fix: handle off-centre masks in convolver and blurring grid padding (#274)

When the mask is near the image edge, the PSF blurring region can extend
beyond the array boundary. The convolver now accounts for asymmetric
mask positions when computing fft_shape, and the blurring grid passes
allow_padding=True so off-centre datasets no longer raise MaskException.

Co-authored-by: Jammy2211 <JNightingale2211@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: 3 people

autoarray/operators/convolver.py

@@ -116,8 +116,14 @@ class determines how masked real-space data are embedded into a padded array,
             s1 + s2 - 1 for s1, s2 in zip(mask_shape, self.kernel.shape_native)
         )
         import scipy.fft
+        from autoarray.mask.mask_2d_util import required_shape_for_kernel
 
-        fft_shape = tuple(scipy.fft.next_fast_len(s, real=True) for s in full_shape)
+        min_blur_shape = required_shape_for_kernel(mask, self.kernel.shape_native)
+
+        fft_shape = tuple(
+            scipy.fft.next_fast_len(max(s, r), real=True)
+            for s, r in zip(full_shape, min_blur_shape)
+        )
 
         self.fft_shape = fft_shape
         self.mask = mask.resized_from(self.fft_shape, pad_value=1)

autoarray/structures/grids/uniform_2d.py

@@ -710,7 +710,7 @@ def blurring_grid_from(
         """
 
         blurring_mask = mask.derive_mask.blurring_from(
-            kernel_shape_native=kernel_shape_native
+            kernel_shape_native=kernel_shape_native, allow_padding=True
         )
 
         return cls.from_mask(