Why

The following code will run in earlier Deephaven versions, but due to stricter type hint enforcing, fails in version 0.32 and up:

from typing import Optional
from deephaven import empty_table


def f1(x: str) -> str:
    return x


def f2(x: Optional[str]) -> Optional[str]:
    return x


t = empty_table(10).update(["X = i%2 == 0 ? null : `A`"])

t2 = t.update("Y = f1(X)")
t3 = t.update("Z = f2(X)")

It throws this error:

RuntimeError: java.lang.RuntimeException: Error in Python interpreter:
Type: <class 'TypeError'>
Value: Argument None is not compatible with annotation {<class 'str'>}
Line: 255
Namespace: _convert_arg

The calculation of t2 fails because the type signature of f1 says that it handles str arguments, but the column being passed to it contains nulls. An error results because the data received does not match the type hint in the function. In most cases, a user wouldn't really want this to work - their type hint doesn't indicate how nulls should be handled, and the error that is thrown makes sure the issue is clearly visible to the user.