User config management made simple and powerful

## Project description

      _                  __
___| |__   ___  _ __  / _|
/ __| '_ \ / _ \| '_ \| |_
| (__| | | | (_) | | | |  _|
\___|_| |_|\___/|_| |_|_|


An early concept for a opinionated configuration framework to give power users the freedom to use whatever config file syntax they like. It's written in python.

Chonf enforces a structure of configuration that can always be translated into a tree, such as nested dictionaries in python. Most of the widely used config file syntaxes work like this: Json, Toml, Yaml, and so on.

# How To

from chonf import load, Option

# Create a configuration model as a dictionary
mymodel = {
"ui_theme": Option("dark"), # default value here is "dark"
"user_info": {
"name": Option(), # empty Option, default here is None
"email": Option(),
}
}

mymodel,
author="mycompany",
name="myprogram",
)

# myconfigs is now a dict of your configurations


A Toml config file, for example, could look like this:

ui_theme = "light"

[user_info]
name = "Tom Preston-Werner"
email = "tomate@hotmail.com"


## Overwriting stuff with environment variables

For Chonf, environment variables have higher precedence than config files such as Toml or Json files, so you can quickly overwrite a configuration set in a config file with a environment variable such as:

foo@bar:~\$ export myprogram__ui_theme="adwaita"


This allows for quick tests in config changes without opening, editing and saving config files multiple times.

The syntax for the env variable names is the following: first the specified env_prefix argument on the load function, than the keys from the root of the dictionary tree down to the name of the option itself, separated by double__underscores.

On the previous example, this is how you would name the environment variables:

• myconfigs["ui_theme"] is named myprogram__ui_theme
• myconfigs["user_info"]["name"] is named myprogram__user_info__name
• myconfigs["user_info"]["email"] is named myprogram__user_info__email

This unusual double underscore syntax allows usage of single underscores as word separators on the option names without creating ambiguity.

Note that the default for environment variables is to use only letters, digits and underscores, so it's advisable to use only these characters for naming the model dictionary keys. Otherwise, users might not be able to easily access those options through the shell.

## Required options

From Chonf you can also import a Required dataclass that will not take a default value, and will cause the load() function to raise a ConfigLoadingIncomplete exception if the option is not found. This exception is also a dataclass that will contain the data that was read, with Required objects left where stuff was not found, a list of all the keys for the unlocated options and also invalid options (see Functional Pre-Processing and Validation). As an example, if your code looks like this:

from chonf import load, Option, Required

model = {
"a": Required(),
"b": Option(),
"c": {
"c_a": Required(),
"c_b": Option(),
}
}



, if the option conf["a"] or conf["c"]["a"] are missing, the load() function will raise a ConfigLoadingIncomplete exception. In case all options are missing (following comments represent output):

try:
print(err.unlocated_keys)
# [["a"], ["c","c_a"]]

# {
#   "a": InvalidOption(value=None, expected=Required()),
#   "b": None,
#   "c": {
#       "c_a": InvalidOption(value=None, expected=Required()),
#       "c_b": None
#   }
# }


# Multiple Config Paths

If you decide to offer more than one option of config file location, pass a list of paths instead:

configs = load(
model = mymodel,
author = "mycompany",
name = "myprogram",
path = ["/home/me/.config/myprogram", "/etc/myprogram"],
)


You can have several config directory options. What comes first in the list will have higher priority. In this example, the user level configurations will be able to shadow the ones defined system-wide.

# Change Environment Variables Prefix

If you would like to use something other than the name of your program as the prefix for the env vars, pass the env_prefix argument, as the name is also used in the path to the default config directories.

configs = load(
model = mymodel,
author = "mycompany",
name = "myprogram",
env_prefix = "mypro",
...


# Functional Pre-Processing and Validation

If some of your options require some specific type of data or any other kind of validation, Chonf provides a functional approach to verifying and pre-processing your options during loading.

Pass to your option a callable object (like a function) as the pre_process argument. The function should be able to receive a single positional argument (the value read from some env var or file) and returning the pre-processed value. If the value is invalid, the function should raise a chonf.InvalidOption exception containing the value received and some info about what was expected.

In the following snippet, we can check if a option is a number and immediately convert it into it's numeric type:

from chonf import load, Option, InvalidOption

def into_number(value):
try:
return float(value)
except ValueError as err:
raise InvalidOption(value, "something compatible with float") from err

model = {"username": Option(), "a_number": Option(pre_process=into_number)}


Future versions of Chonf will implement common predefined pre-process functions.

# Repeating Structures

Sometimes you might want to have a group of many similar configuration structures somewhere in your model. For example, in a text editor, you might have a group of language server definitions, with each one's specific info. A simple example:

[language_servers]

[language_servers.python]
name = "Python Language Server"
command = "pyls"

[language_servers.fortran]
name = "Fortran Language Server"
command = "flang"


Chonf provides a simple way to define such repetitions. Define the repeating sub-model in a Repeat object:

from chonf import load, Required, Repeat

model = {
"language_servers": Repeat({
"name": Required(),
"command": Required()
})
}



Notice how you can have required options inside the repeating structures. The blocks for python and fortran in the previous example are not required, but if they are found, their name and command will be required. Also, Repeat must find a subtree on your configurations, if it finds a leaf node (a value such as a string) it will deem it invalid.

Also, if you know the keys for each block, nothing is stopping you from using dictionary comprehensions.

# Default Config Paths

Chonf should be able to identify the user's OS and locate default locations for user-wide and system-wide configurations, but this feature is not well tested yet. Feedback is very welcome.

To try it, just skip the path argument on the load() function, and it will try to find the config files in your system's default. If you wish to see where this location might be by Chonf's algorithm, call chonf.default_path() or take a look at the paths.py module on the source code.

# Next Steps

This project is still in its early stages, this is what we plan to allow next:

• Procedurally generated options from python scripts
• Yaml and XML (.ini is way too limiting so we probably won't implement it)
• User and System level priority order customization for the end user
• Support for custom relative paths (location of config files inside config folder)
• Function for dumping data back into whatever format the end user defined as preferred
• Make all file formats optional other than Json that already comes with python.
• Figure out versioning and publish on pip

## Project details

Uploaded source
Uploaded py3