oasysdb 0.2.0

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
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

use super::*;

const DIMENSION: usize = 128;
const LEN: usize = 100;

#[test]
fn build_large() {
    let len = 10000;
    let records = Record::many_random(DIMENSION, len);
    let collection = create_collection(&records);
    assert_eq!(collection.len(), len);
}

#[test]
fn insert() {
    let records = Record::many_random(DIMENSION, LEN);
    let mut collection = create_collection(&records);

    // Create a new record to insert.
    let new_record = Record::random(DIMENSION);
    collection.insert(&new_record).unwrap();

    assert_eq!(collection.len(), LEN + 1);

    // Assert the new data is in the collection.
    let id: VectorID = LEN.into();
    assert_eq!(collection.get(&id).unwrap().data, new_record.data);
}

#[test]
fn insert_invalid_dimension() {
    let records = Record::many_random(DIMENSION, LEN);
    let mut collection = create_collection(&records);

    // Create a new record with an invalid dimension.
    let new_record = Record::random(DIMENSION + 1);

    assert_eq!(collection.dimension(), DIMENSION);

    // Assert the new record is not inserted.
    assert_eq!(collection.insert(&new_record).is_err(), true);
}

#[test]
fn insert_data_type_object() {
    let records = Record::many_random(DIMENSION, LEN);
    let mut collection = create_collection(&records);

    // Create a new record with a data of type HashMap.
    let vector = Vector::random(DIMENSION);
    let data = HashMap::from([("key", "value")]);
    let new_record = Record::new(&vector, &data.clone().into());

    collection.insert(&new_record).unwrap();

    assert_eq!(collection.len(), LEN + 1);

    // Assert the new data is in the collection.
    let id: VectorID = LEN.into();
    assert_eq!(collection.get(&id).unwrap().data, data.into());
}

#[test]
fn delete() {
    let records = Record::many_random(DIMENSION, LEN);
    let mut collection = create_collection(&records);

    // Delete a record from the collection.
    let id = VectorID(1);
    collection.delete(&id).unwrap();

    assert_eq!(collection.len(), LEN - 1);
}

#[test]
fn update() {
    let records = Record::many_random(DIMENSION, LEN);
    let mut collection = create_collection(&records);

    // New record to update.
    let id = VectorID(5);
    let record = Record::random(DIMENSION);
    collection.update(&id, &record).unwrap();

    assert_eq!(collection.len(), LEN);
    assert_eq!(collection.get(&id).unwrap().data, record.data);
}

#[test]
fn search() {
    let len = 1000;
    let records = Record::many_random(DIMENSION, len);
    let collection = create_collection(&records);

    // Generate a random query vector.
    let query = Vector::random(DIMENSION);

    // Get the approximate and true nearest neighbors.
    let result = collection.search(&query, 5).unwrap();
    let truth = collection.true_search(&query, 10).unwrap();

    // Collect the distances from the true nearest neighbors.
    let distances: Vec<f32> = truth.par_iter().map(|i| i.distance).collect();

    assert_eq!(result.len(), 5);

    // The search is not always exact, so we check if
    // the distance is within the true distances.
    assert_eq!(distances.contains(&result[0].distance), true);
}

#[test]
fn get() {
    let records = Record::many_random(DIMENSION, LEN);
    let collection = create_collection(&records);

    // Get a record from the collection.
    let id = VectorID(5);
    let record = collection.get(&id).unwrap();

    assert_eq!(record.data, records[5].data);
    assert_eq!(record.vector, records[5].vector);
}