Bug: MultiIndex returns strange search results

[kimihailv]

Describe the bug

MultiIndex (even from one index) gives wrong nearest neighbours.

Steps to reproduce

import numpy as np
from usearch.index import Indexes, Index

np.random.seed(42)
data = np.random.random((10_000, 256))

index = Index(ndim=256)
index.add(np.arange(data.shape[0]), data)

gt = np.arange(1000)[:, np.newaxis]
print('Single index recall@1:', (index.search(data[:1000], 1).keys == gt).sum(axis=1).mean())

multi = Indexes(
    indexes=[index]
)

print('Multi index recall@1:', (multi.search(data[:1000], 1).keys == gt).sum(axis=1).mean())

Output:

Single index recall@1: 0.927
Multi index recall@1: 0.001

Expected behavior

Similar performance

USearch version

2.5.1

Operating System

Ubuntu 23.04

Hardware architecture

x86

Which interface are you using?

Python bindings

Contact Details

No response

Is there an existing issue for this?

• [X] I have searched the existing issues

Code of Conduct

• [X] I agree to follow this project's Code of Conduct

[ashvardanian]

This was just a matter of putting the return statement one line lower 🤦 Thanks!

[ashvardanian]

:tada: This issue has been resolved in version 2.7.1 :tada:

The release is available on GitHub release

Your semantic-release bot :package::rocket: