I recently joined a team where we have an extension module based Python package say: mypackage.so, mypackage.pyd files are loaded as the sole package. So, now I'm taking responsibility to add the stubs for type hints for IDEs. I have tried the two famous automatic stub generation tools - mypy stubgen, pybind11-stubgen. Two of those stub generation tools expose only the top level class symbols but there are enormous other symbols that serve as properties to the classes and other stuff. All of them are not getting generated in the
.pyistub files. Could anyone please provide me a solution for this?
Could anyone please give some ideas or pointers for this to me?
UnboundType from a callback that takes a TypeChecker for context API (e.g. get_method_hook)? I've seen that attrs does this with the anal_type method, but TypeChecker doesn't have that. Is it possible to get the semantic analyser at least and perhaps use that to analyse the UnboundType?
print works fine; I think that changes to the plugin I was tweaking weren't actually being ran by mypy.
None hook, which meant that the plugin I was interested in didn't get to do anything.
I'm really struggling to understand an error
class Parent:
value: int | None = None
class Child(Parent):
def __init__(self, input: int) -> None:
self.input = input
@property
def value(self) -> int | None:
return self.input * 2I'm getting: Signature of "value" incompatible with supertype "Parent"
I can fix it with another property but it seems like overkill
class Parent:
@property
def value(self) -> int | None:
return None
class Child(Parent):
def __init__(self, input: int) -> None:
self.input = input
@property
def value(self) -> int | None:
return self.input * 2
2 replies
Hey cool cats. Maintainer of Flask recommended that this is a bug with mypy. Wanted a sanity check here before I open an issue. What do people think? pallets/flask#4600
TLDR: mypy appears to not consider this to be valid because of the name "bar" as the arg
def foo(bar: SomeType): -> SomeType
return bar
MyCallable = typing.Callable[[SomeType], SomeType]
def my_callback(foo: MyCallable)
foo("something")error: Argument 1 to "foo" of "MyCallable" has incompatible type "Callable[[Arg(SomeType, 'bar')], SomeType]"; expected "Callable[[SomeType], SomeType]"From the maintainer:
Please open this against MyPy first. It looks like they're treating your function as a named argument callable and not matching that to the basic callable in Flask's annotation.
classmethod on an Enum? The way I usually annotate classmethods doesn't seem to work:from enum import Enum
from typing import Set, Type, TypeVar
_Status = TypeVar("_Status", bound="Status")
class Status(Enum):
A = 1
B = 2
C = 3
@classmethod
def subset(cls: Type[_Status]) -> Set[_Status]:
return {cls.A, cls.B}sscce.py:14: error: Argument 1 to <set> has incompatible type "Status"; expected "_Status"
sscce.py:14: error: Argument 2 to <set> has incompatible type "Status"; expected "_Status"
Found 2 errors in 1 file (checked 1 source file)So, this seems to work:
from enum import Enum
from typing import Set
class Status(Enum):
A = 1
B = 2
C = 3
@classmethod
def subset(cls) -> Set["Status"]:
return {cls.A, cls.B}However, cls wouldn't be Status in a subclass, right? (Not that that's allowed haha)
1 reply
TypeVar method work here?
Hi, I have the following code:
from typing import Generic, TypeVar
from abc import ABC
R = TypeVar("R", bound="AbstractRepository")
T = TypeVar("T", bound="AbstractUnitOfWork")
class AbstractRepository(ABC):
pass
class ConcreteRepository(AbstractRepository):
pass
class AbstractUnitOfWork(ABC, Generic[R]):
repo: R
def commit(self: T) -> T:
return self
class ConcreteUnitOfWork(AbstractUnitOfWork[ConcreteRepository]):
def commit(self) -> "ConcreteUnitOfWork":
return super().commit()When running with the strict flag mypy complains that AbstractUnitOfWork in the definition of T is missing a type parameter.
I am not sure what the correct type parameter is here. I tried R and AbstractRepository but that just produces other errors. Does someone know what the correct parameter would be?
7 replies
typing.Concatenate? https://mypy-play.net/?mypy=latest&python=3.10&gist=ace3e59645b2dc98695cbc1e4c189792
1 reply
main.py:4: error: The first argument to Callable must be a list of types or "..."
Callable[P, Any] says error: The first argument to Callable must be a list of types or "...", and this is with mypy 0.950
1 reply
if TYPING or whatever ;-)
@agronholm: Basically imagine:
from typing import Dict, List
def foo(bar: Dict) -> List:
return []of course ignoring my lack of specifying types inside of dict/list but the from typing doesn't actually need to be imported for program runtime, so it could instead go like this:
if TYPE_CHECKING:
from typing import Dict, List
concurrent.futures.Future with a Protocol and get main.py:23: error: Argument "x" to "f" has incompatible type "Future[C]"; expected "Future[P]" https://mypy-play.net/?mypy=latest&python=3.10&gist=dda72d0cfe1a433261a7a5493d57cb1e. C seems to satisfy P when used directly, without the Future layer. is this a variance issue? is there another thing i should be doing instead?
1 reply
is not None checks by providing a callable object that includes the None check. it is specifically a pair of context managers that time the body, one of them asserts a maximum time for testing.
diff --git a/psycopg/psycopg/connection.py b/psycopg/psycopg/connection.py
index abf6bb0f..54b0e446 100644
--- a/psycopg/psycopg/connection.py
+++ b/psycopg/psycopg/connection.py
@ -665,6 +665,9 @@ class Connection(BaseConnection[Row]):
_pipeline: Optional[Pipeline]
+ _SelfRow = TypeVar("_SelfRow", bound="Connection[Row]")
+ _Self = TypeVar("_Self", bound="Connection[TupleRow]")
+
def __init__(
self,
pgconn: "PGconn",
@@ -679,7 +682,7 @@ class Connection(BaseConnection[Row]):
@overload
@classmethod
def connect(
- cls,
+ cls: Type[_SelfRow],
conninfo: str = "",
*,
autocommit: bool = False,
@@ -688,13 +691,13 @@ class Connection(BaseConnection[Row]):
cursor_factory: Optional[Type[Cursor[Row]]] = None,
context: Optional[AdaptContext] = None,
**kwargs: Union[None, int, str],
- ) -> "Connection[Row]":
+ ) -> _SelfRow:
...
@overload
@classmethod
def connect(
- cls,
+ cls: Type[_Self],
conninfo: str = "",
*,
autocommit: bool = False,
@@ -702,7 +705,7 @@ class Connection(BaseConnection[Row]):
cursor_factory: Optional[Type[Cursor[Any]]] = None,
context: Optional[AdaptContext] = None,
**kwargs: Union[None, int, str],
- ) -> "Connection[TupleRow]":
+ ) -> _Self:
...
@classmethod # type: ignore[misc] # https://github.com/python/mypy/issues/11004Connection[Any] instead of Connection[TupleRow]
_Self cannot be parametrizeddiff --git a/psycopg/psycopg/connection.py b/psycopg/psycopg/connection.py
index abf6bb0f..964c777a 100644
--- a/psycopg/psycopg/connection.py
+++ b/psycopg/psycopg/connection.py
@ -665,6 +665,8 @@ class Connection(BaseConnection[Row]):
_pipeline: Optional[Pipeline]
+ _Self = TypeVar("_Self", bound="Connection")
+
def __init__(
self,
pgconn: "PGconn",
@ -679,7 +681,7 @@ class Connection(BaseConnection[Row]):
@overload
@classmethod
def connect(
- cls,
+ cls: Type[_Self],
conninfo: str = "",
*,
autocommit: bool = False,
@@ -688,13 +690,13 @@ class Connection(BaseConnection[Row]):
cursor_factory: Optional[Type[Cursor[Row]]] = None,
context: Optional[AdaptContext] = None,
**kwargs: Union[None, int, str],
- ) -> "Connection[Row]":
+ ) -> _Self[Row]:
...
@overload
@classmethod
def connect(
- cls,
+ cls: Type[_Self],
conninfo: str = "",
*,
autocommit: bool = False,
@@ -702,7 +704,7 @@ class Connection(BaseConnection[Row]):
cursor_factory: Optional[Type[Cursor[Any]]] = None,
context: Optional[AdaptContext] = None,
**kwargs: Union[None, int, str],
- ) -> "Connection[TupleRow]":
+ ) -> _Self[TupleRow]:
...
@classmethod # type: ignore[misc] # https://github.com/python/mypy/issues/11004
Hey everyone, I just found this community after reading up on creating mypy plugins. Our codebase uses mypy to check all code, and we use Pydantic in quite a few places to add some runtime validation as well. We're exploring adding Prefect to our stack, but I'm running into some issues around typing. Namely this: Prefect has chosen to primarily focus on supporting Pyright before Mypy, but Pydantic and Pyright have issues working together. Plus, Pydantic has a whole Mypy integration which offers a bunch of awesome features; they clearly prefer Mypy. Now, I want to continue using Pydantic and Mypy, but I also really want to start using Prefect.
Is anyone else in a similar position to me, and is it possible that a Prefect 2.0 mypy plugin is already being developed? If not, I'm considering exploring this myself.
As an example of something in Prefect that works fine with Pyright, but doesn't work with Mypy: https://github.com/PrefectHQ/prefect/blob/orion/src/prefect/tasks.py#L231-L255
These 3 overloads are defined on Task.__call__, where Task is meant to be a decorator on a function. When I use Task to decorate any of my functions, Mypy tells me that reveal_type(decorated_task) is a PrefectFuture[None, Literal[False]]. Essentially, it never matches the third overload, so it always falls back to the first overload. I cannot for the life of me figure out why this is the case, especially considering Pyright has no trouble here. I thought it might have something to do with PEP 612 support, but Mypy seems to nearly fully support PEP 612 at this point, and this doesn't feel like much of an edge case.
So, hopefully you can see my dilemma here. Checking a Prefect task against Mypy fails on pretty much the most basic use case, but checking Pydantic-driven task functions against Pyright leaves out all of the great features of the mypy plugin. If anybody has seen Mr. Nobody, I feel like I'm the kid at the train having to decide which direction to go. And much like Jared Leto, I feel like I could fully explore both possibilities and return to the train station only to still be utterly confused as to which path to take.
from datetime import timedelta
from time import sleep
from typing import Callable, Optional, Union
def _make_sleeper(
seconds: Optional[Union[int, float, timedelta]] = 30,
) -> Callable[[], None]:
if seconds is not None:
if isinstance(seconds, timedelta):
seconds = seconds.total_seconds()
if seconds < 0:
raise ValueError("seconds must be positive.")
def _sleeper() -> None:
if seconds:
sleep(seconds)
return _sleepersscce.py:17: error: Argument 1 to "sleep" has incompatible type "Union[float, timedelta]"; expected "float"
Found 1 error in 1 file (checked 1 source file)I feel like I keep on running into situations where I'd like to be able to do something like
TaggedStr = NewType("TaggedStr", str, Generic[T]) # I actually hadn't thought about the syntax. Just imagine it's taking the Generic as another base class, or something.
str_tagged_with_int = TaggedStr[int]("my int-related string")Unless I've drastically misunderstood the typing documentation, I don't think there's a way to achieve this with baseline typing. If so, has anyone else run across this and made a plugin for it? I haven't had any luck searching for stuff like "parametric newtype python".
1 reply
sqlite3.Row now that it is generic. it seems that it was genericized against a single type, but many tables will not use the same type for all columns. also, i have setup to use the dict-like rows which would require typed-dict-like hinting i guess.