Failure with the new version of **blockCV**

[rvalavi]

Failure with the new version of **blockCV**

[pat-s]

The data should not matter for this in my view.
You should be able to adapt/copy the test used in test-ResamplingSpCVBuffer.R and replicate it with your own data.

You can see what is being done starting here: https://github.com/mlr-org/mlr3spatiotempcv/blob/main/R/ResamplingSpCVBuffer.R#L119.

The values going into the test can be found here: https://github.com/mlr-org/mlr3spatiotempcv/blob/main/tests/testthat/test-ResamplingSpCVBuffer.R#L57.

[pat-s]

Thanks for the info.

Any idea why numbers change with the new update? Does the updated code apply a different sorting of the results? I think it would be easiest if the sorting of the result would stay as before to avoid having to push an update just because of this.

[rvalavi]

@pat-s yeah of course! Could you provide me with your example data so I can diagnose it?

[rvalavi]

@pat-s I checked and here are a few notes:
1- all blocking functions of v2.x still work (and will work for a while to allow appropriate time for adapting the new code), but a new version of them is written with new (general) names
2- there are a few new features such as hexagonal blocks for spatial cv (cv_spatial function) and the newly proposed algorithm, nearest distance distribution matching (Mila et al 2022) is added (cv_nndm); read some of the updates at https://github.com/rvalavi/blockCV#new-updates-of-the-version-30
3- I revert back function spatialBlock to exactly match the previous version to pass mlr3spatiotempcv tests; somehow the data passed to rsmp didn't work with sf::st_is_longlat for the new version of my function.
4- the values in the test (https://github.com/mlr-org/mlr3spatiotempcv/blob/main/tests/testthat/test-ResamplingSpCVBuffer.R#L57.) are the indices and their order shouldn't matter! that's why tests such as expect_equal(rsp$test_set(16), c(22, 16)) failed; I changed the order in my function for now to pass this test; please change these tests to something like: expect_equal(sort(rsp$test_set(16)), c(16, 22)) in future versions.

[pat-s]

fixed in #222