Configuration#

ArgDigest supports three configuration levels. Use them in this precedence order (highest first):

Explicit decorator arguments (digestion_source, digestion_style, strictness, etc.).
Explicit configuration module (config="mylib._argdigest").
Environment module (ARGDIGEST_CONFIG="mylib._argdigest").
Auto-discovered module (<root_package>._argdigest) when no explicit config is provided.

Recommended `_argdigest.py` template#

# mylib/_argdigest.py

# Axis 2 -- the value contract of each argument.
DIGESTION_SOURCE = "mylib._private.argdigest.argument"
DIGESTION_STYLE = "package"  # package | registry | decorator | auto
STANDARDIZER = "mylib._private.argdigest.argument_names_standardization:argument_names_standardization"
STRICTNESS = "warn"          # warn | error | ignore
SKIP_PARAM = "skip_digestion"

# Axis 1 -- the argument contract of each function.
FUNCTION_SOURCE = "mylib._private.argdigest.function"
DOMAIN_SOURCE = "mylib._private.argdigest.domain"
UNKNOWN_ARGUMENT = "error"   # error | warn | ignore

# Declared argument-name aliases, applied before both axes.
NORMALIZATION_SOURCE = "mylib._private.argdigest.normalization"

Both policies accept the same aliases: raise -> error, warning -> warn, silent/none -> ignore.

The two policies are not the same knob#

	Fires when	Who made the mistake	Default
`STRICTNESS`	a declared parameter has no digester	the library author	`warn`
`UNKNOWN_ARGUMENT`	a keyword is outside the function’s contract	whoever made the call	`error`

A missing digester is a to-do for you, so a warning is right. A keyword nobody declared is a caller’s typo, and warning is not enough: warnings are routinely filtered off exactly where users read output, and the call would then run with the default and return a plausible wrong answer.

FUNCTION_SOURCE and DOMAIN_SOURCE are optional. Without them every closed signature is still held to its own parameters – ArgDigest never ends up more permissive than Python – and only functions taking **kwargs stay unconstrained until you declare the domain their keywords come from.

Using config in decorators#

from argdigest import arg_digest

@arg_digest(config="mylib._argdigest")
def get(...):
    ...

Programmatic defaults (optional)#

You can set defaults programmatically at import time:

import argdigest.config

argdigest.config.set_defaults(
    digestion_source="mylib._private.argdigest.argument",
    digestion_style="package",
    strictness="warn",
)

Practical guidance#

Prefer a single _argdigest.py per library package.
Use explicit decorator overrides only for exceptional endpoints.
Keep configuration values stable and documented for contributors.

Next#

Continue with Configuration Precedence.