This is a Python library that binds to Apache Arrow in-memory query engine DataFusion.
Like pyspark, it allows you to build a plan through SQL or a DataFrame API against in-memory data, parquet or CSV files, run it in a multi-threaded environment, and obtain the result back in Python.
It also allows you to use UDFs and UDAFs for complex operations.
The major advantage of this library over other execution engines is that this library achieves zero-copy between Python and its execution engine: there is no cost in using UDFs, UDAFs, and collecting the results to Python apart from having to lock the GIL when running those operations.
Its query engine, DataFusion, is written in Rust, which makes strong assumptions about thread safety and lack of memory leaks.
Technically, zero-copy is achieved via the c data interface.
The following example demonstrates running a SQL query against a Parquet file using DataFusion, storing the results in a Pandas DataFrame, and then plotting a chart.
The Parquet file used in this example can be downloaded from the following page:
See the examples directory for more examples.
from datafusion import SessionContext
import pandas as pd
import pyarrow as pa
# Create a DataFusion context
ctx = SessionContext()
# Register table with context
ctx.register_parquet('taxi', 'yellow_tripdata_2021-01.parquet')
# Execute SQL
df = ctx.sql("select passenger_count, count(*) "
"from taxi "
"where passenger_count is not null "
"group by passenger_count "
"order by passenger_count")
# collect as list of pyarrow.RecordBatch
results = df.collect()
# get first batch
batch = results[0]
# convert to Pandas
df = batch.to_pandas()
# create a chart
fig = df.plot(kind="bar", title="Trip Count by Number of Passengers").get_figure()
fig.savefig('chart.png')This produces the following chart:
arrow-datafusion-python has bindings which allow for serializing a SQL query to substrait protobuf format and deserializing substrait protobuf bytes to a DataFusion LogicalPlan, PyLogicalPlan in a Python context, which can then be executed.
from datafusion import SessionContext
from datafusion import substrait as ss
# Create a DataFusion context
ctx = SessionContext()
# Register table with context
ctx.register_parquet('aggregate_test_data', './testing/data/csv/aggregate_test_100.csv')
substrait_plan = ss.substrait.serde.serialize_to_plan("SELECT * FROM aggregate_test_data", ctx)
# type(substrait_plan) -> <class 'datafusion.substrait.plan'>
# Alternative serialization approaches
# type(substrait_bytes) -> <class 'list'>, at this point the bytes can be distributed to file, network, etc safely
# where they could subsequently be deserialized on the receiving end.
substrait_bytes = ss.substrait.serde.serialize_bytes("SELECT * FROM aggregate_test_data", ctx)
# Imagine here bytes would be read from network, file, etc ... for example brevity this is omitted and variable is simply reused
# type(substrait_plan) -> <class 'datafusion.substrait.plan'>
substrait_plan = ss.substrait.serde.deserialize_bytes(substrait_bytes)
# type(df_logical_plan) -> <class 'substrait.LogicalPlan'>
df_logical_plan = ss.substrait.consumer.from_substrait_plan(ctx, substrait_plan)
# Back to Substrait Plan just for demonstration purposes
# type(substrait_plan) -> <class 'datafusion.substrait.plan'>
substrait_plan = ss.substrait.producer.to_substrait_plan(df_logical_plan)pip install datafusion
# or
python -m pip install datafusionconda install -c conda-forge datafusionYou can verify the installation by running:
>>> import datafusion
>>> datafusion.__version__
'0.6.0'This assumes that you have rust and cargo installed. We use the workflow recommended by pyo3 and maturin.
The Maturin tools used in this workflow can be installed either via Conda or Pip. Both approaches should offer the same experience. Multiple approaches are only offered to appease developer preference. Bootstrapping for both Conda and Pip are as follows.
Bootstrap (Conda):
# fetch this repo
git clone git@github.com:apache/arrow-datafusion-python.git
# create the conda environment for dev
conda env create -f ./conda/environments/datafusion-dev.yaml -n datafusion-dev
# activate the conda environment
conda activate datafusion-devBootstrap (Pip):
# fetch this repo
git clone git@github.com:apache/arrow-datafusion-python.git
# prepare development environment (used to build wheel / install in development)
python3 -m venv venv
# activate the venv
source venv/bin/activate
# update pip itself if necessary
python -m pip install -U pip
# install dependencies (for Python 3.8+)
python -m pip install -r requirements-310.txtThe tests rely on test data in git submodules.
git submodule init
git submodule updateWhenever rust code changes (your changes or via git pull):
# make sure you activate the venv using "source venv/bin/activate" first
maturin develop
python -m pytestTo change test dependencies, change the requirements.in and run
# install pip-tools (this can be done only once), also consider running in venv
python -m pip install pip-tools
python -m piptools compile --generate-hashes -o requirements-310.txtTo update dependencies, run with -U
python -m piptools compile -U --generate-hashes -o requirements-310.txtMore details here