Right now if you have f(x: int) and you call it with 'str' you get a very confusing and unhelpful error message - it simply tells you that the pattern doesn't match.
Instead, we should try to tell our users more about why its not matching - for instance, if you only have a single overload (which is the majority of cases), you could indicate which terms are not matching, or make a clear indication if you have the wrong number of arguments.
This change needs to be made in both PyFunctionInstance.cpp (see calls to argTupleTypeDescription) and also in python_typed_function_wrapper.py, which currently implements a bunch of different ways of producing this error message that leak compiler details and are not consistent. Ideally, the compiler always produces the same message as the interpreter.
Right now if you have
f(x: int)and you call it with 'str' you get a very confusing and unhelpful error message - it simply tells you that the pattern doesn't match.Instead, we should try to tell our users more about why its not matching - for instance, if you only have a single overload (which is the majority of cases), you could indicate which terms are not matching, or make a clear indication if you have the wrong number of arguments.
This change needs to be made in both PyFunctionInstance.cpp (see calls to argTupleTypeDescription) and also in python_typed_function_wrapper.py, which currently implements a bunch of different ways of producing this error message that leak compiler details and are not consistent. Ideally, the compiler always produces the same message as the interpreter.