, vec![TopNVotingElt::new(1, 1)]);

        let mut candidates = v.winners(&vec![(1, Ok(0.2)), (2, Ok(0.2))]);
        candidates.sort_by(|l, r| l.track_id.partial_cmp(&r.track_id).unwrap());
        assert_eq!(
            candidates,
            vec![TopNVotingElt::new(1, 1), TopNVotingElt::new(2, 1)]
        );

        let mut candidates = v.winners(&vec![
            (1, Ok(0.2)),
            (1, Ok(0.22)),
            (2, Ok(0.21)),
            (2, Ok(0.2)),
            (3, Ok(0.22)),
            (3, Ok(0.2)),
            (4, Ok(0.23)),
            (4, Ok(0.3)),
            (5, Ok(0.24)),
            (5, Ok(0.3)),
            (6, Ok(0.25)),
            (6, Ok(0.5)),
        ]);
        candidates.sort_by(|l, r| l.track_id.partial_cmp(&r.track_id).unwrap());
        assert_eq!(
            candidates,
            vec![
                TopNVotingElt::new(1, 2),
                TopNVotingElt::new(2, 2),
                TopNVotingElt::new(3, 2),
                TopNVotingElt::new(4, 2),
                TopNVotingElt::new(5, 2)
            ]
        );
    }
}