Earth engine is not always exporting results with the same rounded value (we are talking about the 10th decimal). This makes most of the tests very fragile.
I realized that pytest-regression is equipped with a num_regression fixture that is dedicted to compare numerical value with a default tolerance set to e-8 which would make the test more stable.
It can safely be used on every image regressions as we compute a reducer to get the datas.
nice side effect: it will reduce the size of the repository as csv are lighter than yml files.
Earth engine is not always exporting results with the same rounded value (we are talking about the 10th decimal). This makes most of the tests very fragile.
I realized that pytest-regression is equipped with a
num_regression
fixture that is dedicted to compare numerical value with a default tolerance set to e-8 which would make the test more stable.It can safely be used on every image regressions as we compute a reducer to get the datas.
nice side effect: it will reduce the size of the repository as csv are lighter than yml files.