Update default projection matrices

Author: gselzer

src/scenex/model/_nodes/camera.py

@@ -1,13 +1,16 @@
 from __future__ import annotations
 
-from typing import Literal
+from typing import TYPE_CHECKING, Literal
 
 from pydantic import Field
 
-from scenex.model._transform import Transform
+from scenex.utils import projections
 
 from .node import Node
 
+if TYPE_CHECKING:
+    from scenex.model._transform import Transform
+
 CameraType = Literal["panzoom", "perspective"]
 Position2D = tuple[float, float]
 Position3D = tuple[float, float, float]
@@ -35,8 +38,8 @@ class Camera(Node):
         description="Whether the camera responds to user interaction, "
         "such as mouse and keyboard events.",
     )
-    # FIXME: Default should be explained. And z-scale should probably be -1
     projection: Transform = Field(
-        default_factory=Transform,
-        description="Describes how 3D points are mapped to a 2D canvas",
+        default_factory=lambda: projections.orthographic(1, 1, 1),
+        description="Describes how 3D points are mapped to a 2D canvas, "
+        "default is an orthographic projection of a unit cube, centered at (0, 0, 0)",
     )

src/scenex/utils/projections.py

@@ -7,7 +7,7 @@
 from scenex.model._transform import Transform
 
 
-def orthographic(width: float = 2, height: float = 2, depth: float = 2) -> Transform:
+def orthographic(width: float = 1, height: float = 1, depth: float = 1) -> Transform:
     """Creates an orthographic projection matrix.
 
     Note that the resulting projection matrix provides no positional offset; this would
@@ -20,15 +20,15 @@ def orthographic(width: float = 2, height: float = 2, depth: float = 2) -> Trans
     Parameters
     ----------
     width: float, optional
-        The width of the camera rectangular prism. Default 2 (mirroring the side length
+        The width of the camera rectangular prism. Default 1 (mirroring the side length
         of a unit cube).
     height: float, optional
-        The height of the camera rectangular prism. Default 2 (mirroring the side length
+        The height of the camera rectangular prism. Default 1 (mirroring the side length
         of a unit cube).
     depth: float, optional
         The depth of the camera rectangular prism. The near and far clipping planes of
         the resulting matrix become (-depth / 2) and (depth / 2) respectively. Default
-        2, increase (to render things farther away) or decrease (to increase
+        1, increase (to render things farther away) or decrease (to increase
         performance) as needed.
 
         TODO: Is this a good default? May want to consider some large number (1000?)

tests/adaptors/_vispy/test_camera.py

@@ -55,7 +55,7 @@ def test_transform_with_view(view_camera: tuple[snx.Camera, adaptors.Camera]) ->
         [
             [5, 0, 0, 0],
             [0, 5, 0, 0],
-            [0, 0, 1, 0],
+            [0, 0, -1, 0],
             [5, 5, 0, 1],
         ]
     )
@@ -69,7 +69,7 @@ def test_transform_with_view(view_camera: tuple[snx.Camera, adaptors.Camera]) ->
         [
             [5, 0, 0, 0],
             [0, 5, 0, 0],
-            [0, 0, 1, 0],
+            [0, 0, -1, 0],
             [0, 0, 0, 1],
         ]
     )
@@ -82,15 +82,14 @@ def test_projection_with_view(view_camera: tuple[snx.Camera, adaptors.Camera]) -
     node = adaptor._vispy_node
     assert isinstance(node, BaseCamera)
     # Centered at [0, 0], top left [-1, -1], bottom right [1, 1]
-    identity_tform = Transform()
-    assert camera.projection == identity_tform
+    assert camera.projection == Transform().scaled((1, 1, -1))
     # Vispy wants to map [-1, -1] to [0, 0]
     # Vispy wants to map [1, 1] to [10, 10]
     exp_tform_mat = np.asarray(
         [
             [5, 0, 0, 0],
             [0, 5, 0, 0],
-            [0, 0, 1, 0],
+            [0, 0, -1, 0],
             [5, 5, 0, 1],
         ]
     )

tests/utils/test_projections.py

@@ -19,20 +19,20 @@ def test_orthographic() -> None:
     # By default, the return should unproject NDCs to a depth-inverted unit cube
     exp_mat = np.asarray(
         [
-            [1, 0, 0, 0],
-            [0, 1, 0, 0],
-            [0, 0, -1, 0],
+            [2, 0, 0, 0],
+            [0, 2, 0, 0],
+            [0, 0, -2, 0],
             [0, 0, 0, 1],
         ]
     )
     act_mat = orthographic()
     assert np.array_equal(exp_mat, act_mat)
     exp_corners = np.asarray(
         [
-            [-1, -1, 0],
-            [-1, 1, 0],
-            [1, -1, 0],
-            [1, 1, 0],
+            [-0.5, -0.5, 0],
+            [-0.5, 0.5, 0],
+            [0.5, -0.5, 0],
+            [0.5, 0.5, 0],
         ]
     )
     assert np.array_equal(exp_corners, vec_unproject(CORNERS, act_mat))