#103: add regression test and update NEWS

NEWS.md

@@ -21,11 +21,6 @@ Based on a request by user
 
 ## Bug fixes and minor improvements
 
-* If `n_components` was greater than `n_neighbors` then `umap_transform` would
-crash the R session. Thank you to [ChVav](https://github.com/ChVav) for 
-reporting this (<https://github.com/jlmelville/uwot/issues/102>).
-* Some new checks for NA values in input data have been added. Also a warning
-will be emitted if `n_components` seems to have been set too high.
 * A new setting for `init_sdev`: set `init_sdev = "range"` and initial
 coordinates will be range-scaled so each column takes values between 0-10. This
 pre-processing was added to the Python UMAP package at some point after `uwot`
@@ -40,6 +35,17 @@ parameters `a` and `b` are set directly when invoking `umap`, then both
 feature was added in response to a question from 
 [kjiang18](https://github.com/kjiang18) 
 (<https://github.com/jlmelville/uwot/issues/95>).
+* Some new checks for NA values in input data have been added. Also a warning
+will be emitted if `n_components` seems to have been set too high.
+* If `n_components` was greater than `n_neighbors` then `umap_transform` would
+crash the R session. Thank you to [ChVav](https://github.com/ChVav) for 
+reporting this (<https://github.com/jlmelville/uwot/issues/102>).
+* Using `umap_transform` with a model where `dens_scale` was set could cause
+a segmentation fault, destroying the session. Even if it didn't it could give
+an entirely artifactual "ring" structure. Thank you 
+[FemkeSmit](https://github.com/FemkeSmit) for reporting this and providing
+assistance in diagnosing the underlying cause 
+(<https://github.com/jlmelville/uwot/issues/103>).
 
 # uwot 0.1.14
 

tests/testthat/test_transform.R

@@ -246,4 +246,27 @@ test_that("return transform fgraph (#104)", {
   expect_equal(length(test_umap$rho), 10)
 })
 
+# regression tests the bug reported in #103 where ai and aj were transposed and
+# also the data being transformed is larger than the original data
+# this at best leads to a wide ring structure being formed for some of the
+# transformed data (but may also lead to NaN or a seg fault). This test checks
+# that the transformed data doesn't cover a large range, which would be
+# diagnostic of the ring forming: with the error present, the range of
+# coordinates is around c(-40, 40) vs c(-6, 6) otherwise
+test_that("leopold transform (#103)", {
+  iris_species_12 <- iris[1:100, ]
+  iris_species_3 <- iris[101:150, ]
+
+  set.seed(42)
+  iris_s3_leopold <- umap(iris_species_3, dens_scale = 1, ret_model = TRUE)
+
+  set.seed(42)
+  iris_s12_transform <- umap_transform(iris_species_12, iris_s3_leopold)
+
+  transform_range <- range(iris_s12_transform)
+  # as long as the coordinates aren't in the c(-40, 40) range then we are
+  # probably ok
+  expect_gt(transform_range[1], -10.0)
+  expect_lt(transform_range[2], 10.0)
+})