[R] Implement support for keepNA = FALSE in base::nchar()

Author: HyukjinKwon

r/R/dplyr-funcs-string.R

@@ -484,15 +484,21 @@ register_bindings_string_other <- function() {
       if (is.na(keepNA)) {
         keepNA <- !identical(type, "width")
       }
-      if (!keepNA) {
-        # TODO: I think there is a fill_null kernel we could use, set null to 2
+      if (keepNA) {
         arrow_not_supported("keepNA = TRUE")
       }
       if (identical(type, "bytes")) {
-        Expression$create("binary_length", x)
+        result <- Expression$create("binary_length", x)
       } else {
-        Expression$create("utf8_length", x)
+        result <- Expression$create("utf8_length", x)
       }
+
+      if (!keepNA) {
+        # When keepNA = FALSE, NA values should return 2 (length of "NA" as string)
+        result <- Expression$create("coalesce", result, Expression$scalar(2L))
+      }
+
+      result
     },
     notes = "`allowNA = TRUE` and `keepNA = TRUE` not supported"
   )

r/tests/testthat/test-dplyr-funcs-string.R

@@ -1442,6 +1442,16 @@ test_that("nchar with namespacing", {
   )
 })
 
+test_that("nchar with keepNA = FALSE", {
+  df <- tibble(x = c("foo", NA_character_, "bar"))
+  compare_dplyr_binding(
+    .input |>
+      mutate(n = nchar(x, keepNA = FALSE)) |>
+      collect(),
+    df
+  )
+})
+
 test_that("str_trim()", {
   compare_dplyr_binding(
     .input |>