✨ optimize: have linprog return a specialized OptimizeResult

Author: jorenham

scipy-stubs/optimize/_linprog.pyi

@@ -1,5 +1,5 @@
 from collections.abc import Callable, Sequence
-from typing import Final, Literal, LiteralString, TypeAlias, TypedDict, overload, type_check_only
+from typing import Final, Literal, TypeAlias, TypedDict, overload, type_check_only
 from typing_extensions import deprecated
 
 import numpy as np
@@ -15,11 +15,9 @@ __all__ = ["linprog", "linprog_terse_callback", "linprog_verbose_callback"]
 
 _Ignored: TypeAlias = object
 _Max3: TypeAlias = Literal[0, 1, 2, 3]
-_Max4: TypeAlias = Literal[_Max3, 4]
 
 _Int: TypeAlias = int | np.int32 | np.int64
 _Float: TypeAlias = float | np.float64
-_Float1D: TypeAlias = onp.Array1D[np.float64]
 
 @type_check_only
 class _OptionsCommon(TypedDict, total=False):
@@ -83,17 +81,17 @@ __docformat__: Final = "restructuredtext en"  # undocumented
 LINPROG_METHODS: Final[Sequence[MethodLinprog | MethodLinprogLegacy]] = ...  # undocumented
 
 class OptimizeResult(_OptimizeResult):
-    x: _Float1D  # minimizing decision variables w.r.t. the constraints
-    fun: _Float  # optimal objective function value
-    slack: _Float1D  # slack values; nominally positive
-    con: _Float1D  # residuals of equality constraints; nominally zero
-    status: _Max4
-    message: LiteralString
-    nit: int  # >=0
-    success: bool  # `success = status == 0`
-
-def linprog_verbose_callback(res: _OptimizeResult) -> None: ...
-def linprog_terse_callback(res: _OptimizeResult) -> None: ...
+    x: onp.Array1D[np.float64] | None
+    fun: float | None
+    slack: onp.Array1D[np.float64] | None
+    con: onp.Array1D[np.float64] | None
+    success: bool
+    status: Literal[0, 1, 2, 3, 4]
+    message: str
+    nit: int
+
+def linprog_verbose_callback(res: OptimizeResult) -> None: ...
+def linprog_terse_callback(res: OptimizeResult) -> None: ...
 
 #
 @overload  # highs (default)
@@ -105,11 +103,11 @@ def linprog(
     b_eq: onp.ToFloat1D | None = None,
     bounds: Bound | Sequence[Bound] = (0, None),
     method: Literal["highs"] = "highs",
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsHighs | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...
 @overload  # highs-ds
 def linprog(
     c: onp.ToFloat1D,
@@ -120,11 +118,11 @@ def linprog(
     bounds: Bound | Sequence[Bound] = (0, None),
     *,
     method: Literal["highs-ds"],
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsHighsDS | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...
 @overload  # highs-ipm
 def linprog(
     c: onp.ToFloat1D,
@@ -135,11 +133,11 @@ def linprog(
     bounds: Bound | Sequence[Bound] = (0, None),
     *,
     method: Literal["highs-ipm"],
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsHighsIPM | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...
 @overload  # interior-point (legacy, see https://github.com/scipy/scipy/issues/15707)
 @deprecated("`method='interior-point'` is deprecated and will be removed in SciPy 1.17. Please use one of the HIGHS solvers.")
 def linprog(
@@ -151,11 +149,11 @@ def linprog(
     bounds: Bound | Sequence[Bound] = (0, None),
     *,
     method: Literal["interior-point"],
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsInteriorPoint | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...
 @overload  # revised simplex (legacy, see https://github.com/scipy/scipy/issues/15707)
 @deprecated("`method='revised simplex'` is deprecated and will be removed in SciPy 1.17. Please use one of the HIGHS solvers.")
 def linprog(
@@ -167,11 +165,11 @@ def linprog(
     bounds: Bound | Sequence[Bound] = (0, None),
     *,
     method: Literal["revised simplex"],
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsRevisedSimplex | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...
 @overload  # simplex (legacy, see https://github.com/scipy/scipy/issues/15707)
 @deprecated("`method='simplex'` is deprecated and will be removed in SciPy 1.17. Please use one of the HIGHS solvers.")
 def linprog(
@@ -183,11 +181,11 @@ def linprog(
     bounds: Bound | Sequence[Bound] = (0, None),
     *,
     method: Literal["simplex"],
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsSimplex | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...
 @overload  # any "highs"
 def linprog(
     c: onp.ToFloat1D,
@@ -197,8 +195,8 @@ def linprog(
     b_eq: onp.ToFloat1D | None = None,
     bounds: Bound | Sequence[Bound] = (0, None),
     method: MethodLinprog = "highs",
-    callback: Callable[[_OptimizeResult], _Ignored] | None = None,
+    callback: Callable[[OptimizeResult], _Ignored] | None = None,
     options: _OptionsHighs | None = None,
     x0: onp.ToFloat1D | None = None,
     integrality: _Max3 | Sequence[_Max3] | onp.CanArrayND[npc.integer] | None = None,
-) -> _OptimizeResult: ...
+) -> OptimizeResult: ...

tests/.ruff.toml

@@ -1,4 +1,4 @@
 extend = "../pyproject.toml"
 
 [lint]
-extend-ignore = ["B018", "PYI015", "PYI017"]
+extend-ignore = ["B018", "PYI002", "PYI015", "PYI017"]

tests/optimize/test_linprog.pyi

@@ -1,15 +1,24 @@
 from collections.abc import Sequence
 from typing import assert_type
 
+import numpy as np
+
 from scipy.optimize import OptimizeResult, linprog
 from scipy.optimize._typing import Bound
 
 c: list[float]
 
 bound: Bound
-uniformly_bounded = linprog(c, bounds=bound)
-assert_type(uniformly_bounded, OptimizeResult)
-
 bounds: Sequence[Bound]
-variably_bounded = linprog(c, bounds=bounds)
-assert_type(variably_bounded, OptimizeResult)
+
+###
+
+res = linprog(c, bounds=bounds)
+assert_type(res.fun, float | None)
+assert_type(res.x, np.ndarray[tuple[int], np.dtype[np.float64]] | None)
+assert_type(res.fun, float | None)
+assert_type(res.success, bool)
+assert_type(res.message, str)
+
+_1: OptimizeResult = linprog(c, bounds=bound)
+_2: OptimizeResult = linprog(c, bounds=bounds)