deltalake 1.6.1

Native Delta Lake Python binding based on delta-rs with Pandas integration

Deltalake-python

Native Delta Lake Python binding based on delta-rs with Pandas integration.

Example

from deltalake import DeltaTable
dt = DeltaTable("../rust/tests/data/delta-0.2.0")
dt.version()
3
dt.file_uris()
['s3://bucket/table/part-00000-cb6b150b-30b8-4662-ad28-ff32ddab96d2-c000.snappy.parquet',
 's3://bucket/table/part-00000-7c2deba3-1994-4fb8-bc07-d46c948aa415-c000.snappy.parquet',
 's3://bucket/table/part-00001-c373a5bd-85f0-4758-815e-7eb62007a15c-c000.snappy.parquet']

See the user guide for more examples.

Installation

# with pip
pip install deltalake
# with uv
uv add deltalake
# with poetry
poetry add deltalake

NOTE: official binary wheels are linked against openssl statically for remote object store communication. Please file a Github issue to request a critical openssl upgrade.

Tracing and Observability

Delta-rs supports OpenTelemetry tracing for performance analysis and debugging.

Basic Example

import os
import deltalake
from deltalake import write_deltalake, DeltaTable

# Enable logging to see trace output in stdout
os.environ["RUST_LOG"] = "deltalake=debug"

# Initialize tracing (uses default HTTP endpoint or OTEL_EXPORTER_OTLP_ENDPOINT env var)
# For authentication, set OTEL_EXPORTER_OTLP_HEADERS="x-honeycomb-team=your-api-key"
# The HTTP exporter automatically reads OTEL_EXPORTER_OTLP_HEADERS for API keys
deltalake.init_tracing()

# All Delta operations are now traced
write_deltalake("my_table", data)
dt = DeltaTable("my_table")
df = dt.to_pandas()

When you run this code, you'll see trace information in stdout showing operation timings and execution flow.

Build custom wheels

Sometimes you may wish to build custom wheels. Maybe you want to try out some unreleased features. Or maybe you want to tweak the optimization of the Rust code.

To compile the package, you will need the Rust compiler and maturin:

curl https://sh.rustup.rs -sSf | sh -s

Then you can build wheels for your own platform like so:

uvx maturin build --release --out wheels

Note:

• uvx invokes a tool without installing it.
• if you plan to often use maturin, you can install the "tool" with uv tool install maturin.

For a build that is optimized for the system you are on (but sacrificing portability):

RUSTFLAGS="-C target-cpu=native" uvx maturin build --release --out wheels

Cross compilation

The above command only works for your current platform. To create wheels for other platforms, you'll need to cross compile. Cross compilation requires installing two additional components: to cross compile Rust code, you will need to install the target with rustup; to cross compile the Python bindings, you will need to install ziglang.

The following example is for manylinux2014. Other targets will require different Rust target and Python compatibility tags.

rustup target add x86_64-unknown-linux-gnu

Then you can build wheels for the target platform like so:

uvx --from 'maturin[zig]' maturin build --release --zig \
    --target x86_64-unknown-linux-gnu \
    --compatibility manylinux2014 \
    --out wheels

If you expect to only run on more modern system, you can set a newer target-cpu flag to Rust and use a newer compatibility tag for Linux. For example, here we set compatibility with CPUs newer than Haswell (2013) and Linux OS with glibc version of at least 2.24:

RUSTFLAGS="-C target-cpu=haswell" uvx --from 'maturin[zig]' maturin build --release --zig \
    --target x86_64-unknown-linux-gnu \
    --compatibility manylinux_2_24 \
    --out wheels

See note about RUSTFLAGS from the arrow-rs readme.