[Code scan] DataType crashes for tuple dtypes despite documenting them

[njzjz]

This issue is part of a Codex global repository code scan.

DataType.__init__() documents dtype as "type or tuple[type]", and check() uses isinstance(data, self.dtype), which supports tuple dtypes. However, error and repr formatting use self.dtype.__name__, which raises AttributeError when dtype is a tuple.

Affected code:

dpdata/dpdata/data_type.py

Lines 42 to 44 in a7a50bf

	dtype : type or tuple[type]
	data type, e.g. np.ndarray
	shape : tuple[int], optional

dpdata/dpdata/data_type.py

Lines 98 to 101 in a7a50bf

	return (
	f"DataType(name='{self.name}', dtype={self.dtype.__name__}, "
	f"shape={self.shape}, required={self.required}, "
	f"deepmd_name='{self.deepmd_name}')"

dpdata/dpdata/data_type.py

Lines 146 to 149 in a7a50bf

	elif not isinstance(data, self.dtype):
	raise DataError(
	f"Type of {self.name} is {type(data).__name__}, but expected {self.dtype.__name__}"
	)

Minimal reproducer:

from dpdata.data_type import DataType

print(repr(DataType("x", (list, tuple))))

Current behavior:

AttributeError: 'tuple' object has no attribute '__name__'

The same issue occurs in check() when reporting a type mismatch. Either tuple dtypes should be supported in formatting, or the documented accepted type should be narrowed.

[dosubot]

This is a valid bug. The __repr__ method [1] and the check() error message [2] both call self.dtype.__name__, which fails when dtype is a tuple of types.

A straightforward fix would be to replace self.dtype.__name__ with a helper that handles both cases, e.g.:

def _dtype_name(self) -> str:
    if isinstance(self.dtype, tuple):
        return " | ".join(t.__name__ for t in self.dtype)
    return self.dtype.__name__

Then use self._dtype_name() in __repr__ and in the check() error message. Alternatively, the type annotation for dtype in __init__ could be updated from type to type | tuple[type, ...] to match the documented behavior [3].

To reply, just mention @dosu.

---

Docs are dead. Just use Dosu.