Skip to content

Latest commit

 

History

History
191 lines (130 loc) · 6.95 KB

CONTRIBUTING.md

File metadata and controls

191 lines (130 loc) · 6.95 KB

Contributing to geopy

Reporting issues

If you caught an exception from geopy please try to Google the error first. There is a great chance that it has already been discussed somewhere and solutions have been provided (usually on GitHub or on Stack Overflow).

Before reporting an issue please ensure that you have tried to get the answer from the doc: https://geopy.readthedocs.io/.

Keep in mind that if a specific geocoding service's API is not behaving correctly then it probably won't be helpful to report that issue here, see https://geopy.readthedocs.io/en/latest/#geopy-is-not-a-service

The following resources are available for your input:

  1. Stack Overflow with geopy tag. There's a somewhat active community here so you will probably get a solution quicker. And also there is a large amount of already resolved questions which can help too! Just remember to put the geopy tag if you'd decide to open a question.
  2. GitHub Discussions is a good place to start if Stack Overflow didn't help or you have some idea or a feature request you'd like to bring up, or if you just have trouble and not sure you're doing everything right. Solutions and helpful snippets/patterns are also very welcome here.
  3. GitHub Issues should only be used for definite bug reports and specific tasks. If you're not sure whether your issue fits this then please start with Discussions first.

Submitting patches

If you contribute code to geopy, you agree to license your code under the MIT.

The new code should follow PEP8 coding style (except the line length limit, which is 90) and adhere to the style of the surrounding code.

You must document any functionality using Sphinx-compatible RST, and implement tests for any functionality in the test directory.

In your Pull Requests there's no need to fill the changelog or AUTHORS, this is a maintainer's responsibility.

For your convenience the contributing-friendly issues are marked with help wanted label, and the beginner-friendly ones with good first issue.

If your PR remains unreviewed for a while, feel free to bug the maintainer.

Setup

  1. Create a virtualenv

  2. Install geopy in editable mode along with dev dependencies:

    pip install -e ".[dev]"
    
  3. Ensure that tests pass

    make test
    

Running tests

To quickly run the test suite without Internet access:

make test-local

To run the full test suite (makes queries to the geocoding services):

make test

Or simply:

pytest

To run a specific test module, pass a path as an argument to pytest. For example:

pytest test/geocoders/nominatim.py

Before pushing your code, make sure that linting passes, otherwise a CI build would fail:

make lint

Geocoder credentials

Some geocoders require credentials (like API keys) for testing. They must remain secret, so if you need to test a code which requires them, you should obtain your own valid credentials.

Tests in CI from forks and PRs run in test-local mode only, i.e. no network requests are performed. Full test suite with network requests is run only for pushes to branches by maintainers. This helps to reduce load on the geocoding services and save some quotas associated with the credentials used by geopy. It means that PR builds won't actually test network requests. Code changing a geocoder should be tested locally. But it's acceptable to not test such code if obtaining the required credentials seems problematic: just leave a note so the maintainers would be aware that the code hasn't been tested.

You may wonder: why not commit captured data and run mocked tests? Because there are copyright constraints on the data returned by services. Another reason is that this data goes obsolete quite fast, and maintaining it is cumbersome.

To ease local testing the credentials can be stored in a json format in a file called .test_keys located at the root of the project instead of env variables.

Example contents of .test_keys:

{
    "BING_KEY": "...secret key...",
    "OPENCAGE_KEY": "...secret key..."
}

Building docs

make docs

Open docs/_build/html/index.html with a browser to see the docs. On macOS you can use the following command for that:

open docs/_build/html/index.html

Adding a new geocoder

Patches for adding new geocoding services are very welcome! It doesn't matter how popular the target service is or whether its territorial coverage is global or local.

A checklist for adding a new geocoder:

  1. Create a new geocoding class in its own Python module in the geopy/geocoders package. Please look around to make sure that you're not reimplementing something that's already there! For example, if you're adding a Nominatim-based service, then your new geocoder class should probably extend the geopy.geocoders.Nominatim class.

  2. Follow the instructions in the geopy/geocoders/__init__.py module for adding the required imports.

  3. Create a test module in the test/geocoders directory. If your geocoder class requires credentials, make sure to access them via the test.geocoders.util.env object (see test/geocoders/what3words.py for example). Refer to the Geocoder credentials section above for info on how to work with credentials locally.

  4. Complete your implementation and tests! Give TDD a try if you aren't used to it yet! 🎉 Please note that it's possible to run a single test module instead of running the full suite, which is much slower. Refer to the Running tests section above for a command example.

  5. Make sure to document the geocode and reverse methods with all their parameters. The class doc should contain a URI to the Terms of Service.

  6. Add a reference to that class in the docs/index.rst file. Please keep them ordered alphabetically by their module file names. Build the docs (Building docs section above) to make sure that you've done them right!

  7. If your tests use credentials, add their names to the end of the .github/workflows/ci.yml file.

That's all!

Improving a geocoder

If you want to add additional parameters to a geocode or reverse method, the additional parameters must be explicitly specified and documented in the method signature. Validation may be done for type, but values should probably not be checked against an enumerated list because the service could change. The additional parameters should go to the end of the method signature.

Please avoid simply passing through arbitrary parameters (e.g. with **kwargs) to the params of a request to the target service. The target service parameters might change, as well as the service's API, but the geocoder class's public API should stay the same. It's almost impossible to achieve that when a pass-through of arbitrary parameters is allowed.