### Code To start, install maestral with the `dev` extra to get all dependencies required for development: ``` pip3 install maestral[dev] ``` This will install packages to check and enforce the code style, use pre-commit hooks and bump the current version. Code is formatted with [black](https://github.com/psf/black). Coding style is checked with [flake8](http://flake8.pycqa.org). Type hints, [PEP484](https://www.python.org/dev/peps/pep-0484/), are checked with [mypy](http://mypy-lang.org/). You can check the format, coding style, and type hints at the same time by running the provided pre-commit hook from the git directory: ```bash pre-commit run -a ``` You can also install the provided pre-commit hook to run checks on every commit. This will however significantly slow down commits. An introduction to pre-commit commit hooks is given at [https://pre-commit.com](https://pre-commit.com). ### Documentation The documentation is built using [sphinx](https://www.sphinx-doc.org/en/master/) and a few of its extensions. It is built from the develop and master branches and hosted on [Read The Docs](https://maestral.readthedocs.io/en/latest/). If you want to build the documentation locally, install maestral with the `docs` extra to get all required dependencies: ``` pip3 install maestral[docs] ``` The API documentation is mostly based on doc strings. Inline comments should be used whenever code may be difficult to understand for others. ### Tests The test suite uses a mixture of [unittest](https://docs.python.org/3.8/library/unittest.html) and [pytest](https://pytest-cov.readthedocs.io/en/latest/), depending on what is most convenient for the actual test and the preference of the author. Pytest should be used as the test runner. Test are grouped into those which require a linked Dropbox account ("linked") and those who can run by themselves ("offline"). The former tend to be integration test while the latter are mostly unit tests. The current focus currently lies on integration tests, especially for the sync engine, as they are easier to maintain when the implementation and internal APIs change. Exceptions are made for performance tests, for instance for indexing and cleaning up sync events, and for particularly complex functions that are prone to regressions. The current test suite uses a Dropbox access token provided by the environment variable `DROPBOX_ACCESS_TOKEN` or a refresh token provided by `DROPBOX_REFRESH_TOKEN` to connect to a real account. The GitHub action which is running the tests will set the `DROPBOX_ACCESS_TOKEN` environment variable for you with a temporary access token that expires after 4 hours. Tests are run on `ubuntu-latest` and `macos-latest` in parallel on different accounts. When using the GitHub test runner, you should acquire a "lock" on the account before running tests to prevent them from interfering which each other by creating a folder `test.lock` in the root of the Dropbox folder. This folder should have a `client_modified` time set in the future, to the expiry time of the lock. Fixtures to create and clean up a test config and to acquire a lock are provided in the `tests/linked/conftest.py`. If you run the tests locally, you will need to provide a refresh or access token for your own Dropbox account. If your account is already linked with Maestral, it will have saved a long-lived "refresh token" in your system keyring. You can access it manually or through the Python API: ```Python from maestral.main import Maestral m = Maestral() print(m.client.auth.refresh_token) ``` You can then store the retrieved refresh token in the environment variable `DROPBOX_REFRESH_TOKEN` to be automatically picked up by the tests.