Value Error When Applying Multiple Data Validation Decorators to the Same Function

[rohithrockzz]

Description:

I encountered an issue while trying to apply multiple data validation decorators to a single function in the Hamilton DAG framework. Specifically, I am trying to validate different columns of a DataFrame using multiple instances of the @check_output_custom decorator. However, I receive a ValueError indicating that the function cannot be defined more than once.

Steps to Reproduce:

1. Define a function to process a DataFrame.
2. Apply multiple @check_output_custom decorators to the function, each with different validation parameters.
3. Attempt to run the decorated function.

Example code snippet:

1st issue code snippet

@check_output_custom(CompositePrimaryKeyValidatorPySparkDataFrame(columns=["OrderID", "ItemNumber"], importance="fail"))
@check_output_custom(CategoricalValuesValidatorPySparkDataFrame(column="CategoryID", allowed_values=[1, 2, 3], importance="fail"))
def process_order_data(order_data_config: dict, order_filter_template: List) -> DataFrame:
    # Function implementation
    pass

This raises the error:

ValueError: Cannot define function process_order_data_raw more than once. Already defined by function <function process_order_data

2nd issue code snippet

@check_output_custom(CategoricalValuesValidatorPySparkDataFrame(column="CategoryID", allowed_values=[1, 2, 3], importance="fail"))
@check_output_custom(CategoricalValuesValidatorPySparkDataFrame(column="ProductID", allowed_values=[10, 20, 30], importance="warn"))
def process_order_data(order_data_config: dict, order_filter_template: List) -> DataFrame:
    # Function implementation
    pass

This raises the error:

ValueError: Cannot define function process_order_data_CategoricalValuesValidator more than once. Already defined by function <function process_order_data

Expected Behavior

Applying multiple @check_output_custom decorators to a single function should allow for different validation checks on various columns of the DataFrame without raising a ValueError.

Actual Behavior

A ValueError is raised, indicating that the function cannot be defined more than once by the same validator.

Library & System Information

python version = 3.9.5
hamilton library version = 1.65.0

Additional Context:

This issue prevents the application of multiple validators to a single function, which is necessary for comprehensive data validation in our use case. It would be helpful if the framework could support multiple validators on the same function without raising errors.

Thank you for your attention to this issue.

[elijahbenizzy]

Description:

I encountered an issue while trying to apply multiple data validation decorators to a single function in the Hamilton DAG framework. Specifically, I am trying to validate different columns of a DataFrame using multiple instances of the @check_output_custom decorator. However, I receive a ValueError indicating that the function cannot be defined more than once.

Steps to Reproduce:

1. Define a function to process a DataFrame.
2. Apply multiple @check_output_custom decorators to the function, each with different validation parameters.
3. Attempt to run the decorated function.

Example code snippet:

1st issue code snippet

@check_output_custom(CompositePrimaryKeyValidatorPySparkDataFrame(columns=["OrderID", "ItemNumber"], importance="fail")) @check_output_custom(CategoricalValuesValidatorPySparkDataFrame(column="CategoryID", allowed_values=[1, 2, 3], importance="fail")) def process_order_data(order_data_config: dict, order_filter_template: List) -> DataFrame: # Function implementation pass

This raises the error:

ValueError: Cannot define function process_order_data_raw more than once. Already defined by function <function process_order_data

2nd issue code snippet

@check_output_custom(CategoricalValuesValidatorPySparkDataFrame(column="CategoryID", allowed_values=[1, 2, 3], importance="fail")) @check_output_custom(CategoricalValuesValidatorPySparkDataFrame(column="ProductID", allowed_values=[10, 20, 30], importance="warn")) def process_order_data(order_data_config: dict, order_filter_template: List) -> DataFrame: # Function implementation pass

This raises the error:

ValueError: Cannot define function process_order_data_CategoricalValuesValidator more than once. Already defined by function <function process_order_data

Expected Behavior

Applying multiple @check_output_custom decorators to a single function should allow for different validation checks on various columns of the DataFrame without raising a ValueError.

Actual Behavior

A ValueError is raised, indicating that the function cannot be defined more than once by the same validator.

Library & System Information

python version = 3.9.5 hamilton library version = 1.65.0

Additional Context:

This issue prevents the application of multiple validators to a single function, which is necessary for comprehensive data validation in our use case. It would be helpful if the framework could support multiple validators on the same function without raising errors.

Thank you for your attention to this issue.

Thanks for opening! This is limitation I think. E.G. two that have the same name + another complexity. I think we can build a fix, but just to check, if you have them both in the same validator (E.G. as follows) does it work? My guess is not, but worth a try:

@check_output_custom(
    CompositePrimaryKeyValidatorPySparkDataFrame(columns=["OrderID", "ItemNumber"], importance="fail")),
    CategoricalValuesValidatorPySparkDataFrame(column="CategoryID", allowed_values=[1, 2, 3], importance="fail")
)

[skrawcz]

Another thought would be to add another custom validator that takes in multiple validators... 🤔

Otherwise I think a potential avenue to scope would be to include some name_ kwarg to help name the node so it doesn't clash...

[rohithrockzz]

@elijahbenizzy
Yes, its working. Thanks for your help. If it is mention in the documentation, it would be helpful for the new comers.
But the 2nd issue still present which is 'If we pass same data validator twice getting error' like below
@check_output_custom(
CategoricalValuesValidatorPySparkDataFrame(column="ReportingId", allowed_values=[156], importance="fail")),
CategoricalValuesValidatorPySparkDataFrame(column="CategoryID", allowed_values=[1, 2, 3], importance="fail")
)

[skrawcz]

@rohithrockzz could you try installing
pip install sf-hamilton==1.66.1rc0
and see if that fixes your issue please?

[rohithrockzz]

@skrawcz
Yes, it worked. Thank you so much for the quick fix

[skrawcz]

@rohithrockzz great thanks for verifying. I will publish a non-RC version in the morning.

[skrawcz]

@rohithrockzz this has been released under sf-hamilton==1.66.1