Databricks Python Data Sources

Introduced in Spark 4.x, Python Data Source API allows you to create PySpark Data Sources leveraging long standing python libraries for handling unique file types or specialized interfaces with spark read, readStream, write and writeStream APIs.

Data Source Name	Purpose
mcap	Read MCAP (ROS 2 bag) files
mqtt	Stream data from MQTT brokers
zipdcm	Read DICOM files from Zip file archives

Installation

Install the base package:

pip install python-data-sources

Install with specific data source support:

# Install with MCAP support
pip install python-data-sources[mcap]

# Install with MQTT support
pip install python-data-sources[mqtt]

# Install with ZipDCM support
pip install python-data-sources[zipdcm]

# Install with all data sources
pip install python-data-sources[all]

Quick Start

MCAP Data Source

from pyspark.sql import SparkSession
from python_data_sources.mcap import MCAPDataSource

spark = SparkSession.builder.getOrCreate()
spark.dataSource.register(MCAPDataSource)

df = (
    spark.read.format("mcap")
    .option("path", "/path/to/data.mcap")
    .load()
)

MQTT Streaming Data Source

from pyspark.sql import SparkSession
from python_data_sources.mqtt import MqttDataSource

spark = SparkSession.builder.getOrCreate()
spark.dataSource.register(MqttDataSource)

df = (
    spark.readStream.format("mqtt_pub_sub")
    .option("broker_address", "mqtt.example.com")
    .option("topic", "sensors/#")
    .load()
)

ZipDCM Data Source

from pyspark.sql import SparkSession
from python_data_sources.zipdcm import ZipDCMDataSource

spark = SparkSession.builder.getOrCreate()
spark.dataSource.register(ZipDCMDataSource)

df = (
    spark.read.format("zipdcm")
    .load("/path/to/dicom_files.zip")
)

Development

This project uses uv for build and environment management.

Setup

# Install uv
brew install uv

# Create development environment
make dev

Running Tests

# Run tests for a specific submodule
make test-module MODULE=mcap
make test-module MODULE=mqtt
make test-module MODULE=zipdcm

Project Structure

python-data-sources/
├── pyproject.toml              # Unified build configuration
├── src/
│   └── python_data_sources/    # Main package
│       ├── mcap/               # MCAP data source
│       ├── mqtt/               # MQTT streaming data source
│       └── zipdcm/             # ZipDCM data source
├── tests/
|   └── unit/
│       ├── common/             # Common module tests
│       ├── mcap/               # MCAP tests
│       ├── mqtt/               # MQTT tests
│       └── zipdcm/             # ZipDCM tests
└── .github/workflows/
    └── test.yml                # CI/CD workflow

Documentation

Refer to the python-data-sources documentation for detailed information on how to use supplied python data sources, its features, and configuration options.

Contributing

See CONTRIBUTING.md for detailed information about contributing to the Python Data Sources library.

📄 Third-Party Package Licenses

© 2025 Databricks, Inc. All rights reserved. The source in this project is provided subject to the Databricks License [https://databricks.com/db-license-source]. All included or referenced third party libraries are subject to the licenses set forth below.

Datasource	Package	Purpose	License	Source
mcap	mcap	Python API for MCAP files	MIT	https://github.com/foxglove/mcap
mcap	mcap-protobuf-support	Protobuf schema support	MIT	https://github.com/foxglove/mcap
mqtt	paho-mqtt	MQTT client library	EPL-2.0 / EDL-1.0 (BSD-3)	https://github.com/eclipse/paho.mqtt.python
zipdcm	pydicom	Python API for DICOM files	MIT	https://github.com/pydicom/pydicom
zipdcm	pylibjpeg	Decoding / Encoding pixel formats	MIT	https://github.com/pydicom/pylibjpeg
zipdcm	pylibjpeg-openjpeg	J2K, JP2, and HTJ2K plugin for pylibjpeg	MIT	https://github.com/pydicom/pylibjpeg-openjpeg
zipdcm	pylibjpeg-rle	RLE plugin for pylibjpeg	MIT	https://github.com/pydicom/pylibjpeg-rle