Skip to content

Contributing to the Iceberg Python library

For the development, Poetry is used for packing and dependency management. You can install this using:

pip install poetry

Make sure you're using an up-to-date environment from venv

pip install --upgrade virtualenv pip
python -m venv ./venv
source ./venv/bin/activate

To get started, you can run make install, which installs Poetry and all the dependencies of the Iceberg library. This also installs the development dependencies. If you don't want to install the development dependencies, you need to install using poetry install --no-dev.

If you want to install the library on the host, you can simply run pip3 install -e .. If you wish to use a virtual environment, you can run poetry shell. Poetry will open up a virtual environment with all the dependencies set.

To set up IDEA with Poetry:

  • Open up the Python project in IntelliJ
  • Make sure that you're on latest main (that includes Poetry)
  • Go to File -> Project Structure (⌘;)
  • Go to Platform Settings -> SDKs
  • Click the + sign -> Add Python SDK
  • Select Poetry Environment from the left hand side bar and hit OK
  • It can take some time to download all the dependencies based on your internet
  • Go to Project Settings -> Project
  • Select the Poetry SDK from the SDK dropdown, and click OK

For IDEA ≤2021 you need to install the Poetry integration as a plugin.

Now you're set using Poetry, and all the tests will run in Poetry, and you'll have syntax highlighting in the pyproject.toml to indicate stale dependencies.

Installation from source

Clone the repository for local development:

git clone https://github.com/apache/iceberg-python.git
cd iceberg-python
pip3 install -e ".[s3fs,hive]"

Install it directly for GitHub (not recommended), but sometimes handy:

pip install "git+https://github.com/apache/iceberg-python.git#egg=pyiceberg[pyarrow]"

Linting

pre-commit is used for autoformatting and linting:

make lint

Pre-commit will automatically fix the violations such as import orders, formatting etc. Pylint errors you need to fix yourself.

In contrast to the name suggest, it doesn't run the checks on the commit. If this is something that you like, you can set this up by running pre-commit install.

You can bump the integrations to the latest version using pre-commit autoupdate. This will check if there is a newer version of {black,mypy,isort,...} and update the yaml.

Cleaning

Removal of old cached files generated during the Cython build process:

make clean

Helps prevent build failures and unexpected behavior by removing outdated files, ensuring that only up-to-date sources are used & the build environment is always clean.

Testing

For Python, pytest is used a testing framework in combination with coverage to enforce 90%+ code coverage.

make test

By default, S3 and ADLS tests are ignored because that require minio and azurite to be running. To run the S3 suite:

make test-s3

To run the ADLS suite:

make test-adls

To pass additional arguments to pytest, you can use PYTEST_ARGS.

Run pytest in verbose mode

make test PYTEST_ARGS="-v"

Run pytest with pdb enabled

make test PYTEST_ARGS="--pdb"

To see all available pytest arguments, run make test PYTEST_ARGS="--help".

Integration tests

PyIceberg has integration tests with Apache Spark. Spark will create a new database and provision some tables that PyIceberg can query against.

make test-integration

This will restart the containers, to get to a clean state, and then run the PyTest suite. In case something changed in the Dockerfile or the provision script, you can run:

make test-integration-rebuild

To rebuild the containers from scratch.

Code standards

Below are the formalized conventions that we adhere to in the PyIceberg project. The goal of this is to have a common agreement on how to evolve the codebase, but also using it as guidelines for newcomers to the project.

API Compatibility

It is important to keep the Python public API compatible across versions. The Python official PEP-8 defines public methods as: Public attributes should have no leading underscores. This means not removing any methods without any notice, or removing or renaming any existing parameters. Adding new optional parameters is okay.

If you want to remove a method, please add a deprecation notice by annotating the function using @deprecated:

from pyiceberg.utils.deprecated import deprecated


@deprecated(
    deprecated_in="0.1.0",
    removed_in="0.2.0",
    help_message="Please use load_something_else() instead",
)
def load_something():
    pass

Which will warn:

Call to load_something, deprecated in 0.1.0, will be removed in 0.2.0. Please use load_something_else() instead.

If you want to remove a property or notify about a behavior change, please add a deprecation notice by calling the deprecation_message function:

from pyiceberg.utils.deprecated import deprecation_message

deprecation_message(
    deprecated_in="0.1.0",
    removed_in="0.2.0",
    help_message="The old_property is deprecated. Please use the something_else property instead.",
)

Which will warn:

Deprecated in 0.1.0, will be removed in 0.2.0. The old_property is deprecated. Please use the something_else property instead.

Type annotations

For the type annotation the types from the Typing package are used.

PyIceberg offers support from Python 3.9 onwards, we can't use the type hints from the standard collections.

Third party libraries

PyIceberg naturally integrates into the rich Python ecosystem, however it is important to be hesitant adding third party packages. Adding a lot of packages makes the library heavyweight, and causes incompatibilities with other projects if they use a different version of the library. Also, big libraries such as s3fs, adlfs, pyarrow, thrift should be optional to avoid downloading everything, while not being sure if is actually being used.