used coerce_to_class to convert imputation features and samples back …

…to their input classes

R/dim_reduction.R

@@ -311,6 +311,20 @@ impute_missing_values <- function(
     ) %>%
     dplyr::as_tibble()
 
+  # coerce feature and/or sample variables to their original classes
+  # (they will be converted to characters by impute.knn)
+
+  imputed_measurements[[feature_pk]] <- coerce_to_classes(
+    imputed_measurements[[feature_pk]],
+    triple_omic$features[[feature_pk]]
+  )
+
+  imputed_measurements[[sample_pk]] <- coerce_to_classes(
+    imputed_measurements[[sample_pk]],
+    triple_omic$samples[[sample_pk]]
+  )
+
+  # if in-place imputation is desired then remove the old variable
   updated_measurements <- triple_omic$measurements
   if (value_var == impute_var_name) {
     updated_measurements <- updated_measurements %>%

tests/testthat/test-dim_reduction.R

@@ -26,3 +26,19 @@ test_that("Removing missing values works", {
   expect_equal(nrow(filtered_brauer$samples), 34)
   expect_equal(nrow(filtered_brauer$measurements), 15674)
 })
+
+test_that("Matrices keys are reconstructed with appropriate classes", {
+
+  tomic_with_missing_values <- simple_triple
+  tomic_with_missing_values$measurements <- tomic_with_missing_values$measurements %>%
+    dplyr::slice(-c(1:5))
+
+  if (!("impute" %in% rownames(utils::installed.packages()))) {
+    cli::cli_alert("{.pkg impute} is not available for testing; imputation tests will be skipped")
+  } else{
+    imputed_tomic <- impute_missing_values(tomic_with_missing_values)
+    expect_s3_class(imputed_tomic, "triple_omic")
+  }
+
+
+})