Version conflicts between dependencies has been a big issue for GCP Python users. The issue typically happens when a user depends on two libraries A and B, both of which depend on incompatible versions of library C. This can lead to non-deterministic behavior, since only one version of C actually gets loaded into the library.
This repository is providing a toolkit for GCP Python open source projects to bootstrap their development infrastructure, enforcing centralized dependency management, CI, and release process, thus ensuring compatibility across all of our GCP Python open source libraries for our end-users.
Compatibility Lib is a library to get compatibility status and dependency information of Python packages. It contains three tools: compatibility checker, outdated dependency highlighter and deprecated dependency finder. And it also provides utilities to query data from the BigQuery tables (external user will need to set up tables with the same schema that this library is using).
Installation:
pip install compatibility-lib
Compatibility checker gets the compatibility data by sending requests to the Compatibility Server endpoint, or by querying the BigQuery table (if the given package is listed in our configs, which are pre-computed).
Usage like below,
import itertools
from compatibility_lib import compatibility_checker
packages = ['package1', 'package2', 'package3']
package_pairs = itertools.combinations(packages, 2)
checker = compatibility_checker.CompatibilityChecker()
# Get self compatibility data
checker.get_self_compatibility(python_version='3', packages=packages)
# Get pairwise compatibility data
checker.get_pairwise_compatibility(
python_version='3', pkg_sets=package_pairs)
Outdated Dependency Highlighter finds out the outdated dependencies of a Python package, and determines the priority of updating the dependency version based on a set of criteria below:
- Mark “High Priority” if dependencies have widely adopted major release. (e.g 1.0.0 -> 2.0.0)
- Mark “High Priority” if a new version has been available for more than 6 months.
- Mark “High Priority” if dependencies are 3 or more sub-versions behind the newest one. (e.g 1.0.0 -> 1.3.0)
- Mark “Low Priority” for other dependency updates.
Usage:
from compatibility_lib import dependency_highlighter
packages = ['package1', 'package2', 'package3']
highlighter = dependency_highlighter.DependencyHighlighter()
highlighter.check_packages(packages)
Deprecated Dependency Finder can find out the deprecated dependencies that a Python package depends on.
Usage:
from compatibility_lib import deprecated_dep_finder
packages = ['package1', 'package2', 'package3']
finder = deprecated_dep_finder.DeprecatedDepFinder()
for res in finder.get_deprecated_deps(packages):
print(res)
Displaying the compatibility status for your package as a Github Badge.
- Self Compatibility
- Compatibility with Google OSS Python packages
- Dependency version status
See the usage here.
- Set Up Python Environment
https://cloud.google.com/python/setup
- Install py 3.6 (may not be included in previous step)
sudo apt install python3.6
- Clone the cloud-opensource-python project and cd to project
git clone [email protected]:GoogleCloudPlatform/cloud-opensource-python.git
cd cloud-opensource-python
- Fork project and configure git remote settings
git remote add upstream [email protected]:GoogleCloudPlatform/cloud-opensource-python.git
git config --global user.email "[email protected]"
- Create a virtualenv, and source
tox -e py36
source .tox/py36/bin/activate
- Install gcloud SDK and initialize
curl https://sdk.cloud.google.com | bash
gcloud init
- Create new service account key
- In your browser, navigate to Cloud Console
- menu > IAM & admin > Service accounts
- under bigquery-admin, actions > create new key
- Set GOOGLE_APPLICATION_CREDENTIALS
export GOOGLE_APPLICATION_CREDENTIALS=”path/to/service/key.json”
- Build compatibility_lib library from source and install
python compatibility_lib/setup.py bdist_wheel
pip install compatibility_lib/dist/*
We use nox test suite for running tests.
- Install Nox for testing
pip install nox-automation
- Run the tests
nox -s unit # unit tests
nox -s lint # linter
nox -s system # system tests
nox -l # see available options
nox # run everything
Apache 2.0 - See LICENSE for more information.
This is not an official Google product.