DM-45144: Fix support for weight arrays in AccumulatorMeanStack.add_masked_image() #381
Conversation
There was a problem hiding this comment.
Minor improvement suggestions, and I'd like to request one unit test involving a non-trivial array as weight. I'm thinking of a toy case where the image and variance are identical (but have pixel-to-pixel variance) and passing the image as the weight produces the same stacked image as the stacked variance plane where the weight is 1.0
But more generally, what is the use case to support weight that is an array? Neither in adding per-detector warps to make a warp, nor in coadding warps do we want to use a weight that varies from one pixel to another.
| @@ -106,25 +106,37 @@ def add_masked_image(self, masked_image, weight=1.0): | |||
| good_pixels = np.where(((masked_image.mask.array & self.bit_mask_value) == 0) | |||
| & np.isfinite(masked_image.mask.array)) | |||
|
|
|||
| self.sum_weight[good_pixels] += weight | |||
| self.sum_wdata[good_pixels] += weight*masked_image.image.array[good_pixels] | |||
| weight_array = False | |||
There was a problem hiding this comment.
I know this is 100% internal, but i'd suggest calling it is_weight_array to denote that its meant to be boolean.
tests/test_accumulator_mean_stack.py
Outdated
| afw_masked_image.variance.array) | ||
| testing.assert_array_equal(online_masked_image.mask.array, | ||
| afw_masked_image.mask.array) | ||
| for use_weight_arrays in [False, True]: |
There was a problem hiding this comment.
Here and elsewhere, I'd recommend making use_weight_arrays an argument to the test function and use lsst.utils.tests.methodParameters.
|
cc @esheldon for the use of the weight array |
|
It was the 3-year ago me that made you put this comment Eli. I'm sorry, I don't think that was a good suggestion: #263 (comment) |
erykoff
force-pushed
the
tickets/DM-45144
branch
from
1adeb2a
to
0027f26
Compare
No description provided.