Contributing to ahriman

Welcome to ahriman! The goal of the project is to provide the best user experience to manage Arch linux repositories. In order to follow this we set some limitations for the issue creations and heavily restricted code contribution.

Create an issue

Basically just follow the suggested templates:

Bug report requires at least the way to reproduce the issue and behaviour description (expected and actual ones). In order to resolve the bug, the additional questions may be asked, please consider them as lesser evil.
Feature requests basically just require feature description and the purpose why do you want this feature to be implemented. It is required to make sure that the feature you want is going to be implemented in the way you really want it (and to make sure that this feature is not already implemented).
Questions and discussions have free templates, and you are free to ask your question in the way you want.

Code contribution

There are some strict limitation for suggested pull requests:

autopep8, bandit, pylint, mypy must pass.
Test coverage must remain 100%.

Code formatting

In order to resolve all difficult cases the autopep8 is used. You can perform formatting at any time by running make check or running autopep8 command directly.

Code style

Again, the most checks can be performed by make check command, though some additional guidelines must be applied:

Every class, every function (including private and protected), every attribute must be documented. The project follows Google style documentation. The only exception is local functions.

Correct way to document function (if a section is empty, e.g. no notes or there are no args, it should be omitted) is the following:

def foo(argument: str, *, flag: bool = False) -> int:
    """
    do foo. With very very very long
    docstring

    Notes:
        Very important note about this function

    Args:
        argument(str): an argument. This argument has
            long description also
        flag(bool, optional): a flag (Default value = False)

    Returns:
        int: result with
            long description

    Raises:
        RuntimeError: a local function error occurs

    Examples:
        Very informative example how to use this function, e.g.::

            >>> foo("argument", flag=False)

        Note that function documentation is in rST.
    """

Returns should be replaced with Yields for generators.

Class attributes should be documented in the following way:

class Clazz(BaseClazz):
    """
    brand-new implementation of :class:`BaseClazz`

    Attributes:
        CLAZZ_ATTRIBUTE(int): (class attribute) a brand-new class attribute
        instance_attribute(str): an instance attribute

    Examples:
        Very informative class usage example, e.g.::

            >>> from module import Clazz
            >>> clazz = Clazz()
    """

    CLAZZ_ATTRIBUTE = 42

    def __init__(self, *args: Any, **kwargs: Any) -> None:
        """
        default constructor

        Args:
            *args(Any): positional arguments
            **kwargs(Any): keyword arguments
        """
        self.instance_attribute = ""

Type annotations are the must, even for local functions. For the function argument self (for instance methods) and cls (for class methods) should not be annotated.
For collection types built-in classes must be used if possible (e.g. dict instead of typing.Dict, tuple instead of typing.Tuple). In case if built-in type is not available, but collections.abc provides interface, it must be used (e.g. collections.abc.Awaitable instead of typing.Awaitable, collections.abc.Iterable instead of typing.Iterable). For union classes, the bar operator (|) must be used (e.g. float | int instead of typing.Union[float, int]), which also includes typinng.Optional (e.g. str | None instead of Optional[str]).
classmethod should (almost) always return Self. In case of mypy warning (e.g. if there is a branch in which function doesn't return the instance of cls) consider using staticmethod instead.

Recommended order of function definitions in class:

class Clazz:

    def __init__(self) -> None: ...  # replace with `__post_init__` for dataclasses

    @property
    def property(self) -> Any: ...

    @cached_property
    def property_cached(self) -> Any: ...  # cached property has to be treated as normal one

    @classmethod
    def class_method(cls) -> Self: ...

    @staticmethod
    def static_method() -> Any: ...

    def __private_method(self) -> Any: ...

    def _protected_method(self) -> Any: ...

    def usual_method(self) -> Any: ...

    def __hash__(self) -> int: ...  # basically any magic (or look-alike) method

Methods inside one group should be ordered alphabetically, the only exceptions are __init__ (__post_init__ for dataclasses) and __new__ methods which should be defined first. For test methods it is recommended to follow the order in which functions are defined.

Though, we would like to highlight abstract methods (i.e. ones which raise NotImplementedError), we still keep in global order at the moment.

For the most cases there is custom pylint plugin, which performs checks automatically.

Abstract methods must raise NotImplementedError instead of using abc.abstractmethod. The reason behind this restriction is the fact that we have class/static abstract methods for those we need to define their attribute first making the code harder to read.
For any path interactions pathlib.Path must be used.
Configuration interactions must go through ahriman.core.configuration.Configuration class instance.
In case if class load requires some actions, it is recommended to create class method which can be used for class instantiating.
The code must follow the exception safety, unless it is explicitly asked by end user. It means that most exceptions must be handled and printed to log, no other actions must be done (e.g. raising another exception).
Exceptions without parameters should be raised without parentheses, e.g.:
```
raise RuntimeError
```
For the external command ahriman.core.util.check_output function must be used.
Every temporary file/directory must be removed at the end of processing, no matter what. The tempfile module provides good ways to do it.

Import order must be the following:

# optional imports from future module
from __future__ import annotations

# Module import for those which are installed into environment (no matter standard library or not)...
import os
# ...one per line...
import pyalpm
# ...in alphabetical order
import sys

# Blank line between
# ``from module import submodule`` import
from pathlib import Path
# ...again in alphabet order. It is possible to do several imports, but make sure that they are also in alphabetical order.
from pyalpm import Handle, Package

# Blank line again and package imports
from ahriman.core.configuration import Configuration

One file should define only one class, exception is class satellites in case if file length remains less than 400 lines.
It is possible to create file which contains some functions (e.g. ahriman.core.util), but in this case you would need to define __all__ attribute.
The file size mentioned above must be applicable in general. In case of big classes consider splitting them into traits. Note, however, that pylint includes comments and docstrings into counter, thus you need to check file size by other tools.
No global variable is allowed outside of ahriman module. ahriman.core.context is also special case.
Single quotes are not allowed. The reason behind this restriction is the fact that docstrings must be written by using double quotes only, and we would like to make style consistent.
If your class writes anything to log, the ahriman.core.log.LazyLogging trait must be used.

Web API methods must be documented by using aiohttp_apispec library. The schema testing mostly should be implemented in related view class tests. Recommended example for documentation (excluding comments):

import aiohttp_apispec

from marshmallow import Schema, fields  

from ahriman.web.schemas import AuthSchema, ErrorSchema, PackageNameSchema, PaginationSchema
from ahriman.web.views.base import BaseView


class RequestSchema(Schema):

    field = fields.String(metadata={"description": "Field description", "example": "foo"})


class ResponseSchema(Schema):

    field = fields.String(required=True, metadata={"description": "Field description"})


class Foo(BaseView):

    POST_PERMISSION = ...
    ROUTES = ...

    @aiohttp_apispec.docs(
        tags=["Tag"],
        summary="Do foo",
        description="Extended description of the method which does foo",
        responses={
            200: {"description": "Success response", "schema": ResponseSchema},
            204: {"description": "Success response"},  # example without json schema response
            400: {"description": "Bad data is supplied", "schema": ErrorSchema},  # exception raised by this method
            401: {"description": "Authorization required", "schema": ErrorSchema},  # should be always presented
            403: {"description": "Access is forbidden", "schema": ErrorSchema},  # should be always presented
            404: {"description": "Repository is unknown", "schema": ErrorSchema},  # include if BaseView.service() method is called
            500: {"description": "Internal server error", "schema": ErrorSchema},  # should be always presented
        },
        security=[{"token": [POST_PERMISSION]}],
    )
    @aiohttp_apispec.cookies_schema(AuthSchema)  # should be always presented
    @aiohttp_apispec.match_info_schema(PackageNameSchema)
    @aiohttp_apispec.querystring_schema(PaginationSchema)
    @aiohttp_apispec.json_schema(RequestSchema(many=True))
    async def post(self) -> None: ...

It is allowed to change web API to add new fields or remove optional ones. However, in case of model changes, new API version must be introduced.
On the other hand, it is allowed to change method signatures, however, it is recommended to add new parameters as optional if possible. Deprecated API can be dropped during major release.
Enumerations (Enum classes) are allowed and recommended. However, it is recommended to use StrEnum class if there are from/to string conversions and IntEnum otherwise.

Other checks

The projects also uses typing checks (provided by mypy) and some linter checks provided by pylint and bandit. Those checks must be passed successfully for any open pull requests.

Developers how to

Run automated checks

make check tests

Generate documentation templates

make specification

Create release

make VERSION=x.y.z check tests release

The command above will also run checks first and will generate documentation, tags, etc., and will push them to GitHub. Other things will be handled by GitHub workflows automatically.

11 KiB Raw Blame History