Skip to main content

linaldb (Python native bindings)

Embedded native Python bindings for LINALDB — a PyO3 extension linking the same synchronous engine the CLI/REPL runs (TensorDb/execute_line, src/engine/db.rs + src/dsl/mod.rs) directly into your Python process. No server, no network — this is the "use it like SQLite" story, as opposed to clients/python (a thin HTTP client for a running linal serve). See ../EMBEDDED_CONTRACT.md for the exact result shapes this module implements.

Published on PyPI as linaldb:

pip install linaldb

Status

Nine releases on PyPI so far, 0.1.0 through 0.1.8 — most bumped purely to pick up a real engine fix (this crate has no code of its own beyond the PyO3 bindings), each found via genuine end-to-end testing against real data in the linal-hub sibling project rather than isolated unit tests. Two of the more severe: 0.1.4 (2026-09-13) fixed EXPLAIN LINEAGE misattributing ancestry across a zero-copy TRANSPOSE (a transposed matrix and its untransposed source could hash identically); 0.1.5 (2026-09-13) fixed a JOIN's SELECT/WHERE/aggregate expressions silently returning the wrong table's value for a qualified column when both sides shared a bare column name — no error, just wrong data. See CHANGELOG.md for the complete version-by-version history.

Usage

import linaldb

db = linaldb.Db()  # persists to ./data by default, exactly like the CLI
db.execute("CREATE DATASET t COLUMNS (id: Int, score: Float)")
db.execute("INSERT INTO t VALUES (1, 0.9), (2, 0.4)")

result = db.execute("SELECT * FROM t WHERE score > 0.5")
print(result.columns, result.rows)   # ExecuteResult
df = result.to_pandas()              # requires the `pandas` extra

db.execute("SAVE DATASET t")
dataset = db.dataset("t")
df = dataset.to_pandas()             # reads data.parquet directly off disk

See examples/digit_classification_embedded.py and the Jupyter notebook examples/digit_classification_embedded.ipynb for a complete real-data walkthrough — the same real UCI handwritten-digits classification workflow clients/python/examples/digit_classification.py runs over HTTP, ported to embedded mode: no linal serve subprocess at all, just an in-process Db(), an in-process classification query, and an independent numpy recomputation from the dataset export, cross-checked against the SQL result.

Build

This crate vendors HDF5 and pulls in arrow/tokio/axum/zarrs as dependencies of the root linal crate (Cargo.toml here has a path dependency on ../..) — the first build compiles all of that, same as building the CLI itself; a .cargo/config.toml here points the build at the repo-root target/ dir to avoid a second full vendor build if you've already built the root crate.

Requires a Rust toolchain (cmake + a C toolchain for HDF5, per the root CLAUDE.md) and maturin:

python3 -m venv .venv && source .venv/bin/activate
pip install maturin
maturin develop           # builds the extension, installs it into .venv

If your Python interpreter is newer than PyO3's currently-supported maximum CPython version, set PYO3_USE_ABI3_FORWARD_COMPATIBILITY=1 for the build (PYO3_USE_ABI3_FORWARD_COMPATIBILITY=1 maturin develop) — this repo was built and verified against CPython 3.14 that way.

About

LINALDB is built by Gorigami, a software company based in Colombia, and maintained by Nicolás Balaguera. See the project README and LICENSE for the full picture and licensing terms.

Development

source .venv/bin/activate
pip install -e ".[dev]"
maturin develop
pytest tests/

No live server needed for anything here (that's the whole point of embedded mode) — tests and examples all run against an in-process Db().

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

linaldb-0.1.10-cp39-abi3-win_amd64.whl (10.1 MB view details)

Uploaded CPython 3.9+Windows x86-64

linaldb-0.1.10-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (9.6 MB view details)

Uploaded CPython 3.9+manylinux: glibc 2.17+ x86-64

linaldb-0.1.10-cp39-abi3-macosx_11_0_arm64.whl (8.0 MB view details)

Uploaded CPython 3.9+macOS 11.0+ ARM64

File details

Details for the file linaldb-0.1.10-cp39-abi3-win_amd64.whl.

File metadata

  • Download URL: linaldb-0.1.10-cp39-abi3-win_amd64.whl
  • Upload date:
  • Size: 10.1 MB
  • Tags: CPython 3.9+, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for linaldb-0.1.10-cp39-abi3-win_amd64.whl
Algorithm Hash digest
SHA256 986bce9f291e85130b2a73a0ee3889d7c49505c377f85a22a8ea843c1e7f09eb
MD5 8218202b648e6ab6387f3416058594f2
BLAKE2b-256 c763af35ea303c9d35731f91949bf81a7391b89559cf0610216e49538f5f5c55

See more details on using hashes here.

Provenance

The following attestation bundles were made for linaldb-0.1.10-cp39-abi3-win_amd64.whl:

Publisher: pypi-publish-embedded.yml on gorigami/linaldb

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file linaldb-0.1.10-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.

File metadata

File hashes

Hashes for linaldb-0.1.10-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Algorithm Hash digest
SHA256 b7d77ccc8ef2f0a30de574e251c973576b1c304ac04b1b50c6ae0c9b1d49f592
MD5 0d3855c1c9715000e449cebb3575ba5a
BLAKE2b-256 41e06fc7f5024c6859abadb3373c4a2a1dba596a2c116f8ec4d10b8808d6eb9b

See more details on using hashes here.

Provenance

The following attestation bundles were made for linaldb-0.1.10-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl:

Publisher: pypi-publish-embedded.yml on gorigami/linaldb

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file linaldb-0.1.10-cp39-abi3-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for linaldb-0.1.10-cp39-abi3-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 e149d68b7a4cb901201a7e18d36783e840eb8882f0d61ea20519c0fcd987c7e6
MD5 15a76f36df27a12078d25826a2ed4109
BLAKE2b-256 bc5c989d1f575763783789ba554b73df9bffe4e316298775538e7566c52e07ab

See more details on using hashes here.

Provenance

The following attestation bundles were made for linaldb-0.1.10-cp39-abi3-macosx_11_0_arm64.whl:

Publisher: pypi-publish-embedded.yml on gorigami/linaldb

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Release history Release notifications | RSS feed

0.1.12

3 files

0.1.11

3 files

This release

0.1.10 This release

3 files

0.1.9

3 files

0.1.8

3 files

0.1.7

3 files

0.1.6

3 files

0.1.5

3 files

0.1.4

3 files

0.1.3

3 files

0.1.2

3 files

0.1.1

3 files

0.1.0

3 files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page