Closed sjsrey closed 1 year ago
Thanks, must've missed the test failures. This is happening because the order of the rectangle points returned from openCV has changed. They're now counter-clockwise, rather than clockwise.
My original variable names (I think) become a bit misleading, since the points aren't really the max/min x/y anymore.
So, I've just updated the order in the point array & swapped to allclose. Since opencv only works with float32, we'll need higher absolute precision in the allclose. I've updated that to the libpysal.common.ATOL * 100
.
Odd... on my platform, I'm getting different results even with the same version of opencv? I think this can be addressed by testing the numpy.roll
of the bounding box, since the right answer sometimes appears to be just one row offset.
Seems like there was an issue with the build... no failure, just an error from a timeout. I can't trigger a rebuild, as I don't think I'm a maintainer?
With the minimum rotated rectangle in GEOS being exposed soon by PyGEOS, we should move this functionality to use pygeos in a future release.
:exclamation: No coverage uploaded for pull request base (
master@5e2922f
). Click here to learn what that means. The diff coverage isn/a
.
@@ Coverage Diff @@
## master #69 +/- ##
=========================================
Coverage ? 63.15%
=========================================
Files ? 19
Lines ? 2266
Branches ? 0
=========================================
Hits ? 1431
Misses ? 835
Partials ? 0
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
Something changed in the dependency stack as the ordering and precision on the return was causing failures on travis for a a bit now.