vicinity 0.11.1

Approximate nearest-neighbor search
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

from __future__ import annotations

import importlib.util
import json
from pathlib import Path
from types import ModuleType

import numpy as np


def load_script() -> ModuleType:
    script_path = (
        Path(__file__).resolve().parents[1] / "scripts/generate_multiscale_data.py"
    )
    spec = importlib.util.spec_from_file_location(
        "generate_multiscale_data",
        script_path,
    )
    assert spec is not None
    assert spec.loader is not None
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)
    return module


def write_tiny_scale(script: ModuleType, scale_dir: Path, scale: str) -> dict:
    scale_dir.mkdir()
    script.SCALES[scale] = {
        "n_train": 8,
        "n_test": 3,
        "dim": 4,
        "desc": "Tiny test scale",
    }
    settings = script.scale_settings(scale)
    k = settings["ground_truth_k"]
    script.save_vectors(scale_dir / "train.bin", np.zeros((8, 4), dtype=np.float32))
    script.save_vectors(scale_dir / "test.bin", np.zeros((3, 4), dtype=np.float32))
    script.save_neighbors(scale_dir / "neighbors.bin", np.zeros((3, k), dtype=np.int32))
    script.save_labels(scale_dir / "train_topics.bin", np.zeros(8, dtype=np.int32))
    script.save_f32_array(scale_dir / "test_lids.bin", np.zeros(3, dtype=np.float32))
    script.save_labels(scale_dir / "test_difficulty.bin", np.zeros(3, dtype=np.int32))
    script.save_vectors(
        scale_dir / "test_drift.bin",
        np.zeros((3, 4), dtype=np.float32),
    )
    script.save_neighbors(
        scale_dir / "neighbors_drift.bin",
        np.zeros((3, k), dtype=np.int32),
    )
    script.save_vectors(
        scale_dir / "test_filter.bin",
        np.zeros((3, 4), dtype=np.float32),
    )
    script.save_labels(
        scale_dir / "test_filter_topics.bin",
        np.zeros(3, dtype=np.int32),
    )
    script.save_neighbors(
        scale_dir / "neighbors_filter.bin",
        np.zeros((3, k), dtype=np.int32),
    )
    script.write_json_atomic(scale_dir / "metrics.json", {"meta": {"scale": scale}})
    script.write_json_atomic(
        scale_dir / "multiscale_dataset.json",
        script.output_manifest(scale_dir, settings),
    )
    return settings


def test_matching_scale_outputs_accepts_complete_manifest(tmp_path: Path) -> None:
    script = load_script()
    scale_dir = tmp_path / "Z"
    settings = write_tiny_scale(script, scale_dir, "Z")

    assert script.matching_scale_outputs(scale_dir, settings)


def test_output_manifest_records_hashes(tmp_path: Path) -> None:
    script = load_script()
    scale_dir = tmp_path / "Z"
    write_tiny_scale(script, scale_dir, "Z")

    manifest = json.loads((scale_dir / "multiscale_dataset.json").read_text())
    for info in manifest["outputs"].values():
        assert isinstance(info["bytes"], int)
        assert isinstance(info["sha256"], str)
        assert len(info["sha256"]) == 64


def test_matching_scale_outputs_rejects_truncated_payload(tmp_path: Path) -> None:
    script = load_script()
    scale_dir = tmp_path / "Z"
    settings = write_tiny_scale(script, scale_dir, "Z")
    (scale_dir / "train.bin").write_bytes((scale_dir / "train.bin").read_bytes()[:12])

    assert not script.matching_scale_outputs(scale_dir, settings)


def test_matching_scale_outputs_rejects_same_size_payload_drift(
    tmp_path: Path,
) -> None:
    script = load_script()
    scale_dir = tmp_path / "Z"
    settings = write_tiny_scale(script, scale_dir, "Z")
    train_path = scale_dir / "train.bin"
    payload = bytearray(train_path.read_bytes())
    payload[-1] ^= 0x01
    train_path.write_bytes(payload)

    assert not script.matching_scale_outputs(scale_dir, settings)