Skip to content


Repository files navigation


CI Coverage Status Python Versions PyPI Version

Dlint is a tool for encouraging best coding practices and helping ensure Python code is secure.

The most important thing I have done as a programmer in recent years is to aggressively pursue static code analysis. Even more valuable than the hundreds of serious bugs I have prevented with it is the change in mindset about the way I view software reliability and code quality.

For a static analysis project to succeed, developers must feel they benefit from and enjoy using it.

For documentation and a list of rules see docs.


$ python -m pip install dlint

And double check that it was installed correctly:

$ python -m flake8 -h
Usage: flake8 [options] file file ...


Installed plugins: dlint: 0.14.1, mccabe: 0.5.3, pycodestyle: 2.2.0, pyflakes: 1.3.0

Note the dlint: 0.14.1.


Dlint builds on flake8 to perform its linting. This provides many useful features without re-inventing the wheel.


Let's run a simple check:

$ cat << EOF >
$ python
$ python -m flake8 --select=DUO DUO105 use of "exec" is insecure

The --select=DUO flag tells flake8 to only run Dlint lint rules.

From here, we can easily run Dlint against a directory of Python code:

$ python -m flake8 --select=DUO /path/to/code

To fine-tune your linting, check out the flake8 help:

$ python -m flake8 --help

Inline Editor

Dlint results can also be included inline in your editor for fast feedback. This typically requires an editor plugin or extension. Here are some starting points for common editors:


Dlint can easily be integrated into CI pipelines, or anything really.

For more information and examples see 'How can I integrate Dlint into XYZ?'.

Custom Plugins

Dlint's custom plugins are built on a simple naming convention, and rely on Python modules. To make a Dlint custom plugin use the following conventions:

  • The Python module name must start with dlint_plugin_.
  • The linter class name must start with Dlint.
  • The linter class should inherit from dlint.linters.base.BaseLinter.
    • If for some reason you'd like to avoid this, then you must implement the get_results function appropriately and inherit from ast.NodeVisitor.

See an example plugin for further details.


First, install development packages:

$ python -m pip install -r requirements.txt
$ python -m pip install -r requirements-dev.txt
$ python -m pip install -e .


$ pytest


$ flake8


$ pytest --cov


$ pytest -k test_benchmark_run --benchmark-py-file /path/to/ tests/test_benchmark/

Or get benchmark results for linters individually:

$ pytest -k test_benchmark_individual --benchmark-py-file /path/to/ tests/test_benchmark/

Or run against a single linter:

$ pytest -k test_benchmark_individual[DUO138-BadReCatastrophicUseLinter] --benchmark-py-file /path/to/ tests/test_benchmark/