Advertisement

Your Ad could be here. I want to connect my readers to relavant ads. If you have a product targeted at developers, let's talk. [email protected]

There are various ways to configure python tools, config files, code, or environment variables. Let's look at a few projects that allow users to configure them through the use of config files and how they do it.

Motivation

This will not include how they are implemented, I've looked at a few and its not simple. This will focus on where config is placed and the order in which duplicates are resolved.

The motivation of this article is to serve as a bit of a reference guide for those who may want to create their own package that needs configuration.

Flake8

Global

User settings can exist in the users ~/.config/flake8 file to configure how flake8 runs on their machine.

  • ~/.config/flake8

Per-Project

Only One project config file will be considered, but allows for several options. These files all use the ini format and must have a [flake8] section header to be consideered.

Selection of the config file can also be overridden by the --config cli option.

An extra config file may be selected as --append-config. It will be read in last and take highest precedence.

  • tox.ini
  • setup.cfg
  • .pep8
  • .flake8

Example Config

valid in any of the supported files


[flake8]
max-line-length = 88
extend-ignore = E203, W503

Options

The number of options configured through config files is fairly short for flake8.

  • exclude
  • filename
  • select
  • ignore
  • max-line-length
  • format
  • max-complexity

Black

Black only supports TOML file formats for configuration.

Global

Black provides no global config support. If you really needed one I guess you could make a cli alias.

Per-Project

Black states that it includes sane defaults that do not need configured, but if you need to do so it only supports pyproject.toml or cli arguments.

Personally I believe that a lot of work went into making these sane defaults really good. I personally do not make any configuration changes to black.

  • pyproject.toml

Example

pyproject.toml


[tool.black]
line-length = 88
target-version = ['py37']
include = '\.pyi?$'
exclude = '''

(
  /(
      \.eggs         # exclude a few common directories in the
    | \.git          # root of the project
    | \.hg
    | \.mypy_cache
    | \.tox
    | \.venv
    | _build
    | buck-out
    | build
    | dist
  )/
  | foo.py           # also separately exclude a file named foo.py in
                     # the root of the project
)
'''

Resolution

Black will use teh pyproject.toml file for configuration, then make any addional overrides through the use of command line arguments.

MyPy

mypy takes the cake for the most complex configuration. Primarily because you can configure how it treats different modules specifically. These modules may be inside your codebase or installed and imported in.

Per-Project

  • --config-file
  • mypy.ini
  • .mypy.ini

Global

  • $XDG_CONFIG_HOME/mypy/config
  • ~/.config/mypy/config
  • ~/.mypy.ini

Resolution

  • --config-file
  • mypy.ini
  • .mypy.ini
  • setup.cfg
  • $XDG_CONFIG_HOME/mypy/config
  • ~/.config/mypy/config
  • ~/.mypy.ini

Example

mypy.ini


# Global options:

[mypy]
python_version = 2.7
warn_return_any = True
warn_unused_configs = True

# Per-module options:

[mypy-mycode.foo.*]
disallow_untyped_defs = True

[mypy-mycode.bar]
warn_return_any = False

[mypy-somelibrary]
ignore_missing_imports = True

Kedro - framework

Kedro is a unique one here. It offers two distinctly different configurations, one for how the framework behaves and the other for actual project config.

Kedro does utilizes a settings.py and pyproject.toml to define a bit more of the framework settings. These are the outter layer of your project.

These files sit at the root of the project.

pyproject.toml

This replaces much of what used to be specified in run.py.

  • package_name
  • project_name
  • project_version
  • source_dir

Settings.py

  • DISABLE_HOOKS_FOR_PLUGINS
  • HOOKS
  • SESSION_STORE_CLASS
  • SESSION_STORE_ARGS
  • CONTEXT_CLASS

Kedro - project

Within the project generally in the src/conf directory kedro allows you to set both local and base configurations. Local configurations will be git ignored and most commonly used for credentials.

  • catalog
  • logging
  • credentials

Config Loader

Kedro lets you setup the config loader if you choose to do so. You can configure the directories to look in as well as the glob pattern for files.


from kedro.config import ConfigLoader

conf_paths = ["conf/base", "conf/local"]
conf_loader = ConfigLoader(conf_paths)
conf_catalog = conf_loader.get("catalog*", "catalog*/**")

additional envs

Additional to the base and local config, kedro lets you specify an env at runtime through a --env argumet or a KEDRO_ENV variable. setting this will additionally tell kedro to reach into conf/<env-name> for configuration.

Resolution Order

kedro will load each config starting from base, local, then env and will overrite any colllisions along the way.

precedence heirarchy

  • env
  • local
  • base

Jinja Support

As of 0.17.0 kedro supports jinja2 templates in its yml configuration files. This is quite beneficial as catalogs can become incredebly repetative.


fast-trains:
    type: MemoryDataSet

fast-cars:
    type: pandas.CSVDataSet
    filepath: s3://${bucket_name}/fast-cars.csv
    save_args:
        index: true


slow-trains:
    type: MemoryDataSet

slow-cars:
    type: pandas.CSVDataSet
    filepath: s3://${bucket_name}/slow-cars.csv
    save_args:
        index: true

pytest

Currently pytest is configured

resolution order

pytest will look for the existence of each of these files, if its a match it will stop looking for new files, even if the file is empty.

  • pytest.ini
  • pyproject.toml with [tool.pytest.ini_options]
  • tox.ini with [pytest]
  • setup.cfg with [tool:pytest]

Multiple Config

pytest is a bit unique here in that it allows for multiple configs. There is a complex resolution for module specific configuration, but essentially it does the resolution highlighted above through a number of directories and returns the config closest to the test module.

Example pytest config


# pytest.ini
[pytest]
minversion = 6.0
addopts = -ra -q
testpaths =
    tests
    integration

Command Line Options

As far as I am aware every option specified in a config file can also be configured or overridden at the command line.

ipython

Ipython is configured completely at a system level with python scripts within the users ~/.ipython/ directory. The user may have multiple profiles that can be created by running ipython profile create [profilename] or specified by running ipython --profile=[profilename]

Config Directory

By default this is ~/.ipython, but an be configured by setting the IPYTHONDIR environment variable or --ipython-dir=<path> command line option.

Example Config


# sample ipython_config.py
c = get_config()

c.TerminalIPythonApp.display_banner = True
c.InteractiveShellApp.log_level = 20
c.InteractiveShellApp.extensions = [
    'myextension'
]
c.InteractiveShellApp.exec_lines = [
    'import numpy',
    'import scipy'
]
c.InteractiveShellApp.exec_files = [
    'mycode.py',
    'fancy.ipy'
]
c.InteractiveShell.autoindent = True
c.InteractiveShell.colors = 'LightBG'
c.InteractiveShell.confirm_exit = False
c.InteractiveShell.deep_reload = True
c.InteractiveShell.editor = 'nano'
c.InteractiveShell.xmode = 'Context'

c.PromptManager.in_template  = 'In [\#]: '
c.PromptManager.in2_template = '   .\D.: '
c.PromptManager.out_template = 'Out[\#]: '
c.PromptManager.justify = True

c.PrefilterManager.multi_line_specials = True

c.AliasManager.user_aliases = [
 ('la', 'ls -al')
]

CommandLine Overrides

Every configurable value can be overridden from the command line.


ipython --ClassName.attribute=value

Config Magic

Configuration can be overridden at runtime with the %config magic.


%config IPCompleter.greedy = True

Startup

Every ipython profile has a startup directory where it will execute each .py and .ipy file on startup. You can make additional configuration here, import modules you want readily available, execute literally any python code you want to at the startup of that particular profile.