ballista 54.0.0

Python client for Apache Arrow Ballista Distributed SQL Query Engine

Ballista

Ballista support for datafusion python.

This project is tracked under its own Cargo.toml and is intentionally not part of the default Cargo workspace so that it doesn't cause overhead for maintainers of the main Ballista codebase. Its version is bumped in lockstep with the workspace crates by dev/update_ballista_versions.py, and the wheels are built against the in-repo ballista crates via path dependencies (not crates.io), so an RC can produce wheels for an unpublished version.

Creating a SessionContext

[!IMPORTANT] Current approach is to support datafusion python API, there are know limitations of current approach, with some cases producing errors.

We are trying to come up with the best approach to support ballista python interface.

More details could be found at #1142

Creates a new context which connects to a Ballista scheduler process.

from datafusion import col, lit
from datafusion import DataFrame
# we do not need datafusion context
# it will be replaced by BallistaSessionContext
# from datafusion import SessionContext
from ballista import BallistaSessionContext

# Change from:
#
# ctx = SessionContext()
#
# to: 

ctx = BallistaSessionContext("df://localhost:50050")

# all other functions and functions are from
# datafusion module
ctx.sql("create external table t stored as parquet location './testdata/test.parquet'")
df : DataFrame = ctx.sql("select * from t limit 5")

df.show()

Known limitations and inefficiencies of the current approach:

• The client's SessionConfig is not propagated to Ballista.
• Ballista-specific configuration cannot be set.
• Anything requiring custom datafusion_proto::logical_plan::LogicalExtensionCodec.
• No support for UDF as DataFusion Python does not serialise them.
• A Ballista connection will be created for each request.

Example DataFrame Usage

ctx = BallistaSessionContext("df://localhost:50050")
df = ctx.read_parquet('./testdata/test.parquet').filter(col(id) > lit(4)).limit(5)

pyarrow_batches = df.collect()

Check DataFusion python provides more examples and manuals.

Jupyter Notebook Support

PyBallista provides first-class Jupyter notebook support with SQL magic commands and rich HTML rendering.

Install Jupyter extras first:

pip install "ballista[jupyter]"

HTML Table Rendering

DataFrames automatically render as styled HTML tables in Jupyter notebooks:

from ballista import BallistaSessionContext

ctx = BallistaSessionContext("df://localhost:50050")
df = ctx.sql("SELECT * FROM my_table LIMIT 10")
df  # Renders as HTML table via _repr_html_()

SQL Magic Commands

For a more interactive SQL experience, load the Ballista Jupyter extension:

# Load the extension
%load_ext ballista.jupyter

# Connect to a Ballista cluster
%ballista connect df://localhost:50050

# Register .parquet table
%register parquet public.test_data_v1 ../testdata/test.parquet

# Check connection status
%ballista status

# List registered tables
%ballista tables

# Show table schema
%ballista schema my_table

# Execute a simple query (line magic)
%sql SELECT COUNT(*) FROM orders

# Execute a complex query (cell magic)
%%sql
SELECT
    customer_id,
    SUM(amount) as total
FROM orders
GROUP BY customer_id
ORDER BY total DESC
LIMIT 10

You can also store results in a variable:

%%sql my_result
SELECT * FROM orders WHERE status = 'pending'

Execution Plan Visualization

Visualize query execution plans directly in notebooks:

df = ctx.sql("SELECT * FROM orders WHERE amount > 100")
df.explain_visual()  # Displays SVG visualization

# With runtime statistics
df.explain_visual(analyze=True)

Note: Full SVG visualization requires graphviz to be installed (brew install graphviz on macOS).

Progress Indicators

For long-running queries, use collect_with_progress() to see execution status:

df = ctx.sql("SELECT * FROM large_table")
batches = df.collect_with_progress()

Example Notebooks

See the examples/ directory for Jupyter notebooks demonstrating various features:

• getting_started.ipynb - Basic connection and queries
• dataframe_api.ipynb - DataFrame transformations
• distributed_queries.ipynb - Multi-stage distributed query examples

Scheduler and Executor

Scheduler and executors can be configured and started from python code.

To start scheduler:

from ballista import BallistaScheduler

scheduler = BallistaScheduler()

scheduler.start()
scheduler.wait_for_termination()

For executor:

from ballista import BallistaExecutor

executor = BallistaExecutor()

executor.start()
executor.wait_for_termination()

Development Process

Detailed development process explanation can be found in datafusion python documentation. Improving build speed section can be relevant.

Creating Virtual Environment

pip

python3 -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

uv

uv sync --dev --no-install-package ballista

Developing & Building

pip

maturin develop

Note that you can also run maturin develop --release to get a release build locally.

uv

uv run --no-project maturin develop --uv

Or uv run --no-project maturin build --release --strip to get a release build.

Testing

pip

python3 -m pytest

uv

uv run --no-project pytest