envly 0.1.0

Minimal, type-safe environment variable validation for Python.

envly

Minimal, type-safe environment variable validation for Python.

class MyEnv(Environment, prefix="APP_"):
    HOST    = stringVar()
    PORT    = intVar(default=8080)
    DEBUG   = boolVar(default=False)
    TOKEN   = stringVar(secret=True)

env = MyEnv()
print(env.PORT)   # 8080 —> int, not str
print(env.TOKEN)  # SecretStr('**redacted**')

---

Installation

pip install envly

---

Declaring fields

Every field is declared using a factory function. The type is baked into the function, so no annotations are required!

Function	Returns	Notes
stringVar()	str	Supports secret=True
intVar()	int
floatVar()	float
boolVar()	bool	Accepts true/false/1/0/yes/no/on/off
enumVar(*choices)	str	Restricts to allowed values
urlVar()	str	Requires scheme and host
emailVar()	str
pathVar()	Path	Returns pathlib.Path
regexVar(pattern)	str	Must fully match
listVar(of=)	list[T]	Split from a delimited string
jsonVar()	Any	Parsed with json.loads

---

Options

All var functions share these keyword arguments:

PORT = intVar(
    default=8080,                        # used when the var is not set
    validate=lambda x: x < 65536,       # single validator
    validate=(                           # or a tuple of validators
        lambda x: x > 1024,
        lambda x: x < 65536,
    ),
    var_name="SERVICE_PORT",             # override the env var key
)

listVar (sub-coercion per element)

PORTS   = listVar(of=int)                   # "8080,9000" -> [8080, 9000]
TAGS    = listVar(of=str, sep=";")          # "api;bot"   -> ["api", "bot"]
ALLOWED = listVar(of=int, validate=lambda xs: all(x < 65536 for x in xs))

Whitespace around each element is stripped automatically. Errors include the offending index: [PORTS[1]] expected an integer, got 'nope'.

jsonVar (optional type assertion)

SETTINGS = jsonVar()           # any valid JSON
SETTINGS = jsonVar(type=dict)  # asserts the root value is a dict

stringVar(secret=True) (redacted values)

TOKEN = stringVar(secret=True)

Returns a SecretStr, which is a str subclass that redacts itself in __repr__ and __str__, so secrets don't leak into logs. The raw value is accessible via .reveal().

print(env.TOKEN)           # **redacted**
print(env.TOKEN.reveal())  # the-actual-secret

---

Prefix

Pass prefix on the class definition to prepend a namespace to every var name.

class MyEnv(Environment, prefix="APP_"):
    PORT = intVar(default=8080)  # reads APP_PORT

var_name always takes precedence over the prefix, letting you opt individual fields out:

class MyEnv(Environment, prefix="APP_"):
    DB_URL = stringVar(var_name="DATABASE_URL")  # reads DATABASE_URL, not APP_DATABASE_URL

---

Schema

Environment.schema() returns a typed description of every declared field — useful for documentation, startup health checks, or generating .env templates.

for field in MyEnv.schema():
    req = "required" if field["required"] else f"default={field.get('default')!r}"
    print(f"{field['env_var']:<20} {field['type']:<12} {req}")

APP_HOST             str          required
APP_PORT             int          default=8080
APP_DEBUG            bool         default=False
APP_TOKEN            SecretStr    required

---

Inheritance

MyEnvs compose naturally through class inheritance.

class BaseMyEnv(Environment):
    LOG_LEVEL = enumVar("debug", "info", "warn", "error", default="info")
    DEBUG     = boolVar(default=False)

class MyEnv(BaseMyEnv, prefix="APP_"):
    HOST  = stringVar()
    PORT  = intVar(default=8080)

---

Testing

Pass _source to substitute os.environ with a plain dict.

env = MyEnv(_source={"APP_HOST": "localhost", "APP_PORT": "9000"})

---

Immutability

MyEnv instances are frozen after construction. Any attempt to set an attribute raises AttributeError.

env.PORT = 1  # AttributeError: MyEnv instances are immutable after construction.