oasysdb 0.4.0

Fast embedded vector database with incremental HNSW indexing.
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

use super::*;

/// The distance function used for similarity calculations.
#[derive(Debug, Serialize, Deserialize, Clone, Copy)]
#[derive(PartialEq, Eq, Hash, PartialOrd, Ord)]
pub enum Distance {
    /// Dot product function.
    Dot,
    /// Euclidean distance function.
    Euclidean,
    /// Cosine similarity function.
    Cosine,
}

impl Distance {
    /// Creates a new distance function from a string.
    /// Available options:
    /// * `dot`: Dot product function.
    /// * `euclidean`: Euclidean distance function.
    /// * `cosine`: Cosine similarity function.
    pub fn from(distance: &str) -> Result<Self, Error> {
        match distance {
            "dot" => Ok(Distance::Dot),
            "euclidean" => Ok(Distance::Euclidean),
            "cosine" => Ok(Distance::Cosine),
            _ => Err("Distance function not supported.".into()),
        }
    }

    /// Calculates the distance between two vectors.
    pub fn calculate(&self, a: &Vector, b: &Vector) -> f32 {
        assert_eq!(a.0.len(), b.0.len());
        match self {
            Distance::Dot => Distance::dot(a, b),
            Distance::Euclidean => Distance::euclidean(a, b),
            Distance::Cosine => Distance::cosine(a, b),
        }
    }

    // List additional distance functions below.

    fn dot(a: &Vector, b: &Vector) -> f32 {
        let zip = a.0.iter().zip(b.0.iter());
        zip.map(|(x, y)| x * y).sum()
    }

    fn cosine(a: &Vector, b: &Vector) -> f32 {
        let dot = Self::dot(a, b);
        let ma = a.0.iter().map(|x| x.powi(2)).sum::<f32>().sqrt();
        let mb = b.0.iter().map(|y| y.powi(2)).sum::<f32>().sqrt();
        dot / (ma * mb)
    }

    fn euclidean(a: &Vector, b: &Vector) -> f32 {
        let zip = a.0.iter().zip(b.0.iter());
        zip.map(|(a, b)| (a - b).powi(2)).sum::<f32>().sqrt()
    }
}

impl From<&PyAny> for Distance {
    fn from(distance: &PyAny) -> Self {
        let distance = distance.str().unwrap().to_string();
        Distance::from(&distance).unwrap()
    }
}

impl IntoPy<Py<PyAny>> for Distance {
    fn into_py(self, py: Python) -> Py<PyAny> {
        match self {
            Distance::Dot => "dot".into_py(py),
            Distance::Euclidean => "euclidean".into_py(py),
            Distance::Cosine => "cosine".into_py(py),
        }
    }
}