✨ sparse.linalg: complete LaplacianNd & accept in dtypes in LinearOperator

scipy-stubs/sparse/linalg/_interface.pyi

@@ -20,17 +20,18 @@ _NumberT = TypeVar("_NumberT", bound=np.number[Any])
 _Matrix: TypeAlias = np.matrix[Any, np.dtype[_NumberT]]
 
 _ToShape: TypeAlias = Iterable[op.CanIndex]
-_ToDType: TypeAlias = type[_SCT] | onp.HasDType[np.dtype[_SCT]] | np.dtype[_SCT]
+_ToDType: TypeAlias = type[_SCT] | np.dtype[_SCT] | onp.HasDType[np.dtype[_SCT]]
 
+_JustFloat: TypeAlias = opt.Just[float]
 _JustComplex: TypeAlias = opt.Just[complex]
 
 _FunMatVec: TypeAlias = Callable[[onp.Array1D[np.number[Any]] | onp.Array2D[np.number[Any]]], onp.ToComplex1D | onp.ToComplex2D]
 _FunMatMat: TypeAlias = Callable[[onp.Array2D[np.number[Any]]], onp.ToComplex2D]
 
-_SCT = TypeVar("_SCT", bound=np.inexact[Any])
-_SCT_co = TypeVar("_SCT_co", bound=np.inexact[Any], default=np.inexact[Any], covariant=True)
-_SCT1_co = TypeVar("_SCT1_co", bound=np.inexact[Any], default=np.inexact[Any], covariant=True)
-_SCT2_co = TypeVar("_SCT2_co", bound=np.inexact[Any], default=_SCT1_co, covariant=True)
+_SCT = TypeVar("_SCT", bound=np.number[Any])
+_SCT_co = TypeVar("_SCT_co", bound=np.number[Any], default=np.inexact[Any], covariant=True)
+_SCT1_co = TypeVar("_SCT1_co", bound=np.number[Any], default=np.inexact[Any], covariant=True)
+_SCT2_co = TypeVar("_SCT2_co", bound=np.number[Any], default=_SCT1_co, covariant=True)
 _FunMatVecT_co = TypeVar("_FunMatVecT_co", bound=_FunMatVec, default=_FunMatVec, covariant=True)
 
 ###
@@ -57,7 +58,19 @@ class LinearOperator(Generic[_SCT_co]):
     @overload
     def __init__(self, /, dtype: _ToDType[_SCT_co], shape: _ToShape) -> None: ...
     @overload
-    def __init__(self: LinearOperator[np.float64], /, dtype: onp.AnyFloat64DType | type[float], shape: _ToShape) -> None: ...
+    def __init__(
+        self: LinearOperator[np.intp],
+        /,
+        dtype: onp.AnyIntPDType | type[opt.JustInt],
+        shape: _ToShape,
+    ) -> None: ...
+    @overload
+    def __init__(
+        self: LinearOperator[np.float64],
+        /,
+        dtype: onp.AnyFloat64DType | type[_JustFloat],
+        shape: _ToShape,
+    ) -> None: ...
     @overload
     def __init__(
         self: LinearOperator[np.complex128],
@@ -97,7 +110,7 @@ class LinearOperator(Generic[_SCT_co]):
     @overload
     def dot(self, /, x: onp.ToFloat) -> _ScaledLinearOperator[_SCT_co]: ...
     @overload
-    def dot(self, /, x: onp.ToComplex) -> _ScaledLinearOperator: ...
+    def dot(self, /, x: onp.ToComplex) -> _ScaledLinearOperator[_SCT_co | np.complex128]: ...
     @overload
     def dot(self, /, x: onp.ToFloatStrict1D) -> onp.Array1D[_SCT_co]: ...
     @overload

scipy-stubs/sparse/linalg/_special_sparse_arrays.pyi

@@ -1,44 +1,49 @@
-from scipy._typing import Untyped
+from typing import Any, Final, Generic, Literal, TypeAlias, overload
+from typing_extensions import TypeVar
+
+import numpy as np
+import optype.numpy as onp
+from scipy.sparse import bsr_array, coo_array, csc_array, csr_array, dia_array, dok_array, lil_array
 from scipy.sparse.linalg import LinearOperator
 
 __all__ = ["LaplacianNd"]
 
-class LaplacianNd(LinearOperator):
-    grid_shape: Untyped
-    boundary_conditions: Untyped
-
-    def __init__(self, /, grid_shape: Untyped, *, boundary_conditions: str = "neumann", dtype: Untyped = ...) -> None: ...
-    def eigenvalues(self, /, m: Untyped | None = None) -> Untyped: ...
-    def eigenvectors(self, /, m: Untyped | None = None) -> Untyped: ...
-    def toarray(self, /) -> Untyped: ...
-    def tosparse(self, /) -> Untyped: ...
-
-class Sakurai(LinearOperator):
-    n: Untyped
-    def __init__(self, /, n: Untyped, dtype: Untyped = ...) -> None: ...
-    def eigenvalues(self, /, m: Untyped | None = None) -> Untyped: ...
-    def tobanded(self, /) -> Untyped: ...
-    def tosparse(self, /) -> Untyped: ...
-    def toarray(self, /) -> Untyped: ...
-
-class MikotaM(LinearOperator):
-    def __init__(self, /, shape: Untyped, dtype: Untyped = ...) -> None: ...
-    def tobanded(self, /) -> Untyped: ...
-    def tosparse(self, /) -> Untyped: ...
-    def toarray(self, /) -> Untyped: ...
-
-class MikotaK(LinearOperator):
-    def __init__(self, /, shape: Untyped, dtype: Untyped = ...) -> None: ...
-    def tobanded(self, /) -> Untyped: ...
-    def tosparse(self, /) -> Untyped: ...
-    def toarray(self, /) -> Untyped: ...
-
-class MikotaPair:
-    n: Untyped
-    dtype: Untyped
-    shape: Untyped
-    m: Untyped
-    k: Untyped
-
-    def __init__(self, /, n: Untyped, dtype: Untyped = ...) -> None: ...
-    def eigenvalues(self, /, m: Untyped | None = None) -> Untyped: ...
+_SCT = TypeVar("_SCT", bound=np.number[Any])
+_SCT_co = TypeVar("_SCT_co", bound=np.number[Any], default=np.int8, covariant=True)
+
+_BoundaryConditions: TypeAlias = Literal["dirichlet", "neumann", "periodic"]
+_ToDType: TypeAlias = type[_SCT] | np.dtype[_SCT] | onp.HasDType[np.dtype[_SCT]]
+
+# because `scipy.sparse.sparray` does not implement anything :(
+_SpArray: TypeAlias = bsr_array | coo_array | csc_array | csr_array | dia_array | dok_array | lil_array
+
+###
+
+class LaplacianNd(LinearOperator[_SCT_co], Generic[_SCT_co]):
+    grid_shape: Final[onp.AtLeast1D]
+    boundary_conditions: Final[_BoundaryConditions]
+
+    @overload  # default dtype (int8)
+    def __init__(
+        self: LaplacianNd[np.int8],
+        /,
+        grid_shape: onp.AtLeast1D,
+        *,
+        boundary_conditions: _BoundaryConditions = "neumann",
+        dtype: _ToDType[np.int8] = ...,  # default: np.int8
+    ) -> None: ...
+    @overload  # know dtype
+    def __init__(
+        self,
+        /,
+        grid_shape: onp.AtLeast1D,
+        *,
+        boundary_conditions: _BoundaryConditions = "neumann",
+        dtype: _ToDType[_SCT_co] = ...,  # default: np.int8
+    ) -> None: ...
+
+    #
+    def eigenvalues(self, /, m: onp.ToJustInt | None = None) -> onp.Array1D[np.float64]: ...
+    def eigenvectors(self, /, m: onp.ToJustInt | None = None) -> onp.Array2D[np.float64]: ...
+    def toarray(self, /) -> onp.Array2D[_SCT_co]: ...
+    def tosparse(self, /) -> _SpArray: ...