Bug using out with scaled Astra and Tigre projection operators.

[MargaretDuff]

Description

Passing out to the direct or adjoint of a scaled Astra and Tigre projection operator leads to an error

from cil.processors import TransmissionAbsorptionConverter, Slicer
import numpy as np
from cil.plugins.tigre import ProjectionOperator


# Load data
ground_truth = dataexample.SIMULATED_SPHERE_VOLUME.get()

data = dataexample.SIMULATED_CONE_BEAM_DATA.get()

# Consider just a single 2D slice 
data = data.get_slice(vertical='centre')
ground_truth = ground_truth.get_slice(vertical='centre')

absorption = TransmissionAbsorptionConverter()(data)
absorption = Slicer(roi={'angle':(0, -1, 5)})(absorption)

ig = ground_truth.geometry

A = ProjectionOperator(image_geometry=ig, 
                       acquisition_geometry=absorption.geometry)
A_scaled=3*A

with

from cil.plugins.astra.operators import ProjectionOperator

[MargaretDuff]

In the ScaledOperator direct and adjoint it assumes that self.operator.adjoint(x, out=out) and self.operator.direct(x, out=out) returns as well as fills out: https://github.com/TomographicImaging/CIL/blob/5749b38a9f59a8d6730e86a395184e62cc152b44/Wrappers/Python/cil/optimisation/operators/Operator.py#L500C1-L512C19

However, the astra and tigre processors don't return if out is filled:

CIL/Wrappers/Python/cil/plugins/astra/processors/AstraForwardProjector2D.py

Lines 105 to 123 in 5749b38

	def process(self, out=None):
	IM = self.get_input()
	#ASTRA expects a 2D array with shape 1, CIL removes dimensions of len 1
	new_shape_ig = [self.volume_geometry.voxel_num_y,self.volume_geometry.voxel_num_x]
	new_shape_ig = [x if x>0 else 1 for x in new_shape_ig]
	IM_data_temp = IM.as_array().reshape(new_shape_ig)
	sinogram_id, arr_out = astra.create_sino(IM_data_temp, self.proj_id)
	astra.data2d.delete(sinogram_id)
	arr_out = np.squeeze(arr_out)
	if out is None:
	out = AcquisitionData(arr_out, deep_copy=False, geometry=self.sinogram_geometry.copy(), suppress_warning=True)
	return out
	else:
	out.fill(arr_out)