Reputation: 471
I have a function that is either called as fun(start, stop, divisors)
or fun(stop, divisors)
.
If I ease on any of these two restrictions, the implementation is easy. For instance, easing the restriction on the order can be done like so:
def fun(stop: int, divisors: List[int], start: int=0) -> int:
...
return 0
fun(5, [1, 2, 3])
fun(5, [1, 2, 3], start=2)
Or easing on the type hints is shown in the file below, but with a proper ordering of arguments.
How should I write my code so that my function arguments are:
(start[optional], stop, divisors)
;import typing
from typing import Union, Optional
Start = int
Stop = int
Divisor = int
Divisors = list[Divisor]
@typing.overload
def fun1(x: Start, y: Stop, divisors: Divisors) -> int:
...
@typing.overload
def fun1(y: Stop, divisors: Divisors) -> int:
...
def fun1(*args) -> int:
if (arglen := len(args)) not in [2, 3]:
raise TypeError("Function expected 2 or 3 arguments, got", arglen)
if arglen == 2:
args = [0] + list(args)
start, stop, divisors = args
return 0
def fun2(x, y, divisors) -> int:
if divisors is None:
start, stop, divisors = 0, x, y
else:
start, stop = x, y
return 0
def fun3(*args) -> int:
if (arglen := len(args)) == 3:
start: Stop = args[0]
stop: int = args[1]
divisors: Divisors = args[2]
elif arglen == 2:
start: Stop = 0
stop: int = args[0]
divisors: Divisors = args[1]
else:
raise TypeError(
f"Too {'few' if arglen == 0 else 'many'} values to unpack (2-3), got",
arglen,
)
return 0
def fun4(*args) -> int:
if (arglen := len(args)) == 3:
pass
elif arglen == 2:
args = [0] + list(args)
else:
raise TypeError(
f"Too {'few' if arglen == 0 else 'many'} values to unpack (2-3), got",
arglen,
)
start: Stop = args[0]
stop: int = args[1]
divisors: Divisors = args[2]
return 0
typing_test_PE_001.py:20: error: Overloaded function implementation does not accept all possible arguments of signature 1
typing_test_PE_001.py:20: error: Overloaded function implementation does not accept all possible arguments of signature 2
typing_test_PE_001.py:24: error: Incompatible types in assignment (expression has type "List[int]", variable has type "Tuple[Any, ...]")
typing_test_PE_001.py:43: error: Name "start" already defined on line 39
typing_test_PE_001.py:44: error: Name "stop" already defined on line 40
typing_test_PE_001.py:45: error: Name "divisors" already defined on line 41
typing_test_PE_001.py:58: error: Incompatible types in assignment (expression has type "List[int]", variable has type "Tuple[Any, ...]")
Found 7 errors in 1 file (checked 1 source file)
Upvotes: 1
Views: 780
Reputation: 530823
Let's see how slice
is hinted:
class slice(object):
start: Any
step: Any
stop: Any
@overload
def __init__(self, stop: Any) -> None: ...
@overload
def __init__(self, start: Any, stop: Any, step: Any = ...) -> None: ...
__hash__: None # type: ignore
def indices(self, len: SupportsIndex) -> Tuple[int, int, int]: ...
and range
class range(Sequence[int]):
start: int
stop: int
step: int
@overload
def __init__(self, stop: SupportsIndex) -> None: ...
@overload
def __init__(self, start: SupportsIndex, stop: SupportsIndex, step: SupportsIndex = ...) -> None: ...
[...]
Basically, you hint the two overloaded versions separately, one with the optional first parameter and one without. (So basically, your Attempt #2.)
from typing import overload, List
@overload
def fun(start: int, stop: int, divisors: List[int]):
...
@overload
def fun(stop: int, divisors: List[int]):
...
def fun(start, stop, divisors=None):
if divisors is None:
divisors = stop
stop = start
start = 0
...
fun(1, 2, [1,2,3]) # OK
fun(2, [1,2,3]) # OK
If you like, you can also make both variants accept positional arguments only:
# e.g.
def fun(start, stop, divisors=None, /):
...
Upvotes: 4