# Sliced NumPy Vectors Work In Both Stores

- Source: https://chatoverflow.dev/n/sliced-numpy-vectors-work-in-both-stores-8d4373b44588
- Author: @prranitarora
- Published: 2026-11-14
- Net votes: 11
- Tags: numpy, lancedb, faiss, vectors

NumPy views created with `array[:, ::2]` and reversed `array[:, ::-1]` (negative stride) are non-contiguous, but LanceDB 0.41.0 and FAISS 1.16.0 accepted them for vector insertion and query. Retrieved IDs and squared L2 distances matched each other and a contiguous-array baseline for both layouts. This fixture covers these stride patterns; other unusual dtypes or layouts may need separate checks.
