Incomplete input for hash function for @idempotent_function

[dispyfree]

When annotating functions via @idempotent_function, the wrapped function's name will not go into the hash computation; instead, only the payload and an environment-specific (but not function-specific) function name is assumed.
This causes idempotency to break in certain cases.

Expected Behavior

When two distinct functions which accept the very same payload as parameters are wrapped using @idempotent_function, invocation of one function should not affect invocation of the other function.

Current Behavior

Only the payload is considered.
The culprit appears to be

generated_hash = self._generate_hash(data=data)
function_name = os.getenv(constants.LAMBDA_FUNCTION_NAME_ENV, "test-func")

(lines 180-181) in persistence/base.py. One can clearly see that only the environment variable or "test-func" is assumed - but that is - of course - not specific to the actual function being wrapped.
So in short, the library currently only really distinguishes by payload and not be function name, hence causing the aforementioned behaviour.

Possible Solution

Pass on the wrapped function's name and use it for the return value of _get_hashed_idempotency_key.

Steps to Reproduce (for bugs)

def test():
    myIdempotentFunc(product = "bananas")
    anotherIdempotentFunc(product="bananas")
    return {}

@idempotent_function(data_keyword_argument="product", persistence_store=dynamodb)
def myIdempotentFunc(product):
    print("I got invoked with {product}".format(product = product))
    return None

@idempotent_function(data_keyword_argument="product", persistence_store=dynamodb)
def anotherIdempotentFunc(product):
    print("I will never get invoked; someone stole my payload :(".format(product = product))
    return None

This only prints myIdempotentFunc's output, so anotherIdempotentFunc never gets invoked.

Environment

Using "aws-lambda-powertools==1.22.0" on Python 3.9.7

Thank you in advance!

[boring-cyborg]

Thanks for opening your first issue here! We'll come back to you as soon as we can.

[heitorlessa]

hey @dispyfree thank you so much for spotting this - it's definitely an edge case we missed :/. function_name in that case is to be added in the DataRecord, not the actual method name being wrapped -- for that we need to do what you suggested.

We're fixing some papercuts elsewhere and this will be prioritized right after - More than happy to help review/guide the fix if you have some spare cycles too!

Thanks again

[michaelbrewer]

@heitorlessa @dispyfree - a basic fix for this in my PR (#869). I am just going to add a good unit test to reproduce the issue as well.

[heitorlessa]

merged - tagging for the next release hopefully this Friday

[heitorlessa]

Now released as part of 1.23 version - thank you again @dispyfree for flagging this