drytoml’s Docs

Keep your toml configuration centralized and personalizable.

PyPI PyPI - Python Version
Format Lint Test codecov
VSCode Ready-to-Code License: Unlicense

Through drytoml, TOML files can have references to any content from another TOML file. References work with relative/absolute paths and urls, and can point to whole files, a specific table, or in general anything reachable by a sequence of getitem calls, like ["tool", "poetry", "source", 0, "url"]. drytoml takes care of transcluding the content for you.

Inspired by flakehell and nitpick, the main motivation behind drytoml is to have several projects sharing a common, centralized configuration defining codestyles, but still allowing granular control when required.

This is a deliberate departure from nitpick, which works as a linter instead, ensuring your local files have the right content.


drytoml has two main usages:

  1. Use a file as a reference and “build” the resulting one:

    # contents of pyproject.dry.toml
    __extends = "../../common/black.toml"
    target-version = ['py37']
    include = '\.pyi?$'
    # contents of ../../common/black.toml
    line-length = 100
    $ dry export --file=pyproject.dry.toml > pyproject.toml
    # contents of pyproject.toml
    line-length = 100
    target-version = ['py37']
    include = '\.pyi?$'
  2. Use included wrappers, allowing you to use references in your current configuration without changing any file:

    Consider the original pyproject.dry.toml from the example above, an alternative usage for drytoml is shown next. Instead of this:

    $ black .
    All done! ✨ 🍰 ✨
    14 files left unchanged.

    You would run this:

    $ dry black
    reformatted /path/to/cwd/file_with_potentially_long_lines.py
    reformatted /path/to/cwd/another_file_with_potentially_long_lines.py
    All done! ✨ 🍰 ✨
    2 files reformatted, 12 files left unchanged.

    What just happened? drytoml comes with a set of wrappers which

    1. Create a transcluded temporary file, equivalent to the resulting pyproject.toml in the example above

    2. Configure the wrapped tool (black in this case) to use the temporary file

    3. Run black, and get rid of the file on exit.

For the moment, the following wrappers are available (more to come, contributions are welcome):

In the works:

  • [ ] docformatter

  • [ ] pytest


Although the flakehell project was archived, we’re keeping a fork alive from here, availabe as flakeheaven in pypi.



  • A compatible python >3.6.9

  • [recommended] virtualenv

  • A recent pip


Install as usual, with pip, poetry, etc:

  • pip install drytoml

  • poetry add drytoml (probably you’ll want poetry add --dev drytoml instead)


For any command , run --help to find out flags and usage. Some of the most common are listed below:

  • Use any of the provided wrappers as a subcommand, eg dry black instead of black.

  • Use dry -q export and redirect to a file, to generate a new file with transcluded contents

  • Use dry cache to manage the cache for remote references.


Q: I want to use a different key to trigger transclusions

A: In cli mode (using the dry command), you can pass the --key flagcli, to change it. In api mode (from python code), initialize drytoml.parser.Parser using a custom value for the extend_key kwarg.

Q: I changed a referenced toml upstream (eg in github) but still get the same result.

A: Run dry cache clear --help to see available options.


Start by creating an issue, forking the project and creating a Pull Request.

Setting up the development environment

If you have docker, the easiest way is to use the provided devcontainer inside vscode, which already contains everything pre-installed. You must open the cloned directory using the remote-containers extension. Just run poetry shell or prepend any command with poetry run to ensure commands are run inside the virtual environment.

Alternatively, you can use poetry: poetry install -E dev

The next steps assume you have already activated the venv.


  • Commits in every branch must adhere to Conventional Commits. Releases are done automatically and require conventional commits messages compliance.

  • To validate commits, you can install the pre-commit hook

    pre-commit install --hook-type commit-msg
  • With venv activated, you can commit using cz commit instead of git commit to ensure compliance.

Running checks

You can look at the different actions defined in .github/workflows. There are three ways to check your code:

  • Remotely, by pushing to an open PR. You’ll se the results in the PR page.

  • Manually, executing the check from inside a venv

    For example, to generate the documentation, check .github/workflows/docs. To run the Build with Sphinx step:

    sphinx-build docs/src docs/build

    Or to run pytest, from .github/workflows/tests.yml:

    sphinx-build docs/src docs/build

    … etc

  • Locally, with act (Already installed in the devcontainer)

For example, to emulate a PR run for the tests workflow:

act -W .github/workflows/tests.yml pull_request


Check out current development here

Indices and tables