Skip to content

Latest commit

 

History

History
736 lines (589 loc) · 32.6 KB

CONTRIBUTING.md

File metadata and controls

736 lines (589 loc) · 32.6 KB
Raw

Contributing to typeshed

Welcome! typeshed is a community project that aims to work for a wide range of Python users and Python codebases. If you're trying a type checker on your Python code, your experience and what you can contribute are important to the project's success.

The contribution process at a glance

  1. Prepare your environment.
  2. Find out where to make your changes.
  3. Prepare your changes:
    • Small fixes and additions can be submitted directly as pull requests, but contact us before starting significant work.
    • Create your stubs, considering what to include and conforming to the coding style.
  4. Optionally format and check your stubs.
  5. Optionally run the tests.
  6. Submit your changes by opening a pull request.
  7. Make sure that all tests in CI are passing.

You can expect a reply within a few days, but please be patient when it takes a bit longer. For more details, read below.

Preparing the environment

Code away!

Typeshed runs continuous integration (CI) on all pull requests. This means that if you file a pull request (PR), our full test suite -- including our linter, Flake8 -- is run on your PR. It also means that bots will automatically apply changes to your PR (using Black and Ruff) to fix any formatting issues. This frees you up to ignore all local setup on your side, focus on the code and rely on the CI to fix everything, or point you to the places that need fixing.

... Or create a local development environment

If you prefer to run the tests and formatting locally, it's possible too. Follow platform-specific instructions below. For more information about our available tests, see tests/README.md.

Whichever platform you're using, you will need a virtual environment. If you're not familiar with what it is and how it works, please refer to this documentation.

Note that some tests require extra setup steps to install the required dependencies.

Linux/Mac OS

On Linux and Mac OS, you will be able to run the full test suite on Python 3.9, 3.10, or 3.11. To install the necessary requirements, run the following commands from a terminal window:

$ python3 -m venv .venv
$ source .venv/bin/activate
(.venv)$ pip install -U pip
(.venv)$ pip install -r requirements-tests.txt

Windows

If you are using a Windows operating system, you will not be able to run the pytype tests, as pytype does not currently support running on Windows. One option is to install Windows Subsystem for Linux, which will allow you to run the full suite of tests. If you choose to install WSL, follow the Linux/Mac OS instructions above.

If you do not wish to install WSL, run the following commands from a Windows terminal to install all non-pytype requirements:

> python -m venv .venv
> .venv\scripts\activate
(.venv) > pip install -U pip
(.venv) > pip install -r "requirements-tests.txt"

Code formatting

The code is formatted using Black. Various other autofixes and lint rules are also performed by Ruff and Flake8, with plugins flake8-pyi, and flake8-noqa.

The repository is equipped with a pre-commit.ci configuration file. This means that you don't need to do anything yourself to run the code formatters or linters. When you push a commit, a bot will run those for you right away and add any autofixes to your PR. Anything that can't be autofixed will show up as a CI failure, hopefully with an error message that will make it clear what's gone wrong.

That being said, if you want to run the formatters and linters locally when you commit, you're free to do so. To use the same configuration as we use in CI, we recommend doing this via pre-commit:

(.venv)$ pre-commit run --all-files

Where to make changes

Standard library stubs

The stdlib directory contains stubs for modules in the Python standard library — which includes pure Python modules, dynamically loaded extension modules, hard-linked extension modules, and the builtins. The VERSIONS file lists the versions of Python where the module is available.

Third-party library stubs

We accept stubs for third-party packages into typeshed as long as:

  • the package is publicly available on the Python Package Index;
  • the package supports any Python version supported by typeshed; and
  • the package does not ship with its own stubs or type annotations.

The fastest way to generate new stubs is to use scripts/create_baseline_stubs.py (see below).

Stubs for third-party packages go into the stubs directory. Each subdirectory there represents a PyPI distribution, and contains the following:

  • METADATA.toml, describing the package. See below for details.
  • Stubs (i.e. *.pyi files) for packages and modules that are shipped in the source distribution.
  • (Rarely) some docs specific to a given type stub package in README file.

When a third party stub is added or modified, an updated version of the corresponding distribution will be automatically uploaded to PyPI within a few hours. Each time this happens the least significant version level is incremented. For example, if stubs/foo/METADATA.toml has version = "x.y" the package on PyPI will be updated from types-foo-x.y.n to types-foo-x.y.n+1.

Note: In its current implementation, typeshed cannot contain stubs for multiple versions of the same third-party library. Prefer to generate stubs for the latest version released on PyPI at the time of your stubbing.

The METADATA.toml file

The metadata file describes the stubs package using the TOML file format. Currently, the following keys are supported:

  • version: The versions of the library that the stubs support. Two formats are supported:
    • A concrete version. This is especially suited for libraries that use Calendar Versioning.
    • A version range ending in .*. This is suited for libraries that reflect API changes in the version number only, where the API-independent part is represented by the asterisk. In the case of Semantic Versioning, this version could look like this: 2.7.*. When the stubs are updated to a newer version of the library, the version of the stub should be bumped (note that previous versions are still available on PyPI).
  • requires (optional): A list of other stub packages or packages with type information that are imported by the stubs in this package. Only packages generated by typeshed or required by the upstream package are allowed to be listed here, for security reasons. See this issue for more information about what external dependencies are allowed.
  • extra_description (optional): Can be used to add a custom description to the package's long description. It should be a multi-line string in Markdown format.
  • stub_distribution (optional): Distribution name to be uploaded to PyPI. This defaults to types-<distribution> and should only be set in special cases.
  • upstream_repository (recommended): The URL of the upstream repository.
  • obsolete_since (optional): This field is part of our process for removing obsolete third-party libraries. It contains the first version of the corresponding library that ships its own py.typed file.
  • no_longer_updated (optional): This field is set to true before removing stubs for other reasons than the upstream library shipping with type information.
  • upload (optional): This field is set to false to prevent automatic uploads to PyPI. This should only used in special cases, e.g. when the stubs break the upload.
  • partial_stub (optional): This field marks the type stub package as partial. This is for 3rd-party stubs that don't cover the entirety of the package's public API.
  • requires_python (optional): The minimum version of Python required to install the type stub package. It must be in the form >=3.*. If omitted, the oldest Python version supported by typeshed is used.

In addition, we specify configuration for stubtest in the tool.stubtest table. This has the following keys:

  • skip (default: false): Whether stubtest should be run against this package. Please avoid setting this to true, and add a comment if you have to.
  • ignore_missing_stub: When set to true, this will add the --ignore_missing_stub option to the stubtest call. See tests/README.md for more information. In most cases, this field should be identical to partial_stub.
  • apt_dependencies (default: []): A list of Ubuntu APT packages that need to be installed for stubtest to run successfully.
  • brew_dependencies (default: []): A list of MacOS Homebrew packages that need to be installed for stubtest to run successfully
  • choco_dependencies (default: []): A list of Windows Chocolatey packages that need to be installed for stubtest to run successfully
  • platforms (default: ["linux"]): A list of OSes on which to run stubtest. Can contain win32, linux, and darwin values. If not specified, stubtest is run only on linux. Only add extra OSes to the test if there are platform-specific branches in a stubs package.

*_dependencies are usually packages needed to pip install the implementation distribution.

The format of all METADATA.toml files can be checked by running python3 ./tests/check_consistent.py.

Preparing Changes

Before you begin

If your change will be a significant amount of work to write, we highly recommend starting by opening an issue laying out what you want to do. That lets a conversation happen early in case other contributors disagree with what you'd like to do or have ideas that will help you do it.

Format

Each Python module is represented by a .pyi "stub file". This is a syntactically valid Python file, although it usually cannot be run by Python (since forward references don't require string quotes). All the methods are empty.

Python function annotations (PEP 3107) are used to describe the signature of each function or method.

See PEP 484 for the exact syntax of the stub files and below for the coding style used in typeshed.

Auto-generating stub files

Typeshed includes scripts/create_baseline_stubs.py. It generates stubs automatically using a tool called stubgen that comes with mypy.

To get started, fork typeshed, clone your fork, and then create a virtualenv. You can then install the library with pip into the virtualenv and run the script below, replacing $INSERT_LIBRARY_NAME_HERE with the name of the library:

(.venv)$ pip install $INSERT_LIBRARY_NAME_HERE
(.venv)$ python3 scripts/create_baseline_stubs.py $INSERT_LIBRARY_NAME_HERE

When the script has finished running, it will print instructions telling you what to do next.

If it has been a while since you set up the virtualenv, make sure you have the latest mypy (pip install -r requirements-tests.txt) before running the script.

Supported type system features

Since PEP 484 was accepted, there have been many other PEPs that added new features to the Python type system. In general, new features can be used in typeshed as soon as the PEP has been accepted and implemented and most type checkers support the new feature.

Supported features include:

Features from the typing module that are not present in all supported Python versions must be imported from typing_extensions instead in typeshed stubs. This currently affects:

  • TypeAlias (new in Python 3.10)
  • Concatenate (new in Python 3.10)
  • ParamSpec (new in Python 3.10)
  • TypeGuard (new in Python 3.10)
  • Self (new in Python 3.11)
  • Never (new in Python 3.11)
  • LiteralString (new in Python 3.11)
  • TypeVarTuple and Unpack (new in Python 3.11)
  • Required and NotRequired (new in Python 3.11)
  • Buffer (new in Python 3.12; in the collections.abc module)
  • @deprecated (new in Python 3.13; in the warnings module)

Some type checkers implicitly promote the bytearray and memoryview types to bytes. PEP 688 removes this implicit promotion. Typeshed stubs should be written assuming that these promotions do not happen, so a parameter that accepts either bytes or bytearray should be typed as bytes | bytearray. Often one of the aliases from _typeshed, such as _typeshed.ReadableBuffer, can be used instead.

What to include

Stubs should include the complete interface (classes, functions, constants, etc.) of the module they cover, but it is not always clear exactly what is part of the interface.

The following should always be included:

  • All objects listed in the module's documentation.
  • All objects included in __all__ (if present).

Other objects may be included if they are being used in practice or if they are not prefixed with an underscore. This means that typeshed will generally accept contributions that add missing objects, even if they are undocumented. Undocumented objects should be marked with a comment of the form # undocumented. Example:

def list2cmdline(seq: Sequence[str]) -> str: ...  # undocumented

We accept such undocumented objects because omitting objects can confuse users. Users who see an error like "module X has no attribute Y" will not know whether the error appeared because their code had a bug or because the stub is wrong. Although it may also be helpful for a type checker to point out usage of private objects, we usually prefer false negatives (no errors for wrong code) over false positives (type errors for correct code). In addition, even for private objects a type checker can be helpful in pointing out that an incorrect type was used.

What to do when a project's documentation and implementation disagree

Type stubs are meant to be external type annotations for a given library. While they are useful documentation in its own merit, they augment the project's concrete implementation, not the project's documentation. Whenever you find them disagreeing, model the type information after the actual implementation and file an issue on the project's tracker to fix their documentation.

Stub versioning

You can use checks like if sys.version_info >= (3, 12): to denote new functionality introduced in a given Python version or solve type differences. When doing so, only use two-tuples. Because of this, if a given functionality was introduced in, say, Python 3.11.4, your check:

  • should be expressed as if sys.version_info >= (3, 11):
  • should NOT be expressed as if sys.version_info >= (3, 11, 4):
  • should NOT be expressed as if sys.version_info >= (3, 12):

When your stub contains if statements for different Python versions, always put the code for the most recent Python version first.

Stub file coding style

Syntax example

The below is an excerpt from the types for the datetime module.

MAXYEAR: int
MINYEAR: int

class date:
    def __new__(cls, year: SupportsIndex, month: SupportsIndex, day: SupportsIndex) -> Self: ...
    @classmethod
    def fromtimestamp(cls, timestamp: float, /) -> Self: ...
    @classmethod
    def today(cls) -> Self: ...
    @classmethod
    def fromordinal(cls, n: int, /) -> Self: ...
    @property
    def year(self) -> int: ...
    def replace(self, year: SupportsIndex = ..., month: SupportsIndex = ..., day: SupportsIndex = ...) -> Self: ...
    def ctime(self) -> str: ...
    def weekday(self) -> int: ...

Conventions

Stub files are like Python files and you should generally expect them to look the same. Your tools should be able to successfully treat them as regular Python files. However, there are a few important differences you should know about.

Style conventions for stub files are different from PEP 8. The general rule is that they should be as concise as possible. Specifically:

  • all function bodies should be empty;
  • prefer ... over pass;
  • prefer ... on the same line as the class/function signature;
  • avoid vertical whitespace between consecutive module-level functions, names, or methods and fields within a single class;
  • use a single blank line between top-level class definitions, or none if the classes are very small;
  • do not use docstrings;
  • use variable annotations instead of type comments, even for stubs that target older versions of Python.

The primary users for stub files are type checkers, so stub files should generally only contain information necessary for the type checker, and leave out unnecessary detail. However, stubs also have other use cases:

  • stub files are often used as a data source for IDEs, which will often use the signature in a stub to provide information on functions or classes in tooltip messages.
  • stubs can serve as useful documentation to human readers, as well as machine-readable sources of data.

As such, we recommend that default values be retained for "simple" default values (e.g. bools, ints, bytes, strings, and floats are all permitted). Use = ... for more complex default values, rather than trying to exactly reproduce the default at runtime.

Some further tips for good type hints:

  • for arguments that default to None, use Foo | None explicitly for the type annotation;
  • use float instead of int | float for parameter annotations (see PEP 484 for motivation).
  • use built-in generics (list, dict, tuple, set), instead of importing them from typing.
  • use X | Y instead of Union[X, Y] and X | None, instead of Optional[X];
  • import collections (Mapping, Iterable, etc.) from collections.abc instead of typing;
  • avoid invariant collection types (list, dict) for function parameters, in favor of covariant types like Mapping or Sequence;
  • avoid union return types: python/mypy#1693;
  • use platform checks like if sys.platform == 'win32' to denote platform-dependent APIs;
  • use mypy error codes for mypy-specific # type: ignore annotations, e.g. # type: ignore[override] for Liskov Substitution Principle violations.
  • use pyright error codes for pyright-specific suppressions, e.g. # pyright: ignore[reportGeneralTypeIssues].
    • pyright is configured to discard # type: ignore annotations. If you need both on the same line, mypy's annotation needs to go first, e.g. # type: ignore[override] # pyright: ignore[reportGeneralTypeIssues].

Imports in stubs are considered private (not part of the exported API) unless:

  • they use the form from library import name as name (sic, using explicit as even if the name stays the same); or
  • they use the form from library import * which means all names from that library are exported.

When adding type hints, avoid using the Any type when possible. Reserve the use of Any for when:

  • the correct type cannot be expressed in the current type system; and
  • to avoid union returns (see above).

Note that Any is not the correct type to use if you want to indicate that some function can accept literally anything: in those cases use object instead.

Stub files support forward references natively. In other words, the order of class declarations and type aliases does not matter in a stub file. You can also use the name of the class within its own body. Focus on making your stubs clear to the reader. Avoid using string literals in type annotations.

Context managers

When adding type annotations for context manager classes, annotate the return type of __exit__ as bool only if the context manager sometimes suppresses exceptions -- if it sometimes returns True at runtime. If the context manager never suppresses exceptions, have the return type be either None or bool | None. If you are not sure whether exceptions are suppressed or not or if the context manager is meant to be subclassed, pick bool | None. See python/mypy#7214 for more details.

__enter__ methods and other methods that return instances of the current class should be annotated with typing_extensions.Self (example).

Naming

Type variables and aliases you introduce purely for legibility reasons should be prefixed with an underscore to make it obvious to the reader they are not part of the stubbed API.

A few guidelines for protocol names below. In cases that don't fall into any of those categories, use your best judgement.

  • Use plain names for protocols that represent a clear concept (e.g. Iterator, Container).
  • Use SupportsX for protocols that provide callable methods (e.g. SupportsInt, SupportsRead, SupportsReadSeek).
  • Use HasX for protocols that have readable and/or writable attributes or getter/setter methods (e.g. HasItems, HasFileno).

@deprecated

Typeshed uses the @typing_extensions.deprecated decorator (@warnings.deprecated since Python 3.13) to mark deprecated functionality; see PEP 702.

A few guidelines for how to use it:

  • In the standard library, apply the decorator only in Python versions where an appropriate replacement for the deprecated functionality exists. If in doubt, apply the decorator only on versions where the functionality has been explicitly deprecated, either through runtime warnings or in the documentation. Use if sys.version_info checks to apply the decorator only to some versions.
  • Keep the deprecation message concise, but try to mention the projected version when the functionality is to be removed, and a suggested replacement.

Incomplete annotations

When submitting new stubs, it is not necessary to annotate all arguments, return types, and fields. Such items should either be left unannotated or use _typeshed.Incomplete if this is not possible:

from _typeshed import Incomplete

field: Incomplete  # unannotated

def foo(x): ...  # unannotated argument and return type

Incomplete can also be used for partially known types:

def foo(x: Incomplete | None = None) -> list[Incomplete]: ...

Any vs. Incomplete

While Incomplete is a type alias of Any, they serve difference purposes: Incomplete is a placeholder where a proper type might be substituted. It's a "to do" item and should be replaced if possible. Any is used when it's not possible to accurately type an item using the current type system. It should be used sparingly.

"The Any trick"

Consider the following (simplified) signature of re.Match[str].group:

class Match:
    def group(self, group: str | int, /) -> str | Any: ...

The str | Any seems unnecessary and weird at first. Because Any includes all strings, you would expect str | Any to be equivalent to Any, but it is not. To understand the difference, let's look at what happens when type-checking this simplified example:

Suppose you have a legacy system that for historical reasons has two kinds of user IDs. Old IDs look like "legacy_userid_123" and new IDs look like "456_username". The function below is supposed to extract the name "USERNAME" from a new ID, and return None if you give it a legacy ID.

import re

def parse_name_from_new_id(user_id: str) -> str | None:
    match = re.fullmatch(r"\d+_(.*)", user_id)
    if match is None:
        return None
    name_group = match.group(1)
    return name_group.uper()  # This line is a typo (`uper` --> `upper`)

The .group() method returns None when the given group was not a part of the match. For example, with a regex like r"\d+_(.*)|legacy_userid_\d+", we would get a match whose .group(1) is None for the user ID "legacy_userid_7". But here the regex is written so that the group always exists, and match.group(1) cannot return None. Match groups are almost always used in this way.

Let's now consider typeshed's -> str | Any annotation of the .group() method:

  • -> Any would mean "please do not complain" to type checkers. If name_group has type Any, you will get no error for this.
  • -> str would mean "will always be a str", which is wrong, and would cause type checkers to emit errors for code like if name_group is None.
  • -> str | None means "you must check for None", which is correct but can get annoying for some common patterns. Checks like assert name_group is not None would need to be added into various places only to satisfy type checkers, even when it is impossible to actually get a None value (type checkers aren't smart enough to know this).
  • -> str | Any means "must be prepared to handle a str". You will get an error for name_group.uper, because it is not valid when name_group is a str. But type checkers are happy with if name_group is None checks, because we're saying it can also be something else than an str.

In typeshed we unofficially call returning Foo | Any "the Any trick". We tend to use it whenever something can be None, but requiring users to check for None would be more painful than helpful.

Submitting Changes

Even more excellent than a good bug report is a fix for a bug, or the implementation of a much-needed stub. We'd love to have your contributions.

We use the usual GitHub pull-request flow, which may be familiar to you if you've contributed to other projects on GitHub. For the mechanics, see Mypy's git and GitHub workflow help page, or GitHub's own documentation.

Anyone interested in type stubs may review your code. One of the maintainers will merge your pull request when they think it's ready. For every pull request, we aim to promptly either merge it or say why it's not yet ready; if you go a few days without a reply, please feel free to ping the thread by adding a new comment.

To get your pull request merged sooner, you should explain why you are making the change. For example, you can point to a code sample that is processed incorrectly by a type checker. It is also helpful to add links to online documentation or to the implementation of the code you are changing.

As the author of the pull request, it is your responsibility to make sure all CI tests pass and that any feedback is addressed. The typeshed maintainers will probably provide some help and may even push changes to your PR to fix any minor issues, but this is not always possible. If a PR lingers with unresolved problems for too long, we may close it (see below).

Also, do not squash your commits or use git commit --amend after you have submitted a pull request, as this erases context during review. We will squash commits when the pull request is merged. This way, your pull request will appear as a single commit in our git history, even if it consisted of several smaller commits.

Third-party library removal policy

Third-party stubs are generally removed from typeshed when one of the following criteria is met:

  • The upstream package ships a py.typed file for at least six months, or
  • the package does not support any of the Python versions supported by typeshed.

If a package ships its own py.typed file, please follow these steps:

  1. Open an issue with the earliest month of removal in the subject.
  2. A maintainer will add the "removal" label.
  3. Open a PR that sets the obsolete_since field in the METADATA.toml file to the first version of the package that shipped py.typed.
  4. After at least six months, open a PR to remove the stubs.

If third-party stubs should be removed for other reasons, please follow these steps:

  1. Open an issue explaining why the stubs should be removed.
  2. A maintainer will add the "removal" label.
  3. Open a PR that sets the no_longer_updated field in the METADATA.toml file to true.
  4. When a new version of the package was automatically uploaded to PyPI (which usually takes up to 3 hours), open a PR to remove the stubs.

If feeling kindly, please update mypy for any stub obsoletions or removals.

Maintainer guidelines

The process for preparing and submitting changes also applies to maintainers. This ensures high quality contributions and keeps everybody on the same page. Avoid direct pushes to the repository.

When reviewing pull requests, follow these guidelines:

  • Typing is hard. Try to be helpful and explain issues with the PR, especially to new contributors.
  • When reviewing auto-generated stubs, just scan for red flags and obvious errors. Leave possible manual improvements for separate PRs.
  • When reviewing large, hand-crafted PRs, you only need to look for red flags and general issues, and do a few spot checks.
  • Review smaller, hand-crafted PRs thoroughly.

When merging pull requests, follow these guidelines:

  • Always wait for tests to pass before merging PRs.
  • Use "Squash and merge" to merge PRs.
  • Make sure the commit message is meaningful. For example, remove irrelevant intermediate commit messages.
  • The commit message for third-party stubs is used to generate the changelog. It should be valid Markdown, be comprehensive, read like a changelog entry, and assume that the reader has no access to the diff.
  • Delete branches for merged PRs (by maintainers pushing to the main repo).

Marking PRs as "deferred"

We sometimes use the "deferred" label to mark PRs and issues that we'd like to accept, but that are blocked by some external factor. Blockers can include:

  • An unambiguous bug in a type checker (i.e., a case where the type checker is not implementing the typing spec).
  • A dependency on a typing PEP that is still under consideration.
  • A pending change in a related project, such as stub-uploader.

PRs should only be marked as "deferred" if there is a clear path towards getting the blocking issue resolved within a reasonable time frame. If a PR depends on a more amorphous change, such as a type system change that has not yet reached the PEP stage, it should instead be closed.

Maintainers who add the "deferred" label should state clearly what exactly the blocker is, usually with a link to an open issue in another project.

Closing stale PRs

To keep the number of open PRs manageable, we may close PRs when they have been open for too long. Specifically, we close open PRs that either have failures in CI, serious merge conflicts or unaddressed feedback, and that have not seen any activity in three months.

We want to maintain a welcoming atmosphere for contributors, so use a friendly message when closing the PR. Example message:

Thanks for contributing! I'm closing this PR for now, because it still
<fails some tests OR has unresolved review feedback OR has a merge conflict>
after three months of inactivity. If you are still interested, please feel free to open
a new PR (or ping us to reopen this one).