Custom dataclasses with __dataclass_{fields,params}__ fields are not considered DataclassInstance #1987
Description
Summary
Hi!
First of all, thank you very much for this tool!
I use equinox.Module to create Python objects that are both dataclasses and PyTrees (to be compatible with JAX). However, it looks like running ty check emits a lot of errors that should not occur, at least from what I understood: a Python class is a DataclassInstance if it has both __dataclass_fields__ and __dataclass_params__ attributes. Another issue is that ty will raise an error about missing required arguments, whereas they are not missing.
I looked at other issues on this repo and couldn't find a similar one, but maybe I missed it.
FYI, I tried to used the playground but I don't see how we can include external dependencies, and the equinox module is too complex to be included manually.
MWE (check.py)
# /// script
# requires-python = ">=3.11"
# dependencies = [
# "equinox",
# ]
# ///
from dataclasses import dataclass, replace, asdict, field
import equinox as eqx
@dataclass
class MyBuiltinDataClass(eqx.Module):
a: int
b: float
c: str = field(default="default_value")
class MyEqxDataClass(eqx.Module):
a: int
b: float
c: str = eqx.field(default="default_value")
m = MyBuiltinDataClass(a=1, b=2.0)
print(m)
print(replace(m, b=3.5))
print(asdict(m))
print(hasattr(m, "__dataclass_fields__")) # True
print(hasattr(m, "__dataclass_params__")) # True
m = MyEqxDataClass(a=1, b=2.0)
print(m)
print(replace(m, b=3.5))
print(asdict(m))
print(hasattr(m, "__dataclass_fields__")) # True
print(hasattr(m, "__dataclass_params__")) # Trueuv run python check.py's output
MyBuiltinDataClass(a=1, b=2.0, c='default_value')
MyBuiltinDataClass(a=1, b=3.5, c='default_value')
{'a': 1, 'b': 2.0, 'c': 'default_value'}
True
True
MyEqxDataClass(a=1, b=2.0)
MyEqxDataClass(a=1, b=3.5)
{'a': 1, 'b': 2.0, 'c': 'default_value'}
True
True
ty check check.py's output
error[missing-argument]: No argument provided for required parameter `c`
--> check.py:26:5
|
24 | print(hasattr(m, "__dataclass_params__")) # True
25 |
26 | m = MyEqxDataClass(a=1, b=2.0)
| ^^^^^^^^^^^^^^^^^^^^^^^^^^
27 |
28 | print(m)
|
info: rule `missing-argument` is enabled by default
error[invalid-argument-type]: Argument to function `replace` is incorrect
--> check.py:29:15
|
28 | print(m)
29 | print(replace(m, b=3.5))
| ^ Argument type `MyEqxDataClass` does not satisfy upper bound `DataclassInstance` of type variable `_DataclassT`
30 | print(asdict(m))
31 | print(hasattr(m, "__dataclass_fields__")) # True
|
info: Type variable defined here
--> stdlib/dataclasses.pyi:32:1
|
30 | __all__ += ["KW_ONLY"]
31 |
32 | _DataclassT = TypeVar("_DataclassT", bound=DataclassInstance)
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
33 |
34 | @type_check_only
|
info: rule `invalid-argument-type` is enabled by default
error[invalid-argument-type]: Argument to function `asdict` is incorrect
--> check.py:30:14
|
28 | print(m)
29 | print(replace(m, b=3.5))
30 | print(asdict(m))
| ^ Expected `DataclassInstance`, found `MyEqxDataClass`
31 | print(hasattr(m, "__dataclass_fields__")) # True
32 | print(hasattr(m, "__dataclass_params__")) # True
|
info: Matching overload defined here
--> stdlib/dataclasses.pyi:67:5
|
66 | @overload
67 | def asdict(obj: DataclassInstance) -> dict[str, Any]:
| ^^^^^^ ---------------------- Parameter declared here
68 | """Return the fields of a dataclass instance as a new dictionary mapping
69 | field names to field values.
|
info: Non-matching overloads for function `asdict`:
info: (obj: DataclassInstance, *, dict_factory: (list[tuple[str, Any]], /) -> _T@asdict) -> _T@asdict
info: rule `invalid-argument-type` is enabled by default
Found 3 diagnostics
Version
ty 0.0.2