load-environ-typed 0.3.0

You define a class, we load it up from environment in a type safe way

load-environ-typed

This library lets you declare a NamedTuple or dataclass with field types, and gives you a function that will load the values from the environment, validating your types for you.

Getting started

from load_environ_typed import load

from dataclasses import dataclass

@dataclass
class MyEnviron:
	db_host: str
	db_port: int

environ = load(MyEnviron)

FAQ

What types can I use? What about custom types?

By default, any type that takes a single string in it's constructor can be used as type. Think of int, str, float, pathlib.Path, etc.

We've added default loaders for the types below:

• bool - "true" or "false", case insenstive
• datetime.date - Using datetime.date.fromisoformat
• datetime.time - Using datetime.time.fromisoformat
• datetime.datetime - Using datetime.datetime.fromisoformat

How are fields matched to enviroment variables?

The loader assumes the names are the same, except that the class fields are lowercase, and the environment fields are uppercase. If you have different or more complicated rules, you can pass a name conversion function via field_name_to_var_name.

@dataclass
class MyEnviron:
    iso_date: datetime.date

environ = sut.load(MyEnviron, environ={
    'ISO_DATE': '2021-01-01',
})

Can values be optional?

Certainly:

@dataclass
class MyEnviron:
	DB_HOST: Optional[str]

However:

• empty string and "none" (case insensitive) count as None
• defaults take precedence over optionality

What if my type cannot take a string in its constructor?

You can pass so-called loader functions. These take in a string, and are expected to return a value of the given type, or raise a ValueError when the given string is not valid for the given type. This is also the mechanism that we use to support standard Python types such datetime.date, which is shown below.

@dataclass
class MyEnviron:
    ISO_DATE: datetime.date

environ = sut.load(MyEnviron, environ={
    'ISO_DATE': '2021-01-01',
}, loaders={
    'ISO_DATE': datetime.date.fromisoformat,
    # or, if you want ALL `date`s to use this loader:
    datetime.date: datetime.date.fromisoformat,
})

NOTE: date has a default loader, so you don't need to do this for date.

Can I use Unions?

You can, but you need to use a default loader, as you need some way to distinguish between the types, and there is no general way to do so, at least not without enforcing our way of working on you.

How do I work with default values?

If you want default values, it's probably best to have use a dataclass with kw_only=True, as otherwise you have to order your variables based on whether there's a default or not.

@dataclass(kw_only=True)
class MyEnviron:
	DB_HOST: str
	DB_PORT: int = 3306

For most types, you can simply set the default value as you're used to with dataclasses. However, you may not want to instantiate an (expensive) property as default. In those cases, you can pass defaults along using the defaults argument.

@dataclass
class MyEnviron:
	VAR: SomeExpensiveClass

environ = load(MyEnviron, defaults={
	'VAR': '#!serialized.data!#',
})

NOTE: kw_only requires Python3.10 or higher. Below 3.10, you can use the defaults argument or order your variables. Similarly, if you plan on solely using the defaults argument, you don't need kw_only.

Load returns an instance. What if I want a global?

Since environ is available at startup, and doesn't change, it's perfectly valid to just instantiate a global variable. You should probably use a frozen dataclass for this.

@dataclass(frozen=True)
class MyEnviron:
	DB_HOST: str
	DB_PORT: int

ENVIRON = load(MyEnviron)

What if there's an issue with the default loaders?

First, the loaders you pass will be taken before using the default loaders.

Second, if you have more structural issues with the default loaders, simply pass use_default_loaders=False.

Contributing

Check out this repo, and then use the following steps to test it:

python3 -m venv venv
venv/bin/pip install -r requirements-dev.txt
make test

Deploying

First, update pyproject.toml with the new version number, and commit that.

Then:

rm -f dist/* # Clean old build files
venv/bin/python -m build
venv/bin/python -m twine upload dist/*