Reject out-of-bounds field-of-view origins

Author: mmporong

CHANGELOG.md

@@ -6,6 +6,10 @@ This project adheres to [Semantic Versioning](https://semver.org/) since version
 
 ## [Unreleased]
 
+### Changed
+
+- `tcod.map.compute_fov` now raises `ValueError` instead of warning when `pov` is outside the transparency array, as previously announced by the warning.
+
 ## [21.2.2] - 2026-10-22
 
 ### Changed

tcod/map.py

@@ -180,7 +180,7 @@ def compute_fov(
     `pov` is the point-of-view origin point.  Areas are visible if they can
     be seen from this position.  `pov` should be a 2D index matching the axes
     of the `transparency` array, and must be within the bounds of the
-    `transparency` array.
+    `transparency` array.  A `ValueError` is raised if `pov` is out of bounds.
 
     `radius` is the maximum view distance from `pov`.  If this is zero then
     the maximum distance is used.
@@ -240,13 +240,9 @@ def compute_fov(
         msg = "The tcod.map.compute_fov function has changed.  The `x` and `y` parameters should now be given as a single tuple."
         raise TypeError(msg)
     if not (0 <= pov[0] < transparency.shape[0] and 0 <= pov[1] < transparency.shape[1]):
-        warnings.warn(
-            f"Given pov index {pov!r} is outside the array of shape {transparency.shape!r}."
-            "\nThis will raise an error in future versions.",
-            RuntimeWarning,
-            stacklevel=2,
-        )
-    map_buffer: NDArray[np.bool_] = np.zeros(
+        msg = f"Given pov index {pov!r} is outside the array of shape {transparency.shape!r}."
+        raise ValueError(msg)
+    map_buffer: NDArray[np.bool_] = np.empty(
         transparency.shape,
         dtype=[("transparent", bool), ("walkable", bool), ("fov", bool)],
     )

tests/test_tcod.py

@@ -61,14 +61,18 @@ def test_tcod_bsp() -> None:
 
 
 @pytest.mark.parametrize("pov", [(-1, 0), (0, -1), (3, 0), (0, 5)])
-def test_compute_fov_out_of_bounds(pov: tuple[int, int], monkeypatch: pytest.MonkeyPatch) -> None:
+def test_compute_fov_out_of_bounds(pov: tuple[int, int]) -> None:
     transparency = np.ones((3, 5), dtype=bool)
-    # Model an uninitialized allocation containing stale visible cells.
-    monkeypatch.setattr(np, "empty", np.ones)
-    with pytest.warns(RuntimeWarning, match="outside the array"):
-        visible = tcod.map.compute_fov(transparency, pov)
+    with pytest.raises(ValueError, match="outside the array"):
+        tcod.map.compute_fov(transparency, pov)
+
+
+@pytest.mark.parametrize("pov", [(0, 0), (0, 4), (2, 0), (2, 4)])
+def test_compute_fov_at_boundary(pov: tuple[int, int]) -> None:
+    transparency = np.ones((3, 5), dtype=bool)
+    visible = tcod.map.compute_fov(transparency, pov)
     assert visible.shape == transparency.shape
-    assert not visible.any()
+    assert visible.all()
 
 
 @pytest.mark.filterwarnings("ignore:Use map.+ to check for this")