Contributing
Thanks for your interest in contributing to web3.py! Read on to learn what would be helpful and how to go about it. If you get stuck along the way, reach for help in the Python Discord server.
How to Help
Without code:
Answer user questions within GitHub issues, Stack Overflow, or the Python Discord server.
Write or record tutorial content.
Improve our documentation (including typo fixes).
Open an issue on GitHub to document a bug. Include as much detail as possible, e.g., how to reproduce the issue and any exception messages.
With code:
Fix a bug that has been reported in an issue.
Add a feature that has been documented in an issue.
Add a missing test case.
Warning
Before you start: always ask if a change would be desirable or let us know that you plan to work on something! We don’t want to waste your time on changes we can’t accept or duplicated effort.
Your Development Environment
Note
Use of a virtual environment is strongly advised for minimizing dependency issues. See this article for usage patterns.
All pull requests are made from a fork of the repository; use the GitHub UI to create a fork.
web3.py depends on submodules, so when you clone
your fork to your local machine, include the --recursive flag:
$ git clone --recursive https://github.com/<your-github-username>/web3.py.git
$ cd web3.py
Finally, install all development dependencies. The dev dependency group is
installed by default, so uv sync is enough for a local development environment:
$ uv sync
$ uv run prek install
Using Docker
Developing within Docker is not required, but if you prefer that workflow, use the sandbox container provided in the docker-compose.yml file.
To start up the test environment, run:
$ docker compose up -d
This will build a Docker container set up with an environment to run the Python test code.
To run the core tests from your local machine:
$ docker compose exec sandbox bash -c 'pytest tests/core'
The container does not have go-ethereum installed, so you can exclude those tests
by using the -k "not goethereum" flag.
$ docker compose exec sandbox bash -c 'pytest tests/integration -k "not goethereum"'
You can run arbitrary commands inside the Docker container by using the
bash -c prefix.
$ docker compose exec sandbox bash -c 'pwd && ls'
Or, if you would like to open a session to the container, run:
$ docker compose exec sandbox bash
Running The Tests
A great way to explore the code base is to run the tests.
First, install the test dependencies:
$ uv sync --group test
You can run all tests with:
$ pytest
However, running the entire test suite takes a very long time and is generally impractical. Typically, you’ll just want to run a subset instead, like:
$ pytest tests/core/eth-module/test_accounts.py
Linting is also performed by the CI and locally with each commit. You can save yourself some time by checking for linting errors manually:
$ make lint
It is important to understand that each pull request must pass the full test suite as part of the CI check. This test suite will run in the CI anytime a pull request is opened or updated.
Writing Tests
We strongly encourage contributors to write good tests for their code as part of the code review process. This helps ensure that your code is doing what it should be doing.
We strongly encourage you to use our existing tests for both guidance and
homogeneity / consistency across our tests. We use pytest for our tests.
For more specific pytest guidance, please refer to the pytest documentation.
Within the pytest scope, conftest.py files are used for common code
shared between modules that exist within the same directory as that particular
conftest.py file.
Unit Testing and eth-tester Tests
Our unit tests are grouped together with tests against the eth-tester library,
using the py-evm library as a backend, via the EthereumTesterProvider.
These tests live under appropriately named child directories within the
/tests directory. The core of these tests live under /tests/core.
Do your best to follow the existing structure when adding a test and make sure
that its location makes sense.
Integration Testing
Our integration test suite setup lives under the /tests/integration directory.
The integration test suite is dependent on what we call “fixtures” (not to be
confused with pytest fixtures). These zip file fixtures, which also live in the
/tests/integration directory, are configured to run the specific client we are
testing against along with a genesis configuration that gives our tests some
pre-determined useful objects (like unlocked, pre-loaded accounts) to be able to
interact with the client when we run our tests.
The parent /integration directory houses some common configuration shared across
all client tests, whereas the /go_ethereum directory houses common code to be
shared across geth-specific provider tests. Though the setup and run configurations
exist across the different files within /tests/integration, our integration module
tests are written across different files within /web3/_utils/module_testing.
common.pyfiles within the client directories contain code that is shared across all provider tests (http, ipc, and ws). This is mostly used to override tests that span across all providers.conftest.pyfiles within each of these directories contain mostly code that can be used by all test files that exist within the same directory or subdirectories of theconftest.pyfile. This is mostly used to house pytest fixtures to be shared among our tests. Refer to the pytest documentation on fixtures for more information.test_{client}_{provider}.pyfiles (e.g.test_goethereum_http.py) are where client-and-provider-specific test configurations exist. This is mostly used to override tests specific to the provider type for the respective client.
The integration tests are each run in insolation to prevent muddied contexts. Because
they are run in isolation, they can be parallelized with pytest-xdist in order to
speed up the test suite. To run the tests in parallel, you can use the -n flag
with pytest. For example, to run the tests in parallel with 4 workers, you can
use the following command:
$ pytest tests/integration/go_ethereum/path/to/module/or/test -n 4
Working With Test Contracts
Contracts used for testing exist under web3/_utils/contract_sources. These contracts
get compiled via the scripts/compile_contracts.py maintenance tool. To use
this script, simply pass the Solidity version to be used to compile the contracts as an
argument at the command line.
- Arguments for the script are:
- -v or –version Solidity version to be used to compile the contracts. If
blank, the script uses the latest available version from solcx.
- -f or –filename If left blank, all .sol files will be compiled and the
respective contract data will be generated. Pass in a specific
.solfilename here to compile just one file.
The maintenance dependency group provides py-solc-x and Ruff:
$ uv sync --group maintenance
The following example compiles all the contracts and generates their respective
contract data that is used across our test files for the test suites. This data gets
generated within the contract_data subdirectory within the contract_sources
folder.
$ cd ../web3.py/web3/_utils/contract_sources
$ uv run --project ../../.. --group maintenance python ../../../scripts/compile_contracts.py -v 0.8.17
Compiling OffchainLookup
...
...
reformatted ...
To compile and generate contract data for only one .sol file, specify using the
filename with the -f (or --filename) argument flag.
$ cd ../web3.py/web3/_utils/contract_sources
$ uv run --project ../../.. --group maintenance python ../../../scripts/compile_contracts.py -v 0.8.17 -f OffchainLookup.sol
Compiling OffchainLookup.sol
reformatted ...
If there is any contract data that is not generated via the script but is important
to pass on to the integration tests, the _custom_contract_data.py file within the
contract_data subdirectory can be used to store that information when appropriate.
Be sure to re-generate the integration test fixture after running the script to update the contract bytecodes for the integration test suite - see the Generating New Fixtures section below.
Manual Testing
To import and test an unreleased version of web3.py in another context, you can install it from your development directory:
$ python -m pip install -e ../path/to/web3py
Code Style
We use prek to enforce a consistent code style across the library. This tool runs automatically with every commit, but you can also run it manually with:
$ make lint
If you need to make a commit that skips the prek checks, you can do so with
git commit --no-verify.
We use Black as part of our linting. To ignore the commits that introduced Black in git history, you can configure your git environment like so:
$ git config blame.ignoreRevsFile .git-blame-ignore-revs
This library uses type hints, which are enforced by the mypy tool (part of the
prek checks). All new code is required to land with type hints, with the
exception of code within the tests directory.
Documentation
Good documentation will lead to quicker adoption and happier users. Please check out our guide on how to create documentation for the Python Ethereum ecosystem.
Pull requests generate their own preview of the latest documentation at
https://web3py--<pr-number>.org.readthedocs.build/en/<pr-number>/.
Pull Requests
It’s a good idea to make pull requests early on. A pull request represents the start of a discussion, and doesn’t necessarily need to be the final, finished submission.
See GitHub’s documentation for working on pull requests.
Once you’ve made a pull request, check the GitHub Actions status in the GitHub interface and make sure all tests are passing. In general, pull requests that do not pass the CI build yet won’t get reviewed unless explicitly requested.
Use the conventional-commit format for the pull request title. GitHub generates release notes from merged pull requests.
Generating New Fixtures
Our integration tests make use of Geth private networks. When new versions of the client software are introduced, new fixtures should be generated.
Before generating new fixtures, make sure you have the test dependencies installed:
$ uv sync --group test
Note
A “fixture” is a pre-synced network. It’s the result of configuring and running a client, deploying the test contracts, and saving the resulting state for testing web3.py functionality against.
Geth Fixtures
Install the desired Geth version on your machine locally. We recommend py-geth for this purpose, because it enables you to easily manage multiple versions of Geth.
Note that
py-gethwill need updating to support each new Geth version as well. Adding newer Geth versions to py-geth is straightforward; see past commits for a template.If py-geth has the Geth version you need, install that version locally. For example:
$ uv run --group integration python -m geth.install v1.16.7
Specify the Geth binary and run the fixture creation script (from within the web3.py directory):
$ GETH_BINARY=~/.py-geth/geth-v1.16.7/bin/geth uv run --group integration python -m scripts.generate_fixtures.go_ethereum
The output of this script is your fixture, a zip file, which is now stored in
/tests/integration/. The/tests/integration/go_ethereum/conftest.pyfile should be updated automatically to point to this new fixture. Delete the old fixture.Run the tests. To ensure that the tests run with the correct Geth version locally, you may again include the
GETH_BINARYenvironment variable.The
GETH_VERSIONvalues in/.github/workflows/test.yamlshould be automatically updated to match thego-ethereumversion used to generate the test fixture.
CI Testing With a Nightly Geth Build
Occasionally you’ll want to have CI run the test suite against an unreleased version of Geth - e.g. to test upcoming hard fork changes. The workflow described below is for testing only, as updates will only be merged into main once the Geth release is published and the test runs are updated to use the new stable version.
Configure
scripts/generate_fixtures/common.pyas needed.Geth automagically compiles new builds for every commit that gets merged into the codebase. Download the desired build from the develop builds.
Build your test fixture, passing in the binary you just downloaded via
GETH_BINARY. Don’t forget to update the/tests/integration/go_ethereum/conftest.pyfile to point to your new fixture.Our CI runs on Ubuntu, so download the corresponding 64-bit Linux develop build, then add it to the root of your web3.py directory. Rename the binary
custom_geth.In the GitHub Actions geth integration jobs, update the
GETH_VERSIONenvironment value if the fixture generation requires a non-default geth binary.Create a PR and let CI do its thing.
Releasing
Releases are typically done from the main branch, except when releasing a beta (in
which case the beta is released from main, and the previous stable branch is
released from said branch).
Final test before each release
Before releasing a new version, run the test suite from the release branch:
$ git checkout main && git pull
$ uv run --group test pytest tests/core
Review the documentation that will get published:
$ uv run --group docs sphinx-apidoc -o docs/ . "*conftest*" "tests" "scripts" "web3/tools/*"
$ uv run --group docs sphinx-build -W -b html docs docs/_build/html
$ uv run --group docs sphinx-build -W -b doctest docs docs/_build/doctest
$ uv run --group docs sphinx-build -W -b epub docs docs/_build/epub
Push the release to github & pypi
After confirming that the release package looks okay, create a GitHub Release with the version tag to publish. GitHub Actions builds the package from the release tag and publishes to PyPI with trusted publishing. Release notes should be written in the GitHub Release.
setuptools-scm derives the package version from the release tag. Use PEP 440 tags,
for example v8.0.0b3 or v8.0.0.