Envclasses are a thin wrapper around dataclasses which allows for the values to be defined via environment variables rather than explicitly in code. Values are typed and are able to be defaulted.
I got tired of writing code that was configured through environment variables, referencing the environment variable when I needed to instantiate something. This made it difficult to keep up with how I could configure that software that I was writing as I would have to comb through the code and make sure that the documentation was up to date.
Envclasses are an attempt to reduce the sprawl of configuration through environment variables and centralize configuration into a single, document-able class. They are both inspired by, and built on top of dataclasses, which is why their structure is so similar.
Defining an environment class is simple:
from envclasses import EnvClassMeta
class ApplicationConfig(metaclass=EnvClassMeta):
db_url: str
db_username: str
db_password: str
port: int = 5050
mode: str = 'development'
config = ApplicationConfig()
The provided metaclass will turn the ApplicationConfig
into a dataclass with fields defined from os.environ
.
The metaclass will prioritize upper-case versions of fields before lower-case, that is to say DB_URL
would be
prioritized over db_url
. Mixed-case variants are not considered.
If values are not defined, the metaclass will wait until all fields have been tested to report which are missing. In the
event that we should ignore missing fields, the environment variable env_ignore_missing
should be defined as true
or
yes
.
In cases of testing, having the environment read once at initialization of the module its defined in isn't ideal. It can lead to a lot of plumbing work to overwrite mock values and import sequences.
For cases liek this, using Refresh Mode might be appropriate. In this mode, the env class will re-read any environment variable of a given attriubute on access of said attribute. For example:
from envclasses import EnvClassMeta
class FreshLoadingConfig(metaclass=EnvClassMeta):
test_var: int
envclasses.os.environ['ENV_IGNORE_ERRORS'] = 'yes'
envclasses.os.environ['PYENV_CLASS_REFRESH_LOAD'] = 'yes'
config = ApplicationConfig()
print(config.test_var)
# prints "None"
envclasses.os.environ['test_var'] = '100'
print(config.test_var)
# prints "100"
Note that you should use this behaviour in production given the dynamic nature of it. It is only meant for testing ease.