hannoy 0.2.0

HNSW Approximate Nearest Neighbors in Rust, based on LMDB and optimized for memory usage
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
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173

import re
from pathlib import Path

import numpy as np
import pytest

import hannoy
from hannoy import Metric, Reader


@pytest.fixture(scope="function")
def array_db(tmp_path: Path) -> hannoy.Database:
    db = hannoy.Database(tmp_path, Metric.EUCLIDEAN)
    vectors = np.array(
        [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], dtype=np.float32
    )
    with db.writer(3, m=4, ef=10) as writer:
        writer.add_items([0, 1, 2], vectors)
    return db


def test_add_items_from_numpy_array(array_db: hannoy.Database) -> None:
    reader: Reader = array_db.reader(0)
    assert reader.by_vec([1.0, 0.0, 0.0], n=1) == [(0, 0.0)]
    assert reader.by_vec([0.0, 1.0, 0.0], n=1) == [(1, 0.0)]
    assert reader.by_vec([0.0, 0.0, 1.0], n=1) == [(2, 0.0)]


@pytest.mark.parametrize(
    ("vectors", "exc"),
    [
        pytest.param(
            np.array([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]], dtype=np.float32),
            ValueError(
                "add_items requires as many items as the array has rows: got 1 items for 2 rows"
            ),
            id="row-count-mismatch",
        ),
        pytest.param(
            np.array([1.0, 0.0, 0.0], dtype=np.float32),
            # https://github.com/PyO3/rust-numpy/issues/561
            TypeError("'ndarray' object is not an instance of 'ndarray'"),
            id="requires-2d",
        ),
    ],
)
def test_add_items_rejects_invalid_input(
    tmp_path: Path, vectors: np.ndarray, exc: Exception
) -> None:
    db = hannoy.Database(tmp_path, Metric.EUCLIDEAN)
    with pytest.raises(type(exc), match=re.escape(str(exc))):
        with db.writer(3, m=4, ef=10) as writer:
            writer.add_items([0], vectors)


def test_by_array_1d_matches_by_vec(array_db: hannoy.Database) -> None:
    reader: Reader = array_db.reader(0)
    query = np.array([1.0, 0.0, 0.0], dtype=np.float32)

    res = reader.by_array(query, n=2, ef_search=10)
    assert res == reader.by_vec([1.0, 0.0, 0.0], n=2, ef_search=10)


def test_by_array_2d_batches_queries(array_db: hannoy.Database) -> None:
    reader: Reader = array_db.reader(0)
    queries = np.array([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]], dtype=np.float32)

    res = reader.by_array(queries, n=1, ef_search=10)
    assert res == [[(0, 0.0)], [(1, 0.0)]]


def test_by_array_out_1d(array_db: hannoy.Database) -> None:
    reader: Reader = array_db.reader(0)
    query = np.array([1.0, 0.0, 0.0], dtype=np.float32)
    expected = reader.by_array(query, n=2, ef_search=10)

    ids = np.zeros(2, dtype=np.uint32)
    distances = np.zeros(2, dtype=np.float32)
    result = reader.by_array(query, n=2, ef_search=10, out=(ids, distances))

    assert result is None
    assert list(zip(ids.tolist(), distances.tolist())) == expected


def test_by_array_out_2d(array_db: hannoy.Database) -> None:
    reader: Reader = array_db.reader(0)
    queries = np.array([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]], dtype=np.float32)
    expected = reader.by_array(queries, n=1, ef_search=10)

    ids = np.zeros((2, 1), dtype=np.uint32)
    distances = np.zeros((2, 1), dtype=np.float32)
    result = reader.by_array(queries, n=1, ef_search=10, out=(ids, distances))

    assert result is None
    for row, expected_row in zip(range(2), expected):
        assert list(zip(ids[row].tolist(), distances[row].tolist())) == expected_row


def test_by_array_out_pads_missing_hits(array_db: hannoy.Database) -> None:
    reader: Reader = array_db.reader(0)
    query = np.array([1.0, 0.0, 0.0], dtype=np.float32)

    ids = np.zeros(5, dtype=np.uint32)
    distances = np.zeros(5, dtype=np.float32)
    reader.by_array(query, n=5, ef_search=10, out=(ids, distances))

    # only 3 items are indexed, so the last 2 slots are padding
    assert ids[3:].tolist() == [np.iinfo(np.uint32).max] * 2
    assert distances[3:].tolist() == [float("inf")] * 2


@pytest.mark.parametrize(
    ("query", "exc"),
    [
        pytest.param(
            np.array([1.0, 0.0, 0.0], dtype=np.float64),
            # https://github.com/PyO3/rust-numpy/issues/561
            TypeError("'ndarray' object is not an instance of 'ndarray'"),
            id="wrong-dtype",
        ),
        pytest.param(
            np.zeros((2, 2, 2), dtype=np.float32),
            ValueError("by_array requires a 1D or 2D array, got 3D"),
            id="wrong-ndim",
        ),
    ],
)
def test_by_array_rejects_invalid_query(
    array_db: hannoy.Database, query: np.ndarray, exc: Exception
) -> None:
    reader: Reader = array_db.reader(0)
    with pytest.raises(type(exc), match=re.escape(str(exc))):
        reader.by_array(query, n=1)


def _out_with_readonly_ids() -> tuple[np.ndarray, np.ndarray]:
    ids = np.zeros(2, dtype=np.uint32)
    ids.flags.writeable = False
    return (ids, np.zeros(2, dtype=np.float32))


@pytest.mark.parametrize(
    ("out", "exc"),
    [
        pytest.param(
            (np.zeros(2, dtype=np.uint32),),
            ValueError("expected tuple of length 2, but got tuple of length 1"),
            id="wrong-tuple-length",
        ),
        pytest.param(
            (np.zeros(2, dtype=np.int32), np.zeros(2, dtype=np.float32)),
            TypeError("'ndarray' object is not an instance of 'ndarray'"),
            id="wrong-ids-dtype",
        ),
        pytest.param(
            (np.zeros(3, dtype=np.uint32), np.zeros(2, dtype=np.float32)),
            ValueError("out[0] (ids) must have shape [2], got [3]"),
            id="wrong-ids-shape",
        ),
        pytest.param(
            _out_with_readonly_ids(),
            TypeError("The given array is not writeable"),
            id="read-only",
        ),
    ],
)
def test_by_array_out_rejects_invalid_out(
    array_db: hannoy.Database, out: tuple[np.ndarray, np.ndarray], exc: Exception
) -> None:
    reader: Reader = array_db.reader(0)
    query = np.array([1.0, 0.0, 0.0], dtype=np.float32)
    with pytest.raises(type(exc), match=re.escape(str(exc))):
        reader.by_array(query, n=2, out=out)