@@ -1641,6 +1641,17 @@ def test__resample_valid_output():
16411641resample (np .zeros ((9 ,9 )),out )
16421642
16431643
1644+ @pytest .fixture
1645+ def nonaffine_identity ():
1646+ class NonAffineIdentityTransform (Transform ):
1647+ input_dims = 2
1648+ output_dims = 2
1649+
1650+ def inverted (self ):
1651+ return self
1652+ return NonAffineIdentityTransform ()
1653+
1654+
16441655@pytest .mark .parametrize ("data, interpolation, expected" ,
16451656 [(np .array ([[0.1 ,0.3 ,0.2 ]]),mimage .NEAREST ,
16461657np .array ([[0.1 ,0.1 ,0.1 ,0.3 ,0.3 ,0.3 ,0.3 ,0.2 ,0.2 ,0.2 ]])),
@@ -1649,7 +1660,7 @@ def test__resample_valid_output():
164916600.28476562 ,0.2546875 ,0.22460938 ,0.20002441 ,0.20002441 ]])),
16501661 ]
16511662)
1652- def test_resample_nonaffine (data ,interpolation ,expected ):
1663+ def test_resample_nonaffine (data ,interpolation ,expected , nonaffine_identity ):
16531664# Test that equivalent affine and nonaffine transforms resample the same
16541665
16551666# Create a simple affine transform for scaling the input array
@@ -1661,20 +1672,36 @@ def test_resample_nonaffine(data, interpolation, expected):
16611672
16621673# Create a nonaffine version of the same transform
16631674# by compositing with a nonaffine identity transform
1664- class NonAffineIdentityTransform (Transform ):
1665- input_dims = 2
1666- output_dims = 2
1667-
1668- def inverted (self ):
1669- return self
1670- nonaffine_transform = NonAffineIdentityTransform ()+ affine_transform
1675+ nonaffine_transform = nonaffine_identity + affine_transform
16711676
16721677nonaffine_result = np .empty_like (expected )
16731678mimage .resample (data ,nonaffine_result ,nonaffine_transform ,
16741679interpolation = interpolation )
16751680assert_allclose (nonaffine_result ,expected ,atol = 5e-3 )
16761681
16771682
1683+ @check_figures_equal ()
1684+ def test_nonaffine_scaling_to_axes_edges (fig_test ,fig_ref ,nonaffine_identity ):
1685+ # Test that plotting an image with equivalent affine and nonaffine
1686+ # transforms is scaled the same to the axes edges
1687+ data = np .arange (16 ).reshape ((4 ,4 ))% 3
1688+
1689+ # Specifically choose an axes bbox that has a fractional pixel
1690+
1691+ fig_test .set_size_inches (5 ,5 )
1692+ fig_test .set_dpi (100 )
1693+ ax = fig_test .subplots ()
1694+ ax .set_position ([0.2 ,0.2 ,300.5 / 500 ,300.5 / 500 ])
1695+ ax .imshow (data ,interpolation = 'nearest' ,
1696+ transform = nonaffine_identity + ax .transData )
1697+
1698+ fig_ref .set_size_inches (5 ,5 )
1699+ fig_ref .set_dpi (100 )
1700+ ax = fig_ref .subplots ()
1701+ ax .set_position ([0.2 ,0.2 ,300.5 / 500 ,300.5 / 500 ])
1702+ ax .imshow (data ,interpolation = 'nearest' )
1703+
1704+
16781705def test_axesimage_get_shape ():
16791706# generate dummy image to test get_shape method
16801707ax = plt .gca ()