Architecture

ArgDigest is organized around three core concepts:

• Argument digestion: per-argument digesters discovered from user libraries.

• Pipelines: reusable rules registered by kind and executed in order.

• Context and errors: structured data and exceptions for consistent diagnostics.

Key modules:

• argdigest/core/decorator.py: @arg_digest implementation and execution flow.

• argdigest/core/argument_loader.py: discovery of argument digesters. Uses functools.lru_cache to prevent redundant package scanning.

• argdigest/core/argument_registry.py: decorator-based digester registry.

• argdigest/core/registry.py: pipeline registry and execution.

• argdigest/core/context.py: call context container.

• argdigest/core/errors.py: error and warning classes.

Performance Strategy

ArgDigest employs caching at two critical levels to ensure minimal runtime overhead:

1. Digester Discovery: argument_loader._load_from_package is memoized to avoid repeated pkgutil.iter_modules calls.

2. Signature Inspection: decorator.get_digester_metadata caches inspect.signature results for all digesters, preventing redundant parsing of function signatures during import.