ARROW-11756: [R] passing a partition as a schema leads to segfaults

r/R/dataset.R

@@ -64,6 +64,12 @@ open_dataset <- function(sources,
                          partitioning = hive_partition(),
                          unify_schemas = NULL,
                          ...) {
+  # If you aren't explicit with the argument names it looks like everything
+  # works but it will create a segfault and crash the R session, this fixes it
+  if (any(class(schema) %in% "HivePartitioning")) {
+    partitioning <- schema
+    schema <- NULL
+  }
   if (is_list_of(sources, "Dataset")) {
     if (is.null(schema)) {
       if (is.null(unify_schemas) || isTRUE(unify_schemas)) {

r/tests/testthat/test-dataset.R

@@ -194,6 +194,20 @@ test_that("Hive partitioning", {
   )
 })
 
+test_that("Hive partitioning without explicit argument name", {
+  ds <- open_dataset(hive_dir, hive_partition(other = utf8(), group = uint8()))
+  expect_is(ds, "Dataset")
+  expect_equivalent(
+    ds %>%
+      filter(group == 2) %>%
+      select(chr, dbl) %>%
+      filter(dbl > 7 & dbl < 53) %>%
+      collect() %>%
+      arrange(dbl),
+    df2[1:2, c("chr", "dbl")]
+  )
+})
+
 test_that("Partitioning inference", {
   # These are the same tests as above, just using the *PartitioningFactory
   ds1 <- open_dataset(dataset_dir, partitioning = "part")