Darglint2

A functional docstring linter which checks whether a docstring’s description matches the actual function/method implementation. Darglint2 expects docstrings to be formatted using the Google Python Style Guide, or Sphinx Style Guide, or Numpy Style Guide.

Feel free to submit an issue/pull request if you spot a problem or would like a feature in darglint2.

Table of Contents:

Project Status

This is a renamed fork of the original darglint by @terrencepreilly, and I’m maintaining it and accepting bugfixes.

Important: darglint2 is very slow compared to alternatives due to its parser design. darglint2 is only intended as a low-effort replacement for darglint for those old darglint users who need bugfixes and critical features but don’t want to spend the trouble of migrating to a better alternative. See e.g. this discussion for details.

For the 1.8.1 version of darglint, @terrencepreilly added a Project Status section in the README in October 2021:

I no longer work with Python regularly, and I’d like to spend some more time on other projects. So while I’ll continue to maintain darglint, I likely won’t be adding significant new features. That said, I will try to accept pull requests. See the contribution section for more information. Consider it in maintenance mode.

Later, in December 2022, the darglint repository was archived and maintenance was ceased. With a few useful bugfixes and pull requests left out of the final 1.8.1 release, @akaihola decided to adopt the package for maintenance, but copied and renamed it to darglint2 at the request of the original author.

For more background about the fork, see also this discussion attached to commit 0c8a3887 in @9dogs’s fork of the original terrencepreilly/darglint.

Installation

To install darglint2, use pip.

pip install darglint2

Or, clone the repository, cd to the directory, and

pip install .

Configuration

darglint2 can be configured using a configuration file. The configuration file must be named either .darglint2, .darglint, setup.cfg, or tox.ini. It must also have a section starting with the section header, [darglint2]. Finally, the configuration file must be located either in the directory darglint2 is called from, or from a parent directory of that working directory.

Currently, the configuration file allows us to ignore errors, to specify message templates, to specify the strictness of checks and to ignore common exceptions.

Error Configuration

If we would like to ignore ExcessRaiseErrors (because we know that an underlying function will raise an exception), then we would add its error code to a file named .darglint2:

[darglint2]
ignore=DAR402

We can ignore multiple errors by using a comma-separated list:

[darglint2]
ignore=DAR402,DAR103

Instead of specifying error codes to ignore in general one can also specify a regex to exclude certain function names from tests. For example, the following configuration would disable linting on all private methods.

[darglint2]
ignore_regex=^_(.*)

Message Template Configuration

If we would like to specify a message template, we may do so as follows:

[darglint2]
message_template={msg_id}@{path}:{line}

Which will produce a message such as DAR102@driver.py:72.

Finally, we can specify the docstring style type using docstring_style (“google” by default):

[darglint2]
docstring_style=sphinx

Strictness Configuration

Strictness determines how lax darglint2 will be when checking docstrings. There are three levels of strictness available:

  • short: One-line descriptions are acceptable; anything more and the docstring will be fully checked.

  • long: One-line descriptions and descriptions without arguments/returns/yields/etc. sections will be allowed. Anything more, and the docstring will be fully checked.

  • full: (Default) Docstrings will be fully checked.

For example, if we have the following function:

def double(x):
    # <docstring>
    return x * 2

Then the following table describes which errors will be raised for each of the docstrings (rows) when checked against each of the configurations (columns):

┌──────────────────────────────┬──────────────────┬────────────────┬──────────────────┐
│ Docstring                    │  short           │  long          │  full            │
├──────────────────────────────┼──────────────────┼────────────────┼──────────────────┤
│ """Doubles the argument."""  │ None             │ None           │ Missing argument │
│                              │                  │                │ Missing return   │
│                              │                  │                │                  │
│                              │                  │                │                  │
├──────────────────────────────┼──────────────────┼────────────────┼──────────────────┤
│ """Doubles the argument.     │ Missing argument │ None           │ Missing argument │
│                              │ Missing return   │                │ Missing return   │
│ Not very pythonic.           │                  │                │                  │
│                              │                  │                │                  │
│ """                          │                  │                │                  │
│                              │                  │                │                  │
├──────────────────────────────┼──────────────────┼────────────────┼──────────────────┤
│ """Doubles the argument.     │ Missing return   │ Missing return │ Missing return   │
│                              │                  │                │                  │
│ Args:                        │                  │                │                  │
│     x: The number to double. │                  │                │                  │
│                              │                  │                │                  │
│ """                          │                  │                │                  │
└──────────────────────────────┴──────────────────┴────────────────┴──────────────────┘

In short, if you want to be able to have single-line docstrings, and check all other docstrings against their described parameters, you would specify

[darglint2]
strictness=short

In your configuration file.

Ignoring common exceptions

We can specify a list of exceptions that don’t need to be documented in the raises section of a docstring. For example,

[darglint2]
ignore_raise=ValueError,MyCustomError

Logging

When darglint2 fails unexpectedly, you can try to gather more information when submitting a bug by running with logging. For example,

darglint2 --log-level=INFO unexpected_failures.py

Darglint2 accepts the levels, DEBUG, INFO, WARNING, ERROR, and CRITICAL.

Usage

Command Line use

Given a python source file, serializers.py, you would check the docstrings as follows:

darglint2 serializers.py

You can give an optional verbosity setting to darglint2. For example,

darglint2 -v 2 *.py

Would give a description of the error along with information as to this specific instance. The default verbosity is 1, which gives the filename, function name, line number, error code, and some general hints.

To use an arbitrary error format, you can pass a message template, which is a python format string. For example, if we pass the message template

darglint2 -m "{path}:{line} -> {msg_id}" darglint2/driver.py

Then we would get back error messages like

darglint2/driver.py :61 -> DAR101

The following attributes can be passed to the format string:

  • line: The line number,

  • msg: The error message,

  • msg_id: The error code,

  • obj: The function/method name,

  • path: The relative file path.

The message template can also be specified in the configuration file as the value message_template.

darglint2 is particularly useful when combined with the utility, find. This allows us to check all of the files in our project at once. For example, when eating my own dogfood (as I tend to do), I invoke darglint2 as follows:

find . -name "*.py" | xargs darglint2

Where I’m searching all files ending in “.py” recursively from the current directory, and calling darglint2 on each one in turn.

Ignoring Errors in a Docstring

You can ignore specific errors in a particular docstring. The syntax is much like that of pycodestyle, etc. It generally takes the from of:

# noqa: <error> <argument>

Where <error> is the particular error to ignore (DAR402, or DAR201 for example), and <argument> is what (if anything) the ignore statement refers to (if nothing, then it is not specified).

Let us say that we want to ignore a missing return statement in the following docstring:

def we_dont_want_a_returns_section():
  """Return the value, 3.

  # noqa: DAR201

  """
  return 3

We put the noqa anywhere in the top level of the docstring. However, this won’t work if we are missing something more specific, like a parameter. We may not want to ignore all missing parameters, either, just one particular one. For example, we may be writing a function that takes a class instance as self. (Say, in a bound celery task.) Then we would do something like:

def a_bound_function(self, arg1):
  """Do something interesting.

  Args:
    arg1: The first argument.

  # noqa: DAR101 arg1

  """
  arg1.execute(self)

So, the argument comes to the right of the error.

We may also want to mark excess documentation as being okay. For example, we may not want to explicitly catch and raise a ZeroDivisionError. We could do the following:

def always_raises_exception(x):
    """Raise a zero division error or type error.o

    Args:
      x: The argument which could be a number or could not be.

    Raises:
      ZeroDivisionError: If x is a number.  # noqa: DAR402
      TypeError: If x is not a number.  # noqa: DAR402

    """
    x / 0

So, in this case, the argument for noqa is really all the way to the left. (Or whatever description we are parsing.) We could also have put it on its own line, as # noqa: DAR402 ZeroDivisionError.

Type Annotations

Darglint2 parses type annotations in docstrings, and can, optionally, compare the documented type to the actual type annotation. This can be useful when migrating a codebase to use type annotations.

In order to make these comparisons, Darglint2 only accepts types accepted by Python (see PEP 484.) That is, it does not accept parentheses in type signatures. (If parentheses are used in the type signature, Darglint2 will mark that argument as missing. See issue darglint#90.)

Error Codes

  • DAR001: The docstring was not parsed correctly due to a syntax error.

  • DAR002: An argument/exception lacks a description

  • DAR003: A line is under-indented or over-indented.

  • DAR004: The docstring contains an extra newline where it shouldn’t.

  • DAR005: The item contains a type section (parentheses), but no type.

  • DAR101: The docstring is missing a parameter in the definition.

  • DAR102: The docstring contains a parameter not in function.

  • DAR103: The docstring parameter type doesn’t match function.

  • DAR104: (disabled) The docstring parameter has no type specified

  • DAR105: The docstring parameter type is malformed.

  • DAR201: The docstring is missing a return from definition.

  • DAR202: The docstring has a return not in definition.

  • DAR203: The docstring parameter type doesn’t match function.

  • DAR301: The docstring is missing a yield present in definition.

  • DAR302: The docstring has a yield not in definition.

  • DAR401: The docstring is missing an exception raised.

  • DAR402: The docstring describes an exception not explicitly raised.

  • DAR501: The docstring describes a variable which is not defined.

The number in the hundreds narrows the error by location in the docstring:

  • 000: Syntax, formatting, and style

  • 100: Args section

  • 200: Returns section

  • 300: Yields section

  • 400: Raises section

  • 500: Variables section

You can enable disabled-by-default exceptions in the configuration file using the enable option. It accepts a comma-separated list of error codes.

[darglint2]
enable=DAR104

Scope

Darglint2’s primary focus is to identify incorrect and missing documentationd of a function’s signature. Checking style is a stretch goal, and is supported on a best-effort basis. Darglint2 does not check stylistic preferences expressed by tools in the Python Code Quality Authority (through tools such as pydocstyle). So when using Darglint2, it may be a good idea to also use pydocstyle, if you want to enforce style. (For example, pydocstyle requires the short summary to be separated from other sections by a line break. Darglint2 makes no such check.)

Sphinx

Darglint2 can handle sphinx-style docstrings, but imposes some restrictions on top of the Sphinx style. For example, all fields (such as :returns:) must be the last items in the docstring. They must be together, and all indents should be four spaces. These restrictions may be loosened at a later date.

To analyze Sphinx-style docstrings, pass the style flag to the command:

darglint2 -s sphinx example.py
darglint2 --docstring-style sphinx example.py

Alternatively, you can specify the style in the configuration file using the setting, “docstring_style”:

[darglint2]
docstring_style=sphinx

Numpy

Darglint2 now has an initial implementation for Numpy-style docstrings. Similarly to Sphinx-style docstrings, you can pass a style flag to the command:

darglint2 -s numpy example.py
darglint2 --docstring-style numpy example.py

Or set it in a configuration file:

[darglint2]
docstring_style=numpy

The numpy parser and error reporter are not yet fully stabilized. Add issues or suggestions to the tracking bug, Issue #69.

Integrations

Flake8

Darglint2 can be used in conjunction with Flake8 as a plugin. The only setup necessary is to install Flake8 and Darglint2 in the same environment. Darglint2 will pull its configuration from Flake8. So, if you would like to lint Sphinx-style comments, then you should have docstring_style=sphinx in a Flake8 configuration file in the project directory. The settings would be entered under the flake8 configuration, not a separate configuration for Darglint2. E.g.:

[flake8]
strictness=short
docstring_style=sphinx

To see which options are exposed through Flake8, you can check the Flake8 tool:

flake8 --help | grep --before-context=2 Darglint2

SublimeLinter

A plugin for SublimeLinter can be found here. Note that it was built for the original darglint, not darglint2, and its compatibility with darglint2 is yet to be verified.

Pre-commit

Download pre-commit and install it. Once it is installed, add this to .pre-commit-config.yaml in your repository:

repos:
    - repo: https://github.com/akaihola/darglint2
      rev: master
      hooks:
          - id: darglint2

Then run pre-commit install and you’re ready to go. Before commiting, darglint2 will be run on the staged files. If it finds any errors, the user is notified and the commit is aborted. Store necessary configuration (such as error formatting) in .darglint2, .darglint, setup.cfg or tox.ini.

Roadmap

Below are some of the features or efforts from the original darglint project at the time of forking in February 2023. Where a milestone or issue is associated with the idea, it will be mentioned. Some of these ideas were moonshots and may not get implemented. They are ordered roughly according to priority/feasibility stated by the original author.

  • Expose command-line options through sphinx.

  • Robust logging for errors caused/encountered by darglint2.

  • Check class docstrings (See darglint#25).

  • Autoformatting docstrings. (See darglint milestone #3).

  • Optional aggressive style checking through command line flag.

  • ALE support.

  • Syntastic support. (Syntastic is not accepting new checkers until their next API stabilizes, so this may take some time.)

Development and Contributions

Development Setup

Install darglint2. First, clone the repository:

git clone https://github.com/akaihola/darglint2.git

cd into the directory, create a virtual environment (optional), then setup:

cd darglint2/
virtualenv -p python3.6 .env
source .env/bin/activate
pip install -e .

You can install dependencies using

pip install poetry
poetry install

You can run the tests using

python setup.py test

Or, install pytest manually, cd to the project’s root directory, and run

pytest

This project tries to conform by the styles imposed by pycodestyle and pydocstyle, as well as by darglint2 itself.

A dockerfile exists for testing with Python3.4. Although it’s not officially supported (only 3.6+), it’s nice to try to make minor version numbers support it. You would build the dockerfile and test using something like

pushd docker-build
docker build -t darglint2-34 -f Dockerfile.test34 .
popd
docker run -it --rm -v $(pwd):/code darglint2-34 pytest

Tooling and tests

The bin/ folder hold various development utilities for Darglint2. bnf_to_cnf is a utility to convert BNF grammars to CNF grammars), and doc_extract extracts docstrings from repositories and annotates them for use in integration tests.

Note: The order of items in generated Python grammar files may change between runs. It’s ok for the CYK parsing algorithm since it will identify all possible parse trees. If we ever change the parsing algorithm (e.g. LR or LL(K)), we may need to change this method to ensure the order of items is consistent.

There is an integration test framework. Test fixtures are ignored in Git, since the integration tests are only relevant for local development (and even then, mostly just release). The integration tests are as follows:

  • goldens.py: Tests against goldens for individual docstrings. This attempts to ensure that parsed docstrings always contain the expected sections after development. Goldens are generated using the doc_extract utility in the bin/ folder, mostly from large Open Source projects like Django. The format is recorded in the return value of docstringEncoder in bin/doc_extract/static/src/Main.elm. To run, doc_extract -s repos.txt -o docstrings.json accepts local Git repo paths in repos.txt and writes docstrings and metadata into docstrings.json. The Elm webapp (which currently has a radio button bug) accepts docstrings.json, lets you interactively choose the type of each docstring, discard docstrings, and save the result back into output.json which can then be used for integration tests. It’s absent from the repo to avoid needing to handle possible license issues.

  • grammar_size.py: Tests that the grammar size doesn’t increase significantly. Larger grammars will result in longer parse times, and it could be relatively easy to accidentally introduce a much larger grammar.

  • performance.py: Tests performance of the parser against individual docstrings to make sure we don’t introduce a performance regression. Also tests performance for individual files in some repositories.

  • TODO: We still need to add some tests against multiple configurations, and against entire repositories.

Contribution

If you would like to tackle an issue or feature, email me or comment on the issue to make sure it isn’t already being worked on. Contributions will be accepted through pull requests. New features should include unit tests, and, of course, properly formatted documentation.

Also, check out the wiki prior to updating the grammar. It includes a description of darglint2’s parsing pipline.