Fix ratio sharpening not sharing invalid mask between bands #2262
Conversation
|
When I run asv on my own machine (while I'm doing other things) for the last 4 commits where this PR is the last entry, I get this for peakmem on the AHI true_color load test:
So not much of a change. Time:
It may look like it goes up, but those other commits weren't even related to these changes so this is just system variance from my point of view. I think this is safe to merge in that regard. I'm still tracking down another issue that @kathys identified in ssec/polar2grid#558 regarding quality/sharpness of the images, but I'm not sure it is related to these refactorings. I should know more in the next hour or so. |
Codecov Report
@@ Coverage Diff @@
## main #2262 +/- ##
==========================================
+ Coverage 94.12% 94.19% +0.06%
==========================================
Files 295 295
Lines 45376 45379 +3
==========================================
+ Hits 42712 42746 +34
+ Misses 2664 2633 -31
Flags with carried forward coverage won't be shown. Click here to find out more.
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. |
|
CI is having a rough day today. First it was a mamba hiccup, now it is a coveralls connection hiccup. 🤞 |
|
Merging given the simplicity of the fix, the fact that I broke it originally, and the issues it is causing for me downstream. Slap my wrist if you must. 📏 |
Fixes a bug introduced in #2240 where I refactored the duplicate masking behavior because it was already handled in the parent class, GenericCompositor. However, I didn't change the default for the ratio sharpening class to enable this behavior in the parent class so it wasn't being done at all. The tests were not complex enough to pick up this mistake but have been updated in this PR.
The behavior I'm talking about is where a mask is created of the NaN locations in each band, merged, and then applied to all the bands. Without this, you get strange colors where one band has missing data but another doesn't. This is especially necessary on the edges of resampled data where different resolution may "stretch" further geographically than other bands.