Address typing errors (#1316)

Author: fepegar

src/torchio/data/image.py

@@ -135,7 +135,7 @@ def __init__(
         tensor: TypeData | None = None,
         affine: TypeData | None = None,
         check_nans: bool = False,  # removed by ITK by default
-        reader: Callable = read_image,
+        reader: Callable[[TypePath], TypeDataAffine] = read_image,
         **kwargs: dict[str, Any],
     ):
         self.check_nans = check_nans
@@ -617,8 +617,8 @@ def read_and_check(self, path: TypePath) -> TypeDataAffine:
         # Make sure the data type is compatible with PyTorch
         if self.reader is not read_image and isinstance(tensor, np.ndarray):
             tensor = check_uint_to_int(tensor)
-        tensor = self._parse_tensor_shape(tensor)
-        tensor = self._parse_tensor(tensor)
+        tensor = self._parse_tensor_shape(tensor)  # type: ignore[assignment]
+        tensor = self._parse_tensor(tensor)  # type: ignore[assignment]
         affine = self._parse_affine(affine)
         if self.check_nans and torch.isnan(tensor).any():
             warnings.warn(

src/torchio/data/inference/aggregator.py

@@ -63,8 +63,8 @@ def _crop_patch(
         index_ini, index_fin = location[:3], location[3:]
 
         # If the patch is not at the border, we crop half the overlap
-        crop_ini = half_overlap.copy()
-        crop_fin = half_overlap.copy()
+        crop_ini: np.ndarray = half_overlap.copy()
+        crop_fin: np.ndarray = half_overlap.copy()
 
         # If the volume has been padded, we don't need to worry about cropping
         if self.volume_padded:
@@ -78,9 +78,17 @@ def _crop_patch(
         new_index_fin = index_fin - crop_fin
         new_location = np.hstack((new_index_ini, new_index_fin))
 
-        patch_size = patch.shape[-3:]
+        patch_size = np.asarray(patch.shape[-3:], dtype=int)
+        crop_fin = crop_fin.astype(int)
         i_ini, j_ini, k_ini = crop_ini
         i_fin, j_fin, k_fin = patch_size - crop_fin
+        # Make type checkers happy
+        i_ini = int(i_ini)
+        j_ini = int(j_ini)
+        k_ini = int(k_ini)
+        i_fin = int(i_fin)
+        j_fin = int(j_fin)
+        k_fin = int(k_fin)
         cropped_patch = patch[:, i_ini:i_fin, j_ini:j_fin, k_ini:k_fin]
         return cropped_patch, new_location
 

src/torchio/data/io.py

@@ -459,7 +459,7 @@ def ensure_4d(tensor: TypeData, num_spatial_dims=None) -> torch.Tensor:
     return tensor
 
 
-def check_uint_to_int(array):
+def check_uint_to_int(array: np.ndarray) -> np.ndarray:
     # This is because PyTorch won't take uint16 nor uint32
     if array.dtype == np.uint16:
         return array.astype(np.int32)

src/torchio/data/sampler/weighted.py

@@ -184,7 +184,7 @@ def extract_patch(  # type: ignore[override]
         i, j, k = self.get_random_index_ini(probability_map, cdf)
         index_ini = i, j, k
         si, sj, sk = self.patch_size
-        patch_size = si, sj, sk
+        patch_size = int(si), int(sj), int(sk)
         cropped_subject = self.crop(
             subject,
             index_ini,

tox.ini

@@ -36,7 +36,4 @@ commands = ruff format --diff
 dependency_groups =
     types
 commands =
-    mypy \
-        --install-types \
-        --non-interactive \
-        src
+    mypy src