functions decorated with `daft.func` don't respect the input type signature

[universalmind303]

Describe the bug

if you have invalid types for native expressions such as using upper on a list[int], you get a type error during planning stating that the input is not of the correct type.

df = daft.from_pydict({"ints_list": [[1], [1, 2, 3]]})
df = df.select(daft.functions.upper(df["ints_list"]))
df.show()
daft.exceptions.DaftCoreException: DaftError::External Unable to create logical plan node.
Due to: DaftError::TypeError Expects input to 'upper' to be utf8, but received List[Int64]

but if you use a udf, the type signature is ignored, and results in a runtime error.

@daft.func
def my_upper(s: str) -> str: # s is actually a list[int] and the type signature is ignored
    return s.upper()

df = df.select(my_upper(df["ints_list"]))
df.show()
daft.exceptions.DaftCoreException: DaftError::ComputeError Error processing some rows:
0: DaftError::PyO3Error AttributeError: 'list' object has no attribute 'upper'

To Reproduce

see above

Expected behavior

if i have a function like:

@daft.func
def my_upper(s: str) -> str: ...

I'd expect daft to recognize that I expect this to only accept strings and error during planning time instead of failing during execution.

I'd want to get the same error message as you get with the native expressions

Due to: DaftError::TypeError Expects input to 'my_upper' to be utf8, but received List[Int64]

Component(s)

Expressions

Additional context

No response

[universalmind303]

tagging @kevinzwang as you might have some ideas around this.

[kevinzwang]

@universalmind303 we currently do not do static type checking with UDF input type annotations. It's something that would be great to add though!