Tracer.capture_method decorator does not preserve function type hints

[zlalvani]

What were you trying to accomplish?
I want to use the @tracer.capture_method decorator for detailed function level tracing however it's currently unusable because it disables all errors when linting with mypy.

Expected Behavior

See code example below

Current Behavior

Function signatures (argument and return types) are not preserved when decorated with @tracer.capture_method and linting with mypy.

Possible Solution

Mypy can be forced to look at the decorated function signature: https://mypy.readthedocs.io/en/stable/generics.html#declaring-decorators

Steps to Reproduce (for bugs)

1. Use code sample below
2. Run mypy (tested on mypy==0.782)

Environment

• Powertools version used: aws-lambda-powertools==1.10.2
• Packaging format (Layers, PyPi): PyPi

How to enable debug mode**

tracer = Tracer()

@tracer.capture_method
def foo(bar: str) -> str:
    return bar

x = foo('baz')
x.get('this is not a dict')  # should cause mypy error
y = foo(123)  # should cause mypy error

[heitorlessa]

Hey @zlalvani thanks a lot for raising this. While we've had numerous MyPy issues fixed we don't yet have full support for it -- Roadmap item that you can help us +1 to prioritize: aws-powertools/powertools-lambda#3

I'm unsure how to fix this atm as I had thought functools.wraps would ensure some level of correctness here (not at wrapt level, tho) -- This is that part of the code: https://github.com/awslabs/aws-lambda-powertools-python/blob/8724294013e3157de9478f90e7c1283a952de54f/aws_lambda_powertools/tracing/tracer.py#L607

I've been meaning to refactor capture_method as it's the most complex logic we have in the codebase, so it'd be a good exercise nevertheless to get this improved.

I'd appreciate if you have any guidance on how we can satisfy MyPy in a backwards compatible way - happy to provide any guidance if you'd like to try contributing it.

Thank you!

[heitorlessa]

Updating here to let you know the fix will be available in the next release.

Update: Spoke too soon. While this works it breaks when using decorator parameters - error: <nothing> not callable. The closest workaround is this but not suitable to what we're doing -- needs further investigation:

python/mypy#1551 (comment)

offending block: https://github.com/awslabs/aws-lambda-powertools-python/blob/5b87bb195fb154d2a112364a5d1d5c9513898e55/aws_lambda_powertools/tracing/tracer.py#L490

Potential solution is to use an overload like Pydantic does: pydantic/pydantic#2052

[heitorlessa]

Just merged the fix for synchronous functions - Async however requires the new ParamSpec only available in 3.10, so I'll wait for more customers using mypy + async fn as it'll increase the package size by 240K.

This will be available in next week's release (1.18.0) -- I'll ping back here once it's released.

Thank you again for flagging it @zlalvani

[heitorlessa]

Hi @zlalvani - This is now fixed in 1.18.0

More details in the release notes: https://github.com/awslabs/aws-lambda-powertools-python/releases/tag/v1.18.0

[zlalvani]

Awesome thanks for the fix! @heitorlessa